Kharin EngFfractMech2023

Engineering Fracture Mechanics 291 (2023) 109473
Contents lists available at ScienceDirect
Engineering Fracture Mechanics

journal homepage: www.elsevier.com/locate/engfracmech
Crack tip hydrogen diffusion with multiple type traps and

implications for hydrogen assisted cracking
Viktor Kharin
FSIRG, University of Salamanca, Campus Viriato, 49022 Zamora, Spain
A R T I C L E I N F O A B S T R A C T
Keywords: Hydrogen effects on fracture are ruled by its amounts in microstructural features that entrap the
Hydrogen assisted cracking species and are actual sites of the embrittlement events. The paper aims to describe hydrogen
Hydrogen embrittlement delivery to diverse concurring microstructural positions basing on comprehensive diffusion-
Hydrogen diffusion
trapping theory that relies on minimal a priori constraints. The operation of multiple type traps
Hydrogen trapping
near a crack in a model steel system that uptakes hydrogen during loading is elucidated. The
calculations validate the postulate of local lattice-trap equilibrium and the dominance of
hydrogen transport by only lattice-site diffuser flow. The simulations exhaustively describe the
functioning of multifarious microstructural features and reveal their potentiality both to carry out
the mechanistic effects of hydrogen and to retard hydrogen accumulation in the fracture nuclei.
1. Introduction
The influence of hydrogen on the mechanical behaviour of metals has been a long-standing problem in metals science. The effect
was first reported in mid 1870s [1,2]. Shortly after that, hydrogen was declared “an enemy of iron and steel, rendering it brittle” [3].
The phenomenon had been receiving moderate academic attention until its impact on structural integrity came to the fore as critical
engineering problem in 1940s [4]. This has been boosting research efforts on the matter since then [4–6]. Hydrogen harms to metals
are known under various names focusing on specific manifestations, such as hydrogen embrittlement (HE), hydrogen assisted fracture/
cracking (HAF/HAC), hydrogen attack, etc. They are conditioned by the alloy composition and microstructure, hydrogen amount in
metal, stress–strain state and temperature [4–6].
Metal fractures in general, and under the action of hydrogen in particular, originate with the aid of crystal imperfections such as
dislocations [7–9], vacancies [10,11], etc. The stress–strain progression drives the evolution of microstructural imperfections towards
a loss of material integrity [7,8], and hydrogen can enhance the capacity of defects to engender and expand fracture [9–12]. Not all the
hydrogen in metal directly affects fracture, but only its part residing in crystal imperfections that are the sites of actual HAF events
[9,12–14].
Hydrogen delivery to damage locations proceeds by diffusion between empty interstices, which are hollows along the energy relief
for hydrogen in metal as schematised in Fig. 1. Not all walls there are identical. Regular lattice, impurities, vacancies, dislocations and
other microstructural features provide dissimilar interstices [14–16] forming a set Ξ of site types. Each site type A ∈ Ξ is characterised
by the energy GA of hydrogen there and by the site volumic concentration NA. They amount to the total site concentration Ntot = Σ NA .
A∈Ξ
Saddle points with energies EAB separate adjacent sites A and B ∈ Ξ. Comparing to the regular lattice sites L ∈ Ξ, crystal imperfections
E-mail address: gatogris@usal.es.
https://doi.org/10.1016/j.engfracmech.2023.109473
Received 3 March 2023; Received in revised form 7 June 2023; Accepted 5 July 2023
Available online 20 July 2023
0013-7944/© 2023 The Author. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
V. Kharin Engineering Fracture Mechanics 291 (2023) 109473
usually provide deeper energy holes for hydrogen, where its mean residence times are longer. This results in uneven hydrogen oc
cupancy of different microstructural states and moderation of hydrogen transport velocity in metal. Such lower-energy states form the
subset Ξ Tr ≡ {Tr∣Tr ∈ Ξ, GTr < GL} of hydrogen traps [14–19]. The amount of hydrogen allocated to the type A interstices is presented
by the partial volumic concentration CA. The total hydrogen concentration in metal is Ctot = Σ CA . The grade of filling of A-type
A∈Ξ
positions with hydrogen is presented by the occupancy θA = CA/NA.
The occupancies of traps determine the effects of hydrogen on the states of respective crystal imperfections giving rise to hydrogen
sensitive material behaviour [9,12,14]. The microstructural features, which directly participate in the actual fracture micromechanism
and ease it with the aid of hydrogen, form the subset of deleterious traps Ξ X ≡ {X∣X ∈ ΞTr, X carries hydrogen harm}. The set of
occupancies {θX∣X ∈ Ξ X} presents the hydrogen amounts that directly control hydrogen’s detrimental effects. Corresponding partial
concentrations CX = θXNX, too, can be controlling variables, but not straightforwardly. Anyway, neither the hydrogen amounts CA
associated with the states A ∈ Ξ\ΞX that do not participate in per se HAF, nor the total one Ctot are relevant. The local criterion of HAF at
some instant of time t in a material point identified by its position vector x is set as the requirement that the stress–strain state, which is
represented by the tensors of stress σ ̂ (x, t) and accumulated plastic strain ̂ε p (x, t), together with hydrogen amounts in responsible
( )
microstructural features θX(x,t) form critical combination σ ε p , 〈θX 〉 cr , where the tuple-like notation 〈…〉 presents the list of variables
̂, ̂
which are the elements of corresponding set {…}. Similarly to the mechanics notions of yield or damage surfaces, this may be
paraphrased in terms of HAF locus as follows:
( )
Ψ σ ε p , 〈θX 〉 = 0
̂, ̂ (1)
Scalar function Ψ of tensor arguments is representable via tensors’ scalar invariants, such as the maximal normal stress σmax , the
∫ eq n
mean (or hydrostatic) stress σ m and the equivalent stress σ eq, together with the cumulative equivalent plastic strain εeq
p = d̄εp , where
(2 )1/2
d̄εeq
p = 3d ε p : d ε p
̂ ̂ is the incremental equivalent plastic strain corresponding to the plastic strain increment tensor d̂
ε p , and the
maximal principal plastic strain εp (the incompressible plasticity is assumed).
HAF is contingent on hydrogen delivery to the deleterious sites X in potential fracture process zone (FPZ) [9,12,13,20]. The
description of hydrogen transport resolving transient partitioning of the species between microstructural features in metal is crucial for
the understanding, gaining predictive capability and devising strategies against HE [9,20–23]. Already the early criteria for hydrogen
resistant microstructures relied upon the trapping [13–15]. Strong innocuous traps were suggested as general remedy for HE
[13,15,23–25]. Though, it was inferred [13,14] that trap roles might vary depending on circumstances. Later considerations [20–22]
supported this. Few works [9,20,22] addressed concurrent operation of several types of traps in relation to HAF, but they provided
limited data about the trap functioning.
This paper aims to delve into the matter of concurrent operation of distinct traps to elucidate their roles in HAF basing on the
comprehensive diffusion-trapping theory [19] that relies on minimal a priori constraints. To this end, cracked steel specimen under
hydrogen charging during constant extension rate test (CERT) is considered as a model system. Specifically, the experiments on HAC in
pearlitic steel [26–31] are simulated as an instructive example.
2. Basement diffusion-trapping theory
2.1. Transport equations
The adopted theory [19] considers diffuser jumps between adjacent interstices along the potential relief (Fig. 1) at uniform
temperature. An external potential field U(x) can additively alter the energy landscape. In a stressed solid, U(x) = − VHσ m(x), where VH
is the partial molar volume of hydrogen in metal. Each interstice provides a room for a single diffuser particle, i.e., all occupancies are
θA ≤ 1 (A ∈ Ξ). The frequency of thermally activated particle jumping from A- towards B-sites (A,B ∈ Ξ) is υAB = ωA exp(− βΔEAB ),
where ωA is the vibration frequency, ΔEAB = EAB − GA is the activation barrier (Fig. 1), and β = (RT)− 1 incorporates the gas constant R
Fig. 1. Schematic potential energy diagram for hydrogen travelling in metal.
2
and the absolute temperature T. The probability of success of any transition into B site is YB = γ B(1 − θB), where γ B = NB/Ntot is the
number fraction of B-sites. The mass transfer rate by A→B jumps is controlled by the partial diffusivity dAB = d0AB exp(− βΔEAB ), where
d0AB = ωA l2AB /6 involves the diffuser jump length lAB . The partial net diffusion flow by A → B jumps JAB reads:
JAB = − dAB (YB ∇CA − CA ∇YB − βVH YB CA ∇σm ) (2)

3
For a diffuser state A in a material element d x of the solid’s spatial domain Ω, the partial mass balance equation is as follows:
∂CA ∑ ∑{ dBA YA ∇ ⋅ (∇CB − βVH CB ∇σm )− }
= (υBA CB YA − υAB CA YB ) + (3)
∂t B∈Ξ
− dAB CA (∇ ⋅ ∇YB + βVH ∇YB ⋅ ∇σ m )
B∈Ξ,
B∕
=A
The partial differential equation (PDE) of total balance is as usual:

∂Ctot
= − ∇ ⋅ Jtot (4)
∂t
where Jtot = Σ JAB is the total flux. The populations of certain type sites A ∈ Ξ, such as the interstices at dislocations and vacancies
A,B∈Ξ
[9,10,20,32–39], may be altered by plastic straining making their concentrations dependent on the cumulative plastic strain, NA =
NA (εeq eq
p ). Thence, the diffusion-trapping depends on both mechanical fields σ m(x,t) and εp (x, t).
Beyond other diffusion theories [17,18,40–42], this one takes account of the local transitions between distinct sites within a
material element at a place x, which appear in the first sum in (3). Thus, this theory is not bound to the local equilibrium [17,43] that
has often been referred to as the Oriani’s postulate. The local equilibrium is equivalent to the kinetics principle of detailed balance [44]
telling that the rates of forward and backward transitions between A- and B-type sites in d3x are equal, i.e., υAB CA YB = υBA CB YA , which
annihilates respective terms of the addressed sum in (3). On the other side, unlike the McNabb-Foster type models [45,46], the second
sum in (3) accounts for the nonlocal transfers (partial flows) JAB between all kind sites in material elements situated at x and x + dx.
2.2. Equilibria in diffusion-trapping
Diffusion-trapping pursues local and, ultimately, global equilibria, when the net transfers of the species become null, respectively,
between different states A, B, … ∈ Ξ at a location x, and up to between all states throughout the whole system. This corresponds to the
thermodynamic equilibrium representable in terms of the chemical potentials μA of hydrogen residing in the sites A [18,19,43,44],
which with account for the vibrational entropy [47] read:
( )
θA
μA (θA ) = GA − VH σ m + β− 1 ln + β− 1 ln(βωA ) (5)
1 − θA
The parity between the chemical potentials μA(x) = μB(x), which means the equilibrium between the states A and B, implies the
equality
θB (x) θA (x)
= KBA (6)
1 − θB (x) 1 − θA (x)
where KBA = ξBA exp(βEBA

