Role of slip in hydrogen-assisted crack initiation in Ni-based alloy 725

Abstract

We conduct in situ tensile straining experiments to investigate the role of hydrogen and slip in crack initiation in nickel-based alloy 725. Our experiments reveal no tendency for hydrogen to enhance localized slip and no necessity of slip for crack initiation. We use electrochemical charging to introduce hydrogen into samples, melt extraction to measure hydrogen content, and digital image correlation to analyze localized plastic strains during in situ tensile tests in a scanning electron microscope. Cracks initiate both in regions with and without nearby localized slip. Moreover, the fraction of cracks initiating with no nearby slip is greater at higher hydrogen content. Slip-assisted crack initiation generally occurs at locations where intergranular slip is arrested, especially at intersections of slipping coherent twin boundaries with thin twin lamellae. Cracks that initiate without nearby slip occur at a wider variety of microstructural features, including inclusions, triple junctions, and surface flaws.

Increasing H content in alloy 725 reduces slip-assisted crack initiation and increases crack initiation with no nearby slip.

INTRODUCTION

UNS N07725 (alloy 725) is a high-strength, corrosion-resistant, Ni-based alloy (1). When exposed to hydrogen (H), it undergoes hydrogen embrittlement (HE) (2): It loses ductility, thus limiting its performance as a structural material. In flaw-free alloy 725 components, crack initiation is the first link in the chain of events leading to H-assisted failure. Moreover, crack propagation may occur by repeated initiation and link-up of microcracks ahead of a propagating main crack (3). Thus, understanding crack initiation is essential for reliable lifetime predictions of components operating in H-rich environments and for the development of novel, HE-resistant alloys.

Hydrogen-enhanced localized plasticity (HELP) (4–6) and hydrogen-enhanced decohesion (HEDE) (7, 8) are hypothesized mechanisms for HE in Ni-based alloys (4–6). HEDE states that H weakens atomic bonds, while HELP proposes that H enhances localized plasticity and localized plasticity promotes fracture. However, both mechanisms remain under debate. H is known to reduce cohesive strength, e.g., along grain boundaries (GBs) (9–11), but it is not clear whether this reduction is sufficient to explain the degree of experimentally observed embrittlement. While there is evidence that H enhances the intrinsic mobility of individual dislocations in Ni (12), it does not follow that plastic strains in precipitation strengthened materials such as alloy 725, thereby becoming more localized. Moreover, while localized plasticity does promote fracture in some materials [e.g., shear bands in metallic glasses (13)], evidence for its role in HE is usually indirect: It takes the form of slip lines (14) or enhanced dislocation densities (15) observed at fracture surfaces or cracks observed along slip bands (16). Moreover, in both HEDE and HELP, the focus is primarily on explaining propagation of preexisting cracks, rather than initiation of new ones.

Recent research on HE of alloy 725 has focused on understanding the role of microstructural elements and their impact on the alloy’s susceptibility to embrittlement. GBs are preferential sites for crack initiation and propagation in alloy 725 (17). Coherent twin boundaries (CTBs) are especially susceptible to crack initiation (17), with ease of dislocation slip along bands parallel to CTBs thought to be a major cause (18, 19). However, the role of localized slip in crack initiation is generally inferred from postmortem analyses (14, 16), rather than directly observed through in situ experiments. Thus, it remains unknown whether slip causes crack initiation, if initiation causes slip, if initiation and slip have a common cause, or if their colocation is merely incidental. Moreover, the role of H in slip-assisted crack initiation remains obscure.

The research reported here addresses these open questions through a specialized experiment that systematically varies H content. It observes crack initiation and propagation in situ during tensile testing in a scanning electron microscope (SEM) and quantifies surface slip via digital image correlation (DIC).

RESULTS

Hydrogen embrittlement

We investigate initially flaw-free alloy 725 specimens [with identical composition (20) and microstructure (21, 22) as in prior studies] at four different electrochemical H charging levels, ranging from none (i.e., as-prepared) to high, as detailed in the “Specimen preparation” section and Table 1. During charging, H diffuses from sample surfaces into the interior. During the 4-hour span of our in situ tensile test, H continues to diffuse, with a fraction of it outgassing from the sample. We compute the H concentration profiles (detailed in the “H distribution calculation” section) after charging and after a tensile test, as shown in Fig. 1.

Table 1. Charging conditions and outcomes.

Measured central flat thickness, charging current density and time, as-charged surface H content, and calculated H outgassing fractions of the alloy 725 samples. The uncertainty of charging current density and H content is calculated based on the standard error of the mean. N/A, not applicable.

Sample name	Central flat thickness (mm)	Charging current density (mA/cm2)	H charging duration (hours)	H content (1018/cm2)	% H outgassing after 4 hours
H-free	0.205 ± 0.005	N/A	N/A	<0.04	N/A
Low	0.245 ± 0.005	1.245 ± 0.002	50 ± 0.5	0.85 ± 0.11	44.4
Medium	0.280 ± 0.005	4.980 ± 0.007	51 ± 0.5	1.75 ± 0.28	43.7
High	0.315 ± 0.005	24.788 ± 0.051	49 ± 0.5	2.08 ± 0.05	44.6

Fig. 1. H concentration profiles across the central gauge thickness.

(A) Low-H, (B) medium-H, and (C) high-H samples under the as-charged condition (solid blue line) and after 4 hours of in situ tensile testing (red dashed line). The profiles are computed using Eq. 1, with the sum truncated after 1000 terms. h, hours.

