Adsorbents development for hydrogen cleanup from ammonia decomposition in a catalytic membrane reactor

.


Introduction
While hydrogen is regarded as the ideal energy carrier, the difficulties associated to its storage and distribution are challenges that have so far prevented the widespread of industrially promising hydrogen based technologies for power production [1], such as proton exchange membranes fuel cells (PEMFCs).The commercialization of such technologies on an industrial scale, which would allow to make a big step in the direction of a more sustainable future, requires in fact improvements in the infrastructure for hydrogen supply.In order to overcome the challenges related to hydrogen transportation and storage, one alternative consists in storing hydrogen in the form of liquid fuels [2][3][4] which could be then decomposed to produce hydrogen when required [5,6].In this regard, ammonia has recently gained attention as the potential ideal candidate for the storage of hydrogen [3,6,15,[7][8][9][10][11][12][13][14] to be used as feedstock for PEMFCs, [16][17][18][19][20][21][22][23][24] as the absence of carbon in its molecular structure minimizes the risk of CO poisoning of the PEMFCs electrodes [25] and prevents the release of CO 2 .Moreover, its relatively low cost, ease of liquefaction as well as its already existing infrastructure for storage and transportation allow for an economically competitive, relatively easy and safe hydrogen storage and transportation [3,6,[10][11][12].
In order to recover hydrogen from ammonia, ammonia decomposition into hydrogen and nitrogen has to occur according to Eqn. (1) and hydrogen has then to be separated and purified from nitrogen and possible traces of unconverted ammonia.While in conventional systems for hydrogen production from ammonia these two steps are carried out into two separate units, the membrane reactor technology allows them to simultaneously take place in a single device with higher efficiencies as well as lower costs [5,8,9,[26][27][28][29][30][31][32][33][34][35].
Membrane reactors for ammonia decomposition have already been experimentally investigated in literature, generally employing ruthenium-based (Ru-based) catalysts to promote the ammonia decomposition reaction and palladium-based (Pd-based) membranes for hydrogen separation [5,7,8,14,23,28,30,32,36]. It has been demonstrated that with a membrane reactor it is possible to achieve higher ammonia conversions than in conventional system even at lower temperatures while simultaneously recovering large part of the hydrogen produced [8,28,30].However, the NH3 concentration in the hydrogen stream to be used as feedstock in PEMFCs must not exceed 0.1 ppm as higher concentrations would poison the fuel cell electrodes.Since this specification is not always met when hydrogen is produced via ammonia decomposition in a membrane reactor [5,8,14], strategies must be implemented in order to increase the hydrogen purity by reducing the residual ammonia concentration in the hydrogen stream.In our previous work [7], it was experimentally demonstrated that a possible alternative to produce a ultra-pure hydrogen stream consists in a small adsorption unit downstream the membrane reactor.Specifically, a bed of zeolite 13X at ambient conditions was demonstrated to be effective for hydrogen purification from residual ammonia.Objective of this work is to investigate the performance of different sorbent materials for hydrogen purification from ammonia and to ultimately propose the best performing one in terms of adsorption capacity, regenerability and stability for its further scale-up.The purification of hydrogen produced via ammonia decomposition from residual NH 3 concentration to achieve the 0.1 ppm threshold has in fact not been widely studied yet in literature.Miyaoka et al. [37] demonstrated that a lithium exchanged X-type zeolite (LiX zeolite) can be used to purify a hydrogen stream produced via ammonia decomposition containing about 1000 ppm of ammonia while being stable along 30 cycles when thermally regenerated at 623 K.The adsorption capacity of LiX zeolite at 1000 ppm and 298 K was measured to be 3.35 mmol/g sorbent .Extending the work of Miyaoka et al., Ouyang et al. [38] investigated the dynamic adsorption of ammonia with faujasite (FAU) zeolites to purify a hydrogen stream containing 1700 ppm of NH 3 .It was demonstrated that high purity zeolite 13X (13X-HP) and zeolite LiX with low Si/Al ratio (LiLSX) could lower the NH 3 concentration to 0.1 ppm while maintaining a high flow rate and low pressure drops.The adsorption capacity of zeolites 13X-HP and LiLSX at 1700 ppm of NH 3 were measured to be 3.23 and 3.16 mmol/g sorbent , respectively.Luna et al. [39] evaluated the performance of commercially available NH 3 sorbents and ion-exchanged CuY zeolites for the removal of NH 3 from a nitrogen stream containing NH 3 concentrations ranging between 25 and 50 ppm.This study demonstrated that the adsorption capacity of CuY zeolite exceeded the one of commercially available sorbents for ammonia removal, being the adsorption capacity of CuY zeolite measured to be 3.21 and 2.82 mmol/ g sorbent for NH 3 concentrations in the nitrogen stream equal to 50 and 25 ppm, respectively.The stability of this type of zeolite was not investigated, but according to the NH 3 temperature programmed desorption (NH 3 -TPD) spectra presented in the study from Liu et al. [40] the cyclability of CuY zeolite is expected to be possible as its thermal regeneration can take place at moderate temperatures (673 K).Besides zeolites, other types of sorbents such as polymer resins, silica gels, alumina and carbonaceous materials have also been investigated in literature.However, these materials have shown lower NH 3 uptake compared to zeolites at comparable conditions and their regenerability is often problematic as chemical reagents such as HNO 3 and NaOH are required or stability issues occur [41,42].Porous organic polymer (POPs), covalent organic frameworks (COFs), hydrogen organic frameworks (HOFs) and metal organic frameworks (MOFs) have also been demonstrated to show promising NH 3 adsorption capacities at more favourable regeneration conditions compared to zeolites (373 -473 K).However, these materials require extensive post-modification and their high production costs limit their application in industry [41].Based on an extensive literature review including the abovementioned results, different forms of X-and Y-zeolites have been selected for the specific application of this work and particularly the copper-exchanged (Cuexchanged) form of the sodium and hydrogen from of Y-zeolite (NaY and HY zeolites) and the copper-exchanged form of zeolite 13X.Next to showing promising performance for ammonia removal from a hydrogen stream, these types of zeolites are in fact also relatively abundant and available at low-cost while being at the same time regenerable at moderate temperature.Moreover, as they all belong to the FAU family of zeolites, NaY, HY and 13X zeolites share the same framework structure which has some peculiar features that make them very interesting to be investigated for hydrogen purification form NH 3 .Primarily, they have a 3D porous framework consisting of sodalite cages which are connected through six-membered rings (6R) to form double-six rings (D6R) or hexagonal prisms.Their pores, which are created by adjacent supercages each one of which is reachable by twelve-membered rings (12R), have a size of 7.4 Å × 7.4 Å [43] and ammonia molecules, having a kinetic diameter of 2.6 Å, can therefore enter the pores avoiding diffusion limitations.Furthermore, the framework of FAU zeolites contains a broad selection of cation positions which can be substituted by Cu 2+ ions.Ion-exchange procedure can therefore be used to control some features of FAU zeolites such as their pore size, acidity and stability and consequently improve their performance properties for NH 3 removal from hydrogen.Different forms of ion-exchanged 13X, NaY and HY zeolites have therefore been synthesised, characterized and, subsequently, their performance in terms of total and useful capacity were assessed by means of NH 3 adsorption-desorption tests performed via Temperature Programmed Desorption (TPD) technique.As benchmark for their performance, their parent zeolites -i.e. the non-ion-exchanged form of zeolites NaY, HY and 13X -were used.Moreover, since the one-time use of sorbent materials is not economically viable, the regeneration process of the sorbents and their multicycle activity and stability were also evaluated.The most promising sorbent was then tested in a small scale fixed bed reactor in which its performance was assessed under relevant operating conditions emulating the outlet stream of a membrane reactor for ammonia decomposition.

