Changes in hydrogen states after interactions between hydrogen and stress-induced martensite transformation for Ni–Ti superelastic alloy

ABSTRACT Changes in hydrogen states after dynamic interactions between hydrogen and a stress-induced martensite transformation have been investigated for Ni–Ti superelastic alloy. After homogenised hydrogen concentration by aging, the interactions change the hydrogen thermal desorption behaviour. Hydrogen in a solid solution probably changes to hydrogen trapped in damage regions induced by the interactions, as reported previously. With time after the interactions, the amount of hydrogen desorbed at low temperatures (200–400°C) markedly increases, clearly indicating that hydrogen states further change from the states immediately after the interactions. In cyclic tensile deformation in the stress plateau region caused by stress-induced martensite and reverse transformations, the number of cycles to fracture substantially increases with time after the interactions. The present results indicate that hydrogen states further change to a nominally interacting state after the interactions despite the absence of dynamic structural changes and a hydrogen concentration gradient, thereby suppressing hydrogen embrittlement related to the martensite transformation of the alloy.


Introduction
Changes in hydrogen states during dynamic lattice changes such as martensite transformation play an essential role in the hydrogen embrittlement of Ni-Ti superelastic alloy [1][2][3][4].During dynamic interactions between hydrogen and stress-induced martensite transformation, hydrogen states probably change irreversibly from hydrogen in solid solution to hydrogen trapped in defects and/or damage regions induced by the interactions [2,3].These changes in hydrogen states presumably result from the accommodation of dynamic structural changes induced by martensite transformation, including lattice invariant shear [2][3][4].Hydrogen in solid solution strongly interacts with martensite transformation compared with hydrogen trapped in defects and/or damage [2,3].As a result of these interactions, a large amount of damage, including vacancy clusters and unique dislocation structures, which leads to the degradation of mechanical properties, is generated [5][6][7][8].The first interactions introduce the most damage, thereby significantly enhancing the hydrogen embrittlement [2][3][4].The cyclic stress-induced martensite transformation accumulates defects resulting in no marked degradation of mechanical properties, in the absence of hydrogen, thereby decreasing the critical stress for martensite transformation [9][10][11][12].In the presence of hydrogen, cyclic interactions accumulate damage regions rather than defects, although the effects of subsequent interactions on embrittlement are relatively small on account of hydrogen trapping in the damage regions [2,3].It is likely that with changes in hydrogen states, the interaction process perpetually affects each transformation.
Not only during but also after the interactions, the hydrogen states possibly change further.For instance, when a specimen subjected to the interactions induced by the martensite transformation immediately after hydrogen charging is aged at room temperature, hydrogen trapped in damage regions at the surface layer of the specimen becomes detrapped and diffuses into the centre of the specimen [8,13,14].In this case, the driving force of the changes in hydrogen states (hydrogen diffusion) is presumably the hydrogen concentration gradient [15][16][17][18].However, when the concentration gradient is eliminated by aging [15,18], that is by homogenisation, before the interactions, it is not clear whether hydrogen trapped in damage regions changes further after the interactions.In general, hydrogen trapped in defects or damage regions is considered to be stable and negligibly changes its state [19,20].When the trapped hydrogen further changes after the interactions even after a short time, the state is not necessarily stable but strongly trapped.Since the state of the further changed hydrogen is unknown, the effects of rechanging hydrogen states on hydrogen embrittlement cannot be predicted at the present time.Therefore, it is important to investigate whether the further changed hydrogen states enhance or suppress the hydrogen embrittlement related to the martensite transformation.The findings will be useful for further elucidating the hydrogen embrittlement mechanism.
The objective of the present study is to evaluate, in the absence of dynamic structural changes and the hydrogen concentration gradient, any additional changes in the hydrogen states after interactions between hydrogen and the stress-induced martensite transformation of Ni-Ti superelastic alloy.If the hydrogen states substantially change further after the interactions, the effects will be reflected in the hydrogen thermal desorption and hydrogen embrittlement behaviour.

