Corrosion Science xxx (xxxx) xxx–xxx Contents lists available at ScienceDirect Corrosion Science journal homepage: www.elsevier.com/locate/corsci Eﬀects of alloying elements on the corrosion behavior of Ni-based alloys in molten NaCl-KCl-MgCl2 salt at diﬀerent temperatures ⁎ Hua Sun , Peng Zhang, Jianqiang Wang Shanghai Institute of Applied Physics, Chinese Academy of Sciences, 2019 Jialuo Road, 201800 Shanghai, PR China A R T I C LE I N FO A B S T R A C T Keywords: Ni-based alloy Molten chloride salt Corrosion Alloying element The eﬀects of alloying elements on corrosion of Ni-based alloys in NaCl-KCl-MgCl2 under Ar in the temperature range of 600∼800 °C were investigated by immersion test, SEM/EDS, EPMA and XRD techniques. A thermodynamic potential-pO2− diagram for Cr, Ni at 700 °C was constructed. The alloy corrosion is attributed to the dissolution of Cr to form Cr chlorides into salt. The elements Cr, Mo and W can aﬀect the alloy corrosion. The corrosion is accelerated and the intergranular corrosion becomes more pronounced with temperature. Furthermore, the eﬀects of the alloying elements on corrosion have a strong temperature dependence. The related mechanisms are discussed. 1. Introduction Molten chloride salts are promising candidates of heat transfer ﬂuid (HTF) and thermal energy storage (TES) media for next generation concentrating solar power (CSP) plants because of their beneﬁcial characteristics such as low cost, low melting point, high boiling point and good heat transfer property [1–4]. However, a main challenge using molten chloride salts in next generation CSP plants is the corrosion problem of structural materials at the aim operating temperatures (> 600 °C) [1,4,5]. Therefore, the material degradation caused by corrosion is one of key issues related to next generation CSP plants. Ni-based alloys are considered as candidate structural materials for next generation CSP plants due to combined good mechanical properties and corrosion resistance [4,6]. Much work was denoted to investigate the corrosion behavior of various Ni-based alloys in molten chloride salts in order to evaluate their corrosion resistance [7–15]. Vignarooban et al.  investigated that the corrosion behavior of Hastelloy N, C276, C22 in molten NaCl-KCl-ZnCl2 at 250 °C and 500 °C and found that Hastelloy N showed the highest corrosion rate due to its low Cr content. However, Ambrosek  evaluated the corrosion resistance of various Ni-based alloys including Hastelloy N, Inconel 617, Inconel 625, Inconel 718 and Haynes 230 etc. in molten KCl-MgCl2 at 850 °C and pointed out that Hastelloy N exhibited the least attack depth because of its low Cr content. Alkhamis  compared the corrosion performance of Hastelloy C276 and Haynes 230 in molten MgCl2-KCl and NaCl-KCl-ZnCl2 at 800 °C and found that Hastelloy C276 with high Mo showed better corrosion resistance than Haynes 230 with high W. ⁎ On contrary, Oryshich and Kostyrko  demonstrated that Mo and W were similar in improving the corrosion resistance of Ni-based alloys in molten chloride salts. It can be seen that the alloying elements such as Cr, Mo and W can aﬀect the resistance of Ni-based alloys against molten chloride salts. Nevertheless, no systematic studies have been performed to clarify the relationship between the alloying elements and the corrosion behavior of Ni-based alloys in molten chloride salt. Since the CSP plants will use mature commercial alloys with various complex alloying elements, understanding how the alloying elements aﬀect the corrosion behavior of commercial alloys is crucial. In addition, it is generally recognized that the corrosion of the alloys in molten chlorides salt is attributed to the selective dissolution and/or oxidation of the element Cr depending on the diﬀerent experimental conditions. The corrosion products covered on the alloy surfaces are usually characterized as Cr-rich oxides such as Cr2O3, MgCr2O4, NiCr2O4 or LiCrO2 according to the diﬀerent molten salt systems [11–13]. Nevertheless, the major existing form of Cr dissolved into the molten salt is unknown due to the diﬃculties associated with characterizing the species with low concentration in molten salt. Therfore, the corrosion mechanism of the alloys in molten chloride salts, especially in the molten salts under inert gas is not well understood. In fact, it is convenient to predict the corrosion products of the alloys in molten salt by constructing the potential-oxoacidity (E-pO2−) relationship [16–19]. The objective of this work is to investigate the eﬀects of the alloying elements Cr, Mo and W on the corrosion behavior of commercial Nibased alloys in molten chloride salt under Ar atmosphere in the Corresponding author. E-mail address: firstname.lastname@example.org (H. Sun). https://doi.org/10.1016/j.corsci.2018.08.021 Received 19 November 2017; Received in revised form 31 July 2018; Accepted 8 August 2018 0010-938X/ © 2018 Elsevier Ltd. All rights reserved. Please cite this article as: Sun, h., Corrosion Science (2018), https://doi.org/10.1016/j.corsci.2018.08.021 Corrosion Science xxx (xxxx) xxx–xxx H. Sun et al. Table 1 Concentration of main impurities in as-received salts (ppm, by weight). Salt Cr Fe Ni Zn Co Mo SO42− PO43− NO3− NaCl KCl MgCl2 1.01 1.12 0.32 1.64 1.39 1.88 0.52 < 0.005 / 0.50 0.48 0.49 1.0 1.0 1.06 1.49 1.48 1.23 < 25 < 25 < 25 < 25 < 25 < 25 < 25 < 25 < 25 temperature range of 600–800 °C by means of weight change measurement and microstructure characterization techniques. A thermodynamical E-pO2− diagram of Cr, Ni is constructed. The related mechanisms are also discussed. 