Stress corrosion of copper in sulfide solutions: Variations in pH‐buffer, strain rate, and temperature

Copper is the intended canister material for the disposal of spent nuclear fuel in Sweden. At repository depth the groundwater may contain dissolved sulfide. The main goal for this work is to study the tendency for stress corrosion of copper in sulfide solutions and examine the influence of various experimental parameters on stress corrosion. Slow strain rate testing was performed on copper test rods in solutions with 1.0 mM sulfide. The pH was kept near neutral with phosphate or borate buffer. The test matrix included variations in temperature, strain rate, and duration of the tests as well as salt and buffer concentrations. Cross‐sections of the specimens after testing were investigated using scanning electron microscope/energy dispersive X‐ray spectroscopy detector. Stress–strain curves do not reveal any signs of stress corrosion. However, intergranular corrosion in the shape of crack and pit‐like features developed in all tests with 1.0 mM sulfide. The length of the deepest features in all these tests was of the same order of magnitude (10–20 µm). The suggested mechanism proposes that crack‐like features originate at the surface of the copper metal from the oxidation of grain boundaries that behave as slightly less noble.


| INTRODUCTION
Copper is one of the intended materials from the multibarrier principle as the final repository for spent nuclear fuel in Sweden. [1]Cylindrical canisters manufactured from copper will encapsulate the spent fuel and will be placed in granite bedrock, enclosed by bentonite clay.The canister consists of a 5 cm copper shell as corrosion barrier and a cast iron insert to withstand the mechanical loads. [1,2]In the postclosure safety assessment for the repository [1,3] one mode of potential failure of a copper canister is stress corrosion cracking (SCC).
The main goal of this work is to study the influence of various test parameters on the tendency for SCC of copper in sulfide solutions.The work is mainly concerned about the copper quality intended for use by the Swedish Nuclear Fuel and Waste Management Co (SKB), referred to as SKB-copper.This material can be described as phosphorus-doped (30-100 ppm), oxygenfree copper (Cu-OFP). [2]he investigation was performed in RISE, Kista, Sweden, from 2018 through 2022 on the stress corrosion of copper in neutral sulfide solutions, with the results reported in SKB technical reports. [3,4]The experimental results from previous work, by us and by others, on slow strain rate test (SSRT) testing of SKB-copper are summarized in earlier work. [5]Figure 1 illustrates the results with respect to pH and sulfide concentration.The upper diagram illustrates the data points studied at room temperature and the lower diagram illustrates data points studied at higher temperatures.Open symbols were used to show that no signs indicating SCC were found, as interpreted by the respective authors.Symbols filled with red indicate that signs suggesting SSC were indeed found.
The work of Taniguchi and Kawasaki, [8] which in a way was the starting point of the issue of stress corrosion cracking in sulfide, was not included in Figure 1.They studied copper with a different composition, in terms of impurities and dopants, and found signs of SCC at conditions corresponding to those studied by Taxén et al. [5] and by Bhaskaran et al. [6] at 80°C.The different experimental details of these studies were compared and discussed previously. [5]he diagrams represent the experimental conditions only in terms of two levels of temperature and scales for pH and sulfide concentration.The composition of the background electrolytes has also varied between the studies.All researchers added sulfide in the form of Na 2 S to the test solutions.The natural pH of these solutions tends to be high.Bhaskaran et al. partially neutralized the solution in some experiments at room temperature. [6]ecker and Öijerholm added a phosphate buffer to about pH 7.2. [7]ecent tests for stress corrosion have been performed elsewhere, with constant strain rates [9,10] and by constant load. [11]Superficial intergranular corrosion was observed after both types of tests but no apparent degradation of the mechanical properties of copper was found.After a test at constant load, crack-like surface defects (a few tens of μm in length) were observed by scanning electron microscope (SEM) in surfaces of specimens tested in air, in water without sulfide, and in sulfide containing environments.However, general corrosion in the sulfide solution tended to decrease the depth of the initial cracklike surface defects. [11]tomic hydrogen in copper, from corrosion in sulfide, as a possible cause for embrittlement or cracking of copper has been suggested or implied. [12]Corrosion of copper in sulfide was shown to increase the concentration of hydrogen in copper, [9,10] and stress and strain has been shown to increase the hydrogen uptake compared to unloaded specimen. [10]

