Experimental and Numerical Study of Pd/Ta and PdCu/Ta Composites for Thermocatalytic Hydrogen Permeation

The development of stable and durable hydrogen (H2) separation technology is essential for the effective use of H2 energy. Thus, the use of H2 permeable membranes, made of palladium (Pd), has been extensively studied in the literature. However, Pd has considerable constraints in large-scale applications due to disadvantages such as very high cost and H2 embrittlement. To address these shortcomings, copper (Cu) and Pd were deposited on Ta to fabricate a composite H2 permeable membrane. To this end, first, Pd was deposited on a tantalum (Ta) support disk, yielding 7.4 × 10−8 molH2 m−1 s−1 Pa−0.5 of permeability. Second, a Cu–Pd alloy on a Ta support was synthesized via stepwise electroless plating and plasma sputtering to improve the durability of the membrane. The use of Cu is cost-effective compared with Pd, and the appropriate composition of the PdCu alloy is advantageous for long-term H2 permeation. Despite the lower H2 permeation of the PdCu/Ta membrane (than the Pd/Ta membrane), about two-fold temporal stability is achieved using the PdCu/Ta composite. The degradation process of the Ta support-based H2 permeable membrane is examined by SEM. Moreover, thermocatalytic H2 dissociation mechanisms on Pd and PdCu were investigated and are discussed numerically via a density functional theory study.


Introduction
The reliance on fossil fuels since the 1950s has caused a stable increase in greenhouse gas emissions and triggered the greenhouse gas effect, the main driver of global warming [1]. One of the approaches to mitigate global warming is to develop clean and renewable energy sources, such as wind, water, solar, and geothermal energy. However, sustainable and eco-friendly energy sources vary by region and lack the required infrastructure worldwide. Hydrogen (H 2 ) is one of the most promising energy carriers, given its high gravimetric energy density and its clean conversion byproduct, namely, water [2]. The chemical energy stored in H 2 can be directly converted into electricity by using fuel cells and/or heat by combustion [3]. To date,~50% of all H 2 is produced by natural gas reforming, whereas a mixture of gases is released as the product [4,5]. Utilization of other H 2 carriers and sources such as ammonia [6], a liquid organic hydrogen carrier (LOHC) [7], methanol [8], and biomass [9] also requires exhaust purification before the end-use. Therefore, the separation technique is essential for obtaining pure H 2 from a mixture consisting of H 2 , nitrogen (N 2 ), carbon monoxide (CO), carbon dioxide (CO 2 ), and water vapor (H 2 O) [10]. However, commercial methods such as pressure swing adsorption (PSA) are meation performance was delineated. Finally, density functional theory (DFT) was also applied to evaluate and explain the experimental trends obtained herein.
The novelty of this study can be summarized as the following: 1.
Providing evidence that fabrication of nanometer-thick Pd and PdCu on a dense support is achievable via plasma sputtering.

2.
Analysis of temporal stability of Pd/Ta and PdCu/Ta membranes.

Materials
Ta sheets with a thickness of 250 µm were purchased from Koralco Corporation (Gwangju, Republic of Korea). They were then wire-cut into disks with a diameter of one inch. Hydrochloric acid (HCl) and phosphoric acid (H 3 PO 4 ) were supplied by Samchun Chemicals (Seoul, Republic of Korea). Tin chloride (SnCl 2 ), palladium chloride (PdCl 2 ), hydrazine, and ethylenediaminetetraacetic acid (EDTA) were purchased from Sigma-Aldrich (St. Louis, MO, USA). Tetraamminepalladium dichloride monohydrate (Pd(NH 3 ) 4 Cl 2 ·H 2 O) was purchased from Sigma-Aldrich (St. Louis, MO, USA). All chemicals were used asreceived and without further purification unless stated otherwise.

Membrane Preparation
The preparation process is shown in Figure S1 (in Supplementary Material). The Pd layer was electroless-plated (ELP) on pre-treated Ta discs and tubes. The surface of the Ta support was polished in the following order. First, Ta supports were polished using 800, 1200, 1500, 4000, and 7000 grit sandpaper. Subsequently, as mentioned in the literature [17], the impurities on the membrane were removed using a basic solution, and organic substances were removed through acid treatment using stepwise immersion in HCl and H 3 PO 4 aqueous solutions. The Pd ELP is based on a well-established metal-metal galvanic exchange technique [36]. To this end, the surface was activated by sequential dipping of the membrane in 1.0 g/L SnCl 2 and 1.0 g/L PdCl 2 aqueous solutions, each containing 0.01 M HCl as the stabilizing agent. This process was repeated three times, and each step was conducted for 5 min. Finally, the ELP was performed at a temperature of 60 • C in a Pd(NH 3 ) 4 Cl 2 bath following the details mentioned in [17].
For the PdCu/Ta membrane, the sample was prepared using ELP and plasma sputtering (SPT). First, Pd was plated using the method described above. Then, Cu was deposited on the Pd layer by using a magnetron sputtering system (Korea Vacuum Tech, Gimpo-si, Korea). Subsequently, the prepared membrane was treated at 480 • C for 1 h in H 2 atmosphere for alloying. The co-sputtered (co-SPT) PdCu/Ta membrane was also prepared by using a plasma sputtering system to avoid the adverse effect of Sn residue from SnCl 2 during the activation process. The sputtering conditions were 10 cm (distance between the target and substrate), 25 W (Pd, DC), 18-45 W (Cu, RF) in 20 NmL/min of Ar stream, 2 mTorr of working pressure, and deposition temperatures of the room (~20 • C, R.T.) and 400 • C (Pd). The PdCu/Ta membranes were also prepared with ELP Pd and SPT Cu, successively. After deposition, Pd and Cu were alloyed at 480 • C for 1 h in a H 2 atmosphere.

