2022 Volume 62 Issue 10 Pages 2107-2117
It is well known that the presence of hydrogen causes deterioration of the mechanical properties of steel, which appears in the forms of reduced fracture toughness, shorter fatigue life, etc., and these phenomena are recognized as hydrogen embrittlement. Here, the effect of hydrogen on crack initiation in fracture toughness tests was investigated using a combination of experimental and computational approaches. Tempered lath martensitic steel was subjected to fracture toughness tests with a monotonically rising load in air and high-pressure hydrogen gas environments. While cracking propagated continuously within grains in the air environment, cracking in the hydrogen environment grew by linking of isolated interfacial failures ahead of the main crack tip. To understand the nucleation mechanism of isolated failure in the presence of hydrogen, tensile simulations of twist grain boundaries (TGBs) rotated around the <110> axis at various misorientation angles were conducted using molecular dynamics (MD) simulations. While dislocation emission from TGB rotated 70° is the dominant deformation mode in the absence of hydrogen, rupture along TGB rotated 110° and 170° without stress relaxation due to dislocation emission is the dominant deformation mode in the presence of hydrogen. As a consequence, it is indicated that the origin of hydrogen-induced isolated crack initiation in the vicinity of a fatigue precrack is rupture along the block boundaries within the martensitic structure due to hydrogen-induced inhibition of dislocation emission from grain boundaries (GBs).
For the realization of a hydrogen society, the development of steel materials for infrastructure facilities such as pressure vessels for high pressure hydrogen gas and hydrogen pipelines is expected. However, it is well known that hydrogen in steel causes deterioration of mechanical properties such as reduction of elongation at break, reduction of fracture toughness and increase of the fatigue crack propagation rate by hydrogen addition. This phenomenon is called hydrogen embrittlement, and is particularly remarkable in high strength steels.1,2,3) In infrastructure facilities for high pressure hydrogen gas, design which considers not only fracture due to high internal pressure but also fracture due to hydrogen embrittlement is necessary.4,5) As one safety measure for preventing accidents in pressure vessels and piping, the need to satisfy “leak-before-break” (LBB) has been pointed out. LBB is the concept of preventing large-scale accidents by reducing the internal pressure by leakage of the internal fluid at the same time as a crack at the inside surface of a pressure vessel or piping propagates and penetrates in the plate thickness direction. To achieve this, the fracture toughness value of the residual part in the pressure vessel or piping must to be higher than the fracture driving force at crack initiation. Fracture toughness values in the presence of hydrogen are evaluated by fracture toughness tests in a high pressure hydrogen gas environment3) and cathode hydrogen charging environment,6) and it is known that fracture toughness values decrease remarkably in the presence of hydrogen.
Fracture toughness means the resistance to fracture when a macroscopic mechanical load is applied to a specimen with a crack-like defect. The dominant factor in the extreme reduction in fracture toughness that occurs in the presence of hydrogen is considered to be hydrogen-induced cracking (HIC). To consider the mechanism of hydrogen-induced cracking, it is important to divide HIC into the initiation and propagation of microcracks. Although the martensite microstructure is indispensable in high strength steels, martensite contains various lattice defects, and it is essential to identify the lattice defect that becomes the origin of microcrack initiation. In this study, monotonic tensile-type fracture toughness tests were conducted in a high pressure hydrogen gas environment, and crack initiation was defined as the point where a load corresponding to the macroscopic mechanical response begins to decrease rapidly in the load-crack opening displacement curve. The moment of microcrack initiation was captured by stopping the loading, and the microscopic fracture phenomenon at a pre-induced fatigue crack tip was observed. To clarify the mechanism of crack initiation in a hydrogen gas environment, the relationship between the microscopic fracture phenomenon and the microstructure of the steel was investigated.
Mainly three mechanisms are discussed as the typical mechanisms contributing to hydrogen embrittlement: HELP (hydrogen-enhanced localized plasticity),7,8) in which hydrogen contributes to the interaction between dislocations, HEDE (hydrogen-enhanced decohesion),9,10) in which hydrogen reduces the interstitial bonding force, and HESIV (hydrogen-enhanced strain-induced vacancies),11,12) in which hydrogen stabilizes the formation of atomic vacancies and vacancy clusters. These proposed mechanisms refer to the interaction of hydrogen with lattice defects such as dislocations, grain boundaries and atomic vacancies. However, there are still many unclear points regarding the initiation and propagation processes of fracture in hydrogen embrittlement phenomena. The effects of hydrogen embrittlement are various, depending on the strength level, physical properties and mechanical properties of the material. Recently, the idea that these mechanisms act in a complex manner has become the mainstream. Therefore, it is important to clarify the processes leading to microscopic fracture including the precursor stage, and to understand the main roles that hydrogen plays in each process.
