Hydrogen Induced Abrupt Structural Expansion at High Temperatures of a Ni32Nb28Zr30Cu10 Membrane for H2 Purification

Ni-Nb-Zr amorphous membranes, prepared by melt-spinning, show great potential for replacing crystalline Pd-based materials in the field of hydrogen purification to an ultrapure grade (>99.999%). In this study, we investigate the temperature evolution of the structure of an amorphous ribbon with the composition Ni32Nb28Zr30Cu10 (expressed in atom %) by means of XRD and DTA measurements. An abrupt structural expansion is induced between 240 and 300 °C by hydrogenation. This structural modification deeply modifies the hydrogen sorption properties of the membrane, which indeed shows a strong reduction of the hydrogen capacity above 270 °C.


Introduction
For purification of hydrogen to an ultrapure grade (>99.999%), crystalline Pd and Pd-Ag (100-200 µm thickness) membranes have been employed for several decades [1,2]. Usually, one side of the membrane is exposed to the pressure of a gas containing H 2 . The membrane splits the hydrogen molecule, and the atomic hydrogen is absorbed in the metal and diffuses towards the lower H 2 concentration side. On the low-pressure face, the hydrogen atoms recombine to form H 2 molecules and pure hydrogen is released thanks to the high selectivity of the metallic membranes. For the Pd-based membranes, H 2 flux values in excess of 1 mol·m −2 ·s −1 have been reported [3].
However, Pd is an expensive material and in certain conditions can suffer from embrittlement, due to the formation of hydrides. In order to prevent embrittlement, alloying of Pd with Ag, Au or Ru is usually performed [4]. To mitigate the economic impact of Pd, thin layers of Pd-Ag alloys (about 50 µm) were shown to have both complete selectivity to hydrogen and good durability [5]. More recently, a different approach was proposed, in which small particles of Pd (200 nm) were mixed with silica substrates [6]. Nevertheless, palladium is also a strategic material and, therefore, starting from the amorphous phase, heating the sample above the crystallization temperature, usually more than one crystalline phase is formed [23].
Furthermore, Yamaura et al. [28] investigated the changes induced by the addition of a different element in the Ni-Nb-Zr ternary alloys on hydrogen permeability; among all the considered elements (Al, Co, Cu, P, Pd, Si, Sn, Ta and Ti), the best permeability results were obtained by the copper addition. The copper substitution does not increase the crystallization temperature, which is considered a key parameter for the potential use of these membranes; however, for (Ni 0.6 Nb 0.4 ) 45 Zr 50 Cu 5 alloy, the copper substitution increases the permeability significantly. Yamaura et al. [28] showed that hydrogen permeability decreases with the increasing Vickers hardness and concluded that the Ni-Nb-Zr-X (X = Co or Cu) amorphous alloys have high potential as hydrogen-permeable membranes. Indeed, at 400 • C the permeability increases from 0.11-1.30 × 10 −8 mol·m −1 ·s −1 ·Pa −0.5 for (Ni 0.6 Nb 0.4 ) 50 Zr 50 to 2.34 × 10 −8 mol·m −1 ·s −1 ·Pa −0.5 for (Ni 0.6 Nb 0.4 ) 45 Zr 50 Cu 5 , which is comparable to the permeability of conventional Pd-Ag alloys.
In order to obtain materials with high permeability but low hydrogen embrittlement, knowledge of their hydrogenation properties is important as it can provide information about the hydrogen effect on the material structure. In this framework, in this paper, we will report a study of the structure of a Ni 32 Nb 28 Zr 30 Cu 10 amorphous membrane, performed between 25 and 400 • C, by means of high temperature XRD measurements (HTXRD), either in a hydrogen or in a helium atmosphere. We will show that upon hydrogenation, the membranes display an abrupt expansion of the structure around 300 • C. In correspondence with this structural modification, the hydrogen absorption properties change and the membrane absorbs much less hydrogen in the high temperature regime than in the low temperature one.

