Introduction

Hydrogen has a longstanding reputation as a damaging element in a broad range of metals and alloys, especially in terms of in-service degradation of structural components.1,2,3,4 In particular, the contribution of hydrogen to the acceleration of fatigue failure is well-documented.5,6,7,8,9,10 Several mechanisms of hydrogen-assisted cracking (HAC) of steels have been postulated, including the longstanding decohesion theory introduced by Troiano11,12 and Oriani,3,13 and the hydrogen-assisted deformation mechanism first proposed by Beachem14 and verified by Birnbaum et al.15,16,17 using in situ transmission electron microscopy (TEM) straining experiments in H2. Birnbaum and Sofronis18 further developed these concepts into the hydrogen-enhanced localized plasticity (HELP) mechanism. Alternatively, Lynch19,20 introduced the adsorption induced localized slip process (also known as adsorption induced dislocation emission), whereby adsorbed hydrogen weakens interatomic bonds facilitating dislocation nucleation at the crack tip. Dislocation injection on different crystallographic planes serves to advance the crack, while dislocation activity ahead of the crack tip can lead to voids which can then interact with the crack growing by the alternate slip mechanism to advance the crack front. More recently, another mechanism whereby hydrogen-induced deformation twins play a significant role in the acceleration of fatigue crack growth rates in the presence of hydrogen was suggested.8 The picture that emerges from the collective volume of research is that, when demonstrating and modeling the damaging effect of hydrogen, most studies only report the bulk concentration of hydrogen present in the test specimen measured by methods such as thermal desorption spectroscopy,21,22 or in the test atmosphere. Only a small volume of work is presented regarding the localized distribution of hydrogen around crack tips and other microstructural features such as grain boundaries, dislocations, twins, stacking faults, and inclusions. These include older studies with TEM replica-based tritium microautoradiography technique capable of spatial resolutions of 300 nm pioneered by Tiner and Co-workers23,24,25 and refined by Lacombe et al. at Universite Paris Sud26,27,28 and secondary ion mass spectrometry (SIMS) ion microscopy29,30 studies with lateral resolutions limited to 1 μm. Emerging more recently is work using atom probe tomography (APT), which has atomic resolution but samples a much smaller volume (~100,000 nm3 compared to greater than 1 × 109 nm3 for NanoSIMS),31,32 as well as scanning probe methods33,34 that show promise for quantification but whose spatial resolution is of the order of hundreds of microns. NanoSIMS imaging bridges the gap between these two methods, providing a better spatial resolution than tritium autoradiography (and also dispensing with the need to work with radioactive materials) while simultaneously sampling larger volumes than APT, while also offering the potential for absolute quantification of hydrogen concentration using well characterized matrix-matched standards.

Results

In this study, we directly show the distribution of hydrogen, in the form of deuterium, around primary and tertiary fatigue crack tips as well as MnS inclusions in two type 316L/316LN stainless-steel alloys. Both alloys are similar in composition (differing mainly in S and N content as shown in Table 1) and contain a microstructure consisting of large (50–150 μm) equiaxed grains; however, the 316L steel contains MnS particles due to it elevated level of S (0.012 wt%). Prior to fatigue testing, compact tension specimens (0.5 T) where precharged with deuterium using the gas phase charging method. The precharging conditions (300 °C/138 MPa for 124 days) insured that a concentration of ~140 wppm of deuterium was present (as calculated using relations).35 The measured fatigue crack growth rates, 3.84 × 10−5 mm/cycle for heat AS192 and 1.99 × 10−5 mm/cycle for heat AS198, were between 1.8× and 2.6× higher than those measured during comparable testing performed on uncharged material in ambient temperature air. The fatigue testing resulted in primarily transgranular cracking with a relatively uniform crack front. The fracture surfaces are faceted with indications of planar slip and “fissuring” which is suggestive of cracking on secondary slip planes.

Table 1 Composition (wt%) of the two type 316 stainless steel alloys used in this study
Full size table

Analysis of hydrogen in magnetic sector SIMS, though experimentally possible, is fraught with artefacts and thus difficult to interpret. The omnipresence of gaseous hydrogen-containing molecules even in the best vacuum systems invariably adsorb on the sample surface and are subsequently measured by the mass spectrometer. A large fraction of the hydrogen signal therefore originates from the vacuum system and not the sample. The best approach to alleviate this potential for artefact is to use deuterium in the experimental design. With a natural isotopic abundance of only 0.0115%, the fraction of deuterium in the ultra high vacuum analysis chamber will be negligible. A second difficulty not unique to hydrogen but to all SIMS analysis is the issue of quantification. Secondary ion yields are known to vary over several orders of magnitude depending upon the material chemistry, and true quantification requires matrix-matched standards, which are often difficult to produce. However, relative differences in concentration can be measured by normalizing the signal of interest to a signal representing the matrix. In this work, we have used oxygen as a matrix signal as it has a high secondary ion yield, is an elemental ion so issues of molecular secondary ion formation are avoided, is distributed uniformly in the steel, and is also present in the cracks which aids in their delineation, especially at the fine crack tips, from the stainless steel matrix.

