2025 Volume 65 Issue 5 Pages 644-649
The interaction of hydrogen with lattice defects plays a crucial role in the hydrogen embrittlement mechanism, but the origin of hydrogen-related defects remains unclear. In this study we investigate the formation process of hydrogen-induced vacancies in austenitic stainless steel SUS 304 by positron annihilation lifetime spectroscopy. Positron lifetime measurements of hydrogen-charged samples subjected to tensile testing by different strains show that the formation of hydrogen-induced vacancies first appears when a strain of about 5% is applied. Electropolishing of the hydrogen-charged layer reveals the generation of vacancy-hydrogen complexes in the bulk underneath the hydrogen-charged layer, which develop into vacancy clusters by further application of stress. From the PALS results and complementary X-ray diffraction analysis, the dislocation density required for the formation of hydrogen-induced vacancies is quantitatively determined. Clarification of the conditions for the formation of hydrogen-induced vacancies provides important input and reference data for models of the hydrogen embrittlement in stainless steels.
A hydrogen-based society in which hydrogen is utilized in everyday life and industrial activities represents one step in the path towards ultimately achieving a decarbonized society. However, hydrogen embrittlement (HE) represents a major hinder to the deployment of hydrogen energy in the society. Despite the fact that a unified understanding of the HE mechanism has not yet been reached, in recent years, the hydrogen-enhanced strain-induced vacancy (HESIV) model has gain particular interest.1) According to this model, hydrogen can accumulate in areas of high tensile stress, such as grain boundaries, where dislocation cutting during tensile testing creates atomic vacancies. Hydrogen binds to the vacancies, stabilizes them and promotes their aggregation. These vacancies, in turn, form clusters, which can destabilize the plastic deformation. The HESIV model highlights the importance of strain-induced vacancies, which are generated in the presence of hydrogen2) due to the reduction of the vacancy formation energy.3,4) In light of the crucial role of hydrogen-induced vacancies, it has become clear that it is necessary to correctly grasp and evaluate materials not only from a phenomenological perspective but also from an atomistic perspective, such as lattice defects and their interaction with hydrogen.
Recent studies of hydrogen-charged austenitic stainless steels by low-temperature thermal desorption spectroscopy have shown that hydrogen trap sites are increased in strained specimens, where enhanced hydrogen desorption from vacancy-type defects was found (see, e.g.,5,6) and references therein). Positron annihilation spectroscopy, which is a non-destructive and powerful technique for the detection of lattice defects,7) has also revealed that hydrogen-induced vacancies do not form simply by addition of hydrogen, but that stress load is a condition for their generation.8,9) In particular, hydrogen-induced vacancy clusters as large as 13 atomic vacancies have been detected in the fractured specimens of the most common type of austenitic stainless steel AISI 304.10) In addition, a high density of vacancies and their clustering has been revealed in other stainless steels, iron and nickel.11,12,13) These vacancies, which are understood to develop in localized high-strain regions, reduce the energy barrier for crack propagation and promote crack nucleation and growth.5) Although these hydrogen-enhanced strain-induced vacancies are thought to be at the heart of HE, their precise formation process and their role in the HE mechanism remain mostly unknown. Clarifying how and under which conditions hydrogen-induced vacancies form represents an important piece of the puzzle in understanding the HE mechanism.
In this research, we investigate the formation process of hydrogen-induced vacancies in austenitic stainless steel 304 by positron annihilation lifetime spectroscopy (PALS) with the ultimate aim of shedding light on their role in the HE mechanism. Samples are strained by different strains up to 10% after hydrogen charge and PALS measurements are carried out at each strain value to determine the strain load at which hydrogen-induced vacancies appear and their formation process. In addition, the mechanical conditions necessary for the generation of hydrogen-induced vacancies are estimated using a simplified model.
Austenitic stainless steel SUS 304 plates purchased from Nilaco Co. were used after processing into dumbbell-shaped specimens with a gauge length of 20 mm, width of 10 mm and thickness of 0.2 mm for the tensile tests and PALS measurements. To remove any initial defects in the samples, solution treatment was performed using an infrared gold image furnace. The samples were annealed at 900°C for 5 min under a continuous flux of Ar gas, after raising the temperature at a heating rate of 4°C/min over 4 hours. The samples were then gradually cooled down to room temperature over a couple of hours, before extraction from the furnace. Preliminary PALS measurements on the samples after solution treatment evidenced a single positron lifetime of 110 ± 2 ps, which is consistent with the known bulk lifetime value of iron-based materials such as 304 steel. This indicates that the solution treatment performed under the above conditions was successful in removing the initial defects.