Preliminarily DTA Measurements
As the occurrence of crystallization changes the absorption properties of the Ni-Nb-Zr membranes, we preliminary investigated the crystallization process by DTA measurements on the Ni 32 Nb 28 Zr 30 Cu 10 sample. In the similar compound Ni 42 Nb 28 Zr 30 , which does not contain Cu, it has been reported that crystallization occurs at temperatures higher than 480 • C [20,23,29]; however, an investigation of the presently investigated composition is still missing. Figure 1 reports the DTA curves measured at different temperature rates, ranging between 15 and 30 • C/min. phase, heating the sample above the crystallization temperature, usually more than one crystalline phase is formed [23]. Furthermore, Yamaura et al. [28] investigated the changes induced by the addition of a different element in the Ni-Nb-Zr ternary alloys on hydrogen permeability; among all the considered elements (Al, Co, Cu, P, Pd, Si, Sn, Ta and Ti), the best permeability results were obtained by the copper addition. The copper substitution does not increase the crystallization temperature, which is considered a key parameter for the potential use of these membranes; however, for (Ni0.6Nb0.4)45Zr50Cu5 alloy, the copper substitution increases the permeability significantly. Yamaura et al. [28] showed that hydrogen permeability decreases with the increasing Vickers hardness and concluded that the Ni-Nb-Zr-X (X = Co or Cu) amorphous alloys have high potential as hydrogen-permeable membranes. Indeed, at 400 °C the permeability increases from 0.11-1.30 × 10 −8 mol·m −1 ·s −1 ·Pa −0.5 for (Ni0.6Nb0.4)50Zr50 to 2.34 × 10 −8 mol·m −1 ·s −1 ·Pa −0.5 for (Ni0.6Nb0.4)45Zr50Cu5, which is comparable to the permeability of conventional Pd-Ag alloys.
In order to obtain materials with high permeability but low hydrogen embrittlement, knowledge of their hydrogenation properties is important as it can provide information about the hydrogen effect on the material structure. In this framework, in this paper, we will report a study of the structure of a Ni32Nb28Zr30Cu10 amorphous membrane, performed between 25 and 400 °C, by means of high temperature XRD measurements (HTXRD), either in a hydrogen or in a helium atmosphere. We will show that upon hydrogenation, the membranes display an abrupt expansion of the structure around 300 °C. In correspondence with this structural modification, the hydrogen absorption properties change and the membrane absorbs much less hydrogen in the high temperature regime than in the low temperature one.

Preliminarily DTA Measurements
As the occurrence of crystallization changes the absorption properties of the Ni-Nb-Zr membranes, we preliminary investigated the crystallization process by DTA measurements on the Ni32Nb28Zr30Cu10 sample. In the similar compound Ni42Nb28Zr30, which does not contain Cu, it has been reported that crystallization occurs at temperatures higher than 480 °C [20,23,29]; however, an investigation of the presently investigated composition is still missing. Figure 1 reports the DTA curves measured at different temperature rates, ranging between 15 and 30 °C/min.  They present three subsequent thermally activated peaks, starting above 480 • C, centered around 520, 550 and 570 • C, when measured at ∆T/∆t = 15 • C/min. The inset of Figure 1 displays the Kissinger plot for the three peaks. The calculated activation energies of the three processes are 370 ± 30, 441 ± 2 and 470 ± 70 kJ/mol. Such values are on the same order of magnitude of those already reported for other Ni-Nb-Zr membranes [20,22,23,29]. The DTA measurements indicate that the crystallization process, occurring in subsequent steps, starts above 480 • C.

Hydrogen Absorption Properties
The hydrogen absorption properties of the membranes were measured by obtaining pressure-composition isotherms at different temperatures in the range between 158 and 400 • C; these temperatures are well below the crystallization temperature obtained by the previously reported DTA experiments.
