Engineering Fracture Mechanics 157 (2016) 56–71

Contents lists available at ScienceDirect

Engineering Fracture Mechanics

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

A uniform hydrogen degradation law for high strength steels

Haiyang Yu a, Jim Stian Olsen a, Antonio Alvaro b, Vigdis Olden b, Jianying He a,
Zhiliang Zhang a,⇑
a
Department of Structural Engineering, Norwegian University of Science and Technology, 7491 Trondheim, Norway
b
SINTEF Materials and Chemistry, 7456 Trondheim, Norway

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

Article history: The degrading effect of hydrogen on high strength steels is well recognized. The hydrogen
Received 26 November 2015 degradation is dependent not only on hydrogen content, but also on geometric constraints
Received in revised form 22 January 2016 or equivalently, level of stress triaxiality, which means the hydrogen degradation locus is
Accepted 1 February 2016
not likely to be a unique material property. Experimental data on notched tensile tests
Available online 12 February 2016
reported by Wang et al. are analyzed via cohesive zone modeling, and a cohesive strength
based uniform hydrogen degradation law is proposed upon normalization of hydrogen
Keywords:
degradation loci with different specimen geometries. Since the effects of hydrogen content
Hydrogen embrittlement
High strength steel
and geometric constraints are decoupled during normalization, the proposed law is appli-
Cohesive zone modeling cable to all the specimen geometries as a material property. This law is subsequently
Hydrogen degradation law applied to simulate the constant loading tests performed on the same material. Excellent
Constant loading test agreement is observed between the simulation and test results in terms of incubation time
for fracture initiation and highest permissible initial hydrogen content. The inconsistency
observed in one of the cases is discussed, suggesting that the effects of strain rate and stress
relaxation need to be taken into account in order to improve the transferability of the
degradation law calibrated from tensile tests to constant loading situations.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction

Hydrogen induced degradation of mechanical properties, often referred to as hydrogen embrittlement (HE), is a well rec-
ognized problem for structural steels [1,2]. Extensive studies have been done both experimentally [3–9] and numerically
[10–13], yielding a number of models accounting for this phenomenon. The real mechanism behind HE, however, remains
controversial due to the complexity in microstructures, hydrogen absorption and trapping sites and fracture modes [14].
Two representative mechanisms with different natures are the hydrogen enhanced decohesion model (HEDE) and the hydro-
gen enhanced localized plasticity model (HELP). The HEDE model assumes that dissolved hydrogen reduces the cohesive
strength of the iron lattice [15]; The HELP model, on the other hand, assumes that solute hydrogen enhances dislocation
motion leading to localized plastic deformation at the crack tip [16]. Either model is supported by a number of atomistic cal-
culations [17–19] and experimental observations [20–24] and is regarded as the dominating factor in particular fracture sce-
narios [25]. While the HEDE mechanism is commonly assumed to be dominant in cases where brittle intergranular fracture
surfaces are observed, the HELP mechanism is correlated to the situations where shear localization bands, slip traces or small
shallow dimples are present [24,26]. In reality, both brittle and ductile characteristics could be observed on the same fracture

⇑ Corresponding author.
E-mail address: zhiliang.zhang@ntnu.no (Z. Zhang).

http://dx.doi.org/10.1016/j.engfracmech.2016.02.001
0013-7944/Ó 2016 Elsevier Ltd. All rights reserved.
 H. Yu et al. / Engineering Fracture Mechanics 157 (2016) 56–71 57

Nomenclature

dc critical cohesive separation

g level of stress triaxiality
hC L i lattice hydrogen concentration
rc critical cohesive stress, or cohesive strength
rh local hydrostatic stress
r1
h remote hydrostatic stress
rc;H¼0 cohesive strength without hydrogen
f viscosity for numerical regularization
CI initial homogenous hydrogen concentration
CL local lattice hydrogen concentration at failure
CLT constant loading test
CSRT conventional strain rate tensile test
CZM cohesive zone modeling
D diffusion coefficient
HE hydrogen embrittlement
HEDE hydrogen enhanced decohesion
HELP hydrogen enhanced localized plasticity
R gas constant
r notch radius
SSRT slow strain rate tensile test
TZ absolute zero temperature
TSL traction separation law
VH partial molar volume of hydrogen

surface [4,7], indicating a possible combined effect of both mechanisms. In such situations, the part played by the HELP
model is usually more pronounced at lower levels of hydrogen content.
Despite the lack of a universally accepted micro-mechanism for HE, considerable progress has been made on the contin-
uum level. Hardie and Liu [3] investigated the effect of geometrical constraints on HE susceptibility by performing notched
tensile tests in gaseous hydrogen. Wang et al. [4–7] performed notched tensile tests as well as constant loading tests on
hydrogen pre-charged specimens made from AISI 4135 high strength steel. Moro et al. [8] performed notched tensile tests
on X80 steel in high pressure gaseous hydrogen environment and discussed the influence of strain rate. Nanninga et al. [9]
performed similar tests and compared the HE susceptibility among three pipeline steels. Olden et al. performed constant
loading tests in sea water under cathodic protection on single edge notched tension (SENT) specimens made from 25%Cr
duplex stainless steel [1] and API X70 pipeline steel [12]. The experimental work leads to a general conclusion that the
HE susceptibility increases with increasing levels of strength and increasing levels of stress concentration, or equivalently,
stress triaxiality. Some established failure criteria in hydrogen free situations can be adapted in the engineering HE failure
analysis with modifications accounting for the hydrogen effect. Enos and Scully [27] adapted the so-called stress modified
critical strain criterion in HE prediction by calibrating the hydrogen degradation law on critical strain from notched tensile
tests in presence of hydrogen. Since the effects of stress triaxiality is incorporated in the original stress modified critical
strain criterion [28], this method can inherently account for the geometry effects in HE prediciton. Similarly, Wang et al.
[29] adapted the strain based failure criterion with a hydrogen degradation law on critical strain calibrated for a boron bear-
ing steel. For a different high strength steel AISI4135, they applied the stress based failure criterion and presented a hydrogen
degradation law on critical stress [4,7]. In both situations, the degradation relation was proposed as power law and was con-
cluded as triaxiality independent. Most recently, Ayas et al. [13] re-analyzed the test data reported by Wang et al. [4,7] and
Hagihara et al. [30] and treated the critical axial stress based hydrogen degradation law as a unique material property. Recall
the micro-mechanisms discussed above, we can see that the hydrogen affected critical strain criterion is closely related to
the HELP model while the hydrogen affected critical stress criterion to the HEDE model.
By utilizing the cohesive zone modeling (CZM) technique, it is possible to incorporate contributions from HELP as well as
from HEDE. In CZM, damage is processed inside a layer of cohesive elements that are inserted along the fracture path
between solid elements. The constitutive behavior of cohesive elements is described by the so-called traction separation
law (TSL) [31]. The TSL is usually characterized by the critical cohesive stress rc which gives the strength of the cohesive
element and the critical separation dc which defines the failure point. The area below the traction separation curve is called
the separation energy or the cohesive energy Cc . For detailed information regarding CZM and TSL, the readers are referred to
[32,33]. CZM provides a phenomenological continuum framework for failure analysis [34–36]. Different failure scenarios and
factors affecting the failure behavior can be accounted for by manipulating the TSL. The bilinear TSL, for instance, is often
employed for brittle fracture [37,34,38] and the trapezoidal TSL for ductile situations [39,40]. The effects of stress triaxiality
[41,42] and strain rate [11] have also been considered by adding new items into the TSL. The degradation effect of hydrogen
58 H. Yu et al. / Engineering Fracture Mechanics 157 (2016) 56–71