Experimental procedure
A commercially available Ni-Ti superelastic alloy (Ni: 56.01 mass%, Ti: balance) wire with a diameter of 0.5 mm, the same as used in the previous study [8], was cut into specimens.The critical stress for martensite transformation, which is defined as the average stress of the stress plateau region, and the tensile strength at a strain rate of 8.33 × 10 -3 s -1 and room temperature (25 ± 2°C) were 597 ± 4 and 1335 ± 9 MPa, respectively [8].The cyclic stress-strain curves for the uncharged specimen have been reported previously [8]; no fracture occurred within 2000 cycles.The martensite start (M s ), finish (M f ), reverse transformation start (A s ) and finish (A f ) temperatures of the specimen were determined to be 14.5, -27.5, -16.0 and 20.5°C by differential scanning calorimetry (DSC) at a scan rate of 10°C min -1 under automated control [21].The specimens were carefully finished with 600-grit SiC paper and ultrasonically cleaned with acetone for 5 min.
A flowchart of the preparation procedure for six types of specimen is shown in Figure 1.Cathodic hydrogen charging was performed at a current density of 10 A/m 2 for 1 h in 0.9% NaCl aqueous solution at room temperature.To homogenise enriched hydrogen at the surface layer of the specimen, aging was carried out at room temperature for 240 h in air immediately after hydrogen charging.During aging, the hydride TiNiH (tetragonal; a = 0.6221 nm, c = 1.2363 nm [22,23]) decomposes and is not detected by X-ray diffractometry analysis [15,17,24,25].In this stage, hydrogen diffuses into the centre part of the specimen (the hydrogen distribution is homogenised) without diffusing out from the specimen, and most of the hydrogen probably exists in solid solution [3,26,27].The specimen was subjected to the pre-interactions between hydrogen charging and the stress-induced martensite transformation, i.e. a single tensile loading followed by unloading in the stress plateau region produced by the stress-induced martensite (up to 0.1 strain) and reverse transformation at a strain rate of 8.33 × 10 -3 s -1 at room temperature.As a result, most of the hydrogen in solid solution presumably changes to hydrogen trapped in damage induced by the pre-interactions [2,3].To investigate any further changes in hydrogen states after the pre-interactions, the specimens were left for 0, 6, 24, and 240 h at room temperature in air.To compare the hydrogen states after and without the pre-interactions, a specimen additionally aged for 240 h at room temperature in air after homogenising the hydrogen concentration, i.e. aged for a total of 480 h, was also prepared.
To evaluate further changes of the hydrogen states of these specimens, hydrogen thermal desorption analysis (TDA) was conducted using a gas chromatograph at a heating rate of 100°C h -1 from room temperature to 700°C.The sample gas was analysed at 5 min intervals using Ar as the carrier gas.The amount of desorbed hydrogen was defined as the integrated peak intensity.The amount of charged hydrogen was calculated by subtracting the amount of pre-dissolved hydrogen (8 mass ppm [8]) from the amount of desorbed hydrogen.
A cyclic tensile deformation test in the stress plateau region was performed until fracture at room temperature at the above strain rates.The standard deviation was calculated from the results of at least three specimens.The fracture surfaces were examined by field emission scanning electron microscopy (FE-SEM).

