WELDING

RESEARCH
SUPPLEMENT TO THE WELDING JOURNAL, OCTOBER 2023
Sponsored by the American Welding Society

Understanding and Controlling the Weld

Microstructure of Steels

The analysis of microstructure development in steel weld metals is addressed

BY T. KOSEKI

Abstract Introduction
In this article, selected studies are reviewed with a It was a great pleasure to receive the 2022 Comfort A.
focus on the analysis of microstructure development Adams Lecture Award and to be given this opportunity to
in steel weld metals. In the study of austenitic present a review article about my welding research for the
stainless steel weld metals, microstructure Welding Journal. Over the past 40 years, I have been involved
in various research projects on the welding of different steels.
development in the primary ferrite solidification
The major research interest has been understanding micro-
mode (FA mode) was clarified and related to why FA- structure development during welding because it is always
mode welds are resistant to hot cracking. In studies critical to achieving the desired properties and performances
of duplex stainless steel weld metals and high-Cr of weld metals.
ferritic stainless steel weld metals, nitrogen-driven When I started my research on the weldability of stainless
microstructure development and TiN-assisted steels in the 1980s, research on the solidification and micro-
grain refinement, respectively, were described, and structure of austenitic stainless steel weld metals was actively
discussions about the mechanism of equiaxed grain being driven by pioneering studies conducted in the late ’70s
formation in the weld metals were added in the to early ’80s (Refs. 1–10). Classification of the solidification
latter. In the study of low-alloy steel weld metals, mode and microstructure, prediction of the solidification
the roles of titanium oxide and titanium nitride mode and ferrite content, and the effect of the mode and
(TiN) inclusions on intragranular ferrite formation microstructure on weld performance, such as hot-cracking
and the refinement of weld microstructure were susceptibility, corrosion resistance, and cryogenic toughness,
were investigated by pioneering studies; those studies were
described based on crystallographic analysis and
expanded by subsequent studies for better understanding.
the first principles calculation. At the end, the My initial work (Refs. 11, 18) was among the subsequent stud-
potential importance of the application of different ies being stimulated by the pioneering studies.
multiscale, multiphysics simulations to welding In this context, this article starts with reviewing my
research was pointed out. research on the solidification and microstructure of aus-
tenitic stainless steel weld metals, and then the review is
Keywords extended to those of duplex and ferritic stainless steel and
low-alloy steel weld metals. The main focus is placed on the
■ Steel mechanism of microstructure development in those welds
■ Microstructure and also on how weld microstructure can be controlled. At
■ Solidification the end, some comments are added regarding a future direc-
■ Transformation tion of welding research based on my research activities and
■ Heterogeneous Nucleation recent research trends.

https://doi.org/10.29391/2023.102.018
OCTOBER 2023 | 235-s
 Fig. 1 — Schematic presentation of the solidification mode of austenitic stainless steel weld metals by Brooks
and Thompson (Ref. 12). Note: Modes A, AF, FA, and F were added to the original by the author.

Hot-Cracking Susceptibility of FA-Mode

Austenitic Stainless Steel Weld Metals
The solidification of austenitic stainless steel weld metals
is classified into four modes, A, AF, FA, and F, which are
schematically summarized in Fig. 1, presented by Brooks
and Thompson (Ref. 12). Both the A and AF modes start with
the primary phase of austenite. A completes as austenite
single phase, but AF completes with the formation of fer-
rite in the last stage of solidification in the interdendritic
region by eutectic reaction. In FA mode, solidification starts
with a primary ferrite, and austenite subsequently forms in
the course of solidification by either a peritectic or eutectic
reaction. The primary ferrite partially transforms to austenite
during solidification and postsolidification cooling, which
results in retained ferrite at the core of dendrites with dif-
ferent morphologies in the final microstructure, as depicted
in Fig. 1. In F mode, solidification starts and completes as
ferrite, and austenite subsequently forms from the ferrite
grain boundaries during postsolidification cooling.
Among the four solidification modes of austenitic stainless Fig. 2 — Effect of solidification mode and impurities
steel weld metals, FA mode is the most complicated, and, (P + S) on weld hot-cracking susceptibility (Ref.
thus, its microstructure development has been examined by 14). Note: Modes A, AF, FA, and F were added to the
many researchers (Refs. 12, 13). On the other hand, FA mode original by the author.
is practically important, offering much lower hot-cracking
susceptibility over the A and AF modes (as seen in Fig. 2 by
Kujanpaa et al. [Ref. 14]), which has led to the fact that many 3. Increased solubility of impurities in ferrite that reduces
welding consumables for welding austenitic stainless steels the amount of the low-melting impurity-containing phase.
are alloy designed so the welds can solidify with FA mode.
The understanding of FA mode is important not only for
Brooks and Thompson (Refs. 12, 13) suggested the following
the understanding of fundamental weld behavior but also for
three reasons FA mode offers low hot-cracking susceptibility:
the practical alloy design of welding materials. The author
1. Irregular grain boundaries, resulting in a more- and his colleagues were also interested in and worked on
complicated crack this issue (Refs. 15, 16) with a particular emphasis on crys-
2. Lower surface energies of ferrite/austenite boundaries tallographic analysis of FA-mode welds that had not been
that prevent the boundary from being wetted by impurity conducted sufficiently.
containing liquid

236-s | WELDING JOURNAL

 A B

Fig. 3 — Microstructure of the FA-mode weld metals. A — Skeletal or vermicular ferrite; B — lacy ferrite.

