Metallic Supported Pd-Ag Membranes for Simultaneous Ammonia Decomposition and H₂ Separation in a Membrane Reactor: Experimental Proof of Concept

Abstract

The use of ammonia as a hydrogen carrier requires efficient cracking technology. A promising solution is the use of a membrane reactor (MR), which enables both ammonia decomposition and hydrogen separation to take place within the same device, providing advantages in terms of efficiency and compactness compared to conventional systems. The literature reports that ceramic-supported double-skinned Pd-Ag membranes show outstanding performance for hydrogen separation as well as good stability of the separation layer during ammonia decomposition. However, their sealing in the reactor may result in leakage increase, while their mechanical stability remains an unresolved issue. To circumvent these limitations, the use of metallic supported Pd-based membranes is recommended, due to their higher mechanical stability and ease of sealing and integration in the reactor. In this work, we propose the development of robust metallic supported hydrogen-selective membranes for integration in membrane reactors for ammonia cracking. A conventional Pd-Ag membrane was prepared on a low-cost porous Hastelloy X tubular filter, modified with α-Al₂O₃/γ-Al₂O₃ to reach the desired surface quality. The membrane was then tested for ammonia decomposition in a MR configuration, showing the ability to reach >99% NH₃ conversion above 475 °C with H₂ feed recovery >60%. The results achieved pave the way towards a possible substitute for the ceramic-supported alternatives.

1. Introduction

Hydrogen is a key component in many industrial processes and is a clean energy carrier that has the potential to replace fossil fuels in various applications [1,2,3,4,5]. However, the current methods employed for its production are primarily based on non-renewable sources and emit significant amounts of greenhouse gases [6]. To address this issue, alternative and more sustainable pathways for hydrogen production are nowadays explored. While green hydrogen can be easily produced via water electrolysis, hydrogen storage and transportation still pose challenges due to the lack of infrastructure and subsequent requirement of high-pressure storage.

A promising route for the storage, transportation and on-site production of sustainable hydrogen consists in the decomposition of green ammonia [7,8,9]. This process, which involves the breakdown of ammonia (NH₃) into nitrogen (N₂) and hydrogen (H₂), has in fact several advantages compared to conventional systems for hydrogen production, the most important being that the process is not responsible for carbon emissions [10,11,12,13]. Furthermore, hydrogen production from ammonia decomposition allows to overcome the challenges related to hydrogen storage and distribution. The ease of liquefaction of ammonia compared to compressed hydrogen, its lower cost per unity of energy stored and its already existing infrastructure for storage and transportation make in fact this molecule a promising energy carrier and make ammonia decomposition an attractive solution for decentralized hydrogen production [8,9,10]. Particularly, the fact that ammonia is a carbon-free molecule, makes ammonia attractive for the production of carbon-free hydrogen to be used in carbon-sensitive applications such as PEM fuel cells [10,11,12,13,14,15,16,17,18,19].

To produce hydrogen from ammonia, first ammonia must be decomposed into hydrogen and nitrogen (according to Equation (1)), then hydrogen must be separated from nitrogen and unconverted ammonia. While these two steps require dedicated units in a conventional process, they can be simultaneously performed in a membrane reactor with advantages in terms of compactness of the system. Moreover, while ammonia conversion in a conventional reactor is favored at high temperature and low pressure, the use of a membrane reactor has proved to reduce the reactor operating temperature and increase the operating pressure compared to a conventional system, with advantages in terms of energy efficiency (due to low temperature) and compactness of reactor (due to high pressure) [15,20,21,22,23]. The selective separation of hydrogen from the reaction zone - which is favoured at high pressure - enhances in fact the reaction kinetics and shifts the reaction equilibrium towards the reaction products thereby increasing the conversion of the feedstock.

$${\mathrm{N}\mathrm{H}}_3\rightleftarrows 0.5{\mathrm{N}}_2+1.5{\mathrm{H}}_2\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}{∆\mathrm{H}}_{\mathrm{f}}^{\mathrm{o}}=45.92\frac{\mathrm{k}\mathrm{J}}{\mathrm{m}\mathrm{o}\mathrm{l}}$$

(1)

Hydrogen production via ammonia decomposition has already been experimentally investigated in several works available in literature [15,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35] and the best performance in terms of NH₃ conversion, hydrogen recovery and purity of hydrogen produced have been obtained when using a Ru-based catalyst to promote the ammonia decomposition reaction and ultra-thin ceramic supported “double skin” Pd-based membranes for hydrogen separation. Ceramic supported membranes in fact generally show very good performance for pure hydrogen separation, as the smooth surface of the ceramic support (with accurate control on porosity and narrow pore sizes distributions up to a few nanometers) facilitates the deposition of ultra-thin and defect-free palladium layers.

