Hydrogen Distribution Permeated through a Duplex Stainless Steel Detected by Hydrogen Microprint Technique

Abstract

To provide reliable relationship between hydrogen embrittlement (HE) and hydrogen distribution, a duplex stainless steel (DSS: JIS SUS329J4L) annealed and electrolytically hydrogen-charged was investigated by means of hydrogen microprint technique (HMPT), where distribution of hydrogen was examined on the opposite side of the charged surface. Quantitative analysis was made by classifying the site of detected hydrogen into three categories: ferrite matrix, austenite grain and phase boundary. The HMPT was performed on the 1.5 h and 24 h charged specimens with two holding times in the ambient air for 0.5 (as quick as possible) and 300 h. In the 1.5 h charged and 0.5 h held specimen, hydrogen atoms were mostly detected on the phase boundary. When charging time was increased to 24 h, relative fraction of hydrogen desorbed in the austenite phase against the ferrite matrix and phase boundary increased. The relative fraction of hydrogen atoms in the austenite phase was also increased by increasing holding time to 300 h irrespective of the charging time. During the holding, hydrogen atoms inside the ferrite matrix were presumed to preferentially diffuse out from the specimen or transferred to the phase boundary, while hydrogen atoms already trapped at the phase boundary will move into the interior of the austenite phase. The results obtained in the present study where experimental conditions are systematically selected can be rationally interpreted only with the higher solubility and smaller diffusivity of hydrogen in the austenite phase, rather than considering the binding energy of hydrogen with phase boundary.

1. Introduction

Duplex stainless steels (DSSs) are composed of body-centred cubic ferrite phase and face-centered cubic austenite phase, and have higher strength and lower cost together with higher resistance to stress corrosion cracking than austenitic stainless steels that are being widely used and attractive materials for the oil and gas industry, chemical plants and structures in marine environments, etc.^1,2,3) On the other hand, hydrogen in metals has a detrimental effect on their mechanical properties reducing both ductility and fracture resistance, and all phenomena associated with its deleterious effects are known as hydrogen embrittlement (HE), to which many reported failures have been related.^4,5,6) In DSS, HE is also prone to occur^{3,7,8,9,10,11,12)} due to the higher diffusivity and higher permeation rate of hydrogen in the ferrite phase.^4,13,14)

Several mechanisms were proposed to explain HE in metals such as hydrogen-enhanced decohesion (HEDE),^15,16,17) hydrogen-enhanced localized plasticity (HELP),¹⁸⁾ hydride formation¹⁹⁾ and hydrogen-enhanced strain-induced vacancy (HESIV).^20,21,22) Despite several studies^{3,7,8,9,10,11,12)} that have been made so far, HE in DSS has not been well understood yet even in terms of phenomenology, not to mention the mechanism. Moreover, hydrogen behaviour, the key issue to clarify the mechanism for HE, has not been clarified yet, either.

Currently, several methods are used to investigate the behaviour of hydrogen in metals: thermal desorption spectroscopy (TDS),^23,24) hydrogen microprint technique (HMPT),^25,26) secondary ion mass spectroscopy (SIMS),^27,28) tritium autoradiography (TARG),^29,30) etc.

Regarding the hydrogen distribution in DSS investigated by means of HMPT, Overiero-García²⁵⁾ claimed that the phase boundary acts as diffusion path of hydrogen in the DSS from a result that hydrogen atoms were most frequently detected at the phase boundary, while Yalçì and Edmonds³¹⁾ suggested that the phase boundary will not act as a preferential diffusion path for hydrogen, from a result that shorter charging time caused a decrease in the number of hydrogen atoms detected on the phase boundary. On the other hand, Luu et al.¹¹⁾ claimed that more silver grains can be observed on ferritic phase, from which they concluded that the faster permeation rate in the ferrite than the austenite. Although they did not mention, un-negligible amount of hydrogen can be found on the phase boundary and austenite phase in the image they showed.

As described in the above paragraph, there has been inconsistency on the hydrogen distribution in DSS so far. The reason for this seems to be lack of quantification and different experimental conditions by the different authors, particularly emulsion covering side with respect to the charged side and holding time after the emulsion covering. The authors’ group claimed in the previous paper³²⁾ that most of hydrogen was observed in the ferrite/austenite phase boundary by applying HMPT on the charged side as a function of charging and holding time and by analysing the results quantitatively.