Adsorbent materials preparation
Cu-exchanged sodium and hydrogen forms of an Y-zeolite, referred in this work as NaYCu and HYCu, respectively, were prepared following the procedure described in the work of Liu and Aika [40].Copper ions were introduced as extra framework cations into commercially available NaY and HY zeolites (SiO 2 /Al 2 O 3 = 5.1) supplied in powder form by Zeolyst International and Alfa Aesar, respectively.The preparation procedure consists of the following steps.At first, 2.5 g of zeolite powders are treated in a 250 mL round-bottom flask with 100 mL of an aqueous solution containing 0.1 M copper(II)nitrate trihydrate provided by Merck Emsure.The suspension is continuously stirred and kept at 343 K for 3 h using a thermocouple for temperature control.Subsequently, the suspension is transferred into several falcon tubes, centrifuged for 7 min at 5000 rpm and then the liquid remaining in the falcon tubes is decanted.The ion-exchanged sample in the falcon tube is then washed two times in distilled water and centrifuged under the same conditions.The falcon tubes are then placed in an oven at 353.15 K for 12 h and, once dried, the sample is transferred from the falcon tubes to a ceramic bowl, and further calcined in air at 773.15 K using a temperature ramp rate of 4 K/min to prevent thermal stress of the molecular sieve [44].The calcinated samples are pelletized using a laboratory pellet press, crushed using a mortar, and sieved to produce 0.2-0.5 mm granules to prevent significant pressure drops during adsorption experiments.In order to increase the reliability and show the reproducibility of the experimental results, the full ion-exchange procedure was followed twice, resulting in the preparation of two samples of NaYCu zeolite (refereed in this work as NaYCu1 and NaYCu2) and two samples of HYCu zeolite (refereed in this work as HYCu1 and HYCu2).Moreover, to further increase the copper loading of the sorbents, one sample of both zeolite NaY and zeolite HY was also prepared by consecutively repeating the ion-exchange procedure 5 times (samples NaYCu5x and HYCu5x).A schematic representation of the procedure for the preparation of ion-exchanged forms of NaY and HY zeolites is given in the supplementary material.
As the Cu-exchanged sodium form of X-zeolite (zeolite 13X) is reported in literature as not stable at temperatures above 473 K when ionexchange procedure is performed in presence of copper(II)nitrate trihydrate aqueous solution, another procedure proposed by Cui et al. [45] has been followed for the preparation of this form of the X-zeolite.Following this procedure, 2.5 g of zeolite 13X are dissolved in a 0.2 M copper-ammonia solution, which is formed by adding 15 mL of 28 vol% ammonium hydroxide (NH 4 OH) solution to 110 mL of 0.2 M Cu(NO 3 ) 2 • 3H 2 O aqueous solution.The suspension is continuously stirred at room temperatures for 48 h.The solution is then decanted, washed three times using distilled water and centrifuged using the same centrifuging conditions as for the ion-exchange procedure adopted to prepare NaYCu and HYCu samples.The falcon tubes are then placed in an oven at 353.15 K for 12 h and subsequently the samples are calcined at 673 K for 10 h in inert atmosphere (He).A schematic representation of the procedure for the preparation of the ion-exchanged form of NaX zeolite is given in the supplementary material.
To investigate the effect of a different calcination environment during ion-exchange procedure, a sample of zeolite 13X ion-exchanged by means of a copper-ammonia-solution was prepared performing calcination in air at 673 K for 10 h (sample 13XCuII) and a sample of zeolite NaY ion-exchanged 5 times by means of Cu-solution was prepared performing calcination in inert atmosphere (sample NaYCu5xI).Due to the low performance in terms of (multi-cycling) ammonia adsorption capacity found in this work for zeolites HY, the effect of calcination atmosphere was not investigated for this family of adsorbents.A summary of the samples that have been synthesised and used in this work is presented in Table 1.

Adsorbent materials characterization
Inductively coupled plasma optical emission spectroscopy (ICP-OES) has been used to determine the bulk chemical composition of the sorbent samples in a Spectroblue Analyzer by Spectro Analytical Instruments.The samples were prepared according to the following procedure.25 mg of sorbent were dissolved in 1.5 mL of a 1:1:1 acid mixture containing HF (40 wt%), HNO3 (65 wt%) and H2O.The solution was then further diluted with demineralized water to achieve 50 mL of solution, homogenized and finally diluted again 10 times.20-25 mL of the prepared solution were then transferred to a falcon tube for the ICP-OES measurement.
The chemical composition of the surface of the sorbents was examined using X-ray Photoelectron Spectroscopy (XPS) in a K-Alpha XPS system supplied by Thermo Scientific.Few milligrams of sorbent were deposited on a double-sided carbon tape and then transferred into the analysis chamber of the XPS machine at room temperature.The chamber was evacuated to achieve ultra-vacuum conditions (10 -8 mbar).Spectra were collected using monochromatic Al Kα X-rays at 1486.5 eV running at 72 W and a spot size of 400 μm.Spectra were acquired using a floodgun source to account for surface charging.Survey and high-resolution scans for Al, C, Cu, Na, O and Si were performed for all the sorbents.The pass energy of the survey and high-resolution scans were set at 50 eV.A total number of 30 scans in combination with a dwell time of 50 s were made for each high-resolution scan and the data was analyzed using the software CasaXPS.Before analysis of the data, the binding energy data was referenced to the C1s line (284.8eV) for charge correction.To determine the Al/Cu and Al/Na ratios as well as possible valence states of the copper species at the sorbents surface, peak deconvolution of the high-resolution spectra was performed using a symmetric pseudo-Voigt function, referred to as GL (90), GL(80) and GL (30) in CasaXPS software, after subtracting the Shirley background.The curve fitting procedure was taken from the work of Biesinger et al. [46].
X-ray Powder Diffraction (XRD) has been used to qualitatively evaluate the crystalline phases present in the sorbents.The X-ray diffraction patterns were obtained by means of a Rigaku Miniflex diffractometer using CuKα radiation (λ = 1.5418Å) at ambient conditions.Data were collected at 2θ between 3 • and 70 • using increments of 0.02 • .
The surface area, average pore size and pore size distribution of the different sorbents used in this work have been measured by N2 adsorption/desorption while applying the Brunauer-Emmett-Teller (BET) theory.Specifically, the nitrogen adsorption equilibrium isotherm at 77 K was measured using an automatic Micromeritics volumetric adsorption instrument (ASAP 2020).Before performing the pure N2 gas adsorption measurements, the sorbent samples (150 mg) were degassed by increasing the temperature to 523.15 K at a heating ramp rate of 10 K/min and by holding this temperature for 4 h.
Transmission electron microscopy (TEM) (JEOL ARM 200F) was used to study the surface morphology and particle size distribution of formed copper crystallites in the copper ion-exchanged zeolite samples.Images were recorded at magnifications ranging between 100.000 and 280.000 times.
The surface morphology of sorbent materials was investigated by scanning electron microscopy (SEM) and the bulk chemical composition of each present element was analyzed with an integrated energy dispersive X-ray detector (EDX).A Phenom ProX Desktop SEM was used with an electron optical magnification range between 160 and 5.000 times and acceleration voltage of 20.5 kV are used to obtain clear topology images.
The environment of both Cu + and Cu 2+ has been investigated performing CO adsorption measurements at 313 and 105 K, respectively, by means of Fourier-transform infrared spectroscopy (FTIR).While the CO partial pressure in the FTIR was gradually increased introducing CO pulses, 32 scans were performed at an optical resolution of 2 cm − 1 and storing data from 900 to 4000 cm − 1 in order to create and average spectra.For the CO adsorption experiments at 313 K, the pulses sequence consisted of 9 pulses of a 1:9 CO/He mixture (containing 2.5 μL of CO), 5 pulses of 25 μL of pure CO, 5 sequences of 5 consecutive pulses of pure CO and the CO adsorption experiment was then finalized with manifold pulses.For the CO adsorption experiments at 105 K, the pulses sequence consisted of 10 pulses of a 1:9 CO/He mixture (containing 2.5 μL of CO) followed by 10 sequences of 5 consecutive pulses of pure CO.
Both at 313 and 105 K, after complete saturation of the sample, the IR cell was evacuated at vacuum conditions and, subsequently, an IR measurement was conducted to record the CO irreversibly adsorbed in the sample.

