1. Introduction

One of the superior gas separation materials, carbon membranes have focused in the last decade because of their high permeation performances and separation factors, which jump over the polymeric membranes’ upper bound[1-3]. In addition, carbon membranes have the excellent thermal and chemical stabilities as compared with polymeric membranes[4-5].

Up to now, many researchers studied the carbon membranes using various polymer precursors, such as polyimide[6-11], phenolic resin[12-13], and poly(furfuryl alcohol)[14]. Also, the pyrolysis conditions[10,11,15,16] and the chemical structures of polymer precursors having side pendent groups[15-18] significantly affected their gas separation performances. Basically, carbon membranes are divided into carbon molecular sieve (CMS) membranes and adsorption-selective carbon (ASC) membranes, which depend on the pore diameters. The former with pore diameters 3-5 Å have ideal separation factor and the latter have the pore diameters in the range of 5-7 Å[19]. Therefore, many researchers focused on the optimization of the pyrolysis conditions using various homo- or co-polymer precursors to control the porosity of carbon structures and successively prepared the carbon membranes having high permeabilities and selectivities.

On the other hand, Ozaki et al.[20] developed membranes by controlling the pores of carbon membranes using two kinds of polymers, one giving a high carbon yield and the other creating a very low carbon yield. According to this concept, Hatori et al.[21] prepared carbon membranes derived from polyimide/poly(ethylene glycol) blended polymer and characterized the mesoporous carbon structures using nitrogen-adsorption isotherm method. The properties of porous carbon structures (pore size, pore size distribution, pore volume, and bulk density) using polymer-blending method importantly are depended on the phase separation behavior between two polymers because the phase separation domain of thermally labile polymer acted as a pore-forming agent.

In our previous studies[3,22-24], we prepared the carbon membranes containing silica derived from poly(imide siloxane). The pyrolytic carbon-silica (C-SiO₂) membrane was consisted of the carbon-rich phase and siloxane-rich phase because of the micro-phase separation between the polyimide (carbon-rich aromatic phase) domain and the polyimide siloxane (siloxanerich phase) domain. Furthermore, we fabricated the CMS membranes using polyimide/polyvinylpyrrolidone blended polymer. This CMS membrane did not exhibit any characteristics of mesoporous carbon structures by thermal decomposition of polyvinylpyrrolidone evidenced from the nitrogen adsorption isotherms even though polyimide was hydrophobic and polyvinylpyrrolidone showed hydrophilic. Consequently, the permeation performance of the CMS membranes was enhanced with the less reduction of their separation factor[25].

To control the porosity of carbon membranes using polymer-blending method, the miscibility of two polymers is a major concerning element to control phase separation behavior. In the polymer blends, polymer- polymer interaction is an important factor for determining the phase separation of blends since phase behavior of blends is dominated by these interactions [26]. Therefore, the micro-phase separation domain of the thermal labile polymer might be related with the characteristics of porous carbon structures during the pyrolysis. In this study, we blended hydrophilic polyvinylpyrrolidone (PVP) in hydrophobic poly(imide siloxane) (Si-PI). In addition, the gas separation properties of the fabricated C-SiO₂ membranes derived from Si-PI/PVP blended polymers were investigated using single gases, He, CO₂, O₂, and N₂ as functions of the ratio of Si-PI/PVP and pyrolysis conditions, such as final pyrolysis temperature and isothermal time. It is also our intention to compare the sorption isothermal behaviors as well as the gas separation performance between carbonized membranes prepared from PI/PVP and Si-PI/PVP.

2. Experimental

2.1. Materials

Pyromellitic dianhydride (PMDA) and 4,4’-oxydianiline (ODA) were purchased from Tokyo Kasei Co., Inc., Tokyo, Japan and α,ω-aminopropyl poly(dimethyl siloxane) (PDMS, Shinetsu Co. Inc., Tokyo, Japan), which had M_n = 900 and n = 10 was used. Polyvinylpyrrolidone (PVP, M.W. 10,000, Aldrich Chem. Co., Milwaukee, WI, USA) was used as a thermally labile polymer. Tetrahydrofuran (THF) and N-methylpyrrolidinone (NMP) were obtained from Aldrich Chem. Co., Milwaukee, WI, USA and used after drying in 4 Å molecular sieves

