Optimization of CO2/H2 Separation over Ba-SAPO-34 Zeolite Membrane Synthesized by Microwave Heating

CO2/H2 separation using membrane technology is an important research area in order to obtain high purity hydrogen as one source of clean energy. Finding a suitable inorganic membrane is one of the critical issues, which needs to be explored for CO2/H2 separation. In the present study, Ba-SAPO-34 zeolite membrane was synthesized and followed by a modification process. CO2/H2 separation of the membrane was investigated by varying the independent process variables (CO2 % in the feed, pressure difference across the membrane and temperature). Modeling and optimization for the responses (CO2/H2 separation selectivity and CO2 permeance) was performed by applying response surface methodology and central composite design, which is available in Design Expert software. The accuracy of the models in predicting the response was tested by comparing with the experimental value of response and the two values were in good agreement. The optimization of the models gave CO2 permeance of 19.23 × 10−7 mol/m2 s Pa and CO2/H2 separation selectivity of 11.6 at 5% CO2 in the feed, a pressure difference of 100 kPa, and temperature of 30 °C for Ba-SAPO-34 zeolite membrane.


Introduction
In recent years, gas separation has received enormous attention among researchers due to the issues of energy security and global climate change. Numbers of articles for gas separation processes have been published [1][2][3][4][5]. Hydrogen (H 2 ) separation technologies have gained increasing importance nowadays since hydrogen is one of the important chemical sources for industries. It is also one the main energy sources for transportation fuel and electrical power generation [6]. Separation and purification are important technologies in processes for H 2 production, such as thermochemical processes. In order to obtain high purity H 2 , separation of H 2 from carbon dioxide (CO 2 ) is one such important area [5].
Common separation techniques for H 2 separation from CO 2 are physical absorption with solvents, pressure swing adsorption and cryogenic distillation [5,7]. However, these processes have a number of drawbacks such as complexity of the system, high energy consumption for solvent regeneration, equipment corrosion and flow problems caused by viscosity of solvent [8,9]. Membrane-based technology appears to be a potential alternative for H 2 separation in view of its advantages such as sustainable operation and relatively low energy consumption [7]. Palladium membranes have been extensively studied for H 2 separation due to its high hydrogen selectivity [10][11][12]. However, the usage of palladium and its alloys have a number of disadvantages, including a high sensitivity to chemicals (i.e., sulphur, chlorine and carbon monoxide in most applications) and their extremely high cost [13]. Polymeric membrane are other candidates for separation of H 2 in view of their low cost and low energy requirement [14,15]. However, the application of polymeric membrane in H 2 separation is limited by the disadvantages such as the low mechanical stability of rubbery polymers [16].
Design of Experiments is commonly used to perform optimization for the process parameters [47]. Response surface methodology (RSM), available in Design of Experiments, is a statistical tool that could allow reduction in the required numbers of experiments and could be used to investigate the effect of the significant process variable and the effect of the variables' interaction on the process [48]. There have been numerous studies that have applied RSM and hence showed RSM as an effective tool for optimization of the process [49][50][51][52].
Our previous work [53] has shown that CO 2 /CH 4 separation was selectivity improved from 30 for the H-SAPO-34 zeolite membrane to 103 for the Ba-SAPO-34 zeolite membrane. In the present work, Ba-SAPO-34 zeolite membrane was formed by modifying the presynthesized H + -form of SAPO-34 (H-SAPO-34) zeolite membrane. The Ba-SAPO-34 zeolite membrane was subjected to the CO 2 /H 2 separation process by varying three process variables, which are CO 2 % (concentration) in the feed, pressure difference and temperature.
The objective of the current study was to perform optimization on the operating process conditions of the membrane separation for the CO 2 /H 2 separation selectivity and CO 2 permeance. In current work, CO 2 permeance was reported instead of H 2 permeance because the Ba-SAPO-34 membrane was found to be CO 2 -selective over the ranges of the process variables studied.

