Speciﬁc Structure and Properties of Composite Membranes Based on the Torlon ® (Polyamide-imide) / Layered Perovskite Oxide

: The use of perovskite-type layered oxide K 2 La 2 Ti 3 O 10 (Per) as a modiﬁer of the Torlon ® polyamide-imide (PAI) membrane has led to the formation of an speciﬁc structure of a dense nonsymmetrical ﬁlm, namely, a thin perovskite-enriched layer (3–5 µ m) combined with the polymer matrix (~30 µ m). The PAI / Per membrane structure was studied by SEM in combination with energy dispersive microanalysis of the elemental composition which illustrated di ﬀ erent compositions of top and bottom surfaces of the perovskite-containing membranes. Measurement of water and alcohol contact angles and calculation of surface tension revealed hydrophilization of the membrane surface enriched with perovskite. The transport properties of the nonsymmetrical PAI / Per membranes were studied in the pervaporation of ethanol-ethyl acetate mixture. The inclusion of 2 wt.% Per in the PAI gives a membrane with a high separation factor and increased total ﬂux.


Introduction
Progressive development worldwide depends on technologies that provide energy conservation, as well as minimizing waste and having zero emissions. The concept of an ecological future pushes into the foreground the use of membrane technologies as a favorable substitute for traditional separation processes [1][2][3][4].
The key element determining the effectiveness of membrane processes is the membrane. Most processes are carried out at elevated temperatures, therefore the use of membranes based on polymers of a heteroaromatic structure with enhanced thermal and chemical resistance is preferable [5][6][7][8]. We chose Torlon ® polyamide-imide (PAI), a thermoplastic that is characterized by high mechanical strength, a wide temperature range of operation (from −270 • C to +250 • C), chemical resistance, and relative ease of manufacturability. These unique properties cause an interest in the manufacturing membranes based on PAI for ultra-and nanofiltration [9], gas separation [10], and pervaporation [11][12][13][14][15][16]. However, published works on pervaporation with PAI membranes had aimed alcohol dehydration and data on the performance that were not impressive. To enhance performance, the PAI membranes have been modified by blending with polyesterimide [8], polyimide [9], and gelatin [10]. In the present work, Torlon ® PAI membrane and its modified form were studied in pervaporation of organic mixture for the first time. We developed a new composite membrane by PAI modification with the perovskite-type layered oxide K 2 La 2 Ti 3 O 10 (Scheme 1). pervaporation of organic mixture for the first time. We developed a new composite membrane by PAI modification with the perovskite-type layered oxide K2La2Ti3O10 (Scheme 1).
Perovskite oxides are mixed metal compounds with a well-defined cubic structure and the general formula ABO3, where A and B are various cations, while the B cation is surrounded by an octahedron of O anions [17]. Perovskite-type layered oxide represents solid crystalline species consisting of two-dimensional nanosized plates of perovskite alternating with cations or cationic structural elements. Oxides of this type exhibit ion exchange and intercalation properties including reversible hydration of interlayer space [18,19]. These materials are attractive as components of composites due to the possibility of rearrangement of the metal sections, as well as their stability at elevated temperatures [20][21][22][23]. Currently, inorganic membranes based on perovskites are widely used for oxygen purification in catalytic membrane reactors and there are several attempts to use exfoliated perovskite nanosheets as membrane material [24]. However, there is no data on the use of perovskites as modifiers for polymer membranes for pervaporation. Perovskite-type layered oxides seem promising materials for incorporation into membranes due to combination of two factors: (i) the possibility of creating selective transport channels through the interlayer space of the perovskite structure-the presence of ionic compounds in the polymer matrix leads to an increase in the solubility of polar molecules and a decrease in the solubility of nonpolar media, and (ii) the possibility of increasing the distance between polymer chains, which should increase the permeability of the membranes due to the frame of inorganic materials that can prevent the tight packing of polymer chains [25][26][27].
This work aims to develop a method for manufacturing PAI/Per composites and dense film membranes based on them. The new composite membrane developed in this work is intended for the membrane process of pervaporation, which is used to separate liquid mixtures, including azeotropic, closely boiling, and heat-sensitive liquids [28]. The study on pervaporation separation of Currently, inorganic membranes based on perovskites are widely used for oxygen purification in catalytic membrane reactors and there are several attempts to use exfoliated perovskite nanosheets as membrane material [24]. However, there is no data on the use of perovskites as modifiers for polymer membranes for pervaporation. Perovskite-type layered oxides seem promising materials for incorporation into membranes due to combination of two factors: (i) the possibility of creating selective transport channels through the interlayer space of the perovskite structure-the presence of ionic compounds in the polymer matrix leads to an increase in the solubility of polar molecules and a decrease in the solubility of nonpolar media, and (ii) the possibility of increasing the distance between polymer chains, which should increase the permeability of the membranes due to the frame of inorganic materials that can prevent the tight packing of polymer chains [25][26][27].
This work aims to develop a method for manufacturing PAI/Per composites and dense film membranes based on them. The new composite membrane developed in this work is intended for the membrane process of pervaporation, which is used to separate liquid mixtures, including azeotropic, closely boiling, and heat-sensitive liquids [28]. The study on pervaporation separation of the ethanol-ethyl acetate mixture is of particular interest [29][30][31] because these liquids have close boiling points (EtOH-78.3 • C, EtOAc-77.0 • C) and form an azeotropic mixture of 26 wt.% ethanol and 74 wt.% ethyl acetate [32]. The impact of perovskites on morphology and physical properties are thoroughly discussed.