Permeation Testing
The tests were conducted by using high-purity (99.999%) H 2 gas for the permeation test and Ar gas for purging during the heating and cooling processes. The gas flow rate was adjusted using a thermal mass flow controller (MFC, Bronkhorst High-Tech BV, Ruurlo, Netherlands), while the pressure at the membrane terminals was adjusted using an electric pressure controller (EPC). The permeate gas flow rate was measured by a mass flow meter (MFM, Bronkhorst High-Tech BV, Ruurlo, Netherlands) and a bubble flow meter (BFM, Horiba, Kyoto, Japan). The measurement system is described in detail in Figure 1. The supplied gas was either passed through the H 2 separation membrane in the reactor (permeate) or separated to escape the furnace (retentate). In the experiment, H 2 permeability was measured at 400, 425, 450, 475, and 500 • C (ramp-up rate: 5 • C/min) and the Figure 1. The supplied gas was either passed through the H2 separation membrane in the reactor (permeate) or separated to escape the furnace (retentate). In the experiment, H2 permeability was measured at 400, 425, 450, 475, and 500 °C (ramp-up rate: 5 °C/min) and the pressure range of 1-5 bar. A temporal stability test for H2 permeation was performed at the 500 °C and 5 bar conditions until the membranes were broken.

2.4.. Material Characterizations
The surface and cross-sectional images of the prepared membranes were retrieved through scanning electron microscopy (SEM, Inspect F-50, FEI Company, Hillsboro, Oregon, United States), and elemental mapping was performed using an energy dispersive spectrometry detector (EDS, AMETEK Inc., Berwyn, Pennsylvania, United States). An accelerating voltage of 15 kV was used for SEM analysis unless stated otherwise. Before retrieving the cross-sectional images, the samples had been molded and cured in epoxy resin for one day. Besides these characterizations, the prepared membranes were characterized using an X-ray diffractometer (XRD, D' Max 2500, Rigaku, Tokyo, Japan) to investigate the lattice of alloyed metals and Rutherford backscattering spectroscopy (RBS, National Electrostatics Corporation, Middleton, Wisconsin, United States) to determine the composition of the alloyed metal prepared by ELP and SPT, respectively. XRD scanning range was 10-90 deg with a step size of 0.02 (2θ).

DFT Modeling
Spin-polarized density functional theory (DFT) calculations with the aid of the Vienna Ab-initio Simulation Package (VASP, University of Vienna, Vienna, Austria) [37] were also conducted, and the exchange-correlation function was described using the Perdew Burke-Ernzerhof generalized gradient approximation (GGA) method [38]. The projector augmented wave (PAW) method was applied to substitute complicated ionic potentials caused by the interaction between the ion and electron cores [39]. A plane wave

Material Characterizations
The surface and cross-sectional images of the prepared membranes were retrieved through scanning electron microscopy (SEM, Inspect F-50, FEI Company, Hillsboro, OR, USA), and elemental mapping was performed using an energy dispersive spectrometry detector (EDS, AMETEK Inc., Berwyn, PA, USA). An accelerating voltage of 15 kV was used for SEM analysis unless stated otherwise. Before retrieving the cross-sectional images, the samples had been molded and cured in epoxy resin for one day. Besides these characterizations, the prepared membranes were characterized using an X-ray diffractometer (XRD, D' Max 2500, Rigaku, Tokyo, Japan) to investigate the lattice of alloyed metals and Rutherford backscattering spectroscopy (RBS, National Electrostatics Corporation, Middleton, WI, USA) to determine the composition of the alloyed metal prepared by ELP and SPT, respectively. XRD scanning range was 10-90 deg with a step size of 0.02 (2θ).

