J Porous Mater (2008) 15:625–633 DOI 10.1007/s10934-007-9142-2
Characterization of highly selective microporous carbon hollow fiber membranes prepared from a commercial co-polyimide precursor E. P. Favvas Æ E. P. Kouvelos Æ G. E. Romanos Æ G. I. Pilatos Æ A. Ch. Mitropoulos Æ N. K. Kanellopoulos
Published online: 20 July 2007 Springer Science+Business Media, LLC 2007
Abstract In this work, the preparation of gas separating carbon hollow fiber membranes based on a 3,3¢4,4¢benzophenone tetracarboxylic dianhydride and 80% methylphenylene-diamine + 20% methylene diamine co-polyimide precursor (BTDA-TDI/MDI, R84 Lenzing GmbH), their permselectivity properties as well as details of the carbon nanostructure are reported. Hollow fibers were initially prepared by the dry/wet phase inversion process in a spinning set-up, while the spinning dope consisted of P84 as polymer and NMP as solvent. The developed polymer hollow fibers were further carbonized in nitrogen at temperatures up to 1173 K. Thermogravimetric analysis was used to investigate the weight loss during the carbonization process. The nitrogen, methane and carbon dioxide adsorption capacity of the prepared materials was determined gravimetrically at 273 and 298 K and hydrogen adsorption experiments were performed at 77 K up to 1 bar. Scanning electron microscopy was used to elucidate the morphological characteristics and the nanostructure while H2 sorption at 77 K was applied to evaluate the microporosity of the developed carbon hollow fiber membranes. In all cases, the permeability (Barrer) of He, H2, CH4, CO2, O2 and N2 were measured at atmospheric pressure and temperatures 313, 333 and 373 K and were found higher than those of the precursor. Moreover, the calculated permselectivity values were significantly E. P. Favvas (&) E. P. Kouvelos G. E. Romanos G. I. Pilatos N. K. Kanellopoulos Institute of Physical Chemistry, NCSR ‘‘Demokritos’’, 153 10 Agia Paraskevi, Attikis, Greece e-mail:
[email protected] A. Ch. Mitropoulos Department of Petroleum and Natural Gas Technology, Cavala Institute of Technology, 654 04 St. Lucas, Cavala, Greece
improved. The developed carbon fibers exhibit rather low H2 permeance values (8.2 GPU or 2.74 · 10–9 mol/m2sPa) with a highest H2/CH4 selectivity coefficient of 843 at 373 K. Keywords Co-polyimide precursor Carbon hollow fibers Permselectivity Kinetic selectivity
1 Introduction Gas separation via distillation is a highly energy consuming process. Pressure swing adsorption (PSA) although very efficient, is one of the most energy consuming and costly alternative approaches for gas separation. For these reasons gas separation technology is currently focused on membrane processes. Furthermore, much attention has been paid lately to the development of carbon molecular sieve membranes (CMSM) [1, 2] due to their very promising performance characteristics. The separation layer of these membranes consists of a carbon structure accommodating nanopores with size equal to one to three molecular diameters of the gases to be separated. In general, carbon molecular sieves exhibit higher thermal and chemical stability than polymer membranes [3] and moreover it has been shown that their separation performance is superior as compared to this of their polymeric precursor systems. This has led to an extensive research in the field of CMSMs, with an emphasis on the precursor material selection [4, 5] and the pyrolysis conditions [6]. Amongst the polymer precursors applied for the preparation of carbon membranes the most frequently used are polyfurfuryl alcohol (PFA) [7], polyvinylidene chloride (PVDC) [8, 9], cellulose [10, 11], phenolic resins [12],
123
626
polyacrylonitrile (PAN) [13, 14], polyetherimides [15, 16] and polyimides [17–20]. Polyimides are categorized among the most stable classes of polymers. They can be used at temperatures higher than 573 K, and usually decompose before reaching their melting point. They are considered to be excellent precursors for glassy carbon [21] because they do not go through a melting phase transition and thus do not lose their shape. Common commercial polyimides used for the preparation of CMS membranes are Matrimid, Kapton and P84 polyimides. Fuertes et al. pyrolyzed Kapton and Matrimid polyimides at 723–973 K [4] and reported permselectivities of CO2/CH4 16 and 33, CO2/N2 9 and 15, O2/N2 4 and 5 for CMSM obtained from Kapton and Matrimid precursors respectively. CMS membranes from Matrimid precursors are, in general, less permeable but more selective. Tin et al. reported that the gas permeability of carbon membranes obtained from P84 polyimide [22], increases dramatically about 2–3 orders after carbonization as compared to their polymer precursor. For instance high CO2/CH4 ideal selectivity of 89 is obtained for such a membrane carbonized at 1073 K in nitrogen environment. The present work describes a method to produce polymeric hollow fiber membranes using the dry/wet phase inversion spinning method and thereinafter their carbon molecular sieve analogues with enhanced gas selectivity properties. The challenge was to develop a P84 co-polyimide hollow fiber as membrane precursor and apply a single pyrolysis protocol for the preparation CMS membranes with enhanced separation performance. The characterization and the evaluation of performance of the final products were carried out by means of gas adsorption (CO2, CH4, H2, N2), and single-phase permeability (CO2, CH4, H2, He, N2, O2) measurements.
