Effect of polymer precursors on carbon molecular sieve structure and separation performance properties

CARBON

4 8 ( 2 0 1 0 ) 4 4 3 2 –4 4 4 1

available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/carbon

Effect of polymer precursors on carbon molecular sieve structure and separation performance properties Mayumi Kiyono a, Paul J. Williams b, William J. Koros a b

a,*

Department of Chemical & Biomolecular Engineering, Georgia Institute of Technology, Atlanta, GA 30332-0001, USA Shell Projects and Technology, 3333 Highway 6 South, Houston, TX 77082, USA

A R T I C L E I N F O

A B S T R A C T

Article history:

Polymer precursor and processing method have a significant effect on the separation per-

Received 29 April 2010

formance of carbon molecular sieve (CMS) membranes. The authors previously developed

Accepted 2 August 2010

a polymer processing method involving oxygen exposure during pyrolysis using synthe-

Available online 6 August 2010

sized polyimide, 6FDA/BPDA-DAM. The objectives of this work were (i) to demonstrate the generality of the oxygen doping method with a commercially available polymer Matrimid, (ii) to investigate resultant CMS membrane structures, and (iii) to engineer the CMS performance observed with Matrimid precursor by tuning the pyrolysis temperature. The investigation of the pore structures is challenging due to their amorphous structures. Various researchers investigated using traditional characterization methods, such as XRD and adsorption, yet molecular sieving structure in ultramicropore region is still not known. Here, the authors investigated using gas molecules as a probe. By interpolating the characterization results, hypothetical ultramicropore size distributions for each CMS membranes are suggested. The results are used to explain dramatically different separation performance trends observed between 6FDA/BPDA-DAM and Matrimid CMS membranes and to adapt the doping method on Matrimid CMS membrane for better performance.  2010 Elsevier Ltd. All rights reserved.

1.

Introduction

Carbon molecular sieves (CMS) have been investigated for various applications, such as pressure swing adsorption (PSA) systems [1] and membrane separation systems [2,3]. In particular, CMS membranes are known to have attractive separation performance, well above the ‘‘upper bound curve’’ for soluble polymeric materials [2,4], for challenging gas separations, such as O2/N2, CO2/CH4, C3H6/C3H8 [5,6]. Previously, a polymer processing method which exposes trace oxygen to polymeric membranes during pyrolysis was developed to tune the CMS structure as well as gas separation performance using 6FDA/BPDA-DAM [7]. This method potentially allows one to enhance productivity (permeability) by more than

100 times while doubling efficiency (selectivity) compared with polymeric membranes. For the case considered, the synthesized polymer, 6FDA/BPDA-DAM contains halocarbons, which may be both environmentally and economically unfavored [8]. This study seeks to demonstrate an ability to successfully adapt the inert pyrolysis method with other polyimide materials for attractive separation performance in economically with non-halogen-containing precursors. Due to relatively lower material cost and ease of access, a commercially available polymer, Matrimid, was chosen as a model precursor. Polymeric CMS membranes are amorphous and hence difficult to investigate with regard to morphologies, especially in ultramicropore regions. Here, we investigate ultramicropore

* Corresponding author: Fax: +1 404 385 2683. E-mail address: [email protected] (W. J. Koros). 0008-6223/$ - see front matter  2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.carbon.2010.08.002

CARBON

4 8 ( 20 1 0 ) 4 4 3 2–44 4 1

structures by using various gas molecules as probes to help reveal the structural differences between 6FDA/BPDA-DAM and Matrimid CMS membranes. The results were further used to improve the efficacy of the inert pyrolysis method on separation performance of Matrimid CMS membranes.

2.

Background and theory

2.1.

Transport in carbon molecular sieve membranes

The sorption–diffusion mechanism is used to explain gas transport through carbon molecular sieve (CMS) membranes. Gas molecules sorb on the membrane at the upstream, diffuse into the membrane under the influence of a chemical potential gradient, and desorb from the membrane at the downstream. There are two parameters commonly used to measure separation performance of membranes. Permeability is used to measure a membrane material’s intrinsic productivity and equals the pressure and thickness normalized flux, as described below: Pi ¼

ni  l : Dpi

ð1Þ

In Eq. (1), ni represents the flux of gas molecule component ‘‘i’’ through the membrane of the membrane thickness, l, and Dpi is the transmembrane partial pressure difference that acts as the driving force for component i across the membrane. The most common unit for permeability is the Barrer which is defined as: 1 Barrer½¼1010

cm3 ðSTPÞ  cm : cm2  s  cm Hg

ð2Þ

An additional popular unit is kmol m m2 s1 kPa1, which can be converted to Barrers by multiplication by 2.99 · 1015. The second parameter is selectivity. Selectivity is a measure of a membrane’s separation efficiency and equals the ratio of the component permeabilities for the case of a negligible downstream permeate pressure. Permeability can also be described in terms of the governing kinetic and thermodynamic parameters, namely the diffusion coefficient, Di, and the sorption coefficient, Si, by the following equation: Pi ¼ Di  Si :

ð3Þ

The diffusion coefficient, Di, in CMS membranes is a strong function of penetrate size, and ideally enables molecular sieving discrimination between similarly sized penetrants. The sorption coefficient equals the concentration of gas sorbed divided by the penetrant partial pressure at equilibrium and depends on the condensability of the gas penetrant and its interactions with the membrane material. For carbon molecular sieves with rigid saturatable capacities, a Langmuir isotherm is commonly used, so the sorption coefficient can be expressed as: Si ¼

Ci C0Hi bi ¼ : pi 1 þ bi pi

ð4Þ

where Ci is the equilibrium uptake of penetrant i by the sorbent, pi is the partial pressure, C0Hi is the Langmuir hole filling capacity and bi is the Langmuir affinity constant for component i.

