Fatigue and Crack Initiation - George C. Sih

Multiscale Fatigue Crack Initiation and Propagation of
Engineering Materials: Structural Integrity and

Microstructural Worthiness
SOLID MECHANICS AND ITS APPLICATIONS
Volume 152
Series Editor: G.M.L. GLADWELL

Department of Civil Engineering
University of Waterloo
Waterloo, Ontario, Canada N2L 3GI
Aims and Scope of the Series

The fundamental questions arising in mechanics are: Why?, How?, and How much?
The aim of this series is to provide lucid accounts written by authoritative researchers
giving vision and insight in answering these questions on the subject of mechanics as it
relates to solids.
The scope of the series covers the entire spectrum of solid mechanics. Thus it includes
the foundation of mechanics; variational formulations; computational mechanics;
statics, kinematics and dynamics of rigid and elastic bodies: vibrations of solids and
structures; dynamical systems and chaos; the theories of elasticity, plasticity and
viscoelasticity; composite materials; rods, beams, shells and membranes; structural
control and stability; soils, rocks and geomechanics; fracture; tribology; experimental
mechanics; biomechanics and machine design.
The median level of presentation is the first year graduate student. Some texts are
monographs defining the current state of the field; others are accessible to final year
undergraduates; but essentially the emphasis is on readability and clarity.
For a list of related mechanics titles, see final pages.

G.C. Sih
Editor
Multiscale Fatigue Crack

Initiation and Propagation of
Engineering Materials: Structural
Integrity and Microstructural
Worthiness
Fatigue Crack Growth Behaviour of Small and Large Bodies
G.C. Sih
East China University of Science
Shanghai, China
and
Institute of Fracture and Solid Mechanics
Lehigh University
Bethlehem, Pennsylvania, USA
ISBN-13: 978-1-4020-8519-2 e-ISBN-13: 978-1-4020-8520-8
Library of Congress Control Number: 2008xxxxx
© 2008 Springer Science+Business Media, B.V.

No part of this work may be reproduced, stored in a retrieval system, or transmitted
in any form or by any means, electronic, mechanical, photocopying, microfilming, recording
or otherwise, without written permission from the Publisher, with the exception
of any material supplied specifically for the purpose of being entered
and executed on a computer system, for exclusive use by the purchaser of the work.
Printed on acid-free paper
987654321
springer.com
Table of Contents
Foreword vii
Contributors xi
Application of Virtual Testing for Obtaining Fracture Allowable of

Aerospace and Aircracft Materials 1
B. Farahmand
An Equivalent Block Approach to Crack Growth 23

R. Jones, S. Pitt and D. Peng
Prediction of Fatigue Crack Growth Rates in Ti-6Al-4V Alloy 47

A. M. Korsunsky, D. Dini and M. J. Walsh
Some Practical Implications of Exponential Crack Growth 65

L. Molent, S. Barter and R. Jones
Fatigue Behaviour of FS, LB and MIG Welds of AA6061-T6 and

AA6082-T6 85
P. M. G. P. Moreira, V. Richter-Trummer and P. M. S. T. de Castro
Fatigue Damage from Surface to Bulk 113

C. A. Rodopoulos
Microcracking in High Temperature Low Cycle Fatigue 133

S. L. Mannan and M. Valsan
Invariant Form of Micro-/Macro-Cracking in Fatigue 181

G. C. Sih
Fatigue Crack Growth Rate of Cable-Stayed Portion of Runyang Bridge:

Part I – Cable Crack Growth Due to Disproportionate Cable Tightening/
Loosening and Traffic Loading 209
G. C. Sih and X. S. Tang
v
vi Table of Contents
Fatigue Crack Growth Rate of Cable-Stayed Portion of Runyang Bridge:

Part II – Steel Wire Crack Growth Due to Disproportionate Cable Tightening/
Loosening and Traffic Loading 249
G. C. Sih and X. S. Tang
Fatigue of Small-Scale Metal Materials: From Micro- to Nano-Scale 275

G. P. Zhang and Z. G. Wang
Assessment of Fatigue Damage in Heterogeneous Materials by Application

of a Novel Compliance Technique 327
H. Mughrabi and H. W. Höppel
Fatigue Crack Growth of Aircraft Aluminum Alloys 345

Sp. G. Pantelakis and Al. Th. Kermanidis
Author Index 367
Subject Index 369

Foreword
What can be added to the fracture mechanics of metal fatigue that has not
already been said since the 1900s? From the view point of the material and
structure engineer, there are many aspects of failure by fatigue that are in
need of attention, particularly when the size and time of the working
components are changed by orders of magnitude from those considered by
traditional means. The 21st century marks an era of technology transition
where structures are made larger and devices are made smaller, rendering
the method of destructive testing unpractical. While health monitoring
entered the field of science and engineering, the practitioners are
discovering that the correlation between the signal and the location of
interest depends on a priori knowledge of where failure may initiate. This
information is not easy to find because the integrity of the physical system
will change with time. Required is software that can self-adjust in time
according to the monitored data. In this connection, effective application of
health monitoring can use a predictive model of fatigue crack growth.
Earlier fatigue crack growth models assumed functional dependence on
the maximum stress and the size of the pre-existing crack or defect. Various
possibilities were examined in the hope that the data could be grouped such
that linear interpolation would apply. The idea of associating fatigue test
data with design was made when the data led to a straight line relationship
with the range of mode I stress intensity factor on a log-log plot. The
y-intercept and slope of this line were found by curve fitting and considered
to be material specific even though they also depended on specimen size.
Wide acceptance of this approach was soon adopted due to the simplicity of
integrating the two-parameter relation for determining the critical crack
length for different specimen configuration. Moreover, the stage of slow
and stable crack growth was able to be delineated from the onset of unstable
and rapid crack propagation. Even at this date efforts are being made to
better understood the physical meaning of the crack growth rate data outside
the range of the two-parameter straight line zone referred to as region II.
Region I and III are associated, respectively, with fatigue crack initiation
and fast crack propagation, not to mention the two transitional zones where
region II connects to regions I and III. Despite the exhaustive undertakings
vii
viii Foreword
to examine microphotographs the phenomenon of multiscale fatigue crack

growth remains elusive.
One of the main objectives of this book is to examine size-scale
dependent microstructure evolution of fatigue damage affiliated with small
specimens ranging from micrometer to nanometer. These details enhance
the understanding of small scale fatigue damage mechanisms. The scale of
investigation is also expanded by several orders of magnitude to include slip
mismatch due to differences in grain size, orientation, micro-texture, etc
which arise because of the irregular propagation of short cracks. The fatigue
damage map method (FDMM) was used to quantify the results which are
compared with multiscale test data. In addition the effects of strain rate,
temperature and hold time on the high temperature on the low cycle fatigue
of austenitic and ferritic steels and Nimonic PE16 superalloy are also
examined to study their microstructural changes and crack initiation and
propagation behavior.
Modeling crack growth rate data under simple and complex load spectra
differing from the conventional approaches invoking crack closure or
constraint effects are attempted. Micro to macro range of crack sizes in a
variety of metals are used to correlate data on a straight line referred to a
linear scale covering the regions I, II and III with two empirical parameters.
A large number of problems are considered. They include the 1969
Lockheed F-111 wing fatigue test; the F/A-18 final test program; the round
robin helicopter test program; cracking in composite patched panels; USAF
C-141 repair; and NASA centre notch panel tests.
Analytical dual scale model is presented to describe the transitory
behavior of micro-/macro-cracking for the fatigue of 7075-T6 and 2024-T3
aluminum pre-cracked panels. Accounted for are small cracks that can close
in compression and open in tension as in the case of alternating fatigue
loading. Interactive effects of load, geometry and material are considered by
using three independent micro/macro parameters to render the fatigue
growth rate data form invariant in a log-log plot. The straight line
correlation enables the connection of data referred to micro- and
macro-cracking in fatigue. The same approach was applied to analyze the
fatigue life of cables and steel wires of the Rungyang cable-stayed bridge
based on the design data with and without traffic. The wide variations of
predicted life for the 52 cables can be problematic for the future
maintenance of the bridge as some cables may need early replacement while
others can still be used.
Improved fatigue damage tolerance behavior of aerospace high purity
2024 aluminum alloy is discussed in connection with load sequence and
spectrum effects. Of equal concern for aerospace application is
thermo-mechanical fatigue in the presence of oxidation and creep strain
Foreword ix
accumulation. Virtual testing methodology is explored for fatigue crack

growth associated with the safe-life assessment of aerospace structural
components. Fatigue data of butt welded joints of aluminium alloys for
6061-T6 and 6082-T6 coupon specimens are found and needed to quantify the
influence of three different welding techniques. They are the metal inert gas
welding, laser beam welding and friction stir welding.
It should be said the consideration of length alone is not enough for
distinguishing micro- from macro-defect since the change of defect
morphology may take place during scale transition. Creation of
macro-fracture free surface is caused predominantly by mechanical load
giving rise to symmetric or skew-symmetric geometric pattern whereas
asymmetry typifies the creation of micro-fracture that are obstructed by the
highly irregular material microstructure. Microcracks tend to wonder by
curving and branching because of the inherent different constraint of the
adjoining crack surface. Micro-photos of fatigue striations suggest that the
very tip of a microcrack always remains open on account of geometric
asymmetry. This additional tip distance suggests a micro-stress singularity
that can differ from the macro-stress singularity. It so happen that a
micro-/macro-crack can possess a double singularity: one weak and one
strong. The dual singularity gave allowance to scale transition where a
microcrack can turn to a macrocrack or vice versa. This is precisely what
occurs in tension/compression alternating fatigue loading. Micro- and
macro-cracking correspond, respectively, to crack opening (tension) and
closing (compression). Reference can be made to macromechanical and
microstructural to accentuate the influence of scale with cause.
Nanochemical would refer to imperfection size resulting from chemical
reaction.
The cause of fatigue can be traced to the loss of electrons from metal
atoms in chemical reaction. Such a process in the presence of repeated
mechanical loading even for small amplitude can greatly reduce the life of
metals. The destruction can involve oxidation that increases the
susceptibility of the local region to nanocrack initiation. The grain boundary
is a likely location where the atoms are more loosely packed. A simple
approach would be to use a mean characteristic length to describe a nano
defect with the corresponding energy density weighed by a scale factor
related to inhomogeneity of the atoms or the atomic lattices. The same can
be done at the micro and macro scale level. When needed scale divisions
can be further refined by introducing meso regions to fill the gaps from one
scale range to another. Cause-and-effect studies of material damage in
fatigue require a sufficiently long stretch of scaling such that nano data can
be brought up to the macroscopic scale. Use can be made of the
overwhelming amount of micrograph data presented in this volume. They
x Foreword
demonstrate the different fatigue crack growth details due to strain rate,
temperature, loading type, material, processing technique and other factors.
In this way the practitioner can benefit. Although the foregoing task is still
far from being accomplished, it is essential to start focusing attention in that
direction.
No work on fatigue crack growth can possibly claim to be complete. The
interpretation of the same data can vary as the physical models will differ.
The field of fatigue damage being as important as it is to material and
structure application, the entailing course of future development is
inextricably linked to the social, political, economic and ideological
predilection of the establishment. The limited attempt made here can only
represent the views of the contributors to whom gratitude is due. Special
thanks go to Ren Huifang who design the book cover and to Shen Shufang
who help to format the manuscripts.
Shanghai, China, 2007

G. C. Sih
Contributors
Barter S
Air Vehicles Division, Defence Science and Technology Organisation,
506 Lorimer Street, Fishermans Bend 3207, Australia
Simon.Barter@defence.gov.au
de Castro PMST
Faculdade de Engenharia da Universidade do Porto Rua Dr. Roberto Frias,
200-465 Porto Portugal
ptcastro@fe.up.pt
Dini D
Department of Mechanical Enginineering, Imperial College London South,
Kensington Campus, London SW7 2AZ, UK
d.dini@imperial.ac.uk
Farahmand B
The Boeing Company, IDS, 5301 Bolsa Ave, Huntington Beach,
CA 92647, USA
Bob.Farahmand@Boeing.com
Höppel HW
Department of Materials Science and Engineering, Universität
Erlangen-Nuernberg, Martensstrasse 5, D-91058 Erlangen, Germany
hoeppel@ww.uni-erlangen.de
Jones R
Cooperative Research Centre for Integrated Engineering Asset Management
(CIEAM), Department of Mechanical Engineering, Monash University, P.O.
Box 31, Monash University, Victoria, 3800, Australia
Rhys.Jones@eng.monash.edu.au
Kermanidis AlTh
Department of Mechanical and Industrial Engineering, University of
Thessaly, 38334 Volos, Greece
akermanidis@mie.uth.gr
xi
xii Contributors
Korsunsky AK
Department of Engineering Science, University of Oxford, Parks Road,
Oxford OX13PJ, UK
Alexander.korsunsky@eng.ox.ac.uk
Mannan SL
National Engineering College, Kovilpatti, TN, India
sardariBmannan@yahoo.com
Molent L
Lorrie.Molent@dsto.defence.gov.au
Moreira PMGP
Faculdade de Engenharia da Universidade do Porto Rua Dr. Roberto Frias,
4200-465 Porto, Portugal
pmgpm@fe.up.pt
Mughrabi H
Department of Materials Science and Engineering, Universität
Erlangen-Nuernberg, Martensstrasse 5, D-91058 Erlangen, Germany
Hael.Mughrabi@ww.uni-erlangen.de
Pantelakis SpG
Laboratory of Technology and Strength of Materials, Department of
Mechanical Engineering and Aeronautics, University of Patras, Greece
pantelak@mech.upatras.gr
Peng D
DSTO Centre of Expertise for Structural Mechanics, Department of
Mechanical Engineering, Monash University, P.O. Box 31, Monash
University, Victoria, 3800, Australia
Daren.Peng@eng.monash.edu.au
Pitt S
506 Lorimer Street, Fishermans Bend, Victoria 3207, Australia
monsafe@bigpond.net.au
Richter-Trummer V
Combustion Systems Engineering, Rolls-Royce plc, P.O. Box 31,
Derby DE24 8BJ, UK
valentine@fe.up.pt
Contributors xiii
Rodopoulos CA
Laboratory of Technology and Strength of Materials, Department of
Mechanical Engineering and Aeronautics, University of Patras,
Panepistimioupolis, Patras 26500, Greece
Rodopoul@mech.upatras,gr
Sih GC
International Center for Sustainability, Accountability and Eco-
Affordability of Large Structures (ICSAELS) Advanced Technology for
Large Structural Systems (ATLSS), Lehigh University, Bethlehem, PA
18015, USA; and Department of Mechanical Engineering and Mechanics,
Lehigh University, Bethlehem, PA 18015, USA
gcs8866@yahoo.com
Tang XS
School of Bridge and Structural Engineering, Changsha University
of Science and Technology, Changsha, Hunan 410076, China
Tang-xuesong231@sohu.com
Valsan M
Indira Gandhi Centre for Atomic Research, Kalpakkam, TN, India
Walsh MJ
Derby DE24 8BJ, UK
Michael.Walsh@Rolls-Royce.com
Wang ZG
Shenyang National Laboratory for Materials Science, Institute of Metal
Research, Chinese Academy of Sciences, 72 Wenhua Road, Shenyang
110016, China
zhgwang@imr.ac.cn
Zhang GP
110016, China
gpzhang@imr.ac.cn
Application of Virtual Testing for Obtaining Fracture
Allowable of Aerospace and Aircraft Materials
B. Farahmand
The Boeing Company, IDS
5301 Bolsa Ave, Huntington Beach, CA 92647, USA
Email: bob.farahmand@boeing.com
Abstract
Fracture toughness and fatigue crack growth rate data are two key
parameters which are necessary for conducting safe life analysis of fracture
critical parts used in space and aircraft structures. Currently, these
allowables are obtained through the ASTM testing standards which are
costly and time consuming. In many occasions, due to budget limitations
and deadlines set forth by the customer, it is not possible to conduct fracture
related tests in time. A proposed numerical approach has been developed
[1-3] that is based on the extended Griffith theory and can predict fracture
allowables for a variety of alloys. The simplicity of the concept is based on
the use of basic, and in most cases available, uniaxial full stress-strain data
to derive material fracture toughness values. Because the fracture toughness
is thickness dependent, its value for several thicknesses can be calculated
and used in the residual strength capability analysis. More importantly, its
value is used to predict the region III of the fatigue crack growth rate curve.
Regions I & II of the da/dN versus 'K curve can be estimated separately and
will be connected to region III to establish the total fatigue crack growth rate
data. As the result of this work two computer codes, fracture toughness
determination (FTD) and fatigue crack growth (FCG), were generated under
the NASA contract. Results of fracture toughness and fatigue crack growth
rate data calculated by this approach were compared with numerous test
data in the NASGRO database [4]. Excellent agreements between analyses
and test data were found, which will validate the FTD and FCG
methodology. This novel approach is referred to as the virtual testing
technique. It enables engineers to generate fracture allowable analytically
by eliminating unnecessary tests. Yet, there is another innovative approach
in the virtual testing arena that relies on the multiscale modeling and
simulation technique. This technique is becoming popular in the field of
computational materials, where failure mechanism is defined from the
bottom up approach. The methodology is based on the ab-initio concept
where it is assumed that the failure of material will initiate from the
atomistic level, where nanocracks under the applied load will grow (as the
G.C. Sih (ed.), Multiscale Fatigue Crack Initiation and Propagation of Engineering Materials:
Structural Integrity and Microstructural Worthiness, 1–22.
© Springer Science+Business Media B.V. 2008
2 B. Farahmand
result of poor bonding) and advance toward the micro, meso, and macro
level, which thereafter will cause total structural failure. It should be noted
that in most cases microscopic cracks already pre-existed in the part during
the machining operation, assembling parts, or rough handling. Section 6 will
briefly review the application of multiscale modeling and simulation
technique and its application on aerospace and aircraft parts.
1. Introduction & Background
The requirements of all fracture control plans for space and aircraft
structures dictate that fracture critical components must have adequate life
during their service operation. The safe-life assessment of these components
requires information on material allowables such as: the plane stress (Kc),
plane strain (KIc), part-through fracture toughness (KIe), and the fatigue
crack growth rate properties. Fracture tests under the ASTM testing
standard procedures require detailed specimen preparation, pre-fatiguing
the notch, fatigue crack growth rate measurements and interpretation of raw
data, which are all costly and time consuming. Furthermore, in order for
valid material property data to be produced for each case-specific material
batch, a substantial number of tests will typically need to be performed.
Therefore, any method that can improve and optimize this process, and
reduce the number of experimental tests, will help reduce costs without
compromising any safety issues. The proposed analytical technique uses
fundamental fracture-mechanics-based reasoning to establish a link
between plastic damage development in uniaxial stress-strain data and the
material’s fracture mechanics allowables. Even though the methodology is
working for a variety of aerospace alloys, further validation for a wider
range of materials may be required. In this work, the concept uses steel,
aluminum and titanium based alloys from a pedigree database. Furthermore,
it quantifies material property sensitivity to the predictions for KIc and Kc
and the subsequent estimation of 'Kth threshold and the Paris constants, C
and n values. The prediction of KIc and Kc is based on the energy balance
approach. It has been shown previously that material residual strength
capability curve (a plot of fracture stress versus half a crack length) can be
generated through the Griffith theory for an elastic media. This has been
extended to account for the presence of plastic deformation at the crack tip
[1, 3]. The energy absorption rate at the crack tip process zone could be
assumed to have a similar mode of deformation of uniaxial specimens.
Hence the material stress-strain curve of a uniaxial specimen can reflect the
local energy dissipation profile within the process zone and can describe the
fracture behavior of a cracked specimen as long as the stress and strain
distribution and the subsequent energy balance is satisfied. Similar ideas of
Virtual Testing for Fracture of Aerospace and Aircraft Materials 3
a process zone have been previously considered at high temperatures [5]

whereby creep cracking within a process zone can be described by the
failure behavior of uniaxial creep test. In this proposed model, the failure
range in the process zone is due to the elastic/plastic loading in a uniaxial
tensile test specimen, whereas in the creep case the criteria for the damage
process is a time dependant elastic/plastic/creep phenomenon. However, it
has been shown that creep and plasticity can be treated in a self-similar
manner using non-linear fracture mechanics concepts.
In a tensile case, the total energy per unit thickness absorbed in plastic
straining around the crack tip, UP, can be written as the sum of two energy
terms (UP = UF + UU) where UF and UU are the energy absorbed per unit
thickness in plastic straining of the material beyond the ultimate at the crack
tip and below the ultimate stress near the crack tip, respectively. The rate of
energy absorbed at the crack tip in terms of available and consumed energy
terms, UE, US, UF and UU, can be rewritten by w [UE - US - UF - UU]/wc = 0.
Determining the rate at which energy is absorbed in crack tip plastic
deformation (wUF/wc and wUU/wc), the residual strength capability equation
for ductile material can be established. The quantity wUS/wc is the energy
rate for creating two new crack surfaces. Having fracture stress and half
critical crack length on hand, the critical value of the stress intensity factor
(fracture toughness, Kc) can be calculated. More importantly, the variation
of material fracture toughness, Kc, as a function of part thickness, can be
established, which is extremely useful when conducting a safe-life
assessment of fracture critical components of aircraft and space hardware
where parts thickness can vary significantly. Section 2 will include the
detail description of the extended Griffith theory.
Fracture toughness values calculated for a variety of aerospace alloys
were in excellent agreement with test data which will be discussed in section
3. Moreover, this technique was further extended to predict the fatigue
properties of aerospace metallic material under a cyclic loading
environment. Fatigue crack growth rate tests under the ASTM E645 testing
standard procedures are costly and time consuming. In the proposed
technique, each region of the da/dN curve will be established separately and
then connected to obtain the total fatigue crack growth rate data. It uses the
predicted Kc and KIc values to derive the Paris constants, C & n. Section 4
includes a detailed explanation related to this topic.
2. Detail Description of the Extended Griffith Theory
This theory assumes the fracture characteristics of a metal, local to the crack
tip, are directly related to its ability to deform. Crack tip straining is
assumed to be of two kinds: local strainability at the crack tip (the region of
4 B. Farahmand
highly plastic deformation) and uniform strainability near the crack tip.
Thus, fracture behavior can be characterized by two energy released terms
representing the absorbed energy at and near the crack tip. Both terms can
be shown to be determinable from the full uniaxial stress-strain curve. The
two deformation regions are illustrated in Fig. 1. The total energy per unit
thickness absorbed in plastic straining of the material around the crack tip,
UP, can be written as:
U P = U F + UU (1)
where UF and UU are the energy absorbed per unit thickness in plastic
straining of the material beyond the ultimate at the crack tip and below the
ultimate stress near the crack tip, respectively. Eq. (1) can be rewritten as:
∂[UE − UF − UU ]/∂c = 0 (2)
where wUF/wc and wUU/wc are the rates at which energy is absorbed in plastic
straining beyond the ultimate stress at the crack tip and below the ultimate
stress near the crack tip, respectively. To obtain a relationship between the
applied stress and critical crack length, the energy absorption rates, wUF/wc
and wUU/wc, for the two plastic regions formed around the crack tip must be
determined. Sections 2.1 and 2.2 will discuss the derivation of wUF/wc and
wUU/wc terms, respectively. Applied stress and crack length relationship will
be discussed in section 2.3.
Fig. 1 The crack tip plastic zone and different region of the stress-strain curve.
2.1 Determination of wUF/wc term

To obtain the energy absorption rate for the highly strained region at the
crack tip, the extended Griffith equation in terms of its components can be
rewritten as:
πσ 2 c ∂UF ∂UU
= 2T + + (3)
E ∂c ∂c
where 2T=wUS/wc =E̓ED is the surface tension of the material, that is, the
work done in breaking the atomic bonds, D is the atomic spacing in
angstroms, E is the material modulus of elasticity, and E ҏis the correction
factor (3.94x10-10 inch/angstrom). The quantity wUF/wc in Eq. (3), is set
equal to WFhF, where WF is equal to the unrecoverable energy density
(energy per unit volume) represented by the area under the plastic uniaxial
engineering stress-strain curve (from the stress at which necking begins to
the stress at the fracture point) for an uncracked tensile specimen. For a
given alloy, the area under the stress-strain curve from the stress at which
necking begins to the stress at the fracture point is a constant for a wide
range of thicknesses, and therefore, the value of WF is a constant. It is
assumed that essentially all of the energy represented by WF is absorbed at
the crack tip in a single dominant coarse slip band. The quantity hF is the
size of the highly strained portion of the small region at the crack tip as a
result of WF. Its minimum value is considered to be equal to the effective
o
height of a coarse slip band oriented to make an angle of 45 with the plane
of the crack under monotonic load. At room temperature the quantity hF =
0.00056 in. (approximately 10 micrometers). An empirical relationship
based on test data [1,3] has been developed for tough metals with large
necking strains that gives a higher hF value and better correlation with the
test data than the value hF = 0.00056 in. It is assumed that the quantity hF is
directly related to the energy per unit volume, WF, for the material in
consideration, and was formulated as:
hF = γ (E 2 /σU3 )WF (4)
where J =Ҟ (8/S)Dhmin, VU is the material ultimate allowable stress,

hmin=0.00056 in. and D ҏis material atomic spacing. The atomic spacing, D,
for Aluminum, Titanium alloys, and Iron and Nickel are 2.86 and 2.48
Angstrom, respectively. The quantity WF in Eq. (4) is the area under the
6 B. Farahmand
stress strain curve from necking up to fracture (see Fig. 1) and is

approximately equal to:
WF = σ̄U F εP N (5)
where V UF is the neck stress and its value is taken at the centroid of the
plastic energy bounded on the top by the stress-strain curve from the
beginning of necking to fracture and on the bottom by a straight line from
the beginning of necking to fracture. For material with a negligible amount
of necking the neck stress VUF = (VUlt + VF)/2. The plastic straining, HPN,
(Eq. 5) at the onset of necking is fully discussed in reference [1]. Eq. (3) in
terms of hF, VUF , and HPN can be expressed as:
πσ c ∂UU
= 2T + hF σ̄U F εP N + (6)
E ∂c
Fig. 2 The quantity hF is the effective height of the highly strained portion of the
small region at the crack tip as a result of WF.
2.2 Determination of wUU/wc term
The quantity w UU can be expressed as WUhU, where WU is the

wc
unrecoverable energy density represented by the area under the plastic
uniaxial stress-strain curve from the elastic limit stress, VL, to the ultimate
allowable stress, VU, for an un-cracked specimen. Eq. (6) becomes:
πσ 2 c
= 2T + hF σ̄U F εP N + WU hU (7)
E
The effective height of the volume in which WU is absorbed in the uniform

plastic region at the crack tip is hU. Steps taken to derive the energy rate
term, w UU , for the plane stress condition are as follows:
wc
x Obtain the true crack tip stresses (VTr,̓ VTT, and VTrT) in terms of
applied stress and crack length in polar coordinates by using the
linear elastic fracture mechanics. These stresses are responsible for
material deformation at the crack tip region.
x Use the octahedral shear stress theory [1, 3] to correlate the biaxial
tension stress with the true uniaxial tension stress, VT.
x Apply the crack tip uniaxial tension strain, VT, in the energy
absorption rate equation, to obtain the w UU term (or WUhU).
wc
The term w UU in its final form is given by:
wc
−

∂UU n+1 εF T εT L εT U
=h WU ∗ −1 β (8)
∂c n−1 εT U εT εT U
where h=0.00056 inch. The unrecoverable energy density WU is:

n σT L n+1
WU = εT P U σT U 1 − (9)
n+1 σT U
2.3 Applied stress and crack length relationship
Eq. (3) describing the extended Griffith theory in terms of crack length, c:
E
c= [2T + WF hF + WU hU ] (10)
πσ 2
For a thin plate the value of E=1.3 describing the thickness correction for
the plane stress conditions. Expressing half a crack length, c, in terms of its
components, WUhU and WFhF, for wide plate becomes:
8 B. Farahmand

E n σT n+1
c= 2T + σ̄U F εP N hF k + h ∗ σT U εT U 1 −
πσ 2 µ n−1 σT U
(11)
εT F ε T L εT U n−1/n
∗ ∗ −1 β
εT U εT εT L
where n is the strain hardening coefficient. For a finite width condition the
width correction factor Y is included:

E n σT U εT U σT n+1
c= 2T + σ̄U F εP N hF k + h ∗ 1−
πσ 2 µY 2 n−1 Y6 σT U

εT F εT L εT U n−1/n (12)
∗ ∗ −1 β
εT U εT εT L
πc
−1
where Y = cos . Other terms in Eq. (12) are related to true stress
W
and strain and are defined as:
c Half crack length

n Strain hardening coeficient
V Applied stress
V T True stress
V TU True stress at ultimate
V UF Stress between ultimate & fracture
H T True strain
H PN Plastic strain at necking
H TU , H TL , H TL True strain at ultimate, limit & fracture
k , P , E Thickness corrections
Parameters P, k and E are factors that are used to account for the thickness
correction and they are applied to both sides of the components of Eq. (12)
where energy is available and consumed for plastic deformation. Fig. 3
describes the shear lip formation in the plane-stress fracture, whereas a
plane-strain fracture is characterized by a flat fracture surface. It is
assumed that the energies released and absorbed in plane-stress and
plane-strain are in proportion to these thicknesses. Let t0 be the maximum

thickness for plane-stress fracture. The thickness corrections k & P are
applied to two strained regions at the crack tip. These quantities, thickness
parameters are expressed in terms of maximum thickness for plane stress
fracture, t0 (the amount of shear lip), total thickness, t, (Fig. 3) and Poisson’s
ratio Q.

t0 t − t0 t0 t − t0 t0 t − t0
k= + 2 ,µ= + (1 − ν )
2
, β = 0.13 + 0.12
t t t t t t
Fig. 3 The shear lip formation in the plane-stress fracture. The plane-strain fracture
is characterized by a flat fracture surface.
3. Fracture Toughness Calculation
To obtain the fracture toughness value of aerospace alloys through the

virtual testing approach (by applying the proposed technique), the equation
of the stress intensity factor of a center crack specimen can be used,
K=YVSc, where half crack length, c, can be calculated via Eq. (12). It
should be noted that a full stress-strain curve for the material under
investigation must be available as shown in Fig. 1. To validate the result of
this technique, test data from the NASGRO material database [4] were
selected and the variations of Kc versus part thickness were plotted for two
aluminum alloys using the FTD software [6], Fig. 4. The effect of plate
width and crack length on fracture toughness (narrow and wide plates) was
compared with the data from the NASGRO database [4].
It should be noted that the NASGRO curve for a given material is not
based on numerous tests at different thicknesses through which a smooth
curve passes. Rather, it is based on a relationship between Kc and thickness
10 B. Farahmand
observed from multiple materials (Fig 5). This relationship is available via
an empirical equation proposed by [7] as a function of KIc and material yield
value, which represent the lower bound of Kc values for a given part
thickness. Fig. 5 contains numerous test data which umbrellas the whole
range of plate width and crack lengths, a. The estimated empirical
NASGRO curve fit represents the lower bound and in some cases the typical
values of Kc. The upper bound value of data shown in Fig. 5 must be
associated with larger cracks and wider plates. For this reason the FTD has
the option of plotting the fracture toughness versus thickness variations for
plates of different width. The narrow plate represents the fracture toughness
associated with small cracks where the residual strength can be as high as
75% of the material yield value. Good agreement between the physical
testing and analysis can be seen when analytical data were compared with
typical values of test data.
(a) 2219-T87 (b) 2014-T6

Fig. 4 Example of Fracture toughness versus plate thickness for two aluminum
alloys compared to NASGRO data [4]).
Fig. 5 Measured Fracture Toughness versus thickness for several crack lengths of
2219-T87 aluminum [1]).
4. Generating Fatigue Crack Growth Rate Curve without Testing
4.1 Establishing region I (fracture toughness)
The testing procedure in the ASTM E647 is currently used to obtain fatigue
crack growth rate data through physical testing. The proposed technique
identifies the relevant parameters affecting the micro-mechanical behavior
within a plastic process zone and relates it to plastic damage in uniaxial
tensile tests. It then uses the data from the uniaxial stress/strain test to derive
the materials’ fracture toughness values via the extended Griffith theory.
The fracture toughness value is used to establish region III of the da/dN
versus 'K curve (the accelerated region). The total curve will then be
plotted by estimating the other two regions of the curve (Fig. 6). Using data
derived from the analysis of a wide range of tensile data it has been
established [8] that a relationship exists between Kth ('Kth for R=0) and the
plane strain fracture toughness for aluminum, KIc, giving D = KIc/'Kth.
4.2 Establishing region II (Paris region)
To estimate the Paris region, the two quantities Kc & Kth must be available.
These two quantities will be helpful to establish two points in the Paris
region. The two points in Fig. 6 (the lower and upper points of the Paris
region corresponding to points 1 and 2) have unique properties which are
common among many aluminum alloys. The lower point in the steady crack
growth rate region, just before getting into the threshold zone of the fatigue
curves (point 1 of Fig. 6), has a material independent property. That is, the
ratio of the stress intensity factor, K, at the mean lower bound point and the
threshold value, Kth, ('K/'Kth for R=0) is ~1.125 for the crack growth rate
per cycle, da/dN ~2.54E-6 mm/cycle (~1.0E-7 in./cycle). In the upper
region of the da/dN curve (at the end of the steady crack growth Paris
region, point 2 of Fig. 6), the ratio of the upper bound stress intensity factor,
K, and its critical value, Kc, ('K/Kc for R=0) is found to be ~0.9 for the
da/dN ~0.127 mm/cycle (~0.005 in./cycle) from the data. These two unique
points in the Paris region were found when plotting numerous fatigue crack
growth rate data for aluminum as shown in Fig. 7. Therefore, by having the
two quantities Kc and Kth available, the two points in the Paris region can be
12 B. Farahmand
Fig. 6 Three regions of the da/dN curve (threshold, Paris and accelerated regions).
Fig. 7 Fatigue crack growth rate curve for several aluminum alloys where the two
points in the Paris region are unique.
C(1 − f ) n K n 1 − Kth p
da K
=
q (13)
dN (1 − R)n 1 − (1−R)Kc
K
generated. Hence, the C and n of fatigue crack growth rate constants shown
in Eq. (13) (region II) can be determined [4]. The above assumptions used
for establishing the Paris region are also applicable to Titanium and Steel
alloys.
4.3 Establishing region III (threshold stress intensity factor, Kth)
The methodology used to estimate the threshold stress intensity factor

analytically is based on studying the threshold values of more than one
hundred tests from different metallic alloys and relating it to several key
static and fracture allowables. The variation of Kth values for R=0 for each
alloy were compared with the variation of other parameters such as material
yield, Fty, final elongation, Hf, and the plane strain fracture toughness, KIc.
The quantity Kth was found to be related to the material plane strain fracture
toughness for aluminum alloys, where KIc.= DKth. In the latter portion of this
section we will also assess the relationship between the Kth and final
elongation, Hf, for aluminum, titanium and steel alloys separately. Fig. 8
shows the NASGRO value of Kth for several aluminum, titanium and steel
alloys. Fig. 9 shows the plot of D=KIc/Kth for the data points shown in Fig. 8.
From Figs 8 and 9 it can be seen that the NASGRO value of Kth is material
dependent and has varying ranges for aluminum, titanium and steel alloys.
From the experimental Kth values in Fig. 8, mean values of 'Kth=2.5
MPa¥m for Aluminum alloys, 'Kth=4 MPa¥m for the Titanium alloys and
'Kth=5 MPa¥m for steels are found. The scatter for the steels are seen to be
highest because of different types of steel with wide range of tensile
properties and the scatter for Aluminum is seen to be the lowest.
Looking at the stress-strain curves for several aluminums, the final
elongation for most aluminums fall between 8 and 12% with KIc values
range from 20 to 27 MPam. This is one reason for aluminums alloys
having the minimum amount of scatter as indicated in Fig. 8 when compared
with the steel. The range of D for aluminums and titanium falls
approximately between 7-14 and 9-16, respectively (see Fig. 9). The range
of D for variety of steels is also shown in Fig. 9 and their lower and upper
values are 5 and 25, respectively.
14 B. Farahmand
Fig. 8 Threshold stress intensity factor values for more than one hundred alloys.
Fig. 9 The plot of D=KIc/Kth for the data points shown in Fig. 8.
To understand the physical relationship that may exist between tensile and
fracture behavior, correlations between them has been established for the
different alloys. Fig. 10 shows the relationship between tensile final
elongation, Hf, and Kth for the data of different alloys. Fig. 10 suggests that
the tensile elongation could be an important factor in understanding the
fracture properties in these alloys and improving the predictions for fatigue
and fracture. Using the best fit equations shown in the Fig. 10, it would
possible to get the Kth using available tensile ductility for use in the FCG
code [6] to calculate da/dN data. The best fit equations for Kth give:
Kt h = −0.6337 + 0.3148εf (14)

Fig. 10 Best fit curve for the variation of Kth versus elongation, Hf .
Eq. (14) gives a very good indication of a linear relationship between Kth
and Hf. This will allow FCG to use the estimated Kth from Eq. (14) to
generate the total da/dN curve. The total predicted fatigue crack growth
data, covering all the regions of the da/dN curve, can then be plotted using
Eq. (13), where the fracture parameters and constants are taken from the
estimated Kc, Kth, and the Paris constants C and n values. For any other
range of R-ratios the Newman closure equation, f, [4] can be used to
establish the full fatigue crack growth rate curve when Rz0. In all cases the
constants p and q of equation 13 were taken as 0.5 and 1, respectively.
5. Comparison of Predicted Bounds with Experimental Fatigue Crack

Growth Data
Established and then compared with fatigue crack growth test data in
NASGRO database. The computer program, [6] which can run the
simulations of the model, and the World Wide Web were made available in
order to verify a number of test cases. The input of data can be treated as
deterministic data or as probabilistic bounds of the data. Based on the
above-mentioned assumptions (construction of accelerated, Paris, and
threshold regions), the fatigue crack growth curves for three alloys are e.
The use of the statistical method gives further confidence to the
methodology and is therefore crucial to any sensitivity analysis that would
be needed in design and life estimation methods. In obtaining material
properties data through physical or virtual testing, it is always expected
some level of scatter on fracture toughness and fatigue crack growth values.
The scatter is due to material variations through heat lots when the material
16 B. Farahmand
is processed. A lesser amount of variation can also be observed in test

coupons that have been machined from a given plate of a given
manufacturer by a specified heat lot. For comparison with test data, the
upper and lower bounds for alpha, D, for each of the alloy sets were taken
from Fig. 9 and compared with test data available in NASGRO. Figs. 11-13
show the predicted bounds of fatigue crack growth curves for the Al, Ti and
the steel family of alloys. In all cases the experimental data fall between the
two limits of Hf ҏvalue. Numerous data comparisons between analyses and
test data are available and are reported in [8].
Fig. 11 FCG and NASGRO da/dN curves for 6061-T62 aluminum.
For life assessment analysis of structural components it will be

appropriate to use the lower bound da/dN data corresponding to the upper
bound D values. The upper bound values of D (lower Kth value) can provide
a conservative life assessment. It may be possible to avoid unnecessary
conservatism by tuning the threshold value to the material ductility, using
the correlations shown by Eq. (10), in order to obtain a better agreement
with the test data.
6. Fracture Mechanics & Multiscale Modeling & Simulation
The virtual testing methodology discussed in the previous sections of this

chapter was based on the continuum mechanics approach, where the linear
Fig. 12 FCG and NASGRO da/dN curves for Ti-2.5 Cu Titanium.
Fig. 13 FCG and NASGRO da/dN curves for Inconel 706 Steel.
elastic fracture mechanics were used to describe the crack tip behavior.
Furthermore, fracture mechanics allowables were obtained analytically
through the extended Griffith theory, which can enable engineers to
generate fracture related data without conducting ASTM testing. Both the
fracture toughness and fatigue crack growth rate data can be generated via
the FTD & FCG computer codes and can be used in life assessments of
18 B. Farahmand
fracture critical parts. It should be noted that all the assumptions used in the
derivation of Kc and the da/dN data were based on large crack behavior.
However, in reality the initial cracks embedded in material prior to their
service usage are much smaller in length and the laws of continuum
mechanics may not apply in order to assess their behavior. The largest flaw
size that escapes inspection after machining parts, prior to their service
usage, depends on the capability of NDI technique. If the NDI inspection
shows the surface of the part is free from cracks, the analyst will use the
standard NDI flaw size as the initial flaw in their damage tolerant analysis
[4]. The standard NDI initial flaw size is different for different NDI method
and it is related to the capability of the inspection technique. The life
assessment analysis results based on this approach are too conservative and
in some cases lead to redesigning the part and adding unnecessary weight
into the structure. Experimental and analytical work showed that the typical
flaw size found on the surface of the part after machining operation and
prior to service operation is much smaller than the standard NDI flaw size,
typically of the order of a few microns in dimensions. Clearly, the safe-life
analysis for this crack will result in a much larger number of cycle-
to-failure when fatigue crack growth analysis is performed.
To assess microscopic crack growth when subjected to a load varying
environment is challenging, and the conventional linear elastic fracture
mechanics based on large crack methodology is not applicable for
predicting fracture allowables. The virtual testing methodology for crack
growth behavior must be assessed by the multiscale modeling and
simulation approach. Recently, industries are transitioning from
conventional metallic alloys to super light high strength material such as the
next generation of high-performance structural and multifunctional
composites reinforced with nanoparticles. Weight has been and always will
be the driver for aircraft and aerospace structure designs. Depending on
the application, the cost savings associated with the weight reduction of
structural parts is estimated to range from $100/lb to $10,000/lb.
Aerospace and other industries have made significant investments in
nanomaterials in an attempt to modify proven material systems to have
superior mechanical-/thermal/electrical properties. Successes with
nanocomposite materials have been seen in limited applications across the
composites sporting goods industry or with lower-grade polymers.
Incorporating nanoparticles into existing material systems is proving to be
more difficult that one would anticipate.
Carbon nanotubes have many remarkable stand-alone physical
characteristics such as novel electronic properties, exceptionally high axial
strengths, and outstanding axial Young’s moduli (~an order of magnitude
above metallic alloys). However, once incorporated into a
composite-materials system, these benefits, in some cases, are negated if

bonding between the resin and nanotubes are weak. Proper dispersion and
functionalization (establishing strong bonds between Carbon nanotubes and
polymer) of nanotubes in the polymer matrix can enhance the mechanical
and fracture toughness of nanocomposites. In most cases the failure
initiation in nanocomposites is at the interface area, where a lack of bonds
between the polymer matrix and nanotubes are weak. Fig. 14 shows the
nanocrack initiation at the interface that will contribute to the material
degradation and premature failure of the part when subjected to a service
load.
Fig. 14 Weak interface bonds between nanotubes and polymer can be the source of
damage initiation.
The multiscale modeling of resin and carbon nanotubes can be conducted

through the coarse-graining technique to assess the localized regions of Fig.
14. A beads-and-springs concept will be used to represent the polymer and
the nanotubes within the matrix, which can reduce the computational time
and scaling considerably vs. the use of explicitly atomistic models [9-21].
As shown in Fig. 15, the ~200+ atoms will be replaced by a few
beads-and-springs, where each bead represents several atoms of polymer
and nanotubes, including the interface interactions. Nano-cracks are
incorporated into the modeling by allowing them to have various sizes,
where their length can be as small as a few nanometers and as large as a few
microns (mimicking the width and length of nanotubes).
6.1 Brief description of the coarse grain technique
Multiscale modeling simulation technique will be implemented to assess the

global properties of the nanocomposites through the coarse grain technique
[9-19]. The Materials Studio of Accelrys software [19] is a powerful tool
that can be used to bridge the length scale describing the interface bonds
20 B. Farahmand
Fig. 15 The coarse grain technique approach linking atomistic model to the
airframe parts.
behavior between nanotubes and the resin to micro, meso and the airframe
parts. The scale linking is possible via implementation of several powerful
modules in the Accelrys as well as the Materials Visualizer tool for
modeling the polymer of choice. A few examples of Material Studio
modules are Amorphous Cell Construction, Blends, DMole, Dissipative
Particle Dynamics (DPD), Discover, and Synthia. For example, the Synthia
calculates polymer properties using advanced Quantitative
Structure-Property Relationships (QSPRs). It allows researchers to rapidly
screen candidate polymers for a wide range of properties, and allows the
property prediction of copolymer blends. Or, the DPD module will help
analyst to obtain the global properties of nanocomposite through the coarse
grain technique by establishing super cells containing several molecules of a
given polymer.
7. Conclusion
The proposed analytical approach (also referred to as the virtual testing) can
provide a useful tool for engineers to derive fracture toughness and fatigue
crack growth data for classical metal alloys used in the aerospace industry
when only a few or no test data are available. The methodology is based on
continuum mechanics and basic material properties. It was shown that
fracture properties can be predicted by the energy under the uniaxial tensile
test via the extended Griffith theory. Two computer codes (FTD & FCG)
have been developed that can estimate material fracture toughness and
fatigue crack growth rate data. Excellent agreements between analytical
results and test data were found for variety of alloys used in aerospace. In
addition to the deterministic calculations in the program a probabilistic
module has been developed to allow the user to reach a better understanding
of the effect of scatter on material reliability, and life prediction. It has been
found that material plane strain and stress fracture toughness are sensitive to
material variations observed in the full stress-strain curve. Moreover,
material fatigue crack growth curve is sensitive to parameters that
contribute to the threshold, Paris, and accelerated regions. The proposed
method takes into account both these effects in the predictions for fracture
toughness and fatigue crack growth rates.
A multi-scale computational modeling approach that will dramatically
reduce computational time was also discussed. This approach outlines a
methodology to overcome one of the primary challenges associated with
hierarchical modeling of materials: namely, the accurate prediction of
physical/chemical properties and behavior from the nanoscale to the
macroscale without a loss of intrinsic structural information. Due to large
computational-power requirement, the computational modeling of large
atomistic systems is not possible and the coarse-grain modeling technique
for computational material study can be used. This technique will
dramatically reduce computational complexity where computation times
can be reduced by several orders of magnitude.
The proposed computational method will use the coarse-grain simulation
technique to bridge these length scales and to address the mechanical
behavior of NanoComposite materials based fully or in part on the
fundamental laws of physics (classic and quantum). This will enable the
material scientist to alter key variables, via modeling, by varying parameters
such as bond capability, NanoTube length, nanotube strength, and
single-walled NanoTube versus multiwalled tubes, etc. before attempting to
conduct laboratory tests, which can be costly and time consuming. The
impact of each the above elements on global properties can be addressed
through a reliable software tool that can be applied to fully assess all the
regions of the time and length scale.
To remain competitive in the world market, it is imperative that the U.S.
aerospace community invest in fundamental technologies and novel
applications. The increased reliance on validated simulation methods
linked to rapid, interdisciplinary design is a clear indicator that a
Computational Materials program is required in order to take advantage of
the multifunctional benefits of nano-structured materials in aerospace
structures.
References
[1] Farahmand B. The fracture mechanics of ductile metals theory. In: Fatigue &
fracture mechanics of high risk parts. Chapman & Hall, New York, (1997)
289-322.
22 B. Farahmand
[2] Farahmand B. The fracture mechanics of ductile metals theory. In: Fracture
mechanics of metals, composites, welds, and bolted joints. Kluwer Academic
Publisher, Boston, (2000)237-277.
[3] Farahmand B. Predicting Fracture and fatigue crack growth properties using
tensile properties. J Eng Frac Mech, 2007, Accepted for publication.
[4] JSC, SRI, ESA, FAA. Fatigue crack growth computer program. NASGRO 4.0,
2000.
[5] Nikbin KM, Smith DJ and Webster GA. Prediction of creep crack growth from
uniaxial data. Proceedings of the Royal Society of London, 1984.
[6] NASA/LaRC Contract # NAS-01067. Virtual Testing. NASA Langley
Research Center, Alpha STAR Corporation and Boeing Aerospace Company,
2000.
[7] Vroman GA. Material thickness effect on critical stress intensity factor.
Monograph #16, TRW Space and Technology Group, 1983.
[8] Farahmand B, Xie D and Abdi F. Estimation of fatigue and fracture allowables
for metallic materials under cyclic loading. AIAA Conference, 2007.
[9] Baschnagel J, Binder K and Doruker P. Bridging the gap between atomistic
and coarse-grained models of polymers. Adv. Polym. Sci., (2000) 41-156.
[10] Carmesin I and Kremer K. Monte Carlo and molecular dynamics simulations
of polymers. Macromolecules, 21 (1988) 2819.
[11] Paul W, Binder K, Kremer K, Heermann D. Dynamics of polymer solutions
and melts. Reptation predictions and scaling of relaxation times. J of Chemical
Phys, (1991) 7726-7740.
[12] Baschnagel J, Binder K, Paul W, Laso, Suter UW and Batoulis I. On the
construction of coarse-grained models for linear flexible polymer chains. J.
Chem. Phys, (1991) 6014-6025.
[13] Doruker P and Mattice W. A second generation of mapping/reverse mapping of
coarse-grained and fully atomistic models of polymer melts Macromolecules.
Institute of Polymer Science, 1997.
[14] Groot, R and Warren P. Dissipative particle dynamics: Bridging the gap
between atomistic and mesoscopic simulation. J. Chem. Phys., (1997)
4443-4435.
[15] Tartakovsky A M and Meakin P. A smoothed particle hydrodynamics model
for miscible flow in three-dimensional fractures and the two-dimensional
Rayleigh-Taylor instability. J. Comput. Phys., (2005) 610-624.
[16] Groot RD andRabone KL. Mesoscopic simulation of cell membrane damage,
morphology change and rupture by nonionic surfactants. J. of Biophys., (2001)
725-736.
[17] Reith D, Meyer H and Muller F. Deriving effective mesoscale potentials from
atomistic simulations. Max-Planck-Institut fur Polymerforschung,
Ackermannweg,( 2003) 1625-1636 .
[18] Gates TS, Odegard GM, Frankland JV and Clancy TC. Multi-scale modeling
and simulation of nanostructured materials. Composites Science and
Technology, (2005) 2416-2434.
[19] Accelrys . Materials studio release notes, Release 4.1, Accelrys Software, Inc.,
San Diego, 2006.
An Equivalent Block Approach to Crack Growth
R. Jones1,2*, S. Pitt3 and D. Peng2
1
Cooperative Research Centre for Integrated Engineering Asset
Management (CIEAM), Department of Mechanical Engineering, Monash
University, P.O. Box 31, Monash University, Victoria, 3800, Australia
2
DSTO Centre of Expertise for Structural Mechanics, Department of
Mechanical Engineering, Monash University, P.O. Box 31, Monash
University, Victoria, 3800, Australia.
3
Department of Mechanical Engineering, Monash University, P.O. Box 31,
Monash University, Victoria, 3800, Australia
*Email: Rhys.Jones@eng.monash.edu.au
Abstract
This work is concerned with the recent developments in the formulation

and application of the generalised Frost-Dugdale crack growth law and il-
lustrates how it can be used to predict crack growth, from near micron size
initial flaws, under variable amplitude load spectra.
1. Introduction
The problems of global warming, which led to the Koyoto agreement, and
the potential depletion of oil reserves have focused attention on the need for
even greater fuel efficiency. This has led Boeing to design their latest civil
aircraft the B-787 Dreamliner to achieve a 20% reduction in fuel. In con-
trast in 2006 the USAF used ~$6 billion worth of jet fuel, and as a result is
now actively seeking ways to reduce its carbon footprint. In military and
civil aircraft the majority of the heavy primary structural members are
dominantly metallic. Thus a means for lightening these life limiting metal-
lic parts is essential. However, all new aircraft design and structural
changes made to in-service aircraft require a damage tolerance analysis, as
outlined in the US Joint Services Structural Guidelines JSSG-2006. The
purpose of this requirement is to ensure that any cracks present in the struc-
ture will not cause loss of the structure for some predetermined period of
in-service operation. Thus the challenge is to reduce weight without com-
promising structural integrity. This requires tools that can accurately esti-
mate the entire fatigue crack growth history under complex variable ampli-
tude load spectra. To meet this challenge it is necessary to have the
capability to:
24 R. Jones et al.
(1) Rapidly and accurately assess the durability, and therefore the safety, of
existing and new designs and structural modifications under both current
and future operational environments; and
(2) To allow a reduction in weight without compromising safety or durabil-
ity.
One possible approach is to use an equivalent block method whereby crack
growth data either generated from the measured load spectra, or one man-
dated by the certifying authority, is used to develop an expression for the
average crack growth per load block. The development of one such equiva-
lent block method, which is based on the generalised Frost-Dugdale law
[1,6], for predicting fatigue crack growth under variable amplitude loading
is a focal point of this Chapter. The ability of this law to model cracks
growing under constant amplitude loading was studied in [1-6], and its
applicability to predict crack growth in both specimens and in F-111 and
F/A-18 aircraft full-scale fatigue tests was illustrated in [1, 6].
A background to the prognosis tools discussed in this Chapter is presented-
prior to assessing crack growth under variable amplitude loading. At this
stage it should be mentioned that most current approaches are based on the
concept of similitude. This Chapter reveals that this hypothesis is invalid in
Region I.
2. Crack Growth
The recent papers [1-6] have shown that the generalized Frost-Dugdale law
can be applied to a wide class of engineering problems and that there is of-
ten a near linear relationship between the log of the crack size and the life,
i.e. the number of cycles, or blocks. These observations are reinforced by
the work presented in [7] which studied the growth of physically small
cracks, which ranged in size from 20 microns to several mm’s in length, in
an austenitic-ferritic 2205 duplex stainless steel and a range of other mate-
rials and concluded that over these length scales da/dN was proportional to
the crack length and that the crack growth history could be written as per
Frost & Dugdale, in the form:
a = ai eYN (1)
where Y is a load and geometry dependent parameter, and ai is the initial

flaw size. A near linear relationship between ln(a) and N has also been re-
ported, for physically small cracks, by a large number of researchers [8-14].
This conclusion, i.e. a near linear relationship between ln(a) and N, was
also one of the primary findings reported in an early USAF/General Dy-
namics [15] study into small crack growth undertaken as part of the F-16
research and development program. The observation that da/dN was often
proportional to the crack length led to the conjecture [1-6] that in the low to
mid range 'K region, i.e. in Region I and part way into Region II, the in-
crement in the crack length per cycle (da/dN) could be described by the
generalised Frost-Dugdale law, viz:
da/dN = C c ( a/a*)(1-J/2) ('N)J- (da/dN)0

= C* a(1-J/2) ('N)J - (da/dN)0 (2)
where 'N, is the crack driving force, C c , a*, and J are constants, C*
= C c /a*(1-J/2) and (da/dN)0 reflects both the fatigue threshold and the nature
of the notch (defect/discontinuity) from which cracking initiates. Note that
this law has da/dN dependent on both the crack driving force and the crack
length.
For the special case of a small crack in a large centre cracked panel under
remote constant amplitude loading with R ~ 0.0 equation (2) implies that
da/dN is proportional to the quantity a'VJ. This relationship agrees with
that derived by Tomkins [13] who assumed that crack growth was propor-
tional to the accumulated plastic strain, which was averaged over a charac-
teristic length ahead of the crack length.
Equation (2), which is a generalized form of the Frost-Dugdale law in that
J is not assumed to be 3 as in [6], also resembles the fractal based crack
growth law proposed in [16-18], and has da/dN proportional to both 'VJ
and the crack length a as per [7].
A similar equation was developed in [19] which also found that the in-
crement in the crack length per cycle (da/dN) was a function of both the
stress intensity factor K and the crack length. A similar crack growth law
was also proposed in [20]. Here they considered a crack to have a series of
dislocations ahead of the crack tip, which they termed a super dislocation,
and that crack growth occurred as a result of the formation of micro-cracks
ahead of the crack tip. These assumptions led to a crack growth equation of
the form:
da/dN = F 'V(1/O-1) ('K -'Kth)2/(K2Ic-'K2) (3)
where F, and O are material constants and KIc is the plane strain Mode I
fracture toughness. It should be noted that for small cracks 'K is propor-
tional to 'V Sa so that when 'K << KIc , and O = 1/(J -1) then equations
(2) and (3) have a similar form. However, the question of what to use for
26 R. Jones et al.
'V when the stress field is complex means that equation (2) has greater
generality than equation (3).
In this context it should be noted that [21], which presented an analytical
solution for a flaw growing with a non sharp notch, also found that the si-
militude hypothesis does not hold in Region I. Indeed, for the range of
problems considered in [20] their non-similitude growth law had the same
form for flaws that ranged from near microns to approximately 10 mm’s.
These findings together with the realisation, presented in [22] that “in the
threshold regime, there is something missing either in the (crack closure)
model” and the statement in [23] that “a small crack may not be closed for
as much of the load cycle as a larger crack”, i.e. that crack closure effects
are less significant for small cracks, led reference [4], to question the valid-
ity of the similitude hypothesis, that underpins most current crack growth
models, in Region I. (In this chaper the hypothesis of similitude is taken to
be as first proposed by [24], viz: that for constant amplitude loading the in-
crement in crack length per cycle (da/dN) takes the form:
da/dN =f(Kmax, R) (4)
where f(Kmax, R) is only a function of Kmax, the maximum value of the

stress intensity factor in the cycle, and R (= Kmin/Kmax), where Kmin is the
minimum value of the stress intensity factor in the cycle, and where 'K =
Kmax - Kmin= Kmax (1-R). This means that: Two different cracks growing in
identical materials with the same stress intensity factor range 'K, and the
same Kmax, will grow at the same rate.
As remarked in [1, 6] a hypothesis, in this case similitude i.e. equation
(4), can be disproved by a single counter example. A number of counter ex-
amples were presented in [1]. One example presented involved the work of
[25] that presented crack growth for a Grade 1 austempered ductile iron ob-
tained using a compact test specimen as per the ASTM guideline1 (E647-
95a). Figure 1 presents a conventional da/dN versus Ҟ'K plot of the data
whilst Figure 2 presents an analysis of the data associated with Region I,
and also the lower portion of Region II, which reveals a (near) linear rela-
tionship between da/dN and 'K3/a1/2 as per Equation (1) with J = 3. As
such the da/dN was a function of both 'K and the crack length a, and hence
similitude does not follow.
1Information on the crack length, load da/dN, and K at each point in this test is
contained in the spreadsheet associated with Figure 1.
Fig.1 Crack Growth in L6B-2, from [25].
(a) First set data

28 R. Jones et al.
(b) Second set data

Fig. 2 Crack growth in Region I in L6B-2, adapted from [25].
Crack growth data for a Grade 1 austempered ductile iron obtained using
a compact test specimen as per the ASTM guidelines (E647-95a) was also
analysed in [1], see Figures 3 and 4. (As in the previous example informa-
tion on the crack length, load da/dN, and K at each point in this test is con-
tained in the spreadsheets associated with Figures 3 and 4.) The test speci-
mens were supplied by ADI Engineering, Process and Heat Treatment Pty.
Ltd. and had a chemical composition in %wt of; 2.50C, 0.28Mn, 2.6Si and
less than 0.1P. Analysis of the data associated with Region I again reveals a
(near) linear relationship between da/dN and 'K3/a1/2, see Figure 4a). As
such the crack growth rate is a function of both 'K and the crack length a
and similitude does not hold. In this case it was found that J = 3, C* = 9.42
10-10, and (da/dN)0 = -2.76 10-6. It is again interesting to note that, as shown
in Figure 4 b), this relationship extends into Region II.
In each of the examples presented the experimental results reveal that in
Region I and into Region II the increment in the crack length per cycle
da/dN has the form:
da/dN = 'KJ a1-J/2 – (da/dN)o (5)
As such these counter examples disprove the similitude hypothesis in Re-

gion I. A more extensive collection of counter examples is given in [1, 6].
Fig. 3 Crack Growth in an austempered ductile iron, from Khan and Chen [25].
Fig. 4 Crack growth in Region I in an austempered ductile iron, adapted from Khan
and Chen [25].
The test data presented in [26], which presented crack growth data ob-
tained using the ASTM standard constant stress ratio test method for middle
crack tension, compact tension and eccentrically loaded edge-crack tension
specimens, also revealed that similitude does not hold in Region I. In this
study it was found that the increment in the crack length per cycle (da/dN)
30 R. Jones et al.
was not a unique function of 'K and R, see Figure 5. Thus since for con-
stant R ratio tests Kmax = 'K/(1-R) this means that the similitude hypothesis
first proposed in [24] was invalid.
Fig. 5 Non similitude crack growth in D6ac steel, from Forth et al. [26].
The crack growth data presented above, which cover several quite differ-
ent rail industry steels as well an aircraft industry steel (D6ac), confirm that
in Region I, and also in the lower portion of Region II, da/dN is a function
of both 'K, and Kmax and the crack length a. As such these various counter
examples reveal that the similitude hypothesis inherent in the the majority
of crack growth models is invalid in Region I. In contrast the examples pre-
sented above and those presented in [1, 5, 6, 27 ] show that in Region I, and
also in the lower portion of Region II, crack growth follows the generalised
Frost-Dugdale crack growth law, which as shown in [1] accurately predicts
crack growth from near micron size flaws.
3. An Equivalent Block Method for Predicting Fatigue Crack Growth
Many practical engineering problems, i.e. cracking in rail and aircraft struc-
tures, involve complex load spectra that can be approximated by a number
of repeating load blocks. In certain circumstances these repeated blocks of
loads can be treated as equivalent to load cycles. To this end let us consider
the case of block loading, where each block consists of a spectrum with n
cycles that have peak stresses of Vi, i =1… n, with the associated cyclic
ranges being ¨Vi, i =1… n. Also assume that:
i) The slope of the a versus block curve has a minimal number of discon-
tinuities.
ii) There are a large number of blocks before failure.
With these assumptions reference [28] built on equation (2) to derive the
following expression for the crack growth per block, da/dB, viz:
da/dB = C*KmaxJ a1-J/2 - da/dB0 (6)

~
where C is a spectra dependent constant, and Kmax is the maximum value
of the stress intensity factor in the block. However, this formulation can be
derived from a number of assumed forms for the crack driving force. To
this end let us assume that crack growth per cycle can be expressed as:
da/dN = C*a(1-J/2) ( f(R) ¨K ) J (7)
where J is a constant, and f(R) is a function of the R ratio. Setting f(R) = 1

yields a non-similitude variant of the Paris growth law, whilst setting:
f(R) = 1/(1-R)p (8)
where p is a constant, yields a non-similitude variant of the Walker crack

growth law [29], and setting yields a non-similitude variant of Elber’s crack
closure law [30].
f(R) = (0.5+0.4R) (9)
If there are a large number of blocks before failure then, to a first ap-
proximation, the crack length a can be considered to be a constant in any
given block, and hence E can also be assumed to be a constant within that
block, and that 'Ki can be expressed as:
'Ki = E 'Vi(Sa) (10)
With these assumptions the crack growth per block, da/dB, can be written
as
n
da/dB = ¦
i 1
C* a(1-J/2) (f(Ri) ¨Ki) J (11)
32 R. Jones et al.
which on substituting for 'Ki as given by equation (10) can be expressed

as:
n
da/dB= ¦i 1
C*a(1-J/2) (f(Ri) E 'ViS1/2a1/2) J
n
= (ES1/2) J a ¦ i 1
C* (f(Ri) 'Vi) J (12)
where Ri is the R ratio associated with the i’th cycle.Equation (12) can be
written in a number of forms. One such form is:
n
da/dB= (Vmax ES1/2) J a ¦i 1
C* (f(Ri) 'Vi/ Vmax) J
n
=[ ¦ i 1
C* (f(Ri) 'Vi/ Vmax) J] KmaxJ/aJ/2-1
KmaxJ/aJ/2-1
=C (13)
where
= [C*(f(Ri) 'Vi /Vmax) J ]

C (14)
and
Kmax = Vmax E (S a) (15)
where Vmax is the maximum stress seen in the block, and Kmax is the corre-
sponding is the maximum value of the stress intensity factor seen in the
block.
This formulation represents a non-similitude variant of that proposed in
[31-34]. Researchers at Boeing (Seattle) [35] also developed a related (non-
similitude) approach whereby instead of equation (13) da/dB was expressed
as:
da/dB = C (K/g(a/t))m (16)
where the function g(a/t), which is a function of ratio of the crack length (a)
to the thickness (t) of the specimen, was experimentally determined. This
formulation was necessary to enable the predictions to match the measured
crack length histories, which as shown in [5] followed the generalised

Frost-Dugdale crack growth law.
At this stage it is important to note that:
i) As shown in [2] for cracks with very small initial sizes, in some cases
the initial flaw sizes were of the order of several microns, growing under
variable amplitude loading there was a near linear relationship between the
average increment in the crack length per block and the crack length that
held up till (near) failure. This implies that any crack growth equation
should apply to Regions I, II, and III.
ii) When analysing crack growth in the F/A-18 wing attachment centre
barrel fatigue test [1] the best results were obtained when the generalised
Frost-Dugdale law was used well into Region II.
iii) The review paper [36] on crack growth and similitude concluded that
similitude was lost during fatigue crack growth under variable amplitude
loading.
iv) Quantitative fractography [37, 38] has shown that the mechanisms un-
derpinning crack growth under constant and variable amplitude loading dif-
fer.
As a result, although under constant amplitude loading the generalised
Frost-Dugdale law is only strictly applicable to Region I and to the lower
portion of Region II, this “equivalent block” variant, i.e. Equation (13), can
be extended to have a form consistent with both Regions I, II and III, viz:
da/dB = (a1- J/2 KmaxJ - da/dBo)/(1.0 - Kmax/Kc) (17)

~
where a is now the average crack length in the block, C is a spectra de-
pendent constant, and Kc is the apparent cyclic fracture toughness. Here, as
described in [1, 6], the term da/dBo reflects the both nature of the disconti-
nuity from which the crack initiates and the apparent fatigue threshold for
this particular block loading spectra. However, it should be noted that this
variant of the generalized Frost-Dugdale law is only applicable to crack
growth data where the slope of the a versus Block curve has minimal dis-
continuities.
4. Examples
Case 1) Consider the crack growth data presented in [39] for Al7010-
T73651 compact tension specimens tested under variants of the Rotarix
helicopter load spectrum. These specimens had a thickness of 17.5mm and
a width of 70mm. Four different variants of the Rotarix spectrum (termed
34 R. Jones et al.
Rot 16, 20, 24, 32) were tested, see [39]. The Rotarix spectrum used was
representative of that experienced by a typical rotary wing structure. It had
a large number of small cycles at high R-values. The baseline Rotarix spec-
trum represented 140 flights, which corresponded to 190.5 hours of flight,
for more details see [39]. In these spectra cycles were progressively re-
moved from the original baseline Rotarix spectrum to produce different
spectra, see [39]. The Rot 16 spectrum was the largest spectrum and con-
tained 99.4% of the original spectrum. The Rot 20 and 24 spectra were very
similar and only the Rot 16, Rot 20 and the Rot 32 spectra were analysed.
The relationship between the maximum stress intensity factor in the block
Kmax and the crack length is given in Figure 7.
Fig. 6 Crack growth rate for the Rotarix 20 spectra, adapted from [39].
The resultant experimental and the computed crack length histories, ob-
tained using Equation (15) with the value of J fixed at three and the values
~
of C and Kc as given in Table 1, are shown in Figure 8 where we again see
good agreement between the computed and the measured crack length his-
tory. (In these problems the left hand side of Equation (15) is da/dCycles.)
As in the previous examples information on the crack length, load da/dN,
and K at each point in this test is contained in the associated spreadsheet.
Fig. 7 Stress intensity factor versus crack length from [39].
Fig. 8 Experimental and computed crack length history under Rotarix spectra, from
[39].
Table 1 Values of C and Kc for specimens tested under a Rotarix spectra.

Spectrum C x 10-8 (fitted) Kc
Rot 16 .01 30
Rot 20 .50 30
Rot 32 1.0 30
36 R. Jones et al.
Fig. 9 Crack growth rate under gust load spectra.
Case 2) Crack growth data for growth from a 20 mm diameter hole in a

2mm thick 7075-T6 aluminium alloy sheet tested under a Fokker-F27 gust
spectra was presented in [40]. This work presented a plot of the crack
growth (m/flight) against Krms, the root mean square value of K seen in a
block, which shows a typical sigmoidal shape, see Figure 9. However, as
shown in Figure 10 the average increment in the crack length per flight con-
forms to equation (17). As in the previous examples information on the
crack length, load da/dN, and K at each point in this test is contained in the
associated spreadsheets.
5. Fatigue Crack Growth Analysis of a High-Frequency Loaded

Helicopter Coupon – The AHS Round-Robin Problem
The objective of the Helicopter round-robin study outlined in [41] was to

assess the capability of current structural analytical tools to predict fatigue
crack growth. The round-robin test specimen was chosen to represent a
complex aerospace component, subjected to a typical rotorcraft spectrum
(Asterix). The geometry was a flanged plate with a central lightning hole,
see Figure 11. The study required the participants to predict crack growth
from a 2 mm (radius) corner defect on the inner edge of the large central
hole to the failure, where failure was defined as a projected surface crack
length of 25 mm. The crack growth path is shown in Figure 12.
Fig. 10 Crack growth rate correlation to the Frost-Dugdale law.
Fig. 11 Round-Robin crack configuration [41].

38 R. Jones et al.
Fig. 12 Schematic diagram of the round-robin crack growth evolution (all dimen-
sions are in mm).
The crack was found to grow for a considerable part of its life as a part
through crack. The crack subsequently transitioned from the 6 mm rein-
forced hole-edge into the 2 mm thick web, and later into the larger thick-
ness as it approached a flange. The component was made from 7010-
T73651 aluminum alloy, and as in [41, 42], was assumed to have a Youngs
modulus of 70,000 MPa, a Poisson’s ratio of 0.3, and a fracture toughness
of 34 MPa m.
The Asterix load sequence used in this study was derived from strain data
measured on a helicopter lift frame, see [42]. The maximum and minimum
stress points in the spectrum were given as 100 and -4 respectively. The As-
terix spectrum represented 190.5 flights, or 140 sorties, and was consisted
of 371,610 cycles. The largest stress in the spectrum was to 130MPa. A text
file named “Helicopter Round Robin Spectra” containing this spectra is
available with this chapter.
5.1 Finite element analysis
A 3D finite element model of the round-robin component was first created.