2. Experimental methods Fig. 1. Schematic diagram of the experimental setup. 2.1. Preparation of salt mixture and specimens Afterwards, the Al2O3 crucibles were divided into three groups according to the experimental temperatures, and subsequently the Al2O3 crucibles of the same group were encapsulated into the same 316SS crucible. Ar arc welds were used to seal the 316SS crucibles. Notably, all of these procedures were performed inside the glove box full of highpurity Ar (99.999%), where H2O and O2 contents were 2 ppm and 0.5 ppm (mole fraction), respectively and the system pressure was about 1 atm. Correspondingly, H2O and O2 partial pressures in the experimental system were calculated to be 2 × 10−6 and 5 × 10-7 atm, respectively. After the crucibles were sealed, the corrosion tests were performed in the high temperature resistance furnaces outside the glove box. The furnaces were heated to 600 °C, 700 °C and 800 °C, respectively and held at each temperature for 400 h. After the tests, the furnaces were cooled to room temperature. The crucibles were removed and cut using a mechanical cutting machine. The specimens were retrieved and cleaned ultrasonically in deionized water and alcohol to remove the residual salt. The specimens were then dried and weighed to obtain the weight changes with a accuracy of ± 0.01 mg. The surface and cross-sectional images of the specimens were examined using scanning electron microscopy (SEM, Merlin Compact). The elemental distributions of the specimens were analyzed by electron probe micro-analysis (EPMA, Shimadzu EPMA-1720). The precipitated phases on the specimens were identiﬁed by energy dispersive X-ray spectroscopy (EDS, LEO 1530 V P) and X-ray diﬀraction (XRD, D8 Advance). It should be noted that the specimens for crosssectional analyses were embedded into epoxy resin and again abraded and polished. Before measurement, the prepared cross-sections were sprayed Au or Pt to enhance the conductance of the specimens. The ternary eutectic NaCl-KCl-MgCl2 (33-21.6–45.4 mol%) was selected as the experimental molten salt. Before tests, the salt mixture was prepared using commercial NaCl, KCl and MgCl2 (greater than 99.5 wt % purity), supplied by Sinopharm Chemical Reagent Corporation (Shanghai, China). The concentration of main impurities in as-received salts is listed in Table 1, which was analyzed by inductively coupled plasma-optical emission spectroscopy (ICP-OES) and ion chromatography (IC). The salts were weighed and mixed according to the above ratio, and then placed in Al2O3 crucibles. The salts were oven dried at 300 °C for 24 h. After that, the salts were transferred rapidly into a glove box full of high-purity Ar (99.999%) to minimize moisture absorption. Four commercial Ni-based alloys including three Ni-Mo-Cr alloys (Hastelloy N, C276 and C22) and one Ni-W-Cr alloy (Haynes 230) were selected as the alloys tested according to their chemical compositions, as seen in Table 2. The Mo content (16.8 wt.%) of Hastelloy N is similar to that (15.6 wt.%) of C276, while the Cr content (7.01 wt.%) of Hastelloy N is obviously lower than that (16.2 wt.%) of C276. The Cr content (21.6 wt.%) of Hastelloy C22 is similar to that (21.6 wt.%) of Haynes 230, while the Mo content (13.1 wt.%) of C22 is replaced by the W content (13.4 wt.%) of Haynes 230. The specimens with a size of 10 mm × 10 mm × 2 mm were abraded with emergy paper to 1500 grit and polished with 0.05 μm Al2O3 power, followed by cleaning with deionized water and alcohol and then drying. 2.2. Experimental setup and procedure The corrosion tests were performed in an experimental setup with double layer crucibles, as shown in Fig. 1. Al2O3 crucibles were chosed as the inner layer crucibles because of their chemical inertness in molten chloride salts [5,20]. 316 stainless steel (316SS) crucibles with lids were chosed as the outer layer crucibles in order to avoid the eﬀects of H2O and O2 from the external environment on the experimental results. Before corrosion tests, Al2O3 crucibles, 316SS crucibles and lids were cleaned with deionized water and alcohol and then dried at 120 °C for 24 h to remove residual moisture. Three specimens of each alloy were placed horizontally in each Al2O3 crucible in order to ensure the same contact surfaces between the specimens and the molten salt. Solid NaCl-KCl-MgCl2 (3321.6–45.4 mol%) salt of 40 g was then added into each Al2O3 crucible. 3. Results 3.1. Weight change of the alloys Fig. 2 shows the weight changes of the Ni-based alloys exposed in molten NaCl-KCl-MgCl2 under Ar atmosphere in the temperature range of 600∼800 °C for 400 h. All alloys underwent weight loss, except that Haynes 230 showed weight gain at 800 °C. The weight loss of the alloys at 700 °C was obviously larger than that at 600 °C, indicating a degraded corrosion resistance. In addition, the weight loss of Haynes 230 was obviously larger than that of the other Ni-based alloys at 600 °C and Table 2 Chemical compositions of the alloys tested (wt.