can be introduced in several manners. Ahn et al. [24] assumed that the material softening occurs due to presence of hydro-
gen, which influences the void process close to the crack tip. They performed unit cell analysis with the softened material
and used the results to modify the TSL. Apparently, HE is introduced through CZM on the basis of the HELP mechanism
in this study. Serebrinsky et al. [10], on the other hand, kept dc for a brittle linearly decreasing TSL constant while lowering
rc according to the increase in hydrogen content. In this study, the cohesive strength and the separation energy are lowered
in presence of hydrogen, corresponding to the concept of HEDE. Liang and Sofronis [43] built the hydrogen influence into a
polynomial TSL in the manner of HEDE and introduced material softening to the continuum elements at the same time, thus
combining the mechanisms of HEDE and HELP. Among these three approaches, the second one is the most straightforward
since no change in the TSL formulation or extra material softening model is required. Furthermore, since CZM is a phe-
nomenological approach in nature, it captures the failure behavior by suitable adjustment of the cohesive parameters
regardless of the specific micro-mechanism. Based on these considerations, the HEDE concept is often built into the CZM for-
mulation to model the HE phenomena. Olden et al. [38,1,12] developed a three-step finite element procedure comprised of
elastic–plastic stress analysis, stress driven hydrogen diffusion analysis and hydrogen informed CZM analysis for HE mod-
eling. In their analysis, the hydrogen content at each time increment were passed into the cohesive elements and used to
reduce the cohesive strength according to a degradation law. The hydrogen degradation law used by Serebrinsky et al.
[10] was employed. Satisfactory agreement with the experimental results was achieved by suitable adjustment of initial
cohesive parameters. Alvaro et al. [44,2] then extended this approach to the 3D situation.
For all the HE modeling techniques and failure criteria discussed above, a proper hydrogen degradation law is of funda-
mental importance. Serebrinsky et al. [10] proposed a hydrogen degradation law for the cohesive strength based on the first
principle simulation results reported by Jiang and Carter [18], which was adopted by other researchers [38,1,44,2]. It is also
possible to determine the hydrogen influenced mechanical properties through nano-mechanical testing methods [23,45,46].
While atomistic calculation and nano-mechanical testing are powerful tools in investigating the mechanisms behind HE, it’s
questionable whether the hydrogen degradation laws retrieved by these methods are directly transferable to the engineering
components, considering the large difference in time and length scales as well as the impurities introduced in the manufac-
turing process. It is therefore favorable, from the engineering point of view, if one can calibrate a hydrogen degradation law
directly from conventional material tests. Wang et al., as mentioned earlier, calibrated the hydrogen degradation relation for
the critical stress [4,7] and critical strain [29] based on notched tensile tests. Similar practices can be seen elsewhere
[27,47,30]. A hydrogen degradation law for the cohesive strength rc calibrated from tensile tests, however, has not been
reported.
In the present work, the procedure for calibrating a CZM based hydrogen degradation law from notched tensile tests is
established. A hydrogen degradation law is given for rc by analyzing the tensile test data on the AISI 4135 high strength steel
reported by Wang et al. [4,7]. The law is subsequently built into the three-step HE simulation scheme reported by Olden et al.
[38,1] and validated against the constant loading tests performed on the same material [6]. Throughout this paper it is
assumed that fracture in high strength steels happens in a HEDE manner [48] and that only lattice hydrogen contributes
to the decohesion behavior [13]. It will be shown that the hydrogen degradation loci become independent of the specimen
geometry upon normalization, which indicates the possibility of decoupling the effects of stress triaxiality and hydrogen on
material cohesive strength. Different from the opinion that the failure locus in terms of peak axial stress versus local hydro-
gen concentration is a unique material property [13], we treat the uniform CZM based hydrogen degradation law as a mate-
rial property. Based on these results, procedures for calibrating the hydrogen degradation law from a simple series of tensile
tests and for implementing the locus in practice are established, which are straightforward and easy to implement in engi-
neering failure analysis. The proposed methodology is satisfactorily verified by the constant loading tests for both materials
investigated here. Inconsistency between the simulation and the test results observed in one of the cases is attributed to a
combined contribution from the strain rate effect and the stress relaxation behavior, indicating that these two factors need to
be considered in order to improve the applicability of the proposed hydrogen degradation locus.

2. Formulation and numerical procedure

2.1. Material and test

Here is a brief summary of the experiments performed by Wang et al. [4,6,7]. These tests were performed on two cate-
gories of specimens made from AISI 4135 high strength steel, denoted B15 and B13, respectively. The difference lies in the
tempering temperature during heat treatment: B15 (tempered at 633 K) developed a tensile strength of rB ¼ 1450 MPa and
B13 (tempered at 733 K) of rB ¼ 1320 MPa. Detailed material properties are given in [6].
A homogenous hydrogen distribution was introduced into the specimens, after which the specimens were electroplated
to prevent hydrogen release [4]. The so-called slow strain rate tensile tests (SSRT) were then performed on hydrogen pre-
charged smooth and notched specimens, giving nominal stress–strain curves at different levels of initial hydrogen content.
The notched specimens are with radii of r ¼ 0:1 mm and r ¼ 0:8 mm, representing high and medium stress triaxiality cases,
respectively. The smooth specimen represents the case with the lowest stress triaxiality. The initial hydrogen content was
measured by the thermal desorption spectrometry (TDS) analysis. More details regarding the SSRT and TDS are found in
[4,7].
 H. Yu et al. / Engineering Fracture Mechanics 157 (2016) 56–71 59

In another set of tests, the pre-charged specimens were loaded up to a net section stress rN ¼ 0:9rB and then kept at this
load level for 6000 min, giving the incubation time for delayed fracture at different levels of initial hydrogen content. These
tests were referred as the constant loading tests (CLT) [6].
In the finite element (FE) analyses, J2 flow theory with isotropic hardening is applied, and the material hardening behav-
iors are determined by the results of the smooth tensile tests in absence of hydrogen. The stress–strain relations as given in
[13] for B13 and B15 are employed in the current study.

2.2. Cohesive zone modeling approach

Cohesive zone modeling (CZM) is performed on the test results. Axi-symmetric elements in ABAQUS 6.14 [49] are used to
model the specimen with a layer of cohesive elements inserted at the mid-section, as shown in Fig. 1(a). Only a quarter of the
specimen needs to be modeled due to symmetry. There are two possible ways of applying symmetric boundary conditions to
the cohesive elements, as discussed by Brocks et al. [50]. On one hand, axial displacements of the nodes on the lower cohe-
sive surface can be constrained allowing only half-opening of the cohesive elements, in which the critical separation dc rep-
resents half of the actual value while the cohesive strength rc is the same as the actual one. Alternatively, symmetric full-
opening of the cohesive elements can be realized by establishing linear constraint equations on the axial displacements of
corresponding nodes on the upper and lower cohesive surfaces, and the critical separation dc here is the same as the actual
one while the cohesive strength rc represents half of the actual value. The effectiveness of both practices has been validated
against the simulation results of the whole specimen model in a preliminary case study not shown here. The latter approach
is employed in the present work since it results in a larger dc and a smaller rc compared to the former approach, which is
beneficial for convergence from the numerical point of view. The polynomial TSL introduced by Needleman [51] is applied in
the current CZM simulations [1]
8
2
< 27
rðdÞ ¼ 4
rc ddc 1  ddc d < dc
ð1Þ
:
0 d > dc

where rðdÞ and d are stress and separation in the cohesive element at the current time increment. An illustration of this TSL
is found in Fig. 1(b).
It is very common to have convergence problem in CZM, that is, at the point of instability, implicit FE computations are
unable to converge to an equilibrium solution, which usually terminates the calculation and makes it impossible to follow
the post-instability behavior [52]. This problem is particularly pronounced in our case where failure initiates in a material
interface across which there is generally low stress gradients due to lack of an initially sharp crack. In order to overcome this
problem, we applied the technique suggested by Gao and Bower [52] where a viscosity-like item is built into the constitutive
behavior of the cohesive elements
 2  
27 d d d d
rðdÞ ¼ rc 1 þf ð2Þ
4 dc dc dt dc

f is the viscosity-like parameter which is introduced not to model the physical energy dissipation process but to regularize
the numerical instability [52]. To some extent, this technique is analogous to that employed by ABAQUS [49] as a solution for
the convergence problems with its embedded cohesive elements.
Upon choosing a f value as small as possible, this method effectively overcomes the convergence problem with negligible
influence on global and local stress states, as verified in a preliminary study shown in Fig. 2. This figure illustrates the results
of CZM simulations on the notched tensile bars with r ¼ 0:8 mm, in terms of axial stress distribution close to the notch root.

