See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/357214908

Thermomechanical simulation of the heat-affected zones in welded ultra-

high strength steels: Microstructure and mechanical properties

Article  in  Materials and Design · December 2021

DOI: 10.1016/j.matdes.2021.110336

CITATIONS READS

4 219

7 authors, including:

Shahriar Afkhami Vahid Javaheri

Lappeenranta – Lahti University of Technology LUT University of Oulu
20 PUBLICATIONS   309 CITATIONS    79 PUBLICATIONS   594 CITATIONS   

SEE PROFILE SEE PROFILE

Mohsen Amraei Tuomas Skriko

University of Turku Lappeenranta – Lahti University of Technology LUT
22 PUBLICATIONS   210 CITATIONS    33 PUBLICATIONS   277 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Slurry Erosion of Ultra-hard Induction Hardened Wear Resistant Steel View project

Advanced steels for a green planet View project

All content following this page was uploaded by Vahid Javaheri on 31 December 2021.

The user has requested enhancement of the downloaded file.

 Materials & Design 213 (2022) 110336

Contents lists available at ScienceDirect

Materials & Design

journal homepage: www.elsevier.com/locate/matdes

Thermomechanical simulation of the heat-affected zones in welded

ultra-high strength steels: Microstructure and mechanical properties
Shahriar Afkhami a, Vahid Javaheri b, Mohsen Amraei c,f,⇑, Tuomas Skriko d, Heidi Piili c,f, Xiao-Ling Zhao e,
Timo Björk a
a
Laboratory of Steel Structures, LUT University, P.O. Box 20, 52850 Lappeenranta, Finland
b
Materials and Mechanical Engineering, University of Oulu, 90014, Finland
c
Laboratory of Laser Materials Processing and Additive Manufacturing, LUT University, Lappeenranta 53850, Finland
d
Laboratory of Welding Technology, LUT University, P.O. Box 20, 53851 Lappeenranta, Finland
e
The School of Civil and Environmental Engineering, UNSW Sydney, NSW 2052, Australia
f
Department of Mechanical and Materials Engineering, University of Turku, Turku, 20520, Finland

h i g h l i g h t s g r a p h i c a l a b s t r a c t

Thermomechanical simulation using

Gleeble proved to be a reliable
approach to replicate and reconstruct
HAZ subzones.

Ferrite and granular bainite as the
primary cause behind the HAZ
softening.

Acceptable reliability of the hardness
models based on microstructures to
predict soft zones.

Controversial simultaneous decrease
in hardness and toughness was
associated with carbide precipitation
in tempered phases.

a r t i c l e i n f o a b s t r a c t

Article history: Ultra-high strength steels (UHSS) have a determining role in construction and industry. Furthermore,
Received 20 September 2021 welding as the primary joining process for steel has a similar role in promoting its applications.
Revised 13 December 2021 Therefore, welded UHSS have a vital role in related applications. However, due to their complex
Accepted 19 December 2021
microstructures, these steels are more prone to harmful effects of welding heat input on the mechanical
Available online 21 December 2021
properties compared to mild steels. Thus, identifying the correlations between the microstructural trans-
formations triggered by the heat input and the mechanical properties can lead to new insights and hin-
Keywords:
dering the drawbacks. This study investigates the microstructures and mechanical properties of S960
Ultra-high strength steel
Heat-affected zone
(with a severe softening after welding) and S1100 (with a negligible decrease of the mechanical proper-
Mechanical properties ties after welding) to understand the mechanisms behind the softening of welded UHSS. Microstructural
Microstructure analysis showed the formation of soft phases, e.g., ferrite and granular bainite, as the primary reason for
the softening. Furthermore, tempered forms of martensite and bainite resulted in the simultaneous
decrease of hardness and notch toughness. Finally, the applicabilities of two experimental approaches
to predict hardness from microstructural constituents were evaluated for welded S960 and S1100 and
proved to have relatively good reliability to detect their HAZ softened spots.
Ó 2021 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).

⇑ Corresponding author.
E-mail address: [email protected] (M. Amraei).

https://doi.org/10.1016/j.matdes.2021.110336
0264-1275/Ó 2021 The Authors. Published by Elsevier Ltd.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
S. Afkhami, V. Javaheri, M. Amraei et al. Materials & Design 213 (2022) 110336