b ) is the trapping strength of B-sites relative to A-ones, which depends on the frequency ratio ξBA = ωA/ωB and
the relative binding energy EBA
b = GA − GB . Local equilibrium may be either partial, when (6) holds for some subset {A,B,…} ⊂ Ξ, or
complete, when it does for all site types, i.e., μA(x) = μB(x) = … = μ(x). When ξBA = 1, equation (6) matches the Oriani’s equilibrium
[17,20–22].
Considering regular lattice sites L as the reference state, the indication L may be, when convenient, dropped in the site parameter
notations. Relevant characteristics of traps are then the next: the activation energies for trapping L → Tr and detrapping1 Tr → L,
respectively, ΔETr Tr Tr
t = ELTr − GL and ΔEd = ETrL − GTr , the trap binding energies Eb = GL − GTr defining the trap strengths KTr =
( Tr ) 0
ξTr exp βEb . The partial diffusivity dLL = dLL exp( − βΔELL ) corresponding to L → L jumps is the lattice diffusivity DL ≡ dLL.
At complete local equilibrium, all partial concentrations CA are presented in terms of the lattice one CL as follows:
C A = ΦA (CL )CL (7)
where the contribution ΦA of A-type sites to the metal’s hydrogen absorptivity (solubility) reads:
NA KA
ΦA (CL ) = (note : ΦL ≡ 1) (8)
NL + (KA − 1)CL
The total concentration is
1
They are also named the “saddle point” and “trap activation” energies, respectively [48].
3
Ctot = Φ(CL )CL (9)
where the total trap-induced alteration of solubility with respect to the perfect lattice – the total relative absorptivity – reads:
∑ ΦL ≡1 ∑
Φ= ΦA = 1 + ΦTr (10)
A∈Ξ Tr∈Ξ Tr
Global equilibrium for metal in H2 gas means the equality between all chemical potentials (5) throughout a solid, i.e., ∀A ∈ Ξ and
∀x ∈ Ω, and the gas state chemical potential μH2 , which is [44]
1 ( / )
μH2 = β− 1 ln fH2 f0 + μ0 (11)
2
where fH2 is the gas fugacity, f0 and μ0 are the reference values. At global equilibrium of stressed metal in H2 gas, the equilibrium partial
concentrations (partial solubilities) CeqA according to (5) and (11) satisfy the next equations:
( ) √̅̅̅̅̅̅
CAeq x, fH2 = SA Sσ (x)[1 − θeq
A (x)] fH2 (12)
GL ) − 1/2
where the partial solubility factors (the Sieverts-type coefficients) are SA = KA NA eβ(μ0 − f0 , Sσ (x) = eβVH σm (x) presents the stress
effect on solubility, and θeq eq
A = CA /NA .
2.3. Hydrogen entry into metal
When hydrogen is supplied from an environment, its influx to the metal through the exposed surface Γ ⊂ ∂ Ω should be specified via
pertinent boundary conditions (BCs). This may be done similarly to the development of the adopted diffusion-trapping theory [19]
supplementing the bulk site set Ξ with hydrogen states at the metal-environment interface and considering the species transitions
between all implicated states. To date, diverse interfacial states of the species have been discerned, and particular balance equations
applicable to definite charging systems may be encountered elsewhere [46,49–54].
Albeit its limitations, a suitable approximation for the present modelling is to assume that interface reactions are so fast that they
keep at the interface the local equilibrium between contiguous states of hydrogen in the environment and in the adjacent lattice
[43,46]. Corresponding BC reads:
CL (x, t)|x∈Γ = CLeq (x, fin )|x∈Γ (13)
where the equilibrium lattice concentration Ceq L is determined by (12) using the input fugacity fin of incoming hydrogen in Oriani’s
terms [43], i.e., the fugacity that represents the charging system producing the same equilibrium concentration of lattice-dissolved
hydrogen, as does H2 gas at given temperature.
3. Prototype experiments (the modelling case)
The simulations address the experiments on hot-rolled patented eutectoid pearlitic steel bars [26–31], which composition and
ambient-temperature hydrogen-free mechanical characteristics are presented in Tables 1 and 2. The steel microstructure was char
acterised [29] as lamellar pearlite having cementite interlamellar spacing s = 0.1 μm.
The specimens of hydrogen-free steel were pre-cracked by fatigue in an inert environment, and afterwards subjected to CERTs
under electrolytic charging. The charging is represented by the input fugacity fin = 20.3 MPa (200 atm) corresponding to moderately
harsh cathodic hydrogenation [16,55]. The temperature T = 300 K is set.
The tests are supposed to comply with the fracture mechanics requirements of the crack tip small scale yielding (SSY) near a plane-
strain normal opening (mode I) crack. The crack tip elastoplastic stress–strain field is controlled then by the stress intensity factor (SIF)
KI [7,50]. Focusing on the crack tip processes, the loading is presented in terms of the applied SIF pattern KI = KI(t). The results of HAC
tests depended on the fatigue pre-cracking parameters [26–28,30,31,50]. As well, pre-cracking procedure affects the crack tip
stress–strain field [27,30,31,56,57], which is the mechanical background of diffusion. Then, the entire loading history including fa
tigue pre-cracking is worth taking into account. The considered loading route consists of constant-amplitude cycling between KI =
0 and KI = KImax followed by monotonic loading at constant SIF rate K̇I (the overdot indicates time derivatives). Specifically, the tests
performed at KImax = 0.45KIC and K̇I = 0.0025 MPa⋅m1/2/s are taken as the prototype for the modelling. In these tests, HAC initiation
was detected at the SIF value KHAC = 0.6KIC [30,31] setting the reference for HAF condition assessment.
Table 1
Chemical composition of the steel (mass %) [26–30].
C Mn Si P S Cr Ni Mo Fe
0.74 0.70 0.20 0.016 0.023 0.01 0.01 0.001 balance
4
Table 2
Mechanical properties of the steel [30].
Young modulus E Yield stress σY Fracture toughness KIC Tensile strain ε–stress σ fit using the strength coefficient K and strain hardening exponent n: ε =
(GPa) (MPa) (MPa⋅m1/2) σ/E + (σ/K)n
Stage I: εp ≤ 1.07 Stage II: εp > 1.07
K (MPa) n K (MPa) n
195 725 53 2120 5.8 2160 17
4. Diffusion-trapping parameter assignments
Steels are full of crystalline defects providing hydrogen traps of multiple types [9,14–16,20,23,24,58,59]. Ideally, all them should
be characterized. This requires multifarious investigations that are hard to complete [9,20,24,25,48,60–64]. In the simulations, the
involved traps and their parameters usually are subject to the data availability and the trap’s alleged importance for the considered
phenomenon. Former simulations of diffusion-trapping in hydrogen permeation, degassing and HE tests dealt with one
[17,32,40–42,45,54], two [21,58,64,65], and up to five [9,20,22,65,66] trap types. Nowadays, the involvement of numerous trap
types in calculations is not a complication. In contrast, trap characterisation is the difficulty.
Apropos pearlitic microstructures, fracture events have been associated with ferrite dislocations, cementite lamellae and vacancies
[10,11,38,67–70] pointing them up as potentially deleterious positions. They have also been quoted [14–16,22,40,41,59,61,62] as the
main traps regulating hydrogen diffusion. Though, all other traps, too, do matter in hydrogen transport in metal, and thence in
approaching the HAF event, despite they could be innocuous as regards per se hydrogenous damage [13,21,70]. So, all traps are worth
accounting for in the context of HE. Microstructural features affording diverse hydrogen states in the present steel are the next: (i)
ferrite lattice, (ii) ferrite dislocations, (iii) cementite lamellae, (iv) inclusions, (v) vacancies and (vi) solution impurities.
The disparity of values attributed to hydrogen state characteristics in metals is commonplace. Experimental determination of site
parameters has never been direct, but subdued to theoretical interpretations solving the involved parameters by way of fitting the
experimental and the theoretical calculated behaviours [9,10,24,25,37,48,61–66]. With this respect, it is worthy of a reminder that
“with the large number of parameters necessary to describe hydrogen trapping, several theoretical curves with widely different sets of
parameters could exhibit the same shape”, so that “any supposed good fit of the experimental curves cannot be said to be unique, and
could well be of limited value” [60]. On the other hand, ab-initio calculations manifested substantial discrepancies [24,71–79], too.
Overall, disparate magnitudes have been assigned with uncertain trustworthiness to nominally the same traps [70,80].
Lacking specific data, usual assumption [17,59,81–83] that all saddle point energies are equal one to another is adopted, i.e., EAB =
ELL. This reduces the assortment of the energy landscape characteristics to ΔELL and ETr b . As regards the interstitial vibration frequencies
ωA, their diverseness has been explored [76,83,84] considering them adjustable parameters to match theoretical models and exper
iments. Nevertheless, the matter remains uncertain. The assumption of equality between all them [17,18,81,82,84,85] is adopted
making the trap strengths KTr = exp(βETr b ). Analogous assumption of equality between jump lengths makes the partial diffusivities
dAB = dLL /KA .
BCC-iron has octahedral and tetrahedral interstices. Dual occupancy of them is inevitable, but the shares of occupied states depend
on hydrogen content and temperature [58,72,86–88]. Apropos typical HE conditions in carbon steels, currently accepted wisdom is
that hydrogen in ferrite predominantly resides in the tetrahedrons [16,58,73,88], which are the ground-state lattice sites L. Their
concentration is NL = 5.09×1029 m− 3 [81]. The diffusion activation energy and the pre-exponential factor for iron are ΔELL = 8.37 kJ/
mol and d0LL = 6.70×10–8 m2/s [87], and the Sieverts’ solubility is SL = 6.25×1023exp(− βΔH) m− 3⋅Pa− 1/2, where ΔH = 28.6 kJ/mol
[16,40]. For the present temperature and charging intensity, they bring DL = 2.34×10–9 m2/s and the equilibrium concentration (12)
⃒ 22 − 3
in the stress-free (σm = 0, i.e., Sσ = 1) lattice C0L = Ceq
L (fin ) σ m =0 = 2.95×10
⃒ m . The latter seems reasonable for moderately intensive
cathodic charging [21]. The partial molal volume is VH = 2 cm3/mol [16].
The trap parameters are assigned basing on the comprehensive scrutiny of experimental and ab-initio modelling results (see the
Appendix). The involved traps belong to ferrite dislocations, cementite-ferrite interfaces (CeFeIs), MnS particle-ferrite interfaces
(MnSFeIs), vacancies and solute elements in the steel. Table 3 gives the inventory of the adopted trap characteristics.
The total relative absorptivity Φ brought about by the whole of traps and some particular contributions ΦTr are displayed in Fig. 2.
At global equilibrium of the stress-free steel under considered charging, the assigned trapping parameters render the total hydrogen
concentration (9) of 0.29 mass ppm when εeq eq
p = 0, and 2.9 mass ppm when εp ≥ 3.3. These values fit the experimental data range
[11,69,70,89,90].
5
Table 3
Trap characteristics.
Trap type (trap’s label Tr) ETr
b
KTr NTr
(kJ/mol) at T = 300 K (1024/m− 3)
Ferrite dislocation (⊥) 35 1.2×106 from 5.9 to 148

depending on εeq
p
§
3
Regular CeFeI (Ce) 20 3.0×10 146
CeFeI misfit dislocation (Ce⊥) 35 1.2×106 11.4
Regular MnSFeI (MnS) 49 3.4×108 3.1×10− 2
MnSFeI misfit dislocation (MnS⊥) 62 6.2×1010 1.5×10− 2
Vacancy (vac) 50 5.1×108 from 0.11 to 2.8
depending on εeq
p
§
Mn solute (Mn) 8 24.7 4.3×103

Si solute (Si) 3 3.3 2.7×103
P solute (P) 0 irrelevant
Cr solute (Cr) 4 4.97 73
Ni solute (Ni) 4 4.97 65
Mo solute (Mo) 4.8 6.9 4.0
§
They are proportional to the strain-dependent dislocation density ρ⊥ (εeq
p ), see equations (A2) and (A9) in the Appendix.
5. Diffusion-trapping model reduction
In the analysed case, the governing equations (2) and (3) can be reduced taking account of the next observations.
(I) The total trap concentration is Σ NTr < 0.01NL (see Table 3), i.e., Ntot ≈ NL, so that γL = NL/N ≈ 1, and all trap number fractions
Tr∈Ξ Tr
are small, γ Tr ≪ 1 (NTr ≪ Ntot).

(II) The sure upper bound of the crack-tip mean stress is [57] ̃σ m = 4σY , using which in the solubility relation (12) yields the upper
̃ L = C0 exp(βVH ̃
bound of the lattice concentration C σ m ) = 10.2C0 ; corresponding upper bound of the lattice occupancy is ̃θL =
L L
7
C
̃ L /NL = 5.9×10− ≪ 1.
The smallness of θL allows neglecting it in comparison with unity. Local equilibria (6) then take the next equivalent forms:
KTr θL NTr KTr
θTr = and CTr = CL (14)
1 + KTr θL NL + KTr CL
and the partial relative absorptivities (8) read:

γTr KTr
ΦTr = (15)
1 + KTr θL
Equation (12) now defines the equilibrium lattice concentration in stress-free metal in the Sieverts law form, such as C0L = SL fin1/2 ,
and the equilibrium lattice concentration in stressed solid reads:
Fig. 2. Total relative absorptivity Φ (black) and the contributions ΦTr of MnSFeI misfit dislocations (magenta), vacancies (blue), ferrite dislocations
(red) and the cumulative contribution of all weaker traps having EbTr ≤ 20 kJ/mol (green): (a) undeformed steel; (b) deformed steel with saturated
trap concentrations.
6
CLeq (x, t) = CL0 (fin )eβVH σm (x,t) (16)
5.1. Is local trap-lattice equilibrium applicable?
At a given position x, diffuser jumps between distinct sites pursue local equilibria. These transitions are described by ordinary
differential equations (ODEs) that result from the balance equations (3) omitting there the nonlocal movements between x and x + dx
that involve field gradients. When traps are sparse, γTr ≪ γ L, the lattice sites are dominant neighbours of each trap. Then, trap-lattice
transitions matter the most among the local ones within a material element d3x containing the concentration CL(x,t). Using the mean
value theorem for integrals, the solution [19] of corresponding ODE renders the genuine partial concentration CTr(t) as follows:
[ ∫t ]
⌢ K ⌢
(17)
Tr
CTr (t) = e− t/ τ eξ/ τ CL γ Tr dξ + CTr0
τ0 0
where τ0 = KTr l2LL /(6DL ) is the relaxation time of the trap emptying into hydrogen-free lattice (i.e., at CL = 0), CTr0 = CTr |t=0 , and τ =
⌢
⌢ ⌢ ∫
τ0 /(1 + KTr θ L ), in which θ L (t) = t − 1 0t θL (ξ)dξ. Integration by parts in (17) yields the trap-lattice imbalance as follows:
⎛ ⎞
∫t ⌢
ξ/ τ
[ ] ⌢ ⌢ e γ Ċ
(18)
⎝ Tr L
LE
CTr LE
(CL ) − CTr (t) = CTr (CL0 ) − CTr0 e− t/ τ + KTr e− t/ τ + CL γ̇ Tr ⎠dξ
0 1 + KTr θL 1 + KTr θL
where CLE Tr (CL ) is the concentration of trapped hydrogen corresponding to a given CL by the local equilibrium equality (14), and CL0 =
CL |t=0 . This uncovers the imbalance origins, which are the next:
(a) the initial trap-lattice imbalance [CLE

Tr (CL0 ) − CTr0 ], which effect vanishes as exp( − t/ τ ) where the trap relaxation time τ = τ (KTr )
⌢ ⌢ ⌢
increases with KTr implying that stronger traps approach the local equilibrium slower;
(b) the lattice hydrogen and trap amounts CL and γ Tr ∝ NTr with their rates ĊL and γ̇Tr ∝ ṄTr , so that the larger any of them the larger
the imbalance.
The rate ĊL (x, t) comes from the evolution of CL (x, t) under given hydrogen charging, and the rates γ̇ Tr ∝ ṄTr come from the
generation of traps NTr = NTr (εpeq ) under plastic straining at the rate ε̇eq eq eq
p rendering ṄTr = (dNTr /dεp )ε̇p .
To obtain an upper bound estimate of the lag [CLE
Tr (t) − CTr (t)] from (18), one may use for CL(t) the solution for lattice-only diffusion,
which is faster than the trap-retarded one and provides an upper bound of ĊL . Concerning the diffusion in front of a crack tip towards
( √̅̅̅̅̅̅̅)
the depth x in initially hydrogen-free metal at BC (13), an overvalued closed form approximation [91] CL (x, t) = C ̃ L erfc 1x/ DL t is
2
employed. As far as the imbalance is larger for stronger traps, and it is enhanced by both the trap amounts γTr and generation rates
γ̇Tr ∝ ε̇eq
p , it is expedient to check the matter for the strongest or the most numerous non-multiplying traps, as well as for the strain-
multiplying ones having the largest strengths and numbers or generation rates. In the present case, the furthest from equilibrium
are MnSFeI misfit dislocations (the strongest taps), ferrite dislocations (larger γ Tr ∝ NTr and γ̇Tr ∝ ṄTr ) and vacancies (larger KTr), see
Table 3 and equations (A2) and (A9) in the Appendix. To benefit from the quadrature solution (18), the upper-bound near-tip strain
rate is taken from the closed form Hutchinson’s solution [92] as follows:
Fig. 3. Genuine partial concentrations of trapped hydrogen CTr(t) in comparison with the local equilibrium ones for the strongest non-multiplying
traps (MnSFeI misfit dislocations, magenta), the less-strong more-numerous traps that multiply under plastic straining (vacancies, blue) and the
most numerous and fast multiplying medium-strength traps (ferrite dislocations, red) during CERT at the rate K̇I = 1.0 MPa⋅m1/2/s.
7
( )n+1
n
2n (σ Y )n KI (t) n
n− 1
ε̇ =eq
p λ K̇ I (19)
n+1 K xσ 2Y
where λ = 1/(3π) is the factor in the Irwin’s plane-strain plastic zone size estimation rY = λ(KI /σY )2 [7,56]. To assess a worse case trap-
lattice imbalance during CERTs, an overvalued loading rate K̇I = 1.0 MPa⋅m1/2/s is tried.
These inputs render in-excess estimations of the discrepancies between the genuine partial concentrations of trapped hydrogen
CTr (t) and the ones obeying local equilibrium CLE LE
Tr (CL ) at the prescribed CL (t) as shown in Fig. 3. The local equilibria CTr (t) = CTr (t) are
reached there within ~10− 2 s. This time is negligible comparing to the timescale of even the fast (short-lasting, near 1 min.) CERT at K̇I
= 1.0 MPa⋅m1/2/s. The attainment of local equilibrium occurs earlier for weaker traps, as well as under slower loading and more
realistic (i.e., slower) rates ĊL .
Performed calculations amply validate the applicability of the complete local equilibrium for the present traps. This postulate has
been in use in numerous diffusion-trapping analyses [9,20,22,40–42], but it has received only rough rationale so far [19,63,65,81].
Anyhow, the present solution (18) shows that the postulate may be not valid because of too large magnitude of any of these factors: the
trap strength, the lattice concentration and the trap density together with their evolution rates. This implies that the circumstances of a
specific HE case, in addition to the intrinsic material characteristics, determine the applicability of the Oriani’s postulate.
5.2. Do the diffuser movements passing by traps matter?
The above confine (II) and the local equilibrium equalities (14) allow recasting the fluxes (2) corresponding to non-local lattice-
lattice, lattice-trap and trap-trap movements as follows:
⎫
JLL = − DL CL ∇ln(CL /Sσ ) ⎬
JLTr + JTrL = 2γTr (1 − θTr )JLL , Tr ∈ Ξ Tr (20)
⎭
JTr1 Tr2 + JTr2 Tr1 = 2γTr1 γTr2 (1 − θTr1 )(1 − θTr2 )JLL , Tr1 , Tr2 ∈ ΞTr
The total flux then reads:

