Large and Highly Selective and Permeable CHA Zeolite Membranes

: Large (100 cm 2 membrane area) tubular chabazite (CHA) zeolite membranes (450 nm thick) were experimentally evaluated for the separation of CO 2 /CH 4 in an industrial laboratory. An industrially relevant feed flow rate of 250 Ndm 3 / min was used. The feed pressure and temperature were varied in the ranges of 5 − 18 bar and 292 − 318 K, respectively. For a CO 2 / CH 4 feed with a molar ratio of 1:1, the experimental CO 2 /CH 4 selectivity was high at 205, and the CO 2 permeance arrived at 52 × 10 − 7 mol/(m 2 · s · Pa) at 5 bar and 292 K. As far as we know, there is no report in the literature on large CHA membranes with such high permeability and selectivity. A high CO 2 /CH 4 selectivity was also observed for a 1:4 CO 2 /CH 4 feed. However, as indicated by mathematical modeling, concentration polarization was still an issue for membrane performance, especially at high operating pressures, even though the flow rate of the feed was relatively high. Without concentration polarization, the theoretical CO 2 /CH 4 selectivity was 41% higher than the experimental value for a 1:1 CO 2 / CH 4 feed at 18 bar. The corresponding CO 2 permeance without concentration polarization was 23% higher than the experimentally observed value, reaching 34 × 10 − 7 mol/(m 2 · s · Pa). CHA membrane processes for the removal of CO 2 from CH 4 were designed, and the electricity consumption and module cost of the process were also estimated. All of the results in this study indicate a great potential of the large CHA membranes for biogas and natural gas upgrading; however, concentration polarization should be minimized in industrial processes.