1. Introduction microstructures should be recognized in the first place to assist

this approach [3,11].
Steel, as a pivotal engineering material, is vital to metal struc- According to the literature, HAZ in welded UHSS can be divided
tures. Various sectors such as transportation, shipbuilding, and into several subzones depending on their microstructures and dis-
the construction are heavily dependent upon this material. These tances from the fusion line. The distance determines the peak tem-
resource-intensive industries are currently more focused on effec- perature (Tpeak) and cooling rate each subzone goes through to
tively protecting the environment and sustainable development reach its final microstructure and mechanical properties. These
due to the contemporary natural crisis associated with climate subzones usually include coarse-grain (CG), fine-grain (FG), inter-
change and high carbon emissions. In order to achieve these goals, critical (IC), and sub-critical (SC) HAZs. Each subzone could have
natural resources should be used more efficiently, energy con- its unique physical and mechanical characteristics differing from
sumption and carbon emission should be reduced, and material those of others and might act as a weak point through the welded
waste has to be minimized. Thus, modern steel structures and area. Hence, it is crucial to evaluate each subzone independently to
components should be more energy-efficient and withstand higher thoroughly analyze the behavior of HAZ as a whole [2,3,11,20].
extreme natural loads to extend service life [1,2]. This necessity Most of the investigations in the literature are focused on the
accentuates the importance of the new generation of micro- overall performance of welded joints. Therefore, in a preceding
alloyed steels with ultra-high strength levels and simple chemical study, the mechanical properties of each individual HAZ subzone
compositions, along with efficient and economical manufacturing following the unique thermal cycle it experienced through a gas
processes of these steels, since they make it possible to build metal arc welding (GMAW) procedure, as a commonly used indus-
cost-efficient and robust steel structures with significant trial welding process, were investigated [11]. However, the
strength-to-weight ratios [3]. microstructural transformations and their mechanisms were
According to the current definition, steels with minimum nom- excluded since they were out of the scope of that research. Hence,
inal yield strength higher than 700 MPa are known as ultra-high the current study examines the microstructure of each subzone
strength steels (UHSS). As a type of advanced high strength steels individually to evaluate the HAZ. In the present study, the
(AHSS), these metals consist of complex multiphase microstruc- microstructural transformations, their resultant microstructures,
tures that obtain good combinations of strength and ductility by and their correlation to the mechanical properties are probed via
grain refinement through thermomechanical processing tech- independent thermomechanically reconstructed samples. Conse-
niques, e.g., a reiterative combination of rolling, tempering, or quently, the link between the welding HI and the microstructures
quenching cycles. The resultant microstructure usually is a combi- is assessed to understand the harmful effects of welding on UHSS.
nation of ferrite, bainite, martensite, and retained austenite [4–6]. For this purpose, two types of UHSS, S960 (with a relatively high
These steels have become popular in typical industrial applica- softening at its HAZ) and S1100 (with a relatively negligible HAZ
tions; However, the available range of their strength is not fixed softening), are chosen to provide the reader with a reasonable
yet due to their constant development. The current yield limit of comparison to understand which microstructural transformations
UHSS has reached above 1400 MPa, and it improves progressively. resulted in the mechanical deterioration. These two UHSS grades
Therefore, their domain of applications becomes wider every day have different alloying and manufacturing processes. While the
[3,6]. S960 is manufactured via direct quenching, the S1100 has gone
To leverage the most from UHSS in demanding applications, through a quenched and tempering process.
welding as a common joining technique for steel structures and
components is vital. However, due to the thermomechanical
strengthening mechanisms and complex nature of the microstruc- 2. Materials and methods
tures of these steels, UHSS are more prone to the adverse effects of
welding on their properties compared to conventional carbon In this study, the HAZs of S960 and S1100 structural steels are
steels with lower strength ranges [2,3,7]. Welding heat input (HI) divided into seven distinct subzones for close examination based
exposes UHSS to microstructural transformations that are not in on their distance from the weld fusion line. The peak temperature
favor of the mechanical performance of their welded joints. These and cooling rate associated with each subzone are presented in
transformations affect the hardness, ductility, or strength of the Table 1. A Gleeble 3800 thermomechanical simulator was used to
materials in their heat-affected zone (HAZ) [8–10]. Therefore, char- simulate these subzones. Further information on the time–temper-
acterizing the microstructures and identifying their influential evo- ature data of the reconstructed regions is presented in [11]. The as-
lutions before and after welding are essential to recognize the received materials were cut into 55 mm  10 mm  5 mm blocks
adverse effects of this process and hinder them while joining UHSS to fit in the simulator copper grips. The free span was of about
[11]. 22 mm and the thermocouple was placed exactly in the mid-
Numerous studies have been recently focused on welding of length of specimen. The thermal cycles of each subzone were sim-
UHSS and the effects of HI on their microstructures and properties ulated based on the welding parameters presented in Table 2.
[2,3,11–15]. According to these investigations, a decrease in These welding parameters represent an actual GMAW process typ-
strength and hardness in the HAZ, generally known as HAZ soften- ically used in industrial applications. The hardness values of the
ing, is the most severe drawback associated with welded UHSS,
causing up to 60% loss in the hardness or strength [11,16]. This Table 1
negative effect is especially highlighted in low alloy direct- Time-temperature data of the simulated HAZ subzones [11].
quenched types of UHSS. This degradation of the mechanical prop-
Subzone Estimated distance from the Approximated peak Dt8/5
erties is due to the microstructural changes that result in undesir- number fusion line (mm) temperature (°C) (s)
able phases in the HAZ, which are triggered by the welding HI.
1 0.0 1350 10.3
Furthermore, some other properties of welded joints in UHSS,
2 0.5 1154 10.4
e.g., toughness and load-bearing capacity, are also proved to rely 3 1.0 1000 10.5
on HI and its resultant microstructure [2,7,11,17–19]. However, it 4 1.5 890 10.7
is possible to minimize the harmful effects of HI by controlling it 5 2.0 800 11.6
6 2.5 730 –
to prevent or limit unwanted microstructural transformations.
7 3.0 670 –
Therefore, the undesirable transformations and their resultant
2
S. Afkhami, V. Javaheri, M. Amraei et al. Materials & Design 213 (2022) 110336

Table 2
Welding variables determining the Gleeble thermal cycles [11].

Variable Welding potential(V) Welding current(A) Welding speed(mm/s) Linear heat input(kJ/mm)
Value 25 220 6.2 0.7