Synthesis of Perovskite
The perovskite K 2 La 2 Ti 3 O 10 (Per) was synthesized by a hydrothermal method from commercially available precursors. The starting materials were La 2 O 3 , 99.9% (Vekton, Saint Petersburg, Russia), TiO 2 P-25 (Evonik, Essen, Germany) and KOH, 99.9% (Vekton, Saint Petersburg, Russia) as both precursor and reaction media. The synthesis was performed in optimized conditions based on the method described earlier [33]. The stoichiometric amounts of TiO2 and La2O3 calculated for 250 mg of the product were weighed and suspended in 40 mL of 4 M KOH. The suspension was sonicated for 15 min and placed in a 50 mL Teflon-lined vessel. The vessel was then inserted into a stainless steel autoclave and heated at 230 • C. After the 72 h reaction period the vessel was naturally cooled in the air for~3 h and the precipitate was separated by centrifugation. The obtained slurry material was washed once with 40 mL of distilled water and twice with 40 mL of ethanol and dried in air overnight.
As K 2 La 2 Ti 3 O 10 is known to undergo partial substitution of interlayer potassium cations by protons in water solutions and humid air atmosphere [34], to obtain the stable phase product for membrane preparation the as-synthesized K 2 La 2 Ti 3 O 10 ·yH 2 O powder was maintained in distilled water for 24 h forming a partly substituted stable form.

Perovskite Characterization
The phase compositions characterization was studied by powder XRD-analysis using a Rigaku MiniFlex II diffractometer with Cu Kα radiation source in the range of 3-60 • , and at a scan speed of 10 • min −1 . The results of XRD-analysis show the formation of a well-crystallized, single-phase compound after the hydrothermal procedure (see Figure S1).
The morphologies of the obtained samples were carried out by scanning electron microscopy (SEM) (Zeiss Merlin, Oxford Instruments).

Membrane Preparation
The PAI/Per composites were prepared by mixing solutions of 3 wt.% PAI in NMP and Per in 10 mL of NMP in different amounts to obtain 1-3 wt.% Per in the resulting membranes. The produced solution was sonicated by ultrasound disperser Hielscher UP200St with a 7 mm probe (Hielscher Ultrasonics, Teltow, Germany) for 20 min at 40% amplitude. Then it was coated on a glass plate and left in an oven at 50 • C for a solvent evaporation. In 72 h, membranes were peeled off and placed in a vacuum oven at 60 • C to constant weight. The thicknesses of the resulted membranes were found to bẽ 33 µm.