INTRODUCTION
The CO 2 in natural gas and biogas is a big problem because it reduces the calorific value and could cause corrosion in transport pipelines.Therefore, the removal of the CO 2 from CH 4 is important.Membrane technology is considered a potential competitor to the current industrial separation technologies. 1−4 This decreases the life span of polymeric membranes and limits their applications.For CO 2 /CH 4 mixtures, polymeric membranes commonly applied in industry (Table 1) displayed low selectivities (20−67) and CO 2 permeances, rendering a costly process. 5In addition, the commercially used polymeric membranes in Table 1 were evaluated at low feed CO 2 concentrations owing to the plasticization of the membranes.
Zeolite membranes are ceramic membranes with excellent thermal and chemical stabilities, alongside great potential for gas separation applications because of their well-defined micropores, which result in high selectivity and permeability. 10−14 For a 1:1 CO 2 /CH 4 mixture, a reported CO 2 /CH 4 mixture selectivity was 447 on Deca-dodecasil 3 Rhombohedral (DDR) membranes at 298 K, but the CO 2 permeance was low at 1.6 × 10 −7 mol/(m 2 •s•Pa). 11DR membranes are also highly CO 2 -selective at an elevated temperature. 14The high selectivity was ascribed to the oval shape of the pore opening of DDR zeolite with the dimensions 0.36 nm × 0.44 nm, which allowed CO 2 molecules (kinetic diameter 0.33 nm) to easily permeate the membrane, while the larger CH 4 molecules (kinetic diameter 0.38 nm) permeated very slowly. 15In addition, the high CO 2 adsorption selectivity of the DDR zeolite also contributed to its high mixture selectivity. 15Mobil five (MFI) zeolite membranes with a relatively large pore size (around 0.55 nm), including Zeolite Socony Mobil−5 (ZSM-5) with high Si/Al ratios, 13 silicalite-1 with infinite Si/Al ratio, 16 and titanium silicalite-1 (TS-1), 17 have been studied intensively for gas separations.ZSM-5 membranes gave a high CO 2 permeance of approximately 70 × 10 −7 mol/(m 2 •s•Pa) for 1:1 CO 2 /CH 4 at 250 K. 13 The thin film thickness of approximately 450 nm was also one of the reasons for the high CO 2 permeance. 13However, both CO 2 and CH 4 molecules could easily pass through the larger pore size of MFI zeolite leading to a low CO 2 /CH 4 selectivity of approximately 9, which was significantly lower than that for the DDR membranes.In addition, a lower adsorption selectivity of CO 2 /CH 4 on MFI zeolite than on DDR zeolite was also observed. 18The larger pore size coupled with lower adsorption selectivity explains the lower CO 2 /CH 4 mixture selectivity of MFI membranes compared to DDR membranes.
Chabazite (CHA) zeolite (0.38 nm pore size) has shown great potential for CO 2 /CH 4 separation, owing to the pore size lies between the size of CO 2 and CH 4 molecules, selective adsorption toward CO 2 , and high surface permeability selectivity of CO 2 /CH 4 . 19Moreover, the surface permeability is a result of the surface barrier that may be a surface diffusion process with higher activation energy than the surface diffusion process in the pores. 20The disk and tubular CHA membranes developed by our group displayed high permeability and selectivity for the separation of CO 2 /CH 4 feeds. 12,21Notably, the developed tubular CHA membranes (50 cm in length, 100 cm 2 membrane area, industrially relevant size) showed excellent separation performance for dry and humid CO 2 / CH 4 feeds in our laboratory. 22For the humid CO 2 /CH 4 feed at 6 bar and 293 K, we observed a CO 2 /CH 4 mixture selectivity of approximately 198 coupled with a CO 2 permeance of 14 × 10 −7 mol/(m 2 •s•Pa). 22However, since the maximum feed flow rate of our separation setup was only 60 Ndm 3 /min, severe concentration polarization occurred for the dry mixture separation.Therefore, for the dry mixture, a relatively low CO 2 /CH 4 mixture selectivity of approximately 61 was observed at 5 bar and 293 K.However, the permeance of CO 2 (52 × 10 −7 mol/(m 2 •s•Pa)) was about 4 times higher than that of the humid mixture, despite the severe concentration polarization. 22Recently, we reported that highly permeable and selective zeolite membranes are promising materials for biogas upgrading. 5For instance, for biomethane (CO 2 4%, CH 4 96%) production from a 1:1 CO 2 /CH 4 feed (1000 N m 3 /h) using a polymeric membrane, a three-stage process is required.Our CHA zeolite membrane was able to perform the same process at 20 bar pressure and room temperature in just two stages, with low CH 4 loss (∼0.5%).The membrane area required was 4.3 m 2 for the zeolite membrane process, while a much greater membrane area of 7.3 × 10 3 m 2 was required for polymeric membrane process. 5ecause of the lower selectivity of the membranes and their larger recycling ratio, the polymeric membrane process required approximately twice the electricity consumption than that of the corresponding zeolite membrane process. 5urthermore, the high permeability of zeolite membranes means fewer expensive membrane modules are required to achieve the desired separation, thereby much lower cost. 5n this study, defect-free large (100 cm 2 membrane area) tubular CHA zeolite membranes were investigated for CO 2 / CH 4 separation in an industrial laboratory.The industrially relevant feed flow rate was 250 Ndm 3 /min.To simulate the compositions of biogas and high CO 2 content natural gas, CO 2 /CH 4 feed compositions of 1:1 and 1:4 were tested.Because the water content in the real natural gas could be as low as 1.3 ppm after pretreatments in the natural gas processing plants, only dry CO 2 /CH 4 feeds were used for separation in this study.Additionally, the temperature of the raw biogas leaving the digester is around 310 K and the pressure of the gas is near-atmospheric.However, after the removal of H 2 S and water from the gas, the temperature of the gas may be higher or lower than 310 K depending on the techniques used.Consequently, the membrane temperature was varied between 292 and 318 K in the present work.To arrive at a high flux and, thereby, low cost for the membranes, a high feed pressure is desirable.Polymeric membranes are sensitive to high partial pressure of CO 2 and the maximum operational pressure is limited.On the contrary, zeolite membranes are insensitive to the high partial pressure of CO 2 , and consequently, the membrane experiments were carried out in a wide pressure range of 5−18 bar (highest possible pressure in the setup) in the present study.Mathematical models were used to analyze the separation data and evaluate the effects of the concentration polarization.