2.2. Fabrication of C-SiO₂ membranes

Poly(imide siloxane) (Si-PI) copolymer was synthesized using PMDA, ODA, and PDMS. The chemical composition of PMDA/ODA/PDMS was 10/8/2 by mmol for the Si-PI. Firstly, PDMS was dissolved in tetrahydrofuran (THF, Aldrich Chem. Co., Milwaukee, WI, USA) and then N-methylpyrrolidinone (NMP, Aldrich Chem. Co., Milwaukee, WI, USA) solution of ODA was added. Finally, powder of PMDA was added to THF/NMP solution of PDMS and ODA, and then the reaction mixture was stirred under a nitrogen atmosphere at room temperature for 6 h to obtain homogeneous siloxane-containing poly(amic acid)(Si-PAA) solution. The solid concentration of Si-PAA was kept at 15% by weight. To prepare the Si-PAA/PVP solutions, powders of PVP were added 0% (denoted SiPI-P0), 2% (denoted SiPI-P2), and 5% (denoted SiPI-P5) by weight in the each of Si-PAA solutions. The Si-PAA/PVP solutions were cast onto a glass plate and then thermally imidized. The thermal imidization protocol used in this study was illustrated in our previous study[22]. The synthesized Si-PI/PVP precursors were kept in the vacuum oven to isolate from the moisture due to the hydrophilic properties of PVP and then used to prepare the pyrolytic C-SiO₂ membranes.

Finally, the carbon-silica (C-SiO₂) membranes were prepared by pyrolysis of the Si-PI/PVP precursors with the ratio of Si-PI/PVP and the final pyrolysis temperature (550, 600, 700, and 800°C) under argon purge flow rate of 300 [cc(STP)/min] in a tube furnace supported on an alumina holding plate. The pyrolysis protocol used in this study was also identical with that of our previous work[27]. Additionally, we employed an isothermal time of 60, 120, and 240 min at the final pyrolysis temperature of 550°C.

2.3. Characterization methods

Differential scanning calorimetry (DSC, TA Instruments, New Castle, DE, USA) measurements were used from -50 to 350°C under a nitrogen gas flow with a heating rate of 10 °C/min to measure the glass transition temperatures (T_g) of Si-PI/PVP polymer blends, which is a useful parameter in evaluating the phase separation behavior of the two polymers. Thermogravimetric analysis (TGA) was performed using a TA Instruments TGA2050 thermogravimetric analyzer (TA Instrument, New Castle, DE, USA) to study the thermal properties of PVP and Si-PI. To investigate the structural changes of C-SiO₂ membranes derived from Si-PI/PVP precursor by the final pyrolysis temperature and isothermal condition with evaluated average d-spacing values, wide-angle X-ray diffraction (WAXD) measurement was carried out using a Rigaku Denki model RAD-C (Rigaku, Tokyo, Japan) with Cu Kα radiation of wavelength 1.54 Å. To confirm the gas adsorption characteristics of the C-SiO₂ membranes derived from Si-PI/PVP precursors with the ratio of Si-PI/PVP, nitrogen adsorption isotherms [BET (Brunauer-Emmett-Teller) method] were observed at -196°C.

2.4. Gas permeation measurement

The permeation properties of the C-SiO₂ membranes derived from Si-PI/PVP precursors were measured using the high vacuum time-lag method at feed temperature of 25°C with the pressure difference of 8.11 × 106 Pa (760 Torr) between the feed side and the permeate side. To probe the molecular sieving capability of the C-SiO₂ membranes prepared with the thermal labile polymer and the pyrolysis conditions (pyrolysis temperature and isothermal time), the small gas molecules [He (2.6 Å), CO₂ (3.3 Å), O2 (3.46 Å), and N₂ (3.64 Å)] were used with the high purity of 99.99%. To measure the permeabilities of the flat-type C-SiO₂ membranes with the area of 2 × 2 cm² and the thickness of 30 μm, we masked the membrane using impermeable alumina tape, and then epoxy sealant was applied at the interface between the tape and the C-SiO₂ membranes to prevent any gas leak.

The measurement of gas permeability coefficients through the C-SiO₂ membranes was illustrated in our previous work[15,27]. The permeability is usually expressed in Barrers, where 1 Barrer is 10^-10 cm³(STP)cm/cm²⋅s⋅cmHg. The ideal separation factor is defined as the ratio of the pure gas permeabilities of each component (P_A and P_B) from the following equation[28]:

To obtain the reasonable gas permeation results of the C-SiO₂ membranes, all the gas permeation experiments were measured more than three time, and the error in the permeation measurements was less than 5%.

3. Results and Discussion

3.1. Thermal properties and phase behaviors of Si-PI/PVP

Fig. 1 shows the thermal properties of Si-PI (sample SiPI-P0) and Si-PI/PVP (sample SiPI-P5) precursors. While the sample SiPI-P0 exhibited two-step weightloss curves, the sample SiPI-P5 showed three-step decomposition profiles. From the results, the first weigh-loss appeared by the PVP decomposition was well agreed with TGA data in our previous study[26], the second and the third weight-loss curves exhibited the decomposition of PMDA-PDMS segments and PMDA-ODA segments. This trend followed well with the results of siloxane-containing polyimides[22].

