J. Phys. Chem. B 2005, 109, 22141-22146
22141
Factors Influencing the Transport of Short-Chain Alcohols through Mesoporous γ-Alumina Membranes Sankhanilay Roy Chowdhury,† Dave H. A. Blank, and Johan E. ten Elshof* Inorganic Materials Science, MESA+ Institute for Nanotechnology, Faculty of Science & Technology, UniVersity of Twente, P.O. Box 217, 7500 AE Enschede, The Netherlands ReceiVed: August 23, 2005; In Final Form: October 3, 2005
The pressure-driven transport of water, ethanol, and 1-propanol through supported γ-alumina membranes with different pore diameters is reported. Water and alcohols had similar permeabilities when they were transported through γ-alumina membranes with average pore diameters of 4.4 and 6.0 nm, and the permeability coefficient was found to be proportional to the square of pore size, in accordance with a viscous flow mechanism. For transport through membranes with an average diameter of 3.2 nm, the behavior of water was in accordance with the viscous flow mechanism, but the permeability of the membrane for ethanol and 1-propanol was much smaller than expected and could not be explained in terms of viscous flow. Although the low permeability of the membrane with 3.2 nm pores for ethanol and 1-propanol was partly due to the presence of small amounts of water in the alcohols, the permeability coefficients were still substantially smaller when water was absent. This intrinsic difference between water and alcohol may be due to differences in molecular size, chemisorption of alcohols on the oxide pore wall, which would lead to a reduction of the effective pore size, and/or a certain degree of translational ordering of the alcohol molecules inside the membrane pores, which leads to an effectively higher viscosity and, therefore, to a higher transport resistance.
1. Introduction Liquid permeation through nanometer-dimensioned pores has been receiving attention since the 1950s.1 In recent years, the transport of hydrophilic and hydrophobic solvents through porous media has become a very active field of research in membrane separation and catalysis. The permeation of alcohols through porous membranes is of major importance in two membrane separation processes, namely pervaporation2 and nanofiltration.3 In pervaporation, alcohols are transported through a membrane in vapor form,2 whereas in nanofiltration the transport occurs in liquid form.3 The permeability of solvents through porous membranes can depend strongly on specific membrane material-solvent interactions and the physicochemical properties of the permeating solvents. It is thus important to understand the solvent transport mechanism through the nanodimensioned pores, as the properties of solvents often deviate from their bulk properties in such small dimensions.4 Sandwich-type ceramic composite membranes with nanofiltration (NF) characteristics typically have pore radii in the range of 0.5-3.5 nm.5,6 Under moderate process conditions the pore size and structure of ceramic membranes is fixed. Inorganic NF membranes have been prepared from γ-alumina,5,7 titania,8,9 and silica-zirconia10,11 and were employed for the rejection of large organic molecules11 and the retention of small9,10 and complex ions.11,12 The transport on nonaqueous liquids through NF membranes is receiving considerable attention in recent years.13-17 In principle, when the transport of liquid through a membrane occurs by viscous flow, the volume flux J through a membrane * Corresponding author. Phone: +31-53-489-2695. Fax: +31-53-4894683. E-mail:
[email protected]. † Current address: Ecole Nationale Supe ´ rieure de Chimie Mulhouse, 3, rue Alfred Werner, 68093 Mulhouse Cedex, France.
is proportional to the applied pressure ∆P and inversely proportional to the viscosity η of the liquid, as expressed by Darcy’s law:
1 J ) - km∆P η
(1)
Here, km is the overall permeability coefficient of the membrane. Under conditions where Darcy’s law is valid, the membrane permeability coefficient of a single membrane layer ks with pore radius r is defined as
ks )
r2 8τL
(2)
where is the porosity of the membrane material, τ the tortuosity of the pore structure, and L the membrane layer thickness.18 From eq 2 it is clear that ks is a constant that depends only on the pore structure of the membrane, not on the physical properties of the permeating liquid. However, Tsuru et al. showed that the permeability of microporous/mesoporous silica-zirconia membranes for methanol, ethanol, and 1-propanol deviated from Darcy’s law when the pore diameter was between 1.2 and 3.5 nm, while “normal” viscous flow was observed when membranes with a pore diameter of 70 nm were used.19 The permeability decreased with increasing chain length of the permeating alcohol. They attributed this behavior to steric effects, i.e., the molecular size of 0.36-0.45 nm of the above-mentioned alcohols is considerable compared with the pore size, but they suggested that the affinity between the membrane and the alcohol molecules may also play an significant role. Since oxide membranes have OHfunctional pore walls, hydrogen bond interactions between an alcohol headgroup and surface hydroxyl groups may exert a
10.1021/jp054743o CCC: $30.25 © 2005 American Chemical Society Published on Web 11/02/2005
22142 J. Phys. Chem. B, Vol. 109, No. 47, 2005
Roy Chowdhury et al.