However, ceramic supported membranes are not easy to seal and couple to stainless-steel reactor modules [36]. Moreover, having the ceramic support a thermal expansion coefficient significantly different compared to that of palladium, they can be relatively easily damaged while working at high temperature, as cracks can form at the Pd/support interface [37]. These limitations make the commercialization of membrane reactors for ammonia decomposition more challenging. On the other hand, these challenges could be overcome using metallic supported membranes.

Steel and steel alloys materials have been investigated as potential candidates for a new generation of Pd-based membranes supports [38,39,40,41]. They in fact require only a simple weld to be sealed and are way less prone to breakage or crack formation thanks to their high mechanical stability. However, metal-based supports are characterized by rough surfaces, large superficial pore mouths, and wide pore size distribution, making membrane gas-tightness and H₂ selectivity quite difficult to achieve during the Pd deposition step [42]. Metallic supports with the desired superficial characteristics therefore require extremely low media grades and specific supplier operated treatments, which increase their final costs. The challenging prices of metallic supports with suitable surface characteristics make the development of pre-treatments on low-cost steel-based filters a necessity for the commercialization of the MR ammonia cracking technology [33]. For this reason, this work focuses on modifying raw and cheaper tubular metallic filters and on converting them into a membrane support by filling the large, non-uniform pores with α-alumina particles of decreasing size.

Additionally, steel-based membrane supports present strong interaction with the deposited Pd film, a phenomenon known as intermetallic diffusion [43] (or Pd-support interaction). The diffusion of Pd in the support and vice versa can therefore hinder membrane H₂ permeation performance, as well as long-term stability [41,43,44]. This additional challenge can be overcome by the addition of a ceramic-based (commonly Al₂O₃ [45,46], CeO [47,48], YSZ [38,49], ZrO₂ [50,51]), interdiffusion barrier between the Pd film and the metallic support itself [52], which proves essential for the membrane’s functioning in reaction environment. Moreover, the addition of dopant metals, (e.g., Au, Cu) to the selective Pd layer could improve the membrane’s stability to a few ppm of feedstock pollutants such as H₂S. However, it would not influence its thermal stability or separation performance, having the same effect as Ag addition [53,54]. In this work, we characterize a promising preparation procedure for Pd-Ag membranes on low-cost, rough Porous Hastelloy X filters, consisting in a support filling step with α-Al₂O₃ of decreasing size, and a boehmite based coating as interdiffusion barrier. By applying all the selected pre-treatments, we aim to demonstrate that hydrogen selective membranes fabricated from rough, porous Hastelloy X filters can be used for ammonia cracking applications, offering a first alternative to conventional ceramic supported membranes. The large pore size distribution and surface roughness of the selected filters renders the pre-treatments of the supports particularly challenging. However, the ceramic filler particle size and amount of filling cycles prove easily tunable parameters for the further improvement of membrane permeance, paving the way towards a standardized development procedure for Pd-based membranes supported on this kind of filters. Following the procedure characterization, we then test one of the membranes to assess its suitability for ammonia decomposition applications in a fixed bed membrane reactor.

2. Results and Discussion

2.1. Membrane Preparation

In Figure 1, the surface evolution of the membrane support filter with each pre-treatment is shown. The untreated filter presented superficial pore mouths larger than 50 µm, which were partially closed during the wet-polishing treatment (Figure 1a,b). At this stage, the surface roughness of the filter (Ra = 6.1 µm) was considerably reduced (Ra = 1.1 µm). However, the superficial porosity of the material was lost due to plastic deformation. In order to recover the gas permeance through the media and preserve the reduction in surface roughness, the polished layer was cracked open via chemical etching, as confirmed by the superficial cracks in Figure 1c. These cracks promote gas permeation while preserving the smoothening and closing effect of the polishing treatment. However, some of the largest pore mouths remained untouched by the selected pre-treatments. The depth of these valleys was filled with α-Al₂O₃ of decreasing size, promoting both surface leveling and pore size distribution reduction (Figure 1c). After the deposition of the interdiffusion barrier (Figure 1d), the average surface roughness, Ra, was reduced to 0.8 µm. In Figure 2 the cumulative flow distribution through the filter, obtained via CFP after each modification step, is shown. After the polishing and etching pre-treatments, the filter presented wide flow distribution and high permeance. However, after asymmetrically filling with α-Al₂O₃, the detected average pore diameter was 230 nm. The deposition of the interdiffusion barrier promoted further sharpening of the flow distribution towards pores of 90 nm. This value proved comparable to the target of 100 nm, representing the average pore diameter of a ceramic support for highly selective Pd-Ag membranes [55] for ammonia decomposition.