(a) (b)

Fig. 1. Illustrations of (a) the specimen geometry and (b) a polynomial TSL.
60 H. Yu et al. / Engineering Fracture Mechanics 157 (2016) 56–71

Fig. 2. Effect of the viscosity-like item on the axial stress distribution at failure initiation.

When f ¼ 0, i.e. no viscosity is applied, the simulation can’t proceed beyond the failure initiation point because of the con-
vergence problem, which can be solved by applying a non-zero viscosity. A slightly higher level of axial stress at the point of
failure initiation is observed upon introduction of viscosity. Meanwhile, this effect becomes less pronounced with a smaller
value of viscosity. For the case with f ¼ 8:2E  9, the influence of viscosity is practically negligible.
The FE simulation is performed in ABAQUS 6.14 [49] and the cohesive process is handled by a user element subroutine
(UEL) originally developed by Scheider et al. [53] and modified by the current authors for introduction of the viscosity-like
item.
The CZM approach presented above is used to analyze the SSRT data reported by Wang et al. [4,7]. Tensile test results in
absence of hydrogen were analyzed first and suitable combinations of cohesive parameters ðrc ; dc Þ that are able to capture
both the global and local stress characteristics at failure are determined for the smooth specimen and the notched specimens
with r ¼ 0:8 mm and r ¼ 0:1 mm. Subsequently, the critical displacement dc determined in the hydrogen-free situation is
kept constant while the cohesive strength rc is adjusted to fit the results of the tensile tests performed on hydrogen pre-
charged specimens. Note that the hydrogen effect is not built into the numerical procedure at this stage.

2.3. Treatment of hydrogen

The CZM based cohesive strength in presence of hydrogen can be determined based on the technique presented in the last
subsection. Another important aspect of calibrating a proper hydrogen degradation law is the treatment of hydrogen, since
the degradation law is to be given as a relation between the cohesive strength and the hydrogen concentration. Initial con-
centrations C I were explicitly given for the homogenously distributed hydrogen in the tests. Such global values, however,
cannot be directly used in the calibration, since we are addressing hydrogen induced degradation locally and it has to be
correlated with local conditions. This has been shown in [4,7], where the global failure stress-hydrogen concentration plots
 H. Yu et al. / Engineering Fracture Mechanics 157 (2016) 56–71 61

were obviously scattered and couldn’t be rationalized with a general expression while plots in terms of local values were
fitted to a fracture locus which was taken as a material property.
The local hydrogen population is very complicated and is generally categorized in two communities: the lattice intersti-
tial hydrogen that accumulates at sites of increased stress due to the dilatation of the lattice [54] and the trapped hydrogen
that resides at the so-called trapping sites such as carbide particles, dislocations and grain boundaries [13]. According to the
principle by Oriani [55], the lattice hydrogen and trapped hydrogen are always in equilibrium. Based on this principle, a
number of diffusion models are proposed [54,56–58], reflecting the competition between hydrostatic stress field governing
the lattice hydrogen content and the plastic strain fields that is correlated to the trapped hydrogen content. In the situation
with large scale yielding, plastic strain, hence the trapping effect will dominate the total hydrogen concentration and has to
be considered, which is the case with low and mediate strength steels. For details of consideration and numerical implemen-
tation of the trapped hydrogen in CZM simulations, the reader is referred to [45,1]. In the case of high strength steels where
little plasticity develops before failure, especially in presence of hydrogen, the contribution from the trapped hydrogen is
often neglected [48,4,7]. Most recently, Ayas et al. [13] showed that the tensile test data on the AISI4135 steel could only
be rationalized with the lattice hydrogen concentration instead of the trapped hydrogen concentration.
Based on the considerations discussed above, the maximum value of the lattice hydrogen concentration at failure C L ,
referred to as the local lattice hydrogen concentration hereinafter, is used to construct the hydrogen degradation locus
for the targeted AISI4135 steel. A common practice of calculating the lattice hydrogen concentration is to utilize the follow-
ing equation [4,7,13]
 
ðrh  r1h ÞV H
hC L i ¼ C I exp ð3Þ
RT
where hC L i is the lattice hydrogen concentration, C I the initial homogenous hydrogen concentration across the pre-charged
specimen, rh and r1 h the local hydrostatic stress and the remote hydrostatic stress at failure, V H the partial molar volume of
hydrogen in bcc-Fe, R the gas constant and T the testing temperature in degree Kelvin. rh and r1 h can be obtained by FE
simulation.
In the present study, lattice hydrogen concentration is determined by the stress-driven mass diffusion analysis with ABA-
QUS [49], in order to be consistent with the three-step HE simulation to be elaborated later. Mass diffusion in ABAQUS is
based on the modified Fick’s law accounting for the effect of hydrostatic stress gradient [2]
@hC L i DV H DV H
¼ Dr2 hC L i þ rhC L irrh þ hC L ir2 rh ð4Þ
@t RðT  T Z Þ RðT  T Z Þ
where D is the diffusion coefficient and T Z the absolute zero temperature. An elastic–plastic stress analysis is performed for
each specimen, giving the stress field until failure. The stress information is then fed back to ABAQUS for the stress driven
diffusion analysis. Sufficiently long time is specified in these analyses to give saturated hydrogen concentration, since the
loading rate in the real tests was small enough for equilibrium states to be established [13]. The parameters related to
the mass diffusion analyses are summarized in Table 1.
The cohesive strengths calibrated in the earlier step are then plotted against the maximum lattice hydrogen concentra-
tion at failure C L obtained here, yielding the CZM based hydrogen degradation loci for different specimen geometries.

2.4. Three-step HE simulation procedure

To validate the effectiveness of the calibrated hydrogen degradation law, CZM simulation coupled with hydrogen effect is
performed in this part. The so-called three-step HE simulation procedure [45,1] is employed and a brief summary is given
below

(I) Elastic–plastic FE simulation. The specimen is loaded up to and then kept at a certain load level, yielding detailed
information of the stress field throughout the loading history.
(II) Stress driven mass diffusion analysis based on Eq. (4). The homogeneously pre-charged hydrogen re-distributes
according to the stress field information obtained in the previous step, yielding hydrogen concentrations at individual
nodes over the loading process.
(III) Elastic–plastic FE analysis with addition of user-defined cohesive elements inserted over the mid-section. The influ-
ence of hydrogen is accounted for by a decrease of the cohesive strength according to the calibrated hydrogen degra-
dation law. Hydrogen content information obtained in the previous step is used to update the cohesive strength.

Table 1
Parameters used in the mass diffusion analysis [6].

V H (m3/mol) D (m2/s) R (J/mol) T (K)

6 11 8.314 298
2:1  10 2:5  10

D ¼ 4  1011 m2/s for B13.