Membrane Characterizations
Membrane morphology was observed using scanning electron microscope SEM Zeiss SUPRA 55VP (Carl Zeiss AG, Oberkochen, Germany). Before the test samples were coated with a platinum 20 nm layer by the Q150R S Plus coater (Quorum Technologies Ltd., Lewes, UK). The platinum layer of 20 nm thick was coated on the sample surface by cathode sputtering using the Quorum 150 (Great Britain) installation.
Energy dispersive microanalysis (EDS) for study on the elemental composition of samples was performed using a system of microanalysis INCA Energy with an XMax 80 OXFORD detector. Spectra of film surface were examined for phase identification in a sample. Quantitative analysis was performed using the method of fundamental parameters. The standard spectrum accumulation time was 60 s. To determine small amounts of elements, the accumulation time was increased to 300 s.
The contact angle measurements were carried out by a DSA 10-drop shape analyzer (KRÜSS GmbH, Hamburg, Germany) at room temperature and atmospheric pressure using water and n-propanol. Based on measured contact angles, the critical surface tension (σ s ) including dispersion σ d s and polar σ p s components were calculated by the Owens-Wendt method [35]: Membrane density (ρ) was measured by an immersing a sample (0.05-0.10 g) in chloroform-isopropanol solutions of various compounds in which the membrane remains suspended. Experiments were performed five times with at least two samples.
Fractional free volumes, FFV, of the PAI and composites PAI/Per were calculated by the following equations, respectively [36,37]: where V 0 = 1/ρ is the polymer specific volume and V w is the van der Waals volume of the PAI macromolecule estimated via Askadskii's group contribution method [38], ρ PAI , ρ Per , and ρ C are densities of polymer, perovskite, and composites, respectively, and wt Per is the weight fraction of the Per modifier in the polymer matrix.

Pervaporation
The pervaporation experiment was carried out on a laboratory-scale unit and the details of the apparatus have been described in [39].
The pervaporation results were described in terms of total flux and separation factor. The total flux (J) was determined as the amount of liquid penetrated through membrane area per time interval: where Q is the total weight of the permeate collected at time interval t and S is the effective surface area of the membrane. The separation factor (α) was calculated with the following equation: where x and y are the ethanol concentration in the feed and in the permeate, respectively.

