Forensic Investigation of Stainless Steel 316 Hydrogen-Membrane and Ammonia-Cracking Reactors Through Mechanical Testing

Abstract

Knowledge of alloy behavior under industry-relevant conditions is critical to hydrogen production and processing, yet it is currently limited. To understand more about the impact of hydrogen damage on stainless steel 316 under realistic in-service conditions, we conducted a forensic investigation of two reactors exposed to various hydrogen-processing conditions. We examined samples of reactor walls exposed to hydrogen-containing atmospheres for >100 and ~1000 h at elevated temperatures during hydrogen separation and ammonia cracking. The samples were characterized by tensile testing, stretch–bend testing, and three-point bending. A loss in ductility and strength was observed for the reactor wall material compared with both untreated materials and materials annealed in neutral atmospheres at the same temperatures used during reactor operation. The three-point bend testing, which was conducted on inner and outer pipe-surface material extracted via electrical discharge machining, showed larger changes in the flexural modulus of exposed reactors but increases in the elastic limit. Microstructural observations revealed that hydrogen may play a role in stress relaxation, possibly promoting normalization at lower-than-expected temperatures. We also observed that materials exposed to ammonia undertake more damage from nitriding than from hydrogen.

1. Introduction

Hydrogen is an emerging green fuel expected to form part of power generation for the 21st century and beyond [1]. It is generated from processes such as gasification, syngas processing, methane reforming, and ammonia cracking [2]. This has challenged engineers who are being asked to make large-scale, long-life (10–30 years), reliable infrastructure that can withstand the damage caused by hydrogen. In metals, especially stainless steels, hydrogen embrittlement leads to the formation of inclusions, hydride phases, and gas-containing voids. This type of damage creates flaws, which initiate cracks surrounded by embrittled material, resulting in accelerated crack propagation and significantly reduced fatigue life at both low and high temperatures [3,4,5].

This is not a new problem for engineers. Damage from hydrogen diffusion into stainless steels, including stainless steel 316 (SS316), was being investigated as early as 1951 [6]. Since then, most mechanical testing of stainless steels in hydrogen environments has focused on tensile strength and fatigue behavior. This has revealed that hydrogen gas causes [3,4,7]:

• Hydrogen embrittlement, when dissolved hydrogen within the metal causes a loss of ductility, which occurs at low temperature < 100 °C.
• Hydrogen stress cracking, when applied stress drives the diffusion of atomic hydrogen to cause localized embrittlement and cracking. This occurs in the absence of a corrosion process, i.e., when the material is polarized cathodically. When a corrosion reaction is present, i.e., the material is anodically polarized, cracking localized embrittlement/pits/inclusions is referred to as hydrogen stress corrosion cracking.
• Hydrogen blistering, through which hydrogen diffuses into the metal and accumulates at voids or inclusions to form subsurface planar cavities. The atomic hydrogen reforms into H₂ gas and creates localized gas pressure/stress.
• Hydrogen-induced cracking, when cracks form to connect hydrogen blisters.
• Metal hydride embrittlement, when the diffusion of atomic hydrogen leads to the formation of brittle metal hydride phases in the material.
• High-temperature hydrogen attack, when atomic hydrogen reacts with carbides, sometimes forming methane in the steel and leading to decarburization and internal fissuring.

Looking at these damage mechanisms, hydrogen has the ability to diffuse into and interact with metals, causing significant microstructural damage that leads to sudden and catastrophic failure. At temperatures of 500–1200 K, the diffusion of hydrogen into austenitic stainless steel is as low as 59 kJ/mol, and recombination from atomic to gaseous hydrogen is 47 kJ/mol [8]. Compared with SS316, chromium forms both the protective Cr₂O₃ oxide layer and strengthening MX carbides, as it has an activation energy of 243 kJ/mol: the lowest activation energy of the elements within the materials [9]. In a hydrogen-rich environment, the low activation energy of hydrogen means it starts to damage the material under conditions far milder than those needed to activate the inherit protective and strengthening mechanisms characteristic of stainless steel [10].