Because direct in-situ observation of the behavior of lattice defects such as dislocations, grain boundaries and atomic vacancies in steel under applied stress is extremely difficult, the role of lattice defects in microcrack initiation was investigated by molecular dynamics (MD) calculation,13,14,15) which makes it possible to analyze the activity of lattice defects under stress in the presence of hydrogen by simplifying and modeling the lattice defects. For example, in a previous study by Wan et al.16) using the MD calculation approach, tensile simulations of twist grain boundaries (TGBs) in the presence of hydrogen were conducted. That study clarified the fact that the elementary processes of dislocation emission and absorption on the grain boundary, void formation and growth on the grain boundary, and fracture along the grain boundary occurred in order. However, many of the MD calculations in previous studies focused on highly symmetrical grain boundaries and did not consider the correspondence with the martensite microstructure. Although the details will be described in Chapter 2, detailed observations of fracture toughness tests in a hydrogen gas environment suggested that microcracks were initiated not on the prior austenite grain boundaries, but along large angle grain boundaries inside prior austenite grains. Therefore, simplified modeling of block boundaries, which are the most primitive high angle grain boundaries in the hierarchical martensite microstructure, was investigated using MD calculations. It is known that a crystallographic correspondence exists in the crystal orientation relationship between each block with different variants in the martensite structure.17) The misorientation between blocks is limited to the specific misorientation angle around the 〈110〉 axis in the body centered cubic lattice. Therefore, in order to discuss whether fracture induced from block boundaries can be a dominant factor in hydrogen-induced microcrack initiation, the fracture behavior of grain boundaries with a crystallographic misorientation relationship on blocks and non-blocks was compared by using MD calculations of bicrystal models with systematically changed misorientation angles around the 〈110〉 axis.
The purpose of this study is to propose a method for investigating the relationship between the fracture origin and steel microstructure in hydrogen-induced cracking of high strength steels by combining an experimental approach using fracture toughness tests to clarify the phenomenon of microcrack initiation, and a computational approach using MD calculations to elucidate the deformation and fracture behavior of a single grain boundary without other lattice defects.
The chemical composition of the JIS-SCM 435 steel investigated in this study is shown in Table 1. Figure 1 shows an optical microscope (OM) and scanning electron micrograph (SEM) image of the microstructure of the specimen at a thickness of 1/4 t. The sample shows a tempered lath martensite structure in which cementite is formed in the lath boundary and laths. Table 2 shows the mechanical properties of the sample material.
| C | Mn | Si | P | S | Cr | Mo |
|---|---|---|---|---|---|---|
| 0.34 | 0.21 | 0.74 | 0.023 | 0.005 | 1.06 | 0.17 |
(a) OM image and (b) SEM image of microstructure of JIS-SCM435 steel (nital etching).
| YS MPa | TS MPa | EL % | RA % |
|---|---|---|---|
| 815 | 935 | 23.1 | 67.8 |
A compact tension (CT) specimen, as shown in Fig. 2, was prepared from the center of the plate thickness so that the crack propagation direction was perpendicular to the rolling direction. Fatigue precracks were introduced in all CT specimens up to a crack length of 30.5 mm under the fatigue test conditions of a frequency of 5 Hz and load ratio of 0.1 in the air environment. Then, a monotonic tensile-type fracture toughness test (RL test) was conducted at room temperature in air and 40 MPa hydrogen gas using a material testing facility for high pressure hydrogen gas environments. The opening displacement speed in the RL tests was 2.5 μm/min.
Dimensions (mm) of compact tension (CT) specimen used in monotonic tensile-type fracture toughness tests.
In order to investigate the crack initiation process in the RL test, in this study, tensile deformation was stopped before and after crack initiation, and CT specimens subjected immediately to unloading were obtained. The CT specimen was cut in half in the thickness direction by wire cut electro-discharge machining. One half-cut specimen was polished to remove the heat-affected zone and etched with nital. The relationship between crack initiation and the microstructure was then investigated by the electron backscatter diffraction (EBSD) method. The other half-cut specimen was forcibly ruptured immediately after immersion in liquid nitrogen, and the fracture surface was observed with an OM and SEM.
