A simple approach to hierarchical ceramic ultrafiltration membranes

A simple approach to hierarchical ceramic ultrafiltration membranes

Journal of Membrane Science 212 (2003) 29–38 A simple approach to hierarchical ceramic ultrafiltration membranes Kimberly A. DeFriend a,b , Andrew R...

607KB Sizes 1 Downloads 16 Views

Journal of Membrane Science 212 (2003) 29–38

A simple approach to hierarchical ceramic ultrafiltration membranes Kimberly A. DeFriend a,b , Andrew R. Barron a,b,c,∗ a Department of Chemistry, Rice University, 6100 Main Street, Houston, TX 77005, USA Center for Biological and Environmental Nanotechnology, Rice University, Houston, TX 77005, USA Department of Mechanical Engineering and Materials Science, Rice University, Houston, TX 77005, USA b

c

Received 29 May 2002; received in revised form 16 August 2002; accepted 16 September 2002

Abstract Asymmetric alumina ultrafiltration membranes with a hierarchical structure have been fabricated using carboxylic acid surface stabilized alumina nanoparticles (alumoxanes). Pre-formed hollow ␣-alumina spheres (3 ␮m nominal diameter) were prepared, using 2, 5, or 8 wt.% alumoxane solutions, and suspended in an aqueous solution of acetic-alumoxane (A-alumoxane) nanoparticles. This suspension was contacted with an ␣-alumina support. Firing to 600 ◦ C gave a defect-free alumina hierarchical membrane with a total thickness of ca. 2 ␮m. The flux and permeability for the membrane containing alumina spheres derived from 2 wt.% solution of A-alumoxane is comparable to the porous support, while those derived from 5 and 8 wt.% solution of alumoxane are similar to a “flat” alumoxane-derived membrane. An alternative route to increasing the flux and permeability of the asymmetric membranes was developed whereby colloidal polystyrene beads of either 0.75, 3, or 15 ␮m diameter, were suspended in an aqueous solution of either A-alumoxane (1 wt.%) or methoxy(ethoxyethoxy)acetic acid alumoxane (MEEA-alumoxane, 10 wt.%). The surface of an ␣-alumina support was dip coated in the polystyrene/alumoxane colloidal solution, dried then fired to 600 ◦ C, resulting in an asymmetric alumina membrane with a hierarchical tertiary structure. As the polystyrene out-gasses during pyrolysis the top of the coated spheres burst, resulting in a macroporous membrane in which the ceramic walls have a pore size and hence MWCO defined by the alumina formed from the alumoxane nanoparticles (A-alumoxane versus MEEA-alumoxane) rather than the macroporous structure of the membrane itself. The permeabilities of these membranes are equivalent or better than the support. © 2002 Elsevier Science B.V. All rights reserved. Keywords: Ceramic; Hierarchical; Alumina; Membrane; Nanoparticle; Ultrafiltration

1. Introduction The application of ceramic membranes in pollution prevention, resource recovery and waste treatment activities is increasing due to their excellent mechani∗ Corresponding author. Tel.: +1-713-348-5610; fax: +1-713-348-5619. E-mail address: [email protected] (A.R. Barron). URL: http://www.rice.edu/barron

cal strength and tolerance to solvents, as well as pH, oxidation, and temperature extremes [1,2]. In addition, the amphoteric properties of ceramic surfaces result in uniquely versatile membranes for water and waste-water treatment [3,4]. Alumina sol–gel-derived membranes are presently the most accepted routes to making alumina ultrafiltration filters. Leenaars et al. first developed the technique of using sol–gel processes to make alumina ultrafiltration membranes [5–8]. These filters, along

0376-7388/02/$ – see front matter © 2002 Elsevier Science B.V. All rights reserved. PII: S 0 3 7 6 - 7 3 8 8 ( 0 2 ) 0 0 4 5 0 - 7

