Desalination,70(1988)395-404 Elsevier Science Publishers B.V., Amsterdam-Printed
395 in The Netherlands
AND SOME PROPERTIES OF TUBULAR ALPHA ALUMINA
CERAMIC MEMBRANES FOR MICROFILTRATION AND AS A SUPPORT FOR ULTRAFILTRATION AND GAS SEPARATION MEMBRANES
R.A. Terpstra, B.C. Bonekamp and H.J. Veringa Netherlands Energy Research Foundation, ECN, P.O. Box 1, 1755 ZG PETTEN, The Netherlands
ABSTRACT The tubular ceramic membrane system prepared by us consists of a substrate which is made by means of extrusion, and a microfiltration layer which is about 30 pm thick and which is applied by a film coating technique. This two layer system can also be used as a substrate for an ultrafiltration or gas separation membrane layer which is then also applied by a coating technique. The water fluxes for different substrate-MF layer combinations are given and the influence of the substrate permeability on the total permeability is discussed. It is also shown by the permeability of some gases that the three layer system can be used for gas separation. INTRODUCTION In this paper we describe the preparation, characterization and some properties of ceramic alumina membranes. The membrane system is tubular (outer diameter = 20 mm, length = 90 cm) and consists of a porous substrate with one microporous layer on the inside of the tube. This layer has microfiltration properties but can also act as a support for an ultrafiltration (UF) or gas separation (GS) layer. As is well known, ceramic membranes have a high mechanical, chemical and thermal resistance and so they can be used in applications which involve high temperatures, agressive environments or for which cleaning of the membranes with high backflush pressures or with steam is necessary. The development of the substrate tube and microfiltration layer is carried out in close cooperation with Hoogovens Industrial Ceramics and for the work on the UF/GS layer there is a cooperation with the University of Twente.
PREPARATION Substrate tube The substrate tube acts as a support for the actual membrane layer. It has to be mechanically strong and its resistance to fluid flow must be very low.
The substrate tube is formed by extrusion of a ceramic paste. This is a well known forming method for all kinds of plastics and it is also used for the forming of ceramics, for instance in brick making. In our case the ceramic paste consists of a ceramic a-Al,O, powder, an organic and an inorganic binder, a lubricant and water. After mixing for seve ral hours, the paste obtained is put in the extrusion cylinder and after evacuation the piston forces the paste through a die. The pressure on the piston can be varied giving different extrusion velocities. A characteristicvelocity-pressure diagram is shown in fig. 1. This type of behaviour is to be 0.0
Fig. 1. Characteristic velocity-extrusionpressure diagram.
expected for (pseudo)plasticfluids. We find that for the paste of fig. 1 visual defects appear in the extruded tube at pressures of 37 bar or above. The effect of the extrusion pressure on the properties of the final tube does not seem to be very large, although at high extrusion pressures the permeability and final strength seem to increase. This can be due to a better (more dense) packing of the particles at high extrusion pressures and velocities which introduce high shear forces that can break up eventual agglomerates (ref. 1). But further research is required here. The extruded tubes are dried under controlled conditions to keep them straight, and then sintered at 1600'~ in air.
Microfiltration layer The microfiltration layer is applied at the inside of the substrate tube
using a filmcoating technique which uses a suspension of a-Al,O, powder particles in water. The thickness of the coated layer can be adjusted by adjusting the suspension viscosity. This can be done for instance by changing the solid contents of the supensions. The main problems which we have had so far in coating a smooth MF-layer on a macroporous substrate layer are the forming of pinholes and microcracks. Both problems have been overcome. To prevent the forming of pinholes, the capillary suction of the pores is suppressed. Microcracks are prevented by using other substrate material with different thermal properties. In fig. 2 it can be seen how the layer thickness depends on the speed with which the coating is applied which, apart from the suspension viscosity, is
Fig. 2. Characteristic layer thickness-velocitydiagram for film coating of the microfiltration layer using a colloidal stable a-Al,O, suspension. also a parameter to adjust the layer thickness. After coating the layer is dried under laboratory conditions and then sintered at 1200°C in air.
Ultrafiltration ana gas separation layer The Y-A1201 ultrafiltration or gas separation layer is obtained by coating with a boehmite (Y-AlOOH) sol. The sol is made by hydrolysis of aluminium secondary butoxyae at 80-90-c (ref. 2). During the coating process the sol is in
contact with the MF-layer which now acts as a substrate. Due to the capillary action of the pores in the MF-layer a membrane layer is slip cast on the MF-layer. It is found that the layer thickness increases with the square root of the contact time being in accordance with current models for slip casting. After drying the cast layer is sintered at 600'~ in air. At about 4OO'C boehmite is transformed to Y-Al,O,. The boehmite particles are thought to be platelike and stacked like bricks so that the pores are determined by the "interbrick" distances. In table 1 some X-ray data of unsupported membrane sheets of boehmite and Y-A1,OI are shown. The differences in relative intensities between the measured and the JCPDS values for powdered material imply that there is a texture in both sheets, which is in accordance with a bricklike
TABLE 1 X-ray data (obtained with Ni-filtered Cu Ku radiation) for a boehmite (I-AlOOH) sheet and a Y-AltO3 sheet (sheet surface parallel to the sample holder surface)
X-ray data for Y-AlOOH
X-ray data for Y-Al,O,
020 120 200 051 080
100 6 [ 5 7
100 65 25 30 6
220 311 222 400 440
rl: 3 100
20 80 50 100 100
stacking. For boehmite the (020) face is preferentially oriented parallel to the sheet surface, and this face is also expected to be the large surface area facet of the boehmite crystals. From the X-ray data of the sheets we also find that (020) of boehmite is parallel to (220) of Y-Al,O, after the transformation. A similar (020) texture in a boehmite sheet was also reported by Leenaars et al. (ref. 2) who have also developed the bricklike stacking theory to explain the narrow pore size distribution of the membrane. We will come back to it further on.
