Fluorine Doping of Microporous Organosilica Membranes for Pore Size Control and Enhanced Hydrophobic Properties.

Fluorine-doped organosilica membranes for gas and pervaporation (PV) separation were fabricated using a sol-gel method. NH4F and bis(triethoxysilyl)methane (BTESM) were selected as the dopant and Si precursor, respectively, for the fabrication of fluorine-doped organosilica membranes. Doping with fluorine was evaluated for its effect on the physicochemical properties of organosilica (hydrophobicity/hydrophilicity and network size). Fluorine doping dramatically eliminated the formation of Si-OH groups in the sol, so that the condensation of Si-OH groups during the calcination process was suppressed. It is possible that fluorine doping enlarged the network pore sizes in organosilica, because the F-BTESM (F/Si = 1/9) membrane showed superior He and H2 permeance with a low H2/N2 permeance ratio that corresponded to the network pore size by comparison with an undoped BTESM membrane. The F-BTESM (F/Si = 1/9) membranes clearly showed a high level of C3H6 permeance (>3.0 × 10-7 mol m-2 s-1 Pa-1) with a high C3H6/SF6 permeance ratio (∼250), which suggests that the network pore size of F-BTESM is suitable for the separation of large molecules such as hydrocarbon gases (C3/C4, C4 isomer, etc.). Organosilica membranes both with and without fluorine doping showed stable PV performance because of the fact that H2O permeance and each permeance ratio under different separation systems was approximately constant over 10 h at 70 °C. Fluorine doping enhanced the hydrophobic nature of the organosilica, which was confirmed by the H2O adsorption and PV properties.

Introduction

Amorphous n class="Chemical">SiO2 membranes with pore sizes in the sub-nanometer range have great potential for application to gas/liquid sehemical">paration.[1−3] Membrane performance (selectivity, permeability) depends on the thickness and pore size of the sehemical">paration layer, so that a thin microporous hemical">pan class="Chemical">silica layer is coated onto an intermediate layer (γ-alumina,[1,3] SiO2–ZrO2[2]) to form a porous substrate, which makes it possible to form a layer-by-layer structure to decrease the permeation resistance. The utilization of silsesquioxane (RSiO3/2) categorized as either a pendant-type[4−8] or a bridged-type[8−14] Si precursor makes it possible to control the pore size of an amorphous structure as well as the hydrophobic/hydrophilic properties, respectively, because of the presence of nonhydrolyzable organic groups. The network structure of bridged-type organosilica utilizing bis(triethoxysilyl)methane (BTESM) and bis(triethoxysilyl)ethane (BTESE) via a sol–gel method is suitable for H2/organic compounds[8,11] and C3H6/C3H8[10] separation, respectively, because the spacer (Si–CH2–Si, Si–C2H4–Si unit) can work as a minimum unit in networks for the construction of a loose network structure.

The effect that a linking unit with a pan class="Chemical">carbon number larger than C3 (Si–hemical">pan class="Chemical">C3H6–Si) exerts on network pore size and on a microporous structure was also evaluated by fabricating organosilica membranes such as [bis(triethoxysilyl)propane (BTESP), bis(trimethoxysilyl)hexane (BTMSH), bis(triethoxysilyl)benzene, and bis(triethoxysilyl)octane (BTESO)].[14] The network size determined by gas permeation properties became larger as the carbon number increased between 2 Si atoms, and H2 selectivity was decreased as the carbon number between 2 Si atoms increased, but despite having larger network pore sizes, the BTESP (Si–C3H6–Si), BTMSH (Si–C6H12–Si), and BTESO (Si–C8H16–Si) membranes showed smaller values for gas permeance compared with either the BTESM or the BTESE membranes. The increased flexibility of the linking units in the structures of Si–C3H6–Si, Si–C6H12–Si, and Si–C8H16–Si membranes tended to collapse the micropores and decrease the porosity, which was confirmed by the measurement of N2 adsorption properties. This tendency suggests that increased flexibility might not be appropriate for the formation of highly permeable membranes.