Membrane Structure
The morphology of the PAI dense films modified with different amounts of perovskite was studied by scanning electron microscopy (SEM). In Figure 1 micrographs of the cross-section of PAI/Per membranes containing 0, 1, 2, and 3 wt.% Per are presented. The inclusion of perovskite into the PAI changes the film cross-section images-the morphology of the membrane is complicated. Particular  Figure 2e shows SEM images of as-prepared K 2 La 2 Ti 3 O 10 oxide particles which have a plate-like morphology typical for layered compounds with lateral sizes of about 2-5 µm and~4-5 nm thickness. According to Figure 2b-d), the sizes of the perovskite inclusions increase and plate profiles of perovskite inclusions (similar to initial oxide particle morphology) appear.
The elemental composition of the samples was investigated by energy dispersive analysis. It was observed that PAI/Per membranes made of composite materials have different compositions of the top and bottom surfaces. Table 1 shows the data of electron dispersion spectra (EDS) of the elemental analysis of the PAI/Per (2%) membrane along lines (1-5) of the cross-section near the top and bottom surfaces of the membrane. Figure 3 presents a micrograph of the cross-section of the PAI/Per (2%) membrane, where the position of these (1-5) spectral lines is indicated. For all PAI/Per membranes, the EDS of the top surface contains elements (C, N, O) that constitute the chemical formula of PAI. The composition of the bottom surface, in addition to reduced amounts (C, N, O), contains elements K, Ti, and La, which are components of the perovskite. Thus, the data of Table 1 and Figure 3 indicate the nonuniform distribution of perovskite in the PAI matrix along with the depth of the membrane. The spectrum along line 1 (top) does not contain perovskite, which appears in spectrums 3 and 4, and spectrum 5 (bottom) contains the maximum amount of perovskite elements. The filler is concentrated on one side of the membrane, in a layer~3 µm thick. This may be due to a significant difference in the density of the membrane components: PAI~1.38 g/cm 3 ; Per~4.0 g/cm 3 . When preparing the casting solution using an ultrasonic treatment, a uniform transparent PAI/Per (2%) solution is formed. However, after the solution coating to the glass plate, during the evaporation of NMP at 50 • C for 3-4 h, perovskite diffuses to the bottom under the action of gravity. uniform top and a two-component bottom surfaces. The morphology of the top surface of all membranes remains identical while the morphology of the bottom surface changes after modification by perovskites. Figure 2e shows SEM images of as-prepared K2La2Ti3O10 oxide particles which have a plate-like morphology typical for layered compounds with lateral sizes of about 2-5 µ m and ~4-5 nm thickness. According to Figure 2b-d), the sizes of the perovskite inclusions increase and plate profiles of perovskite inclusions (similar to initial oxide particle morphology) appear. The elemental composition of the samples was investigated by energy dispersive analysis. It was observed that PAI/Per membranes made of composite materials have different compositions of the top and bottom surfaces. Table 1 shows the data of electron dispersion spectra (EDS) of the elemental analysis of the PAI/Per (2%) membrane along lines (1-5) of the cross-section near the top and bottom surfaces of the membrane. Figure 3 presents a micrograph of the cross-section of the PAI/Per (2%) membrane, where the position of these (1-5) spectral lines is indicated. For all PAI/Per membranes, the EDS of the top surface contains elements (C, N, O) that constitute the chemical formula of PAI. The composition of the bottom surface, in addition to reduced amounts (C, N, O), contains elements K, Ti, and La, which are components of the perovskite. Thus, the data of Table 1 and Figure 3 indicate the nonuniform distribution of perovskite in the PAI matrix along with the depth of the membrane. The spectrum along line 1 (top) does not contain perovskite, which appears in spectrums 3 and 4, and spectrum 5 (bottom) contains the maximum amount of perovskite elements. The filler is concentrated on one side of the membrane, in a layer ~3 µ m thick. This may be due to a significant difference in the density of the membrane components: PAI ~1.38 g/cm 3 ; Per ~4.0 g/cm 3 . When preparing the casting solution using an ultrasonic treatment, a uniform transparent PAI/Per (2%) solution is formed. However, after the solution coating to the glass plate, during the evaporation of NMP at 50 °C for 3-4 h, perovskite diffuses to the bottom under the action of gravity.  A similar EDS analysis of the cross-section was carried out for membranes containing also 1 and 3 wt.% perovskite. The same tendency for an increased concentration of perovskite elements was observed on the bottom of all membranes; and in the membrane containing 3 wt.% Per, the concentration of these elements was the highest.  A similar EDS analysis of the cross-section was carried out for membranes containing also 1 and 3 wt.% perovskite. The same tendency for an increased concentration of perovskite elements was observed on the bottom of all membranes; and in the membrane containing 3 wt.% Per, the concentration of these elements was the highest. Figure 4 shows the results of the study on surfaces enriched in perovskite for membranes with 0, 1, 2, and 3 wt.% Per. With an increase of the perovskite content in the membrane, the concentration of C and N elements, which are included only in the PAI formula, decreases ( Figure 4a); and vice versa, the concentration of K, Ti, and La elements that are components of perovskite increases (Figure 4b).
In this case, the concentration of oxygen (O), which is a common element for the both PAI and perovskite, remains unchanged, during decreasing (with the polymer) and increasing (with the filler). A similar EDS analysis of the cross-section was carried out for membranes containing also 1 and 3 wt.% perovskite. The same tendency for an increased concentration of perovskite elements was observed on the bottom of all membranes; and in the membrane containing 3 wt.% Per, the concentration of these elements was the highest. Figure 4 shows the results of the study on surfaces enriched in perovskite for membranes with 0, 1, 2, and 3 wt.% Per. With an increase of the perovskite content in the membrane, the concentration of C and N elements, which are included only in the PAI formula, decreases (Figure 4a); and vice versa, the concentration of K, Ti, and La elements that are components of perovskite increases ( Figure  4b). In this case, the concentration of oxygen (O), which is a common element for the both PAI and perovskite, remains unchanged, during decreasing (with the polymer) and increasing (with the filler).   It was found that PAI/Per membranes have an extraordinary nonsymmetrical structure, unexpected for a manufacturing process such as casting on the plate. The thin perovskite-enriched layer (3-5 µm) is probably bonded by hydrogen and coordination links the polymer component (PAI) of the membrane (~30 µm), as evidenced by the mixed composition of the intermediate layers.