After solution treatment, hydrogen was added to the samples by the cathodic electrolytic hydrogen charging method. An electrolytic solution made of a 3% NaCl aqueous solution and 3 g/L NH4SCN was used. The hydrogen charge was carried out with a current density of 50 A/m2 at room temperature for 48 hours. Under these conditions, the hydrogen concentration in the samples is 178 ppm and hydrogen diffuses up to a depth of about 10 μm, with a concentration profile that decreases exponentially from the surface.10) Despite the relatively small hydrogen diffusion volume, hydrogen-charged 304 steel samples are hydrogen embrittled, as demonstrated by tensile tests in earlier studies.8,10)
Subsequently, the samples were subject to tensile testing up to various strain values in the range from 0 to 10%, i.e., spanning both the elastic region (strains smaller than about 3%) and the plastic region (strains >3%). The reason for preparing specimens with strains up to 10% only in this study, is that the formation of hydrogen-induced defects in specimens with strains of 10% or higher has already been confirmed in previous studies.8,10) Tensile tests in the plastic region were performed at a strain rate of 4.17×10−4 s−1 using a small tensile tester manufactured by Deben Co. (2 kN model). For the tensile tests in the elastic region, a benchtop tensile testing machine (model FTN1-13A) manufactured by Aiko Engineering Co. was used to enable PALS measurements in situ under stress load (see below). In this case the applied strain rate was 4.17×10−3 s−1.
Electrolytic polishing was also conducted on the hydrogen-charged samples strained by 6%, to confirm the formation of vacancy-hydrogen complexes in the bulk underneath the hydrogen-charged layer. A mixed solution of sulfuric acid and phosphoric acid (volume ratio 2:3) was used as the electrolytic solution and electropolishing was done at room temperature with an electric current of 4 A for 180 s. As a result, a surface layer with a thickness of about 10 ± 2 μm was removed.
In this study, PALS measurements were performed on all samples to analyze the generation of hydrogen-related defects in the samples. PALS is a method that enables to detect lattice defect in materials non-destructively and with high sensitivity, and has been described in detail elsewhere.7) Here, a 22Na radioisotope embedded between two thin Kapton foils was used as the positron source, which was sandwiched in between two identical samples for the measurements. In this configuration, positrons probe a depth of up to ~130 μm in the sample. Positron lifetime spectra were collected using a digital positron lifetime spectrometer with a time resolution function of the device of about 180 ps full width at half maximum. The lifetime spectra were analyzed with the method of least squares using the PALSfit software,14) by fitting the spectrum with the sum of multiple exponential decay functions, each representing a different lifetime component from positrons annihilating in the bulk or defects. All the measured lifetime spectra were successfully fitted with two or three lifetime components, with the variance of the fits ranging from 1.0 to 1.2. The background and the source component stemming from positrons annihilating in the Kapton foils were subtracted from the spectra before analysis. The results reported here are the average of at least three measurements and the uncertainties represent the standard errors of those measurement. In addition, note that the PALS measurements on the samples strained by 0 to 3% were performed in situ while applying stress load to the sample, because this strain interval corresponds to the elastic region. For specimens strained by more than 3%, the PALS measurements were performed after unloading the samples after the tensile deformation.
From a comparison of the measured positron lifetimes with those calculated from first principles for iron,15) one can determine the type of defects in 304 stainless steel. Moreover, from the intensity of each lifetime component, the relative concentration of each defect type can be estimated. For reference, here we recall that, in bcc-Fe, the positron lifetime of a perfect crystal (bulk) is about 110 ps, that of dislocations ranges from around 120 to 160 ps, while that of a monovacancy is 180 ps. However, since monovacancies are unstable at room temperature, lifetime values of about 140 to 170 ps are generally estimated to be dislocations, since the lifetimes of dislocations and monovacancies are close and it is often impossible to resolve them in the lifetime analysis. Positron lifetimes longer than 190 ps are ascribed to vacancy clusters and the longer the lifetime, the larger the vacancy cluster (although the relationship is not linear).