2 Experimental
J Porous Mater (2008) 15:625–633
Fig. 1 Chemical structure of P84 co-polyimide
393 K under vacuum prior to be used for the production of the hollow fibers. The solvent applied was N-methylpyrolidone (NMP). The P84 polymeric hollow fiber precursors were prepared by the dry/wet phase inversion process in a spinning set-up. The spinning dope, consisting only of P84 polymer and NMP as solvent, was mixed overnight at 323 K in a stainless-steel vessel of 3 liters in order to prepare a homogeneous solution. In a second step the solution was filtered through a 15 lm metal filter to remove impurities existing in the raw polymers. Both vessels as well as the spinneret were thermostated at 323 K in order to facilitate the flow of the polymer solution. After filtering, the dopes were allowed to degas inside a second stainless steel vessel for 2 days. The bore liquid was a degassed mixture of NMP and deionized water. The polymer solution (28.5% w/w P84/NMP) and bore fluid (70% w/v NMP/H2O) were simultaneously pumped through a tube-in-orifice spinneret using gear pumps. The i.d. of the spinneret was 500 lm and the o.d. 700 lm. The extruded fibers passed first through a 6 cm air gap before entering to the coagulation bath, which was filled with tap water at room temperature. The nascent fibers were oriented by means of two guiding wheels and pulled by a third wheel into a collecting reservoir. In order to remove residual NMP, the produced fibers were washed with tap water overnight and then solvent exchanged in plastic containers with ethanol for 6 h. This experimental set-up was applied to produce three batches of fibers. The post treatment (pyrolysis process) and the techniques involved for the characterization of the developed materials are following described.
2.1 Materials 2.2 Thermogravimetric analysis The precursor was an asymmetric hollow fiber prepared from commercial R84 co-polyimide. P84 (BTDA-TDI/ MDI) was obtained from Lenzing and is a thermally stable co-polyimide of 3,3¢4,4¢-benzophenone tetracarboxylic dianhydride and 80% methylphenylene-diamine + 20% methylene diamine. Its chemical structure is shown in Fig. 1, while the calculated density is about 1.30 g/cm3. The glass transition temperature (Tg) of this polymer is 588 K as determined by differential scanning calorimetry (heating rate of 10 K/min) under dry nitrogen environment. The polymer was dried overnight at
123
The thermal degradation of P84 precursor was monitored by Thermogravimetric analysis (TGA) in dry N2 environment. The instrumentation consists of three digital mass flow controllers (DMFC––Bronkhorst B.V. Holland), a CI microbalance MK2M5 (CI electronics LTD) equipped with an aluminium enclosure head upgraded for pressures up to 10 bar, a high temperature (up to 1373 K) furnace (Thermawatt S.A.) with Eurotherm PID Temperature controller and heat resistant SS sample chamber with quartz rods and pans.
J Porous Mater (2008) 15:625–633
627
2.3 Pyrolysis procedure
2.6 Single gas permeation
Carbon hollow fibers were prepared using a simple inert environment pyrolysis protocol, at a final temperature of 1173 K. A pre-weighed batch of polyimide hollow fibers was placed in the quartz glass tube of a Carbolite CTF 12/ 75 High Temperature Tube Furnace, equipped with a Eurotherm 2408CP temperature controller. The quartz tube was fed with 150 cm3/min pure N2, while the flow-rate was kept constant throughout the pyrolysis process. The temperature was increased from 298 to 373 K with a ramp of 2 K/min. The samples were left at 373 K for 30 min in order to remove absorbed water molecules. Subsequently, the stabilized fibers were re-weighted in order to evaluate the initial mass loss. Then, the following pyrolysis protocol was applied: heating rate of 5 K/min up to 1173 K, stabilization at the maximum temperature for 5 min and then controlled quenching with a cooling rate of 10 K/min down to 298 K. The carbon fibers were then removed from the furnace and immediately weighted in order to estimate the final mass loss.