The incorporation of Si–F bonds in the pan class="Chemical">SiO2 structure via the addition of hemical">pan class="Chemical">NH4F and/or the utilization of triethoxyfluorosilane was an innovative strategy to control the physicochemical properties dictating the amorphous nature of SiO2, which includes hydrophilicity/hydrophobicity and network pore size.[15−21] Hydrophobic surface modification of porous media [mesoporous silica (MCM-41), zeolite] was successfully achieved by replacing the Si–OH groups with Si–F groups.[15−17] The incorporation of Si–F bonds affected the Si–O–Si bond angle,[18−20] and this change in the SiO2 structure made it possible to form a loose and uniform structure rather than that of SiO2.[19−21] Highly permeable gas separation membranes such as CO2/CH4 (CO2 permeance: 4.1 × 10–7 mol m–2 s–1 Pa–1, CO2/CH4 permeance ratio: 300 at 35 °C)[19] and C3H6/C3H8 (C3H6 permeance: 2.2 × 10–7 mol m–2 s–1 Pa–1, C3H6/C3H8 permeance ratio: 42 at 35 °C)[21] were tailored via a fluorine-induced SiO2 matrix. The pore size distribution obtained by single-gas permeation showed that densification of the network structure for a fluorine-induced SiO2 matrix was suppressed even under high temperatures and/or a hydrothermal atmosphere.[20,21]

In the present study, n class="Chemical">NH4F and hemical">pan class="Chemical">BTESM were selected as the dopant and Si precursor, respectively, to fabricate fluorine-doped BTESM membranes that could then be used to evaluate the effect that fluorine doping exerts on the organosilica network size as well as on the gas permeation and pervaporation (PV) properties. The fluorine doping of a network structure was characterized via X-ray photoelectron spectroscopy (XPS), Fourier transform infrared (FT-IR) spectra, and N2 adsorption properties. The hydrophobicity/hydrophilicity of organosilica was evaluated via H2O adsorption.

Results and Discussion

Effect of Fluorine Doping on the Physicochemical Properties of Organosilica

Figure shows the FT-IR spectra for n class="Chemical">BTESM (a) and hemical">pan class="Chemical">F-BTESM (F/Si = 1/9) (b) films before/after calcination at 350 °C. Si–O–Si asymmetric stretching vibrations were detected at around 1050–1100 cm–1[22,23] in both samples, irrespective of fluorine doping. Grill and Neumayer[22] reported that the peak at around 1100 cm–1 can be ascribed to the larger-angle Si–O–Si bonds in a cage structure with a bond angle of approximately 150°, while the peak at around 1050 cm–1 can be ascribed to the stretching of the smaller-angle Si–O–Si bonds in a network structure. The peak at 960–980 cm–1 can be ascribed to the stretching of Si–OH bonds.[23] Both asymmetric and symmetric C–H stretching vibrations of −CH2– fragments were detected in both samples at 2932 and 1350 cm–1,[24,25] respectively, irrespective of fluorine doping and with/without calcination. This indicated that organosilica networks with linking units (Si–CH2–Si) were formed by the hydrolysis/condensation of the ethoxy and silanol groups, respectively. Interestingly, the F-BTESM (F/Si = 1/9) film showed fewer chemisorbed H2O molecules (3400 cm–1)[23] compared with the BTESM film.

Figure 1

FT-IR spectra of BTESM (a) and F-BTESM (F/Si = 1/9) (b) films before/after calcination at 350 °C under an air atmosphere.

FT-IR spectra of pan class="Chemical">BTESM (a) and hemical">pan class="Chemical">F-BTESM (F/Si = 1/9) (b) films before/after calcination at 350 °C under an air atmosphere.

