CH4 separation: Preparation, characterization, separation performance and economic evaluation

CH4 separation: Preparation, characterization, separation performance and economic evaluation

Journal of Membrane Science 487 (2015) 141–151 Contents lists available at ScienceDirect Journal of Membrane Science journal homepage: www.elsevier...

2MB Sizes 2 Downloads 244 Views

Journal of Membrane Science 487 (2015) 141–151

Contents lists available at ScienceDirect

Journal of Membrane Science journal homepage: www.elsevier.com/locate/memsci

SAPO-34 Membranes for N2/CH4 separation: Preparation, characterization, separation performance and economic evaluation Shiguang Li a,n, Zhaowang Zong b, Shaojun James Zhou a, Yi Huang c,d, Zhuonan Song c,d, Xuhui Feng b, Rongfei Zhou e, Howard S. Meyer a, Miao Yu c,d,nn, Moises A. Carreon b,nnn a

Gas Technology Institute, 1700 S Mount Prospect Road, Des Plaines, IL 60018, United States Chemical and Biological Engineering Department, Colorado School of Mines, Golden, CO 80401, United States c Department of Chemical Engineering, University of South Carolina, Columbia, SC 29208, United States d SmartState Center of Catalysis for Renewable Fuels, University of South Carolina, Columbia, SC 29208, United States e State Key Laboratory of Materials-Oriented Chemical Engineering, College of Chemistry and Chemical Engineering, Nanjing Tech. University, Nanjing 210009, PR China b

art ic l e i nf o

a b s t r a c t

Article history: Received 30 January 2015 Received in revised form 25 March 2015 Accepted 29 March 2015 Available online 6 April 2015

SAPO-34 membranes were synthesized by several routes towards N2/CH4 separation. Membrane synthesis parameters including water content in the gel, crystallization time, support pore size, and aluminum source were investigated. High performance N2-selective membranes were obtained on 100nm-pore alumina tubes by using Al(i-C3H7O)3 as aluminum source with a crystallization time of 6 h. These membranes separated N2 from CH4 with N2 permeance as high as 500 GPU with separation selectivity of 8 at 24 1C for a 50/50 N2/CH4 mixture. Nitrogen and CH4 adsorption isotherms were measured on SAPO-34 crystals. The N2 and CH4 heats of adsorption were 11 and 15 kJ/mol, respectively, which lead to a preferential adsorption of CH4 over N2 in the N2/CH4 mixture. Despite this, the SAPO-34 membranes were selective for N2 over CH4 in the mixture because N2 diffuses much faster than CH4 and differences in diffusivity played a more critical role than the competitive adsorption. Preliminary economic evaluation indicates that the required N2/CH4 selectivity would be 15 in order to maintain a CH4 loss below 10%. For small nitrogen-contaminated gas wells, our current SAPO-34 membranes have potential to compete with the benchmark technology cryogenic distillation for N2 rejection. & 2015 Elsevier B.V. All rights reserved.

Keywords: Gas separation Inorganic membrane Zeolite Nitrogen rejection N2/CH4 separation

1. Introduction The typical U.S. natural gas pipeline specification for inert gases is less than 4%. Approximately 14% of U.S. natural gas contains 44% nitrogen, and thus nitrogen rejection technologies are required [1]. Most of the nitrogen rejection plants in the U.S. use a cryogenic distillation process. The basic principle is to liquefy the natural gas to the point where most of the gas is in the liquid phase and then distill the resulting liquid. Cryogenic plants are most suited to large natural gas wells that can deliver 50–500 MMSCFD (million standard cubic feet per day) for 10–20 years [1]. These large wells allow the high capital cost of the cryogenic plant to be defrayed over several years. Principal variations in the cryogenic process involve whether to use one or two columns (at different pressures). The cost of cryogenic process is high and

n

Corresponding author. Tel.: þ 1 847 544 3478. Co-corresponding author. Tel.: þ 1 803 777 5207. Co-corresponding author. Tel.: þ1 303 273 3329. E-mail addresses: [email protected] (S. Li), [email protected] (M. Yu), [email protected] (M.A. Carreon). nn

nnn

http://dx.doi.org/10.1016/j.memsci.2015.03.078 0376-7388/& 2015 Elsevier B.V. All rights reserved.

sensitive to the plant size. For a plant capacity of 20 MMSCFD assuming $4/MSCF (thousand standard cubic feet) gas price to the plant and 15 vol% N2 in feed, the cost is approximately $0.80/MSCF, whereas for a plant capacity of 1 MMSCF/day, the cost is as high as about $2.40/MSCF [2]. In addition to the cryogenic processing plants, there are a few pressure swing adsorption (PSA) plants using adsorbents such as Nitrotec Energy's carbon molecular sieves, and Engelhard's molecular gate absorbents (now part of the BASF Group). The costs are about $1.50–1.60/MSCF for a plant capacity of 1 MMSCFD [2]. Membrane separation offers the advantage of low energy cost relative to the more established gas separation processes such as adsorption and cryogenic distillation. For membrane separation, the challenge is to develop membranes with the necessary nitrogen/methane separation characteristics. Two types of membranes can be used in N2 rejection: CH4-selective or N2-selective membranes (Fig. 1). CH4-selective membranes: For CH4-selective membranes, CH4 is more permeable than N2 so that it is concentrated in the lowpressure permeate side, and needs to be recompressed to pipeline pressure before entering the pipeline (Fig. 1a). Meyer and Henson

142

S. Li et al. / Journal of Membrane Science 487 (2015) 141–151

Feed stream 15% N2 85% CH4

Feed stream Reject stream

Product stream (4% N2, 96% CH4)

Product stream (4% N2, 96% CH4)

15% N2 85% CH4 Reject stream

Fig. 1. Membrane processes for N2 rejection using (a) CH4-selective membranes, and (b) N2-selective membranes.