In Figure 3, the surface of a twin membrane is observed via confocal laser and optical microscopy. Particularly, the selected superficial section is fully closed by the deposited Pd-Ag layer. This is possible due to the filter pre-treatments and two-step modification, which guarantee the formation of the fully dense Pd-Ag layer on the selected filters.

2.2. Single Gas Permeation Tests

The measured hydrogen and nitrogen permeances at 400, 450 and 500 °C and 1 bar pressure difference across the membrane are listed in

Table 1. H₂ and N₂ permeance and ideal H₂/N₂ selectivity of the membrane used in this work at 400, 450, 500 °C and 1 bar pressure difference across the membrane.

Temperature [°C]	H2 Permeance [mol/s/m2/Pa]	N2 Permeance [mol/s/m2/Pa]	H2/N2 Ideal Perm-Selectivity [-]
400	5.8 × 10−7	1.1 × 10−10	5287
450	6.6 × 10−7	1.1 × 10−10	5892
500	7.3 × 10−7	1.9 × 10−11	38,839

together with the corresponding ideal perm-selectivity. The experimental results show that H₂ permeation significantly increases with increasing temperature, whereas N₂ permeation is less significantly affected by a temperature variation. This results in increasing H₂/N₂ ideal-perm-selectivity when increasing temperature.

In Figure 4 the hydrogen flux through the membrane at different temperatures is represented as a function of the transmembrane pressure difference. The best fit for the hydrogen fluxes and hydrogen partial pressure difference across the membrane (${∆P}_{H_2}^n)$ was found for a pressure exponent value of 0.69. In Pd-based membranes, the pressure exponent is 0.5 when the rate limiting step is diffusion through the bulk of palladium. A value of n deviating from 0.5 may indicate the presence of a contribution of the metallic support to the H₂ transport mechanism, external mass transfer limitations, or limitations in the surface reactions [56,57].

From the hydrogen fluxes through the membrane measured in the temperature range between 400 °C and 500 °C the activation energy for hydrogen permeation through the membrane was then estimated to be 9.1 kJ/mol from the Arrhenius’ plot reported in Appendix A (Figure A3). This value is well in agreement with previously reported values of apparent activation energy for Pd-based membranes. The calculated value for activation energy lumps the effect of both the activation energy required for hydrogen permeation through the selective layer of the membrane and of the activation energy required for hydrogen to permeate through the membrane support.

2.3. Binary Mixture (H₂/N₂) Permeation Tests

In Figure 5a the effect of a different hydrogen concentration in the feed mixture on the separation performance of the investigated Pd-based membrane is shown. Specifically, the permeation flux through the membrane is represented as a function of the hydrogen partial pressure difference across the membrane for a H₂/N₂ feed mixture containing a hydrogen concentration ranging between 50% and 95%. The results of the single gas hydrogen permeation tests are also reported and used as benchmark for comparison. The membrane separation performance is different compared to the case in which the pure hydrogen permeation test was performed. When N₂ concentration in the feed mixture increases, a lower amount of hydrogen is available for separation and therefore, at constant pressure difference across the membrane, the hydrogen permeation flux through the membrane decreases. Moreover, a lower hydrogen partial pressure in the feed mixture results in lower driving force for hydrogen separation which in turn leads to lower hydrogen permeation. This discrepancy in the results arises from mass transfer limitation phenomena, known as concentration polarization, from which Pd-Ag membranes suffer. In line with literature, mass transfer limitations become more remarkable when the hydrogen concentration in the feed mixture decreases as well as when the pressure difference across the membrane increases. At higher total pressure difference across the membrane, the mass transfer limitation increases due to the higher flow through the membrane and the higher recovery. As a result of concentration polarization, a lower amount of hydrogen is recovered at constant hydrogen partial pressure difference across the membrane.

To investigate the effect of concentration polarization on hydrogen permeation, a relative flux was calculated as the ratio between the hydrogen flux measured when feeding the H₂/N₂ mixture and the hydrogen flux measured under pure hydrogen permeation tests. The relative fluxes as a function of the concentration of N₂ in the feed are plotted in Figure 5b. It can be concluded that the membrane suffers from quite reduced flux even for low N₂ concentration in the feed. Moreover, the higher the temperature the higher the mass transfer limitation effect. All Pd-based membranes show in fact higher hydrogen fluxes for increasing temperatures and the higher amount of hydrogen that permeates the higher the mass transfer limitation effect.

2.4. Permeation Tests under Reactive Conditions

The effect of temperature and pressure at which the ammonia decomposition reaction takes place on the performance of the membrane reactor has been investigated and the results are presented in

Table 2. NH₃ conversion and H₂ recovery at different reaction temperatures.