EXPERIMENTAL SECTION
2.1.Membrane Characterization.ZeoMem Sweden AB provided the membrane tubes.The length of the membrane tubes was 500 mm, with a functional membrane area of approximately 100 cm 2 .The inner and outer diameters of the membrane tube were 7 and 10 mm, respectively.The membrane preparation procedure has been described in a patent assigned to ZeoMem Sweden AB 23 and in a previous publication in a scientific journal. 21A stainless steel cell was used to install the membrane for the permeation experiments (Figure S1 in the Supporting Information).A field emission scanning electron microscope (FEI Magellan 400) at Luleå Material Imaging and Analysis (LUMIA) was used to record the scanning electron microscopy (SEM) images.
2.2.Separation.An industrial laboratory performed the separation experiments.The schematic of the experiment is presented in Figure 1.
Before the experiments, the membranes were dried in an oven (Heraeus UT 6060).The drying conditions were 473 K and 24 h in a nitrogen atmosphere.After drying, the membrane cell was immediately closed with Swagelok end-caps to prevent recontamination with moisture and installed in the setup.
Mixtures of CO 2 /CH 4 (1:1 and 1:4) with a flow rate of 250 Ndm 3 /min were used as feeds.The separation experiments were conducted at feed pressures from 5 to 18 bar and temperatures from 293 to 318 K.The temperature was controlled by a heater before the membrane unit.Atmospheric pressure was maintained on the permeate side.A diaphragm gas flow meter was employed to record the permeate flow rate.Measurement for each data point was running for approx- Industrial & Engineering Chemistry Research imately 2 h, during which all flows, temperatures, and pressures were constant.After passing through the membrane, the retentate and permeate flows were mixed in a low-pressure gas holder (Eisenbau Heilbronn GmbH, 3.5 m 3 ) with a variable volume.The volume of the gas holder was large enough to ensure proper mixing of the gases so they could be compressed and used as feed.Due to this recycle, the gas composition was maintained constant during the experiment.The total gas volume in the setup loop was about 4 m 3 .The only loss during the separation experiment was caused by the analyzer stream for GC (approximately 1 mL sent to the vent and the other gas sent back to the setup loop from the GC sampling loop), which was negligible compared to the total gas volume.An example of data collection and curation can be found in the Supporting Information.Equation 1 was used to calculate the separation factor where x and y represent the concentrations in the mole on the feed and permeate sides, respectively.The flux J i and permeance Π i for component i were calculated as follows where F i (mol/s) represents the permeate flow rate of i, A m (m 2 ) is the actual membrane area for separation, and ΔP i (Pa) is the difference in the transmembrane pressure of i.Because the partial pressures of CO 2 and CH 4 varied along the large membrane on the feed side, an incremental membrane model was used to calculate the permeance.The partial pressure of each component at each increment was estimated by using the incremental membrane model.The membrane was divided into 10 area increments, and the permeances were assumed to be constant for all increments.The composition on the feed side of each increment was calculated by using the material balance.Equation 4 was used to calculate the mixture selectivity (S i/j ) of component i over j

