Mixed Matrix Membranes prepared from polysulfone and Linde Type A zeolite

Abstract This work focuses on dip coating and further phase inversion prepared polysulfone/LTA zeolite mixed matrix membranes (MMMs). The Linde Type A (LTA) zeolite synthesized under hydrothermal conditions by the organic free method was introduced as fillers at 10 and 20 wt.% loadings into the polysulfone polymer matrix to obtain MMMs. The x-ray diffraction (XRD) and scanning electron microscopy (SEM) analysis indicated that the as-synthesized LTA zeolite samples were crystalline and mainly composed of crystal of predominantly cubic shape. Textural characterisation using Ar adsorption/desorption data of LTA zeolite shown the existence of mesoporous. Atomic force microscopy (AFM) in combination with the SEM characterised the membrane morphology and the dispersion of zeolite fillers. The effect of the zeolite loading on the performance of the MMMs was analysed. It points out that N2 permeability was increased with the increment of zeolite filler, also the high membranes permeability and the weak dependence upon transmembrane pressure and therefore its high selectivity. The average membrane thickness was 150 μm.


Introduction
The first synthesis of LTA zeolite took place in 1956 at Union Carbide Labs by hydrothermal synthesis under the efforts of Breck and co-workers [1]. Since this challenge and before, many other zeolite structures have been obtained by different methods, but well-known zeolites as the LTA structure keep the interest under new perspectives due to its different fields of application on the commercial scale and in terms of scientific research [2].
Many are the essential properties of zeolite membranes, among them, highly ordered crystalline structure, thermal stability that makes the membranes useful at hightemperature operations, chemical resistance, controlled host-sorbate interaction and uniform molecular size pores. Taking into consideration all of them, different studies have been carried out on the preparation and characterization of zeolite membranes during the last thirty years [3][4][5].
In general, preparation of membranes with zeolite use one of the two following routes [6]: (a) Crystallization of zeolite film by growing the zeolite layer on porous support like ceramic, metal or glass or self-supported membranes by direct grows without any support or growing the zeolite layer over temporary support. (b) Use of zeolite crystal previously synthesised as inorganic fillers to produce zeolite-filled membranes by embedding zeolite crystal into different kinds of matrices.
However, zeolite membranes is still a challenge nowadays because their fragile structures make the development of large-scale inorganic membranes hard, especially to produce reliable defect-free membrane and costeffective fabrication processes [7]. For this reason, despite the results in the development of zeolite membranes, only a few industrial applications are considered nowadays [4].
The first synthetic membranes from organic nature were developed with cellulose and its derivative products, but their performance under the chemical environment or temperature was too low [8]. The development of new or-ganic membranes using polymers as polyamide, polyacrylonitrile, and others with better resistance to chemical environment and better glass transition temperatures (Tg) in comparison with cellulose membranes have allowed their inclusion in new processes for polymeric membranes never seen. Other polymers as polysulfone, polyester sulfone, etc., in addition to chemical stability and Tg, compared to those before, have excellent mechanical properties.
The possibility of incorporating polymers and inorganic components into a material has been explored since the early industrialization of the polymers. Insertion of inorganic materials as fillers in the system could improve optical and mechanical properties [9]. The development of inorganic chemistry in recent decades and the introduction of the concept of "organic-inorganic hybrid materials" combined with multi-functionalization [10] has allowed the development of custom products. In these products, the coexistence on the nanoscale of organic and inorganic phases makes them candidates for applications such as nonlinear optics, electronics, sensors, and catalysts [11].
The first research about mixed matrix membranes (MMMs) was developed around 1970 with the addition of 5A zeolite as fillers into the polymeric matrix showing the best gas separation performance in comparison to organic membranes made by polymer as a single-phase [12]. Synthetic zeolite as the dispersed phase is the most inorganic materials used as fillers into the MMMs, improving the possibilities of this system for gas separations [11], but only the membrane modification processes allow to eliminate or reduce the compatibility lack in the interphase between zeolite particles and polymer chains [13].