The images in Fig. 1 are the deuterium and oxygen isotopic maps (Fig. 1a, b). The oxygen image clearly shows the location of the crack. The derived deuterium/oxygen (D/O) ratio image is displayed (Fig. 1c) using a hue-saturation-intensity (HSI) transformation, where the hue or color represents the actual ratio value, and the intensity is representative of the counting statistics in that same area. Thus, our eye is guided to areas of significant enrichment and away from those with poor counting statistics, which will be less statistically reliable. As a result of the high oxygen signal from the crack tip, the D/O ratio is significantly lower in the crack itself, demonstrated by the blue hue in the HSI image, and thus clearly delineates the crack from the matrix stainless steel. Several regions of enhanced D/O ratios are noted by the hot spots in the yellow–pink hue range, in the wake of the crack and near the tips of both the primary and a tertiary crack. These regions indicate a D/O ratio at least 50% higher than in the majority of other areas in the bulk of the sample. However, to completely eliminate any effect of objective biasing of the ratio image, Fig. 1d shows that D/O ratio image once again but with a set of regions of interest also drawn. These are a result of a segmentation process whereby the ratio image is analyzed for contiguous regions of a minimum area (16 pixels2 in this case, with a pixel size of 78 nm) with a mean ratio value at least 50% higher than the background level measured in areas away from the crack tip in the same image. We note several of these regions of interest (ROIs) directly adjacent the crack tip, some along the wake of the crack, and a third cluster at the tip of a tertiary crack. The areas of these ROIs range between approximately 0.12 and 0.34 μm2 in size. Comparing Fig. 1c, d shows a slight difference in where the apparent enrichments are located, and demonstrates the importance of having a robust method to extract regions of enrichment that are not biased by visual perception but given by the actual measurement values.

Fig. 1
figure 1

NanoSIMS isotopic images and derived ratio images at crack tip from sample heat AS192. a 2H image b 16O image c 2H/16O ratio displayed as HSI image. Scale bar in a–c is 5 μm. d 2H/16O ratio ratio showing locations of ROI’s generated by segmentation algorithm showing contiguous regions with a minimum area of 16 pixels2 (pixel size = 78 nm) and a mean ratio at least 50% higher than the background taken from an area away from the crack tip. Scale bar is 1 μm

Full size image

A similar set of images is presented in Fig. 2, but from another crack tip at much higher spatial resolution. The HSI D/O image in Fig. 2c again directs our attention to two larger “hot spots”, with almost twice the ratio values, adjacent and in front of the crack tip.

Fig. 2
figure 2

Higher resolution NanoSIMS isotopic images and derived ratio images at crack tip from sample heat AS198. a 2H image b 16O image c 2H/16O ratio displayed as HSI image. Scale bar is 1 μm

Full size image

Following NanoSIMS analysis, the samples were lightly re-polished to remove damage caused by the primary Cs+ ion beam. TEM samples were prepared from the crack tip region using the focused ion beam lift out technique and examined using a 200 kV accelerating voltage. The presence of a cellular structure (Fig. 3a), whose boundaries were determined to be formed by networks of dislocations (Fig. 3b) was observed. These images were acquired approximately 4–5 μm ahead of the crack tip and show highly localized regions of dislocation tangles. The region examined is in fact quite similar to the area shown in Fig. 1, and where several clusters of enriched D/O ratio values are observed. Magnifying these regions to bring them to the same magnification as the TEM images, as in Fig. 3c, the cellular appearance of the image is strikingly similar to that of the TEM image in Fig. 3a, strongly suggesting that the dislocation tangles are strong trap sites for deuterium.