Fig. 4 — Solidification front of a FA-mode (19 wt-% Cr-11 wt-% Ni) stainless steel weld obtained by quenching
the weld pool using liquid tin during welding, along with pole figures of primary ferrite (δ) and interdendrite
austenite (γ) (Ref. 16).

An important result from the crystallographic analysis along with pole figures of the phases. A common feature of
of FA-mode welds is that there is usually a parallel relation- the two FA-mode welds is the parallel relationship between
ship between <100> of ferrite and <100> of austenite in <100> of ferrite and <100> of austenite. To examine where
FA-mode welds. Figure 3 shows two typical microstructures of this parallel relationship starts, the solidification front of a
FA-mode welds having different ferrite morphologies, where FA-mode weld was observed by quenching the weld pool
dark (black) is ferrite and bright (white) is austenite matrix, using liquid tin during welding. The microstructure of the

OCTOBER 2023 | 237-s

 Fig. 5 — Optical microstructure of a FA-mode (19 wt-% Cr-11 wt-% Ni) stainless steel weld and pole figures from
ferrite (δ, black) and austenite (γ, white), where ferrite grows upward from the bottom and changes growth
direction in the middle (as shown with green to blue) while austenite is matrix (as shown with red) (Ref. 16).

weld is shown in Fig. 4 (Ref. 16), along with pole figures of upward from the bottom to the middle of the figure, and
the phases. Ferrite dendrites grow into the liquid (i.e., weld then the growth direction changes to right upward, both of
pool) from the right to left of the figure, which is parallel to the which are also confirmed by the change of <100> direction
heat flow from left to right, and the crystallographic growth in the pole figures attached. On the other hand, as the pole
direction of the ferrite is identified as <100>, the preferential figure for austenite suggests, it seems that austenite (white)
growth direction for the solidification of the body-centered continuously grows upward from the bottom to the top of
cubic (bcc). At the same time, in the interdendritic regions, the figure even after ferrite changes the growth direction in
austenite is observed to grow as if independently, and the the middle. This supports the idea that ferrite and austenite
growth direction is identified as <100>, the preferential grow independently and indicates that the change of growth
growth direction for the solidification of the face-centered direction of ferrite does not occur at the same time as that
cubic (fcc). Therefore, it was concluded that the parallel rela- of austenite.
tionship between <100> of ferrite and <100> of austenite in The independent growth of ferrite and austenite is more
FA-mode welds originally comes from the solidification stage, clearly verified when multiple grains are evaluated by elec-
and this seems to indicate that, in FA-mode solidification, tron backscatter diffraction (EBSD), as shown in Fig. 6 (Ref.
ferrite and austenite grow independently, each growing with 16), where the optical microstructure of the observed area
its preferential growth direction without any restriction on is shown in A and the corresponding EBSD phase mapping
each other. for ferrite and austenite are shown in B and C, respectively.
The next question is whether the parallel relationship It should be noted that the same orientation area results in
between <100> of ferrite and <100> of austenite is con- the same color in each mapping. From ferrite morphology
tinued when the growth direction is changed by the change in A and the corresponding color of ferrite in B, ferrite grain
of heat flow direction with a moving heat source. Figure 5 boundaries are indicated by the dotted lines added in A. How-
(Ref. 16) shows a region in which the growth direction of ever, the austenite grain boundaries identified in C do not
ferrite changes. Skeletal ferrite (black) indicates that it grows match the ferrite boundaries, and, moreover, they are not

238-s | WELDING JOURNAL

 A B C

Fig. 6 — A — Optical microstructure of a FA-mode (19 wt-% Cr-11 wt-% Ni) stainless steel weld; B, C — EBSD
phase mapping of ferrite (δ) and austenite (γ), where ferrite and austenite grow downward from the top and
change growth direction in the middle (as shown with arrows). Note: Arrows and borders with dotted lines
were added to the original figure from Ref. 16 for clarification.

along grain boundaries, those in FA-mode welds were short

and blunt even in the Varestraint test, which should be the
result from irregular austenite grain boundaries in the FA
mode. It is said that hot-cracking susceptibility increases
again when the solidification mode changes from FA to F,
fully ferritic solidification (Ref. 17), although the level of sus-
ceptibility is low as compared with those of the A and AF
modes. This can also be rationalized by the irregular aus-
tenite boundaries achieved by the two-phase independent
solidification in the FA mode.
The other reasons number 2 and 3, proposed by Brooks and
Thompson (Ref. 12), should not be ruled out, but it seems that
they offer additional benefits to the first reason. Hot cracking
in FA-mode welds should take place at austenite-austenite
boundaries, the last part to solidify with low-melting-point
liquid, and, thus, low interfacial energy between ferrite and
austenite may not be the primary reason for low hot-cracking
Fig. 7 — Phosphorus distribution at the solidification
front in A-mode and FA-mode welds. susceptibility. The third reason should lead to decreased
segregation of impurity elements, such as phosphorus and
sulfur, and thereby to reduced cracking susceptibility. Figure
7 (Ref. 18) compares the distribution of phosphorous at the
smooth or straight but rather irregular. This suggests that
solidification front of A-mode and FA-mode welds, both
the grain structure developments of ferrite and austenite
being obtained by the liquid-tin quenching during welding.
are independent of each other, as considered in the previous
Phosphorous is rejected to and enriched in the interdendritic
figure. In addition, the irregular boundaries seem to result
regions in A-mode solidification, while the enrichment of
from the pinning of austenite solidification growth by preex-
phosphorous immediately becomes indistinct behind the
isting primary ferrite. Because the final part of solidification
solidification front due to rapid diffusion in FA-mode solid-
is along austenite grain boundaries, the irregular austenite
ification. This significant back diffusion should be helpful in
boundaries could lead to discontinuous remnant liquid in
reducing phosphorus concentration at the last part to solidify,
the last stage of solidification.
thus reducing the occurrence of cracking there. However, as
It seems that these observations confirm the above- seen in Fig. 2 (Ref. 14), low cracking susceptibility in FA mode
mentioned reason number 1 suggested by Brooks and is still held even with increasing impurity levels, which implies
Thompson (irregular grain boundaries, resulting in a that high solubility and diffusivity of impurity elements in
more-complicated crack) for the low cracking susceptibility ferrite could not be the primary reason for low cracking
of the FA mode. Brooks and Thompson (Ref. 12) reported that susceptibility. Therefore, the irregular solidification grain
while hot cracks in A-mode welds were long and extended