Fig. 2 shows the DSC thermograms for the blend polymers with the ratio of Si-PI/PVP. To define the phase separation of a polymer blend, DSC is the most conventionally used to determine T_g or two T_gs. A single Tg is the criterion for the miscibility of a polymer blend. On the other hand, an immiscible polymer blend exhibits more than one T_g[29]. As shown in the figure, the T_g of the Si-PI/PVP polymer blends were significantly upper-shifted with decreasing the amount of PVP. In the literatures[30,31], PVP has a T_g in the range of 126-174°C while those of Si-PI was observed at 221°C due to the aromatic imide domain. From these results, it was concluded that Si-PI and PVP formed the homogeneously mixed polymer blends. However, we deduced that PVP formed the micro- phase separation domain in Si-PI even though the hydrophobic Si-PI and the hydrophilic PVP were homogeneously mixed. Therefore, the effect of the micro- phase separated PVP domains on the properties of porous carbon structures during the pyrolysis is studied with the N₂ adsorption isotherms and their gas separation properties.

3.2. Nitrogen adsorption isotherms

Gas adsorption isotherm is one of the most widely used techniques to characterize porous carbon structures[ 32]. A knowledge of adsorption mechanisms in different sizes of porosity is necessary to explain the shapes of isotherms. The isotherm shapes are divided into six major classes according to IUPAC classification. One of the physical adsorption according to IUPAC classification, type I isotherm is given by the microporous materials having pores no wider than ~2 nm whereas type IV isotherm is related with the mesoporous materials with pores in the range 2-50 nm. As shown in Fig. 3, Fig. 4, and Fig. 5, the C-SiO₂ membrane derived from sample SiPI-P0 precursor showed the type I isotherm. On the other hand, the isotherm shapes of the C-SiO₂ membranes prepared by pyrolysis of samples SiPI-P2 and SiPI-P5 precursors were changed to the type IV isotherm shapes with increasing content of PVP.

According to the nitrogen adsorption isotherms, type I isotherms are related with the microporous solids having a small external area. The very steep region at low P/P0 is due to the filling of very narrow pores and the limiting uptake is dependent on the accessible micropore volume[33]. Therefore, the C-SiO₂ membrane derived from sample SiPI-P0 precursor might have the microporous carbon structures. However, the C-SiO₂ membranes prepared by pyrolysis of Si-PI containing PVP showed type IV isotherms possessing a significant hysteresis loop with increasing the amount of PVP, which is associated with mesoporous materials. That is, the mesoporous carbon structures were created by the thermal decomposition of phase separated PVP domain during the pyrolysis. Interestingly, we did not observe the type IV isotherms showing a hysteresis loop in the carbon molecular sieve (CMS) membranes derived from polyimide containing PVP in our previous study [26]. Carbon membranes from PI/PVP showed the typical type I isotherm.

As compared with the N₂ adsorption isotherms of the CMS membranes derived from PI/PVP and the C-SiO₂ membranes derived from Si-PI/PVP, the properties of porous carbon structures were mainly affected by the phase separation behaviors between two polymers. In the literature, the miscibility of polymers is determined by the interaction between different polymers[ 34]. The miscibility of PVP and Si-PI decreased due to the hydrophobic nature of PMDA-PDMS segments even though two polymers were homogenously mixed. That is, the phase separated PVP domain was mixed with Si-PI, and also the larger phase separated domain size of PVP was formed in Si-PI than the phase separation domain of PVP in polyimide. In addition, the larger phase separation PVP was affected by the incremental PVP contents. Therefore, the C-SiO₂ membranes derived from Si-PI/PVP polymer blends obtained the mesoporous carbon structures caused by the thermal decomposition of PVP as confirmed by the nitrogen adsorption isotherm results. The relationship between the characteristics of N₂ adsorption isotherms and the gas permeation results of the C-SiO₂ membranes is described with detailed explanation in the next section.

3.3. Molecular sieving properties

Until now, we characterized the porous structures of the C-SiO₂ membranes prepared by the pyrolysis of blend polymers containing the thermal labile polymer using N₂ adsorption isotherms. Basically, the thermal labile polymer, PVP acted as a pore-forming agent in the CMS membranes during the pyrolysis even though the porous carbon structures were influenced by the miscibility of two blending polymers. In the next section, the results of gas permeation performances and separation properties of C-SiO₂ membranes prepared with the ratio of Si-PI/PVP and the pyrolysis conditions, such as the final pyrolysis temperature and the isothermal time are described.

