Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences
You have accessReview article

A review of cohesive zone modelling as an approach for numerically assessing hydrogen embrittlement of steel structures

L. Jemblie

L. Jemblie

Department of Engineering Design and Materials, Norwegian University of Science and Technology, 7491 Trondheim, Norway

Google Scholar

Find this author on PubMed

, and
O. M. Akselsen

O. M. Akselsen

Department of Engineering Design and Materials, Norwegian University of Science and Technology, 7491 Trondheim, Norway

SINTEF Materials and Chemistry, 7456 Trondheim, Norway

Google Scholar

Find this author on PubMed

    Abstract

    Simulation of hydrogen embrittlement (HE) requires a coupled approach; on one side, the models describing hydrogen transport must account for local mechanical fields, while, on the other side, the effect of hydrogen on the accelerated material damage must be implemented into the model describing crack initiation and growth. This study presents a review of coupled diffusion and cohesive zone modelling as a method for numerically assessing HE of a steel structure. While the model is able to reproduce single experimental results by appropriate fitting of the cohesive parameters, there appears to be limitations in transferring these results to other hydrogen systems. Agreement may be improved by appropriately identifying the required input parameters for the particular system under study.

    This article is part of the themed issue ‘The challenges of hydrogen and metals’.

    1. Introduction

    Hydrogen-induced degradation of mechanical properties, often termed hydrogen embrittlement (HE), is a well-recognized threat for structural steels. It manifests as loss in ductility, strength and toughness, which may result in unexpected and premature catastrophic failures. The phenomenon was first reported by Johnson [1], and was later extensively researched both experimentally [216], yielding a number of models accounting for the phenomenon. However, no consensus about the basic mechanisms responsible for HE has been reached yet. Two theories have been advanced as the more accepted ones for the case of hydrogen degradation in steel: hydrogen-enhanced decohesion (HEDE), in which interstitial atomic hydrogen reduces the bond strength and thus the necessary energy to fracture [17,18]; and hydrogen-enhanced localized plasticity (HELP), in which atomic hydrogen accelerates dislocation mobility through an elastic shielding effect which locally reduces the shear stress [19,20]. Today, it is seemingly recognized that no single mechanism can comprehensively explain all the phenomena associated with HE. Rather it appears that different mechanisms apply to different systems, and that a combination of mechanisms is more likely in many cases.

    In recent years, cohesive zone modelling (CZM) has gained increasing interest as a suitable method for modelling HE [1012,14,16], with the possibility of providing increased understanding of the involved process and their interactions combined with reduced time and costs compared with experimental programmes. The damage process is classically described by interface elements, the constitutive relation of which is defined by a cohesive law (traction separation law). Simulation of hydrogen-induced degradation requires a coupled approach, including modelling of transient mass transport, plastic deformation, fracture and their interactions. On one side, the models describing hydrogen diffusion must account for local mechanical field quantities; i.e. hydrostatic stress and plastic strain. On the other side, the effect of hydrogen on the accelerated material damage must be implemented into the cohesive law.

    The present work reviews CZM as a method for numerically assessing the HE susceptibility of steel structures. Especially approaches for implementing hydrogen influence the cohesive model and the coupling aspect between hydrogen transport and CZM is presented, followed by a discussion on some practical applications.

    2. Hydrogen transport

    To date, models of transient hydrogen diffusion generally account for trapping by dislocations and hydrostatic drift. Recent approaches include capturing the effect of multiple trap sites and hydrogen transport by dislocations [13,21,22]. An extensive review of hydrogen transport models is beyond the scope of this work, and only a short summary covering the most governing aspects is given. For a more thorough description, the reader is referred to Sofronis and co-workers [8,21,22] and Krom et al. [9].