62 H. Yu et al. / Engineering Fracture Mechanics 157 (2016) 56–71

Detailed information on numerical implementation of this approach is found in [45] and it should be noted that the part
in the original procedure accounting for the trapped hydrogen concentration is excluded in the present study. The results of
the HE simulation are then compared with those of the constant loading test [6].

3. Results and discussion

As elaborated in the last section, the main purpose of this work is to calibrate a CZM based hydrogen degradation law for
the target high strength steel from SSRT and to explore the applicability of the proposed equation to the CLT. The results of
these two parts are presented separately.

3.1. Calibration of the hydrogen degradation law

The stress–strain relations at different initial hydrogen concentrations C I were reported in literature [7], which are
extracted and used as the reference for determining cohesive parameters. CZM simulations with trial combinations of cohe-
sive parameters are first performed on hydrogen free cases, yielding the proper set of ðrc ; dc Þ that gives the right description
of both the nominal stress–strain curve and local stress distribution. The local stress distribution is determined by elastic–
plastic FE simulations without cohesive elements. An example of parameter calibration for the smooth specimen without
hydrogen is given in Fig. 3(a). The CZM result captures the main characteristics of the nominal stress–strain curve given
by experiment with good accuracy and the cohesive parameters are determined as rc ¼ 1700 MPa and dc ¼ 0:02 mm for this
case. The test data of notched tensile specimens are analyzed in the same manner, yielding TSLs at C I ¼ 0, as shown in Fig. 4.
Clearly, the cohesive parameters are dependent on the level of stress triaxiality: for the case of a larger notch radius which
indicates a lower triaxiality, dc is larger while rc is lower; as the triaxiality becomes higher, dc decreases while rc grows. The
cohesive energy decreases as the stress triaxiality increases. Such trends are in accordance with those found in literature
[59,41,11,42].
The cohesive parameters with hydrogen are obtained on the basis of those determined without hydrogen, in a HEDE man-
ner: for each specimen geometry, the corresponding critical separation dc is kept constant while the cohesive strength rc is
lowered in order to fit the corresponding nominal stress–strain curve and to capture the local stress state. An example of
calibration for the r ¼ 0:1 mm notched specimen with an initial hydrogen concentration C I ¼ 0:015 wppm is given in
Fig. 3(b), where the cohesive parameters are determined as rc ¼ 2922 MPa and dc ¼ 0:0005 mm. A summary of the cases
selected for CZM calibration are shown in Table 2.
The stress field at failure is analyzed for each case and is fed to ABAQUS for stress-driven hydrogen diffusion analysis. The
pre-charged hydrogen concentration C I is applied as the initial value all across the specimen. The applied diffusion time is
sufficiently long for the hydrogen diffusion to reach the steady (saturated) state. The hydrogen degradation loci can thus be
obtained by plotting the CZM based cohesive strength against corresponding C L values for each geometry, as shown in Fig. 5,
for the case of B15. An obvious effect of specimen geometry, or stress triaxiality, is reflected here. The degradation loci of
different specimen geometries are separated and show an increasing trend with the level of stress triaxiality. Although
the difference between the cases of r ¼ 0:1 mm and r ¼ 0:8 mm is minor, significant difference is observed when the
r ¼ 1 case is taken into account. Similar situation is discussed by Ayas et al. [13], who re-analyzed the same sets of test data
and presented the hydrogen degradation loci in terms of maximum principal stress. In their study, the degradation loci of the
notched specimens were rationalized as a unique material property while the obviously lower locus of the smooth specimen
was correlated to the weakest link theory. The weakest link theory was originally developed by Weibull for describing the
strength of brittle materials [60], it was then extended to consider the probabilistic failure of ceramic materials [61,62]. It is

(a) (b)

Fig. 3. (a) An example of CZM parameter fitting for the smooth specimen with C I ¼ 0; (b) CZM parameter fitting for the r ¼ 0:1 mm notched specimen with
C I ¼ 0:015 wppm; both cases are for B15 [7].
 H. Yu et al. / Engineering Fracture Mechanics 157 (2016) 56–71 63

Fig. 4. Calibrated TSLs for different specimens geometries in absence of hydrogen (C I ¼ 0).

Table 2
Overview of the cases selected for CZM calibration.

Material Notch radius r (mm) Intial hydrogen concentration C I (wppm)

B15 1 0, 0.27, 0.50, 1.00, 1.30
B15 0.8 0, 0.10, 0.15, 0.28, 1.06
B15 0.1 0, 0.015, 0.03, 0.40, 1.02
B13 0.8 0, 0.21, 0.42, 0.85, 1.17, 1.60, 2.30
B13 0.1 0, 0.11, 0.24, 0.44, 0.94, 1.40, 2.20

also applicable in failure scenarios showing certain amount of ductility, such as cleavage. Due to its statistical nature, this
method has the merit of rationalizing the scatter in experimental data [63]. While the weakest link theory seemed capable
of explaining the weaker response in smooth specimens [13], it is still somewhat questionable to use the degradation locus
of the notched specimen as a material property irrespective of the obvious geometry dependence. In the present work, the
fracture loci are rationalized in a different manner which is at least more straightforward.
It is noted that although lying apart, the hydrogen degradation loci in Fig. 5 actually follow similar trends, except the last
data point of the smooth case. This abnormal point might be caused by errors in the experiment and is discarded here, which
is the same practice as in [13]. We then perform the so-called normalization technique on these three degradation loci: the
cohesive strength rc at each data point is divided by its corresponding value at C L ¼ 0 (denoted rc;H¼0 ), as shown in Fig. 6.
Disregarding the abnormal point, it is obvious that the three loci collapse into one upon normalization and this normalized
degradation locus can be expressed by a single equation
rc
¼ 0:421e2:227C L þ 0:579 ð5Þ
rc;H¼0
This equation can be reformulated in a more general manner

rc ðdc ; g; HÞ ¼ Rðdc ; gÞ  DðHÞ ð6Þ

where g represents the contribution from stress triaxiality and H the contribution from hydrogen (lattice hydrogen in the
present case); Rðdc ; gÞ corresponds to rc;H¼0 in Eq. (5) and DðHÞ to the items on the right. The cohesive strength is thus gov-
erned by two factors, the stress triaxiality and the local hydrogen content, whose influences can be treated separately. In this
sense, the hydrogen degradation effect is ‘‘decoupled” from the stress triaxiality, thus independent of specimen geometry.
Based on these considerations, we suggest to use Eq. (5) or equivalently DðHÞ, instead of the individual degradation locus
in Fig. 5, as a material property regarding hydrogen degradation effect. This relation is named as the CZM based uniform
hydrogen degradation law, since it is summarized based on normalization of the loci with different geometries.
There are several implications with the proposed hydrogen degradation law. An apparent advantage of this law compared
with those deduced from the fracture loci of the notched specimens [4,7,13] is that it is applicable to all geometries including
the smooth specimen, since the effects of geometry and hydrogen are decoupled. In the current law, an exponential form is
applied and the maximum value of cohesive strength is explicitly given in Eq. (5) as rc;H¼0 . The degradation relations pro-
posed by Wang et al. [4,7], however, were in the form of a power law and indicated infinite strength at a local hydrogen con-
centration close to zero. In this aspect, Eq. (5) is more realistic since it gives the upper boundary of the cohesive strength and
correctly reflects the degrading effect on the cohesive strength due to hydrogen. Further, we believe that the current degra-
dation law agrees better with the microscopic observations. Fracture surfaces were observed on a scanning electron micro-
64 H. Yu et al. / Engineering Fracture Mechanics 157 (2016) 56–71

Fig. 5. Hydrogen degradation loci for the case of B15.