At all charging levels, H is concentrated into ~50-μm-thick surface layers immediately after charging, with peak H contents of 1870, 3627, and 5120 atomic parts per million (a.ppm) in the low–, medium–, and high–H content samples, respectively, while the sample center is nearly H-free. This H distribution motivates us to report H content in Table 1 as the number of atoms per unit sample surface area. After in situ testing, sites of maximum H concentration are located ~15 μm beneath the free surfaces, with peak values of 1033, 2022, and 2823 a.ppm. The fraction of H that outgasses during the test is around 40%. However, the peak hydrogen concentration after outgassing remains higher than the value reported to initiate brittle cracks in Ni-based alloy 718 (23) and comparable to or higher than the reported H content in H-embrittled alloy 725 (24).

Consistent with HE, elongation to failure decreases with increasing H content (Fig. 2A). Moreover, the reduction in cross-sectional area at failure decreases from 35% for the low-H sample to 28% for the high-H sample. The small, intermittent stress drops in Fig. 2A correspond to periodic pauses in loading for the acquisition of SEM images. The plot in Fig. 2A also suggests a slight decrease in modulus and yield strength with increasing H content.

Fig. 2. Tensile test and fracture surface.

(A) Engineering stress versus elongation/gauge length plots for all four test specimens. (B) SEM micrograph of the fracture surface in the medium–H content sample. The solid white rectangle indicates dimples at the sample midpoint, while yellow dashed rectangles show near-surface regions of cleavage-like fracture.

Figure 2B shows a representative fracture surface observed on the medium–H content sample. The surface and immediate subsurface exhibit cleavage-type fracture, while the midsection shows dimples. This is a typical fracture morphology observed in prior fractography studies (24–26). It demonstrates that near-surface regions undergo H embrittlement, but the interiors do not. Plane view images of fracture surfaces in all samples are provided in the Supplementary Materials (fig. S1). The entire fracture surface of the H-free sample is dimpled (no cleavage facets). By contrast, all H-charged samples exhibit a transition from cleavage to dimples around 65 μm beneath the sample surface. The degree of H charging does not affect the location of this transition. This outcome is consistent with the calculated H distributions in Fig. 1, which show that 65 μm corresponds to the maximum depth of H penetration.

Analysis of surface slip along GBs

We observe numerous slip bands on the sample surface during in situ tensile testing. The slip bands are identified through analysis of surface strains. As detailed in the Materials and Methods, surface strains are obtained through DIC of sample surfaces with predeposited Ag particles. Figure 3 (B to E) shows maps of deviatoric surface strain ε_dev = ∣ε − tr(ε)/2∣, where ε is the two-by-two local surface strain tensor determined by DIC, for all four samples at a tensile elongation/gauge length of 0.1067. The full surface strain evolution throughout the entire loading process may be viewed in movies S1 to S4. In all cases, regardless of H content, we observe qualitatively similar strain distributions: Most of the surface strains are near or below the average deviatoric strain, ${\overline{\mathrm{\epsilon }}}_{\text{dev}}$ , while isolated bands have strains well above ${\overline{\mathrm{\epsilon }}}_{\text{dev}}$ . These long, linear bands correspond to traces of slip bands intersecting the surface.

Fig. 3. Deviatoric strains at different H contents.

(A) Histograms of ${\mathrm{\epsilon }}_{\text{dev}}/{\overline{\mathrm{\epsilon }}}_{\text{dev}}$ and maps of ε_dev (overlayed on pretest SEM micrographs) for (B) H-free ( ${\overline{\mathrm{\epsilon }}}_{\text{dev}}=0.0404$ ), (C) low-H ( ${\overline{\mathrm{\epsilon }}}_{\text{dev}}=0.0367$ ), (D) medium-H ( ${\overline{\mathrm{\epsilon }}}_{\text{dev}}=0.0492$ ), and (E) high-H ( ${\overline{\mathrm{\epsilon }}}_{\text{dev}}=0.0357$ ) samples. Dashed boxes denote crack initiation sites, white arrows point to two examples of slip bands along GBs, and pink circles illustrate two examples of slip within grain interiors. All data are from elongation/gauge length of 0.1067.

To compare the slip distributions in Fig. 3 (B to E) quantitatively, Fig. 3A plots distributions of ${\mathrm{\epsilon }}_{\text{dev}}/{\overline{\mathrm{\epsilon }}}_{\text{dev}}$ for all four samples. Uniform plasticity is expected to yield sharply peaked distributions centered on unity. Broader ones—especially with long tails at high ${\mathrm{\epsilon }}_{\text{dev}}/{\overline{\mathrm{\epsilon }}}_{\text{dev}}$ values—indicate greater localized flow. The distributions for all four samples are relatively broad, with widths spanning an order of magnitude below and above ${\overline{\mathrm{\epsilon }}}_{\text{dev}}$ , consistent with our qualitative observation that flow is localized into discrete slip bands. The sample with no H has the lowest peak height and appears to be skewed toward higher ${\mathrm{\epsilon }}_{\text{dev}}/{\overline{\mathrm{\epsilon }}}_{\text{dev}}$ than the other three. These differences suggest that, compared to the H-charged samples, the H-free sample may contain a larger number of slip bands with lower strains per band. We attribute these differences to the fact that localized slip occurs primarily within grain interiors in the H-free sample, while in H-charged samples, it occurs primarily at GBs, as detailed below. Thus, differences in the strain distribution with and without H may be due to unlike slip band behavior in grain interiors and at GBs. Differences between the distributions for the H-charged samples are not as pronounced. Moreover, there are no monotonic trends in the shapes of these distributions as a function of H content. Therefore, the degree of localized plasticity in all three H-charged samples is comparable.