The pressure-composition isotherms of the Ni 32 Nb 28 Zr 30 Cu 10 membrane are reported in Figure 2. We indicate the hydrogen concentration in the sample as H/M, i.e., the ratio between the number of hydrogen atoms and the number of metal atoms. The measurements at low temperatures were possible thanks to the fast kinetics of the sorption process in this membrane (at least 10 times faster than in Ni 42 Nb 28 Zr 30 ). Figure 2 shows the absence of any pressure plateau in the whole temperature range. A similar behavior has been already observed for Ni-Nb-Zr amorphous ribbons [29]. Pressure plateaus are, on the contrary, typically observed in crystalline hydrides. Moreover, the hydrogen capacity of the Ni 32 Nb 28 Zr 30 Cu 10 membrane is higher at lower temperatures or, equivalently, the hydrogen solubility at a fixed p decreases as T increases; for example, for p = 1 bar one observes a hydrogen content of 0.52 H/M at 158 • C and only~0.27 H/M at 400 • C. Some single points along the pressure-composition isotherms were duplicated in order to verify the reproducibility of the measurements. They present three subsequent thermally activated peaks, starting above 480 °C, centered around 520, 550 and 570 °C, when measured at ΔT/Δt = 15 °C/min. The inset of Figure 1 displays the Kissinger plot for the three peaks. The calculated activation energies of the three processes are 370 ± 30, 441 ± 2 and 470 ± 70 kJ/mol. Such values are on the same order of magnitude of those already reported for other Ni-Nb-Zr membranes [20,22,23,29]. The DTA measurements indicate that the crystallization process, occurring in subsequent steps, starts above 480 °C.

Hydrogen Absorption Properties
The hydrogen absorption properties of the membranes were measured by obtaining pressure-composition isotherms at different temperatures in the range between 158 and 400 °C; these temperatures are well below the crystallization temperature obtained by the previously reported DTA experiments.
The pressure-composition isotherms of the Ni32Nb28Zr30Cu10 membrane are reported in Figure 2. We indicate the hydrogen concentration in the sample as H/M, i.e., the ratio between the number of hydrogen atoms and the number of metal atoms. The measurements at low temperatures were possible thanks to the fast kinetics of the sorption process in this membrane (at least 10 times faster than in Ni42Nb28Zr30). Figure 2 shows the absence of any pressure plateau in the whole temperature range. A similar behavior has been already observed for Ni-Nb-Zr amorphous ribbons [29]. Pressure plateaus are, on the contrary, typically observed in crystalline hydrides. Moreover, the hydrogen capacity of the Ni32Nb28Zr30Cu10 membrane is higher at lower temperatures or, equivalently, the hydrogen solubility at a fixed p decreases as T increases; for example, for p = 1 bar one observes a hydrogen content of ~0.52 H/M at 158 °C and only ~0.27 H/M at 400 °C. Some single points along the pressure-composition isotherms were duplicated in order to verify the reproducibility of the measurements. Only a limited comparison of the hydrogen absorption properties of the presently investigated Ni32Nb28Zr30Cu10 membrane with the previous literature is possible. Indeed, in many cases Ni-Nb ribbons were hydrogenated by means of electrolytic charging [30,31], instead of the direct reaction with H2 gas. Conic et al. [32] measured only the absorption kinetics at various temperatures of Zr alloys containing Nb and Ta mixtures (10 wt % Nb, 12 wt %Ta and 10 wt %Nb and 12 wt %Ta). Hao et al. reported pressure-composition curves measured at 300 °C for Ni60Nb40, (Ni0.6Nb0.4)50Zr50 and (Ni0.6Nb0.4)70Zr30 [33]. The hydrogen content increases as the Zr content increases, reaching p ≈ 0.5 MPa ~1.1 mass % in (Ni0.6Nb0.4)50Zr50 and ~0.7 mass % in (Ni0.6Nb0.4)70Zr30 [16]. Also Yamaura et al. measured a single pressure-composition curve of (Ni0.6Nb0.4)70Zr30 at 300 °C [18]. Recently, we investigated the sorption properties of melt-spun (Ni0.6Nb0.4−yTay)100−xZrx with y = 0, 0.1 and x = 20, 30, Only a limited comparison of the hydrogen absorption properties of the presently investigated Ni 32 Nb 28 Zr 30 Cu 10 membrane with the previous literature is possible. Indeed, in many cases Ni-Nb ribbons were hydrogenated by means of electrolytic charging [30,31], instead of the direct reaction with H 2 gas. Conic et al. [32] [33]. The hydrogen content increases as the Zr content increases, reaching p ≈ 0.5 MPa~1.1 mass % in (Ni 0.6 Nb 0.4 ) 50 Zr 50 and~0.7 mass % in (Ni 0.6 Nb 0.4 ) 70 Zr 30 [16]. Also Yamaura et al. measured a single pressure-composition curve of (Ni 0.6 Nb 0.4 ) 70 Zr 30 at 300 • C [18].