30

K.A. DeFriend, A.R. Barron / Journal of Membrane Science 212 (2003) 29–38

with the vast majority of those reported in the literature [9–12], were made by the controlled hydrolysis of aluminum alkoxides to form alumina. The preparation techniques used by various researchers vary the drying or sintering conditions, which result in small changes in porosity or pore size. The membrane selectivity is primarily dependent upon the pore-size distribution; the narrower the pore’s size distribution, the more selective the membrane. However, for sol–gel the pore size is generally limited to the sizes of the precursor particles before sintering, which is difficult to control. Furthermore, sol–gels must be used immediately after preparation to avoid aggregation or precipitation. We have previously reported the fabrication of asymmetric alumina ultrafiltration membranes using carboxylic acid surface stabilized alumina nanoparticles (carboxylate-alumoxanes) [13,14]. A comparison with membranes derived from sol–gel methods showed their properties to be favorable, and we have recently reported further improvements [15]. The synthesis of the alumina nanoparticles is simple and low cost, producing a defect-free membrane in a one-step process. For example, the carboxylate-alumoxane nanoparticles may be prepared, in sufficient quantity for 100–200 m2 of finished membrane, in a single laboratory-scale batch with the cost of raw material being less that US$ 5. Once prepared, the carboxylate-alumoxanes are stable for months in solution, or may be dried and redissolved as desired, without changes in particle size. For both sol–gel and our carboxylate-alumoxanederived membranes supporting a relatively thin selective membrane onto a thicker, more permeable substrate yielding an asymmetric membrane enhances mechanical integrity and permeability (Fig. 1a). Unfortunately, despite this approach, and due to the small pore size of the alumoxane-derived membranes (ca. 2 nm) [13,14], the permeability of the asymmetric membranes is significantly lower than that of the support. In order to approach the permeability of the support, the ultrafiltration membrane must be as thin as possible. Sol–gel-derived membranes often require multiple dip-fire sequences to ensure integrity [5–12]. In contrast, we have found that a single-step process is sufficient for the alumoxane nanoparticle approach. Unfortunately, in order to ensure a defect-free membrane a thickness of 1–2 ␮m is required. Thus, an

alternative approach must be used in both processes to increase flow [16,17]. A typical approach for a sol–gel membrane is to layer materials of different porosity such that the thinnest possible layer of the “effective” ultrafiltration membrane is provided. However, this also required multiple process steps and each layer may result in a decrease in flux. If decreasing membrane thickness is not practical, an alternative approach is to increase the macroscopic surface area of the membrane. By analogy with biological membranes, one proposal is the creation of a hierarchical structure, in which the macroscopic structure evolves through ever decreasing sizes. A good example of such a structure would be the mammalian lung. A hierarchical approach has previously been used for organic membranes [18] and mesoporous materials [19]. In this paper, we present a simple approach to a simple second-order hierarchical asymmetric alumina membrane cast on porous ceramic supports (Fig. 1b and c).

2. Experimental procedure 2.1. Materials Acetic acid and methoxy(ethoxyethoxy)acetic acid functionalized alumina nanoparticles (A-alumoxane and MEEA-alumoxane, respectively) were prepared by previously published methods [20–22]. Aqueous solutions of alumoxane were degassed before use. RefractronTM ␣-alumina supports were obtained from the Refractron Technologies Corp. (Newark, NJ) and were heated to 600 ◦ C prior to use to remove surface grease. Colloidal polystyrene beads of 0.75, 3, or 15 ␮m diameter and 3 ␮m spheres in the dry form, were obtained from Polysciences Inc. Flat alumina membranes were deposited by dip coating the surface of an alumina support by the previously published method [13,14]. 2.2. Characterization methods Scanning electron microscopy (SEM) studies were performed on a Phillips XL-30 ESEM scanning microscope. The samples were attached to a metal mount using carbon tape. Due to the insulating nature of the

K.A. DeFriend, A.R. Barron / Journal of Membrane Science 212 (2003) 29–38

31

Fig. 1. Schematic representation of (a) a simple asymmetric filter and alternative designs of a hierarchical asymmetric filter using, (b) pre-formed alumina spheres, and (c) polystyrene-derived voids.

materials, a thin layer of gold was applied as a coating to provide a conducting surface. AFM images, grain size analysis, and surface roughness analysis of samples were obtained using a Nanoscope IIIa Scanning Probe Microscope (Digital Instruments, Santa Barbara, CA) in tapping mode. FESP tips were used with a pyramidal shape and

end radius of 5–10 nm (Digital Instruments). Samples were attached to 15 mm magnetic specimen disks with carbon tape. Roughness and cross-section analysis were determined by the accompanying Nanoscope IIIa software. Porosity, surface area, pore volume, and pore-size distributions were obtained using a Coulter SA3100.