3.3.1. Effect of thermal labile polymer

To study the gas permeation performance of the C-SiO₂ membranes derived from blend polymers with the ratio of Si-PI/PVP, the gas permeabilities through the C-SiO₂ membranes were measured using pure single gases, such as He, CO₂, O2, and N₂ at 25°C. Fig. 6 shows gas permeabilities and Fig. 7 shows selectivities of the C-SiO₂ membranes prepared by the pyrolysis of samples SiPI-P0, SiPI-P2, and SiPI-P5 precursors at 550°C. As shown in Fig. 6, all of the C-SiO₂ membranes showed that their permeabilities increased with the order of gas molecular size [He (2.6 Å) > CO₂ (3.3 Å) > O2 (3.46 Å) > N₂ (3.64 Å)]. In general, the permeabilities of small gas molecules through the microporous CMS membranes are basically determined by the kinetic diameters of gas molecules. That is, the gas transport mechanism through the C-SiO₂ membranes derived from Si-PI/PVP precursors followed the molecular sieving mechanism.

The main objective of this study was to study the effect of the thermal labile polymer on the gas permeation performances of the C-SiO₂ membranes. To achieve this purpose, we introduced the PVP in the Si-PI with the amount of 2 and 5% by weight. From the results, all of the gas permeabilities through the C-SiO₂ membranes increased remarkably with incremental PVP contents. These gas permeation results are well explained with the results of N₂ adsorption experiments as introduced above. The isotherm of sample CSi550-P5 illustrates that the adsorption rapidly increased at P/P0 > 0.8 and shows the hysteresis during the desorption process, which indicates that a number of ink-bottle-like pores are formed by the thermal decomposition of PVP during the pyrolysis. Furthermore, the C-SiO₂ membranes showed the higher surface area with the incremental PVP contents in the order: 96.48 m²/g (sample CSi550-P0) < 113.59 m²/g (sample CSi550-P2) < 136.04 m²/g (sample CSi550-P5). Also, the total pore volume increased in the order: 0.353 cm³/g (sample CSi550-P0) < 0.365 cm³/g (sample CSi550-P2) < 0.513 cm³/g (sample CSi550-P5). In our previous study[26], we also confirmed that the gas permeabilities through the CMS membranes derived from PI/PVP increased with the higher PVP contents.

The selectivities of all of the C-SiO₂ membranes are shown in Fig. 7 as a function of PVP contents in Si-PI. The He/N₂, CO₂/N₂, and O₂/N₂ selectivities of the C-SiO₂ membranes derived from Si-PI/PVP decreased with increasing the contents of PVP. In particular, the O2 permeabilities of the C-SiO₂ membranes increased from 135 to 1080 Barrer and the O₂/N₂ selectivities decreased from 9 to 8 with the increasing ratio of Si-PI/PVP.

3.3.2. Effect of pyrolysis temperature

In CMS membranes, the gas permeabilities and separation properties are mainly depended on the pyrolysis conditions, such as pyrolysis temperature[11,15,35,36], heating rate[11], isothermal time[9], and pyrolysis atmosphere[ 9,11] because the pore size, pore size distribution, and bulk density of porous carbon structures are determined by the pyrolysis conditions. In the present study, we fabricated the C-SiO₂ membranes derived from sample SiPI-P5 with the final pyrolysis temperature from 550 to 800°C under argon flow and measured their gas permeabilities. As shown in Fig. 8 and Fig. 9, all of the gas permeabilities of the C-SiO₂ membranes decreased with increasing pyrolysis temperature while the selectivities increased. In the literatures [11,35,36], the pyrolysis temperature determined the permselectivities of carbon molecular sieve membranes due to the enhanced compactness of the turbostratic carbon structure and the reduction of effective pore sizes, which brought about the lower permeabilities and the higher selectivities.

The permeability (P = DS) of non-porous polymeric membrane and CMS membrane is determined by the diffusion coefficient (D) and the solubility coefficient (S), and the permselectivity (α = P_A/P_B) is terms of the diffusion selectivity (DA/DB) and the solubility selectivity (SA/SB). In the CMS membranes, the diffusion coefficient is a main determining factor for their permeation performances. The O2 and N₂ diffusion coefficients and the O₂/N₂ diffusion selectivities of the C-SiO₂ membranes prepared by varying the pyrolysis temperature are listed in Table 1. The diffusion coefficient (D = L²/6θ) can be estimated from the membrane thickness (L) and the time lag (θ) of a transient permeation measurement. As listed in the table, the O₂/N₂ permselectivities through C-SiO₂ membranes particularly prepared by the higher pyrolysis temperature were mainly determined by the diffusion selectivities. In the meantime their solubility selectivities were the range of 1.2-1.8 which was the general trend of CMS membranes. These results are consistent with the reduction of effective pore sizes at higher pyrolysis temperature.