PALS measurements were performed on hydrogen-charged specimens strained by 0–10% to investigate the generation behavior of hydrogen-induced defects and at which tensile strain value the hydrogen-related defects appear. Similarly, for reference, PALS measurements were also conducted on hydrogen-free samples subjected to the same tensile strain. Figure 1 shows the results of the one-component analysis of the lifetime spectra, namely the average positron lifetime as a function of the tensile strain for the hydrogen-charged and hydrogen-free specimens. The average positron lifetime is an important figure of merit, because from a knowledge of its deviation from the known positron lifetime value of the bulk, one can obtain information on the average amount of defects. The results show that the average positron lifetime generally increases with strain in a similar manner for both the hydrogen-free samples and the hydrogen-charged samples. This is because strain-induced defects, such as dislocations, are introduced and their concentration gradually increases with strain. In the region of strains up to 3%, the average positron lifetime value for both hydrogen-free and hydrogen-charged samples remains almost constant at about 110 ps, i.e., the bulk lifetime. This is because this range of strains consists in the elastic region and the material has not undergone plastic deformation yet, so no defects are introduced. However, for strains higher than 3% the average positron lifetime value of the hydrogen-charged samples is slightly higher than that of the hydrogen-free samples. This indicates that additional defects are introduced in the hydrogen-charged samples with respect to the hydrogen-free samples. These additional defects appeared in a hydrogen environment under the application of stress, so they can be regarded as hydrogen-enhanced strain-induced defects.
A multi-component analysis of the PALS spectra was done to determine the nature of these hydrogen-enhanced strain-induced defects. The results of the two- and three-component analyses of the hydrogen-charged and hydrogen-free samples as a function of strain are presented in Fig. 2. For the hydrogen-free samples strained by 4–10%, two-component analysis was sufficient to fit the lifetime spectra of all measured samples. The first positron lifetime component is in the range 55 to 95 ps, which corresponds to the reduced bulk lifetime (due to the introduction of defects), while the second lifetime component is around 155–165 ps, reflecting the presence of strain-induced dislocations. In addition, the relative intensity of the dislocation component increases with strain, which indicates an increase in the concentration of dislocations. This is the result of increasingly larger plastic deformation due to the application of gradually higher stress.
For the hydrogen-charged specimens, the positron lifetime spectra were well fitted with two lifetime components up to a strain of 4% and with three components at strains higher than that. In the strain interval 1–4%, only lifetime components corresponding to the bulk and dislocations were detected. Similar to the hydrogen-free samples, the intensity of the dislocation component increases with strain (even beyond 4%), highlighting a rise in the density of strain-induced dislocations with increasingly larger plastic deformation. On the other hand, from a strain of 5% or more, a further positron lifetime component was detected in addition to those of the bulk and dislocations. The positron lifetime value of this component is 285 ps, which indicates a vacancy cluster with a size of about 9 atomic vacancies from comparison with theoretical calculations.15) Since this component was not detected in the hydrogen-free samples, these vacancy cluster can be considered to be hydrogen-induced defects that appear only in a hydrogen environment under stress application. In addition, the positron lifetime of the vacancy clusters hardly changes with strain, whereas the relative intensity increases. This result indicates an increase in the concentration of hydrogen-induced vacancies with rising strain, that is hydrogen promotes the formation of vacancies, which is in agreement with earlier theoretical predictions.16)
From these results we have demonstrated that hydrogen-induced vacancies are first generated when a tensile strain of about 5% is applied to the material. This is quite early in the deformation process of the material, so that we can conclude that hydrogen-induced vacancies form as soon as the plastic deformation begins in a hydrogen environment. We note that actually vacancy clusters may have been generated even at lower strains, such as 4%. This is related to the sensitivity of PALS measurements, where defects can be detected in concentrations as small as 10−7 of atomic ratios. If smaller concentrations of vacancy clusters than this value existed in the 4%-strained sample, they may not have been detected or resolved in the analysis because they were below the detection limit of the PALS measurements. As a result, the detectable formation strain of hydrogen-induced vacancies in 304 steel can be regarded to be 5%.