DFT Modeling
Spin-polarized density functional theory (DFT) calculations with the aid of the Vienna Ab-initio Simulation Package (VASP, University of Vienna, Vienna, Austria) [37] were also conducted, and the exchange-correlation function was described using the Perdew Burke-Ernzerhof generalized gradient approximation (GGA) method [38]. The projector augmented wave (PAW) method was applied to substitute complicated ionic potentials caused by the interaction between the ion and electron cores [39]. A plane wave expansion with a cutoff energy of 400 eV was used to express the valence electrons. A 5 × 5 × 1 Monkhorst-Pack mesh k-point was utilized to determine the optimal geometries and total energy with sufficient accuracy [40]. 50 at.% Pd to understand the H 2 dissociation on the surface and the diffusion of atomic H into the membrane. Each slab was modeled using 2 × 2 six-layer supercells, with all the layers relaxed. Although Pd exists as the FCC crystal structure in the standard state, BCC Pd is considered for comparison against BCC PdCu. The quantified lattice parameters of FCC Pd, FCC PdCu, BCC Pd, and BCC PdCu were, 3.94 Å, 3.81 Å, 3.23 Å, and 2.99 Å, respectively. These estimates showed good agreement with the empirical estimates [a Pd = 3.89 Å, a (FCC Pd52Cu48) = 3.77 Å and a (BCC Pd47Cu53) = 2.97 Å] (see Table 1). Moreover, the climbing image nudged elastic band (CI-NEB) method [41] was applied to quantify the energy barriers for H 2 dissociation on the modeled Pd and PdCu surfaces. Six images between the initial and final adsorption geometries were generated for this purpose. Equation (1) was used to calculate the binding energy (E bind ) of H 2 (or H): where E H 2 (or H)/slab , E H 2 (or H) , and E slab are the total energy of the H 2 (or H)-adsorbed slab, gaseous H 2 (or H), and pure slab systems, respectively.

Pd/Ta Membrane
In the Pd/Ta H 2 separation membrane, the permeability was measured at 450-500 • C and 1-5 bar of pressure difference, as shown in Figure 2a. For these metal membranes, H 2 permeates in the following order: adsorption, dissociation, volumetric diffusion, association, and desorption. Equation (2) establishes the permeate flux of these membranes based on Sievert's law: where Q f represents the permeability of the membrane, l is the thickness of the membrane, P feed is the pressure of the front part before permeation, and P perm is the pressure of the latter part after permeation. In the H 2 separation membranes, n = 0.5 if the permeation rate is determined by diffusion through the metal layer, while n = 1 if the rate is determined by the H 2 dissociation/association reaction on the surface, and 0.5 < n < 1 if both apply [45]. It is generally thought that H 2 penetration in a metal layer containing Pd follows the Sievert's law, which means n = 0.5 [46]. In this case, the atmosphere was close to the ideal gas conditions, and the rate of H 2 permeation was mainly conducted through the metal lattice. Figure 2 shows the permeability of the Pd/Ta composite membrane at various temperatures and pressures. The H 2 permeability of the Pd/Ta membrane was measured as 16.18 cm 3 ·cm −2 ·min −1 . This illustrates a 7.4 × 10 −8 mol H 2 m −1 s −1 Pa −0.5 permeability which is well within the range reported previously for Pd/Ta membranes [34]. At 500 • C, H 2 permeated through the membrane at n = 0.5. However, n was close to one in the 475 • C and 450 • C experiments. This phenomenon could be driven by the degradation of the membrane surface with the increasing experimental time. In this experiment, H 2 permeability was measured from high to low temperatures.
ured from high to low temperatures. Figure 2b illustrates the comparison of the permeability of H2 separation membranes prepared by various methods at 500 °C and a pressure difference of 5 bar. In the H2 separation membrane, fabricated by ELP and SPT at 400 °C, the H2 permeability exhibited nearly similar values. The H2 separation membrane, deposited at room temperature, exhibited a significantly lower estimate than the other two membranes. This phenomenon is attributable to the difference in the density of the Pd surface. In general, due to the sintering effect, the metal deposited at high temperatures is denser than the one sputtered at room temperature. When Pd is deposited on the substrate at a high temperature, Pd atoms easily move on the substrate, thereby forming a denser layer during sputtering deposition [47]. This was confirmed by the SEM images of the samples before the H2 permeation test, as shown in Figure 3. As seen, the membranes prepared by ELP and SPT at 400 °C have denser structures than the room-temperature  Figure 2b illustrates the comparison of the permeability of H 2 separation membranes prepared by various methods at 500 • C and a pressure difference of 5 bar. In the H 2 separation membrane, fabricated by ELP and SPT at 400 • C, the H 2 permeability exhibited nearly similar values. The H 2 separation membrane, deposited at room temperature, exhibited a significantly lower estimate than the other two membranes. This phenomenon is attributable to the difference in the density of the Pd surface.
