Metallurgical Modelling of Welding

Metallurgical
Modelling
of Welding
SECOND EDITION
0YSTEIN GRONG
Norwegian University of
Science and Technology,
Department of Metallurgy,
N-7034 Trondheim, Norway
MATERIALS MODELLING SERIES
Editor: H. K. D. H. Bhadeshia
The University of Cambridge
Department of Materials Science
and Metallurgy
T H E INSTITUTE OF MATERIALS
Book 677
First published in 1997 by
The Institute of Materials
1 Carlton House Terrace
London SWlY 5DB
First edition (Book 557)

Published in 1994
The Institute of Materials 1997

All rights reserved
ISBNl 86125 036 3
Originally typeset by
PicA Publishing Services
Additional typesetting and corrections by
Fakenham Photosetting Ltd
Printed and bound in the UK at
The University Press, Cambridge
TO TORHILD, TORBJ0RN AND HAVARD
without your support, this book would never have been finished.
Preface to the second edition
Besides correcting some minor linguistic and print errors, I have in the second edition in-
cluded a collection of different exercise problems which have been used in the training of stu-
dents at NTNU. They illustrate how the models described in the previous chapters can be used
to solve practical problems of more interdisciplinary nature. Each of them contains a 'prob-
lem description' and some background information on materials and welding conditions. The
exercises are designed to illuminate the microstructural connections throughout the weld
thermal cycle and show how the properties achieved depend on the operating conditions ap-
plied. Solutions to the problems are also presented. These are not complete or exhaustive, but
are just meant as an aid to the reader to develop the ideas further.
Trondheim, 28 October, 1996
0ystein Grong
Preface to the first edition
The purpose of this textbook is to present a broad overview on the fundamentals of welding
metallurgy to graduate students, investigators and engineers who already have a good back-
ground in physical metallurgy and materials science. However, in contrast to previous text-
books covering the same field, the present book takes a more direct theoretical approach to
welding metallurgy based on a synthesis of knowledge from diverse disciplines. The motiva-
tion for this work has largely been provided by the need for improved physical models for
process optimalisation and microstructure control in the light of the recent advances that have
taken place within the field of materials processing and alloy design.
The present textbook describes a novel approach to the modelling of dynamic processes in
welding metallurgy, not previously dealt with. In particular, attempts have been made to ra-
tionalise chemical, structural and mechanical changes in weldments in terms of models based
on well established concepts from ladle refining, casting, rolling and heat treatment of steels
and aluminium alloys. The judicious construction of the constitutive equations makes full use
of both dimensionless parameters and calibration techniques to eliminate poorly known ki-
netic constants. Many of the models presented are thus generic in the sense that they can be
generalised to a wide range of materials and processing. To help the reader understand and
apply the subjects and models treated, numerous example problems, exercise problems and
case studies have been worked out and integrated in the text. These are meant to illustrate the
basic physical principles that underline the experimental observations and to provide a way of
developing the ideas further.
Over the years, I have benefited from interaction and collaboration with numerous people
within the scientific community. In particular, I would like to acknowledge the contribution
from my father Professor Tor Grong who is partly responsible for my professional upbringing
and development as a metallurgist through his positive influence on and interest in my re-
search work. Secondly, I am very grateful to the late Professor Nils Christensen who first
introduced me to the fascinating field of welding metallurgy and later taught me the basic
principles of scientific work and reasoning. I will also take this opportunity to thank all my
friends and colleagues at the Norwegian Institute of Technology (Norway), The Colorado School
of Mines (USA), the University of Cambridge (England), and the Universitat der Bundeswehr
Hamburg (Germany) whom I have worked with over the past decade. Of this group of people,
I would particularly like to mention two names, i.e. our department secretary Mrs. Reidun
0stbye who has helped me to convert my original manuscript into a readable text and Mr.
Roald Skjaerv0 who is responsible for all line-drawings in this textbook. Their contributions
are gratefully acknowledged.
Trondheim, 1 December, 1993
0ystein Grong
Contents
Preface to the Second Edition ........................................................ xiii
Preface to the First Edition ............................................................. xiv
1. Heat Flow and Temperature Distribution in Welding ........... 1

1.1 Introduction ............................................................................... 1
1.2 Non-steady Heat Conduction .................................................... 1
1.3 Thermal Properties of Some Metals and Alloys ........................ 2
1.4 Instantaneous Heat Sources ..................................................... 4
1.5 Local Fusion in Arc Strikes ........................................................ 7
1.6 Spot Welding ............................................................................. 10
1.7 Thermit Welding ........................................................................ 14
1.8 Friction Welding ........................................................................ 18
1.9 Moving Heat Sources and Pseudo-steady State ...................... 24
1.10 Arc Welding ............................................................................... 24
1.10.1 Arc Efficiency Factors .................................................. 26
1.10.2 Thick Plate Solutions ................................................... 26
1.10.2.1 Transient Heating Period ............................. 28
1.10.2.2 Pseudo-steady State Temperature
Distribution ................................................... 31
1.10.2.3 Simplified Solution for a Fast-moving High
Power Source .............................................. 41
1.10.3 Thin Plate Solutions ..................................................... 45
1.10.3.1 Transient Heating Period ............................. 48
1.10.3.2 Pseudo-steady State Temperature
Distribution ................................................... 49
This page has been reformatted by Knovel to provide easier navigation. vi

Contents vii
1.10.3.3 Simplified Solution for a Fast Moving High

Power Source .............................................. 56
1.10.4 Medium Thick Plate Solution ....................................... 59
1.10.4.1 Dimensionless Maps for Heat Flow
Analyses ...................................................... 61
1.10.4.2 Experimental Verification of the Medium
Thick Plate Solution ..................................... 72
1.10.4.3 Practical Implications ................................... 75
1.10.5 Distributed Heat Sources ............................................. 77
1.10.5.1 General Solution .......................................... 77
1.10.5.2 Simplified Solution ....................................... 80
1.10.6 Thermal Conditions during Interrupted Welding .......... 91
1.10.7 Thermal Conditions during Root Pass Welding ........... 95
1.10.8 Semi-empirical Methods for Assessment of Bead
Morphology .................................................................. 96
1.10.8.1 Amounts of Deposit and Fused Parent
Metal ............................................................ 96
1.10.8.2 Bead Penetration ......................................... 99
1.10.9 Local Preheating .......................................................... 100
References ......................................................................................... 103
Appendix 1.1: Nomenclature ............................................................ 105
Appendix 1.2: Refined Heat Flow Model for Spot Welding .............. 110
Appendix 1.3: The Gaussian Error Function .................................... 111
Appendix 1.4: Gaussian Heat Distribution ....................................... 112
2. Chemical Reactions in Arc Welding ...................................... 116

2.1 Introduction ............................................................................... 116
2.2 Overall Reaction Model ............................................................. 116
2.3 Dissociation of Gases in the Arc Column .................................. 117
2.4 Kinetics of Gas Absorption ........................................................ 120
2.4.1 Thin Film Model ........................................................... 120
2.4.2 Rate of Element Absorption ......................................... 121
This page has been reformatted by Knovel to provide easier navigation.

viii Contents
2.5 The Concept of Pseudo-equilibrium .......................................... 122

2.6 Kinetics of Gas Desorption ........................................................ 123
2.6.1 Rate of Element Desorption ......................................... 123
2.6.2 Sievert’s Law ............................................................... 124
2.7 Overall Kinetic Model for Mass Transfer during Cooling in
the Weld Pool ............................................................................ 124
2.8 Absorption of Hydrogen ............................................................ 128
2.8.1 Sources of Hydrogen ................................................... 128
2.8.2 Methods of Hydrogen Determination in Steel
Welds ........................................................................... 128
2.8.3 Reaction Model ............................................................ 130
2.8.4 Comparison between Measured and Predicted
Hydrogen Contents ...................................................... 131
2.8.4.1 Gas-shielded Welding .................................. 131
2.8.4.2 Covered Electrodes ..................................... 134
2.8.4.3 Submerged Arc Welding .............................. 138
2.8.4.4 Implications of Sievert’s Law ....................... 140
2.8.4.5 Hydrogen in Multi-run Weldments ............... 140
2.8.4.6 Hydrogen in Non-ferrous Weldments .......... 141
2.9 Absorption of Nitrogen .............................................................. 141
2.9.1 Sources of Nitrogen ..................................................... 142
2.9.2 Gas-shielded Welding .................................................. 142
2.9.3 Covered Electrodes ..................................................... 143
2.9.4 Submerged Arc Welding .............................................. 146
2.10 Absorption of Oxygen ................................................................ 148
2.10.1 Gas Metal Arc Welding ................................................ 148
2.10.1.1 Sampling of Metal Concentrations at
Elevated Temperatures ............................... 149
2.10.1.2 Oxidation of Carbon ..................................... 149
2.10.1.3 Oxidation of Silicon ...................................... 152
2.10.1.4 Evaporation of Manganese .......................... 156
2.10.1.5 Transient Concentrations of Oxygen ........... 160

Contents ix
2.10.1.6 Classification of Shielding Gases ................ 166

2.10.1.7 Overall Oxygen Balance .............................. 166
2.10.1.8 Effects of Welding Parameters .................... 169
2.10.2 Submerged Arc Welding .............................................. 170
2.10.2.1 Flux Basicity Index ....................................... 171
2.10.2.2 Transient Oxygen Concentrations ............... 172
2.10.3 Covered Electrodes ..................................................... 173
2.10.3.1 Reaction Model ............................................ 174
2.10.3.2 Absorption of Carbon and Oxygen .............. 176
2.10.3.3 Losses of Silicon and Manganese ............... 177
2.10.3.4 The Product [%C] [%O] ............................... 179
2.11 Weld Pool Deoxidation Reactions ............................................. 180
2.11.1 Nucleation of Oxide Inclusions ..................................... 182
2.11.2 Growth and Separation of Oxide Inclusions ................. 184
2.11.2.1 Buoyancy (Stokes Flotation) ........................ 185
2.11.2.2 Fluid Flow Pattern ........................................ 186
2.11.2.3 Separation Model ......................................... 188
2.11.3 Predictions of Retained Oxygen in the Weld Metal ...... 190
2.11.3.1 Thermodynamic Model ................................ 190
2.11.3.2 Implications of Model ................................... 192
2.12 Non-metallic Inclusions in Steel Weld Metals ........................... 192
2.12.1 Volume Fraction of Inclusions ...................................... 193
2.12.2 Size Distribution of Inclusions ...................................... 195
2.12.2.1 Effect of Heat Input ...................................... 196
2.12.2.2 Coarsening Mechanism ............................... 196
2.12.2.3 Proposed Deoxidation Model ....................... 201
2.12.3 Constituent Elements and Phases in Inclusions .......... 202
2.12.3.1 Aluminium, Silicon and Manganese
Contents ...................................................... 202
2.12.3.2 Copper and Sulphur Contents ..................... 202
2.12.3.3 Titanium and Nitrogen Contents .................. 203
2.12.3.4 Constituent Phases ...................................... 204

x Contents
2.12.4 Prediction of Inclusion Composition ............................. 204

2.12.4.1 C-Mn Steel Weld Metals .............................. 204
2.12.4.2 Low-alloy Steel Weld Metals ........................ 206
References ......................................................................................... 212
Appendix 2.2: Derivation of Equation (2-60) .................................... 219
3. Solidification Behaviour of Fusion Welds ............................ 221

3.1 Introduction ............................................................................... 221
3.2 Structural Zones in Castings and Welds ................................... 221
3.3 Epitaxial Solidification ............................................................... 222
3.3.1 Energy Barrier to Nucleation ........................................ 225
3.3.2 Implications of Epitaxial Solidification .......................... 226
3.4 Weld Pool Shape and Columnar Grain Structures .................... 228
3.4.1 Weld Pool Geometry .................................................... 228
3.4.2 Columnar Grain Morphology ........................................ 229
3.4.3 Growth Rate of Columnar Grains ................................. 230
3.4.3.1 Nominal Crystal Growth Rate ...................... 230
3.4.3.2 Local Crystal Growth Rate ........................... 234
3.4.4 Reorientation of Columnar Grains ............................... 239
3.4.4.1 Bowing of Crystals ....................................... 240
3.4.4.2 Renucleation of Crystals .............................. 242
3.5 Solidification Microstructures .................................................... 251
3.5.1 Substructure Characteristics ........................................ 251
3.5.2 Stability of the Solidification Front ................................ 254
3.5.2.1 Interface Stability Criterion ........................... 254
3.5.2.2 Factors Affecting the Interface Stability ....... 256
3.5.3 Dendrite Morphology ................................................... 260
3.5.3.1 Dendrite Tip Radius ..................................... 260
3.5.3.2 Primary Dendrite Arm Spacing .................... 261
3.5.3.3 Secondary Dendrite Arm Spacing ............... 264

Contents xi
3.6 Equiaxed Dendritic Growth ....................................................... 268

3.6.1 Columnar to Equiaxed Transition ................................. 268
3.6.2 Nucleation Mechanisms ............................................... 272
3.7 Solute Redistribution ................................................................. 272
3.7.1 Microsegregation ......................................................... 272
3.7.2 Macrosegregation ........................................................ 278
3.7.3 Gas Porosity ................................................................ 279
3.7.3.1 Nucleation of Gas Bubbles .......................... 279
3.7.3.2 Growth and Detachment of Gas Bubbles .... 281
3.7.3.3 Separation of Gas Bubbles .......................... 284
3.7.4 Removal of Microsegregations during Cooling ............ 286
3.7.4.1 Diffusion Model ............................................ 286
3.7.4.2 Application to Continuous Cooling ............... 286
3.8 Peritectic Solidification .............................................................. 290
3.8.1 Primary Precipitation of the γp-phase ........................... 290
3.8.2 Transformation Behaviour of Low-alloy Steel Weld
Metals .......................................................................... 290
3.8.2.1 Primary Precipitation of Delta Ferrite ........... 290
3.8.2.2 Primary Precipitation of Austenite ................ 292
3.8.2.3 Primary Precipitation of Both Delta
Ferrite and Austenite ................................... 292
References ......................................................................................... 293
4. Precipitate Stability in Welds ................................................. 301

4.1 Introduction ............................................................................... 301
4.2 The Solubility Product ............................................................... 301
4.2.1 Thermodynamic Background ....................................... 301
4.2.2 Equilibrium Dissolution Temperature ........................... 303
4.2.3 Stable and Metastable Solvus Boundaries .................. 304
4.2.3.1 Equilibrium Precipitates ............................... 304
4.2.3.2 Metastable Precipitates ............................... 308

xii Contents
4.3 Particle Coarsening ................................................................... 314

4.3.1 Coarsening Kinetics ..................................................... 314
4.3.2 Application to Continuous Heating and Cooling ........... 314
4.3.2.1 Kinetic Strength of Thermal Cycle ............... 315
4.3.2.2 Model Limitations ......................................... 315
4.4 Particle Dissolution .................................................................... 316
4.4.1 Analytical Solutions ...................................................... 316
4.4.1.1 The Invariant Size Approximation ................ 319
4.4.1.2 Application to Continuous Heating and
Cooling ........................................................ 322
4.4.2 Numerical Solution ....................................................... 325
4.4.2.1 Two-dimensional Diffusion Model ................ 326
4.4.2.2 Generic Model ............................................. 328
4.4.2.3 Application to Continuous Heating and
Cooling ........................................................ 329
4.4.2.4 Process Diagrams for Single Pass 6082-
T6 Butt Welds .............................................. 332
References ......................................................................................... 334
5. Grain Growth in Welds ........................................................... 337

5.1 Introduction ............................................................................... 337
5.2 Factors Affecting the Grain Boundary Mobility .......................... 337
5.2.1 Characterisation of Grain Structures ............................ 337
5.2.2 Driving Pressure for Grain Growth ............................... 339
5.2.3 Drag from Impurity Elements in Solid Solution ............ 340
5.2.4 Drag from a Random Particle Distribution ................... 341
5.2.5 Combined Effect of Impurities and Particles ................ 342
5.3 Analytical Modelling of Normal Grain Growth ........................... 343
5.3.1 Limiting Grain Size ....................................................... 343
5.3.2 Grain Boundary Mobility ............................................... 345

Contents xiii
5.3.3 Grain Growth Mechanisms .......................................... 345

5.3.3.1 Generic Grain Growth Model ....................... 345
5.3.3.2 Grain Growth in the Absence of Pinning
Precipitates .................................................. 347
5.3.3.3 Grain Growth in the Presence of Stable
Precipitates .................................................. 348
5.3.3.4 Grain Growth in the Presence of Growing
Precipitates .................................................. 351
5.3.3.5 Grain Growth in the Presence of
Dissolving Precipitates ................................. 356
5.4 Grain Growth Diagrams for Steel Welding ................................ 360
5.4.1 Construction of Diagrams ............................................ 360
5.4.1.1 Heat Flow Models ........................................ 360
5.4.1.2 Grain Growth Model ..................................... 361
5.4.1.3 Calibration Procedure .................................. 361
5.4.1.4 Axes and Features of Diagrams .................. 363
5.4.2 Case Studies ............................................................... 364
5.4.2.1 Titanium-microalloyed Steels ....................... 364
5.4.2.2 Niobium-microalloyed Steels ....................... 367
5.4.2.3 C-Mn Steel Weld Metals .............................. 370
5.4.2.4 Cr-Mo Low-alloy Steels ................................ 372
5.4.2.5 Type 316 Austenitic Stainless Steels ........... 375
5.5 Computer Simulation of Grain Growth ...................................... 380
5.5.1 Grain Growth in the Presence of a Temperature
Gradient ....................................................................... 380
5.5.2 Free Surface Effects .................................................... 382
References ......................................................................................... 382
6. Solid State Transformations in Welds ................................... 387

6.1 Introduction ............................................................................... 387
6.2 Transformation Kinetics ............................................................ 387
6.2.1 Driving Force for Transformation Reactions ................ 387

xiv Contents
6.2.2 Heterogeneous Nucleation in Solids ............................ 389

6.2.2.1 Rate of Heterogeneous Nucleation .............. 389
6.2.2.2 Determination of ∆Ghet.* and Qd ................... 390
6.2.2.3 Mathematical Description of the C-curve ..... 392
6.2.3 Growth of Precipitates .................................................. 396
6.2.3.1 Interface-controlled Growth ......................... 396
6.2.3.2 Diffusion-controlled Growth ......................... 397
6.2.4 Overall Transformation Kinetics ................................... 400
6.2.4.1 Constant Nucleation and Growth Rates ...... 400
6.2.4.2 Site Saturation ............................................. 402
6.2.5 Non-isothermal Transformations .................................. 402
6.2.5.1 The Principles of Additivity ........................... 403
6.2.5.2 Isokinetic Reactions ..................................... 404
6.2.5.3 Additivity in Relation to the Avrami
Equation ...................................................... 404
6.2.5.4 Non-additive Reactions ................................ 405
6.3 High Strength Low-alloy Steels ................................................. 406
6.3.1 Classification of Microstructures .................................. 406
6.3.2 Currently Used Nomenclature ...................................... 406
6.3.3 Grain Boundary Ferrite ................................................ 408
6.3.3.1 Crystallography of Grain Boundary
Ferrite .......................................................... 408
6.3.3.2 Nucleation of Grain Boundary Ferrite .......... 408
6.3.3.3 Growth of Grain Boundary Ferrite ................ 422
6.3.4 Widmanstätten Ferrite .................................................. 427
6.3.5 Acicular Ferrite in Steel Weld Deposits ........................ 428
6.3.5.1 Crystallography of Acicular Ferrite ............... 428
6.3.5.2 Texture Components of Acicular Ferrite ...... 429
6.3.5.3 Nature of Acicular Ferrite ............................. 430
6.3.5.4 Nucleation and Growth of Acicular
Ferrite .......................................................... 432
6.3.6 Acicular Ferrite in Wrought Steels ............................... 444

Contents xv
6.3.7 Bainite .......................................................................... 444

6.3.7.1 Upper Bainite ............................................... 444
6.3.7.2 Lower Bainite ............................................... 447
6.3.8 Martensite .................................................................... 448
6.3.8.1 Lath Martensite ............................................ 448
6.3.8.2 Plate (Twinned) Martensite .......................... 448
6.4 Austenitic Stainless Steels ........................................................ 453
6.4.1 Kinetics of Chromium Carbide Formation .................... 456
6.4.2 Area of Weld Decay ..................................................... 456
6.5 Al-Mg-Si Alloys .......................................................................... 458
6.5.1 Quench-sensitivity in Relation to Welding .................... 459
6.5.1.1 Conditions for β’(Mg2Si) Precipitation
during Cooling .............................................. 459
6.5.1.2 Strength Recovery during Natural
Ageing ......................................................... 461
6.5.2 Subgrain Evolution during Continuous Drive
Friction Welding ........................................................... 464
References ......................................................................................... 467
Appendix 6.2: Additivity in Relation to the Avrami Equation ............ 475
7. Properties of Weldments ........................................................ 477

7.1 Introduction ............................................................................... 477
7.2 Low-alloy Steel Weldments ....................................................... 477
7.2.1 Weld Metal Mechanical Properties .............................. 477
7.2.1.1 Weld Metal Strength Level ........................... 478
7.2.1.2 Weld Metal Resistance to Ductile
Fracture ....................................................... 480
7.2.1.3 Weld Metal Resistance to Cleavage
Fracture ....................................................... 485
7.2.1.4 The Weld Metal Ductile to Brittle
Transition ..................................................... 486

xvi Contents
7.2.1.5 Effects of Reheating on Weld Metal

Toughness ................................................... 491
7.2.2 HAZ Mechanical Properties ......................................... 494
7.2.2.1 HAZ Hardness and Strength Level .............. 495
7.2.2.2 Tempering of the Heat Affected Zone .......... 500
7.2.2.3 HAZ Toughness ........................................... 502
7.2.3 Hydrogen Cracking ...................................................... 509
7.2.3.1 Mechanisms of Hydrogen Cracking ............. 509
7.2.3.2 Solubility of Hydrogen in Steel ..................... 513
7.2.3.3 Diffusivity of Hydrogen in Steel .................... 514
7.2.3.4 Diffusion of Hydrogen in Welds ................... 514
7.2.3.5 Factors Affecting the HAZ Cracking
Resistance ................................................... 518
7.2.4 H2S Stress Corrosion Cracking .................................... 524
7.2.4.1 Threshold Stress for Cracking ..................... 524
7.2.4.2 Prediction of HAZ Cracking Resistance ....... 525
7.3 Stainless Steel Weldments ....................................................... 527
7.3.1 HAZ Corrosion Resistance .......................................... 527
7.3.2 HAZ Strength Level ..................................................... 529
7.3.3 HAZ Toughness ........................................................... 530
7.3.4 Solidification Cracking .................................................. 532
7.4 Aluminium Weldments .............................................................. 536
7.4.1 Solidification Cracking .................................................. 536
7.4.2 Hot Cracking ................................................................ 540
7.4.2.1 Constitutional Liquation in Binary Al-Si
Alloys ........................................................... 541
7.4.2.2 Constitutional Liquation in Ternary Al-Mg-
Si Alloys ....................................................... 542
7.4.2.3 Factors Affecting the Hot Cracking
Susceptibility ................................................ 544
7.4.3 HAZ Microstructure and Strength Evolution during
Fusion Welding ............................................................ 547

Contents xvii
7.4.3.1 Effects of Reheating on Weld Properties ..... 547

7.4.3.2 Strengthening Mechanisms in Al-Mg-Si
Alloys ........................................................... 548
7.4.3.3 Constitutive Equations ................................. 548
7.4.3.4 Predictions of HAZ Hardness and
Strength Distribution .................................... 550
7.4.4 HAZ Microstructure and Strength Evolution during
Friction Welding ........................................................... 556
7.4.4.1 Heat Generation in Friction Welding ............ 556
7.4.4.2 Response of Al-Mg-Si Alloys and Al-SiC
MMCs to Friction Welding ............................ 557
7.4.4.3 Constitutive Equations ................................. 558
7.4.4.4 Coupling of Models ...................................... 558
7.4.4.5 Prediction of the HAZ Hardness
Distribution ................................................... 560
References ......................................................................................... 564
8. Exercise Problems with Solutions ......................................... 571

8.1 Introduction ............................................................................... 571
8.2 Exercise Problem I: Welding of Low Alloy Steels ...................... 571
8.3 Exercise Problem II: Welding of Austenitic Stainless Steels ..... 583
8.4 Exercise Problem III: Welding of Al-Mg-Si Alloys ...................... 587
Index .............................................................................................. 595
Author Index ................................................................................. 602

1
Heat Flow and Temperature Distribution in
Welding
1.1 Introduction
Welding metallurgy is concerned with the application of well-known metallurgical principles

for assessment of chemical and physical reactions occurring during welding. On purely prac-
tical grounds it is nevertheless convenient to consider welding metallurgy as a profession of its
own because of the characteristic non-isothermal nature of the process. In welding the reac-
tions are forced to take place within seconds in a small volume of metal where the thermal
conditions are highly different from those prevailing in production, refining and fabrication of
metals and alloys. For example, steel welding is characterised by:
High peak temperatures, up to several thousand 0 C.

High temperature gradients, locally of the order of 103 0C mm"1.
Rapid temperature fluctuations, locally of the order of 103 0C s 1 .
It follows that a quantitative analysis of metallurgical reactions in welding requires detailed

information about the weld thermal history. From a practical point of view the analytical
approach to the solution of heat flow problems in welding is preferable, since this makes it
possible to derive relatively simple equations which provide the required background for an
understanding of the temperature-time pattern. However, because of the complexity of the
heat flow phenomena, it is always necessary to check the validity of such predictions against
more reliable data obtained from numerical calculations and in situ thermocouple measure-
ments. Although the analytical models suffer from a number of simplifying assumptions, it is
obvious that these solutions in many cases are sufficiently accurate to provide at least a quali-
tative description of the weld thermal programme.
An important aspect of the present treatment is the use of different dimensionless groups
for a general outline of the temperature distribution in welding. Although this practice in-
volves several problems, it is a convenient way to reduce the total number of variables to an
acceptable level and hence, condense general information about the weld thermal programme
into two-dimensional (2-D) maps or diagrams. Consequently, readers who are unfamiliar with
the concept should accept the challenge and try to overcome the barrier associated with the use
of such dimensionless groups in heat flow analyses.
1.2 Non-Steady Heat Conduction
The symbols and units used throughout this chapter are defined in Appendix 1.1.
Since heat losses from free surfaces by radiation and convection are usually negligible in
welding, the temperature distribution can generally be obtained from the fundamental differ-
ential equations for heat conduction in solids. For uniaxial heat conduction, the governing
equation can be written as:1
(i-D
where T is the temperature, t is the time, x is the heat flow direction, and a is the thermal
diffusivity. The thermal diffusivity is related to the thermal conductivity X and the volume
heat capacity pc through the following equation:
(1-2)
For biaxial and triaxial heat conduction we may write by analogy:1
d-3)
and
(1-4)
The above equations must clearly be satisfied by all solutions of heat conduction problems,
but for a given set of initial and boundary conditions there will be one and only one solution.
1.3 Thermal Properties of Some Metals and Alloys
A pre-condition for obtaining simple analytical solutions to the differential heat flow equa-
tions is that the thermal properties of the base material are constant and independent of tem-
perature. For most metals and alloys this is a rather unrealistic assumption, since both X, a,
and pc may vary significantly with temperature as illustrated in Fig. 1.1. In addition, the
thermal properties are also dependent upon the chemical composition and the thermal history
of the base material (see Fig. 1.2), which further complicates the situation.
By neglecting such effects in the heat flow models, we impose several limitations on the
application of the analytical solutions. Nevertheless, experience has shown that these prob-
lems to some extent can be overcome by the choice of reasonable average values for X, a and
pc within a specific temperature range. Table 1.1 contains a summary of relevant thermal
properties for different metals and alloys, based on a critical review of literature data. It should
be noted that the thermal data in Table 1.1 do not include a correction for heat consumed in
melting of the parent materials. Although the latent heat of melting is temporarily removed
during fusion welding, experience has shown this effect can be accounted for by calibrating
the equations against a known isotherm (e.g. the fusion boundary). In practice, such correc-
tions are done by adjusting the arc efficiency factor Tq until a good correlation is achieved
between theory and experiments.
Carbon steel
Hx-H0 = PC(T-T0 ),J/mm3
Temperature, 0C
Fig. 1.1. Enthalpy increment H7-H0 referred to an initial temperature T0 = 200C. Data from Refs.
2-4.
Table 1.1 Physical properties for some metals and alloys. Data from Refs 2 - 6 .
Material (WrTIm-10C-1) (mm2 s"1) (Jmnr 3 0C"1) (0C) (J mnr 3 ) (J mnr 3 )
Carbon 0.040 8 0.005 1520 7.50 2.0

Steels
Low Alloy 0.025 5 0.005 1520 7.50 2.0
Steels
High Alloy 0.020 4 0.005 1500 7.40 2.0
Steels
Titanium 0.030 10 0.003 1650 4.89 1.4
Alloys
Aluminium 0.230 85 0.0027 660 1.73 0.8
(> 99% Al)
Al-Mg-Si 0.167 62 0.0027 652 1.71 0.8
Alloys
Al-Mg 0.149 55 0.0027 650 1.70 0.8
Alloys
Does not include the latent heat of melting (AH1n).

(a)
X9 W/mm 0C
Temperature, 0C
(b)
High alloy steel
X, W/mm 0C
Temperature, 0C
Fig. 1.2. Factors affecting the thermal conductivity X of steels; (a) Temperature level and chemical
composition, (b) Heat treatment procedure. Data from Refs. 2-4.
1.4 Instantaneous Heat Sources
The concept of instantaneous heat sources is widely used in the theory of heat conduction.1 It
is seen from Fig. 1.3 that these solutions are based on the assumption that the heat is released
instantaneously at time t - 0 in an infinite medium of initial temperature T0, either across a
plane (uniaxial conduction), along a line (biaxial conduction), or in a point (triaxial conduc-
tion). The material outside the heat source is assumed to extend to x = + °° for a plane source
in a long rod, to r = °° for a line source in a wide plate, or to R = °° for a point source in a heavy
slab. The initial and boundary conditions can be summarised as follows:
T-T0 = oo for t = O and x = O (alternatively r = O or R = O)
T-J 0 = O for t = O and x * O (alternatively r > O or 7? > O)
7-T 0 = O for O < t < oo when x = ± oo (alternatively r = oo or R = oo).
It is easy to verify that the following solutions satisfy both the basic differential heat flow
equations (1-1), (1-3) and (1-4) and the initial and boundary conditions listed above:
(i) Plane source in a long rod (Fig. 1.3a):

d-5)
where Q is the net heat input (energy) released at time t = O, and A is the cross section of the
rod.
(ii) Line source in a wide plate (Fig. 1.3b):
(1-6)
where d is the plate thickness.
(iii) Point source in a heavy slab (Fig. 1.3c):

(1-7)
Equations (1-5), (1-6) and (1-7) provide the required basis for a comprehensive theoretical
treatment of heat flow phenomena in welding. These solutions can either be applied directly
or be used in an integral or differential form. In the next sections a few examples will be given
to illustrate the direct application of the instantaneous heat source concept to problems related
to welding.
(a) T
Fig. 1.3. Schematic representation of instantaneous heat source models; (a) Plane source in a long rod.
(b) T
X y
(C)
Fig. 1.3.Schematic representation of instantaneous heat source models (continued); (b) Line source in a
wide plate, (c) Point source in a heavy slab.
1.5 Local Fusion in Arc Strikes
The series of fused metal spots formed on arc ignition make a good case for application of
equation (1-7).
Model
The model considers a point source on a heavy slab as illustrated in Fig. 1.4. The heat is
assumed to be released instantaneously at time t = 0 on the surface of the slab. This causes a
temperature rise in the material which is exactly twice as large as that calculated from equation
(1-7):
(1-8)
In order to obtain a general survey of the thermal programme, it is convenient to write

equation (1-8) in a dimensionless form. The following parameters are defined for this pur-
pose:
— Dimensionless temperature:
(1-9)
where Tc is the chosen reference temperature.
— Dimensionless time:
d-10)
where tt is the arc ignition time.
— Dimensionless operating parameter:
(1-11)
where qo is the net arc power (equal to Qlt(), and (Hc-Ho) is the heat content per unit volume at
the reference temperature.
— Dimensionless radius vector:
(1-12)
By substituting these parameters into equation (1-8), we obtain:
(1-13)
Heat source
Isotherms
3-D heat flow
Fig. 1.4. Instantaneous point source model for assessment of temperatures in arc strikes.
e/n Linear time scale

0Zn1
T1
î
Fig. 1.5. Calculated temperatures in arc strikes.
Equation (1-13) has been solved numerically for different values ofCT1and T1. The results
are presented graphically in Fig. 1.5. Due to the inherent assumption of instantaneous release
of heat in a point, it is not possible to use equation (1-13) down to very small values OfCT1 and
T1. However, at some distance from the heat source and after a time not much shorter than the
real (assumed) time of heating, the calculated temperature-time pattern will be reasonably
correct. Note that the heavy broken line in Fig. 1.5 represents the locus of the peak tempera-
tures. This locus is obtained by setting 3In(OAi1VdT1 = 0:
from which
Substituting this into equation (1-13) gives:
(1-14)
where Qp is the peak temperature, and e is the natural logarithm base number.
Example (1.1)
Consider a small weld crater formed in an arc strike on a thick plate of low alloy steel. Calcu-
late the cooling time from 800 to 5000C (Af875), and the total width of the fully transformed
region adjacent to the fusion boundary. The operational conditions are as follows:
where r| is the arc efficiency factor. Relevant thermal data for low alloy steel are given in
Table 1.1.
Solution
In the present case it is convenient to use the melting point of the steel as a reference tempera-
ture (i.e. 0 = 0m = 1 when Tc = TJ. The corresponding values OfZi1 and 9 (at 800 and 5000C,
respectively) are:
Cooling time At8/5

Since the cooling curves in Fig. 1.5 are virtually parallel at temperatures below 800 0 C, Af875
will be independent of Cr1 and similar to that calculated for the centre-line ((J1 = 0). By rear-
ranging equation (1-13) we get:
and
Total width offully transformed region

Zone widths can generally be calculated from equation (1-14), as illustrated in Fig. 1.6. Tak-
ing the Ac3-temperature equal to 8900C for this particular steel, we obtain:
and
Alternatively, the same information could have been read from Fig. 1.5. Although it is
difficult to check the accuracy of these predictions, the calculated values for Ats/5 and ARlm are
considered reasonably correct. Thus, because the cooling rate is very large, in arc strikes a
hard martensitic microstructure would be expected to form within the transformed parts of the
HAZ, in agreement with general experience.
1.6 Spot Welding
Equation (1-6) can be used for an assessment of the temperature-time pattern in spot welding
of plates.
Model
The model considers a line source which penetrates two overlapping plates of similar thermal
properties, as illustrated in Fig. 1.7. The heat is assumed to be released instantaneously at time
Heat source
Fig. 1.6. Definition of isothermal zone width in Example (1.1).

Electrode
Heat source d
Fig. 1.7. Idealised heat flow model for spot welding of plates.
t = 0. If transfer of heat into the electrodes is neglected, the temperature distribution is given
by equation (1-6).
This equation can be written in a dimensionless form by introducing the following group of
parameters:
— Dimensionless time:
(1-15)
where th is the heating time (i.e. the duration of the pulse).
(1-16)
where dt is the total thickness of the joint.
— Dimensionless radius vector:
(1-17)
By substituting these parameters into equation (1-6), we get:
(1-18)
where 6 denotes the dimensionless temperature (previously defined in equation (1-9)).

Linear time scale
e/n2
6/n2
T
2
T2
Fig. 1.8. Calculated temperature-time pattern in spot welding.
Figure 1.8 shows a graphical representation of equation (1-18) for a limited range of a 2 and
T2. A closer inspection of the graph reveals that the temperature-time pattern in spot welding
is similar to that observed during arc ignition (see Fig. 1.5). The locus of the peak tempera-
tures in Fig. 1.8 is obtained by setting d\n{^ln7}ldx2 - 0.
which gives
and
(1-19)
Example (1.2)
Consider spot welding of 2 mm plates of low alloy steel under the following operational con-
ditions:
Calculate the cooling time from 800 to 5000C (Af8/5) in the centre of the weld, and the cooling
rate (CR.) at the onset of the austenite to ferrite transformation. Assume in these calculations
that the total voltage drop between the electrodes is 1.6 V. The M^-temperature of the steel is
taken equal to 475°C.
Solution
If we use the melting point of the steel as a reference temperature, the parameters n2 and 6 (at
800 and 5000C, respectively) become:
Cooling time Atg/5

The parameter A%5 can be calculated from equation (1-18). For the weld centre-line (CT2 = 0),
we get:
and
Cooling rate at 475 0C

The cooling rate at a specific temperature is obtained by differentiation of equation (1-18) with
respect to time. When (J2 = 0 the cooling rate at 9 = 0.3 (475°C) becomes:
and
Since the cooling curves in Fig. 1.8 are virtually parallel at temperatures below 8000C (i.e.
for QZn2 < 0.15), the computed values of Ar8/5 and CR. are also valid for positions outside the
weld centre-line. In the present example the centre-line solutions can be applied down to
(°"2m)2 ~ 2. According to equation (1-19), this corresponds to a lower peak temperature of:
which is equivalent with:

It should be emphasised that the present heat flow model represents a crude oversimplifica-
tion of the spot welding process. In a real welding situation, most of the heat is generated at the
interface between the two plates because of the large contact resistance. This gives rise to the
development of an elliptical weld nugget inside the joint as shown in Fig. 1.9. Moreover, since
the model neglects transfer of heat into the electrodes, the mode of heat flow will be mixed and
not truly two-dimensional as assumed above. Consequently, equation (1-18) cannot be ap-
plied for reliable predictions of isothermal contours and zone widths. Nevertheless, the model
may provide useful information about the cooling conditions during spot welding if the effi-
ciency factor if] and the voltage drop between the electrodes can be estimated with a reasonable
degree of accuracy.
A more refined heat flow model for spot welding is presented in Appendix 1.2.
1.7 Thermit Welding
Thermit welding is a process that uses heat from exothermic chemical reactions to produce
coalescence between metals and alloys. The thermit mixture consists of two components, i.e.
a metal oxide and a strong reducing agent. The excess heat of formation of the reaction prod-
uct provides the energy source required to form the weld.
Model
In thermit welding the time interval between the ignition of the powder mixture and the com-
pletion of the reduction process will be short because of the high reaction rates involved.
Assume that a groove of width 2L1 is filled instantaneously at time t = 0 by liquid metal of an
initial temperature Tt (see Fig. 1.10). The metal temperature outside the fusion zone is T0. If
heat losses to the surroundings are neglected, the problem can be treated as uniaxial conduc-
tion where the heat source (extending from -L 1 to +L1) is represented by a series of elementary
sources, each with a heat content of:
(1-20)
At time t this source produces a small rise of temperature at position JC, given by equation (1 -5):
(1-21)
The final temperature distribution is obtained by substituting u = (x-xy(4at)m (i.e. dx'-

- du(4at)m) into equation (1-21) and integrating between the limits JC'= -L 1 and x'- +L1. This
gives (after some manipulation):
(1-22)
Isl'srau*'*'=]
Fusion zone
Fig. 1.9. Calculated peak temperature contours in spot welding of steel plates (numerical solution). Op-
erational conditions: / = 23kA, 64 cycles. Data from Bently et al1
Fusion zone
Fig. 1.10. Idealised heat flow model for thermit welding of rails.
where erf(u) is the Gaussian error function. The error function is defined in Appendix 1.3*.
Because of the complex nature of equation (1-22), it is convenient to present the different
solutions in a dimensionless form by introducing the following groups of parameters:
*The error function is available in tables. However, in numerical calculations it is more convenient to use the
Fortran subroutine given in Appendix 1.3.
Dimensionless temperature:
(1-23)
Dimensionless time:
(1-24)
Dimensionless jc-coordinate:
(1-25)
Substituting these parameters into equation (1-22) gives:
(1-26)
Equation (1-26) has been solved numerically for different values of Q and T3. The results
are presented graphically in Fig. 1.11. As would be expected, the fusion zone itself (Q < 1)
cools in a monotonic manner, while the temperature in positions outside the fusion boundary
(Q > 1) will pass through a maximum before cooling. The locus of the HAZ peak temperatures
in Fig. 1.11 is defined by 3673T3 = 0. Referring to Appendix 1.3, we may write:
which gives
(1-27)
The peak temperature distribution is obtained by solving equation (1-27) for different com-
binations of Qm and T3m and inserting the roots into equation (1-26).
Example (1.3)
Consider thermit welding of steel rails (i.e. reduction of Fe2O3 with Al powder) under the
following operational conditions:
Calculate the cooling time from 800 to 5000C in the centre of the weld, and the total width
of the fully transformed region adjacent to the fusion boundary. The Ac3-temperature of the
steel is taken equal to 8900C.
Definition of parameters:
91
T
3
Fig. 1.11. Calculated temperature-time pattern in thermit welding.
Solution
For positions along the weld centre-line (Q. = 0) equation (1-26) reduces to:
Cooling time At 8/5

From the above relation it is possible to calculate the cooling time from Tt = 22000C to 800 and
5000C, respectively:
and
By rearranging equation (1-24), we obtain the following expression for Ar875:

The computed value for A/8/5 is also valid for positions outside the weld centre-line, since
the cooling curves at such low temperatures are reasonably parallel within the fusion zone.
Total width of fully transformed region

The fusion boundary is defined by:
The locus of the 8900C isotherm in temperature-time space can be read from Fig. 1.11.
Taking the ordinate equal to 0.40, we get:
By inserting this value into equation (1-27), we obtain the corresponding coordinate of the
isotherm:
The total width of the fully transformed HAZ is thus:
Unfortunately, measurements are not available to check the accuracy of these predictions.
Systematic errors would be expected, however, because of the assumption of instantaneous
release of heat immediately after powder ignition and the neglect of heat losses to the sur-
roundings. Nevertheless, the present example is a good illustration of the versatility of the
concept of instantaneous heat sources, since these solutions can easily be added in space as
shown here or in time for continuous heat sources (to be discussed below).
1.8 Friction Welding
Friction welding is a solid state joining process that produces a weld under the compressive
force contact of one rotating and one stationary workpiece. The heat is generated at the weld
interface because of the continuous rubbing of the contact surfaces, which, in turn, causes a
temperature rise and subsequent softening of the material. Eventually, the material at the
interface starts to flow plastically and forms an up-set collar. When a certain amount of up-
setting has occurred, the rotation is stopped and the compressive force is maintained or slightly
increased to consolidate the weld.
Model (after Rykalin et al.5j

The model considers a continuous (plane) heat source in a long rod as shown in Fig. 1.12(a).
The heat is liberated at a constant rate q'o in the plane x = 0 starting at time / = 0. If we
subdivide the time t during which the source operates into a series of infinitesimal elements dt/
(Fig. 1.12b), each element will have a heat content of:
(1-28)
(a)
Continuous heat source
(b) q
t
Fig. 1.12. Idealised heatflowmodel for friction welding of rods; (a) Sketch of model, (b) Subdivision of
time into a series of infinitesimal elements dt'.
At time / this heat will cause a small rise of temperature in the material, in correspondance
with equation (1-5):
(1-29)
If we substitute t"=t-1'into equation (1 -29), the total temperature rise at time t is obtained
by integrating from t"= t (t'= 0) to /"= 0 (t'= t):
(1-30)
In order to evaluate this integral, we will make use of the following mathematical transfor-
mation:
where
and
Hence, we may write:
The latter integral can be expressed in terms of the complementary error function* erfc{u)
by substituting:
and integrating between the limits u = x I (4at)l/2 and w = <*>.

This gives (after some manipulation):
(1-31)
If the temperature of the contact section at the end of the heating period is taken equal to Th,
equation (1-31) can be rewritten as:
(1-32)
where t'h denotes the duration of the heating period (t < t'h). Measured contact section tem-
peratures for different metal/alloy combinations are given in Table 1.2.
Equation (1-32) may be presented in a dimensionless form by the use of the following
groups of parameters:
Dimensionless temperature:
(1-33)
Dimensionless time:
(1-34)
The complementary error function is defined in Appendix 1.3.

Table 1.2 Measured contact section temperatures during friction welding of some metals and alloys.
Data from Tensi et al.10
Metal/Alloy Measuring Temperature Level Partial

Combination Method [0C] Melting
Steel Thermocouples 1080-1340 No

1
Steel-Nickel Direct readings 1260-1400 No/Yes
Steel-Titanium Direct readings1 1080 No
1
Copper-Al Direct readings 548 Yes
1
Copper-Nickel Direct readings 1083 Yes
Al-Cu-2Mg Thermocouples 506 Yes
Al-4.3Cu Thermocouples 562 Yes
Al-12Si Thermocouples 575 Yes
Al-5Mg Thermocouples 582 Yes
Based on direct readings of the voltage drop between the two work-pieces.
— Dimensionless .^-coordinate:
(1-35)
(1-36)
Equation (1-36) describes the temperature in different positions from the weld contact sec-
tion during the heating period. However, when the rotation stops, the weld will be subjected to
free cooling, since there is no generation of heat at the interface. As shown in Fig. 1.13(a) this
can be accounted for by introducing an imaginary heat source of power +qo at time t = t'h
which acts simultaneously with an imaginary heat sink of negative power -q o. It follows from
the principles of superposition (see Fig. 1.13b) that the temperature during the cooling period
is given by:9
(1-37)
where 6"(x4) and 6"(T 4 - 1) are the temperatures calculated for the heat source and the heat
sink, respectively, using equation (1-36).
Equations (1-36) and (1-37) have been solved numerically for different values of Q'and T4.
The results are presented graphically in Fig. 1.14. Considering the contact section (Q'= 0), the
temperature increases monotonically with time during the heating period, in correspondance
with the relationship:
(1-38)
q
(a)
Imaginary heat source

Real heat
source
t
Imaginary heat sink
e"
(b)
Heating
period
$ffl9
Fig. 1.13. Method for calculation of transient temperatures during friction welding; (a) Sketch of
imaginary heat source/heat sink model, (b) Principles of superposition.
Similarly, for the cooling period we get:
(1-39)
Outside the contact section (Q / > 0), the temperature rise will be smaller and the cooling
rate lower than that calculated from equations (1-38) and (1-39).
Heating Cooling
e"
Fig. 1.14. Calculated temperature-time pattern in friction welding.
Example (1.4)
Consider friction welding of 026mm aluminium rods (Al-Cu-2Mg) under the following con-
ditions:
Calculate the peak temperature distribution across the joint. Assume in these calculations
that the thermal diffusivity of the Al-Cu-2Mg alloy is 70mm2 s"1.
Solution
Readings from Fig. 1.14 give:
Next Page
In this particular case, it is possible to check the accuracy of the calculations against in situ
thermocouple measurements carried out on friction welded components made under similar
conditions. A comparison with the data in Fig. 1.15 shows that the model is quite successful in
predicting the HAZ peak temperature distribution. In contrast, the weld heating and cooling
cycles cannot be reproduced with the same degree of precision. This has to do with the fact
that the present analytical solution omits a consideration of the plastic straining occurring
during friction welding, which displaces the coordinates and alters the heat balance for the
system.
1.9 Moving Heat Sources and Pseudo-Steady State
In most fusion welding processes the heat source does not remain stationary. In the following
we shall assume that the source moves at a constant speed along a straight line, and that the net
power supply from the source is constant. Experience shows that such conditions lead to a
fused zone of constant width. This is easily verified by moving a tungsten arc across a sheet of
steel or aluminium, or by moving a soldering iron across a piece of lead or tin. Moreover,
zones of temperatures below the melting point also remain at constant width, as indicated by
the pattern of temper colours developed on welding ground or polished sheet.
It follows from the definition of pseudo-steady state that the temperature will not vary with
time when observed from a point located in the heat source. Under such conditions the tem-
perature field around the source can be described as a temperature 'mountain' moving in the
direction of welding (e.g. see Fig. 15 in Ref. 11). For points along the weld centre-line, the
temperature at different positions away from the heat source (which for a constant welding
speed becomes a time axis) may be presented in a two-dimensional plot as indicated in Fig.
1.16. Specifically, this figure shows a schematic representation of the temperature in steel
welding from the base plate ahead of the arc to well into the solidified weld metal trailing the
arc. If we consider a fixed point on the weld centre-line, the temperature will increase very
rapidly during the initial period, reaching a maximum of about 2000-22000C for positions
immediately beneath the root of the arc.11 When the arc has passed, the temperature will start
to fall, and eventually (after long times) approach that of the base plate. In contrast, an ob-
server moving along with the heat source will always see the same temperature landscape,
since this will not change with time according to the presuppositions.
It will be shown below that the assumption of pseudo-steady state largely simplifies the
mathematical treatment of heat flow during fusion welding, although it imposes certain re-
strictions on the options of the models.
1.10 Arc Welding
Arc welding is a collective term which includes the following processes*:
- Shielded metal arc (SMA) welding.

- Gas tungsten arc (GTA) welding.
- Gas metal arc (GMA) welding.
*The terminology used here is in accordance with the American Welding Society's recommendations. 12
Previous Page
In this particular case, it is possible to check the accuracy of the calculations against in situ
thermocouple measurements carried out on friction welded components made under similar
conditions. A comparison with the data in Fig. 1.15 shows that the model is quite successful in
predicting the HAZ peak temperature distribution. In contrast, the weld heating and cooling
cycles cannot be reproduced with the same degree of precision. This has to do with the fact
that the present analytical solution omits a consideration of the plastic straining occurring
during friction welding, which displaces the coordinates and alters the heat balance for the
system.
1.9 Moving Heat Sources and Pseudo-Steady State
In most fusion welding processes the heat source does not remain stationary. In the following
we shall assume that the source moves at a constant speed along a straight line, and that the net
power supply from the source is constant. Experience shows that such conditions lead to a
fused zone of constant width. This is easily verified by moving a tungsten arc across a sheet of
steel or aluminium, or by moving a soldering iron across a piece of lead or tin. Moreover,
zones of temperatures below the melting point also remain at constant width, as indicated by
the pattern of temper colours developed on welding ground or polished sheet.
It follows from the definition of pseudo-steady state that the temperature will not vary with
time when observed from a point located in the heat source. Under such conditions the tem-
perature field around the source can be described as a temperature 'mountain' moving in the
direction of welding (e.g. see Fig. 15 in Ref. 11). For points along the weld centre-line, the
temperature at different positions away from the heat source (which for a constant welding
speed becomes a time axis) may be presented in a two-dimensional plot as indicated in Fig.
1.16. Specifically, this figure shows a schematic representation of the temperature in steel
welding from the base plate ahead of the arc to well into the solidified weld metal trailing the
arc. If we consider a fixed point on the weld centre-line, the temperature will increase very
rapidly during the initial period, reaching a maximum of about 2000-22000C for positions
immediately beneath the root of the arc.11 When the arc has passed, the temperature will start
to fall, and eventually (after long times) approach that of the base plate. In contrast, an ob-
server moving along with the heat source will always see the same temperature landscape,
since this will not change with time according to the presuppositions.
It will be shown below that the assumption of pseudo-steady state largely simplifies the
mathematical treatment of heat flow during fusion welding, although it imposes certain re-
strictions on the options of the models.
1.10 Arc Welding
Arc welding is a collective term which includes the following processes*:
- Shielded metal arc (SMA) welding.

- Gas tungsten arc (GTA) welding.
- Gas metal arc (GMA) welding.
*The terminology used here is in accordance with the American Welding Society's recommendations. 12
- Flux cored arc (FCA) welding.
- Submerged arc (SA) welding.
The main purpose of this section is to review the classical models for the pseudo-steady
state temperature distribution around moving heat sources. The analytical solutions to the
differential heat flow equations under conditions applicable to arc welding were first presented
(a)
Heating period Cooling period

Temperature, 0C
Predicted heating and cooling cycles

for the contact section (x=0)
Time, s
(b)
Peak temperature, 0C
Observed relationship
Predicted relationship
Distance from contact section, mm
Fig. 1.15. Comparison between measured and predicted temperatures in friction welding of Al-Cu-2Mg
alloys; (a) Temperature-time pattern, (b) Peak temperature distribution. Data from Tensi et al.10
by Rosenthal,1314 but the theory has later been extended and refined by a number of other
inve stigators .9'n*15"20
1.10.1 Arc efficiency factors

In arc welding heat losses by convection and radiation are taken into account by the efficiency
factor r\, defined as:
(1-40)
where qo is the net power received by the weldment (e.g. measured by calorimetry), / is the
welding current (amperage), and U is the arc voltage.
For submerged arc (SA) welding the efficiency factor (r\) has been reported in the range
from 90 to 98%, for SMA and GMA welding from 65 to 85%, and for GTA welding from 22
to 75%, depending on polarity and materials.11
A summary of ranges is given in Table 1.3.
7.70.2 Thick plate solutions
Model (after Rykalin9)

According to Fig. 1.17, the general thick plate model consists of an isotropic semi-infinite
body at initial temperature T0 limited in one direction by a plane that is impermeable to heat.
At time t = 0 a. point source of constant power qo starts on the surface at position O moving in
the positive x-direction at a constant speed U The rise of temperature T- T0 in point P at time
t is sought.
During a very short time interval from ^'to t'+ dt'the amount of heat released at the surface
is dQ = qodt'. According to equation (1-7) this will produce an infinitesimal rise of tempera-
ture in P at time t:
(1-41)
where is the time available for conduction of heat over the distance
to point P.
For a convenient presentation of the pseudo-steady state solution, the position P should be
referred to that of the moving heat source. This is achieved by changing the coordinate system
from O to O'(see Fig. 1.17):
and

T
Weld pool
Root of arc
Solidified
weld metal
X=Vt
Relative position along weld centre-line
Fig. 1.16. Schematic diagram showing weld centre-line temperature in different positions from the heat
source during steel welding at pseudo-steady state.
Table 1.3 Recommended arc efficiency factors for different welding processes. Data from Refs 11,21.
Arc efficiency factor j]

Welding Process Range Mean
SA welding (steel) 0.91-O.99 0.95
SMA welding (steel) 0.66-0.85 0.80
GMA welding (CO2-steel) 0.75-0.93 0.85
GMA welding (Ar-steel) 0.66-0.70 0.70
GTA welding (Ar-steel) 0.25-0.75 0.40
GTA welding (He-Al) 0.55-0.80 0.60
GTA welding (Ar-Al) 0.22-0.46 0.40

(1-42)
where
The total rise of temperature at P is obtained by substituting:
and
into equation (1-42), and integrating between the limits u = (R2l4af)m and u = o°.This gives
(after some manipulation):
(1-43)
It is well-known that:
Hence, the general thick plate solution can be written as:
(1-44)
If u is sufficiently small (i.e. when welding has been performed over a sufficient period), we
obtain the pseudo-steady state temperature distribution:
(1-45)
This equation is often referred to as the Rosenthal thick plate solution,1314 in honour of
D. Rosenthal who first derived the relation by solving the differential heat flow equation di-
rectly for the appropriate boundary conditions.
1.10.2.1 Transient heating period

It follows from the above analysis that the pseudo-steady state temperature distribution is
Fig. 1.17. Moving point source on a semi-infinite slab.
attained after a transient heating period. The duration of this heating period is determined by
the integral in equation (1-44).
Taking the ratio between the real and the pseudo-steady state temperature equal to K1, we
have:
(1-46)
Equation (1-46) can be expressed in terms of the following parameters:
Dimensionless radius vector:
(1-47)
Dimensionless time:
(1-48)
Substituting these into equation (1-46) gives:
(1-49)
where and
Equation (1-49) has been solved numerically for a limited range of cr3 and x. The results are
presented graphically in Fig. 1.18. A closer inspection of Fig. 1.18 reveals that the duration of
the transient heating period depends on the dimensionless radius vector a 3 . In practice this
means that the Rosenthal equation is not valid during the initial period of welding unless the
distance from the heat source to the observation point is very small. It should be noted, how-
ever, that a dimensionless distance o~3 may be 'short' for one combination of welding speed
and thermal diffusivity, while the same position may represent a 'long' distance for another
combination of V and a. Similarly, the dimensionless time T may be 'short' or 'long' at a
chosen number of seconds, depending on the ratio v/2a.
Example (1.5)
Consider stringer bead deposition on a thick plate of aluminium at a constant welding speed of
5 mm s"1. Calculate the duration of the heating period when the distance from the heat source
to the point of observation is 17 mm.
Solution
Taking a = 85 mm2 s"1, the dimensionless radius vector becomes:
Ki = (T-T0)/(T-T0)p.s.
T = v2t/2a
Fig. 1.18. Ratio between real and pseudo-steady state temperature in thick plate welding for
different combinations ofCT3and T.
It is seen from Fig. 1.18 that the pseudo-steady state temperature distribution is approached
when T ~ 3, which gives:
This corresponds to a total bead length of:
The above calculations show that the Rosenthal equation is not valid if the ratio between R
and L2 exceeds a certain critical value (typically 0.15 to 0.30 for aluminium welds and 0.4 to
0.6 for steel welds). This important point is often overlooked when discussing the relevance of
the thick plate solution in arc welding.
1.10.2.2 Pseudo-steady state temperature distribution

The Rosenthal equation gives, with the limitations inherent in the assumptions, full informa-
tion on the thermal conditions for point sources on heavy slabs. Accordingly, in order to obtain
a general survey of the pseudo-steady state temperature distribution, it is convenient to present
the different solutions in a dimensionless form. The following parameters are defined for this
purpose:11
(1-50)
Dimensionless jc-coordinate:
(1-51)
Dimensionless ^-coordinate:
(1-52)
Dimensionless z-coordinate:
(1-53)

(1-54)
where 6 and a 3 are the dimensionless temperature and radius vector, respectively (previously
defined in equations (1-9) and (1-47)).
Equation (1-54) has been solved numerically for chosen values of a 3 and £. A graphical
presentation of the different solutions is shown in Fig. 1.19. These maps provide a good
(a)
e/n3
%
Fig, /./P.Dimensionless temperature maps for point sources on heavy slabs; (a) Vertical sections paral-
lel to the ^-axis.
overall indication of the thermal conditions during thick plate welding, but are not suitable for
precise readings. Consequently, for quantitative analyses, the following set of equations can
be used:11
Isothermal zone widths

The maximum width of an isothermal enclosure is obtained by setting 3ln(0M3)/9a3 = 0:
From the definition of a 3 we have:

(b)
Fig. /.iP.Dimensionless temperature maps for point sources on heavy slabs (continued): (b) Isothermal
contours in the £-\}/-plane for different ranges of 0/n3.
Partial differentiation of the Rosenthal equation gives:

and
(1-55)
Equation (1-55) can be used for calculations of isothermal zone widths V|/w and cross sec-
tional areas A1. From Fig. 1.20 we have:
(1-56)
and
(1-57)
A graphical presentation of equations (1-55), (1-56), and (1-57) is shown in Fig. 1.21.
Length of isothermal enclosures

Referring to Fig. 1.20, the total length of an isothermal enclosure £r is given by:
(1-58)
where £' and £"are the distance from the heat source to the front and the rear of the enclosure,
respectively.
x,S
Heat source
y.v
z, C
Fig. 1.20.Three-dimensional graphical representation of Rosenthal thick plate solution (schematic).

Vm,O3m,Al
1
V9P
Fig. 1.21. Dimensionless distance a3m, half width \|/m and cross sectional area A1 vs n3 /Qp.
The coordinates £' and £" are found by setting a 3 = ± ^ in equation (1-54). This gives:
(1-59)
and
(1-60)
Volume of isothermal enclosures

Since the assumption of a point heat source involves semi-circular isotherms in the \|/-£ plane,
the volume of an isothermal enclosure is obtained by integration over the total length from £"
tor:
(1-61)
The former integral is readily evaluated by substituting:

which follows from a differentiation of equation (1-54). Hence, we may write:
(1-62)
Noting that <r3 = - £"= n3 /0 at the lower limit of integration, we obtain:
(1-63)
The dimensionless volume T is related to the real volume of the enclosure V(mm3) through
the following equation:
(1-64)
Cooling conditions close to weld centre-line

For points located on the weld centre-line behind the heat source \|/ = £ = 0, and
Hence, equation (1-54) reduces to*:
d-65)
Equation (1-65) provides a basis for calculating the cooling time within a specific tempera-
ture interval (e.g. from O1 to B2):
(1-66)
For welding of low alloy steel the cooling time from 800 to 5000C is widely accepted as an
adequate index for the thermal conditions under which the austenite to ferrite transformation
takes place. From equation (1-66), we have:
(1-67)
(1-69)
Similarly, the cooling rate at a specific temperature is obtained by differentiating equation

(1-65) with respect to time:
(1-70)
By multiplying equation (1-70) with the appropriate conversion factor, we get:
(1-71)
Example (1.6)
Consider stringer bead deposition (GMAW) on a thick plate of low alloy steel under the fol-
lowing conditions:
Sketch the contours of the fusion boundary and the Ac3-isotherm (9100C) in the £-\|/ (x-y)
plane at pseudo-steady state.
Solution
As shown in Fig. 1.22(a) it is sufficient to calculate the coordinates in four different (character-
istic) positions to draw a sketch of the isothermal contours. If we neglect the latent heat of
melting, the d/n3 ratio at the melting point becomes, according to equations (1-9) and (1-50):
End-points
The end-points are readily obtained from equations (1-59) and (1-60):
and
Maximum width
The maximum width of an isothermal enclosure can generally be calculated from equations
(1-55) and (1-56), or read from Fig. 1.21. Taking O^ /n3 = QIn3 = 0.088, we obtain:
(a) V
Pos. (3)
Pos. (2)
Pos. (4) Pos.(1)
%
(b) ¥
\
x(mm)
y(mm)
Fig. 1.22. Calculation of isothermal contours from Rosenthal thick plate solution; (a) Calculation proce-
dure, (b) Sketch of fusion boundary and Acj-isotherm in position £ = 0 (Example 1.6).
and
Intersection point with y/ (y)-axis
In this case £ = £ = O, and cr3 = \|/. Hence, equation (1-54) reduces to:
which gives
Similarly, the contour of the Ac3-isotherm can be determined by inserting BZn3= 0.052 into
the same set of equations. Figure 1.22(b) shows a graphical representation of the computed
isothermal contours.
Example (1.7)
Consider GTA welding on a thick plate of low alloy steel under the following conditions:
Calculate the weld pool volume, the weld bead cross section, the width of the fully trans-
formed HAZ, the cooling time from 800 to 5000C, and the cooling rate at the onset of the
austenite to ferrite transformation (e.g. at 6500C). The arc efficiency factor is taken equal to
0.5.
Solution
If we neglect the latent heat of melting, the GM3 ratio at the melting point becomes:
Weld pool volume

The coordinate £' may be calculated from equation (1-59):
We can now use equations (1-63) and (1-64) to calculate the weld pool volume. This gives:
and
Weld bead cross section
The weld bead cross section can be read from Fig. 1.21. Taking n3 /6 p = 1/0.445 = 2.196, we
obtain:
and
Width of fully transformed HAZ

When Tp = 9100C, the n3 /Qp ratio becomes:
From Fig. 1.21 we have:
which gives
Cooling time from 800 to 5000C

The cooling time, Af8/5, can be obtained from equation (1-68):
This value is also valid for positions outside the weld centre-line, since the cooling curves
at such low temperatures are reasonably parallel down to Tp ~ 9100C (see Fig. 1.19(a)).
Cooling rate at 6500C

The cooling rate at a specific temperature is given by equation (1-71). Taking the transforma-
tion start temperature equal to 6500C, we get:
1. 10.2.3 Simplified solution for a fast-moving high power source

It follows from Fig. 1.19(b) that the isotherms behind the heat source become increasingly
elongated as the arc power q0 and the welding speed v increase. In the limiting case, when
qo —> oo? v —» oo and qo A) remains finite, the isotherms will degenerate into surfaces which are
parallel to the welding x direction as shown in Fig. 1.23(a). Conduction of heat will then occur
exclusively in directions normal to the jc-axis.
In a short time interval dt, the amount of heat released per unit length of the weld is equal to:
(1-72)
According to the assumptions this amount of heat will remain in a slice of thickness dx due
to the lack of a temperature gradient in the welding direction. Since symmetry demands that
the isotherms in the y-z plane are semi-circles*, the situation becomes identical to the tempera-
ture field around a linear instantaneous heat source in a thin plate, provided that the space
above the slab is replaced by solid material and the strength of the source is doubled (see Fig.
1.23(b)). The solution is then given by equation (1-6) if we replace QId by Iq0 Iv:
d-73)
where r* is the two-dimensional radius vector in the y-z plane.

Equation (1-73) represents the simplified solution for a fast moving high power source on a
semi-infinite slab, and is valid within a limited range of the more general Rosenthal equation
for three-dimensional heat flow (equation (1-45)). By introducing the dimensionless radius
vector Cr4 = Dr*12a, equation (1-73) transforms to:
(1-74)
A graphical presentation of equation (1-74) gives a family of curves which resembles the
thermal cycles shown in Fig. 1.19(a). Although the cooling conditions close to the weld cen-
tre-line are similar to those calculated from the Rosenthal equation, the predicted width/depth
of the isotherms will always be greater than that inferred from the general thick plate solution
as illustrated in Fig. 1.24 due to the assumption of 2-D heat flow. The parameter o~4m in Fig.
1.24 is obtained by differentiating equation (1-74) with respect to time:
*The isotherms must meet the plate surface at right angles in the absence of a temperature gradient across the bound-
ary.
(a)
(b)
Symmetry
plane
Fig. 1.23. Fast moving high power source on a semi-infinite slab; (a) Sketch of model, (b) Analogy be-
tween a fast moving high power source and an instantaneous line source.
4m m
Asymptote
0 /n
P 3
Fig. 1.24. Theoretical width of isotherms under 2-D and 3-D heat flow conditions, respectively at pseudo-
steady state (thick plate welding).
which gives
Substituting this into equation (1-74) gives:
(1-75)
It is interesting to note that the dimensionless width a4m will approach \j/ m when the dp /n3
ratio becomes sufficiently small (i.e. less than about 0.1). Under such conditions the isotherms
will be strongly elongated in the welding direction (see Fig. 1.19(b)), which forces the heat to
flow primarily in directions normal to the x-axis.
A general graphical representation of the weld thermal programme can be obtained by
combining equations (1-74) and (1-75):
(1-76)
Equation (1-76) has been plotted in Fig. 1.25. This graph provides a basis for calculating
the retention time within specific temperature intervals under various welding conditions.
e/ep
W ° J
Fig. 1.25. Temperature-time pattern in thick plate welding at high arc power and high welding speed.
Example (1.8)
Consider SA welding on a thick plate of low alloy steel under the following conditions:
Calculate the retention time within the austenite regime (T > 9100C) for points located lmm
outside the fusion boundary.
Solution
If we use the melting point of the steel as a reference temperature, the parameter n3 becomes:
A comparison with Fig. 1.24 shows that the assumption of 2-D heat flow is justified when
dp < 1. Hence, the width of the fusion zone (8p = 1) can be calculated from equation (1-75):
which gives
The peak temperature lmm outside the fusion boundary is thus:
The total time spent in the thermal cycle from 0 = 0.59 (T= 9100C) to G^ = 0.78 and down
again to 0 = 0.59 can now be read from Fig. 1.25. Taking the ordinate 0/0^ equal to 0.76, we
obtain:
which gives
and
1.10.3 Thin plate solutions

As shown in Fig. 1.26, the general thin plate model considers a line source in a wide sheet of
thickness d and initial temperature T0. At time t = 0 the source starts to move at a constant
speed D in the positive x-direction. The rise of temperature T-T0 in point P at time t is sought.
According to equation (1-6) the elementary source dQ = qo df released at position Vt' will
cause a small rise of temperature dT in point P at time t:
(1-77)
where is the time available for conduction of heat over the distance
to point P.
If we refer the position P to that of the heat source at time t, we shall expect a solution
independent of time at pseudo-steady state. This is achieved by changing the coordinate sys-
tem from O to O' (see Fig. 1.26):
Fig. 1.26. Moving line source in a thin sheet.
K0(U)1K1(U)
U
Fig. 1.27. Graphical representation of the Bessel functions Ko(u) and Kx{u).
d-78)
where
For integration of all contributions from to , we introduce:
from which
Substituting these parameters into equation (1-78) give:
d-79)
It is well-known that:
where K0(G5) is the modified Bessel function of the second kind and zero order. A graphical
representation of A^(W) is shown in Fig. 1.27.
Hence, the general thin plate solution can be written as:
(1-80)
When w is sufficiently large (i.e. when welding has been performed over a sufficient pe-
riod), we obtain the pseudo-steady state temperature distribution:
(1-81)
Equation (1-81) is often referred to as the Rosenthal thin plate solution, in memory of its
originator D. Rosenthal.1314 It follows that this model is applicable to all types of welding
processes (including electron beam, plasma arc and laser welding), provided that a full through-
thickness penetration is achieved in one pass.
1.10.3.1 Transient heating period

In thin plate welding the duration of the transient heating period is determined by the inte-
gral in equation (1-80). Taking the ratio between the real and pseudo-steady state tempera-
ture equal to K2, we obtain:
(1-82)
where w = x/2, and dw = dill.

Equation (1-82) has been solved numerically for a limited range of cr5 and x. The results are
presented graphically in Fig. 1.28.
Example (1.9)
Consider butt welding of a thin aluminium plate at a constant travel speed of 5mm s"1. Calcu-
late the duration of the transient heating period when the distance from the heat source to the
point of observation is 17mm.
Solution
Taking a = 85 mm2 s"1, the dimensionless radius vector becomes:
It follows from Fig. 1.28 that the pseudo-steady state temperature distribution is approached
when T ~ 5, which gives:
and
This minimum bead length is nearly twice as large as that calculated for 3-D heat flow for
the same combination of welding speed, thermal diffusivity and radius vector (see Example
K2=(T-To)/(T-To)p.s.
t = v2t/2a
Fig. 1.28. Ratio between real and pseudo-steady state temperature in thin plate welding for different
combinations of a 5 and T.
(1.5)). Consequently, the duration of the transient heating period is significantly longer in thin
plate welding than in thick plate welding due the pertinent differences in the heat flow condi-
tions.
1.10.3.2 Pseudo-steady state temperature distribution

A graphical presentation of the pseudo-steady state temperature distribution similar to that
shown in Fig. 1.19 for three-dimensional heat flow may be obtained by introducing the
dimensionless plate thickness 8 = vdlla, which is a measure of the relative speeds of the arc
and the heat flow in the material. By rearranging equation (1-81), we get:
(1-83)
Plots of this equation are shown in Fig. 1.29. It follows that the pseudo-steady state tem-
perature distribution in thin plate welding depends on the parameter 68//?3. In practice, this
means that the shape of the isotherms is not influenced by the welding speed, since both 8 and
n3 are proportional to v.
Isothermal zone widths

The maximum width of an isothermal enclosure is obtained by setting 3(68 /n3) /die, = 0. Not-
ing that (d/du)Ko(u) - -Kx(u), where K1 (u) is the modified Bessel function of the second kind
and first order (shown in Fig. 1.27), we get:
(1-84)
\
\
%
\
V
Fig. 1.29. Dimensionless temperature maps for line sources in thin plates for different ranges of 98/n3.
Heat source
Fig. 1.30. Graphical representation of Rosenthal thin plate solution (schematic)
This gives:
(1-85)
and
(1-86)
Equation (1-86) can be used for calculations of isothermal zone widths (v|/m) and cross
sectional areas (A2) in thin plate welding. Referring to Fig. 1.30, we have:
d-87)
and
(1-88)
Figure 1.31 shows a graphical presentation of equations (1-86), (1-87), and (1-88).
Length of isothermal enclosures

The distance from the heat source to the front (£') and the rear (^') of an isothermal enclo-
sure is obtained by substituting o~5 = ± £ into equation (1-83).
This gives:
(1-89)
and
(1-90)
^5m.Vm-A 2 /8
n3/6p5
Fig. l.Ji.Dimensionless distance a5m, half width \|/m and cross sectional area A2/S vs n3/Qph.
s\-?.*t
n 3 /98
Fig. 1.32. Dimensionless distance from heat source to front (£') and rear (£") of isothermal enclosure vs
«3/68 (thin plate welding).
A graphical presentation of equations (1-89) and (1-90) is shown in Fig. 1.32. Included is
also a plot of the total length of the enclosure £t vs the parameter n3 /08.

For points located on the weld centre-line behind the heat source a 5 = -£ = X. When x is larger
than about 1, (i.e. t > IaIv1), it is a fair approximation to set K0 (x)« exp(-x)Vrc/2x (see Fig.
1.27). Hence, equation (1-83) reduces to:
(1-91)
Equation (1-91) provides a basis for calculating the cooling time within a specific tempera-
ture interval (e.g. from O1 to 02):
(1-92)
The dimensionless cooling time from 800 to 5000C is thus given by:
(1-93)
from which the real cooling time is obtained:
(1-94)
Taking as average values X = 0.025 W mm"1 0C"1, pc = 0.005 J mm"3 0C"1, and T0 = 200C
for welding of low alloy steels, we have:
d-95)
Similarly, the cooling rate at a specific temperature is obtained by differentiating equation

(1-91) with respect to time:
(1-96)
By multiplying equation (1-96) with the appropriate conversion factor, we get:

(1-97)
For welding of low alloy steels, the cooling rate becomes:
(1-98)
Example (LlO)
Consider GTA butt welding of a 2mm thick sheet of cold-rolled aluminium (Al-Mg alloy)
under the following conditions:
Sketch the contours of the fusion boundary and the Ar-isotherm in the £-\|/ (x-y) plane at
pseudo-steady state. The recrystallisation temperature Ar of the base material is taken equal to
2750 C. Calculate also the cross sectional area of the fully recrystallised HAZ and the cooling
rate at 2750C for points located within this region.
Solution
Referring to Fig. 1.22(a) (Example (1.6)) it is sufficient to calculate the coordinates in four
different (characteristic) positions to sketch the contour of the fusion boundary. If we neglect
the latent heat of melting, the n3/68 ratio at the melting point becomes:
End-points
The end-points can be read from Fig. 1.32:
and
Maximum widths
The maximum width of an isothermal enclosure can generally be calculated from equations
(1-86) and (1-87) or read from Fig. 1.31. When n3/QpS = 0.84, we obtain:
and
Intersection point with y/(y)-axis
In this case £ = 0 and cr5 = \j/. Hence, equation (1-83) reduces to:
which gives
Similarly, the contour of the Ar-isotherm can be determined by inserting n3/db = 2.08 into
the same set of equations. Figure 1.33 shows a graphical representation of the calculated
isothermal contours.
§
x(mm)
y(mm)
Fig. 1.33. Calculated contours of fusion boundary and Ar-isotherm in GTA butt welding of a 2mm thick
aluminium plate (Example 1.10).
Cross sectional area of fully recrystallised HAZ
In general, cross sectional areas can be read from Fig. 1.31. Taking the n3/OpS ratio equal to
0.84 (Qp= 1) and 2.08 (8 p = 0.48), respectively, we have:
which gives
Cooling rate at 275 0C

The cooling rate at a specific temperature can be calculated from equation (1-97). In the present
case, we obtain:
1.10.3.3 Simplified solution for a fast moving high power source

It follows from Fig. 1.29 that the isotherms behind the heat source become increasingly elon-
gated as the 08M3 ratio decreases. In the limiting case the isotherms will degenerate into
surfaces which are parallel to the welding x direction, as shown in Fig. 1.34.
In a short time interval dt the amount of heat released per unit length of the weld is equal to:
(1-99)
According to the assumptions this amount of heat will remain in a rod of constant cross
sectional area due to the lack of a temperature gradient in the welding direction. Under such
conditions the mode of heat flow becomes essentially one-dimensional, and the temperature
distribution is given by equation (1-5):
(1-100)
Equation (1-100) represents the simplified solution for a fast moving high power source* in
a thin sheet, and is valid within a limited range of the more general Rosenthal equation for two-
dimensional heat flow (equation (1-81)).
By substituting the appropriate dimensionless parameters into equation (1-100), we obtain:
* Since the shape of a given isotherm in the x-y plane is determined by the qjd ratio, the minimum welding speed
which is required to maintain 1-D heat flow increases with decreasing qjVd ratios. Hence, the term 'fast moving high
power source' is also appropriate in the case of the thin plate welding.
Fig. 1.34. Fast moving high power source in a thin plate.
(1-101)
The locus of the peak temperatures is readily evaluated from equation (1-101) by setting
3ln(e8M3)/3T = 0:
which gives
and
(1-102)
It is evident from the plot in Fig. 1.35 that the predicted width of the isotherms is always
greater than that inferred from the general thin plate solution (equation (1-83)) due to the
assumption of one-dimensional heat flow. However, such deviations become negligible at
very small Qpb/n3 ratios because of a small temperature gradient in the welding x direction
compared to the transverse y direction of the plate.
A general graphical representation of the weld thermal programme (similar to that shown in
Fig. 1.25 for a fast moving high power source on a heavy slab) can be obtained by combining
equations (1-101) and (1-102):
¥m(1-D)/¥m(a-D)
Asymptote
0 p 8/n 3
Fig. 1.35. Theoretical width of isotherms under 1-D and 2-D heat flow conditions, respectively at pseudo-
steady state (thin plate welding).
e/ep
2t/(vm)2
Fig. 1.36. Temperature-time pattern in thin plate welding at high arc power and high welding speed.
Next Page
(1-103)
Equation (1-103) has been plotted in Fig. 1.36.
Example (LU)
Consider butt welding of a 2mm thin plate of austenitic stainless steel with covered electrodes
(SMAW) under the following conditions:
Calculate the retention time within the critical temperature range for chromium carbide
precipitation (i.e. from 650 to 8500C) for points located at the 8500C isotherm.
Solution
If we use the melting point of the steel as a reference temperature, the parameter n3/5 becomes:
Qp< 1. Hence, the total time spent in the thermal cycle from 6 = 0.43 (T = 6500C) to 0p = 0.56
(Tp = 8500C) and down again to 0 = 0.43 can be read from Fig. 1.36. Taking the ordinate 0/0p
equal to 0.76, we obtain:
which gives
and
1.10.4 Medium thick plate solution

In a real welding situation the assumption of three-dimensional or two-dimensional heat flow
inherent in the Rosenthal equations is not always met because of variable temperature gradi-
ents in the through thickness z direction of the plate.
Model (after Rosenthal14)

The general medium thick plate model considers a point heat source moving at constant speed
across a wide plate of finite thickness d. With the exception of certain special cases (e.g.
watercooling of the back side of the plate), it is a reasonable approximation to assume that the
Previous Page
(1-103)
Equation (1-103) has been plotted in Fig. 1.36.
Example (LU)
Consider butt welding of a 2mm thin plate of austenitic stainless steel with covered electrodes
Calculate the retention time within the critical temperature range for chromium carbide
precipitation (i.e. from 650 to 8500C) for points located at the 8500C isotherm.
Solution
If we use the melting point of the steel as a reference temperature, the parameter n3/5 becomes:
Qp< 1. Hence, the total time spent in the thermal cycle from 6 = 0.43 (T = 6500C) to 0p = 0.56
(Tp = 8500C) and down again to 0 = 0.43 can be read from Fig. 1.36. Taking the ordinate 0/0p
equal to 0.76, we obtain:
which gives
and
1.10.4 Medium thick plate solution

In a real welding situation the assumption of three-dimensional or two-dimensional heat flow
inherent in the Rosenthal equations is not always met because of variable temperature gradi-
ents in the through thickness z direction of the plate.
Model (after Rosenthal14)

The general medium thick plate model considers a point heat source moving at constant speed
across a wide plate of finite thickness d. With the exception of certain special cases (e.g.
watercooling of the back side of the plate), it is a reasonable approximation to assume that the
plate surfaces are impermeable to heat. Thus, in order to maintain the net flux of heat through
both boundaries equal to zero, it is necessary to account for mirror reflections of the source
with respect to the planes of z = 0 and z = d. This can be done on the basis of the 'method of
images' as illustrated in Fig. 1.37. By including all contributions from the imaginary sources
...2q__2 , 2g_i , 2q\ , 2q2 ,...located symmetrically at distances ± 2id below and above the upper
surface of the plate, the pseudo-steady state temperature distribution is obtained in the form of
a convergent series*:
(1-104)
where
Note that equation (1-104) is simply the general Rosenthal thick plate solution (equation
(1-45)) summed for each source.
Fig. 1.37. Real and imaginary point sources on a medium thick plate.
*The number of imaginary heat sources necessary to achieve the required accuracy depends on the chosen values of
R0 and vd/2a.
By substituting the dimensionless parameters defined above into equation (1-104), we ob-
tain:
(1-105)
where
It follows from equation (1-104) that the thermal conditions will be similar to those in a
thick plate close to the centre of the weld. Moreover, Rosenthal1314has shown on the basis of
a Fourier series expansion that equation (1-104) converges to the general thin plate solution
(equation (1-81) for points located sufficiently far away from the source. However, at interme-
diate distances from the heat source, the pseudo-steady state temperature distribution will de-
viate significantly from that observed in thick plate or thin plate welding because of variable
temperature gradients in the through-thickness direction of the plate. Within this 'transition
region', the thermal programme is only defined by the medium thick plate solution (equation
(1-104)).
1.10.4.1 Dimensionless maps for heat flow analyses

Based on the models described in the previous sections, it is possible to construct a series of
dimensionless maps which provide a general outline of the pseudo-steady state temperature
distribution during arc welding.20
Construction of the maps

The construction of the maps is done on the basis of the medium thick plate solution (equation
(1-105)). This model is generally applicable and allows for the plate thickness effect in a
quantitative manner. Since the other solutions are only valid within specific ranges of this
equation, they will have their own characteristic fields in the temperature-distance or the tem-
perature-time space. The extension of the different fields can be determined from numerical
calculations of the temperature distribution by comparing each of these models with the me-
dium thick plate solution, using a conformity of 95% as a criterion.
Similarly, when the 95% conformity is not met between the respective solutions, the fields
are marked 'transition region'. Since any combination of dimensionless temperature, operat-
ing parameter, and plate thickness locates a point in a field, it means that the dominating heat
flow mechanism can readily be read off from the maps.
Peak temperature distribution

The variation of peak temperature with distance in the \j/(^j-direction has been numerically
evaluated from equation (1-105) for different values of the dimensionless plate thickness (8 =
vdlla). The results are shown graphically in Fig. 1.38(a) and (b) for the two extreme cases of
£ = 0 (z = 0) and £ = 8 (z = d), respectively.
An inspection of the maps reveals that the temperature-time pattern in stringer bead
weldments can be classified into three main categories:
(a)
Thin plate solution

(2-D heat flow) (1-D heat flow)
VV
Y
m
(b)
Thin plate solution
(2-D heat flow) 1-D heat flow
V"3
Y
m
Fig. 1.38. Peak temperature distribution in transverse direction (\|/ = \\fm) of plate; (a) Upper plate surface
(^ = 0), (b) Lower plate surface (J = 8).
1. Close to the heat source, the thermal programme will be similar to that in a thick plate
(Fig. 1.38(a)), which means that the temperature distribution is determined by equation (1-
54). For large values of the dimensionless plate thickness, the mode of heat flow may
become essentially two-dimensional. This corresponds to the limiting case of a fast mov-
ing high power source in a thick plate (equation (1-74)). Under such conditions the slope of
the Qp/n3-\ym curves in Fig. 1.38(a) attains a constant value of-2.
2. With increasing distance from the heat source, a transition from three-dimensional to
two-dimensional heat flow may occur, depending on the dimensionless plate thickness and
the operational conditions applied. Considering the upper surface of the plate (Fig. 1.38(a)),
the extension of the transition region is seen to decrease with increasing values of 8 as the
conditions for thick plate welding are approached. The opposite trend is observed for the
bottom plate surface (Fig. 1.38(b)), since a small dimensionless plate thickness generally
results in a more rapid equalisation of the temperature gradients in the t,(z) direction. When
the curves in Fig. 1.38(b) become parallel with the jc-axis, the temperature at the bottom of
the plate reaches its maximum value. Note that within the transition region, reliable predic-
tions of the pseudo-steady temperature distribution can only be made from the medium
thick plate solution (equation (1-105)).
3. For points located sufficiently far away from the heat source, the temperature gradi-
ents in the through-thickness direction of the plate become negligible. This implies that the
temperature distribution at the upper and lower surface of the plate is similar, and can be
computed from the thin plate solution (equation (1-83)). When the conditions for one-
dimensional heat flow are approached (equation (1-101)), the slope of the dp/n3-\\fm curves
in Fig. 1.38(a) and (b) attains a constant value o f - 1 .

Figure 1.39 contains a plot of the cooling programme for points located on the weld centre-line
(\j/ = £ = 0), as calculated from equation (1-105). A closer inspection of Fig. 1.39 reveals that
the slope of the cooling curves increases gradually from -1 to -0.5 with increasing distance
from the heat source. This corresponds to a change from three-dimensional to one-dimen-
sional heat flow.
From Fig. 1.39 it is possible to read-off the cooling time within specific temperature inter-
vals for a wide range of operational conditions. These results are also valid for positions
outside the weld centre-line, since the cooling curves are virtually parallel in the transverse \|/
direction of the plate. A requirement is, however, that the peak temperature of the thermal
cycle is significantly higher than the actual temperature interval under consideration.
Retention times at elevated temperatures

The retention time, Axn is defined as the total time spent in a thermal cycle from a chosen
reference temperature 0 to the peak temperature dp and down again to 9. This parameter can
readily be computed from equation (1-105) by means of numerical methods. The results of
such calculations (carried out in position £ = 0) are shown graphically in Fig. 1.40 for 9 = 0.5
An inspection of Fig. 1.40 reveals a complex temperature-time pattern. In this case it is not
possible to determine the exact field boundaries between the respective solutions, since the
Thin plate solution
(2-D heat flow) (1-D heat flow)
e/n3
e/n3
T= ^
X
Fig. 1.39. Cooling programme for points located on the weld centre-line (\\f = £ =0).
e/n 3
AXf
Fig. 1.40. Total time spent in a thermal cycle from 9 through 9p to 9 for a chosen reference temperature of
9 = 0.59p.
mode of heat flow may vary within a single thermal cycle. Hence, the extension of the differ-
ent fields is not indicated in the graph. The results in Fig. 1.40 provide a systematic basis for
calculating the retention time within specific temperature intervals under various welding con-
ditions.
Isothermal contours
Because of the number of variables involved, it is not possible to present a two-dimensional
plot of the isotherms without first specifying the dimensionless plate thickness. Examples of
calculated isotherms in different planes are shown in Figs. 1.41 and 1.42 for 8 equal to 0.5 and
5, respectively. It is evident that an increase in the dimensionless plate thickness from 0.5 to 5
has a dramatic effect on the shape and position of the isothermal contours. However, in order
to explain these observations in an adequate manner, it is necessary to condense the results into
a two-dimensional diagram. As shown in Fig. 1.43, this can be done by plotting the calculated
field boundaries in Fig. 1.38(a) at maximum width of the isotherms vs the parameters Qp/n3
and vdlla.
It is seen from Fig. 1.43 that a large plate thickness generally will favour three-dimensional
heat flow. With decreasing values of Qp/n3, the conditions for a fast moving high power source
are approached before the transition from the thick plate to the thin plate solution occurs. In
such cases the isotherms at the bottom of the plate will be strongly elongated in the welding £
direction and shifted to positions far behind the heat source. The opposite trend is observed at
small values of vdlla, since a rapid equalisation of the temperature gradients in the through-
thickness direction of the plate will result in elliptical isotherms at both plate surfaces, located
in an approximately equal distance from the heat source. In either case the temperature at
which the cross-sectional isotherms approach a semi-circle or become parallel with the XXz)-
axis can be obtained from Fig. 1.43 by reading-off the intercept between the line for the
dimensionless plate thickness and the respective field boundaries.
Limitations of the maps

Since the maps have been constructed on the basis of the analytical heat flow equations, it is
obvious that they will apply only under conditions for which these equations are valid. The
simplifying assumptions inherent in the models can be summarised as follows:
(a) The parent material is isotropic and homogeneous at all temperatures,

and no phase changes occur on heating.
(b) The thermal conductivity, density, and specific heat are constant and
independent of temperature.
(c) The plate is completely insulated from its surroundings, i.e. there are no
heat losses by convection or radiation from the boundaries.
(d) The plate is infinite except in the directions specifically noted.
(e) The electrode is not consumed during welding, and all heat is concen-
trated in a zero-volume point or line.
(a) ¥
(b)
%
(C)
Fig. 1.41. Computed isothermal contours in different sections for 8 = 0.5;

(a) Front view (\j/ = \|/m),
(b) Side view (\|/ = 0),
(c) Top view (£ = 0) and bottom view (£ = 8).
(a)
¥
(b)
%
(C)
\-
Fig. 1.42.Computed isothermal contours in different sections for 8 = 5;

(a) Front view (\|/ = \|/m);
(b) Side view (\|/ = 0);
(c) Top view (£ = 0) and bottom view (£ = 8).
Thick plate solution
(3-D heat flow)
ep/n3
Thin plate
solution
1-D heat flow
Thick plate
solution
(2-D heat flow)
5 = vd/2a
Fig. 1.43. Heat flow mechanism map showing calculated field boundaries in transverse direction (i|/
= i|/m) of plate vs Qp/n3 and 8 = vdlla.
(f) Pseudo-steady state, i.e. the temperature does not vary with time when
observed from a point located in the heat source.
In general, the justification of these assumptions relies on a good correlation between theory
and experiments. However, since the analytical solutions ignore the important role of arc
energy distribution and directed metal currents in the weld pool, predictions of the weld ther-
mal programme should be restricted to positions well outside the fusion zone where such
effects are of less importance (to be discussed below).
Example (1.12)
Consider stringer bead welding (GMAW) on a 12mm thick plate of aluminium (> 99% Al)
Based on Fig. 1.43, sketch the peak temperature contours in the transverse section of the
weld at pseudo-steady state.
Solution
If we neglect the latent heat of melting, the parameter n3 at the chosen reference temperature
(Tc = Tm) becomes:
Similarly, when v = 3mm s l and a = 85mm2 s"1 we obtain the following value for the di-
mensionless plate thickness:
Readings from Fig. 1.43 give:
ep ^, (0C) Model System Comments
0.50 —• 1.0 340 —• 660 Medium thick Heat flow in x and y directions,
plate solution partial heat flow in z direction
0.17 -> 0.50 130 -> 340 Thin plate solution Heat flow in x and y directions,
(2-D heat flow) negligible heat flow in z direction
< 0.17 < 130 Thin plate solution Heat flow in y direction, negligible
(1 -D heat flow) heat flow in x and z directions
A sketch of the HAZ peak temperature contours in the transverse section of the weld is
shown in Fig. 1.44.
Case Study (Ll)

Consider stringer bead GMA welding on 12.5mm thick plates of low alloy steel and aluminium
(i.e. a Al-Mg-Si alloy), respectively under the conditions E = 1.5 kJ mm"1 and r\ = 0.8. Details
of welding parameters for the four series involved are given in Table 1.4. Computed peak
temperature contours in the transverse section of the welds are given in Figs. 1.45 and 1.46.
Arrows indicate heat flow directions
Fusion
zone
Fig. 1.44. Schematic diagram showing specific peak temperature contours within the HAZ of an alu-
minium weld at pseudo-steady state (Example 1.12).
(a) WELD A1
y(mm)
z(mm)
(b) WELD A2
y(mm)
z(mm)
Fig. 1.45. Computed peak temperature contours in aluminium welding at pseudo-steady state (Case study
1.1); (a) Weld Al, (b) Weld A2. Black regions indicate fusion zone.
Aluminium welding
In general, the maximum width of the isotherms at the upper and lower surface of the plate can
be obtained from Fig. 1.38(a) and (b), although these maps are not suitable for precise read-
ings. A comparison with the computed peak temperature contours in Fig. 1.45(a) and (b)
reveals a strong influence of the welding speed on the shape and position of the cross-sectional
isotherms at a constant gross heat input of 1.5 kJ mm"1. It is evident that the extension of the
fusion zone and the neighbouring isotherms becomes considerably larger when the welding
speed is increased from 2.5 to 5 mm s"1. This effect can be attributed to an associated shift
from elliptical to more elongated isotherms at the plate surfaces (e.g. see Fig. 1.43), which
reduces heat conduction in the welding direction. It follows from Fig. 1.43 that the upper left
corner of the map represents the typical operating region for aluminium welding.
Because of the pertinent differences in the heat flow conditions, the temperature-time pat-
tern will also vary significantly between the respective series as indicated by the maps in Figs.
1.39 and 1.40. Hence, in the case of aluminium welding the usual procedure of reporting arc
power and travel speed in terms of an equivalent heat input per unit length of the bead is highly
questionable, since this parameter does not define the weld thermal programme. In general,
the correct course would be to specify both qo, v and d, and compare the weld thermal history
on the basis of the dimensionless parameters n3 and 8.
Steel welding
If welding is performed on a steel plate of similar thickness, the operating region will be
shifted to the lower right corner of Fig. 1.43. Under such conditions, the isotherms adjacent to
the fusion boundary will be strongly elongated in the x-direction even at very low welding
speeds (see Fig. 1.42). This implies that the thermal programme approaches a state where the
temperature distribution is uniquely defined by the net heat input r\E, corresponding to the
(a) W E L D S1
y(mm)
z(mm)
(b) WELD S2
y(mm)
z(mm)
Fig. 7.46.Computed peak temperature contours in steel welding at pseudo-steady state (Case study 1.1);
(a) Weld Sl, (b) Weld S2. Black regions indicate fusion zone.
Table 1.4 Operational conditions assumed in Case study (1.1).
qo V d E n3 8
Material Series (W) (mras"1) (mm) (kJmrrr 1 )
Al-Mg-Si Al 6000 5 12.5 1.5 0.36 0.50

alloy A2 3000 2.5 12.5 1.5 0.09 0.25
Low alloy Sl 9600 8 12.5 1.5 32.6 10
steel S2 4800 4 12.5 1.5 8.2 5
limiting case of a fast moving high power source. As a result, the calculated shape and width
of the fusion boundary and neighbouring isotherms are seen to be virtually independent of
choice of qo and v as illustrated in Fig. 1.46(a) and (b).
1.10.4.2 Experimental verification of the medium thick plate solution

It is clear from the above discussion that the medium thick plate solution provides a systematic
basis for calculating the temperature distribution within the HAZ of stringer bead weldments
under various welding conditions. In the following, the accuracy of the model will be checked
against extensive experimental data, as obtained from in situ thermocouple measurements and
numerical analyses of a large number of bead-on-plate welds.
Weld thermal cycles

Examples of measured and predicted weld thermal cycles in aluminium welding are presented
in Fig. 1.47. It is evident that the medium thick plate solution predicts adequately the HAZ
temperature-time pattern under different heat flow conditions for fixed values of the peak
temperature. This, in turn, implies that the model is also capable of predicting the total time
spent in a thermal cycle within a specific temperature interval as shown in Fig. 1.48.
Weld cooling programme

At temperatures representative of the austenite to ferrite transformation in mild and low alloy
steel weldments, the conditions for a fast moving high power source are normally approached
before the transition from thick plate to thin plate welding occurs (see Fig. 1.39). In such
cases, it is possible to present the different solutions for Ax875 (at \|/ = £ = 0) in a single graph by
introducing the following groups of variables*:
Ordinate:
Abscissa:
Relevant literature data for the cooling time between 800 and 5000C are given in Fig. 1.49.
A closer inspection of the figure reveals a reasonable agreement between observed and pre-
dicted values in all cases. For welding of thick plates, the ordinate attains a constant value of
1. Similarly, in thin sheet welding, the slope of the curve becomes equal to 1.
In aluminium welding, the thermal conditions will be much more complex because of the
resulting higher base plate thermal diffusivity (see Fig. 1.39). Hence, it is not possible to
describe the weld cooling programme in terms of equations (1-74) and (1-101), which apply to
fast moving high power sources. The plot in Fig. 1.50 confirms, however, that the medium
thick plate solution is also capable of predicting the cooling time within specific temperature
intervals (e.g. from 300 to 1000C) in aluminium weldments.
These groups of variables can be obtained from equations (1-74) and (1-101).
Measured
Predicted
GMAW: q 0 = 3872 W, v = 8.8 mm/s,

d =8 mm
Temperature (0C)
Time (sec)
Fig. 1.47. Comparison between measured and predicted weld thermal cycles in aluminium welding for
fixed values of Tp. Data from Myhr and Grong.20
A tr(s), observed
T
P
A tr(s), predicted
Fig. /.4#. Comparison between measured and predicted retention times in aluminium welding for fixed
values of Tp. Data from Myhr and Grong.20
SMAW
SMAW
SAW
SAW
THICK PLATE
1
SOLUTION
n 3 [-i--4-
8/5
AT
THIN PLATE
SOLUTION
-2Hr + Ir 1
5 6SOO 6SOO
Fig. 1.49. Comparison between observed and predicted cooling times from 800 to 5000C in steel welding
(solid lines represent theoretical calculations). Data from Myhr and Grong.20
GMAW (Al+2.5 wt% Mg)

(sec), observed
At
T(0C)
t(sec)
At (sec), predicted
Fig. 1.50. Comparison between observed and predicted cooling times from 300 to 1000C in aluminium
welding. Data from Myhr and Grong.20
Peak temperatures and isothermal contours
Figure 1.51 shows a comparison between measured and predicted HAZ peak temperatures in
aluminium welding. It is evident that the relative positions of the HAZ isotherms can be
calculated with a reasonable degree of accuracy from the medium thick plate solution, pro-
vided that the equation is precalibrated against a known isotherm (i.e. the fusion boundary).
Additional information on the HAZ peak temperature distribution in aluminium welding
can be obtained from the data of Koe and Lee21 reproduced in Fig. 1.52. These numerical
calculations* showed a good correlation with experimental measurements. A comparison with
the medium thick plate solution in Fig. 1.52 reveals a fair agreement between numerically and
analytically computed isothermal contours. It is interesting to note that even though the plate
thickness is small (i.e. 3.2mm), the mode of heat flow becomes essentially three-dimensional
close to the fusion boundary. This important point is often overlooked when discussing the
relevance of the analytical heat flow models in thin sheet welding.
LlOAJ Practical implications

The following important conclusions can be drawn from the results presented in Figs. 1.38-
1.52:
GMAW (Al +2.5 wt% Mg)

Tp, 0C , observed
Tp, 0C , predicted
Fig. 1.51. Comparison between observed and predicted HAZ peak temperatures in aluminium welding.
Data from Myhr and Grong.20
Based on the finite difference method (FDM).

y (mm)
Z (mm)
Fig. 1.52. Comparison between numerically and analytically computed peak temperature contours in
GTA welding of a 3.2mm thin aluminium sheet. (Broken lines: numerical model; solid lines: analytical
model.) Data for welding parameters and material properties are given in Ref. 21.
1. Considering heat flow and temperature distribution in fusion welding, there exists no
defined plate thickness which can be regarded as 'thick' or 'thin'. Accordingly, in a real
welding situation, the mode of heat flow will vary continuously with increasing distance
from the heat source.
2. Close to the centre of the weld, the thermal conditions will be similar to those in a
heavy slab. This means that the temperature distribution is approximately described by the
Rosenthal thick plate solution.
3. At intermediate distances from the heat source, the temperature distribution will de-
viate significantly from that observed in thick plate welding because of variable tempera-
ture gradients in the through-thickness direction of the plate. Within this transition region,
reliable predictions of the pseudo-steady state temperature distribution can only be made
from the medium thick plate solution.
4. For points located sufficiently far away from the heat source, the temperature gradi-
ents in the through-thickness direction of the plate become negligible. Under such condi-
tions, the weld thermal programme is approximately defined by the Rosenthal thin plate
solution.
5. In general, a full description of the weld thermal history requires that both the arc
power qo, the travel speed V, and the plate thickness d are explicitly specified. Hence, the
usual procedure of reporting welding variables in terms of an equivalent heat input per unit
length of the bead, £(kJ mm"1), is highly questionable, since this parameter does not define
the weld thermal programme. An exception is welding of thick steel plates, where the
temperature distribution approaches that of a fast moving high power source because of a
low thermal diffusivity of the base metal.
6. A comparison between theory and experiments shows that the medium thick plate
solution predicts adequately both the peak temperature distribution and the temperature-
time pattern within the HAZ of stringer bead weldments for a wide range of operational
conditions (including aluminium and steel welding). A requirement is, however, that the
equation is calibrated against a known isotherm (e.g. the fusion boundary) due to the sim-
plifying assumptions inherent in the model.
1.10.5 Distributed heat sources
In some cases it is also necessary to consider the important influence of filler metal additions,
arc energy distribution, and convectional heat flow in the weld pool on the resulting bead
morphology to obtain a good agreement between theory and experiments. Of particular inter-
est in this respect is the formation of the so-called weld crater/weld finger, frequently observed
in GMA and SA stringer bead weldments (see Fig. 1.53). Although the nature of these phe-
nomena is very complex, they can readily be accounted for by applying an empirical correc-
tion for the effective heat distribution in the weld pool.
1.10.5.1 General solution
Model (after Myhr and Grong22)

The heat distribution used to simulate the weld crater/weld finger formation is shown
schematically in Fig. 1.54(a). Here we consider two discreate distributions of elementary
point sources, which extend in the y- and z-direction of the plate, respectively. The contribu-
tion from each source to the temperature rise in an arbitrary point P located within the plate is
calculated on the basis of the "method of images", as shown in Fig. 1.54(b) and (c). For a heat
source displaced in the y-direction (Fig. 1.54(b)), the temperature field is given by equation (1 -
104):
(1-106)
where
y
Crater
Finger
HAZ"
Fusion line
Fig. 1.53. Schematic diagram showing the weld crater/weld finger formation during stringer bead weld-
ing.
(a) y
(b)
Z
Fig. 1.54. General heat flow model for welding on medium thick plates; (a) Physical representation of the
heat distribution by elementary point sources, (b) Method for calculating the temperature field around an
elementary point source displaced along the y-axis.
(C)
Fig. 1.54. General heat flow model for welding on medium thick plates(continued): (c) Method for cal-
culating the temperature field around an elementary (submerged) point source displaced along the z-axis.
Similarly, for a submerged point source located along the z-axis (Fig. 1.54(c))> we obtain:
(1-107)
where
and
Note that equation (1-107) correctly reduces to equation (1-106) when A approaches zero.
The total temperature rise in point P is then obtained by superposition of the temperature
fields from the different elementary heat sources, i.e.:
(1-108)
where
In practice, we can subdivide the heat distributions into a relatively small number of el-
ementary point sources, and usually a total number of 8 to 10 sources is sufficient to obtain
good results (i.e. smooth curves). However, the relative strength of each heat source and their
distribution along the y- and z-axes must be determined individually by trial and error by
comparing the calculated shape of the fusion boundary with the real (measured) one.
Figure 1.55 shows the results from such calculations, carried out for a single pass (bead-in-
groove) GMA steel weld. It is evident that the important effect of the weld crater/weld finger
formation on the HAZ peak temperature distribution is adequately accounted for by the present
model. A weakness of the model is, of course, that the shape and location of the fusion bound-
ary must be determined experimentally before a prediction can be made.
1.10.5.2 Simplified solution

Similar to the situation described above, the point heat source will clearly not be a good
model when the heat is supplied over a large area. Welding with a weaving technique and sur-
facing with strip electrodes are prime examples of this kind.
Model (after Grong and Christensen19)

As a first simplification, the Rosenthal thick plate solution is considered for the limiting case
of a high arc power qo and a high welding speed D, maintaining the ratio qo Iv within a range
applicable to arc welding. Consider next a distributed heat source of net power density qo I2L
extending from -L to +L on either side of the weld centre-line in the y-direction*, as shown
schematically in Fig. 1.56. It follows from equation (1-73) that an infinitesimal source dqy
located between y and y + dy will cause a small rise of temperature in point P at time f, as:
(1-109)
where and
Alternatively, we can use a Gaussian heat distribution, as shown in Appendix 1.4.

z (mm)
Shaded region indicates

fusion zone
y (mm)
Fig. 1.55. Calculated peak temperature contours in the transverse section of a GMA steel weldment (Op-
erational conditions: / = 450A, U - 30V, v = 2.6mm s"1, d = 50mm).
2-D heat flow
Fig. 1.56. Distributed heat source of net power density qJ2L on a semi-infinite body (2-D heat flow).
(1-110)
where erf(u) is the Gaussian error function (previously defined in Appendix 1.3).
Peak temperature distribution

Because of the complex nature of equation (1-110), the variation of peak temperature with
distance in the y-z plane can only be obtained by numerical or graphical methods. Accord-
ingly, it is convenient to present the different solutions in a dimensionless form. The following
parameters are defined for this purpose:
(1-111)
Dimensionless time:
(1-112)
Dimensionless y-coordinate:
(1-113)
Dimensionless z-coordinate:
(1-114)
(1-115)
where 0 is the dimensionless temperature (previously defined in equation (1-9)).