32

K.A. DeFriend, A.R. Barron / Journal of Membrane Science 212 (2003) 29–38

Helium was used to determine the free space in the sample tube and nitrogen as the absorbate gas. All samples were out-gassed at 300 ◦ C for 2 h under a stream of dry nitrogen using a Coulter SAPrep. Calculations were based on the cross-sectional area of nitrogen using the value of 0.162 nm2 . Surface area was calculated using the Brunauer–Emmett–Teller (BET) equation with five data points. Pore-size distributions were determined using the Barrett–Joyner–Halenda (BJH) technique using 65 data points from the nitrogen adsorption isotherm [23]. Pore volume calculation was performed at a relative pressure of 0.9814. Pure water flux was measured on both the carboxylate-alumoxane-derived filters and the hierarchical carboxylate-alumoxane-derived filters at a pH close to the isoelectric point of alumina. Samples were placed in a Nalgene, model 300-4000 dead end filtration cell. A zero-air tank was connected to the cell for pressure, and a regulator was used to maintain a constant pressure of 10 psi. Permeate was collected at atmospheric pressure, so that the pressure at the end of the regulator was equal to the transmembrane pressure. Permeate volume was measured over time to calculate flux and permeability. 2.3. Formation of hollow α-alumina spheres Hollow ␣-alumina spheres were prepared by an adaptation of previously published methods [15]. Dry-form polystyrene beads (3 ␮m) were coated with an aqueous solution of A-alumoxane (1, 2, 5, 8, or 10 wt.%). The solution was pipetted onto the beads that were placed in coated ceramic crucible, and allowed to dry in air. The coating process was conducted in a ceramic firing crucible so to minimize the amount of agitation subjected to the coated spheres. The spheres were coated three times to achieve a uniform coating as determined by SEM. The alumoxane coated polystyrene beads were fired to 220 ◦ C for 40 min to burn off organic substituents to allow for the dissolution of the polystyrene, not the alumoxane in toluene. The pre-ceramic coated beads were stirred in toluene for 1 h and then vacuum filtered. The washing process was performed five times before firing to 1000 ◦ C to set the alumina shell. It is important to conduct multiple washes to remove all the polystyrene, because the polystyrene solution tends to “gum up” the surface of the alumina shell preclud-

ing removal of additional polystyrene. To separate the free-standing spheres from any extra alumina resulting from the coating process, the fired (1000 ◦ C) material was placed in water, centrifuged and filtered. 2.4. Formation of hierarchical membranes from pre-formed hollow α-alumina spheres Hollow ␣-alumina spheres (3 ␮m nominal diameter) prepared as described above were suspended in an aqueous solution of either A-alumoxane (1 wt.%) or MEEA-alumoxane (10 wt.%). The surface of a RefractronTM ␣-alumina support was dip coated in the resulting colloidal solution. The newly formed filter was allowed to dry overnight before firing to 600 ◦ C for 6 h with a dwell time of 5 h. 2.5. Formation of hierarchical membranes Colloidal polystyrene beads with either 0.75, 3, or 15 ␮m diameter, Polysciences Inc., were suspended in an aqueous solution of either A-alumoxane (1 wt.%) or MEEA-alumoxane (10 wt.%). The surface of a RefractronTM ␣-alumina support was dip coated in the polystyrene/alumoxane colloidal solution. The newly formed filter was allowed to dry overnight before firing to 600 ◦ C for 6 h with a dwell time of 5 h.

3. Results and discussion 3.1. Incorporation of pre-formed hollow alumina spheres onto an alumina membrane Pre-formed hollow ␣-alumina spheres (Fig. 2) were prepared using polystyrene beads of 3 ␮m diameter as substrates and coating with a 2, 5 or 8 wt.% aqueous solution of the acetic acid functionalized alumina nanoparticles (A-alumoxane). Drying at 200 ◦ C, extraction of the polystyrene with toluene, and sintering to 1000 ◦ C, resulted in polycrystalline ␣-alumina spheres with a wall thickness of ca. 1 ␮m. Previous results have shown that firing A-alumoxane to 1000 ◦ C results in an average pore size of ca. 12 nm [15]. The pre-formed hollow ␣-alumina spheres were incorporated into a ceramic membrane formed from a 1 wt.% A-alumoxane aqueous solution (see Section 2). Since the 2, 5, and 8 wt.% A-alumoxane solutions

K.A. DeFriend, A.R. Barron / Journal of Membrane Science 212 (2003) 29–38

33

Fig. 2. SEM image of hollow ␣-alumina spheres after sintering to 1000 ◦ C.