Modeling.
The method for estimating the concentration polarization index (CPI) was described in previous work. 24Briefly, the CPI was calculated using where n m and n b are the molar concentrations of the species that preferentially permeate through the membrane at the interface of the gas bulk and zeolite film, and in the gas bulk, respectively.The corrected separation results, i.e., the separation without the effect of concentration polarization, were calculated by assuming that the observed fluxes were proportional to the partial pressure difference over the membrane, as described in our previous work. 22For completeness, the separation performance without the contribution from concentration polarization was also estimated by a mathematical model developed in our previous study, as the model was fitted to experimental data recorded on a small disk membrane without any significant concentration polarization. 20quation 6 was employed to calculate the flux J in the model Here, α (m/s) is the surface permeability, D (m 2 /s) represents the diffusivity, L (m) is the thickness of the membrane (450 nm), and the pore volume fraction ε is 0.382 for CHA zeolite. 15C eq is the gas concentration in the pores of the zeolite at equilibrium, which was calculated from the ideal adsorbed solution theory. 25Superscripts f and p indicate the feed and permeate sides, respectively.Equation 7 was used to calculate the surface permeability α i k j j j j i k j j j y where α* (m/s) is the surface permeability when the loading was zero at 300 K, C sat (kmol/m 3 ) is the saturated adsorption capacity, and E α (kJ/mol) represents the activation energy for surface permeability.The values (in Table 2) of E α , α*, and C sat were taken from a previous study. 20e dependence of the diffusivity on temperature was described using the following equation i k j j j j i k j j j y Diffusivity could probably be affected by other factors, e.g., pressure apart from temperature; 26 however, in this study, the diffusivity was assumed to be pressure independent as in our  previous work. 19,20In this equation, D 0 (300) is the diffusivity at 300 K and E diffusion is the activation energy for diffusion.Table 2 shows the values of the parameters for the mathematical model in this study.An example of the mathematical model calculation can be found in the Supporting Information.

Cost Analysis.
The electricity need of the process and the cost of the membrane modules were estimated for zeolite membranes and polymeric membranes for the removal of CO 2 from a feed of 1000 N m 3 /h biogas comprising a 50:50 CO 2 / CH 4 mixture.The retentate was biomethane comprising a 4:96 CO 2 /CH 4 mixture, which is suitable as a vehicle fuel.Processes were designed to arrive at a methane slip of less than 1%.The zeolite membrane process was designed based on the experimental data observed in the present study, while the polymeric membrane process was designed based on data available in the literature. 5The processes were designed for a feed pressure of 10 bar.The total membrane area was estimated by using an incremental model, and the membrane area of each increment was 100 cm 2 .A mass balance was used to calculate the composition of the feed and permeate as well as the flow for each increment.
The compressor power was estimated using the following equations In eq 9, 27 T 1 is the initial temperature of the gas and T 2 is the temperature of the gas after compression, which was calculated using eq 10.The constant C p represents the specific heat capacity of the mixture at constant pressure, which was calculated by using eq 11.The isentropic (η c ) and the motor (η m ) efficiencies are 0.80 and 0.75, 28 respectively.The parameter ṁis the mass flow rate.
In eq 10, P 1 is the pressure before compression and P 2 is the pressure after compression.The constant γ is 1.3, 29 which is equal to C p /C v .The constant C v is the specific heat capacity at a constant volume.
In eq 11, C pCHd 4 is 2.25 kJ/(kg•K) and C pCOd 2 is 0.85 kJ/(kg• K). 29 For small-scale production in a nonautomated process, the production cost of zeolite membranes including modules has been estimated to 32 200 USD/m 2 membrane area by Dr.-Ing.Hannes Richter and Jan-Thomas Kuḧnert at Fraunhofer IKTS, which corresponds to 36 000 USD/module. 5 The production costs in an automated process would, of course, be much lower, and on the other hand, a vending price resulting in a good margin is necessary.Consequently, a vending price of 36 000 USD/module was assumed for the zeolite membranes.For industrially produced polymeric membranes including modules, the vending price is approximately 100 USD/m 2 . 30,31nce a polymeric membrane module may accommodate approximately 150 m 2 polymeric membranes, 5 the vending price would be 15 000 USD/module.