Further changes in hydrogen states from hydrogen thermal desorption behaviour
The representative hydrogen thermal desorption curves of the specimens are shown in Figure 2. The amount of desorbed hydrogen was approximately 200 mass ppm irrespective of specimen preparation conditions.Thus, no hydrogen diffused out of the specimen during aging or with time after the pre-interactions at room temperature, as reported previously [8,16,17,21].For the specimen aged at room temperature for 240 h after hydrogen charging for 1 h (Figure 2a), the primary desorption peak (550-600°C) and the shoulder of the peak on the low-temperature side (400-500°C) were observed.Upon additional aging for 240 h (Figure 2b), a small broad peak appeared in the low-temperature region (200-400°C), but the basic desorption behaviour was almost the same, such as the appearance of the primary desorption peak and the shoulder of the peak observed in Figure 2a.These changes in desorption behaviour are possibility related to further hydrogen homogenisation.For the specimen immediately after the pre-interactions (Figure 2c), the intensity of the primary desorption peak decreased, but the shoulder of the peak was larger.The probable origin of this is that hydrogen states change from hydrogen in solid solution to hydrogen trapped in damage induced by the dynamic interactions between hydrogen and stress-induced martensite transformation [2,3,27].In our previous study [17], the desorption behaviour in ultrahigh vacuum negligibly changed after one interaction (pre-interaction) between hydrogen and stress-induced martensite transformation for the specimen subjected to hydrogen charging followed by aging for 240 h.This difference is perhaps attributed to differences in the measurement conditions (in atmospheric pressure Ar gas or in ultrahigh vacuum) and/or the microstructure of the specimen [28][29][30].In addition, similar desorption behaviours do not necessarily indicate similar hydrogen states, although different desorption behaviours reflect different hydrogen states [3,8].Accordingly, the hydrogen states appeared to also be changed by one interaction even in our previous study [17].
For the specimen after the pre-interactions and then left to stand for 6 h (Figure 2d), a new broad peak appeared in the low-temperature region (200-400°C), although the intensity of the primary desorption peak decreased and the peak shoulder was smaller.At present, it is difficult to identify the details of the hydrogen states from only the changing behaviour of desorption.Nevertheless, these marked changes in desorption behaviour are clear evidence of further changes in hydrogen states with time after the pre-interactions despite the absence of dynamic structural changes or a hydrogen concentration gradient.The cause of the further changes in hydrogen states may be that internal stress relaxation moves hydrogen or strain fields near damage induced by the interactions further attract hydrogen during standing time, i.e.Cottrell atmosphere like.Fan et al. [31] also showed that dislocations induced by phase transformation move hydrogen from twin boundaries, as indicated by the changes in the relaxation peak of internal friction for Ni-Ti alloys, although the effects of standing time on hydrogen movement were not clarified.Hence, it appears that hydrogen states during or immediately after the interactions are temporary.With a longer standing time of 24 h (Figure 2e) or 240 h (Figure 2f) after the pre-interactions, the desorption behaviour changed negligibly.A slight further change in hydrogen states may not always be detected.Nonetheless, at least in the early stage after the pre-interactions, it is clear that hydrogen states further change considerably.

Suppression of hydrogen embrittlement by further changes in hydrogen states
Figure 3 shows the number of cycles to fracture as a function of standing time for the specimen without and after pre-interactions.For reference, the numbers of cycles to fracture for the specimens, left for 3 or 12 h are also shown.In addition, the representative stress-strain curves of the specimens are also shown from the cycle after pre-interactions.For the specimen aged for 240 h after hydrogen charging, the mean value and standard deviation of the number of cycles to fracture were 460 ± 136.The critical stress for the martensite transformation at the first cycle, i.e. the pre-interaction cycle, was 626 ± 10 MPa, which is approximately 25 MPa higher than that for the uncharged specimen (597 ± 4 MPa).The increase in critical stress is probably ascribed to the suppression of martensite transformation owing to hydrogen in solid solution [3,8,15,17,18,21,27,30,[32][33][34].In the second cycle, the critical stress decreased to 609 ± 8 MPa.At 300 cycles, the critical stress was 461 ± 4 MPa, which decreased from that of the first cycle by approximately 160-170 MPa.This decrease in critical stress is presumably attributed to the formation of damage related to cyclic interactions [3,9].Upon additional aging for 240 h (total 480 h), the number of cycles to fracture (511 ± 97) increased negligibly.
The critical stresses at the first cycle and at 300 cycles were 626 ± 9 MPa and 460 ± 5 MPa, respectively, which are almost the same as that of the specimen aged for 240 h (without additional aging) after hydrogen charging.These results are consistent with no fundamental changes in hydrogen desorption behaviour (Figure 2b).It is likely that the marked changes in hydrogen concentration and/or states rarely occur during additional aging, and eventually, they have almost no effect on the hydrogen embrittlement behaviour.Fracture mainly occurred in the elastic deformation region of martensite phase.However, in rare cases, fracture occurred during martensite transformation (stress plateau region).The martensite phase inhomogeneously nucleates in the specimen owing to the relatively high strain rate [35].As a result, in the present study, fracture may occur in the stress plateau region as well as the elastic deformation region of the martensite phase.
The specimen left for 0 h after the pre-interactions, i.e. the specimen immediately after the pre-interactions, is consequently the same as the specimen aged for 240 h after hydrogen charging in the cyclic tensile deformation test.Even for a short standing time such as 3 h or 6 h after the pre-interactions, the number of cycles to fracture increased.In the case of the specimen left for 6 h after the pre-interactions, the number of cycles to fracture was 600 ± 135.For 24 h after the pre-interactions, the number of cycles to fracture further increased to 746 ± 141, which is approximately 60% larger than that for 0 h.Thus, hydrogen states were continuously rechanging even for 24 h, although no significant change in desorption behaviour was observed (Figure 2e).However, for 240 h after the pre-interactions, the increase in the number of cycles to fracture (725 ± 70) saturated.Irrespective of standing time, the critical stresses at the second cycle (first cycle after the pre-interactions) and at 300 cycles were 605-615 MPa and 455-465 MPa, respectively.Thus, the martensite transformation behaviour itself is only slightly affected by standing time after the pre-interactions.Note that standing time substantially increases the number of cycles to fracture even for the short time.Judging from the suppression of hydrogen embrittlement, hydrogen states probably further change to a nominally interacting state during standing time.Even in the case of nominally interacting states of hydrogen, the self-multiplication of damage induced by the pre-interactions leads to the degradation of mechanical properties and eventually to fracturing with almost no influence from hydrogen [8].Details of further changed hydrogen states are beyond the scope of the present study.Nevertheless, on the basis of rechanging even with a short standing time, hydrogen states do not change drastically such as jumping trap sites but change slightly.For example, further changed hydrogen states appear to be tightly re-trapped in the vicinity of the original damage when hydrogen is trapped in the damage immediately after the pre-interactions.
The representative FE-SEM images of fracture surfaces of the specimens subjected to cyclic deformation tests are shown in Figure 4.The fracture surfaces macroscopically exhibited a brittle mode with almost no reduction in area.From the magnified views of the fracture initiation point, a microroughness pattern consisting of sub-micrograins was observed for all specimens.These features appear to be related to the damage induced by cyclic interactions between hydrogen and stress-induced martensite transformations, as reported previously [8,13,21].However, some relationships between hydrogen states and fracture surface morphology remain unclear, as reported previously [21,36].Similarly, in the present study, no significant differences in the characteristics of the fracture surface were observed among hydrogen states.It is unlikely that further changes in hydrogen states after the interactions substantially affect the fracture surface morphology.One of the reasons for this may be that because the effects of cyclic interactions until fracture are great, the effects of standing time after the pre-interactions, i.e. effects of the initial stage of the fracture processes, are offset on the fracture surface.Moreover, since fracture always occurs after cyclic interactions (after martensite transformation) rather than during the interactions, the process during the interactions may have a weaker direct effect on fracture surface.