Adsorbent materials testing 2.3.1. Adsorption/desorption of NH 3 via temperature Programmed desorption (TPD) technique
The total sorption capacity of the sorbents, which is defined as the maximum loading of NH 3 per unit mass of sorbent, and their adsorption/ desorption cyclability were studied performing multiple adsorption/ desorption cycles at lab-scale.Specifically, ten adsorption/desorption cycles were performed and evaluated by means of multiple NH 3 -Temperature Programmed Desorption (NH 3 -TPD) measurements performed in sequence at ambient pressure (AutoChem 2920).Before starting an experiment, each sample consisting of 60 mg of sorbent was degassed to ensure the removal of impurities.This was done by increasing the sample temperature up to 623 K at a heating rate of 5 K/min under a He flow rate of 25 mL N /min, holding for 3 h and finally cooling the sample down to ambient conditions with a cooling rate of 5 K/min.After degassing, each adsorption/desorption cycle was performed according to the following procedure.Each sample was saturated with NH 3 by feeding for 3 h at room temperature a stream of 100 mL N /min of a He/ NH 3 mixture containing 3000 ppm of NH 3 .Such relatively high concentration was chosen to limit the duration of cycling experiments for this initial screening of the adsorbent materials⋅NH 3 desorption was then carried out by increasing the sorbent bed temperature up to 623 K under a flow rate of 10 mL N /min of He with a heating rate of 5 K//min and by subsequently keeping the regeneration temperature for 30 min.The TCD signal was recorded over time both during adsorption and regeneration and, since it provides information on the NH 3 concentration at the outlet of the packed bed of sorbent, from the TCD signal recorded during adsorption a breakthrough curve was obtained.Such curve was then used to determine the total saturation and useful capacities of each sorbent being them the total area above the entire breakthrough curve and the area above the breakthrough curve before a particular threshold of NH 3 concentration at the adsorbent bed outlet is achieved (0.1 ppm in the specific case of this work), respectively.Specifically, the total saturation capacity (W A ) was calculated according to Eqn. (2) and the useful capacity (W U ) was calculated according to Eqn. (3).In these equations t end is the adsorption/regeneration time (min), c NH3 is the concentration of NH 3 (ppm) at the outlet of the packed bed of sorbent, c NH3,inlet is the concentration of NH 3 at the inlet of the packed bed of sorbent (3000 ppm), t b is the breakthrough time (min), ϕ NH3/He is the flow rate at the inlet of the packed bed of sorbent, F is a factor for the conversion from ppm to g/cm 3 and m is the total mass of sorbent sample (g).

Adsorption/desorption of NH3 in a small-scale fixed bed reactor
The sorbent showing the best performance both in terms of cyclability and useful capacity during NH 3 -TPD was further tested at ambient pressure in a dedicated setup in which it was exposed to a gas stream simulating a typical composition of the permeate stream leaving a Pdbased membrane reactor for the production of NH 3 -derived hydrogen [7,8].A schematic representation of the experimental setup used can be found in the supplementary material of this work.200 mg of sorbent were packed between two layers of glass wool in a ceramic tube (ID 8.85 mm) located inside an electrical oven.The feed gases are controlled by mass flow controllers from Bronkhorst and fed to the ceramic tube from the top.To measure the ammonia concentration of the stream leaving the adsorbent bed, the bottom side of the ceramic tube is connected to a Fourier-Transform Infrared Spectrometer (FTIR) from Shimadzu which mounts a 5 m gas cell from Specac and an MCT (Mercury-Cadmium-Telluride) detector, allowing the detection of NH 3 concentrations as low as 0.75 ppm.The gas stream leaving the FTIR is then sent to a water absorption unit before being vented to atmosphere.
Before starting an experiment, the sorbent was heated up to 623 K at a heating rate of 5 • C/min in N 2 atmosphere (200 mL N /min) and kept for 3 h at 623 K for degassing.The temperature of the bed was then decreased to ambient conditions with a cooling rate of 5 K/min in N atmosphere (200 ml N /min) and, subsequently, a FTIR background scan was performed after flushing its gas cell for 2 h with 700 mL N /min of H 2 .The sorbent adsorption and regeneration experiments were then started.Based on the results of our previous works [7,8], the permeate of a typical membrane reactor for hydrogen production via ammonia decomposition was selected to be a stream of 0.7 L N /min of hydrogen containing 10.0 ppm of NH 3 .In the experimental tests carried out in this work, this stream was reproduced by appropriately diluting a stream of hydrogen containing 95.7 ± 5 ppm NH 3 with a pure hydrogen stream.The NH 3 concentration downstream the adsorbent bed was continuously monitored with automatically performed FTIR scans until the sorbent saturation was achieved.Sorbent saturation was considered to be achieved when NH 3 concentration at the outlet of the reactor was measured to closely approach the NH 3 concentration of the feed stream (10.0 ppm).Once the sorbent saturation was achieved, the setup was first flushed at room temperature for 3 h with 700 mL N /min of H 2 and subsequently flushed with 200 mL N /min of N 2 for 30 min.The regeneration of the sorbent was then performed by heating the sorbent sample up to 623 K with a ramp rate of 5 K/min in N 2 atmosphere and keeping the temperature at 623 K for 30 min.After regeneration, the sample was cooled down to room temperature with a ramp rate of 5 K/min.Finally, the cyclability of the sorbent was investigated performing two further complete adsorption/regeneration cycles in which the sorbent was saturated under a flow of 700 mL N /min of a hydrogen stream containing 86.5 ppm of NH 3 from a calibration bottle and later regenerated as during the first adsorption/regeneration cycle.The choice to perform the saturation of the sorbent with a higher NH 3 concentration during the second and third cycles was taken to limit the duration of cycling experiments.

Bulk and surface composition
In Table 2 the bulk chemical composition of all the prepared samples obtained via ICP-OES analysis is presented together with the SiO 2 /Al 2 O molar ratio and the degree of ion-exchange of each sample.From the experimental results, it is possible to assess that copper can be loaded on zeolites 13X, NaY and HY by means of ion-exchange procedure.