The sorption isotherms reported in Figure 2 are almost parallel to one another; however, one can observe a large displacement between the curve corresponding to 242 • C and that corresponding to 270 • C, and between the latter one and the isotherm measured at 300 • C, while all other curves display a constant shift toward lower H/M values as T increases (see afterwards for a quantitative analysis). Such a change in the absorption properties of the membranes is even more evident considering the van't Hoff plot, as reported in Figure 3. between 300 and 400 °C [29]. Compared to (Ni0.6Nb0.4)70Zr30, the presently investigated membranes display a higher solubility by ~0.02 H/M at 300 °C and ~0.04 H/M at 400 °C. The sorption isotherms reported in Figure 2 are almost parallel to one another; however, one can observe a large displacement between the curve corresponding to 242 °C and that corresponding to 270 °C, and between the latter one and the isotherm measured at 300 °C, while all other curves display a constant shift toward lower H/M values as T increases (see afterwards for a quantitative analysis). Such a change in the absorption properties of the membranes is even more evident considering the van't Hoff plot, as reported in Figure 3. In the case of crystalline materials, for which a pressure plateau is observed in the p-c isotherms, the van't Hoff plot reports ln(p) vs. 1/T, where p is the plateau pressure at temperature T (K) [26], and the slope of the best fit lines provides the hydrogenation enthalpy, ΔHhyd. However, the pressure-composition isotherms of the presently studied amorphous materials do not exhibit any plateau. In this case, one can still construct a van't Hoff plot reporting ln(p) vs. 1/T, where p is the pressure at a fixed value of H/M at the various temperatures. Also, in this case, one can estimate the hydrogenation enthalpy, ΔHhyd, from the slope of the plot [34]. In order to obtain the values of p at fixed values of H/M at the various temperatures, an interpolation of the experimental data using beta splines was performed by means of a program implemented with the Labview routines. The van't Hoff plot for the Ni32Nb28Zr30Cu10 membrane calculated for H/M from 0.30 to 0.42 in steps of 0.02 is reported in Figure 3, together with the best fit lines. Figure 3 clearly shows two different slopes, one in the temperature range between 158 and 242 °C and another between 300 and 400 °C; however, an evident jump is present between 242 and 300 °C. Even the hydrogenation enthalpy of the material changes when passing from the low T region to the high temperature one. For example, for H/M = 0.30, ΔHhyd passes from 36 ± 1 kJ/mol at low T to 23 ± 1 kJ/mol at high temperature. It can be noted that ΔHhyd depends on the particular H/M considered, as reported in Table 1.  In the case of crystalline materials, for which a pressure plateau is observed in the p-c isotherms, the van't Hoff plot reports ln(p) vs. 1/T, where p is the plateau pressure at temperature T (K) [26], and the slope of the best fit lines provides the hydrogenation enthalpy, ∆H hyd . However, the pressure-composition isotherms of the presently studied amorphous materials do not exhibit any plateau. In this case, one can still construct a van't Hoff plot reporting ln(p) vs. 1/T, where p is the pressure at a fixed value of H/M at the various temperatures. Also, in this case, one can estimate the hydrogenation enthalpy, ∆H hyd , from the slope of the plot [34]. In order to obtain the values of p at fixed values of H/M at the various temperatures, an interpolation of the experimental data using beta splines was performed by means of a program implemented with the Labview routines. The van't Hoff plot for the Ni 32 Nb 28 Zr 30 Cu 10 membrane calculated for H/M from 0.30 to 0.42 in steps of 0.02 is reported in Figure 3, together with the best fit lines. Figure 3 clearly shows two different slopes, one in the temperature range between 158 and 242 • C and another between 300 and 400 • C; however, an evident jump is present between 242 and 300 • C. Even the hydrogenation enthalpy of the material changes when passing from the low T region to the high temperature one. For example, for H/M = 0.30, ∆H hyd passes from 36 ± 1 kJ/mol at low T to 23 ± 1 kJ/mol at high temperature. It can be noted that ∆H hyd depends on the particular H/M considered, as reported in Table 1. The dependence of the hydrogenation enthalpy on the hydrogen content has already recently been reported for some other amorphous membranes for hydrogen purification, namely (Ni 0.6 Nb 0.4−y Ta y ) 100−x Zr x (x = 0.2, 0.3; y = 0, 0.1) membranes [29]. In the case of (Ni 0.6 Nb 0.4 ) 70 Zr 30 the enthalpy decreases from 41 kJ/mol for H/M = 0.34 to 33 kJ/mol for H/M = 0.42 [29]. This trend was attributed to the different interstitial sites available for hydrogen trapping during the progression of the absorption process: at the beginning, the deepest energy levels are occupied, while only shallower energy levels are available at the higher hydrogen content [29].