( )
∑ ∑
J tot = JLL 1 + 2 (1 − θTr )γTr + 2 (1 − θTr1 )(1 − θTr2 )γ Tr1 γTr2 (21)
Tr∈ΞTr Tr1 ,Tr2 ∈ΞTr
Thence, the partial fluxes, which pass by traps, do not count in comparison with the lattice flow JLL when the traps are close to
saturation, θTr → 1. The same occurs when γ Tr ≪1. This deriving reveals the conditions when partial flows passing by traps become
negligible reducing the total flux Jtot to the lattice one. This occurs when the complete local equilibrium holds and the traps either are
close to saturation or their concentrations are small. The identity Jtot ≡ JLL has been postulated in numerous diffusion-trapping an
alyses [9,20,22,40–42,46,54] since its adoption by McNabb and Foster [45], but it has not been validated so far.
5.3. Case-adapted diffusion-trapping equations
Performed deductions allow reduction of the PDE system (3) to the equations of local equilibria (14) complemented with the single
PDE (4), where Jtot ≡ JLL. Taking JLL from (20) and Ctot in the form (9), the total balance (4) reads:
∂CL
H = − ∇ ⋅ JLL − Qε̇eq
p CL (22)
∂t
Fig. 4. Total hydrogen capacity H (black) and the contributions HTr of MnSFeI misfit dislocations (magenta), vacancies (blue), ferrite dislocations
(red) and the cumulative contribution of all traps having ETr
b ≤ 20 kJ/mol (green): (a) undeformed steel; (b) deformed steel with saturated trap
concentrations. Dashed verticals show the upper bound of the lattice concentration in the analysed HE case.
8
where H = ∂Ctot /∂CL = ∂(ΦCL )/∂CL is the volumic hydrogen capacity presenting the variation of total volumic hydrogen content per
unit change of the lattice concentration under complete local equilibrium, and Q(CL , εeq eq eq
p ) = ∂Φ(CL , εp )/∂εp engenders the sink-like
term associated with the total hydrogen absorptivity rise due to trap generation by plastic straining. This is essentially the same
equation that has been implemented in other studies [9,20,22,40–42]. The present derivation validates its basement postulates of local
equilibrium and lattice diffusion controlled total flux for definite range of metal-hydrogen interaction episodes. Overall, the soundness
of equation (22) is case specific.
Following the definitions (10) and (15), the total capacity H comprises the contributions HA from particular type sites as follows:
∑ HL ≡1 ∑
H= HA = 1 + HTr (23)
A∈Ξ Tr∈Ξ Tr
where
∂CTr γTr KTr
HTr = = (24)
∂CL (1 + KTr θL )2
Plastic straining generates in the present case only ferrite dislocations and vacancies implying the next:
∑ ( KTr dγTr
)
Q= eq (25)
Tr=⊥,vac
1 + KTr θL dεp
Fig. 4 displays the total capacity H and the partial contributions HTr of some traps.
Alternatively, the equation of diffusion-trapping based on the total balance (4) with Jtot ≡ JLL can be reformulated in terms of Ctot
instead of CL by way of inverting (9) as CL = CL (Ctot , 〈NTr 〉) and recasting JLL of (20) via the chain rule for the derivatives of CL . The
result reads:
[ ( )]
H ∑ ∂lnΦ ∂lnSσ
Jtot = − D* Ctot ∇lnCtot − ∇NTr + ∇σ m (26)
Φ Tr∈ΞTr ∂NTr ∂σm
where D* = DL/H is the Oriani’s apparent diffusivity (this was also named effective [18]). Present deriving extends the Oriani’s model
[17] by incorporation of variable trap concentrations NTr(x,t) and non-uniform stress field σ m(x,t). The diffusivity D* depends on the
hydrogen and trap concentrations as illustrated in Fig. 5. Similarly, mere substitution of CL with Ctot /Φ(Ctot , 〈NTr 〉 ) leads to the
expression of total flux of the non-uniform solubility model [19,50,57,91,93].
Anyhow, diffusion-trapping equations in terms of Ctot are inept for calculations with multiple trap types because it is not then
feasible to solve (9) for CL = CL(Ctot) and derive PDE for Ctot in closed form. Such handicap disappears when KTrθL ≪ 1 for all traps. This
linearises the flux equation (26), where the effective/apparent diffusivity maintains constant infimum value D¯* =
( )− 1
DL 1 + ΣγTr (εeq
p )KTr [19]. Corresponding linear diffusion-trapping problem is handleable as described elsewhere [31,57,91].
Tr
6. Calculation model setup
Elastoplastic mechanics-and-diffusion modelling of cracks at SSY under monotonic loading customarily has been implemented via
the Rice’s boundary-layer formulation considering convenient (usually circular) domain with a crack under mechanical BC determined
Fig. 5. Oriani’s apparent diffusivity D* in the undeformed steel (solid line) and after deformation up to the trap concentration saturation (dash-dot
line). Dashed horizontal and vertical indicate, respectively, the lattice diffusivity DL and the upper bound of the lattice concentration in the ana
lysed case.
9
Fig. 6. Schematics of the model geometry and loading: (a) the simulated domain; (b) the model of a crack; (c) the loading route (time t ≥ 0 cor
responds to the CERT).
by the KI-asymptotic elastic solution [22,31,40–42,54,94]. Although under monotonic loading such formulation is theoretically
indifferent to the choice between BCs set by elastic KI-controlled displacements or stresses, it does not when cyclic loading generates
plastic strains. Having no intention to resolve this ambiguity, the straightforward approach akin to the former analyses [56,91,95] is
adopted, which outline is as follows.
The edge-crack panel is loaded in plane strain by remote tractions σ (Fig. 6a). To ensure the KI-controlled SSY [56,57,91,95], the
crack length a and the panel sizes w with h are taken large comparing to the crack tip plastic zone upper-bound dimension ̃rY =
λ(KIC /σY )2 (indeed, all these sizes exceed ∼ 30̃rY ). Thin notch with semicircular tip (Fig. 6b) has been esteemed suitable model of
undeformed crack in large-deformation elastoplastic analyses [22,31,40–42,54,56,57,91,94,95]. The adopted original crack width b0
= 5 μm agrees with the estimates for cracks in medium–high strength steels [30,57]. The loading route σ(t) is calibrated to produce the
desired SIF pattern KI (t)∝a1/2 σ (t). It comprises several constant-amplitude load cycles between KI = 0 and KImax followed by mono
tonic loading up to the reference level KIR, and keeping KI = KIR afterwards (Fig. 6c). Adjusting to the prototype experiments, the values
KImax = 0.45KIC, KIR = KIC and monotonic loading rate K̇I = 0.0025 MPa⋅m1/2/s were set. Higher value K̇I = 1.0 MPa⋅m1/2/s was also
tried to examine the rate effect (or the effect of testing/diffusion timescale tR = KIR/K̇I ) on the crack tip hydrogenation.
Apropos the mechanical background for the stress–strain affected diffusion, two potential effects of hydrogen, which are the
hydrogen induced metal dilation [16] that generates mechanical stresses and the hydrogen effect on material’s elastoplastic
stress–strain response, are disregarded because of their minor significance in the analysed case [31]. Accordingly, the stress–strain
analysis is not coupled with hydrogen charging. The constitutive model of rate-independent J2-elastoplasticity with mixed isotropic-
kinematic hardening along the stress–strain mastercurve of hydrogen-free material, which is characterised in Table 2, is adopted [57].
Thereupon, a shape of the loading route in time t is irrelevant, but only KImax and KIR matter. The solution of this mechanics problem
Fig. 7. Undeformed FE mesh with the details of its refinement near the crack tip.
10
provides the input for diffusion calculations.

Hydrogen charging proceeds during CERTs at t ≥ 0 (Fig. 6c). Considering the steel to be hydrogen free before tests, the initial
condition reads:
CL (x, t) |t=0 = 0 for x ∈ Ω (27)
Hydrogen enters the metal through the crack surface Γ. Accounting for (16), the BC (13) is particularised as follows:
⃒
CL (x, t)|x∈Γ = CL0 eβVH σm (x,t) ⃒x∈Γ for t⩾0 (28)
Diffusion BCs at remote edges of the model domain Ω (Fig. 6a) are irrelevant. Indeed, comparing the distances r of diffusion from
Fig. 8. Distributions of the mean stress σ m (red), crack opening stress σ yy (blue) and cumulative plastic strain (black) in front of the crack tip at
indicated load levels KI (and corresponding loading times) during CERTs after fatigue pre-cracking. The position x̄c of the maximum normal stress at
KI = KIC defines the FPZ supremum size.
11
these boundaries, which are r ≥ a, with those from the crack tip surface towards potential fracture locations within the crack tip plastic
zone at r < ̃rY , the ratios of these distances exceed a/̃rY ~ 30. Respective diffuser travel times from the remote and crack-tip surfaces
towards fracture loci near the tip, which are t ∝ r2 [16], differ more than (a/r Y )2 ~ 302 times. In consequence, hydrogen populations
∼
near the crack tip should attain the equilibrium (16) much earlier than hydrogen from remote surfaces could approach the crack tip
vicinity. As computationally convenient option, nil hydrogen flux is prescribed at the remote edges ∂Ω\Γ of the analysis domain.
Exploiting the problem symmetry, the calculation domain is reduced to the panel half (shadowed in Fig. 6a) with nil vertical
displacement and hydrogen flux on its bottom edge.
Large-deformation elastoplasticity problem was solved applying the finite element (FE) procedure described elsewhere
[56,57,91,95]. The case-adapted diffusion-trapping model was implemented into corresponding FE procedure via the weak formu
lation [31,40–42,57,91] of PDE (22). The discretisation of solution domain Ω with four-node FEs (Fig. 7) employed highly refined
mesh near the crack tip to treat steep-gradient stress–strain and concentration fields. To ensure the solution mesh convergence, several
mesh refinements were tried. Presented results were generated with the mesh having average undeformed size of the smallest elements
about 0.02b0. The FE procedure was verified through comparison with the published solution [22].
7. Crack tip stress–strain field
Fig. 8 displays distributions in the crack plane of the mean and crack opening stresses, and of the plastic strain depending on the
depth x from its tip at definite load levels KI during CERTs after fatigue pre-cracking (at t ≥ 0 in Fig. 6c). The indicated load levels are
convertible into CERT times t = KI /K̇I obeying t/tR = KI/KIR. The stresses in Fig. 8 manifest the closest to the crack tip relative maxima
m (KI ) = max{σ m (x, KI )} and σ yy (KI ) = max{σ yy (x, KI )}, which are situated at respective distances xm (KI ) and xyy (KI ).
σ+ + + +
x x
These results have been commonplace in similar analyses [22,31,40–42,54,94,95]. Additional insight into the relevant features of
the stress–strain state is provided in Fig. 9. It shows the advancement of the stress relative maxima and of their positions during CERTs
from the compressive stresses situated at x̄+ m,yy = 0 to the tensile ones ahead of the tip. When KI reaches KIC, these stresses and their
positions attain respective global maxima σ̄m = 3.6σY, σ̄ yy = 4.5σY and x̄+
m ≈ x̄yy = 13.5 μm. The latter sets the FPZ size supremum x̄c =
+
x̄+
yy . Whilst KI < ~0.1KIR, the stresses along the FPZ x ⩽ x̄c are dominated by the compression caused by cyclic pre-loading (Fig. 8).
m,yy turn tensile at load minute advance beyond ~0.05KIR, and they become absolute
Notwithstanding, the stress relative maxima σ +
extrema soon after that. Moreover, Fig. 10 presents the evolutions of the mean stress (the crack opening stress behaves similarly) and
accumulated plastic strain in the determinate material points Pi (i = 1,2,3), which positions xi = x(Pi) stay during the tests near 7, 13
and 25 μm, respectively.
8. Diffusion-trapping computation results
Spatial distributions of the lattice concentration CL (x, KI ) at the loading rates K̇I of 0.0025 and 1.0 MPa⋅m1/2/s are shown in Fig. 11
for the indicated load levels KI (and corresponding stress–strain states presented in Fig. 8), which are attained at respective test times
t = KI /K̇I (recall, t/tR = KI/KIR). Fig. 12 displays the concentration evolutions CL (Pi , t) in the formerly indicated material points Pi (i =
1,2,3) during slow and fast CERTs. The equilibrium concentration Ceq L (x,t), which is determined according to (16) by the stress σ m(x,t)
shown in Figs. 8 and 10a, is presented in Figs. 11 and 12, too.
Under local equilibrium, CL via relations (14) and (9) determines the partial and total hydrogen concentrations CTr and Ctot. Their
spatial distributions along the crack plane and time evolutions in the determinate material points Pi (i = 1,2,3) during slow and fast
CERTs are displayed in Figs. 13 and 14.
Figs. 11 to 14 describe the near crack diffusion-trapping more comprehensively than it was done in other works [22,40–42,54],
which generally (excepting [22]) dealt with a single trap type. Moreover, the present modelling elucidates remarkable effects of the
Fig. 9. Evolutions of the extrema of mean (red) and crack-opening (blue) stresses near the crack tip during CERT after fatigue pre-cracking: (a) the
closest to the tip stress relative maxima, (b) their respective positions. Dotted line in (a) presents the plastic strain at the position of maximum crack
opening stress. Rhomb markers are to distinguish the curves that partially overlap.
12
Fig. 10. Advancement of the crack tip stress–strain state characteristics in the material points, which positions x stay close to 7, 13 and 25 μm (red,
blue and black lines, respectively) during CERT after fatigue pre-cracking: (a) the mean stress; (b) the plastic strain.
residual stresses induced by fatigue pre-cracking and of the loading rate.