Although much mechanical testing of stainless steels in hydrogen environments has been conducted and much learned, these tests do not necessarily represent all in-service conditions. Significant focus has been placed upon the effect of hydrogen on fatigue failure, but far less so on creep. Well known in the power-generation industry, creep is the slow plastic deformation of materials at low stresses, static loads, and elevated temperatures. The plasticity arises from diffusion and dislocation processes that can be influenced by the hydrogen-damage mechanisms identified above for fatigue.

Of the works found for stainless steels JIS SUS304, JIS SUS304L [11], JIS SUS304L [12], AISI310S [13], SS304, and SS304L [14], a hydrogen environment was found to reduce creep life, but it did not show embrittlement. Indeed, the samples tested under a hydrogen environment had higher ductility, leading to increased strain rates and greater reductions in final cross-sectional area. The specimens also failed from ductile fracture, rather than intergranular fracture. The increase in ductility has been proposed to be linked to hydrogen-enhanced, localized plasticity and decarburization, but recent work suggests this may not be the case [12]. Therefore, it remains unknown why the long-term plasticity of stainless steels in a hydrogen environment results in mechanical behavior contrary to that expected for hydrogen embrittlement.

Having a plant operating at constant pressure (low stress) and at high temperatures for long periods of time in a hydrogen-rich environment is inevitable. Therefore, engineers need a greater understanding of material behavior. In this work, we conducted the forensic testing of equipment used in two well-controlled scientific experiments conducted using CSIRO: hydrogen-membrane and ammonia-cracking reactors. The hydrogen-membrane reactor was used in experiments with H₂ flux permeating through palladium-coated, tubular vanadium membranes by applying the constant pressure method under 99.99% H₂ [15]. The ammonia-cracking reactor was used in an ammonia-cracking system tested under a variety of conditions, and it produced high-purity H₂ from NH₃ using a commercial ruthenium-based catalyst and CSIRO membrane technologies [16]. Both applications undergo constant low stress and high temperatures, i.e., creep conditions. In both applications, the SS316 inner surface of the pipe is exposed to a slightly cooler, hydrogen-rich environment, while the outer surface is hotter and exposed to air. This is a realistic, in-service condition in which the sample material is uniformly exposed to hydrogen with no temperature gradient.

We conducted our investigation of the two reactors to gain some understanding of the impact of hydrogen damage on SS316 under realistic in-service conditions. The hydrogen-membrane (HM) reactor’s inner surface was exposed to 100% H₂ atmosphere (1 bar) at 320–350 °C for ~100 h, and the outer surface was exposed to air (1 bar) at 350–360 °C. The ammonia-cracking (AC) reactor’s inner surface was exposed to a H₂ + N₂ + HN₃ atmosphere (H₂ (vol%) ~70%, NH₃ (vol%) ~10%, P = 5–6 bar), at 500–520 °C for ~1000 h, and the outer surface was exposed to air (1 bar) at 450–500 °C. Samples were extracted from the reactor and subjected to tensile, three-point bend, and stretch–bend tests. To identify any impact of hydrogen on the mechanical properties of the reactors’ material, the results of the test were compared with those of the same size and shapes of unexposed (new) material, and with the material exposed at the same temperature only. We analyzed the samples via optical microscopy and a scanning electron microscope with energy dispersive spectroscopy (SEM/EDS) to understand possible mechanisms of hydrogen and ammonia interaction with the material.

2. Materials and Methods

2.1. Materials

The HM tube and AC pipe were cut into strips in the longitudinal direction (Figure 1a). The strips were 10 mm wide × 100 mm long for the HM reactor, and 20 mm wide × 200 mm long for the AC reactor. These samples are referred to as ‘used’ throughout this paper.

Two types of SS316 reactor walls were investigated:

• HM reactor used for hydrogen separation. The tube was 25.4 mm (1”) in diameter and 1.2 m long, with 1.8 mm wall thickness (Figure 1b). The inner surface of the tube was exposed to 100% H₂ atmosphere (1 bar) at 320–350 °C for ~100 h. The outer surface was exposed to air (1 bar) at 350–360 °C.
• AC reactor: 25.4 mm (1”) diameter pipe, 1.5 m long, with 6.0 mm wall thickness (Figure 1c). The inner surface of the pipe was exposed to a H₂ + N₂ + NH₃ atmosphere (H₂ (vol%) ~70%, NH₃ (vol%) ~10%, P = 5–6 bar), at 500–520 °C for ~1000 h. The outer surface was exposed to air (1 bar) at 450–500 °C.