2.2.

4433

Structure of carbon molecular sieve membranes

Carbon molecular sieve (CMS) membranes are formed by thermal decomposition of polymer precursors. As a result of the thermal treatment, CMS membranes consist of an almost pure carbon material [9]. Such carbon materials have a highly aromatic structure comprised of disordered sp2 hybridized carbon sheets with angles of disorientation, called a turbostratic structure. The structure can be envisioned to comprise roughly parallel layers of condensed hexagonal rings with no three-dimensional crystalline order. Pores are formed from packing imperfections between microcrystalline regions in the material [10,11]. The pore structure in CMS membranes is described as ‘‘slit-like’’ with an ‘‘idealized’’ pore structure. Furthermore, CMS membranes have a bimodal pore distribu˚ connected by tion, which consists of large pores of 6–20 A smaller pores known as ‘‘ultramicropores’’ [12]. Such a combination of ultramicropores and micropores is believed to provide the combined molecular sieving function and high permeability characteristic of these unusual materials. The disordered structure of the carbon materials is different from zeolites, which have a uniform, well defined set of pores. With the distribution of the ultramicropores, CMS materials offer the important advantage of facile formation of defect free membranes for use in gas separation applications. Structures of CMS membranes have been investigated by many researchers using traditional techniques, such as X-ray diffraction (XRD), transmission electron microscopy (TEM), and adsorption. Unfortunately, due to the amorphous nature of CMS, it has been difficult to determine the structure, especially the ultramicropore region that governs the molecular sieving process. When XRD was performed, broad peaks were observed due to its amorphous structure; therefore, the precision with which one can deconvolute information to estimate actual pore structure was poor [5,13,14]. The results of high resolution TEM were also inconclusive due to the amorphous structure of CMS, and the image hardly showed any pores of the material [12,15]. Chen et al. conducted an argon adsorption isotherm to obtain a pore size distribution of CMS membranes, but the argon molecule was too large to analyze the selective ultramicropore region [16]. Campo and Mendes concluded that the pore size distribution derived from CO2 adsorption equilibrium may not be enough to explain performance [13]. In the current study, various gas molecules were used as probes to interpolate ultramicropore distribution.

2.3. Tuning transport properties of carbon molecular sieve membranes Many variables influence resulting separation performance and structure of CMS membranes, including pyrolysis atmosphere [7,10,15]. Previously, the authors developed a tuning method which utilizes oxidation during the pyrolysis process. The method utilizes the fact that oxygen may chemisorb and form a stable bond at the selective ultramicropore windows at high temperatures. The bond is strong and stable and does not break unless it experiences the same heating treatment responsible for its original formation [7]. Therefore, this doping process makes the sieving structure less permeable and

4434

CARBON

4 8 ( 2 0 1 0 ) 4 4 3 2 –4 4 4 1

more selective when the amount of doping is increased. Synthesized polymer, 6FDA/BPDA-DAM, was used to test this method. Results in Fig. 1 support the hypothesis that permeability decreases with increase in selectivity up to a point when excessive doping makes the sieve lose both permeability and selectivity. These data show a specific example, but this study focuses on exploring the generality of the tuning method for a very different commercially available polymer, Matrimid.

3.

Experimental

3.1.

Materials

The polymer precursor Matrimid 5218 was provided by Vantico, Inc. and 6FDA/BPDA-DAM was synthesized in-house. The chemical structure and the characteristics of the polymer are shown in Table 1. The synthesis of 6FDA/BPDA-DAM was conducted via a polycondensation reaction by addition of the dianhydride, 5,5 0 -[2,2,2-trifluoro-1-(trifluoromethyl)ethylidene]bis-1,3-isobenzofurandione (6FDA), and diamines, 2,4,6-trimethyl-1, 3-phenylene diamine (DAM) and 3,3 0 ,4,4 0 -biphenyl tetracarboxylic dianhydride (BPDA), in solution using the solvent, n-methylpyrrolidone (NMP). All monomers were purified by sublimation prior to use. This sublimation was essential for the end product to have high molecular weight and lower polydispersity index (PDI). In this study, the reaction stoichiometry was adjusted to have the ratio of BPDA to DAM of 1:1. The polycondensation is sensitive to water and minimal

Polymer precursor CMS (Vacuum pyrolysis) CMS (Inert pyrolysis) Robeson curve 08 30 ppm O2/Ar

100

8 ppm O2/Ar

2

50 ppm O /Ar 2

Polymeric dense films were prepared by first drying the polymer powder in a vacuum oven at 110 C for >12 h to remove moisture. Immediately after removal from the oven, 3– 5 wt% of polymer solution was prepared by dissolving it in dichloromethane (Sigma–Aldrich) and was placed on a rolling apparatus for >12 h. A solution casting method was used to prepare polymer dense films in a glove bag to achieve a slow solvent evaporation rate. After solvent was evaporated, (in usually 3–4 days), films were removed from the casting plate and placed in the vacuum oven at 110 C for >12 h to remove residual solvent. Once the films were removed from the oven, they were cut into small discs with a diameter of 2.54 cm. All films had a thickness of approximately 60 ± 10 lm.