	Conventional System	Membrane Reactor
Temperature [°C]	Thermodynamic Equilibrium Conversion [%]	NH3 Conversion [%]	H2 Recovery [%]
425	97.0	84.2	37.6
450	97.8	98.2	55.5
475	98.3	99.2	60.7
500	98.7	99.3	62.9
Reaction pressure = 5 bar, NH3 feed flow rate = 0.5 LN/min

and

Table 3. NH₃ conversion and H₂ recovery at different reaction pressures.

	Conventional System	Membrane Reactor
Pressure [bar]	Thermodynamic Equilibrium Conversion [%]	NH3 Conversion [%]	H2 Recovery [%]
3	99.0	99.2	42.7
4	98.7	99.2	51.9
5	98.3	99.2	60.7
6	98.0	99.1	66.1
Reaction temperature = 475 °C, NH3 feed flow rate = 0.5 LN/min

, respectively.

From Table 2 it is possible to see that a temperature increase results in higher ammonia conversion as well as in higher recovery. The increase in ammonia conversion can be ascribed to the more favorable kinetics and thermodynamics of reaction when operating the reactor at high temperature, whereas the improved hydrogen recovery results from the higher hydrogen partial pressure in the reactor which in turn leads to higher driving force for hydrogen separation. Moreover, from Table 3 and Figure 6, it can be seen that NH₃ conversion in the membrane reactor is higher than the calculated thermodynamic equilibrium conversion of the conventional reactor (without H₂ separation membrane) for temperatures higher than 450 °C. Overall, these results are in agreement with other literature studies [15,20,21,22,23].

More specifically, a comparison between the results shown in Figure 6 and the ones obtained in our previous work [23] under similar operating conditions using a ceramic supported membrane is presented in

Table 4. Comparison between the experimental results achieved in this work and in the work of Cechetto et al. [23].

	This Work	Cechetto et al. [23]
Membrane
Configuration	Supported tubular Pd-based membrane	Supported tubular Pd-based membrane with a porous Al2O3-YSZ protective layer
Support	Metallic (Hastelloy X)	Ceramic (Al2O3)
Selective layer composition	Pd-Ag	Pd-Ag
Selective layer thickness [μm]	$~$6–8	$~$6–8
Length [mm]	90	195
H2 permeance at 450 °C and 1 bar [mol/s/m2/Pa]	6.6 × 10−7	1.2 × 10−6
H2/N2 ideal perm-selectivity at 450 °C and 1 bar [-]	5890	68,960
NH3 conversion [%] *
T = 425 °C	84.2	96.5
T = 450 °C	98.2	99.7
T = 475 °C	99.2	99.8
T = 500 °C	99.3	99.8
H2 recovery [%] *
T = 425 °C	37.6	79.5
T = 450 °C	55.5	87.5
T = 475 °C	60.7	88.9
T = 500 °C	62.9	88.9

* Reaction pressure = 5 bar, NH₃ feed flow rate = 0.5 L_N/min, GHSV = 120 mL/(g_cat·h).

. The lower H₂ permeance and length of the membrane used in this work (10 cm, due to ease of preparation and handling) result in a decreased H₂ recovery and NH₃ conversion—the latter being particularly evident at the lowest temperature of 425 °C, while reaching >99% at T > 475 °C. Highly selective metallic supported membranes display in fact lower H₂ permeance compared to the ceramic supported equivalents. This difference is to be attributed to the intrinsic difference in supports porosity, the necessity for a filler to reduce the metallic support’s large pore size distribution and, finally, the addition of a barrier to prevent Pd-support interaction phenomena [49,58].

In view of these results, this paper demonstrates for the first time that, despite the need to scale-up membrane length and possibly improve their H₂ permeation properties, the proposed pre-treated metallic supports are a suitable alternative to ceramic ones for the fabrication of Pd-based, H₂ selective membranes used in ammonia decomposition membrane reactors.

As it is possible to see from Table 3, hydrogen recovery can be improved increasing the reactor operating pressure, i.e., increasing the driving force for hydrogen separation. It is worth noticing that, while in a conventional system a pressure increase has a significant negative impact on NH₃ conversion, in the investigated pressure range in the membrane reactor the decrease in NH₃ conversion due to increasing pressure is minor. The decrease in NH₃ conversion that is expected at high pressure according to the Le Châtelier’s principle is in fact counterbalanced by the fact that a pressure increase improves the hydrogen removal from the reaction zone resulting in faster kinetics and shifted thermodynamics which in turn enhance ammonia conversion.

In case of application in PEM fuel cells, according to ISO 14687:2019 [59], NH₃ concentration in the hydrogen stream must not exceed 0.1 ppm. While the residual NH₃ concentration in the hydrogen stream has not been evaluated in this work, it has already been demonstrated in literature that commercially available adsorbent materials [12,16,31,60,61,62,63] or ion-exchange forms of different type of zeolites [63,64,65,66] can be used to reduce the residual ammonia concentration in the hydrogen stream to levels that are suitable for PEM fuel cell application.