Surface Tension Properties
The most common characteristic of solid surface tension is the contact angle. Table 2 presents the contact angles of water and n-propanol on the surfaces of the PAI/Per membranes containing 0, 1, 2, and 3 wt.% Per. PAI is a rather hydrophilic polymer since the water contact angle on its surface is less than 90 • . It should be noted that all measurements were performed on the surface of PAI/Per membranes enriched with perovskite. The inclusion of perovskite reduces the contact angles of water and alcohol for the membranes. The data on contact angles were used for calculation of the critical surface tension of PAI/Per membranes. The polar (σ p s ) and dispersion (σ d s ) contributions to the critical surface tension (σ s ) were estimated separately. Figure 5 illustrates the dependence of the critical surface tension and its polar and dispersion components on perovskite content in the membrane. As the loading of perovskites was increased, the polar contribution increased but the dispersion contribution decreases. The critical surface tension (σ s ) increases in all cases. Thus, the inclusion of perovskite resulted in a hydrophilization of the composite membrane.
is less than 90°. It should be noted that all measurements were performed on the surface of PAI/Per membranes enriched with perovskite. The inclusion of perovskite reduces the contact angles of water and alcohol for the membranes. The data on contact angles were used for calculation of the critical surface tension of PAI/Per membranes. The polar (σ p s) and dispersion (σ d s) contributions to the critical surface tension (σs) were estimated separately. Figure 5 illustrates the dependence of the critical surface tension and its polar and dispersion components on perovskite content in the membrane. As the loading of perovskites was increased, the polar contribution increased but the dispersion contribution decreases. The critical surface tension (σs) increases in all cases. Thus, the inclusion of perovskite resulted in a hydrophilization of the composite membrane.  Fraction free volume (FFV) was calculated based on data from the densities of PAI/Per membranes measured by flotation method. FFV characterizes the elements of the free volume distributed in the polymer matrix. Their presence facilitates the diffusion of molecules of separating mixtures through polymer films and influence on flux in membrane processes. Figure 6 shows FFV values calculated for the PAI/Per films by Equation (3). The FFVs of the composites are higher than that of PAI. This can be referred to a reorganization of polymer chains because of Per incorporation and formation having of not-so-dense macromolecule packaging in the composite membranes as Fraction free volume (FFV) was calculated based on data from the densities of PAI/Per membranes measured by flotation method. FFV characterizes the elements of the free volume distributed in the polymer matrix. Their presence facilitates the diffusion of molecules of separating mixtures through polymer films and influence on flux in membrane processes. Figure 6 shows FFV values calculated for the PAI/Per films by Equation (3). The FFVs of the composites are higher than that of PAI. This can be referred to a reorganization of polymer chains because of Per incorporation and formation having of not-so-dense macromolecule packaging in the composite membranes as against the PAI membrane. This fact will play a significant role in the study of transport properties of the membranes.
Symmetry 2020, 12, x FOR PEER REVIEW 9 of 14 against the PAI membrane. This fact will play a significant role in the study of transport properties of the membranes.