In general, due to the sintering effect, the metal deposited at high temperatures is denser than the one sputtered at room temperature. When Pd is deposited on the substrate at a high temperature, Pd atoms easily move on the substrate, thereby forming a denser layer during sputtering deposition [47]. This was confirmed by the SEM images of the samples before the H 2 permeation test, as shown in Figure 3. As seen, the membranes prepared by ELP and SPT at 400 • C have denser structures than the room-temperature SPT surface with many defects. The SPT Pd layer deposited at RT has a relatively larger size (~10 nm, thus, agreeing with the available literature [48,49]) and a larger number of pores than the SPT Pd deposited at 400 • C and ELP Pd. On top of that, almost no pores were identified on the surface.
Moreover, the morphology of the H2 separation membranes was changed after the permeation test for 10 h. The porous structure of the SPT samples was changed to a smoother surface due to sputtering at the operating temperature. Large pore islands were formed by the agglomeration of Pd on the surface. After long exposure to a high temperature, Pd becomes more aggregated, revealing the Ta support layer [34]. This degradation is a critical weakness in the H2 separation membranes during long-term use. To address this, PdCu alloys were also prepared and tested in this study.  Figure 4 shows the SEM images of the cross-sections of the ELP and SPT Pd/Ta membranes after the H2 permeation experiment. In contrast to Figure 1d (for Pd/Ta before the H2 permeation test), the delamination was identified on the cross-section of the membrane after the experiment. This finding confirms that separation occurred precisely at the interface between Ta and Pd. As the experiment was conducted at a high temperature (500 °C), the surface of Ta, directly exposed to H2, increases as Pd aggregates on the surface. Then, delamination occurs, given the difference in the H2 embrittlement between the two metals. Unlike Pd with generally low H2 embrittlement at high temperatures, Ta exhibits the opposite trend [23,50]. In addition, the Sn used in electroless plating can also affect delamination. Moreover, the morphology of the H 2 separation membranes was changed after the permeation test for 10 h. The porous structure of the SPT samples was changed to a smoother surface due to sputtering at the operating temperature. Large pore islands were formed by the agglomeration of Pd on the surface. After long exposure to a high temperature, Pd becomes more aggregated, revealing the Ta support layer [34]. This degradation is a critical weakness in the H 2 separation membranes during long-term use. To address this, PdCu alloys were also prepared and tested in this study. Figure 4 shows the SEM images of the cross-sections of the ELP and SPT Pd/Ta membranes after the H 2 permeation experiment. In contrast to Figure 1d (for Pd/Ta before the H 2 permeation test), the delamination was identified on the cross-section of the membrane after the experiment. This finding confirms that separation occurred precisely at the interface between Ta and Pd. As the experiment was conducted at a high temperature (500 • C), the surface of Ta, directly exposed to H 2 , increases as Pd aggregates on the surface. Then, delamination occurs, given the difference in the H 2 embrittlement between the two metals. Unlike Pd with generally low H 2 embrittlement at high temperatures, Ta exhibits the opposite trend [23,50]. In addition, the Sn used in electroless plating can also affect delamination.
Furthermore, the Sn residue in the ELP method reduces the adhesion of Pd to the Ta substrate, thus, promoting the delamination of Pd (see Figure S2). This phenomenon prevents a uniform supply of hydrogen atoms to Ta, subsequently leading to a higher degree of delamination, which can ultimately reduce the stability of the H 2 separation membrane in long-term operations. The delamination occurs in the membranes prepared by sputtering without Sn, but this delamination occurs at smaller scales and more sporadically. This indicates that the H 2 embrittlement and Sn residue factors both affect the degradation of the metal-support-based H 2 separation membrane.  Furthermore, the Sn residue in the ELP method reduces the adhesion of Pd to the Ta substrate, thus, promoting the delamination of Pd (see Figure S2). This phenomenon prevents a uniform supply of hydrogen atoms to Ta, subsequently leading to a higher degree of delamination, which can ultimately reduce the stability of the H2 separation membrane in long-term operations. The delamination occurs in the membranes prepared by sputtering without Sn, but this delamination occurs at smaller scales and more sporadically. This indicates that the H2 embrittlement and Sn residue factors both affect the degradation of the metal-support-based H2 separation membrane.