[3] evaluated the economics for CH4-selective membranes. Various single-, two- and three-stage systems were analyzed in their study. The estimated costs were too high to be of commercial interest for the state-of-the-art CH4-selective membranes unless the CH4/N2 selectivity could be increased to greater than 7. N2-selective membranes: For a process using N2-selective membranes, CH4 remains in the retentate stream (Fig. 1b) at high pressures, and thus there is a significant saving in recompression cost as compared to CH4-selective membranes. However, the state-of-theart N2-selective polymeric membranes display low N2 permeabilities, typically less than 10 barrers (1 barrer¼10  10 cm3 (STP)  cm/(cm2 s cm Hg)) at a selectivity of approximately 3, as shown in the “Robeson” plot [4]. Recently, some inorganic materials have been made into N2selective membranes and shown N2/CH4 separation performances above the so-called upper bound observed for dense polymers [4]. Ning and Koros [5], using a carbon molecular sieve (CMS) membrane, obtained a N2/CH4 selectivity of 7.7 and a N2 permeability of 6.8 barrers at 24 1C. Li et al. [6,7] prepared membranes with SAPO-34 deposited as a continuous layer on stainless steel tubular supports. The membranes separated N2 from CH4 with a selectivity of 4 at 22 1C. The N2 permeance was 2.1  10  8 mol/ (m2 s Pa) (¼ 63 GPU (Gas Permeation Unit), 1 GPU ¼3.348  10  10 mol/(m2 s Pa)). Very recently, the N2/CH4 separation performance through the SAPO-34 membranes has been improved by the same research group. The N2/CH4 selectivity was 5–7 and the N2 permeance was  1  10  7 mol/(m2 s Pa) (¼300 GPU) for a feed pressure of 350 kPa as reported by Wu et al. [8], for membranes on alumina supports. They also prepared SSZ-13 zeolite membranes for N2/CH4 separation and obtained selectivity of 13 and N2 permeance of 2.2  10  8 mol/(m2 s Pa) (¼66 GPU) at 20 1C for a feed pressure of 270 kPa. Compared to the SSZ-13 membranes, SAPO-34 membranes are more attractive in a real-world N2 rejection for the natural gas processing because of their high N2 permeance. SAPO-34 is a silicoaluminophosphate having the composition SixAlyPzO2 where x ¼0.01–0.98, y ¼0.01–0.60, and z¼0.01–0.52 [9]. The SAPO-34 structure is formed by substituting silicon for phosphorous in the AlPO4, which has a neutral framework and exhibits no ion exchange capacity [9]. Most of the SAPO-34 membrane studies have been focused on separating CO2 (kinetic diameter: 0.33 nm) from CH4 (kinetic diameter: 0.38 nm) [10–18]. Compared to CO2/CH4 separation, the separation of N2/CH4 mixtures using SAPO-34 membranes is much more challenging because the kinetic diameters of N2 (0.364 nm) and CH4 differ by less than 5% and both gases adsorb weakly on SAPO-34 [11]. The objective of the current study was to develop SAPO-34 membranes towards N2/CH4 separation and determine the potential of SAPO-34 membranes for N2 rejection in natural gas processing. In both N2/CH4 separation studies by Li et al. [6,7] and Wu et al. [8], the SAPO-34 membranes were prepared by single template of tetraethylammonium hydroxide (TEAOH) as structure directing agent. In the current study, SAPO-34 membranes were prepared by dual-template TEAOH and dipropylamine (DPA) for N2/CH4 separation. In addition, synthesis parameters including water content in

the gel, crystallization time, support pore size, and aluminum source were investigated and optimized. The SAPO-34 prepared by the optimized conditions has much higher N2 permeability and slightly higher N2/CH4 selectivity than the previously reported membranes [6–8]. Nitrogen and CH4 adsorption isotherms were measured on SAPO-34 crystals collected from the best SAPO-34 membrane. The heats of adsorption for N2 and CH4 on SAPO-34 were also measured to explain the effect of competitive adsorption in mixture separation. The dependences of the N2/CH4 separation performance on temperature and feed pressure were also studied. In addition to membrane preparation, characterization, and separation performance studies, Aspen HYSYS software was used to model the one-stage membrane process to see the potential of our SAPO-34 membranes for the N2 rejection in natural gas processing. The sensitivities of the N2 rejection costs to the N2 permeance and the N2/CH4 selectivity were evaluated and discussed. Based on this preliminary economic study, future membrane improvement strategy towards N2/CH4 separation is also presented.

2. Experimental methods 2.1. Membrane preparation The membranes were synthesized by seeding (secondary growth) onto the inside surfaces of porous alumina supports with pore sizes of 50 nm, 100 nm and 200 nm. These tubes were typically 6-cm long with effective areas between 6.5 and 7.3 cm2. About 1 cm on each end of the alumina supports was glazed to prevent feed bypass and to provide a sealing surface for O-rings. Before synthesis, the supports were boiled in deionized water for three times and each lasted for 40 min. They were dried at 70 1C overnight. 2.1.1. Synthesis of SAPO-34 seeds The seeds were prepared with a molar composition of 1.0 Al2O3: 2.0 P2O5: 0.6 SiO2: 4.0 TEAOH: 150 H2O. In a typical synthesis, aluminum isopropoxide (Al(i-C3H7O)3, Aldrich 99.99%) and deionized water were stirred for 1 h to form a homogeneous solution, followed by addition of phosphoric acid (H3PO4, SigmaAldrich 85 wt% aqueous solution) for another 2 h of stirring. This mixing order was different from previous studies reported by Wu et al. [8] and Funke et al. [19] where these three chemicals were added at the same time. Next, Ludox AS-40 colloidal silica (40 wt.% suspension in water) was added, and the resulting solution was stirred for another 3 h. Tetraethylammonium hydroxide (TEAOH, Aldrich 35 wt% aqueous solution) was then added, and the solution was stirred overnight at room temperature ( 22 1C). The gel was placed in an autoclave and heated in a microwave oven (CEM Mars 5 Microwave Reaction System with XP-1500 plus control vessel) at 180 1C for 7 h. After the solution was cooled to room temperature, it was centrifuged at 3300 rpm for 10 min to collect the seeds, which were followed by deionized water washing. The process was repeated three times, and the resulting precipitate was dried overnight at 100 1C.

S. Li et al. / Journal of Membrane Science 487 (2015) 141–151

2.1.2. Preparation of gel for growth of membranes Several gel compositions containing Al2O3, P2O5, and SiO2 sources, templates and H2O were mixed, stirred and aged for growth of membranes. Al(i-C3H7O)3 and Al(OH)3 were used as Al2O3 sources, whereas H3PO4 and Ludox AS-40 were used as P2O5 and SiO2 sources, respectively. TEAOH was used as pivotal template for the formation of SAPO-34 structure and DPA was employed as secondary structure directing agent. For example, a typical synthesis using Al(i-C3H7O)3 as Al2O3 source and dual templates had a gel composition of 1.0 Al2O3: 1.0 P2O5: 0.3 SiO2: 1.0 TEAOH: 1.6 DPA: 150 H2O. To prepare the gel, Al(i-C3H7O)3 and deionized water were stirred for 1 h to form a homogeneous solution, followed by addition of H3PO4 into the gel for another 2 h of stirring. Then Ludox AS-40 colloidal silica was added, and the resulting solution was stirred for another 3 h. The TEAOH (Aldrich 35 wt% aqueous solution) was then added, followed by addition of DPA (Acros 99%) after 1 h. The gel was stirred at 45–50 1C for 4 days before it was used in membrane growth.