    (a) Trapping of hydrogen

    Atomic hydrogen is generally considered to either reside at normal interstitial lattice sites (NILS) or be trapped at microstructural defects such as dislocations, carbides, grain boundaries and interfaces. Given a metal lattice, the hydrogen concentration in NILS is given by [8]

    Display Formula
    2.1
    where θL is the lattice site occupancy, NL is the density of solvent atoms and β is the number of NILS per solvent atom. Similarly, the concentration of hydrogen trapped at a given site is [8]
    Display Formula
    2.2
    where θT is the occupancy, NT is the density of the specific trap site (dislocation, carbide, etc.) and α is the number of sites per trap. According to Oriani’s theory [23], hydrogen in NILS and hydrogen in reversible trapping sites are always in local equilibrium, such that
    Display Formula
    2.3

    with EB being the trap binding energy. This approach is valid in the domain of rapid trap filling and escape kinetics.

    For microstructural defects such as carbides and grain boundaries, the trap densities are often assumed constant throughout the material. For dislocations, however, the trap density varies pointwise depending on the local plastic strain. Based on experimental work by Kumnick & Johnson [2] on hydrogen trapping in deformed iron, Sofronis & McMeeking [8] proposed the following relationship between the dislocation trap density and the equivalent plastic strain εp:

    Display Formula
    2.4
    An alternative theoretical approach has been proposed by Sofronis and co-workers [24,25], assuming one trap site per atomic plane threaded by a dislocation, maintaining that this is consistent with the experimental work of Thomas [26]. The dislocation trap density is then expressed as a function of the dislocation density ρ and the lattice parameter a
    Display Formula
    2.5
    The dislocation density (measured in dislocation line length per cubic metre) is considered to vary linearly with the equivalent plastic strain according to
    Display Formula
    2.6
    where ρ0=1010 line length m−3 denotes the dislocation density at zero plastic strain, and γ=2.0×1016 line length m−3. Using the lattice parameter of BCC iron a=2.86 Å , the trap densities according to the data from Kumnick & Johnson [2] and the model by Sofronis and co-workers [24,25] are compared in figure 1. It can be concluded that the model by Sofronis and co-workers yields a dislocation trap density about three orders of magnitude larger than the data by Kumnick and Johnson.
    Figure 1.

    Figure 1. Dislocation trap densities according to the work by Kumnick & Johnson [2] (dashed lines) and the model by Sofronis and co-workers [24,25] (solid lines). In calculating CT, it is assumed that αθT=1, which accordingly gives the maximum possible hydrogen concentration trapped at dislocations.

    (b) Hydrogen diffusion

    The governing model for hydrogen transport, as developed by Sofronis & McMeeking [8], yields

    Display Formula
    2.7
    where
    Display Formula
    2.8
    is the hydrogen flux of diffusion particles through NILS, motivated by chemical potential gradients. Inline Formula is the partial molar volume of hydrogen and p is the hydrostatic pressure. Despite their small size, dissolved hydrogen atoms induce a distortion in the steel lattice, resulting in the formation of hydrostatic compressive stresses and volume dilation [27]. The chemical potential of hydrogen is therefore lowered in regions of tensile hydrostatic stresses, and consequently a hydrogen flux is generated towards these regions. The opposite effect occurs on encountering compressive hydrostatic stresses. Substituting equation (2.8) in (2.7), under the assumption of Oriani’s theory of equilibrium [23], provides the governing model for the transient hydrogen transport model, derived by Sofronis & McMeeking [8] and later modified by Krom et al. [9]:
    Display Formula
    2.9
    This model accounts for trapping by dislocations and hydrostatic drift. The last term is the plastic strain rate factor, accounting for the effect of the strain rate on the transient hydrogen concentrations. It disappears in the absence of dislocation trap sites (dNT/dεp=0).

    3. A cohesive zone modelling approach to hydrogen embrittlement

    Cohesive models were first formulated by Barenblatt [28] and Dugdale [29], who introduced finite nonlinear cohesive tractions in front of an existing crack as a means to overcome the crack tip stress singularity. To date, the cohesive model is extensively applied for crack propagation analysis using the finite-element method. Among the various approaches available, it is appealing in that it requires few parameters and in its universality of applicability [30].

