Sie sind auf Seite 1von 6


3 Membrane Preparation
Different methods of polymer membrane preparation have been covered in several reviews [59]. Membranes can be classified, according to their morphology as shown in Fig. 3.1. Dense homogeneous polymer membranes are usually prepared (i) from solution by solvent evaporation only or (ii) by extrusion of the melted polymer. However, dense homogeneous membranes only have a practical meaning when made of highly permeable polymers such as silicone. Usually the permeant flow across the membrane is quite low, since a minimal thickness is required to give the membrane mechanical stability. Most of the presently available membranes are porous or consist of a dense top layer on a porous structure. The preparation of membrane structures with controlled pore size involves several techniques with relatively simple principles, but which are quite tricky. Commercial membranes were already produced in Germany by Sartorius in the early 1920s. However, they had only a limited application on a laboratory






Fig. 3.1 Membrane classification according to the morphology.

Membrane Technology in the Chemical Industry. Edited by Suzana Pereira Nunes and Klaus-Viktor Peinemann Copyright 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim ISBN: 3-527-31316-8


3 Membrane Preparation

scale. The breakthrough of the membrane technology came first in the 1960s with the development of the asymmetric porous membranes by Loeb and Sourirajan [10]. The asymmetric membranes combine high permeant flow, provided by a very thin selective top layer and a reasonable mechanical stability, resulting from the underlying porous structure. An asymmetric structure characterizes most of the presently commercially available membranes, which are now produced from a wide variety of polymers. By far the most common method used in generation of asymmetric structures in membranes is the phase-inversion process. Other methods used to form pores in membranes will be discussed in the following sections. Particularly in the case of microfiltration, several techniques other than phase inversion are currently applied in the industry.

3.1 Phase Inversion

The phase-inversion process consists of the induction of phase separation in a previously homogeneous polymer solution either by temperature change, by immersing the solution in a nonsolvent bath (wet process) or exposing it to a nonsolvent atmosphere (dry process). In the thermal process [11, 12], a low molecular weight component usually acts as a solvent at high temperature and as a nonsolvent at low temperature. It is then removed after formation of the porous structure. Although the thermal process can be applied to a wide range of polymers, it is especially interesting for those with poor solubility, such as polypropylene, which can hardly be manufactured into a porous membrane by other phase-inversion processes. An isotropic microporous structure is usually formed. The isothermal phase inversion is commercially more widespread. Usually the polymer solution is immersed in a nonsolvent bath (wet process) and a solventnonsolvent exchange leads to phase separation. The polymer-rich phase forms the porous matrix, while the polymer-poor phase gives rise to the pores. The morphology is usually asymmetric, with a selective skin on the surface, as shown in Fig. 3.2. The pore structure is generated by phase separation. Phase separation in this case is mainly a liquidliquid demixing process, although solidliquid demixing may also play an additional role in systems containing a crystallizable polymer, such as cellulose acetate and poly(vinylidene fluoride). After immersion in a nonsolvent bath, the solventnonsolvent exchange brings the initially thermodynamically stable system into a condition for which the minimum Gibbs free energy is attained by separating into two coexisting phases. The predominant mechanism of phase inversion leading to pore formation and the thermodynamics involved are the subject of a fruitful and sometimes controversial discussion in the literature [1327], as well as the most probable paths in the phase diagram. A simplified diagram is shown in Fig. 3.3. Basically, the mechanism of phase separation depends on the crossing point into the unstable region. If the solventnonsolvent exchange brings the system first to a metastable condi-

3.1 Phase Inversion


Fig. 3.2 Asymmetric porous membrane.

tion (Path A), the nucleation and growth mechanism (NG) is favored. A dispersed phase consisting of droplets of a polymer-poor solution is formed in a concentrated matrix. If no additional nonsolvent influx or temperature change in the system were induced, the composition inside the nuclei would, in the early stages, be practically the same as expected at the equilibrium and would, practically, not change with time. Only the size of the droplets increases with time. If the demixing path crosses the critical point, going directly into the unstable region (Path B), spinodal decomposition (SD) predominates. A concentration fluctuation appears in the initially homogeneous system and progresses with increasing amplitude, leading to a separation into two cocontinuous phases. Here again, the polymer-poor phase will form the pores. The initial steps of phase separation, either by NG and SD can be relatively well described according to theories of phase separation. However, at later stages, both NG and SD usually progress to a phase coalescence and the final structure can only be predicted with difficulty. At least as important as the starting mechanism of phase separation is the point where the developing structure is fixed. Parallel to demixing, as the concentration of the polymer solution changes, by solventnonsolvent exchange, the mobility of the system decreases. Reasons for this may vary from physically unfavorable polymersolvent (or nonsolvent) interaction, leading to stronger polymerpolymer contacts, to vitrification of the polymer concentrated phase, as the solvent concentration decreases, and also in some cases partial crystallization. If the system gels and solidifies directly after