Scanning electron microscopy (SEM) was used to determine the asymmetric structure and the dimensions of the precursor and carbon hollow fibers. The instrument used was a JEOL JSM 6300 scanning electron microscope.
Permeation measurements of various gases (He, H2, CH4, CO2, O2 and N2) were performed using the variable pressure method in a high-pressure (70 bar) stainless steel permeation rig [23]. In Fig 2 the design of the cell used for the permeation experiments is illustrated. Three fibers, each one 3 cm long, were inserted into a 3/8† stainless steel holder. The open ends of the fibers were sealed using a low vapor pressure epoxy resin (Torr seal, Varian). The effective permeation area for the hollow fiber membrane modules was approximately 1 cm2. Due to the temperature limited performance of the epoxy resin (up to 393 K), just before the sealing procedure, all membranes were heat treated at 523 K for at least 24 h under helium atmosphere, in order to eliminate any trace of adsorbed molecules (H2O, CO2, HC). The sealing and placing of the membranes into the permeation cell was subsequently performed in a specially designed glove box under inert (He) atmosphere and finally, the cell bearing the activated membranes was mounted into the permeation rig. Gas was admitted to the high-pressure section of the rig, while the low-pressure side remained isolated under vacuum. Permeance experiments were performed by continuously monitoring the pressure increase in the low-pressure side of the rig by means of an accurate differential pressure transducer. The permeance, Pe, value (GPU) and Permeability, K, (Barrer) were determined by using the following equations:
2.5 Hydrogen adsorption at 77 K
Pe ¼ 6 104
The hydrogen adsorption isotherm at 77 K of the carbon microporous membrane was acquired using a commercial volumetric porosimeter (Quantachrome Autosorb-1, with Krypton upgrade). Before measurement, the sample was degassed overnight at 573 K under high vacuum (~10–7 mbar).
where, Vlow is the collection volume (cm3), dPlow/dt is the pressure increase rate (mbar/min) in the collection volume, U the total active area of membrane sample (cm2), DP
2.4 Scanning electron microscopy
Vlow ðdPlow =dtÞ ðGPUÞ; DP U Texp Vlow ðdPlow =dtÞ ‘ K ¼ 6 108 ðBarrerÞ; DP U Texp
Fig. 2 Home-made metal cell and metal rig used for the permeation experiments
123
628
(mbar) the pressure head, Texp (K) the experimental temperature and ‘ (cm) the active layer thickness. 2.7 Gravimetric measurements Carbon dioxide, nitrogen and methane high-pressure adsorption-desorption isotherms were produced gravimetrically in a pressure range 0–20 bar, at 273 and 298 K, by means of an Intelligent Gravimetric Analyser (IGA, Hiden Isochema Ltd). The samples were outgassed at 523 K overnight, under ultrahigh vacuum (10–7 mbar). The experimental results have been corrected for buoyancy effects taking into account the compressibility of gases. The mass and density of hang-down systems (Au chains), sample and counterbalance holders (porous stainless steel and glass, respectively), and counterbalance (stainless steel rods) have been pre-determined, while the skeletal density of the samples was measured by helium pycnometry. No buoyancy corrections for the adsorbed phase were carried out; thus, the results refer to excess isotherms. High purity carbon dioxide (99.998%), Nitrogen (99.9995%) and methane (99.95%) were used for the experiments.
J Porous Mater (2008) 15:625–633
the phase inversion between polyimide polymer solution and coagulation liquid. The separating layer is at the outer surface of the fibers for both polymeric and carbon membranes. As observed in Fig. 4, after carbonization process the asymmetric morphology of the membrane was preserved and the microporous separating layer was at the outer surface of carbon membrane. 3.2 TGA analysis The thermal degradation of P84 precursor was monitored by TGA in dry N2 environment. Figure 5 illustrates the weight variation of polymer during the heating process up to 1173 K. According to the TGA thermo-diagram, the degradation temperature (Td) is 798 K and was defined as the temperature corresponding to 5% weight loss. This indicates the high thermal resistance of P84 co-polyimide. The total weight loss at 1173 K, with 5 K/min heating rate, was approximately 34%. The polymer began to decompose after 623 K (Tg of P84 polyimide is 588 K) and this decomposition was accompanied by a total mass loss of 34.1%. The enhanced thermal stability of the P84 polyimide highlights its quality as a precursor for the preparation of carbon molecular sieve membranes.