Fig. 6. Normalized hydrogen degradation locus for the case of B15.

scope (SEM) in the original tests [4,7]. While the area fraction of intergranular fracture increased with increasing diffusible
hydrogen content, the shear lip regions consisted of fine dimples were present in varying degrees on almost all the fracture
surfaces, and this phenomenon was more pronounced at lower levels of hydrogen content and stress triaxiality, which indi-
cated a possible combined action of HELP and HEDE mechanisms. Again, it is then questionable to describe these failure
behaviors as purely brittle, especially in the case of smooth specimen at low hydrogen concentration. In the current model,
the cohesive strength is reduced from an initial value determined in hydrogen free situation rc;H¼0 or Rðdc ; gÞ. These initial
values are ductile in nature, since the triaxiality dependence agrees well with that in a ductile process [64] and more con-
vincingly, the variation of the calibrated TSLs in Fig. 4 agrees very well with the trend shown by unit cell analyses [59,41]. In
the present model, the possible contribution of ductility is not completely ruled out at low levels of hydrogen concentration,
which might help to explain the triaxiality dependence of the degradation loci in Fig. 5. A step further, we may even postu-
late that some interactions between hydrogen and void process, as discussed by Ahn et al. [24] and Liang et al. [26], are
implicitly included. At high levels of hydrogen concentration, this model implies a lower boundary of the cohesive strength,
which might indicate a completely brittle failure scenario that can be accounted for with the weakest link theory. These con-
clusions are based on reasonable correlations between the test data and simulation results instead of on mechanisms inten-
tionally built into the model. Again, it is worth noting that CZM is a phenomenological technique where no physical
mechanisms of HE is explicitly represented.
On the basis of the proposed hydrogen degradation law in Eq. (6) and the dicussions above, a practical issue is implied.
Since the hydrogen effect is decoupled from that of the geometric constraints, DðHÞ can be determined by only one series of
notched tensile tests with the same geometry with hydrogen. In this situation, the smooth specimen would be the best can-
didate for calibration of DðHÞ. The fracture of smooth specimens precharged with hydrogen is usually way before obvious
necking, indicating very small stress gradient and thus negligible hydrogen redistribution up to failure, it is therefore accept-
able to use the initial hydrogen concentration C I and the nominal failure stress to construct the degradation loci in Figs. 5 and
6. Meanwhile, it should be noted that obvious necking before fracture could still occur in cases of sufficiently small C I values,
where the effect of stress gradient needs to be considered. The calibrated hydrogen degradation law DðHÞ can then be taken
 H. Yu et al. / Engineering Fracture Mechanics 157 (2016) 56–71 65

as a unique material property. In CZM based engineering HE failure assessments, one needs only to find the right cohesive
parameters ðdc ; rc Þ for the current geometry before incorporating the hydrogen degradation law as elaborated in the previ-
ous section. Actually, even this part of effort can be alleviated through implementation of the so-called triaxiality dependent
CZM technique [11,42] where the effect of triaxiality Rðdc ; gÞ is explicitly incorporated. In the present CZM framework, how-
ever, the effect of triaxiality is not incorporated.
Exactly the same procedure is done for B13 steel and the results are shown in Fig. 7. The results of the smooth tensile test
were not given for this material, instead, an additional set of notched tensile test results at r ¼ 0:25 mm was reported [4].
Only the notched tensile results of r ¼ 0:1 mm and r ¼ 0:8 mm are analyzed here since the additional case is believed to con-
tribute little to final conclusions. The last data point is discarded in interpretation of Fig. 7 due to its obvious abnormality.
Clearly, the degradation loci of different notch radii are separated (Fig. 7(a)) and collapse into one upon normalization
(Fig. 7(b)). A uniform hydrogen degradation law is thus proposed for B13 in the same manner as for B15

rc ðdc ; g; HÞ ¼ Rðdc ; gÞ  DðHÞ ð7Þ

DðHÞ ¼ 0:431e0:543C L þ 0:6 ð8Þ

This equation has the same form as Eq. (5), the hydrogen degradation law for B15. A comparison between these two equa-
tions reveals a larger and faster decrease with hydrogen concentration in the case of B15. Recall the difference in mechanical
properties between B13 and B15, we recognize a more pronounced HE susceptibility in the steel with higher initial strength,
which is in accordance with the conclusion in [6].

3.2. Application of the hydrogen degradation law

This part presents the application of the proposed hydrogen degradation law. A large amount of HE induced fracture
occurs under the context of constant loading in engineering practice. In the case of subsea pipelines, for instance, constant
loading scenario is created by internal fluid pressure, hydrogen developed during the cathodic protection process will accu-
mulate in steel and may lead to HE fracture when the critical hydrogen content is reached after a long time [6,1]. HE assess-
ment under the constant loading condition is twofold: to predict the incubation time for delayed fracture at a given level of
hydrogen content and to determine a critical value of initial hydrogen content C IC below which the component is safe for a
certain load level. Apparently, both aspects need combination of a proper hydrogen degradation law and the evolution of
local hydrogen content.
The hydrogen degradation law proposed in last section is built into the three-step HE simulation framework elaborated
earlier. The load level and diffusion parameters used by Wang et al. [6] in the CLT performed on B13 and B15 are applied
and the simulation results are compared against the test data. Since the real time for the loading-up stage in CLT were not
given, we use the hydrogen saturation time reported in [6] as a reference to scale the step time in ABAQUS to the real
time.
The results of CZM simulation that contains the proposed hydrogen degradation law and the hydrogen diffusion are plot-
ted against the CLT results in Fig. 8, for the case of r ¼ 0:8 mm. Satisfactory agreement between these two sets of data is
observed: the CZM simulation captures the time to fracture as well as the critical hydrogen concentration C IC below which
no fracture is observed at the current load level. Note that the points marked with an arrow in the figure indicate that no
failure is observed after 6000 min which is the loading time in the CLT [6]. Similar situation stands for the case of B13 with
r ¼ 0:1 mm, as shown in Fig. 9, where the simulation results agree with the test data very well. These observations demon-
strate the effectiveness of the proposed hydrogen degradation laws and indicate that the laws calibrated from SSRT could be
transferable to CLT.

(a) (b)

Fig. 7. (a) Hydrogen degradation loci for the case of B13; (b) Normalized hydrogen degradation locus for the case of B13.
66 H. Yu et al. / Engineering Fracture Mechanics 157 (2016) 56–71

Fig. 8. Comparison between the CZM simualtion and the test results, for the case of B15 with r ¼ 0:8 mm.

Fig. 9. Comparison between the CZM simualtion and the test results, for the case of B13 with r ¼ 0:1 mm.

Fig. 10. Comparison between the CZM simualtion and the test results, for the case of B15 with r ¼ 0:1 mm.