For the presently studied Ni 32 Nb 28 Zr 30 Cu 10 membrane, the calculated ∆H hyd both in the lowtemperature and in the high T regions decreases as H/M increases, similar to the other (Ni 0.6 Nb 0.4−y Ta y ) 100−x Zr x membranes [29]. However, for the low temperature phase, ∆H hyd is much higher than that of the high temperature phase, showing the different sorption properties of the compound in the two temperature ranges. Indeed, in Figure 2 we report a piece of the p-c curve at 300 • C calculated as the extrapolation of the low temperature curves, i.e., using ∆H hyd obtained for the low T region. It is well evident that the experimental curve is shifted to a lower H/M by about 0.2-0.23.
The observed change of the absorption properties around 240 • C could be related to a structural modification of the sample. However, the membrane should retain its amorphous structure, at least within the detection limit of the previously reported DTA measurements, which do not show any peak in this temperature range. To further study this hypothesis, we performed an XRD study of the high temperature evolution of the sample structure.

XRD Study of the High Temperature Evolution of the Structure
The XRD pattern of the pristine membrane, measured at room temperature in a helium atmosphere, is reported in Figure 4. It presents the typical features of an amorphous structure, with a broad peak centered around 2θ~39 • . This confirms the amorphous state of the investigated membrane, which, indeed, was produced by melt-spinning, in order to intentionally disrupt the crystalline order and render its structure amorphous (see Section 3). The dependence of the hydrogenation enthalpy on the hydrogen content has already recently been reported for some other amorphous membranes for hydrogen purification, namely (Ni0.6Nb0.4−yTay)100−xZrx (x = 0.2, 0.3; y = 0, 0.1) membranes [29]. In the case of (Ni0.6Nb0.4)70Zr30 the enthalpy decreases from 41 kJ/mol for H/M = 0.34 to 33 kJ/mol for H/M = 0.42 [29]. This trend was attributed to the different interstitial sites available for hydrogen trapping during the progression of the absorption process: at the beginning, the deepest energy levels are occupied, while only shallower energy levels are available at the higher hydrogen content [29].
For the presently studied Ni32Nb28Zr30Cu10 membrane, the calculated ΔHhyd both in the lowtemperature and in the high T regions decreases as H/M increases, similar to the other (Ni0.6Nb0.4−yTay)100−xZrx membranes [29]. However, for the low temperature phase, ΔHhyd is much higher than that of the high temperature phase, showing the different sorption properties of the compound in the two temperature ranges. Indeed, in Figure 2 we report a piece of the p-c curve at 300 °C calculated as the extrapolation of the low temperature curves, i.e., using ΔHhyd obtained for the low T region. It is well evident that the experimental curve is shifted to a lower H/M by about 0.2-0.23.
The observed change of the absorption properties around 240 °C could be related to a structural modification of the sample. However, the membrane should retain its amorphous structure, at least within the detection limit of the previously reported DTA measurements, which do not show any peak in this temperature range. To further study this hypothesis, we performed an XRD study of the high temperature evolution of the sample structure.

