Effect of partial Ni substitution in V85Ni15 by Ti on microstructure, mechanical properties and hydrogen permeability of V-based BCC alloy membranes

Vanadium-based alloy membranes with body-centred-cubic (BCC) structure are considered as one of the leading alternatives to Pd-based alloys for hydrogen separation applications due to their lower cost and higher permeability. As permeability and mechanical properties depend on what microstructure can be produced mainly by alloy composition under same processing conditions, the effect of alloy composition on microstructure, mechanical properties and hydrogen permeability has been investigated for the V85Ni15 and V85Ni10Ti5 (at%) alloys prepared by a same process route. All Ni atoms dissolve into the V-matrix to form a single highly supersaturated solid solution with dendritic segregation of Ni-solute atoms in the binary alloy. A part of Ni replacement with 5 at% Ti leads to the formation of small interdendritic phases NiTi and NiTi2 in addition to major phase of V-based solid solution. The mechanical property testing shows that the ultimate strength of the ternary alloy is higher than that of the binary alloy, but the elongation and rollability are lower due to a combination of solid solution hardening and particle strengthening effect. The addition of Ti can greatly increase permeability about 4 times greater than the binary alloy at a permeation testing of 400 °C. But the presence of small amounts of interdendritic compounds provides a barrier to hydrogen migration, resulting in a relatively lower hydrogen diffusion coefficient. In theory, the diffusivity and solubility of hydrogen atom in the presence of alloying element Ti is higher than that in the presence of alloying element Ni in vanadium. This is demonstrated using first principles calculation which further explains the mechanism of hydrogen permeation.


Introduction
As a non-renewable energy, the utilization of fossil fuel energy produces a lot of pollution gases, which affects the air quality and people's health. Hydrogen energy is considered to be one of the most important new energy sources in the future due to its renewability and environmental friendliness. Hydrogen will be required much in the near future, especial ultra-high purity hydrogen, which mainly used in semiconductor and fuel cell industries [1]. The production and purification of H 2 is an important link of its application in industrial production [2][3][4][5]. Membrane separation of H 2 is considered as one of the most promising separation technologies because of its low cost and energy consumption, and high purity of separated H 2 [6][7][8]. The commercial application of Pd-Ag alloy membranes in the field of H 2 separation was realized in 1960s, but its large-scale application in membrane separation was limited due to the scarcity and price of Pd. For nonpalladium based alloy membranes, the permeability of VB group metals (V, Nb and Ta, etc) with BCC structure Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI. is 20-100 times higher than that of FCC structure metals (Pd, Cu and Ni, etc) [9]. However, H transport through pure V, Nb and Ta metals leads to lattice expansion and hydride formation due to high H solubility in V, Nb and Ta metals, resulting in hydrogen embrittlement. Alloying is one of the most effective methods to solve hydrogen embrittlement [10][11][12][13][14][15].
V-based solid solutions (Vss) with BCC structure can be formed with a wide range of alloy elements [16]. The addition of Nickel to Vanadium has been reported to increase resistance to embrittlement and maintain a higher permeability than commercially available Pd-Ag alloy membranes [17]. This finding makes V-Ni alloys highly attractive for membrane applications. Of these alloys, a composition of V 85 Ni 15 (at%) has been proposed as a promising candidate for hydrogen separation membranes. Partial substitution of Ni by Ti can result in a multiphase microstructure. A new alloy V 85 Ni 10 Ti 5 (at %) has been reported to be ∼4 times greater than V 85 Ni 15 in hydrogen permeability at 400°C [16]. This ternary alloy consisted of a primary BCC V-rich solid solution and small amount of secondary phase particles as evidenced by SEM. Unfortunately, the identification of secondary phase has not been determined since then. In the present work, we used EBSD Kikuchi patterns and micro-XRD to determine the types of all secondary-phase compounds in the V-matrix interdendritic region, and also established a relationship of the multi-phase microstructure with mechanical properties for the development of high-strength alloy membranes. In order to investigate the rollability of thinner sheet alloy membrane fabrication, cold rolling was carried out on these two alloys to allow a higher hydrogen flux. In addition, based on the permeability and solubility data obtained, and combined with the first principle calculation, we gain an insight into the understanding of the effect of Ti which is opposite to Ni on the solubility and diffusivity of hydrogen in vanadium. This is very helpful to design a new composition in the V-Ni-Ti ternary system for the development of both high-permeability and high embrittlement-resistant alloy membranes.