The development of organic-inorganic membranes presents improvements in the properties of permeability, selectivity as well as their thermal and mechanical properties. Due to the differences between the properties of the polymer and the inorganic phases, and the strong tendency of nanofillers to form aggregates, it is challenging to prepare MMMs. Because of the weak adhesion between the polymer and the inorganic particles, this type of membranes usually presents a variety of defects in the polymerparticle interphase [14][15][16].
The main objective has been to prepare mixed matrix membranes, where LTA zeolite was used as a dispersed phase in a polysulfone matrix. The physical and chemical properties of the zeolite and the resultant membranes were studied using different characterisation techniques, including the analysis of the influence of zeolite content on the MMMs and the final membrane performance under 1-6 bar N 2 transmembrane pressure difference.

LTA zeolite synthesis
Preparation of the LTA zeolite powders by a direct hydrothermal cycle was done following the best synthesis conditions from previous optimisation studies on the effect of the reaction times and its significant results on the zeolite crystal structures [17]. Briefly, the starting aluminosilicate gel molar composition 34 Na 2 O:Al 2 O 3 :3SiO 2 :462 H 2 O was obtained by mixing an aluminate solution with a silicate solution under magnetic stirring. The aluminate solution was prepared by dissolving 4.36 g of sodium aluminate in 80 g of deionised water at 55 ∘ C up to apparent total solubility and then the addition of 24 g of sodium hydroxide, all under vigorous magnetic stirring. The silicate solution was obtained by mixing 12.76 g of sodium silicate in 80 g of deionised water at 55 ∘ C, followed by the addition of 24 g of sodium hydroxide until a clear solution, all under vigorous magnetic stirring. Finally, mixing both previous solutions under vigorous magnetic stirring at 85 ∘ C.
The final light white aluminosilicate hydrogel was stirred for more 10 min at this temperature before being transferred to a teflon coated wall stainless steel reactor and then, placed in an oven at a temperature of 100 ∘ C. The synthesis was carried out by static hydrothermal method at autogenous pressure at the mentioned temperature with a fixed crystallisation time of 75 min. Once the crystallisation time was over, the reactor was cooled and the assynthesised product recovered by filtration, washed many times with deionised water up to constant pH of wash waters and then dried in air at 60 ∘ C for 12 h.

Mixed matrix membrane preparation
Two different samples of polymer/zeolite composite membranes were prepared by the phase inversion technique, both with the same solvent and polysulfone matrix but different amounts of zeolite fillers. Similar conditions (normal pressure and ambient temperature) were used at this step to avoid any external influence on the mechanical properties and performance.
Five steps were considered during the preparation of the MMMs: (a) Preparation of a fraction of dispersed zeolite suspension in the solvent: Dispersed zeolite powders were sieved through-out a 150 µm sieve and subject to a suspension procedure in proportions of 10 % and 20 % by weight relative to 60 cm 3 of the organic solvent methyl-2-pyrrolidone. The zeolite powders were first stirred in the solvent and then ultra-sonicated in a 50W ultrasonic bath for 15 min interval.

Characterisation
The LTA zeolite powders were characterized by XRD using an Advance Bruker D8 diffractometer (CuKα radiation at 40KV and 30mA, 2θ range from 5 ∘ to 50 ∘ , step size of 0.05, room temperature), for confirming the final phases developed during the present study as well as the degree of zeolite powders crystallinity. Scanning electron microscopy (SEM, HITACHI S-4700) was used to analyse morphology aspects of this structure. Thermogravimetry analysis was performed using STA 409 (Netzsch, Deutschland), designed for simultaneous TGA-DTA measurements at a heating rate of 5 ∘ C·min −1 under N 2 atmosphere to observe the zeolite behaviour as a function of temperature and its thermal properties. Fourier-transform infrared spectroscopy (FTIR, Bruker Optics IFS 66V/S Vacuum FT-IR) with a resolution of 2cm −1 over a wavenumber range of 4000 to 400 cm −1 in transmission setup and 64 scans was used to study the location of the different vibrational bands of the zeolite as the source of specific structural in-formation. The samples were prepared by the KBr pellets. Finally, gas adsorption-desorption isotherm (Micromeritics ASAP 2020) was used to measure the specific surface area as well as other parameters using N 2 and Ar as the adsorbate. Before isotherms measurements, the samples were degassed at 10 −1 Pa to be activated at 350 ∘ C for a minimum of 12h (ISO Standard 9277:2010-09).