| CHEMICAL BACKGROUND
The relevant solutions contain chloride as well as sulfide.Stability diagrams for solid and dissolved species may help to understand the effect of pH and for the interpretation of the results.Figures 2 and 3 show stability diagrams for copper in 1 mM sulfide with 100 mM chloride at 25°C and 90°C, respectively.The diagrams were calculated from thermodynamic data in Puigdomenech and Taxén. [13]Equilibrium potentials from Nernst equation were calculated for each of the substances assuming that metallic copper is present, and that sulfide is present as S(-II), although this oxidation state is not the most stable form of sulfur in all ranges of potential and pH.Complex constants for Cu(HS) 2 − and CuHS(aq) were taken from Mountain and Seward, [14,15] as those publications contain data from a later date, with  [6] Taxén et al, [5] Becker and Öijerholm, [7] and this work. [4]Color figure can be viewed at wileyonlinelibrary.com] temperature dependencies reported there.Table 1 shows the stability constants for the dominating complexes that the diagrams are based on.The solid black line in Figures 2 and 3 shows the maximum potential that can arise at a hydrogen pressure of 1.0 atm if there are no oxidants other than water.The 1.0 atm pressure and the location of the black line are not absolute upper limits for the potential that hydrogen evolution can generate, but it is a useful reference limit.At lower levels of dissolved H 2 slightly higher potentials can arise, and the dashed black line corresponds to a partial pressure of 0. , respectively at concentrations of 1.0 µM.These concentrations are not in equilibrium with the solid Cu 2 S(s) nor with one another.If they were, the lines would coincide with the line for Cu(s)/Cu 2 S(s).
The selected concentrations of Cu(HS) 2 − and CuHS(aq) require higher potential to form than the solid copper sulfide.The hypothetical solution with these concentrations would be supersaturated with respect to Cu 2 S(s).
If we focus on Figure 3 at 90°C, the curves for the dissolved copper sulfides require higher potentials to form at the selected concentration, than hydrogen evolution to 1.0 atm can give at high pH, since water is a too weak oxidant.However, at values below pH 8, hydrogen-evolving corrosion can create the selected concentration.It would not be the lowest energy state, which would be equilibrium with Cu 2 S(s), but it would be a downhill process.Nevertheless, if the solid Cu 2 S(s) was formed first, dissolution cannot generate concentrations of the selected level of 1.0 µM.That would be an uphill process.
The black circle at pH 7.4 represents a result from the present work at 90°C. [3]The corrosion potential, about -0.609 V E H (−0.850 V vs. SCE + 0.2412 V [16] ) is located below all the lines for dissolved concentrations of 1.0 µM.In the present experiments at about pH 7.2-7.4,dissolved concentrations slightly lower than 1.0 µM can arise from corrosion of copper with its usual properties.The measured corrosion potential is well below the solid black line for hydrogen evolution at 1.0 atm, leaving a significant margin for activation potential to actually generate hydrogen.At pH 12 the solid black line intersects the solid red line, representing the stability limit for Cu(s)/Cu 2 S(s) in Figure 3.While Cu(s)/Cu 2 S(s) may coexist with H 2 at 1.0 atm, there would be no or very small margin for the necessary activation potential to actually generate hydrogen.At potentials significantly higher than the intersection point at pH 12, only small amounts of hydrogen can form until local equilibrium limits production.At a potential significantly lower than the intersection point at pH 12, only very small amounts of dissolved copper sulfide and no solid Cu 2 S can form.The potential at the intersection point seems to be a reasonable prediction of the corrosion potential in 1 mM sulfide at pH 12 and 90°C.The predicted corrosion potential is about 200 mV lower than the equilibrium potential for a 1.0 µM concentration of Cu 2 S(HS) 2 2− , which is the most stable form of dissolved copper at pH 12. Very low concentrations of dissolved copper and no solid corrosion products can form at pH-values higher than pH 12 at 1.0 atm H 2 .
The decrease in oxidizing property of water at high pH is a main factor.At a neutral pH there is sufficient activation potential that can produce Cu 2 S as well as dissolved copper in the form of sulfide complexes, but not at alkaline pH.
At relatively high potential there are one or three lines in blue.These lines represent the potentials required to form the dissolved copper chloride complexes CuCl(aq), CuCl 2 − , and CuCl 3 2− at concentrations of 1.0 µM.The lines are located at potentials much higher than the black line.The lines for CuCl(aq) and CuCl 3 2− are located at a potential higher than 0 V at 25°C and falls outside the scale in Figure 2. Hydrogen evolution to 1.0 atm cannot produce concentrations of 1.0 µM for any of the chloride species, at a chloride concentration of 0.1 M. The cluster compound Cu 3 S 3 has also been identified in a corroding system, [17] but does not appear in the equilibrium diagrams as it contains sulfur in the formal oxidation state S(-I).