The variation of peak temperature Qp with distance in the y-z plane has been numerically
evaluated from equation (1-115) for chosen values of P and 7 (i.e. P = 0, (3 = 3/4, P = I , and
7 = 0). The results are presented graphically in Figs. 1.57 and 1.58 for the through thickness
(z = zm) and the transverse (y = ym) directions, respectively. These figures provide a systematic
basis for calculating the shape of the weld pool and neighbouring isotherms under various
welding conditions.
In Fig. 1.59 the weld width to depth ratio has been computed and plotted for different
combinations of nw and 9p. It is evident that the predicted width of the isotherms generally is
much greater, and the depth correspondingly smaller than that inferred from the point source
model. Such deviations tend to become less pronounced with decreasing peak temperatures
(i.e. increasing distance from the heat source). At very large value of nw, the theoretical shape
of the isotherms approaches that of a semi-circle, which is characteristic of a point heat source.
Example (1.13)
Consider GMA welding with a weaving technique on a thick plate of low alloy steel under the
following conditions:
Sketch the contours of the fusion boundary and the Ac r isotherm (71O0C) in the y-z plane.
Compare the shape of these isotherms with that obtained from the point heat source model.
Solution
If we neglect the latent heat of melting, the operating parameter at the chosen reference tem-
perature (Tc = Tm) becomes:
Fusion boundary
Here we have:
Readings from Figs. 1.57 and 1.58 give:
Ac j-temperature
In this case the peak temperature should be referred to 7200C, i.e.:
ep/nw
V -
Fig. /.57. Calculated peak temperature distribution in the through-thickness direction of the plate at dif-
ferent positions along the weld surface.
ep/nw
y m /L
Fig. 1.58. Calculated peak temperature distribution in the transverse direction of the plate at position y
(Z) = 0.
y m /z m
nw
Fig. 1.59. Effects of nw and dp on the weld width to depth ratio.
from which
Readings from Figs. 1.57 and 1.58 give:
Similarly, equation (1-75) provides a basis for calculating the width of the isotherms in the
limiting case where all heat is concentrated in a zero-volume point. By rearranging this equa-
tion, we obtain:
which gives
6.3 mm when Bp = 1(Tp = 15200C)

and
when
Figure 1.60 shows a graphical presentation of the calculated peak temperature contours.
Implications of model
It is evident from Fig. 1.60 that the predicted shape of the isotherms, as evaluated from equa-
tion (1-110), departs quite strongly from the semi-circular contours required by a point heat
source. Moreover, a closer inspection of the figure shows that inclusion of the heat distribution
also gives rise to systematic variations in the weld thermal programme along a specific
isotherm, as evidenced by the steeper temperature gradient in the v-direction compared with
the z-direction of the plate. This point is more clearly illustrated in Fig. 1.61, which compares
the HAZ temperature-time programme for the two extreme cases of z = 0 and y = 0, re-
spectively. It is obvious from Fig. 1.61 that the retention time within the austenite regime is
considerably longer in the latter case, although the cooling time from 800 to 5000C, Af8/5, is
reasonably similar. These results clearly underline the important difference between a point
heat source and a distributed heat source as far as the weld thermal programme is concerned.
Model limitations
In the present model, we have used the simplified solution for a fast moving high power
source (equation 1-73)) as a starting point for predicting the temperature-time pattern. Since
the equations derived later are obtained by integrating equation (1-73), they will, of course,
apply only under conditions for which this solution is valid.
Moreover, a salient assumption in the model is that the heat distribution during weaving
can be represented by a linear heat source orientated perpendicular to the welding direction.
Although this is a rather crude approximation, experience shows that the assumed heat dis-
z, mm
Fig. /.60. Predicted shape of fusion boundary and Acpisotherm during GMA welding of steel with an
oscillating electrode (Example 1.13). Solid lines: Distributed heat source; Broken lines: Point heat source.
Temperature, 0C
Time, s
Fig. L61. Calculated HAZ thermal cycles in positions y = 0 and z = 0 (Example 1.13).
tribution is not critical unless the rate of weaving is kept close to the travel speed. However,
for most practical applications weaving at such low rates would be undesirable owing to an
unfavourable bead morphology.
Case Study (1.2)

Surfacing with strip electrodes makes a good case for application of equation (1-110). Specifi-
cally, we shall consider SA welding of low alloy steel with 60mm X 0.5mm stainless steel
electrodes. The operational conditions employed are listed in Table 1.5.
It is evident from the metallographic data presented in Fig. 1.62 that neither the bead pen-
etration nor the HAZ depth (referred to the plate surface) can be predicted readily on the basis
of the present heat flow model when welding is carried out with a consumable electrode,
owing to the formation of a reinforcement. This situation arises from the simplifications made
in deriving equation (1-110). The problem, however, may be eliminated by calculating the
depth of the Ac3 and Ac1 regions relative to the fusion boundary, i.e. Azm = zm(Qp) - zm (Qp = 1),
or Aym = ym (8p) - ym (6p = 1), for specific positions along the weld fusion line, as shown by the
solid curves in Fig. 1.62 for Qp = 0.54 and 0.45, respectively. An inspection of the graphs
reveals satisfactory agreement between theory and experiments in all three cases, which im-
plies that the model is quite adequate for predicting the HAZ thermal programme as far as strip
electrode welding is concerned. This result is to be expected, since the assumption of two-
dimensional heat flow is a realistic one under the prevailing circumstances.
Case Study (1.3)

As a second example we shall consider GTA welding (without filler wire additions) at various
heat inputs and amplitudes of weaving within the range from 1 to 2.5 kJ mm"1 and 0 to 15mm,
respectively. Data for welding parameters are given in Table 1.6.
Stainless steel!
y, mm
Weld S3
Fusion line
z, mm
Stainless steel
y, mm
Weld S4
z, mm
Stainless steel!
y, mm
Weld S5
z, mm
Fig. 1.62. Comparison between observed and predicted Ac3 and Ac1 contours during strip electrode weld-
ing (Case study 1.2). Data from Grong and Christensen.19
Table 1.5 Operational conditions used in strip electrode welding experiments (Case study 1.2).
Base metal/ Weld / U v 2L nw

filler metal No. (A) (V) (mms"1) (mm) Cn = 0.7)
combination
S3 730 27 1.8 60 0.34

Low alloy steel/
S4 73 27 2 2
stainless steel ° - 60 0.28
S5 730 27 2.5 60 0.24
Table 1,6 Operational conditions used in GTA welding experiments (Case study 1.3).
I U v 2L nw
1 12
WeIdNo. (A) (V) (mms" ) (mm) "V=O- T] = 0.23
Bl 200 13.5 2.6 9.5 0.20 0.40

B2 200 14.0 1.1 15.0 0.20 0.40
B3 200 13.5 2.5 -
B4 200 12.5 1.0
Calibration procedure
In general, a comparison between theory and experiments requires that the arc efficiency fac-
tor can be established with a reasonable degree of accuracy. Unfortunately, the arc efficiency
factor for GTA welding has not yet been firmly settled, where values from 0.25 up to 0.75 have
been reported in the literature (see Table 1.3). Additional problems result from the fact that
only a certain fraction of the total amount of heat transferred from the arc to the base plate is
sufficiently intense to cause melting. This has led to the introduction of the melting efficiency
factor T]m, which normally is found to be 30-70% lower than the total arc efficiency of the
process, depending on the latent heat of melting, the applied amperage, voltage, shielding gas
composition, or electrode vertex angle.23 Consequently, since these parameters cannot readily
be obtained from the literature, the following reasonable values for r\m and r\ have been as-
sumed to calculate nw in Table 1.6, based on a pre-evaluation of the experimental data: j \ m =
0.12 (fusion zone), TI = 0.23 (HAZ).
It should be noted that the above values also include a correction for three-dimensional heat
flow, since the assumption of a fast moving high power source during low heat input GTA
welding is not valid. Hence, both the arc efficiency factor and the melting efficiency factor
used in the present case study are seen to be lower than those commonly employed in the
literature.
Full weaving (welds Bl and B2)

The results from the metallographic examination of the two GTA welds deposited under full
weaving conditions are presented graphically in Fig. 1.63. Note that the shape of the fusion
boundary as well as the Ac3 and the Ac1 isotherms can be predicted adequately from the present
model for both combinations of E and L (an exception is the HAZ end points in position z = 0),
provided that proper adjustments of i\m and Tj are made. The good correlation obtained in Fig.
1.63 between the observed and the calculated peak temperature contours justifies the adapta-
tion of the model to low heat input processes such as GTA welding, despite the fact that the
assumption of two-dimensional heat flow is not valid under the prevailing circumstances.
No weaving (welds B3 and B4)

For the limiting case of no weaving (Fig. 1.64), the concept of an equivalent amplitude of
weaving has been used in order to calculate the peak temperature contours from the model.
This parameter (designated Leq) takes into account the effects of convectional heat flow in the
weld pool on the resulting bead geometry, and is evaluated empirically from measurements of
the actual weld samples. At low heat inputs (Fig. 1.64(a)), the agreement between theory and
y, mm
z, mm
Weld B1
Fusion line
y, mm
Weld B2 z, mm
Fig. 1.63. Comparison between observed and predicted fusion line, Ac3 and Ac1 contours during GTA
welding under full weaving conditions (Case study 1.3). Data from Grong and Christensen.19
experiments is largely improved by inserting 2Leq = 7.5mm into equation (1-110), when com-
parison is made on the basis of the point source model. In contrast, at a heat input of 2.5 kJ
mm"1 (Fig. 1.64(b)), the measured shape of the HAZ isotherms is seen to approach that of a
semi-circle, and hence the deviation between the present model and the simplified solution for
a fast moving high power point source is less apparent.
Intermediate weaving
At intermediate amplitudes of weaving (2L = 5 and 7.5mm, respectively), convectional heat
flow in the weld pool will also tend to increase the bead width to depth ratio beyond the
theoretical value predicted from the present model, as shown in Fig. 1.65. The plot in Fig. 1.65
Weld B3
y, mm
Fusion line
z, mm
Weld B4 y, mm
z, mm
Fig. 1.64. Comparison between observed and predicted fusion line, Ac3 and Ac1 contours during GTA
welding with a stationary arc (Case study 1.3). Solid lines: Distributed heat source, Broken lines: Point
heat source. Data from Grong and Christensen.19
includes all data obtained in the GTA welding experiments with an oscillating arc, as reported
by Grong and Christensen.19 These results suggest that the applied amplitude of weaving must
be quite large before such effects become negligible. Consequently, adaptation of the model to
the weld series considered above would require an empirical calibration of the weaving ampli-
tude similar to that performed in Fig. 1.64 for stringer bead weldments to ensure satisfactory
agreement between theory and experiments.
1.10.6 Thermal conditions during interrupted welding

Rapid variations of temperatures as a result of interruption of the welding operation can have
an adversely effect on the microstructure and consequently the mechanical properties of the
weldment.
Width
Depth
Width to depth ratio
Theoretical
curve
n
w
Fig. 1.65.Comparison between observed and predicted weld width to depth ratios during GTA welding
with an oscillating arc (Case study 1.3). Data from Grong and Christensen.19
T,e
t,t
Fig. 1.66. Idealised heat flow model for prediction of transient temperatures during interrupted welding.
The situation existing after arc extinction may be described as shown in Fig. 1.66. From time
t = t* there is no net heat supply to the weldment. This condition is satisfied if the real source
q0 is considered maintained by adding an imaginary source +qo and sink -qo of the same
strength at t*. The temperature at some later time t** in a given position R0 (measured from the
origin 0") is then equal to the difference of temperatures due to the positive heat sources qo
and +qo and the negative heat sink -qo. Each of these temperature contributions will be a
product of a pseudo-steady state temperature Tps, and a correction factor K1 or K2 (given by
equations (1-49) and (1-82), respectively). Hence, for 3-D heat flow, we have:
(1-116)
where
and
Similarly, for 2-D heat flow, we get:
(1-117)
where
(ro is the position of the weld with respect to the imaginary heat source at time f* in the x —y
plane).
Example (1.14)
Consider repair welding of a heavy steel casting with covered electrodes under the following
conditions:
Suppose that a 50mm long bead is deposited on the top of the casting. Calculate the tem-
perature in the centre of the weld 5 s after arc extinction.
Solution
The pseudo-steady state temperature for points located on the weld centre-line (i|/ = £= 0) can
be obtained from equation (1-65). When t** - t* = 5 s, we get:
Referring to Fig. 1.67, the position of the weld with respect to the imaginary heat source at
time f* is 10mm, which gives:
Fig. 1.67. Sketch of weld bead in Example 1.14.
(a)
3-D heat flow
(b)
3-D heat flow
(C)
3-D heat flow
Fig. /.6#. Recommended correction factor/for some joint configurations; (a) Single V-groove, (b) Dou-
ble V-groove, (c) T-joint.
Moreover, the dimensionless times T** and T** - x* are:
and
At these coordinates, the correction factor K1 is seen to be 1 and 0.62, respectively (Fig.
1.18). The temperature in the centre of the weld 5 s after arc extinction is thus:
which is equivalent to
LlOJ Thermal conditions during root pass welding

During conventional bead-on-plate welding the angle of heat conduction is equal to 180° due
to symmetry effects (e.g. see Fig. 1.23). In order to apply the same heat flow equations during
root pass welding, it is necessary to introduce a correction factor,/, which takes into account
variations in the effective heat diffusion area due to differences in the joint geometry. Taking
/equal to 1 for ordinary bead-on-plate welding (b.o.p.), we can define the net heat input of a
groove weld as:9
(1-118)
Recommended values of the correction factor/for some joint configurations are given in
Fig. 1.68.
Example (Ll5)
Consider deposition of a root pass steel weld in a double-V-groove with covered electrodes
Calculate the cooling time from 800 to 5000C (Ar875), and the cooling rate (CR.) at 6500C
in the centre of the weld when the groove angle is 60°.
Solution
The cooling time, Ar875, and the cooling rate, CR., can be obtained from equation (1-68) and
(1-71), respectively:
Cooling time, Atm

Cooling rate at 6500C
The above calculations show that the thermal conditions existing in root pass welding may
deviate significantly from those prevailing during ordinary stringer bead deposition due to
differences in the effective heat diffusion area. These results are in agreement with general
experience (see Fig. 1.69).
1.10.8 Semi-empirical methods for assessment of bead morphology

In fusion welding fluid flow phenomena will have a strong effect on the shape of the weld
pool. Since flow in the weld pool is generally driven by a combination of buoyancy, electro-
magnetic, and surface tension forces (e.g. see Fig. 3.10 in Chapter 3), prediction of bead mor-
phology from first principles would require detailed consideration of the current and heat flux
distribution in the arc, the interaction of the arc with the weld pool free surface, convective
heat transfer due to fluid flow in the liquid pool, heat of fusion, convective and radiative losses
from the surface, as well as heat and mass loss due to evaporation.
Over the years, a number of successful studies have been directed towards numerical weld
pool modelling, based on the finite difference, the finite element, or the control volume ap-
proach.24"31 Although these studies provide valuable insight into the mechanisms of weld pool
development, the solutions are far too complex to give a good overall indication of the heat-
and fluid-flow pattern. The present treatment is therefore confined to a discussion of factors
affecting the nominal composition of single-bead fusion welds. This composition can be ob-
tained from an analysis of the amount of deposit D and the fused part of the base material B,
from which we can calculate the mixing ratio BI(B + D) or DI(B + D). Methods have been
outlined in the preceding sections for handling such problems by means of point or line source
models. The following section gives a brief description of procedures which can be used for
predictions of the desired quantities in cases where the classic models break down, or where
the calculation will be too tedious.
1.10.8.1 Amounts of deposit and fused parent metal

The heat conduction theory does not allow for the presence of deposited metal. The rate of
deposition, dMwldt, is roughly proportional to the welding current /, and is often reported as a
coefficient of deposition, defined as:
(1-119)
Since the area of deposited metal D is frequently wanted, we may write:
(1-120)
where p is the density, and v is the welding speed.

Plate thickness: Groove angle: <>
| = 60°
Cooling time, AtQ/5(s)
Heat input, E (kJ/mm)
Fig. 1.69. Comparison between observed and predicted cooling times from 800 to 5000C in root pass
welding of steel plates (groove preparation as in Fig. 1.68(b)). Data from Akselsen and Sagmo.34
Recommended values of k'/p for some arc welding processes are given in Table 1.7. In
practice, the deposition coefficient k'/p will also vary with current density and electrode stickout
due to resistance heating of the electrode. Consequently, the numbers contained in Table 1.7
are estimated averages, and should therefore be used with care.
Example (1.16)
Consider stringer bead deposition (S AW) on a thick plate of low alloy steel under the follow-
ing conditions:
Table 1.7 Average rates of volume deposition in arc welding. Data from Christensen.32
Welding Process k'/p (mm3 A"1 s l)
SMAW 0.3-0.5
GMAW, steel 0.6-0.7
GMAW, aluminium -0.9
SAW, steel -0.7
Calculate the mixing ratio Bf(B + D) at pseudo-steady state.
Solution
The amount of fused parent material can be obtained from equation (1-75). If we include an
empirical correction for the latent heat of melting, the dimensionless radius vector a4m be-
comes:
This gives:
Similarly, the amount of deposited metal can be calculated from equation (1-120). Taking
ifc'/p equal to 0.7mm3 A"1 s"1 for SAW (Table 1.7), we get:
The mixing ratio is thus:
Example (L 17)
Consider stringer bead deposition with covered electrodes (SMAW) on a thick plate of low
alloy steel under the following conditions:
Calculate the mixing ratio BI(B + D) at pseudo-steady state.
Solution
In this particular case the conditions for a fast moving high power source are not met. Thus, in
order to eliminate the risk of systematic errors, the amount of fused parent metal should be
calculated from the general Rosenthal thick plate solution (equation (1-45)) or read from Fig.
1.21. When Tc = Tm (i.e. 8^ = 1), we obtain:
Reading from Fig. 1.21 gives:

and
Moreover, the amount of deposited metal can be calculated from equation (1-120). Taking
k'/p equal to 0.4mm3 A"1 s~l for SMAW (Table 1.7), we get:
and
The above calculations indicate a small difference in the mixing ratio between SA and
SMA welding, but the data are not conclusive. In practice, a value of BI(B + D) between 1/3
and 1/2 is frequently observed for SMAW, while the mixing ratio for SAW is typically 2/3 or
higher. The observed discrepancy between theory and experiments arises probably from diffi-
culties in estimating the amount of fused parent metal from the point heat source model.
1.10.8.2 Bead penetration

It is a general experience in arc welding that the shape of the fusion boundary will depart quite
strongly from that of a semi-circle due to the existence of high-velocity fluid flow fields in the
weld pool.24"31 For combinations of operational parameters within the normal range of arc
welding, a fair prediction of bead penetration h can be made from the empirical equation
derived by Jackson et al.:33
(1-121)
A summary of Jackson's data is shown in Table 1.8. It is seen that the constant C in equa-
tion (1-121) has a value close to 0.024 for SAW and SMAW with E6015 type electrodes, and
about 0.050 for GMAW with CO2 -shielding gas. Penetration measurements of GMA/Ar + O2,
GMA/Ar, and GMA/He welds, on the other hand, show a strong dependence of polarity, and
shielding gas composition, to an extent which makes the equation useless for a general predic-
tion. Such data have therefore not been included in Table 1.8.
Example (1.18)
Based on the Jackson equation (equation (1-121)), calculate the bead penetration for the two
specific welds considered in Examples (1.16) and (1.17). Use these results to evaluate the
applicability of the point heat source model under the prevailing circumstances.
Solution
From equation (1-121) and Table 1.8, we have:
Table 1.8 Recommended bead penetration coefficients for some arc welding processes. Data from
Jackson.33
Welding Process C Comments
SAW, steel -0.024 Various types of fluxes

(Zz from 3 to 15 mm)
SMAW, steel -0.024 Wide range of /, U, and v
(E6015) (h from 0.7 to 5mm)
GMAW, steel -0.050 Electrode positive
(CO2 - shielding) (h from 6.5 to 8mm)
and
The corresponding values predicted from the point heat source model are:
and
Provided that the Jackson equation gives the correct numbers, it is obvious from the above
calculations that the point heat source model is not suitable for reliable predictions of the bead
penetration during arc welding. This observation is not surprising.
1.10.9 Local preheating

So far, we have assumed that the ambient temperature T0 remains constant during the welding
operation (i.e. is independent of time). The use of a constant value of T0 is a reasonable ap-
proximation if the work-piece as a whole is subjected to preheating. In many cases, however,
the dimensions of the weldment allow only preheating of a narrow zone close to the weld.
This, in turn, will have a significant influence on the predicted weld cooling programme, par-
ticularly in the low temperature regime where the classic models eventually break down when
T approaches T0.
Model (after Christensen n)

The idealised preheating model is shown in Fig. 1.70. Here it is assumed that the weld centre-
line temperature is equal to the sum of the contributions from the arc and from the field of
preheating. The former contribution is given by equation (1-45) for R = -x = Vt, provided that
the plate thickness is sufficiently large to maintain 3-D heat flow. Similarly, the temperature
field due to preheating can be calculated as shown in Section 1.7 for uniaxial heat conduction
from extended sources (thermit welding). By combining equations (1-45) and (1-22), we
obtain the following relation for the weld centre-line:
(1-122)
T
Temperature profile
att = 0
Weld
Preheated
zone
z
Fig. 1.70. Sketch of preheating model.
where T* is the local preheating temperature, and L* is the half width of the preheated zone.
Equation (1-122) can be written in a general form by introducing the following groups of par-
ameters:
— Dimensionless temperature:
(1-123)
Time constant:
(1-124)
Dimensionless time:
(1-125)
Dimensionless half width of preheated zone:
(1-126)
By inserting these parameters into equation (1-122), we obtain:
(1-127)
It is evident from the graphical representation of equation (1-127) in Fig. 1.71 that the
predicted weld cooling programme falls within the limits calculated for Q"-^ 0 (no preheating)
and Q /7 -> oo (global preheating). The controlling parameter is seen to be the dimensionless
half width of the preheated zone Q", which depends both on the actual width L*, the base plate
thermal properties a, X, and the net heat input qo Iv.
Example (1.19)
Consider stringer bead deposition with covered electrodes (SMAW) on a thick plate of low
alloy steel under the following conditions:
Calculate the cooling time from 800 to 50O0C (A%5), and the cooling time £1Oo measured
from the moment of arc passage to the temperature in the centre of the weld reaches 10O0C.
Solution
First we calculate the time constant to from equation (1-124):
from which we obtain

e*-
^6
Fig. 1.71. Graphical representation of equation (1-127).

Next Page
Cooling time, At8/5

The dimensionless temperatures conforming to 800 and 5000C are:
from which
This cooling time is only slightly longer than that calculated from equation (1-68) for T0 =
200C (6.9s), showing that moderate preheating up to 1000C is not an effective method of
controlling Ar875.
Cooling time, t]00

When T=T0* = 1000C, the dimensionless temperature 9* = 1. Reading from Fig. 1.71 gives
T 6 - 10, from which:
The above value should be compared with that evaluated from the numerical data of Yurioka
et al.,35 replotted in Fig. 1.72 (see p.104). It follows from Fig. 1.72 that the weld cooling
programme in practice is also a function of the plate thickness d, an effect which cannot read-
ily be accounted for in a simple analytical treatment of the heat diffusion process. For the
specific case considered above the parameter ^100 varies typically from 500 to 900s, depending
on the chosen value of d. This cooling time is significantly shorter than that calculated from
equation (1-127), indicating that the analytical model is only suitable for qualitative predic-
tions.
References
1. H.S. Carslaw and J.C. Jaeger: Conduction of Heat in Solids; 1959, Oxford, Oxford University
Press.
2. British Iron and Steels Research Association: Physical Constants of some Commercial Steels
at Selected Temperatures; 1953, London, Butterworths.
3. R. Hultgren, R.L. Orr, RD. Anderson and K.K. Kelly: Selected Values of Thermodynamic
Properties of Metals and Alloys; 1963, New York, J. Wiley & Sons.
4. E. Griffiths (ed.): J. Iron and Steel Inst., 1946,154, 83-121.
5. J.E. Hatch (ed.): Aluminium — Properties and Physical Metallurgy; 1984, Metals Park (Ohio),
American Society for Metals.
6. Metals Handbook, 9th edn., Vol. 2, 1979, Metals Park (Ohio), American Society for Metals.
Previous Page
Cooling time, At8/5

The dimensionless temperatures conforming to 800 and 5000C are:
from which
This cooling time is only slightly longer than that calculated from equation (1-68) for T0 =
200C (6.9s), showing that moderate preheating up to 1000C is not an effective method of
controlling Ar875.
Cooling time, t]00

When T=T0* = 1000C, the dimensionless temperature 9* = 1. Reading from Fig. 1.71 gives
T 6 - 10, from which:
The above value should be compared with that evaluated from the numerical data of Yurioka
et al.,35 replotted in Fig. 1.72 (see p.104). It follows from Fig. 1.72 that the weld cooling
programme in practice is also a function of the plate thickness d, an effect which cannot read-
ily be accounted for in a simple analytical treatment of the heat diffusion process. For the
specific case considered above the parameter ^100 varies typically from 500 to 900s, depending
on the chosen value of d. This cooling time is significantly shorter than that calculated from
equation (1-127), indicating that the analytical model is only suitable for qualitative predic-
tions.
References
1. H.S. Carslaw and J.C. Jaeger: Conduction of Heat in Solids; 1959, Oxford, Oxford University
Press.
2. British Iron and Steels Research Association: Physical Constants of some Commercial Steels
at Selected Temperatures; 1953, London, Butterworths.
3. R. Hultgren, R.L. Orr, RD. Anderson and K.K. Kelly: Selected Values of Thermodynamic
Properties of Metals and Alloys; 1963, New York, J. Wiley & Sons.
4. E. Griffiths (ed.): J. Iron and Steel Inst., 1946,154, 83-121.
5. J.E. Hatch (ed.): Aluminium — Properties and Physical Metallurgy; 1984, Metals Park (Ohio),
6. Metals Handbook, 9th edn., Vol. 2, 1979, Metals Park (Ohio), American Society for Metals.
Heat input:
E=1.7 kJ/mm
Cooling time, t100(s)
Preheating temperature, T^ (0C)
Fig. 1.72. Cooling time to 1000C, tm, in steel welding for different combinations of T0*, L*, d and E.
Data from Yurioka et a/.35
7. K.P. Bentley, J.A. Greenwood, R McKnowlson and R.G. Bakes: Brit. Weld. J., 1963,10, 613-
619.
8. N.N. Rykalin, A.I. Pugin and V.A. Vasil'eva: Weld. Prod., 1959, 6, 42-52.
9. N.N. Rykalin: Berechnung der Warmevorgdnge beim Schweissen; 1953, Berlin, VEB Verlag
Technik.
10. H.M. Tensi, W. Welz and M. Schwalm: Aluminium, 1981, 58, 515-518.
11. N. Christensen, V. de L. Davis and K. Gjermundsen: Brit. Weld. J., 1965,12, 54-75.
12. Welding Handbook, 8th edn., Vol. 2, 1991, Miami (Florida), American Welding Society.
13. D. Rosenthal: Weld. / , 1 9 4 1 , 20, 220s-234s.
14. D. Rosenthal: Trans. ASME, 1946, 68, 849-866.
15. CM. Adams: Weld. J., 1958, 37, 210s-215s.
16. RS. Myers, O.A. Uyehara and G.L. Borman: Weld. Res. Bull., 1967, 123, 1-46.
17. T.W Eagar and N.S. Tsai: Weld. J., 1983, 62, 346s-355s.
18. M.F. Ashby and K.E. Easterling: Ada Metall, 1984, 32, 1935-1948.
19. 0. Grong and N. Christensen: Mater. ScL Tech., 1986, 2, 967-973.
20. O.R. Myhr and 0 . Grong: Acta Metall. Mater., 1990, 38, 449-460.
21. S. Kou and Y. Le: Metall. Trans., 1983,14A, 2245-2253.
22. O.R. Myhr and 0. Grong: Unpublished work, 1990, University of Trondheim, The Norwegian
Institute of Technology.
23. R.W. Niles and CE. Jackson: Weld. J., 1975, 54, 25s-32s.
24. G.M. Oreper, T.W. Eagar and J. Szekely: Weld J., 1983, 62, 307s-312s.
25. Y.H. Wang and S. Kou: Proc. Int. Conf. on Trends in Welding Research, Gatlinburg, TN, May,
1986, pp. 65-69, Publ. ASM International.
26. S.A. David and J.M Vitek: Int. Mater. Rev., 1989, 34, 213-245.
27. K.C. Mills and BJ. Keene: Int. Mater. Rev., 1990, 35, 185-216.
28. R.L. UIe, Y. Joshi and E.B. Sedy: Metall. Trans., 1990, 21B, 1033-1047.
29. T. Zacharia, S.A. David, J.M. Vitek and H.G. Kraus: Metall. Trans., 1991, 22B, 243-257.
30. A. Matsunawa: Proc. 3rd Int. Conf. on Trends in Welding Research, Gatlinburg, TN, 1992,
pp.3-16, Publ. ASM International.
32. N. Christensen: Welding Metallurgy Compendium, 1985, University of Trondheim, The Nor-
wegian Institute of Technology.
33. CE. Jackson: Weld. J., 1960, 39, 226s-230s.
34. O.M. Akselsen and G. Sagmo: Technical Report STF34 A89147, 1989, Trondheim (Norway),
Sintef-Division of Metallurgy.
35. N. Yurioka, M. Okumura, S. Ohshita and S. Saito: HW Doc. XII-E-10-81, 1981.
Appendix 1.1
Nomenclature
General symbols finite difference method
thermal diffusivity (mm2 s"1) heat content per unit volume at Tc

(J mm"3)
cross section (mm2)
enthalpy increment referred to an
start temperature of ferrite to initial temperature T0 (J mm' 3 )
austenite transformation (0C)
latent heat of melting (J mm"3)
end temperature of ferrite to
austenite transformation (0C) amperage (A)
recrystallisation temperature (0C) modified Bessel function of second

kind and zero order
cooling rate (0C s"1)
modified Bessel function of second
plate thickness (mm) kind and first order
natural logarithm base number integration parameter
Gaussian error function integration parameter
complementary Gaussian error start temperature of austenite to

function martensite transformation (0C)
integration parameter y-axis/transverse direction (mm)
net power (W) z-axis/through thickness direction

(mm)
net heat input (J)
efficiency factor
two-dimensional radius vector
(mm) dimensionless temperature
locus of peak temperature in T-r dimensionless temperature con-

space (mm) forming to 8000C
three-dimensional radius vector dimensionless temperature con-

(mm) forming to 5000C
locus of peak temperature in T-R dimensionless temperature con-

space (mm) forming to the melting point
isothermal zone width (mm) dimensionless peak temperature
temperature (0C) volume heat capacity

(J mm-3 0C-1)
reference temperature (0C)
thermal conductivity
ambient temperature (0C) (W mm"1 0C-1)
melting point (0C) dimensionless time
peak temperature (0C) dimensionless cooling time
time (s) dimensionless cooling time from

800 to 5000C
time variable (s)
dimensionless cooling time from
time variable (s) 300 to 1000C
cooling time (s)

Specific symbols
0
cooling time from 800 to 500 C (s)
Local Fusion in Arc Strikes
0
cooling time from 300 to 100 C (s)
dimensionless operating parameter
integration parameter
arc ignition time (s)
voltage (V)
isothermal zone width (mm)
integration parameter
dimensionless R-vector
x-axis/welding direction (mm)
locus of peak temperature in G-CT1
space
dimensionless isothermal zone locus of peak temperature in 0'-£2
width space (star denotes a specific peak
temperature)
dimensionless time
dimensionless time
locus of peak temperature in 0-T1
space dimensionless cooling time from
^ to 5000C
dimensionless cooling time
locus of peak temperature in (T-T3
space
Spot Welding
thickness of overlapping plates Friction Welding

(mm)
integral in equation (1-30)
net power generation at weld in-
heating time (s) terface (W)
dimensionless r-vector duration of heating period (s)
locus of peak temperature in 0-cr2 contact section temperature at the

space end of heating period (0C)
dimensionless time dimensionless temperature
locus of peak temperature in 0-T2 dimensionless peak temperature

space
dimensionless jc-coordinate
dimensionless cooling time
locus of peak temperature in 6"-
Q.' space
Thermit Welding dimensionless time
half width of groove (mm)
Arc Welding
initial temperature of liquid metal
(0C)
amount of fused parent metal
(mm2)
distance from reference point to in-
finitesimal source (mm)
constant in Jackson equation
dimensionless temperature
amount of deposited metal (mm2)
dimensionless peak temperature
gross heat input per unit length of
weld (kJ mm"1)
dimensionless .^-coordinate
correction factor for the net heat
input during root pass welding
FCAW flux cored arc welding imaginary heat source of net arc
power qo, qa, or qb (W)
GMAW gas metal arc welding
maximum intensity of distributed
GTAW gas tungsten arc welding (Gaussian) heat source (W mm"1)
bead penetration (mm) power density of distributed

(Gaussian) heat source (W mm"1)
integer variables.... -1,0, L...
infinitesimal heat source (W)
constant in heat distribution func-
tion (mm"2) two-dimensional radius vector in
y-z plane (mm)
coefficient of weld metal deposi-
tion (g A"1 s"1) locus of peak temperature in T-r*
space (mm)
amplitude of weaving or half width
of strip electrode (mm) distance from infinitesimal heat
source to point P in x-y or y-z plane
half width of preheated zone (mm) (mm)
equivalent amplitude of weaving position of weld end-crater with

(mm) respect to imaginary heat source at
time t** in x-y plane (mm)
half width of linear source in
Gaussian heat distribution model distance from infinitesimal heat
(mm) source to point P in x-y-z space
(mm)
length of weld bead (mm)
k distances from real and imaginary
mass of weld metal (g) heat sources to point P in x-y-z
space (mm)
position of weld with respect to
dimensionless operating parameter imaginary heat source at time f**
in weaving model in x-y-z space (mm)
reference point in stationary coor- SAW submerged arc welding

dinate system
SMAW shielded metal arc welding
reference point in moving coordi-
nate system time at moment of arc extinction
(s)
arbitrary reference point
time constant in preheating model
arbitrary point of observation (S)
strength of elementary heat cooling time to 1000C (s)

sources (W)
time constant in heat distribution displacement of elementary heat
function (s) source in z-direction (mm)
time referred to moment of arc ig- groove angle

nition (s)
ratio between real and pseudo-steady
retention time (s) state temperature (thick plate
welding)
preheating temperature (0C)
ratio between real and pseudo-
welding speed (mm s"1) steady state temperature (thin plate
welding)
volume of isothermal enclosure
(mm3) dimensionless temperature in
preheating model
welding direction in stationary co-
ordinate system (mm) dimensionless cross sectional area
of isothermal enclosure (thick plate
^-coordinate at maximum width of welding)
isotherm (mm)
dimensionless cross sectional area
transverse direction in stationary of isothermal enclosure (thin plate
coordinate system (mm) welding)
distance from infinitesimal heat density of weld metal (g mm 3 )

source to point P in j-direction
(mm) dimensionless distance from real
and imaginary heat sources to point
^-coordinate at maximum width of P
isotherm (mm)
dimensionless /^-vector
isothermal zone width (mm)
dimensionless R0-vector
through-thickness direction in sta-
tionary coordinate system (mm) locus of peak temperature in 0-a3
space
z-coordinate at maximum width of
isotherms (mm) dimensionless r*-vector
isothermal zone width (mm) locus of peak temperature in 0-04

space
dimensionless y-coordinate in
weaving model dimensionless r-vector
dimensionless z-coordinate in dimensionless r^-vector

weaving model
locus of peak temperature in 8-CT5
dimensionless plate thickness space
dimensionless jc-axis dimensionless zo -axis
dimensionless xo -axis dimensionless z-coordinate at

maximum depth of isotherm
dimensionless jc-coordinate at
maximum width of isotherm melting efficiency factor
dimensionless length of isothermal dimensionless volume of isother-

enclosure mal enclosure
dimensionless distance from heat dimensionless half width of

source to front of isothermal enclo- preheated zone
sure
dimensionless time in weaving
dimensionless distance from heat model
source to rear of isothermal enclo-
sure dimensionless time in preheating
model
dimensionless y-axis
locus of peak temperature in 6-x
dimensionless yo -axis space
dimensionless ^-coordinate at dimensionless retention time

maximum width of isotherm
dimensionless time at moment of
dimensionless isothermal zone arc extinction
width
dimensionless time referred to mo-
dimensionless z-axis ment of arc ignition
Appendix 1.2
Refined Heat Flow Model for Spot Welding
The refined model is based on the assumption that all heat is released instantaneously at time
t = 0 in a point located at the interface between the two overlapping plates, which implies that
equation (1-7) is valid. However, in order to maintain the net flux of heat through both plate
surfaces equal to zero, it is necessary to account for mirror reflections of the source with
respect to the planes z = dt/2 and z = - dt/2. This can be done on the basis of the method of
images, as illustrated in Fig. A 1.1. By including all contributions from the imaginary sources
"Q-2 >Q-i ,Q\,Qi >••• located symmetrically at distances ± idt below and above the centre-axis
of the joint, the temperature distribution is obtained in the form of a convergent series:
(Al-I)
where
y
Z
Fig. Al.l.Refined heat flow model for spot welding of plates.
and i is an integer variable (...-1, 0, 1...)-
A numerical solution of equation (Al.l) gives a peak temperature distribution similar to

that shown in Fig. 1.9.
Appendix 1.3
The Gaussian Error Function
The eiTor function erf(u) and the complementary error function erfc(u) are special cases of the
incomplete gamma function. Their definitions are:
and
The functions have the following limiting values and symmetries:
and
The following Fortran subroutine can be used for calculations of the error functions with a
fractional error less than 1.2 X 10~7:
FUNCTION ERFC(U)
Z=ABS(U)
T=l./(l.+0.5*Z)
ERFC=T*EXP(-Z*Z-1.26551223+T*(1.00002368+T*(.37409196+
* T%09678418+T*(-.18628806+T%27886807+T*(-l.13520398+
* T*(1.48851587+T*(-.82215223+T*.17O87277)))))))))
IF (U.LT.O.) ERFC=2.-ERFC
RETURN
END
Appendix 1.4
Gaussian Heat Distribution
Following the treatment of Rykalin,9 the situation may be described as shown in Fig. Al.2.
Here we consider a distributed heat source of net power density (in W mm"1):
(Al-2)
The total power of the source qo is obtained by integration of equation (Al-2). Substituting
and integrating from u = -©© to u = +<*>, gives:
from which
(Al-3)
It follows from equation (1-73) that an infinitesimal source dqy> located between j ' a n d
will cause a small rise of temperature dTy> in point P at time t, as:
(Al-4)
where
and
Integration of equation (Al-4) between the limits y'= -°o and >>'= +00 gives:
(Al-5)
q(y)
P
2-D heat flow
z
Fig, Al.2. Distributed heat source of net power density q(y) on a semi-infinite body.
where is a time constant)and n = Aat.
The latter integral can be evaluated by substituting:
from which
and integrating between the limits w = -°° and
This gives (after some manipulation):
(Al-6)
If we replace the Gaussian heat distribution by a linear source of the same strength, which
extends from -L0 to +L0 on either side of the weld centre-line in the y-direction (see Fig. A1.3),
we may write:
By rearranging this equation, we obtain:
(Al-7)
In practice, the parameter L0 has the same physical significance as the weaving amplitude
L in equation (1-110). Consequently, these solutions are equivalent in the sense that they
predict a similar temperature-time pattern.
q(y)
Fig. Al.3. Physical representation of a Gaussian heat distribution by a linear source of width 2LO.
2
Chemical Reactions in Arc Welding
2.1 Introduction
The weld metal composition is controlled by chemical reactions occurring in the weld pool at
elevated temperatures, and is therefore influenced by the choice of welding consumables (i.e.
combination of filler metal, flux, and/or shielding gas), the base metal chemistry, as well as the
operational conditions applied. In contrast to ladle refining of metals and alloys where the
reactions occur under approximately isothermal conditions, a characteristic feature of the arc
welding process is that the chemical interactions between the liquid metal and its surroundings
(arc atmosphere, slag) take place within seconds in a small volume where the metal tempera-
ture gradients are of the order of 1000°C mm"1 with corresponding cooling rates up to
1000°C s"1. The complex thermal cycle experienced by the liquid metal during transfer from
the electrode tip to the weld pool in GMA welding of steel is shown schematically in Fig. 2.1.
As a result of this strong non-isothermal behaviour, it is very difficult to elucidate the
reaction sequences during all stages of the process. Consequently, a complete understanding
of the major controlling factors is still missing, which implies that fundamentally based pre-
dictions of the final weld metal chemical composition are limited. Additional problems result
from the lack of adequate thermodynamic data for the complex slag-metal reaction systems
involved. However, within these restrictions, the development of weld metal compositions
can be treated with the basic principles of thermodynamics and kinetic theory considered in
the following sections.
2.2 Overall Reaction Model
In ladle refining of metals and alloys, the reaction kinetics are usually controlled by mass
transfer between the liquid metal and its surroundings (slag or ambient atmosphere). Exam-
ples of such kinetically controlled processes are separation of non-metallic inclusions from a
deoxidised steel melt or removal of hydrogen from liquid aluminium. In welding, the reaction
pattern is more difficult to assess because of the characteristic non-isothermal behaviour of the
process (see Fig. 2.1). Nevertheless, experience shows that it is possible to analyse mass
transfer in welding analogous to that in ladle refining by considering a simple two-stage reac-
tion model, which assumes:1
(i) A high temperature stage, where the reactions approach a state of local
pseudo-equilibrium.
(ii) A cooling stage, where the concentrations established during the initial stage
tend to readjust by rejection of dissolved elements from the liquid.
Gas nozzle
Contact tube
Shielding gas
Filler wire
Arc plasma
temperature~10000°C Electrode tip droplet
(1600-20000C)
Cold part of Falling droplet (24000C)
weld pool (< 19000C)
Hot part of weld pool
(1900-22000C)
Weld pool Base plate

retention
time 2-1Os
Fig. 2.1. Schematic diagram showing the main process stages in GMA welding. Characteristic average
temperature ranges at each stage are indicated by values in parenthesis.
As indicated in Fig. 2.2 the high temperature stage comprises both gas/metal and slag/metal
interactions occurring at the electrode tip, in the arc plasma, or in the hot part of the weld pool,
and is characterised by extensive absorption of elements into the liquid metal. During the
subsequent stage of cooling following the passage of the arc, a supersaturation rapidly in-
creases because of the decrease in the element solubility with decreasing temperatures. The
system will respond to this supersaturation by rejection of dissolved elements from the liquid,
either through a gas/metal reaction (desorption) or by precipitation of new phases. In the latter
case the extent of mass transfer is determined by the separation rate of the reaction products in
the weld pool. It should be noted that the boundary between the two stages is not sharp, which
means that phase separation may proceed simultaneously with absorption in the hot part of the
weld pool.
In the following sections, the chemistry of arc welding will be discussed in the light of this
two-stage reaction model.
2.3 Dissociation of Gases in the Arc Column
As shown in Table 2.1, gases such as hydrogen, nitrogen, oxygen, and carbon dioxide will be
widely dissociated in the arc column because of the high temperatures involved (the arc plasma
temperature is typically of the order of 10 0000C or higher). From a thermodynamic stand-
point, dissociation can be treated as gaseous chemical reactions, where the concentrations of
the reactants are equal to their respective partial pressures. Hence, for dissociation of diatomic
gases, we may write:
(2-1)
where X denotes any gaseous species.

'Hot1 part of 'Cold' part of weld pool
weld pool
Solid weld metal

Peak temperature
Solid weld metal

Temperature
Grey j
zonei
Absorption Rejection of
of elements dissolved elements
Solid weld metal

Peak concentration
Solid weld metal
Concentration
Equilibrium concentration at melting point
Time
Fig. 2.2. Idealised two-stage reaction model for arc welding (schematic).
Table 2.1 Temperature for 90% dissociation of some gases in the arc column. Data from Lancaster.2
Gas Dissociation Temperature (K)
CO 2 3800
H2 4575
O2 5100
N2 8300
Next, consider a shielding gas which consists of two components, i.e. one inert component
(argon or helium) and one active component X2. When the fraction dissociated is close to
unity, the partial pressure of species X in the gas phase px is equal to:
(2-2)
where H1 and nx are the total number of moles of components / (inert gas) and X, respectively
in the shielding gas, andptot is the total pressure (in atm).
It follows from equation (2-1) that two moles of X form from each mole of X2 that dissoci-
ates. Hence, equation (2-2) can be rewritten as:
(2-3)
where nXl is the total number of moles of component X2 which originally was present in the
shielding gas.
If nXl and H1 are proportional to the volume concentrations of the respective gas compo-
nents in the shielding gas, equation (2-3) becomes:
(2-4)
Taking vol% / = (100 - vol% X2) andp,ot = 1 atm, we obtain the following expression for
Px-
(2-5)
Similarly, if X2 is replaced by another gas component of the type YX2, we get:
(2-6)
and
(2-7)
It is evident from the graphical representations of equations (2-5) and (2-7) in Fig. 2.3 that
the partial pressure of the dissociated component X increases monotonically with increasing
concentrations of X2 and YX2 in the shielding gas. The observed non-linear variation of px
arises from the associated change in the total number of moles of constituent species in the gas
phase due to the dissociation reaction. Moreover, it is interesting to note that the partial pres-
sure px is also dependent on the nature of the active gas component in the arc column (i.e. the
stoichiometry of the reaction). This means that the oxidation capacity of for instance CO2 is
only half that of O2 when comparison is made on the basis of equal concentrations in the
shielding gas (to be discussed later).
Px
Vol%)^ f VoRGYX2
Fig. 2.3. Graphical representation of equations (2-5) and (2-7).
2.4 Kinetics of Gas Absorption
In general, mass transfer between a gas phase and a melt involves:3
(i) Transport of reactants from the bulk phase to the gas/metal interface.
(ii) Chemical reaction at the interface.
(iii) Transport of dissolved elements from the interface to the bulk of the metal.
2.4.1 Thin film model

In cases where the rate of element absorption is controlled by a transport mechanism in the gas
phase (step one), it is a reasonable approximation to assume that all resistance to mass transfer
is confined to a stagnant layer of thickness 8 (in mm) adjacent to the metal surface, as shown in
Fig. 2.4. Under such conditions, the overall mass transfer coefficient is given by:2
(2-8)
where Dx is the diffusion coefficient of the transferring species X (in mm2 s~*).
Although the validity of equation (2-8) may be questioned, the thin film model provides a
simple physical picture of the resistance to mass transfer during gas absorption.
Partial pressure
Distance
Fig. 2.4. Film model for mass transfer (schematic).
2.4.2 Rate of element absorption

Referring to Fig. 2.5, the rate of mass transfer between the two phases (in mol s"1) can be
written as:
(2-9)
where A is the contact area (in mm 2 ), R is the universal gas constant (in mm3 atm K"1 mol"1),
T is the absolute temperature (in K), px is the partial pressure of the dissociated species X in the
bulk phase (in atm), and px is the equilibrium partial pressure of the same species at the gas/
metal interface (in atm).
Based on equation (2-9) it is possible to calculate the transient concentration of element X
in the hot part of the weld pool. Let m denote the total mass of liquid weld metal entering/
leaving the reaction zone per unit time (in g s"1). If Mx represents the atomic weight of the
element (in g mol"1), we obtain the following relation w h e n / ? x » p°x:
(2-10)
It follows from equation (2-10) that the transient concentration of element X in the hot part
of the weld pool is proportional to the partial pressure of the dissociated component X in the
plasma gas. Since this partial pressure is related to the initial content of the molecular species
X2 or YX2 in the shielding gas through equations (2-5) and (2-7), we may write:
Arc column
Bulk gas
phase
Stagnant gaseous
boundary layer
Gas/metal interface
Metal phase
Hot part of
weld pool
Fig. 2.5. Idealised kinetic model for gas absorption in arc welding (schematic).
(2-11)
and
(2-12)
where C1 and C2 are kinetic constants which are characteristic of the reaction systems under
consideration.
2.5 The Concept of Pseudo-Equilibrium
Although the above analysis presupposes that the element absorption is controlled by a
transport mechanism in the gas phase, the transient concentration of the active component X in
the hot part of the weld pool can alternatively be calculated from chemical thermodynamics by
considering the following reaction:
X(gas) X (dissolved) (2-13)
By introducing the equilibrium constant K{ for the reaction and setting the activity coeffi-
cient to unity, we get:
(2-14)
This equation should be compared with equation (2-10) which predicts a linear relationship
between wt% X and px. If the above analysis is correct, one would expect that the partial
pressure px at the gas/metal interface is directly proportional to the partial pressure of the
dissociated component in the bulk phase. Unfortunately, the proportionality constant is diffi-
cult to establish in practice.
2.6 Kinetics of Gas Desorption
During the subsequent stage of cooling following the passage of the arc, the concentrations
established at elevated temperatures will tend to readjust by rejection of dissolved elements
from the liquid. When it comes to gases such as hydrogen and nitrogen, this occurs through a
desorption mechanism, where the driving force for the reaction is provided by the decrease in
the element solubility with decreasing metal temperatures.
2.6.1 Rate of element desorption

Consider a melt which first is brought in equilibrium with a monoatomic gas of partial pres-
sure px at a high temperature T1, and then is rapidly cooled to a lower temperature T2 and
immediately brought in contact with diatomic X2 of partial pressure pXl (see Fig. 2.6). Under
such conditions, the rate of element desorption (in mol s"1) is given by:
(2-15)
where k'd is the mass transfer coefficient (in mm s 1X and p°x is the equilibrium partial pres-
sure of component X2 at the gas/metal interface (in atm).
Bulk gas
phase
Stagnant gaseous
boundary layer
Gas/metal interface
Metal phase
Cold part of
weld pool
Fig. 2.6. Idealised kinetic model for gas desorption in arc welding (schematic).
The partial pressure pX2 can be calculated from chemical thermodynamics by considering
the following reaction:
2X(dissolved) = X2 (gas) (2-16)
from which
(2-17)
where K2 is the equilibrium constant, and [wt% X] is the concentration of element X in the
liquid metal (in weight percent). Note that the activity coefficient has been set to unity in the
derivation of equation (2-17).
The equilibrium constant K2 may be expressed in terms of the solubility of element X in the
liquid metal at 1 atm total pressure Sx. Hence, equation (2-17) transforms to:
(2-18)
By combining equations (2-15) and (2-18), we get:
(2-19)
Data for the solubility of hydrogen and nitrogen in some metals up to about 22000C are
given in Figs. 2.7 and 2.8, respectively. It is evident that the element solubility decreases
steadily with decreasing metal temperatures down to the melting point. This implies that the
desorption reaction is thermodynamically favoured by the thermal conditions existing in the
cold part of the weld pool.
2.6.2. Sievert's law

It follows from equation (2-19) that desorption becomes kinetically unfeasible when
Px2 ~ Px2' corresponding to:
(2-20)
Equation (2-20) is known as the Sievert's law. This relation provides a basis for calculating
the final weld metal composition in cases where the resistance to mass transfer is sufficiently
small to maintain full chemical equilibrium between the liquid metal and the ambient (bulk)
gas phase.
2.7 Overall Kinetic Model for Mass Transfer during Cooling in the Weld Pool
Because of the complexity of the rate phenomena involved, it would be a formidable task to
derive a complete kinetic model for mass transfer in arc welding from first principles. How-
(a) (b)
Aluminium
ml H2/100 g fused metal
ml H2/100 g fused metal

Copper
Solid Cu
Temperature, 0C Temperature, 0C
(C) (d)
ml H2/100g fused metal
ml H2/100g fused metal
Iron Nickel
Temperature, 0C Temperature, 0C
Fig. 2.7. Solubility of hydrogen in some metals; (a) Aluminium, (b) Copper, (c) Iron, (d) Nickel. Data
compiled by Christensen.4
ever, for the idealised system considered in Fig. 2.9, it is possible to develop a simple math-
ematical relation which provides quantitative information about the extent of element transfer
occurring during cooling in the weld pool. Let [%X]eq denote the equilibrium concentration of
element X in the melt. If we assume that the net flux of element X passing through the phase
boundary A per unit time is proportional to the difference ([%X] - [%X]eqX the following
balance is obtained:3
(2-21)
where V is the volume of the melt (in mm 3 ), kd is the overall mass transfer coefficient (in
mm s"1), and A is the contact area between the two phases (in mm 2 ).
Temperature, 0C
Iron
log (wt% N)
104AT1 K
Fig. 2.8. Solubility of nitrogen in iron. Data from Turkdogan.5
Phase I i
Contact area (A)

Distance
Net flux of X
Phase i Volume (V)
Concentration
Fig. 2.9. Idealised kinetic model for mass transfer in arc welding (schematic).
By rearranging equation (2-21) and integrating between the limife [%X]( (att = O) and [%X]
(at an arbitrary time t\ we get:
(2-22)
where to is a time constant (equal to VI kjA).

It is evident from the graphical representation of equation (2-22) in Fig. 2.10 that the rate of
mass transfer depends on the ratio Vl kji, i.e. the time required to reduce the concentration of
element X to a certain level is inversely proportional to the mass transfer coefficient kd. This
type of response is typical of a first order kinetic reaction.
Although the above model refers to mass transfer under isothermal conditions, it is also
applicable to welding if we assume that the weld cooling cycle can be replaced by an equiva-
lent isothermal hold-up at a chosen reference temperature. Thus, by rearranging equation (2-
22), we get:
(2-23)
It follows that the final concentration of element X in the weld metal depends both on the
cooling conditions and on the intrinsic resistance to mass transfer, combined in the ratio t/to.
When [%X]eq is sufficiently small, equation (2-23) predicts a direct proportionality between
[%X] and [%X\t (i.e. the initial concentration of element X in the weld pool). This will be the
case during deoxidation of steel weld metals where separation of oxide inclusions from the
weld pool is the rate controlling step. Moreover, when t/t0 » 1 (small resistance to mass
transfer), equation (2-23) reduces to:
(2-24)
Under such conditions the final weld metal composition can be calculated from simple
chemical thermodynamics.
Because of this flexibility, equation (2-23) is applicable to a wide range of metallurgical
problems at the same time as it provides a simple physical picture of the resistance to mass
transfer during cooling in the weld pool.
(X-X^)Z(X1-Xeq)
t,s
Fig. 2.10. Graphical representation of equation (2-22).

2.8 Absorption of Hydrogen
Some of the well-known harmful effects of hydrogen discussed in Chapters 3 and 7 (i.e. weld
porosity and HAZ cold cracking) are closely related to the local concentration of hydrogen
established in the weld pool at elevated temperatures due to chemical interactions between
the liquid metal and its surroundings.
2.8.1 Sources of hydrogen

Broadly speaking, the principal sources of hydrogen in welding consumables are:6
(i) Loosely bound moisture in the coating of shielded metal arc (SMA) electrodes and in
the flux used in submerged arc (SA) or flux-cored arc (FCA) welding. Occasionally, moisture
may also be introduced through the shielding gas in gas metal arc (GMA) and gas tungsten arc
(GTA) welding.
(ii) Firmly bound water in the electrode coating or the welding flux. This can be in the
form of hydrated oxides (e.g. rust on the surface of electrode wires and iron powder), hydro-
carbons (in cellulose), or crystal water (bound in clay, astbestos, binder etc.).
(iii) Oil, dirt and grease, either on the surface of the work piece itself, or trapped in the
surface layers of welding wires and electrode cored wires.
It is evident from Fig. 2.11 that the weld metal hydrogen content may vary strongly from
one process to another. The lowest hydrogen levels are usually obtained with the use of low-
moisture basic electrodes or GMA welding with solid wires. Submerged arc welding and flux-
cored arc welding, on the other hand, may give high or low concentrations of hydrogen in the
weld metal, depending on the flux quality and the operational conditions applied (note that the
former process is not included in Fig. 2.11). The highest hydrogen levels are normally associ-
ated with cellulosic, acid, and rutile type electrodes. This is due to the presence of large
amounts of asbestos, clay and other hydrogen-containing compounds in the electrode coating.
Table 2.2 (shown on page 132) gives a summary of measured arc atmosphere compositions
in GMA and SMA welding. Included are also typical ranges for the weld metal hydrogen
content.
2.8.2 Methods of hydrogen determination in steel welds

Hydrogen is unlike other elements in weld metal in that it diffuses rapidly at normal room
temperatures, and hence, some of it may be lost before an analysis can be made. This, coupled
with the fact that the concentrations to be measured are usually at the parts per million level,
means that special sampling and analysis procedures are needed. In order that research results
may be compared between different laboratories and can be used to develop hydrogen control
procedures, some international standardisation of these sampling and analysis methods is nec-
essary.
Three methods are currently being used, as defined in the following standards:
Potential hydrogen level
FCAW
Very Low Medium High

low
Weld hydrogen level
Fig. 2.11. Ranking of different welding processes in terms of hydrogen level (schematic). The diagram is
based on the ideas of Coe.6
(i) The Japanese method (JIS Z 313-1975), which has been adopted with important ad-
justments from the former ASTM designation A316-48T. This method involves collection of
released hydrogen from a single pass weld above glycerine for 48h at 45 0 C. The total volume
of hydrogen is reported in ml per 10Og deposit. Only 5 s of delay are allowed from extinction
of the arc to quenching.
(ii) The French method (N.F.A. 81-305-1975) where two beads are deposited onto core
wires placed in a copper mould. Hydrogen released from this bead is collected above mercury,
and the volume is reported in ml per 10Og fused metal (including the fused core wire metal).
(iii) The International Institute of Welding (HW) method (ISO 3690-1977), where a single
bead is deposited on previously degassed and weighed mild steel blocks clamped in a quick-
release copper fixture. The weldment is quenched and refrigerated according to a rigorously
specified time schedule. Hydrogen released from the specimens is collected above mercury
for 72 h at 25°C, and the results are reported in ml per 10Og deposit, or in g per ton fused metal.
To avoid confusion, it is recommended to use the symbol HDM for the content reported in terms
of deposited metal (ml per 10Og deposit), and HFM for the content referred to fused metal (ml
per 100 g or g per ton fused metal). The relationship between HDM and HFM is shown in Fig.
2.12.
As would be expected, these three methods do not give identical results when applied to a
given electrode. Approximate correlations have been established between the HW criteria
HDM and HFM and the numbers obtained by the Japanese and the French methods (designated
HJIS and HFR, respectively). For covered electrodes tested at various hydrogen levels, we
have:7
Fig. 2.12. The relation between HDM and HFM (0.9 is the conversion factor from ml per 10Og to g per ton).
(2-25)
(2-26)
The conversion factor from HFR to HFM applies to a ratio of deposited to fused metal,
DI(B + D), equal to 0.6, which is a reasonable average for basic electrodes.
The use of HFM in preference of HDM is normally recommended, because it is a more ra-
tional criterion of concentration. Moreover, HDM values would be grossly unfair, if applied to
high penetration processes like submerged arc welding. In GTA welds made without filler
wire HDM cannot be used at all, since there is no deposit.
It should be noted that the present HW procedure gives the amount of 'diffusible hydrogen'.
For certain purposes the total hydrogen content may be wanted. It is obtained by adding the
content of 'residual hydrogen' determined on the same samples by vacuum or carrier gas
extraction at 6500C. A very small additional amount may be observed on vacuum fusion of the
sample, tentatively labelled 'fixed hydrogen'. There is no clear line of demarcation between
these categories of hydrogen. As will be discussed later, the extent of hydrogen trapping
depends both on the weld metal constitution and the thermal history of the metal. In single-
bead basic electrode deposits the diffusible fraction is usually well above 90%.
2.8.3 Reaction model

Normally, measurements of hydrogen in weld metals are carried out on samples from solidi-
fied beads. Due to the rapid migration of hydrogen at elevated temperatures, such data do not
represent the conditions in the hot part of the weld pool. Quenched end crater samples would
be better in this respect, but they are not representative of normal welding. Further complica-
tions arise from the presence of hydrogen in different states (e.g. diffusible or residual hydro-
gen) and the lack of consistent sampling methods.
Nevertheless, experience has shown that pick-up of hydrogen in arc welding can be inter-
preted on the basis of the simple model outlined in Fig. 2.13. According to this model, two
zones are considered:
(i) An inner zone of very high temperatures which is characterised by absorption of atomic
hydrogen from the surrounding arc atmosphere.
Electrode
Hot part of weld pool Cold part of weld pool

Absorption of atomic hydrogen Desorption of hydrogen
(controlled by pH in the arc column) (controlled by pH2 in ambient gas phase)
Hydrogen
trapped in
weld metal
Weld pool
Fig. 2.13. Idealised reaction model for hydrogen pick-up in arc welding.
(ii) An outer zone of lower temperatures where the resistance to hydrogen desorption is
sufficiently small to maintain full chemical equilibrium between the liquid weld metal and the
ambient (bulk) gas phase.
Under such conditions, the final weld metal hydrogen content should be proportional to the
square root of the initial partial pressure of diatomic hydrogen in the shielding gas, in agree-
ment with Sievert's law (equation (2-20)).
2.8.4 Comparison between measured and predicted hydrogen contents

It is evident from the data in Table 2.2 that the reported ranges for hydrogen contents in steel
weld metals are quite wide, and therefore not suitable for a direct comparison of prediction
with measurement. For such purposes, the welding conditions and consumables must be more
precisely defined.
2.8.4.1 Gas-shielded welding

In GTA and GMA welds the hydrogen content is usually too low to make a direct comparison
between theory and experiments. An exception is welding under controlled laboratory condi-
tions where the hydrogen content in the shielding gas can be varied within relatively wide
limits. The results from such experiments are summarised in Fig. 2.14, from which it is seen
that Sievert's law indeed is valid. A closer inspection of the data reveals that the weld metal
hydrogen content falls within the range calculated for chemical equilibrium at 1550 and 20000C,
depending on the applied welding current. This shows that the effective reaction temperature
is sensitive to variations in the operational conditions.
An interesting effect of oxygen on the weld metal hydrogen content has been reported by
Matsuda et al.9 Their data are reproduced in Fig. 2.15. It is evident that the hydrogen level is
significantly higher in the presence of oxygen. This is probably due to the formation of a thin
(protective) layer of slag on the top of the bead, which kinetically suppresses the desorption of
hydrogen during cooling.
Table 2.2 Measured arc atmosphere compositions in steel welding. Also included are typical ranges for
the weld metal hydrogen content. Data compiled by Christensen.4
Arc Atmosphere Composition Weld Metal

(vol%) Hydrogen
Content (ppm)
Method Primary Source
of Hydrogen CO2 CO H 2 +H 2 O Range Average
GMAW* Moisture introduced 98-80 2-20 <0.02 1-5 3

(CO2) through the shielding gas
SMAW Firmly bound water in -4 -34 -62 10-30 25
(acid) the electrode coating
SMAW Firmly bound water in ~4 -42 -54 10-30 25
(rutile) the electrode coating
SMAW Loosely bound water in -19 -77 -4 2-10 3-5
(basic) the electrode coating
FCAW Firmly bound water 10-20
(rutile) influx
FCAW Loosely bound water 2-5
(basic) influx
SAW Loosely bound water 2-10
(basic) influx
*The arc atmosphere composition can vary within wide limits, depending on the operational conditions applied.
GTAW (low-alloy steel)

ml H2/100 g fused metal, HpM
GTAW (low-alloy steel)
Welding conditions: 300A-18V-2.5 mm/s
ml H2 /100 g fused metal, HJ|S
Weld metal oxygen content, wt%

Fig. 2.15. Hydrogen pick-up in GTA welding at different levels of oxygen in the weld metal. Data from
Matsuda et al.9
Example (2.1)
Consider GTA welding (Ar-shielding) on a thick plate of low-alloy steel under the following
conditions:
/ = 200A, U = 15V, v = 3 mm s"1, TI = 0.5, T0 = 20°C
The shielding gas contains 0.1 vol% moisture (H2O) and is supplied at a rate of 15Nl mhr 1 .
Calculate the 'potential' hydrogen level, assuming that all hydrogen introduced through the
shielding gas is absorbed in the weld metal.
Solution
First we calculate the total mass of hydrogen per mm:
The resulting bead cross section and total mass of weld metal per mm can be estimated
from the Rosenthal equation by considering the dimensionless operating parameter at the melting
point (equation (1-50)):

Taking the density of the steel equal to 7.85 X 10 3 g mm 3, we obtain:
The 'potential' hydrogen level is thus:
It is evident from the above calculations that the 'potential' hydrogen level is at least one
order of magnitude higher than the expected weld metal hydrogen content (1 to 3 ppm). This
shows that the hydrogen pick-up in GTA welding is not determined by the total amount of
hydrogen which is introduced through the shielding gas, but is mainly controlled by the result-
ing partial pressure of hydrogen in the ambient (bulk) gas phase.
2.8.4.2 Covered electrodes

In SMA welding the partial pressure of hydrogen is more difficult to assess due to the presence
of trapped moisture and hydrogen-containing compounds in the electrode coating. Such com-
pounds will loose their identity at the stage of introduction into the arc atmosphere. Since very
little information is available on the species present in the arc column, we shall base our esti-
mate on a simple thermodynamic approach, including only the molecular species H2 and H2O
which can be determined by analysis (see data in Table 2.2). It follows that the combined
partial pressure of H2 and H2O in the gas phase is given by:
(2-27)
The parameter pw can be estimated on the basis of combustion measurements of the elec-
trode coating, assuming that no carbon is picked up or lost from the system in excess of the
amount calculated from an analysis of the base plate and the electrode wire. For a recorded
content of mw g H2O and mc g CO2 per 100 g of electrode coating, we obtain:
(2-28)
From a thermodynamic standpoint, replacement of pHl b y / ^ in the expression for Sievert's

law requires the use of a modified solubility of hydrogen, defined as:
(2-29)
where K3 is the equilibrium constant for the H 2 O-H reaction, and [%O] is the weld metal
oxygen content. In practice, the correction term ^ K3/(K3+[%0]) does not depart signifi-
cantly from unity, which means that Sw ~ SH.
During welding with basic covered electrodes considerable amounts of CO2 may form as a
result of decomposition of calcium carbonate, according to the reaction:
(2-30)
Modern basic electrodes contain between 20 to 40 weight percent CaCO3, which is equiva-
lent with a CO2 content of 9 to 18 percent. Taking as an average mc equal to 15 g CO2 per 100 g
electrode coating, we obtain:
(2-31)
In Fig. 2.16 the validity of equation (2-31) has been checked against relevant literature data
(compiled by Chew10). A closer inspection of the data reveals that the weld metal hydrogen
content falls within the range calculated for chemical equilibrium at 1520 to 2000°C, taking Sw
equal to the solubility of hydrogen in pure iron at the indicated temperatures (i.e. 27 and 40 ml
H2 per 100 g fused metal, respectively). Although the observed scatter in the effective reaction
temperature is admittedly large, equation (2-31) points out a very interesting effect, namely
that the hydrogen content of SMA steel weld metals is controlled by the combined partial
pressure of H2 and H2O in the ambient gas phase. For this reason it is frequently recom-
mended that calcium carbonate is added to the electrode coating, which on decomposition
produces considerable amounts of shielding gas in the form of CO2. Hydrogen shielding can
also be achieved by additions of volatile alkali-fluorides, which on heating will evaporate and
dilute the atmosphere with respect to hydrogen.
SMAW (low-alloy steel)

ml H2/100 g fused metal, HpM •
Water content in electrode coating, wt%

Fig. 2.16. Hydrogen pick-up in SMA welding at different water contents in the electrode coating. Data
compiled by Chew.10
Example (2.2)
Consider SMA welding on mild steel with basic covered electrodes. The electrode coating
contains 35 wt% CaCO3 and 0.5 wt% H2O in the as-received condition. After drying at 3500C
for 1 h the water content is reduced to 0.2 wt% H2O. Estimate the weld metal hydrogen
content (in ppm) both before and after drying of the electrode. Assume in these calculations an
effective reaction temperature of 18000C.
Solution
First we calculate the CO2 content per 100 g of electrode coating. Taking the atomic weight of
CaCO3 and CO2 equal to 100.1 and 40.0, respectively, we obtain:
The combined partial pressure pw can now be estimated from equation (2-28). Before
drying we have:
After drying of the electrode, the partial pressure pw becomes:
From Fig. 2.7(c) it is evident that the solubility of hydrogen in liquid iron at 18000C is
about 37 ml H2 per 100 g fused metal. This corresponds to a modified solubility Sw (in ppm)
of:
Substituting this value into the expression for Sievert's law gives:
(before drying)
(after drying)
It follows from the above calculations that a low weld metal hydrogen level requires the use
of 'dry' basic electrodes. In practice, this can be achieved by protecting the electrodes against
moisture pick-up during storage (see Fig. 2.17). However, in certain cases it is necessary to
differentiate between strongly bound and loosely adsorbed moisture in the coating of basic
electrodes. This point is more clearly illustrated in Fig. 2.18, which shows the HDM content of
hydrogen in basic electrode deposits at various levels of coating moisture. It is seen that water
remaining from an insufficient baking treatment is more dangerous than moisture picked up by
exposure of a properly dried coating. This has to do with the fact that loosely adsorbed mois-
Water content in electrode coating, wt%
Exposure time, days
Fig. 2.17. Moisture content in basic electrode coating as a function of exposure time and relative humid-
ity (R.H.) in ambient gas phase. Data from Evans.11

high
ml H2 /100 deposit, HDM
medium
low
very low
Water content, wt%

Fig. 2.18. Hydrogen pick-up in SMA welding at different levels (states) of adsorbed water in the elec-
trode coating. Data from Evans and Bach.12
ture will tend to evaporate during the welding operation (before it enters the arc column)
because of resistance heating of the electrode, a process which is not feasible when the water is
bound in rust on the surface of the electrode wire or the iron powder.
2.8.4.3 Submerged arc welding

This method is usually classified as a pure slag-shielded process, because carbonates or other
gas-producing compounds are not present in large quantities. A closed arc cavity does exist,
however, as indicated by the falling volt-ampere curve characteristic of open arcs, and by
observations made by probes inserted through the flux cover.
It is reasonable to assume that the gas contained within this enclosure consists of metal
vapour, volatile constituents originating from the flux, and relatively small fractions of carbon
monoxide and water vapour. Acid fluxes of the calcium silicate type will probably generate
silicon monoxide, while agglomerated fluxes bonded with alkali silicate will produce volatile
alkali fluorides. In addition, carbon monoxide may be present as a result of oxidation of car-
bon, or decomposition of carbonates.
A small but important contribution to the cavity atmosphere is the trace of moisture remain-
ing in the flux even after careful drying. No direct measurements of partial pressures are avail-
able, and the gas composition must therefore be inferred from observations of hydrogen ab-
sorption in the weld metal. Hydrogen pick-up during SA welding has been examined by Evans
and Bach.12 Their data are replotted in Fig. 2.19. The shape of the observed curve of hydrogen
vs residual water content would seem to indicate a relationship similar to that predicted by
Sievert's law. In fact, a very close fit can be obtained through empirical calibration of the
dilution term in equation (2-28). This, however, implies unreasonable amounts of CaCO3.
Carbon monoxide in addition to that delivered by carbonates could be formed by oxidation of
carbon. Again, an unreasonable amount of carbon loss would be required. Therefore, it must
be concluded that further research is needed for a proper interpretation of the factors control-
ling hydrogen pick-up in SA welding.
SAW (low-alloy steel)

Hydrogen content, HFM (ppm)
ml H2/100 deposit, HDM
Water content, wt%

Fig. 2.19. Hydrogen pick-up in SA welding at different water contents in the flux. Data from Evans and
Bach.12
Example (2.3)
Consider SA welding on a thick plate of low-alloy steel under the following conditions:
The flux contains 0.04 wt% H2O and is consumed at a rate of 0.6 g per g weld deposit.
Estimate both 'potential' and 'equilibrium' hydrogen levels when the total oxidation loss of
carbon in the weld pool is 0.03 wt%.
Solution
First we calculate the total amount of fused parent metal and weld deposit formed on welding.
From equations (1-75) and (1-120), we have:
and
When the dilution ratio DI(B + D) is known, it is possible to calculate the total flux con-
sumption per gram fused weld metal:
The 'potential' hydrogen level is thus:
If we assume that all CO produced by reactions between dissolved carbon and oxygen is
infiltrated in the arc column, the following balance is obtained:
Total number of moles of CO per g fused weld metal:
Total number of moles of H2O per g fused weld metal:

This gives:
Since the effective reaction temperature of hydrogen absorption in SA welding is not known,
the maximum solubility of hydrogen at 1 atm total pressure is taken equal to 33 ppm, similar to
that in the previous example. By inserting this value in the expression for Sievert's law, we
obtain:
In practice, the 'potential' hydrogen level represents an upper limit for the hydrogen con-
centration which cannot be exceeded. Thus, the contradictory results obtained in the present
example clearly illustrate the difficulties involved in estimating the effective partial pressure
of hydrogen in SA welding.
2.8.4.4 Implications of Sievert's law

An important implication of Sievert's law is that the fraction of hydrogen picked up from the
arc atmosphere is very high at low hydrogen pressures:
(2-32)
As seen from equation (2-32), the first traces of hydrogen added to the atmosphere are
completely absorbed in the metal. At increasing partial pressures the fraction of hydrogen
picked up in the metal will gradually decrease, finally attaining a threshold of (SH/2) in the
case of pure H2. This shows that the concept of 'potential' hydrogen content frequently used to
characterise filler materials (see Fig. 2.11) is a dangerous one, since the rates of absorption are
so different in the high and low ranges of the hydrogen potential.
2.8.4.5 Hydrogen in multi-run weldments

So far, no standardised method is available for the determination of hydrogen in multi-layer
welds. Early measurements by Roux,13 using an arrangement similar to that subsequently
adopted in French standards, indicate a constant ratio of extracted hydrogen to the mass of
fused metal, regardless of the number of passes. If hydrogen is reported on the basis of depos-
ited metal, this ratio may vary by a factor of 2.5 when comparing a deposit made in five passes
to a single bead.
Exploratory measurements of local hydrogen contents in large-size joints have been made
by Skjolberg,14 who butt welded a 40 mm plate with a self-shielding flux cored wire at an
interpass temperature of 2000C. Samples were cut from a refrigerated part of the weldment at
mid-thickness, including positions in the weld metal close to the fusion line and samples in the
HAZ. His results are summarised in Table 2.3.
Normal testing of the filler wire according to ISO 3690 gave fused metal hydrogen contents
of 3.3 ppm (diffusible) and 1.7 ppm (residual). A comparison with Table 2.3 shows that the
multi-run content of diffusible hydrogen is much lower than the corresponding ISO value,
probably as a result of a high interpass temperature which facilitates loss of hydrogen to the
surroundings through diffusion.
Table 2.3 Measured hydrogen contents in multi-run FCA steel weldment. Data from Skjolberg.14
HAZ
Distance from fusion line (mm)
Condition Weld Metal Oto 5 5 to 10 10 to 15
As-welded 0.6 ppm diffusible 0.25 ppm 0.15 ppm

0.9 ppm residual diffusible diffusible
PWHT* 0.35 ppm diffusible 0.15 ppm 0.15 ppm 0.10 ppm
(4h/150°C) 2.25 ppm residual diffusible diffusible diffusible
*Post weld heat treated.
2.8.4.6 Hydrogen in non-ferrous weldments

The solubility of hydrogen in metals and alloys of industrial importance increases with tem-
perature, and passes through a maximum in the vicinity of the boiling point, where the oppos-
ing trends of increasing solubility and increasing dilution by metal vapour balance. Solubility
curves for hydrogen in aluminium, copper, and nickel up to about 22000C have previously
been presented in Fig. 2.7.
Since all these metals can dissolve considerable amounts of hydrogen, the risk of hydrogen
absorption during welding is imminent if moisture is present in the shielding gas. Results
obtained from arc melting experiments with Cu, Al, Ni in Ar-H 2 gas atmospheres indicate that
hydrogen is absorbed at a high temperature zone under the arc and is transported by fluid flow
to the outer, cooler regions of the pool.15 Rejection of the gas in the supersaturated outer
regions is slower than the absorption in the hot zone, so the gas content throughout the pool
approximates to that in the absorption zone. Typical estimates of the effective reaction tem-
perature of hydrogen desorption (based on the Sievert's law) gave the following result:15
Copper: 16500C
Aluminium: 19000C
Nickel: 19000C
At present, it is not known whether these reaction temperatures also apply to conventional
GTA or GMA welding of the same materials or are mainly restricted to the operational condi-
tions employed in the arc melting experiments.
2.9 Absorption of Nitrogen
It is generally recognised that interstitial nitrogen embrittles steel (e.g. see discussion in Chap-
ter 7). In steel weld metals the associated loss of toughness due to free nitrogen has been
attributed to solid solution hardening and dislocation locking effects. In addition, excessive
nitrogen pick-up can cause porosity in steel weldments because of gas evolution during solidi-
fication.
2.9.7 Sources of nitrogen
Since the total nitrogen level in most welding consumables and shielding gases is quite low,
the main source of nitrogen contamination is air infiltrated in the arc column. For this reason,
the weld metal nitrogen content is very sensitive to variations in the operational conditions
(e.g. arc length, electrode stick-out, shielding gas flow rate etc.). The overall reaction of nitro-
gen absorption is similar to that of hydrogen:
(dissolved) (2-33)
By introducing the equilibrium constant K4 for the reaction, we get:
(2-34)
where SN is the maximum solubility of nitrogen at 1 atm total pressure,/^ is the activity coef-
ficient, and pNl is the resulting partial pressure of diatomic nitrogen in the gas phase.
The solubility of nitrogen in liquid iron is approximately given by:
(2-35)
where T is the temperature in K.

At 1600 and 20000C, this equation gives equilibrium concentrations of 446 and 465 ppm,
respectively. In alloyed steel containing large amounts of nitride-forming elements (e.g.
austenitic stainless steel), the activity coefficient of nitrogen fN is about 1/4 and hence, the
solubility will be about 4 times higher than that calculated from equation (2-35).
From a primitive model of pseudo-equilibrium between gaseous N2 and dissolved N a maxi-
mum solubility of about 465 ppm would be expected in welding under 1 atm total pressure.
Thus, the maximum pick-up of nitrogen in deposition of bare wire in air would be of the order
of 465A/OT8 ppm or 416 ppm. If a tentative estimate of air infiltration in the arc column is
made at 1 vol% N2, the expected pick-up of nitrogen would be 465 VoToT or about 47 ppm.
A comparison with the data in Table 2.4 shows that the measured weld metal nitrogen
contents are much higher than predicted from Sievert's law. This implies that the mechanism
of nitrogen desorption is different from that of hydrogen.
2.9.2 Gas-shielded welding

Information on the factors controlling nitrogen pick-up may be obtained from the work of
Kobayashi et a/.,16 who examined the GMA welding process in a systematic manner. Some of
their results are shown in Fig. 2.20.
Figure 2.20(a), for low-alloy steel, reveals that the square root relationship is a fair approxi-
mation only for welding in mixtures of N 2 and H 2 (curve No. 5). Mixtures of N2 + Ar (curve
No. 3), N 2 + CO2 (curve No. 2) and N2 + O 2 (curve No. 1) show increasing deviation from the
predicted behaviour. Pure N2 under reduced pressure gives a curve (No. 4) of an entirely differ-
ent shape including a maximum at pN ~ 0.05.
Table 2.4 Summary of measured weld metal nitrogen contents. Data compiled by Christensen.4
Welding Method Material Nitrogen Content (ppm)
SMAW (basic electrodes) Low-alloy steel 60-180

Stainless steel 550-650
SMAW (rutile electrodes) Low-alloy steel 200-350
Stainless steel 600-750
SAW Low-alloy steel 40-140
FCAW Low-alloy steel 125-275
GMAW Low-alloy steel 50-200
Similar features are seen from Fig. 2.20(b) for welding of stainless steel. Again, the devia-
tion becomes more pronounced as the oxidation potential of the gas mixture is increased in the
sequence H 2 -Ar-CO 2 -O 2 . Moreover, a comparison with Fig. 2.20(a) reveals that the displace-
ment of the nitrogen concentrations in the presence of chromium is larger than expected from
the calculated reduction of the nitrogen activity coefficient.
The trends observed in Fig. 2.20 have been confirmed by O'Brien and Jordan17 who studied
nitrogen pick-up during CO2-shielded welding of low-alloy steel. As can be seen from Fig.
2.21 (a) their curves are similar to those of Kobayashi et al.16 for short circuiting metal transfer,
while a mixed spray/globular transfer gives a sharp rise of nitrogen absorption up to pNi = 0.3
followed by a constant or slightly decreasing concentration (Fig. 2.21(b)). Both patterns are
clearly not in accordance with predictions based on Sievert's law (equation (2-34)).
An interpretation of the observed trends should be made with a view to absorption of hy-
drogen, where the concept of pseudo-equilibrium has proved useful for a semiquantitative
prediction. In both cases the molecular species H2 and N 2 are known to dissociate in the arc
column (see Table 2.1), and would therefore dissolve in the metal to an extent far beyond the
solubility controlled by pH or PN . The excess of dissolved hydrogen is probably released as
gas at weld pool temperatures. This will also be the case with nitrogen in the absence of oxy-
gen, as shown previously in Fig. 2.20(a) and (b). However, under oxidising conditions the
desorption of gaseous nitrogen becomes suppressed by the presence of oxygen at the gas/metal
interface and hence, nitrogen is retained at a level which by far exceeds the solubility limit at
1 atm total pressure of N2. This has been confirmed experimentally by Uda and Ohno18 in their
classic work on surface active elements (i.e. oxygen, sulphur and selenium) in liquid steel. A
similar phenomenon was quoted in Section 2.8.4.1 from the work of Matsuda et al9 even in
the case of hydrogen, where increased entrapment of hydrogen was observed in the presence
of oxygen (see Fig. 2.15).
It appears thus that excessive absorption of nitrogen (and in some cases also hydrogen)
should be interpreted as a state of incomplete release of solute, as described previously in
Sections 2.6 and 2.7. As a consequence, Sievert's law cannot be used for an estimate of nitro-
gen pick-up in steel welding, unless the weld metal oxygen content is extremely low.
2.9.3 Covered electrodes

The nitrogen content of SMA weld deposits is known to be sensitive to variations in the arc
(a)
GMAW (low-alloy steel)

Nitrogen content, wt%
Vol% N 2 in shielding gas
(b)
GMAW (stainless steel)

Fig. 2.20. Nitrogen pick-up in GMA welding at different concentrations of N2 in the shielding gas;
(a) Low-alloy steel, (b) Stainless steel. Data from Kobayashi et al.16
(a)
Low-alloy steel
Experiment

(b)
Low-alloy steel
Experiment Fig. 2.21. Nitrogen pick-up in GMA weld-

ing at different concentrations of N2 in the
shielding gas; (a) Short circuting metal
transfer, (b) Mixed and free flight metal
transfer. Data from O'Brien and Jordan.17
Vol% N2in shielding gas

length (voltage) because of the risk of air infiltration in the arc column. This point is more
clearly illustrated in Fig. 2.22, which shows that the resulting weld metal nitrogen level may
vary significantly from one weld to another, depending on the operational conditions applied.
Consequently, the use of long arcs in SMAW should be avoided in order to prevent excessive
pick-up of nitrogen from the surrounding atmosphere.
2.9.4 Submerged arc welding

In submerged arc welding the risk of air infiltration in the arc column is less imminent, since
welding is performed under the shield of a flux. Hence, in multipass welds the filler wire itself
will be the main source of nitrogen (see Fig. 2.23), while in single pass weldments the base
plate nitrogen content is more important because of the high dilution involved. The latter point
is illustrated by the following numerical example.
Example (2.4)
Consider SA (single pass) welding on a thick plate of low-alloy steel under the following
conditions:
Based on the 'rule of mixtures', calculate the weld metal nitrogen content. Assume in these
calculations that the nitrogen content of the base plate and the filler wire is 0.005 and 0.012
wt%, respectively.

4 and 5 mm basic covered electrodes
Nitrogen content, ppm
A B C D
Welder No.
Fig. 2.22. Natural fluctuations in nitrogen pick-up during SMA welding due to variations in the arc
length. Data from Morigaki et al.19
Next Page
SAW (multipass steel weldments)

Weld metal nitrogen content, ppm
loss
gain
Nitrogen content in electrode wire, ppm
Fig. 2.23. Nitrogen pick-up in SA welding at different levels of nitrogen in the electrode wire. Data from
Bhadeshia et a/.20
Solution
First we calculate the total amount of fused parent metal and weld deposit formed on welding.
From equations (1-75) and (1-120), we have:
and
The 'rule of mixtures' gives us the nominal weld metal nitrogen content, which is defined
as:
The above calculations show that the nitrogen content of single pass SA steel welds is close
to that of the base plate because of the high dilution involved. This is in agreement with gen-
eral experience.
Previous Page
2.10 Absorption of Oxygen
Partial oxidation almost invariably accompanies the welding of steel. It is well established that
considerable interaction takes place between the liquid weld metal and its surroundings (arc
atmosphere, slag) when welding is performed in the presence of oxygen. For slag-protected
processes, the flux is the main source of oxygen because of its content of easily-reduced ox-
ides, such as iron oxide, manganese oxide, silica, or rutile. In gas metal arc (GMA) welding,
oxygen is often deliberately introduced through the shielding gas to improve the arc stability
and bead morphology, but at the expense of an increased oxygen content in the weld metal and
intensified losses of alloying elements. The oxidation reactions proceed very rapidly under the
prevailing conditions owing to the high metal temperatures and the large interfacial contact
area available for interactions.
A general survey of oxygen contents in fusion welds is shown in Table 2.5.
2.10.1 Gas metal arc welding

The GMA welding process offers a special advantage in the way that it allows the reactions to
be studied through variations of the shielding gas composition without the complicating pres-
ence of a flux. Gas-shielded welding is also interesting form a practical point of view, since the
process is readily mechanised and can be applied for welding of nearly all types of steel.
Moreover, many of the reactions in the weld pool will be similar to those occurring during
welding with covered electrodes or submerged arc fluxes.
The amount of oxygen transferred to the metal during the high-temperature stage will clearly
depend on the atmosphere. In GMA welding with active gases such as CO2, Ar + O2 or Ar + O2 +
CO2 the extent of element absorption is controlled by the oxygen potential of the shielding gas.
This type of gas/metal interaction will be discussed below.
The analytical weld metal oxygen content is also affected by the presence of deoxidisers,
because the transition from the high-temperature stage to the cooling stage is not sharp. As was
Table 2.5 Summary of measured weld metal oxygen contents. Data compiled by Christensen.4
Welding Method Material Oxygen Content (ppm)
SMAW (basic electrodes) Low-alloy steel 250^40

SMAW (rutile electrodes) Low-alloy steel 500-1300
SMAW (acid electrodes) Low-alloy steel 700-1400
GTAW (pure Ar or He) Low-alloy steel 50-100
GMAW (Ar-O 2 mixtures) Low-alloy steel 200-1300*
GMAW (CO2) Low-alloy steel 500-800
FCAW (self-shielded cored wire) Low-alloy steel 130-200
SAW (calcium silicate fluxes) Low-alloy steel 450-2600
SAW (manganese silicate fluxes) Low-alloy steel 400-750
SAW (basic fluxes) Low-alloy steel 180-350
Depends on the oxygen potential of the shielding gas.
discussed briefly in the introduction to Chapter 2, a separation of non-metallic inclusions may
occur towards the end of stage one and possibly at the beginning of stage two. A necessary
condition for removal of oxygen introduced into the metal is, of course, the formation of oxy-
gen-rich phases and compounds (e.g. microslag or carbon monoxide). These matters will be
considered in a separate section on weld metal deoxidation reactions.
2.10.1.1 Sampling of metal concentrations at elevated temperatures

In order to understand the extent and direction of the oxidation reactions, it is not sufficient to
characterise the initial and final conditions (i.e. consumables/parent plate and the weld metal).
Accordingly, especially designed experiments are required for assessment of the reactions
taking place at the four main process stages:
(1) Electrode tip

(2) Arc column
(3) Hot part of the weld pool
(4) Cold part of the weld pool.
Proper sampling techniques are needed for basic studies of this kind, which allow sampling
of the falling droplets during their flight through the arc column. An additional requirement
is that the speed of quenching is sufficiently high to freeze-in the metal composition estab-
lished at elevated temperatures. If not, spontaneous reactions and subsequent losses of dis-
solved elements due to CO gas formation and manganese silicate slag precipitation may take
place on cooling.
To overcome these problems, a special 'melt spinning' technique has been developed by
Grong and Christensen,1 utilising the same principles as those employed in production of amor-
phous alloy ribbons. By using a water-cooled, fast rotating copper wheel (spinner) as a cath-
ode, rapid crystallisation of droplets released from the electrode tip can be obtained in the
absence of a weld pool*. The reactions which normally occur in the pool during regular multi-
run deposition may then be assessed by comparison of analytical data for chilled metal (i.e.
falling droplets) and normal multi-layer weldment, respectively.
2.10.1.2 Oxidation of carbon

Filler wires intended for GMAW are usually overalloyed with respect to carbon, silicon, and
manganese to compensate for heavy oxidation losses of these elements during the welding
operation. It is evident from the data presented in Fig. 2.24(a) and (b) that virtually identical
carbon contents are obtained for chilled and multilayer weld metal, except when welding is
performed in extremely oxidising atmospheres (beyond 20 vol% O2 in Ar). This implies that
oxidation of carbon does not normally take place in the weld pool, but is located to one or both
of the preceding stages, i.e. electrode tip or arc column. However, since the carbon oxidation at
the electrode tip is more than sufficient to account for the observed total loss1, it is likely that
little, if any, carbon is lost in the latter stage.
The fact that carbon is found to oxidise readily at the electrode tip but not in later stages,
indicates that the supply of oxygen and/or the conditions of nucleation are more favourable in
*The estimated cooling rate is of the order of 105 0C s"1 or higher.1

f
The measured electrode tip carbon content also includes oxidation after arc extinction. This accounts for the observed
loss of carbon beyond that recorded in chilled metal or multi-layer weld deposit.
(a)
Electrode tip
Chilled metal (falling droplet)
Multi-layer weld metal
Electrode wire
Wt% C
Vol% O 2 in Ar
(b)
Electrode tip
Chilled metal (failing droplet)
Electrode wire
wt%c
V o l % C O 2 in Ar
Fig. 2.24. Measured carbon contents in electrode tip droplets, chilled metal and multi-layer weld deposit
vs the oxygen potential of the shielding gas; (a) Ar-C>2 gas mixtures, (b) Ar-CC>2 gas mixtures. Data from
Grong and Christensen.1
the former case. Normally, homogeneous nucleation of CO gas within the liquid metal is con-
sidered impossible, which means that the CO nucleation in practice must take place heteroge-
neously. However, the most probable site for CO evolution during droplet formation at the
electrode tip will be the gas/metal interface itself, which allows carbon to be oxidised simulta-
neously with silicon and manganese. It is reasonable to assume that most of the observed
carbon oxidation is located to the hot layers facing the arc, where the reaction is thermody-
namically favoured. At other surface positions Si and Mn are expected to prevent carbon from
reacting due to a rather low metal temperature, stated to be only slightly above the melting
point at the time of detachment.21
It is evident from the data in Fig. 2.24(a) and (b) that the carbon losses increase with in-
creasing O2 or CO2 contents in the shielding gas up to a certain critical level. Hence, supply of
oxygen to the tip droplet surface is the rate controlling step for oxidation of carbon at low
oxygen potentials. This conclusion is also consistent with calculations made by Corderoy et
al.,22 who found that transport of atomic oxygen through a stagnant gaseous boundary layer
close to the metal surface controls the oxidation rate of alloying elements at this stage of the
process.
The carbon oxidation will gradually decline with increasing oxygen concentrations in the
shielding gas, probably as a result of build-up of carbon monoxide in the surrounding atmos-
phere. When the critical CO gas pressure is reached, the carbon reaction is blocked, silicon
(and manganese) now exercising control of the oxygen level, as indicated in Fig. 2.25. For Ar-
O 2 gas mixtures this critical pco pressure is attained at about 10 to 15 vol% O 2 in the shielding
gas, corresponding roughly to 0.05 wt% C oxidised in chilled metal. When welding is per-
formed in Ar-CO 2 mixtures the reaction is blocked at a much earlier stage of carbon oxidation
(equal to about 0.02 wt% C lost in chilled metal), since dissociation of CO2 in this case will
produce an additional amount of CO to concentrate in the surrounding gas phase.
atm
Silicon control
Carbon control
Pco>
Temperature, 0C
Fig. 2.25. Break even equilibrium partial pressure of CO vs temperature for silicon control of oxygen
level at 0.8 wt% Si and silica saturation. Data from Elliott et al. 23
The mechanism indicated above is supported by the data presented in Fig. 2.26, which
show that a CO2-rich atmosphere even may act carburising if the initial carbon content of the
electrode wire is sufficiently low. Moreover, it is interesting to note that the carbon-oxygen
reaction is also influenced by the rate of droplet detachment. Since the highest carbon oxida-
tion losses are normally associated with a coarse globular droplet transfer mode, this suggests
that the reaction time is more important than the effective contact area available for interaction
which depends on the droplet size.
2.10.1.3 Oxidation of silicon

It is evident from the data in Fig. 2.27 that loss of silicon mainly take place in the weld pool, as
indicated by the difference between the measured silicon content in chilled and multi-layer
weld metal. In Ar-O 2 mixtures the Si loss increases steadily with increasing oxygen potential
of the shielding gas. Thus, at 30 vol% O2 in Ar it amounts to 0.59 wt% Si (or 0.67 wt% O)
removed from the weld pool as a result of deoxidation reactions. A similar situation exists in
the case of CO2-shielded welding up to about 20 vol% CO2 in Ar. At higher CO2 contents, the
Si loss tends to drop off, finally attaining an upper limit of approximately 0.30 wt% Si corre-
sponding to 0.34 wt% O removed from the weld pool.
In comparison, the amount of Si lost in the two preceding stages (i.e. electrode tip and arc
column) is much smaller, as shown by the data for the chilled metal Si content. Since no slag is
formed under the conditions of rapid cooling, silicon must escape in the form of a gaseous

Weld metal carbon content, wt%
(Gain)
(Loss)
Filler wire carbon content, wt%

Fig. 2.26. Correlation between filler wire and weld metal carbon contents in CO2-shielded welding.
Data from Ref.24.
(a)

Electrode wire
Wt% Si
Vol% O 2 in Ar
(b)
Electrode wire
Wt% Si
Vol% CO 2 in Ar
Fig. 2.27. Measured silicon contents in chilled and multi-layer weld metals vs the oxygen potential of the
shielding gas; (a) Ar-02 gas mixtures, (b) Ar-CC>2 gas mixtures. Data from Grong and Christensen.1
product. Evaporation losses can in this case be excluded due to a very low vapour pressure of
silicon at the prevailing temperatures. It is therefore reasonable to assume that SiO(g) forms as
a result of chemical reactions occurring at the electrode tip. The observed decrease in chilled
metal Si content with increasing O2 and CO2 contents in the shielding gas is probably caused
by the presence of CO at the gas/metal interface, which facilitates SiO formation according to
the reaction:
(slag) (gas) (gas) (gas) (2-36)
Figure 2.28 shows a plot of the equilibrium partial pressure of SiO vs temperature at silica
saturation for three different CO levels. Note that stoichiometric amounts of CO2 have been
assumed to form in order to calculate psi0. It is evident from the thermodynamical data pre-
sented in Fig. 2.28 that the formation of SiO is strongly dependent on the CO partial pressure
at the slag/metal interface. Thus, from a thermodynamic standpoint the silicon loss at the elec-
trode tip should be most pronounced during CO2-shielded welding due to the resulting higher
CO pressure. The data shown in Fig. 2.27(a) and (b) support this assumption. It is seen that the
silicon loss is increased by a factor of 3 to 5 in presence of CO2 when comparison is made on
the basis of equal oxygen potential of the shielding gas (i.e. equal loss of deoxidants in the
weld pool). Similar observations have also been made by Heile and Hill27 from determination
of silicon in collected GMA welding fumes. The recorded chilled metal Si loss in Fig. 2.27(a)
and (b) is in good agreement with the reported fume formation rates of silicon. Since all CO
consumed in reaction (2.36) is expected to be regenerated immediately by decomposition of
CO2 at the metal surface, the high CO partial pressure required for SiO formation at the elec-
trode tip is maintained even in the case of extensive silicon losses. Consequently, the net reac-
tion can be written as:
PSiO,10-3atm
Temperature, 0C
Fig. 2.28. Equilibrium partial pressure of SiO vs temperature at different CO levels. Data from Refs. 25
and 26.
(slag) (dissolved) (gas) (2-37)
When welding is performed in Ar-O 2 gas mixtures, pSiO andpco may be taken proportional
to the recorded loss of silicon and carbon in chilled metal (see data in Fig. 2.29). It is evident
from this plot that the silicon loss is directly proportional to the corresponding loss of carbon
up to a certain critical level. Thus, during the initial period of carbon oxidation at the electrode
tip the SiO formation is probably controlled by the resulting partial pressure of CO at the slag/
metal interface, according to reaction (2-36). When the carbon reaction is blocked, the chilled
metal silicon loss becomes independent of the CO partial pressure (see reaction (2-37)), since
all CO consumed in the SiO formation will immediately be recirculated within the system.
However, at the break even point for silicon control of the oxygen level at the electrode tip, the
CO pressure in the surrounding gas phase will be the same for both Ar-O 2 and Ar-CO 2 gas
mixtures. Hence, the recorded loss of silicon in chilled metal at 20 vol% CO2 in Ar is seen to be
similar to that in Ar + 10 vol% O2, as indicated by the heavy broken line in Fig. 2.29.
At the temperatures where liquid steel is normally deoxidised, silicon and manganese have
a strong affinity to oxygen. Their ability to form stable oxides decreases rapidly with increas-
ing temperature, and above approximately 180O0C silicon and manganese do no longer act as
efficient deoxidation agents. Precipitation of manganese silicate slags is therefore favoured by
the lower metal temperatures prevailing at the electrode tip and in the cold part of the weld
pool. At higher temperatures, these oxides become unstable. Consequently, as a result of the
metal superheating occurring during droplet transfer through the arc column, the macroscopic
slag phase formed earlier at the electrode tip surface (as reported by Corderoy et ah12) will
redissolve in the metal. This gives rise to a relatively high chilled metal oxygen content (to be
discussed below).
Ar+20 vol% CO2

Loss of silicon (%)
Ar+10vol%O2
Loss of carbon (%)

Fig. 2.29. Correlation between loss of carbon and silicon in chilled metal at different O 2 levels in the
shielding gas. The corresponding loss of silicon at 20 vol% CO2 in Ar is indicated by the heavy broken
line in the graph. Data from Grong and Christensen.1
Example (2.5)
Consider CO2-shielded welding on a thin sheet of low-alloy steel with a 0.8mm dia. electrode
wire under the following conditions:
Based on the data presented in Fig. 2.27(b), calculate the fume formation rate (FFR) of
silicon (in mg per min) due to SiO formation at the electrode tip. The wire feed rate is 125mm s"1.
Solution
The total loss of silicon due to SiO formation may be taken equal to the observed difference
between the filler wire and the chilled metal silicon contents. For welding in pure CO2, we get:
The corresponding fume formation rate of silicon (in mg min"1) can readily be calculated
when the wire feed rate (WFR) is known. Taking the density of the steel equal to 7.85mg mm"3,
we obtain:
A comparison with the measured FFR of silicon in Table 2.6 (at / = 13OA) shows that the
calculated value is reasonable correct. Moreover, these data support our previous conclusion
that the SiO formation is favoured by a high CO2 content in the shielding gas due to the
dissociation reaction. In fact, more detailed studies of the reaction kinetics have confirmed that
the rate of SiO formation is proportional to the resulting partial pressure of CO at the gas/metal
interface,28 in agreement with equation (2-36).
2.10.1.4 Evaporation of manganese

It is seen from the data in Fig. 2.30(a) and (b) that the amount of manganese lost in chilled
metal is virtually independent of the oxygen potential of the shielding gas, as indicated by the
constant difference of about 0.35 wt% between the filler wire and the chilled metal Mn con-
tents in both graphs. This implies that significant amounts of manganese are lost during
droplet transfer through the arc column as a result of evaporation.
At the prevailing temperatures, the vapour pressure of iron is also high due to the almost
unity activity of Fe. If the average arc metal temperature is taken equal to about 24000C,29 the
data in Fig. 2.31 indicate that the vapour pressures of iron and manganese (at 1.27 wt% Mn)
are nearly identical and close to 0.05 atm. In the hot surface layers of liquid metal facing the
arc the temperature will be even higher, which means that iron vapour will dominate.
Measurements of collected GMA welding fume reported by Heile and Hill27 (see data in
Table 2.6) show a substantial higher loss of iron than that derived from simple thermodynami-
cal calculations taking the rate of element loss proportional to the vapour pressure. From their
results a reasonable value of the average mass ratio Fe to Mn in dust is about 5. Consequently,
as a preliminary estimate the loss of iron may be taken 5 times the amount of manganese lost
Table 2.6 Measured fume formation rates in GMA welding of ferrous materials. Data from Heile and
Hill.27
Fume Formation Rate

Shielding Gas Current Voltage (mg min"1)
(A) (V) Mn Si Fe
250 29 1 1 22
Argon 300 31 1 0 12
350 35 5 4 51
150 28 15 8 134
Ar+ 2 % O 2 200 28 12 10 75
300 29 5 4 35
400 34 18 16 86
100 28 29 33 273
Ar+ 5% O 2 200 28 16 22 129
300 28 10 23 76
100 23 11 20 75
Ar + 25% CO2 150 27 13 35 105
300 35 29 62 191
130 27 13 59 86
150 30 14 63 120
Pure CO2 200 30 19 73 126
250 30 31 112 216
300 30 36 125 214
in chilled metal, corresponding to a total loss of 1.7 wt% (Fe + Mn) or 500mg (Fe + Mn) per
min. The reported fume formation rate of Fe and Mn under nearly similar conditions is only
one half of that stated above, which indicates that the calculated flux of metal vapour may be
somewhat overestimated. However, if these two values are considered to represent borderline
cases, the volume of metal vapour is in the range from 300 to 500 times that of the droplets
and will therefore be more than sufficient to protect the metal from arc atmosphere oxidation
at this stage of the process. This conclusion is consistent with statements made by Distin et al?1
who claim that iron vapour acts as an effective oxygen getter already at about 19000C.
Very large amounts of manganese are also lost in the weld pool stage, as shown by the
difference between the measured Mn contents in chilled and multi-layer weld metals. This
situation appears to be quite similar to that of silicon. For Ar-O 2 mixtures the manganese loss
increases steadily with increasing oxygen contents in the shielding gas, and reaches a value of
about 1.08 wt% Mn (or 0.31 wt% O) removed from the weld pool at 30 vol% O2 in Ar. In the
case OfAr-CO2 shielded welding the Mn oxidation loss starts to drop off at a CO2 content of
about 20 vol%, finally attaining an upper limit of about 0.50 wt% Mn corresponding to 0.15
wt% O removed from the weld pool.
A more detailed discussion of oxidation reactions in GMA welding is given in Section
2.10.1.5.
Example (2.6)
Consider GMA welding low-alloy of steel under conditions similar to those in Example 2.5.
(a)
Chilled metal
Wt% Mn
(falling droplet)

Electrode wire
Vol% O 2 in Ar
(b)
Wt% Mn

Electrode wire
Vol%CO 2 inAr
Fig. 2.30. Measured manganese contents in chilled and multi-layer weld metals vs the oxygen potential
of the shielding gas; (a) Ar-02 gas mixtures, (b) Ar-CO2 gas mixtures. Data from Grong and Christensen.1
Based on the data presented in Fig. 2.30(b), calculate the fume formation rate (FFR) of manga-
nese due to evaporation losses occurring during droplet transfer through the arc column. Esti-
mate also the effective mass transfer coefficient for manganese evaporation under the prevail-
ing circumstances by utilising the vapour pressure data in Fig. 2.31. The surface temperature
of the falling droplets is assumed constant and equal to 26000C.
LogpMn,atm
PMr/PFe
Temperature, 0C
Fig. 2.31. Equilibrium manganese vapour pressure and corresponding vapour pressure ratio pMn to PFe vs
temperature at 1.27 wt% Mn in iron. Data from Kubaschewski and Alcock. 30
Solution
The total loss of manganese due to evaporation may be taken equal to the observed difference
between the filler wire and the chilled metal manganese contents. For welding in pure CO2, we
have:
The corresponding fume formation rate of manganese (in mg min"1) can readily be calcu-
lated when the wire feed rate (WFR) is known. Taking the density of the steel equal to 7.85mg
mm~3, we obtain:
A comparison with the data in Table 2.6 reveals that the reported fume formation rate of
manganese (at / = 13OA) is much lower than computed in the present example. This discrep-
ancy can probably be attributed to differences in the filler wire manganese content. Assuming
that the evaporation loss of manganese is controlled by a transport mechanism in the gas phase,
we can estimate the effective mass transfer coefficient from equation (2-15) by inserting a
reasonable average value for the manganese vapour pressure at the gas/metal interface (pMn in
the bulk phase is taken equal to zero). Reading from Fig. 2.31 (at T= 26000C) gives:
If the diameter of the falling droplets is taken equal to the diameter of the filler wire, it is
possible to calculate the total loss of manganese associated with one droplet (in mol):
Since the average flight time of large, globular droplets through the arc column is of the
order of one second, the corresponding flux of manganese vapour per unit time is close to 9.97
X 10~8 mol s"1. Thus, by rearranging equation (2-15), we obtain the following value for the
effective mass transfer coefficient:
Although the above value is rather uncertain, the calculated mass transfer coefficient is of
the expected order of magnitude. This supports our previous conclusion that the evaporation
kinetics are controlled by a transport mechanism in the gas phase.
2.10.1.5 Transient concentrations of oxygen

It is evident from the data summarised in Fig. 2.32 that the oxygen content in both chilled and
multi-layer weld metal increases with increasing oxygen potential of the shielding gas. The
chilled metal analysis is representative of the oxygen absorption occurring at the electrode tip
due to the lack of gas/metal interaction in the arc column. Measurements of manganese oxida-
tion losses during droplet formation have been performed by Corderoy et al.22 From the curve
presented for 20ms tip melting cycles, a reasonable estimate for the manganese oxidation loss
in Ar-O 2 shielding gas mixtures may be:
(2-38)
Taking the Si to Mn mass ratio in precipitated slag equal to 0.66,22 the corresponding loss of
silicon is equal to:
(2-39)
Based on equations (2-38) and (2-39) it is possible to calculate the oxygen absorption at the
electrode tip which is associated with the MnO and the SiO2 slag formation. If corrections also
are made for the amount of oxygen simultaneously removed as iron oxide*, we obtain:
(2-40)
*An approximate correction can be made from an analysis of the Fe to Mn mass ratio in precipitated slag, which under
the prevailing circumstances is close to 0.16.22
(a)
Wt% O Electrode wire
Vol%O2inAr
(b)

Electrode wire
Wt% O
Vol% CO2 in Ar
Fig. 2.32. Measured oxygen contents in chilled and multi-layer weld metals vs the oxygen potential of
the shielding gas; (a) Ar-C>2 gas mixtures, (b) Ar-CO2 gas mixtures. Data from Grong and Christensen.1
At 30 vol% O2 in Ar the absorption of oxygen is roughly 0.27 wt% O, which is reasonably

close to that recorded in the chilled metal analysis (about 0.23 wt% O). The rate controlling
step for metal oxidation at this stage of the process is believed to be transport of atomic oxygen
through a stagnant gaseous boundary layer of Ar and/or CO adjacent to the metal surface.22
On the other hand, the measured chilled metal oxygen contents are much too low to account
for the heavy oxidation losses of deoxidants observed in multi-layer welds. This indicates that
considerable amounts of oxygen are introduced in the hot part of the weld pool immediately
beneath the root of the arc. Although the gas/metal interfacial contact area available for reac-
tion is much smaller than that of the electrode tip (or falling) droplets, the strong turbulence
existing in the hot part of the pool will provide an effective circulation of the liquid metal
through the reaction zone.32 Moreover, absorption of oxygen at this stage of the process will be
favoured by the increased time available for gas/metal interaction. If the diameter of the reac-
tion zone is taken equal to about 3mm, the time available for oxygen absorption in the weld
pool is of the order of one second for a typical welding speed of 3mm s"1. In comparison, the
corresponding reaction time during droplet formation at the electrode tip is only 20 to 50 ms.
Calculations of the total oxygen absorption can be done on the basis of the measured differ-
ence between silicon and manganese contents in chilled and multi-layer weld metals, desig-
nated A[%S7] and A[%Mn], respectively*. Moreover, corrections should be made for the amount
of oxygen simultaneously removed as iron oxide from the weld pool. For deoxidation with
silicon and manganese the mass ratio wt% Fe to wt% Mn in precipitated slag is equal to about
O.I.1 This leads to the following balance:
(2-41)
At 30 vol% O2 in Ar the total oxygen absorption amounts to:
It is seen from the graphical representation of equation (2-41) in Fig. 2.33(a) and (b) that
most of the oxygen pick-up takes place in the weld pool. However, at present it is not clear
whether the calculated values represent a real transient concentration in the hot part of the
weld pool or a number in concentration units representing precipitation of manganese silicate
slags. According to Fischer and Schumacher33 the solubility of oxygen in liquid iron is 0.94
wt% at 19000C and 1.48 wt% at 20000C This temperature range is probably relevant with re-
spect to the hot part of the weld pool, although the surface metal temperature will be even
higher.34 However, in steel weld metals the oxygen concentrations should be well below this
solubility limit in the presence of silicon and manganese. Taking the average weld pool silicon
content equal to 0.50 wt%, the corresponding oxygen content in equilibrium with a silica-
saturated slag is roughly 0.10 and 0.20 wt% at 1900 and 20000C, respectively.23 On the other
hand, under the prevailing circumstances there will be no solid/liquid interface available for
nucleation of oxide particles, and hence homogeneous nucleation is the only possibility.
According to Sigworth and Elliott35 this requires a relatively high degree of supersaturation,
which means that the oxygen concentration most likely will exceed the Si-Mn-O equilibrium
value before slag precipitation occurs.
The data in Fig. 2.33(a) and (b) indicate that the weld pool oxygen absorption is controlled
by a complex transport mechanism in the gas phase. However, since mass transfer in gas-jet/
liquid systems is not fully understood,3 we shall only consider the limiting case where the
resistance to mass transfer is confined to a stagnant gaseous boundary layer adjacent to the
metal surface. Under such conditions equation (2-10) predicts that the transient oxygen con-
centration is determined by the partial pressure of atomic oxygen in the plasma gas, as shown
*Oxygen consumed in CO and SiO formation is not included.

(a)
Calculated total oxygen
absorption
Analytical weld metal
oxygen content
Wt% O
Vol% O2 in Ar
(b)
Calculated total oxygen
absorption
Analytical weld metal
oxygen content
Wt% O
Vol% CO2 in Ar
Fig. 2.33. Calculated total oxygen absorption in GMA welding at different oxygen potentials of the
shielding gas; (a) Ar-O 2 gas mixtures, (b) Ar-CO 2 gas mixtures. The analytical weld metal oxygen
content is indicated by the broken lines in the graphs. Data from Grong and Christensen.1
by the plots in Fig. 2.34. If we as a borderline case assume that all oxygen present in the plasma
gas is immediately absorbed in the liquid metal, the slope of the curve in Fig. 2.34 for Ar-O 2
gas mixtures (equal to about 2.67 wt% dissolved oxygen at one atmosphere total pressure of
atomic oxygen) is representative of the total amount of gas passing through the arc column.
Taking the mass of liquid metal leaving/entering the reaction zone equal to 3Og min"1 under
the prevailing circumstances,1 the following gas flow rate is obtained (in Nl min"1).
This value corresponds to about 6 per cent of the total shielding gas flow rate.
In contrast to the situation described in Fig. 2.34 for Ar-O 2 gas mixtures, a large deviation
from the expected relationship is observed for welding in CO2-rich atmospheres. At present,
the reason for this shift in the reaction kinetics is not known. However, it is evident from the
data presented in Fig. 2.34 that the effective mass transfer coefficient decreases by a factor of
about 2.5 when the CO2 content in the shielding gas increases from 10 to 100 vol%. This
implies that the oxidation capacity of pure CO2 is comparable with that of Ar + 13 vol% O2,
although the partial pressure of atomic oxygen in the plasma gas is equivalent with an oxygen
content of about 33 vol%.
Example (2.7)
Consider GMA welding of low-alloy steel in Ar + 20 vol% CO2 and pure CO2, respectively.
Based on the results in Figs. 2.27(b) and 2.30(b), calculate the total weight of top bead slag (in
gram per 100 gram weld deposit) which forms as a result of deoxidation reactions. Assume in
these calculations that the iron to manganese mass ratio in precipitated slag is equal to 0.1.
[%O]t0t,wt%
Pure CO2
P0, atm
Fig. 2.34. Calculated total oxygen absorption in GMA welding at different partial pressures of atomic
oxygen in the plasma gas. Data from Grong and Christensen.1
Solution
The amount of silicon and manganese lost as a result of deoxidation reactions is equal to the
observed difference in the chilled and multi-layer weld metal Si and Mn contents. Assuming
that these elements are removed from the weld pool as SiO2 and MnO, respectively, the fol-
lowing balance is obtained:
Thus, for welding in Ar + 20 vol% CO2, the total weight of slag amounts to:
Similarly, for CO2-shielded welding, we get:
A comparison with the experimental data in Fig. 2.35 shows that there is a fair agreement
between the amount of slag recorded by weighing and that calculated from a simple mass
balance of silicon, manganese and iron.
Calculated weight of slag

Measured weight of slag
g slag/10Og deposit
Vol% CO2 in Ar
Fig. 2.35. Comparison between measured and calculated weight of top bead slag in CO2-shielded weld-
ing. Data from Grong and Christensen.1
2.10.1.6 Classification of shielding gases
The data in Fig. 2.34 provide a basis for evaluating the oxidation capacity of various shielding
gases. For welding in Ar + O2 and Ar+CO2 gas mixtures up to 10 vol% CO2 in argon, the total
oxygen absorption is approximately given by the following equation:
(2-42)
Similarly, for welding with CO2-rich shielding gases (i.e. between 10 and 100 vol% CO2),
we obtain:
(2-43)
Equal oxidation capacity means that the total weld metal oxygen absorption is the same for
both shielding gas mixtures. Hence, we may write:
(2-44)
when the CO2 content in the shielding gas is less than 10 vol%, and
(2-45)
when welding is performed with CO2-rich shielding gases (more than 10 vol% CO2 in Ar).
Based on equations (2-44) and (2-45) it is possible to compare the oxidation capacity of
various shielding gases (see Table 2.7). Included in Table 2.7 is also a slightly modified ver-
sion of the International Institute of Welding (HW) classification system,36 which is based on
an evaluation of retained (analytical) oxygen in the weld deposit. It is evident from these data
that both systems are applicable and mutual consistent, although the former one utilises a more
rational criterion for the shielding gas oxidation capacity.
2.10.1.7 Overall oxygen balance

In GMA welding with solid wires, the CO content in the exhaust gas provides a direct measure
of the extent of gas/metal interaction. This CO content should be compatible with that calcu-
lated from an overall oxygen balance for the reaction system.37
Example (2.8)
Consider GMA welding of low-alloy steel under the following conditions:
Table 2.7 Proposed shielding gas classification scheme for GMA welding of low-alloy steel according to
equations (2-44) and (2-45). Included is also a modified version of the corresponding IIW's classification
system.36
Shielding Gas Composition [%O] tot

Vol%CO2 Vol%O 2 Vol%Ar (wt%) IIW's Terminology*
0-2 balance Lightly oxidising

0-4 - balance <0.10 (<0.02)
0-2 0-1 balance
2-5 balance Oxidising
4-10 - balance 0.1-0.25 (0.025-0.035)
2-5 1-2V2 balance
5-8 balance Oxidising
10-25 - balance 0.25-0.40 (0.035-0.045)
- 8-13 balance Strongly oxidising
25-100 - balance 0.40-0.60 (0.045-0.07)
> 13 balance >0.60 Extremely oxidising
(> 0.07)
The analytical weld metal oxygen content (in wt%) is given by the values in brackets.
The shielding gas is pure CO2 and is supplied at a constant rate of 15Nl min *. Based on the
composition data in Table 2.8 calculate the resulting CO content in the welding exhaust gas.
Solution
First we calculate the nominal weld metal chemical composition by neglecting oxidation loss
of alloying elements due to chemical reactions:
The dilution ratios BI(B + D) and DI(B + D) can be estimated from the classic heat flow
theory presented in Chapter 1. From equations (1-75) and (1-120), we have:
Table 2.8 Chemical composition of filler wire, base plate and weld metal used in Example 2.8.
C O Si Mn
Element (wt%) (wt%) (wt%) (wt%)
Filler wire 0.10 0.01 0.93 1.52

Baseplate 0.14 0.007 0.40 1.30
Weld metal 0.09 0.065 0.35 0.81
and
This gives:
The extent of gas/metal interaction can then be evaluated from the observed concentration
displacements:
Calculated values for the concentration displacements of carbon, oxygen, silicon and man-
ganese utilising the composition data in Table 2.8 are given below.
Element C O Si Mn
\-%X\nom. 0
- 1 2 0
- 0 0 8 0
' 6 3 L 3 9
[A%X] -0.03 0.057 -0.28 -0.58
The total CO evolution (in mol min l) can now be computed from an overall oxygen bal-
ance for the reaction system. In these calculations we shall assume that Si and Mn lost as
SiO(g) and Mn(vap.) immediately react with CO2 to form SiO2 and MnO, respectively*. Tak-
ing the density of steel equal to 7.85 X 10~3 g mm"3, the total mass of weld metal produced per
unit time amounts to:
Overall oxygen balance (i.e. consumption of CO2)
Oxidation of carbon:
Oxidation of silicon:
:
CO 2 consumed in oxidation of iron vapour is disregarded.
Oxidation of manganese:
Increase in oxygen content:
Total CO evolution (sum):(13.4 + 53.5 + 28.3 + 9.6) X 10~3 mol CO min - 1 = 104.8 X 10~3 mol
CO min' 1
Based on this information it is possible to calculate the resulting CO content in the welding
exhaust gas:
A comparison with the data in Table 2.2 shows that a CO content of about 15 vol% is
reasonably close to that determined by analysis.
2.10.1.8 Effects of welding parameters

So far, gas/metal interactions in GMA welding has mainly been discussed in terms of the
oxygen potential of the shielding gas. In the following, some consideration will be given to the
effects of welding parameters on the weld metal chemistry.
Amperage
When the welding current is raised, the time available for interaction decreases due to the more
rapid detachment of the electrode tip droplets. At the same time the interfacial contact area
increases as the average droplet size becomes smaller. From measurements of fume formation
rates in GMA welding,27 it has been shown that these two counteracting effects will almost
cancel, i.e. the total amount of emitted dust (in mg per g deposit) is found to be constant and
nearly independent of the applied amperage. On the other hand, the total fume formation rate
is probably not a reliable index for the burn-off of Si and Mn, since the evolution of iron
vapour during droplet transfer will tend to conceal the corresponding loss of alloying ele-
ments.
The effect of amperage (or more correct the droplet detachment frequency) on the burn-off
of carbon, silicon, and manganese in CO2-shielded welding has been investigated by Smith et
al.38 They found that the recovery of alloying elements in the weld deposit increased with
increasing welding current (i.e. droplet detachment frequency). In view of the previous discus-
sion, it is reasonable to assume that the higher weld metal carbon and silicon contents reported
by Smith et al3S are a result of a reduced CO and SiO gas evolution at the electrode tip due to
the shorter time available for chemical interaction. In the case of manganese reduced evapora-
tion losses because of a more rapid transfer of the droplets through the arc column offers a
reasonable explanation to the increased element recovery. This shows that the weld metal
chemistry is sensitive to variations in the welding current.
Arc voltage
Since the arc voltage neither affects the melting rate nor the droplet size to any great extent,39
variations in the arc voltage should only have a minor effect on the weld metal chemistry. This
conclusion is apparently in conflict with observations made by Lindborg,40 who found that the
oxidation reactions in GMA welding were strongly voltage dependent and at the same time
independent of the welding current, the droplet detachment frequency, and the mode of metal
transfer (spray or short-circuiting). Consequently, further investigations are required to ex-
plain these discrepancies.
Welding speed
It can be inferred from the data of Grong and Christensen1 that the analytical weld metal
carbon and oxygen contents are virtually independent of the welding speed v within the nor-
mal range of GMAW (i.e. from 0.4 to 6 mm s"1). However, the intensified losses of silicon and
manganese observed at low welding speeds indicate that more oxygen is absorbed in the weld
pool under such conditions. This point is more clearly illustrated in Fig. 2.36 which shows a
plot of [%O]tot vs v for a series of multi-pass GMA welds deposited under the shield of Ar +
10 vol% O2. It is evident that the total oxygen absorption increases nearly by a factor of two
when the welding speed decreases from 6 to 0.4mm s"1. This shows that the welding speed has
a marked effect on the transient oxygen pick-up in the hot part of the weld pool during GMA
welding, since it controls the time available for element absorption.
2.10.2 Submerged arc welding

In flux-shielded processes the reaction pattern is much more difficult to assess because of the
Calculated total oxygen absorption

[%O]tot,wt%
Travel speed, mm/s

Fig. 2.36. Calculated total oxygen absorption in GMA welding at different travel speeds. Data from
Grong and Christensen.1
complicating presence of the slag. For this reason most investigators have chosen to analyse
empirically slag/metal reactions in SA welding. Nevertheless, some authors have been able to
interpret their results on more theoretical grounds in spite of the complex reaction systems
involved.41"*5 Unfortunately, these thermodynamic approaches give, at best, only a qualitative
description of the compositional changes occurring during the welding operation.
Recently, a kinetic model has been developed by Mitra and Eagar46 to account for varia-
tions in the element recovery in both single-pass and multi-pass SA steel weldments. From
their work it is evident that the transfer of alloying elements between the slag phase and the
weld metal cannot be adequately described by means of a primitive model of pseudo-equilib-
rium without including a more detailed analysis of the reaction kinetics. This shows that the
conditions existing in SA welding are quite similar to those prevailing during GMA welding,
although the experimental and theoretical challenges are much greater in the former case due
to the complicating presence of a macroscopic slag phase.
2.10.2.1 Flux basicity index

During SA welding of steel, oxygen may be transferred from the slag to the weld metal due to
decomposition of easily reduced oxides at elevated temperatures according to the overall reac-
tion:
MxOy = xM (dissolved) + y O (dissolved) (2-46)
where MxOy denotes any oxide component in the slag phase (e.g. SiO2, MnO or FeO).
The basicity index (B.I.), originally adopted from steel ladle refining practice, is most fre-
quently employed for assessment of oxygen pick-up in SA welding, since it gives an approxi-
mate measure of the flux oxidation capacity. A number of different expressions exists in the
literature, but for the purpose of convenience the basicity index defined by Eagar47 has been
adopted here:
basic oxides"
non- basic oxides
(2-47)
where the concentration of each flux component is given in weight percent.