Experimental procedure 2.1. Alloy preparation
The V 85 Ni 15 and V 85 Ni 10 Ti 5 (at%, nominal composition) samples were made from high purity (99.9%) Vanadium, Nickel and Titanium metals. Vanadium, Nickel and Titanium pellets were mixed and melted on a water-cooled copper hearth using a vacuum/argon arc melter. The melting chamber was evacuated and then high-purity argon gas was backfilled to ensure the ingots not to be oxidized. In order to ensure compositional homogeneity, the arc-melted (AM) ingots (total weight 60 g for each) were melted several times.

Hardness and tensile testing, and rolling
The samples for testing properties and observing microstructures were cut using electrical discharge machining (EDM) from the arc-melted ingots. Vickers hardness of both samples were measured using a digital microhardness tester with loads of 500 g and a dwell time of 10 s. The final micro-hardness was the average value of 10 measurements performed on the samples surface. The samples surfaces for measurement were ground and then polished using diamond pastes, followed by etching in solution containing HF and HNO 3 . Tensile specimens were sectioned from the arc-melted ingots with gauge length of 10 mm, width of 2 mm and thickness of 1.5 mm. Tensile tests were carried out on a tensile machine at the room temperature and constant speed of 0.01 mm s −1 .
Rolling process was carried out on a 120-mm diameter rolling mill with a roller speed of 0.1 ms −1 at the room temperature. The initial thickness of the samples is 1.5 mm. Ultimate reduction rate of the first pass rolling (r up ) was used to evaluate the rollability of V 85 Ni 15 and V 85 Ni 10 Ti 5 alloys. The maximum reduction of the first pass rolling was measured using the method referring to [18]. The ultimate reduction rate of the first pass rolling (r up ) is defined as follows. where t 0 and t are the thickness of the AM sample and the rolled sample with the maximum reduction of the first pass rolling, respectively.

Microstructural analysis
The microstructures of the alloy samples were observed in back-scattered electron (BSE) mode using a Quanta 400 field emission gun scanning electron micro-scope (FEG-SEM) equipped with x-ray Energy Dispersive Spectroscopy (EDS). Secondary electron (SE) imaging mode in the SEM was also employed for fracture surface morphology observation. A JEOL 8500F electron microprobe was employed for quantitative composition microanalysis of different areas both in V 85 Ni 15 and V 85 Ni 10 Ti 5 alloys. The intensity measurements were conducted at 12 kV accelerating voltage and a beam current of 30 nA.
X-ray diffraction (XRD) patterns were obtained by a Bruker D8 Advance x-ray Diffractometer employing CuKα radiation (40 kV, 40 mA) and a graphite (002) monochromator. The samples were scanned over the 2-theta range 30°to 100°with a step size of 0.02°and a count time of 4 s per step. Bruker XRD search match program EVA was used to analyze the collected XRD data. Crystalline phases were identified from the powder diffraction database (ICDD-JCPDS). For lattice parameter measurements, the Quartz (SiO 2 ) was used as an external standard for the XRD instrument alignment. The whole diffraction patterns corresponding to the V-phase solid solution were fitted using the Bruker TOPAS software to obtain the lattice parameters.
Minor phases in V 85 Ni 10 Ti 5 alloy which levels are below the detection limit of bulk XRD were identified by Micro-XRD. The Micro-XRD patterns were obtained using a Bruker General Area Detection Diffraction (GADDS) x-ray micro-diffractometer equipped with a HiStar area detector, CuKα radiation (40 kV, 40 mA), and cross-coupled Goebel monochromating mirrors. A pinhole collimator was employed to collimate the x-ray beam to a 300 μm spot with a count time of 300 s. Position-sensitive proportional counter (PSPC) was used to collect two-dimensional intensity patterns which were converted to conventional XRD patterns.
The presence of minor phases in V 85 Ni 10 Ti 5 alloy was confirmed by electron back-scatter diffraction (EBSD) technique. A JEOL 7001F FEG-SEM was used to collect Kikuchi patterns under specimen inclination of 70°and operating at 20 kV. Oxford Instruments Channel 5 HKL software was employed to index the acquired EBSD patterns for phase determination.