PdCu/Ta Membrane
Besides the Pd/Ta membrane, a layer of PdCu alloys was deposited on Ta supports to mitigate the disadvantages of Pd, including the weak durability and high cost. PdCu/Ta membranes were prepared using two methods: (i) electroless-plated (ELP) Pd, followed by sputtered Cu (SPT), and (ii) co-sputtered PdCu (co-SPT) on the Ta disk. Figure 5 indicates that the formation of a uniform Pd-based coating is achievable on dense supports, which is a silicon wafer in this case. The thickness of the co-sputtered layer is measured to be 446.8 nm, and it is at least one order of magnitude lower than the Pd coating on the porous supports reported by previous studies [45]. The advantage of thinner Pd-based alloy could also be supported by the work of Ramachandran et al. [51], where they illustrated that H2 permeability is inversely correlated with Pd-based membrane thickness.

PdCu/Ta Membrane
Besides the Pd/Ta membrane, a layer of PdCu alloys was deposited on Ta supports to mitigate the disadvantages of Pd, including the weak durability and high cost. PdCu/Ta membranes were prepared using two methods: (i) electroless-plated (ELP) Pd, followed by sputtered Cu (SPT), and (ii) co-sputtered PdCu (co-SPT) on the Ta disk. Figure 5 indicates that the formation of a uniform Pd-based coating is achievable on dense supports, which is a silicon wafer in this case. The thickness of the co-sputtered layer is measured to be 446.8 nm, and it is at least one order of magnitude lower than the Pd coating on the porous supports reported by previous studies [45]. The advantage of thinner Pd-based alloy could also be supported by the work of Ramachandran et al. [51], where they illustrated that H 2 permeability is inversely correlated with Pd-based membrane thickness.  Figure 6 shows the characterization of the co-SPT alloy membrane by using XRD and RBS. X-ray diffractograms demonstrate that Pd and Cu exist as alloys and not as separate phases. Note that quantitative analysis of the alloys was possible by using RBS, which confirmed the alloy composition present in the membrane. The PdCu alloy has different metal lattices depending on the component ratio. The highest performance ratio is the 6:4  Figure 6 shows the characterization of the co-SPT alloy membrane by using XRD and RBS. X-ray diffractograms demonstrate that Pd and Cu exist as alloys and not as separate phases. Note that quantitative analysis of the alloys was possible by using RBS, which confirmed the alloy composition present in the membrane. The PdCu alloy has different metal lattices depending on the component ratio. The highest performance ratio is the 6:4 weight ratio (47:53 mol ratio), where both FCC and BCC structures exist simultaneously [29]. In general, the FCC lattice has high solubility, and the BCC lattice has high diffusivity. Figure 6b shows that the optimum alloying ratio has been achieved for the as-prepared PdCu(co-SPT)/Ta membrane.   Figure 7 shows the H2 permeability of PdCu/Ta membranes at various temperatures (400-500 °C) and pressure gradients (0.41-1.45 bar 0.5 ). As expected, an increase in either temperature or pressure gradient results in higher H2 permeability. The Pd(ELP)Cu(SPT)/Ta membrane exhibited 9.66 cm 3 ·cm −2 ·min −1 at 500 °C and a pressure difference of 5 bar, which is three times higher than that of the PdCu(co-SPT)/Ta membrane. If Pd and Cu are both deposited by ELP, the plated Pd will detach when Cu is plated. This occurs because Cu replaces the Pd atoms when plated under higher pH conditions than Pd plating. Previous studies have shown that Pd precursors can be used to activate the surface of supports to conduct Cu ELP, with an appropriate pH of 11 [52,53]. Moreover, in the case of co-sputtering Pd and Cu, some practical challenges emerge. Although the alloy composition can be easily controlled by co-sputtering, we identified a low H2 permeation problem. It has been likely driven by less dense structures at room temperature deposition (see Figure 7a). In contrast, the side effects of high-temperature deposition,  Figure 7 shows the H 2 permeability of PdCu/Ta membranes at various temperatures (400-500 • C) and pressure gradients (0.41-1.45 bar 0.5 ). As expected, an increase in either temperature or pressure gradient results in higher H 2 permeability. The Pd(ELP)Cu(SPT)/Ta membrane exhibited 9.66 cm 3 ·cm −2 ·min −1 at 500 • C and a pressure difference of 5 bar, which is three times higher than that of the PdCu(co-SPT)/Ta membrane. If Pd and Cu are both deposited by ELP, the plated Pd will detach when Cu is plated. This occurs because Cu replaces the Pd atoms when plated under higher pH conditions than Pd plating. Previous studies have shown that Pd precursors can be used to activate the surface of supports to conduct Cu ELP, with an appropriate pH of 11 [52,53]. Moreover, in the case of co-sputtering Pd and Cu, some practical challenges emerge. Although the alloy composition can be easily controlled by co-sputtering, we identified a low H 2 permeation problem. It has been likely driven by less dense structures at room temperature deposition (see Figure 7a). In contrast, the side effects of high-temperature deposition, such as the formation of unwanted alloy phases, could simultaneously emerge. Meanwhile, the method of sputtering Cu on an electroless-plated Pd could solve the problem above, revealing a stable high H 2 permeability, as shown in Figure 7b. Moreover, the maximum permeability obtained for the PdCu/Ta membrane was 4.4 × 10 −8 molH2 m −1 s −1 Pa −0.5 which is more than the reports on PdCu membranes with permeabilities ranging from 0 to 2.75 × 10 −8 molH2 m −1 s −1 Pa −0.5 [54]. The H2 permeability in the Pd(ELP)Cu(SPT)/Ta membrane was ~60% of that in the Pd/Ta membrane. However, it exhibits excellent durability at high temperatures, where the n value does not change with the experimental time. The constant value of n indicates that the rate of H2 molecule decomposition on the surface remains constant during the experiment. This phenomenon can be interpreted as a lower degree of surface degradation compared with the Pd/Ta membrane. This phenomenon can be confirmed by the SEM images of the Pd(ELP)Cu(SPT)/Ta membrane before and after the temporal stability test, as shown in Figure 8. As a result, the surface of the Pd(ELP)Cu(SPT)/Ta membrane has not degraded after the test, unlike the Pd/Ta membranes (see Figure 3). Only slight agglomeration by Moreover, the maximum permeability obtained for the PdCu/Ta membrane was 4.4 × 10 −8 mol H 2 m −1 s −1 Pa −0.5 which is more than the reports on PdCu membranes with permeabilities ranging from 0 to 2.75 × 10 −8 mol H 2 m −1 s −1 Pa −0.5 [54]. The H 2 permeability in the Pd(ELP)Cu(SPT)/Ta membrane was~60% of that in the Pd/Ta membrane. However, it exhibits excellent durability at high temperatures, where the n value does not change with the experimental time. The constant value of n indicates that the rate of H 2 molecule decomposition on the surface remains constant during the experiment. This phenomenon can be interpreted as a lower degree of surface degradation compared with the Pd/Ta membrane. This phenomenon can be confirmed by the SEM images of the Pd(ELP)Cu(SPT)/Ta membrane before and after the temporal stability test, as shown in Figure 8. As a result, the surface of the Pd(ELP)Cu(SPT)/Ta membrane has not degraded after the test, unlike the Pd/Ta membranes (see Figure 3). Only slight agglomeration by heat and no bare Ta surface islands were identified in the Pd(ELP)Cu(SPT)/Ta case. This indicates the absence of degradation due to the high temperature. This result is consistent with the constant n value in the H 2 permeation test. As Ta does not directly interact with H 2 , the deterioration due to H 2 embrittlement of Ta is considered to be averted. with the constant n value in the H2 permeation test. As Ta does not directly interact with H2, the deterioration due to H2 embrittlement of Ta is considered to be averted. Lastly, calculations assuming the cost of Pd at (48,226 $/kg [33]) and Cu at (6.11 $/kg [33]) show that Pd coating with the as-fabricated dimensions (~4 μm in thickness and 25.4 mm in diameter) costs ~1157.4 $/m2, whereas the as-proposed PdCu alloy costs ~57% less, well within the United States Department of Energy's standard of <5400 $/m2 [55].

