Design of (Nb, Mo)40Ti30Ni30 alloy membranes for combined enhancement of hydrogen permeability and embrittlement resistance

The effect of substitution of Nb by Mo in Nb40Ti30Ni30 was investigated with respect to microstructural features and hydrogen dissolution, diffusion and permeation. As-cast Nb40−xMoxTi30Ni30 (x = 0, 5, 10) alloys consist of primary bcc-Nb phase and binary eutectic (bcc-Nb + B2-TiNi). The substitution of Nb by Mo reduces the hydrogen solubility in alloys, but may increase (x = 5) or decrease (x = 10) the apparent hydrogen diffusivity and permeability. As-cast Nb35Mo5Ti30Ni30 exhibits a combined enhancement of hydrogen permeability and embrittlement resistance as compared to Nb40Ti30Ni30. This work confirms that Mo is a desirable alloying element in Nb that can contribute to a reduction in hydrogen absorption and an increase in intrinsic hydrogen diffusion, thus improving embrittlement resistance with minimal permeability penalty.

The mass production of high purity hydrogen is one of the central topics of hydrogen economy, which has attracted much attention in recent years. Membrane separation is the benchmark technology used to separate and purify hydrogen from a mixed or contaminated gas, typically produced by methane steam reforming [1][2][3][4][5] or coal gasification [6][7][8] . Foremost among the possible choices of metallic membrane materials are Pd-based alloys due to their high catalytic surface property, reasonable hydrogen permeability, good durability and tolerance to syngas species. It was reported in refs 9 and 10 that a nominal hydrogen production capacity of 40 Nm 3 /h and an energy efficiency of 70-76% could be reached by a modern steam reforming system with Pd-based alloy membranes. Recently, Pd-based alloy membranes were extended to produce ultra-pure hydrogen by catalytic reforming of olive mill wastewater (OMW) 11,12 . OMW is a biomass by-product of the olive oil industry, which results in environmental issues because of its poor biodegradability and high phytotoxicity. A treatment of OMW in a Pd-Ag membrane reactor demonstrated the capability to produce up to 3.25 kg of ultra-pure hydrogen per ton of OMW 11 . A higher capability in hydrogen production of 12.3 kg per ton of OMW was reached by catalytic reforming of the combined OMW and methane in a Pd-Ag membrane reactor 12 . However, the cost of Pd is uneconomically high and subject to dramatic changes, which restricts its large scale application. This stimulates the development of pinhole-free Pd-based alloy membranes with ever decreasing thickness 13 , or the development of new less expensive metal membranes.
Because of higher hydrogen permeability but much lower cost and richer natural resources, group V metals, such as vanadium (V), niobium (Nb) and tantalum (Ta) are among the leading candidates to replace Pd-based alloys for hydrogen separation. In pure form, they are sensitive to hydrogen embrittlement (HE) induced by excessive hydrogen absorption. Much of current research is focused on compositional or/and structural modification to inhibit HE. This generates mainly two types of notable hydrogen permeable alloys. The first type comprises V 14-20 , Nb [21][22][23][24] and Ta 25 solid solution alloys formed by selective alloying for group V metals with the aim to reduce hydrogen solubility. Such alloys exhibit an obviously improved HE resistance as compared to the pure counterpart, and also show distinctly higher hydrogen permeability than the leading Pd-based alloys. The other type comprises dual-phase alloys such as Nb-Ti-Ni 26-31 , Nb-Ti-Co 32-36 , V-Ti-Ni 37-39 , Ta-Ti-Ni 40, 41 etc. These alloys generally consist of primary bcc-Nb/V/Ta for hydrogen permeation and secondary eutectic (bcc-Nb/V/ Ta + B2-TiNi/TiCo) with high HE resistance. After pertinent modulation of the dual-phase structure, these alloys achieve an excellent balance between hydrogen permeability and embrittlement resistance.