Correlating slip band locations with the underlying microstructure before tensile testing (see fig. S2) reveals that, in the no-H sample, slip occurs primarily in grain interiors. Meanwhile, in all three H-charged samples, slip occurs both along GBs and within grain interiors. Slip within grain interiors takes the form of parallel bands, many of which span the entire grain width and impinge upon GBs. Slip along GBs is due to a relative displacement of adjacent grains along the GB plane (see fig. S3B for illustration). To investigate the relationship between GB character and propensity for slip, we extract the following quantities for the three H-charged samples:

1) n_GB: The total number of GBs in the field of view of each sample.

2) n_GB5: The total number of GBs whose surface traces are at least 5 μm in length. These traces are sufficiently long to perform further quantitative analysis, including determining local slip values (see Supplementary Text) and GB crystallographic character.

3) n_s: The total number of slipping GBs, defined as those whose traces are at least 5 μm in length and, at any given elongation, exhibit local deviatoric strain at least twice the average deviatoric strain.

4) n_sCTB: The total number of slipping GBs that are CTBs.

5) n_{111}: The total number of GBs (including CTBs) whose traces are at least 5 μm in length and whose surface traces are consistent with at least one of the adjacent grains having a {111}-type GB facet. Since electron backscatter diffraction (EBSD) cannot determine the inclination of GB planes with respect to the sample surface, n_{111} is an upper bound on the true number of GBs with at least one of the adjacent grains having a {111}-type GB facet. Slip that appears to occur along the plane of such GBs may actually be slip along a {111}-type plane in the interior of one of the adjacent grains.

The outcome of our analysis is shown in Table 2. Each sample contains between 100 and 200 GBs in the field of view, with 60 to 80% long enough to perform further analysis of slip and crystallographic character. Approximately 20 to 40% of these latter GBs undergo slip. Of these, the clear majority (65 to 75%) are CTBs. While this analysis yields different outcomes for each of the three samples, there is no apparent trend related to H content.

Table 2. Analysis of GBs in all three H-charged specimens after tensile testing.

Columns from left to right: sample label (by H content, see Table 1), n_GB5, n_GB, n_s, n_sCTB, n_{111}, the maximum number of slipping GBs with at least one of the adjacent grains having a {111}-type GB facet, and the minimum number of slipping GBs with no {111} facets.

Sample	nGB5/nGB	n s	n sCTB	n {111}	n{111} ∩ ns	ns – (n{111} ∩ ns)
Low	100/123	25	18	81	24	1
Medium	102/169	40	29	85	36	4
High	76/111	21	14	63	20	1

Most of slipping GBs that are not CTBs have traces consistent with at least one of the neighboring grains having a {111}-type GB facet. Thus, because {111}-type planes are dislocation glide planes in face-centered cubic (fcc) crystals, it is not possible to conclude whether the slip observed at these boundaries is due to shearing of the GB plane or to slip along a glide plane near the GB plane within one of the adjacent grains.

However, we also find several slipping non-CTBs whose traces are not consistent with either of the adjacent grains having a {111}-type GB facet. Since EBSD cannot determine the inclination of GB planes with respect to the sample surface, this number is a lower bound on the true number of slipping GBs where neither of the adjacent grains has a {111}-type GB facet. Thus, {111} facets are not essential for localized GB slip. This observation suggests that, in some cases, the slip does occur directly along the GB plane rather than within a grain nearby the GB plane.

Quantitative analysis of registered DIC and EBSD images allows us to determine slip values on individual GBs. For any point on a slipping boundary, we define the slip value as the discontinuity across the boundary plane of the displacement component tangent to the boundary, ∆d_t (see Supplementary Text and fig. S3). We perform this analysis at multiple locations along all slipping GBs for different tensile elongations. We find that slip values vary with elongation and with location along the boundary. General trends in slip behavior are as follows: (i) As global elongation increases, slip values increase; and (ii) high localized strains correlate with local slip values. Comparing slip analyses in all three H-charged samples, we find no trend in average slip behavior with total H content (see fig. S4). Thus, H does not enhance localized surface plasticity.

Analysis of crack initiation events

Samples with H concentrated near surfaces are particularly suitable for crack initiation studies because cracks readily form on the brittle surface, while the relatively more ductile central portion impedes crack propagation. Consequently, many individual crack initiation events may be observed before sample failure. As shown in Fig. 4 (B to E), multiple cracks initiate in our samples (see movies S5 to S8; white arrows point to the crack initiation events). Table S1 lists all crack initiation events in all three H-charged samples and provides the average deviatoric strain of the DIC field of view at which each event occurs. We count the number of discrete crack initiation events in each in situ tensile test until ~1.3% strain before failure and plot it as a function of H content in Fig. 4A. At no H charging, no surface cracks are initiated. As H content increases, so does the number of initiated surface cracks.

Fig. 4. Crack initiation events in all samples.

(A) Number of discrete crack initiation events as a function of H content. (B to E) SEM micrographs of the surface immediately before final failure with all crack initiation sites marked for (B) uncharged, (C) low-H, (D) medium-H, and (E) high-H samples. Dashed squares represent crack initiation events near high-slip regions, while dotted squares denote crack initiation without nearby slip.

Examination of local slip patterns in the vicinity of crack initiation sites leads us to distinguish two types of initiation events, illustrated in Fig. 5: ones that follow the onset of localized slip nearby, as shown in Fig. 5 (A to C), and those that occur with no prior nearby slip, exemplified in Fig. 5 (D to F). Table S1 identifies each crack initiation event as occurring with or without nearby slip. It also lists the inclination angle of each initiated crack with respect to the tensile axis.

Fig. 5. Crack initiation with and without nearby slip.