The obtained CL-solutions determine via (14) the trap occupancies θTr, which are presented in Figs. 15 and 16. They provide in
sights into the behaviours of multiple concurrent traps and enable deductions about the trap functioning with implications for HE
according to the trap roles both in fracture events and in hydrogen transport, which are reflected by respective foundational equations
(1) and (21)-(25).
In Figs. 13 and 14, the traps combining considerable binding energies and large densities are the main holders of the total hydrogen
content Ctot. Anyhow, most magnitudes of concentrations are rather comparable within about an order of magnitude, and they are
transient. This hardly helps to distinguish between potentially harmful and innocuous traps. On the other hand, the trap occupancies in
Figs. 15 and 16 are disparate, and definite ones are stationary due to saturation. These data allow better screening of traps to identify
potentially harmful ones. Anyhow, the partial concentrations do not directly represent the species amounts in microstructural features,
which rule HE.
9. Implications for HAC
9.1. Deleterious and innocuous hydrogen states
The key premise [9,12,14] is that HE relies on the entrapping of hydrogen in definite crystalline imperfections making them
microfracture origins. Such traps X are deleterious, and other ones are innocuous. To perpetrate hydrogen-related mechanistic effects,
X-traps must accumulate certain amounts of hydrogen represented by the occupancies θX, and HAF events are fundamentally related to
the critical hydrogen contents θcr
X . On the other hand, all traps slowdown hydrogen transport in metal helping to delay (but not avoid)
hydrogen delivery to the deleterious sites. The underlying philosophy of hydrogen-resisting alloy design [14] is to try
(I) to increase the critical values θcr

X of fracture nucleating sites and
(II) to reduce the actual hydrogen amounts θX on them.
The first point is about microstructural mechanisms of fracture and their triggering by hydrogen, which is out of the scope of the
present work, and the second one relies on the hydrogen transport.
To derive implications for HAF, deleterious traps X must be identified and their occupancies θX must be analysed. To proceed,
generic criterion of HAF (1) needs concretisation. To this end, the mechanistic effect of hydrogen is deemed to ease the stress-
controlled fracture ruled by the maximum normal stress σmax n . Corresponding fracture criterion has equivalent implicit and explic
itly resolved with respect to some suitable variable forms as follows:
( ) ( max ) ′
Ψ σmax max
n , 〈θX 〉 = 0 ⇔ σ n = σ cr cr ″ ′
n (〈θX 〉) ⇔ θX′ = θX′ σ n , 〈θX″ 〉 (X ∈ Ξ X , {X } = Ξ X X ) (29)
where σcrn is the critical stress at the fracture event.

Near the crack, the fracture controlling stress σmax
n = σyy . To engender damage even in the presence of hydrogen, the stress must be
tensile. Compressive near-tip stresses generated by cyclic pre-loading (see Fig. 8) exclude fracturing at the very beginning of CERTs. At
loading minute advance of ~0.05KIR, the stress local extremum σ + yy turns tensile, and becomes the global maximum shortly afterwards
yy (KI ) the plausible location of HAF event xc = xyy . The occupancies of representative traps in the
(Figs. 8 and 9a) making its position x+ +
potential fracture location xc (KI ) during CERTs are shown in Fig. 17.
From Fig. 17, the observations about the trap operation modes in per se HAF are as follows.
13
Fig. 11. Distributions of the lattice hydrogen concentration in front of the crack tip at the indicated load levels KI (and corresponding diffusion
times) during CERTs at the loading rates of 0.0025 and 1.0 MPa⋅m1/2/s (red and blue lines, respectively) after fatigue pre-cracking. Dashed lines
present the equilibrium lattice concentration corresponding to the instantaneous stress σ m(x,KI).
• The occupancy of relatively weak regular CeFeI traps (KCe ~ 103) stays minute (θCe < ~10− 3) in both fast and slow tests. Weaker
sites (KTr < KCe) attract according to (14) even less hydrogen. Occupying one among more than 103 sites of this kind, hydrogen
hardly can alter the mechanistic roles of such microstructural features reducing σcr n . With respect to per se HAF, such sites form the
functional category of null traps ΞnTr ≡ {nTr∣nTr ∈ ΞTr, θnTr ≪ 1}. To fall into this category, the traps according to (14) must satisfy
KnTrθL ≪ 1.
• On the other extreme, rather strong traps, such as vacancies (Kvac ~ 108), become saturated shortly after the initiation of booth slow
and fast CERTs. The level θvac (xc ) = 0.95 is surpassed at the times t0.95 ≈ 0.05tR, i.e., within the very initial ~5%-segments KI ≤ ∼
t0.95 K̇I = 0.05KIR of the CERTs. Stronger traps (KTr > Kvac) approach the saturation even earlier. The compression caused by cyclic
pre-loading (see Figs. 8 to 10a) dominates the stress in the crack tip FPZ x ⩽ x̄c during the CERT initial phase KI ≤ ~0.05KIR (t ≤
~t0.95). Fracture cannot occur there even with the aid of hydrogen. Then, such segments of CERT routes are disregardable. In the
remaining determinative segments KI ≥ 0.05KIR (t ≥ 0.05tR) of the tests, when the crack tip stresses are tensile, these traps stay
virtually full, θTr ≈ 1. Such hydrogen-saturated microstructural faults may have a role in material’s mechanics, but their situations
lack the transitoriness, so that these traps cannot match the HE rate (time) sensitivity. Such sites form the functional category of
permanently jammed traps ΞjTr ≡ {jTr∣jTr ∈ ΞTr, θjTr ≈ 1}. They satisfy the condition KjTrθL ≫ 1 ensuring θjTr ≈ 1 by (14). Virtually
constant occupancies θjTr = 1 cease to be the variables operating in HAF criterion (29).
14
Fig. 12. Evolutions of the lattice hydrogen concentration in the material points, which stay at the indicated distances x from the crack tip, during
CERTs at the loading rates of 0.0025 and 1.0 MPa⋅m1/2/s (red and blue lines, respectively) after fatigue pre-cracking. Dashed lines present the
equilibrium lattice concentration corresponding to the instantaneous stress σm(x,KI).
• As regards the intermediate strength traps, such as ferrite dislocations and CeFeI misfit ones (K⊥,Ce⊥ ~ 106), they do not attain
saturation. Their occupancies during CERTs are visibly transient and become substantial, θ⊥Ce⊥ ~ 10− 1. The attainment of 1 at.H
per every few sites and perceptible dependence of the occupancies on the loading rate (process time) prompt to associate these traps
with hydrogen-sensitive material behaviour that conforms with experienced load rate sensitivity (time dependence) of HE. Such
operation mode differentiates these sites as the category of effectual traps ΞeTr ≡ {eTr∣eTr ∈ ΞTr, θnTr ~ 10− 1}. Having confirmed
their involvement in per se HAF micromechanism, such traps present the deleterious states and their occupancies come to be the
controlling variables in HAF criterion (29).
This categorisation of traps by their operation modes puts focus on their faculty to regulate hydrogen effects on fracture. In contrast
to other classifications relying on the trap binding energy, strength or “(ir)reversibility” [13–15,21–23,60–66,96,97], it helps to
identify potentially deleterious traps in considered HE episodes. The occupancies of traps from the union ΞeTr ∪ ΞjTr are eligible to
define HAF criterion (29). Though, the trapping states of actual plenitude degrade the hypersurface (29) to its section by the hy
perplanes θjTr = 1.
Turning to pearlitic microstructures, the experiments [10,11,70] implied that harmful hydrogen resided in the traps, which were
assimilable to the present eTr-category, whereas stronger traps assignable to the jTr-class resulted innocuous. Then, the deleterious
states are among misfit CeFeI and ferrite dislocations. This conforms to the inferences about ferrite dislocations and cementite as the
microstructural features that engender fracture [67–70,90,98,99], where hydrogen induced decohesion and localised plasticity may
act synergistically [9,12,20,25,69]. The arguments about the key role of ferrite dislocations in the microfracture in pearlitic steels
[67–69] via microcrack initiation by dislocation pile-ups [7,8,9,12] point out ferrite dislocations as deleterious sites.
Now, the HAF controlling variables are σmax
n yy and θX = θ⊥, and the fracture locus (29) becomes a contour in the (σ yy ,θ⊥ )-plane.
= σ+ +
The state of affairs in the potential fracture position xc(KI) during CERT may be presented in this plane by corresponding trajectory
1/2
yy |KI =KI (t) ,θ⊥ | x=xc (t) ), see Fig. 18. HAF event in the prototype experiment at K̇I = 0.0025 MPa⋅m
(σ+ /s occurred at the SIF KHAC = 0.6KIC
[30,31] attained at the time tHAC = KHAC /K̇I = 0.6tR . The maximal normal stress at this instant is σHAC
yy yy (KHAC ) = 4.2σ Y, and it
= σ+
cr
takes place at xc = x+
yy (KHAC ) = 7.5 μm (see Figs. 8 and 9). The critical occupancy of deleterious traps at HAF event is θ⊥ = θ⊥ (xc ,tHAC )
cr
= 0.29 (Figs. 17a and 18). This sets the point (σ HAC
yy , θ⊥ ) of fracture locus in the (σ yy ,θ⊥ )-plane. The point (σ̄ yy , 0) corresponds there to
+ +
hydrogen-unassisted fracture at KI = KIC . At faster rate K̇I = 1.0 MPa⋅m1/2/s, the former values KHAC and σHAC
yy are attained at shorter
time t = KHAC /K̇I < tHAC . The trap occupancy at this instant is less, θ⊥ = 0.22 (Figs. 17a and 18), and the point (σ+
yy , θ⊥ )
|KI =KHAC ,t=KHAC /K̇I does not belong to HAF locus. The test must continue towards fulfilment of HAF criterion at larger KI and σ+
yy (KI ), see
Fig. 18. This accords with commonly experienced rate-dependence of HE. Additional CERTs at different rates could provide more
points of the HAF locus.
9.2. On the withdrawal of hydrogen from deleterious sites by innocuous traps
When hydrogen source is unlimited as is the case in environmental HE, traps under local equilibrium do not deprive one another of
hydrogen. At given CL, which is furnished by the hydrogen diffusion flow fed by the environment, the presence of other traps is
inconsequential for the occupancy of whichever one, see (14). Particularly, the trapping cannot selectively reduce hydrogen amounts
θX in the deleterious sites X by way of the species withdrawal therefrom and re-allocation to innocuous sites.
Lattice hydrogen concentration CL(x,t) asymptotically pursues the equilibrium level Ceq L (x, t), which sets the instantaneous upper
bound for the actual one: CL (x, t) ⩽ CeqL (x, t). According to (16), C
eq
L depends on the hydrogen input fugacity fin, the instantaneous stress
σm and the temperature T, Ceq L = C eq
L (fin , σ m , T), but not on traps. The lattice concentration supremum C̄L = CeqL (σ̄ m ), and thence, the
15
(caption on next page)
16
Fig. 13. Distributions of the partial and total hydrogen concentrations in front of the crack tip at the indicated load levels KI (and corresponding
diffusion times) during CERTs at the loading rates of 0.0025 and 1.0 MPa⋅m1/2/s (left- and right-hand plots, respectively) after fatigue pre-cracking:
Ctot (black), CMnS⊥ (magenta), Cvac (blue), C⊥ (red) and the cumulative concentration corresponding to all weaker traps having EbTr ≤ 20 kJ/mol
(green). Dashed lines present the lattice concentration CL for reference.
Fig. 14. Evolutions of the partial and total hydrogen concentrations in the material points, which stay at the indicated distances x from the crack
tip, during CERTs at the loading rates of 0.0025 and 1.0 MPa⋅m1/2/s (left- and right-hand plots, respectively) after fatigue pre-cracking: Ctot (black),
CMnS⊥ (magenta), Cvac (blue), C⊥ (red) and the cumulative concentration corresponding to all weaker traps having EbTr ≤ 20 kJ/mol (green). Dashed
lines present the lattice concentration CL for reference.
suprema of hydrogen amounts in the deleterious sites, which by (14) are θ̄X = θX (C̄L ), are not affected by trapping, either. Then, these
vulnerable microstructural features X alone define the alloy’s resistance to HAF in terms of the threshold stress (and the threshold SIF for
cracks). It defines the upper bound safe load corresponding to the maximum content of the pernicious hydrogen under given charging
intensity.
All opposites apply in the case of internal HE, i.e., when insulated metal holds fixed amount of hydrogen. Traps do compete there
for hydrogen in a manner that stronger sites deprive weaker ones of the species [20,21,41,81]. Strong innocuous traps could then
17
(caption on next page)
18
Fig. 15. Distributions of the trap occupancies in front of the crack tip at the indicated load levels KI (and corresponding diffusion times) during
CERTs at the loading rates of 0.0025 and 1.0 MPa⋅m1/2/s (left- and right-hand plots, respectively) after fatigue pre-cracking: θMnS⊥ (magenta), θvac
(blue), θ⊥ (red) and θCe (green); rhomb markers are to distinguish the lines tending to overlap.
Fig. 16. Evolutions of the trap occupancies in the material points, which stay at the indicated distances x from the crack tip, during CERTs at the
loading rates of 0.0025 and 1.0 MPa⋅m1/2/s (left- and right-hand plots, respectively) after fatigue pre-cracking: θMnS⊥ (magenta), θvac (blue), θ⊥ (red)
and θCe (green).
beneficiate the HE resistance until they get full.
9.3. On the slowdown by traps of hydrogen access to the deleterious states
Although traps under local equilibrium cannot gain hydrogen one at the expense of another because all the occupancies univocally
correspond to the current CL(x,t), the trapping reduces the lattice hydrogenation rate ĊL , and thence, retards hydrogen accumulation in
the deleterious positions. Hydrogen capacity H may be adopted as rate reduction index (RRI) as regards the lattice charging: ĊL =
|∇ ⋅ JLL |/H, see (22). (As well, H determines the reduced apparent diffusivity D* =DL/H in Ctot-description (26) of hydrogen transport.)
According to (23)-(24), the decelerating effect of traps is proportional to their quantities and strengths. The utmost RRI is H̄ = 1 +
Σ γTr KTr . This implies up to ~103 times deceleration of the lattice charging in the present case (see Fig. 4). The downgrade of the
Tr
19
Fig. 17. Evolutions of the vacancy, ferrite dislocation and regular CeFeI trap occupancies θvac (blue), θ⊥ (red) and θCe (green) in the potential
fracture location during CERTs after fatigue-pre-cracking at (a) slow and (b) fast loading rates of 0.0025 and 1.0 MPa⋅m1/2/s, respectively.
Fig. 18. Evolutions of the situation “maximum normal stress – dislocation hydrogen content” in the potential fracture location during CERTs at the
rates of 0.0025 and 1.0 MPa⋅m1/2/s (red and blue plots, respectively; cross markers are to distinguish the curves that may overlap); dashed line
presents a likely HAF locus, where red and black rhombs correspond to the fracture events at slow-rate HAC test (KI = KHAC) and in fracture
toughness test with no hydrogen effect (KI = KIC), respectively.
Fig. 19. Oriani’s apparent diffusivity D* in front of the crack tip on the way to potential HAF locus at the times t/tR (loads KI/KIR) equal to 0.05
(red), 0.1 (green), 0.3 (blue) and 0.6 (black) during CERTs at (a) slow and (b) fast loading rates of 0.0025 and 1.0 MPa⋅m1/2/s, respectively. Dashed
horizontals show the lattice diffusivity DL. Rhomb markers are to distinguish the curves close to overlapping.
apparent diffusivity D* from DL ~ 10− 9 m2/s to ~10− 12 m2/s (Fig. 5), too, reflects this maximal slowdown of hydrogen transport. Such
large retardation owes to strong traps (KTr > ~108). However, they are efficient retarders only whilst they stay far from saturation. This
becomes evident after recasting the RRI (23) with the use of (14) as follows:
20
Fig. 20. Equilibrium trap occupancy diagrams (from stronger to weaker traps): MnSFeI misfit dislocations (magenta), vacancies (blue), ferrite
dislocations (red) and regular CeFeI sites (green). The frame delimits the domain corresponding to the considered HE episodes.
∑
H = 1+ γTr KTr (1 − θTr )2 (30)
Tr
Overall, the occupancies of traps regulate their ability to retard hydrogen transport in metal in a manner that the closer the trap to
saturation the less is its decelerating effect, and the trap fullness (θTr = 1) turns the slowdown off. Keeping θL ≪ 1/KTr preserves θTr ≪ 1
and ensures the trap’s maximal retarding capability. Strong traps lose it at all at minute lattice charging.
Concerning the analysed CERTs, the trap occupancies near the crack tip (Figs. 15 and 16) suggest the discrimination of trap roles on
the delay of hydrogen supply to the deleterious states, which manifests certain affinity with the formerly considered trap operation
modes as regards per se HAF events as follows.
• The nTr-type traps (the weaker-trap segment), which occupancies near the crack tip stay negligible (θnTr ≪ 1). Accordingly, their
cumulative contribution to the RRI (30) stays maximal, but minor: it is ΣHTr = Σγ Tr KTr = 1.1 (see Fig. 4). Larger trap amounts γ Tr
nTr nTr
∝ NTr could enhance the effect.