    (a) The cohesive model

    The cohesive theory of fracture is a purely phenomenological continuum framework, not representative of any physical material. The constitutive response of the material is divided into two parts: an arbitrary material law relating the stresses and strains in the bulk regions adjacent to the crack faces, and a cohesive law characterizing the separation process by describing the forces opposing crack formation (tractions) as a function of the incipient crack surfaces’ separation distance. Common to most cohesive laws is that they can be described by two independent parameters out of the following three: the cohesive strength σC, the critical separation δC and the cohesive energy ΓC. Figure 2 displays three commonly applied cohesive laws, plotted as normalized traction versus separation: a linear decreasing law suggested for brittle materials by Hillerborg et al. [31], a polynomial law suggested by Needleman [32] for ductile materials and, more recently, a versatile trapezoidal law suggested by Scheider [33] also for ductile materials. The area surrounded by the curve represents the cohesive energy. A more thorough compilation of cohesive laws can be found in literature, e.g. Shet & Chandra [34] or Brocks et al. [35].

    Figure 2.

    Figure 2. Cohesive laws by Hillerborg et al. [31] (solid line), Needleman [32] (long dashed line) and Scheider [33] (short dashed line).

    An intrinsic disconnection exists between atomistic and engineering cohesive descriptions, where the fundamental formulation by Barenblatt [28] is equivalent to the atomistic conception of the cohesive zone. Typically, the work of separation differs by orders of magnitude, suggesting the engineering description contains elements of the plastic work of fracture. In CZM, the cohesive energy can physically be understood as the total energy dissipated by the cohesive element during separation.

    The influence of the shape of the cohesive law on the results is controversial; while Scheider & Brocks [36] found significant effect on their calculated results, Tvergaard & Hutchinson [37] concluded that such an influence is negligible. Irrespective, the cohesive law has to be chosen in relation to the actual micromechanical damage mechanism leading to failure. Values of the cohesive parameters should be chosen so that they do not affect the overall compliance of the system [15]. Alvaro et al. [15] point out the importance of this in relation to modelling HE. A choice of cohesive parameters which infers low values of the initial stiffness will result in lower values of hydrostatic stress and equivalent plastic strain, consequently affecting the lattice and trapped hydrogen populations.

    Despite cohesive zone simulation being straightforward, it has limitations when it comes to modelling crack nucleation, failing to produce a converged solution at the point where the crack first nucleates. These problems, which are especially prominent in performing a coupled hydrogen transport and cohesive analysis, are attributed to a snap-back instability that occurs just after the stress reaches the peak strength of the interface [38]. Gao & Bower [38] found that adding a small viscosity term in the cohesive relation significantly increases the numerical stability. Yu et al. [39] have applied the viscosity term by Gao & Bower [38] in a three-step, un-coupled, hydrogen-informed cohesive zone model under constant displacement, and found the viscous regularization to be effective in solving the convergence problem with good accuracy. In relation to performing a coupled hydrogen transport and cohesive analysis, there is still some uncertainty as to whether a model containing this viscosity term is able to accurately predict the time to fracture.

    (b) Implementing hydrogen influence

    Most of the known attempts to implement hydrogen influence in the cohesive model are through the HEDE principle [11,15,16,4042]—hydrogen reduction of the cohesive energy at fracture. In its most simplistic approach, the critical hydrogen-dependent cohesive stress σC(C) is assumed to decrease linearly with increasing hydrogen concentration

    Display Formula
    3.1
    where σC(0) is the critical cohesive stress with no hydrogen influence and ξ is a softening parameter, often found by fitting to experimental results [4244]. This formulation predicts a hydrogen-influenced fracture toughness KIC=0 and, thus, complete decohesion upon the attainment of a certain critical hydrogen concentration.