OCTOBER 2023 | 239-s

 Fig. 8 — Microstructure and distribution of Cr, Ni, Mo, and N in a 22 wt-% Cr-6 wt-% Ni-3 wt-% Mo-0.12 wt-%
N duplex stainless steel base metal (Ref. 20). It is noted the figures are all in monochrome in Ref. 20, but the
original color results are shown here.

boundaries we confirmed in an investigation by the author and increasing nitrogen content is the key to the improvement
and his colleagues (Refs. 15, 16) should be the primary reason of weld microstructure and performance. Figures 8 and 9 (Ref.
for the low cracking susceptibility of FA-mode welds. 20) show the microstructure and the distribution of Cr, Ni,
Mo, and N in the base metal and its autogenous gas tungsten
arc weld metal of a duplex stainless steel containing 22 wt-%
Ferrite-to-Austenite Transformation in Cr, 6 wt-% Ni, 3 wt-% Mo, and 0.12 wt-% N. In the base metal,
F-Mode Duplex Stainless Steel Weld Metals the microstructure was a so-called microduplex structure
consisting of alternately layered ferrite and austenite. Ferrite-
The solidification of duplex stainless steel welds cor- forming elements Cr and Mo were enriched in ferrite, and
responds to the F-mode (Fig. 1), which completes with austenite-forming elements Ni and N were enriched in aus-
the ferrite single phase, and austenite is subsequently tenite. In the weld metal, on the other hand, the microduplex
formed in the solid state from the ferrite grain boundar- structure of the base metal was destroyed by remelting and
ies during postsolidification cooling (Ref. 19). The phase solidification, and the microstructure consisted of coarse
balance between ferrite and austenite in duplex stainless ferrite grains with allotriomorphic or Widmansttaten aus-
steels is usually kept equal to achieve good combinations tenite along the grain boundaries. Nitrogen was significantly
of corrosion resistance and mechanical properties, but the enriched in austenite, while the partitioning of Cr, Ni, and
austenite content decreases when duplex stainless steels are Mo between ferrite and austenite was not identified clearly.
welded without filler metals. This is because the ferrite-to- Also, N content was depleted in the vicinity of austenite. It
austenite transformation is suppressed during relatively was concluded that the austenite formation is controlled
rapid postsolidification cooling in the weld, which leads to and stabilized by N in N-bearing duplex stainless steel weld
lower performance of the welds than that of base metals. metals, and the depletion of N in the vicinity of austenite
It was found by the author and his colleague (Ref. 20) that indicates that the transformation to austenite is controlled
ferrite-to-austenite transformation is controlled by nitrogen, by N diffusion. It was also noted that segregation of Ni is

240-s | WELDING JOURNAL

Fig. 9 — Microstructure and distribution of Cr, Ni, Mo, and N of a 22 wt-% Cr-6 wt-% Ni-3 wt-% Mo-0.12 wt-% N
duplex stainless steel autogenous GTA weld (Ref. 20). It is noted the figures are all in monochrome in Ref. 20,
but the original color results are shown here.

A B C

Fig. 10 — Microstructure of duplex stainless steel welds having different Ni and nitrogen contents. A — 22 Cr-9
Ni-3 Mo-0.03 N; B — 22 Cr-6 Ni-3 Mo-0.12 N (almost the same Ni equivalent as A); C — 22 Cr-6 Ni-3 Mo-0.18 N
(Ref. 20).

OCTOBER 2023 | 241-s

 A B

Fig. 11 — Microstructure of high-Cr ferritic stainless steel GTA welds having an alloy composition of: A — 17 wt-%
Cr-0.007 wt-% Ti-0.0097 wt-% N; B — 17 wt-% Cr-0.3 wt-% Ti-0.0092 wt-% N (Ref. 22).

B
A

Fig. 12 — A — Solidification front of 17 Cr-0.3 Ti-0.0092 N ferritic stainless steel weld metal, obtained by the
liquid-tin quenching of the weld pool during welding; B — equiaxed grain formation observed in the pool ahead
of the solidification front; C — TiN at the center of the equiaxed grain and its AES analysis (Ref. 22).