2.1.3. Preparation of SAPO-34 membranes The inside of the supports was seeded by rubbing with SAPO34 seeds (prepared under 2.1.1). The outside of the supports was wrapped with Teflon tape to avoid growth of membranes on the outside surface. After seeding, the supports were placed in an autoclave filled with the synthesis gel (prepared under 2.1.2). Usually two seeded supports were placed in one autoclave, which was then filled with the synthesis gel to about 0.5 cm above the top of the supports. Hydrothermal treatment was carried out in a conventional oven at 220 1C for 6–24 h. After this step, the membranes were washed for 15 min with flowing tap water and dried at 100 1C for  2 h. The membranes were calcined at 400 1C for 4 h with heating and cooling rates of 0.8 1C/min under a weak vacuum (pressure of 300 Pa). Six approaches were employed in the current study with synthesis conditions including pore size of the support, gel composition, alumina source, and crystallization time shown in Table 1. The number of membranes prepared for each approach is also provided.

2.2. Material and membrane characterization Seeds and crystals collected from the bottom of the support tubes during membrane synthesis were used in various analyses including powder X-ray diffraction (PXRD) using a Siemens Kristalloflex 810 diffractometer operating at 30 kV and 25 mA with Cu Kα1 radiation (λ¼ 1.54059 Å). Some SAPO-34 membranes were

143

broken and analyzed by a scanning electron microscope (SEM, JEOL JSM-7000F). Gas adsorption isotherms were measured by a volumetric method using a home-built adsorption system. In these measurements, ultra-high purity N2 (99.999%, Airgas) and CH4 (99.999%, Airgas) were used. SAPO-34 crystals ( 0.20 g) were first outgassed at 200 1C for 2 h. Helium was then introduced to calibrate the volume of adsorption cell containing SAPO-34 crystals. After volume calibration, the system was evacuated, and then N2 or CH4 was introduced to the sample cell. The pressure change was measured using an ASHCROFT K17MO242C13000 model transducer. Based on the pressure change and volume calibrated, the amount of gas adsorbed at that equilibrium pressure was calculated. 2.3. Gas permeation measurement Single-gas and mixture permeations were measured in a flow system (Fig. 2). The membrane was mounted in a stainless steel module and sealed at each end with silicone O-rings. The pressure on each side of the membrane was independently controlled. Fluxes were measured using a bubble flow meter. For mixture separation, a pre-mixed 50/50 N2/CH4 mixture was used as feed. The total feed flow rate was 100 mL/min. The testing temperature was 23 1C, and the feed pressure was 223 kPa and the transmembrane pressure drop was 138 kPa if not indicated otherwise. The compositions of the feed, retentate and permeate streams were measured, after attaining the steady state, using a gas chromatograph (SRI instruments, 8610C) equipped with a thermal conductivity detector and HAYESEP-D packed column. The oven, injector and detector temperatures in the GC were kept at 40 1C, 50 1C and 150 1C, respectively. The permeance of the component i, P i , is defined as Pi ¼

Ji Δpln ;i

ð1Þ

where J i is the flux through the membrane for component i. For the cross-flow configuration, because one component preferentially permeates through the membrane, the partial pressures in the feed and retentate are quite different. Therefore, a Log-mean pressure drop, Δpln ;i , is calculated by Δpln ;i ¼

ðpf ;i  pr;i Þ ln ½ðpf ;i  pp;i Þ=ðpr;i  pp;i Þ

ð2Þ

where pf ;i , pr;i , andpp;i are partial pressures for component i, in feed, retentate, and permeate sides, respectively. The permeability is the permeance multiplied by membrane thickness. The ideal selectivity

Table 1 Membrane preparation conditions. Approach Pore size of the support (nm)

1

100

2

100

3

100

4

50

5

200

6

100

Membrane synthesis

Number of membranes prepared

Gel composition

Al source

Crystallization time (h)

1.0 Al2O3: H2O 1.0 Al2O3: H2O 1.0 Al2O3: H2O 1.0 Al2O3: H2O 1.0 Al2O3: H2O 1.0 Al2O3: H2O

Al(iC3H7O)3 Al(iC3H7O)3 Al(iC3H7O)3 Al(iC3H7O)3 Al(iC3H7O)3 Al(OH)3

24

2

24

2

6

3

6

2

6

2

6

1

1.0 P2O5: 0.3 SiO2: 1.0 TEAOH: 1.6 DPA: 77 1.0 P2O5: 0.3 SiO2: 1.0 TEAOH: 1.6 DPA: 150 1.0 P2O5: 0.3 SiO2: 1.0 TEAOH: 1.6 DPA: 150 1.0 P2O5: 0.3 SiO2: 1.0 TEAOH: 1.6 DPA: 150 1.0 P2O5: 0.3 SiO2: 1.0 TEAOH: 1.6 DPA: 150 1.0 P2O5: 0.3 SiO2: 1.0 TEAOH: 1.6 DPA: 150

144

S. Li et al. / Journal of Membrane Science 487 (2015) 141–151

MFC Pressure transducers 250 PSI

Back pressure regulator

2000 PSI

Data Processor

MFM

To Atmosphere

Retentate

Retentate out Stainless steel module

GC Helium (carrier gas)

membrane 250

PSI

Permeate

Bubble flow meter

MFM

Gas Chromatograph

Permeate out Soap solution

Fig. 2. Gas permeation testing system.

Fig. 3. (a) SEM image of the seeding crystals synthesized by microwave treatment and (b) XRD pattern of the SAPO-34 seeds.

yer Zeolite lay 8.2 µm

Fig. 4. (a) Surface, and (b) cross-sectional SEM micrographs of a membrane prepared by approach 1. Membranes were synthesized at 220 1C for 24 h onto 100-nm-pore Al2O3 supports with a gel composition of 1.0 Al2O3: 1.0 P2O5: 0.3 SiO2: 1.0 TEAOH: 1.6 DPA: 77 H2O. Al(i-C3H7O)3 was used as the Al2O3 source.

is the ratio of the single-gas permeances, and the separation selectivity, αsep i=j , is the ratio of the permeances for mixtures.

3. Results and discussion 3.1. Seeds characterization Fig. 3a shows an SEM image of the seeding crystals prepared by microwave heating. These crystals are composed of nanosheets with lengths and widths between 200 and 800 nm, and thicknesses between 50 and 220 nm. XRD confirms the CHA topology typical of SAPO-34 (Fig. 3b).