3.4. 1 100

1000

104

Formation of polymer precursor films

Pyrolysis

The polymer films were pyrolyzed on a corrugated quartz plate at a temperature of 550 C. They were heated at a rate of 13.3 C/min from a room temperature to 250, 3.85 C/min from 250 to 535 C, 0.25 C/min from 535 to 550 C, and kept at 550 C for 2 h. For vacuum pyrolysis, a pump (Edwards) was used to create a low pressure, and a liquid nitrogen trap was used to prevent any back diffusion of oil vapor. For inert pyrolysis, inert gas was purged with a desired flow rate for at least 12 h before the experiments to ensure the system was at steady state. Between runs, the quartz tube and plate were rinsed with acetone (Aldrich) and burned in the air at 800 C to remove any deposited materials which could affect consecutive runs.

4 ppm O2/Ar

4

CO /CH Selectivity

3.2.

3.3.

1000

10

exposure to moisture was ensured. The reaction produced polyamic acid which is the precursor to the polyimide. Thermal imidization was chosen to dehydrate the polyamic acid to form a cyclic polyimide. The reaction solution was heated to a high temperature of 180 C in order for the imidization to occur [17]. Gel permeation chromatography (GPC) showed the polyimide product with molecular weight of 103,000 and PDI of 2.2 as a result of the polymer synthesis. Thermal gravimetric analysis (TGA) was conducted to characterize decomposition temperature (Tdecom). The TGA combined with Fourier transform infrared spectroscopy (TGA-FTIR, provided from Netzsch, STA 409 PC Luxx TGA/DSC) was performed to characterize the decomposition process of polymers. Using these polymers, polymeric dense films were prepared and pyrolyzed to form carbon molecular sieve membranes.

105

CO Permeability (Barrer) 2

Fig. 1 – Separation performance of 6FDA/BPDA-DAM CMS dense films [7]. A circular data point represents polymeric properties and square points represent properties of CMS membranes. A line with an arrow represents a trend between separation performance and oxygen concentration during pyrolysis. The Robeson curve shows the trade-off from currently available polymer [2].

Permeation

Once CMS membranes were prepared, they were immediately loaded into permeation cells. The permeability was measured using a constant volume–variable pressure method [18,19]. Both upstream and downstream of the permeation system were evacuated for at least 12 h before the measurement, and a leak rate was confirmed to be always less than 1% of the permeate rate of the slowest gas. Once the whole system was evacuated, the upstream was pressurized with a test gas while the downstream was maintained at vacuum. The pressure rise in a standard volume on the downstream was

CARBON

4435

4 8 ( 20 1 0 ) 4 4 3 2–44 4 1

Table 1 – Chemical structures and characteristics of polyimide discussed in this study. Polymer

Chemical structure

Tdecom

6FDA/BPDA-DAM

450 C CF3

O

CF3

O

N

O

CH 3 N

O CH3

O

O

N H 3C

CH3 N

O CH3

O

H3 C Y

X

2,4,6-trimethyl-1,3-phenylene diamine (DAM), 3,3 0 ,4,4 0 -biphenyl tetracarboxylic dianhydride (BPDA), and 5,5 0 -[2,2,2-trifluoro-1-(trifluoromethyl)ethylidene] bis-1,3-isobenzofurandione (6FDA). The ration of X to Y is 1:1. BTDA-DAPI (Matrimid)

425 C H3 C

CH3 O

H3 C

O

O N

N O

O n

monitored with time by LabView (National Instruments) and permeability was calculated using Eq. (1). The system was again evacuated each time before experiments with different gases for at least 12 h.

3.5.

Sorption

Gas sorption measurements were made using a pressure decay method [20,21], where the equilibrium sorbed concentration at a given pressure can be used to calculate the solubility coefficient. Once the CMS films were loaded, the system was evacuated for 24 h. A feed reservoir was pressurized with a certain amount of gas and allowing the system to equilibrate thermally. The entire system was kept in a heated water bath with a circulator to maintain a constant temperature. Once the feed reservoir came to equilibrium, the pressure valve between the feed and the sample cell was opened and then quickly closed to introduce a dose of the feed gas into the cell. The pressure in both chambers was monitored with pressure transducers. The amount of gas sorbed was then calculated using a mole balance.

4.

Results and discussions

4.1. Oxygen membranes

doping

method

on

Matrimid

CMS

CMS dense films were prepared by pyrolyzing Matrimid polymeric membranes under (i) vacuum (0.005 torr) and (ii) argon inert with specific amount of oxygen (3, 10, 30, 50, and 100 ppm O2/Ar, provided by AirGas). After pyrolysis, each CMS membrane was characterized with FTIR. CMS membranes prepared under vacuum pyrolysis showed no significant peaks, which is consistent with literature [11]. The spectra in Fig. 2 shows C@O group appearing in the vicinity of 1700 cm1 on CMS samples prepared by inert

Fig. 2 – FTIR spectra of Matrimid CMS membranes.