A comparison between the experimental results achieved in this work and others available in recent literature is presented in

Table 5. Overview of different types of materials used in literature as supports for the fabrication of membranes for hydrogen separation in membrane reactors for ammonia decomposition.

MEMBRANE					REACTOR OPERATING CONDITONS				CATALYST	PERFORMANCE MR		Ref.
Selective Layer Composition	Selective Layer Thickness	Length [mm]	Type of Support	Support Material (Thickness)	Temperature [°C]	Reaction Pressure [bar]	Permeate Pressure [bar]	GHSV [mL/(gcat h)]		NH3 Conversion [%]	H2 Recovery [%]
Pd	6.2	N/A	Ceramic	YSZ (130 mm)	400	5	1	N/A	Ru (impregnated in the membrane support)	98	87.5	[20]
Pd	3	N/A	Ceramic	Al2O3 (N/A)	500	5	1	2000	Ni/La-Al2O3—6 g	>99 *	92 *	[15]
Pd-Ag	$~$6–8	195	Ceramic	α-Al2O3 (3.5 mm)	500	6	1	120 **	(2 wt.%) Ru/Al2O3—250 g	99.8	91.6	[31]
Pd-Ag	4.61	202	Ceramic	α-Al2O3 (2 mm)	400	4	Vacuum	120 **	(2 wt.%) Ru/Al2O3—250 g	99.3	93.5	[23]
Pd-Au	8	186 (*)	Ceramic	N/A	485	5	1	1200 **	(0.5 wt.%) Ba-CoCe—10 g	>99 *	92 *	[32]
Pd	6.5	N/A	Metallic	Stainless steel (+MnCO3)	400	3	1	1880 **	(5 wt.%) Ru/MgO—1.5 g	99.8	N/A	[29]
Pd-Ag	1.8	100 (*)	Ceramic	N/A	450	7	1	5000	(3 wt.%) Ru/Y/K/Al2O3—3 g	99.11	90.6	[30]
Pd	200	65	NS	N/A	450	1	Vacuum	1164 **	(5 wt.%) Ru/SiO2—0.5 g	87 *	59 *	[21]
Pd	2	90	Ceramic	α-Al2O3 (N/A)	375	1	Vacuum	680 **	(2 wt.%) Ru/Al2O3—0.88 g	>99 *	-	[22]
Pd	$~$5	450	Metallic	Inconel 600 (N/A)	430	5	Vacuum	N/A	(2 wt.%) Ru/Al2O3—200 ml	99.4	97.5	[33]
Pd-Ag/V-Fe	~0.2 µm Pd-Ag ~100 µm V-Fe	N/A	Metallic	V-10 mol.%-Fe alloy (~100 µm V-Fe)	350	3	1	3000 **	(5 wt.%) Ru/Cs2O/Pr6O11—0.2 g	89 *	89 *	[35]
Pd/Ta	~ 0.4 µm Pd ~250 µm Ta	N/A	Metallic	Tantalum (~250 µm)	450	6.5	1	30,000	(1.6 wt.%) Ru/La-Al2O3—6 g	>99.5	N/A	[28]
Pd/Ta/Pd	$~$1–2 µm Pd ~250 µm Ta	N/A	Metallic	Tantalum ($~$250 µm)	500	5	1	6000	(0.65 wt.%) Ru/La-Al2O3—1 g	95 *	86 *	[27]
Pd	$~$13	156	Ceramic	α-Al2O3 (0.5 mm)	500	3	1	135 **	(70 wt.% ) Ni/γ-Al2O3—29 g	99 *	80 *	[25]
Pd	4.23	73	Ceramic	YSZ (130 mm)	450	5	1	1200 **	(0.5 wt.%) Ru/Al2O3—5 g in the catalyst bed (1.9 wt.%) Ru/YSZ—Ru impregnated in the membrane support	>99	>90	[34]
Pd	N/A	N/A	Metallic	N/A	450	5	1	N/A	(-) Ru/Al2O3—N/A	>99 *	91 *	[24]
Pd-Ag	$~$6–8	$~$90	Metallic	Hastelloy X	500	5	1	1200	(2 wt.%) Ru/Al2O3—250 g	99.1	66.1	This study

* Data not directly reported in the publication and retrieved from graphic representation of experimental results; ** Not directly reported in the publication. Calculated based on provided information about catalyst and flow rates used; NS = Non Supported; N/A = Not Available.