Permeation and solubility testing
A constant pressure method reported elsewhere [16] was used to measure hydrogen permeability of the samples. Both sides of the samples for the testing were chemically cleaned and radio frequency (RF) sputtered with a Pd layer of 500 nm to promote H 2 dissociation. The samples were sealed in a custom-designed apparatus using circular copper gaskets (internal diameter of 16 mm). The feed-side was supplied with a mixture of 90% H 2 +10% CO 2 at a pressure of 4.0 bar (a) and 200 ml min −1 during hydrogen permeation testing. The permeate side was supplied with argon stream at 200 ml min −1 . The hydrogen concentration in the Ar stream was determined using a gas chromatograph, and hydrogen flux across the membrane was calculated from this resulting concentration. Detection of CO 2 in the permeate stream was confirmed the integrity of the membrane. Permeability was calculated as the gradient of hydrogen flux against ( -P P 1 0.5 2 0.5 ), where P 1 and P 2 are the feedside pressure and permeate-side pressure, respectively.
The hydrogen uptake of two alloys with varying hydrogen pressure at 400°C and up to 12 bar (a) was measured by the Sieverts' technology to determine solubility. Only the data of the increasing pressure cycle is showed due to little difference between the increasing and decreasing pressure cycles.

Microstructure
The BSE-SEM images of the microstructures of arc-melted V 85 Ni 15 and V 85 Ni 10 Ti 5 alloys were showed in figure 1. Coarse grains were observed and no secondary phase was present in the arc-melted V 85 Ni 15 alloy (as shown in figure 1(a)). Combined with the XRD results (figure 2), it is clear that the arc-melted V 85 Ni 15 alloy is a single-phase Vanadium-based solid solution (Vss). It should be noted that dendritic segregation of Ni-solute atoms in Vss of the arc-melted V 85 Ni 15 alloy has been observed, suggesting compositional inhomogeneity of elemental Ni distribution. It can be seen from the V-Ni phase diagram in figure 3 [19], the maximum solubility of Ni in the vanadium solid solution is 24 at% at peritectic temperature of 1280°C and shows a pronounced temperature dependence decreasing to 10 at% at peritectoid temperature of 900°C and around 2 at% at 200°C. For a given composition of 15 at% Ni located between 10 and 24 at%, the resulting equilibrium microstructure should be a two-phase mixture of Vss and NiV 3 compound below the peritectoid temperature. However, the arc-melting process belongs to non-equilibrium rapid solidification, and thus prevents the kinetically slow precipitation of NiV 3 compound due to rapid cooling, forming a highly supersaturated solid solution with dendritic segregation of Ni-solute atoms.
Substitution of Ni by 5 at% Ti can result in a multi-phase microstructure. As shown in figure 1(b), primary Vss phase in the as-cast V 85 Ni 10 Ti 5 alloy has a dendritic microstructural feature with small amounts of secondary phases formed in the interdendritic regions. Pores at the grain boundaries and between dendrites were formed by irregular shrinkage during solidification of the alloy, but these are discontinuous and do not result in leakage of gases across the alloy membrane. The presence of minor secondary phases in the interdendritic regions was not detected by the conventional XRD instrumentation which provides bulky information about the structure of crystalline materials (see figure 2). However, the micro-XRD patterns detected from three different spots by the position-sensitive proportional counter reveals the presence of NiTi 2  compound in the cast V 85 Ni 10 Ti 5 alloy (as shown in figure 4). Actually, there are two types of intermetallic compounds as shown in the higher magnification SEM microstructure ( figure 5(a)). It can be seen in more details that the interdendritic region mainly consisted of white-phase and dark-phase compounds in addition to primary V-rich BCC solid solution. The EDS was used in semi-quantitative mode to determine chemical composition of these phases observed in the SEM microstructure, as shown in figure 6. The energy peaks correspond to the various elements in the three phases, and their peak intensity indicate that the white-phase compound contains more Ni and Ti, whereas the dark-phase particles is a Ti-rich intermetallic compound. The EDS results are correlated with the crystallographic structure of each of three phases measured by electron   Table 2 compares the hardness from bulk (500 g load) and Vss (100 g load) in the V 85 Ni 15 and V 85 Ni 10 Ti 5 alloys, bulk hardness is equal to Vss hardness of single-phase V 85 Ni 15 alloy. The bulk and Vss hardness of V 85 Ni 15 alloy are lower than that of V 85 Ni 10 Ti 5 alloy, and Vss hardness of V 85 Ni 10 Ti 5 alloy is lower than that of the bulk. Atomic size difference between Ti and V is twice larger than that between Ni and V, Ti addition generates more distortion of the Vss lattice, resulting in a higher Vss hardness of V 85 Ni 10 Ti 5 alloy than that of V 85 Ni 15 alloy. The bulk hardness of V 85 Ni 10 Ti 5 alloy is higer than Vss hardness due to hard and brittle NiTi 2 phases. The difference of the bulk and Vss hardness of V 85 Ni 10 Ti 5 alloy is small, which suggests that the Vss is a major factor controlling the change in hardness.