Preparation of Zeolite Membrane
H-SAPO-34 membrane was deposited on α-alumina disc and then followed by a modification to the Ba-SAPO-34 membrane by following the procedures described in our previous work [43,53]. The synthesis precursor with the molar composition of Al 2 O 3 :P 2 O 5 :1.2T-EAOH:0.3SiO 2 :80H 2 O was prepared by mixing deionized water, aluminium isopropoxide (Al(i-C 3 H 7 O) 3 , 98%, Merck, Darmstadt, Germany), tetraethylammonium hydroxide (TEAOH, 35 wt%, Sigma-Aldrich, St. Louis, MI, USA), phosphoric acid (H 3 PO 4 , 85%, Sigma-Aldrich) and Ludox AS-40 colloidal silica sol (40 wt%). The synthesis precursor was poured into a Teflon-lined vessel with α-Alumina disc placed in the vessel. The filled Teflon-lined vessel was heated at 200 • C for 2 h in a microwave oven (MARS 5, CEM Corporation, Matthews, NC, Canada). When the microwave heating was done, rinsing and drying were performed on the membrane. The procedures for heating, rinsing and drying were repeated three times. Calcination was performed on the H-SAPO-34 membrane at 400 • C for 15 h in a furnace. In order to modify the H-SAPO-34 membrane to the Ba-SAPO-34 membrane, ion-exchange was performed on the H-SAPO-34 membrane at 70 • C for 5 h by using ion-exchange solution containing Ba 2+ . Rinsing the membrane with ethanol and followed by drying the membrane at 100 • C overnight were then carried out.
The characterization works of the Ba-SAPO-34 membrane were described in our previous work [43,53].

Design of Experiments
By using Design Expert software version 6.0.6 (STAT-EASE Inc., Minneapolis, MN, USA), Design of Experiments was applied for investigating the CO 2 /H 2 separation. The modeling and analysis of problems, which include the generation of model equations by using experimental data, determination of the effect of variables and variables' interaction on the responses and optimization studies on the responses, were performed by using RSM coupled with central composite design (CCD) [54,55].
Three independent variables were studied for CO 2 /H 2 separation in the current study, which include CO 2 % in the feed, pressure difference and temperature, as shown in Table 1. The factor code for CO 2 % in the feed, pressure difference and temperature is C, B and A, respectively. As shown in Table 1, the low level and high level are represented by −1 and +1, respectively. CO 2 separation selectivity and CO 2 permeance are the responses that were investigated in the current study. Equations (1) and (2) show the polynomial that can be investigated by using Design Expert software for the approximation for the relationship between response (y) and the set of independent variables [50,54,56]: First order model: Second order model: where y is the response, x i and x j are the independent variables, x i x j is the first order interaction between x i and x j , β 0 , β i , β ii and β ij is the regression coefficient for intercept, linear, quadratic and interaction terms, respectively, n is the number of independent variables and ε is the error.

CO 2 /H 2 Gas Separation Studies
The Ba-SAPO-34 membrane was sealed in a stainless steel module using silicone gaskets and was subjected to CO 2 /H 2 separation studies. Mass flow controllers were used to feed CO 2 and H 2 gases to the membrane module The CO 2 concentration in the feed was varied. The permeate pressure was kept at atmospheric pressure. Back pressure regulator was used to adjust the feed pressure so that the pressure difference across Ba-SAPO-34 membrane can be varied. The temperature for gas separation was varied by changing the temperature of an electronic-controlled oven where the membrane module was located. Online gas chromatography (PERKIN ELMER, CLARUS 500) equipped with CARBOXEN-1010 column and thermal conductivity detector, was used to analyze the composition of the permeate and retentate exit streams.
Permeance, P i (mol/m 2 s Pa) of the gas was determined by using Equation (3).
where ∆p i is the partial pressure difference of gas i across the membrane (Pa), J i is the flux of gas i (mol/m 2 s), the gas i may corresponds to CO 2 or H 2 . The CO 2 /H 2 separation selectivity, α CO 2 /H 2 was determined by using Equation (4).