The case of B15 with r ¼ 0:1 mm, however, is somewhat different, see Fig. 10. At the same initial hydrogen concentration
C I , failure occurs earlier in the CZM simulation with hydrogen degradation than in the constant loading tests. Note that the
points with arrows pointing left indicate failure occurs within less than 0.1 min while those with arrows pointing right
indicate no failure after 6000 min. The critical value C IC below which no failure occurs was predicted to be in the range of
 H. Yu et al. / Engineering Fracture Mechanics 157 (2016) 56–71 67

0.035–0.06 wppm from the constant loading tests, whereas simulation yields C IC ¼ 0:028 wppm. The simulation based on
the current hydrogen degradation law seems to give conservative predictions compared to the real life situation. In engineer-
ing practice, we are usually more concerned about the C IC value than the time to fracture, hence the current CZM simulation
still seems acceptable due to its fair prediction of C IC and its conservative nature.
The possible reasons for the disagreement observed between the simulations and the tests shown in Fig. 10 are discussed,
in a qualitative manner. There are two possible ways to account for this problem, from the aspect of fracture driving force or
from the aspect of material resistance.
The driving force is reflected by the local stress level which is determined by elastic–plastic FE simulation where Von
mises material is assumed. In this situation, the local stress states developed in SSRT and in CLT are exactly the same at a
given level of remote load and no variation in the local stress state is expected during constant loading stage. In order to
account for the difference between SSRT and SLT, we introduce the concept of creep relaxation of stress [65] where certain
amount of stress re-distribution occurs due to creep behavior during the constant loading period [66,67]. Although the creep
behavior is commonly known to come into effect at high temperatures for the case of steel, it could also exert significant
effects in case of high stress concentration at room temperatures, which is referred as the cold creep [68,69]. This mechanism
tends to give an increasingly homogenous stress distribution across the specimen, which naturally lowers the peak stress at
stress concentration sites due to imposition of the force equilibrium, as shown in [70]. In order to given a more detailed illus-
tration, we re-analyzed the stress state in the constant loading test with FE simulation where visco-plastic material model is
employed [67]. The two-step loading method and creep constants as suggested by Nuez and Glinka [67] are applied, and the
results for the case of B15 material with r ¼ 0:1 mm are shown in Figs. 11 and 12. Clear effects of stress relaxation and redis-
tribution can be seen in these figures. The maximum axial stress is lowered by approximately 400 MPa by the end of the
constant loading stage, which is more than enough to account for the longer incubation time for the delayed fracture in real
tests. These calculations, of course, are of illustrative purpose only, since there has been no test performed on this kind of
material to validate the existence of the so-called cold creep effect, let alone proper calibration of the creep constants. Nev-
ertheless, the concept of stress re-distribution during constant loading tests may provide a qualitative explanation which is
feasible at least from the numerical point of view.
Another aspect concerns the calibration and application of the hydrogen degradation law that is the fundamental indica-
tor and dictator of the material resistance during CZM simulation. Except for the distinction in the constant loading stage,
there exists another difference between CLT and SSRT that is used as the basis for calibrating the hydrogen degradation
law: the loading rate during the loading-up stage in CLT is larger than that in SSRT. Although the loading rate or loading time
is not explicitly given for the CLT [6], it can be retrieved by the target load level and the time scale deduced from the hydro-
gen diffusion simulation as mentioned earlier. The approximated loading rate in the CLT is comparable to that in the so-
called conventional rate tensile tests (CSRT) reported by Hagihara et al. [30] that is 200 times as large as in the SSRT utilized
here. Considering that the effect of strain rate on high strength steels charged with hydrogen has been reported in a number
of studies [48,71], it is reasonable to examine the transferability of the current hydrogen degradation law to high strain rate
situations. A comparison between the degradation loci constructed from SSRT and CSRT was given in [30], as cited here in
Fig. 13. The locus determined from the constant loading tests [6] are also plotted in the same figure. These failure loci,
although constructed on the basis of maximum axial stress at failure, provide some clues for our problem since similar trends
apply to the CZM based loci in the present work. A general trend can be seen from Fig. 13 that material strength seems a bit
higher in CRST than in SSRT, which is more pronounced for the case with the smaller notch radius. Further, the failure points
determined from CLT seems to agree better with the loci of CSRT instead of SSRT. A general conclusion can be drawn that

Fig. 11. (a) Distribution of the axial stress close to the notch root when the constant load level is first achieved; (b) distribution of the axial stress after stress
relaxation during the constant loading stage.
68 H. Yu et al. / Engineering Fracture Mechanics 157 (2016) 56–71

Fig. 12. Redistribution of the axial stress across the mid-section during the constant loading stage.

Fig. 13. The axial stress based hydrogen degradation loci reported in the literature [4,6,30].

higher resistance to fracture is developed in cases with higher strain rates, which is in accordance with that in earlier studies
[48,71]. Such observation might be responsible for the current problem, to some extent: since the hydrogen degradation law
is calibrated from SSRT with very low strain rates, it actually underestimates the material strength in the CLT where the
material is ”reinforced” during the loading-up stage with a much higher strain rate, thus reduced incubation time and critical
C IC values are predicted in CZM simulations with constant loading. Since the hydrogen degradation relation for the current
steel B15 exhibits much sharper decrease at lower hydrogen contents, as shown in Fig. 6, the error caused by the underes-
timation is more pronounced for the case of r ¼ 0:1 mm that fails at a lower hydrogen content than the case of r ¼ 0:8 mm
that fails at a higher hydrogen content.
Both explanations are able to account for the inconsistency shown in Fig. 10, at least qualitatively. However, due to the
lack of further experimental evidence for ”cold creep” and to the scattering of test data in Fig. 13, it is very difficult to quan-
tify either of them, which is out of the scope of current study. From the phenomenological perspective, we propose the com-
bined contribution of these two aspects as the final explanation to the current problem: on one hand, stress redistribution
occurs in high stress concentration region in real tests, which lowers the peak value of axial stress as the driving force; mean-
while, material in CLT actually possesses a higher strength than that specified by the hydrogen degradation law which is cal-
ibrated in the slow strain rate condition. Although minor in the individual case, the combined effect of these two might
produce a visible influence as seen in Fig. 10. For the case of B13 steel with r ¼ 0:1 mm at a smaller remote load (0:9rB ),
the stress level developed at the notch tip is lower than that in B15, hence smaller degrees of stress redistribution; in addi-
tion, since the hydrogen degradation for B13 is not so pronounced as for B15, the deviation caused by the underestimation of
material resistance is less pronounced. This may explain why no divergence is observed in the case of B13.
Finally, it should be noted that although much effort is dedicated to finding a proper explanation for the inconsistency in
the case of B15 with r ¼ 0:1 mm, the results shown in Fig. 10 can still be considered acceptable, in the sense that a good
approximation of the critical C IC value is given and most importantly, in a conservative manner. Considering the good agree-
ments observed in other two cases, we conclude that the proposed uniform hydrogen degradation law gives a satisfactory
prediction of the constant loading condition for this specific material. Meanwhile, a word of caution should be given that the
 H. Yu et al. / Engineering Fracture Mechanics 157 (2016) 56–71 69

loading rate in the loading-up stage and possible stress redistribution at sharp notch or crack tip in the constant-load stage
need to be addressed in order to improve the accuracy of engineering HE failure assessment with the current methodology.