Membrane Morphology.
The SEM image of the top view in Figure 2 shows a clean, uniform, and well-intergrown zeolite film, without any cracks or pinholes.The zeolite film layer was approximately 450 nm thick, and the pores of the support remained open.The uniformity of the thickness and surface of the membrane was also confirmed by SEM images.Figure S2 in the Supporting Information shows SEM images taken at different positions of a membrane.
3.2.Separation of a 1:1 CO 2 /CH 4 Feed.The filled blue squares in Figure 3 illustrate the experimentally observed mixture selectivity, permeance, separation factor, and flux for a CO 2 /CH 4 feed with a molar ratio of 1:1 at 292 K and 5−18 bar.A CO 2 /CH 4 mixture selectivity was very high at 205 at 5 bar.The high selectivity for the mixture was ascribed to the high surface permeability selectivity of CO 2 /CH 4 . 19Slightly lower experimental separation selectivities were observed at higher feed pressures; however, a high selectivity of approximately 179 was still observed at 18 bar.At 5 and 18 bar, the corresponding experimental permeances of CO 2 were 52 × 10 −7 and 28 × 10 −7 mol/(m 2 •s•Pa).Under similar conditions, we observed significantly higher permeances compared to the values reported by other groups for CHA membranes 32,33 because of the thin zeolite film layer and clean support, as shown in the SEM images.At 18 bar, the experimental flux and separation factors for CO 2 and CO 2 /
The experimentally observed separation results from 292 to 318 K at 18 bar are presented in Figure 4.At 292 K, the experimental permeance of CO 2 was 33 × 10 −7 mol/(m 2 •s•Pa) and separation selectivity of CO 2 /CH 4 was 148.The high mixture selectivity was ascribed to the high surface permeability selectivity as well as the high adsorption selectivity of CO 2 /CH 4 at low temperatures.As reported  Industrial & Engineering Chemistry Research earlier, the surface permeability selectivity and adsorption selectivity decreased with increasing temperature. 19This resulted from the higher surface permeability activation energy of CH 4 compared to that of CO 2 .In this case, the mass transport of CH 4 was more favorable when the temperature was increased.Thus, the mixture selectivity for CO 2 /CH 4 was lower at higher temperatures.The experimental selectivity for the mixture decreased from 148 to 69 when the temperature increased from 292 to 318 K, but a slightly higher CO 2 permeance at approximately 36 × 10 −7 mol/(m 2 •s•Pa).