    In recent years, quantum-mechanical approaches by first principle calculations have been increasingly used to quantify the effect of hydrogen on decohesion [45,46]. A key factor is that hydrogen strongly prefers to stay on the surface rather than in the bulk, which provides a driving force for decohesion and, thereby, embrittlement. Using an equilibrium thermodynamic description, Van der Ven & Ceder [45] have obtained a complete set of traction-separation curves for decohesion along the Al(111) planes with a hydrogen coverage between 0 and 1 (1 representing the saturation value). The results are displayed in figure 3a, which reveals a decrease in the cohesive energy with increasing hydrogen coverage. The critical separation, however, was found to be insensitive to hydrogen throughout the given range.

    Figure 3.

    Figure 3. Hydrogen effect on decohesion by quantum-mechanical approaches. (a) Traction separation curves for decohesion along Al(111) planes with a hydrogen coverage between 0 and 1, by Van der Ven & Ceder [45]. (b) Cleavage energy for decohesion along Al(111) and Fe(110) as a function of hydrogen coverage, by Jiang & Carter [46].

    Jiang & Carter [46] have calculated the ideal cleavage energy (equal to twice the surface energy, γ) of Fe and Al in the presence of various amounts of hydrogen within the framework of a Born–Haber thermodynamic cycle. The main idea is that hydrogen dissolved in metals quickly segregates to the incipient crack surfaces as a crack begins to form. An almost linear decrease in cleavage energy with increasing hydrogen coverage is observed for both Al(111) and Fe(110), as displayed in figure 3b. A fit to the data for the H/Fe system yields [11]

    Display Formula
    3.2
    where θH is the surface hydrogen coverage, γ(θH) is the hydrogen-dependent surface energy and γ(0) is the surface energy with no hydrogen influence. The data fit is illustrated by the red line in figure 3b. The definition of hydrogen coverage follows the Langmuir–McLean isotherm [47], relating it to the bulk hydrogen concentration C (unit mol H/mol Fe) through
    Display Formula
    3.3
    where Inline Formula is the Gibbs energy difference between surface and bulk material, surface being any microstructural interface like a crystallographic plane, grain boundary, etc. The hydrogen coverage as a function of hydrogen concentration is plotted in figure 4 for various levels of Gibbs energy ranging between 10 and 60 kJ mol−1. It is evident that a given value of Gibbs energy covers a concentration range of about four orders of magnitude, where the lower bound represents a hydrogen concentration threshold for embrittlement and the upper bound represents a corresponding saturation level.
    Figure 4.

    Figure 4. Hydrogen coverage as a function of hydrogen concentration, for various levels of Gibbs energy (kJ mol−1). Plotted according to the Langmuir–McLean isotherm [47].

    Extrapolation of nanometre-scale quantum-mechanical calculations to macroscopic-scale continuum models entails some difficulty. Atomistic predictions of peak stresses are of the order of the theoretical strength of the crystal, while opening displacements are only a few angstroms [11,48,49]. Further, the cohesive zone sizes attendant to first principle calculations are on the nanometre scale, making finite-element calculations unfeasible, as the mesh must fully resolve the cohesive zone in order to obtain a converged solution. Using a renormalization procedure described by Nguyen & Ortiz [48] and Hayes et al. [49] to scale the atomic-level cohesive properties up to the continuum scale, Serebrinsky et al. [11] have developed a cohesive model of fracture, accounting for the effect of hydrogen segregation by a quantum-mechanical treatment. Based on the relation in equation (3.2), the following coupling between hydrogen coverage and the critical hydrogen-dependent cohesive stress σC(θH) is suggested for BCC iron [11]:

    Display Formula
    3.4
    The critical separation δC is deemed constant, insensitive to the hydrogen coverage, based on the results from Van der Ven & Ceder [45] in figure 3a. The influence of hydrogen, in terms of hydrogen coverage, on the cohesive strength and consequently on the cohesive energy is illustrated in figure 5a for the polynomial cohesive law by Needleman [32]. Using the coupling between hydrogen coverage and bulk concentration as supplied by the Langmuir–McLean isotherm, Serebrinsky et al. [11] suggested ΔG0b=30 kJ mol−1, which represents the trapping energy of hydrogen at an Fe grain boundary, yielding a threshold concentration of about 0.001 wppm and an embrittlement saturation level of about 5 wppm. Hence, a concentration level close to the theoretical solubility of hydrogen in iron (about 10−4 wppm [50]) should not induce any effect on the cohesive properties, implying the importance of trapped hydrogen.
    Figure 5.