%). Alloy Ni Mo Cr Fe Co Mn Al Ti Si C W HastelloyN C276 C22 Haynes230 70.56 58 Bal Bal 16.8 15.6 13.1 1.2 7.01 16.2 21.6 21.6 4.16 5.9 3.7 1.17 0.002 0.17 0.56 0.24 0.52 0.4 0.26 0.47 0.15 0.002 0.36 0.015 0.055 0.004 < 0.001 0.1 3.2 2.9 13.4 2 0.39 < 0.01 0.37 Corrosion Science xxx (xxxx) xxx–xxx H. Sun et al. Fig. 4 shows the cross-sectional EPMA mappings of the alloys exposed in molten NaCl-KCl-MgCl2 under Ar atmosphere at 600 °C for 400 h. Hastelloy N, C276 and C22 exhibited little Cr depletion with a depth less than 5 μm, while the Cr depletion depth of three Ni-Mo-Cr alloys seemed to increase in the following order: Hastelloy N < C276 < C22. Haynes 230 showed non-uniform Cr depletion and its maximum Cr depletion depth was up to 30 μm. No signiﬁcant Mo and W depletion was detected on the Ni-Mo-Cr and Ni-W-Cr alloys, suggesting that Mo and W exhibited the excellent resistance against molten chlorides. Hosoya et al.  also pointed out that Mo and W can not be attacked in molten chloride salts. It should be noted that Mo was enriched in the Cr depletion regions of C276 and C22, which could be attributed to the diﬀusion of Mo from the alloy base to the surface because C276 and C22 suﬀered more severe corrosion than Hastelloy N. In addition, some Mo-rich and W-rich phases were distributed randomly on Hastelloy N and Haynes 230 matrices, respectively, which were likely related to Mo-rich and W-rich carbides. Fig. 5 shows the cross-sectional EPMA mappings of the alloys exposed in molten NaCl-KCl-MgCl2 under Ar atmosphere at 800 °C for 400 h. The corrosion of the alloys was still attributed to the Cr depletion. Compared with 600 °C, the Cr depletion depths of all tested alloys inceased and the Cr depletion along the grain boundaries became more marked. The results reveal that the corrosion of the alloys is accelerated and shifted gradually from general corrosion to intergranular corrosion with increasing temperature in accordance with the above SEM results. The extent of Cr depletion of four alloys was very diﬀerent at 800 °C. The Cr depletion depth of Hastelloy N was about 30 μm, whereas that of C276 was increased to about 50 μm. Although there was no signiﬁcant diﬀerence in the Cr depletion depths of C22 and C276, more Cr was dissolved in the Cr depletion layer of C22. The Cr depletion layer of Haynes 230 was almost throughout the whole specimen, and Cr depletion primarily occurred along the grain boundaries in addition to the uniform Cr depletion near the surface of the alloy. The EPMA results were also good consistent with the above SEM results, that is, the alloying elements Cr, Mo and W can aﬀect the corrosion of Ni-based alloys, resulting that the corrosion resistance of the alloys degraded in the following order: Hastelloy N > C276 > C22 > Haynes 230. Furthermore, the eﬀects of the alloying elements on corrosion have a strong temperature dependence. Compared with 600 °C, a number of Mo-rich precipitated phases were observed along the grain boundaries of the Ni-Mo-Cr alloys at 800 °C. It should be noted that the Mo-rich phases near the Hastelloy N surface reduced versus the alloy base, whereas the Mo-rich phases may be concentrated on the surfaces of C276 and C22. In contrast, no signiﬁcant W-rich phases were detected at the grain boundaries of Haynes 230. Furthermore, the W-rich phases in the near-surface region of the alloy obviously reduced at 800 °C, suggesting that the W-rich phases could be dissolved during corrosion. Otherwise, the Ni and W depletion layers with a depth of about 20 μm were also detected in the nearsurface region of Haynes 230 at 800 °C, indicating that the elements Ni and W could suﬀer slight dissolution. Besides, the signiﬁcant amounts of O element were observed in the Cr depletion regions of the alloys exposed in molten NaCl-KCl-MgCl2 under Ar atmosphere at 800 °C for 400 h, as shown in Fig. 6. Therefore, it may be assumed that the enriched O mainly came from the insoluble species in molten salt instead of the corrosion products of the alloys, which caused the weight gains of the alloys. Furthermore, the amount of enriched O increased with the accelerated corrosion, resulting that the alloys exhibited smaller weight loss and even weight gain even though they suﬀered more severe corrosion as evidenced in Fig. 2. The concrete reasons for the O enrichment will be further discussed. As we known, the composition and structure of the precipitated phases at grain boundaries are closely related to the corrosion behavior of the alloys. Therefore, the cross-sections of various alloys were etched with metallographic etchant to further study the diﬀerences of the precipitated phases using SEM/EDS analysis, and the corresponding Fig. 2. Weight changes of Ni-based alloys exposed in molten NaCl-KCl-MgCl2 under Ar in the temperature range of 600∼800 °C for 400 h. 700 °C, suggesting that the Ni-W-Cr alloy exhibited worse corrosion resistance than the Ni-Mo-Cr alloys. However, the weight changes of four Ni-based alloys followed the diﬀerent tendency as temperature increased from 700 °C to 800 °C. The weight loss of Hastelloy N increased, and the weight of Hastelloy C276 and C22 showed little change, while Haynes 230 shifted from weight loss to weight gain. As a result, it is not possible to accurately assess the corrosion resistance of the alloys only according to the result of weight changes. The further microstructure analyses are necessary. 3.2. Cross-section analyses Fig.3 shows the cross-sectional SEM images of the alloys exposed in molten NaCl-KCl-MgCl2 under Ar atmosphere in the temperature range of 600∼800 °C for 400 h. At 600 °C, Hastelloy N showed slight corrosion attack in the form of shallow subsurface voids with a depth of about 4 μm. At 700 °C, the alloy exhibited obvious corrosion with a depth of about 15 μm. The voids were distributed uniformly in the nearsurface region of the alloy, while particularly pronounced along the grain boundaries close to the alloy base. At 800 °C, the alloy corrosion was further accelerated. Apart from small amounts of surface voids, Hastelloy N exhibited intergranular corrosion and the whole corrosion depth was about 36 μm. These results reveal that the corrosion of Hastelloy N is accelerated and the intergranular corrosion becomes pronounced with