Arrhenius Plot
According to the van't Hoff-Arrhenius equation, the relationship between H2 permeability and temperature can be expressed by Equation (3): where P is the permeability (mol/m 2 s Pa 0.5 ), P0 is the pre-exponential coefficient, Ea is the apparent activation energy, R is the ideal gas constant (J/mol K), and T is the absolute temperature. The activation energies can be quantified from the slope of the lnP˗1000/T graphs, as shown in Figure 9. The estimates are 20.2 kJ/mol for the Pd(ELP)/Ta disk, 26.4 kJ/mol for the Pd(ELP)Cu(SPT)/Ta disk, and 38.4 kJ/mol for the PdCu(co-SPT)/Ta disk. The H2 decomposition reaction on the membrane surface is endothermic. The reaction on the Pd surface has lower activation energy than that on the PdCu alloy surface. These values are consistent with the experimental permeations, where the Pd disk has the fastest permeation rate and the lowest activation energy. This confirms that the H2 permeation through palladium is faster than that of the PdCu alloy. Moreover, the difference in the activation energy between the PdCu alloy membranes in the H2 permeability results was also identified; compared with Pd/Ta, the difference was 1.45 times and 2.15 times higher for Pd(ELP)Cu(SPT)/Ta and PdCu(Co-SPT)/Ta, respectively. This is in good agree- Lastly, calculations assuming the cost of Pd at (48,226 $/kg [33]) and Cu at (6.11 $/kg [33]) show that Pd coating with the as-fabricated dimensions (~4 µm in thickness and 25.4 mm in diameter) costs~1157.4 $/m 2 , whereas the as-proposed PdCu alloy costs~57% less, well within the United States Department of Energy's standard of <5400 $/m2 [55].

Arrhenius Plot
According to the van't Hoff-Arrhenius equation, the relationship between H 2 permeability and temperature can be expressed by Equation (3): where P is the permeability (mol/m 2 s Pa 0.5 ), P 0 is the pre-exponential coefficient, E a is the apparent activation energy, R is the ideal gas constant (J/mol K), and T is the absolute temperature. The activation energies can be quantified from the slope of the lnP-1000/T graphs, as shown in Figure 9. fabricated by co-SPT. It is considered that the non-dense alloy structure on the membrane surface, observed through SEM, acts as a constraint with regards to the H2 permeation and increases the activation energy. Godbole et al. used NiO thin films and concluded that the activation energy of the reaction could be increased as the surface roughness is increased [56]. In other words, the activation energies are different even with the same alloy.  The H 2 decomposition reaction on the membrane surface is endothermic. The reaction on the Pd surface has lower activation energy than that on the PdCu alloy surface. These values are consistent with the experimental permeations, where the Pd disk has the fastest permeation rate and the lowest activation energy. This confirms that the H 2 permeation through palladium is faster than that of the PdCu alloy. Moreover, the difference in the activation energy between the PdCu alloy membranes in the H 2 permeability results was also identified; compared with Pd/Ta, the difference was 1.45 times and 2.15 times higher for Pd(ELP)Cu(SPT)/Ta and PdCu(Co-SPT)/Ta, respectively. This is in good agreement with the results in Figure 7, where the membrane made by successive ELP and SPT exhibited a significantly higher (up to 5.9 times) H 2 permeability value than the membrane fabricated by co-SPT. It is considered that the non-dense alloy structure on the membrane surface, observed through SEM, acts as a constraint with regards to the H 2 permeation and increases the activation energy. Godbole et al. used NiO thin films and concluded that the activation energy of the reaction could be increased as the surface roughness is increased [56]. In other words, the activation energies are different even with the same alloy.