Symmetry considerations mean that it was only necessary to model half of
the structure, as shown in Figure 13. The model was hex-meshed, and con-
sisted of 61,962 9-noded CHEXA elements with a total of 268,478 nodes. It
was subjected to a peak (remote) load which corresponded to a net section
stress at the plane of symmetry of 130 MPa. The model was then imported
into the NEi/NASTRAN to compute the resultant stress field. The computa-
tional analysis took approximately 1352 seconds CPU time on a PIV per-
sonal computer running under a Linux operating system.
Fig. 13 Nominal stress distribution at the crack face.
The generalised Frost-Dugdale crack growth law was used [27] to predict
the crack growth history from an initial 2 mm quadrant flaw to a surface
crack length of 25 mm. In the first instance N was taken to be the stress in-
tensity factor range 'K. For the purpose of simplicity in the present analy-
sis, the value of C was determined by using a simple curve fitting tech-
nique, where C was manually adjusted to match the time to grow from a
2 mm to a 4 mm corner crack. This approach yielded a value of C 1.28 u
10-11, and J = 3.
When predicting crack growth from 2 to 5 mm the weight function tech-
nique was used to compute the stress intensity factors from the stress field
obtained using the (uncracked) NEi-Nastran finite element model. How-
ever, in the thinner section the 2D stress intensity factor solution given in
[41, 42] was used, for more details see [27]. The resultant Beta factor solu-
tion used in the analysis is shown in Figure 14 along with that given in the
original proposal [41, 42] and the Beta factor used in [43].
40 R. Jones et al.
Fig. 14 Beta factors used in the analysis.
5.2 Results and discussion
The crack length histories for two test specimens was presented in [42]. The
starting defect in specimen number 1 was a spark machined quarter circular
corner crack of 2 mm radius located at the edge of the hole. In test 2, the
defect radius was 1.5 mm, and the defect was grown as a fatigue crack un-
der the Asterix spectrum up to 2 mm radius and crack growth data were re-
corded from this point up until failure. From Figure 15 it can be seen that
there was little difference in the crack growth histories. Both started at a
moderate growth rate which steadily increased to a fairly rapid rate and was
then followed by a much slower rate of crack growth.
In the round-robin study most of the predictions submitted to the organiz-
ers did not match the test data particularly well, see [42]. However, as can
be seen in Figure 15 the predictions obtained using the generalised Frost-
Dugdale law were in good agreement with the measured data.
This problem was also analysed using the average block variant, viz:
da/dB= a(1-J/2)KmaxJ (18)
with J = 3 and C = 0.92 10-15 and the results are also presented in Figure
15 which again shows an excellent agreement between the predicted and
the measured crack length history. As in the previous example information
on the crack length, load da/dN, and K at each point in this test is contained
in the spreadsheet associated with Figure 15.
Fig. 15 Computed surface crack length history.
It is informative to contrast this solution with the conclusions reached in

[43], who used FASTRAN,viz:
the early stages of crack growth compared very well with the test data.
But, again, the crack-growth rates in the 2-mm thick section were too
fast.
This shortcoming is largely removed in the present study, see Figure 15.
It should be noted that interest in this problem was kindled due to the con-
clusions reached by US Army researchers [44], who used AFGROW and
NASGRO to predict the crack length history for this problem, who stated:
i) “It is evident that the crack growth was under-predicted for the crack
length below 5 mm and is over-predicted when crack length is more than 10
mm.”
ii) “More research; especially in the area of modeling crack growth near
the crack growth threshold and determining experimentally the crack
growth threshold values are needed.”
Indeed, the present crack growth law was specifically formulated to ad-
dress point (ii), and its ability to predict crack growth history for this com-
plex and challenging problem is particularly pleasing. Its ability to predict
crack growth for a range of other complex problems is shown in [1, 27, 45,
46].
42 R. Jones et al.
6. Conclusion
This work has shown that in Region I crack growth data reveals the simili-
tude hypothesis to be invalid. It has summarised recent work on the gener-
alised Frost-Dugdale crack growth law and revealed how it can be used to
predict crack growth under both constant and variable amplitude loading.
The various examples presented show that this approach:
i. is generally consistent with experimental results,

ii. provides a reasonable representation of macroscopic observations of
crack growth in a range of alloys,
iii. can be used to predict crack growth from near micron size initial flaws,
and
iv. has the potential to accurately predict crack growth in real air-
craft structures under variable amplitude load spectra.
References
[1] Jones R., Molent L. and Pitt S. Crack growth of physically small flaws, Inter-
national Journal of Fatigue, 29 (2007) 1658-1667.
[2] Molent L. Jones R. Barter S. and Pitt S., Recent Developments In Fatigue
Crack Growth Assessment, International Journal of Fatigue, 28(12) (2006)
1759-1768.
[3] Barter S., Molent L. Goldsmith N. and Jones R. An experimental evaluation
of fatiguecrack growth, Journal Engineering Failure Analysis, 12(1) (2005)
99-128.
[4] Jones R. Barter S. Molent L. and Pitt S. A multi-scale approach to crack
growth, Multiscaling in molecular and continuum mechanics: biology, elec-
tronics and material science, Edited by G. C. Sih, Springer, October, 2006.
[5] Jones R. Molent L., Pitt S. and Siores E. Recent developments in fatigue
crack growth modelling, Proceedings 16th European Conference on Fracture,
Alexandropoulos, Greece,July 2006.
[6] Jones R. Chen B. and Pitt S. Similitude: cracking in steels, Theoretical and
Applied Fracture Mechanics, 48(2) (2007) 161-168.
[7] Polak J. and Zezulka P. Short crack growth and fatigue life in austenitic-
ferritic duplex stainless steel, Fatigue Fract Engng Mater Struct, 28 (2005)
923–935.
[8] Harkegard G. Denk J. and Stark K. Growth of naturally initiated fatigue
cracks in ferritic gas turbine rotor steels, International Journal of Fatigue,
(2005) 1–12.
[9] Nisitani H. Goto M. and Kawagoishi N. A small-crack growth law and its re-
lated phenomena, Engineering Fracture Mechanics, 41(4) (1992) 499-513.
[10] Kawagoishi N. Chen Q. and Nisitani H. Significance of the small crack

growth law and its practical application, Metallurgical and Materials Transac-
tions A, 31A (2000) 2005 2013.
[11] Caton MJ.,Jones JW. Boileau JM. and Allison, JE. The effect of solidification
rate on the growth of small fatigue cracks in a cast 319-type aluminum alloy,
Metallurgical and Materials Transactions A, 30A (1999) 3055-3068.
[12] Murakamia Y. and Miller K.J. What is fatigue damage? A view point from
the observation of low cycle fatigue process, International Journal of Fatigue,
(2005) 1–15.
[13] Tomkins B. Fatigue crack propagation-an analysis, Phil. Magazine, 18 (1968)
10411066.
[14] Polák J. Interpretation of the coffin-manson law in terms of short crack
growth, Proceedings Fatigue 06, Atlanta, USA, May 2006.
[16] Potter JM. and Yee BGW. Use of small crack data to bring about and quantify
improvements to aircraft structural integrity, Proceedings AGARD Specialists
Meeting on Behavior of Short Cracks in Airframe Structure, Toronto , Canada
, 1982.
[17] Carpinteri A. Scaling laws and renormalization groups for strength and
toughness of disordered materials. Int. J. Solids Struct., 31 (1994) 291–302.
[18] Carpinteri A. and Spagnoli An. A fractal analysis of size effect on fatigue
crack growth, Int. J. Fatigue, 26 (2004) 125–133.
[19] Spagnoli An. Self-similarity and fractals in the Paris range of fatigue crack
growth. Mechanics of Materials, 37 (2005) 519–529.
[20] Dong P. Hong JK. and Cao Z, Stresses and stress intensities at notches:
‘anomalous crack growth’ Revisited, International Journal of Fatigue, 25
(2003) 811–825.
[21] Ding HZ. and Xing XS. A plastic flow-induced fracture theory for fatigue
crack growth, Journal Of Materials Science, 31 (1996) 4099–4103.
[22] Sih GC. and Tang XS. Form-invariant representation of fatigue crack growth
rate enabling linearization of multiscale data, Theoretical and Applied Frac-
ture Mechanics, 47(2007) 1–14.
[23] Newman JC. Jr. Brot A. and Matias C. Crack-growth calculations in 7075-
T7351 aluminum alloy under various load spectra using an improved crack-
closure model, Engineering Fracture Mechanics , 71 (2004) 2347–2363.
[24] Newman JC.Jr. Phillips EP. and Swain MH. Fatigue life prediction methodol-
ogy using small crack theory, International Journal of Fatigue, 21 (1999) 109-
119.
[25] Paris PC. Gomez RE. and Anderson WE. A rational analytic theory of fa-
tigue. The Trend in Engineering, 13(1) (1961) 9-14.
[26] Khan AS. and Chen B.K. Evaluation of austempered ductile iron for applica-
tion to rail cast components, Asia Pacific Industrial Engineering and Man-
agement Systems Conference,Gold Coast, Australia, Dec 2004.
[27] Forth SC. Johnston WM. and Seshadri BR. The effect of the laboratory
specimen on fatigue crack growth rate, Proceedings 16th European Confer-
ence on Fracture, Alexandropoulos, Greece, July, 2006.
44 R. Jones et al.
[28] Tiong U.H., Jones R. and Molent L., Damage Tolerance Analysis of a Heli-
copter Component, Proceedings 1st International Conference on Damage Tol-
erance Assessment of Aircraft Structure, Delft, The Netherlands, September
25th-28th, 2007.
[29] Jones R. and Pitt S., Crack patching: revisited, Composite Structures,
7667(32) (2006) 218-223.
[30] Walker E.K., The effect of stress ratio during crack propagation and fatigue
for 2024-T3 and 7076-T6 aluminum. In: Effect of Environment and Complex
Load History on Fatigue Life, ASTM STP 462, Philadelphia: American Soci-
ety for Testing and Materials, 1970, 1-14.
[31] Elber W., The significance of fatigue crack closure, Damage Tolerance of
Aircraft Structures, ASTM STP-486 (1971), 230-242.
[32] Schijve J., Fatigue crack growth under variable-amplitude loading, Engineer-
ing Fracture Mechanics, Engineering Fracture Mechanics, (1979) 207-221.
[33] Gallagher J.P. and Stalnaker H.D., Developing normalised crack grwth curves
for tracking damage in aircraft, American Institute of Aeronautics and Astro-
nautics, Journal of Aircraft, 15(2) (1978) 114-120.
[34] Miedlar P.C., Berens A.P., Gunderson A. and Gallagher J. P., Analysis and
support initiative for structural technology (ASIST), AFRL-VA-WP-TR-
2003-3002.
[35] Barsom, J.M. and Rolfe, S. T.: Fracture and fatigue control in structures: Ap-
plications of fracture Mechanics, Butterworth-Heinemann Press, November
1999.
[36] Miller M., Luthra V.K. and Goranson U.G.: Fatigue crack growth characteri-
zation of jet transport structures, Proc. of 14th Symposium of the International
Conference on Aeronautical Fatigue (ICAF), Ottawa, Canada, June 10-12,
1987.
[37] Davidson D.L., How Fatigue Cracks Grow, Interact with Microstructure, and
Lose Similitude, Fatigue and Fracture Mechanics: 27th Volume, ASTM STP
1296, R. S. Piascik, J. C. Newman, and N. E. Dowling, Eds.: American Soci-
ety for Testing and Materials, (1997) 287-300.
[38] Molent L., McDonald M., Barter S. and Jones R., Evaluation of spectrum fa-
tigue crack growth using variable amplitude data. International Journal of Fa-
tigue, in press.
[39] White P., Barter S. and Molent L., Observations of crack path changes under
simple variable amplitude loading in AA7050-T7451, Int. Journal of Fatigue,
(In press).
[40] Zitounis V., Fatigue crack growth rates under variable amplitude load spectra
containing tensile underloads, PhD Thesis, Cranfield, 2004.
[41] Wanhill R.J.H., Gust severity effects on fatigue crack propagation in alumin-
ium alloy sheet materials, International Journal of Fatigue, 1(3) (1979) 118-
123.
[42] Cansdale R. and Perrett B., The Helicopter Damage Tolerance Round Robin
Challenge, Workshop on Fatigue Damage of Helicopters, University of Pisa,
September 12-13, 2002.
[43] Irving P.E., Lin J. and Bristow J.W., Damage tolerance in helicopter - Report
on the round robin challenge, presented at American Helicopter Society 59th
Annual Forum, Phoenix, Arizona, 2003.
[44] Newman J.C., Irving P.E., Lin J. and Le D.D., Crack growth predictions in a
complex helicopter component under spectrum loading, Fatigue Fract Engng
Mater Struct, 29 (2006) 949-958.
[45] Vaughan R.E. and Chang J.H., Life prediction for high cycle dynamic com-
ponents using damange tolerance and small threshold cracks, presented at
American Helicopter Society 59th Annual Forum, Phoenix, Arizona, 2003.
[46] Jones R., Peng D. and Tiong U.H., An Equivalent Block Approach To Fa-
tigue Crack Growth, 5th Australasian Congress on Applied Mechanics,
ACAM 2007, 10-12 December 2007, Brisbane, Australia.
[47] Molent L., Jones R. and Pitt S., Understanding crack growth in fuselage lap
joints, Theoretical and Applied Fracture Mechanics, available on line.
Prediction of Fatigue Crack Growth Rates in
Ti-6Al-4V Alloy
A. M. Korsunsky1*, D. Dini2 and M. J. Walsh3

1
Department of Engineering Science, University of Oxford, Parks Road,
Oxford OX1 3PJ, UK
2
Department of Mechanical Engineering, Imperial College, Exhibition
Road, London SW7 2AZ , UK
3
Derby DE24 8BJ, UK
*Email: alexander.korsunsky@eng.ox.ac.uk
Abstract
Crack initiation under thermo-mechanical fatigue is analyzed with consid-

eration given to oxidation and creep strain accumulation. Transitions and
interactions between different mechanisms, crack initiation and crack
propagation rates display significant scaling and size effects. The objective
of this work is to present experimental data for crack initiation and propa-
gation in aerospace alloys, and to offer some means of describing these
data so as to develop improved capabilities for life prediction in aerospace
materials and assemblies.
1. Introduction
Reliable prediction of fatigue crack growth rates in aerospace materials

and components underpins the so-called defect-tolerant approach to lifing.
In this methodology the presence or appearance of defects and cracks in
components is accepted. However, safe operation is guaranteed by regular
inspections and health monitoring, and ensuring (by means of reliable
modeling) that no crack may grow far enough to reach the critical size in
the interval between inspections.
Under such circumstances it is clear that particular attention has to be
paid to the development and validation of predictive modeling capabilities
for fatigue crack propagation. The situation is complicated by the fact that
it is often a challenge to represent correctly the in-service loading experi-
enced by a cracked component. In practice, on top of the major cycles as-
sociated with each flight (LCF component), cycles of higher frequency and
lower amplitude are also present (HCF component). Sensitivity to dwell at
maximum load is also often observed. Furthermore, it is well established
48 A. M. Korsunsky et al.
that complex load sequences involving overloads and underloads result in

fluctuations of fatigue crack growth rates (retardation and acceleration)
that must be accounted for in crack growth calculations.
Damage tolerant design (DTD) has become accepted as a method of fa-
tigue lifing of airframe structures since early 1970’s, when military speci-
fication MIL-A-83444 was introduced in 1971 [0]. In view of the substan-
tial improvements in safe component lives and the associated cost savings,
damage tolerant design soon became the subject of much academic re-
search and industrial development. In 1982 the Advisory Group for Aero-
space Research and Development (AGARD) headquartered in France
asked its Structures and Materials Panel to form sub-committees devoted
to this subject [0]. One of these two sub-committees, SMP/SC.33, was
formed in 1983 to organize a cooperative test programme between NATO
countries with the purposes of (i) promoting familiarity of laboratories
with test techniques for DTD of engine disc materials, (ii) standardizing
test specimen geometries and test techniques, and (iii) calibrating the re-
sults from participating laboratories through a round robin test programme.
The programme focused on the use of typical engine disc materials, Ti-
6Al-4V, IMI 685 and Ti-17, and the use of realistic engine loading se-
quences. The data were intended to provide a basis for the evaluation of
different modeling techniques for the prediction of fatigue crack growth
rates (FCGRs) and component lives. The Core Programme focused on Ti-
6Al-4V using detailed test procedures [0]. The Ti-6Al-4V material for the
round robin test programme was provided by Rolls-Royce from RB211 fan
disc forgings. There were in total 216 tests carried out by 12 different labo-
ratories. Good repeatability of crack growth testing was confirmed, with
potential drop technique proving extremely accurate in measuring crack
sizes. The database of fatigue crack growth data for Ti-6Al-4V material
was deemed to be large enough to be used in the Supplemental Programme
for life prediction modeling.
The predictive models used in the Supplemental Programme all relied on
Elber’s plasticity-induced crack closure concept [0] and the effective stress
intensity factor range 'K eff , except for the Rolls-Royce approach. The lat-
ter approach used the Walker model [0] in the form
da m n
dN
'K (1 R) , (1)
where R denotes the ratio of minimum to maximum stress (stress intensity

factor).
Prediction of Fatigue Crack Growth in Rates in Ti-6Al-4V Alloy 49
2. Experimental Programme and Results
The specimens used in the programme were made from material in the
form of solution treated and aged compressor disc forgings provided by
Rolls-Royce. Typical mechanical properties of Ti-6Al-4V material were:
0.2% yield stress 870 MPa, ultimate tensile strength 970 MPa, elongation
to failure 11%, reduction in area 27%, fracture toughness 70 MPa·m1/2.
Typical grain size of this material was 25Pm [0].
Specimens designation used in the testing programme was composed us-
ing a six letter code, as follows: LA-GE-##, where LA indicates the labo-
ratory where testing was carried out, GE indicates the crack geometry, and
## indicates the specimen test number in a particular series.
Laboratories contributing to the testing programme on Ti-6Al-4V were
the following:
AF Air Force Materials Laboratory (AFML), WPAFB, Dayton, Ohio,

USA
CE Centre d’Essais Aéronautique de Toulouse (CEAT), France
PI University of Pisa, Pisa, Italy
Two specimen geometries were used for fatigue crack propagation testing.
ASTM compact tension specimens, designated CT, were used for the con-
sideration of long crack growth regime where typically the assumption of
two-dimensional through-thickness crack geometry can be used. The sec-
ond specimen geometry, designated CC, corresponded to the corner crack
specimen developed by Rolls-Royce [0] in order to simulate corner crack
flaw geometries subjected to three-dimensional stress fields typically
found in disc bore locations and at fastener holes [0]. Geometry of both
CT and CC specimens are shown in Figure 1. Crack length data were col-
lected using potential drop technique.
Loading sequences considered in the programme included two constant
amplitude sequences characterised by R-ratios (ratio of minimum to
maximum applied load) of R = 0.1 and R = 0.7. The waveform shape was
trapezoidal with the frequency of 0.25 Hz (cycle duration 4 seconds). In
addition to the constant amplitude sequences, further ‘simple sequences’
were used. Three of these were selected to study the effect of minor cycles
on a single major cycle, and designated SS1, SS2 and SS3, respectively.
As illustrated in Figure 2(a), these sequences can be considered as constant
amplitude trapezoidal cycles with a dwell at peak load that is replaced with
10 minor cycles. Minor cycle amplitudes were 10% unload from peak load
for SS1, 30% unload for SS2, and 50% unload for SS3, respectively. For
SS1, 30% unload for SS2, and 50% unload for SS3, respectively. For these
three loading sequences the R-ratio of major cycles was set to zero, and all
loading and unloading rates in major and minor cycles were constant, re-
sulting in cycle duration of 13 seconds for SS1, SS2 and SS3.
(a) Compact tension specimen (b) Corner crack specimen
Fig. 1 Compact tension (CT) and corner crack (CC) specimen geometries [0].
Fig. 2 Illustration of the trapezoidal waveform used for constant amplitude tests
(CST) [0].
Loading sequence SS4 (Figure 2b) contained a single overload with the
ratio of 1.7 after every 1000 constant amplitude triangular cycles. The pur-
pose of this test was to investigate the retardation effect due to overload.
Alongside the ‘simple sequences’ outlined above, complex sequences
were also utilised, namely, the full Cold TURBISTAN sequence [0] and
three modified versions. TURBISTAN loading sequence is a variable am-
plitude loading standard and comprises 100 individual flights with an av-
erage of 154 end points per flight. Four TURBISTAN sequences used con-
tained minor cycle omission levels of 0%, 10%, 30% and 50%,
respectively. The loading sequences resulting from various levels of omis-
sion are illustrated in Figure 4.
Experimental results from the AGARD report [0] used in this study are
illustrated in Figures 3 and 4 as Paris plots of fatigue crack growth rate
(FCGR) vs applied stress intensity factor range for compact tension (CT)
specimens. Figure 3 shows significant divergence between results obtained
using different loading sequences, suggesting that using applied stress in-
tensity factor 'K appl alone does not allow obtaining adequate correlation
for FCGR. Further detailed analysis revels that the data for loading se-
quences SS1, SS2 and constant amplitude cycling (CST) collapse well.
However, significant deviation is observed in the remaining simple loading
sequence data, particularly for sequence SS4 (containing periodic overload
every 1000 cycles), and, to a lesser extent, for sequence SS3 (containing
minor cycles with 50% unloads).
(a-1). Simple loading sequence type 1 (SS1).
(a-2). Simple loading sequence type 2 (SS2)

(a-3). Simple loading sequence type 3 (SS3).
(b) Simple loading sequence type 4 (SS4).

Fig. 3 Simple loading sequence type 1,2,3 and 4 (SS1,SS2,SS3 and SS4) [0]
(a) Original sequence (b) 10% omission level
(c) 30% omission level (d) 50% omission level
Fig. 4 Example of TURBISTAN flight number 1 loading sequence with various

omission levels [0].
Fig. 5 Experimental results for fatigue crack growth rate (FCGR) vs applied stress
intensity factor range for compact tension (CT) specimens used in the AGARD
test series [0]. Simple loading sequences (SS1, SS2, SS3 and SS4) were applied.
Analysis of the results presented in Figure 5 for TURBISTAN loading

sequences with different percentages of omission shows that acceptably
good correlation is achieved, perhaps with the exception of some sensitiv-
ity to loading details for conditions approaching the lower threshold.
Review of the results contained in the AGARD addendum report [0]
draws, amongst others, the conclusions that (i) for simple loading se-
quences, using R=0.1 as baseline, the addition of minor cycles does not in-
fluence fatigue crack growth rate (FCGR) until the magnitude of the minor
cycle reaches 50% (as in SS3) of the major cycle; and that (ii) the overload
sequence, SS4, compared to the baseline R=0.1 data, retards the fatigue
crack growth rate by a factor of eight at 'K =10 MPa·m1/2, and a factor of
four at 'K =30 MPa·m1/2.
Consider the suitability of a recently proposed model of. [0,0] for the
prediction of fatigue crack growth rates (FCGR’s) in Ti-6Al-4V material
using a simplified description of plastic deformation in the vicinity of
crack tip, in combination with the use of the two-parameter crack driving
force. The model is introduced, and its application to the set of AGARD
data is illustrated below.
Fig. 6 Experimental results for fatigue crack growth rate (FCGR) vs applied stress
intensity factor range for compact tension (CT) specimens used in the AGARD
test series [0]. TURBISTAN loading sequences were applied.
3. The Two-Parameter Crack Driving Force
The work in [0, 0] presented a two-parameter approach to fatigue crack

growth prediction in steel, aluminium and titanium alloys. The purpose of
this approach is to account for the influence of stress ratio on fatigue crack
growth through detailed semi-analytical consideration of the development
and effect of residual stress around the tip of the growing fatigue crack. In
all cases the driving force is expressed as a combination of the maximum
stress intensity factor, K max , and the stress intensity range, 'K , where
both parameters are corrected for the presence of residual stress. The pur-
pose of this approach was to correlate fatigue crack growth data obtained
under a wide range of load ratios and fatigue crack growth rates from near-
threshold to high growth rate regimes. The principal advantage of the ap-
proach lies in the fact that no finite element modeling of the near-tip de-
formation is required, since material response is described in terms of
Ramberg-Osgood strain-stress relations, in combination with the Neuber
rule. In [0] the model was successfully verified using fatigue crack growth
data for Ti–6Al–4V titanium alloy with load ratios ranging from -1 to 0.7.
Use is made of the above approach to correlate the data obtained in the
AGARD programme, and discuss the results in view of the suitability of
this method for damage tolerant design calculations.
The starting point for the analysis is the introduction of inherent small,
but finite crack tip blunting radius,̓ U * . The consequence of this assump-
tion is two-fold. On the one hand, under tensile loading it may now be
noted that the crack tip stress-strain field is no longer singular; in fact, in-
stead of the classical stress intensification expressions, the stress distribu-
tion can be described using the formulae given in [0], of the form
K § U* ·
Vy ¨1 ¸ ... (2)
2S x © 2 x ¹
The maximum stress at the crack tip is attained at x U * / 2 and is equal
to
2K
V ymax . (3)
SU *
On the other hand, the presence of finite radius ensures that upon the ap-
plication of compressive stress to the crack its shores cannot immediately
close simultaneously (as might be the case for a crack with an infinitely
sharp tip). Instead, close to the tip the crack would remain open, forming a
cavity that can be approximated as a circle, with the possibility of some
reverse plastic flow due to stress concentration. Interesting aspect of this
simplifying assumption is then the fact that the asymmetry of deformation
response to tensile and compressive applied stresses is explicitly captured
in the micromechanical description. The authors of the Noroozi et al.
analysis note that the origin of such view can be traced back to Neuber’s
concept of ‘microsupport’, that postulates the existence of the smallest ma-
terial-specific length scale (notch radius) that is still perceived to be differ-
ent from a sharp crack. The same material parameter is also used to define
the length dimension of elementary material volumes within which failure
processes occur. It is interesting to observe that this view can also serve as
a basis for developing an explanation for the KT fatigue threshold diagram
[0].
Crack propagation is considered to be due to sequential failure of ele-
mentary material volumes occupying the segments between
xi U * (2i 1) / 2 and xi 1 U * (2i 1) / 2 with respect to the tip centre
position [0]. Average stress in each such volume is computed as follows:
xi 1
1 K § U* · K\ y ,i
V ey ,i ³ ¨1 ¸ dx , (4)
( xi 1 xi ) xi 2S x © 2 x ¹ 2SU *
where therefore
xi 1
2 ª U* º
\ y ,i «
( xi 1 xi ) ¬« x
x U * / 2 » , \ y ,1 1.63299 , etc. (5)
¼» xi
Further development of the micromechanical model of Noroozi et al. [0]

proceeds by taking into account the reverse plastic deformation in the vi-
cinity of the crack tip and its effects on residual stress formation, and
hence on the stress intensity factor modification. Material deformation is
described using a simplified strain energy equivalence procedure, also due
to [0], that allows the prediction of inelastic strains and stresses within
elementary material volumes ahead of the crack tip. Note that this task can
be accomplished without recourse to detailed finite element analysis, al-
though the results obtained in this way show reasonable agreement with
such simulations [0]. Cyclic uniaxial material response can be conven-
iently described by the Ramberg-Osgood additive strain-stress law,
1/ n '
§V · § V ·
H ¨ ¸¨ ¸ . (6)
© E ¹ © K '¹
The two terms represent elastic and plastic strains, respectively. Parame-
ters K’ and n’ can be readily identified by fitting functional relationship in
equation (2) to material’s uniaxial macroscopic cyclic stress-strain curve.
A further aspect of the Noroozi et al approach is the use of the Coffin-
Manson strain-life curve in combination with Smith-Watson-Topper pa-
rameter [0], in the form:
2
'H (V ' f ) 2b bc
V max 2N V ' f H ' f 2N . (7)
2 E
Here N is the number of cycles to failure; V max and 'H refer to the maxi-
mum stress and strain range in the stabilized fatigue cycle, respectively;
V ' f and H ' f are fatigue stress and strain parameters, respectively; and b
and c are strain-life fatigue law exponents.
Analytical predictions for fatigue crack growth rate (FCGR) were ex-
pressed in [0] in the following form:
da
C 'N J
dN (8)
p 1 p
'N K max,tot 'K tot
Here for the case of predominantly elastic behaviour (near the lower fa-
tigue crack growth threshold) the following parameters are used:
pel 0.5; J el 1 b
1 2b
ª \ y ,1 2 º
*« »
C 2U in Plane Stress (9)
«¬ 4SU *V 'f 2 »¼
1 2b
2U * ª \ y ,1 2 º
C « » in Plane Strain
1 Q 2 1 2b «¬ 4SU *V 'f 2 »¼
For the case of predominantly plastic behaviour (i.e. at higher values of
stress intensity factor range) the following parameters are used:
n' 2
p pl ; J pl
n ' 1 bc
§ 1 ·
¨ ¸
C
ª
2 U * « n ' 3
\ y ,1 2 º
»
© bc ¹
in Plane Stress (10)
« * ' '
»
«¬ 2 n '1 SU V f H f E »¼
§ 1 ·
¨ ¸
ª 2 º © bc ¹
C
2U *
«
\ y ,1
» in Plane Strain
1 Q 2 1 2b « 2 nn ''13 SU *V ' H ' E »
¬« »
f f ¼
In order to be able to plot both the predictions obtained assuming pre-
dominantly elastic and predominantly plastic behaviour on the same graph
and compare them to the experimental data, we need to write the fatigue
crack driving force, 'N, for the predominantly elastic behaviour as a func-
tion of the applied stress ratio and the driving force for the predominantly
plastic behaviour. If we neglect the effect of residual stresses on the total
stress intensity factors and assume Rappl0, then
p p
'N el 1 Rappl pl el 'N pl (11)
It should be noted that all results discussed below are plotted with respect
to the fatigue crack force for the predominantly plastic behaviour, i.e.
'N='Npl.
After calculating the residual stress intensity factor, Kr, as described in [0],
it is necessary to modify the applied maximum stress intensity factor,
Kmax,appl and the applied stress intensity factor range, Kappl. This modifica-
tion accounts for the effect of the residual stress on the fatigue crack
growth. The interactions of the stress intensity factor, the plastic zone and
the residual stress manifest themselves through the change (decrease) of
the resultant maximum stress intensity factor, Kmax,tot, without significant
changes in the resultant minimum stress intensity factor, Kmin,tot. It is as-
sumed that the residual stress intensity factor, Kr, contributes mainly to the
change (decrease) of the resultant maximum stress intensity factor, Kmax,tot,
and subsequently to the resultant stress intensity range, 'Ktot. However, the
magnitude of the residual stress effect depends on the applied stress ratio
and it has to be treated differently for positive and negative applied stress
ratios, Rappl.
Figure 7 illustrates several consequences of applying the model to
AGARD fatigue crack growth rate data for simple loading sequences.
Firstly, comparison with the same data plotted in Figure 5 in a conven-
tional form (da/dN vs Kappl) reveals that better collapse of the results to-
wards a master curve is achieved. This is primarily the consequence of tak-
ing into account stress intensity factor modification in fatigue cycles with
the R ratio less of equal to 0.5 (for minor cycles) that arises from the pres-
ence of residual stresses. Secondly, the quality of agreement with the data
achieved by the plastic regime prediction is on the whole satisfactory, with
the slope and location of the trend line capturing the overall location of the
data points. However, the quality of agreement in the predominantly elas-
tic regime leaves room for improvement. Note that the effect of the resid-
ual stresses was neglected in deriving the modified stress intensity driving
force, 'N, for predominately elastic behaviour. Furthermore, in the limited
version of the model used here the collapsing of the data points is per-
formed making the assumption that the specimens are subjected to uniform
loading histories. Therefore, the influence of overload, underload and
mixed LCF/HCF cycling (replicating in-service conditions) are not ade-
quately captured. It is important to note, however, that improved, more de-
tailed models have been proposed that capture the history effect by calcu-
lating the stress-strain response at the crack tip for each consecutive load
reversal. The magnitude and distribution of residual compressive stresses
is then continuously monitored and updated during cycling, and serve as
the basis for predicting the response during the next load reversal. Fur-
thermore, a memory model has been proposed that postulates the influence
of past stress history on the growing fatigue crack.
Fig. 7 Correlation obtained for FCGR’s using two-parameter fatigue crack driving
force for simple loading sequences.
force for TURBISTAN loading sequences.
Figure 8 illustrates the predictions obtained using the predominantly elastic

and plastic regimes for the TURBISTAN data. Good clustering of data can
be noted in this figure, as well as in Figure 6, where the same results are
plotted in a conventional way. However, the quality of predictions shown
in Figure 8 for both predominantly elastic and plastic regimes requires fur-
ther improvement.
4. Discussion
Consider the application of [0] methodology to a substantial set of experi-

mental fatigue crack growth rate (FCGR) data in Ti-6Al-4V alloy. Most of
the loading sequences considered in the experimental study were not con-
stant amplitude sequences, but in fact contained combinations of major and
minor cycles with different values of the R ratio, and also included com-
plex TURBISTAN sequences with different levels of omission.
force for simple loading sequences.
Predictive description of fatigue crack growth rates is most importantly

used for design calculations. It is therefore important to express the growth
rate in the most compact form suitable for easy incorporation into numeri-
cal models. From this point of view, analytical expressions possess distinct
advantages. It is therefore proposed here to identify a functional form for
the dependence of FCGR on the two-parameter fatigue driving force, as

described below.
The fitting ‘knee’ function is sought in the form of a merging between
two distinct power law functions that prevail in the predominantly elastic
and plastic regimes, for the large and small values of 'N respectively. The
form of this function is:
da 'N J el
C fit (12)
dN ª J el J pl º F
«1 § 'N · F »
«¬ ©¨ 'N 0 ¹¸ »¼
where
1
ªª 1 2b º J pl J el
2 º
« « \ y ,1 » 1 »
« « 4SU *V 'f 2 » J
pel p pl º el »
'N 0 «¬ ¼ ¬ª1 R ¼ » (13)
« ¨
§ 1 · »
« 2 ¸ »
ª \ y ,1 º © bc ¹
« « » »
« « n ' 3 * ' ' » »
«¬ n '1 SU V H E »¼
¬« 2 f f ¼ »
1 2b
2U * ª \ y ,1 2 º
C fit « » in Plane Stress
J el
ª¬1 R pel p pl º¼ «¬ 4SU *V 'f 2 »¼
1 2b
2U * ª \ y ,1 2 º
C fit « » in Plane Strain
J el
ª¬1 R pel p pl º¼ 1 Q 2 1 2b «¬ 4SU *V 'f 2 »¼
and Ȥ is a parameter representing the smoothness of the transition between

the predominantly elastic and predominantly plastic behavior. This con-
stant can be adjusted to provide the best match to experimental data. In the
example shown in Figure 9 this parameter has been set to Ȥ=4.5. It is ap-
parent that the use of this function leads to a close agreement between pre-
diction and the experimental data.
5. Conclusions
The original model of [0] for the correlation between fatigue crack growth
rates and the two-parameter fatigue driving force has been applied directly
to the data contained in the AGARD addendum report [0] with minimal
degree of adjustment. Material properties of alloy Ti-6Al-4V were taken

from experimental cyclic dog-bone sample tests or from published litera-
ture [0]. Nevertheless, the quality of agreement achieved in this way was
encouraging.
More reliable prediction of the effects of overloads and underloads pre-
sent within loading sequences on the fatigue crack growth rates requires
more careful analysis of the detailed interaction between local plastic de-
formation and damage at the crack tip. The process of crack advance leads
to a gradual attenuation of the effects of overloads and underloads on fur-
ther growth. The interaction of prior deformation history with subsequent
plastic deformation is particularly intricate, and is probably most effi-
ciently described by some phenomenological relationship [17].
A merging ‘knee’ function has been proposed for the description of tran-
sition between predominantly elastic and plastic regimes, respectively. The
function provides a convenient means of incorporating the correlation in
design calculations.
Table 1 Ti-6Al-4V specimen designations and loading sequences.
SPECIMEN SEQUENCE
AF-CT-07 SS1, R=0.0
AF-CT-08 SSl, R=0.0
AF-CT-09 SS2, R=0.0
AF-CT-10 SS3, R=0.0
AF-CT-11 SS3, R=0.0
AF-CT-12 SS4, R=0.l
AF-CT-13 SS4, R=0.1
AF-CT-14 10% TURBISTAN
AF-CT-15 10%
AF-CT-16 30%
AF-CT-17 30%
AF-CT-18 50%
AF-CT-19 FULL TURBISTAN
CE-CC-07 CST AMP. R=0.7
CE-CC-08 SS1, R=0.0
CE-CC-09 SS2, R=0.0
CE-CC-10 CST AMP. R=0.1
CE-CT-11 CST AMP, R=0.1
CE-CT-12 CST AMP, R=0.7
CE-CT-13 SS1, R=0.0
CE-CC-14 SS3, R=0.0
CE-CC-15 SS3, R=0.0

CE-CC-16 SS4, R=0.1
CE-CT-17 SS4, R=0.1
CE-CT-18 10% TURBISTAN
CE-CT-19 FULL TURBISTAN
CE-CC-20 10%
CE-CC-21 10%
CE-CC-22 30%
CE-CC-23 30%
CE-CC-24 50%
PI-CC-07 10%
PI-CC-08 10%
PI-CC-09 30%
PI-CC-10 30%
PI-CC-11 50%
PI-CC-12 FULL TURBISTAN
PI-CC-13 FULL TURBISTAN
PI-CT-14 10%
PI-CT-15 30%
PI-CT-16 30%
PI-CT-17 50%
PI-CT-18 FULL TURBISTAN
PI-CT-19 FULL TURBISTAN
RR-CT-28 CST AMP, R=0.1
References
[1] Military Specification MIL-A-83444, Airplane Damage Tolerance equire-

ments (U.S.) (1971).
[2] AGARD Conference proceedings 393, Damage tolerance concepts for criti-
cal engine components,(1985).
[3] Mom AJA. Working document for the AGARD cooperative test programme
on titanium alloy engine disc material, NLR TR 84022L (1984).
[4] Elber W. The significance of fatigue crack closure. In: Damage tolerance in
aircraft structures, STP 486, ASTM, Philadelphia, PA (1971) 230–42.
[5] Walker EK. The effect of stress ratio during crack propagation and fatigue
for 2024-T3 and 7076-T6 aluminum. In: Effect of environment and complex
load history on fatigue life. STP 462, ASTM, Philadelphia, PA (1970) 1–14.
[6] AGARD report R-766 (addendum), AGARD engine disc cooperative test
programme, (1993).
[7] Pickard AC. Brown CW. and Hicks MA. The development of advanced
specimen testing and analysis techniques applied to fracture mechanics lifing
of gas turbine components, ASME Intl Conf on Advances in Life Prediction

Methods, Woodford DA. and Whitehead JR. eds., ASME (1983).
[8] A.A. ten Have, Cold TURBISTAN . Final definition of a standardized fa-
tigue test loading sequence for tactical aircraft cold section engine discs,
Combined report of the MTU, IABG, LBF, RWTH Aachen, Rolls-Royce,
RAE, SNECMA, CEAT, ONERA, NLR, Univ. of Utah and NRC. National
Aerospace Laboratory TR87054L, Holland (1987).
[9] Noroozi AH. G. Glinka G. and Lambert S. A two parameter driving force for
fatigue crack growth analysis, International Journal of Fatigue, 27 (2005)
1277–1296.
[10] Noroozi AH. G. Glinka G. and Lambert S. A study of the stress ratio effects
on fatigue crack growth using the unified two-parameter fatigue crack
growth driving force, International Journal of Fatigue 29 (2007) 1616–1633.
[11] Creager M. and Paris PC. Elastic field equations for blunt cracks with refer-
ence to stress corrosion cracking. Int J Fract Mech 3 (1967) 247–252.
[12] Neuber H., Kerbspannugslehre. Berlin: Springer (1985).
[13] Kitagawa H.and Takahashi S. Applicability of fracture mechanics to very
small cracks in the early stage. Second Intl Conf on Mechanical Behaviour
of Materials, Metals Park, American Society for Metals (1976) 627–631.
[14] Neuber H. Theory of stress concentration for shear-strained prismatic bodies
with arbitrary nonlinear stress–strain law, ASME J Appl Mech 28 (1961)
544–551.
[15] Pommier S. A study of the relationship between variable level fatigue crack
growth and the cyclic constitutive behaviour of steel. Int J Fatigue 23 (2001)
111–118.
[16] Smith KN. Watson P. and Topper TH. A stress–strain function for the fa-
tigue of metals, J Mater 5 (1970) 767–778.
[17] Glinka G. private communication, 2007.
Some Practical Implications of Exponential
Crack Growth
L. Molent1*, S. Barter1 and R. Jones2
1
2
Department of Mechanical Engineering, Monash University, P.O. Box 31,
Victoria, 3800, Australia
*Email: Lorrie.Molent@defence.gov.au
Abstract
A review of experimental data show that for many lead fatigue cracks in
service components loaded with service spectra, exponential growth (i.e.
log crack depth versus cycles or hours) applies for the majority of the life.
This behaviour is shown to extend from the micro to macro range of crack
sizes in a variety of metals. As a consequence of this, it will also be shown
that the crack growth rate is directly proportional to the crack depth. By
combining these observations with traditional fracture mechanics ap-
proaches to crack growth modelling, a model that is a function of the stress
intensity factor (K) with a fixed crack depth influence (non-similitude for
the K parameter alone) is proposed. It will then be shown that this model
allows for Region I to be smoothly integrated with Region II of the con-
stant amplitude da/dN data. Further, it will be shown that for variable am-
plitude crack growth data, crack growth ranging from microns to many
millimetres can be modelled using this single model.
This modelling approach is of particular importance in structural integ-
rity analysis where fatigue cracking cannot always be avoided and the ma-
jority of the fatigue life of highly stressed, nominally gross defect free
structure is spent growing physically small cracks from initiating disconti-
nuities (i.e. loads in Region I for constant amplitude loading growth rates)
up to the point of loss in acceptable strength.
1. Introduction
Where fatigue cracking is inevitable, as is the case for many critical struc-
tures such as aircraft, trains and rails, oil rigs and bridges etc., it is gener-
ally accepted that the majority of the fatigue life of metallic components is
governed by the growth of the fatigue cracks while they are small. It is
also true that the cyclic loading leading to fatigue cracking is of variable
amplitude (VA), whereas fatigue life predictions are usually based on con-
66 L. Molent et al.
stant loading cycle data. Generally these cracks start from some form of
discontinuity in the material and over time grow to a size where the com-
ponent can no longer support the applied loading. There has been so much
research into fatigue crack growth (CG) that it is virtually impossible to
summarize. It therefore may seem surprising that a reliable and accurate
means for characterizing the growth of cracks under variable loading (VA)
is still unavailable [0]. One possible explanation for this may be that the
characterisation of so-called small CG, that aligns with Region I on a Paris
type constant amplitude (CA) CG rate da/dN versus stress intensity (K)
range ( 'K ) plot (where a is the crack depth and N is a load cycle) and
Region II (Paris Region) of the same type of plot, is yet to be fully unified.
These Regions are indicated on a typical data set shown in Fig. 1.
Unification of the R ‘ egion I and II’ growth has not yet been fully
achieved, despite numerous models that have been developed to achieve
this goal. These models include the El Haddad model [2] in which a crack
Fig. 1 Typical long crack CA da/dN data for Al7050-T7451 for cracking in the L-
T orientation taken from AFGROW [4]. In terms of long crack data the three Re-
gions of growth are indicated approximately.
size dependency on the threshold stress intensity range, Kthes is added to

the physical crack length, so that 'K is proportional to a a 0 instead of
a ; and the FASTRAN model which adds a fictitious plastic zone size
dependent crack size to the physical crack [3].
A problem for prediction of the fatigue life of a structural component is

that the data used for life calculation are inevitably based on cracking typi-
cally tens of millimetres long, which is used to produce the Paris-type CA
plot as shown in Fig. 1. These data are known as long crack data. In con-
trast, most critical fatigue cracking initiates from small discontinuities in a
material, and these cracks can spend a significant proportion of their lives
at sizes below about 1mm in depth for small discontinuities and high
stresses. These differences can lead to significant differences in the CG
rate calculated using long crack Region 1 data and that measured experi-
mentally for a given 'K. This is the so called small or short crack inconsis-
tency [5]. These inconsistencies occur at low 'K values, and this behav-
iour is sometimes (incorrectly) referred to as anomalous CG [5]. This
variation in the threshold values and CG rates in Region I is a serious
problem when it comes to crack life prediction for cracks that grow from
small initial discontinuity sizes. As stated by others (e.g. [6]), it will be
shown here that it is the short crack or low crack tip 'K regime that gov-
erns a large portion of the life of many fatigue critical components. In this
chapter a simple model will be described and shown to be applicable to the
prediction of CG for small cracks or low crack tip K levels for VA CG
ranging from microns to many millimetres. The model is shown to bridge
both Region I and II for CA loading, or small cracks through to many mil-
limetres regime for VA loading. The model is explicitly dependant on
crack length. The equivalent Region III CG where the crack is approaching
failure under static modes of crack extension is also considered despite this
period usually only representing a small portion of the total life in many
components.
2. Crack Growth Model
The CG model described below is based on observations indicating that

the Regions I and II (or equivalent for VA loading) produce consistent and
regular CG in practical circumstances. An example of this is shown in Fig.
2, where the surface of a fatigue crack that was grown in an ex-service
F/A-18 aircraft component with the mini-FALSTAFF VA loading spec-
trum [7] is presented. The component was made from aluminium alloy
(Al)7050-T7451. Each of the repeating marks indicated is a block of the
mini-FALSTAFF spectrum. The peak load in this spectrum generated a
Kmax of less than 1MPa(m)0.5 at the initial crack size of 0.01mm shown,
while this value was about 10 MPa(m)0.5 by the end (upper edge) of the
view in this Figure. These K values can be compared to those presented in
Fig. 1 (for CA data) i.e. spanning the equivalent Region I and II.
68 L. Molent et al.
From Fig. 2 it can be seen that the crack grows from a surface crack only
a few microns deep (starting from a small crack that had grown during ser-
vice) in a fashion that appears regular and systematic, being free of the in-
consistencies suggested by short crack effects. This indicates that for such
a spectrum of loads, the small to relatively large CG regimes should be
easily predicted.
Fig. 2 An optical view of a fatigue crack surface in an Al7050-T7451 compo-

nent after post service fatigue testing [8]. Between each of the repeating marks
(arrows) is a block of growth due to the application of a single block of the mini-
FALSTAFF load spectrum.
The simple model discussed in this chapter is based on four primary ob-
servations, namely:
1) The growth of naturally occurring surface cracks commences shortly

after the introduction into service (i.e. the initiation phase is an insig-
nificant portion of the total life). Thus CG can be assumed to occur
from the first load cycle and the total life is represented by the CG from
the initiating defect size to the point of loss in residual strength at the
critical crack size;
2) The growth of critical (i.e. the fastest) surface initiating cracks in near
ambient conditions loaded under VA loading will produce approxi-
mately exponential growth, which when presented on log crack depth
versus linear life (in load cycles, or blocks of cycles) plot, gives straight
line. This appears to consistently arise in the absence of fluctuating
stress fields (residual stress, stress gradients, load shedding etc.), sig-
nificant changes in geometry (i.e. finite width effects are negligible)
and/or significant changes in environment. This behaviour has been
shown to apply for a wide range of problems [9];
3) If the CG history data are generated on a spectrum block-by-block basis,
the rate of CG (i.e. da/dB, where B is one block of the loading spectrum),
is directly proportional to the crack depth [10], again implying exponen-
tial CG based on crack depth. Thus, when testing, it is important to gen-
erate data for many blocks of crack growth (| 10 or more blocks); and
4) The da/dN or da/dB for a given spectrum is approximately proportional
to the applied stress cubed. This was first noted for CA loading [11]
and later for VA loading [9].
Whilst a CG model could be developed based on the crack depth and ap-
plied stress (see [9], [11]), there is an advantage in using K; as it includes a
geometry factor (E), for which many geometries have been analysed and
the results of which are available in the open literature e.g. [12]. In addi-
tion, much CA CG data of the form of da/dN versus K for many materials
already exists. The Paris and related formulations rely on an exponent of
K to describe the CG rate found from long crack growth experiments.
However, as will be shown in this chapter, the VA growth is typically ex-
ponential based on the crack length. Therefore the power on crack length a
needs to be fixed (with an exponent of unity) while the power on K should
be allowed to vary along with the long crack data. This leads to a Paris-like
related model of the form [9] for CA data:
da
C * a (1 m*/ 2 ) ('K ) m* (1)
dN
or for VA data:
da
C * a (1 m*/ 2 ) ( K ref ) m* (2)
dB
where, C*, and m* are constants for a given material, spectrum and applied
stress. Here it should be noted that da/dN is a function of both a and ¨K
(which when combined give the CG a ¨K dependence with a correction
that makes the CG exponentially dependent on crack depth).
While Equation (1) may be suitable for CA data, a further complication
occurs in that s‘ pectrum effects’ su ch as overloads and underloads, when
applied as part of a group of loads may have a significant influence on the
70 L. Molent et al.
overall damage created by the spectrum or block of loads. The damage

may be greater or lesser than the sum of the damage created by the loads
when applied constantly as occurs in CA loading. But if a block of loads,
which includes the overloads and/or underloads is treated singularly in the
same manner as a single cycle of CA loading (i.e. on a block-by-block ba-
sis) the same formulae (in the form of Equation (2)) will be shown to be
applicable with the exception that a reference K value such as Kmax (the
peak K applied in the b‘ lock’ of loading) is used, as will be seen in this
chapter.
This model was derived to predict exponential crack growth when m*
was approximately equal to 3 (i.e. cubic stress dependency). Further Re-
gion III (or growth due to quasi-static modes of failure) can be accounted
for by incorporating Kc in the formulation, where here blocks (B) of loads
are considered instead of single loads:
~
da/dB = ( C a1- m*/2 Kmaxm*)/(1.0 - Kmax/Kc) (3)
~
where C is material, stress and spectrum dependent, Kc is the apparent
mode I fracture toughness depending on the plane stress or plane strain or
a mix of the values depending on the apparent fracture toughness of the
component with the fatigue crack just prior to failure. Kc may simply be
calculated using the K formulation, which for a block-by-block analysis
can be based on the maximum stress in the block, Vmax:
Kc = Vmax E (S a) (4)
In the remainder of this chapter the model is shown to fit several exam-
ples of CG, where the cracks have grown in components with complex ge-
ometries and loading. The E were taken from standard handbook solutions.
3. Low Threshold Constant Amplitude Crack Growth
Much has been written on the so-called short crack effect [13] for CA
data. To adequately reinvestigate this region of growth, data needs to be
assessed which includes the a versus N along with the corresponding ¨K.
Unfortunately, published data with the a explicitly included is rare, al-
though one such data set (Fig. 3: crack depth versus cycles, and Fig. 4:
crack growth rate versus ¨K) is available for Al2024-T351 [14], and is in-
vestigated here. These data were generated from standard middle through
crack coupons of length = 152mm, thickness = 12.7mm, width = 76.2mm
and initial notch width 2a = 12.7mm, tested at R=0.1. The crack lengths
reported were the average value of the extension from both sides of the
notch.
From Fig. 4 it can be seen that the data appear to produce the classic
sigmoidal shape that represents growth from Region I through to Region
III. Now, if it is considered that the low CG rate data extends up to ap-
proximately 1 x 10-7 m/cycles as indicated on Fig. 4, and the rate data are
expressed as a function of a, which is shown in Fig. 5, then it can be seen
that this region represents crack lengths of up to 10mm. From Fig. 3 this
may be seen as the majority of the life of these specimens.
When Equation 1 is applied to these data, it can be seen in Fig. 6 that a
near linear relationship, regardless of the region, is produced suggesting a
unification of Regions I to III.
Fig. 3 Crack growth history data (measured from notch edge).

72 L. Molent et al.
Fig. 4 Crack rate data for Al2024-T4 from [14]. Note log-log scale. Region
Iappears to end at about the line indicated.
Fig. 5 Crack rate data versus crack depth.

Fig. 6 Crack growth data as per Equation (1). Note the linear-linear scale.
4. Exponential Crack Growth for Describing Spectra Data
An examination of the CG of fatigue cracks grown with many spectra has

found that the CG curves suggest exponential growth [6]. This has been
found to be the case for most of the life of cracks grown in service or from
full-scale fatigue tests [16], in addition to coupon tests carried out under
service spectra [9]. The following formulations describe this growth:
ln a \N L ln a 0 or a a 0 e\N L (5)
where \ is a parameter that is geometry, material and load dependent, a is

the crack depth at time NL, and a0 is the initial crack-like flaw size (depth
of the crack at the start of loading, N=0), and NL is the number of cycles to
grow from a0 to a.
By way of an example of this CG behaviour, some CG history data for
Al7050 material tested under a VA service fighter aircraft spectrum and
measured by quantitative fractography (QF), is shown in Fig.7. Equation 5
74 L. Molent et al.
describes the growth depicted well and the extent of this agreement spans
from approximately 0.01 to several millimetres. In this fatigue test pro-
gram [15] up to five h‘ our glass’ (very low Kt) Al7050 specimen sets were
tested at up to four stress levels each. The specimens were 6 .3 5 mm
thick and 28mm wide at their thinnest width with material properties as
listed in Table 1. The coupon is shown schematically in Fig. 8. Five com-
plex service load sequences typical of fighter aircraft spectra, were applied
(two sets of CG curves covering two of the spectra at one stress level are
shown in the figure). The CG curves shown in Fig.7 demonstrate exponen-
tial growth.
Fig.7 Graph of QF crack depth against flight hours for cracks in aluminium alloy
7050-T7451 test coupons loaded with two F/A-18 usage spectra, from [15].
Table 1 Material properties of the Al7050-T7451 coupons used to generate the

crack growth shown in Fig.7.
UTS UTS UTS YS YS YS
Coupon L LT ST L LT ST Hardness Conductivity Plate
properties MPa MPa MPa MPa MPa MPa HV10 %IACS Thickness
532 542 510 480 480 435 181 39 144mm
Fig. 8 Schematic of the coupons used to generate the crack growth shown in Fig.7.
In addition to the coupon results shown above, the same form of crack-
ing (exponential) can be shown to exist in full scale structure. The crack
surface shown in Fig. 2 was one of several cracks that occurred in a bulk-
head from an F/A-18 aircraft that was tested to failure in the Flaw Identifi-
cation through Application of Loads (FINAL) series of tests [8] after being
withdrawn from service. The test article in the test rig, along with the lo-
cations of the cracks measured is shown in Fig. 9. Eight cracks were
measured using QF; four from each side of the bulkhead. The CG curves
are presented in Fig. 10 where exponential growth over a crack depth
range from less than 0.01mm to several millimetres is evident.
Lower beam lower

flange failure locations
Fig. 9 An F/A-18 CB mounted in the FINAL rig. The locations where the meas-
ured cracks occurred have been highlighted in red.
76 L. Molent et al.
(a) First set
(b) Second set

Fig. 10 Crack growth curves for the four largest fatigue cracks from the failure on
either side of the lower beam of an Y488 bulkhead from the F/A-18 aircraft tested
in the ‘FINAL’ series of tests [8]. The curves were derived from measurement of
the growth per block of mini-FALSTAFF on the fracture surface. The flaw depth
includes the growth from service.
5. Crack Growth Rate for Spectra Data
Taking the CG data from the two spectra shown in Fig.7 [15], the CG rates
were calculated and these are presented in Fig. 11 in the form of CG rate;
da/dB versus a data. Here the trend lines represent the average of all the
CG rates in the tests at this single stress level. The units of da/dB are
mm’s per block and the crack sizes cover the range shown in Fig.7. (For
these cracks the E factor remained essentially constant.) These data show
the advantage of analysing crack growth data on a block-by-block basis,
since such an analysis inherently captures any spectrum sequence effects.
Fig. 11 Apparent growth rate versus the crack depth a, adapted from [15]. Some
crack growth data shown in Fig.7.
Fig. 11 also shows the trend line for each spectrum and reveals a near
linear relationship between da/dB and the crack depth (a). The slope of this
curve is the exponent of the power curves fitted by linear regression. This
Paris-like exponent is approximately 2 for this alloy. This implies expo-
nential growth as has been found to be the case for a wide range of other
fatigue cracks [10].
Using Equation 5, Fig. 12 presents the slopes of the QF CG curves from
[15] derived from the results of CG measurement of coupons tested at the
four different maximum stress levels (including those presented in Fig.7),
plotted against peak applied stress. It can be seen that by fitting power
curves of the form O =C1Vmax where C1 is a constant and [ is the exponent,
to these data the exponent was found to be about 3, thus suggesting that a
cubic describes the crack growth variation with stress.
78 L. Molent et al.
Fig. 12 Crack growth slopes versus notch stress for two spectra, from [15].
Using this cubic power relationship, as suggested by the fits of the data
shown in Fig. 12, along with work presented in the literature [10],[11], it
can be shown to allow prediction of growth rates at one stress level (V1)
from those of another stress level (V2) given the a0 for the new crack, the E
for the original crack growth and that the loading sequence is the same:
3
§ V2 ·
¨¨ ¸¸ \ 1 N
© V1 ¹
a2 a 20 e (6)
To illustrate the predictive capabilities of this simple transfer function,

CG data for two different stress levels taken from an F4 aircraft wing cou-
pon test program presented in [17] is reproduced here in Crack growth
curves for the 207 and 248 MPa peak loaded coupons are shown in this
Figure. The growth of 248 MPa loaded crack is predicted using the data
from the 207 MPa loaded crack, using Equation 5. An initial crack (ao)
size of 0.154mm was used. Other examples are given in [9]. Note that
both crack growth curves are essentially exponential.
One practical outcome of this is that Equations 5 and 6 appear to provide
a means for scaling the results of fatigue testing performed at one stress
level to a different stress level (for the same spectrum).
Fig. 13 Crack growth data for F4 wing, Figure 9.2.1a in [17]. A prediction of the
248 MPa crack has been made from the 207 MPa crack using 0.154mm as the ao.
This section has demonstrated that assumption that m* in Equations 1, 2

and 3 is reasonable.
6. Modeling Crack Growth Histories
In order to demonstrate the validity and flexibility of Equation 2 a number

of case studies are considered in this Section.
Case 1: Crack Growth in Centre Notched Panels
The fatigue performance of 110mm wide, 5mm thick centre notch Al7050-
T736 panels under either a gust loading spectrum (MINITWIST) with an
in-flight stress of 55 MPa; or manoeuvre spectrum loading (FALSTAFF)
with a maximum stress of 171.3 MPa are considered. These data are drawn
from Reference [18]. The measured and computed results, using Equation
(3) (with da/dN substituted by da/dFlights) with the best fit estimates of
C* = 1.4x 10-8 and J = 3 and Kc of 40 MPa m, for the MINITWIST spec-
trum are presented in Fig. 14, and the measured and computed results, us-
ing the best fit estimates of C* =1.8 x 10-8 and m* = 3 and an apparent
toughness Kc of 60 MPam for the FALSTAFF spectrum are presented in
Fig. 15. In both cases we see good agreement between the measured and
the computed crack length histories.
80 L. Molent et al.
Fig. 14 Measured and computed crack growth histories under MINITWIST load-
ing, adapted from [18].
Fig. 15 Measured and computed crack growth histories under FALSTAFF load-
ing, adapted from [18].
Case 2: Crack growth for a range of fighter aircraft spectra
The next set of data considered was chosen specifically because it had
been generated for a modern aircraft design and included a full range of
spectrum types likely to be experienced in the aircraft’s lifetime [19]. The
spectrum characteristic (from [19]) and estimated constants are presented
in Table 2 and the experimental and computed (using Equation 2) data are
presented in
Fig. 16(a) and (b). The material was an Al 7xxx series (not specified) and
the geometry of the coupons was not specified [19]. However it can be in-
ferred that the testing conformed to ASTM E 647 with spectrum loading
applied. It can be seen that for all the spectra types considered the CG
conforms reasonably well to that predicted by Equation 3 and the predic-
tions are reasonable.
Table 2 Crack growth constants and spectrum characteristics.
Estimated
Coupon Kc (ksi
Estimated Max
in) No. of
(identifier) ~ 13 m* stress Mean R
Cycles
C x10 (MPa MPA
m (%DLS)
Fuselage
Bulkhead
103.5 293.3.
Wing At- 2.9 3 0.014 26388
(93) (0.493)
tach
(FBWA)
Fuselage
Bulkhead 74.0 154.9
4.9 3 0.374 35529
Upper Cap (68) (0.934)
(FBUC)
Wing Spar
98. 275.3
Lower Cap 0.76 3 -2.431 29214
(90) (0.557)
(WSLC)
Vertical
Tail Tip 45.7 175.1
2.9 3 -2.266 28083
Rib Attach (43) (0.54)
1 (VTRA)
Vertical
37.0 169.7
Tail Skin 1.8 3 0.182 21896
(34) (0.876)
(VTS)
82 L. Molent et al.
(a) Five sets measured and three sets predicted.
(b) Six sets measured and six sets predicted.
Fig. 16 Measured and computed crack growth for the control points
7. Discussion
It would appear that for practical applications when modelling CG histo-

ries using Equation 3 there are essentially only two constants that when de-
termined allow simple predictions to be achieved for variable amplitude
loading cases. This can be seen from the examples presented in here. Fur-
ther, the constant m* | 3 was shown to be a good approximation for the
examples given here and has been found to be applicable for a wide range
of problems [9]. Consequently, since m* may be approximated as 3 then
~
for prediction a single parameter, C needs to be derived from CG data or
estimated by other methods. These methods, which may be used to derive
~
C for untested spectra are beyond the scope of this chapter and will be
the subject of future work.
8. Conclusions
A three parameter crack growth model was shown to adequately model the
growth of cracks under variable amplitude block loading from microns
through to many millimetres in length i.e. bridging Region I through to
Region III for constant amplitude data.
This is a particularly important development in structural integrity analy-
sis where fatigue cracking cannot always be avoided and the majority of
the fatigue life of nominally gross defect free structure is spent growing
physically small cracks from initiating discontinuities (Region I for con-
stant amplitude loading) up to the point of loss in acceptable strength. The
model described in this chapter was based on the observation that for lead
surface initiating cracks the majority of the life follows approximately ex-
ponential crack growth. The Region III, or the onset of static failure
modes, was also accounted for by considering the fracture toughness of the
material.
By assuming a crack growth rate that is an approximate function of the
applied stress cubed then the model simplifies to only a single parameter
~
( C ) problem given that the fracture toughness of interest is known. How-
~
ever, methods of defining C require further research. This model was
shown to fit experimental crack growth data produced under a wide range
of variable amplitude loading. By considering the crack growth data on a
spectrum block-by-block approach sequence effects are inherently cap-
tured. This simple model involving only one parameter has significant ad-
vantages in the understanding and prediction of crack growth.
References
[1] Schijve J. Fatigue of structures and materials in the 20th century and the state
of the art. International Journal of Fatigue 25 (2003) 679-702.
[2] El Haddad MH, Smith KN and Topper TH. Fatigue crack propagation of
84 L. Molent et al.
short cracks. Engineering Materials and Technology, Trans. ASME, (1979)

42.
[3] Newman JC Jr. FASTRAN II - A fatigue crack growth structural analysis
program, NASA TM-104159.
[4] AFGROW (2006) Version 4.11.14.0. http://www.siresearch.info/index.php
[5] Pearson S. Initiation of fatigue cracks in commercial aluminium alloys and
the subsequent propagation of very short cracks. Eng. Fract. Mech., 1975:
235.
[6] Davidson DL and Lankford J. Fatigue crack growth in metals and alloys:
mechanisms and mircomechanics, International Materials Reviews 37/2
(1992) 45-76.
[7] van Dijk GM and de Jonge LB. Introduction to a fighter aircraft loading stan-
dard for fatigue evaluation F ‘ ALSTAFF’, Proc. Eighth ICAF Symposium,
Emmen, Switzerland, Swiss Federal Aircraft Establishment (F+ W), 3(61)
(1975) 1-39.
[8] Molent L, Dixon B., Barter SA., Medved, J. & Nguyen, Q. (2005) The FINAL
program of enhanced teardown for a fighter aircraft. proc. ICAF 2005 Ham-
burg, Jun 8-10, 2005.
[9] Barter S, Molent L, Goldsmith N and Jones R. An experimental evaluation of
fatigue crack growth. Engng Fail Anal 12(1) (2005) 99-128.
[10] Molent L, Jones R, Barter S and Pitt S. Recent developments in crack growth
assessment, J of Fatigue 28(12) (2006) 1759-1768.
[11] Frost NE and Dugdale DS. The propagation of fatigue cracks in test speci-
mens. J Mechanics and Physics of Solids 6 (1958) 92-110.
[12] Sih GC. Handbook of Stress-Intensity Factors, Institute of Fracture and Solid
Mechanics, Lehigh Univ. USA (1973).
[13] Suresh S and Ritchie RO. Propagation of short fatigue cracks. International
Metals Reviews, 6 (1984) 445.
[14] Forth SC, Wright CW and Johnston WM Jr. 7075-T6 and 2024-T351 Alumi-
num Alloy Fatigue Crack Growth Rate Data”, NASA/TM-2005-213907,
2005.
[15] Pell RA, Mazeika PJ and Molent L. The Comparison Of Complex Load Se-
quences Tested At Several Stress Levels By Fractographic Examination, En-
gineering Failure Analysis12(4) (2005)586-603.
[16] Molent L and Barter SA. A comparison of crack growth behaviour in several
full-scale airframe fatigue tests, Int J Fatigue 29 (2007) 1090-1099.
[17] USAF Damage Tolerance Design Handbook. 1998. Wright Patterson AFB,
USA.
[18] Wanhill RJH. Damage tolerance engineering property evaluations of aero-
space aluminium alloys with emphasis on fatigue crack growth, NLR TP
94177, 1995.
[19] Ball DL and Norwood DS and TerMaath SC. Joint Strike Fighter Airframe
Durability and Damage Tolerance Certification, Proceedings 47th
AAIA/ASME/ASCEI Structures, Structural Dynamics, and Materials Confer-
ence, pp 1-17, 1 - 4 May 2006, Newport, Rhode Island, USA.
Fatigue Behaviour of FS, LB and MIG Welds of
AA6061-T6 and AA6082-T6
P. M. G. P. Moreira*, V. Richter-Trummer and P. M. S. T. de Castro
Faculdade de Engenharia da Universidade do Porto
Rua Dr. Roberto Frias, 4200-465 Porto, Portugal
*Email: pmgpm@fe.up.pt
Abstract
The increasing use of aluminium alloys in transportation industry, not only

in aeronautics but also in automotive industry, creates the need for re-
search on more efficient and reliable welding processes to be used. In or-
der to allow the industry to use novel manufacturing techniques as Laser
Beam Welding and Friction Stir Welding, which promise high efficiency,
research work on S-N and crack growth fatigue data from the weld zone is
required to provide tools to assess the damage tolerance. This chapter is a
contribution to this effort, contrasting the fatigue behaviour of joints made
using a traditional process, Metal Inert Gas welding, with those made with
Friction Stir and Laser Beam Welding.
1. Introduction
Notwithstanding the widespread interest in the possibilities offered by fric-

tion stir and laser beam welding, data comparing the fatigue behavior of
joints obtained using these processes is still needed. In this study the fa-
tigue behavior of joints made using a traditional process (metal inert gas
welding), was contrasted with the behavior of joints made using laser
beam welding and friction stir welding.
Friction stir welding [1], a solid-state joining process developed and pat-
ented by The Welding Institute, emerged as a welding technique to be used
in high strength alloys that were difficult to join with conventional tech-
niques. The process was developed initially for aluminum alloys, but since
then FSW was found suitable for joining a large number of materials. In
aeronautics, for instance, riveting is the preferred manufacturing process
for aircraft fuselage structures; nevertheless, FSW is emerging as an ap-
propriate alternative technology due to low distortion, high strength of the
joint and high processing speeds.
Conventional fusion welding of aluminum alloys produces a weld prone
to defects such as porosity, consequence of entrapped hydrogen gas not be-
ing able to escape from the weld pool during solidification. An example of
86 P. M. G. P. Moreira et al.
a fusion process is the Gas Metal Arc Welding (GMAW) process, MIG
(Metal Inert Gas) welding process [2]. In the MIG welding process the arc
and the weld are protected from atmospheric contamination by a gas
shield, and an electric potential is established between the electrode and
the workpiece causing a current flow, which generates thermal energy in
the partially ionized inert gas [3].
The Laser Beam Welding (LBW) process is also a thermal effecting
melting process which causes metallurgical changes in the weld metal as
well as in the heat affected area beside the weld seam. Effects such as oxi-
dation reactions are comparable to other thermal treatments in the case of
aluminum; but because of the short cycle time oxidation of the weld seam
is very low [4].
Because of the large temperature interval between melting and solidifica-
tion in the heat affected zone the material is partially molten in the solid
aluminum matrix. In the transition area between the weld seam and heat
effected zone small defects (solidification cracks and pores) occur during
the solidification process. Because of their size there is no special negative
influence on the mechanical properties. The thermal influence on the mate-
rial structure beside the weld seam itself is very small. Welding an alloy
which does not require additional wire because of metallurgical effects
shows that there is no change in chemical composition compared to the
base material. The main reasons for this feature are: the high welding
speed; the short cycle time of the melting and solidification process; the
deep penetration effect of the keyhole with a very large ratio of depth to
width; and the turbulent melt pool dynamic.
In Friction Stir Welding (FSW) the interaction of a non consumable and
rotating tool with the workpieces being welded creates a welded joint
trough frictional heating and plastic deformation at temperatures below the
melting temperature of the alloys being joined.
Studied is the fatigue behavior of friction stir and laser beam butt welds
of two 3mm thickness age hardenable aluminum, 6082-T6 and 6061-T6 al-
loys, is presented. For the case of the laser beam welds the aluminum
6061-T6 was the only material tested. For comparison, MIG butt welds of
the same alloys were also performed and tested. Tensile tests and micro-
hardness measurements of weld joints and base materials were performed
in order to determine the influence of each welding process in the me-
chanical properties. The fatigue behavior (S-N curves) of specimens of all
welding process was analyzed. Microstructure was examined and corre-
lated with the macroscopic mechanical behavior. Scanning electron mi-
croscopy was carried out and the fractographic features of friction stir
welds and MIG welds compared.
Fatigue of FS, LB & MIG Welds of AA6061-T6 & AA6082-T6 87
2. Material and Welding Processes
The alloys 6082-T6 and 6061-T6 are high strength Al–Mg–Si alloys that
contain manganese to increase ductility and toughness. The alloy was solu-
tion heat treated at §540°C, followed by quenching to room temperature
which leads to a supersaturated solution. If the material is stored at room
temperature natural ageing takes place (T4 condition). The T6 condition is
obtained through artificial ageing at a temperature of approximately 180°C
[5].
The MIG welding parameters used were: 128A, 17.1V, 700mm/min and
Argon at a 20l/min flow rate. A filler wire AWS ER5356 (with the desig-
nation ESAB OK Autrod 18.15) with a diameter of 1mm [6] was used.
Welds were performed using an automated welding robot, GMF
ROBOTICS – Arc Mat Sr.
The laser beam welds were performed using a ABB IRB6600 robot in
conjunction with a Nd:YAG laser, with a power source of 4kW. The fol-
lowing parameters were used in tests: welding speed 50mm/seg; focal dis-
tance 200mm; 40l/min He flow; gap between plates 0.2mm; filler wire of
1.2mm diameter with a feeding speed of 5.5m/min ER4043.
The friction stir welds were performed in a prototype machine with a ca-
pability of 6m weld length developed to be used in a Portuguese shipyard.
For both aluminum alloys were used the same parameters: welding speed
of 800mm/min; pitch angle of 2º; rotating speed of 1500rpm. The FSW
process of the aluminum 6082-T6 was performed using a tool with a 6mm
diameter threaded pin and the shoulder had 15mm diameter. For the alumi-
num 6061-T6 a tool with a 4mm diameter threaded pin and a shoulder of
10mm diameter were used. Since the optimal welding speed depends on
several factors: alloy type, penetration depth and joint type being the most
important these parameters were chosen by trial and error attempts until no
visually detected defects could be identified. The penetration depth was
adapted to fully penetrated butt joint in a material of 3mm thickness.
3. Tensile Tests
Tensile tests were performed to determine the mechanical properties of the

welded and un-welded material (yield stress ıyield, rupture stress ıupt and
Young modulus E). The average values of ıyield, ırupt and E of tests per-
formed in base material specimens of both aluminum 6082-T6 and alumi-
num 6061-T6 alloy are presented in Table 1. The aluminum 6061-T6 has
an ultimate tensile stress about 6% higher than the aluminum 6082-T6. All
base material (BM) specimens failed in the same manner, 45º shear plane.
The average values of ıyield, ırupt and E of tests performed in base mate-
rial specimens of both aluminum 6082-T6 and aluminum 6061-T6 alloy
welded by MIG, LBW and FSW technique are also presented in Table 1.
Since the rupture stress of the MIG welded specimens is lower than the
yield stress of the parent material the elongation measured in the fractured
surface will only reflect the measurement of the material affected by the
welding process. This resulted in lower values of elongation when com-
pared with the base material.
Table 1 Material properties obtained in tensile tests.