Separation of a 1:4 CO 2 /CH 4 Feed.
The separation for the 1:4 CO 2 /CH 4 feed was evaluated in the pressure range of 5−18, with the temperature fixed at 292 K (Figure 5).At 5 bar, the experimental CO 2 /CH 4 selectivity was 97 and a slightly lower selectivity of 82 was observed at 18 bar.At 5 and 18 bar, the CO 2 permeances observed in the separation experiments were quite high at 49 × 10 −7 and 35 × 10 −7 mol/ (m 2 •s•Pa), respectively.Although the CO 2 permeances decreased at higher feed pressures, they were still much higher than the permeances reported by other groups (Table 3).The separation factors of the CO 2 /CH 4 and CO 2 flux observed in the experiment were higher at higher feed pressures.Notably, the CO 2 flux was 8 times higher when the feed pressure was increased by a factor of 3.6 (Figure 5d).
The separation of a 1:4 CO 2 /CH 4 feed was also investigated at temperatures from 292 to 318 K and 18 bar (Figure 6).At 292 K, the mixture selectivity of CO 2 /CH 4 and corresponding CO 2 permeance observed in the experiment were 82 and 35 × 10 −7 mol/(m 2 •s•Pa), respectively.The separation factor of CO 2 /CH 4 and CO 2 flux at the corresponding experiment conditions were 55 and 0.8 mol/(m 2 •s), respectively.As in Section 3.2, the selectivity toward CO 2 decreased with increasing temperature.However, the experimentally observed CO 2 permeance remained almost constant in the studied temperature range.
The separation results of the studied CHA membrane also outperform those of the polymeric membranes used in industry (Table 1).These results are also summarized in Figure 7 for a better comparison.For instance, for the most commonly used cellulose acetate membranes in the industry, the reported CO 2 permeance and CO 2 /CH 4 mixture selectivity   1) and CHA membranes studied in this work.were 0.8 × 10 −9 mol/(m 2 •s•Pa) and 20, respectively. 7Matrimid membranes with a higher CO 2 permeance of approximately 3.8 × 10 −9 were developed; however, it was still much lower than that of zeolite membrane.An even higher CO 2 permeance of 2.4 × 10 −7 mol/(m 2 •s•Pa) has been observed for ethylene oxide-containing thin-film composite (TFC) membranes in a research laboratory, but low CO 2 /CH 4 selectivity of 33. 34.4.Discussion of the Effects of Concentration Polarization.The CPI reached a maximum of 0.94 at 5 bar for the 1:1 CO 2 /CH 4 feed (Figure 8) when the feed flow rate was 250 Ndm 3 /min.Figure 8 shows that the CPI values were lower at higher feed pressures, owing to the higher CO 2 flux and lower gas velocity in the membrane tube at higher pressures.Note that the data presented in Figure 8a,b represent two experimental series, and consequently, there is a small difference for the experimental points recorded at 18 bar and 292 K.The exact CPI values can be found in the Supporting Information.
Subsequently, the corrected selectivity and permeance without a contribution from concentration polarization were estimated.The unfilled dots in Figures 3−6 illustrate the results.Figure 3 shows the corrected mixture selectivity, permeance, separation factor, and flux for the 1:1 CO 2 /CH 4 feed.Evidently, the corrected separation performance was much higher than that of the experimental results.For instance, at 18 bar, increases of 23 and 41% in CO 2 permeance and CO 2 /CH 4 mixture selectivity were obtained without contribution from concentration polarization.The corrected separation factor of CO 2 /CH 4 and corresponding corrected CO 2 flux were also much higher, at approximately 227 and 2.53 mol/(m 2 •s), respectively.These results indicate that the separation was affected more severely by concentration polarization at higher pressures owing to the lower feed velocity and higher flux at elevated pressures.
For a 1:4 CO 2 /CH 4 feed at 18 bar and in the temperature range 292−318 K, the separation performance was severely affected by concentration polarization, as indicated by the low CPI values in Figure 8. Figure 6 shows the corrected CO 2 /CH 4 mixture selectivity and CO 2 permeance were 130 and 53 × 10 −7 mol/(m 2 •s•Pa), respectively, at 292 K.These values are 51 and 59% higher, respectively, than the experimentally observed values at the same temperature.The corrected CO 2 /CH 4 mixture selectivity reached 84 at the high separation temperature of 315 K, which was 65% higher than the experimentally observed result.
The separation performance for the 1:4 CO 2 /CH 4 feed at 292 K and 5−18 bar was also corrected for concentration polarization, as shown by the open squares in Figure 5.At 5 bar, the corrected CO 2 /CH 4 selectivity was 111, which was 14% higher than the experimentally observed value, with a corrected CO 2 permeance of 55 × 10 −7 mol/(m 2 •s•Pa).Even larger differences can be seen between the corrected separation selectivities and the experimental selectivities at higher feed pressures.For instance, at 18 bar, the corrected mixture selectivity was 1.6 times the experimentally observed mixture selectivity, owing to the low CPI of 0.76.At 5 bar, the high CPI of 0.98 resulted in similar values for both the corrected and experimentally observed CO 2 /CH 4 separation factor and CO 2 flux (Figure 5c,d).However, at 18 bar, owing to the low CPI of 0.76, the corrected CO 2 /CH 4 separation factor reached 87, which was 58% higher than the experimental separation factor.In this case, the corrected CO 2 flux was also 50% higher than the experimental value, arriving at 1.2 mol/(m 2 •s).