3.2. Membrane preparation and characterization Water content in the gel: Approaches 1 and 2 used the same synthesis conditions except that the water mole ratio was 77 in approach 1, whereas in the approach 2 nearly twice the amount of water (mole ratio of 150) was used. The SEM images of the membrane surface and cross section for approaches 1 and 2 are shown in Figs. 4 and 5. The surface SEM images for membranes prepared by both approaches (Figs. 4a and 5a) show well intergrown zeolite crystals. The crystals obtained via the approach 2 (Fig. 5a) look more uniform than those prepared by approach 1 (Fig. 4a). Moreover, as expected, the dilute gel led to a thinner dense zeolite layer (7.3 μm) as shown in the cross-sectional SEM micrographs (Fig. 5b vs. Fig. 4b).

S. Li et al. / Journal of Membrane Science 487 (2015) 141–151

145

Zeolite layer 7.3 µm

Fig. 5. (a) Surface, and (b) cross-sectional SEM micrographs of a membrane prepared by approach 2. Membranes were synthesized at 220 1C for 24 h onto 100-nm-pore Al2O3 supports with a gel composition of 1.0 Al2O3: 1.0 P2O5: 0.3 SiO2: 1.0 TEAOH: 1.6 DPA: 150 H2O. Al(i-C3H7O)3 was used as the Al2O3 source.

Table 2 Comparison of the N2/CH4 separation performance for membranes prepared by approaches 1 and 2. Approach Gel composition

Membrane thickness (μm) Number of membranes prepared N2/CH4 separationa N2 permeance (GPU) Selectivity

1.0 Al2O3: 1.0 P2O5: 0.3 SiO2: 1.0 TEAOH: 1.6 DPA: 77 H2O 8.2 1.0 Al2O3: 1.0 P2O5: 0.3 SiO2: 1.0 TEAOH: 1.6 DPA: 150 H2O 7.3

1 2 a

320 7120 370 752

2 2

2.8 7 1.1 6.0 7 0.49

All the 7 values are standard deviations.

Zeolite layer 6.2 µm

Fig. 6. (a) Surface, and (b) cross-sectional SEM micrographs of a membrane prepared by approach 3. Membranes were synthesized at 220 1C for 6 h onto 100-nm-pore Al2O3 supports with a gel composition of 1.0 Al2O3: 1.0 P2O5: 0.3 SiO2: 1.0 TEAOH: 1.6 DPA: 150 H2O. Al(i-C3H7O)3 was used as the Al2O3 source.

Mixture (50/50 N2/CH4) permeations were measured for the membranes prepared by approaches 1 and 2. Note that the N2 (kinetic diameter: 0.364 nm) and CH4 (kinetic diameter: 0.38 nm) molecules can permeate through both SAPO-34 pores (0.38 nm) and defects larger than 0.38 nm. The SAPO-34 pores are selective, whereas the defects are generally not selective for light gas mixtures [10–12]. Table 2 shows that as the water mole ratio increased from 77 to 150, the N2/CH4 selectivity increased from 2.8 71.1 to 6.0 70.49, indicating the fraction of the defects of the membranes decreased. A diluted gel formed a thinner membrane layer, and thus the average N2 permeance obtained for the approach 2 was higher than that for the approach 1. Crystallization time: The crystallization time used in approaches 1 and 2 was 24 h. In approach 3, the crystallization time was reduced to 6. The shortened crystallization time led to a thinner dense zeolite layer (6.2 μm) as indicated by the cross-sectional SEM image (Fig. 6b). The SEM image of the membrane surface (Fig. 6a) still shows well intergrown zeolite crystals. When the crystallization time was reduced from 24 to 6 h, the average N2 permeance increased by 19% (Table 3). This was mainly because the membrane thickness decreased by 15%. The average N2/CH4 selectivity was also improved from 6 to 7.3 (Table 3).

Pore size of the support: Rubbing was used in the current study to seed the SAPO crystals onto the support. Achieving a highquality seeding layer is important because it affects the property of the resulting membrane. The quality of the seeding layer is usually influenced by the size match between the seeding crystals and the support pores as wells as the roughness of the inside surface of the support. Approaches 3, 4 and 5 used the same preparation conditions except that membranes were synthesized on alumina supports with pore size of 100 nm, 50 nm and 200 nm, respectively. Intergrown zeolite crystals were formed for all these supports (Figs. 6a, 7a and 8a). The membrane thicknesses estimated from SEM analysis of the membrane cross-sections decreased with increasing pore size of the support. As shown in Table 4, among the three types of supports, the seeding crystals have been implanted the best onto the 100-nm-pore support for secondary growth of a membrane as the highest N2/CH4 selectivity was observed for the 100-nm-pore supported membranes (Table 4). Alumina source: Al(OH)3 was used instead of Al(i-C3H7O)3 as the Al2O3 source in approach 6. Other synthesis conditions were the same as approach 3. The SEM images of the membrane surface show poor intergrown zeolite crystals (Fig. 9a). The membrane thicknesses estimated from SEM analysis of the membrane

146

S. Li et al. / Journal of Membrane Science 487 (2015) 141–151

Table 3 Comparison of the N2/CH4 separation performance for membranes prepared by approaches 2 and 3. Approach

Crystallization time (h)

2 3

24 6 a

Membrane thickness (μm)

7.3 6.2

Number of membranes prepared

2 3

N2/CH4 separationa N2 permeance (GPU)

Selectivity

3707 52 4407 74

6.0 7 0.49 7.3 7 0.21

All the 7 values are standard deviations.

Zeolite layer 7.7 µm

Fig. 7. (a) Surface, and (b) cross-sectional SEM micrographs of a membrane prepared by approach 4. Membranes were synthesized at 220 1C for 6 h onto 50-nm-pore Al2O3 supports with a gel composition of 1.0 Al2O3: 1.0 P2O5: 0.3 SiO2: 1.0 TEAOH: 1.6 DPA: 150 H2O. Al(i-C3H7O)3 was used as the Al2O3 source.

Zeolite layer 3.6 µm

Fig. 8. (a) Surface, and (b) cross-sectional SEM micrographs of a membrane prepared by approach 5. Membranes were synthesized at 220 1C for 6 h onto 200-nm-pore Al2O3 supports with a gel composition of 1.0 Al2O3: 1.0 P2O5: 0.3 SiO2: 1.0 TEAOH: 1.6 DPA: 150 H2O. Al(i-C3H7O)3 was used as the Al2O3 source.