The peak area ratios for Si–OH (960–980 cm–1)/Si–O–Si (1050–1100 cm–1) in both samples before/after calcination at 350 °C are summarized in Table to demonstrate the effect of pan class="Chemical">fluorine doping on Si–OH groups. The hemical">pan class="Chemical">BTESM film showed a higher value for the Si–OH/Si–O–Si peak area ratio compared with that of the F-BTESM film (F/Si = 1/9), but the peak area ratio was largely decreased after calcination. On the other hand, the F-BTESM (F/Si = 1/9) film showed a lower value for the Si–OH/Si–O–Si peak area ratio, and only a slight decrease was confirmed after calcination at 350 °C. Thus, fluorine doping eliminated the formation of Si–OH groups in the sol (before calcination), and the condensation of Si–OH groups was suppressed during the calcination process.

Table 1

Peak Area Ratio of Si–OH (960–980 cm–1)/Si–O–Si (1050–1100 cm–1) for BTESM and F-BTESM (F/Si = 1/9) Films before/after Calcination at 350 °C under an Air Atmosphere

		peak area ratio (-)
		Si–OH/Si–O–Si
BTESM	before calcination	0.24
	after calcination	0.06
F-BTESM (F/Si = 1/9)	before calcination	0.03
	after calcination	0.02

Figure shows the XPS spectra in the range of 0–1200 eV (a) and from 680 to 695 eV (b) for n class="Chemical">F-BTESM (F/Si = 1/9) powders calcined at 350 °C. The F/Si molar ratio was calculated from each peak area ratio: Si 2p at 104.6 eV; F 1s and Si–F at 688 eV;[26] and C–F at 685–687.5 eV.[27,28] When hemical">pan class="Chemical">NH4F was doped into an amorphous SiO2 structure derived from TEOS, the doped F was present as Si–F bonds, which could be detected at 688 eV.[19,20] On the other hand, when NH4F was doped into the organosilica structure derived from BTESM (Si–C–Si unit), only a slight peak at 688 eV was detected, and the largest peak was detected in the range of 685–687.5 eV, which can be ascribed to semi-ionic and covalent C–F bonds. F-BTESM (F/Si = 1/9) calcined at 350 °C showed approximately the same fluorine concentration as the fluorine-doped concentration (10 mol %) in the sol, which confirmed no direct decomposition of NH4F at 145–225 °C.[29]

Figure 2

XPS spectra in the range of 0–1200 (a) and 680–695 eV (b) for F-BTESM (F/Si = 1/9) powders calcined at 350 °C under air.

XPS spectra in the range of 0–1200 (a) and 680–695 eV (b) for pan class="Chemical">F-BTESM (F/Si = 1/9) powders calcined at 350 °C under air.

Figure a shows the n class="Chemical">N2 adsorption isotherms at 77 K (0 < P/Ps < 0.1) for hemical">pan class="Chemical">BTESM and F-BTESM (F/Si = 1/9) powders calcined at 350 °C under an air atmosphere. Both samples showed a trend whereby the adsorbed amount of N2 increased with increase in the relative pressure, indicating microporous properties. However, the slope of F-BTESM (F/Si = 1/9) was higher than that of BTESM. Fluorine doping effectively increased the Brunauer–Emmett–Teller (BET) surface area and pore volume, as summarized in Table .

Figure 3

N2 adsorption isotherm at 77 K (0 < P/Ps < 0.1) (a) and H2O adsorption isotherm at 25 °C (b) for BTESM and F-BTESM (F/Si = 1/9) powders calcined at 350 °C under an air atmosphere.

Table 2

Pore Volume (MP) and BET Surface Area for BTESM and F-BTESM (F/Si = 1/9) Powders Calcined at 350 °C under an Air Atmosphere

	surface area (BET) (m2 g–1)	pore volume (MP) (cm3 g–1)
BTESM	597	0.34
F-BTESM (F/Si = 1/9)	893	0.65

n class="Chemical">N2 adsorption isotherm at 77 K (0 < P/Ps < 0.1) (a) and hemical">pan class="Chemical">H2O adsorption isotherm at 25 °C (b) for BTESM and F-BTESM (F/Si = 1/9) powders calcined at 350 °C under an air atmosphere.