• The jTr-type traps (the stronger-trap segment), which fullness near the crack tip (θjTr ≈ 1) annihilates respective terms (1 – θjTr)2 in
(30) along the CERT determinative phase KI ≥ 0.05KIR (t ≥ 0.05tR). Such traps do not modulate hydrogen transport rate. (Only
during disregardable initial portion of the test they can stay far from fullness rendering the cumulative contribution to the RRI
comparable with the supremum Σ γ Tr KTr ≈ 2×103, see Fig. 4.).
jTr
• The eTr-type traps (the intermediate-strength traps), which occupancies become considerable, but stay appreciably less than unity
during substantial segments of the tests. They bring major contributions HeTr = γ eTr KeTr (1 − θeTr )2 , which dominate the total RRI.
These traps produce up to ~102-fold deceleration of ĊL within the present case range CL < CL (see Fig. 4). The apparent diffusivity
∼
D* ≈ 0.03DL within FPZ x ⩽ x̄c and in its surroundings (Fig. 19) evinces the delaying effect of these traps, too.
9.4. Case dependence of the potentiality of traps to mitigate hydrogen embrittlement
The trap roles both in per se events of HAF and in the slowdown of hydrogen delivery to the fracture sites are directly ruled by the
trap occupancies θTr. At local equilibrium (14), θTr-s are set by the lattice concentration CL as shown in Fig. 20. The belonging of a trap
to definite functional category is conditioned by the effectual range CL ⩽ CL ⩽ C̄L within potential FPZ in a given HE episode. In the
present case, the lattice concentration supremum is C̄L = C0L exp(βVH σ̄ m ) = 8.3C0L . The lower bound CL ≈ 0.6C0L is set according to the
CL solutions in the FPZ x ⩽ x̄c at the onset of the CERT determinative phases KI ≥ 0.05KIR (t ≥ 0.05tR), see Figs. 11, 13 and 14. Which
traps go into this or that functional category depends not only on the binding energies, but also on the extrinsic factors, such as the
hydrogen charging intensity, the stress field and the temperature, which determine the trap circumstance – the case-dependent CL
range in the FPZ. That is, the adopted trap categorisation by the operation modes is case specific, as do the trap exposures to hydrogen.
1/2
In particular, less severe charging, i.e., lower C0L ∝ fin , shifts all occupancy-related curves in Figs. 2, 4, 5, and 20 to the right,
whereas the bounds CL /C0L and C̄L /C0L remain there immutable. Depending on the shift magnitude, distinct characteristic segments of
the diagrams may get in the factual CL-range implying changes of the trap operation modes in per se HAF event and increasing (more
notably or less) their slowdown effect on hydrogen delivery. Elevation of the hydrogen charging intensity has the opposite conse
quences. The stress state and temperature, too, affect the ranges of CL and the trap exposures to hydrogen θTr(CL). The stress level
depends on solid’s geometry, loading and material’s mechanical properties, like, e.g., the yield strength and strain hardening regulate
the top mean stress σ̄ m [56,57,94,95], and thence C̄L . The temperature regulates both the stress effect on CL and on the trap strengths
with consequences on their occupancies. Overall, the potentiality of every trap to mitigate HE is case dependent.
21
10. Conclusions
The concurrent participation of multiple microstructural traps in hydrogen transport towards potential origins of fracture near the
crack tip is considered in the framework of a comprehensive diffusion-trapping theory. Following the underlying principle that trap
exposures to hydrogen regulate the functioning of respective microstructural features in hydrogen embrittlement (HE), the occu
pancies of traps by hydrogen are addressed. The analysed system is steel specimen with a crack under hydrogen charging during
constant-rate extension. The approach is applicable to environmental hydrogen assisted cracking (HAC) in diverse material systems.
The main outcomes of the analysis are as follows.
• The solution of trap-lattice local exchange kinetics shows negligible lagging of traps behind the local equilibrium. This validates the
Oriani’s local equilibrium postulate.
• Closed-form deriving demonstrates how the filling of traps with hydrogen or the smallness of their quantities can make negligible
the partial diffusion fluxes passing by traps under local equilibrium. This validates the commonplace assumption that the total
hydrogen flow in metal equals the lattice diffusion flux.
• Overall, the trap-lattice local equilibrium is case-sensitive, and its corollaries are, too. The fulfilment of local equilibrium depends
on trap strengths, hydrogen charging and loading.
Calculated trap occupancies reveal the potentiality of respective microstructural features both to carry out the mechanistic effects
of hydrogen and to modulate the rate of hydrogen delivery to the fracture loci. As regards the modes of trap operation in HE, three
functional categories of traps are discerned as follows.
• Traps, which occupancies stay negligible (the weaker-trap segment). They are null traps having no role in per se embrittlement
events. They produce permanent reduction of the hydrogen transportation rate, which is proportional to the trap quantities and
strengths. Though, it ordinarily stays rather minor.
• Traps, which get saturated upon a minute intake of the lattice-residing hydrogen in a relevant process zone shortly after the very
beginning of the HAC episode (the stronger-trap segment). Such jammed traps may produce constant hydrogen-related decrement
of material strength, if they are vulnerable microstructural features, but they cannot match the manifestations of HAC rate (time)
dependence. They do not modulate the rate of hydrogen accumulation in the process zone overall.
• Traps, which occupancies in the process zone attain appreciable magnitudes staying remarkably below the saturation (the
intermediate-strength trap segment). If they are vulnerable microstructural features, such effectual traps can produce transient
hydrogen-related decrement of material strength compatible with the manifestations of HE rate (time) dependence. As well, they
reduce the rate of hydrogen supply to fracture loci by orders of magnitude depending on the trap numbers, strengths and lattice
hydrogen concentration.
Whether the trap operation mode falls into this or that functional category is not determined solely by the trap strength, but also by
the range of lattice concentration, which depends on particular loading and hydrogen charging conditions. Thence, it is case specific.
Overall, hydrogen trapping at microstructural features has only a limited scope as a strategy to improve the HAC resistance via
hindering the hydrogen delivery to potential fracture sites. The hydrogen transport retardation by traps vanishes as they approach
saturation. In the case of HAC under moderately intense environmental charging, strong traps having the most deceleration poten
tiality become full in the crack tip vicinity fairly promptly, which aborts their rate reduction effect. Substantial delay of HAC may be
attained with the stronger the better traps under the condition that they stay sufficiently far from saturation in given metal-hydrogen
interaction circumstances. Anyway, trapping can be beneficial for the delay, but not prevention of hydrogen accumulation in fracture
nucleating sites. A way towards more comprehensive protection against hydrogen embrittlement would be the creation of micro
structures with elevated intrinsic resistance to hydrogen where the triggering of microfracture would require more hydrogen in the
vulnerable microstructural features.
CRediT authorship contribution statement
V. Kharin: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Software, Methodology,
Investigation, Formal analysis, Data curation, Conceptualization.
Declaration of Competing Interest
The author declares that he has no known competing financial interests or personal relationships that could have appeared to
influence the work reported in this paper.
Data availability
Data will be made available on request.
22
Acknowledgements
This work was supported by the Spanish Ministry for Economy and Competitiveness [grant number BIA2011-27870]; and Junta de
Castilla y León [grant number SA132G18].
Appendix. Diffusion-trapping parameter assignments
The characterisation of traps in alloys is usually uncertain, and reported theoretical and experimental values for nominally the same
traps often disagree [14,63,74]. As well, the identifications of traps to assign the experimentally assessed characteristics may be
deceptive [14,32]. The experimental trap energy data for (nearly) eutectoid pearlitic steels are clustered around 20, 35 and 60 kJ/mol
as the likely expectations [59,69,70]. Intensive examination of data affords the following characterisation of traps in the considered
steel.
A1. Ferrite dislocations
The cores and near-core strained lattice of edge, screw and mixed dislocations [14–17,23,32,70,71,73,74,78] provide multifarious
low energy states. Experimental binding and detrapping energies from ~17 to ~60 kJ/mol were ascribed to dislocations
[9,14–17,23,32,46,48,58–63,70,100,101]. Though, the declared trap identities may be often regarded as tentative. Ab-initio models
revealed diverse irregular interstices with the binding energies from 20 to 45 kJ/mol at edge dislocations and from 16 to 43 kJ/mol at
screw ones [71,73,74,78,102–106]. The dislocation core traps at ambient temperature are at least ~103 times stronger than farther
ones [71,74,78,102–105]. These dominant states are considered here the single type of dislocation trap labeled by ⊥. Adjusting the
cited results to the mentioned modes of the experimental data set [59,69,70], a reasonable binding energy assignment is E⊥ b = 35 kJ/
mol.
Direct measurements of dislocation trap concentration N⊥ in the steel are not available. Presuming that the adjacent deep sites at
dislocation cores are blocked by H-H repulsion within ~0.2 nm [107], atomistic analyses [71,73,74,104] revealed three traps per each
√̅̅̅
atomic plane threaded by dislocations of primary slip systems {110}〈1̄11〉, i.e., per every aFe / 2 period along dislocation lines
[20,22,81], where aFe = 0.287 nm is the ferrite lattice cell size. The measurements of dislocation density ρ⊥ for the present steel are
absent. Analogous steels [33–39] manifested its increase with plastic deformation towards saturation. The bilinear fit [22] is adopted
basing on these measurements as follows:
⎧
( ) ⎨ ρ + ρsat − ρ0 εeq at εeq < 3.3
(A1)
0 p p
eq
ρ⊥ εp = 3.3
⎩
ρsat otherwise
where ρ0 = 4×1014 m− 2 and ρsat = 1016 m− 2