Tensile properties
The ultimate strength and elongation of V 85 Ni 15 and V 85 Ni 10 Ti 5 alloys obtained are shown in table 3. From  table 3, it can be seen that the ultimate strength of V 85 Ni 10 Ti 5 alloy is higher than that of V 85 Ni 15 alloy, but the elongation is lower. It should be noticed that the elongations of both alloys are lower than 15%, which refer to supersaturated solid solution of Vss due to rapid cooling. The fracture surfaces of the specimens are shown in figure 7. As can be seen, a number of uneven size dimples appear on the fracture of V 85 Ni 15 alloy ( figure 7(a)) and local areas with no dimple are relatively smooth, while the fracture of V 85 Ni 10 Ti 5 alloy ( figure 7(b)) is rough and a number of extremely shallow dimples were observed, which is consistent with poor plastic (low elongation). High hardness Vss and hard-brittle NiTi 2 in V 85 Ni 10 Ti 5 alloy are predominant factor controlling the tensile properties. Table 4 shows the comparison of ultimate reduction rate of the first pass rolling (r up ) for V 85 Ni 15 and V 85 Ni 10 Ti 5 alloys at room temperature. It can be noted that the rollability of single-phase V 85 Ni 15 alloy is higher than that of multiphase V 85 Ni 10 Ti 5 alloy, and the r up increases about 21%. The Vss and NiTi 2 phases in the V 85 Ni 10 Ti 5 alloy make the alloy has high hardness and low plastic. The interface between the Vss and NiTi 2 phase in V 85 Ni 10 Ti 5 alloy is difficult to coordinate for deformation during rolling, and therefore the stress concentration occurs easily at the interface after dislocations movement is obstructed, resulting in forming of the cracks.   [16], while the substitution of partial Ni by 5 at% Ti increases this value to 9.30×10 −8 . The role of Ti in increasing the permeability of V is clearly demonstrated by contrast with the role of Ni. The hydrogen permeability of multiphase V-Ni-Ti alloys with a higher proportion of intermetallic compounds in compositions have been examined previously by other groups [14], and the microstructure of the alloys consist of discrete BCC phases within the Ni-Ti compounds which greatly affected the hydrogen permeation. Large amounts of Ni-Ti compounds are the barriers to H migration. In contrast, our V 85 Ni 10 Ti 5 alloy contains a large amount of BCC matrix as a major phase (see figure 1(b)) for hydrogen diffusion and thus the hydrogen permeability of this alloy is vastly greater than that reported in [14]. That is to say the BCC V-rich phase is a predominant factor controlling the hydrogen permeability.    The relationship of hydrogen solubility (expressed as H/M) with pressure for pure V, V 85 Ni 15 and V 85 Ni 10 Ti 5 alloys at 400°C was established as displayed in figure 8. It is clearly shown that alloying V with Ni significantly reduced the hydrogen solubility, while Ti increased the hydrogen solubility in the alloy. The diffusion coefficients for H through pure V, V 85 Ni 15 and V 85 Ni 10 Ti 5 alloys at 400°C were calculated by combining the permeability and solubility data, and listed in table 6 for comparison. Pure V displayed the highest hydrogen diffusion coefficient (1.20×10 −8 at 400°C). The addition of 15at% Ni reduces this value to 0.96×10 −8 . The addition of the third element Ti further reduces the diffusion coefficient to 0.96×10 −8 .