Figure b shows the n class="Chemical">H2O adsorption isotherms at 25 °C for hemical">pan class="Chemical">BTESM and F-BTESM (F/Si = 1/9) powders calcined at 350 °C. The amounts of H2O adsorbed onto both samples were normalized by the BET surface area. The adsorbed amounts of H2O on BTESM gel powders increased in an approximately linear fashion with relative pressure increases in the range of 0–0.2, which showed a type III adsorption isotherm. This is because BTESM displays hydrophobic properties by comparison with conventional SiO2 because of the presence of organic linking units (Si–CH2–Si). A similar H2O adsorption property was reported in hydrophobic carbonized-template molecular sieve silica.[30] F-BTESM (F/Si = 1/9) gel powder showed a very low amount of H2O adsorption volume, despite having a larger pore volume compared with that of BTESM. Thus, fluorine doping significantly increased the hydrophobic property of organosilica because the density of Si–OH groups in F-BTESM (F/Si = 1/9) was much smaller than that in BTESM.

Gas Permeation Properties of Fluorine-Doped Organosilica Membranes

Figure shows the molecular size dependence of gas permeance at 200 °C for pan class="Chemical">BTESM and hemical">pan class="Chemical">F-BTESM (F/Si = 1/9) membranes calcined at 350 °C. Both membranes showed values for H2 permeance that were higher than 10–6 mol m–2 s–1 Pa–1 and permeance ratios for H2/SF6 that were above 1000, which was much higher than the Knudsen ratio (H2/SF6: 8.54). The F-BTESM (F/Si = 1/9) membrane clearly showed lower H2 selectivity (H2/N2, H2/C3H8) compared with that of the BTESM membrane. For example, the F-BTESM (F/Si = 1/9) membrane showed H2/N2 and H2/C3H8 permeance ratios of 4.9 and 11, while those for BTESM were 14 and 690, respectively. The BTESM membrane showed a C3H6 permeance of 2.0 × 10–8 mol m–2 s–1 Pa–1 with a C3H6/C3H8 permeance ratio of 13 at 200 °C, which indicated the appropriate network pore size for C3H6/C3H8 separation, as discussed in a previous paper.[10] On the other hand, the F-BTESM (F/Si = 1/9) membrane was less selective for C3H6/C3H8 (C3H6/C3H8 permeance ratio: 2.2) because of the enlarged network pore size, which decreased the molecular sieving properties.

Figure 4

Molecular diameter dependence of gas permeance (a) and dimensionless permeance based on He permeance (b) at 200 °C for BTESM and F-BTESM (F/Si = 1/9) membranes.

Molecular diameter dependence of gas permeance (a) and dimensionless permeance based on He permeance (b) at 200 °C for pan class="Chemical">BTESM and hemical">pan class="Chemical">F-BTESM (F/Si = 1/9) membranes.

Figure b shows the molecular size dependence of dimensionless permeance based on He permeance at 200 °C for pan class="Chemical">BTESM and hemical">pan class="Chemical">F-BTESM (F/Si = 1/9) membranes. The calculated dimensionless permeance based on He under the Knudsen mechanism is shown by the broken lines in the same figure. The dimensionless permeance for a BTESM membrane appeared to roughly follow the range of a Knudsen mechanism for He (0.26 nm) and CO2 (0.33 nm). The values of dimensionless permeance for molecules larger than N2 were smaller than those calculated under a Knudsen mechanism, and each value was clearly decreased with increases in the molecular size. On the contrary, the value of dimensionless permeance for the F-BTESM (F/Si = 1/9) membrane was approximately the same as that calculated in the range of He and C3H6, which indicated that gas molecules with molecular sizes ranging from 0.26 to 0.5 nm permeated the fluorine-organosilica network structure by Knudsen diffusion. The values of dimensionless permeance for C3H8 and SF6 molecules were smaller than those under Knudsen diffusion, which indicated that molecular sieving dominated the large-molecule permeation. Thus, fluorine doping is viable as a method that can be used to enlarge the network pore sizes of organosilica.