Considering that fragmental results based on different experimental conditions (detecting side with respect to the charging side, charging conditions and holding times) will not provide conclusive information, we have moved step forward in this study, and applied HMPT on the opposite side of the charged surface to trace the hydrogen diffusion path and trapping sites in DSS more accurately. In this case, the experimental procedures and assessment of the experimental results will be approximately the same as in the authors’ previous study³²⁾ carrying out HMPT on the charged side, i.e., the detected site of hydrogen will be classified into three categories: in the ferrite matrix, in the austenite grains and on the phase boundary. The amount of the hydrogen detected in the three sites will be discussed quantitatively.

2. Experimental Procedure

2.1. Material and Hydrogen Charging Method

The material used in this study is the same as in the previous study,^7,8,32) but the previous mill sheet data were found to be slightly different from the correct data, which are as follows: JIS SUS329J4L sheet with 1 mm thickness annealed at 1050°C for 1 min and then water-cooled. The chemical compositions and the microstructure of the material used are listed in Table 1 and Fig. 1, respectively. According to the phase map (Fig. 1(a)) taken by electron back scattering diffraction (EBSD), the two phases are both elongated in the rolling direction with measured phase fractions of 51.2% of austenite and 44.4% of ferrite. By ignoring the fraction of undefined area, i.e., considering the microstructure to be composed of purely two phases, phase fraction is turned to be 53.6% of austenite and 46.4% of ferrite. From the IPF map (Fig. 1(b)), it is found that the austenite grains have various orientations whereas the normal of the ferrite grains mostly have directions close to [111].

Table 1. Chemical composition of the SUS329J4L specimen in mass%.

C	Si	Mn	Cr	Mo	N	Ni	Fe
0.019	0.40	0.75	24.53	3.22	0.17	6.39	Bal.

Fig. 1.

Results of EBSD analysis of the specimen. (a) phase map, (b) IPF map. RD: rolling direction, TD: transvers direction. The colour scale indicated below IPF map is common between the two cubic phases.

Test pieces, with a gauge portion of 12 mm in length and 5 mm in width, were cut from the sheet in the longitudinal direction by electric discharge machining (EDM), ground with waterproof abrasive paper up to 3000 grit, mirror-finished by buffing with diamond paste up to 0.25 μm, and finally etched with aqua regia. The surfaces were rinsed with distilled water, cleaned ultrasonically in acetone, and then dried quickly by warm air. Electrolytic hydrogen charging was performed onto the opposite of the etched surface of the gauge portion of the test pieces (the other surfaces were insulation-coated) for 1.5 and 24 h at room temperature (RT) with platinum anode in a sulfuric acid aqueous solution with pH=2.5 containing 0.1 mass% of NH₄SCN as a hydrogen recombination inhibitor with a current density of 100 A·m⁻² and DC voltage of 10 V.

2.2. HMPT

The etched surface (the opposite of the charged surface) of the test pieces was covered with nuclear emulsion (Ilford L4 diluted with distilled water by 4 times) containing AgBr in a darkroom with wire loop method, as quickly as possible (0.5 h) after the charging. For some of the test pieces, the emulsion covering was conducted after holding the test pieces for 300 h at RT from the end of the charging. Furthermore, to investigate the hydrogen distribution and diffusion path in the middle-thickness portion, the 24 h charged test pieces were cut, and their cross section was etched and covered with the emulsion.

After holding all the above samples in the darkroom at RT for 24 h from the emulsion covering, the specimen was placed into the fixing solution (Super Fuji Fix) for totally 15 min (2 min in the darkroom and 13 min in the lab environment), and then rinsed with running water for 15 min, dried naturally and observed with an SEM (Hitachi S 3400) equipped with an EDX device to confirm the detected particle to be silver. The procedures of hydrogen charging, holding, emulsion covering, etc. are illustrated schematically in Fig. 2.

Fig. 2.

Flow of HMPT procedure. RT: room temperature.

3. Results and Discussion

In the previous research, the authors³²⁾ applied HMPT onto the hydrogen-charged surface. In the present research, to compare and observe preferential trapping sites with high binding energy, diffusion path and distribution of hydrogen in the microstructure, the emulsion was coated on the opposite of hydrogen charged surface. Figures 3(a) and 3(b) show the resultant HMPT/SEM images of the test pieces held 0.5 h after charging for1.5 h and 24 h. The light area looking like matrix is known to be ferrite phase, while the dark area frequently looking like an isolated precipitate is austenite phase.^8,32) The small white spots observed are fine silver grains, confirmed by EDX, indicating the location of hydrogen desorption. The number of silver grains naturally increases with increasing the charging time. It is revealed that most of hydrogen atoms are detected on the phase boundary while some of the particles are seen both in ferrite and austenite grains. It is also to be noted that most of the particles in the austenite phase is located close to the phase boundary, as indicated by the arrows in Fig. 3(b).