. The table lists information about the configuration of the membrane used for hydrogen separation, operating conditions of the membrane reactor, catalyst employed, and the best results achieved in each work. By “best results” we intend the results that facilitated efficient membrane reactor operation, which includes NH₃ conversion greater than 99% and H₂ recovery above 90%. Alternatively, when optimal reactor operation was not achieved, the results that were the closest to this target were reported. The results achieved in this work demonstrate that hydrogen selective membranes fabricated from low-cost porous Hastelloy-X filters offer an alternative to conventional ceramic and other metallic supported membranes. Similarly to other works available in literature, in fact, NH₃ conversion >99% was achieved. On the other hand, the H₂ recovery achieved in this work is sensibly lower compared to the one achieved in other studies. This can be related to the combination of multiple aspects, which are mainly related to the fact that the performance of a MR for H₂ production from NH₃ not only depend on the reactor operating conditions, but also on the membrane properties and separation performance. Specifically, while the large pore size of the filter requires filler introduction to ensure membrane selectivity, an additional barrier is needed to prevent metallic interdiffusion, which contributes to relatively low hydrogen permeation. Furthermore, the membrane area installed in this study is relatively low (membrane length ~10 cm), which also results in lower H₂ recovery factors. Given the different operating conditions at which the experimental results reported in Table 5 have been achieved, a direct quantitative comparison between the results achieved in this work and works available in literature proves challenging. Nonetheless, H₂ recoveries comparable to values achieved in literature could be achieved by compensating for the reduced H₂ permeance with larger installed membrane area. Additionally, as the NH₃ conversion enhancement compared to a system in which no membrane is used is proportional to the H₂ permeation through the membrane, the installation of a larger membrane area is also expected to further increase NH₃ conversion, which in turn would also contribute to favor H₂ recovery.

Lastly, in order to show the impact of membrane exposure at high temperature and to ammonia decomposition reaction on the membrane separation performance, in Figure 7 the ideal H₂/N₂ perm-selectivity of the membrane at 450 °C and 4 bar(a) is represented as a function of time. As it is possible to see, the ideal H₂/N₂ perm-selectivity decreases over time, indicating that during the experimental campaign defects have formed on the membrane surface or on the sealings, resulting in permeation of a higher amount of impurities (N₂) through the membrane walls. Specifically, over a time of 360 h the H₂/N₂ ideal perm-selectivity decreased from 4807 to 3017. This decrease could be improved both by tuning the support pre-treatments (e.g. the α-Al₂O₃ filler particle size and filling cycles combination) and interdiffusion barrier (e.g. layer thickness) or via deposition of a “double-skin” mesoporous layer (γ-Al₂O₃-YSZ) onto the Pd surface, similarly to the membrane tested in our previous work ([5]) in order to fully close possible defects forming on the Pd layer.

3. Materials and Methods

3.1. Membrane Preparation

A commercial unrefined porous Hastelloy X filter with an outer diameter of 1.2 cm, average surface roughness (Ra) of 6.1 µm, and 0.5 µm nominal media grade (MG) was acquired from Hebei Golden Flame Wire Mesh Co, China. The filter was cut to 10 cm length and welded to dense stainless steel (AISI316L) tubes, to achieve a dead-end configuration (Figure 8).

The filter’s surface was then modified according to the procedure reported below and illustrated in Figure 9:

• The rough filter was polished in an industrial surface finishing machine via wet-polishing mechanism (ERBA EVT-170) for 6 h, delivering a suitable trade-off between surface roughness reduction and gas permeation preservation [42]. The polished support was then vertically submerged in aqua regia for 30 s, to recover the lost superficial porosity. After the acid attack, the filter was thoroughly rinsed with deionized water to remove all mordant residuals. An oxidation in static air atmosphere was then performed in a furnace for 1 h at 750 °C, with heating rate 2 °C/min.
• The support’s superficial pore size was improved by filling asymmetrically with α-Al₂O₃ of decreasing particle size via immersion in a magnetically stirred α-Al₂O₃-H₂O suspension, improved by addition of HNO₃ (67 vol.%) dropwise. The filler was pulled through the superficial pores via vacuum-assisted dip coating, with a lower wait time of 60 s per immersion cycle. Between each cycle the support was gently rinsed with distilled water. 20 aspiration cycles were performed with alumina 18 µm (AA-18, Sumitomo), 10 aspiration cycles with alumina 5 µm (AA-5, Sumitomo) and 10 with alumina 1.5 µm (AA-1, Sumitomo).
• Finally, a mesoporous smoothening interdiffusion barrier was deposited to complete the improvement of support’s surface uniformity. A solution with boehmite loading 0.9 wt.% was prepared in distilled water, incorporating a water-based solution of organic additives, namely 3.5 wt.% polyvinyl alcohol (PVA) (MW 130,000) and 1 wt.% polyethylene glycol (PEG) (MW 400). The deposited layer was dried under rotation in a climate chamber at 40 °C and 60% relative humidity for 1 h and sintered for 1 h at 550 °C in a static air furnace.