samples were measured with a Struers DuraScan 70 by applying 5 3. Results

kgf for 15 s. After that, the middle sections of the samples were
ground sequentially with different abrasive pads (up to 2000 grit) The microstructures of the base metals and their simulated HAZ
and polished using 1 lm colloidal silica. Nital etchant with 15 s subzones comprised several microstructural features, including
soaking time was used to reveal the microstructural features of (polygonal) ferrite, various bainitic morphologies (i.e., granular,
the samples by scanning electron microscopy (SEM). SEM was per- upper, and lower bainite), lath-structured martensite, and retained
formed with a SU3500 microscope from Hi-Tech Instruments. Con- austenite. After welding, the constituent fractions of these features
tinuous cooling transformation (CCT) diagrams of the materials depended upon the base metal microstructures and their prior
were modeled by the JMatPro software package using the chemical austenite grains (as material dependant parameters) and the dis-
compositions of S960 and S1100 provided as certificates by their tance from the fusion line determining the thermal gradients
manufacturer (Table 3). [11]. Data achieved via SEM, FESEM, and EBSD are presented in
A field emission scanning electron microscope (FESEM-ZEISS, Sections 3.1 and 3.2. These results were simultaneously employed
SIGMA) was employed for the electron backscatter diffraction to analyze and quantify microstructural features from each sub-
(EBSD) measurements. In addition to the described sample prepa- zone accurately.
ration for SEM, an additional polishing step using 0.04 mm colloidal
silica was carried out on the EBSD samples. The EBSD measure- 3.1. Microstructural features
ments were performed using an accelerating voltage of 15 kV, a
working distance of 15 mm, and a scanning step size of 0.2 mm. Microstructures of S960 and S1100 in their as-received con-
The data were acquired from an area size of 220 mm  220 mm. dition are presented in Figs. 2 and 3, respectively. According to
Image quality (IQ) data from the EBSD results (EBSD-IQ) was Fig. 2, S960 consisted of a mixture of bainite (B), martensite
used, in addition to the SEM qualitative analysis, to have a more (M), and small sporadic islands of retained austenite (RA)
reliable identification and quantification of the microstructural arranged along the prior austenite grain boundaries (PAGB).
features. As reported in [21–23], different phases and microstruc- PAGBs are marked by red dashed lines in all the figures.
tural constituents produce distinct IQ values in EBSD measure- According to [27], martensite in microalloyed steels with
ments. The distinction is because of the varying dislocation complex chemical compositions, e.g., S960 and S1100, is an
networks and densities of each microstructural feature. Hence, ultra-fine mixture of martensite/austenite (MA) with a distinct
quantitative analysis of microstructures can be performed by ana- un-etched smooth appearance in SEM images. In such a mix-
lyzing the numerical values of EBSD-IQ. For this purpose, based on ture, retained austenite is trapped between the small laths of
the previous work by DeArdo [24,25], IQ histograms were normal- martensite. However, these fine features are not visible under
ized and then deconvoluted into multiple peaks as the representa- magnifications commonly used in SEM analysis. Thus, these
tives of different phases by assuming a normal distribution. Then, microstructural features are marked as martensite (M) in the
each unique peak was assigned to a specific microstructure compo- images from the current study since SEM data are used for
nent according to its position in the normalized IQ value axis. Then, microstructural evaluations. Similar to S960, the microstructure
the fraction of each microstructural constituent was calculated as of S1100 comprised a mixture of bainite and martensite with
the ratio of the area below each peak to the total normalized IQ small islands of retained austenite scattered along the bound-
area. aries (Fig. 3).
Regarding the quasi-static tensile tests, the blocks processed Microstructural features of HAZ subzones from S960 are pre-
with the Gleeble 3800 were mechanically machined to test speci- sented in Figs. 4 and 5, following their sequence and distance
mens following the dimensions shown in Fig. 1 (a). These dimen- from the fusion line. According to these figures, the subzones of
sions were selected since they are compatible with the sample S960 comprised different mixtures of ferrite (F), bainite, lath
size limitations associated with the Gleeble machine. The tests martensite, tempered martensite (TM), bainitic ferrite (BF), gran-
were carried out at room temperature ( 20 °C), with a 150 kN test ular bainite (GB), and retained austenite. However, the fractions
rig, under the constant strain rate of 5  10-4 s1. The digital image of these features were different in each subzone following the
correlation (DIC) technique was utilized using an ARAMIS system unique thermal cycle it endured through the simulation proce-
to compare local strain values in the center of the gauge areas. dure using the Gleeble machine. Microstructural features of HAZ
Finally, Charpy impact toughness tests were conducted at room subzones from S1100 are presented in Figs. 6 and 7. The subzones
temperature on specimens with dimensions shown in Fig. 1 (b). of S1100 were different mixtures of martensite, bainite, tempered
Due to the limited thickness of the base materials in their original martensite, autotempered martensite (AM), and bainitic ferrite.
forms (8 mm thick plates), sub-size samples were used for both All the microstructural features mentioned for S960 and S1100
tensile and Charpy tests. The Charpy specimens were per the HAZ subzones were confined between prior austenite grain
Annex A3 of ASTM E23 [26]. boundaries.

Table 3
Chemical compositions of the base materials according to their certificates.

Material C (wt%) Si (wt%) Mn P S V Cu Cr Ni Mo CEV

(wt%) (wt%) (wt%) (wt%) (wt%) (wt%) (wt%) (wt%) (wt%)
S960 0.088 0.040 1.130 0.010 0.000 0.009 0.011 1.000 0.050 0.117 0.590
S1100 0.133 0.192 1.490 0.006 0.001 0.152 0.437 1.340 0.998 0.388 0.850

3
S. Afkhami, V. Javaheri, M. Amraei et al. Materials & Design 213 (2022) 110336

Fig. 1. Schematic views of sub-size (a) tensile and (b) Charpy specimens (lengths and angles are in mm and degree, respectively).

Fig. 2. Microstructure of S960 in its as-received condition: (a) 2000  and (a’) 5000.

Fig. 3. Microstructure of S1100 in its as-received condition: (a) 2000  and (a’) 5000.

3.2. EBSD results measurements can potentially reveal more structural features than
secondary electron images since they represent the quality of
Inverse pole figure (IPF) maps of the subzones from S960 and acquired diffraction patterns [21,22]. IQ values might be affected
S1100 are presented in Fig. 8. The size of PAGBs from different sub- by microstructural characteristics, surface topology, local chem-
zones can be evaluated and compared accurately by the average istry, or residual plastic deformations [28,29]. Consequently, com-
values of equivalent circle diameters (ECD). The ECD measure- pared to typical SEM images, IQ analysis can provide more
ments from the HAZ subzones in this study are presented in information about the microstructural features. In general, accu-
Fig. 9. According to the results, the microstructure of each subzone mulated strains in a crystal lattice are the primary parameter that
differed from those of the others in terms of the type and size of alters IQ values throughout the lattice microstructure. These
the microstructural features and the size of the PAGBs. The differ- strains are predominantly caused by dislocations, and there is a
ences in these features from the subzones can be attributed to their direct relationship between the strain values and dislocation
different thermal histories. Each HAZ subzone had its unique peak densities. Consequently, changes in IQ values can also represent
temperature and cooling rate according to its distance from the fluctuations in dislocations densities. Hence, a microstructure
fusion line (Table 1). component with a high dislocation density results in a low
In order to characterize the final microstructure and quantify IQ-valued region in its EBSD measurements and vice versa.
the fraction of each microstructural component, the EBSD-IQ data In steels similar to S960 and S1100, under continuous cooling
were analyzed using the method developed by DeArdo [24]. regimes (similar to those that the materials experience through
Pattern or image quality (IQ) maps obtained from the EBSD welding procedures), the IQ values of the transformed

4
S. Afkhami, V. Javaheri, M. Amraei et al. Materials & Design 213 (2022) 110336

Fig 4. Microstructural features from the simulated HAZ of S960: subzones 1 (a, a’), 2 (b, b’), 3 (c, c’), and 4 (d, d’).