are the initial and saturated dislocation densities, respectively. The dislocation site
concentration reads:
√̅̅̅ ( )/
N⊥ = 3 2ρ⊥ εeq p aFe (A2)
A2. Cementite lamellae
It is taken for verified that cementite-ferrite interface (CeFeI), but not bulk cementite, traps hydrogen at ambient temperature
[69,75,77,108,109]. Attributed to CeFeIs experimental binding and detrapping energies range from ~10 to ~60 kJ/mol
[69,70,82,85,108,110,111]. They refer to diverse cementite morphologies, which alter the interface conditions [98,112]. The binding
energies close to 20 kJ/mol were associated, in particular, with lamellar pearlite [70,108,111]. The CeFeI site concentrations have not
been measured. Ab-initio calculations furnished the CeFeI trap binding energies and planar concentrations for some interface orien
tation relationships (ORs) [75–78]. The Isaichev OR having the habit planes (101)Ce ‖(21̄1)Fe seems favourable for lamellar pearlite
[76,113]. There were revealed three deep traps with the binding energies about 24 kJ/mol per cementite (101)Ce face cell area
S(101)Ce = 0.81×0.51 nm2. Supposing these traps sufficiently distant as to not block one another, the trap surface density is MCe = 3/
S(101)Ce = 7.3×1018 m− 2. However, forced interface coherency, which is unlikely in real CeFeIs [75,113], vitiates these calculations.
Incoherencies at interfaces are compensable by misfit dislocations [113,114], which engender specific traps [75]. Misfit disloca
tions at Isaichev CeFeI form two arrays, which respective line and Burgers vectors Li and Bi (i = 1,2) in the BCC-iron lattice basis are
L1 = − B2 = 12[1̄1̄1]aFe and L2 = B1 = [01̄1̄]aFe , and array spacings are d1 = 13.3 nm and d2 = 10.4 nm [113]. Assuming one trap per
iron atom spacing along each dislocation [20,22,81], the surface density of CeFeI dislocation traps is MCe⊥ = (|L1 |d1 )− 1 + (|L2 |d2 )− 1 =
5.4×1017 m− 2. Their cores cover small fraction ∼ (|B1 |/d1 + |B2 |/d2 ) = 0.05 of the interface area. Then, the number of “regular”
interface traps out of the misfit dislocation cores is approximately MCe.
Generically, the surface density MZ of interface sites Z (Z is a placeholder for the interface and site labelling) determines their
volumic concentration as follows:
23
NZ = MZ Σ Z (A3)
where Σ Z is the interface area per unit volume. Taking into account that the cementite lamella widths, lengths and spacings are much
larger than their thickness s, CeFeI specific area is Σ Ce = 2/s = 2×107 m− 1 [76]. This yields the volumic concentrations of CeFeI regular
and misfit dislocation traps, respectively, NCe = 1.46×1026 m− 3 and NCe⊥ = 1.14×1025 m− 3.
The trapping properties of neither site are known for sure. Because hydrogen binding energy to strained interstices is symbatic with
the lattice strain [73], the forcing of CeFeI coherence by iron lattice stretching in the atomistic simulations should overvalue the
regular CeFeI binding energy ECe b . The mentioned modes of experimental data set and the estimate [108] prompt the assignment
ECe
b = 20 kJ/mol. Adapting to the general trend that trapping is the stronger the worse is the interface coherency [24,112,115], the next
value from the cited experimental energy data modes may be tentatively associated with CeFeI misfit dislocations. The binding energy
ECe⊥
b = 35 kJ/mol is adopted for them. Anyway, trap types corresponding to the measured trap energies have never been identified for
certain.
A3. Inclusions
Particles containing oxides, sulphides, etc., are usually present in steels, in pearlitic ones in particular [67–69,116]. These particles
often are some cores in the MnS encasements [116,117]. As well, per se MnS inclusions are typical in pearlitic steels [67,69,99].
Hydrogen is believed to be insoluble in MnS [14,16,61,79], and then only MnS-ferrite interface (MnSFeI) matters.
From the data for analogous steel [116], inclusions in the present one can be assumed globular MnS particles with average radius
rMnS = 1 μm. The molar content of sulphur (Table 1) determines the maximum mass amount of MnS in the steel, being excessive Mn the
impurity. The handbook density of MnS ρMnS = 4 g/cm3 yields then the volume fraction and average spacing of MnS particles,
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
respectively, φMnS = 1.3×10− 3 and dMnS = rMnS 3 43π/φMnS = 15 μm. The latter agrees with reported values [69]. The MnSFeI area per
steel unit volume is Σ MnS = 3φ/rMnS = 3.8×103 m− 1.
Experimental detrapping energies ascribed to MnS particles range from ~60 to ~90 kJ/mol [15,61,96,97,118]. Neither the trap
binding energies, nor their concentrations have been measured for certain. The ab-initio modelling [79,118] of the forcedly coherent
MnSFeI with cube-on-cube OR (001)MnS||(001)Fe and [100]MnS||[100]Fe hardly resembles real MnSFeIs, which must be semi-coherent
[119,120]. Although MnS shares the halite crystal form with extensively studied alloying element carbides MeC [24,60,84,121,122],
the Baker-Nutting OR (001)MeC||(001)Fe and [100]MeC||[110]Fe of MeC-ferrite interfaces is ruled out for MnSFeI, because its larger
lattice parameter aMnS ≈ 0.52 nm vs. that of carbides aMeC ~ 0.44 nm makes the MnSFeI mismatch at such OR too large [119]. Anyhow,
neither of the cited ORs has been corroborated for MnSFeIs, which structure identifications are controversial [79,119,120,123,124].
Plausible MnSFeI structure may be deduced using energetic-geometric arguments [113,125]. The habit planes “most apt to occur in
the act of crystallization … are …, in general, those which reticular densities are the most considerable” [126]. The ordering of MnS
and ferrite habit planes from more to less (≻) close-packed ones is as follows:
{
{ 001}MnS ≻ { 011}MnS ≻ { 111}MnS ≻ { 112}MnS ≻ …
(A4)
{ 110}Fe ≻ { 100}Fe ≻ { 112}Fe ≻ { 111}Fe ≻ …
The favourable combination of habit planes is conditioned by the smallness of lattice mismatch, which implies smaller energy of
strains fitting crystal planes one to another. Candidate plane pairs can be selected from (A4) basing on the Bramfitt’s mismatch [127]
3 ( )
1∑ h[klm]MnS cosϕ − h[klm]Fe
δB = (A5)
3 i=1 h[klm]Fe i
where h[klm]* (* is a wildcard for MnS or Fe) are in-plane atomic spacings taken along three (i = 1,2,3) low-index directions [klm]* within
a right angle quadrant, and ϕ is the angle between the involved directions [klm]MnS and [klm]Fe . Combinations of the planes from (A4)
render δB > ~5% impeding coherent interfaces, but some of them keep δB < ~20% allowing semi-coherent ones, where mismatches are
compensable by misfit dislocations [127,128]. The misfit dislocation theory [113,129–131] enables calculation of the interface
dislocation net content B and its decomposition into the chosen lattice dislocations Bi (i = 1,2) using the Frank-Bilby equation that
reads:
( −1 ) ∑(n × li )
B= ̂ S Fe − Ŝ− 1 p = ⋅ p Bi (A6)
MnS
i=1,2
di
− 1
where ̂ S * are the mappings of, respectively, the adjacent Fe and MnS lattices from their natural states onto some reference interface
lattice, n and l i = L i /|Li | are the unit vectors of, respectively, the interface normal and the dislocation lines Li, di are their spacings, and
p is arbitrary probe vector. Known the reference lattice and Bi-s, the solution of (A6) is straightforward [114]. However, the reference
lattice identification is a complicated problem of energy minimum [113,125,129–132], which has not been solved for MnSFeI. Here, a
simplified approach is implemented as follows.
Several combinations of habit planes from (A4) are examined, and few of them with smaller disregistries are chosen as the
candidate ORs. Ferrite lattice is set as the reference one [114,130] reducing S ̂ Fe to the identity transformation and the candidate Bi-s to
the ferrite dislocations. S
̂ MnS , li and di (i = 1,2) can then be calculated [114] from (A6). The optimal OR among the candidates must
24
minimise the interface energy. Its evaluation is difficult, and geometric parameters symbatic with the energy, such as the planar
density of misfit dislocations ρMnS⊥ = 1/d1 +1/d2 or the Ecob-Ralph parameter among others [125,129,130], may be adopted as
approximate indicators. The prelogarithmic factor u⊥ of the elastic energy of the mixed interface dislocations seems more suitable
[114,125,130,133] because it reflects the difference between screw and edge dislocation contributions to the energy [129,130]. It
reads [134]:
∑χ (Bi × li )2 + χ (Bi ⋅ li )2
u⊥ = e s
(A7)
i=1,2
di
where the factors associated with, respectively, edge and screw components of dislocations
[ ]
μ μ 1 1 μFe μMnS
χ e = Fe MnS + and χ s = (A8)
2π μFe (3 − 4νMnS ) + μMnS μMnS (3 − 4νFe ) + μFe 2π(μFe + μMnS )
depend on the shear moduli μ* and Poisson coefficients ν* , which are μMnS = 39 GPa with νMnS = 0.2, and μFe = 77 GPa with νFe =
0.3. Accordingly, the smallest Bi-s are not a priori the best option, which depends on the array’s edge-screw character determined by the
chosen Bi-s and the calculated for them Li-s. This makes it expedient to compare candidate ORs trying various Bi-s.
Being aware of limitations of the adopted OR favourability indicator, this procedure was accomplished for several combinations of
habit planes from (A4) revealing the most likely OR (110)MnS ‖(001)Fe and [1̄10]MnS ‖ [110]Fe , which is similar to the one suggested
elsewhere [79,122]. Calculated misfit dislocation arrays have B1 = [100]aFe with L1 = [0.39, 1, 0]aFe , B2 = [01̄0]aFe with L2 =
⃒ ⃒
[1, 0.39, 0]aFe , and the spacings d1,2 = 5.8⃒B1,2 ⃒.
Presuming mutual blocking of adjacent interstices within ~0.2 nm [107], reasonable assumption is that regular MnSFeI has one
trap per (001)Fe face cell area S(001)Fe = a2Fe . Corresponding planar trap concentration is MMnS ≈ (1 − a Fe ρMnS⊥ )/S(001)Fe = 8.3×1018
m− 2. One trap per iron interplane space along misfit dislocation is assumed. Corresponding planar density of misfit dislocation traps is
MMnS⊥ = (d−1 1 l1 ⋅ [010] + d−2 1 l2 ⋅ [1̄00])/aFe = 3.9×1018 m− 2. By (A3), the volumic concentrations of MnSFeI regular and misfit
dislocation traps, respectively, are NMnS = 3.1×1022 m− 3 and NMnS⊥ = 1.5×1022 m− 3.
Ascribed to micrometric MnS particles experimental detrapping energies scatter around ~70 kJ/mol [14,15,61,96,97,118,121].
The adopted assumption about saddle point energies implies the binding energy of 62 kJ/mol, which agrees with the maximum of the
formerly mentioned apparent modes of the experimental data set for (nearly) eutectoid pearlitic steels. Misfit dislocation traps are
expected to be stronger ones at MnSFeI, and this topmost value is assigned to them, EMnS⊥ b = 62 kJ/mol. Having no clues about regular
MnSFeI traps, the ratios of the binding energies of the two types of interface sites may be assumed similar for both MnSFeI and CeFeI,
i.e., EMnS⊥
b /EMnS
b ∼ ECe⊥ Ce
b /Eb . Therefrom, Eb
MnS
= 49 kJ/mol can be tentatively assigned.
A4. Vacancies
Trapping characteristics of vacancies are uncertain. Reported trap energies range from ~40 to ~60 kJ/mol
[14,73,74,78,105,106,135,136]. An intermediate value of the binding energy to vacancyEvac b = 50 kJ/mol is assigned. Measurements
of the vacancy concentration nvac for the considered steel are absent. The data for analogous steel [36] verify the linear fit nvac = ςρ ⊥ at
ς = 1.4×108 m− 1. Atomistic analyses showed two deep binding sites per vacancy [73]. Accordingly, the vacancy trap concentration
reads:
( ) ( )
Nvac εeq
p = 2nvac = 2ζρ⊥ εeq
p (A9)
A5. Solute impurities
Assuming that all the carbon and the sulphur in the steel are consumed to form cementite and MnS, the remaining elements from
Table 1 form the set of substitutional solutes Θ ≡ {Mn,Si,P,Cr,Ni,Mo}. The experiments imply that most cited elements are traps
[15,137]. The solute volumic concentrations nY (Y ∈ Θ) are determined by the steel composition. According to the ab-initio modelling
[136,138], tetrahedral interstices situated on the third coordination spheres near solutes are the most favourable trap locations. There
are 48 sites at each shell, but the H-H repulsion disables the interstices within ~0.2 nm [107] around the occupied positions. This
leaves about one sixth of the shell sites available for trapping. Accordingly, trap concentrations are NY = 8nY. The assignments of
binding energies EYb (Y ∈ Θ) based on experiments and ab-initio calculations [15,23,137,138] are listed in the Table 3.
References
[1] Johnson WH. II. On some remarkable changes produced in iron and steel by the action of hydrogen and acid. Proc Roy Soc London 1875;23:168–79. https://
doi.org/10.1098/rspl.1874.002.
25
[2] Reynolds O. On the effect of acid on the interior of iron wire. J Frankl Inst 1875;9:70–2. https://doi.org/10.1016/0016-0032(75)90215-X.
[3] Hughes DE. Note on some effects produced by the immersion of steel and iron wires in acidulated water. Nature 1880;21:602–3. https://doi.org/10.1038/
021602a0.
[4] Cotterill P. The hydrogen embrittlement of metals. Progr Mater Sci 1961;9:205–50. https://doi.org/10.1016/0079-6425(61)90005-6.
[5] Hydrogen degradation of ferrous alloys. Oriani RA, Hirth JP, Śmiałowski M, editors. Park Ridge: Noyes Publications; 1985.
[6] Gaseous hydrogen embrittlement of materials in energy technologies. Vols. 1 and 2. Gangloff RP, Somerday BP, editors. Oxford: Woodhead Publishing; 2012.
[7] Knott JF. Fundamentals of fracture mechanics. London: Butterworths; 1973.
[8] Kharin VS. Nucleation and growth of microcracks: an improved dislocational model and implications for ductile/brittle behaviour analysis. In: Blauel JG,
Schwalbe K-H, editors. Defect assessment in components − fundamentals and applications. London: MEP; 1991. p. 489–500.
[9] Novak P, Yuan R, Somerday BP, Sofronis P, Ritchie RO. A statistical, physical-based, micro-mechanical model of hydrogen-induced intergranular fracture in
steel. J Mech Phys Solids 2010;58(2):206–26. https://doi.org/10.1016/j.jmps.2009.10.005.
[10] Nagumo M. Hydrogen related failure of steels – a new aspect. Mater Sci Technol 2004;20:940–50. https://doi.org/10.1179/026708304225019687.
[11] Doshida T, Takai K. Dependence of hydrogen-induced lattice defects and hydrogen embrittlement of cold-drawn pearlitic steels on hydrogen trap state,
temperature, strain rate and hydrogen content. Acta Mater 2017;79:93–107. https://doi.org/10.1016/j.actamat.2014.07.008.
[12] Panasyuk VV, Andreikiv AE, Kharin VS. A model of crack growth in deformed metals under the action of hydrogen. Sov Mater Sci 1987;23(2):111–24. https://
doi.org/10.1016/j.actamat.2014.07.008.
[13] Pressouyre GM. Trap theory of hydrogen embrittlement. Acta Met 1980;28:895–911. https://doi.org/10.1016/0001-6160(80)90106-6.
[14] Bernstein IM, Pressouyre GM. The role of traps in the microstructural control of hydrogen embrittlement of steels. In: Oriani RA, Hirth JP, Śmiałowski M,
editors. Hydrogen degradation of ferrous alloys. Park Ridge: Noyes Publications; 1985. p. 641–85.
[15] Pressouyre GM. A classification of hydrogen traps in steel. Met Trans 1979;A10:1571–3. https://doi.org/10.1007/bf02812023.
[16] Hirth JP. Effects of hydrogen on the properties of iron and steel. Met Mater Trans 1980;A11:861–90. https://doi.org/10.1007/bf02654700.
[17] Oriani RA. The diffusion and trapping of hydrogen in steel. Acta Met 1970;18(1):147–57. https://doi.org/10.1016/0001-6160(70)90078-7.
[18] Kirchheim R. Solubility, diffusivity and trapping of hydrogen in dilute alloys. Deformed and amorphous metals − II. Acta Met 1982;30:1069–78. https://doi.
org/10.1016/0001-6160(82)90003-7.
[19] Toribio J, Kharin V. A generalised model of hydrogen diffusion in metals with multiple trap types. Phil Mag 2015;95:3429–51. https://doi.org/10.1080/
14786435.2015.1079660.
[20] Nagao A, Dadfarnia M, Somerday BP, Sofronis P, Ritchie RO. Hydrogen-enhanced-plasticity mediated decohesion for hydrogen-induced intergranular and
“quasi-cleavage” fracture of lath martensitic steels. J Mech Phys Solids 2018;112:403–30. https://doi.org/10.1016/j.jmps.2017.12.016.
[21] Turnbull A. Perspectives on hydrogen uptake, diffusion and trapping. Int J Hydrogen Energy 2015;40:16961–70. https://doi.org/10.1016/j.
ijhydene.2015.06.147.
[22] Dadfarnia M, Sofronis P, Neeraj T. Hydrogen interaction with multiple traps: can it be used to mitigate embrittlement? Int J Hydrogen Energy 2011;36:
10141–8. https://doi.org/10.1016/j.ijhydene.2011.05.027.
[23] Bhadeshia HKDH. Prevention of hydrogen embrittlement in steels. ISIJ Int 2016;56:24–36. https://doi.org/10.2355/isijinternational.ISIJINT-2015-430.
[24] Di Stefano D, Nazarov R, Hickel T, Neugebauer J, Mrovec M, Elsässer C. First-principles investigation of hydrogen interaction with TiC precipitates in alpha-Fe.
Phys Rev B 2016;93:184108. https://doi.org/10.1103/physrevb.93.184108.
[25] Barrera O, Bombac D, Chen Y, Daff TD, Galindo-Nava E, Gong P, et al. Understanding and mitigating hydrogen embrittlement of steels: a review of
experimental, modelling and design progress from atomistic to continuum. J Mater Sci 2018;53(9):6251–90. https://doi.org/10.1016/j.ijhydene.2011.05.027.
[26] Toribio J, Lancha AM, Elices M. Characteristics of the new tearing topography surface. Scr Met Mater 1991;25:2239–44. https://doi.org/10.1016/0956-716X
(91)90007-N.
[27] Toribio J, Lancha AM. Effect of cold drawing on susceptibility to hydrogen embrittlement of prestressing steel. Mater Struct 1993;26:30–7. https://doi.org/
10.1007/BF02472235.
[28] Toribio J, Lancha AM. Overload retardation effects on stress corrosion behaviour of prestressing steel. Constr Build Mater 1996;10:501–5. https://doi.org/
10.1016/0950-0618(96)00015-3.
[29] Toribio J, Lancha AM. Anisotropic stress corrosion cracking behaviour of prestressing steel. Mater Corros 1998;49:34–8. https://doi.org/10.1002/(sici)1521-
4176(199801)49:1<34::AID-MACO34>3.0.CO;2-A.
[30] Toribio J, Kharin V. Role of crack tip mechanics in stress corrosion cracking of high-strength steels. Int J Fract 2004;126:L57–62. https://doi.org/10.1023/b:
frac.0000026588.31654.e1.
[31] Toribio J, Aguado L, Lorenzo M, Kharin V. Hydrogen assisted cracking in pearlitic steel rods: the role of residual stresses generated by fatigue precracking.
Materials 2017;10:485. https://doi.org/10.3390/ma10050485.
[32] Kumnick AJ, Johnson HH. Deep trapping states for hydrogen in deformed iron. Acta Met 1980;28:33–9. https://doi.org/10.1016/0001-6160(80)90038-3.
[33] Borchers C, Al-Kassab T, Goto Sh, Kirchheim R. Partially amorphous nanocomposite obtained from heavily deformed pearlitic steel. Mater Sci Eng 2009;A502:
131–8. https://doi.org/10.1016/j.msea.2008.10.018.
[34] Zhang X, Godfrey A, Huang X, Hansen N, Liu Q. Microstructure and strengthening mechanisms in cold-drawn pearlitic steel wire. Acta Mater 2011;59:
3422–30. https://doi.org/10.1016/j.actamat.2011.02.017.
[35] Yang F, Fan YY, Jin YM. Measurement of dislocation density in cold-drawn steel wires by means of X-ray Bragg profiles analysis. Appl Mech Mater 2012;
184–185:1054–9. https://doi.org/10.4028/www.scientific.net/AMM.184-185.1054.
[36] Chen YZ, Csiszár G, Cizek J, Westerkamp S, Borchers C, Ungár T, et al. Defects in carbon-rich ferrite of cold-drawn pearlitic steel wires. Met Mater Trans 2013;
44(8):3882–9. https://doi.org/10.1007/s11661-013-1723-x.
[37] Chakraborty J, Ghosh M, Ranjan R, Das G, Das D, Chandra S. X-ray diffraction and Mössbauer spectroscopy studies of cementite dissolution in cold-drawn
pearlitic steel. Phil Mag 2013;93:4598–616. https://doi.org/10.1080/14786435.2013.838010.
[38] Borchers C, Kirchheim R. Cold-drawn pearlitic steel wires. Progr Mater Sci 2016;82:405–44. https://doi.org/10.1016/j.pmatsci.2016.06.001.
[39] Hirakami D, Manabe T, Ushioda K, Noguchi K, Takai K, Hata Y, et al. Effect of aging treatment on hydrogen embrittlement of drawn pearlitic steel wire. ISIJ Int
2016;56(5):893–8. https://doi.org/10.2355/isijinternational.ISIJINT-2015-735.