V. Cechetto et al.
Moreover, the copper loading of NaY ion-exchanged samples is found to be approximatively-three times higher than the one of HY ionexchanged samples confirming that Na + cations more readily exchange copper compared to H + cations [47].Besides by an increase of copper loading, ion-exchange is also confirmed by a decrease in the Na 2 O loading of the zeolites after ion-exchange procedure.This decrease indicates in fact that some of the Na + cations available in the sorbent structure have been substituted by Cu 2+ ions.From Table 2 it can be concluded that consecutive ion-exchanges procedures do not significantly increase the copper loading of the samples.The degree of ionexchange of samples NaYCu1 -NaYCu2 and HYCu1 -HYCu2 do not significantly differ indicating that the adopted method for ion-exchange is reproducible.Finally, from Table 2 it can be observed that the ionexchanged form of NaY and HY zeolites show a lower Al content compared to their parent zeolite.This is in line with the results presented in the work of Wang et al. [48], in which it is demonstrated that dealumination takes place when ion-exchange is carried out in presence of a Cu 2+ solution followed by calcination in air.On the other hand, the introduction of ammonia hydroxide during ion-exchange procedure of zeolite 13X as well as inert calcination atmosphere seem to prevent dealumination of the sample, as after ion-exchange procedure the Al content of zeolite 13X increases and the SiO 2 /Al 2 O 3 molar ratio is not significantly affected in comparison to NaY and HY zeolites.This trend is also confirmed by the Al 2 O 3 content of the five times ion-exchanged form of NaY zeolite calcined in inert atmosphere (NaYCu5xI), which indicates that the inert atmosphere for calcination is beneficial to prevent sample dealumination.
In Table 3 the surface elemental composition including the Al/Cu, the Cu 2+ /(Cu 0 + Cu + ) and the Al/Na atomic ratios measured through XPS analysis are shown.Since it is reported in literature that Cu 0 and Cu + species cannot be accurately differentiated, the ratio of Cu species is defined as Cu 2+ /(Cu 0 + Cu + ) [49].Table 3 reports also the bulk Al/Cu and Al/Na atomic ratios as measured via ICP-OES for comparison.Since the experimental procedure for the preparation of the ion-exchange form of NaY and HY zeolites was proven to be reproducible, samples NaYCu2 and HYCu2 have been used to evaluate the performance of the one-time ion-exchanged form of NaY and HY zeolites, respectively, and are referred from now on in this work as NaYCu and HYCu.
As expected, the Al/Na surface ratio of all the HY zeolites exceeds the one of NaY and 13X zeolites.Moreover, since as previously explained in HY zeolites Na + species significantly decrease during ion-exchange procedure, the Al/Na surface ratio of the sample increases significantly per ion-exchange procedure for ion-exchanged HY zeolites.Although in a less significant way compared to HY zeolites, an increase in the Al/ Na surface ratio is also detected for the ion-exchanged forms of NaY and 13X zeolites as Cu substitutes some of the Na + cations at the surface.From Table 3 the Al/Cu surface ratio is found to be higher than the Al/ Cu bulk ratio for all the samples.Since after ion-exchange procedure the copper content on the zeolite surface increases and the Al/Cu surface ratio would be consequently expected to decrease, this result supports the hypothesis that the Al detaching from the zeolite structure during aqueous ion-exchange procedure and calcination reattaches at the external surface of the ion-exchanged zeolites as EFAl [43,48].Furthermore, the higher copper loading arising from consecutive ion-exchanges procedures results in Al/Cu atomic ratio decreasing for consecutive ionexchange procedures, indicating that after each ion-exchange procedure more Cu species are deposited at the surface of the zeolite.From Table 3 it is possible to see that for some ion-exchanged forms of zeolites NaY and 13X, the Al/Na surface ratio is lower than the Al/Na bulk ratio, which seems not to justify the hypothesis that part of the Al detaching from the zeolite structure during ion-exchange procedure and calcination reattaches at the external surface of the zeolite.However, these zeolites show a Al/Na surface ratio lower than a Al/Na bulk ratio already prior ionexchange and it is therefore possible that, despite the overall Al/Na ratio increase, the Al/Na ratio increase at the surface due to copper exchange and Al migration is still not sufficient to overcome the increase in the Al/ Na bulk ratio.On the other hand, it can also be noticed that Al/Na surface ratio is lower than the Al/Na bulk ratio only for the pristine form of NaY and 13X and for the ion-exchanged forms of the same zeolites calcined in air, whereas for the ion-exchanged forms of zeolites NaY and 13X for which calcination was performed in He, namely samples NaYCu5x and 13XCuI, the Al/Na surface ratio is higher than the Al/Na bulk ratio.Since samples NaYCu5x and 13XCuI have been prepared following the same procedure of samples NaYCu5x and 13XCuII, respectively, with the only exception of a different calcination environment, this finding suggests that the calcination environment next to preventing dealumination of the sample (as indicated in Table 2) plays a significant role in the distribution of the different species in the zeolite structure.It can be inferred that an inert calcination environment allows for a higher degree of Al migration to the external surface of the zeolite and/or for a  migration to the bulk of some Na + cations that are still available at the sample surface after ion-exchange procedure.
With the only exception of samples NaYCu5x and 13X-CuII, all the samples show a Cu 2+ /(Cu 0 + Cu + ) ratio lower than 1.This indicates that for these samples the amount of copper that is available in non-oxidized form (metallic Cu 0 or Cu + ions) is higher than the amount of copper available in its oxidized form (Cu 2+ ).Since this happens also for samples that have undergone calcination in air it is possible to conclude that when calcination is performed in air not all the Cu species available in the zeolite structure after ion-exchange procedure react with oxygen contained in air to forming Cu (2+) O (2-) .On the other hand, although some Cu 2+ is still present at the surface of the samples that have been calcined in He, namely NaYCu5xI and 13X-CuI, it is possible to infer that part of the Cu species introduced in the zeolites structure via ionexchange procedure is successfully auto-reduced during calcination in He.
The XRD patterns for all the forms of NaY, HY and 13X zeolites investigated in this work are presented in Fig. 1.All the Cu-exchanged form of zeolites show a XRD pattern that resembles the crystal phase of their parent zeolite confirming the stability of the microporous structure of the zeolites undergoing ion-exchange procedure followed by calcination, either in air or He.However, since a decrease in relative intensity of the ion-exchanged samples is detected in comparison to pristine zeolites, the degree of crystallinity for Cu-exchanged zeolites decreases with the increasing number of ion-exchange procedures.This indicates that, although most of the crystallinity is maintained for the ion-exchanged zeolites, some parts of the crystal structure have collapsed.The structural degradation can be attributed to the dealumination occurring during aqueous ion-exchange at lower temperatures (<373 K) [48], but no amorphous phase could be detected for all the Cu-FAU zeolites.Finally, from Fig. 1 it is possible to conclude that the Cu species are highly dispersed as isolated Cu + /Cu 2+ cations in the zeolite framework and/or as small CuO x clusters that are below the XRD detection limit ( 5 The complete nitrogen isotherms of all the materials investigated in this work are presented in Fig. 2. The curves obtained for pristine 13X, NaY and HY zeolites resemble a combination of type I and IV adsorption isotherm according to the IUPAC classification [50].Specifically, type I behavior is observed at low relative pressure where micropore filling occurs, while the strong increase in the N 2 adsorption at higher pressures can be explained by pore condensation into larger mesopores and macropores.A significant decrease in N 2 adsorption can be seen for all the samples prepared via ion-exchange procedure and this is due to the fact that part of the crystal structure collapses during ion-exchange procedure as well as to the fact that the blockage of some pores occurs while introducing copper in the zeolite structure [51].The behavior of 13X-CuII zeolite deviates from the one of all the other Cu-exchanged zeolite samples by exhibiting a hysteresis loop corresponding to the one indicated in the IUPAC classification as type H 2 .This refers to a complex pore structure in which pore blocking/percolation effects significantly contribute.As this hysteresis loop is missing for both the ion-exchanged forms of NaY and HY calcined in air, the formation of this copper network could be ascribed to the introduction of ammonia hydroxide solution during sorbent preparation.It is however worth to be mentioned that, in a strict sense, the BET method is not valid for estimating the surface area of the prepared microporous zeolites as this method relies on assumptions that are not entirely justified for microporous materials.The value for the monolayer capacity that is derived from the BET plot can therefore be misleading and the choice of N 2 as standard adsorptive might lead to a different area occupied by the adsorbed N 2 molecule compared to the customary value of 0.162 nm 2 for polar sorbents.Argon might be used in future studies to omit the latter deviation as it should be less sensitive to polar sorbents due to its absence of quadrupole moment.Nevertheless, it is chosen in this study to apply the BET method to compare equivalent BET surface area of different sorbents, as summarized in the legend of Fig. 2. According to the experimental results, the surface area further decreases upon repetitive ion-exchange procedure and this trend is more remarked for HY zeolites than for NaY zeolites as most of the crystal structure remains intact as well as the copper loading remains low.The surface area of 13X-CuI, with a BET surface area of 642.3 m 2 /g being similar to its parent zeolite, surpasses the one of all the other zeolites.