In light of these results and of earlier theoretical studies on the interaction of hydrogen with vacancies in iron, which reported that hydrogen binds to vacancies and stabilizes them16) and that hydrogen reduces the formation energy of vacancies,17,18) we can imagine the formation behavior of hydrogen-induced vacancies as follows. When tensile stress is applied to a hydrogen-charged material, dislocations form and monovacancies are generated, e.g., by dislocation cutting. At room temperature, monovacancies are unstable, but hydrogen atoms can be trapped at these vacancies and stabilize them, forming vacancy-hydrogen complexes. However, as further strain is applied to the material or due to thermal desorption of hydrogen at room temperature, hydrogen can be detrapped from the complexes, leaving monovacancies behind. Monovacancies can easily diffuse and disappear, e.g., at grain boundaries, or aggregate with nearby vacancies, forming vacancy clusters. This process would explain the generation of hydrogen-induced vacancies starting from the formation of vacancy-hydrogen complexes.
In our earlier study, the formation of vacancy-hydrogen complexes in hydrogen-charged austenitic stainless steel 304 strained by 10% was demonstrated.10) However, it is unclear whether these complexes form already at a strain of around 5%, i.e., when the hydrogen-induced vacancy clusters are first detected. To confirm that the vacancy-hydrogen complexes develop into vacancy clusters and to prove the vacancy formation model hypothesized above, the 10-μm-thick hydrogen-charged layer of the hydrogen-charged samples strained by 6% was electropolished and PALS measurements were carried out. In our previous work,10) it was found that the hydrogen-induced vacancy clusters are generated only in the hydrogen-charged layer, so that a similar finding is expected in the sample strained by 6% as well. Therefore, by removing that layer, it is possible to observe whether the precursor defects of the vacancy clusters, namely the vacancy-hydrogen complexes, are actually formed in the bulk of the sample below the hydrogen-charged layer. The PALS results are presented in Fig. 3 and exhibit only two lifetime components: one with a positron lifetime of about 110 ps, corresponding to the bulk, and another with a positron lifetime of 180 ps. No vacancy cluster component was detected in this sample, which confirms the absence of vacancy clusters underneath the hydrogen-charged layer. The 180 ps component has a longer positron lifetime than that of dislocations and is similar to that of monovacancies. However, monovacancies are unstable at room temperature, so that hydrogen atoms must be trapped at monovacancies and stabilize them, for a component with such a lifetime value to be detected. Hence, this component can be attributed to vacancy-hydrogen complexes, as argued in our earlier study.10) These complexes are formed in the bulk of the sample in a region deeper than the hydrogen-charged layer. This is thought to be due to effects, such as dislocation drag, in which hydrogen is transported from the hydrogen-charged layer into the bulk by binding to strain-induced dislocations. After all, it is known that hydrogen enhances the dislocation mobility.19,20) Alternatively, it is possible that hydrogen diffused into the bulk through the martensite phase, which was likely formed in the hydrogen-charged layer, due to the higher hydrogen diffusivity of the martensite phase compared to the austenite phase. In addition, from Fig. 3 we note that the relative intensity of the vacancy-hydrogen complex component is as high as 32%, i.e., higher than that of the vacancy clusters shown in Fig. 2. As this value is unusually high, it suggests that this lifetime component is actually a mixed component of vacancy-hydrogen complexes and dislocations, which could not be resolved in the lifetime analysis due to the positron lifetime of these two types of defects being too close. We note that in a sample strained by 6%, dislocations would be naturally expected to exist.
The hydrogen-charged, 6%-strained, 10-μm electropolished sample was then strained until fracture and the PALS results of the fractured sample are also reported in Fig. 3. In this specimen, two lifetime components were detected: a first component of 155 ps, corresponding to dislocations, and a second component with a lifetime of 285 ps, related to vacancy clusters. Since a vacancy-hydrogen complex component was no longer detected and vacancy clusters were formed, we have demonstrated that the vacancy clusters develop from the vacancy-hydrogen complexes by application of tensile stress in a hydrogen environment or, in other words, that the vacancy-hydrogen complexes are the precursors of the vacancy clusters.