Material ıyield ırupt Elongation
Specimen type E [GPa]
[MPa] [MPa] [%]
Base material 6082-T6 276.2 322.9 67.1 17.5
6061-T6 306.3 342.0 68.5 17.1
MIG 6082-T6 176.8 210.0 - 4.0
6061-T6 156.3 221.2 - 4.2
FS welded 6082-T6 140.5 226.1 - 5.5
6061-T6 158.7 241.5 - 4.0
LB welded 6061-T6 171.0 265.4 - 3.7
It was found that the friction stir welded specimens have lower yield
stress values than MIG specimens but higher values of the rupture stress.
Also the elongation of the friction stir welded specimens presented higher
values.
For the case of MIG welded specimens rupture occurred outside the
welding seam in the HAZ (heat affected zone) as presented in Figure 1.
Similar observations are presented by Ericsson et al. [5] where a 45º frac-
tured surface was also found. The aluminum 6082-T6 MIG welded speci-
mens present a yield stress and a rupture stress of 65% of the base mate-
rial. The yield stress obtained in this work is 20% higher than the results
presented by [5]. The aluminum 6061-T6 MIG welded specimens has a
yield stress of 51% of the base material and a rupture stress of 65% of the
base material. Comparing MIG weldments of both alloys it was found that
the yield stress of the 6082-T6 is higher but the rupture stress is higher for
the 6061-T6 MIG welded specimens.
Fig. 1 Fracture surface of a MIG welded specimen of AA6061-T6.
In the case of the 6082-T6 friction stir welded specimens fracture oc-
curred near the weld edge, Figure 2, where a decrease of hardness occurs
[7]. The fracture surface presents a 45º angle, as presented by Svensson et
al. [8]. In the case of the aluminum 6061-T6 the fracture started at the
weld root indicating that a lack of penetration occurred during welding
(root flaw). Since the fracture surface occurred at the weld middle line it is
possible to identify the different layers of material that were formed by
each rotation of the welding tool. Dickerson et al. [9] suggested that root
flaws up to 0.35mm deep do not cause degradation in mechanical perform-
ance when compared to flaw-free welds. The elongation of all friction stir
welded specimens (4% to 5%) is approximately 25% of the base material
(17%).
Fig. 2 Crack path of AA6082-T6 FS welded specimen.
The aluminum 6082-T6 friction stir welded specimens present a yield

stress of 51% and a rupture stress of 70% of the base material. Scialpi et
al. [7] obtained also a relation of rupture stress of 76% between the base
material and friction stir welded specimens. Values of the same magnitude
are also reported by Ericsson [5], Harris [10], and Nicholas [11] in their
literature reviews.
The aluminum 6061-T6 friction stir welded specimens have a yield
stress of 52% and a rupture stress of 71% of the base material. Hong et al.
[12] performed friction stir weld of 4mm thick aluminum 6061-T651 plates
and found lower values for the yield stress and rupture stress when com-
pared with those obtained in this study.
From all type of welded specimens, laser beam welded specimens pre-
sent the higher yield and rupture stress. The aluminum 6061-T6 laser beam
welded specimens have a yield stress of 56% and a rupture stress of 78%
of the base material. These results are higher than the ones presented in [4,
13]. Nevertheless, the elongation presents the lowest values. In the 6061-
T6 laser beam welds fracture occurred near the weld bead edge with a 45º
angle, Figure 3.
Fig. 3 LB weld crack orientation.
4. Microhardness Profiles
The hardness profiles can assist in the interpretation of the weld micro-
structure and mechanical properties. Microhardness tests were performed
to characterize the Vickers hardness profile in the vicinity of the weld area.
Measurements were performed at the specimens’ middle thickness using a
100gf load. Figure 4 illustrates the hardness profiles of the MIG welded
aluminum 6061-T6 and 6082-T6 specimens. The major softened areas are
the weld centre line, and the two transitions zones just at the limit of the
HAZ. In these areas the hardness reaches a minimum value near 50HV and
the base material has values over 90HV. Minimum average values are
found in the aluminum 6082-T6 specimen.
In the friction stir welded aluminum 6082-T6 and 6061-T6 a hardness
decrease is identified in the TMAZ (thermo-mechanically affected zone).
The average hardness of the nugget zone was found to be significantly
lower than the hardness of the base alloy. There is a zone outside the nug-
get zone which has the lower hardness value. The welding process sof-
tened the material reducing the hardness to 33% of the parent material, as
shown in [7]. The hardness minimum values are obtained in the welding
retreating side, e.g. [8]. As it also suggested in [10] the variation of the mi-
crohardness values in the welded area and parent material is due to the dif-
ference between the microstructure of the base alloy and weld zone. A di-
rect comparison of measurements of both friction stir welded alloys is pre-

sented in Figure 5.
Fig. 4 Microhardness profiles of the MIG welded 6061-T6 and 6082-T6 speci-
mens.
Fig. 5 Microhardness profile of the FS welded 6082-T6 and 6061-T6 (6082-T6:

6mm diameter pin, shoulder 15mm; 6061-T6: 4mm diameter pin, shoulder 10mm).
As described in the literature in alloys such as 6082 and 6061 the main
strengthening precipitate is ȕƎ-Mg5Si6 which is stable at temperatures <
200°C [14]. This precipitate exists in the unaffected base material but is
absent in the nugget and in the HAZ. In the friction stir weld temperatures
are expected to be over 200-250°C, during heating, the ȕƎ is easily dis-
solved [8]. This ȕƎ precipitate is the mainly responsible for hardening. It
was reported [8] that no other fine scale precipitation was found in the
nugget, while in the HAZ precipitation of ȕ’-Mg1,7Si occurred on the Al–
Mn–Si dispersoids. On cooling, precipitation of ȕƍ is favorable to ȕƎ, and
since the ȕƍ precipitates have less of a strengthening effect compared to ȕƎ,
a lower hardness is obtained. The ȕ’ will act as a nucleation sites for the
precipitates. Following the diagram of continuous cooling precipitation
(CCP) presented in 14], in the HAZ where temperatures are near or less
than 300ºC the precipitation of ȕ’ is very high, it occurs the transition from
ȕ’’ to ȕ’ by dissolution. In the weld nugget, where the temperature is ex-
pected to be higher and so the MgSi precipitates go into solution. On cool-
ing, the time for precipitation is limited, and so only a small volume frac-
tion of the ȕƍ precipitates form in the weld nugget. The nugget hardness
recovery is due to recrystallization of a very fine grain structure.
The microhardness profile, at mid thickness, of the aluminum 6061-T6
laser beam weld is presented in Figure 6. The minimum hardness value
found was obtained inside the weld bead, near its limit, and has a value of
approximately 60HV. The bead limit presented a width of 2.5mm and the
HAZ, which corresponds to hardness values lower than the base material,
has a width of approximately 4.5mm from each side of the bead limit.
Fig. 6 Microhardness profile of the LB welded AA6061-T6.
As presented in Figure 7, when compared with the friction stir welds and
MIG welds, the laser beam welds presented the smaller area (width) af-
fected by the welding process, approximately 12mm. For the friction stir
welds the width of the welding affected area was approximately 15mm,
and in the MIG welding process 24mm.
Fig. 7 Microhardness profile of the LB, MIG and FS welded AA6061-T6.
5. S-N Fatigue Data
Fatigue tests were carried out in a servo-hydraulic MTS testing machine

and specimens’ dimensions were chosen according to the ASTM standard
E466 [15]. The weld was perpendicular to the load direction in the S-N
tests and to the material rolling direction. The maximum stress levels used
were selected as a function of the yield stress for each type of joint. Values
of 120% to 40% of the yield stress were chosen. A stress ratio of R=0.1
was used and the frequency was in the interval of 7–26Hz depending upon
the stress level. Life was defined as the number of cycles to failure and the
number of cycles considered as a threshold for infinite life was 107 cycles.
An extensive program was performed analyzing all type of specimens of
aluminum 6082-T6 and 6061-T6: base material specimens, MIG speci-
mens, friction stir welded specimens and laser beam welded specimens (of
aluminum 6061-T6 only). All specimens were tested in the as-welded con-
figuration.
For the MIG welding specimens the geometrical detail of the transition
from the weld seam to the plate surface of the base material can be inter-
preted as a geometrical defect that enhances the initiation of a first crack.
For the MIG weld specimens fracture occurred in the HAZ near the notch
created during the deposition of filler material (Figure 8). Figure 9 presents
examples of the fracture surface of MIG specimens.
(a) Fracture weld after fatigue (b) Side view

Fig. 8 Fracture location of MIG specimens AA6082.
(a) AA6082-T6 MIG specimen with (b) AA6061-T6 MIG specimen with
fracture surface fracture surface
Fig. 9 Fracture surface for the fatigue tested MIG specimens.
Some MIG welded specimens presented defects due to the welding proc-
ess. Due to the fusion nature of this process, during the welding the mate-
rials entrapped pores appear which influences the specimen mechanical
behavior. An example of this phenomenon is identified in the fractograph
presented in Figure 10. This fractograph corresponds to a MIG welded
6061-T6 specimens tested at 75% of its yield stress and had a fatigue life
of 16492 cycles (six times lower than the average fatigue life for this re-
mote load). Results for the fatigue tests are presented in Figure 11.
Fig. 10 Fractograph of a MIG welding specimen containing entrapped pores.

Fig. 11 S-N fatigue data of the MIG welded specimens.
Data was fitted using a power equation and some fatigue scatter was
found in the as-welded MIG specimens. The 6061-T6 MIG welded speci-
mens presented higher fatigue lives than the MIG 6082-T6 specimens.
Friction stir welded specimens of both base materials were fatigue tested.
The fracture location of the frictions stir welded specimens is shown in
Figure 12. In the friction stir welded aluminum 6082-T6 specimens frac-
ture propagated at the centre of the weld and in the aluminum 6061-T6
near the shoulder limit. These fracture locations were also identified in the
tensile tests. Figure 13 presents examples of the fracture surface for all
types of friction stir welded specimens where a more irregular surface is
identified in the aluminum 6082-T6.
(b) AA6061-T6 FS welded fa-

(a) AA6082-T6 FS welded fatigue
tigue tested specimen, fracture
tested specimen, fracture location
location
Fig. 12 Fracture location of FS welded specimens.
(a) AA6082-T6 FS welded

(b)AA6061-T6FSwelded specimen
specimen
Fig. 13 Fracture surface for the fatigue tested FS welded specimens.
Fig. 14 S-N fatigue data of the FS welded specimens.
Results of fatigue life obtained for the friction stir welded specimens are
presented in Figure 14.
In tests of friction stir welded specimens, data shows a narrow scatter
and was fitted using a power equation. For the friction stir welded 6082-T6
it was found that for 65% and 60% of the yield stress the fatigue life is
considered infinite. In friction stir welded specimens fatigue cracks have
propagated at the centre of the weld, as verified in [8] when studying fric-
tion stir welded 5mm and 10mm thick 6082-T6 specimens. Also, Ericsson
et al. [5] in their study on the fatigue life assessment of the same alloy ob-
tained a fatigue life of 5u105 cycles to failure with a stress range of about
90MPa, at a stress ratio, R=0.5. In the present study, the fatigue life of
5u105 cycles is obtained at a stress range of 105MPa. The friction stir
welded 6061-T6 specimens presented lower lives than the friction stir
6082-T6 specimens when tested at stresses lower than 130MPa. For the
friction stir welded 6061-T6 specimens’ cracks have propagated near the
limit of the tool shoulder.
Laser beam welded specimens of aluminum 6061-T6 were fatigue tested.
The laser beam welds were carried out using additional wire and there was
a seam overfill of approximately 1mm at the top of the weld seam. The
fracture location of the laser beam welded specimens occurred at the end
of the overfill. Results of fatigue life obtained for the friction stir welded
specimens are presented in Figure 15.
Results of fatigue tests obtained in this study are significantly higher
than the ones presented on [4]. When compared to its base material, there
is a decrease of approximately 60% of the fatigue strength. The reasons for
this decrease in the properties are the reduced strength capabilities in the
weld metal and heat affected zone, and also a strong influence of geomet-
rical discontinuities (especially in fatigue).
Fig. 15 S-N fatigue data of the LB welded AA6061-T6 specimens.
Base materials specimens were also fatigue tested at stresses between

90% and 65% of their yield stress. A comparison between the data from
the base materials tests and all welded specimens is presented in Figure 16.
The base materials specimens presented much higher fatigue live than any
type of welded specimens, being the aluminum 6061-T6 the one that per-
formed better.
Fig. 16 S-N data of base material and welded specimens.

6.1 MIG welded specimens
6.1.1 MIG welded AA6082-T6
The macrostructure of the aluminum 6082-T6 MIG weld is presented in

Figure 17. The identification of the two weld beads is straightforward, and
the transition between the HAZ and base material is also identified. Exam-
ples of microstructures are shown in Figure 18.
Fig. 17 Macrostructure of the MIG welded AA6082-T6.
(a) Microstructure 1, weld (b) Microstructure 2, weld
Parent material
Welded material
(c) Microstructure 3, weld (d) Microstructure 4, base material
Fig. 18 MIG welded AA6082-T6 microstructures.

Microstructure 1 shows the structure in the centre of the weld. Micro-

pores with a diameter no larger than 100µm, dispersed in the matrix are
identified in microstructures 2 and 3. These pores are due to the fusion na-
ture of the MIG welding technique. The transition between the material
that suffered fusion to the material only affected by the heat generated dur-
ing welding is shown in microstructure 2. A structure of larger grains is
identified in the left side of microstructure 3 which corresponds to material
that suffered fusion. The HAZ material has a structure similar to the base
material, fine grains with dispersed precipitates. Microstructure 4 shows
the base material.
6.1.2 MIG welded AA6061-T6
The macrostructure of the aluminum 6061-T6 MIG weld is presented in

Figure 19. Also, the identification of the two weld beads is straightfor-
ward, and the transition between the HAZ and base material is also identi-
fied. Examples of microstructures are shown in Figure 20.
In microstructures 1 and 2, solidification structure, centre of the weld are
identified micropores with approximately a maximum of 100µm of diame-
ter disperse in the matrix. In microstructure 3 is identified the transition be-
tween the material that suffered fusion (right side) to the material only af-
fected by the heat generated during welding (left side). In the virtual line
that defines the transition between the HAZ material and the fused material
it is possible to identify a structure containing grains formed due recrystal-
lization. Microstructure 4 shows the base material matrix where grains of
approximately 80µm and some precipitates are present.
Fig. 19 Macrostructure of the MIG welded AA6061-T6.

(a) Microstructure 1, weld (b) Microstructure 2, centre of the weld
(c) Microstructure 3, limit of the weld (d) Microstructure 4, base material

bead
Fig. 20 MIG welded AA6061-T6 microstructures.
6.2 LBW welded specimens
The macrostructure of the aluminum 6061-T6 LB weld is presented in

Figure 21. The identification of the weld bead is straightforward, and the
transition between the HAZ and base material is also identified. Examples
of microstructures are shown in Figure 22.
Fig. 21 Macrostructure of the LBW welded AA6061-T6.

(a) Microstructure 2,centre of the (b) Microstructure 3,limit of

weld detail the weld bead
(c) Microstructure 4, limit of the (d) Microstructure 5, HAZ

weld bead detail
Fig. 22 LBW welded AA6061-T6 microstructures.
The microstructural analysis revealed a structure similar to the one found

for MIG welding, but since the heat was very localized the area that ex-
perienced microstructural changes is much smaller in LBW. In microstruc-
tures 1 and 2, solidification structure, centre of the weld is almost absent of
pores. In microstructure 3 is identified the transition between the material
that suffered fusion to the material only affected by the heat generated dur-
ing welding (HAZ). Microstructure 5 shows grain structure in the HAZ.
6.3 FSW specimens
6.3.1 FS welded AA6082-T6 specimens
In Figure 23 the macrostructure of the friction stir weld of the aluminum

6082-T6 alloy is presented. At the centre is possible to identify the weld
nugget (NZ) with its typical shape of onion rings. The weld nugget experi-
ences high strain and is prone to recrystallization. Immediately at its side is
the TMAZ which ends at the tool shoulder delimited by the dashed lines.
After the TMAZ appears a zone affected only by the heat generated during
the welding process, e.g. [7, 17]. Friction stir welding gives rise to micro-
structure changes.
Fig. 23 Macrostructure of the FS welded AA6082-T6.
Microstructures obtained in the different locations are shown in Figure

24. Friction stir welding gives rise to microstructure changes. As observed
in [8] the base material contains two sizes of grains owing to partial recrys-
tallization. The recrystallized grains are approximately 20 P m in size while
the non-recrystallized grains can be larger than 100 P m. Microstructure 3,
structure at the weld nugget, shows dynamic recrystallized grains much
smaller and equiaxed when compared to the elongated base metal micro-
structure. The grain evolution between the base material and the welded
affected material is evident comparing microstructures 3 and 4. The grains
in the nugget zone are equiaxed [8]. In the HAZ the grain size is similar to
the base metal. Second-phase particles in the workpiece are essentially
stirred into the weld zone. Similar observations are presented in [7]. The
dynamic continuous recrystallization microstructure which characterizes
the friction stir welding process of this alloy is well documented in [18].
(a) Microstructure 1, weld (b) Microstructure 2, weld
(c) Microstructure 3, weld (d) Microstructure 4, BM

Fig. 24 FS welded AA6082-T6 microstructures.
6.3.2 FS welded AA6061-T6 specimens
In Figure 25 the macrostructure of the friction stir weld of the aluminum

6061-T6 alloy is presented. At the centre is possible to identify the weld
nugget. The thermo-mechanically affected zone ends at the tool shoulder
delimited by the dashed red lines. After the thermo-mechanically affected
zone appears the heat affected zone. In this weld sample, particularly at the
advancing side, a high penetration of the shoulder is identified creating a
small notch effect that could affect mainly the fatigue lives.
Fig. 25 Mcrostructure of the FS welded AA6061-T6.
For the case of the friction stir welded aluminum 6061-T6 the micro-
structures obtained in the different locations are shown in Figure 26.
(a)Microstructure 1, weld (b) Microstructe 2, limit of TMAZ to HAZ

nugget
(c) Microstructure 3, base material (d) Microstructure 4 , weld nugget

Fig. 26 FS welded AA6061-T6.
In the welded zone grain shape and dimensions evolution is quite evi-
dent. In the heat affected zone the grain size is similar to the base metal.
Second-phase particles in the workpiece are essentially stirred into the
weld zone. Similar observations are presented in [7]. The dynamic con-
tinuous recrystallization microstructure which characterizes the friction stir
welding process of this alloy is well documented by Liu et al. [18]. As in
the aluminum 6082-T6, the grains in the nugget zone have an approximate
equiaxed shape and a large difference between the base material and the
nugget zone is easily identified as already shown by Liu et al. [18].
7. SEM Analysis
Fractured fatigue test specimens were analyzed by SEM. Measurements

were performed considering the crack initiation site as the origin and
measurements were obtained through the crack length. The result in a co-
ordinate is the average of four measurements of the number of striations in
a perpendicular line to their orientation.
(a) AA6082-T6 fractograph 4, crack (b) AA6082-T6 fractograph 18 crack

length of 0.503mm length of 3.763mm
(c) AA6061-T6 fractograph 6, crack (d) AA6061-T6 fractograph 19, crack

length of 0.377mm length of 3.401mm
Fig. 27 Fatigue striations for different crack lengths.
7.1 Base material striations spacing
A fatigue test specimen of aluminum 6082-T6 and another of aluminum

6061-T6 subjected to fatigue tests with a maximum stress of 70% of the
yield stress (193.3MPa for the aluminum 6082-T6 and 214,4MPa for the
aluminum 6061-T6) and R=0,1 were analyzed. The 6082-T6 base material
specimen analyzed had a fatigue life of 485858 cycles and the 6061-T6 a
fatigue life of 783586 cycles.
For the aluminum 6082-T6 the fatigue area has 12.6mm2 and a crack
length of 4.38mm. For the aluminum 6061-T6 the fatigue area has 13.4mm2
and a crack length of 4.60mm. Figure 27 shows fatigue striations at differ-
ent crack lengths. The striation spacing vs. crack length for such specimens
is presented in [19], where some of the data concerning MIG and FSW
joints is also presented.
7.2 MIG welded specimens, striations spacing
Two MIG butt welded specimens in the as-welded configuration were ana-
lyzed after fatigue testing (R=0.1). The specimen of aluminum 6082-T6
was fatigue tested at a maximum stress of 60% of the yield stress
(106.1MPa), and the specimen of aluminum 6061-T6 was tested at a
maximum stress of 70% of the yield stress (109.4MPa). The aluminum
6082-T6 specimen presented a fatigue life of 46645 cycles and the 6061-
T6 specimen had a fatigue life of 28476 cycles. In the MIG welded speci-
mens fatigue cracks appeared at the end of the weld seam, near the V notch
effect due to the extra material deposited by the melted feed wire.
For the aluminum 6082-T6 the fatigue area has 24.8mm2 and a crack
length of 12.56mm. For the aluminum 6061-T6 the fatigue area has
22.4mm2 and a crack length of 12.50mm.
In the aluminum 6082-T6 MIG welded specimen the striations identifi-
cation process was very hard to carry out. The fatigue crack surface pre-
sented a heterogenic structure with different structures randomly dispersed.
Figure 28 shows SEM images of fatigue striations at two crack lengths for
the aluminum 6082-T6 MIG welded specimen. For example in fractograph
9, presented in Figure 28, besides the larger striations at the top of the frac-
tograph a finer type of striations are identified. This leads induces that the
coarser striations are of a second order that contain several finer striations.
Figure 29 shows SEM images of fatigue striations at two crack lengths
for the aluminum 6061-T6 MIG welded specimen.
(a) Fractograph 4, crack length of (b) Fractograph 9 crack length of

0.361mm 2.582mm
Fig. 28 Fatigue striations for two crack lengths, AA6082-T6 MIG welded speci-
men.

1.893mm 10.250mm
men.
In this specimen the striations identification process was easier than in

the aluminum 6082-T6 specimen. Nevertheless the fatigue crack surface
was a heterogenic structure with several different structures dispersed ran-
domly. For example, in fractograph 8 presented in Figure 29 fine striations
are identified in an irregular structure. In fractograph 17, almost at the end
of the fatigue crack are, coarse striations are found.
Figure 30 presents two examples of features identified in the SEM analy-
sis of the aluminum 6061-T6 MIG welded specimen. In fractograph 6, be-
sides some fatigue striations, some gaps in the material continuity are iden-
tified. These gaps were probably formed in the heating and cooling during
welding leading to the appearance of internal cracks. These gaps are mi-
cro-cracks that can be signs of small hot cracks. Hot cracks can be formed
in AlMgSi fusion welds when the low melting MgSi eutectic in the grain
boundaries remains liquid during weld solidification and subsequent mate-
rial shrinkage [5]. Heat treatable aluminum alloys are sensitive to hot short
cracking, which results form HAZ liquidation during welding [20]. This
specimen presents no less micropores than the aluminum 6082-T6 MIG
specimen. In fractograph 7 striations with a V shape orientation were
found. The image shows the random distributions that can be found in
striations if the fatigue crack is observed with high magnifications.
Figures 31 and 32 present the striation spacing versus crack length for
the aluminum 6082-T6 and 6061-T6 MIG welded specimens.
(a) Fractograph 6, crack length of (b) Fractograph 7, crack length of

0.611mm 1.794mm
Fig. 30 SEM analysis of the AA6061-T6 MIG welded specimen.
Fig. 31 Fatigue striation spacing vs. crack length AA6082-T6 MIG welded speci-
men.
Fig. 32 Fatigue striation spacing vs. crack length AA6061-T6 MIG welded speci-
men.
7.3 FS welded specimens, striations spacing
The specimens were fatigue tested at a maximum stress of 70% of their

yield stress (98.4MPa for the aluminum 6082-T6 and 111.1MPa for the
aluminum 6061-T6). The 6082-T6 specimen presented a fatigue life of
258827 cycles and the 6061-T6 specimen had a fatigue life of 173499 cy-
cles. For the aluminum 6082-T6 the fatigue area has 20.4mm2 and a crack
length of 10.29mm. For the aluminum 6061-T6 the fatigue area has
24.8mm2 and a crack length of 12.50mm.
The friction stir welded aluminum 6082-T6 specimen has an irregular
structure in its entire fracture surface preventing the identification of
measurable striations.
For the aluminum 6061-T6 friction stir welded the fatigue crack propa-
gated near the shoulder edge. Despite the heterogeneity found in the frac-
tured surface the striations measurement process was feasible. In fracto-
graph 5 presented in Figure 33 several striations sets with different
orientations are identified. In fractograph 21, almost at the end of the fa-
tigue crack, are found coarser striations.
Figure 34 presents the striation spacing versus crack length for the alu-
minum 6061-T6 friction stir welded specimen.
Fatigue of FS, LB &
MIG Welds of AA6061-T6 &
AA6082-T6 109

0.877mm 9.544mm
men.
Fig. 34 Fatigue striation spacing vs. crack length AA6061- T6 FS welded speci-
men.
8. Conclusions
Sound welds have been obtained using MIG and FSW for 3mm plate
thickness of two aluminum alloys 6082-T6 and 6061-T6. Tensile testing of
the weld joints and base material produced adequate tensile strength val-
ues.
Yield and rupture stress of friction stir welded and MIG welded speci-
mens are lower than for base material. All welding processes lead to a de-
crease of the material mechanical properties, more pronounced in the MIG
specimens. Detailed hardness examination revealed lower hardness values
in the MIG welded specimens.
When compared with the friction stir welds and MIG welds, the laser
beam welds presented the smaller area (width) affected by the welding
process.
The friction stir welded 6061-T6 specimens presented lower lives than
the friction stir 6082-T6 specimens when tested at stresses lower than
130MPa. Fatigue scatter is somewhat higher in the MIG welded speci-
mens. Its fatigue lives are lower than friction stir welded specimen. The
MIG welded 6061-T6 specimens presented higher fatigue lives than the
MIG 6082-T6 specimens.
The general microstructure was in good agreement with previous pub-
lished researches. In friction stir welding the nugget has recrystallized
grains smaller than the base material. A SEM analysis of fatigue striation
measurements was performed for fatigue specimens of both welding proc-
esses and fatigue crack growth rate can be estimated from using these
measurements.
Acknowledgments
The work was partially supported by PhD scholarship FCT

SFRH/BD/19281/2004, FP6 project AST3-CT-2004–516053 and FCT
PTDC/EME-TME/66362/2006. The collaboration of Rui Louro of ISQ,
Miguel Figueiredo, Sérgio Tavares and Rui Silva of FEUP is gratefully
acknowledged.
References
[1] Thomas W. Nicholas E. Needham J. Murch M. Templesmith P. and Dawes

C. Improvements Relating to Friction Welding, TWI, Editor. 1992.
[2] PLC Air Products, Welder’s Handbook, For Gas Shielded Arc Welding, Oxy
Fuel Cutting & Plasma Cutting 3rd edn ed, ed. A.P. PLC. 1999.
[3] Wahab M. Painter M. and Davies M. The prediction of the temperature dis-
tribution and weld pool geometry in the gas metal arc welding process. Jour-
nal of Materials Processing Technology, 77(1-3) (1998) 233-239.
[4] Behler K. Berkmanns J. Ehrhardt A and Frohn W. Laser beam welding of
low weight materials and structures. Materials & Design, 18(4-6) (199) 261-
267.
[5] Ericsson M. and Sandstrom R. Influence of welding speed on the fatigue of
friction stir welds, and comparison with MIG and TIG. International Journal
of Fatigue, 25(12) (2003) 1379-1387.
[6] AB ESAB. ESAB Welding Handbook -Filler Materials for Manual and
Automatic Welding. 2003. Goteborg Sweden.
[7] Scialpi A. de Filippis L and Cavaliere P. Influence of shoulder geometry on

microstructure and mechanical properties of friction stir welded 6082 alu-
minium alloy. Materials & Design, 28(4) (2007) 1124-1129.
[8] Svensson L. Karlsson L. Larsson H. Karlsson B. Fazzini M. and Karlsson J.
Microstructure and mechanical properties of friction stir welded aluminium
alloys with special reference to AA 5083 and AA 6082. Science and Tech-
nology of Welding & Joining, 5(5) (2000) 285-296(12).
[9] Dickerson T and Przydatek J. Fatigue of friction stir welds in aluminium al-
loys that contain root flaws. International Journal of Fatigue, 25(12) (2003)
1399-1409.
[10] Harris D. and Norman A. Properties of Friction Stir Welded Joints: A Re-
view of the Literature. in EUROSTIR. 2003. Progress report presented at the
6th PSG Meeting.
[11] Nicholas E and Kallee S. Friction Stir Welding - a decade on. in IIW Asian
Pacific International Congress. 2000. Sydney.
[12] Hong S. Kim S. Lee C. and Kim S. Fatigue crack propagation behavior of
friction stir welded Al-Mg-Si alloy. Scripta Materialia, 55(11) (2006) 1007-
1010.
[13] Weston J. Jones I. and Wallach E. Laser welding of aluminium alloys using
different laser sources, in CISFFEL 6 - International Conference on Laser
and Power Beam Welding. 1998: Toulon, France.
[14] Grong Ø. Metallurgical Modelling of Welding. second edt ed. Materials
Modelling Series, ed. T.I.o. Materials. 1997, London.
[15] ASTM, Constant Amplitude Axial Fatigue Tests of Metallic Materials. [16]
ASM, ASM Handbook: Metallography and microstructures. Vol. 9. 1995,
ASM, Materials, Park, OH.
[17] Woo W. Choo H. Brown D. Vogel S. Liaw P. and Feng Z. Texture analysis
of a friction stir processed 6061-T6 aluminum alloy using neutron diffrac-
tion. Acta Materialia, 54(15) (2006) 3871-3882.
[18] Liu G. Murr L. Niou C. McClure J. and Vega F. Microstructural aspects of
the friction-stir welding of 6061-T6 aluminum. Scripta Materialia, 37(3)
(1997)355-361.
[19] Moreira P. de Figueiredo M and de Castro P. Fatigue behaviour of FSW and
MIG weldments for two aluminium alloys, Theoretical and Applied Fracture
Mechanics, 48(2) (2007) 169-177.
[20] Davis J. Aluminum and aluminum alloys. ASM Specialty Handbook. 1993,
Materials Park, OH: ASM International. 784.
Fatigue Damage from Surface to Bulk
C. A. Rodopoulos
Laboratory of Technology and Strength of Materials, Department of Me-
chanical Engineering and Aeronautics, University of Patras, Panepistimi-
oupolis, Patras 26500, Greece
Email: Rodopoul@mech.upatras.gr
Abstract
The work continuous with the Fatigue Damage Map Method (FDMM) that
attempts to unify the propagation rates with the concept of the effective
plastic strain. Reference is made to the local strain characterizing the crack
tip modified to include the variation of flow resistance as a function of
crack length. Comparison with experimental data of multiscale crack
growth rates is provided.
1. Introduction
Two mechanical properties are strongly affiliated to fatigue damage (FD);

these are the maximum tensile part of the fatigue or endurance limit of
smooth specimens for fully reverse loading, ıFL,MAX(R=—1), and the cyclic
yield stress, V yc . The first defines the maximum permissible tensile stress
below which, an inherited or newly formed fatigue crack constrained by
the first major microstructural barrier, will never experience an incre-
mental tendency of its growth rate and hence it will reach propagation
rates equal or smaller than one Burger’s vector (crack arrest), see Figure 1.
Here, the size of this crack cannot be larger than the largest grain found on
the surface. The condition stands for the natural tendency of the large
grains to exhibit maximum deformation.
The cyclic yield stress on the other hand, quantifies as the bulk resistance
of the material towards plastic flow. Plastic flow is a rather precise term
because its rate can either increase or decrease for the same amount of en-
ergy per cycle depending on the nature of the material for hardening or
softening. Inevitably, plastic flow saturation will be reached indicating that
the majority of grains in the particular plane have achieved plastic defor-
mation (Figure 2). The choice of fully reverse loading should not be ig-
nored since it contains vital information regarding internal materials’ plas-
tic flow controllers (Bauschinger).
The fundamental difference between the fatigue limit and plastic flow re-
sistance is scale. The first is related to the response of the surface and the
114 C. A. Rodopoulos
second to the response of the bulk to cyclic loading. In the text that follows
characterisation of such a difference is attempted.
Fig. 1 Arrested fatigue crack by grain boundary in 2024-T351 aluminium alloy.
Fig. 2 Regular plastic deformation in Super Pure Aluminium after 800 cycles at
0.1% plastic strain.
2. An Experimental Fact
To facilitate the distinction between surfaces and bulk it is important to jog

the memory of the reader to a rather well known experimental fact. In a
specimen failed by natural initiation of fatigue cracks, striations are only
found in the bulk of the specimen and never on the surface. Herein the sur-
face defines the region near the crack nuclei. Figures 3a, b are showing
such a case. The reader should refrain from associations with pre-cracked,
especially through-thickness specimens. An explanation is provided later

in the text.
(a) Lack striation (b) Striation in bulk material

Fig. 3 Lack of striations at the region of crack nuclei in a 7150-T651 aluminum
alloy and well defined striations inside the bulk of the material.
To appreciate the phenomenon it is imperative to conceptualize the

mechanism of fatigue striations and identify those parameters controlling
it. From the several models proposed over the years, to explain the mecha-
nism of fatigue striations, those in [1, 2] are the most prominent. Both pos-
tulate that there is a continuous, following the loading waveform, blunting
and re-sharpening of the crack front due to a symmetrical alternating shear
on two dominant slip planes. Similarly the work in [3] used symmetrical
shear stresses to explain crack propagation through dislocation generation
and movement. In all the above cases, a symmetrical deformation field is
needed to dictate for the steady-state development of fatigue damage. De-
spite the fact that the term steady-state is somehow loosely defined outside
the strict boundaries of elasticity, it can be simplistically linked to the pro-
portionality instituted between the monotonic and cyclic crack tip plastic-
ity. Where the first, is the product of single sign dislocations, while the
second, is the result of dislocation dipoles (two opposite signs). It is there-
fore important to associate the lack of fatigue striations found near the re-
gion of the crack with some sort of absence of symmetrical deformation.
3. Defining the Absence of Symmetry
Grain boundaries and other microstructural barriers (martensitic laths, eu-

tectoid colony, ferrite-pearlite boundary, and so on) are known to con-
straint the spread of crack tip plasticity (the term tip is introduced because
it is not known whether barriers can homogeneously constrain the volu-
metric crack plasticity) and hence the propagation of the crack [3, 4]. The
term “spread” is rather precise since it reflects the tendency of the crack tip
plasticity to expansively broaden if it was not for some internal resistance
by the material. In the classical fracture mechanics this is expressed by the
yield stress of the material. Herein the reader should question whether the
macroscopic yield stress contains sufficient scale range to be portrayed as
the internal material’s resistance. It is therefore rational to assume that the
size of crack tip plasticity is more likely to be controlled by microstruc-
tural barriers rather than by a more wide-ranging interaction of surface and
bulk displacements. The strength of these barriers and especially of the
grain boundaries is encapsulated in the parameter k of the Petch-Hall equa-
tion [5]. Its value depends on the crystallographic system, solute additions
(dislocation pinning) and work hardening (dislocation interaction) [5]. In a
number of modelling works [6,7], it was reported that there is critical sur-
face crack length or a critical surface crack tip plasticity size, considering
that its end is bounded by the barriers, beyond which the near-tip strain
field is strong enough to: a) underpin dislocations through the boundary; b)
activate a new dislocation source in an adjacent grain and c) to produce an
internal crack in an adjacent grain which will then coalescence with the
main crack [8]. In either case, once the critical crack length is reached, the
constrained crack tip plasticity expands and the crack is now partially free
to propagate until the process repeats itself. Such rationale explains the
ladder type propagation of short cracks, as seen from the surface, Figure.4.
Fig. 4 Typical propagation behaviour of short crack in 2024-T351.
Yet, under the assumption of coherent crack growth, a homogeneous

spread of crack tip plasticity in the adjacent grains along the crack front is
required. The above comes against differences in the crystal orientation,

size and shape of the adjacent grains, angle distribution of the grain
boundaries and creates an additional difficulty towards the homogeneous
spread of crack tip plasticity [9]. Armstrong et al in 1962 [10], initially
tackled the phenomenon also known as slip mismatch, in terms of the
strength of the grain boundary to constraint plasticity and by Navarro and
de los Rios in terms of crack arrest through the term “grain orientation fac-
tor” [11]. A typical depiction of the slip mismatch is shown in Figure 5.
(a)Crack in large grain (b) Crack triples(23 grs) (c) Crack covers 60 grs
Fig. 5 In sequence crack in solid line occupies the largest surface grain, the region
with tripling surface length with crack tip plasticity zone (more than 23 grains)
and a crack length 6 times its initial size covering more than 60 grains.
As is easily understood, the rate, describing the dynamic probability of

engulfing a “new/different” grain, will eventually become infinitesimal in-
dicating conditions of fully polycrystalline behaviour (the bulk mechanical
properties are in effect). Slip mismatch is known to follow a rectangular
hyperbolic distribution function reflecting the rate of the number grains
engaged by crack tip plasticity per unit crack length (slip mismatch rate).
Of course it is easily understood that the number of cycles required to pro-
duce the unit crack length is inversely proportional to crack length. In sim-
ple terms attendance of fully polycrystalline behaviour signifies that the
average value of the slip mismatch rate remains the same. In general, the
grain orientation factor is expected to define saturated polycrystalline be-
haviour of the material when asymptotically reaching values of 3.07 for
FCC (Taylor) and 2.0 for BCC (Sachs) structures. Without doubt, slip
mismatch is a rather complex phenomenon to be modelled in terms of fa-
tigue since additional parameters, like grain boundary orientation, micro-
texture, favourable slip planes, slip behaviour, grain size distribution, and
so on, are actively present [12, 13]. Yet, their effect is somehow included
into the Kitagawa-Takahashi or crack arrest diagram, see Figure 6.
Fig. 6 Experimental crack arrest data from 2024-T351 at stress ratio 0.1.
Herein effects coming from the dynamic interaction between crack

lengths, far-field stress, history of crack tip plasticity and crack shapes are
integrated into a single response controlling crack growth. To demonstrate
the difficulty in attaining the shear stress symmetry, postulated by elastic-
ity, it is imperative to engage a graphical narrative. Let us consider a short
2D crack as appears in Figure 5. When the crack is microstructurally short
(Figure 7a) coherent crack growth emanating from the symmetrical alter-
nating shear is possible since both component of stress stretch within the
same grain. The process can explain the plateau found in the crack arrest
data for crack lengths smaller or equal to the largest grain. Herein the grain
orientation factor is considered as unity. Expansion of the crack beyond
the first grain (Figures 7b,c) defines conditions for unsymmetrical alternat-
ing shear since the dominant slip planes can be found in more than one dif-
ferent grains. This asymmetry tends to modify the crack shape and particu-
larly its aspect ratio. The process its energy consuming and can cause
changes in the crack path or bifurcation. If however the crack shape is to
be maintained, the dislocation density, the parameter can be used to define
the level of alternating shear, needs to be comparable. The process necessi-
tates the deceleration of the crack tip as that approaches the grain boundary
and the continuous increase in dislocation density. The process will con-
tinue until the dislocation density in the two slip planes is high enough as
to be considered comparable. Once done, the crack will accelerate and the
process repeats itself. The phenomenon manifests itself by the saw cut
crack growth behaviour as seen in Figure 4. Of course it is easily under-
stood that the difference in the dislocation density as this defines the un-
symmetrical alternating shear, will end once the probability of the crack tip
slip planes to enter a largely different grain diminishes, Figure 7.
Fig. 7 Schematic representation of the unsymmetrical and symmetrical alternating

shear.
To tackle the problem is imperative to consider that every polycrystal

acts as an aggregate with grain boundaries providing internal strengthen-
ing. The notion immediately points towards the know Petch-Hall equation.
Herein, the flow resistance of the material is known the increase inversely
to the average grain size. The average grain size is a well-known material
property and yet conceals a number of important elements.
4. Scaling Fatigue Damage – Microstructural Worthiness
In [14] it was suggested that in order to identify the relationship between

the bulk yield stress of polycrystals and the Petch-Hall equation there is a
minimum volume (number) of grains that should be used to express the
grain size distribution. Considering what is already known in statistics, it
was suggested that since a polycrystal behaves as an aggregate there is a
critical number of grains necessary to accurately identify a selected grain
size distribution and extract average values. If the number is smaller than
the critical, the bulk yield stress will show large scatter. Expanding the
above notion in the case of fatigue damage such number will depend on
the size of the crack tip plasticity. Hence, it can be postulated “the plastic
flow resistance depends on the number of grains sampled by the perimeter
of the crack (in the case of crack arrest-including the history of plastic flow
resistance when the crack was propagating as in threshold tests) or by the
crack tip plasticity area (in the case of present crack growth)”. The above
points towards the following rationale. The flow resistance of the material
depends on the number of grains sampled by the crack tip plasticity and
has the tendency to increase to a saturated value (bulk yield stress), once
the number is large enough as to qualify for average value. Herein, it is
important to note that the crack will follow a particular route, laid by the
largest possible grains within the spectrum of available paths defined by
the maximum stress level. As a result the flow resistance as a function of
crack length will appear below that given by its bulk value. It is also im-
perative to consider that high stress levels will endorse the crack with en-
ergy reserves, which associated with the low flow resistance, will allow the
crack to propagate considerable faster compared to a value calculated via
the symmetry of alternating stress. In addition, with the crack having to
choose from a large spectrum of crack paths the potential scatter in the
number of cycles to failure will be low. At low stress levels the mechanism
is somehow different. First the number of available paths is small while
they are not necessarily the easiest. Secondly the energy reserves are just
sufficient (for stress levels slightly over the fatigue limit) and hence the
crack can experience significant delays in its growth rate.
Based on the above, it is reasonable to ascertain that the crack arrest test
represents the minimisation of crack tip plasticity starting from different
values of flow resistance especially. Its profile is defined by the grain size
distribution since the latter defines the available crack paths. To better un-
derstand the influence of grain size on crack paths it is necessary to keep in
mind that fatigue damage will always start from the largest possible grain
found on the surface. Progression will take place through a path, which al-
lows growth of the crack with minimum energy needs. Therefore it is not
difficult to understand that the crack will follow a path laid by the largest
possible grain allowed by the material. For this reason it is imperative to
examine some statistical distributions of grain size. Figure 8 shows the
grain size distribution of 7150-T651 aluminium alloy and low carbon steel.
The analysis reveals, for the aluminium alloy, a maximum grain size of
132Pm, an average grain of 60.1Pm and standard deviation of 26.4Pm.
Whether the 132Pm grain will accommodate the crack origin is a question
having to do with its probability and the size of the specimen/component.
In a more broadly accepted case, it is statistically correct to consider that
the spectrum of crack paths, bringing the range-average of the number of
grains embedded inside the area of crack tip plasticity from its maximum
value to its average, is controlled by the standard deviation. Here it is im-
portant to note that BCC structures are exhibiting smaller standard devia-
tion in their grain size than FCC structures. A comparison between Figures
8a and 8b reveals the case. A theoretical analysis was published in [15].
(a) 7150-T651
(b) Low carbon steel
Fig. 8 Grain size distribution.
The rationale mentioned earlier motivated the work in [16] which sug-
gested a simplified solution based on three engineering boundary condi-
tions for the case of surface crack arrest defined by the following equation,
ª 2V max FL( R 1 )V UTS º

« » V1
1 mi ¬« 2V UTS 1 R 2 1 R V FL R »
1 ¼
'V arrest d V 1 , -f R 1
Y m1 2a / D
(1)
where Y is the crack correction factor to be used only when the crack has
LEFM conditions, VmaxFL(R=-1) is the maximum tensile part of the fatigue
limit for fully reverse loading, VUTS is the ultimate tensile strength, R is the
stress ratio (far-field stress ratio between the minimum and maximum
stress in a predetermined loading cycle), V1(a) is the crack closure stress as
a function of the crack length a, D is the average grain size and mi/m1 is
the grain orientation factor. The subscript i defines the number of half
grain constituting the fatigue damage (crack length and crack tip plasticity,
see [16]). The first engineering condition considers that the orientation fac-
tor of the first grain is always set at 1, to encapsulate the high probability
of finding a favourably oriented grain for slip along the surface. This trans-
lates into,
mi
1 for i 1, 'V arrest V max FL R 1
(2)
m1
The second condition considers that polycrystalline saturation behaviour

should take place, at the latest, at the point where crack arrest is deter-
mined by small scale yielding conditions and hence the long crack thresh-
old is in effect. The above is converted into,
mi VY
3.07 (FCC) or 2.0 (BCC), 'V arrest d (3)
m1 3
The third condition protects the physical tendency of the crack arrest
capacity of the material to exhibit maxima in the region of microstructural
short cracking (the fatigue limit is the highest permissible crack arrest
stress). The conditions is written as,
i t 1, 'V arrest d V max FL R 1

(4)
VY
In Eq. (3) the condition 'V arrest d , where Vy stands for the mono-
3
tonic yield stress, implies the truthful application of LEFM to describe the
elastic stress field ahead of the crack tip with an error not exceeding 7%
[17]. Incorporation of such boundary condition into Eq. (1) yields,
mi
3.07 (FCC) or 2.0 (BCC),
m1
(5)
81V max FL R=1 2 36V max FL R=-1 2
it (FCC) or i t (BCC)
VY 2 VY 2
where i in the case of crack arrest is the crack length, i=2a/D .
Based on the above, the following equations have been determined [16],
§ mi · 2.31
¨ ¸ 3.07 1 (6)
© m1 ¹ FCC § 2a · 0.34
¨1 0.05 ¸
© D¹
for Face-Centred Cubic (FCC) crystal structures, and
§ mi · 1.2
¨ ¸ 2.0 1
© m1 ¹ BCC § 2a · 0.34
¨1 0.05 ¸
© D¹ (7)
for Body-Centred Cubic (BCC) metals.
5. The KT Diagram
The above passage points towards a realisation of the K-T diagram in

terms of the grain size distribution, as shown in Figure 9. Herein, the first
part denoted by the fatigue limit is controlled by the largest grain found on
the surface. The transition towards the threshold condition is controlled by
the distribution of the grain size which directly affects the probable num-
ber of grains sampled by the crack perimeter of a particular geometric
shape. In other words, the number of grains sampled within this area is be-
low the critical one expressing the material’s bulk flow resistance. Finally,
threshold stress through the use of the bulk property, Vy or Vcy, implies the
use of the average grain size taken from a number of grains exceeding the
critical value. The parameter Vcy stands for the cyclic yield stress.
Fig. 9 Schematic showing the Kitagawa-Takahashi crack arrest diagram influ-

enced by the grain size distribution incorporated into the crack length.
Modeling the effect of grain size on the crack arrest diagram requires ex-
tensive analysis of three critical features. It is known that different materi-
als, not just those of a different crystallographic structure, exhibit a broad
spectrum of grain size distributions types. The second feature emanates
from the sampling assortment. In other words, should it be a particular di-
rection for sampling probabilities? The third case comes from the defini-
tion of a condition to portray the critical number of grains, which directly
incorporate crack geometry effects. Based on the above, the previous
boundary conditions need to be redrawn,
CONDITION 1:
mi 2a
1 for i 1, i= o 'V arrest V max FL R 1
(8)
m1 Dmax
CONDITION 2:
mi 2a Vy
3.07( FCC ) or 2.0( BCC ), i o 'V arrest d (9)
m1 D 3
or, to include the elasto-plastic behaviour of the material,
mi 2a Vc
3.07 (FCC) or 2.0 (BCC), i= o 'V arrest d Y (10)
m1 D 3
The above yields,
max 2 max 2
2a 81V FL R=1 2a 36V FL R=-1
t (FCC) or t (BCC) (11)
D V cY 2 D V cY 2
and CONDITION 3:
81V max FL R=1 2 2a

1 i d c 2
, i= , Dmax ! DZ t D
V Y DZ
(12)
V cy
o V max FL R 1
! 'V arrest !
3
for FCC and
36V max FL R=1 2 2a

1 i d c 2
, i= , Dmax ! DZ t D
V Y DZ
(13)
V cy
o V max FL R 1
! 'V arrest !
3
for BCC. Note that DZ is the average size of the number of grains sampled
within a selected probability range and intercepted by the crack perimeter.
Solution of the above boundary conditions in terms of constitute equations
(no fitting exercises have been performed to identify their interaction and
the system is solved in a singular manner) for a variety of crack lengths
and mechanical properties, resulted into,
§ mi · 2.31
¨ ¸ 3.07 1
(14)
© m1 ¹ FCC § 2a · 0.34
¨ 1 0.05 ¸
© Dz ¹
§ mi · 1.2
¨ ¸ 2.0 1
(15)
© m1 ¹ BCC § 2a · 0.34
¨ 1 0.05 ¸
© Dz ¹
The parameter Dz can either be extracted through a generalised function
to describe different types with some degree of error but maintaining its
nature as to be hold under the same equation umbrella, or, via discrimina-
tion of materials into several types of probability density functions at the
expense of a universal nature.
6. Crack Propagation Rate
In a number of works [17-20], it is known that the number of available dis-

location defining plastic deformation is proportional to the grain size. Con-
sidering that the propagation rate of the crack can be described by the
elasto-plastic crack tip opening displacement, Gtipѽ as,
da
IG tup (16)
dN
with I a dimensionless parameter representing the fraction of dislocations

in the slip band participating in the process of crack extension and takes
values between 0 and 1, it is easily understood that a crack path engaging a
number n of the largest possible grains, will endorse the crack with the
highest possible propagation rates. In other words, the energy needs of the
crack per unit new surface will be kept at minimum. The likelihood of
such potential optimum condition is controlled by the grain size distribu-
tion itself and especially by the probability value of the largest grain. To
incorporate the potential of the largest grain to self determine the number
of grains sampled by the crack perimeter, the parameter n is written as,
Pc
nt (17)
Dmax
and the proportion or right tail estimate of the size distribution is,
n Pc
t (18)
N NDmax
The parameter Pc stands for the crack tip perimeter. Determination of the
value that satisfies the proportion value, denoted by Dz, is achieved by,
Dmax
Pc
³
Dz
f ( D )dD |
NDmax
(19)
Since Dz is now known, a range-average (or range mid-point) needs to be

calculated. Such value, defined as DZ is given as,
§ P ·
Dmax ¨ c 1¸ Dz
Dz © Dmax ¹ (20)
Pc
Dmax
It is worth noting that for values of (Pc, N, Dmax) such as Dz D , Eq.

(20) reduces to Dz D by construction. Such boundary condition signi-
fies the critical number of grains further which, the mid-point equals the
average grain size related to N. In other words, Eq. (20) provides the dis-
tribution of the number of sampled grains towards their critical number
(corresponding to the bulk properties) as a function of the crack shape. So-
lution for the case of the low carbon steel and the 7150-T651 aluminium
alloy is depicted in Figures 10-11 for the grain orientation factor and crack
arrest stress range.
From the above analysis assume that slip mismatch relates the cyclic
yield stress and the fatigue limit, for conditions of crack arrest, through the
relationship,
Fig. 10 Profile of grain orientation factor for 7150-T651 aluminium alloy consid-
ering a semi-elliptical surface crack and a through-thickness crack (thickness
5mm).
Fig. 11 Crack-arrest diagram for 7150-T651 according to Eq.(1) and R=-1.

ª º
« »
« 2.31 »
V arrest FCC flow «3.07 1 » V FL R 1
(21)
0.34
« § 2a · »
« ¨ 1 0.05 ¸ »
¬ © Dz ¹ ¼
ª º
« »
« 1.2 »
V arrest BCC flow « 2.0 1 » V FL ( R 1) (22)
0.34
« § 2a · »
« ¨ 1 0.05 ¸ »
¬ © Dz ¹ ¼
Such case provides a rational explanation towards the transition from

large to small scale yielding conditions as a function of crack length, ge-
ometry and material. It can also be easily seen that in the case of the fa-
tigue limit (arrest of a crack with length no larger than the largest grain, the
value of plastic flow resistance is equal to the far field stress (fatigue
limit).
7. Conclusions
The work considers that the scale effects related to fatigue damage are en-
capsulated inside the crack arrest diagram. Herein, the initial stress plateau
corresponds to the fatigue limit and extends to crack length related to the
largest grain found in the material. From there on there is an irregular but
continuous drop of the stress with crack length. The phenomenon will con-
tinue up to the point where the stress degradation and the size of the crack
obtain some sort of equilibrium denoted by a constant gradient. The gradi-
ent is usually found at –0.5 signifying potential association with conditions
of Linear Elastic Fracture Mechanics or symmetrical alternating shear. Be-
tween these two limiting cases, a lot of research has been denoted to the
explanation of such a potentially complex situation. The concept of short
crack behaviour has emerged and consumed a very large portion of re-
search funding. Of particular interest is the fact that there was and possibly
still is, a severe incision within the research community between those ap-
proaching the problem from the viewpoint of the material and those using
traditional mechanics. The role of the devil’s advocate was played by ex-
perimentalists who found particular cases where neither the case of short
cracking was always identified and neither solutions given by the mechan-
ics was always applicable. This work represents a rather innovative way of
attacking the problem and steams from the fact that the above discrepan-
cies should be somehow explained and incorporated. Taking into account
the role of the material and particularly of the grain size distribution, the
corresponding crack paths simulating the effect of stress and the role of the
crack shape and geometry tends to provide a platform for integration. The
work concludes that the flow resistance of the material which controls the
plastic deformation and hence the available energy which needs to be
minimised in order for the crack to arrest, is not constant and tends to start
from a value being the fatigue limit and saturate to a more macroscopic
value that of the cyclic yield stress with crack length.
References
[1] Laird C. The influence of metallurgical structure on the mechanisms of fa-

tigue crack propagation, In Fatigue Crack Propagation, STP, 415 (1967) 131-
168, ASTM, Philadelphia.
[2] Newman P. Coarse slip model of fatigue, Acta Metall, 17 (1969) 1219-1225.
[3] Sadananda K and Glinka G. Dislocation process that affect kinetics of fa-
tigue crack growth, Phil. Mag, 85(2-3) (2005) 189-203.
[4] Fang B, Hong Y and Bai Y. Experimental and theoretical study on numerical
density evolution of short fatigue cracks, Acta Mech. Sinica, 11(2) (1995)
144-152.
[5] Miller K J. Metal Fatigue-Past, Current and Future, The John Player Lecture,
Proc. IMECHE, London, 205(5) (1991) 291-304.
[6] Petch N J. The cleavage strength of polycrystals, J. Iron Steel Inst, 174 (1953)
25-28.
[7] de los Rios ER, Tang Z and Miller KJ. Short crack fatigue behaviour in a
medium carbon steel, Fatig.Fract.Engng.Mater.Struct, 7(20) 1984) 97-108.
[8] Navarro A and de los Rios ER. Fatigue crack growth modelling by succes-
sive blocking of dislocations, Proc. R. Soc. London, A437 (1992) 375-390.
[9] Li GS. A three-dimensional finite element analysis for a crystallographic
crack near the interface of an incompatible bicrystal, Fatig. Fract. Engng.
Mater. Struct., 16(1) (1993) 21-35.
[10] Peralta P, Dickerson R, Dellan N, Komandur K and Jameel MA. Effects of
local grain orientation on fatigue crack growth in multicrystalline FCC me-
tallic materials, J Engng. Mater. Techn., 127 (2005) 23-32.
[11] Armstrong RW, Codd I, Douthwaite RM and Petch NJ. The plastic deforma-
tion of polycrystalline aggregates, Phil, Mag, 7 (1962) 45-58.
[12] de los Rios ER, Navarro A. Considerations of grain orientation and work
hardening on short-fatigue crack modelling, Phil. Mag. A, 61(3) (1990) 435-
449.
[13] Zhai T, Wilkinson AJ and Martin JW. A crystallographic mechanism for fa-
tigue crack propagation through grain boundaries Acta Mater., 48 (2000)
4917-4927.
[14] Kurzydzsowski KJ. A model for the flow stress dependence on the distribu-
tion of grain size in polycrystals, Scr. Metall. Mat., 24 (1990) 879-883.
[15] Rodopoulos CA, de los Rios ER. Theoretical Analysis on short fatigue
cracks, Inter. J. of Fatigue, 24 (2002) 719-724.
[16] Rodopoulos CA. Evolution of fatigue damage using the fatigue damage map
method, Theo. Appl. Fract. Mech., 45 (2006) 252-265.
[17] Brown M.W Interfaces between short, long and non-propagating cracks, In:
The Behaviour of Short Fatigue Cracks, K. J. Miller and E. R. de los Rios
eds, (Mechanical Engineering Publications, London, (1986) 423-439.
[18] Pelloux RM. Crack extension by alternating shear, Engng. Fract. Mech., 1
(1970) 697-704.
[19] Eastbrook JN. A dislocation model for the rate of initial growth of Stage I
fatigue cracks, Inter. J. of Fracture, 24 (1984) R43-R49.
[20] de los Rios ER, Tang Z and Miller KJ. Short crack fatigue behaviour in a
medium carbon steel, Fatig. Fract. Engng. Mater. Struct., 7(2) (1984) 97-
108.
Microcracking in High Temperature Low Cycle
Fatigue
S. L. Mannan1* and M. Valsan2
1
National Engineering College, Kovilpatti, TN, India
2
Indira Gandhi Centre for Atomic Research, Kalpakkam, TN, India
*Email: Sadari_mannan@yahoo.com
Abstract
At elevated temperatures, fatigue deformation and fracture are influenced

by synergistic actions of several time dependent damage mechanisms.
These include dynamic strain ageing, oxidation, creep and phase transfor-
mations. Studied are the influence of dislocation density, precipitates,
strain rate, temperature and hold time on the high temperature low cycle
fatigue behaviour of austenitic and ferritic steels and their weld joints and
a Nimonic PE16 superalloy. Cycling hardening and softening response in
various alloys would be discussed in terms of accompanying substructural
changes.
1. Introduction
Low Cycle Fatigue (LCF) is an important consideration in the design of

many high temperature operating equipments such as aircraft gas turbines,
heat exchangers and power plant components like steam turbines, pipings
and reactor vessel. In this context, many researchers have attempted the
evaluation and characterization of LCF properties of alloys widely used in
service at elevated temperatures [1-5]. LCF damage occurs in these com-
ponents due to large plastic strains introduced during start-up/ shut-down
operations or power transients. Further on-load periods introduce creep-
fatigue interactions which is more deleterious than either creep or fatigue
acting separately. At high temperatures the fatigue deformation and life are
influenced by several time dependent mechanisms such as differences in
slip mode, dynamic strain ageing (DSA), oxidation, creep and phase trans-
formations. This has been reviewed in detail by several investigators [6-9].
These damage processes can lead to premature failure when compared
with fatigue failure under time-independent conditions. The time-
dependent mechanisms mentioned above may influence the fatigue life ei-
ther synergetically or independently depending on test conditions.
134 S. L. Mannan and M. Valsan
Considered are the factors that influence the formation and growth of
microcracks in high temperature fatigue would be examined by conducting
tests on smooth samples. The role of initial microstructure and the micro-
structure that evolves during testing would be studied by tests on samples
of different size and volume fractions of the Jc precipitates in Nimonic
PE16 Superalloy. Further, the influence of various test conditions like
temperature, strain rate, dwell-time on the formation and growth of micro-
cracks is examined based on tests conducted on this material and on
316L(N) Stainless steel, 316 weld metal and a modified 9Cr-1Mo ferritic
steel [10-20]. In certain situations cyclic deformation led to strain local-
ization resulting in rapid crack initiation and propagation. The mechanisms
affecting crack initiation and propagation are examined to understand
variations in high temperature LCF life as a function of microstructure and
test variables in smooth specimens. A fatigue crack generally initiates at
the surface of the smooth specimen first, grows along crystallographic slip
direction and then crack growth occurs at 90o to the stress axis marked by
fatigue striations. Fatigue failure towards last few cycles is reflected in a
drastic reduction in load drop in the tension part of the hysteresis loop and
cusp formation in the compression part. A 20% drop in the tensile load is
taken as the failure life in the present study. Smooth specimens employed
in this study have been considered to be part of the plastic zone around the
notch in actual components and the fatigue life of the specimens has been
taken as the crack initiation life in actual components [21].
1.1 Material and heat treatment
The materials used in this study are candidate materials for the primary
and secondary sodium circuits of the fast breeder reactors. Nimonic PE-16
finds application as wrapper tubes and clad tubes in fast reactors apart
from applications at high temperatures as a material for aircraft ducting
systems, gas turbine flame tubes, missile hot components and superheater
tubes. LCF studies on this relatively simple precipitation hardened system
(with a low volume fraction, 11% maximum, of J') would also facilitate
fundamental understanding of the operative deformation and fracture
mechanisms and correlation with fatigue life in this class of superalloys.
The prior microstructure of the alloy is chosen with two different J' sizes
namely 18nm and 35nm such that the former favours precipitate shearing
and the latter Orowan looping. Consequently the influence of these on
strain localization and micro-cracking can be evaluated.
316L(N) stainless steel and 316 weldmetal find application in the pri-
mary sodium circuit of the fast reactor as reactor vessel, pipings and heat
exchangers. The choice of this alloy is based on its excellent high-
temperature tensile, creep, fatigue and creep-fatigue strengths in combina-
tion with good fracture toughness and fabricability. 316L(N) has lesser
susceptibility to sensitization and associated intergranular corrosion.
Modified 9Cr-1Mo ferritic steel (with alloying additions of Niobium and
Vanadium and controlled amount of nitrogen) is extensively used as a
structural material at elevated temperatures up to 873 K in fossil-fired
power plants, petrochemical industries and as a material for tubing in the
reheater and superheater portions and as thick-section tube sheet material
in the steam generators of Liquid Metal Cooled Fast Breeder Reactor [22].
High thermal conductivity and low thermal expansion coefficient coupled
with enhanced resistance to stress corrosion cracking in steam-water sys-
tems are important considerations in the selection of this steel for these ap-
plications. The alloy also exhibits good weldability and microstructural
stability over very long periods of exposure to high temperature service
conditions.
These alloys were tested under different temperature and strain rate
combinations including those which promote strain localization and en-
hanced transgranular and intergranular cracking. Further intergranular and
transgranular cracking assisted by creep and oxidation are evaluated by
conducting tests with hold times at tension and compression peaks in the
strain cycle.
1.1.1 Nimonic PE-16 Superalloy
Nimonic PE-16 superalloy is a nickel-iron base superalloy that is strength-

ened by chromium and molybdenum in solid solution and also by ordered
face centered cubic Ni3 (Al,Ti) precipitates. The chemical composition of
the alloy is given in Table 1. The alloy in the form of 20 mm diameter rods
was solution treated at 1313 K for 4 hours. Solution annealing was found
to dissolve majority of the preexisting carbides and all the J' except some
TiCN precipitates. This was designated as Microstructure A.
Table 1 Chemical composition of nimonic PE 16 superalloy (in wt%).

Ni Cr Ti Al C Mo Zr S B Mn Cu Fe
43.2 16.5 1.2 1.2 0.07 3.3 0.03 0.004 0.0015 0.04 0.06 33.83
PE-16 alloy is generally subjected to two-step ageing treatments to

achieve the specified mechanical properties. The rationale for the choice of
different temperatures for the intermediate and final ageing treatments is as
follows: the solvus temperature of Jc phase for the PE-16 alloy varies in the
temperature range 1133-1148 K depending on the specific composition of
the alloy [23]. Hence the two intermediate temperatures of ageing were
chosen such that 1173 K is above the solvus temperature and 1073 K is be-
low. Thus ageing at 1073K would lead to the precipitation of both Jc and
carbides whereas ageing at 1173 K would cause the precipitation of car-
bides alone. Intermediate ageing at 1173 K resulted in the precipitation of
MC carbides primarily along the grain boundaries and occasionally within
the matrix. A comparison of the effect of final ageing treatments namely
973 K and 1023 K, on the tensile properties of the alloy after intermediate
ageing revealed that the maximum strength values are attainable at 973 K.
This has been attributed to the maximum volume fraction of Jc formed at
973 K compared to that obtained at 1023 K for the same duration [24]. Op-
timum tensile properties at room temperature have been obtained by heat
treatment 1313 K/4h + 1073 K/2h + 973 K/16h ( yield strength :556 MPa,
UTS : 983 MPa, % elongation : 26).This was designated as Microstructure
B which consisted of maximum volume fraction of J' precipitate (0.11). J'
precipitate was uniformly distributed in the matrix with a size of 18nm.
Aging at 1073 K for 2 hours had also led to the precipitation of M23C6 type
carbides in both intra and intergranular locations. Another batch of the so-
lution annealed samples was subjected to a double aging treatment of
1173K/1h followed by 1023 K/8 h. This was designated as Microstructure
C which consisted of intra and intergranular MC and M23C6 type carbides
and J' of 35nm diameter, 0.08 volume fraction.
1.1.2 316L(N) Stainless steel and welds
The chemical composition of 316L(N) SS and 316 weld metal used in this
investigation is given Table 2. 316L(N) material obtained in mill-annealed
condition was given a solutionising treatment at 1373 K/1h followed by a
water quench. This heat treatment yielded an average grain size of 85 Pm.
Evaluation of elevated temperature LCF behaviour of 316L(N) stainless
steel has received much attention in the recent years [12,15,19,25-31]. Ni-
trogen addition has been reported to be beneficial, and the LCF life has
been found to saturate around 0.12% N [28].
Table 2 Chemical composition of 316L(N) base metal and 316 weld metal (in
wt%).
Alloy C Mn Ni Cr Mo N S P
316L(N) 0.002 1.75 12.0 17.0 2.4 0.078 0.002 0.023
Base
316 weld metal 0.06 1.42 11.9 18.8 2.0 0.05 0.01 0.009
Welds are the weak links in structures. Most of the service failures are
found to occur either in the HAZ or in the weld metal, which are more fre-
quently associated with the presence of defects or microstructural inho-
mogeneities compared to the base metal. In austenitic stainless steel welds,
ferrite introduced to reduce their tendency to hot cracking and micro fis-
suring, transforms to a hard and brittle phase known as phase, when these
materials are exposed to elevated temperatures (773-1173 K) for extended
periods of time, leading to low ductility creep ruptures when sufficiently
high stresses are applied at elevated temperatures. Weld metal specimens
were machined from weld pads prepared by shielded metal arc welding
process using 316 electrodes. X-ray radiography was used for assessing the
soundness of the welds followed by G-ferrite measurements using a
magne-gauge. The average ferrite number in the weld metal was between 4
and 5.
1.1.3 Modified 9Cr-1Mo Steel
Modified 9Cr-1Mo steel was obtained in the form of hot-forged rods of 70

mm diameter and also in the form of extruded tube of 30 mm wall thick-
ness. The chemical compositions of these two product forms are given in
Table 3. The normalizing treatment of this steel was carried out at 1313 K
for 1 hour plus air cooling and tempering was done at 1033 K for 1 hour
plus air cooling which resulted in a tempered martensitic structure.
Table 3. Chemical composition of modified 9Cr-1Mo ferritic steel (in wt%).

Product Cr Mo C V Nb Mn Si S P N Ni Fe
Forged bar 8.720.90 0.096 0.22 0.08 0.46 0.32 0.006 0.012 0.051 0.1 Bal.
Extruded pipe8.2 0.92 0.11 0.13 0.09 0.48 0.3 0.005 0.018 0.051 0.12 Bal
2. Experimental
2.1 Low cycle fatigue testing
Fully reversed total axial strain controlled LCF tests were conducted at
773, 823 and 873 K in air on the 316L(N) base metal, 316 weld metal and
modified 9Cr-1Mo specimens using a servo hydraulic machine equipped
with a radiant heating facility. Cylindrical smooth specimens of 25mm
gauge length and 10mm diameter were used in these tests. Tests were car-
ried out with total strain amplitudes in the range r0.25% to r1.0% with a
strain rate of 3u10-3 s-1. Temperature effect was evaluated on 316L(N)
stainless steel and Modified 9Cr-1Mo steel in the temperature range 300-
923 K at a strain range of r0.6%. In Nimoinic PE-16 superalloy LCF tests
were conducted at 723, 823 and 923 K at various strain amplitudes. In all
these materials strain rate effects were evaluated in the range 3u10-5 s-1 to
3u10-2 s-1 at a strain amplitude of r0.6%. Creep-fatigue interaction experi-
ments were conducted by introducing tension/compression holds and ten-
sion plus compression holds in the range varying from 1 min. to 90 min. at
873 K at strain amplitudes r1.0% and r0.6%.
2.2 Metallography
The tested samples were sectioned parallel to the loading direction, pol-
ished, etched and examined under an optical microscope. The 316L(N)
base metal and Nimonic PE-16 superalloy were etched using 70% HNO3
while the etching of weld metal was done using a modified Murakami’s
reagent (30g. of KOH, 30g. of K3Fe(CN)6 in 150 ml water) at 363 K for 30
seconds. Fractography of the failed specimens was carried out using a
scanning electron microscope and substructural changes were studied by
transmission electron microscope. Samples for transmission electron mi-
croscopy (TEM) for 316L(N) base metal and Nimonic PE-16 superalloy
were obtained from thin slices cut at a distance of 3 mm away from the
fracture surface. These samples were first mechanically polished down to
250Pm, and then electropolished in a solution containing 20% perchloric
acid and 80% methanol at 243 K with a d.c. voltage of 10 V. The slip-band
spacing measurements were performed on the longitudinal sections of the
fatigue-tested samples using an SEM or an optical microscope with a
graduated eye-piece depending upon the resolution needed for measuring
the slip band spacing.
3. Results and Discussion
The results of this investigation are discussed under the following sections:
x Influence of microstructure – role of precipitates
x Strain rate and temperature effects
x Creep – fatigue – environment interactions – hold time effects
3.1 Influence of microstructure –role of precipitates
The initial and evolving microstructural changes would be reflected in the

cyclic stress response curves. These changes have a pronounced influence
on the microcracking. In this section, the role of precipitates in influencing
deformation (planar slip versus uniform deformation) and cracking - re-

sulting from slip localization due to shearing of small size J' precipitates
and grain boundary cracking due to slip band impingement on grain
boundaries - would be illustrated by examining results of tests on three mi-
crostructural conditions of Nimonic PE-16 superalloy. The role of carbide
precipitates would also be brought out. The prior microstructure after dou-
ble ageing (Microstructure B) consisted of ordered J' precipitate of average
size 18nm (Fig. 1).
The cyclic stress response was characterized, generally by an initial
hardening to a peak stress followed by a softening regime under all testing
conditions at 923 K (Figs. 2(a) through (c)). The initial hardening during
cycling has been attributed to dislocation-dislocation and dislocation-J'
precipitate interactions. The operative deformation mechanisms are found
to depend on the size of the J': namely, shearing for very fine precipitates
and Orowan looping for coarse precipitates (for J' with d= 35nm). In Mi-
crostructure A the contribution to hardening comes from dislocation – dis-
location interaction. However, when shearing of J' precipitates takes place,
the contribution to hardening is derived also from dislocation accumulation
on J' precipitates initially and order hardening contribution from the pre-
cipitate once it is subjected to shearing. When precipitates are coarse, they
are circumvented by Orowan looping process, as in the case of Microstruc-
ture C [4]. The dislocation-precipitate interaction displaces the dislocations
from the primary slip planes. This process repeats, eventually resulting in
homogenization of slip due to linkup in three dimensions.
Various mechanisms have been presented in the literature to explain the
cyclic softening in precipitation hardened systems. The two mechanisms
which are widely accepted are (1) disordering of J' due to mechanical
scrambling [32,33] and (2) reduction in the size of the precipitate due to
repeated shearing [34]. The mechanism underlying the cyclic softening
observed in fatigue testing of Nimonic PE-16 was examined earlier by the
present authors [11]. Based on TEM studies on samples interrupted to
various fractions of fatigue life, it has been established that the reduction in
the size of J' is responsible for cyclic softening. Figure 3 shows the cyclic
stress response of Microstructure B fatigue tested at 823 K at strain ampli-
tude of r0.6%. TEM microstructure of samples fatigue tested to various
fractions of fatigue life are superimposed in this diagram. These studies
revealed that dislocations are accumulated in the first few cycles as shown
in samples interrupted after 10 cycles representing the hardening stage in
the cyclic stress response and partial slicing of the Jc precipitates are ob-
served after 80 cycles (representing the softening stage). However, after
300 cycles (representing the saturation stage) all the J' precipitates are
completely sliced by the dislocations and spherical particles are no longer

present in the sample. It is observed that the precipitates are completely
offset and they cease to act as obstacles to the motion of dislocations. Thus
cyclic softening is attributed to reduction in the size of Jc.
Fig. 1 Dark field micrograph showing Jc precipitate morphology before testing.
450 Nimonic PE-16

Microstructure A
Tensile stress amplitude (MPa)
400 923 K
350
300
250
Total strain amplitude
200 +0.25%
+0.4%
150 +0.6%
+0.8%
100 +1.0%
0 1 2 3 4
10 10 10 10 10
Number of cycles N
(a) Microstructure A at 923 K
700
Nimonic PE-16
Microstructure B
600
923 K
500
400
300 Total strain amplitude

+0.25%
200 +0.4%
+0.6%
100 +0.8%
+1.0%
0 1 2 3 4
10 10 10 10 10
Number of cycles N
(b)Microstructure B at 923 K
650
Nimonic PE-16
600 Microstructure C
923 K
550
500
450
400
350
300
Total strain amplitude
+0.4%
250 +0.6%
+1.0%
200
0 1 2 3
10 10 10 10
. Number of cycles N
(c) Microstructure C at 923 K
Fig. 2 Cyclic Stress Response of Nimonic PE-16.