These results indicate that the ultimate separation performance of high-flux zeolite membranes can be realized only after minimizing concentration polarization.The appropriate design of the membrane modules could increase the flow rate and turbulence to minimize concentration polarization or, alternatively, a spacer could be used to reduce the volume and increase the turbulence on the feed side of the tubular membranes.Our estimates show that for a 1:1 CO 2 /CH 4 feed mixture, the CPI value at a feed pressure of 18 bar would increase from 0.83 to 0.91 if the feed velocity would increase 100% by e.g., increasing the feed flow rate 100%.
A mathematical model (eqs 5−7) was also used to estimate the separation performance, and the dashed curves in Figures 3−6 illustrate the results.As shown in Figure 3, the correlation is strong between the corrected experimental data and the data obtained from the mathematical model for the 1:1 CO 2 /CH 4 feed.As illustrated in, e.g., Figure 3a, the model predicts an increased selectivity with increased feed pressure.This is a result of a faster increase of the surface permeability α for CO 2 , due to a faster increase of C eq for CO 2 with increasing pressure, see eq 7.This is an effect of the higher b-value for CO 2 than for CH 4 . 20The mathematical model failed to predict reasonable data because of the very small partial pressure difference of CO 2 over the membrane for the 1:4 CO 2 /CH 4 feed at feed pressures lower than 8 bar.This is a result of small errors in the estimated adsorbed concentrations by the IAST.Therefore, no data for the model are shown in Figure 5 for pressures lower than 8 bar.It should be pointed out that the model was fitted to experimental data for small disk membranes (2.8 cm 2 membrane area) reported elsewhere, 20 and no fitting was carried out in the present work.Still, the model can describe most of the experimental data observed in the present work surprisingly well, while the model deviates from some experimental data observed in the present work.Moreover, in the present work, we successfully estimated the separation selectivities and permeances of the large membrane for the 1:1 CO 2 /CH 4 feed under all investigated conditions, and for the 1:4 feed at feed pressures higher than 8 bar, with decent agreement between the experimental, corrected, and modeled results.
The observed separation results on the large-size CHA membranes in this work were much better than the results reported on small-and large-size membranes by other groups, as summarized in Table 3.For instance, for a CO 2 /CH 4 feed with a molar ratio of 1:1 at 293 K and 1.4 bar, a CHA membrane exhibited a high CO 2 /CH 4 mixture selectivity of 153 and a high CO 2 permeance of 48 × 10 −7 mol/(m 2 •s•Pa).However, the membrane area was small, about 8 cm 2 , and the used feed pressure was also low.A relatively larger CHA membrane (tube, 45 cm 2 membrane area) showed a higher permeance of 16 × 10 −7 mol/(m 2 •s•Pa) for CO 2 at a feed pressure of 20 bar; however, still, the CO 2 /CH 4 mixture selectivity was relatively low about 72 at 298 K. 42 In a separate study, for a 1:1 CO 2 /CH 4 feed at 3 bar and 298 K, Zhou et al. observed a selectivity of 132 for 1:1 CO 2 /CH 4 mixture on SSZ-13 membranes (85 cm 2 membrane area), but the permeance of CO 2 was only 4.6 × 10 −7 mol/(m 2 •s•Pa). 41.5.Estimation of Electricity Consumption and Module Cost.We designed a CHA membrane process for the removal of CO 2 from biogas based on the experimental data reported in the present study.For comparison, a polymeric membrane process (Figure 9a) was designed based on data reported in the literature.We assumed that the processes were operated at feed pressure and feed temperature of 10 bar and 20 °C, respectively.For the CHA membrane process, a CO 2 permeance of 45 × 10 −7 mol/(m 2 •s• Pa) and a CO 2 /CH 4 selectivity of 200 were assumed based on the experimental data observed for feeds of 1:1 CO 2 /CH 4 and 1:4 CO 2 /CH 4 mixtures (see Figures 3 and 5).For the polymeric membrane, a CO 2 permeance of 5 × 10 −9 mol/(m 2 • s•Pa) and a CO 2 /CH 4 selectivity of 50 were assumed based on data reported for the best polymeric membranes in the literature. 5Figure 9 shows the process schemes, and the   4.
For the CHA zeolite membrane process, a two-stage process with recycling was necessary to arrive at a methane slip of less than 1%.For this process, the estimated total membrane area was 7.8 m 2 .Due to the lower selectivity of polymeric membranes, a three-stage process is needed to arrive at a methane loss of less than 1%.Due to the low permeability of polymeric membranes, the total membrane area for the three stages is as much as 15.5 × 10 3 m 2 .In previous work, we designed zeolite membrane modules based on 1.2 m singlechannel membrane tubes with 43 tubes in each module, with a membrane area in each module of 1.12 m 2 . 5Consequently, 7 modules are sufficient assuming the same module design as reported previously. 5In this case, the total module cost for this membrane process was about 0.5 million USD.For bundles of polymeric membrane capillaries, we also designed modules with an outer diameter of 12 cm and a length of 1.7 m with a membrane area of 152 m 2 in each module.Consequently, 102 modules with a cost of approximately 1.5 million USD would be needed.Due to the low recycle ratio of 18% for the zeolite membrane process, the estimated electricity use was as low as 0.20 kWh/Nm 3 raw biogas.This was only about 63% of the electricity used for the polymeric membrane process, with a much higher recycle ratio of 57%.The methane slip is as low as 0.82% for the two-stage CHA membrane process as shown in Figure 9b.Due to additional membrane modules and stage for the polymeric membrane process, the estimated methane slip would be as low as 0.28%, for a selectivity of 50.