Rollability
The intrinsic permeability of the alloy involves both hydrogen solubility and diffusivity. Since P=SD, where P, S and D are permeability, solubility coefficient and diffusion coefficient, respectively, high permeability can be reached by increasing either solubility or diffusion. In order to better understand the effect of alloying elements with different atomic size on solubility and diffusion of hydrogen in the V-matrix solid solution, the obtained data about the solution energy (E sol ) and diffusion energy barrier (E a ) of hydrogen in pure V, V-Ni and V-Ti using the first-principle calculation [22] were used as a measure of the ease of solubility and diffusion (see table 7). The more negative E sol means more H dissolution into the V-matrix solid solution, and is favorable to reach a higher H solubility. The solution energy of H in pure V was calculated to be −0.218 eV, whereas the presence of Ni in V increased the hydrogen solution energy to a value of −0.107 eV, which is about 2 times greater than that of H in pure V, making the hydrogen dissolution more difficult. For the V-Ti system alloy, the hydrogen solution energy (−0.294 eV) is lower than that of H in pure V. The more negative value indicates that    According to the calculated diffusion energy barrier data, the E a value of H in V-Ni alloy is higher than that of H in pure V, which suggests that the diffusion rate of hydrogen through the V-Ni (Ni-induced V lattice contraction) is slower than through pure V. This theoretical calculation is consistent with the experimental results obtained from pure V and V 85 Ni 15 alloy. Compared to pure V and V-Ni, the V-Ti alloy has a relatively lower E a value of hydrogen. This means the presence of Ti in V increases hydrogen transport through the V-Ti solid solution with larger lattice constant than pure V. The dissolution of Ti into the host V 85 Ni 15 BCC alloy would therefore be expected to obtain a higher diffusion coefficient of hydrogen though the major V-Ni-Ti solid solution (V 92.1 Ni 5.1 Ti 2.8 ) phase in the V 85 Ni 10 Ti 5 alloy than through the V 85 Ni 15 alloy. In fact, our result is conflict with the expectation. This should be attributed to the formation of non-BCC hydrogen diffusion phases such as NiTi and NiTi 2 compounds in the multi-phase microstructure. These compounds act as barriers to inhibit hydrogen transport, resulting in the reduction of hydrogen diffusion coefficient through the bulk alloy. Further work needs to entirely make the BCC single-phase alloy of the V 92.1 Ni 5.1 Ti 2.8 composition in order to verify the first-principle calculation of the V-Ni-Ti ternary BCC structure.
The solubility and diffusion coefficient of hydrogen are both necessary to increase in order to increase the hydrogen permeability through a given alloy membrane. However, excessive hydrogen solubility will lead to hydrogen embrittlement and make the degradation of mechanical properties [23]. As increasing diffusivity does not induce a mechanical penalty, therefore, it is desirable to develop an alloy with large D and low H/M. Our results showed that although partial substitution of Ni with 5 at% Ti increased permeability mainly due to increasing the hydrogen solubility in the alloy, the diffusion coefficient value is higher than that of Pd (∼0.55×10 −8 m 2 s −1 at 400°C [24]). In order to better develop the promising vanadium alloys for H 2 selective membrane application, further work to maximizing hydrogen diffusivity needs to construct a multi-component model against the experimental results using the first-principle calculation. This type modelling would help us to understand how two or more alloying elements together exert a synergistic effect on solubility and diffusion behavior of hydrogen in the multi-component BCC alloys.

Conclusion
The effect of alloying elements Ni and Ti on microstructure, mechanical properties and hydrogen permeability has been investigated for both V 85 Ni 15 and V 85 Ni 10 Ti 5 alloys prepared by the same process route. The conclusions can be drawn as follows: (1) All Ni atoms dissolve into the V-matrix to form a single highly supersaturated BCC solid solution with dendritic segregation of Ni-solute atoms in V 85 Ni 15 alloy. The addition of Ti into the base V 85 Ni 15 alloy results in the formation of multi-phase microstructure consisting of primary BCC V-matrix solid solution and small amounts of intermetallic compounds. These compounds are visualised and identified as NiTi and NiTi 2 using EBSD technique providing simultaneous information of crystal structure and chemical composition for excellent phase identification.
(2) The mechanical property testing shows that the ultimate strength of the V 85 Ni 10 Ti 5 alloy is higher than that of the V 85 Ni 15 alloy, but the elongation and rollability are lower due to a combination of solid solution hardening and particle strengthening effect. The ultimate reduction rate of the first pass rolling (r up ) for both V 85 Ni 15 and V 85 Ni 10 Ti 5 alloys at room temperature are above 35%, showing good formability to allow for the fabrication of thinner membranes for high-efficiency hydrogen permeation. Table 7. Calculated lattice constant and hydrogen transport parameters of pure V, V-Ni and V-Ti alloys using first-principles method [22].

Alloy Lattice constant (Å) Solution energy (eV)
Diffusion energy barrier (Ea) at 673 K (eV) (3) The major V-Ni-Ti solid solution (V 92.1 Ni 5.1 Ti 2.8 ) phase in the V 85 Ni 10 Ti 5 alloy can greatly increase hydrogen permeability and solubility, but doesn't provide larger hydrogen diffusion coefficient than the V 85 Ni 15 single-phase alloy, even though the first-principle calculation demonstrates that alloying element Ti has both lower solution energy and diffusion energy barrier of hydrogen in V than alloying element Ni. The main reason should be attributed to the formation of non-BCC hydrogen diffusion phases NiTi and NiTi 2 compounds in the multi-phase microstructure. These compounds act as barriers to inhibit hydrogen transport through the bulk alloy.