CHARACTERIZATIONAND PROPERTIES The pore properties of the substrate and MF-layer are characterized by
mercury porosimetry, and of the UF/GS layer by N, adsorption/desorption.In table 2 the results are given for some different types of substrates and MF-layers as well as for the UF/GS layer. In fig. 3 the distribution of the pore size of the various layers is shown. In fig. 4 a SEM micrograph is shown of a three layer system (substrate no. 5.
MF layer no. 2 according to table 2).
The top layer has a narrow pore size distribution in the UF/GS range. It will be shown further on that this layer has gas separation properties.
TABLE 2 Some pore properties of the different substrate, MF- and UF-layers. (E = porosity; r = mean hydraulic radius; L = thickness) h Substrate tube
(mm) 1 2
L (um) 26-35 30
0.002 - 3
0: log 39-42 0.122 36-45
n IOWI rhm
Fig. 3. Pore size (radius) distribution for the top layer (UF/GS), a no. 2 microfiltration layer and a no. 5 substrate (see table 2).
400 Fluxes For the substrates with and without one of the MF layers the water fluxes or permeabilities are given in table 3.
Fig. 4. SEM-micrograph of a three layer membrane system for ultrafiltrationor gas separation (substrate no. 5, microfiltration layer no. 2, table 2). Fig. 5 shows the fluxes of some gases through a three layer membrane. The permeabilities are plotted versus the mean pressure in the membrane. The intercept is a measure for the transport by Knudsen diffusion, this is the segregative flux. the slope is a measure for laminar flow which of course is non segregative (ref. 3). As was stated before, the function of the substrate is to give strength to the system. Its permeability should not lower the total permeability very much. For a MF (2 layer) system the permeability should mainly be determined by the permeability of the MF layer. Fig. 6 shows the (calculated)relation between the MF layer permeability and the total permeability for different substrates. The permeability of the MF layer is calculated by assuming a series resistance (- l/permeability)model for the two layer system. Taking into account that in general the strength of a porous material decreases with increasing permeability, the information obtained from fig. 6 can be used to choose for a certain application an optimum MF layer-substrate combination. For different MF layer-substratecombinations the influence of the substrate permeability on the total permeability is given in the last column of table 3 (compare also fig. 6).
TABLE 3 Water permeabilities for the different substrates and substrate-MF combinations
Permeability l/h mr bar
l/h mz bar
1 2 2 2 2
- 1200** 24000 24000 24000 24000
2 1 ?
1300 800 5700
37000 25000 25000 25000 25000 88500
9100 900 5000 1300
25 4 20 5
2 1 I
_ 700 900 1400 6000 go00
5: 6 25 38
Only preliminary experiments have yet been carried out using yeast suspensions. These experiments have been performed with a module containing 18 ceramic tubes (0, = 20 mm, ei = 14 mm) of 90 cm length each. A substrate no. 1 with a 20 pm MF coating no. 2 (table 2) were used in this experiment. The permeability versus time is shown in fig. 7. The permeate was always clear. The initial permeability is low compared to the figures of table 3. This is due to the presence of small a-Fe*O, particles from one of the pumps in the system which formed a thin layer on the membranes. In the experiment the pres___-____-__
Here the influence of the substrate permeability (Psubstr) on the total permeability is given in percentage, calculated from (1 - Ptot /PMF) * 100%. The permeabilities can be related as follows, using a series resistance model: l/Ptot = l/Psubstr + l/PMF.
+* Commercially obtained substrate.
Fig. 5. The permeabilities of H,, He, N, and CO, through a three layer system (qbstrate no. 5, MF layer no. 2, table 2) as a function of the mean pressure (p) in the membrane.
Fig. 6. The calculated relationship between the permeability (expressed in l/h m2 bar) of the microfiltration layer and of the total system for different substrate permeabilities.
403 sure across the membrane was 3 bar and the cross flow velocity was 2.9 m/s. We have also started to apply cleaning by back flushing, with pressure differences up to 5 bar. But also these experiments are yet quite preliminary, and no characteristic results can be given.
Fig. 7. Decrease of permeability with time in a cross flow microfiltration experiment using a 3 weight 1:yeast suspension. The pressure difference across the membrane is 3 bar, the cross flow velocity is 2.9 m/s. CONCLUSIONS Substrate tubes with different properties were fabricated and were used to support microfiltration layers with different properties. The influence of the resistance of the substrate tube on the permeability of the system is calculated. These results can be used to choose the optimum MF layer-substrate combination taking into account that in general the strength of the substrate decreases with increasing permeability. It is also shown that the substrate-MF system can be used as a carrier system for an ultrafiltration or gas separation membrane. The separation or enrichment in case of a gas mixture occurs due to a difference in Knudsen diffusion which means that the larger the difference in molecular masses of the components of the mixture the higher the separation factor will be. The performance of the membrane system will be tested extensively in the near future. Only preliminary tests have been carried out so far.
This study is financially supported by Hoogovens Groep B.V. and the Dutch Ministry of Economic Affairs (via the programme PBTS).
N.McN. Alford, J.D. Birchall and K. Kendall, High-strength ceramics through collodial control to remove defects, Nature vol. 330 (1987), p. 51-53.
A.F.M. Leenaars. Thesis University of Twente. The Netherlands (1984).
R.A. Terpstra, B.C. Bonekamp. H.M. van Veen. A.J.G. Engel, R. de Rooy and H.J. Veringa. Preparation and properties of tubular ceramic Al,O, membranes for gas separation, presented at Science of Ceramics 14, Canterbury, UK (1987).