Characterization Results of Ba-SAPO-34
The Ba-SAPO-34 membrane was characterized by using Scanning Electron Microscopy (SEM) and High-Resolution Transmission Electron Microscopy (HRTEM) in our previous work [43]. The top view SEM image and cross-sectional view SEM image of Ba-SAPO-34 membrane can be found in our previously published works [43]. It was observed from the top view SEM image of the Ba-SAPO-34 membrane that the membrane consists of orthorhombic zeolite crystals with a size of approximately 1 µm [43]. Meanwhile, the cross-sectional view SEM image of the Ba-SAPO-34 membrane showed that Ba-SAPO-34 membrane layer thickness is 4 µm approximately [43]. On the other hand, the HRTEM image of Ba-SAPO-34 can also be found in our previously published works [43] and the HRTEM showed Ba-SAPO-34 with pore channel diameter of less than 0.5 nm.

Experiment Design Matrix
In the current study, a total of 20 experiment runs for sets of independent variables (C: CO 2 % in the feed, B: pressure difference and A: temperature) was suggested by CCD for the CO 2 /H 2 gas separation studies as shown in Table 2. The values of CO 2 /H 2 separation selectivity and CO 2 permeance, which were obtained from experimental work, are shown in Table 2 as well. The CO 2 permeance was in the range of 1.96 to 19.23 × 10 −7 mol/m 2 s Pa and the CO 2 /H 2 separation selectivity was in the range of 2.3 to 12.2. The experimental runs at the temperature of 105 • C, pressure difference of 300 kPa and 27.5% CO 2 in the feed were repeated another five times (run 15-20 as shown in Table 2) in order to check for the reproducibility of the data. The low values of standard deviations (0.01 for CO 2 permeance and 0.05 for CO 2 /H 2 separation selectivity) for the repeated runs indicates good reproducibility of the responses.

Response Surface Modeling
"Inverse" transformation was used to analyze the responses of CO 2 permeance and CO 2 /H 2 separation selectivity as defined in Equation (5).
where y is the value of the response and y' is the transformed value. "Inverse" transformation was applied because this function was able to model and predict the experimental data very well. The CO 2 permeance and CO 2 /H 2 separation selectivity was modeled, analyzed in the form of 1/(CO 2 permeance) and 1/(CO/H 2 separation selectivity) respectively.

Response Surface Modeling of CO 2 Permeance
The Analysis of Variance (ANOVA) of the CO 2 permeance is shown in Table 3. Equation (6) shows the chosen quadratic model to reflect the relationship between the independent variables and the response.
where C, B and A correspond to the coded value of CO 2 % in the feed, pressure difference and temperature, respectively. In order to have the terms of the model to be significant at the 95% confidence level, the values of probability should be less than 0.0500 ("Prob > F" less than 0.0500). In this case, all the terms (A, B, C, A 2 , B 2 , C 2 , AB, AC and BC) were found to be significant for the model of 1/(CO 2 permeance). The "Lack of Fit F-value" of 1.16 implied that the Lack of Fit was not significant relative to the pure error. It is good to have non-significant Lack of Fit. Figure 1 presents the comparison between predicted 1/(CO 2 permeance) attained by using Equation (6) with the experimental 1/(CO 2 permeance). Good agreement between predicted 1/(CO 2 permeance) and experimental 1/(CO 2 permeance) is indicated by the correlation coefficient value (R 2 ) of 1.000. Hence, this reflects the high accuracy of the generated model Equation (6) to predict the 1/(CO 2 permeance) in current work.  Figure 2 that the 1/(CO 2 permeance) increased with temperature for all three levels of pressure difference. A similar trend was reported by Li et al. [42]. When the separation temperature increased from 30 to 180 • C, the surface coverage declined and the CO 2 diffusivity increased. The decline in surface coverage prevailed the increment in diffusivity when the temperature increased. Subsequently, this resulted in decline in CO 2 permeance, and hence increment in the 1/(CO 2 permeance) when the temperature increased. The 1/(CO 2 permeance also increased with pressure difference as shown in Figure 2. In the current study, gas permeance was calculated by dividing the gas flux with partial pressure difference across the membrane. The increase in CO 2 partial pressure gradient was more than the increase in CO 2 surface coverage gradient, hence leading to a decrease in CO 2 permeance with an increase in pressure difference of 100-500 kPa across the membrane in the current study. Figures 3 and 4 show that the 1/(CO 2 permeance) increased with an increase in CO 2 % in the feed from 5 to 50%. When the CO 2 % in the feed increased, the CO 2 permeance decreased. The increase in the CO 2 loading approached saturation in the membrane and thus resulted in drop in CO 2 permeance as observed and reported by Hong et al. [39].