4. Concluding remarks

Experimental data on hydrogen degraded AISI4135 high strength steel [4,6,7] are analyzed with CZM approach. CZM
based cohesive strength is determined for each specimen geometry according to the nominal stress–strain curves yielded
by the notched tensile tests performed on hydrogen pre-charged specimens. Stress driven mass diffusion analysis is con-
ducted to determine the local lattice hydrogen concentration at failure. These two sets of data are then utilized to construct
the hydrogen degradation loci for different specimen geometries. Upon rationalization of the degradation loci, a uniform
hydrogen degradation law is proposed for the current material, whose effectiveness is then verified against the constant
loading tests.
The hydrogen degradation law is proposed on the basis that the hydrogen degradation loci at different levels of stress
triaxiality collapse into one upon normalization with corresponding hydrogen free conditions. Since the effects of geometry
and hydrogen are naturally decoupled during the normalization procedure, the hydrogen degradation law becomes indepen-
dent of geometry (or stress triaxiality), hence taken as the material property. Both cases of notched specimen and smooth
specimen can be unified into this law without resorting to further assumptions such as the weakest link theory. This law
keeps the contribution from stress triaxiality in the HE process, see Rðdc ; gÞ in Eq. (6), which is analogous to the triaxiality
dependence in a ductile failure scenario. It is therefore claimed that certain amount of contribution from ductile failure
mechanism is implied by the current model, which agrees better with the microscopic observation where some ductile fea-
tures were present during the entire failure process for low hydrogen concentrations.
The uniform hydrogen degradation law is then built into the CZM framework and used to predict the incubation time for
delayed fracture and the critical initial hydrogen concentration C IC at a given constant load level, which is similar to the engi-
neering application of subsea equipment with cathodic protection. Excellent agreement is observed for two of the three cases
examined; larger deviations from the test results, in a conservative manner, are seen in the case of B15 with r ¼ 0:1 mm,
which is qualitatively attributed to combined effects of the stress redistribution at crack tip during the constant loading stage
and of the strain rate strengthening during the loading-up stage. Despite the deviations in that case, general conclusions are
made that satisfactory predictions are achieved with the proposed methodology and that the effectiveness of the uniform
hydrogen degradation law is verified.
As discussed in last section, there exists some difference between the constant loading test and the tensile test in that
stress relaxation is possible for the former situation, especially at large levels of stress concentration, i.e. sharp notches
and cracks; furthermore, distinct strain rates during the loading-up stage may exert certain influences as well. It is thus
important to account for these issues in a quantitative manner in order to improve the transferability of the hydrogen degra-
dation laws from the slow strain rate tensile tests to the constant loading conditions, which is one aspect of the future work.
In addition, the current three-step approach is only a consequential methodology since the updated stress information due
to CZM process cannot be communicated backward to influence hydrogen diffusion. Therefore, another aspect of our future
work is to develop a framework where the hydrogen diffusion and the CZM process are fully coupled.

Acknowledgements

The financial support from Aker Solutions and NTNU via the Integrity of Ni-Alloys for Subsea Applications (INASA) project
is greatly acknowledged. We also want to thank the Research Council of Norway for funding through the Hydrogen-induced
degradation of offshore steels in ageing infrastructure – models for prevention and prediction (HIPP). Contract No. 234130/
E30.

References

[1] Olden V, Thaulow C, Johnsen R, stby E, Berstad T. Influence of hydrogen from cathodic protection on the fracture susceptibility of 25 sent testing and fe-
modelling using hydrogen influenced cohesive zone elements. Eng Fract Mech 2009;76:827–44.
[2] Alvaro A, Olden V, Akselsen OM. 3d cohesive modelling of hydrogen embrittlement in the heat affected zone of an x70 pipeline steel part ii. Int J
Hydrogen Energy 2014;39:3528–41.
[3] Hardie D, Liu S. The effect of stress concentration on hydrogen embrittlement of a low alloy steel. Corrosion Sci 1996;38:721–33.
[4] Wang M, Akiyama E, Tsuzaki K. Effect of hydrogen and stress concentration on the notch tensile strength of aisi 4135 steel. Mater Sci Eng A
2005;398:37–46.
[5] Wang M, Akiyama E, Tsuzaki K. Crosshead speed dependence of the notch tensile strength of a high strength steel in the presence of hydrogen. Scripta
Mater 2005;53:713–8.
[6] Wang M, Akiyama E, Tsuzaki K. Determination of the critical hydrogen concentration for delayed fracture of high strength steel by constant load test
and numerical calculation. Corrosion Sci 2006;48:2189–202.
[7] Wang M, Akiyama E, Tsuzaki K. Effect of hydrogen on the fracture behavior of high strength steel during slow strain rate test. Corrosion Sci
2007;49:4081–97.
[8] Moro I, Briottet L, Lemoine P, Andrieu E, Blanc C, Odemer G. Hydrogen embrittlement susceptibility of a high strength steel x80. Mater Sci Eng A
2010;527:7252–60.
[9] Nanninga NE, Levy YS, Drexler ES, Condon RT, Stevenson AE, Slifka AJ. Comparison of hydrogen embrittlement in three pipeline steels in high pressure
gaseous hydrogen environments. Corrosion Sci 2012;59:1–9.
70 H. Yu et al. / Engineering Fracture Mechanics 157 (2016) 56–71