For a complete understanding of the generation behavior of the hydrogen-induced vacancies, it is also important to know the mechanical conditions under which these defects begin to form. These conditions include not only the tensile strain, which was found to be around 5%, but also the dislocation density associated with vacancy generation. This is a crucial parameter, for example, for types of deformation other than tensile tests. There are several conventional methods for calculating the dislocation density of materials, such as electron backscattering diffraction21) and X-ray diffraction.22) Here, the dislocation density was estimated from the PALS results using simple calculations18) based on the trapping model.23) In principle, the trapping model can only be applied when the defect distribution is uniform. However, in these samples, there is a discrepancy in the defects formed in the hydrogen-charged layer and in the bulk. Although the maximum penetration depth of positrons in iron is about 130 μm, almost half of the positrons annihilate within the depth of the hydrogen-charged layer.24) Therefore, as a first approximation, we may use the PALS results to obtain an estimate of the average dislocation density. The dislocation density can be calculated from the positron lifetimes and intensities using the following equation:
| (1) |
Here, C is the defect density, τ is the positron lifetime, I is the relative intensity, and μ is the specific positron trapping rate at the defect. The subscripts f, d, and v refer to the perfect lattice, dislocations and vacancy clusters, respectively. The positron lifetime in the perfect lattice τf was set to 108 ps for pure iron. The literature value for the specific positron trapping rate μd = 5.1×10−4 m2 s−1 in pure iron was used.25) Figure 4 shows the strain dependence of the calculated dislocation density for the hydrogen-charged and hydrogen-free samples. In general, the dislocation density increases with strain, as expected due to the application of increasing tensile stress. However, the dislocation density of the hydrogen-charged samples is much higher than that of the hydrogen-free samples. This result confirms that hydrogen not only promotes the formation of vacancies but also enhances the dislocation density in the hydrogen-charged layer, possibly owing to a hydrogen-induced increase in dislocation mobility as suggested by earlier work.26) In addition, the dislocation density in the hydrogen-charged specimen strained by 5%, where hydrogen-induced vacancy clusters in the hydrogen-charged layer were first detected, is 2.95×1015 m−2. In other words, it was found that hydrogen-induced vacancy clusters start to form when the dislocation density reaches this value. This result is very important because the formation of vacancies is influenced by dislocation shear during plastic deformation and the phenomenon of localization of plastic strain, which is one of the characteristics of HE, coincides with the formation behavior of vacancy clusters due to hydrogen.
Previous studies have reported that in hydrogen-charged strained austenitic stainless steel 304 typical dislocation densities can range from 1014 to 1015 m−2, although this value can significantly increase with higher levels of strain and hydrogen content, reaching densities as high as 5×1015 m−2 due to the increased dislocation generation caused by the interaction of hydrogen with the crystal lattice.27,28) In comparison, the dislocation densities of the hydrogen-charged strained samples reported in Fig. 4 appear quite high, as they reach an order of magnitude of 1016 m−2 at higher strains. To evaluate the dislocation density using an independent experimental technique, X-ray diffraction (XRD) patterns for selected samples were measured using Cu Kα radiation at 40 kV, 40 mA. Those results are shown in Fig. 5 for a 2θ range of 40° to 55°. Profile analysis was carried out on the γ (1 1 1) peak at around 43.7° using the modified Williamson–Hall method22) and yielded dislocation densities for the hydrogen-free 6% strained sample and the hydrogen-charged 6% strained samples of 1.9×1014 m−2 and 2.3×1014 m−2, respectively. These results indicate an increase of about 20% in dislocation density due to hydrogen addition. Comparison of these values with the dislocation densities calculated from the PALS results suggests that the latter might be somewhat overestimated. This discrepancy is possibly due to the limitations in the approximations assumed in the application of the trapping model, which do not entirely reflect the actual heterogeneous defect distribution in the samples. After all, it is known that accurately estimating dislocation density can be challenging, because different characterization techniques provide varying results depending on the method used and the limitations of each technique, often leading to discrepancies in measured absolute values depending on which method is employed.29)
In this study, the hydrogen-induced vacancy formation process in austenitic stainless steel 304 was investigated by PALS. PALS measurements were carried out on hydrogen-charged samples subject to tensile testing at various strains up to 10%. Hydrogen-induced vacancy clusters were first detected at a strain of 5%. By electropolishing the hydrogen-charged layer of a hydrogen-charged 6%-strained sample, the formation of vacancy-hydrogen complexes was confirmed and these developed into vacancy clusters by further application of stress. In addition, the minimum dislocation density required for the hydrogen-induced vacancy clusters to form was found to be 2.95×1015 m−2 from the PALS results or, alternatively, 2.3×1014 m−2 from the XRD analysis. These findings are very important for understanding the complex dynamics of vacancy formation in the presence of hydrogen and provide a useful experimental reference point for dynamic simulations and models of the HE mechanisms, as well as material design.
The authors declare no conflict of interest.
Financial support from the 31st Research Promotion Grant of the Iron and Steel Institute of Japan is greatly acknowledged.