    Figure 5. Reduction in cohesive energy at different levels of hydrogen coverage for the polynomial cohesive law by Needleman [32], where (a) illustrates hydrogen influence on the cohesive strength only (single) and (b) illustrates hydrogen influence on both the cohesive strength and the critical separation (double).

    Raykar et al. [42] have proposed a hydrogen damaging effect both on the cohesive strength and on the critical separation, basing this choice on the experimentally observed reduction in both ultimate tensile strength and percentage elongation in the presence of hydrogen. A linear dependence on hydrogen concentration according to equation (3.1) was chosen for both parameters. The same approach has also been applied by Gobbi et al. [51], however, here with a hydrogen dependence according to the work by Serebrinsky et al. [11]. Figure 5b illustrates hydrogen influence, in terms of hydrogen coverage, according to equation (3.4), on both the cohesive strength and the critical separation, for the polynomial cohesive law by Needleman [32]. A comparison of the effect of single and double hydrogen influence on the critical cohesive energy at fracture is made in figure 6, where the cohesive energy is plotted as a function of the hydrogen coverage for the two cases in figure 5, displaying an enhanced hydrogen damaging effect with double hydrogen influence. Although this approach was found to display a reasonable fit with experimental data [42,51], no quantification of any effect of hydrogen on the critical separation is found to date.

    Figure 6.

    Figure 6. Relationship between critical cohesive energy at fracture and hydrogen coverage for the polynomial cohesive law by Needleman [32]. Single hydrogen influence denotes hydrogen reduction of the critical cohesive stress. Double hydrogen influence denotes hydrogen reduction of the critical cohesive stress and of the critical separation.

    Liang & Sofronis [10] have proposed an alternative model for hydrogen decohesion, based on work by, among others, Rice & Wang [52,53], who estimated the effect of segregated hydrogen on interface cohesion from a general thermodynamic framework. The resulting hydrogen-dependent cohesive strength is expressed for two limiting cases of interfacial separation: separation at constant hydrogen concentration (denoted as fast separation) given by equation (3.5), and separation at constant hydrogen chemical potential (denoted as slow separation) given by equation (3.6),

    Display Formula
    3.5
    and
    Display Formula
    3.6
    where Inline Formula and Inline Formula are the Gibbs energy of segregation for the interface and free surface, respectively, Γ/Γmax is the interfacial hydrogen coverage and Inline Formula. The proposed model was used to simulate separation along a chromium carbide/FCC matrix (nickel alloy 690) interface. The resulting range of polynomial cohesive laws [32] for various interfacial hydrogen coverage values is presented in figure 7. Using parameters representative of Fe (110), (2γint)0=4.86 J m−2 and Γmax=5.85×10−5 mol m−2 [46], assuming Inline Formula [13], the hydrogen-dependent cohesive stress for the fast separation case can be estimated. The result is plotted as the green dotted line in figure 3b, representing a good fit with the result by Jiang & Carter [46].
    Figure 7.

    Figure 7. Hydrogen-influenced cohesive laws from the decohesion model by Liang & Sofronis [10]; Inline Formula is the normal traction and q is a non-dimensional separation parameter.

    (c) Coupling of diffusion and mechanical models

    The Langmuir–McLean isotherm defines the necessary coupling between the hydrogen diffusion model in §2 and the hydrogen-dependent cohesive law described in the previous section. The coupling takes place in two ways: first, hydrogen accelerates material damage by building up over the cohesive zone, as indicated by equations (3.3) and (3.4). Second, hydrogen transport is influenced by the local hydrostatic stress and plastic strain fields, according to equation (2.9).