Fig. 3 Cyclic stress response of Nimonic PE-16, Microstructure B at 823 K (insert

(a) Dislocation accumulation in slip bands after 10 cycles, (b) dark field micro-
graph of J' after 80 cycles and (c) dark field of micrograph of J' after 300 cycles).
In Microstructure A, softening has been attributed to shearing of fine J'

formed during testing at 923 K. In Microstructure C, the size of the J' par-
ticles (35nm) is too large to permit shearing. Hence, Orowan looping is
predominant. However, the evidences have been reported for the fresh pre-
cipitation of fine J' at 923 K. The lower volume fraction of J' in Micro-
structure C (0.08) has resulted in the fresh precipitation of J' during testing.
The occurrence of superdislocation pairs, Fig. 4, indicates that cyclic sof-
tening in this microstructure is due to shearing of fine J' formed during
testing.
Fatigue life and fracture behaviour depend on the deformation processes
operating during cycling. Fatigue life of Nimonic PE-16 alloy was found
to be influenced by the amount and size of the precipitates (as it deter-
mines the slip processes), Fig. 5. Material in solution annealed condition
(A) showed the highest fatigue life followed by B and then C. The superior
fatigue life of Microstructure A can be attributed to the planar slip mecha-
nism, Fig. 6(a) which promotes increased reversibility in slip processes.
Microstructure A exhibits transgranular fracture, Fig. 6(b) and the reversi-
bility in slip increases the resistance of A to both transgranular crack initia-
tion and propagation and leads to the highest fatigue life.
Fig. 4 Superdislocation pairs in Microstructure C, strain amplitude +1.0% -

indicating precipitate shearing and strain localization.
-2
10
Nimonic PE-16
Temperature: 923 K
Microstructure
Plastic strain amplitude 'Hp/2
A
B
C
-3
10
2 3 4
10 10 10
Number of reversals to failure 2Nf
Fig. 5 Fatigue life plot for Nimonic PE-16 microstructure A, B and C, 923 K.
(a) Planar dislocation arrangement, microstructure A, strain ampl. +0.4%

(b) Transgranular cracking, microstructure A, 923 K, strain ampl. +0.4%

Fig. 6 Microstructure arrangement.
The beneficial effects of planar slip on fatigue life have been attributed
to a high degree of slip reversibility at the propagating crack tip [35]. A
lower crack propagation rate when Jc particles are sheared during deforma-
tion has also been reported in an Fe-Ni-Al alloy [36].
However, the beneficial effect of planar slip may get eclipsed in situa-
tions where the microstructure is such that (1) stress concentrations associ-
ated with the dislocation pileups cause intergranular cracks and (2) local-
ized deformation in widely spaced slipbands cause transgranular cleavage
cracks. In such situations, homogeneity of the slip process will be benefi-
cial to fatigue life. The two-slope behaviour observed in fatigue life plot,
Fig. 5 and cyclic stress strain curve, Fig. 7 for Microstructure B has been
rationalized on the above basis. In Microstructure B, precipitate shearing
mechanism is found to be predominant in planar slipbands, since Jc size is
only 18nm. This enhances the reversibility of the slip processes. Neverthe-
less, over the strain amplitudes investigated, the variation of the slipband
spacing with plastic strain amplitude for Microstructure B shows a devia-
tion from the general trend, especially at low strain regimes, Fig. 8. At low
strains, slip is rather concentrated in specific bands, producing inhomoge-
neous deformation, Fig. 9(a). These slipbands produce pile-ups at the grain
boundaries, causing stress concentration and mixed mode fracture Fig.
9(b). Furthermore, in Microstructure B, the grain boundaries are decorated
with M23C6 carbide precipitates with accompanying reduction in the ductil-
ity. Hence, the stress concentrations associated with the pileups cause car-
bide-matrix interface decohesion and intergranular cracking. At higher
strains, the dislocation substructure becomes more homogeneous because
of the operation of multiple slip. This results in lower stress concentrations
at the grain boundaries and a reduced susceptibility to intergranular failure,

Fig. 9(c).
3
10
Nimonic PE-16
Tensile stress amplitude 'V/2 (MPa)
923 K
Microstructure
A
B
C
2
10
0.1 1
Fig. 7 Cyclic Stress-Strain Curves of Nimonic PE-16, Microstructures A, B & C at

923 K.
The two-slope behaviour has also been reported in Nimonic PE-16 at

room temperature[37]. It has been found that at high strains, intersections
of densely packed dislocations in three different slipbands led to a break-
down of slip planarity, which reduced the stress concentration due to pile-
ups. The two-slope behaviour in the Coffin-Manson life plot has also
4.0
Nimonic PE-16
3.5 Temperature 923 K
-3 -1
Strain Rate: 3 x 10 s
Slip band spacing i (Pm)
3.0
2.5
2.0
1.5
1.0 Microstructure
A
0.5 B
C
0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2
Plastic strain amplitude at peak stress (%)
Fig. 8 Variation of slip band spacing (i) with plastic strain amplitude for Nimonic
PE-16, microstructures A, B and C.
(a) Planar dislocation in microstructure B with, strain amplitude +0.4%
(b) Mixed mode cracking at low strain in microstructure B, +0.4%
(c) Transgranular fracture at high strain for microstructure B, +1.0%
Fig. 9 Microstructure in Nimonie PE-16.

Fig. 10 Facetted Fracture in Microstructure C, 923 K, strain amplitude +0.6%.
been reported in various other alloy systems, e.g. dual -phase steels [38],
aluminium alloy[39] and Nimonic 80 A[34]. In these investigations, the
change in the Coffin-Manson slope has been ascribed to a change in de-
formation mechanism from slipband formation at low strains to twinning
and grain boundary rotation at high strains.
In double aged condition C precipitate shearing with superdislocation
pair was observed (Fig. 4). Further, evidences for in-situ precipitation were
found in several areas [4]. Deformation in Microstructure C is clearly de-
marcated by regions where Orowan looping mechanism (homogeneous de-
formation) is operating and places where strain localization due to precipi-
tate shearing are predominant. Since overcoming of large Jc particles
requires either cross slip or Orowan loop formation, slip is not reversible
as in the cutting process. The Orowan looping process in Microstructure C
is, therefore, consistent with a higher crack propagation rate. In addition,
the microstructure deforms in a heterogeneous manner and exhibits large
slipband spacings, Fig. 8. The wider slipband spacing in Microstructure C
is considered to result from the strain localization effects in regions where
fine Jc precipitates form during testing. Since shearing of the fine Jc is more
favourable mechanism compared to looping of coarse Jc (originally present
in the alloy), slip will be mainly restricted to these soft regions. Accord-
ingly, in Microstructure C, due to the precipitation of fine Jc in widely
spaced regions, it behaves like a heterogeneous matrix consisting of a soft
region and a hard region, and strain will be unequally distributed among
them. This localization is expected to have two consequences: (1) faceted
fracture resulting from channelization of dislocations due to precipitate
shearing, Fig. 10 and (2) de-cohesion of grain boundary carbides caused by
dislocation pileups. Thus, in Microstructure C, secondary cracks are cre-
ated in the matrix and on the grain boundaries, leading to an accelerated
fatigue crack propagation. The reduced fatigue life of the Microstructure
C, is therefore, attributed to the presence of a large number of internal
cracks and reduced slip reversibility associated with the Orowan looping
process.
The effects of testing temperature and frequency/strain rate on the LCF
behaviour of structural alloys used in high temperature applications have
been studied using balanced loading waveforms(equal ramp rates in ten-
sile-going and compressive-going directions) as these parameters influence
cracking behaviour and hence life [10,15-18, 40,41]. With decrease in fre-
quency/strain rate at high testing temperatures deformation mechanisms
such as dynamic strain ageing (caused by solute dislocation interactions)
and oxidation become predominant.
3.2 Strain rate and temperature effects on cyclic properties and fatigue
life
3.2.1 Role of dynamic strain ageing
Cyclic stress response (CSR) as a function of strain rate on 316L(N) SS at

823 K is shown in Fig. 11(a). At 773 and 823 K the degree of hardening
increased significantly with decreasing strain rate and maximum hardening
is observed at the lowest strain rate of 3 u10-5 s-1 , Fig.11(a) [15]. The
negative strain rate stress response observed at these temperatures and
strain rates is also reflected in a negative strain rate dependence of the half-
life stress, Fig. 11(b). Further, negative temperature dependence of CSR in
316L(N) SS in the temperature range 573 K to 873 K is shown in Fig.
11(c). The negative strain rate stress response in cyclic stress response
curves has also been noticed in Nimonic PE-16 Superalloy at various tem-
peratures, (typically at 823 K, Fig. 11(d)). These are the various manifesta-
tions of DSA operating in these alloys.
Dynamic strain ageing was found to influence the deformation and frac-
ture behaviour of the 316L(N) SS and Nimonic PE-16 under specific strain
rate and temperature conditions. The increase in stress response in the
DSA regime has been found to be related to an increase in the dislocation
density in between the slip bands. Figures 12(a) and (b) give a comparison
between dislocation densities observed at higher strain rates and lower
strain rates respectively in between slip bands. Condition conducive to
DSA results in higher dislocation densities, Fig. 12(b). In 316L(N)
stainless steel planar slip bands are observed in the DSA operating range
whereas cell structure is noticed in the non-DSA regime (Figs. 13(a) and
(b) respectively). Enhancement of slip planarity (presence of planar slip) in
the DSA range has also been reported in AISI 310 steel [42]. Further, DSA
also results in increased degree of localization of slip which is reflected in
an increase in slip band spacing. Typical slip band spacing values as a
function of strain rate and temperature as observed in Nimonic PE-16 su-

peralloy for Microstructures A and B are shown in Figs. 14(a) and (b) re-
spectively. It is found that slip band spacing is larger in the DSA regime
(lower strain rates) indicating enhanced slip localization. The solute dislo-
cation interaction during DSA restricted cross slip of dislocations and in-
creased slip planarity. Impingement of planar slip bands on the grain
boundary leads to intergranular cracking typically as shown in Fig. 15. In-
tergranular cracking associated with slip band impingement in the DSA
range was also observed in 304 stainless steel [40]. Enhanced intergranu-
lar cracking associated with slip band impingement led to reduced fatigue
life in the DSA regime for Nimonic PE-16 at low strain rates at 723 and
823 K, Fig. 16. At 923 K oxidation reduces the fatigue life at low strain
rates in this alloy. The role of oxidation in reducing life at 923 K will be
discussed in the subsequent sections.
316L(N) SS
Strain amplitude: + 0.6%
400 Temperature: 823 K
300
Strain Rate
200 -5 -1
3 x 10 s
-4 -1
3 x 10 s
-3 -1
3 x 10 s
-2 -1
3 x 10 s
100
0 1 2 3
10 10 10 10
Number of cycles N
(a) Cyclic stress response at 823 K
316L(N) SS
Strain Amplitude : + 0.6%
400
Half-life stress amplitude 'V/2 (MPa)
350
300 773 K
823 K
873 K
-5 -4 -3 -2 -1
10 10 10 10 10
-1
Strain rate (s )
(b) Variation of half-life stress amplitude for 316L(N) SS
316L(N) SS
-3 -1
Strain rate: 3 x 10 s
400 Strain amplitude: + 0.60%
300
Temperature
298 K
200
573 K
673 K
773 K
873 K
923 K
100
0 1 2 3
10 10 10 10
Number of cycles N
(c) Negative temperature dependence on cyclic stress response
450 Nimonic PE-16

Microstructure A
400 823 K, +0.6%
350
300
250
Strain Rate
-2 -1
200 3 x 10 s
-3 -1
3 x 10 s
150 -4 -1
3 x 10 s
-5 -1
100 3 x 10 s
0 1 2 3
10 10 10 10
Number of cycles N
(d) Cyclic stress of Nimonic PE-16 with microstructure A at 823 K
Fig. 11 Strain rate effect.
(a) Low dislocation density

(b) High dislocation density
Fig. 12 Microstructure at 823 K, 3 x 10-3 s-1 and strain amplitude r0.6%.
(a) 873 K
(b) 573 K
Fig. 13 Well developed cell microstructure at strain amplitude of r0.6%.

5
Microstructure A
Strain amplitude: +0.6%
Slip band spacing i (Pm) 4
1
723 K
823 K
923 K
0
-5 -4 -3 -2 -1
10 10 10 10 10
-1
Strain rate (s )
(a) Microstructure A
5
Microstructure B
Temperature: 923 K
Slip band spacing i (Pm)
4
823 K
0
-5 -4 -3 -2
10 10 10 10
-1
Strain rate (s )
(b) Microstructure B
Fig. 14 Variation of average slip band spacing Nimonic PE-16 with strain ampli-
tude r0.6%.
Fig. 15 Grain boundary cracks associated with impingement of slip bands in PE-
16 Superalloy, Microstructure A, 3 u 10-5 s-1, strain amplitude r0.6%.
1600
923 K Microstructure A
823 K Strain amplitude: +0.6%
1400
723 K
Number of cycles to failure Nf
1200
1000
800
600
400
-5 -4 -3 -2
10 10 10 10
-1
Strain rate (s )
Fig. 16 Variation of fatigue life with strain rate in Nimonic PE-16, strain ampli-
tude r0.6%.
In 316L(N) SS a significant reduction in fatigue life is noticed when

testing strain rate is reduced to 3 u 10-5 s-1 from 3 u 10-2 s-1 at temperatures
773, 823 and 873 K ( Fig. 17). It is noticed from this figure that with in-
crease in temperature, fatigue resistance generally decreases except for
specific cross overs at strain rates < 3 u 10-3 s-1. That is at strain rates be-
low 3 u 10-3 s-1 , fatigue resistance is lower at 823 K compared to that at
873 K. These observations can be correlated with the influence of DSA on
the deformation and fracture behaviour.
The influence of planar slip on cracking can be assessed by measuring
the total length of intergranular and transgranular secondary crack density
in the DSA regime. Crack density (crack length per unit area) as a function
of strain rate at 773 K (typically for 316L(N) SS) is shown in Fig. 18(a)
and crack length distribution in Fig. 18(b). The secondary crack density
was found to be maximum in the DSA regime (low strain rates) for this al-
loy. With decrease in strain rate both intergranular and transgranular crack
density increased indicating accelerated crack initiation and propagation.
Further, very long cracks (a consequence of crack coalescence) are seen at
low strain rates. More number of longer cracks were observed at 823 K at
lower strain rates compared to 873 K. (Figs. 18(c) and (d). The stress con-
centration associated with the intersection of planar slip bands with the
grain boundaries has contributed to the enhanced internal grain boundary
cracking and reduced lives at low strain rates in these alloys. Since crack
coalescence can lead to faster crack propagation, a reduced fatigue life is
observed at strain rate – temperature conditions where long cracks are ob-
served.
1600
316L(N) SS
Strain amplitude: + 0.6%
1200
800
400
773 K
823 K
873 K
0
-5 -4 -3 -2 -1
10 10 10 10 10
-1
Strain rate (s )
Fig. 17 Variation of fatigue life with strain rate in type 316L(N)SS, strain ampli-
tude +0.6%.
It must be pointed out that within the strain rate employed, DSA operates
over the temperature range of approximately 573 to 873 K in 304 SS [40].
DSA enhances the degree of inhomogeneity of deformation during LCF by
solute locking of slow moving dislocations between slip bands [9]. Pre-
sumably, the dislocation velocities inside the slip bands were too high for
dynamic ageing of mobile dislocations to take place and consequently
DSA enhanced the partitioning of strains into separate regions character-

ized by high and low amplitudes of dislocation movement leading to en-
hanced inhomogenity of deformation [43]. Further slow moving disloca-
tions become aged by the solute atmospheres and additional dislocations
are generated to maintain the imposed deformation rate. This process
caused an increase in total dislocation density. The negative strain rate de-
pendence of cyclic stress response over the temperature and strain rate
range where DSA operates results from an increase in total dislocation
density during deformation. The matrix was hardened during DSA, caus-
ing an increase in flow stress needed to impose the same total strain during
successive cycles.
316L(N), 773 K, +0.6%
Strain rate
-2 -1
300 3 x 10 s
-3 -1
3 x 10 s
Crack density (Pm/mm )
2
-4 -1
3 x 10 s
200
100
0
Transgranular Intergranular
(a) Crack density as a function of strain rate at 773 K
316L(N), 773 K, +0.6%

14 Strain Rate
-2 -1
3 x 10 s
12 -3 -1
3 x 10 s
-4 -1
3 x 10 s
10
Number of cracks
0
<1 1-2 2-3 3-4 4-5 5-6 6-7 7-8 >8
Crack length x 100 (Pm)

(b) Crack length distribution as a function of strain rate at 773 K
20
-2 -1
18 3 x 10 s
-3 -1
3 x 10 s
16
-4 -1
3 x 10 s
14
Number of cracks
12
10
0
1 1-2 2-3 3-4 4-5 5-6 6-7 7-8 >8
(c) Crack length distribution as a function of strain rate at 823 K
30
28 -2 -1
3 x 10 s
26 -3 -1
3 x 10 s
24 -4 -1
3 x 10 s
22
20
Number of cracks
18
16
14
12
10
8
6
4
2
0
1 1-2 2-3 3-4 4-5 5-6 6-7 7-8 >8
(d) Crack length distribution as a function of strain rate at 873 K
Fig. 18 Crack distribution for different temperation at strain amplitude r0.6%.
The influence of DSA on number of cycles to transgranular crack initia-

tion and evolution of the microcrack density with number of cycles have
been investigated in Alloy 800H [44]. In this study, a comparison has been
made with test results in vacuum to eliminate the effect of environment on
crack initiation. It has been found that DSA leads to: (a) an overall de-
crease in the number of cycles to crack initiation, (b) an appreciably higher
crack density, and (c) reduction in fatigue life. The enhanced transgranular
crack density observed under DSA conditions may be a consequence of the
increased slip step height formed at the free surface. At medium to high
strain levels, slip steps play an important role in planar slip alloys, in
which fine homogeneous distribution of slip confers optimum resistance to
crack initiation. The increased tendency for transgranular crack initiation
in DSA regime could be due to the increased slip step height at the surface.
Dynamic strain ageing has also been reported to cause a faster reduction
in life by way of rapid crack propagation where transgranular fracture is
dominant. Higher response stresses developed during cyclic deformation
can lead to a larger stress concentration at the crack tip, which would ac-
count for increased crack growth rates and hence a reduced number of cy-
cles in the crack propagation stage [10,15].
3.2.2 Role of oxidation at high temperatures and low strain rates
In order to assess the influence of oxidation at high temperatures and low

strain rates detailed studies have been conducted on Modified 9Cr-1Mo
steel. The dependence of fatigue life on test temperature (300-873 K) at a
constant strain amplitude of r0.6% and strain rate of 3 u 10–3 s-1 is plotted
in Fig. 19. The alloy exhibited a decrease in fatigue life with increasing
temperature. The effect of temperature on life was more pronounced at
lower strain amplitudes, as seen in Fig. 20. Further, the variation of fatigue
life with strain rate showed a maximum at the intermediate strain rate of 3
u 10-3 s-1 at all the temperatures investigated (Fig. 21).
Extensive branching of transgranular cracks is noticed on the tested
samples associated with second phase boundaries especially at high strain
amplitudes of testing, Fig. 22(a). Further, microcracking perpendicular to
the loading direction in the vicinity of the main crack is noticed at a strain
rate of 3 u 10-2 s-1, Fig. 22(b). However, at lower strain rate of 3 u 10-3 s-1,
secondary crack density was found to be much lower, Fig. 22(c). Hence
the reduction of the fatigue life at high strain amplitudes and higher strain
rates could be rationalized based on the accelerated crack initiation and ac-
celerated crack propagation associated with crack linkage observed in this
alloy.
With decrease in strain amplitude from r1.0% to r0.25%, crack branch-
ing was reduced.[16]. Ingress of oxygen and oxidation of surface-
connected grain boundaries and slip bands were observed at lower strain
amplitudes and lower strain rates of testing. Failure at 873 K was found to
be strongly influenced by oxidation, particularly at lower strain amplitude
of testing (r0.25%). Figure 23(a) shows a SEM fractograph depicting
crack initiation and propagation being assisted by oxidation at the strain

amplitude of r0.25%. The optical micrograph in Fig. 23(b) shows cracks
filled with oxide debris obtained in low strain rate testing conditions. It
may be mentioned that the lower strain rates at elevated temperature pro-
vide adequate time for the environmental interaction to take place, which
accelerates both the crack initiation and propagation phases. In contrast,
under conditions of low temperatures and high strain rates, oxidation ef-
fects were not significant.
The detrimental effect of oxidation was reflected in transgranular crack
initiation and stage I and stage II crack propagation. The reduction in the
LCF life under high temperature and low strain rate conditions could thus
be ascribed to the enhanced environmental effects. Both crack initiation
and propagation were seen to be transgranular under all the testing condi-
tions investigated.
1500 Modified 9Cr-1Mo steel (Forged bar)
-3 -1
1400 Strain rate: 3 x 10 s
1300
1200
1100
1000
900
800
300 400 500 600 700 800 900
Temperature (K)
Fig. 19 Temperature effect on fatigue life of Modified 9Cr-1Mo steel (forged
bar).
0.010 Modified 9Cr-1Mo steel (forged bar)
-3 -1
Strain rate: 3x10 s
0.001
773 K
823 K
873 K
3 4
10 10
Number of reversals to failure 2Nf
Fig. 20 Fatigue lives at various temperatures and strain amplitudes of Modified
9Cr-1Mo steel.
Modified 9Cr-1Mo Steel (forged bar)

1200 Strain amplitude: +0.6%
1000
800
773 K
823 K
600 873 K
-3 -2
10 10
-1
Strain rate (s )
Fig. 21 Variation of LCF life with strain rate at different temperatures, Modified
9Cr-1Mo steel.
(a) At r1.0%, 3u10-3 s-1
(b) 823 K, 3u10-2 s-1, r0.6% (c) 823 K, 3u10-3 s-1, r0.6%
Fig. 22 Crack propagation at 873 K (arrows indicate the loading direction), Modi-
fied 9Cr-1Mo steel (forged bar).
Measurement of the distribution of the secondary crack number density

versus crack length at strain amplitudes of r0.6% and r1.0% in the tem-
perature range of 300-873 K is shown in Figs. 24(a) and (b), respectively.
It is observed that the total number of secondary cracks was higher at
higher strain amplitudes (r1.0%, Fig. 24(b)) of testing compared to r0.6%
(Fig. 24(a)). However, contrary to expectations, it was noticed that the
number of long secondary cracks decreased with increase in the test tem-
perature from 300 to 873 K, Fig. 24(a). This could be understood from the
fact that these measurements were taken at failure. At 873 K, since the al-
loy was exposed to lesser number of cycles before failure (which is as-
sisted by oxidation) and experienced lower stresses compared to that at
300 K, the secondary cracks at higher temperatures (773, 873 K) did not
grow in length. However, at larger strain amplitudes (r1.0%) where oxida-
tion effects were less apparent, significant differences in the crack density
with temperature did not exist.
(a)Oxidation-assisted crack initiation and propagation at r0.25%
(b)Modified 9Cr-1Mo steel (forged bar) at r0.6%.
Fig. 23 Cracking at 873 Kand 3u10-4 s-1.

14 Modified 9Cr-1Mo steel (forged bar)

-3 -1
Strain rate : 3x10 s
Number of cracks 12 Strain amplitude: +0.6%
10 RT
773 K
8 873 K
0
0-0.5 0.5-1 1-1.5 1.5-2 2-2.5 2.5-3 3-3.5 3.5-4 4-4.5 4.5-5 >5
(a) At r0.6%
50 Modified 9Cr-1Mo steel (forged bar)

-3 -1
Strain rate: 3x10 s
40
Number of cracks
30
20
773 K
873 K
10
0
0-0.5 0.5-1 1-1.5 1.5-2 2-2.5 2.5-3 3-3.5 3.5-4 4-4.5 4.5-5 5-5.5

(b) At r1.0%
Fig. 24 Frequency distribution of secondary cracks in Modified 9Cr-1Mo steel

(forged bar) .
3.3 Creep-fatigue-environment interaction - Hold time effects
Hold times are introduced in a strain cycle either in tension, compression

or tension plus compression peaks. This leads to creep-fatigue interaction.
However, in air testing oxidation effects also play a role in degrading the
fatigue life of the material at high temperatures. Tension hold is found to

be more damaging than compression hold in stainless steels namely SS
304[45], 304L[46], 316[47], 316L[48], 316L(N)[49] and Alloy D9 [50]
whereas, the converse holds good for 2.25 Cr–1Mo[51, 52], IMI 829 [53],
nickel base alloys [54,55] and Modified 9Cr-1Mo [18, 56-58]. In the fol-
lowing section studies conducted by the present authors to understand the
hold time effects on 316L(N) SS and its welds are discussed first followed
by the influence of hold time on Mod.9Cr-1Mo steel.
3.3.1 316L(N) Stainless steel base metal and weld metal
The introduction of strain hold at the peak strain in tension/compression in

total strain controlled testing causes stress relaxation leading to creep-
fatigue interaction. The creep-fatigue lives of the 316L(N) base metal and
316 weld metal as a function of the length of the hold time are shown in
Fig. 25. It could be seen that, at 873 K, in both continuous cycling and
hold time tests, 316 weld metal showed a higher fatigue endurance com-
pared to the base metal. It was also evident that the hold time effect on fa-
tigue life was dependent on the position as well as the duration of hold.
Compared to continuous cycling conditions, imposition of hold at peak
strain was found to decrease the fatigue life. Tensile hold was observed to
be more damaging than the compression hold. Further, a significant reduc-
tion in fatigue life was observed by increasing the duration of tensile hold.
The greater creep fatigue life of the weld metal compared to the base
metal at 873 K can be correlated to the crack propagation differences be-
tween the base metal (Fig.26(a)) and weld metal, Fig. 26(b). In base metal
under 1 min. tension hold the crack initiation occurs in intergranular mode
and crack propagation by mixed mode, Fig. 26(a). However in weld metal
microstructural changes influences the microcrack growth (crack growth
slowed down by crack deflection), Fig. 26(b). The fine duplex austenite-
ferrite microstructure in 316 weld metal with its many transformed phase
boundaries at these testing conditions, offered a greater resistance to the
extension of the fatigue cracks by causing deflection of the crack paths
compared to the base metal. G ferrite got transformed to brittle V phase
during testing. Crack deflection at the transformed G phase boundaries
was found to increase the crack propagation resistance causing an en-
hanced fatigue life.
At a given strain amplitude, the relaxed stress during hold time at half
life for the 316L(N) base metal is provided in Table 4. In all the hold time
tests, rapid stress relaxation occurred in the first few seconds of the strain
hold, followed by a slower rate of stress relaxation during the rest of the
hold period. During stress relaxation, conversion of elastic to plastic strain
took place and the strain rates were typically of the order of 10-4 s-1 to 10-8
s-1 during the slow relaxation period. The build up of tensile inelastic strain
led to the accumulation of grain boundary creep damage in the form of
cavities. With increase in the duration of the hold time a significant
amount of stress relaxation took place, leading to enhanced build up of in-
tergranular creep damage. This conforms with the magnitude of Vr devel-
oped during stress relaxation (Table 4) i.e. Vr increases with increase in the
length of the hold time and is greater in tension hold compared with the
compression hold.
Tension hold +0.6%, base

800 Comp. hold, +0.6%, base
Tension hold, +1.0%, base
Tension hold, +0.6% weld
600
400
200
-3 -1
316L(N) SS, 873 K, 3 x 10 s
0
0 20 40 60 80 100 120
Hold time (min.)
Fig. 25 Fatigue life as a function of hold time.
It must be pointed out that the absolute magnitude of the Vr alone could
not be associated with the damage that determines the creep-fatigue life.
As a function of the strain amplitude, it was observed that Vr was rela-
tively large at high strain amplitudes. However, the degree of reduction in
life during hold time tests, defined as N/Nf (N, the fatigue life during hold-
time tests and Nf the corresponding fatigue life in continuous cycling) was
found to be larger at lower strain amplitudes, compared to higher strain
amplitudes of testing (Table 4). The strain rates during relaxation at higher
strain amplitudes were generally higher than those observed at low strain
amplitudes of testing. In general, the relaxation strain rates of magnitude >
10-4 s-1 observed at high strain amplitudes are typically expected to cause
matrix deformation, while those observed at low strain amplitudes namely,
< 10-4 s-1, corresponded to that of creep deformation. It has been suggested
that relaxation strain rates < 10-4 s-1 generally contribute to grain boundary
damage and cause a greater reduction in life [59, 60].
(a) Mixed mode crack growth in 1 min. tension hold, 316L(N) base metal
(b) Crack growth in 316 weld metal, 1 min. T hold and +0.6%
Fig. 26 Crack morphology at 873 K.

Table 4 Effects of hold time on LCF properties at 873 K.
Hold 'Ht/2 'V/2 V r N N/Nf

(min) (%) (MPa) (MPa) (Cycles)
0 ±0.6 328 - 580(Nf) 1
1t ,, 307 51 475 0.819
1c ,, 323 45 510 0.879
10t ,, 291 57 409 0.705
30t ,, 288 74 330 0.635
90t ,, 274 94 235 0.405
0 ±1.0 370 - 130(Nf) 1
10t ,, 342 68 140 1.08
30t ,, 322 96 110 0.846
where,
t : Tensile hold ; c : Compression hold;
'Ht/2 : Total strain amplitude
'V/2 : Half-Life stress amplitude
N : Number of cycles to failure (with hold)
Nf : Number of cycles to failure (without hold)
V r : Relaxed stress during hold
The grain boundary damage developed during relaxation changes the

modes of crack initiation and propagation. In 1 min. compression hold, the
crack initiation and propagation occurs by transgranular mode (Fig.27(a))
similar to continuous cycling conditions, Fig. 27(b). However, under ten-
sion hold conditions crack initiation becomes intergranular and propaga-
tion is mixed mode (trans- + intergranular), Fig. 26(a).
It is seen that the surface-connected slip bands are oxidized under 10
min. tension hold conditions, Fig. 28(a) and the fracture surface is marked
periodically by fatigue striations and intergranular cracks , Fig. 28(b). Oxi-
dation interaction is found to be more as the length of the hold time is in-
creased to 90 minutes (Fig. 29(a)). Both crack initiation and propagation
seem to be strongly assisted by oxidation and fracture surface is com-
pletely covered by a thick oxide layer, Fig. 29(b).
(a) Within 1 min. compression hold for 316L(N) base metal
(b)Include crack initiation at 873 K:continuous cycling at

strain amplitude r0.6%
Fig. 27 Microstructure for transgranular crack propagation.

(a) Oxidation assisted crack initiation from slip bands
(b) Mixed mode crack propagation assisted by oxidation
Fig. 28 Microsdtructure for 316L(N) SS with 10 minutes tension hold at 873 K.
(a) Oxide location A

(b) Oxide location B
Fig. 29 Oxidation induced Intergranular Cracking in 90 min. Tension Dwell Tests,

316L(N) Base Metal, strain amplitude r0.6%.
The influence of the predominant fracture mode on fatigue can be as-

sessed from a quantitative measurement of crack density i.e. cumulative
length of the trans- and intergranular secondary cracks on the longitudinal
section per unit area, as shown in Fig. 30(a) along with the frequency dis-
tribution of the crack length, Fig. 30(b). It was seen that intergranular
crack density was the lowest in compression hold test, and crack propaga-
tion was mainly transgranular. This behaviour could be attributed to the
following reasons. During stress relaxation in compression hold, the
growth of internal grain boundary cavities was very unlikely, since this
needed both shear and normal tensile stresses across the grain boundary.
Thus for cycles containing compression hold periods, failure was domi-
nated by transgranular crack initiation and growth mechanisms and fatigue
lives were longer, compared with tensile hold periods. However, bulk
creep damage was favoured under tension hold conditions. Crack density
measurements on this steel, Fig.30(a) indicated that the transgranular
crack density decreased with an increase in the length of the hold-time in
tension and crack propagation was mainly intergranular. Moreover, it was
noted that crack linkage becomes extensive as the length of the hold time
increased and long intergranular cracks were observed in 10 and 90 min-
utes tension-hold tests, Fig. 30(b). Thus the reduced fatigue life for longer
hold-times could be ascribed to the occurrence of enhanced creep and oxi-
dation damage at grain boundaries that facilitates accelerated intergranular
crack initiation and propagation.
316L(N), 873 K, +0.6%

350
300 1 Min. compression

90 Min. tension
Crack density (Pm/mm )
2
250
10 Min. tension
1 Min. tension
200
150
100
50
0
Transgranular Intergranular
(a) Crack Density
316L(N), 873 K, +0.6%

14 1 Min. compression
90 Min. tension
12 10 Min. tension
1 Min. tension
Number of cracks
10
0
<1 1-2 2-3 3-4 4-5 5-6 6-7 7-8 >8
(b) Crack Length
Fig. 30 Crack distribution as a function of hold time with 873 K at r0.6%.

3.3.2 Modified 9Cr-1Mo steel
The influence of tensile and compressive hold times on LCF behaviour of

modified 9Cr-1Mo steel was investigated at 873 K at a strain amplitude of
r0.6%. The variation of fatigue life with the duration of tensile hold time
up to 30min on Mod.9Cr-1Mo steel (forged condition) at 873 K is shown
in Fig. 31 and a comparison of the effect of tensile, compressive and ten-
sile plus compressive hold time on fatigue life at 873 K of Modified 9Cr-
1Mo material (extruded tube) is shown in Fig. 32. It was seen that the fa-
tigue life decreased with increase in the length of hold time up to 30 min.
Dwell in compression was found to be more deleterious than dwell in ten-
sion and tension-plus-compression dwell was found to be the most detri-
mental. Enhanced dwell sensitivity in compression has also been reported
in this alloy by other investigators [56,57]. Further, it has been reported
that tension hold is more damaging than compression hold in vacuum test-
ing whereas the converse to be true for air testing [58]. Large tensile dwell
sensitivity is essentially observed in materials exhibiting creep cavitation
[45-50]. However, enhanced compressive dwell sensitivity, can be attrib-
uted to one of the following reasons, namely, development of tensile mean
stress, shape and size of cavities and oxidation behaviour.
The greater vulnerability of 2.25Cr-1Mo steel to compression hold has
been interpreted in terms of the behaviour of oxides [61]. It was shown
that in compression after a tensile hold, the oxide spalls, thus revealing a
950
Modified 9Cr-1Mo Steel
900
Temperature 873 K
-3 -1
850 Strain Rate: 3 x 10 s
Strain Amplitude: +0.6%
800
750
700
650
600
550
500
450
0 5 10 15 20 25 30 35
Hold time (min.)

Fig. 31 Influence of hold time on fatigue life at 873 K.
500
Modified 9Cr-1Mo Steel
Temperature 873 K
450
-3 -1
Strain Rate: 3 x 10 s
Strain Amplitude: +0.6%
400
350
300
250
Tension Hold
200
Compression Hold
Tension + Compression Hold
150
0 5 10
Hold time (min.)

Fig. 32 Fatigue life as a function of hold time, Mod. 9Cr-1Mo steel (extruded
tube), 873 K.
(a)Through (b) 1minute tension hold
(c) 10 minutes tension hold (d) 1 minute compression hold, 873

K, + 0.6%, Modified 9Cr-1Mo steel
(extruded tube)
Fig. 33 Fractograph showing slip band cracking under continuous cycling.
fresh surface, enabling the removal of surface damage. Conversely, in ten-

sion after a compression hold, the oxide film cracks facilitating early crack
nucleation in the underlying material. This behaviour was related to the
tensile strain in the oxide [62].
In Modified 9Cr-1Mo steel no appreciable tensile mean stress in com-
pression hold has been noticed. Further, creep cavitation was also not ob-
served during hold time experiments. Fractographs of the fatigue tested
samples under continuous cycling, 1 min tension, 10 min tension and 1
min compression shown in Fig. 33(a) through (d) respectively showed
clearly transgranular crack initiation from surface connected slip bands and
transgranular crack propagation marked by fatigue striations. This clearly
negates the role of intergranular cavities in degrading the fatigue life of
this material under hold time conditions.
Detailed metallography conducted on longitudinal sections of the fa-
tigue tested samples (under 1 min tension, 10 min tension and 1 min com-
pression depicted in Figs. 34 (a) through (c ) respectively) showed multi-
ple crack initiation sites and oxidation assisted crack initiation and
propagation. Also the number density of secondary cracks was more in the
case of compression hold compared to tension hold. This clearly indicated
that the larger compression dwell sensitivity in Modified 9Cr-1Mo ferritic
steel essentially arises from the oxidation effects. However, unlike 2.25Cr-
1Mo steel no oxide spallation was noticed in this steel during tension hold.
It is conceived that in modified 9Cr-1Mo steel an oxide layer formed on
the surface of the specimen cracks which on further oxidation during ten-
sion hold leads to crack tip blunting, while during compression hold they
stay sharp [56]. Moreover, it has also been proposed that the wedge effect
of the exfoliated oxide cause oxidation-assisted cracking in Modified 9Cr-
1Mo steel. Further, the oxide layer formed on the surface of the specimen
was a weak barrier that could be easily overcome by slip and several crack
initiation sites were observed in air-tested specimen compared to that in
vacuum [63]. The reduced life in this steel, under creep-fatigue interaction
conditions has thus been attributed to oxidation assisted crack initiation
and propagation.
4. Concluding Remarks
Discussed are a few aspects of how microstructure, both initial and that
developed subsequently during LCF deformation influence the formation
and propagation of microcracks which eventually determine fatigue life,
from examples of authors’ work on materials of interest to fast breeder re-
actor technology. In Nimonic PE-16 shearing of J' precipitates leads to cy-
clic softening. Shearing of precipitates under heterogeneous distribution
(a) 1 minute tension hold
(b) 10 minutes tension hold
(c) 1 minute compression hold, modified 9Cr-1Mo steel (extruded tube),

873 K, +0.6%.
Fig. 34 Optical Metallograph of longitudinal section of fatigue tested samples de-

picting secondary cracks.
can lead to localization of strain resulting in enhanced transgranular and

intergranular cracking and a reduced fatigue life. Intergranular fracture is
accentuated by the presence of carbide precipitates on the grain bounda-

ries. Dynamic strain ageing is found to influence fatigue deformation and
fracture at high temperatures and low strain rates conducive to dislocation
solute atom interaction. DSA leads to enhanced localization of slip leading
to inhomogeneity of deformation and accelerated intergranular and trans-
granular cracking. This caused a reduction in life. In austenitic stainless
steel base metal and weld metal fatigue life decreased with the introduction
of hold. Tension hold was found to be more deleterious than compression
hold. Further, fatigue life decreased with increase in the length of the hold
time. Creep and oxidation contributed to reduction in fatigue life with hold
time. In situations where crack deflection associated with the G ferrite
transformation occurred, 316 weld metal showed a higher fatigue life
compared to the base metal. In Modified 9Cr-1Mo ferritic steel compres-
sion hold was found to be more deleterious than tension hold. It is sug-
gested that the thin oxide layer formed on the surface is cracked by slip
processes in this steel. Further, crack tip blunting by oxidation during ten-
sion hold reduces the crack propagation rate, whereas in compression hold
the crack remains sharp leading to enhanced crack propagation rate. Un-
derstanding of mechanism of microcracking in different alloy systems un-
der a variety of test conditions is important for the development of micro-
structures resistant to fatigue cracking under specific test conditions.
Further considering the fact that many approaches are available for life
prediction at elevated temperatures, understanding of the mechanisms of
deformation and damage would lead to realistic life assessment and predic-
tion techniques.
References
[1] Coffin LF. Effect of frequency and environment on fatigue crack growth of
A286 at 1100 F. In: Carden AE, Mc Evily AJ, Wells CH (eds) Fatigue at ele-
vated temperatures. STP 520, ASTM, Philadelphia, PA, (1973) 112-122.
[2] Gell M and Leverant GR. Mechanism of high temperature fatigue. In:
Carden AE, Mc Evily AJ, Wells CH (eds) Fatigue at elevated temperatures.
STP 520, ASTM, Philadelphia, PA, (1973,) 37-67.
[3] Antolovich SD, Liu S and Baur R. Low cycle fatigue behaviour of Rene 80
at elevated temperature. Metall Trans A, 12 (1981) 473-481.
[4] Valsan M, Parameswaran P, Bhanu Sankara Rao K, Vijayalakshmi M, Man-
nan SL and Sastry DH. High-temperature low cycle fatigue behaviour of a
Nimonic PE-16 superalloy - correlation with deformation and fracture. Met-
all Trans A, 23 (1992) 1751-1761.
[5] Mannan SL, Valsan M High temperature low cycle fatigue, creep-fatigue
and thermomechanical fatigue of steels and their welds. Int J Mechanical
Sciences 48 (2006) 160-175.
[6] Sidey D and Coffin LF Jr. Low cycle fatigue damage mechanisms at high
temperature. In: Fong JT (ed) Symp. on fatigue mechanisms. ASTM STP,
675 (1979) 528-570.
[7] Wells CH, Sullivan CP and Gell M. Mechanism of fatigue in creep range, In:
Manson SS (ed) Metal fatigue damage: mechanism, detection, avoidance and
repair. ASTM STP, 495 (1971) 61-122.
[8] Bhanu Sankara Rao K. Time dependent low cycle fatigue behaviour of high
temperature alloys. Trans Indian Inst Met, 42 (1989) 61-82.
[9] Mannan SL. Role of dynamic strain ageing in low cycle fatigue. Bull of Ma-
ter Sci, 16(6) (1993) 561-582.
[10] Valsan M, Sastry DH, Bhanu Sankara Rao K and Mannan SL. Effect of
strain rate on high-temperature low-cycle fatigue properties of a Nimonic
PE-16 superalloy. Metall Trans A, 25 (1994) 159-171.
[11] Valsan M, Sundararaman D, Ray SK and Mannan SL. Low cycle fatigue de-
formation behaviour of a Nimonic PE-16 superalloy in a double-aged micro-
structural condition. Trans Indian Inst Met, 49(4) (1996) 471-477.
[12] Srinivasan VS, Sandhya R, Bhanu Sankara Rao K, Mannan SL and Ragha-
van KS. Effects of temperature on the low cycle fatigue behaviour of nitro-
gen alloyed type 316L stainless steel. Int J of Fatigue, 13(6) (1991) 471-478.
[13] Valsan M, Sundararaman D, Bhanu Sankara Rao K and Mannan SL. A com-
parative evaluation of low cycle fatigue behaviour of type 316L(N) base
metal, 316 weld metal and 316L(N)/316 weld joint. Metall and Mater Trans-
actions A, 26 (1995) 1207-1219.
[14] Nagesha A, Valsan M, Bhanu Sankara Rao K and Mannan SL. Strain con-
trolled fatigue of type 316L(N) base metal and 316 SS weld metal at elevated
temperatures. In: Xue-Ren Wu, Wang ZG (eds) Proceedings of the seventh
International fatigue conference, FATIGUE ’99,2. Beijing, China, (1999)
1303-1308.
[15] Srinivasan VS, Valsan M, Sandhya R, Bhanu Sankara Rao K, Mannan SL
and Sastry DH. High temperature time-dependent low cycle fatigue behav-
iour of a type 316L(N) stainless steel. Int J of Fatigue, 21 (1999) 11-21.
[16] Nagesha A, Valsan M, Kannan R, Bhanu Sankara Rao K and Mannan SL.
Influence of temperature on the low cycle fatigue behaviour of a modified
9Cr-1Mo ferritic steel. Int J of Fatigue, 24(12) (2002) 1285-1293.
[17] Mannan SL, Bhanu Sankara Rao K, Valsan M and Nagesha A. Strain con-
trolled low cycle fatigue and creep-fatigue interaction behaviour of Modified
9Cr-1Mo ferritic steel. Trans Indian Inst Met, 58(2-3) (2005) 159-168.
[18] Vani Shankar, Valsan M, Kannan R, Bhanu Sankara Rao K, Mannan SL and
Pathak SD. Low cycle fatigue behavior and microstructural evolution of 9Cr-
1Mo-V-Nb ferritic steel. Mat Sci and Engg, 437(2) (2006) 413-422.
[19] Srinivasan VS, Valsan M, Bhanu Sankara Rao K, Mannan SL and Raj B.
Low cycle fatigue and creep-fatigue interaction behaviour of 316L(N)
stainless steel and life prediction by artificial neural network approach. Int J
of Fatigue, 25 (2003) 1327-1338.
[20] Srinivasan VS, Sandhya R, Valsan M, Bhanu Sankara Rao K and Mannan
SL. Comparative evaluation of strain controlled low cycle fatigue behaviour
of solution annealed and prior cold worked 316L(N) stainless steel. Int J of
Fatigue, 26 (2004) 1295-1302.
[21] Coffin LF Jr. Fatigue at high temperature. In: Carden AE, Mc Evily AJ,
Wells CH (eds) Fatigue at elevated temperatures. ASTM STP, 520 (1973) 1-
34.
[22] Brinkman CR, Gieseke BG and Maziasz PJ. The influence of long term
thermal ageing on the microstructure and mechanical properties of modified
9Cr-1Mo steel. In: Liaw PK, Viswanathan R, Murty KL, Simonen EP, Frear
D (eds) Microstructure and mechanical properties of ageing material. The
Minerals, Metals and Materials Society, (1993), 107-115.
[23] White RJ and Fisher SB. The precipitation and growth kinetics of Jc in Ni-
monic PE-16. Mat Sci and Engg, 33 (1978) 149-157.
[24] Seetharaman V, Bhanu Sankara Rao K, Mannan SL and Rodriguez P. Micro-
structure and mechanical properties of a Nimonic PE-16 superalloy subjected
to double ageing treatment. Mat Sci and Engg, 63 (1984) 35-50.
[25] Taillard R, Degallaix S and Foct J. Effect of nitrogen upon the low cycle fa-
tigue behaviour of 316L stainless steels. In: Rie KT (ed) Low cycle fatigue
and elasto-plastic behaviour of materials. Elsevier, Munich, (1987) 83-88.
[26] Sandstrom R, Engstrom J, Nilsson JO and Nordgren A. Elevated temperature
low cycle fatigue of the austenitic stainless steel type 316 and 253 MA – in-
fluence of microstructure and damage mechanisms. High Temp Technol, 7
(1989) 2-9.
[27] Nilsson JO. The influence of nitrogen on high temperature low cycle fatigue
behaviour of austenitic stainless steel. Fat Fract Engg Mater Struct 7, (1984)
55-64.
[28] Vogt B, Degallaix S and Foct J. Low cycle fatigue life enhancement of 316L
stainless steel by nitrogen alloying. Int J Fatigue, 6(4) (1984) 211-215.
[29] Nilsson JO. The effect of slip behaviour on the low cycle fatigue behaviour
of two austenitic stainless steels. Scripta Metall, 17 (1983) 593-596.
[30] Taillard R and Foct J. Mechanisms of the action of nitrogen interstitials upon
low cycle fatigue behaviour of 316 stainless steels. In: Foct J, Hendry A
(eds) Proc Int conf on high nitrogen steels 88. The Institute of Metals and the
Socitie francaise de Metallurgie , Lillie, France, (1988) 387-391.
[31] Nilsson JO and Thorvaldsson T. The influence of nitrogen on microstructure
and strength of a high-alloy austenitic stainless steel. Scand J Metall, 15
(1985,) 83-89.
[32] Stoltz RE and Pineau AG. Dislocation-precipitate interaction and cyclic
stress-strain behaviour of a J' strengthened superalloy. Mater Sci Eng, 34
(1978) 275-284.
[33] Calabrese C and Laird C. Cyclic stress-strain response of two phase alloys
part I. Microstructure containing particles penetrable by dislocations. Mater
Sci Engg, 13 (1974) 141-157.
[34] Lerch A and Gerold V. Cyclic hardening mechanisms in Nimonic 80A. Met-
all Trans A, 18 (1987) 2135-2141.
[35] Kim WH and Laird C. Crack nucleation and stage I propagation in high
strain fatigue – II mechanism. Acta Metall, 26 (1978) 789-799.
[36] Hornbogen E and Zum Gahr KH. Microstructure and fatigue crack growth in
a J Fe-Ni-Al alloy. Acta Metall, 24 (1976) 581-592.
[37] Arbuthnot CHD. The Micro-mechanisms of fatigue deformation in Nimonic
PE-16. In: Maurer KL, Matzer FE (eds) Proc 4th european conf on fracture
and role of microstructure 2. Leoben, Austria, The Chameleon press Ltd.,
London, United Kingdom, (1982) 407-413.
[38] Mediratta SR, Ramaswamy V and Rama Rao P. Two stage cyclic work hard-
ening and two-slope Coffin-Manson relationship in dual phase steels. Scripta
Metall, 20 (1986) 555-558.
[39] Sanders TH Jr and Starke EA Jr. The relationship of microstructure to mono-
tonic and cyclic straining of two age hardening aluminum alloys. Metall
Trans, 7A (1976) 1407-1418.
[40] Bhanu Sankara Rao K, Valsan M, Sandhya R, Mannan SL and Rodriguez P.
Synergistic interactions during high temperature fatigue of type 304 stainless
steel-grain size dependence. Trans Indian Institute of Metals, 44(3) (1991)
255-270.
[41] Bhanu Sankara Rao K, Valsan M, Sandhya R, Mannan SL and Rodriguez P.
Grain size dependence of dynamic strain ageing during low cycle fatigue of
type 304 stainless steel. In: Salama K, Ravi Chandar K, Taplin DMR, Rama
Rao P (eds) ICF-7, Advances in fracture research 2, Pergamon, Oxford,
(1989) 1323-1330.
[42] Kanazawa K, Yamaguchi K and Nishijima S. Mapping of low cycle fatigue
mechanisms at elevated temperatures for an austenitic stainless steel. In:
Solomon HD, Halford GR, Kaisand LR, Leis BN (eds) Low cycle fatigue.
ASTM STP 942, Philadelphia, (1988) 519-530.
[43] Abdel-Raouf H, Plumtree A and Topper TH. Effect of temperature and de-
formation rate on cyclic strength and fracture of low carbon steel. In: Coffin
LF, Krempl E (eds) Cyclic stress-strain behaviour- analysis, experimentation
and failure prediction. ASTM STP, 519 (1971) 28-57.
[44] Bressers J. Fatigue and microstructure. In: Marriott JB, Merz M, Nihoul J,
Ward J (eds) High temperature alloys, their exploitable potential. Elsevier
Applied Science, Amsterdam, (1987) 385-410.
[45] Lee SB and Miller AK. A phenomenological model for intergranular fatigue
by r-type and wedge type cavitation. J Engg Mater Tech, 117 (1995) 311-
321.
[46] Halford GR, Johnson JR and Brown JA. In: Richmond RJ, Wu ST (eds) Ad-
vanced earth – to – orbit propulsion technology. NASA Conf publications,
(1986) 2437-2450.
[47] Yamaguchi K and Kanazawa K. Effect of strain wave shape on high tem-
perature fatigue of type 316 steel and application of strain range partitioning
method. Metall Trans, 11A (1980) 2019-2027.
[48] Shi HJ and Pluvinage G. Cyclic stress strain response during isothermal and
thermomechanical fatigue. Int J Fatigue, 16 (9) (1994) 549-557.
[49] Srinivasan VS, Nagesha A, Valsan M, Bhanu Sankara Rao K, Mannan SL
and Sastry DH. Effect of hold time on low cycle fatigue behaviour of nitro-
gen bearing 316L stainless steel. Int J of Pressure Vessels and Piping, 76
(1999)863-870.
[50] Sandhya R, Bhanu Sankara Rao K and Mannan SL. Creep fatigue interaction
behaviour of a 15Cr-15Ni Ti Modified austenitic stainless steel as a function
of Ti/C ratio and microstructure. Mat Sci and Engg A, 392 (2005) 326-334.
[51] Brinkman CR, Strizak JP, Brooker MK and Jaske CEJ. Time dependant
strain controlled fatigue behaviour of annealed 2.25Cr-1Mo steel for use in
nuclear steam generator design. J Nucl Mater, 62 (2/3) (1976) 181-204.
[52] Goswami T. Creep-fatigue: Paper 1. Compilation of data and trends in the
creep fatigue behaviour of low alloy steels. High Temp Mater Proc, 14(1)
(1995) 1-19.
[53] Plumbridge WJ and Stanley M. Low cycle fatigue of a Titanium 829 alloy.
Int J Fatigue, 8 (1986) 209-216.
[54] Plumbridge WJ and Ellison EG. Low cycle fatigue behaviour of superalloy
blade materials at high temperatures. Mater Sci Tech, 3 (1987) 706-715.
[55] Nazmy MY. High temperature low cycle fatigue of Inconel 783 and applica-
tion of strain range partitioning. Metall Trans, 14A (1983) 449-461.
[56] Aoto K, Komine R, Ueno F, Kawasaki H and Wada Y. Creep-fatigue evalua-
tion of normalized and tempered modified 9Cr-1Mo. Nucl Engg Design, 153
(1994) 97-110.
[57] Van Den Avyle JA and Jones WB. Substructure and strengthening mecha-
nisms in 2.25Cr-1Mo steel at elevated temperatures. Metall Trans, 11A
(1980) 1275-1286.
[58] Kim S and Weertman JR. Investigations of microstructural changes in a fer-
ritc steel caused by high temperature fatigue. Metall Trans A, 19 (1988) 999-
1007.
[59] Bhanu Sankara Rao K, Meurer HP and Schuster H. Creep-fatigue interaction
of Inconel 617 at 950 qC in simulated nuclear helium. Mat Sci and Engg,
104A (1988) 37-51.
[60] Rie KT, Schmidt RM, Ilschner B and Sam SW. A model for predicting low
cycle fatigue life under creep-fatigue interaction. In: Solomon HD, Halford
GR, Kaisand LR, Leis BN (eds) Low cycle fatigue. ASTM STP 942, Phila-
delphia, (1988) 313-328.
[61] Teranishi H and McEvily AJ. On the fatigue crack initiation and propagation
at elevated temperatures. In: Francois O (ed) ICF 5, Advances in Fracture
Research, Pergamon Press, Paris, (1980) 2439-2445.
[62] Challenger KD. Elevated temperature fatigue of 2.25Cr-1Mo steel. In: Val-
luri SR, Taplin DMR, Rama Rao P, Knott JF, Dubey R (eds) ICF6, New
Delhi, India, (1984) 2239-2248.
[63] Ebe G and McEvily AJ. Effect of processing in high temperature low cycle
fatigue properties of modified 9Cr-1Mo steel. Fatigue and Fracture Engg
Mat Struct, 17 (1984) 299-314.
Invariant Form of Micro-/Macro-Cracking in Fatigue
G. C. Sih 1, 2 *
1
International Center for Sustainability, Accountability and
Eco-Affordability of Large Structures (ICSAELS)
Advanced Technology for Large Structural Systems (ATLSS)
2
School of Mechanical and Power Engineering, East China
University if Science and Technology, Shanghai 200237, China
*Email: gcs8866@yahoo.com, gcs1@lehigh.edu
Abstract
The size/time transitory character of fatigue cracking can be hidden in the

data or revealed explicitly depending on the manner of presentation. The
former can correspond to the crack length while the latter to the crack
growth rate variations, both referred to the number of fatigue cycles.
Conventionally speaking, the sigmoidal curve for the plot of log da/dN
against log 'K consists of three regions known as I, II and III. They are
referred to, respectively, as crack initiation, stable crack growth and fast
fracture although the transition from one region to another involving scale
shifting is seldom discussed. Even though damage by very small defects are
recognized in fatigue but their representation may become dubious when
the parameter “a” in the crack growth rate expression da/dN may have to
adopt a different physical interpretation. Using the line crack configuration
for describing crack growth in metals, reference will be made to three
approximate crack size range of a |10-80 mm (macromechanical) for
regions II and III, a |10-3 to 10-1mm (microstructural) for region I and
presumably a |10-4 mm or smaller (nanochemical) for the lower scale
regions.
The objective of developing a multiscale fatigue crack growth model is to
determine the increment of a stress intensity or energy density parameter
such that the data in regions I, II and III can be related. One way is to have
the crack growth rate da/dN to lie on a straight line such that linear
interpolation can be applied to relate the nano, micro and macro fatigue
data. This behavior is referred to as form-invariant. The proposed scheme
accounts for two succeeding scales such as micro/macro and nano/micro. In
this way. the scaling shifting capability can be extended, say from nano to
macro. The dual scale micro/macro line crack model will be presented. It
consists of three physical parameters d* (micro/macro length), P*
182 G. C. Sih
(micro/macro material constants) and V* (external/internal stresses). The

model can also delve into the details of the material microstructure when
needed. For structural applications many of the incidental material
parameters can be absorbed by the macro empirical parameters which will
be limited to two such that the present findings can be incorporated into the
current fracture control methodology.
Using the available fatigue crack growth data for the 7075-T6 and
2024-T3 aluminum panels, the sigmoidally-curved regions I, II and III will
be shown to become a straight line whose slope and y-intercept in a log-log
plot will depend on the mean stress Vm and stress amplitude Va that are
accounted for simultaneously by the dual scale energy density expression.
The slopes of lines line referred to the energy density plots ranged from
2.890 to 3.019 when Vm varied from 9 to 12 kg/mm2 and Va from 2.5 to 6.5
kg/mm2. The slopes obtained a wider scatter of 3.532 to 4.224 when the
energy density range was replaced by the stress intensity range. This was
expected because the stress intensity approach leaves out the mean stress
effect. The present work is offered as the first step to include scaling effects
so that micro material behavior can enter into the macro fatigue crack
growth process.
1. Introduction
The macro mechanical properties are known to be affected by the material

microstructure which can be changed by heat treatment and/or high
pressure. Various processing techniques have been developed and applied
to alter the strength of uniaxial specimens. The results presumably can be
extended to the design of structural members under multi-axial stress states.
As the test data scatter increased with increasing material strength, it was
found necessary to account for the effect of pre-existing defects. The
classical notion of plane stress for thin bodies and plane strain for thick
sections were realized to be inadequate as they entail only global scale
dimensions. The size of failure initiation locality relative to global
instability is fundamental. This has led to crack initiation and propagation
as the two events considered in the facture mechanics discipline [1-3].
Scale segmentation was also necessary such that analyses can be divided
into different ranges referred to as nanoscopic, microscopic and
macroscopic. The inter-connecting gap has been known as the mesoscopic
[4,5] region. These distinctions were necessary to avoid considering the
process of non-equilibrium [6,7] where the material properties changed not
only from location to location but also in elapsed time. By tradition, the
open thermodynamic system has been simplified to a closed system where
the conditions on the boundary are assumed to be known and hence

specified. The fluctuations or movements at the interface are assumed to be
localized for large bodies and ignored. As the body size is decreased, the
average and bulk properties are no longer representative of the physical
behavior. At the nano scale, the physical laws can differ [8,9] and the
description of matter can adopt an uncertain character as advocated in
quantum mechanics where the origin of failure may initiate from the
non-uniform distribution of the electrons [10].
Advancement in micro/macro crack interaction in fatigue has been
crawling at the a snail’s pace because not enough attention has been
directed to understand the mechanism of crack opening followed by closing
in the same cycle repeated many times. Most of the data are collected for
metals undergoing low amplitude loads alternating at one-half of the yield
strength according to the ASTM standard. This corresponds to the situation
where the elements ahead of a macroscopic crack would undergo
permanent deformation such that the local hyteresis loops would be
measurable. The cracked material at the microscale has sufficient restraint
to close upon unloading, A restraining force or stress is said to prevail that
changes with the crack length. This mechanism is believed to be
fundamental in describing the transitory character of
micro-/macro-cracking that correspond to regions I and II of the
conventional sigmoidal fatigue crack growth rate curve. The dual scale
crack model [11,12] provided the impetus for developing a fatigue crack
model that can switch from a microcrack to a macrocrack and vice versa.
Such a crack prevails in a non-homogeneous medium where the top and
bottom surface of the crack would subject to different constraint that can be
modeled by the free-fixed boundary condition. That is the crack path would
in general wonder about without symmetry. This led to the discovery of a
crack with double stress singularity [13]. The weak singularity of order
1/r0.25 can model the microcrack behavior and the strong singularity of order
1/r0.75 can model the macrocrack. The solution was validated by satisfying
the equations of continuum mechanics with finite displacements at the
singular crack tip. In addition, the microstructure effects of the material
were also shown to affect the angular distribution of the double singularity
local stress field. Hence, two additional parameters M1 and M2 would affect
the angular variations of the stresses where the coordinate angle T is
measured from the line extended from the original crack plane. The details
can be found in [14]. They reveal the complexities of the analysis when the
microstructural effects of the material are included.
The influence of material microstructure on fatigue crack growth has
received much interest in recent years [15-20] although the approaches
184 G. C. Sih
varied from the empirical to the semi-analytical models. The results when
presented in terms of the crack growth rate da/dN can be irratic if the spatial
and temporal scale is not carefully observed. Variance in data presentation
cannot be avoided if no agreement can be found in defining nano-, micro-
and macro-crack size in relation to time. Consistency in data presentation
appears to be needed before progress can be made to understand the
underlying physical mechanisms that govern fatigue crack growth.
2. Statement of Fatigue Crack Growth Rate Models
The advent of linear elastic fracture mechanics was followed by the

development of the two-parameter fatigue crack growth relation [21]
da
=A(ǻK) p (1)
dN
In Eq.(1), A and p were found empirically from a log-log plot of da/dN

versus 'K which is known as the macro-stress intensity factor such that it
depends on the stress amplitude range and the square root of the half crack
length a for a panel with a central crack of length 2a. To be more specific,
reference can be to the plot in Fig.1 for aluminum alloys [22]. Region II
refers to long cracks for a from 10 to 40mm. For short cracks with a in the
range of 1 to 10mm, the curve acquires a different slope and has been
referred to as the “threshold”. Under alternating tension and compression
loading, the crack mouth opens and closes for each cycle. It can change
from invisible (microscopic) to visible (macroscopic). The corresponding
data switches from region I to region II. For lower da/dN, the crack size is
even smaller with dimensions of 10-2 to 10-1mm. The recent concern is
whether the classical 'K correlation should be replaced by other quantities
to account for the change of small to large cracks. This would no doubt alter
the shape of the da/dN curve, in addition to a possible reinterpretation of the
physical mechanism(s) that is causing this change.
Although countless empirical forms of 'K in Eq.(1) has been proposed to
offer a better fit of the data, few of the models revealed any significant
progress. The exception is the work in [23,24] that gave an exponent of p=2
using a self-similitude argument in contrast to p=4 [21] arrived from a plot
of a multitude of fatigue data for aluminum alloys. By in large; the form of
Eq.(1) has been retained while the exponent p may vary in application. In
what follows, the classical form of 'K will be replace by ǻK macro micro such
that
da
C('K macro
micro )
n
(2)
dN
The superscript “macro” and subscript “micro” are introduced to show that
ǻK macro
micro can represent the intensity factor for the microcrack or the
macrocrack depending on the tightness ratio V* of the contacting crack
surfaces in addition to the micro/macro shear modulus ratio P* and a
relative length parameter d* that defines the relative size of the local crack
tip region as shown in Fig.2. The restraining stress Vo that controls the
tightness of the crack surfaces is normalized to the applied stress Vf to
define V*. In situations where the mean stress Vm plays an important role in
addition to the stress amplitude Va, Eq.(2) will be further replaced by the
range of the micro/macro energy density factor 'S macro micro offering another
form of the fatigue crack growth relation:
da
B('Smacro
micro )
m
(3)
dN
Fig. 1 Fatigue crack size increase with growth [22].

186 G. C. Sih
Fig. 2 Strong and weak stress singularity for the dual scale crack model.
3. Stress Intensification Approach
The strength of stress singularity near a crack tip played a fundamental role
in the development of the linear elastic fracture mechanics model as it can
be related to the energy released by a unit extension of a self-similar crack.
Such a simple interpretation ceases to prevail for a microcrack where
anti-symmetry becomes the rule rather than the exception. A brief
description of the double singularity stress field will be made for the sake of
continuity.
3.1 Double singularity stress field
The double singularity solution can be found in [13]. Adopted in this

discussion is a crack with free-fixed surface condition. Two singularities of
different order prevail at the same crack tip. There is the strong singularity
with order 1/r0.75 and the weak singularity with order 1/r0.25 having the
respective intensities of (K S ) macro W macro
micro and ( K ) micro as shown by
(K S ) macro (K W ) macro
(V ij ) macro
micro
micro
f ij (M1 , T) micro
g ij (M 2 , T) (4)
2Sr 0.75 2Sr 0.25
The method of scale multiplier can be used [13] to give

(K S ) macro
micro d (K W ) macro
micro (5)
Eq. (1) can thus be re-written only in terms of (K S ) macro

micro as
macro (K S ) micro
macro
r
(ıij ) micro = [fij (ĳ1 ,ș)+ g (ĳ 2 ,ș)]+ (6)
2ʌ r 0.75
d ij
To be reminded is the notation ( ) macro micro that applies to the dual scale
model to denote that the crack can be micro or macro depending on the ratio
d/r (or d*) and the restraining stress Vo to applied stress ratio Vf ratio given
by V* that is contained in (K S ) macro *
micro . Large V would close the crack
making it invisible and the opposite would result if the restraining stress is
small. Refer to Fig.1. This effect is particularly important in fatigue where
loading and unloading tend to open and close the crack continuously. As it
can be seen from Eq. (6), the local crack tip stresses depend on the distance
on the relative distance r/d. Such a character is also necessary for
addressing micro/macro effects. Since both (K S ) macro W macro
micro and ( K ) micro are
related by Eq.(5), it suffices to consider (K S ) macro

micro only and superscript s can
be dropped to give
16 ʌ (1-Ȟ macro )µ micro ı f c 2 -a 2 2ı a

K macro
micro = 0.25
(1- o sin -1 ) (7)
f(ĳ1 ,ĳ 2 )µ macro d ʌı f c
provided that
f(M1, M2)=5M1+7M2+12 (8)
The relation between c and a can be found in Fig.1. Eq. (4) can be found
in [13] and will not be elaborated any further.
3.2 Crack growth relation based on stress intensity
Assuming that the maximum and minimum values of applied and crack tip
stresses occur at the same time without phase change, then the increment
change of the local stress intensification K macro
micro can be denoted by
188 G. C. Sih
'K macro
micro (K macro macro
micro ) max ( K micro ) min (9)
Making use of Eq.(7) and letting Vf equal to Vmax at maximum stress and
to Vmin at minimum stress, it is found that
16 S (1 Q macro )P micro c 2 a 2 2V o a
(K macro
)
micro max 0.25
(1 sin 1 )V max
f (M1 , M 2 )d P macro SV f c
(10)
16 S (1 Q macro )P micro c 2 a 2 2V a
(K macro
)
micro min 0.25
(1 o sin 1 )V min
(11)
Inserting Eqs. (10) and (11) into Eq. (9), there results
'K macro
micro 0.25
(1 o sin 1 )(V max V min )
(12)
Recall that V*=Vo/Vf and P*=Pmicro/Pmacro. The length d=ca is defined in

Fig.1 and the quantities M1 and M2 can be identified with different
microcrack branching patterns [14]. They may also be related to fissures
and striations that are observed by micrographs in the fatigue of aluminum
alloys. Although these microscopic effects involving the Ms can be included
in Eq. (12), it is more pertinent to focus attention on the invariant form of
da/dN by first expressing Eq.(2) in the form
da
C1 ('K macro
micro )
n
(13)
dN
At this stage, the incidental variables M1 and M2 or the function f(M1,M2) can
be absorbed into the macro-parameters. More specifically, the quantity
3f(M1,M2)/40 may be factored out of Eq.(13) after eliminating ǻK macro micro by
using Eq.(12). This renders a new ǻK macro
micro without f(M1,M2) that will be
understood to be
'K macro
micro 0.25
(1 o sin 1 )(V max V min )
5d P macro SV f c
(14)
A new parameter C can now be defined as
n
ª 40 º
C « » C1 (15)
¬ 3f (M1 , M 2 ) ¼
Application of Eqs.(14) and(15) can convert, Eq.(13) to the form of Eq.(2)

which will be used for the calculation of numerical data. There should be no
confusion between the use of ǻK macro
micro in Eq.(12) with f(M1,M2) and the use
of ǻK macro
micro in Eq.(14) without f(M1,M2). Both form of da/dN involves only
two empirical parameters C1 and n or C and n. The foregoing exercise
shows explicitly that microscopic effects can be hidden in the macroscopic
empirical constants C, C1 and n.
4. Energy Density Intensification Approach
The local energy density factor can be used to include the effect of mean
stress. The form of dW/dV that can be applied to micro-/macro-cracking
given by
macro
§ dW · [S(r, T)]macro
micro
¨ ¸ (16)
© dV ¹ micro r
4.1 Strength of energy density singularity
The 1/r singularity for dW/dV applies also for the double stress singularity
field of Eq.(4) or (6). Knowing that K is related to S [3], it can be shown
that [13]
(1 Q macro ) 2 (c 2 a 2 ) P 2V a
S macro
micro 2 0.5 0.5
( micro V f ) 2 (1 o sin 1 ) 2 F(T)
128f (M1 , M 2 )P micro d r P macro SV f c
(17)
Note that
190 G. C. Sih
1
F(T) [f r (T) f T (T)]2 2[f rT (T)]2 (18)
2
in which fr(T), fT(T) and frT(T) are complicated functions of T and Mj with
j=1, 2. They can be found in [13] and will not be repeated here since they
can be absorbed into the empirical parameters to be shown later. The energy
density intensification is non-uniform around the crack and S macro micro in
Eq.(17) can depend on the radial distance r from the crack tip and the
microstructural details via f(M1, M2).
4.2 Crack growth relation based on energy density intensification
While non-local effects are included in 'K macro micro of Eq.(2), 'K of Eq.(1)
applies locally only in the limit that the distance r approaches zero. Scaling
for crack growth has also been excluded in the classical 'K. This limitation
does not exist in the 'S macro
micro which can be regarded as the energy released
when the crack extends by the amount r='a. Now let (S macro
micro ) max be
associated with Vmax and (S macro

micro ) min associated with Vmin such that
'S macro
micro (S macro macro
micro ) max (S micro ) min (19)
and using Eq. (17), it can be shown that
(1 Q macro ) 2 (c 2 a 2 ) P 2V a
'Smacro
micro ( micro ) 2 (1 o sin 1 ) 2 F(T)(V 2max V 2min )
128f 2 (M1 , M 2 )P micro d 0.5 r 0.5 P macro SV f c
(20)
A two parameter fatigue crack growth rate relation can thus be obtained:
da
B1 ('S macro
micro )
m
(21)
dN
This is reminiscence of the classical strain energy density approach to

fatigue in [3] if 'S macro
micro in Eq. (21) is replaced by 'S for macrocracking
only. A simplified version of Eq. (20) or (21) may be used by leaving out
the details of the incidental parameters M1 and M2. This can be

accomplished by letting
F(T)
B [ 2
]m B1 (22)
32f (M1 , M 2 )(1 2Q micro )
and defining a new 'S macro

micro without f(M1, M2). Again, there should be no
confusion between Eq.(20) with f(M1, M2).and Eq.(23) without f(M1, M2). In
what follows, Eqs.(22) and (23) will be used such that da/dN can be
computed from Eq.(3). This completes the derivation of the dual scale line
crack model based on the energy density intensification approach.
Calculation of numerical results will follow.
(1 2Q micro )(1 Q macro ) 2 (c 2 a 2 ) P micro 2 2V a

'Smacro
micro 0.5 0.5
( ) (1 o sin 1 ) 2 (V 2max V 2min )
4P micro d r P macro SV f c
(23)
5. Invariant Form of Two-Parameter Fatigue Crack Growth Rate

Relation
The two parameter crack growth rate relation of Eq. (1) was first proposed
in [21] and discussed in [23, 24] with reference to the value of the exponent
p on 'K. Although much work has been done in the ways that A and p in
Eq.(1) are affected by microscopic effects, it is rather recent that da/dN
were made reference to small cracks [25,26]. While crack sizes can be
addressed with reference the three regions I, II and III in Fig.3, it is not
clear how the crack size/time effects can be identified with the 'K since the
two parameter fit applies only to region II by the equation
da
log = plog(ǻK)+logA (24)
dN
With data fitted to a straight line, A can be determined from y-intercept

and p from the slope of the line in Fig. 3. The three regions I, II and III are
also referred to the micro, macro and fast crack growth in Fig.4(a). Use of
'K macro macro
micro or 'S micro can make the fatigue data to lie on a straight line as
shpwn in Fig. 4(b). In this way, linear interpolation can be used to connect
the results of regions I, II amd III and Eqs.(2) nd (3) are said to be form
invariant for microcracking and macrocracking in fatigue.
192 G. C. Sih
Fig. 3 Fatigue data representation of Eq.(1) for through cracks in two dimensions.
(a) Crack growth data (b) Invariant form of 'K or 'S

Fig. 4 Invariant form of crack growth rate in relation to region I, II and III: (a)
Crack growth data; (b)Linear line plot of 'K or 'S.
5.1 Crack surface tightness stress ratio
As mentioned earlier, accommodation for the opening and closing of the

crack surfaces is a major consideration for modeling the fatigue crack
growth behavior. This implies that the restraining stress Vo in relation to the
applied stress Vf or the ratio Vo/Vf should vary with the crack half length a.
Assuming that
0.65, for a 3mm

Vo °
®0.48421, for a 10mm (25)
Vf °0.01049, for a 65mm
¯
the bilinear relationship between the stress ratio Vo/Vf and the crack half
length a is adopted.
Vo 0.023684a 0.72105, for 1 d a d 10mm

® (26)
Vf ¯ 0.008613a 0.57034, for 10 d a d 65mm
The unit of length a is millimeter. The two stage bilinear relationship is

displayed in Fig. 5. The corresponding numerical values of the curve in Fig.
5 are summarized in Table 1.
5.2 Microcrack tip open segment ratio
As the micro/macro crack extends, the tip opening segment length d can
also have an effect. That is local intensification will depend on the relative
distance r/d which will increase with a as the microcrack becomes
macroscopic. This ratio is not sensitive for very small cracks. The behavior
Fig. 5 Bilinear relationship for stress ratio Vo/Vf versus crack half length a.
194 G. C. Sih
Table 1 Values of stress ratio Vo/Vf with the crack half length a.
Fig. 6 Length ratio r/d versus crack half length a.
Table 2 Values of length ratio r/d with the crack half length a.
a(mm) d8 10 40 60 65
r/d 0.73 1 5 7.7 8.3
is shown in Fig. 6 and the numerical results can be found in Table 2. The
'K macro
micro model of Eq.(14) is independent of the radial distance r. A value of
d=1mm will be used. The curve in Fig. 6 applies only to the 'S macromicro of Eq.
(23).
It suffices to know the trends of the curves in Figs. 5 and 6 for application
of the dual scale model. Moreover, only Fig. 5 is needed to make the form
of the crack growth rate relation in Eq.(14) invariant. However, better

correlation of fatigue data is obtained using Eq. (23) where Fig. 6 is also
used. This because Eq.(23) corrects for the mean stress effects whereas Eq.
(14) does not.
6. Linearization of Micro-/Macro-Data for Fatigue Crack Growth Rate
Referring to the data in [22] for the fatigue tests of 2024-T3 and 7075-T6
aluminum pre-cracked panels, the dual scale models based on 'K macromicro in
Eq. (14) and 'S macro

micro in Eq. (23) will be used such that data in regions I, II
and III as shown in Fig.3 would lie on a straight line without any bends. The
shear modulus ratio Pmicro/Pmacro will be set at two because its value is not
known to change with the crack length. The parameters V* and d* will
account for the transition of microcracking to macrocracking and vice versa.
Use will be made of the macroscopic properties of 2024-T3 and 7075-T6
given in Table 3. Data [22] for six different cases will be analyzed. They are
referred to the means stress Vm= (Vmax+Vmin)/2 and the stress amplitude Va=
VmaxVmin as shown in Table 4 with R being the mean stress ratio Vmin/Vmax.
Table 3 Mechanical properties of 2024-T3 and 7075-T6 aluminum alloy.