Two cracks in the medium-H sample: (A) SEM micrograph of crack D, (B) deviatoric strain map before crack initiation, and (C) deviatoric strain map at initiation. (D) SEM micrograph of crack I, (E) deviatoric strain map before crack initiation, and (F) deviatoric strain map at initiation. Crack D initiates with intense localized slip (yellow in the deviatoric strain color map) along a nearby CTB (indicated by the red line), while crack I initiates in a grain interior with no nearby slip: The high deviatoric strain region in (F) is due to the crack opening displacement. Bright speckles in (A) and (D) are predeposited Ag nanoparticles for DIC tracking. The gray rectangle at the top of (D) is a fiducial marker prepared by focused ion beam (FIB) milling. White arrows in (B) and (D) point to slipping boundaries. Red lines in (B), (C), (E), and (F) represent twin boundaries, and black dashed lines are nontwin boundaries.

At low H content, all cracks initiate near high-slip regions. As H content increases, so does the proportion of cracks initiating with no nearby slip. At the highest H content, most initiation events are not preceded by nearby slip. All cracks that initiate near regions of prior slip are located at GBs. Of the cracks that initiate with no prior slip, only two are not located at GBs: one in the medium-H sample and one in the high-H sample. Thus, four of six observed instances of crack initiation with no nearby slip occur at GBs, although an overwhelming majority of regions with no prior slip are within grain interiors. This finding is consistent with H-assisted crack initiation in alloy 725 being predominantly intergranular, as reported in previous investigations (25, 27, 28).

The foregoing observations suggest that, at high H content, localized slip is not necessary for crack initiation in alloy 725. To assess this inference quantitatively, we perform a P value analysis for the null hypothesis “slip does not affect a GB’s susceptibility to crack initiation.” The analysis is limited to cracks that initiate at GBs, i.e., it excludes the two observed instances of crack initiation with no prior slip in grain interiors. It uses the data in Table 2, which shows that the fraction of slipping GBs is 25% in the low-H sample, 39% for medium-H sample, and 28% for high-H sample. As detailed in the Supplementary Materials (the “Analysis of statistical significance” section), the resulting probability (P value) of observing the number of crack initiation events that we saw in our experiments, if the null hypothesis holds, is 0.001 in the low-H, 0.0006 in the medium-H, and 0.15 in the high-H samples. Following accepted practice, whereby a P value lower than 0.05 implies that the null hypothesis is falsified (29), we conclude that slip does influence a GB’s susceptibility to crack initiation in the low– and medium–H content samples. However, for the high–H content sample, the null hypothesis is not falsified, leading to the conclusion that slip plays an ever-smaller role in crack initiation as H content increases.

Cracks that initiate with no nearby slip generally have higher inclination angles relative to the tensile axis, suggesting that normal stresses play a more substantial role than shear in their initiation. As detailed in the Supplementary Materials (table S1), the fraction of initiated cracks whose inclination with respect to the tensile axis is larger than 60° increases from ~20% in the low– and medium–H content samples to 55% in the high–H content sample, indicating a tendency toward increased mode I fracture in the latter.

Crack initiation near high slip regions

Figures 5 (A to C) and 6 (B and C) are two examples of cracks that initiate in the vicinity of a slipping GB in the medium–H content sample. In initiation event D (Fig. 5, A to C), the slipping boundary is a CTB. The crack initiates at a location where the slipping CTB intersects with a 3-μm twin lamella. Similarly, in event A (Fig. 6, B and C), a crack also initiates at the intersection of a thin twin lamella with a slipping CTB.

Fig. 6. Slip at GBs.

(A) Tangential displacement discontinuity ∆d_t plotted against distance from the crack initiation location ∆x for the crack initiation site labeled A in Figs. 3D and 4D. The legend in (A) lists global elongation/gauge length. (B) Map of ε_dev around this crack initiation site at 0.0800 elongation/gauge length. The star is the crack initiation location, while the red arrow shows the orientation of the x axis of the plot in (A) [for illustration purposes, the red arrow is drawn parallel and offset from the slipping GB analyzed in (A)]. (C) Inverse pole figure along the image normal direction from EBSD showing a thin twin lamella intersecting the slipping boundary at the crack initiation location (red dashed lines indicate CTBs).

Figure 6A plots the tangential displacement discontinuity ∆d_t versus distance from the crack initiation site ∆x at several stages of tensile elongation for the crack initiation site defined in Fig. 6B. Slip generally increases at all locations along the GB with increasing tensile elongation. Examining slip as a function of distance from the initiation site, we see that slip first increases, reaches a maximum at 10 to 20 μm from the crack initiation location, and then drops. Thus, the crack does not initiate at the location of maximum slip along the boundary. Rather, it initiates where slip is arrested: in this case, at the intersection between the slipping boundary and a ~4-μm-thick twin lamella, shown in Fig. 6C.

These features are characteristic of all cracks initiating near regions of high slip in our tests: All form at locations where slip along a GB is arrested. Furthermore, over 60% of them form at intersections between a slipping boundary and a thin twin lamella, suggesting that twin lamellae are especially effective obstacles to slip. Among the twin lamellae in all three samples, 38 to 50% are thinner than 4 μm (fig. S2). Conversely, no crack initiation occurs on slipping boundaries that do not encounter an obstacle that arrests slip. Crack initiation at locations of slip arrest occurs even without H in some metals (30) and is attributed to the generation of stress concentrations at these locations (31). Thus, the likely role of slip in our tests is to create these stress concentrations, and the likely role of H is to reduce cohesive strength sufficiently for decohesion to occur at these locations.