Adsorption and regeneration experiments
The total saturation capacity and the desorption profile as function of temperature of each sample after the first adsorption step are depicted in Fig. 3 and Fig. 4, respectively.
The NH 3 -TPD profiles of pristine zeolites NaY, HY and 13X can be deconvoluted either into two pronounced peaks -one at relatively low temperature (T < 400 K) and one at relatively high temperature (T > 500 K) -or into a unique pronounced peak at low temperature (T < 400 K) combined with a longer tail at elevated temperatures.The low temperature (LT) peak refers to NH 3 that is weakly bounded to the adsorbent material and can be attributed to the presence of Lewis acid sites, extra framework aluminum (EFAl) and aluminium (Al) defects in the zeolite structure.The high temperature (HT) peak or the extended tail are on the other hand related to the desorption of protonated NH 3 from Brønsted acid sites and the position of this peak or tail shows the relative strength of these Brønsted acid sites [52,53].As it is possible to see from Fig. 4, the Cu-exchanged form of each zeolite shows LT peaks with higher intensity compared to the one of their respective parent zeolite and this is due to the fact that the generated EFAl and isolated Cu species can contribute to this peak [52,53].Moreover, as it is clearly visible from Fig. 4(a) for the specific case of NaY-type of zeolites, the NH 3 -TPD profile of the Cu-FAU zeolites shows an additional peak at intermediate temperature (IT) which can be attributed to ammonia adsorbed in the acid sites provided by copper complexation.These experimental results show that the Cu 2+ exchange provides additional NH 3 adsorption sites to the zeolite which contribute to an increase of the zeolite saturation capacity.The intensity of this intermediate peak is a direct measure of the copper loading of the adsorbent materials.As it is possible to see from Fig. 4, shift towards higher temperatures as well as a broadening of the HT peak is also observed for Cu-FAU.This indicates that the strength of the Brønsted acid sites that remain in the sorbent structure after Cu exchange increases and that the ion-exchange procedure has led to an increase of the heterogeneity of these sites [53].
Compared to their parent zeolite, the saturation capacity of most of the Cu-FAU zeolites is enhanced although the increase in saturation capacity does not proportionally increase with the copper loading (Fig. 3).Moreover, consecutive ion exchange procedures lead to an increase of the saturation capacity of Cu-exchanged HY and NaY samples.The saturation capacities of HY, HYCu and HYCu5x are in fact 1.95 ± 0.11 mmol/g sorbent , 2.48 ± 0.01 mmol/g sorbent , 2.81 ± 0.04 mmol/ g sorbent , respectively; whereas the ones of NaY, NaYCu and NaYCu5x are 2.34 ± 0.06 mmol/g sorbent , 4.33 ± 0.85 mmol/g sorbent , 5.44 ± 0.40 mmol/g sorbent , respectively.Since NaYCu5xI shows a saturation capacity of 5.32 ± 0.09 mmol/g sorbent which is lower than the one of both NaY-Cu5x and NaYCu, it can be inferred that an inert calcination atmosphere is not beneficial for an increase of the saturation capacity of NaY zeolites.Conversely, being the saturation capacity of samples 13X, 13X-CuI and 13X-CuII 3.88 ± 0.25 mmol/g sorbent, 4.50 ± 0.16 mmol/g sorbent and 3.59 ± 0.22 mmol/g sorbent , respectively, an inert calcination atmosphere seems to be beneficial for an increase in the saturation capacity of zeolite 13X.In Fig. 5 the useful and saturation capacities of all the sorbents at the first and last adsorption/desorption cycle are represented.As it is possible to see, the pristine form of all the zeolites shows good regenerability, as the saturation and useful capacities do not significantly decrease over the cycles.On the other hand, it can be noticed that despite its highest useful and saturation capacity compared to other adsorbent materials after the first adsorption/desorption cycle, NaY-Cu5x cyclability is not optimal.From Fig. 6(a), in which the breakthrough curves obtained over the ten cycles are depicted, it can in fact be seen that after a first cycle the adsorption performance rapidly decreases to afterwards stabilize for all the cycles following the second one.This pattern is observed for all the Cu-FAU zeolites with the only exception of zeolite 13X-CuI for which, as it is possible to see from Fig. 5 and Fig. 6 (b), the useful capacity remains nearly constant over the 10 cycles, being measured to be 3.02 ± 0.39 mmol/g sorbent during cycle 1 and 3.07 ± 0.01 mmol/g sorbent during cycle 10.The breakthrough plots for all the other adsorbent materials can be found in the supplementary material.From the results of these experiments it can be concluded that being not only the most stable adsorbent material over ten adsorption/desorption cycles, but also the one with highest useful and saturation capacity after 10 adsorption/desorption cycles, zeolite 13X-CuI can be regarded among all the investigated materials as the most promising material for ammonia traces removal from hydrogen produced via ammonia decomposition in a membrane reactor.
13X-CuI was therefore tested for the removal of 10 ppm of ammonia in H 2 in a fixed bed reactor, simulating the composition of the permeate stream leaving a palladium based membrane reactor for NH 3 cracking as obtained in our previous work [7,8].Fig. 7 depicts the first adsorption breakthrough curve: in the first part of the experiment, the concentration of NH 3 at the bed outlet was measured to be below the accuracy limit of the FTIR (0.75 ppm), indicating that 13X-CuI is certainly able to reduce the concentration of ammonia in the hydrogen stream from 10 ppm to < 0.75 ppm.However, the fact that the absorbance band recorded prior breakthrough based on the extrapolation of the calibration curve (available in the supplementary material of this work) well resembles the noise of the background scanning, makes it possible to infer that the pre-breakthrough outlet stream is likely to be virtually free from NH 3 .Based on the breakthrough time to achieve 0.75 ppm of NH 3 , the useful capacity of 13X-CuI has been calculated to be 2.22 ± 0.00 mmol/g sorbent .The sorbent therefore shows a significatively lower value of useful capacity compared to the one calculated during breakthrough experiments using 3000 ppm of NH 3 in He (3.07 ± 0.01 mmol/g sorbent ).This can be explained by the dependence of the adsorption equilibrium on the adsorbate partial pressure (or concentration) in the gas phase [54].Furthermore, other two factors influencing the result are the presence of N 2 in the system as regeneration gas instead of He and the presence of H 2 instead of He in the adsorption step.Although H 2 , N 2 and He share the absence of any dipole moment, the significant difference in their polarizability and quadruple moment has in fact an impact on the useful capacity of the material which is given by the fact that the polarizability of the non-polar gasses results in the interaction between the strong electric field found in zeolite cavities and the induced dipole of the adsorbate while at the same time the quadruple moment of the molecules results in the interaction between electric field gradient found in zeolite cavities and quadruple moment of the adsorbate [55].Accordingly, since the polarizability (ρ) and quadruple moment of the adsorbate gasses are as ρ N2 ≫ρ H2 > ρ He [56], a higher adsorption capacity for N 2 is expected when during degassing and regeneration (and cooling down) of the sorbent N 2 is used instead of He.The use of N 2 as a regeneration gas rather than He might have therefore led to the inhibition of some of the adsorption sites in the sorbents during the breakthrough experiments, lowering the useful capacity of 13X-Cu at realtime conditions.Additionally, due to the slightly higher interactions of H 2 compared to He with the sorbent, some hydrogen is also expected to adsorb during hydrogen purification, which might also show some minor influence on the measured useful capacity.
In order to show the cyclability and stability of 13X-CuI when using  The regeneration of all the sorbents was performed at 623 K for 30 min at a heating rate of 5 K/min using 10 mL/min of He of purge gas.
N 2 as purge gas, two additional adsorption/desorption cycles have been performed and the useful capacities obtained during the three adsorption/desorption cycles are represented in Fig. 8.After the first cycle the sorbent sample shows slightly improved useful capacity, but this can be explained by the fact that different NH 3 concentration (partial pressures) are present in the feed (10.0 ppm of NH 3 during cycle 1 and 86.5 during cycle 2 and 3), affecting the adsorption equilibrium conditions.On the other hand, no significant decrease in adsorption performance is detected between the second and third cycle, showing the sorbent cyclability and stability to be good also when N 2 is used as purge gas.