XRD Study of the High Temperature Evolution of the Structure
The XRD pattern of the pristine membrane, measured at room temperature in a helium atmosphere, is reported in Figure 4. It presents the typical features of an amorphous structure, with a broad peak centered around 2θ ~39°. This confirms the amorphous state of the investigated membrane, which, indeed, was produced by melt-spinning, in order to intentionally disrupt the crystalline order and render its structure amorphous (see Section 3). The temperature dependence of the structure of the Ni32Nb28Zr30Cu10 membrane was investigated by means of X-ray diffraction, heating the sample in a hydrogen ( Figure 5) atmosphere between 25 and 400 °C in subsequent steps. The temperature dependence of the structure of the Ni 32 Nb 28 Zr 30 Cu 10 membrane was investigated by means of X-ray diffraction, heating the sample in a hydrogen ( Figure 5) atmosphere between 25 and 400 • C in subsequent steps. Heating the sample in H2 (Figure 5), the angular position of the peak remains unaltered between 25 and 242 °C, while abruptly, at 270 °C, the peak shifts to a lower angular position, which afterward stays fixed up to 400 °C. It is worth noting that the persistence of the single broad feature in the diffractogram even at high temperatures suggests that the amorphous nature of the membrane is retained up to the highest T investigated here (400 °C) and excludes a transition toward a crystalline state. For comparison, we conducted similar temperature-dependent XRD experiments on a Ni32Nb28Zr30Cu10 membrane under a helium atmosphere (p ~ 1 bar) and the results are displayed in Figure 6; in this case, no abrupt peak shift is detectable with the increasing temperature. This is a clear indication that the modifications occurring in the sample structure are mainly induced by the presence of hydrogen.
The angular position of the broad HTXRD peak is linked to the mean distance between atoms in the amorphous structure and the displacement towards lower 2θ values suggests an abrupt expansion of the structure at high temperatures.  Heating the sample in H 2 ( Figure 5), the angular position of the peak remains unaltered between 25 and 242 • C, while abruptly, at 270 • C, the peak shifts to a lower angular position, which afterward stays fixed up to 400 • C. It is worth noting that the persistence of the single broad feature in the diffractogram even at high temperatures suggests that the amorphous nature of the membrane is retained up to the highest T investigated here (400 • C) and excludes a transition toward a crystalline state. For comparison, we conducted similar temperature-dependent XRD experiments on a Ni 32 Nb 28 Zr 30 Cu 10 membrane under a helium atmosphere (p~1 bar) and the results are displayed in Figure 6; in this case, no abrupt peak shift is detectable with the increasing temperature. This is a clear indication that the modifications occurring in the sample structure are mainly induced by the presence of hydrogen. Heating the sample in H2 (Figure 5), the angular position of the peak remains unaltered between 25 and 242 °C, while abruptly, at 270 °C, the peak shifts to a lower angular position, which afterward stays fixed up to 400 °C. It is worth noting that the persistence of the single broad feature in the diffractogram even at high temperatures suggests that the amorphous nature of the membrane is retained up to the highest T investigated here (400 °C) and excludes a transition toward a crystalline state. For comparison, we conducted similar temperature-dependent XRD experiments on a Ni32Nb28Zr30Cu10 membrane under a helium atmosphere (p ~ 1 bar) and the results are displayed in Figure 6; in this case, no abrupt peak shift is detectable with the increasing temperature. This is a clear indication that the modifications occurring in the sample structure are mainly induced by the presence of hydrogen.
The angular position of the broad HTXRD peak is linked to the mean distance between atoms in the amorphous structure and the displacement towards lower 2θ values suggests an abrupt expansion of the structure at high temperatures.  The angular position of the broad HTXRD peak is linked to the mean distance between atoms in the amorphous structure and the displacement towards lower 2θ values suggests an abrupt expansion of the structure at high temperatures.
To obtain a more quantitative picture, it must be noticed that the peak position at the highest intensity represents the main average interatomic distance, in either clusters or atoms in the Ni-rich amorphous matrix.
The 2θ values and the average interatomic distances obtained at the highest intensity are reported in Table 2, for both the X-ray diffraction patterns measured under H 2 or He. In Figure 7, the average interatomic distances measured under hydrogen are plotted as a function of temperature: a significant expansion of the structure is observed between 200 and 300 • C, in agreement with the temperature (242 • C), where a change in the hydrogen absorption properties is detected by the previously reported pressure-composition isotherms. On the contrary, when heating in He, a normal thermal expansion, without abrupt changes, is detected (see Table 2). To obtain a more quantitative picture, it must be noticed that the peak position at the highest intensity represents the main average interatomic distance, in either clusters or atoms in the Ni-rich amorphous matrix.