| EXPERIMENTAL
The choice of experimental conditions studied in the present work was based on the observations from the literature and from the assessment of the chemical systems.The results in Figure 1 imply that the pH of the solution may be an important factor, besides the sulfide concentration, for the appearance of signs that can be interpreted as SCC.The presence of phosphate in the test solution may potentially also have other effects than to act as a pH-buffer.
Test rods were cut from oxygen-free phosphorousdoped copper (Cu-OFP) supplied by SKB, by spark cutting followed by turning into shape.After that, the test rods were fully annealed with molten salt treatment at 600°C for 10 min and quenching in water for 60 s.
F I G U R E 3 Stability diagram for copper in 1 mM sulfide and 100 mM chloride at 90°C.The dissolved copper-containing species are shown at a concentration of 1 µM.The black circle at pH 7.4 represents a result from the present work at 90°C as reported in the results section. [3][Color figure can be viewed at wileyonlinelibrary.com]Traces of salt were removed using sandpaper (grit 320) on a lathe at slow speed.The section exposed to the solution was further polished up to 600 grit paper.Finally, the rods were degreased using ethanol and rinsed in deionized water immediately before mounting in the SSRT cell.
After exposure the rods were examined using a light optical microscope (LOM) and SEM brand Zeiss Gemini 450 equipped with an energy dispersive X-ray spectroscopy detector (EDS).Cross sections were prepared using common metallographic techniques and examined via SEM/EDS.Crack or pit-like features were recorded in numbers on one side of the longitudinal cross-section.A representative sample of the recorded features was further analyzed using the image software ImageJ [18] to estimate their length.The length was measured from a subjective baseline at the interface between the corrosion products and the metallic copper, following the shape of the crack or pit in a straight line up to the tip.
The experimental setup for this set of experiments is explained in detail elsewhere. [5]In that previous study, the sulfide stock solution was fed through the SSRT-cell continuously.In the current study, [3,4] two stock solutions were fed using a tube pump.In the SSRT-cell, the pH 7.2 phosphate or borate buffer solution was mixed with the alkaline sulfide stock solution.Both solutions were fed at the same rate.Figure 4 shows a collection of test rods before testing.The length of the narrow section is 20 mm and the smallest diameter is 7 mm.
Figure 5 shows a schematic of the experimental setup.
Table 2 shows the detailed conditions of the exposures.Two different sets of experiments with comparable conditions were carried out in the years 2018 [3] and 2021. [4]The matrix was designed to study the role of different parameters (e.g., temperature, sulfide concentration, chloride concentration, and strain rate).All the solutions contained phosphate buffer, 10 or 1.0 mM, with the exception of Run 2021.6, in which a borate-containing solution was utilized as buffer.For interrupted tests (2, 4, or 14 days of duration), the test rods were unloaded by reversing the SSRT machine so that the strain was reduced by the same rate as they were loaded.
The sulfide concentration was sampled at the sampling valve in Figure 5 and immediately analyzed its sulfide content by means of a CHEMetrics sulfide analyzing vacuum glass kit. [19]The pH of the sample was also measured.The potential of the test rod and the potential of a Pt was measured periodically.This section contains the results from the exposures defined in Table 2.The first part of the results compiles the tests conducted until rupture, with stress-strain curves shown in Figure 6 and measured corrosion potentials shown in The sulfide concentration remained constant, as far as could be judged from the indicator color scale.A small dip in the concentration at about 170 h of exposure was generally observed which corresponded to a change and refill of the stock solutions.