3 Results and discussion 3.3 Sorption measurements 3.1 Morphology and structure of the membrane The morphological characteristics of the asymmetric polymeric P84 co-polyimide hollow fiber precursor and carbon molecular sieve hollow fiber membrane were obtained from SEM analysis and are presented in Fig. 3 and 4 respectively. The respective magnifications were 120 for Fig. 4c, 1000 for Fig. 3 and Fig. 4a and 2000 for Fig. 4b and Fig. 4d. The macro-voids shown at Fig. 3 were generated during the wet spinning process as the result of
Fig. 3 SEM micrograph of the precursor cross-section
123
The following two assumptions were made to interpret the results of adsorption experiments: a) the dense polymer is structurally homogeneous all over the hollow fiber, (regardless the existence of macropores) and b) the dense polymer part of the precursor is homogeneously pyrolyzed to produce the same micropore structure all over the carbon membrane. Based on the above and after considering that the macropores contribution to the gas uptake should be negligible, the adsorption results must be characteristic of the micropore layer. Adsorption experiments of N2 at 77 K were initially conducted in order to define the BET surface area and the microporosity of the carbon hollow fibers. However the pertinent measurement has revealed negligible uptake of N2 at 77 K. Since carbon molecular sieves have pores and constrictions with dimensions comparable to the size of N2 molecules [24–26] it was a challenge for us to investigate if the cause of this negligible uptake was either the very slow diffusion [27, 28] of nitrogen molecules into the micropores of the CMSM membrane or if during pyrolysis the carbon porous structure has completely collapsed giving rise to the development of a completely dense carbon material. For this reason H2 adsorption-desorption measurements (77 K) [29] were carried out. The small size of H2 and the fact that the 77 K is by far above its supercritical temperature ensures that even the smallest pores, possibly not accessible to other
J Porous Mater (2008) 15:625–633
629
Fig. 4 (a) and (b): Inert layer, (c) and (d): cross layer structure of the carbon membrane
molecules, would easily be monitored, while equilibration kinetics are expected to be fast enough for reliable measurements. Although, Argon at 77 K or CO2 at 173 K could be alternative approaches for the evaluation of the BET surface they were not actually applied since Ar physical properties are not far removed from those of Nitrogen and also in the case of CO2, the high quadrupole moment renders its adsorption isotherm very sensitive to the presence of polar groups in the surface of the solid. The results of H2 adsorption measurements (77 K) revealed significant uptake as shown in Fig. 6 and confirmed the ultra-micropore structure of our carbon membrane. Since H2 is in supercritical state at 77 K, the calculation of the specific surface area (s.s.a.) via the BET interpretation is inconsistent. However we have calculated the Langmuir ˚ 2, as s.s.a., by using a molecular surface area of 12.4 A proposed in literature [30]. This calculation led to an s.s.a.