It is evident from Fig. 2.37, which shows a typical correlation between the weld metal
oxygen content and B.L, that the oxygen level of welds produced under acid fluxes (i.e. low
B.I.) is strongly dependent on the basicity index. In contrast, the oxygen concentrations of
welds deposited under basic fluxes are seen to be essentially independent of B.I., as indicated
by the horizontal part of the curve in Fig. 2.37. It should be noted that this analysis gives no
information about the extent of slag/metal interaction, since it is based on data for retained
oxygen in the weld deposit. Consequently, because of the empirical nature and limited applica-
bility of the basicity index, its role in the choice of welding fluxes for SA welding is a keenly
debated question.
Basic fluxes
Oxygen content, wt%
Acid fluxes
Flux basicity index

Fig. 2.37. Correlation between retained oxygen and flux basicity in SA welding. Data from Eagar.47
2.10.2.2 Transient oxygen concentrations

In SA welding of C-Mn steels, the transient flux of oxygen passing through the weld pool can
be estimated from the observed concentration displacements of silicon and manganese, which
may be taken equal to the difference between absorbed and rejected Si and Mn, respectively:
(2-48)
and
(2-49)
If we assume that rejection of Si and Mn in the weld pool occurs as a result of MnSiO3
microslag precipitation and subsequent phase separation, [%Mn]rej^ is bound to [%Si]rej. through
the following stoichiometric relationship:
(2-50)
Taking the ratio between absorbed Mn and Si in the weld metal equal to k, a combination of
equations (2-48), (2-49) and (2-50) gives:
(2-51)
The value of k is difficult to evaluate in practice, but in view of the reported mass transfer
coefficients for Mn and Si a reasonable estimate would be about 0.5 in the case of manganese
silicate fluxes.46 Under such conditions the total oxygen absorption, [%O]abs is given by:
(2-52)
where [A%O] is the observed concentration displacement of oxygen in the weld metal, and
[%O]rej. is the amount of oxygen rejected from the weld pool as a result of deoxidation reac-
tions.
Based on equation (2-52) it is possible to estimate the total oxygen absorption during SA
welding of C-Mn steels from an analysis of measured concentration displacements of oxygen,
silicon, and manganese in the weld metal. The results of such calculations are shown graphi-
cally in Fig. 2.38, using data from Indacochea et a/.44 It is evident from this plot that the total
oxygen absorption during SA welding is much larger than that inferred from an analysis of
retained oxygen in the weld deposit. The situation is thus quite similar to that observed experi-
mentally in GMA welding (see Fig. 2.33).
It should be noted that the calculated values for [%O]abs. in Fig. 2.38 may be encumbered
by systematic errors due to the number of simplifying assumptions inherent in equation (2-52).
However, this does not affect our main conclusion regarding the significance of the oxygen
absorption, since more refined calculations give a pattern similar to that observed above (see
Fig. 2.39).
2.10.3 Covered electrodes

Chemical reactions during SMA welding have been studied by several investigators in the
Experimental MnO-FeO-SiO2 fluxes

Oxygen content, wt%
Retained oxygen
[A% Si]0.5 - [A%Mn]

Fig. 2.38. Calculated total oxygen absorption in SA welding with experimental MnO-FeO-SiO2 fluxes.
Data from Indacochea et al.AA
Flux Type CS: Bead on Plate
» — : Two Wires
Filled Symbols: Flux FB
Oxygen content, wt%
Absorbed
oxygen
Retained
oxygen
Silicon content, wt%

Fig. 2.39. Calculated total oxygen absorption in SA welding with commercial calcium silicate (CS) and
fluoride-basic (FB) fluxes. Data from Christensen and Grong.45
past.47"51 Most of these investigators have interpreted their results as a high-temperature equi-
librium between the slag and the weld metal, but a verified quantitative understanding of the
transfer of elements during welding is lacking. This situation arises mainly from the lack of
adequate thermodynamic data for the complex slag/metal systems involved.
2.10.3.1 Reaction model

The reaction model presented here is restricted to welding with basic covered electrodes.
During SMA welding gases are generated by decomposition of compounds present in the
electrode coating. In the case of basic covered electrodes, the decomposition of limestone
results in an atmosphere consisting predominantly of carbon monoxide and carbon dioxide,
containing only small amounts OfH2 and H2O (see data in Table 2.2). The characteristic high
concentrations of CO and CO2 in the arc atmosphere would be expected to lead to extensive
absorption of carbon and oxygen in the weld metal. Under the prevailing circumstances, it is
reasonable to assume that these reactions approach a state of local pseudo-equilibrium during
droplet transfer through the arc column.
During the subsequent stage of cooling in the weld pool, a supersaturation with respect to
the various deoxidation reactions is initially increasing, which is released when the conditions
for nucleation of the respective reaction products are reached. Since carbon is a much stronger
deoxidant than silicon and manganese at temperatures above about 17000C,23 it is reasonable
to assume that carbon will be in control of the oxygen level during the initial stage of cooling*,
in accordance with the reaction:
C (dissolved) + O (dissolved) = CO(gas) (2-53)

* Although gases such as CO and CO2 are widely dissociated and ionised in the arc column, from a thermodynamic
standpoint, there is no objection to the choice of molecular species as components for the system, provided that
equilibrium is maintained down to temperatures where such species are stable.
Carbon boil in the weld pool has been detected experimentally during welding with cov-
ered electrodes,49 which implies that heterogeneous nucleation of CO is kinetically feasible
under the prevailing circumstances. Possible nucleation sites for CO are gas bubbles present in
the macroscopic slag layer covering the metal, created by the vigorous stirring action of the arc
plasma jet. It should be noted that this behaviour is in sharp contrast to experience with GMA
welding, where little or no oxidation of carbon takes place in the weld pool, as shown previ-
ously in Section 2.10.1.2. It is tentatively suggested that the apparent difference between SMA
and GMA welding regarding the possibilities for CO nucleation in the weld pool arises from
the lack of a macroscopic slag layer in the latter case.
Unlike carbon, the deoxidation capacity of silicon (and manganese) increases rapidly with
decreasing metal temperatures (se Fig. 2.40), which means that carbon oxidation becomes
gradually suppressed during cooling in the weld pool. Upon reaching the critical temperature
indicated in Fig. 2.40, the carbon reaction is blocked, silicon and manganese now control the
oxygen level. An unknown but significant fraction of the manganese silicate inclusions pre-
cipitated in the hot part of the weld pool beneath the root of the arc are brought by convection
currents to the interface between the macroslag and the metal, where they are readily absorbed.
The remaining fraction formed in the cold and unstirred part of the weld pool is trapped in the
metal solidification front in the form of finely dispersed oxide particles. This results in a high
and rather unpredictable weld metal oxygen content.
The above reaction model has been tested experimentally against data obtained from a
series of hyperbaric welding experiments carried out in a remotely controlled pressure cham-
ber with basic covered electrodes containing various levels of ferrosilicon in the electrode
coating (see Table 2.9). Welding under hyperbaric conditions offer the special advantage of
assessing the reactions through variations in the ambient pressure without changing the com-
position of the electrode coating or the core wire. Consequently, if the proposed reaction model
Oxygen content, wt%
0.1 wt%Cat10bar.
0.1 wt% C at 1 bar
Temperature,°C
Fig. 2.40. The break even equilibrium temperature for silicon control of oxygen level at 0.1 wt% C and
0.3 wt% Si. Data from Elliott et al.23
Table 2.9 Contents of ferrosilicon and iron powder in the electrode coating of experimental
consumables used in the hyperbaric welding experiments.
Electrode FeSi (76 wt% Si) Iron Powder
R* 4.5 wt% 31wt%

A 5.5 wt% 30 wt%
B 6.5 wt% 29 wt%
C 7.5 wt% 28 wt%
^Reference electrode (E8018-C1 type electrode).
Electrode R
A
B
C
Carbon content, wt%
Low FeSi levels
Total pressure, bar
Fig. 2.41. Carbon absorption in hyperbaric SMA welding. Data from Grong et al.51
is at least qualitatively correct would expect a correlation between the weld metal carbon
content and the concentrations of oxygen, silicon and manganese, both under atmospheric and
hyperbaric welding conditions. The main effect of pressure on weld metal chemistry is thus to
suppress the carbon-oxygen reaction in the weld pool at the expense of intensified oxidation
losses of silicon and manganese, as indicated by the thermodynamic data in Fig. 2.40.
2.103.2 Absorption of carbon and oxygen

It is evident from the data presented Fig. 2.41 that the weld metal carbon content increases
monotonically with pressure from 1 to 31 bar for all four electrodes involved. This indicates
that the carbon oxidation in the weld pool is systematically suppressed under hyperbaric weld-
ing conditions. Moreover, Fig. 2.41 reveals a small but important effect of electrode deoxidation
capacity on the weld metal carbon content. Since ferrosilicon itself is an insignificant source of
carbon, the observed increase in the carbon concentrations with increasing additions of
ferrosilicon to the electrode coating is an indication that carbon oxidation in the weld pool is
blocked at an earlier stage of the process at high silicon levels, according to the reaction:
Si (dissolved) + 2CO(gas) = 2 C (dissolved) + SiO2 (slag) (2-54)
This interpretation is further supported by the results from the oxygen determination con-
tained in Fig. 2.42. Although there is considerable scatter in the data in this figure, it is evident
that the recorded enhancement of the weld metal carbon content at high ferrosilicon levels in
the electrode coating is accompanied by a corresponding reduction in the oxygen concentra-
tions.
2.10.3.3 Losses of silicon and manganese

Suppression of carbon oxidation in the weld pool at elevated pressures gives rise to intensified
oxidation losses of silicon and manganese, as shown in Figs. 2.43 and 2.44. Moreover, it is
apparent that increased additions of ferrosilicon to the electrode coating result in a correspond-
ing increase in both the silicon and the manganese concentrations. This finding suggests that
the final weld metal content of the deoxidants is controlled by the reaction:
Si (dissolved) + 2MnO (slag) = 2 Mn (dissolved) + SiO2 (slag) (2-55)
Assuming the activity ratio (tf Mn0 ) / (aSio2 )in precipitated slag to be constant and inde-
pendent of pressure, equation (2-55) may be rewritten as:
(2-56)
In Fig. 2.45 the weld metal manganese content has been plotted versus the square root of
the silicon content by inserting data from Figs. 2.43 and 2.44. As it appears from Fig. 2.45, the
Oxygen content, ppm
Electrode R
A
B
C
Total pressure, bar
Fig. 2.42. Oxygen absorption in hyperbaric SMA welding. Data from Grong et al.51
Electrode R
A
B
C
Total pressure, bar

Fig. 2.43. Silicon oxidation in hyperbaric SMA welding. Data from Grong et al.51
Electrode R
A
B
C
Manganese content, wt%
High FeSi levels
Low FeSi levels
Total pressure, bar
Fig. 2.44. Manganese oxidation in hyperbaric SMA welding. Data from Grong et al.51
experimental data cluster around a straight line passing through the origin, which confirms
that the silicon and manganese concentrations are balanced by a reaction according to equation
(2-55).
Electrode R
A
B
C
1/2
[Silicon content, wt%]
Fig. 2.45. Correlation between weld metal manganese and silicon contents. Data from Grong et a/.51
2.103.4 The product [%C] [%O]

From steelmaking practice, the product [%C] [%O] is generally accepted as an adequate index
of the interaction between carbon and oxygen during the refining stage. This product is related
to the equilibrium content of dissolved carbon and oxygen in contact with carbon monoxide of
a partial pressure pco:
(2-57)
Here K5 is the equilibrium constant for reaction (2-53) (equal to about 2.0 X 10"3 at 16000C
and 2.6 X 10~3 at 20000C),23 Nco is the mole fraction of CO in the reaction product (equal to
the partial pressure of CO at 1 bar), andptotis the total ambient pressure.
During the initial stage of cooling in the weld pool, the oxygen content in an assumed
equilibrium with carbon would be expected to be higher than the analytical values. This situa-
tion applies in particular to welds made under hyperbaric conditions, where significant quan-
tities of oxygen clearly are removed from the weld pool in the form of oxide inclusions after
the completion of the carbon oxidation.
The concentration of dissolved oxygen at the break even temperature for silicon control of
the oxygen level can be estimated from the measured concentration displacements of oxygen,
silicon and manganese in the weld deposit with increasing pressures, relative to 1 bar (desig-
nated A[%(9], A[%Si] and A[%Mrc], respectively). If the total amount of oxygen which reacts
with silicon and manganese at 1 bar, as a first approximation, is taken equal to the analytical
weld metal oxygen content, the following balance is obtained:
(2-58)
Next Page
Here [%0]eq. is the oxygen concentration in an assumed equilibrium with carbon at a given
pressure, and [%O]anaL is the analytical weld metal oxygen content. For this correlation, minor
vaporisation losses of manganese as well as possible reactions between oxygen and liquid iron
have been neglected.
In Fig. 2.46 the product m = [%C] [%O] is plotted vs the total ambient pressure. Calcula-
tions of m have been done both on the basis of [%O]anaL and [%O]eq. It can be seen from Fig.
2.46 that the former set of data (i.e. open symbols in the graph) cannot be represented by a
straight line passing through the origin, which should apply to a true equilibrium reaction.
However, when proper corrections are made for the amount of oxygen removed from the weld
pool after the completion of the carbon oxidation, such a correlation may be obtained as shown
by the solid line in Fig. 2.46. No clear effect of the electrode deoxidation capacity (i.e. ferrosilicon
content) on the product m = [%C] [%O] can be observed within the precision of measure-
ments. This result is to be expected if the weld metal carbon content is controlled by a local
equilibrium with oxygen established at elevated temperatures in the weld pool. Also, inspec-
tion of the slope of the curve (i.e. heavy solid line in Fig. 2.46) indicates that the product
K5Nco is about 1.14 X 10~3 under the prevailing circumstances. If a reasonable average value
for the equilibrium constant K5 of 2.3 X 10~3 is assumed within the specific temperature range
of the reaction, we get:
Nco~0.5 (2-59)
The above calculations suggest that the controlling partial pressure of CO in the reaction
product is significantly lower than the ambient pressure under hyperbaric welding conditions.
This probably arises from an extensive infiltration of helium in the nucleating bubbles at the
slag/metal interface which, thermodynamically, will enhance the deoxidation capacity of car-
bon according to Le Chatelier's Principle. The conditions existing in hyperbaric SMA welding
thus appear to be similar to OBM/Q-BOP steelmaking, where simultaneous injection of oxy-
gen and inert gas from the bottom of the convenor during the decarburisation stage results in a
steel carbon content which is typically below the value calculated for equilibrium between
oxygen and carbon at 1 atm partial pressure of CO.52
2.11 Weld Pool Deoxidation Reactions
During cooling, the metal concentrations established at high temperatures due to dissolution of
oxygen tend to readjust by precipitation of new phases. Accordingly, a supersaturation with
respect to the various deoxidation reactants initially increases and thus provides the driving
force for nucleation of oxides. Subsequently, the deoxidation reactions will proceed rapidly
through growth of nuclei above a critical size. Equilibrium conditions will finally establish the
limits for the degree of deoxidation that can be achieved.
In spite of the fact that large amounts of oxygen are removed from the weld pool during the
deoxidation stage, the analytical weld metal oxygen content exceeds by far the value predicted
from chemical thermodynamics, assuming that equilibrium conditions are maintained down to
the solidification temperature (see data in Table 2.5). This situation cannot be ascribed to a
large deviation from chemical equilibrium, but is mainly a result of an incomplete phase sepa-
ration. Consequently, due consideration must be given to the kinetics. The three basic consecu-
tive steps in steel deoxidation are shown in Fig. 2.47.
Previous Page
Here [%0]eq. is the oxygen concentration in an assumed equilibrium with carbon at a given
pressure, and [%O]anaL is the analytical weld metal oxygen content. For this correlation, minor
vaporisation losses of manganese as well as possible reactions between oxygen and liquid iron
have been neglected.
In Fig. 2.46 the product m = [%C] [%O] is plotted vs the total ambient pressure. Calcula-
tions of m have been done both on the basis of [%O]anaL and [%O]eq. It can be seen from Fig.
2.46 that the former set of data (i.e. open symbols in the graph) cannot be represented by a
straight line passing through the origin, which should apply to a true equilibrium reaction.
However, when proper corrections are made for the amount of oxygen removed from the weld
pool after the completion of the carbon oxidation, such a correlation may be obtained as shown
by the solid line in Fig. 2.46. No clear effect of the electrode deoxidation capacity (i.e. ferrosilicon
content) on the product m = [%C] [%O] can be observed within the precision of measure-
ments. This result is to be expected if the weld metal carbon content is controlled by a local
equilibrium with oxygen established at elevated temperatures in the weld pool. Also, inspec-
tion of the slope of the curve (i.e. heavy solid line in Fig. 2.46) indicates that the product
K5Nco is about 1.14 X 10~3 under the prevailing circumstances. If a reasonable average value
for the equilibrium constant K5 of 2.3 X 10~3 is assumed within the specific temperature range
of the reaction, we get:
Nco~0.5 (2-59)
The above calculations suggest that the controlling partial pressure of CO in the reaction
product is significantly lower than the ambient pressure under hyperbaric welding conditions.
This probably arises from an extensive infiltration of helium in the nucleating bubbles at the
slag/metal interface which, thermodynamically, will enhance the deoxidation capacity of car-
bon according to Le Chatelier's Principle. The conditions existing in hyperbaric SMA welding
thus appear to be similar to OBM/Q-BOP steelmaking, where simultaneous injection of oxy-
gen and inert gas from the bottom of the convenor during the decarburisation stage results in a
steel carbon content which is typically below the value calculated for equilibrium between
oxygen and carbon at 1 atm partial pressure of CO.52
2.11 Weld Pool Deoxidation Reactions
During cooling, the metal concentrations established at high temperatures due to dissolution of
oxygen tend to readjust by precipitation of new phases. Accordingly, a supersaturation with
respect to the various deoxidation reactants initially increases and thus provides the driving
force for nucleation of oxides. Subsequently, the deoxidation reactions will proceed rapidly
through growth of nuclei above a critical size. Equilibrium conditions will finally establish the
limits for the degree of deoxidation that can be achieved.
In spite of the fact that large amounts of oxygen are removed from the weld pool during the
deoxidation stage, the analytical weld metal oxygen content exceeds by far the value predicted
from chemical thermodynamics, assuming that equilibrium conditions are maintained down to
the solidification temperature (see data in Table 2.5). This situation cannot be ascribed to a
large deviation from chemical equilibrium, but is mainly a result of an incomplete phase sepa-
ration. Consequently, due consideration must be given to the kinetics. The three basic consecu-
tive steps in steel deoxidation are shown in Fig. 2.47.
Electrode R
A
B
C
[%C][%O]x102
Total pressure, bar
Fig. 2.46. The product [%C][%0] in hyperbaric SMA welding. Solid symbols: calculations based on
[%O]eq.. Open symbols: calculations based on [%O]anai.- Data from Grong et al. 51
Diffusion of reactants in the
Heterogeneous nucleation
Homogeneous nucleation
melt to the oxide nuclei
Surface tension
Separation Stage
Average particle size
Particle coalescence
Buoyancy
Nucleation Stage
Stirring
Growth Stage
Reaction time
Fig. 2.47. The three major consecutive steps in steel deoxidation (schematic).
Although rate phenomena in ladle refining of liquid steel are extensively investigated and
reported in the literature,53"55 only recently attempts have been made to include such effects in
an analysis of deoxidation reactions in arc welding. 1 ' 56 ' 57
2.11.1 Nucleation of oxide inclusions

During ladle-refining of liquid steel, it is well established that homogeneous nucleation of
oxide inclusions may occur in certain regions of the melt where the supersaturation is suffi-
ciently high.55'58 Over the composition range normally applicable to deoxidation of steel, the
number of nuclei formed at the time of addition of deoxidisers is approximately 105 mirr 3 . 55
However, in steel weld metals, the number of oxide nuclei formed during the initial stage of
deoxidation must be considerably higher to account for the observed inclusion number density
of about 107 mm~3 to 108 mm"3.56'57 This implies that the supersaturation established in the
weld pool on cooling as a result of rapid temperature fluctuations (~103 0C s"1) exceeds by far
that obtained by additions of deoxidisers to a liquid steel melt under approximately isothermal
conditions.
There are several theories available for treating nucleation phenomena, but for the purpose
of convenience, a simplified version of the model of Turpin and Elliott58 has been adopted
here. Consider a steel melt which is brought to a state of supersaturation by first equilibrating
it with pure MxOy at a high temperature T\ and then rapidly cooling it to a lower temperature
T2, as shown schematically in Fig. 2.48. It follows from the classical theory of homogeneous
nucleation that the required temperature difference T1-T2 necessary to achieve spontaneous
precipitation of MxOy is approximately given by the following relationship:
(2-60)
where AH° is the standard enthalpy of reaction, G is the oxide-steel interfacial energy (as-
sumed to be constant and independent of temperature), and Vm is the molar volume of the
nucleus.
The derivation of equation (2-60) is shown in Appendix 2.2.
Example (2.9)
Assume that precipitation of manganese silicates in the weld pool occurs according to the
following reaction:
Si (dissolved) + Mn (dissolved) + 3 O (dissolved) = MnSiO3 (slag) (2-61)
where
AG°(J) = 858620 + 3457.
Based on equation (2-60), calculate the critical temperature interval of subcooling for ho-
mogeneous nucleation OfMnSiO3. Typical physical data for liquid steel and manganese sili-
cate slags are given in Table 2.10.
£
Time
Fig. 2.48. Idealised model for homogeneous nucleation of oxide inclusions in steel weld metals (sche-
matic).
Table 2.10 Physical data for liquid steel and manganese silicate slags at 16000C. Data from Refs. 3 and
53.
Density Viscosity Interfacial Energyt

Property (kgnr 3 ) (kgm^s- 1 ) (Jm" 2 )
Steel 6900 4.3 X 10~3

Silicate slag 2300 - 0.8
1
In contact with liquid steel.
Solution
First we estimate the molar volume of the nucleus:
By inserting the appropriate values for Vm, AH°, and a in equation (2-60), we obtain:
from which
If we assume that the supersaturation is released at T2 = 16000C (1873K), the initial tem-
perature of the liquid T1 becomes:
The critical temperature interval of subcooling is thus:
Similar calculations can also be carried out for other types of oxide inclusions, e.g. FeO(I),
SiO2(s), and Al2O3(S). The results of such computations are presented graphically in Fig. 2.49,
using data from Refs 55 and 58. It is evident from these plots that the critical temperature
interval of subcooling depends on the interfacial energy, G. Although data for oxide-steel
interfacial energies are scarce, the following average values are frequently used in the litera-
ture, 5558 i.e. a(FeO-Fe) = 0.3 J m"2; G(SiO2-Fe) = 0.9 J m~2; and G(Al2O3-Fe) - 1.5 J mr2. If
these values are accepted, the results in Fig. 2.49 indicate that the critical temperature interval
of subcooling for homogeneous nucleation of FeO (1), SiO2(s), and Al2O3(S) is of the order of
200 to 30O0C. Considering the fact that the liquid weld metal spans a temperature range of
about 2200 to 15000C,34 it is not surprising to find that nucleation of oxide inclusions occurs
readily in the weld pool during cooling. It should be noted that the quoted data for G are
representative of ladle-refined steel deoxidised at 16000C. At higher metal temperatures, in the
presence of large amounts of dissolved oxygen, the oxide-steel interfacial energies would be
expected to be significantly lower.59 Hence, it is reasonable to assume that the actual tempera-
ture interval of subcooling required for spontaneous oxide precipitation in a weld pool is well
below 2000C.
Critical temperature interval, AT (0C)
Interfacial energy,o(J/m2)
Fig. 2.49. Critical temperature interval of subcooling for homogeneous nucleation of oxide inclusions in
steel weld metals at 16000C as a function of the interfacial energy a.
2.11.2 Growth and separation of oxide inclusions

In practice, there are three major growth processes in steel deoxidation:55
(i) Collision
(ii) Diffusion
(iii) Ostwald ripening.
From deoxidation and ladle refining of liquid steel it is well established that the flotation
rate of the oxides generally depends on their growth rates, since large inclusions separate
much more rapidly than small ones, in agreement with Stokes law.55 Growth of the oxides can
proceed either through diffusion of reactants in the melt to the oxide nuclei or by collision and
coalescence of ascending inclusions, and is therefore influenced by factors such as the number
density of the nuclei, interfacial tensions, and the extent of melt stirring.53"55 The last factor is
of particular importance in welding, because the stirring action will increase the possibilities
for collision and coalescence of inclusions and, depending on the direction of flow, can give
rise to circulation of inclusion-laden metal to the surface. As a result, the separation of small
oxide particles, i.e. microslag, is strongly favoured by the turbulent conditions existing in the
hot part of the weld pool immediately beneath the root of the arc.
2.11.2.1 Buoyancy (Stokes flotation)

Assessment of the role of buoyancy (Stokes flotation) in the separation of deoxidation prod-
ucts from the weld pool can be done on the basis of the experimental data of Grong et ai56
reproduced in Fig. 2.50. It is evident that a change in the welding position (i.e. from flat to
overhead position) has no significant effect on the weld metal oxygen content. This shows that
the buoyancy effect does not play an important role in the separation process of oxide inclu-
sions during welding.
Upward welding
Downward welding
Oxygen content, wt%
Horizontal
Welding orientation, degrees

Fig. 2.50. Effect of welding position on retained oxygen in GMA steel weld deposits. Data from Grong et
a/. 56
The above conclusion is also in agreement with predictions based on the Stokes law, which
gives the terminal velocity of the ascending particles (u) relative to the liquid:55
(2-62)
where g is the gravity constant, dv is the particle diameter, Ap is the difference in densities
between the liquid steel and the inclusions, and |x is the steel viscosity.
Taking the Stokes parameter, gAp/18|ji, equal to 0.6 jjinr1 s"1 for manganese silicate slags in
steel (Table 2.10), equation (2-62) becomes:
(2-63)
Generally, the majority of non-metallic inclusions in steel weld metals are of a diameter
below 2 jxm.56'57 According to equation (2-63), such particles have a relative velocity less than
2.4 |xm s"1. This implies that the buoyancy effect alone is far too insignificant to promote
flotation of the inclusions out of the weld pool before solidification, when it is recognised that
the average fluid flow velocity in the weld pool is four to five orders of magnitude higher (to
be discussed later).
Stokes law is based on the assumption that the inclusions are completely wetted by the
liquid steel, i.e. there is no slip at the oxide/metal interface.55 Normally, interfacial tension
effects promote slip at the particle/metal interface, which, in turn, enhances the flotation rate
of the ascending oxides (often referred to as the Plockinger effect).60 The concept of a wetting
angle has been used in this context. But, the slip phenomenon is probably perceived better in
terms of a secondary flow in the interfacial region between the liquid metal and the solid
precipitate, which is produced by gradients in interfacial tension. In the rare case of no wetting
(0 = 180°), it has been shown that the average terminal velocity of the inclusions is approxi-
mately 50% higher than that given by equation (2-62).55 For silica slags, the wetting angle (in
the presence of air) is close to 115°,54 indicating that the correction in the particle terminal
velocity due to the slip at the oxide/metal interface in less than about 30%. Consequently,
interfacial tension effects between the slag and the steel do not significantly affect the flotation
rate of the particles, and, therefore, can be ignored.
2.11.2.2 Fluid flow pattern

It is evident from the above discussion that the separation of deoxidation products during arc
welding is controlled by the fluid flow fields set up within the weld pool. The typical flow
pattern in a submerged arc weld pool is shown in Fig. 2.51. It can be seen from the figure that
a depression is formed at the forward edge of the pool, which forces the melted metal to flow
underneath and on either side of the depression, following the arrows in Fig. 2.51. At the rear
of the weld pool the flow direction is reversed, and the metal streams back along the pool
surface. A similar flow pattern has also been observed in GMA weld pools.61 As a result, the
inclusions precipitated in the turbulent part of the weld pool are rapidly brought to the upper
surface behind the arc due to the high-velocity flow field created within the liquid metal, and
disengaged by the surface tension effects in the pool.
The drag force exerted on the particles because of the liquid flow-field velocity can be
estimated from published data for weld pool flow velocities following electromagnetic fluid
mechanics theory. Normally, for weld pools the Peclet number for heat transfer lies within the
range from 10 to 5000, which indicates that the heat flow is predominately convectional.62 The
limits quoted above for the Peclet number correspond to a range in the average weld pool fluid
velocity from about 0.025 to 0.4 m s"1 in the case of SA welding.6364 Unfortunately, specific
information about the fluid flow velocity gradients in the weld pool are lacking, which pre-
vents a more complete analysis of variations in the flow pattern with increasing distance from
the root of the arc.
Generally, the drag force, Fd, acting on a spherical particle in relative motion to a fluid can
be expressed as:3
(2-64)
where Cd is the drag coefficient, p is the fluid density, u is the bulk velocity of the fluid
relative to the particle, and dv is the particle diameter. The drag coefficient is, in turn, a func-
tion of the particle Reynolds number, A^,:3
(2-65)
where |x is the fluid viscosity. Typical physical-property data for liquid steel are given in Table
2.10. By inserting the values for p and JJL from Table 2.10, then using the relative velocity
calculated in equation (2-63) (2.4 |xm s"1) as one extreme and the fluid flow velocity (0.4 m s"1) as
the other (i.e. assuming a stationary particle), one can demonstrate that the particle Reynolds
numbers obtained never exceed 1.3 for most weld metal inclusions. The limit calculated above
for the Reynolds number is within the so-called creeping flow region where Stokes law is
indeed valid. Under such conditions, the ratio between the particle drag force Fd and the corre-
sponding gravity force Fg is given by the expression:3
(2-66)
Note that equation (2-66) is the basis for obtaining Stokes law (equation (2-62)).
For steady motion of the particle, Fd and Fg are, of course, equal. However, the significance
of Fd can also be interpreted in a transient sense, by considering the limiting case where the
particle is stationary and its instantaneous motion is governed by the forces acting on it. In this
case, the ratio of Fd (max) acting on the particle relative to the net gravity force (given by
Electrode
Pulsating cavity
Base plate
Fig. 2.51. Typical flow pattern in SA weld pools. For clarity, the arc, slag and flux have been omitted.
The sketch is based on the ideas of Lancaster.32
equation (2-66)) can determine the dominant force responsible for the particle's trajectory. In
Fig. 2.52 the above ratio has been calculated and plotted against the particle diameter, dv, for
values of u' equal to 0.025 and 0.4 m s"1, which represent the typical velocities reported for
SA steel weld pools. 63 ' 64 It can be seen from the figure that the drag force is always several
orders of magnitude greater than the gravity force for particles within the typical size range of
weld metal inclusions (i.e. less than 2 |xm in diameter). This result is to be expected, since the
relative velocity, based on Stokes law and calculated in equation (2-63), is negligible com-
pared with the liquid velocities in the weld pool (2.4 \xm s"1 vs 0.02 to 0.4 m s"1). This calcu-
lation thus supports our previous conclusion that the separation of the oxide inclusions is con-
trolled solely by the fluid flow behaviour in the weld pool. The fact that the phase separation
proceeds under strongly turbulent conditions is also evidenced by the large number of iron
droplets being mechanically dispersed in the top bead slag of GMA steel welds, as shown by
the optical micrographs in Fig. 2.53.
In the case of GMA welding, the non-metallic inclusions that are brought to the upper
surface behind the arc coalesce rapidly to form large slag clusters that float on the top of the
bead. Generally, re-entrapment of the slag does not occur owing to the decrease in the total
surface free energy of the system, which is caused by the emergence of the inclusions from the
weld metal.54 Consequently, the slag will remain floating on the top of the bead even when
welding is performed in the overhead position, as shown previously in Fig. 2.50.
2.11.2.3 Separation model

Based mainly on experience with the GMA welding process, a simple model for the assess-
ment of the sequence of deoxidation reactions in arc welding has been proposed,1 and is shown
schematically in Fig. 2.54. The model is based on the assumption that equilibrium between the
reacting elements and precipitated slag is maintained down to low metal temperatures, and
divides the weld pool into the following two main reaction zones:
*ci(max/Fg
Inclusion diameter, |i.m

Fig. 2.52. The ratio between maximum particle drag force, Fornax) and the corresponding gravity force,
F8, vs the inclusion diameter at two different flow velocities in the weld pool. Calculations are based on
equation (2-66).
(a) (b)
Fig. 2.53. Optical micrographs showing characteristic "comet's tails" of trapped iron droplets (light
areas) in collected top bead slags of GMA steel welds; (a) Ar + 5vol%O2, (b) Ar + 5vol%CO2.
Electrode
(a)
Cold part of weld pooN Arc Hot part of weld pool
Base plate
(b) S e c t i o n A-A
Cold part of weld pool: Hot part of weld pool: Final weld metal,
Deoxidation/phase oxygen concentration
Temperature
Deoxidation/incomplete separation
phase separation
Oxygen content
Fig. 2.54. Schematic diagrams showing the sequence of reactions occurring during weld metal deoxidation;
(a) Longitudinal section of weld pool, (b) Cross section of weld pool along A-A. The diagrams are based
on the ideas of Grong and Christensen.1
(i) The hot part of the weld pool, characterised by simultaneous oxidation and deoxidation
of the metal, where the separation of microslag takes place continuously as a result of highly
turbulent flow conditions.
(ii) The cold part of the weld pool, where precipitated slag will largely remain in the metal
as finely dispersed particles as a result of inadequate melt stirring.
Under such conditions equation (2-23) predicts a direct correlation between absorbed and
retained (analytical) oxygen in the weld metal, i.e.:
(2-67)
in agreement with experimental observations (see plots in Figs. 2.33, 2.38, and 2.39). Typi-
cally, the proportionality constant C4 varies between 0.1 to 0.2, which corresponds to a range
in the t/to ratio from 2.3 to 1.6. This shows that the boundary between 'hot' and 'cold' parts of
the weld pool is not well defined, but depends on the welding system under consideration as
well as on the operational conditions applied.
2.11.3 Predictions of retained oxygen in the weld metal

Although the weld metal oxygen content is controlled by a transport mechanism in the weld
pool, the concept of pseudo-equilibrium can still be used for an assessment of slag/metal reac-
tions in arc welding.
2.11.3.1 Thermodynamic model
In the case of silicon deoxidation of steel weld metals, we may write:
Si (dissolved) + 2 O (dissolved) = SiO2 (slag) (2-68)
On introduction of the equilibrium constant for the reaction, we obtain:

(2-69)
To allow for the decrease in the silica activity with increasing manganese-to-silicon ratios,
it is essential to establish a correlation that links the activity of silica to the concentrations of
the deoxidation elements in the weld metal. A semi-empirical corr61ation of this kind has been
presented by Walsh and Ramachandran,65 derived from a re-analysis of activity data for silica
in the Fe-Mn-Si-O system previously published by Hilty and Crafts.66 Within the tempera-
ture range from 1550 to 16500C, they showed that the silica activity in the deoxidation product
can be approximately expressed as:
(2-70)
where K1 represents the manganese-to-silicon ratio at which the activity of silica becomes
unity for a given temperature. A check of this equation against more recent data for silica
activities in MnO-SiO2 slags reported by Turkdogan67 supports the findings of Walsh and
Ramachandran65 that the activity of silica is approximately given by equation (2-70) for a wide
range in the steel manganese-to-silicon ratio (i.e. from 0.1 to 50). By combining equations (2-
69) and (2-70), it is possible to obtain an expression for the equilibrium oxygen content, solely
in terms of the silicon and manganese concentrations:
(2-71)
where K^ is a temperature-dependent parameter equal to (K1I K6)05.

The temperature dependence of the Si-O reaction (equation (2-68)) is well established and
is approximately given by the relationship:23
(2-72)
when pure SiO2 is used as the standard state for the silica activity.
For a rough estimate of the temperature dependence of equation (2-70), the results of
Turkdogan55 can be used. It should be noted that Walsh and Ramachandran65 did calculate the
temperature dependence of K1 within the range from 1550 to 16500C. However, because equa-
tion (2-70) is empirical, the function cannot be extrapolated beyond these temperature limits.
The data quoted in Ref. 55 are derived directly from the Si-Mn reaction (equation (2-55)) and
activity data for MnO at silica saturation. On introduction of the equilibrium constant for
equation (2-55), we obtain:
(2-73)
By using data from Ref.55, the initial [%Mn]2/[%Si] ratio for precipitation of silica satu-
rated slags (equal to K9(aMnO)2) at 1500,1550,1600 and 16500C has been recorded and replotted
against temperature, as shown in Fig. 2.55. The figure shows that the critical [%Mn]2/[%Si]
ratio for precipitation of silica saturated slags is temperature dependent and decreases from
about 5 at 15000C to below 1.5 at 16500C. On the basis of a crude extrapolation of the data to
higher metal temperatures, it can, however, be argued that the ratio would approach a constant
value of approximately 0.75 at temperatures beyond about 17500C (indicated by the broken
horizontal line in Fig. 2.55). The above observation reflects the fact that the Si-Mn-O reaction
equilibrium (equation (2-55)) is not very sensitive to a change in temperature (i.e. the enthalpy
for the reaction is small). Over the composition range normally applicable to Si-Mn deoxidation
of steel weld metals the observed threshold for the critical [%Mn]2/[%Si] ratio for precipita-
tion of silica-saturated slags in Fig. 2.55 corresponds to a manganese-to-silicon ratio closely
equal to unity. Consequently, at temperatures higher than about 17500C, K1 can, as a first
approximation, be taken constant and independent of temperature (i.e. K1- Y). The tempera-
ture dependence of equation (2-71) is thus simply (1/^ 6 ) 05 or:
(2-74)
for temperatures higher than about 17500C (2023K).

[%Mn]2/[%Si]
Temperature, 0C
Fig. 2.55. The critical [%Mn]2/ [%Si] ratio for precipitation of silica-saturated slags as a function of
temperature. Data from Turkdogan.55
2.11.3.2 Implications of model

Figure 2.56 shows plots of retained (analytical) oxygen in GMA and SMA steel weld metals vs
the theoretical deoxidation parameter ([%Si][%Mn])~°25, using relevant literature data. 5668
This parameter allows for the inherent decrease in the silica activity with increasing weld
metal Mn to Si ratios. A closer inspection of the slopes of the curves reveals that the effective
reaction temperature falls within the range calculated for chemical equilibrium between sili-
con, manganese, and oxygen at 1800 to 19000C. Although the boundary between 'hot' and
'cold' parts of the weld pool for possible inclusion removal in practice is not sharp, the results
in Fig. 2.56 literally suggest that all oxides precipitated above 1800 to 19000C are simultane-
ously removed from the weld pool under the prevailing turbulent flow conditions. At lower
temperatures, the degree of melt stirring is too low to promote separation of the deoxidation
products out of the pool before solidification and hence, they are trapped in the metal solidifi-
cation front in the form of finely dispersed inclusions. Moreover, the observed difference in
the effective reaction temperature between GMA and SMA welding supports our previous
conclusion that any change in welding parameters, flux or shielding gas composition, which
alters the fluid flow pattern in the weld pool, will often have a stronger influence on the weld
metal oxygen content than variations in the deoxidation practice. This implies that control of
the weld metal oxygen level through additions of deoxidants, in practice, is difficult to achieve.
2.12 Non-Metallic Inclusions in Steel Weld Metals
Inclusions commonly found in steel weldments will either be exogenous or indigenous, de-
pendent on their origin. The first type arises from entrapment of welding slags and surface
scale, while indigenous inclusions are formed within the system as a result of deoxidation
reactions (oxides) or precipitation reactions (nitrides, sulphides). The latter group is almost
always seen to be heterogeneous in nature both with respect to chemistry (multiphase parti-
GMA Welding
Oxygen content, wt% SMA Welding
([%SI] [%Mn]f°"25
Fig. 2.56. Examples of pseudo-equilibrium in GMA and SMA welding of C-Mn steels. The solid lines in
the graph represent thermodynamical calculations at indicated temperatures. Data from Refs. 56 and 68.
cles), shape (angular or spherical particles), and crystallographic properties as a result of the
complex alloying systems involved69 (see Fig. 2.57). An exception may be C-Mn steel welds,
where the oxide inclusions will be predominately glassy, spherical, manganese silicates.1
A survey of important weld metal inclusion characteristics is given in Table 2.11.
2.12.1 Volume fraction of inclusions

It is evident from Table 2.11 that the volume fraction of non-metallic inclusions in steel weld
metals normally falls within the range from 2 X 10~3 to 8 X 10~3, depending on the type of
weld under consideration. Based on simple stoichiometric calculations it is possible to convert
the analytical weld metal oxygen and sulphur contents to an equivalent inclusion volume frac-
tion when the chemical composition of the reaction products is known. This is shown below.
Table 2.11 Summary of weld metal inclusion characteristics. The data are compiled from miscellane-
ous sources.
Size Distribution* Chemical Composition

Type of Weld Constituent Reported
Vv X 10"3 dv NvX 107 Sv elements phases
C-Mn steel Si, Mn, O, S SiO2,

weld metals 3-8 0.3-0.6 1-10 10-50 (traces of Al, MnOSiO2,
Ti, and Cu) MnS, (CuxS)
Low-alloy Al, Ti, Si, MnOAl2O3,
steel weld 2-6 0.3-0.7 0.5-5 10-40 Mn, O, S, N 7-Al2O35TiN,
metals (Cu) MnOSiO2, SiO2,
a-MnS, (3-MnS,
(Cu^S)
*Vy: volume fraction; dv: arithmetic mean (3-D) particle diameter (jxm); Nv: number of particles per unit volume
(mm"3); Sv\ total particle surface area per unit volume (mm2 per mm3).
Fig. 2.57. Digital STEM brightfield image and Si, Al, S, Mn and Ti X-ray images of a multiphase weld
metal inclusion. After Kluken and Grong.57
Solution
First we calculate the total weight of retained MnOSiO2 per 100 g weld metal:
This corresponds to an equivalent volume fraction of:
Similarly, we can calculate the weight and volume fraction of MnS in the weld metal:
and
The total volume fraction of MnOSiO2 and MnS is thus:
In practice, the stoichiometric conversion factors for oxygen and sulphur are virtually con-
stant for a wide spectrum of inclusions70 and hence, they can be regarded as independent of
composition. Taking the solubility of sulphur in solid steel equal to 0.003 wt%, the following
relationship is obtained for steel weld metals:57'70
(2-75)
The validity of equation (2-75) has been confirmed experimentally by comparison with
microscopic assessment methods.5771
In steel weld metals the majority of the inclusions will be in the submicroscopic range
owing to the limited time available for growth of the oxides. From the histogram in Fig. 2.58 it
is seen that particles with diameters between 0.3 to 0.8 |xm contribute to nearly 50 percent of
the total inclusion volume fraction. This trend is not significantly changed by additions of
strong deoxidisers, such as aluminium and titanium, or by a moderate increase/decrease in the
heat input.57
2.12.2 Size distribution of inclusions

As shown in Fig. 2.59 the majority of the three-dimensional (3-D) inclusion diameters fall
within the range of 0.05 to 1.5 |xm, with a characteristic peak in the particle frequency at about
0.4 to 0.5 |jim. These data obey the log-normal law, since a plot of the frequencies against the
logarithms of the diameters approximately gives a symmetrical curve. Considering specific
inclusion size classes, deoxidation with aluminium generally results in a higher fraction of
coarse particles (> 1 |xm) due to incipient clustering of Al2O3.57' 69 However, the observed
particle clustering has no significant influence on the arithmetic mean 3-D inclusion diameter,
as shown by the data in Fig. 2.59.
Relative volume fraction (%)
Particle diameter (jim)
Fig. 2.58. Percental contribution of different size classes to the total volume fraction of non-metallic
inclusions in a low-alloy steel weld metal. Data from Kluken and Grong.57
2.12.2.1 Effect of heat input

In contrast to the situation described above, the 3-D inclusion size distribution is strongly af-
fected an increase in the heat input (see Fig. 2.60). At 1 kJ mm"1, the measured 3-D inclusion
diameters fall within the range from 0.05 to 1 |xm, with a well-defined peak in the particle fre-
quency at about 0.3 |xm. When the heat input is increased to 8 kJ mm"1, the content of coarse
inclusions will dominate (>0.5 |xm), which results in a broader distribution curve and a shift
in the peak frequency towards larger particle diameters. A comparison with Fig. 2.61 reveals
that the arithmetic mean 3-D inclusion diameter is approximately a cube-root function of the
heat input. This result is to be expected if Ostwald ripening is the dominating coarsening
mechanism in the cold part of the weld pool (to be discussed below).
2.12.2.2 Coarsening mechanism

As already mentioned in Section 2.11.2 there are three major growth processes in steel deox-
idation, i.e. (i) collision, (ii) diffusion, and (iii) Ostwald ripening. In the cold part of the weld
pool, particle growth by collision can be excluded in the absence of adequate melt stirring be-
cause of a low collision probability of inclusions while ascending in the molten steel within the
regime of Stokes law.72 In addition, the diffusion-controlled part of the deoxidation reaction
(which involves diffusion of reactants in the melt to the oxide nuclei) would be expected to
be essentially complete within a fraction of a second when the number of nuclei is greater
than 107 mm"3.55 This implies that the observed increase in the inclusion diameter with in-
creasing heat inputs (Fig. 2.61) can be attributed solely to Ostwald ripening effects.
Before discussing details of the inclusion growth kinetics, it is essential to clarify the tem-
perature level in the 'cold' part of the weld pool. As shown by the results in Fig. 2.62, the liquid
metal temperature in the trailing edge of the weld pool is fairly constant and slightly above
the melting point of the steel. Accordingly, inclusion growth in welding (at a fixed volume frac-
tion) can be treated as an isothermal process, where the time dependence of the mean par-
ticle diameter dv is approximately given by the Wagner equation:
(a)
Low Al (0.018 wt%)
Low Ti (0.005 wt%)
Frequency (%)
Particle diameter (fxm)
(b)
High Al (0.053 wt%)
High Ti (0.053 wt%)
Frequency (%)
Particle diameter, p,m
Fig. 2.59. Three-dimensional (3-D) size distribution of non-metallic inclusions in two different low-
alloy steel weld metals; (a) Low weld metal aluminium and titanium levels, (b) High weld metal alu-
minium and titanium levels. Data from Kluken and Grong.57
(2-76)
Here do is the initial particle diameter, a is the oxide-steel interfacial energy, Dm is the
element diffusivity, Cm is the element bulk concentration, V'm is the molar volume of the oxide
per mole of the diffusate, and t is the retention time.
Frequency (%)
Particle diameter, p,m

Fig. 2.60. Effect of heat input on the 3-D inclusion size distribution in low-alloy steel weld metals. Data
from Kluken and Grong.57
Particle diameter, u,m
Heat input, kJ/mm

Fig. 2.61. Variation of arithmetic mean 3-D inclusion diameter with heat input during SA welding. Data
For welding of thick plates, the time available for growth of particles in the 'cold' part of
the weld pool can be estimated from the Rosenthal equation, i.e. equation (1-45) in Section
1.10.2 (Chapter 1). If the characteristic length of the cooling zone is taken equal to the weld
ripple lag (defined in Fig. 2.63), the retention time t is approximately given by the following
relationship:
Temperature, C°
"Tpeak = 1538°C (about 8 mm from edge of weld pool)
Time, seconds
Fig. 2.62. Measured temperature level in the trailing edge of the weld pool during GMA welding. Data
Top view of weld crater (z = 0)
Retention time
Weld ripple lag
Welding speed
Max. width
Heat source
Fusion boundary
Fig. 2.63. Definition of weld ripple lag xm, and retention time t.
(2-77)
where J\ is the arc efficiency (equal to about 0.95 for SAW and 0.80 for GMAW /SMAW), and
E is the gross heat input (kJ mm"1). Note that equation (2-77) assumes constant values for the
steel thermal diffusivity and volume heat capacity (5 mm2 s"1 and 0.0063 J mm"3 0C"1, respec-
tively), and no preheating (T0* = 200C).
In Fig. 2.64 the Ostwald ripening theory has been tested against relevant literature data,
which may be considered representative of the 3-D particle size distribution. Although there is
some scatter in the data, the observed inclusion growth rates fall within the range calculated
for oxygen diffusion-controlled coarsening of SiO2 and Al2O3 at 15500C, using the Wagner
equation. In these calculations, a reasonable average value for the bulk diffusivity of oxygen
has been assumed (i.e. 10~2 mm2 s"1).55 If the effective inclusion growth rate constant for low-
alloy steel weld metals is taken equal to the slope of the curve in Fig. 2.64, the following
relationship is achieved:
GMAW
SAW
SMAW
dv,jLim
•1/3
i i sJ/3
Fig. 2.64. Relation between arithmetic mean (3-D) inclusion diameter dv and retention time t for different
arc welding processes. Data compiled by Kluken and Grong.57
1
'Hot' part of CoId1 part of
weld pool weld pool
Solid weld metal

Solid weld metal
Temperature
Incomplete phase;
Phase separation
Deoxidation
Deoxidation
separation
Growth of
inclusions
Oxidation
Gas/metal- Homogeneous
slag/metal nucleation Ostwald ripening
reactions
Time
Fig. 2.65. Proposed deoxidation model for steel weld metals (schematic). The diagram is based on the
ideas of Kluken and Grong.57
(2-78)
By substituting equation (2-77) into equation (2-78), we obtain a direct correlation between
the arithmetic mean 3-D inclusion diameter dv and the net heat input T\E:
(2-79)
Equation (2-79) predicts that dv is a simple cube root function of E, in agreement with the
experimental data in Fig. 2.61.
It should be noted that the measured shape of the particle distributions (see Figs. 2.59 and
2.60) deviates somewhat from that required by the Wagner equation, which assumes a quasi-
stationary distribution curve, and that the maximum stable particle diameter is about 1.5 times
the mean diameter of the system.73 Although the origin of this discrepancy remains to be
resolved, this suggests that particle clustering is also a significant process in steel weld metals
as it is in other metallurgical systems. In fact, such effects would be expected to be most
pronounced at high aluminium levels because of a large interfacial energy between Al2O3 and
liquid steel, in agreement with experimental observations.57'69
2.12.2.3 Proposed deoxidation model

Referring to Fig. 2.65, the sequence of reactions occurring during weld metal deoxidation can
be summarised as follows. In general, nucleation of oxide inclusions occurs homogeneously
as a result of the supersaturation established during cooling in the weld pool. The diffusion-
controlled deoxidation reactions (i.e. diffusion of reactants to the oxide nuclei) will be essen-
tially complete when the liquid metal temperature attains a constant level of about 15500C at
some distance behind the root of the arc. Growth of the particles may then proceed under
approximately isothermal conditions at a rate controlled by the Wagner equation until the
temperature reaches the melting point of the steel. Since retention times in welding generally
depend on the heat input, it follows that choice of operating parameters will finally determine
the degree of particle coarsening to be achieved.
Example (2.11)
Consider SA welding with a basic flux on a thick plate of low-alloy steel under the following
conditions:
Previous experience has shown that this steel/flux combination gives a weld metal oxygen
and sulphur content of 0.035 and 0.008 weight percent, respectively. Based on the stereometric
relationships given below, calculate the total number of particles per unit volume Nv, the total
number of particles per unit area Na, the total particle surface area per unit volume Sv, and the
mean particle centre to centre volume spacing Xv in the weld deposit:74'75
(2-80)
(2-81)
(2-82)
(2-83)
Solution
First we calculate the total volume fraction of oxide and sulphide inclusions from equation
(2-75):
Vv = 10-2 [5.0(0.035) + 5.4(0.008 -0.003)] = 2.0 X 10~3
The arithmetic mean (3-D) inclusion diameter can then be evaluated from equation (2-79):
This gives:
particles per mm3 particles per mm3
particles per mm 2 particles per mm 2
mm 2 per mm3 mm 2 per mm3
2.123 Constituent elements and phases in inclusions

It is evident from Table 2.11 that non-metallic inclusions commonly found in low-alloy steel
weld metals may contain a considerable number of constituent elements and phases.
2.12.3.1 Aluminium, silicon and manganese contents

Figure 2.66 shows examples of measured X-ray intensity histograms for silicon, manganese,
and aluminium in inclusions extracted from a low-alloy steel weld metal. These results have
been converted into arbitrary elemental weight concentrations by normalising the collected X-
ray data for all preselected elements (i.e. Al, Si, Mn, Ti, Cu, and S) to 100%. The characteristic
normal distribution curves recorded for silicon, manganese, and aluminium show that the con-
tent of each oxide phase may vary significantly from one particle to another. This observation
is not surprising, considering the complex chemical nature of the weld metal inclusions (see
Fig. 2.57).
2.12.3.2 Copper and sulphur contents

In addition, the inclusions may contain significant levels of both copper and sulphur in addi-
tion to aluminium, silicon, manganese and titanium. Sulphide shells around extracted inclu-
sions have frequently been observed in SA and SMA steel welds, often in combination with
copper. This has been taken as an indication of copper sulphide formation.
However, based on the wavelength dispersive X-ray (WDX) data reported by Kluken and
Grong,57 it can be argued that copper sulphide is a rather unlikely reaction product in steel
weld metals as it is in ordinary ladle-refined steel.76 From their data it is evident that the total
Silicon Manganese Aluminium
26 wt%
28 wt%
20 wt%
Frequency (%)
Arbitrary elemental weight concentrations, wt%

Fig. 2.66. Measured X-ray intensity histograms for silicon, manganese and aluminium in inclusions
extracted from a low-alloy steel weld metal. Arrows indicate average composition. Data from Kluken
and Grong.57
number of counts recorded for copper in discrete particles is not significantly higher than the
corresponding matrix value, which shows that the copper content of the inclusions is low.
Since these measurements were carried out on mechanically polished specimens and not on
carbon extraction replicas, as done in the EDX analysis, the indications are that the higher
inclusion copper level observed in the latter case mainly results from surface copper contami-
nation inherent from the extraction process. In contrast, the WDX analysis of sulphur revealed
evidence of sulphur enrichment in most of the particles. Considering the fact that these parti-
cles also contained significant amounts of manganese, it is reasonable to assume that most of
the sulphur is present in the form of MnS (possibly with some copper in solid solution).76
2.12.3.3 Titanium and nitrogen contents

From the literature reviewed, it is apparent that conflicting views are held about the role of
titanium in weld metal deoxidation. From a thermodynamic standpoint, Ti2O3 is the stable
reaction product in titanium deoxidation,76 but this phase has not yet been detected experimen-
tally in steel weldments (only in continuous cast steel).77 However several authors have
reported the presence of crystalline patches containing titanium toward the edges of
inclusions.78"80 This constituent has a cubic crystal structure with a lattice parameter close to
0.42 nm, conforming to either 7-TiO, TiN, or TiC (note that the 7-T1O phase is the high-
temperature modification of the titanium monoxide).81
In general, formation of titanium monoxide requires strongly reducing conditions, which
implies that the 7-TiO constituent is not an equilibrium reaction product in steel deoxidation.55'76
Hence, it is reasonable to assume that the titanium-rich crystalline patches observed toward
the edges of weld metal inclusions are titanium nitride. This conclusion has later been con-
firmed experimentally by Kluken and Grong.57
2.12.3.4 Constituent phases

Based on the above discussion, it is possible to rationalise the formation of primary and sec-
ondary reaction products (i.e. oxides, sulphides and nitrides) during cooling in the weld pool,
as shown in Fig. 2.67. In general, the inclusions will consist of an oxide core which is formed
during the primary deoxidation stage. The chemical composition of the deoxidation product
can vary within wide limits, depending on the activities of Al, Ti, Si, Mn, and O in the weld
metal. The surface of the oxides will partly be covered by MnS and TiN (see also STEM
micrograph in Fig. 2.57). Precipitation of these phases occurs after the completion of the weld
metal deoxidation, probably during solidification, where the reactions are favoured by solute
enrichment in the interdendritic liquid. Additional precipitation of TiN may occur in the solid
state as a result of diffusion of titanium and nitrogen to the surface of the inclusions.
2.12.4 Prediction of inclusion composition

Since the diffusion-controlled deoxidation reactions are completed within a fraction of a sec-
ond when the number of nuclei is of the order of 107 mm"3 or higher,55 the average chemical
composition of the inclusions should be compatible with that calculated for chemical equilib-
rium at temperatures close to the melting point (e.g. 15500C).
2.12.4.1 C-Mn steel weld metals

Over the composition range normally applicable to silicon-manganese deoxidation of steel
weld metals (i.e. between 0.4 to 0.7 wt% Si and 0.8 to 1.5 wt% Mn) the equilibrium reaction
product at 15500C should be silica-saturated slags with a mole fraction of SiO2 close to 0.55. 5565
Since the two other slag components are MnO and FeO, we may write:
(2-84)
The activity coefficients for MnO and FeO in the ternary system SiO 2 -MnO-FeO can be
computed from the equations presented by Sommerville et al.:S2
(2-85)
and
(2-86)
For the specific case of silica-saturated slags, we obtain:
(2-87)
and
(2-88)
TiN (secondary reaction product)
Oxide Core (primary reaction product]
MnS (secondary reaction product)
Fig. 2.67. Schematic diagram showing the presence of primary and secondary phases in weld metal
inclusions.
Under such conditions the activity ratio aMnO/aFeO in the slag is given by:
(2-89)
The activity ratio aMnO/aFeO can also be estimated from the equilibrium constant for the Fe-
Mn reaction at 15500C,23 i.e.:
(2-90)
The corresponding mole fractions of MnO and FeO in the slag phase are then obtained by
combining equations (2-89) and (2-90):
(2-91)
and
(2-92)
Equations (2-91) and (2-92) provide a basis for calculating the chemical composition of the
inclusions under different deoxidation conditions. A requirement is, however, that the weld
metal Si to Mn ratio is sufficiently high to promote precipitation of silica-saturated slags at
15500C.
and
This gives the following chemical composition of the inclusions (in wt%):
The above calculations should be compared with the composition data presented in Fig.
2.68. It is evident from this plot that the agreement between predictions and experiments is
reasonably good both at high and low weld metal manganese levels. This justifies the
simplifications made in deriving equations (2-91) and (2-92).
2.12.4.2 Low-alloy steel weld metals

In principle, a procedure similar to that described above could also be used to establish a
theoretical basis for predicting the chemical composition of the inclusions in the case low-
alloy steel weld metals. Unfortunately, adequate activity data for the Fe-Al-Ti-Si-Mn-O sys-
tem are not available in the literature.
An alternative approach would be to calculate the average inclusion composition from sim-
ple mass balances, assuming that all oxygen combines stoichiometrically with the various
deoxidation elements to form stable oxides in the order Al2O3, Ti2O3, SiO2, and MnO, accord-
ing to their oxygen affinity in liquid steel (see Fig. 2.69). If reasonable average values for the
inclusion and steel densities are used (i.e. 4.2 and 7.8 g cm"3, respectively), the following set of
equations* can be derived from a balance of O, Al, Ti, S, Si, and Mn and data for acid soluble
aluminium and titanium in the weld metal:57
Aluminium
The average aluminium content of the inclusions, [%Al]inch can be estimated from the meas-
ured difference between total and acid soluble aluminium in the weld metal, (k%Al)we[d. This
difference is, in turn, equal to the total mass of aluminium in the inclusions:
(2-93)
where mind and Vv (cal) are the total mass and volume fraction of non-metallic inclusions in
the weld deposit, respectively.
*The numerical constants in the constitutive equations given below could alternatively be expressed in terms of
atomic weights etc. to bring out more clearly their physical significance (e.g. see the treatment of Bhadeshia and
Svensson83).
Ar - O2 gas mixtures
Ar - CO2 gas mixtures
Mole fraction
Fig. 2.68. Comparison between measured and predicted microslag composition in GMA welding of
C-Mn steels. Solid lines represent theoretical calculations based on equations (2-91) and (2-92). Data
from Grong and Christensen.1
Iron oxide in solution at 16000C

Oxygen in solution, wt%
Mn
Si
Ti
Al
Deoxidizer in solution, wt%
Fig. 2.69.Deoxidation equilibria in liquid steel at 1600°C.Data from Turkdogan.55
By rearranging equation (2-93), we get:
(2-94)
However, since data for acid soluble aluminium (and titanium), in practice, may contain
large inherent errors, the following restriction applies:
(2-95)
where [%O]anaL is the analytical weld metal oxygen content.
Titanium
Similarly, the average titanium content of the inclusions, [%Ti]incl, can be estimated from the
measured difference between total and acid soluble titanium, (A%Ti)weld. However, since TiN
dissolves readily in strong acid, it is necessary to include an empirical correction for the amount
of titanium nitride which simultaneously forms at the surface of the inclusions during solidifi-
cation. This can be done on the basis of published data for the solubility product of TiN in
liquid steel.84 If we assume that the nitrogen content of the inclusions is proportional to the
calculated difference between total and dissolved nitrogen at the melting point (15200C), the
following relationship is obtained:
(2-96)
where [%N]anai is the analytical weld metal nitrogen content, and [%Ti]soL is acid soluble
titanium. Note that the correction term for TiN, in practice, neither can be negative nor exceed
[%Ti]soh
In this case, the maximum amount of titanium which can be bound as Ti2O3 is determined
by the overall oxygen balance:
(2-97)
Sulphur
If the solubility of sulphur in solid steel is taken equal to 0.003 wt%,70 the average sulphur
content of the inclusions, [%S]ind is given by:
(2-98)
where [%S]anai is the analytical weld metal sulphur content.
Silicon and manganese

From the experimental data of Saggese et al.S5 reproduced in Fig. 2.70, it is evident that the
mass ratio between SiO2 and MnO in the oxide phase may be considered constant and virtually
independent of composition (i.e. equal to about 0.94). This implies that the average silicon
content of the inclusions, [%Si]ind, can be calculated from a balance of oxygen:
SiO2
MnO Al2O3
wt% AI2O3
Fig. 2.70. Measured inclusion compositions in low-alloy steel weld metals. Data from Saggese et alP
(2-99)
Considering manganese, proper adjustments should also be made for the amount of MnS
formed at the surface of the inclusions during solidification. Hence, the average manganese
content of the inclusions, [%Mn]inci, is given by the sum of the oxygen and the sulphur contri-
butions:
(2-100)
Experimental verification of model

In Fig. 2.71, the accuracy of the model has been tested against the experimental SA/GMA
inclusion data reported by Kluken and Grong,57 taking the sum (%A1 + %Ti + %S + %Si +
%Mn) equal to 100%. A closer inspection of the graphs reveals a reasonable agreement be-
tween calculated and measured average inclusion compositions in all cases, which confirms
that the model is sound. Included in Fig. 2.71 is also a collection of data for SMA low-alloy
steel weld metals (3 kJ mm"1 — basic electrodes). Since these results follow the same pattern,
it implies that the model is generally applicable and, therefore, can be adopted to all relevant
arc welding processes.
Aluminium
Sulphur
Measured composition, wt%
Marked symbols: SMAW
Calculated composition, wt%
Titanium
Manganese
Measured composition, wt%
Silicon
Marked symbols: SMAW
Calculated composition, wt%

Fig. 2.71. Comparison between measured and predicted inclusion compositions; (a) Aluminium and
sulphur, (b) Titanium, manganese and silicon. Data from Kluken and Grong57'86
Implications of model
It can be inferred from equations (2-94) and (2-95) that the chemical composition of the in-
clusion oxide core is directly related to the aluminium to oxygen ratio in the weld metal.
Referring to Fig. 2.72, the fraction of MnOAl2O3 and 7-Al2O3 in the inclusions is seen to in-
crease steadily with increasing (A%Al)wdJ[%O]anal ratios until the stoichiometric compo-
sition for precipitation of aluminium oxide is reached at 1.13. At higher ratios, the deoxida-
tion product will be pure Al2O3, since aluminium is present in an over-stoichiometric amount
with respect to oxygen. When titanium is added to the weld metal, titanium oxide (in the form
of Ti2O3) may also enter the reaction product. At the same time both TiN and a-MnS form
epitaxially on the surface of the inclusions during solidification. Consequently, in Al-Ti-Si-Mn
deoxidised steel weld metals the total number of constituent phases within the inclusions may
approach six. The kinetics of inclusion formation are further discussed in Ref. 87.
SiO2
MnO
Fig. 2.72. Coexisting phases in inclusions at different weld metal aluminium-to-oxygen ratios. Shaded
region indicates the approximate composition range for the oxide phase. The diagram is constructed on
the basis of the model of Kluken and Grong57 and relevant literature data.
Example (2.13)
Consider SA welding of low-alloy steel with two different basic fluxes. Data for the weld
metal chemical compositions are given in Table 2.12. Based on Fig. 2.72, estimate the total
number of constituent phases in the inclusions in each case.
Solution
It follows from Fig. 2.72 that the chemical composition of the deoxidation product is deter-
mined by the aluminium to oxygen ratio in the weld metal. For weld A, we have:
Since this ratio is higher than the stoichiometric factor of 1.13, all oxygen is probably tied up
as aluminium oxide. In addition, weld A contains small amounts of titanium and sulphur,
which may give rise to precipitation of TiN and MnS at the surface of the inclusions during
solidification. Hence, the three major constituent phases in the weld metal inclusions are
7-Al2O3, TiN, and ot-MnS.
Table 2.12 Chemical composition of SA steel weld metals considered in Example 2.13.
Weld C O Si Mn S N Al* Ti*

No. (wt%) (wt%) (wt%) (wt%) (wt%) (wt%) (wt%) (wt%)
A 0.09 0.021 0.45 1.52 0.01 0.006 0.028 0.010

(0.003) (0.009)
B 0.09 0.045 0.45 1.52 0.01 0.006 0.028 0.028
(0.003) (0.010)
*Data for acid soluble Al and Ti are given by the values in brackets.
In the case of weld B the situation is much more complex due to a higher content of oxygen
and titanium. From Table 2.12, we have:
and
These calculations show that Al and Ti are not present in sufficient amounts to tie-up all oxy-
gen. Under such conditions Fig. 2.72 predicts that the total number of constituent phases in the
inclusions is six, i.e.: SiO2, MnOAl2O3, 7-Al2O3, Ti2O3, a-MnS and TiN.
References
1 0 . Grong and N. Christensen: Scand. J. Metall, 1983,12, 155-165.

2. J.F. Lancaster: Metallurgy of Welding, 3rd Edn, 1980, London, George Allen & Unwin Ltd.
3. J. Szekely and NJ. Themelis: Rate Phenomena in Process Metallurgy, 1971, New York, John
Wiley & Sons, Inc.
4. N. Christensen: Welding Metallurgy Compendium, 1985, University of Trondheim, The Nor-
wegian Institute of Technology, Trondheim, Norway.
5. E.T. Turkdogan: Physical Chemistry of Oxygen Steelmaking, Thermochemistry and Thermo-
dynamics, 1970, United States Steel Corporation.
6. F.R. Coe: Welding Steels without Hydrogen Cracking, 1973, Abington (Cambridge), The Welding
Institute.
7. Doc. IIS/IIW-532-77: Weld. World, 1977,15, 69-72.
8. B. Chew and R. A. Willgoss: Proc. Int. Conf Weld Pool Chemistry and Metallurgy, London,
April 1980, Paper 25,155-165. Publ. The Welding Institute (England).
9. F. Matsuda, H. Nakagawa, K. Shinozaki and H. Kihara: Trans. JWRI, 1978, 7, 135-137.
10. B. Chew: Weld. J., 1973, 52, 386s-391s.
11. GM. Evans: Schweissmitteilungen (Oerlikon), 1977, No. 79, 4-8.
12. GM. Evans and H. Bach: HW Doc. IIA-363-74, 1974.
13. R. Roux: Rev. Met. 1954,192-209.
14. E.M. Skjolberg: M.Sc. thesis, 1980, University of Trondheim, The Norwegian Institute of Tech-
nology, Trondheim, Norway.
15. D.G. Howden and D.R. Milner: Brit. Weld. J., 1963,10, 304-316.
16. T. Kobayashi, T. Kuwana and T. Kikuchi: HW Doc. XII-265-65, 1965.
17. J.E. O'Brien and M.F. Jordan: Met. Constr. Brit. Weld. J., 1971,3, 299-303.
18. M. Uda and S. Ohno: Trans. Nat. Res. Inst. for Metals, 1973,15, 20-28.
19. O. Morigaki, T. Tanigaki, M. Kuwabara and K. Fujibayashi: HW Doc. 11-746-75, 1975.
20. H.K.D.H. Bhadeshia, L.E. Svensson and B. Gretoft: J. Mater. ScL Lett, 1988, 7, 610-612.
21. E. Halm0y: Proc. Int. Conf. on Arc Physics and Weld Pool Behaviour, London, May, 1979,49-
57. Publ. The Welding Institute (England).
22. D.J.H. Corderoy, B. Wills and G.R. Wallwork: Proc. Int. Conf. on Weld Pool Chemistry and
Metallurgy, London, April 1980, 147-153. Publ. The Welding Institute (England).
23. J.F. Elliott, M.Gleiser and V. Ramakrishna: Thermochemistry for Steelmaking (Vol. II), 1963,
London, Addison-Wesley Publ. Company (Pergamon Press).
24. Doc. IIS/IIW-343-70: Weld. World, 1970, 8, 28-35.
25. I. Barin and O. Knacke: The Thermochemical Properties of Inorganic Substances, 1973, Ber-
lin, Springer Verlag.
26. Janaf, Thermochemical Tables - 2nd Edn (NSRDS); 1971, New York, National Bureau of
Standards.
27. R.F. Heile and D.C. Hill: Weld. J., 1975, 54, 201s-210s.
28. B. Ozturk and RJ. Fruehan: Metall Trans. B, 1985,16B, 801-806.
29. G. Jelmorini, G.W. Tichelaar and GJ.P.M. Van den Heuvel: HW Doc. 212-411-77, 1977.
30. O. Kubaschewski and C B . Alcock: Metallurgical Thermochemistry, 5th Edn, 1979, Oxford,
Pergamon Press.
31. PA. Distin, S.G. Whiteway and CR. Masson: Can. Metall. Quart., 1971,10, 13-18.
32. J.F. Lancaster: The Physics of Welding, 1984, Oxford, HW-Pergamon Press, 204-267.
33. WA. Fischer and J.F. Schumacher: Arch. Eisenhiittenwesen, 1978, 49, 431-435.
34. N. Christensen, V. de L. Davis and K. Gjermundsen: Brit. Weld. J., 1965,12, 54-75.
35. G.K. Sigworth and J.F. Elliott: Metall. Trans., 1973, 4,105-113.
36. D.N. Shackleton and A.A. Smith: HW Doc. XII-832-84, 1984.
37. 0. Grong, N.H. Rora and N. Christensen: Scand. J. Metall., 1984,13,154-156.
38. A.A. Smith et al: Weld. World., 1970, 8, 28-35.
39. A.A. Smith: CO2-Shielded Consumable Electrode Arc Welding, 2nd Edn, 1965, Brit. Weld.
Res. Assoc.
40. U. Lindborg: Met. Constr. and Brit. Weld. J., 1972, 4, 52-55.
41. T.H. North, H.B. Bell, A. Nowicki and I. Craig: Weld. J., 1978, 57, 63s-71s.
42. CS. Chai and T. W. Eagar: Metall. Trans., 1981,12B, 539-547.
43. U. Mitra and T.W. Eagar: Metall. Trans., 1984,15A, 217-227.
44. J.E. Indacochea, M. Blander, N. Christensen and D.L. Olson: Metall. Trans., 1985, 16B, 237-
245.
45. N. Christensen and 0 . Grong: Scand. J. Metall, 1986,15, 30-40.
46. U. Mitra and T.W. Eagar: Metall Trans., 1991, 22B, 65-100.
47. T.W. Eagar: Weld. J., 1978, 57, 76s-80s.
48. N. Christensen and J. Chipman: Weld. Res. Coun. Bull. Series, 1953, No. 15, New York, Weld-
ing Research Council (USA).
49. A. Apold: Carbon Oxidation in the Weld Pool, 1962, Oxford, Pergamon Press.
50. D.A. Wolstenholme: Proc. Int. Conf on Trends in Steel and Consumables for Welding, Lon-
don, 1978,123-134. Publ. The Welding Institute (England).
51. 0 . Grong, D.L. Olson and N. Christensen: Metal Constr., 1985,17, 810R-814R.
52. P. Nilles, P. Dauby and J. Claes: Proc. Int. Conf. Basic Oxygen Steelmaking —A New Technol-
ogy Emerges, London, 1978, 60-72, The Metals Society (England).
53. U. Lindborg and K. Torsell: Trans. TMS-AIME, 1968, 242, 94-102.
54. N.F. Grevillius: Jernkont. Ann., 1969,153, 547-572.
55. E.T. Turkdogan: Proc. Int. Conf. on Chem. Metall. of Iron and Steel, Sheffield, July 1971,153-
170, Publ. The Iron and Steel Institute (England).
56. 0 . Grong. T.A. Siewert, G.P. Martins and D.L. Olson: Metall. Trans. A, 1986,17A, 1797-1807.
57. A.O. Kluken and 0 . Grong: Metall. Trans. A, 1989, 2OA, 1335-1349.
58. M.L. Turpin and J.F. Elliott: J. Iron Steel Inst., 1966, 204, 217-225.
59. E.T. Turkdogan: Physicochemical Properties of Molten Slags and Glasses, 1983, London, The
Metals Society.
60. E. Plockinger and M. Wahlster: Stahl und Eisen, 1960, 80, 659-669.
61. BJ. Bradstreet: HW Doc. 212-138-68, 1968.
62. J.F. Lancaster: Phys. Technol, 1984,15, 73-79
63. F. Eickhorn and A. Engel: HW Doc. 212-201-70, 1970.
64. N. Mori and Y. Horii: HW Doc. 212-188-70, 1970.
65. R.A. Walsh and S. Ramachandran: Trans. TMS-AIME, 1963, 227, 560-562.
66. D.C. Hilty and W. Crafts: Trans.-AIME, 1950,188, 425-436.
67. E.T. Turkdogan: Trans. TMS-AIME, 1965, 233, 2100-2112.
68. G.M. Evans: HW Doc. IIA-630-84, 1984.
69. 0 . Grong and D.K Matlock: Int. Met. Rev., 1986, 31, 27-48.
70. A.G. Franklin: J. Iron Steel Inst., 1969, 207, 181-186.
71. A.O. Kluken, 0. Grong and J. Hjelen: Mat. ScL Technol., 1988, 4, 649-654.
72. L.M. Hocking: Quart. J. Royal Meterol. Soc, 1959, 85, 44-50.
73. C. Wagner: Z Electrochemie, 1961, 65, 581-591.
74. E.E. Underwood: Quantitative Stereology, 1970, London, Addison-Wesley Publ. Co.
75. M.F. Ashby and R. Ebeling: Trans. TMS-AIME, 1966, 236, 1396-1404.
76. R. Kiessling: Non-Metallic Inclusions in Steel, 1978, London, The Metals Society (TMS).
77. H. Homma, S. Ohkita, S. Matsuda and K. Yamamoto: Weld. J., 1987, 66, 301s-309s.
78. G. Thewlis: HW Doc. IIA-736-88, 1988.
79. J.M. Dowling, J.M. Corbett and H.W. Kerr: Metall. Trans. A, 1986,17A, 1611-1623.
80. G.M. Evans: Metal Constr., 1986,18, 631R-636R.
81. J.L. Murray and H.A. Wriedt: Bull. Alloy Phase Diagr., 1987, 8, 148-165.
82. LD. Sommerville, I. Ivanchev and H.B. Bell: Proc. Int. Conf Chem. Metall of Iron and Steel,
Sheffield, July 1971, 23-25, Publ. The Iron and Steel Inst. (1973).
83. H.K.D.H. Bhadeshia and L.E. Svensson: Mathematical Modelling of Weld Phenomena (Eds
H. Cerjak and K.E. Easterling), 1993, London, The Institute of Materials, 109-180.
84. S. Matsuda and N. Okumura: Trans. ISIJ, 1978,18, 198-205.
85. M.E. Saggese, A.R. Bhatti, D.N. Hawkins and J.A. Whiteman: Proc. Int. Conf on the Effect of
Residual, Impurity and Micro-Alloy ing Elements on Weldability and Weld Properties, London,
Nov. 1983, Paper 15, Publ. The Welding Institute (England).
86. A.O. Kluken and 0 . Grong: Report No. STF34 F87125, 1987, Sintef, Trondheim, Norway.
87. S.S. Babu, S.A. David, J.M. Vitek, K. Mundra and T. DebRoy: Mater. ScL Technol., 1995,11,186-
199.
Appendix 2.1
Nomenclature
contact area (mm2) activity coefficient of
element X
difference between
total and acid soluble Al in gravity constant
weld metal (wt%) (9.81ms- 2 )
thermal diffusivity standard free energy of

(mm2 s 1 ) reaction (J mol"1 or
kJ mol-1)
activity of arbitrary slag
component energy barrier for homoge-
neous nucleation
cross section of fused (kJ mol-1)
parent metal (mm2)
driving force for precipita-
flux basicity index tion of oxide inclusions
(J nr 3 )
kinetic constants
drag coefficient
GTAW gas tungsten arc welding
molar concentration of
element X in the liquid standard enthalpy of
(mol m~3) reaction (J mol-1 or kJ
mol-1)
cross section of deposited
metal (mm2) hydrogen content related
to deposited metal (ml per
element diffusivity in liquid 10Og deposit)
phase (m2 s"1 or mm2 s"1)
hydrogen content related
diffusivity of element X in to fused metal (ml per 10Og
gas phase (mm2 s"1) or g per ton)
gross heat input (kJ mm"1) hydrogen content related

to French practice (g per
FCAW flux cored arc welding ton)
drag force acting on heat content per unit

spherical particle in relative volume at the melting point
motion to a fluid (N) (J mm"3)
gravity force acting on a hydrogen content related

spherical particle in relative to Japanese practice (ml per
motion to a fluid (N) 10Og deposit)
fume formation rate amperage (A)

(mg miir 1 )
nucleation rate of oxide total mass of liquid metal
inclusions in the weld pool leaving/entering the
(nuclei nr 3 s"1) reaction zone per unit time
(g s-1)
constant in equation (A2-1)
(nuclei rrrV 1 ) arbitrary flux or slag
component
ratio between absorbed Si
and Mn in the weld metal concentration displacement
of manganese referred to
coefficient of weld metal different standard states
deposition (g A-1S"1) (wt%)
mass transfer coefficient for total manganese absorption

gas absorption (mm s"1) in the weld metal due to
slag/metal interactions
mass transfer coefficient for (wt%)
gas desorption (mm s"1)
rejected manganese during
overall mass transfer cooling in the weld pool
coefficient (mm s"1) (wt%)
equilibrium constants oxidation loss of manga-

nese at electrode tip (wt%)
the product [%C][%O]
atomic weight of element X
mass of CO2 per 100 g of
electrode coating (g or dimensionless operating
wt%) parameter
total mass of inclusions per total number of moles of

100 g weld deposit (g) component /
mass of H2O per 100 g of absorption/desorption rate

electrode coating (g or of element X in the weld
wt%) pool (mol s"1)
total weight of retained total number of moles of

MnOSiO2 (g per 100 g weld component X
deposit)
total number of moles of
total weight of precipitated component X2
microslag (in g per 100 g
weld deposit) number of particles per
unit area (mm"2)
total weight of precipitated
MnS (g per 100 g weld Avogadro constant (6.022
deposit) X 1023 mor 1 )
mole fraction of arbitrary combined partial pressure
slag component of H2 and H2O (atm)
analytical weld metal post weld heat treatment

nitrogen content (wt%)
parts per million (g per ton)
mole fraction of CO in gas
phase net arc energy (W)
Reynolds number universal gas constant

(J K-1 mol-1)
number of particles per unit
volume (No. per mm"3) relative humidity (%)
calculated oxygen absorp- analytical weld metal

tion in electrode tip or sulphur content (wt%)
falling droplets (wt%)
standard entropy of
analytical weld metal reaction (J K -1 mol"1)
oxygen content (wt%)
concentration displacement
of silicon referred to
equilibrium weld metal different standard states
oxygen content (wt%) (wt%)
rejected oxygen during total silicon absorption in

cooling in the weld pool the weld metal due to slag/
(wt%) metal interactions (wt%)
total oxygen absorption in rejected silicon during

hot part of weld pool (wt%) cooling in the weld pool
(wt%)
total pressure (atm or bar)
oxidation loss of silicon at
partial pressure of compo- electrode tip (wt%)
nent X in bulk gas phase
(atm) solubility of element X at 1
atm total pressure (ml per
equilibrium partial pressure 10Og, ppm or wt%)
of component X at gas/
metal interface (atm) total particle surface area
per unit volume (mm2 per
partial pressure of compo- mm3)
nent X2 in bulk gas phase
modified solubility of
(atm)
hydrogen at 1 atm total
pressure (ml per 100 g, ppm
equilibrium partial pressure
or wt%)
of component X2 at gas/
metal interface (atm)
SAW submerged arc welding volume fraction of
inclusions
SMAW shielded metal arc welding
concentration of compo-
STEM scanning transmission nent / (vol%)
electron microscope
concentration of compo-
time (s) nent X2 (vol%)
time constant (s) concentration of

component YX2 (vol%)
acid soluble titanium in
weld metal (wt%) WDX wavelength dispersive X-
ray analysis
difference between total
and acid soluble Ti in weld WFR wire feed rate (mm s -1 or
metal (wt%) m min"1)
temperature (0C or K) arbitrary element or

gaseous species
ambient temperature (0C or
K) weld ripple lag (mm)
preheating temperature (0C concentration of element X

or K) in the weld metal (wt%)
reference temperatures (0C concentration displacement

or K) of element X referred to
nominal weld metal
rising velocity of ascending composition (wt%)
particles relative to the
liquid (|jLms-1) equilibrium concentration
of element X in the weld
bulk velocity of the fluid pool (wt%)
relative to the particles
(ms- 1 ) initial concentration of
element X in the weld pool
voltage (V) (wt%)
welding (travel) speed average content of element

(mm s"1) X in inclusions (wt%)
volume of melt (mm3) nominal concentration of

element X in weld metal
molar volume of nucleus (wt%)
(m3 mol-1)
concentration of element X
molar volume of oxide per in base plate (wt%)
mole of the diffusate
(m3 mol-1)
concentration of element X thickness of stagnant
in filler wire (wt%) gaseous boundary layer
(mm)
arbitrary element or
gaseous species wetting angle
density (kg irr 3 or mean particle centre to

g mm"3) centre volume spacing
(jxm)
difference in density
between liquid steel and dimensionless y-coordinate
inclusion (kg irr 3 or at maximum width of
g mm"3) isotherm
slag/metal interfacial activity coefficient for

energy (J nr 2 ) MnO in slag phase
arc efficiency factor activity coefficient for

SiO2 in slag phase
viscosity (kg nr 1 s"1)
Appendix 2.2
Derivation of equation (2-60)
The nucleation rate / as a function of temperature can be expressed as:
(A2-1)
where J0 is a constant (with the unit nuclei per m3 and s) and AG* is the energy barrier for
nucleation.
By rearranging equation (A2-1) and inserting reasonable values for J and J0 for the specific
case of homogeneous nucleation of oxide inclusions in liquid steel,55 we obtain:
(A2-2)
From the classic theory of homogeneous nucleation AG* is given by:
(A2-3)
where NA is the Avogadro constant, o is the interfacial energy between the nucleus and the
liquid (in J nr 2 ) and AGV is the driving force for the precipitation reaction (in J nr 3 ).
The parameter AGV can be expressed as:
(A2-4)
where AH° and AS° are the standard entalpy and entropy of the precipitation reaction, respec-
tively and Vm is the molar volume of the nucleus (in m3 mol"1).
It is evident from Fig. 2.48 that AGV = 0 when T=Tu which gives \S° = A//7 T\. Hence,
equation (A2-4) may be rewritten as:
(A2-5)
By combining equations (A2-2), (A2-3) and (A2-5), we obtain the following relationship
between T\ and T^-
(A2- 6)
3
Solidification Behaviour of Fusion Welds
3.1 Introduction
Inherent to the welding process is the formation of a pool of molten metal directly below the
heat source. The shape of this molten pool is influenced by the flow of both heat and metal,
with melting occurring ahead of the heat source and solidification behind it. The heat input
determines the volume of molten metal and, hence, dilution and weld metal composition, as
well as the thermal conditions under which solidification takes place. Also important to solidi-
fication is the crystal growth rate, which is geometrically related to weld travel speed and weld
pool shape. Hence, weld pool shape, weld metal composition, cooling rate, and growth rate
are all factors interrelated to heat input which will affect the solidification microstructure.
Some important points regarding interpretation of weld metal microstructure in terms of these
four factors will be discussed below.
Since the properties and integrity of the weld metal depend on the solidification microstruc-
ture, a verified quantitative understanding of the weld pool solidification behaviour is essen-
tial. At present, our knowledge of the chemical and physical reactions occurring during solidi-
fication of fusion welds is limited. This situation arises mainly from a complex sequence of
reactions caused by the interplay between a number of variables which cannot readily be ac-
counted for in a mathematical simulation of the process. Nevertheless, the present treatment
will show that it is possible to rationalise the development of the weld metal solidification
microstructure with models based on well established concepts from casting and homogenising
treatment of metals and alloys.
3.2 Structural Zones in Castings and Welds
During ingot casting, three different structural zones can generally be observed, as shown
schematically in Fig. 3.1. The chill zone is produced by heterogeneous nucleation in the re-
gion adjacent to the mould wall as a result of the pertinent thermal undercooling. These grains
rapidly become dendritic, and dendrites having their <100> direction (preferred easy growth
direction for cubic crystals) parallel to the maximum temperature gradient in the melt will
soon outgrow those grains that do not have this favourable orientation. Competitive growth
occurring during the initial stage of the solidification process leads to an alignment of the
crystals in the heat flow direction and eventually to the formation of a columnar zone. 12 Fi-
nally, an equiaxed zone may develop in the centre of the casting, mainly as a result of growth
of detached dendrite arms within the remaining, slightly undercooled liquid.
A similar situation also exists in welding, as indicated in Fig. 3.2 However, in this case the
chill zone is absent, since the partly melted base metal grains at the fusion boundary act as seed
crystals for the growing columnar grains.3 In addition, the growth direction of the columnar
Shrinkage pipe
Chill zone
Columnar zone
Equiaxed zone
Mould
Fig. 3.1. Transverse section of an ingot showing the chill zone, the columnar zone and the equiaxed zone
(schematic).
grains will change continuously from the fusion line towards the centre of the weld due to a
corresponding shift in the direction of the maximum temperature gradient in the weld pool.
This change in orientation may result in a curvature of the columnar grains (Fig. 3.2(a)). Alter-
natively, new grains can nucleate and grow in a columnar manner, producing a so-called 'stray'
structure as shown schematically in Fig. 3.2(b). Finally, if the conditions for nucleation of new
grains are favourable, an equiaxed zone will form near the weld centreline similar to that
observed in ingots or castings (see Fig. 3.2(b)).
Although the process of weld pool solidification is frequently compared with that of an
ingot in 'miniature', a number of basic differences, already mentioned, exist which strongly
influence the microstructure and properties of the weld metal. Of particular importance is also
the disparity in cooling rate between a fusion weld and an ingot (see Fig. 3.3). For conven-
tional processes such as shielded metal arc (SMA), gas metal arc (GMA), submerged arc (SA)
or gas tungsten arc (GTA) welding the cooling rate may vary from 10 to 103 0 C s"1, while for
modern high energy beam processes such as electron beam (EB) and laser welding the cooling
rate is typically of the order of 103 to 106 0 C s"1.4 Consequently, to appreciate fully the impli-
cations of these differences in general solidification behaviour between a weld pool and an
ingot, it is necessary to consider in detail the sequence of events taking place in the solidifying
weld metal beginning with the initiation of crystal growth at the fusion boundary.
3.3 Epitaxial Solidification
It is well established that initial solidification during welding takes place epitaxially, where the
partly melted base metal grains at the fusion boundary act as seed crystals for the columnar
grains. This process is illustrated schematically in Fig. 3.4.
(a)
Welding direction
(b)
Columnar zone
Equiaxed zone Welding direction
Columnar zone •
Fig. 3.2. Examples of structural zones in fusion welds (schematic); (a) Curved columnar grains,
(b) Stray grain structure.
Cooling rate, °C/s
Rapid solidification technology

Electron beam welding
Laser welding
GMAW, GTAW
SMAW, SAW,
Process
Fig. 3.3. Disparity in cooling conditions between casting, welding and rapid solidification.
boundary
Fusion
HAZ Weld metal
Fig. 3.4. Schematic illustration showing epitaxial growth of columnar grains from partly melted base
metal grains at fusion boundary.
Liquid (L)
Substrate (S)
Fig. 3.5. Schematic representation of heterogeneous nucleation.

3.3.1 Energy barrier to nucleation
During epitaxial solidification, a solid embryo (nucleus) of the weld metal forms at the melted-
back surface of the base metal grain. Assuming that the interfacial energy between the embryo
and the liquid is isotropic, it can be shown, for a given volume of the embryo, that the interfa-
cial energy of the whole system is minimised if the embryo has the shape of a spherical cap.
Under such conditions, the following relationship exists between the interfacial energies (see
Fig. 3.5):
(3-1)
where (3 is the wetting angle.

The change in free energy, AGhet, accompanying the formation of a solid nucleus with this
configuration is given by:5
(3-2)
where VE is the volume of the solid embryo, AGV is the free energy change associated with the
embryo formation, AEL and AES are the areas of the embryo-liquid and embryo-substrate inter-
faces, respectively, and/(P) is the so-called shape factor, defined as:
(3-3)
The critical radius of the stable nucleus, r / , is found by differentiating equation (3-2) with
respect to rs and equating to zero:
(3-4)
By substituting equation (3-4) into equation (3-2), we obtain the following expression for
the energy barrier to heterogeneous nucleation (AG^ r ):
(3-5)
where AHm is the latent heat of melting, Tm is the melting point, and AJT is the undercooling.
It is easy to verify that the first term in equation (3-5) is equal to the energy barrier to
homogeneous nucleation, AGôm. Hence, we may write:
(3-6)
Equation (3-6) shows that AG^ is a simple function of the wetting angle O). Since the
chemical composition and the crystal structure of the two solid phases are usually very similar,
we have:6
Under such conditions equation (3-1) predicts that the wetting angle 3 ~ 0 (cos(3 ~ 1),
which implies that there is a negligible energy barrier to solidification of the weld metal (№}*het
~ 0), i.e. no undercooling of the melt is needed, and solidification occurs uniformly over the
whole grain of the base metal. This is in sharp contrast to conventional casting of metals and
alloys where some undercooling of the melt is always required to overcome the inherent en-
ergy barrier to solidification (see Fig. 3.6).
3.3.2 Implications of epitaxial solidification

Since the initial size of the weld metal columnar grains is inherited directly from the grain
growth zone adjacent to the fusion boundary, the solidification microstructure depends on the
grain coarsening behaviour of the base material. This is particularly a problem in high energy
processes such as submerged arc and gas metal arc welding, where grain growth of the base
metal can be considerable. In such cases the size of the columnar grains at the fusion boundary
will be correspondingly coarse, as indicated by the data in Fig. 3.7.
Moreover, during multipass welding the columnar grains can renucleate at the boundary
between for instance the first and the second weld pass and subsequently grow across the
entire fusion zone, as illustrated in Fig. 3.8. This type of behaviour is usually observed in
weldments which do not undergo transformations in the solid state (e.g. aluminium, certain
titanium alloys, stainless steel etc.). In practice, the problem can be eliminated by additions of
inoculants via the filler wire, which facilitates a refinement of the columnar grain structure
through heterogeneous nucleation of new (equiaxed) grains ahead of the advancing interface
(to be discussed later).
AG
r
s
Casting
Welding Homogeneous
nucleation
Fig. 3.6. The free energy change associated with heterogeneous nucleation during casting and weld
metal solidification, respectively (schematic). The corresponding free energy change associated with
homogeneous nucleation is indicated by the broken curve in the graph.
Fusion
line
prior austenite grain size (jim)
HAZ Weld
Weld metal
metal
HAZ prior austenite grain size (jim)
Fig. 3.7. Correlation between HAZ prior austenite grain size at the fusion boundary and the correspond-
ing weld metal prior austenite grain size. Data from Grong et al?
2. pass
1. pass
HAZ
Base metal
Fig. 3.8. Optical micrograph showing renucleation of columnar grains during multipass GMA welding
of a P-titanium alloy.
3.4 Weld Pool Shape and Columnar Grain Structures
Growth of the columnar grains always proceeds closely to the direction of the maximum ther-
mal gradient in the weld pool, i.e. normal to the fusion boundary. This implies that the colum-
nar grain morphology depends on the weld pool geometry.
3.4.1 Weld pool geometry

The weld pool geometry is a function of the welding speed and the balance between the heat
input and the cooling conditions, as influenced by the base plate thermal properties. At pseudo-
steady state, these conditions establish a dynamic equilibrium between heat supply and heat
extraction so that the shape of the weld pool remains constant for any given speed. Following
the treatment in Chapter 1, the weld pool geometry depends on the dimensionless operating
parameter n3, defined as:
(3-7)
where qo is the net arc power, v is the welding speed, a is the thermal diffusivity of the base
plate, and Hm-Ho is the heat content per unit volume at the melting point.
As shown in Fig. 3.9(a), a tear-shaped weld pool is favoured by a high n3 value, which is
characteristic of fast moving high power sources. In contrast, at a low arc power and a low
welding speed the shape of the weld pool becomes more elliptical because of a shift in the
mode of heat flow (see Fig. 3.9(b)). Note, however, that the thermal properties of the base
metal is also of importance in this respect, since the n3 parameter is a function of both a and
Hm-Ho. Consequently, a tear-shaped weld pool is usually observed in weldments of a low
thermal diffusivity (e.g. austenitic stainless steel), whereas an elliptical or spherical weld pool
is more likely to form during aluminium welding owing to the resulting higher thermal diffu-
sivity of the base metal.
In addition to the factors mentioned above, the geometry of the weld pool is also affected by
convectional heat transfer due to the presence of buoyancy, electromagnetic or suface tension
gradient forces. Recently, attempts have been made to include such effects in heat flow mod-
els for welding.8"11 Referring to Fig. 3.10(a) the buoyancy force will promote the formation of
a shallow, wide weld pool because of transport of 'hot' metal to the surface and 'cold' metal to
the bottom of the pool. In the presence of the electromagnetic force the flow pattern is re-
versed, since the latter force will tend to push the liquid metal in the central part of the pool
downward to the root of the weld. This makes the weld pool deeper and more narrow, as
shown in Fig. 3.10(b).
Moreover, it is generally accepted that surface tension gradients can promote circulation of
liquid metal within the weld pool from the region of low surface tension to the region of higher
surface tension.9 In the absence of surface active elements such as oxygen and sulphur, the
surface tension decreases with increasing temperature as illustrated in Fig. 3.10(c), which forces
the metal to flow outwards towards the fusion boundary. This results in the formation of a
relatively wide and shallow weld pool. However, if oxygen or sulphur is present in sufficient
quantities a positive temperature coefficient of the surface tension may develop, which facili-
tates an inward fluid flow pattern and an increased weld penetration (see Fig. 3.10(d)). The
important influence of surface active elements on the resulting bead morphology is well docu-
HAZ isotherms
Fusion boundary
V
Weld
pool
(a)
HAZ isotherms
Fusion boundary
V
Weld
pool
(b)
Fig. 3.9. Theoretical shape of fusion boundary and neighbouring isotherms under different operational
conditions; (a) High n3 values, (b) Low ^-values.
merited for ordinary GTA austenitic stainless steel welds. 1 2 1 3 The indications are that such
effects become even more important under hyperbaric welding conditions. 14
3.4.2 Columnar grain morphology

It is evident from the above discussion that a change in the weld pool geometry, caused by
variations in the operational conditions, may strongly alter the weld metal solidification micro-
structure. In fact, more than nine different grain morphologies have been observed during
fusion welding.15 The two most important are shown in Fig. 3.11. Referring to Fig. 3.11(a) a
spherical or elliptical weld pool will reveal curved and tapered columnar grains owing to a
shift in the direction of the maximum thermal gradient in the liquid from the fusion boundary
towards the weld centre-line. In contrast, a tear-shaped weld pool yields straight and broad
Electrode
Arc
Weld pool
(a)
Electrode
Arc
Weld pool
(b)
Fig. 3.10. Schematic diagrams illustrating the major fluid flow mechanisms operating in a weld pool;
(a) Buoyancy force (b) Electromagnetic force.
columnar grains as shown in Fig. 3.11(b), since the direction of the maximum temperature
gradient in the melt does not change significantly during the solidification process. The lat-
ter condition is known to promote formation of centre-line cracking because of mechanical
entrapment of inclusions and enrichment of eutectic liquid at the trailing edge of the weld
pool.
3.4.3 Growth rate of columnar grains

The growth rate of the columnar grains is geometrically related to the weld travel speed and
the weld pool shape.
3.4.3.1 Nominal crystal growth rate

Since the shape of the weld pool remains constant during steady state welding, the growth rate
of the columnar grains must vary with position along the fusion boundary. This point is more
clearly illustrated in Fig. 3.12 which shows a sketch of a single columnar grain growing paral-
lel with the steepest temperature gradient in the weld pool. Taking the angle between the
Surface tension
Temperature
Electrode
Arc
Weld pool
(C)
Surface tension
Temperature
Electrode
Arc
Weld pool
(d)
Fig. 3.10. Schematic diagrams illustrating the major fluid flow mechanisms operating in a weld pool
(continued); (c) Surface tension gradient force (negative gradient); (d) Surface tension gradient force
(positive gradient).
Heat source v
(a)
Heat source v
(b)
Fig. 3.11. Schematic comparison of columnar grain structures obtained under different welding condi-
tions; (a) Elliptical weld pool (low n3 values), (b) Tear-shaped weld pool (high n3 values). Open arrows
indicate the direction of the maximum temperature gradient in the weld pool.
Fusion boundary
Heat source
Crystal
Fig. 3.12. Definition of the nominal crystal growth rate RN.
growth direction and the welding direction equal to a, the steady state growth rate, R N , be-
comes:
(3-8)
where v is the welding speed.

Considering spherical or elliptical weld pools, the nominal crystal growth rate is lowest at
the edge of the weld pool (a—>90°, cosa-^0) and highest at the weld centre-line where R N
approaches v (a->0, c o s a ^ l ) . In contrast, columnar grains trailing behind a tear-shaped weld
pool will grow at an approximately constant rate which is significantly lower than the actual
welding speed (a » 0), since the direction of the maximum temperature gradient in the weld
pool does not change during the solidification process. This is also in agreement with practical
experience (see Fig. 3.13).
(a)
Nominal growth rate (RN), mm/s Niobium (1 mm plate thickness)
Relative position from edge of weld pool (%)•
(b)
Stainless steel (1 mm plate thickness)
Nominal growth rate (RN), mm/s
Equiaxed zone
Relative position from edge of weld pool (%)
Fig. 3.13. Measured crystal growth rates in thin sheet electron beam welding; (a) Niobium, (b) Stainless
steel. Data from Senda et al.16
Example (3.1)
Consider electron beam (EB) welding of a lmm thin sheet of austenitic stainless steel under
the following conditions:
Estimate on the basis of the Rosenthal thin plate solution (equation 1-83) the steady state
growth rate of the columnar grains trailing the weld pool.
Solution
The contour of the fusion boundary can be calculated from the Rosenthal thin plate solution
according to the procedure shown in Example (1.10). If we include a correction for the latent
heat of melting, the QbZn3 ratio at the melting point becomes:
Substitution of the above value into equation (1-83) gives the fusion boundary contour
shown in Fig. 3.14. It is evident from Fig. 3.14 that the weld pool is very elongated under the
prevailing circumstances due to a constrained heat flow in the ^-direction. This implies that the
angle a will not change significantly during the solidification process. Taking a as an average,
equal to about 70°, the steady-state crystal growth rate R N becomes:
This value is in reasonable agreement with the measured crystal growth rates in Fig. 3.13(b).
3.4.3.2 Local crystal growth rate

Equation (3-8) does not take into account the inherent anisotropy of crystal growth. For fac-
eted materials the dendrite growth directions are always those that are 'capped' by relatively
slow-growing (usually low-index) crystallographic planes.1 Figure 3.15 shows examples of
faceted cubic crystals delimited by {100} and {111} planes, respectively. If the {111} planes
are the slowest growing ones, the {100} planes will grow out, leaving the {111} facets and a
new crystal growing in the <100> directions as shown schematically in Fig. 3.15(b).
Although most metals and alloys do not form faceted dendrites, the anisotropy of crystal
growth is still maintained during solidification.2 In fact, experience has shown that the major
dendrite growth direction is normally the axis of a pyramid whose sides are the most closely
packed planes with which a pyramid can be formed.1 These directions are thus <100> for
body- and face-centred cubic structures, < 1010 > for hexagonal close-packed structures, and
<110> for body-centred tetragonal structures.
Because of the existence of preferred growth directions, the local growth rate of the crystals
RL will always be higher than the nominal growth rate R N defined in equation (3-8). Consider
now a cubic crystal which grows along the steepest temperature gradient in the weld pool, as
shown schematically in Fig. 3.16. If § denotes the angle between the interface normal and the
<100> direction, the following relationship exists between RN and RL:
-y(mm)
Columnar zone Heat source
Equiaxed zone
+x(mm)
Columnar zone
Fusion boundary
+yjmm)
Fig. 3.14. Predicted shape of fusion boundary during electron beam welding of austenitic stainless steel
(Example (3.1)).
(a)
(b)
Fig. 3.25.Examples of faceted cubic crystals; (a) Crystal delimited by {100} planes, (b) Crystal delim-
ited by {111} planes.
Columnar grain
Welding direction (x)
Fig. 3.16. Definition of the local crystal growth rate RL.