The MMMs were characterised by SEM (HITACHI S-4700) and by atomic force microscopy (AFM). For the SEM studies, the MMMs were torn at room temperature to check their cross-section. Samples were gold-sputtered before testing. The AFM experiments were performed with a commercial AFM (Nanotec Electrónica S.L.) in a dry N 2 (g) atmosphere [18]. Under these conditions, the environmental humidity was held below 10% to avoid capillary forces. Si cantilevers (PPP-NCHR Nanosensors) with a force constant K = 40N/m, tip radius R tip = 10 nm and resonance frequency f 0 = 300 kHz were used. Topographic images were acquired in tapping mode with a scanning rate between 0.5 and 1 Hz and analysed by the WSxM software [18]. Besides, N 2 permeation using a lab experimental setup [19] where the mass flow controller was used to control the N 2 gas to the membrane. Finally, the measurements of membrane thickness using a digital micrometre (Mitutoyo, IP65) with precision up to 0.001 mm. Three points of the membrane effective area over 10 membranes were measured and the average thickness calculated. Figure 1 shows the XRD pattern of the as-synthesized LTA zeolite with the most critical 2θ reflection peaks at 7.2 ∘ , 10.1 ∘ , 12.5 ∘ , 16.0 ∘ , 22 ∘ , 23.9 ∘ , 27.0 ∘ , 30.1 ∘ and 34 ∘ , in relation with the characteristic peaks reported by Treacy [20]. The sharp peaks of the XRD pattern point to the high crystallinity of the as-synthesized structure. For indexing purposes, the obtained pattern was correlated with the Collection of Simulated XRD Powder Patterns for Zeolites from the Structure Commission of the International Zeolite Association [20] observing the same peaks position but variations in the intensities, perhaps due to preferred orientation or hydration grade. Although, no peaks more than the corresponding to hydrated LTA zeolite structure with a good crystallinity were observed under the synthesis conditions in the as-synthesized dispersed materials  Figure 2 shows the thermal behaviour of the LTA zeolite powders. The thermogravimetric analysis (TGA) curve depicts the hydration behaviour with a continuous water loss as increasing temperature, with a total weight loss of about 17%. The stronger evidence of water loss as the heating processes take place between 40 ∘ C and 100 ∘ C is associated with the water desorption from the surface of grains in the powdered sample. In the range from 100 ∘ C to 200 ∘ C are in connections with the external, loosely bound zeolite water [21]. The losses at higher temperatures are related to the slow desorption of water from the more tightly bound zeolite water molecules in the sodalite cages [22]. These results are correlated with the differential thermal analysis (DTA) curves that first shows an endothermic peak at 150 ∘ C attributed to initial desorption of physisorbed water with the initial formation of the anhydrous phase at this temperature. The endothermic behaviour at 385 ∘ C in DTA curve corresponds to trapped water molecules in the mentioned cages. With the temperature increase around 500 ∘ C and above the TGA, the curve shows a slight mass loss with a tendency to stabilise around 700 ∘ C, indicating the completion of the dehydration process. Above this temperature, the Na-LTA framework starts to collapse. No further weight loss up to 800 ∘ C is observed. Two exothermic peaks observed in the DTA curve over 800 ∘ C are correlated with the structural breakdown of LTA zeolite and the phase transformation to another structure. Szostak [23] associated this fact with the formation of cristobalite or low crystalline carnegiete which structure is based on cristobalite typestructure [24], then thermal stability is limited as much up to 750 ∘ C.