Crack initiation without nearby slip

Figure 5 (D to F) shows crack initiation event I from the medium-H sample, which occurred without prior slip (Fig. 5E). This crack is initiated far from any GB traces on the surface facet of an individual grain and around 4 μm below a focused ion beam (FIB) marker. Using x-ray computed tomography (fig. S5), we observed that this crack is not branched and there are no additional disconnected subsurface cracks or cuboidal carbide inclusions (20) in its vicinity. At first, the microcrack has the shape of a zigzagging polyline, with the average polyline oriented 90° to the tensile axis. Localized strain is generated around the crack after initiation (Fig. 5F). Crack initiation events without nearby slip are only observed in the medium– and high–H content specimens. In general, these events occur at locations where there is a surface flaw or impurity, such as FIB marks (crack I in Fig. 3D), inclusions (crack A in Fig. 3E), or impurity particles (crack I in Fig. 3E).

Crack propagation

Once cracks initiate, they may subsequently propagate along nearby bands of prior slip. Figure 7 shows this crack propagation for the initiation event in Fig. 6. Figure 8 illustrates the propagation of the crack whose initiation is shown in Fig. 5 (A to C), and this crack propagates along a low-angle GB (LAGB) formed during tensile staining. This observation aligns with the findings of Xie et al. (32), who reported that LAGBs formed under a H atmosphere are susceptible to cracking. Cracks may propagate along slipping CTBs (as in Fig. 6), but we did not observe any cases of propagation along the boundaries of thin twin lamellae that intersect slipping GBs.

Fig. 7. Crack propagation along nearby bands of prior slip.

SEM micrograph of the location shown in Fig. 6 (B and C) (crack A of medium-H sample, see Fig. 3D). Initiation at (A) 0.0933 elongation/gauge length and propagation at (B) 0.1333, (C) 0.1867, and (D) 0.2267 elongation/gauge length. The star indicates the crack initiation location. The red arrow shows the direction of crack propagation along the slipping GB.

Fig. 8. Crack propagation along an LAGB.

(A) SEM micrograph of the crack shown in Fig. 6 (A to C) (crack D of medium-H sample) after propagation along a nearby LAGB; (B) corresponding map of grain orientations. The star indicates the crack initiation location. The white arrow shows the direction of crack propagation along an LAGB.

DISCUSSION

Using a combination of in situ mechanical testing, DIC, and melt extraction, we have shown how surface crack initiation varies with H content in initially flaw-free alloy 725. Our findings are based on analysis of multiple crack initiation events within a field of view containing more than 100 GBs. Thus, each sample provides adequate statistics for analysis of trends. In H-free (uncharged) samples, we do not observe any crack initiation events. As H content increases, so does the number of crack initiation events per unit sample surface area. At lower H content, all observed events occur in the vicinity of bands of localized intergranular slip. However, at higher H content, some cracks initiate without any nearby localized slip. The fraction of crack initiation events without nearby slip increases with H content. Thus, while localized slip is essential for crack initiation at lower H concentration, it is not essential at higher concentration.

In the case of slip-assisted initiation (i.e., cracks initiating near bands of localized slip), the initiation site is not located at the position of maximum slip but rather where the slip is blocked by an obstacle. In our study, the locations most vulnerable to slip-assisted crack initiation (i.e., ones where slip-assisted cracks initiate most frequently) are intersections of intergranular slip bands with ~4-μm-thick twin lamellae. We conclude that these lamellae are especially potent obstacles to slip. This view agrees with the micropillar compression experiments conducted by Liebig et al. (33) in copper and α-brass. They found that CTBs block slip transmission, leading to dislocation pileups and strain hardening.

Intergranular crack initiation at locations where localized plasticity is arrested has previously been observed in irradiated steels (34) and γ-TiAl (35). In these cases, the expected role of slip is to increase hydrostatic stress and strain energy density at the location of slip arrest (36), eventually causing bond breaking to occur there. We surmise that such a mechanism is also responsible for slip-assisted crack initiation in our experiments. This mechanism contrasts with the one reported in HE UNS N07718 (alloy 718) by Zhang et al. (16, 37). They observe crack initiation by void growth and coalescence at high-slip locations (as opposed to locations of slip arrest) within slip bands in grain interiors. The difference between slip initiation mechanisms in alloys 718 and 725 reinforces the view that HE does not occur by the same mechanism in all materials (38).

Our interpretation of slip-assisted crack initiation in alloy 725 may at first appear inconsistent with the findings of Lu and Wang (39), who performed compression tests on H-charged micropillars of alloy 725, including single-crystal and bicrystal specimens (the latter contain GBs). Their investigation revealed localized slip bands and cases of slip arrest at GBs. However, none of their tests resulted in crack initiation. We attribute this outcome primarily to the micropillar dimensions (diameters < 5 μm), which are too small to permit slip discontinuities comparable to those seen in our experiments: ~0.5 μm at the point of crack initiation, as shown in Fig. 6A. By contrast, the slip steps observed by Lu and Wang (39) appear to be on the level of tens of nanometers. Nanometer-scale slip steps are apparently insufficient to generate the stresses and strain energies required for crack initiation at the location of slip arrest.

Pristine, defect-free CTBs have low GB energy, high work of separation, and low H solubility (40). However, localized slip along CTBs may deposit a high density of dislocation debris along the boundary (41), enhancing H solubility, reducing cohesion, and promoting the extension of cracks along slip bands coincident with CTBs. In ~75% of slip-assisted crack initiation events in our study, the slip that led to crack initiation occurred along a CTB. While the location of crack initiation is at the point of slip arrest, once a crack is initiated, it may extend along the slip band responsible for its initiation. This conclusion agrees with the work of Seita et al. (17), who reported that CTBs are preferential crack initiation sites in HE alloy 725.