Fig. 3. SEM image of the surface of an alumina membrane incorporating hollow alumina spheres into the top layer.

produced hollow alumina spheres with good shape retention [15], spheres using these concentrations were incorporated into the membranes. A flat ceramic membrane, of approximately 1 ␮m thickness, prepared using A-alumoxane [14] on a porous alumina support, was used as a base for the macroporous membrane. The base membrane was used so as to ensure that any macroscopic holes or cracks in the macroporous membrane would not lead to failure of the asymmetric membrane. A schematic representation of the macroporous membrane is shown in Fig. 1b. The surface of a flat ceramic membrane was brought into contact with a suspension of alumina spheres in an aqueous solution of A-alumoxane solution for 2–5 s. The newly made filter was dried in air overnight before firing to 600 ◦ C. The total thickness of the membrane was designed to be comparable to the flat 2 ␮m thick membranes we have previously described [14]. SEM images of the surface show the incorporation of the hollow spheres into an alumina matrix (Fig. 3). A

summary of flow, flux, and permeability characteristics for these sphere-containing membranes is given in Table 1. The flow, flux, and permeability parameters for the membrane with the alumina spheres derived from 5 and 8 wt.% solution of A-alumoxane, show a flux comparable to the flat membrane reported previously [14]. In contrast, the flux and permeability for the membrane containing alumina spheres derived from 2 wt.% solution of A-alumoxane is comparable to the porous support (see Table 1). The concentration of the alumoxane solution determines the wall thickness of the hollow alumina spheres. The flux and permeability measurements suggest that the thicker the wall of the pre-formed alumina sphere the more restricted the flow through the spheres and/or the flow between the spheres, reducing the overall available cross-section of the membrane surface. As may be seen from Table 1, there is an inverse correlation between the permeability and the size of the spheres.

Table 1 Comparison of alumina membranes containing ␣-alumina spheres with traditional flat membrane and the porous support

A-alumoxane used for spheres (wt.%) Flow (ml min−1 ) Flux (×10−6 m s−1 ) Permeability (nm2 ) Pore volume (ml g−1 ) Surface area (m2 g−1 )

Alumina support

“Flat” alumina membrane

Membrane containing pre-formed ␣-alumina spheres

n/a 0.12 1.44 37.0 0.024 3.5

n/a 0.71 0.86 22.1 0.32 111.0

2 0.108 1.3 33.7 0.47 245.5

5 0.065 0.78 20.4 0.48 224.8

8 0.06 0.73 18.7 0.50 254.24

34

K.A. DeFriend, A.R. Barron / Journal of Membrane Science 212 (2003) 29–38

Clearly, the presence of the spheres as part of a membrane system offers a route to improved membrane performance, however, the pore size (ca. 30 nm) of these ␣-alumina spheres formed at 1000 ◦ C is significantly larger than the matrix formed at 600 ◦ C [13,14]. It would be desirable to create a similar membrane structure, but in which the entire membrane has a uniform small pore size. 3.2. Increasing flux and permeability by formation of macroporous membranes The above results suggest that the presence of designed macroporous voids can increase the flux through a ceramic membrane, but further increases in the flow rate would be desirable. One method to achieve this goal is to increase the void space present in the membrane, as well as the surface area, without sacrificing the pore size provided by the alumoxane-derived alumina [14,15]. We have investigated the one-step formation of macroporous voids within an alumina membrane by the use of the colloidal polystyrene beads rather than the alumina spheres. Polystyrene beads suspended in an aqueous solution of an alumoxane were cast onto a 1 ␮m base membrane prepared using A-alumoxane on a porous alumina support. After drying the polystyrene was removed by pyrolysis and the resulting membrane was heated to 600 ◦ C to form a macroporous alumina membrane. The pyrolysis/sintering temperature was chosen to optimize the pore size and pore-size distribution of the resulting alumina membrane [15]. A schematic representation of the resulting asymmetric membrane is shown in Fig. 1c. The identity of the alumoxane, the concentration of alumoxane solution, and the diameter of the polystyrene were investigated to determine their effects of the membrane structure and the flow/flux performance of the membrane. Aand MEEA-alumoxanes were used; the former providing a higher ceramic yield and smaller average pore size than the latter. The sizes of the polystyrene investigated were 0.75, 3.0, and 15.0 ␮m, as well as a mixture of the three. The use of higher concentrations of alumoxanes (10 wt.%) was found to be detrimental for the optimization of the flux and permeability due to the thickness of the final membrane. However, higher concentrations do allow for alignment of the poly-