Table 4 Comparison of the N2/CH4 separation performance for membranes prepared by approaches 3, 4 and 5. Approach Pore size of the support (nm)

3 4 5

100 50 200 a

Membrane thickness (μm)

6.2 7.7 3.6

Number of membranes prepared

3 2 2

N2/CH4 separationa

N2 Permeance (GPU)

Selectivity

4407 74 580 7 42 6107 25

7.3 7 0.21 2.4 7 0.92 3.2 7 1.2

(OH)3, which led to a more homogenous gel solution for membrane synthesis, and thus better membrane quality. Because the membranes prepared by approach 3 had the highest N2/CH4 separation selectivity and moderate N2 permeance, they were used for further study to investigate the separation mechanisms. 3.3. Adsorption isotherms and heats of adsorption

All the 7 values are standard deviations.

cross-sections (5.6 μm, Fig. 9b) was close to that of the membrane prepared by approach 3. N2 permeance as high as 2600 GPU was obtained for this membrane, but the selectivity was only 2.3 (Table 5), indicating a poor quality of the membrane when using Al(OH)3 as the alumina source. During the synthesis, we found that Al(i-C3H7O)3 was more soluble in aqueous solution than Al

SAPO-34 powder was collected from the bottom of the membranes prepared by approach 3. The powder was calcined and used for adsorption study. Adsorption isotherms for N2 and CH4 at various temperatures are shown in Figs. 10 and 11. The amounts of CH4 adsorbed at high pressures were 2.1–2.9 times the amounts of N2 adsorbed. For light gases on zeolites, Langmuir single-site adsorption can be assumed: θ¼

q bp ¼ qsat 1 þ bp

ð3Þ

S. Li et al. / Journal of Membrane Science 487 (2015) 141–151

147

Zeolite layer 5.6 µm

Fig. 9. (a) Surface, and (b) cross-sectional SEM micrographs of a membrane prepared by approach 6. Membranes were synthesized at 220 1C for 6 h onto 100-nm-pore Al2O3 supports with a gel composition of 1.0 Al2O3: 1.0 P2O5: 0.3 SiO2: 1.0 TEAOH: 1.6 DPA: 150 H2O. Al(OH)3 was used as the Al2O3 source.

-5

Table 5 Comparison of the N2/CH4 separation performance for membranes prepared by approaches 3 and 6.

3

Membrane thickness (μm)

Al(i6.2 C3H7O)3 Al(OH)3 5.6

6 a

Number of membranes prepared

N2/CH4 separationa

N2 Permeance (GPU)

Selectivity

3

4407 74

7.3 70.21

1

2600

2.3

-6

ln (bqsat)

Approach Al source

N2 -7

All the 7 values are standard deviations.

CH4

y = 15.025x -11.494

y = 11.24x -11.072

-8 0.25

0.3

0.35

0.4

0.45

N2 adsorption capacity (mmol/g)

1000/RT

0.16

20°C

where θ is the fractional coverage, q is the coverage, qsat is the saturation coverage, p is the pressure. Eq. (3) can be converted to

0.12

65°C

0.08

1 1 1 1 ¼ U þ q bqsat p qsat

ð4Þ

110°C 0.04

150°C

0 0

20

40 60 80 Pressure (kPa)

100

120

Fig. 10. Adsorption isotherms for N2 on SAPO-34 powder at various temperatures.

CH4 adsorption capacity (mmol/g)

Fig. 12. Adsorption equilibrium constants as a function of inverse temperature for N2 and CH4 adsorption on SAPO-34.

0.5

Through which bqsat can be obtained by plotting 1=q verse 1=p using the adsorption isotherm data. Note that the saturation coverage is the maximum number of molecules that can fit in the zeolite pores, and thus it should be a constant for a certain molecule absorbed on a certain zeolite. The adsorption equilibrium constant, b, can be written in terms of the heat and entropy of adsorption:   ΔS ΔH ads b ¼ exp  ð5Þ R RT

20°C 0.4

and thus, 

0.3

bqsat ¼ exp

65°C

0.2

110°C 150°C

0.1 0 0

20

80 40 60 Pressure (kPa)

100

120

Fig. 11. Adsorption isotherms for CH4 on SAPO-34 powder at various temperatures.

 ΔS ΔH ads  U qsat R RT

ð6Þ

Eq. (6) can be further converted to ln bqsat ¼ 

ΔH ads ΔS þ þ ln qsat R RT

ð7Þ

The heats of adsorption (  ΔH ads ) determined from Arrhenius plots (Fig. 12) of the adsorption equilibrium constants are 11 kJ/ mol for N2 and 15 kJ/mol for CH4, indicating CH4 adsorbs more strongly than N2 on SAPO-34 crystals.

S. Li et al. / Journal of Membrane Science 487 (2015) 141–151

The transport mechanism through zeolite membranes is based on adsorption–diffusion. Depending upon the type of zeolite, the mixture system, and the operating conditions, mixtures are separated by at least one of the following three mechanisms: 1) molecular sieving—larger molecules cannot fit into the pores, and thus the smaller molecules preferentially permeate; 2) differences in diffusivity—the smaller, less hindered type of molecule in a mixture diffuses faster than the larger ones; and 3) competitive adsorption—one type of molecule is more strongly adsorbed on the zeolite and thus can dramatically inhibit permeation of the second molecule. Both N2 (kinetic diameter: 0.364 nm) and CH4 (kinetic diameter: 0.38 nm) can fit into the SAPO-34 pores (0.38 nm), which rules out the separation by molecular sieving. The CH4 adsorbs more strongly than N2 on SAPO-34 crystals based on the heats of adsorption obtained from Fig. 12. Thus, the preferential adsorption of CH4 would favor separating CH4 over N2 in the mixture. On the other hand, the smaller molecule N2 diffuses faster than the larger molecule CH4. For example, Li et al. [6] reported that the diffusivity of the N2 was 24 times of that of the CH4 through a SAPO-34 membrane prepared by the single-template method. The differences in diffusivity would favor separating N2 over CH4 in the mixture. Single-gas and mixture permeations were thus measured to determine the roles of differences in diffusivity and competitive adsorption. Fig. 13 shows single-gas permeances of N2 and CH4 through membrane M-100nm-6h-1 as a function of temperature. The M-100nm-6h-1 was one of the membranes prepared by the approach 3. The N2 single gas permeance decreased with increasing temperature, whereas CH4 permeance was almost constant. The highest N2/CH4 ideal selectivity was 8.4 (observed at 23 1C), and it decreased to 6.2 at as temperature increased to 75 1C. Similar to the single gas behavior, the N2 permeance for a N2/CH4 mixture (50/50) decreased with temperature, whereas the CH4 permeance was essentially independent of temperature for the membrane (Fig. 14). Thus, the N2/CH4 separation selectivity decreased with increasing temperature, but was lower than the ideal selectivity. The highest N2 permeance for this membrane was 480 GPU and its highest N2/CH4 separation selectivity was 7.5. The N2 molecule permeates faster than CH4 molecule in both single-gas and mixture permeation because N2 is smaller (diffuses faster). The higher heat of adsorption for CH4 than N2 causes the preferential adsorption of CH4 in the N2/CH4 mixture. Thus, CH4