    Experimental results investigating the effect of hydrogen on fracture generally display a weaker effect of hydrogen with increasing concentration [4,6,13]. Thomas et al. [4] found that the threshold stress intensity factor for HE in AERMET 100 steel decreased sharply with an increasing diffusible hydrogen concentration up to 2 wppm, and more modestly with higher concentrations. The result is displayed in figure 8 for a normalized threshold stress intensity factor, together with the normalized hydrogen-dependent cohesive stress according to the linear model in equation (3.1) and the model by Serebrinsky et al. [11], with the hydrogen concentration calculated according to the Langmuir–McLean isotherm for ΔG0b=30 kJ mol−1. The model by Serebrinsky et al. [11] captures the exponential embrittlement effect of hydrogen, attaining a saturation level at high concentrations. The linear model, fitted to the initial part of the experimental data, gives a reasonable approximation at low concentrations only. The results confirm the necessity of a saturating HE law, as also pointed out by Serebrinsky et al. [11].

    Figure 8.

    Figure 8. Normalized threshold stress intensity factor and normalized hydrogen-dependent cohesive stress as a function of hydrogen concentration, according to experimental data by Thomas et al. [4] (diamonds), the linear decohesion model (dashed line) and the exponential decohesion model by Serebrinsky et al. [11] (solid line), with ΔG0b=30 kJ mol−1.

    4. Practical applications of the coupled cohesive model

    The capability of the model to reliably predict hydrogen-induced crack nucleation and propagation in structural steel applications is of key importance for further developments. An engineering tool, able to partly replace time-consuming and costly experimental programmes, should be of general validity and provide robustness and transferability to other material systems and environments. While most studies are able to reproduce single experimental results by appropriate fitting to the cohesive parameters, there still appears to be limitations on transferring these results to other hydrogen systems [15,16]. Moriconi et al. [16] have developed a cohesive model based on coupled effects between mechanical cyclic loading and hydrogen diffusion. Simulated fatigue crack growth was compared with experimental measurements on martensitic stainless steel under gaseous hydrogen. The results indicate that, while the model was able to reasonably predict the fatigue crack growth behaviour under low hydrogen pressure, it failed to account for the enhanced crack growth observed at high pressures. Limitations in the model, particular in the case of lattice diffusion, were pointed out as possible explanations; however, no conclusions were drawn.

    Dadfarnia et al. [22] have extended the hydrogen transport model by Sofronis & McMeeking [8] and Krom et al. [9] (equation (2.9)) to account for hydrogen transport by dislocations. Moving dislocations represent moving traps that carry hydrogen atoms. Thus, hydrogen is transported by both diffusion through NILS and by mobile dislocations. Results from numerical simulations indicate that dislocation transport can contribute to an elevation of the local hydrogen concentration above levels predicted by the classical diffusion model, with the effect being larger for materials with a lower hydrogen diffusion coefficient and higher dislocation trap binding energy.

    Brocks and co-workers [14,54] have developed a model of hydrogen-induced cracking, which, in addition to the coupled interactions of hydrogen diffusion and reduced cohesive strength, also includes the effect of surface kinetics on hydrogen absorption and hydrogen-induced softening of the local yield strength (HELP mechanism). A thorough description of the model can be found in [14,54]. By including both local hydrogen softening and hydrogen-induced lowering of the local cohesive strength, the model describes an attempt in the direction of including both the HEDE and the HELP degradation mechanisms and their interactions. Simulated fracture resistance curves (CTOD-R) were compared with experimental results on high-strength low-alloy structural steel, with appropriate fitting of the cohesive parameters and their dependence on the lattice hydrogen concentration. The results are displayed in figure 9 for various deformation rates, where the two mid curves (10 μm h−1 and 100 μm h−1) represent real predictions, capturing the rate dependence of the R-curves due to hydrogen diffusion quite well. While numerous experimental measurements are necessary in order to determine the required input parameters, the authors argue the model may, to some extent, replace expensive laboratory testing, especially considering its transferability to other systems by identifying the required parameters.