Tip temperature, 0C
Liquidus temperature
Tip velocity, mm/s-

Fig. 3.17. Calculated dendrite tip temperature vs dendrite growth velocity for an Fe-15Ni-15Cr alloy.
The undercooling of the dendrite tip is given by the difference between the liquidus temperature and the
solid curve in the graph. Data from Rappaz et alP
(3-9)
which gives:
(3-10)
Equation (3-10) shows that the local growth rate increases with increasing misalignment of
the crystal with respect to the direction of the maximum temperature gradient in the weld pool.
Since such crystals cannot advance without a corresponding increase in the undercooling ahead
of the solid/liquid interface (see Fig. 3.17), they will soon be outgrowed by other grains which
have a more favourable orientation. Fusion welds of the fee and bcc type will therefore de-
velop a sharp <100> solidification texture in the columnar grain region, similar to that docu-
mented for ingots and castings. The weld metal columnar grains may nevertheless be sepa-
rated by 'high-angle' boundaries, as shown in Fig. 3.1&, due to a possible rotation of the grains
in the plane perpendicular to their <100> length axes.
Example (3.2)
Consider electron beam welding of a 2mm thick single crystal disk of Fe-15Ni-15Cr under
The orientation of the disk with respect to the beam travel direction is shown in Fig. 3.19.
Calculate on the basis of the minimum velocity (undercooling) criterion the growth rate of
the dendrites trailing the weld pool under steady state welding conditions (assume 2-D
heat flow). Make also schematic drawings of the solidification microstructure in different
sections of the weld. Relevant thermal properties for the Fe-15Ni-15Cr single crystal are
given below:
Solution
Since the base metal is a single crystal, separate columnar grains will not develop. Neverthe-
less, under 2-D heat flow conditions growth of the dendrites can occur both in the [100] and
the [010] (alternatively the [010]) direction. Referring to Fig. 3.20 the growth rate of the
[100] and the [010] deridrites is given by:
and
Fig. 3.18. Spatial misorientation between two columnar grains growing in the <100> direction (sche-
matic).
Heat source
Weld
Fig. 3.19. Orientation of the single crystal Fe-15Ni-15Cr disk with respect to beam travel direction (Ex-
ample (3.2)).
Welding direction
Fig. 3.20. Schematic diagram showing the pertinent orientation relations between the fusion boundary
interface normal and the dendrite growth directions (Example (3.2)).
From this it is seen that the velocity of the [100] dendrites is always equal to that of the heat
source v. In contrast, the growth rate of the [010] dendrites depends both on v and a, and will
therefore vary with position along the fusion boundary. It follows from minimum velocity
criterion that the [100] dendrites will be selected when the interface normal angle a is less than
45°, while the [010] dendrites will develop at larger angles. This is shown graphically in Fig.
3.21.
At pseudo-steady state the fusion boundary can be calculated from the Rosenthal thin plate
solution (equation (1-83)) according to the procedure shown in Example (1.10). If we include
dendrites
/V
hk ,
R
a, degrees
Fig. 3.21.Normalised minimum dendrite tip velocity vs interface normal angle a (Example 3.2)).
a correction for the latent heat of melting, the QbIn3 ratio at the melting point becomes:
Substitution of this value into equation (1-83) gives the fusion boundary contour shown in
Fig. 3.22(a). Included in Fig. 3.22 are also schematic drawings of the predicted solidification
microstructure in different sections of the weld.
The results in Fig. 3.22 should be compared with the reconstructed 3-D image of the solidi-
fication microstructure in Fig. 3.23, taken form Rappaz et al.17 Due to partial heat flow in the
z-direction, [001] dendrite trunks will also develop. Nevertheless, these data confirm the gen-
eral validity of equations (3-8) and (3-10) relating crystal growth rate to welding speed and
weld pool shape.
3.4.4 Reorientation of columnar grains

In principle, there are two different ways a columnar grain can adjust its orientation during
solidification in order to accommodate a shift in the direction of the maximum temperature
gradient in the weld pool, i.e.:
(i) Through bowing

(ii) Through renucleation.
(a) -y (mm)
dendrites Heat source
dendrites +x (mm)
dendrites
+y (mm)
Fusion boundary
(b) Fusion zone (4.4 mm)
Base plate
dendrites dendrites dendrites
Fig. 3.22. Schematic representation of the weld metal solidification micro structure (Example 3.2));
(a) Top view of fusion zone, (b) Transverse section of fusion zone.
3.4.4.1 Bowing of crystals

A continuous change in the crystal orientation due to bowing will result in curved columnar
grains, as shown previously in Fig. 3.2(a). This type of grain morphology has been observed
in for instance electron beam welded aluminium and iridium alloys.34 Normally, the adjust-
ment of the crystal orientation is promoted by multiple branching of dendrites present within
the grains. Alternatively, the reorientation can be accommodated by the presence of surface
defects at the solid/liquid interface, e.g. screw dislocations, twin boundaries, rotation bounda-
ries, etc. The latter process presumes, however, a faceted growth morphology, and is therefore
of minor interest in the present context.
Example (3.3)
Consider a curved columnar grain of iridium which grows from the fusion boundary towards
the weld centre-line along a circle segment of length L, as shown schematically in Fig. 3.24.
Based on the assumption that the bowing is accommodated solely by branching of [010] dendrites
in the [100] direction, calculate the maximum local growth rate of the crystal during solidifica-
tion.
y
X
2
Fig. 3.23. Reconstructed 3-D image of solidification microstructure in an electron beam welded Fe-15Ni-
15Cr single crystal. The letters (a), (b) and (c) refer to [100], [010] and [001] type of dendrites, respec-
tively. After Rappaz et al.17
Weld centre-line
Fusion line
Fig. 3.24. Sketch of curved columnar grain in Example (3.3).
Solution
In principle, the solution to this problem is identical to that presented in Example (3.2). Refer-
ring to Fig. 3.24 the growth rate of [100] and the [010] dendrite stems is given by:
and
It follows from Fig. 3.21 that growth will occur preferentially in the [010] direction as long
as the interface normal angle a is larger than 45°, while the [100] direction is selected at
smaller angles. This means that the local growth rate of the dendrites, in practice, never will
exceed the welding speed v.
3.4.4.2 Renucleation of crystals

In ingots and castings, three different mechanisms for nucleation of new grains ahead of the
advancing interface are operative:12
(i) Heterogeneous nucleation

(ii) Dendrite fragmentation
(iii) Grain detachment.
The former mechanism is of particular importance in welding, since the weld metal often
contains a high number of second phase particles which form in the liquid state. These parti-
cles can either be primary products of the weld metal deoxidation or stem from reactions
between specific alloying elements which are deliberately introduced into the weld pool through
the filler wire. The latter process is also known as inoculation.
Nucleation potency of second phase particles

In general, the effectiveness of individual particles to act as heterogeneous nucleation sites can
be evaluated from a balance of interfacial energies, analogous to that described in Section
3.3.1 for epitaxial nucleation. It follows from the definition of the wetting angle (3 in Fig. 3.5
that the energy barrier to heterogeneous nucleation is a function of both the substrate/liquid
interfacial energy ySL, the substrate/embryo interfacial energy yES, and the embryo/liquid in-
terfacial energy yEL. Complete wetting is achieved when:
(3-11)
Under such conditions, the nucleus will readily grow from the liquid on the substrate. Un-
fortunately, data for interfacial energies are scarce and unreliable, which makes predictions
based on equation (3-11) rather fortuitous.18
In pure metals, experience has shown that the solid/liquid interfacial energies are roughly
proportional to the melting point, as shown by the data in Fig. 3.25. On this basis, it can be
expected that the higher melting point phases will reveal the highest ySL values, and thus be
nucleants for lower melting phases. A similar situation also exists in the case of non-metallic
inclusions in liquid steel, where the high-melting point phases are seen to exhibit the highest
solid/liquid interfacial energies (see Fig. 3.26).
In contrast, very little information is available on the substrate/embryo interfacial energy
yES. For fully incoherent interfaces, yES would be expected to be of the order of 0.5 to 1 J m~2.5
However, this value will be greatly reduced if there is epitaxy between the inclusions and the
nucleus, which results in a low lattice disregistry between the two phases. In general, assess-
ment of the degree of atomic misfit between the nucleus n and the substrate s can be done on
Melting point, K
Interfacial energy, J / m 2
Fig. 3.25. Values of solid/liquid interfacial energy ySL of various metals as function of their melting points.
Data from Mondolfo.18
lnterfacial energy, J/m2
Melting point, 0 C
Fig. 3.26. Values of interfacial energy 7 5L for different types of non-metallic inclusions in liquid steel at
16000C as function of their melting points. Data compiled from miscellaneous sources.
the basis of the Bramfitt's planar lattice disregistry model :19
(3-12)
where a low-index plane of the substrate;

a low-index direction in (hkl)s
a low-index plane in the nucleated solid;
a low-index direction in (hkl)n;
the interatomic spacing along [wvw]n;
the interatomic spacing along [WVH>]5; and
the angle between the [wvw]^ and the [wvw]w.
In practice, the undercooling Ar (which is a measure of the energy barrier to heterogeneous

nucleation) increases monotonically with increasing values of the planar lattice disregistry, as
shown by the data in Fig. 3.27. This means that the most potent catalyst particles are those
which also provide a good epitaxial fit between the substrate and the embryo. Examples of
such catalyst particles are TiAl3 in aluminium 18 and TiN in steel.19 Nucleation of delta ferrite
at titanium nitride will be considered below.
Undercooling, 0C
« „ « >
Fig. 3.27. Relationship between planar lattice disregistry and undercooling for different nucleants in
steel. Data compiled from miscellaneous sources.
Example (3.4)
In low-alloy steel weld metals, titanium nitride can form in the melt due to interactions be-
tween dissolved titanium and nitrogen. Assume that the TiN particles are faceted and delim-
ited by {100} planes. Calculate on the basis of equation (3-12) the minimum planar lattice
disregistry between TiN and the nucleating delta-ferrite phase under the prevailing circum-
stances. Indicate also the plausible orientation relationship between the two phases. The
lattice parameters of delta ferrite and TiN at 15200C may be taken equal to 0.293 and 0.43 lnm,
respectively.
Solution
Titanium nitride has the NaCl crystal structure, while delta ferrite is body-centred cubic, as
shown in Fig. 3.28(a) and (b). It is evident from Fig. 3.29(a) that a straight cube-to-cube ori-
entation relationship between TiN and 8-Fe will not result in a small lattice disregistry.
However, the situation is largely improved if the two phases are rotated 45° with respect to
each other (see Fig. 3.29(b)), conforming to the following orientation relationship:
The resulting crystallographic relationship at the interface is shown schematically in Fig.

3.29(c). Since the lattice arrangements are similar in this case, equation (3-12) reduces to:
A comparison with the data in Fig. 3.27 shows that the calculated lattice disregistry con-
forms to an undercooling of about 1 to 2°C. This value is sufficiently small to facilitate hetero-
geneous nucleation of new grains ahead of the advancing interface during solidification.
Considering other inclusions with more complex crystal structures, the chances of obtain-
ing a small planar lattice disregistry between the substrate and the delta ferrite nucleus are
Fe- atoms
N-atoms
Ti- atoms
(a) (b)
Fig. 3.28. Crystal structures of phases considered in Example (3.4); (a) Titanium nitride, (b) Delta ferrite.
TiN
(a)
TiN
(C)
(b)
Ti atoms N atoms 8-Fe atoms
Fig. 3.29. Possible crystallographic relationships between titanium nitride and delta ferrite (Example (3.4));
(a) Straight cube-to-cube orientation, (b) Twisted cube-to-cube orientation, (c) Details of lattice arrange-
ment along coherent TiN/d-Fe interface.
rather poor (see Fig. 3.27). Nevertheless, such particles can act as favourable sites for hetero-
geneous nucleation if 7 ^ is sufficiently large compared with yEL and 7 ^ . This is illustrated by
the following example:
Example (3.5)
In low-alloy steel weld metals 7-Al2O3 inclusions can form during the primary deoxidation
stage as discussed in Section 2.12.4.2 (Chapter 2). Based on the classic theory of heteroge-
neous nucleation, evaluate the nucleation potency of such inclusions with respect delta ferrite.
Solution
It is readily seen from Fig. 3.27 that the planar lattice disregistry between delta ferrite and
Al2O3 is very large, which indicates of a fully incoherent interface (i.e. yES « 0.75 J m" 2 ).
Moreover, readings from Figs. 3.25 and 3.26 give the following average values for the delta fer-
rite/liquid and the inclusion/liquid interfacial energies:
and
According to equation (3-11) complete wetting is achieved when ySL > yES + yEL. This
requirement is clearly met under the prevailing circumstances.
Similar calculations can also be performed for other types of non-metallic inclusions in
steel weld metals. The results are presented graphically in Fig. 3.30. It is evident that the
nucleation potency of the inclusions increases in the order SiO2-MnO, Al 2 O 3 -Ti 2 O 3 -SiO 2 -
MnO, Al2O3, reflecting a corresponding increase in the inclusion/liquid interfacial energy ySL.
The resulting change in the weld metal solidification microstructure is shown in Fig. 3.31,
from which it is seen that both the average width and length of the columnar grains decrease
with increasing Al2O3-contents in the inclusions. This observation is not surprising, consider-
ing the characteristic high solid/liquid interfacial energy between aluminium oxide and steel
(see Fig. 3.26). The important effect of deoxidation practice on the weld metal solidification
microstructure is well documented in the literature.320"22
Rate of heterogeneous nucleation

It can be inferred from the classic theory of heterogeneous nucleation that the nucleation rate
p (degrees)
No wetting
Complete wetting
G*he/AGhom
Embryo'
A
Inclusion
(Y sf Y ES )/7 EL
Fig. 3.30. Nucleation potency of different weld metal oxide inclusions with respect to delta ferrite.
(a) AI2O3 content (wt%)
Average width of grains, î m 95% confidence limit
Pure AI2O3
(A%AI)weld/[%O]anaL
(b) AI2O3 content (wt%)
95% confidence limit

Average length of grains, ^m
Pure AI2O3
<A%AIW[%0W
Fig. 3.31.Effect of deoxidation practice on the columnar grain structure in low-alloy steel weld metals;
(a) Average width of columnar grains, (b) Average length of columnar grains. Data from Kluken et al.22
Nhet(i.e. number of nuclei which form per unit time and unit volume of the melt) is interre-
lated to the energy barrier AG^ through the following equation:5
(3-13)
whereZ1 is a frequency factor, Nv is the density of nucleation sites per unit volume of the melt,
AGD is the activation energy for diffusion of atoms across the interface, and k is the Boltzmann
constant.
Since AGD is often negligible compared with AGêt in liquids, equation (3-13) reduces to:
(3-14)
Equation (3-14) shows that the nucleation rate Nhet depends both on Nv and &Ghe{. Hence,
under full wetting conditions (AGhet ~ 0), the number of nuclei which form per second and mm3
ahead of the advancing interface is directly proportional to the instantaneous concentration of
catalyst particles in the melt. Examples of such particles are TiAl3 in aluminium and TiN/Al2O3
in steel. The important effect of controlled titanium additions and subsequent TiAl3 precipi-
tation on the columnar grain structure in 1100 aluminium welds is illustrated in Fig. 3.32.
Critical cell/dendrite alignment angle

It follows from the minimum growth rate criterion and the definition of the local crystal growth
rate in equation (3-10) that reorientation of the columnar grains will occur when the cell/
dendrite alignment angle $ reaches a critical value <\>*. The value of 4>* will, in turn, depend on
the nucleation potency of the catalyst particles and can be estimated for different types of
welds. If growth of the columnar grains is assumed to occur along a circle segment of length
L (see Fig. 3.33), the critical cell/dendrite alignment angle is given by:
(3-15)
where co is the total grain rotation angle, and / is the average length of the columnar grains (in
mm).
Average width of grains, (x m
Titanium content, wt%
Fig. 3.32. Effect of titanium on the columnar grain structure in 1100 aluminium welds. The value Y is
the fractional distance from fusion line to top surface of weld metal. Data from Yunjia et alP
Weld metal
Fig. 3.33. Characteristic growth pattern of columnar grains in bead-on-plate welds (schematic).
By introducing reasonable average values for co and / in the case of SA welding of low-
alloy steel,22 we obtain:
(3-16)
Calculated values for 4>* in steel weld metals are presented in Fig. 3.34, using data from
Kluken et al.22 An expected, the critical cell/dendrite alignment angle in fully aluminium
deoxidised steel welds is seen to be very small (of the order of 2°), reflecting the fact that
nucleation of delta ferrite occurs readily at Al2O3 inclusions. The value of 4>* increases gradu-
ally with decreasing Al2O3 contents in the inclusions and reaches a maximum of about 4° for
Si-Mn deoxidised steel weld metals. This situation can be attributed to less favourable nucle-
ating opportunities for delta ferrite at silica-containing inclusions, which reduces the possibili-
ties of obtaining a change in the crystal orientation during solidification through a nucleation
and growth process.
Dendrite fragmentation
In principle, nucleation of new grains ahead of the advancing interface can also occur from
random solid dendrite fragments contained in the weld pool. Although the source of these
solid fragments has yet to be investigated, it is reasonable to assume that they are generated by
some process of interface fragmentation due to thermal fluctuations in the melt or mechanical
disturbances at the solid/liquid interface.3 At present, it cannot be stated with certainty whether
grain refinement by dendrite fragmentation is a significant process in fusion welding.26
Grain detachment
Since the partially melted base metal grains at the fusion boundary are loosely held together by
liquid films between them, there is also a possibility that some of these grains may detach
themselves from the base metal and be trapped in the solidification front.26 Like dendrite
fragments, such partially melted grains can act as seed crystals for the formation of new grains
in the weld metal during solidification if they are able to survive sufficiently long in the melt.
Next Page
R L / v cos a
Calculated from
equation (3-10)
Critical cell/dendrite alignment angle ($*)

Fig. 3.34. Critical cell/dendrite alignment angle ()>* for reorientation of delta ferrite columnar grains dur-
ing solidification of steel weld metals. Data from Kluken et al.22
3.5 Solidification Microstructures
So far, we have discussed growth of columnar grains without considering in detail the weld
metal solidification microstructure. In general, each individual grain will exhibit a substruc-
ture consisting of a parallel array of dendrites or cells. This substructure can readily be re-
vealed by etching, also in cases where it is masked by subsequent solid state transformation
reactions (as in ferrous alloys).2224
3.5.1 Substructure characteristics

A cellular substructure within a single grain consists of an array of parallel (hexagonal) cells
which are separated from each other by 'low-angle' grain boundaries, as shown schematically
in Fig. 3.35. In the presence of solute, these boundaries respond to etching even in the absence
of segregation. When the cellular to dendritic transition occurs, the cells become more dis-
torted and will finally take the form of irregular cubes, as indicated by the optical micrograph
in Fig. 3.36. This is actually a dendritic type of substructure, where the formation of secondary
and tertiary dendrite arms is suppressed because of a relatively small temperature gradient in
the transverse direction compared with the longitudinal (growth) direction. Fully branced
dendrites may, however, develop in the centre of the weld if the thermal conditions are favour-
able. Branching will then occur in specific crystallographic directions, e.g. along the three
<100> easy growth directions for bcc and fee crystals, as illustrated in Fig. 3.37.
Besides the difference in morphology, the distinction between cells and dendrites lies pri-
marily in their sensitivity to crystalline alignment. Cells do not necessarily have the <100>
axis orientation, while dendrites do.2 Hence, cells can grow with their axes parallel to the heat
flow direction, regardless of the crystal orientation. This important point is often overlooked
when discussing competitive grain growth in fusion welding.
Previous Page
R L / v cos a
Calculated from
equation (3-10)
Critical cell/dendrite alignment angle ($*)

Fig. 3.34. Critical cell/dendrite alignment angle ()>* for reorientation of delta ferrite columnar grains dur-
ing solidification of steel weld metals. Data from Kluken et al.22
3.5 Solidification Microstructures
So far, we have discussed growth of columnar grains without considering in detail the weld
metal solidification microstructure. In general, each individual grain will exhibit a substruc-
ture consisting of a parallel array of dendrites or cells. This substructure can readily be re-
vealed by etching, also in cases where it is masked by subsequent solid state transformation
reactions (as in ferrous alloys).2224
3.5.1 Substructure characteristics

A cellular substructure within a single grain consists of an array of parallel (hexagonal) cells
which are separated from each other by 'low-angle' grain boundaries, as shown schematically
in Fig. 3.35. In the presence of solute, these boundaries respond to etching even in the absence
of segregation. When the cellular to dendritic transition occurs, the cells become more dis-
torted and will finally take the form of irregular cubes, as indicated by the optical micrograph
in Fig. 3.36. This is actually a dendritic type of substructure, where the formation of secondary
and tertiary dendrite arms is suppressed because of a relatively small temperature gradient in
the transverse direction compared with the longitudinal (growth) direction. Fully branced
dendrites may, however, develop in the centre of the weld if the thermal conditions are favour-
able. Branching will then occur in specific crystallographic directions, e.g. along the three
<100> easy growth directions for bcc and fee crystals, as illustrated in Fig. 3.37.
Besides the difference in morphology, the distinction between cells and dendrites lies pri-
marily in their sensitivity to crystalline alignment. Cells do not necessarily have the <100>
axis orientation, while dendrites do.2 Hence, cells can grow with their axes parallel to the heat
flow direction, regardless of the crystal orientation. This important point is often overlooked
when discussing competitive grain growth in fusion welding.
Cell wall' Fig. 3.35. Schematic representation of the cellular substructure.
Fig. 3.36. Optical micrograph showing the characteristic cellular-dendritic substructure in a low-alloy
steel weld. The metallographic section is normal to the columnar grain growth direction. After Kluken
etal.22
Fig. 3.37. Schematic representation of the development of

primary, secondary and tertiary dendrite arms along <100>
directions in cubic crystals.
A characteristic feature of cellular and cellular-dendritic growth is also that the boundary
between two adjacent columnar grains will closely follow the contours of the original cell
boundaries. An illustration of this point is contained in Fig. 3.38. Consequently, since all cell
walls are preferential sites for segregation during solidification (see ion micrograph in Fig.
3.39), the presence of solute at the columnar grain boundaries can strongly alter the kinetics of
subsequent solid state transformation reactions. The indications are that for instance phospho-
rus segregations at prior austenite grain boundaries will promote the formation of grain bound-
ary ferrite in low-alloy steel weld metals during the 7 to a transformation because of the
associated increase in the A<?3-temperature.22'25 We will return to this question in Chapter 6.
(a)
(b)
Fig. 3.39. Ion (SIMS) micrograph showing evidence of phosphorus segregations at primary solidifica-
tion (cell) boundaries in a low-alloy steel weld. After Kluken et al22
3.5.2 Stability of the solidification front

The stability of the solid/liquid interface is critical in determining the microstructural charac-
teristics of the weld metal.
3.5.2.1 Interface stability criterion

The simplified treatment given here is based on the constitutional undercooling criterion which
ignores the important effect of capillarity on the interface stability.
Consider a simple binary system, the phase diagram of which is given in Fig. 3.40(a).
During solidification, a solute-rich layer will form in front of the growing interface, as shown
schematically in Fig. 3.40(b). At steady state, the mass of solute transferred from the liquid to
the interface, R1(C1Jp is equal to the mass of solute which accumulates in the solid, R1(C8)^
minus the amout which diffuses back into the liquid, D1(O1C1Za1X)1. Hence, we may write:
(3-17)
Provided that equilibrium exists at the solid/liquid interface, i.e. (Cs)t = IcJC1J1, equation
(3-17) can be rewritten as:
(3-18)
where DL is the diffusivity of the solute in the liquid, and ko is the equilibrium partition coeffi-
cient (Jco < 1).
(a)
Temperature
Concentration
(b)
Liquid composition
Distance (x)
(C)
Temperature
Liquidus temperature (TL)
Constitutionally undercooled region
Distance (x)
Fig. 3.40. Constitutional undercooling in alloy solidification; (a) Schematic representation of binary phase
diagram, (b) Build-up of solute-enriched layer in front of solid/liquid interface, (c) Undercooled region
ahead of solid/liquid interface.
With the aid of the phase diagram in Fig. 3.40(a) it is easy to verify that the equilibrium
liquidus temperature increases with distance from the interface because of the lower solute
content. Referring to Fig. 3.40(c) the latent heat of melting AHm will diffuse away from the
interface (thereby stabilising possible interface protuberances) if the actual temperature gradi-
ent in the liquid (dTA/dx)t is less than the equilibrium liquidus temperature gradient (3T1ZdX)1.
The latter gradient can be expressed as:
(3-19)
where (dTL/dCL)i denotes the slope of the liquidus curve in the phase diagram (designated mL).
By combining equations (3-18) and (3-19), we obtain the following criterion for the inter-
face stability:
(3-20)
Taking (C1). = C0Ik0 and Rj = RL this equation can alternatively be written as:
(3-21)
3.5.2.2 Factors affecting the interface stability

It follows from equation (3-21) that the stability of the weld metal solidification front is con-
trolled by the extent of constitutional undercooling ahead of the advancing interface, and is
therefore influenced by factors such as the total amount of alloying or impurity elements present,
C0, the local crystal growth rate, RL, and the thermal gradient in the weld pool, GL. The
combined effects of alloy content, growth rate, and thermal gradient are presented schematically
in Fig. 3.41. It can be seen from the figure that a planar solidification front is favoured by a
high GL/RL ratio, in agreement with predictions based on equation (3-21). At lower GLIRL
ratios, the morphology of the interface changes to cellular, cellular-dendritic or dendritic, de-
pending on the degree of constitutional undercooling ahead of the advancing interface.
Normally, the GL/RL ratio close to the fusion line is large enough to facilitate cellular solidi-
fication. The temperature gradient in the weld pool decreases, however, with distance from
the fusion boundary and hence, a cellular-dendritic type of substructure is often observed in
the central areas of the columnar grain region.22 This substructure may change to dendritic
close to the weld centre-line if the degree of constitutional undercooling in front of the advanc-
ing interface is sufficiently large. The various types of growth products that may develop
during normal solidification of fusion welds are shown schematically in Fig. 3.42.
Example (3.6)
In aluminium welding, binary Al-Si and Al-Mg alloys are frequently used as filler metals.
Consider stringer bead deposition with an Al-4.5wt%Si filler wire at a constant welding speed
of 1.5mm s"1. Based on equation (3-21) calculate the critical (minimum) temperature gradient
Alloying level (C )
Planar
GL/RL
Fig. 3.41. Schematic representation of the combined effect of crystal growth rate RL and melt thermal
gradient GL on the weld metal solidification microstructure.
Fusion zone
v Larger R L, Weld centre-line

Small RL, Small GL Equiaxed-dendritic
Large G1 Cellular-dendritic
Cellular
HAZ grains
Fig. 3.42. Schematic diagram illustrating structural variations in the weld metal solidification micro-
structure across the fusion zone.
in the weld pool which gives a planar solidification front. Relevant physical data for the Al-Si
system are given below:2
Solution
When mL, ko and D1 are known the critical temperature gradient (G1)cr can readily be calcu-
lated from equation (3-21). Taking RL ~ v, we obtain:
The above calculations show that the temperature gradient in the weld pool must be ex-
tremely large in order to promote a planar solidification front. The calculated value for (GL)cr
corresponds to a cooling rate of about 92 4000C s"1 (CR. = G1R1). Besides low-heat input
electron beam and laser welding, such high cooling rates are rarely observed in fusion welding
(see Fig. 3.3). On this basis it is not surprising to find that the weld metal solidification micro-
structure is normally of the cellular or the dendritic type.
For a rough evaluation of the weld metal solidification microstructure the diagram in Fig.
3.43 can be used. This diagram summarises the various microstructures which can be obtained
(using a typical alloy with a melting range of 500C) when the imposed temperature gradient GL
and crystal growth rate RL are varied. Moving from the lower left to the upper right along the
lines at 45° leads to a refinement of the structure without changing the morphology (constant
GL/RL ratio). Crossing these lines by passing from the upper left to the lower right leads to
changes in the morphology from planar to cellular or dendritic growth, while the scale of the
microstructure remains essentially the same. The gray bands define the regions over which
one structure changes into another. An example of the application of this diagram is given
below.
Example (3.7)
Consider GTA butt welding of a 2mm thin sheet of aluminium (Al-Mg alloy) under the fol-
lowing conditions:
Based on the diagram in Fig. 3.43, estimate the weld metal solidification microstructure in
different positions from the fusion boundary. In these calculations we shall assume that the
actual thermal gradient in the weld pool is equal to the average thermal gradient within the
temperature interval from 680 to 5500C, as evaluated from the Rosenthal thin plate solution
(equation (1-83)). Relevant thermal data for the Al-Mg alloy are given in Table 1.1 (Chapter
1).
Solution
The contours of the fusion boundary and the 680 and the 5500C isotherms can be calculated
according to the procedure shown in Example (1.10). If we neglect the latent heat of melting,
the corresponding 68//?3 ratios at these temperatures become:
GL,°C/mm
RL, mm/s
Fig. 3.43. Variation of weld metal solidification microstructure with GL and RL. The diagram is based on
the ideas of Kurz and Fisher.2
Substitution of the above values into equation (1-83) gives the isothermal contours shown
in Fig. 3.44. It follows from Fig. 3.44 that the thermal gradient decreases from about 25°C
mnr 1 at the fusion boundary to approximately 100C mm"1 close to the weld centre-line. This
occurs parallel with an increase in the nominal crystal growth rate from 0.8 to 4mm s"1.
According to Fig. 3.43 the calculated values of GL and RL conform to a cellular-dendritic
solidification microstructure within the central regions of the weld and an equiaxed-dendritic
microstructure close to the weld centre-line. Both types of substructures are commonly ob-
served in aluminium weldments.2728
Weld pool (680 0C)
Fusion boundary (650 0C)
HAZ (550 0C)
+x (mm)
+Y (mm)
Fig. 3.44. Predicted shape of fusion boundary and neighbouring isotherms during thin plate aluminium
welding (Example (3.6)).
3.5.3 Dendrite morphology
Within the fully dendritic region of the weld, the dendrite morphology remains largely un-
changed over a wide range of cooling rates. Nevertheless, it will become finer as the heat is
extracted at greater rates.
3.5.3.1 Dendrite tip radius

A combination of stability analysis with diffusion considerations and the specification of a
dendrite tip shape and a radius selection criterion can lead to an unique solution for the mor-
phology and solute distribution around a growing dendrite.2930 Based on such analyses it is
possible to calculate the dendrite tip radius rd as function of the growth rate RL and the thermal
gradient GL in the weld pool for a given alloy system.17
Referring to Fig. 3.45, the dendrite growth is seen to occur between two stability limits. At
low speeds, the rd vs R1 curve reaches a threshold where the tip radius increases rapidly and
finally approaches an infinite value corresponding to a planar solidification front. At high
speeds, an absolute stability limit is also found. This corresponds to the limiting case where
the transport of solute across the interface becomes too sluggish to keep pace with the advanc-
ing solidification front. However, within the central areas of the curve, the dendrite tip radius
is seen to decrease steadily with increasing values of RD typically as l/(RL)m. Note that the
thermal gradient in the weld pool GL will have a strong influence on the low growth rate
stability limit. In contrast, the absolute stability limit is essentially independent of GL, as
indicated in Fig. 3.45.
Tip radius,\irr\
Growth rate, mm/s

Fig. 3.45. Predicted relationship between dendrite tip radius rd, crystal growth rate RL and melt thermal
gradient GL for a Fe-15Cr-15Ni alloy. Data from David and Vitek.4
3.5.3.2 Primary dendrite arm spacing
The primary dendrite arm spacing (defined in Fig. 3.46) is an important characteristic of the
solidification microstructure and has a marked effect on the weld metal mechanical properties.
If it is assumed that the dendrite envelope, representing the mean cross-section of the trunk,
can be described by an ellipse (see Fig. 3.47), the radius of curvature of the ellipse is given by:2
(3-22)
In general, the half width b of the dendrite stem is proportional to the primary arm spacing
X1. In cases where the arrangement of the dendrite trunks can be represented by a simple
close-packed hexagonal array, the value of b is exactly equal to X1 IV3 -2 Similarly, the total
length of the dendrite stem g can be calculated by considering the difference between the tip
temperature Ttip and the root temperature Troot, as shown in Fig. 3.48. Taking the thermal
gradient in the mushy zone equal to GL, we obtain:
(3-23)
This gives the following relationship between the primary dendrite arm spacing X1, the tip
radius rd and the thermal gradient GL:
(3-24)
rd
b
Fig. 3.46. Definition of primary X1 and secondary X2 dendrite tig. 3.47.Geometrical relationship
arm spacings. between the radius of curvature rd
and the width to length ratio of an
elliptical dendrite.
Ttip
Troot
9 X
Heat flow
Cell/dendrite
Fig. 3.48. Definition of the dendrite tip Ttip and root Troot temperatures.
If we also take into account that the dendrite tip radius rd is inversely proportional to the
square root of the crystal growth rate RL within the central range of RL (see Fig. 3.45), equation
(3-24)reduces to:
(3-25)
where c{ is a kinetic constant which is characteristic of the alloy system under consideration.
It is evident from the above analysis that the primary dendrite arm spacing cannot readily
be characterised by one single parameter (e.g. the cooling rate), since its dependence on GL
and R1 have different exponents.
Example (3.8)
Based on equation (3-25), show that the following relationships exist between the primary
dendrite arm spacing X1, the net arc power qo, and the weld travel speed v during thick plate
and thin plate welding, respectively:
Thick plate welding:

(3-26)
Thin plate welding:

(3-27)
where c2 and c3 are proportionality constants.

Solution
Under 3-D heat flow conditions, the pseudo-steady state temperature distribution is given by
the Rosenthal thick plate solution (equation (1 -45)). For points located on the weld centre-line
behind the heat source y = z = 0, and R* = -x. Hence, equation (1-45) reduces to:
The thermal gradient GL in the mushy zone close to the weld centre-line can be obtained
by differentiating of the above equation with respect to the x-coordinate:
(3-28)
By inserting the appropriate expressions for GLandRL into equation (3-25) (noting that RL=v
at the weld centre-line), we arrive at an expression for X1 which is identical with the one
presented in equation (3-26).
Similarly, under 2-D heat flow conditions, the pseudo-steady state temperature distribution
is given by the Rosenthal thin plate solution (equation 1-81). For points located on the weld
centre-line behind the heat source y = 0 and r = -x = vt. If IxI is sufficiently large, it is a fair
approximation to set Ko(u) ~ exp(-u)^n!2u . Hence, equation (1-81) reduces to:
The thermal gradient GL is then given as:
(3-29)
From this we see that the primary dendrite arm spacmg X1 during thin plate welding is
interrelated to qo and v through a relationship of the type shown in equation (3-27).
In Fig. 3.49 the validity of equation (3-26) has been checked against the experimental data
of Jordan and Coleman,27 who measured the primary dendrite arm spacing in different GMA
Al-Mg-Mn welds. It is evident from this plot that their data-points can approximately be
represented by straight lines passing through the origin, as required by the theory. Moreover,
a closer inspection of the figure shows that the dendrite arm spacing close to the centre-line
varies systematically from the bottom to the top of the weld. This observation is not surpris-
ing, considering the fact that the rate of solidification increases progressively from the toe to
the surface of the plate.
Example (3.9)
Consider GTA butt welding of a 2mm thin aluminium sheet (Al-Mg alloy) under conditions
similar to those employed in Example (3.7). Based on equation (3-25) calculate the relative
change in the primary dendrite arm spacing X1 from the fusion boundary to the weld centre-
line during solidification.
GMAW (Al-Mg-Mn alloys)
Just above toe of weld
Level with surface of
Primary dendrite arm spacing, p,m
plate
Parallel to plate
surface
Just above toe.
of weld
[(qo)1/2/(v)1/4],(WS1/2/mm1/2)1/2
Fig. 3.49. Experimental verification of equation (3-26). Data from Jordan and Coleman.27
Solution
As shown in Fig. 3.44, the thermal gradient, GL, decreases from about 25°C mm"1 at the
fusion boundary to approximately 10°C mm"1 close to the weld centre-line. This occurs
parallel with an increase in the nominal crystal growth rate from 0.8 to 4mm s"1. Taking the
primary dendrite arm spacing at the fusion boundary equal to X1*, the ratio X1 /X1* in different
positions of the weld becomes:
Central region of columnar zone:
Weld centre-line:
In contrast to that predicted above, the smallest dendrite arm spacing is normally observed at
the weld centre-line.26 This has to do with the fact that the constant C1 in equation (3-25), in
practice, decreases with increasing distance from the fusion boundary due to solute segrega-
tion, which gradually reduces the coarsening rate of the dendrites.
3.5.3.3 Secondary dendrite arm spacing

It follows from the above analysis that the primary dendrite arm spacing, once it has been
established, will remain constant during subsequent cooling of the weld. This is not true of the
secondary arms (see definition in Fig. 3.46), which undergo a continuous ripening process.
Hence, their size and morphology will change with time as the thicker branches grow larger at
the expense of the smaller ones.2
The driving force for the ripening process is the disparity in chemical potential (interfacial
energy) between branches with different curvature. Since this process is analogous to the
Ostwald ripening of precipitates (see equation (2-76)), the spacing of the branches, X2, will be
a simple cube root function of the local solidification time fo:1'2
(3-30)
where M is a mobility term defined as:2
(3-31)
The parameter Q in equation (3-31) refers to the Gibbs-Thomson coefficient and is given as:
(3-32)
where a is the solid/liquid interfacial energy, and AS, is the entropy of fusion.
In practice, the value of M can easily vary by an order of magnitude. Nevertheless, its effect
on the secondary dendrite arm spacing is rather weak, since X2 is proportional to the cube root
of M. Hence, a plot of the secondary dendrite arm spacing vs the local solidification time in a
logarithmic diagram will normally reveal a straight-line relationship between X2 and to with a
slope close to 0.33, as shown in Fig. 3.50.
GTAW (Stainless steel)

X2^m
Slopes 0.33
to'5
Fig. 3.50. Relation between secondary dendrite arm spacing X2 and local solidification time to in type
AISI 310 stainless steel welds. Data from Kou and Lee.31
Example (3.10)
Based on equation (3-30), derive a relationship between the secondary dendrite arm spacing
X2, the net arc power qo, and the weld travel speed v during thick plate welding. Calculate then
the secondary dendrite arm spacing in the centre of a thick GTA Al-Si weld deposited under
Relevant physical data for the Al-Si system are given below:2
Solution
The local solidification time to is the time for the dendrite array to pass an arbitrary point in the
weld and is therefore a measure of the solidification rate. Referring to Fig. 3.48, the local
solidification time is defined as:
(3-33)
By inserting the appropriate expressions for GL (equation (3-28)) and RL in position y = z = 0,

we obtain the following relationship between to, qo and v during thick plate welding:
(3-34)
A combination of equations (3-30) and (3-34) gives:
(3-35)
In the present example, the numerical values of M and X2 are:

and
Although reliable experimental data are not available for a direct comparison, the calcu-
lated value for X2 is considered reasonably correct. In Al-Cu castings a secondary dendrite
arm spacing of about 4(im corresponds to a cooling rate of the order of 103 °C s"1, as shown in
Fig. 3.51. In the present example, the cooling rate at the solid/liquid interface is close to:
This value is compatible with a secondary dendrite arm spacing of 4|am.
Example (3.11)
The operating parameters given in Example (3.10) are also applicable to single pass butt weld-
ing of thin aluminium plates. Based on equation (3-30), estimate the secondary dendrite arm
spacing in the centre of a 2mm thick Al-Si butt weld deposited under such conditions.
Solution
In thin plate welding the local solidification time to is obtained by combining equations (3-29)
and (3-33). Noting that RL = v at the weld centre-line, we get:
(3-36)
AI-4.5 wt% Cu
X 2 ,nm
Cooling rate, °C/s
Fig. 3.57.Relation between secondary dendrite arm spacing X2 and cooling rate CR. in Al-Cu castings.
Data compiled by Munitz.32
The secondary dendrite arm spacing is thus given as:
(3-37)
3
Taking pc equal to 0.0027 J mm C ] in the case of Al-Si alloys, we obtain:
It is evident from the above calculation that the secondary dendrite arm spacing is signifi-
cantly coarser in thin plate welding compared with thick plate welding. This observation is not
surprising, considering the pertinent difference in the heat transfer mode and thus the cool-
ing rate between these two types of weldments. In the former case, we have:
A comparison with the data in Fig. 3.51 shows that a cooling rate of about 78°C s"1 is
compatible with a secondary dendrite arm spacing of 13 jim.
3.6 Equiaxed Dendritic Growth
As already stated in the introduction of the chapter, an equiaxed zone is often observed close to
the weld centre-line. This zone can in certain cases be very dominating and completely over-
ride the columnar grain zone, as shown in Fig. 3.52.
As long as growth occurs in a columnar manner the crystals are in contact with solid metal.
The heat will therefore be conducted through the crystals in a direction which is parallel and
opposite to their growth direction, as illustrated in Fig. 3.53(a). In equiaxed dendritic growth
the situation is different. Here the melt will be slightly undercooled because of extensive seg-
regation of solute to the weld centre-line where the solidification fronts growing from each
side impinge. This makes the crystals hotter than the liquid and gives rise to a radial heat flow
away from the crystals in the same direction as that of growth (see Fig. 3.53(b)).
3.6.1 Columnar to equiaxed transition

In fusion welds, the GL/RL ratio decreases continuously from the fusion boundary to the weld
centre-line during the crystallisation process. For a given alloy system containing a fixed
number of heterogenous nucleation sites it is reasonable to assume that the columnar to equiaxed
FL CL
FL CL
Fig. 3.52.Example of grain refinement in an Al-Mg-Si plasma arc weld due to TiAl3-precipitation (FL:
fusion line, CL: centre-line). Courtesy of M.I. Onsoien, SINTEF, Trondheim, Norway.
transition occurs when the GJR1 ratio drops below a certain critical value fcn. Hence, we may
write:
(3-38)
During thick plate welding (3-D heat flow) the temperature gradient at the weld centre-line
is given by equation (3-28). Taking RL = v, a combination of equations (3-28) and (3-38) gives:
(3-39)
where c4 is a kinetic constant which is characteristic of the alloy system under consideration.
Similarly, for welding of medium thick plates (mixed heat flow), we get:
(3-40)
where c5 is a new kinetic constant, and n is an exponent which varies between zero and unity
depending on the mode of heat flow (i.e. thin, medium thick or thick plate welding).
(a) (b)
Heat flow
Fig. 3.53. Schematic diagrams showing different dendrite growth morphologies in castings; (a) Colum-
nar dendritic growth, (b) Equiaxed dendritic growth. The diagrams are based on the ideas of Kurz and
Fisher.2
Equation (3-40) predicts that the columnar to equiaxed transition in fusion welds occurs at
critical combinations of the net arc power qo and the welding speed v. Thus, a decrease in qo
must always be compensated by a corresponding increase in v in order to maintain equiaxed
dendritic growth at the weld centre-line. This is also in agreement with general experience
(see Fig. 3.54).
Example (3.12)
Based on the experimental data in Fig. 3.54, calculate the critical G1IR1 ratio which provides
equiaxed dendritic growth during thick plate welding of Al-Mg alloys (5083 series). Relevant
thermal data for the Al-Mg system are given in Table 1.1 (Chapter 1).
Plate thickness: 10-12 mm
Net arc power, kW
Welding speed, mm/s
Fig. 3.54. Effect of net arc power qo and welding speed v on the columnar to equiaxed transition in
different aluminium welds. Data from Matsuda et al.33
Solution
The critical G1IR1 ratio is obtained by rearranging equation (3-39):
By inserting data from Fig. 3.54 and Table 1.1, we get:
By multiplying fcr with the welding speed we see that the calculated value corresponds to
a critical temperature gradient of about 928°C mm"1. This means that the thermal conditions
existing in welding will favour growth of equiaxed dendrites close to the weld centre-line. A
requirement is, of course, that the melt contains a sufficient number of seed crystals to facili-
tate heterogeneous nucleation of new grains ahead of the advancing solid/liquid interface.
Next Page
3.6.2 Nucleation mechanisms

Several nucleation mechanisms have been proposed to explain the columnar to equiaxed tran-
sition during weld pool solidification, including:26

The former mechanism is particularly relevant to welding, since the weld metal often con-
tains a high number of second phase particles which form in the liquid state. As already
mentioned in Section 3.4.4.2, these particles can either be primary products of the weld metal
deoxidation or stem from reactions between specific alloying elements which are deliberately
introduced into the weld pool through the filler wire. The important effect of deoxidation
practice (inclusions) on the columnar to equiaxed transition in ferritic stainless steel GTA
welds is shown in Figs. 3.55 and 3.56.
3.7 Solute Redistribution
During solidification of fusion welds, alloying and impurity elements tend to segregate exten-
sively to the centre parts of the intercellular or interdendritic spaces under the conditions of
rapid cooling. This, in turn, alters the kinetics of the subsequent solid state transformation
reactions.
3.7.1 Micro segregation

In general, microsegregation in ingots and castings is caused by the interplay between a number
of variables which cannot readily be accounted for in a mathematical simulation of the proc-
ess 12 Nevertheless, several important deductions can be made from the well known non-
equilibrium lever rule or Scheil equation which, in spite of its simple nature, gives a reason-
able description of the segregation pattern during weld metal solidification.4'34
The Scheil equation applies to directional solidification of long bars and is based on the
following assumptions:
(i) Uniform liquid composition (i.e. complete mixing in the liquid state).
(ii) A flat solid/liquid interface.
(iii) Local equilibrium at the solid/liquid interface (ko is constant).
(iv) Negligible solid-state diffusion.
(v) Equal solid and liquid densities.
Under such conditions it is fairly simple to derive an expression for the solute concentration
in the metal as a function of the fraction solidified. Referring to Fig. 3.57, a mass balance
gives:
(3-41)
Since dz « z and Cs = ko C1, equation (3-41) can be rewritten as:

Previous Page
3.6.2 Nucleation mechanisms

Several nucleation mechanisms have been proposed to explain the columnar to equiaxed tran-
sition during weld pool solidification, including:26

The former mechanism is particularly relevant to welding, since the weld metal often con-
tains a high number of second phase particles which form in the liquid state. As already
mentioned in Section 3.4.4.2, these particles can either be primary products of the weld metal
deoxidation or stem from reactions between specific alloying elements which are deliberately
introduced into the weld pool through the filler wire. The important effect of deoxidation
practice (inclusions) on the columnar to equiaxed transition in ferritic stainless steel GTA
welds is shown in Figs. 3.55 and 3.56.
3.7 Solute Redistribution
During solidification of fusion welds, alloying and impurity elements tend to segregate exten-
sively to the centre parts of the intercellular or interdendritic spaces under the conditions of
rapid cooling. This, in turn, alters the kinetics of the subsequent solid state transformation
reactions.
3.7.1 Micro segregation

In general, microsegregation in ingots and castings is caused by the interplay between a number
of variables which cannot readily be accounted for in a mathematical simulation of the proc-
ess 12 Nevertheless, several important deductions can be made from the well known non-
equilibrium lever rule or Scheil equation which, in spite of its simple nature, gives a reason-
able description of the segregation pattern during weld metal solidification.4'34
The Scheil equation applies to directional solidification of long bars and is based on the
following assumptions:
(i) Uniform liquid composition (i.e. complete mixing in the liquid state).
(ii) A flat solid/liquid interface.
(iii) Local equilibrium at the solid/liquid interface (ko is constant).
(iv) Negligible solid-state diffusion.
(v) Equal solid and liquid densities.
Under such conditions it is fairly simple to derive an expression for the solute concentration
in the metal as a function of the fraction solidified. Referring to Fig. 3.57, a mass balance
gives:
(3-41)
Since dz « z and Cs = ko C1, equation (3-41) can be rewritten as:

Fraction equiaxed grains (%)
(a)
Aluminium content, wt%

Average grain size, Jim
(b)
Aluminium content, wt%
Fig. 3.55. Effect of deoxidation practice (aluminium additions) on the columnar to equiaxed transition in
ferritic stainless steel GTA welds; (a) Average fraction of equiaxed grains observed at the surface of the
welds vs aluminium content, (b) Surface equiaxed grain size vs aluminium content. Data from Villafuerte
etal21
(3-42)
Integration of this equation gives:
(3-43)
(a)
Fraction equiaxecl grains (%)
(b)
Average grain size, JI m
Fig. 3.56. Effect of deoxidation practice (titanium additions) on the equiaxed transition in ferritic stain-
less steel GTA welds; (a) Average fraction of equiaxed grains observed at the surface of the weld vs
titanium content, (b) Surface equiaxed grain size vs titanium content. Data from Villafuerte et al.21
from which
(3-44)
and
(3-45)
where fs is the fraction solidified (equal to z/L*).

Solid Liquid
Concentration, C
Distance, z
Fig. 3.57. Solute redistribution during non-equilibrium freezing according to the Scheil equation.
Equation (3-44) is valid up to CL = Ceut where the remaining melt solidifies in the form of
intercellular or interdendritic eutectics. The eutectic fraction feut is, in turn, given as:
(3-46)
Figure 3.58 shows how the Scheil equation can be used for an evaluation of the
microsegregation pattern in binary alloy systems by considering a small volume element of
length L* which solidifies perpendicular to the cell/dendrite growth direction.
Since the Scheil equation does not allow for solid state diffusion during solidification, a
slightly refined version of this equation also exists in the literature:1
(3-47)
where a* is a dimensionless diffusion parameter, defined as:
(3-48)
Solid
Liquid
Fig. 3.58.Idealised model for microsegregation in ingots and castings (schematic).
Equation (3-47) has been used by Brody and Flemings34 to evaluate the effect of solid
diffusion on the amount of eutectic in different cast structures. The extent of this diffusion
depends on the dimensionless product a* ko and it becomes significant only for values of
a* Jc0 greater than about O.I.1 Note that in cellular or cellular-dendritic growth the primary
dendrite arm spacing X1 provides a measure of the diffusion length. Taking L* = X ,/2, equation
(3-48) can be rewritten as:
(3-49)
In contrast, during equiaxed dendritic growth the secondary dendrite arm spacing X2 is a
more appropriate dimension for the solidification microstructure, since the back diffusion process
here occurs mainly between secondary arms and not between primary trunks.2 In such cases
we may write:
(3-50)
Example (3.13)
Consider GTA welding of an Al-2wt%Cu alloy under the following solidification condi-
tions:
Welding speed
Cooling rate
Primary dendrite arm spacing
Estimate on the basis of the Scheil equation the degree of microsegregation occurring dur-
ing weld metal solidification. Relevant physical data for the Al-Cu system are given below:
Solution
The local solidification time can be calculated from equation (3-33). Taking RL = v, we get:
Since the solidification conditions in this case facilitate the formation of a cellular-dendritic
type of substructure close to the weld centre-line (see Fig. 3.43), the characteristic diffusion
length L* is determined by the primary dendrite arm spacing X1. The product a* ko is then
given as:
Because the numerical value of a* ko is very small, the contribution from diffusion in the
solid state can be neglected. Hence, the extent of microsegregation occurring during solidifi-
cation can be evaluated from equation (3-45). Taking C0 = 2wt% and ko = 0.17, we obtain:
The results are presented graphically in Fig. 3.59.

As expected, the copper concentration is seen to increase monotonically from the core to
the periphery (surface) of the dendrite stem. When the eutectic composition is reached, the
remaining fraction solidifies as Al(5.6wt%Cu) + CuAl2. According to equation (3-46) the
eutectic fraction is equal to:
These results should be compared with the experimental data of Brooks and Baskes35
replotted in Fig. 3.60. It is evident that the measured copper concentration profile for the Al-
2wt%Cu GTA weld is similar to that inferred from the Scheil equation, although the observed
dendrite core concentration lies significantly above the predicted one. Consequently, the Scheil
equation gives a reasonable description of the segregation pattern during weld metal solidifi-
cation, in spite of the simplifying assumptions inherent in the model.
z.^m
Cu concentration, wt%
Normalized distance (z/L*)

Fig. 3.59.Predicted Cu concentration profile based on the Scheil equation (Example (3.13)).
GTAW (Al-2wt% Cu)

Cu concentration, wt%
Position, ]im
Fig. 3.60. Electron microprobe analysis of Cu across primary solidification (cell) boundaries in an Al-
2wt% Cu GTA weld. Data from Brooks and Baskes.35
3.7.2 Macro segregation

Macrosegregation in the form of solute banding is also frequently observed after weld metal
solidification.34 This type of segregation arises from a periodic enrichment and/or depletion
of solute elements, caused by a non-steady advancement of the solidification front (see Fig.
Fig. 3.6/. Example of solute banding (dark stripes) in a phosphorus-containing copper weld. After Gar-
land and Davis.36
3.61). Although the origin of the phenomenon is not yet fully understood, it is reasonable to
assume that the pertinent fluctuations in the solidification rate occurs as a result of frequent
variations in the heat flux during welding. Direct experimental evidence for such a correlation
can be obtained from the data of Garland and Davis.36
Moreover, there is a pronounced tendency for alloying and impurity elements to segregate
to the weld centre-line where the columnar grains growing from each side impinge. This, in
turn, may produce hot tearing as a result of the formation of low-melting eutectics between the
dendrite arms. In general, the risk of hot tearing decreases with increasing width to depth ratio
of the weld because of a more favourable crystal growth mode, as illustrated in Fig. 3.62.
3.7.3 Gas porosity

Troublesome impurities in fusion welds are those which precipitate, alone or in combination
with other elements, to form various gaseous reaction products.37 These reaction products may
be simple diatomic gases such as H2 and N2 or more complex gaseous compounds like CO, or
H2O. An illustration of gas porosity due to hydrogen evolution is contained in Fig. 3.63.
Following the discussion in Chapter 2, absorption of gases occurs readily in the hot part of
the weld pool during welding because of interactions with the surrounding arc atmosphere.
During the subsequent stage of cooling a supersaturation rapidly increases due to the associ-
ated decrease in the element solubility with decreasing temperatures (see Figs. 2.7 and 2.8).
The system will respond to this supersaturation by rejection of dissolved elements from the
liquid, either through a gas/metal reaction (desorption) or by precipitation of new phases. The
latter incident may result in porosity if the gas bubbles become trapped in the weld metal
solidification front.
3.7.3.1 Nucleation of gas bubbles

In liquid metals a bubble will be stable if the gas pressure inside the bubble is sufficiently high
to balance the external forces. These external forces are the liquid/vapour interfacial energy
(a) Correct width to depth ratio (b) Incorrect width to depth ratio
Width Width
Depth
Depth
Fig. 3.62. Effect of weld width to depth ratio on the tendency to centre-line cracking; (a) Correct width to
depth ratio, (b) Incorrect width to depth ratio.
Fig. 3.63. Example of gas porosity in a GTA 7106 aluminium weld. After D'annessa.39
Transport of dissolved Liquid

gaseous species
Gas bubble
Solid
Fig. 3.64. Growth of a gas bubble due to diffusion in the liquid phase (schematic).
a*, the metallostatic pressure head/?m, and the ambient pressure pa. Thus, for a stable bubble,
we may write:1
(3-51)
where pg is the total gas pressure inside the bubble, and rg is the radius of the gas bubble.
For shallow welds, the contribution from the metallostatic pressure head pm can be ig-
nored. Hence, equation (3-51) reduces to:
(3-52)
Since a* is typically of the order of 1 J m 2 (9.87 atm |im) for most gas-metal systems, we
may write:
(3-53)
It is evident from equation (3-53) that the interfacial energy term is negligible at large
values of r . However, if the radius of curvature becomes sufficiently small, extremely large
pressures are required to maintain a stable bubble. Thus, there is a bubble nucleation prob-
lem, which is formally similar to that of nucleation of a solid from a liquid (discussed in
Section 3.3.1). In fact, it can be shown on the basis of classic nucleation theory that the driving
force normally associated with rejection of dissolved gases in liquid metals is by far too small
to allow for homogeneous nucleation of gas bubbles in the weld pool during cooling. This, in
turn, implies that solid particles (e.g. inclusions) entrained in the liquid metal will be the most
probable sites for gas bubble formation in fusion welds.
3.7.3.2 Growth and detachment of gas bubbles

As shown in Fig. 3.64, growth of gas bubbles in liquids is a diffusion-controlled process where
the time dependence of the mean bubble radius is given by:38
(3-54)
Here Q.* is the growth constant, defined as:
(3-55)
where CL is the molar concentration of solute in the supersaturated liquid, C6 is the equilibrium
molar concentration of the solute at the gas/liquid interface, and p^ is the gas density (in the
same units as CL and C6).
Equations (3-54) and (3-55) may be used to estimate the growth rate of a bubble while it is
still attached to the solid/liquid interface. The bubble becomes detached when the buoyancy
force, which is pushing it upwards, exceeds the surface tension force, which tends to keep it
attached to the solid surface. The bubble radius at which detachment occurs is given by the so-
called Fritz equation:38
(3-56)
where gc is the gravity constant, and P is the wetting angle (in degrees).
Based on equations (3-54) and (3-56) it is possible to evaluate the conditions for growth
and detachment of gas bubbles during weld metal solidification. This is a subject of consider-
able importance in welding, since the pore formation will inevitably affect the mechanical
integrity of the weldment.
Example (3.14)
Consider GTA butt welding of a 3mm thin Al-Mg sheet under the following conditions:
Suppose that gas bubbles form at the solid/liquid interface during solidification due to re-
jection of dissolved hydrogen from a supersaturated liquid. Based on equations (3-54) and
(3-56) estimate the maximum theoretical radius of the gas bubbles and the critical radius at
which the bubbles detach themselves from the solid/liquid interface during welding. Relevant
physical data for the Al-Mg system are given below:
Solution
Since particles located at the solid/liquid interface are the most probable sites for hydrogen
gas evolution, the local solidification time t0 provides a conservative estimate of the growth
time t in equation (3-54). From equation (3-36), we have:
In order to calculate the growth constant from equation (3-55), it is necessary to convert the
concentration driving force to molar units:
The molar density of the gas p is obtained from the ideal gas law:
The value of the growth constant is thus:
The maximum theoretical radius of the gas bubbles can now be evaluated from equation
(3-54) by inserting the appropriate values for Q*, DH, and to:
Similarly, the critical radius at which the hydrogen bubbles become detached may be esti-
mated from equation (3-56). Since PM > > PH 2 , we obtain:
By inserting this value into equation (3-54) it is also possible to estimate the average bubble
detachment frequency under the prevailing circumstances:
Since r (crit.)« r (max), the maximum pore radius will probably be closer to 0.7mm than
30mm in a real welding situation. This is also in agreement with practical experience (see Fig.
3.65).
GTAW (Aluminium)
Cumulative probability (%)
Diameter of pores, mm
Fig. 3.65. Measured distribution of pore diameters in some GTA aluminium welds deposited with differ-
ent hydrogen-containing shielding gases. Data from Tomii et al.40
3.7.3.3 Separation of gas bubbles

It follows that the gas bubbles will start to migrate towards the surface of the weld immediately
after they become detached from the solid/liquid interface. Small, spherical bubbles (charac-
terised by a bubble Reynold number less than 2) will rise at a terminal velocity determined by
Stokes law:38
(3-57)
where dg is the diameter of the gas bubbles, and jLl is the viscosity of the liquid.
Larger gas bubbles (characterised by a bubble Reynold number between 2 and 400) will
also rise in a rectilinear manner, but their terminal velocity may be as much as 50% greater
than that predicted from Stokes law.38 Depending on their flotation rate, such ascending gas
bubbles will either escape to the weld surface or be trapped in the weld metal solidification
front in the form of macroscopic gas porosity (see Fig. 3.63).
Example (3.15)
Based on Stokes law (equation (3-57)) calculate the rising velocity of a 0.2mm large hydrogen
bubble ascending in liquid aluminium. Relevant physical data for liquid aluminium are given
below:
Solution
Since p z » p , we may write:
It is evident from the above calculations that the flotation rate of such gas bubbles is quite
high and of the same order of magnitude as the weld pool fluid flow velocity (discussed in
Section 2.11.2). Hence, the buoyancy force would be expected to play a significant role in the
separation process of detached gas bubbles in the weld pool. On this basis it is not surprising to
find that a change in the welding position (e.g. from flat to overhead) results in a dramatic
increase in the volume of porosity during GTAW of aluminium alloys (see Fig. 3.66).
Although a great deal has been reported on the causes and effects of porosity in weld metals
(see Ref. 37 for an excellent discussion), little is known about the mechanism of pore forma-
tion relative to solidification mechanics, nucleation, growth and transport of gas bubbles in the
weld pool. Consequently, a more fundamental approach to the porosity problem in fusion
welding (along the lines indicated above) is necessary in order to obtain a verified, quantitative
understanding of the phenomenon.
GTAW
Overhead position
Volume of pores, ml/10Og
Flat position
Welding orientation, degrees
Fig. 3.66. Porosity in GTA welds deposited on 2mm sheets of aluminium at various orientations from flat
to overhead. Data compiled by Devletian and Woods.37
3.7.4 Removal of microsegregations during cooling
As shown in Section 3.7.1, the characteristic growth pattern of cellular and dendritic solidifi-
cation, in combination with the rapid cooling rates normally associated with fusion welding,
lead to extensive segregation of alloying and impurity elements to the intercellular or
interdendritic spaces. Segregation produced by this means is remarkably persistent, and can in
certain cases only be eliminated by prolonged high-temperature heat treatment. A simplified
analysis of homogenisation of microsegregations in fusion welds is given below.
3.7.4.1 Diffusion model

It is a reasonable approximation to regard microsegregations in cast structures as periodic,
where the concentration at any point and time along an arbitrary line, C(x,t), is given by:41
where C and Cmax are as indicated in Fig. 3.67, lavg is the average distance between adja-
cent maxima and minima, and Ds is the diffusivity of the solute in the solid.
Equation (3-58) states that the concentration remains constant and equal to Cavg at positions
JC = 0, JC = lavg, x = 2/ etc., while the peak of the sine wave is attained at distances x = lavg /2,
x = 5lavg /2 etc. during the decay. If only the peak concentration is considered, the sine term
becomes equal to unity and equation (3-58) reduces to:
(3-59)
from which the homogenisation time during isothermal heat treatment thom can be obtained:
(3-60)
Since the diffusion length I , in practice, is equal to the half dendrite arm spacing (X1 /2 or
X2/2), equation (3-60) predicts that the homogenisation time is proportional to the square OfX1
or X2. The latter parameters are, in turn, determined by the thermal conditions existing within
the mushy zone during weld metal solidification, and are therefore sensitive to variations in
welding variables such as the net arc power qo and the travel speed v.
3.7.4.2 Application to continuous cooling

During cooling of the weld in the solid state, some equalisation of microsegregations will
occur through diffusion. The extent of this diffusion can be reported in terms of an equivalent
isothermal homogenisation time at a chosen reference temperature Tr. If Qs denotes the activa-
tion energy for diffusion of the solute in the solid, the equivalent isothermal homogenisation
time tr at Tr is given by:
(3-61)
Dendrite arm spacing
Concentration
Fig. 3.67. Relaxation of sinusoidal distribution of solute during isothermal annealing.
The integral on the right-hand side of equation (3-61) represents the kinetic strength of the
weld thermal cycle with respect to homogenisation (solid state diffusion), and can be deter-
mined by means of numerical methods when the cooling programme is known. The extent of
solute diffusion may then be evaluated from equation (3-60) by inserting representative
values for t and / :
T UY'g.
(3-62)
Example (3.16)
[n low-alloy steel, both carbon, phosphorus, and manganese are known to segregate to the
interdendritic spaces during solidification. Consider SA welding on a thick plate of steel
Based on equations (3-61) and (3-62) calculate the extent of homogenisation occurring
within the solid weld metal during cooling in the austenite regime (i.e. from 1520 to 7000C).
Relevant data for the diffusivity of carbon, phosphorus, and manganese in austenite are given
below:
Solution
During thick plate welding, the cooling programme can be calculated from equation (1-45).
For points located on the weld centre-line behind the heat source y = z = 0, and R* = -x = vt.
Hence, equation (1-45) reduces to:
By inserting representative values for T0, qo, v and \ , we obtain:
From this it is seen that the actual transformation temperatures (i.e. 1520 and 7000C) are
reached after 20.9 and 46.2s, respectively. If 13500C is used as a reference temperature, the
kinetic strength of the cooling cycle with respect to solute diffusion can be expressed as:
Numerical integration of this equation over the weld cooling cycle gives:
At 13500C (1623 K) the diffusivities of C, P and Mn in austenite are:
The extent of homogenisation is thus:

It is evident from the above calculations that interdendritic segregations of carbon are
readily removed during cooling of the weld from the solidification temperature due to a high
diffusivity of C in austenite. In contrast, segregations of phosphorus and manganese are much
more persistent, since the diffusion cannot keep pace with the falling temperature. In the lat-
ter case the initial concentration gradients will largely be maintained down to temperatures
where the austenite to ferrite transformation occurs (see ion micrograph of phosphorus seg-
regations in Fig. 3.39). This, in turn, can promote the formation of different types of trans-
formation products (ranging from ferrite to martensite depending on the nature of the segre-
gants) along the primary solidification boundaries due to local variations in the steel
hardenability. An example of martensite banding in an AISI 4340 SA steel weld is shown in
Fig. 3.68.
It should be noted that the conditions for equalisation of interdendritic segregations (e.g.
phosphorus) are not dramatically altered by a change in the heat input as long as the tr I l\vg
ratio in equation (3-62) remains fairly constant. For instance, if the net arc power qQ in the
above example is reduced by a factor of four (conforming to welding with covered electrodes),
the corresponding change in / and J1350 (P) will be:
and
from which
This value is reasonably close to that obtained during high-heat input SA welding
(0.036 s jirrr2), as shown in the above example.
Fig. 3.68. Banding of martensite (M) along primary solidification boundaries in an AISI 4340 SA steel
weld. After Burck.42
3.8 Peritectic Solidification
Referring to the schematic phase diagram in Fig. 3.69, crystallisation within peritectic systems
starts with primary precipitation of the 8p-phase from the liquid. At the peritectic temperature
T a new solid phase 7 forms, according to the reaction:
5p+ liquid -> yp <3"63)
Because of the nature of this reaction, there is a strong tendency for the secondary phase 7
to grow along the 8^/liquid interface and, thus, to isolate the primary phase from contact with
the liquid.43 Depending on the growth mode, the decomposition of the primary 8^-phase is
said to occur either through a peritectic reaction or by a peritectic transformation,44 as
shown schematically in Fig. 3.70 and 3.71, respectively.
3.8.1 Primary precipitation of the y -phase

Primary precipitation of the 7^-phase from the liquid is also possible if the temperature is
gradually decreasing during solidification.43 As shown by Kerr et al.44^6 this type of precipi-
tation is favoured by the presence of solid particles within the liquid metal which can act as
seed crystals for the 7^-phase in a selective manner. Particularly at high cooling rates, the for-
mation of the primary 8^-phase can be completely suppressed, thereby allowing the secondary
7 -phase to nucleate and grow directly from the supercooled liquid.
3.8.2 Transformation behaviour of low-alloy steel weld metals

In general, low-alloy steels undergo multiple phase transformations during solidification and
subsequent cooling. Depending on the cooling rate, carbon or the substitutional alloy content,
the primary solidification product will either be delta ferrite 8 Fe , austenite 7 Fe or a mixture of
both. 47 ^ 9
3.8.2.1 Primary precipitation of delta ferrite
When a low-alloy steel is slowly cooled below the crystallisation temperature, delta ferrite
will be the first phase to form.50 In fusion welds, the delta ferrite grains will reveal an anisotropic
columnar morphology, with their major axes aligned in the direction of the steepest tempera-
ture gradient in the weld pool. On further cooling below the peritectic temperature, austenite
nucleates epitaxially at the primary delta ferrite grain boundaries, as shown schematically in
Fig. 3.72, since these sites provide the lowest energy barrier against heterogeneous nucleation
(requires a Kurdjumow-Sachs or a Nishiyama-Wasserman type of orientation relationship
between 8Fe and 7Fe).51 Subsequent growth of the austenite into the delta ferrite may then
proceed by a peritectic transformation at a rate which is controlled by diffusion of carbon in
the austenite. This reaction pattern is analogous to that documented for growth of 7 F e -
Widmanstatten sideplates in duplex stainless steel.5253
Since the austenite is bound by an orientation relationship with the delta ferrite,51 the
austenite cannot grow across the primary delta ferrite solidification boundaries. Hence, after
the peritectic transformation the columnar 7Fe grains will adopt the original 8Fe morphology,
as shown in Fig. 3.72. Direct experimental evidence for such a correspondence can be ob-
tained from the optical micrographs in Fig. 3.38 and the texture data presented in Section 6.3.5
(Chapter 6).
Liquid (L)
Temperature
Fig. 3.69.Peritectic phase diagram (schematic).
Growth direction
Liquid Diffusion of
solute atoms
Fig. 3.70.Peritectic reaction by which the secondary 7p-phase grows along the surface of the primary
8p-phase (schematic).
Liquid
Fig. 3.7L Peritectic transformation involving long-range diffusion of solute atoms through the secondary
7^-phase (schematic).
Delta ferrite
Liquid
Solute segregations
Austenite
Austenite grain
boundary
Fig. 3.72. Primary delta ferrite solidification with subsequent growth of austenite along the boundaries of
the primary 8Fe-phase (schematic).
3.8.2.2 Primary precipitation of austenite

Under certain welding conditions the weld metal can solidify directly as austenite without any
primary precipitation of delta ferrite (see Fig. 3.73). This solidification mode has been ob-
served during welding with covered electrodes on high-carbon steels.54 Because of competi-
tive growth occurring during the initial stage of the solidification process, the columnar austenite
grains will have one of their <100> axes aligned in the direction of the maximum temperature
gradient in the weld pool. Consequently, the weld metal will develop a solidification texture
which is similar to that observed during primary precipitation of delta ferrite.51
3.8.2.3 Primary precipitation of both delta ferrite and austenite

In certain low-alloy steel weld metals the austenite grain boundaries will systematically cross
the original delta ferrite solidification boundaries, as indicated in Fig. 3.74 because of a shift in
the mechanism of the peritectic transformation.3'22'2454"56 This point is more clearly illus-
trated in Fig. 3.75, which shows ion (SIMS) micrographs of phosphorus and boron segregations
Austenite
Columnar
grain
Liquid
Fig. 3.73. Primary austenite solidification (schematic).

Fig. 3.74. Optical micrograph showing austenite grain boundaries crossing primary delta ferrite solidifi-
cation boundaries in a low-alloy SA steel weld (indicated by arrows). After Kluken et al.22
at prior solidification (cell) and austenite grain boundaries*, respectively in a SA steel weld
metal deposited under the shield of a basic flux. It is obvious from Fig. 3.75 that there is no
matching between the two types of boundaries in this particular case.
It follows from the analysis of Kluken et al?2 that the observed shift in the mechanism of
the peritectic transformation can probably be attributed to heterogeneous nucleation of
austenite at inclusions (e.g. Al2O3), which is energetically more favourable than nucleation at
8Fe/8Fe grain boundaries. Under such conditions, the austenite is not bound by an orientation
relationship with the delta ferrite and is thus free to grow across the original delta ferrite
columnar grain boundaries, as shown schematically in Fig. 3.76. The austenite grains will
therefore adopt a morphology which is different from the columnar one.245556
References
1. M.C. Flemings: Solidification Processing, 1974, New York, McGraw-Hill Book Company.
2. W. Kurz and DJ. Fisher: Fundamentals of Solidification, 3rd Edn, 1989, Aedermannsdorf
(Switzerland), Trans. Tech. Publications.
3. GJ. Davis and J.G. Garland: Int. MetalL Rev., 1975, 20, 83-106.
4. S.A. David and J.M. Vitek: Int. Mater. Rev., 1989, 34, 213-245.
5. D. A. Porter and K.E. Easterling: Phase Transformations in Metals and Alloys, 1981, Wokingham
(England), Van Nostrand Reinhold Co. Ltd.
6. K.E. Easterling: Introduction to the Physical Metallurgy of Welding, 1983, London, Butterworths
& Co (Publisher) Ltd.
7. 0. Grong, TA. Siewert and G.R. Edwards: Weld. J., 1986, 65, 279s-288s.
8. G.M. Oreper, TW. Eagar and J. Szekely: Weld. J., 1983, 62, 307s-312s.
9. Y.H. Wang and S. Kou: Proc. Int. Conf. on Advances in Welding Science, Gatlinburg, TN,
May, 1986, 65-69, Publ. ASM International.
*Because of a similar atomic weight of BO2~ and AlO~, both types of ions will be sampled in the SIMS analysis.
Consequently, the white spots observed within the interior of the grains are probably traces of aluminium oxide
inclusions.
(a)
(b)
Fig. 3.75. Ion (SIMS) micrographs showing austenite grain boundaries crossing primary delta ferrite
solidification boundaries in a low-alloy SA steel weld; (a) Phosphorus segregations at primary solidifica-
tion (cell) boundaries, (b) Boron segregations at prior austenite grain boundaries (same area as in Fig.
3.75(a)). After Kluken et alP
Delta ferrite.
Liquid
Solute segregations
Inclusions (e.g. AI2O3)
Austenite Austenite grain

boundary
Fig. 3.76. Primary delta ferrite solidification with subsequent nucleation of austenite at inclusions (sche-
matic).
10. K.C. Mills and BJ. Keene: Int. Mater. Rev., 1990, 35, 185-216.
11. T. Zacharia, S.A. David, J.M. Vitek and H.G. Kraus: Metall. Trans., 1991, 22B, 243-257.
12. CR. Heiple and J.R. Roper: Weld. J., 1982, 61, 97s-102s.
13. CR. Heiple and P. Burgardt: Weld. J., 1985, 64, 159s-162s.
14. T. Habrekke, H.O. Knagenhjem and J.O. Berge: Proc. 9th. Int. Conf. on Offshore Mech. and
Arctic Engin., Houston, TX, Feb., 1990, 511-515, Publ. ASME.
15. H. Biloni: Physical Metallurgy, 3rd Edn (Eds R.W. Chan and P. Haasen), 1983, Amsterdam,
North-Holland Physics Publ., 478-579.
16. T. Senda, F. Matsuda and M. Kato: Techn. Reports of the Osaka University (Japan), 1970,20,
527-558.
17. M. Rappaz, S.A. David, J.M. Vitek and L.A. Boatner: Metall. Trans. A, 1989,2OA, 1125-1138.
18. L.F. Mondolfo: Mat. ScL and TechnoL, 1989, 5, 118-122.
19. B.L. Bramfitt: Metall. Trans., 1970, 1, 1987-1995.
20. G.N. Heintze and R. McPherson: Weld. /., 1986, 65, 71s-81s.
21. J.C. Villafuerte, E. Pardo and H.W. Kerr: Metall. Trans. A, 1990, 21A, 2009-2019.
22. A.O. Kluken, 0. Grong and G. R0rvik: Metall. Trans. A, 1990, 21 A, 2047-2058.
23. H. Yunjia, R.H. Frost, D.L. Olson and G.R. Edwards: Weld. J., 1989, 68, 280s-289s.
24. CE. Cross, 0. Grong, S. Liu and J.F. Capes: Applied Metallography, (Ed. G.F. Vander Voort),
1986, New York, Van Nostrand Reinhold Inc., 197-210.
25. A.O. Kluken and 0 . Grong: Proc. Int. Conf. on Recent Trends in Welding ScL and TechnoL
(TWR 89), Gatlinburg, TN, May, 1989, 781-786, Publ. ASM International.
26. S. Kou: Welding Metallurgy, 1987, New York, John Wiley & Sons.
27. M.F. Jordan and M.C. Coleman: Brit. Weld. J., 1968,15, 553-558.
28. CE. Cross and G.R. Edwards: Treatise on Mat. ScL TechnoL, 1989, 31, 171-187.
29. W. Kurz, B. Giovanola and R. Trivedi: Acta Metall., 1986, 34, 823-830.
30. R. Trivedi and W. Kurz: Acta MetalL, 1986, 34, 1663-1670.
31. S. Kou and Y. Lee: MetalL Trans., 1982,13A, 1141-1152.
32. A. Munitz: MetalL Trans., 1985,16B, 149-161.
33. F. Matsuda, K. Nakata, K. Tsukamoto and K. Arai: Trans. JWRI, 1983,12, 81-87.
34. H.D. Brody and M.C. Flemings: Trans. AIME, 1966, 236, 615-623.
35. J.A. Brooks and M.I. Baskes: Proc. Int. Conf. on Trends in Welding Research (TWR 86),
Gatlinburg, TN, May, 1986, 93-99, Publ. ASM International.
36. J.G. Garland and GJ. Davis: Brit. Weld. J., 1970,17, 171-175.
37. J.H. Devletian and W.E. Woods: Weld. Res. Council Bull., 1983, 290, 1-18.
38. J. Szekely and NJ. Themelis: Rate Phenomena in Process Metallurgy, 1971, New York, John
Wiley & Sons, Inc.
39. A.T. D'annessa: Weld. J., 1967, 46, 491s-499s.
40. Y. Tomii, A. Sakaguchi and M. Mizuno: Proc. 4th Int. Conf. on Aluminium Weldments, Tokyo
(Japan), 1988, 46-60, Publ. Japan Light Metal Welding and Construction Association.
41. J.D. Verhoeven: Fundamentals of Physical Metallurgy, 1975, New York, John Wiley & Sons,
Inc.
42. P. Burck: M.Sc. thesis, 1984, Colorado School of Mines, Golden, Colorado (USA).
43. M. Hillert: Solidification and Casting of Metals, 1979, London, The Metals Society, 81-87.
44. H. W. Kerr, J. Cisse and G.F. Boiling: Acta MetalL, 1974, 22, 677-686.
45. J. Cisse, G.F. Boiling and H.W. Kerr: /. Cryst. Growth, 1972,13/14, 777-781.
46. J. Cisse, H.W. Kerr and G.F. Boiling: MetalL Trans., 1974, 5, 633-641.
47. H. Fredriksson: Met. ScL, 1976,10, 77-86.
48. H. Fredriksson and J. Stjerndahl: MetalL Trans., 1977, 8A, 1107-1115.
49. H. Fredriksson and J. Stjerndahl: Met. ScL, 1982,10, 575-585.
50. N.S. Pottore, CL Garcia and AJ. DeArdo: MetalL Trans., 1991, 22A, 1871-1880.
51. A.O. Kluken, 0. Grong and J. Hjelen: MetalL Trans., 1991, 22A, 657-663.
52. N. Suutala, T. Takalo and T. Moisio: MetalL Trans., 1979,10A, 1183-1190.
53. S.A. David: Weld. J., 1981, 60, 63s-71s.
54. A.A.B. Sugden and H.K.D.H. Bhadeshia: MetalL Trans., 1987,19A, 669-674.
55. J.G. Garland and RR. Kirkwood: Proc. Int. Symp. on Welding of Line Pipe Steels, St. Louis
(USA), 1977, 176-227, Publ. Welding Research Council (New York).
56. R.C. Cochrane: Weld, in the World, 1983, 21, 16-29.
Appendix 3.1
Nomenclature
thermal diffusivity difference between total and

(mm2 s-1) acid soluble Al in weld metal
(wt%)
lattice parameter (nm)
b half width of dendrite stem
equilibrium 7/a solvus tem- (|Lim)
perature (°C)
bcc body-centred cubic structure
area of embryo-liquid inter-
face (m2) various kinetic constants and
temperature-dependent pa-
area of embryo-substrate in- rameters
terface (m2)
average solute concentration eutectic fraction
in solid phase (wt%)
fraction solidified
initial alloy concentration
(wt%) shape factor
equilibrium concentration of face-centred cubic structure

solute at gas/liquid interphase
(wt% or mol cirr3) length of dendrite stem (jam)
eutectic concentration (wt%) gravity constant (m s"2)
concentration of alloying el- temperature gradient in weld

ement in liquid phase (wt% pool (°C mm"1 or K irr 1 )
or mol cm"3)
maximum solute concentra-
tion in solid phase (wt%) GTAW gas tungsten arc welding
concentration of alloying el- activation energy for diffu-

ement in solid phase (wt%) sion of atoms across the in-
terface (J)
solute concentration at posi-
tion x and time t (wt%) energy barrier to heterogene-
ous nucleation (J)
cooling rate (0C s~l or K s"1)
energy barrier to homogene-
plate thickness (mm) ous nucleation (J)
the interatomic spacing along free energy change associ-

[MVW] n ated with the embryo forma-
tion (J irr 3 )
the interatomic spacing along
[uvw\ heat content per unit volume
at the melting point (J mm"3)
diffusion coefficient in liquid
(mirr s [) latent heat of melting (J irr 3 )
diffusion coefficient in solid a low-index plane in the nu-

(mm2 s-1) cleated solid
diffusion coefficient of arbi- a low-index plane of the

trary element X substrate
(mm2 s~!)
amperage (A)
EBW electron beam welding
Boltzmann constant (1.381
frequency factor (s"1) X 10" 23 JK- 1 )
detachment frequency of gas equilibrium partition coeffi-

bubbles (s"1) cient
critical GJR1 ratio which fa- modified Bessel function of

cilitates equiaxed dendritic second kind and zero order
growth(°Csmm" 2 )
length of columnar grain radius of spherical nucleus
(mm) (m)
average distance between ad-

critical radius of stable nu-
jacent maxima and minima
concentrations (|im) cleus (m)
total length of circle segment universal gas constant (8.314

(mm) J K-1 mol-1)
total length of volume ele- three-dimensional radius

ment (mm or Jim) vector (mm)
slope of the liquidus curve dendrite growth rate in hkl

(°C per wt%) direction (mm s"1)
mobility term (mm3 s"1) steady-state growth rate

of solid/liquid interface
exponent (mm s"1)
dimensionless operating pa- local crystal growth rate (mm

rameter
nominal crystal growth rate

rate of heterogeneous nuclea- (mm s'1)
tion (nuclei per s and mm3)
SAW submerged arc welding
density of nucleation sites per
unit volume of melt (nuclei SIMS secondary ion mass
per mm3) spectrometry
analytical weld metal oxygen SMAW shielded metal arc welding

content (wt%)
entropy of fusion per unit
ambient pressure (atm) volume (J nr 3 Kr1)
total gas pressure in bubbles time (s)

(atm)
integration limits (s)
metallostatic pressure head
(atm) homogenisation time (s)
net arc power (W) local solidification time (s)
activation energy for diffu- equivalent isothermal ho-

sion of solute in solid (kJ mogenisation time (s)
mol-1)
temperature (°C or K)
two-dimensional radius vec-
tor (mm) ambient temperature (°C or
K)
dendrite tip radius (Jim)
actual temperature in the liq-
radius of gas bubble (mm) uid (°C or K)
liquidus temperature accord- primary peritectic phase
ing to phase diagram (°C or
K) efficiency factor
melting point (°C or K) interface normal angle (de-

grees)
peritectic temperature (°C or
K) dimensionless diffusion pa-
rameter
reference temperature (°C or
K) molar density of gas (mol
cm"3)
dendrite root temperature (°C
or K) density of liquid (kg n r 3 or
g cm"3)
dendrite tip temperature (°C
or K) volume heat capacity
(J mm~3 0C"1)
undercooling (°C or K)
cell/dendrite alignment angle
temperature difference be- (degrees)
tween tip and root of dendrite
stem (°C or K) critical cell/dendrite align-
ment angle (degrees)
a low-index direction in
(UcQn Gibbs-Thomson coefficient
(mm °C or mK)
a low-index direction in
W), dimensionless growth con-
stant
voltage (V)
dimensionless temperature
rising velocity of gas bubbles
(mm s"1) dimensionless temperature at
the melting point
volume of solid embryo (m3)
wetting angle (degrees)
welding speed (mm s"1)
viscosity of liquid metal (kg
Jt-axis/welding direction nr 1 s-1)
(mm)
angle between [MVW]^ and
y-axis/transverse direction [«vw]n (degrees)
(mm)
austenite
z-axis/through thickness di-
rection (mm) secondary peritectic phase
dimensionless plate thick- total grain rotation angle (de-

ness grees)
delta ferrite solid/liquid interfacial energy

(J mr2)
lattice disregistry between a
nucleus n and a substrate s liquid/vapour interfacial en-
ergy (J nr 2 )
embryo/liquid interfacial en- primary dendrite arm spacing
ergy (J rrr2) (Mm)
embryo/substrate interfacial primary dendrite arm spacing

energy (J nr 2 ) close to fusion boundary
(Mm)
substrate/liquid interfacial
energy (J rrr2) secondary dendrite arm spac-
ing (Mm)
thermal conductivity (W
mm-10C"1)
4
Precipitate Stability in Welds
4.1 Introduction
Precipitate stability is an important aspect of welding metallurgy. Normally, modern structural

steels and aluminium alloys derive their balanced package of high strength, ductility and tough-
ness via optimised thermomechanical processing to produce a fine-grained, precipitation
strengthened matrix. This delicate balance of microalloy precipitation and microstructure,
however, is significantly disturbed by the heat of welding processes, which, in turn, affects the
mechanical integrity of the weldment.
When a commercial alloy is subjected to welding or heat treatment several competitive
processes are operative which may contribute to a change in the volume fraction and size
distribution of the base metal precipitates. The two most important are:1
(i) Particle coarsening (Ostwald ripening)

(ii) Particle dissolution (reversion)
Referring to Fig. 4.1, particle coarsening occurs typically at temperatures well below the
equilibrium solvus Te of the precipitates, while particle dissolution is the dominating mecha-
nism at higher temperatures. On the other hand, there exists no clear line of demarcation
between these two processes, which means that particle coarsening can take place simultane-
ously with reversion in certain regions of the weld where the peak temperature of the thermal
cycle falls within the 'gray zone' in Fig. 4.1. Nevertheless, it is important to regard them as
separate processes, since the reaction kinetics are so different (coarsening is driven by the
surface energy alone, whereas dissolution, which involves a change in the total volume frac-
tion, is driven by the free energy change of transformation).
4.2 The Solubility Product
The solubility product is a basic thermodynamic quantity which determines the stability of
the particles under equilibrium conditions. Because of its simple nature, the solubility product
is widely used for an evaluation of the response of grain size-controlled and dispersion-hard-
ened materials to welding and thermal processing.23
4.2.1 Thermodynamic background

In general, the solubility product can be derived from an analysis of the Gibbs free energy AG°
of the following dissolution reaction:
Temperature
Increasing heating rate

Particle dissolution
'Grey zone1
Particle coarsening |
%B
Fig. 4.1. Schematic diagram showing the characteristic temperature ranges where specific physical
reactions occur during reheating of grain size-controlled and dispersion-hardened materials.
(4-1)
At equilibrium, we have:
(4-2)
where AH° and AS° are the standard enthalpy and entropy of reaction, respectively. The other
symbols have their usual meaning (see Appendix 4.1).
When pure An Bm is used as a standard state, the activity of the precipitate {aAn Bm) is equal
to unity. In addition, for dilute solutions it is a fair approximation to set aA~[%A] and aB~[%B],
where the matrix concentrations of elements A and B are either in wt% or at%*. Hence, the
solubility product can be written as:
(4-3)
*For the solute, the standard state is usually a hypothetical 1 % solution. This implies that the activity coefficient is
equal to unity as long as Henry's law is obeyed.
Table 4.1 gives a summary of equilibrium solubility products for a wide range of precipi-
tates in low-alloy steels and aluminium alloys.
In addition to the compounds listed in Table 4.1, different types of mixed precipitates may
form within systems which contain more than two alloying elements.3"6 However, since the
presence of such multiphase particles largely increases the complexity of the analysis, only
pure binary intermetallics will be considered below.
4.2.2 Equilibrium dissolution temperature

Based on equation (4-3) it is possible to calculate the equilibrium dissolution temperature Td of
the precipitates. By rearranging this equation, we get:
(4-4)
where [%A]o and [%B]o refer to the analytical content of elements A and B in the base metal,
respectively.
Equation (4-4) shows that the equilibrium dissolution temperature increases with increas-
ing concentrations of solute in the matrix. This is in agreement with the Le Chatelier's princi-
ple.
Table 4.1 Equilibrium solubility products for different types of precipitates in low-alloy steels and
aluminium alloys. Data compiled from miscellaneous sources.
Material/ phase Type of log [%A]n [%B]m

Precipitate C* = AS°/R' D* = Mi0IR'
TiN 0.32 8000
TiC 5.33 10475
NbN 4.04 10230
Low-alloy steel NbC 2.26 6770
(austenite)t VN 3.02 7840
VC 6.72 9500
AIN 1.79 7184
Mo2C 5.0 7375
Al-Mg-Si^ Mg2Si 5.85 5010
Al-Cu-Mg$ CuMg 6.64 4005
MgZn 5.33 2985
Al-Zn-Mgij: Zn2Mg 7.72 4255
All concentrations in wt%

All concentrations in at.%
Example (4.1)
Consider a low-alloy steel with the following chemical composition:
Calculate on the basis of the reported solubility products in Table 4.1 the equilibrium disso-
lution temperature of each of the following three nitride precipitates, i.e. NbN, AlN, and TiN.
Solution
The equilibrium dissolution temperature of the precipitates can be computed from equation (4-
4) by inserting the correct values for C* and D* from Table 4.1:
It is evident from these calculations that precipitates of the NbN and the AlN type will
dissolve readily at temperatures above 1050 to 11000C, while TiN is thermodynamically sta-
ble up to about 14500C. In practice, however, a certain degree of superheating is always
required to overcome the inherent kinetic barrier against dissolution, particularly if the heating
rate is high. Consequently, in a real welding situation the actual dissolution temperature of the
precipitates may be considerably higher than that inferred from simple thermodynamic calcu-
lations based on the solubility product (to be discussed later).
4.2.3 Stable and metastable solvus boundaries

Due to the lack of adequate phase diagrams for the complex alloy systems involved, thermo-
dynamic calculations based on the solubility product represent in many cases the only practi-
cal means of estimating the solid solubility of alloying elements in commercial low-alloy steels
and aluminium alloys.
4.2.3.1 Equilibrium precipitates

In the case of large, incoherent precipitates (where the Gibbs-Thomson effect can be neglected),
the concentration of element A in equilibrium with pure An Bm at different temperatures can be
inferred directly from equation (4-3). If we replace / ^ by R (i.e. switch from common to
natural logarithms), this equation yields:
(4-5)
Equation (4-5) describes the solvus surface within the solvent-rich corner of the phase dia-
gram. However, when a pure binary compound dissolves the concentration of elements A and
B in solid solution is fixed by the stoichiometry of the reaction. The following relationship
exists between [%B] and [%A]:
(4-6)
or
where MA and MB are the atomic weight of elements A and B, respectively.

Figure 4.2 shows a graphical representation of equations (4-5) and (4-6), and the corre-
sponding change in the matrix concentrations during dissolution of pure AnBm for a given set
of starting conditions.
Alternatively, we can express T as function of the product [%A]n [%B]m by utilising equa-
tions (4-3) and (4-6). The combination of these equations provides a mathematical description
of the 'solvus boundary' of an equilibrium precipitate in a multi-component alloy system. It is
evident from the graphical representation in Fig. 4.3 that the solid solubility will always in-
[%A]0
[%B]
[%B]o
Excess B
[%A]
Fig. 4.2. Concentration displacements during dissolution of binary intermetallics (equilibrium condi-
tions).
I Increased dissolution temperature
Temperature
Increased additions
Nominal alloy composition
_of element B
Reduced solid solubility
of element A
Concentration of element A
Fig. 4.3. Factors affecting the solid solubility of a binary intermetallic compound in a multi-component
alloy system (schematic).
crease with increasing temperature when AH° is positive. This type of behaviour is character-
istic of intermetallics in metals and alloys, since the dissolution process in such systems is
endothermic.7 As a result, increased additions of a second alloying element B will also reduce
the solubility of the first alloying element A by shifting the 'solvus boundary' towards higher
temperatures when an intermetallic compound between A and B is formed.
With the aid of Fig. 4.3 it is easy to verify that the equilibrium volume fraction of the
precipitates/^ at a fixed temperature is given by:
(4-7)
where fmax is the maximum possible volume fraction precipitated at absolute zero.
Equation (4-7) provides a basis for estimating the equilibrium volume fraction of binary
intermetallics in complex alloy systems at different temperatures in cases where the concentra-
tion of element B is sufficiently high to tie-up all A in the form of precipitates. Similarly, if A
is present in an overstoichiometric amount with respect to B, we may write:
(4-8)
Example (4.2)
In Al-Mg-Si alloys the equilibrium Mg2Si phase may form during prolonged high tempera-
ture annealing. Consider a pure ternary alloy which contains 0.75 wt% (0.83 at.%) Mg and 1.0
wt% (0.96 at.%) Si. Estimate on the basis of the solubility product the equilibrium volume
fraction of Mg2Si at 4000C. Make also a sketch of the Mg2Si solvus in a vertical section
through the ternary Al-Mg-Si phase diagram. Relevant physical data for the Al-Mg-Si sys-
tem are given below:
Solution
The maximum possible volume fraction of Mg2Si precipitated at absolute zero (fmax) can be
estimated from a simple mass balance by considering the stoichiometry of the reaction:
Moreover, the solubility product [at.% Mg]2 [at.% Si] at 4000C (673K) can be obtained
from equation (4-3) by utilising data from Table 4.1:
from which
If we also take into account the stoichiometry of the reaction, the solubility product can be
expressed solely in terms of the Mg-concentration. Substituting
into the above equation gives [at.% Mg] ~ 0.20. The equilibrium volume fraction OfMg2Si at
4000C is thus:
Similarly, the equilibrium Mg2Si solvus can be calculated from the solubility product by
substituting
into equation (4-3). By inserting data from Table 4.1 and rearranging this equation, we get:
It is seen from the graphical representation of the above equation in Fig. 4.4 that the Mg2Si
compound is thermodynamically stable up to about 3000C. At higher temperatures the phase
will start to dissolve until the process is completed at 5600C.
It is obvious from these calculations that the microstructure of overaged Al-Mg-Si alloys
should be very persistent to the heat of welding processes. In practice, only a narrow solutionised
zone forms adjacent to the fusion boundary. However, within this zone significant strength
recovery may occur after welding due to reprecipitation of hardening phases from the
supersaturated solid solution. Consequently, in such weldments the ultimate HAZ strength
level is usually higher than that of the base metal, as illustrated in Fig. 4.5.
4.2.3.2 Metastable precipitates

In practice, the solid solubility is also affected by the size of the particles. If, for instance, a
Dissolution temperature: 5600C

Temperature, 0C
Nominal alloy composition
at% Mg
at%Mg Si
Fig. 4.4. Solubility of Mg2Si in aluminium (Example (4.2)).

Strength level
After artificial ageing
After natural ageing
Immediately after welding
HAZ Unaffected base metal
Distance from fusion line

Fig. 4.5. Response of overaged Al-Mg-Si alloys to welding and subsequent heat treatment (schematic).
spherical precipitate is acted on by an external pressure of say 1 atm, the same precipitate is
also subjected to an extra pressure AP due to the curvature of the particle/matrix interface, just
as a soap bubble exerts an extra pressure on its content (see Fig. 4.6(a)). The pressure AP is
given as:8
(4-9)
where 7 is the particle/matrix interfacial energy, and r is the radius of the precipitate.
Because of this extra pressure, the Gibbs energy of a small precipitate will be higher than
that of a large one, which, in turn, increases its solubility (see Fig. 4.6(b)). The important
influence of particle curvature on the solid solubility has been extensively investigated and
reported in the literature.18 Usually, the phenomenon is referred to as the capillary or the
Gibbs-Thompson effect.
In the following we shall assume that the thermodynamic and crystallographic properties
of the metastable precipitates are similar to those of the equilibrium phase and that the re-
duced thermal stability is only associated with capillary effects. For single phase precipitates
in binary alloy systems, it is fairly simple to show that the concentration of solute across
a curved interface, [%A]r, is interrelated to the equilibrium concentration of solute across a
planar interface, [%A], through the following equation:8
(4-10)
where Vn is the molar volume of the precipitate (in m3 mol"1), and Q is the contribution of the
interface curvature to the reaction enthalpy (equal to IyVJr).
(a)
Matrix
Atmospheric
pressure
(b)
Small precipitate
Gibbs energy
Matrix
Large precipitate
[%A] [%A] r
Concentration
Fig. 4.6. Effect of interfacial energy on the solubility of small particles; (a) Schematic representation of
spherical particles embedded in a metal matrix, (b) Integral molar Gibbs energy of matrix and precipi-
tates at a constant temperature.
Assuming that this relationship also holds in the case of binary intermetallics, a combina-
tion of equations (4-5) and (4-10) gives:
(4-11)
where
(4-12)
or
Alternatively, we can express T as a function of the product [%A]rn [%B]rm. This gives the
following expression for the solvus temperature of metastable precipitates T'eq:
(4-13)
It is evident from the graphical representation of equation (4-13) in Fig. 4.7 that the solid
solubility at a given temperature is significantly increased at small particle radii. Taking as an
example 7 = 0.5 J n r 2 , Vm = 10~5 m3 moH, R = 8.314 J Kr1 moH, T = 500 K, we obtain from
equation (4-10):
or
where r is the particle radius in nm.

From this it is seen that quite large solubility differences can arise for particles in the range
from r = 1 - 50nm.
Large (equilibrium) precipitates
Small (metastable) precipitates

Temperature
Concentration
Fig. 4.7. Graphical representation of equation (4-13) (schematic).

Example (4.3)
In Al-Mg-Si alloys metastable (hardening) p"(Mg2Si)-precipitates m a y form during artificial
ageing in the temperature range from 160-2000C. Consider a T6 heat treated ternary alloy
which contains 0.75 wt% (0.83 at.%) Mg and 1.0 wt% (0.96 at.%) Si. Based on equation (4-
13) make a sketch of the metastable P "(Mg2Si) solvus in a vertical section through the ternary
Al-Mg-Si phase diagram. In these calculations we shall assume that the thermodynamic
properties of the metastable (3"(Mg2Si) phase are similar to those of the equilibrium (3 (Mg2Si)
phase, i.e. the reduced thermal stability is only related to the Gibbs-Thomson effect. Relevant
physical data for the Al-Mg-Si system are given below:
Solution
First we estimate the molar volume of the precipitate:
The contribution of the particle curvature to the reaction enthalpy is thus:
The metastable [3"(Mg2Si) solvus can now be calculated from the solubility product by
substituting
into equation (4-13). By inserting data from Table 4.1 and rearranging this equation, we get:
It is evident from the graphical representation of the above equation in Fig. 4.8 that the
particle curvature has a dramatic effect on the solid solubility. A comparison with Fig. 4.4
shows that the dissolution temperature drops from about 5600C in the case of the equilibrium
Mg2Si phase to approximately 225°C for the metastable |3"(Mg2Si)-phase. On this basis it is
not surprising to find that artificially aged (T6 heat treated) Al-Mg-Si alloys suffer from se-
vere softening in the HAZ after welding, as shown schematically in Fig. 4.9. Moreover, it is
Dissolution temperature: 225 0C
Temperature, 0C
composition A
Nominal alloy
Metastable solvus
boundary
at% Mg
at% Mg 2 Si
Fig. 4.8. Solubility of (3"(Mg2Si) in aluminium (Example (4.3)).

Strength level
After artificial ageing
After natural ageing
Immediately after welding
HAZ
Fig. 4.9. Response of artificially aged Al-Mg-Si alloys to welding and subsequent heat treatment (sche-
matic).
evident that the characteristic low dissolution temperature of the precipitates also gives rise to
the formation of a heat affected zone which is significantly wider than that observed during
welding of overaged Al-Mg-Si alloys.9 This shows that the response of age-hardenable alu-
minium alloys to welding and thermal processing depends strongly on the initial base metal
temper condition.
With the aid of equation (4-11) it is also possible to calculate an average (apparent) metastable
solvus boundary enthalpy for hardening |3"(Mg2Si)-precipitates in Al-Mg-Si alloys. A closer
evaluation of the exponent gives:
This value is in close agreement with the reported solvus boundary enthalpy for (3"(Mg2Si)-
precipitates in 6082-T6 aluminium alloys. 910
4.3 Particle Coarsening
When dispersed particles have some solubility in the matrix in which they are contained, there
is a tendency for the smaller particles to dissolve and for the material in them to precipitate on
larger particles. The driving force is provided by the consequent reduction in the total interfa-
cial energy and ultimately, only a single large particle would exist within the system.
4.3.1 Coarsening kinetics

The classical theory for particle coarsening was developed independently by Lifshitz and
Slyovoz11 and by Wagner.12 The kinetics are generally controlled by volume diffusion through
the matrix. At steady state, the time dependence of the mean particle radius r is found to
be:11'12
(4-14)
where ro is the initial particle radius, 7 is the particle-matrix interfacial energy, Dm is the ele-
ment diffusivity, Cm is the concentration of solute in the matrix, Vm is the molar volume of the
precipitate per mole of the diffusate, and t is the retention time.
Although the classic Lifshitz-Wagner theory suffers from a number of simplifying assump-
tions, experimental observations usually reveal a cubic growth law of the form given by equa-
tion (4-14).13
4.3.2 Application to continuous heating and cooling

Ion, Easterling and Ashby14 have shown how equation (4-14) can be applied to continuous
heating and cooling. In their analysis equation (4-14) was used in a more general form:
(4-15)
where c{ is a kinetic constant, and Qs is the activation energy for the coarsening process (for
binary intermetallics Qs may be taken equal to the activation energy for diffusion of the less
mobile constituent atom of the precipitates in the matrix).
4.3.2.1 Kinetic strength of thermal cycle

It follows that the extent of particle coarsening occurring during a weld thermal cycle can be
calculated by integration of equation (4-15) between the limits t = t{ and t = t2:
(4-16)
The integral on the right-hand side of equation (4-16) represents the kinetic strength of the
thermal cycle with respect to particle coarsening, and can be determined by means of numer-
ical methods when the weld thermal (T-t) programme is known. The resulting radius of the
precipitates may then be evaluated from equation (4-16) by inserting representative values
for the constants ro and C1 (e.g. obtained from quantitative particle measurements).
4.3.2.2 Model limitations

A salient assumption in the classic Lifshitz-Wagner theory is that the particles coarsen at
almost constant volume fraction, i.e. no solute is lost to the surrounding matrix during the
coarsening process. Consequently, equation (4-16) should only be applied in cases where the
peak temperature of the thermal cycle is well below the equilibrium solvus of the precipitates.
Example (4.4)
Consider stringer bead deposition (GMAW) on a thick plate of a Ti-microalloyed steel under
Assume that the base metal contains a fine dispersion of TiN precipitates in the as-received
condition. Calculate on the basis of equation (4-16) and the Rosenthal thick plate solution
(equation (1-45)) the extent of particle coarsening occurring within the fully transformed heat
affected zone during welding. Relevant physical data for titanium-microalloyed steels are
given below:
(activation energy for diffusion of Ti in austenite)
Solution
In the present example the problem is to calculate the size of the TiN precipitates in different
positions from the fusion boundary. This requires detailed information about the weld thermal
programme, as shown in Fig. 4.10(a). By substituting the appropriate values for qo, X, a and v
into the Rosenthal thick plate solution, the governing heat flow equation becomes:
where /?* refers to the three-dimensional radius vector in the moving coordinate system (desig-
nated R in equation (1-45)), while x is the welding direction (equal to vt at pseudo-steady
state).
Since titanium nitride is thermodynamically stable up to the melting point of the steel,
equation (4-16) can be used to calculate the extent of particle coarsening occurring within the
transformed parts of the HAZ. In the present example, we may write:
where the times tx and t2 are defined in Fig. 4.10(a).

The kinetic strength of the weld thermal cycle with respect to particle coarsening can now
be evaluated from these two equations by utilising the numerical integration procedure shown
in Fig. 4.10(b). The results from such computations are presented graphically in Fig. 4.11.
It is evident from this figure that significant coarsening of the precipitates occurs within the
HAZ during welding, particularly in regions close to the fusion boundary where the peak
temperature of the thermal cycle is high. A comparison with the experimental data of Ion et
al.l4 (reproduced in Fig. 4.12) shows that the theory gives a fairly good prediction of particle
size as a function of the peak temperature, provided that the kinetic constant C1 in equation (4-
16) can be estimated with a reasonable degree of accuracy. In practice, however, the numerical
value of C1 will vary significantly with the chemical composition and thermal history of the
base metal (see Fig. 4.13). This means that empirical calibration of equation (4-16) to experi-
mental data is always required to avoid systematic deviations between theory and experiments.
4.4 Particle Dissolution
During welding, the thermal pulse experienced by the heat affected zone adjacent to the fusion
boundary can result in complete dissolution of the base metal precipitates. Since this may give
rise to subsequent strength loss and grain growth, it is important to understand how variations
in welding parameters and operational conditions affect the dissolution rate. In the following,
the kinetics of particle dissolution will be discussed from a more fundamental point of view.
4.4.1 Analytical solutions

Over the years, several analytical models have been developed which describe the kinetics of
particle dissolution in metals and alloys at elevated temperatures.16 None of these solutions are
exact, since they represent different approximations to the diffusion field around the dissolv-
(a)
Weld metal
HAZ
Temperature
Y~ regime
Time
(b)
(i/r)exp(-Qs/RT)
Time
Fig. 4.10. Kinetic strength of weld thermal cycle with respect to particle coarsening (Example (4.4)); (a)
HAZ temperature-time programme (schematic), (b) Numerical integration procedure (schematic).
Fully transformed HAZ •
Partly transformed HAZ

Particle radius, nm
Weld metal
Fig. 4.11. Coarsening of TiN during steel welding (Example (4.4)).
Rosenthal thick plate heat cycle:

Frequency, %•
Particle radius, nm
Fig. 4.12. Measured size distribution of TiN before (broken lines) and after (full lines) weld thermal
simulation. Operational conditions as in Example (4.4). Data from Ion et al.14
ing precipitates. Nevertheless, it will be shown below that at least some of them are suffi-
ciently accurate to capture the essential physics of the problem and to give valuable quantita-
tive information on the extent of particle dissolution occurring during the weld thermal cycle.
Annealing temperature: 1350 0C
Particle radius, nm
Steel A
Steel B
Steel C
Annealing time, s
Fig. 4.13. Effects of annealing time and steel chemical composition on the mean particle size of TiN.
Data from Matsuda and Okumura.15
4.4. L1 The invariant size approximation

The model described here is due to Whelan.17 Consider a spherical particle embedded in an
infinite matrix, as shown schematically in Fig. 4.14. The corresponding matrix concentration
profile is plotted in the lower part of the figure. In this case the concentration of the constituent
element A is higher close to the particle/matrix interface than in the bulk. Hence, there is a
tendency for the element to diffuse away from the particle and into the surrounding matrix (i.e.
the particle dissolves). Based on the assumption that the particle/matrix interface is stationary
(i.e. the diffusion field has no memory of the past position of the interface), Whelan17 arrived
at the following expression for the dissolution rate of a spherical precipitate at a constant
temperature:
(4-17)
where r is the particle radius, a is the dimensionless supersaturation (defined in Fig. 4.14), and
Dm is the element bulk diffusivity.
The term Hr on the right-hand side of equation (4-17) stems from the steady-state part of
the diffusion field, while the (1 A/7) term arises from the transient part. Because of the com-
plex form of equation (4-17) it cannot be integrated analytically and hence, numerical
methods must be applied. However, if the transient part of equation (4-17) is neglected (con-
forming to the solution after long times), it is possible to obtain a simple expression for the
particle radius as a function of time:
(4-18)
where ro is the initial particle radius.

Equation (4-18) is identical with the so-called invariant-field solution developed independ-
Concentration
Distance
Fig. 4J4. Schematic representation of the concentration profile around a dissolving spherical particle in
an infinite matrix.
ently by Aaron et al.16 and is valid after a certain period of time, provided that there is no
impingement of diffusion fields from neighbouring precipitates. As shown in Fig. 4.15, this
simplified solution gives a reasonable description of the dissolution kinetics of small spherical
precipitates in steel during reheating above the AC1 -temperature.
Following the treatment of Agren,18 the time required for complete dissolution of a spher-
ical precipitate td can be obtained from equation (4-18) by setting r = 0:
(4-19)
Moreover, the volume fraction of the precipitates/as a function of time is given by:
(4-20)
where fo is the initial particle volume fraction.

The former equation shows that the dissolution time td depends strongly on the initial parti-
cle size rQ.
Example (4.5)
The following example illustrates the direct application of equations (4-18) and (4-19). Con-
sider a niobium-microalloyed steel which contains a fine dispersion of NbC precipitates. Pro-
vided that impingement of diffusion fields from neighbouring particles can be neglected, cal-
culate the total time required for complete dissolution of a 100 nm large NbC precipitate at
Particle radius, jam
Numerical solution (Agren)
Simplified analytical
solution (Whelan)
Time, s
Fig. 4.15. Dissolution kinetics of spherical cementite particles in austenite at 8500C. Data from Agren.18
135O°C. Data for the steel chemical composition and the diffusivity of Nb in austenite at
13500C are given below:
Steel chemical composition:
Diffusivity of Nb in austenite at 13500C:
Atomic weight of Nb:

Atomic weight of C :
Solution
In the present example it is reasonable to assume that the dissolution rate of the precipitate is
controlled by diffusion of Nb in austenite. For a single NbC precipitate embedded in a Nb-
depleted matrix, the dimensionless supersaturation becomes:
The equilibrium concentration of niobium at the particle/matrix interface can be estimated

from the solubility product by utilising data from Table 4.1. If we assume that the carbon
concentration at the interface is constant and equal to the nominal value of 0.12 wt% (i.e. the
stoichiometry of the dissolution reaction is neglected), equation (4-5) reduces to:
Moreover, the concentration of Nb in the precipitate is equal to:
This gives:
The dissolution time td can now be calculated from equation (4-19) by inserting the appro-
priate values for ro, aNt)J and Dm\
A comparison with Fig. 4.16 shows that the predicted value is off by a factor of about 4
compared with that obtained from more sophisticated numerical calculations. This degree of
accuracy is acceptable and justifies the use of equation (4-18) for prediction of the dissolution
rate of spherical precipitates under different thermal conditions provided that the model is
calibrated against experimental data points.
4.4.1.2 Application to continuous heating and cooling
Application of the model to continuous heating and cooling requires numerical integration of
equation (4-18) over the weld thermal cycle:
(4-21)
Under such conditions the volume fraction of the precipitates is given by:
(4-22)
Equations (4-21) and (4-22) provide a basis for predicting the extent of particle dissolution
occurring within the HAZ during welding in the absence of impingement of diffusion fields
from neighbouring precipitates.
Dissolution time, s
Particle diameter, nm
Fig. 4.16. The dissolution time of NbC in austenite at 13500C as function of initial particle diameter lro
for different Nb and C levels (numerical solution). Data from Suzuki et al.6
Example (4.6)
Consider stringer bead deposition (SAW) on a thick plate of a Nb-microalloyed steel (0.10
wt% C - 0.03 wt% Nb) under the following conditions:
Assume that the base metal contains a fine dispersion of NbC precipitates in the as-received
condition. Calculate on the basis of equation (4-22) and the Rosenthal thick plate solution
(equation (1-45)) the extent of particle dissolution occurring within the fully transformed HAZ
during welding. Relevant physical data for Nb-microalloyed steels are given below:
Solution
In the present example the problem is to calculate the variation in the//fo ratio across the fully
transformed HAZ. By substituting the appropriate values for qo, X, a, and v into the Rosenthal
thick plate solution, the governing heat flow equation becomes:
Since it is reasonable to assume that the dissolution rate of the precipitates is controlled by
diffusion of Nb in austenite, the dimensionless supersaturation reduces to:
As shown in example (4.5), the equilibrium concentration of niobium at the particle/matrix

interface (in wt%) can be estimated from the solubility product by utilising data from Table
4.1. If we assume that the carbon concentration at the interface is constant and equal to the
nominal value of 0.10 wt%, equation (4-5) becomes:
Moreover, the concentration of Nb in the precipitate is equal to:
This gives:
By substituting the appropriate expressions for aNb and DNb into equation (4-22), we ob-
tain:
Here the lower and upper integration limits refer to the total time spent in the thermal cycle
from Ac3 to T and down again to Ac3.
The extent of particle dissolution occurring within the HAZ during welding can now be
calculated in an iterative manner by numerical integration of the above equation over the weld
thermal cycle. The results from such computations are presented graphically in Fig. 4.17.
It is evident from these data that NbC starts to dissolve when the peak temperature of the
thermal cycle T exceeds the equilibrium dissolution temperature Td of the precipitate. The
process is completed when T approaches 13300C, conforming to a temperature interval of
1900C. This shows that considerable superheating is required in order to overcome the inher-
ent kinetic barrier against particle dissolution under the prevailing circumstances.
4.4.2 Numerical solution

In the previous treatment, no consideration is given to impingement of diffusion fields from
neighbouring precipitates or the position of the particle/matrix interface during the dissolution
process. In certain cases, however, such phenomena will have a marked effect on the dissolu-
tion kinetics.18"22 A good example is Al-Mg-Si alloys where the hardening P"(Mg2Si)-phase
forms a very fine distribution of needle-shaped precipitates along <100> directions in the
aluminium matrix. These precipitates are closely spaced and will therefore interact strongly
with each other during dissolution (coupled reversion).
No dissolution-
f/fo
Complete dissolution
Fig. 4.17. Dissolution of NbC during steel welding (Example (4.6)).

4.4.2.1 Two-dimensional diffusion model
For rod or needle-shaped precipitates in a finite, depleted matrix, the rate of dissolution can be
calculated by numerical methods from a simplified two-dimensional diffusion model. Assum-
ing that the precipitates are mainly aligned in one crystallographic direction, it is reasonable to
approximate their distribution by that of a face-centered cubic (close-packed) space lattice, as
shown in Fig. 4.18(a). If planes are placed midway between the nearest-neighbour particles,
they enclose each particle in a separate cell. Since symmetry demands that the net flux of
solute through the cell boundaries is zero, the dissolution zone is approximately defined by an
inscribed cylinder whose volume is equivalent to that of the hexagonal cell. The modelling
principles outlined in Fig. 4.18(a) and (b) have previously been used by a number of other
investigators to describe particle dissolution during isothermal heat treatment.18"22 Conse-
quently, readers who are unfamiliar with the concept should consult the original papers for
further details.
It follows from Fig. 4.18(b) that the rate of reversion can be reported as:
(4-23)
where ro is the initial cylinder (particle) radius.
(a)
(b)
Concentration
Distance
Fig. 4.18. Numerical model for dissolution of rod-shaped particles in a finite, depleted matrix; (a) Disso-
lution cell geometry, (b) Particle/matrix concentration profile (moving boundary).
For a specific alloy, the ratio between ro and L (the mean interparticle spacing) can be
calculated from a simple mass balance, assuming that all solute is tied-up in precipitates. Tak-
ing this ratio equal to 0.06 for rod-shaped precipitates in diluted alloys,9 the kinetics of particle
dissolution during isothermal heat treatment have been examined for a wide range of opera-
tional conditions. These results are presented in a general form in Fig. 4.19 by the use of the
following groups of dimensionless parameters:
Dimensionless time (4-24)
Dimensionless supersaturation (4-25)

(a is defined previously in Fig. 4.14).
The data in Fig. 4.19 suggest that the reaction kinetics during the initial stage of the process
are approximately described by the relation:
(4-26)
where c2 is a kinetic constant, and n{ is a time exponent (assumed constant and equal to 0.5
under the prevailing circumstances).
The rate of particle dissolution will gradually decline with increasing values of T as a result
of impingement of diffusion fields from neighbouring precipitates which reduces a. In prac-
tice, this is seen as a continuous decrease in the slope of the flfo-x curves in Fig. 4.19 (nx < 0.5).
In such cases equation (4-26) will only be valid within small increments of X.
log <1 - f/fo)
log T
Fig. 4.19. Dissolution kinetics of rod-shaped particles in a finite, depleted matrix. Data from Myhr and
Grong.9
4.4.2.2 Generic model
Myhr and Grong9 have shown how this model can be applied to specific alloy systems.
From equation (4-26) we have:
(4-27)
where n{ < 0.5.