4436

CARBON

4 8 ( 2 0 1 0 ) 4 4 3 2 –4 4 4 1

1000 Polymer precursor CMS (Vacuum pyrolysis) CMS (Inert pyrolysis, O concentration in Ar purge is listed) 2

100 3 ppm 10 ppm

4

CO /CH Selectivity

Robeson curve 08

2

30 ppm 50 ppm 100 ppm

10

1 1

10

100

3

10

4

10

5

10

CO Permeability (Barrer) 2

Fig. 3 – Separation performance of Matrimid CMS dense films. Tests were conducted at 35 C with an upstream pressure of 50 psia. A circular data point represents polymeric properties and square points represent properties of CMS membranes. A dash arrow line represents a trend between separation performance and oxygen concentration during pyrolysis.

pyrolysis. This indicates that the oxygen ‘‘doping’’ method was successfully applied to produce CMS membranes. The result of CO2/CH4 separation performance is shown in Fig. 3. It demonstrates that the attractive separation performance, which is above the upper bound curve, can be achieved with pyrolysis under (i) vacuum and (ii) inert pyrolysis when the pyrolysis atmosphere is properly controlled: this

Fig. 5 – TGA-FTIR result of Matrimid. Y-axis in arbitrary intensity units.

result supports the observation with 6FDA/BPDA-DAM that oxygen exposure has a correlation with separation performance. The result in Fig. 3 shows that the oxygen doping method caused decreases in both permeability and selectivity. This trend is different from the trend the authors observed with the 6FDA/BPDA-DAM CMS membranes (shown in Fig. 1): we speculate that the discrepancy is caused by the difference in their intrinsic CMS structures. We hypothesize that Matrimid CMS membranes, created from an intrinsically lower free volume precursor, has more closed and less selective intrinsic structure than 6FDA/BPDA-DAM CMS membranes. Therefore, oxygen doping not only reduces pore sizes of ultramicropore and micropores but also further closes selective pores, leading to a decrease in both permeability and selectivity. In order to test the hypothesis, several characterizations were conducted to investigate the structure of each CMS membrane.

4.2. TGA-FTIR on 6FDA/BPDA-DAM and Matrimid CMS membranes

Fig. 4 – TGA-FTIR result of 6FDA/BPDA-DAM [11]. Y-axis in arbitrary intensity units.

The decomposition process of each polymer was investigated using TGA-FTIR. Polymeric samples were heated under argon purge, and the evolved gases were sent to FTIR chamber to analyze the chemical composition. Figs. 4 and 5 show decomposition results of 6FDA/BPDA-DAM and Matrimid. In both cases, majority of by-products evolved after an onset of their decomposition. Polymer 6FDA/BPDA-DAM produced HF (4250– 4500 cm1), CH4 (3017 cm1), CHF3 (1150, 1178 cm1), CO2 (2110 cm1), and CO (2190 cm1). Matrimid produced mainly CH4 (3017 cm1), CO2 (2110 cm1) and CO (2190 cm1). There are three stages of pyrolysis when polymers are heated up to 1200 C: (i) the precarbonation, (ii) the carbonization, and (iii) dehydrogenation [22]. A temperature range of

4437

4 8 ( 20 1 0 ) 4 4 3 2–44 4 1

100

80

3

60

3

100 C to polymer decomposition temperatures, 450 C for 6FDA/BPDA-DAM and 425 C for Matrimid, would be considered as the precarbonation stage. During this first stage, molecules, such as excess monomer and solvent are removed. Polymeric films turn black and linear conjugated C–C systems start to form near the decomposition temperature. In the carbonization stage, rapid weight loss is observed due to the removal of entities, such as oxygen, nitrogen, and CF3. As Figs. 4 and 5 show, major by-products evolve on this regime. An exact temperature range of this stage is difficult to define. However, it is speculated to be between the decomposition temperatures and the temperature where the rate of weight loss is significantly reduced (an elbow of decomposition weight loss curves). Evolution of CO, CO2, and CH4 were also observed for Matrimid by Barsema et al. [23]. At the end of this stage, a loose network of linear conjugated systems is formed [22]. In the dehydrogenation stage, hydrogen is gradually eliminated, typically in between the temperature 500 and 1200 C. Jenkins and Kawamura note that the rate of the removal is a characteristic of a given heat-treatment temperature [22]. Elemental analysis on carbon membranes pyrolyzed in similar manners show 95–99% of aromatic carbon content [6,15,24,25]; moreover, the percentage of carbon element is dependent of pyrolysis temperature [22]. As the large fluorinated compound are produced and diffuse out of membrane films more open ultramicropore structures are believed to be formed with 6FDA/BPDA-DAM than with Matrimid, which lacks these fluorinated moieties.

C (cm (STP)/cm CMS)

CARBON

40

20

0 0

100 150 200 Pressure (psia)

250

300

Fig. 6 – Sorption isotherms for 6FDA/BPDA-DAM CMS membranes prepared with 1 ppm O2/Ar inert pyrolysis. The isotherms were obtained with six different gases: s, CO2; d, CH4; j, O2; h, N2; D, He; m, SF6. The experiments were repeated and fitted with Langmuir isotherm model described in dot lines. Experiments were repeated and had less than 5% of deviation.