A layer of Pd-Ag alloy is deposited onto the treated support via a 5 h electroless plating procedure, reported in a previous work by Tanaka et al. [67,68] up to a 6–8 µm thick Pd-Ag layer.

3.2. Membrane Characterizations

The evolution of support’s surface parameters was monitored through the membrane’s preparation procedure with different techniques. The surface roughness of the support was measured via contact profilometry (MarSurf PS 10). The superficial pore diameter distribution of the untreated, pre-treated and modified Hastelloy support were measured by capillary flow porometry (CFP or gas-liquid displacement) technique, in the setup reported in Appendix A, Figure A1. The surface of twin sample supports was observed via Scanning Electron Microscopy (SEM, Phenom Pro, ThermoFisher, The Netherlands). In order to preserve the metallographic structure of the porous metal, the samples were prepared via scoring and breakage of the membrane and observed as is. The membrane surface after Pd deposition was then observed via optical-laser confocal microscopy (VKX-3000, Keyence, Japan), to avoid sample breakage.

3.3. Experimental Setup for Permeation and Ammonia Decomposition

Ammonia decomposition tests have been performed in a tubular membrane reactor in which the membrane was fully immersed in a packed bed of catalyst. The reactor is made of stainless-steel and has an inner diameter of 4.5 cm and a length of 28 cm. Since ammonia decomposition is an endothermic reaction, heat is supplied to the reactor by an electrical split oven with three heating sections independently controlled. A porous stainless-steel plate at the bottom of the reactor was used to ensure uniform gas distribution at the reactor inlet, whereas at the freeboard of the catalytic bed the reactor has a conical shape to reduce the gas velocity and minimize the risk of catalyst particle escape during the experiments. The feed gases (NH₃, H₂ and N₂) were controlled by mass flow controllers (Bronkhorst) and the pressure of the system was controlled by means of a back pressure regulator. The permeate side of the membrane was connected to a film-flow meter (Horiba Stec VP3/VP4) to determine the permeation flux. The retentate side of the membrane, after passing through a gas filter which aims at protecting the downstream equipment from fine particles, was sent to a μ-GC (Varian CP-4900) to measure its composition. The retentate and permeate lines were subsequently mixed and sent to a water absorption unit in which possible traces of NH₃ are absorbed preventing their release in atmosphere. In this study, for hydrogen separation the metallic supported Pd-based membrane described in the previous section has been used. The membrane was fully immersed in a packed bed of 250 g of a commercial (2 wt.%) Ru/Al₂O₃ catalyst in pelletized form (3 mm) from Alfa Aesar. Characterization of such a catalyst before and after test has been previously discussed (supplementary data of [23]). A schematic representation of the setup used in this work is given in Appendix A, Figure A2.

3.4. Experimental Methods

Before its integration in the reactor, the membrane was sealed with graphene tape and a pressed 1.3 cm stainless steel ring, in order to exclude any possible leaks from the welding cord at the dead-end cap. The membrane was then tested in helium/ethanol in order to verify the absence of undesired leakages from both the sealing and Pd-Ag surface. As no leakages were detected, the membrane was subsequently installed in the membrane reactor. The reactor was then heated up to 500 °C at a heating rate of 2 °C/min in N₂ atmosphere and activated by feeding the reactor with a hydrogen stream, as indicated in [69,70]. The activation was considered complete when hydrogen permeation was measured to be stable. The following tests have subsequently been performed: single gas (H₂ and N₂) and binary mixture (H₂/N₂) permeation tests, as well as permeation tests under reactive conditions. More details about the membrane testing procedure are given in the following sub-sections and an overview of the experimental conditions investigated in this work is presented in

Table 6. Overview of the experimental conditions investigated in this work.

Single gas permeation tests
Single gases investigated	H2, N2
Temperature [°C]	400, 425, 450, 475, 500
Retentate pressure [bar]	2, 3, 4, 5, 6
Permeate pressure [bar]	1
Binary mixture permeation tests
Binary mixture	H2/N2
Temperature [°C]	400, 425, 450, 475, 500
Retentate pressure [bar]	2, 3, 4, 5, 6
Permeate pressure [bar]	1
Ammonia decomposition
Temperature [°C]	425, 450, 475, 500
Retentate pressure [bar]	3, 4, 5, 6
Permeate pressure [bar]	1
NH3 feed flow rate [LN/min]	0.5

. In order to assess the membrane’s stability during exposure at high temperature and to ammonia decomposition reaction, single gas H₂ and N₂ permeation tests have been periodically performed at 450 °C and 4 bar(a) during the experimental campaign.