This equation can further be simplified if we assume that and
(4-28)
For isothermal heat treatment at a chosen reference temperature (Tn), the rate of particle
dissolution is determined by the retention time tr{. Let f* denote the maximum hold time
required for complete dissolution of the precipitates. It follows that equation (4-28) can be
written in a general form by normalising tn with respect to t*x. The parameter t*n is obtained
by setting flfo = 0:
(4-29)
where C3 is a new kinetic constant.

If heat treatment is carried out at a different temperature (T * Tn) , the maximum hold time
t* is simply given by:
(4-30)
By inserting the approximate expressions for C1 and Dm into equation (4-30) (see previous
examples), and rearranging equation (4-28), we obtain:
(4-31)
and
(4-32)
where Q'apP. is the apparent (metastable) solvus boundary enthalpy (defined in Section 4.2.3.2),
and Qs is the activation energy for diffusion of the less mobile constitutive atom of the precipi-
tates.
Equations (4-31) and (4-32) exploit some good modelling techniques. For example, the use
of a dimensionless time eliminates an unknown kinetic constant which premultiplies t and ^1*
in the derivation of equation (4-32). Moreover, by raising the dimensionless time to a power
H1 means that the premultiplying constant, here unity, is independent of the value of nv and is
itself also dimensionless. Finally, the form of equation (4-31) eliminates further unknown
kinetic constants, and may readily be calibrated using an experimental time r* at a reference
temperature.
Figure 4.20 shows the variation inflfo with time (on log axes), from a range of isothermal
experiments carried out on 6082-T6 aluminium alloys, using hardness (or electrical conduc-
tivity) measurements to evaluate/7/\ The curve (equation (4-32)) extrapolates back to a slope
of 0.5 (the exponent n{) for the case of the early stages of dissolution before impingement of
the diffusion fields. The exponent nx is seen to fall to lower values when the proportion dis-
solved is higher, in agreement with the theoretical curves in Fig. 4.19.
4.4.2.3 Application to continuous heating and cooling

Myhr and Grong9 have also shown how this model can be applied to situations where the
temperature varies with time (as in welding).
In order to obtain a general description of particle dissolution under non-isothermal condi-
tions, it is convenient to introduce the related concepts of an isokinetic reaction and the kinetic
strength of a thermal cycle.23 A reaction is said to be isokinetic if the increments of transfor-
mation in infinitesimal isothermal time steps are additive. Christian24 defines this mathemati-
cally by stating that a reaction is isokinetic if the evolution equation for some state variable X
may be written in the form:
log(i-f/f 0 )
log (t/t*)
Fig. 4.20. 'Master-curve' for dissolution of hardening p"(Mg2Si)-precipitates in 6082-T6 aluminium

alloys. Data from Myhr and Grong.9
(4-33)
where G(X) and H{T) are arbitrary functions of X and T, respectively.

For a given thermal history T(t), this essentially means that the differential equation con-
tains separable variables of X and T. The same criterion may also be applied to the models
derived above. In the case of coupled reversion, we may write:
(4-34)
and
(4-35)
Since dfldlx and t\ are unique functions of/and T, respectively, the additivity condition is
satisfied. Consequently, when the temperature varies with time, we replace the term 11 t*x in
equation (4-34) by dt I t*x and integrate over the thermal cycle, giving:
(4-36)
This integral is called the kinetic strength of the thermal cycle with respect to reversion.
The resulting volume fraction of the precipitates following a heating cycle is then found by
evaluating the integral Z1 numerically (e.g. by utilising input data from Table 4.2) and replac-
ing 111* with Z1 in equation (4-32), yielding a value for flfQ from the master curve of Fig. 4.20.
Case Study (4.1)
By utilising equation (4-36) and the general heat flow model for welding on medium thick
plates (i.e. equation (1-104)), it is possible to calculate the variation in the f/fo ratio (i.e. the
solute distribution) across the HAZ of single pass 6082-T6 aluminium weldments for a wide
range of operational conditions (see Table 4.3). The results from such computations are pre-
sented graphically in Fig. 4.21 -4.23.
When stringer bead deposition is carried out on a plate of medium thickness, the solute
distribution in the transverse y direction is expected to vary with distance from the plate sur-
face due to a continuous change in the heat flow conditions. A closer inspection of Figs. 4.21
and 4.22 shows that this is correctly accounted for in the present model. In contrast, a full
penetration butt weld will always reveal a similar solute distribution in the transverse direction
of the weld, as shown in Fig. 4.23. This situation arises from the lack of a temperature gradient
in the through-thickness z direction of the plate.
Table 4.2 Basic input data in dissolution model for hardening p" (Mg2Si)- precipitates in 6082-T6
aluminium alloys. Data from Myhr and Grong.9
Parameter Q'app Qs nx t\
(starting value) (375°C)
Value 3OkJm0I-1 13OkJm 0 I- 1 0.5 600 s
(a)
Wo
Comolete dissolution
Aym ,mm
(b)
f/fo
ym,mm
Fig. 4.21. Dissolution of p"(Mg2Si)-precipitates during aluminium welding (Weld 1); (a) Upper plate
surface, (b) Lower plate surface. Operational conditions as in Table 4.3.
Table 4.3 Operational conditions used in aluminium welding experiments (Case study 4.1).
Weld Material Plate thickness Net arc power Welding speed

No.* (mm) (kW) (mms"1)
1 AA6082 15 9.1 4.2
2 AA6082 15 5.7 9.1
3 AA6082 13 14.0 5.8
* Calibration of heat flow model is done by including an empirical correction for heat consumed in melting of the
parent material (thermal data for the AA6082 alloy are given in Table 1.1, Chapter 1).
(a)
Mo
Aym,mm
(b)
f/fo
ym,mm
Fig. 4.22. Dissolution of (3"(Mg2Si)-PrCCiPiIaIeS during aluminium welding (Weld 2); (a) Upper plate
surface, (b) Centre of plate. Operational conditions as in Table 4.3.
4.4.2.4 Process diagrams for single pass 6082-T6 butt welds

For single pass butt welding of plates, the heat flow model (equation 1-104)) can largely be
simplified if the net arc power is kept sufficiently high compared with the plate thickness (i.e.
q /d>0.5 kW mm"1). Under such conditions the mode of heat flow becomes essentially one-
dimensional, and the temperature distribution is determined by the ratio qo/vd, kJ ram"2 (see
Sections 1.10.3.3 and 1.10.4.1 in Chapter 1).
Wo
Ay ,mm
m
Fig. 4.23. Dissolution of p"(Mg2Si)-precipitates during aluminium welding (Weld 3). Operational con-
ditions as in Table 4.3.
Figure 4.24 shows plots of the variation in the flfo ratio across the HAZ of 6082-T6 alu-
minium weldments for different values of qo Ivd. It follows that a narrow width of the dissolu-
tion zone requires the use of a low energy per mm2 of the weld. In practice, this can be
achieved by the use of an efficient welding process (e.g. electron beam or laser welding) which
facilitates deposition off a full penetration butt weld without employing a groove preparation
(i.e. eliminates the need for filler metals).
f/fo
Scale:
Fig. 4.24. Process diagram showing the solute distribution within the HAZ of single-pass 6082-T6 alu-
minium butt welds for different values of qo Ivd.
References
1. R.D. Doherty: Physical Metallurgy, 3rd Edn (Eds R.W. Chan and R Haasen), 1983, Amster-
dam, North-Holland Physics Publ., 934-1030.
2. K.E. Easterling: Introduction to the Physical Metallurgy of Welding, 1983, London, Butterworths
& Co., Ltd.
3. H. Adrian and RB. Pickering: Mater. ScL TechnoL, 1991, 7, 176-182.
4. B. Loberg, A. Nordgren, J. Strid and K.E. Easterling: MetalL Trans., 1984,15A, 33-41.
5. J. Strid and K.E. Easterling: Acta MetalL, 1985, 33, 2057-2074.
6. S. Suzuki, G.C. Weatherly and D.C. Houghton: Acta. MetalL, 1987, 35, 341-352.
7. J.L. Petty-Galis and R.D. Goolsby: J. Mater. ScL, 1989, 24, 1439-1446.
(England), Van Nostrand Reinhold Co. Ltd.
9. O.R. Myhr and 0. Grong: Acta MetalL Mater., 1991, 39, 2693-2702; ibid., 2703-2708.
10. H.R. Shercliff and M.F. Ashby: Acta MetalL Mater., 1990, 38, 1789-1802; ibid., 1803-1812.
11. J.M. Lifshitz and V.V. Slyozov: /. Phys. Chem. Solids, 1961,19, 35-50.
12. C. Wagner: Z. Electrochem., 1961, 65, 581-591.
13. L.C. Brown: Acta MetalL Mater., 1992, 40, 1293-1303.
14. J.C. Ion, K.E. Easterling and M.F. Ashby: Acta MetalL, 1984, 32, 1949-1962.
15. S. Matsuda and N. Okumura: Trans. ISIJ, 1978,18, 198-205.
16. H.B. Aaron, D. Fainstein and G.R. Kotler: J. Appl. Phys., 1970, 41, 4404-4410.
17. MJ. Whelan: Metal ScL J., 1969, 3, 95-97.
18. J. Agren: Scand. J. MetalL, 1990,19, 2-8.
19. R.A. Tanzilli and R.W. Heckel: Trans. Met. Soc. AIME, 1968, 242, 2313-2321.
20. H.B. Aaron and R. Kotler: MetalL Trans., 1971, 2, 393-408.
21. R. Asthana and S.K. Pabi: Mat. ScL Eng., 1990, A128, 253-258.
22. U.H. Tundal and N. Ryum: MetalL Trans., 1992, 23A, 433-444; ibid., 445-449.
23. H.R. Shercliff, 0. Grong, O.R. Myhr and M.F. Ashby: Proc. 3rd Int. Conf. on Aluminium
Alloys — Their Physical and Mechanical Properties, Trondheim, Norway, June 1992, Vol. Ill,
pp. 357-369, The University of Trondheim, The Norwegian Inst. of Technol.
24. J.W. Christian: Phase Transformations in Metals and Alloys, 1975, Oxford, Pergamon Press.
Appendix 4.1
Nomenclature
thermal diffusivity (mm2 s l) end temperature of ferrite to

austenite transformation (0C)
activity of element A in alloy
equilibrium concentration of
activity of precipitate An Bm element A in matrix (wt% or
at.%)
in alloy
analytical content of element
activity of element B in alloy A in alloy (wt% or at.%)
arbitrary alloying element equilibrium concentration of
element A across a curved
start temperature of ferrite to particle/matrix interface
austenite transformation (0C) (wt% or at.%)
arbitrary alloying element standard free energy of reac-
tion (J mol"1 or kJ mol"1)
element B in matrix (wt% or GMAW gas metal arc welding
at.%)
H(T) arbitrary function of T
analytical content of element
B in alloy (wt% or at.%) standard enthalpy of reac-
tion (J mol"1 or kJ mol"1)
element B across a curved amperage (A)
particle/matrix interface
(wt% or at.%) kinetic strength of thermal
cycle with respect to rever-
various kinetic constants and sion
temperature-dependent pa-
rameters equilibrium constant
normalised (dimensionless) mean interparticle spacing in

entropy of reaction dissolution model (m, Jim or
nm)
concentration of solute at
particle/matrix interface integer
(wt% or at.%)
atomic weight of element A
nominal alloy composition (g mol"1)
(wt% or at.%)
concentration of solute in atomic weight of binary pre-

matrix (mol irr 3 , wt% or cipitate (g mol"1)
at.%)
atomic weight of element B
concentration of solute inside (g mol-1)
the precipitates (wt% or
at.%) integer
plate thickness (mm) time exponent in dissolution

model
normalised enthalpy of reac-
tion (K) pressure caused by curvature
effects (J nr 3 )
element diffusivity (mm2 s"1
or m2 s"1) net arc power (W)
particle volume fraction apparent (metastable) solvus

boundary enthalpy (kJ mor 1 )
initial particle volume frac-
tion activation energy for diffu-
sion (kJ moH)
equilibrium particle volume
fraction particle radius (m, Jim or nm)
maximum particle volume initial particle radius (m, (im

fraction at absolute zero or nm)
arbitrary function of X
universal gas constant artificially aged condition
(8.314JK-1InOl-1)
voltage (V)
universal gas constant multi-
plied by InIO (19.14J Kr1 welding speed (mm s"1)
mol"1)
welding direction (mm)
three-dimensional radius
vector in Rosenthal equation state variable
(mm)
transverse direction (mm)
standard entropy of reaction width of HAZ referred to fu-
(J K mol"1) sion boundary (mm)
SAW submerged arc welding through-thickness direction

(mm)
time (s)
molar volume of precipitate
time necessary for complete (m3 mol 1 )
dissolution of precipitate(s)
molar volume of precipitate
integration limits (s) per mole of the diffusate (m3
mol-1)
time necessary for complete
reversion at T(s)
density of precipitate An Bm
retention time at T1. (s) (g cm"3 or kg rrr3)
time necessary for complete thermal conductivity (W

reversion at Trl(s) mm-1 0C-1)
temperature (0C or K) dimensionless supersaturat-

ion in dissolution model
ambient temperature (0C or
K) density (g cm"3 or kg nr 3 )
equilibrium dissolution tem- arc efficiency factor

perature of precipitate (0C or
K) particle/matrix interfacial
energy (J m~2)
melting point (0C or K)
dimensionless time in disso-
peak temperature of thermal lution model
cycle (0C or K)
contribution of particle cur-
chosen reference temperature vature to reaction enthalpy (J
in dissolution model (0C or mol"1 or kJ mol"1)
K)
equilibrium phase in Al-Mg-
equilibrium solvus tempera- Si alloys
ture (0C or K)
hardening precipitate in Al-
solvus temperature of Mg-Si alloys
metastable precipitates (0C or
K)
5
Grain Growth in Welds
5.1 Introduction
Grain growth is an important aspect of welding metallurgy. Normal grain growth in metals
and alloys is a thermally activated process driven by the reduction in the grain boundary en-
ergy. Physically, it occurs by growth of the larger grains at the expense of the smaller ones
which tend to shrink.
Under isothermal heat treatment conditions, normal grain growth is well described by the
following empirical equation:1
(5-1)
where D is the mean grain size (diameter), D0 is the initial grain size, n is the time exponent,
/ is the isothermal annealing time, Qapp. is the apparent activation energy for grain growth, and
C1 is a kinetic constant. The other symbols have their usual meaning.
For most metals and alloys the time exponent n in equation (5-1) varies typically in the
range from 0.1 to 0.4, as shown in Fig. 5.1. Only in the case of ultrapure metals annealed at
very high temperatures the time exponent may approach a constant value of 0.5. This corre-
sponds to the limiting case where the grain boundary migration rate is directly proportional to
the driving pressure 7 /D (7 denotes the grain boundary interfacial energy per unit area, while
D is the grain size).
It is well recognised that alloying and impurity elements both in the dissolved state and in
the form of inclusions or second phase particles will retard grain growth.2"* Consequently, a
comprehensive theoretical treatment of grain growth in welds must include a consideration of
such effects. The present analysis will therefore start with a closer examination of factors
affecting the grain boundary mobility in metals and alloys under conditions applicable to weld-
ing.
5.2 Factors Affecting the Grain Boundary Mobility
5.2.7 Characterisation of grain structures

A critical aspect of modelling grain growth is the quantitative description of grain structures,
which is essential in making a comparison between theoretical predictions and experimental
(U) JU8U0CJX9 91U|1
•Brass
Z.R.AI
Homologous temperature (T/Tm)

Fig. 5.1. Temperature-dependence of the time exponent in isothermal grain growth (Z.R.: zone refined,
H.R: high purity). Data compiled by Hu and Rath.1
measurements. Different parameters are used to describe the size and the shape of individual
grains. In three dimensions, individual grain volumes cannot be determined directly from
measurements made in single cross sections through the structure. Therefore, certain geomet-
ric assumptions must be employed to obtain these quantities.
Since most grain size measurements seek to correlate the interaction of grain boundaries
with specific properties (e.g. the transformation behaviour), an estimate of the grain boundary
area per unit volume Sv is often required. This parameter can be calculated without assump-
tions concerning grain shape and size distribution from measurements of the mean linear grain
intercept D*.5
(5-2)
If the mean grain diameter D is required from 5V, this may be obtained by assuming a
spherical grain shape. Noting that each boundary is shared by two adjacent grains, we obtain:
(5-3)
from which
(5-4)
It follows that the mean linear grain intercept D* is always smaller than the actual grain size
D.
A common observation in metals and alloys is that the size distributions of grain aggregates
at different annealing times become equivalent when the measured grain size parameter, D, is
normalised (scaled) by the time-dependent average of this metric, D (see Fig. 5.2). This
means that grain structures can be completely characterised, in a statistical sense, by simple
probability functions of the standard deviation of the distribution together with the time de-
pendence of the average size scale D.
5.2.2 Driving pressure for grain growth

The thermodynamic driving pressure for motion of a spherically curved element of grain bound-
ary is given by:7
(5-5)
where y is the grain boundary interfacial energy, and p* is the radius of curvature.
It is conventional practice to replace p* in equation (5-5) with some measure of the average
grain size, such as the mean linear grain intercept D*, or with the diameter of the equivalent
spherical volume of some geometrically modelled average grain size. Experimental measure-
ments performed on high purity aluminium indicate that p* ~ 3.23 D*.7 This observation is
consistent with the model of Hellman and Hillert,8 which predicts that the curvature of the
most critical element of grain boundary that must be stabilised is p* = 3 D . Under such condi-
tions, the driving pressure becomes:
(5-6)
Zone refined iron

F (arbitrary units)
Rayleigh distribution
Log-normal
distribution
D/D
Fig. 5.2. Comparison of measured grain size data in iron with the Rayleigh and log-normal distribution
functions F. The similarity of the size distributions at different annealing times illustrates the self-similar
scaling behaviour of normal grain growth. After Pande.6
In practice, the numerical constant in equation (5-6) can vary by, at least, a factor of three,
depending on the assumptions of the models. Consequently, in the general case the average
driving pressure is given by:
(5-7)
where c2 is a constant which is characteristic of the system under consideration.
5.2.3 Drag from impurity elements in solid solution

In the dissolved state impurity and alloying elements will retard grain growth through elastic
attraction of the atoms towards the open structure of the grain boundary. The boundary must
then either drag the impurity atoms along (so that its speed is limited by the diffusion rate of
these atoms) or break away if the impurity concentration is sufficiently small or the driving
pressure or temperature is high enough.
An analysis of such effects can be done on the basis of the classic impurity drag theories of
Chan2 and Lucke and Stiiwe,3 which deal with the following two extreme cases:
(i) A low velocity limit, where the rate of grain boundary migration is controlled by diffu-
sion of impurity atoms perpendicular to the boundary.
(ii) A high velocity limit, where the grain boundary migration process is mainly governed
by the diffusion of solvent atoms across the boundary (i.e. controlled by the rate of boundary
self diffusion).
The low velocity limit is associated with either a low driving pressure PG or a high impurity
level C0, and is characterised by a linear type of relationship between the grain boundary
migration rate Va and PG:2
(5-8)
Here e denotes the intrinsic drag coefficient, while 1F is a parameter depending on the
diffusivity and the interaction energy between the grain boundary and the impurity atoms.
For the other extreme (i.e. the high velocity limit), the grain boundary migration rate Vb is
described by the relationship:2
(5-9)
where 4Vp2 denotes another complex function of the impurity diffusivity and the interaction
energy between the grain boundary and the impurity atoms.
In the case of spherical grains, the classic grain growth equation predicts that the grain
boundary migration rate V is a power function of the driving pressure (y /D). This is readily
seen by differentiating equation (5-1) and inserting 7 (the grain boundary interfacial energy)
into the resulting expression:
(5-10)
where c3 is a new kinetic constant.
Note that at very high driving pressures or low impurity levels, equation (5-9) approaches
the limiting case where the migration rate Vb becomes directly proportional to PG, correspond-
ing to a time exponent n = 0.5 in the grain growth equation. At this point the grain boundary
will break away from its surrounding impurity atmosphere and migrate at a rate close to the
rate of boundary self diffusion. In most cases, however, the relationship between V and PG
derived from equation (5-10) will be different from the theoretical one due to the empirical
nature of the grain growth equation.
For high driving pressures and intermediate impurity concentrations, the classic impurity
drag theories predict a discontinuous transition from the high to the low velocity limit. It has,
however, been argued by Vandermeer9 that the observed transition may be considered as con-
tinuous. Thus, in a log V vs log PG plot it would appear as a steep curve connecting the lines
for the high and low velocity extremes together, as illustrated schematically in Fig. 5.3.
5.2.4 Drag from a random particle distribution

The retardation of grain growth by second phase particles was first theoretically investigated
by Zener.10 There seems to be general agreement that the maximum pinnig force Fp exerted by
a single particle of radius r on a grain boundary is given by: 410
(5-11)
High velocity limit

log V
Transition region
Low velocity limit
log PG
Fig. 5.3. Schematic variation of grain boundary migration rate V with driving pressure PQ according to
the classic impurity drag theories of Cahn2 and Liicke and Stiiwe.3 The diagram is based on the ideas of
Vandermeer.9
If the number of interacting particles per unit area of the grain boundary is taken equal to na,
the resulting retardation pressure becomes:
(5-12)
Assuming that only one half of the particles which touches the grain boundary will interact
with a maximum force, na is related to Nv (the number of particles per unit volume) through the
following equation:410
(5-13)
Given that the particles are spherical and of uniform size, Nv can be expressed as:
(5-14)
where/is the particle volume fraction.
A combination of equations (5-12), (5-13), and (5-14) leads to the well-known expression
for the so-called Zener drag (or Zener retardation pressure):410
(5-15)
In practice, the numerical constant in equation (5-15) can vary by, at least, a factor of five,
depending on the assumptions of the models.11 Consequently, in the general case the Zener
retardation pressure is given by:
(5-16)
where C4 is a constant which is characteristic of the system under consideration.
5.2.5 Combined effect of impurities and particles

As already stated in the introduction of this chapter, the time exponent n is a measure of the
resistance to grain boundary motion in the presence of impurity and alloying elements in solid
solution. Based on equation (5-1) Hu and Rath 112 have shown that the grain boundary migra-
tion rate V is related to the effective driving pressure APG and the time exponent n through the
following equation:
(5-17)
where M is the grain boundary mobility.

In alloys containing grain boundary pinning precipitates, the effective driving pressure APG
is defined as the numerical difference between PG and Pz. By inserting the correct expressions
for PG and Pz into equation (5-17), we obtain:
(5-18)
It follows from equation (5-18) that the grain boundary migration rate V becomes inversely
proportional to the average grain size D when n = 0.5 a n d / = 0. This corresponds to the
limiting case where the grain boundary will break away from its surrounding impurity atmos-
phere and migrate at a rate which is controlled by the diffusion of solvent atoms across the
boundary. In most cases, however, the observed relationship between V and APG will be
different from the theoretical one due to drag from second phase particles (f> 0) or impurity
elements in solid solution (n < 0.5).
5.3 Analytical Modelling of Normal Grain Growth
By substituting V = V2 (dD/dt) and M = M0 exp (- QappIRT) into equation (5-18), it is

possible to obtain a simple differential equation which describes the variation in the average
grain size D with time t and temperature T in the presence of impurities and grain boundary
pinning precipitates:
(5-19)
Equation (5-19) can be written in a more general form by setting and
(5-20)
From this it is seen that the parameters M0 and k are true physical constants which are
related to the grain boundary mobility and the pinning efficiency of the precipitates, respec-
tively.
5.3.1 Limiting grain size

Equation (5-20) shows that the grain structure is stabilised when (d D ldi) - 0. The stable
(limiting) grain size is given by:
(5-21)
The parameter k (which in the following is referred to as the Zener coefficient) is defined as
the ratio between the numerical constants in equations (5-7) and (5-16), respectively. In the
original Zener's model k = 4/3, while other investigators have arrived at different results.811"14
As shown in Fig. 5.4, the limiting grain size may vary by over one order of magnitude, depend-
ing upon the assumptions of the models. This makes it difficult to apply equations (5-20) and
(5-21) for quantitative grain size analyses without further background information on the Zener
coefficient.
Diim.nm
Gladman
r/f, um
Fig. 5.4. Relation between limiting grain size Dum., particle radius r, and volume fraction/predicted by
different models.
Example (5. J)
Consider multipass GMA welding on a thick steel plate under the following conditions:
Based on the models of Zener,10 Hellman and Hillert,8 and Gladman13 estimate the limiting
austenite grain size Dlim in the transformed parts of the weld HAZ when the oxygen and
sulphur contents of the as-deposited weld metal are 0.04 and 0.01 wt%, respectively.
Solution
As shown in Chapter 2 of this textbook the volume fraction of oxide and sulphide inclusions
can be calculated from equation (2-75):
Similarly, the average radius of the grain boundary pinning inclusions can be obtained from
equation (2-79):
This gives the following values for the limiting austenite grain size:
Zener:
Hellman and Hillert:
Gladman:
and
As expected, the limiting austenite grain size is seen to vary by more than one order of
magnitude, depending on the assumptions of the models. In practice, the Zener coefficient in
low-alloy steel weld metals falls within the range from 0.32 to 0.93, as shown in Fig. 5.5. The
average value of A: is close to 0.52, which is the same as that inferred from the Gladman model
(upper limit). When it comes to intermetallic compounds such as titanium nitride, the Zener
coefficient varies typically between 0.75 and 0.25 during grain growth in the austenite re-
gime. 1617 This suggests that k ~ 0.50 is a reasonable estimate of the grain boundary pinning
efficiency of oxides and nitrides in steel.
5.3.2 Grain boundary mobility

Direct application of equation (5-20) requires also reliable information on the time exponent n
and the grain boundary mobility M. When n = 0.5 and/= 0, the classic impurity drag theories
predict that the activation energy Qapp, should be close to the value for boundary self diffusion
in the matrix material.2'3 This borderline case is approximately attained in steel welding, as
shown in Fig. 5.6(a) and (b), since the driving pressure for austenite grain growth immediately
following the dissolution of the pinning precipitates is usually so large that the grain boundary
migration rate approaches the higher velocity limit defined in equation (5-9).18 On this basis it
is not surprising to find that Qapp falls within the range reported for lattice self diffusion (284
kJ mol"1) and boundary self diffusion (170 kJ mol~l) in pure 7-iron19 during welding.18 In
most cases, however, the activation energy will be different from the theoretical one due to
complex interactions between impurity atoms and grain boundaries (characterised by a time
exponent n < 0.5). Under such conditions, the value of Qapp has no physical meaning.1
5.3.3 Grain growth mechanisms

Equation (5-20) provides a basis for evaluating the grain growth inhibiting effect of impurity
elements and second phase particles under different thermal conditions. This also includes
situations where the grain boundary pinning precipitates either coarsen or dissolve during the
heat treatment process.
5.3.3.1 Generic grain growth model

Equation (5-20) can readily be integrated to give the average grain size D as a function of
time. In the general case we may write:
(5-22)
(a)
SA steel weld metal
Austenite grain size, Jim
Annealing temperature, 0 C
(b)
GMA and SA steel
weld metals
D||m. Hm
r/f, urn
Fig. 5.5. Evaluation of the Zener coefficient in steel weld metals containing stable oxide and sulphide
inclusions; (a) Determination of Dum. from isothermal grain growth data (holding time: 30 min),
(b) Variation in Dum. with the inclusion rlf ratio. Data from Skaland and Grong.15
where D{im is the limiting grain size (defined in equation (5-21)).

The integral I1 on the right-hand side of equation (5-22) represents the kinetic strength of the
thermal cycle with respect to grain growth and can be determined by numerical methods when
the temperature-time programme is known. In practice, however, it is not necessary to solve this
integral to evaluate the grain growth mechanisms. Consequently, the left-hand side of equation
(a)
Steel A
Log (DxZD1) Slope: n = 0.4
Log [number of cycles]
(b)
Steel A
Steel B
LogD7
Fig. 5.6. Evaluation of the time exponent n and the activation energy Q for austenite grain growth in
steel under thermal conditions applicable to welding; (a) Time exponent n, (b) Activation energy Qapp.
Data from Akselsen et ah18
(5-22) can be solved explicitly for different values of DUm, n, and Z1. The results may then be
presented in the form of novel diagrams which show the competition between the various
processes that lead to grain growth during heat treatment of metals and alloys.
A more thorough documentation of the predictive power of the model and its applicability
to welding is given in Section 5.4.
5.3.3.2 Grain growth in the absence ofpinning precipitates

In the absence of grain boundary pinning precipitates, we have:/= 0, Dlim —> ~ , and (1/ DlitrL) = 0.
Under such conditions, equation (5-22) reduces to:
(5-23)
After integration this equation yields:
(5-24)
Referring to Fig. 5.7, the average grain size D becomes a simple cube root function of Z1
when n = 0.5 and D 0 = 0. In other situations (n < 0.5), the grains will coarsen at a slower rate
due to drag from alloying and impurity elements in solid solution. This is seen as a general
reduction in the slope of the D-Ix curves in Fig. 5.7.
The important austenite grain growth inhibiting effect of phosphorus and free nitrogen in
steel following particle dissolution is shown in Fig. 5.8.
5.3.3.3 Grain growth in the presence of stable precipitates

If grain growth occurs in the presence of stable precipitates, the limiting grain size {Dlim) in
equation (5-22) becomes constant and independent of the thermal cycle. In the specific case
when n = 0.5 the integral on the left-hand side of equation (5-22) has the following analytical
solution:
(5-25)
D, ^m
I1W"
Fig. 5.7. Predicted variation in average grain size D with /, and n f o r / = 0 and D0 = 0 ('free' grain
growth).
(a)
Steel A
D Y ,fim
Number of cycles
(b)
Steel B
D y ,um
Number of cycles -
Fig. 5.8. Illustration of the austenite grain growth inhibiting effect of phosphorus and free nitrogen in
low-alloy steel during reheating above the Ac^ temperature (multi-cycle weld thermal simulation);
(a) Steel A (50ppm P, 20ppm N), (b) Steel B (180ppm P, 80ppm N). Data from Akselsen et a/.18
from which the average grain size D is readily obtained. In other cases, numerical methods
must be employed to evaluate D.
It is evident from the graphical representation of equation (5-25) in Fig. 5.9 that the grain
growth inhibiting effect of the precipitates is very small during the initial stage of the process
when D « D lim. Under such conditions the grains will coarsen at a rate which is comparable
with that observed for free grain growth (n = 0.5,/= 0). The grain coarsening process becomes
gradually retarded as the average grain size increases because of the associated reduction in
the effective driving pressure APG until it comes to a complete stop when AP0 = 0 (i.e. D =
D Hm)-
D, jim
I 1 ^m 2
Fig. 5.9. Predicted variation in average grain size D with Z1 and Dnm. for n = 0.5 and D0 = 0 (stable
precipitates).
D, (im
I 1 , [nm]1/n
Fig. 5.10. Predicted variation in average grain size D with Z1 and n for Dum. = 250|Jin and D0 = 0
(stable precipitates). Dotted curves correspond to grain growth in the absence of pinning precipitates.
If grain growth at the same time occurs under the action of a constant drag from impurity
elements in solid solution, the situation becomes more complex. As shown in Fig. 5.10, a
decrease in the time exponent from say 0.5 to 0.2 gives rise to a marked reduction in the slope
of the D-I 1 curves, similar to that observed in Fig. 5.7 for particle-free systems (/= 0). How-
ever, the predicted grain coarsening rate is lower than that evaluated from equation (5-24) due
to the extra drag exerted by the grain boundary pinning precipitates. This leads ultimately to a
stabilisation of the microstructure when D = DUm.
5.3.3.4 Grain growth in the presence of growing precipitates

Very little information is available in the literature on the matrix grain growth behaviour of
metals and alloys in the presence of growing second phase particles. So far, virtually all
modelling work has been carried out on two phase a-(3 titanium alloys.14 Unfortunately, none
of these models can be extended to more complex alloy systems such as steels or aluminium
alloys.
When grain growth occurs in the presence of growing second phase particles, Dum. will no
longer be constant due to the associated increase in the particle rlf ratio with time. As shown in
Chapter 4 of this textbook, the Lifshitz-Wagner theory2021 provides a basis for modelling
particle growth during welding and heat treatment of metals and alloys in cases where the peak
temperature of the thermal cycle is kept well below the equilibrium solvus of the precipitates.
Under such conditions, the particles will coarsen at almost constant volume fraction (f=fo), in
accordance with equation (4-16):
(5-26)
where Qs is the activation energy for the coarsening process, C5 is a kinetic constant, and I2 is
the kinetic strength of the thermal cycle with respect to particle coarsening. The other symbols
have their usual meaning.
If the base metal contains particles of an initial radius ro and volume fraction/^, the limiting
grain size at I2 = 0 (D° lim) can be defined as:
(5-27)
from which
(5-28)
Similarly, when I2 > 0, we may write:
(5-29)
By combining equations (5-26), (5-28), and (5-29), we arrive at the following relationship
between (D Um ) and I2:
(5-30)
It is seen from equation (5-30) that the limiting grain size in the presence of growing
particles depends on the product (k/fo)3I1. In practice, the grain boundary pinning effect of the
precipitates is determined by the relative rates of particle coarsening and grain growth in the
material, i.e. whether the grain boundary mobility is sufficiently high to keep pace with the
increase in DUm during heat treatment. Generally, the pinning conditions are defined by the
(k/fo)3 I1IIx ratio, which after substitution and rearranging yields:
(5-31)
In cases where the parameters c5 ,Qs, M0*, and Qapp are known, the average grain size D
can readily be evaluated from equations (5-22), (5-30), and (5-31) by utilising an appropriate
integration procedure. However, since Qs normally differs from Qapp^ the (klfo)3I1I Ix ratio
will depend on the thermal path during continuous heating and cooling. Consequently, solu-
tion of these coupled equations generally requires stepwise integration in temperature-time
space via a fourth heat flow equation. This problem will be dealt with in Section 5.4.
The situation becomes much simpler if heat treatment is carried out isothermally. Under
such conditions the product (k/fo)311 will only differ from Ix by a proportionality constant m,
which is characteristic of the system under consideration. Accordingly, equation (5-30) can be
rewritten as:
(5-32)
From this we see that the two coupled equations (5-22) and (5-32) can be solved explicitly
for different values of D°nm., n, m, and Z1. Hence, it is possible to present the results in the form
of novel 'mechanism maps' which show the competition between particle coarsening and
grain growth during isothermal heat treatment for a wide range of operational conditions.
Examples of such diagrams are given in Figs. 5.11 and 5.12.
It is evident from these figures that the grain coarsening behaviour during isothermal heat
treatment is very sensitive to variations in the proportionality constant m. For large values of
m, the matrix grains will coarsen at a rate which is comparable with that observed in Fig. 5.7
for particle-free systems (f = 0). This corresponds to a situation where the grain boundary
pinning precipitates will completely outgrow the matrix grains. It is interesting to note that
particle outgrowing is more likely to occur if the time exponent n is small, as shown in Fig.
5.12, because of the associated reduction in the grain boundary mobility in the presence of
impurity elements in solid solution. In other systems, where the proportionality constant m is
closer to unity, the reduced coarsening rate of the precipitates gives rise to a higher Zener
retardation pressure and ultimately to a stagnation in the matrix grain growth. In the limiting
case, when m = 0, the grain growth behaviour becomes idential to that observed in Figs. 5.9
and 5.10 for stable precipitates.
Time exponent n = 0.5
D.jim
I 1 ,^m 2
Fig. 5.11. Predicted variation in average grain size D with Ix and m for D°um. = 50jLim, n = 0.5, and
D0 =0 (growing precipitates).

D,fxm
I1^m1'"
Fig. 5.12. Predicted variation in average grain size D with I1 and m for D°um. - 50|im, n = 0.3, and
D0 = 0 (growing precipitates).
Example (5.2)
Consider a titanium-microalloyed steel with the following chemical composition:
Ti(total): 0.016 wt%, Ti(soluble): 0.009 wt%, N: 0.006 wt%
Assume that the base metal contains an uniform dispersion of TiN precipitates in the as-
received condition, conforming to a limiting austenite grain size T>°um. of 50 |iim. Provided that
boundary drag from impurity elements in solid solution can be neglected (i.e. n ~ 0.5), esti-
mate on the basis of Fig. 5.11 the average austenite grain size D1 in the material after 25 s of
isothermal annealing at 13000C. Relevant physical data for titanium-microalloyed steels are
given below:
(activation energy for diffusion of Ti in austenite)
Solution
The initial volume fraction of TiN in the material can be estimated from simple stoichiometric
calculations by considering the difference between total and soluble titanium. Taking the
atomic weight of Ti and N equal to 47.9 and 14.0 g mol"1, respectively, we obtain:
From this we see that the initial radius of the TiN precipitates in the base metal is close to:
Since heat treatment is carried out under isothermal conditions, the parameters m and Z1 can
be obtained directly from equations (5-31) and (5-22) without performing a numerical integra-
tion:
Similarly, in the case of Z1 we get:
The average austenite grain size can now be read from Fig. 5.11 by linear interpolation
between the curves for m = 10 and 100 Jim. This gives:
Although experimental data are not available for a direct comparison, the predicted grain
size is of the expected order of magnitude. From this it is obvious that considerable austenite
grain growth may occur in titanium-microalloyed steels because of particle coarsening, in
spite of the fact that TiN, from a thermodynamic standpoint, is stable up to the melting point of
the steel. The process can, to some extent, be counteracted by the use of a finer dispersion of
TiN precipitates in the material. For example, if the initial particle radius is reduced by a factor
of five (conforming to a change in I W from 50 to 10 Jim), the austenite grain size of the
annealed material decreases from 75 to 65 jLim, as shown in Fig. 5.13. Nevertheless, since
particle coarsening is a physical phenomenon occurring during high temperature heat treat-
ment of metals and alloys, austenite grain growth cannot be avoided. This explains why, for
instance, conventional titanium-microalloyed steels are not suitable for high heat input weld-
ing due to their tendency to form brittle zones of Widmanstatten ferrite and upper bainite in the
coarse grained HAZ region adjacent to the fusion boundary.22

D,jim
Stable particles
I1-Hm2
Fig. 5.13. Predicted variation in average grain size D with Z1 and m for D°Hm. = 10 um, n = 0.5, and
Do = 0 (growing precipitates).
5.3.3.5 Grain growth in the presence of dissolving precipitates
Little information is available in the literature on the matrix grain growth behaviour of metals
and alloys in the presence of dissolving precipitates. As shown in Chapter 4, the model of
Whelan23 provides a basis for calculating the dissolution rate of single precipitates embedded
in an infinite matrix. If the transient part of the diffusion field is neglected, the variation in the
particle radius r with time t at a constant temperature is given by equation (4-18):
(5-33)
where a is the dimensionless supersaturation (defined in Fig. 4.14), and Dm is the element
diffusivity.
Application of the model to continuous heating and cooling requires numerical integration
of equation (5-33) over the weld thermal cycle:
(5-34)
where I3 is the kinetic strength of the thermal cycle with respect to particle dissolution.
From this relation the following expression for the particle volume fraction can be derived
(see equation (4-22), Chapter 4):
(5-35)
where fo is the initial particle volume fraction.

By substituting Dlim - k(rlf) and D°nm. = Kr0If0) into equations (5-34) and (5-35), it is
possible to obtain a simple mathematical relation which describes the variation in the limiting
grain size with I3 during particle dissolution. After some manipulation, we obtain:
(5-36)
It is seen from equation (5-36) that the limiting grain size increases from D°um. at I3 = 0 to
infinite when I3 = (fo Ik)2 (D°um. )2 • Since the magnitude of the Zener drag, in practice, depends
on the relative rates of grain growth and particle dissolution in the material, the pinning condi-
tions are defined by the (klfo)2131 Ix ratio:
(5-37)
Equation (5-37) shows that the (k/fo)2I3111 ratio is contingent upon the thermal path during
continuous heating and cooling. Consequently, application of the model to welding generally
requires numerical integration of the coupled equations (5-22), (5-36), and (5-37) over the
weld thermal cycle.
However, the integration procedure is largely simplified if heat treatment is carried out
isothermally. In such cases the product (k/fo)213 will only differ from Ix by a proportionality
constant m*, which is characteristic of the system under consideration. By substituting m*Ix
into equation (5-36), we obtain:
(5-38)
From this we see thaUhe two coupled equations (5-22) and (5-38) can be solved explicitly
for different values of Dun., n,m*, and I1. Hence, it is possible to present the results in the
form of novel 'mechanism maps' which show the competition between particle dissolution
and grain growth during isothermal heat treatment for a wide range of operational conditions.
Examples of such diagrams are given in Figs. 5.14 and 5.15.
As expected, the stability of the second phase particles is sensitive to variations in the
proportionality constant m*. Normally, the precipitates will exert a drag on the grain bounda-
ries as long as they are present in the metal matrix. However, when the dissolution process is
completed, the matrix grains are free to grow without any interference from precipitates. This

Complete particle
dissolution
D,jim
Stable particles
I 1 ,nm 2
Fig. 5.14. Predicted variation in average grain size D with Z1 and ra* for D°um. = 50 um, n = 0.5, and
D0 =0 (dissolving precipitates).
Complete particle dissolution
D,jim
I 1 ^m 1 ' 0
Fig. 5.15. Predicted variation in average grain size D with I1 and m* for D°um. - 50 Jim, n - 0.3, and
Do =0 (dissolving precipitates).
means that the grains, after prolonged high temperature annealing, will coarsen at a rate which
is comparable with that observed in Fig. 5.7 for particle-free systems. In the limiting case,
when m* = 0, the grain growth behaviour becomes identical to that shown in Figs. 5.9 and 5.10
for stable precipitates.
Example (5.3)
Consider a niobium-microalloyed steel with the following composition:
Nb(total): 0.025 wt%, Nb(soluble): 0.010 wt%, C: 0.10 wt%
Assume that the base metal contains a fine dispersion of NbC precipitates in the as-received
condition, conforming to a limiting austenite grain size Dnm. of 50 jim. Provided that the
boundary drag from impurity elements in solid solution can be neglected (i.e. n ~ 0.5), esti-
mate on the basis of Fig. 5.14 the average austenite grain size D 7 in the material after 25 s of
isothermal annealing at 13000C. Relevant physical data for niobium-microalloyed steels are
given below:
Solution
The initial volume fraction of NbC in the material can be estimated from simple stoichiometric
calculations by considering the difference between total and soluble niobium. Taking the
atomic weight of Nb and C equal to 92.9 and 12.0 g mol"1, respectively, we obtain:
From this we see that the radius of the NbC precipitates in the base metal is close to:
As shown in Example 4.6 (Chapter 4), the dimensionless supersaturation of niobium a ^

adjacent to the particle/matrix interface during dissolution can be written as:
By substituting this value into the expression for the proportionality constant m*, we obtain:
Moreover, at 1300°C the value OfZ1 becomes:
The average austenite grain size can now be read from Fig. 5.14 by interpolation between
the curves for m* = 1 and/= 0 (free grain growth). This gives:
Since the calculated value of D1 is reasonably close to that observed for a particle-free
system, it means that the presence of a fine dispersion of NbC in the base metal has no signifi-
cant effect on the resulting austenite grain size under the prevailing circumstances. Other
Next Page
types of niobium microalloyed steels may reveal a different grain coarsening behaviour, de-
pending on the chemical composition, size distribution, and initial volume fraction of the base
metal precipitates. However, the pattern remains essentially the same, i.e. the growth inhibi-
tion is always succeeded by grain coarsening as long as the precipitates are thermally unstable.
5.4 Grain Growth Diagrams for Steel Welding
In welding the temperature will change continuously with time, which makes predictions of
the HAZ grain coarsening behaviour rather complicated. The method adopted from Ashby et
^ 24,25 j s b asec i o n m e jd e a o f integrating the elementary kinetic models over the weld thermal
cycle where the unknown kinetic constants are determined by fitting the integrals at certain
fixed points to data from real or simulated welds. Although the introduction of the Zener drag
in the grain growth equation largely increases the complexity of the problem, the methodology
and calibration procedure remain essentially the same. This means that the results from such
complex computations can be presented in the form of simple grain growth diagrams which
show contours of constant grain size in temperature-time space.
5.4.1 Construction of diagrams

A grain growth model for welding consists of two components, i.e. a heat flow model, and a
structural (kinetic) model.
5.4.1.1 Heat flow models

As a first simplification, the general Rosenthal equations26 are considered for the limiting case
of a high net power qo and a high welding speed v, maintaining the ratio qo/v within a range
applicable to arc welding. It has been shown in Chapter 1 that under such conditions, where no
exchange of heat occurs in the .^-direction, the following equations apply:
Thick plate welding (2-D heat flow)
(5-39)
Thin plate welding (1-D heat flow)

Previous Page
types of niobium microalloyed steels may reveal a different grain coarsening behaviour, de-
pending on the chemical composition, size distribution, and initial volume fraction of the base
metal precipitates. However, the pattern remains essentially the same, i.e. the growth inhibi-
tion is always succeeded by grain coarsening as long as the precipitates are thermally unstable.
5.4 Grain Growth Diagrams for Steel Welding
In welding the temperature will change continuously with time, which makes predictions of
the HAZ grain coarsening behaviour rather complicated. The method adopted from Ashby et
^ 24,25 j s b asec i o n m e jd e a o f integrating the elementary kinetic models over the weld thermal
cycle where the unknown kinetic constants are determined by fitting the integrals at certain
fixed points to data from real or simulated welds. Although the introduction of the Zener drag
in the grain growth equation largely increases the complexity of the problem, the methodology
and calibration procedure remain essentially the same. This means that the results from such
complex computations can be presented in the form of simple grain growth diagrams which
show contours of constant grain size in temperature-time space.
5.4.1 Construction of diagrams

A grain growth model for welding consists of two components, i.e. a heat flow model, and a
structural (kinetic) model.
5.4.1.1 Heat flow models

As a first simplification, the general Rosenthal equations26 are considered for the limiting case
of a high net power qo and a high welding speed v, maintaining the ratio qo/v within a range
applicable to arc welding. It has been shown in Chapter 1 that under such conditions, where no
exchange of heat occurs in the .^-direction, the following equations apply:
(5-39)

5.4.1.2 Grain growth model
The use of equation (5-22) for prediction of the HAZ grain structure requires quantitative
information about the different kinetic constants entering the model. In the following we shall
assume that data for the time exponent n, the activation energies Qam and Qs, the temperature
dependence of the dimensionless supersaturation a and the element diffusivity Dm as well as
the initial and limiting grain sizes D0 and Dum. are available in the literature. From knowl-
edge of these parameters it is possible to calibrate the model against data from real or simu-
lated welds by adjusting the remaining kinetic constants so that a good agreement is obtained
between theory and experiments.
5.4.1.3 Calibration procedure

The calibration procedure involves the following basic steps:
Evaluation of experimental data

Supposejhat the mean grain size at two different locations in the HAZ is known (designated
D1 and Z)2, respectively). At each of these locations the peak temperature T and the cooling
time within a specific temperature range (e.g. from 800 to 5000C), Af8/5, are given. The thermal
cycles for the points can then be computed from either equation (5-39) or (5-40). Using these
temperature profiles and an appropriate value for the activation energy Q , the I1IM* ratio
is calculated from the integral:
(5-41)
where the limits tx and t2 refer to the total time spent in the thermal cycle from the chosen
reference temperature Tc to the peak temperature Tp and down again to Tc.
Tuning of coarsening model

When the parameters n, Qapp., Qs, D0, and D°lim are known, it is possible to evaluate the
remaining (unknown) kinetic constants from equations (5-22) and (5-30) by an iterative proce-
dure. The following parameter is defined for this purpose:
(5-42)
The next step is to calculate the integral
(5-43)
for the same pairs of values of Tp and Ar875 as above. The differential grain growth equation
can now be solved by selecting an appropriate starting value for M0* and evaluating the corre-
sponding Q-value which conforms to a mean grain size of D\ and D2, respectively. The
computations are repeated by adjusting M0* until a contour in Mo*-Q space is built up for each
grain size. The accepted values of M* and Q are then found by considering the intersection
point between the two curves, as shown schematically in Fig. 5.16(a).
(a) MS.
Accepted,
"" value
" ~vaTue~
Accepted^
a
(b) MJ
Accepted.
~ value" "
Acce|3ted_
value" "
Q*
Fig. 5.16. Method for calibrating unknown kinetic constants to experimental grain growth data;
(a) M*o-Q. (coarsening model), (b) M*o-Q* (dissolution model).
Tuning of dissolution model

In this case the unknown kinetic constants are combined in a single calibration parameter £2*,
defined as:
(5-44)
where a° and D°m include all constants entering the expressions for the dimensionless
supersaturation a and the element diffusivity Dm, respectively.
Under such conditions equation (5-37) becomes:
(5-45)
where AH* is the standard enthalpy of the dissolution reaction per mole of the diffusate (de-
fined by the solubility product in equation (4-5), Chapter 4), and Qd is the activation energy for
diffusion of the less mobile constituent atom of the precipitates in the matrix.
By calculating the integral in equation (5-45) for the same pairs of values of Tp and Af8/5 as
above and selecting an appropriate starting value for Mo* in equation (5-41), it is possible to
build up a contour in Mo*-£T space for each grain size that satisfies the differential grain
growth equation. The accepted values of M* and Q* are then found by considering the inter-
section point between the curves representing D1 and D2 in Fig. 5.16(b).
5.4.1.4 Axes and features of diagrams

The microstructural information calculated using the mathematical models described in the
previous sections may be plotted on various kinds of welding diagrams. The graphical rep-
resentation chosen here has been adopted from Ashby et al?^25 The process diagrams have
axes of weld input energy and peak temperature, and display contours of constant grain size
along with information about the thermal stability of the grain boundary pinning precipitates.
The axes can, in turn, be converted into an equivalent cooling time, Atm, and isothermal zone
width Ar* or Aym through equations (5-39) and (5-40):
(5-46)
(5-47)
where Tp < Tn, (Tn, is the melting point).
(5-48)
(5-49)
where Tp < Tn,.

Welding diagrams of this kind are very useful, since they summarise the effect of the impor-
tant process variables in a systematic and illustrative manner at the same time as they provide
a good overall indication of the grain growth behaviour of materials during welding. In addi-
tion, the diagrams can be used for quantitative predictions of the austenite grain size across the
HAZ of steel welds for a wide range of operational conditions. This will be illustrated below
in various numerical examples.
5.4.2 Case studies

The following section describes grain growth diagrams for different types of steels. The pa-
rameters used to construct the maps are either tabulated or included in the text. Some of these
are taken from the literature, while others are arrived at by fitting the theory of the previous
sections to data from real or simulated welds according to the procedure shown in Fig. 5.16.
To obtain a consistent basis, all grain sizes reported here conform to the mean linear inter-
cept.28 Conversion to three-dimensional grain sizes may then be done through equation (5-4)
or by the use of other appropriate conversion factors (e.g. 1.776 as recommended by Ashby et
a/.24'25).
5.4.2.1 Titanium-microalloyed steels

Titanium-microalloyed steels are widely used in welded structures. From a thermodynamic
standpoint, additions of small amounts of titanium to steel would be expected to impede austenite
grain coarsening during welding by virtue of its ability to form stable nitrides even at high
temperatures (see discussion in Chapter 4). However, certain restrictions must be adhered
during casting and subsequent thermomechanical processing of the steel so that the number
density of TiN particles is sufficiently high to retard grain growth.22 This currently limits the
use of titanium for austenite grain size control to continuously cast and controlled rolled steels.
Typical HAZ grain growth diagrams for Ti-microalloyed steels can be constructed on the
basis the experimental data reported by Ion et al.25 Tables 5.1 and 5.2 contain information
about the steel chemical composition and the parameters used in the computations, while Fig.
5.17 gives an overall indication of the accuracy of the predictions after calibration of the model
against two experimental data points.
Table 5.1 Chemical composition of Ti-microalloyed steel used by Ion et al.25 (in wt%).
C Si Mn P S Al Ti N
0.12 0.23 1.53 0.005 0.006 0.02 0.011 0.009
Table 5.2 Data used to construct welding maps for Ti-microalloyed steel (compiled from miscellane-
ous sources).
" I ~PP I Q? I »°Hm. I K I ^* I «

(kJmol-1) (kJmol-1) (|xm) (jun) (IJLm2S-1) (1JLm3Ks-1)
0.5 224 240 20 9 7.7 X 109 9.9 X 1014
!Activation energy for diffusion of Ti in austenite.

Ti-microalioyed steel
E
=i
s
(a
I
Oi
I
i
Calibration point
Predicted grain size,jim

Fig. 5.17. Comparison between measured and predicted HAZ austenite grain sizes after calibration of
model to data reported by Ion et al.25 for Ti-microalloyed steel (simulated thick plate welds).
The response of the base material to welding under 2-D and 1-D heat flow conditions is
shown in Fig. 5.18(a) and (b), respectively. As expected, the presence of TiN particles is seen
to retard austenite grain growth within the heat affected zone during welding. However, since
particle coarsening is a physical process occurring at temperatures well below the equilibrium
solvus of the precipitates, the problem cannot be eliminated. This means that a coarse grained
region will always form adjacent to the fusion boundary, even at very low heat inputs, as
indicated by the nomograms in Fig. 5.18(a) and (b).
Example (5.4)
Consider SA welding on a thick plate of a titanium-microalloyed steel under the following
conditions:
Evaluate on the basis of the nomograms in Fig. 5.18(a) the variation in the austenite grain
size across the fully transformed HAZ after welding. Estimate also the total width of the HAZ
(referred to the fusion boundary) under the prevailing circumstances.
Solution
First we calculate the net heat input per unit length of the weld:
Readings from Fig. 5.18(a) give the HAZ austenite grain size profile shown in Fig. 5.19.
(a) Thick plate welding (2-D heat flow)
Prior austenite grain size (^m)
Relative size of pinning precipitates (r/ro)
Net heat input (qo/v), kJ/mm
Cooling time, At B/5 , s

(b) Thin plate welding (1-D heat flow)

Prior austenite grain size dim)
Relative size of pinning precipitates (r/ro)
Net heat input (qo/vd), kJ/mm2
FZg. 5.18. HAZ grain growth diagrams for titanium-microalloyed steel; (a) Thick plate welding (2-D heat
flow), (b) Thin plate welding (1-D heat flow). No preheating (T0 = 200C).
SAW (Ti-microalloyed steel)
Austenite grain size, ^m
Fig. 5.19. Predicted variation in austenite grain size across the fully transformed HAZ of a Ti-microalloyed
steel weld (Example 5.4).
The total width of the fully transformed HAZ can now be estimated from equation (5-47)
by using data from Table 1.1 (Chapter 1). Taking the Ac3 -temperature equal to 9100C, we
obtain:
Based on the quoted value of £2 in Table 5.2, it is also possible to estimate the initial volume
fraction of TiN in the base metal. Taking C5 « 6.67 X 104 |im3 K s"1 and k « 0.5 for titanium-
microalloyed steels, we obtain from equation (5-42):
A comparison with the experimental data of Ringer et aill shows that a volume fraction of
2 X 10"4 is reasonably close to that measured by microscopic assessment methods.
5.4.2.2 Niobium-microalloyed steels

This class of steel is used for a variety of applications, including ship building, pressure ves-
sels, oil platforms and bridges. The steels contain small amounts of microalloying elements
such as niobium, vanadium, and aluminium which during thermomechanical processing com-
bine with carbon and nitrogen to form fine dispersions of grain boundary pinning precipitates.
They are readily weldable, but suffer from severe HAZ grain coarsening due to dissolution of
the carbo-nitrides at elevated temperatures.1822
Typical HAZ grain growth diagrams for Nb-microalloyed steels (with chemical composi-
tion as in Table 5.3) can be constructed on the basis of the experimental data of Ion et al. 25 The
parameters used to compute the diagrams are listed in Table 5.4. Figure 5.20 shows a correla-
tion between predicted and observed grain sizes after calibration of the model against two
experimental data points.
It is evident from the nomograms in Fig. 5.21 (a) and (b) that considerable austenite grain
growth occurs during welding of niobium-microalloyed steels because of dissolution of the
grain boundary pinning precipitates. On the average, the HAZ austenite grain size adjacent to
the fusion boundary is four to six times larger than that observed for titanium-microalloyed
steels. This gives rise to a high HAZ hardenability, which facilitates formation of low tem-
perature transformation products such as martensite and bainite during welding ?2
Example (5.5)
In Table 5.4 the quoted value for the effective grain boundary pinning constant Q* is 5.0 X
1018 |im 2 s"1. Based on equations (5-27) and (5-44), estimate the initial volume fraction/, and
radius ro of NbC in the parent material under the prevailing circumstances.
Table 5.3 Chemical composition of Nb-microalloyed steel used by Ion et al.25 (in wt%).
C Si Mn P S Al Nb N
0.12 0.16 0.91 0.002 0.005 0.04 0.021 0.011
Nb-microalloyed steel
SAW
GMAW
Laser welding
Weld simulation
Measured grain size, urn
Calibration point
Predicted grain size, jam
model to data reported by Ion et al.25 for Nb-microalloyed steel (real and simulated thick plate welds).
(a)
Prior austenite grain size (urn)
Relative volume fraction of pinning precipitates (f/fo)
Cooling time, M8^, s


Net heat Input (qo/vd), kJ/mm2
Fig. 5.21. HAZ grain growth diagrams for niobium-microalloyed steel; (a) Thick plate welding (2-D heat
Table 5.4 Data used to construct welding maps for Nb-microalloyed steel (compiled from miscellane-
ous sources).
n I Qapp. I ^*f I 0} I D°lim. I ^ I K* I "^

(kJmol-1) (kJmol-1) (kJ mol-1) ((Jim) (|xm) (1Xm2S-1) (1Xm2S-1)
0.5 224 130 343 11 6 37 X 109 5.0 X 1018
fEstimated from the solubility product of NbC in austenite.

Âctivation energy for diffusion of Nb in austenite.
Reasonable average values for a 0 , D°m, k, and D°nm. are given below:
20.5 (estimated from the solubility product of NbC in austenite, Example

4.6)
5.9 X 1010 Jim2 s"1 (diffusion constant for Nb in austenite)
0.5 (Zener coefficient)
11 |Lim (from Table 5.4)
Solution
The initial volume fraction of NbC can be estimated from equation (5-44). After rearranging
this equation, we obtain:
The corresponding radius of the precipitates can now be obtained from equation (5-27):
Although experimental data are not available for a direct comparison, the calculated values
for/o and ro are reasonable and of the expected order of magnitude (see Example 5.3).
5.4.2.3 C-Mn steel weld metals

In C-Mn steel weld metals the volume fraction of non-metallic inclusions is considerably
higher than for normal cast steel products because of the limited time available for growth and
separation of the deoxidation products (see discussion in Chapter T). For the same reason, the
weld metal inclusions are also significantly smaller in dimension and more finely dispersed.
Since oxides and sulphides will neither coarsen nor dissolve during the weld thermal cycle,
they will ultimately lead to a stabilisation of the austenite grain structure within the reheated
regions of multipass steel weld metals. This is illustrated schematically in Fig. 5.22.
Grain growth diagrams for C-Mn steel weld metals (with chemical compostion as in Table
5.5) can be constructed on the basis of the theory described in the previous sections by select-
ing reasonable average values for the kinetic constants in equation (5-22). The combination of
parameters listed in Table 5.6 gives a fair agreement between predictions and measurements,
as shown in Fig. 5.23.
The response of the material to reheating above the Ac3-temperature under 2-D and 1-D
heat flow conditions is illustrated in Fig. 5.24(a) and (b). As expected, the presence of finely
Weld metal
Base plate
Fully transformed HAZ

AustenJte grain size
Peak temperature
Fig. 5.22. Schematic diagram illustrating the effect of non-metallic inclusions on the weld metal grain
coarsening behaviour.
Table 5.5 Chemical composition of C-Mn steel weld metal used by Kluken et al.29 (in wt%).
C O Si Mn P S N Nb V Al Ti
0.09 0.034 0.48 1.86 0.01 0.01 0.005 0.004 0.02 0.018 0.005
Table 5.6 Data used to construct welding maps for C-Mn steel weld metal (compiled from
miscellaneous sources).
" I ~pp. I K
1
I bHm. I K?
(kJmol" ) (|xm) (|xm) (JJLm2S-1)
0.5 224 10 95 20X109
dispersed inclusions within the weld metal gives rise to a strong austenite grain boundary
pinning effect, similar to that documented for TiN in microalloyed steels. However, since no
particle coarsening occurs in the present case, it means that a small austenite grain size
(< 95jnm) is preserved at all relevant peak temperatures, irrespectively of the applied heat
input. Accordingly, the weld metal grain growth behaviour is seen to be quite different from
that of the base metal, even when the nominal chemical composition has not been significantly
changed by the welding process.
Austenite grain size, jim
C-Mn steel weld metal

Peak temperature: —1350° C
Cooling time, A t . . s
8/5
Fig. 5.23. Comparison between measured and predicted austenite grain sizes after calibration of model
to data reported by Kluken et al.29 for C-Mn steel weld metal (simulated thick plate heat cycles).
Example (5.6)
Consider deposition of a cap layer (GMAW) on the top of a thick multipass Nb-microalloyed
steel weld under the following conditions:
Estimate on the basis of the nomograms in Figs. 5.21(a) and 5.24(a) the variation in the
prior austenite grain size across the fully transformed HAZ at different locations along the
periphery of the weld after arc extinction. The situation is illustrated in Fig. 5.25.
Solution
Readings from Figs. 5.21(a) and 5.24(a) give the HAZ austenite grain size profiles shown
in Fig. 5.26. From this we see that the prior austenite grain size adjacent to the fusion bound-
ary of a cap layer will vary significantly with position along the periphery of the weld, depend-
ing on the type of material sampled (i.e. base plate or weld metal).
5.4.2.4 Cr-Mo low-alloy steels

Chromium-molybdenum low-alloy steels are widely used in the petroleum industry and in
high-power-generating equipment because of their good corrosion and oxidation resistance
(a)
Cooling time, At 8/5 , s

(b) Thin plate welding (1 -D heat flow)

Fig. 5.24. Grain growth diagrams for reheated C-Mn steel weld metal; (a) Thick plate welding (2-D heat
HAZ Base plate
Fig. 5.25. Sketch of multipass steel weld in Example 5.6.

Austenite grain size, ^m
Dissolution of NbC
Fig. 5.26. Predicted variation in austenite grain size across the fully transformed HAZ at different loca-
tions along the periphery of the weld (Example 5.6).
and their high creep strength. They are readily weldable, although reheat cracking and cold
cracking in the HAZ and fusion zone may be a problem.30 In practice, these difficulties can be
overcome by the choice of an appropriate welding procedure (e.g. preheating in combination
with a low heat input), which reduces austenite grain growth and maximises the proportion of
the HAZ refined by subsequent weld passes.3031
The grain growth data of Miranda and Fortes31 provide a basis of calibrating the kinetic
constants in equations (5-22) and (5-45), as shown in Fig. 5.27. Information about steel chemical
composition and parameters used to construct the maps are contained in Table 5.7 and 5.8,
respectively.
It is seen from the nomograms in Fig. 5.28(a) and (b) that considerable austenite grain
growth occurs in Cr-Mo low-alloy steel weldments due to dissolution of molybdenum carbide
(Mo2C) at elevated temperatures. However, the maximum HAZ austenite grain size is signifi-
Table 5.7 Chemical composition of Cr-Mo steel used by Miranda and Fortes31 (in wt%).
C Si Mn P S Cr Mo Ni Cu Al V
0.10 0.24 0.46 0.01 0.01 2.19 0.94 0.23 0.1 0.01 0.01
Calibration points
Predicted grain size, jam
Symbol * w s >
Measured grain size, |im

model to data reported by Miranda and Fortes31 for Cr-Mo low-alloy steel (SAW-thick plates).
Table 5.8 Data used to construct welding maps for Cr-Mo low-alloy steel (from Refs. 24 and 31).
n I QapP. I A//*+ I QP I b°lm T ~ F I MO* I QT~

(kJmol-1) (kJmol-1) (kJmor"1) (^m) (fim) (IJLm1711S-1) (IXm2S-1)
0.32 180 71 300 20 8 1.3 X 109 7.8 X 1017
f Estimated from the solubility product of Mo2C in austenite.

f f Activation energy for diffusion of Mo in austenite.
cantly smaller than that observed during welding of Nb-microalloyed steels, in spite of the fact
that Mo2C is less stable than NbC. This situation can probably be attributed to drag from
alloying and impurity elements in solid solution, which retard grain growth through elastic
attraction of the atoms towards the open structure of the grain boundary (indicated by a time
exponent n = 0.32 in Table 5.8).
5.4.2.5 Type 316 austenitic stainless steels

The 316 series of stainless steels have good general corrosion resistance, but are sensitive to
stress corrosion cracking in the presence of chlorides. They are used extensively in the chemi-
cal industry and in power plants, particularly in advanced nuclear technology. During welding
dissolution of the base metal chromium carbides (e.g. C^C 6 ) will occur in parts of the HAZ
where the peak temperature of the thermal cycle has been above about 8500C. This gives rise
to considerable austenite grain growth.24
HAZ grain growth diagrams for this type of steel (with chemical composition as in Table
5.9) can be constructed on the basis of the experimental data of Ashby and Easterling,24 using
the parameters listed in Table 5.10. Figure 5.29 shows a correlation between predicted and
observed grain sizes after calibration of the model against two experimental data points.
(a) Thick plate welding (2-D heat flow)
Relative volume fraction of pinning precipitates (Vi0)
Cooling time, At 8751 s


Relative volume fraction of pinning precipitates (f/y
Fig. 5.28. HAZ grain growth diagrams for Cr-Mo low-alloy steel; (a) Thick plate welding (2-D heat
Table 5.9 Chemical composition of type 316 austenitic stainless steel used by Ashby and Easterling24
(in wt%).
C Mn Si Cr Ni Mo P S
0.05 2.0 1.0 19 11 2.3 0.045 0.030
Table 5.10 Data used to construct welding maps for type 316 austenitic stainless steel (compiled from
miscellaneous sources).
n I Qapp. 1
I Atf*+ I Q/ I D°lm.
1 1
I K I M
o* I ^*
(kJmol- ) (kJmol- ) (kJ mol" ) (|jim) (fxm) (1JLm2S-1) (1Xm2S-1)
0.5 224 60 240 18 18 3.4 X IQ9 2.1 X IQ14
!Estimated from data quoted by Kou32 for Cr23C6 in 304 stainless steel.
Âctivation energy for diffusion of Cr in austenite.
Measured grain size, u.m
Calibration point
Predicted grain size, ^m
model to data reported by Ashby and Easterling24 for type 316 stainless steel (simulated thick plate
welds).
It follows from the nomograms in Fig. 5.30(a) and (b) that the presence of Cr23C6 has little
influence on the HAZ grain coarsening behaviour because it dissolves at a fairly low tempera-
ture. This results in a rather coarse austenite grain structure adjacent to the fusion boundary,
which in certain cases may exceed 100 jim.
Austenite grain size ^x m)
Cooling time, At8/5, s

Thin plate welding (1 -D heat flow)

Austenite grain size (JLIITI)
Fig. 5.30. HAZ grain growth diagrams for type 316 austenitic stainless steel; (a) Thick plate welding
(2-D heat flow), (b) Thin plate welding (1-D heat flow). No preheating (T0 = 200C).
Example (5.7)
Consider GTA butt welding of a 2mm thin sheet of type 316 austenitic stainless steel under the
Provided that the conditions for one dimensional (1-D) heat flow are met, estimate on the
basis of the nomograms in Fig. 5.30(b) the variation in the austenite grain size across the HAZ
after welding. Estimate also the total width of the HAZ (referred to the fusion boundary)
under the prevailing circumstances.
Solution
First we calculate the net heat input per mm2 of the weld:
Readings from Fig. 5.30(b) give the HAZ austenite grain size profile shown in Fig. 5.31.
The total width of the grain growth zone can be estimated from equation (5-49) by using
data from Table 1.1 (Chapter 1). Taking the peak temperature for incipient dissolution of
Cr23C6 equal to 9000C, we obtain:
Austenite grain size, jim
Fig. 5.31. Predicted variation in austenite grain size across the HAZ of a 316 austenitic stainless steel
weld (Example 5.7).
5.5 Computer Simulation of Grain Growth
The multivariable characteristics of grain growth impose several restrictions on the use of
analytical modelling techniques for a mathematical description of the grain evolution during
welding. However, many of these restrictions can be relaxed by employing novel computa-
tional techniques in mainframe facilities which enable the modelling of exact grain shapes in
topological connected microstructures.
Historically, the application of the computer simulation technique to poly cry stalline micro-
structures has evolved along two different paths. The first approach treats the grain boundaries
as continuous interfaces, which are governed by determistic equations of motion.33"37 The
other approach is a Monte Carlo-based technique that takes into account explicitly the interac-
tions among individual grains by discretising the microstructure to construct an image of the
grain aggregates in the computer.38"43 Both approaches have been applied to the modelling of
grain growth phenomena in welding 374445 with the objective of incorporating important side-
effects which cannot readily be accounted for in a simple analytical treatment of the process.
5.5.1 Grain growth in the presence of a temperature gradient

In comparing HAZ grain growth in thermally simulated and real welds it is sometimes found
that the maximum grain size is larger in simulated specimens than in a weld HAZ when com-
parison is made on the basis of a similar temperature-time programme.16 A possible explana-
tion to this observation is the phenomenon of 'thermal pinning' which can be attributed to the
presence of steep temperature gradients in a weld HAZ. This effect has been disregarded in the
previous analytical analysis, but can readily be incorporated in a numerical model.
Following the treatment of Fortes and Soares37 the effect of 'thermal pinning' can be simu-
lated by considering 1-D grain growth in a thin polycrystal with a Gaussian grain distribution.
As an illustration of principles, a constant mobility gradient was assumed, the extreme mobilities
being in the ratio of 5:1 and 10:1 in different simulations. Simulation with a uniform grain
mobility (UNIF) was then carried out for comparison. The polycrystal was subsequently di-
vided into three regions of the same size, i.e. cold (C), intermediate (IM), and hot (H) regions,
respectively, and the grain size distributions were obtained for each of these regions and for the
entire polycrystal (G: global region). The simulations were carried out until the total number
of grains were reduced from initially 5000 to 1500 grains. The results of the computations are
summarised in Fig. 5.32.
It is evident from Fig. 5.32(a) that grain growth is slower in a mobility gradient than under
isothermal heat treatment conditions (UNIF). This effect becomes more pronounced as the
mobility gradient increases. Moreover, a closer inspection of Figs. 5.32(b) and (c) reveals that
the grain growth rate of the entire crystal (global region: G) is slower than the corresponding
coarsening rate of the intermediate region (IM), which again indicates that grain growth is
slower in a mobility gradient. Although the simulation results in Fig. 5.32 are not directly
applicable to welding (in welding the temperature, and hence the mobility, at a fixed point in
the crystal will not be constant but vary with time), these findings provide a strong motivation
for incorporating the 'thermal pinning' effect in future refinements of the HAZ grain growth
models.
(a)
D x 10 (arbitrary units)
t1/2 (arbitrary units)
(b)
(C)
Fig. 5.32. Effect of 'thermal pinning' on grain growth; (a) Plot of D vs t for different simulation condi-
tions, (b) Coarsening kinetics of different thermal regions for a mobility gradient of 5:1, (c) Same as in
(b) for a mobility gradient of 10:1. Data from Fortes and Soares.37
5.5.2 Free surface effects
In a weld HAZ the fusion line represents a physical barrier against grain growth which cannot
be exceeded. In principle, it can be regarded as a free surface, which means that the grain
boundaries must meet the fusion line at right angles.
The simulation results of Saetre and Ryum36 provide a basis for evaluating to what extent
grain growth under isothermal conditions is affected by the presence of a free surface. Figure
5.33 shows the evolution of a 2-D Voronoi structure with time. In Fig. 5.33(a) all grain boundaries
are straight lines, but the triple line junctions are not in equilibrium. However, adjustments of
the triple line junctions into equilibrium positions lead to local curvatures of the grain bounda-
ries near the junctions, and the grain growth process is thus initiated. Initially, the Voronoi
structure contained 485 grains, but this number is gradually reduced during the coarsening
process (Figs. 5.33(b), (c) and (d)). A qualitative inspection of the diagrams reveals, on the
other hand, no clear difference in the grain size in the radial position, which indicates that the
constrain provided by the free surface is only of minor importance in the present context. It
should be noted that this does not exclude the possibility that the HAZ grain size is influenced
by the presence of a fusion boundary, since other effects such as surface grooving and solute
segregation (not studied here) can impose additional restrictions on the system by contributing
to physical pinning of the grain boundaries. Consequently, further modelling work is required
to explore these possibilities.
References
1. H. Hu and B.B. Rath: Metall Trans., 1970,1, 3181-3184.

2. J.W. Cahn: Acta Metall, 1962,10, 789-798.
3. K. Lucke and H. Stuwe: in Recovery and Recrystallization of Metals, 1963, New York,
Interscience.
4. E. Nes, N. Ryum and O. Hunderi: Acta Metall, 1985, 33, 11-22.
5. E.E. Underwood: Quantitative Stereology, 1970, London, Addison-Wesley Publ. Co.
6. C S . Pande: Acta Metall, 1987, 35, 2671-2678.
7. B.R. Patterson and Y. Liu: Metall Trans., 1992, 23A, 2481-2482.
8. P. Hellman and M. Hillert: Scand. J. Metall, 1975, 4, 211-219.
9. R.A. Vandermeer: Acta Metall, 1967,15, 447-458.
10. C. Zener (quoted by C S . Smith): Trans. AIME, 1948,175, 15-51.
11. CH. Worner and P.M. Hazzledine: JOM, 1992, 44 (No. 9), 16-20.
12. B.B. Rath and H. Hu: Trans. TMS-AIME, 1969, 245, 1243-1252, ibid., 1577-1585.
13. T. Gladman: Proc. Roy. Soc, 1966, 294A, 298-309.
14. G. Grewal and S. Ankem: Acta Metall Mater, 1990, 38, 1607-1617.
15. T. Skaland and 0. Grong: University of Trondheim, The Norwegian Institute of Technology,
Trondheim Norway (unpublished work).
16. K.E. Easterling: Introduction to the Physical Metallurgy of Welding, 2nd Edn, 1992, Oxford,
Butterworth-Heinemann Ltd.
17. S.P. Ringer, W.B. Li and K.E. Easterling: Acta Metall, 1989, 37, 831-841.
18. O.M. Akselsen, 0. Grong, N. Ryum and N. Christensen: Acta Metall, 1986, 34, 1807-1815.
19. A.M. Brown and M.F. Ashby: Acta Metall, 1980, 28, 1085-1101.
20. J.M. Lifshitz and V.V. Slyozov: J. Physics Chem. Solids, 1961,19, 35-50.
21. C. Wagner: Z. Electrochem., 1961, 65, 581-591.
22. 0. Grong and O.M. Akselsen: Metal Construction, 1986,18, 557-562.
(a) (b)
(C) (d)
Fig. 5.33. The evolution of 2-D grain structures during normal grain growth; (a) Initial Voronoi structure,
485 grains, (b) 456 grains, (c) 349 grains, (d) 251 grains. Data from Saetre and Ryum.36
23. MJ. Whelan: Metal ScL J., 1969, 3, 95-97.

24. M.F. Ashby and K.E. Easterling: Acta MetalL, 1982, 30, 1969-1978.
25. J.C. Ion, K.E. Easterling and M.F. Ashby: Ada MetalL, 1984, 32,1949-1962.
26. D. Rosenthal: Trans. ASME, 1946, 68, 849-866.
27. O.R. Myhr and 0. Grong: Acta MetalL Mater., 1990, 38, 449-460.
28. ASTM Standard, El 12-84 (1984).
29. A.O. Kluken, 0. Grong and H. Hemmer: Technical Report STF34 F87093, 1987, Trondheim
(Norway), Sintef-Division of Metallurgy.
30. RJ. Alberry and W.K.C. Jones: Metals Technology, 1977, 4, 45-51; ibid., 360-364; ibid., 557-
566.
31. R.M. Miranda and M.A. Fortes: Mater. ScL Eng., 1989, A108, 1-8.
32. S. Kou: Welding Metallurgy, 1987, Toronto, John Wiley & Sons.
33. V.E. Fradkov and D.G. Udler: in Simulation and Theory of Evolving Microstructures (Eds M.R
Anderson and A.D. Rollett), 1990, TMS, Warrendale, Pa, pp. 15-29.
34. HJ. Frost, CV. Thompson and D.T. Walton: ibid., pp.31-39.
35. K. Kawasaki, T.Nagai and S. Ohta: ibid., pp.65-77.
36. T.O. Saetre and N. Ryum: in Modelling of Coarsening and Grain Growth (Eds C S . Pande and
S.R Marsh), 1993, The Minerals, Metals & Materials Society, pp.281-294.
37. M.A. Fortes and A. Soares: ibid., pp. 257-270.
38. A.D. Rollett, DJ. Srolovitz and M.R Anderson: Acta Metall, 1989, 37, 1227-1240.
39. G.S. Grest, M.R Anderson, DJ. Srolovitz and A.D. Rollett: Scripta Metall Mater., 1990, 24,
661-665.
40. DJ. Srolovitz, G.S. Grest and M.R Anderson: Acta Metall, 1986, 34, 1833-1845; ibid, 1988,
36,2115-2128.
41. T.O. Sartre, O. Hunderi and E. Nes: Acta Metall, 1986, 34, 981- 987.
42. K. Marthinsen, O. Lohne and E. Nes: Acta Metall, 1989, 37, 135-145.
43. A.D. Rollett, MJ. Lutony and DJ. Srolovitz: Acta Metall. Mater., 1992, 40, 43-55.
44. A. Kern, W. Reif and U. Schriever: Proc. Int. Conf. on Grain Coarsening, 18-21 June, 1991,
Rome, Mater. ScL Forum, 1992, vols 94-96, 709-714, Trans. Tech Publications.
45. B. Radhakrishnan and T. Zacharia: Proc. Int. Conf. on Modeling and Control of Joining Proc-
esses, 8-10 Dec, 1993, Orlando, FL, American Welding Society.
Appendix 5.1
Nomenclature
thermal diffusivity (mm2 s"1) limiting grain size at h - 0 (|im or m)
end temperature of ferrite to austenite element diffusivity (jum2 s"1, mm2 s"1
transformation (0C or K) or m2 s"1)
nominal alloy composition (wt%, at.%

kinetic constant in expression for Dm
or mol rrr3)
(|Ltm2 s"1, mm2 s"1 or m2 s"1)
various kinetic constants and tempera-
ture-dependent parameters particle volume fraction
plate thickness (mm) initial particle volume fraction
grain size (jLirn or m) maximum pinning force exerted by a

single particle ( J m 1 )
mean grain size (\im or m)
mean linear grain intercept (JJJTI or m)
GTAW gas tungsten arc welding
initial grain size (jim or m)
standard enthalpy of dissolution reac-
limiting grain size (jim or m) tion per mole of the diffusate (J mol"1
or kJ mol"1)
amperage (A) activation energy for particle coarsen-
ing (J mor 1 or kJ mor 1 )
kinetic strength of thermal cycle with
respect to grain growth (jam1/n) particle radius (nm, jam or m)
kinetic strength of thermal cycle with initial particle radius at t - 0 (nm, jam
respect to particle coarsening (fxm3) or m)
kinetic strength of thermal cycle with two-dimensional radius vector in y-z

respect to particle dissolution (jim2) plane (mm)
Zener coefficient universial gas constant (8.314 J Kr1

mor 1 )
constant related to the relative rates of
grain growth and particle coarsening in grain boundary area per unit volume
the material (jum3"1/n) (jam2 per jum3 or m2 per m3)
constant related to the relative rates of submerged arc welding

grain growth and particle dissolution in
the material (jam2"I/n) time (s or min)
grain boundary mobility (variable units) integration limits (s)
kinetic constant in expression for M cooling time from 800 to 5000C (s)
(variable units)
temperature (0C or K)
modified kinetic constant in expression
for M(J^m1711S-1) reference temperature (0C or K)
time exponent in grain growth equation ambient temperature (0C or K)
number of interacting particles per unit melting point (0C or K)

area of grain boundary (nr 2 )
peak temperature (0C or K)
number of particles per unit volume
(nr 3 ) voltage (V)
driving pressure for grain growth welding speed (mm s"1)

(J m-3)
grain boundary migration rate (jam s~{
effective driving pressure for grain or m s"1)
growth (J nr 3 )
grain boundary migration rate conform-
retardation pressure due to second ing to the low velocity limit (|Lim s 1 or
phase particles (J irr 3 ) ms- 1 )
net arc power (W) grain boundary migration rate conform-

ing to the high velocity limit (jam s~l or
apparent activation energy for grain ms" 1 )
growth (J mor 1 or kJ mor 1 )
x-axis/welding direction (mm)
activation energy for element diffusion
in dissolution model (J moT 1 or y-axis/transverse direction (mm)
kJ mor 1 )
z-axis/through thickness direction (mm) grain boundary interfacial energy
(J m-2)
dimensionless supersaturation in disso-
lution model calibration constant in coarsening
model (|Lim3 K s"1)
kinetic constant in expression for a
calibration constant in dissolution
parameter depending on the diffusivity model (|Lim2 s"1)
and the interaction energy between the
grain boundary and the impurity atoms density (g cm"3 or kg rrr3)
another complex function of the impu- radius of curvature of a spherical grain

rity diffusivity and the interaction en- (jum or m)
ergy between the grain boundary and
the impurity atoms volume heat capacity (J mm"3 0C"1)
intrinsic drag coefficient thermal conductivity (W mm"1 0C"1)
arc efficiency
6
Solid State Transformations in Welds
6.1 Introduction
The majority of phase transformations occurring in the solid state take place by thermally
activated atomic movements. In welding we are particularly interested in transformations that
are induced by a change in temperature of an alloy with a fixed bulk composition. Such
transformations include precipitation reactions, eutectoid transformations, and massive trans-
formations both in the weld metal and in the heat affected zone.
Since welding metallurgy is concerned with a number of different alloy systems (including
low and high alloy steels, aluminium alloys, titanium alloys etc.), it is not possible to cover all
aspects of transformation behaviour. Consequently, the aim of the present chapter is to pro-
vide the background material necessary for a verified quantitative understanding of phase
transformations in weldments in terms of models based on thermodynamics, kinetics, and
simple diffusion theory. These models will then be applied to specific alloy systems to illumi-
nate the basic physical principles that underline the experimental observations and to predict
behaviour under conditions which have not yet been studied.
6.2 Transformation Kinetics
In order to understand the extent and direction of a transformation reaction, it is essential to

know how far the reaction can go and how fast it will proceed. To answer the first question we
need to consider the thermodynamics, whereas kinetic theory provides information about the
reaction rate.
6.2.1 Driving force for transformation reactions

In practice, solid state transformations require a certain degree of undercooling, which is
essential to accommodate the surface and strain energies of the new phase.1 Generally, this
minimum molar free energy of transformation, AG, can be written as a balance between the
following four contributions:
(6-1)
Here AGy (the volume free energy change associated with the transformation) and AGD
(free energy donated to the system when the nucleation takes place heterogeneously) are nega-
tive, since they assist the transformation, while AGS (increase in surface energy between the
two phases) and AGE (increase in strain energy resulting from lattice distortion) are both posi-
tive because they represent a barrier against nucleation. It follows that the transformation
Molar free energy Stable (B) Stable (a)
Temperature
Fig. 6.1. Schematic representation of the molar free energies of two solid a and P phases as a function of
temperature (allotrophic transformation — no compositional change).
reaction can proceed when the driving force AG becomes greater than the right-hand side of
equation (6-1).
For an allotropic transformation, in which there is no compositional change, AG will be a
simple function of temperature, as illustrated in Fig.6.1. For alloys the situation is slightly
more complex, since there is an additional variable, i.e. the composition. In such cases the
temperature at which the a-phase becomes thermodynamically unstable (Teq) corresponds to a
fixed point on the a-(3 solvus boundary in the equilibrium phase diagram, as shown
schematically in Fig. 6.2. Since phase diagrams are available for many of the important indus-
trial alloy systems, it means that the driving force for a transformation reaction can readily be
obtained from such diagrams in the form of a characteristic undercooling (AT).
Temperature
%B
Fig. 6.2. Schematic representation of the a-(3 solvus boundary in a simple binary phase diagram.
6.2.2 Heterogeneous nucleation in solids
In general, solid state transformations in metals and alloys occur heterogeneously by nuclea-
tion at high energy sites such as grain corners, grain boundaries, inclusions, dislocations and
vacancy clusters. The potency of a nucleation site, in turn, depends on the energy barrier
against nucleation (AG*^) which is a function of the 'wetting' conditions at the substrate/
nucleus interface.1
It can be seen from Fig. 6.3 that nucleation at for instance inclusions or dislocations is
always energetically more favourable than homogeneous nucleation (AG*het < tsG*hom ) but
less favourable than nucleation at grain boundaries or free surfaces. As a result, the transfor-
mation behaviour is strongly influenced by the type and density of lattice defects and second
phase particles present within the parent material.
6.2.2.1 Rate of heterogeneous nucleation

Whereas every atom is a potential nucleation site during homogeneous nucleation, only those
associated with lattice defects or second phase particles can take part in heterogeneous nuclea-
tion. In the latter case the rate of nucleation (Nhet) is given by:1'2
(6-2)
where v is a vibration frequency factor, Nv is the total number of heterogeneous nucleation

sites per unit volume, AG^, is the energy barrier against nucleation, and Qd is the activation
energy for atomic migration across the nucleus/matrix interface.
Grain boundary
Free
surface Inclusion
horn.
Dislocations/stacking faults
AG* /AG*
Vacancy clusters
het.
Inclusions
boundaries
Grain
corners
Grain
surfaces
Free
Nucleation site
Fig. 6.3 Schematic diagram showing the most potent sites for heterogeneous nucleation in metals and
alloys.
It follows from the graphical representation of equation (6-2) in Fig. 6.4 that the nucleation
rate Nhet is highest at an intermediate temperature due to the competitive influence of
undercooling (driving force) and diffusivity on the reaction kinetics. This change in Nhet with
temperature gives rise to corresponding fluctuations in the transformation rate, as shown
schematically in Fig. 6.5. Note that the peak in transformation rate is due to two functions,
growth and nucleation (which peak at different T) whereas peak in Nhet is due to nucleation
only.
6.2.2.2 Determination of AGhet and Qd

During the early stages of a precipitation reaction, the reaction rate may be controlled by the
nucleation rate Nhet. Under such conditions, the time taken to precipitate a certain fraction of
the new phase t* is inversely proportional to Nhet\
(6-3)
where C1 and C2 are kinetic constants.

By taking the natural logarithm on both sides of equation (6-3), we obtain:
(6-4)
If the complete C-curve is known for a specific transformation reaction, it is possible to

evaluate AG*het and Qd from equation (6-4) according to the procedure described by Ryum.3
In general, a plot of In t* vs HT will yield a distorted C-curve with well-defined asymptotes, as
shown in Fig. 6.6. At high undercoolings, when &G*het is negligible, the slope of the curve
becomes constant and equal to QdIR. The mathematical expression for this asymptote is:
T T
Low undercooling
High diffusivity
High undercooling
Low diffusivity
%B Nhet.
Fig. 6.4. Schematic diagram showing the competitive influence of undercooling (driving force) and
diffusivity on the heterogeneous nucleation rate.
T
logt
Fraction transformed
logt
Fig. 6.5. Fraction transformed as a function of time referred to the C-curve (schematic).
(6-5)
At the chosen reference temperature Tr the time difference between the real C-curve and
the extension of the lower asymptote amounts to (see Fig. 6.6):
(6-6)
from which
(6-7)
It follows that equations (6-5) and (6-7) provide a systematic basis for obtaining quantita-
tive information about Qd and AGhet from experimental microstructure data through a simple
graphical analysis of the shape and position of the C-curve in temperature-time space.
1/r
T
lnt
Fig. 6.6. Determination of AG*het and Qd from the C-curve (schematic).
6.2.2.3 Mathematical description of the C-curve

In order to obtain a full mathematical description of the C-curve, we need to know the varia-
tion in the energy barrier AG*het with undercooling AT. For heterogeneous nucleation of pre-
cipitates above the metastable solvus, the strain energy term entering the expression for AG*het
can usually be ignored. In such cases the energy barrier is simply given as:1
(6-8)
where TV4 is the Avogadro constant, ^ n is the interfacial energy per unit area between the nu-
cleus and the matrix, AGV is the driving force for the precipitation reaction (i.e. the volume free
energy change associated with the transformation), and 5(0) is the so-called shape factor which
takes into account the wetting conditions at the nucleus/substrate interface.
For a particular alloy, AGV is for small Ar proportional to the degree of undercooling:l
(6-9)
where C3 is a kinetic constant. This equation follows from the definition of AGv in diluted alloy
systems and the mathematical expression for the solvus boundary in the binary phase dia-
gram.
By substituting equation (6-9) into equation (6-8), we get:
(6-10)
It follows that A0 is a characteristic material constant which is related to the potency of the
heterogeneous nucleation sites in the material. The value of A0 is, in turn, given by equations
(6-7) and (6-10):
(6-11)
In cases where A0 is known, it is possible to obtain a more general expression for t* by
substituting equation (6-10) into equation (6-3):
(6-12)
Equation (6-12) can further be modified to allow for compositional and structural varia-
tions in the parent material by using the calibration procedure outlined in Fig. 6.7. Let tr
denote the time taken to precipitate a certain fraction of (3 at a chosen reference temperature
T= Tr in an alloy containing Nv nucleation sites per unit volume. If we take the corresponding
solvus temperature of the (3-phase equal to T*q , the expression for t* becomes:
(6-13)
A combination of equations (6-12) and (6-13) then yields:
(6-14)
Equation (6-14) provides a basis for predicting the displacement of the C-curve in tempera-
ture-time space due to compositional or structural variations in the parent material. In gen-
T
C-cun/e(Nv=N^)
logt
Fig. 6.7. Method for eliminating unknown kinetic constant in expression for t*.
eral, an increase in Nv will shift the nose of the C-curve to the left in the diagram (i.e. towards
shorter times), as shown schematically in Fig. 6.8, because of the resulting increase in the
nucleation rate. Moreover, in solute-depleted alloys the critical undercooling for nucleation
will be reached at lower absolute temperatures where the diffusion is slower. This results in a
marked drop in Nhet, which displaces the C-curve towards lower temperatures and longer times
in the IT-diagram, as indicated in Fig. 6.9.
Example (6.1)
Isothermal transformation (IT) or continuous cooling transformation (CCT) diagrams are avail-
able for many of the important alloy systems.4 In the case of aluminium, so-called tempera-
ture-property diagrams exist for different types of wrought alloys.45 Suppose that the C-curve
in Fig. 6.10 conforms to incipient precipitation of [3'(Mg2Si) particles at manganese-contain-
ing dispersoids in 6351 extrusions. Use this information to estimate the values of A0 and Qd in
equation (6-3) when the solvus temperature of (3'(Mg2Si) is 5200C.
Solution
The parameters A0 and Qd can be evaluated from the C-curve according to the procedure
shown in Fig. 6.6. Referring to Fig. 6.11, the value of AG^ at the chosen reference tempera-
ture Tr = 35O°C (623K) is equal to:
When AGhet is known, the parameter A0 can be obtained from equation (6-11):
iogt
Fig. 6.8. Effect of Nv on the shape and position of C-curve in temperature-time space (schematic).
T T T
%B \e, logt
Fig. 6.9. Effect of solute content on the shape and position of C-curve in temperature-time space (sche-
matic).
Similarly, Qd can be read from Fig. 6.11 by considering the slope of the lower asymptote:
This value is in good agreement with the reported activation energy for diffusion of magne-
sium in aluminium.6
AA 6351 - T6
Temperature, 0C
Time, s
Fig. 6.10. C-curve for 99.5% maximum yield strength of an AA6351-T6 extrusion. After Staley.5
Solvus temperature: 520 0C
103/T, K"1
lnt
Fig. 6.11. Determination of kG*heU and Qd from the C-curve in Fig. 6.10 (Example 6.1).
6.2.3 Growth of precipitates

If the embryo is larger than some critical size, it will grow by a transport mechanism which
involves diffusion of solute atoms from the bulk phase to the matrix/nucleus interface.
6.2.3.1 Interface-controlled growth

When transfer of atoms across the a/(3-interface becomes the rate-controlling step, the reac-
tion is said to be interface-controlled. This growth mode is therefore observed during the
initial stage of a precipitation reaction before a large, solute-depleted zone has formed around
the particles. In the case of incoherent precipitates, the variation in the particle radius r with
time is given by:7
(6-15)
where M1 is a mobility term, C0 is the concentration of solute in matrix, Ca is the concentration

of solute at the particle/matrix interface, and Cp is the concentration of solute inside the pre-
cipitate.
In general, the mobility of incoherent interfaces is high, since the solute atoms can easily
'jump' across the interface and find a new position in the particle lattice, as shown schematically
in Fig. 6.12(a). In contrast, a coherent interface is essentially inmobile because transfer in this
case involves trapping of atoms in an intermediate lattice position, as indicated in Fig. 6.12(b).
As a result, semi-coherent precipitates are forced to grow by lateral movement of ledges along
a low energy interface in a direction where the matrix is incoherent with respect to the particle
lattice (see Fig. 6.13). In such cases the thickening rate of the precipitates U*aj^ is given by:3'7
Incoherent Interface
Coherent interface
(a)
(b)
Fig. 6J2. Schematic illustration of atom transfer across different kinds of interfaces; (a) Incoherent
interface, (b) Coherent interface.
U
a/p
Lateral move-
ment of incoherenf
interface
Fig. 6.13. Thickening of plate-like precipitates by the ledge mechanism (schematic).
(6-16)
where M1* is a new mobility term, and / is the interledge spacing.
6.2.3.2 Diffusion-controlled growth

For growth of incoherent precipitates above the metastable solvus, the rate-controlling step
will be diffusion of solute in the matrix. If precipitation of the P-phase occurs from a
supersaturated a, the reaction proceeds by diffusion of solute to the growing p-particle, as
shown schematically in Fig. 6.14. On the other hand, when the (3-phase is formed by rejection
of solute from the a-phase, the transformation occurs by diffusion of atoms away from the P-
particle, as indicated in Fig. 6.15.
Aron et al.s have presented general solutions for diffusion-controlled growth of both flat
plates and spheres under such conditions. In the former case the half thickness AZ of the plate
is given by:
(6-17)
The parameter E1 in equation (6-17) is frequently referred to as the one-dimensional para-

bolic thickening constant, and is defined as:
Liquid
Temperature
%B
Concentration
Diffusion of solute
Distance
Fig. 6.14. Schematic representation of concentration profile ahead of advancing interface during precipi-
tation of (B from a supersaturated a-phase.
Temperature
%B
Concentration
Diffusion of solute
Distance
Fig. 6.15. Schematic representation of concentration profile ahead of advancing interface during growth
of solute-depleted P into a metastable a-phase.
(6-18)
where Dm is the diffusivity of the solute in the matrix, and erfc(u) is the complementary error
function (defined previously in Appendix 1.3, Chapter 1).
Similarly, for growth of spherical precipitates, the variation in the radius r with time can be
written as:8
(6-19)
where e 2 *s t n e corresponding parabolic thickening constant for a spherical geometry, defined

as:
(6-20)
The parabolic relations in equations (6-17) and (6-19) imply that the growth rate slows
down as the (3-phase grows. This is due to the fact that the total amount of solute partitioned
during growth decreases with time when the diffusion distance increases. Moreover, the form
of equations (6-18) and (6-20) suggests that the maximum in the growth rate is achieved at an
intermediate temperature because of the competitive influence of undercooling (driving force)
and diffusivity on the reaction kinetics. Consequently, a plot of E1 or £2 vs temperature will
reveal a pattern similar to that shown in Fig. 6.4 for the nucleation rate, although the thicken-
ing constants generally are less temperature-sensitive.
In addition to the models presented above for plates and spheres, approximate solutions
also exist in the literature for thickening of needle-shaped precipitates, based on the Trivedi
theory for diffusion-controlled growth of parabolic cylinders.9 However, because of space limi-
tations, these solutions will not be considered here.
6.2.4 Overall transformation kinetics

The progress of an isothermal phase transformation may be conveniently represented by an IT-
diagram of the type shown in Fig. 6.5. Among the factors that determine the shape and posi-
tion of the C-curve are the nucleation rate, the growth rate, the density and the distribution of
the nucleation sites as well as the physical impingement of adjacent transformed volumes.
Due to the lack of adequate kinetic models for diffusion-controlled precipitation, we shall
assume that the overall microstructural evolution with time can be described by an Avrami-
type of equation:10
(6-21)
where X is the fraction transformed, n is a time exponent, and k is a kinetic constant which
depends on the nucleation and growth rates.
The exponential growth law summarised in the Avrami equation is valid for linear growth
under most circumstances, and approximately valid for the early stages of diffusion-controlled
growth.10 Table 6.1 gives information about the value of the time exponent for different ex-
perimental conditions.
In general, the value of n will not be constant, but change due to transient effects until the
steady-state nucleation rate is reached and n attains its maximum value. Subsequently, the
nucleation rate starts to decrease as the sites become filled with nuclei and eventually ap-
proach zero when complete saturation occurs. This is because the heterogeneous nucleation
sites are not randomly distributed in the volume, but are concentrated near other nucleation
sites leading to an overall reduction in n. From then on, the transformation rate is solely
controlled by the growth rate.
6.2.4.1 Constant nucleation and growth rates

For a specific transformation reaction, the value of k in equation (6-21) can be estimated from
Table 6.1 Values of the time exponent n in the Avrami equation. After Christian.10
Polymorphic changes, discontinuous precipitation, eutectoid reactions, interface

controlled growth, etc.
Increasing nucleation rate
Constant nucleation rate
Decreasing nucleation rate
Zero nucleation rate (saturation of point sites)
Grain edge nucleation after saturation
Grain boundary nucleation after saturation
Diffusion controlled growth
All shapes growing from small dimensions, increasing nucleation rate
All shapes growing from small dimensions, constant nucleation rate
All shapes growing from small dimensions, decreasing nucleation rate
All shapes growing from small dimensions, zero nucleation rate
Growth of particles of appreciable initial volume
Needles and plates of finite long dimensions, small in comparison with their separation
Thickening of long cylinders (needles) (e.g. after complete end impingement)
Thickening of very large plates (e.g. after complete edge impingement)
Precipitation on dislocations (very early stages)
kinetic theory, using the classic models of nucleation and growth described in the previous
sections. In practice, however, this is a rather cumbersome method, particularly if the base
metal is of a heterogeneous chemical nature. Alternatively, we can calibrate the Avrami equa-
tion against experimental microstructure data, e.g. obtained from generic IT-diagrams. A con-
venient basis for such a calibration is to write equation (6-21) in a more general form:
(6-22)
where k* is a new kinetic constant (equal to kr1/n). In the latter equation the parameter k* can be
regarded as a time constant, which is characteristic of the system under consideration. Note
that this form of the Avrami equation is mathematically more appropriate, as the dimensions of
the k* constant are not influenced by the value of the time exponent n.
During the early stages of a transformation reaction, the reaction rate is controlled by the
nucleation rate. Let f denote the time taken to precipitate a certain fraction of P (X = Xc) at
an arbitrary temperature T (previously defined in equation (6-14)). It follows from equation
(6-22) that the value of k* in this case is given as:
(6-23)
A combination of equations (6-22) and (6-23) then gives:
(6-24)
from which
(6-25)
Equation (6-25) represents an alternative mathematical description of the Avrami equation,

and is valid as long as the nucleation and growth rates do not change during the transforma-
tion. It has therefore the following limiting values and characteristics: X=O when t = 0, X =
Xc when t = t*, and X—>1 when r—> <*>.
6.2.4.2 Site saturation

If the nucleation rate is considered to be zero by assuming early site saturation, the subsequent
phase transformation simply involves the reconstructive thickening of the p-layer. In the one-
dimensional case, the process can be modelled in terms of the normal migration of a planar
a/p interface, as shown schematically in Fig. 6.16. Let Aa/^ denote the interfacial area be-
tween a and (3 per unit volume and Ua/^ the growth rate of the incoherent a/p-interface. From
Fig. 6.16 we see that the volume fraction of the transformed (3-phase is given as:
(6-26)
By using the standard Johnson-Mehl correction for physical impingement of adjacent trans-
formation volumes, we may write in the general case:
(6-27)
which after integration yields:

(6-28)
This specific form of the Avrami equation is valid under conditions of early site saturation
where the a/p-interface is completely covered by P nuclei at the onset of the transformation.
6.2.5 Non-isothermal transformations

So far, we have assumed that the phase transformations occur isothermally. This is, of course,
a rather unrealistic assumption in the case of welding where the temperature varies continu-
ously with time. From the large volume of literature dealing with solid state transformations in
Fig. 6.16, Schematic illustration of the planar geometry assumed in the site saturation model.
metals and alloys, it appears that the bulk of the research has been concentrated on modelling
of microstructural changes under predominantly isothermal conditions.1"411 In contrast, only
a limited number of investigations has been directed towards non-isothermal transforma-
tions. 51012 " 18 However, these studies have clearly demonstrated the advantage of using ana-
lytical modelling techniques to describe the microstructural evolution during continuous cool-
ing, instead of relying solely on empirical CCT-diagrams.
6.2.5.1 The principles of additivity

From the literature reviewed it appears that there is considerable confusion regarding the ap-
plication of isothermal transformation theory for prediction of non-isothermal transformation
behaviour. These difficulties are mainly due to the independent variations of the nucleation
and growth rate with temperature. In fact, it can be shown on theoretical grounds that the
problem is only tractable when the instantaneous transformation rate is a unique function of
the fraction transformed and the temperature.10 This leads to the additivity criterion described
below.
The principles of additivity are based on the theory advanced by Scheil.12 He proposed that
the start of a transformation under non-isothermal conditions could be predicted by calculating
the consumption of fractional incubation time at each isothermal temperature, with the trans-
formation starting when the sum is equal to unity. The Scheil theory has later been extended to
phase transformations to predict continuous cooling transformation kinetics from isothermal
microstructure data.10'17'18
Let t* again denote the time taken to precipitate a certain fraction of P (X - Xc) at an arbi-
trary temperature T. If the reaction is additive, the total time to reach Xc under continuous
cooling conditions is obtained by adding the fractions of time to reach this stage isothermally
until the sum is equal to unity. Noting that t* varies with temperature, we may write in the
general case:
(6-29)
A schematic illustration of the Scheil theory is contained in Fig. 6.17.

T
Cooling curve
Subdivision of time into

infinitesimal steps of iso-
thermal heat treatments.
logt
Fig. 6.17. Schematic illustration of the Scheil theory.
6.2.5.2 Isokinetic reactions
The concept of an isokinetic reaction has previously been introduced in Section 4.4.2.3 (Chap-
ter 4). A reaction is said to be isokinetic if the increments of transformation in infinitesimal
isothermal time steps are additive, according to equation (6-29). Christian10 defines this math-
ematically by stating that a reaction is isokinetic if the evolution equation for some state vari-
able X may be written in the form:
(6-30)
where G(X) and H(T) are arbitrary functions of X and T, respectively.

For a given thermal history, T(t), this essentially means that the differential equation con-
tains separable variables of X and T.
6.2.5.3 Additivity in relation to the Avrami equation

The concept of an isokinetic reaction can readily be applied to the Avrami equation. Differen-
tiation of equation (6-22) with respect to time leads to the following expression for the rate of
transformation:
(6-31)
In a typical diffusion-controlled nucleation and growth process, the fraction transformed X

will not be independent of temperature, since the equilibrium volume fraction of the precipi-
tates decreases with temperature (e.g. see equation (4-7) in Chapter 4). However, for dilute
alloys it is a fair approximation to neglect this variation as the solvus boundary becomes in-
creasingly steeper and in the limiting case approaches that of a straight (vertical) line. Thus, if
n is constant and k* depends only on the transformation temperature, the reaction will be iso-
kinetic in the general sense defined by Christian.10
Because of the independent variations of the nucleation and growth rate with temperature,
the transformation rate will not be a simple function of temperature. However, by considering
the form of the constitutive equations, it is obvious that the change in the nucleation rate with
temperature is far more significant than the corresponding fluctuations in the growth rate.
This point is more clearly illustrated in Fig. 6.18, which shows the temperature-dependency of
the nucleation and growth rates of grain boundary ferrite in a C-Mn steel. It is evident from
these data that the change in the parabolic thickening constant £, is negligible compared with
the fluctuations in the nucleation rate. Consequently, in transformations that involve continu-
ous cooling it is sufficient to allow for the variation of Nhet with temperature, provided that
site saturation has not been reached. Thus, if n is constant we can apply the Scheil theory
directly and rewrite equation (6-25) in an integral form:
(6-32)
In equation (6-32) Z1 represents the kinetic strength of the thermal cycle with respect to P-
precipitation. This parameter is generally defined by the integral:
-2 -1
Thickening constant Ce1),|ims"

Nucleation rate (N*het), cm s
Temperature, 0C
Fig. 6.18. Predicted variation in N*het and E1 with temperature during the austenite to ferrite transfor-
mation in a C-Mn steel (0.15 wt% C, 0.40 wt% Mn). Data from Umemoto et al.19
(6-33)
where dt is the time increment at T, and f is the corresponding hold time required to reach Xc
at the same temperature (given by equation (6-14)). The derivation of equation (6-32) is shown
in Appendix 6.2
The principles of additivity are also applicable under conditions of early site saturation. If
only U^p varies with temperature, it is possible to rewrite equation (6-28) in an integral form:
(6-34)
This equation can readily be integrated by numerical methods when the temperature-time
programme is known.
6.2.5.4 Non-additive reactions

If the additivity condition is not satisfied, it means that the fraction transformed is dependent
on the thermal path, and the differential equation has no general solution. This, in turn, implies
that the C-curve concept breaks down and cannot be applied to non-isothermal transforma-
Next Page
tions. Solution of the differential equation then requires stepwise integration in tempera-
ture-time space, using an appropriate numerical integration procedure. As already pointed
out, this will generally be the case for diffusion-controlled precipitation reactions, since the
evolution parameter X is a true function of temperature. Under such conditions, experimen-
tally based continuous cooling transformation (CCT) diagrams must be employed.
6.3 High Strength Low-Alloy Steels

High-strength low-alloy steels are typically produced with a minimum yield strength in the
range 300-500 MPa, depending on the plate thickness.2021 During welding microstructural
changes take place both within the heat affected zone (HAZ) and the fusion region, which, in
turn, affect the mechanical integrity of the weldment.2122 In the HAZ, for instance, nitrides
and carbides coarsen and dissolve, and grain growth occurs to an extent that depends on the
distance from the fusion boundary and the exposure time characteristic of the welding process.
This can have a profound effect on the subsequent structure and properties of the weld by
displacing the CCT curve to longer times, thereby producing more Widmanstatten ferrite, or
increasing the possibility of bainite and martensitic transformation products on cooling. The
formation of such microstructures may reduce the toughness of the weld and increase the risk
of hydrogen cracking.2122
6.3.1 Classification of microstructures

It is appropriate to start this section with a detailed classification of the various microstructural
constituents commonly found in low-alloy steel weldments.
During the austenite to ferrite transformation, a large variety of microstructures can de-
velop, depending on the cooling rate and the steel chemical composition. Normally, the micro-
structure formed within each single austenite grain after transformation will be a complex
mixture of two or more of the following constituents, arranged in approximately decreasing
order of transformation temperature:
(i) grain boundary (or allotriomorphic) ferrite (GF)

(ii) polygonal (or equiaxed) ferrite (PF)
(iii) Widmanstatten ferrite (WF)
(iv) acicular ferrite (AF)
(v) upper bainite (UB)
(vi) lower bainite (LB)
(vii) martensite (M).
The microstructural constituents listed above are indicated in Fig. 6.19, which shows
photomicrographs of typical regions within low-alloy steel weldments.
6.3.2 Currently used nomenclature

Quantification of microstructures in steel welds is most commonly done by means of optical
microscopy. Several systems have been introduced throughout the years for the classification
of the various constituents, with each system reflecting different investigator's views and
Previous Page
tions. Solution of the differential equation then requires stepwise integration in tempera-
ture-time space, using an appropriate numerical integration procedure. As already pointed
out, this will generally be the case for diffusion-controlled precipitation reactions, since the
evolution parameter X is a true function of temperature. Under such conditions, experimen-
tally based continuous cooling transformation (CCT) diagrams must be employed.
6.3 High Strength Low-Alloy Steels

High-strength low-alloy steels are typically produced with a minimum yield strength in the
range 300-500 MPa, depending on the plate thickness.2021 During welding microstructural
changes take place both within the heat affected zone (HAZ) and the fusion region, which, in
turn, affect the mechanical integrity of the weldment.2122 In the HAZ, for instance, nitrides
and carbides coarsen and dissolve, and grain growth occurs to an extent that depends on the
distance from the fusion boundary and the exposure time characteristic of the welding process.
This can have a profound effect on the subsequent structure and properties of the weld by
displacing the CCT curve to longer times, thereby producing more Widmanstatten ferrite, or
increasing the possibility of bainite and martensitic transformation products on cooling. The
formation of such microstructures may reduce the toughness of the weld and increase the risk
of hydrogen cracking.2122
6.3.1 Classification of microstructures

It is appropriate to start this section with a detailed classification of the various microstructural
constituents commonly found in low-alloy steel weldments.
During the austenite to ferrite transformation, a large variety of microstructures can de-
velop, depending on the cooling rate and the steel chemical composition. Normally, the micro-
structure formed within each single austenite grain after transformation will be a complex
mixture of two or more of the following constituents, arranged in approximately decreasing
order of transformation temperature:
(i) grain boundary (or allotriomorphic) ferrite (GF)

(ii) polygonal (or equiaxed) ferrite (PF)
(iii) Widmanstatten ferrite (WF)
(iv) acicular ferrite (AF)
(v) upper bainite (UB)
(vi) lower bainite (LB)
(vii) martensite (M).
The microstructural constituents listed above are indicated in Fig. 6.19, which shows
photomicrographs of typical regions within low-alloy steel weldments.
6.3.2 Currently used nomenclature

Quantification of microstructures in steel welds is most commonly done by means of optical
microscopy. Several systems have been introduced throughout the years for the classification
of the various constituents, with each system reflecting different investigator's views and
(a) (b)
(C) (d)
Fig. 6.19. Optical micrographs showing various microstructural constituents commonly found in low-
alloy steel weldments; (a) Coarse grained HAZ (low heat input welding), (b) Coarse grained HAZ (high
heat input welding), (c) As-deposited weld metal (low heat input welding), (d) Reheated weld metal (low
heat input welding). Letters in micrographs are defined in the text.
discretions. This controversy in terminology has been a source of confusion, and the work by
Sub-Commission IXJ of the International Institute of Welding (HW)23 for developing guide-
lines for quantification of microstructures is, therefore, an important step towards a standard-
ised system of nomenclature.
The IIW recommendations are based on the scheme originally proposed by Abson and
Dolby.23 The IIW system involves a simplified classification procedure compared with the
outline used in Fig. 6.19, since the distinction between acicular ferrite and the various sideplate
structures is based on features such as aspect ratio, relative lath size, and number of parallel
laths. This has led to the introduction of the FS-constituent (ferrite with aligned second phase),
which, in principle, comprises both Widmanstatten ferrite and upper bainite.
In contrast to the HW approach to classifying microstructural elements based on their ap-
pearance in the optical microscope, other investigators rank the various constituents solely in
terms of their transformation behaviour, according to the scheme originally proposed by Dube
et ai26 From a scientific point of view, this classification system is more correct, since it does
not violate common terminology based on thermodynamics and kinetics of transformation
reactions. However, with the omission of the FS-constituent grouping utilised by the HW, the
Dube system is more inconvenient to use in practice because the different transformation prod-
ucts often cannot readily be identified on the basis of their transformation characteristics. Con-
sequently, both classification systems appear to have their weaknesses, which, in turn, limit
their applicability.
6.3.3 Grain boundary ferrite

Grain boundary (or allotriomorphic) ferrite is the first phase to form on cooling below the Ae3-
temperature. It nucleates preferentially at austenite grain corners and boundaries, since these
sites generally provide the lowest energy barrier against nucleation (see Fig. 6.3). The funda-
mental aspects of grain boundary ferrite have been reviewed in detail by Bhadeshia,26 where
many of the original references can also be found.
6.3.3.1 Crystallography of grain boundary ferrite

The grain boundary ferrite allotriomorphs nucleate having a Kurdjumow-Sachs type orienta-
tion relationship with one of the austenite grains:20
This orientation relationship, which lies within the so-called Bain orientation region,27 is
adopted in order to minimise the increase in the strain energy resulting from lattice distortion
AGEby formation of a low-energy interface between the ferrite nucleus and the parent austenite
phase.1 Subsequent growth of the ferrite may then occur into the adjacent austenite grain with
which the ferrite has a random orientation relationship,28 since a disordered (incoherent) inter-
face generally has a higher mobility than an ordered (coherent/semi-coherent) interface at low
undercoolings.
6.3.3.2 Nucleation of grain boundary ferrite

As shown in Fig. 6.20, allotriomorphic ferrite can nucleate both at grain corners, grain edges
and grain faces, the former ones being the most potent sites for ferrite nucleation.1 If we
assume that ferrite nucleation occurs preferentially at austenite grain faces, the A^_/ Nv term
in equation (6-14) may be replaced by the inverse grain size ratio Dy /Dy, where Dy and Dy
refer to the austenite grain size in the actual and the reference material, respectively. This
leads to the following expression for t*:
(6-35)
Edge
Corner
Fig. 6,20. Sketch of an austenite grain showing different sites for ferrite nucleation.
Based on equation (6-35) it is possible to predict the displacement of the C-curve in tem-
perature-time space due to structural or compositional variations in the parent material. As an
illustration, we shall assume that the parameters listed in Table 6.2 are representative of nu-
cleation of grain boundary ferrite in a low-alloy steel with an initial austenite grain size of 10
jim. In addition, we need information about the Ae3-temperature in the equilibrium phase
diagram. This temperature can readily be obtained from thermodynamic calculations, even for
multicomponent systems.29 Alternatively, we can use the empirical relationship quoted by
Leslie:30
(6-36)
where all compositions are given in weight %.