2.2. Results of RL TestFigure 3 shows the load-opening displacement curves obtained from the RL tests conducted in the air and 40 MPa hydrogen gas environments. In the air environment, crack mouth opening displacement (CMOD) increased linearly, transitioned gradually to a nonlinear response over approximately 30 kN, and decreased gradually after the maximum load was exceeded. It is considered that the microcrack initiates in the transition from the linear region to the nonlinear region. The microcrack initiation point of the RL test in the air environment was at CMOD=0.8 mm. Non-initiation of microcracking when the RL test was stopped at CMOD=0.7 mm was also confirmed by observing the cross section at the center of thickness of the sample obtained under that condition. In the hydrogen gas environment, the load decreased rapidly at approximately 13 kN, which was less than half of the load at the microcrack initiation point in the air environment. In the hydrogen gas environment, it is considered that the microcrack initiates at the starting point of load reduction. The microcrack initiation point in the RL test in the hydrogen gas environment was at CMOD=0.24 mm. As in the air environment, in the RL test in the H2 gas environment, the central cross section of the sample from the test stopped at CMOD=0.20 mm was also observed, and it was confirmed that no microcrack initiation had occurred. The points examined to investigate microcrack initiation and to confirm non-initiation of microcracking are indicated by arrows in Fig. 3. Here, in the air environment, both the initiation point and the non-initiation point were in the nonlinear region, but in the hydrogen gas environment, the non-initiation point was in the linear region, while the initiation point was in the nonlinear region.
Load-CMOD curves obtained by tests conducted in air and 40 MPa H2 gas environments. The open and closed arrows indicate the positions examined to confirm non-initiation of microcracking and to observe the position of crack initiation, respectively. (Online version in color.)
Figure 4 shows the results of OM and SEM observation of the cross section of the microcrack initiation point determined as described above. In the RL test in the air environment, it was found that the tip of the fatigue precrack had a blunted shape and a large opening, and a microcrack with no opening had initiated from the tip of the blunted fatigue precrack. A completely different aspect was confirmed at the microcrack initiation point in the RL test in the hydrogen gas environment. The shape of the fatigue precrack tip was very sharp and not clearly opened. In addition, the microcracks connected continuously from the tip of the fatigue precrack observed in the air environment were not observed in the hydrogen gas environment. In the hydrogen gas environment, isolated microcracks with a length of several tens of μm were found to initiate discretely over a region about 300 μm in front of the fatigue precrack tip. Here, the size of the plastic zone18) in front of the fatigue precrack tip s=(K/σys)2/6π is estimated by assuming the plane strain state in linear fracture mechanics. K is the stress intensity factor at CMOD=0.24 mm and σys is the YS value in Table 2. The calculated size was s=221 μm. Thus, linear fracture mechanics suggests that the microcracks in the hydrogen gas environment are initiated beyond the plastic region of the precrack tip. In addition, it is also important to understand the plastic deformation behavior in the plastic region of the crack tip affected by hydrogen.19)
(a, b) OM and (c) SEM images of cross-sections at crack initiation (nital etching). (a) In the air environment, loading was stopped at CMOD=0.80 mm. (b, c) In the 40 MPa H2 gas environment, loading was stopped at CMOD=0.24 mm. (Online version in color.)
Figures 5 and 6 show OM and SEM images of the fracture surfaces of microcrack initiation obtained by forced fracture under a liquid nitrogen atmosphere for the specimens tested in the air and hydrogen gas environments, respectively. In the OM observation results in Figs. 5(a) and 6(a), the fracture surfaces induced by the precrack introduction test, the RL test and low temperature forced rupture, respectively, could be clearly distinguished. It was confirmed that the fracture surfaces induced by the precrack introduction test and low temperature forced rupture were similar in the air and hydrogen gas environments. It was also found that the fracture surface induced by the RL test in the air environment was uniform, whereas the fracture surface in the hydrogen gas environment had two kinds of fracture surfaces in which completely different aspects coexisted. In the fracture surface morphology in the RL test in the air environment, as shown in Figs. 5(b) and 5(c), a stretch zone was observed in the vicinity of the precrack, which is commonly reported in high toughness steels, and dimples were observed in front of it. On the other hand, the fracture surface morphology in the RL test in the hydrogen gas environment in Fig. 6(c) displayed a mixture of “Quasi-cleavage (QC)” fracture surfaces and flat fracture surfaces. This fracture surface is different from the “QC” fracture surface induced by low temperature forced rupture shown in Figs. 5(d) and 6(d). Thus, the “QC” fracture surface observed in the RL test in the hydrogen gas environment is considered to be a hydrogen-induced “QC” fracture surface. It may also be noted that the morphology of the hydrogen-induced “QC” in this study is consistent with those in previous studies.20,21)
Fracture surface of sample ruptured in liquid nitrogen using specimen loaded to CMOD=0.80 mm in air environment, observed by (a) OM and (b)–(d) SEM.