inhibits the permeation of N2 in the mixture. That is why the N2/CH4 separation selectivity was lower than the ideal selectivity (Fig. 14). However, the gaps between these two selectivities were within 11%, indicating the preferential adsorption of CH4 in the N2/CH4 mixture does not dominate the separation. Differences in diffusivity played a more critical role than the competitive adsorption in the mixture. Fig. 15 shows permeances and selectivity of a N2/CH4 mixture (50/50) at 23 1C as a function of trans-membrane pressure drop for the SAPO-34 membrane M-100nm-6h-1. The permeate side pressure was kept at 85 kPa. The N2 permeance decreased with pressure drop, whereas the CH4 permeance was essentially independent of pressure drop. As a result, the N2/CH4 separation selectivity decreased with increasing trans-membrane pressure drop. Note that these permeation tests mentioned above were conducted at Colorado School of Mines (Golden, Colorado, USA). The atmospheric pressure at that elevation is approximately 85 kPa. Mixture separation was also carried out at a permeate side pressure of the standard atmosphere (101 kPa). The N2 permeance was 500 GPU and the selectivity was 8 at 23 1C for a transmembrane pressure drop of 138 kPa. The highest feed pressure tested in the current study was 0.64 MPa. The real natural gas feeds can be 4–7 MPa. Testing of the SAPO-34 membranes at higher pressures will be our future focus.

500

N2 permeance

12

400 9

Ideal selectivity

300

6 200

Separation selectivity

100

N2/CH4 selectivity

3.4. Gas permeation and separation mechanism

Permeance (GPU)

148

3

CH4 permeance

0

0 20

30

40

50

60

70

80

Temperature (oC) Fig. 14. Permeances and selectivity of a CO2/CH4 mixture (50/50) as a function of temperature for SAPO-34 membrane M-100nm-6h-1. The feed pressure was 223 kPa and the trans-membrane pressure drop was 138 kPa. The N2/CH4 ideal selectivity is shown for comparison.

10

9 300 6

Ideal selectivity

200 100

3

CH4 permeance

0

0 20

40

60

80

Temperature (oC) Fig. 13. Single gas permeances for N2 and CH4, and N2/CH4 ideal selectivity through SAPO-34 membrane M-100nm-6h-1 as a function of temperature. The feed pressure was 223 kPa and the pressure drop was 138 kPa.

8 400

Selectivity

300

6 4

200

2

CH4 permeance

100 0 0

100

200

300

400

N2/CH4 selectivity

400

Permeance (GPU)

12

N2 permeance

N2 permeance

500

N2/CH4 selectivity

Permeance (GPU)

500

500

0 600

Pressure drop (kPa) Fig. 15. Permeances and selectivity of a N2/CH4 mixture (50/50) at 23 1C as a function of trans-membrane pressure drop for SAPO-34 membrane M-100nm-6h-1. The permeate pressure was kept at 85 kPa.

S. Li et al. / Journal of Membrane Science 487 (2015) 141–151

3.5. Comparison to the membranes reported in the literature Table 6 compares N2/CH4 selectivities and N2 permeances for our membrane with other N2-selective inorganic membranes reported in the literature. Among these materials, our SAPO-34 membrane showed the highest N2 permeance. With the same type of material (SAPO-34), the N2 permeance is 67% higher than that reported by Wu et al. [8] When comparing the two types of SAPO34 membranes, one should consider the fact that they were synthesized by different routes. Wu et al. [8] used single template TEAOH in their synthesis, whereas in our study, dual templates (TEAOH and DPA) were used as structure directing agents. They used Al(OH)3 as the Al source whereas we used Al(i-C3H7O)3. In addition, the gel compositions, especially Si/Al ratios, were different. All these led to a different crystalline size as the thickness of our membrane was 2–3 times of that of their membrane (2–3 μm). Interestingly, the permeance of N2 was much higher for our SAPO34 than Wu et al.'s membrane, even though our membrane was thicker. This may be due to the different framework compositions that affect adsorption–diffusion properties of the SAPO-34 membrane. Maple and Williams [20] performed adsorption studies of N2 and CH4 on SAPO-34 and SAPO-18 crystals. They found that specific retention volumes and enthalpies of adsorption are highly dependent on pore structure and framework composition. Nitrogen permeabilities (permeance  membrane thickness) for our SAPO-34 membranes were calculated using the N2 permeance obtained from the mixture permeation and the thickness measured by SEM. The calculated N2 permeability for our best SAPO-34 membrane was as high as 3100 barrers. It is anticipated that the membrane can be made with a thickness less than 2 μm (as reported by Wu et al. [8]) by further optimizing membrane synthesis conditions. This would lead to a N2 permeance greater than 1500 GPU, making SAPO-34 membranes viable for highly economic N2 rejection in natural gas processing, as discussed below. 3.6. Preliminary economic evaluation of one-stage membrane process for N2 rejection Aspen HYSYS software was used to model the membrane process for N2/CH4 separation. The major economic evaluation bases are: 1) $4 per MSCF gas price to plant, 2) 4% N2 in the residue gas, 3) membrane module costs of $400/m2 in a skid, and 4) membrane lifetime of 15 years. The reported costs for commercial 8-inch diameter polymer membrane modules in a gas separation skid including membrane, steel vessels, flanges, valves and pipes were approximately $500/m2, and the costs can be reduced by using larger diameter modules since total costs of vessels, flanges, valves, and pipes on the system are the major fraction of the final membrane skid cost [1]. Thus, module costs of $400/m2 in a skid are achievable for polymer membranes. The current costs of zeolite membranes are higher than those of polymers membranes mainly because largescale membrane preparation procedure is immature and the

Table 6 Comparison of N2/CH4 separations through inorganic membranes. Material

α

Carbon molecular sieve SSZ-13 SAPO-34 (a)

7.7  0.1 13 66 5– 300 7 8 500

SAPO-34 (b) prepared by approach 3

N2 permeance (GPU)

Thickness (μm)

Ref.