T A B L E 2 Summary of the test conditions.Runs denominated "2018" correspond to exposures published in Taxén et al. [3] while "2021" correspond to exposures published in Moya Núñez et al. [4] Run Temperature (°C)  Figure 8 shows a photo of one part of the test rod after Run 2018.5.The blackened surface as well as the location of the final rupture close to the middle of the narrow section was typical of all specimens drawn until rupture in 1.0 mM sulfide.
Figure 9 shows a sequence of SEM-images of longitudinal cross sections of test rods from Run 2018.1 in the left-hand column with similar images from Run 2018.2, for comparison, in the right-hand column.
The images in the left-hand column of Figure 9 show cracks or crack-like features that extend to about 20-30 µm from the surface.The term crack is used somewhat loosely in this paper for any wedge-shaped cavity.The images in the right-hand column show no such cracks but only an unevenly corroded surface.The cracks in the left-hand column in Figure 9 show some solid particles.Some cracks seem very full of solids whereas others contain only a few grains.
Figure 10 from Run 2018.1 shows the distribution of cracks in the final rupture.There are cracks over the whole length of the gauge section, but the highest number of cracks is found in the transition zone to the larger diameter.The initial gauge length, 20 mm was strained to about 36 mm in the test.The computed middle of the gauge length after testing would then be 18 mm.The final rupture occurred at a distance of 51 − 32 = 19 mm into the gauge section which gives a location close to the middle of the gauge section.
A series of interrupted tests were performed to study when, during the tests, the cracks appeared.As an added benefit, it was realistic to study a much lower strain rate.Figure 11 shows stress-strain curves for tests that were interrupted after a predetermined time.At that time the machine was reversed so that the sample could be loosened.The small variations in strain between 18% and 20% strain in Figure 11 are due to an adaption to the working hours of the operator.
The measured corrosion potentials of the copper rods in this set of experiments were similar to those in Figure 7 and range from about −870 mV (vs.SCE) to −840 mV, where the highest potentials were measured for the experiment performed at 30°C, Run 2021.2.
Figure 12 shows a photo of the test rod from Run 2018.7, that was run for 4 days.
Longitudinal cross-sections of the test rods were studied using SEM and analyzed for the appearance of individual cracks and for the abundance and distribution along a profile represented as C-D or B-A in Figure 12. Figure 13 shows a selection of SEM-images for the test rod used in Run 2021.1 which was strained to about 20% during 4 days.These images can be compared to the selection in Figure 14 from Run 2021.3where the rod was strained to about 13% during 14 days.Figure 13 shows cavities that seem more compactly filled with solids than do the cavities in Figure 14.A common feature is the appearance of large grains of corrosion product outside but relatively near the cavities.
Figure 15 shows histograms for the number of cracks found after Run 2018.6 (2 days), while Figure 16 shows the corresponding data after Run 2018.7 (4 days).Traces of oxygen were frequently found at analyses of corrosion products at test rods after testing.Figure 8 shows a slight greenish tint at the outer surface.SEM-analyses indicate an increased presence of oxygen also deep inside corrosion attacks. [3] few such sites were studied by TEM and show maximum depths similar to maximum (secondary) crack depths at rupture, after approximately 14 days.
• Necking and final rupture occurs close to the middle of the gauge section.• Variations in chloride concentration, phosphate buffer concentration or a change to borate buffer had only minor, if any, effects on crack abundance and depth.• Solid corrosion products, in many cases identified as Cu 2 S, is frequently seen as detached particles rather than as thin films on the surface.