By lowering the A<?3-temperature, the austenite stabilising elements C, Mn, Ni and Cu re-
duce the undercooling and hence, the driving force for transformation of austenite, at any
lower temperature. On the same basis, the ferrite stabilising elements will assist the transfor-
mation because they increase the undercooling. In particular, the very large ferrite stabilising
effects of P, Al and Ti are obvious from the above relationship.
Effect of austenite grain size

A closer inspection of equation (6-35) reveals that the position of the C-curve shifts by a factor
of (Dy I Dy) when the austenite grain size changes from D1 to Dy. This is due to the fact
that the austenite grain boundary area per unit volume is inversely proportional to the grain
size. Consequently, grain growth which occurs during welding will have a profound effect on
the subsequent transformation behaviour of the weld. An illustration of this point is given
below.
Table 6.2 Input data used to construct C-curve for allotriomorphic ferrite in reference steel.
Parameter I T I T I T*eq I T0 I Q] I Dy
(K) (s) (K) (Jmol"1) (kJmol-1) (jim)
Value 823 0.6 1108 700 135 10
f Activation energy for diffusion of carbon in austenite.

Example (6.2)
Consider SA welding on a thick plate of a Nb-microalloyed steel under the following condi-
tions:
Evaluate on the basis of the grain growth diagram in Fig. 5.21 (a) (Chapter 5) and equations
(6-32), (6-33), and (6-35) the conditions for ferrite formation at two different positions within
the HAZ corresponding to a peak temperature Tpof 13500C and 10000C, respectively. As-
sume in these calculations that the ferrite may form within the temperature range from 800 to
6000C, and that the equilibrium volume fraction of ferrite (f^qFe) in the fully transformed
steel is 0.9.
Solution
Readings from Fig. 5.21 (a) give the following HAZ grain sizes:
By substituting data from Table 6.2 into equation (6-35), we obtain:
Grain refined region ( Dy - 1 OfJm):
Grain coarsened region ( Dy = 150 jlm):
As expected, the theoretical C-curves in Fig. 6.21 reveal a strong effect of the austenite
grain size on the HAZ transformation kinetics. By considering the superimposed weld cool-
ing curve, it is possible to estimate the volume fraction of ferrite/""7^ which forms in each case
from equations (6-32) and (6-33). Taking the time exponent n in the Avrami equation equal
to 5/2 for nucleation of ferrite at austenite grain boundaries10 and 1 — Xc = 0.98, we obtain:
Grain refined region (Dy = 10 jlm):

Temperature, 0C
Time, s
Fig. 6.21. Effect of austenite grain size on the HAZ transformation kinetics (Example 6.2). The superim-
posed cooling curve corresponds to a cooling time, A%5, of 21s.
and
Grain coarsened region ( Dy = 150 jim):
and
From this we see that polygonal ferrite dominates the microstructure within the grain re-
fined region, whereas ferrite hardly forms within the grain coarsened HAZ under the prevail-
ing circumstances. Although experimental data are not available for a direct comparison, the
predicted effect of the austenite grain size on the HAZ transformation kinetics is reasonable
and consistent with general experience (e.g. see experimental CCT-diagrams in Fig. 6.22).
Effect of austenite and ferrite stabilising elements

In practice, the transformation behaviour of steel weldments is affected both by the prior
austenite grain size and by alloying additions. In welding metallurgy, the combined effects are
often discussed by considering their influence on hardenability. Broadly speaking, additions of
hardenability elements may serve two purposes:
(i) To ensure the desired strength level by solid solution or precipitation strengthening;
(ii) To control the microstructure through modification of the nucleation and growth rates of
proeutectoid ferrite.
Temperature, 0C
Time, s
Fig. 6.22. CCT-diagrams for a low-carbon Cu-Ni containing steel. Superimposed on the CCT-diagrams
are two cooling curves corresponding to Af875 of 10 and 100 s, respectively. Austenitising conditions;
Heavy solid lines: 9000C for 5 min, Heavy broken lines: 13000C for 5 s. Data from Cross et al?x
In the latter case the effect is related to a shift in the Ae3-temperature of the steel, which
alters the undercooling and hence, the driving force for the austenite to ferrite transformation.
This point is illustrated by the following example.
Example (6.3)
Consider SA welding on a thick plate of a Nb-microalloyed steel under conditions similar to
those employed in Example (6.2). Based on equations (6-32), (6-33), and (6-35) estimate the
volume fraction of grain boundary ferrite in the grain coarsened HAZ (Tp = 13500C) after
welding when the A^-temperature of the steel is 863°C (1136K).
Solution
By substituting data from Table 6.2 into equation (6-35), we arrive at the following expression
for r *50:
It is evident from the graphical representation of the above equation in Fig. 6.23 that an in-
crease in the v4e3-temperature (e.g. from 1108 to 1136K) displaces the C-curve towards higher
temperatures and shorter times in the IT-diagram. This, in turn, gives rise to improved con-
ditions for ferrite nucleation. Taking the time exponent n in the Avrami equation equal to 5/2
and (1 — Xc) = 0.98 as in the previous example, we obtain after integration:
Temperature, 0C
Time, s
Fig. 6.23.Effect of steel chemical composition (Ae3-temperature) on the HAZ transformation kinetics
(Example 6.3). The superimposed cooling curve corresponds to a cooling time, Ar875, of 21s.
and
Although the calculated value offa'Fe is rather uncertain, the trends predicted in the present
example are reasonable and consistent with general experience.
Effect of boron alloying

The role of boron in steel transformation kinetics has been a subject of research for many
years. A number of hardenability mechanisms have been proposed to explain the behaviour of
boron in steels. Of these, only four have survived to the present.32 All assume that boron
influences hardenability by increasing the energy barrier against ferrite nucleation at austenite
grain boundaries* and that it does not influence the thermodynamic properties of the austenite
and ferrite phases, i.e. reduces the A<?5-temperature of the steel.32 Moreover, it is now gener-
ally accepted that impurity-boron interactions are responsible for most of the observed erratic
behaviour of boron steels with respect to hardenability.32"34 In particular, if the steel contains
small amounts of dissolved oxygen and nitrogen, these elements may combine with boron to
reduce the free, diffusible boron content.32 The hardenability effect of boron is therefore sen-
sitive to variations in the applied steel deoxidation and alloying practice.32"37
*Based on classic nucleation theory it can be argued that a decrease in the y-y grain boundary energy due to boron
segregations will suppress the formation of ferrite at these sites. Further enhancement of the energy barrier against
ferrite nucleation through an increase in the total strain energy of the embryo is possible if the free grain boundary
volume becomes filled with either boron atoms or borocarbide precipitates (e.g. Fe23(B9C)6).
During welding, quantitative information about the extent of boron segregation which oc-
curs to the austenite grain boundaries under various thermal programmes can be obtained on
the basis of a well established theoretical model for quench-induced segregation of boron in
steel.3839 At peak temperatures above 1000 to 11000C, borocarbides and -nitrides present in
the base plate will rapidly dissolve in the matrix,4041 leading to a significant increase in the
amount of free diffusible boron. Generally, solute atoms in a crystal lattice will have an as-
sociated strain energy,38 which implies that it is energetically feasible to pair the solute boron
atom with a vacancy. Since the formation of vacancies is a thermally activated process, it fol-
lows that the fraction of boron occupying such sites, [B]v, increases exponentially with tem-
perature:39
(6-37)
Here m contains various geometric and entropy terms, Zy is the vacancy formation energy,
Eb is the vacancy-boron binding energy, and [B] is the bulk concentration of free boron. By
substituting reasonable average values for ra, Ep and Eb into equation (6-37), we arrive at the
following expression for [B]V:39A2
(6-38)
During cooling, the vacancy concentration initially established at elevated temperatures

tends to readjust by elimination of excess point defects at grain boundaries through diffusion.
This, in turn, results in an associated flow of solute boron to the austenite grain boundaries (the
extent of which is controlled by the peak temperature and the bulk boron concentration), pro-
vided that the diffusivity of the boron-vacancy complexes is higher than that of the vacancies
and the boron atoms at all relevant temperatures.
Consider now the limiting case where the boron-vacancy complex diffusion to the grain
boundaries occurs sufficiently rapid to keep pace with the falling temperature; i.e. the equilib-
rium concentration, [B]v, is maintained from the peak temperature Tp down to the start tem-
perature of the austenite to ferrite transformation Ae3. In view of the high diffusivity of boron-
vacancy complexes in austenite, this is not an unrealistic assumption when the cooling rate is
of the order of 500C s"1 (representative of a weld HAZ).39 Under such conditions, the amount
of boron which segregates to the austenite grain boundaries within the temperature interval
from Tp to Ae3 on cooling, [B]gb, can approximately be written as:42
when Tp » Ae3.
In addition to the diffusible boron content, the grain size is also an important variable in
steel hardenability. A quantitative estimate of the combined effect of boron segregations and
austenite grain size on the HAZ transformation kinetics can be obtained by assuming that the
ferrite nucleates primarily on grain faces. In such cases the total number of heterogeneous
nucleation sites per unit volume Nvis given as:
(6-40)
where na is the number of nucleation sites per unit grain boundary area, and Sv is the grain
boundary surface area per unit volume (equal to 2/Dy).
If we, as a second approximation, assume that na is inversely proportional to the amount of
boron which diffuses to the grain boundaries on cooling, the N* / Nv ratio in equation (6-14)
can be written as:
(6-41)
Equation (6-41) predicts that the position of the ferrite C-curve in temperature-time space
depends on the l[B]gb Z)7) I l[B]*gb Z>y J ratio, as shown schematically in Fig. 6.24. Consequently,
this ratio can be regarded as a measure of the HAZ hardenability during welding of boron-
containing steels.
In Fig. 6.25 the microstructure data of Akselsen et al.42 have been replotted vs the
hardenability parameter [B]gbDy, taking the product [B]*gbDy in the reference steel equal to
unity for a direct comparison between theory and experiments. It is evident from the graph
that the HAZ martensite content of the two boron-containing steels can be represented by one
single curve under the prevailing circumstances. This result is to be expected if the displace-
ment of the ferrite C-curve in temperature-time space is determined by a relationship of the
type shown in equation (6-41).
High values of [B] D

Temperature
Low values of [Bl D

gb Y Cooling curve
log time
Fig. 6.24. Effect of boron alloying on the shape and position of ferrite C-curve in temperature-time space
according to the site blocking mechanism (schematic).
Martensite, vol%
Steel A (11 ppm B)

Steel B (26 ppm B)
Hardenability parameter ([B] D )
Fig. 6.25. Relation between the volume fraction of martensite and the hardenability parameter
[B]gb Dy for simulated weld HAZs ([B]gb in ppm, Dy in urn, [B]gbDy = 1). Data from Akselsene£#/.42
Example (6.4)
Consider GMA welding on a thick plate of a boron-containing steel under the following con-
ditions:
/=300A, U = 30V, v = 4mm s-1, TI = 0.8, T0 = 200C
Suppose that the free (diffusible) boron content of the steel at elevated temperatures is 40
ppm. Estimate on the basis of the theory outlined in the previous sections the conditions for
ferrite/martensite formation in the grain refined HAZ (Tp = 11000C) when the austenite grain
size is 15|Lim. In these calculations we shall assume that the [B]*gb Dy product in the reference
steel (with thermodynamic properties as in Table 6.2) is close to unity.
Solution
First we estimate the amount of boron which segregates to the austenite grain boundaries
during cooling from equation (6-39). When Tp = 11000C (1373K), we get:
This gives the following value of the Af* / Nv ratio (equation 6-41):
The resulting displacement of the ferrite C-curve can now be calculated from equation (6-
14), using input data from Table 6.2:
As expected, the theoretical C-curves in Fig. 6.26 reveal a strong effect of boron alloying
on the HAZ transformation kinetics. By considering the superimposed weld cooling curve, it
is possible to estimate the volume fraction of ferrite f^~Fe which forms from equations (6-32)
and (6-33). Taking the time exponent n in the Avrami equation equal to 5/2 for nucleation of
ferrite at austenite grain boundaries and (1 - Xc) = 0.98 as in the previous examples, we ob-
tain after integration within the temperature range from 800 to 600° C:
and
If the same calculations are performed for the reference steel in Table 6.2 (characterised by
[*]gb = iBfgb )> we get:
and
From this we see that boron, even in small quantities, can have a dramatic effect on the
HAZ transformation kinetics by promoting the formation of bainite and martensite at the ex-
pense of grain boundary ferrite. This is in good agreement with general experience (see ex-
perimental CCT-diagrams in Fig. 6.27).
Effect of solidification-induced segregation

As shown in Chapter 3 of this textbook, the characteristic solidification pattern of mild and
low-alloy steel weld metals leads to extensive segregation of alloying and impurity elements
to grain boundaries and interdendritic spaces. Of particular interest in this respect is phospho-
rus segregations at columnar austenite grain boundaries. Referring to Figs. 3.38 and 3.39
(Chapter 3), the phosphorus-rich zone is seen to extend about ±7jam on either side of the grain
boundaries. Since phosphorus is among the strongest ferrite stabilising elements in steel (see
equation (6-36)), the existence of such solidification-induced segregations would be expected
Temperature, 0C
Time, s
Fig. 6.26. Predicted displacement of ferrite C-curve in temperature-time space due to segregation of
boron to prior austenite grain boundaries (Example 6.4). The superimposed cooling curve corresponds
to a cooling time, Ats/5, of 9.2s.
to enhance the nucleation rate of grain boundary ferrite at these sites due to the associated
increase in the Ae3-temperature. The above phenomenon should not be confused with equilib-
rium segregation of phosphorus to austenite grain boundaries during heat treatment of steel,
which stems form attraction of the atoms towards the open structure of the boundary. In the
latter case phosphorus may act as a hardenability element by occupying favourable sites for
ferrite nucleation at the austenite grain boundaries analogous to that documented for boron in
steel.43'44
Example (6.5)
Suppose that the local phosphorus content adjacent to the columnar austenite grain boundaries
in a low-alloy steel weld metal is 500 ppm, whereas the bulk concentration of phosphorus is
100 ppm. Estimate on the basis of the theory developed in the previous sections the resulting
displacement of the ferrite C-curve in temperature-time space when the columnar austenite
grain size is 80|jun. In these calculations we shall assume that the transformation characteristic
of the bulk metal is similar to that of the reference steel in Table 6.2.
Solution
First we estimate the actual Ae3-temperature within the phosphorus-rich region adjacent to the
columnar austenite grain boundaries from equation (6-36). Taking the A<?3-temperature of the
bulk phase equal to 835°C (1108K), the local phase boundary temperature becomes:
By substituting data from Table 6.2 into equation (6-35), we arrive at the following expres-
sion for ^ 0 :
(a)
Temperature, 0C
Steel A (11 ppmB)
Cooling time, At0,., s

(b)
Steel B (26 ppm B)
Temperature, 0C
Cooling time, A t 8/5 , s
Fig. 6.27. CCT-diagrams for boron-containing steels; (a) Steel A (llppm B), (b) Steel B (26ppm B).
Data from Akselsen et al.42
It is evident from the graphical representation of the above equation in Fig. 6.28 that the
observed increase in the yl^-temperature from 1108 to 1136K displaces the ferrite C-curve
towards higher temperatures and shorter times in the IT-diagram. The resulting effect on the
weld metal transformation kinetics is obvious, since an increase in the nucleation rate of fer-
rite will favour early site saturation at the austenite grain boundaries. On this basis it is not
surprising to find that allotriomorphic ferrite in low-alloy steel weld metals tends to form con-
tinuous veins of blocky ferrite along the columnar austenite grain boundaries, as shown in Fig.
6.19(c). Outside the fusion zone the conditions for early site saturation are less favourable,
since modern steelmaking practice implies that solidification-induced segregations are re-
moved by prolonged high-temperature annealing prior to the welding operation. Hence, fer-
rite veining of the type shown in Fig. 6.19(c) is not commonly observed within the reheated
regions of the base plate, unless the heat input is extremely large (see Fig. 6.19(a) and (b)).
Temperature, 0C
Time, s
Fig. 6.28. Effect of solidification-induced phosphorus segregations on the austenite to ferrite transforma-
tion in low-alloy steel weld metals (Example 6.5).
Effect of cooling rate on ferrite grain size

In steel metallurgy it is well accepted that accelerated cooling refines the ferrite grain size, and
thus improves both strength and toughness of the parent material. 1920 If the austenite is al-
lowed to recrystallise before it transforms, the final ferrite grain size Da in the base plate will
be an unique function of the prior austenite grain size D1 and the cooling rate through the
critical transformation temperature range for ferrite formation (CR.).
The fundamental aspects of ferrite grain refinement by accelerated cooling have been con-
sidered by Umemoto et al.19 By allowing for the variation in the nucleation and growth rates
of allotriomorphic ferrite with temperature (see Fig. 6.18), they arrived at the following theo-
retical expression for Da when austenite grain faces are the dominant nucleation site of fer-
rite:
(6-42)
where D1 is in |im and (CR.) is in 0C s"1.

In practice, Umemoto et al.19 observed a discrepancy between theory and experiments due
to competitive nucleation of ferrite at austenite grain edges and corners (characterised by a
theoretical grain size dependence of D12/3 and D1, respectively). Consequently, the real fer-
rite grain size in the steel varied with austenite grain size and cooling rate as:
(6-43)
where C4 is a kinetic constant which is characteristic of the alloy system under consideration.
An indication of the applicability of equation (6-43) to welding can be obtained from the
micro structure data of Evans,45 reproduced in Fig. 6.29. In this plot, the reported heat inputs
have been converted into an equivalent cooling rate at 7000C via equation (1-71) (Chapter 1).
It is evident from the figure that the observed variation in the ferrite grain size with cooling
rate is consistent with calculations based on equation (6-43). Hence, the model of Umemoto et
Ferrite grain size,jim
Equation (6-43)
Weld A (0.04 wt% C, 0.58 wt% Mn)

Weld C (0.05 wt% C, 1.33 wt% Mn)
Cooling rate at 700 0C1 °C/s

Fig. 6.29. Variation of ferrite grain size with cooling rate in reheated C-Mn steel weld metals. Data from
Evans.45
al19 can readily be employed for prediction of the ferrite grain size within the grain refined
region of both single pass and multipass steel welds, provided that the austenite grain size and
the cooling rate can be estimated with a reasonable degree of accuracy.
Example (6.6)
Consider multipass welding with covered electrodes (SMAW) on a thick plate of low-alloy
steel under the following conditions:
Previous experience has shown that polygonal ferrite forms within the low-temperature
reheated regions of the weld metal, typically 1.0 to 1.7 mm beneath the surface (fusion bound-
ary) of subsequent weld passes. Estimate on the basis of the grain growth diagram in Fig.
5.24(a) (Chapter 5) and equation (6-43) the maximum variation in the ferrite grain size across
the weld HAZ under the prevailing circumstances. Thermal data for low-alloy steels are given
in Table 1.1 (Chapter 1).
Solution
First we need to convert the depths at which polygonal ferrite appears beneath the fusion
boundary to an equivalent (characteristic) peak temperature range. If we neglect the contribu-
tion from heat flow in the welding direction, this conversion can be done on the basis of
equation (5-47) in Chapter 5:
Taking Ar*m equal to 1 and 1.7mm, respectively, we get:
and
The prior austenite grain size Dy at these two locations can now be read from Fig. 5.24(a)
and Table 5.6, respectively:
Since the cooling rate (CR.) at a given temperature is essentially the same across the weld
HAZ, the maximum variation in the ferrite grain size can be evaluated directly from equation
(6-43) without further background information:
From this we see that the variation in the ferrite grain size is significantly smaller than the
corresponding change in the prior austenite grain size. This result is in good agreement with
general experience.
6.3.3.3 Growth of grain boundary ferrite

If the austenite grain boundaries become rapidly decorated with a continuous layer of ferrite
(so that the subsequent transformation involves the reconstructive thickening of these layers),
the evolution of allotriomorphic ferrite is determined solely by its growth kinetics.37 As shown
in Example 6.5, this is a realistic assumption in the case of as-deposited steel weld metals,
where the presence of solidification-induced phosphorus segregations at the columnar austenite
grain boundaries favours an early site saturation.
Several investigators have modelled the evolution of allotriomorphic ferrite in low-alloy
steel weld metals along the lines indicated above.46"48 The most thorough analysis is probably
that of Bhadeshia et al.46 who were able to account for the combined effect of temperature and
steel chemical composition on the growth kinetics. However, in order to illustrate the compe-
tition between the different variables that contribute to the formation of grain boundary ferrite
in as-deposited steel weld metals, the simplified treatment of Liu and Olson47 and Fleck et al.4S
has been adopted here.
As a starting point the Avrami equation17 is considered for the limiting case of zero nuclea-
tion rate by assuming early site saturation. If subsequent growth of the grain boundary ferrite
allotriomorphs occurs in both directions perpendicular to the austenite grain boundaries at a
rate which is controlled by diffusion of carbon in austenite, the time dependence of the ferrite
thickness, AZa, is given by equation (6-17):
(6-44)
where E1 is the one-dimensional parabolic thickening constant (defined in equation (6-18)).
The corresponding growth rate of allotriomorphic ferrite Ua is then obtained by differenti-
ating equation (6-44) with respect to time:
(6-45)
From a stereological standpoint, the grain boundary surface area per unit volume Sv cannot
be calculated without further assumptions regarding the shape of the columnar austenite grains.
Bhadeshia et al.31*46*49 solved this problem by representing the grain morphology by a uniform,
space-filling array of hexagonal prisms. However, with the precision aimed at here, it is suffi-
cient to assume that Sv is equal to the surface area per unit volume of an inscribed cylinder
whose volume is equivalent to that of the hexagonal prisms (see Fig. 6.30). Noting that each
grain boundary is shared by two adjacent grains, the expression for Sv (in the absence of end
effects) becomes:47'48
(6-46)
where dy and I1 are the diameter and length of the inscribed cylinder, respectively.
By substituting the above expressions for Ua and Sv into equation (6-34), we obtain:
(6-47)
In practice, the one-dimensional parabolic thickening constant in equation (6-47) varies

both with temperature and steel chemical composition, as shown by the data of Bhadeshia et
aL46 reproduced in Fig. 6.31. However, if we instead use a reasonable average value for E1
within the characteristic transformation temperature range for allotriomorphic ferrite (i.e.
E1 = E1), the integral I2 in equation (6-47) has the following analytical solution:
x Growth direction of columnar grains
Fig. 6.30. Hexagonal prism model for the columnar austenite grain morphology in low-alloy steel weld
metals. The inscribed cylinder in the figure has approximately the same surface to volume ratio as the
hexagonal prisms.
Parabolic thickening constant (E1), ^m/s1/2
Temperature, 0C
Fig. 6.31. Effect of temperature and steel chemical composition on the one-dimensional parabolic thick-
ening constant for allotriomorphic ferrite. (1): 0.03 wt% C, (2): 0.06 wt% C, (3): 0.08 wt% C and (4):
0.10 wt% C. Data from Bhadeshia et al.46
(6-48)
from which
(6-49)
Equation (6-49) predicts that the volume fraction of grain boundary ferrite in the as-depos-
ited weld metal depends on the combined action of the following three main variables:
(i) The one-dimensional parabolic thickening constant ei which is determined by the weld
metal chemical composition (i.e. the content of austenite and ferrite stabilising elements).
(ii) The columnar austenite grain size dy which is controlled by the weld metal solidification
microstructure (i.e. the weld metal chemistry, the weld pool geometry, and the thermal
conditions under which solidification occurs).
(iii) The retention time within the critical transformation temperature range for allotriomorphic
ferrite, as determined by the applied heat input and the mode of heat flow (i.e. thick plate,
medium thick plate, or thin plate welding, respectively).
It can be seen from the microstructure data of Grong et al.50 reproduced in Fig. 6.32 that the
influence of these variables are adequately accounted for in the present model. However, the
calculated volume fractions of grain boundary ferrite are consistently lower than the measured
ones. This discrepancy can probably be attributed to the use of constant values for the one-
dimensional parabolic thickening constant. Consequently, if proper corrections are made for
the inherent variation in E1 with temperature and steel chemical composition, the agreement
between theory and experiments is significantly improved, as shown by the data of Bhadeshia
et al.46 reproduced in Fig. 6.33.
Example (6.7)
Consider GMA welding on a thick plate of a low-alloy steel under the following conditions:
Suppose that the grain boundary allotriomorphs form within the temperature range from
750 to 6000C at a constant rate ei of 3 jam s~1/2. Estimate on the basis of the theory developed
in the previous section the volume fraction of allotriomorphic ferrite in the weld deposit when
the columnar austenite grain size is 80|Lim. In these calculations we shall assume that the
equilibrium volume fraction offerritef®q~Fe in the fully transformed steel is 0.9. Thermal data
for low-alloy steels are given in Table 1.1 (Chapter 1).
Solution
The situation is described in Fig. 6.34. In this case the problem is to estimate the retention time
/Str within the critical temperature range for ferrite formation from the Rosenthal thick plate
solution (Chapter 1). From equation (1-66), we have:
Regression line
Volume fraction of GF
C-Mn steel weld metals

Si content: 0.50-1.03 wt%
Mn content: 0.33-2.72 wt%
x * . ^ )
Fig. 6.32. Experimental verification of equation (6-49). Data from Grong et al.5Q
C-Mn steel weld metals
Measured volume fraction of GF

Line of unit slope
Calculated volume
fraction of GF
Fig. 6.33. Comparison between measured and predicted volume fractions of grain boundary ferrite in
C-Mn steel weld metals. Data from Bhadeshia et al.46
Temperature
log time
Fig. 6.34. Conditions for allotriomorphic ferrite formation in low-alloy steel weld metals (Example 6.6).
By inserting this value into equation (6-49), we obtain:
Although experimental data are not available for a direct comparison, the calculated value
of fa'Fe is reasonable and of the expected order of magnitude.
6.3.4 Widmanstdtten ferrite

With increasing degree of undercooling the redistribution of carbon becomes insufficient to
maintain a planar growth mode, and hence, further growth of the ferrite can only take place by
lateral movement of ledges along a low-energy interface*.120 This, in turn, implies a
Kurdjumow-Sachs-type orientation relationship between the austenite and the ferrite, i.e.
{Ill }1_Fe parallel with {110}a_F6, and <110> 7 .^ parallel with <11 l>a_Fe, which is a character-
istic feature of the Widmanstatten ferrite structure. The ferrite sideplates, once nucleated,
grow very rapidly under the prevailing conditions because of an efficient redistribution of
carbon to the sides of the advancing interface.1 As a result, parallel arrays of ferrite laths of
high aspect ratios (typically 10:1 to 20:1) are often found in the areas adjacent to the austenite
grain boundaries, as shown by the optical micrographs in Fig. 6.19(b) and (c).
Following the treatment of Bhadeshia et al.,46 the lengthening rate of Widmanstatten ferrite
can be estimated using the Trivedi theory9 for diffusion-controlled growth of parabolic cylin-
ders with correction for the assumed displacive character of the Widmanstatten ferrite trans-
formation. As shown in Fig. 6.35, the pertinent growth rates are normally so large that the
formation of Widmanstatten ferrite is essentially complete within a fraction of a second. This
implies that the transformation, for all practical purposes, can be treated as being isothermal.37
When the growth rate is known, the volume fraction of Widmanstatten ferrite in the as-
deposited weld metal may be estimated by assuming that the ferrite sideplates nucleate at a
constant rate at the yFelaFe boundaries and subsequently grow into the interior of the colum-
nar austenite grains until they physically impinge with intragranularly nucleated acicular fer-
rite or allotriomorphic ferrite.46 It is seen from the microstructure data of Bhadeshia et al.46
reproduced in Fig. 6.36 that the calculated volume fractions of Widmanstatten ferrite are in
reasonable agreement with experiments, although the discrepancy in certain cases is admit-
tedly large.
*A different view is suggested by Bhadeshia et al.31'46'51'52 who claim that growth of Widmanstatten ferrite occurs in
a displacive manner analogous to that documented for martensite in steel, with the exception that carbon must diffuse
during growth.
Next Page
a b
Growth rate of WF, u,m/s
Temperature, 0C
Fig. 6.35. Predicted growth rates of Widmanstatten ferrite in C-Mn steel weld metals; (a) Growth rate
calculations for weld compositions listed in Fig. 6.31, (b) Growth rate calculations after modifying com-
positions to allow for carbon enrichment due to grain boundary ferrite formation. Data from Bhadeshia
etai46
6.3.5 Acicular ferrite in steel weld deposits

Simultaneously with or immediately after the formation of Widmanstatten ferrite at the austenite
grain boundaries, acicular ferrite may start to nucleate intragranularly at non-metallic inclu-
sions. This phase is commonly observed in low-alloy steel weld metals, where the fine disper-
sion of oxide inclusions provides favourable sites for heterogeneous nucleation.3653 There
seems to be general agreement that microstructures primarily consisting of acicular ferrite
provide optimum weld metal mechanical properties, both from a strength and toughness point
of view, by virtue of its small lath size and high dislocation density. 365354 Consequently, the
formation of this particular microconstituent is of significant commercial importance and has
therefore attracted substantial research interest over the years. 363746 " 58 In spite of all this ef-
fort, the acicular ferrite transformation in low-alloy steel weld metals is still a subject of con-
siderable controversy.
6.3.5.1 Crystallography of acicular ferrite

It is well established that acicular ferrite nucleates in the transformation temperature range
between Widmanstatten ferrite and lower bainite.55"59 Based on conventional diffraction pat-
tern analyses in the transmission electron microscope (TEM), Bhadeshia 6tf a/.54'59have shown
that the acicular ferrite plates exhibit an orientation relationship with the austenite grain in
which they grow. The observed orientation relationship lies within the Bain orientation re-
gion27 and can approximately be described as:
Previous Page
a b
Growth rate of WF, u,m/s
Temperature, 0C
Fig. 6.35. Predicted growth rates of Widmanstatten ferrite in C-Mn steel weld metals; (a) Growth rate
calculations for weld compositions listed in Fig. 6.31, (b) Growth rate calculations after modifying com-
positions to allow for carbon enrichment due to grain boundary ferrite formation. Data from Bhadeshia
etai46
6.3.5 Acicular ferrite in steel weld deposits

Simultaneously with or immediately after the formation of Widmanstatten ferrite at the austenite
grain boundaries, acicular ferrite may start to nucleate intragranularly at non-metallic inclu-
sions. This phase is commonly observed in low-alloy steel weld metals, where the fine disper-
sion of oxide inclusions provides favourable sites for heterogeneous nucleation.3653 There
seems to be general agreement that microstructures primarily consisting of acicular ferrite
provide optimum weld metal mechanical properties, both from a strength and toughness point
of view, by virtue of its small lath size and high dislocation density. 365354 Consequently, the
formation of this particular microconstituent is of significant commercial importance and has
therefore attracted substantial research interest over the years. 363746 " 58 In spite of all this ef-
fort, the acicular ferrite transformation in low-alloy steel weld metals is still a subject of con-
siderable controversy.
6.3.5.1 Crystallography of acicular ferrite

It is well established that acicular ferrite nucleates in the transformation temperature range
between Widmanstatten ferrite and lower bainite.55"59 Based on conventional diffraction pat-
tern analyses in the transmission electron microscope (TEM), Bhadeshia 6tf a/.54'59have shown
that the acicular ferrite plates exhibit an orientation relationship with the austenite grain in
which they grow. The observed orientation relationship lies within the Bain orientation re-
gion27 and can approximately be described as:
Measured volume fraction of WF
Calculated volume
fraction of WF
Fig. 6.36. Comparison between measured and predicted volume fractions of Widmanstatten ferrite in C-
Mn steel weld metals. Data from Bhadeshia et al.46
This corresponds to the well-known Kurdjumow-Sachs (K-S) orientation relationship* and

suggests that growth of the AF plates occurs either by a ledge mechanism or by a pure shear
transformation similar to that reported for upper bainite in steel. 1^0-60"62
6.3.5.2 Texture components of acicular ferrite

Kluken et al.63 have studied the development of transformation textures in as-deposited steel
weld metals containing acicular ferrite, using the electron backscattering pattern (EBSP) tech-
nique. Referring to the (200) stereographic projection in Fig. 6.37, the measured acicular
ferrite orientations form a symmetrical and consistent pattern and disclose evidence of a cou-
pled solidification/solid state transformation texture in the weld metal. A closer examination
of the data reveals the existence of three major texture components within the acicular ferrite
microstructure, i.e. one <100> component and two complementary <111> components.63 The
former component includes acicular ferrite plates which have a <100> direction approximately
*The orientation relationship can alternatively be described by the Nishiyama-Wasserman (N-W) correspondence
which also lies within the Bain orientation region. However, since the K-S and the N-W orientation relationships
only differ from each other by a 5.26° rotation of the close packed planes, they can be regarded as equivalent.
(a) (b)
Fig. 6.37. The development of transformation textures in as-deposited steel weld metals containing acicular
ferrite; (a) (200) pole figure showing the crystallographic orientations of acicular ferrite referred to the
original cell/dendrite growth direction, (b) Backscattered electron channeling contrast image of delta
ferrite/austenite columnar grain (the metallographic section is normal to the cell/dendrite growth direc-
tion). After Kluken et al.63
parallel with the cell/dendrite growth direction (Fig. 6.38(a)). The other plates have a <111>
direction aligned in the same crystal growth direction (Fig. 6.38(b)). These data can be repre-
sented by two sub-components which are displaced with respect to each other by a 60° rotation
about a common <111> axis.
It is evident from the measurements of Kluken et al.63 that the acicular ferrite plates in as-
deposited steel weld metals exhibit an orientation relationship with both the austenite and the
prior delta ferrite columnar grains in which they grow. This 'memory' effect arises from the
characteristic solidification pattern and transformation behaviour of low-alloy steel welds (e.g.
see discussion in Section 3.8.2, Chapter 3). As shown schematically in Fig. 6.39, the columnar
grain region will exhibit a sharp <100> solidification texture which has its origin in the phe-
nomenon of preferred crystal growth. At the onset of the peritectic reaction, the austenite
adopts a K-S type of orientation relationship with the delta ferrite in order to minimise the
energy barrier against nucleation.64 The austenite subsequently grows around the periphery of
the primary phase until impingement occurs on neighbouring columnar grain boundaries. During
the yFe to aFe transformation, this memory effect gives rise to the formation of acicular ferrite
plates which have a <100> direction approximately parallel with the original cell/dendrite
growth direction. The presence of the two other texture components within the weld metal is
thus a result of complementary crystal rotations taking place within the same orientation re-
gion.
The proposed sequence of reactions is in excellent agreement with the texture analysis of
Hu65 who made theoretical calculations of the resulting orientations of iron after §Fe to yFe and
yFe to aFe transformations in succession according to the scheme outlined in Fig. 6.39.
6.3.5.3 Nature of acicular ferrite

Currently, the mechanisms of acicular ferrite formation in low-alloy steel weld metals are not
Cell/dendrlte growth direction
(a) (b)
Fig. 6.38. Schematic diagrams showing the three main texture components in acicular ferrite according
to the Kurdjumow-Sachs orientation relationship; (a) The <100> texture component, (b) The two com-
plementary <111> texture components. After Kluken et a/.63
fully understood. However, detailed TEM studies performed by Bhadeshia et aL37>59>62>66 have
clearly demonstrated that acicular ferrite is a form of intragranularly nucleated bainite. In
practice, this means that the microconstituent may be present either as 'upper' or 'lower'
acicular ferrite in the weld deposit (depending on the carbon concentration), as shown
schematically in Fig. 6.40.
In general, the ferrite component of upper bainite is composed of groups of thin parallel
laths (subplates) with a well-defined crystallographic habit.2061 Although the growth mecha-
nism of upper bainite is still a subject of considerable controversy, it has been postulated that
the subplates advance into the austenite with their own tip configurations. One model is shown
in Fig. 6.41, where each subplate forms as a ledge upon the adjacent subplates through a
nucleation and growth process. These ferrite laths possess the same variant of the K-S orien-
tation relationship, which means that they are separated by low-angle grain boundaries. A
typical austenite grain will contain numerous sheaves of bainitic ferrite exhibiting different
variants of the K-S orientation relationship. This implies that the boundary between adjacent
plates of acicular ferrite should alternately be of the low-angle and high-angle type, a feature
which also has been observed experimentally (see data in Fig. 6.42). Hence, both the mor-
phology and the crystallography of acicular ferrite bear a close resemblance to upper bainite.
Columnar grain
Fig. 6.39. Schematic diagram showing the sequence of reactions occurring during cooling of a low-alloy
steel weld through the critical transformation temperature ranges. After Kluken et al.63
6.3.5.4 Nucleation and growth of acicular ferrite

Since the thickening rate of plate-like precipitates is very high, the acicular ferrite laths, once
nucleated, will grow into the austenite phase until they physically impinge with neighbouring
plates. At present, there exist no models which allow the acicular ferrite content to be calcu-
lated from first principles. The reason for this is quite obvious, since the final weld metal
microstructure generally depends on complex interactions between a number of different vari-
ables, including:36
(i) The total alloy content.

(ii) The concentration, chemical composition, and size
distribution of non-metallic inclusions,
(iii) The solidification microstructure.
(iv) The prior austenite grain size,
(v) The weld thermal cycle.
Carbon supersaturated plate
Carbon diffusion Carbon diffusion into

into austenite austenite and carbide
precipitation in ferrite
Carbide precipitation
from austenite
'Upper' acicular 'Lower' acicular

ferrite ferrite
Fig. 6.40. Schematic illustration of the transition from 'upper' to 'lower' acicular ferrite in low-alloy
steel weld metals. The diagram is based on the ideas of Bhadeshia and Svensson.37
Successive
nucleation and
growth of
parallel plates
Fig. 6.41. Proposed model for nucleation and growth of upper bainite in steel (schematic). After
Verhoeven.61
Misorientation, degrees
Plate number
Fig. 6.42. Measured spatial misorientation between adjacent plates of acicular ferrite in a low-alloy steel
weld. Data from Kluken et a/.63
A more realistic approach would be to estimate the volume fraction of acicular ferrite
via the equation:37
(6-50)
where f%FFe and f%FFe are the corresponding volume fractions of grain boundary ferrite and
Widmanstatten ferrite, respectively (note that in equation (6-50) the formation of microphases
has been disregarded).
The method outlined above has shown to work well for numerous welds (e.g. see Fig. 6.43),
but fails when the primary microstructure consists of a mixture of acicular ferrite and martensite,
as is the case in high strength steel weld deposits6768 In spite of this shortcoming, equation (6-
50) expresses in an explicit manner the real essence of the problem, namely that the evolution
of the acicular ferrite microstructure depends on the interplay between several competing nu-
cleation and growth processes which occur consecutively during cooling from the Ae3-tem-
perature. This important point is often overlooked when discussing the conditions for acicular
ferrite formation in low-alloy steel weld metals.
Size effects in heterogeneous nucleation

The important influence of second-phase particles on the austenite to ferrite transformation
has been examined theoretically by Ricks et al69 using classical nucleation theory. By assum-
ing inert, incoherent, and non-deformable inclusions and constant values for the volume free
energy change and the surface free energy of both the yFe /yFe and yFe /aFe boundaries, the
normalised energy barrier to nucleation has been calculated and plotted against the particle
radius as shown in Fig. 6.44. It can be seen from the figure that the nucleation of ferrite at
inclusions is always energetically more favourable than homogeneous nucleation, but less
favourable than nucleation at austenite grain boundaries, irrespectively of the inclusion size.
The most potent nucleation sites are particles of a radius greater than about 0.2-0.5 jum, which
Allotriomorphic
ferrite
Volume fraction
Widmanstatten
ferrite
Acicular
ferrite
Carbon content, wt%

Fig. 6,43. Predicted variation in microstructure as a function of carbon concentration in C-Mn steel
weld deposits (SMAW-IkJ mm~ l). Data from Bhadeshia and Svensson.37
horn.
AG* /AG*
Inclusion
het.
, Austenite
grain boundary
Inclusion radius, jim

Fig. 6.44. Effect of particle radius on the normalised energy barrier against ferrite nucleation at inclu-
sions. The corresponding energy barrier against nucleation of ferrite at austenite grain boundaries is
indicated by the horizontal broken line. Data from Ricks et al.69
is within the typical size range of most weld metal inclusions (see Figs. 2.57-2.61 in Chapter
2). This finding is in excellent agreement with the results of Barbaro et al.70 reproduced in Fig.
6.45, showing that a certain minimum inclusion size (say 0.2-0.3 jam) is required for acicular
ferrite nucleation in steel weld deposits.
It should be noted, however, that Ricks et al69 omitted a consideration of the effects of
plastic strain produced as a result of differences in thermal contraction between the austenite
and the particles as well as the possibility for the ferrite to adopt reasonable orientation rela-
tionships with both the austenite and the catalyst particles. Based on nucleation theory it can
be argued that these factors will influence the transformation process.1 This, in turn, may
explain why certain types of inclusions appear to be more favourable nucleation sites for acicular
ferrite than other (to be discussed below).
Catalyst effects in heterogeneous nucleation

As mentioned above, there is considerable circumstantial evidence available in the scientific
literature that intragranular nucleation of acicular ferrite is preferentially associated with spe-
cific types of non-metallic inclusions (i.e. 7-Al2O3, MnOAl2O3, TiN). 365571 " 78 Different mecha-
nisms have been proposed over the years to explain these phenomena, including:3655'73'76'77
(i) Nucleation resulting from a small lattice disregistry between the inclusions and the
ferrite.
(ii) Nucleation in the vicinity of inclusions caused by local compositional inhomogeneity

in the steel matrix.
(iii) Nucleation in the vicinity of inclusions resulting from favourable strain or dislocation
arrays due to differences in the thermal contraction between the particles and the
matrix.
Probability of nucleation
Inclusion radius, JLX m

Fig. 6.45. Effect of inclusion size on the probability of acicular ferrite nucleation in steel weld deposits.
Data from Barbara et al.10
Because of the complexity of the weld metal inclusions, and the experimental difficulties
involved in performing controlled in situ measurements, it cannot be stated with certainty
which of these three mechanisms that are operative during the acicular ferrite transformation.
However, based on simple theoretical calculations it can be argued that the contribution from
the elastic strain fields around the particles due to differential contraction effects probably is
too insignificant to influence the free energy of transformation and that the resulting punching-
stress at the particle/matrix interface is well below the critical value required to generate new
dislocations in the austenite.47'81'82 Moreover, detailed STEM/EDS microanalyses have failed
to reveal detectable variations in the matrix composition in the vicinity of the inclusions.83
Hence, nucleation resulting from a small lattice disregistry between the inclusions and the
ferrite appears to be the most likely explanation to the observed effects of deoxidation practice
on the weld metal transformation behaviour.
From a theoretical standpoint, the development of a faceted ferrite nucleus which exhibits a
rational orientation relationship with both the austenite and the inclusions would require that
the substrate and the austenite have similar crystal structures and identical lattice orientations.
The catalyst particles must therefore be cubic and bear an orientation relationship with the
austenite which lies within the Bain region.27 However, considering the fact that the weld
metal inclusions form in the liquid state prior to the solidification process, the latter require-
ment cannot generally be met.37 Nevertheless, even if the orientation of the inclusions were
perfectly random, it is apparent that orientation relations within the Bain region would be
observed purely by chance. In view of the high symmetry of the cubic system, the probability
of this happening must be calculated. An approximative estimate is given below for single
phase cubic inclusions, based on the method described by Ryder et al?4
Figure 6.46 contains a standard stereographic projection of the austenite crystal, showing
the <100> rFe -poles (squares) and the <110>rFe-poles (dots). The Bain orientation region is
represented by small circles of radius 11° centered on the <100> r / v - and the <110> rFe -poles.84
Fig. 6.46. Stereographic projection of Bain regions represented by 11° circles, round <100>7_Fe-poles
(squares) and <110>7_Fe-poles (dots) of the austenite lattice. After Ryder et aiS4
The austenite/inclusion(i) orientation relationship is within the orientation region derived

from the Bain correspondence if one <100>rpole lies within a <100>7_Fe-region and the other
two <100>rpoles lie within <110>rF<?-regions. Assuming a random inclusion orientation, the
probability P 1 that a given <100>rpole lies within a given <100> r ^-region is (o/47T, where o>
is the solid angle enclosed by one <100> r ^-region (equal to 27t(l-cos 11°) = 0.115). Since
there are three <100>/ directions and six <100>7_Fe-regions, the probability that at least one
<100>rpole lies within a <100>7_F£,-region is equal to:
(6-51)
Imagine now that the inclusion lattice is rotated through 360° about the <100>raxis which
lies within a <100> y . Fe -region. The other two <100> r poles will then lie within
<110>7.Fe-regions for at the most 4 X 22° = 88° of this rotation, since the diameter of the
<100>7_Fe-regions is 22°. Hence, the probability P2 that, if one <100> r pole lies within a
<100>7.Fe-region, the other two will lie within <110>7.Fe-regions is given in the upper limit
by:
(6-52)
The total probability that a given orientation relationship lies within the Bain region purely
by chance is thus:
(6-53)
Therefore, assuming random orientation, about 4% of the weld metal inclusions would lie
within the Bain region purely by chance if they were single phase cubic crystals. In practice,
however, inclusions commonly found in low-alloy steel weld metals are of a very heterogene-
ous chemical and crystalline nature. As shown in Fig. 2.72 in Chapter 2, a typical inclusion
may contain up to six different constituent phases, including the three cubic phases 7-Al2O3,
MnOAl2O3, and TiN. This implies that at least 12% of the inclusions may contain a cubic
phase which lies within the Bain orientation region.
Measurements of orientation relationships between specific inclusion constituent phases
(i.e. 7-Al2O3, MnOAl2O3, and TiN) and contiguous acicular ferrite plates performed by Grong
et a/.85 support the above interpretation (see Table 6.3). Referring to the standard stereographic
projections in Fig. 6.47, a very high proportion of the ferrite/inclusion orientations falls within
the Bain region. In addition, two other variants (i.e. No. 5 and 6) are indicated for TiN which
do not meet this requirement. They are therefore regarded as spurious (in the sense that the
observed orientation relationships do not stem from a catalyst nucleation event) and should be
ignored. Hence, it may be concluded that the observed orientation relationships between acicular
ferrite and specific inclusion constituent phases are not fully reproducible in the true meaning
of the word, since only those combinations which satisfy the inherent crystallography of the
acicular ferrite microstructure are acceptable.
An interesting observation from the data in Table 6.3 is that nucleation of acicular ferrite on
inclusions is always associated with low-index planes of the {100} or the {110} type, which
indicates a faceted growth morphology of the inclusions. Faceted growth may occur as a result
of anisotropy in the growth rates between high-index and low-index crystallographic planes,
and can in the extreme case lead to a morphology of the type shown in Fig. 6.48. Conse-
quently, formation of faceted inclusions in the liquid steel during deoxidation appears to be an
intrinsic feature of low-alloy steel weld metals.
Simple verification on the basis of classic nucleation theory shows that the associated re-
duction of the energy barrier to nucleation, AGlet , is the primary cause for the ferrite nucleus
to develop epitaxial orientation relationships with the substrate and the austenite.8687 Refer-
ring to Fig. 6.49, a qualitative ranking of the different inclusion constituent phases with respect
to nucleation potency of acicular ferrite can be made from the data presented in Table 6.3. It is
evident that both 7-Al2O3, MnOAl2O3, and TiN reveal a good lattice matching with the ferrite
phase in one crystallographic direction. In addition, nucleation of acicular ferrite at TiN offers
Table 6.3 Observed orientation relationships between acicular ferrite and different inclusion constitu-
ent phases in a SA low-alloy steel weld. Data from Grong et al.85
Substrate Orientation Relationship Interplanar Spacing
(s) Variant No. Plane Combinations Ratio^
7-Al2O3 (100) s ~//(011) a _ Fe 1 {011}a_Feand{400}s 1.02

(distorted spinel) [011]s ~ // [533] a . Fe
MnOAl2O3 (200) s ~//(110) a F e 2 {110}a_Feand {400}s 0.99
(spinel) (011)s // (010)a Fe 3 {200}a Fe and {440}s 0.99
[Oil],//[QOlLp 6
(TlO) 8 // (100)a Fe {200}a Fe and {220}s 0.97
(112)s//(011)a_Fe 4 {011} a . Fe and{112} s 1.17
[lll] s //[011] a . F e
TiN (101) s //(103) a _ Fe {310} a . Fe and{330} s 0.91
(NaCl) (320)s//(112)a_Fe 5 {112}a.Fe and {320}s 0.99
[232] s //[351] a _ Fe
(101) s //(133) a _ Fc 6 {133} aFe and{550} s 1.10
[221]s//[200]a_Fe {200}a.Feand {221 }s 1.01
Defined as d(hkl)a.Fe /d(hkl)s

Fig. 6.47. Standard (100) stereographic projections of the orientation relationships listed in Table 6.3; (a)
Variant (1) and (3), (b) Variant (4), (5) and (6). The Bain orientation region is indicated by the 11° circles
in the graphs (see Fig. 6.46 for details).
the advantage of partial lattice coherence in a second (independent) direction, which further
contributes to a reduction of AGê/ through a minimisation of the interfacial energy between
the two phases. This makes TiN an extremely efficient nucleant for acicular ferrite.
Microstructure data available for submerged arc (SA) steel weld deposits clearly support
the above findings that nucleation of acicular ferrite occurs preferentially at inclusions which
contain aluminium or titanium. As shown in Fig. 6.50, a high volume fraction of acicular
ferrite is always achieved when sufficient amounts of titanium are added either through the
filler wire or the flux, irrespectively of the aluminium and oxygen concentrations. This is in
sharp contrast to welds produced with welding consumables containing low levels of titanium,
where the acicular ferrite content drops rapidly with decreasing [A%Al]weld/[%O]anal ratios
Fig. 6.48, Example of a faceted crystal delimited by {100} and {110} planes (schematic).
Austenite grain boundary

;et.
AG
MnOAI2O3
T-Al2O3
TiN
Nucleation site
Fig. 6.49. Qualitative ranking of different inclusion constituent phases with respect to nucleation po-
tency of acicular ferrite.
due to the presence of lower fractions of 7-Al2O3 and MnOAl2O3 in the inclusions (see Fig.
2.72 in Chapter 2). Similar observations have also been made by other investigators.3672"
76,78,79
It should be noted that the weld metal transformation behaviour in practice depends on
complex interactions between a number of important variables, including alloying and
deoxidation practice, the solidification microstructure, the prior austenite grain size, and the
weld thermal cycle.36'5358 This means that the presence of 7-Al2O3, MnOAl2O3 or TiN at the
surface of the inclusions is perhaps a necessary but not a sufficient criterion for formation of
acicular ferrite in steel weld metals.
Example (6.8)
Consider a partly Ti-Al deoxidised steel weld metal which contains a total number of 4 X 107
inclusions per mm 3 . Based on Fig. 2.72 in Chapter 2 and the theory developed above, estimate
an upper limit for the volume of a typical plate of acicular ferrite when the weld metal [A%Al]weld/
\y°O\anaL r a t i o i s °- 80 -
Vol% AF
[A%A|lweld/[%o]anal
Fig. 6.50. Effect of deoxidation practice (inclusion chemistry) on the acicular ferrite transformation in
low-alloy steel weld metals. Data compiled from miscellaneous sources.
Solution
From Fig. 2.72 it is seen that the total number of constituent phases in the inclusions is six,
including the three cubic phases 7-Al2O3, MnOAl2O3, and TiN. If we assume a random ori-
entation and only one nucleation event per inclusion, the following upper limit for the acicular
ferrite plate volume is obtained:
The above volume corresponds to an acicular ferrite plate which has the shape of a square
lath of side lOjim and thickness 2|im. Although this estimate is in reasonable agreement with
experimental observations,37 the prediction is conservative in the sense that it assumes only
one nucleation event per inclusion. In practice, an oxide inclusion which is orientated within
the Bain region has the capability of nucleating several acicular ferrite plates, as shown by the
SEM micrograph in Fig. 6.51. In addition, the acicular ferrite plates may nucleate
autocatalytically at aFe /yFe boundaries, a process which also is referred to as sympathetic
nucleation in the literature.36'37'7083 At present, it is not clear to what extent autocatalytic
nucleation plays a role in the development of the acicular ferrite microstructure.
Hardenability effects
Since acicular ferrite is one of the last phases to form after the growth of allotriomorphic and
Widmanstatten ferrite, it is bound to be influenced by the prior transformation products, as
indicated by equation (6-50). The strong dependence of the acicular ferrite content on the
austenite grain size must therefore be understood on this basis. 3747 ' 48 ' 7088 " 90 It is evident from
the data of Barbaro et al.70 reproduced in Fig. 6.52 that a coarse austenite grain size favours
intragranular nucleation of acicular ferrite at the expense of formation of allotriomorphic and
Fig. 6.51. SEM micrograph of a carbon extraction replica showing evidence of multiple nucleation of
acicular ferrite at a weld metal inclusion.
Widmanstatten ferrite. This effect is most pronounced during slow cooling, since the combi-
nation of a small austenite grain size and a slow cooling rate implies that much of the yFe-
phase already has transformed to allotriomorphic ferrite before the temperature for intragranular
nucleation of acicular ferrite is reached.
Example (6.9)
Consider a low-alloy steel weld metal which contains a total number of 5 X 107 inclusions per
mm3 with an average radius of 0.25 Jim. Use this information to evaluate the conditions for
acicular ferrite formation within the as-deposited weld metal and the low-temperature reheated
region of the weld when the austenite grain size is 100 and 10 |iim, respectively. In these
calculations we shall assume that the equilibrium volume fraction of ferrite ff~Fe in the fully
transformed steel is 0.9.
Solution
First we need to estimate the total surface area per unit volume available for ferrite nuclea-
tion at austenite grain boundaries, Sv (GB), and non-metallic inclusions, Sv (/), respectively.
As-deposited weld metal
Reheated weld metal
(as before)
Volume fraction of AF
Austenjte grain size,jim
Fig. 6.52. Effect of austenite grain size on the acicular ferrite transformation in low-alloy steel weld
deposits. Data from Barbara et al.70
From the above calculations it is apparent that the conditions for intragranular nucleation of
acicular ferrite at inclusions are particularly favourable within the as-deposited weld metal,
since SV(I) > SV(GB), whereas nucleation and growth of ferrite at austenite grain boundaries
will dominate within the reheated region of the weld (SV(GB)» SV(I)). This conclusion is also
consistent with predictions based on the Avrami equation (equation (6-49)). If we, as an illus-
tration, assume that the volume fraction of grain boundary ferrite in the as-deposited weld
metal is 0.3, the corresponding fraction of GF in the reheated weld metal becomes:
On this basis it is not surprising to find that the microstructure within the grain refined
region of low-alloy steel welds is usually polygonal ferrite, while the as-deposited weld metal
also contains high proportions of acicular ferrite (see Fig. 6.19(c) and (d), respectively).
It is important to realise, however, that the presence of allotriomorphic ferrite at the austenite
grain boundaries has the benefical effect of suppressing the formation of bainitic sheaves at the
austenite grain boundaries, which, in turn, allows the acicular ferrite to develop on intragranular
nucleation sites.9091 Consequently, due to the number of competing nucleation and growth
processes involved, the volume fraction of acicular ferrite is often seen to pass through a maxi-
mum when the weld metal hardenability is successively increased by additions of alloying
elements, as shown in Fig. 6.53.
Microstructural component, %
GF and PF
WF Fertile with aligned
second phase (FS)
Chromium content, wt%

Fig. 6.53. Effect of chromium additions on the C-Mn primary weld metal microstructure. After Babu
and Bhadeshia.91
6.3.6 Acicular ferrite in wrought steels

In recent years a new class of low-carbon microalloyed steels has emerged to meet the need for
improved weldability, particularly with respect to the HAZ toughness. 70'92 These steels are
not aluminium-killed (they usually contain less than 30 ppm Al), but are instead deoxidised
with titanium (120-140 ppm) to produce a relatively coarse distribution of submicroscopic Ti
oxide inclusions within the base metal (presumably Ti 2 O 3 ). During high heat input welding (>
5 kJ mm"1), the oxides will not retard austenite grain growth but instead act as favourable
nucleation sites for acicular ferrite within the interior of the austenite grains, as shown
schematically in Fig. 6.54(a). However, since the inclusion number density is quite low (<106
particles per mm3), the microstructure is significantly coarser than that normally observed in
steel weld deposites.92 Nevertheless, their presence contributes to suppress the formation of
Widmanstatten ferrite and upper bainite within the grain coarsened HAZ, which is a common
problem with the traditional Al-Ti microalloyed steels22 (see Fig. 6.54(b)).
63.7 Bainite
Bainitic microstructures (besides acicular ferrite) are frequently observed during welding, par-
ticularly in the HAZ of low-carbon microalloyed steels,22'93"95 but also within the fusion re-
gion of the weld if the nucleation conditions are favourable.37'55'5991 Two main forms can be
identified, i.e. upper and lower bainite, as indicated in Fig. 6.19(a).
6.3.7.1 Upper bainite

In general, the morphology of upper bainite bears a close resemblance to Widmanstatten fer-
rite (both are typical sideplate structures), while its crystallography is more like that of low-
carbon martensite (the K-S orientation relationship is usually less precise than for WF).20 In
spite of this similarity, the formation of upper bainite cannot be fully understood in terms of
HAZ
WF/UB
HAZ
Fig. 6.54. Schematic illustration of the HAZ transformation behaviour during high heat input welding;
(a) Ti-oxide containing steel, (b) Conventional Al-Ti microalloyed steel. The diagrams are based on the
ideas of Homma et al.92
the classic theory of martensite nucleation assuming a pure invariant plane strain deforma-
tion.1'20'61
HAZ transformation behaviour

As shown in Fig. 6.19(a), upper bainite is frequently observed within the grain coarsened HAZ
of low-carbon microalloyed steels. The bainite sheaves nucleate preferentially at austenite
grain boundaries and subsequently grow into the interior of the grains until they physically
impinge with other transformation products such as martensite or Widmanstatten ferrite. Be-
cause of these competitive nucleation and growth processes, the highest volume fractions of
upper bainite are normally attained at an intermediate cooling rate, as indicated by the micro-
structure data in Fig. 6.55.
Weld metal transformation behaviour

Upper bainite may also form within the columnar grain region of high heat input SA steel
welds, as shown by the optical micrograph in Fig. 6.56. The indications are that the observed
Ferrite with
iUBj aligned second - WF
dhase (FS)
Mill
Cooling time, At 8/5 , s
Fig. 6.55. Effect of cooling time, Af8/5, on the grain coarsened HAZ transformation behaviour (simulated
thick plate heat cycles, Tp « 13500C). Data from Akselsen et al96
Fig. 6.56. Optical micrograph showing formation of upper bainite within the columnar grain region of a
SA steel weld.
shift in the weld metal transformation behaviour is related to a change in the deoxidation
practice which alters the kinetics of the subsequent solid state transformation reactions through
a modification of the solidification microstructure.64 Solidification induced phosphorus
segregations are of particular interest in this respect, since previous examinations have shown
that phosphorus can strongly enhance the formation of grain boundary ferrite by raising the
local Ae3 temperature of the steel97 (see also Example 6.5).
Following the discussion in Section 6.3.5.4, the evolution of allotriomorphic ferrite at the
austenite grain boundaries has the beneficial effect of suppressing the formation of bainitic
sheaves at these sites, which, in turn, allows the acicular ferrite to develop on intragranular
Carbides
Lower bainite
Martensite
Fig. 6.58. TEM micrograph showing the formation of lower bainite within the HAZ of a low-carbon
microalloyed steel.
6.3.8 Martensite
At very high undercoolings, the austenite decomposes to martensite by means of an invariant
plane strain deformation, which implies that there is no change in the steel chemical composi-
tion. The reaction product will either be lath or plate (twinned) martensite, depending on the
alloy level. Lath martensite is commonly found in plain carbon and low-alloy steels up to
about 0.5wt% C, and is formed by a slip mechanism, as shown schematically in Fig. 6.59(a).
When the carbon content exceeds this threshold, the martensite transformation occurs rather
by formation of deformation twins (Fig. 6.59(b)). The crystal structure of plate martensite is
bet (body-centred tetragonal), while lath martensite reveals a bcc (body-centred cubic) struc-
ture,61 which becomes increasingly distorted with increasing steel carbon contents.20 Both
transformation products exhibit the characteristic Kurdjumow-Sachs orientation relationship
with the austenite, but this relationship tends to be less precise at high carbon levels.1'20
6.3.8.1 Lath martensite

Lath martensite is frequently observed within the grain coarsened HAZ during low heat input
welding of microalloyed steels. As shown in Fig. 6.60, the microstructure will be fully
martensitic for values of Ar875 up to about 5-10s, depending on the steel hardenability. The
hardness of the martensite formed is usually below 400 VPN, which is significantly lower than
the corresponding peak value achieved after water-quenching.98 This means that considerable
autotempering occurs during cooling from the M^-temperature, which in the case of low-car-
bon microalloyed steels can be as high as 5000C.
6.3.8.2 Plate (twinned) martensite

In the intercritical HAZ (i.e. partly transformed region), carbon-rich austenite, formed by de-
composition of pearlite or isolated carbides, may transform to high carbon (twinned) martensite
on cooling, as shown schematically in Fig. 6.61. However, since the Mf temperature in this
case will be far below room temperature, significant amounts of retained austenite may be
present in the areas adjacent to the martensite islands. This transformation product is therefore
Slip Twin
a b
Fig. 6.59. Mechanisms of martensite formation in steel (schematic); (a) Slip along parallel planes, (b)
Generation of deformation twins. The diagram is based on the ideas of Verhoeven.61
Steel A
Steel B
Martensite content, vol%
Steel C
Cooling time, At0._ ,s

8/5
Fig. 6.60. Effect of cooling time, Af8/5, on the volume fraction of lath martensite within the grain coars-
ened HAZ of microalloyed steels (simulated thick plate heat cycles). Data from Akselsen et «/.98
frequently referred to as the martensite-austenite (M-A) constituent in the scientific litera-

ture.99"103
Kinetics of austenite formation

In general, the formation of austenite during intercritical annealing of low-carbon microalloyed
steels can be separated into three main stages:99
(i) Rapid growth of austenite into pearlite until the dissolution process is completed.
(ii) Slower growth of austenite into ferrite at a rate which is either controlled by carbon
diffusion in the austenite or by diffusion of substitutional elements such as manga-
nese in the ferrite, depending on the applied annealing temperature.
T T
t wt% C
Fig. 6.61. Schematic illustration of the formation of plate (twinned) martensite within the intercritical
HAZ of low-carbon microalloyed steels; Heating leg of thermal cycle: Pearlite -> -yFe (>0.5 wt% C),
Cooling leg of thermal cycle: yFe(> 0.5 wt% C) -> a'Fe (twinned martensite).
(iii) Very slow final equilibration of solute concentration gradients within the austenite or
the ferrite through diffusion.
At temperatures above say 7700C, the rate of pearlite decomposition is sufficiently high for
complete dissolution, even during low-heat input welding. Hence, in hot rolled and normal-
ised steels all pearlite will transform to austenite when the peak temperature of the thermal
cycle exceeds this threshold. However, if the starting microstructure is a mixture of martensite
and upper bainite (as often will be the case in multi-pass welding), the austenite will nucleate
both at prior austenite grain boundaries and along the interfaces between laths of bainite or
martensite, as shown by the optical micrographs in Fig. 6.62. A similar pattern has also been
observed during intercritical annealing of dual-phase steels.104'105 Considering the kinetics,
carbides precipitated within autotempered low-carbon martensite or between the laths of bainite
would be expected to dissolve at a rate comparable with that of pearlite.105 Hence, it is reason-
able to assume that no retained carbides will be present intragranularly after reheating to say
Tp « 7700C when the starting microstructure is a mixture of martensite and bainite.
Further growth of the austenite into the ferrite requires, however, that the peak temperature
of the thermal cycle is raised significantly above 7700C in order to reach the kinetic
(paraequilibrium) stage where the reaction is controlled by diffusion of carbon in austenite."
Consequently, the possibilities of obtaining growth of the austenite colonies within the low
peak temperature regions of the intercritical HAZ are strongly limited, which implies that the
carbon content of the austenite islands in these regions should be close to the saturation level
of carbon in austenite at the temperature of dissolution.
Conditions for M-A formation

On subsequent cooling, the carbon-rich austenite can decompose to a variety of microstruc-
tures (ranging from twinned martensite to pearlite) or remain untransformed, depending on the
base plate hardenability, the weld cooling programme, and the initial size of the austenite
islands.102103'106 When the starting microstructure is a mixture of ferrite and pearlite (as fre-
quently observed in normalised steels), twinned martensite forms readily along the prior base
(a)
(b)
Fig. 6.62. Optical micrographs showing favourable sites for austenite formation during two-pass weld
thermal simulation. (First cycle: Tp « 13500C, Atm « 12 s, Second cycle: Tp « 775°C, Atm « 12 s);
(a) Intergranular, (b) Intragranular. After Akselsen et aim
metal pearlite bands or within isolated pearlite colonies at high cooling rates, as shown by the
TEM micrographs in Fig. 6.63(a) and (b). With decreasing cooling rates, however, the trans-
formation product shifts from twinned martensite to predominatly pearlite (see Fig. 6.63(c)
and (d)) in the absence of strong hardenability elements such as molybdenium and boron which
can stabilise the M-A constituent.102'107"109
In controlled rolled and accelerated cooled steels, where the carbides are mainly present in
the form of submicroscopic colonies or isolated particles located at ferrite/ferrite grain boundaries
(see Fig. 6.64(a)), the situation becomes slightly more complex. Under such conditions, the
majority of the austenite islands formed within the low peak temperature region of the
intercritical HAZ will be of a size below 1 jam. It has been verified experimentally that such
(a) (b)
(C) (d)
Fig. 6.63. Examples of transformation products formed after intercritical thermal cycling of a ferritic/
pearlitic starting microstructure to a peak temperature of 775°C; (a) M-A islands (black areas) sur-
rounded by ferrite (At6/4 ~ 5 s), (b) Close-up of twinned martensite within a M-A island (At6/4 ~ 5 s),
(c) Isolated pearlite colony formed at an intermediate cooling rate (Ar674 ~ 12 s), (d) Pearlite colonies
formed during slow cooling (At614 ~ 35 s). After Akselsen et al.103
small particles do not readily transform to martensite, but will largely remain in the steel in the
form of retained austenite.100'103'110'111 This can be explained by the lack of nucleation oppor-
tunities for martensite or by the volume restraint provided by the surrounding ferrite matrix.1 x x
Consequently, twinned martensite in controlled rolled and accelerated cooled steel is seen to
form within a few, relatively large austenite islands which stem from decomposition of single
carbide colonies, as illustrated by the TEM micrograph in Fig. 6.64(b).
During multipass welding, the initial carbide distribution in the base plate will be of less
importance because of the transformations imposed by the heat of previous weld passes. Hence,
after full reaustenitising of the steel the M-A constituent may form both intergranularly and
intragranularly, depending on the starting microstructure, as shown previously in Fig. 6.62(a)
and (b). An example of intergranularly nucleated twinned martensite is contained in Fig. 6.65.
Volume fraction of M-A constituent

At present, an adequate kinetic model for partial reaustenitising of steel during continuous
heating and cooling is lacking. However, in view of the previous discussion it is obvious that
the volume fraction of the M-A constituent within the intercritical HAZ of single pass steel
weldments must be closely related to the initial base plate pearlite content. This is also in
agreement with experimental observations (see Fig. 6.66).
Next Page
(a)
(b)
Fig. 6.64. Conditions for martensite formation within the intercritical HAZ of a controlled rolled and
accelerated cooled steel containing copper and nickel; (a) TEM micrograph of the initial base plate car-
bide distribution, (b) TEM micrograph showing evidence of twinned martensite within a l(im large
austenite colony formed during intercritical thermal cycling to T « 775°C (Ar674 « 35 s). After Akselsen
etal^
6.4 Austenitic Stainless Steels
Simple austenitic stainless steels generally contain between 18 and 30 wt% chromium, 8 to 20
wt% nickel, and between 0.03 to 0.1 wt% carbon.20 Since the solubility of carbon decreases
rapidly with temperature (see Fig. 6.67), reheating of the steels within the temperature range
from 500 to 9000C will lead to the rejection of carbon from solid solution, usually by chro-
mium carbide precipitation (Cr23C6). These carbides nucleate preferentially at the austenite
grain boundaries, which, in turn, results in depletion of the regions adjacent to the grain bounda-
ries with respect to chromium, as shown schematically in Fig. 6.68. The presence of such
chromium-depleted regions within the steel makes it sensitive to intergranular corrosion in
service.20'112
Previous Page
(a)
(b)
Fig. 6.64. Conditions for martensite formation within the intercritical HAZ of a controlled rolled and
accelerated cooled steel containing copper and nickel; (a) TEM micrograph of the initial base plate car-
bide distribution, (b) TEM micrograph showing evidence of twinned martensite within a l(im large
austenite colony formed during intercritical thermal cycling to T « 775°C (Ar674 « 35 s). After Akselsen
etal^
6.4 Austenitic Stainless Steels
Simple austenitic stainless steels generally contain between 18 and 30 wt% chromium, 8 to 20
wt% nickel, and between 0.03 to 0.1 wt% carbon.20 Since the solubility of carbon decreases
rapidly with temperature (see Fig. 6.67), reheating of the steels within the temperature range
from 500 to 9000C will lead to the rejection of carbon from solid solution, usually by chro-
mium carbide precipitation (Cr23C6). These carbides nucleate preferentially at the austenite
grain boundaries, which, in turn, results in depletion of the regions adjacent to the grain bounda-
ries with respect to chromium, as shown schematically in Fig. 6.68. The presence of such
chromium-depleted regions within the steel makes it sensitive to intergranular corrosion in
service.20'112
Fig. 6.65. TEM micrograph showing evidence of intergranularly nucleated twinned martensite in dou-
ble-cycled specimens (First cycle: Tp ~ 13500C, A;6/4 -12s, Second cycle: Tp - 775°C, A;6/4 - 12s) After
Akselsen et al.103
Content of M-A constituent, vol%-
Base plate pearlite content, v o l % •
Fig. 6.66. Effect of base plate pearlite content on the volume fraction of M-A constituent in thermally
cycled specimens (simulation conditions: Tp -775°C, Atm « 5s). Data from Akselsen et al.m
Temperature, 0C
Carbon content, wt%
Fig. 6.67. Solvus temperatures for different types of carbides in 304 austenitic stainless steel as a func-
tion of carbon content. Data from Kou.112
Cr depleted zone Grain boundary
Cr carbide precipitate
Fig. 6.68. Schematic illustration OfCr23C6 precipitation at grain boundaries in austenitic stainless steels.
The diagram is based on the ideas of Kou.112
6.4.1 Kinetics of chromium carbide formation
Precipitation of Cr23C6 in austenitic stainless steels is a typical nucleation and growth process.
The temperature-time transformation curve is therefore C-shaped, and can be modelled ac-
cording to the general principles outlined in Section 6.2.2.3. As an illustration, we shall as-
sume that the parameters listed in Table 6.4 are representative of nucleation of Cr23C6 in a 304
type austenitic stainless steel containing 0.07 wt% C. If we at the same time allow for the
variation in austenite grain size with distance from the fusion boundary, the shape and position
of the C-curve in temperature-time space at different locations within the HAZ is given by
equation (6-35):
(6-54)
A graphical representation of equation (6-54) in Fig. 6.69 shows that the nose of the
C-curve for the reference steel (characterised by Teq = T*eq and D1 = D 7 ) is strongly shifted to
the left in the IT-diagram, thereby providing favourable conditions for Cr23C6 precipitation in
the heat affected zone during welding. The problem becomes less imminent if the base plate
carbon content is reduced from say 0.07 to 0.04 wt%. In that case the associated reduction in
the solvus temperature Teq from 920 to 8000C will displace the C-curve towards longer times
in the diagram which by far exceed the duration of a typical weld thermal cycle, as shown in
Fig. 6.69.
6.4.2 Area of weld decay

Experience shows that corrosion attact within the HAZ of single pass austenitic stainless steel
welds usually starts in regions where the peak temperature of the thermal cycle has been close
to the nose of the C-curve in Fig. 6.69. This observation cannot be explained by just consider-
ing the HAZ temperature-time pattern, since the retention time within the critical temperature
range for chromium carbide precipitation is virtually the same for both high and low peak
temperature thermal cycles, as indicated in Fig. 6.70. The explanation is, of course, that the
C-curve becomes increasingly shifted to the right in the IT-diagram as the peak temperature
increases due to austenite grain growth occurring during welding. This point is more clearly
illustrated below.
Example (6.10)
Consider single pass butt welding of 2mm sheets of 304 austenitic stainless steels with cov-
ered electrodes under the following conditions:
Evaluate on the basis of equations (6-32), (6-33), and (6-54) in combination with the grain
growth diagram in Fig. 5.30(b) (Chapter 5) the conditions for chromium carbide formation
within the HAZ during welding when the base plate carbon content is 0.07 and 0.04 wt% C,
respectively. Thermal data for austenitic stainless steels are given in Table 1.1 (Chapter 1).
Temperature, 0C
Time, s
Fig. 6.69. Effect of carbon content on the isothermal precipitation of C ^ C 6 in 304 austenitic stainless
steels. Broken curve: 0.07 wt% C, Solid curve 0.04 wt% C. The diagrams are constructed on the basis
of equation (6-54).
T
Grain growth Area of weld

zone decay
Fig. 6.70. Schematic illustration of thermal cycles and sensitisation region within the HAZ of an
unstabilised austenitic stainless steel.
Table 6.4 Input data used to construct C-curve for Cr23C6 precipitates in reference steel.
Parameter T tr T*q A0 Q/ £>*
(K) (s) (K) (Jmol- 1 ) (kJmol-1) (jxm)
Value 923 60 1193 150 200 18
[• Activation energy for grain boundary diffusion of Cr in austenite.

Solution
If we, as a first approximation, neglect the contribution from heat flow in the welding direc-
tion, the temperature-time pattern is given by equation (1-100) in Chapter 1. Taking n = 4 and
(1 — Xc) = 0.98 (assuming interface-controlled growth), the extent of chromium carbide pre-
cipitation occurring within the HAZ of the weld can be calculated from equations (6-32) and
(6-33) by numerical integration over the actual thermal cycles. The results of such computa-
tions are presented graphically in Fig. 6.71.
As expected, precipitation of chromium carbides occurs readily within the low peak tem-
perature region of the weld when the carbon content is 0.07 wt%. For points located within
the grain growth zone, the resulting displacement of the C-curve towards longer times in the
CCT-diagram will gradually retard the reaction, which implies that the fraction transformed
starts to decline when the peak temperature of the thermal cycle exceeds about 11500C. In
contrast, the low-carbon steel reveals no sign of chromium carbide precipitation within the
HAZ. This result is in good agreement with general experience, and explains why the carbon
content of modern austenite stainless steels has been gradually lowered to values below 0.03
wt % in step with the progress in the steel manufacturing technology.20112
6.5 Al-Mg-Si Alloys
Age-hardenable Al-Mg-Si alloys are widely used as structural components in welded assem-
blies. They offer tensile strength values higher than 350 MPa in the artificially aged T6 condi-
tion owing to the presence of very fine, needle shaped (3"(Mg2Si) precipitates along <100>
directions in the aluminium matrix.113 Although Al-Mg-Si alloys are readily weldable, they
suffer from severe softening in the heat affected zone (HAZ) because of reversion (dissolu-
tion) of the 3"(Mg2Si) precipitates during the weld thermal cycle. 6112 This type of mechanical
impairment represents a major problem in engineering design.114
Unaffected base metal
Austenite grain size, pm

Grain growth zone
Fig. 6.71. Conditions for Cr23C6 precipitation within the HAZ of a single pass austenitic stainless steel
butt weld (Example 6.10).
6.5.1 Quench-sensitivity in relation to welding
High strength alloys such as AA 6082 contains manganese in addition to magnesium and
silicon. Manganese is added to control recrystallisation and grain growth in the material dur-
ing hot forming. The disadvantage is that it increases the quench sensitivity of the alloy.115
The reason for this is that the Mn-bearing dispersoids (which form during homogenisation)
provide favourable nucleation sites for the non-hardening (3'(Mg2Si) phase, as shown by the
TEM micrograph in Fig. 6.72. The resulting reduction in the solute content leads to a reduced
HAZ strength both in the naturally aged TA and peak aged T6 conditions. 56116
6.5.1.1 Conditions for (3 '(Mg2Si) precipitation during cooling

Myhr and Grong6 have shown how the quench-sensitivity concept can be applied to welding
of Al-Mg-Si alloys. Their model has later been refined and extended to heat treatment of AA
6082 alloys by Shercliff et al.ni and Bratland et a/.118119
As an illustration, we shall assume that the parameters listed in Table 6.5 are representative
of nucleation of (3'(Mg2Si) particles in an AA 6082 alloy homogenised at 5800C. If we further
assume that the P'(Mg2 Si) particles only form at Mn-bearing dispersoids, the (N* /Nv) term
in equation (6-14) may be replaced by the corresponding (5* / Sv) ratio, where S* and Sv refer
to the total surface area per unit volume of dispersoids in the reference and the actual alu-
minium alloy, respectively. This leads to the following expression for t*:
Fig. 6.72. TEM micrograph showing nucleation of non-hardening (3'(Mg2Si) precipitates at Mn-bearing
dispersoids in an AA 6082 alloy.
Table 6.5 Input data used to construct C-curve for p' (Mg2Si) precipitates in reference aluminium alloy.
Parameter Tr t\ T*q. A0 Qj S^
(K) (s) (K) (Jmol"1) (kJmol"1) (mm2 per mm3)
Value 573 20 766 350 130 100
f" Activation energy for diffusion of Mg in aluminium.

(6-55)
A graphical representation of equation (6-55) in Fig. 6.73 shows that the nose of the C-
curve for the reference alloy (characterised by Teq = r and Sv = S^ is strongly shifted to
the left in the IT-diagram, thereby providing favourable conditions for (3'(Mg2Si) formation
dunng welding. In general, an increase in T or Sv will enhance the quench-sensitivity of the
material because of the resulting increase in the nucleation rate. This will be the case if the
alloy contams large amounts of excess silicon in solid solution or is homogenised at a tem-
perature lower than 5800C.118119
Example (6.11)
Consider plasma-MIG butt welding of a 10 mm thick Al-Mg-Si plate under the following
conditions:
qo = 10 kW, v = 10 mm s~\ T0 = 200C
Evaluate on the basis of equations (6-32), (6-33), and (6-55) the conditions for P'(Mg,Si)
precipitation within the high peak temperature region of the HAZ during welding (T7 > T )
In these calculations we shall assume that the transformation behaviour of the base'metafis
similar to that of the reference alloy in Table 6.5. Relevant thermal data for Al-Mg-Si alloys
are given in Table 1.1 (Chapter 1).
Solution
Referring to Fig. 1.43 in Chapter 1, the mode of heat flow becomes essentially one-dimen-
sional if the net arc power is kept sufficiently high compared with the plate thickness (i.e.
Temperature,°C
Time, s
Fig. 6.73. C-curve for precipitation of 3'(Mg2Si) in the reference AA 6082 alloy. The diagram is con-
structed on the basis of equation (6-55).
qold>0.05 kW mm"1). Since this requirement is met in the present case, the HAZ temperature-
time pattern is given by equation (1-100).Taking n = 0.75 and (1 - Xc) = 0.84 for precipitation
of p'(Mg2Si) in AA6082 aluminium alloys,118 we obtain after integration of equations (6-32)
and (6-33) over the weld cooling cycle:
and
The above calculations show that precipitation of (3'(Mg2Si) particles at dispersoids is, in-
deed, a significant process under the prevailing circumstances. Since the mode of heat flow
during single pass butt welding of aluminium plates is essentially one-dimensional, it is possi-
ble to construct general transformation diagrams which give the fraction transformed as a
function of the applied heat input. An example of such a diagram is contained in Fig. 6.74.
6.5.1.2 Strength recovery during natural ageing

Myhr and Grong6 have shown how equation (6-15) can be applied for modelling of the HAZ
room temperature ageing characteristics. Figure 6.75 shows a typical natural ageing curve for
AA 6082 alloys after full solution heat treatment. Due to enhanced precipitation of GP-zones,
the hardness will increase from about 42 VPN to a maximum of 80 VPN after a period of 5 to
7 days. However, this ultimate hardness is significantly reduced if solute is lost during the
weld cooling cycle because of precipitation of non-hardening (3'(Mg2Si) particles at
dispersoids.6117 The relationship between the solute content, C0, and the maximum fraction of
hardening precipitates which form during natural ageing, AXp, can be obtained from a simple
logt
Net heat input ( q /vd), kJ/mm2
Fig. 6.74. Conditions for 0'(Mg2Si) precipitation within the HAZ of single pass AA 6082 butt welds
(Tp>Teq).
Water-quenched
specimens
Hardness (VPN)
5-7 days
Log time
Fig. 6.75. Typical ageing curve for an AA 6082 aluminium alloy at room temperature (schematic).
2-D kinetic (cell) model, assuming that the reaction is interface-controlled. Let r denote the
radius of the growing precipitates (defined in Fig. 6.76(a)). Since we are only interested in the
terminal value of r at a fixed temperature, the time t in the expression for r can be regarded as
constant. Hence, we may write (when C0 » Ca and Cp » C0):
(6-56)
where C4 is a kinetic constant.
If the distribution of the precipitates is approximated by that of a 2-D face-centered cubic
space lattice (see Fig. 6.76(b)), the parameter, AXp, is simply given as:
(6-57)
where rm denotes the maximum particle radius which forms within the system if all alloying
elements are present in solid solution at the onset of the ageing reaction (C0 = C*).
Because of the stoichiometry of the precipitation reaction, C0 and C* in the expression for
AXp may be taken proportional to the magnesium concentration in solid solution. Hence, we
may write:
(6-58)
Equation (6-58) predicts that AXp is a simple power function of CMg.

The magnesium concentration in solid solution can be obtained from electrical resistivity
measurements, and is given by the following equation:6118
(6-59)
where pm is the measured resistivity, pss is the resistivity in the as-quenched condition, while p0
is the corresponding resistivity in the fully annealed condition (i.e. when all Mg and Si are tied
up in precipitates).
A comparison between equation (6-58) and the electrical resistivity data in Fig. 6.77 con-
firms the relevance of this power-law-relationship, although the deviation in certain cases in
admittedly large.
Example (6.12)
Consider plasma-MIG butt welding of a 10 mm thick Al-Mg-Si plate under conditions similar
to those employed in Example 6.11. Estimate on the basis of equation (6-58) the relative
(a) (b)
Concentration
Distance
Fig. 6.76. Simplified 2-D kinetic (cell) model for precipitation of hardening particles in Al-Mg-Si alloys
during natural ageing; (a) Particle/matrix concentration profile, (b) Cell geometry.
AXp
C
Mg /C*Mg
Fig. 6.77. Experimental verification of equation (6-58).
fraction of hardening precipitates which forms within the fully reverted region of the HAZ
after prolonged room temperature ageing.
Solution
Referring to Example (6.11), the relative fraction of (3'(Mg2Si) precipitates which forms at
dispersoids during the weld cooling cycle amounts to:
from which
This gives:
Since the resulting precipitation strength increment is directly proportional to AXp,ul loss
of solute in the form of (3'(Mg2Si) particles during the weld cooling cycle will inevitably lead
to a reduced HAZ strength in the naturally aged (T4) condition.6 We shall return to this ques-
tion in Chapter 7 (Section 7.4.3).
6.5.2 Sub grain evolution during continuous drive friction welding

Continuous drive friction welding is a solid-state joining process that produces coalescence by
the heat developed between two surfaces by mechanically induced rubbing motion. When the
appropriate rotation velocity is reached, the two workpieces are brought together and an axial
force is applied. The two surfaces are held under pressure, and due to the heat developed, a
plasticised layer forms at the interface. After a predetermined time the rotation stops and the
pressure is increased to facilitate forging or local upsetting of the heated metal. Filler metal,
flux, or shielding gas is not required with this process.
The structural changes within the fully plasticised region of friction welded Al-Mg-Si
alloys arise from the combined action of work hardening and softening due to dynamic recovery.
When the steady state conditions for deformation in friction welding are reached, the rate of
dislocation generation in the plasticised material balances the dislocation annihilation rate,
which means that large strains can be imposed without any changes in theflowstress or subgrain
size. Referring to Fig. 6.78, the subgrain structure within the fully plasticised region consists
of sheaves of virtually equiaxed grains (2-3 |jim in size) that are separated from each other by
low-angle grain boundaries. This type of microstructure is characteristic of hot worked alu-
minium alloys.120
In general, the Zener-Hollomon (Zh) parameter provides a basis for evaluating the evolu-
tion of the subgrain structure during hot working. As shown by McQueen and Jonas,121 the
following relationship exists between the subgrain diameter ds and the Z^-parameter in
aluminium alloys:
(6-60)
Partly deformed Fully plasticized Partly deformed
region region region
(a)
(b)
Fig. 6.78. Micrographs showing the subgrain structure within the fully plasticised and partly deformed
region of a friction welded Al-Mg-Si alloy; (a) Overview, (b) Close-up of the subgrain structure (EBSP
image).
where
(6-61)
The peak temperature T and the strain rate i distributions within the fully plasticised re-
gion of a friction weld can be computed on the basis of the generic models developed by
Midling and Grong.122 Examples of such calculations are shown in Fig. 6.79(a). Plots of the re-
sulting Zener-Hollomon parameter and subgrain diameter at different locations within the
HAZ are contained in Fig. 6.79(b).
Fully plasticized region Partly deformed region
Radial position:
Strain rate, s"1
Axial distance, mm
l
Zener-Hollomon parameter, s
Subqrain diameter, urn

Radial position:
Fully plasticized region Partly deformed region
Axial distance, mm
Fig. 6.79. Modelling of the subgrain evolution in the fully plasticised region of a friction welded AA
6082 aluminium alloy; (a) Predicted peak temperature Tp and strain rate (e) distributions, (b) Plots of
the Zener-Hollomon Zh parameter and resulting subgrain diameter ds at different locations within
the HAZ. Data from Midling and Grong.122
It is evident from the graphs that the value of the Zener-Hollomon parameter is of the order
of 1014-1012 s"1 within the fully plasticised region of the HAZ during continuous drive friction
welding of Al-Mg-Si alloys. The predicted range in Zh corresponds to a subgrain size of 2-3
jam, in agreement with experimental observations (see EBSP image in Fig. 6.78(b)). Outside
the fully plasticised region the Zener-Hollomon parameter drops from about 1012 to 1010 s"1
due to a sudden change the strain rate from 102 to 10° s"1 as the contribution from the material
flow field in the rotational direction ceases. This value is outside the validity range of equation
(6-60), and leads to the formation of coarse subgrains at the boundary between the fully plas-
ticised and the partly deformed region, as shown by the TEM micrographs in Fig. 6.80.
Fig. 6.80. TEM micrographs showing evidence of coarse subgrains at the boundary between the fully
plasticised and partly deformed region of a friction welded AA 6082 aluminium alloy.
References
(England), Van Nostrand Reinhold.
2. M.E. Fine: Phase Transformations in Condensed Systems, 1964, New York, The Macmillian Com-
pany.
3. N. Ryum: Compendium in Physical Metallurgy, 1973, Trondheim (Norway), The University of
Trondheim, The Norwegian Institute of Technology.
4. G.F. Vander Voort (ed.): Atlas of Time-Temperature Diagrams for Nonferrous Alloys, 1991, Ma-
terials Park (Ohio), ASM International.
5. J.T. Staley: Mater. ScL TechnoL, 1987, 3, 923-935.
6. O.R. Myhr and 0. Grong: Acta Metall Mater., 1991, 39, 2693-2702; ibid., 2703-2708.
7. R.D. Doherty: In Physical metallurgy, 3rd Edn (Eds R.W. Cahn and P. Haasen), 1983, Amster-
dam, 934-1030, North-Holland Physics Publishing.
8. H.B. Aaron, D. Fainstein and G.R. Kotler: J. Appl. Phys., 1970, 41, 4404-4410.
9. R. Trivedi: Metall. Trans., 1970, 1, 921-927.
10. J. W. Christian: The Theory of Phase Transformations in Metals and Alloys -Part I, 1975, Oxford
(England), Pergamon Press.
11. Atlas of isothermal transformation and cooling transformation diagrams, 1977, Materials Park
(Ohio), ASM International.
12. E. Scheil: Arch. Eisenhuttenwes., 1934/35, 8, 565-571.
13. J.S. Kirkaldy: Metall. Trans., 1973, 4, 2327-2333.
14. M.B. Kuban, R. Jayaraman, E.B. Hawbolt and J.K. Brimacombe: Metall Trans., 1986, 17A,
1493-1503.
15. LA. Wierszykkowski: Metall Trans., 1991, 22A, 993-999.
16. R.G. Kamat, E.B. Hawbolt, L.C. Brown and J.K. Brimacombe: Metall. Trans., 1992, 23A, 2469-
2480.
17. M. Avrami: J. Chem. Phys., 1939, 7, 1103-1112; ibid, 1940, 8, 212-224; ibid, 1941, 9, 177-184.
18. J.W. Chan: Acta Metall, 1956, 4, 572-575.
19. M. Umemoto, Z.H. Guo and I. Tamura: Mat. ScL Technol, 1987, 3, 249-255.
20. R.W.K. Honeycombe: Steels — Microstructure and Properties, 1980, London, Edward Arnold
(Publishers) Ltd.
21. H. Suzuki: Weld. World, 1982, 20, 121-148.
22. 0. Grong and O.M. Akselsen: Metal Construction, 1986, 18, 557-562.
23. R.C. Cochrane: HW Doc. IXJ-102-85 (1985).
24. DJ. Abson and R.E. Dolby: HW Doc. IXJ-29-80 (1980).
25. CA. Dube, H.I. Aaronson and R.F. Mehl: Rev. Met, 1958, 55, 201-210.
26. H.K.D.H. Bhadeshia: Progress in Mater. ScL, 1985, 29, 321-386.
27. E.C. Bain: Trans. Am. Inst. Min. Eng., 1924, 70, 25-46.
28. S.S. Babu and H.K.D.H. Bhadeshia: Mater. ScL Eng., 1991, A142, 209-219.
29. H.K.D.H. Bhadeshia and D.V. Edmonds: Acta Metall, 1980, 28, 1265-1273.
30. W.C. Leslie: The Physical Metallurgy of Steels, 1981, London (England), McGraw-Hill.
31. CE. Cross, 0 . Grong, S. Liu and J.F. Capes: In Applied Metallography, (Ed. G.F. Vander Voort),
1986, New York, Van Nostrand Reinhold, 197-210.
32. J.E. Morral and T.B. Cameron: Proc. Int. Symp. on Boron in Steels, Milwaukee (Wisconsin),
September, 1979, 19-32, Publ. The Metall. Soc. AIME (1980).
33. S. Pakrasi, E. Just, J. Betzold and F. Hollrigl-Rosta: ibid, pp. 147-164.
34. L. F. Porter: ibid., pp. 199-211.
35. J.H. Devletian: Weld. J., 1976, 55, 5s-12s.
36. 0. Grong and D.K. Matlock: Int. Met. Rev., 1986, 31, 27-48.
37. H.K.D.H. Bhadeshia and L.E. Svensson: In Mathematical Modelling of Weld Phenomena, (Eds
H. Cerjack and K.E. Easterling), 1993, London, The Institute of Materials, 109-180.
38. E.D. Hondros and M.P. Seah: In Physical metallurgy, (Eds R.W. Cahn and P. Haasen), 1983,
Amsterdam, North Holland Physics Publishing B.V., 855-931.
39. T.M. Williams, A.M. Stoneham and D.R. Harris: Met. ScL J., 1976,10, 14-19.
40. R.W. Fountain and J. Chipman: Trans. AIME, 1962, 224, 599-606.
41. Ph. Maitrepierre, D. Thivellier and R. Tricot: Metall Trans., 1975, 5A, 287-301.
42. O.M. Akselsen, 0 . Grong and PE. Kvaale: Metall Trans., 1986,17A, 1529-1536.
43. CJ. McMahon: Metall. Trans., 1980, HA, 531-535.
44. FA. Jackobs and G. Krauss: J. Heat Treating, 1981, 2, 139-146.
45. G.M. Evans: HW Doc. II-A-469-79 (1979).
46. H.K.D.H. Bhadeshia, L.E. Svensson and B. Gretoft: Acta Metall., 1985, 33, 1271-1283.
47. S. Liu and D.L. Olson: Weld. J., 1986, 65, 139s-15Os.
48. N.A. Fleck, 0. Grong, G.R. Edwards and D.K. Matlock: Weld. J., 1986, 65, 113s-121s.
49. H.K.D.H. Bhadeshia, L.E. Svensson and B. Gretoft: J. Mat. ScL, 1986, 21, 3947-3951.
50. 0. Grong, T.A. Siewert and G.R. Edwards: Weld. J., 1986, 65, 279s-288s.
51. H.K.D.H. Bhadeshia: Mater. ScL Forum, 1990, 56-58, 263-274.
52. A. AIi and H.K.D.H. Bhadeshia: Mater. ScL Technol., 1990, 6, 781-784.
53. DJ. Abson and RJ. Pargeter: Int. Met. Rev., 1986, 31, 141-194.
54. J.R. Yang and H.K.D.H. Bhadeshia: Weld. /., 1990, 69, 305-307.
55. RL. Harrison and R.A. Ferrar: Int. Mater. Rev., 1989, 34, 35-51.
56. R.C. Cochrane: Weld, in the World, 1983, 21, 16-29.
57. J.M. Dowling, J.M. Corbett and H.W. Kerr: Metall. Trans., 1986,17A, 1611-1623.
58. DJ. Abson: Weld, in the World, 1989, 27, 11-28.
59. M. Strangwood and H.K.D.H. Bhadeshia: Proc. Int. Conf. on Advances in Welding Science and
Technology, Gatlinburg, TN, May 1986, pp. 209-213. Publ. ASM INTERNATIONAL, Metals
Park, OH, 1986.
60. R.F. Hehemann, K.R. Kinsman and H.I. Aaronson: Metall. Trans., 1972, 3, 1077-1094.
61. J.D. Verhoeven: Fundamentals of Physical Metallurgy, 1975, New York, John Wiley & Sons.
62. H.K.D.H. Bhadeshia and J.W. Christian: Metall. Trans., 1990, 21A, 767-797.
63. A.O. Kluken, 0. Grong and J. Hjelen: Metall. Trans., 1991, 22A, 657-663.
64. A.O. Kluken, 0 . Grong and G. R0rvik: Metall. Trans., 1990, 21A, 2047-2058.
65. H. Hu: Trans. AIME, 1965, 233, 1071-1075.
66. A.A.B. Sugden and H.K.D.H. Bhadeshia: Metall. Trans., 1989, 2OA, 1811-1818.
67. P. Deb, K.D. Challenger and A.E. Therrien: Metall. Trans., 1987,18A, 987-999.
68. H.K.D.H. Bhadeshia and L.E. Svensson: J. Mater. ScL, 1989, 24, 3180-3188.
69. R.A. Ricks, G.S. Barritte and PR. Howell: Proc. Int. Conf. on Solid-Solid Phase Transformations,
Pittsburg (Pennsylvania), August 1981, pp. 463-468. Publ. The Metall. Soc. AIME (1982).
70. FJ. Barbaro, P. Krauklis and K.E. Easterling: Mater. ScL Technol., 1989, 5, 1057-1068.
71. S. Liu: In Ferrous Alloy Weldments, (Eds D.L. Olson and T.H. North), 1992, Zurich (Switzer-
land), Trans. Tech. Publications, pp. 1-18.
72. 0. Grong and A.O. Kluken: ibid., pp. 21 -46.
73. L. Devillers, D. Kaplan, B. Marandet, A. Ribes and PV. Ribound: Proc. Int. Conf on Effects of
Residual, Impurity and Alloying Elements on Weldability and Weld Properties, London, Novem-
ber 1983, Paper 1, Publ. The Welding Institute (England).
74. C. Bonnet and F.P. Charpentier: ibid., Paper 8.
75. M.E. Saggese, D.N. Hawkins and J.A. Whiteman: ibid., Paper 15.
76. A.R. Mills, G. Thewlis and J.A. Whiteman: Mater. ScL Technol., 1987, 3, 1051-1062.
77. R.A. Farrar and PL. Harrison: J. Mat. ScL, 1987, 22, 3812-3820.
78. G.M. Evans: Weld. J., 1992, 71, 447s-454s.
79. M. Es-Souni, PA. Beaven and G.M. Evans: Mater. ScL Eng., 1990, A130, 173-184.
80. S. Liu and D.L. Olson: /. Mater. Eng., 1987, 9, 237-251.
81. S. Liu: PhD thesis, 1984, Dept. of Metall. Eng., Colorado School of Mines, Golden, Colorado
(USA).
82. 0 . Grong and A.O. Kluken: Unpublished work.
83. R.A. Ricks, PR. Howell and G.S. Barritte: J. Mat. Sci, 1982,17, 732-740.
84. PL. Ryder, W. Pitsch and R.F. Mehl; Acta Metall., 1967,15, 1431-1440.
85. 0 . Grong, A.O. Kluken, H.K. Nylund, A.L. Dons and J. Hjelen: Metall. Mater. Trans., 1995,26A,
525-534.
86. WC. Johnson, C. L. White, RE. Marth, RK. Ruf, S.M. Tuominen, K.D. Wade, K.C. Russell and
H.I. Aaronson: Metall Trans., 1975,6A, 911-919.
87. RE. Marth, H.I. Aaronson, G.W. Lorimer, T.L. Bartel and K.C. Russell: Metall. Trans., 1976, 7A,
1519-1528.
88. CM. Dallam and D.L. Olson: Weld. J., 1989, 68, 198s-2O5s.
89. G. Thewlis: HW Doc. IXJ-165-90 (1990).
90. R.A. Farrar, Z. Zhang, S.R. Bannister and G.S. Barrite: J. Mater. ScL, 1993, 28, 1385-1390.
91. S.S. Babu and H.K.D.H. Bhadeshia: Mater. ScL Technol, 1990, 6, 1005-1020.
92. H. Homma, S. Ohkita, S. Matsuda and K. Yamamoto: Weld. J., 1987, 66, 301s-309s.
93. R.E. Dolby: Weld. J., 1979, 58, 225s-238s.
94. B.C. Kim, S. Lee, NJ. Kim and D.Y. Lee: Metall Trans., 1991, 22A, 139-149.
95. R.H. Phillip: Weld. J., 1983, 62, 12s-18s.
96. O.M. Akselsen, J.K. Solberg, G. R0rvik and AJ. Paauw: Technical Report STF34 F87013, 1987,
Trondheim (Norway), Sintef-Division of Metallurgy.
97. A.O. Kluken and 0. Grong: Proc. Int. Conf. on Recent Trends in Welding Science and Technology,
Gatlinburg, TN, May, 1989, pp. 781-786. Publ. ASM International, Materials Park, OH, 1990.
98. O.M. Akselsen, G. R0rvik, M.I. Ons0ien and 0. Grong: Weld. J., 1989, 68, 356s-362s.
99. G.R. Speich, V.A. Demarest and R.L. Miller: Metall. Trans., 1981,12A, 1419-1428.
100. C.A.N. Lanzillotto and FB. Pickering: Metal ScL, 1982, 16, 371-382.
101. J.H. Chen, Y. Kikuta, T. Araki, M. Yoneda and Y. Matsuda: Acta Metall, 1984, 32, 1779-1788.
102. M. Ramberg, O.M. Akselsen and 0 . Grong: Proc. Int. Conf. on Advances in Welding Science and
Technology, Gatlinburg, TN, May, 1986, pp. 679-684. Publ. ASM International, Materials Park,
OH, 1986.
103. O.M. Akselsen, 0. Grong and J.K. Solberg: Mater. ScL Technol, 1987, 3, 649-655.
104. X, -L. Cai, AJ. Garratt-Reed and W.S. Owen: Metall. Trans., 1985, 16A, 543-557.
105. JJ. YI, LS. Kim and H.S. Choi: Metall. Trans., 1985, 16A, 1237-1245.
106. O.M. Akselsen, J.K. Solberg and 0 . Grong: Stand. J. Metall, 1988,17, 194-200.
107. A.R. Marder: Metall. Trans., 1981, 12A, 1569-1579.
108. X.P Shen and R. Priestner: Metall Trans., 1990, 21 A, 2547-2553.
109. NJ. Kim and YG. Kim: Mater. ScL Eng., 1990, A129, 35-44.
110. K.R. Kinsman, G. Das and R.F Hehemann: Acta Metall, 1977, 25, 359-365.
111. N.K. Balliger andT. Gladman: Met. ScL, 1981, 15, 95-108.
112. S. Kou: Welding Metallurgy, 1987, Toronto (Canada), John Wiley & Sons.
113. J.E. Hatch (ed.): Aluminium — Properties and Physical Metallurgy, 1984, Metals Park (Ohio),
114. FM. Mazzolani: Aluminium Alloy Structures, 1985, Boston (USA), Pitman Publish. Inc.
115. O. Lohne and A.L. Dons: Scand. J. Metall, 1983, 12, 34-36.
116. R.P Martukanitz: Proc. Int. Conf. on Advances in Welding Science and Technol, Gatlinburg
(Tenn.), May, 1986, pp. 193-201. Publ. ASM International, Materials Park (OH).
117. H.R. Shercliff, 0. Grong, O.R. Myhr and M.F Ashby: Proc. 3rd. Int. Conf. on Aluminium Alloys
—Their Physical and Mechanical Properties, Trondheim (Norway), June, 1992, Vol. Ill, pp. 357-
369. Publ. The Norwegian Institute of Technology, Trondheim, Norway.
118. D.H. Bratland, 0. Grong, O.R. Myhr and H.R. Shercliff: Proc. Int. Conf on Computer-Assisted
Materials Design and Process Simulation, Tokyo (Japan), September, 1993, pp. 135-141. Publ.
The Iron and Steel Institute of Japan.
119. D.H. Bratland, 0 . Grong, H.R. Shercliff, O.R. Myhr and S.Tjotta:,4cta Metall Mater., Overview
No. 124,1997,45,1-22.
120. CM. Sellars: Proc. 3rd Int. Conf. on Aluminium Alloys—Their Physical and Mechanical
Properties, Trondheim (Norway), June, Vol. Ill, pp. 89-106. Publ. The Norwegian Institute of
Technology, Trondheim, Norway.
121. HJ. McQueen and JJ. Jonas: Treatise on Mater. ScL Technol, 1975,6, 393-493.
122. O.T. Midling and 0 . Grong: Acta Metall Mater., 1994,42,1595-1609: ibid., 1611-1622.
Appendix 6.1
Nomenclature
start temperature of ferrite to concentration of solute at parti-

austenite transformation (0C or cle/matrix interface (at.% or
K) wt%)
end temperature of ferrite to concentration of solute inside the

austenite transformation (0C or precipitates (at.% or wt%)
K)
concentration of solute in matrix
equilibrium 7/a solvus tempera- (at.% or wt%)
ture (0C or K)
maximum concentration of sol-
interfacial area per unit volume
ute in matrix (at.% or wt%)
(jLim2 per |um3 or m2 per m3)
continuous cooling transforma-
material constant related to the tion
potency of the heterogeneous
nucleation sites (J mol"1) cooling rate (0C s"1 or K s 1 )
difference between total and acid plate thickness (mm)

soluble aluminium in weld metal
(wt%)
subgrain diameter (jum or m)
acicular ferrite
average diameter of columnar
integration parameter austenite grains (\xm or m)
bulk concentration of free boron

ferrite grain size (\xm or m)
in steel (ppm)
amount of boron which segre- austenite grain size (jum or m)

gates to austenite grain bounda-
ries on cooling (ppm)
austenite grain size in reference
material (jum or m)
amount of boron which segre-
gates to austenite grain bounda-
element diffusivity (JLim2 s"1 or
ries in reference steel on cool- !!TV1)
ing (ppm)
complementary error function
fraction of boron atoms associ-
ated with vacancies (ppm) vacancy-boron binding energy (J
mol"1 or kJ mol"1)
various kinetic constants and
temperature-dependent param- vacancy formation energy (J
eters mor 1 or kJ mol"1)
electron backscattering pattern arbitrary function of (T)
electron dispersive system amperage (A)
volume fraction of transforma- kinetic strength of thermal cycle

tion product with respect to p-precipitation
equilibrium volume fraction of modified Z1 integral

transformation product
integral in equation (6-47)
ferrite with aligned second phase
molar free energy of a-phase (J kinetic constant in Avrami equa-

mol"1 or J m"3) tion (variable units)
molar free energy of p-phase (J time constant in Avrami equation

mol"1 or J nr 3 ) (s)
arbitrary function of X Kurdjumow-Sachs orientation

relationship
molar free energy of transforma-
tion (J mol"1 or J m~3) isothermal transformation
free energy donated to the sys- interledge spacing (nm)

tem when nucleation occurs het-
erogeneously (J mol"1 or J nr 3 ) length of columnar austenite
grains (|Lim or m)
increase in strain energy
resulting from lattice distortion lower bainite
(J mol 1 or J m 3 )
proportionality constant in boron
increase in surface energy due to diffusion model
nucleus formation (J mol"1 or J
ITT3)
martensite
volume free energy change as- mobility parameters related to

sociated with transformation (J the jump-frequency of atoms
mol"1 or J m 3 ) across a particle/matrix interface
(variable units)
energy barrier against heteroge- end temperature of martensite

neous nucleation (J or J mol"1) transformation (0C or K)
energy barrier against homoge- start temperature of martensite

transformation (0C or K)
neous nucleation (J or J mol"1)
martensite-austenite constituent
grain boundary ferrite
metal inert gas
gas metal arc welding
time exponent in Avrami equa- two-dimensional radius vector in
tion y-z plane (mm)
number of nucleation sites per

unit grain boundary area (irr 2 or isothermal zone width referred
finr 2 ) to the fusion boundary (mm)
universal gas constant (8.314 J

Avogadro constant (6.022 X mol-1 K"1)
1023 moH)
grain boundary or particle sur-
rate of heterogeneous nucleation face area per unit volume (m2
(m-3 s-1) per m3 or mm2 per mm 3 )
rate of grain boundary nuclea- shape factor related to the

tion (rrr2 s"1 or CITT2 s"1) wetting conditions at nucleus/
substrate interface
total number of heterogeneous submerged arc welding

nucleation sites per unit volume
(rrr3) shielded metal arc welding
total number of heterogeneous scanning transmission electron

nucleation sites per unit volume microscope
in reference material (m~3)
time (s)
Nishiyama-Wasserman orienta-
tion relationship time taken to precipitate a certain
fraction of P at an arbitrary tempera-
ture T (s)
analytical weld metal oxygen
content (wt%)
time taken to precipitate a cer-
tain fraction of (3 at a chosen ref-
probability
erence temperature Tr (s)
polygonal ferrite integration limits (s)
net arc power (W or kW) cooling time from 600 to 4000C

(S)
activation energy for diffusion (J

cooling time from 800 to 5000C
mol"1 or kJ mol 1 )
(S)
radius of precipitates (nm, Jim or retention time (s)

m)
temperature (0C or K)
maximum particle radius follow-
ing prolonged room temperature undercooling or temperature in-
ageing (nm, fim or m) terval (0C or K)
equilibrium solvus temperature integration parameter
(0C or K)
half thickness of plate (jam, m)
ambient temperature (0C or K)
Zener-Hollomon parameter (s"1)
0
peak temperature ( C or K)
frequency factor (s"1)
chosen reference temperature
(0C or K) interfacial energy between nu-
cleus and matrix (J irr 2 )
naturally aged condition
wetting angle (degrees)
peak-aged condition
non-hardening (metastable) pre-
transmission electron micro- cipitates in Al-Mg-Si alloys
scope
hardening (metastable) precipi-
voltage (V) tates in Al-Mg-Si alloys
migration rate of incoherent

strain rate (s~!)
interface (|im s"1)
migration rate of semi-coherent parabolic thickening constant for

interface (jam s"1) plate-like precipitates (jam s~1/2
or m s-1/2)
upper bainite
average value of E1 within a
welding speed (mm s" ) 1 specific temperature range
(|im s~1/2 or m s"1/2)
volume (|iim3, mm3, or m3)
parabolic thickening constant for
Vickers pyramid number spherical precipitates (|im s~1/2
or m s"1/2)
Widmanstatten ferrite
arc efficiency
fraction transformed (corrected
for physical impingement of ad- solid angle enclosed by one
jacent volumes) <100>7_Fe-pole in a standard
stereographic projection of the
fraction transformed (not cor- Bain region
rected for physical impingement
of adjacent volumes) measured electrical resistivity
(nQm)
fraction transformed defined by
the C-curve electrical resistivity in solute-
depleted matrix (nQm)
maximum fraction of hardening
precipitates which form during electrical resistivity in as-
natural ageing quenched condition (nQm)
Appendix 6.2
Additivity in relation to the Avrami Equation
In order to prove that equation (6-32) is the isokinetic version of the Avrami equation, we first
need to rewrite it in a differential form. From equation (6-25), we have:
(A6-1)
Substituting
(A6-2)
(A6-3)
(A6-4)
into equation (A6-1) then gives:
(A6-5)
Provided that equation (A6-5) contains separable variables of X and t* (T), it can be inte-
grated as follows:
(A6-6)
Taking b = 1 - XIn, y = 1 - X and dy = - dX, we get:

Rearranging equation (A6-7), gives:
(A6-8)
where
(A6-9)
Equation (A6-8) is identical with equation (6-32) in the text.