Fig. 4. SEM image of polystyrene beads deposited onto an alumina membrane from a 10 wt.% MEEA-alumoxane solution.

styrene beads onto the surface of the membrane support (see Fig. 4). AFM of the colloids deposited in the alumoxane films shows that there is a difference in spacing between the spheres depending on the identity of the alumoxane solution they are dispersed in. When mixed with MEEA-alumoxane, the spheres appear to be touching (Fig. 5a), however, A-alumoxane results in the spheres separated by about 4 nm (Fig. 5b).

Fig. 5. AFM-derived surface morphology of polystyrene beads deposited onto an alumina membrane from (a) a 10 wt.% MEEAalumoxane solution, and (b) a 10 wt.% A-alumoxane solution.

K.A. DeFriend, A.R. Barron / Journal of Membrane Science 212 (2003) 29–38

Our goal was to form a nanoporous membrane with macroporous features by out-gassing the colloidal polystyrene by firing the film to 600 ◦ C, the firing temperature used to form the membrane, as previously re-

Fig. 6. SEM images of the surface of an asymmetric membrane produced by the pyrolysis of polystyrene beads in (a) 10 wt.% A-alumoxane solution, (b) 10 wt.% MEEA-alumoxane solution, and (c) 1 wt.% A-alumoxane solution.

35

ported. Although the high concentration (10 wt.%) of A-alumoxane allowed for the colloidal beads to align on the surface, they prohibit the out-gassing of the decomposition products of the polystyrene by insulating the colloids (Fig. 6a). The resulting membranes were similar in appearance to those described above and were formed using the pre-formed alumina spheres. In contrast, the use of 10 wt.% MEEA-alumoxane solutions allowed for alignment of the polymer beads, but resulted in the rupture of the ceramic spheres upon pyrolysis to give a macroporous high surface area membrane (Fig. 6b). MEEA-alumoxane has a lower ceramic yield (37%) than A-alumoxane (76%) and therefore results in a thinner more porous coating of the polymer beads that allows the volatiles to out-gas. Even though 10 wt.% MEEA-alumoxane solutions produced a membrane with the desired macroscopic features, the average pore size of a membrane derived

Fig. 7. SEM images showing the structure of the macroporous features 1 wt.% A-alumoxane with 0.75 ␮m polystyrene beads used to derive the alumina membrane.

36

K.A. DeFriend, A.R. Barron / Journal of Membrane Science 212 (2003) 29–38

Fig. 8. SEM image across the surface of an alumina membrane derived from 1 wt.% A-alumoxane and 0.75 ␮m polystyrene beads.

from this alumoxane is larger (10 nm) and has a large pore-size distribution (5–30 nm) than membranes formed from A-alumoxane. Since the thickness of a flat membrane derived from 10 wt.% MEEAalumoxane was found to be similar to those made from a 1 wt.% A-alumoxane, the latter may be used to provide uniformly small pore size and pore-size distributions. The use of A-alumoxane results in a pore size of 7 nm as determined from BET measurements. The MWCO of the filter prepared using the A-alumoxanes gave an 80% rejection of molecular weights of between 9000 and 10,000 g mol−1 , corresponding to a pore diameters of >4 nm. Since we are using a base membrane prepared from A-alumoxane it is not as important to maintain the pore size of the membrane with the macroporous features with regard to molecular weight cut-off performance. A compari-