Transport Properties
For nonsymmetrical PAI/Per membranes, transport properties were studied in the pervaporation of a two-component ethanol-ethyl acetate mixture. Separation of this mixture may be of independent interest since these liquids form an azeotropic mixture of 26 wt.% ethanol and 74 wt.% ethyl acetate [25].
In addition, this separation task arises during the manufacture of ethyl acetate by esterification (ethanol + acetic acid ↔ ethyl acetate + water) and the manufacture of butyl acetate by transesterification (ethyl acetate + n-butanol ↔ butyl acetate + ethanol). Considering the peculiarity of the ethyl acetate synthesis, to isolate the target component it is necessary to separate the ethyl acetate from ethanol impurities. In the butyl acetate production, ethanol and ethyl acetate act as byproducts and subsequent regeneration of these important organic solvents is required to make them reusable.
Furthermore, the separation of the alcohol-ester system is interesting in view of the assessment of the possibility to shift a chemical equilibrium in the four-component system that occurs during the chemical reaction. This can be done by isolation of the low molecular weight component (alcohol) and thereby increasing the yield of the target product. Table 3 lists some physical properties of the studied liquids. Ethanol and ethyl acetate have close boiling points and significantly different density, volume, and solubility parameters. Such a characteristic as a solubility parameter, δ, is often used to predict the interaction of a polymer and an organic solvent. In accordance with the solubility theory [40,41], the solubility of a liquid in a polymer is greater, the smaller the difference in their solubility parameters |Δδ|. The solubility parameter δ for pure PAI Torlon® is equal to 23.0 (J/cm 3 ) 1/2 . Thus, the solubility of ethanol in the studied membranes should be higher than the solubility of esters.  Figure 7 shows the dependences of ethanol concentration in the permeate on ethanol concentration in the feed for the pervaporation of ethanol-ethyl acetate mixture by using PAI and