considerable influence over the mobility of the alcohol inside a membrane pore. More recently, we showed that the permeability of mesoporous ceramic γ-alumina membranes for nonpolar liquids such as toluene and n-hexane does not obey the viscous flow mechanism below a certain critical pore diameter when only traces of water are present in the solvent.20 While viscous flow was observed when the average pore diameter was 5.9 nm, substantial deviations from Darcy’s law were observed with membranes with pore sizes of 3.5-4.4 nm. Molar fractions of dissolved water of typically ∼5 × 10-4 decreased the permeability of toluene and n-hexane by a factor of 2-4. Water-free hydrocarbons followed the viscous flow mechanism for all investigated pore sizes. The presence of water is thought to result in capillary condensation of water from the hydrocarbon phase, which leads to pore blocking of the smaller pores. Because of the low solubility of water in nonpolar solvents, even ppm levels of water have a dramatic effect. This behavior is not expected when solvents with a high or unlimited solubility for water are used, such lower alcohols. In the present paper, we report on the liquid permeation behavior of water, ethanol, and 1-propanol through γ-alumina membranes with different pore sizes. It will be shown that ethanol and 1-propanol behave qualitatively similar as toluene and n-hexane, although the effect on permeability is less pronounced and occurs only below ∼4 nm pore diameter. Following the literature available in different related fields, we propose an explanation for the observed phenomena. 2. Experimental Section 2.1. Preparation of γ-Alumina Membranes and Powders. The γ-alumina membrane consists of a macroporous R-alumina support and a thin mesoporous γ-alumina layer. The support was made by colloidal filtration of 0.3 µm of R-alumina particles (AKP-30, Sumitomo). The dispersion was stabilized by peptizing with nitric acid. After being dried at room temperature, the filter compact was sintered at 1100 °C. Flat disks of Ø 39 mm and 2.0 mm thickness were obtained by machining and polishing. The porosity of the final supports is ∼30%, and the average pore diameter is in the range of 80-120 nm. Three different mesoporous γ-alumina membranes of ∼3 µm thickness were prepared by dip-coating the above-mentioned porous R-alumina supports in a boehmite solution, followed by drying and calcining at 450, 600, or 800 °C (heating/cooling rates 0.5 °C/min), respectively, for 1 h.20 The deposition/calcination cycle was carried out twice. Powders of γ-alumina phase were prepared by drying the boehmite solution in air, followed by calcination at 450, 600, or 800 °C (heating/cooling rates 0.5 °C/min), respectively, for 1 h. These powders were subjected to nitrogen adsorption/ desorption experiments (Micromiretics) at 77 K. The membranes and powder are hereafter designated as γ-450, γ-600, and γ-800, respectively, depending on calcination temperature. 2.2. Solvent Permeation Experiments. Liquid flux measurements on R-alumina supports were carried out using demineralized water (Q2 water), ethanol (Merck), and n-hexane (Merck) in a dead-end nanofiltration cell.20 Liquid flux measurements on supported γ-alumina membranes were carried out with demineralized water, ethanol, and 1-propanol (Merck). The volume of the cell is 1000 mL, and the operating pressure range was kept in the range of 2-14 bar. The stirring speed in the cell was kept constant at 200 rpm throughout all experiments. The water content of the alcohols was measured using Karl
Figure 1. Nitrogen sorption isotherms of γ-alumina powders calcined at (a) 450 °C, (b) 600 °C, and (c) 800 °C, respectively.
Fischer titration (Metrohm KFT 756). The water content of the as-received alcohols was 0.03 mol/L for ethanol and 0.025 mol/L for 1-propanol. Alcohol feeds with lower water concentrations were made by drying them using molecular sieves (Type 3A) and anhydrous CaCl2 as drying agents for different periods of time. Alcohol feeds with higher water contents were made by addition of water.
Transport of Short-Chain Alcohols through γ-Alumina Membranes
J. Phys. Chem. B, Vol. 109, No. 47, 2005 22143
Figure 2. Pore size distribution of γ-alumina powders calcined at 450 °C, 600 °C, and 800 °C, respectively.