| DISCUSSION
Stress-strain curves do not reveal any signs of stress corrosion.The time to final rupture and the elongation at rupture are independent of the test conditions used.Thus, the variation in chemistry had no effect on the mechanical behavior of the rods under SSRT testing.
The presence of oxygen in the corrosion products, at some sites is believed to be mainly due to atmospheric oxidation after the test.A comparison of the measured corrosion potentials to the equilibrium diagrams shows that the measured values lie below the line for hydrogen evolution as the cathodic reaction.The black circle in the diagram in Figure 3 is representative of conditions in 1.0 mM sulfide.This circle is located about 100 mV lower than the equilibrium curve for hydrogen evolution.The relatively high overpotential was taken as an indication that the reduction of water to hydrogen was the main cathodic reaction.The exceptions were tests with 0.02 mM sulfide where a slight leakage of oxygen was the likely source of the cathodic reaction.
The location of the first crack-like attacks at the ends of the transition zone can be explained by local higher surface stress caused by the discontinuity of the geometry. [20]It was concluded that the effect of localizing an excess surface stress to the transition zone would disappear at a sufficiently large radius of the fillet. [20]It is observed that after a strain of about 10%-20%, the effective radius has increased from the initial 9 mm (Figure 4).The increase in effective radius with strain is consistent with the fact that the distribution of cracks becomes more even at higher total strains.The more even distribution of cracks after longer duration of the test implies that a number of cracks are formed, or at least become visible, only after 4 days of exposure.
Figure 17 with measurements of the deepest corrosion attacks for Runs 2021.1 through 2021.6 is an attempt at visualizing the quantifying of the corrosion morphology.The depths were measured on the longitudinal cross-section along one of the lines a-b or c-d in Figure 12.There is, of course, an infinite number of ways to cut such a cross-section.The differences between the results for the conducted experiments should not be overestimated since the measurements are somewhat subjective and some of the apparent corrosion attacks may actually be flaws in the material present from start.However, the presence of crack-like features after all these tests is significant and the temperature dependence showing deeper attacks after test at 75°C and 60°C than at 30°C.In addition, as stated above, there are cracks found after longer duration at sites where none were found after 4 days of testing.The maximum depth after the test to rupture at 90°C was estimated to be 20-30 µm which is consistent with the trend at lower temperatures in Figure 17.However, there were no interrupted tests for 90°C, therefore the reason for the deeper attack at 90°C than at lower temperatures could be the longer duration of the test as well as the higher temperature.Tests to rupture at 60°C resulted in no attacks deeper than 20 µm, which is consistent with the trend of higher temperatures giving deeper attacks.
Comparing the prevalence and depths reported in Figures 15-17 to the observation that initial crack-like features tend to decrease in depth due to general corrosion, [11] reveals an apparent conflict.Possible explanations are differences in strain and strain rate or differences in the shape of the test rod.Low strain rate and low strain, as well as the rectangular cross-section of the test rod used, [11] may all contribute to allowing a more or less adhering layer of Cu 2 S to form.With higher strain rates and higher strains, as studied here, a brittle layer of Cu 2 S probably breaks frequently and has poor barrier properties.Possibly the contraction of a circular cross-section during strain contributes to the loss of adherence.As a consequence of stronger barrier action, the sulfide concentration at the copper surface may have been lower in the referenced work than in the present work.Additionally, lower sulfide concentrations are likely to produce a more even corrosion attack because of the lower solubility of Cu 2 S; more of the corrosion products stay at the site of corrosion because the aqueous species there are less soluble.Whatever the explanation, the results [11] suggest that if our tests had contained a final exposure period without any further strain, the depth of observable crack-like features would be reduced because of general corrosion.