4.3. Sorption on 6FDA/BPDA-DAM and Matrimid CMS membranes

80 70 60 50 40

3

3

C (cm (STP)/cm CMS)

Gas sorption was examined to characterize sorption coefficients, Langmuir hole filling capacity and affinity constant. Each polymeric film was pyrolyzed under a flow of 200 cc/ min 1 ppm O2/Ar to produce CMS membranes. The lowest concentration of oxygen in argon mixture available was chosen to prepare CMS membranes with structures closest to their ‘‘intrinsic structures.’’ The sorption coefficients depend on the micropore sizes of carbon material, the critical temperature of the molecular which measures condensability, and affinity of penetrants to the material. If the ultramicropore size is small enough to effectively exclude one gas molecule while passing through the separating gas molecule, true molecular sieving occurs. In this case, no uptake of the sieved component would be shown on the sorption isotherm. For carbon material, however, a distribution of ultramicropore exists. Therefore, only regions connected with true molecular sieving pore windows would be inaccessible to a larger gas molecule. Figs. 6 and 7 show sorption isotherms measured with six gases, He, CO2, O2, N2, CH4, and SF6. The sorption isotherms were fitted to the Langmuir equation model in Eq. (4). The hole filling capacities, C0H , and affinity constants, b, are shown in Tables 2 and 3, along with the kinetic diameters of gases. In all gas measurements, 6FDA/BPDA-DAM CMS membranes showed higher sorption coefficients and hole filling capacities than Matrimid CMS membranes. This indicates a larger available micropore volume in 6FDA/BPDA-DAM CMS membranes than for Matrimid CMS membranes. The ratio of

50

30 20 10 0

0

50

100 150 200 Pressure (psia)

250

300

Fig. 7 – Sorption isotherms for Matrimid CMS membranes prepared with 1 ppm O2/Ar inert pyrolysis. The isotherms were repeated and obtained with six different gases: s, CO2; d, CH4; j, O2; h, N2; D, He; m, SF6. The experiments were repeated and fitted with Langmuir isotherm model described in dot lines. Experiments were repeated and had less than 5% of deviation.

ðC0H ÞCO2 values for 6FDA/BPDA-DAM vs. Matrimid CMS membranes’ is about 1.3. This ratio is similar to the reported value

4438

CARBON

4 8 ( 2 0 1 0 ) 4 4 3 2 –4 4 4 1

Table 2 – Langmuir hole filling capacity C0H and Langmuir affinity constant b calculated based on 6FDA/BPDA-DAM CMS sorption isotherms. Gas He CO2 O2 N2 CH4 SF6

Kinetic ˚ ) [29] diameter (A

C0H (cc(STP)/ ccCMS)

b (psia1)

13.0 98.0 72.2 69.4 78.0 12.7

0.005 0.06 0.006 0.01 0.025 0.002

2.6 3.3 3.46 3.64 3.8 5.5

Table 3 – Langmuir hole filling capacity C0H and Langmuir affinity constant b calculated based on Matrimid CMS sorption isotherms. Gas He CO2 O2 N2 CH4 SF6

Kinetic ˚ ) [29] diameter (A

C0H (cc(STP)/ ccCMS)

b (psia1)

11.7 74.3 68.5 27.1 57.9 11.6

0.003 0.02 0.006 0.01 0.005 0.0005

2.6 3.3 3.46 3.64 3.8 5.5

of 1.4 for the ratio of micropore volume for the 6FDA/BPDADAM vs. Matrimid CMS samples by DFT analysis in Steel and Koros [26], which is consistent with our data. The trend of isotherms for both sets of CMS membranes is similar. The similarity can also be seen with the order of magnitudes on Langmuir constants in both CMS membranes. This is likely caused by the fact that CMS samples were treated in the same manner and resulted in similar overall gross structure. As previously mentioned, pyrolysis treatment above 550 C results in more than 95% carbon in the final structure [15,25], and the percentage presumably depends upon the intensity of heat treatment [22]. Despite its small molecular size, the sorption coefficients for He were significantly lower than other gases. This is attributed to its non-condensable nature. On the other hand, the lowest sorption coefficients observed for both CMS materials was found for the highly condensable gas, SF6. It may be caused by the small popula˚ . In tion of ultramicropores in the range between 5 and 6 A fact, bimodal pore size distributions seen by various researchers suggest that the minimum for typical CMS materials lies around the size of an SF6 molecule [13,27,28]. This explains the order of Langmuir affinity constants in both CMS membranes with being SF6 the lowest, because access of a large spherical molecule SF6 to the regular microvoids is simply

too low to enable accurate measurement of the affinity constants for typical regular micropores. A comparison of R2 values for the affinity constants on SF6 vs. other gases shows poor fitting of the model: 0.6 and 0.95 respectively, providing further evidence of this observation.