3. Results and Discussion Figure 1 shows the nitrogen sorption isotherms of γ-450, γ-600, and γ-800 powders. The shape of the hysteresis loop becomes more elongated with increasing calcination temperature. The shape of the isotherm clearly indicates differences in pore shape/geometry in the different powders. As the calcination temperature is increased the hysteresis loop changes from type H2 to type H1.21 The γ-800 powders show steep adsorption and desorption branches (type H1). This is characteristic of tubular shaped capillaries with both ends open. The γ-450 powder shows a delayed adsorption branch and a steep desorption branch (Type H2). This may arise from the same type of open capillary as in type H1, but the effective radii of the bodies of the pores are heterogeneously distributed and the narrow entrances are of equal size. The γ-600 powder shows the intermediate behavior of type H1 and type H2, which reflects the transition from H2 to H1 with increasing calcination temperature. It is known that the crystallite shape of the primary oxide particles change with calcination temperature. The morphology of the γ-alumina crystallites changes from platelet-like into spherical with increasing calcination temperature.22 This change of shape also changes the shape of the space between two crystallites. At lower calcination temperatures, the pores are more or less slit shaped, but with increasing calcination temperature it changes into a more spherical shape. The pore size distributions of the γ-450, γ-600, and γ-800 powders as calculated from the N2 desorption isotherms using the BJH method are shown in Figure 2. The pore diameters are 1.8-4.2 nm (average 3.2 nm) for γ-450, 2-6 nm (average 4.4 nm) for γ-600, and 3-8 nm (average 6.0 nm) for γ-800.20 Thus, the pore size increases and the pore size distribution becomes broader with increasing calcination temperature. Figure 3 shows the product of viscosity and flux Jη for water, ethanol (as-received), and 1-propanol (as-received) through γ-800, γ-600, and γ-450 membranes. For γ-800 and γ-600 membranes the trend is in full agreement with Darcy’s law, eq 1. The slopes of the fitted straight lines in Figure 3a,b are the overall permeability coefficients kov of the stacked membranes. It is clear that the transport of water, ethanol, and 1-propanol through γ-800 and γ-600 membranes with an average pore diameter of 4-8 nm is by viscous flow. The behavior is different for the γ-450 membrane shown in Figure 3c. This membrane
Figure 3. Product of volume flux and viscosity of water, ethanol, and 1-propanol through (a) γ-800, (b) γ-600, and (c) γ-450 membranes.
has a different permeability for water than for the alcohols. It also seems that a minimum pressure of ∼0.7 bar has to be applied before a net flow of alcohols through the membrane occurs. The permeability coefficient kγ of an individual γ-alumina layer can be calculated from kov with the relationship17 -1 -1 k-1 ov ) kγ + kR
(3)
where kR is the permeability coefficient of the bare R-alumina
22144 J. Phys. Chem. B, Vol. 109, No. 47, 2005
Figure 4. Permeability coefficient kγ of γ-450 layer versus membrane pore radius. The permeability values of the alcohols are for as-received liquids.
TABLE 1: Permeability Coefficients kγ of Mesoporous γ-Alumina Membrane Layer for Water, Ethanol, and 1-Propanol (As Received) solvent
permeability coefficient kγ (10-14 m) γ-800 γ-600 γ-450
water 5.7 ( 0.2 2.8 ( 0.3 2.2 ( 0.09 ethanol (0.03 mol/L of H2O) 5.6 ( 0.1 2.6 ( 0.4 0.89 ( 0.06 1-propanol (0.025 mol/L of H2O) 5.6( 0.4 2.6 ( 0.6 0.86 ( 0.09
support. Permeation measurements with water, hexane, and ethanol through an R-alumina membrane indicated viscous flow behavior with kR ) (1.17 ( 0.08) × 10-14 m.17 Table 1 lists the permeability coefficients of the γ-alumina membrane layer kL for different membrane-solvent combinations. Figure 4 shows the kγ values versus the γ-alumina membrane pore radius on a log-log scale. The figure shows that a straight line with slope ∼2 can be drawn through most of the data points. This indicates that the permeability coefficient kγ is proportional to the second power of pore radius. This is in accordance with eq 2, provided that all other parameters remain more or less constant. The only deviation from this behavior concerns the alcohols at the smallest pore size of ∼3.2 nm. The effect of water present in the alcohols on the permeability coefficient of the γ-450 layer is shown in Figure 5. It is clear that the permeability of the γ-450 membrane for ethanol and 1-propanol increases with decreasing water concentration. The effect of permeability is most pronounced when the H2O concentration is lowered to below 0.02 mol/L. For each solvent with a certain water concentration, the volume flux was measured at different pressures. A linear fit of the experimental data of flux versus pressure did not seem to pass the origin, i.e., extrapolating the experimental trend to zero flux indicated that a small but non-zero threshold pressure had to be exceeded before net flow of alcohol occurred. The estimated values are listed in Table 2. It is known that the behavior of liquids in spaces confined to a few molecular diameters can differ strongly from the behavior of bulk liquids. The anomalous behavior of the γ-450 membrane in comparison with γ-600 and γ-800 is therefore likely related to specific features of its pore structure. The pores of γ-450 alumina membranes can be considered as two rough surfaces facing each other at an average distance of ∼3.5 nm. As Figure
Roy Chowdhury et al.
Figure 5. Influence of water concentration on permeability coefficient kγ of γ-450 layer for ethanol and 1-propanol.
TABLE 2: Threshold Pressure for Transport of Ethanol and 1-Propanol with Different Water Concentrations through a γ-450 Membrane solvent ethanol
1-propanol
water concentration (mol/L)
threshold pressure (bar)
0.003 0.009 0.02 0.03 0.20 0.002 0.007 0.012 0.025 0.40
0 0 0.3 ( 0.3 0.6 ( 0.4 0.4 ( 0.1 0 0 0.4 ( 0.3 0.7 ( 0.5 0.6 ( 0.4
2 shows, a small but substantial fraction of pores in γ-450 has a pore size