2.2. Mechanical Testing and Analysis

Tensile testing was conducted using ASTM A370 (AS 1391) [17] for non-standard sample geometries at 23 °C in air, using Beijing Time High Technology Ltd. model WDW–100E (Beijing, China). HM samples had a gauge section with a 1.8 × 5.4 mm cross section and a 25 mm length, while AC reactor samples had a gauge section with a 6 × 12.7 mm cross section and 50 mm in length. The test was conducted on single samples due to the large sample size and limited material at the 7.5 mm/min cross-head speed. Although hardness testing is often used to quantify embrittlement, it only measures localized plastic deformation, requires significant surface preparation, potentially destroying critical features, and can only be conducted on flat samples, rather than a pipe. Therefore, a stretch–bend test (SBT) or clamped beam bend test [18,19] was conducted to understand the impact of embrittlement on the plasticity of the bulk material. This test is used to understand the ductility and fracture behavior of thin metal [20]. The strain measured in an SBT test is dependent on non-metallic inclusions, content, or shape, therefore providing more insight into the impact of microstructural damage caused by a hydrogen environment. SBT was conducted using an Instron 1000HDX by LMATS (Norwood, MA, USA). A U-shaped indenter with a 5 mm radius was used, with the span between clamp being 210 mm. The HM sample had a cross section of 1.8 × 5.4 mm, while the AC reactor sample had a cross section of 6 × 12.7 mm. Tests were conducted at 1 mm/min (repeated two to three times for new and used samples) and 0.1 mm/min (single sample) to investigate the impact on the elastic limit.

Three-point bend tests were conducted on 1.8 × 1.8 mm samples extracted using electrical discharge machining (EDM) cutting. Full, through-thickness samples were used for the HM material, whereas, for the AC material, samples were extracted from the inner and outer surfaces. Test pieces were placed on the machine with the exposed surface (hydrogen/air) experiencing compression. Testing was undertaken according to ASTM E290 [21] at 1.0 mm/min at 23 °C in air, using Instron 5544A (Norwood, MA, USA).

Stress, strain, and modulus were calculated using the rectangular formula, as the samples from the pipe were sufficiently small to negate the curvature of the pipe:

$$Bendingstress={\displaystyle \frac{3Fl}{{2bd}^2}}$$

$$Bendingstrain={\displaystyle \frac{6Dd}{L^2}}$$

$$Flexuralmodulus={\displaystyle \frac{L^{3}m}{{4bd}^3}}$$

where F is force (N), L is length between supports (mm), b is breadth of sample (mm), d is depth of sample (mm), D is the deflection at the maximum point (mm), and m is the initial slope of the load deflection curve (N/mm).

All mechanical tests were conducted in standard laboratory conditions (room temperature, air). Mechanical tests have a force error of ±0.5 N and a displacement error of ±0.005 mm from the crosshead. A strain rate of 1.0 mm/min was chosen for the three-point bend test, as it produced the most contrast between new and used samples. Bend tests were conducted in duplicates showing less than 5% difference between measurements.

For comparison, new or unused SS316 tube/pipe of the same diameter was also sampled and is designated as ‘new’ throughout this paper. According to the time–temperature–transformation diagram for SS316 [22], the conditions for the HM reactor will not induce any significant microstructural changes; therefore, the new material was tested in the as-received condition for the HM reactor. As the temperature of the AC reactor is higher and in the range that results in sensitization, the new material for the AC reactor was heat-treated in N₂ at 520 °C for 1000 h, corresponding to the operating temperature of the AC reactor. The heat-treated samples are designated as ‘new heat-treated’, while the exposed HM and AC reactors’ samples are designated as ‘used’ throughout this paper.

2.3. Microstructural Analysis

Mechanically processed samples were cut with a water-cooled abrasive wheel saw. The cross sections were hot-mounted, ground, and polished to a 1 µm finish, with final polishing performed using colloidal silica (0.3 µm). The polished samples were electrolytically etched in a 10% w/v oxalic acid solution. Samples were then analyzed via optical microscopy under a Leica DMi8A metallographic inverted microscope (Wetzlar, Germany) and under a field-emission SEM (JEOL7001F, Tokyo, Japan) at 15 Kv with EDS Oxford XMax 80 mm² window and AZtecOne 7.0 software.