increasing temperature. The similar phenomena were also observed on C276 and C22. However, the corrosion grow rates of three Ni-Mo-Cr alloys were different with temperature. At 600 °C, the corosion depths of C276, C22 and Hastelloy N were very similar. However, the corrosion depths of C276 were increased to 25 μm at 700 °C and more than 50 μm at 800 °C, respectively, which were much higher than that of Hastelloy N. The corrosion depths of C22 were the same as that of C276 at 700 °C and 800 °C, whereas its void density and the general corrosion depth were signiﬁcantly higher than that of C276. Haynes 230 underwent more severe corrosion than the above NiMo-Cr alloys. Its local maximum corrosion depth increased sharply from 25 to 100 μm as temperature increased from 600 °C to 700 °C, while the corrosion voids were almost throughout the whole specimen at 800 °C. It is concluded that the corrosion resistance of the Ni-Mo-Cr alloys in molten chloride salt increases with an increase of Mo content and a decrease of Cr content. Moreover, three Ni-Mo-Cr alloys show better corrosion resistance than the Ni-W-Cr alloy. Furthermore, the diﬀerences in the corrosion resistance of the alloys become more signiﬁcant with temperature. 3 Corrosion Science xxx (xxxx) xxx–xxx H. Sun et al. Fig. 3. Cross-sectional SEM images of Hastelloy N(a,b,c), C276(d,e,f), C22(g,h,i) and Haynes 230(j,k,l) exposed in molten NaCl-KCl-MgCl2 under Ar for 400 h at 600 °C(a,d,g,j), 700 °C(b,e,h,k) and 800 °C (c,f,i,l). observed on the surfaces of the Ni-Mo-Cr alloys. As a replacement, numerous voids were distributed uniformly on the alloy surfaces at 600 °C (Fig. 8a, c and e), while the voids were mainly presented at the grain boundaries of the alloys at 800 °C (Fig. 8b, d and f). The results were consistent with the cross-sectional SEM and EPMA results. Otherwise, many Mo-rich phases were segregated around the voids on the surfaces of the Ni-Mo-Cr alloys based on the surface SEM/EDS results. The XRD analysis showed these Mo-rich phases primarily presented as MoNi3 and small amounts of Mo2C, which were diﬀerent from those distributed on the alloy bases (Fig. 7). Combining with the surface and cross-section results, therefore, it can be assumed that the element Mo can diﬀuse from the alloy base to the surface and subsequently form MoNi3 and Mo2C around the voids during the corrosion process of the Ni-Mo-Cr alloys. Luo et al.  also found that Mo can transport onto the surface of Inconel 617 exposed in molten LiF-MgF2-KF, which can be attributed to the vacancy diﬀerence between the alloy base and the Cr depletion layer. Similarly, Yin et al.  also found the Mo-rich phases within the Cr depleted regions of Hastelloy N after corrosion in FLiNaK salt at 700 °C. Moreover, the amount of Mo-rich phases in Hastelloy N increased with the acceleration of corrosion. The Mo-rich phases around the voids are taken as a reason for better resistance of the Ni-Mo-Cr alloys. It is worth mentioning that the Ni enrichment was also found in the general corrosion layers of C276 and C22 (Fig. 4b, c and 5 results are shown in Fig. 7. It can be seen that the signiﬁcant continuous precipitated phases were found at the grain boundaries of four Ni-based alloys, even though they were not observed at the grain boundaries of Haynes 230 by the above EPMA analysis. According to the EDS results, the precipitated phases on the alloys consisted of carbides, whereas their chemical compositions were diﬀerent. The carbides on the Ni-MoCr alloys were composed of Ni, Mo, Cr and C, and the Cr content of carbides increased with increasing the Cr content of the alloys. In general, however, the carbides on the Ni-Mo-Cr alloys were enriched in Mo and Ni, and the maximum Cr content of the carbides was not above 20 wt.%. In contrast, the carbides on the Ni-W-Cr alloy consisted of Ni, W, Cr and C. Furthermore, the carbides were enriched in Cr (above 45 wt.%) which was far more than that of the carbides at the grain boundaries of the Ni-Mo-Cr alloys. The diﬀerences in the chemical compositions of the carbides presented at the grain boundaries were probably responsible for the diﬀerent resistance of various alloys against the molten chloride salt. 3.3. Surface analyses Figs. 8 and 9 show the surface SEM/EDS results and XRD patterns of various alloys exposed in NaCl-KCl-MgCl2 under Ar atmosphere for 400 h at 600 °C and 800 °C, respectively. No corrosion products were 4 Corrosion Science xxx (xxxx) xxx–xxx H. Sun et al. Fig. 4. Cross-sectional EPMA mappings of various alloys exposed in molten NaCl-KCl-MgCl2 under Ar at 600 °C for 400 h: (a)Hastelloy N; (b) C276; (c) C22; (d) Haynes 230. and XRD results, while the Mo-rich phases decreased near the surface of Hastelloy N at 800 °C based on the EPMA results. The concrete reason for the discrepancy will be analyzed in the future work. The Ni-W-Cr alloy, i.e. Haynes 230 suﬀered intergranular corrosion b, c). Furthermore, the thickness of Ni-rich layers increased with temperature, which could be due to the leftovers of base element Ni caused by Cr depletion. In addition, many Mo-rich phases can be observed on the surface of Hastelloy N at 600 °C and 800 °C based on the SEM/EDS 5 Corrosion Science xxx (xxxx) xxx–xxx H. Sun et al. Fig. 5. Cross-sectional EPMA mappings of various alloys exposed in molten NaCl-KCl-MgCl2 under Ar at 800 °C for 400 h: (a)Hastelloy N; (b) C276; (c) C22; (d) Haynes 230. ascribed to MgO combining with the XRD results (Fig. 9d). While the irregular phases distributed uniformly on the alloy surface mainly consisted of Ni element, which could be attributed to the corroded alloy