When discussing the preferential sites for crack initiation, Harris et al. (42) distinguish two types of preference: frequency preference (percentage of cracks initiated at a particular location at the end of a mechanical test) and strain preference (locations where cracks initiate earliest in a mechanical test, i.e., at the smallest overall strain). The postmortem analysis of Seita et al. (17) demonstrates a frequency preference for cracks to initiate at CTBs. In their work, Harris et al. (42) questioned whether there is also a strain preference for crack initiation at CTBs. The data provided in table S1 confirm that there is also a strain preference for crack initiation at CTBs: Intergranular microcracks initiate at lower strains adjacent to slipping CTBs than any other microstructural features.

In addition to elucidating the role of slip in crack initiation, our observations also yield insight concerning the role of H in alloy 725. One likely effect of H is to weaken cohesion, especially at defects and flaws. This role is especially apparent in the case of cracks initiating without nearby localized slip. The preferred locations for these initiation events are surface flaws, including elements of the material microstructure, such as inclusions and GBs, as well as extrinsic features generated during sample preparation, e.g., surface impurity particles or FIB markers. These flaws may have reduced cohesion due to preferential H trapping (43). At low H content, blocked slip is needed to generate stress concentrations high enough to initiate cracks, even at flaws. However, at high H content, cohesion is reduced enough to initiate cracks at lower local stress levels.

Another role of H concerns its influence on slip. We find no tendency for H to enhance localized slip. Localization of flow into discrete slip bands is common in precipitation-strengthened Ni-based alloys, even in the absence of H (44). In our experiments, introducing H neither increases the displacement discontinuity across slip bands nor reduces the number of active bands. Instead, the primary influence of H appears to be to shift the location of localized slip from grain interiors to GBs. We find that most slipping GBs in H-charged samples have at least one {111} facet, suggesting that slip typically does not occur directly on GB planes but rather near GBs along dislocation glide planes within grain interiors.

The findings of Zhang et al. (19) provide a partial explanation for H-assisted flow localization along GBs in precipitation-strengthened alloy 725. They report that CTBs in this material are flanked by zones denuded of γ″ precipitates and that these zones are channels for easy dislocation glide. This observation explains the preference for flow localization along CTBs in H-charged samples. By extension, our work further indicates that, in alloy 725, other GBs with {111} facets (besides CTBs) may also feature precipitate-denuded zones, accounting for their ease of slip. γ″ precipitates cannot be resolved with SEM-DIC, so their actual role in slip localization at non-CTBs requires further characterization by transmission electron microscopy.

However, these considerations do not explain why slip localization along GBs requires H. In the absence of H, we hypothesize that incipient slip bands within precipitate-denuded zones harden rapidly by the mechanism described by Argon and East (45). In this mechanism, segments of opposite sign screw dislocations gliding along parallel planes within a slip band cross-slip and annihilate, leaving behind forests of edge segments. These edge segments are obstacles to further glide, resulting in hardening. Hardening eventually elevates the flow resistance within GB slip bands beyond that of bands within grain interiors, thereby favoring further slip along the latter while the former inactivates.

In this context, one possible way that H may promote localized slip at GBs rather than within grain interiors is to reduce hardening within GB slip bands by impeding cross-slip. In fcc metals, screw dislocations dissociate into Shockley partials separated by a stacking fault (46). The Shockley partials cannot cross-slip because they have mixed (edge/screw) character. Cross-slip therefore requires a local constriction of the faulted area, allowing the Shockley partials to recombine into a screw dislocation (47). In Ni, H reduces stacking fault energy (48–50), thereby increasing the equilibrium distance between Shockley partials and hindering the constrictions required for cross-slip. With reduced hardening due to H, GB slip bands retain lower flow resistance than slip bands within grain interiors and consequently remain active to higher strains. However, the viability of this explanation ultimately depends on whether the concentration of H segregated to dislocations near GBs is large enough to elevate the stacking fault energy notably. Solute drag exerted on dislocations by H may also affect the relative hardening tendency of slip bands near GBs and in grain interiors (51).

Theories of HE, such as HELP (4–6) and HEDE (7, 8), are primarily concerned with crack propagation, not initiation. Thus, an investigation of crack initiation cannot discriminate between them. Nevertheless, it may shed light on the assumptions underpinning these theories. In particular, our observations that H does not enhance localized plasticity and that cracks do not initiate at locations of maximum localized slip are at odds with the assumptions underlying HELP. While our work suggests that reduced cohesion likely plays a role in crack initiation, quantitative comparisons between local stresses and cohesive strengths are not yet available.

Classical theories of fracture examine the conditions under which preexisting cracks extend (52). The role of crack initiation in failure of initially crack-free materials is comparatively less well understood. Crack initiation is an essential step in theories of failure via H-induced fast fracture (53). In some materials, crack propagation is thought to proceed through repeated crack initiation and coalescence (3, 54). On the other hand, copious initiation of new cracks can also have a shielding effect, hindering the propagation of preexisting cracks (55). Further research is needed to elucidate the connection between crack initiation and failure.

Last, our tensile tests suggest that H may reduce Young’s modulus in alloy 725. These observations are in qualitative agreement with studies on pure Ni by Müller et al. (56) and Lawrence et al. (57), both of which reported H-induced stiffness reductions. Elucidating the effect of H on modulus requires further investigation.

MATERIALS AND METHODS

Specimen preparation

A forged bar of precipitation strengthened alloy 725 from Special Metals was provided by the Centro Sviluppo Materiali. Samples taken from the same bar were used in several prior investigations (22, 28). Some heats of alloy 725 are known to form the brittle, intergranular F phase (24, 27). However, no F phase was detected in the material used for this study. From this bar, we fabricate single-gauge test specimens by electrical discharging machining (EDM) using a Mitsubishi MD + PRO III wire EDM tool. A rectangular section on one end of each specimen serves as a connection to the electrochemical cell, and it is cut off before tensile testing.