Overall, there still remain several challenges for the above mentioned V, Nb and Ta based alloys to meet the US DOE performance targets 42,43 for hydrogen separation and purification applications. Meeting hydrogen permeation flux and durability targets are the key issues to be solved for these materials. One strategy is compositional modification for tailoring hydrogen solubility and diffusivity. The desirable effect is a reduction in hydrogen solubility and an increase in hydrogen diffusivity, thus improving the HE resistance with no or minimal permeability penalty. However, it was found that the hydrogen diffusion through V-Ni 44 , V-Ni-Ti 45 , Nb-Ti-Co 36 etc. exhibits a significantly concentration dependent behavior. Decreasing hydrogen solubility is generally coupled with a reduced hydrogen diffusivity. This is in contradiction with the above mentioned aim of compositional modification.
Recently, Yukawa et al. 46,47 reported that the addition of W and Mo in Nb could reduce hydrogen solubility and simultaneously increase hydrogen diffusivity. They presented an Nb-8W-8Mo alloy membrane that shows significantly enhanced HE resistance and hydrogen flux, which is, however, only applicable in a limited hydrogen pressure range. This inspired us to investigate whether there exists a similar effect on hydrogen solubility and diffusivity for the substitution of Nb by Mo in Nb 40 Ti 30 Ni 30 (all compositions in atomic percent). Nb 40 Ti 30 Ni 30 is an attractive hydrogen permeable alloy in dual-phase Nb-Ti-Ni systems 26 . It exhibits strong HE resistance in a much larger hydrogen pressure range than that for Nb-based solid solution alloys. However, the hydrogen permeability of Nb 40 Ti 30 Ni 30 is not quite satisfactory for the practical applications 48 . In this work, the substitution of Nb by Mo in Nb 40 Ti 30 Ni 30 was investigated with respect to microstructural features and hydrogen dissolution, diffusion and permeation. It is expected that this compositional modification may improve the permeability and also enhance the HE resistance.

Results and Discussion
Microstructure. X-ray diffraction (XRD) patterns of as-cast Nb 40−x Mo x Ti 30 Ni 30 (x = 0, 5, 10) are shown in The microstructure of as-cast Nb 40−x Mo x Ti 30 Ni 30 (x = 0, 5, 10) is shown in Fig. 2. The white regions represent bcc-Nb and the dark regions represent B2-TiNi. Typical hypoeutectic microstructures, i.e. primary bcc-Nb dendrites surrounded by eutectic (bcc-Nb + B2-TiNi), are observed in all three as-cast alloys, as shown in Fig. 2(a,c and e). The substitution of Nb by Mo does not lead to a change in the microstructural sequence, but induces a change in growth morphology of bcc-Nb dendrites with less developed side branches. This may be attributed to the partial remelting of the primary bcc-Nb phase due to a retrograde solidus surface of the primary phase. In addition, the substitution of Nb by Mo results in a decrease in volume fraction of bcc-Nb in the eutectic, and a gradual increase in fraction of B2-TiNi, as seen in Fig. 2(b,d and f). Correspondingly, the eutectic structure changes from a typical lamellar ( Fig. 2(b)) to a lamellar/degenerate ( Fig. 2(e))) to a fully degenerate pattern ( Fig. 2(f)). Hydrogen is mainly dissolved in the bcc-Nb of the dual-phase alloys, and Nb is one of the typical exothermic occluders 43 whose hydrogen solubility clearly decreases with increasing temperature. Schmidt et al. 49 reported that TiNi also absorbs hydrogen exothermically, but the amount of absorbed hydrogen is negligible in the present pressure and temperature conditions. From Fig. 3, it can also be seen that the substitution of Nb by Mo decreases the hydrogen solubility. This should be associated with the change in the elastic properties and the electronic structure induced by the substitutional Mo atoms in the host bcc-Nb lattice. The dissolution of Mo in Nb contracts the host bcc-Nb lattice (see Fig. 1), and a larger energy for elastic deformation is required for the absorption of H atoms. This corresponds to the interstitial site blocking mechanism 19 that reduces hydrogen solubility (for an example in a V-Pd alloy, see ref. 19). The electronic structure is another important factor in establishing hydrogen solubility in metals. The electronegativity of Mo (2.16) is higher than that of Nb (1.6), but both values are lower than that of hydrogen (2.2). Hydrogen gains charge from each of the nearest neighbor metal atoms in the bcc-Nb lattice 50 . It can be expected that less charge is shared between the lattice and hydrogen when the neighboring Nb atoms are replaced by Mo. Correspondingly, the binding energy of hydrogen with the bcc-Nb lattice is reduced by the substitution of Nb by Mo. This inherently contributes to the reduction in hydrogen solubility.