Fig. 3.

SEM images of HMPT of the specimens hydrogen-charged for 1.5 h (a) and 24 h (b), after which the opposite side of the charged surface was covered with nuclear emulsion, held for 24 h and then fixed. The time taken from the finish of charging to emulsion covering was 0.5 h.

The silver grains as shown in Fig. 3 were classified according to the location in the microstructure into those on phase boundaries, inside ferrite and austenite grains. The area fraction of silver grains is illustrated in Fig. 4 with the result of the relative area fraction in Fig. 5 to investigate the effect of charging time on hydrogen trapping site and diffusion path. Figure 4 shows that the amount of desorbed hydrogen (total area fraction of silver grains) is naturally increased from 1.8% to 3.5%. as the charging time increases from 1.5 h to 24 h. The same trend was observed in the authors’ previous study where the hydrogen charging and emulsion covering were conducted on the same side. An important feature in Fig. 5 is that the silver grains have been detected mostly (66.6%) on the phase boundary of the specimen with the shortest charging time, while in the 24 h charged sample, silver particles observed in the ferrite matrix and austenite phase increases in relative comparison with those in the interphase region. The increase in the number of the hydrogen atoms detected in austenite grains can be attributed to the increase in the diffusion distance inside the austenite grains. Based on the reported diffusivity of hydrogen in austenite phase at 20°C,¹⁴⁾ about 10% of the surface concentration will be attained at a distance of 8 μm from the surface for a diffusion time of 24 h. In the case of the present paper, from the large differences in the diffusivity and solubility between ferrite and austenite phases,¹³⁾ the hydrogen detected in the austenite phase is presumed to diffuse from the charged surface initially inside ferrite matrix, trapped at the phase boundary and then diffuse into the austenite grain.

Fig. 4.

Area fraction of the silver grains in the HMPT images, corresponding to the conditions in Fig. 3. Viewing area and hydrogen-charging time: (a) 1318 μm² and 1.5 h, (b) 2164 μm² and 24 h, respectively.

Fig. 5.

Relative area fraction of the silver grains in the HMPT images, corresponding to the conditions in Fig. 4.

Figures 6(a) and 6(b) show the HMPT/SEM images of the specimens hydrogen-charged under the same charging condition of the above specimens but held for 300 h at the ambient air prior to emulsion covering. The area percentage and relative fraction of silver grains in the viewing area is illustrated in Figs. 7 and 8, respectively. It is noticed from Fig. 7 that the total fraction of silver grains is naturally increased by increasing the charging time, as is the same in Fig. 4. The increase in the relative fraction of the hydrogen desorbed in austenite phase can be also seen in Fig. 8, as was found in Fig. 5. By comparing Fig. 7 with Fig. 4, it is noticed that the total fraction of silver grains decreased by increasing the holding time from charging to emulsion covering in each charging time compared to the 0.5 h held specimen. This means that the amount of remaining hydrogen atoms was naturally decreased by the holding at RT for 300 h. From Fig. 5, it was found that the increase in the relative fraction of the silver particles in the ferrite matrix is small with increasing charging time from 1.5 h to 24 h. In contrast, the increase in the relative fraction in the austenite phase with increasing charging time was marked, while the relative fraction on the phase boundary was decreased correspondingly. By comparing Fig. 8 with Fig. 5, the relative fraction of silver particles on the boundary is decreased by increasing holding time from 0.5 h to 300 h both in the cases of 1.5 and 24 h charging, but still the largest among the three sites. The hydrogen atoms in the trapping site with low binding energy such as in the ferrite matrix will diffuse out during the holding of 300 h or move to phase boundary that is regarded as a trapping site with high binding energy. Also, the hydrogen atoms already trapped at the phase boundary will move into the interior of the austenite phase. Thus, the feature in the specimen 24 h charged and 300 h held approaches the distribution of trapping site with high binding energy for hydrogen in the austenite phase and phase boundary.

Fig. 6.