[10] Serebrinsky S, Carter EA, Ortiz M. A quantum-mechanically informed continuum model of hydrogen embrittlement. J Mech Phys Solids
2004;52:2403–30.
[11] Anvari M, Scheider I, Thaulow C. Simulation of dynamic ductile crack growth using strain-rate and triaxiality-dependent cohesive elements. Eng Fract
Mech 2006;73:2210–28.
[12] Olden V, Alvaro A, Akselsen OM. Hydrogen diffusion and hydrogen influenced critical stress intensity in an api x70 pipeline steel welded joint
experiments and fe simulations. Int J Hydrogen Energy 2012;37:11474–86.
[13] Ayas C, Deshpande VS, Fleck NA. A fracture criterion for the notch strength of high strength steels in the presence of hydrogen. J Mech Phys Solids
2014;63:80–93.
[14] Sadananda K, Vasudevan AK. Review of environmentally assisted cracking. Metall Mater Trans A 2011;42:279–95.
[15] Oriani RA. A mechanistic theory of hydrogen embrittlement of steels. Berichte der Bunsengesellschaft fr physikalische Chemie 1972;76:848–57.
[16] Birnbaum HK, Sofronis P. Hydrogen-enhanced localized plasticity a mechanism for hydrogen-related fracture. Mater Sci Eng A 1994;176:191–202.
[17] Hoagland R, Heinisch H. An atomic simulation of the influence of hydrogen on the fracture behavior of nickel. J Mater Res 1992;7:2080–8.
[18] Jiang DE, Carter EA. First principles assessment of ideal fracture energies of materials with mobile impurities: implications for hydrogen embrittlement
of metals. Corrosion Rev 2004;52:4801–7.
[19] Lu G, Zhang Q, Kioussis N, Kaxiras E. Hydrogen-enhanced local plasticity in aluminum: an ab initio study. Phys Rev Lett 2001;87:095501. PRL.
[20] Wada M, Akaiwa N, Mori T. Field evaporation of iron in neon and in hydrogen and its rate-controlling processes. Philos Mag A 1987;55:389–403.
[21] Lee TC, Robertson IM, Birnbaum HK. An hvem in situ deformation study of nickel doped with sulfur. Acta Metall 1989;37:407–15.
[22] Abraham D, Altstetter C. Hydrogen-enhanced localization of plasticity in an austenitic stainless steel. Metall Mater Trans A 1995;26:2859–71.
[23] Barnoush A, Vehoff H. Electrochemical nanoindentation: a new approach to probe hydrogen/deformation interaction. Scripta Mater 2006;55:195–8.
[24] Ahn DC, Sofronis P, Dodds Jr R. Modeling of hydrogen-assisted ductile crack propagation in metals and alloys. Int J Fract 2007;145:135–57.
[25] Robertson I, Sofronis P, Nagao A, Martin ML, Wang S, Gross DW, et al. Hydrogen embrittlement understood. Metall Mater Trans B 2015;46:1085–103.
[26] Liang Y, Ahn DC, Sofronis P, Dodds Jr RH, Bammann D. Effect of hydrogen trapping on void growth and coalescence in metals and alloys. Mech Mater
2008;40:115–32.
[27] Enos DG, Scully JR. A critical-strain criterion for hydrogen embrittlement of cold-drawn, ultrafine pearlitic steel. Metall Mater Trans A
2002;33:1151–66.
[28] Kim N-H, Oh C-S, Kim Y-J, Yoon K-B, Ma Y-H. Comparison of fracture strain based ductile failure simulation with experimental results. Int J Pressure
Vessels Piping 2011;88:434–47.
[29] Wang M-Q, Akiyama E, Tsuzaki K. Fracture criterion for hydrogen embrittlement of high strength steel. Mater Sci Technol 2006;22:167–72.
[30] Hagihara Y, Ito C, Hisamori N, Suzuki H, Takai K, Akiyama E. Evaluation of delayed fracture characteristics of high strength steel based on csrt method.
Tetsu-to-Hagane 2008;94:215–21.
[31] Needleman A. An analysis of decohesion along an imperfect interface. Int J Fract 1990;42:21–40.
[32] Needleman A. Some issues in cohesive surface modeling. Proc IUTAM 2014;10:221–46.
[33] Park K, Paulino GH. Computational implementation of the ppr potential-based cohesive model in abaqus: educational perspective. Eng Fract Mech
2012;93:239–62.
[34] Pezzotta M, Zhang ZL, Jensen M, Grande T, Einarsrud MA. Cohesive zone modeling of grain boundary microcracking induced by thermal anisotropy in
titanium diboride ceramics. Comput Mater Sci 2008;43:440–9.
[35] Pezzotta M, Zhang ZL. Effect of thermal mismatch induced residual stresses on grain boundary microcracking of titanium diboride ceramics. J Mater Sci
2010;45:382–91.
[36] Ren XB, Zhang ZL, Nyhus B. Effect of residual stress on cleavage fracture toughness by using cohesive zone model. Fatigue Fract Eng Mater Struct
2011;34:592–603.
[37] Song SH, Paulino GH, Buttlar WG. A bilinear cohesive zone model tailored for fracture of asphalt concrete considering viscoelastic bulk material. Eng
Fract Mech 2006;73:2829–48.
[38] Olden V, Thaulow C, Johnsen R, stby E, Berstad T. Application of hydrogen influenced cohesive laws in the prediction of hydrogen induced stress
cracking in 25 steel. Eng Fract Mech 2008;75:2333–51.
[39] Tvergaard V, Hutchinson JW. The relation between crack growth resistance and fracture process parameters in elastic-plastic solids. J Mech Phys Solids
1992;40:1377–97.
[40] Scheider I, Brocks W. Simulation of cupcone fracture using the cohesive model. Eng Fract Mech 2003;70:1943–61.
[41] Siegmund T, Brocks W. A numerical study on the correlation between the work of separation and the dissipation rate in ductile fracture. Eng Fract
Mech 2000;67:139–54.
[42] Banerjee A, Manivasagam R. Triaxiality dependent cohesive zone model. Eng Fract Mech 2009;76:1761–70.
[43] Liang Y, Sofronis P. Toward a phenomenological description of hydrogen-induced decohesion at particle/matrix interfaces. J Mech Phys Solids
2003;51:1509–31.
[44] Alvaro A, Olden V, Akselsen OM. 3d cohesive modelling of hydrogen embrittlement in the heat affected zone of an x70 pipeline steel. Int J Hydrogen
Energy 2013;38:7539–49.
[45] Olden V, Thaulow C, Johnsen R. Modelling of hydrogen diffusion and hydrogen induced cracking in supermartensitic and duplex stainless steels. Mater
Des 2008;29:1934–48.
[46] Alvaro A, Thue Jensen I, Kheradmand N, Lvvik OM, Olden V. Hydrogen embrittlement in nickel, visited by first principles modeling, cohesive zone
simulation and nanomechanical testing. Int J Hydrogen Energy 2015.
[47] Kim Y, Chao YJ, Pechersky M, Morgan M. On the effect of hydrogen on the fracture toughness of steel. Int J Fract 2005;134:339–47.
[48] Gangloff RP. Hydrogen assisted cracking of high strength alloys. Report, DTIC Document; 2003.
[49] A.U. Manual, Version 6.14-1, Dassault Systmes Simulia Corp., Providence, RI; 2014.
[50] Brocks W, Arafah D, Madia M. Exploiting symmetries of fe models and application to cohesive elements. Milano/Kiel; 2013.
[51] Needleman A. A continuum model for void nucleation by inclusion debonding. J Appl Mech 1987;54:525–31. http://dx.doi.org/10.1115/1.3173064.
[52] Gao YF, Bower AF. A simple technique for avoiding convergence problems in finite element simulations of crack nucleation and growth on cohesive
interfaces. Modell Simul Mater Sci Eng 2004;12:453.
[53] Scheider I. Cohesive model for crack propagation analyses of structures with elasticplastic material behavior foundations and implementation. GKSS
Research Center, Geesthacht; 2001.
[54] Sofronis P, McMeeking RM. Numerical analysis of hydrogen transport near a blunting crack tip. J Mech Phys Solids 1989;37:317–50.
[55] Oriani RA. The diffusion and trapping of hydrogen in steel. Acta Metall 1970;18:147–57.
[56] Krom AHM, Koers RWJ, Bakker A. Hydrogen transport near a blunting crack tip. J Mech Phys Solids 1999;47:971–92.
[57] Taha A, Sofronis P. A micromechanics approach to the study of hydrogen transport and embrittlement. Eng Fract Mech 2001;68:803–37.
[58] Dadfarnia M, Martin ML, Nagao A, Sofronis P, Robertson IM. Modeling hydrogen transport by dislocations. J Mech Phys Solids 2015;78:511–25.
[59] Siegmund T, Brocks W. Prediction of the work of separation and implications to modeling. Int J Fract 1999;99:97–116.
[60] Liu J. Eakest link theory and multiaxial criteria, vol. 25. Elsevier; 1999. p. 55–68. <http://www.sciencedirect.com/science/article/pii/
S1566136999800077>. doi: http://dx.doi.org/10.1016/S1566-1369(99)80007-7.
[61] Batdorf SB, Crose JG. A statistical theory for the fracture of brittle structures subjected to nonuniform polyaxial stresses. J Appl Mech 1974;41:459–64.
http://dx.doi.org/10.1115/1.3423310.
[62] Batdorf SB. Some approximate treatments of fracture statistics for polyaxial tension. Int J Fract 1977;13:5–11.
[63] Ayas C, Fleck NA, Deshpande VS. Hydrogen embrittlement of a bimaterial. Mech Mater 2015;80(Part B):193–202.
 H. Yu et al. / Engineering Fracture Mechanics 157 (2016) 56–71 71

[64] Zhang ZL, Niemi E. Studies on the ductility predictions by different local failure criteria. Eng Fract Mech 1994;48:529–40.
[65] Bassani JL, McClintock FA. Creep relaxation of stress around a crack tip. Int J Solids Struct 1981;17:479–92.
[66] Garcia-Granada AA, Lacarac VD, Holdway P, Smith DJ, Pavier MJ. Creep relaxation of residual stresses around cold expanded holes. J Eng Mater Technol
2000;123:125–31. http://dx.doi.org/10.1115/1.1310305.
[67] Nuez JE, Glinka G. Analysis of non-localized creep induced strains and stresses in notches. Eng Fract Mech 2004;71:1791–803.
[68] Fishgoit AV, Kolachev BA. A model of subcritical crack growth under the action of cold creep and dissolved hydrogen. Soviet Mater Sci: A Transl Fiziko-
khimicheskaya mekhanika materialov/ Acad Sci Ukrainian SSR 1986;21:295–7.
[69] Brookfield DJ, Moreton DN, Moffat DG. Shakedown and cold creep of stainless steel type 316 torispherical drumheads subjected to internal pressure. J
Pressure Vessel Technol 1986;108:289–96. http://dx.doi.org/10.1115/1.3264788.
[70] Ellison EG, Harper MP. Creep behaviour of components containing cracks a critical review. J Strain Anal Eng Des 1978;13:35–51.
[71] Thomas RS, Scully J, Gangloff R. Internal hydrogen embrittlement of ultrahigh-strength aermet 100 steel. Metall Mater Trans A 2003;34:327–44.