Fig. 3
figure 3

Dislocation structure at crack tip from sample heat AS192. a TEM bright field image in crack wake region approximately 4–5 μm from crack. Scale bar is 500 nm. b Same region imaged at higher magnification showing dislocations clusters are responsible for forming the cellular network observed in a. Scale bar is 200 nm. c 2H/16O ratio displayed as HSI image in the crack wake region 4–5 μm from crack, with the same magnification as in (a). Note areas of high and low deuterium also display cellular-like structure with cell sizes very similar to those observed in (a). Scale bar is 500 nm

Full size image

Along with imaging deuterium in the vicinity of crack tips, we have also used this imaging capability to examine the potential penetration of hydrogen in MnS inclusions. For this set of experiments the detectors were positioned to measure 32S−, 32S1H−, and 32S2H−. Other mass interferences based on the three isotopes of sulfur or molecular ions stemming from O2 are easily separated with the instrument. Typical images are presented in Fig. 4, showing the secondary electron image and the aforementioned isotopic images and the 32S2H−/32S− HSI ratio image. The enrichment around the MnS inclusion is striking, and the fact that the 32S1H−, and 32S2H− appear so different rules out any possible mass interference effect, as the 33S1H− image would look identical to the 32S1H−, but weaker in signal owing to the much lower natural abundance of 33S−. Although similar findings have been observed using tritium autoradiography in cathodically charged samples,36,37,38 our data provide direct evidence that these inclusions trap hydrogen in gaseous environments as well. Further longer term exposure studies would be required on higher S content alloys to observe if the idea proposed by Hanninen9,10 postulating that MnS inclusions act as hydrogen trap sites in low alloy steels and their subsequent dissolution in aqueous environments (as was recently directly observed through in situ TEM/EDS examination)39 changes crack tip electrochemistry and enhances hydrogen absorption also holds in gaseous environments.

Fig. 4
figure 4

NanoSIMS isotopic images and derived ratio images from MnS inclusion in sample heat AS192 (higher sulfur content) matrix away from crack tip. a Secondary electron image b 32S− image c 32S1H− image d 32S2H− image, and e 32S2H−/32S− ratio displayed as HSI image. Scale bar is 5 μm

Full size image

Discussion

The distribution of deuterium “hot spots” in localized regions within the cyclic plastic zone, as evident in Figs. 1 and 2, are potentially significant to how hydrogen may influence the deformation process and help elucidate which mechanism of HAC is responsible for the fatigue crack growth, at least under these test conditions. Fatigue loading is known to evolve complex deformation structures with locally high densities of dislocations,40,41,42 as are similarly observed in the present study (Fig. 3). The localization of hydrogen to these deformation structures, as is strongly suggested by Fig. 3, is consistent with hydrogen-enhanced dislocation nucleation and/or localization. As described by the HELP mechanism,18 hydrogen segregated to dislocations can reduce their interaction energy with other dislocations through modification of the elastic stress field. Therefore, the ability of dislocations to overcome obstacles will be enhanced and it directly follows that dislocation mobility will be increased at lower applied stress levels. Obstacles need not only be solute, but also can be other dislocations, and so a further manifestation of this effect is a smaller dislocation–dislocation distance in regions of high dislocation density, which can affect local stress states and potentially influence crack propagation. If this were indeed the case, we would also expect to see regions of higher hydrogen concentration resulting from the enhanced transport of hydrogen by dislocations which is clearly the case in Figs. 1–3. These results and method of analysis are of particular interest in light of the modeling of hydrogen transport by dislocations recently described by Dadfarnia et al.43 Although their model does not describe a fatigue test, their predicted transport of hydrogen away from the crack tip as well as loss of pre-charged hydrogen from the crack tip site are in line with what is observed in this study. Thus, the methodology presented herein provides a means to experimentally validate these and future models, in response to one of the future research directions suggested in the recent review of advances on hydrogen embrittlement of structural materials.44

Methods

Samples for NanoSIMS analysis were prepared from the compact tension specimens using standard metallographic procedures with a final polish using an oxide polishing suspension. No etchant was used thereby avoiding sources of potential contamination. High spatial resolution SIMS images mapping the distribution of the elemental 1H−, 2H−, 12C−, 16O−, and 28Si− and the molecular 16O2H− and 32S2H− secondary ions were acquired with a Cameca NanoSIMS 50L ion microprobe employing a Cs+ primary ion beam. The beam with approximately a 100 nm spot size is rastered across the sample surface and the secondary ions produced by sputtering processes are filtered first for energy and subsequently for mass by a magnetic sector with a fixed magnetic field. Seven electron multiplier detectors are stationed at the appropriate radii to simultaneously detect and measure the secondary ions of interest from the same sputtered volume. By synchronizing the detection with the raster of the primary Cs+ beam, images are formed analogous to image acquisition in a scanning electron microscope. In fact, secondary electrons are also produced during the sputtering process and are used to produce a secondary electron image. Up to several hundred image acquisitions from the same area are acquired for each analysis, and the stack of images produced are aligned to a specific prominent feature in the image and then summed to create a single mass image with statistically relevant counts. More detailed description of the instrumentation can be found in Lechene et al.45 All data was analyzed with the open source OpenMIMS46 software plugin for ImageJ.47

Data availability

The NanoSIMS image data that support the findings are available from the corresponding author upon reasonable request.