Yield Ultimate Shear modulus Poisson’s
strength strength Pmacro(kg/mm2) ratio
Material type
Vyd Vult (kg/mm2) Qmacro
(kg/mm2)
2024-T3 37.1 48.3 2812 0.33
7075-T6 47.3 53.0 2742 0.33
Table 4 Case classification of fatigue crack growth data.
Case Vm (kg/mm2) Va (kg/mm2) R
I 12 2.5 0.655
II 12 4.0 0.500
III 12 6.5 0.291
IV 9 2.5 0.565
196 G. C. Sih
V 9 4.0 0.385
VI 9 6.0 0.161
6.1 Discussion of results using Eq.(2)
Based on the numerical results provided earlier, the log of da/dN versus
log of 'K macro
micro may be plotted. The curves for 2024-T3 and 7075-T6
aluminum alloy are displayed in Fig. 7. Note that the data in all three
regions I, II and III can now be adequately represented by straight lines.
The slope for 2024-T3 signified by n is 3.532 which is smaller than that of
7075-T6 with n=4.179. This means that the crack growth rate of 2024-T3 is
slower than that of 7075-T6 which is to be expected. Not to be misled is
that the x-coordinate in Fig. 7 has been shifted and did not start from zero.
The full scale is exhibited in Fig. 8 such that the y-intercepts give log
C=-6.967 for 7075-T6 and -6.710 for 2024-T3. When the stress amplitude
Va is raised from 2.5 to 4.0 kg/mm2 for the same mean stress Vm= 12
kg/mm2 although the stress ratio R is reduced from 0.655 to 0.500, a slight
increase in the crack growth rate da/dN for both materials are seen. This is
Fig.7 Log crack growth rate da/dN v.s. log micro/macro stress intensity range
'K macro
micro for Case I.
Fig.8 Coordinates redefined for log crack growth rate da/dN v.s. log micro/macro
stress intensity range 'K macro
micro for Case I.
'K macro
micro for Case II.
the results of increase in the slopes of the straight lines for 2024-T3
(n=3.632) and 7075-T6 (n=4.297). The straight line fit in Fig. 9 is good
because most of the data points pertained to half crack length in the range
10 to 40 mm corresponded to region II.
198 G. C. Sih
A further increase of the stress amplitude to Va = 6.5 kg/mm2 with Vm= 12

kg/mm2 leads to additional reduction in R to 0.297. This shows that the
slopes of da/dN acquire further gain with n=4.224 for 2024-T3 and n=4.742
for 7075-T6 although the data gained more scatter as shown in Fig.10. The
detrimental effect of increase stress amplitude on fatigue life is exemplified
as in the use of the classical S-N curve where no consideration to crack
growth was made. Reduction of the maximum stress also has a noticeable
effect on the crack growth rate. Consider Case IV in Table 4 where Va = 2.5
kg/mm2 with Vm= 9 kg/mm2 and R = 0.565 being similar to that in Case I. A
complete set of data points for all three regions were obtained as exhibited
in Fig. 11. Decrease in the slopes of the curves confirms a reduction of the
crack growth rate when the stress amplitude is decreased. The effect of
stress amplitude starts to take over in Case V when Va = 4.0 kg/mm2 with
Vm= 9 kg/mm2 and R = 0.385. Fig. 12 shows that the slopes of the curves
are increased to n=3.725 for 2024-T3 and n=3.781 for 7075-T6. Further
aggravation of the crack growth rate is seen by the curves in Fig. 13 when
the stress amplitude is raised to Va = 6.0 kg/mm2. Now the slopes attain
even higher values of n=3.812 for 2024-T3 and n=4.511 for 7075-T6.
'K macro
micro for Case III.
'K macro
micro for Case IV.
'K macro
micro for Case V.
200 G. C. Sih
Fig.13 Log crack growth rate da/dN v.s.log micro/macro stress intensity range
'K macro
micro for Case VI.
The findings in Figs. 7 and 13 can be summarized in Table 5 for a

comparison of the data for the 2024-T3 and 7075-T6 material, particularly
with reference to the parameters C and n in Eq. (2) for the 'K macro
micro model.
The stress amplitude is seen to play a dominant role by the fatigue data in
Table 5. Cases I to III show that the increase in stress amplitude tends to
increase the crack growth rate by raising both the y-intercept log C and the
slope n. The reduction of maximum stress reflected by lowering the mean
stress Vm for Cases IV to VI did lower the da/dN as shown by Case IV but
the further increase in the stress amplitude again raised the absolute log C
and n values as summarized in Table 5.
Table 5 Stress amplitude and mean stress effects on C and n in Eq. (2).
7075-T6 2024-T3
Case
Log C n Log C n
I -6.961 4.179 -6.770 3.532
II -7.336 4.297 -7.067 3.632
III -8.907 4.742 -8.492 4.224
IV -6.455 3.765 -7.139 3.707
V -6.748 3.781 -7.410 3.725
VI -8.530 4.511 -7.854 3.812
6.2 Discussion of results using Eq.(3)
The difference between 'K macro macro

micro in Eq. (14) and 'S micro in Eq. (23) is
not insignificant and will vary for different test conditions and materials.
Based on the same crack growth data as those used for obtaining the curves
in Figs. 7 to 13, Fig. 14 displays a plot of log(da/dN) versus log( 'S macro
micro ).
All data points fell on straight lines, one for 2024-T3 and one for 7075-T6
with the respective slopes of m=2.690 and 3.394. These values are smaller
than those of n=3.532 and 4.179 in Table 5 for Case I of the 'K macro micro
model. The difference in slopes for the two materials using the two models
is about the same. Fig. 15 shows similar trends for Case II where the data
are more scattered as the stress amplitude was increased from 2.5 to 4.0
kg/mm2. Much more scatter of the data resulted in Case III where Va is
increased to 6.5 kg/mm2 with Vm kept at 12 kg/mm2. This is also
accompanied by an increase in the crack growth rates as illustrated in Fig.
16. The values of the slope m in Figs. 14 to 16, however, are much closer
together when the same data are represented by the 'K macro micro model shown
by the curves in Figs. 7 to 10 inclusive. A good fit of the data is found in Fig.
17 for Case IV. The two straight lines for 2024-T3 and 7075-T6 are nearly
parallel with slopes of m=3.036 and 3.055. The superiority of the fatigue
crack growth resistance for 2024-T3 does not show up for the applied mean
stress and stress amplitude. The 7075-T6 aluminum would have served
equally well. This illustrates the location specificity when selecting the use
of materials in non-uniform stress and/or energy density fields. The crack
growth behavior of the two materials start to deviate from one another as
the stress amplitude is raised to 4.0 kg/mm2 with a mean stress of 9 kg/mm2
as in Case V where the plot of 'S macro micro is given in Fig. 18. Case VI for
Va=6.0 kg/mm2 corresponds to the results in Fig. 19. Data for large stress
amplitudes tend to have more scatter. It is worthwhile to note the slopes m
in Table 6 and compare them to the values n in Table5. An important
conclusion is that the variations of the m values are much less than those for
n as the mean stresses are change. This is to be expected because the
'S macro
micro model accounts for the mean stress effect whereas the 'K micro
macro
model does not. This concludes the presentation of the numerical results for
the fatigue crack growth rate where the data were transformed onto a
straight line relationship. To be noted is that the position of the x-axis in
Figs. 14 to 19 has been shifted and did not start from the origin. The
difference in reference can be seen from the curves in Figs. 7 and 8. The
202 G. C. Sih
locations of the y-intercepts in Table 6 should now be clear. Moreover, the

log 'S macro macro
micro in Figs 14 to 19 are negative while 'S micro in Eq. (23) is
positive definite.
Fig. 14 Log crack growth rate da/dN vs. log micro/macro energy density range
'S macro
micro for Case I.
'S macro
micro for Case II.
'S macro
micro for Case III.
'S macro
micro for Case IV.
204 G. C. Sih
'S macro
micro for Case V.
'S macro
micro for Case VI.
Table 6 Stress amplitude and mean stress effects on B and m in Eq. (21).
7075-T6 2024-T3
Case
Log B m Log B m
I 8.314 3.394 6.286 2.890
II 8.098 3.394 6.224 2.888
III 7.268 3.153 7.052 3.303
IV 7.711 3.055 6.885 3.036
V 7.035 2.947 6.741 3.015
VI 7.992 3.326 6.632 3.019
7. Concluding remarks
Even though metal fatigue has been researched for more than a century,
full confidence is still lacking when prediction of fatigue life of structural
components is at stake.. By in large, the effort has been concerned with
relating the onset of rapid fracture to the initiation of cracks under low
stress but repeated many times. In more recent times, attention has been
focused on explaining the size and time scaling effects associated with the
fatigue crack growth rate da/dN models with reference to the commonly
known regions I, II and III. In particular, the use of the macroscopic stress
intensity factor range 'K for region I is questioned. The concerns are
whether data for micron size and smaller cracks could still be adequately
correlated by the two parameter relation. Hence, a fourth region for crack
nucleation has been proposed [25.26]. This region is effectively concerned
with nanometer size cracks with length a | 10-4 mm or smaller. Crack
nucleation may have to be identified with a fifth region. What this indicates
is that the discovery of defect initiating at smaller and smaller time and size
scales necessitates a more refined procedure for modeling the evolution of
crack initiation to propagation. Several issues are in need of additional
attention:
(1) Is it possible to retain the two parameter da/dN relation using 'K macro
micro
or some other parameter and describe nano-scale cracks assuming that
the line crack configuration can be retained?
(2) If the answer to (1) is positive, then will it be possible to establish an
invariant form of da/dN using 'K macro macro
micro or 'S micro ?
206 G. C. Sih
What has been learned from experience is that through crack specimens
are more conducive to the aircraft structure engineers who entail crack
length a |10-80 mm (macromechanical) that falls into region II. Material
engineers are more concerned with surface cracks due material
microstructure effects associated with region I where a |10-3 to10-1mm
(microstructural). The fourth region would presumably fall into the scale
range most sensitive to chemical reactions. This corresponds to a |10-4 mm
or smaller (nanochemical) [27]. Defects in this size range appear very early
and are deeply trapped to the interior of the material. The inherent coupling
of size and time scale possesses a hierarchy in the damage process that can
be referred to as macromechanical, microstructural and nanochemical.
Subatomic effects will no doubt be of interest to the biologists that may be
further classified with the prefix pico or femto. The point to be made is that
the creation of different imperfections, defects or cracks are intimately
related to the dissipated energy density of a particular physical process.
They are inherently paired in size and time.
It is clear that length alone is not adequate for distinguishing micro- and
macro-cracks. In fatigue, the time variable is accounted for by dN/dt which
is in fact the frequency of the fatigue cycles. Hence data can be represented
in terms of da/dt just as well as da/dN. Tightness of the crack mouth,
however, is regarded important because it can change microcracking to
macrocracking or the other way around under alternating tension and
compression loading. The clue that there prevails a very small segment of
the microcrack tip that does not close because of micro-asymmetry
prompted the discovery of double singularity stress field [13]. Two
competing singularities prevail one being stronger with the order 1/r0.75 and
the other weaker with the order 1/r0.25. The simultaneous consideration of
the double singularity is responsible for explaining the transitory behavior
of micro- and macro-cracking. In retrospect, any multiscale crack models
should have the capability to alter the morphology of the defect during scale
transition. At the same time, connection of the results at the different scales
can be best achieved by linear interpolation if the da/dN data can be made
to lie on a line. Such a possibility can also be realized in a transformed
plane that is often used in space transformation. An immediate extension of
the present approach would be the development of a triple scale model of
macro/micro/nano cracks.
References
[1] Griffith A A. The phenomenon of rupture and flow in solids, Philosophical

Transactions, Royal Society of London, Series A221 (1921) 163-198.
[2] Irwin GR. Fracture Mechanics, In: Structural Mechanics, Pergamon Press,
New York, (1960) 557-592.

[3] Sih GC. Mechanics of fracture initiation and propagation. The Netherlands:
Kluwer Academic Publishers, 1991.
[4] Sih G.C. Crack front behavior of mesofracture in non-equilibrium mechanics,
Journal of Physical Mesomechanics, 1(1) (1998) 77-87.
[5] Popov VL. and Kroner E. On the role of scaling in elastoplasticity, Journal of
Physical Mesomechanics, 1(1) (1998) 103-112.
[6] Sih GC. Thermomechanics of solids: nonequilibrium and irreversibility, J. of
Theoretical and Applied Fracture Mechanics, 9(3) (1988) 175-198.
[7] Sih GC. Some basic problems in nonequilibrium thermomechanics, eds. S.
Sienietyez and P. Salamon, Taylor and Francis, New York, 1992, 218-247.
[8] Schwab K. Henrikson E. A., Worlock J. M. and Roukes M. L. Measurement
of the quantum of thermal conductance, Nature, 404 (2000) 974-976.
[9] Roukes ML. Plenty of room indeed. In: The Edge of Physics, Scientific
America (2001) 92-100.
[10] Sih GC. Signatures of rapid movement of electrons in valence band region:
interdependence of position, time and temperature, J. of Theoretical and
AppliedFracture Mechanics, 45(1) (2006) 1-12.
[11] Sih GC. and Tang X. S. Dual scaling damage model associated with weak
singularity for macroscopic crack possessing a micro/mesoscopic notch tip, J.
of Theoretical and Applied Fracture Mechanics, 42(1) (2004) 1-24.
[12] Tang X.S. and Sih G. C. Weak and strong singularities reflecting multiscale
damage: micro-boundary conditions for free-free, fixed-fixed and free-fixed
constraints, J. of Theoretical and Applied Fracture Mechanics, 43(1) (2005)
5-62.
[13] Sih G.C. and Tang X. S. Simultaneous occurrence of double micro/macro
stress singularities for multiscale crack model, J. of Theoretical and Applied
Fracture Mechanics, 46(2) (2006) 87-104.
[14] Tang XS. and Sih GC. Evaluation of microstructural parameters for micro-/
macro-line crack damage model, J. of Theoretical and Applied Fracture
Mechanics,46(3) (2006) in press.
[15] Sih GC. and Tang XS. Simultaneous occurrence of double micro/macro stress
singularities for multiscale crack model, J. of Theoretical and Applied
Fracture Mechanics, 46(2) (2006) 87-104.
[16] Vasudevan AK. Sadananda K. and Louat N. A review of crack closure,
fatigue crack threshold and related phenomena, Material Science and
Engineering, A188 (1994) 1-22.
[17] Newman Jr. JC. Brot A. and Matias C. Crack-growth calculations in
7075-T7351 aluminum alloy under various load spectra using an improved
crack-closure model, Engineering Fracture Mechanics 71 (2004) 2347–2363.
[18] Noroozi AH. Glinka G. and Lambert S. A two parameter driving force for
fatigue crack growth analysis, International Journal of Fatigue, 27 (2005)
1277-1296.
[19] Jones R. Barter S. Molent L. and Pitt S. Crack growth at low K's and the
Frost-Dugdale law, Journal of the Chinese Institute of Engineers, Special
Issue inHonour of Professor G. C. Sih, 27(6) (2004) 869-875.
208 G. C. Sih
[20] Jones R. and Pitt S. On the Frost-dugdale law and the two parameter driving
force for fatigue crack growth analysis, Sixth Int. Conf on Fatigue Damage of
Structural Materials: Fatigue Damage VI, Hyannis, Massachusetts, Sept.
17-22, 2006.
[21] Paris PC. The growth of cracks due to variations in load, Ph. D. Dissertation,
Department of Mechanics, Lehigh University (1962).
[22] Broek D. and Schijve J. The influence of the mean stress on the propagation
of fatigue cracks in aluminum alloy sheets, National Aeronautics and
Astronautics Research Institute NLR-TN M. 21111, Amsterdam (1963) 1-57.
[23] Liu HW. Fatigue crack propagation and applied stress range, ASME Trans., J.
Basic Eng. 85D(1) (1963) 116-122.
[24] Liu HW. A review of fatigue crack growth analyses, J. of Theoretical and
Applied Fracture Mechanics 16 (1991) 91-108.
[25] McDowell DL. Multiaxial small fatigue crack growth in metals, Int. J.
Fracture, 19(1) (1997) S127-S135.
[26] Wang CH. Effect of stress ratio on short fatigue crack growth, ASME Journal
of Engineering Materials and Technology, 118 (1996) 362-366.
[27] Sih GC. Crack tip system for environment assisted failure of nuclear reactor
alloys: multiscaling from atomic to macro via mesos, J. of Pressure
Equipment and Systems, 3(2005)1-25.
Fatigue Crack Growth Rate of Cable-Stayed Portion of
Runyang Bridge: Part I–Cable Crack Growth Due to
Disproportionate Cable Tightening/Loosening and
Traffic Loading
G. C. Sih 1, 2 * and X. S. Tang 3
1
2
University of Science and Technology, Shanghai 200237, China
3
School of Bridge and Structural Engineering, Changsha University of
Science and Technology, Changsha, Hunan 410076, China
*
Email: gcs8866@yahoo.com, gcs1@lehigh.edu
Abstract
Relatively little is known about the fatigue crack growth behavior of

bridge cables. Such information can become increasingly more important
in time for the cable-stayed bridges whose structural integrity can be
greatly affected by the degradation of the individual cable material due to
aging as well as defect growth. Moreover, crack initiation and propagation
are intimately related such that they have to be treated as one of the same
process. This requires a dual scale fatigue crack growth model involving
both micro- and macro-cracking where the interaction of three mi-
cro/macro material, geometric and load parameters is considered. Based on
the design data for the stresses in the 52 cables of the Runyang cable-
stayed bridge, three typical cables referred to as #28, #38 and #50 are se-
lected for analyzing fatigue crack growth under traffic and no traffic condi-
tions. The initial tension in the cable can vary by using D=1 as the refer-
ence such that deviation from the design condition may be regarded as
tightening and loosening. Under traffic, initial tightening and loosening of
the cable can enhance and impede fatigue crack growth, respectively.
When traffic is removed, initial tightening and loosening of the cable can
impede and enhance fatigue crack growth. Opposing effects of fatigue
crack growth are observed for cables with and without traffic. These re-
sults are discussed in terms of crack growth rate and crack length as a
function of the fatigue cycles assuming that the cable has a life of two mil-
lion cycles. Substantial variance in fatigue crack growth is found. Crack
210 G. C. Sih and X. S. Tang
length can vary from a few mm to 50 mm and more before reaching the
onset of rapid fracture. The stiffness of the cable is also affects crack
growth. Each cable is found to behave as a structure of its own. Replace-
ment of cables appears to be eminent during the life span of cable-stayed
bridge because of the wide variance of fatigue lives of the cables. This
means that some cables will fail while others may still have ample remain-
ing life.
1. Introduction
The critical components in a cable-stayed bridge are no doubt the cables

each of them behaves differently and acts as a structure in itself. The ef-
fects of the main load carried by the bundles of high strength steel wires
are susceptible to change in time because aging and degrading of the mate-
rial property will occur in addition to corrosion caused by the environment
conditions. The current construction of the cables inhibits and prevents the
successful use of non-destructive testing despite extensive efforts are being
made in health monitoring. For bridges that have been in service after a
decade or more the situation becomes even more uncertain because dam-
age accumulates at the different scales, say micro and macro. This fact was
recognized from the earlier research on dual scaling [1,2]. Empirical mod-
els may not be adequate because they are unable to know the presence of
tightly closed cracks even though the material may have separated locally.
In this case, there is no signal to transmit for detection. To reiterate, this is
especially true for situations where the faults can closed and remain dor-
mant but they can open to inflict serious damage under overload or unex-
pected accidents such as earthquakes or vehicle impact. The undetected
flaws can alter a situation of fail-safe to one that triggers disaster. On site
health monitoring is only as effective as knowing the locations where
gages should be placed. The critical sites are often hidden and inaccessible
for inspection until damage becomes excessive. Not enough emphasizes
can be placed on knowing the potential sites of damage for a given design
and how damage would develop in service. An understanding of fail-safe
in the design of large structures [3] is vital for the mitigation or prevention
of potential injury/life and property. The period for the inspection of
bridges is every two years. This is a mandatory requirement of the US Na-
tional Bridge Inspection Standard (NBIS). The recent disaster of the I-35W
bridge in Minnesota [4] is indicative of the ineffectiveness of implement-
ing the standards.
Although the current state-of-the-art of bridge maintenance and inspec-
tion recognizes the importance of flaw detection caused by fatigue for steel
structure bridges [5,6] but relatively little attention [7-9] has been devoted
Fatigue Crack Growth Rate of Runyang Bridge: Part I 211
to the long term fatigue of cables for suspension and cable-stayed bridges.
Most of the cable fatigue works address the initial design requirement
[10,11] which can cast false confidence on the interpretation of field data,
particularly when uncertainties prevail with the assumed boundary condi-
tions in design. This problem is particularly acute for the cable-stayed
bridge that may possess more than one hundred cables and each of them
can have widely different behavior. It means that some of them may be on
verge of breaking while others may still have ample remaining life. The
study in [9] has shown such effects to be prevalent by using the design
data of the cable-stayed portion of the Runyang bridge [12]. The diversifi-
cation of cables’behavior is typical as they can experience loosening and
tightening every hours of the day where the traffic may vary from full to
no load, not counting for wind that also causes variation in cable tension.
This becomes increasingly problematic for aging cable-stayed bridges as
the cables do not deteriorate at the same rate. Some may have ample re-
maining life while others are in need of replacement. The present moni-
tored data are not capable of sorting out the degree of cable damage be-
cause no confidence can be placed on their association with the actual
physical damage. This problem cannot be alleviated by embedding sensors
into the cables during fabrication. There is not a direct correspondence be-
tween the sensor and the local damaged area. It is necessary to have a vali-
dated software that can address fatigue damage for interpreting the sensor
data, preferably over an area that is large in comparison with the damage.
Not enough emphasis can be placed on the latter, especially when micro-
cracking and macrocracking are both mingled in the signal. It was demon-
strated in [9] that the crack growth behavior in two of the Runyang bridge
cables differed widely when fatigued under traffic and no traffic condi-
tions. The discrepancy can be the difference of stable crack growth and the
onset of rapid fracture. Additional work is carried out in this communica-
tion to offer a more complete discussion of the fatigue crack growth be-
havior of cables for the Runyang bridge. Considered in particular is the
crack growth rates of some typical cables showing that the variance may
be relevant for the future development of maintenance and inspection pro-
cedures.
2. Problem Statement
The Runyang bridge consists of two portions as shown in Fig. 1(a) for the
cabled-stayed portion and Fig. 1(b) for the suspension portion. A close
view of the cables for the cable-stayed portion can be found in Fig. 2. Note
that the cables are mounted onto the two towers and anchored at the
girders of the bridge. The distance between the two compressed towers is
406m. The structure is sufficiently flexible such that it can sway. Because
of the load variations, tightening and loosening of the cables is expected.
Considered are conditions for the bridge under traffic and no-traffic condi-
tions. The large variations of fatigue lives for the cables are identified for
several different scenarios of cable tightening and loosening that can be
caused by climate changes, a particular concern in recent years. The results
are needed for the development of inspection procedures for the health
monitoring of the Runyang bridge. Similar analysis should be done for the
suspension portion of the bridge.
(a) Cable-stayed portion (b) Suspension portion

Fig. 1 Close view of the cable-stayed and suspension portions of Runyang bridge.
Fig. 2 The Runyang cable-stayed portion of the bridge with 406m span.
2.1 Cable forces with and without traffic
Assuming symmetry, only 52 of the total of 104 cables of the Runyang ca-
ble-stayed bridge need to be shown. The tension in the cable with no traf-
fic and with traffic is as shown, respectively, in Figs. 3 and 4. The position
of each cable is shown and numbered from left to right. The largest forces
prevail in cables #3 and #50. They are located, respectively, third from the
left and right end position. The small differences in the forces in cables
#1~#26 and #27~#52 can be ignored. To be examined are the fatigue crack
growth characteristics in cables #28, #38 and #50 where cable #38 corre-
sponds to the minimum tension.
Fig. 3 Load in cables without traffic.
Fig. 4 Maximum and minimum load in cables with traffic.
2.2 Multiscale application of fracture mechanics
Modern bridge design takes into account the influence of material proper-
ties such as high to very high strength, high temperature and corrosive re-
sistance. This implies that the microstructural effect of the material comes
into play. From the fracture mechanics view point, it is necessary to con-
sider the behavior of small cracks that might be dormant under one condi-
tion and grow to large cracks under another condition. Microcracks and its
transition to macrocracks must be accounted for. The traditional sigmoidal
curve as illustrated in Fig. 5 can no longer be represented by the two-
parameter crack growth rate relation [13] because it remains straight only
for the portion Region II for macrocrack growth (or long cracks). Micro-
cracks grow in Region I referred to as the threshold. Region III pertains to
fast crack propagation heading towards the onset of global instability.
Non-destructive testing should therefore be designed to detect the stable
macrocrack growth in Region II.
Fig. 5 Traditional two-parameter fatigue crack growth rate.
Fig. 6(a) shows the crack growth curves partitioned with reference to the
crack size from nano to macro. In order to preserve the traditional frame-
work of fracture control of using only two empirical parameters, a multis-
cale model of crack growth was advanced [14,15] by replacing the range
of stress intensity factor 'K [13] for macrocracks only with 'K macro
micro
or
'S macro
micro
that can account for the transition of microcracks to macrocracks
or vice versa. The extension to triple scale involving nano, micro and
macro is straightforward. In general the use of 'S macro
micro
is preferred because
it accounts for the effect of mean stress. Examples of using 'K macro
micro
or
'S macro
micro
can be found in [14,15]. In a log da/dN versus log 'K macro
micro
or
log 'S macro
micro
domain, the three Regions I, II and III no longer appear as sig-
moidal. Instead, a straight line is obtained. This is shown in Fig. 6(b). This
enables the shifting of the results from nano to micro and from micro to
macro or directly from nano to macro.
(a) Crack growth curve (b) Crack growth rate

Fig. 6 Form invariant property of multiscale crack growth rate relation.
2.3 Stress amplitude and mean stress
The simultaneous consideration of stress amplitude Va=VmaxVmin and

mean stress Vm=(Vmax+Vmin)/2 suggests the application of 'S macro
micro
which
will be made use of in the sequel. Here, Vmax and Vmin are the maximum
and minimum cyclic applied stress, respectively. Further distinction must
be made because they are different under traffic and no-traffic conditions.
The superscripts D refers to traffic condition and o to traffic absent condi-
tion. The values of V omax and V Dmax for the 52 cables of the Runyang bridge
can be found in Table A1 of Appendix I of [9]. Two Cases I and II will be
considered; they are defined as
DV Dmax V omax
Case I: V a DVDmax V omax , V m (1)
2
V Dmax DV omax
Case II: V a V Dmax DVomax , V m (2)
2
Case I in Eq. (1) refers to tightening of the cables that correspond to
D=1.0, 1.2, 1.4 and 1.6 with D=1 (normal tension) as the reference. Nu-
merical values of Va, Vm and VaVm for the 52 cables in Runyang bridge
with D=1.0, 1.2 and 1.6 are given in Tables A2, A3 and A4, Appendix II of
[9]. Case II in Eq. (2) refers to loosening of the cables that correspond to
V=0.91, 0.94, 0.97, 1.00 (normal tension). Henceforth, ĮıĮmax will be re-
ferred to as “max. stress with traffic” and Įıomax as “max. stress without
traffic.” Remember that D=1.0 stands for normal tension while Dt1ҏ and
Dd1ҏrefer, respectively to more and less cable tension. Cable tightening and
loosening are defined in terms of Va and Vm as given by Eqs.(1) and (2).
2.3.1 Case I: cable tightening and loosening with traffic
Fig. 7 displays a plot of the product VaVm for the cables as a function of the
cable position from left to right numbered as 1 to 52. Note that the product
VaVm fluctuates according to the variations of the tension in the cables due
to varying traffic as indicated in Fig. 4. The end cables peak with a dip for
the middle cable. The minima tension occur in cables #14 and #38. Sym-
metry about cable #26 is not quite preserved. The cable tightens with in-
creasing D from 1.0 to 1.6 with the maximum VaVm at D=1.6 which will
tend to reduce fatigue life. The peaks of the curves corresponding to cable
#3 and #50 rise very quickly with the parameter D.
Fig. 7 Product VaVm of cables versus position of cables for Case I (tightening) with
D=1.0, 1.2, 1.4 and 1.6 under varying traffic.
Numerical values of Va, Vm and VaVm for the 52 cables of the Runyang
bridge with D=0.97 and 0.91 can be found in [9] as mentioned earlier.
Loosening of the cable under traffic has a beneficial effect to fatigue life.
This can be seen from Fig. 8 where decreasing D from 1.00 (normal ten-
sion) to 0.91 (less tension) reduces the amplitude of VaVm although the
amount of reduction in Fig. 8 is not as pronounced as the rise in Fig. 7 for
cable tightening. It can be said that cable tightening under varying traffic
tends to increase the product VaVm while cable loosening under varying
traffic decreases the product VaVm that is beneficial.
Fig. 8 Product VaVm of cables versus position of cables for Case I (loosening)with
D=0.91, 0.94, 0.97 and 1.00 under varying traffic.
2.3.2 Case II: cable tightening and loosening without traffic
Consider now Case II in Eq. (2) where the bridge is not under traffic while
the tension in the cables can still vary that is accounted for by D as defined
in Eq. (2). Numerical values of Va, Vm and VaVm for the 52 cables of the
Runyang bridge with D=1.00, 1.04 and 1.12 are given in Tables A7, A8
and A9 of Appendix IV [9]. Case II refers to max. stress without traffic
while Dt1 corresponds to more tension in the cables. It can be concluded
from the results in Fig. 9 that tightening of the cable will now have a bene-
ficial effect as the product VaVm will decrease when D is increased from
1.00 to 1.12 corresponding to “tightening”. Fig. 10 provides the results for
loosening of the cables without traffic. This increases VaVm as D is de-
creased from 1.0 to 0.7. Therefore loosening of the cable is of no help in
the absence of traffic. Numerical values of Va, Vm and VaVm for the 52 ca-
bles of the Runyang bridge with D=0.9 and 0.7 are given, respectively, in
Appendix V of Tables A10 and A11 [9]. The results in Figs. 7 to 10 inclu-
sive are summarized in Table 1.
Table 1. Effect of cable tightening and loosening on VaVm with and without traf-
fic.
Cases VaVm VaVm
Case I (traffic) Decrease (loosening) Increase (tightening)
Case II (no traffic) Decrease (tightening) Increase (loosening)
Fig. 9 Product VaVm of cables versus position of cables for Case II (tightening)
with D=1.00, 1.04, 1.08 and 1.12 in absence of traffic.
Fig. 10 Product VaVm of cables versus position of cables for Case II (loosening)
with D=0.7, 0.8, 0.9 and 1.0 in absence of traffic.
3. Multiscale Micro-/Macro-Crack Model
Multiscale crack growth models [7-9] must not only distinguish the size
and time effects of small and large cracks but they must also be able to
treat the transition region behavior of the cracks, say the transfer of micro-
to macro-cracking if only two scales are involved. Hence both the micro-
and macro-scale material properties must be considered in addition to mi-
cro- and macro-length that must be differentiated. A dual scale model must
contain at least three parameters d*, P* and V* such as
d , * P micro , * V o (3)
d* P V
do P macro Vf
in which d* is the normalized characteristic dimension compared to the mi-
cro length do. A comparison between the micro- and macro-modulus is re-
flected by P* while V* accounts for the ratio of the material restraining
stress Vo to the applied stress Vf such that Vo/Vf must always be less than
one. This corresponds to the threshold for the onset of crack growth. Such
a behavior manifests the closing of the microcrack in contrast to the mac-
rocrack that is defined to possess a macro-distance crack opening. The
transition of microcracking to macrocracking can be essential in fatigue
where the applied load switches repeatedly from tension to compression or
vice versa for each cycle.
3.1 Fatigue crack growth of cable
The bridge cable is a composite made of strands of steel wires wrapped in

polymeric tubes with spacing filled by polymers. Although the properties
of the cable will be obtained by approximation but they will be different
from those for the wire. The details of which are discussed in [9]. It is
more pertinent to consider the growth rate da/dN of a fatigue crack with
length a (cm) in a cable as given by
da \
<o ( 'S macro
micro )
(4)
dN
where <o and \ can be found by test data in a regular fatigue test. Here,
<o is the y-intercept and \ the slope of the curve for the straight line por-
tion of the log-log plot of Eq. (4). The crack length can be computed by in-
tegrating Eq. (4). The results for \=1 is given by
a(N)=a o eq(N-No ) , for \=1 (5)
In Eq. (5), N stands for the load cycle as specified by the traffic. By mak-
ing the contractions
q <q o , for \=1 and q o (1 2Q micro )P * (1 V* ) 2 d * (6)
and using the expression for 'S macro

micro
given as [9]
2(1 Q macro ) 2 V a V m d o
'S macro
micro q oa (7)
P macro r
Under these considerations, there results
2(1 Q macro ) 2 V a V m d o , for \=1 (8)
< <o
P macro r
Eq.(4) can thus be simplified to the form

da
qa , for \=1 (9)
dN
Keep in mind that the radial distance from the crack tip r in Eq. (7) and (8)
can be micro or macro in size.
3.2 Fatigue properties of cable obtained from steel wire
Although fatigue tests of cable and steel wire have known to be made, but
the data are not readily available except for the work in [16,17] that was
made use of in [8,9] to study fatigue crack growth in a steel wire. For this
material, the log-log plot of the da/dN versus 'K data based on the classi-
cal stress intensity range can be found in Fig. 11. Again the conversion of
the log-log da/dN versus 'K to log-log da/dN versus 'S macromicro
has already
been given in [9] and will not be repeated. Since Eq. (4) can be written as
da
log( ) log(<o ) \ log('S macro
micro )
(10)
dN
The y-intercept <o and the slope \ of the line represented by Eq.(10) can
be found as
\=1, <o=4.2u10-8 for the cable. (11)
The relevant material and geometric parameters for the cable can be ob-
tained from
Ec = EEw, for Ed1 (12)
where Ec and Ew denote the Young’s moludus for cable and wire, respec-
tively.
The following numerical values are used:
P micro =2, * V o =0.3, * d =1, d =10-3mm, P
P* V d o micro=0.4, r/do=1
P macro Vf do
(13)
and Pmacro=Pw, Emacro=Ew and Qmacro=Qw while the initial values of ao=
0.01mm, No=0.1u106 apply to fatigue. For the wire with diameter
Dw=5mm, it follows that
Ew=199.82GPa, Qw=0.3, Pw=Ew/[2(1+Qw)]=76.85GPa. (14)
The normalized quantities in Eqs.(13) together with do and Qmicro are as-
sumed to be valid for both the cable and wire. Appropriate value of E in
Eq. (12) such as E=1.0, 0.9, 0.8, 0.7, --- can be selected together with the
data in Table 2 to study the crack growth rate relation for the cable. The
subsequent results are generated for cables with an assumed life of two
million cycles, i.e., Nf=2u106 cycles.
Fig. 11 Log-log plot of da/dN versus 'K for high strength steel data from [16].
Table 2 Mechanical properties of bridge steel wire [16,17].

Elastic modulus Ultimate stress Ultimate strain
Ew (GPa) w w
V ult (MPa) H ult (%)
199.82 1722.55 5.5
3.3 Equivalent surface crack growth
Recall that <o and \ in Eq. (11) were found empirically from a plate
specimen and not that of the assumed shape of a round bar for the cable or
the wire. This, however, would not affect the qualitative feature of the re-
sults since the shape factor can be absorbed into <o quantitatively. To this
end, an equivalent crack length a can be used regardless of the specific
crack shape. More precisely, the clam shaped fatigue surface crack in Fig.
12(a) can be replaced by a straight edge crack in Fig. 12(b). For a homo-
geneous cross section cable, the diameter Dc is approximated by a single
edge crack specimen of thickness Dc. This completes the determi-nation of

the empirical parameters for the crack growth rate relation of the cable.
(a) Clam shell surface crack (b) Equivalent edge crack

Fig. 12 Surface crack simplified as through crack in cable and wire.
4. Crack Growth Rate of Bridge Cable
According to Eq. (12), the parameter E controls the difference between the
stiffness of the cable as compared with that of the steel wire with its
modulus Ew as given in Table 2. Since the cable stiffness is less than that
of a single steel wire E as defined in Eq. (12) is always less than one since
the cable is not completely filled by wires. For practical consideration,
E=0.7 will be assumed for the discussion of numerical results. Making use
of the numerical values of the relevant parameters as given by Eqs. (13)
and (14). The crack growth rate da/dN can be computed directly from Eq.
(4) with the help of Eq. (11). The quantity D can be made to vary depend-
ing on whether the cable is tight for Dt1 or loose for Dd1. These condi-
tions are specified by Eqs. (1) and (2).
4.1 Case I: Cable tightening and loosening for E=0.7
The variations of the crack growth rate with the number of fatigue cycles
N will be examined for three typical cables which are referred to as #28,
#38 and #50 under traffic and no traffic conditions.
4.1.1 Cable #28
Fig. 13 displays the variations of da/dN with N for cable #28. Note that for
D less than 1.2, the crack growth rate is negligibly small up to the two mil-
lion cycles. As D is gradually increased, da/dN tends to increase with N.
For Dt1.6, the slopes of the curves in Fig. 13 tend to become vertical when
N=1.4u106 cycles is approached. This implies that tightening of the cable

would result in earlier unstable growth which is undesirable.
Fig. 13 Crack growth rate da/dN versus number of cycles N for cable #28 and
Case I (tightening) with Dt1 and E=0.7 under traffic.
Case I (loosening) with Dd1 and E=0.7 under traffic.
Loosening of the cable for Dd1 on the other hand leads to beneficial effects
because it decreases the crack growth as illustrated in Fig. 14 for the same
material and geometric parameters as those used to obtain the data in Fig.
13. With the same D of 1.6, it can be seen that the slope of the top curve in
Fig. 14 still correspond to stable crack growth after N=2u106 cycles while
the same curve in Fig. 13 has nearly reached unstable crack growth at
N=1.4u106 cycles. This difference is significant and indicates that the ten-
sion of the cables for cable-stayed bridges should be monitored and hence
can be adjusted when the need arises.
4.1.2 Cable #38
The situation in cable #38, however, is entirely different. Even for D=1.6,
Fig. (15) shows that da/dN | 1.9u10-6mm/c at N|2u106 cycles as the fa-
tigue crack is still propagating in a stable manner. This corresponds to the
tightening condition of cable #38. Fig. 16 indicates that further slow down
of the crack results when D is made to be less than one for loosening of the
cables.
Case I (tightening) with Dt1 and E=0.7 under traffic condition.
Case I (loosening) with Dd1 and E=0.7 under traffic condition.
4.1.3 Cable #50
The crack growth behavior of cable #50 in Fig. 17 is more severe than that
of cable #28 in Fig. 13. Comparing the curves for D=1.6 and N|1.3u106
cycles, Fig. 17 gives an approximate crack growth rate of da/dN|50u10-5
mm/c and Fig. 13 gives da/dN|10u10-5mm/c, a factor of 5 difference. That
is the crack growth rate in cable #50 is five times greater than that in cable
#28. The non-uniformity of the tension in cables #28, #38 and #50 are ex-
hibited.
Case I (tightening) with Dt1, E=0.7 with traffic.
Case I (loosening) with Dt1, E=0.7 with traffic.
Loosening of the cables #28 and #50 under traffic can impede crack
growth. The disparity between the results is similar and can be seen from
Figs. 14 and 18.
4.2 Case II: Cable tightening & loosening for E=0.7
Instead of varying the cable tension under the traffic condition, the varia-
tion of cable tension without traffic can be different. Hence, the effects
must be studied separately. Even though the same symbol D is used for
D V Dmax and D V omax , the individual values of D are different. This will be
shown for the results of cables #28, #38 and #50.
4.2.1 Cable #28

Plotted in Fig. 19 are the crack growth rate as a function of N for cable #28
where Eq. (2) has been implemented with D=1.04, 1.08 and 1.12 that cor-
respond to increasing of initial cable tension. This increase is seen to be
beneficial as the curves with D=1.08 and 1.12 give lower da/dN than those
with curves with D=1.00 and 1.04 corresponding to less tension. For
N|2u106 cycles, da/dN|1.8u10-6mm/c for D=1.00 as compared to da/dN|
0.6u10-6mm/c for D=1.12. That is higher initial tension cable under no traf-
fic can reduce the crack growth rate by three or 3.00 times after two mil-
lion cycles. This effect is not significant for low fatigue cycles. Decrease
of initial cable effects without traffic are shown in Fig. 20 for D=0.7, 0.8
0.9 and 1.00. Again the effects become more pronounced with increasing
fatigue cycles. The maximum difference is about 5.3 times in da/dN at
N|2u106 cycles.
Fig. 19 Crack growth rate da/dN versus number of cycles N for cable #28and Case
I (tightening) with Dt1, E=0.7 without traffic.
Case I (loosening) with Dt1,E=0.7 without traffic.
4.2.2 Cable #38

The variations of da/dN with N in cable #38 is quite different. First of all,
the crack growth rate as shown in Fig. 21 for each D from 1.00 to 1.12 is
nearly constant for all fatigue cycles from 105 to 2u106 cycles. They differ
from the range (0.75-1.23)u10-7 to (0.85-1.51)u10-7mm/c for Dt1. That is
when the initial cable tension is increased. The results for decreasing the
cable tension with D=1.00, 1.04, 1.08 and 1.12 can be found in Fig. 22.
The change of da/dN with N is similar to the trends of the curves in Fig. 21
except that the D values are now smaller than one and the curves tend to
increase more as N is increased.
Case II (tightening) with Dt1 and E=0.7 in absence of traffic.
Case II (loosening) with Dd1 and E=0.7 in absence of traffic.
4.2.3 Cable #50
Cable #50 behaves more like cable #28 where the change in da/dN is small
for small N and the difference increase with increasing N as shown in Fig.
23. For N|2u106 cycles, da/dN|3.8u10-6mm/c for D=1.00 in contrast to
da/dN|1.3u10-6mm/c for D=1.12. That is higher initial tension cable under
no traffic can reduce the crack growth rate by 2.93 times after two million
cycles. This factor is slightly lower than 3.00 for cable #28 in Fig. 19. Fig.
24 shows the results for decreasing the initial tension of cable #50 which
corresponds to Dd1. The four curves with D=0.7, 0.8, 0.9 and 1.00 indicate
that there are hardly any differences in da/dN for small N but the curves
start to spread out appreciably at N=2u106 cycles, at which point the dif-
ference in the crack growth rate reaches a factor of 6.67. This is signifi-
cant. It again reflects the non-uniformity of tension in the cable can lead to
appreciable difference in the crack growth rate. This implies some cables
may fail by fatigue much earlier than others.
5. Crack Length as a Function of Fatigue Cycle
The da/dN curves in Figs. 19 to 24 may be integrated to yield the results

for crack length as a function of fatigue cycles. Such information can be
used to check with test data and used in fracture control. The following
discussions pertain to cables #28, #38 and #50.
Fig. 23 Crack growth rate da/dN versus number of cycles N: for cable #50 and
Case II tightening) with Dt1, E=0.7 without traffic.
Case II (loosening) with Dd1 and E=0.7 in absence of traffic.
5.1 Case I: Tightening & loosening for E=0.7
In view of Eq. (9), the crack length a can also be obtained directly from
da/dN without integration since q in Eq. (9) can be found from Eqs. (6)
and (8). This procedure happens to apply for the present case of \=1 as
stated in Eq. (11).
Fig. 25 Crack depth a versus number of cycles N for cable #28 and Case I (tight-
ening) with Dt1 and E=0.7 under traffic.
Fig. 26 Crack depth a versus number of cycles N for cable #28 and Case I (loosen-
ing) with Dd1 and E=0.7 under traffic.
5.1.1 Cable #28
The reference crack growth curve is given by D=1.0 as shown in Fig. 25. It
remained nearly constant for the range of fatigue cycles N up to two mil-
lion. As the cable is tightened gradually with D=1.1, 1.2, ---, 1.6, the
curves start to deviate from the reference with increasing crack length. At
N=1.4u106 cycles and D=1.5, the crack depth is 25mm. This length is dou-
bled to 50mm at the same fatigue cycles when the cable tension in raised
with D=1.6. In this case, increase cable tension tends to enhance crack
growth. On the other hand, a decrease of the initial cable tension with Dd1,
crack growth can be impeded. Fig. 26 shows the four curves for D=1.00,
0.97, 0.94 and 0.91. A crack depth of 3.1 mm at N=2u106 cycles and
D=1.00 reduces to 1.8 mm when the initial cable tension is decreased with
D=0.91. These effects are different for different cables.
3.2
3 Case I (tightening), Cable 38
-6
2.8 \=1<R=1.06X10 , E=0.7
Max. stress with traffic •
Crack depth a (mm)
2.6
2.4
2.2 D
D
2 D
D
1.8 D
1.6 D
D
1.4
1.2
1
0.8
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
6
Number of cycles N X 10
Fig. 27 Crack length a versus number of cycles N for cable #38 and Case I (tight-
ening) with Dd1 and E=0.7 under traffic.
1.4
Case I (loosening), Cable 38
-6
\=1<R=1.06X10 , E=0.7
1.3 Max. stress with traffic•
Crack depth a (mm)
1.2
1.1
D
D
1 D
D
0.9
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
6
Fig. 28 Crack length a versus number of cycles N for cable #38 and Case I (loos-
ening) with Dd1 and E=0.7 under traffic.
5.1.2 Cable #38
The results for cable #38 are summarized in Fig. 27 for Case I where the
initial cable tension is increased with D=1.0, 1.1, ---, 1.6. All the curves
start with an initial crack length of 1 mm. They then increase in amplitude
with increasing N. The magnitude of the reference curve D=1.00 increase
very slowly with N while the other curves increase faster as D is increased.
A factor of 3.2 in crack depth is gained at N=2u106 cycles for D=1.6. Fig.
28 further shows the effects of decreasing the initial cable tension with
D=1.00, 0.97, 0.94 and 0.91. Not much change can be observed as all of
the curves in Fig. 28 are bunched together.
5.1.3 Cable #50
Crack growth is most pronounced in cable #50 and the results are sensitive
to the initial cable tension. For D=1.6, it took 1.1u106 cycles to advance the
crack from 1 mm to 50 mm whereas 1.5u106 cycles were required for the
crack to growth the same depth if D is reduced from 1.6 to 1.4. An addition
of 400,000 cycles are needed if the initial increase in cable tension were
reduced by the equivalent amount of 0.2 in D. Again the change in crack
growth by decreasing the initial tension is not significant. This can be seen
from the curves in Fig. 30.
5.2 Case II: Tightening & loosening for E=0.7
Similar behavior can be demonstrated when the cable tension is altered

under no traffic conditions although the effects are not as dramatic.
ening) with Dt1 and E=0.7 under traffic.
Fig. 30 Crack depth a versus number of cycles N for cable #50 and Case I (loosen-
ing) with Dd1 and E=0.7 under traffic.
Fig. 31 Crack depth a versus number of cycles N for cable #28 and Case II (tight-
ening) with Dt1 and E=0.7 in absence of traffic.
5.2.1 Cable #28
Fig. 31 plots the crack depth a as a function of the cycle number N for
D=1.00, 1.04, 1.08 and 1.12. First of all, the change in crack depth is rela-
tively small even when N reached two million cycles. Moreover, the crack
growth ranged from 1.9 to 3.1 mm which is small compared to 50 mm and
beyond under traffic conditions. Refer to Fig. 25 for cable #28. The oppo-
site effects are seen when comparing the results of Fig. 26 with traffic to
those in Fig. 32 without traffic. For D=0.7, 0.8, 0.9 and 1.0, Fig. 32 shows
that the crack growth range is from 2.8 to 7.5 mm at N=2u106 cycles as
compared to the range of 1.8 to 3.1 mm in Fig. 26 for the same cable under
traffic. Contrary to intuition, more crack growth occurred for cable tension
that corresponds to no traffic conditions.
Fig. 32 Crack depth a versus number of cycles N for cable #28 and Case II (loos-
ening) with Dd1 and E=0.7 in absence of traffic.
5.2.2 Cable #38
Maximum crack growth in cable #38 ranged from 1.14 to 1.25 mm at

N|2u106 cycles under no traffic conditions. With traffic, the crack growth
range was found to be about 1.20 to 3.20 mm at N|2u106 cycles as shown
by Fig. 27. This trend, however, is not always observed as it was seen for
cable #28 and will also be different for the case when the initial cable ten-
sion is decreased, i.e, for Dd1. Referring to Fig. 34, the crack growth range
of 1.23 to 1.50 mm at N|2u106 cycles under no traffic turns out to be lar-
ger than that of 1.12 to 1.26 mm in Fig. 28 when the cable is under traffic.
Up to now, it is seen that each cable behaves differently, a very typical fea-
ture of the cable-stayed bridge.
1.4
Case II (tightening), Cable 38
-6
\=1<R=1.06X10 , E=0.7
1.3 Max. stress without traffic•
Crack depth a (mm)
1.2
1.1
D
D
1 D
D
0.9
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
6
1.6
Case II (loosening), Cable 38
-6
1.5 \=1<R=1.06X10 , E=0.7
Max. stress without traffic•
Crack depth a (mm)
1.4
1.3
1.2
1.1 D
D
D
1 D
0.9
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
6
5.2.3 Cable #50
The behavior of cable #50 without traffic differs significantly from that
under traffic. For Dt1 without traffic, crack growth is very limited as it can
be seen from Fig. 35 where the range of crack growth after N|2u106 cy-
cles is only 2.54 to 4.54 mm. This is significantly different from the corre-
sponding results under traffic in Fig. 29 where crack growth extended into
the range of 50 mm and beyond at D=1.3 even though N has reached only
1.9 million cycles.
For Dd1 and N|2u106 cycles, the crack growth range of 4.50 to 14.1 mm
in Fig. 36 without traffic is greater than the range of 2.35 to 4.55 mm in
Fig. 30 with traffic.
6. Cable Stiffness Effects
With a knowledge that the fatigue life can be greatly influenced the cable
tension, it goes without saying that cable stiffness can have similar effects,
particularly when the mechanical properties of the bridge cables are fabri-
cation specific. Using the stiffness Ew of the steel wire as the base, Ec for
the cable will be regarded as certain percentage of Ew via a parameter E as
indicated by Eq. (12) such that Ed1 will always hold as the cable can only
be partially filled with wires. The maximum cable stiffness corresponds to
E=1. Crack growth will be studied for cables #28, #38 and #50 with E=0.7,
0.8, 0.9 and 1.0.
6.1 Case I: Tightening and loosening for varying E
Under traffic, two cases will be considered: increase of initial tension to

D=1.6 and decrease of initial tension to D=0.9. In what follows, the nu-
merical results will be presented.
6.1.1 Cable #28: D=1.6 (tightening) and D=0.91 (loosening)
With an initial starter line crack of depth 1 mm, the depth extends to 50
mm as N reaches 2 million cycles. This corresponds to E=1 in Fig. 37. For
E=0.9, the crack depth has propagated way beyond the onset rapid fracture
before N=2u106 cycles. Even earlier crack instability applies to E=0.7 and
0.8. A decrease of the initial tension with D=0.9, Fig. 38 shows that the
crack depth is still sub-critical at N=2u106 cycles. For E=0.7, 0.8, 0.9 and
1.0, the crack has extended to 1.85, 1.70, 1.60 and 1.53 respectively for the
situation when the initial cable tension is decreased below the design
value. It can thus be said that change in the cable stiffness does not seem to
exert appreciable effect on crack growth.
ening) with D=1.6 and E=0.7, 0.8, 0.9 and 1.0 under traffic.
Crack growth behavior for Cable #38 is not significant. Fig. 39 shows that
as E changes from to 0.7, the largest difference in crack depth is from 2.15
to 3.18 mm with D=1.6. That is when the initial tension was increased. De-
creasing the initial tension with D=0.91, the change in the crack depth is
even less as shown by the results in Fig. 40.
Appreciable crack growth occurred in cable #50 for D=1.6 and E=0.7. The
crack depth reached 50 mm when the fatigue cycles are only 1.1u106 as il-
lustrated in Fig. 41. Decrease of initial cable tension with D=0.91 does
not exert appreciable influence on crack growth. This can be seen from the
curves in Fig. 42 that undergo little changes as N is increased.
6.2 Case II: Tightening & loosening for varying E
Suppose that the traffic is now removed so that the tension in the cables
will be reduced accordingly. However, the distribution will not be altered
in proportion such that each cable will have to be analyzed individually.
Moreover the amount of initial tension will be dictated by Eq. (2) which
differs from that of Eq. (1) for the case when traffic is present.
Summarized in Figs. 43 and 44 are the crack growth characteristics of ca-

ble #28 in the absence of traffic load. Decrease in cable stiffness tends to
enhance crack growth. For an increase of the initial cable tension with
D=1.12, the crack depth increased from 1.60 to 1.99 mm after two million
cycles as E is decreased from 1.00 to 0.7, respectively. The results corre-
spond to those in Fig. 43.The trend of the curves in Fig. 44 is similar to
that in Fig. 44 except that the crack depth now changes from 4.0 to 7.5 mm
at N=2u106 cycles with D=0.7 which is a measure of the initial decrease of
cable tension.
ening) with D=1.12 and E=0.7, 0.8, 0.9 and 1.0 without traffic.
By the same token, Figs. 45 and 46 exhibit the crack growth results for ca-
ble #38 with D=1.12 and 0.7, respectively, when the traffic load is not pre-
sent. No direct comparison can be made with the results in Figs. 39 and 40
1.18 Case II (tightening), Cable 38

-6
\=1<R=1.06X10 , D=1.12
1.14 Max. stress without traffic•
Crack depth a (mm)
1.1
1.06
1.02
E
E
0.98 E
E
0.94
0.9
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
6
1.6
Case II (loosening), Cable 38
-6
1.5 \=1<R=1.06X10 , D=0.7
Max. stress without traffic•
Crack depth a (mm)
1.4
1.3
1.2
E
E
1.1 E
E
0.9
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
6
with the load present because the amount initial tension governed by D is
not the same. They seem to have the opposite trend. That is the range of
crack growth with 2.15 to 3.18 mm for D=1.6 in Fig. 39 and 1.09 to 1.14
mm for D=0.91 in Fig. 40 are contrasted with 1.10 to 1.15 mm for D=1.12
in Fig. 45 and 1.32 to 1.50 mm for D=0.70 in Fig. 46.
The influence of traffic load on cable #50 is most significant. With traffic,
Fig. 41 for D=1.6 showed that crack grew up to a depth of 50 mm for
E=0.7 after a million cycles. A stiffer cable with E=1.0 can sustain the
same crack growth after 1.5u106 cycles. This is a substantial increase in fa-
tigue cycles. Without traffic, Fig. 47 shows that crack growth is reduced
considerably with D=1.12. A crack depth of 1.95 mm is found for E=1.0
and 2.65 mm for E=0.7. The results for a decrease of initial tension with
D=0.7 are given in Fig. 48. The corresponding crack growth is 1.2 mm for
E=1.0 and 1.44 mm for E=0.7. The effect of traffic loading in this case
shown by Fig. 42 for D=0.91 is not significant since the increase in crack
growth is only 1.85 mm for E=1.0 and 2.45 mm for E=0.7. This completes
the discussion of crack growth in relation to the influence of initial cable
tension.
Based on the design data supplied for the Runyang cable-portion of the
bridge, the fatigue crack growth characteristics for the 52 cables are ana-
lyzed and the results are presented for three typical cable behavior referred
to as cable #28, #38 and #50. Assessments are made for varying initial in-
crease and decrease of the cable tension using the design condition as a
reference under traffic and no traffic loading. Use is made of a dual scale
fatigue crack growth model where microscopic and macroscopic effects
can be distinguished. This includes the transitory behavior of micro- and
macro-cracking, commonly referred to as Region I for fatigue crack initia-
tion and II for fatigue crack propagation in a plot of log da/dN the crack
growth rate against the log of the increment of the stress intensity factor
'K that applies to linear elastic fracture mechanics (LEFM). The advent of
modern materials with higher strength and toughness has delayed macro-
scopic crack growth by converting the available energy to extend the range
of macrocracking. A description of fatigue cracking would thus entail
damage at both the micro- and macro-scale, if not at an even lower scale.
To this end, modern fatigue crack growth models must necessarily con-
sider multiscale.
Under traffic, fatigue cracking in cable #3 or #50 (because of symmetry)
is most severe as compared to those in #26 or #28 which are the next high-
est in crack growth rate. Examples are provided to show the effect of in-
crease in the initial cable tension by using a parameter D to quantify the
deviation from the design condition at D=1. For D=1.6 and N|1.3u106 cy-
cles, the crack growth rate was found to be da/dN|50u10-5 mm/c for cable
#50 and da/dN|10u10-5mm/c for cable #28 while all other conditions re-
main the same. This is difference of five times. Increase of initial cable
tension under traffic condition enhances fatigue crack growth which is not
beneficial. The opposite effect occurs when the initial cable tension is de-
creased. That is the fatigue crack growth can be impeded. The above re-
sults can also be expressed in terms of crack extension. Again for D=1.6,
the crack advanced from 1 mm to 50 mm after 1.1u106 cycles. A reduction
of D from 1.6 to 1.4 required an addition of 400,000 cycles to yield the
same range of crack extension.
The behavior of cable #50 without traffic differs significantly from that
under traffic. For an increase of initial cable tension with Dt1 and no traf-
fic, the crack grew only from 2.54 to 4.54 mm after N|2u106 cycles. For a
decrease in initial cable tension with Dd1, the crack growth range is 4.50 to
14.1 mm after N|2u106 cycles. The difference in fatigue crack growth
with and without traffic can differ by at least one order of magnitude or
more based on an upper limit of two million cycles for the cable life. It can
thus be stated that
Cable fatigue life can be enhanced and impeded, respectively, by
decreasing and increasing the initial cable tension when the traffic is
present.
Cable fatigue life can be enhanced and impeded, respectively, by
increasing and decreasing of the initial cable tension when the traffic
is absent.
In contrast to earlier design of bridge cables, higher strength steel wires
have been used to improve the cable strength. It is therefore useful to de-
velop a relation between the fatigue life of the cable and that of the wires
that are the basic elements of the cable. In this way, the test data for the
cable and wire may correlated. It is not obvious that the same fatigue life
of two million cycles should hold for both the cable and wire. It is also
necessary to realize that the present tests are not representative of the ac-
tual situation of the bridge where the length of the cable and hence the
wire comes into play. It is well known that the high strength material is
sensitive to pre-existing surface defects that would be an inherent property
of long wires. To this end, Part II of this investigation will be devoted to
associating the fatigue behavior of the cable and the wire in the hope that
the analytical work would encourage more extensive fatigue tests to which
the analytical findings may be used as a guide.
References
[1] Sih GC and Tang XS (2004) Dual scaling damage model associated with
weak singularity for macroscopic crack possessing a micro/mesoscopic
notch tip. J. of Theoretical and Applied Fracture Mechanics, 42(1): 1-24.
[2] Sih GC and Tang XS (2006) Simultaneous occurrence of double mi-
cro/macro stress singularities for multiscale crack model. J. of Theoretical
and Applied Fracture Mechanics, 46(2): 87-104.
[3] Sih GC (2007) Multiscaling aspects of fail-safe for material and structure.
In: Sih GC. Tu SD. and Wang ZD. (eds.), Integrated Approaches for Materi-
als and Structural Safety. East China University of Science and Technology
Press, Shanghai, pp 1-13.
[4] Sih GC (2007) Collapse simulation of I-35W bridge: a dynamic finite ele-
ment analysis. International Center for Sustainability, Accountability and
Eco-Affordability of Large Structures (ICSAELS), College of Civil Engi-
neering, Southeast University, Nanjing, 210096, China.
[5] Fisher JW (1980) Fatigue behavior of full-scale welded bridge attachments.
Transportation Research Board, US National Research Council.
[6] Fisher JW (1984) Fatigue and Fracture of Steel Bridges. Wiley Interscience,
New york.
[7] Mahmoud KM (2007) Fracture strength for a high strength steel bridge cable
wire with a surface crack. J. of Theoretical and Applied Fracture Mechanics,
48(2): 152-160.
[8] Sih GC, Tang XS, Mahmoud KM and Kassir MK (2007) Effect of crack
shape and size on estimating the fracture strength and crack growth fatigue
life of bridge cable steel wires. J. of Bridge Structures, in press.
[9] Sih GC, Tang XS, Li AQ, Li ZX and Tang KK (2007) Fatigue crack growth
behavior of cables and steel wires for the cable-stayed portion of Runyang
bridge: disproportionate loosening and/or tightening of cables. J. of Theo-
retical and Applied Fracture Mechanics, in press.
[10] Cohen S, Bakst B, Borenstein S, Lohn M, Foley R and Krawczynski J
(2007) Minnesota bridge problems uncovered in 1990, Associated Press,
Aug. 3.
[11] Sih GC (2007) Lecture on “bridge design and construction: criteria for
strength and efficiency”. College of Civil Engineering. Southeast University,
Nanjing, China, August.
[12] The Collection of Runyang Yangtze River Bridge. In: Wu Shengdong (ed.).
Construction Volume 4, Cable Bridge, People’s Communications Press,
Bejing, China (2005), pp33-47.
[13] Paris PC (1962) The growth of cracks due to variations in load. Ph. D. Dis-
sertation, Department of Mechanics, Lehigh University.
[14] Sih GC and Tang XS (2007) Form-invariant representation of fatigue crack
growth rate enabling linearization of multiscale data. J. of Theoretical and
Applied Fracture Mechanics, 47(1): 1-14.
[15] Sih GC (2007) Fatigue crack growth rate behavior of polyvinylchloride hid-
den by scaling: Reinterpreted by micro/macrocrack model. J. of Theoretical
and Applied Fracture Mechanics, 47(2): 87-101.
[16] Wang SY and Dang ZJ (2002) Research on heavy load high fatigue stress
amplitude cable and anchorage of cable-stayed bridge. J. of Bridge Con-
struction, 3: 14-16 (Chinese).
[17] Practical Handbook of Engineering Materials. Chinese Standard Press, Bei-
jing, Vol. 1 (1988) 375 (Chinese).
Fatigue Crack Growth Rate of Cable-Stayed Portion
of Runyang Bridge: Part II – Steel Wire Crack
Growth Due to Disproportionate Cable
Tightening/Loosening and Traffic Loading
G. C. Sih 1, 2 *and X. S. Tang 3
1
2
University of Science and Technology, Shanghai 200237, China
3
School of Bridge and Structural Engineering, Changsha University of
Science and Technology, Changsha, Hunan 410076, China
*
Email: gcs8866@yahoo.com, gcs1@lehigh.edu
Abstract
Having completed an investigation in Part I that deals with the fatigue

crack growth in the cables of the Runyang cable-stayed bridge based the
design data for the cables, Part II of this work is concerned with the fatigue
crack growth of the steel wires being the constituents of the cables. The
objective is not only to check the compatibility of the fatigue crack growth
behavior of the cable and wire but also the independent design fatigue lim-
its imposed on both the cable and wire. In particular, assumptions are
made to relate the mechanical and fatigue properties of the cable in rela-
tion to those for the steel wire. The mean stress and stress amplitude are
accounted for simultaneously such that the increase and decrease of the
initial cable tension on the fatigue cracking of steel wires can be assessed.
The increase use of higher strength wire materials calls for consideration
of damage at both the micro- and macro-scale. That is more advanced ma-
terial can absorbed more energy at the microscopic scale in contrast to en-
ergy dissipated by macrocracking. A dual scale fatigue crack model is
therefore required to address the micro/macro interaction for the mechani-
cal and fatigue properties of the material and their influence on crack
growth.
Under traffic and an increase of the initial wire tension of D=1.6 with
D=1 being the design reference for the cable, a defect of 0.1mm in the wire
can extend more than 60% across a 5mm diameter wire after 9u105 cycles
which is about 45% of the fatigue life of the wire. According to the cable
design specification, the defect should extend only 0.1mm after 2u106 cy-
cles. This shows that the fatigue life of wires is sensitive to initial defects
and to the initial wire tension. These conditions are not considered in the
present design of the cable nor the wire. What this means is that inspection
and maintenance procedures are not thought of at the design stage; they are
established during the life span of the bridge. Even more of a disadvantage
is that there are no basis for comparing the monitored data with wire
and/or cable damage analyses in order to gain an understanding of the fail-
ure mechanisms of wire and cable.
1. Introduction
The risk of bridge failure has been the aging of material and increase in
traffic. Both factors contribute to fatigue crack growth that cannot always
be monitored because the locations of damage are not always obvious and
they can and often escapes the attention of the inspector.
The collapse of the eight-lane steel-deck truss bridge in Minneapolis has
called attention to deficiencies in many older spans around the nation. Al-
though the feeling is that better material and design technologies should
make bridges safer, relatively little efforts are being made to understand
the basic mechanisms governing fatigue crack growth, particularly when
the higher strength and tougher materials are now able to sustain micro-
cracking in addition to macrocracking. This means that failure analysis
would involve at least damage at both the microscopic and macroscopic
scale, if not at an even lower scale. Such a need has not caught up to the at-
tention of those who determine the rules of bridge design.
In recent times, the US Federal Highway Administration has started to
enforce new rules for bridge design in order to make new structures more
efficient, more reliable, safer and longer lasting. Among other improve-
ments, the rules emphasize the ability of bridges to withstand peak traffic
loads, brutal weather, in addition to events like ship collisions and earth-
quakes. The American Association of State Highway and Transportation
also indicated that the new rules should encourage a more reliable way of
designing a bridge. Relatively little is known with regard to the durability
of cable-stayed structures which utilize steel cables as its primary load-
carrying members. Traffic and wind loads induce wide stress variations in
these cables that could make them vulnerable to failure by fatigue. Each
cable consists of multiple high strength steel strands arranged in a bundle
and encased in a polyethylene pipe filled with cement grout. The ends of
each cable are anchored at the bridge deck level, while the center of the
cable is draped over the top of the bridge tower or pylon, on a curved sup-
port called a "saddle." The cable is designed to withstand two million cy-
Fatigue Crack Growth Rate of Runyang Bridge: Part II 251
cles of axial load and fluctuating bending stress induced through vertical
displacement of the saddle. Tests are usually performed at a minimum fre-
quency of two Hz. The specification-required acceptance criteria limit the
number of broken wires in fatigue testing to two percent, or five out of a
total of some 250 and more wires. The combined capabilities of the axial
and the saddle fixture tests are those commercially available in the U.S. for
fatigue performance evaluation of bridge stay cables. The trend for the past
few decades has been to increase the strength from approximately 1500 to
1700 MPa [1] which is known to have a trade-off effect with the fracture
toughness. The design requirements for the limits set for the fatigue cycles
of the cable and wire, however, appear to be arbitrary and their relation-
ship is at best fortuitous. To begin with the length of the cable or wire
tested does not coincide with the full structure. High strength wire is sensi-
tive to pre-existing surface defects that are non-uniformly distributed over
the full length of the wire. It is well known for high strength silicone glass
fibers that the fiber strength is length dependent. Although no such infor-
mation prevails for the high strength steel wire, it is likely that the test data
for steel wire can have a large range of scatter and that the location failure
is location dependent along the wire length. This would indicate the exis-
tence of surface defects in that would affect the tensile strength. It is be-
lieved that the results of cable breakage for long and short wires can be
widely different, not to mention the difference in the stress states. Dissect-
ing fatigue damaged cables should provide information on fracture initia-
tion dependency on surface defects that would be scattered along the wire.
The testing of single steel wire with different length should prove the
point. Hence, long wires are susceptible to fatigue stress corrosion and hy-
drogen embrittlement that can result in a reduction of the fatigue life of
wires [2-4] and hence the cable [5] in the long run.
2. Statement of Problem
The driving force for fatigue cracks has been attributed to the accumula-
tion of energy density which is in fact the area inscribed by the hysteresis
loops although the criteria may be expressed in different forms. Compari-
son of analytical models with test results, however, can vary, often due to
insufficient information on the size and time scale of damage initiation and
termination. A specimen can be regarded as globally reversible but locally
irreversible on account of permanent deformation near the tip of a macro-
crack. The local damage can still be macroscopic owing to the accumula-
tion of energy density next to the crack tip. This corresponds to the ASTM
recommendation of fatigue tests for pre-cracked metal specimens where
the alternating stress amplitude should about 50% of the yield strength so
that the local macroscopic element would be deforming in the plastic

range. This should be distinguished from damage done to a microscopic
element where permanent microscopic deformation may have occurred for
applied stress amplitude below 50% of the yield strength.
To be more specific, Fig. 1(a) shows a schematic of the cable that con-
sists of multiple high strength steel strands arranged in a bundle and en-
cased in a polyethylene pipe filled with cement grout. The ends of each
cable are anchored at the bridge deck level, while the center of the cable is
draped over the top of the bridge tower or pylon. The single steel wire con-
sidered in this work is 5mm in diameter as shown in Fig. 1(b).
Illustrated in Fig. 2 is the fracture surface appearance of a steel bolt that
is formally similar to that of a steel wire tested under similar conditions.
When the specimen is subjected to sufficiently low stress amplitude but
repeated many
(a) Single cable (b) Single steel wire

Fig. 1 Bridge cable consisting of steel wires in polymer pipes and matrix.
Fig. 2 Fatigue crack growth in steel bolt from initiation to termination.