Hydrogen flux and permeability. In Fig. 4, the measured hydrogen permeation flux (J') as a function of time (t) is shown for the as-cast Nb 40 Ti 30 Ni 30 at 673 K as a representative case. The flux evidently increases with increasing the upstream pressure (P u ). At each fixed P u , the flux is essentially stable after a short initial transient. This stable average flux is referred to as steady-state flux (J) under the given pressure and temperature conditions. Correspondingly, the relation between the J and P u can be built, considering that the downstream hydrogen pressure (P d ) is permanently maintained at 0.1 MPa.
In Fig. 5, the steady-state hydrogen permeation fluxes (J) at 523…673 K for membranes of as-cast Nb 40    all temperatures and pressures, which were testified by the permeation test using Ar at room temperature. This indicates that all alloys are not susceptible to HE.
In earlier work on Nb-based dual-phase alloy membranes [26][27][28][29][30][31][32][33][34][35][36] , the hydrogen flux (J) was generally transformed to permeability (Φ) (in mol H 2 m −1 s −1 Pa −0.5 ) using the relationship Φ = J·d/ΔP 0.5 . In Fig. 6(a-c), the calculated hydrogen permeabilities of the as-cast Nb 40−x Mo x Ti 30 Ni 30 (x = 0, 5, 10) alloys are plotted against the upstream hydrogen pressure. The Φ values of each membrane evidently increase with increasing temperature. At a given temperature, Φ is essentially constant at a P u of 0.2…0.4 MPa for all membranes, but gradually exhibits pressure dependence at higher P u values (>0.4 MPa), decreasing with increasing P u . The constant Φ value is attributed to the linear relation between C and P 0.5 (i.e. C = K · P 0.5 + α) and the linear relation between J·d and ΔP 0.5 in the pressure range of 0.1…0.4 MPa at a given temperature. This is commonly found in Nb-based dual-phase alloy membranes 27,31,[34][35][36] . The temperature dependence of Φ in the Arrhenius plot follows a straight line for each membrane in the low hydrogen pressure range, as shown in Fig. 6(d). The Φ values of as-cast Nb 40 Ti 30 Ni 30 are slightly higher than those of pure Pd at high temperatures. The substitution of Nb by Mo increases the Φ values for 5 at.% Mo, but distinctly reduces the values for 10 at.% Mo. The as-cast Nb 35 Mo 5 Ti 30 Ni 30 exhibits an attractive hydrogen permeability, particularly 3.13 × 10 −8 mol H 2 m −1 s −1 Pa −0.5 at 673 K. This is ~1.96 times that of pure Pd. At higher P u values (>0.4 MPa), the decrease of Φ with increasing P u indicates a deviation from ideal Sieverts'-type behavior, which is expected from the hydrogen absorption data.
Hydrogen diffusion. The above results clearly show that the hydrogen permeability of (Nb, Mo) 40 Ti 30 Ni 30 alloys is sensitive to the Mo content. This suggests that the substitution of Nb by Mo also induces different hydrogen diffusion behavior in alloys. The apparent hydrogen diffusion coefficient can be calculated by D = J·d/ΔC according to Fick's first law. Here, ΔC is the difference of hydrogen concentration between the upstream and downstream surfaces of the membrane. D is the concentration dependent diffusion coefficient that can be further associated with the concentration independent diffusion coefficient (D*): T

1/2
Here, f(r) is the thermodynamic factor that can be calculated by the hydrogen absorption data in Fig. 3, as described previously 36 . The concentration dependent apparent diffusion coefficient (D) takes into account the chemical potential gradient of hydrogen as the thermodynamic driving force for diffusion, which reveals the average/overall diffusion behavior of hydrogen in membranes. The concentration independent hydrogen diffusion coefficient (D*) is known as Einstein's diffusion coefficient or intrinsic diffusion coefficient. D* is strictly only applicable in infinitely dilute systems with a random component in the movement of hydrogen atoms, which accounts for the interaction of hydrogen and the metal lattice. Dolan et al. 44 first put forward a method of calculating concentration dependent and concentration independent hydrogen diffusion coefficients of V-Ni alloy membranes by correlating hydrogen absorption and flux data. This method was extended to evaluate hydrogen diffusion behavior of several other V-based hydrogen permeable alloys 51 and also of the present (Nb, Mo) 40 Ti 30 Ni 30 alloys.