Fracture surface of sample ruptured in liquid nitrogen using specimen loaded to CMOD=0.24 mm in 40 MPa H2 gas environment, observed by (a) OM and (b)–(d) SEM.
Based on the results of cross-sectional observation at microcrack initiation and the fracture surface observation results obtained by low temperature forced rupture, the mechanism of microcrack initiation and propagation in fracture toughness tests will be discussed. In the air environment, it is considered that a microcrack is initiated at the tip of the opened fatigue precrack, and this microcrack propagates continuously. In contrast, in the hydrogen gas environment, the fatigue precrack does not open, and an isolated microcrack is initiated in the front region away from the tip of the fatigue precrack, after which the isolated crack discretely induces the next initiation of an isolated microcrack, and the crack propagates by this process. In the vicinity of the deviation point where the load-CMOD curve in the hydrogen gas environment deviates from the curve in the air environment, as shown in Fig. 3, the load does not change to a decrease immediately after the deviation point at CMOD=0.22 mm. The gradient of the load becomes more moderate after the deviation point and negative where the maximum load is reached at CMOD=0.28 mm. The change of the gradient after the deviation point is thought to correspond to the existence of a ligament part during initiation of the discrete microcracks. Resistance to the macroscopic load also remains due to the existence of the ligament, but further loading ruptures the ligament, eliminating the resistance to the macroscopic load, and as a consequence, the load decreases. Based on this, it is considered that the region from the starting point of the gradient change to the maximum load point is microcrack initiation, and the region after the maximum load point is microcrack propagation.
In order to clarify the correspondence between the microcrack initiation site and the steel microstructure in the RL test, the vicinity of the microcrack tip in the air and H2 gas environments was analyzed by the EBSD method. The isolated microcrack initiated at the most distant position from the fatigue precrack tip was selected as the microcrack tip in the hydrogen gas environment. The inverse pole figure (IPF) and kernel average misorientation (KAM) maps are shown in Fig. 7. From the IPF map, the microcrack in the air environment was initiated inside a prior austenite grain. However, the isolated microcracks in the hydrogen environment were initiated along the high angle grain boundaries inside the prior austenite grain. From the KAM map, in the air environment, a dense region of high KAM values exists around the microcrack tip, and an extension of large misorientation region corresponding to the plastic deformation was observed. In the hydrogen gas environment, regions with high KAM values were observed sparsely in the vicinity of the microcrack, suggesting that plastic deformation such as dislocation activation was not clearly induced.
EBSD analysis of vicinity of crack tips generated in (a, b) air and (c, d) 40 MPa H2 gas environments. (a) and (c) show IPF maps, and (b) and (d) are KAM maps corresponding to the areas of (a) and (c), respectively. (Online version in color.)
From the results of the fracture surface observation in the RL test in the hydrogen gas environment in Fig. 6(c), hydrogen-induced “QC” occupies a large area, and from the results of the cross-sectional observation by EBSD in Fig. 7(d), it is considered that microcracks were initiated inside the prior austenite grain. This suggests that microcrack initiation is induced by lattice defects inside the prior austenite grain. In particular, understanding the behavior of high-angle grain boundaries as candidates for lattice defects is important from the following two viewpoints. First, growth and accumulation of dislocations at the tip of a fatigue precrack like that which occurs in the air environment are less likely to occur in the hydrogen gas environment. Second, plastic restraint is thought to act strongly in the hydrogen gas environment since the fatigue precrack tip is very sharp and the radius of curvature is very small, at several μm. Therefore, the characteristics of block boundaries, which are the most frequent high-angle boundaries in prior austenite grains, were analyzed using MD calculations.