H 2 Permeation Modeling
For a deeper understanding of the H 2 permeation difference between PdCu/ and Pd membranes, DFT calculations were performed on the reaction energy/barrier for H 2 dissociation and the energy for H migration in FCC Pd (111), BCC Pd (110), FCC PdCu (111), and BCC PdCu (110) slabs. Figure 10 and Table 2 display the optimized adsorption configuration and binding energy (indicated by E bind ) for one H 2 molecule and two H atoms on the four membrane models. On the surface of each membrane, H 2 was adsorbed onto the top site and then dissociated into two hollow sites: two fcc sites (fcc-fcc) for FCC Pd (111) (Figure 10a), one fcc1 site and one hcp1 site (associated with two Pd and one Cu atom) for FCC PdCu (111) (Figure 10b), two hollow sites for BCC Pd (110) (Figure 10c), and two hollow1 sites (associated with two Pd and one Cu atom) for BCC PdCu (110) (Figure 10d). As shown in Table 2, the activation energies for H 2 dissociation into two H atoms on the FCC Pd (111) and BCC Pd (110) surfaces were lower than those on the FCC PdCu (111) and BCC PdCu (110) surfaces. This finding indicates that the pure Pd membrane has higher catalytic activity in the initial formation of H atoms on the surface than in the PdCu case.

H2 Permeation Modeling
For a deeper understanding of the H2 permeation difference between PdCu/ and Pd membranes, DFT calculations were performed on the reaction energy/barrier for H2 dissociation and the energy for H migration in FCC Pd (111), BCC Pd (110), FCC PdCu (111), and BCC PdCu (110) slabs. Figure 10 and Table 2 display the optimized adsorption configuration and binding energy (indicated by Ebind) for one H2 molecule and two H atoms on the four membrane models. On the surface of each membrane, H2 was adsorbed onto the top site and then dissociated into two hollow sites: two fcc sites (fcc-fcc) for FCC Pd (111) (Figure 10a), one fcc1 site and one hcp1 site (associated with two Pd and one Cu atom) for FCC PdCu (111) (Figure 10b), two hollow sites for BCC Pd (110) (Figure 10c), and two hollow1 sites (associated with two Pd and one Cu atom) for BCC PdCu (110) ( Figure 10d). As shown in Table 2, the activation energies for H2 dissociation into two H atoms on the FCC Pd (111) and BCC Pd (110) surfaces were lower than those on the FCC PdCu (111) and BCC PdCu (110) surfaces. This finding indicates that the pure Pd membrane has higher catalytic activity in the initial formation of H atoms on the surface than in the PdCu case.      (110)] during the migration process from the top surface layer to the inside layer of slab. This suggests that the Pd membrane is more conductive than the PdCu membrane for H 2 permeation from the energy point of view, which is also supported by the work of Huang and Chen [57]. This finding resonates well with the experimental observations.  (110)] during the migration process from the top surface layer to the inside layer of slab. This suggests that the Pd membrane is more conductive than the PdCu membrane for H2 permeation from the energy point of view, which is also supported by the work of Huang and Chen [57]. This finding resonates well with the experimental observations.  (110); tetrahedral sites associated with two Pd/two Cu atoms are referred to Tr, T, and Tt along the diffusion path.

Comparative Temporal Stability Tests
A temporal H2 permeation test was performed using the sample with the highest permeability, namely Pd(ELP)/Ta and Pd(ELP)Cu(SPT)/Ta H2 separation membranes. H2 permeability was continuously measured, while maintaining the hardest conditions of the H2 permeation test conditions, namely, 500 °C and a pressure difference of 5 bar. Figure   Figure 11. Variations of H binding energy as an H atom migrates through the Pd and PdCu 6-layer slabs. (a) FCC Pd (111); octahedral and tetrahedral sites are denoted as O and T, respectively. (b) FCC PdCu (111); octahedral sites associated with four Pd/two Cu atoms and two Pd/four Cu atoms are referred to as O1 and O2, respectively. Tetrahedral sites surrounded by two Pd/one Cu atoms and one Pd/two Cu atoms are denoted as T1 and T2, respectively. (c) BCC Pd (110); tetrahedral sites are denoted as Tt, Tr, and T along diffusion path. (d) BCC PdCu (110); tetrahedral sites associated with two Pd/two Cu atoms are referred to Tr, T, and Tt along the diffusion path.