SEM images of HMPT of the specimens hydrogen-charged for 1.5 h (a) and 24 h (b), and then held in ambient environment for 300 h, after which the opposite side of the charged surface was covered with nuclear emulsion, held for 24 h and then fixed.

Fig. 7.

Area fraction of the silver grains in the HMPT images, corresponding to the conditions in Fig. 6. Viewing area and hydrogen-charging time: (a) 1756 μm² and 1.5 h, (b) 1642 μm² and 24 h, respectively.

Fig. 8.

Relative area fraction of the silver grains in the HMPT images, corresponding to the conditions in Fig. 7.

The hydrogen detected on the opposite side should have diffused through the ferrite matrix and desorbed from the three kinds of sites. The hydrogen from ferrite is attributable to the high diffusivity in this phase, while that from the phase boundary may be to its high binding energy with hydrogen. However, Olden et al. claimed from finite element simulation that the hydrogen distribution can be explained considering the solubility and diffusivity of hydrogen in the austenite phase that are larger and smaller by factors of 1×10³ and 2×10⁻⁶ times,¹³⁾ respectively, without considering the high binding energy of the phase boundary. Although it cannot be discussed quantitatively on the binding energy of hydrogen with the three sites dealt with in the present study, it is rational that hydrogen atoms along the phase boundary can diffuse into the austenite phase because of the high solubility in this phase, and detected by HMPT at locations close to the boundary, examples of which were indicated by the arrows in Fig. 3(b). Thus, unlike the previously reported fragmental HMPT results,^11,25,31) the results obtained in the present study where experimental conditions are systematically selected can be rationally interpreted only with the higher solubility and smaller diffusivity of hydrogen in the austenite phase, rather than considering the binding energy.

Figure 9 shows the distribution of hydrogen in the cross section (middle thickness) of the specimen for the charging time of 24 h, which is referred to from the previous paper for reference.³²⁾ Here, the horizontal direction of the image was corrected to RD from TD in the previous paper, and the time taken from the finish of charging to emulsion covering was corrected to about 2 h from 0.5 h described in the previous paper. It is revealed that most of the silver particles are seen on the phase boundaries. Figure 10 shows the comparison of the relative area fraction of the three locations: charged-side surface, cross section (middle thickness region) and the opposite side (present result); the former two are quoted from the previous paper.³²⁾ It is obvious that fraction of hydrogen atoms in the phase boundary decreases as the distance from the charged side is increased, while those in the austenite phase and ferrite matrix increase. This should be discussed by considering that the charging and holding condition is almost the same for the three locations. The increase in the fraction in the austenite phase with the distance from the charged surface cannot be explained directly by considering that the solubility in the austenite phase is larger than in the ferrite matrix.

Fig. 9.

SEM image of HMPT of the specimen hydrogen-charged for 24 h, cut to reveal the cross section, which was ground, polished, etched and then covered with nuclear emulsion, held for 24 h and then fixed.³²⁾ The time taken from the finish of charging to the emulsion covering was about 2 h.

Fig. 10.

Relative area fraction of the silver grains in the HMPT images of the test pieces after 24 h charging, (a) charged side surface,³²⁾ (b) cross section³²⁾ and (c) opposite side surface. (Area fraction of total of the silver particle in each area indicated at the top of the columns).

Firstly, extremely sharp concentration gradient after finishing the charging must be taken into account. During charging, hydrogen concentration adjacent to the surface should be in local equilibrium with high fugacity of hydrogen in the electrolyte side, which will be far greater than the solubility at ambient pressure, but right after the charging is stopped the surface concentration will be fallen to the solubility. This extremely sharp concentration gradient will cause much faster outward diffusion in the depth (thickness) direction than the other directions such as from the phase boundary to the inside of austenite grain. Resultantly, the hydrogen detected by HMPT will reflect just the distribution at the end of charging.

Secondary, it should be considered that the oxide layer may act as barrier against the hydrogen desorption. If this is the case, hydrogen atoms that has reached the opposite side will stay near this surface and has sufficient time to diffuse into the austenite phase. In aluminium, dense oxide film is known to be the barrier for hydrogen diffusion.³³⁾ Since in stainless steels, the protective oxide layer is known to be also dense,³⁴⁾ it will act as the barrier against hydrogen desorption as well. On the other hand, some paths for hydrogen introduction must have been formed during charging because of high fugacity produced by electrolytic charging in the surface oxide film. Although there is no evidence, these kinds of paths may be formed preferentially at the phase boundary because phase boundaries are prone to form local cell.