[40] Sofronis P, McMeeking RM. Numerical analysis of hydrogen transport near a blunting crack tip. J Mech Phys Solids 1989;37:317–50. https://doi.org/10.1016/
0022-5096(89)90002-1.
[41] Krom AHM, Koers RWJ, Bakker A. Hydrogen transport near a blunting crack tip. J Mech Phys Solids 1999;47:971–92. https://doi.org/10.1016/s0022-5096
(98)00064-7.
[42] Kanayama H, Shingoh T, Ndong-Mefane S, Ogino M, Shioya R, Kawai H. Numerical analysis of hydrogen diffusion problems using the finite element method.
Theor Appl Mech Jap 2008;56:389–400. https://doi.org/10.11345/nctam.56.389.
[43] Oriani R.A. The physical and metallurgical aspects of hydrogen in metals. In: Passell TO, editor. Proceedings of the fourth international conference on cold
fusion. Lahaina, 1993. Palo Alto: EPRI; 1994, 42 pp. [http://www.lenr-canr.org/acrobat/OrianiRAthephysica.pdf accessed 14 February 2023].
[44] Fukai Y. The metal-hydrogen system. Berlin: Springer; 2005.
[45] McNabb A, Foster PK. A new analysis of the diffusion of hydrogen in iron and ferritic steels. Trans Met Soc AIME 1963;227:618–27.
[46] Turnbull A. Modelling of environment assisted cracking. Corros Sci 1993;34:921–60. https://doi.org/10.1016/0010-938X(93)90072-O.
[47] Fultz B. Vibrational thermodynamics of materials. Progr Mater Sci 2010;55:247–352. https://doi.org/10.1016/j.pmatsci.2009.05.002.
[48] Choo WY, Lee JY. Thermal analysis of trapped hydrogen in pure iron. Met Mater Trans 1982;A13:135–40. https://doi.org/10.1007/BF02642424.
[49] Shanabarger MR. The isothermal kinetics of hydrogen adsorption onto iron films observed with the chemisorption-induced resistance change. Surface Sci
1985;150:451–79. https://doi.org/10.1016/0039-6028(85)90658-2.
26
[50] Toribio J, Kharin V. Factors affecting the intrinsic character of the crack growth kinetics curve in stress corrosion cracking: a review. Corros Rev 2001;19:
207–51. https://doi.org/10.1515/CORRREV.2001.19.3-4.207.
[51] Protopopoff E, Marcus Ph. Surface effects on hydrogen entry into metals. In: Marcus Ph, editor. Corrosion mechanisms in theory and practice. 2nd ed. New
York: Marcel Dekker; 2002. p. 53–96.
[52] Pisarev AA. Hydrogen adsorption on the surface of metals. In: Gangloff RP, Somerday BP, editors. Gaseous hydrogen embrittlement of materials in energy
technologies, Volume 2. Oxford: Woodhead Publishing; 2012. p. 3–26. https://doi.org/10.1533/9780857095374.1.3.
[53] Liu Q, Atrens AD, Shi Zh, Verbeken K, Atrens A. Determination of the hydrogen fugacity during electrolytic charging of steel. Corros Sci 2014;87:239–58.
https://doi.org/10.1016/j.corsci.2014.06.033.
[54] Martínez-Pañeda E, Díaz A, Wright L, Turnbull A. Generalised boundary conditions for hydrogen transport at crack tips. Corros Sci 2020;173:108698. https://
doi.org/10.1016/j.corsci.2020.108698.
[55] Venezuela J, Tapia-Bastidas C, Zhou Q, Depover T, Verbeken K, Gray E, et al. Determination of the equivalent hydrogen fugacity during electrochemical
charging of 3.5NiCrMoV steel. Corros Sci 2018;132:90–106. https://doi.org/10.1016/j.corsci.2017.12.018.
[56] Toribio J, Kharin V. Finite-deformation analysis of the crack-tip fields under cyclic loading. Int J Solids Struct 2009;46:1937–52. https://doi.org/10.1016/j.
ijsolstr.2009.01.006.
[57] Toribio J, Kharin V. Role of cyclic pre-loading in hydrogen assisted cracking. In: Kane RD, editor. Environmentally assisted cracking: predictive methods for
risk assessment and evaluation of materials, equipment, and structures. West Conshohocken: ASTM STP 1401; 2000. p. 329–42. https://doi.org/10.1520/
STP10228S.
[58] Turnbull A. Hydrogen diffusion and trapping in metals. In: Gangloff RP, Somerday BP, editors. Gaseous hydrogen embrittlement of materials in energy
technologies, Volume 2. Oxford: Woodhead Publishing; 2012. p. 89–128. https://doi.org/10.1533/9780857095374.1.89.
[59] Ebihara K, Suzudo T, Kaburaki H, Takai K, Takebayashi Sh. Modeling of hydrogen thermal desorption profile of pure iron and eutectoid steel. ISIJ Int 2007;47:
1131–40. https://doi.org/10.2355/isijinternational.47.1131.
[60] Pressouyre GM, Bernstein IM. A quantitative analysis of hydrogen trapping. Met Trans 1978;A9:1571–80. https://doi.org/10.1007/bf02661939.
[61] Lee JL, Lee JY. Hydrogen trapping in AISI 4340 steel. Metal Sci 1983;13:426–32. https://doi.org/10.1179/030634583790420619.
[62] Li D, Gangloff RP, Scully JR. Hydrogen trap states in ultrahigh-strength AERMET 100 steel. Met Trans 2004;A35:849–64. https://doi.org/10.1007/s11661-
004-0011-1.
[63] Song EJ, Suh D-W, Bhadeshia HKDH. Theory for hydrogen desorption in ferritic steel. Comput Mater Sci 2013;79:36–44. https://doi.org/10.1016/j.
commatsci.2013.06.008.
[64] Kadin Yu, Gamba M, Faid M. Identification of the hydrogen diffusion parameters in bearing steel by evolutionary algorithm. J Alloys Comp 2017;705:475–85.
https://doi.org/10.1016/j.jallcom.2017.02.208.
[65] Leblond JB, Dubois D. A general mathematical description of hydrogen diffusion in steels − II. Numerical study of permeation and determination of trapping
parameters. Acta Met 1983;31:1471–8. https://doi.org/10.1016/0001-6160(83)90143-8.
[66] Turnbull A, Saenz de Santa Maria M, Thomas ND. The effect of H2S concentration and pH on hydrogen permeation in AISI 410 stainless steel in 5% NaCl.
Corros Sci 1989;29:89–104. https://doi.org/10.1016/0010-938X(89)90082-6.
[67] Porter DA, Easterling KE, Smith GDW. Dynamic studies of the tensile deformation and fracture of pearlite. Acta Met 1978;26:1405–22. https://doi.org/
10.1016/0001-6160(78)90156-6.
[68] Lewandowski JJ, Thompson AW. Micromechanisms of cleavage fracture in fully pearlitic microstructures. Acta Met 1987;35:1453–62. https://doi.org/
10.1016/0001-6160(87)90091-5.
[69] Enos DG, Scully JR. A critical-strain criterion for hydrogen embrittlement of cold-drawn, ultrafine pearlitic steel. Met Mater Trans 2002;A33:1151–66. https://
doi.org/10.1007/s11661-002-0217-z.
[70] Takai K, Watanuki R. Hydrogen in trapping states innocuous to environmental degradation of high-strength steels. ISIJ Int 2003;43:520–6. https://doi.org/
10.2355/isijinternational.43.520.
[71] Wen M, Fukuyama S, Yokogawa K. Atomistic simulations of effect of hydrogen on kink-pair energetics of screw dislocations in bcc iron. Acta Mater 2003;51:
1767–73. https://doi.org/10.1016/S1359-6454(02)00575-X.
[72] Jiang DE, Carter EA. Diffusion of interstitial hydrogen into and through bcc Fe from first principles. Phys Rev B 2004;70:064102. https://doi.org/10.1103/
PhysRevB.70.064102.
[73] Ramasubramaniam A, Itakura M, Carter EA. Interatomic potentials for hydrogen in α-iron based on density functional theory. Phys Rev B 2009;79:174101.
https://doi.org/10.1103/PhysRevB.79.174101.
[74] Kimizuka H, Ogata Sh. Slow diffusion of hydrogen at a screw dislocation core in α-iron. Phys Rev B 2011;84:024116. https://doi.org/10.1103/
PhysRevB.84.024116.
[75] Kawakami K, Matsumiya M. Ab-initio investigation of hydrogen trap state by cementite in bcc-Fe. ISIJ Int 2013;53:709–13. https://doi.org/10.2355/
isijinternational.53.709.
[76] Mirzoev AA, Verkhovykh AV, Okishev KY, Mirzaev DA. Hydrogen interaction with ferrite/cementite interface: ab initio calculations and thermodynamics.
Molecular Phys 2018;116:482–90. https://doi.org/10.1080/00268976.2017.1406161.
[77] McEniry EJ, Hickel T, Neugebauer J. Ab initio simulation of hydrogen-induced decohesion in cementite-containing microstructures. Acta Mater 2018;150:
53–8. https://doi.org/10.1016/j.actamat.2018.03.005.
[78] Kholtobina AS, Pippan R, Romaner L, Scheiber D, Ecker W, Razumovskiy VI. Hydrogen trapping in bcc Iron. Materials 2020;13(10):2288. https://doi.org/
10.3390/ma13102288.
[79] Xiao H, Huang F, Peng Z, Fan L, Liu J. Sequential kinetic analysis of the influences of non-metallic inclusions on hydrogen diffusion and trapping in high-
strength pipeline steel with Al-Ti deoxidisation and Mg treatment. Corros Sci 2022;195:110006. https://doi.org/10.1016/j.corsci.2021.110006.
[80] Galindo-Nava EI, Basha BIY, Rivera-Díaz-del-Castillo PEJ. Hydrogen transport in metals: integration of permeation, thermal desorption and degassing. J Mater
Sci Technol 2017;33:1433–47. https://doi.org/10.1016/j.jmst.2017.09.011.
[81] Krom AHM, Bakker A. Hydrogen trapping models in steel. Met Mater Trans 2000;B31:1475–82. https://doi.org/10.1007/s11663-000-0032-0.
[82] Ramunni VP, De Paiva Coelho T, de Miranda PEV. Interaction of hydrogen with the microstructure of low-carbon steel. Mater Sci and Eng 2006;435–436:
504–14. https://doi.org/10.1016/j.msea.2006.07.089.
[83] Mirzaev DA, Yakovleva IL, Tereshchenko NA, Tabatchikova TI, Okishev KYu, Mirzoev AA, et al. Possibilities of hydrogen atom trapping in steels by ferrite/
cementite interfaces. 2. Adsorption theory. Bul South Ural State Univ Ser Metallurgy 2014;14:30-9 (in Russian). [http://dspace.susu.ac.ru/xmlui/handle/
0001.74/5028 accessed 14 February 2023].
[84] Wei FG, Hara T, Tsuzaki K. Precise determination of the activation energy for desorption of hydrogen in two Ti-added steels by a single thermal-desorption.
Met Mater Trans 2004;B35:587–97. https://doi.org/10.1007/s11663-004-0057-x.
[85] Hong G-W, Le J-Y. The interaction of hydrogen and the cementite-ferrite interface in carbon steel. J Mater Sci 1983;18:271–7. https://doi.org/10.1007/
BF00543835.
[86] da Silva JRG, Stafford SW, McLellan RB. The thermodynamics of the hydrogen-iron system. J Less-Common Met 1976;49:407–20. https://doi.org/10.1016/
0022-5088(76)90052-7.
[87] Kiuchi K, McLellan RB. The solubility and diffusivity of hydrogen in well-annealed and deformed iron. Acta Met 1983;31:961–84. https://doi.org/10.1016/
0001-6160(83)90192-X.
[88] Castedo A, Sanchez J, Fullea J, Andrade MC, de Andres PL. Ab initio study of the cubic-to-hexagonal phase transition promoted by interstitial hydrogen in iron.
Phys Rev B 2011;84:094101. https://doi.org/10.1103/PhysRevB.84.094101.
[89] Jeng HW, Chiu LH, Johnson DL, Wu JK. Effect of pearlite morphology on hydrogen permeation, diffusion, and solubility in carbon steels. Met Mater Trans
1990;A21:3257–9. https://doi.org/10.1007/BF02647319.
27
[90] Yu S-H, Lee S-M, Lee S, Nam J-H, Lee J-S, Bae C-M, et al. Effects of lamellar structure on tensile properties and resistance to hydrogen embrittlement of pearlitic
steel. Acta Mater 2019;172:92–101. https://doi.org/10.1016/j.actamat.2019.04.040.
[91] Toribio J, Kharin V. Role of fatigue crack closure stresses in hydrogen-assisted cracking. In: McClung RC, Newman JC, editors. Advances in fatigue crack
closure measurement and analysis: second volume. West Conshohocken: ASTM STP 1343; 1999. p. 440–58. https://doi.org/10.1520/STP15773S.
[92] Hutchinson JW. Singular behaviour at the end of a tensile crack in a hardening material. J Mech Phys Solids 1968;16:13–31. https://doi.org/10.1016/0022-
5096(68)90014-8.
[93] Leblond JB, Dubois D. A general mathematical description of hydrogen diffusion in steels − I. Derivation of diffusion equations from Boltzmann-type transport
equations. Acta Met 1983;31:1459–69. https://doi.org/10.1016/0001-6160(83)90142-6.
[94] McMeeking RM. Finite deformation analysis of crack tip opening in elastic-plastic materials and implications for fracture. J Mech Phys Solids 1977;25:357–81.
https://doi.org/10.1016/0022-5096(77)90003-5.
[95] Toribio J, Kharin V. Simulations of fatigue crack growth by blunting–re-sharpening: plasticity induced crack closure vs. alternative controlling variables. Int J
Fatigue 2013;50:72–82. https://doi.org/10.1016/j.ijfatigue.2012.02.019.
[96] Beidokhti B, He P, Kokabi AH, Dolati A. Control of hydrogen cracking in the welded steel using microstructural traps. Mater Sci and Technol 2017;33:408–14.
https://doi.org/10.1080/02670836.2016.1219499.
[97] Peng Zh, Liu J, Huang F, Hu Q, Cao Ch, Hou Sh. Comparative study of non-metallic inclusions on the critical size for HIC initiation and its influence on
hydrogen trapping. Int J Hydrogen Energy 2020;45:12616–28. https://doi.org/10.1016/j.ijhydene.2020.02.131.
[98] Ju CP, Rigsbee JM. Interfacial coherency and hydrogen damage in plain carbon steel. Mater Sci Eng 1988;102:281–8. https://doi.org/10.1016/0025-5416(88)
90583-6.
[99] Lewandowski JJ, Thompson AW. Microstructural effects on the cleavage fracture stress of fully pearlitic eutectoid steel. Met Trans 1986;A17:1769–86. https://
doi.org/10.1007/BF02817275.
[100] Hagi H, Hayashi Y. Effect of dislocation trapping on hydrogen and deuterium diffusion in iron. Trans Jap Inst Met 1987;28:368–74. https://doi.org/10.2320/
matertrans1960.28.368.
[101] Findley KO, O’Brien MK, Nako H. Critical Assessment 17: mechanisms of hydrogen induced cracking in pipeline steels. Mater Sci Technol 2015;31:1673–80.
https://doi.org/10.1080/02670836.2015.1121017.
[102] Matsumoto R, Taketomi Sh, Matsumoto S, Miyazaki N. Atomistic simulations of hydrogen embrittlement. Int J Hydrogen Energy 2009;34:9576–84. https://
doi.org/10.1016/j.ijhydene.2009.09.052.
[103] Zhao Y, Lu G. QM/MM study of dislocation-hydrogen/helium interactions in α-Fe. Model Simul Mater Sci Eng 2011;19:065004. https://doi.org/10.1088/
0965-0393/19/6/065004.
[104] Itakura M, Kaburaki H, Yamaguchi M, Okita T. The effect of hydrogen atoms on the screw dislocation mobility in bcc iron: a first-principles study. Acta Mater
2013;61:6857–67. https://doi.org/10.1016/j.actamat.2013.07.064.
[105] Bombac D, Katzarov IH, Pashov DL, Paxton A. Theoretical evaluation of the role of crystal defects on local equilibrium and effective diffusivity of hydrogen in
iron. Mater Sci Technol 2017;33:1505–14. https://doi.org/10.1080/02670836.2017.1310417.
[106] Tehranchi A, Zhang X, Lu G, Curtin WA. Hydrogen-vacancy-dislocation interactions in α-Fe. Model Simul Mater Sci Eng 2017;25(2):025001. https://doi.org/
10.1088/1361-651X/aa52cb.
[107] Westlake DG. Stoichiometries and interstitial site occupation in the hydrides of ZrNi and other isostructural intermetallic compounds. J Less-Common Met
1980;75:177–85. https://doi.org/10.1016/0022-5088(80)90115-0.
[108] Hagi H. Effect of interface between cementite and ferrite on diffusion of hydrogen in carbon steels. Mater Trans JIM 1994;35:168–73. https://doi.org/
10.2320/matertrans1989.35.168.
[109] Adachi N, Ueno H, Onoe K, Morooka S, Todaka Y. Hydrogen permeation property of bulk cementite. ISIJ Int 2021;61:2320–2. https://doi.org/10.2355/
isijinternational.ISIJINT-2021-134.
[110] Choo WY, Lee JY. Hydrogen trapping phenomena in carbon steel. J Mater Sci 1982;17:1930–8. https://doi.org/10.1007/BF00540409.
[111] Robertson WM, Thompson AW. Permeation measurements of hydrogen trapping in 1045 steel. Met Mater Trans 1980;A11:553–7. https://doi.org/10.1007/
BF02670691.
[112] Robertson WM, Thompson AW. Authors’ reply. Met Mater Trans 1983;A14(1):158. https://doi.org/10.1007/BF02643752.
[113] Guziewski M, Coleman SP, Weinberger CR. Interface energetics and structure of the pearlitic microstructure in steels: an atomistic and continuum
investigation. Acta Mater 2018;155:1–11. https://doi.org/10.1016/j.actamat.2018.05.051.
[114] Bhadeshia HKDH. Geometry of crystals, polycrystals, and phase transformations. Boca Raton: CRC Press; 2018.
[115] Shi R, Ma Y, Wang Z, Gao L, Yang X-S, Qiao L, et al. Atomic-scale investigation of deep hydrogen trapping in NbC/α-Fe semi-coherent interfaces. Acta Mater
2020;200:686–98. https://doi.org/10.1016/j.actamat.2020.09.031.
[116] Rodríguez R. Daño microestructural producido durante la fabricación de aceros de pretensado y consecuencias en su comportamiento en fractura. Zamora:
Universidad de Salamanca; 2015. PhD Thesis.
[117] van der Eijk .C, Walmsley J., Grong Ø. Effects of titanium containing oxide inclusions on steel weldability. In: David S, DebRoy T, Lippold J, Smartt H, Vitek J,
editors. Trends in welding research, proceedings of the 6th international conference, Callaway Gardens Resort, Phoenix, 2002, Metals Park: ASM Int; 2003. p.
730-5.
[118] Peng Z, Liu J, Huang F, Hu Q, Cheng Z, Liu S, et al. Effect of submicron-scale MnS inclusions on hydrogen trapping and HIC susceptibility of X70 pipeline steels.
Steel Res Int 2018;89(7):1700566. https://doi.org/10.1002/srin.201700566.
[119] Mills AR, Thewlis G, Whiteman JA. Nature of inclusions in steel weld metals and their influence on formation of acicular ferrite. Mater Sci Technol 1987;3:
1051–61. https://doi.org/10.1179/mst.1987.3.12.1051.
[120] Yong Q, Sun X, Yang G, Zhang Z. Solution and precipitation of secondary phase in steels: phenomenon, theory and practice. In: Weng Y, Dong H, Gan Y,
editors. Advanced Steels. Berlin, Heidelberg: Springer; 2011. p. 109–17. https://doi.org/10.1007/978-3-642-17665-4_14.
[121] Lee J-Y, Lee SM. Hydrogen trapping phenomena in metals with B.C.C. and F.C.C. crystal structures by the desorption thermal analysis technique. Surf Coat
Technol 1986;28:301–14. https://doi.org/10.1016/0257-8972(86)90087-3.
[122] Wei FG, Tsuzaki K. Quantitative analysis on hydrogen trapping of TiC particles in steel. Met Mater Trans 2006;A37:331–53. https://doi.org/10.1007/s11661-
006-0004-3.
[123] Lin C-K, Pan Y-C, Su Y-H-F, Lin G-R, Hwang W-S, Kuo J-C. Effects of Mg-Al-O-Mn-S inclusion on the nucleation of acicular ferrite in magnesium-containing low-
carbon steel. Mater Charact 2018;141:318–27. https://doi.org/10.1016/j.matchar.2018.05.005.
[124] Yamada T, Terasaki H, Komizo Y. Microscopic observation of inclusions contributing to formation of acicular ferrite in steel weld metal. Sci Technol Weld Join
2008;13:118–25. https://doi.org/10.1179/174329308X271797.
[125] Bollmann W. Crystal defects and crystalline interfaces. Berlin, Heidelberg: Springer; 1970.
[126] Bravais MA. Études cristallographiques. Paris: Gauthier-Villars; 1866. p. 168.
[127] Bramfitt BL. The effect of carbide and nitride additions on the heterogeneous nucleation behavior of liquid iron. Met Trans 1970;1:1987–95. https://doi.org/
10.1007/BF02642799.
[128] Turnbull D, Vonnegut B. Nucleation catalysis. Ind Eng Chem 1952;44(6):1292–8. https://doi.org/10.1021/ie50510a031.
[129] Knowles KM. The dislocation geometry of interphase boundaries. Phil Mag 1982;46:951–69. https://doi.org/10.1080/01418618208236943.
[130] Hirth JP, Pond RC, Hoagland RG, Liu X-Y, Wang J. Interface defects, reference spaces and the Frank-Bilby equation. Progr Mater Sci 2013;58:749–823. https://
doi.org/10.1016/j.pmatsci.2012.10.002.
[131] Wang J, Zhang R, Zhou C, Beyerlein I, Misra A. Characterizing interface dislocations by atomically informed Frank-Bilby theory. J Mater Res 2013;28:1646–57.
https://doi.org/10.1557/jmr.2013.34.
28
[132] Chen Y, Shao S, Liu X-Y, Yadav SK, Li N, Mara N, et al. Misfit dislocation patterns of Mg-Nb interfaces. Acta Mater 2017;126:552–63. https://doi.org/10.1016/
j.actamat.2016.12.041.
[133] Lothe J. Dislocations interacting with surfaces, interfaces or cracks. In: Indenbom VL, Lothe J, editors. Elastic strain fields and dislocation mobility.
Amsterdam: North-Holland; 1992. p. 329–89. https://doi.org/10.1016/B978-0-444-88773-3.50012-1.
[134] Matthews JW. Chapter 8 − Coherent interfaces and misfit dislocations. In: Matthews JW, editor. Epitaxial growth, Part B. New York: Academic Press; 1975.
p. 559–609. https://doi.org/10.1016/B978-0-12-480902-4.50011-2.
[135] Iwamoto M, Fukai Y. Superabundant vacancy formation in iron under high hydrogen pressures: thermal desorption spectroscopy. Mater Trans JIM 1999;40:
606–11. https://doi.org/10.2320/matertrans1989.40.606.
[136] Counts WA, Wolverton C, Gibala R. First-principles energetics of hydrogen traps in alpha-Fe: point defects. Acta Mater 2010;58:4730–41. https://doi.org/
10.1016/j.actamat.2010.05.010.
[137] Hagi H. Effect of substitutional alloying elements (Al, Si, V, Cr, Mn Co, Ni, Mo) on diffusion coefficient of hydrogen in α-Iron. Mater Trans JIM 1992;33:472–9.
https://doi.org/10.2320/matertrans1989.33.472.
[138] Mirzaev DA, Mirzoev AA, Rakitin MS. Alloying effects on thermodynamic characteristics of hydrogen in bcc iron. Bull South Ural State Univ Ser Metallurgy
2016;16(04):40–53. https://doi.org/10.14529/met160405.
29