3.3.3. Effect of isothermal time

One of the pyrolysis conditions, the isothermal time (soaking time) was varied to prepare C-SiO₂ membranes derived from sample SiPI-P5 from 60 to 240 min at 550°C and measured their permeation performances. Koros et al.[36] reported on the effect of isothermal time on the permeabilities and selectivities of CMS membranes. In particular, the reduction of isothermal time effectively affected the improved permeability while the selectivity was maintained.

Fig. 10 and Fig. 11 show the results of permeabilities and selectivities of the C-SiO₂ membranes prepared with the incremental isothermal time of 60, 120, and 240 min at 550°C, respectively. The permeabilities of all the gas molecules through the C-SiO₂ membranes decreased with increasing the isothermal time while their selectivities increased. This result coincided with the Koros’ report. Among the gas separation properties of the C-SiO₂ membranes, the O₂/N₂ se-

lectivity of the C-SiO₂ membrane prepared with the isothermal time of 240 min is 110% higher than those of the C-SiO₂ membranes prepared with the isothermal time of 60 min while the permeability is 26% lower. In this study, very high separation factors of the C-SiO₂ membranes can be achieved by the varying isothermal time conditions.

The relationship between the permeation results and the porous C-SiO₂ membranes was studied by calculating average d-spacing values using WAXD patterns. In general, WAXD is used to measure the difference in interlayer spacing between carbon materials[37]. The average d-spacing values of the C-SiO₂ membranes were calculated from the Bragg equation as follows:

where d is the d-spacing value, θ the diffraction angle, λ the X-ray wavelength, and n an integral number. As shown in Fig. 12, broad peaks of the C-SiO₂ membranes were observed at 2θ = 22.24, 23.48, and 23.91° with an incremental isothermal time. As listed in Table 2, the calculated average d-spacing values of the C-SiO₂ membranes decreased with increasing the isothermal time. From the WAXD patterns, the separation factors of the C-SiO₂ membranes were reasonably explained with the average d-spacing values. That is, the constructed initial porous carbon structures derived from polymer blends shrank under the isothermal condition. Therefore, the carbon layers were stacked densely, which led to the reduced pore size between the carbon layers. Consequently, the C-SiO₂ membranes prepared under the isothermal conditions showed the higher selectivities and reduced permeabilities.

Fig. 13 shows performances of the C-SiO₂ membranes prepared by the thermal labile polymer and the pyrolysis conditions as compared with other carbon membranes derived from the PMDA/ODA-based polyimide precursors. From the results, the performances of the C-SiO₂ membranes derived from Si-PI/PVP precursors located upper right position as compared with other CMS membranes. In particular, the C-SiO₂ membranes showed the similar slope with the upper bound of polymeric membranes between the O2 permeabilities and the O₂/N₂ selectivities with the pyrolysis temperature in this study. Park et al.[17] also reported that the slope of the upper bound between the permeability and the selectivity of CMS membrane was depended on the final pyrolysis temperature if the same precursors were used. Consequently, the C-SiO₂ membranes derived from Si-PI containing PVP showed the O2 permeabilities of 30-1,080 Barrer and the O₂/N₂ selectivities of 8-17 in this study depending a pyrolysis temperature and isothermal time.

4. Conclusions

In this study, the C-SiO₂ membranes were prepared by pyrolyzing Si-PI/PVP polymer blends and by varying the ratio of Si-PI/PVP and the pyrolysis conditions. The characteristics of porous carbon structures derived from polymer blends was investigated using nitrogen adsorption isotherms. While the C-SiO₂ membranes derived from Si-PI showed the typical type I isotherm according to IUPAC classifications, the C-SiO₂ membranes with the higher content of PVP showed the type IV isotherm with the hysteresis loop. It is believed that the phase separated domain of the thermal labile polymer PVP in Si-PI affected the microporous carbon structures containing mesopores. From the molecular sieving properties of the C-SiO₂ membranes, the gas permeabilities increased with the content of PVP in Si-PI while their selectivities decreased. In addition, the performances of the C-SiO₂ membranes were significantly related with the pyrolysis conditions, such as the final pyrolysis temperature and the isothermal time. In particular, the C-SiO₂ membranes pyrolyzed at 550°C with the incremental isothermal times showed the 110% enhanced O₂/N₂ selectivity while the O2 permeability was reduced by about 26%.