base. It appeared that most of alloy base was dissolved and therefore and no corrosion products were found on the alloy surface at 600 °C (Fig. 8g). In contrast, two diﬀerent phases were observed on the surface of Haynes 230 at 800 °C (Fig. 8h). The large particles on the top surface with diameters of 10∼20 μm were composed of Mg and O, which were 6 Corrosion Science xxx (xxxx) xxx–xxx H. Sun et al. Fig. 6. EPMA mappings of O element on various alloys exposed in molten NaCl-KCl-MgCl2 under Ar at 800 °C for 400 h: (a)Hastelloy N; (b) C276; (c) C22; (d) Haynes 230. activity in molten salt, i.e. pO 2 − = −log α O2 − . The calculation procedure is described in detail elsewhere [17,18]. The diagram is divided into three regions: immunity, oxidation and dissolution. The domain of immunity on the diagram deﬁnes the combination of potential and pO2− where corrosion does not occur. The oxidation domain corresponds to the combination of potential and pO2− where the alloying elments are oxidized to form their oxides. The dissolution domain means that the combination of potential and pO2− where the alloying elements are dissolved to form their chlorides. This suggests that the stability domains of Cr and Ni can be established as a function of potential and pO2−. In the present work, H2O and O2 partial pressures were evaluated to be 6.5 × 10−6 and 1.6 × 10−6 atm at 700 °C, respectively, according to the ideal gas law pV = nRT . H2O and O2 were converted into HCl and Cl2 by the following reactions (1) and (2) [24–27]. signiﬁcant Ni and W depletion was also detected in the near-surface region of the alloy as observed in the EPMA results (Fig. 5d). The result further indicated that the substrate of Haynes 230 suﬀered severe corrosion at 800 °C. 4. Disscusion 4.1. Thermodynamic analysis about the alloy corroison The SEM and EPMA results revealed that the corrosion of the alloy was mainly ascribed to the dissolution of Cr. Similar results were also reported in the works of Ambrosek, Alkhamis and Cho et al. [8,9,14] about the corrosion of Ni-based alloys in the molten chloride salt under inert atmosphere. However, there is little discussion on the chemical composition of Cr dissolved in molten chloride salt. Indeed, the corrosion of the alloys in molten salt is electrochemical process in natural . As already mentioned, the stable corrosion products of the alloys in molten salt can be predicted by the thermodynamic E-pO2− relationship. In order to better understand the corrosion mechanisms, therefore, the E-pO2- diagram for the active element Cr and base element Ni at 700 °C in molten chloride salt was calculated based on the thermodynamics software of HSC 6.0, as shown in Fig. 10. Where E represents the equilibrium electrode potentials of various electrochemical reactions versus standard chlorine reduction electrode (Cl2/Cl−), and pO2− represents the negative logarithm of O2− ion MgCl2 (l) + H2 O(g) = MgO(s) +2HCl(g) (1) 4HCl(g) + O2 (g) = 2Cl2 (g) +2H2 O(g) (2) Because H2O partial pressure was about four times larger than that of O2 and moisture in salts can also hydrolyze the salt by reaction (1), forming HCl. Therefore, it can be assumed that O2 was converted completely into Cl2, resulting that the cover gas above the molten salt was mainly composed of Ar, HCl and Cl2. In this case O2 in the molten salt was absent, and correspondingly O2− in the molten salt was close to zero. 7 Corrosion Science xxx (xxxx) xxx–xxx H. Sun et al. Fig. 7. SEM/EDS analysis of the precipitated phases at the grain boundaries of various alloys exposed in molten NaCl-KCl-MgCl2 under Ar at 800 °C for 400 h, etched with 50 vol.% HCl + 50 vol.% alcohol + small amount of H2O2: (a) Hastelloy N, (b) C276; (c) C22; (d) Haynes 230. chlorine partial pressure. Due to the most negative oxidation potential of Cr, the element Cr at the alloy/salt interface was preferentially electrochemically dissolved to form CrCl2 into the salt. Studies by Zahs et al.  also showed that Cr was attacked preferentially in chloridizing atmospheres, followed by Fe and Ni. Based on the above analyses, the corrosion reactions of Ni-based alloys in the molten chloride salt under the present conditions are inferred as follows: Anodic reaction: The Cl2 as the depolarized species was involved in the cathodic process. The Cl2 partial pressure was calculated to be 3.2 × 10−6 atm at 700 °C according to reaction (2) and Cl− ion activity in the molten chloride salt was assumed to be 1. According to the Nernst equation, the equilibrium electrode potential of Cl2 reduction at 700 °C was calculated to be -0.53 V (vs. Cl2/Cl−). In addition, HCl may be dissolved partially into the melt and corresponding H+ ion could act as a depolarized species to participate in the cathodic reaction. However, the equilibrium electrode potential of H2 evolution can not be calculated due to the unknown H+ ion activity and H2 partial pressure. Generally, the cathodic reaction of corrosion depends on the depolarized species with the lowest negative electrode potential , suggesting that the electrode potential of cathodic reaction under the present conditions will be equal or above -0.53 V (vs. Cl2/Cl−). As a result, the corrosion potential of Ni-based alloys was between the potentials of Cr oxidation and Cl2 reduction as labelled in Fig. 10 or shifted to more positive potential direction. It can be seen that Cr and Ni are dissolved to form CrCl2/CrCl3 and NiCl2 under the present conditions, respectively. Zahs, Kawahara and Bender et al. [26,27,29] also pointed out that the elements such as Cr, Fe and Ni tended to form their chlorides in oxidizingchloridizing atmospheres with low oxygen partial pressure and high Cr(s) → Cr 2+ (l) +2e (3) Cathodic reaction: Cl2 (g) +2e → 2Cl− (l) (4) and/or 2H+ (l) +2e → H2 (g) (5) Overall corrosion reactions: Cr(s) +Cl2 (g) = CrCl2 (l) (6) and/or Cr(s) +2HCl(l) = CrCl2 (l) + H2 (g) 8 (7) Corrosion Science xxx (xxxx) xxx–xxx H. Sun et al. Fig. 8. Surface SEM images and corresponding EDS analysis results of Hastelloy N(a,b), C276(c,d), C22(e,f) and Haynes 230(g,h) exposed in molten NaCl-KCl-MgCl2 under Ar for 400 h at 600 °C(a,c,e,g) and 800 °C(b,d,f,h). 