seen in ferrite in the weld metal in Fig. 9. Nickel is probably N help austenite formation even during rapid postsolidifica-
enriched in interdendritic regions of ferrite solidification, tion cooling of welds, while slow diffusion and redistribution
but Ni enrichment does not affect austenite formation as an of Ni does not work effectively. Figure 10C shows the micro-
austenite former, which also supports the idea that N is more structure of the weld containing 0.18% N, indicating further
influential to austenite formation. Those measurements and increase of austenite content over the 0.12% N-containing
analyses were reported for the first time in the research of weld in Fig. 10B. The increase of austenite content and the
duplex stainless steels by the author and his colleague in the improvement of ferrite-austenite balance improve the tough-
Welding Journal (Ref. 20), where the measurement results ness and ductility of welds and corrosion resistance. In this
could not be published in color but in monochrome; thus, in context, most filler materials for the welding of duplex stain-
this article, the originals are shown in color. less steels are so designed that the weld deposits contain
Figure 10 (Ref. 20) demonstrates the beneficial effect of high N content.
N on austenite formation in duplex stainless steel welds.
Figures 10A and B compare the microstructure of two welds Grain Refinement of F-Mode Ferritic
having compositions of 22% Cr-3% Mo-9% Ni-0.03% N and
22% Cr-3% Mo-6% Ni-0.12% N, respectively. They contain
Stainless Steel Weld Metals
different amounts of Ni and N, but their Ni equivalents are
High-chromium ferritic stainless steels solidify with the
almost identical; austenite formation from the ferrite grain
ferritic single phase and complete without any subsequent
boundaries is more obvious in the N-enriched weld, which
phase transformation to austenite; hence, the final micro-
confirms the effect of N. Rapid diffusion and redistribution of

242-s | WELDING JOURNAL

 Fig. 13 — Solubility limit of titanium and nitrogen Fig. 14 — A schematic of equiaxed grain formation in
in 17 Cr steel calculated by Thermo-Calc, showing undercooled melt ahead of growing solid (dendrites)
equiaxed grain formation is associated with TiN (Ref. 26).
formation in the weld pool (Ref. 22).

solubility limit, TiN precipitates at temperatures much higher

structure consists of coarse, elongated ferrite grains growing than the liquidus temperature. Then TiN precipitates away
from the fusion boundaries (Refs. 21–23), as shown in Fig. from the solidification front and could grow by coarsening
11A (Ref. 22). Because this microstructure causes the degra- and ripening mechanisms, and the number of TiN supplied
dation of toughness and ductility, the refinement of the weld to the solidification front could decrease. In summary, it is
microstructure was investigated and found to be achieved by clarified that the refining of the solidification microstructure
the addition of titanium to welds (Refs. 21–23), as shown in in Ti-added high-Cr ferritic stainless steel welds is achieved
Fig. 11B (Ref. 22). The author and his colleague investigated by equiaxed solidification with a good nucleating agent, TiN,
the mechanism of the refinement and clarified the equiaxed and the refinement is fortified when the Ti and N contents
solidification in weld solidification. of steel are so designed as to lead to TiN precipitation at or
By a liquid-tin quenching experiment, it was found that slightly above the liquidus temperature of the steel.
equiaxed grains form in the liquid just ahead of the solidifi- Because weld solidification proceeds with a steep tem-
cation front, and titanium nitride (TiN) exists at the center perature gradient ahead of a growing solid, the possibility
of the grains, as shown in Fig. 12 (Ref. 22). TiN has a sodium of equiaxed solidification is supposed to be quite limited in
chloride-type crystal structure and good lattice coherency weld solidification unless high-potential nucleating agents
with bcc-Fe when (100) and <100> of TiN are parallel to are well supplied just in front of the growing solid in the weld
(100) and <110> of bcc-Fe, known as the Baker-Nutting pool, and in this case, the nucleating agent was confirmed
(B-N) relationship (Ref. 24). Therefore, it is considered that as TiN. Figure 14 is a schematic of columnar-to-equiaxed
TiN forming in the melt works as a nucleating agent for the transition (CET) by Gäumann and Kurz (Ref. 26), where △TN
equiaxed grains of ferrite. Figure 13 (Ref. 22) shows the and △Td are the undercooling necessary for solid nucleation
solubility limit of Ti and N in Fe-17 mass% Cr melt at its liq- on the nucleating agent of interest and tip undercooling of
uidus temperature, which was calculated using Thermo-Calc dendritic growth, respectively. It is noted that tip undercool-
(Ref. 25), and the experimental observation results of equi- ing, △Td, is nearly equal to constitutional undercooling, △TC, in
axed grain formation are superposed in the figure. When welding, and △TC can be calculated from steel chemistry and
the product of Ti and N contents is more than the solubility solidification condition. From Fig. 14, it is obviously inferred
limit (i.e., higher content side of the solubility curve in the that if △Td (i.e., △TC) is greater than △TN, the equiaxed grain
figure), TiN can precipitate in the melt; the contents of Ti can nucleate in the melt ahead of a growing solid. This sit-
and/or N become higher, and the formation temperature uation was simulated by the author and his colleagues by
of TiN in the melt becomes greater. A comparison between using the Monte Carlo method, as shown in Fig. 15 (Ref. 23),
the experimental observation of equiaxed grain formation where it was assumed that equiaxed grain can probabilisti-
and the solubility limit curve suggests that equiaxed grain cally nucleate with △TN of 1.8 K on TiN dispersed randomly in
formation is fortified when the contents of Ti and N are more the melt and the density of TiN was changed; details of the
than the solubility limit and close to the limit. It is inferred that simulation are given elsewhere (Ref. 23). It is shown that
the precipitation of TiN in the melt is a prerequisite for equi- columnar grains initially grow from the fusion boundary and
axed grain formation, and TiN precipitation near the liquidus undergo the transition to equiaxed grains in the middle in the
temperature of the steel (i.e., near the solidification front) presence of TiN, while the transition does not occur without
is more effective to enhance equiaxed grains. This means TiN. The transition happens earlier and the number of equi-
that when the contents of Ti and N are much higher than the axed grains becomes larger as the number density of TiN

OCTOBER 2023 | 243-s

 Fig. 16 — Relationship between (N0)1/3 △TC and the
fraction of equiaxed grain area in weld metals
obtained experimentally and from the Monte Carlo
simulation (Ref. 23).