The author considered that deformation inside blocks is suppressed due to the high multiaxial state at the fatigue crack tip. Therefore, in the MD calculations in this study, a boundary condition which constrains lattice deformation in the direction perpendicular to the tensile direction was applied. Since blocks are groups of the same habit planes, the ratio of the area occupied by the twist grain boundary (TGB) is considered to be higher than that of the tilt grain boundary, and for this reason, the MD calculations focused on TGBs. Although the interaction between high-angle grain boundaries and dislocations is also an important characteristic in real materials with intrinsic dislocations and dislocation sources inside blocks, it should be noted that the MD calculations in this study were performed with a simplified model that includes only grain boundaries as lattice defects in order to understand the characteristics of the TGBs themselves. In addition, the amount of hydrogen added in the MD calculations is approximately 100 wt ppm, which is larger than the amount of hydrogen measured in experiments. This is the same level of hydrogen as that at which the effect of hydrogen appeared remarkably in the previous study using MD calculation.16) On the other hand, for example, the amount of hydrogen permeated into the sample held in 40 MPa high pressure hydrogen gas without loading is approximately 0.3 wt ppm. However, when stress is applied, it is predicted that the amount of hydrogen penetration near the crack tip under stress will increase extremely due to the increased size of the interstitial sites and the large local stress concentration in the vicinity of lattice defects such as dislocations and grain boundaries. Although quantitative measurement of the amount of hydrogen in samples under the stress is difficult, this is a problem for future study. In Chapter 3, MD calculations were carried out to understand the characteristics of the grain boundaries under a condition in which a qualitatively large amount of hydrogen penetrates as a stress concentration source at the crack tip and the effect of hydrogen appears remarkably.
Calculation models of the TGBs described in Section 2.3 were prepared. In order to construct a simplified system, in this study, one grain was fixed and the other grain was rotated in a two-grain boundary model, as shown in Fig. 8(a), and the misorientation angles of the TGBs were systematically changed. MD calculations were performed for the grain boundary with block misorientation relationship and other misorientations. In the modeling, the x-axis of grain A was the [111] direction and the z-axis was the [110] direction. Here, grain B was rotated in a positive half-hour clockwise direction with respect to the z-axis. In order to match the periodicity of grains A and B along the x- and y-axes, the size of the calculation model was changed at each misorientation angle. The length along the z-axis was uniform. The misorientation angles and sizes of the calculation models are shown in Table 3. In addition to the twin relationship, the orientation relationships of 50°, 60°, −10°, and −60° around the 〈110〉 axis correspond to the relationships of blocks.17) In this study, the twin relationship corresponds to the misorientations of 70° and 110°. Misorientation relationships of −10° and −60° correspond to misorientations of 170° and 120° around the 〈110〉 axis, respectively.
(a) Schematic of atomic modeling of twist grain boundary in [110] direction. Initial atomic model for MD calculations displayed by using (b) central symmetry parameter and (c) corresponding distribution of hydrogen atoms. (Online version in color.)
| Rotation angle (°) | Model size (nm) | ||
|---|---|---|---|
| θ | x | y | z |
| 10 | 14.9 | 14.0 | 20.2 |
| 30 | 11.9 | 14.0 | 20.2 |
| 50 | 14.9 | 14.0 | 20.2 |
| 60 | 14.9 | 16.8 | 20.2 |
| 70 | 14.9 | 14.0 | 20.2 |
| 90 | 11.9 | 14.0 | 20.2 |
| 110 | 14.9 | 14.0 | 20.2 |
| 120 | 14.9 | 14.0 | 20.2 |
| 130 | 14.9 | 14.0 | 20.2 |
| 150 | 14.9 | 14.0 | 20.2 |
| 170 | 14.9 | 14.0 | 20.2 |
The Large-scale Atomic/Molecular Massively Parallel Simulator (LAMMPS) code22) was applied for the MD calculations using the embedded-atom method potential function developed by Ramasubramaniam et al.23) as the interatomic potential between Fe and H. To simulate the behavior at finite temperatures, a calculation cell corresponding to the lattice constant at the temperature concerned was created. The temperature was controlled by the Nosé-Hoover method24,25) so as to maintain a constant temperature at all times. The temperature was 300 K and the time step was 1 fs. Periodic boundary conditions were also applied.
Atomeye26) was used as a visualization software for the calculation models. Lattice defects were analyzed using the center symmetry parameter (CSP).27) CSP is a parameter that quantitatively calculates the symmetry between an atom and its neighboring atoms, where lower symmetry of the atoms corresponds to a higher CSP. Since it is thought that lattice defects are composed of atoms with a local structure without periodicity, lattice defects were identified by displaying only atoms with high CSP. In this study, atoms with CSP of more than 0.015 were visualized to identify lattice defects. It should be noted that in the presence of hydrogen, not only lattice defects, but also the atoms affected by lattice distortion due to the trapped hydrogen within interstitial sites, were visualized. In addition, relaxed configurations of the grain boundary models were obtained by forming the initial velocity of the atoms corresponding to 1 K based on the normal distribution after rotating the grain B shown in Fig. 8(a) at an arbitrary angle, heating it to 1000 K and holding it for 500 ps, and then lowering the temperature to 300 K and holding it for 500 ps. In the presence of hydrogen, the same temperature history was given to the state in which the hydrogen atoms were arranged by random selection from tetrahedral sites in the body-centered cubic lattice. For example, Figs. 8(b) and 8(c) show a calculation model rotated by 70°. It can be seen that the accumulation of hydrogen on the grain boundaries can be obtained due to the temperature history.
