A thin porous ceramic membranemade by a method quite similar to that proposed previously2] was used for gasphase separation of alcohol/water gaseous mixtures (methanol/water, ethanol/water and isopropanol/water systems) at their normal boiling points. The results show that this kind of ceramic membrane is quite efficient for bypassing the azeotropic point in alcohol/water separation and also for further enrichment of alcohols. The separation factor observed for these systems was larger than 5.

In the previous work1} a method was proposed to make a thin ceramic membranefor gas separation. Later it was found that the membrane could be improved by keeping it in hot humid air for a few days after being treated with a dilute aqueous solution of sodium silicate, and that this kind of membrane is fairly good for alcohol/water separation. The separation factor was found in the previous work2) to be dependent on the alcohol concentration, and the smallest value was about 5 for methanol/water, 8 for ethanol/water and 1 1 for isopropanol/water gaseous mixture.The previous results anticipated a possibility of gas separation by a mechanismsomewhatlike molecular sieving with this kind of ceramic membrane. It was found that the previous, rather small separation factor at high alcohol concentrations was mainly due to some large pinholes left in the membrane. In this work an improvement was made to the membrane, which was used for separation of gaseous alcohol/ water mixtures to obtain relatively large separation factors.1. Experimental Ceramic membraneA coarse porous ceramic cylinder (Multipoaron supplied by Mitsui Grinding Wheel Company) was used as the substrate to support a thin ceramic membraneon its outer surface. The ceramic cylinder was cut to 150-mm length. One end of the substrate was connected to a glass tube and the other was closed with alumina cement. As mentioned in the previous work,1} the substrate was dipped in an alumina sol and was dried at 80°C. Then it was fired at around 450°C for about 20 minutes. These procedures were repeated several times to form a quite thin alumina membraneof about 10 /mi in thickness.Then aluminum isopropoxide dissolved in an organic solvent was used to makethe pores smaller. It was finally treated with sodium silicate aqueous Received July 3, 1985. Correspondence concerning this article should be addressed to M. Asaeda.84 solution, which gave stability to the membrane.Most of the membrane is made of bohmite, but the surface of pores can be considered to have a structure of silica-alumina, though this could not be clearly detected by an X-ray analyzer. One of the most important steps in these procedures was found to be the first simple step of making a porous thin alumina membrane.Somelarger pinholes after this step were found to have a large effect on the performance of the final membrane. These relatively large pinholes were filled by pressuring the alumina sol into the membrane module through the glass tube while it was dried from outside.

The authors previously proposed a method of making a thin porous ceramic membranesupported on a coarse porous solid for gas separation. The membrane was applied to dehumidification of air in the previous work,1} in which leakage of air was found to be fairly large at relatively high temperatures and at low relative humidity.In the present work, the performance of an improved ceramic membraneused as a dehumidifier was studied experimentally and the separation mechanism was considered. ExperimentalThin porous membranes of silica-alumina were formed at the outer surfaces of coarse porous ceramic cylinders by the methods proposed in the previous works.1~3) Some relatively large pin-holes in the preliminary step of the membrane-making procedures were found to have a large effect on performance of the final membrane. These large holes were shown in the previous work3) to be filled to some extent by pressuring alumina sol into the porous cylindrical membrane module while it was dried. A membrane module (module A) used in this work was prepared by the method described in the previous work.3) Another module (module B) was made in the same way as module A except for being kept in a dilute aqueous solution of sodium silicate at 100°C for about 10 min just before its washing in boiling water.The experimental apparatus and procedures are quite similar to those described in the previous work.1* Only two small improvements were added to the apparatus to control and to measure the inside pressure of the membranemodule during dehumidification measurements. Permeabilities of pure gases in membrane moduleAwere also measured. 2. Results and Discussion In Fig. 1 are shown some observed results of apparent permeabilities of gases with membrane module A and its substrate. The observed permeability of Received July 18, 1985. Correspondence concerning this article should be addressed to M. Asaeda. 238helium is reduced to nearly half that reported in the previous work,2) where some relatively large pinholes remained in the membrane.If the flow mechanism is considered to diverge from Knudsen flow at around 200kPa, the pore diameter must be about 10 nanometers, which would not have given the large separation factors (larger than 60) for alcohol/water gaseous mixtures in the previous work3) and the quite small air leakage in the dehumidification of air mentioned below. The reason for this divergence of flow mechanism from Knudsen's near atmospheric pressure is probably due to the pore structure of constriction.The dotted lines in this figure show the estimated Knudsen permeabilities assuming that helium is nonadsorptive and that the flow mechanismis Knudsen's for each gas. The observed permeabilities for nitrogen, argon, propane and propylene are larger than

No abstract