Microstructure Refinement of Low-Alloy

Fig. 15 — Monte Carlo simulation of the solidification Steel Weld Metals
of 17 Cr-0.4 Ti-N stainless steel weld containing
different densities of TiN as a nucleating agent for In the welding of low-alloy steels, refinement of the final
equiaxed ferrite grain. Note that the solidification microstructure is crucial to the toughness of the weld metal
starts from the base metal of the right side and that and the heat-affected zone (HAZ) of the base metal. The
△TN is assumed to be 1.8 K (Ref. 23). final ferritic and/or bainitic microstructure is produced by
solid-state transformation during cooling from the high-
temperature phase, austenite, and therefore, to refine the
increases. According to the model proposed by Gäumann final microstructure (either the refinement of starting aus-
and Kurz (Ref. 26), CET takes place when the temperature tenite or the increase of nucleation of ferrite or bainite in
gradient in liquid ahead of the growing solid, G, becomes: austenite if needed) (Refs. 27, 28). In the weld metal, it is
difficult to achieve the refinement of austenite due to weld
!
∆$#
solidification, and, therefore, the latter choice — the increase
𝐺𝐺 < 𝑎𝑎(𝑁𝑁! )" ∆𝑇𝑇" (1 −
∆$$
) (1) of the nucleation of ferrite or bainite — has been attempted
by dispersing nitride and/or oxide inclusions as heteroge-
neous nucleation sites in the weld metal (Refs. 27, 28). The
solidification of low-alloy steel weld metals starts with pri-
where N0 is the number of nucleating sites of solid (i.e., the mary ferrite, which changes to austenite by the peritectic
number of TiN in this case) and a is constant. The model reaction in the middle part of solidification. At the same time,
means that CET is more likely as N0 and △TC increase and △TN the primary ferrite transforms to austenite. After solidifi-
decreases. Figure 16 (Ref. 23) plots the relationship between cation, the microstructure eventually becomes austenite
(N0)1/3 △TC and the fraction of the equiaxed grain area in weld single phase, which undergoes significant grain coarsening
metals obtained experimentally and by a Monte Carlo sim- and ripening. Subsequent transformation takes place in the
ulation. The fraction of equiaxed grain area increases with coarse austenite grains.
increasing (N0)1/3 △TC in both experiments and simulations,
For the nucleation of ferrite and/or bainite in austenite of
and it was concluded that CET in welds occurs in accordance
low-alloy steel welds, inclusions, such as TiN and titanium
of the theoretical CET model (Ref. 26) when good nucleating
oxide (TiO), have been identified as effective nucleating
agents offering sufficiently small △TN are present in the weld
agents (Refs. 27, 28). Both TiN and TiO are B1 compounds,
pool.
having a sodium chloride crystal structure, and, thus, offer a
low-energy coherent interface with bcc-Fe in the B-N orien-
tation relationship at the nucleation of ferrite, as previously
described (Ref. 24). There is a tendency that TiN works for
the nucleation of ferrite at relatively high temperatures in

244-s | WELDING JOURNAL

 A B

Fig. 17 — Intragranular ferrite nucleating on inclusions, nitrides, or oxides. A — Polygonal ferrite; B — acicular
ferrite.

A B C

Fig. 18 — Orientation relationship between the embedded compounds and ferrite developed from the
compounds: A — For TiN; B — for TiO; C — for the reference of B-N orientation relationship with an allowance of
5 deg misalignment to be compared with A and B (Ref. 31).

relatively low-carbon-equivalent (i.e., relatively low-strength) Before our study, whether the ferrite could satisfy both orien-
steel welds. The ferrite is often called polygonal ferrite due tation relationships simultaneously had not been examined.
to its morphology; an example is shown in Fig. 17A. On the Single-crystal TiN and TiO were embedded in a low-alloy
other hand, TiO is likely to encourage the nucleation of more steel, and ferrite formation from the embedded compounds
bainitic ferrite at relatively low temperatures in relatively was examined in the steel after a welding thermal cycle with
high-carbon-equivalent steel welds, where the ferrite pro- a peak temperature in the fully austenite region was given
duced is often called acicular ferrite due to its morphology; (Refs. 31, 32). Figure 18 (Ref. 31) shows the orientation rela-
an example is shown in Fig. 17B. It was not clear why TiN and tionship between the embedded compounds and ferrite
TiO showed different behaviors in ferrite formation, which developed from the compounds: A for TiN, B for TiO, and C
became the motivation for our study. Also, when the ferrite for the reference of the B-N orientation relationship with
formation becomes bainitic, such as Widmanstätten fer- an allowance of 5 deg misalignment with circles to be com-
rite, the ferrite may need to have a low-energy orientation pared with A and B. Figure 18A was constructed by EBSD data
relationship with the mother phase, austenite, such as the from 53 ferrite grains, mostly polygonal, and Fig. 18B from
Kurdjumov-Sachs (K-S) orientation relationship (Refs. 29, 30) 49 ferrite grains, mostly acicular. It was shown that ferrite
as well as the B-N orientation relationship (Ref. 24) with TiO. grains generated from TiN and TiO have a B-N orientation
relationship with the compounds, but it seems that ferrite

OCTOBER 2023 | 245-s

 A B

Fig. 19 — First principles calculation results showing differential charge density at the interfaces between bcc-
Fe and B1 compounds: A — TiN; B — TiO with a B-N orientation relationship, {100}bcc-Fe //{100}B1 and <100>bcc-Fe//
<110>B1. Note that for the calculation of the TiO-Fe interface, MgO was used as the base oxide for the stability
of the calculation, and one Mg atom was replaced by the Ti atom (Ref. 31).