4.4. Structural difference between 6FDA/BPDA-DAM and Matrimid CMS membranes Pore size distributions in ultramicropore region were investigated in this study using various gas molecules as probes. Diffusion coefficients were obtained from permeation and sorption experiments using Eq. (3). Similar to sorption isotherm experiments, CMS membranes were prepared under inert pyrolysis of 200 cc/min 1 ppm O2/Ar to produce CMS membranes with close to ‘‘intrinsic’’ structures. Once samples were prepared, the experiments were conducted at 35 C with a pressure of 50 psia using the test gases listed in Tables 2 and ˚ ), O2 (3.46 A ˚ ), N2 (3.64 A ˚ ), CH4 ˚ ), CO2 (3.3 A 3, namely He (2.6 A ˚ ˚ (3.80 A), and SF6 (5.5 A). Table 4 lists permeability coefficients, Table 5 lists sorption coefficients, and Table 6 shows diffusion coefficients for 6FDA/BPDA-DAM and Matrimid CMS membranes. Permeabilities of 6FDA/BPDA-DAM CMS membranes are higher than for Matrimid CMS membranes; possibly due to higher free volume of the polymer precursor and resultant larger C0H as a result of evolution of CF3 group during the heating process as seen on TGA-FTIR. Table 5 confirms that transport in CMS membranes is not sorption dominant with regard to selectivity. For CO2/CH4 separation, sorption selectivity is less than half of diffusion selectivity. In the case of O2/N2 separation on both 6FDA/BPDA-DAM and Matrimid CMS membranes, the sorption selectivity is in the range between 0.9 and 2, which is similar to results observed by Singh and Koros [30]. The results of diffusion coefficients suggest that 6FDA/ BPDA-DAM CMS membranes have a larger average number of accessible ultramicropore windows than Matrimid CMS membranes. For all tested gases, diffusion coefficients are much higher for 6FDA/BPDA-DAM CMS membranes than that of Matrimid CMS membranes. For instance, DHe is more than double and DCO2 is more than six times higher than Matrimid CMS membranes. In addition, diffusion selectivity indicates that 6FDA/BPDA-DAM CMS membranes have more selective pore structures. For CO2/CH4 separation, the diffusion selectivity is 17 on 6FDA/BPDA-DAM CMS membranes while it is 11 for Matrimid CMS membranes. By interpreting results of diffusion coefficients, effective semi-quantitative pore size distributions for the ultramicropore region were constructed and are shown in Fig. 8. The distributions were drawn to match the ratio of diffusion coefficients relative to the area of accessible ultramicropores for each respective molecule

Table 4 – Permeability of 6FDA/BPDA-DAM and Matrimid CMS membranes pyrolyzed under 1 ppm O2/Ar inert gas. Tests were conducted at 35 C with a pressure of 50 psia. Units are in Barrer with % deviation of less than 10%. Polymer precursor 6FDA/BPDA-DAM Matrimid

He

CO2

O2

N2

CH4

SF6

2530 605

7170 1049

1530 301

204 63

247 17

0.6 0.13

CARBON

4439

4 8 ( 20 1 0 ) 4 4 3 2–44 4 1

Table 5 – Sorption coefficients of 6FDA/BPDA-DAM and Matrimid CMS membranes in ccSTP/(ccCMS-psia). Experiments were repeated and had less than 10% of deviation. Polymer precursor 6FDA/BPDA-DAM Matrimid

He

CO2

O2

N2

CH4

SF6

0.049 0.031

1.41 1.28

0.43 0.37

0.46 0.18

0.82 0.23

0.018 0.006

Table 6 – Diffusion coefficients of 6FDA/BPDA-DAM and Matrimid CMS membranes in 108 cm2/s. Experiments were repeated and had less than 10% of deviation. Polymer precursor 6FDA/BPDA-DAM Matrimid

He

CO2

O2

N2

CH4

SF6

2680 1020

262 42.4

237 42.1

22.8 18.0

15.5 3.85

1.72 1.14

Fig. 8 – Diffusion coefficient based ultramicropore size distribution. The dot line represents the distribution of 6FDA/BPDADAM CMS membranes and the solid line represents the distribution of Matrimid CMS membranes. The X-axis is linearly scaled with an indication of molecule’s kinetic diameters.

for the gas separations among CO2, O2, N2, CH4 gases. In addition, the total area under the curve was adjusted to be about 2.6 times larger for 6FDA/BPDA-DAM CMS membranes than that of Matrimid CMS membranes to reflect the relative diffusion coefficients of He in the two polymers. This is based on an assumption that He samples all pores accessible to any gas molecule in both CMS membranes from the two precursors. Details are described in Appendix A. As seen from the diffusion coefficient data in Table 6 and the diffusion coefficient based ultramicropore size distribution in Fig. 8, 6FDA/BPDA-DAM CMS membranes have a larger number of large pores, but also a more selective pore structure than Matrimid CMS membranes.

4.5. Engineering performance of CMS membranes with pyrolysis temperature Previous sections have shown that Matrimid CMS membranes have closed and less selective intrinsic pore structures

that result in a decrease in both permeability and selectivity with an increase in oxygen exposure. In order to utilize the oxygen doping method on the Matrimid precursor, experiments were designed to create CMS membranes with more open intrinsic pore structures. Researchers have shown that higher temperatures tend to produce more selective yet less permeable CMS membranes [12,31,32] presumably due to systematic relaxation of the CMS matrix. This suggests that higher pyrolysis temperatures result in more selective and less permeable CMS structure. In principle, lowering the pyrolysis temperature should lead to more open intrinsic CMS structure so that one can take advantage of the oxygen doping method. In order to test this hypothesis, a slightly lower pyrolysis temperature of 500 C with 1 ppm oxygen in argon gas was chosen to demonstrate the effect of temperature and the doping process. Results are shown in Fig. 9. As predicted, the lower pyrolysis temperature produces more permeable, but less selective CMS membranes with ‘‘nearintrinsic’’ structures, when CMS membranes are exposed to

4440

CARBON

4 8 ( 2 0 1 0 ) 4 4 3 2 –4 4 4 1

4

10

Table 7 – Diffusion selectivity among challenging gas separations based on Table 6. This was referenced to construct ultramicropore size distribution curves.