Characterisation of the as-synthesized zeolite
To discuss the Fourier transform infrared (FTIR) spectroscopy spectra of samples shown in Figure 3 and the bands assigned for the structure of LTA zeolite was considered the most reported point of view [25][26][27][28], which is based on the interpretation of the theoretical spectra corresponding to the vibrations of the D4R structural units in synthetic LTA zeolite [26]. The Figure 3 shows characteristic bands of as-synthesized LTA zeolite in the range 4000 to 400cm −1 from the absorption of defined frequencies of these zeolites. The 3430 cm −1 bands are close to the typical band 3400 cm −1 assigned to the stretching of H bridges attributed to the interaction that occurs in the zeolite cavities by physically adsorbed water by the solid and surface oxygen [22,23,[25][26][27]. Vibrations around 1650 cm −1 are attributed to the bending of the OH group in adsorbed water [22,29]. Vibrations reflected at 1000 cm −1 are assigned to the asymmetric stretching vibrations characteristic of νas Si-O (Si) and νas Si-O (Al) bridge bonds in TO 4 tetrahedra belonging to aluminosilicates with zeolite or sodalite structure [23,24,26] [25,28]. The results in the mid-infrared region are consistent with the XRD study, confirming the LTA zeolite synthesis results.  Figure 4 shows the crystallisation habits of the as-synthesized zeolite dispersed material with complete and precise crystal particles, focusing on a big crystal. In general, the sample consists of particles aggregates and isolated ones with welldeveloped faces and defined edges showing cubic and in some cases cubic-octahedral shapes. The crystal shape observed is consistent with the assigned space group Fm3c for LTA zeolite. Textural properties of the as-synthesized LTA zeolite summarised in Table 1 were evaluated using N 2 and Ar as adsorbate at the temperature of −195.8 ∘ C. The value for the surface area according to Brunauer-Emmett-Teller (BET) model using N 2 is relatively low, but also the obtained in the case of Ar, as adsorbate, is comparatively lower than the usual one reported for cation exchanged LTA zeolites [30]. Although it is not frequent, the use of Ar can be justified, not only because the kinetic diameter of the N 2 molecule is comparable with the effective diameter of the zeolite A channel, but also because the presence of a quadrupole moment in N 2 can result in a more significant interaction with the heterogeneous surface of the structure of the zeolite. This fact, in turn, would entail greater difficulty in the differentiation between zeolites of different pore sizes. Besides, if the material were microporous, the adsorption of N 2 in micropores occurs at values of relative pressures lower than Ar adsorption, the latter being more favourable for precise measurements of smaller micropores [31]. The results in Table 1 can be explained if we consider that the molecules of the adsorbate have had difficulty diffusing through the pores of the Na-LTA zeolite adsorbent. As in the case of the use of N 2 as adsorbate, already mentioned above, attributable not only to the shape of the N 2 molecule but to the presence of Na + ions that occupy sites in the cavities and the manifestation of the strong electrostatic field associated with them in the internal surface and its relation to the adsorption properties [32]. Also from the table results, it could be inferred that the difference in the value of the specific surface favourable to the use of Ar as adsorbate, is not determined both by the size of the pores, as by the greater volume of pores to which this adsorbate can access. However, the adsorption of Ar at −195.8 ∘ C shows a limited application for the determination of mesoporous size since, under these conditions, the temperature of the system is lower than the corresponding to the triple point [33], showing the following drawbacks: a) A complete analysis of the micro and mesoporosity is not possible, because, at −195.8 ∘ C, the Ar is approximately 6.5 degrees below its triple point, b) The analysis only makes sense for pore diameters less than 15 nm, because capillary condensation cannot be observed above that pore size. Figure 5 shows the SEM images from the top surface of the MMMs with different inorganic fillers weight content: (a) 10% and (b) 20% to the organic solvent methyl-2-pyrrolidinone. Note that zeolite particles (bright spot in the image) are embedded into the polymer and distributed to all over the membrane with good dispersion. Figure 6 shows cross-section morphology SEM images of the MMMs with 20 wt.% and the extent of zeolite particles dispersion within the polymer matrix. The compar-ison with the previous Figure 5 allows expressing that fillers are also distributed in the volume of the membrane. No apparent voids were observed at the zeolite-polymer interface so, the shaded area around in the vicinity of zeolite agglomerates cannot be estimated or assigned to membrane defects considering the SEM by itself is not enough to elucidate membranes defects and also the stress on the membrane during the manually tear.