4.2. Eﬀects of the alloying elements on the alloy corrosion The thermodynamic prediction agrees well with the experimental results, that is, the corrosion of the alloys primarily depends on the dissolution of Cr to form Cr chloride into the salt, resulting in the formation of voids at the subsurface of the alloys and weight loss. It should be mentioned that MgO with low solubility in the molten chloride salt can be formed based on reaction (1), which can deposit into the voids of the corroded specimens. Accordingly, the alloys showed the abnormal weight changes with the accelerated corrosion (Fig. 2) and the O enrichment was observed in the Cr depletion regions (Fig. 6). The present investigation found that the Ni-Mo-Cr alloys showed better corrosion resistance than the Ni-W-Cr alloy. This result implies that the element Mo is favorable to enhance the corrosion resistance of Ni-based alloys in molten chloride salts under Ar gas. As before mentioned, Alkhamis  also found Hastelloy C276 with high Mo showed better corrosion resistance than Haynes 230 with high W in molten MgCl2-KCl and NaCl-KCl-ZnCl2 at 800 °C. Additionally, Bender et al.  and Galetz et al.  also revealed that Mo containing Ni-based alloys were almost inert against chlorine induced corrosion in reducing atmospheres. However, a comparison of the Gibbs free energy of 9 Corrosion Science xxx (xxxx) xxx–xxx H. Sun et al. Fig. 9. XRD patterns of the surfaces of various alloys exposed in molten NaCl-KCl-MgCl2 under Ar for 400 h at 600 °C and 800 °C: (a)Hastelloy N; (b) C276; (c) C22; (d) Haynes 230. phases, restricting the outward diﬀusion of Cr from the alloy base because of the nobility of Mo element (Fig. 11a). In contrast, it is diﬃcult for W to diﬀuse and segregate around the voids of the Ni-W-Cr alloy due to the large atom size of W [31,32], so that the outward diﬀusion of Cr can not be prevented during the corrosion process of the Ni-W-Cr alloy (Fig. 11d). On the other hand, the chemical compositions of carbides at the grain boundaries of various alloys are diﬀerent. According to Fig. 7, the Ni-Mo-Cr alloys are prone to form the Mo-rich and Ni-rich carbides at the grain boundaries, while the Ni-W-Cr alloy facilitates the formation of Cr-rich carbides at the grain boundaries. A comparision of Gibbs free energy of formation of various divalent metal chlorides at 800 °C (CrCl2 -265.2 kJ/mol; FeCl2 -214.2 kJ/mol; NiCl2 -145.6 kJ/mol; MoCl2 -129.9 kJ/mol; WCl2 -135.7 kJ/mol) indicates that Cr is the most thermodynamically favored element in chloride salts compared to the other alloying elements, suggesting that the high Cr contents are very detrimental for the resistance of carbides against molten chloride salt. Galetz et al.  also pointed out that Mo carbides were much more resistant than Cr carbides to chlorinating attack. Therefore, the carbides at the grain boundaries of Haynes 230 can be dissolved during corrosion, leading to severe intergranular corrosion (Fig. 11e and f). In contrast, the Cr contents of carbides at the grain boundaries of the NiMo-Cr alloys are obviously smaller than that of Haynes 230, and therefore the Ni-Mo-Cr alloys show better resistance to the intergranular corrosion than Haynes 230 (Fig. 11b and c). Meanwhile, the carbides with better compatibility with the molten salt can play a better role in preventing the outward diﬀusion of Cr from the alloy base, which can further protect the alloy base from corrosion. The present investigation also found that the corrosion resistance of the Ni-Mo-Cr alloys increased with increasing Mo content and decreasing Cr content. This was mainly attributed to the diﬀerent inﬂuence of Mo and Cr on the corrosion, that is, Mo can help to improve the corrosion resistance of Ni-based alloys, while Cr tends to be dissolved preferentially in molten chlorides. Additionally, the Cr content of Fig. 10. Potential-oxoacidity (E-pO2−) diagram calculated for Cr, Ni in molten NaCl-KCl-MgCl2 at 700 °C. formation for divalent Mo and W chlorides indicates that Mo and W show similar resistance against molten chloride salt from the thermodynamic viewpoint (600 °C: MoCl2 -157.9 kJ/mol; WCl2 -155.5 kJ/mol; 700 °C: MoCl2 -143.8 kJ/mol; WCl2 -145.5 kJ/mol; 800 °C: MoCl2 -129.9 kJ/mol; WCl2 -135.7 kJ/mol). Based on the present experimental results, the diﬀerent inﬂuence of Mo and W on the corrosion resistance of Ni-based alloys mainly depends on two reasons, as shown schematically in Fig. 11. On the one hand, the elements Mo and W show diﬀerent ability in restricting the outward diﬀusion of Cr. According to the EPMA, SEM/ EDS and XRD results (Figs. 4, 5, 8 and 9), Mo can diﬀuse from the alloy base to the surface during the corrosion process of the Ni-Mo-Cr alloys and be then segregated around the voids as a form of new Mo-rich 10 Corrosion Science xxx (xxxx) xxx–xxx H. Sun et al. Fig. 11. Schematic of corrosion of the Ni-based alloys in molten NaCl-KCl-MgCl2 under Ar gas:(a,d) the early stage of corrosion; (b,e) the later stage; (c,f) observed situation. but also along line and surface defects such as grain boundaries. According to the literature [33,34], the lattice and grain boundary diﬀusion coeﬃcients of Cr in Ni-based alloys are expressed as respectively: carbides increases with Cr content of the alloy bases (Fig. 7), which is also responsible for worse corrosion resistance of the Ni-Mo-Cr alloys with higher Cr content. 