In the MD calculations, a tensile simulation in the z-axis direction by strain control was conducted. The strain rate was 1.0×108. In order to verify the effect of the constraint of the lattice deformation (cell length elongation) in the direction perpendicular to the tensile direction (x- and y-axis directions) described in Section 2.3, a comparison between tensile simulations with and without constraint was conducted using a grain boundary with a misorientation angle of 70°. The stress-strain curves obtained by the MD calculations are shown in Fig. 9. In the simulation without constraint, stress relaxation behavior was confirmed. Here, a dislocation was released from one grain boundary upon stress reduction and was absorbed into the other grain boundary. This stress relaxation was repeated as the tensile simulation progressed, and although the simulation was carried out up to strain of 12%, it did not lead to fracture. Stress relaxation due to dislocation emission and absorption from the grain boundary was also confirmed in the simulation with constraint, but the frequency of dislocation emission was less than that in the simulation without constraint. Stress increased again after dislocation emission, and finally, fracture was confirmed at strain of 8%. In the simulation without constraint, the interaction between grain boundaries and dislocations was the dominant factor for deformation, indicating that a simulation with constraint is preferable for evaluating characteristics in situations where plastic deformation is difficult to occur due to a multiaxial stress state.
Stress-strain curves of tensile MD calculations on twist grain boundary with 70° rotation angle in pure iron system. (Online version in color.)
As a result of the tensile simulations of the TGBs with the misorientation angles shown in Table 3, the results could be broadly classified into four patterns based on the differences in deformation and fracture behavior. As typical results, a stress-strain curve at the grain boundary with the misorientation angle of 70° is shown in Fig. 10. The stress-strain curves at the misorientation angles of 110°, 90° and 60° are shown in Fig. 11.
(a) Stress-strain curves of tensile MD calculations for twist grain boundaries with a 70° rotation angle in absence and presence of hydrogen atoms. (b) to (e) Snapshots of MD calculations displayed by the central symmetry parameter to investigate lattice defects. The yellow arrows in (b) to (e) indicates the twist grain boundary. (Online version in color.)
Stress-strain curves of tensile MD calculations for twist grain boundaries with rotation angles of (a) 110°, (b) 90° and (c) 60°. (Online version in color.)
First, the result for the grain boundary with the misorientation angle of 70° is shown in Fig. 10, and is called pattern A. From Fig. 10(b), in the absence of hydrogen, dislocation was nucleated from the grain boundary at strain of 3.0%. Then, a stress rise was confirmed after stress relaxation. Fracture along the grain boundary was observed at strain of 9.1%. In the presence of hydrogen, tensile deformation proceeded without dislocation emission and stress relaxation, and fracture along the grain boundary was observed at strain of 8.4%. At the moment of fracture in the presence of hydrogen, nucleation of a large number of dislocations from the grain boundaries was confirmed. This suggests that cleavage along grain boundaries is not a simple phenomenon. In pattern A, the results indicated that dislocation emission was suppressed by the addition of hydrogen.
Second, the result for the misorientation angle of 110° is shown in Fig. 11(a) and is called pattern B. In the absence of hydrogen, deformation proceeded linearly and fracture along the grain boundary occurred at strain of 8.6% without dislocation emission. However, this did not mean that the whole plane was broken. That is, the existence of a partial remaining ligament was confirmed. In the presence of hydrogen, a stress rise was confirmed after deviation from the linear region with dislocation emission at strain of 4.4%, and fracture along the grain boundary occurred at strain of 8.3%. In pattern B, it was confirmed that dislocation emission was promoted by the addition of hydrogen. In other words, the effect of hydrogen was exactly the opposite as in pattern A.
Third, the result for the misorientation angle of 90° is shown in Fig. 11(b), and is called pattern C. The dislocation emission confirmed in patterns A and B was confirmed in both the presence and the absence of hydrogen, and it was found that stress was kept almost constant. Fracture along the grain boundary was also observed in both calculations. Thus, addition of hydrogen did not change the fracture process. Fracture along the grain boundary with dislocation emission is indicated in pattern C.