For H content measurements, all surfaces are polished with 9-, 3-, and 1-μm diamond particles and finished with 0.05-μm colloidal silica to remove EDM recast layers. For tensile testing, we introduce a second gauge on a 1.5-mm by 1.5-mm central flat of the as-machined specimens, as shown in fig. S6. The flat is ground down to a thickness of ~350 μm, followed by polishing with 9-, 3-, and 1-μm diamond particles, and finished with 0.05-μm colloidal silica vibratory polishing, resulting in a final thickness of 200 to 300 μm. The exact final thickness of the central flat in each specimen is listed in Table 1. Because this flat has the lowest cross-sectional area perpendicular to the tensile axis, all plastic deformation during loading occurs within it, while the remainder of the specimen is below the elastic limit. Our surface preparation procedure completely removes the EDM recast layer on the observed side of the flat and yields a mirror finish suitable for EBSD characterization.

DIC pattern deposition

We obtain SEM-DIC trackable surface patterns via the nanofilm remodeling method of Di Gioacchino and Quinta da Fonseca (58). We deposit 30-nm-thick Ag films on the polished central flat of our tensile test specimens using a Lesker PVD 75 DC sputter deposition system. The specimens are held for ~1.5 hours at ~120°C on a hot plate while exposed to vapor from a 0.25 M NaCl solution, causing the Ag nanofilm to remodel into 230 ± 110–nm-diameter islands. For ease of identification during subsequent in situ tensile testing, regions with the highest-quality surface patterns are marked with two square-shaped fiducial markers and two numbers: 1 and 2. These numbers are used for aligning and correlating SEM micrographs, DIC maps, and EBSD images. The markers and numbers are 100-μm distant from each other. Each one has dimensions of ~7 μm by 7 μm. The markers and numbers are created by FIB deposition in a Tescan LYRA3 FIB-SEM. The crystal orientation of the polished surface with Ag particles is EBSD detectable.

Microstructure characterization

EBSD was performed using a Tescan FERA-3 SEM and Oxford Nordlys EBSD detector with an accelerating voltage of 20 kV and an acquisition speed of 40 Hz. Grain orientations and GB characters were analyzed using the Aztec and channel 5 software by Oxford Instruments. We performed EBSD twice: first, after Ag deposition before the in situ tensile testing and, next, after the tensile test. Following Seita et al. (18), we identified CTBs as Σ3 boundaries whose surface traces have low curvature and orientation consistent with {111}-type boundary facets in both grains.

Electrochemical charging

H is introduced into test specimens by cathodic electrochemical charging in a three-electrode Gamry multiport cell following the procedure detailed in (59). We use a Pt wire counterelectrode (anode) and a saturated calomel reference electrode. The electrolyte is 1 M H₂SO₄ with NaCl (1 g/liter) added to improve conductivity. We use a VersaSTAT3F galvanostat/potentiostat from AMETEK to control and monitor current and overpotential.

We investigate samples at four different charging conditions: one in an as-prepared (uncharged) state and three charged for ~50 hours at different current densities: 1.5, 5, or 25 mA/cm². The exact charging parameters for each specimen are listed in Table 1. After removal from the electrolyte, samples were gently dried using Kimtech wipes, rinsed in deionized water, and ultrasonicated in acetone. Since both H content measurement and tensile testing are destructive, two specimens were charged at each condition: one for each test type.

H content measurement

Total H content is measured using a Bruker G8 Galileo melt extraction instrument. As discussed in the “H distribution calculation” section, the H in our samples is not uniformly distributed but concentrated in thin layers along the surfaces. Therefore, Table 1 reports the H content as the number of H atoms per unit surface area. Samples for H testing were cut into three pieces along the length of the gauge, and the H content in each piece was determined separately. The variance in H content among these three pieces is used as an estimate of the H content uncertainty for each charging condition.

To determine the effect of the deposited Ag surface pattern on H uptake, we compared the H concentration in a sample charged at 25 mA/cm² (the high H case in Table 1) with and without a Ag surface pattern. The H content in the sample with a Ag surface pattern is ~4% lower than in a sample with no pattern. This difference in H uptake is not notable enough to affect the conclusions of our study. We also compared H-assisted crack formation on samples with and without Ag particles (fig. S7) and found no qualitative differences.

H distribution calculation

During charging, H diffuses from sample surfaces into the interior. The charging time (t_charge) is insufficient to achieve a uniform H distribution. The final internal H distribution may be calculated using the analytical solution for one-dimensional Fickian diffusion into a symmetrical slab (60)

$$C(x,t)=C_0\frac{4}{\mathrm{\pi }}{\displaystyle {\sum }_{n=0}^{\infty }}\frac{1}{2n+1}\text{sin}\left[(2n+1)\mathrm{\pi }\frac{x}{L}\right]e^{[-\frac{(2n+1){\mathrm{\pi }}^2\mathit{Dt}}{L^2}]}$$ (1)

The slab has a thickness of L and an initially uniform H concentration of zero. At time t = 0, a constant H concentration of C₀ is applied to the free surfaces at x = 0 and x = L. The solution provides C as a function of depth x within the slab (measured from one of the surfaces) for all t > 0. D is H diffusivity, and we use D = 5.8 μm²/hour, the room-temperature value reported for alloy 725 (24).