Polymer o

500 C CMS o

550 C CMS Robeson 08

Polymer precursor

He/N2

CO2/CH4

O2/N2

6FDA/BPDA-DAM Matrimid

117 56.7

16.9 11.0

10.3 2.33

30 ppm O /Ar 2

100

4

CO /CH Selectivity

1000

2

1 ppm O /Ar 10

100 ppm O /Ar

3 ppm O2/Ar

2

2

1 1

10

3

100 10 CO Permeability (Barrer)

4

10

5

10

2

Fig. 9 – Separation performance of Matrimid CMS membranes pyrolyzed by different temperatures with oxygen ‘‘doping’’ process.

appear to have ‘‘intrinsically’’ fewer and smaller pores; therefore, in simple terms, the oxygen doping method closes previously productive and selective ultramicropores but leaves less selective ones open. In addition to the semi-quantitative pore size distributions, experiments were conducted to create more open intrinsic pore structures, by applying lower pyrolysis temperatures, to take an advantage of the doping process on Matrimid CMS membranes. The resulting separation performance shows that a combination of pyrolysis temperature and oxygen exposure can provide tools to control the structure and the separation performance of CMS membranes. By knowing the intrinsic structures, created in low oxygen exposure pyrolysis conditions at a given temperature, one can predict the trends in separation performance with the doping method.

low amounts of oxygen. In addition, exposure to higher oxygen concentration enables use of the ‘‘doping’’ method, since the selectivity enhancement was more than double from 1 ppm O2/Ar to 30 ppm O2/Ar.

Acknowledgement

5.

Appendix A

Conclusions

A method developed previously on 6FDA/BPDA-DAM was adapted to prepare attractive performance of Matrimid CMS membranes. FTIR results showed C@O bond formed as amount of chemisorbed oxygen was increased during pyrolysis. Moreover, the separation performance of CMS membrane indicates that the method was successfully adapted, but with a different overall trend. Specifically, for Matrimid precursor, the trend observed on the separation performance was different from what previously observed for the 6FDA/BPDA-DAM precursor, giving rise to selectivity and permeability decrease as oxygen doping increased for the Matrimid case. Investigation of ultramicropore structure suggested that 6FDA/BPDADAM CMS membranes have more open and selective pore structures that Matrimid CMS membranes. This was investigated by a series of characterizations. TGA-FTIR revealed large amount of CF3 compounds evolved during pyrolysis on 6FDA/BPDA-DAM material. This seems to create a pore structure with more and larger ultramicropores than Matrimid. Sorption isotherms showed 6FDA/BPDA-DAM CMS membranes have larger sorption capacities, which indicate larger volume of sorptive micropore. Study of permeation, sorption, and diffusion coefficients with various gas molecules demonstrated that 6FDA/BPDA-DAM CMS membrane have larger and more selective ultramicropore structures. The diffusion coefficient based ultramicropore size distributions were for ‘‘intrinsic’’ CMS structures were constructed. The distribution is used to explain the difference on separation trend observed by the oxygen doping method. Matrimid CMS membranes

The authors acknowledge the financial support of Shell Global Solutions (US) Inc. (Houston, TX).

As noted earlier, diffusion coefficient based ultramicropore size distributions in the ultramicropore region for the two CMS membrane materials were drawn in Fig. 8 based on the diffusion coefficients in Table 6. The diffusion coefficients were calculated with permeability values and sorption coefficients using Eq. (3). First, the overall shape of the curve was built with an assumption that the number of ultramicropores that are accessible to the SF6 molecule provides a useful metric of the minimal number of large size ultramicropores to which a value of unity was assigned. Then the area which represents number of additional ultramicropores accessible for the rest of the gas molecules was scaled to be proportional to the diffusion coefficients. Finally, distribution curves were drawn to satisfy the ratios of diffusion coefficients among challenging separation gas pairs listed in Table 7. The shape ˚ ) and CO2 of the distribution in Fig. 8 between He (2.6 A ˚ ) was drawn for convenience with uncertainty and is (3.3 A not important for this discussion. Clarification of the shape in this range requires additional work with probes between ˚. 2.6 and 3.3 A R E F E R E N C E S

[1] Ruthven DM, Farooq S, Knaebel KS. Pressure swing adsorption. 1st ed. New York: VCH Publishers, Inc.; 1994. [2] Robeson LM. The upper bound revisited. J Membr Sci 2008;320(1–2):390–400.