(a) Clam shell cracks (b) Equivalent edge cracks

Fig. 3 Clam shell cracks modeled as edge cracks.
times, a small crack next to the surface would initiate at the top and grow
slowly and then rapidly towards to the bottom of the specimen until it ter-
minates. The crack front beach marks are shown and they are sketched in
Fig. 3(a) where the clam shell cracks would first grow in a stable manner
until the crack front curvature starts to deviate from the clam shell pattern.
The shaded portion of the specimen represents fast fracture. To simplify
the analysis, an equivalent edge crack as indicated in Fig. 3(b) will be as-
sumed where the crack would grow in segments until the onset of unstable
fracture.
3. Dual Scale Fatigue Crack Model for Steel Wire
The discovery of double singularity [6,7] at the same crack tip had led to
the development of a dual scale crack model. The strong and weak stress
singularity can refer, respectively, to the macrocrack and microcrack
which can be distinguished by the crack opening distance. That is a tightly
closed crack that cannot be seen by the naked eye would be regarded as
microscopic. This is particularly relevant in fatigue where the crack sur-
face would open in tension and close in compression. Switching from mi-
cro to macro and vice versa occurs every full cycle. Such a behavior can be
accounted for by the relation
da I
) o ( 'S macro
micro )
(1)
dN
with )o and I being the empirical constants assumed in the two-parameter
model. The superscript “macro” and subscript “micro” are introduced to
indicate that the increment of the energy density factor 'S accounts for the
microscopic and macroscopic effects. This can be best illustrated by con-
sidering the case of I=1.
3.1 Equivalent fatigue crack growth relation for I=1
Without loss in generality, it can be shown that

da
pa (2)
dN
In Eq.(2), p stands for
p ) p o , p o (1 2Q micro )P* (1 V* ) 2 d * (3)
such that
2(1 Q macro ) 2 V a V m d o
'S macro
micro poa (4)
P macro r
can be replaced by having
2(1 Q macro ) 2 V aw V mw d o (5)
) )o
P macro r
Note that V aw V mw refers to the effect of the stress amplitude and mean
stress felt by the wire which is not the same as that for the cable. All quan-
tities in Eq. (5) refer to the wire. Defining / as the parameter relating VaVm
for the cable to V aw V mw for the wire, there results
Va V m /V aw V wm (6)
Since p is independent of N, a(N) can be obtained from Eq. (2) by direct

integration as
a ( N ) a o e p( N o N ) , for I=1 for the wire (7)
3.2 Mechanical and fatigue properties of wire
The mechanical properties of the steel wire can be found in Table 1 [8]
such that Pmacro and Qmacro in Eq. (4) or (5) can be replaced, respectively, by
Pw and Qw as follows:
/=0.8, Ew=199.82GPa, Pmacro=Pw=Ew/[2(1+Qw)]=76.85GPa. (8)
Table 1. Mechanical properties of bridge steel wire [8].

Elastic modulus Ultimate stress Ultimate strain
Ew (GPa) w w
V ult (MPa) H ult (%)
199.82 1722.55 5.5
Furthermore, the quantities in the second of Eqs. (3) are taken as
P micro V o =0.3, * d =1, do=10-3mm, Qmacro=0.3, Qmicro=0.4

P* =2, V* d
P macro Vf do
(9)
6
together with ao=0.01mm, and No=0.1u10 and r/do=1. The normalized
quantities in Eqs. (9) together with do and Qmicro are assumed to be valid for
both the cable and wire. The range of crack length is a=1~2.6mm with fa-
tigue cycles in the range N=0.1~2.0u106. The fatigue properties based on
the classical stress intensity range 'K are given in Fig. 4 [9]. The data can
be substituted into
da
log( )=log(ĭ o )+Ilog(ǻSmacro
micro )
(10)
dN
while ǻSmacro
micro
can be reduced to 'K with I=1. This gives
I=1, )o=2.15u10-6 for the wire (11)
Numerical values of the crack growth rate and crack length can now be
computed for studying fatigue cracking behavior of the steel wire in bridge
cables.
Fig. 4 Log-log plot of da/dN versus 'K for high strength steel data from [9].
4. Crack Growth Rate versus Fatigue Cycles for Steel Wire
Examined will be the effects of initial tension of cables inflicted to the fa-
tigue crack growth characteristics of the wire under traffic (Case I) and no
traffic (Case II) which are defined as
DV Dmax V omax
Case I: V a DVDmax V omax , V m (12)
2
V Dmax DV omax
Case II: V a V Dmax DVomax , V m (13)
2
The superscript parameter D is used to vary the tension in the cable in
Eqs. (12) for the traffic condition and in Eqs. (13) for the no-traffic condi-
tion The quantity V omax denotes the cable stress without traffic. Moreover,
it is more expedient to use the stress amplitude Va and mean stress Vm.
Hence, V omax and V Dmax will be converted to the mean stress Vm=(Vmax+Vmin)
/2 and the stress amplitude Va= VmaxVmin with Vmax and Vmin being the
maximum and minimum cyclic applied stress. This will be done for the
cables and effects exerted on the wire can be found by using Eq. (6) for
fixed values of /.
4.1 Case I: wire tightening & loosening under traffic
Numerical values of Va, Vm and VaVm for the 52 cables of the Runyang
bridge with D=1.0, 1.2 and 1.6 are found from Eqs. (12). The values are
tabulated only for cable #28, #38 and #50 as shown in Tables 2, 3 and 4,
respectively. The conversion to Vaw V mw is then made by application of Eq.
(6). The highest Va, Vm and VaVm occur in cable #50 followed by cable #28
while cable #38 attains the lowest stress amplitude and mean stress.
Similarly, the numerical values of Va, Vm and VaVm for cable #28, #38
and #50 with D=0.97 and 0.91 can be found in Tables 5 and 6, respec-
tively. In what follows, the crack growth rate in the wires will be dis-
cussed.
Table 2 Values of Va, Vm and VaVm for Cable #28, #38 and #50 for Case I with
D=1.0.
Cable No. Va (MPa) Vm (MPa) VaVm u105 (MPa2)
28 196.46 797.03 1.56583
38 90.52 358.60 0.32461
50 238.73 900.23 2.14916
D=1.2.
28 375.51 886.55 3.32909
38 171.29 398.99 0.68343
50 424.25 1002.19 4.43623
D=1.6.
28 733.61 1065.60 7.81738
38 332.84 479.76 1.59680
50 850.49 1206.11 10.25785
D=0.97.
28 169.60 783.60 1.32899
38 78.40 352.54 0.27641
50 208.15 884.93 1.84196
D=0.91.
28 115.89 756.74 0.87695
38 54.17 340.43 0.18442
50 146.97 854.35 1.25564
4.1.1 Wire in cable #28 for Dt1 and Dd1
Plotted in Fig. 5 are the variations of the crack growth rate da/dN with the
number of fatigue cycles N for the wire in cable #28 with D=1.0, 1.1, ---,
1.6 where the design condition D=1.0 serves as a reference. Note that
da/dN remained negligibly small for D up to 1.3 over the design life of
N=2u106 cycles. One order of magnitude increase of da/dN is seen at
N=1.80u106 cycles for D=1.5 and N=1.48u106 cycles for D=1.6. This is a
difference of 320,000 cycles.
Shown in Fig. 6 are curves for D=0.91, 0.94 and 0.97. Compared to the
reference curve D=1.0 for da/dN|75u10-7 mm/c after two million cycles,
they all decrease in the crack growth rate with lowest being 16u10-7 mm/c.
This occurred at D=0.91 and N=2u106 cycles.
Fig. 5 Wire crack growth rate da/dN versus number of cycles N for wire in Cable
#28 and Case I with Dt1.
#28 and Case I with Dd1.
The situation for the wire in cable #38 is not as severe. Fig. 7 shows that
the crack growth rates for D=1.0, 1.1 and 1.2 are nearly constant as N in-
creased up to two million cycles at which time da/dN remained within the
range of 0.5 to 1.0u10-8 mm/c. This is several orders of magnitude smaller
than the crack growth rate in Fig. 5 for the wire in cable #28.
A significant reduction of da/dN can be seen from Fig. 8 for the wire in
cable #38 when the initial cable tension is decreased with D=1.0, 0.97,
0.94 and 0.91. The corresponding crack growth rate ranged from 14 to
33u10-10 mm/c
The outer most distanced cables of the bridge tend to be more stresses.
This applies also to the wires in these cables such as #50. Fig. 9 indicates
that da/dN reached about 50u10-5 mm/c for D=1.5 at N|1.60u106 cycles
and D=1.6 at N|1.33u106 cycles. At these high crack growth rates, there
prevails the likelihood of fatigue crack growth failure. When the initial
tension in cable #50 is decreased, the corresponding wires are les stressed
since D in Eq. (13) would be less than unity. According to Fig. 10, da/dN
at N=2u106 cycles reduced to 4u10-7 mm/c at D=0.91 and to 22u10-7 mm/c
at D=1.0.
4.2 Case II: wire tightening & loosening without traffic
Consider Eq. (13) where the traffic load is removed. The stress amplitude
should be reduced accordingly. The maximum and minimum applied
stresses determine the stress amplitude. The results of Va, Vm and VaVm for
cable #28, #38 and #50 are given, respectively, in Tables 7, 8 and 9 for
D=1.00, 1.04 and 1.12. Case II refers to max. stress without traffic while
Dt1 corresponding to increase of initial cable tension.
Tables 10 and 11 summarize the Va, Vm and VaVm for D=0.9 and 0.7
where the initial cable tension is decreased below the reference design
level of D=1.0. Contrast to intuition, the values of Va, Vm and VaVm in Ta-
bles 10 and 11 are higher than those in Tables 5 and 6. Hence, decreasing
the initial cable tension without traffic does not decrease the stress ampli-
tude nor the mean stress.
Table 7 Values of Va, Vm and VaVm for Cable #28, #38 and #50 for Case II with
D=1.0.
28 196.46 797.03 1.56583
38 90.52 358.60 0.32461
50 238.73 900.23 2.14916
D=1.04.
28 168.51 811.00 1.36659
38 77.99 364.87 0.28455
50 207.50 915.85 1.90038
D=1.12.
28 112.60 838.95 0.94469
38 52.92 377.40 0.19972
50 145.03 947.08 1.37356
D=0.9.
28 266.34 762.09 2.02973
38 121.85 342.93 0.41788
50 316.82 861.19 2.72842
D=0.7.
28 406.10 692.21 2.81103
38 184.52 311.60 0.57497
50 472.99 783.10 3.70401
The crack growth rates in Fig. 11 for the wire in cable #28 are found to de-
crease with increasing D from 1.00 to 1.12. After two million cycles da/dN
#28 and Case II with Dt1.
#28 and Case II with Dd1.
decreased from 75u10-7 mm/c at D=1.00 to 17.5u10-7 mm/c at D=1.12. If

the initial cable tension is decreased by letting D=0.7, 0.8, 0.9 and 1.0, then
the da/dN in the wire becomes 66u10-7 mm/c at D=0.7 and 5u10-7 mm/c at
D=1.0. These results can be found in Fig. 12.
The wire crack growth rate in Fig. 13 refers to cable #38 with Dt1 without
the influence of traffic. The curves rise slowly with N and they are
nearly parallel. The range of da/dN at N=2u106 cycles is about 16u10-10

mm/c at D=1.12 to 33u10-10 mm/c at D=1.00. Fig. 14 plots the wire crack
growth rate as a function of N for D=0.7, 0.8, 0.9 and 1.0 when the initial
tension in cable #38 is decreased. An increase in the wire da/dN is seen at
N=2u106 cycles. The range of wire da/dN for D=1.0 to 0.7 is from
27.5u10-10 to 80u10-10 mm/c.
The wire da/dN in cable #50 as shown in Fig. 15 increases by more than
one order of magnitude for D=1.00, 1.04, 1.08 and 1.12 without traffic.
The curve rise slowly for low cycles and then increases more quickly as N
approaches the limit of two million cycles. The beneficial effect is that the
wire da/dN decreases as the initial cable tension is increased to D=1.12.
The final drop in wire da/dN is from 22u10-8 at D=1.00 to 44u10-9 mm/c at
D=1.12. The effect of decreasing cable tension in the absence of traffic is
shown in Fig. 16. The curves are found to rise more quickly with decreas-
ing D as N is increased. A wire da/dN of 15u10-9 mm/c at D=1.00 increases
to 28u10-8 mm/c at D=0.7.
5. Crack Length as a Function of Fatigue Cycle for Steel Wire
The wire da/dN can be used to find the wire crack length as a function of
the fatigue cycles. The will also be exhibited for the wire crack growth
characteristics corresponding to cable #28, #38 and #50 with and without
traffic.
5.1 Case I: wire tightening & loosening under traffic
Crack length serves as an indicator for the failure of structural components

by cracking or fracture. Its variation with time or fatigue cycles consists of
three stages: stable and slow growth, transition region and unstable and
rapid propagation. Such information can be monitored and compared with
the initial design condition to determine the current state of the structure
system such that a decision could be made with reference to continuing
service or interrupting service for repair. The intention of this study is to
determine the critical states of fatigue cracks in the wire by assuming that
under service the cable tension can deviate from the design conditions.
The wire crack depth corresponding to cable #28 is plotted as a function of

the fatigue cycles N in Fig. 17 when traffic is present. No growth occurs
Fig. 17 Wire crack length a versus number of cycles N for Cable #28 and Case I
with Dt1.
with Dd1.
for the reference state D=1. As D is increased in constant increment of 0.1

from 1.0 to 1.6, the wire crack depth starts to increase with increasing N.
The curves for D=1.4 to 1.6 rise very quickly after one million cycles. All
three curves reach a crack depth of 3 mm before the design life of
N=2u106 cycles although these results correspond to abnormal conditions
where the initial cable tension were higher than the design value. For lower
initial cable tension with Dd1, Fig. 18 shows that the wire crack depth is
much smaller, they are .027 mm for D=0.91 and 0.071 mm for D=1.00.
Wire crack depth for cable #38 is negligibly small. Fig. 19 shows that even
after two million cycles for all Dt1, the crack depth a changes very little
from its original size of 10x10-2 mm for D=1.0 and 1.1 although increase
in the crack depth is seen as N increases. At N=106 cycles, a | 75x10-2 mm
for D=1.6. The variations in the crack depth in terms of D is much less for
Dd1. The results are shown in Fig. 20. Note that the fatigue cracking be-
havior of the wires in cable #38 differed widely from those in cable #28.
This illustrates the possibility for using different wire material in different
cables such that the design in terms of wire fatigue life owing to crack
growth could even out and would not undergo such wide variations.
with Dt1.
with Dd1.
Crack depth for the wires in cable #50 is even more appreciable than those
in cable #28 indicated by Fig. 17. Possible unstable wire crack instability
may have occurred for D as small as 1.2 at or before the design
with Dt1.
with Dd1.
fatigue life. For D=1.6, the crack depth may nave reached a critical state
after only 860,000 cycles. By decreasing the initial tension of the wires in
cable #50 with Dd1, a reduction in crack depth is seen in Fig. 22 where the
largest crack depth at D=1.0 is only 0.15 mm after two million cycles.
5.2 Case II: wire tightening & loosening without traffic
Removing the traffic leads to small crack growth. Summarized in Figs. 23

to 28 inclusive are the crack depth results for the wires in cable #28, #38
and #50.
Figs. 23 and 24 show that the wire crack depth increase slowly with the fa-
tigue cycles N for Dt1 and Dd1 as the initial tension of the wires are in-
creased and decreased, respectively. Starting with the same initial crack
length of ao= 10-2 mm, all the curves for N up to about 4x105 cycles are
almost unchanged; they start to deviate with increasing N. Referring to
Fig.23, the wire crack depth changed from 32x10-3 to 72x10-3 mm after
N=2u106 cycles as D changed from 1.12 to 1.00, respectively. Increasing D
Fig. 23 Wire crack length a versus number of cycles N for Cable #28 and Case II
with Dt1.
decreased the crack depth. Fig.24 exhibits the results for Dd1 in the same
cable with the traffic absent. In this case, decreasing D resulted in greater
crack depth. This trend can be found from the four curves in Fig.24 with
reference curve with D=1 giving the lowest crack depth. At N=2u106 cy-
cles, the crack depth is 60x10-3 mm for D=1 and 36x10-2 mm for D=0.7.
Significant change in the crack depth is observed. The crack growth be-
havior is seen to depend sensitively on loosening and tightening of the
wire.
with Dd1.
Even smaller crack depth occurs for the wire in cable #38. This can be
seen from the results in Figs. 25 and 26. No further discussion is required
for this case.
with Dt1.
with Dd1.
Finally, Figs. 27 and 28 show the crack depth for the wires in cable #50 for
Dt1 and Dd1, respectively, in the absence of traffic. The wire crack in-
creased about one order of magnitude as the fatigue cycles increased to
two million cycles. The largest crack depth is 1.15 mm occurred at D=0.7
in Fig. 28 when the initial wire tension is reduced below the design value.
This completes the discussion of the wire crack growth characteristics.
with Dt1.
with Dd1.
An attempt has been made to relate the fatigue crack growth characteristics
of the steel wires that make up the bridge cables. This is analogous to that
of relating the composite bundle strength from the properties of the fibers,
a problem that is not only unresolved up to now but still lacks understand-
ing. A possible attribute to the difficulties is incorporating the surface de-
fects being related to the fiber or wire length because past tests have re-
vealed the length dependency of the wire strength on its length. Since it is
unpractical to test the full length of the bridge cables or wires, little is
known on the validity of the present fatigue test data since they are based
on specimens that are considerably shorter than the full structure. Despite
the simplifying assumptions made in the present analytical model, the es-
sential dual scale material and geometric parameters are considered such
that their influence on fatigue crack growth is exhibited on a relative base.
A wide variance of fatigue crack growth has been observed for the wires in
the different cables. The results can vary by one order of magnitude or
more according to the design cable data for the Runyang cable-stayed
bridge. They show that the wires in the highly stresses cables can fracture
by fatigue at only one-half of design fatigue life of two million cycles
while others are still within the safe limits. The highly non-uniformly dis-
tributed cable tensions appear to a typical characteristic in the design of the
present-day cable-stayed bridges. Not enough is known at present whether
cable tension non-uniformity would over ride the beneficial aspects of ca-
ble-stayed bridge design when weighed in financial terms. Little efforts are
made to this basic issue, rather the trend seems to rely on health monitor-
ing to which some remarks are in order.
Recent works [10-12] have emphasized the necessity to correctly iden-
tify the monitored signals with the physical damage that is expected to be
observed. Firstly, the scenarios selected for the design loads must be de-
fined before deciding on the hardware (e.g. the type and accuracy of sen-
sors) and software (model relating the signal to the type and size of dam-
age) used for health monitoring. Keep in mind that software is needed to
convert many of parameters such as stress, temperature, etc. from the
change of displacement or electrical resistance. In addition, the non-
equilibrium behavior of an open system introduces fundamental difficul-
ties to the in-field measurements. More attention should be placed on the
relative changes rather than the values of the specific parameters. Too
much emphasis cannot be said with reference to the danger of placing false
confidence on unqualified monitored data. Furthermore, it should also be
said that analytical models are seldom devised to predict the physical out-
come but they are made available to guide tests and to minimize the num-
ber of tests by relating the results of two independent tests.
References
[1] Elices M. Llorca J. and Astiz MA. Fatigue of steels for concrete reinforce-
ment and cables. In: Carpinteri, A. (Ed.), Handbook of Fatigue Crack Propa-
gation in Metallic Structures. Elsevier, Amsterdam (1994) 191-220.
[2] Toribio J. and Lancha AM. Stress corrosion behavior of high-strength steel:
design on the basis of the crack growth kinetics curve. Materials & Design,
16(5) (1995) 283-288.
[3] Mahmoud KM. Fracture strength for a high strength steel bridge cable wire
with a surface crack. J. of Theoretical and Applied Fracture Mechanics,
48(2) (2007)152-160.
[4] Sih GC. Tang XS. Mahmoud KM and Kassir MK, Effect of crack shape and
size on estimating the fracture strength and crack growth fatigue life of
bridge cable steel wires. J. of Bridge Structures (2007) in press.
[5] Sih GC. and Tang XS. Fatigue crack growth rate of cable-stayed portion of
Runyang bridge: Part I–cable crack growth due to disproportionate cable
tightening/ loosening and traffic loading. In: G. C. Sih (Ed.), Multiscale Fa-
tigue Crack Initiation and Propagation of Engineering Materials: Structural
Integrity and Microstructural Worthiness, Springer (2008) 207-244.
[6] Sih GC. and Tang XS. Dual scaling damage model associated with weak Sin
gularity for macroscopic crack possessing a micro/mesoscopic notch tip. J.
of Theoretical and Applied Fracture Mechanics, 42(1) (2004) 1-24.
[7] Tang XS and Sih GC. Evaluation of microstructural parameters for micro-/
macro line crack damage model. J. of Theoretical and Applied Fracture Me-
chanics, 46(3) (2006)175-201.
[8] Practical Handbook of Engineering Materials. Chinese Standard Press,
Bejing, Vol. 1 (1988) 375 (Chinese).
[9] Wang SY and Dang ZJ, Research on heavy load high fatigue stress ampli-
tude cable and anchorage of cable-stayed bridge. J. of Bridge Construction, 3
(2002) 14-16 (Chinese).
[10] Sih GC, Signal recognition of fatigue crack growth in bridge structures con-
nected to specimen behavior. J. of Bridge Struct. 2(3): (2006) 133-145.
[11] Sih GC, Dual scale monitoring of fatigue cracks based on time rate change
of local compliance. In: Li AQ. Sih GC, Nied HF and Li ZX (eds.), Health
Monitoring of Structure, Material and Environment, Southeast University
Press, Nanjing, China, (2007) 17-36.
[12] Sih GC, Tang XS. Li ZX. Li AQ. and Tang KK. Fatigue crack growth be-
havior of cables and steel wires for the cable-stayed portion of Runyang
bridge: disproportionate loosening and/or tightening of cables. J. of Theo-
retical and Applied Fracture Mechanics (2007) in press.
Fatigue of Small-Scale Metal Materials: From Micro-
to Nano-Scale
G. P. Zhang* and Z. G. Wang
110016, P. R. China
*Email: gpzhang@imr.ac.cn
Abstract
This work considers currently-developed fatigue testing techniques and

evaluation methods for the small-scale metal materials. Then, a number of
studies on fatigue damage behavior and fatigue properties are presented
and highlighted as a function of the length scales. The relationship be-
tween length-scale-dependent microstructure evolution and fatigue behav-
ior is addressed to understand fatigue damage mechanisms at small scales.
Finally, the perspectives of studies on fatigue of the small-scale metal ma-
terials are proposed.
1. Introduction
Fatigue, as one of the most important failure modes in engineering struc-

tural materials, has been investigated for over one hundred years. In recent
years, the transition from the traditional engineering to micro/nano tech-
nologies drives research interests in fatigue toward small-scale materials
[1]. These small-scale materials are widely used in Micro-Electro- Me-
chanical Systems (MEMS)/microsystems, etc. high-tech fields [2]. For ex-
ample, in microelectronics and communication thin metal films on a sub-
strate are basic functional components. In automobile industry, MEMS
accelerometers and rate gyroscopes are used in safety airbags and antilock
braking systems to replace larger and more costly components. Small pres-
sure sensors are used in engine fuel injection systems to monitor fuel and
air pressure. In the biomedical field, some pressure and chemical sensors
in pumps, respirators and dialysis machines are the most applicable
MEMS-based devices.
In general, the small-scale materials can be classified into three catego-
ries according to their dimensions: (i) thin films whose dimension in one
direction is close to micrometers or less; (ii) thin wires/fibers whose di-
mensions in two directions are very small; and (iii) small particles. On the
other hand, these materials can also be roughly classified into two types
276 G. P. Zhang and Z. G. Wang
based on confinement: (i) free-standing thin films, small dimensional foils

(Fig. 1(a)) and thin wires (Fig. 1(b)), and (ii) thin films confined by a sub-
strate (Fig. 1(c)). No matter what kind of these materials are, at least one
dimension of the material should be in the range from micrometers to na-
nometers.
(a) Thin foil/film
(b) Fine wire
(c) Substrate
Fig.1 Schematic of the small-scale materials (top) free-standing thin foils/films,
(middle) fine wires and (bottom) thin films confined by a substrate.
As metallization interconnects, thin metal films confined by a substrate

are widely used in microelectronic devices and integrated circuits in semi-
conductor industry. The thin films are unavoidably subjected to thermal fa-
tigue loading during temperature variation, which is caused by the differ-
ence in thermal expansion coefficients between film and underlying
substrate. Temperature variation typically occurs during film preparation,
micro-fabrication and also during device operation [3]. That leads to stress
and strain cycles which can cause the onset of fatigue damage. For free-
standing films or foils, they are usually microfabricated into small func-
tional and/or structural components in MEMS/Microsystems. These small-
dimensional components would be subjected to mechanical fatigue loading
at different frequency in all kinds of environment. For example, micro ra-
dio frequency (RF) switches, etc., are constantly subjected to a cyclic load
with high frequency during the period of their service. Fatigue usually
causes a gradual change in the properties of the small-scale material [4, 5].
The formation of microcracks or damage in microcomponents due to fa-
tigue would gradually change the resonant frequency and electrical resis-
tance of microdevices, degrade the sensor output and, most seriously, lead
to the failure of MEMS devices. Therefore, the failure induced by me-
chanical fatigue and thermal fatigue are the potential reliability threats to
these systems. Understanding the underlying mechanisms of fatigue be-
havior and measuring fatigue strengths of these small-scale materials are
Fatigue of Small-Scale Metal Materials 277
great challenges and may help us to improve the overall reliability of entire
devices.
Today, the high integration density in micro/nano-systems is leading to
material dimensions (geometrical and microstructural dimensions) shrink-
ing toward the submicrometer and nanometer scale. To evaluate fatigue
properties of such materials it is necessary to consider the following im-
portant questions corresponding to fundamental and experimental aspects.
1. Which fatigue testing method is suitable to evaluate the fatigue prop-

erties of small-scale materials like thin metal films with thickness of
several micrometers or less?
2. Do the fatigue behavior and mechanism change with varying length
scale? For example, what happens when the film thickness or grain
size is smaller than the dimension of a typical fatigue-induced dislo-
cation structure as observed in bulk materials?
3. How does a fatigue crack initiate and propagate in a small-scale mate-
rial? What is the mechanism and size effect?
In this work, newly-developed experimental methods will be described

for performing fatigue tests of the small-scale materials, and then fatigue
failure process including crack initiation and growth will be introduced in
sections 3 and 4, respectively. Furthermore, underlying physical mecha-
nisms of fatigue damage of the small-scale materials will be discussed with
consideration of size effects in section 5.
2. Fatigue Testing and Evaluation Methods of Small-Scale

Materials
As the dimensions of the material decrease into the micrometer, submi-

crometer or even nanometer regime, new experimental methods have to be
developed to measure fatigue properties accurately. It is suggested that the
following factors are needed to be considered.
1. Ultra-low force load cell, closed-loop high resolution displacement

actuator and PC-controlled data acquisition system.
2. Reliable clamping method that can overcome the difficulty in mount-
ing thin films without damaging.
3. Application of a simple and homogenous stress state which makes ex-
traction of fatigue data and comparison with bulk materials easier.
4. It is also anticipated to in-situ observe fatigue damage process, such
as fatigue crack initiation and growth during fatigue loading.
5. Simple sample fabrication process that allows the fabrication of

small-scale materials suitable for fatigue testing. Also the samples are
easily prepared into transmission electron microscopy (TEM) samples.
(a) Uniaxial cyclic loading
(b) Dynamic bending

Fig. 2 Schematic illustration of fatigue machine for small-scale materials through
different testing methods.
Table 1 Summary of fatigue testing methods of small-scale materials.

Method Materials Sample shape Reference
Uniaxial tension- Cu Free-standing thin wires, [6-10]
tension dog-bone shaped foils
Dynamic bending Ni-P amorphous alloys Micro-sized cantilever [11-13]
Stainless steel beam
Silver
Resonant vibration Cu/Nb foils [14]
Uniaxial tension- Cu/Polyimide Dog-bone shaped films [15]
compression confined by a substrate
Thermal-cyclic Cu Patterned metal lines [16]
Concerning the points mentioned above, a couple of testing methods

have been developed in recent years. Table 1 summaries the fatigue testing
methods and the corresponding materials conducted recently. In general,

these methods are roughly classified into uniaxial cyclic loading and dy-
namic bending based on the loading mode. Like the conventional method
used for bulk materials, the most common test method is uniaxial cyclic
loading, which can directly measure the cyclic stress-strain behavior of the
samples. Following this idea, a simple apparatus for fatigue of thin metal
wires has been constructed in [6]. It applies a constant strain amplitude on
the sample by means of an electromagnetic vibrator (a hard bass loud-
speaker) being excited by a frequency generator. The force is measured by
a load cell, and the strain is measured by an inductive strain gauge. A mi-
crometer is used to apply the mean stress on the sample. After amplifica-
tion, both signals are registered by a memory oscilloscope. Similarly, a
new tension-tension fatigue apparatus was developed in [7], which could
apply cyclic load by an electromagnetic actuator controlled by a frequency
generator. High resolution piezoelectric sensors were used to measure the
load. Fatigue tests were conducted successfully with free-standing Cu
films in the thickness range of 20 Pm-100 Pm. After that, a couple of new
methods are developed for uniaxial cyclic loading of free-standing metal
foils [9, 10] through replacing the magnetic vibrator with piezoelectric ac-
tuator. The basic principle of a fatigue machine for small-scale materials is
depicted in Fig. 2(a). A set-up in [10] is a piezoelectric-driven uniaxial
stress-strain measurement system to evaluate fatigue properties of thin Ni
films. The force calibration was linear over 0-1.50 N with a resolution of
0.2 mN. The displacement range of the closed-loop piezoelectric actuator
in the test apparatus is 0-90 Pm, with a displacement measurement resolu-
tion of 1.8 nm. The frequency response of the piezoelectric actuator is up
to 103 Hz, which is sufficient for fatigue testing. To facilitate grip sample,
three-axis micropositioning system and a CCD with remote microscope
were used to locate the specially-designed glass grip appropriately. Fur-
thermore, a MEMS-based apparatus was set up to conduct fatigue testing
of very small samples.
Even though the fatigue properties of the metal foils have been meas-
ured, ideal tension-compression cyclic loading for very thin free-standing
films (less than one micrometer) may not work using this method. Alterna-
tively, the dynamic bending method was developed recently. A fatigue
machine for small-scale materials was developed in [11] as schematically
illustrated in Fig. 2(b). The machine has a displacement resolution of 5 nm
and a load resolution of 10 PN. The various waveforms for cyclic loading
and frequencies up to 100 Hz can be applied by using this machine. The
cyclic loading of a free-standing Ni-P amorphous alloy and SUS304
stainless steel foils was conducted in [11,12] by using dynamic bending of
micro-cantilever beams, respectively. Before testing, the diamond tip of

the indenter was precisely located at the loading point of the microbeam by
the repeated calibrations using the CCD camera, and then the microcanti-
lever beam was dynamically bent under constant load amplitude control.
According to the elastic beam theory, when the beam was deflected dy-
namically, the top surface at the fixed end of the microbeam would be sub-
jected to a pull–pull stress and the rear surface of the microbeam to a
push–push stress. During dynamic bending tests, the microbeam stiffness
was recorded automatically. Fatigue life of the free-standing small samples
was determined based on the variation of the microbeam stiffness during
dynamic bending because crack initiation would cause a decrease in the
stiffness.
Fig. 3 Schematic of the bilayer Ag-SiO2 beam deflection experiment. The indenter
tip is used to deflect the beam [13].
Fig.4 Schematic of the resonant frequency device to perform fatigue tests with
fully reversed stress amplitudes on thin films [14].
A method to conduct cyclic tension-compression loading of thin Ag

films deposited onto a SiO2 substrate was proposed in [13] through a
nanoindenter with a function of continuous stiffness measurement, as
shown in Fig. 3. The dynamically elastic bending of the film/SiO2 sub-
strate composite causes cyclic tension-compression loading to the thin Ag

film at the top of the substrate. Fatigue failure of the Ag film was moni-
tored by the variation of the stiffness of the microbeam. Recently, a device
based on the resonant frequency method was constructed in [14], as shown
schematically in Fig. 4. By this kind of dynamic bending method, fatigue
properties of self-supported Cu/Nb multilayers with a total thickness of 40
Pm and individual layer thickness of 40 nm were measured. Their results
show that the resonant frequency method is efficient for fatigue testing of
self-supported thin film materials. This method provides a new fatigue
testing technique to investigate fatigue damage behaviors of metallic mul-
tilayers and/or thin films under a very high frequency and a relatively easy
way to monitor fatigue failure lifetime compared with other methods.
Even though dynamic bending methods through repeated indentation or
resonant vibration can obtain fatigue properties of the small-scale materi-
als, the disadvantages of this method are the requirement for micro-
fabrication of the samples and the nonuniform stress distribution at the
cross-section of the microbeam does not facilitate extraction of fatigue
data.
To overcome the disadvantages pointed out above, a new fatigue testing
method was proposed in [15] through applying a uniaxial load to metal
films deposited on a compliant substrate with a high elasticity and me-
chanical stability, as shown in Fig. 5(a). The film/substrate-composite is
strained by tensile loading and unloading while the film is deformed elasti-
cally and plastically in tension and in compression on loading and unload-
ing, respectively. In addition, to evaluate the applied plastic strain range,
the cyclic stress-strain curves of the samples can also be measured in situ
by X-ray diffraction goniometer, as shown in Fig. 5(b). The advantages of
this method are:
1. A cyclic tension-compression load can be applied to the very thin

metal films with thickness even down to tens of nanometers.
2. The bulk-like fatigue sample of metal film/substrate composite can be
fabricated easily and fixed to the fatigue machine by relatively large
clamps.
3. Fatigue tests can be controlled either by constant load range or total
strain range.
4. Samples for Transmission Electron Microscopy (TEM) can be easily
obtained from the fatigued specimens by conventional preparation
methods.
Therefore, this method seems to be the most suitable one for fatigue in-
vestigations of thin films at room temperature. Following such a method, a
number of studies of fatigue behaviors of thin metal films and metallic

multilayers are conducted.
(a) Cu film/polyimide specimen (b)Film stress

Fig. 5 Cu film/polyimide substrate composite for fatigue testing and film stress Vx
as a function of substrate strain during the second and tenth cycle of the test on a
0.7 Pm thick copper film. The arrows indicate the development of the film stresses
during loading and unloading [15].
Fig. 6 Schematic illustration of the alternating current testing setup [16].
Thermal fatigue of the metal film/substrate system is a potential threat

to the reliability of chip-level packaging. Since the film is bonded to a sub-
strate that is much thicker than the film, the mismatch in coefficient of
thermal expansion between the metal film and the substrate leads to
changes in film strain upon changes in temperature. A direct method to
generate thermal cyclic strain in the films is through repeated heating-
cooling by a furnace. However, this slow process would not be suitable to
generate a high enough number of thermal fatigue cycles for the evaluation
of thermal fatigue damage of the films.
A method for thermal fatigue testing of thin metal films and lines was
proposed in [16]. Fig. 6 schematically illustrates the setup of the in-situ al-
ternating current (AC) testing system. The AC was generated by a signal
generator and applied to the Cu film by two metallic probes. The actual
temperature in the Cu line was monitored by measuring the voltage and
current in the lines by a four point measurement during the cycles. Know-
ing the resistance-temperature relationship of a line, the temperature can be
determined by calculating the resistance changes from the measured cur-
rent and voltage signals. Joule heating due to the AC causes temperature
cycles, which result in cyclic strain and stress in the film bonded to a Si
substrate. Different strain amplitudes could be adjusted by varying the am-
plitude of the electrical signal. This test system was installed in an SEM
chamber and allowed in-situ observations of the fatigue damage evolution
under different testing conditions. For example, cycling the temperature
between 100 and 250 qC resulted in a total cyclic strain range of about
0.21% and caused damage in a 300 nm thin Cu lines after 3x107 cycles at
200 Hz. Such a testing system can be of significance not only for the in-
situ thermal-fatigue testing of thin metal films but also for the evaluation
of the reliability of the actual metal interconnects by simulating the real
working conditions of interconnect structures.
Even though several kinds of fatigue testing methods for small-scale
materials have been developed, none of them has enough merits as men-
tioned previously. Thus, up till now there is a long way to find novel ex-
perimental methods and characterization ways special for the small-scale
materials, especially for free-standing ultra-thin films and nano-scale
wires. Nonetheless, a number of studies on fatigue properties and damage
behavior have been carried out by these methods. Fatigue behavior differ-
ent from that of bulk materials was found and will be reported in the fol-
lowing sections.
3. Fatigue Crack Initiation
It is well known that fatigue damage in a bulk metal usually originates

from the surface of a material where there exists a locally-high stress con-
centration. These stress concentration sites on the material surface can be
either surface flaws caused by mechanical machining or the places where
fatigue extrusions/intrusions at the surface of the bulk materials are formed
due to the development of cyclic strain localization. Beneath fatigue extru-
sions/intrusion, persistent slip bands (PSB) with typical dislocation struc-
tures can develop through the formation of dislocation walls. It is widely
evidenced that the height of the fatigue extrusions/intrusions is about sev-
eral micrometers and the spacing of the typical dislocation walls is about
1.3 Pm. When the dimensions of the small-scale material is close to and
even smaller than the characteristic dimensions of the fatigue extrusion,
such as height, and the typical fatigue dislocation structures, such as dislo-
cation walls in PSBs, it is questionable whether fatigue extrusions

/intrusions and typical dislocation structures similar to that observed in
bulk materials can form in the small-scale materials and what the relation-
ship between fatigue damage and the length scale of the material is?
Table 2 Fatigue tests of free-standing metal foils.

Material Preparing h d Testing condi- Contents Reference
method [Pm] [Pm] tions
Cu Rolling & HT 20-100 100 T-T S-N curve [7]
R=0, f=70 Hz
Cu ED & 25 LCF FD [8]
Wrought 35
Cu EBD 1.1 0.98 [9]
Cu Rolling & HT 9-250 15-45 T-C; R=-1; f=20 FCGR [17]
kHz FD
Cu ED 35, 105 2-10 T-C; R=-1; f=20 FCGR [17]
kHz FD
Al Rolling and HT 50-250 140-218 T-C; R=-1; f=20 FCGR [17]
kHz FD
Cu, Mo, Rolling and HT 20-250 T-C; R=-1; f=20 FCGR [18]
Al kHz
Cu ED & Wrought 10-25 1-10 Flex fatigue S-N curve [19]
SUS304 Rolling and HT 25 2.89 DB S-N; FD [12,20]
R=0.2 FCGR
TiAl Extruded 20 R=0.1, f=78-90 S-N; FD [21]
Hz
Ni LIGA 70, 270 2-5, 5- T-T S-N; FD [22]
10 R=0.1, f=10 Hz
Ni LIGA 200-400 T-T S-N [23]
R=0.1, f=200 Hz
Ni LIGA 10 1.1 S-N; [24]
FCGR
Ni LIGA 70, 250 5, 0.015 T-C FD [25]
R=-1, f=20 Hz
Ni-P ED 12 - R=0.1, 0.3, 0.5; FCGR; FD [11,26]
f=10 Hz
Ni-P ED 12 - R=0.1, 0.3, 0.5; S-N curve [27]
f=10 Hz
Cu/Nb PVD 40 40 Resonance S-N [14]
T-C Tension-Compression; T-T Tension-tension; DB Dynamic bending; HT Heat-

treatment; ED Electro-deposition; EBE Electron-beam evaporvated; PVD Physical
Vapor Deposition; LIGA Lithographic, Galvanoformung, and Abformung; FD Fa-
tigue damage; LCF Low-cycle fatigue; FCGR Fatigue crack growth rate
Table 3 Fatigue tests of thin metal films confined by a substrate.
Materials Preparing h [Pm] d (Pm) Testing Contents* Reference

method conditions
Ag Magnetron 1.5-0.3 1.0 Dynamic FD [13,29]
sputtering bending
Cu Magnetron 3.0-0.4 1.0 R=0, f=1 Hz FD; Lifetime [30-32]
sputtering curve
Cu Magnetron 0.2 1.0 R=0, f=1 Hz FD [32,33]
sputtering
Cu Magnetron 3.0-0.4 0.78-0.28 R=0, f=1 Hz FD; Lifetime [34,35]

sputtering curve
Cu Magnetron 0.128-0.050.02-0.03 R=0, f=10 Hz FD; Lifetime [36]
sputtering curve
Cu/Ta Magnetron 0.5 0.1 R=0, f=10 Hz FD [38,39]
sputtering
Cu Magnetron 0.2 1.0 Thermal FD [16]
sputtering fatigue
Table 4 Fatigue tests of thin metal wires.
MaterialsPreparing method h d Testing condi- Contents Reference

[Pm] [Pm] tions
Cu Drawing and 30, 50, 15-45 R=0; f=20 Hz S-N [6]
Annealing 90 curve
Au Drawing and 25, 100 140-218R=0; f=20 Hz S-N [6]
Annealing curve
Cu Drawing and An- 10-125 10-125 T-C, S-N [28]
nealing curve
A number of studies of fatigue properties of free-standing thin metal

foils [7-10, 11,12,14, 17-27], fine wires [6, 28], thin metal film
[13,15,16,29-39] confined by a substrate have been carried out in recent
years. These studies are summarized in Table 2 for the free-standing metal
foils, Table 3 for thin metal films confined by a substrate and Table 4 for
fine metal wires.
Fig. 7 summarizes some typical small-scale materials that have been
studied based on their geometrical (foil/film thickness and wire diameter)
and microstructural (grain size) dimensions. It is clear that some studies
mainly concentrate on the foils/wires with geometric dimension being lar-
ger than grain size, while the others are the cases of the geometric dimen-
sion of the films/wires being smaller than grain size, especially the films
confined by a substrate.
Fig. 7 A summary of the materials tested under cyclic loading based on the
geometrical (foil/film thickness and wire diameter) and microstructural (grain
size) dimensions.
For the free-standing thin foils/wires, a study in [7] showed that the 100
µm thick Cu foils contained a large number of well-defined and sharp slip
bands with in some cases well-developed extrusions. However, for the 20
µm thick Cu foils they were almost free of extrusions and contained only a
few grains with a low density of faint slip bands. These results indicated
that the microstructural development toward failure is somehow delayed
for the thinner foils. Especially, extrusions and microcracks in the thinner
foils have been observed mainly at the end of fatigue life. Nevertheless,
the microcrack observed is probably due to an extrusion. For the fine metal
wires, it was found [6] that the absence of extrusion-like features on thin
wire surfaces up to 105 cycles may be interpreted as the absence of PSBs
of the type observed in marcoscopic specimens. It was reported [28] that
slip bands corresponding to extrusions and intrusions formed on the sur-
face of the bamboo-structured 20 µm and 50 µm wires shown in Fig. 8(a)
and (b), respectively. Fatigue fracture in the wires occurred as a result of
crack initiation at the slip bands in a deformed grain which propagated
mostly in a transcrystalline mode, as shown in Fig. 8(c). The thinner wires
failed in a transcrystalline manner with a pronounced crack path parallel to
the slip bands, while the thicker wires showed cases of intercrystalline and
transcrystalline failure.
The fatigue damage behavior of 25 µm thick 304 type stainless steel

foils with fine grains of 2.89 Pm was investigated in [12,20]. They found
that the significant intrusion-like slip traces formed at the micrometer-
sized foil surface due to the development of strain localization zone (SLZ),
as shown in Fig. 9. The FIB examination of the SLZ indicates that the ex-
trusions/intrusions-like damage formed in the fine grain, as indicated by an
arrow in the upper right image of Fig. 9. Further characterization of the ex-
trusion/intrusion-like damage by an atomic force microscopy (see the
lower right image of Fig. 9) shows that severe strain localization has de-
veloped to form extrusions/intrusions with about several hundreds of na-
nometers dimensions (peaks/valleys), which may become the preferential
site for crack initiation. Obviously, the strain localization by forming ex-
trusion/intrusion-like damage in the 25 Pm thick stainless steel foil is still
preferred, and that is quite similar to that in the bulk materials.
Thus, the crack initiation was suggested to originate from the intrusion-
like slip traces preferentially. Based on the microscopic observations, it is
concluded that the fatigue damage process of the 25 µm thick foils experi-
enced the following stages similar to the bulk 304 stainless steel: (1) cyclic
strain localization and accumulation in the grains oriented preferentially;
(2) extrusion/intrusion-induced crack nucleation and (3) crack growth from
the nucleation site by the transgranular mode.
(a) Intrusions (b)Slip bands (c) Crack site
Fig. 8 SEM micrographs of fatigued Cu wires showing (left) inrusions on de-

formed grains of a microwire of 20 Pm after N=107 cycles (middle) slip bands
formation on the surface of a bamboo-structured 50 Pm wire after N=108 cycles
and (right) crack site in a bamboo-structured microwire of 125 Pm after N=6u106
cycles [28].
Fig. 9 Microscopic observations of the fatigue damage behavior of the mi-

crobeams: the left image is an overview of the fatigue sample at the stress
amplitude of 234 MPa after 4.5u105 cycles, imaged by SEM, the upper right
image is a close observation of the strain-localized zone (SLZ) imaged by FIB, the
lower right image is an AFM characterization of the SLZ [20].
On the other hand, a few studies of high cycle fatigue properties of

LIGA (Lithographic, Galvanoformung, and Abformung) Ni foils with
thicknesses ranging from several tens of micrometers to hundreds of
micrometers have been performed [10, 22-25]. Reported in [25] is the
micrometer-thick Ni films with columnar grains, clear persistent slip bands
(PSBs) occurred. Especially, it is found that crack nucleated from surface
thickened oxides forming at slip bands. It was found in [40] that fatigue
failure of 26 Pm thick LIGA Ni foils initiated in zones of localized
extrusions and intrusions associated with PSBs. Furthermore, the TEM
examination of the deformation zone reveals that there existed an
unexpected thick oxide on the surface of the PSBs. This PSB oxide
thickening mechanism appears to be the source of crack initiation.
In general, the above studies of fatigue damage behavior of free-
standing metal foils/wire with thickness/diameter larger than 10 Pm clearly
reveal that a fatigue crack can still originate from fatigue
extrusions/intrusions, which act as a micro-notch with a high stress
concentration. This process seems to be similar to that happening in the
bulk materials.
However, for the much thinner films with a thickness less than 10 Pm, it
was [9] argued that the absence of clear slip steps on the sample surfaces
and of dislocation cells in 1.1 Pm thick Cu films indicated that dislocations
moved individually or in small groups. A systematic investigation of fa-
tigue damage behavior of metallic films and multilayers confined by a sub-
strate has recently been conducted in [13,29,30-39]. In particular, the fa-
tigue damage behavior of Ag films was conducted in [30] with a thickness
ranging from 1.5 Pm to 0.3 Pm through dynamic microbeam deflection.
The results show that fatigue failure of thin Ag films is related to the for-
mation of voids, which were observed close to the film-substrate interface
beneath extrusions. These extrusions were preferentially observed in
(100)-oriented grains, whereas (111)-oriented grains were to a large extent
undamaged after fatigue testing. This was suggested to be due to smaller
flow stress in (100)-oriented grains. The extrusions appear as a narrow rib-
bon of squeezed-out material located in the interior of a single grain. The
height of the extrusions was in the range of the film thickness with a regu-
lar surface structure. Beneath the extrusions voids were detected, as indi-
cated by “C” in Fig. 10. The occurrence of voids beneath the extrusions
was observed in all of the experiments performed, whereas closely spaced
slip bands have only been observed one time. The void structure has been
found to extend from the film-substrate interface towards the surface.
These observations imply that the film-substrate interface may act as a
blocking boundary to dislocations and to prevent them from leaving the
film. The top surface is expected to act as a free surface, through which
dislocations can leave the film and form surface steps. Therefore, fatigue
damage still originates from small extrusions formed through dislocation
glide.
Furthermore, low cycle fatigue tests of thin Cu films with a thickness
ranging from 3.0 Pm to 0.4 Pm under an almost constant plastic strain
range (0.83-0.91%) were carried out in [30]. Since the Cu film was
deposited onto an elastic polyimide substrate, the film/polyimide-
composite is strained by tensile loading and unloading while the film is
deformed elastically and plastically in tension and in compression on
loading and unloading, respectively. The Cu films were actually subjected
to tension-compression load. The results show that surface extrusions and
voids at the film/substrate interface dominate fatigue damage.
(a) Cross-section location (b) Dislocation motion

Fig. 10 FIB image of a cross-section that was cut into the extrusion (left)
Schematic of dislocation motion and annihilation as possible mechanism resulting
in the formation of voids and extrusions. Full and open dislocation symbols
correspond to dislocations of opposite signs, produced during loading and
unloading, respectively [13].
Fig. 11 Schematic of the relationship between geometric length scale (film

thickness) and microstructural length scale (mean grain size) in the thin films
studied [32].
The fatigue behavior of 0.2 Pm thick Cu films under the same fatigue
conditions (Plastic strain range, ~0.8%) as used in [30] was investigated
[33]. All these Cu films have almost columnar grains with a size about 1
Pm in diameter. It was found that fatigue crack initiation behavior of the
0.2 Pm thick Cu films is quite different from that of the 3.0 Pm thick Cu
films. While the work in [35] conducted low cycle fatigue tests of thin Cu
films with mean grain size ranging from 0.78 Pm to 0.28 Pm, which is
smaller than the film thickness. A clear fatigue extrusion was found in the
large grain of the 3.0 Pm thick Cu films. To thoroughly understand fatigue
damage behavior in these thin Cu films, systematic work of microscopic
examination was conducted through focused-ion-beam microscopy (FIB)
and TEM [32]. First of all, the metal films were classified into two catego-
ries, namely type I and type II films, as defined in Fig. 11. For the type I
films the grain size is about one micrometer scale, which is typically larger
than the film thickness, and for the type II films, the grain size is normally
submicrometer scale, which is smaller than the film thickness. The micro-
structure parameters and fatigue testing conditions of the above Cu films
were summarized in Table 5.
For the type I films with a grain size about 1 Pm, some clear size effects
are found and summarized below based on FIB plan-view observations of
the film surfaces after damage:
1. Fatigue extrusions/intrusions can still occur in the thin Cu films with

a thickness ranging from 3.0 Pm to 0.2 Pm, as shown in Fig. 12;
2. However, with decreasing the film thickness the dimensions of the fa-
tigue extrusions including height and width decrease (Figs. 12(a)-(c)).
Especially, there are almost no fatigue extrusions/intrusions in the 0.2
Pm thick films, as shown Fig. 12(d);
3. Fatigue damage in the 3.0 Pm thick films was found to originate from
the extrusions/intrusions, which is quite similar as of the bulk
materials (Fig. 12(a)), while fatigue cracking in the 0.2 Pm thick
films (Fig. 12(d)) preferentially initiate along the grain boundaries
and twin boundaries rather than fatigue extrusions/intrusions. As a
result, a transition of fatigue damage behavior from the extrusion-
induced mode to the boundary-induced mode occurs.
(a) 3.0 Pm thick film (b) 1.0 Pm thick film after

after 5u103 cycles after 5u103 cycles
(C) 0.4 Pm thick film (d) 0.2 Pm thick film

after 1u104 cycles after 2u104 cycles
Fig. 12 Extrusions at the surfaces of the fatigued type I Cu thin films with micron-
sized grains imaged by FIB at an angle of 45o. The tensile axis is the horizontal
direction [32].
Table 5 Summary of microstructural parameters and fatigue testing conditions of

the type I and II films reported in [32].
Thin Cu films h (Pm) d (Pm) 'Ht (%) 'Hpl (%) N
3.0 2.19r1.04 1.0 0.91 5u103
Type I
1.0 1.51r0.74 1.0 0.87 5u103
0.4 1.00r0.52 1.0 0.86 1u104
0.2 1.21r0.55 1.0 0.83 2u104
3.0 0.78r0.39 1.3 1.01 6u104
Type II 1.5 0.48r0.22 1.3 0.92 6u104
0.4 0.28r0.09 1.7 1.22 6u104
(a) Extrusion height (b)Extrusion width (c) Extrusion/grain area ratio
Fig. 13 Extrusion height, width and ratio of extrusion to grain area as a function
of the film thickness in the type I Cu films. A best fit with Eq. (6.10) to the
extrusion widths is included as a solid line. The extrusion height, width, and the
extrusion area are the mean values of the measurements, while the error bars are
the standard deviations [32].
Fatigue-extrusion features inside the grains gradually disappear with

decreasing the film thickness even though the applied plastic strain ranges
are almost the same for all films. Quantitative measurement of the height,
width of the extrusions and the ratio of the extrusion area to the grain area
clearly reveals a clear trend of the decrease in extrusion dimension with
decreasing film thickness, as shown in Fig. 13. The variation of the
morphology of fatigue extrusions indicates that localized accumulation of
cyclic plastic strain within grains through the development of fatigue
extrusions is restricted in the very thin films (0.2 Pm).
For the type II films shown in Fig. 14, it was found that fatigue extru-
sions can not form in the submicrometer-sized grains, on the contrary the
extensive cracking along grain boundaries and the surface roughening
were found in the fatigue failured Cu films. A careful comparison of the
film surface before and after fatigue loading (Fig. 14(b)) shows that the
surface roughening in the 1.5 Pm thick films with submicron grains may
result from grain boundary grooving and the potential grain boundary slid-
ing. The findings in the type II films further demonstrates that interface-
induced fatigue damage becomes more preferred in the films with small
length scales either film thickness or grain size being about 200 nm.
(a) Before fatigue
(b)After 6u104 cycles
(c) Damage thin film

Fig. 14 FIB images of fatigue damage in the 1.5 Pm thick type II films imaged by
FIB at an angle of 45o. The tensile axis is the horizontal direction [32].
Summarizing observations of fatigue crack initiation behavior in the

metal foils [7-10,13,17-20,22-25,40], wires [6,28] and thin films [29-39]
confined by a substrate, it is noted that fatigue damage in the small-scale
metal materials can still originate from the strain localization sites where
fatigue extrusions/intrusions develop like that in the bulk materials when
the length scales of the material (film thickness and grain size) are more
than 1 Pm, while as the length scale is less than 1 Pm the suppression of
the formation of fatigue extrusions/intrusions results in the interface-
induced fatigue damage. It is suggested that the decrease in length scale
inhibits the localized accumulation of plastic strain within grains, such as
at extrusions/intrusions and in extended dislocation structures, and pro-
motes the formation of damage such as cracks at twin and grain boundaries
during fatigue. Thus, it is predicted that the fatigue failure in the small-
scale materials with length scale of only a few nanometers would be pref-
erentially caused by interface-related damage.
4. Fatigue Crack Growth
Fatigue crack growth rate is an important factor to determine fatigue

reliability and a necessary value to conduct damage-tolerance design of
bulk materials. Conventionally, fatigue crack growth rate (da/dN) can be
described by the Paris law as:
da / dN C('K ) m (1)
C and m are constants, which are related to microstructures, loading

frequency and waveform, environment, temperature and load ratio. For a
ductile alloy, m is about 2~4. 'K is stress intensity factor range at the
crack tip. Concerning fatigue crack growth of a small-scale material, up till
now few studies have been conducted. The complexity of the fatigue crack
growth behavior in small-scale materials may result from effects of the
ratio of grain size (d) to film thickness (h) and the stress state (plane stress
and plane strain) at the crack tip.
Studies [12, 20, 41] on fatigue crack propagation of 25 Pm thick 304
type stainless steel foils with mean grain size of 2.89 Pm indicate that
fatigue crack growth of the material still exhibited the typical
characteristics from the stage I (shear crack) to stage II (tensile crack),
which is similar to that in bulk metals. Fatigue crack growth threshold
from a notch was determined as 6.8 MPm1/2 [41], which is quite close to
that of the bulk counterpart [42]. The size effects on fatigue crack growth
behavior of rolled Cu, Al and Mo foils about several tens of micrometers
thick were examined [18]. For the foils with thickness up to about 150 Pm
intermittent crack arrest due to interaction with microstructural barriers
and a negative curvature of the crack growth curve were observed, as
shown in Fig. 15. It was explained the results in terms of LEFM based on
the crack resistance and the ratio (rp/h) of the prevalent radius (rp) of
plastic zone to the foil thickness (h) criteria resulting in a transition from
plane stress to plane strain. The criteria are given as follows:
1. rp /h<0.025 for cracking in plane strain

2. rp /h>1 for cracking in plane stress
Fig. 15 Crack length as function of number of loading cycles of Cu foils

(recrystallized) with varying thickness [18].
The radius of the plastic zone may be determined by
1 K max
rp ( ) (2)
2S V y
where Kmax is the maximum stress intensity factor during a loading cycle;
Vy is the yield stress after saturated cyclic hardening. The calculation of the
rp /h of the fatigued foils clearly indicates that the observed variation of
crack growth curve in foils is well correlated with the transition from a
state of plane stress to a state of plane strain with increasing foil thickness.
The fatigue testing of micron-scale Ni-P amorphous alloy thin foils was
conducted through dynamic bending [26]. The small-scale cantilever
beams of Ni-P amorphous alloy were fabricated by FIB. To examine fa-
tigue crack growth behavior of the small-scale films, a micronotch with a
depth of 3 Pm was introduced into samples by FIB milling technique.
SEM observation of the fracture surface of the fatigued Ni-P thin film
shows that striation-like markings have appeared and were deduced to be
fatigue striations. This may suggest that the fatigue crack has propagated
by cyclic plastic deformation at the crack tip, i.e. blunting and resharpen-
ing mechanism like that of bulk materials. Furthermore, the fatigue crack
growth rate under different stress ratios as a function of 'K of the crack tip
was estimated based on the measurement of the striation spacing, which
was assumed to be equivalent to the fatigue crack propagating distance per
cycle. A clear effect of stress ratio on fatigue crack growth rate was found
in Fig. 16 through the fact that the fatigue crack growth rates of the mate-
rial under stress ratios of 0.3 and 0.5 are comparable although, in contrast,
the fatigue crack growth rate at a stress ratio of 0.1 is lower at a given
value of 'K. The fact that the effect of stress ratio on the fatigue crack
growth rate observed here is quite similar to the findings in bulk metals
suggests that crack closure effects may still occur even in such micro-sized
specimens, and thereby affect the fatigue crack growth behavior. It is
worth noting that fatigue loading of micro-sized samples in their study has
led to the extension of cracks by only 2-3 Pm, which should be a short
crack for a bulk metal, as schematically illustrated in Fig. 17(a). In the
bulk metal with a short crack less than 100 Pm, the crack closure effect is
generally less pronounced. However, compared with the specimen size
such a 2-3 Pm crack in the present micro-sized sample should be suffi-
ciently long and may actually not be a short crack, as schematically illus-
trated in Fig. 17(b). As a result, the crack closure effect is likely to be pro-
nounced even for micro-sized specimens.
In general, a physically short crack in a bulk material would become a
long crack in a small scale material, which potentially affects fatigue crack
growth rate and behavior. On the other hand, the stress state at the crack
tip such as plane stress and plane strain is also important factor controlling
fatigue growth rate. However, more future work is needed to be conducted
to understand microscopic process of the crack growth in small-scale ma-
terials.
Fig. 16 Fatigue crack growth resistance curves for micro-sized specimens at

different stress ratios [26].
(a) Ordinary size specimen (b) Micro-size specimen
Fig. 17 Schematic drawing of plastic zone at crack tip for ordinary-sized and
micro-sized specimens [26].
5. Fatigue Strength
Fatigue strength of all kinds of small-scale materials including metals and

amorphous alloys has been given more attention and extensively
investigated for recent years. To understand fatigue strength of the small-
scale materials and their size effects, two categories of small-scale materi-
als, i.e. free-standing thin foils/wires and thin films confined by a
substrate, are presented as follows.
5.1 Fatigue strength of free-standing foils/wires
For metallic wires, tension-tension fatigue tests were conducted [6] (Table
4) for annealed Cu wires with a diameter of 30 Pm, 50 Pm and 95 Pm, re-
spectively. It was found that the number of cycles to failure of the Cu
wires increases with decreasing diameter of the wires. The longer fatigue
life of the 30 Pm-diameter wire was consistent with the larger value for its
yield strength. Systematic studies [28] of fatigue properties of Cu wires
with four different diameters from 10 Pm to 125 Pm were carried out by
using ultrasonic resonance fatigue testing system. The loading frequency is
f=20 kHz and stress ratio R is -1. As shown in Fig. 18, for a certain num-
ber of loading cycles to failure the strain amplitude significantly increases
for the 10 Pm-diameter wires even though the strain amplitude of the 50
Pm diameter wires is lower than that of the 125 Pm-diameter wires. The
improved fatigue properties of the 10 Pm-diameter wire is suggested to be
a result of the increase in the yield stress by a factor of two and the pro-
nounced effect of the surface oxide layer of the thinnest wires, which re-
sults in dislocation pile-ups. Following the same assumptions described for
the uniaxial testing the crack prohibiting effect of the surface oxide layer
for the wires of 10 Pm seems to be more effective. In addition, the im-
proved fatigue performance of the thicker wires, that is the fatigue resis-
tance of the 125 Pm-diameter wires is larger than that of the 50 Pm-
diameter wires, could also be a result of higher tensile strength and fracture
elongation values, dependence of crack propagation paths of the wire di-
ameter as well as the more pronounced notch effects leading to fatal cracks
of thinner wires. The tendency of the variation in fatigue strength with the
wire diameter ranging from 20 Pm to 125 Pm found in [28] is just contrary
to the findings reported in [6]. The differences may be attributed to purity
and tensile properties of the wire, loading mode and number of loading cy-
cles, thus the result of both studies may not be comparable.
Fig. 18 Relationship between strain amplitude and the number of cycles to failure
for Cu micro wires with different diameter [28].
Fig. 19. S-N curves of specimens 100 Pm and 20 Pm thick foils. The frequency
was 70 Hz [7].
Concerning the free-standing foil metallic materials, two kinds of

materials are commonly produced. One is prepared by electro-deposition,
which normally gets equiaxed grains, the other is produced through
rolling, wrought, which mostly gives pancaked/elongated grains. It was
reported in [7] that under R=0 and f=70 Hz and 700 Hz conditions the 100
Pm thick Cu foils always failed after a number of cycles which was 10-30
times lower than that for the 20 Pm thick samples, while at the same
number of cycles, the fatigue resistance of the 20 Pm thick foil is about 1.4
times higher than that of the 100 Pm thick foil (Fig. 19). These results
clearly reveal size effects on fatigue strength of free-standing metal films.
Such size effects are attributed to the proximity of the surface for well-
recrystallized samples, processing parameters and texture for as-rolled
non-annealed samples.
(a) Stress amplitude vs fatigue cycle (b) Fatigue strength

Fig. 20 Cycles to fatigue crack initiation for 304 stainless for R=0.15 and com-
parison of fatigue strength between micrometer-thick 304 stainless steel foil and
bulk 304 stainless steel with different grain size under R= -1 [20].
The fatigue strength of 304 type stainless steel foils with thickness of 25
Pm was examined through dynamic bending microbeams [20]. They
obtained stress amplitude-fatigue life data of the material under the
condition of f=10 Hz and R=0.15, as shown in Fig. 20(a). Considering the
mean stress effect due to non-fully reversed loading here (stress ratio
R=0.15), the Basquin relation modified in [43], which allows one to
predict fatigue response for arbitrary values of various R, was used to fit
the fatigue data. The Morrow relation is given as [43]
Va (V f V m )(2 N f ) b (3)
V a (1 R )
Vm (4)
1 R
where Nf is fatigue life under a certain stress amplitude (Va) and Vm is the
mean stress. Vf and b are parameters. The fitting of the stress amplitude-
fatigue life under R=0.15 (see Fig. 20(a)) by using Eqs. (3) and (4) yields
Vf and b as 1471 MPa and -0.123, respectively. In order to compare the
thin 304 stainless steel foil data with the values of bulk 304 stainless steel,
which is obtained at the condition of Vm = 0 and R = -1, the fatigue life of
the 304 SS foils was predicted based on the Morrow relation, Eq. (3) for
Vm =0 and R =-1 and the obtained fatigue parameters (Vf and b). Fig. 20(b)
presents a comparison of the predicted fatigue life and the literature values
for bulk 304 stainless steel with different grain sizes [44]. It can be seen
that the fatigue strength of the 25 Pm thick 304 stainless steel foil is much
larger than that of the bulk 304 stainless steel with grain size from 1 Pm to
17 Pm. This comparison implies that the potential contribution of the
geometrical dimensions to the enhanced fatigue strength should not be
neglected in the micron scale material. The fatigue endurance limit of the
thin films is 247 MPa, which is about 31% the ultimate tensile strength of
the thin foils (VUTS = 802 MPa). The relation between the fatigue limit and
the tensile strength is consistent with that of the bulk material, that is, the
fatigue limit is generally about one-third of the tensile strength [45]. Even
though fatigue damage behavior in the 25 Pm thick 304 stainless steel foils
still follows that of the bulk material as describe in section 3, but the
fatigue strength has been enhanced due to the influences of both fine grain
size and small geometrical dimensions.
The enhanced fatigue strength in the thin foils subjected to dynamic
bending may be attributed to both microstructural and geometrical dimen-
sions. Firstly, the 2.89 Pm-sized grains of the thin 304 stainless steel foil
promotes yield strength, and thus fatigue strength is improved. This trend
is similar to the finding in thin wires [6, 28] mentioned before. Secondly,
the contribution of the geometrical dimensions may result from two as-
pects. (1) The increase in the ratio of surface to volume may change dislo-
cations activity near the surface of the small scale sample. The increased
fatigue strength in thinner Cu wires (30 Pm in diameter) was attributed to
the absence of the persistent slip bands (PSB) and walls structures [6]. The
dislocations, which have a chance to contribute to the PSB walls by inter-
section processes and dipole formation, were presumably attracted by the
image forces of the nearest surface and would escape more easily from the
sample surface. That may affect the formation of dislocation structures in
the small scale samples. This can be evidenced by the fact that the disloca-
tion cell structures in the grains close to the beam surface disappeared [13].
Therefore, it is suggested that the enhanced fatigue strength should be as-
sociated with the small geometrical dimensions of the samples. (2) The de-
formation gradient induced by bending is expected to increase the disloca-
tion interactions both between statistically stored dislocations (SSDs) and
between geometrically necessary dislocations (GNDs) and SSDs [46]. That
has been manifested by the fact that the bending yield strength (584 MPa)
of the thin foil is higher than the tensile yield strength (454 MPa) of the
thin foil [13]. Therefore, it is expected that the 25 Pm thick 304 stainless
steel foil would have a higher fatigue strength if they are cyclically de-
formed under a non-uniform loading condition. The dynamic bending-
induced fatigue properties may be of importance for the fatigue reliability
design of microbeam-type structural components in MEMS devices.
The low-cycle fatigue tests on 25 Pm thick Cu foils electrodeposited in
the additive-free solution (AF) and the additive-containing solution (AC)
and 33 Pm thick wrought foils were performed in [8] under stress-
controlled conditions. Based on the Basquin equation [45], the relation
between fatigue strength and number of cycles to failure can be described
as
V a V f' ( N f ) b (5)
where Va is the applied stress amplitude, V’f is fatigue strength coefficient,