Response Surface Modeling of CO 2 /H 2 Separation Selectivity
The ANOVA of the CO 2 /H 2 separation selectivity is shown in Table 4. Equation (7) shows the chosen quadratic model to reflect the relationship between the independent variables and the response.
where C, B and A correspond to the coded value of CO 2 % in the feed, pressure difference and temperature, respectively. The model was significant in view of its F-value of 391.84. The significance of variable's effect on CO 2 /H 2 separation selectivity decreased in the order of temperature (A) > CO 2 % in the feed (C) > pressure difference (B) with the F-value in the order of 2209.94 > 586.84 > 5.03, respectively. It is shown in Table 4 that A, B, C, A 2 , C 2 , AB, AC, BC were significant terms for the model of 1/(CO 2 /H 2 separation selectivity). However, the term of B 2 was included in Equation (7) to obtain a hierarchy model. The "Lack of Fit F-value" of 4.39 implied that the Lack of Fit was not significant relative to the pure error due to noise. Figure 5 presents the comparison of the predicted 1/(CO 2 /H 2 separation selectivity) attained by using Equation (7) with the experimental 1/(CO 2 /H 2 separation selectivity). Good agreement between predicted 1/(CO 2 /H 2 separation selectivity) and experimental 1/(CO 2 /H 2 separation selectivity) is indicated by the correlation coefficient value (R 2 ) of 0.9972. Hence, this reflects the high accuracy of the generated model Equation (7) to predict the 1/(CO 2 /H 2 separation selectivity) in the current work.  Figure 6 that the 1/(CO 2 /H 2 separation selectivity) increased when the temperature increased from 30 to 180 • C. CO 2 adsorbed more strongly on Ba-SAPO-34 membrane than H 2 , and hence permeated faster through the membrane pore despite its larger molecule kinetic diameter than H 2 [57]. Therefore, the CO 2 /H 2 separation selectivities obtained in the current study were more than 1. The values became less than 1 when the CO 2 /H 2 separation selectivities were inversed. At temperature as low as 30 • C, strong CO 2 adsorption on the membrane inhibited the adsorption and permeance of H 2 . The degree of CO 2 inhibition toward H 2 adsorption reduced due to lower CO 2 surface coverage when the temperature increased [39]. The CO 2 permeance decreased but the H 2 permeance increased, led to a drop in CO 2 /H 2 separation selectivity or in other words, an increase in 1/(CO 2 /H 2 separation selectivity) when the temperature increased. When the CO 2 % in the feed increased, the 1/(CO 2 /H 2 separation selectivity) increased, as can be seen in Figures 7 and 8.