Another rather curious feature in Fig. 10 is that the area fraction of the total silver particles indicated on the top of the figure is naturally highest at the charged surface, but that it is smallest in the middle thickness region, which may be attributable to the minimum holding time required for the emulsion covering. To observe the middle-thickness area, cross section was prepared by cutting from the charged specimen, ground, polished, etched and then covered with emulsion. It took about 2 h from the end of hydrogen charging to emulsion covering, roughly four times longer than in the case of the surfaces. Some of the hydrogen atoms should have diffused out during this handling period.

4. Conclusion

Hydrogen was introduced onto one surface of DSS by electrolytically hydrogen charging method. Hydrogen distribution in relation to the microstructure was observed on the opposite side of charging surface of the specimen immediately after charging and after holding for 300 h by means of HMPT. Quantitative analysis has been made in terms of the sites (ferrite matrix, austenite phase and phase boundary) where hydrogen atoms were desorbed and detected by HMPT as a function of charging and holding times.

The HMPT revealed that the percentage of silver particles was largest at phase boundary, next inside ferritic phase, and lowest inside the austenite phase immediately after charging. After 300 h holding from hydrogen charging, the austenite phase acted as a second trapping site with high binding energy after phase boundary. During the holding, hydrogen atoms inside the ferrite matrix were presumed to be preferentially diffused out from the specimen or transferred to the phase boundary, while hydrogen atoms already trapped at the phase boundary will move into the interior of the austenite phase. The results obtained in the present study where experimental conditions are systematically selected can be rationally interpreted only with the higher solubility and smaller diffusivity of hydrogen in the austenite phase, rather than considering binding energy of hydrogen with phase boundary.

Acknowledgement

The authors would like to thank Professor K. Tsuzaki in Kyushu University for providing the access to EBSD analysis.