Figure 14 after 4 days of strain shows crack-like features partly filled with corrosion products.Aqueous transport of species between the crack-tip and the bulk solutions seems possible but somewhat impeded by the solids in the crack.Compared to the appearance of the cracks in Figure 13 after 14 days of strain, the latter are relatively filled with more solids, thus leaving only very narrow channels between grains of corrosion product for aqueous transport.This suggests progressive intergranular corrosion that gradually fills up the cracks with solid corrosion products.The solids form not only at crack-walls but also as grains that seem separated from the metal.This indicates that the solids form from a supersaturated solution and thus are secondary corrosion products, while dissolved copper sulfide species are the primary corrosion products.Similar transport of corrosion products from the site of corrosion to a remote site of deposition was found by Chen et al. [17] Although in their system the main carrier was identified as Cu 3 S 3 .
The exchange Cu(s) ↔ Cu + + e − is known to be very fast. [21]It may not be unreasonable to expect that this reaction is in electrochemical equilibrium.Aqueous equilibrium between complexes comprising HS − and Cu + also seems reasonable.Thus, the conditions illustrated in the equilibrium diagrams in Figures 2 and 3 seem relevant.Supersaturated solutions usually appear when there are no suitable sites for particle growth, so the nucleation of new grains may be slow while the growth of existing grains may be much faster.Volumes with many existing grains lead to lower degrees of supersaturation than the same volume with only a few grains.In the limit, in a volume without existing grains, all corroded copper diffuses out from the crack towards a growing grain outside the crack acting as a sink for copper.These observations allow a mechanism for the development of the cracks to be formulated as illustrated in Figure 18.
The first step in the process is that the surface film of copper sulfide breaks under the strain.The radial contraction of the test rod is probably a strong contributing factor to the loss of adherence and breaking of the initial surface film.
Corrosion initiates susceptible grain boundaries and produces small cavities.The intergranular corrosion [22] is aggravated by strain that pulls apart the grains that form the flanks of the cavity, resulting in a wedge-shaped cavity.The corrosion produces mainly dissolved copper sulfide complex species.The local solution produced by corrosion at the grain boundary becomes supersaturated with respect to solid copper sulfide, Cu 2 S. Precipitation of Cu 2 S takes place mostly outside the cavity although some Cu 2 S is formed also in the cavity.The net consumption of sulfide in the cavity is relatively small and the supply of sulfide (reactant) to the propagating front does not seem to limit the propagation rate.The rate of propagation is instead limited by the diffusion of the dissolved copper sulfide complex species (product).The sink for the dissolved copper is growing particles of Cu 2 S(s) that maintain the concentration gradient from supersaturation at the front of the cavity, toward near equilibrium outside the cavity.Eventually, grains of Cu 2 S begin to grow also inside the cavity.These grains provide shorter diffusion paths from sites of propagation (high concentration) to the sink (low concentration).
The walls of the cavity may attain a smooth appearance without any adhering films of Cu 2 S, at least not at a thickness visible in SEM.The low corrosion rate at the walls of the cavity can be explained in part because corrosion at the grain boundary has left a near-perfect crystal surface, [23] in part also because of near electrochemical equilibrium between the perfect crystals and the copper-containing solution.The absence of a visible copper sulfide film may be due to an unfavorable interaction between metal and Cu 2 S. [24] However, a smooth appearance without any adhering films may also indicate a fresh intergranular attack corresponding to the left image in Figure 18.At sites where the corrosion rate at crack walls is not so low, the initial crack-like attack may attain a more pit-like shape.
Intergranular corrosion is observed in the present study and the deepest intergranular attacks seem to be adequately explained without referring to other mechanisms of SCC, with the exception of superficial film rupture.If other mechanisms have been operative, the effects are overshadowed by intergranular corrosion.