value of around 560 m2/g, which is in full accordance with the BET and Langmuir calculations carried out at the high pressure CO2 isotherms at 273 and 298 K (see below). Nitrogen, methane and carbon dioxide adsorption–– desorption isotherms of the carbon membrane sample at higher temperatures, where diffusion is fast enough to ensure adsorption equilibrium, are shown in Fig. 7. The shape of the isotherms for the three gases in CMS membrane is of type I, typical for microporous solids [31]. In Fig. 7 it is shown that the carbon dioxide adsorption capacity (4.2 and 4.5 mmol/g for 298 and 273 K respectively at 20 bar) is greater than the uptake of nitrogen and methane. In addition the adsorption and desorption branch of nitrogen and carbon dioxide isotherms coincide, whereas the methane isotherms exhibit hysteresis in the desorption branch especially in the low-pressure region. This is characteristic of hindrance effects leading to adsorption points that are
Fig. 5 Thermogravimetric analysis of P84 precursor
Fig. 6 Low pressure hydrogen adsorption/desorption at 77 K
123
630
Fig. 7 Carbon dioxide, methane and nitrogen adsorption/desorption uptakes
steady state rather than equilibrium ones. The hindrance effects especially for the case of CH4 that is a rather spherical molecule with kinetic diameter of 0.36 nm as compared to the linear CO2 molecule with kinetic diameter of 0.33 nm, lead to remarkable permselectivity values for the pair CO2/CH4 (PeCO2/PeCH4 = 52) and will be further discussed in the following sect. (3.4). Although methane and nitrogen are at supercritical conditions their isotherms adsorption curves approach a plateau at 20 bars providing an evidence for the existence of microporosity. The above
Fig. 8 Kinetic selectivities of: (a) CO2/CH4 at 273 K, (b) CO2/ N2 at 273 K, (c) CO2/CH4 at 298, (d) CO2/N2 at 298 K
123
J Porous Mater (2008) 15:625–633
conclusion is also supported from the low-pressure hydrogen adsorption curve at 77 K (Fig. 6). From the highpressure CO2 isotherms at 273 and 298 K we have calculated that the specific surface areas (BET and Langmuir) between 525 and 625 m2/g, by using a molecular surface ˚ 2 [30]. These values are in accordance with area of 22.2 A the H2 calculated s.s.a. From the point of the kinetic selectivity, which is the most important factor in PSA applications, the obtained values for the pairs CO2/N2, CO2/CH4 at 273 K and 298 K and for pressure steps up to 250 mbar, are presented in the following Fig. 8. The maximum selectivity of the pair CO2/CH4 during the first isotherm step (0–50 mbar) was obtained at 4.3 and 4.7 min for 273 and 298 K respectively, while the corresponding time for the pair CO2/N2 was 6.3 and 8.5 min. The diffusion coefficients (D=4‘2 ) obtained by the familiar solution of the transient sorption curve for slabs: 1 mt 8X 1 ¼1 2 exp½Dð2n þ 1Þ2 p2 t=4‘2 ; p n¼0 ð2n þ 1Þ2 m
where, mt/m, the uptake versus time divided by the maximum uptake during the pressure step, D(cm2/sec) the diffusion coefficient, t(sec), ‘(cm) the slab dimension, are presented in the following Table 1.
J Porous Mater (2008) 15:625–633
631
Permeance values expressed in GPU rather than permeability coefficients (Barrer) are adopted due to the fact that the thickness of the asymmetric structure, which totally controls the gas transport, could be just approximately defined by SEM images. However as it is evidenced in recent literature concerning the fabrication of similar materials [32, 33] the thickness of the polymeric skin layer is approximately 0.5 lm, while the thickness of the developed microporous carbon layer is about 2.5 lm. By adopting these permeability values in Barrer units are also included for both the precursor and carbon membrane in Tables 2 and 3. Under the experimental conditions applied, permeance increases with temperature for all the gases examined, a fact revealing that micropore or else ‘‘activated diffusion’’ is the prevailing mechanism of gas transport. Such an activated process implies that molecules have to overcome energy barriers in order to enter, diffuse and exit from the pore system (due to the small size of the pores and/or the existence of constrictions) and is common in all microporous solids. In this respect the observation of activated diffusion confirms that our membranes are indeed defect free and moreover reveals that the pore sizes are comparable to the molecular sizes highlighting the sieving
Table 1 Diffusion coefficient for pressure steps up to 1 bar 2
–4
Pressure steps (mbar)
D/4‘ (1/sec) · 10 CH4 273 K
CH4 298 K
100–150
5.20
4.5
8.8
11
–
9.8
150–200
4.40
3
9.4
10
8
8.8
200–250
3.6
3.2
8.9
10.4
7.2
9.2
250–500
5.2
6.1
9.4
10
8.3
9
500–750
4.1
5.5
8.4
10
7.9
8.6
750–1000
3.9
5.3
8.7
7.7
9
CO2 273 K
CO2 298 K
9.4
N2 273 K
N2 298 K
3.4 Single gas permeation Tables 2 and 3 present the permeation properties of He, H2, CH4, CO2, N2, and O2 for the carbon membrane and polymeric precursor respectively while Tables 4 and 5 illustrate the ideal selectivities (permselectivities, i.e. calculated as the ratio of permeances) of the same materials for six binary gas mixtures. The gas permeance values of carbon membrane are quite similar with those referred in literature [17].