Optimization Studies
The goal set for the responses and variables, that need to be satisfied simultaneously for optimizing the responses by using Design Expert, is presented in Table 5. It was the goal for the responses to minimize the 1/(CO 2 /H 2 separation selectivity) and the 1(CO 2 permeance), or in other words, to maximize the CO 2 /H 2 separation selectivity and the CO 2 permeance. Design Expert generated solutions (optimum conditions) with different total desirability as shown in Table 6. The desirability function approach was applied in the RSM to optimize the operating conditions in the present work. This approach was applied to determine the operating condition that results in response to the highest desirability. Each estimated response variable was transformed into an individual desirability value, d i , using the desirability function [58]. The value of the desirability varies over the range 0 ≤ d i ≤ 1 (8) where (d i = 1) reflects a completely ideal response value and (d i = 0) reflects a completely undesirable response value. Then, the individual desirability values were combined in order to determine the value of the total desirability, D, as shown in Equation (9).
where m is the number of responses. In view of the highest total desirability of 0.996 displayed by Solution 1 as shown in Table 6. Solution 1 was selected as the optimum condition.
With the optimum condition of Solution 1, 1/(CO 2 /H 2 separation selectivity) value of 0.086 and the 1/(CO 2 permeance) value of 0.052 (×10 −7 mol/m 2 s Pa) −1 , were attained, which are equivalent to CO 2 /H 2 separation selectivity of 11.6 and CO 2 permeance of 19.23 × 10 −7 mol/m 2 s Pa. An additional five experiments were conducted at the optimum operating condition (temperature of 30 • C, pressure difference of 100 kPa, 5% CO 2 in the feed) generated by Design of Experiments in order to check the accuracy of the Design of Experiments. The separation result is presented in Table 7. The experimental values of CO 2 /H 2 separation selectivity and CO 2 permeance were compared with the values predicted by using the models. The attainment of mean error of 3.64% for CO 2 /H 2 separation selectivity and the mean error of 1.46% for CO 2 permeance reflects good agreement between the predicted values and the experimental values. This implies that Design of Experiments with RSM is an accurate tool used to model and predict CO 2 /H 2 separation performance of the membrane in the current work.

Comparison of CO 2 /H 2 Separation Performance with the Other Zeolite Membranes Reported in the Literature
The Ba-SAPO-34 zeolite membrane in the present study was also compared for its CO 2 /H 2 gas separation performance with the other zeolite membranes reported in the literature, and the comparison is shown in Table 8. Yin et al. [59] prepared a stainless-steelnet-supported P/NaX composite, which displayed CO 2 /H 2 selectivity of~4.1-6.4 and CO 2 permeance of~0.18-1.68 × 10 −7 mol/m 2 s Pa. Mirfendereski et al. [60] reported H 2 /CO 2 selectivity of 0.11-0.22, which is equivalent to CO 2 /H 2 selectivity of~9.1-4.5 for ZSM-5 zeolite membranes. In the other studies reported by Aydani et al. [61], the SSZ-13 membrane was synthesized by dynamic rub coating. CO 2 permeance of 5.8 × 10 −7 mol/m 2 s Pa and CO 2 /H 2 selectivity of 17 were obtained for the synthesized SSZ-13 membrane [61]. On the other hand, CO 2 /H 2 selectivity of 17 was reported for the DDR membrane by Zito et al. [62]. Xu et al. [63] prepared Na-LTA and Cs-LTA membranes. Xu et al. [63] reported an H 2 /CO 2 separation factor of 5.9 and 8 for Na-LTA and Cs-LTA membranes, respectively. Moreover, Li et al. [64] reported the synthesis of the AlPO 4 -LTA membrane and obtained an H 2 /CO 2 separation factor of 7.3 for the AlPO 4 -LTA membrane. The reported H 2 /CO 2 separation factor values of greater than one for these membranes indicates that these membranes are H 2 -selective.

Conclusions
In this study, the CO 2 /H 2 separation process over Ba-SAPO-34 zeolite membrane was investigated. Modeling and optimization for the responses (CO 2 /H 2 separation selectivity and CO 2 permeance) as a function of the independent process variables (CO 2 % in the feed, pressure difference and temperature) was performed by applying response surface methodology and central composite design, which is available in Design Expert software. The obtained model equations were able to predict the responses over the ranges of CO 2 % in the feed, pressure difference and temperature studied. In addition, optimum CO 2 permeance of 19.23 × 10 −7 mol/m 2 s Pa and CO 2 /H 2 separation selectivity of 11.6 were obtained at 5% CO 2 in the feed, pressure difference of 100 kPa and temperature of 30 • C for the Ba-SAPO-34 zeolite membrane.