A B

Fig. 20 — A schematic summary of intragranular ferrite formation in austenite. A — Polygonal ferrite on TiN;
B — acicular ferrite on TiO.

from TiN shows a stricter B-N relationship than that from TiO. with a B-N orientation relationship: {100}bcc-Fe //{100}B1 and
Ferrite grains from TiO show relatively larger misalignment <100>bcc-Fe//<110>B1. In the figures, an increase in charge
from the B-N relationship. Examining orientation relation- density is show in green and a decrease in blue. It is noted that
ships between ferrite and prior austenite, it was found that in the calculation of the TiO-Fe interface, MgO was used as the
ferrite from TiN did not show a specific relationship, while base oxide, and one Mg atom was replaced by a Ti atom for the
most of the acicular ferrite from TiO showed a K-S relation- stability of the calculation. At the TiN-Fe interface, seen in Fig.
ship. It was considered that ferrite from TiN can select a 19A, there was not any significant charge interaction across
strict B-N relationship without the restriction of austenite, the interface, and interfacial energy sharply increased with
while ferrite from TiO needs to compromise in terms of a B-N increasing misalignment from the B-N atom position. On the
relationship to satisfy the K-S relationship with the mother other hand, at the TiO-Fe interface, there was a charge inter-
phase. Figure 19 (Ref. 31) shows the results of a first princi- action between Fe and Ti atoms. In Fig. 19B, charge density
ples calculation showing differential charge density at the increased at the TiO-Fe interface, as indicated by the green
interfaces between bcc-Fe and B1 compounds, TiN and TiO, areas, which suggests that Ti behaves like metal in TiO. This

246-s | WELDING JOURNAL

 A B

C D E

Fig. 21 — MD simulation of hot press for steel-steel bonding showing changes in the following: A — Total
potential energy during isothermal holding; B–E — lattices annealed at 0.5 Tm after compression (Ref. 34).

contributes to reducing interfacial energy in addition to a B-N

relationship, and, thereby, some misalignment from the B-N
atom position can be allowed for the low-energy interface.
Ferrite formation from B1 compounds is summarized in
Fig. 20. Ferrite from TiN tends to have polygonal morphology,
which is a product of diffusional transformation, and has a
B-N orientation relationship with TiN without a specific orien-
tation relationship with austenite, as shown in Fig. 20A. This
indicates that the dispersion of TiN is more useful in relatively
low-carbon-equivalent (i.e., relatively low-strength) steel
welds and the HAZ, where the austenite-to-ferrite transfor-
mation temperature is relatively high. Ferrite from TiO, on the
other hand, tends to have acicular morphology, indicating the
product of relatively low-temperature transformation, and
has both a B-N relationship with TiO and a K-S relationship
with austenite. The difference from the ferrite from TiN is that
a B-N relationship could be lax at the TiO-ferrite interface
to concurrently satisfy the K-S relationship. Because of the
characteristics, TiO dispersion, which has been successfully
Fig. 22 — Evolution of interfacial strength under applied in various high-strength low-alloy steel weld metals,
isothermal holding time after 5% compression for should be useful in relatively high-carbon-equivalent, high-
bonding (Ref. 34). strength steel weld metals.