microstructural components decrease following their transforma- approach, the areal ratio of each curve under the normalized IQ
tion temperatures. The IQ values of (polygonal) ferrite are the high- curve determines the fraction of its representative phase or mor-
est, and the values drop sequentially from ferrite to granular phology, as shown in Fig. 10 for subzones of S960 as an example.
bainite, upper bainite, lower bainite, and martensite. Thus, the IQ Phase fractions achieved via this approach are presented in
curves can be deconvoluted to distinct sub-curves depending on Fig. 11 (a) and (b) for S960 and S1100, respectively.
the number of potential microstructural components [24,30]. The
position of each deconvoluted peak in the normalized IQ axis (from 3.3. Local strains, hardness, and notch toughness
0 to 100), which was used to identify the components, can be
described as follows: 70–80 for (Polygonal) ferrite and granular The local strains of HAZ subzones under fixed quasi-static loads,
bainite, 50–70 for upper bainite, 30–40 for lower bainite and in addition to their hardness and notch toughness values, were
martensite, and 20–30 for tempered initial microstructure. In this measured to compare these regions regarding their mechanical
5
S. Afkhami, V. Javaheri, M. Amraei et al. Materials & Design 213 (2022) 110336

Fig. 5. Microstructural features from the simulated HAZ of S960: subzones 5 (a, a’), 6 (b, b’), and 7 (c, c’).

properties. The comparison made it possible to understand a corre- and some amount of ferrite. All these features are confined within
lation between the mechanical behavior of the subzones and their coarse prior austenite grains resulted from the very high peak tem-
microstructures, as this is discussed in Section 4.2. The local strain perature of this subzone. This evaluation agrees with the SEM data
ranges under a fixed quasi-static load are shown for S960 and shown in Fig. 4 (a) and (a’).
S1100 in Figs. 12 and 13, respectively. The value of the fixed load After region 1, subzones 2, 3, and 4 had the highest peak tem-
was chosen following the tensile strength of the weakest specimen peratures sequentially and experienced microstructural evolutions
(subzone) for each steel grade. The hardness and notch toughness similar to subzone 1. Consequently, they also comprised a mixture
values were extracted from the preceding study introduced in Sec- of martensite, bainite, and ferrite. However, due to their different
tion 1 and are presented in Fig. 14 (a) and (b), respectively [11]. peak temperatures, the prior austenite grain size decreased from
zone 1 to 4 (Fig. 4 (a), (b), (c), and (d)). This expectation agrees with
the experimental results shown in Fig. 4 (d, d’) and 9 (b). The sum-
4. Discussion mation of these microstructural modifications (changes in prior
austenite grain sizes and phase ratios) did not cause any significant
The expected microstructural transformations and features in changes in the hardness values, and only limited fluctuations were
the HAZs of S960 and S1100 can be identified via their CCT dia- observed in these areas (further discussion can be found in Sec-
grams. The estimated CCT diagrams of S960 are presented in tion 4.2) [11].
Fig. 15. Considering the austenitization-finish temperature (A3) The second type of HAZ subzones consists of zones 5, 6, and 7.
of S960, HAZ subzones of this steel can be divided into two groups. These areas did not reach the full austenitization temperature
The first group consisting of subzones 1 to 4, with peak tempera- range. Although Subzone 5 had a peak temperature close to A3,
tures higher than A3, experienced full austenitization, while the the cooling rate was not high enough to favor martensite or lower
rest had partial (subzones 5–6) to no austenitization (subzone 7) bainite formation. Thus, austenite to ferrite, granular bainite, and
depending on their peak temperatures. According to Fig. 15 (a), upper bainite transformations were more expected. Consequently,
subzone 1 is expected to comprise a mixture of martensite, bainite, this area was a combination of ferrite/granular bainite, fragments
6
S. Afkhami, V. Javaheri, M. Amraei et al. Materials & Design 213 (2022) 110336

Fig. 6. Microstructural features from the simulated HAZ of S1100: subzones 1 (a, a’), 2 (b, b’), 3 (c, c’), and 4 (d, d’).

of tempered martensite or bainitic ferrite, a small amount of fine- Regarding subzone 6, the maximum temperature barely passed
textured fresh bainite (as a result of partial austenitization and low A1. Thus, a minor austenitization is expected to occur in this area.
cooling rate), and a small number of retained austenite islands spo- Furthermore, the limited cooling rate of the area is more in favor of
radically scattered throughout the microstructure (Fig. 5 (a, a’)). bainite formation than martensite. Consequently, the microstruc-
The ferrite/granular bainite formation and decline in martensite/ ture consists of tempered martensite and bainitic ferrite originat-
lower bainite contents in this area resulted in a sudden drop in ing from the base metal and fragments of fresh bainite resulting
its hardness and the existence of a soft spot in the HAZ (generally from the limited partial austenitization (Fig. 5 (b, b’)). Although
known as softened zone). the hardness drop from zone 5 recovered a little due to the absence

7
S. Afkhami, V. Javaheri, M. Amraei et al. Materials & Design 213 (2022) 110336

Fig. 7. Microstructural features from the simulated HAZ of S1100: subzones 5 (a, a’), 6 (b, b’), and 7 (c, c’).

of ferrite in region 6, tempered microstructure did not cause a sig- ferrite nucleation and growth through the cooling down of S1100
nificant increase in the hardness since tempered martensite and prevented subzones 1 to 5 from any hardness decrease compared
bainite are known to be relatively soft mixtures compared to their to the base material (further discussion is provided in section 4.2).
original states. Furthermore, the combination of bainite and martensite as the
Finally, subzone 7, as the boundary between HAZ and the base microstructures of these zones resulted in a moderate increase in
metal, experienced peak temperatures as high as 670 °C and did the hardness values. In addition, by decreasing the peak tempera-
not pass any critical temperature through its heating cycle. Thus, ture from zone 1 to 5, the ratio of martensite to bainite con-
the microstructural features of subzone 7 consisted of tempered stituents and prior austenite grain size increased and decreased,
martensite and bainitic ferrite, in addition to some amount of respectively (Figs. 6 and 7). These phenomena resulted in a hard-
martensite and bainite remaining from the initial microstructure ness increase in subzone 1 and relatively small hardness fluctua-
(Fig. 5. (c, c’)). This mixture resulted in hardness values similar to tions from zone 1 to 5. Finally, some traces of autotempered
that of subzone 6. In addition, white precipitates arranged along martensite was also detected in these regions. However, this con-
the high angle boundaries in subzones 6 and 7, especially 7 in stituent did not cause any drastic changes in the mechanical prop-
Fig. 5 (c’), are due to the tempered microstructural features in these erties of these subzones, as shown in section 4.2.
areas. Regarding subzone 6, the peak temperature is slightly higher
The CCT diagrams of S1100 are presented in Fig. 16. Similar to than A1, and the material is partially austenitized through its heat-
S960, HAZ subzones of S1100 can be divided into two groups of ing stage. Thus, the final microstructure is expected to be a mixture
fully austenitized (subzones 1, 2, 3, 4, 5) and not fully austenitized of fresh martensite originated from the partial austenitization and
(6 and 7) areas, depending on their heating cycles and peak tem- a mixture of tempered martensite and bainitic ferrite due to the
peratures. According to Fig. 16, there is no austenite to ferrite tempering of the initial microstructure (Fig. 7 (b, b’). The tempered
transformation through the cooling stage of S1100, unlike S960. microstructural features resulted in a sudden drop in the hardness
Thus, the microstructures of these subzones mainly comprised value in subzone 6 compared to its predecessors. However, the
mixtures of bainite and martensite. In addition, the absence of hardness decrease is not too severe to cause a softened spot in