CARBON

4 8 ( 20 1 0 ) 4 4 3 2–44 4 1

[3] Ismail AF, David LIB. A review on the latest development of carbon membranes for gas separation. J Membr Sci 2001;193(1):1–18. [4] Robeson LM. Correlation of separation factor versus permeability for polymeric membranes. J Membr Sci 1991;62(2):165–85. [5] Steel KM, Koros WJ. Investigation of porosity of carbon materials and related effects on gas separation properties. Carbon 2003;41(2):253–66. [6] Vu DQ, Koros WJ, Miller SJ. High pressure CO2/CH4 separation using carbon molecular sieve hollow fiber membranes. Ind Eng Chem Res 2002;41(3):367–80. [7] Kiyono M, Williams PJ, Koros WJ. Effect of pyrolysis atmosphere on separation performance of carbon molecular sieve membranes. J Membr Sci 2010;359(1–2):2–10. [8] Wuebbles DJ. Weighing functions for ozone depletion and greenhouse gas effects on climate. Annu Rev Energy Environ 1995;20:45–70. [9] Pierson HO. Handbook of carbon, graphite, diamond, and fullerenes. New York: Noyes Publication; 1993. [10] Geiszler VC, Koros WJ. Effects of polyimide pyrolysis conditions on carbon molecular sieve membrane properties. Ind Eng Chem Res 1996;35(9):2999–3003. [11] Williams PJ. Analysis of factors influencing the performance of CMS membrane for gas separation. Ph.D. thesis, Georgia Institute of Technology, Atlanta, GA, USA; 2006. [12] Steel KM. Carbon membranes for challenging gas separations. Ph.D. thesis, The University of Texas at Austin, Austin, TX, USA; 2000. [13] Campo MC, Magalhaes FD, Mendes A. Comparative study between a CMS membrane and a CMS adsorbent: Part I. Morphology, adsorption equilibrium and kinetics. J Membr Sci 2010;346(1):15–25. [14] Park HB, Kim YK, Lee JM, Lee SY, Lee YM. Relationship between chemical structure of aromatic polyimides and gas permeation properties of their carbon molecular sieve membranes. J Membr Sci 2004;229(1–2):117–27. [15] Suda H, Haraya K. Gas permeation through micropores of carbon molecular sieve membranes derived from Kapton polyimide. J Phys Chem B 1997;101(20):3988–94. [16] Chen J, Loo LS, Wang K, Do DD. The structural characterization of a CMS membrane using Ar sorption and permeation. J Membr Sci 2009;335(1–2):1–4. [17] Omole IC. Crosslinked polyimide hollow fiber membranes for aggressive natural gas feed streams. Ph.D. thesis, Georgia Institute of Technology, Atlanta, GA, USA; 2008. [18] Pye DG, Hoehn HH, Panar M. Measurement of gas permeability of polymers: I. Permeabilities in constant

[19]

[20]

[21]

[22]

[23]

[24]

[25]

[26]

[27]

[28]

[29]

[30]

[31]

[32]

4441

volume/variable pressure apparatus. J Appl Polym Sci 1976;20(7):1921–31. Pye DG, Hoehn HH, Panar M. Measurement of gas permeability of polymers: II. Apparatus for the determination of permeabilities of mixed gases and vapors. J Appl Polym Sci 1976;20(2):287–301. Costello LM, Koros WJ. Temperature dependence of gas sorption and transport properties in polymers: measurement and applications. Ind Eng Chem Res 1992;31(12):2708–14. Koros WJ, Paul DR. Design considerations for measurements of gas sorption in polymers by pressure decay. J Polym Sci Polym Phys Ed 1976;14(10):1903–7. Jenkins GM, Kawamura K. Polymeric carbons – carbon fiber, glass and char. London: Cambridge University Press; 1976. Barsema JN, Klijnstra SD, Balster JH, van der Vegt NFA, Koops GH, Wessling M. Intermediate polymer to carbon gas separation membranes based on Matrimid PI. J Membr Sci 2004;328(1–2):93–102. Jones CW, Koros WJ. Carbon molecular sieve gas separation membranes: I. Preparation and characterization based on polyimide precursors. Carbon 1994;32(8):1419–25. Singh A. Membrane materials with enhanced selectivity: an entropic interpretation. Ph.D. thesis, The University of Texas at Austin, Austin, TX, USA; 1997. Steel KM, Koros WJ. An investigation of the effects of pyrolysis parameters on gas separation properties of carbon materials. Carbon 2005;43:1843–56. Lee H, Suda H, Haraya K. Characterization of the postoxidized carbon membranes derived from poly(2,4-dimethyl1,4-phenylene oxide) and their gas permeation properties. Sep Purif Technol 2008;59(2):190–6. Stoeckli F, Centeno TA. On the characterization of microporous carbons by immersion calorimetry alone. Carbon 1997;35(8):1097–100. Mulder M. Basic principles of membrane technology. 2nd ed. Dordrecht, Netherlands: Kluwer Academic Publishers; 1997. Singh A, Koros WJ. Significance of entropic selectivity for advanced gas separation membranes. Ind Eng Chem Res 1996;35(4):1231–4. Barsema JN, van der Vegt NFA, Koops GH, Wessling M. Carbon molecular sieve membranes prepared from porous fiber precursor. J Membr Sci 2002;205(1–2):239–46. Vu DQ. Formation and characterization of asymmetric carbon molecular sieve and mixed matrix membranes for natural gas purification. Ph.D. thesis, University of Texas at Austin, Austin, TX, USA; 2001.