    Figure 9.

    Figure 9. CTOD-R curves for various deformation rates, comparing experimental tests (symbols) and simulation results (lines). (From Brocks et al. [14].) (Online version in colour.)

    Recent work by the present authors includes coupled hydrogen diffusion and cohesive zone simulations, with the aim of reproducing fracture mechanical tests of bi-metallic compact tension specimens performed in air and under cathodic protection (3.5% NaCl solution at −1050 mV SCE). The developed model is based on the presented work by Serebrinsky et al. [11], with ΔG0b=30 kJ mol−1. A best fit of the cohesive parameters to the experimental results in air was achieved for σC=1188 MPa (3.5σy) and δC=0.005 mm. Taking into account the effect of hydrogen on the critical cohesive stress, the initial lattice hydrogen concentration was varied from 0.00034 to 1 wppm, corresponding to the theoretical solubility of hydrogen in ferrite [50] and a 3% NaCl [11,55] aqueous solution, respectively. The resulting load–CMOD curves for various hydrogen concentrations are displayed in figure 10, where the dotted, vertical lines indicate the points of fracture for the respective concentration. The corresponding traction versus separation curves are compared in figure 11. Fracture is defined as the encountering of divergence, just after the stress reaches the peak strength of the interface.

    Figure 10.

    Figure 10. Load–CMOD curves according to coupled hydrogen diffusion and cohesive simulations of a bi-metallic compact tension specimen, comparing various initial lattice hydrogen concentrations. The vertical lines indicate points of fracture for the respective concentration levels.

    Figure 11.

    Figure 11. Comparing traction versus separation curves for various initial lattice hydrogen concentrations.

    The model is clearly able to account for a decrease in the fracture toughness with increasing hydrogen concentration, while also maintaining the stiffness of the cohesive element. It fails, however, to provide a solution comparable to experimentally measured hydrogen concentrations. For experimental testing under cathodic protection, failure is observed at a CMOD value of approximately 0.3 mm, which according to figure 10 would indicate a lattice hydrogen concentration in the range 0.001–0.01 wppm. This is two to three orders of magnitude below the measured concentration of 1.5–2.5 wppm [55]. Serebrinsky et al. [11] suggested that agreement for high hydrogen concentrations might be improved by considering different adsorption sites at the cracking interface, with a distribution of adsorption energies. Novak et al. [13] reported that high-binding-energy traps cannot account for the loss in strength observed on hydrogen charged steel, because these traps remain saturated with hydrogen regardless of loading conditions and/or hydrogen exposure conditions. Thus, lowering the Inline Formula level could be justified, identifying parameters according to the particular system in question.

    5. Conclusion

    A coupled mass transport and CZM approach for simulating hydrogen-induced cracking is described and discussed. To date, CZM approaches have proved to be able to reproduce single experimental results by appropriate fitting of the cohesive parameters. However, there appears to be limitations in transferring these results to other hydrogen systems. Agreement may be improved by appropriately identifying the required input parameters for the particular system under study.

    Authors' contributions

    All authors have contributed to this review article.

    Competing interests

    The authors declare that they have no competing interests.

    Funding

    The present work was financed by the Research Council of Norway (Petromaks 2 programme, contract no. 234110/E30), Statoil, Gassco, Technip, POSCO and EDF Induction and performed within the ROP project (www.sintef.no/rop).

    Acknowledgements

    The authors gratefully acknowledge the valuable input from Antonio Alvaro, Philippe Mainçon and Vidar Osen.

    Footnotes

    One contribution of 24 to a discussion meeting issue ‘The challenges of hydrogen and metals’.

    Published by the Royal Society. All rights reserved.

    References