8
S. Afkhami, V. Javaheri, M. Amraei et al. Materials & Design 213 (2022) 110336

Fig. 8. IPF maps of subzones 1–7 for S960 (a-g) and S1100 (a’-g’) along with the pick temperatures indicated on the side arrows.

the HAZ as the hardness value is close to that of the base material 1 and 4 are highlighted by dashed red lines on the CCT diagrams of
(Section 4.2). Finally, heating S1100 up to 670 °C and cooling it S960 and S1100 (Figs. 15 and 16, respectively). According to the
down did not cause any significant microstructural changes in this calculated CCT diagrams, the austenitization-start temperature
material in subzone 7 (Fig. 7 (c, c’)). (A1) for S960 was 722 °C which was slightly higher than that of
S1100 (716 °C). However, the difference in austenitization-finish
4.1. EBSD and quantitative phase analysis temperatures (A3) was significant as they were estimated at
826 °C for S960 and 783 °C for S1100. Accordingly, as illustrated
Depending on the distance from the fusion line, each subzone in Fig. 8 using the IPF maps, in the case of S960, only subzones 1
experienced a unique thermal cycle, peak temperature, and Dt8/5, to 4 experienced full austenitization during their heating. Regard-
as described in Table 1. As examples, the cooling paths of subzones ing S1100, in addition to subzones 1 to 4, subzone 5 also reached
9
S. Afkhami, V. Javaheri, M. Amraei et al. Materials & Design 213 (2022) 110336

Fig. 9. The reconstructed IPF maps showing prior austenite grain size for the subzones 1, 4, 6: S960 (a-c) and S1100 (a’-c’).

the temperature range of full austenitization. The extra fully temperature–time-dependent phenomenon, prior austenite grain
austenitized subzone in S1100 occurred due to the lower A3 of this size decreased gradually by decreasing the peak temperature in
steel than S960. both alloys. However, due to composition differences, S1100
According to EBSD results, the elongated texture of the BM showed a more refined grain structure in all the conditions. The
was eliminated in the fully austenitized subzones, and new ECD values from the measurements can be found in Fig. 9. In
equiaxed grains were formed throughout these regions in either addition, both 0.2% proof stress and tensile strength of the sub-
of the base materials. For a better illustration, the prior austenite zones 1 to 4 progressively increased by the gradual decrease in
structures of subzones 1, 4, and 6 are reconstructed from the IPF the prior austenite grain size in both of the materials, as shown
maps and presented in Fig. 9. Since grain growth is a in section 4.2.

10
S. Afkhami, V. Javaheri, M. Amraei et al. Materials & Design 213 (2022) 110336

Fig. 10. IQ data of subzones 1–7 (a- g) along with three examples of the normalized IQ values and their phase fraction analysis for (h) subzone 1, (i) subzone 4, and (j) subzone
5.

Regarding the IQ data, the IQ images and their subsequent IQ martensite due to its significantly lower martensite-start tempera-
curves varied by differing the applied thermal cycles on the differ- ture (Ms) than that of S960 (critical temperature of martensite for-
ent subzones, as shown in Section 3.2. The image qualities of the mation can be found in Figs. 15 and 16).
S1100 samples were analyzed in the same way as S960. The results
for both alloys are summarized in Fig. 11. In addition to the large 4.2. Correlations between the microstructural evolutions and
fraction of upper and lower types of bainite observed in both mechanical properties
alloys, S960 consisted of a considerable amount of high
transformation-temperature products of ferrite/granular bainite HAZ subzones of both S960 and S1100 consisted of areas with
compared to S1100 (which did not reveal any sign of ferrite forma- different microstructural features according to their peak temper-
tion). Instead, S1100 was comprised of a notable amount of fresh ature and cooling rate during the Gleeble-simulated welding
11
S. Afkhami, V. Javaheri, M. Amraei et al. Materials & Design 213 (2022) 110336

Fig. 11. Phase fraction analysis in detail according to the IQ evaluation for (a) S960 and (b) S1100.

Fig. 12. Local equivalent Mises strains of S960 subzones under the fixed load value of 25.8 kN: (a) subzone 1, (b) subzone 2, (c) subzone 3, (d) subzone 4, (e) subzone 5, (f)
subzone 6, (g) subzone 7, and (h) the base metal.

12
S. Afkhami, V. Javaheri, M. Amraei et al. Materials & Design 213 (2022) 110336

Fig. 13. Local equivalent Mises strains of S1100 subzones under the fixed load value of 36.4 kN: (a) subzone 1, (b) subzone 2, (c) subzone 3, (d) subzone 4, (e) subzone 5, (f)
subzone 6, (g) subzone 7, and (h) the base metal.

Fig. 14. (a) Hardness and (b) notch toughness values of the simulated HAZ subzones [11].

process. Different microstructures result in different mechanical load, except the base material, and this trend demonstrates
properties; thus, these subzones are expected to show different softening behavior. This behavior is crucial for designing
behavior under mechanical loads. Fig. 17 for S960 shows that all welded structures since it poses a potential for premature failure
the HAZ subzones had lower hardness values than the base mate- at the HAZ.
rial (the hardness ratios were smaller than one). However, the Contrary to S960, all the subzones of S1100, except number six,
drastic hardness decrease occurred in subzone 5, with an approxi- had hardness values higher than that of the base material (with
mately 30% drop due to ferrite/granular bainite formation. Next, hardness ratios greater than one in Fig. 17). Although the hardness
subzones 6 and 7 had the lowest hardness values due to tempered drop of subzone 6 was not significant compared to the results of
martensite and bainitic ferrite formation. The quasi-static tensile S960, the maximum local plastic strain in this subzone was signif-
behavior of the subzones also followed their hardness trend. Sub- icantly higher than that of others. According to Fig. 13, the maxi-
zone 5, which had the lowest hardness, revealed the lowest tensile mum plastic strains and their distribution for both subzone 6
strength. Hence, the highest local plastic strain under a fixed load is and the base material are relatively similar. This behavior contrasts
associated with this subzone, as shown in Fig. 12. The fixed load is the assumption on welded UHSS where softened HAZ is considered
the load at the tensile strength of the weakest specimen for each to govern the failure and dictate the overall performance of the
steel grade. All the specimens underwent plastic strains at this joint. As proved by Amraei et al. [7], considering the von Mises

13
S. Afkhami, V. Javaheri, M. Amraei et al. Materials & Design 213 (2022) 110336

Fig. 15. Theoretical CCT diagrams of S960 for the points with the highest (a) and lowest (b) full austenitization temperatures (subzones 1 and 4, respectively) plotted via
JMatPro.

criterion for the failure, the surrounding material of the soft region HV m ¼ 127 þ 949C wt% þ 27Siwt% þ 11Mnwt% þ 16Cr wt%
plays a vital role in creating a boundary condition that can enhance þ 8Niwt% þ 21log10 V R ð1Þ
the overall performance of the weldment. The size of the soft
region is also of importance preventing the plastic strains from
local concentration and piling up. Hence, a firm boundary condi- HV b ¼ 323 þ 185C wt% þ 330Siwt% þ 153Mnwt% þ 144Cr wt%
tion surrounding a narrow softened subzone can result in failure þ 191Mowt% þ 65Nwt% þ log10 V R ð89 þ 53C wt%
outside the soft area, i.e., the base material. In welded S1100, sub-  55Siwt%  22Mnwt%  20Cr wt%  33Mowt%  10Niwt% Þ ð2Þ
zones 1–5 have tensile and hardness values higher than the base
material, and only subzone 6 shows a minor softening. Thus, the
HV f ¼ 42 þ 223C wt% þ 53Siwt% þ 30Mnwt% þ 7Cr wt%
final failure in welded S1100 typically occurs at the base material
[2]. þ 19Mowt% þ 12:6Niwt%
The hardness value of carbon steel can be estimated after its þ log10 V R ð10   19Siwt% þ 8Cr wt% þ 4Niwt% þ 130V wt% Þ
continuous cooling based on the chemistry, microstructure, and
ð3Þ
thermal condition of the material; and, there are various models
available for this purpose [31–33]. In this section, two models, a where HVm, HVb, and HVf are the Vickers hardness values of
regression type and a best-fit one, are tested against the experi- martensite, bainite, and ferrite, respectively, and VR is the cooling
mental data. The regression model developed by Maynier et al. rate after austenitization (°C/h). The following linear rule of mix-
[32] can be expressed as follows: tures gives the final hardness of steel:
14
S. Afkhami, V. Javaheri, M. Amraei et al. Materials & Design 213 (2022) 110336

Fig. 16. Theoretical CCT diagrams of S1100 for subzones 1 (a) and 4 (b) plotted via JMatPro.

Fig. 17. Hardness fluctuations for different HAZ subzones [11].

15
S. Afkhami, V. Javaheri, M. Amraei et al. Materials & Design 213 (2022) 110336

HV ¼ f f ðHV f Þ þ f b ðHV b Þ þ f m ðHV m Þ ð4Þ Regarding the notch toughness values, the results seemed con-
troversial at first sight. Typically, Charpy notch toughness and
where ff, fb, and fm are the fractions of ferrite, bainite, and
hardness have a reverse correlation with each other. However, in
martensite, respectively. In this study, fractions of granular bai-
this study, subzones with the minimum hardness also had the low-
nite were also accounted for in ff calculations due to the similar
est notch toughness values, as shown in Fig. 14. This behavior can
properties and characteristics of granular bainite and ferrite.
be attributed to the existence of tempered martensite and bainitic
The presence of lower bainite and martensite were also
ferrite in the softened areas. Although these microstructural fea-
accounted for together in fm calculations due to the same
tures have relatively low hardness values due to the existence of
reason.
ferrite, carbides in the tempered martensite precipitated along
The best-fit model, developed by Pohjonen et al. [33] for bearing
the bainitic ferrite boundaries and encouraged more brittle frac-
steel with 0.2 Cwt%, is as follows:
tures in the Charpy test. Thus, areas such as subzone 5 in S960
and 6 in S1100 suffered from the simultaneous decrease of hard-
HV ¼ f b ðC þ Alog10 V R Þ þ f m HV m ð5Þ ness and notch toughness values.
where C and HVm are the fitting parameters, and A is a material
dependant constant that can be calculated via Eq. (6): 5. Conclusions

A ¼ 89 þ 53C wt%  55Siwt%  22Mnwt%  10Niwt% Thermo-mechanical simulations using Gleeble 3800 were per-
formed to replicate seven HAZ subzones of S960 and S1100 struc-
 20Cr wt%  33Mowt%
tural steels. Then, the microstructural transformations, their
Similar to the regression model, granular bainite and lower bai- mechanisms, and the resultant microstructures were studied. The
nite contributions in the hardness values were accounted for in the correlations between the microstructure and mechanical proper-
fb and fm fractions, respectively. The fitted C and HVm values were ties were also comprehensively studied, as schematically shown
30 and 370 for S960, and 400 and 400 for S1100, respectively. The in Fig. 19, based on the peak temperatures and cooling rates expe-
effects of austenite grain size and homogeneity are neglected in rienced during the welding process. According to the results, the
both models since they do not consider austenitizing temperature following conclusions are drawn:
and heating time in their calculations.
Fig. 18 shows comparisons of the experimental results pre-
Depending on the distance from the fusion line and its subse-
sented in Section 3.3 with the predictions made by the models. quent peak temperature, HAZ subzones of UHSS with properties
According to the results and comparisons, different thermal cycles different from their base metals can fit into three groups: fully
of the different subzones lead to different microstructures and, austenitized (Tpeak > A3), partially austenitized (A1 < Tpeak < A3),
consequently, different hardness levels. Pohjonen’s model showed and tempered (Tpeak < A1).
slightly better results than the other model; however, both models
In the fully austenitized subzones, the prior austenite grain size
deviated from the experimental measurements regarding the sub- decreased by increasing the distance from the fusion line.
zones with Tpeak between A1 and A3 or lower than A1. Despite the
Decrease in the contents of lower bainite and martensite, as
deviations, both models seem to be equally acceptable for the prac- hard microstructural mixtures, and their replacement by softer
tical use to detect the softened HAZ of either welded structural mixtures of upper bainite, granular bainite, and ferrite resulted
steels and UHSS. in soft spots in the HAZ, typically known as softened HAZ.

Fig. 18. The calculated hardness values for the different subzones along with the average value from the experimental measurements.

16
S. Afkhami, V. Javaheri, M. Amraei et al. Materials & Design 213 (2022) 110336

Fig. 19. Schematic illustration of the prior austenite grain structure changes at the different subzones for both alloys and the resulted hardness value based on the final
microstructure (an example of the final microstructures is illustrated for subzone 1).

According to the CCT diagrams, hindering and avoiding ferrite (improving the boundary condition). Finally, the simultaneous
formation in UHSS can preserve the mechanical properties after decrease in toughness and hardness in the softened HAZ might
welding, e.g., the case of S1100. raise another concern regarding welded UHSS, which requires fur-

Analyzing EBSD-IQ data from UHSS with complex microstruc- ther research, especially in cryogenic applications. Regarding this
tures proved to be a relatively reliable approach to quantita- matter, thorough investigations similar to the current study but
tively identifying their microstructures. in cryogenic conditions seem vital to approve the reliability of

There was a good agreement between the hardness values and welded UHSS on a broader domain of industrial applications.
local strains of the subzones. Regarding the notch toughness, Data Availability Statement
this property reached its minimum level in the softest subzones. The raw/processed data required to reproduce these findings
The controversial simultaneous decrease of hardness and can be obtained upon request from the corresponding author.
toughness in such areas was attributed to the tempered
martensite and bainite in these regions. Declaration of Competing Interest

Hardness models proposed by Maynier et al. [1] and Phjonen
et al. [2] effectively predicted the locations of the softened spots The authors declare that they have no known competing finan-
in the HAZs of UHSS. However, there were slight fluctuations cial interests or personal relationships that could have appeared
between the actual and predicted values at the regions where to influence the work reported in this paper.
peak temperatures were below A3.
Acknowledgment
In conclusion, although hardness decrease seems to be a draw-
back for certain welded UHSS grades, it can be hindered by delay- The authors wish to thank SSAB Europe Co. for providing the
ing ferrite formation in these steels. Modifications in chemical base materials used in this research. The help and support of Mr.
composition and initial microstructure by adjusting the manufac- Matti Koskimäki and Prof. Jari Larkiola for managing the experi-
turing process of steel might assist in achieving such a purpose, mental procedures, Mr. Juha Uusitalo in performing the Gleeble
similar to S1100. In addition, it is possible to minimize the adverse simulations, and Mr. Toni Väkiparta in conducting the SEM are
effect of the softened zone on the overall performance of welded highly appreciated. The authors also extend special thanks for
joints by reducing the size of the subzone and restricting it the technical support provided by the staff members at the
between its adjacent regions with better mechanical properties laboratories of LUT Steel Structures and Welding Technology.
17
S. Afkhami, V. Javaheri, M. Amraei et al. Materials & Design 213 (2022) 110336

Vahid Javaheri would also like to thank Jenny and Antti Wihuri [16] H. Jiao, X.L. Zhao, A. Lau, Hardness and compressive capacity of longitudinally
welded very high strength steel tubes, J. Constr. Steel Res. 114 (2015) 405–416,
Foundation for the personal financial support.
https://doi.org/10.1016/j.jcsr.2015.09.008.
[17] K. Wallin, S. Pallaspuro, I. Valkonen, P. Karjalainen-Roikonen, P. Suikkanen,
References Fracture properties of high performance steels and their welds, Eng. Fract.
Mech. 135 (2015) 219–231, https://doi.org/10.1016/j.
[1] B.A. Burgan, M.R. Sansom, Sustainable steel construction, J. Constr. Steel Res. engfracmech.2015.01.007.
62 (11) (2006) 1178–1183, https://doi.org/10.1016/j.jcsr.2006.06.029. [18] S. Pallaspuro, H. Yu, A. Kisko, D. Porter, Z. Zhang, Fracture toughness of
[2] M. Amraei, A. Ahola, S. Afkhami, T. Björk, A. Heidarpour, X.L. Zhao, Effects of hydrogen charged as-quenched ultra-high-strength steels at low
heat input on the mechanical properties of butt-welded high and ultra-high temperatures, Mater. Sci. Eng. A. 688 (2017) 190–201, https://doi.org/
strength steels, Eng. Struct. 198 (2019), https://doi.org/10.1016/j. 10.1016/j.msea.2017.02.007.
engstruct.2019.109460 109460. [19] L. Tong, L. Niu, S. Jing, L. Ai, X.L. Zhao, Low temperature impact toughness of
[3] S. Afkhami, T. Björk, J. Larkiola, Weldability of cold-formed high strength and high strength structural steel, Thin-Walled Struct. 132 (2018) 410–420,
ultra-high strength steels, J. Constr. Steel Res. 158 (2019) 86–98, https://doi. https://doi.org/10.1016/j.tws.2018.09.009.
org/10.1016/j.jcsr.2019.03.017. [20] H. Bhadeshia, R. Honeycombe, Heat Treatment of Steels: Hardenability, in:
[4] S. Mandal, N.K. Tewary, S.K. Ghosh, D. Chakrabarti, S. Chatterjee, Thermo- Steels Microstruct. Prop., Elsevier, 2017: pp. 217–236. 10.1016/b978-0-08-
mechanically controlled processed ultrahigh strength steel: Microstructure, 100270-4.00008-1.
texture and mechanical properties, Mater. Sci. Eng. A. 663 (2016) 126–140, [21] V. Javaheri, A. Pohjonen, J.I. Asperheim, D. Ivanov, D. Porter, Physically based
https://doi.org/10.1016/j.msea.2016.03.127. modeling, characterization and design of an induction hardening process for a
[5] S. K., R. Kawall, Ultrahigh Strength Steel: Development of Mechanical new slurry pipeline steel, Mater. Des. 182 (2019), https://doi.org/10.1016/
Properties Through Controlled Cooling, in: Heat Transf. - Eng. Appl., InTech, j.matdes.2019.108047 108047.
2011. 10.5772/26514. [22] V. Javaheri, S. Pallaspuro, A. Kaijalainen, S. Sadeghpour, J. Kömi, D. Porter,
[6] S. Keeler, M. Kimchi, P.J. Mooney, Advanced High-Strength Steels Application Promising bending properties of a new as-rolled medium-carbon steel
Guidelines Version 6 (2017). https://www.worldautosteel.org/projects/ achieved with furnace-cooled bainitic microstructures, Mater. Sci. Eng. A.
advanced-high-strength-steel-application-guidelines/. 796 (2020), https://doi.org/10.1016/j.msea.2020.140011.
[7] M. Amraei, M. Dabiri, T. Björk, T. Skriko, Effects of Workshop Fabrication [23] H. Tervo, A. Kaijalainen, V. Javaheri, S. Kolli, T. Alatarvas, S. Anttila, J. Kömi,
Processes on the Deformation Capacity of S960 Ultra-high Strength Steel, J. Characterization of coarse-grained heat-affected zones in al and ti-deoxidized
Manuf. Sci. Eng. Trans. ASME. 138 (2016), https://doi.org/10.1115/1.4033930. offshore steels, Metals (Basel). 10 (2020) 1–18, https://doi.org/10.3390/
[8] F. Farrokhi, J. Siltanen, A. Salminen, Fiber Laser Welding of Direct-Quenched met10081096.
Ultrahigh Strength Steels: Evaluation of Hardness, Tensile Strength, and [24] A.J. DeArdo, C.I. Garcia, K. Cho, M. Hua, New method of characterizing and
Toughness Properties at Subzero Temperatures, J. Manuf. Sci. Eng. Trans. quantifying complex microstructures in steels, Mater. Manuf. Process. 25 (1-3)
ASME. 137 (2015), https://doi.org/10.1115/1.4030177. (2010) 33–40, https://doi.org/10.1080/10426910903143415.
[9] W. Guo, D. Crowther, J.A. Francis, A. Thompson, Z. Liu, L. Li, Microstructure and [25] J. Wu, P.J. Wray, C.I. Garcia, M. Hua, A.J. Deardo, Image quality analysis: A new
mechanical properties of laser welded S960 high strength steel, Mater. Des. 85 method of characterizing microstructures, ISIJ Int. 45 (2005) 254–262, https://
(2015) 534–548, https://doi.org/10.1016/j.matdes.2015.07.037. doi.org/10.2355/isijinternational.45.254.
[10] W. Guo, L. Li, D. Crowther, S. Dong, J.A. Francis, A. Thompson, Laser welding of [26] ASTM Int. (2018) 1–26, https://doi.org/10.1520/E0023-18.
high strength steels (S960 and S700) with medium thickness, J. Laser Appl. 28 [27] A. Navarro-López, J. Hidalgo, J. Sietsma, M.J. Santofimia, Characterization of
(2016), https://doi.org/10.2351/1.4944100 022425. bainitic/martensitic structures formed in isothermal treatments below the Ms
[11] M. Amraei, S. Afkhami, V. Javaheri, J. Larkiola, T. Skriko, T. Björk, X.L. Zhao, temperature, Mater. Charact. 128 (2017) 248–256, https://doi.org/10.1016/
Mechanical properties and microstructural evaluation of the heat-affected j.matchar.2017.04.007.
zone in ultra-high strength steels, Thin-Walled Struct. 157 (2020), https://doi. [28] S.I. Wright, M.M. Nowell, EBSD Image Quality Mapping, Microsc. Microanal. 12
org/10.1016/j.tws.2020.107072 107072. (01) (2006) 72–84, https://doi.org/10.1017/S1431927606060090.
[12] J. Brabec, Š. Ježek, L. Beneš, A. Kříž, P. Majrich, Suitability Confirmation for [29] X. Tao, A. Eades, Alternatives to Image Quality (IQ) Mapping in EBSD, Microsc.
Welding Ultra-High Strength Steel S1100QL Using the RapidWeld Method, Microanal. 8 (S02) (2002) 692–693, https://doi.org/10.1017/
Manuf. Technol. 21 (2021) 29–36. 10.21062/mft.2021.014. S1431927602106465.
[13] A. Su, Y. Liang, O. Zhao, Experimental and numerical studies of S960 ultra-high [30] A.J. DeArdo, M.J. Hua, K.G. Cho, C.I. Garcia, On strength of microalloyed steels:
strength steel welded I-section columns, Thin-Walled Struct. 159 (2021), an interpretive review, Mater. Sci. Technol. 25 (9) (2009) 1074–1082, https://
https://doi.org/10.1016/j.tws.2020.107166 107166. doi.org/10.1179/174328409X455233.
[14] A.K. Pramanick, H. Das, J.W. Lee, Y. Jung, H.H. Cho, S.T. Hong, M. Shome, A.K. [31] J. Trzaska, Calculation of the steel hardness after continuous cooling, Arch.
Pramanick, Texture analysis and joint performance of laser-welded similar and Mater. Sci. Eng. 61 (2013) 87–92.
dissimilar dual-phase and complex-phase ultra-high-strength steels, Mater. [32] P. Maynier, B. Jungmann, J. Dollet, Creusot-Loire system for the prediction of
Charact. 174 (2021), https://doi.org/10.1016/j.matchar.2021.111035 111035. the mechanical properties of low alloy steel products, in, Hardenability
[15] T. Skriko, K. Lipiäinen, A. Ahola, H. Mettänen, T. Björk, Fatigue strength of Concepts with Appl. to Steel (1997) 518–545.
longitudinal load-carrying welds in beams made of ultra-high-strength steel, J. [33] A. Pohjonen, M. Somani, D. Porter, Modelling of austenite transformation along
Constr. Steel Res. 179 (2021), https://doi.org/10.1016/j.jcsr.2021.106563 arbitrary cooling paths, Comput. Mater. Sci. 150 (2018) 244–251, https://doi.
106563. org/10.1016/j.commatsci.2018.03.052.

18

View publication stats