7
Properties of Weldments
7.1 Introduction
Weldments are prime examples of components where the properties obtained depend upon the
characteristics of the microstmcture. Since failure of welds often can have dramatic conse-
quences, a wealth of information is available in the literature on structure-property relation-
ships. However, in order to fit some of the apparently conflicting results into a more consistent
picture, a theoretical approach is adopted here rather than a review of the literature. This
procedure also involves the use of phenomenological models for the quantitative description
of structure-property relationships in cases where a full physical treatment is not possible.
7.2 Low-Alloy Steel Weldments
The symbols and units used throughout the chapter are defined in Appendix 7.1.
The major impetus for developments in high-strength low-alloy (HSLA) steels has been
provided by the need for: (i) higher strength, (ii) improved toughness, ductility, and formability,
and (iii) increased weldability. In order to meet these contradictory requirements, the steel
carbon contents have been progressively lowered to below 0.10 wt% C. The desired strength
is largely achieved through a refinement of the ferrite grain size, produced by the additions of
microalloying elements such as aluminium, vanadium, niobium, and titanium in combination
with various forms for thermomechanical processing.1 This procedure has made it possible to
improve the resistance of steels to hydrogen-assisted cold cracking, stress corrosion cracking,
and brittle fracture initiation in the weld heat-affected zone (HAZ) region, without sacrificing
base metal strength, ductility, or low-temperature toughness.2 Controlled rolled HSLA steels
are currently produced with a minimum yield strength in the range from 350-550 MPa. Above
this strength level, quenched and tempered steels are commonly employed.
72.1 Weld metal mechanical properties

The recent progress in steel plate manufacturing technology has, in turn, called for new devel-
opments in welding consumables to produce weld metal deposits with mechanical properties
essentially equivalent to the base metal.3 From the large volume of literature dealing with
HSLA steel filler metals, it appears that the bulk of weld metal research over the past decade
has been concentrated on the achievement of a maximum toughness and ductility for a given
strength level by control of the weld metal microstructure.34 There seems to be general agree-
ment that microstructures primarily consisting of acicular ferrite provide optimum weld metal
mechanical properties, both from a strength and toughness point of view, by virtue of its high
dislocation density and small lath size. The formation of large proportions of upper bainite,
Widmanstatten ferrite, or grain boundary ferrite, on the other hand, are considered detrimental
to toughness, since these structures provide preferential crack propagation paths, especially
when continuous films of carbides are present between the ferrite laths or plates. Attempts to
control the weld metal acicular ferrite content have led to the introduction of welding
consumables containing complex deoxidisers (Si, Mn, Al, Ti) and balanced additions of vari-
ous alloying elements (Nb, V, Cu, Ni, Cr, Mo, B).
The final weld metal microstructure will depend on complex interactions between several
important variables such as:3"5
(i) The total alloy content.

(ii) The concentration, chemical composition, and size distribution of non-metallic
inclusions.
(iii) The solidification microstructure.
(iv) The prior austenite grain size,
(v) The weld thermal cycle.
Although the microstructural changes taking place within the weld metal on cooling through
the critical transformation temperature range in principle are the same as those occurring dur-
ing rolling and heat treatment of steel, the conditions existing in welding are significantly
different from those employed in steel production because of the characteristic strong non-
isothermal behaviour of the arc welding process. For example, in steel weld deposits the
volume fraction of non-metallic inclusions is considerably higher than that in normal cast steel
products because of the limited time available for growth and separation of the particles. Oxy-
gen is of particular interest in this respect, since a high number of oxide inclusions is known to
influence strongly the austenite to ferrite transformation both by restricting the growth of the
austenite grains as well as by providing favourable nucleation sites for various types of
microstructural constituents (e.g. acicular ferrite). Moreover, during solidification of the weld
metal, alloying and impurity elements tend to segregate extensively to the centre parts of the
interdendritic or intercellular spaces under the conditions of rapid cooling.67 The existence of
extensive segregations further alters the kinetics of the subsequent solid state transformation
reactions. Accordingly, the weld metal transformation behaviour is seen to be quite different
from that of the base metal, even when the nominal chemical composition has not been signifi-
cantly changed by the welding process.3"5 This, in turn, will affect the mechanical integrity of
the weldment.
7.2.1.1 Weld metal strength level

In low-alloy steel weld metals there are at least four different strengthening mechanisms which
may contribute to the final strength achieved. These are:
(i) Solid solution strengthening,

(ii) Dislocation strengthening,
(iii) Grain boundary strengthening,
(iv) Precipitation strengthening.
The relative contribution from each is determined by the steel chemical composition and
the weld thermal history. Because of the number of variables involved, a full physical treat-
ment of the problem is not possible. Consequently, the simplified treatment of Gladman and
Pickering8 has been adopted here.
Figure 7.1 shows the individual strength contributions in low-carbon bainite, which is the
dominating microconstituent in as-deposited steel weld metals (includes both upper and lower
bainite as well as acicular ferrite). Firstly, there are the solid solution strengthening increments
from alloying and impurity elements such as manganese, silicon and uncombined nitrogen,
which in the present example correspond to a matrix strength of about 165 MPa. Secondly, the
grain size contribution to the yield stress is shown as a very substantial component, the magni-
tude of which is determined by the bainite lath size. Finally, a typical increment for dispersion
strengthening is indicated. This contribution is negligible at large lath sizes typical of upper
bainite, but becomes significant at small grain sizes because of a finer intralath carbide disper-
sion.8 Hence, in steel weld deposits containing high proportions of acicular ferrite or lower
bainite carbides will make a direct contribution to strength, even at relatively low carbon lev-
els.
The results in Fig. 7.1 are of significant practical importance, since they show the inherent
limitations of the system with regard to the maximum strength that can be achieved through
control of the microstructure. As shown in Section 6.3.5.4 (Chapter 6), the typical lath size
(width) of acicular ferrite in low-alloy steel weld metals is about 2 jam. According to Fig. 7.1,
this corresponds to a maximum yield strength of approximately 650 MPa, which is in good
agreement with the observed threshold strength of acicular ferrite containing steel weld depos-
its.9 If higher strength levels are required, it is necessary to decrease the grain (lath) size
through a refinement of the microstructure, i.e. by replacing acicular ferrite with either lower
p 1/4 -f/o
Number of carbides per mm Nv (mm" )
Yield strength, MPa
Dispersion
Grain size
Matrix strength
-1/2
Bainitic ferrite grain size, mm
Fig. 7.1. Contributions to strength in low-carbon bainite. Data from Gladman and Pickering.8
bainite or martensite. Development along these lines has led to the introduction of a new
generation of high strength steel weld metals with a yield strength in the range from 650 to 900
MPa.10'11
7.2.1.2 Weld metal resistance to ductile fracture

It is well established that the weld metal resistance to ductile fracture is strongly influenced by
the volume fraction, shape, and size distribution of non-metallic inclusions.12"15 Although a
verified quantitative understanding of the fracture process is still lacking, there seems to be
general agreement that it involves the following three basic steps:16
(i) Nucleation of internal cavities during plastic flow, preferentially at non-metallic

inclusions.
(ii) Growth of these cavities with continued deformation,
(iii) Final coalescence of the cavities to produce complete rupture.
Details of these three stages may vary widely in different materials and with the state of
stress existing during deformation. Similarly, the fractographic appearance of the final frac-
ture surface is also influenced by the same factors.
Effect of inclusion volume fraction

The primary variables affecting the true strain at fracture 8/ are the inclusion diameter dv, and
the inclusion volume fraction Vv. The relation between Ef and Vv has been determined experi-
mentally for a wide variety of materials, and can most simply be expressed as:17
(7-1)
where c\ is an empirical constant.
The tensile test data of Widgery12 reproduced in Fig. 7.2 reveal a strong dependence of £/
on Vv, but the relationship appears to be linear rather than non-linear, as predicted by equation
(7-1). Due to a similar fracture mechanism, a correlation also exists between the Charpy V-
notch (CVN) upper shelf energy and the true fracture strain in tensile testing, as shown in Fig.
7.3. For this reason, the weld metal impact properties are normally seen to decrease with
increasing oxygen concentrations when testing is performed in the upper shelf region. From
Fig. 7.4 we see that the CVN upper shelf energy is a linear function of the weld metal oxygen
content. This observation is not surprising, considering the fact that the inclusion volume
fraction is directly proportional to the oxygen level (see equation (2-75) in Chapter 2).
Effect of inclusion size distribution

Void nucleation may occur both by cracking of the inclusions and by interface decohesion. In
the former case, the critical stress for particle cracking ap is given by:16
(7-2)
where 7^ is the surface energy of the particle, Ep is the Young's modulus of the particle, A is the
stress concentration factor at the particle, and dv is the particle diameter.
GMAW
(E=1.6kJ/mm)
True fracture strain
Inclusion volume fraction

Fig. 7.2. Variation of true fracture strain £/with inclusion volume fraction Vv. Data from Widgery.12
SAW and FCAW

CVN upper shelf energy, J
Fig. 7.3. Correlation of CVN upper shelf energy with true fracture strain in tensile testing. Data from
Akselsen and Grong.20
Equation (7-2) predicts that large inclusions will tend to form voids first as the stress re-
quired for initiation is proportional to (l/dv )1/2. This result is also in agreement with experi-
mental observations. As shown in Fig. 7.5, the size distribution of inclusions located in the
centre of voids at the fracture surface is significantly coarser than the corresponding particle
size distribution in the material. In particular, large, angular shaped aluminium oxide (AI2O3)
SAW
Oxygen content, wt%
Fig. 7.4. Correlation of CVN upper shelf energy with analytical weld metal oxygen content. Data from
Devillers et aL 13
inclusions appear to be preferential nucleation sites for microvoids in low-alloy steel weld
metals (see Fig. 7.5(b)). Although the combined effect of particle size and local stress concen-
tration on the ductile fracture behaviour cannot readily be accounted for in a mathematical
simulation of the process, the CVN data in Fig. 7.6 suggest that the content of large inclusions
(e.g. of a diameter greater than about 1.5 Jim) should be minimised in order to maintain a high
resistance against dimpled rupture. In practice, this requires careful control of the weld metal
aluminium-oxygen balance and the heat input applied during welding (see Section 2.12 in
Chapter 2).
Effect of strength level

The toughness of a material reflects its ability to absorb energy in the plastic range. One way of
looking at toughness is to assume that it scales with the total area Uj under the stress-strain
curve. Several mathematical expressions for this area have been suggested. For ductile mate-
rials we may write:19
(7-3)
where Rm is the ultimate tensile strength (UTS).

If Uj is regarded as a material constant, one would expect that Rm and £y are reciprocal
Total inclusion SAW
population
Inclusions associated
with dimples
Frequency, %
Frequency, %
(a) Inclusion diameter, jum
Total inclusion SAW

population
Frequency, %
Frequency, %
Inclusions associated
with dimples
(b) Inclusion diameter, jum
Fig. 7.5. Histograms showing the size distribution of non-metallic inclusions in the weld metal and in the
centre of microvoids at the fracture surface, respectively; (a) Low aluminium level (Al-containing man-
ganese silicate inclusions), (b) High aluminium level (AI2O3 inclusions). Data from Andersen.18
High Ti levels
Medium Ti levels
Low Ti levels
SAW
Nv (d v >1.5 um)-105
Fig. 7.6. Correlation of CVN upper shelf energy with number of particles per mm3 greater than 1.5 urn,
Nv(dv > 1.5 um). Data from Grong and Kluken.15
SAW and FCAW

Ultimate tensile strength, MPa
Fig. 7.7. Correlation of true fracture strain with ultimate tensile strength (low-alloy steel weld metals).
Data from Akselsen and Grong.20
quantities, i.e. an increase in Rm is always associated with a corresponding decrease in Ef,

according to the equation:
(7-4)
where c^ is a constant which is characteristic of the alloy system under consideration.

It is evident from the tensile test data in Fig. 7.7 that the fracture strain is a true function of
Rm, although the relationship appears to be linear rather than non-linear, as predicted by equa-
tion (7-4). These results are of considerable practical importance, since they imply that the
upper shelf energy absorption, and hence, the shape of the CVN transition curve is strongly
affected by the weld metal strength level. Accordingly, control of the weld metal microstruc-
ture becomes particularly urgent at high strength levels to avoid problems with the cleavage
fracture resistance (to be discussed below).
7.2.1.3 Weld metal resistance to cleavage fracture

Cleavage fracture is characterised by very little plastic deformation prior to the crack propaga-
tion, and occurs in a crystallographic fashion along planes of low indicies, i.e. of high atomic
density.1 Body-centred cubic (bcc) iron cleaves typically along {100} planes, which implies
that the cracks must be deflected at high angle grain (or packet) boundaries, as shown
schematically in Fig. 7.8. Consequently, in steel weld metals the ferrite grain size and the
bainite packet width are the main microstructural features controlling the resistance to cleav-
age crack propagation.
Since the microstructure which forms within each single austenite grain will not be uniform
but a complex mixture of two or more constituent phases, it is difficult, in practice, to define
a meaningful grain size or packet width. For this reason, most investigators have attempted to
correlate toughness with the presence of specific microconstituents in the weld metal.3"5 For
example, an increase in the volume fraction of acicular ferrite will result in a corresponding
increase in toughness (i.e. decrease in the CVN transition temperature), as shown in Fig. 7.9.
(a)
(b)
Fig. 7.8. Schematic diagrams showing cleavage crack deflection at interfaces; (a) High angle ferrite-
ferrite grain boundaries, (b) High angle packet boundaries (bainitic microstructures).
SAW (E = 5.2-6.2 kJ/mm) Al: 0.018-0.062 wt%
Ti: 0.005 - 0.065 wt%
Transition temperature, 0C
O: 0.018-0.058 wt%
Acicular ferrite content, vol%

Fig. 7.9. Correlation between the weld metal 35J CVN transition temperature and the acicular ferrite
content. Data from Grong and Kluken.15
This observation is not surprising, considering the extremely fine lath size of the acicular
ferrite microstructure (typically less than 5jim).
Moreover, results obtained from fractographic examinations of SMA and FCA steel weld
metals have demonstrated that large non-metallic inclusions (> l|im) can strongly influence
the cleavage fracture resistance, either by acting as cleavage cracks themselves of by provid-
ing internal sites of stress concentration which facilitate carbide-initiated cleavage in the adja-
cent matrix.21'22 In the former case, the critical stress required for crack propagation in the
matrix, Cf(M), is given by the Griffith's equation:19
(7-5)
where En is the Young's modulus of the matrix, ye^ is the effective surface energy (equal to
the sum of the ideal surface energy and the plastic work), and c is the half crack length.
Since c is proportional to the particle diameter dv, equation (7-5) predicts that welds con-
taining large inclusions should be more prone to cleavage cracking than others. This result is
also in agreement with general observations. For example, in self-shielded FCA steel weld
metals it has been demonstrated that cleavage crack initiation is usually associated with large
aluminium-containing inclusions which form in the molten pool before solidification (see Fig.
7.10). Consequently, control of the inclusion size distribution is essential in order to ensure an
adequate low-temperature toughness.
7.2.1.4 The weld metal ductile to brittle transition

In addition to the parameters mentioned above, there are several other factors, some interre-
lated, which play an important part in the initiation of cleavage fracture. These are:1
(i) The temperature dependence of the yield stress.

(a)
(b)
(C)
Fig. 7.10. Initiation of cleavage fracture in a self-shielded FCA steel weld from an aluminium-containing
inclusion; (a) Initiation site short distance ahead of the notch, (b) Detail of initiation site showing cracked
inclusion, (c) Detail of cracked inclusion (remnants of particle are left in the hole).
(ii) Dislocation locking effects caused by interstitials or alloying elements in solid

solution (e.g. nitrogen and silicon),
(iii) Nucleation of cracks at twins,
(iv) Nucleation of cracks at carbides.
In general, this picture is too complicated to establish a physical framework within which
the various theoretical models for the ductile to brittle transition in steel can be embedded. We
are therefore forced to base our judgement and understanding of how key parameters affect the
position and shape of the CVN transition curve solely on scattered phenomenological observa-
tions and empirical models (e.g. see the reviews of Grong and Matlock3 or Abson and Pargeter4).
An example of how far the latter approach has been developed is given below.
Akselsen and Grong20 have established a series of empirical equations which relates tough-
ness to the weld metal acicular ferrite content and the ultimate tensile strength (UTS). Figures
7.11 and 7.12 show how each of these parameters influences the CVN transition curve. It is
evident from the diagrams that control of the weld metal acicular ferrite content becomes
particularly important at high strength levels to avoid problems with the fracture toughness. In
cases where undermatch is aimed at (i.e. a weld metal to base plate strength ratio less than
unity), the weld metal tensile strength is typically of the order of 450 to 550 MPa. Within this
range a volume fraction of acicular ferrite beyond 25 vol% will generally be sufficient to meet
current toughness requirements (35 J at -40 0 C). If overmatch is desired, the volume fraction
of acicular ferrite becomes more critical, partly because of a higher weld metal strength level
and partly because of more stringent toughness requirements (e.g. 45 J rather than 35 J at
-40 0 C). Process diagrams of the type shown in Figs. 7.11 and 7.12 can therefore serve as a
basis for proper selection of consumables for welded steel structures.
It should be noted that Akselsen and Grong20 in their analysis omitted a consideration of the
important influence of free (uncombined) nitrogen and non-metallic inclusions on the CVN
transition curve. Based on the experimental data in Fig. 7.13 it can be argued that such
compositional variations can be equally detrimental to toughness as a decrease in the acicular
ferrite content. Consequently, further refinements of the models are required if a verified
quantitative understanding of the ductile to brittle transition in low-alloy steel weld metals is
to be obtained.
Example (7.1)
Consider multipass FCA steel welding with two different electrode wires, one with titanium
additions and one without. Table 7.1 contains a summary of weld metal chemical composi-
tions. Provided that the microstructure and the inclusion size distribution are similar in both
cases, use this information to evaluate the low-temperature toughness of the welds, as revealed
by CVN testing.
Solution
Since the nitrogen content of both welds is quite high (0.011 wt%), the risk of a toughness
deterioration due to strain ageing is imminent, particularly at low Ti levels. Taking the atomic
weight of titanium and nitrogen equal to 47.9 and 14.0 g mol"1, respectively, the stoichiometric
amount of titanium that is necessary to tie-up all nitrogen as TiN can be calculated as follows:
WeIdA
In weld A most of the nitrogen is free (uncombined) due to an unbalance in the titanium con-
tent. This means that the risk of a toughness deterioration due to strain ageing is high.
(a)
Tensile strength: 600 MPa
Absorbed energy, J
Vol% acicular ferrite
35 Joules
Test temperature, 0C
(b)
Tensile strength: 800 MPa
Absorbed energy, J
Vol% acicular ferrite
35Joules.
Fig. 7.11. Predicted effect of weld metal acicular ferrite content on the CVN transition curve at two
different tensile strength levels; (a) Rm = 600 MPa, (b) Rm = 800 MPa. Data from Akselsen and Grong.20
WeIdB
Weld B contains 0.030 wt% Ti, which is not far from the stoichiometric amount of titanium
necessary to tie-up all nitrogen. Although some titanium also is bound as Ti2O3, it is reason-
able to assume that the free nitrogen content in this case is sufficiently low to eliminate prob-
lems with strain ageing. Consequently, weld B would be expected to exhibit the highest tough-
ness (i.e. the lowest CVN transition temperature) of the two, as indicated in Fig. 7.14.
(a)
25 vol% acicular ferrite
Absorbed energy, J UTS
35 Joules
(b)
75 v o l % acicular ferrite
Absorbed energy, J
UTS
-..35J.Q.ute$-
Test temperature, 0 C
Fig. 7.12. Predicted effect of weld metal ultimate tensile strength (UTS) on the CVN transition curve at
two different volume fractions of acicular ferrite; (a) 25 vol% AF, (b) 75 vol% AF. Data from Akselsen
and Grong.20
Table 7.1 Chemical composition of FCA steel weld metals considered in Example (7.1).
Element
Weld wt% C wt% Si wt% Mn wt% Al wt% Ti wt% S wt% N wt% O
A 0.10 0.40 1.50 0.005 0.006 0.008 0.011 0.031

B 0.10 0.40 1.50 0.005 0.030 0.008 0.011 0.031
(a)
Absorbed energy, J SMAW (basic electrodes) 95% confidence interval
Testing temperature: -400C
Nitrogen content, ppm
(b)
Low content of coarse inclusions

Absorbed energy, J
High content of coarse

inclusions ( > 1}im )
Self-shielded FCA steel

weld metals
Fig. 7.13. Effect of impurities on weld metal CVN toughness; (a) Nitrogen content, (b) Inclusion level.
Data from ESAB AB (Sweden) and Grong et al. 22
7.2.1.5 Effects of reheating on weld metal toughness

In principle, improvement of weld properties can be achieved through a post-weld heat treat-
ment (PWHT). This may have the benefits of:3
(i) Enhancing the fatigue strength through a general reduction of welding residual
stresses.
Absorbed energy
WeIdB WeIdA
35 J
Fig. 7.14. Schematic drawings of the CVN transition curves for welds A and B (Example (7.1)).
(a)
(b)
Fig. 7.15. Typical low-temperature fracture modes of Ti-B containing steel weld metals; (a) Quasi-
cleavage (as-welded condition), (b) Intergranular fracture (after PWHT).
(ii) Increasing the toughness by recovery (i.e. removal of strain-aged damage) and
martensite tempering.
For these reasons local PWHTs are commonly required for all structural parts above a
specified plate thickness (e.g. 50 mm according to current North Sea offshore specifications).
Post-weld heat treatment is usually carried out in the temperature range from 550 to 6500C.
In practice, however, the toughness achieved will depend on the weld metal chemical com-
position, and in some cases deterioration rather than improvement of the impact properties is
observed after PWHT. In such cases the reduction in toughness can be ascribed to:3'4
(i) Precipitation hardening reactions. Present experience indicates that elements such
as vanadium, niobium, and presumably titanium can produce a marked deteriora-
tion in toughness because of precipitation of carbonitrides in the ferrite, provided
that these elements are present in the weld metal in sufficiently high concentra-
tions.
(ii) Segregation of impurity elements (e.g. P, Sn, Sb and As) to prior austenite grain
boundaries, which, in turn, can give rise to intergranular fracture. The indications
are that this type of embrittlement is strongly enhanced by the presence of second
phase particles at the grain boundaries.
Experience shows that Ti-B containing steel weld metals often fail by intergranular frac-
ture in the columnar grain region after PWHT,23 as evidenced by the SEM fractographs in Fig.
7.15. The observed shift in the fracture mode is associated with a significant drop in toughness
(Fig. 7.16) and arises from the combined action of solidification-induced phosphorus
segregations and borocarbide precipitation along the prior columnar austenite grain bounda-
SAW Open symbols: 5 - 8 ppm B

Filled symbols: 20 - 25 ppm B
ACVN, J
Base line

Fig. 7.16. Observed displacement in the CVN toughness after PWHT (ACVN) as a function of the weld
metal boron and titanium contents. Negative values indicate loss of toughness. Data from Kluken and
Grong.23
Fig. 7.17. TEM micrograph showing precipitation of borocarbides, Fe23(B,C)6, along prior austenite
grain boundaries in a Ti-B containing steel weld metal after PWHT (600 0 C-Ih).
ries (Fig. 7.17). Since borocarbides are brittle and partly incoherent with the matrix, they can
be regarded as microcracks (of length dp) ready to propagate. In such cases there is virtually
no plastic deformation occurring before crack propagation, which implies that the intergranular
fracture stress is given by the Griffith's equation:24
(7-6)
where 7 ^ is again the effective surface energy (equal to the sum of the ideal surface energy
and the plastic work), and dp is the particle diameter.
Although the value of yeg_ would be expected to be low in the presence of solidification-
induced phosporus segregations,24 this alone is not sufficient to initiate intergranular fracture
in the weld metal. However, during PWHT the borocarbides will start to grow from an ini-
tially small value to a maximum size of about 0.1 to 0.2jim (Fig. 7.17), following the classic
growth law for grain boundary precipitates dpatl/4?5 This implies that the intergranular frac-
ture stress, Oj(I), will gradually decrease with increasing annealing times, as indicated in Fig.
7.18. When the matrix fracture strength, Cj(M), is reached, the fracture mode shifts from
predominantly quasi-cleavage in the as-welded condition (Fig. 7.15(a)) to intergranular rup-
ture after PWHT (Fig. 7.15(b)). This is observed as a marked reduction in the CVN toughness,
as shown previously in Fig. 7.16.
7.2.2 HAZ mechanical properties

The last twenty years have seen a revolution in the metallurgical design of steel. Whereas old
steels relied on the use of carbon for strength, the trend today is to rely more on grain refine-
ment in combination with microalloy precipitation to meet the current demand for an im-
proved weldability. This includes both the sensitivity to weld cracking and the HAZ mechani-
cal properties required by service conditions and test temperatures. The latter aspect is of
particular interest in the present context and will be discussed later.
Stress
Intergranular
fracture mode
Quasi-cleavage
fracture mode
Particle diameter
[Annealing time]174
Fig. 7.18. Schematic illustration of the mechanisms of temper embrittlement in Ti-B containing steel
weld metals (Gf(M): matrix fracture strength, (*/(/): intergranular fracture strength).
7.2.2.1 HAZ hardness and strength level

The HAZ hardness and strength level is of significant practical importance, since it influences
both the cracking resistance and the toughness. Although the peak strength is mainly control-
led by the martensite content (see Fig. 7.19), the relationship is generally too complicated to
allow reliable predictions to be made from first principles. This implies that our understanding
of the HAZ strength evolution, at best, is semi-empirical.
Martensite content, vol%
p0.2> R nv M P a
HV 5 ,kp/mm 2
Cooling time, At 8 / 5 , s
Fig. 7.19. Structure-property relationships in the grain coarsened HAZ of low-carbon microalloyed steels
(vol% M: martensite content, Rp : 0.2% proof stress, Rm: ultimate tensile strength, HV5: Vickers hard-
ness, A%5.* cooling time from 800 to 5000C). Data from Akselsen et al.26
A number of different empirical models exist in the literature for prediction of HAZ peak
hardness and strength.26"31 However, the aptness of some of these models is surprisingly
good, which justifies construction of iso-hardness and iso-strength diagrams for specific grades
of steels.32 Examples of such diagrams are given in Fig. 7.20. It is evident from Fig. 7.20 that
the HAZ peak strength is controlled by two main variables, i.e. the steel chemical composition
and the weld cooling programme. The compositional effect is allowed for by the use of an
empirical carbon equivalent, which ranks the influence of the various alloying elements on the
steel hardenability. According to Yurioka et al.,2* the CEn-equivalent is given as:
(7-7)
where all compositions are given in wt%.

Moreover, the cooling time from 800 to 5000C, Af8/5, is used as an abscissa in Fig. 7.20.
This parameter is widely accepted as an adequate index for the weld cooling programme, and
can be read from nomograms of the type shown in Fig. 1.49 (Chapter 1). The axes of Fig. 1.49
are dimensionless, but they can readily be converted into real numbers through the use of the
following conversion factors:33
Ordinate:
(7-8)
Abscissa:
(7-9)
The different parameters in equations (7-8) and (7-9) are defined in Appendix 7.1.
The results in Fig. 1.49 are interesting, since they show that the cooling time, A%5, depends
on the mode of heat flow during welding. In this case the transition from 'thick' to 'thin'
plates, corresponding to an abscissa of about 0.64, is clearly not represented by a single plate
thickness d, but will be a function of both the net heat input r\E and the ambient temperature T0.
Accordingly, the HAZ strength level is seen to vary between wide limits, depending on the
steel chemical composition and the operational conditions applied (Fig. 7.20).
Example (7.2)
Consider stringer bead deposition (GMAW) on two low-alloy steel plates of similar composi-
tion but different thickness under the following conditions:
I = 250A, U = 30V, v = 5mm s"1, r| = 0.8, T0 = 200C
According to the steel mill certificate the CEn carbon equivalent is equal to 0.46 wt%. Use
this information together with the diagrams in Figs. 1.49 and 7.20 to estimate the peak HAZ
strength level when the plate thickness is 10 and 30 mm, respectively.
(a)
CEn, wt%
Cooling time, A t 8 / 5 , s
(b)
CE||f wt%
Cooling time, At 8 7 5 , s
Fig. 7.20. HAZ iso-property diagrams for HSLA steels; (a) Iso-hardness contours, (b) Iso-yield strength
contours. Data from Kluken et al.32
Solution
First we calculate the net heat input per unit length of the weld r\E:
From equation (7-9) we have:
Readings from Fig. 1.49 then give:
d = 10 mm:
from which
d = 30 mm:
from which
We can now use the diagrams in Fig. 7.20(a) and (b) to obtain the peak HAZ hardness and
yield strength, respectively. This gives:
d = 10 mm:
d = 30 mm:
It is evident from the above calculations that the HAZ strength level is sensitive to varia-
tions in the welding conditions. Normally, the HAZ hardenability is high enough to facilitate
a local strength increase adjacent to the fusion boundary, as shown in Fig. 7.21. An exception
is high heat input welding on quenched and tempered steels (Fig. 7.2l(b)), where the presence
of large amounts of Widmanstatten ferrite and polygonal ferrite within the grain coarsened and
grain refined region, respectively can lead to a severe HAZ softening. This type of mechanical
impairment represents a problem in engineering design, since it puts a restriction on the use of
high strength steels in welded structures.
(a) Medium strength High strength

steels steels.
R p02 and R m ,
Gf[R.
GCR
GCR
GRR
BM"
BM
JfL
IR"
Low heat input welding: E^ 1 -2 kJ/mm
(b) Medium strength High strength

steels steels
p0.2 and R m>
MPa
R
QQR.
^GRR.
GCR
"GRR
TR"
SR"
BM"
JfL,
,BM
[SR
High heat input welding: E^4 kJ/mm
Fig. 7.21. Effects of steel chemical composition and welding conditions on the HAZ strength level (BM:
base metal, SR: subcritical region, IR: intercritical region, GRR: grain refined region, GCR: grain coars-
ened region); (a) Low heat input, (b) High heat input. Data from Akselsen and R0rvik.34
7.2.2.2 Tempering of the heat affected zone
Certain regulations for offshore structures require that no part of the welded joint shall be
harder than a specified limit, e.g. 280, 300 or 325 VPN, to reduce the risk of hydrogen crack-
ing. Such requirements cannot always be met by a suitable choice of preheating and welding
conditions.
In practice, a reduction in the HAZ strength level can be achieved by applying a PWHT.
The tempering effect of different temperature-time combinations can be described by the
Hollomon-Jaffe parameter:35
(7-10)
where T is in K (absolute temperature).

In Fig. 7.22 the isothermal hardness data reported by Olsen et al.36 have been plotted against
the empirical Hollomon-Jaffe parameter. In this particular case the best fit is obtained if the
constant B* in equation (7-10) is equal to 16.5 (with t in seconds). It is evident from Fig. 7.22
that tempering at, say, 6000C for 1 h is more than sufficient to reduce the HAZ peak hardness
to values below 280 VPN. This implies that PWHT is an effective (but expensive) way of
reducing the HAZ strength level.
Deposition of temper weld beads has been suggested as an alternative means of reducing
the hardness of the HAZ.36"38 This procedure is indicated schematically in Fig. 7.23, showing
two temper beads (black) in the lower sketch. If the beads are properly positioned with respect
to the fusion line, the outer Ac\ contour of the HAZ produced by the temper bead should just
touch the fusion line of the last filler pass, as indicated in the upper sketch of Fig. 7.23. The
material reaustenitised by the temper bead would then be weld metal, while the HAZ remain-
ing from the last filler pass would be tempered below the transformation range.
Filled symbols:
t = 10 seconds
Vickers hardness, VPN
Steel chemical composition (wt%)
F> = T(16.5 + logt)
Fig. 7.22. Hollomon-Jaffe type plot of isothermal hardness data. After Olsen et al.36
Temper bead
Fusion line
Last Ac3 line
filler pass Ac1 line
Fig. 7.23. Schematic illustration of weld bead tempering.

Since the Hollomon-Jaffe parameter is an empirical criterion developed for isothermal tem-
pering of medium and high carbon steels, it cannot readily be applied to pulsed tempering. A
better approach would be to use the so-called Dorn parameter,39 which in an integral form can
be written as:39'40
(7-11)
where Qapp. is the apparent activation energy for the controlling diffusion reaction.
The Dorn parameter has proved useful to compare isothermal and pulsed tempering data on
the assumption that the kinetics of softening, in the actual range of hardness, are controlled by
diffusion of carbon in ferrite. Qualitatively, the aptness of equation (7-11) can be illustrated in
a plot of measured hardness against the diffusional parameter P^ ( s e e Fig- 7.24). It is evident
from Fig. 7.24 that the isothermal data points can be represented by a smooth curve which
coincides with the upper boundary of the scatter band obtained in pulsed tempering. The
slightly higher hardness observed after isothermal tempering arises probably from a brief pe-
riod of heating that makes the effective time somewhat less than 10 s.
Case Study (7.1)

As an illustration of principles, Fig. 7.25 shows a case of identical welding parameters for the
last filler pass and the temper bead, the latter one being positioned so as to give a peak tem-
perature of 7200C at the fusion line of the former one. The temperature field around the two
Isothermal 10 s
Series 1
" 2
Hardness ratio HV/HVmax, %

" 3
" 4
Vickers hardness. VPN

Double pulse
1 2
^s '
Fig. 7.24. Measured hardness ratio HVIHVmax. vs the Dorn parameter P2 (Qapp. = 83.14 kJ mol *). Data
fromOlsentf/tf/.36
beads is clearly the same. In Fig. 7.25 an estimate has been based on the simplified Rykalin
thick plate solution, which applies to a fast moving high power source on a semi-infinite body
(see equation (1-73) in Chapter 1). At T-T0 ~ 15000C, a fusion line radius of about 4.4 mm is
obtained for a net heat input of 0.8 kJ mm"1. The corresponding Ac\ radius is 6.5 mm.
The temperature-time pattern is shown in the lower left graphs of Fig. 7.25 for three differ-
ent positions in the HAZ, i.e. y = 0 (former fusion line), y = 1 mm, and y = 2 mm (z = 0). The
corresponding plots of dP2 ldt vs t are shown to the right. Taking the area P2 under each curve
and reading the hardness ratio at TJP^ from Fig. 7.24, an expected hardness profile is ob-
tained, as shown in the upper diagram of Fig. 7.25. The expected effect of tempering is seen to
range from a hardness of about 65% (HV « 265 VPN) at the fusion line to about 80% (HV «
340 VPN) close to the outer boundary of the HAZ (y = 2 mm). If the centre-line displacement
had been different from the chosen optimum of 2.1 mm (e.g. say 3 mm), the predicted hardness
curve would be shifted to about 75% and 90% of the peak hardness at y = 0 and y = 2 mm,
respectively. On the other hand, if the centre-line distance had been shorter, say 1 mm, a
narrow zone of the original HAZ would be re-austenitised and therefore about as hard as
before deposition of the temper bead.
The results from the above modelling exercise show that the HAZ hardness of weld toes
and cap layers can be reduced by applying an appropriate temper bead technique. However,
this requires an extremely good process control, since the temper beads must be positioned
very precisely for a successful result. Consequently, the use of temper beads for improvement
of the HAZ properties has not found a wide application in the industry.3641
7.2.2.3 HAZ toughness

In spite of the recent developments in steel plate manufacturing technology, there is still con-
cern about the HAZ toughness of low-carbon microalloyed steels because of their tendency to
max
HV/HV
last filter pass
temper bead
y, mm
HA2
We d metal Parent plate
T, 0C
106exp(-10000/T)
t,s t,s
Fig. 7.25. Application of Dorn parameter to weld bead tempering (Case Study (7.1)).
form brittle microstructures within specific thermal regions of the weld. 4142 Moreover, im-
provement of the HAZ toughness through PWHT is sometimes found to be difficult in contrast
to experience with more traditional C-Mn steels.41'43 Consequently, the increasing use of low-
carbon microalloyed steels in various welded structures has introduced new problems related
to the HAZ brittle fracture resistance which formerly did not appear to be of particular con-
cern.44
Fully transformed region

Specific effects of peak temperature on HAZ toughness, as assessed on the basis of thermally
cycled CVN specimens, are shown in Fig. 7.26. It is apparent from the graph that embrittlement
in the fully transformed part of the HAZ is often located in the grain coarsened region adjacent
SR R
I G
RR G
CR
-3
(D
c
O
I Sn
igel cycel
0
Peak e
tmperau
tre,C
Fig. 7.26. Effects of peak temperature on the CVN energy absorption at -400C (SR: subcritical region,
IR: intercritical region, GRR: grain refined region, GCR: grain coarsened region). Data from Akselsen et
a/.45
to the fusion boundary where the peak temperature of the thermal cycle has been above about
12000C. The problem can mainly be ascribed to the presence of low-toughness microstruc-
tures such as upper bainite and Widmanstatten ferrite which form typically at intermediate and
slow cooling rates (see Fig. 6.55 in Chapter 6). In contrast, the grain refined region will almost
always exhibit a satisfactory low-temperature toughness owing to the characteristic fine po-
lygonal ferrite microstructure.41 An exception is low heat input welds produced from steels
with a heavily banded pearlite/ferrite microstructure, where the risk of a toughness deteriora-
tion is imminent due to martensite formation along the prior base metal pearlite bands.45'46
In recent years a new class of low-carbon microalloyed steels has emerged which is charac-
terised by an excellent low temperature HAZ toughness, even at high heat inputs (see Fig.
7.27). This particular grade is frequently referred to as Ti-O steels due to their content of
indigenous titanium oxide inclusions (presumably Ti2O3). Although the mechanisms involved
are not yet fully understood, it is reasonable to assume that the improved toughness at high
heat inputs arises from a refinement of the HAZ microstructure, as discussed previously in
Section 6.3.6 (Chapter 6). It is interesting to note that the major effect of the titanium oxide
inclusions in this case appears not to be control of the austenite grain size (which in some cases
can exceed 500 |im at the fusion boundary), but is rather to act as favourable nucleation sites
for acicular ferrite intragranularly.4748 Similar phenomena are well known from transforma-
tion kinetics of low-alloy steel weld deposits, where non-metallic inclusions play an important
role in the development of the acicular ferrite microstructure.3"5
Intercritical region
The microstructural evolution in the intercritical HAZ of low-carbon steels has previously
been discussed in Section 6.3.8.2 (Chapter 6).
In order to understand the origin of embrittlement in the intercritical region, consideration
must be given to the stress fields and the transformation strains developed in the ferrite matrix
Ti-O steel
Ti-N steel
Transition temperature, 0C
Fig. 7.27. Response of modern Ti-O steels and traditional Ti-N steels to CVN testing following weld
thermal simulation. Data from Homma et al.41
as a result of the martensite formation.49 It follows from Fig. 7.28 that the hard martensite-
austenite (M-A) islands will give rise to significant stress concentrations at the martensite/
ferrite interface owing to the pertinent difference in the yield strength (stiffness) between the
two phases. At the same time, the volume expansion associated with the austenite to martensite
transformation leads to significant elastic and plastic straining of the ferrite.50 At moderately
high temperatures and deformation strains, many of the matrix dislocations will be mobile,
which means that the ferrite will maintain its ductility, while the stiffer M-A islands are ex-
posed to cracking and debonding. With increasing strain, the cracks can grow into voids and
further develop into deep holes, until final rupture occurs by hole/void coalescence due to
internal necking.49 However, when mechanical testing is performed at subzero temperatures
under high strain rate conditions (> 102 s"1 for CVN testing), the flow strength of the ferrite
increases significantly because of the reduced mobility of the screw dislocations.51 In addition,
strain partitioning between the M-A islands and the ferrite may also occur, which further
enhances the stress concentrations at the M-A/ferrite interface.52 Accordingly, the local stress
level at the interface will eventually exceed the cleavage strength of the ferrite, with conse-
quent initiation of brittle fracture. This conclusion is consistent with observations made from
tensile testing of dual-phase steels, showing that failure of dual-phase microstructures often is
caused by fracture in the ferrite region.52"54
Because the intercritical HAZ toughness is closely related to the volume fraction of the
M-A constituent in the matrix, 4 5 5 1 5 5 embrittlement can normally be avoided by decreasing
the cooling rate through the critical transformation temperature range to facilitate pearlite for-
mation (see Fig. 7.29). An exception is boron-containing steels, where the HAZ hardenability
is high enough to stabilise the M-A constituent, even at slow cooling rates (see CVN data for
steel B in Fig. 7.29).
Normalized stress
Stiff
particle
Normalized distance
Fig. 7.28. Stress distribution in matrix caused by stiff inclusion (or: radial stress, (5$: tangential stress,
tmax.' maximum shear stress). Data from Chen et al.49
Open symbols:
Filled symbols:
Steel A
Absorbed energy, J
(T-L)
(L-T)
Steel B
(T-L)
Cooling time, At 6 / 4 , s
Fig. 7.29. Effect of cooling time Ar674 on the intercritical HAZ toughness at -20 0 C (thermally cycled
specimens). Steel A: 11 ppm B, Steel B: 26 ppm B. Data from Ramberg et al.55
Effect of PWHT
Considering the intercritical HAZ, a significant improvement of the CVN toughness can be
achieved by applying a PWHT, as shown by the data of Akselsen et al51 This effect arises
partly from a reduction of the stress concentrations at the M-A/ferrite interface as a result of
martensite tempering and partly from relaxation of transformation strains within the ferrite
matrix.51 Such recovery reactions will start to occur when the temperature is raised above
about100 0 C.
(a)
(b)
CTOD at -1O0C, mm
P content, wt%
Fig. 7.30. Effects of PWHT on the grain coarsened HAZ toughness; (a) Example of intergranular frac-
ture along prior austenite grain boundaries after PWHT (6000C - 1 h), (b) Measured CTOD vs base
plate phosphorus content for post weld heat treated specimens (6000C - 4 h). Data quoted by Grong and
Akselsen.41
In contrast to the behaviour described above for the intercritical HAZ, the reported effect of
PWHT on the grain coarsened HAZ toughness is much more complicated and rather confus-
ing. However, experience has shown that particularly niobium-vanadium containing steels
are sensitive to PWHT due to the strong precipitation hardening potential of Nb(C,N) and
V(C,N).43'56 In addition, a toughness deterioration may occur as a result of segregation of
impurity elements such as phosphorus, tin, and antimony to prior austenite grain boundaries.
This, in combination with a tempered martensitic microstructure, can lead to intergranular
fracture when testing is performed at subzero temperatures (see Fig. 7.30(a)). The detrimental
effect of phosphorus on the HAZ toughness of low carbon microalloyed steels after PWHT is
shown in Fig. 7.30(b).
Example (7.3)
Consider procedure test SA welding on a thick plate of a Nb-microalloyed steel under the
/ = 500A, U = 30V, v = 6mm s"1, r\ = 0.95, T0 = 200C
Table 7.2 contains data from CVN testing of the base plate and thermally cycled specimens.
The weld thermal simulation experiments were carried out at three different peak tempera-
tures (i.e. 13500C, 10000C, and 7800C) under cooling conditions similar to those employed in
the SA welding trial. Based on the data in Table 7.2 and the simplified Rykalin thick plate
solution (equation (1-73) in Chapter 1), estimate the locations of the brittle zones (referred to
the fusion boundary) within the HAZ of the SA procedure test weld considered above.
Solution
It is evident from the CVN data in Table 7.2 that the HAZ toughness would be expected to be
low in positions of the weld where the peak temperature has been close to 780 and 13500C,
conforming to the intercritical and grain coarsened region, respectively. Based on the simpli-
fied Rykalin thick plate solution, the following expression can be derived for an arbitrary
isothermal zone width, Ar*m, referred to the fusion boundary (see equation (5-47) in Chapter
5):
Taking pc and Tm equal to 0.005 J mm"3 0C"1 and 15200C, respectively for low-alloy steels
(from Table 1.1 in Chapter 1), we obtain:
Table 7.2 Results from CVN testing of base metal, thermally cycled specimens, and procedure test
weld (Example (7.3))
Test results Absorbed energy at -40 0 C (J)

Base metal 320, 310, 305; average: 312
0
Thermally cycled Tp = 780 C 40, 36, 34; average: 37
specimens T = 10000C 225, 220, 219; average: 221
7;= 13500C 50, 46, 40; average: 46
Weld HAZ* GCR: 63, GRR: 225, IR: 53
1
GCR: grain coarsened region; GRR: grain refined region; IR: intercritical region.
Next Page
Intercritical HAZ (Tp « 7800C):
Grain coarsened HAZ (Tp « 13500C):
From this we see that the brittle zones are located 3.5 and 0.5 mm from the fusion bound-
ary, respectively. A comparison with the procedure test results in Table 7.2 shows that the
measured CVN toughness after welding at these locations is slightly higher than that inferred
from the weld thermal simulation experiments. This observation is not surprising, considering
the fact the CVN specimens extracted from the procedure test weld, in practice, include a
wide spectrum of thermal regions which have undergone highly different temperature-time
programmes, whereas the microstructure within the thermally cycled CVN specimens is more
homogeneous due to a similar temperature-time pattern across the whole gauge length (see
Fig. 7.31). Hence, weld thermal simulation cannot replace procedure testing carried out on
real welds. Nevertheless, it is a useful method of evaluating the microstructural stability and
mechanical response of materials to reheating, as experienced in welding.
7.2.3 Hydrogen cracking

Hydrogen embrittlement as a problem is mainly associated with ferritic steels and the risk of
crack initiation in the grain coarsened HAZ following welding.5758 As shown in Fig. 7.32,
these cracks are usually situated at weld toes, weld root, or in an underbead position.
Occationally, hydrogen cracks can also develop in the weld metal. A characteristic feature of
hydrogen-induced cracking is that the process is time-dependent, i.e. the crack may first ap-
pear after several minutes or hours from the time of arc extinction. Consequently, the phenom-
enon is also referred to delayed cracking or cold cracking in the scientific literature.
7.2.3.1 Mechanisms of hydrogen cracking

Hydrogen embrittlement in steels in characterised by:59'60
(i) The crystal structure dependence

Hydrogen embrittlement is mainly associated with materials which exhibit a bcc
or a bet crystal structure, i.e. ferritic and martensitic steels. Austenitic stainless
steels and aluminium alloys with a fee crystal structure are usually not sensitive to
hydrogen.
(ii) The microstructure dependence

A martensitic steel is generally more prone to hydrogen cracking than a ferritic
steel, but a martensitic microstructure is not a requirement for crack initiation.
Previous Page
Intercritical HAZ (Tp « 7800C):
Grain coarsened HAZ (Tp « 13500C):
From this we see that the brittle zones are located 3.5 and 0.5 mm from the fusion bound-
ary, respectively. A comparison with the procedure test results in Table 7.2 shows that the
measured CVN toughness after welding at these locations is slightly higher than that inferred
from the weld thermal simulation experiments. This observation is not surprising, considering
the fact the CVN specimens extracted from the procedure test weld, in practice, include a
wide spectrum of thermal regions which have undergone highly different temperature-time
programmes, whereas the microstructure within the thermally cycled CVN specimens is more
homogeneous due to a similar temperature-time pattern across the whole gauge length (see
Fig. 7.31). Hence, weld thermal simulation cannot replace procedure testing carried out on
real welds. Nevertheless, it is a useful method of evaluating the microstructural stability and
mechanical response of materials to reheating, as experienced in welding.
7.2.3 Hydrogen cracking

Hydrogen embrittlement as a problem is mainly associated with ferritic steels and the risk of
crack initiation in the grain coarsened HAZ following welding.5758 As shown in Fig. 7.32,
these cracks are usually situated at weld toes, weld root, or in an underbead position.
Occationally, hydrogen cracks can also develop in the weld metal. A characteristic feature of
hydrogen-induced cracking is that the process is time-dependent, i.e. the crack may first ap-
pear after several minutes or hours from the time of arc extinction. Consequently, the phenom-
enon is also referred to delayed cracking or cold cracking in the scientific literature.
7.2.3.1 Mechanisms of hydrogen cracking

Hydrogen embrittlement in steels in characterised by:59'60
(i) The crystal structure dependence

Hydrogen embrittlement is mainly associated with materials which exhibit a bcc
or a bet crystal structure, i.e. ferritic and martensitic steels. Austenitic stainless
steels and aluminium alloys with a fee crystal structure are usually not sensitive to
hydrogen.
(ii) The microstructure dependence

A martensitic steel is generally more prone to hydrogen cracking than a ferritic
steel, but a martensitic microstructure is not a requirement for crack initiation.
(a)
Specimen holder
[Homogeneous zone
CVN-specimen
Notch location
(b) Notch location Weld metal
CVN-
specimen Base metal
Grain coarsened HAZ

Grain refined HAZ
Fig. 7.31. Methods for evaluation of HAZ toughness (schematic); (a) Weld thermal simulation, (b) Weld
procedure testing.
(a) (b)
HAZ Transverse crack
Toe
Root irack
Toe crack Underbead
crack crack
Underbead HAZ
crack
Fig. 7.32. Schematic diagrams showing hydrogen-induced cracks in different types of welds; (a) Fillet
weld, (b) Butt weld. The diagrams are based on the ideas of Coe.57
5%/min 104%/min
Uncharged
Uncharged
Charged
Charged
5x105%/min 1.9x106%/min
Uncharged
Charged
Charged
Uncharged
Fig. 7.33. Variation of true fracture strain with nominal strain rate and test temperature for charged and
uncharged specimens. Data from Brown and Baldwin.59
(iii) The strain rate dependence

Hydrogen embrittlement is most prominent at low strain rates typical of tensile
testing, as shown in Fig. 7.33. At high strain rates the hydrogen diffusion is not
fast enough to keep pace with the fracture development.
(iv) The temperature dependence

Hydrogen cracking occurs usually within the temperature range from -150 to
+2000C. This temperature dependence reflects the fact that both the hydrogen
concentration and the stress intensity at the crack tip must exceed some critical
value before crack propagation occurs.
(v) The time dependence

Since hydrogen embrittlement is a diffusion-controlled process, the cracks will
propagate in a stepwise manner to allow for supply of hydrogen from the sur-
rounding matrix to the crack tip (see Fig. 7.34).
Over the years a number of mechanisms have been proposed to explain the origin of hydro-
gen embrittlement. The three most important are:
Fracture
Resistance change x 10 ,Q.
-8
Fracture
Applied stress:
1240MPa
Applied stress:
1100MPa
Time, min
Fig. 7.34. Example of stepwise crack propagation in notched tensile specimens, as inferred from electri-
cal resistivity measurements. Data from Steigerwald et a/.60
(a) The hydrogen gas pressure model, originally proposed by Zapfee and Sims,61 which
postulates that atomic hydrogen will diffuse to microvoids where it recombines to
form molecular hydrogen. In ferritic steels the equilibrium H2(g) pressure within
the microvoids is typically of the order of 106 to 107 atm, which is more than
sufficient to bring about a local fracture development.
(b) The surface energy model (Petch62). According to this model hydrogen will re-
duce the effective surface energy of the crack. Under such conditions the crack
can propagate at a lower nominal stress in the presence of hydrogen, in agreement
with the Griffith's theory (equation (7-5)).
(c) The slip softening model of Beachem,63 which accounts for the experimental ob-
servation that hydrogen-charged specimens generally exhibit a lower flow stress
than hydrogen-free specimens. This suggests that hydrogen interfers with dislo-
cations in a manner which facilitates different types of fracture, including micro void
coalescence (or dimpled rupture), quasicleavage fracture, and intergranular frac-
ture.64
Currently, it cannot be stated with certainty which of these three mechanisms that are opera-
tive under the conditions existing in welding. However, this question is of minor importance
in the present context, since we here are mainly concerned with the factors responsible for
hydrogen cracking in steel weldments.
7.2.3.2 Solubility of hydrogen in steel
Since hydrogen is the smallest of all atoms, it is readily soluble in iron. In general, both
octahedral and tetrahedral lattice sites are potential traps for interstitials, as indicated in Fig.
7.35. In the case of hydrogen it is believed that the dissolved atoms are mainly present in
tetrahedral positions in the form of protons.65 Because of the pertinent difference in the size of
the fee and the bcc interstices (see Fig. 7.35), the hydrogen solubility in iron will change
stepwise with temperature following the bfe —> yFe and yFe —> aFe transformations, as shown
previously in Fig. 2.7(c) (Chapter 2).
(a)
Metal atom Metal atom

Octahedral interstices Tetrahedra! interstices
(Size: 0.15 aQ) (Size: 0.08 aQ)
(b)
Metal atom Metal atom

Octahedral interstices Tetrahedral interstices
(Size: 0.07 a0) (Size: 0.13 a 0)
Fig. 7.35. Schematic representation of octahedral and tetrahedral lattice sites in different crystal struc-
tures; (a) Face-centred cubic (fee) structures, (b) Body-centred cubic (bcc) structures.
In addition to the interstitial fraction, hydrogen may be present in the form of molecular
(gaseous) hydrogen trapped in micro voids or plane lattice defects. This amount is frequently
referred to as residual hydrogen, and can in many cases overshadow the equilibrium hydrogen
content. For example, at room temperature the maximum solubility of atomic hydrogen in the
iron lattice is estimated to be 0.001 to 0.01 ppm, while the analytical hydrogen content of
steels varies typically from 1 to 10 ppm. This supersaturation is formidable and provides the
necessary driving force for trapping of gaseous hydrogen in the microstructure.
7.2.3.3 Diffusivity of hydrogen in steel

Published data for the diffusivity of hydrogen in steels are summarised in Fig. 7.36. At high
temperatures, the diffusivity of hydrogen in ferritic steels is in reasonable agreement with the
reported value for lattice diffusion of hydrogen in bcc iron. However, when the temperature
drops below say 2000C, both the scatter and the discrepancy become more apparent due to the
phenomenon of hydrogen trapping. Inclusion of the trapping effect has led to the introduction
of an apparent diffusion coefficient for hydrogen in ferritic steels, D%pp , which according to
Oriani66 is given by:
(7-12)
where D^ is the lattice diffusion coefficient for hydrogen in bcc iron, K is the density of trap
sites (i.e. number of trap sites per number of lattice sites), and EB is the binding energy be-
tween hydrogen and the trap site.
A graphical representation of equation (7-12) is shown in Fig. 7.36. A closer inspection of
the graph reveals that the predicted temperature dependence of the apparent diffusion coeffi-
cient is in fair agreement with the reported diffusivity data for hydrogen in steel. Moreover, it
is interesting to note that the hydrogen diffusion coefficient in austenite is nearly two orders of
magnitude lower than the corresponding value for the ferrite phase at a given temperature.
This observation is not surprising, considering the pertinent difference in the packing density
between the fee iron lattice and the bcc iron lattice (74% and 68%, respectively). Thus, for
diffusion of hydrogen in austenite, we have:57
(7-13)
where T is the absolute temperature (in K).
7.2.3.4 Diffusion of hydrogen in welds

The thermodynamics and kinetics of hydrogen absorption in the weld pool have previously
been discussed in Section 2.8 (Chapter 2).
Since hydrogen is a very mobile atom (and therefore is easily lost to the surroundings), the
hydrogen concentration will vary both in the longitudinal direction and in the through thick-
ness direction of the weld, as shown in Fig. 7.37. This process will continue even after the
weld has cooled down to room temperature due to the characteristic high diffusivity of atomic
hydrogen in ferrite (see Fig. 7.38).
Temperature, 0C
Hydrogen diffusion coefficient, m2/s
Ferritic steels
Trapping
theory
Austenitic
steels
1000/T1K'1
Fig. 7.36. Summary of reported diffusion coefficients of hydrogen in iron and steel. Data compiled by
Coe57 and Yurioka and Suzuki.58
Several successful attempts have been made in the past to model hydrogen diffusion in
welds by means of numerical methods.68"70 Unfortunately, none of these solutions are simple
enough to get a good overall indication of the hydrogen redistribution during cooling and
subsequent PWHT. As an illustration of principles, we shall therefore present a simplified
analytical solution to the hydrogen diffusion problem in welding, based on an analogy be-
tween diffusion and heat conduction.
Diffusion model
The idealised model considers a butt weld of uniform hydrogen concentration in the longitudi-
nal direction, as shown in Fig. 7.39. The width of the fusion zone is 2L, while the initial
hydrogen concentration at the time of solidification (i.e. at t = 0) is Q. The hydrogen con-
centration in the base metal outside the fusion zone is C0. If element losses to the surroundings
are neglected, the problem can be treated as uniaxial diffusion in an isotropic solid analogous
to that described in Section 1.7 (Chapter 1) for heat conduction in thermit welding. Thus, in
the limiting case where the diffusion coefficient can be regarded as constant, the hydrogen
concentration ( Q as a function of time (t) and distance (y) is given by equation (1-22):
Deposited metal
ml H 2 /10Og
Fused metal
Position x, mm
ml H 2 /100g
Mean value
Diffusible
Residual
10 mm
Fig. 7.37. Measured longitudinal and lateral distributions of hydrogen in a single pass SMA weld quenched
right after welding. Data from Christensen et al.61
(7-14)
where D** is the hydrogen diffusivity, and erf(u) is the Gaussian error function (defined previ-
ously in Appendix 1.3, Chapter 1).
In practice, it is necessary to rewrite equation (7-14) in a differential form to allow for the
variation in the hydrogen diffusion coefficient with temperature. After some manipulation, we
obtain:
(7-15)
H I/cm3
15 mm
a)
p. I/cm3
b)
Fig. 7.38. Redistribution of hydrogen following welding (numerical calculations); (a) Right after weld-
ing, (b) After 12 h at room temperature (ljil cm"3 = 0.0115 ppm). Data from Christensen.68
Fusion zone
Fig. 7.39. Sketch of idealised hydrogen diffusion model.

This differential equation can be integrated numerically in temperature-time space when
the weld thermal programme is known. The boundary conditions are as follows:
when when
when when
Case Study (7.2)

Although the above model does not give a true physical picture of the hydrogen redistribution
in butt welds, it may provide valuable quantitative information about the extent of hydrogen
diffusion occurring during cooling from the solidification temperature under different welding
conditions. Figure 7.40(a) shows a sketch of a 2mm thick single pass butt steel weld made by
means of the GTA process. For the purpose of convenience we shall assume that the tempera-
ture field around the heat source is given by the simplified Rykalin thin plate solution (equa-
tion (1-100) in Chapter 1). Thus, at (T-T0)= 15000C a total width of the fusion zone of about
4.3mm is obtained for a net heat input of 67.2 J mm"2. The corresponding distance from the
weld centre-line to the 13500C HAZ isotherm is 2.5mm.
Figure 7.40(b) shows computed temperature and hydrogen concentration profiles at the
centre of the weld (y = 0) and in the grain coarsened HAZ (y = 2.5mm) for a chosen ambient
temperature of 200C (no preheating). As expected, the hydrogen concentration within the
fusion zone itself (v < L) decreases in a monotonic manner as the weld cools down. In contrast,
the hydrogen concentration outside the fusion zone (y > L) is seen to pass through a local
maximum. In the absence of preheating this maximum is attained after very long times, which
may initiate hydrogen cracking in the grain coarsened HAZ if the microstructure is martensitic.
The picture is completely changed if the ambient temperature is raised to 1000C (moderate
preheating). As shown in Fig. 7.40(c), the main effect of preheating is to decrease the cooling
rate in the low-temperature regime (i.e. below 5000C) after the completion of the austenite to
ferrite transformation. The HAZ microstructure is therefore not significantly altered, but in-
stead more hydrogen is allowed to diffuse out of the weld region before the temperature drops
below the critical value where hydrogen cracking may occur. This is seen as a shift in the peak
hydrogen concentrations towards lower absolute values and shorter times in Fig. 7.40(c).
7.2.3.5 Factors affecting the HAZ cracking resistance

Safety against hydrogen cracking is an important aspect of weldability. In spite of the knowl-
edge accumulated and the improvements made over the past decades, the current trends to-
wards stronger steels and heavier sections still require continuous attention to the risk of cracking.
Cold cracking test methods
In principle, a proper weldability criterion should enable the user to select combinations of
steel, consumables and operational conditions that will ensure sufficient crack safety at a mini-
mum of total cost. It should also enable him to examine the effects of these main variables
separately and establish a quantitative grading system for safety which takes into account the
consequences of failure.
In recognition of this situation, a number of empirical cracking test methods has been de-
Tp =
HAZ iaiMMlTOMsl 1 3500C
(b)
Temperature-time Solid curves: Centre - line
programme Broken curves: HAZ
T0 = 200C
(C-C0)Z(C1-C0)
Temperature, 0C
Hydrogen
concentration
Time, s
(C)
Temperature-time Solid curves: Centre - line
programme Broken curves: HAZ
T0 = IOO0C
(C-C0)Z(C1 -C 0 )
Temperature, 0C
Hydrogen
concentration1
Time, s
Fig. 7.40. Computed temperature and hydrogen concentration profiles during GTA butt welding of a
2mm thin steel sheet (Case Study (7.2)); (a) Sketch of weld, (b) Redistribution of hydrogen in the ab-
sence of preheating, (c) Redistribution of hydrogen after preheating to 1000C.
veloped over the years to study the mechanisms of hydrogen cracking in weldments (e.g. see
the review of Yurioka and Suzuki58). Broadly speaking, the cold cracking tests fall into either
one of the two categories, i.e. self-restrained tests or externally loaded tests. Examples of the
former type are the Tekken (oblique Y-groove) cracking test, the CTS (controlled thermal
severity) cracking test, and the cruciform cracking test. Well-known externally loaded tests
are the implant cracking test, the TRC (tensile restraint cracking) test, and the RRC (rigid
restraint cracking) test.
The implant method

The implant technique is a good example of a cracking test method which allows separate
assessment of the various metallurgical and operational factors that contribute to hydrogen
cracking in welds. As shown in Fig. 7.41 the Scandinavian version of the implant test employs
a cylindrical test bar made from the steel to be examined which is notched (threaded) at some
distance from one of its ends.71 The bar is inserted into a reamed hole in the backing plate of
a similar grade of steel, so that the threaded end is flushed with the plate surface. In order to
execute a test, the hole is sealed off with a single weld bead deposited on the plate under
carefully controlled conditions. After a specified delay of 60 s per kJ mm"1 gross heat input,
the implant assembly is quenched and put under a static tensile load at room temperature. The
stress applied is subsequently reported on the nominal cross-section of the bar. After a speci-
fied loading time (if no rupture has occurred), the assembly backing plate with the test bar may
be sectioned and examined with respect to microcracks in the HAZ.
The implant test may be performed at different stress levels, cooling rates and hydrogen
contents. The implant rupture strength, RIR, is obtained in a programme of stepwise loading,
going down or up by a fixed amount depending on rupture or not. It is then defined as the
nominal stress at which the statistically probability of rupture is 50% when the load is applied
for a long period.
Implant test results

The critical stress for rupture in the Scandinavian version of the implant test may be written in
a differential form as a sum of four contributions:
Test weld
Implant test Backing

plate
Tensile load
Fig. 7.41. Schematic representation of the implant test method.

(7-16)
The CEW parameter in equation (7-16) refers to the so-called HW carbon equivalent, origi-
nally developed for C-Mn steels:
(7-17)
Moreover, A/333 is a hydrogen diffusional parameter which takes into account variations in
the measured implant rupture strength after various thermal treatments (including preheating
and PWHT). According to Christensen and Simonsen,71 the extent of hydrogen diffusion
which occurs in the low-temperature regime can be reported in the form of an equivalent
isothermal hold time at 6O0C (or 333K), defined as:
(7-18)
where Tc refers to the local HAZ temperature at the moment of quenching (usually taken as
1000C).
It follows from equation (7-16) that the two first members reflect the influence of micro-
structure upon the implant rupture strength, and is therefore related to the HAZ peak hardness.
The two last members take into account the effect of analytical HFM and local hydrogen con-
centrations. As shown in to Fig. 7.42, the numerical values of the partial derivatives dRIR/
dCEw, dRIR/dAts/5, 3RIR/3 log HFM, and dRIR I BÂt333 may vary within relatively wide limits,
depending on the steel chemical composition and the operational conditions applied. Never-
theless, the concept is still useful for quantitative predictions of the HAZ cracking resistance,
as illustrated below.
Example (7.4)
Experience has shown that conventional pipeline steels with carbon equivalent CEW up to
0.4% can be welded with basic electrodes (Af8/5 ~ 8-9 s) without the use of external preheating,
provided that the weld metal hydrogen content is kept sufficiently low (HFM ~ 4 ppm). Sup-
pose that the same procedure shall be employed in hyperbaric welding of pipeline steels at a
depth of 320 m (33 bar total pressure). Based on the implant test data in Fig. 7.42, estimate the
minimum reduction in the steel carbon equivalent (ACEW) which must be incorporated in the
specifications to compensate for the increased hydrogen absorption observed at such depths
(#FM-10ppm).
Solution
The concept of partial derivatives implies that we will have the same safety against hydrogen
cracking if there is no net change in the implant rupture strength (i.e. ARm = 0). Since the weld
(a)
HSLA
R|R,MPa
steels
CEW,%
(b)
HSLA steel
R|R, MPa
At8/5, S
Fig. 7.42. Examples of implant test results: (a) Rm vs CEW, (b) RIR vs Ar8/5.
cooling programme is similar in both cases, the variation in A/8/5 and Ât333 can be neglected.
Hence, equation (7-16) reduces to:
(C)
HSLA steel
R1R, MPa
HFM, ppm
(d)
R|R, MPa
Quenched and tempered steels
^ • • "
Fig. 7.42. Examples of implant test results (continued); (c) RIR vs HFM, (d) RIR vs -Âr333 . Data from
Christensen and Simonsen.71
In the present example the total change in the weld metal hydrogen content between 1 and
33 bar is equal to:
Moreover, the numerical values of dRIR/dCEw and dRIR/d log HFM can be read from Fig.
7.42(a) and (c), respectively. When A/8/5 ~ 8.6 s, we obtain:
and
This gives:
The above calculations suggest that the CEW carbon equivalent of pipeline steels should not
exceed 0.35% if hydrogen cracking is to be avoided under hyperbaric welding conditions.
7.2.4 H2S stress corrosion cracking

Hydrogen sulphide (H2S) stress corrosion cracking is a well-known phenomenon taking place
in steels in environments containing sour oil and gas. As shown in Fig. 7.43 this type of
cracking arises from corrosion reactions with subsequent absorption of hydrogen in the metal.
The embrittlement mechanisms are therefore similar to those reported for hydrogen cracking
in steels, and can be evaluated from standard test methods.72
7.2.4.1 Threshold stress for cracking

For a given combination of steel, microstructure and H2S concentration there exists a lower
limit for the imposed stress where cracking no longer will occur. Experience has shown that
the threshold stress, vth, is related to the yield strength RpQ2 through the following equation:73
(7-19)
Equation (7-19) predicts that the threshold stress (and thus the steel cracking resistance)
passes through a local maximum as the yield strength increases. The locus of this peak stress
is obtained by setting dath/dRpo2 = 0, which gives Rpo2 ~ 600MPa and crth (max) ~ 360MPa.
In practice, a hardness criterion rather than a yield strength criterion is used for ranking of
steels with regard to H2S stress corrosion cracking resistance. According to Dieter,19 the fol-
lowing relation exists between Rpo and HV:
(7-20)
where m is the strain hardening exponent in the Ludwik equation.

The value of m may vary between wide limits, depending on the steel chemical composi-
tion and the heat treatment conditions applied, but for HSLA steels m ~ 0.15 is a reasonable
compromise.19 In that case the observed maximum in the cracking resistance at Rp02 ~600MPa
Short distance between
anode and cathode sites
Stress
Anode:
Crack Embrittled zone
Cathode:
Stress
Fig. 7.43. Mechanisms of hydrogen absorption in cathodic stress corrosion cracking (schematic).
corresponds to a hardness of about 250VPN. This value should be compared with the maxi-
mum hardness level of 22HRC (Rockwell C) or 248VPN incorporated in many offshore speci-
fications.
7.2.4.2 Prediction of HAZ cracking resistance

Prediction of the H2S stress corrosion cracking resistance based on equation (7-19) requires
quantitative information about the HAZ strength level. The peak strength for various combi-
nations of steels and welding conditions can easily be read from diagrams of the type shown in
Fig. 7.20 or calculated from diverse empirical formulae.2674 Figure 7.44 shows examples of
computed cr^ - At^ profiles for three different types of steels spanning a range in the CEn
carbon equivalent from 0.41 to 0.50%. It is evident from these plots that the cooling time
required to obtain the maximum threshold stress depends on the steel chemical composition.
In general, ultra-low-carbon steels will exhibit the highest resistance against stress corrosion
cracking, since a low HAZ hardenability will eliminate problems with martensite formation in
the grain coarsened region during the 7 to a transformation. On the other hand, these steels
suffer from a severe HAZ softening during high heat input welding, with consequent reduction
in the threshold stress. Under such conditions it may be safer to use ordinary low-carbon
microalloyed steels or C-Mn steels with a higher HAZ hardenability to compensate for the
observed strength loss.
Equation (7-19) can also be employed for assessment of the relevance of current hardness
requirements. Figure 7.45 shows the same types of plots as in the preceding figure (on a
normalised scale) where typical ranges for the HAZ hardness are indicated. It is evident that
the threshold stress is rather insensitive to small variations in the hardness level. Only in cases
where the HAZ hardness exceed 300VPN or drops below say 220VPN a significant deteriora-
tion in the H2S stress corrosion cracking resistance is to be expected. This suggests that the
maximum hardness requirement of 248VPN incorporated in many offshore specifications is
too stringent, since it imposes severe restrictions on the use of steels in welded structures
without improving the service performance to any great extent.
Next Page
Threshold stress, MPa
Ultra-low-carbon steel
Low-carbon steel
C-Mn steel
A
WS
Fig. 7.44. Computed a f/l -A% 5 profiles for selected steels.
Normalized threshold stress
Ultra-low-carbon steel
Low-carbon steel
C-Mn steel
A s
w
Fig. 7.45. Effect of peak hardness on the HAZ stress corrosion cracking resistance.
Previous Page
7.3 Stainless Steel Weldments
Stainless steels are widely used in various industries where corrosion is of particular concern.
These materials can be classified into four main categories, based on their microstructure:
(i) Austenitic stainless steels.

(ii) Austenitic-ferritic (duplex) stainless steels.
(iii) Ferritic stainless steels.
(iv) Martensitic stainless steels.
As shown in Table 7.3, the welding of stainless steels is encumberred by a number of differ-
ent metallurgical problems, including solidification cracking, hydrogen cracking, precipita-
tion reactions, and grain growth. Some of these problems will be discussed below in the light
of information available in recent literature.75"78
7.3.1 HAZ corrosion resistance

Austenitic stainless steels containing about 0.07 wt% carbon or more are often susceptible to
intergranular corrosion in the weld HAZ due to chromium carbide precipitation (see Section
6.4 in Chapter 6). This phenomenon is frequently referred to as the 'weld decay' in the scien-
tific literature.76
Table 7.3 Characteristics of stainless steel weldments.
Chemical Composition
Material Major Elements Minor Elements Welding Problems
Austenitic Cr: 17-25 wt% • HAZ chromium carbide precipitation

stainless steels1"7' Ni: 8-20 wt% Mo, Ti, Al, Nb • HAZ grain growth
Mn: ~2 wt% • Solidification cracking
C: 0.03-0.15 wt%
Duplex stainless Cr: 18-27 wt% • Shift in the HAZ austenite/ferrite
steels1"'*'' Ni: 7-8 wt% Ti, N balance
Mo: \-A wt% • Reduced HAZ toughness
C:<0.1 wt% • Precipitation of o-phase, Cr2N etc.
Ferritic stainless Cr: 12-30 wt% • HAZ grain growth
steels1"* Mn: 1-1.5 wt% Al, V, Zr • Low HAZ toughness
C:<0.15wt% • HAZ chromium carbide precipitation
Martensitic Cr: 11-18 wt% • Hydrogen cracking
stainless steels1"'*'5 Mn:~1.0wt% Ni, Mo, W, V • High HAZ hardness and
C: 0.15-1.2 wt% transformation stresses
• Low HAZ toughness
High resistance against general corrosion

High resistance against stress corrosion cracking (in the presence of chlorides)
High strength
Adequate low-temperature toughness
Like weld decay, knife-line attack is also related to intergranular C^Q-precipitation, but it
differs from the former type in that the corrosion attack occurs in a narrow region adjacent to
the fusion boundary following multipass welding of Ti/Nb stabilised stainless steels. As shown
in Fig. 7.46, the origin of knife-line attack can be attributed to dissolution of TiC or NbC
during the initial weld thermal cycle in regions of the HAZ where the peak temperature has
been above say 12500C. Since the cooling rate within the high temperature regime is very
high, reprecipitation of TiC or NbC does not take place on cooling. This leaves carbon free to
react with chromium during subsequent reheating to peak temperatures of about 700-8000C
following deposition of the second layer (filler pass).
Knife-line corrosion attack is usually observed in weldments where the second pass (top)
(a)
Dissolution of
TiC or NbC
Temperature
C-curve for precipitation

of TiC or NbC
1st thermal
cycle
Time
(b)
Temperature

of TiC or NbC

OfCr C
23 6
2nd thermal
cycle
Time
Fig. 7.46. Mechanisms of knife-line corrosion attack in austenitic stainless steel weldments (schematic);
(a) Dissolution of TiC or NbC during the initial weld thermal cycle, (b) Precipitation of Cr23C6 in the
low peak temperature region of the weld HAZ following deposition of the second layer. (The
corresponding C-curves for precipitation of TiC/NbC and Cr23C6 in the grain growth zone are displaced
far to the right in the diagram.)
overlaps the lower part of the fusion boundary of the first (root) pass, as indicated in Fig.
7.47(a). In practice, the problem can be eliminated by simply reversing the welding sequence
or by changing the welding procedure so that the second pass overlaps the middle rather than
the lower part of the root pass. The latter point is illustrated in Fig. 7.47(b).
732 HAZ strength level

In single phase materials such as austenitic stainless steels, the primary grain size will make a
direct contribution to strength. In general, the grain size dependence of the yield stress (<3O) is
well described by the Hall-Petch relation:19
(7-21)
where ct is the friction stress (representing the overall resistance of the crystal lattice to dislo-
cation movement), k is the locking parameter (which measures the relative hardening contri-
bution of the grain boundaries), and D is the average grain diameter.
It follows from the analysis in Section 5.4.2.5 (Chapter 5) that the grain size across the HAZ
of austenitic stainless steel weldments may vary by a factor of three to five, depending on the
(a)
Cr-depleted
region
Knife-line corrosion attack
(b)
Cr-clepleted
region
No corrosion attack
Fig. 7.47. Effect of welding performance on the corrosion resistance of Ti/Nb-stabilised austenitic stain-
less steels; (a) Low resistance against knife-line corrosion attack, (b) High resistance against knife-line
corrosion attack. The diagrams are based on the ideas of Kou.76
applied heat input. This means that a permanent soft zone will form adjacent to the fusion
boundary after welding, which may reduce the overall load-bearing capacity of the joint.
Example (7.5)
Consider plasma arc butt welding of a 5mm thick plate of type 316 austenitic stainless steel
Provided that the conditions for one dimensional heat flow are met, estimate on the basis of
the nomograms in Fig. 5.30(b) (Chapter 5) the variation in the austenite grain size across the
HAZ after welding. Calculate then via the Hall-Petch relation (equation (7-21)) the expected
reduction in the HAZ strength level due to this change in the microstructure.
Input data:
Base metal yield strength: 300 MPa
Locking parameter in Hall-Petch relation: k = 227 MPa (im1/2
Solution
First we need to calculate the net heat input per mm2 of the weld:
Readings from the nomograms in Fig. 5.30(b) give the HAZ grain size profile shown in Fig.
7.48. In order to obtain the resulting HAZ strength distribution, it is necessary to fix the value
of the friction stress, Oi, in the Hall-Petch relation. In the present example, we have:
from which
It follows from the graphical representation of the Hall-Petch relation in Fig. 7.48 that the
variation in the yield strength across the HAZ is rather small under the prevailing circum-
stances. In fact, the maximum HAZ strength reduction which may occur in such materials
because of grain growth is about 18%, corresponding to high heat input welding conditions.
This implies that welding does not impose severe restrictions on the design stress as long as the
steel is used in the fully annealed condition.
7.3.3 HAZ toughness

All body-centred cubic metals (including ferritic and martensitic stainless steels) show a marked
temperature dependence of the fracture toughness, as indicated in Fig. 7.49. At high tempera-
tures fracture occurs normally by ductile rupture, whereas cleavage is the dominating fracture
mode at low temperatures. This type of behaviour makes both ferritic and martensitic stainless
Austenite grain size, \im
Yield strength, MPa

Fig. 7,48. Computed HAZ grain size and yield strength profiles in a 5mm thick butt weld of type 316
austenitic stainless steel (Example (7.5)).
Absorbed energy
Test temperature
Fig. 7.49. Effect of temperature on notch toughness (schematic).
steels unsuitable for many structural applications where the HAZ toughness is of particular
concern.
In contrast, medium- and high-strength fee metals have usually such high toughness that
brittle fracture is not a problem at low temperatures, unless there is some special reactive
chemical environment. Austenitic stainless steels and aluminium alloys fall within this cat-
egory. When it comes to duplex stainless steel weldments, the situation is more complex.
Here the HAZ toughness is determined by the austenite/ferrite balance in the weld, which, in
turn, depends on the steel chemical composition and the operational conditions applied.78"80
As shown in Fig. 7.50, complete ferritisation is normally achieved in regions close to the
fusion boundary during the initial heating leg of the thermal cycle. Provided that the cooling
rate through the critical transformation temperature range is kept reasonably low, a significant
proportion of the ferrite may retransform back to austenite on cooling (see Fig. 7.51). This will
contribute to a high HAZ toughness, even at subzero temperatures, as indicated by the CVN
transition curves in Fig. 7.52. In general, an austenite content of about 30 vol% is sufficient to
avoid problems with the HAZ toughness in duplex stainless steel welds.
7.3.4 Solidification cracking

When a high alloy steel solidifies, a variety of microstructures can developed, depending on
the steel chemical composition and the cooling conditions applied. Referring to the Fe-Ni-Cr
phase diagram in Fig. 7.50, compositions on the Ni-rich side of the peritectic/eutectic liquidus
solidify as primary austenite, while those on the Cr-rich side solidify as primary delta ferrite.
The former condition is known to promote solidification cracking in welds due to partitioning
of low-melting-point segregates of sulphur and/or phosphorus to the ^Fe-IFe grain bounda-
ries.76"78
Since the solubility of sulphur and phosphorus is higher in delta ferrite than in austenite
(0.18 and 2.8 wt%, respectively vs 0.05 and 0.25 wt% in austenite), the cracking susceptibility
can normally be reduced if delta ferrite is present in sufficient amounts.76 At the same time,
the interfacial Vetting' conditions are significantly improved owing to the formation of a low-
Weld metal Liquid (L)

Partly melted zone
Delta ferrite grain growth zone
Peak temperature
9
Partly transformed zone §
I
I
Chromium content, wt%
Fig. 7.50. Schematic diagram defining different thermal regions within the HAZ of a single pass duplex
stainless steel weld.
30-45sat1300°C Fe-26.0 wt% Cr - 6.9 wt% Ni
Temperature, 0C A4- temperature: 1260 0C
% 5 - ferrite at RT
Time, s
Fig. 7.51. CCT-diagram for a duplex stainless steel. Data from Mundt and Hoffmeister.79
energy interface between yFe and o>e, which prevents spreading of the liquid along the grain
boundaries.82 For these reasons, a minimum weld metal delta ferrite content of about 5 to 10
vol% is usually specified for austenitic stainless steels.
The quantitative relationship between the delta ferrite content and the weld metal chemical
composition in austenitic stainless steels has been determined first by Schaeffler83 and later by
Delong et a/.84'85 The constitution diagram of Delong is shown in Fig. 7.53. Here the alloying
elements are grouped into ferrite formers (i.e. Cr, Mo, Si, and Nb) and austenite formers (i.e.
Ni, C, N, and Mn) to determine the corresponding chromium and nickel equivalents for a
given alloy. The Delong diagram differs from the Schaeffler diagram in that the important
nitrogen contribution also is included in the former, thus allowing a more accurate prediction
of the weld metal delta ferrite content.
Example (7.6)
Consider plasma arc butt welding of a 5mm thick plate of austenitic stainless steel under con-
ditions similar to those employed in Example (7.5). Data for the base metal (BM) and the
filler wire (FW) chemical compositions are given in Table 7.4. Use this information together
with the Delong diagram in Fig. 7.53 to determine which of the two filler wires (I or II) that
provides the highest resistance against weld metal solidification cracking under the prevailing
circumstances. In the present example we shall assume that the mixing ratio B/(B+D) is 0.57
(the mixing ratio is defined in Section 1.10.8, Chapter 1).
Solution
In the absence of oxidation losses, the weld metal chemical composition is given by the 'rule
of mixtures':
3-D heat flow Peak temperature: 13000C
conditions
Absorbed energy, J
Fig. 7.52. CVN transition curves for a duplex stainless steel after weld thermal simulation. Data from
Videm.81
Table 7.4 Chemical composition of base plate and filler wires used in Example (7.6).
Material1 wt% C wt% Cr wt% Ni wt% Mo wt% Mn wt% Si wt% N
Baseplate 0.03 17.0 12.0 2.0 2.0 0.9

Filler wire I 0.03 18.0 11.5 2.7 0.8 0.9 0.07
Filler wire II 0.03 18.0 15.5 - 0.8 0.9 0.07
Max. 0.015 wt% S and 0.03 wt% P
This leads to the weld metal chemical compositions shown in Table 7.5. We can now
calculate the Cr- and Ni-equivalents for both welds:
Nickel equivalent (%Ni+30x%C+30x%N+0.5x%Mn)
Austenite
Schaeffler A+M line
Austenite+
martensite Austenite+ferrite
Chromium equivalent (%Cr+%Mo+1.5x%Si+0.5x%Nb)
Fig. 7.53. The Delong diagram85 showing the relationship between delta ferrite content and weld metal
chemical composition for stainless steels.
Table 7.5 Computed weld metal chemical compositions (Example (7.6)).
Weld No. wt% C wt% Cr wt% Ni wt% Mo wt% Mn wt% Si wt% N
I 0.03 17.4 11.8 2.3 1.5 0.9 0.03

II 0.03 17.4 13.5 1.1 1.5 0.9 0.03
Weld I : About 93 vol% austenite and 7 vol% delta ferrite

Weld IP. 100 % austenite
From this we see that wire I provides the highest resistance against weld metal solidifica-
tion cracking for the combination of steel and operational conditions considered above.
It should be emphasised that the Delong diagram gives no information about the real solidi-
fication microstructure, since it is based on measurements of retained delta ferrite at room
temperature. Also, the important effect of cooling rate on the weld metal transformation be-
haviour is neglected in the present analysis. Consequently, the use of such empirical diagrams
for selection of steel and welding consumables is a keenly debated question in the scientific
literature.
7.4 Aluminium Weldments
Aluminium alloys are to an increasing extent used as structural components in welded assem-
blies because of their high strength, low density, and good resistance against general corro-
sion. In certain cases the application of aluminium is restricted by a low HAZ strength level
due to softening reactions occurring during welding. In other cases the cracking resistance or
the fatigue strength becomes the limiting factor, depending on the design criterion. Table 7.6
summarises typical problems associated with welding of aluminium and its generic alloys. In
the following, we shall focus on the structural and mechanical response of age-hardenable
aluminium alloys to the heat of welding processes, with particular emphasis on Al-Mg-Si
alloys and Al-SiC metal matrix composites.
7.4.1 Solidification cracking

Conventional filler metals for aluminium welding are based on the binary Al-Si or Al-Mg
systems. It follows from Fig. 7.54 that maximum cracking sensitivity occurs somewhat be-
tween pure Al and 2 to 4 wt% Si and Mg in aluminium, respectively. This behaviour arises
from the competitive influence of two different processes.76 In pure aluminium, on the one
hand, the solidification cracking problem does not exists, since there is no low-melting-point
eutectic present at the grain boundaries. In solute-rich aluminium alloys, on the other hand, the
eutectic liquid is so abundant that it backfills and 'heals' the incipient cracks, which results in
a low cracking sensitivity. Somewhere in between the two composition limits, the amount of
eutectic liquid may be just large enough to form continuous films at the grain boundaries. This,
in combination with high shrinkage or thermal contraction stresses, may lead to solidification
cracking, as indicated in Fig. 7.55.
Normally, aluminium-based filler metals contain between 4 to 5 wt% Si or Mg. This solute
content is usually high enough to prevent solidification cracking during welding. Problems
Table 7.6 Characteristics of aluminium weldments.
Chemical Composition
Material Major Elements Minor Elements Welding Problems
Al-Mg alloys Mg: 1-5 wt% Si, Fe, Mn • Solidification cracking

(5 XXX-series) • Reduced HAZ strength level
(work-hardened materials)
Al-Mg-Si alloys Mg: 0.5-1.3 wt% Mn, Cr, Fe • Solidification cracking
(6xxx-series) Si: 0.4-1.4 wt% • Hot cracking
• Reduced HAZ strength level
Al-Zn-Mg alloys Mg: 0.5^4 wt% Cu, Cr, Mn, Fe • Solidification cracking
(7xxx-series) Zn: 2-8 wt% • Hot cracking
• Low HAZ stress corrosion cracking
resistance
Al-SiC metal Si: -7.5 wt% Fe, Ti • Conventional fusion welding
matrix composites Mg: -0.5 wt% is not recommended
(10-15% vol% SiC) • High-quality joints can be
produced by means of friction
welding or diffusion bonding
AI-Si system
Relative crack sensitivity
Al-Mg system
Composition of weld, wt%

Fig. 7.54. Solidification cracking sensitivity of binary Al-Si and Al-Mg alloys. Data from Dudas and
Collins.86
may arise, however, if the weld metal becomes heavily depleted with respect to Si or Mg due
to dilution with the parent metal. Welds of a high B/(B+D) ratio fall within this category. In
such cases it may be necessary to use over-alloyed filler wires to obtain crack-free welds.
Example (7.7)
Consider plasma arc butt welding of a 10mm thick aluminium plate of type AA 6082-T6
(containing 0.7 wt% Mg and 0.9 wt% Si) under the following conditions:
qo = 2OkW, v = 10mm s"1, T0 = 200C
Experience shows that the bead reinforcement amounts to 10% of the groove cross section
(details of the groove geometry are given in Fig. 7.56). Two different filler wires are available,
wire I with 5 wt% Si and wire II with 5 wt% Mg. Use this information along with the diagrams
in Fig. 7.54 to determine which of the two filler wires (I or II) that provides the highest resist-
ance against weld metal solidification cracking under the prevailing circumstances. In these
calculations we shall assume that the temperature field around the heat source is given by the
simplified Rykalin thin plate solution (equation (1-100) in Chapter 1). Thermal data for Al-
Mg-Si alloys are contained in Table 1.1.
Solution
First we need to estimate the mixing ratio BI(B+D). The total width of the fusion zone (2ym)
can be obtained from equation (1-100):
(a)
(b)
Fig. 7.55. Examples of solidification cracking in aluminium weldments (Varestraint test coupons);
(a) Al-I wt% Si, (b) Al-I wt% Mg. After Cross.87
Fig. 7.56. Groove geometry for a single pass Al-Mg-Si butt weld (Example (7.7)).
This gives:
and
from which
In the absence of oxidation losses, the weld metal chemical composition is given by the
'rule of mixtures':
If wire I is used, the weld metal Mg and Si concentrations become:
Similarly, in the case of wire II, we get:
A comparison with Fig. 7.54 shows that wire I provides the highest safety against weld
metal solidification cracking under the prevailing circumstances, while wire II is unaccept-
able.
The lower cracking resistance of the Al-Mg wire compared with the Al-Si wire at high
dilution ratios arises from the pertinent difference in the fraction of eutectic liquid which forms
during weld metal solidification. For pure binary alloys the eutectic fraction feuL is given by
equation (3-46) in Chapter 3:
where CeuL is the eutectic concentration, and ko is the equilibrium partitioning coefficient (given
by the binary Al-Si and Al-Mg phase diagrams).
If the contribution from the accompanying alloying element is neglected, the values of feuL
become:
Weld I (1.80 wt% Si):
Weld II (1.65 wt% Mg):
From this we see that the fraction of eutectic liquid in weld I is so abundant that it backfills
and 'heals' all incipient cracks, while feut in weld II is just large enough to form continuous
films at the columnar grain boundaries which, in turn, promotes solidification cracking.
7.4.2 Hot cracking

Hot cracking is a phenomenon occurring within the high peak temperature regions of the HAZ
during welding of aluminium alloys. In spite of the difference in location, hot cracking like
solidification cracking is intergranular and arises from the combined action of grain boundary
liquation and stresses induced by solidification shrinkage and thermal contraction. The forma-
tion of liquid phases within a weld HAZ can readily be explained with the help of a simple
binary phase diagram of the type shown in Fig. 7.57. If the composition of the binary alloy is
Liquid (L)
Temperature
Composition
Fig. 7.57. Schematic binary phase diagram defining the equilibrium conditions for partial melting during
reheating.
higher than Cmax^ and the alloy is heated to a temperature above TeuU partial melting will
occur. Since the eutectic phase is usually located at the grain boundaries, these sites become
immediately covered with liquid films if the wetting conditions are favourable.
7.4.2.1 Constitutional liquation in binary Al-Si alloys

Experience shows that even binary alloys with a nominal composition lower than Ceut can
undergo incipient melting at the eutectic temperature. This will obviously be the case if the
alloy contains segregations so that the concentration locally exceeds the critical composition.
Another possibility is melting due to 'constitutional liquation', a mechanism originally pro-
posed by Pepe and Savage8889 for the formation of hot cracking in 18-Ni maraging steel
weldments at temperatures well below the bulk solidus of the alloy. The same theory has later
been applied by Reiso et al.90~93 and Lohne and Ryum94 to explain incipient melting during
homogenisation heat treatment of aluminium alloys. In the following, the treatment of Lohne
and Ryum94 is adopted to illustrate the principles of constitutional liquation in binary Al-Si
alloys under conditions applicable to welding. The theory is later extended to ternary Al-Mg-
Si alloys.
Figure 7.58 shows a section of the Al-rich corner of the binary Al-Si phase diagram. Con-
sider next a pure Al-1.2 wt% Si alloy that first is brought to equilibrium at a lower temperature
to form 10|um large Si particles (Fig. 7.59(a)) and then is rapidly heated to a higher tempera-
ture slightly above Teut. If local equilibrium is maintained at the particle/matrix interface
during heating, the composition is given by the Al-Si solvus boundary in Fig. 7.58. When Ceut
is reached incipient melting will take place by formation of eutectic liquid (according to the
reaction Al + Si —» liquid), as shown in Fig. 7.59(b). The process will continue until the whole
Si particle has dissolved (Fig. 7.59(c)). Depending on the density/location of the Si particles
and the interfacial wetting conditions, grain boundary liquation may eventually occur, as indi-
cated in Fig. 7.59(d).
Liquid
Liquid+solid
Temperature, 0C
ss
Al+Si
Fig. 7.58. Section of the Al-rich comer of the binary Al-Si phase diagram. Data from Ref. 95.
(a) (b)
(C) (cO
Fig. 7.59. Optical micrographs showing the microstructural evolution during homogenisation heat treat-
ment of binary Al-Si alloys; (a) Isolated Si particle embedded in a matrix of Al, (b) Partially melted Si
particle surrounded by Al-Si eutectic (25s at 582°C), (c) Globular Al-Si eutectic structure formed after
complete dissolution of the Si particle (60s at 582°C), (d) Spreading of eutectic liquid along a grain
boundary (25s at 582°C). Courtesy of O. Lohne, Sintef - Division of Metallurgy, 7034 Trondheim,
Norway.
7.4.2.2 Constitutional liquation in ternary Al-Mg-Si alloys

Constitutional liquation in ternary Al-Mg-Si alloys has been investigated by Reiso et al.96
Figure 7.60 shows a section of the Al-rich corner of the quasi-binary Al-Mg2Si phase diagram.
When an alloy of composition C < Ceut is heated up rapidly, as in the case of welding, the
primary Mg2Si particles do not have enough time to dissolve. When the quasi-binary eutectic
temperature is reached at about 593°C, incipient melting will occur, according to the reac-
tion:96
liquid (7-22)
Further heating to a higher temperature (T > 593°C) provides additional time for dissolu-
tion of Mg2Si and formation of more liquid of variable composition. In practice, reaction (7-
Temperature, 0C
Mg2Si concentration, wt%
Fig. 7.60. Quasi-binary section of the Al-rich corner of the ternary Al-Mg-Si phase diagram. Data from
Reiso et al.96 and Phillips.97
22) may also proceed below the quasi-binary eutectic temperature Teut, since the diffusion rate
of Mg is higher than that of Si.96 By considering the kinetics it can be shown that the interface
concentration will gradually move towards the silicon side of the quasi-binary line in the phase
diagram as the Mg2Si particles dissolve, thereby reducing the temperature at which liquation
occurs (from 593 down to about 5800C). Moreover, in alloys containing excess amounts of
silicon, two other side reactions can take place:96
liquid (7-23)
and
liquid (7-24)
The former is analogous to the melting reaction in binary Al-Si alloys (Fig. 7.58), while the
latter corresponds to the eutectic reaction in the ternary Al-Mg-Si system (Teut varies from
555 to 559°C, depending on the source). Since this eutectic temperature represents the lowest
temperature at which a melt can exist within the system, it means that local melting of second
phase particles cannot take place below, say, 555 to 559°C during welding of Al-Mg-Si al-
loys.
7.4.2.3 Factors affecting the hot cracking susceptibility
In addition to the thermodynamic and kinetic effects mentioned above, there are several other
factors, some interrelated, which play an important part in the formation of hot cracks in
Al-Mg-Si weldments. These are:76
(i) The number density and size distribution of Mg2Si and Si particles in the base
metal.
(ii) The total grain boundary area per unit volume available for absorption of eutectic
liquid (determined by the HAZ grain size).
(iii) The interfacial wetting conditions.
(iv) The local tensile stress level in the partially melted region.
It follows that the hot cracking susceptibility may be significantly altered by a change in
one of these parameters, but, in practice, there is very little that can be done to prevent grain
boundary liquation if the base material already contains second phase particles. In fact, the
only useful way of eliminating the cracking problem is to reduce the tensile stress in the par-
tially melted HAZ through proper selection of welding consumables. This is because the
composition of the weld metal can be adjusted so that solidification is completed first in the
partially melted region and then in the fusion zone, thus avoiding hot cracking in the former.
As an illustration of principles, the diagrams of Gittos and Scott98 (reproduced in Fig. 7.61)
will be considered. These diagrams show the variation of the weld metal solidus temperature
with base metal dilution for two commercial filler wires (wire I: Al-5 wt% Si, and wire II: Al-
5 wt% Mg). It is evident that the risk of hot cracking is highest when wire II is used, particu-
larly at high B/(B+D) ratios, since solidification occurs first in the weld metal and then in the
partially melted region. In contrast, wire I provides a good HAZ cracking resistance over the
whole composition range because of the resulting lower solidus temperature. In the latter case
the weld metal solidification and thermal contraction stresses are imposed on the HAZ at a
stage where liquid no longer exists at the grain boundaries.
Example (7.8)
Consider plasma arc butt welding of a 10mm thick aluminium plate of AA6082-T6 (contain-
ing 0.7 wt% Mg and 0.9 wt% Si) under conditions similar to those employed in Example (7.7).
Use the diagrams in Fig. 7.61 to determine which of the two filler wires (I or II) that provides
the highest resistance against hot cracking in the partially melted region under the prevailing
circumstances.
Solution
In the previous example the base metal dilution ratio, BI(B+D), was found to be 0.78. Under
such conditions wire I provides the highest resistance against hot cracking, since solidification
first occurs in the partially melted region and then in the weld metal. This explains why Al-Si
based filler wires are usually recommended for single pass butt welding of Al-Mg-Si extru-
sions if the strength level is not of particular concern.
(a)
Solidus temperature, 0C
liquid
B/(B+D), %
(b)
Solidus temperature, 0C
liquid
B/(B+D), %
Fig. 7.61. Variation of weld metal solidus temperature with dilution for 6082 aluminium alloys (Al-0.7
wt% Mg-0.9 wt% Si); (a) Wire I (Al-5 wt% Si), (b) Wire II (Al-5 wt% Mg). The melting temperature
for different base metal constituent phases are indicated by the horizontal broken lines in the diagrams.
Data from Gittos and Scott98 and Reiso et al.96
(a) Peak temperature
Partly reverted region:
Fully reverted region

Hardness
HAZ-
(b)
C-curve for
Temperature
precipitation
of (3'(Mg2 Si)
Weld thermal cycle
Time
Fig. 7.62. Schematic diagrams showing the sequence of reactions occurring in the HAZ of 6082-T6
aluminium welds; (a) Hardness distribution following p"(Mg2Si) dissolution; (b) Precipitation of P'(Mg2Si)
at dispersoids during the weld cooling cycle.
7A3 HAZ microstructure and strength evolution during fusion welding
The age-hardenable Al-Mg-Si alloys have been widely studied, to the extent that most of the
underlying physical processes are well established. They offer tensile strength values higher
than 350 MPa in the artificially aged (T6) condition owing to the presence of very fine, needle-
shaped 3"(Mg2Si) precipitates along <100> directions in the aluminium matrix." Although
Al-Mg-Si alloys are readily weldable, they suffer from severe softening in the heat affected
zone (HAZ) because of reversion (dissolution) of the (3"(Mg2Si) precipitates during the weld
thermal cycle.76100"103 This type of mechanical impairment represents a major problem in en-
gineering design, since it reduces the load-bearing capacity of the joint.104
7.4.3.1 Effects of reheating on weld properties

Microstructural changes in the HAZ of Al-Mg-Si alloys have been examined by several in-
vestigators,100"103 and the main results are summarised in Fig. 7.62.
It is evident from Fig. 7.62(a) that reversion of [3"(Mg2Si) precipitates will occur to an
increasing extent in the peak temperature range from 250 to 5000C. This is associated with a
continuous decrease in the HAZ hardness until the dissolution process is completed. During
cooling of the weld, some solute recombines to form coarse, metastable (3'(Mg2Si) precipitates
which do not contribute to strengthening (Fig. 7.62(b)). However, close to the fusion bound-
ary a large fraction of alloying elements will remain in solid solution at the end of the thermal
cycle, thereby giving conditions for extensive age-hardening at room temperature over a pe-
riod of 5 to 7 days (Fig. 7.62(c)).
In general, enhanced HAZ strength recovery can be achieved by the use of artificial ageing
in the temperature range from 150 to 1800C,76 but this possibility will not be considered here.
(C) Peak temperature
Contribution;
from natural:
ageing \
Hardness
Resulting hardness
profile
Fig. 7.62. Schematic diagrams showing the sequence of reactions occurring in the HAZ of 6082-T6
aluminium welds (continued); (c) Hardness distribution after prolonged room temperature ageing.
7.4.3.2 Strengthen ing mechan isms in A l-Mg-Si alloys
Due to the lack of experimental evidence of coherency strains around (3"(Mg2Si) precipitates
in Al-Mg-Si alloys," it has been suggested that the increased resistance to dislocation motion
accompanying the presence of these structures arises from the high energy required to break
Mg-Si bonds in the particles as dislocations shear through them. Assuming that this strength-
ening effect is associated with order hardening, the net precipitation strength increment, Aop,
can be calculated from the equation originally derived by Kelly and Nicholson:105
(7-25)
where 7/ is the internal interface (or antiphase) boundary energy, b is the Burgers vector,/is
the particle volume fraction, and c$ is a kinetic constant.
By introducing the relative particle volume fraction, f/fo, we obtain:
(7-26)
where fo is the initial volume fraction of (3"(Mg2Si) precipitates in the alloy, and C4 is a new
kinetic constant (equal to c 3 / o ).
It is evident from equation (7-26) that Aop = Aap (max) = c4 when/7/ o = 1. Hence, this
equation can be rewritten as:
(7-27)
Here amin denotes the intrinsic matrix strength after complete particle dissolution, while
dmax is the original base metal strength in the artificially aged (T6) condition.
Provided that a linear relationship exists between yield strength and hardness, equation (7-
27) can be rewritten as:102
(7-28)
Equation (7-28) provides a basis for assessing the reaction kinetics through simple hardness
measurements. Typical values for amax, HVmax, omin, and HVmin are given in Table 7.7.
7.4.3.3 Constitutive equations

The kinetics of (3"(Mg2Si) dissolution during reheating of 6082-T6 aluminium alloys have
been considered in Section 4.4.2 (Chapter 4).
Table 7.7 Properties of some Al-Mg-Si and Al-SiC weldments.
HVmax HVmin omax <5min

Type of Weld Material [VPN] [VPN] [MPa] [MPa] ® W
Fusion welds 6082-T6 110 42 282 78 0.56

Friction welds 6082-T6 115 47 297 93 0.49 0.43
Al-SiC-T6 137 55 342 83 0.55 0.43
Reversion model
When the 3"(Mg2Si) particles dissolve, the volume fraction falls from its initial value f0, ac-
cording to equation (4-36). A combination of equations (4-36) and (7-28) leads to the follow-
ing expression for the dimensionless strength parameter within the partly reverted region of
the HAZ:
(7-29)
where t\ is the maximum hold time required for complete particle dissolution at a given tem-
perature (defined by equation (4-31) in Chapter 4), and n is a time exponent (< 0.5). The
variation of n with/// o is shown in Fig. 4.20.
Natural ageing model

In general, the fraction of hardening (3"(Mg2Si) precipitates which forms during natural ageing
(flfo) depends on the amount of remnant solute present in the matrix material after cooling of
the weld. By considering the kinetics of the C-curve for precipitation of essentially non-
hardening 3'(Mg2Si) particles at dispersoids during the weld cooling cycle, Myhr and Grong102
arrived at the following relationship for flfo'
(7-30)
where
(7-31)
Here O is a material constant (defined in Table 7.7), and t\ is the critical time required to
precipitate a certain fraction of $'(X = Xc) at an arbitrary temperature (T). The variation of
t*2 with temperature is given by equation (6-55) in Chapter 6.
The net precipitation increment following natural ageing (a2) can therefore be written as:
(7-32)
Coupling of models
Based on equations (7-29) and (7-32) it is possible to calculate the HAZ strength distribution
after welding and subsequent natural ageing when the weld thermal programme is known.
Figure 7.63 shows a sketch of the superimposed hardness profiles, as evaluated from these
equations. Since the resulting strength level in the partly reverted region depends on the inter-
play between two competing processes (i.e. dissolution and reprecipitation), it is convenient to
define the 'boundary' between the two models on the basis of the intersection point in Fig.
7.63 where a\ = a 2 , i.e.:
(7-33)
and
(7-34)
Peak temperature
Partly reverted
Fully reverted
region
region
Reversion model
Hardness
Intersection
point
, Natural ageing model

Fig. 7.63. Coupling of reversion and natural ageing models.
It follows that this locus also defines the minimum HAZ strength level, which is an impor-
tant parameter in engineering design.
7.4.3.4 Predictions of HAZ hardness and strength distribution

The predictions are based on computer programmes which utilise the medium thick plate heat
flow solution described in Section 1.10.4 (Chapter 1) and the kinetic models outlined above to
calculate the HAZ hardness distribution for specific welding conditions (details are given in
Ref. 106).
Accuracy of predictions
Examples of measured and predicted HAZ hardness profiles are shown in Figs. 7.64 and 7.65.
When stringer bead welding is carried out on a plate of medium thickness, the hardness
distribution in the transverse y direction will vary with distance from the plate surface due to a
continuous change in the heat flow conditions. A comparison between observed and predicted
hardness profiles in Fig. 7.64 shows that such effects are readily accounted for in the present
model. In contrast, a full penetration butt weld will always reveal a similar HAZ hardness
distribution in the transverse section of the weld, as shown in Fig. 7.65. This situation arises
from the lack of a temperature gradient in the through-thickness z direction of the plate. Moreo-
ver, it is evident from Figs. 7.64 and 7.65 that the final dimensions of the HAZ are strongly
influenced by variations in welding parameters and operational conditions. Hence, it is diffi-
cult to justify the use of a constant safety factor for the width of the HAZ as recommended in
current design rules for welded Al-Mg-Si alloys.104
Aptness of models
Based on the kinetic models described in the proceding sections, it is possible to construct
two-dimensional (2-D) maps which show characteristic hardness and peak temperature con-
tours in the HAZ of 6082-T6 aluminium weldments. Examples of such diagrams are given in
(a)
HV (predicted)
Hardness, VPN
Aym, mm
(b)
HV (predicted)
Peak temperature. 0C
Hardness, VPN
ym,mm
Fig. 7.64. Comparison between measured and predicted HAZ hardness profiles in a stringer bead GMA
weld; (a) Upper plate surface, (b) Lower plate surface. The peak temperature distribution is indicated by
the broken lines in the graphs. (Operational conditions: qo = 9.1 kW, v = 5.1 mm s"1, d = 15mm). Data
from Myhr and Grong.102
Fig. 7.66. Included is also a 3-D plot of the HAZ hardness distribution in the transverse
section of the weld.
The results in Fig. 7.66 reveal a direct relationship between the HAZ isothermal contours
on the one hand and the resulting HAZ hardness/strength distribution on the other. In this
particular example the soft zone closely follows the contour of the 4000C isotherm. This, in
turn, implies that the minimum HAZ strength level is fairly constant and virtually independent
of choice of welding parameters (i.e. close to 60 VPN for single pass welds).
HV (predicted)
Peak temperature. 0C
Hardness, VPN
A y m , mm
Fig. 7.65. Comparison between measured and predicted HAZ hardness profiles in a single pass plasma
arc butt weld. The peak temperature distribution is indicated by the broken line in the graph. (Opera-
tional conditions: qo = 14.0 kW, v = 5.8 mm s"1, d = 13mm). Data from Myhr and Grong.102
In practice, the HAZ hardness can be converted into an equivalent yield or ultimate tensile
strength through the following regression formulae:102
(7-35)
and
(7-36)
From equation (7-35) we see that a minimum HAZ hardness of about 60 VPN corresponds
to a strength reduction factor of:
This value is in good agreement with the recommended strength reduction factor of 0.49
incorporated in many welding specifications and standards.104'107
Process diagrams for single pass butt welds

Because of the complex temperature-time pattern in aluminium welding, it is not possible to
condense general information about the HAZ strength distribution into 2-D process diagrams.
An exception is single pass butt welding of plates, where the medium thick plate solution
described in Section 1.10.4 (Chapter 1) can largely be simplified if the net arc power is kept
sufficiently high compared with the plate thickness (e.g. qold > 0.5 kW mm"1). Under such
conditions the mode of heat flow becomes essentially one-dimensional, and the temperature
distribution is determined by the ratio qo/vd, kJ mm"2 (see equation (1-100) in Chapter 1).
y, mm
Hardness code
z, mm
Fusion zone
(a)
(b)
Fig. 7.66. Computed HAZ hardness and peak temperature contours in the transverse section of a stringer
bead GMA weld; (a) 2-D graphical representation, (b) 3-D graphical representation. (Operational condi-
tions as in Fig. 7.64). Data from Myhr and Grong.102
Figure 7.67 shows plots of the HAZ hardness/strength profiles for different values of qo Ivd.
It follows that a narrow width of the HAZ requires the use of a low energy input per mm2 of the
weld. In practice, this can be achieved by the choice of an efficient welding process (e.g.
electron beam or laser welding) which allows deposition of a full penetration butt weld with-
out employing a groove preparation (i.e. eliminates the need for filler metals).
Multipass welding
The present process model can also be extended to multipass welding if it is assumed that
reversion of indigenous (3'(Mg2Si) precipitates occurs instantaneously on reheating above the
phase boundary solvus temperature (here taken equal to 5200C).
Hardness, VPN V2[MPa] Rm [MPa]
Scale:
10 mm

Fig. 7.67. Process diagram for single pass 6082-T6 butt welds. Data from Myhr and Grong.102
Figure 7.68 shows 2-D and 3-D plots of computed HAZ hardness and peak temperature
contours for a simulated two-pass butt weld. This model system consists of two imaginary
stringer beads which are placed symmetrically on each side of a 15mm thick plate. Compara-
ble hardness data for a single pass weld of same thickness are contained in Fig. 7.66. It is
evident from Figs. 7.66 and 7.68 that deposition of a second pass will neither increase the
width nor reduce the strength of the HAZ to any great extent. The explanation lies in the fact
that the temperature field around each heat source tends to overlap with increasing distance
from the fusion boundary due to symmetry effects (see Fig. 7.68(a)), which prevents excessive
strength loss during reheating of the weld. Similar observations have also been made from
actual testing of multipass Al-Mg-Si weldments.108
Example (7.9)
Consider plasma arc butt welding of a 10mm thick aluminium plate of type AA 6082-T6 under
conditions similar to those employed in Example (7.7). Estimate on the basis of the process
diagram in Fig. 7.67 and the simplified Rykalin thin plate solution (equation (1-100) in Chap-
ter 1) both the minimum HAZ strength level, the total width of the reduced strength zone after
welding, as well as the lower temperature limit for dissolution of the (3"(Mg2Si) precipitates
during the weld thermal cycle. Thermal data for Al-Mg-Si alloys are given in Table 1.1
(Chapter 1).
Solution
First we need to calculate the net heat input per mm2 (q ol vd). In the present example, we
have:
y, mm
Hardness code
z, mm
Fusion zone
(a)
(b)
Fig. 7.68. Computed HAZ hardness and peak temperature contours for a simulated two-pass butt weld
(the second pass is deposited immediately after cooling of the first pass); (a) 2-D graphical representa-
tion, (b) 3-D graphical representation. (Operationalconditions: qo = 9A kW, v = 5.1 mms~ l ,d- 15mm).
Data from Myhr and Grong.102
and
Similarly, by considering the extension of the HAZ and the corresponding length of the
scale bar in Fig. 7.67, the total width of the reduced strength zone becomes:
The relationship between peak temperature T- Tp and distance y = ym from the heat source
can now be obtained by differentiating equation (1-100) with respect to time. After some ma-
nipulation, we obtain:
By substituting qo/vd = 200 J ram"2, pc = 0.0027 J mm"3 0C"1, and Tm = 652°C into the
above equation, the following temperature for incipient dissolution of the p"(Mg2Si) precipi-
tates is obtained:
It is obvious from the above calculations that the degree of HAZ softening occurring during
welding is substantial under the prevailing circumstances. This explains why, for instance,
high heat input deposition is usually not recommended for Al-Mg-Si alloys.
7.4.4 HAZ microstructure and strength evolution during friction welding