Finally, the result for the misorientation angle of 60° is shown in Fig. 11(c), and is called pattern D. There was no dislocation emission in the presence or absence hydrogen, and fracture along the grain boundary occurred during linear deformation. In the presence of hydrogen, resistance due to the ligament remained slightly after the stress drop at strain of 9.3%. Since this resistance finally vanished, it is considered not to be stress relaxation by dislocation emission, but to be fracture during linear deformation. As in pattern C, the fracture process in pattern D was not changed by hydrogen addition, but it is interesting that the reverse fracture process was suggested in terms of the existence of dislocations.
The stresses of the first phenomenon occurring in time series at each misorientation angle are summarized and shown in Fig. 12. When a fracture was induced after a dislocation emission, the stress on the dislocation emission was plotted. In all tensile simulations, the final fracture was fracture along the grain boundary with dislocation emission, regardless of the presence or absence of dislocation emission before fracture. Pattern D was observed with the largest number of misorientation angles calculated in this paper. Patterns A and B were confirmed only at grain boundaries with one misorientation angle each, and it is interesting that these were both misorientation angles with a twin relationship. In order to discuss the origin of fracture, here, it is considered that the phenomenon induced by the lowest stress in Fig. 12 is preferentially induced among the misorientation angles in this study. It is suggested that dislocation emission at the 70° grain boundary occurs preferentially in the absence of hydrogen, while fracture in the 120° and 170° grain boundaries occurs preferentially in the presence of hydrogen. It is interesting that the grain boundary misorientations of 70°, 120° and 170° are all block misorientation relationships. This indicates that fracture originating from the block boundaries is preferentially induced in the presence of hydrogen, while plastic deformation originating from the block boundaries is preferentially induced in the absence of hydrogen.
Stress of preferentially-activated event in MD calculations with various misorientation angles of twist grain boundaries. A, B, C and D indicate the qualitative classification of the stress-strain curves corresponding to Figs. 10(a), 11(a), 11(b) and 11(c), respectively. (Online version in color.)
In these MD calculations, the modeling and conditions were very limited, in that only the hydrogen concentration of 100 wt ppm and the twist grain boundary rotated around the 〈110〉 axis were examined. As future problems, it will be necessary to consider the hydrogen concentration dependence, other rotation axes and misorientation angles, and tilt grain boundaries. In this paper, the misorientation angles of the TGBs rotated around the 〈110〉 axis, which represents the majority of cases of the block relationship, were changed systematically, and it was found that the characteristics of the grain boundary differed depending on the misorientation angle. Therefore, it is suggested that remarkable anisotropy at the grain boundaries of blocks is an essential factor for clarifying the mechanism of hydrogen-induced fracture.
In Chapter 2, the phenomenon of hydrogen embrittlement was examined from the macroscopic viewpoint by using fracture toughness tests, and the origin of fracture was investigated from the microscopic viewpoint by SEM and EBSD observation. In Section 2.2, initiation of microcracks in the presence of hydrogen was defined as a phenomenon in which multiple isolated microcracks are initiated discretely in the region from the deviation point from the linear region in the load-CMOD curve to the load maximum point, and propagation of microcracks was defined as the phenomenon in which the isolated microcracks connect in the region after the load maximum point. In Chapter 3, simulations of the deformation and fracture characteristics of TGBs rotated around the 〈110〉 axis were conducted from the viewpoint of lattice defects using MD calculations.
First, the initiation of microcracks in the presence of hydrogen will be discussed. The results of the MD calculations indicated that a high-angle grain boundary with a misorientation relationship of the blocks in the martensite structure fractured most preferentially among TGBs rotated around the 〈110〉 axis. Since the model size used in the MD calculations was several 10 nm, it should be noted that the fracture in the tensile simulations does not directly correspond one-to-one with microcrack initiation, which is a macroscopic mechanical response. Nevertheless, the fracture in the MD calculations captures the primitive damage in the precursor stage of isolated microcracks with a length of several 10 μm observed in Fig. 4(c). It is considered that the nucleation site of this damage is on the grain boundaries between blocks, and the size corresponds to the size of the blocks. A mechanism of microcrack initiation in the presence of hydrogen is also suggested, in which the primitive damage grows to a size of several 10 μm and remains discontinuously in the form of isolated microcracks. When a load is applied to an isolated microcrack, the stress field with plastic constraint at both ends of the isolated microcrack is estimated to be a factor in the initiation of a new isolated microcrack forward from the original microcrack in a crack propagation direction, as is also seen in the case of a fatigue precrack tip. Thus, it is considered that the origin of the isolated microcrack at the microcrack initiation point in the fracture toughness test in the presence of hydrogen is the primitive damage induced on block boundaries due to suppression of dislocation emission from the grain boundary by the hydrogen effect. The decrease in the macroscopic stress slope is considered to be due to a reduction of the stressed area resulting from multiple initiation of isolated microcracks at the point of deviation from the linear region of the load-CMOD curve in the presence of hydrogen. However, a quantitative analysis by a finite element analysis is needed, and this is a problem for future study. Since it is not possible to consider the change of the macroscopic property of fracture toughness only by MD calculations, it is important to combine atomistic approaches with other techniques which can handle different scales.