CONCLUSIONS
In summary, a relatively high feed flow rate was used to evaluate the CO 2 /CH 4 separation by large tubular CHA membranes in an industrial laboratory.The observed membrane performance was excellent under all of the studied conditions.Owing to the relatively high feed flow rate, the effect of concentration polarization was low at low feed pressures; unfortunately, concentration polarization greatly affected the separation at high feed pressures.The corrected experimental results without the effect of concentration polarization showed that a higher membrane performance could be obtained if concentration polarization was controlled.Despite this, the separation factor and CO 2 flux observed in this study outperformed the separation performance of other reported CHA membranes, and laboratory-studied and commercially available polymeric membranes.The cost analysis showed a low module cost and low electricity consumption of the CHA membrane processes for biogas upgrading.This work demonstrates that large tubular CHA membranes are very promising for industrial CO 2 /CH 4 separation.
Photograph of a CHA membrane tube and a stainless steel cell; SEM images recorded from different positions of a CHA membrane tube; separation data and curation; CPI values at different feed conditions; an example of how to correct for concentration polarization; and an example of the mathematical model calculation (PDF)

Figure 1 .
Figure 1.Scheme of the membrane separation process in the industrial laboratory.

Figure 2 .
Figure 2. SEM images from the membrane top (a) and cross section (b) in the middle of the 500 mm membrane tube.

Figure 3 .
Figure 3. Mixture selectivity (a), permeance (b), separation factor (c), and flux (d) for a CO 2 /CH 4 feed with a molar ratio of 1:1 at 292 K. Filled squares represent the experimentally observed results, open squares represent the results after correcting for concentration polarization, and dashed lines represent the results from the mathematical model (eqs 5−7).

Figure 4 .
Figure 4. Mixture selectivity (a), permeance (b), separation factor (c), and flux (d) for a CO 2 /CH 4 feed with a molar ratio of 1:1 at 18 bar.Filled squares represent the experimentally observed results, open squares represent the separation results after correction for concentration polarization, and dashed lines represent the results from the mathematical model (eqs 5−7).

Figure 5 .
Figure 5. Mixture selectivity (a), permeance (b), separation factor (c), and flux (d) for the 1:4 CO 2 /CH 4 feed at 292 K. Filled squares are the experimentally observed results, open squares represent the separation results after correcting for concentration polarization, and dashed lines represent the results from the mathematical model (eqs 5−7).

Figure 6 .
Figure 6.Mixture selectivity (a), permeance (b), separation factor (c), and flux (d) for a 1:4 CO 2 /CH 4 feed at 18 bar.Filled squares represent the experimentally observed results, open squares represent the separation results after correcting for concentration polarization, and dashed lines represent the results from the mathematical model (eqs 5−7).

Figure 7 .
Figure 7. CO 2 permeance and CO 2 /CH 4 separation selectivity of polymeric membranes commonly used in industry (data from Table1) and CHA membranes studied in this work.

Table 1 .
Selectivity and Permeance for CO 2 /CH 4 Mixtures Reported for Polymeric Membranes Commonly Used in Industry

Table 2 .
20mmary of the Parameters Used in the Mathematical Model20

Table 3 .
Summary of Reported Permeances and Selectivities for CO 2 /CH 4 Feeds Using CHA-type Zeolite Membranes in the Literature and in This