The 2θ values and the average interatomic distances obtained at the highest intensity are reported in Table 2, for both the X-ray diffraction patterns measured under H2 or He. In Figure 7, the average interatomic distances measured under hydrogen are plotted as a function of temperature: a significant expansion of the structure is observed between 200 and 300 °C, in agreement with the temperature (242 °C), where a change in the hydrogen absorption properties is detected by the previously reported pressure-composition isotherms. On the contrary, when heating in He, a normal thermal expansion, without abrupt changes, is detected (see Table 2).  From the present measurements, it is not possible to ascertain whether the entire icosahedra composing the amorphous structure or only some bonds between certain atoms expand in correspondence with the abrupt thermal expansion, but it is plausible that it may be due to an amorphous-to-amorphous phase transition. This transition would be induced by hydrogen, as it is not observed in the sample heated in He. An expansion of the Zr-Zr bond length induced by hydrogenation was already reported at room temperature [18], but in the present case, we are able to show that an abrupt and unexpected change occurs as the membrane is heated at high temperature under H2 atmosphere. Moreover, in the present work, we show that the two phases, at low and at From the present measurements, it is not possible to ascertain whether the entire icosahedra composing the amorphous structure or only some bonds between certain atoms expand in correspondence with the abrupt thermal expansion, but it is plausible that it may be due to an amorphous-to-amorphous phase transition. This transition would be induced by hydrogen, as it is not observed in the sample heated in He. An expansion of the Zr-Zr bond length induced by hydrogenation was already reported at room temperature [18], but in the present case, we are able to show that an abrupt and unexpected change occurs as the membrane is heated at high temperature under H 2 atmosphere. Moreover, in the present work, we show that the two phases, at low and at high temperature, present different hydrogen sorption properties, with the lower temperature phase displaying a much higher value for the hydrogenation enthalpy.
The X-ray diffraction patterns were obtained using a PANalytical X'Pert pro θ-θ diffractometer (Almelo, The Netherlands). The ribbons were not coated with Pd. This instrument was equipped with a heating stage with Anton Paar XRK 900 (Graz, Austria). Hydrogen gas was introduced into the heating chamber via a home-made volumetric apparatus. The sample was placed inside the chamber and first evacuated using the turbo pump by Pfeiffer Vacuum model number TSH071E (Asslar, Germany).
Simultaneous TGA-DTA measurements were conducted by means of a Setaram Sensys Evolution 1200 TGA system (Caluire, France) [35] under a high purity argon flux (60 mL/min) at ambient pressure. For each experiment, a sample mass of~10 mg was used. Different temperature rates, between 10 and 30 • C/min, were used for each sample in order to calculate the activation energy of the crystallization process.
The hydrogen absorption curves were recorded by the home-made Sieverts apparatus at Sapienza University of Rome described in Reference [23,36]. The experiments were performed on specimens with a mass of~300 mg.

Conclusions
The hydrogen adsorption properties and the amorphous structure of a Ni 32 Nb 28 Zr 30 Cu 10 membrane are studied by means of pressure-composition isotherms and XRD measurements, respectively. In the temperature range between 242 and 300 • C, an abrupt decrease of hydrogen absorption is observed; correspondingly, the hydrogen concentration measured at 7 bar is reduced from~0.6 H/M at 242 • C to~0.45 H/M at 300 • C. The X-ray diffractograms confirm that the membrane remains amorphous also at the highest investigated temperature (400 • C), and clearly indicate an abrupt increase of the mean interatomic distance from~2.31 Å at 200 • C to~2.38 Å occurs. This abrupt expansion is evidenced only when the samples are heated in a hydrogen atmosphere, while a much lower and non-abrupt thermal expansion is observed when samples are heated in an inert atmosphere. Here, for the first time, an abrupt change of the physical properties of amorphous membranes when heated in hydrogen at high temperatures is reported. Incidentally, these findings suggest that we should avoid extrapolations from properties measured close to room temperature.