In Fig. 7, the concentration dependent D for as-cast Nb 40−x Mo x Ti 30 Ni 30 (x = 0, 5, 10) is shown. The D values for all membranes increase with increasing temperature. At each temperature, the D values obviously increase with increasing hydrogen concentration (r). This originates from the change in the electronic structure of bcc-Nb with the gradual addition of H. It was found that the d bands of Nb were modified by the interaction with H, resulting in the appearance of H-induced new energy states 52 . This potentially reduces the activation energy for interstitial hopping of H in the bcc-Nb lattice. In addition, the mobility of hydrogen in the metal lattice is sensitive to the hydrogen concentration in terms of the self-trapping mechanism 53 . At low hydrogen concentrations, the relaxation of the lattice tends to be complete, and hydrogen mainly resides at deep potential sites. As the hydrogen concentration increases, lattice relaxation turns to be less advanced, implying that the average hydrogen potential becomes less deep, so that hydrogen can overcome the potential barrier more easily. Thus, the mobility of hydrogen increases with hydrogen concentration. The reduction in the diffusion activation energy and the increase in mobility altogether contribute to the gradual increase of D with r.
From Fig. 7, it can also be seen that the substitution of Nb by Mo increases the apparent hydrogen diffusion coefficient (D) for 5 at.% Mo, but reduces the values for 10 at.% Mo. This should be associated with the responses of D* and f(r) to the substitution according to equation (1). Fig. 8 shows the concentration independent D* for the three alloy membranes at 523…673 K. The D* values are almost constant at a fixed temperature for each membrane, but there is a slight trend toward higher D* values with increasing r. The average D* in each case follows a linear temperature dependence in the Arrhenius plot, as shown in Fig. 9. The activation energy (Q) of hydrogen diffusion for a dilute hydrogen concentration, and the diffusion pre-exponential factor (D 0 ) can be determined. From Fig. 9 it is concluded that the addition of Mo in Nb increases the average D*, in a more pronounced manner for x = 5 than for x = 10. This particular change mainly originates from the difference in the D 0 values, since there is no significant difference in the Q values for the three alloys. D 0 can be expressed as , where λ, Z, v, S B , S W and k are jump width, correlation number, jump frequency, entropy of vacancy formation, entropy of vacancy migration and Boltzmann's constant, respectively. The substitution of Nb by Mo must induce diverse effects on the parameters in the expression of D 0 , such that the D 0 and

Hydrogen embrittlement (HE). The hydrogen permeation tests in Figs 4 and 5 indicate that as-cast Nb 40−
x Mo x Ti 30 Ni 30 (x = 0, 5, 10) alloy membranes exhibit an appropriate tolerance to HE at 523…673 K. A further evaluation of HE sensitivity was conducted by hydrogen permeation under controlled cooling conditions. Such evaluations have technical relevance, particularly concerning membranes during sustained operation with intermittent thermal cycling. Figure 10 shows the measured hydrogen flux as a function of temperature for the membranes. The flux decreases with decreasing temperature in all cases. An abrupt increase in flux is observed for Nb 40  This indicates that the substitution of Nb by Mo improves the HE resistance, which can straightforwardly be attributed to the decrasing hydrogen solubility (see Fig. 3).