son of flux measurements will change by altering the alumoxane. As the products from the pyrolysis of the polystyrene out-gas, the top of the coated spheres burst, resulting in a macroporous membrane in which the ceramic walls have a pore size defined by the alumoxane. The resulting “divots” resemble a honeycomb pattern (Fig. 7). An oblique view of these surface craters is shown in Fig. 8. The structured array of the macroporous structured membrane is controlled by the size of the polystyrene beads. Membranes formed with the 0.75 ␮m polystyrene beads produced an evenly distributed honeycomb array. The macropores are spaced 100 nm apart, and the shell wall thickness corresponding to 1 ␮m. As the template diameter increases (i.e. larger polystyrene beads are used), the regularity of the array decreases. Presumably this is due to either the ease of packing or the decreased quantity of beads per volume (of solution) for the larger beads. Regardless of polystyrene bead size, the final ceramic wall thickness remains constant and is defined by the identity and concentration of the alumoxane solution. Table 2 summarizes the flux and permeabilities achieved with the varied macro featured membranes. The flow, flux, and permeability are the highest for the membrane with the largest macroporous features. These membranes also exhibit the largest increase in the surface area. The membranes derived from MEEAalumoxane showed higher flow rates than those derived from A-alumoxane due to the larger average pore size (10 nm) and broader pore-size distribution (5–30 nm) of the MEEA-alumoxane-derived macroporous features.

Table 2 Characterization of macrostructured alumina membranes Alumoxane

Polystyrene (␮m)

Flow (ml min−1 )

Flux (×10−6 m s−1 )

Permeability (nm2 )

Surface area (m2 g−1 )

Pore volume (ml g−1 )

Support A-alumoxane A-alumoxane A-alumoxane A-alumoxane A-alumoxane MEEA-alumoxane MEEA-alumoxane MEEA-alumoxane MEEA-alumoxane

– – 0.75 3.0 15 Mixed 0.75 3.0 15 Mixed

0.116 0.071 0.103 0.106 0.095 0.072 0.102 0.159 0.159 0.121

1.40 0.85 1.25 1.28 1.15 0.87 1.23 1.92 1.92 1.46

37.18 22.15 32.21 32.98 29.68 22.48 31.77 49.57 49.66 37.71

3.5 111.3 267.0 265.1 272.1 285.6 218.52 231.75 333.50 202.14

0.02 0.32 0.50 0.56 0.57 0.42 0.53 0.56 0.81 0.29

K.A. DeFriend, A.R. Barron / Journal of Membrane Science 212 (2003) 29–38

4. Conclusions The efficiency, such as flow and permeability, of an ultrafiltration membrane can be improved by developing a hierarchical membrane, by increasing the surface area. This is of importance because ultrafiltration membranes have pore sizes down to 2 nm. Pore sizes of this diameter will inhibit the flow of a solution through the filter. Two methods to increase the surface area have been investigated, depositing hollow spheres in the membrane, forming a convex shaped membrane, or templating the membrane with polystyrene micro spheres forming a macroporous, concave, featured membrane. In both of these cases, the flow through the membrane increased. However, for the hollow spheres deposited into the membrane, those with the thinnest shell, increased the flow, compared to the “flat” membrane because the surface area has doubled. As the shell thickness increased, the flow through the membrane slowed to values less than the “flat” alumina membrane. As a side benefit, the synthesized hollow spheres show exceptional strength, and can be doped with various transition metals [21]. With this in mind, formation of a stronger ceramic composite can be accomplished. The macroporous templated membrane also increased the flow through the membrane by doubling the surface area of the membrane. However, the flow cannot exceed the flow of the support, therefore, it is desirable for the values to approach the support values as close as possible.

[3] [4] [5]

[6]

[7]

[8]

[9]

[10]

[11]

[12]

[13]

[14]

Acknowledgements Financial support for this work was provided jointly by the Environmental Protection Agency (EPA) and the National Science Foundation (NSF) under the Technology for a Sustainable Environment Program.