Characterisation of mixed matrix membranes
The AFM images of the MMMs with 20wt.% content is posted in Figure 7, Figure 7a     the dark spots on the images are assigned to the membrane pores, and the brightest portions represent the highest parts of the membrane surface. Figure 7c corresponds to the height profile along the black line marked in Figure 7a. It shows the height differences and sizes of the concerned pores. By inspecting the line profile, it is possible to estimate the sizes of the pores with a funnel shape. The average outer diameter value was around and 0.4 µm and 0.5 µm in the membranes with a content of 10% wt and 20% wt respectively. Finally, Figure 7d corresponds to a 3D-view of a cubic zeolite grain embedded in the polymeric matrix. In this work, by increasing the zeolite concentration from 10 wt.% to 20 wt.%, the surface roughness parameter calculated by AFM decrease from 26.3nm to 20.3nm. It is mentioned in several scientific papers that a certain amount of filler in the polymer matrix at least causes a reduction of membrane surface roughness and it depends on factors as the hydrophilic properties of the filler [34,35]. However, other authors show otherwise concerning the roughness [36,37]. These results indicate that the addition of LTA zeolite as filler, at least affected the structure of the surface and therefore the mechanism of formation of the MMMs. Figure 8 outlines the behaviour on single N 2 gas flow through the two studied MMMs (Polysulfone matrix with different zeolite loading) as a function of the transmembrane pressure difference. Considering Darcy's law [38] when the data points can fit in a straight line, the flow rate is proportional to the transmembrane pressure difference, so there is no presence in these membranes of an ultralow flow rate of N 2 indicating that the MMMs, both at 10wt% and 20 wt.% zeolite loading exhibit high permeability. At 20wt.% zeolite loading, the MMMs N 2 flow permeability increased 130% over the 10 wt.% loading even when the thickness of the MMMs 20wt.% loading is quite similar to the other one (157 µm vs 154 µm). It is attributed to the increased fractional free volume caused by the filler assuming good compatibility between the permeability in the polymer phase and the molecular sieves, and no induced plastic deformation of polysulfone takes place in the range of the experimental feed gas pressure. Considering the slope of the lines, the permeability is smaller in 10wt.% composition, the MMMs with 10wt.% zeolite loading has a smaller pore size and therefore greater selectivity.

Conclusions
Two specific mixed matrix membranes based on different weight concentrations of zeolite in zeolite/polysulfone composite material (10wt% and 20wt%) with relation to the solvent were synthesised by phase inversion technique.
The main conclusions obtained are: • The as-synthesized LTA zeolite is characterised by thermal stability up to 750 ∘ C. • The effect of zeolite concentration modified the surface morphology at least, by altering the surface roughness. • The increment of molecular sieves loading contributes to the permeability improvement. • In general, the developed polysulfone/LTA zeolite MMMs can exhibit high permeability and good selectivity.
It is the first step to investigate the feasibility of polysulfone-zeolite based mixed matrix membranes fabricated in our lab facing future approaches to membrane research.