Regarding microcrack propagation in the presence of hydrogen after multiple initiation of isolated microcracks, it is suggested that the reduction of the loaded area due to the rupture of the ligaments between the isolated microcracks decreases the macroscopic stress in the fracture toughness test. Thus, the microscopic stress state in microcrack propagation is different from that in the macro tensile direction, and the mechanism changes accordingly. Because the ligaments are affected not only by macroscopic tensile stress, but also by shear stress, it is not sufficient merely to conduct tensile simulations of the grain boundaries by MD calculations when investigating hydrogen-induced crack initiation in the fracture toughness test.
The MD calculations confirmed that the fracture mode was not simple cleavage on the surface of block boundaries. On the high-angle grain boundaries without dislocation emission induced during deformation, dislocation emission from the grain boundaries was induced at the moment of rupture, and the atomic configuration of the original grain boundary was not maintained. As a consequence, newly-formed surfaces were nucleated with emission of lattice defects such as dislocations. Therefore, it is unlikely that the cross section of the fracture surfaces will be sharp enough to be identified as a block boundary when observed experimentally. On the other hand, previous studies20,28,29,30) on hydrogen-induced “QC” fracture using mechanical tests showed that fracture occurred along {110} planes, although there was a difference in the crack propagation path between fracture along laths and fracture within laths. Since the habit plane of block boundaries is the {110} plane, the suggestion that the block boundaries are the initiation site of isolated microcracks is not necessarily contradictory. However, these MD calculations using only high-angle grain boundaries do not include the dislocation source and lath boundary within a prior austenite grain. Their defects are important factors in discussing the origin of fracture, and are also a subject for future examination. In any case, the present study is expected to contribute to the elucidation of the mechanism of the hydrogen embrittlement phenomenon, because basic knowledge focusing only on the grain boundary can be grasped by applying the MD calculations as in this study. Although also noted in Chapter 3, there are some elements which are not considered in the MD calculations, and a systematic accumulation of study of those elements is considered important in the future.
Monotonic tensile-type fracture toughness tests were conducted in air and high pressure hydrogen gas environments. The phenomena of microcrack initiation induced at the deviation point in the macroscopic mechanical response were observed in detail. In the presence of hydrogen, it was found that the precrack tip did not open clearly and isolated microcracks were initiated discretely. In addition, from a comparison of the KAM values in EBSD observations of the initiated microcrack tip, plastic deformation was suppressed in the presence of hydrogen.
MD calculations using models of twist grain boundaries (TGBs) rotated around the 〈110〉 axis with various misorientation angles were conducted assuming the absence or presence of hydrogen. To investigate fracture along the grain boundary, calculation conditions with constrained lattice deformation in the direction perpendicular to the tensile direction were applied in tensile simulations. In the absence of hydrogen, it was found that stress relaxation was induced preferentially at the grain boundary with the misorientation angle of 70° due to dislocation emission from the grain boundary. However, in the presence of hydrogen, fracture along the grain boundary without stress relaxation due to dislocation emission from the grain boundary was induced preferentially at the grain boundaries with misorientation angles of 120° and 170°. It is suggested that the suppression of dislocation emission from the grain boundaries by hydrogen addition induces fracture along the grain boundaries.
From the findings obtained from the fracture toughness tests and MD calculations, it was suggested that the origin of hydrogen-induced microcrack initiation at the deviation point of the load-CMOD curve is micro-scale damage induced on block boundaries by the suppression of dislocation emission from grain boundaries. This paper proposes a method of utilizing MD calculations to understand the mechanism of fracture phenomena in hydrogen embrittlement of steels.
I would like to express my heartfelt gratitude to Dr. Nobuyuki Ishikawa and Dr. Shusaku Takagi of JFE Steel Corporation for their invaluable guidance and advice in writing this article.