OCTOBER 2023 | 247-s

 Future Direction the shrinkage of pores during isothermal holding, which was
relatively slow compared with the first increase of strength
As a future direction of welding and joining research, the due to the relatively long-range diffusion of atoms required
author would like to repeat the importance of the use of var- for pore shrinkage. In summary, the use of the MD calculation
ious numerical simulations to understand complex welding was useful to understanding bonding behavior and interface
behaviors and to better predict weld performance, although formation and also helpful in interpreting the observation of
this importance has been recognized by many researchers the bonding experiment. Its application should be increased
and emphasized by many international conferences and in future welding and bonding research.
seminars, such as Trends in Welding Research and Mathe- It is important to note that multiphysics and multiscale
matical Modelling of Welding Phenomena. The author has simulations are needed to further increase the understanding
applied the Monte Carlo simulation and first principles cal- of welding and joining behaviors because of their complexity.
culation in his research, as in some results shown in Figs. 14 In other words, the integration of simulations of different
and 19, as well as conventional finite difference and element physics and different scales will make it possible to pre-
methods (FDM and FEM) and the calculation of the phase dict properties and performances of welds and joints more
diagram (CALPHAD) method. The Monte Carlo simulation correctly in future welding research. In the integration, exper-
method is quite useful for the modeling of microstructure imental data can be connected to simulations to make each
development, grain growth, and other probabilistic behav- simulation better by data science techniques. The movement
iors during welding and joining, being combined with kinetic of the integration and connection of multiple simulations
simulation of heat conduction, solute diffusion, and parti- along with experimental data has already started in many
tioning, etc., using the FDM. The combination of the FDM areas, most actively in materials science areas. A notewor-
or FEM with the CALPHAD method is also useful because thy example is seen in PRISMS (Ref. 37), which is short for
CALPHAD can supply thermodynamic information to kinetic predictive integrated structural materials science and was
simulation, such as phase stability, driving forces for trans- started by the Materials Genome Initiative (MGI) under Pres-
formation and diffusion, and two-phase equilibria for local ident Barack Obama’s plan as a scientific platform that will
equilibrium assumption. The author and his colleagues enable accelerated predictive materials science for structural
developed a numerical simulation model for weld solidifi- metals. The author also started a similar project in Japan
cation of multicomponent steels that combined the FDM for under the Japanese government’s financial support, which
solidification kinetics and the CALPHAD software Thermo- was called the Materials Integration (MI) project (Refs. 38,
Calc in the early 1990s (Ref. 33), which was the first attempt 39). In this project, different-scale simulations were collected
at this in welding research and was successfully applied to and integrated in the MI system, while experimental data
predict the weld solidification of different austenitic stain- of materials from public and private sectors were collected
less steels. and stored in the MI database to be applied to the simula-
As another important simulation method, the author and tions using data assimilation. The project is still continuing
his colleagues employed a molecular dynamics (MD) calcula- to expand the system and database, and an application of
tion to understand interface development in the bonding and the system was attempted to predict weld performance of
joining of steels and dissimilar metals (Refs. 34–36). Figure 21 structural steels (Ref. 38). Also, there have been many sim-
(Ref. 34) shows a result of a MD calculation that was applied ilar research projects and research activities that focus on
to hot press bonding of iron and iron, the calculation condi- numerical modeling using different simulations.
tions of which were described elsewhere. It was seen that the In summary, the author believes that better understanding
lattice near the bonding interface was significantly distorted, and control of welding and joining behaviors and perfor-
but soon after, short-range rearrangement of atoms started mances of welds and joints would come from those projects
under isothermal holding and nearly completed before the and research activities. Also, it would be nice for young weld-
shrinkage of pores started by surface and boundary diffusion. ing researchers and engineers to participate in those projects
This calculation accounts for the experimental result shown and activities and to make full use of the results.
in Fig. 22 (Ref. 34), in which steel-to-steel hot press bond-
ing was conducted and the evolution of bonding strength Acknowledgments
was measured as a function of isothermal holding time. A
two-step increase in bonding strength was found, and it was The author would like to thank the colleagues and stu-
considered that the first increase of strength resulted from dents who belonged to his research group at Nippon Steel or
the decrease of severe distortion near the bonding interface the University of Tokyo over the past 40 years. In particular,
by short-range rearrangement of atoms, as the MD calculation thanks to my prior colleagues, Prof. Hiroshige Inoue, Osaka
suggests. In fact, both hardness measured by nano-indenter University, and Profs. Junya Inoue and Shoichi Nambu, Uni-
and Kernel Average Misorientation (KAM) values measured versity of Tokyo. Without their efforts and contributions, the
by EBSD near the bonding interface were found to decrease research shown in this article would not have been achieved.
with time (Ref. 35), while the area of bonding interface did Also, thanks to all of my friends in the welding research com-
not increase during the first increase of bonding strength, munity in the United States and all over the world for fruitful
all of which are consistent with the MD calculation and sup- discussions, encouragement, and help in research and related
port the conclusion that the first increase was caused by the activities as well as for their friendship over the years.
decrease of severe distortion near the surface. The second
increase of the bonding interface with time was related to the
observation of the increase of the bonded area resulting from