Theoretical analysis was used to estimate the pore sizes of n class="Chemical">BTESM and hemical">pan class="Chemical">F-BTESM (F/Si = 1/9) membranes. Figure S1 shows the relationship between k0,i1/3 and di for BTESM and F-BTESM (F/Si = 1/9) membranes calcined at 350 °C. The pore sizes for BTESM and F-BTESM (F/Si = 1/9) membranes were calculated at 0.49 and 0.62 nm, respectively, based on a modified gas translation model. It should be noted that the average pore size for a F-BTESM (F/Si = 1/9) membrane by nanopermporometry measurement was 0.5 nm.

It is difficult to estimate the membrane quality of microporous membranes with a large network pore size, because molecular sieving only works for large molecules, as shown in Figure b. In the present study, the relationship between H2 permeance and the H2/pan class="Chemical">SF6 permeance ratio is generally used as a benchmark for the sehemical">paration of H2 over organic compounds such as hemical">pan class="Chemical">toluene and methylcyclohexane. Because of the relatively large molecular sizes, the C3H6/SF6 permeance ratio (C3H6: 0.468 nm, SF6: 0.55 nm) was used to estimate the membrane quality.

Figure a shows the trade-off curve at 200 °C (H2 permeance vs H2/n class="Chemical">SF6 permeance ratio) for hemical">pan class="Chemical">organosilica (BTESM, BTESE,[8,11,14,31] and BTESO[14,31]), F-BTESM (F/Si = 1/9), and zeolite (DDR, CHA, FAU, MFI)[32−35] membranes. The solid line in this figure represents the trade-off between the BTESM and F-BTESM (F/Si = 1/9) membranes. The BTESE (Si–C2H4–Si) membranes showed high levels of H2/SF6 permeation properties. The BTESO membranes with flexible linking units (Si–C8H16–Si) showed lower values for H2 permeance (<10–6 mol m–2 s–1 Pa–1), which suggests that increased flexibility is inappropriate for the formation of highly permeable membranes.[14] F-BTESM (F/Si = 1/9) membranes showed higher levels of H2 permeance with moderate H2/SF6 permeance ratios compared with BTESM membranes, but the plotted results fell approximately within the trade-off curve. Thus, the network pore sizes of BTESM and F-BTESM (F/Si = 1/9) membranes were inadequate for the separation of H2 over large molecules.

Figure 5

Trade-off curve at 200 °C [H2 permeance vs H2/SF6 permeance ratio (a); C3H6 permeance vs C3H6/SF6 permeance ratio (b)] for organosilica (BTESM, BTESE,[8,11,14,31] and BTESO[14,31]), F-BTESM (F/Si = 1/9), and zeolite (DDR, CHA, FAU, and MFI)[32−35] membranes.

Trade-off curve at 200 °C [H2 permeance vs H2/n class="Chemical">SF6 permeance ratio (a); hemical">pan class="Chemical">C3H6 permeance vs C3H6/SF6 permeance ratio (b)] for organosilica (BTESM, BTESE,[8,11,14,31] and BTESO[14,31]), F-BTESM (F/Si = 1/9), and zeolite (DDR, CHA, FAU, and MFI)[32−35] membranes.

Figure b shows the trade-off curve at 200 °C (n class="Chemical">C3H6 permeance vs hemical">pan class="Chemical">C3H6/SF6 permeance ratio) for organosilica (BTESM, BTESE[11]) and F-BTESM (F/Si = 1/9) membranes. In general, the C3H6/SF6 permeance ratio decreases as the C3H6 permeance increases because of the presence of interparticle pores in a membrane. The solid line in this figure marks the trade-off for organosilica (BTESM, BTESE[11]) membranes. F-BTESM (F/Si = 1/9) membranes clearly showed the properties that dictate a high level of C3H6/SF6 permeation. Values for the kinetic diameter[36] and L-J length constant[37] for hydrocarbon gases (C3, C4) lay approximately between the sizes of C3H6 (0.468 nm) and SF6 (0.55 nm), and these are generally used as the effective molecular size for porous media, which makes the network pore size of F-BTESM suitable for the separation of hydrocarbon gases (C3/C4, C4 isomer, etc.).