Deactivation of Cu-FAU zeolites
While among all the investigated materials zeolite 13X-CuI is proven to be the most promising one for the removal of ammonia traces from hydrogen produced via ammonia decomposition in a membrane reactor, a better understanding of the deactivation mechanism of the other Cu-FAU zeolites is required.In this regard, further investigation was carried out on both fresh and used zeolite NaYCu5x.The choice to study the deactivation mechanism of Cu-FAU zeolites while investigating zeolite NaYCu5x derives from the fact that NaYCu5x showed both the best performance during the first adsorption/desorption cycle as well as the Fig. 6.Breakthrough curve for NaYCu5x (a) and 13X-CuI (b) for multiple adsorption/desorption cycles at 3000 ppm NH 3 /He at room temperature.The regeneration of both the sorbents was performed at 623 K for 30 min at a heating rate of 5 K/min using 10 mL/min He as purge gas.Fig. 7. First breakthrough curve for zeolite 13X-CuI at 10 ppm NH 3 in H 2 at 700 mL N /min, simulating the membrane permeate during NH 3 decomposition.The insert shows the official breakthrough, exceeding the threshold and accuracy limit of the FTIR of 0.1 and 0.75 ppm, respectively.most remarkable deactivation.Moreover, since all the zeolites samples with the only exception of 13X-CuI seem to have the same deactivation behavior, an explanation on the deactivation mechanism of zeolite NaYCu5x could also be extended to the one of all the other deactivating Cu-FAU zeolites.
As it was possible to see from Fig. 5(a), the useful and saturation capacities of zeolite NaYCu5x drastically decrease after the first adsorption/desorption cycle.To exclude that this is due to incomplete regeneration of the material after saturation, 5 adsorption/desorption cycles are repeated under the same operating conditions at which the initial breakthrough curve experiments (10 adsorption/desorption cycles) have been performed, but this time the sorbent was regenerated at higher temperatures and increasing the duration of the regeneration phase to favor NH 3 desorption.Specifically, the material is regenerated holding for 60 min a temperature of 673 K for two cycles and a temperature of 723 K for other three cycles.In Fig. 9 the useful capacities achieved during these experiments are compared to one found during the initial breakthrough curve experiments.The increasing regeneration time and temperature have shown to have negligible effect on the useful capacity, ruling out the incomplete regeneration of the material as the reason for lower adsorption capacity from cycle 2.
To further investigate the deactivation mechanism of Cu-FAU zeolites, potential structural changes have been studied by means of XRD and BET.From the XRD patterns of pristine and ion-exchanged NaY and 13X zeolites (Fig. 10), it is possible to see from a qualitative point of view that all the NaY and 13X type zeolites show a decrease in crystallinity over the cycles.The largest decrease in crystallinity is observed for the spent sorbents containing the highest copper loading and calcined in air, namely NaYCu5x and 13X-CuII.As indicated in Table 4, the decrease in crystallinity is accompanied by a decrease in S BET for all the samples except for NaYCu5xI and 13X-CuI.As the increase in surface area for these samples is remarkable (about 2 % and 18 % for NaYCu5xI and 13X-CuI, respectively), it can be inferred that inert atmosphere during calcination makes it possible for redistribution and further dispersion of the Cu species to take place during breakthrough experiments.
The NH 3 -TPD profiles obtained during ten consecutive adsorption/ desorption cycles for pristine NaY and 13X samples (Fig. 11(a) and Fig. 11 (b)) show no significant changes in terms of sorption capacity and peaks shift.On the other hand, the intensity of the NH 3 -TPD profile peaks that are detected for NaYCu5x (Fig. 11 (c)) markedly change after the first adsorption/desorption cycle.This might be due to relatively small morphology changes which should however be confirmed in a future work by means of SEM-EDX analysis.The decrease in the intensity of the HT peak might be attributed to framework Al removal, as a similar phenomenon has already been observed for Cu-exchanged zeolite ZSM-5 during NH 3 selective catalyst reduction (NH 3 -SCR) [57], whereas the decrease in the intensity of the LT peak -which is detected despite Al removal during ion-exchange results in an increase in EFAl species which should directly contribute to additional NH 3 adsorption sites -and the change in curvature of the IT peak after the first cycle -which indicates a change in the behavior for the Cu speciesare believed to be attributed to the formation of copper oxides (CuO x ) and copper alluminate (CuAlO x ) species due to Cu and Al migration [57].These copper aluminate species (CuAlO x ) are in fact thermally stable, hydrophobic and show a low surface acidity, which decreases the number of available NH 3 adsorption sites decreasing in turn the performance of the sorbents [58].In contrast to the NH 3 -TPD profile of NaYCu5x (Fig. 11(c)), from Fig. 11(d), it can be seen that the change in curvature is almost negligible for 13X-CuI.The NH 3 -TPD profiles obtained during ten consecutive adsorption/desorption cycles for all the adsorbent materials can be found in the supplementary material of this work.TEM images presented in Fig. 11 show indeed that both fresh and spent samples of NaYCu5x and 13X-CuI contain CuO x and CuAlO x clusters, although the average cluster size was measured to be slightly larger for NaYCu5x than for 13X-CuI both before and after cycling.The size of the largest clusters is found in the range between 1 and 5 nm.Fig.12..
To further investigate the presence of CuAlO x clusters in the Cu-FAU zeolites structure, the Al/Cu surface and the Cu 2+ /(Cu 0 + Cu + ) atomic ratios on the surface of both the fresh and the used NaY and 13X type zeolites were measured through XPS analysis and the results are presented in Table 5.Since the migration of Al species towards Cu species at the surface can result in the formation of CuAlO x at the material surface, an increase in Al/Cu surface atomic ratio is expected for the spent Cu-FAU zeolites that show the same deactivation patterns as NaYCu5x.For all the Cu-FAU zeolites, except for 13X-CuI, the Al/Cu surface atomic ratio after cycling is found to be higher compared to the fresh ones, confirming the hypotesis that the materials deactivation could be ascribed to the formation of CuAlO x species at the material surface.Moreover, the results presented in Table 5 show also that for all Cu-FAU zeolites, with the exception of NaYCu5xI and 13X-CuI, most of the Cu species reduce to Cu + or Cu 0 resulting in a decrease of the Cu 2+ /(Cu 0 + Cu + ) ratio.This occurs due to the fact that the sample autoreduction occurs during the regeneration phase in inert atmosphere and due to the fact that during regeneration ammonia can act as a reducing agent [59].As NaYCu5xI and 13X-CuI undergo autoreduction already during calcination procedure, the Cu 2+ /(Cu 0 + Cu + ) ratio remains constant for both the fresh and used zeolites.
To confirm the actual formation of CuAlO x clusters, the environment of Cu species has been studied performing CO adsorption in combination Fig. 8. Useful capacity calculated on the breakthrough time to achieve 0.75 ppm of NH 3 for zeolite13X-CuI over three consecutive adsorption-desorption cycles.During the experiments, saturation was carried out feeding 700 mL N / min of an NH 3 /H 2 mixture containing 10 ppm of NH 3 in the first cycle and 86.5 ppm of NH 3 in the second and third cycles, while regeneration was performed at 623 K with He in the first cycle and with N 2 in the second and third cycles.with FTIR spectroscopy on both fresh and used samples of zeolites NaYCu5x and 13X-CuI.At room temperature, in fact, the probe molecule CO solely interacts with Cu + species and the adsorbed CO can be monitored by FTIR spectroscopy.Since in all the Cu-FAU zeolites Cu + cations are present in the form of exchangable bare cations and Cu + oxide species [60], when CO comes in contact with an exchangable cationic Cu + at room temperature, monocarbonyl species (Cu + (CO)) are in fact formed and can be detected as intensive band in the wavelength range between 2250 and 2050 cm − 1 [61].The position of the adsorbance band of the Cu + (CO) species depends on the nature of the Cu + species.Fig. 13 depicts the results obtained for fresh and used NaYCu5x and 13X-CuI.For low CO doses, two peaks are identified for NaYCu5x ((Fig.13(a)) at 2146 and 2157 cm − 1 .According to the result of a study carried out by Podobinski et al. [62], the peak identified at 2146 cm − 1 can be ascribed to CO interacting with Cu + species in the S II* sites (Cu + S II* -CO), whereas the one identified at 2157 cm − 1 can be attribute to CO interacting with Cu + species in the S II sites (Cu + S II -CO).Being the Cu + S II* -CO bonds less strong than the Cu + S II -CO ones, Cu + S II* are less stable than Cu + S II : as a consequence Cu + S II* can be found at the center of an oxygen ring where it is surrounded by oxygen atoms, while Cu + S II can be found just ouside the oxygen ring above the trigonal oxygen pyramid.The more effective neutralization of Cu + S II* results in the fact that its absorbance band can be found at lower frequencies compared to the one of Cu + S II .At higher CO doses, an additional band at 2180 cm − appears.This band is attributed to the symmetric stretching of Cu + (CO) dycarbonyls [63].From a comparison between the results obtained with fresh and used NaYCu5x (i.e. by comparing Fig. 13(a) and Fig. 13(b)), after 10 adsorption/desorption cycles the adsorbance bands related to Cu + (CO) x are red-shifted, indicating an increase in AlO 4 -species in close proximity to all Cu + species.Additionally, as the intensity of the peaks attributed to Cu + S II -CO bonds is higher compared to the one attributed to the Cu + S II* -CO bonds, whereas the trend is opposite in the case of fresh NaYCu5x, it is possible to infer that a part of the Cu + species located at S II + sites in the fresh material moves towards S II* sites during cycling.This finding can be the consequence of the fact that upon removal of water and in presence of NH 3 , the mobility of Cu + species allows them to coordinate more firmly towards framework zeolite oxygen in S II* [64].Fig. 13(c) and Fig. 13(d) show on the other hand that for low CO doses one single peak is identified for 13X-CuI at 2152 cm − 1 .This peak can be assigned to the formation of Cu + S II -CO carbonyl.The red-shift of this band compared to the position of the same band for NaYCu5x can be explained by the lower Si/Al ratio of 13X-CuI in   comparison to the one of NaYCu5x (confirmed also by the ICP-OES measurements already presented Table 2).The absence of the Cu + S II* -CO absorbance band for 13X-CuI might support the hypothesis that the introduction of ammonium hydroxide in the structure of zeolite 13X coordinates the Cu + species towards the supercage preventing their migration to hidden sites [43].From a comparison of the results obtained before and after ten adsorption/desorption cycles, it is possible to see that the intensity and the position of all the absorbance bands obatined with 13X-CuI remain similar, indicating no increase in AlO 4 -species in close proximity of all Cu + species during cycling.As on the other hand the intensity and the position of the same peaks are different for fresh and used NaYCu5x, the deactivation of NaYCu5x might be attributed to the formation of CuAlO x species.
The environment of Cu species on fresh and used samples of zeolites NaYCu5x and 13X-CuI has been studied also performing CO adsorption in combination with FTIR spectroscopy at temperature lower than room temperature, specifically at 105 K.At this temperature, the CO adsorbance patterns show in fact a higher amount of peaks compared to the ones obtained at 323 K as next to Cu species the zeolite framework itself and additional Na + species contribute to CO adsorption [65,66].Fig.
depicts the results obtained for NaYCu5x and 13X-CuI.As it is possible to see from Fig. 14(a), for low CO pressures, two peaks are identified for NaYCu5x at 2170 and 2142 cm − 1 , whereas at higher CO pressures next to these two peaks, which respectively contribute in the wavelength ranges of 2171-2161 cm − 1 and 2138-2131 cm − 1 , an additional peak is found at 2191 cm − 1 .The position of these peaks can be assigned to CO interacting with Cu 2+ and forming Cu 2+ -CO complexes [67].Specifically, the peak in the wavelength range between 2171 and 2161 cm − can be assigned to Cu + (CO) 2 and Cu 2+ (CO) 2 species formed at ionexchange positions, to the free CO that weakly adsorbes in the zeolite channels (H-bonded CO) and to Na + -(CO) x complexes, whereas the one  in the wavelength range between 2138 and 2131 cm − 1 can be ascribed to Cu + (CO) x species and the one at 2191 cm − 1 to Cu + (CO) 3 species as well as to EFAl and Cu 2+ (CO) 3 [66][67][68][69].Similar trends are observed for NaYCu5x undergoing ten adsorption/desorption cycles (Fig. 14(b)), but in this case the intensity of the peaks is significantly increased.The largest increase in adsorbance intensity is found for the peak at 2167 cm − 1 , which corresponds to the strong increase in the Cu + species after cycling as confirmed by the results obtained through XPS analysis (Table 5).An intensification of the peak at 2191 cm − 1 also occurs and the fact that this peak is assigned to EFAl and Cu 2+ (CO) 3 in combination with the fact that according to the results of XPS analysis a decrease in the Cu 2+ is expected when NaYCu5x is used makes it possible to conclude that this intensification is due to EFAl and minor contribution of Cu + species.Pristine and spent NaYCu5x show the same adsorbance band after evacuation at 105 K.This adsorbance band can be deconvoluted into two peaks, where the main peak at 2141 cm − 1 and the shoulder correspond to the Cu + (CO) 2 species and to residual Cu 2+ (CO) species which are known to slowly decompose during evacuation at low temperature [66].Although the increase in adsorbance intensities together with the XPS results support the increase of Cu + species after cycling, the similary of adsorbance band for both fresh and used NaY-Cu5x indicates that the Cu + species forming over cycling do not form in turn Cu + (CO) 2 species resistant to evacuation.The increase of AlO species in close proximity of Cu + species might have weakened the interaction between CO and the metal preventing the formation of stable Cu + (CO) 2 species.The CO absorbance pattern of sample 13X-CuI (Fig.  sample NaYCu5x, although their position is slightly shifted towards lower wavelengths since zeolite 13X-CuI has a lower Si/Al ratio compared sample NaYCu5x.The fourth shift is detected around 2100 cm − 1 and, as it almost coincides with the peak identified at 2110 cm − 1 for zeolite Na-ZSM-5 [70,71], can be attributed to isocarbonyl species in which the CO molecules interact with a single Na + cation via an oxygen atom.The fact that this peak seems to be nearly absent for zeolite NaYCu5x might be explained by the fact that 13X-CuI contains a significantly higher amount of Na + cations compared NaYCu5x.The peak around 2184 cm − 1 is not visible for the spent 13X-CuI sample indicating that most of the most of the Cu 2+ species are reduced to Cu + species.This is also confirmed by the small increase of the absorbance band intensity at 2165 cm − 1 .Since this peak represent also EFAl species, the absence of this peak for the spent 13X-CuI sample also insinuates that that the degree of dealumination is minimized for this sample.The intensity of the peaks at 2140 and 2100 cm − 1 of the spent sorbent are also lower compared to the ones of the fresh sorbent and the absorbance band measured after evacuation at 105 K for the spent sorbent is completely different than the one measured for fresh sample.This might indicate that the peaks around 2140 and 2100 cm − 1 are not solely related to Cu + (CO) species and Na + cations, but that other factors might play a role.However, since this phenomenon has not yet been reported in literature, further experimental work should be addressed in order to better explain this finding.Nevertheless, it is worth mentioning that according to the result of this work this change in copper environment seems not to affect and jeopardize the adsorption performance of zeolite 13X-CuI.