248-s | WELDING JOURNAL

References 21. Villafuerte, J. C., Pardo, E., and Kerr, H. W. 1990. The effect of
alloy composition and welding conditions on columnar-equiaxed
1. Lippold, J. C., and Savage, W. F. 1979. Solidification of austenitic transitions in ferritic stainless steel gas-tungsten arc welds. Met-
stainless steel weldments: Part I - A proposed mechanism. Welding allurgical Transactions A 21A: 2009–2019.
Journal 58(12): 362-s to 374-s. 22. Koseki, T., and Inoue, T. 2001. Equiaxed solidification of steel
2. Lippold, J. C., and Savage, W. F. 1980. Solidification of austenitic nucleating on titanium nitride. Journal of the Japan Institute of
stainless steel weldments: Part 2 — The effect of alloy composition Metals 65(7): 644–651.
on ferrite morphology. Welding Journal 59(2): 48-s to 58-s. 23. Koseki, T., Inoue, H., Fukuda, Y., and Nogami, N. 2003. Numer-
3. Lippold, J. C., and Savage, W. F. 1982. Solidification of austen- ical simulation of equiaxed grain formation in weld solidification.
itic stainless steel weldments: Part 3 ­— The effect of solidification Science and Technology of Advanced Materials 4(2): 183–195.
behavior on hot-cracking susceptibility. Welding Journal 61(12): 24. Bramfitt, B. L. 1970. The effect of carbide and nitride additions
388-s to 396-s. on the heterogeneous nucleation behavior of liquid iron. Metallur-
4. Takalo, T., Suutala, N., and Moisio, T. 1979. Austenitic solidi- gical Transactions 1: 1987–1995.
fication mode in austenitic stainless steel welds. Metallurgical 25. Sundman, B., Jansson, B., and Andersson, J.-O. 1985. The
Transactions A 10A(8): 1173–1181. Thermo-Calc databank system. CALPHAD 9(2): 153–190.
5. Suutala, N., Takalo, T., and Moisio, T. 1980. Ferritic-austenitic 26. Gäumann, M., and Kurz, W. 1998. Why is it so difficult to pro-
solidification mode in austenitic stainless steel welds. Metallurgical duce an equiaxed microstructure during welding? In Mathematical
Transactions A 11A(5): 717–725. Modelling of Weld Phenomena 4, edited by Cerjak, H., 125–136.
6. David, S. A. 1981. Ferrite morphology and variations in ferrite London: IOM.
content in austenitic stainless steel welds. Welding Journal 60(4): 27. Koseki, T. 2005. A review on inclusion-assisted microstructure
63-s to 71-s. control in C-Mn and low alloy steel welds. Welding in The World
7. Kou, S., and Le, Y. 1982. The effect of quenching on the solid- 49: 22–28.
ification structure and transformation behavior of stainless steel 28. Koseki, T., and Thewlis, T. 2005. Inclusion assisted microstruc-
welds. Metallurgical Transactions A 13A: 1141–1152. ture control in C-Mn and low alloy steel welds. Materials Science &
8. Lundin, C. D., DeLong, W. T., and Sponds, D. F. 1975. Ferrite- Technology 21: 867–879.
fissuring relationship in austenitic stainless steel weld metals. Weld- 29. Kurdjumov, G., and Sachs, G. 1930. On the mechanism of
ing Journal 54(8): 241-s to 246-s. hardening of steel. Z. Physik. 64: 325–343.
9. Brooks, P. S., and Lambert, F. J., Jr. 1978. The effect of phos- 30. Yardley, V. A., and Payton, E. A. 2014. Austenite-martensite/
phorus, sulfur and ferrite content on weld cracking of Type 309 bainite orientation relationship: Characterization parameters and
stainless steel. Welding Journal 57(5): 139-s to 143-s. their application. Materials Science and Technology 30(9): 1125–
10. Cieslak, M. J., Ritter, A. M., and Savage, F. 1982. Solidification 1130.
cracking and analytical electron microscopy of austenitic stainless
31. Kasai, K., Lee, C., Nambu, S., Inoue, J., and Koseki, T. 2010.
steel weld metals. Welding Journal 61(1): 1-s to 8-s.
Transformation behavior of ferrite at steel/B1 compounds interface.
11. Ogawa, T., and Koseki, T. 1988. Weldability of newly developed Tetsu-to-Hagané 96(3): 123–128.
austenitic stainless alloys for cryogenic service: Part II – High-
32. Lee, C., Nambu, S., Inoue, J., and Koseki, T. 2011. Ferrite for-
nitrogen stainless steel weld metal. Welding Journal 67(1): 8-s to 17-s.
mation behaviors from B1 compounds in steels. ISIJ International
12. Brooks, J. A., and Thompson, A. W. 1991. Microstructural 51(12): 2036–2041.
development and solidification cracking susceptibility of austenitic
33. Koseki, T., Matsumiya, T., Yamada W., and Ogawa, T. 1994.
stainless steel welds. International Materials Reviews 36(1): 16–44.
Numerical modeling of solidification and subsequent transforma-
13. Brooks, J. A., Williams, J. C., and Thompson, A. W. 1983. Micro- tion of Fe-Cr-Ni Alloys. Metallurgical & Material Transactions A 25A:
structural origin of the skeletal ferrite morphology of austenitic 1309–1321.
stainless steel welds. Metallurgical Transactions A 14A(6): 1271–1281.
34. Pongmorakot, K., Nambu, S., Shibuta, Y., and Koseki, T. 2017.
14. Kujanpaa, V., Suutala, N., Takalo, T., and Moisio, T. 1979. Cor- Investigation on the mechanism of steel/steel solid-state bonding
relation between solidification cracking and microstructure in at low temperatures. Science and Technology of Welding and Joining
austenitic and austenitic-ferritic stainless steel welds. Welding 22(3): 257–263.
Research International 9(2): 55–76.
35. Pongmorakot, K., Nambu, S., Shibuta, Y., and Koseki, T. 2018.
15. Inoue, H., Koseki, T., Ohkita, S., and Fuji, M. 2000. Formation Effects of compressive strain on the evolution of interfacial strength
mechanism of vermicular and lacy ferrite in austenitic stainless of steel/nickel solid-state bonding at low temperature. Science and
steel weld metals. Science and Technology of Welding and Joining Technology of Welding and Joining 23(4): 344–350.
5: 385–396.
36. Pongmorakot, K., Nambu, S., and Koseki, T. 2018. Numer-
16. Inoue, H., and Koseki, T. 2017. Solidification mechanism of ical analysis of effects of compressive strain on the evolution of
austenitic stainless steels solidified with primary ferrite. Acta Mate- interfacial strength of steel/nickel solid-state bonding. Materials
rialia 124: 430–436. Transactions 59(4): 568–574.
17. Westin, E. M. 2022. Hot cracking in duplex stainless steel weld- 37. PRISMS, Electronic References. Retrieved August 14, 2023,
ments – A review. Welding in the World 66: 1483–1499. from prisms-center.org/#/home.
18. Koseki, T. 1989. Unpublished data. 38. Demura, M., and Koseki, T. 2020. SIP – materials integration
19. Suutala, N., Takalo, T., and Moisio, T. 1979. Single-phase fer- project. Materials Transactions 61(11): 2041–3046.
ritic solidification mode in austenitic-ferritic stainless steel welds. 39. National Institute for Materials Science. Materials Data Plat-
Metallurgical Transactions A 10A(8): 1183–1190. form. nims.go.jp/eng/research/platform/materials-data
20. Ogawa, T., and Koseki, T. 1989. Effect of composition profiles
on metallurgy and corrosion behavior of duplex stainless steel weld TOSHIHIKO KOSEKI (koseki.toshihiko@kuas.ac.jp) is formerly
metals. Welding Journal 68(5): 181-s to 191-s. with the University of Tokyo and currently with the Kyoto University
of Advanced Science, Kyoto, Japan.

OCTOBER 2023 | 249-s