Nf is the cycles to failure. b is a constant. V’f of the electrodeposited Cu
foils with and without additives and wrought foils were obtained. It was
found that V’f of the wrought thin foils is quite close to that of the bulk Cu,
while that of the electrodeposited Cu foils (AF and AC foils) is higher than
that of the bulk Cu. Again strong fatigue size effects can be evidenced
here. Such a size effect was attributed to the difference in microstructures,
such as grain size. Furthermore, the work in [8] also reported that the
electrodeposited Cu foils exhibited cyclic hardening behavior.
Fig. 21 Effect of core thickness and annealing on low cycle fatigue life (Nf) for the
electro-deposition (DF8) foil [19].
The effects of grain structure, foil thickness and anneal softening on the
flex fatigue of 12 to 35 Pm thick rolled and electrodeposited copper foils
were examined over a wide range of strain amplitude [19]. The results
showed that at a fixed strain range the number of cycles to failure of the
electrodeposited Cu foil with a grain size of about 1 Pm in diameter
increased with decreasing foil thickness, as shown in Fig. 21. Also the
fatigue life was improved by annealing of the foils.
Furthermore, the fatigue testing of free-standing Cu film with a
thickness of 1.1 Pm prepared by electron beam evaporation was conducted
in [9]. The surfaces of the film were covered by 50 nm thick Ti layers. The
Cu film has hundreds of grains (about 0.98 Pm) across the width but only
one or a few through the thickness. A tension-tension fatigue testing with
R=0.1 was performed. The results shows that fatigue resistance of the Cu
films is superior to that previously reported for a set of Al thin films. The
Cu data fall between the data for unnotched and notched stainless steel and
Al sheet metal results reported in the literature. In fact, the Cu film system
used in [9] is a Ti/Cu/Ti multilayered composite. It is believed that in
addition to the decrease in Cu film thickness, the increase in fatigue
strength of the 1.1 Pm thick Cu film may result from the existence of the
50 nm thick Ti layers even though Read did not mention that. That was
demonstrated by subsequent fatigue experiment of metallic multilayers.
Another kind of free-standing thin films is LIGA Ni film, which has

been paid more attention due to its applications for MEMS structures, such
as the case of accelerometers used for the deployment of airbags in cars. It
was revealed in [22] that the thinner and the stronger 70 Pm Ni films have
a higher endurance limits than the 270 Pm thick Pm films. This difference
was attributed partially to the effects of the columnar microstructures [23].
The improved stress-life behavior of the thinner (70 Pm) structures is
comparable to that of hardened bulk Ni. Furthermore, LIGA Ni thin films
with nano-scale equiaxed grains have much greater strengths and fatigue
resistance than LIGA Ni films with micro-scale columnar microstructures
[25].
Above studies mainly concentrate on fatigue properties of metallic free-
standing thin foils/wires with foil thickness/diameter of more than 10 Pm.
However, fatigue properties of free-standing thin films with a thickness of
about several micrometers or less have not been investigated as
extensively as that of thin foils/wires. The main cause should be attributed
to difficulties in experimental technique for such a thin film, as mentioned
in section 2.
5.2 Thin metal films confined by a substrate
For thin metal films confined by a substrate, a systematic work on fatigue

resistance of thin Ag films with a SiO2 substrate was conducted by using
dynamic microbeam deflection method [29]. Fig. 22 presents a damage
map, which shows the relationship between the stress amplitude and the
film thickness of the fatigued Ag films at various mean stress levels
ranging from 126 to 600 MPa. The upper right part of the graph indicates
the formation of fatigue damage of the Ag films within 3.8×106 cycles,
whereas no damage was found for conditions in the lower left corner.
Damaged beams are denoted by open symbols, those that were not
damaged by solid ones. It is clear that with decreasing film thickness the
critical stress amplitude for damage increases.
Furthermore, systemic studies on fatigue strength of thin Cu films with
columnar grains (Type I film defined in section 3) [30, 33] and equiaxed
grains (Type II film) [35] were performed. Fig. 23 presents a comparison
of fatigue strength of the thin Cu films with different grain size and film
thickness. In Fig. 23, the solid symbols indicate type I Cu films with thick-
nesses ranging from 3.0 Pm to 0.4 Pm [30], 0.2 Pm [33], while the half-
solid symbols is type II Cu films [35]. Based on the comparison, three im-
portant results can be drawn:
1. Under almost constant plastic strain range (0.83-0.91%) the number

of cycles to failure of the Cu films increases with decreasing film
thickness of the type I films. This indicates that fatigue resistance is
enhanced in very thin film. In this case, the enhanced fatigue
strength of the thinner films was attributed to strong confinement of
a film/substrate interface on threading dislocation motion due to the
columnar grain structure of the type I film.
2. For the type II films under almost the same plastic strain range, the
number of cycles to failure for the 3.1 Pm thick films is larger than
that of the 1.1 Pm thick films, but both the 3.1 Pm and 1.1 Pm thick
films have less fatigue resistance than 0.4 Pm films.
3. A comparison of fatigue resistance between the type I Cu films and
type II films shows that the number of cycles to failure for the type
II films is larger than that of the type I films except the type I 0.2
Pm thick films. This may be attributed to the fact that the type II
films have submicrometer-sized grains, while the type I films have
micrometer-sized grains. In addition to the confinement of the
substrate, grain boundary strengthening mechanism is also expected
to be involved in the improvement of fatigue resistance of the type II
film.
Fig. 22 A damage map of the Ag films showing the stress amplitude vs. the film
thickness at various mean stress levels ranging from 126 to 600 MPa. Damaged
beams are denoted by open symbols, those that were not damaged by solid ones
[29].
Fig. 23 Comparison of fatigue properties of thin Cu films with coarse- and fine-
grains.
A further investigation of fatigue strength of thin Cu films with thick-

nesses ranging from 50 nm to 128 nm and nanometer-sized grains has re-
cently been performed in [36]. The ultra-thin films were deposited onto a
polyimide substrate. Measurement of fatigue lifetime of the ultrathin Cu
films as a function of the film thickness under 'Ht=1.1% shows that the fa-
tigue lifetime still increases with decreasing film thickness even if the film
thickness is in the range of several tens of nanometers. The comparison of
the nanometer-grained Cu film [36] with the type II films [35] exhibits that
the fatigue lifetimes of the present 50 nm, 60 nm and 90 nm thick Cu films
are higher than those of the type II micron-thick films under 'Ht|1.1%, but
less than that of the 0.4 Pm thick film. This implies that there may be a po-
tential tendency of the decrease in the fatigue resistance of the nanometer-
thick films as compared with that of the submicrometer-thick films. Of
course, it should be noted that the nanometer-thick Cu films were not made
in the same sputter system and conditions as that reported in [35], which
may result in the difference in fatigue life due to the differences in some
microstructural features, such as texture, grain size, purity etc.
Concerning fatigue strength of ultrathin film with nanometer scale
thickness, it was reported [37] that a 5 Pm thick Cu/Ni multilayer with
equal individual layer thickness of 20 nm on a Cu substrate can effectively
suppress fatigue crack initiation through decreasing surface roughening
and enhance fatigue strength of the Cu substrate. This result reveals a ten-
dency that the nanoscale multilayer may have a high fatigue resistance,
which, as a coating material, can effectively improve fatigue strength of
the substrate. Direct evaluation of fatigue strength of free-standing 40 Pm
thick Cu/Nb multilayers with individual layer thickness of 40 nm was per-
formed recently in [14], who used the resonant frequency method as intro-
duced in section 2. Fig. 24 shows the normalized fatigue strength of the
Cu/Nb multilayers. The results show that the ratio of the fatigue endurance
limit to the ultimate tensile strength was about 0.35.
Fig. 24 Normalized S-N curve for the Cu/Nb multilayers with a total thickness 40
Pm and individual layer thickness 40 nm. The normalization is defined as ratio of
maximum stress to ultimate tensile stress [14].
Quite recently, the Cu/Ta multilayers with individual layer thicknesses

ranging from 10 nm to 500 nm were deposited onto a polyimide substrate
through radio-frequency magnetron sputtering method [38,39]. The multi-
layers have a grain size from about 60 nm to 120 nm. Following the same
fatigue testing method adopted in [30,33] fatigue properties of the nanome-
ter-scale multilayers were examined under f=10 Hz and R|0 [39]. The
Cu/Ta multilayers were actually subjected to tension-compression loading
due to the stretch of the elastic polyimide substrate. Fatigue strength of the
multilayer was defined as the critical stress range to cause fatigue cracking
within 1u106 cycles. Fig. 25 presents the critical stress range as a function
of individual layer thickness of the multilayer. It is found that fatigue
strength gradually increases with decreasing individual layer thickness of
the multilayer when the individual layer thickness is more than 20 nm,
while it begins to drop when the individual layer thickness is less than 20
nm. Correspondingly, the applied strain range for the fatigue cracking also
shows a similar trend with the individual layer thickness. The remarkable
individual layer thickness dependent fatigue strength of the metallic multi-
layers found here is similar to that of yield strength of the multilayers.
Fig. 25 Fatigue strength and the total strain range of Cu/Ta multilayer as function
of individual layer thickness. The trend is similar to the relationship of yield stress
and individual layer thickness in monotonic deformation [39].
6. Length-Scale Dependent Fatigue Mechanism
6.1 Dislocation activity at small scale
As mentioned in sections 3 and 4, extensive experimental results have

clearly revealed the effects of film thickness and sample size on fatigue
damage behavior and fatigue strength of the small-scale materials regard-
less of that the foils/films are confined by a substrate or not. However, the
physical mechanism for the size effects of the small-scale materials con-
fined and unconfined may not be the same. Furthermore, the minimum
length scale between geometric and microstructural dimensions will be the
dominant controlling factor.
It can be seen from sections 3-4 that the free-standing foils/wires with a
length scale more than 10 Pm exhibit the similar fatigue damage behavior
as the bulk materials. This may be attributed to the fact that the minimum
length scale of the material is still larger than the fatigue characteristic
length scale, such as dimensions of fatigue extrusions/intrusions, the spac-
ing of dislocation walls and the size of dislocation cells. Nevertheless, the
fact that fatigue strength of the free-standing foils/wires is higher than that
of the bulk counterpart and increases with decreasing the foil thick-
ness/wire diameter still reveals clear size effects.
An explanation was proposed in [6] that the absence of extrusion-like
features on thin wire surfaces up to 105 cycles may be interpreted as the
absence of PSBs similar to that observed in macroscopic specimens. It is
argued that because the space available for periodic slip motion is limited
by the small diameter, especially before sliding dislocation segments have

a chance to contribute to the "knitting" of PSB walls by intersection proc-
esses and dipole formation, they are attracted by image forces of the near-
est surface and annihilated or changed in configuration. Thus, the PSB
wall structure will either not form at all or only after very high number of
cycles. This could explain the longer lifetime of the 25 Pm and 30 Pm wire
as compared to the thicker ones.
The study in [7] showed that the thicker specimens contained a large
number of well-defined and sharp slip bands with in some cases well-
developed extrusions, while the thinner specimens were almost free of ex-
trusions and contained only occasionally a few grains with a low density of
faint slip bands. This indicates that the microstructural development to-
ward failure is somehow delayed for the thinner samples of 20 Pm thick.
TEM examination of the fatigued 25 Pm thick 304 type stainless steel foils
conduced in [12] shows that no PSB dislocation structures were observed
in the micrometer-sized material with mean grain size of 2.89 µm.
The TEM observations of the fatigued 1.1 Pm Cu films [9] concluded
that the absence of clear steps on the specimen surfaces, combined with the
absence of dislocation cells in the fatigued material, indicated that the dis-
locations move individually or in small groups and escape at the specimen
surface. This is different from the behavior of bulk materials, where slip
bands and dislocation cells are commonly reported. It was found in [28]
that vein structure in one grain of a fatigued Cu wire of 50 Pm diameter
was formed. Depending on the amount of plastic strain in individual grains
of the examined wire, a vein structure, elongated cells and regions of cellu-
lar structure near the crack site as well as regions of lower dislocation den-
sity can also appear.
In general, even though a significant size effect on fatigue strength was
found in free-standing foils/wires, typical dislocation structures, such as
vein structure, PSB walls and cell structure can still form in the micron-
scale materials since their minimum dimensions are larger than the fatigue
characteristic length scale. However, the finding in [9] presents an impor-
tant indication that individual dislocations start to move instead of a group
of dislocations to mediate cyclic plastic strain in the 1.1 Pm thick film.
For the thin metal films confined by a substrate, fatigue mechanisms
may not be the same as the bulk materials and the free-standing foils/wires.
Physical reasons may originate from the constraint of a substrate, which
strongly confines dislocation motion in the film channel. It is well known
that dislocations in a thin film confined by a substrate can nucleate within
a film, from grain boundary, film/substrate interface and a film surface.
Whatever it is, the bowing-out dislocation loop will meet the film/substrate
interface and deposit a segment of dislocation line to the film/substrate in-

terface. This will result in an additional work for the threading dislocations
to move in the channel. In view of this, the flow stress to move a threading
dislocation in the channel of an epitaxial film can be calculated in the
model [3] as
sin I b ª 2G f G s § E s ·º
V Nix « ln¨ h ¸» (6)
cos I cos O 4S(1 Q)h ¬ G f G s © b ¹¼
I is the angle between the film normal and the inclined {111} plane , O the
angle between the film normal and Burgers vector b. h is the film thick-
ness. Gf and Gs are the shear moduli of the film and substrate, respectively.
Es=1-2 is a numerical constant defining the cutoff radius of the stress field
of the interfacial dislocation. Eq. (6) clearly indicates that the flow stress
of the epitixial film increases with decreasing film thickness. Concerning a
polycrystalline film, the yield stress for dislocation channeling can be ex-
pressed as [47]
Gb § D ·§ 2 sin I ·
Wy ln¨ ¸¨ ¸ (7)
4S(1 Q) © b ¹© D h ¹
where M is the same meaning as Eq. (7). D is the grain size of the polycrys-
talline film. This equation results from comparing the mechanical work
done in dislocation glide with the energy associated with depositing dislo-
cations in the interfaces. From Eq. (7) one immediately finds an increase in
yield stress of the film with the decrease in the grain size and film thick-
ness.
For the confined metal films, as shown in section 3.1 fatigue damage
behavior of the Cu films with different thicknesses and grain sizes is
strongly controlled by the length-scale [32]. The main findings can be
summarized by the fact that extensive extrusions/intrusions-induced dam-
age occur in the large grains and micrometer-thick films, while boundary-
induced damage dominates fatigue damage of the 0.2 Pm thick Cu films.
(a) 3.0 Pm thick (b) 1.0 Pm thick (c) 0.4 Pm thick
Fig. 26 Plan-view TEM images of the dislocation structures in the fatigued type I
Cu films [32].
The TEM cross-sectional sample of a fatigued 3.0 Pm thick Cu films

was prepared through FIB milling technique [30]. It was found that there
exist less dislocation structures in the films. Subsequently, the TEM plan-
view observations of the fatigued Cu films with different film thickness
and grain size ranging from micrometers to submicrometers were per-
formed [32]. As shown in Fig. 26, a clear effect of grain dimensions on
dislocation structures that develop during fatigue has been observed. As
the film thickness of the type I film is decreased from 3 Pm to below 1 Pm,
the characteristic dislocation wall/cell structures (Fig. 26(a)) are replaced
with loose tangles of dislocations (Fig. 26(b)) and then with individual dis-
locations (Fig. 26(c)). Furthermore, for the type I film with the same thick-
ness, a similar transition of dislocation structure with film thickness was
found in the type II film. Fatigue dislocation structures are generated by
slip that takes place predominately on a single slip system, whereas the in-
dividual dislocations observed here are emitted from grain boundaries and
are active on all three slip systems. Whether the grain dimensions are fixed
by the film thickness or by the in-plane grain size makes no difference in
the dislocation microstructure. Also, whether the grains are columnar or
non-columnar is immaterial; in all cases the smallest grain dimension con-
trols the type of dislocation structure.
(a) Dislocation walls near larger (b) Tangled dislocations near

extrusion in 3 Pm film extrusion in 1.0 Pm film
(c) Long dislocations near smaller (d) Dislocations near crack close
extrusion in 0.4 Pm film to twin in 0.2 Pm film
Fig. 27 TEM observations of the dislocations near fatigue damage regions in the
type I thin films [32].
In fatigued bulk metals, PSBs are typically found near extrusions and
are believed to be the source of the dislocations that generate the extru-
sions [45]. In order to correlate the extrusions and other surface damage in
the thin films with dislocations, the dislocation structures near fatigue
damage regions were also examined [32]. The 3.0 Pm thick type I film
provided the only evidence of a correlation between extrusions and dislo-
cation structures. For example, Fig. 27(a) shows {100} dislocation walls
close to an extrusion in the 3.0 Pm thick film. Cracks are also evident in
the film at the intersection of the extrusion with the GBs (arrows in Fig.
27(a)). For thinner films, such as the 1.0 Pm thick type I film, tangled dis-
locations with a cell-like structure are found near extrusions (Fig. 27(b)).
When the film thickness is decreased to 0.4 Pm, only individual disloca-
tions are observed near extrusions. For example, in Fig. 27(c) many dislo-

cations (marked “1”) with Burgers vector ½[ 110 ] lie on the (111) slip
plane parallel to the plane of the narrow extrusion and to a crack that has

formed on this plane. Dislocations with Burgers vector ½ [ 1 01 ] close to a
crack along a TB were found in a 0.2 Pm thick type I film as shown in Fig.
27(d). This is reminiscent of the dislocation pile-ups near GBs in the 0.2
Pm thick films reported in [33]. For the type II films, intergranular crack-
ing was frequently observed, for a 1.5 Pm thick film. The bowed out dislo-
cations reveal that dislocation emission has occurred from the triple junc-
tions and from the crack at the GB. In the 0.4 Pm thick film, the lighter
regions along GBs and triple junctions indicate that extensive cracking,
thinning, and possibly grain boundary sliding have occurred. No disloca-
tions were found in the neighborhood of the GB grooves in the thinnest
films.
In general, the careful TEM examination also reveals length-scale de-
pendent dislocation structures. Well-defined dislocation walls and cells as
observed in bulk material form when the minimum dimension of the crys-
tal is at least 3 Pm. With length scales ranging from 1 to 3 Pm, rather dif-
fuse, cell-like dislocation structures and small groups of tangled disloca-
tions are observed. Only individual dislocations exist in crystals smaller
than 1 Pm. Possible reasons for this are considered below:
1. Proximity of free surface: In the thin films investigated here, both the
film thickness and the grain size were varied between hundreds of
nanometers and a few microns. No distinct dislocation structures
were found in the thinner films and/or smaller grains, where the dis-
locations are closer to the film surface or grain boundaries. The dislo-
cation image forces are very large close to the free surface and tend to
draw dislocations toward the surface. Combined with the fact that the
glide distance to the free surface is smaller, dislocations may rear-
range and escape more easily in thin films. In fact, such an effect has
been suggested to account for the absence of extrusion-like features
in fatigued thin Cu wires [6]. However, it is unlikely to be the only
explanation for the observed behavior since it is found that grain
boundaries play a similar role to free surfaces in controlling fatigue
dislocation microstructure.
2. Minimum length scale for dislocation self-organization: As we
mentioned before, in bulk metals self-organized dislocation patterns
with dimensions ranging from microns to sub-microns evolve during
fatigue. The spacing between PSB walls in Cu is about 1.3 Pm, while
the spacing of dislocation walls in a labyrinth structure is about 0.75
Pm [48,49]. The minimum size of dislocation cells is observed to be

about 0.5 Pm [50]. All of these imply that the formation of disloca-
tion structures requires a minimum length lpat, which is around 1 Pm.
An investigation of fatigued bulk fine-grained Ni showed that 1 Pm is
the critical grain size for the formation of dislocation structures [51].
In addition, a theoretical evaluation of the effect of grain size on dis-
location patterning indicates that the repeat length of the dislocation
structures approaches the grain size when the grain size is close to 1.0
Pm [52].
For equiaxed grains, the available slip distance of dislocations along a
{111} slip plane is equal to the grain size, and dislocation structures
are expected to form if the grain size is larger than lpat. For columnar
grains in a film of thickness h, the dislocation slip distance L in the
grain can be expressed as
h
L (8)
sinM
where M is the angle of the slip plane normal with the out-of-plane
direction of the film (Fig. 11). Therefore, the minimum film thickness
required for dislocation structure formation in columnar grains is lpat
sinM. Setting lpat equal to 1.3 Pm, which is the PSB wall spacing along
the <110> direction in Cu, the minimum film thickness h required for
the formation of a pair of PSB’s walls can be estimated as 1.1 and 1.2
Pm for <111> (M =70.5q) and <100> (M = 54.7q) out-of-plane
oriented grains, respectively. As a result, either a film thickness or
grain size of less than about 1 Pm would not provide a large enough
glide distance for the formation of PSB walls.
3. Constraints on dislocation nucleation and motion: In bulk Cu single
crystals subjected to cyclic loading, dislocations accumulate on the
primary slip plane, where mutual trapping of edge dislocations and
annihilation of dislocations with opposite signs cause the formation of
vein structures and eventually PSB wall structures [53]. A sufficiently
high dislocation density is needed on the primary slip plane for
dislocation interactions and the formation of structures. However, the
availability and activation of dislocation sources, either within the
grains, at GBs, or at the film/substrate interface, becomes more lim-
ited with decreasing length scale [54]. Furthermore, it has been ob-
served that small grain size and film thickness constrain dislocation
motion [15,55], which has been explained by a number of models in-
cluding “dislocation channeling” [3, 5, 54]. Supporting evidence for

dislocation channeling has been found in fatigued 0.2 Pm thick films
[33]. Both the inhibition of dislocation sources and dislocation
motion mean that at a given applied strain, there are fewer
dislocations available on the primary slip system to interact and form
dislocation structures. Instead, the TEM images support the idea that
multiple slip systems are activated within the grains and GBs become
more important either as barriers to dislocation motion or as
dislocation sources.
6.2 Theoretical analysis
6.2.1 Length scale dependent strain localization
The findings of the decrease in extrusion width with decreasing film thick-
ness and grain size mentioned in section 3 indicate that the localized ac-
cumulation of plastic strain in extrusions is hindered in small dimensions.
Furthermore, as shown in section 5.1 the observation of dislocation walls
near extrusions in the 3.0 Pm thick film is in agreement with the behavior
of bulk materials. However, the smaller extrusions in the 0.4 Pm thick film
are associated with individual dislocations that glide to the film surface.
Presumably, dislocation structures are more effective at achieving plastic
strain than individual dislocations, and the extrusions are smaller when the
dislocation structures are not present. Even in grains where there are no
dislocations structures, the extrusion dimensions decrease with film thick-
ness and grain size, presumably due to the length scale constraints on dis-
location nucleation and motion discussed in the previous section.
The trend in extrusion dimensions with the film thickness and grain size
can be evaluated in terms of the extrusion widths shown in Fig. 13(b).
Suppose the cyclic plastic strain applied to the thin film goes entirely to the
formation of extrusions (i.e., it is assumed there are no other mechanisms
for plastic strain accommodation). Each extrusion consists of a number of
activated slip planes Nslip along which the dislocations glide. The applied
plastic strain can be expressed as:
H pl N slip H slip (9)
where Hslip is the strain caused by dislocation motion in a given slip plane.
Assuming that Į dislocations sweep across the slip plane and are deposited
at the film/substrate interface and at the grain boundaries, as schematically

illustrated in Fig. 11, the plastic strain from a given slip plane is given by
b sin O
H slip D (10)
D
where Į is the number of dislocations that have traversed the slip plane, b
is the Burgers vector, O is the angle of the slip direction with the out-of-
plane direction of the film, and D is the grain diameter. It has been pro-
posed that the distance between active slip planes is equal to the trapping
distance for dislocation dipoles Sd [57] so that
Wext ( N slip 1)S d (11)
where Wext is the extrusion width. The trapping distance of dislocation

dipoles is given by [57]
Gb
Sd (12)
8S(1 Q)W s
where G is the shear modulus, Q is Poisson’s ratio, and Ws is the saturation

shear stress of the material. Therefore, the extrusion width can be obtained
by combining Eqs. (9)-(12) as
§ H pl D · Gb
Wext ¨¨ 1¸¸ (13)
© Db sin O ¹ 8S(1 Q)W s
Thus, for a given plastic strain and saturation shear stress, the extrusion
width should increase with the grain size.
Fig. 28 A schematic of geometry for irreversible slip leading to extrusion forma-

tion in thin films.
The saturation shear stress in a thin film depends on film thickness and
grain size, but has not been measured. For the sake of simplicity, we use
the yield stress Wy instead, which effectively ignores possible contributions
from cyclic work hardening or softening. This is probably a reasonable as-
sumption when no dislocation structures are formed, such as for the thin-
ner and smaller grained films. It is not likely to be accurate for the thickest
films, where dislocation structures form and change the flow stress. For a
polycrystalline film, the yield stress for dislocation channeling as illus-
trated in Fig. 28 can be expressed as [47]
Gb § D ·§ 2 sin M ·
Wy ln¨ ¸¨ ¸ (14)
4S(1 Q) © b ¹© D h ¹
where M is the angle of the slip plane normal with the out-of-plane direc-
tion of the film. This equation results from comparing the mechanical
work done in dislocation glide with the energy associated with depositing
dislocations in the interfaces. Substituting Eq. (6.9) into Eq. (6.8), one can
obtain for the extrusion width as
§ H pl D · 1
Wext ¨¨ 1¸¸ (15)
© Db sin O ¹ 2 ln§¨ D ·¸§¨ 2 sin M ·¸
© b ¹© D h ¹
Based on Eq. (6.10), the extrusion width for the films with different
thicknesses for a given plastic strain was estimated in [32]. Here the plastic
strain Hpl is set to be the mean of the plastic strains for all the films as
measured by in-situ X-ray diffraction [30], Hpl |0.85%. The grain size was
set to be the mean grain size (D=3.58 Pm) within which extrusions formed.
Other material parameters in Eq. (15) are set to be G=45 GPa, Q=0.324,
and b=0.256 nm. O and M are 35.3q and 70.5q, respectively for an <111>
out-of-plane oriented grain. Thus, the only fitting parameter is D, the num-
ber of dislocations active on a single slip plane in a single cycle. The best
fit to the extrusion width data using Eq. (15) is shown by a solid line in
Fig. 13(b), and is obtained with a value of 8 for D. This number is some-
what smaller than the 14 dislocations expected in a pileup of this size and
at this stress. It can be seen that the model (Eq. (15)) proposed in [32] does
qualitatively capture the experimentally observed variation of extrusion
width with film thickness for film thicknesses less than 1.0 Pm. However,
for the 3.0 Pm thick film the predicted value is less than the experimental
average although it does fall within the standard deviation. This may be
due to the fact that the actual grains containing extrusions in the 3.0 thick
Pm film are much larger than the assumed average of D=3.58 Pm, or to the
fact that dislocation structures have formed in these grains and the satura-
tion shear stress is not equal to the yield stress.
In addition, the extrusion width was also predicted by Eq. (15) and is pre-
sented in Fig. 29 as a function of film thickness and grain size for a fixed
plastic strain. A clear trend shows again that the fatigue extrusion width
gets gradually smaller with decreasing film thickness or grain size. As the
dimensions are decreased, the yield stress increases and the slip planes
within a given extrusion become more closely spaced so that the plastic
strain is localized in a narrower region. Eventually the extrusions will ei-
ther be too small to be observed or the stresses required to generate them
will become so large that other mechanisms, such as interface-induced
damage, will take over.
Fig. 29 Calculation of fatigue extrusion width as predicted by Eq. (6.10) as a

function of film thickness and grain size for a fixed plastic strain of 0.85% [32].
6.2.2. On fatigue cracking
Fatigue cracking in bulk Cu polycrystals always occurs along extru-

sions/intrusions within grains [48, 58], or places where PSBs impinge GBs
[58, 59], when the GB properties and geometry are suitable [60, 61]. In the
larger grains of the films thicker than 3.0 Pm, cracks were observed along
extrusions and at the intersection of extrusions with GBs, just as in bulk
materials. This means that the presence of the substrate and the columnar
grain structure do not change the mechanisms for crack formation down to
a length scale of roughly 3 Pm. But when the film thickness or grain size is
further decreased, the stress required to cause plastic strain increases. As a
result, other stress relief mechanisms than extrusion formation may be ac-
tivated, such as crack formation in the absence of extrusions along GBs
and TBs [32, 33].
The previous TEM results [33] show that one stress relief mechanism in
the thinnest films and the smaller grains is the activation of individual dis-
locations on multiple slip systems. Multiple slip will lead to relatively ho-
mogenous deformation within the grain and the dislocations will interact
with the GBs and TBs rather than with each other. It is also frequently ob-
served that the individual dislocations are piled up near GBs and/or TBs.
These pile-ups generated large stresses which may initiate the formation of
cracks at the GBs and TBs. Thus, as film thickness and grain size are de-
creased, the interaction of dislocations with boundaries is expected to in-
crease and cracking along GBs or TBs will become more prevalent.
6.2.3. A mechanism map of length scale-controlled fatigue behavior
Based on the observation of the fatigue damage behavior in the previous

publications [29-31, 33-35], a mechanism map is summarized in Fig. 30 as
a function of film thickness and grain size [32]. The map presents the ex-
perimental observations of surface damage (extrusions as filled triangles
versus interface cracking as open triangles) and dislocation microstructure
(dislocation walls/cells as filled circles, diffuse cells as half-filled circles
and individual dislocations as open circles). The data summarized in Fig.
30 were acquired using somewhat different strain ranges and different cy-
cle numbers, but nonetheless are with each other. Based on the success of
Eq. (6.10) in describing the dependence of the extrusion widths on film
thickness, the same equation is used to divide the mechanism map into
three regimes. The largest grains of the 3.0 Pm thick film showed the larg-
est extrusions in addition to well-defined dislocation structures. Therefore
this extrusion width (2.41 Pm) is used in Eq. (15) to predict the separation
between Regime I for “bulk-like behavior”, characterized by dislocation
structures and surface extrusions, and Regime II for “transition behavior”,

characterized by dislocation tangles and small surface extrusions or intense
slip bands. In the 0.2 Pm thick film interface-mediated damage had begun
to prevail and the extrusions were only just observable. Thus, this extru-
sion width (0.13 Pm) is used in Eq. (15) to predict the separation between
Regime II and Regime III for “small volume” behavior characterized by
individual dislocations and interface-mediated damage behavior such as
cracking along GBs and TBs, voiding at GBs, and grain boundary diffu-
sion.
Fig. 30 Fatigue damage mechanism map for Cu thin films. Experimental observa-
tions are summarized with data symbols: presence of extrusions (S) or absence of
extrusions (U), dislocation structures (z), dislocation tangles , or individual
dislocations ({). The behavior is divided into three regimes described in the text.
Regime I indicates bulk-like behavior, regime II is a transition region, regime III
indicates “small volume” behavior. The lines separating the different regimes
were obtained from Equation (6.10) for extrusion widths of 2.41 Pm and 0.13 Pm
[32].
Fig. 30 can also be separated into two large regimes by a 45q line. The
upper left corner regime corresponds to length scales for which fatigue be-
havior is predominately controlled by the film thickness, whereas the
lower right corner regime corresponds to grain size controlled fatigue be-
havior. However it is important to remember that a wide distribution of
grain sizes in a given film means that there will be competitive and/or cou-
pled mechanisms. The different mechanism regimes as predicted by Eq.
(15) are in reasonable agreement with both the experimentally observed
surface morphology and the dislocation microstructure data. This success
provides support for the idea that the change in surface damage morphol-
ogy and in dislocation microstructure are coupled effects and due to the
same underlying change in mechanism: As it becomes more difficult to
create and move dislocations to form dislocation structures and extrusions,
individual dislocations begin to dominate and other stress relief mecha-
nisms such as interface cracking and diffusion begin to take over.
7. Conclusion and Perspective
The main findings mentioned in previous sections are summarized below.
1. Current fatigue testing techniques to some extent can obtain fatigue

properties of small-scale materials and reveal size effects of fatigue
behavior of small-scale materials.
2. No matter whether small-scale materials are confined by substrates or
not, the fatigue strength of the material is significantly enhanced with
decreasing length scale of the material. This is generally attributed to
the increase in yield strength with decreasing length scales including
both microstructural and geometric dimensions.
3. For thin metal films confined by a substrate, convincible
experimental evidence shows that when either the film thickness or
the grain size is decreased below roughly 1 Pm, the typical
dislocation wall and cell structures found in fatigued coarse-grained
bulk materials no longer develop and are replaced by individual
dislocations. Similarly, the typical surface damage of fatigued bulk
metals, such as extrusions and cracks near extrusions, is gradually
suppressed and replaced by damage that is localized at interfaces,
such as cracks, grooves, and voids along grain and twin boundaries.
This gradual transition from damage characteristic of bulk metals to
damage localized at interfaces is attributed to constraints on
dislocation activity at submicron length scales.
4. Fatigue crack growth rate is strongly influenced by the geometric
dimension of the material, and is well correlated with the transition
from a state of plane stress to a state of plane strain with increasing
foil thickness. It is worth noting that a physically-short crack in a bulk
material would become a long crack in a small-scale material, which
potentially affects fatigue crack growth rate and behavior.
5. Physical mechanism of fatigue damage of the small-scale materials
confined and un-confined by substrates is different. For the free-
standing small-scale materials, the short dislocation glide distance to
the material surface and small grain size are suggested to be main
reasons for the variation of fatigue damage behavior and the
enhanced strength, while for those confined by substrates the strong

constraint of the film/substrate interface on dislocation nucleation and
motion is a dominant mechanism.
6. A mechanism map of the dependence of fatigue behavior on length
scale is proposed based on extensive experimental evidence and a
theoretical analysis for extrusion formation. The suggested extrusion
model not only successfully predicted the change in surface
morphology but also the change in dislocation microstructure with
length scale. It supports the idea that the two transitions are caused by
a single mechanism change. This mechanism is proposed to be the
inhibition of dislocation creation and motion in small volumes.
The strong length-scale dependent fatigue properties and damage

mechanisms of the small-scale materials indicates that the design of small
structures in micro- and nano-systems and the evaluation of their reliability
under fatigue load should not simply follow conventional theories devel-
oped for bulk materials. Instead it is recommended that the input values for
a reliability assessment should be determined by the standard experimental
testing methods specially developed for small-scale materials, and that re-
liability should be estimated by taking the microstructural and interfacial
effects into account. In view of these aspects, further systematic and in-
depth studies on fatigue of small-scale materials with different length
scales and interface structures are necessary to be conducted in the future.
1. Testing technigue. The development of a standard fatigue testing

technique is still needed for small-scale materials with geometric
dimension less than 1 Pm.
2. Characterization methods. Through the standard testing technique the
corresponding characterization methods for the determination of
fatigue strength, crack growth rate, and S-N curves are expected to be
proposed so that the experimental results are comparable.
3. Free-standing small-scale material. Systematic fundamental work on
fatigue of free-standing single crystal metal foils/wires with
thicknesses ranging from micrometers to nanometers should be a
great challenge for one to get insight into length-scale-controlled
fatigue mechanism of small-scale materials. For example, how do the
strain localization and fatigue damage develop when the length scale
of the free standing metal foils/wires is less than 1 Pm?
4. Thin film on a substrate. Much fundamental work on the evolution of
fatigue damage of the metal films with a thickness in a range of 1000
nm to 100 nm and less than 100 nm is required. In the case of the
length scale between 1000 nm to 100 nm, what is the basic
mechanism for the formation of the small fatigue-induced

extrusions/-intrusions in absence of typical fatigue dislocation
wall/cell structures? For the thin films with a thickness less than 100
nm, how do the cyclic strain localization and fatigue damage develop?
5. Dislocation structure. It is necessary to understand the evolution of
dislocation structures with decreasing length scale from micrometer
to nanometer scales, establish the relation between fatigue properties
and dislocation activity in the confined metal films.
6. Interface & surface. There is an indication that the decrease in the
length scale down to submicron-scale or less makes the interfaces
(including grain boundary, twin boundary, film/substrate interface as
well as other heterogeneous interfaces) and surfaces of the materials
more active to be involved in fatigue properties. How do the
interfaces with different structures affect fatigue behavior and
dislocation activities of the materials?
7. Multilayers with multi-physical scales and heterogenous interfaces. A
mutilayer alternatively consists of two or more constituents with
different crystal structure and layer thickness. In such a multilayered
composite, heterogeneous interfaces including the constituent/ con-
stituent interfaces and constituent/substrate interface and different in-
dividual layer thickness from micro- to nano-scales would strongly
change fatigue properties of the materials.
8. Fatigue data base. For engineering design and theoretical modeling,
data base of fatigue strength of all kinds of small-scale materials is
needed to be obtained by the standard testing techniques.
9. Theoretical & simulation. A combination of fatigue experiments and
multiphysics simulations would be helpful to understand and predict
the reliability of the real small-scale structures used in micro/nano
systems. Furthermore, unified understanding and more suitable
theoretical models for fatigue behavior of the small-scale materials
through extending conventional fatigue theories to the micro/nano-
scale field are eventually needed.
References
[1] Ohring M. Reliability and Failure of Electronic Materials and Devices. Aca-
demic Press, San Diego, 1998.
[2] Menz W. and Paul O. Microsystem Technology. Wiley-VCH,Weinheim,
2001.
[3] Nix WD. Mechanical Properties of Thin Films. Metall Trans 20 (1989) 2217
2245.
[4] Spearing SM. Materials issues in microelectromechanical systems (MEMS)

Acta Mater 48 (2000) 179-196.
[5] Arzt E. Size effects in materials due to microstructural and dimensional con-
straints: a comparative review. Acta Mater 46 (1998) 5611-5626.
[6] Hofbeck R. Hausmann K. Ilschner B. and Kunzi HU. Fatigue of Very Thin
Copper and Gold Wires. Scripta Metall 20 (1986) 1601-1605.
[7] Judelewicz M., Kunzi HU., Merk N. and Ilschner B. Microstructural devel-
opment during fatigue of copper foils 20-100 Pm thick. Mater Sci Eng A
18(1996) 62-62.
[8] Hong S. and Weil R. Low cycle fatigue of thin copper foils. Thin Solid Films
283 (1996) 175-181.
[9] Read DT. Tension-tension fatigue of copper thin films. Int J Fatigue 20
(1998): 203-209.
[10] Son DKJ. Lim TW. and Kwon D. Evaluation of fatigue strength of LIGA
nickel film by microtensile tests. Scripta Mater 50 (2004) 1265-1269.
[11] Takashima K. Higo Y. Sugiura S. and Shimojo M., Fatigue crack growth be-
havior of micro-sized specimens prepared from an electroless plated Ni-P
amorphous alloy thin film. Mater Trans 42 (2001) 68-73.
[12] Zhang GP. Takashima K. Shimojo M. and Higo Y. Fatigue behavior of mi-
crosized austenitic stainless steel specimens. Mater Lett 57 (2003) 1555-
1560.
[13] Schwaiger R. and Kraft O. High cycle fatigue of thin silver films investi-
gated by dynamic microbeam deflection. Scripta Metall 41 (1999) 823-829.
[14] Wang YC. Misra A. and Hoagland RG. Fatigue properties of nanoscale
Cu/Nb multilayers. Scripta Metall 54 (2006) 1593-1598.
[15] Hommel M. Kraft O. and Arzt E. A new method to study cyclic deformation
of thin films in tension and compression. J Mater Res 14 (1999) 2373-2376.
[16] Mönig R. Keller RR. and Volkert CA. Thermal fatigue testing of thin metal
films. Rev Sci Inst 75 (2004) 4997-5004.
[17] Klein M. Hadrboletz A. Weiss B. and Khatibi G. The ‘size effect’ on the
stress–strain, fatigue and fracture properties of thin metallic foils. Mater Sci
Eng A319 (2001) 924-928.
[18] Hadrboletz A. Weiss B. and Khatibi G. Fatigue and fracture properties of
thin metallic foils. Int J Fract 109 (2001) 69-89.
[19] Merchant HD. Minor MG. and Liu YL. Mechanical fatigue of thin copper
foil. J Electron Mater 28 (1999) :998-1007.
[20] Zhang GP. Takashima K. and Higo Y. Fatigue strength of small-scale type
304 stainless steel thin films. Mater Sci Eng A 426 (2006) 95-100.
[21] Halford TP. Takashima K. and Higo Y. Fatigue testing of microsized sam-
ples of J-TiAl based material. Mater Res Soc Sym Proc 842 (2005) S6.9.1-
S6.9.6.
[22] Allameh SM. Lou J. Kavishe F. Buchheit T. and Soboyejo WO. An investi-
gation of fatigue in LIGA Ni MEMS thin films. Mater Sci Eng A 371 (2004)
256-266.
[23] Cho H.S. Hemker KJ. Lian K. Goettert J. and Dirras G. Measured mechani-
cal properties of LIGA Ni structures. Sensors and Actuators A: Physical 103
(2003)59-63.
[24] Son D. Kim J. Kim J.Y. and Kwon D. Tensile properties and fatigue crack
growth in LIGA nickel MEMS structures. Mater Sci Eng A 406 (2005) 274-
278.
[25] Yang Y. Imasogie BI. Allameh SM. Boyce B. Lian K. Lou J and Soboyejo
WO. Mechanisms of fatigue in LIGA Ni MEMS thin films. Mater Sci Eng
A 444(2007) 39-50.
[26] Takashima K. and Higo Y. Fatigue and fracture of a Ni-P amorphous alloy
thin film on the micrometer scale. Fatigue Fract Engng Mater Struct 28
(2005)703-710.
[27] Maekawa S. Takashima K. Shimojo M. Higo Y. Sugiura S. Pfister B. and
Swain MV. Fatigue test of ni-p amorphous alloy microcantilever beams. Jpn
J.Appl Phys 38 (1999) 7194-7198.
[28] Khatibi G. Betzwar-Kotas A. Gröger V. and Weiss B. A study of the me-
chanical and fatigue properties of metallic microwires. Fatigue Fract Engng.
Mater Struct 28 (2005) 723-733.
[29] Schwaiger R. and Kraft O. Size effects in the fatigue behavior of thin Ag
films. Acta Mater 51 (2003) 195-206.
[30] Schwaiger R. Dehm G. and Kraft O. Cyclic deformation of polycrystalline
Cu films. Phil Mag 83 (2003) 693-710.
[31] Zhang GP. Schwaiger R. Volkert CA. and Kraft O. Effect of film thickness
and grain size on fatigue-induced dislocation structures in Cu thin films. Phil
Mag. Lett 83 (2003) 477-483.
[32] Zhang GP. Volkert C.A. Schwaiger R. Wellner P. Arzt E. and Kraft O.
Length-scale-controlled fatigue mechanisms in thin copper films. Acta Mater
54(2006) 3127-3139.
[33] Zhang .P. and Volkert C. Damage behavior of 200-nm thin copper films un-
der cyclic loading. J Mater Res 20 (2005) 201-207.
[34] Kraft O. Schwaiger R. and Wellner P. Fatigue in thin films: lifetime and
damage formation. Mater Sci Eng A 319-321 (2001) 919-923.
[35] Kraft O. Wellner P. Hommel M. Schwaiger R. and Arzt E. Fatigue behavior
of polycrystalline thin copper films. Z Metallkd 93 (2002) 392-400.
[36] Zhang GP. Sun K.H. Zhang B. Gong J. Sun C. and Wang ZG. Tensile and
fatigue strength of ultrathin copper films. Mater Sci Eng A
http://dx.doi.org/10.1016/j.msea.2007.02.132, (2007) Article in press.
[37] Stoudt MR. Cammarata RC. and Ricker RE. Suppression of fatigue cracking
with nanometer-scale multilayered coatings. Scripta Metall 43 (2000) 491-
496.
[38] Zhu XF, Zhang G.P. Tan J. and Zhu SJ., Cracking behavior of Cu/Ta bilay-
ered films under cyclic loading. J Mater Res 22 (2007) 2478-2482.
[39] Zhu XF. and Zhang GP. Fatigue properties and damage mechanism of Cu/Ta
multilayers. (2007) Unpublished research.
[40] Boyce BL., Michael JR. and Kotula PG. Fatigue of metallic microdevices
and the role of fatigue-induced surface oxides. Acta Mater 52 (2004) 1609-
1619.
[41] Zhang GP. Takashima K. and Higo Y. Size effects on deformation and fa-
tigue behavior of a micron sized stainless steel. Acta Metall Sinica 41
(2005)337-341.
[42] Kelegtemur MH. and Chaki TK. The effect of various atmospheres on the
threshold fatigue crack growth behavior of AISI 304 stainless steel. Int J. Fa-
tigue 23 (2001) 169-174.
[43] Morrow JD. Fatigue Design Handbook-Advances in Engineering. Society of
Automotive Engineers, Warrendale, PA. 1968.
[44] Di Schino A. and Kenny JM. Grain size dependence of the fatigue behaviour
of a ultrafine-grained AISI 304 stainless steel. Mater Lett 57 (2003) 3182-
3185.
[45] Suresh S. Fatigue of Materials. Cambridge University Press, Cambridge,
1998.
[46] Gao H. Huang Y. Nix WD. and Hutchinson JW. Mechanism-based strain
gradient plasticity-I. Theory. J Mech Phys Solids 47 (1999) 1239-1263.
[47] Thompson AW. and Backofen WA. The effect of grain size on fatigue. Acta
Metall 19 (1971) 597-606.
[48] Essmann U. Gosele U. and Mughrabi H. A model of extrusions and intru-
sions in fatigued metals I. Point-defect production and the growth of extru-
sions. Phil Mag A 44 (1981) 405-426.
[49] Ackermann F. Kubin L.P. Lepinoux J. and Mughrabi H. The Dependence of
Dislocation Microstructure on Plastic Strain Amplitude in Cyclically
Strained Copper Single Crystals. Acta Mater 32 (1984) 715-725.
[50] Mughrabi H. Lowe T.C. and Valiev RZ. Investigations and Applications of
Severe Plastic Deformation. In: Lowe TC and Valiev RZ (eds) Nato science
Series, Kluwer Academic Publishers, Norwell, (2000) 241.
[51] Thiele E. Holste C. and Klemm R. Influence of size effect on microstructural
changes in cyclically deformed polycrystalline nickel. Z Metallkd 93 (2002)
730-736.
[52] Glazov MV. and Laird C. Size effects of dislocation patterning in fatigued
metals. Acta Metall Mater 43: (1995) 2849-2857.
[53] Laird C. Charsley P. and Mughrabi H. Low energy dislocation structures
produced by cyclic deformation. Mater Sci Eng 81 (1986) 433-450.
[54] Blanckenhagen B. Gumbsch P. and Arzt E. Dislocation sources and the flow
stress of polycrystalline thin metal films. Phil Mag Lett 83 (2003) 1-8.
[55] Keller RM. Baker S.P. and Arzt E. Quantitative analysis of strengthening
mechanisms in thin Cu films: effects of film thickness, grain size, and pas-
sivation.J Mater Res 13 (1998) 1307-1317.
[56] Nix WD. Yielding and strain hardening of thin metal films on substrates.
Scripta Metall 39 (1998) 545-554.
[57] Neumann P. Fatigue. In: Cahn RW and Hasen P (eds) Physical Metallurgy,
Elsevier science, Amsterdam, (1983) 1554–1593.
[58] Ma B.T. and Laird C. Overview of fatigue behavior in copper single crystals
-I. Surface morphology and stage I crack initiation sites for tests at constant
train amplitude. Acta Metall 37 (1989) 325-336.
[59] Mughrabi H. Wang R. Differt K. and Essmann U. Fatigue crack initiation by
cyclic slip irreversibilites in high-cycle fatigue. In: Lankford J. Davidson
DL. Morris WL and Wei RP (eds) Fatigue mechanisms: Advances in quanti-
tative measurement of physical damage, ASTM STP, Philadelphia,PA,
(1983) 5-45.
[60] Kim WH. and Laird C. Crack nucleation and stage I propagation in high
strain fatigue--II. mechanism. Acta Metall 26 (1978) 789-799.
[61] Liu W. Bayerlein M. Mughrabi H. Day A and Quested P.N., Crystallo-
graphic features of intergranular crack initiation in fatigued copper polycrys-
tals. Acta Metall Mater 40 (1992) 1763-1771.
Assessment of Fatigue Damage in Heterogeneous
Materials by Application of a Novel Compliance
Technique
H. Mughrabi* and H. W. Höppel
Department of Materials Science and Engineering
Universität Erlangen-Nürnberg
Martensstrasse 5, 91058 Erlangen, Germany
*Email: mughrabi@ww.uni-erlangen.de
Abstract
It is well known that the reduction of the load-bearing cross section caused
by fatigue damage gives rise to a characteristic extrinsic enhancement of
the elastic compliance of the material. It is less well known that, in
particular at higher strains, the elastic compliance effects stemming from
fatigue damage are modified by the superposition of the intrinsic non-linear
elastic compliance of the material. The latter can be assessed quantitatively
and separated from the overall compliance effect, thus enabling the study
of the extrinsic damage-induced compliance effects. The novel technique
yields information on crack propagation, crack opening and crack closure
and has the advantage of being applicable to plain unnotched specimens. It
is particularly suited for the study of fatigued heterogeneous materials in
which early fatigue damage occurs in the form of many statistically
distributed microcracks some of which then coalesce into a larger main
crack. Combined with a parallel replica study of surface microcracks and
their coalescence, a deeper insight into the cracking process can be
achieved. In the following, the potential of this new approach will be
demonstrated by application to specific experimental situations. It will be
shown that under favourable conditions the novel approach is a useful
non-destructive tool to study the crack propagation process in fatigued
plain unnotched specimens in its early and later stages.
1. Introduction & Background
There are many different ways in which fatigue cracks are initiated and
propagate in different materials. The propagation of fatigue cracks has been
studied by a variety of techniques, including compliance measurements
[1,2]. However, compliance measurements are usually performed on
notched specimens in which one single crack has been induced to
propagate. This situation differs considerably from that encountered in
328 H. Mughrabi and H. W. Höppel
fatigued plain unnotched specimens. In this case, early fatigue damage

frequently occurs in the form of a large number of microcracks which grow
and one of which then propagates as the fatal life-determining crack.
Gradual spreading of fatigue damage in the form of microcracks which then
coalesce is very characteristic of fatigued heterogeneous materials in which
pores, inclusions and other microstructural heterogeneities provide a large
number of favourable crack initiation sites [3-9].
This work will focus on the experimental study of fatigue crack initiation
and early propagation in unnotched specimens. Special cases considered
will include subsurface fatigue crack initiation in heterogeneous materials
and the development of concepts to define fatigue damage and to assess the
characteristics of fatigue crack propagation in the case of heterogeneous
materials containing internal defects. It will be shown that suitably
conducted elastic compliance measurements which separate the intrinsic
non-linear elastic behaviour of the material from the damage-induced
elastic compliance effects can be applied advantageously in the case of
fatigued plain (unnotched!) specimens.
2. Subsurface Fatigue Crack Initiation
In recent years, the fatigue failure mechanisms in materials fatigued in the

ultrahigh-cycle fatigue (UHCF) regime, with fatigue lives in excess of 109
cycles to failure, have been studied in some detail [7-9], compare for
example the Proceedings of the recent 4th International Conference on
Very High Cycle Fatigue [10]. Many of these studies dealt with internal
failures in high-strength materials containing inclusions. In these materials,
when fatigued in the range of very low loading amplitudes and ultrahigh
fatigue lives, fatigue cracks initiate at internal inclusions and lead to failure
by so-called “fish-eye” fracture, compare [7-10]. Subsurface fatigue
cracking can also occur at other internal heterogeneities such as pores or
cavities, as found for example in cast materials. Examples of subsurface
cracking at cast porosity are shown in metallographic sections from a study
on the fatigued cast magnesium alloy AZ91 in Fig. 1(a). This mode of crack
initiation leads to a large number of microcracks at the surface which then
grow and coalesce. The subsequent propagation of the main crack then
occurs along the interdendritic areas or through the dendrites, as shown in
Fig. 1(b).
Fatigue Damage of Heterogeneous Material 329
Fig. 1 Fatigue cracks in magnesium alloy AZ91: (a) Initiation at (subsurface) cast
porosity. (b) Propagation along interdendritic areas or through the dendrites. After
[6,11,12].
Fig. 2 Finite element calculation showing high local strains in the narrow bridge of
material between a spherical pore and the surface of a stressed material. After [13].
Fig. 3 Results of finite element calculation, showing maximum axial stress Vmax
normalized with respect to applied stress V for different locations of a pore. The
stress concentration is largest, when the (spherical) pore just touches the free
surface and is higher than that near an interior pore or an open pore at the surface.
From [13].
In a finite element study [13], an attempt has been made to clarify why
fatigue cracks seem to initiate more easily at a pore lying just beneath the
surface than at an open pore lying at the surface. The computations were
performed for a body under an axial stress V, containing spherical pores of
radius r with their centres lying at different distances d from the free
surface. Figure 2 shows as an example the local equivalent strains Hequiv for a
pore located at a distance d | 0.82 r from the surface, corresponding to a
value 0.45 of the “location parameter” r/(r + d). There is a marked strain
localization in the thin bridge of material separating the pore from the
surface. This makes it plausible that, during cyclic deformation, this strain
localization could lead to the initiation of a fatigue crack and ultimately to
failure. It is interesting to consider the local stresses acting in the vicinity of
pores located at different distances from the surface, as plotted in Fig. 3.
Here, the maximum local axial stresses Vmax, normalized with respect to V,
are plotted for different locations of the pore, as shown schematically for
three locations at the bottom. It should be noted that the local maximal
stresses are indeed largest next to pores that just touch the surface (r/(r + d)
= 0.5) and smaller for pores lying deeper in the material (r/(r + d) < 0.5) and
also for “open” pores at the surface (r/(r + d) = 1).
3. Intrinsic Non-Linear Elasticity and Elastic Compliance
3.1 Theoretical basis, stress dependence of Young’s modulus
The differential Young’s modulus of an undamaged material is defined as
dı
E D ı (1)
dİ el
where V is the stress and Hel the elastic strain. In the limit of vanishing stress,
ED approaches the constant value E0 asymptotically. In that case, Hooke’s
law is valid in its linear form:
ı E 0 İ el (2)
and E0 then is the constant Young’s modulus that is commonly used.

However, at higher stresses, ED becomes noticeably stress-dependent in the
sense that ED decreases (increases) with increasing tensile (compressive)
stresses, as has been demonstrated experimentally [14,15]. This
stress-dependence of the elastic modulus originates from the anharmonicity
of the atomic potential. As the material is strained to larger and larger
elastic strains, the anharmonicity of the potential is probed increasingly,
giving rise to a non-linear elastic behaviour. The latter can be taken into
account through an extension of the linear Hooke’s law by adding a
second-order term (corresponding to a third-order term in the potential
energy). One then obtains [14]:
ı E 0 İ el k İ el2 (3)
The constant k describes the deviation from linear elastic behaviour. It is

always negative. According to the work in [16], the constant k can be
expressed in terms of the third order elastic constants. From Eqs. (1) and (3)
the differential Young’s modulus can be expressed as
ED E 02 4kı (4)
The parameters Eo and k are real material constants which are closely
related to the atomic potential curve of the material considered. In alloys or
in more complex multiphase materials, the constants E0 and k should be
viewed as effective material constants averaged over the constituents of the
material. The constants E0 and k can be obtained experimentally in a cyclic
deformation test on an undamaged specimen from the slopes of the elastic
unloadings after the stress reversals in tension and compression and/or by
intermittent elastic unloadings within a closed cycle, as will be shown in
Section 4.2.
3.2 Combined representation of damage-induced compliance effects and

intrinsic non-linear elastic behaviour
The non-linear elastic behaviour observed after damage has developed can
still be described by eq. (3). However, the differential tangent modulus ED
and the parameters E0 and k can now no longer be considered as real
material constants and begin to deviate from their original values, reflecting
an increasingly non-linear behaviour. The latter is characterized by reduced
values of ED as long as the cracks (cavities) are open. At the same time, a
decrease of Eo and an increase in the magnitude of the negative parameter k
are noted [6,11,17,18-20]. With the difference 'ED between the intrinsic
value E0 of ED of the undamaged specimen and the corresponding modified
value of the damaged specimen with opened cracks, 'ED/E0 can be defined
as a suitable damage parameter [6,11,15,17,18] which corresponds
approximately to the relative loss of load-bearing cross section. This
approach is related to earlier work by Lloyd [21] on damaged material. As
can be shown easily, the quantity 'ED can be expressed as
ǻE D E 02 4kı T E D ı T (5)
Here, VT is the stress at the tensile reversal point, and ED(VT) is the stiffness
of the damaged specimen at that point. As stated before, 'ED can also be
obtained from a series of elastic unloadings within a closed cycle, as will be
shown in Section 3.3 and 4.2.
3.3 Identification of intrinsic material parameters
In order to apply the concepts described above, it is essential to first

determine the intrinsic material parameters ED(V), E0 and k. This is done in
a straightforward manner by analysis of the hysteresis loop of an
undamaged specimen. In order to achieve maximum accuracy and as

explained elsewhere [11,14,15,19], it is expedient to plot the stress not
against the total strain Ht but against the plastic strain Hpl = Ht -V/E. In this
expression, the elastic strain has been taken into account in the linear Hooke
approximation. In the V vs. Hpl-plot, the elastic line becomes vertical in the
limit of vanishing stresses for ED(V) = E0. At larger positive or negative
stresses, deviations from the vertical slope occur which indicate lower and
higher values of ED(V), respectively. From the slopes which can be
evaluated accurately, the differential Young’s moduli can be obtained as
described in the Appendix. An example of measurements, showing a series
of elastic unloadings and reloadings in a closed cycle, is shown in Fig. 4.
The material investigated was the metal matrix composite (MMC)
AA6061-Al2O3-15p-T6 which is reinforced with a volume fraction fp =
15% of Al2O3 particles. The unloading-reloading plots exhibit easily
recognizable deviations from the vertical slope, indicating that ED(V) < E0
for V > 0 and ED(V) > E0 for V < 0, as expected. In a first approximation, the
differential Young’s moduli at the tensile and compressive peak stresses VT
and VC can be evaluated from the slopes of the elastic curves immediately
after stress reversal. With these two values, an optimal fit to eq. (3) yields E0
and k. Alternatively, the elastic unloading-reloading data can be evaluated
to yield values of ED for the stresses from which the unloadings were
performed. An optimal fit of these data to eq. (3) yields accurate values of
E0 and k. For the example shown in Fig. 4, the following values were
obtained: E0 = 88.6 r 0.3 GPa and k = -318 r 20 GPa [20]. Fig. 5 shows an
example of the fit of the actually measured stress dependent values ED(V) to
eq. (3) for the case of a bainitic variant of the roller bearing steel SAE 52100
[14].
Once the elastic data of the undamaged material are known, the
development of fatigue damage can be followed continuously during the
fatigue test by repeating the procedures described and noting the changes in
ED(V), E0 and k compared with the values of the undamaged specimen. This
will be explained in more detail in the examples to be discussed
subsequently.
Fig. 4 Hysteresis curve of undamaged fatigued MMC AA6061-Al2O3-15p-T6,

with elastic unloadings and reloadings. Note that the stress V is plotted against the
plastic strain Hpl (obtained in the linear Hooke approximation), providing higher
resolution. After [20].
Fig. 5 Stress dependence of differential Young’s modulus ED of a bainitic variant of

the roller bearing steel SAE 52100, as a consequence of non-linear elasticity
effects. From [14].
4. Examples of Assessment of Fatigue Damage Based on Changes of

Non-Linear Elastic Compliance
4.1 Fatigue damage in a particulate-reinforced metal-matrix composite
The study of fatigued particulate-reinforced metal matrix composites is

suitable to demonstrate the usefulness of compliance measurements and the
assessment of the variation of the parameters E0 and k as a measure of the
developing fatigue damage [17,18,20]. Figure 6 shows an example of such
a study on the fatigued MMC AA6061-Al2O3-20p-T6 which contains a
volume fraction fp = 20% of Al2O3 particles. The data refer to different
values of the total strain amplitude 'Ht/2 in strain-controlled fatigue tests at
half the fatigue life (N = Nf/2, where N is the number of cycles and Nf the
number of cycles to failure). In the fatigued MMC studied, fatigue damage
has been identified metallographically as the fracture of the particles,
compare also [21], which occurs when the total strain amplitude 'Ht/2
exceeds a value of ca. 0.004. It is evident that, with increasing damage, E0
which is initially constant (97.9 GPa) loses its meaning as a material
constant and starts to decrease as a result of the decreasing effective
load-bearing cross section. In parallel, the elastic behaviour becomes
increasingly more non-linear. This is reflected in the fact that the magnitude
of the parameter k which is initially constant (-502 GPa) increases as the
damage develops.
In the case of MMCs, it is difficult to relate the changes in the
non-linear compliance quantitatively to the loss of the load-bearing cross
section. One reason is that the Young’s moduli of the matrix (| 68 GPa) and
the particles (| 380 GPa) differ very strongly. Another reason is that the
damage is internal and distributed more or less homogeneously throughout
the volume and not in the form of discrete cracks. As will be shown in the
next section, the approach described can be applied more specifically and
with better accuracy to materials in which the damage prevails in the form
of cracks that are initially confined to the surface region and then propagate
into the interior.
Fig. 6 Evolution of fatigue damage in particulate-reinforced metal-matrix

composite AA6061-Al2O3-20p-T6, shown the variation of the parameters Eo and k
with increasing damage (particle fracture) as a function of total strain amplitude
'Ht/2 at half fatigue life, N = Nf/2. An increasingly non-linear elastic behaviour
(with increasing _k_) is found as the particles fracture at 'Ht/2 > 0.004. From [17].
4.2 Combined study of propagation of surface cracks in fatigued

magnesium alloy AZ91 by replica techniques and compliance
measurements
In Section 2, the initiation and propagation of fatigue cracks in the cast

magnesium alloy AZ91 had been discussed qualitatively. Here, it will be
shown that the compliance technique introduced before lends itself to study
in detail specific features of the crack propagation process in the alloy
AZ91. This approach can be combined advantageously with the study of
crack evolution at the surface by a suitable replica technique, as introduced
in [22]. In that case, the fatigue test is interrupted repeatedly in order to take
acetate replicas of the surface of the gauge length. Before viewing these
replicas at fairly high magnification in a scanning electron microscope, a
thin layer of gold is evaporated onto them in order to enhance the contrast
and to provide electrical conductivity. By inspection of the sequences of
replicas taken, one can then trace back from the fatal crack to the origin of
microcrack nucleation and follow the process of crack coalescence and
propagation as seen on the surface. Fig. 7 is an example showing the same
location of the surface of a fatigued specimen of the alloy AZ91 with the
main crack, as seen in two (composite) replicas taken after different
numbers of cycles. From series of traces of the crack distribution in fatigued

specimens of the alloy AZ91 it can be concluded how the main crack
originated by a coalescence of a larger number of microcracks., as seen in
the example shown in Fig. 8.
Fig. 7 Example of surface replicas taken from same location of fatigued

magnesium alloy AZ91after different numbers of cycles. Plastic strain amplitude
'Hpl/2 = 9u 10-5, room temperature. After [6,11].
Fig. 8 Evolution of surface crack growth and coalescence in fatigued magnesium

alloy AZ91, as obtained from successive replicas. Total strain amplitude 'Ht/2 =
5u 10-3, room temperature. From [6,11].
(a)Early in fatigue life; N/Nf | 10% (b)Later in fatigue life; N/Nf | 80%
Fig. 9 Hysteresis loops of fatigued magnesium alloy AZ91. These are plots of
stress V as a function of plastic strain in linear Hooke approximation. 'Ht/2 = 2.25 u
10-3, room temperature. After [6,11,12,15].
Complementary compliance measurements made during intermittent

elastic unloadings and reloadings during a closed cycle reveal the loss of the
load-bearing cross section and provide valuable additional information.
This becomes evident when one compares the hysteresis loops with the
unloadings/reloadings, as recorded early in fatigue life, see Fig. 9 (a), and
near the end of fatigue life, shown in Fig. 9 (b). Early in the fatigue test,
there is no influence from cracks. The slopes of the unloading/reloading
paths are vertical in the limit of vanishing stresses and deviate marginally
(to a large positive value) at higher tensile stresses and in the opposite sense
at higher compressive stresses, as is typical of the intrinsic non-linear
elastic behaviour. Once cracks have developed and grown over a larger
portion of the cross section, a completely different situation is observed
which reflects the superposition of an appreciable additional non-linear
component due to the fatigue damage (Fig. 9 (b)). Now the slopes of the
unloading/reloading paths deviate strongly and in opposite sense in the
tensile and compressive phases of the cycle, respectively, from the initially
vertical slope. The variation of the differential elastic modulus ED during
the cycle can be evaluated (with appropriate values of the constants E0 and
k) and plotted against the stress as shown in Fig. 10. In this figure, two
important features can be recognized, namely the intrinsic stress
dependence of ED when the cracks are closed (upper inclined line and
dashed extension) and the stress dependence due to damage, when the
(main) crack has opened in tension (lower inclined line). Moreover, there is
a jump 'ED between the two lines which allows to recognize the transition
from crack opening to crack closure (at ca. -25 MPa) and vice versa. Crack
opening can be seen to occur at a tensile stress of about +12.5 MPa. This
behaviour is appropriately described by the damage parameter 'ED/E0, as
will be discussed next.
Fig. 10 Differential elastic modulus ED (stiffness), determined by intermittent

elastic unloadings, see Fig. 9 (b), showing crack opening at ca. +12.5 MPa and
crack closure at ca. -25 MPa. 'Ht/2 = 2.25 u 10-3, N/Nf | 80%. After [6,11,12,15].
Fig. 11 Evolution of crack length (main crack) of fatigued magnesium alloy AZ91
at the surface (replicas) and in the bulk (see text), as derived from damage
parameter 'ED/Eo. in different ways. After [6,11,12,15].
Based on the assumption that, after the initial stage of microcrack

initiation, there is one main fatal crack, crack growth can be evaluated in
different ways by assessing the length of the main crack as a function of the
number of cycles. Data thus obtained are plotted in Fig. 11. The damage
parameter 'ED/E0 is in good approximation proportional to the length of the
surface crack, as determined from the replicas, up to about 80% of fatigue
life, whereupon it deviates to higher values, as the crack propagates deeper
into the material [6,11]. The continuous curve represents an automatic
calculation of the damage parameter 'ED/E0 from the stiffness values at the
tensile reversal points of the hysteresis loops (with 'ED computed
according to eq. (5)). There is a remarkably good agreement between these
values and those determined by the elastic unloadings/reloadings. In
summary, it can be said that the compliance measurements, evaluated with
appropriate consideration of the intrinsic non-linear elastic properties, can
yield detailed information on the crack propagation, including crack
opening and crack closure stresses, in fatigued plain unnotched specimens
of materials such as the cast magnesium alloy AZ91. Further
complementary information can be obtained by combining the compliance
technique with a replica study of the evolution of fatigue cracks at the
surface. In closing, it is mentioned that a mesomechanical model has been
developed which describes fatigue crack propagation and fatigue life of the
alloy AZ91 very satisfactorily [23].
5. Summary and Conclusions
The studies discussed above lead to the following conclusions:

x The non-linear compliance of a fatigued plain unnotched specimen
includes contributions from the intrinsically non-linear elastic
material properties and from the damage-induced effects on the
compliance.
x The intrinsic non-linear elastic behaviour reflects the
anharmonicity of the atomic potential. It can be assessed
experimentally in cyclic deformation tests and evaluated in terms
of the Young’s modulus in the limit of vanishing stresses and a
parameter that characterizes the deviations from linearity at higher
stresses.
x Compliance measurements, conducted on fatigued plain unnotched
specimens and evaluated with appropriate consideration of the
intrinsic non-linear elastic properties, are a useful tool to monitor
the evolution of fatigue damage. Thus, detailed information on
crack propagation in fatigued plain unnotched specimens,
including crack opening and crack closure stresses, can be

obtained.
x The technique is particularly suitable to study fatigued materials in
which one main fatal crack develops out of the initially large
number of microcracks.
x The compliance technique can be combined advantageously with a
replica study of the evolution of fatigue cracks at the surface to
yield complementary information.
x The experimental approach described is novel and has the merit of
providing crack growth data in a non-destructive manner during a
standard strain-controlled fatigue test on a plain unnotched
specimen without the need to use a standardized fracture mechanics
specimen.
Appendix: Determination of differential Young’s modulus ED from

slopes of elastic unloading curves in stress vs. plastic strain plots
If the elastic unloading curve in the V vs. Hpl-plot (with plastic strain defined
as previously, i.e. Hpl = Ht - V/Hel) has a stress-dependent slope S(V) at a stress
level V, then this slope is related to the differential elastic modulus ED(V) in
the V vs. Ht-plot and the Young’s modulus E0 in the limit of vanishing stress
as
1 1 1
(A1)
SV E D V E0
Resolving for ED(V), it follows that the stress-dependent differential

Young’s modulus can be derived from the slope S(V) by the relation
E 0 SV
E D V (A2)
E 0 SV
References
[1] Saxena A. and Muhlstein CL. Fatigue crack growth testing.In: ASM
Handbook,Vol. 19, Fatigue and Fracture, ASM International (1996) 168-184.
[2] Rroek D. Elementary Engineering Fracture Mechanics, Sijthoff & Noordhoff,
Alphen aan den Rijn (1978) 316-321.
[3] Mayer HR. Lipowsky H. Papakyriacou M. Rösch R. Stich A. Zettl B. and

Tanzl-Tschegg S. Application of ultrasound for fatigue testing of lightweight
alloys, Fatigue Fract Eng Mater Struct 22, (1999) 591-599.
[4] Ting JC. and Lawrence FV. Modeling the long-life fatigue behavior of a cast
aluminum alloy, Fatigue Fract Eng Mater Struct 16, (1993) 631-647.
[5] Couper MJ. Neeson AE. and Griffiths JR. Casting defects and the fatigue
behaviour of an aluminium casting alloy, Fatigue Fract Eng Mater Struct 13,
(1990) 213-227.
[6] Eisenmeier G. Holzwarth. Höppel HW. and Mughrabi H. Mater Sci Eng
319-321 A, (2001) 578-582.
[7] Murakami Y. Nomoto T. and Ueda T. Factors influencing the superlong
fatigue failure in steels, Fatigue Fract Eng Mater Struct 22, (1999) 581-590.
[8] Bathias C. There is no infinite fatigue life in metals, Fatigue Fract Eng Mater
Struct 22, (1999) 559-565.
[9] Mughrabi H. Specific features and mechanisms of fatigue in the ultrahigh
cycle regime, Int J Fatigue 28, (2006) 1501-1508.
[10] Allison JE. Jones JW. Larsen JM. and Ritchie RO. Eds: Proceedings of 4th
International Conference on Very High Cycle Fatigue (VHCF 4), TMS (The
Minerals, Metals & Materials Society) (2007).
[11] Eisenmeier G. Ermüdungsverhalten der Magnesiumlegierung AZ91:
Experimentelle Untersuchungen und Lebensdauervorhersage, Doctoral
Thesis, Universität Erlangen-Nürnberg (2001).
[12] Eisenmeier G. Höppel HW. Ding HZ. and Mughrabi H. Ermüdungs-
lebensdauer der Magnesiumlegierung AZ91: Experiment und Modellierung.
In: DFG 2. Kolloquium Lebensdauervorhersage, Berlin, Deutscher Verband
für Materialforschung und –prüfung e.V., (2002) 169-178.
[13] Borbély A. Mughrabi H. Eisenmeier G. and Höppel HW. A finite element
modelling study of strain localization in the vicinity of near-surface cavities as
a cause of subsurface fatigue crack initiation, Int. J. Fracture 115, (2002)
227-232.
[14] Sommer C. Christ HJ. and Mughrabi H. Non-linear elastic behaviour of the
roller bearing steel SAE 52100 during cyclic loading, Acta Metall Mater 39,
(1991) 1177-1187.
[15] Mughrabi H. Assessment of fatigue damage on the basis of nonlinear
compliance effects. In: Handbook of Materials Behavior Models, Ed J.
Lemaitre, Academic Press, (2001) 622-632.
[16] Wasserbäch W. Non-linear elastic behaviour and third-order elastic constants
of a roller-bearing steel and some standard steels, Phys. Stat. Sol. (b) 164,
(1991) 121-125.
[17] Hartmann O. Biermann H. and Mughrabi H. Cyclic stress-strain and fatigue
behaviour of particulate-reinforced Al-matrix composites. In: Low Cycle
Fatigue and Elasto-Plastic Behaviour of Materials, Eds K. T. Rie & P. D.
Portella, Elsevier Science Ltd., (1998) 431-436.
[18] Biermann H. Kemnitzer M. and Hartmann O. On the temperature dependence

of the fatigue and damage behaviour of a particulate-reinforced metal-matrix
composite, Mater. Sci. Eng A 319-312, (2001)671-674.
[19] Mughrabi H. Some aspects of fatigue of engineering materials. In: Fracture
Mechanics: Applications and Challenges, Invited Papers presented at the 13th
European Conference on Fracture, ESIS Publication 26, M. Fuentes, M.
Elices, A. Martin-Meizoso, J. M. Martinez-Esnaola, Eds, Elsevier, (2000)
13-28.
[20] Biermann H. Beyer G. and Mughrabi H. Spannungs-Dehnungs-Hyterese-
kurven wechselverformter Proben eines etall-Matrix-Verbundwerkstoffes. In:
Verbundwerkstoffe und Werkstoffverbunde, Ed. G. Ziegler, DGM
I nformationsgesellschaft, Oberursel, (1996) 197-200.
[21] Lloyd DJ. Aspects of fracture in particulate reinforced metal matrix
composites, Acta Metall Mater 39, (1991) 59-71.
[22] Ebi G. Ausbreitung von Mikrorissen in duktilen Stählen, Doctoral Thesis,
Rheinisch Westfälische Technische Hochschule Aachen (1987).
[23] Ding HZ. Eisenmeier G. and Mughrabi H. A mesomechanical model of
fatigue crack initiation and growth with application to the low-cycle fatigue
life prediction of the cast magnesium alloy AZ91. In: Proceedings of Fourth
International Conference FATIGUE 2000: Fatigue and Ductility Assessment
of Materials, Components and Structures, Eds. M. R. Bache et al, EMAS Ltd,
( 2000) 209-217.
Fatigue Crack Growth of Aircraft Aluminum Alloys
Sp. G. Pantelakis1* and Al Th. Kermanidis2
1
Laboratory of Technology and Strength of Materials (LTSM) Depart-
ment of Mechanical Engineering & Aeronautics University of Patras,
26500 Patras, Greece
2
Department of Mechanical and Industrial Engineering, University of
Thessaly, 38334 Volos, Greece
*Email: pantelak@mech.upatras.gr
Abstract
The microstructural characteristics and the influence of existing corrosion

on fatigue crack growth in 2024 T351 aircraft aluminium alloy are investi-
gated. Examined is the elongated grain morphology due to the rolling
process that exhibited higher crack growth resistance than the microstruc-
ture of the sheet material with smaller equiaxed grains. Moreover, plate
microstructure with high purity composition offered better damage toler-
ance characteristics in contrast to the conventional sheet microstruc-
ture.Shown is the need to account for the influence of pre-existing corro-
sion on the aluminum alloys in order to obtain more reliable results from
fatigue and damage tolerance analyses of components with corroded areas.
1. Introduction
Damage tolerance analyses and reliable prediction of crack growth under

fatigue loading rely heavily on experimental data and the comprehensive
understanding of the underlying crack growth mechanisms. Fatigue dam-
age in the material can be divided into a number of subsequent phases
characterized by cyclic slip, crack nucleation, micro-crack growth and
macro-crack growth up to the final material failure. The different physical
processes, which prevail in the gradual fatigue damage accumulation dur-
ing the fatigue life of a metallic component, are complex and interrelated.
They develop with increasing number of fatigue cycles from atomic to
macro-scale damage mechanisms and also entail a host of material, geo-
metric and loading parameters [1]. Their interaction cannot be easily as-
sessed in quantitative terms. Mechanistically, the fatigue damage phases
mentioned above may be summarized in the crack initiation and crack
growth stage, with the latter starting once cracks become visible. Different
quantities are essential to assess crack initiation or crack growth respec-
346 Sp. G. Pantelakis and Al Th. Kermanidis
tively. For crack initiation relevant is the severity of stress concentration.