707 15 7.8 2.0

[5] [8] [8]

6.2

This study

149

membranes are typically prepared on single-channel tubes. The packing densities of the single-channel tubes are usually less than 100 m2/m3. Using such low packing density tubes for zeolite membrane module production to achieve a large membrane area for real application makes the material cost and corresponding energy and labor cost high. Ping et al. [21] reported that SAPO-34 membranes can be scaled up by preparing membranes on sevenchannel monolith alumina supports and the monolith-supported SAPO-34 membranes showed similar CO2/CH4 separation performance to SAPO-34 membranes on single-channel tubes. Monoliths are more practical than discs or single-channel tubes because they provide higher packing densities and a low pressure drop. Spiral wound and hollow fiber modules dramatically improved polymeric membranes cost/performance characteristics because of their higher packing density. The multi-channel monolith geometry has the potential to benefit SAPO-34 membranes in the same way. Therefore, module costs of $400/m2 in a skid may be achievable for SAPO-34 membranes when monoliths are used as the supports. In the current preliminary economic evaluation, membrane module costs of $400/m2 in a skid were assumed in calculating the processing costs. Linear extrapolation for processing costs may apply if the real membrane costs are eventually higher than $400/m2. Membrane technology competes most directly against cryogenic distillation. Lokhandwala et al. [22] reported the development of methane-selective membranes with high permeances and methane/nitrogen selectivities of approximately 3–3.5 for treating natural gas containing high concentrations of nitrogen. They concluded that multi-step/multi-stage membrane systems were the lowest cost nitrogen removal technology in many applications despite the design complexity and compression requirements. In contrast, our SAPO-34 membranes are N2-selective. Also, one-stage membrane process is simple, contains no rotating equipment, and requires minimal maintenance. The purpose of the current preliminary economic evaluation was to find out the required N2 permeance and N2/CH4 selectivity for a one-stage membrane process to compete with cryogenic distillation where the costs are $2.40/MSCF for a plant capacity of 1 MMSCF/day and $0.80/MSCF for a plant capacity of 20 MMSCF/day natural gas plant. These costs are for a 15 vol% N2 in the feed. In N2 rejection processing, in addition to achieving 4% N2 in the residue (requirement of the U.S. natural gas pipeline specification), maintaining a low methane loss is also important, especially for large N2-contaminated gas wells. For cryogenic plants, the methane loss is a function of whether it is a one or two column design, nitrogen concentration in the fuel, feed gas pressure and the like. The methane losses are usually defined as actual losses (such as vents) plus losses due to methane needed for engine fuel for compressors. The losses are approximately 6.7% for a twocolumn design at 20 MMSCFD feed gas and 14 vol% N2 in feed. The CH4 losses for one-column design are higher than those of the two-column design, but generally are under 10%. The methane loss in membrane process, in principal, is not affected by N2 permeance, but N2/CH4 selectivity. Fig. 16 shows methane loss to the permeate as a function of N2/CH4 selectivity for different N2 feed concentrations. It decreases with increasing N2/CH4 selectivity and decreasing N2 feed concentration. Note that for the N2 feed concentration of 15 vol%, when N2/CH4 selectivity is greater than 8.5, the CH4 concentration in the permeate side is typically less than 40%. Without further treatment, such stream may not be used as fuel due to the low thermal value. It must be flared, which represents a revenue loss. Fig. 16 shows that to meet a similar CH4 loss to cryogenic distillation (o 10%), one-stage membrane process should have a membrane with the N2/CH4 selectivity greater than 15 for the N2 feed concentration of 15 vol%. For membranes with selectivities

S. Li et al. / Journal of Membrane Science 487 (2015) 141–151

less than 15, process optimization and staging will be required to achieve the CH4 loss below 10%. For example, the permeate gas can be recompressed and passed through a second CH4-selective membrane stage where majority of the CH4 is recovered. In spite of the higher methane loss, membranes with selectivities lower than 15 can still compete with the cryogenic distillation in terms of cost, especially for small nitrogen-contaminated gas wells where cryogenic distillation costs about $2.40/MSCF for a plant capacity of 1 MMSCF/day. Note that many small N2-contaminated gas wells have been shut in for lack of suitable smallscale nitrogen rejection technology [1]. Fig. 17 shows the N2 rejection costs as a function of N2 permeance for membranes with N2/CH4 selectivities of 8 and 15. Note that the costs here include membrane processing cost and the revenue loss due to the methane loss to the permeate. For both selectivities, the costs decrease sharply as the N2 permeance increases from 100 to 500 GPU. As the N2 permeance increases further, the costs decrease more slowly. Therefore, it is critical to develop membranes with N2 permeances approaching or exceeding 500 GPU to reduce the N2 rejection cost. This preliminary economic evaluation indicates the cost for our SAPO-34 membrane (N2 permeance of 500 GPU and selectivity of 8) could be approximately $1.13/MSCF if our membrane technology is used for N2 rejection in the natural gas processing. This cost is less than half of the cryogenic distillation for a plant capacity of 1 MMSCF/day ( $2.40/MSCF).

70% 15 vol.% nitrogen in feed

CH4 loss to permeate

60%

40 vol.% nitrogen in feed

N2 Permeance = 1,000 GPU N2 Permeance = 2,000 GPU

1.2

0.8

0.4

0

0

4

8

12 16 20 N2/CH4 selectivity

24

28

32

Fig. 18. N2 rejection cost as a function of N2/CH4 selectivity for membranes with N2 permeances of 500, 1000, and 2000 GPU. The N2 feed concentration is 15 vol%.

Fig. 18 presents more cost sensitivity studies. When N2 permeance is 500 GPU, an increase in N2/CH4 selectivity from our current 8 to 30 causes a decrease in cost by 45% (from $1.13/MSCF to $0.62/MSCF) for a N2 feed concentration of 15 vol%. Again, further increasing N2 permeance warrants further reduction in N2 rejection cost; the cost is as low as $0.28/MSCF for membranes with selectivity of 30 and N2 permeance of 2000 GPU. In summary, our current SAPO-34 membranes can possess cost advantage over cryogenic distillation for small nitrogencontaminated gas wells. Our future work is to further improve selectivity or/and permeance to improve the competitiveness of membrane technology.

4. Conclusion

40% 30% 20% 10% 4

0

8

20 12 16 N2/CH4 selectivity

24

28

32

Fig. 16. Methane loss to the permeate as a function of N2/CH4 selectivity.