4.3. Eﬀects of temperature on the alloy corrosion -1 340kJmol ⎞ 2 -1 Dv = 1.8 × 10−2 exp ⎛⎜− ⎟m s RT ⎝ ⎠ The experimental results found that the corrosion depth of the alloys increased and the intergranular corrosion became more outstanding with temperature. Futhermore, the eﬀects of the alloying elements on the corrosion of Ni-based alloys have a strong temperature dependence. The corrosion rate of alloy is mainly controlled by the outward diﬀusion of Cr from the alloy base to the surface. The diﬀusion distance of Cr in Ni could be estimated by assuming  x = (Dt )1/2 (9) -1 335kJmol ⎞ 3 -1 δDgb = 8.2 × 10−8 exp ⎜⎛− ⎟m s RT ⎝ ⎠ (10) where Dv is lattice (volume) diﬀusion coeﬃcient, Dgb is grain boundary diﬀusion coeﬃcient, δ is grain boundary width (assume δ = 0.5 nm), R is the gas constant and T is the temperature. Based on Eqs. (9) and (10), Dv and Dgb of Cr in Ni-based alloys at temperatures investigated can be calculated and are shown in Fig. 12. It is clear that Dv and Dgb increase with temperature, resulting that the Cr diﬀusion depths increase. Accordingly, the corrosion of the alloys is accelerated with temperature. (8) where x is the diﬀusion distance, D the diﬀusion coeﬃcient and t the immersion time. At the same immersion time, x is mainly related to D. The solid diﬀusion occurs not only along point (or lattice) defects, 11 Corrosion Science xxx (xxxx) xxx–xxx H. Sun et al. carbides are precipitated at the grain boundaries of the alloys with increasing temperature, which also accelerates the intergranular corrosion and enlarges the diﬀerences in the corrosion resistance of various alloys. Acknowledgement This work was supported by the Strategically Leading Program of the Chinese Academy of Sciences (Grant No. XDA02040000). References  M. Mehos, C. Turchi, J. Vidal, M. Wagner, Z. Ma, C. Ho, W. Kolb, C. Andraka, A. Kruizenga, Concentraing Solar Power Gen3 Demonstration Roadmap, National Renewable Energy Laboratory, NREL/TP-5500-67464 (2017).  X.K. Xu, G. Dehghani, J.X. Ning, P.W. Li, Basic properties of eutectic chloride salts NaCl-KCl-ZnCl2 and NaCl-KCl-MgCl2 as HTFs and thermal storage media measured using simultaneous DSG-TGA, Sol. Energy 162 (2018) 431–441.  P.D. Myers Jr, D. Yogi Goswami, Thermal energy storage using chloride salts and their eutectics, Appl. Therm. Eng. 109 (2016) 889–900.  M. Sarvghad, S.D. Maher, D. Collard, M. Tassan, G. Will, T.A. Steinberg, Materials compatibility for the next generation of concentrated solar power plants, Energy Storage Mater. 14 (2018) 179–198.  J.C. Gomez-Vidal, R. Tirawat, Corrosion of alloys in a chloride molten salt (NaClLiCl) for solar thermal technologies, Sol. Energy Mater. Sol. Cells 157 (2016) 234–244.  A.M. Kruizenga, Corrosion Mechanisms in Chloride and Carbonate Salts, Sandia National Laboratories, Sandia Report, SAND 2012-7594 (2012).  K. Vignarooban, P. Pugazhendhi, C. Tucker, D. Gervasio, A.M. Kannan, Corrosion resistance of Hastelloys in molten metal-chloride heat-transfer ﬂuids for concentrating solar power applications, Sol. Energy 103 (2014) 62–69.  J.W. Ambrosek, Molten Chloride Salts for Heat Transfer in Nuclear Systems, PHD, University of Wisconsin-Madison, Madison, 2011.  M. Alkhamis, Stability of Metals in Molten Chloride Salt at 800°C. Master Thesis, University of Arizona, Arizona, 2016.  I.V. Oryshich, O.S. Kostyrko, Inﬂuence of molybdenum, tungsten, and cobalt on the corrosion of high-temperature strength nickel alloys in molten salts, Met. Sci. Heat Treat. 27 (1985) 740–746.  M. Hofmeister, L. Klein, H. Miran, R. Rettig, S. Virtanen, R.F. Singer, Corrosion behaviour of stainless steels and a single crystal superalloy in a ternary LiCl-KClCsCl molten salt, Corros. Sci. 90 (2015) 46–53.  B. Liu, X.L. Wei, W.L. Wang, J.F. Lu, J. Ding, Corrosion behavior of Ni-based alloys in molten NaCl-CaCl2-MgCl2 eutectic salt for concentrating solar power, Sol. Energy Mater. Sol. Cells 170 (2017) 77–86.  S.H. Cho, J.S. Zhang, Y.J. Shin, S.W. Park, H.S. Park, Corrosion behavior of Fe–Ni–Cr alloys in the molten salt of LiCl – Li2O at high temperature, J. Nucl. Mater. 325 (2004) 13–17.  H.S. Cho, J.W. Van Zee, S. Shimpalee, B.A. Tavakoli, J.W. Weidner, B.L. GarciaDiaz, M.J. Martinez-Rodriguez, L. Olson, J. Gray, Dimensionless analysis for predicting Fe-Ni-Cr alloy corrosion in molten salt systems for concentrated solar power systems, Corros. 72 (2016) 742–760.  J.W. Wang, C.Z. Zhang, Z.H. Li, H.X. Zhou, J.X. He, J.C. Yu, Corroison behavior of nickel-based superalloys in thermal storage medium of molten eutectic NaCl-MgCl2 in atmosphere, Sol. Energy Mater. Sol. Cells 164 (2017) 146–155.  A. Nishikata, H. Numata, T. Tsuru, Electrochemistry of molten salt corrosion, Mater. Sci. Eng. A 146 (1991) 15–31.  S. Delpech, C. Cabet, C. Slim, G.S. Picard, Molten ﬂuorides for nuclear applications, Mater. Today 13 (2010) 34–41.  S. Rouquette-Sanchez, G.S. Picard, Chalcogenide chemistry in molten salts. I. Selenium (IV) acido-basic and redox properties in the LiCl-KCl eutectic melt at 450, 500, 550 and 600°C, J. Electroanl. Chem. 572 (2004) 173–183.  Y.L. Wang, Q. Wang, H.J. Liu, C.L. Zeng, Eﬀects of the oxidants H2O and CrF3 on the corrosion of pure metals in molten (Li,Na,K)F, Corros. Sci. 103 (2016) 268–282.  I.N. Ozeryanaya, Corrosion of metals by molten salts in heat-treatment processes, Met. Sci. Heat Treat. 