Comparative Temporal Stability Tests
A temporal H 2 permeation test was performed using the sample with the highest permeability, namely Pd(ELP)/Ta and Pd(ELP)Cu(SPT)/Ta H 2 separation membranes. H 2 permeability was continuously measured, while maintaining the hardest conditions of the H 2 permeation test conditions, namely, 500 • C and a pressure difference of 5 bar. Figure 12 shows the results of the H 2 permeation for 14 h. In this graph (red line), H 2 permeability decreased rapidly with time for Pd(ELP)/Ta. This trend confirms that Pd deteriorates quickly at high temperatures. Notably, after~8 h, the H 2 permeability converged to 0. In other words, it takes~8 h to lose the H 2 permeation ability as Pd aggregates exposing the Ta substrate to H 2 , which in turn reduces the catalytic activity of the Pd layer and increases the susceptibility of the exposed Ta to mechanical breakage due to H 2 embrittlement. On the other hand, the trend of the temporal permeation test result for the Pd(ELP)Cu(SPT)/Ta (black line) membrane is noticeably different. Even though the overall trend was decreasing, similar to Pd(ELP)/Ta, the decreasing trend was considerably slower. This finding is in line with the surface degradation of the Pd(ELP)Cu(SPT)/Ta, which was delayed in the H 2 permeation experiment. As a result, higher permeability of H 2 was maintained for twice as long as that of the Pd(ELP)/Ta membrane.

Conclusions
In this study, Pd/Ta and PdCu/Ta composite membranes were prepared using different synthesis methods. The morphology (as-prepared and post-mortem), H2 permeability, and durability of the samples were empirically evaluated. The effect of Cu alloying on the catalytic performance and activation energy of the composite membrane were theoretically studied and discussed. Our study reports several important findings summarized as the following: First, compared with sputtering, electroless plating yields a more uniform and defectless Pd coating on the Ta substrate. Moreover, a higher temperature is more conducive for a lower degree of porosity in sputtering. Second, Pd/Ta prepared by electroless plating and high-temperature (400 °C) sputtering exhibited a similar trend in the H2 permeation rate. Third, the activation energy, calculated for Pd/Ta, was lower than that for PdCu/Ta, indicating a lower catalytic activity as expected by the lesser amount of Pd in the alloy. PdCu/Ta, prepared by electroless Pd plating and sputtered Cu, revealed lower activation energy than co-sputtered PdCu. The DFT modeling results confirmed the lower catalytic activity of PdCu/Ta compared with that of the Pd/Ta membrane. Moreover, unlike using porous supports, developing a Pd or PdCu film with a sub-micron thickness (~500 nm) on a dense Ta substrate is possible while using the sputtering (former) or co-sputtering (latter) techniques. Ultimately, the Pd layer on the Pd/Ta membrane agglomerated into a honeycomb shape was experimentally demonstrated. As a result, Ta decelerates the decomposition of H2 molecules on the surface thanks to the lower catalytic activity. At the same

Conclusions
In this study, Pd/Ta and PdCu/Ta composite membranes were prepared using different synthesis methods. The morphology (as-prepared and post-mortem), H 2 permeability, and durability of the samples were empirically evaluated. The effect of Cu alloying on the catalytic performance and activation energy of the composite membrane were theoretically studied and discussed. Our study reports several important findings summarized as the following: First, compared with sputtering, electroless plating yields a more uniform and defectless Pd coating on the Ta substrate. Moreover, a higher temperature is more conducive for a lower degree of porosity in sputtering. Second, Pd/Ta prepared by electroless plating and high-temperature (400 • C) sputtering exhibited a similar trend in the H 2 permeation rate. Third, the activation energy, calculated for Pd/Ta, was lower than that for PdCu/Ta, indicating a lower catalytic activity as expected by the lesser amount of Pd in the alloy. PdCu/Ta, prepared by electroless Pd plating and sputtered Cu, revealed lower activation energy than co-sputtered PdCu. The DFT modeling results confirmed the lower catalytic activity of PdCu/Ta compared with that of the Pd/Ta membrane. Moreover, unlike using porous supports, developing a Pd or PdCu film with a sub-micron thickness (~500 nm) on a dense Ta substrate is possible while using the sputtering (former) or co-sputtering (latter) techniques. Ultimately, the Pd layer on the Pd/Ta membrane agglomerated into a honeycomb shape was experimentally demonstrated. As a result, Ta decelerates the decomposition of H 2 molecules on the surface thanks to the lower catalytic activity. At the same time, the permeability of PdCu/Ta was maintained twice as long as that of Pd/Ta. Though the temporal stability test shown here is for 14 h due to the compromised mechanical stability of embrittled membranes, one can apply the findings of this study, in terms of the materials and synthesis method, in different membrane geometries and configurations, e.g., a tubular membrane reactor, to achieve a longer service life. Finally, we can conclude that alloying with Cu can open a window to long-term H 2 operation compared with the bare Pd membrane. This greatly contributes to the development of cost-effective and durable membranes for thermocatalytic H 2 separation and purification.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.

Conflicts of Interest:
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to have influenced the work reported in this paper.