3. Results and Discussion

3.1. Tensile Tests

The tensile test results for the new and used samples are shown in Figure 2a and Figure 2b for the HM and AC reactor, respectively.

Figure 2a shows that the used tube underwent an overall reduction in strength and a mild loss in ductility in the HM reactor. In Figure 2b, we see that the as-received and the heat-treated pipe have similar properties. The used AC reactor tube exhibits a smaller difference in the stress–strain behavior between the new and used material despite its having been in the hydrogen environment for 10 times longer than the HM reactor.

Looking more closely at the results reveals that

Table 1. Tensile properties of new and used hydrogen-membrane (HM) and ammonia-cracking (AC) reactor SS316.

Sample	Young’s Modulus (GPa)	Elastic Limit (MPa)	Ultimate Tensile Strength (MPa)	Strain at Failure (%)
HM new	730	415	659	42
HM used	522	320	545	36
AC new as-received	169	182	644	88
AC new heat-treated	189	193	619	91
AC used	138	167	510	75

quantifies the tensile mechanical properties. Looking firstly at Young’s modulus shows that the high value for the HM material is due to its being a cold-drawn tube, rather than a hot-drawn pipe. For both materials, there was a reduction in the Young’s modulus and elastic limit after exposure to a hydrogen environment. This has been widely observed, and is known as hydrogen-enhanced localized plasticity. It arises from the hydrogen having greater mobility than the dislocations, thus causing softening, as there is a reduction in elastic interaction between dislocations [23]. In terms of ultimate tensile strength, we see drops of 17.3% and 17.6%, which are also matched with reductions in overall ductility.

3.2. Stretch-Bend Tests

The next tests performed were SBT on the thin HM samples. SBT testing is similar to three-point bend testing, but it requires the samples to be clamped at either end. As the sample is loaded, it experiences biaxial stresses transversely (bending) and longitudinally (stretching). Although not commonly conducted, SBT is used mainly on plate metal to predict the location and mode of failure.

Figure 3 shows the load–displacement results of the new and used HM SS316 samples, with samples tested with the inner surface downward, i.e., exposed surface in tension. The new material is initially softer and has a higher maximum load, while the linear elastic portion of both tests is similar. Looking at

Table 2. Stretch-bend test results for hydrogen-membrane material.

Sample	Maximum Load, kN	Displacement at Maximum Load, mm
Displacement at 1 mm/min
New	23.46–24.18	33.41–34.16
Used	18.15–20.21	25.2–33.81
Displacement at 0.1 mm/min
New	16.84	31.09
Used	12.96	21.96

, which lists multiple results, we see the higher strain rate tests show a drop in maximum loading of ~19.5% and a reduction in ductility of ~12.7% in used samples. As the slow strain rate test is dominated by plasticity mechanisms and is more sensitive to the yield of the material, we see a ~23% reduction in the maximum load and a ~29.4% reduction in ductility in the used samples. The loss in ductility becomes more apparent when the fractured samples in Figure 4 are examined. The new material shows a clean 45°-angle rupture along the shear plane, while the used material shows an almost 90° angle with characteristic brittle fracture, including multiple rupture sites.

3.3. Bend Tests

To understand the impact of hydrogen embrittlement, bend tests were conducted on samples extracted using EDM. This allowed for small 1.8 × 1.8 mm samples to be extracted from the new and used HM material matching the thickness of the pipe. For AC material, 1.8 × 1.8 mm samples were taken from the inner exposed surface and the outer surface. Tests were stopped at 5% strain as samples began to slip, rather than rupture. Figure 5 shows the bend test results, and

Table 3. Bend-test results for hydrogen-membrane reactor and ammonia-cracking reactor material.

Sample	Flexural Modulus (GPa)	Elastic Limit (MPa)	Stress at 5% Strain (MPa)
HM new	68.9 ± 1.5	211.1 ± 4.6	519.4 ± 11.4
HM used	108.5 ± 2.4	361 ± 7.6	620.4 ± 12.9
AC new inner	102.9 ± 1.9	179.6 ± 3.4	406.5 ± 7.7
AC used inner	73.7 ± 1.5	244.4 ± 5.1	431.5 ± 8.8
AC new outer	105.7 ± 2.1	180.5 ± 3.6	426.8 ± 8.5
AC used outer	79.9 ± 1.8	207.4 ± 4.5	445.4 ± 9.2

shows the modulus, elastic limit, and stress at 5% strain.