27 (1985) 184.  Y. Hosoya, T. Terai, T. Yoneoka, S. Tanaka, Compatibility of structural materials with molten chloride mixture at high temperature, J. Nucl. Mater. 248 (1997) 348–353.  A. Luo, D.L. Jacobson, R. Ponnappan, Compatibility of Inconel 617 alloy with LiFMgF2-KF thermal energy storage salts and vacuum at high temperature, J. Mater. Eng. Perform. 1 (1992) 755–762.  H.Q. Yin, J. Qiu, H.J. Liu, W.G. Liu, Y. Wang, Z.J. Fei, S.F. Zhao, X.H. An, J.H. Cheng, T. Chen, P. Zhang, G.J. Yu, L.D. Xie, Eﬀect of CrF3 on the corrosion behavior of Hastelloy-N and 316L stainless steel alloys in FLiNaK molten salt, Corros. Sci. 131 (2018) 355–364.  W.J. Ding, A. Bonk, J. Gussone, T. Bauer, Cyclic voltammetry for monitoring corrosive impurities in molten chlorides for thermal energy storage, Energy Procedia 135 (2017) 82–91.  W.J. Ding, H. Shi, Y.L. Xiu, A. Bonk, A. Weisenburger, A. Jianu, T. Bauer, Hot corrosion behavior of commercial alloys in thermal energy storage material of molten MgCl2/KCl/NaCl under inert atmosphere, Sol. Energy Mater. Sol. Cells 184 (2018) 22–30. Fig. 12. Diﬀusion coeﬃcient of Cr in Ni-based alloys at diﬀerent temperatures. Dgb is about 105 times larger than Dv at temperatures investigated according to Fig. 12, and therefore the corrosion occurs preferentially along grain boundaries. However, it must be noted that no obvious Cr depletion along the grain boundaries of the Ni-Mo-Cr alloys was detected at 600 °C (Fig. 4a–c). This behavior is probably due to the fact that the Cr diﬀusion distance (< 5 μm) is far smaller than the average grain size of the alloys (about 40∼50 μm). Only the Cr atoms in the near-surface region of Ni-based alloys can diﬀuse into the melt, which is a kind of short-circuit diﬀusion similar to the Cr diﬀusion along grain boundaries. Therefore, the Ni-Mo-Cr alloys show general corrosion at 600 °C. In contrast, the advantages of the grain boundary as a preferential diﬀusion path become more pronounced with the acceleration of corrosion and the increasement of Cr diﬀusion depth, as observed on Haynes 230 at all temperatures investigated (Fig. 4d and 5d) as well as the Ni-Mo-Cr alloys at 800 °C (Fig. 5a–c). Additionally, many carbides can be precipitated at the grain boundaries of various alloys with increasing temperature (Figs. 5 and 7). The carbides are rich in Cr versus the alloy base, and therefore the carbides can be dissolved preferentially, which also contributed to the shift of corrosion pattern from general corrosion to intergranular corrosion. Furthermore, the sensitivity of carbides on various alloys to molten chloride salt is diﬀerent, which also enlarges the diﬀerence in the corrosion resistance of the alloys at higher temperatures. 5. Conclusions (1) Hastelloy N, C276, C22 and Haynes 230 suﬀer corrosion in molten NaCl-KCl-MgCl2 under Ar atmosphere in the temperature range of 600–800 °C for 400 h. The corrosion of the alloys is mainly attributed to the selective dissolution of Cr to form soluble Cr chlorides into the melt. (2) The Ni-Mo-Cr alloys exhibit better corrosion resistance than the NiW-Cr alloy and the corrosion resistance of the Ni-Mo-Cr alloys increases with an increase of Mo and a decrease of Cr. These diﬀerences are mainly caused by the diﬀerent reactivity of formation of chlorides and the diﬀerent roles of the alloying elements in preventing the outward diﬀusion of Cr from the alloy base, as well as the alloying elements induced diﬀerences in the chemical compositions of carbides at the grain boundaries. (3) As temperature increases, the corrosion of Ni-based alloys is accelerated and the intergranular corrosion becomes more pronounced. Furthermore, the eﬀects of the alloying elements on the corrosion resistance of Ni-based alloys have a strong temperature dependence, which is attributed to the temperature induced the increasement of Cr diﬀusion coeﬃcient. Additionally, more 12 Corrosion Science xxx (xxxx) xxx–xxx H. Sun et al.  A. Zahs, M. Spiegel, H.J. Grabke, Chloridation and oxidation of iron, chromium, nickel and their alloys in chloridizing and oxidizing atmosoheres at 400-700°C, Corros. Sci. 42 (2000) 1093–1122.  Y. Kawahara, High temperature corrosion mechanisms and eﬀect of alloying elements for materials used in waste incineration environment, Corros. Sci. 44 (2002) 223–245.  C.N. Cao, Theory for Corrosion Electrochemistry, ﬁrst ed., Chemical Industry Press, Beijing, 1985.  R. Bender, M. Schutze, The role of alloying elements in commercial alloys for corrosion resistance in oxidizing-chloridizing atmosphere Part I: Literature evaluation and thermodynamic calculations on phase stabilities, Mater. Corros. 54 (2003) 567–586.  M.C. Galetz, B. Rammer, M. Schutze, Refractory metals and nickel in high temperature chlorine-containing environments-thermodynamic prediction of volatile     13 corrosion products and surface reaction mechanisms: a review, Mater. Corrros. 66 (2015) 1206–1214. Q. Wu, S.S. Li, Y. Ma, S.K. Gong, First principles calculations of alloying element diﬀusion coeﬃcients in Ni using the ﬁve-frequency model, Chin. Phys. B 21 (2012) 109102. J. Veverkova, A. Strang, G.R. Marchant, G.M. McColvin, H.V. Atkinson, High Temperature Microstructural Degradation of Haynes Alloy 230, TMS (The minerals, metals & Materials Society), 2008, pp. 479–488. F.Y. Quyang, C.H. Chang, J.J. Kai, Long-term corrosion behaviors of Hastelloy-N and Hastelloy-B3 in moisture-containg molten FLiNaK salt environments, J. Nucl. Mater. 446 (2014) 81–89. T.F. Chen, G.P. Tiwari, Y. Iijima, K. Yamauchi, Volume and grain boundary diﬀusion of chromium in Ni-base Ni-Cr-Fe alloys, Mater. Trans. 44 (2003) 40–46.