Conclusions
We have demonstrated that for Ni-Ti superelastic alloy, hydrogen states changed by the interactions between hydrogen and stress-induced martensite transformation further change at room temperature with standing time after the interactions despite the absence of dynamic structural changes and a hydrogen concentration gradient.Hydrogen states further change even with a short time.As one evidence of further changes in hydrogen states, the amount of hydrogen desorbed at low temperatures increases with standing time after the pre-interactions.As a result of further changes in hydrogen states, the number of cycles to fracture substantially increases in cyclic tensile deformation, that is, hydrogen embrittlement related to martensite transformation is suppressed.It is likely that further-changed hydrogen states are primarily nominally interacting states, such as that of hydrogen tightly trapped in damage.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Funding
This work was supported by JSPS KAKENHI [grant number 18K04780].

Notes on contributors
Ryosuke Hayashi is a Graduate student in the Department of Materials Science and Engineering, Kyushu Institute of Technology, Japan.
Ken'ichi Yokoyama is an Associate Professor in the Department of Materials Science and Engineering, Kyushu Institute of Technology, Japan.

Figure 1 .
Figure 1.Preparation of six types of specimen: aged for 240 h after hydrogen charging, aged an additional 240 h, and left to stand for 0, 6, 24, and 240 h after pre-interactions.

Figure 2 .
Figure 2. Typical hydrogen thermal desorption curves of specimens (a) aged for 240 h after hydrogen charging, (b) aged an additional 240 h, and left to stand for (c) 0 h, (d) 6 h, (e) 24 h, and (f) 240 h after pre-interactions.The numerical values are the amounts of desorbed hydrogen.

Figure 3 .
Figure 3.Typical cyclic stress-strain curves and number of cycles to fracture as a function of additionally aged or standing time for the specimen without and after pre-interactions.The numerical values are the number of cycles to fracture.

Figure 4 .
Figure 4. FE-SEM images of fracture surfaces: general and magnified views of specimens subjected to cyclic tensile deformation (a) aged for 240 h after hydrogen charging, (b) aged an additional 240 h, and left to stand for (c) 6 h, (d) 24 h and (e) 240 h after pre-interactions.The numerical values are the numbers of cycles to fracture.