Looking the results in Table 3, we see that, for the AC materials with hydrogen exposure, the elastic limit increases, and the flexural modulus decreases. The HM material is a thin-walled pipe, where we a similar trend with the elastic limit increasing by 41.5% and an increase in the maximum recorded stress of 16.3% when used. For the AC material, we see larger changes for the inner material, which is exposed to ammonia, compared with the outer material, which is exposed to oxygen. The largest change is in the elastic limit, where the inner material showed an increase of 36.5%, while the outer material only showed an increase of 13%. The point of difference between the thin-walled hydrogen exposed (HM) and the thick-walled ammonia exposed (AC) samples is the flexural modulus. For the AC material the softening of the modulus is most likely a result of the metal spending time at high temperatures, as both the ammonia- and oxygen-exposed samples have a softer modulus after use. For the HM material, the cause of the modulus increase is not obvious, but the literature suggests that this can occur due to hydrogen impacting dislocation behavior, and it could be influenced by the strain rate of the test [24].

3.4. Microstructure

Overall, the tensile surface of the used material (Figure 6a,d) shows rupture from the inner/hydrogen-exposed surface, reflecting the results of the tensile and SBT tests. The appearance of fracture surfaces after tensile testing is consistent with ductile rupture at room temperature along the 45° shear plane as shown in Figure 6. This contrasts with the rupture shown in Figure 4, which shows a 90° shear plane for the used material, indicating that a more severe stress state may be required to forensically determine the impact of hydrogen embrittlement on ductility.

Looking more closely at the HM material, the microstructure of the new sample showed deformation bands at the outer surface (Figure 7a) but less at the inner surface (Figure 7c), as expected from pipe-forming processes. Comparing the outer surfaces of the new and used material (Figure 7c,d), we see very little difference in grain structure or carbides. The inner surfaces (Figure 7a,b) show a change in the shape of the grain beyond that expected from exposure to heat. The used material shows a far higher population of rounded grains compared with the new material. For Figure 7a,b, the number of grains with an aspect ratio under 10:1 within 500 μm of the surface is 6.4 times higher in the used inner pipe surface. This indicates that the hydrogen is assisting or accelerating heat-treatment effects to produce a microstructure similar to that which has been annealed/normalized at a temperature 200 °C lower than expected. Creating larger, rounded grains and stress relieving the work hardening from the pipe-manufacturing process, reduces the tensile strength but it can increase ductility, toughness, and compression strength, as seen in the bend test results. This is supported by the work of Turnbull and Zhou [25] and Tien et al. [26], who have observed that hydrogen can induce stress relaxation in steels.

For the AC material, the used sample displayed a more pronounced banded structure than the new sample, as seen in Figure 8. Looking at the outer surface, we see the effects of in-service conditions, i.e., exposure to heat and oxygen. Oxidation impacts the surface roughness, and carbide precipitates slightly increase (indicated with blue arrows). At the mid-wall (Figure 8c,d), the new material has a typical elongated grain profile consistent with pipe material, whereas the used material shows the development of some large, round grains with greater numbers at the bottom of the image, which is closer to the ammonia-exposed surface. The inner wall shows the greatest difference, with the used material showing a higher number of carbides, as well as the presence of rounder grains and deformation-induced martensite.

The new heat-treated (520 °C in N₂ atmosphere for ~1000 h) AC material and the used AC material were then analyzed using SEM/EDS. Figure 9a,b show the fracture and general surface condition of the new heat-treated AC material exposed to N₂. Figure 9c,d show the fracture and general surface of used AC material exposed to N₂ + H₂ + NH₃. The EDS data are presented in

Table 4. Average composition of randomly selected areas of new and used tensile tested ammonia-cracking reactor material (wt.%) obtained via a scanning electron microscope with energy-dispersive spectroscopy.