3.4.1. Single Gas (H₂ and N₂) Permeation Tests

Pure H₂ and N₂ gas permeation measurements were carried out in order to determine the ideal H₂/N₂ perm-selectivity of the membrane. The H₂ and N₂ permeation fluxes were measured for temperatures ranging between 400 °C and 500 °C and pressures ranging between 2 and 6 bar(a), keeping the permeate side of the membrane at atmospheric conditions. The H₂ and N₂ permeances, the ideal H₂/N₂ perm-selectivity as well as the activation energy for hydrogen permeation through the membrane were then calculated.

3.4.2. Binary Mixture (H₂/N₂) Permeation Tests

Binary mixture permeation tests were performed in order to compare the hydrogen flux available at the permeate side of the membrane at different hydrogen concentrations in the feed mixture. Specifically, permeation tests were carried out for H₂/N₂ mixture containing H₂ concentrations ranging between 50 vol.% and 95 vol.%. For each binary mixture composition, the hydrogen permeation through the membrane was evaluated for temperatures and pressures ranging between 400 °C and 500 °C, and 2 bar(a) and 6 bar(a), respectively. The permeate side of the membrane was kept at atmospheric conditions.

3.4.3. Permeation Tests under Reactive Conditions

Ammonia decomposition has been performed for temperatures ranging between 425 °C and 500 °C, pressures in the range between 3 bar(a) and 6 bar(a) and under a feed flow rate of 0.5 L_N/min of pure ammonia. The permeate side of the membrane was kept at atmospheric conditions. The experiments have been carried out according to the following procedure. At the desired temperature, pure ammonia was fed into the reactor and the operating pressure was adjusted. The reaction performance was then monitored until steady state operation was observed. The permeate flow rate and the composition of the stream leaving at the retentate side of the membrane were then measured 5 times and from each measurement NH₃ conversion and H₂ recovery were calculated. The values of NH₃ conversion and H₂ recovery reported in this work for a given combination of experimental conditions are the average of the 5 calculated values of NH₃ conversion and H₂ recovery. NH₃ conversion was calculated according to equation 2 as the ratio between the amount of ammonia that in the reactor decomposes into hydrogen and nitrogen and the amount of ammonia that is fed to the reactor, whereas H₂ recovery was calculated according to equation 3 as the ratio between the amount of hydrogen permeating through the membrane and the total amount of hydrogen that is fed into the reactor in the form of ammonia.

$${\mathrm{NH}}_3\mathrm{conversion}=\frac{{\mathrm{N}\mathrm{H}}_{3,\mathrm{i}\mathrm{n}}-{\mathrm{N}\mathrm{H}}_{3,\mathrm{o}\mathrm{u}\mathrm{t}}}{{\mathrm{N}\mathrm{H}}_{3,\mathrm{i}\mathrm{n}}}$$

(2)

$${\mathrm{H}}_2\mathrm{recovery}=\frac{{\mathrm{H}}_{2,\mathrm{p}\mathrm{e}\mathrm{r}\mathrm{m}\mathrm{e}\mathrm{a}\mathrm{t}\mathrm{e}}}{1.5{\mathrm{N}\mathrm{H}}_{3,\mathrm{i}\mathrm{n}}}$$

(3)

4. Conclusions

An innovative membrane, consisting of a Pd-Ag selective layer successfully supporting a 6 μm rough metallic filter with 0.5 μm media grade and 50 μm pore mouths, has been prepared. The selected pre-treatments proved suitable to achieve a support with a superficial pore size of ~90 nm, resulting in the membrane’s H₂/N₂ ideal selectivity at 500 °C and 1 bar of ~38,000. The H₂/N₂ mixture permeation tests showed an effect of concentration polarization on the H₂ permeation across the membrane, resulting in lower H₂ fluxes when N₂ concentration in the feed is increased, well in agreement with Pd-based membrane behavior reported in the literature. The tests in an ammonia cracking environment showed the ability to overcome the conventional thermodynamic conversion of NH₃ when the selected membrane is introduced, reaching NH₃ conversions >99% for temperatures including and above 475 °C. The hydrogen recovered from the feed amounts to >60% for the same temperature range, considering the membrane being solely 10 cm long.

Finally, a stability test performed at 4 bar for 360 h showed a decrease in the corresponding membrane ideal H₂/N₂ selectivity from ~4800 to ~3000, showing possible defect formation in long-term conditions, highlighting the need for further tuning of the metallic support prior to treatments.

Overall, the promising results show that the selected membrane preparation procedure is proven suitable for the further development of Pd-based membranes for NH₃ cracking on steel/steel alloy supports.