Figure shows a scn class="Chemical">hematic image of the effect of n class="Chemical">fluorine doping on an hemical">pan class="Chemical">organosilica network structure before/after calcination. As described in the previous section, fluorine doping was present as Si–F and C–F (covalent and semi-ionic, respectively) bonds in the organosilica network. Fluorine doping eliminated the formation of Si–OH groups in the sol, and the condensation of Si–OH groups during the calcination process was suppressed. Wang et al.[38] evaluated the effect of calcination temperature (100–300 °C) on the network pore size of organosilica (BTESE) and concluded that He selectivity and the H2O/alcohol separation factor, corresponding to network pore size, were increased largely as calcination temperature was increased from 100 to 200 °C because of the condensation of Si–OH groups. Thus, the suppression of condensation from the Si–OH groups during the calcination process would have been a factor in the formation of a loose network structure via fluorine doping into organosilica. Because the incorporation of Si–F bonds affected the Si–O–Si bond angle, this change in the organosilica structure could also have helped promote the formation of looser organosilica and F–SiO2 network structures.[19−21]

Figure 6

Schematic image of the effect of fluorine doping on an organosilica network structure before/after calcination.

Schematic image of the effect of pan class="Chemical">fluorine doping on an hemical">pan class="Chemical">organosilica network structure before/after calcination.

PV Performance of Fluorine-Doped Organosilica Membranes

Figure shows the time courses for permeance and the permeance ratios through n class="Chemical">BTESM (a) and hemical">pan class="Chemical">F-BTESM (F/Si = 1/9) (b) membranes in EtOH/H2O, IPA/H2O, and n-BuOH/H2O systems at 70 °C under PV. In the present study, each alcohol concentration in the feed was controlled at 90 wt %. The BTESM membrane showed H2O permeance of approximately 2.0 × 10–6 mol m–2 s–1 Pa–1, irrespective of the separation systems (EtOH/H2O, IPA/H2O, and n-BuOH/H2O), but the permeance of each alcohol was decreased as the molecular size increased (molecular size of alcohol systems: EtOH: 0.43 nm, IPA: 0.47 nm, and n-BuOH: 0.49 nm).[39] For example, the permeance values for EtOH, IPA, and n-BuOH through the BTESM membrane were approximately 4.0 × 10–8, 5.0 × 10–9, and 3.0 × 10–9 mol m–2 s–1 Pa–1, which were approximately 1/50, 1/400, and 1/700 that of H2O permeance, respectively, and indicated the differences in the molecular sieving mechanisms of these molecules. The permeance values for IPA and n-BuOH reflected their similar molecular sizes.[39]

Figure 7

Time course for each permeance and permeance ratio through BTESM (a) and F-BTESM (F/Si = 1/9) (b) membranes in EtOH/H2O, IPA/H2O, and n-BuOH/H2O systems at 70 °C under PV.

Time course for each permeance and permeance ratio through n class="Chemical">BTESM (a) and hemical">pan class="Chemical">F-BTESM (F/Si = 1/9) (b) membranes in EtOH/H2O, IPA/H2O, and n-BuOH/H2O systems at 70 °C under PV.

PV performances were similar for the n class="Chemical">F-BTESM (F/Si = 1/9) membrane, and the permeance of hemical">pan class="Chemical">H2O was approximately constant irrespective of the separation system, but the permeance of EtOH was higher than that of either IPA or n-BuOH. It should be noted that organosilica membranes showed stable PV performance, irrespective of fluorine doping, because of the fact that H2O permeance and each permeance ratio under different separation systems was approximately constant over the course of 10 h at 70 °C.

The permeance values for n class="Chemical">H2O, hemical">pan class="Chemical">EtOH, IPA, and n-BuOH were plotted as the function of each molecular size together with those of gas molecules at 200 °C, as shown in Figure . The permeance of H2O through the BTESM membrane was higher than that of either He or H2, despite the larger molecular size. The differences in permeation temperature between single-gas permeation (200 °C) and PV (70 °C) can be explained via the adsorption of H2O molecules by organosilica.[38,40,41] The H2O molecules adsorbed by organosilica networks permeate via a surface diffusion mechanism. A similar trend was confirmed in BTESE (Si–C2H4–Si) membranes,[38,40,41] which show a hydrophobic nature by comparison with BTESM membranes. The values of EtOH, IPA, and n-BuOH fell approximately within the curve of molecular size dependence for gas permeance at 200 °C.