Conclusions
Cu-FAU zeolites have been investigated in this work as promising adsorbent materials for the removal of residual NH 3 impurities from hydrogen produced via ammonia decomposition in a Pd-based membrane reactor to be used as feedstock for PEMFCs.The adsorption and regeneration performance of the adsorbent materials have been assessed at lab-scale by monitoring the breakthrough curves obtained during multiple NH 3 adsorption/desorption and were subsequently compared to the ones of the materials' parent zeolites.All the adsorbent materials have also been characterized by means of several techniques prior and after breakthrough experiments in order to assess possible changes in their structure over the cycles which could explain possible variations in their adsorption performance.The introduction of copper species in the zeolites framework was demonstrated to enhance the NH 3 useful and saturation capacities of the copper exchanged zeolites compared to their parent zeolite.However, the Cu-FAU zeolites synthesized in presence of copper(II)nitrate and calcined in airi.e.all the Cu-FAU zeolites with the only exception of the one-time ion-exchanged form of zeolite 13Xsignificantly deactivated after the first adsorption/desorption cycle.XRD analysis demonstrated that for all the FAU zeolites a decrease in crystallinity occurs over ten adsorption/desorption cycles and that the Cu species are well dispersed in the material structure.On the other hand, NH 3 -TPD, XPS, TEM and FTIR spectrometry measurements revealed that the deactivation of Cu-FAU zeolites can be attributed to significant dealumination of the material which is on the other hand prevented when ion-exchange procedure occurs in presence of ammonium hydroxide and calcination takes place in inert atmosphere.The only ion-exchanged material that did not undergo deterioration phenomena was found to be the one-time ion-exchanged form of zeolite 13X.Its useful capacity, which remained stable over the cycles, was measured to be 3.07 ± 0.39 mmol/g sorbent (19 % higher compared to its parent zeolite) when saturation was performed with a He/NH 3 mixture containing 3000 ppm of ammonia and regeneration was performed in He.Stable performance of this sorbent have also been assessed under operating conditions emulating the outlet stream of a membrane reactor for ammonia decomposition.The useful capacity and saturation capacity were measured to be 2.18 ± 0.02 mmol/g sorbent and 4.33 ± 0.95 mmol/g sorbent , respectively, when saturation was performed with a H 2 / NH 3 stream containing 86.5 ppm of ammonia and regeneration was performed in N 2 .Through FTIR spectrometry it was confirmed that the one-time ion-exchanged form of zeolite 13X can be regarded as a suitable adsorbent material to effectively purify NH 3 -derived H 2 for PEMFC application.However, while so far it has been demonstrated that the one-time ion-exchanged form of zeolite 13X shows good stability, whereas other Cu-FAU zeolites deactivate due to agglomeration of Al and Cu species, it has not been fully unraveled how the introduction of ammonium hydroxide in the ion-exchange solution for the preparation of the ion-exchanged version of 13X zeolite as well as the adoption of inert calcination atmosphere have prevented this from happening for this specific material.Further work should therefore be addressed to investigate the reason why the ion-exchanged form of zeolite 13X prepared performing ion-exchange in ammoniated solution and calcination in inert atmosphere does not undergo deterioration phenomena over cycling.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
nm).No diffraction lines in the XRD patterns correspond in fact to the position corresponding to different copper species at 2θ = 35.5;38.7; 48.7; 58.3; 61.8 • .