References

• 1)   J.-O.  Nilsson: Mater. Sci. Technol., 8 (1992), 685. https://doi.org/10.1179/mst.1992.8.8.685
• 2)   T.  Maki,  T.  Furuhara and  K.  Tsuzaki: ISIJ Int., 41 (2001), 571. https://doi.org/10.2355/isijinternational.41.571
• 3)   A. M.  Elhoud,  N. C.  Renton and  W. F.  Deans: Int. J. Hydrog. Energy, 35 (2010), 6455. https://doi.org/10.1016/j.ijhydene.2010.03.056
• 4)   H. K. D. H.  Bhadeshia: ISIJ Int., 56 (2016), 24. https://doi.org/10.2355/isijinternational.ISIJINT-2015-430
• 5)   M.  Nagumo: ISIJ Int., 41 (2001), 590. https://doi.org/10.2355/isijinternational.41.590
• 6)   S.  Takagi,  Y.  Hagihara,  T.  Hojo,  W.  Urushihara and  K.  Kawasaki: ISIJ Int., 56 (2016), 685. https://doi.org/10.2355/isijinternational.ISIJINT-2015-566
• 7)   A.  Yousefi and  G.  Itoh: ISIJ Int., 58 (2018), 561. https://doi.org/10.2355/isijinternational.ISIJINT-2017-631
• 8)   A.  Yousefi,  G.  Itoh,  Z.  Ghorani and  S.  Kuramoto: ISIJ Int., 59 (2019), 2319. https://doi.org/10.2355/isijinternational.ISIJINT-2019-159
• 9)   T.  Zakroczymski,  A.  Glowacka and  W.  Swiatnicki: Corros. Sci., 47 (2005), 1403. https://doi.org/10.1016/j.corsci.2004.07.036
• 10)   S.-L.  Chou and  W.-T.  Tsai: Mater. Chem. Phys., 60 (1999), 137. https://doi.org/10.1016/S0254-0584(99)00077-2
• 11)   W. C.  Luu,  P. W.  Liu and  J. K.  Wu: Corros. Sci., 44 (2002), 1783. https://doi.org/10.1016/S0010-938X(01)00143-3
• 12)   F.  Zucchi,  V.  Grassi,  C.  Monticelli and  G.  Trabanelli: Corros. Sci., 48 (2006), 522. https://doi.org/10.1016/j.corsci.2005.01.004
• 13)   V.  Olden,  A.  Saai,  L.  Jemblie and  R.  Johnsen: Int. J. Hydrog. Energy, 39 (2014), 1156. https://doi.org/10.1016/j.ijhydene.2013.10.101
• 14)   E.  Owczarek and  T.  Zakroczymski: Acta Mater., 48 (2000), 3059. https://doi.org/10.1016/S1359-6454(00)00122-1
• 15)   R. A.  Oriani and  P. H.  Josephic: Acta Metall., 22 (1974), 1065. https://doi.org/10.1016/0001-6160(74)90061-3
• 16)   R. A.  Oriani and  P. H.  Josephic: Acta Metall., 25 (1977), 979. https://doi.org/10.1016/0001-6160(77)90126-2
• 17)   W. W.  Gerberich,  R. A.  Oriani,  M. J.  Lji,  X.  Chen and  T.  Foecke: Philos. Mag. A, 63 (1991), 363. https://doi.org/10.1080/01418619108204854
• 18)   H. K.  Birnbaum and  P.  Sofronis: Mater. Sci. Eng. A, 176 (1994), 191. https://doi.org/10.1016/0921-5093(94)90975-X
• 19)   D. S.  Shih,  I. M.  Robertson and  H. K.  Birnbaum: Acta Metall., 36 (1988), 111. https://doi.org/10.1016/0001-6160(88)90032-6
• 20)   M.  Nagumo,  M.  Nakamura and  K.  Takai: Metall. Mater. Trans. A, 32 (2001), 339. https://doi.org/10.1007/s11661-001-0265-9
• 21)   M.  Nagumo: Mater. Sci. Technol., 20 (2004), 940. https://doi.org/10.1179/026708304225019687
• 22)   K.  Takai,  H.  Shoda,  H.  Suzuki and  M.  Nagumo: Acta Mater., 56 (2008), 5158. https://doi.org/10.1016/j.actamat.2008.06.031
• 23)   T.  Izumi and  G.  Itoh: Mater. Trans., 52 (2011), 130. https://doi.org/10.2320/matertrans.L-M2010825
• 24)   K.  Ebihara and  H.  Kaburaki: ISIJ Int., 52 (2012), 181. https://doi.org/10.2355/isijinternational.52.181
• 25)   J.  Ovejero-García: J. Mater. Sci., 20 (1985), 2623. https://doi.org/10.1007/BF00556094
• 26)   K.  Ichitani,  S.  Kuramoto and  M.  Kanno: Corros. Sci., 45 (2003), 1227. https://doi.org/10.1016/S0010-938X(02)00218-4
• 27)   K.  Kawamoto,  Y.  Oda,  H.  Noguchi,  H.  Fujii,  T.  Izumi and  G.  Itoh: J. Solid Mech. Mater. Eng., 3 (2009), 898. https://doi.org/10.1299/jmmp.3.898
• 28)   A.  Nishimoto,  M.  Koyama,  S.  Yamato,  Y.  Oda,  T.  Awane and  H.  Noguchi: ISIJ Int., 55 (2015), 335. https://doi.org/10.2355/isijinternational.55.335
• 29)   T.  Otsuka,  H.  Hanada,  H.  Nakashima,  K.  Sakamoto,  M.  Hayakawa,  K.  Hashizume and  M.  Sugisaki: Fusion Sci. Technol., 48 (2005), 708. https://doi.org/10.13182/FST05-A1022
• 30)   H.  Hanada,  T.  Otsuka,  H.  Nakashima,  S.  Sasaki,  M.  Hayakawa and  M.  Sugisaki: Scr. Mater., 53 (2005), 1279. https://doi.org/10.1016/j.scriptamat.2005.07.031
• 31)   H. K.  Yalçì and  D. V.  Edmonds: Mater. Charact., 34 (1995), 97. https://doi.org/10.1016/1044-5803(94)00056-Q
• 32)   Z.  Ghorani,  A.  Yousefi and  G.  Itoh: ISIJ Int., 59 (2019), 1901. https://doi.org/10.2355/isijinternational.ISIJINT-2019-121
• 33)   T.  Izumi,  G.  Itoh,  N.  Itoh and  Y.  Sasaki: Mater. Sci. Forum, 519–521 (2006), 1265. https://doi.org/10.4028/www.scientific.net/msf.519-521.1265
• 34)   G.  Okamoto: Corros. Sci., 13 (1973), 471. https://doi.org/10.1016/0010-938X(73)90031-0