Figure 8

Molecular diameter dependence of each permeance in gas permeation and PV for BTESM (a) and F-BTESM (Fi/Si = 1/9) (b) membranes. The permeance values for He, H2, CO2, N2, CH4, C3H8, C3H6, and SF6 were measured at 200 °C for single-gas permeation, while those of H2O, EtOH, IPA, and n-BuOH were measured at 70 °C under PV.

Molecular diameter dependence of each permeance in gas permeation and PV for pan class="Chemical">BTESM (a) and hemical">pan class="Chemical">F-BTESM (Fi/Si = 1/9) (b) membranes. The permeance values for He, H2, CO2, N2, CH4, C3H8, C3H6, and SF6 were measured at 200 °C for single-gas permeation, while those of H2O, EtOH, IPA, and n-BuOH were measured at 70 °C under PV.

The n class="Chemical">F-BTESM (F/Si = 1/9) membrane showed He and H2 permeance values that were higher than 2.0 × 10–6 mol m–2 s–1 hemical">pan class="Chemical">Pa–1, which was approximately twice that of the BTESM membrane because of the enlarged network pore size. The H2O permeance for the F-BTESM (F/Si = 1/9) membrane, however, was approximately the same as that for the BTESM membrane. This result can be explained by the hydrophobic nature of F-BTESM, which impedes H2O permeation, although the effective molecular size of H2O was sufficiently small to allow permeation through the loose F-BTESM network. The permeance of H2O through the F-BTESM (F/Si = 1/9) membrane was lower than that for either He or H2, which indicates that molecular sieving dominates permeation by H2O molecules. The F-BTESM (F/Si = 1/9) membrane showed higher values for alcohol permeance by comparison with a BTESM membrane. Because there were no large differences in the alcohol adsorption properties between F-BTESM (F/Si = 1/9) and BTESM, as shown in Figure S2, the increased alcohol permeance through F-BTESM (F/Si = 1/9) could have been caused mainly by decreases in the molecular sieving properties because of loose F-BTESM networks.

Conclusions

n class="Chemical">Fluorine-hemical">pan class="Chemical">doped organosilica membranes for gas/liquid separation were fabricated via a sol–gel method. NH4F and BTESM were selected as the dopant and Si precursor, respectively, to fabricate fluorine-doped organosilica membranes that were used to evaluate the effect of doped fluorine on the physicochemical properties of organosilica (hydrophobicity/hydrophilicity, network size). Fluorine doping was present as Si–F and C–F (covalent and semi-ionic, respectively) bonds in the organosilica network. Fluorine doping eliminated the formation of Si–OH groups in the sol, so that the condensation of Si–OH groups during the calcination process was suppressed.

n class="Chemical">Fluorine doping enlarged the network pore sizes of hemical">pan class="Chemical">organosilica, and the F-BTESM (F/Si = 1/9) membrane showed He and H2 permeance that was superior that of a BTESM membrane. The F-BTESM membrane also had a lower H2/N2 permeance ratio that corresponded to the network pore size. F-BTESM (F/Si = 1/9) membranes clearly showed high C3H6/SF6 permeation properties (C3H6 permeance: >3.0 × 10–7 mol m–2 s–1 Pa–1, C3H6/SF6 permeance ratio: ∼250), which shows that the network pore sizes of the F-BTESM are suitable for the separation of hydrocarbon gases (C3/C4, C4 isomer, etc.).

n class="Chemical">Organosilica membranes both with and without hemical">pan class="Chemical">fluorine doping showed a stable PV performance, and both H2O permeance and permeance ratios under different separation systems were approximately constant over 10 h at 70 °C. Fluorine doping enhanced the hydrophobic nature of organosilica, which was confirmed by the results of H2O adsorption and PV.