Sample Identification	Fe	Cr	Ni	Mo	Mn	Si	N
Composition of new heat-treated SS316	69	17	10	2	2	<1	<1
(1) New heat-treated, mid-wall	69	17	10	2	2	<1	<1
(2) New heat-treated, external surface	69	17	10	2	1	<1	<1
(3) New heat-treated, internal surface	69	16	10	2	2	<1	<1
(4) Used, outer surface	69	18	10	2	2	<1	<1
(5) Used ACR, mid-wall	69	17	10	2	2	<1	<1
(6) Used ACR, inner subsurface	69	18	10	2	2	<1	<1
(7) Used ACR, inner surface	66	15	10	1	2	<1	4
(8) Used ACR, grain boundary void	68	17	10	2	1	<1	<1
(9) Used ACR, grain boundary void	69	16	10	2	2	<1	<1

. The composition of each sample represents the average of 10 randomly selected areas (within dashed boxes). The standard deviation of the results (0.2–0.3 wt.%) is much smaller than the accuracy of EDS measurements (1–3 wt.%).

Figure 9 and Table 4 show that the new heat-treated material experiences a typical pipe microstructure and has a consistent composition from the inner surface to mid-wall and the external surface. No carbides/precipitates or obvious oxide layers are present. The used AC material shows the effect of ammonia on the inner surface, with 50 μm of well-established corrosion and at least another 25 μm of degraded corroding material. Looking at Table 4 (Sample 7), we see a drop in Fe, Cr, and Mo and an increase in N. The areas with the decohesion of grain boundaries between elongated grains (yellow arrows (8) and (9) in Figure 9c, inset), which are caused by tensile stress followed by etching, indicate a weakening of grain boundaries. This was more severe in the used sample. The possible source of weakening grain boundaries could be the dissolution of hydrogen at elevated temperatures under the conditions of both H₂ separation (100% H₂) and ammonia cracking (H₂ + N₂ + some NH₃). Although primarily concerned with hydrogen embrittlement, EDS indicates nitridation attack is the primary damage mechanism with nitrogen, N, through the Cr-rich oxide layer, forming brittle Cr₂N precipitates, rather than strengthening Cr₂₃C₆ precipitates. Without the Cr-rich protective oxidation layer, the Fe and Mo are impacted; in some cases, Fe has been observed to start forming its own oxidation layers [27].

When correlating the microstructural observations with the tensile and three-point bending tests and SBT, the reduction in elastic and flexural modulus is possible if the material is experiencing an accelerated annealing driven by the in-service temperatures and presence of hydrogen resulting in grain growth, as seen in vanadium [28]. This softens the material by relieving the residual stresses from the pipe-manufacturing process that work-harden the material [29]. The increase in elastic limit for the bend tests could also arise from this phenomenon, as those tests were performed with the exposed side under compression, and hydrogen-assisted annealing could have reduced hardness.

4. Conclusions

In this article, we have presented a forensic investigation into material extracted from hydrogen-membrane and ammonia-cracking reactors. The number of samples and tests conducted reflect the material available. This limits the number of tests that can be performed and the conclusions that can be drawn. Tensile tests provided expected results, with a decrease in Young’s modulus, elastic limit, and ductility after exposure to a hydrogen environment. To investigate the loss in ductility, a stretch-bend test was performed, which produced similar results to the tensile tests. Three-point bend testing produced more dramatic results, with more distinct changes to flexural modulus but increases in elastic limit in hydrogen-exposed materials. Microscopy revealed that hydrogen may play a role in stress relaxation, possibly promoting normalization at lower-than-expected temperatures. The ammonia-exposed material showed more damage from nitriding than from hydrogen. Our work has identified the following key issues:

• More investigation is needed into how hydrogen interacts with metallic materials and their strengthening mechanisms under in-service conditions. The effect of heat treatment resulting from in-service conditions needs to be accounted for when assessing the impact of hydrogen embrittlement. The potential of stress relaxation caused by hydrogen and its impact on grain morphology also needs more understanding.
• Bend testing provides more pronounced results on the mechanical properties of an exposed material, as the loading conditions target surface damage. Bend testing may be a more informative test than tensile testing for forensic investigation.

Although much research focuses on simulating or reproducing hydrogen embrittlement and studying its impact on materials, there is significant value in forensically investigating existing equipment, as new knowledge can be gained that is needed to ensure the long-term success of hydrogen as fuel.