Fig. 1 .
Fig. 1.XRD patterns of all NaY (a), HY (b) and 13X (c) forms of zeolites used in this work.

Fig. 5 .
Fig.5.Overview of all saturation and useful capacities for the first and last cycles at 3000 ppm NH 3 /He at room temperature for all the studied sorbents.The regeneration of all the sorbents was performed at 623 K for 30 min at a heating rate of 5 K/min using 10 mL/min of He of purge gas.

Fig. 9 .
Fig. 9. Useful capacity found for sample NaYCu5x during 5 adsorption/ desorption cycles performed saturating the sorbent bed with a 100 mL N /min flow rate of an NH 3 /H 2 mixture containing 3000 ppm of NH 3 and regenerating the bed using 10 mL N /min of He.

Fig. 13 .
Fig. 13.CO adsorption on (a) fresh NaYCu5x, (b) used NaYCu5x, (c) fresh 13X-CuI and (d) used 13X-CuI at 323 K.The absorbance has been recorded for successive doses of CO and the colored bar shows the measured pressure inside the FTIR cell.The sorbent and FTIR cell background are subtracted and the intensity of the absorbance is normalized by pellet mass.
14(c)  andFig.(d)) consists of four peaks.Three of these peaks are similar to the ones of

Fig. 14 .
Fig. 14.CO adsorption on (a) fresh NaYCu5x, (b) used NaYCu5x, (c) fresh 13X-CuI and (d) used 13X-CuI at 105 K.The absorbance has been recorded for successive doses of CO and the colored bar shows the measured pressure inside the FTIR cell.The sorbent and FTIR cell background are subtracted and the intensity of the absorbance is normalized by pellet mass.

Table 1
List of adsorbent materials prepared and used in this study.

Table 2
Elemental composition analysis of all the sorbents performed by means of ICP-OES. 5 % of the total sample mass is assumed to consist of water, b the SiO 2 content is estimated on the basis of the Al 2 O 3 , CuO and Na 2 O content, c molar ratio, d calculated using Na 2 O/Al 2 O 3 weight ratio as proposed by Liu and Aika[40], e calculated using CuO/(2 • Al 2 O 3 ) weight ratio. a

Table 3
Al/Cu, Cu 2+ /(Cu 0 + Cu + ) and Al/Na atomic ratios as measured via XPS analysis and Al/Cu and Al/Na atomic ratios as calculated by ICP-OES.

Table 4 S
BET for fresh and spent NaY and 13X related samples.

Table 5
Atomic ratio Al/Cu, Al/Na and Cu 2+ /(Cu 0 + Cu + ) measured via XPS analysis for fresh and spent (ten adsorption/desorption cycles using 3000 ppm NH 3 /He for saturation) NaY and 13X-type zeolites.