Transport Properties
For nonsymmetrical PAI/Per membranes, transport properties were studied in the pervaporation of a two-component ethanol-ethyl acetate mixture. Separation of this mixture may be of independent interest since these liquids form an azeotropic mixture of 26 wt.% ethanol and 74 wt.% ethyl acetate [25].
In addition, this separation task arises during the manufacture of ethyl acetate by esterification (ethanol + acetic acid ↔ ethyl acetate + water) and the manufacture of butyl acetate by transesterification (ethyl acetate + n-butanol ↔ butyl acetate + ethanol). Considering the peculiarity of the ethyl acetate synthesis, to isolate the target component it is necessary to separate the ethyl acetate from ethanol impurities. In the butyl acetate production, ethanol and ethyl acetate act as by-products and subsequent regeneration of these important organic solvents is required to make them reusable.
Furthermore, the separation of the alcohol-ester system is interesting in view of the assessment of the possibility to shift a chemical equilibrium in the four-component system that occurs during the chemical reaction. This can be done by isolation of the low molecular weight component (alcohol) and thereby increasing the yield of the target product. Table 3 lists some physical properties of the studied liquids. Ethanol and ethyl acetate have close boiling points and significantly different density, volume, and solubility parameters. Such a characteristic as a solubility parameter, δ, is often used to predict the interaction of a polymer and an organic solvent. In accordance with the solubility theory [40,41], the solubility of a liquid in a polymer is greater, the smaller the difference in their solubility parameters |∆δ|. The solubility parameter δ for pure PAI Torlon ® is equal to 23.0 (J/cm 3 ) 1/2 . Thus, the solubility of ethanol in the studied membranes should be higher than the solubility of esters.  Figure 7 shows the dependences of ethanol concentration in the permeate on ethanol concentration in the feed for the pervaporation of ethanol-ethyl acetate mixture by using PAI and PAI/Per (2%) membranes. The behavior of the permeate concentration versus feed for pervaporation is significantly different from the vapor-liquid equilibrium curve (VLE) for the ethanol-ethyl acetate mixture. In pervaporation, the permeate is enriched with ethanol, and ethanol content in the permeate increases with ethanol concentration in the feed. For each composition of the feed, the position of PAI/Per (2%) membrane is slightly lower than that of the PAI membrane (88 and 86 wt.% ethanol in permeate for PAI and PAI/Per (2%) correspondingly in pervaporation of the azeotropic mixture).
permeate increases with ethanol concentration in the feed. For each composition of the feed, the position of PAI/Per (2%) membrane is slightly lower than that of the PAI membrane (88 and 86 wt.% ethanol in permeate for PAI and PAI/Per (2%) correspondingly in pervaporation of the azeotropic mixture).     From the data in Figure 8 it follows that the total flux of the membranes increases with the growth of ethanol concentration in the feed, while the flux is higher up to 1.5 times for PAI/Per (2%) membrane as compared with the unmodified PAI membrane. An opposite relationship was obtained for the separation factor (αEtOH/EtOAc) which decreases with the growth of ethanol concentration in the feed. According to αEtOH/EtOAc calculated using data on component concentrations in the feed and permeate, the selectivity of the membrane containing 2% Per is slightly lower compared with the unmodified PAI membrane.
Transport properties of membranes developed in this work were compared with literature data (not numerous) for the pervaporation of the ethanol-ethyl acetate mixture. Table 4 lists data on total flux and separation factors that have been obtained by the use of different membranes [23][24]34]. It should be noted that No 3 is nonpolymeric membranes [24]. Membranes from polydimethylsiloxane (PDMS) exhibit high total flux and a small separation factor [30]. The separation of ethanol-ethyl From the data in Figure 8 it follows that the total flux of the membranes increases with the growth of ethanol concentration in the feed, while the flux is higher up to 1.5 times for PAI/Per (2%) membrane as compared with the unmodified PAI membrane. An opposite relationship was obtained for the separation factor (α EtOH/EtOAc ) which decreases with the growth of ethanol concentration in the feed. According to α EtOH/EtOAc calculated using data on component concentrations in the feed and permeate, the selectivity of the membrane containing 2% Per is slightly lower compared with the unmodified PAI membrane.
Transport properties of membranes developed in this work were compared with literature data (not numerous) for the pervaporation of the ethanol-ethyl acetate mixture. Table 4 lists data on total flux and separation factors that have been obtained by the use of different membranes [23,24,34]. It should be noted that No 3 is nonpolymeric membranes [24]. Membranes from polydimethylsiloxane (PDMS) exhibit high total flux and a small separation factor [30]. The separation of ethanol-ethyl acetate mixture is a difficult task. According to Table 4, separation factors of known membranes are equal to (1.1-3.25) however the results of our work are superior to these data. The separation factor of the ethanol-ethyl acetate mixture, even of the azeotropic composition, is very high for the PAI and PAI/Per (2%) membranes and equal to 21 and 17, respectively. The total flux is higher for PAI/Per (2%) as compared with the unmodified PAI membrane.

Conclusions
A method for producing novel composite membranes by inclusion of up to 3 wt.% perovskite-type layered oxide K 2 La 2 Ti 3 O 10 into the Torlon ® PAI matrix was developed. This method consists of mixing solutions of these components in NMP, sonication, casting on a glass plate, followed by evaporation of the solvent.
The membrane structure was studied by SEM in combination with energy dispersive analysis of the elemental composition. It was found that PAI/Per membranes have an extraordinary nonsymmetrical structure, unexpected for a manufacturing process such as casting on the plate; and a thin perovskite-enriched layer (3-5 µm) combined with the polymer component of the membrane (PAI) (~30 µm).
The creation of a layer containing perovskite in the composite membrane leads to hydrophilization of the membrane surface, as evidenced by a decrease in the contact angles for water and alcohol with an increase in the perovskite content in the membrane.
The transport properties of the nonsymmetrical PAI/Per membrane were studied during the pervaporation of ethanol-ethyl acetate mixture and compared with literature data. It was found that PAI/Per 2% membrane has a high separation factor, mainly removing ethanol, even from a mixture of azeotropic composition. The inclusion of perovskite in the Torlon ® PAI membrane leads to an increase of the total flux.

Conflicts of Interest:
The authors declare no conflict of interest.