Kharin EngFfractMech2023

Uploaded by

Copyright:

Available Formats

Kharin EngFfractMech2023

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Kharin EngFfractMech2023

Uploaded by

Copyright:

Available Formats

Engineering Fracture Mechanics 291 (2023) 109473

Contents lists available at ScienceDirect

Engineering Fracture Mechanics

Crack tip hydrogen diffusion with multiple type traps and

E-mail address: gatogris@usal.es.

2. Basement diffusion-trapping theory

2.1. Transport equations

Fig. 1. Schematic potential energy diagram for hydrogen travelling in metal.

JAB = − dAB (YB ∇CA − CA ∇YB − βVH YB CA ∇σm ) (2)

The partial differential equation (PDE) of total balance is as usual:

2.2. Equilibria in diffusion-trapping

where KBA = ξBA exp(βEBA

Ctot = Φ(CL )CL (9)

2.3. Hydrogen entry into metal

CL (x, t)|x∈Γ = CLeq (x, fin )|x∈Γ (13)

3. Prototype experiments (the modelling case)

0.74 0.70 0.20 0.016 0.023 0.01 0.01 0.001 balance

195 725 53 2120 5.8 2160 17

4. Diffusion-trapping parameter assignments

Ferrite dislocation (⊥) 35 1.2×106 from 5.9 to 148

Mn solute (Mn) 8 24.7 4.3×103

5. Diffusion-trapping model reduction

are small, γ Tr ≪ 1 (NTr ≪ Ntot).

and the partial relative absorptivities (8) read:

CLeq (x, t) = CL0 (fin )eβVH σm (x,t) (16)

5.1. Is local trap-lattice equilibrium applicable?

(a) the initial trap-lattice imbalance [CLE

5.2. Do the diffuser movements passing by traps matter?

The total flux then reads:

5.3. Case-adapted diffusion-trapping equations

6. Calculation model setup

provides the input for diffusion calculations.

7. Crack tip stress–strain field

8. Diffusion-trapping computation results

residual stresses induced by fatigue pre-cracking and of the loading rate.

9. Implications for HAC

9.1. Deleterious and innocuous hydrogen states

(I) to increase the critical values θcr

where σcrn is the critical stress at the fracture event.

9.2. On the withdrawal of hydrogen from deleterious sites by innocuous traps

(caption on next page)

(caption on next page)

beneficiate the HE resistance until they get full.

9.3. On the slowdown by traps of hydrogen access to the deleterious states

∝ NTr could enhance the effect.

9.4. Case dependence of the potentiality of traps to mitigate hydrogen embrittlement

CRediT authorship contribution statement

Declaration of Competing Interest

Data will be made available on request.

Appendix. Diffusion-trapping parameter assignments

A1. Ferrite dislocations

where ρ0 = 4×1014 m− 2 and ρsat = 1016 m− 2

A2. Cementite lamellae

A5. Solute impurities

You might also like