Material properties and loading conditions control principally the behav-
iour of long cracks. Crack growth is dominated by the intensity of the
crack-tip stress distribution. The volume energy density criterion, [2-4],
which first has been used to predict failure initiation under monotonic
loading, can also be applied to predict crack initiation and crack growth
under cyclic loading. It assumes that failure occurs in a critical element
when the volume energy density reaches a certain threshold. The crack ini-
tiation model in [2] applies also to crack propagation [3,4], because atten-
tion is focused on a material element in a specimen without initial crack or
in a local region ahead of a pre-existing crack. According to the crack clo-
sure theory [5], crack growth is dependent to crack closure, which is corre-
lated to several features, such as fracture surface roughness, residual plas-
ticity in the wake or ahead of the crack tip, oxide debris etc. Herein, the
role of microstructure and material properties ahead of the crack tip is im-
portant. For instance, fracture surface roughness is correlated to the grain
morphology of the alloy; the amount of residual plasticity in the wake of
the crack is dependent on the material’s yield stress. The same property in-
fluences crack growth retardation induced by the occurrence of an over-
load.
While these aspects are important in the development of damage toler-
ance analysis tools, the availability of sufficient experimental data is a pre-
requisite.
In the present chapter experimental fatigue crack growth data for se-
lected aircraft aluminum alloys are presented and discussed with the sup-
port of metallographic analyses. The work presented, resulted from two
extensive studies. In the first study the effect of variation in microstructure
of the 2024 T351 alloy due to the aluminum product form (sheet or plate)
as well as the purity content (Fe-Si content) of the alloy on the fatigue
crack growth were investigated [6]. In the second study the fatigue crack
growth behaviour of pre-corroded 2024 T351 alloy was investigated and
discussed under the view of a synergetic effect of corrosion and corrosion-
induced hydrogen embrittlement [7].
1.1 Aircraft aluminum alloys
Aluminum alloys are the primary material for aircraft structures since
1930. Despite the increasing use of advanced polymer matrix composites
in aircraft structures, aluminum alloys remain the main choice for fuselage,
wings and supporting structure of commercial airliners. High specific
properties, superior damage tolerance behaviour and good corrosion resis-
tance along with well known performance characteristics, design experi-
ence and established manufacturing methods represent the main reasons

for the continued wide use of aluminum alloys in modern aircrafts. The
2xxx, 6xxx and 7xxx aluminum series as well as some Al-Li alloys enjoy
the widest use in aircraft structural applications.
The 2xxx series alloys contain copper as the primary alloying element,
which produces high strength but reduced corrosion resistance. The 2024
alloy is probably the most widely used alloy in aircraft applications. The
lower yield stress of the 2xxx series alloys compared to the 7xxx series is
compensated by improved damage tolerance, thus making this alloy series
the first choice for damage tolerant primary structure applications. The
2xxx series alloys are not easily weldable but new alloy types along with
advanced welding technologies make welding of these alloys manageable.
The 6xxx series contains magnesium and silicon, which form the mag-
nesium silicide (Mg2Si). The alloys of this series offer a good balance be-
tween corrosion resistance and strength along with a good weldability. The
most popular alloy of the 6xxx series for use in aircraft applications is the
6061.
In the 7xxx group the primary alloying element is zinc. In this group
two types of alloys exist: the aluminum –zinc –magnesium alloys and the
aluminum – zinc-magnesium-copper alloys, with the second type being
less corrosion resistant. This series includes some of the most high strength
aluminum alloys such as 7075 and 7178, which are exploited for aircraft
frames and structural components. All above mentioned aluminum alloy
series are heat treatable.
Aluminum –Lithium alloys are attractive for lightweight aircraft appli-
cations due to the element Lithium, which is the lightest metallic element.
They are often alloyed with copper, magnesium, zirconium and other ele-
ments to improve properties. They have been used in limited applications
so far (2090 and 8090 are typical alloys used) mainly due to low corrosion
resistance and anisotropic behaviour. New weldable Al-Li alloys of the
2xxx-series (copper being the greatest alloying element) seem to be attrac-
tive for further applications.
2. Microstructural Effects on Fatigue Crack Growth
It is known that microstructural features influence fatigue crack growth in

the near threshold (low ¨Ȁ) regime where the slip characteristics are sig-
nificant [8]. In the low ǻȀ regime crack growth has been found to be sen-
sitive to crack closure, namely plasticity induced crack closure, corrosion
oxide closure and surface roughness crack closure [5, 9-13]. In [14] the
effect of sheet/plate microstructure on the 2090 aluminum–lithium
alloy on fatigue crack growth in the near threshold region has been
demonstrated. At low ǻȀ values in the order of 10-10 m/cycle the sheet
material exhibits significantly higher crack growth rates compared to the
plate material.
In the crack growth region corresponding to medium ¨Ȁ values, the
crack plane becomes normal to the applied nominal stress and crack
growth sensitivity on microstructure is considered to be less significant.
Nevertheless, a correlation between microstructural features and closure
mechanisms in long crack growth behaviour of aluminum alloys seems to
exist. Specifically, surface roughness induced closure (RICC) mechanisms
have been found to influence Stage II fatigue crack growth rates [10-11,
14-15]. The closure mechanisms have been attributed to grain morphology
characteristics. Additionally, crack deflection and consequent crack clo-
sure from wedging of fracture-surface asperities, which are related to
variations in the degree of recrystallization, grain morphology and texture
(evident between sheet and plate product forms) may increase the crack
growth resistance of the material [14].
When fatigue load interaction effects are present the complex damage
accumulation phenomena make a correlation between fatigue crack growth
mechanisms and microstructure even more difficult. Suresh argued in [16]
that the post-overload crack propagation behavior is influenced by micro-
structure as well as crack tip/grain boundary interactions. In the same work
it is reported that surface roughness may lead to sustained crack growth re-
tardation following the application of overloads by reducing the effective
stress intensity range. However limited data can be found in the open lit-
erature on the post-overload fatigue crack response of the same alloy with
variation in microstructure.
This work will first discussed the influence of variations in micro-
structure resulting from sheet and plate product form, as well as from
variations in purity composition (Fe-Si content), on the long crack growth
behaviour of the conventional 2024 T351 aluminum alloy is presented.
The experimental results provide evidence that the effect of microstruc-
tural features (e.g. grain size and morphology) can influence the medium
ǻȀ fatigue crack propagation, which is of high technological interest for
damage tolerant wrought aluminium alloys design.
2.1 Material
The material used in the experimental investigation was aluminum alloy

2024 T351 in plate and sheet product configuration with a thickness of 15
and 3.2 mm respectively. Both plate and sheet materials were tested in
conventional and high purity (HP) composition, which refers to the 2024
material with reduced Fe-rich and Mg2Si phases. The volume fractions of
the respective phases are given in Table 1 for HP and reference material.
In Figure 1 micrographs of the grain structure of the materials used in the
present investigation are shown [6]. The grain morphology obtained for the
plate and sheet product is the result of the different degree of rolling proc-
ess in order to achieve the required thickness (15mm and 3.2 mm respec-
tively). As a result a different grain microstructure was obtained which
corresponds to larger, elongated grains for the plate materials compared to
the fine equiaxed grains in the sheet microstructure.
(a) AA2024 (b) AA2024 (c) 2024HP

Fig. 1 Micrographs showing grain morphology in the materials investigated [6].
Grain size measurements in terms of equivalent circle diameter in

planes L-ST, LT-ST and L-LT are displayed in Table 2. The average grain
size in the plate material was noticeably larger in all-3 planar dimensions
measured exhibiting a coarse and elongated microstructure compared to
the finer equiaxed grain structure in the sheet material. The largest grain
sizes were measured in the plate with high purity composition, which for
the case of the L-LT planar dimension reaches 700µm. The respective
value for the sheet high purity alloy was 30 µm.
Table 1 Calculated volume fractions of Fe-rich and Mg2Si phases in reference and
HP material [6]
Phase 2024 plate HP plate 2024 sheet HP sheet
Fe-rich 1 0.5 1.26 0.65
Mg2Si 0.16 0.06 0.22 0.13
balance 1.16 0.56 1.48 0.78
Table 2 Average planar grain size measured in the four alloys [6].
Plane direction and circle equivalent diameter (µm)
L-ST LT-ST L-LT average
2024HP sheet 18,3 18,2 28,5 21,2
2024 plate 49,4 48,3 101,1 62,3
2024 HP plate 175,0 96,0 349,0 180,3
2024 sheet 97,0 62,0 707,0 162,0
2.2 Fatigue testing, results and fractographic observations
The fatigue tests carried out were constant amplitude fatigue crack growth
tests as well as constant amplitude fatigue crack growth experiments in-
cluding single and periodic tensile overloads. For the fatigue crack propa-
gation experiments center cracked specimens of 3.2 mm thickness were
machined according to ASTM E647 standard [17]. To avoid thickness ef-
fects in the investigation specimens were cut from the mid-plane of the
plate product and were machined to the required thickness of 3.2 mm. The
fatigue crack growth experiments performed are summarized in Table 3.
Table 3 Summary of fatigue crack growth tests.
Constant Amplitude Tests
Material Maximum Frequency Stress Ratio Overl.ratio
stress ımax (Hz) R=ımin/ ımax ıov/ ımax
(MPa)
2024HP sheet 130 25 0.1 1.6
2024 plate 130 25 0.1 1.6
2024 HP plate 130 25 0.1 1.6
2024 sheet 130 25 0.1 1.6
Tests with single overload
Material Stress Ratio Overl.ratio Crack length at overload
R=ımin/ ımax ıov/ ımax (mm)
2024HP sheet 0.1 1.6 8.5,11(2 tests)
2024 plate 0.1 1.6 9,11(2 tests)
2024 HP plate 0.1 1.6 10,12(2 tests)
2024 sheet 0.1 1.6 7.5
Tests with periodic overloads
Material Stress Ratio Overl.ratio Crack length at overload
R=ımin/ ımax ıov/ ımax (mm)
2024HP sheet 0.1 1.6 6.5, 8, 10
2024 sheet 0.1 1.6 6.5, 7.5, 10
Fig. 2 Constant amplitude fatigue crack growth results.
In Fig. 2 constant amplitude fatigue crack propagation curves of the ma-

terials tested are displayed in terms of half crack length versus stress inten-
sity factor range. The higher crack growth resistance for specimens taken
from plate product is reflected through the lower crack growth rates com-
pared to the sheet material in the medium ¨Ȁ region (10-30 MPam1/2). For
both rolling products the material with high purity composition exhibited
superior crack growth behaviour in the mentioned ǻȀ region.
Optical micrographs of the fractured specimen surfaces with plate mi-
crostructure revealed increased roughness and discontinuities along the
crack path direction (Fig.3). The surface roughness was evaluated by
means of the surface profile parameter RL according to the ASM guide-
lines [18]. RL values measured for plate specimens were higher compared
to the sheet material. The RL values measured at specific segments taken at
the same crack path location for plate and sheet 2024 material were
RLPLATE =1.18 and RLSHEET =1.07 respectively (Fig. 4).
(a) Sheet (b) Plate material

Fig. 3 Micrographs showing segments of the free edge of fractured surfaces at the
same location of crack propagation with different degree of surface roughness.
Fig. 4 Profile roughness parameter for sheet and plate material.
The obtained response under constant stress amplitude may be correlated

to pronounced crack closure effects of specimens with plate microstructure
due to increased surface roughness, which produces pronounced crack sur-
face contact. The above observation is in agreement with results reported
in [14-15,19]. The increased surface roughness along the crack path direc-
tion is promoted by the coarse and elongated grain morphology, as a result
of the different rolling process involved. Macroscopic fractographic ex-
amination revealed a fractured surface region perpendicular to the load fol-
lowed by shear fracture (Fig. 5). For all materials tested double shear lips
are evident along the fracture path.
Fig. 5 Fractured surfaces macrographs of different materials subjected to constant

amplitude fatigue tests.
As shown in Fig. 6 in the case of the plate product the fractured surface
region which is perpendicular to the load is smaller and the developed
shear lips are wider in comparison to the sheet material. On the fracture
surfaces of the plate materials pronounced black debris could be observed,
which according to [20] could be evidence of increased lower and upper
crack surfaces contact during fatigue crack growth. By careful examination
of fractured surfaces in Fig. 6 it can be observed that crack surface contact
marks are more pronounced on the shear lips of plate material specimens.
The described mechanism can lead to increased crack closure.
In order to assess the fatigue crack retardation response following ap-
plication of overloads, constant stress amplitude fatigue crack growth ex-
periments including single and periodic tensile overloads were carried out.
The testing parameters were the same as in the previous tests. The applied
overload ratio was ıov/ımax= 1.6, while application of overload was per-
formed at similar ǻȀ levels (at approximately 20 MPam1/2). From the ex-
perimental crack growth curves obtained (Figs. 6-7) it was observed that
crack growth rates measured at comparable ¨Ȁ levels in the post – over-
load stage were lower in the specimens with plate microstructure. The post
overload crack growth behaviour observed is consistent with the results
obtained in the constant amplitude crack propagation curves showing an
increased crack growth resistance for the plate microstructures both in the
prior and post overload stage. Figures 8, 9 and 10 are showing a sustained
retardation of the alloy with high purity composition for the case of single
and periodic overloads. Micrographs of fracture surfaces of the specimens
subjected to tensile overloads revealed similar characteristics to the con-
stant amplitude tests in terms of shear lip development and surface
(a) Crack length (b) Crack growth rate

Fig. 6 Crack length and fatigue crack growth rates after tensile overload for sheet
and plate material.

and plate material.

and plate material.

Fig. 9 Crack length and fatigue crack growth rates after tensile overload for con-
ventional and HP sheet material.
roughness. In the case of single and multiple overloads the load interaction
phenomena present lead to significant plasticity effects. Thus, crack
growth mechanisms become more complex and the results obtained cannot
be discussed solely under the viewpoint of pronounced crack closure ef-
fects due to fracture surface characteristics.

Fig. 10 Crack length and fatigue crack growth rates after periodic tensile overload
for conventional and HP sheet material
3. Effect of existing corrosion damage on Fatigue Crack Growth
It is without doubt that the Aloha Airlines accident in 1988 showed the
global aviation industry the dangers of the structural degradation of aging
components. In the years that followed intensive investigations have
shown that the occurrence of corrosion appreciably degrades the mechani-
cal performance of aluminum alloys [e.g.21].
Corrosion damage of aluminum alloys has been classically attributed to
the complex process of oxidation. In recent years experimental efforts on a
series of aircraft alloys have provided evidence that corrosion is also the
cause for diffusion controlled material hydrogen embrittlement [21-29].
This hydrogen embrittlement phenomenon is reflected into an appreciable
reduction of tensile ductility of the corroded material areas [21-22,28–29].
The hydrogen embrittlement process is explained in [30] through a hydro-
gen induced local microplasticity. Both processes, corrosion and hydrogen
embrittlement, are diffusion-controlled, hence the degradation of the mate-
rial’s mechanical performance is expected to occur in local scale. However
at present there are neither experimental data nor an established methodol-
ogy for assessing the values of the local material properties of the corroded
areas of the material. Notice that experimental investigations on the inter-
action of corrosion and fatigue usually refer to fatigue and fatigue crack
growth tests which are performed in a certain corrosive environment [31-
32] and not to tests performed on pre-corroded material whereby, the latter
represents a different and, for a series of practical cases in older airplanes,
more relevant situation.
Structural degradation due to existing corrosion has been considered as a
load capacity decrease of the structural component [33]. Furthermore, it
has been recognized that the occurrence of corrosion facilitates the onset of
fatigue cracks [34-35] and, hence causes the decrease in fatigue life due to
corrosion-pitting damage [34], as well as multi-site damage scenarios from
corrosion pits [35]. In Figure 11 the S-N curves of uncorroded and pre-
corroded 2024 T351 aluminum alloy specimens are displayed [36]. The re-
sults reveal the essential degradation of fatigue life in the corroded cou-
pons as a result of premature onset of crack initiation due to corrosion pits.
The degradation is reflected in a significant decrease in the material’s fa-
tigue limit of the order of 28.5 % compared to the uncorroded material.
Fig.11 S-N curves of uncorroded and pre-corroded 2024 T351 aluminum alloy
specimens
On the other hand, the effect of prior corrosion damage on the fa-
tigue crack growth behaviour of aircraft structures is presently disregarded
and a few experimental data exist correlating the effect of prior corrosion
on the fatigue crack growth rate [37].
The fatigue and damage tolerance behaviour of pre-corroded 2024 T351
aluminum alloy specimens is presented and discussed under the viewpoint
of a synergetic effect of corrosion and corrosion-induced hydrogen embrit-
tlement. The results demonstrate the essential effect of existing corrosion
on the fatigue and damage tolerance behaviour of the 2024 alloy, as well
as the need to account for the effect of corrosion on the mechanical proper-
ties in fatigue and damage tolerance analyses of the corroded areas of a

structure.
3.1 Experimental procedure, testing and results
The experiments were conducted on bare 2024 T351 aluminium alloy of

1.6 mm thickness. The performed experiments included fatigue crack
growth tests for different stress ratio R on corroded and uncorroded mate-
rial. An overview of the performed tests is given in Table 4. ȉhe selected
fatigue stress ratios were R=ımin/ımax 0.01, 0.1, 0.5 țĮȚ 0.7 and the test fre-
quency 20 Hz. In order to evaluate the R-effect the stress range ǻı was
kept constant at a ımax between 101 and 180 MPa. Machining of the
specimens was made according to ASTM E 647-93 specification. All
specimens were cut in longitudinal (L) orientation relative to the rolling di-
rection. The experimental investigation included uncorroded and pre-
corroded specimens in exfoliation corrosion environment for 36 hours ac-
cording to ASTM standard G34-90 [38]. Exfoliation corrosion environ-
ment has been selected as the environment that satisfactory simulates the
effect of long duration outdoor corrosion exposure on the material’s me-
chanical properties [34]. After exposure, the specimens were cleaned ac-
cording to ASTM G34-90. During all fatigue crack growth tests crack
growth was recorded via a DC Potential Drop measurement method.
Table 4 Summary of fatigue crack growth tests.

Corrosion ex- Stress Ratio Frequency Max. stress ımax No.
posure R=ımin/ ımax (Ǿz) (MPa) tests
none 0.01 20 109 2
EXCȅ 36 hrs 0.01 20 109 2
none 0.1 20 101 2
EXCȅ 36 hrs 0.1 20 101 2
none 0.5 20 180.4 2
EXCȅ 36 hrs 0.5 20 180.4 1
none 0.7 20 176.1 2
EXCȅ 36 hrs 0.7 20 176.1 2

The fatigue crack growth results are summarized in Tables 5-7 and Figs.
12-16. In Table 5 the values of ǻȀ and Ȁmax referring to the critical crack
length are given. Calculation of ǻȀ and Ȁmax was made with implementa-
tion of the equation [17]:
K EV SD (1)
In the interpretation of results the effect of corrosion on the correction fac-
tor ȕ in Eqn. (1) was not taken into account. In Table 6 the number of cy-
cles corresponding to specimen failure are displayed together with the
critical crack length Įcr defined at a time step that corresponds to 3 seconds
prior to specimen failure. The critical crack length was defined in this
manner because of the unfeasibility due to the data recording frequency of
measuring the actual value at fracture. The critical crack length corre-
sponds to the last data point of the curves in Figs. 12a-15a.
Table 5 ǻȀ țĮȚ Kmax values at critical crack length.

Stress int. factor Max. stress int.
Corrosion Stress ratio range ǻK (at Įcr ) factor Kmax (at Įcr)
exposure R=ımin/ımax
[ MPa m ] [ MPa m ]
none 0.01 29.74 30.04
ǼXCO 0.01 22.87 23.09
none 0.1 32.54 36.16
ǼXCO 0.1 25.28 28.09
none 0.5 38.74 77.48
ǼXCO 0.5 16.94 33.88
none 0.7 11.70 40.32
ǼXCO 0.7 11.25 37.54
Table 6 Fatigue crack growth data of corroded and uncorroded specimens.

Fatigue life Critical crack
Corrosion Stress ratio Max. stress
Nf (cycles- length Įcr (mm-
exposure R=ımin/ımax (MPa)
(mean values) (mean values)
46640/47460 20.02/16.46
None 0.01 109
(47050) (18.24)
42700
ǼXCO 36 hrs 0.01 109 12.59 (12.59)
(42700)
81620/69100 24.8/23.2
None 0.1 101
(75360) (24.0)
58640/51960 15.59/21.31
ǼXCO 36 hrs 0.1 101
(55300) (18.45)
43960/49820 19.4/35.73
None 0.5 180.4
(46890) (27.56)
30560/29560 9.72/10.92
ǼXCO 36 hrs 0.5 180.4
(30060) (10.32)
213540 14.17
None 0.7 176.1
(213540) (14.17)
139880/121260 11.75/13.67
ǼXCO 36 hrs 0.7 176.1
(130570) (12.71)
Table 7 Paris coefficients of corroded and uncorroded material.

Corrosion Stress ratio Mean
n C Mean value
exposure R=ımin/ımax value
-11
0.01 3.09 7.38*10
3.13 4.69*10-11
3.16 2.0*10-11
0.1 3.35 5.56*10-11
3.21 9.38*10-11
3.07 1.32*10-10
none
0.5 3.0 1.94*10-10
3.12 1.60*10-10
3.23 1.26*10-10
0.7 3.6 9.25*10-11
3.6 9.25*10-11
-
0.01 - -
2.85 4.88*10-11
2.85 4.88*10-11
0.1 3.16 1.62*10-10
3.11 1.5*10-10
ǼXCO 36 3.05 1.38*10-10
hrs 0.5 3.04 2.38*10-10
3.21 4.37*10-10
3.38 1.99*10-10
0.7 3.16 5.03*10-10
3.47 2.9*10-10
3.78 7.61*10-10
In Figs. 12-16 the fatigue crack growth curves together with the crack
growth rate measurements versus stress intensity factor range ǻȀ are dis-
played for the corroded and un-corroded specimens at different stress ra-
tios.

Fig. 12 Crack length and fatigue crack growth rates of corroded and un-corroded
specimens at R=0.01.

Fig. 13 Crack length and fatigue crack growth rates of corroded and uncorroded
specimens at R=0.1.

Fig. 14 Crack length and fatigue crack growth rates of corroded and uncorroded
specimens at R=0. 5.

Fig. 15 Crack length and fatigue crack growth rates of corroded and un-
corroded specimens at R=0.7.
(a) Uncorroded (b) Corroded

Fig. 16 Fatigue crack growth rates of specimens at R=0.01, 0.1, 0.5, 0.7.
From the obtained results the deteriorating effect of corrosion on fatigue

life of pre-corroded specimens is revealed for all stress ratios examined.
Fatigue life reduction, corresponding to fatigue crack propagation cycles,
which as expected is promoted by the existence of localized corrosion sites
acting as stress concentration locations, is ranging from 12% (R=0.01) to
38.8% (R=0.7). The effect of corrosion on fatigue crack propagation is
more pronounced for small values of R. The observed behaviour can be
explained by considering the mechanisms of partial embrittlement of the
specimens as a result of corrosion exposure. For small values of ǻȀ dam-
age accumulation mechanisms seem not to be influenced to the same de-
gree by the partially embrittled material. At higher tensile stresses (higher
ǻȀ values) embrittlement may have a more pronounced effect in the abil-
ity of the material to deform plastically ahead of the crack tip.
By examining Table 6 the effect of corrosion damage on the critical

crack length to failure is illustrated. The critical crack length of the cor-
roded specimens is reduced by a magnitude of 45% compared to the re-
spective value of the uncorroded specimens, while the stress intensity fac-
tor values corresponding to Įcr is reduced by 34% with regard to the
uncorroded material. The stress intensity factor reduction is attributed to
the reduction of the critical crack length due to fracture toughness degrada-
tion of the corroded alloy.
From the experimental findings obtained in Figs. 12-16 it can be con-
cluded that the crack propagation resistance at high ǻȀ values suffers a
significant degradation compared to the medium ǻȀ region where corro-
sion attack does not have the same impact. The latter observation is sup-
ported by the calculated differences in the coefficients of the Paris equa-
tion. The differences in the coefficients for the corroded and uncorroded
specimens, derived from the da/dN–ǻK plots for stress intensity ranging
from 10– 18 MPa m1/2 are compared in Table 7. The coefficients do not
differ appreciably. This is consistent with the results in [35]. At stage III
crack growth (high ǻȀ), the calculated crack growth rates are high and
cannot be described by the Paris law. At this stage Paris equation cannot
assess the influence of corrosion on fatigue crack growth.
The obtained decrease in fatigue life of corroded specimens is attributed to
the reduction of the critical crack length at failure. The latter is dependent
on the material’s fracture toughness. These parameters are influenced by
the partially embrittled material due to corrosion. The fatigue crack growth
behaviour described above is not easy to explain by means of the classical
interpretation of corrosion. Fractographic analyses performed in [29] on
1.6 mm thick specimen subjected to exfoliation corrosion exposure for 24
h confirm the embrittlement of the material during the corrosion exposure.
The fracture surface appearance varies from the specimen surface to the
specimen interior. At the specimen surface the fracture is intergranular as
shown in Fig. 17. Immediately below there is a zone of quasi-cleavage
fracture and further below the fracture turns to entirely ductile (dimples) as
shown in Fig. 17. The results underline the need to consider the fatigue
crack growth behaviour of corroded 2024 as the result of the synergetic ef-
fect of corrosion and corrosion induced hydrogen embrittlement.
Specimen surface
Intergranular
Quasicleavage
D til f
Fig. 17 Local embrittled areas in corroded 2024 specimens.
The fatigue crack growth behaviour of 2024 T351 aircraft aluminium alloy
(i) with variation in microstructural characteristics and (ii) under the influ-
ence of existing corrosion damage has been investigated.
The experimental findings concerning the microstructural effects on fa-
tigue crack growth revealed the following:
(1) The effect of microstructural features (e.g. grain size and morphology),
as well as variations in purity composition can influence fatigue crack
propagation in the medium ǻȀ region, which is of high technological in-
terest for damage tolerance design.
(2) Plate microstructure with elongated grain morphology induced by the
rolling process exhibited higher crack growth resistance than the micro-
structure of the sheet material with smaller equiaxed grains. This behav-
iour could be attributed to pronounced roughness induced crack closure ef-
fects in the plate microstructure.
(3) The experimental results showed that the plate microstructure with
high purity composition provides better damage tolerance characteristics
compared to the conventional sheet microstructure. With regard to the ef-
fect of existing corrosion on the fatigue crack growth the following con-
clusions can be made:
(4) The synergetic effect of corrosion and corrosion-induced hydrogen
embrittlement resulted in an appreciable decrease in fatigue resistance and
damage tolerance of the corroded material.
(5) The experimental results have demonstrated the need to account for the
influence of pre-existing corrosion on the material’s properties for the reli-
able fatigue and damage tolerance analyses of components involving cor-
roded areas.
References
[1] Suresh S. Fatigue of Materials, second ed., Cambridge University Press,

1998, ISBN 0521578477, 700 pages
[2] Sih GC. and Chao CK. Failure initiation of unnotched specimens subjected
to monotonic and cyclic loading, Theor. Appl. Fract. Mech. 2(1), (1984)
67-73
[3] Sih GC. and Moyer ET. Jr., Path dependent nature of fatigue crack growth,
Inter. J. Eng. Fract. Mech. 7, (1983), 269-280
[4] Sih GC.and Chao CK. Influence of load amplitude and uniaxial tensile pre-
ties on fatigue crack frowth, Theor. Appl. Fract. Mech. 2(3), (1984) 247-
257
[5] Elber W. Damage Tolerance in Aircraft Structures, ASTM Spec.
Tech.Publ.486, (1981) 230
[6] Investigation on Damage Tolerance Behaviour of Aluminum Alloys,
IDA,European Project, Project No: GRD2001-40120, Final Report (2005)
[7] Kermanidis AlTh. Effect of Corrosion on the Structural Integrity of Light-
weight Aircraft Structures, PhD Thesis, University of Patras, Department of
Mechanical and Aeronautical Engineering, Patras, Greece, (2003)
[8] Garrett GG. and Knott JF. Crystallographic Fatigue Crack Growth in Alu-
minum Alloys, Acta Metallurgica, 25, (1975) 841-849
[9] Suresh S. and Ritchie RO. Metall. Trans. A, 13, (1982) 1627
[10] Zaiken E. and Ritchie RO. Effects of Microstructure on Fatigue Crack
Propagation and Crack Closure Behavior in Aluminum Alloy 7150 Materi-
als Science and Engineering, 70 (1985) 151-160.
[11] E. Zaiken E. and Ritchie RO. On the Location of Crack Closure and the
Threshold Condition for Fatigue Crack Growth, Scripta Metallurgica, 18
(1984) 847-850
[12] Tzou JL.Suresh S. Ritchie RO. Fatigue Crack Propagation in Oil Environ-
ments-. 1 Crack Growth Behaviour in Silicone and Parafin Oils, Acta Met-
all. 33 (1) (1985) 105-116
[13] Minakawa K. and McEvily AJ. On Crack Closure in the near Threshold
Region, Scripta Metallurgica, 15 (1981) 633-636
[14] Venkateswara KT. Rao RJ. Bucci KV. Jata. and Ritchie RO. A comparison
of fatigue crack propagation behavior in sheet and plate aluminum-lithium
alloys, Materials Science and Engineering, A141 (1991) 39-48
[15] Henk F. and Jong De. Thickness Direction Inhomogeneity of Mechanical
Properties and Fracture Toughness as observed in Aluminum 7075 T651
Plate Material, Enginneering Fracture Mechnanics 13 (1980) 175-192
[16] Suresh S. Micromechanisms of fatigue Crack Growth Retardation Follow-
ing Overloads, Engineering Fracture Mechanics 18. (3) (1983) 577-593,
[17] ASTM E647-93, Standard Test Method for Measurement of Fatigue Crack
Growth Tests, Annual Book of ASTM Standards, Metals –Test Methods and
Analytical Procedures, Section 3, Vol. 03.01, Metals-Mechanical Testing;
Elevated Low-Temperature Tests; Metallography, (1994)
[18] ASM Handbook, Volume1 2, Fractography, The Materials Information

Society, (2000)
[19] Pao PS. Cooley LA. Imam MA. And Yoder GR. Fatigue Crack Growth in
2090 Al-Li Alloy, Scripta Metallurgica, 23, (1989) 1455-1460
[20] Schijve J. Fatigue Cracks, Plasticity Effects and Crack Closure, Four Lec-
tures on Fracture Mechanics, 1l. No. 1-L, (1978)
[21] Pantelakis SpG. Vassilas NI. and Daglaras PG. Effects of corrosive envi-
ronment on the mechanical behaviour of the advanced Al–Li alloys 2091
and 8090 and the conventional aerospace alloy 2024, Metall, 47 (1993)
135–141.
[22] Pantelakis SpG. DaglarasPG. and Apostolopoulos ChAlk. Tensile and en-
ergy density properties of 2024, 6013, 8090 and 2091 aircraft aluminum al-
loy after corrosion exposure, J. Theor. Appl. Fract. Mech. 33 (2000) 117–
134
[23] Papanikos P. and Kermanidis AlTh. Effect of corrosioninduced hydrogen
embrittlement on the fracture toughness of 2024-T3 aluminum alloy, in:
Proceedings of the 4th International Conference on New Challenges in Me-
somechanics, Aalborg University, Denmark, (2002)
[24] Smiyan OD. Coval MV. and Melekhov RK. Local corrosion damage of
aluminum alloys, Soviet Mater. Sci. 19 (1983) 422
[25] Tuck CDS. Evidence for the formation of magnesium hydride on the gra
boundaries of Al–Zn–Mg alloys during their exposure to water vapour, in:
Proceedings of the Third International Conference of Hydrogen on the Be-
havior of Materials, Jackson, Wyoming, (1980) 503– 511
[26] Scamans GM. R. Alani R. and Swann PR. Pre-exposure embrittlement and-
stress corrosion failure in Al–Zn–Mg alloys, Corros. Sci. 16 (1976) 443
[13] E. Charitidou E. Papapolymerou G, Haidemenopoulos GN. Hassiotis N.
and Bontozoglou HV. Characterization of trapped hydrogen in exfoliation
corroded aluminium alloy 2024, Script. Mater. 41 (1999) 1327
[27] Haidemenopoulos GN, Hassiotis N, Papapolymerou G. and Bontozoglou
HV. Hydrogen absorption into aluminum alloy 2024-T3 during exfoliation
and alternate immersion testing, Corrosion 54 (1998) 73
[28] Petroyiannis PV. Kermanidis AlTh. Papanikos P. and Pantelakis SpG, Cor-
rosion-induced hydrogen embrittlement of 2024 and 6013 aluminium al-
loys, Special issue: Al.Th. Kermanidis et al. / Theoretical and Applied
Fracture Mechanics 43 (2005) 121–132 131 mesofracture mechanics: cur-
rent approaches to material damage at different size and time scales, Theor.
Appl. Fract. Mech. 41 (1–3) (2004) 173
[29] P.V. Petroyiannis PV. Kermanidis AlTh. E. Kamoutsi E. Pantelakis SpG.
Bontozoglou HV. and Haidemenopoulos GN. Evidence on the corrosion-
induced hydrogen embrittlement of the 2024 aluminum alloy, Fatigue
Fract. Eng. Mater. Struct., 28 (2005) 565-574
[30] Birnbaum HK. and Sofronis P. Hydrogen-enhanced localized plasticity-a
mechanism for hydrogen-related fracture, Mat. Sci. Eng. A 176 (1994) 191
[31] Ruiz J. and Elices M. The role of environmental exposure in the fatigue be-
haviour of an aluminium alloy, Corros. Sci. 39 (12) (1997) 2117
[32] Lin CK. and Yang ST.Corrosion fatigue behavior of 7050 aluminum alloys
in different tempers, Eng. Fract. Mech. 59 (6) (1998) 779
[33] Inman ME. Kelly RG. S.A. Willard SA. and Piascik RS. in: Proceedings of
the FAA-NASA Symposium on the Continued Airworthiness of Aircraft
Structures, Virginia, USA, (1997) 129
[34] Zamber JE. and Hillberry BM. Probabilistic approach to predicting fatigue
lives of corroded 2024-T3, AIAA J. 37 (10) (1999) 1311.
[35] Bray GH. Bucci RJ. Colvin EL. and Kulak M. Effects of prior corrosion on
the S/N fatigue performance of aluminium sheet alloys 2024-T3 and 2524-
T3, in: W.A. Van der Sluys, R.S. Piascik, R. Zawierucha (Eds.), Effects of
the Environment on the Initiation of Crack Growth, ASTM STP 1298,
American Society for Testing and Materials, (1997) 89
[36] Kermanidis AlTh. Petroyiannis PV. and Pantelakis SpG. Fatigue and dam-
age tolerance behaviour of corroded 2024 T351 aircraft aluminum alloy,
Theoretical and Applied Fracture Mechanics 43 (2005) 121–132
[37] Chubb JP. Morad TA. Hockenhull BS. and Bristow JW. The effect of exfo-
liation corrosion on the fatigue behaviour of structural aluminium alloys,
Structural Integrity of Ageing Airplanes, Springer-Verlag, Berlin, (1991)
87
[38] ASTM G34-90, Standard Test Method for Exfoliation Corrosion Suscept-
bility in 2xxx and 7xxx Series Aluminum Alloys, (EXCO Test), Annual
Book of ASTM Standards, Metals –Test Methods and Analytical Proce-
dures, Section 3, Vol. 03.02, Wear and Erosion; Metal Corrosion, (1995)
Author Index
Barter SA, 65 Pantelakis SpG, 345

de Castro PMST, 85 Peng D, 23
Dini D, 47 Pitt S, 23
Farahmand B, 1 Richter-Trummer V, 85
Höppel HW, 327 Rodopoulos CA, 113
Jones R, 23, 65 Sih GC, 181, 209, 249
Kermanidis ATh, 345 Tang XS, 209, 249
Korsunsky AK, 47 Valsan M, 133
Mannan SL, 133 Walsh MJ, 47
Molent L, 65 Wang ZG, 275
Moreira PMGP, 85 Zhang GP, 275
Mughrabi H, 327
367
Subject Index
aluminum long,
17050-T7451, surface,
2014-T6, termination,
2024-T3, transgranular
2219-T87, crack growth
6061-T62, block,
6061-T6, fatigue,
6082-T6, growth,
7010-T73651, non-similitude,
7075-T6, rate,
7150-T651, stable/unstable,
pure (super), crystallographic position,
bridge damage
cable-stayed, fatigue,
Runyang, dislocation density,
suspension, surface bulk,
cable ductile iron,
construction, elastic modulus,
loosen/tighten, shear,
mechanical property, Young’s,
compact tension, energy
compliance, 2024-T351,
corroded/uncorroded, absorbed,
crack released,
corner, total,
creep-fatigue, extrusion,
density, fatigue
initiation, creep,
intergranular, damage map,
length, low cycle,
369
370 Subject Index
small specimen, invariant,

SN-curve, length,
striation, macro/micro
thermal, shifting,
film, size/time,
foil, triple,
frequency, sigmoidal,
fracture, slip,
mechanics, slope,
surface, steel
weld, 9Cr-1Mo,
fracture thoughness, 316L(SS),
grain inconel,
morphology, low carbon,
size, SAE 52100,
heterogeneity, strain energy density
interface, factor,
invariant form, function,
loading, strain rate,
const. amlitude, stress intensity factor,
indentor tip, applied,
sequence, maximum,
symmetric effect, range,
traffic/no-traffic, stress singularity,
macrocrack, double,
macromechanical, strong/weak,
magnesium AZ91, stress type
microcrack, amplitude,
microhardness, applied,
microstructural, max./min.,
nimonic PE106, mean,
nanochemical, restraining,
nano-tube, ultimate,
necking, yield,
non-destructive test, temperature elevated,
overload, time (hold),
oxidation, titanium ally
plane strain/stress, Ti-2.5Cu,
plastic zone, Ti-6Al-4V,
pore, threshold (fatigue),
region I, II, III, transition,
scale virtual testing,
dual, wave form,
Subject Index 371
simple loading, wire,

trapezoidal,
weld, Y-intercept
Mechanics
Series Editor: G.M.L. Gladwell
Aims and Scope of the Series
The fundamental questions arising in mechanics are: Why?, How?, and How much? The aim of
this series is to provide lucid accounts written by authoritative researchers giving vision and insight
in answering these questions on the subject of mechanics as it relates to solids. The scope of the
series covers the entire spectrum of solid mechanics. Thus it includes the foundation of mechanics;
variational formulations; computational mechanics; statics, kinematics and dynamics of rigid and
elastic bodies; vibrations of solids and structures; dynamical systems and chaos; the theories of
elasticity, plasticity and viscoelasticity; composite materials; rods, beams, shells and membranes;
structural control and stability; soils, rocks and geomechanics; fracture; tribology; experimental
mechanics; biomechanics and machine design.
1. R.T. Haftka, Z. Gürdal and M.P. Kamat: Elements of Structural Optimization. 2nd rev.ed., 1990
ISBN 0-7923-0608-2
2. J.J. Kalker: Three-Dimensional Elastic Bodies in Rolling Contact. 1990 ISBN 0-7923-0712-7
3. P. Karasudhi: Foundations of Solid Mechanics. 1991 ISBN 0-7923-0772-0
4. Not published
5. Not published.
6. J.F. Doyle: Static and Dynamic Analysis of Structures. With an Emphasis on Mechanics and
Computer Matrix Methods. 1991 ISBN 0-7923-1124-8; Pb 0-7923-1208-2
7. O.O. Ochoa and J.N. Reddy: Finite Element Analysis of Composite Laminates.
ISBN 0-7923-1125-6
8. M.H. Aliabadi and D.P. Rooke: Numerical Fracture Mechanics. ISBN 0-7923-1175-2
9. J. Angeles and C.S. López-Cajún: Optimization of Cam Mechanisms. 1991
ISBN 0-7923-1355-0
10. D.E. Grierson, A. Franchi and P. Riva (eds.): Progress in Structural Engineering. 1991
ISBN 0-7923-1396-8
11. R.T. Haftka and Z. Gürdal: Elements of Structural Optimization. 3rd rev. and exp. ed. 1992
ISBN 0-7923-1504-9; Pb 0-7923-1505-7
12. J.R. Barber: Elasticity. 1992 ISBN 0-7923-1609-6; Pb 0-7923-1610-X
13. H.S. Tzou and G.L. Anderson (eds.): Intelligent Structural Systems. 1992
ISBN 0-7923-1920-6
14. E.E. Gdoutos: Fracture Mechanics. An Introduction. 1993 ISBN 0-7923-1932-X
15. J.P. Ward: Solid Mechanics. An Introduction. 1992 ISBN 0-7923-1949-4
16. M. Farshad: Design and Analysis of Shell Structures. 1992 ISBN 0-7923-1950-8
17. H.S. Tzou and T. Fukuda (eds.): Precision Sensors, Actuators and Systems. 1992
ISBN 0-7923-2015-8
18. J.R. Vinson: The Behavior of Shells Composed of Isotropic and Composite Materials. 1993
ISBN 0-7923-2113-8
19. H.S. Tzou: Piezoelectric Shells. Distributed Sensing and Control of Continua. 1993
ISBN 0-7923-2186-3
20. W. Schiehlen (ed.): Advanced Multibody System Dynamics. Simulation and Software Tools.
1993 ISBN 0-7923-2192-8
21. C.-W. Lee: Vibration Analysis of Rotors. 1993 ISBN 0-7923-2300-9
22. D.R. Smith: An Introduction to Continuum Mechanics. 1993 ISBN 0-7923-2454-4
23. G.M.L. Gladwell: Inverse Problems in Scattering. An Introduction. 1993 ISBN 0-7923-2478-1
Mechanics
24. G. Prathap: The Finite Element Method in Structural Mechanics. 1993 ISBN 0-7923-2492-7
25. J. Herskovits (ed.): Advances in Structural Optimization. 1995 ISBN 0-7923-2510-9
26. M.A. González-Palacios and J. Angeles: Cam Synthesis. 1993 ISBN 0-7923-2536-2
27. W.S. Hall: The Boundary Element Method. 1993 ISBN 0-7923-2580-X
28. J. Angeles, G. Hommel and P. Kovács (eds.): Computational Kinematics. 1993
ISBN 0-7923-2585-0
29. A. Curnier: Computational Methods in Solid Mechanics. 1994 ISBN 0-7923-2761-6
30. D.A. Hills and D. Nowell: Mechanics of Fretting Fatigue. 1994 ISBN 0-7923-2866-3
31. B. Tabarrok and F.P.J. Rimrott: Variational Methods and Complementary Formulations in
Dynamics. 1994 ISBN 0-7923-2923-6
32. E.H. Dowell (ed.), E.F. Crawley, H.C. Curtiss Jr., D.A. Peters, R. H. Scanlan and F. Sisto: A
Modern Course in Aeroelasticity. Third Revised and Enlarged Edition. 1995
ISBN 0-7923-2788-8; Pb: 0-7923-2789-6
33. A. Preumont: Random Vibration and Spectral Analysis. 1994 ISBN 0-7923-3036-6
34. J.N. Reddy (ed.): Mechanics of Composite Materials. Selected works of Nicholas J. Pagano.
1994 ISBN 0-7923-3041-2
35. A.P.S. Selvadurai (ed.): Mechanics of Poroelastic Media. 1996 ISBN 0-7923-3329-2
36. Z. Mróz, D. Weichert, S. Dorosz (eds.): Inelastic Behaviour of Structures under Variable
Loads. 1995 ISBN 0-7923-3397-7
37. R. Pyrz (ed.): IUTAM Symposium on Microstructure-Property Interactions in Composite
Materials. Proceedings of the IUTAM Symposium held in Aalborg, Denmark. 1995
ISBN 0-7923-3427-2
38. M.I. Friswell and J.E. Mottershead: Finite Element Model Updating in Structural Dynamics.
1995 ISBN 0-7923-3431-0
39. D.F. Parker and A.H. England (eds.): IUTAM Symposium on Anisotropy, Inhomogeneity and
Nonlinearity in Solid Mechanics. Proceedings of the IUTAM Symposium held in Nottingham,
U.K. 1995 ISBN 0-7923-3594-5
40. J.-P. Merlet and B. Ravani (eds.): Computational Kinematics ’95. 1995 ISBN 0-7923-3673-9
41. L.P. Lebedev, I.I. Vorovich and G.M.L. Gladwell: Functional Analysis. Applications in Mech-
anics and Inverse Problems. 1996 ISBN 0-7923-3849-9
42. J. Menčik: Mechanics of Components with Treated or Coated Surfaces. 1996
ISBN 0-7923-3700-X
43. D. Bestle and W. Schiehlen (eds.): IUTAM Symposium on Optimization of Mechanical Systems.
Proceedings of the IUTAM Symposium held in Stuttgart, Germany. 1996
ISBN 0-7923-3830-8
44. D.A. Hills, P.A. Kelly, D.N. Dai and A.M. Korsunsky: Solution of Crack Problems. The
Distributed Dislocation Technique. 1996 ISBN 0-7923-3848-0
45. V.A. Squire, R.J. Hosking, A.D. Kerr and P.J. Langhorne: Moving Loads on Ice Plates. 1996
ISBN 0-7923-3953-3
46. A. Pineau and A. Zaoui (eds.): IUTAM Symposium on Micromechanics of Plasticity and
Damage of Multiphase Materials. Proceedings of the IUTAM Symposium held in Sèvres,
Paris, France. 1996 ISBN 0-7923-4188-0
47. A. Naess and S. Krenk (eds.): IUTAM Symposium on Advances in Nonlinear Stochastic
Mechanics. Proceedings of the IUTAM Symposium held in Trondheim, Norway. 1996
ISBN 0-7923-4193-7
48. D. Ieşan and A. Scalia: Thermoelastic Deformations. 1996 ISBN 0-7923-4230-5
Mechanics
49. J.R. Willis (ed.): IUTAM Symposium on Nonlinear Analysis of Fracture. Proceedings of the
IUTAM Symposium held in Cambridge, U.K. 1997 ISBN 0-7923-4378-6
50. A. Preumont: Vibration Control of Active Structures. An Introduction. 1997
ISBN 0-7923-4392-1
51. G.P. Cherepanov: Methods of Fracture Mechanics: Solid Matter Physics. 1997
ISBN 0-7923-4408-1
52. D.H. van Campen (ed.): IUTAM Symposium on Interaction between Dynamics and Control in
Advanced Mechanical Systems. Proceedings of the IUTAM Symposium held in Eindhoven,
The Netherlands. 1997 ISBN 0-7923-4429-4
53. N.A. Fleck and A.C.F. Cocks (eds.): IUTAM Symposium on Mechanics of Granular and Porous
Materials. Proceedings of the IUTAM Symposium held in Cambridge, U.K. 1997
ISBN 0-7923-4553-3
54. J. Roorda and N.K. Srivastava (eds.): Trends in Structural Mechanics. Theory, Practice, Edu-
cation. 1997 ISBN 0-7923-4603-3
55. Yu.A. Mitropolskii and N. Van Dao: Applied Asymptotic Methods in Nonlinear Oscillations.
1997 ISBN 0-7923-4605-X
56. C. Guedes Soares (ed.): Probabilistic Methods for Structural Design. 1997
ISBN 0-7923-4670-X
57. D. François, A. Pineau and A. Zaoui: Mechanical Behaviour of Materials. Volume I: Elasticity
and Plasticity. 1998 ISBN 0-7923-4894-X
58. D. François, A. Pineau and A. Zaoui: Mechanical Behaviour of Materials. Volume II: Visco-
plasticity, Damage, Fracture and Contact Mechanics. 1998 ISBN 0-7923-4895-8
59. L.T. Tenek and J. Argyris: Finite Element Analysis for Composite Structures. 1998
ISBN 0-7923-4899-0
60. Y.A. Bahei-El-Din and G.J. Dvorak (eds.): IUTAM Symposium on Transformation Problems
in Composite and Active Materials. Proceedings of the IUTAM Symposium held in Cairo,
Egypt. 1998 ISBN 0-7923-5122-3
61. I.G. Goryacheva: Contact Mechanics in Tribology. 1998 ISBN 0-7923-5257-2
62. O.T. Bruhns and E. Stein (eds.): IUTAM Symposium on Micro- and Macrostructural Aspects
of Thermoplasticity. Proceedings of the IUTAM Symposium held in Bochum, Germany. 1999
ISBN 0-7923-5265-3
63. F.C. Moon: IUTAM Symposium on New Applications of Nonlinear and Chaotic Dynamics in
Mechanics. Proceedings of the IUTAM Symposium held in Ithaca, NY, USA. 1998
ISBN 0-7923-5276-9
64. R. Wang: IUTAM Symposium on Rheology of Bodies with Defects. Proceedings of the IUTAM
Symposium held in Beijing, China. 1999 ISBN 0-7923-5297-1
65. Yu.I. Dimitrienko: Thermomechanics of Composites under High Temperatures. 1999
ISBN 0-7923-4899-0
66. P. Argoul, M. Frémond and Q.S. Nguyen (eds.): IUTAM Symposium on Variations of Domains
and Free-Boundary Problems in Solid Mechanics. Proceedings of the IUTAM Symposium
held in Paris, France. 1999 ISBN 0-7923-5450-8
67. F.J. Fahy and W.G. Price (eds.): IUTAM Symposium on Statistical Energy Analysis. Proceedings
of the IUTAM Symposium held in Southampton, U.K. 1999 ISBN 0-7923-5457-5
68. H.A. Mang and F.G. Rammerstorfer (eds.): IUTAM Symposium on Discretization Methods in
Structural Mechanics. Proceedings of the IUTAM Symposium held in Vienna, Austria. 1999
ISBN 0-7923-5591-1
Mechanics
69. P. Pedersen and M.P. Bendsøe (eds.): IUTAM Symposium on Synthesis in Bio Solid Mechanics.
Proceedings of the IUTAM Symposium held in Copenhagen, Denmark. 1999
ISBN 0-7923-5615-2
70. S.K. Agrawal and B.C. Fabien: Optimization of Dynamic Systems. 1999
ISBN 0-7923-5681-0
71. A. Carpinteri: Nonlinear Crack Models for Nonmetallic Materials. 1999
ISBN 0-7923-5750-7
72. F. Pfeifer (ed.): IUTAM Symposium on Unilateral Multibody Contacts. Proceedings of the
IUTAM Symposium held in Munich, Germany. 1999 ISBN 0-7923-6030-3
73. E. Lavendelis and M. Zakrzhevsky (eds.): IUTAM/IFToMM Symposium on Synthesis of Non-
linear Dynamical Systems. Proceedings of the IUTAM/IFToMM Symposium held in Riga,
Latvia. 2000 ISBN 0-7923-6106-7
74. J.-P. Merlet: Parallel Robots. 2000 ISBN 0-7923-6308-6
75. J.T. Pindera: Techniques of Tomographic Isodyne Stress Analysis. 2000 ISBN 0-7923-6388-4
76. G.A. Maugin, R. Drouot and F. Sidoroff (eds.): Continuum Thermomechanics. The Art and
Science of Modelling Material Behaviour. 2000 ISBN 0-7923-6407-4
77. N. Van Dao and E.J. Kreuzer (eds.): IUTAM Symposium on Recent Developments in Non-linear
Oscillations of Mechanical Systems. 2000 ISBN 0-7923-6470-8
78. S.D. Akbarov and A.N. Guz: Mechanics of Curved Composites. 2000 ISBN 0-7923-6477-5
79. M.B. Rubin: Cosserat Theories: Shells, Rods and Points. 2000 ISBN 0-7923-6489-9
80. S. Pellegrino and S.D. Guest (eds.): IUTAM-IASS Symposium on Deployable Structures: Theory
and Applications. Proceedings of the IUTAM-IASS Symposium held in Cambridge, U.K., 6–9
September 1998. 2000 ISBN 0-7923-6516-X
81. A.D. Rosato and D.L. Blackmore (eds.): IUTAM Symposium on Segregation in Granular
Flows. Proceedings of the IUTAM Symposium held in Cape May, NJ, U.S.A., June 5–10,
1999. 2000 ISBN 0-7923-6547-X
82. A. Lagarde (ed.): IUTAM Symposium on Advanced Optical Methods and Applications in Solid
Mechanics. Proceedings of the IUTAM Symposium held in Futuroscope, Poitiers, France,
August 31–September 4, 1998. 2000 ISBN 0-7923-6604-2
83. D. Weichert and G. Maier (eds.): Inelastic Analysis of Structures under Variable Loads. Theory
and Engineering Applications. 2000 ISBN 0-7923-6645-X
84. T.-J. Chuang and J.W. Rudnicki (eds.): Multiscale Deformation and Fracture in Materials and
Structures. The James R. Rice 60th Anniversary Volume. 2001 ISBN 0-7923-6718-9
85. S. Narayanan and R.N. Iyengar (eds.): IUTAM Symposium on Nonlinearity and Stochastic
Structural Dynamics. Proceedings of the IUTAM Symposium held in Madras, Chennai, India,
4–8 January 1999 ISBN 0-7923-6733-2
86. S. Murakami and N. Ohno (eds.): IUTAM Symposium on Creep in Structures. Proceedings of
the IUTAM Symposium held in Nagoya, Japan, 3-7 April 2000. 2001 ISBN 0-7923-6737-5
87. W. Ehlers (ed.): IUTAM Symposium on Theoretical and Numerical Methods in Continuum
Mechanics of Porous Materials. Proceedings of the IUTAM Symposium held at the University
of Stuttgart, Germany, September 5-10, 1999. 2001 ISBN 0-7923-6766-9
88. D. Durban, D. Givoli and J.G. Simmonds (eds.): Advances in the Mechanis of Plates and Shells
The Avinoam Libai Anniversary Volume. 2001 ISBN 0-7923-6785-5
89. U. Gabbert and H.-S. Tzou (eds.): IUTAM Symposium on Smart Structures and Structonic Sys-
tems. Proceedings of the IUTAM Symposium held in Magdeburg, Germany, 26–29 September
2000. 2001 ISBN 0-7923-6968-8
Mechanics
90. Y. Ivanov, V. Cheshkov and M. Natova: Polymer Composite Materials – Interface Phenomena
& Processes. 2001 ISBN 0-7923-7008-2
91. R.C. McPhedran, L.C. Botten and N.A. Nicorovici (eds.): IUTAM Symposium on Mechanical
and Electromagnetic Waves in Structured Media. Proceedings of the IUTAM Symposium held
in Sydney, NSW, Australia, 18-22 Januari 1999. 2001 ISBN 0-7923-7038-4
92. D.A. Sotiropoulos (ed.): IUTAM Symposium on Mechanical Waves for Composite Structures
Characterization. Proceedings of the IUTAM Symposium held in Chania, Crete, Greece, June
14-17, 2000. 2001 ISBN 0-7923-7164-X
93. V.M. Alexandrov and D.A. Pozharskii: Three-Dimensional Contact Problems. 2001
ISBN 0-7923-7165-8
94. J.P. Dempsey and H.H. Shen (eds.): IUTAM Symposium on Scaling Laws in Ice Mechanics
and Ice Dynamics. Proceedings of the IUTAM Symposium held in Fairbanks, Alaska, U.S.A.,
13-16 June 2000. 2001 ISBN 1-4020-0171-1
95. U. Kirsch: Design-Oriented Analysis of Structures. A Unified Approach. 2002
ISBN 1-4020-0443-5
96. A. Preumont: Vibration Control of Active Structures. An Introduction (2nd Edition). 2002
ISBN 1-4020-0496-6
97. B.L. Karihaloo (ed.): IUTAM Symposium on Analytical and Computational Fracture Mechan-
ics of Non-Homogeneous Materials. Proceedings of the IUTAM Symposium held in Cardiff,
U.K., 18-22 June 2001. 2002 ISBN 1-4020-0510-5
98. S.M. Han and H. Benaroya: Nonlinear and Stochastic Dynamics of Compliant Offshore Struc-
tures. 2002 ISBN 1-4020-0573-3
99. A.M. Linkov: Boundary Integral Equations in Elasticity Theory. 2002
ISBN 1-4020-0574-1
100. L.P. Lebedev, I.I. Vorovich and G.M.L. Gladwell: Functional Analysis. Applications in Me-
chanics and Inverse Problems (2nd Edition). 2002
ISBN 1-4020-0667-5; Pb: 1-4020-0756-6
101. Q.P. Sun (ed.): IUTAM Symposium on Mechanics of Martensitic Phase Transformation in
Solids. Proceedings of the IUTAM Symposium held in Hong Kong, China, 11-15 June 2001.
2002 ISBN 1-4020-0741-8
102. M.L. Munjal (ed.): IUTAM Symposium on Designing for Quietness. Proceedings of the IUTAM
Symposium held in Bangkok, India, 12-14 December 2000. 2002 ISBN 1-4020-0765-5
103. J.A.C. Martins and M.D.P. Monteiro Marques (eds.): Contact Mechanics. Proceedings of the
3rd Contact Mechanics International Symposium, Praia da Consolação, Peniche, Portugal,
17-21 June 2001. 2002 ISBN 1-4020-0811-2
104. H.R. Drew and S. Pellegrino (eds.): New Approaches to Structural Mechanics, Shells and
Biological Structures. 2002 ISBN 1-4020-0862-7
105. J.R. Vinson and R.L. Sierakowski: The Behavior of Structures Composed of Composite Ma-
terials. Second Edition. 2002 ISBN 1-4020-0904-6
106. Not yet published.
107. J.R. Barber: Elasticity. Second Edition. 2002 ISBN Hb 1-4020-0964-X; Pb 1-4020-0966-6
108. C. Miehe (ed.): IUTAM Symposium on Computational Mechanics of Solid Materials at Large
Strains. Proceedings of the IUTAM Symposium held in Stuttgart, Germany, 20-24 August
2001. 2003 ISBN 1-4020-1170-9
Mechanics
109. P. Ståhle and K.G. Sundin (eds.): IUTAM Symposium on Field Analyses for Determination
of Material Parameters – Experimental and Numerical Aspects. Proceedings of the IUTAM
Symposium held in Abisko National Park, Kiruna, Sweden, July 31 – August 4, 2000. 2003
ISBN 1-4020-1283-7
110. N. Sri Namachchivaya and Y.K. Lin (eds.): IUTAM Symposium on Nonlnear Stochastic Dynam-
ics. Proceedings of the IUTAM Symposium held in Monticello, IL, USA, 26 – 30 August,
2000. 2003 ISBN 1-4020-1471-6
111. H. Sobieckzky (ed.): IUTAM Symposium Transsonicum IV. Proceedings of the IUTAM Sym-
posium held in Göttingen, Germany, 2–6 September 2002, 2003 ISBN 1-4020-1608-5
112. J.-C. Samin and P. Fisette: Symbolic Modeling of Multibody Systems. 2003
ISBN 1-4020-1629-8
113. A.B. Movchan (ed.): IUTAM Symposium on Asymptotics, Singularities and Homogenisation
in Problems of Mechanics. Proceedings of the IUTAM Symposium held in Liverpool, United
Kingdom, 8-11 July 2002. 2003 ISBN 1-4020-1780-4
114. S. Ahzi, M. Cherkaoui, M.A. Khaleel, H.M. Zbib, M.A. Zikry and B. LaMatina (eds.): IUTAM
Symposium on Multiscale Modeling and Characterization of Elastic-Inelastic Behavior of
Engineering Materials. Proceedings of the IUTAM Symposium held in Marrakech, Morocco,
20-25 October 2002. 2004 ISBN 1-4020-1861-4
115. H. Kitagawa and Y. Shibutani (eds.): IUTAM Symposium on Mesoscopic Dynamics of Fracture
Process and Materials Strength. Proceedings of the IUTAM Symposium held in Osaka, Japan,
6-11 July 2003. Volume in celebration of Professor Kitagawa’s retirement. 2004
ISBN 1-4020-2037-6
116. E.H. Dowell, R.L. Clark, D. Cox, H.C. Curtiss, Jr., K.C. Hall, D.A. Peters, R.H. Scanlan, E.
Simiu, F. Sisto and D. Tang: A Modern Course in Aeroelasticity. 4th Edition, 2004
ISBN 1-4020-2039-2
117. T. Burczyński and A. Osyczka (eds.): IUTAM Symposium on Evolutionary Methods in Mech-
anics. Proceedings of the IUTAM Symposium held in Cracow, Poland, 24-27 September 2002.
2004 ISBN 1-4020-2266-2
118. D. Ieşan: Thermoelastic Models of Continua. 2004 ISBN 1-4020-2309-X
119. G.M.L. Gladwell: Inverse Problems in Vibration. Second Edition 2004 ISBN 1-4020-2670-6
120. J.R. Vinson: Plate and Panel Structures of Isotropic, Composite and Piezoelectric Materials,
Including Sandwich Construction. 2005 ISBN 1-4020-3110-6
121. Forthcoming.
122. G. Rega and F. Vestroni (eds.): IUTAM Symposium on Chaotic Dynamics and Control of
Systems and Processes in Mechanics. Proceedings of the IUTAM Symposium held in Rome,
Italy, 8–13 June 2003. 2005 ISBN 1-4020-3267-6
123. E.E. Gdoutos: Fracture Mechanics. An Introduction. 2nd edition. 2005 ISBN 1-4020-2863-2
124. M.D. Gilchrist (ed.): IUTAM Symposium on Impact Biomechanics from Fundamental Insights
to Applications. 2005 ISBN 1-4020-3795-3
125. J.M. Huyghe, P.A.C. Raats and S.C. Cowin (eds.): IUTAM Symposium on Physicochemical
and Electromechanical Interactions in Porous Media. 2005 ISBN 1-4020-3864-X
126. H. Ding, W. Chen and L. Zhang: Elasticity of Transversely Isotropic Materials. 2005
ISBN 1-4020-4033-4
127. W. Yang (ed.): IUTAM Symposium on Mechanics and Reliability of Actuating Materials.
Proceedings of the IUTAM Symposium held in Beijing, China, 1–3 September 2004. 2005
ISBN 1-4020-4131-6
Mechanics
128. J.-P. Merlet: Parallel Robots. 2006 ISBN 1-4020-4132-2
129. G.E.A. Meier and K.R. Sreenivasan (eds.): IUTAM Symposium on One Hundred Years of
Boundary Layer Research. Proceedings of the IUTAM Symposium held at DLR-Göttingen,
Germany, August 12–14, 2004. 2006 ISBN 1-4020-4149-7
130. H. Ulbrich and W. Günthner (eds.): IUTAM Symposium on Vibration Control of Nonlinear
Mechanisms and Structures. 2006 ISBN 1-4020-4160-8
131. L. Librescu and O. Song: Thin-Walled Composite Beams. Theory and Application. 2006
ISBN 1-4020-3457-1
132. G. Ben-Dor, A. Dubinsky and T. Elperin: Applied High-Speed Plate Penetration Dynamics.
2006 ISBN 1-4020-3452-0
133. X. Markenscoff and A. Gupta (eds.): Collected Works of J. D. Eshelby. Mechanics of Defects
and Inhomogeneities. 2006 ISBN 1-4020-4416-X
134. R.W. Snidle and H.P. Evans (eds.): IUTAM Symposium on Elastohydrodynamics and Microelast-
ohydrodynamics. Proceedings of the IUTAM Symposium held in Cardiff, UK, 1–3 September,
2004. 2006 ISBN 1-4020-4532-8
135. T. Sadowski (ed.): IUTAM Symposium on Multiscale Modelling of Damage and Fracture
Processes in Composite Materials. Proceedings of the IUTAM Symposium held in Kazimierz
Dolny, Poland, 23–27 May 2005. 2006 ISBN 1-4020-4565-4
136. A. Preumont: Mechatronics. Dynamics of Electromechanical and Piezoelectric Systems. 2006
ISBN 1-4020-4695-2
137. M.P. Bendsøe, N. Olhoff and O. Sigmund (eds.): IUTAM Symposium on Topological Design
Optimization of Structures, Machines and Materials. Status and Perspectives. 2006
ISBN 1-4020-4729-0
138. A. Klarbring: Models of Mechanics. 2006 ISBN 1-4020-4834-3
139. H.D. Bui: Fracture Mechanics. Inverse Problems and Solutions. 2006 ISBN 1-4020-4836-X
140. M. Pandey, W.-C. Xie and L. Xu (eds.): Advances in Engineering Structures, Mechanics
and Construction. Proceedings of an International Conference on Advances in Engineering
Structures, Mechanics & Construction, held in Waterloo, Ontario, Canada, May 14–17, 2006.
2006 ISBN 1-4020-4890-4
141. G.Q. Zhang, W.D. van Driel and X. J. Fan: Mechanics of Microelectronics. 2006
ISBN 1-4020-4934-X
142. Q.P. Sun and P. Tong (eds.): IUTAM Symposium on Size Effects on Material and Structural
Behavior at Micron- and Nano-Scales. Proceedings of the IUTAM Symposium held in Hong
Kong, China, 31 May–4 June, 2004. 2006 ISBN 1-4020-4945-5
143. A.P. Mouritz and A.G. Gibson: Fire Properties of Polymer Composite Materials. 2006
ISBN 1-4020-5355-X
144. Y.L. Bai, Q.S. Zheng and Y.G. Wei (eds.): IUTAM Symposium on Mechanical Behavior and
Micro-Mechanics of Nanostructured Materials. Proceedings of the IUTAM Symposium held
in Beijing, China, 27–30 June 2005. 2007 ISBN 1-4020-5623-0
145. L.P. Pook: Metal Fatigue. What It Is, Why It Matters. 2007 ISBN 1-4020-5596-6
146. H.I. Ling, L. Callisto, D. Leshchinsky and J. Koseki (eds.): Soil Stress-Strain Behavior:
Measurement, Modeling and Analysis. A Collection of Papers of the Geotechnical Symposium
in Rome, March 16–17, 2006. 2007 ISBN 978-1-4020-6145-5
147. A.S. Kravchuk and P.J. Neittaanmäki: Variational and Quasi-Variational Inequalities in Mech-
anics. 2007 ISBN 978-1-4020-6376-3
Mechanics
148. S.K. Kanaun and V. Levin: Self-Consistent Methods for Composites. Vol. 1: Static Problems.
2008 ISBN 978-1-4020-6663-4
149. G. Gogu: Structural Synthesis of Parallel Robots. Part 1: Methodology. 2008
ISBN 978-1-4020-5102-9
150. S.K. Kanaun and V. Levin: Self-Consistent Methods for Composites. Vol. 2: Wave Propagation
in Heterogeneous Materials. 2008 ISBN 978-1-4020-6967-3
151. U. Kirsch: Reanalysis of Structures. A Unified Approach for Linear, Nonlinear, Static and
Dynamic Systems. 2008 ISBN 978-1-4020-8197-2
152. G.C. Sih (ed.): Multiscale Fatigue Crack Initiation and Propagation of Engineering Materials:
Structural Integrity and Microstructural Worthiness. Fatigue Crack Growth Behaviour of Small
and Large Bodies. 2008 ISBN 978-1-4020-8519-2
springer.com

Fatigue and Crack Initiation - George C. Sih

Uploaded by

Copyright:

Available Formats

Fatigue and Crack Initiation - George C. Sih

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

Fatigue and Crack Initiation - George C. Sih

Uploaded by

Copyright:

Available Formats

Multiscale Fatigue Crack Initiation and Propagation of

Engineering Materials: Structural Integrity and

Series Editor: G.M.L. GLADWELL

Aims and Scope of the Series

For a list of related mechanics titles, see final pages.

Multiscale Fatigue Crack

ISBN-13: 978-1-4020-8519-2 e-ISBN-13: 978-1-4020-8520-8

Library of Congress Control Number: 2008xxxxx

© 2008 Springer Science+Business Media, B.V.

Printed on acid-free paper

Application of Virtual Testing for Obtaining Fracture Allowable of

An Equivalent Block Approach to Crack Growth 23

Prediction of Fatigue Crack Growth Rates in Ti-6Al-4V Alloy 47

Some Practical Implications of Exponential Crack Growth 65

Fatigue Behaviour of FS, LB and MIG Welds of AA6061-T6 and

Fatigue Damage from Surface to Bulk 113

Microcracking in High Temperature Low Cycle Fatigue 133

Invariant Form of Micro-/Macro-Cracking in Fatigue 181

Fatigue Crack Growth Rate of Cable-Stayed Portion of Runyang Bridge:

Fatigue Crack Growth Rate of Cable-Stayed Portion of Runyang Bridge:

Fatigue of Small-Scale Metal Materials: From Micro- to Nano-Scale 275

Assessment of Fatigue Damage in Heterogeneous Materials by Application

Fatigue Crack Growth of Aircraft Aluminum Alloys 345

Author Index 367

Subject Index 369

to examine microphotographs the phenomenon of multiscale fatigue crack

accumulation. Virtual testing methodology is explored for fatigue crack

Shanghai, China, 2007

1. Introduction & Background

a process zone have been previously considered at high temperatures [5]

2. Detail Description of the Extended Griffith Theory

∂[UE − UF − UU ]/∂c = 0 (2)

2.1 Determination of wUF/wc term

hF = γ (E 2 /σU3 )WF (4)

where J =Ҟ (8/S)Dhmin, VU is the material ultimate allowable stress,

stress strain curve from necking up to fracture (see Fig. 1) and is

2.2 Determination of wUU/wc term

The quantity w UU can be expressed as WUhU, where WU is the

The effective height of the volume in which WU is absorbed in the uniform

where h=0.00056 inch. The unrecoverable energy density WU is:

2.3 Applied stress and crack length relationship

c Half crack length

plane-strain are in proportion to these thicknesses. Let t0 be the maximum

3. Fracture Toughness Calculation

To obtain the fracture toughness value of aerospace alloys through the

(a) 2219-T87 (b) 2014-T6

4. Generating Fatigue Crack Growth Rate Curve without Testing

4.1 Establishing region I (fracture toughness)

4.2 Establishing region II (Paris region)

4.3 Establishing region III (threshold stress intensity factor, Kth)

The methodology used to estimate the threshold stress intensity factor

Kt h = −0.6337 + 0.3148εf (14)

5. Comparison of Predicted Bounds with Experimental Fatigue Crack

is processed. A lesser amount of variation can also be observed in test

Fig. 11 FCG and NASGRO da/dN curves for 6061-T62 aluminum.

For life assessment analysis of structural components it will be

6. Fracture Mechanics & Multiscale Modeling & Simulation

The virtual testing methodology discussed in the previous sections of this

Fig. 12 FCG and NASGRO da/dN curves for Ti-2.5 Cu Titanium.

composite-materials system, these benefits, in some cases, are negated if

'Ki = E 'Vi(Sa) (10)

= [C*(f(Ri) 'Vi /Vmax) J ]

Kmax = Vmax E (S a) (15)

Table 1 Values of C and Kc for specimens tested under a Rotarix spectra.