3.2

N2/CH4 = 15 N2/CH4 = 8

2.4

Cost ($/MSCF)

N2 Permeance = 500 GPU

25 vol.% nitrogen in feed

50%

0%

1.6

Cost ($/MSCF)

150

SAPO-34 membranes have potential for N2 rejection in natural gas processing. The membranes separated N2 from CH4 with N2 permeance as high as 500 GPU and separation selectivity of 8 at 24 1C for a 50/50 N2/CH4 mixture. This separation performance is superior to those of the state-of-the-art membranes. The heat of adsorption was higher for CH4 (15 kJ/mol) than N2 (11 kJ/mol), leading to a preferential adsorption of CH4 over N2 in the N2/CH4 mixture. On the other hand, the N2 molecule diffused much faster than the CH4 molecule and differences in diffusivity played a more critical role than the competitive adsorption. As such, the SAPO-34 membranes were selective for N2 over CH4 in the mixture. Preliminary economic evaluation indicates that the required N2/CH4 selectivity is 15 to maintain a CH4 loss below 10% in a one stage membrane process with N2-selective membranes. For small nitrogen-contaminated gas wells, our current SAPO-34 membranes have potential to compete with the benchmark technology cryogenic distillation for N2 rejection. Further improving membrane permeance and selectivity may result in further reduction of the N2 rejection cost.

1.6

Acknowledgments 0.8

0

0

500

1000

1500

2000

N2 permeance (GPU) Fig. 17. N2 rejection cost as a function of N2 permeance for membranes with N2/ CH4 selectivities of 8 and 15. The N2 feed concentration is 15 vol%.

The information, data, or work presented herein was funded in part by the Advanced Research Projects Agency-Energy (ARPA-E), U.S. Department of Energy, under Award number DE-AR0000247. We thank Dennis Leppin at GTI for providing helpful information regarding industrial nitrogen rejection in natural gas processing. We also thank Dr. Mark Pouy from Booz Allen Hamilton, Dr. Dane Boysen, Dr. Jason Rugolo, and Mr. Sven Mumme from ARPA-E for their useful discussions.

S. Li et al. / Journal of Membrane Science 487 (2015) 141–151

References [1] R.W. Baker, K. Lokhandwala, Natural gas processing with membranes: an overview, Ind. Eng. Chem. Res. 47 (2008) 2109. [2] L. Echterhoff and V. K. Pathak, Evaluation of Process Costs for Small Scale Nitrogen Removal from Natural Gas, KBR, GRI-91/0092, 1991. [3] H.S. Meyer and M.S. Henson, Methane selective membranes for nitrogen removal from low quality natural gas—high permeation is not enough, in: Proceedings of the Natural Gas Technologies II: Ingenuity and Innovation, Phoenix, AZ, 2004. [4] L.M. Robeson, The upper bound revisited, J. Membr. Sci. 320 (2008) 390. [5] X. Ning, W.J. Koros, Carbon molecular sieve membranes derived from Matrimids polyimide for nitrogen/methane separation, Carbon 66 (2014) 511. [6] S.G. Li, J.L. Falconer, R.D. Noble, R. Krishna, Modeling permeation of CO2/CH4, CO2/N2, and N2/CH4 mixtures across SAPO-34 membrane with the Maxwell– Stefan equations, Ind. Eng. Chem. Res. 46 (2007) 3904. [7] S.G. Li, J.L. Falconer, R.D. Noble, R. Krishna, Interpreting unary, binary, and ternary mixture permeation across a SAPO-34 membrane with loading dependent Maxwell–Stefan diffusivities, J. Phys. Chem. C 111 (2007) 5075. [8] T. Wu, M.C. Diaz, Y. Zheng, R. Zhou, H.H. Funke, J.L. Falconer, R.D. Noble, Influence of propane on CO2/CH4 and N2/CH4 separations in CHA zeolite membranes, J. Membr. Sci. 473 (2015) 201. [9] R. Szostak, Molecular Sieves—Principles of Synthesis and Identification, Van Nostrand Reinhold, New York, NY, 1989. [10] S.G. Li, J.L. Falconer, R.D. Noble, SAPO-34 membranes for CO2/CH4 separation, J. Membr. Sci. 241 (2004) 121. [11] S.G. Li, G. Alvarado, R.D. Noble, J.L. Falconer, Effects of impurities on CO2/CH4 separations through SAPO-34 membranes, J. Membr. Sci. 251 (2005) 59.

151

[12] S.G. Li, J.G. Martinek, J.L. Falconer, R.D. Noble, T.Q. Gardner, High-pressure CO2/ CH4 separation using SAPO-34 membranes, Ind. Eng. Chem. Res. 44 (2005) 3220. [13] S.G. Li, J.L. Falconer, R.D. Noble, Improved SAPO-34 membranes for CO2/CH4 separations, Adv. Mater. 18 (2006) 2601. [14] M. Hong, S.G. Li, H.F. Funke, J.L. Falconer, R.D. Noble, Ion-exchanged SAPO-34 membranes for light gas separations, Microporous Mesoporous Mater. 106 (2007) 140. [15] M.A. Carreon, S. Li, J.L. Falconer, R.D. Noble, SAPO-34 seeds and membranes prepared using multiple templates, Adv. Mater. 20 (2008) 729. [16] S.G. Li, J.L. Falconer, R.D. Noble, SAPO-34 membranes for CO2/CH4 separations: effect of Si/Al ratio, Microporous Mesoporous Mater. 110 (2008) 310. [17] S.G. Li, M.A. Carreon, Y.F. Zhang, H.H. Funke, R.D. Noble, J.L. Falconer, Scale-up of SAPO-34 membranes for CO2/CH4 separation, J. Membr. Sci. 352 (2010) 7–13. [18] R. Zhou, E.W. Ping, H.H. Funke, J.L. Falconer, R.D. Noble, Improving SAPO-34 membrane synthesis, J. Membr. Sci. 444 (2013) 384–393. [19] H.H. Funke, M.Z. Chen, A.N. Prakash, J.L. Falconer, R.D. Noble, Separating molecules by size in SAPO-34 membranes, J. Membr. Sci. 456 (2014) 185–191. [20] M.J. Maple, C.D. Williams, Separating nitrogen/methane on zeolite-like molecular sieves, Microporous Mesoporous Mater. 111 (2008) 627. [21] E. Ping, R. Zhou, H.H. Funke, J.L. Falconer, R.D. Noble, Seeded-gel synthesis of SAPO-34 single channel and monolith membranes, for CO2/CH4 separations, J. Membr. Sci. 415–416 (2012) 770–775. [22] K.A. Lokhandwala, I. Pinnau, Z. He, K.D. Amo, A.R. DaCosta, J.G. Wijmans, R.W. Baker, Membrane separation of nitrogen from natural gas: a case study from membrane synthesis to commercial deployment, J. Membr. Sci. 346 (2010) 270–279.