To obtain the value of C₀, we compute the H content per unit surface area after charging time t_charge

$$n_{\mathrm{S}}=\frac{1}{2}{\displaystyle {\int }_0^L}C(x,t_{\text{charge}})\mathit{dx}={\mathit{LC}}_0{\displaystyle {\sum }_{n=0}^{\infty }}\frac{4e^{[-\frac{(2n+1){\mathrm{\pi }}^{2}D{t}_{\text{charge}}}{L^2}]}}{{(2n\mathrm{\pi }+\mathrm{\pi })}^2}$$ (2)

As described in the preceding subsection, n_S is determined experimentally using the melt extraction method and reported in Table 1. Thus, using these values, we calculate C₀ for all three charging conditions as

$$C_0=\frac{n_{\mathrm{S}}}{L{\displaystyle {\sum }_{n=0}^{\infty }}\frac{4e^{[-\frac{(2n+1){\mathrm{\pi }}^{2}D{t}_{\text{charge}}}{L^2}]}}{{(2n\mathrm{\pi }+\mathrm{\pi })}^2}}$$ (3)

In our calculation, the sum in the denominator of Eq. 3 is truncated after 1000 terms.

Knowing C₀, we compute the H concentration profiles immediately after charging, as shown in Fig. 1. During the 4-hour span of our in situ tensile tests, H continues to diffuse through the sample while also outgassing. Since Fick’s diffusion equation is linear, we compute the internal H concentration at the end of testing using superposition

$$C_{\text{test}}(x,4 \text{hours})=C(x,t_{\text{charge}}+4 \text{hours})-C(x,4 \text{hours})$$ (4)

where C(x, t) is the analytical solution given in Eq. 1. The two superimposed functions on the right-hand side of Eq. 4 use the same value of C₀. Thus, the superposition predicts a surface concentration of zero during in situ testing, as required. The computed H concentration profiles at the end of the tensile test are shown in Fig. 1. By comparing the total calculated H content at the beginning and end of the tensile tests, we find that the exact fraction of H that outgasses during the test for each sample and list in Table 1.

In situ tensile testing and SEM image acquisition

Immediately after charging, we conduct in situ tensile tests in a Tescan FERA-3 FIB-SEM using a Kammrath Weiss 500 tensile/compression module. Tests are conducted at room temperature in a vacuum chamber (1.05 × 10⁻³ to 1.8 × 10⁻³ Pa) with an average strain rate of 4 × 10⁻⁵/s over times of 4 hours or less. During this time, ~44% of the charged H is lost by outgassing from the sample surface (see Table 1 and the “Crack propagation” section). Straining was halted at ~1.3% increments of uniaxial tensile strain to acquire high-resolution SEM images for DIC. After locating the sample area marked with fiducial markers, we use a secondary electron detector to capture a set of four images, each with a 120-μm by 120-μm field of view. These images are offset 100 μm from each other and later stitched together in Photoshop, yielding a ~220-μm by 220-μm full field of view. We use a working distance of 25 mm (the closest achievable with our tensile stage), a dwell time of 10 μs per pixel, a beam voltage of 20 kV, and a beam current of ~1.6 nA. We acquire images at 16 bits per pixel with an image size no smaller than 1280 by 1280 pixels. A total time of ~4 min is required to complete image acquisition after each 1.3% strain increment.

Image analysis

We use Ncorr, a MATLAB-based code and GUI, for DIC analysis (61). On the basis of our previous work (62), the optimal subset size is 29 pixels in diameter, and the optimal strain window is 2 by 2 pixels. This setting guarantees that for any region with dimensions larger than 4 μm, displacements and strains may be captured with no more than 20% deviation from their true values (63). The setting maintains low SEM image correlation noise without overly smoothing the data.

To find the microstructure features that favor crack initiation and generate localized slip, we use MATLAB to register EBSD images taken before tensile testing with DIC surface strain maps expressed in the reference (undeformed) configuration. For example, fig. S8 presents the registered EBSD micrograph and strain map (at a total elongation of 140 μm) for the medium–H content sample. Ncorr calculates strains in the reference configuration using the Green-Lagrange method.

We correct EBSD micrographs for trapezoidal distortion by applying a geometrical transformation that matches four or more selected points in the strain map. These points include prefabricated fiducial markers, individual surface particles, and crack tips. To ensure high matching confidence, we confirm that the selected points are also coincident in the final (maximally deformed) state of the sample. To this end, we compare the EBSD micrographs of each sample after testing to the final strain maps expressed in the current (deformed) configuration (Ncorr calculates these with the Eulerian method). Thus, we correlate slip bands from the DIC strain map with the underlying microstructure obtained by EBSD before tensile testing.

X-ray computed tomography

A high-resolution computed tomography scan was obtained using a nanoVoxel-3000 at Sanying Precision Instruments Co. Ltd., China. Scanning was carried out using x-ray transmission nanofocus imaging with 60-kV transmission voltage, 30-μA current, and 4-hour duration. The sample was rotated 360° to acquire 1440 projections, recorded as 2940 by 2304 pixel, 16-bit images, which were used to reconstruct a 3D rendering with 0.55-μm voxel dimensions. In a 16-bit image, grayscale values range from 0 to 65,536, with the background (air) corresponding to grayscale values near 0, while the alloy has grayscale values around 45,500. Regions with lower x-ray attenuation than the alloy have lower grayscale values. We identify regions containing more than nine connected voxels with grayscale value lower than 23,000 as internal defects (cracks, voids, and inclusions). They are exported to MATLAB for visualization. Alloy 725 contains cuboidal carbide inclusions with an edge length of order of 10 μm (20).

Supplementary Materials

This PDF file includes:

Supplementary Text

Figs. S1 to S8

Table S1

Legends for movies S1 to S8

Other Supplementary Material for this manuscript includes the following:

Movies S1 to S8