Friction welding is a solid state joining process that involves both heating and plastic deforma-
tion of the parent material under extreme thermal and strain rate conditions.109"113 During the
welding operation the material is deformed at high temperatures, initially at low strain rates,
but due to the axial displacement of the specimen, the material is subsequently brought into the
fully plasticised region where the strain rate may exceed 103 s"1 (e.g. see Fig. (6.79) in Chapter
6). Although the resulting microstructural changes are similar to those observed in many hot
working processes, it is obvious that transient effects must play a more dominant role in fric-
tion welding because of the rapid temperature and strain rate fluctuations. This, in turn, in-
creases the complexity of the analysis.
7.4.4.1 Heat generation in friction welding

A major problem in modelling of heat flow phenomena in friction welding is to obtain an
accurate description of the energy input at the faying interface. This is because the friction
coefficient L| L is changing continuously during the welding cycle from |i > 1 at the dry sliding
start, towards zero when the temperature for asperity melting is reached at the interface.113
However, for the ideal case considered in Fig. 7.69, the torque required to rotate two circular
shafts relative to one other under the action of an axial load pressure is given by: 114115
(7-37)
where M is the interfacial torque, R* is the surface radius, and P(r) is the pressure distribution
across the interface (here assumed constant and equal to P).
Pressure Pressure
Fig. 7.69. Schematic arrangement of friction welding of a solid rod.
If all the shearing work at the interface is assumed to be converted into frictional heat, the
average heat input per unit area and time becomes:114
(7-38)
where qo is the net power (in W), P is the friction pressure (in N mm"2), A is the cross section
(in mm2), and wmax. is the maximum surface velocity at the outer edge (in m s"1).
Equation (7-38) provides a basis for estimating the heat generation at the interface during
continuous drive friction welding in the absence of asperity melting.
7.4.4.2 Response ofAl-Mg-Si alloys and Al-SiC MMCs to friction welding

In friction welding of aluminium alloys and Al-SiC metal matrix composites, the deformation
is maintained throughout the steady state period.113 The microstructural changes that take
place at this stage of the process can be classified as dynamic changes, since they occur under
the action of variable plastic straining. Similarly, those taking place after the forge operation
(when the weld cools) are referred to as static changes due to the lack of external plastic
deformation. The former includes the generation of recovered subgrains and reversion of
constituent precipitates in the area of high temperature and high deformation, whereas the
latter involves precipitation reactions, e.g. formation of hardening [3"(Mg2Si) particles in the
heat affected zone following natural ageing.109"113
In order to assemble an adequate model for the microstructural evolution during friction
welding, it is necessary to divide the HAZ into different reaction zones. Referring to Fig. 7.70,
the following three main regions are of specific interest in this context:
(i) The fully plasticised region, ZpL, where the material is able to accommodate the
plastic strain by dynamic recovery (or recrystallisation) of the microstructure.
Contact section
Z
Fig. 7.70. Schematic diagram showing the three main reaction zones within a friction welded component
(Zpi/. fully plasticised region, Zpd\ partly deformed region, Zud: undeformed region).
(ii) The partly deformed region, Zpd., where the degree of plastic deformation is ac-
commodated by an increase in the dislocation density of the matrix grains. In this
region the temperature is sufficiently high to facilitate dissolution of the base metal
hardening precipitates.
(iii) The undeformed region, Z1^., characterised by partial reversion of the base metal
precipitates.
Aspects of HAZ subgrain evolution during continuous drive friction welding have been
described in Section 6.5.2 (Chapter 6). In the following, the structural and mechanical re-
sponse of T6 heat treated Al-Mg-Si alloys and Al-SiC metal matrix composites to the im-
posed heating and plastic deformation will be considered more in detail.
7.4.4.3 Constitutive equations

Since the HAZ strength level in both types of materials is mainly controlled by dissolution and
precipitation reactions occurring within the aluminium matrix during the weld thermal cy-
cle,109"113 we can use equations (7-29) and (7-32) with minor modifications to describe the
reversion and the natural ageing characteristics of the weld components (see Table 7.7). In
addition, it is necessary to allow for the plastic deformation introduced during friction weld-
ing, because dislocations will be generated in the matrix material to accommodate the strain.113
In general, work hardening of metals and alloys is a very complex problem which has not yet
been properly solved. However, with the precision aimed at here it is sufficient to assume that
this strength contribution is given by:
(7-39)
where 1F is a constant which is characteristic of the material under consideration (given in

Table 7.7). The form of equation (7-39) indicates that the hardness is essentially constant
within the plasticised regions of the weld HAZ.
7.4.4.4 Coupling of models

Based on equations (7-29), (7-32), and (7-39) it is possible to calculate the HAZ strength
distribution after friction welding and subsequent natural ageing when the weld thermal pro-
gramme is known. Figure 7.71 (a) and (b) show schematic representations of the superim-
posed hardness profiles, as evaluated from these models. The resulting HAZ hardness distri-
bution is indicated by the solid curve in the graphs for short and long duration thermal cycles,
respectively. Since the present treatment oversimplifies the problem by only considering the
strongest contribution with no interaction, the justification of this assumption relies on a good
correlation between theory and experiments.
It follows from Fig. 7.71 that particle dissolution is the major softening mechanism in fric-
tion welding of Al-Mg-Si alloys and Al-SiC metal matrix composites. At the same time a
substantial strength recovery will occur as a result of external plastic straining in combination
with intrinsic precipitation of hardening (3"(Mg2Si) particles following prolonged room tem-
perature ageing. Depending on the operational conditions applied, this may give rise to differ-
ences in the shape of the HAZ hardness profiles, as shown in Fig. 7.71.
(a)
Reversion model
Hardness
Work hardening model
Natural ageing model
Unaffected base material

Axial distance, Z
(b)
Reversion model
Hardness
Work hardening model
Natural ageing model Unaffected

base material
Axial distance, Z
Fig. 7.71. Schematic representation of the HAZ hardness distribution after friction welding and subse-
quent natural ageing; (a) Short duration thermal cycle, (b) Long duration thermal cycle. The parameters
Zp/., Zp(L and Zud% are defined in Fig. 7.70.
7.4.4.5 Prediction of the HAZ hardness distribution
The predictions are based on computer programmes which utilise the heat and material flow
models described in Ref.113 in combination with the constitutive equations given above to
calculate the HAZ hardness distribution for specific welding conditions.
Accuracy of predictions
Examples of measured and predicted hardness profiles are given in Figs. 7.72 and 7.73.
A closer inspection of the graphs reveals a good agreement between theory and experi-
ments in all three cases. It is interesting to note that there is no clear distinction in the shape of
the HAZ hardness profiles between friction welded Al-Mg-Si alloys and Al-SiC metal matrix
composites when comparison is made on the basis of a similar temperature-time pattern (see
Fig. 7.73). However, large diameter weld components will normally reveal a different hard-
ness distribution, as shown in Fig. 7.72, because longer welding times will increase the total
heat input. Under such conditions the contribution from the plastic deformation becomes
negligible, which means that the resulting HAZ hardness profile will closely resemble that
observed during conventional gas metal arc (GMA) and plasma arc welding of Al-Mg-Si
alloys (see Figs. 7.64 and 7.65).
Process diagrams
Based on the above process model, it is possible to construct a series of diagrams which sum-
marise information about the effect of important welding variables in a systematic and illustra-
tive manner. Examples of such diagrams for 6082-T6 aluminium alloys and T6 heat treated
Al-SiC metal matrix composites are given in Fig. 7.74(a) and (b), respectively.
Hardness, VPN
Predicted
Measured
Unaffected base
material
Axial distance, mm
Fig. 7.72.Comparison between measured and predicted HAZ hardness profiles in a <£26mm Al-Mg-Si
weld component. (Assumed input data: <E> = *F = 0.56, HVmax= 110, HVmin = 42). Operational condi-
tions: qJA = 17W m m 2 and ts = 6s. Data from Midling and Grong.113
(a)
Hardness, VPN
Predicted
Measured
Unaffected base material
Axial distance, mm
(b)
Hardness, VPN
Predicted
Measured
Axial distance, mm
Fig. 7.73.Comparison between measured and predicted HAZ hardness profiles; (a) 4> 16mm Al-Mg-Si
weld component. (Operational conditions: qJA - 25W mrrr2 and ts = 0.9s), (b) <l>16mm friction welded
Al-SiC metal matrix composite. (Operational conditions: qJA = 25W mirr 2 and ts= 3.8s). Data from
Midling and Grong.113
(a) Hardness, VPN
Axial distance, mm •
(b)
Hardness, VPN
Axial distance, mm
Fig. 7.74. Process diagrams for friction welding; (a) 6082-T6 aluminium alloys. (Operational condi-
tions: umax_ = 2.5m s"1 and [i = 0.5), (b) Al-SiC-T6 metal matrix composites. (Operational conditions:
Umax. = 2.5m s"1 and \x = 0.5). Data from Midling and Grong.113
It is evident from these diagrams that the HAZ hardness distribution depends on the total
heat input applied during friction welding. Although the controlling parameters qo IA and ts
(welding time), in practice, are kept within relatively narrow limits, it is obvious that a small
width of the HAZ requires the use of a high specific power (qol A) in combination with a short
duration heating cycle (ts < 2 s). This is also in agreement with general experience.109"112
Example (7.10)
Consider continuous drive friction welding of a T6 heat treated Al-SiC metal matrix compos-
ite under the following conditions:
Use the process diagram in Fig. 7.74(b) to estimate the minimum HAZ hardness level as
well as the total width of the strength reduced zone after welding. In this example we shall
assume that the friction coefficient Ji is equal to 0.5.
Solution
First we need to calculate the frictional heat per unit area of the weld. From equation (7-38),
we have:
A comparison with Fig. 7.74(b) shows that a specific power of 25 W mm"2 corresponds to
a minimum HAZ hardness of about 90 VPN, i.e. a reduction of 45 VPN compared with the
base material. At the same time the total width of the reduced strength zone is seen to be
12 mm.
It should be emphasised that the observed strength loss is not permanent, since the resulting
HAZ strength level is mainly controlled by dissolution reactions taking place within the alu-
minium matrix during the weld thermal cycle. Consequently, a full HAZ strength recovery
can be achieved by the use of an appropriate post weld heat treatment, as shown by the tensile
test data in Table 7.8.
Table 7.8 Mechanical properties of friction welded Al-SiC metal matrix composites. Data from
Midling and Grong109'113
HV R R e
p0 2 m B
Material [VPN] [MPa] [MPa] [%]
Al-SiC-T6 135 315 352 3.6

(base material)
Al-SiC-T6 90 207 268 3.0
(as-welded condition)
Al-SiC-T6 135 313 348 3.1
(PWHT condition)!
Solution heat treated at 535°C for 3 h followed by water-quenching and artifical ageing at 1600C for 10 h.
References
1. R. W.K. Honeycombe: Steels — Micro structure and Properties, 1980, London, Edward Arnold
(Publishers) Ltd.
2. H. Suzuki: Weld. World, 1982, 20, 121-148.
3. 0 . Grong and D.K. Matlock: Int. Met. Rev., 1986, 31, 27-48.
4. DJ. Abson and RJ. Pargeter: Int. Met. Rev., 1986, 31, 141-194.
5. RL. Harrison and R.A. Ferrar: Int. Met. Rev., 1989, 34, 35-51.
6. GJ. Davis and J.G. Garland: Int. Met. Rev., 1975, 20, 83-106.
7. S.A. David and J.M. Vitek: Int. Mater. Rev., 1989, 34, 213-245.
8. T. Gladman and RB. Pickering: In Yield, Flow and Fracture of Poly crystals (Ed. T.N. Baker),
1983, London, Applied Science Publishers, 141-198.
9. A.O. Kluken, M.I. Ons0ien, O.M. Akselsen and G. R0rvik: Joining ScL, 1991,1, 14-22.
10. R Deb, K.D. Challenger and A.E. Therrien: Metall. Trans., 1987,18A, 987-999.
11. RT. Odland, CW. Ramsay, D.K. Matlock and D.L. Olson: WeIdJ., 1989, 68, 158s-168s.
12. DJ. Widgery: Weld J, 1975, 54, 57s-68s.
13. L. Devillers, D. Kaplan, B. Marandet, A. Ribes and RV. Riboud: Proc. Int. Conf. on Effects of
Residual, Impurity and Microalloying Elements on Weldability and Weld Properties, London,
Nov. 1983, Paper 1, Publ. The Welding Institute.
14. R.A. Farrar: Welding and Metal Fabr., 1976, 44, 578-581.
15. 0. Grong and A.O. Kluken: In Ferrous Alloy Weldments, (Eds D.L. Olson and T.H. North),
1992, Zurich (Switzerland), Trans. Tech. Publications, 21-46.
16. R.H. Van Stone, T.B. Cox, J.R. Low and J.A. Psioda: Int. Met. Rev., 1985, 30, 157-179.
17. A.W Thompson: Acta Metall, 1983, 31, 1517-1523.
18. I. Andersen: MSc Thesis, 1989, Division of Metallurgy, The Norwegian Institute of Technol-
ogy, Trondheim, Norway.
19. G.E. Dieter: Mechanical Metallurgy, 3rd edn, 1986, New York, McGraw-Hill Book Company.
20. O.M. Akselsen and 0. Grong: Mater. ScL Eng., 1992, A159, 187-192.
21. D.E. McRobie and J.F. Knott: Mater. ScL TechnoL, 1985,1, 357-365.
22. 0. Grong, A.O. Kluken and B. Bj0rnbakk: Joining and Materials, 1988, l, 164-169.
23. A.O. Kluken and 0 . Grong: Proc. 3rd Int. Conf. on Trends in Welding Research, Gatlinburg,
TN, June, 1992, pp. 569-574. Publ. ASM International, Materials Park, Ohio (1993).
24. J.I. Ustinovshchikov: Acta Metall., 1983, 31, 355-364.
25. AJ. Ardell: Acta Metall., 1972, 20, 601-609.
26. O.M. Akselsen, G. R0rvik, M.I. Ons0ien and 0 . Grong: Weld J, 1989, 68, 356s-362s.
27. H. Suzuki: Trans. Jap. Weld. Soc, 1981,16, 25-32.
28. N. Yurioka, S. Ohshita and H. Tamehiro: Proc. Int. Symposium on Pipeline Welding in the
'80s, March, 1981, pp. 1-15, Publ. Australian Welding Research Association.
29. C.L.M. Cottrell; Metal Constr., 1984,16, 740-744.
30. HJ.U. Cotton: Metal Constr., 1987,19, 217R-223R.
31. T. Kasuya and N. Yurioka: WeIdJ, 1993, 72, 263s-268s.
32. A.O. Kluken, S. Ibarra, S. Liu and D.L. Olson: Proc. 11th Int. Conf. on Offshore Mechanics
and Arctic Engineering, 1992, Publ. ASME, Book No. H0744A-92.
33. O.R. Myhr and 0 . Grong: Acta Metall. Mater., 1990, 38, 449-460.
34. O.M. Akselsen and G. R0rvik: Mater. ScL TechnoL, 1990, 6, 383-389.
35. J.H. Hollomon and L.D. Jaffe: Trans. AIME, 1945,162, 223-249.
36. K. Olsen, D.L. Olson and N. Christensen: Scand. J. Metall, 1982,11, 163-168.
37. PJ. Alberry: Weld J, 1989, 68, 410s-417s.
38. R Ravi Vishnu and K.E. Easterling: In Mathematical Modelling of Weld Phenomena (Eds H.
Cerjak and K.E. Easterling), 1993, London, The Institute of Materials, pp. 241-299.
39. O.D. Sherby and J.E. Dorn: Trans. AIME, 1953,197, 324-330.
40. T. Reti, M. Gergely and R Tardy: Mater. ScL Technol., 1987, 3, 365-371.
41. 0 . Grong and O.M. Akselsen: Metal Constr., 1986,18, 557-562.
42. C. Thaulow, AJ. Paauw, A. Gunleiksrud and OJ. Naess: Metal Constr, 1985,17, 94R-99R.
43. O.M. Akselsen, 0. Grong and G. R0rvik: Scand. J. MetalL, 1990,19, 258-264.
44. C. Thaulow, AJ. Paauw and K. Guttormsen: Weld J, 1987, 66, 266s-279s.
45. O.M. Akselsen, J.K. Solberg and 0. Grong: Scand. J. MetalL, 1988,17, 194-200.
46. O.M. Akselsen, 0 . Grong and P.E. Kvaale: MetalL Trans., 1986,17A, 1529-1536.
47. H. Homma, S. Ohkita, S. Matsuda and K. Yamamoto: Weld J, 1987, 66, 301s-309s.
48. Y-T. Pan and J-L. Lee: Proc. 3rd. Int. Conf. on Trends in Welding Research, Gatlinburg (Ten-
nessee), June, 1992, pp. 539-543. Publ. ASM International, Materials Park, Ohio (1993).
49. J.H.Chen,Y.Kikuta,T.Araki,M.YonedaandY.Matsuda: ActaMetalL, 1984,32,1779-1788.
50. C.A.N. Lanzillotto and RB. Pickering: Met. ScL, 1982,16, 371-382.
51. O.M. Akselsen, 0 . Grong and J.K. Solberg: Mater. ScL Technol., 1987, 3, 649-655.
52. H.P. Shen, T.C. Lei and J.Z. Liu: Mater. ScL Technol., 1986, 2, 28-33.
53. NJ. Kim and G. Thomas: MetalL Trans., 1981,12A, 483-489.
54. A.F. Szewezyk and J. Garland: MetalL Trans., 1982,13A, 1821-1826.
55. M. Ramberg, O.M. Akselsen and 0. Grong: Proc. 1st Int. Conf. on Trends in Welding Re-
search, Gatlinburg, TN, May, 1986, pp. 679-685. Publ. ASM International, Metals Park, Ohio
(1987).
56. N.E. Hannertz: Schweissen & Schneiden, 1976, 28, 379-382.
57. RR. Coe: Welding Steels Without Hydrogen Cracking, 1973, Abington (Cambridge), The
Welding Institute.
58. N. Yurioka and H. Suzuki: Int. Mater. Rev., 1990, 35, 217-249.
59. LT. Brown and W.M. Baldwin: Trans. AIME, 1954, 200, 298-303.
60. E.A. Steigerwald, RW. Shaller and A.R. Troiano: Trans. AIME, 1959, 215, 1048-1052.
61. CA. Zapffe and CE. Sims: Trans. AIME, 1941,145, 225-237.
62. NJ. Petch: Phil. Mag., 1956,1, 331-337.
63. CD. Beachem: MetalL Trans., 1972, 3, 437-451.
64. A.W. Thompson and LM. Bernstein: In Effects of Hydrogen on the Behaviour of Metals (Eds
LM. Bernstein and A.W. Thompson), 1980, Warrendale, PA, Metallurgical Society of AIME,
pp. 291-308.
65. J.D. Fast: Gases in Metals, 1976, London, Macmillan Press Ltd.
66. R.A. Oriani: Ada MetalL, 1970,18, 147-157.
67. N. Christensen, K. Gjermundsen and R. Rose: Brit. WeIdJ., 1958, 5, 272-281.
68. N. Christensen: Svetsen, 1975, 34, 22-32.
69. N. Yurioka and S. Ohshita: HW Doc. IX-1161-80 (1980).
70. C Zhang and J.A. Goldak: HW Doc. IX-1662-92 (1992).
71. N. Christensen and T. Simonsen: Scand. J. MetalL, 1981,10, 120-126.
72. NACE Standard MR-01-75 (1975).
73. M.I. Ons0ien, O.M. Akselsen, 0. Grong and P.E. Kvaale: WeIdJ., 1990, 69 (No. 1), 45-51.
74. H. Suzuki: Trans. Jap. Weld. Soc, 1981,16, 25-32.
75. D.L. Olson: Weld J, 1985, 64, 281s-295s.
76. S. Kou: Welding Metallurgy, 1987, New York, John Wiley & Sons, Inc.
77. R.D. Campbell: In Ferrous Alloy Weldments (Eds D.L. Olson and T.H. North), 1992, Zurich,
Trans. Tech. Publications Ltd., pp. 167-216.
78. I. Varol, J.C Lippold and W.A. Baeslack: ibid., pp. 217-252.
79. R. Mundt and H. Hoffmeister: Arch. Eisenhuttenwes., 1983, 54, 253-256; ibid. 333-336.
80. S. Atamert and J.E. King: Mater. ScL Technol., 1992, 8, 896-911.
81. M. Videm: MSc. Thesis, 1985, Department of Metallurgy, The Norwegian Institute of Tech-
nology, Trondheim, Norway.
82. J.A. Brooks, A.W. Thompson and J.C. Williams: WeldJ, 1984, 63, 71s-83s.
83. A.L. Schaeffler: Metal Prog., 1949, 56, 680-680B.
84. W.T. Delong, G. Ostrom and E. Szumachowski: WeIdJ., 1956, 35, 526s-533s.
85. CJ. Long and W.T. Delong: WeIdJ., 1973, 52, 281s-297s.
86. J.H. Dudas and RR. Collins: Weld J, 1966, 45, 241s-249s.
87. CE. Cross: Ph.D Thesis, 1986, Colorado School of Mines, Golden, Colorado, USA.
88. JJ. Pepe and W.F. Savage: Weld J, 1967, 46, 41 ls-422s.
89. JJ. Pepe and W.F. Savage: Weld J, 1970, 49, 545s-553s.
90. O. Reiso: Proc. 3rd. Int. Conf. on Aluminium Extrusion Technology, Atlanta, GA, 1984, vol. 1,
pp. 31-40. Publ. Aluminium Association (1984).
91. O. Reiso: Proc. 4th Int. Aluminium Extrusion Technology Seminar, Chicago, IL, 1988, vol. 2,
pp. 287-295. Publ. Aluminium Association (1988).
92. O. Reiso, H.G. 0verlie and N. Ryum: Metall. Trans., 1990, 21A, 1689-1695.
93. H. Gjestland, A.L. Dons, O. Lohne and O. Reiso: In Aluminium Alloys — Their Physical and
Mechanical Properties, 1986, Warley (UK), Engineering Materials Advisory Service Ltd., pp.
359-370.
94. O. Lohne and N. Ryum: Proc. 4th Int. Aluminium Extrusion Technology Seminar, Chicago, IL,
1988, vol. 2, pp. 303-308. Publ. Aluminium Association (1988).
95. Metals Handbook, 8th Edition (vol. 8).
96. O. Reiso, N. Ryum and J. Strid: Metall. Trans. A, 1993, 24A, 2629-2641.
97. H.W.L. Phillips: Annotated Equilibrium Diagrams of Some Aluminium Alloy Systems, 1959,
London, The Institute of Metals, pp. 65-71.
98. N.F. Gittos and M.H. Scott: WeIdJ, 1981, 60, 95s-103s.
99. J.E. Hatch (Ed.): Aluminium — Properties and Physical Metallurgy, 1984, Ohio (USA), Ameri-
can Society for Metals.
100. T. Enjo and T. Kuroda: Trans. JWRI, 1982,11, 61-66.
101. S.D. Dumolt: Ph.D Thesis, 1983, Carnegie-Mellon University, USA.
102. O.R. Myhr and 0. Grong: Ada Metall. Mater., 1991, 39, 2693-2702; ibid, 2703-2708.
103. 0 . Grong and O.R. Myhr: In Mathematical Modelling of Weld Phenomena, (Eds H. Cerjak
and K.E. Easterling), 1993, London, The Institute of Materials, pp. 300-311.
104. FM. Muzzolani: Aluminium Alloy Structures, 1985, Boston (USA), Pitman Publishing Inc.
105. A. Kelly and R.B. Nicholson: Progr. Mat. ScL, 1963,10, 151-156.
106. O.R. Myhr, Ph.D Thesis, 1990, Division of Metallurgy, The Norwegian Institute of Technol-
ogy, Trondheim, Norway.
107. European Recommendations for Aluminium Alloy Structures, 1978.
108. G. Steidl and R. Mossinger: Aluminium, 1977, 53, 199-203.
109. O.T. Midling, 0. Grong and M. Camping: Proc. 12th Riso Int. Symp. on Materials Science:
Metal Matrix Composites—Processing, Micro structure and Properties, Roskilde, Denmark,
1991, pp. 529-534. Publ. Riso National Laboratory (1991).
110. O.T. Midling, 0. Grong and D.H. Bratland: Proc. 3rd Int. Conf. on Aluminium Alloys — Their
Physical and Mechanical Properties, Trondheim, Norway, 1992, pp. 99-105. Publ. The Nor-
wegian Institute of Technology, Department of Metallurgy (1992).
111. O.T. Midling and 0. Grong: Proc. 3rd Int. Conf. on Trends in Welding Research, June, 1992,
Gatlinburg, TN, pp. 1147-1151. Publ. ASM International (1993).
112. O.T. Midling and 0. Grong: Proc. Int. Conf Advanced Composites '93, Wollongon, Australia,
February 1993, pp. 1221-1226. Publ. The Minerals, Metals & Materials Society (1993).
113. O.T. Midling and 0. Grong: Acta Metall. Mater., 1994, 42, 1595-1609; ibid, 1611-1622.
114. N.N. Rykalin, A.I. Pugin, V.A. Vasil'eva: Weld. Prod, 1959, 6, 42-52.
115. B. Crossland: Cont. Phys., 1971,12, 559-574.
Appendix 7.1
Nomenclature
lattice parameter (nm) particle diameter (Jim, m)
cross section (mm2, m 2 ) inclusion diameter (jJm, m)
start temperature of ferrite to austenite amount of deposited metal (mm2)

transformation (0C, K)
diffusivity (mm2 s"1, m2 s"1)
end temperature of ferrite to austenite
transformation (0C, K) average grain size ((Xm, m)
Burgers vector (nm, m) lattice diffusion coefficient (mm2 s"1,

mV1)
body-centred cubic
elongation (%)
body-centred tetragonal
gross heat input (kJ mm"1)
amount of fused parent metal (mm2)
binding energy between hydrogen and
kinetic constant in Hollomon-Jaffe pa- trap site (J mol"1, kJ mol"1)
rameter
Young's modulus of matrix (MPa, GN
base metal m-2)
half crack length (jum, m) Young's modulus of particle (MPa, GN

m-2)
various kinetic and empirical constants
Gaussian error function
element concentration (ppm, ml per
100 g, |il cm-3) particle volume fraction
eutectic concentration (wt%) volume fraction of hardening precipi-

tates in natural ageing model
initial element concentration (ppm,
ml per 100 g, |il cnr 3 ) initial particle volume fraction
maximum solid solubility (wt%) eutectic fraction
base metal or weld metal element con- flux-cored arc welding

centration (wt%, ppm, ml per 100 g,
face-centred cubic
jil cm"3)
filler wire
Yurioka carbon equivalent
grain coarsened region
HW carbon equivalent
1
gas metal arc welding
crack tip opening displacement
grain refined region
Charpy V-notch
gas tungsten arc welding
plate thickness (mm)
hydrogen content related to fused metal post weld heat treatment
(ml per 100 g or g per ton)
net power (W)
Rockwell C hardness
apparent activation energy for soften-
Vickers hardness (kg mm"2 or VPN) ing reaction (J mol"1, kJ mol"1)
Vickers hardness in the artificially aged radius vector (mm, m)

condition (kg mm"2 or VPN)
isothermal zone width referred to fu-
matrix Vickers hardness in the absence sion boundary (mm)
of hardening precipitates (kg mm"2 or
VPN) universal gas constant (8.314 J mol"1
K-i)
kinetic strength of thermal cycle
with respect to P' (Mg2Si) formation friction surface radius (mm, m)
intercritical region implant rupture strength (MPa)
locking parameter in Hall-Petch equa- ultimate tensile strength (MPa)

tion (MPa |im 1/2 )
0.2% proof stress (MPa)
equilibrium partitioning coefficient
scanning electron microscope
density of trap site
shielded metal arc welding
half length of fusion zone in hydrogen
diffusion model (mm, m) subcritical region
longitudinal direction (notch perpen- time (s)

dicular to plate rolling direction)
maximum hold time required for com-
strain hardening exponent in Ludwik plete particle dissolution at T (s)
equation
critical hold time required to precipi-
interfacial torque (Nm) tate a certain amount of P'(Mg2Si) at T
(S)
martensite-austenite constituent
welding time (s)
metal matrix composite
cooling time from 600 to 4000C (s)
time exponent
dimensionless operating parameter in
heat flow model
equivalent isothermal hold time at
friction pressure (N mm"2, MPa) 333K (s)
Hollomon-Jaffe parameter temperature (0C, K)
Dorn parameter (s) chosen reference temperature

(0C, K)
pressure distribution across weld inter-
face (MPa) eutectic temperature (0C, K)
melting point (0C, K) strength (MPa)
ambient temperature (0C, K) yield stress (MPa)

0
peak temperature ( C, K) intergranular fracture strength (MPa)
artificially aged condition matrix fracture strength (MPa)
transmission electron microscope
friction stress (MPa)
transverse direction (notch parallel with
plate rolling direction) yield strength in artificially aged con-
dition (MPa)
velocity (m s"1)
matrix yield strength in the absence of
surface velocity (ms"1) hardening precipitates (MPa)
area under stress-strain curve (J m~3) critical stress for particle cracking
(MPa)
ultimate tensile strength
radial stress (MPa)
welding speed (mm s"1)
threshold stress for H2S stress corrosion
inclusion volume fraction cracking (MPa)
Vickers pyramid number tangential stress (MPa)
welding direction (mm, m) effective surface energy (J irr 2 )
transverse direction (mm, m) internal interface (or antiphase) bound-

ary energy (J m~2)
y-coordinate at maximum width of iso-
therm (mm) surface energy of particle (J m~2)
isothermal zone width (mm, m) friction coefficient
through-thickness direction (mm, m) maximum shear stress (MPa)
partly deformed region in friction weld- dimensionless cooling time from 800
ing model (mm) to 5000C
fully plasticised region in friction weld- dimensionless temperature conforming

ing model (mm) to 8000C
undeformed region in friction welding dimensionless temperature conforming

model (mm) to 5000C
stress concentration factor arc efficiency factor
true fracture strain dimensionless supersaturation
dimensionless plate thickness ) non-hardening precipitates in Al-Mg-

Si alloys and Al-SiC MMCs
material constants
) hardening precipitates in Al-Mg-Si
volume heat capacity (J mm"3 0C"1) alloys and Al-SiC MMCs
dimensionless strength parameter net precipitation strength increment
(MPa)
dimensionless strength parameter in
reversion model maximum precipitation strength incre-
ment (MPa)
dimensionless strength parameter in
natural ageing model HAZ strength reduction factor
dimensionless strength parameter in angular velocity (rad/s)

work hardening model
8
Exercise Problems with Solutions
8.1 Introduction
This chapter contains a collection of different exercise problems which the author has
adopted in his welding metallurgy course for graduate (mature) students. They illustrate how
the models described in the previous chapters can be used to solve practical problems of more
interdisciplinary nature. Each of them contains a 'problem description' and some background
information on materials and welding conditions. The exercises are designed to illuminate the
microstructural connections throughout the weld thermal cycle and show how the properties
achieved depend on the operating conditions applied. Solutions to the problems are also pre-
sented. These are not complete or exhaustive, but are just meant as an aid to the reader to de-
velop the ideas further.
8.2 Exercise Problem I: Welding of Low Alloy Steels
Problem description
Consider gas metal arc (GMA) welding of low allow steels under the following conditions:
(i) Tack welding of a T-joint (Fig. 8.1)
(ii) Root pass deposition in a single V-groove (Fig. 8.2)
(iii) Root pass deposition in a X-groove (Fig. 8.3)
(iv) Deposition of cap layer during multipass welding (Fig. 8.4)
The materials to be welded are a C-Mn steel and a Nb-microalloyed low carbon steel with
chemical compositions and properties as listed in Tables 8.1 and 8.2. Details of welding par-
ameters and operational conditions are given in Table 8.3 and 8.4, respectively.
Table 8.1 Exercise problem I: Base plate chemical compositions (in wt%).
Steel C Si Mn P S Nb Al
1
C-Mn 0.20 0.35 1.46 0.003 0.002 - 0.037
1
LC-Nb 0.08 0.26 1.44 0.003 0.003 0.020 0.025
1
Ti: -0.008, N: 0.0027, Ca: 0.0040, B: 0.0002.
Table 8.2 Exercise problem I: Mechanical properties of base materials.

Steel 1 Rp02 (MPa) I Rm (MPa) I El. (%) I CVN - 4 0 (J)
C-Mn 328 525 33 150
LC-Nb 430 525 32 225
Table 8.3 Exercise problem I: Welding parameters.
Parameter / (A) U (V) v (mm s"1)
Value 150 21 4
f
The arc efficiency factor may be taken equal to 0.85 (see Table 1.3). No preheating is applied (T0 = 20 0C).
Table 8.4 Exercise problem I: Operational conditions and filler wire characteristics K
Shielding gas: Pure CO2

Gas flow rate: 15 Nl per min
Wire diameter: 1.0 mm
Wire feed rate: 6.0 m per min
Wire composition: C: 0.1 wt%, Si: 1.0 wt%, Mn: 1.7 wt%
Weld metal*
composition: C: 0.09 wt%, Si: 0.7 wt%, Mn: 1.2 wt%
Weld metal*
properties: Rp02: 460 MPa, Rm: 560 MPa, El.: 26%, CVN _40: 50 J
f
Data compiled from dedicatedfillerwire catalogues and welding manuals.
* Values refer to all weld metal deposit.
Fig. 8.1. Tack welding of a T-joint.
Fig. 8.2. Root pass deposition in a single V-groove.

Fig. 8.3. Root pass deposition in a X-groove.
Fig. 8.4. Deposition of cap layer during multipass welding.
Analysis:
The students should work in groups (3 to 4 persons) where each group select a specific com-
bination of base material and welding conditions (e.g. deposition of a cap layer on the top of
a thick multipass C-Mn steel weld). The problem here is to evaluate the response of the base
material to heat released by the welding arc. The analysis should be quantitative in nature and
based on sound physical principles. The following points shall be considered:
(a) Select an appropriate heat flow model for the system under consideration.
(b) Estimate the minimum bead length which is required to achieve pseudo-steady state (i.e.
a temperature field that does not vary with position when observed from a point located
in the heat source).
(c) Estimate the value of the deposition coefficient kx (in gA " 1 S" 1 ), the weld cross section
areas D and B (in mm2), and the mixing ratio DI(B + D) during welding.
(d) Estimate the weld metal chemical composition. Calculate then the following quantities:
- Total loss of Si and Mn in the arc column

- Total oxygen pick-up in the weld pool
- Residual oxygen level and total amount of oxygen rejected from the weld pool
during deoxidation
- Total amount of slag formed during welding (in g per 100 gram weld metal)
(e) Carry out a total oxygen balance for the system, and estimate the resulting CO content
in the welding exhaust gas.
(f) Estimate the chemical composition, volume fraction, and mean size (diameter) of the
oxide inclusions which form in the cold part of the weld pool. Calculate then the follow-
ing inclusion characteristics:
- Number of particles per unit volume

- Number of particles per unit area
- Total surface area of particles per unit volume
- Mean particle centre to centre volume spacing
(g) Estimate the weld metal solidification mode and the resulting columnar grain mor-
phology. Indicate also the type of substructure which form at different positions from the
weld centre line.
(h) Evaluate the thermal stability of the base metal grain boundary pinning precipitates. At
which temperature will these precipitates dissolve?
(i) Calculate the austenite grain size profile across the HAZ. Estimate also the size of the
columnar austenite grains in the weld metal.
(j) Estimate the primary reaction products which form in the weld metal and the HAZ after
the austenite to ferrite transformation.
(k) Estimate the maximum hardness in the HAZ after welding. Use this information to eval-
uate the risk of hydrogen cracking and H2S stress corrosion cracking during service.
(1) Estimate the CVN toughness both in the weld metal and the HAZ after welding.
(m) Based on the results obtained explain why the carbon content of modern structural steels
has been gradually lowered to values below 0.1 wt% in step with the progress in steel
manufacturing technology.
Solution:
In all cases we can use stringer bead deposition on thick plates as a model system. It follows
from the analysis in Section 1.10.7 (Chapter 1) that the pertinent difference in the effective
heat diffusion area between a bead-on-plate weld and a groove weld may conveniently be ac-
counted for by introducing a correction factor/, which depends on the geometry of the groove
(see Fig. 1.68). Thus, in the general case the net (effective) power of the heat source can be
written as:
In the following, we shall only consider deposition of a cap layer on a thick plate where
/ = 1, but the analysis can readily be applied to other combinations of steels and welding con-
ditions as well (e.g./< 1). In the former case, we get:
Table 1.1 (Chapter 1) contains relevant input data for the steel thermal properties.
(a) The problem of interest is whether we must use the general (but complex) Rosenthal
thick plate solution (equation (1-45)) or can adopt the simplified solution for a fast moving
high power source (equation (1-73)). Fig 1.24 provides a basis for such an evaluation. The
most critical position will be the fusion line. If we neglect the latent heat of melting, the QJn3
ratio at the melting point becomes:
Readings from Fig. 1.24 suggest that the error introduced by neglecting the contribution
from heat flow in the welding direction is sufficiently small that it can be disregarded in the
calculations of the HAZ thermal programme. This means that equation (1-73) can be used in
replacement of equation (1-45) if that is desirable.
(b) The duration of the transient heating period depends on the actual point of observation
(i.e. the distance from the heat source). If we, as an illustration of principles, would like to
apply the pseudo-steady state solution down to a peak temperature of, say, 7000C, the corre-
sponding nJQ ratio at that temperature becomes:
From Fig. 1.21 we see that this ratio corresponds to a dimensionless radius vector a3m of
about 5. The duration of the transient heating period may now be read from Fig. 1.18. A crude
extrapolation gives:
from which
The minimum bead length is thus 25 mm, which is surprisingly short,
(c) The value of the deposition coefficient may be estimated from the data in Table 8.4.
This value corresponds to a kVp ratio of about 0.65 mm 3A 1 S \ which is in excellent

agreement with the data quoted in Table 1.7. The area D of deposited metal thus becomes (see
equation (1-120)):
The corresponding area of fused parent metal is most conveniently read from Fig. 1.21.
Taking the n3/Q ratio at the melting point equal to (1/0.22) ~ 4.5, we obtain:
from which
The mixing ratio is thus:
This value is somewhat lower than the expected mixing ratio, which for low heat input
welding is close to 0.67.
(d) The composition data in Table 8.4 refer to all weld metal deposit. Since the dilution with
respect to the base material in this case is small, the weld metal composition would be ex-
pected to be close to that given in Table 8.4.
An estimate of the total burn-off of alloying elements during welding can be obtained by
considering the difference in chemical composition between the filler wire and the weld
metal. In the present case we get:
Loss of silicon
As shown in Section 2.10.1.3 (Chapter 2), the silicon loss can partly be ascribed to SiO(g) for-
mation in the arc column (with consequent fume formation), and partly to reactions with oxy-
gen in the weld pool during the deoxidation stage (with consequent silicate slag formation).
The former loss can be estimated from the fume formation data presented in Table 2.6. Taking
the fume formation rate (FFF) of silicon equal to 63 mg min"1, the total loss of silicon in the
arc column amounts to:
The corresponding oxidation loss of silicon in the weld pool is thus:
Loss of manganese
As shown in Section 2.10.1.4 (Chapter 2), manganese is partly lost in the arc column due evap-
oration and partly in the weld pool due to deoxidation reactions. Taking the fume formation
rate of manganese equal to 14 mg min"1 (from Table 2.6), the total loss of Mn in the arc col-
umn amounts to:
The corresponding oxidation loss of manganese in the weld pool is thus:
Oxygen pick-up in the weld pool

When the oxidation losses of silicon and manganese in the weld pool are known, it is possible
to calculate the total oxygen pick-up in the hot spot of the pool immediately beneath the root
of the arc, according to the procedure outlined in Section 2.10.1.5 (Chapter 2). However, first
we need to estimate the residual weld metal oxygen content on the basis of the thermo-
dynamic model presented in Fig. 2.56. In the present example, the numerical value of the
deoxidation parameter is:
Reading from Fig. 2.56 gives a residual oxygen content of about 0.07 wt%. The total oxy-
gen pick-up in the weld pool is thus:
Rejected oxygen from the weld pool

The amount of rejected oxygen is equal to the difference between the total and the residual
oxygen level:
From this we see that most of the oxygen which is picked up at elevated temperatures is
rejected again during cooling in the weld pool due to deoxidation reactions and subsequent
phase separation.
Manganese silicate slag formation

The weld pool deoxidation reactions give rise to the formation of a top bead slag, as shown
in Section 2.10.1.5 (Chapter 2). In the present example the amount of slag per 10Og weld
metal is equal to:
A comparison with Fig. 2.35 shows that the calculated weight of slag is in reasonable agree-
ment with experimental observations.
(e) The oxygen balance is carried out in accordance with the procedure outlined in Section
2.10.1.7 (Chapter 2). First we need to estimate the total mass of weld metal produced per unit
time:
The total CO2 consumption is thus:
Oxidation of carbon:
Oxidation of silicon:
Oxidation of manganese:
Increase in the weld metal oxygen content:
The total CO evolution is equal to the sum of these four contributions:
The resulting CO content in the welding exhaust gas is thus:
A comparison with the experimental data in Table 2.2 shows that the calculated CO con-
tent is of the expected order of magnitude.
(f) The deoxidation model in Section 2.12.4.1 (Chapter 2) can be used to estimate the inclu-
sion composition. From Fig. 2.68 we see that the inclusions are essentially pure manganese sil-
icates with an overall composition close to MnSiO3.
When the inclusion composition is known, it is possible to convert the residual weld metal
oxygen content into an equivalent inclusion volume fraction according to the procedure out-
lined in Section 2.12.1.Taking the stoichiometric conversion factor equal to 5.0 X 10~2 for
manganese silicate slags, we obtain:
Moreover, we can use equation (2-79) in Section 2.12.2.2 to calculate the mean diameter
of the inclusions:
The different inclusion characteristics may now be estimated from equations (2-80) to (2-83):
Number of particles per mm3:
Number of particles per mm2:
Total surface area of particles per mm3:
Mean particle centre to centre volume spacing:
A comparison with Table 2.11 shows that the calculated inclusion characteristics are in
reasonable agreement with those reported for C-Mn steel weld metals.
(g) The characteristic growth pattern of columnar grains in bead-on-plate welds is shown
schematically in Fig. 3.33. The first phase to form will be delta ferrite which subsequently de-
composes to austenite via a peritectic transformation (see Fig. 3.72). The important question is
whether re-nucleation of the grains will occur during solidification. In practice, this depends on
the interplay between a number of variables which cannot readily be accounted for in a sim-
plified analysis, including the weld pool geometry, the cooling rate and the nucleation potency
of the non-metallic inclusions. Broadly speaking, the energy barrier associated with nucleation
of delta ferrite at manganese silicates is rather high (e.g. see Fig. 3.30), which suggests that for-
mation of new grains ahead of the advancing solid/liquid interface is not very likely under the
prevailing circumstances. Hence, the columnar grain zone would be expected to extend entirely
from the fusion line towards the centre of the weld, as frequently observed in this type of welds.
Moreover, Fig. 3.43 provides a basis for estimating the substructure of the weld metal
columnar grains. Close to weld centre-line the local crystal growth rate will approach the
welding speed (i.e. RL ~ 4 mm s"1). At the same time a simple analytical solution exists for the
thermal gradient in the weld pool (equation (3-28)):
From this we see that a cellular-dendritic type of substructure is likely to form within the
central parts of the fusion zone, in agreement with general experience (see Fig. 3.36).
(h) Fig. 5.25 shows the location of the cap layer. Since the base plate is a Nb-microalloyed
steel, the important grain boundary pinning precipitates within the HAZ are either NbC, NbN
or a mixture of these. In the former case the equilibrium dissolution temperature may be es-
timated from the solubility product of the pure binary compounds. From equation (4-4) and
Table 4.1, we have:
and
This shows that NbC is thermodynamically more stable than NbN. In practice, the real dis-
solution temperature may be significantly higher than that predicted from equation (4-4) be-
cause of the kinetic superheating (see discussion in Section 4.4, Chapter 4). The grain growth
diagram in Fig. 5.21 (a) provides a basis for estimating the effect of heating rate (heat input)
on the dissolution kinetics. Taking the ordinate qo /v equal to 2678/4000 = 0.67 kJ mm"1, we
obtain:
This corresponds to a kinetic superheating of about 2000C in the case of NbC.
In the HAZ on the weld metal side (see Fig. 5.25), oxide inclusions may act as effective
grain boundary pinning precipitates. These will be thermodynamically stable up to the melt-
ing point of the steel.
(i) The austenite grain size profile across the base plate HAZ can be read from Fig. 5.21(a).
Taking the ordinate q/v equal to 0.67 kJ mm"1, we see that the maximum austenite grain size
at the fusion boundary will exceed 100 /mm because of dissolution of the base metal grain
boundary pinning precipitates. In the HAZ on the weld metal side, the situation is different.
Here the stable weld metal oxide inclusions will impede austenite grain growth to a much
larger extent.The limiting austenite grain size may be calculated from equation (5-21).Taking
the Zener coefficient equal to 0.5 for oxide inclusions in steel (Fig. 5.4), we obtain:
Because of the phenomenon of epitaxial grain growth (see Section 3.3, Chapter 3), the in-
itial size of the weld metal delta ferrite/austenite columnar grains would be expected to be
comparable to the size of the HAZ austenite grains adjacent to the fusion boundary. Since the
latter varies along the periphery of the fusion boundary at the same time as competitive grain
growth leads to a general coarsening of the solidification microstructure with increasing dis-
tance from the fusion boundary, an average columnar austenite grain size of about 50 /mm
seems reasonable under the prevailing circumstances.
(j) As an illustration of principles, we shall assume that the CCT diagram in Fig. 6.27(a) pro-
vides an adequate description of the base plate transformation behaviour during welding. The
cooling time from 800 to 500 0C can be calculated from equation (1-67):
from which
Readings from Fig. 6.27(a) give the following microstructures within the grain coarsened
and grain refined region of the HAZ, respectively:
Grain coarsened region (T ~ 13500C):

Microstructure : 100% lath martensite
Transformation start temperature: ~ 470 0C
Grain refined region (Tp «10000C):

Microstructure : ferrite + pearlite
Transformation start temperature: ~ 600 0C
It follows that the observed difference in the HAZ transformation behaviour can mainly
be attributed to a corresponding difference in the prior austenite grain size, which according
to Fig. 5.21(a) is about 50 /im at Tp « 1350 0C and below 10 ^m at Tp « 10000C.
In addition, small islands of plate martensite will form within the intercritical (partly trans-
formed) HAZ, where the peak temperature of the thermal cycle has been between Ac1 and
Ac3 (see discussion in Section 6.3.8.2, Chapter 6). Just above the Ac1 temperature the volume
fraction of the M-A (martensite-austenite) constituent is approximately equal to the base
plate pearlite content (Fig. 6.66), which in the present case is about 8 vol%, as judged from the
steel carbon content.
Considering the weld metal, the situation is different. Here the oxide inclusions will
strongly affect the microstructure evolution by promoting intragranular nucleation of acicu-
lar ferrite (see discussion in Section 6.3.5, Chapter 6). In practice, the role of inclusions in weld
metal transformation kinetics is difficult to assess and hence, we will take a more simplistic
(pragmatic) approach to this problem by just comparing the total surface area available for
nucleation of ferrite at prior austenite grain boundaries and inclusions, respectively (SJGB)
versus SJI)). The following three regions of the weld are considered:
As-deposited weld metal:
Reheated weld metal (close to fusion line):

Reheated weld metal (far from fusion line):
In this case an estimate will be made for dy = 10 /mi.
From the above calculations it is apparent that the conditions for acicular ferrite formation
are particularly favourable within the as-deposited weld metal (Sx(I) > SJGB)), and some-
what less favourable within the high peak temperature region of the weld HAZ (SJGB) >
SJI)). In contrast, acicular ferrite would not be expected to form within the low peak tem-
perature region of the HAZ, since nucleation of ferrite at austenite grain boundaries in this
case will completely override nucleation at inclusions (SJGB) » SJI)).This is also in agree-
ment with general experience (e.g. see photographs of typical microstructures in Fig. 6.19(c)
and (d)).
(k) The maximum hardness/strength level within the grain coarsened region of the HAZ can
be estimated from the diagrams presented in Section 7.2.2 (Chapter 7) if the steel composition
and welding parameters fall within the specified range. Alternatively, we can use Fig. 7.19,
which applies to low carbon microalloyed steels. Taking the cooling time from 800 to 500 0C,
Ar8/5, equal to 3.3 s, we obtain:
HVmax = ~ 380 VPN and Rp02 (max) = ~ 980 MPa
In general, a hardness rather than a strength criterion is used as a basis for evaluation of the
risk of hydrogen cracking and H2S stress corrosion cracking during service. In the former case
an upper limit of about 300 to 325 VPN is incorporated in many welding specifications to avoid
problems with hydrogen cracking, but this restriction can be relaxed if specific precautionary
actions are taken during the welding operation to reduce the supply of hydrogen as shown in
Section 7.2.3 (Chapter 7). Considering the H2S stress corrosion cracking resistance a maximum
hardness level of 248 VPN is strictly enforced in many welding specifications, as discussed pre-
viously in Section 7.2.4 (Chapter 7). Hence, significant tempering of the martensite would be
required if the weldment is going to be used in environments containing sour oil or gas.
(1) In general, the toughness requirements vary with the type of application, but for offshore
structures a minimum CVN toughness of 35J at — 400C is frequently specified. From the CVN
data in Tables 8.2 and 8.4 it apparent that both the base plate and the weld metal meet this re-
quirement. Moreover, auto-tempered low carbon martensite and polygonal ferrite, which
form within the grain coarsened and grain refined region of the HAZ, respectively are known
to have an adequate cleavage resistance.This means that the intercritical HAZ is the most
critical region of the joint when it comes to toughness due to the presence of high carbon plate
martensite within the ferrite matrix (see Figs. 6.61 through 6.65 and discussion in Section
7.2.2.3, Chapter 7). In practice, the problem may be solved by applying an appropriate post
weld heat treatment (PWHT).
(m) Since the properties of martensite depend on the carbon content, C-Mn steel weldments
will generally be more prone to hydrogen cracking, H2S stress corrosion cracking and brittle
fraction initiation in the HAZ than low carbon microalloyed steel weldments. This explains
why the base plate carbon content has been gradually lowered to values well below 0.1 wt%
in step with the progress in steel plate manufacturing technology.
8.3 Exercise Problem II: Welding of Austenitic Stainless Steels
Problem description:
Consider GTA welding of 2 mm thin sheets of AISI 316 austenitic stainless steel with chemi-
cal composition as listed in Table 8.5. The base plate has an average grain size of 18 /xm in the
fully annealed condition, which conforms to a tensile yield strength of about 300 MPa. The
sheets shall be butt welded in one pass, using a simple I-groove with 3 mm root gap. In this
case the addition of filler wire is adjusted so that the area of the weld reinforcement amounts
to 50% of the groove cross section. Details of welding parameters and operational conditions
are given in Table 8.6 and 8.7, respectively.
Table 8.5 Exercise problem II: Base plate chemical composition (in wt%).
Steel C Mn Cr Ni
AISI316 0.03 2.0 16 12
Table 8.6 Exercise problem II: Welding parameters*.

Parameter / (A) U (V) v (mm s"1)
Value 200 15 5
+ 0
The arc efficiency factor may be taken equal to 0.4. No preheating is applied (T0 = 20 C).
Table 8.7 Exercise problem II: Operational conditions and filler wire characteristics1.
Shielding gas: Argon
Wire composition:
Weld metal*
properties:
Data compiled from dedicatedfillerwire catalogues and welding manuals.

Values refer to all weld metal deposit.
Analysis:
The problem here is to evaluate the response of the base material to welding under the con-
ditions described above. The analysis should be quantitative in nature and based on sound
physical principles. The following input data are recommended:
Specific questions:
(b) Estimate the minimum bead length which is required to achieve pseudo-steady state
down to a peak temperature of 1000 0C.
(c) Estimate the deposition rate (in gA^s" 1 ), the weld cross section areas D and B (in
mm2), and the dilution ratio B/(B + D) during welding.
(d) Estimate the weld metal chemical composition for the given combination of base
plate, filler wire and dilution ratio.
(e) Sketch the contour of the weld pool and the resulting columnar grain morphology in
the x-y plane after solidification. Estimate also the weld metal delta ferrite content.
(f) Evaluate the risk of solidification cracking during welding.
(g) Calculate the austenite grain size profile across the HAZ. Estimate also the size of the
columnar grains in the weld metal.
(h) Evaluate the risk of chromium carbide formation in the HAZ during welding.
(i) Estimate on the basis of the Hall-Petch relation the maximum load bearing capacity of
the joint during service.
Solution:
(a) The problem of interest is whether we must use the general (but complex) Rosenthal thin
plate solution (equation (1-81)) or can adopt the simplified solution for a fast moving high
power source (equation (1-100)). Fig 1.43 provides a basis for such an evaluation. The most
critical position will be the fusion line. If we neglect the latent heat of melting, the BJn^ ratio
at the melting point becomes:
Similarly, the dimensionless plate thickness is equal to:
Readings from Fig. 1.43 show that we are outside the validity range of the simplified 1-D
model close to the fusion line, but that this solution is a good approximation within the low
peak temperature region of the HAZ. Here equation (1-100) may be used in replacement of
equation (1-81).
(i.e. the distance from the heat source). If we would like to apply the pseudo-steady state sol-
ution down to a peak temperature of 1000 0C, the corresponding nJ8B ratio becomes:
From Fig. 1.31 we see that this ratio corresponds to a dimensionless radius vector a5m of
from which
(c) First we need to calculate D:
This gives the following deposition rate:
The total area of fused metal can be read from Fig. 1.31. At the melting point the n3/0p8
ratio is close to 2, which gives:
and
This gives:
Note that in these calculations we have assumed that A2 is equal to the sum of (B+D) in
order to achieve realistic numbers.
(d) The weld metal composition can be calulated from a simple 'rule of mixtures':
By using input data from Tables 8.5 and 8.7, we get:

(e) The bead morphology can be read from Fig. 1.29. Taking the 68In3 ratio at the melting
point equal to 0.5, it is easy to verify that the geometry of the weld pool in this case is tear-
shaped. The columnar grain structure is therefore similar to that shown in Fig. 3.11(b).
When the composition is known the weld metal microstructure can be read from Fig. 7.53
by considering the resulting chromium and nickel equivalents:
This gives a delta ferrite content of about 7 vol%.
(f) Normally, a minimum delta ferrite content of about 5 to 10 vol% is specified to avoid
problems with solidification cracking in the weld metal (see discussion in Section 7.3.4,
Chapter 7). This requirement is clearly met under the prevailing circumstances.
(g) The HAZ austenite grain size in different positions from the fusion boundary can be read
from Fig. 5.30(b). In the present example the net heat input per mm2 of the weld is equal to:
This corresponds to a maximum austenite grain size of about 60/mi close to the fusion
boundary, which also is a reasonable estimate of the weld metal columnar grain size.
(h) The most critical position is the low peak temperature region of the weld HAZ where Tp
is between 800 and 1000 0C, as shown in Section 6.4.2 (Chapter 6). However, it is evident from
Fig. 6.69 that the risk of chromium carbide formation in this case is negligible because of the
low base plate carbon content. Hence, the corrosion resistance will not be significantly affec-
ted by the welding operation.
(i) The minimum HAZ strength level may conveniently be calculated from equation (7-21),
using input data from Example 7.5 (page 530):
This gives the following strength reduction factor for the joint:
8.4 Exercise Problem III: Welding of Al-Mg-Si Alloys
Problem description:
Consider G M A welding of 5 mm A A 6082 extrusions with chemical composition as listed in

Table 8.8. The base material has a Vickers hardness and tensile yield strength of 110 VPN and
280 MPa, respectively in the T6 temper condition. The extrusions shall be butt welded in one
pass, using a simple I-groove with no root gap. Two different filler wires are available, one Al-Si
wire and one Al-Mg wire (in the following designated wire I and II, respectively). Details of
welding parameters and operational conditions are given in Table 8.9 and 8.10, respectively.
Table 8.8 Exercise problem III: Base plate chemical composition (in wt%).
Alloy Si Mg Mn Fe
AA 6082 0.98 0.64 0.52 0.19
Table 8.9 Exercise problem III: Welding parameters1.

Parameter /(A) (/(V) v (mm s"1)
Value 200 28 10
0
|The arc efficiency factor may be taken equal to 0.8. No preheating is applied (T = 20 C).
Table 8.10 Exercise problem III: Operational conditions and filler wire characteristics1.
Shielding gas: Argon
Gasflowrate: 20 Nl per min
Wire diameter: 1.6 mm
Wire feed rate: 5.5 m per min
Wire composition: Wire I : Al + 5 wt% Si
Wire II: Al +5 wt% Mg
Weld metal* Wire I:
properties: Rp02 : 55 MPa, Rn; 165 MPa, El.: 18%
Wire II:
Rp02 : >130 MPa, Rn;. >280 MPa, El.: >17%, CVN+20: >30 J
Data compiled from dedicated filler wire catalogues and welding manuals.
Values refer to all weld metal deposit.
Analysis:
The problem here is to evaluate the response of the base material to welding under the con-
ditions described above. The analysis should be quantitative in nature and based on sound
physical principles. The following input data are recommended:
Specific questions:
Atomic percent silicon
Temperature, 0C
Weight percent silicon
Fig. 8.5. The binary Al-Si phase diagram.
Atomic percent magnesium

Temperature, 0C
Weight percent magnesium
Fig. 8.6. The binary Al-Mg phase diagram.

(b) Estimate the minimum bead length which is required to achieve pseudo-steady state down
to a peak temperature of 200 0C.
(c) Estimate the value of the deposition coefficient k' (in gA^s" 1 ), the weld cross section
areas B and Z) (in mm2), and the dilution ratio BI(B + D) during welding.
(d) Estimate the content of Mg and Si in the weld metal.
(e) Sketch the weld metal columnar grain structure and the segregation pattern during sol-
idification. Indicate also the type of substructure which forms at different positions along
the periphery of the fusion boundary. Relevant binary phase diagrams are given in Figs.
8.5 and 8.6.
(f) Evaluate the risk of solidification cracking during welding.
(g) Evaluate the risk of liquation cracking in the HAZ during welding.
(h) Sketch the sequence of reactions occurring within the HAZ during welding. Then estimate
the following quantities:
- The temperature for incipient dissolution of /3".
- The total width of the HAZ (referred to the fusion boundary).
- The temperature for full dissolution of /3".
- The total width of the fully reverted HAZ (referred to the fusion boundary).
(i) Estimate for each combination of filler wire and parent material an overall strength re-
duction factor which determines the load bearing capacity of the joint.
(j) Imagine now that the same extrusion instead is used in the fully annealed (O- temper)
condition with a Vickers hardness and tensile yield strength of 50 VPN and 100 MPa, re-
spectively. To what extent will the temper condition affect the microstructure and strength
evolution during welding?
Solution:
(a) The problem of interest is whether we must use the general (but complex) Rosenthal thin
plate solution (equation (1-81)) or can adopt the simplified solution for a fast moving high
power source equation (1-100)). Fig 1.43 provides a basis for such an evaluation. The most
critical position will be the fusion line. If we neglect the latent heat of melting, the 6 In3 ratio
at the melting point becomes:
Similarly, the dimensionless plate thickness is equal to:

Readings from Fig. 1.43 show that we are outside the validity range of the simplified 1-D
solution close to the fusion line, but that equation (1-100) may be used (with some reserva-
tions) within the low peak temperature region of the HAZ.
(i.e. the distance from the heat source). If we would like to apply the pseudo-steady state sol-
ution down to a peak temperature of 200 0C, the corresponding n/86p ratio becomes:
From Fig. 1.31 we see that this ratio corresponds to a dimensionless radius vector <r5m of
from which
It follows that the minimum bead length required to achieve pseudo-steady state during
aluminium welding is much longer than in steel welding due to the pertinent differences in the
heat flow conditions (e.g. see Example 1.5, Chapter 1).
(c) The value of the deposition coefficient may be estimated from the data in Table 8.10:
This value corresponds to a A: Vp ratio of about 0.92 mm3A 1S \ which is in excellent agree-
ment with the data quoted in Table 1.7. The area D of deposited metal thus becomes (see
equation (1-120)):
The total area of fused metal can be read from Fig. 1.31. At the melting point the nJ0p8
ratio is close to 0.93, which gives:
and
This gives:
Note that in these calculations we have assumed that A2 is equal to the sum of (B + D) in
order to achieve realistic numbers.
(d) The weld metal composition can be calulated from a simple 'rule of mixtures':
By using input data from Tables 8.8 and 8.10, we get:
Wire I:
Wire II:
(e) The bead morphology can be read from Fig. 1.29. Taking the 68In3 ratio at the melting
point equal to 1, it is easy to verify that the shape of weld pool in this case is elliptical. The
columnar grain structure is therefore similar to that shown in Fig. 3.11 (a).
Moreover, Fig. 3.43 provides a basis for estimating the substructure of the weld metal
columnar grains. Close to weld centre-line the local crystal growth rate will approach the
welding speed (i.e. RL ~ 10 mm s"1). At the same time a simple analytical solution exists for
the thermal gradient in the weld pool (equation (3-29)):
From this we see that a cellular-dendritic type of substructure is likely to form within the
central parts of the fusion zone, in agreement with general experience.
If we only consider the contribution from the major alloying element in each case, the Scheil
equation (equation (3-46)) may be used for an analysis of the segregation pattern during sol-
idification. By using input data from the binary phase diagrams in Figs. 8.5 and 8.6, we get:
Wire I:
Wire II:
From this we see that the amount of eutectic liquid which forms during solidification is sen-
sitive to variations in the filler wire chemical composition (i.e. the Si or Mg content).
(f) Fig. 7-54 provides a basis for evaluation of the hot cracking susceptibility.
Wire I
In this case the fraction of eutectic liquid is so abundant that it backfills and 'heals' all incipi-
ent cracks. Hence, the hot cracking susceptibility is low.
Wire II
When the Al-Mg filler wire is used the fraction of eutectic liquid is just large enough to form
continuous films at the columnar grain boundaries. Hence, the hot cracking susceptibility is
high.
(g) Liquation cracking arises from melting of specific phases present within the base material
(e.g. Mg2Si and Si), as discussed in Section 7. 4.2.1 (Chapter 7). Fig. 7.61 provides a basis for
evaluating the HAZ cracking susceptibility:
Wire I
In this case the risk of liquation cracking is small because the solidus temperature of the weld
metal is lower than the actual melting temperatures of the base metal constituent phases.
Wire II
Due to the high Bi(B + D) ratio involved, the solidus temperature of the weld metal will ex-
ceed the actual melting temperatures of the base metal constituent phases. This may lead to
liquation cracking in parts of the HAZ where the peak temperature is greater than, say, 555
to 559 0C.
(h) The sequence of reactions occurring within the HAZ during welding of AA 6082-T6 alu-
minium alloys is shown in Fig. 7.62. In the present case we can use Fig. 4.24 for a quantitative
analysis of the /3" dissolution kinetics. Taking the net heat input per mm2 qjvd equal to 0.08
kJmm~2, we obtain:
Total width of HAZ
Temperature for incipient dissolution of /3"
First we need to estimate the corresponding if/m -coordinate in the HAZ from Fig. 1.31:
Reading then gives:
from which
Total width of fully reverted HAZ:
Temperature for complete dissolution of /3"
First we need to estimate the corresponding if/m -coordinate in the HAZ from Fig. 1.31:
Reading then gives:
from which
A comparison with the phase diagram in Fig. 4.8 shows that the calculated temperature for
incipient dissolution of /3" is in good agreement with that obtained from the solubility prod-
uct.
(i) The yield strength in HAZ and the weld metal can be obtained from Fig. 7.67 and Table
8.10, respectively:
Wire I:
HAZ: Rp02 (min) « 130 MPa, Weld metal: Rp02 « 55 MPa , Base metal: Rp02 « 280 MPa
Strength reduction factor (weld metal control):
Wire II:
HAZ: Rp02 (min) - 130 MPa7WeId metal: Rp02 > 130 MPa , Base metal: Rp02 « 280 MPa
Strength reduction factor (HAZ control):
From this we see that the Al-Mg filler wire (wire II) yields the best weld metal mechanical
properties and should therefore be used, unless the cracking resistance is of particular con-
cern.
(j) When the material is present in the O-temper condition, it will contain an appreciable
amount of the equilibrium /3-Mg2Si phase. This will tend to accelerate the problem with li-
quation cracking within the HAZ during welding.
In addition, it is evident from Figs. 4.4 and 4.8 that the equilibrium /3-Mg2Si phase is ther-
modynamically much more stable than the metastable /3" phase. In practice, this means that
only a narrow solutionised zone will form adjacent to the fusion boundary. However, within
this zone significant strength recovery may occur after welding due to natural ageing effects
(see Fig. 4.5), which may result in a HAZ hardness and tensile yield strength level of about 80
VPN and 190 MPa, respectively. Hence, for the O-tempered material, we get:
Wire I
HAZ: Rp02 (max) - 190 MPa, Weld metal: Rp02^ 55 MPa, Base metal: Rp02 « 100 MPa
Strength reduction factor (weld metal control):
/=55/100 = 0.55
Wire II
HAZ: Rp()2 (max) - 190 MPa, Weld metal: Rp02 > 130 MPa, Base metal: Rp02 - 100 MPa
Strength reduction factor (base metal control):
/ = 100/100 = 1
Author Index
A Berge, J.O. 229 Cochrane, R.C. 292-3, 407, 428

Aaron, H.B. 320, 326, 398-9 Bernstein, LM. 512 Coe,F.R. 128-9,509-10,515
Aaronson, H.I. 408, 429 Betzold,J. 413 Coleman, M.C 259, 263-4
Abson, DJ. 407, 428, 440, 477-8, Bhadeshia, H.K.D.H. 147, 206, 292, Collins, F.R. 537
485,493, 504 408-9, 413, 422-9, 431, 4 3 3 ^ , Corbett, J.M. 203, 428
Adams, CM. 26 436,441,443-4 Corderoy, DJ.H. 151, 155, 160-1
Adrian, H. 301,303 Bhatti,A.R. 208-10 Cotton, H J.U. 496
Agren, J. 320-1,326 Biloni, H. 229 Cottrell, CL.M. 496
Akselsen, O.M. 97, 345, 347, 349, Bjornbakk, B. 486,491 Crafts, W. 190
367,406,414-15,419,444,446, Blander, M. 171,173 Craig, I. 171
448-54, 481, 484, 488-90, 495- Boiling, G.F. 290 Cross, CE. 251,259,292-3,412,538
6, 499, 502-7, 525 Bonnet, C 435, 440 Crossland, B. 556
Alberry, RJ. 374, 500, 502 Bradstreet, BJ. 186
Alcock, CB. 159 Bramfitt, B.L. 244
D
AIi, A. 427-8 Bratland, D.H. 459-62, 556-8, 562
Dallam, CM. 441
American Society for Testing Mate- British Iron and Steels Research As-
D'annessa, A.T. 280
rials (ASTM) 364 sociation 3
Das. G. 452
Andersen, I. 483 Brody, H.D. 272, 276
Dauby, P. 180
Anderson, M.P. 380 Brooks, J.A. 277-8, 533
David, S. A. 96,99,105,210,222,228,
Anderson, RD. 3 Brown, A.M. 345
236, 239-41, 250, 260, 272, 278,
Ankem, S. 343,351 Brown, LT. 509, 511
290,478
Apold,A. 174-5 Brown, L.C. 314
Davis, GJ. 221, 240, 247, 250, 278,
Araki, T. 505-6 Burck, R 289
279,292,478
Ardeil, AJ. 494 Burgardt, R 229
Davis, V. deL. 162
Ashby, M.F. 26, 201, 314, 318, 329,
DeArdo, AJ. 290
360, 363-4, 375, 377-8, 459,
Deb, P. 434, 480
461,464 C
DebRoy,T.210
Asthana, R. 326 Cahn, J.W. 337, 340-1, 345
Delong,WT.533
Atlas of isothermal transformation Cai, X.-L. 450
Demarest, V. A. 449-50
and cooling transformation dia- Cameron, T.B. 413
Devillers, L. 435,440,480, 482
grams 403 Camping, M. 556-8, 562
Devletian,J.H.279,285,413
Avrami, M. 403, 422 Capes, J.F. 251,292-3,412
Dieter, G.E. 482,486, 524-5, 529
Carslaw, H.S. 2, 4
Distin, PA. 157
Chai,CS. 171
B Doherty,R.D.301,309,396
Challenger, K.D. 434, 480
Dolby, R.E. 407,444
Babu, S.S. 210,408,443-4 Chan, J.W. 403
Dons, A.L. 438,459, 541
Bach, H. 138 Charpentier, F.R 435, 440
Dorn, XE. 501-2
Bain, E.C 408, 427, 436 Chen, J.H. 505-6
Dowling, J.M. 203,428
Bakes, R.G. 15 Chew, B. 132, 135
Dube, CA. 408
Baldwin, W.M. 509,511 Chipman, J. 414
Dudas, J.H. 537
Balliger, N.K. 452 Choi, H.S. 450
Dumolt, S.D. 547
Bannister, S.R. 441,443 Christensen, N. 24, 26-7, 31-2, 80,
Barbara, FJ. 435-6, 441-4 88,90-1,97,100, 116, 125, 132,
Barin, I. 154 143,148-50,153,155,158,162, E
Barrie, G.S. 441,443 170-1, 173-4, 176-9, 181-2, Eagar, T.W. 26, 96, 99, 105, 171-2,
Barritte, G.S. 434-6,441 186,189,193,207,345,347,349, 174,228
Baskes, M.I. 277-8 367,500,502,515-17,520-2 Easterling, K.E. 26, 226, 247, 301,
Beachem,CD. 512 Christian, J.W. 329, 400-1, 403-4, 303, 309, 314, 318, 345, 360,
Beaven, RA. 440 429,431 363-4, 367, 375, 377-8, 380,
Bell, H.B. 171,204-5 Cisse, J. 290 389, 392, 403, 408, 427, 429,
Bentley, K.R 15 Claes, J. 180 435-6, 441-4, 448, 500, 502
Ebeling, R. 201 Grewal, G. 343,351 I
Edmonds, D.V. 409 Griffiths, E. 3,4 Ibarra, S. 497
Edwards, G.R. 227,259,422-3,425, Grong, 0 , 26, 61, 73-5, 77, 80, 88, Indacochea, J.E. 171,173
428,441 90-2,116, 149-50,153,155,158, International Institute of Welding
Eickhorn,F. 187-8 161, 163-6, 170, 174, 176-9, 129, 152
Elliott, J.F. 151, 162, 174-5, 179, 181-2, 185-6, 189, 192-204, Ion, J.C. 314, 318, 360, 3 6 3 ^ , 368
182, 184, 191 206-7, 209-11, 227, 247-8, Ivanchev, I. 204-5
Engel,A. 187-8 250-4, 256, 290, 292-4, 314,
Enjo, T. 547 327-30,345-7,349,355,360,364,
Es-Souni, M. 440 367-8, 371-2, 406, 412-15, 419,
J
European Recommendations for Alu- 422-3, 425, 428, 430-2, 435-6,
Jackobs, F.A. 418
minium Alloy Structures 552 438, 440-1, 444, 446-54, 458-62,
Jackson, CE. 89, 99, 100
Evans, G.M. 137-8, 192-3, 203, 464, 465-6, 477-8, 480-1, 484-6,
Jaeger, J.C. 2, 4
420-1,435,440 488-91, 493, 496, 502-7, 547-9,
Jaffe, L.D. 500
551-8,560-3
Janaf, ?. 154
Gunleiksrud, A. 503
Jelmorini, G. 156
F Guo, Z.H. 405,420-1
Jonas, JJ. 464
Fainstein, D. 320, 398-9
Jones, W.K.C 374
Farrar, R.A. 435, 441, 443,480
Jordan, M.F. 143, 145, 259, 263-4
Fast, J.D. 513
H Joshi, Y. 96, 99, 105
Ferrar, R.A. 428,435,444,478,485,
Habrekke, T. 229 Just, E. 413
504
Fine, M.E. 389, 403 Halmoy,E. 151
Fischer, W. A. 162 Hannertz, N.E. 507
Harris, D.R. 414 K
Fisher, DJ. 221, 234, 242, 251, 259,
Harrison, P.L. 428, 435, 444, 478, Kaplan, D. 435, 440, 480, 482
261,265-6,270,274
485, 504 Kasuya, T. 496
Fleck, N.A. 422-3, 428, 441
Hatch, J.E. 3, 458, 547 Kato, M. 233
Flemings, M.C. 221, 234, 242, 265,
Hawkins, D.N. 208-10, 435, 440 Kawasaki, K. 380
272,275-6
Hazzledine, P.M. 342-3 Keene, BJ. 96, 99, 105,228
Fortes, M.A. 374-5, 380-1
Heckel, R.W. 326 Kelly, A. 548
Fountain, R.W. 414
Hehemann, R.F. 429,452 Kelly, K.K. 3
Fradkov, V.E. 380
Heile, R.F. 154, 156-7, 169 Kern, A. 380
Franklin, A.G. 195,208
Heintze, G.N. 244, 247 Kerr, H.W. 203, 247, 273, 290, 428
Fredriksson, H. 290
Heiple, CR. 229 Kiessling, R. 202^4
Frost, HJ. 380
Hellman, P. 339, 343-5 Kihara,H. 131, 133, 134
Fruehan, RJ. 156
Hemmer,H. 371-2 Kikuchi, T. 1 4 2 ^
Fujibayashi, K. 146
Hilbert, M. 339, 343-5 Kikuta, Y. 505-6
Hill,D.C. 154, 156-7, 169 Kim, B.C. 444
Hillert, M. 290 Kim, LS. 450
G
Hilty, D.C. 190 Kim, NJ. 444,451,505
Garcia, C.I. 290
Kim, YG. 451
Garland, J. 505 Hjelen, J. 195, 292, 430-3, 438
Kinsman, K.R. 429, 452
Garland, J.G. 221,240,247,250,278, Hocking, L.M. 196
Kirkwood, RR. 292-3
279, 292-3, 478 Hollomon, J.H. 465, 500
Kluken, A.O. 182, 186, 194-204,
Garret-Reed, AJ. 450 H6llrigl-Rosta,F.413
206,209-11,247-8,250-4,256,
Gergely, M. 501-2 Homma, H. 203, 444-5, 504-5
290, 292-4, 371-2, 430-3, 435-
Giovanola, B. 260 Hondros,E.D. 414
6,438,440,446-7,479-80,484,
Gittos, N.F. 544-5 Honeycombe, R.W.K. 406,408,420,
486,491,493,497
Gjermundsen, K. 162, 516 429, 431, 444, 447-8, 453, 486 Knacke, O. 154
Gjestland, H. 541 Horii,Y. 187-8 Knagenhjem, H.O. 229
Gladman, T. 343-5, 452, 479 Houghton, D.C. 303, 323 Knott, J.F. 486
Gleiser,M. 151,162,174-5,179,191 Howden, D.G. 141 Kobayashi, T. 1 4 2 ^
Goldak,J.A. 515 Howell, PR. 434-6, 441 Kotler, G.R. 320, 326, 398-9
Goolsby, R.D. 306 Hu, H. 337-8, 342-3, 345,430 Kou, S. 27, 75-6, 96, 99, 105, 228,
Greenwood, J.A. 15 Hultgren, R. 3 250, 264-5, 272, 377, 453, 455,
Grest, G.S. 380 Hunderi, O. 337, 341-2, 380 458
Gretoft, B. 147, 422-8
Krauklis, P. 435-6, 441-4
Grevillius, N.F. 182, 185, 188
Kraus, H.G. 228
Krauss, G. 418 Mills, K.C. 96, 99,105,228 P
Kubaschewski, O. 159 Milner, D.R. 141 Paauw, AJ. 446, 503
Kuroda, T. 547 Miranda, R.M. 374-5 Pabi, S.K. 326
Kurz, W. 221, 234, 242, 251, 259, Mitra, U. 171-2 Pakrasi, S. 413
260-1,265-6,270,274 Mizuno, M. 284 Pan, Y-T. 504
Kuwabara, M. 146 Moisio, T. 290 Pande, C S . 339
Kuwana, T. 142-4 Mondolfo, L.F. 2A2-A Pardo, E. 247, 273
Kvaale, FE. 414-15, 419, 504 Mori, N. 187-8 Pargeter, RJ. 428, 440, 477-8, 485,
Morigaki, 0.146 493, 504
Morral, J.E. 413 Patterson, B.R. 339
L Mossinger, R. 554 Pepe,JJ. 541
Lancaster, J.F. 118, 120, 162, 187 Mundra, K. 210 Petch,NJ. 512
Lanzillotto, C.A.N. 452, 505 Munitz, A. 267 Petty-Galis, J.L. 306
Le, Y. 27, 75-6, 265 Murray, J.L. 203 Phillip, R.H. 444
Lee, D.Y. 444 Muzzolani, F.M. 547,550,552 Phillips, H.W.L. 543
Lee, J.-L. 504 Myers, RS. 26 Pickering, RB. 301, 303, 452, 479,
Lei, T.C. 505 Myhr, O.R. 26, 61, 73-5, 77, 314, 505
Li, W.B. 345, 367 327-30, 360, 458-62, 464, 496, Pitsch, W. 436-7
Lifshitz, J.M. 314, 351 547-9,550-5 Plockinger, E. 186
Lindborg,U. 170, 182-3, 185 Porter, D.A. 247,309,389,392,403,
Liu, J.Z. 505 408,413,427,429,435,448
N
Liu, S. 251, 292-3,412,422-3,428, Pottore,N.S.290
Naess, OJ. 503
435-6,441,497 Priestner, R. 451
Nagai, T. 380
Liu, Y. 339 Pugin, A.I. 556-7
Nakagawa,H. 131, 133, 134
Loberg, B. 303
Nakata, K. 271
Lohne, O. 380,459,541
Nes,E. 337, 341-2, 380
Long, CJ. 533, 535 R
Nicholson, R.B. 548
Lucke,K. 337, 340-1,345 Ramachandran, S. 190-1, 204
Niles, R.W. 89
Lutony, MJ. 380 Ramakrishna, V. 151, 162, 174-5,
Nilles,P. 180
179, 191
Nordgren, A. 303
Ramberg, M. 450-1, 505-6
North, T.H. 171
M Ramsay, CW. 480
Nowicki,A. 171
Maitrepierre, Ph. 414 Rappaz, M. 236, 239, 241, 260
Nylund, H.K. 438
Marandet, B. 435, 440, 480, 482 Rath, B.B. 337-8, 342-3, 345
Marder,A.R.451 Ravi Vishnu, P. 500, 502
Marthinsen, K. 380 O Reif, W. 380
Martins, G.P. 182, 185-6, 192-3 O'Brien, J.E. 143, 145 Reiso, O. 541-3
Martukanitz, R.P. 459 Odland, PT. 480 Reti,T. 501-2
Matlock, D.K. 193, 195, 201, 413, Ohkita, S. 203, 444-5, 504-5 Ribes, A. 435, 440, 480, 482
422-3, 428, 432, 435, 440-1, Ohno, S. 143 Riboud, PV. 480, 482
477-8, 480, 485, 488, 491, 493, Ohshita, S. 103, 104,496,515 Ribound, PV. (Riboud ?) 435, 440
504 Ohta, S. 380 Ricks, R.A. 434-5, 436, 441
Matsuda, F. 131, 133, 134, 233, 271 Okumura, M. 103, 104 Ringer, S.P. 345, 367
Matsuda, S. 203, 208, 319, 444-5, Okumura, N. 208, 319 Rollett, A.D. 380
504-5 Olsen, K. 500, 502 Roper, J.R. 229
Matsuda, Y. 505-6 Olson, D.L. 171,173-4,176-9,181- Rorvik, G. 247-8, 250-4, 256, 292-
Matsunawa, A. 96, 99, 105 2, 185-6, 192-3, 422-3, 428, 4, 430, 446-9, 495-6, 499,503,
Mazzolani, F.M. 458 436,441,480,497,500,502 507, 525
McKnowlson, P 15 Onsoien, M.I. 479, 448-9, 495-6, Rose, R. 516
McMahon, CJ. 418 525 Rosenthal, D. 26, 28, 31, 33, 38, 41,
McPherson, R. 244, 247 Oreper, G.M. 96, 99, 105, 228 48, 51, 56, 59-61, 76, 98, 133,
McQueen, HJ. 464 Oriani, R.A. 514 360
McRobie, D.E. 486 Orr, R.L. 3 Roux, R. 140
Mehl, R.F. 408, 436-7 Ostrom, G. 533 Rykalin, N.N. 18, 21, 26, 41, 45, 56,
Metals Handbook 3, 545 0verlie, H.G. 541 93,95, 556-7
Midling, O.T. 465-6, 556-8, 560-3 Owen, W.S. 450 Ryum, N. 326,337,341-2,345,347,
Miller, R.L. 449-50 Ozturk,B. 156 349,367,380,382,390,396,403,
Mills, A.R. 435, 440 541-3
S T W
Saetre, T.O. 380, 382 Takalo,T.290 Wagner, C 201, 314, 351
Saggese, ME. 208-10, 435,440 Tamehiro, H. 496 Wahlster, M. 186
Sagmo, G. 97 Tamura, 1.405,420-1 Walsh, R. A. 190-1,204
Saito, S. 103, 104 Tanigaki, T. 146 Walton, D.T. 380
Sakaguchi, A. 284 Tanzilli, R.A. 326 Wang, YH. 96, 99, 105, 228
Savage, W.F. 541 Tardy, P. 501-2 Weatherly, G.C. 303, 323
Schaeffler, A.L. 533 Tensi, H.M. 21,25 Welding Handbook 24
Scheil, E. 403 Thaulow, C. 503 Welz, W. 21,25
Schriever, U. 380 Themelis, NJ. 120,162,183,187,281, Whelan,MJ. 319, 356
Schumacher, J.E 162 284 Whiteman, J.A. 208-10, 435, 440
Schwan,M.21,25 Therrien,A.E.434,480 Widgery, DJ. 480-1
Scott, MH. 544-5 Thewlis, G. 203,435,440-1 Willgoss, R.A. 132
Seah,MR414 Thivellier, D. 414 Williams, J.C 533
Seay, E.B. 96, 99, 105 Thomas, G. 505 Williams, TM. 414
Senda, T. 233 Thompson, A.W. 480,512 Wolstenholme, D.A. 174
Shackleton, D.N. 166-7 Thompson, CV. 380 Woods, WE. 279, 285
Shaller,F.W.509,512 Tichelaar, G.W. 156 Worner, CH. 342-3
Shen, H.P. 505 Tjotta, S. 459,460 Wriedt, H.A. 203
Shen,X.P.451 Tomii,Y.284
Sherby, O.D. 501-2 Torsell, K. 182-3,185
Shercliff, H.R. 314,329,459-62,464 Tricot, R. 414 Y
Shinozaki, K. 131, 133, 134 Trivedi, R. 260, 400,427 Yamamoto, K. 203, 444-5, 504-5
Siewert, T.A. 182, 185-6, 192-3, Troiano,A.R.509,512 Yang, J.R. 428-9
227,425,428 Tsai,N.S.26 Yi, JJ. 450
Sigworth, G.K. 162 Tsukamoto, K. 271 Yoneda, M. 505-6
Simonsen, T. 520-2 Tundal, UH. 326 Yurioka, N. 103, 104,496, 509,515,
Sims, CE. 512 Turkdogan, E.T. 126, 182, 184-6, 520
Skaland, T. 346 191-2,195-6,207,214
Skjolberg, E.M. 140-1 Turpin, M.L. 182,184
Slyozov,V.V. 314,351 Z
Smith, A. A. 166-7, 169, 170 Zacharia, T. 96, 99, 105, 228
Soares, A. 380-1 U Zapffe,CA.512
Solberg, J.K. 446, 450-4, 504-6 Uda, M. 143 Zener, C 341-2,344,465
Sommerville, LD. 204-5 Udler, D.G. 380 Zhang, C 515
Speich, G.R. 449-50 UIe, R.L. 96, 99, 105 Zhang, Z. 441,443
Srolovitz, DJ. 380 Umemoto, M. 405, 420-1
Staley, J.T. 394-5, 459 Underwood, E.E. 201, 338
Steidl, G. 554 Unstinovshchikov, J.I. 494
Steigerwald,E.A.509,512
Stjerndahl, J. 290
Stoneham, A.M. 414 V
Strangwood, M. 428-9, 431, 444 Van Den Heuvel, G J P M . 156
Strid, J. 303, 542-3 Van Stone, R.H. 480
Stuwe,H. 337, 340-1,345 Vander Voort, G.F. 394, 403
Sugden, A.A.B. 292,431 Vandermeer, R.A. 341
Suutala, N. 290 Vasil'eva, VA. 556-7
Suzuki, H. 406, 444, 477, 496, 509, Verhoeven, J.D. 286, 429, 431, 433,
515,520 448-9
Suzuki, S. 303, 323 Villafuerte, XC 247,273
Svensson, L.E. 147, 206, 413, 422- Vitek, J.M. 96,99,105,210,222,228,
8, 431, 4 3 3 ^ , 441, 444, 536 240,250,272,278,478
Szekely, J. 96,99,105,120,162,183,
187,228,281,284
Szewezyk, A.F. 505
Szumachowski, E. 533
Index
Index terms Links
A
absorption of elements see hydrogen, nitrogen, oxygen
acicular ferrite in low-alloy steels 428
crystallography of 428
nature of 430
nucleation and growth of 432
texture components of 429
acicular ferrite in wrought steels 444
aluminium as alloying element in steel
effect on inclusion composition 202 206
effect on solidification microstructure 246 272 293
effect on weld properties 481 486
solubility product of precipitates 303
aluminium weldments 458 536
age-hardenable alloys 458
quench sensitivity 459
precipitation conditions during cooling 459
strength recovery during natural ageing 461
subgrain evolution in friction welding 464
characteristics 536
constitutional liquation
in Al-Mg-Si alloys 542
in Al-Si alloys 541
example (7.9) – minimum HAZ strength level 554
example (7.8) – weld metal hot cracking 544
example (7.7) – weld metal solidification cracking 537
example (7.10) – minimum HAZ hardness level 562
HAZ microstructure evolution 547
This page has been reformatted by Knovel to provide easier navigation. 595
596
Index terms Links
aluminium weldments (Continued)
constitutive equations 548 558
during friction welding 555
during fusion welding 547
hot cracking 540
factors affecting 544
solidification cracking 536
strength evolution during welding 547
constitutive equations 548 558
during friction welding 555
hardness and strength distribution 550 560
strengthening mechanisms in alloys 547
amplitude of weaving – definition 80
arc atmosphere composition 132
see also shielding gases
arc efficiency factors 26
definition 26
selected values 27
arc welding 24
definition of processes 24
austenite
grain size in low-alloy steels 409
primary precipitation in fusion welds 292
austenite formation in low-alloy steels 449
conditions for 450
austenitic stainless steels 453
see also stainless steel weldments
characteristics of 527
chromium carbide formation 456
grain growth diagrams for steel welding 375
weld decay area 456

597
Index terms Links
Avrami equation in solid state transformations
see also solid state transformation and transportation kinetics
additivity in 404 475
exponents in 401
B
Bain orientation region 436
bainite in low-alloy steels 444
lower 447
upper 444
bead morphology 96
bead penetration 99
deposit and fused parent metal 96
example (1.16) – SA welding of steel 97
example (1.17) – SMAW welding of steel 98
example (1.18) – Jackson equation 99
Bessel functions – modified 46 47 49
boron in steel
effect on transformation behaviour 413
segregation of 294
weld properties 493 505
bowing of crystal 240
Bramfitt’s planar lattice disregistry model 244
see also solidification of welds
C
carbon equivalents 496 521
carbon as alloying element in steel
weld deposits 424
carbon-manganese steel weld metals, grain growth in 370
casting, structural zones 221

598
Index terms Links
cell/dendrite alignment angle 249
cellular substructure 251
chemical reaction model – overall 116
chromium carbide formation in austenitic stainless steels 456
chromium-molybdenum steel welds, grain growth in 372
columnar grains 228
columnar to equiaxial transition 268
competitive grain growth 234
concentration displacements during welding see oxygen, absorption of
cooling condition during solidification 221
cooling rate, C.R.
thick plate welding 37
thin plate welding 53
cooling time, ∆t8/5
thick plate welding 36
thin plate welding 53
cooling time, t100 103
D
Delong diagram 535
delta ferrite, primary precipitation of 290 292
dendrite arm spacing 261
primary 261
secondary 264
dendrite fragmentation 250

599
Index terms Links
dendrite substructure 252
dendrite tip radius 260
deoxidation reactions in weld pools 180
example (2.9) – homogeneous nucleation of MnSiO3 182
growth and separation of oxide inclusions 184
buoyancy (Stokes flotation) 185
fluid flow pattern 186
separation model 188
nucleation model 182
nucleation of inclusions 182 219
overall deoxidation model 201
deposit – amount of weld metal 96
deposition rate 96
dissociation of gases in arc column 117
distributed heat sources 77 112
general solution 77
simplified solution (Gaussian heat distribution) 112
simplified solution (planar heat distribution) 80
case study (1.2) – surfacing with strip electrodes 87
case study (1.3) – GTA welding with a weaving technique 87
dimensionless operating parameter 82
dimensionless time 82
dimensionless y- and z-coordinates 82
example (1.13) – effect of weaving on temperature distribution 83
implications of model 86
model limitations 86
2-D heat flow model 80
see also heat flow models
Dorn parameter 501

600
Index terms Links
duplex stainless steels 531
HAZ toughness 532
HAZ transformation behaviour 532
E
energy barrier to solidification 225
enthalpy of reaction 302
definition of 302
values 303
entropy of reaction 302
definition 302
values 303
epitaxial solidification 222
equiaxed dentritic growth 268
equilibrium dissolution temperature of precipitates 303
error functions see Gaussian error functions
F
fluid flow pattern in weld pools 186 228
flux basicity index 171
friction welding 18
see also aluminium weldments
dimensionless temperature 20
dimensionless x-coordinate 21
example (1.4) – peak temperature distribution 23
heat flow model 18
temperature-time pattern 23
Fritz equation 281

601
Index terms Links
fume formation, rate of
iron 157
manganese 156
silicon 152
fused parent metal – amount of 98
G
gas absorption, kinetics of 120
rate of element absorption 121
thin film model 120
gas desorption, kinetics of 123
rate of element desorption 123
Sievert’s law 124
gas porosity in fusion welds 279
growth and detachment of gas bubbles 281
nucleation of gas bubbles 279
separation of gas bubbles 283
Gaussian error functions, definition 112
Gaussian heat distribution 112
see also distributed heat sources
Gibbs-Thomson law 309
Gladman equation 344
grain boundary ferrite 408
growth of 422
nucleation of 408
grain detachment 250
grain growth 337
computer simulation 380
diagrams
construction of 360

602
Index terms Links
grain growth (Continued)
axes and features of 363
calibration procedures 361
heat flow models 360
for steel welding 360
case studies 364
C-Mn steel weld metals 370
Cr-Mo low alloy steels 372
niobium-microalloyed steels 367
titanium-microalloyed steels 364
type 316 austenitic stainless steels 375
driving pressure for 339
example (5.3) – austenite grain size in niobium-microalloyed steels 358
example (5.2) – austenite grain size in Ti microalloyed steels 354
example (5.1) – limiting austenite grain size in steel weld metals 344
grain boundary mobility 337
drag from impurities 340 342
drag froma random particle distribution 341
driving pressure for growth 339
grain structures, characteristics 337
growth mechanisms 345
nomenclature 384
normal grain growth 343
size, limiting 343
Griffith’s equation 486 494
gross heat input – definition 37
growth rate of crystals 230
local 234
nominal 230

603
Index terms Links
H
Hall-Petch relation 529
heat flow models
distributed heat sources 77 112
grain growth diagrams 360
instantaneous heat sources 5
local preheating 100
medium thick plate solution 59
thermal conditions during interrupted welding 91
thermal conditions during root pass welding 95
thick plate solutions 26
thin plate solutions 45
heat input see heat flow models
Hellman and Hillert equation 344
Hollomon-Jaffe parameter 500
hydrogen, absorption of 128
content in welds 132
covered electrodes 134
combined partial pressure of 134
example (2.1) – hydrogen absorption in GTAW 133
example (2.2) – hydrogen absorption in SMAW 136
in gas-shielded welding 131
hydrogen determination 128
implications of Sievert’s law 140
reaction model 130
sources of hydrogen 128
in submerged arc welding 138
effect of water content in flux 138
example (2.3) – hydrogen absorption in SAW 139
hydrogen cracking in low-alloy steel weld metals 509
diffusion in welds 514
diffusivity in steel 514

604
Index terms Links
hydrogen cracking in low-alloy steel weld metals (Continued)
HAZ cracking resistance 518
mechanisms of 509
solubility in steel 513
hydrogen in multi-run weldments 140
hydrogen in non-ferrous weldments 141
hydrogen sulphide corrosion cracking in low-alloy steel weld metals 524
prediction of 525
threshold stress for 524
hyperbaric welding 176
I
implant testing 520
see also hydrogen cracking
inclusions in welds – origin 192
constituent elements and phases in inclusions 202
example (2.10) – computation of inclusion volume fraction 194
example (2.12) – computation of total number of constituent phases in
inclusions 211
prediction of inclusion composition 204
size distribution of inclusions 195
coarsening mechanism 196
effect of heat input 196
example (2.11) – computation of number density and size
distribution of inclusions 201
volume fraction 193
stoichiometric conversion factors 194
instantaneous heat sources 5
line source 5
plane source 5
point source 5
interface stability 254

605
Index terms Links
interfacial energies 242
interrupted welding, thermal conditions 91
example (1.14) – repair welding of steel casting 93
heat flow models 93
K
Kurdjumow-Sachs orientation relationship 408 427 429
444 448
L
latent heat of melting 3
lattice disregistry
see Bramfitt’s planar lattice disregistry model
local fusion in arc strikes 7
dimensionless radius vector 7
example (1.1) – weld crater formation and cooling conditions 9
heat flow model 7
low-alloy steel weldments 477
acicular ferrite in 428
nature of 430
nucleation and growth in 432
texture components of 429
austenite formation in 449
conditions for 450
bainite in 444
lower 447

606
Index terms Links
low-alloy steel weldments (Continued)
upper 444
case study (7.1) – weld bead tempering 501
example (7.1) – low-temperature toughness of welds 488
example (7.2) – peak HAZ strength level 496
example (7.3) – location of brittle zones 508
HAZ mechanical properties 494
hardness and strength level 495
tempering 500
toughness 502
hydrogen cracking 509
diffusion in welds 514
diffusivity in steel 514
example (7.4) – hydrogen cracking under hyperbaric welding
conditions 521
HAZ cracking resistance 518
implant testing 520
mechanisms of 509
solubility in steel 513
hydrogen sulphide corrosion cracking 524
prediction of 525
threshold stress for 524
martensite in 447
austenite formation, kinetics of 449
lath 447
M-A formation, conditions for 450
plate (twinned) 447
mechanical properties 477
ductile to brittle transition 486
reheating 491
resistance to cleavage fracture 485
resistance to ductile fracture 480
strength level 478

607
Index terms Links
low-alloy steel weldments (Continued)
transformation behaviour 290 406
solidification
primary precipitation of austenite 292
primary precipitation of delta ferrite 290 292
solid state
acicular ferrite 428
bainite 444
grain boundary ferrite 408
martensite 447
microstructure classification 406
nomenclature for 406
Widmanstatten ferrite 427
Ludwik equation 524
M
magnesium in aluminium alloys
martensite
in low-alloy steels 447
austenite formation, kinetics of 449
lath 447
M-A formation, conditions for 450
plate (twinned) 447
martensitic stainless steels, characteristics of 527
mass transfer in weld pool, overall kinetic model of 124
medium thick plate solution 59
dimensionless maps for heat flow analysis 61
case study (1.1) – temperature distribution in steel and aluminium
weldments 69
construction of maps 61

608
Index terms Links
medium thick plate solution (Continued)
cooling conditions close to weld centre line 63
example (1.12) – aluminium welding 68
isothermal contours 65
limitation of maps 65
peak temperature distribution 61
retention times at elevated temperatures 63
experimental verification 72
peak temperature and isothermal contours 75
weld cooling programme 72
weld thermal cycles 72
general heat flow model 59
practical implications 75
melting efficiency factor 89
mixing ratio 98
moving heat sources 24
net arc power, definition 26
niobium-microalloyed steels, grain growth in 367
nitrogen, absorption of 141
gas-shielded welding 142
nominal composition 147
sources of 142
submerged arc welding 146
example (2.4) – nitrogen content in weld metal deposit 146
N
non-isothermal transformations
additivity principle 403
and Avrami equation 404

609
Index terms Links
non-isothermal transformations (Continued)
isokinetic reactions 404
non-additive reactions 405
non-steady heat conduction
biaxial conduction 2
triaxial conduction 2
uniaxial conduction 2
nucleation, energy barrier to 225
nucleation, homogeneous 182 219
see also deoxidation reactions in weld pools
nucleation, potency of particles 242
nucleation, rate of heterogeneous during solidification 248 272
nucleation in solid state transformation kinetics 389
in C-curve modeling 390
nucleation of gas bubbles in fusion welds 279
nucleation of grain boundary ferrite in low-alloy steels 408
austenite grain size 409
boron alloying 413
factors affecting ferrite grain size 420
solidification-induced segregation 417
O
operating parameter, dimensionless
point and line heat source models 31
weaving model 82
Ostwald ripening see particle coarsening
oxygen, absorption of 148
classification of shielding gases 166
overall oxygen balance 166

610
Index terms Links
oxygen, absorption of (Continued)
absorption of carbon and oxygen 176
loss of silicon and manganese 177
the product [%C] [%O] 179
reaction model 174
effects of welding parameters 169
amperage 169
voltage 170
welding speed 170
example (2.8) – oxygen consumption and total CO evolution during
GMAW 166
gas arc metal welding 148
manganese evaporation 156
example (2.6) – fume formation rate of manganese 157
sampling of elevated concentrations 149
carbon oxidation 149
silicon oxidation 152
example (2.5) – fume formation rate of silicon 156
SiO formation 154
total oxygen absorption 162 173
transient oxygen concentrations 160
example (2.7) – slag formation in GMAW 164
submerged arc welding 170
concentration displacements 172
flux basicity index 171
total oxygen absorption 173
transient oxygen absorption 172
oxygen, retained in weld metal 190
implications of model 192
thermodynamic model of 190

611
Index terms Links
P
particle coarsening 314
applications to continuous heating and cooling 314
example (4.4) – coarsening of titanium nitride in steel 315
kinetics 314
particle dissolution 316
analytical solution 316
case study (4.1) – solute distribution across HAZ 330
example (4.5) – isothermal dissolution of NbC in steel 320
example (4.6) – dissolution of NbC within fully transformed HAZ 323
numerical solution 325
application to continuous heating and cooling 329
process diagrams for aluminium butt welds 332
Peclet number for weld pools 186
peritectic solidification in welds 290
see also low alloy steel weldments
primary precipitation of γp-phase 290
transformation behaviour 290
precipitate growth mechanisms
liquid state 196
solid state 395
diffusion-controlled 397
interface-controlled 396
precipitate stability 301
see also particle coarsening and particle dissolution
example (4.1) – equilibrium dissolution temperature of nitride
precipitates 304
example (4.2) – equilibrium volume fraction of Mg2Si 307
example (4.3) – metastable β”(Mg2Si) solvus 312
nomenclature 334
solubility product 301
equilibrium dissolution temperature 303

612
Index terms Links
precipitate stability (Continued)
stable and metastable solvus boundaries 304
thermodynamic background 301
preheating, local 100
heat flow model 100
dimensionless half width of preheated zone 101
example (1.19) – cooling conditions during steel welding 102
time constant 101
pseudo-equilibrium, concept of 122
pseudo-steady state temperature distribution, definition 24
R
reversion see particle dissolution
example (1.15) – cooling conditions during root pass welding 95
heat flow model 95
Reynold number
definition 187
of gas bubbles 284
of particles 187
root pass welding, thermal conditions in 95
Rosenthal equations
see thick and thin plate solutions
S
Scheil equation 272
modified 276
original 272
separation of gas bubbles in fusion welds 283
shielding gases see oxygen, hydrogen and nitrogen, absorption of
CO-evolution 166

613
Index terms Links
Sievert’s law 124 140
silicon in aluminium
solid state transformations in welds 387
Al-Mg-Si alloys 458
Avrami equation in, additivity in 404 475
high strength low-alloy steels 406
kinetics see transformation kinetics
nomenclature 471
solid state transformation kinetics 387
see also transformation kinetics
driving force for 387
non-isothermal transformations 402
nucleation in solids 389
overall 400
precipitates, growth of 395
solidification cracking in weldments
aluminium 536
stainless steel 532
solidification microstructures 251
columnar to equiaxed transition 268
dendrite tip radius 260
equiaxed dendritic growth 268
example (3.12) – equiaxed dendritic growth in Al-Si welds 270
example (3.13) – application of Scheil equation 276
interface stability criterion 254
example (3.6) – critical temperature gradient for planar solidification
front in Al-Si welds 256
example (3.7) – substructure characteristics of Al-Mg welds 258
primary dendrite arm spacing 261

614
Index terms Links
solidification microstructures (Continued)
example (3.8) – effect of heat input on primary dendrite arm spacing
in welds 262
example (3.9) – variation of primary dendrite arm spacing across
fusion zone 263
secondary dendrite arm spacing 264
example (3.10) – secondary dendrite arm spacing in thick plate GTA
Al-Si welds 266
example (3.11) – secondary dendrite arm spacing in thin plate GTA
Al-Si butt welds 267
local solidification time 265
substructure characteristics 251
cellular 251
dendritic 252
solidification of welds 221
columnar grain structures and morphology 228
epitaxial solidification 222
energy barrier to solidification 225
implications of 226
growth rate of columnar grains 230
example (3.1) – nominal crystal growth rate in thin sheet welding of
example (3.2) – local dendrite growth rate in single crystal welds 237
local crystal growth rate 234
nominal crystal growth rate 230
renucleation of crystals 242
critical cell-dendrite alignment angle 249
dendrite fragmentation 250
example (3.4) – nucleation potency of TiN with respect to delta
ferrite 245
example (3.5) – nucleation potency of γ-Al2O3 with respect to delta
ferrite 246
grain detachment 250

615
Index terms Links
solidification of welds (Continued)
nucleation potency of second phase particles 242
rate of heterogeneous nucleation 247
reorientation of columnar grains 239
bowing of crystal 240
example (3.3) – bowing by dendritic branching 240
structural zones 221
solubility of gases in liquids and solids 125
hydrogen in Al 125
hydrogen in Cu 125
hydrogen in Fe 125 513
hydrogen in Ni 125
nitrogen in Fe 126
see also gas absorption and gas desorption
solubility product 301
equilibrium dissolution temperature 303
stable and metastable solvus boundaries 304
thermodynamic background 301
solute redistribution in welds 272
example (3.14) – formation of hydrogen bubbles in weld pools 282
example (3.15) – separation of hydrogen bubbles in weld pools 284
example (3.16) – solute redistribution during cooling in austenite
regime 287
gas porosity 279
homogenisation of microsegregations 286
macrosegregation 277
microsegregation 272
spot welding 10
dimensionless radius vector 11
example (1.2) – cooling conditions 12

616
Index terms Links
spot welding (Continued)
heat flow model 11
refined model for 110
stainless steel weldments 527
see also austenitic stainless steels
austenitic
characteristics of 527
chromium carbide formation 456
grain growth diagrams for steel welding 375
example (7.5) – variation in HAZ austenite grain size and strength
level 530
example (7.6) – weld metal solidification cracking 533
HAZ corrosion resistance 527
HAZ strength level 529
HAZ toughness 530
solidification cracking 532
weld decay area 456
duplex
HAZ toughness 532
HAZ transformation behaviour 532
stereometric relationships (number of particles per unit volume, number
of particles per unit area, total surface area per unit volume, and
mean particle volume spacing) 201
Stokes law 185 187 284
substructure of welds 251
T
texture in welds
solidification 221 290
solid state 429

617
Index terms Links
thermal properties of metal and alloys 3
conductivity 3
diffusivity 3
heat content at melting point 3
latent heat of melting 3
melting point 3
volume heat capacity 3
thermit welding 14
example (1.3) – cooling conditions 16
heat flow model 14
thick plate solutions 26
pseudo-steady state temperature distribution 31
dimensionless y-coordinate 31
dimensionless z-coordinate 31
distribution of temperatures 31
example (1.5) – duration of transient heating period in aluminium
welding 30
example (1.6) – thermal contours 37
example (1.7) – weld geometry 39
isothermal zone widths 32
length of isothermal enclosures 34
simplified solution 41
example (1.8) – retention time in steel welding 44

618
Index terms Links
thick plate solutions (Continued)
volume of isothermal enclosures 35
transient heating period 29
thin plate solutions 45
example (1.9) – duration of transient heating period in aluminium
welding 48
pseudo-steady state temperature distribution 49
example (1.10) – weld geometry and cooling rate 54
isothermal zone widths 49
length of isothermal enclosures 51
simplified solution 56
example (1.11) – retention time in steel welding 59
transient heating period 29
titanium as alloying element in steel
effect on inclusion composition 203 208
effect on solidification microstructure 244 272
effect on grain growth 354 364
effect on transformation behaviour 435 444
effect on weld properties 488
titanium-microalloyed steels, grain growth in 364
see also low alloy steel weldments
transformation kinetics 387
Avrami equation 400 475
additivity in 404 475
exponents in 401
driving force for 387

619
Index terms Links
transformation kinetics (Continued)
example (6.1) – C-curve analysis 394
example (6.2) – conditions for ferrite formation within HAZ 410
example (6.3) – volume fraction of grain boundary ferrite in HAZ 412
example (6.4) – ferrite/martensite formation in HAZ 416
example (6.5) – displacement of ferrite C-curve due to segregation 418
example (6.6) – variation in ferrite grain size across HAZ 421
example (6.7) – volume fraction of allotriomorphic ferrite in weld
deposit 425
example (6.8) – volume of acicular ferrite plate 440
example (6.9) – conditions for acicular ferrite formation 442
example (6.10) – conditions for chromium carbide formation 456
example (6.11) – conditions for β’(Mg2Si) precipitation 460
example (6.12) – ageing characteristics of aluminium weldments 463
non-isothermal transformations 402
nucleation in solids 389
overall 400
precipitates, growth of 395
type 316 austentitic stainless steels, grain growth in 375
see also stainless steel weldments
V
volume of weld metal 36
volume fraction of inclusions 193
volume heat capacity 3
W
Wagner-Lifshitz equation 196 314 351
water content 137
in electrode coating 137
in welding flux 138
see also hydrogen absorption

620
Index terms Links
weld pool shape and geometry 228
elliptical weld pool 229
tear-shaped weld pool 229
welding processes, definitions 24
see also arc welding processes
wetting conditions 222 242
interfacial energies 242 247
wetting angle 225
Widmanstätten ferrite in low-alloy steels 427
Z
Zener drag, definition of 341
in grain growth 341
Zener equation 342 344
Zener-Hollomon parameter 465
zinc in aluminium

Metallurgical Modelling of Welding

Uploaded by

Copyright:

Available Formats

Metallurgical Modelling of Welding

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Metallurgical Modelling of Welding

Uploaded by

Copyright:

Available Formats

Metallurgical

MATERIALS MODELLING SERIES

First edition (Book 557)

The Institute of Materials 1997

ISBNl 86125 036 3

Trondheim, 28 October, 1996

Trondheim, 1 December, 1993

Preface to the Second Edition ........................................................ xiii

Preface to the First Edition ............................................................. xiv

1. Heat Flow and Temperature Distribution in Welding ........... 1

This page has been reformatted by Knovel to provide easier navigation. vi

1.10.3.3 Simplified Solution for a Fast Moving High

2. Chemical Reactions in Arc Welding ...................................... 116

This page has been reformatted by Knovel to provide easier navigation.

2.5 The Concept of Pseudo-equilibrium .......................................... 122

This page has been reformatted by Knovel to provide easier navigation.

2.10.1.6 Classification of Shielding Gases ................ 166

This page has been reformatted by Knovel to provide easier navigation.

2.12.4 Prediction of Inclusion Composition ............................. 204

3. Solidification Behaviour of Fusion Welds ............................ 221

This page has been reformatted by Knovel to provide easier navigation.

3.6 Equiaxed Dendritic Growth ....................................................... 268

4. Precipitate Stability in Welds ................................................. 301

This page has been reformatted by Knovel to provide easier navigation.

4.3 Particle Coarsening ................................................................... 314

5. Grain Growth in Welds ........................................................... 337

This page has been reformatted by Knovel to provide easier navigation.

5.3.3 Grain Growth Mechanisms .......................................... 345

6. Solid State Transformations in Welds ................................... 387

This page has been reformatted by Knovel to provide easier navigation.

6.2.2 Heterogeneous Nucleation in Solids ............................ 389

This page has been reformatted by Knovel to provide easier navigation.

6.3.7 Bainite .......................................................................... 444

7. Properties of Weldments ........................................................ 477

This page has been reformatted by Knovel to provide easier navigation.

7.2.1.5 Effects of Reheating on Weld Metal

This page has been reformatted by Knovel to provide easier navigation.

7.4.3.1 Effects of Reheating on Weld Properties ..... 547

8. Exercise Problems with Solutions ......................................... 571

Index .............................................................................................. 595

Author Index ................................................................................. 602

This page has been reformatted by Knovel to provide easier navigation.

Welding metallurgy is concerned with the application of well-known metallurgical principles

High peak temperatures, up to several thousand 0 C.

It follows that a quantitative analysis of metallurgical reactions in welding requires detailed

1.2 Non-Steady Heat Conduction

For biaxial and triaxial heat conduction we may write by analogy:1

1.3 Thermal Properties of Some Metals and Alloys

Material (WrTIm-10C-1) (mm2 s"1) (Jmnr 3 0C"1) (0C) (J mnr 3 ) (J mnr 3 )

Carbon 0.040 8 0.005 1520 7.50 2.0

Does not include the latent heat of melting (AH1n).

1.4 Instantaneous Heat Sources

(i) Plane source in a long rod (Fig. 1.3a):

(ii) Line source in a wide plate (Fig. 1.3b):

(iii) Point source in a heavy slab (Fig. 1.3c):

In order to obtain a general survey of the thermal programme, it is convenient to write

where Tc is the chosen reference temperature.

where tt is the arc ignition time.

— Dimensionless operating parameter:

— Dimensionless radius vector: