You are on page 1of 15

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 7 ( 2 0 2 2 ) 6 5 1 e6 6 5

Available online at www.sciencedirect.com

ScienceDirect

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

An FeeNieCreH interatomic potential and


predictions of hydrogen-affected stacking fault
energies in austenitic stainless steels

X.W. Zhou a,*, C. Nowak a, R.S. Skelton a, M.E. Foster a, J.A. Ronevich a,
C. San Marchi a, R.B. Sills b
a
Sandia National Laboratories, Livermore, CA, 94550, USA
b
Department of Materials Science and Engineering, Rutgers University, Piscataway, NJ, 08854, USA

highlights

 A quaternary FeeNieCreH potential has been developed.


 Stacking fault energies (SFE) in alloys is a distribution, not a single value.
 Hydrogen reduces mean stacking fault energy.
 The reduced stacking fault energy may lead to hydrogen embrittlement.

article info abstract

Article history: While FeeNieCr austenitic stainless steels exhibit relatively good resistance to hydrogen
Received 22 February 2021 embrittlement, they still suffer from significant degradation of ductility, fatigue and frac-
Received in revised form ture properties in gaseous hydrogen environments. Experimental studies in the literature
22 June 2021 suggest that hydrogen reduces stacking fault energy in austenitic stainless steels. This
Accepted 30 September 2021 phenomenon causes a large separation of partial dislocations and lower propensity for
Available online 30 October 2021 cross-slip. Whereas lower stacking fault energy does not correlate well with loss of ductility
in the absence of hydrogen, lower stacking fault energy trends toward greater loss of
Keywords: ductility when hydrogen is present. Calculations of stacking fault energy are challenging
Hydrogen embrittlement for austenitic stainless steels. One main issue is that in alloys, stacking fault energy is not a
FeeNieCr stainless Steels single value but rather varies depending on local composition. Herein, we first report an Fe
Molecular dynamics eNieCreH quaternary interatomic potential and then use this potential to perform time-
Interatomic potential averaged molecular dynamics simulations to calculate stacking fault energies for tens of
Stacking faults thousands of realizations of local compositions for selected stainless steels alloys with and
Thermodynamics and kinetics without internal hydrogen. From statistical analyses, our results suggest that hydrogen
reduces stacking fault energy, which likely impacts deformation mechanisms of FeeNieCr
austenitic stainless steels when exposed to hydrogen environments. We then perform
validation MD simulation tests to show that hydrogen indeed statistically increases the
stacking fault widths due to statistically reduced stacking fault energies.
Published by Elsevier Ltd on behalf of Hydrogen Energy Publications LLC.

* Corresponding author.
E-mail address: xzhou@sandia.gov (X.W. Zhou).
https://doi.org/10.1016/j.ijhydene.2021.09.261
0360-3199/Published by Elsevier Ltd on behalf of Hydrogen Energy Publications LLC.
652 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 7 ( 2 0 2 2 ) 6 5 1 e6 6 5

14e50%. One problem is that SFE in local regions of a random


Introduction alloy does not take a single value but varies depending on local
composition even the overall composition of the alloy is fixed.
FeeNieCr austenitic stainless steels are important for This becomes more complex in practice because segregation
hydrogen energy applications because they are relatively of metal elements occurs depending on processing conditions
resistant to hydrogen embrittlement in comparison to other [39e41], which also changes local properties such as austenite
steels. Hydrogen compatibility, however, is still a major stability and fracture resistance. Typically, when local
concern. It is known that austenitic stainless steels form composition changes, the local properties can be represented
planar slip bands associated with dislocation slip, deforma- by a distribution [42]. For simulations, such a distribution can
tion twinning, and martensite phase transformation during be studied through thousands of realizations of the alloy
deformation [1e3]. These slip bands have been implicated in configurations using different random number seeds for the
limited ductility in the presence of internal hydrogen [3,4] same overall composition. Molecular dynamics (MD) simula-
because they result in local stress concentrations where tions can be applied to thousands of configurations but re-
damage initiates in the form of voids and cracks [5e7]. Prior to quires a high-fidelity FeeNieCreH interatomic potential that
straining, high-concentration of internal hydrogen promotes does not yet exist in the literature. The objective of the present
the formation of stacking faults [8]. Upon straining, hydrogen work is twofold: (1) report an FeeNieCreH interatomic po-
increases the density of the slip bands, and promotes the tential we developed recently and provide its characteristics;
formation of ε-martensite at the expense of deformation and (2) use this potential to study the effects of hydrogen on
twinning in the slip bands [8e10]. In the metastable alloys, a0 - the distribution of SFE of typical austenitic stainless steel
martensite may also form at the intersection of ε-martensite compositions. The implications of the predicted SFE on
bands at large strains (~20%) [11]. hydrogen embrittlement are discussed.
Hydrogen embrittlement has been attributed to the influ-
ence of hydrogen on the stress fields [12,13] and mobility of
dislocations [14e16]. On the other hand, lower stacking fault
energy (SFE) correlates with greater separation distance be- Interatomic potential
tween partial dislocations, which promotes slip band forma-
tion by suppressing cross-slip. Furthermore, both deformation The FeeNieCreH embedded-atom method (EAM) potential
twinning and ε-martensite are also related to stacking faults. used in this paper was constructed based on the FeeNieCr
SFE values greater than approximately 40 mJ/m2 produce potential published previously [43]. This FeeNieCr potential
limited partial dislocation separation, whereas SFE values simultaneously satisfies four criteria important to the me-
between 40 mJ/m2 and 20 mJ/m2 result in increasing separa- chanical properties of materials: (1) enable stable MD simu-
tion of the partial dislocation pairs, producing planar slip and lations of both austenitic and ferritic stainless-steel alloys; (2)
mechanical twinning [17]. SFE values below approximately capture closely the experimental stacking fault energies; (3)
18 mJ/m2 result in strain-induced hexagonal-close-packed accurately reproduce the experimental alloy elastic constants;
(hcp) ε-martensite and a0 -martensite formation [17,18]. As a and (4) give reasonable trends of energies and volumes for
result of these trends and observations of hydrogen effects of various compositions. Note that H was added to this FeeNieCr
planarity of slip, correlations between SFE and hydrogen potential in a previous effort [44]. While the previous
embrittlement have been explored [19e21]. One means to FeeNieCreH potential possesses many desired properties for
quantify hydrogen embrittlement [22e24] is through mea- H, it significantly overestimates the hydrogen-vacancy inter-
surement of the relative reduction of area (RRA), i.e., the action energies as compared to the corresponding density-
reduction of area (on the broken cross section) in hydrogen functional theory (DFT) values. Here we improve over the
normalized by the reduction of area in air. In general, SFE previous FeeNieCreH potential. This improvement does not
values greater than approximately 40 mJ/m2 result in RRA affect the metal properties in the absence of hydrogen but
>80% [21]. For materials with SFE values between 40 mJ/m2 enables the hydrogen properties to match better our DFT re-
and 20 mJ/m2, RRA can vary over a wide range: between 20% sults; see Appendix A for computational details. The func-
and 80% [21]. For steels that have an increasing propensity for tional forms and parameters for the improved EAM potential
planar slip with decreasing SFE values between 20 and 40 mJ/ are listed in Appendix B with the electronic version of the
m2, hydrogen exacerbates the planar localization on a reduced potential given in the supplemental material. Here we discuss
number of slip planes and reduces macroscopic ductility the characteristics of the potential in regard to the hydrogen
[1,13,14,17,18,25]. Because of these correspondences, SFE has properties that are most important for capturing hydrogen
been used to screen materials for hydrogen compatibility [21]. embrittlement. These properties are:
Experimental and first principles calculations have been
used to study SFE of austenitic stainless steels [26e30]. These  hydrogen interstitial site, as this determines how hydrogen
studies indicate that SFE increases with temperature [26] and interacts with other atoms.
Ni composition [26,27]. Experiments also suggest that  swelling volume, as this contributes to hydrogen-
hydrogen reduces the stacking fault energy [31,32]. Several dislocation interactions.
studies have attempted to measure the effect of hydrogen on  diffusion energy barrier, as this influences hydrogen
SFE either directly though in-situ and ex-situ TEM [33e37], or transport kinetics.
indirectly through X-ray diffraction [38]. These studies report  relative hydrogen insertion energies with respect to local
either negligible reductions of SFE or reductions of at most composition, as this determines hydrogen distribution.
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 7 ( 2 0 2 2 ) 6 5 1 e6 6 5 653

 hydrogen-vacancy interaction energy, as this is relevant to


hydrogen induced crack formation.

In addition, we also consider the hydrogen-self interstitial


atom (SIA) interaction energy. This property is not important
under conventional conditions where no SIAs is encountered,
but it is important under irradiation conditions where signif-
icant SIA concentrations form [45]. Capturing the hydrogen-
SIA interaction energy, therefore, allows our potential to be
used to study hydrogen transport for nuclear applications.
Based on our FeeNieCreH potential, molecular statics (MS)
simulations indicate that H prefers to occupy the octahedral
interstitial site in the face-centered-cubic (fcc) lattice of all fcc
elemental (Fe, Ni) and alloy (FeeNieCr) systems, in agreement
with our DFT results. The hydrogen-vacancy and hydrogen-
SIA interaction energies, EH-Va and EH-SIA, were calculated for
fcc Fe, Ni, Cr. The swelling volumes U were calculated for a Fig. 1 e Arrhenius plot of hydrogen diffusion in the
hydrogen atom in fcc Fe, fcc Ni, and a local Cr octahedron in Fe0.70Ni0.11Cr0.19 obtained from statistically-averaged MD
fcc Fe. In addition, the nudged elastic band method [46e48] simulations.
was used to calculate the diffusion energy barriers (identified
to be between octahedral and tetrahedral interstitial sites), Q,
experimental value of 0.51 eV (some experimental data are
for a hydrogen atom in fcc Fe, fcc Ni, and a local Cr octahedron
higher [56]) listed in Table 1.
in fcc Fe. The calculated results are reported in Table 1, which
We also determined the hydrogen insertion energies at
shows that our potential is in good agreement with the DFT
different octahedral locations of an alloy's lattice. To study
and experimental results. Note, a local Cr octahedron in fcc Fe,
this, a 512 atom fcc Fe0.70Ni0.10Cr0.20 cell was created, and a
rather than fcc Cr, was necessary because the lattice distorted
hydrogen atom was sequentially inserted into each of the
considerably when modeling a hydrogen interstitial in pure
octahedral interstitial sites. We found that the hydrogen
fcc Cr with DFT. Similarly, the DFT hydrogen-vacancy and
insertion energy increases with increasing Ni and decreasing
hydrogen-SIA interaction energies in fcc Cr are omitted from
Cr nearest neighbor counts. This was concluded by analyzing
Table 1 because the stable phase of Cr is bcc, not fcc.
the change in energy with respect to the composition of the
Note that the Fe0.70Ni0.11Cr0.19 alloy listed in Table 1 has
six metal atoms at the corners of octahedrons. Since Ni and Cr
similar Fe, Ni, Cr compositions to 304L steel although the latter
have opposite effects, we have plotted the relative insertion
also contains other elements including C, N, Si, and Mn. The
energies, using DFT and our potential, for octahedrons
hydrogen diffusion barrier in Fe0.70Ni0.11Cr0.19 cannot be
composed of only FeeNi and FeeCr in Fig. 2. This figure clearly
calculated using the nudged elastic method [46e48] because
illustrates the effect of the local chemical environment and
there is no unique diffusion path in an alloy. Therefore,
demonstrates that hydrogen prefers to be near Cr. Notably,
statistically-averaged MD simulations [51e55] were per-
our potential is in qualitative agreement with DFT results
formed in order to calculate an Arrhenius plot of hydrogen
obtained using the locally self-consistent Green's function
diffusion and the results are shown in Fig. 1. The diffusion
(LSGF) method (see Appendix A for details). Based on these
barrier, corresponding to the slope of the line, is determined to
results, our potential should be able to capture hydrogen in-
be 0.63 eV which is in relatively good agreement with the
teractions with metal atoms such as Fe, Ni, and Cr and with
lattice defects such as dislocations, vacancies, and metal
interstitials.

Table 1 e Swelling volume U (A3), diffusion energy barrier


Q (eV), and hydrogen-vacancy and hydrogen-SIA
interaction energies, EH-Va and EH-SIA (eV). Stacking fault energy calculation methods
Property Material MS DFT Exp.
Face-centered-cubic computational crystals containing 12
U Fe 2.15 2.16 e
(112) planes in the x direction, 27 (111) planes in the y direc-
Ni 2.22 2.23 e
Cr 1.93 1.95 e tion, and 6 (110) planes in the z direction were used. The
Q Fe 0.60 0.62 0.46 [49] systems contain 1296 metal atoms. The desired alloy was
Ni 0.39 0.41 0.41 [50] created by randomly converting the metal atoms to Fe, Ni, Cr
Cr 0.65 0.67 e based on their compositions. As a result, the overall compo-
Fe0.70Ni0.11Cr0.19 0.63 e 0.51 [6]
sitions of the entire computational cell match the desired
EH-Va Fe 0.33 0.36 e
values, say atomic fractions xFe ¼ 0.70, xNi ¼ 0.11, and
Ni 0.31 0.22 e
Cr 0.22 e e xCr ¼ 0.19 for alloy Fe0.70Ni0.11Cr0.19. However, the local com-
EH-SIA Fe 0.22 0.57 e positions can differ significantly depending on the random
Ni 0.06 0.06 e number seeds. Since no vacancies are present in our systems
Cr 0.16 e e to enable substitutional diffusion during our MD simulations,
654 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 7 ( 2 0 2 2 ) 6 5 1 e6 6 5

Fig. 2 e Relative hydrogen insertion energies in FeeNi and FeeCr octahedrons in an fcc Fe0.70Ni0.10Cr0.20 lattice as obtained
from (a) DFT and (b) EAM potential based molecular statics calculations.

this local composition profile remains fixed once created. The


desired atomic fraction (xH) of hydrogen atoms was randomly
inserted into the octahedral interstitial sites. Since H atoms
move during MD simulations, each simulation can fully
sample a large number of possible H locations incorporating
the H segregation effects. To create two crystals of the same
local alloy composition profile with and without stacking
faults, a crystal without stacking faults was first created using
a random number seed to generate one realization of the
computational cell at the desired alloy composition. Three
stacking fault planes were then created by displacing the
crystal by the [112]/6 partial dislocation Burgers vector at three
equal-spaced locations along y, as shown in Fig. 3. Three
stacking fault planes are needed to maintain the periodic
boundary condition along y. Using periodic boundary condi-
tions and an NPT ensemble (number of atoms, pressure, and
temperature are constant), MD simulations were performed to
anneal both systems (with and without the stacking faults) for
a total of 5.5 ns. After ignoring the first 0.5 ns, time-averaged
energies were calculated for the remaining 5.0 ns to statisti-
cally account for various hydrogen locations that are
encountered. Counterintuitively, the time-averaged energies
also have much lower statistical uncertainty [57e59] than the
minimum energy obtained from molecular statics (MS) sim-
ulations. The SFE was finally determined from the energy
difference between the systems with and without the faults,
normalized by the total fault area.
Using the approach described above, 3600 different
random number seeds were used to obtain 3600 values of SFE
for the same overall alloy composition but different local alloy
composition profiles. These 3600 values are sufficient to
examine a meaningful SFE distribution. Such SFE distributions
were determined for eight different cases: for temperature
300 K, Fe0.70Ni0.11Cr0.19 and Fe0.70Ni0.15Cr0.15 alloy compositions
were studied along with three H concentrations xH ¼ 0.00,
0.01, 0.10; for temperature 1200 K, the Fe0.70Ni0.11Cr0.19 alloy
composition was studied along with two H concentrations
xH ¼ 0.00, 0.01. In total, we conducted 57600 MD simulations
(28800 simulations for systems with and without stacking
faults), resulting in 28800 determinations of stacking fault
energy. Since we expect a small temperature effect on SFE, we
compare results between an elevated 1200 K and 300 K in Fig. 3 e Illustration of our computational cell that includes
order to clearly reveal the trends. three stacking fault planes.
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 7 ( 2 0 2 2 ) 6 5 1 e6 6 5 655

Note that bulk hydrogen concentrations of xH ~0.01 can be and the average SFE of the Fe0.70Ni0.15Cr0.15 alloy is reduced by
achieved experimentally when FeeNieCr stainless steels are ~5.4% with xH ¼ 0.01 and by ~53.2% with xH ¼ 0.10. Note that
precharged with hydrogen [6,60]. While stainless steels will for both of these systems, increasing the hydrogen content by
never reach a bulk concentration of xH ¼ 0.1 in gaseous an order of magnitude roughly increases the hydrogen effect
hydrogen environments, H can segregate at dislocation cores by an order of magnitude. Furthermore, this result indicates
to give a high local concentration [6,61,62] and will also that Ni helps reduce the impact of hydrogen on reducing the
approach these concentrations near stress concentrations. SFE. This finding is in agreement with experimental evidence
Also note that the SFE distribution is sensitive to the total that hydrogen compatibility is improved with Ni content
stacking fault area defined by our computational cell and is [21e24,66,67]. At 1200 K, the SFE of the Fe0.70Ni0.11Cr0.19 alloy is
expected to become narrower as the area increases. The mean reduced by ~4.5% with xH ¼ 0.01. We also note that hydrogen
SFE is not dependent upon the cell dimension, however. In the increases the standard deviation (distribution width) of the
limit of the average SFE for an infinite area (i.e., ignore the observed SFE values. In fact, a significant fraction of the sim-
local effect), the distribution should approach a delta-function ulations with xH ¼ 0.1 exhibited negative SFE values; a negative
centered on the mean [43]. As will be shown below, the finite value indicates that the stacking fault is energetically favor-
area adopted in our model would require 1000 simulations to able over the bulk fcc phase, i.e., a stacking fault with spon-
define this mean SFE. taneously grow. As mentioned previously, the distribution
width may depend on the simulation cell size, but it seems to
be clear that internal hydrogen enhances the compositional
Stacking fault energy results sensitivity of defect structures in austenitic stainless steels.
To explore hydrogen effects on specific local compositions,
SFE distribution histograms and running averages were the change of stacking fault energy DE due to the addition of
calculated and the results are summarized in Fig. 4. Fig. 4 hydrogen is calculated for all realizations of the Fe0.70Ni0.11-
confirms that the SFE in stainless steel alloys is not a single Cr0.19 alloy at 300 K. The results are used to construct the
value but rather a distribution over a wide range for a finite distribution of DE in Fig. 5. Fig. 5 confirms that hydrogen can
simulation cell. Fig. 4 also shows that the running average of both increase (DE > 0) and decrease (DE < 0) stacking fault
the SFE converges to a plateau with increasing number of energy. However, the probability of decrease is higher than
simulations. As mentioned above, the mean is well achieved that of increase, and therefore the average stacking fault en-
after 1000 simulations. Interestingly, Fig. 4 suggests that the ergy is reduced by hydrogen. In particular, the 0.10 hydrogen
SFE distribution can be described by a normal distribution concentration can dramatically reduce the average SFE.
function. We emphasize that this distribution arises from the Separate analysis also confirms that low SFE in hydrogen-free
distribution of different local alloy compositions, and the austenite are more likely to have low SFE with hydrogen.
statistical effect of hydrogen population including any To further explore the hydrogen effects, we analyze one
hydrogen segregation to the stacking faults should be well realization of the Fe0.70Ni0.11Cr0.19 metallic system at 300 K.
averaged out in each time-averaged MD simulation. We also This analysis contains four configurations: with and without
note that since our alloys are random, our results do not ac- stacking faults, and with and without 0.10 hydrogen content.
count for the possibility of micro-segregation of Cr and Ni From the MD results of these configurations as described
solutes (e.g., short-range ordering that induces non- above, we get SFE values of 0.00396 and 0.00056 eV/ A2 (63 and
2
randomness in the distribution). 9 mJ/m ) respectively without and with the 0.10 H. This means
Comparing Fig. 4(a) and (b) indicates that increasing tem- that the 0.10 H reduces the SFE by 0.0034 eV/ A2 (54 mJ/m2),
perature increases SFE, in agreement with experimental ob- which is almost 86%. We repeat the MD simulations for the
servations [63,64]. More quantitively, experiments indicated four configurations except that this time we calculate time
that an increase in temperature by 300 K can cause the SFE of average energies and H concentrations for each of the 27 (111)
an Fe0.65Ni0.16Cr0.19 alloy to increase from 25 to 32 mJ/m2 [63], planes in the y direction (see Fig. 3). This allows us to divide
whereas in Fig. 4, an increase in temperature by 900 K causes the total H-induced reduction of SFE, DE, to the contribution
the SFE of an Fe0.70Ni0.11Cr0.19 alloy to increase from ~43.5 to from each of the (111) planes. The results of these contribu-
~50.3 mJ/m2. The magnitude of the SFE increase is similar, but tions are shown in Fig. 6 along with the time averaged H
our temperature increase is three times larger. However, the concentrations in the two Fe0.70Ni0.11Cr0.19-0.1H systems with
experimental SFE appears to reach plateau at ~325  C [63]. and without stacking faults. It can be seen that as expected, DE
Regardless, a reduced temperature sensitivity of our predicted becomes relatively negligible for planes far away from the
stacking fault energy is expected as our metallic bonding re- stacking fault. Also due to the local variation of metallic
mains unchanged at different temperatures whereas in ex- compositions, DE and hydrogen concentration vary from
periments the formation of short-range order and segregation plane to plane. Although planes 18 and 27 have large positive
can change the bonding depending on temperature [41]. DE values, many other planes have negative DE contributions.
Comparing Fig. 4(a) and (c) shows that increasing Ni content The sum of DE over all the planes equals 0.0034 eV/ A2 (54 mJ/
also increases SFE, again in agreement with experimental m2), matching the value calculated directly as discussed
observations [26,41,65]. Hydrogen clearly shifts the SFE dis- above. The hydrogen concentrations on planes 18 and 27 are
tribution to lower values, indicating a reduction in the mean low after stacking fault formation. Since planes 18 and 27
SFE with increased H content. bound stacking faults, this observation indicates that
At 300 K, the average SFE of the Fe0.70Ni0.11Cr0.19 alloy is hydrogen may not prefer the stacking fault plane as would be
reduced by ~6.9% with xH ¼ 0.01 and by ~71.7% with xH ¼ 0.10, envisioned intuitively. Furthermore, hydrogen-rich planes
656 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 7 ( 2 0 2 2 ) 6 5 1 e6 6 5

Fig. 4 e Histograms and running averages of stacking fault energy for (a) Fe0.70Ni0.11Cr0.19 at 300 K; (b) Fe0.70Ni0.11Cr0.19 at
1200 K; and (c) Fe0.70Ni0.15Cr0.15 at 300 K. Note that the mJ/m2 axis may involve some round-off errors because it is converted
from eV/A axis with the same tick points.

may not correlate to a low stacking fault energy because plane resulting from insertion of the stacking faults. For
planes 18 and 27 have high hydrogen concentrations prior to example, planes with large negative DE values tend to have
stacking fault formation but also large positive DE values. One lower hydrogen concentration prior to stacking fault forma-
remarkable finding in Fig. 6 (see inset) is that the DE contri- tion but higher hydrogen concentration after stacking fault
bution from each plane appears to be approximately linearly formation, such as planes 8, 17, 20, 24, 26. Many of these
related to the local change in hydrogen concentration at that planes are second nearest to the stacking fault plane, i.e., 8, 17,
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 7 ( 2 0 2 2 ) 6 5 1 e6 6 5 657

hydrogen effect on SFE is oversimplified. Our finding should


motivate further studies that may help understand why
hydrogen promotes the ε-martensite deformation mechanism
as opposed to the twinning deformation mechanism [5,8,10].

Molecular dynamics validation tests

To validate the SFE calculations presented above, direct MD


simulations were used to visualize stacking fault width be-
tween partial dislocations of an extended edge dislocation.
Fe0.70Ni0.11Cr0.19 alloys with and without xH ¼ 0.0045 hydrogen
were chosen for this study. The selected hydrogen concen-
tration is consistent with precharged experimental samples.
The intention of using this composition is not to compare with
Fig. 4, but rather to prove that even a low H concentration can
impact the stacking fault width. The system geometry is
illustrated in Fig. 7, the previous approach [68] was used to
Fig. 5 e Distribution of hydrogen-induced change of create an edge dislocation along the x direction on the middle
stacking fault energy in the Fe0.70Ni0.11Cr0.19 alloy at 300 K. plane along the y direction, and hydrogen atoms were initially
Note that the mJ/m2 axis may involve some round-off introduced near the dislocation.
errors because it is converted from eV/A  axis with the The crystal shown in Fig. 7 contains 36 (111) planes in the y
direction (~75 A), and 118 (110) planes in the z direction
same tick points.
(~302 A). Two different x dimensions, corresponding to two
different dislocation lengths, are explored. The long disloca-
26. We point out that the actual SFE should be dictated by all
plane contributions but not by the DE from one plane. How- tion contains 405 (112) planes in the x direction (~598 
A), and
ever, the complicated DE distribution among many different the short dislocation includes 21 (112) planes in the x direction
planes suggests that our previous view regarding the (~31 
A).

Fig. 6 e The change of stacking fault energy (DE) in each of the (111) planes along y due to the addition of 0.1 H to an
Fe0.70Ni0.11Cr0.19 alloy realization at 300 K. The two adjacent planes bounding the three stacking faults are indicated by the
six vertical red dash lines. Note that the mJ/m2 axis may involve some round-off errors because it is converted from eV/A 
axis with the same tick points. (For interpretation of the references to color in this figure legend, the reader is referred to the
Web version of this article.)
658 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 7 ( 2 0 2 2 ) 6 5 1 e6 6 5

Fig. 7 e MD Geometry for stacking fault width tests in Fe0.70Ni0.11Cr0.19 alloys.

MD simulations were performed under the periodic


boundary conditions in the x and z directions and a free
boundary condition in the y direction. Since the two partials
have opposite Burgers vector components in the x direction, a
shear stress of txy ¼ 500 MPa was applied to increase the
stacking fault width. This does not impact the study of the
effects of stacking fault energy on stacking fault width
knowing that stacking fault width l f (SFE - txyb)1; in other
words, the applied stress simply imposes a uniform increase
in stacking fault width which helps to visualize the stacking
fault variation induced by H. The shear stress txy was imple-
mented by applying corresponding forces to the top and bot-
tom one monolayer of atoms. Reflect walls were used at the
top and bottom surfaces to prevent hydrogen from escaping
the lattice.
MD simulations were performed for 1.8 ns where an NPT
ensemble was used for the bulk atoms and an NVE (number of
atoms, volume, and energy are constant) ensemble was used
for surface atoms. A high temperature of 800 K was used for
the simulations so that hydrogen atoms can migrate suffi-
ciently to average out statistical effects with respect to
hydrogen locations (we observe that hydrogen atoms diffuse
readily in these simulations). Again, the intention of using
800 K is not to compare with the 300 K data shown in Fig. 4, but
rather to help equilibrate the H population. To get statistical
information, 10 simulations with different random number
seeds were performed for each case. The final configurations
of all 10 replica simulations of the short dislocation both with
and without hydrogen are summarized in Fig. 8. Similar re-
sults for the long dislocation are shown in Fig. 9 using only one Fig. 8 e Configurations of all 10 replicas of the short
example of the 10 replicas. In Figs. 8 and 9, only hcp atoms dislocation cases obtained at time ¼ 1.8 ns. Only the hcp
(red) detected using adaptive common neighbor analysis [69] (red) atoms are shown to reveal the stacking faults, and the
are shown to reveal the stacking faults. dashed lines indicate the initial dislocation locations. (For
Fig. 8 indicates that statistically, the addition of 0.0045 H interpretation of the references to color in this figure
increases the stacking fault width. Note that since the initial
legend, the reader is referred to the Web version of this
hydrogen population is in a high energy state, hydrogen
article.)
redistribution is faster than random walk. Separate MD sim-
ulations indicated that the peak hydrogen concentration at
the dislocation core reached a saturation value at 1.8 ns, Visual comparison of the stacking fault width of the long
meaning that the hydrogen was equilibrated about the dislo- dislocations shown in Fig. 9 is more challenging. Hence, we
cation. As a result, the observed increase in the stacking fault performed a numerical analysis of all 10 replica final config-
width is not an artifact of the initial hydrogen population. urations both with and without hydrogen. Specifically, we
Certainly, the increased stacking fault width agrees well with calculated minimum and maximum local stacking fault width
the reduced SFE results discussed above. along the dislocation lines. The ranges of stacking fault widths
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 7 ( 2 0 2 2 ) 6 5 1 e6 6 5 659

Fig. 9 e Final configurations of one replica example of the long dislocation cases. Only the hcp (red) atoms are shown to
reveal the stacking faults, and the dash line indicates the initial dislocation location. (For interpretation of the references to
color in this figure legend, the reader is referred to the Web version of this article.)

thus obtained are shown for the various replica simulations in


Fig. 10 in the sequence of increasing maximum local stacking
fault width for samples both with and without hydrogen. On
average, H increased the stacking fault width by 10.5%. The
increase of stacking fault width in local regions, however, can
be significantly larger. Fig. 10 indicates that for the ten replica
simulations with long dislocation lines, the minimum and
maximum stacking fault widths are all longer with hydrogen
than without if the one-to-one comparison is according to the
sorted sequence. This further supports the conclusion that
statistically hydrogen reduces local stacking fault energy and
increases local stacking fault width.
To make a connection between our SFE calculations and
our stacking fault width calculations, we assume that the SFE
varies linearly with hydrogen concentration (as suggested by
  Fig. 10 e Stacking fault range from 10 replicas of long
our results) according to SFEðxH Þ ¼ SFE0 þ dSFE
dxH xH where SFE0
dislocation simulations.

is the (average) SFE in the absence of hydrogen and SFEðxH Þ is


the (average) SFE at hydrogen concentration xH . For Fe0.70-
Ni0.11Cr0.19 using our SFE results at 1200 K in Fig. 4(b), we obtain the influence of the applied shear stress in this expression
1 dSFE
z 4.5. According to elasticity theory, the separation because we are mainly focused on changes in stacking fault
SFE0 dxH
width due to hydrogen). Assuming that the stiffness tensor cijkl
distance between the partial dislocations in an extended
dislocation varies as lðgÞ ¼ Aðcijkl Þ=SFE where Aðcijkl Þ is a is independent of hydrogen concentration, the relative change
in partial separation can be expressed as
function of the stiffness tensor cijkl [70] (note that we neglect
660 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 7 ( 2 0 2 2 ) 6 5 1 e6 6 5

xH ¼ 0.0045 at 800 K, we invert Eq. (2) and obtain a core binding


     1 energy of Ec z 0.1073 eV. Hence, according to Eq. (2) the core
Dl 1 dSFE 1 dSFE
¼ xHc 1 þ xHc (1) concentration enhancement factor xHc/xH in the experiments
l0 SFE0 dxH SFE0 dxH
of Pontini and Hermida with xH ¼ 0.000274 is predicted to be
where xHc is the hydrogen concentration in the dislocation ~3648 and ~69 at 77 and 293 K, respectively. Based on our re-
core (i.e., in the stacking fault). Given the value of Dl
l0
z 10.5% sults, we would expect an observed SFE change, DSFE ¼
SFE0
that we obtained in our dislocation simulations in addition to  
our estimate 1 dSFE
z 4.5, Eq. (1) indicates a core concen-
1
SFE0
dSFE
dxH
xHc , in the range of 8.5% to 100% between 293 and
SFE0 dxH

tration of xHc z 0.021, which is ~4.7  elevated above the global 77 K. Clearly this range is quite large, no doubt because we
concentration of 0.0045. This concentration enhancement have assumed the equilibrium distribution of H, which may
factor is consistent with the maximum concentration that we not have been attained experimentally (especially at lower
observe near the dislocation core in our simulations at about temperatures). Indeed the fact that their SFE change was in-
7.5 times above the global concentration. Note that Eq. (1) is dependent of temperature indicates that H atoms were not
based on constant SFE, isolated dislocation, and infinite me- redistributing appreciably during their measurements.
dium, Our dislocation is in a periodic film which may cause Nonetheless, we conclude that a 40% reduction in SFE at
different stacking fault width [59,71]. Nonetheless, this anal- dislocation cores is plausible. However, it must be emphasized
ysis confirms that the change in dislocation partial spacing that the SFE reduction in the bulk will be much smaller, on the
can be related to the H-induced changes in the SFE. order of 1% or less.
As a final note, we compare our SFE changes computed
here with measurements in the literature. Typically, SFE is
measured by examining the influence of hydrogen charging Conclusions
on the geometry of extended dislocations. For example, SFE
has been deduced from the transmission electron microscope We have developed an FeeNieCreH embedded atom method
measurements of the curvature of partial dislocations at potential. This potential is suitable for studying hydrogen ef-
extended dislocation nodes [72] or the separation distance fects on mechanical properties of stainless steels because it
between partials [73]. Bulk techniques, such as x-ray diffrac- captures the correct hydrogen interstitial site, swelling vol-
tion (XRD) [74], have also been applied, but are similarly ume, diffusion energy barrier, relative energies between
measuring changes in stacking faults in dislocation cores. hydrogen and various metal elements, and interaction en-
Hence, the observed changes in SFE correspond to the local H ergies between hydrogen and point defects (specifically va-
concentration at the dislocation core, xHc , rather than the bulk cancy and self-interstitial atom).
one. One issue with TEM-based measurements is that the H Using the FeeNieCreH potential, MD simulations were
concentration in the sample is highly uncertain due to diffu- performed to study the distribution of stacking fault energy
sion of H out of the thin TEM foil or due to interactions be- with and without hydrogen. We found that hydrogen locally
tween H2 molecules and the electron beam in the increases and decreases the stacking fault energy. However,
environmental TEM [75]. For example, Ferreira et al. [34] when considering the statistical effect of local composition in
measured a 19% change in SFE in an austenitic steel in the a bulk material, hydrogen is found to reduce the average
environmental TEM but could not estimate the hydrogen stacking fault energy of stainless steel, in agreement with
concentration. On the other hand, the background concen- experimental data. Counterintuitively, this mean stacking
tration in bulk specimens is well-defined. Pontini and Her- fault energy reduction occurs in the absence of hydrogen
mida [74] used XRD (calibrated against TEM) to measure SFE segregation to the stacking fault plane. Instead, the reduced
changes in a 304L stainless steel charged with a global H stacking fault energy is caused by the redistribution of nearby
concentration of xH ¼ 0.000274 (274 ppm H). From 77 K to 293 K hydrogen atoms before and after the stacking fault formation.
they observed a relatively constant decrease in SFE of about Although the equilibrium hydrogen concentration in stainless
40%. This was consistent with earlier measurements by steels exposed to gaseous hydrogen will generally be less than
Whiteman and Troiano [35]. At first glance, this seems to 0.01, local hydrogen concentration can be enhanced near de-
contradict with our SFE data in Fig. 4 which indicates a much fects, such as dislocation cores, which may affect deformation
smaller SFE reduction at the same background concentration. structures influenced by the SFE. Our analysis of partial
But when the concentration enhancement at the dislocation dislocation separation distances confirm this conclusion,
core is accounted for, the results are brought into better showing that enhanced hydrogen concentration at the dislo-
agreement. To account for segregation to the core, we note cation core does exacerbate hydrogen's influence on the SFE.
that the concentration enhancement factor at the dislocation Finally, our comparisons with experimental measurements of
core when the hydrogen distribution is at equilibrium is given by SFE changes due to hydrogen further validate the strong in-
Ref. [61]. fluence of H segregation to the dislocation core. These results
provide a plausible explanation of observed experimental
xHc 1
¼ (2) trends: the effects of hydrogen on ductility are sensitive to
xH xH þ ð1  xH ÞexpðEc =kB TÞ
hydrogen content, alloy composition and temperature. Gain-
where kB is Boltzmann's constant and Ec is the H binding en- ing insight into these types of fundamental processes will
ergy at the dislocation core (assuming it is a uniform constant motivate new considerations when trying to improve
for simplicity). Using our result above that xHc z 0.021 when hydrogen compatibility.
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 7 ( 2 0 2 2 ) 6 5 1 e6 6 5 661

octahedral sites. A lattice constant of 3.59 


A was used. The
Declaration of competing interest disordered local moment (DLM) method [80] was employed to
model the paramagnetic state of austenitic stainless steel. The
The authors declare that they have no known competing PBE total energies were used to calculate the relative H
financial interests or personal relationships that could have insertion energies.
appeared to influence the work reported in this paper.
B: Embedded-Atom Method Potential

Acknowledgements Our embedded-atom method potential is constructed


using the Finnis-Sinclair formalism [81], which is more gen-
Sandia National Laboratories is a multi-mission laboratory eral than the Daw-Baskes formalism [82]. The pair potential
managed and operated by National Technology and Engi- function 4IJ ðrÞ between species I and J separated by r is written
neering Solutions of Sandia, LLC., a wholly owned subsidiary as
of Honeywell International, Inc., for the U.S. Department of
Energy's National Nuclear Security Administration under 4IJ ðrÞ ¼ fc;IJ ðrÞ40IJ ðrÞ (B1)
contract DE-NA-0003525. The authors gratefully acknowledge Here 40IJ ðrÞ is the main function and fc;IJ ðrÞ is a cutoff func-
research support from the U.S. Department of Energy, Office of tion, which are defined respectively as
Energy Efficiency and Renewable Energy, Hydrogen and Fuel
    
Cell Technologies Office through the H-Mat program. The Eb;IJ fIJ ðrÞ r  r0;IJ r  r0;IJ
40IJ ðrÞ ¼ bIJ exp  aIJ  aIJ exp  bIJ
views expressed in the article do not necessarily represent the bIJ  aIJ r0;IJ r0;IJ
h i
views of the U.S. Department of Energy or the United States þεIJ exp  lIJ ðr  dIJ Þ 2
(B2)
Government.
8  
Appendices < 1 erfc 3:0157332ðr  rs;IJ Þ  0:9061938ðrc;IJ  rÞ ;
>
r < rc;IJ
fc;IJ ðrÞ ¼ 2 rc;IJ  rs;IJ
>
:
A: Density-Functional Theory Method 0; r  rc;IJ
(B3)
The hydrogen-vacancy (EH-Va) and hydrogen-SIA (EH-SIA) where Eb,IJ, aIJ, bIJ, r0,IJ, εIJ, lIJ, dIJ, rs,IJ, and rc,IJ are nine pair pa-
interaction energies, swelling volumes (U) and diffusion bar-
rameters. The electron density rJI ðrÞ produced by a species J at
riers (Q) reported in Table 1, were calculated using 32 atom
the site occupied by a species I that is located at a distance r
cubic fcc unit cells. These calculations were performed with
away from J is expressed as
the Vienna ab initio simulation package (VASP) [76,77] using
 
the Perdew-Burke-Ernzehof (PBE) exchange-correlation func- rJI ðrÞ ¼ rJ0;I exp gJI r fIJ ðrÞ (B4)
tional. Spin-polarization was used with an energy cut-off of
500 eV and a 5  5  5 k-point grid. Full geometry optimiza- Here rJ0,I and gJI are pair parameters, and the cutoff function
tions (cell and atoms positions) were performed until all forces fIJ ðrÞ has a similar form to Eq. (B3) but takes different pair pa-
were below 0.01 eV/ A. NEB calculations, using the climbing rameters rJsr,J, and rJcr,J:
image method [46e48], were performed for a H atom moving 8 2   3
>
> 3:0157332 r  rJsr;I  0:9061938 rJcr;I  r
from the octahedral to the tetrahedral site (5 images along the > 1
< erfc4 5; r<rJ
cr;I
path) to estimate the diffusion barriers. fIJ ðrÞ¼ 2 rJcr;I  rJsr;I
>
>
The DFT relative H insertion energies were determined >
:
0; r  rJcr;I
using the locally self-consistent LSGF method within the
KorringaeKohneRostoker approach using the atomic sphere (B5)
approximation (KKR-ASA) [78,79]. The calculations were per- For Fe, Ni, and Cr, the embedding energy function FI(r) is
formed on a random Fe70Ni10Cr20 alloy composed of 512 atom given as
with a single H atom; the H atom was placed in each of the 512

8X 3  k
>
> r
>
> fI;n;k 1 ; 0  r < rI;n
>
> rI;n
>
> k¼0
>
>
>
>  k
<X 3
r
FI ðrÞ ¼ f I;k  1 ; rI;n  r < rI;m (B6)
> rI;e
> k¼0
>
>
> 2sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 3
>
>  k  2
>
> X1
r r
>
>
>
: fI;e;k  1 þ fI;e;2 4 1 þ hI  1  15; rI;m  r
k¼0
rI;m rI;m
662 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 7 ( 2 0 2 2 ) 6 5 1 e6 6 5

where fI,n,k, fI,k (k ¼ 0e3), fI,e,k (k ¼ 0e2), hI, rI,n, rI,m, and rI,e are
all species dependent parameters. For H, a different embed-
ding energy function is used:

(  hH "   #
hH r þ 106  6
 hH ln rH;e þ 106 þ 1
FH ðrÞ ¼ fH;0   hH ln r þ 10  ; 0r (B7)
rH;e þ 106 hH

where fH,0, hH, and rH,e are species dependent parameters.


Eqs. (B1) e (B7) fully define our EAM potential. All parameters
needed for simulations are listed in Tables BI e BIII.

Table BI e Two-body parameters of pair potential functions (eV).


IJ FeFe FeNi FeCr FeH NiNi
Eb,IJ 0.3685388 0.3168913 0.4809498 0.3287460 0.2176650
aIJ 8.5291410 7.6368167 7.8770454 5.5447290 7.6448347
bIJ 4.9380853 4.4922451 4.7739668 2.9907669 3.8419747
r0,IJ 2.6955317 2.7368862 2.6708121 1.9729416 2.8049458
εIJ 0.0199993 0.0000000 0.0000000 0.0000000 0.0000000
lIJ 74.9391053 96.0000000 96.0000000 96.0000000 96.0000000
dIJ 2.9297363 0.5000000 0.5000000 0.5000000 0.5000000
rs,IJ 4.6950882 4.2039109 4.1824624 3.2660657 4.2897048
rc,IJ 5.6028271 5.5707079 5.2989661 4.3106536 5.4348366

IJ NiCr NiH CrCr CrH HH

Eb,IJ 0.5373532 0.4380634 0.3240388 0.2148044 0.1422920


aIJ 7.5422660 4.7732543 8.2235774 5.0239123 9.1793485
bIJ 4.4366270 2.7466384 4.7790441 3.4018087 4.5955694
r0,IJ 2.5306916 1.7106993 2.7394184 2.1515789 2.4918009
εIJ 0.0000000 0.0000000 0.0200000 0.0000000 0.0000000
lIJ 96.0000000 96.0000000 82.9887506 96.0000000 96.0000000
dIJ 0.5000000 0.5000000 2.9300698 0.5000000 0.5000000
rs,IJ 4.1824624 2.9258270 4.6810274 3.1782926 4.0213754
rc,IJ 5.5707078 4.4965077 5.7095242 4.4699949 4.7815426

Table BII e Two-body parameters of electron density


functions.
J Fe Fe Fe Fe
I Fe Ni Cr H

rJ0,I 167.4186326 167.4186326 167.4186326 7.9188736


gJI 1.8319540 1.8319540 1.8319540 0.9938832
rJsr,I 4.6950882 4.6950882 4.6950882 4.6950882
rJcr,I 5.6028271 5.6028271 5.6028271 5.6028271 Table BIII e One-body parameters for embedding energy
J Ni Ni Ni Ni
functions (eV).
I Fe Ni Cr H
I Fe Ni Cr H
rJ0,I 18.1732058 18.1732058 18.1732058 11.7245653
gJI 1.3697127 1.3697127 1.3697127 1.1190298 f I,n,0 1.4019206 2.5489097 1.4921488 e
rJsr,I 4.2897048 4.2897048 4.2897048 4.2897048 f I,n,1 0.0334217 0.0743739 0.1173299 e
rJcr,I 5.4348366 5.4348366 5.4348366 5.4348366 f I,n,2 0.3170381 0.2052888 0.5586135 e
f I,n,3 1.7523804 2.2692471 0.8162054 e
J Cr Cr Cr Cr
I Fe Ni Cr H fI,0 1.4007496 2.5555286 1.4985411 1.5461554
rJ0,I 474.9956012 474.9956012 474.9956012 32.3380099 fI,1 0.0000000 0.0000000 0.0000000 e
gJI 1.7445455 1.7445455 1.7445455 1.5359186 fI,2 0.0200526 0.2991878 0.6140190 e
rJsr,I 4.6810274 4.6810274 4.6810274 4.6810274 fI,3 1.37152534 0.0334469 0.2520912 e
rJcr,I 5.7095242 5.7095242 5.7095242 5.7095242 f I,e,0 1.3991776 2.5486840 1.4926530 e
J H H H H
f I,e,1 0.0496719 0.1058161 0.1267652 e
I Fe Ni Cr H
f I,e,2 0.2717334 0.1672253 0.2140316 e
rJ0,I 67.2069730 33.0130147 25.4091675 381.5725962 hI 3.8429379 4.9703125 6.0874554 1.2296172
gJI 1.0033850 0.8086450 0.4722319 2.6429194 rI,n 19.1202772 7.7311452 66.7380710 e
rJsr,I 2.0567001 2.0567001 2.0567001 2.0567001 rI,m 23.3692277 10.4597847 81.5687535 e
rJcr,I 5.7947109 5.7947109 5.7947109 5.7947109 rI,e 21.2447524 9.0954650 74.1534123 70.3013724
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 7 ( 2 0 2 2 ) 6 5 1 e6 6 5 663

references [20] Weber S, Martin M, Theisen W. Development of lean alloyed


austenitic stainless steels with reduced tendency to
hydrogen environment embrittlement. Mater Sci Forum
2012;706e709:1041.
[1] Birnbaum HK, Sofronis P. Hydrogen-induced localized
[21] Gibbs PJ, Hough PD, Thurmer K, Somerday BP, San
plasticity e a mechanism for hydrogen-related fracture.
Marchi C, Zimmerman JA. Stacking fault energy based alloy
Mater Sci Eng, A 1994;176:191.
screening for hydrogen compatibility. JOM 2020;72:1982.
[2] Gerold V, Karnthaler HP. On the origin of planer slip in f.c.c.
[22] Zhang L, Wen M, Imade M, Fukuyama S, Yokogawa K. Effect
alloys. Acta Metall 1989;37:2177.
of nickel equivalent on hydrogen gas embrittlement of
[3] Ulmer DG, Altstetter CJ. Hydrogen-induced strain
austenitic stainless steels based on type 316 at low
localization and failure of austenitic stainless-steels at high
temperatures. Acta Mater 2008;56:3414.
hydrogen concentrations. Acta Mater 1991;39:1237.
[23] Marchi CS, Michler T, Nibur KA, Somerday BP. On the
[4] Talonen J, Hanninen H. Formation of shear bands and strain-
physical differences between tensile testing of type 304 and
induced martensite during plastic deformation of metastable
316 austenitic stainless steels with internal hydrogen and in
austenitic stainless steels. Acta Mater 2007;55:6108.
external hydrogen. Int J Hydrogen Energy 2010;35:9736.
[5] Hatano M, Fujinami M, Arai K, Fujii H, Nagumo M. Hydrogen
[24] Marchi CS, Somerday BP, Tang X, Schiroky GH. Effects of
embrittlement of austenitic stainless steels revealed by
alloy composition and strain hardening on tensile fracture of
deformation microstructures and strain-induced creation of
hydrogen-precharged type 316 stainless steels. Int J
vacancies. Acta Mater 2014;67:342.
Hydrogen Energy 2008;33:889.
[6] San Marchi C, Somerday BP, Robinson SL. Permeability,
[25] Koyama M, Okazaki S, Sawaguchi T, Tsuzaki K. Hydrogen
solubility and diffusivity of hydrogen isotopes in stainless
embrittlement susceptibility of Fe-Mn binary alloys with
steels at high gas pressures. Int J Hydrogen Energy 2007;32:100.
high Mn content: effects of stable and metastable
[7] San Marchi C, Nibur KA, Balch DK, Somerday BP, Tang X,
ε-martensite, and Mn concentration. Metall Mater Trans A
Schiroky GH, Michler T. Hydrogen-assisted fracture of
2016;47:2656.
austenitic stainless steels. In: Somerday BP, Sofronis P,
[26] Vitos L, Korzhavyi PA, Johansson B. Evidence of large
Jones R, editors. Proceedings of the 2008 international
Magnetostructural effects in austenitic stainless steels. Phys
hydrogen conference, effects of hydrogen on materials.
Rev Lett 2006;96:117210.
Materials Park OH: ASM International); 2009. p. 88.
[27] Lu S, Hu QM, Johansson B, Vitos L. Stacking fault energies of
[8] Sabisch JEC, Sugar JD, Ronevich J, San Marchi C, Medlin DL.
Mn, Co, and Nb alloyed austenitic stainless steels. Acta Mater
Interrogating the effects of hydrogen on the behavior of
2011;59:5728.
planar deformation bands in austenitic stainless steel. Metall
[28] Lu J, Hultman L, Holmstrom E, Antonsson KH, Grehk M, Li W,
Mater Trans A 2021;52:1516.
Vitos L, Golpayegani A. Stacking fault energies in austenitic
[9] Olson GB, Cohen M. Kinetics of strain-induced martensite
stainless steels. Acta Mater 2016;111:39.
nucleation. Metall Trans A 1975;6:791.
[29] de Bellefon GM, van Duysen JC, Sridharan K. Composition-
[10] Narita N, Altstetter CJ, Birnbaum HK. Hydrogen-related
dependence of stacking fault energy in austenitic stainless
phase-transformations in austenitic stainless-steels. Metall
steels through linear regression with random intercepts. J
Trans A 1982;13:1355.
Nucl Mater 2017;492:227.
[11] San Marchi C, Ronevich JA, Sabisch JEC, Sugar JD, Medlin DL,
[30] Yang SW, Spruiell JE. Cold-worked state and annealing
Somerday BP. Effect of microstructural and environmental
behaviour of austenitic stainless steel. J Mater Sci 1982;17:677.
variables on ductility of austenitic stainless steels. Int J
[31] Barannikova SA, Nadezhkin MV, Mel’nichuk VA, Zuev LB.
Hydrogen Energy 2021;46:12338.
Tensile plastic strain localization in single crystals of
[12] Martin ML, Dadfarnia M, Nagao A, Wang S, Sofronis P.
austenite steel electrolytically saturated with hydrogen.
Enumeration of the hydrogen-enhanced localized plasticity
Tech Phys Lett 2011;37:793.
mechanism for hydrogen embrittlement in structural
[32] Kuprekova EI, Chumlyakov YI, Chernov IP. Dependence of
materials. Acta Mater 2019;165:734.
critical cleavage stresses as a function of orientation and
[13] Delafosse D. In: Gangloff RP, Somerday BP, editors. In Gaseous
temperature in single crystals of Fe-18%Cr-14%Ni-2%Mo
hydrogen embrittlement of materials in energy technologies.
austenitic stainless steel containing hydrogen. Met Sci Heat
1st ed. Philadelphia, PA: Woodhead Publishing; 2012. p. 247.
Treat 2008;50:282.
[14] Birnbaum HK. Hydrogen effects on deformation - relation
[33] Inoue A. The effect of hydrogen on crack propagation
between dislocation behavior and the macroscopic stress-
behavior and microstructures around cracks in austenitic
strain behavior. Scripta Metall Mater 1994;31:149.
stainless steels. Trans. ISIJ. 1979;19:170.
[15] Birnbaum HK, Sofronis P. Hydrogen-enhanced localized
[34] Ferreira P, Robertson IM, Birnbaum H. Influence of hydrogen
plasticity - a mechanism for hydrogen-related fracture.
on the stacking-fault energy of an austenitic stainless steel.
Mater Sci Eng, A 1994;176:191.
Mater Sci Forum 1996;207:93.
[16] Robertson IM. The effect of hydrogen on dislocation
[35] Whiteman MB, Troiano AR. The influence of hydrogen on the
dynamics. Eng Fract Mech 1999;64:649.
stacking fault energy of an austenitic stainless steel. Phys
[17] Allain S, Chateau J, Bouaziz O, Migot S, Guelton N.
Status Solidi 1964;7:K109.
Correlations between the calculated stacking fault energy
[36] Robertson IM. The effect of hydrogen on dislocation
and the plasticity mechanisms in Fe-Mn-C alloys. Mater Sci
dynamics. Eng Fract Mech 2001;68:671.
Eng, A 2004;387:158.
[37] Robertson IM. The effect of hydrogen on dislocation
[18] Sato K, Ichinose M, Hirotsu Y, Inoue Y. Effects of deformation
dynamics. Eng Fract Mech 1999;64:649.
induced phase-transformation and twinning on the
[38] Pontini AE, Hermida JD. X-ray diffraction measurement of
mechanical properties of austenitic Fe-Mn-Al alloys. ISIJ Int
the stacking fault energy reduction induced by hydrogen in
1989;29:868.
an AISI steel. Scripta Mater 1997;37:1831.
[19] Caskey Jr GR. Hydrogen compatibility handbook for stainless
[39] Egels G, Roncery LM, Fussik R, Theisen W, Weber S. Impact of
steels. Aiken: Savannah River Lab; 1983. p. 156.
chemical inhomogeneities on local material properties and
664 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 7 ( 2 0 2 2 ) 6 5 1 e6 6 5

hydrogen environment embrittlement in AISI 304L steels. Int [59] Zhou XW, Foster ME. Character angle effects on dissociated
J Hydrogen Energy 2018;43:5206. dislocation core energy in aluminum. Phys Chem Chem Phys
[40] Briant CL, Andresen PL. Grain boundary segregation in 2020;23:3290.
austenitic stainless steels and its effect on intergranular [60] San Marchi C, Balch DK, Nibur K, Somerday BP. Effect of
corrosion and stress corrosion cracking. Metall Trans A high-pressure hydrogen gas on fracture of austenitic steels. J
1988;19:495. Pressure Vessel Technol 2008;130:041401.
[41] Gallagher PCJ. The influence of alloying, temperature, and [61] Cai W, Sills RB, Barnett DM, Nix WD. Modeling a distribution
related effects on the stacking fault energy. Metall Trans of point defects as misfitting inclusions in stressed solids. J
1970;1:2429. Mech Phys Solid 2014;66:154.
[42] Zhao S, Osetsky YN, Zhang Y. Atomic-scale dynamics of edge [62] San Marchi C, Somerday BP. Thermodynamics of gaseous
dislocations in Ni and concentrated solid solution NiFe hydrogen and hydrogen transport in metals. Mater Res Soc
alloys. J Alloys Compd 2017;701:1003. Symp Proc 1998;1098-HH08e01.
[43] Zhou XW, Foster ME, Sills RB. An Fe-Ni-Cr embedded atom [63] Latanision RM, Ruff Jr AW. The temperature dependence of
method potential for austenitic and ferritic systems. J stacking fault energy in Fe-Cr-Ni alloys. Metall Trans
Comput Chem 2018;39:2420. 1971;2:505.
[44] Zhou XW, Foster ME, Sills RB, Karnesky RA. Towards [64] Molnar D, Sun X, Lu S, Li W, Engberg G, Vitos L. Effect of
molecular dynamics studies of hydrogen effects in Fe-Cr-Ni temperature on the stacking fault energy and deformation
stainless steels. The International Society of Offshore and behaviour in 316L austenitic stainless steel. Mater Sci Eng
Polar Engineers (ISOPE); 2019. 2019;759:490.
[45] Wei CY, Seidman DN. The spatial distribution of self- [65] Schramm RE, Reed RP. Stacking fault energies of seven
interstitial atoms around depleted zones in tungsten ion- commercial austenitic stainless steels. Metall Trans A
irradiated at 10 K. Phil Mag, A 1981;43:1419. 1965;6:1345.
[46] Henkelman G, Jonsson H. Improved tangent estimate in the [66] Zhou C, Hong Y, Zhang L, An B, Zheng J, Chen X. Abnormal
nudged elastic band method for finding minimum energy effect of nitrogen on hydrogen gas embrittlement of
paths and saddle points. J Chem Phys 2000;113:9978. austenitic stainless steels at low temperatures. Int J
[47] Henkelman G, Uberuaga BP, Jonsson H. A climbing image Hydrogen Energy 2016;41:13777.
nudged elastic band method for finding saddle points and [67] Takaki S, Nanba S, Imakawa K, Macadre A, Yamabe J,
minimum energy paths. J Chem Phys 2000;113:9901. Matsunaga H, Matsuoka S. Determination of hydrogen
[48] Nakano A. A space-time-ensemble parallel nudged elastic compatibility for solution-treated austenitic stainless steels
band algorithm for molecular kinetics simulation. Comput based on a newly proposed nickel-equivalent equation. Int J
Phys Commun 2008;178:280. Hydrogen Energy 2016;41:15095.
[49] Mehrer H, editor. Diffusion in solid metals and alloys, lan-dolt- [68] Sills RB, Foster ME, Zhou XW. Line-length-dependent
bornstein new series, group III, vol. 26. Berlin: Springer; 1990. p. 529. dislocation mobilities in an fcc stainless steels. Int J Plast
[50] Katz L, Guinan M, Borg RJ. Diffusion of H2, D2, and T2 in 2020;135:102791.
single-crystal Ni and Cu. Phys Rev B 1971;4:330. [69] Stukowski A. Structure identification methods for atomistic
[51] Zhou XW, Heo TW, Wood BC, Stavila V, Kang S, simulations of crystalline materials. Model Simulat Mater Sci
Allendorf MD. Temperature- and concentration-dependent Eng 2012;20:045021.
hydrogen diffusivity in palladium from statistically- [70] Hirth JP, Lothe J. Theory of dislocations. New York: McGraw-
averaged molecular dynamics simulations. Scripta Mater Hill; 1967.
2018;149:103. [71] Zhou XW. Thermodynamic analysis of dissociation of
[52] Zhou XW, Gabaly F El, Stavila V, Allendorf MD. Molecular periodic dislocation dipoles in isotropic crystals. RSC Adv
dynamics simulations of hydrogen diffusion in aluminum. J 2020;10:35062.
Phys Chem C 2016;120:7500. [72] Howie A, Swann PR. Direct measurements of stacking-fault
[53] Zhou XW, Dingreville R, Karnesky RA. Molecular dynamics energies from observations of dislocation nodes. Philos Mag
studies of irradiation effects on hydrogen isotope diffusion 1961;6:1215.
through nickel crystals and grain boundaries. Phys Chem [73] Kim J, Lee S-J, de Cooman BC. Effect of Al on the stacking
Chem Phys 2018;20:520. fault energy of Fee18Mne0.6C twinning-induced plasticity.
[54] Zhou XW, Jones RE, Gruber J. Molecular dynamics Scripta Mater 2011;65:363.
simulations of substitutional diffusion. Comput Mater Sci [74] Pontini AE, Hermida JD. X-ray diffraction measurement of
2017;128:331. the stacking fault energy reduction induced by hydrogen in
[55] Spataru CD, Heo TW, Wood BC, Stavila V, Kang S, an AISI 304 steel. Scripta Mater 1997;37:1831.
Allendorf MD, Zhou XW. Statistically-averaged molecular [75] Bond GM, Robertson IM, Birnbaum HK. On the determination
dynamics simulations of hydrogen diffusion in magnesium of the hydrogen fugacity in an environmental cell TEM
and magnesium hydrides. Phys. Rev Mater 2020;4:105401. facility. Scripta Mater 1986;20:653.
[56] Langley RA. Hydrogen trapping, diffusion and [76] Kresse G, Furthmuller J. Efficient iterative schemes for ab
recombination in austenitic stainless steels. J Nucl Mater initio total-energy calculations using a plane-wave basis set.
1984;128&129:622. Phys Rev B 1996;54:11169.
[57] Zhou XW, Sills RB, Ward DK, Karnesky RA. Atomistic [77] Kresse G, Furthmuller J. Efficiency of ab-initio total energy
calculations of dislocation core energy in aluminium. Phys calculations for metals and semiconductors using a plane-
Rev B 2017;95:054112. wave basis set. Comput Mater Sci 1996;6:15.
[58] Zhou XW, Ward DK, Zimmerman JA, Cruz-Campa JL, [78] Abrikosov IA, Niklasson AMN, Simak SI, Johansson B,
Zubia D, Martin JE, van Swol F. An atomistically validated Ruban AV, Skriver HL. Order-N Green’s function technique
continuum model for strain relaxation and misfit dislocation for local environment effects in alloys. Phys Rev Lett
formation. J Mech Phys Solid 2016;91:265. 1996;76:4203.
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 7 ( 2 0 2 2 ) 6 5 1 e6 6 5 665

[79] Abrikosov IA, Simak SI, Johansson B, Ruban AV, Skriver HL. [81] Finnis MW, Sinclair JE. A simple empirical N-body potential
Locally self-consistent Green's function approach to the for transition-metals. Phil Mag A 1984;50:45.
electronic structure problem. Phys Rev B 1997;56:9319. [82] Daw MS, Baskes MI. Embedded-atom method: derivation and
[80] Gyo € rffy BL, Pindor AL, Stocks GM, Staunton GM, Winter H. A application to impurities, surfaces, and other defects in
first-principles theory of ferromagnetic phase transitions in metals. Phys Rev B 1984;29:6443e53.
metals. J Phys F Met Phys 1985;15:1337.

You might also like