3 Membrane Preparation

Fig. 3.3 Mechanism of phase separation during membrane formation.

the first steps of phase separation (for instance, at t2), the membrane will have a fine pore structure, which keeps the original characteristics given by the initial demixing mechanism. If NG demixing stops during the initial stages, a morphology of closed cells would be favored. At later NG stages, the nuclei would grow and touch each other forming interconnected pores. The SD demixing would favor the formation of an interconnected pore structure from the beginning. An asymmetric structure is usually formed across the membrane since the solventnonsolvent exchange may lead to different starting conditions for phase separation at layers far from the surface. Besides the NG and SD demixing, other factors influence the morphology. The whole membrane structure usually can be classified as sponge-like or finger-like. Finger-like cavities are formed in many cases, as the nonsolvent enters the polymer solution. This macrovoid structure may contribute to a lack of mechanical stability in membranes to be used at high pressures. A combination of factors is responsible for the formation of macrovoids and this topic has been well reviewed in the literature [28, 29]. For practical purposes, the predominance of a sponge-like or a macrovoid structure can be induced in different ways. Basically, the sponge-like structure is favored by: 1. increasing the polymer concentration of the casting solution 2. increasing the viscosity of the casting solution by adding a crosslinking agent 3. changing the solvent 4. adding solvent to the nonsolvent bath. The growth of a polymer-poor phase by SD or NG is an isotropical process, which takes place as soon as the solventnonsolvent contents supply the thermodynamic condition for demixing. To understand the macrovoid formation, a quite interesting explanation was provided by McKelvey and Koros [28] as de-

3.1 Phase Inversion moving front moving front


solvent non-solvent

Fig. 3.4 Nonisotropic nucleus growth during macrovoid formation in membranes.

picted in Fig. 3.4. For that, the coupling of the (NG or SD) demixing processes with the rapidly moving front of nonsolvent must be considered. If the nonsolvent diffusion rate into the polymer-poor phase being formed exceeds the rate of outward solvent diffusion, the macrovoid formation is favored. The diffusivity of water is usually expected to be one-to-two orders of magnitude higher than the diffusivity of bulkier organic solvents. The main driving force for the nonsolvent (usually water) influx is the locally generated osmotic pressure. This could be, hypothetically, approximately 100 bar with a difference of only 5 mol% nonsolvent concentration between the initial nucleus and the approaching front. As water moves into a polymer-poor nucleus, its wall is deformed, expanding in the form of a tear. If the walls are fragile, the nucleus may rupture giving rise to macrovoids with unskinned walls. If the walls are stronger, as in the case of nuclei growing in a matrix with higher polymer concentration, the deformation can be restrained or even totally inhibited, giving rise to a macrovoid-free structure. Increasing the polymer concentration of the casting solution to suppress macrovoids has been well registered in the literature for a wide spectrum of polymers such as cellulose acetate [29], aromatic polyamide [30], and polyetherimide [31]. Other factors such as the addition of crosslinking agents that can improve the strength of the growing nucleus wall also contribute to a macrovoidfree structure. An example is the addition of amines to polyetherimide casting solutions [32]. Another way to suppress the macrovoid formation is to reduce the osmotic pressure between the nonsolvent moving front and the polymerpoor phase inside the nuclei. This can be achieved by adding nonsolvent to the casting solution or adding solvent to the nonsolvent bath. An example is the addition of dioxane to an aqueous coagulation bath for a CA dioxane solution [29]. Changing the solvent may act in different ways. Solvents with higher diffusivity across the nucleus walls would be able to leave the nucleus faster, while the nonsolvent is added in, which does not favor macrovoid formation. But even


3 Membrane Preparation

Fig. 3.5 Polyetherimide (PEI) membranes prepared from different casting solutions: (a) 17.5 wt% PEI in dimethylacetamide; (b) 17.5 wt% PEI in 5.5 wt% tetraethoxysilane and 77 wt% dimethylacetamide; (c) 15.5 wt% PEI in 28 wt% THF and 56 wt% g-butyro lactone.

more effective in suppressing macrovoid formation are solvents that increase the solution viscosity or even promote a fast gelation, making the nucleus wall stronger and resistant to deformation. Some examples of solvent influence on membrane morphology are to be found in the literature for polyetherimide [31] (Fig. 3.5) and cellulose acetate [33].