References [1] R.W. Baker, Membrane Separation Systems: Recent Developments and Future Directions, Noyes Data Corp., Park Ridge, NJ, 1991. [2] A.J. Burggraaf, in: A.J. Burggraaf, L. Cot (Eds.), Fundamentals of Inorganic Membrane Science and Technology, Mem-

[15] [16] [17] [18]

[19]

[20]

37

brane Science and Technology Series 4, Elsevier, New York, 1996, p. 21. R.G. Gutman, Membrane Filtration: The Technology of Pressure-Driven Crossflow Processes, Adam Hilger, Bristol, 1987. T. Matsuura, Synthetic Membranes and Membrane Separation Processes, CRC Press, Ann Arbor, MI, 1994. A.F.M. Leenaars, K. Keizer, A.J. Burggraaf, The preparation and characterization of high-temperature thermally stable alumina composite membranes, J. Mater. Sci. 19 (1984) 1077– 1088. A.F.M. Leenaars, A.J. Burggraaf, The preparation and characterization of alumina membranes with ultrafine pores. Part 2. The formation of supported membranes, J. Colloid Interface Sci. 105 (1985) 27–40. A.F.M. Leenaars, A.J. Burggraaf, The preparation and characterization of alumina membranes with ultrafine pores. Part 3. The permeability for pure liquids, J. Membr. Sci. 24 (1985) 245–260. A.F.M. Leenaars, A.J. Burggraaf, The preparation and characterization of alumina membranes with ultrafine pores. Part 4. Ultrafiltration and hyperfiltration experiments, J. Membr. Sci. 24 (1985) 261–270. J. Luyten, S. Vercauteren, A. De Wilde, C. Smolders, R. Leysen, Sol–gel preparation of mesoporous Al2 O3 and TiO2 membranes, Ceram. Trans. 51 (1995) 647–651. K. Lindqvist, E. Lidén, Preparation of alumina membranes by tape casting and dip coating, J. Eur. Ceram. Soc. 17 (1997) 359–366. J. Schaep, C. Vandecasteele, B. Peeters, J. Luyten, C. Dotremont, D. Roels, Characterization and retention properties of a mesoporous ␥-Al2 O3 membrane for nanofiltration, J. Membr. Sci. 163 (1999) 229–237. Y.S. Lin, K.J. de Vris, A.J. Burggraaf, Thermal stability and its improvement of the alumina membrane top-layers prepared by sol–gel methods, J. Mater. Sci. 26 (1991) 715–720. D.A. Bailey, C.D. Jones, A.R. Barron, M.R. Wiesner, Characterization of alumoxane-derived ceramic membranes, J. Membr. Sci. 176 (2000) 1–9. C.D. Jones, M. Fidalgo, M.R. Wiesner, A.R. Barron, Alumina ultrafiltration membranes derived from carboxylate-alumoxane nanoparticles, J. Membr. Sci. 193 (2001) 175–184. K.A. DeFriend, A.R. Barron, submitted for publication. R.F. Dream, How to design nano-, ultra- and microfiltration systems, Chem. Eng. 107 (2000) 84–89. R. Singh, Development and design of advanced membrane systems, Chem. Tech. 26 (1996) 46–53. T. Kawasaki, M. Tokuhiro, N. Kimizuka, T. Kunitake, Hierarchical self-assembly of chiral complementary hydrogen-bond networks in water, J. Am. Chem. Soc. 123 (2001) 6792–6800. H.-P. Lin, Y.-R. Cheng, C.-Y. Mou, Hierarchical order in hollow spheres of mesoporous silicates, Chem. Mater. 10 (1998) 3772–3776. R.L. Callender, C.J. Harlan, N.M. Shapiro, C.D. Jones, D.L. Callahan, M.R. Wiesner, R. Cook, A.R. Barron, Aqueous synthesis of water-soluble alumoxanes: environmentally benign precursors to alumina and aluminum-based ceramics, Chem. Mater. 9 (1997) 2421–2433.

38

K.A. DeFriend, A.R. Barron / Journal of Membrane Science 212 (2003) 29–38

[21] R.L. Callender, A.R. Barron, Facile synthesis of aluminum containing mixed metal oxides using doped carboxylate-alu moxane nanoparticles, J. Am. Ceram. Soc. 83 (2000) 1777. [22] A. Kareiva, C.J. Harlan, D.B. MacQueen, R. Cook, A.R. Barron, Carboxylate substituted alumoxanes as processable

precursors to transition metal–aluminum and lanthanide–aluminum mixed metal oxides: atomic scale mixing via a new transmetalation reaction, Chem. Mater. 8 (1996) 2331. [23] S.J. Gregg, K.S.W. Sing, Adsorption: Surface Area and Porosity, 2nd ed., Academic Press, London, 1982.