| CONCLUSIONS
• Intergranular corrosion in the shape of cracks was observed at a level of 1 mM sulfide at 30°C, 60°C, 75°C, and at 90°C.• Tests under identical conditions with 0.02 mM sulfide did not result in such intergranular corrosion but only an uneven, blackened corroded surface.• Stress-strain curves do not reveal any signs of stress corrosion cracking.The time to final fracture and the elongation at ruptures are independent of the test conditions used.A mechanism is proposed to explain the observed mode of corrosion as well as the observed shape and depth of the cavities formed.According to the proposed mechanism, the cavities will be filled with solid copper sulfide, which will act as a barrier between aqueous sulfide and copper metal causing an ever-stronger transport limitation.Further propagation of such a cavity is only possible if the cavity is widened by strain, creating a new path for sulfide.When a cavity is reactivated, it will again tend to fill up with solid copper sulfide.

2 −
1 atm H 2 .The thick red line shows the boundary between the stability domains of copper metal Cu(s), and copper sulfide Cu 2 S(s).The thin red curve and the dashed red curves show equilibrium potentials for the dissolved species Cu(HS)

F
I G U R E 4 A collection of nine test rods were produced.Approximate locations of O-rings used to seal the inner test environment from the atmosphere are shown.[Color figure can be viewed at wileyonlinelibrary.com]F I G U R E 5 Experimental setup of SSRT-cell testing.SSRT, slow strain rate test.[Color figure can be viewed at wileyonlinelibrary.com]

Figure 7 .
Figure 7. Generally, rupture occurred after 14 days of strain in exposures at a constant extension rate of 5 × 10 −7 s −1 .The sulfide concentration remained constant, as far as could be judged from the indicator color scale.A small dip in the concentration at about 170 h of exposure was generally observed which corresponded to a change and refill of the stock solutions.
buffer.F I G U R E 6 Stress-strain curves for copper in near-neutral solutions with sulfide and chloride.Slow strain rate testing lasting until rupture.[Color figure can be viewed at wileyonlinelibrary.com]F I G U E 7 Measured corrosion potential for tests run to rupture.

F I G U R E 8
Image of the fracture of a specimen that was drawn until rupture, exposed to 1.0 mM sulfide and 10 mM chloride at 60°C (Run 2018.5).[Color figure can be viewed at wileyonlinelibrary.com]F I G U R E 9 Scanning electron microscope-images of longitudinal cross sections of test rod from Run 2018.1 (1 mM sulfide) in the left-hand column with similar images from Run 2018.2 (0.02 mM sulfide) in the right-hand column for comparison.

F
Figure 17 with the measured deepest corrosion attacks for Runs 2021.1 through 2021.6 shows a trend with temperature, so that the deepest attacks are found for 75°C and the most shallow for 30°C.Run 2021.6 with borate instead of phosphate as pH-buffer gave slightly lower maximum depth than Run 2021.3 with the phosphate buffer.

F
I G U R E 13 Selected scanning electron microscope images (SE, 15 kV) of features found in the cross-section of the rod tested in Run 2021.1.F I G U R E 14 Selected scanning electron microscope images of features found in the cross-section of the rod tested in Run 2021.3, at different magnifications.Cu 2 O(s) could tentatively be identified by electron diffraction.Summary of observations • Stress-strain curves show no significant dependence on the solution composition.• Intergranular crack or intercrystalline corrosion was observed after all tests in solutions with 1 mM sulfide.• No such cracks were observed after the test in solution with 0.02 mM sulfide and 10 mM chloride at 60°C or 90°C.• The initial cracks are preferentially located at the ends of the gauge sections of the test rods.• Tests at 60°interrupted after 2 or 4 days of testing

F
I G U R E 15 Approximate distribution of cracks along the longitudinal cross-section (Run 2018.6), 2 days.[Color figure can be viewed at wileyonlinelibrary.com]F I G U E 16 Approximate distribution of cracks along the longitudinal cross-section (Run 2018.7), 4 days.[Color figure can be viewed at wileyonlinelibrary.com]

F
I G U R E 17 The depths of the five deepest corrosion attacks for Runs 2021.1 through 2021.6.[Color figure can be viewed at wileyonlinelibrary.com]

F I G U R E 18
Illustration of the stages in the proposed hypothesis.A local intergranular attack is superimposed on general but uneven corrosion.[Color figure can be viewed at wileyonlinelibrary.com] • Intergranular corrosion in the shape of cracks develops early during the test.Tests interrupted after 2 or 4 days of straining revealed a number of cracks preferentially located towards the ends of the gauge length of the test rod.• Necking and final rupture occur close to the middle of the gauge length and not at the location where the first cracks appear.• The maximum depth of the cracks is estimated to about 20-30 µm after final rupture (tests at 60°C and at 90°C).Cracks after 4 days of testing were estimated to be about 10-20 µm deep (tests between 30°C and 75°C).