It can be expected that the three alloy membranes investigated in the present study are robust under the operating conditions representative for industrial applications such as water gas shift reaction (typically at 0.7…1 MPa and 300…500 °C). However, hydrogen should be evacuated from the membrane at temperatures above the critical fracture temperatures mentioned above. Although the substitution of Nb by Mo cannot enable the membranes to be cycled to room temperature under H 2 atmosphere, it distinctly lowers the temperature to which they can be cycled. Overall, Nb 35 Mo 5 Ti 30 Ni 30 exhibits the best balance between hydrogen permeability and HE resistance. This alloy thus has high potential to serve as non-Pd alloy membrane for hydrogen separation.

Conclusions
Nb 40 Ti 30 Ni 30 is one of the competitive alternatives to Pd-based alloy membranes for hydrogen separation and purification, while further enhancement of hydrogen permeability and embrittlement resistance is necessary to ensure its practical application. A promising strategy is compositional modification with the aim of controlling the hydrogen solubility to a point where embrittlement is inhibited, while the hydrogen diffusivity is increased. The present work shows that the substitution of Nb by Mo in Nb 40 Ti 30 Ni 30 does not lead to a change in the microstructural evolution sequence, while obviously hydrogen solubility is reduced. The apparent hydrogen diffusivity and permeability increase with small Mo additions, but decease with larger ones. This can be attributed to the sensitive response of intrinsic hydrogen diffusivity and thermodynamic factor to the Mo content. Overall, Nb 35 Mo 5 Ti 30 Ni 30 exhibits a combined enhancement of hydrogen permeability and embrittlement resistance as compared to Nb 40 Ti 30 Ni 30 . Mo is a desirable alloying element in Nb for an excellent balance of permeability and embrittlement resistance.

Experimental
Samples. About 40 g ingots of Nb 40−x Mo x Ti 30 Ni 30 (x = 0, 5, 10) alloys were prepared by an arc melting method in purified Ar atmosphere using Nb, Mo, Ti and Ni (99.99 mass% purity for all) as raw materials. All ingots were melted 6 times for compositional homogeneity. Phase identification of the as-cast samples was conducted by X-ray diffraction (XRD). Microstructural observations were carried out using scanning electron microscopy (SEM) in the back-scattered electron (BSE) mode.
Hydrogen absorption and permeation. Disk samples of 12 mm in diameter were cut from the as-cast ingots using a spark erosion wire cutting machine. Both surfaces of the samples were ground with SiC paper (400-2000 grid) and then polished using 0.5 μm Al 2 O 3 powder. The final thickness (d) of the samples was 0.6 mm. A pure Pd layer of 190 nm in thickness was then coated on both sides of the sample by a magnetron sputtering device for avoiding oxidation and enhancing the dissociation/recombination of hydrogen molecules during the hydrogen absorption and permeation tests.
Hydrogen absorption of as-cast Nb 40−x Mo x Ti 30 Ni 30 (x = 0, 5, 10) was characterized using a Sieverts-type apparatus. The equilibrium pressure-composition-temperature (PCT) data were measured at pressures in the range of 0.01-1.0 MPa at 523 K, 573 K, 623 K and 673 K, respectively. More details about the experimental procedure can be found in refs 17 and 18.
Samples for the hydrogen permeation tests were sandwiched by two copper gaskets and fixed in a gas permeation device. Both sides of the sample were evacuated, heated to 673 K in a vacuum and kept for 20 minutes. Hydrogen gas (99.99999 mass% purity) was introduced at pressures of 0.2-0.8 MPa and 0.1 MPa to the upstream and downstream sides of the sample, respectively. The hydrogen permeation flux through the sample was measured using a mass flow meter. The measurements were repeated at 623 K, 573 K and 523 K. In order to investigate the sensitivity to HE for the as-cast alloys under H 2 atmosphere, the hydrogen permeation tests were conducted under controlled cooling conditions. The samples were first subjected to hydrogen permeation at a hydrogen pressure difference of 0.7 MPa at 400 °C for 4 h, followed by cooling to room temperature at a cooling rate of 2 °C min −1 . The variation of the hydrogen permeation flux with time and temperature was measured. An abrupt increase in the hydrogen flux indicates the formation of cracks in the sample due to HE 16 .