Table 2 Gas permeance/permeability of carbon molecular sieve hollow fiber membrane, Activation Energies of all gases and heats of sorption for CH4, CO2 and N2 Kinetic diameter (nm)
Gas Temperature (K) 313
0.26
333
Permeance (GPU/10–2)
Permeability (Barrer/10–2)
Permeance (GPU/10–2)
He
290
1015
353
460
1610
599
0.289
H2
0.33
CO2
0.346
Eact (kJ/mol)
373 Permeability (Barrer/10–2)
qst (kJ/mol)
Permeance (GPU/10–2)
Permeability (Barrer/10–2)
883
461
1153
7.4
—
1498
818
2045
9.2
—
49.9
124.8
9.2
29.4
29.8
21.3
53.2
15.9
—
1
2.5
2
5
18.3
18.2
0.85
2.1
0.97
2.4
4.9
23.1
27.6
96.6
40.2
O2
7.9
27.7
11.9
0.364
N2
0.645
2.26
0.38
CH4
0.71
2.49
100
Table 3 Gas permeance/permeability of precursor hollow fiber membrane Gas
Temperature (K) 313 Permeance (GPU)
333 Permeability (Barrer)
Permeance (GPU)
373 Permeability (Barrer)
Permeance (GPU)
Permeability (Barrer)
He
25.4
12.7
33.4
16.7
48.5
24.25
H2
25.3
12.6
33.2
16.6
47.7
23.85
CH4
7.1
3.55
6.7
3.35
6.9
3.45
CO2
5.4
2.7
6.1
3.05
7.2
3.6
O2
5.6
2.8
5.5
2.75
6.1
3.05
N2
6.2
3.1
5.3
2.65
5.5
2.75
123
632
J Porous Mater (2008) 15:625–633
Table 4 Permselectivities coefficients for six binary gas mixtures for carbon membrane Gases/Temperature
H2/CH4
PermSelectivities 313 (K)
333 (K)
373 (K)
648.7
705.06
843.3
O2/N2
12.25
11.9
10.65
CO2/CH4
38.9
47.25
51.4
N2/CH4
0.91
1.18
2.7
CO2/N2
42.8
40.2
24.95
H2/CO2
16.66
14.9
16.4
Table 5 Permselectivities coefficients for six binary gas mixtures for precursor membrane Gases/Temperature
PermSelectivities 313 (K)
333 (K)
373 (K)
H2/CH4
3.56
4.98
7.03
O2/N2
0.9
1.04
1.11
CO2/CH4
0.76
0.91
1.04
N2/CH4
0.87
0.79
0.80
CO2/N2
0.87
1.15
1.31
H2/CO2
4.68
5.44
6.63
capability of the system. The ‘‘apparent’’ [34] (sorption contributed) activation energy values of all gases, produced by implementing the Arrhenius analysis are already presented in Table 2 together with the isosteric heats of sorption calculated by Vant Hoff equation from the sorption experiments at different temperatures. These values are characteristic of many types of microporous membranes (zeolitic, CMSM, NF-Silica) with pore size in the area of 0.4–0.8 nm [35–38]. Moreover the higher selectivity coefficients for the carbon membrane are 843.3, 12.25, 51.44, 2.06, 40.2 and 16.66 for H2/CH4, O2/N2, CO2/CH4, N2/CH4, CO2/N2 and H2/CO2 respectively. As it was expected the selectivity performance of the polymer precursor membrane was by far inferior to this of its carbon analogue. For instance, after carbonization the CO2/N2 and CO2/CH4 permselectivities are 50 times higher, the O2/N2 20 times higher, while H2/CH4 selectivity was enhanced by a factor of 180! Thus, although the prepared polymeric P84 co-polyimide hollow fiber membrane does not have permselectivity properties adequate for its use as a gas separation membrane, it has been proved an excellent precursor for the preparation of highly selective carbon membranes. The high values of the H2/CH4 permeance selectivity coefficient (649–843) are indicative of a molecular sieving separation mechanism. Hydrogen, the smaller gas with
123
˚ [39], permeates faster than kinetic diameter of 2.89 A methane through the microporous membrane structure. Although methane molecules can be adsorbed (Fig. 7), a fact revealing that methane’s molecular size permits its entrance into the pore network, the high H2/CH4 permselectivity implies that there are kinetic restrictions severely hindering the direct entrance of CH4 molecule from the gas phase into the pore. Furthermore its propagation through the pore channels must be considerably delayed, when compared to H2, due to the strong adsorption potential (CH4 qst = 23.1 kJ/mol) arising by the closeness of the pore walls. This high value of the isosteric heat of sorption for methane means that its molecules remain stuck on the occupation sites of the pore walls for longer time than those of H2. It must be mentioned here that in micropores of such small size, the contribution of surface flow that should enhance the flux of CH4 is negligible since we don’t have the formation of an adsorbed liquid like layer on the pore walls as would be the case for a mesoporous solid. On the other hand the H2/CO2 permselectivity presents significantly smaller values (15–16.7) as compared to H2/ CH4 (649–843) and moreover the permselectivity coefficient for CO2/CH4 is quite promising and varies between 39 and 51.5, increasing linearly with temperature. These results are in fact contradictious with the aforementioned considerations of diffusion delay due to the enhanced adsorption potential, since the calculated isosteric heat of sorption for CO2 qst = 29 kJ/mol, is much higher than this for CH4 qst = 23 kJ/mol (Table 2). Thus the satisfactory separation performance for CO2 over CH4 can certainly be attributed to hindrance effects that have a greater impact to the overall microporous diffusion mechanism than adsorption. This argument may be fully justified by also taking into account that apart from its larger kinetic diameter (rCO2 = 0.33 nm < rCH4 = 0.36 nm), the shape of the CH4 molecule is rather spherical as compared to the linear CO2. Additionally since CO2 is preferentially adsorbed on carbon materials it is expected that the ‘real’ (in a binary mixture) CO2/CH4 selectivity will be enhanced and especially at lower temperatures below 373 K due to interface processes [34], [40] occurring on the external surface of the solid. More specific, the CO2 molecules will exhibit a much higher occupation degree of the external surface of the material, compared to their rival CH4 molecules, and thus the probability of been those that will enter into the pores in a much higher population should be thoroughly enhanced. In general as indicative from the results presented in Table 2, permeance drops with molecular kinetic diameter revealing the predominance of a molecular sieving mechanism for diffusion. The prepared carbon molecular sieve membranes can be use in several gas separation
J Porous Mater (2008) 15:625–633
applications such as nitrogen or oxygen production from air, hydrogen purification for use in fuel cell technology, biogas purification, natural gas upgrade etc.
4 Conclusions In this paper the preparation of asymmetric polymeric hollow fiber membranes with a dry/wet spinning method from commercial P84 co-polyimide and the preparation of carbon molecular sieve membranes derived by carbonization of BTDA-TDI/MDI (R84) co-polyimide hollow fiber membranes in inert nitrogen environment up to 1173 K are reported. The adsorption/desorption as well as permeation properties for gases of industrial interest have been extensively studied. The precursor polymer membranes have good permeability properties but medium gas separation efficiencies. On the other hand our results reveal that the developed carbon membranes exhibit higher permeability and they possess significant sieving properties and very high selectivity. For instance the calculated H2 permselectivity for H2/CH4 mixtures was 843 at 373 K. Acknowledgments The authors would like to thank Dr. Theodore Steriotis for the very important contribution of this paper. The authors would like also to thank the ‘‘HYDROCELL–E22’’ research project of the Greek General Secretariat for Research and Technology and the 04–3–001/6 ‘‘Archimedes’’ Research Project of the Greek Ministry of National Education and Religious Affairs for the support of this work.
633
12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23.
24. 25. 26. 27. 28.
29. 30.
References 31. 1. A. Soffer, J.E. Koresh, S. Saggy, Separation Device. U.S. Patent 4,685,940 (1987) 2. N.K. Kanellopoulos, Recent Advances in Gas Separation by Microporous Ceramic Membranes, (Elsevier Sci. B.V., 2000), p. 323 3. J.E. Koresh, A Soffer, Sep. Sci. Technol. 22, 973 (1987) 4. A.B. Fuertes, D.M. Nevskaia, T.A. Centeno, Micropor. Mesopor. Mater. 33(1–3), 115 (1999) 5. C.W. Jones, W.J. Koros, Carbon 32(8), 1419 (1994) 6. V.C. Geiszler, W.J. Koros, Ind. Eng. Chem. Res. 35(9), 2999 (1996) 7. Y.D. Chen, R.T. Yang, Ind. Eng. Chem. Res. 33(7), 3146 (1994) 8. M.B. Rao, S. Sircar, T.C. Golden, Gas separation by adsorbent membranes. U.S. Patent 5,104,425 (1992) 9. T.A. Centeno, A.B. Fuertes, Carbon 38(7), 1067 (2000) 10. J.E. Koresh, A. Soffer, Sep. Sci. Tech. 18, 723 (1983) 11. A. Soffer, M. Azariah, A. Amar, H. Cohen, D. Golub, S. Saguee, et al. Method of Improving the Selectivity of Carbon Membranes
32. 33. 34. 35. 36. 37. 38. 39. 40.