Experimental Section

Preparation of the Fluorine-Doped Organosilica Sol and Membrane Fabrication

A n class="Chemical">fluorine-hemical">pan class="Chemical">doped organosilica sol was prepared via the hydrolysis and polymerization of BTESM with the addition of NH4F in ethanol with water and HNO3. A specified amount of BTESM and NH4F was added to ethanol under vigorous stirring at 25 °C [fluorine/Si = 1/9 (molar ratio)]. Then, a specified amount of HNO3 and water was added to maintain the solution composition at a BTESM/H2O/EtOH/HNO3 molar ratio of 1/200/1393/0.1 under a BTESM wt % of 0.5.

n class="Chemical">Fluorine-hemical">pan class="Chemical">doped organosilica membranes were formed onto porous α-alumina tubes (porosity: 47%, pore size: 2.1 μm) that were provided by the Nikkato Corporation. First, two types of α-alumina particles (average particle diameter: 200 nm, 2 μm) were dispersed in a SiO2–ZrO2 sol (average sol size: 200 nm) and coated onto the substrate, which was followed by calcination at 550 °C under an air atmosphere in order to form a defect-free intermediate layer.[10,11,14] Then, an F-BTESM (F/Si = 1/9) sol diluted to 0.15 wt % was coated onto the intermediate layer, which was followed by calcination at 350 °C under air for 30 min. In the present study, a hot coating method that was originally proposed by Asaeda[42] was applied. In this method, a diluted F-BTESM (F/Si = 1/9) sol was coated onto the intermediate layer at around 120 °C, which can prevent the formation of pores in the mesopore range during the drying process. These procedures were repeated approximately four times to form a thin defect-free separation layer.

Preparation and Characterization of Fluorine-Doped Organosilica Films and Gels

Powdered samples of the n class="Chemical">F-BTESM (F/Si = 1/9) gels were prehemical">pared by drying at 40 °C, which was followed by calcination at 350 °C under air for 30 min, and then, a mortar was used to grind the dried compound. The hemical">pan class="Chemical">N2 and H2O adsorption isotherms of the organosilica powder were measured at 77 K and 25 °C, respectively (BELMAX, BELJAPAN INC.). The condition of both the fluorine doping in the organosilica structure and that of the fluorine concentration for powdered samples was evaluated via XPS (Thermo Fisher Scientific, ESCALAB 250Xi, Al Kα = 1486.6 eV). A FT-IR spectrometer was used to measure the FT-IR spectra of the films coated onto Si-wafers (FT/IR-4100, Jasco, Japan).

Gas Permeation and PV Measurement

Single-gas permeation was conducted using the experimental apparatus shown in Figure a.[10,14] First, a membrane was pretreated at 300 °C under a pan class="Chemical">N2 flow to remove the chemisorbed hemical">pan class="Chemical">water molecules. Pure gases (He, H2, CO2, N2, CH4, C3H8, C3H6, and SF6) were fed to the outside (upstream) of a cylindrical membrane at 200 kPa, while the permeate side was maintained at atmospheric pressure. An electric furnace controlled the temperature of the permeation cell. A bubble film meter was used to measure the permeation rate.

Figure 9

Schematic diagram of the single-gas permeation (a) and PV (b) experiment.

The experimental PV apparatus is shown in Figure b.[43] Fabricated membranes were soaked in an aqueous solution with each pan class="Chemical">alcohol concentration at 90 wt % (hemical">pan class="Chemical">water/EtOH, water/IPA, water/BuOH) under a temperature of 70 °C. A reduction in the effect of temperature and concentration polarization on H2O flux was accomplished by circulating the solution at approximately 1600 rpm. The feed side was maintained at atmospheric pressure, while a vacuum pump (nearly 0 kPa) was used to evacuate the permeate side. A cold trap used liquid nitrogen to collect the permeated vapor, and the compositions of the feed and permeate streams were calculated by gas chromatography using a TCD detector (GC-14B, Shimadzu, Japan; Column: Porapak P). During the PV experiment, water was periodically added to the feed solution to control the concentration of the feed.