by Chemical Carbon Vapor Deposition. U.S. Patent 5,695,818 (1997) T.A. Centeno, A.B. Fuertes, J. Membr. Sci. 160(2), 201 (1999) J.C. Chen, I.R. Harrison, Carbon 40(1), 25 (2002) L.I.B. David, A.F. Ismail, J. Membr. Sci. 213(1–2), 285 (2003) E. Barbosa-Coutinho, M.M. Salim Vera, C.P. Borges, Carbon 41(9), 1707 (2003) M.G. Sedigh, L. Xu, T.T. Tsotsis, M. Sahimi, Ind. Eng. Chem. Res. 38(9), 3367 (1999) H. Hatori, H. Takagi, Y. Yamada, Carbon 42(5–6), 1169 (2004) A. Bu¨rger, E. Fitzer, M. Heym, B. Terwiesch, Carbon 13(3), 149 (1975) T. Takeichi, Y. Eguchi, Y. Kaburagi, Y. Hishiyama, M. Inagaki, Carbon 36(1–2), 117 (1998) H.B. Park, Y.K. Kim, J.M. Lee, S.Y. Lee, Y.M. Lee, J. Membr. Sci. 229(1–2), 117 (2004) V.C. Geiszler. Polyimide Precursors for Carbon Membranes. University of Texas, PhD thesis, (1997) P.S. Tin, T.S. Chung, Y. Liu, R. Wang, Carbon 42(15), 3123 (2004) T.A. Steriotis, F.K. Katsaros, A.K. Stubos, A.C. Mitropoulos, P. Galiatsatou, N. Zouridakis, et al., Rev. Sci. Instrum. 67(7), 2545 (1996) H. Zhonghua, N. Maes, E.F. Vansant, J. Porous Mater. 2, 19 (1995) D. Lozano-Castello´, D. Cazorla-Amoro´s, A. Linares-Solano, Chem. Eng. Technol. 26(8), 852 (2003) S.W. Rutherford, C. Nguyen, J.E. Coons, D.D. Do, Langmuir 19(20), 8335 (2003) S.J. Gregg, K.S.W. Sing, Adsorption, Surface Area and Porosity (Academic Press, London, 1982), p. 228 F. Rodriguez-Reinoso, A. Linares-Solano, in: Chemistry and Physics of Carbon, Vol 21, ed. by P.A. Thrower (Dekker, New York, 1988) , p. 1–146 J. Jagiello, A. Anson, M.T. Martinez, J. Phys. Chem. B 110(10), 4531 (2006) A.L. McClellan, H.F. Harnsberger, J. Colloid Interface Sci. 23(4), 577 (1967) S. Brunauer, L.S. Deming, W.E. Deming, E. Teller, J. Amer. Chem. Soc. 62(7), 1723 (1940) J.N. Barsema, G.C. Kapantaidakis, N.F.A. van der Vegt, G.H. Koops, M. Wessling, J. Membr. Sci. 216, 195 (2003) J.N. Barsema, N.F.A. van der Vegt, G.H. Koops, M. Wessling, J. Membr. Sci. 205, 239 (2002) R.S.A. De Lange, K. Keizer, A.J. Burggraaf, J. Membr. Sci. 104(1–2), 81 (1995) R.S.A. De Lange, J.H.A. Hekkink, K. Keizer, A.J. Burggraaf, J. Membr. Sci. 99(1), 57 (1995) Y. Takata, T. Tsuru, T. Yoshioka, M. Asaeda, Microporous Mesoporous Mater. 54(3), 257 (2002) E.E. McLeary, J.C. Jansen, F. Kapteijn, Microporous Mesoporous Mater. 90, 198 (2006) P. Ciavarella, H. Moueddeb, S. Miachon, K. Fiaty, J.A. Dalmon, Catal. Today 56(1–3), 253 (2000) D.W. Breck, Zeolite Molecular Sieves; Structure, Chemistry and Use (Wiley, New York, 1973), p. 636 R.M. Barrer, J. Chem. Soc. Faraday Trans. 86(7), 1123 (1990)
123
