Journal of Membrane Science 357 (2010) 6–35 Contents lists available at ScienceDirect Journal of Membrane Science journal homepage: www.elsevier.com/locate/memsci Review Stimuli-responsive membranes Daniel Wandera a , S. Ranil Wickramasinghe b , Scott M. Husson a,? a b Department of Chemical and Biomolecular Engineering and Center for Advanced Engineering Fibers and Films, Clemson University, Clemson, SC, USA Department of Chemical and Biological Engineering, Colorado State University, Fort Collins, CO, USA article info abstract Stimuli-responsive membranes change their physicochemical properties in response to changes in their environment. Stimuli-responsive membranes have been designed to respond to changes in pH, temperature, ionic strength, light, electric and magnetic ?elds, and chemical cues. This review covers the design of stimuli-responsive membranes and their ever-expanding range of use. It considers stimuli-responsive changes in membrane structure and surface characteristics that enable novel applications. © 2010 Elsevier B.V. All rights reserved. Article history: Received 14 February 2010 Received in revised form 29 March 2010 Accepted 31 March 2010 Available online 9 April 2010 Keywords: Temperature-responsive pH-responsive Light-responsive Electro-responsive Magneto-responsive Membrane preparation Membrane modi?cation Ion-gating Contents 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1. General overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2. Responsive mechanisms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Design of responsive membranes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1. Membrane processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1.1. Radiation-based methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1.2. Solvent casting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1.3. Interpenetrating polymer networks (IPNs) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1.4. Phase inversion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2. Surface modi?cation using stimuli-responsive functional polymers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2.1. Surface-initiated modi?cation (“grafting from”) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2.2. “Grafting to” method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3. Gating membranes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.4. Summary—design of responsive membranes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Temperature-responsive membranes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1. Membrane formation—temperature-responsive membranes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1.1. Copolymer systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1.2. Interpenetrating network membranes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1.3. Membrane microcapsules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1.4. Nanocomposite membranes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2. Membrane modi?cation—temperature-responsive membranes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.1. Grafting to modi?cation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.2. UV photografting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 7 7 8 9 9 9 9 9 9 9 10 10 10 10 10 10 11 12 12 12 12 13 2. 3. ? Corresponding author. Tel.: +1 864 656 4502; fax: +1 864 656 0784. E-mail address: shusson@clemson.edu (S.M. Husson). 0376-7388/$ – see front matter © 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.memsci.2010.03.046 D. Wandera et al. / Journal of Membrane Science 357 (2010) 6–35 7 4. 5. 6. 7. 3.2.3. Radiation curing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.4. Radiation-induced grafting (non-UV) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.5. Plasma-initiated grafting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.6. Controlled radical grafting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.7. Ion-gating, temperature-responsive membranes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3. Summary—temperature-responsive membranes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . pH- and ionic strength-responsive membranes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1. Membrane formation—pH-responsive membranes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1.1. Copolymer systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1.2. Interpenetrating network systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1.3. Micro/nanocomposites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2. Membrane modi?cation—pH-responsive membranes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.1. Grafting to modi?cation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.2. Grafting from modi?cation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3. Ionic strength-responsive membranes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4. Summary—pH- and ionic strength-responsive membranes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Photo-responsive membranes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1. Photo-chromism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2. Membrane formation—photo-responsive membranes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3. Membrane modi?cation—photo-responsive membranes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.4. Photo-responsive carrier membranes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.5. Summary—photo-responsive membranes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Electric and magnetic ?eld-responsive membranes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.1. Electric ?eld-responsive membranes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2. Magnetic ?eld-responsive membranes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.3. Summary—electric and magnetic ?eld-responsive membranes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Future directions and conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.1. Future directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.2. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Acknowledgement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 14 15 16 17 18 18 19 19 20 21 21 21 22 24 25 25 25 26 27 30 30 30 30 31 32 32 32 32 32 32 1. Introduction 1.1. General overview The rapidly increasing interest in functional materials with reversibly switchable physicochemical properties has led to significant work on the development of stimuli-responsive membranes, for which mass transfer and interfacial properties can be adjusted using external stimuli: temperature, pH, solution ionic strength, light, electric and magnetic ?elds, and chemical cues. Of particular interest in the development of responsive membranes is the fact that the reversible changes occur locally at a fast rate and with high selectivity. Non-porous and porous stimuli-responsive membranes have a large number of already established applications and many more potential applications where they are key components in complex technical systems such as sensors, separation processes, and drug delivery devices. Enabling reversible changes in polarity or conformation, stimuli-responsive polymers generally are considered important materials (building blocks) for developing responsive membrane systems. In this review, we examine in detail the many recent contributions to the signi?cantly important and fast developing ?eld of stimuli-responsive membranes. By membrane we mean a distinct phase that separates two bulk phases. The membrane phase can be homogeneous or heterogeneous and the membrane could also be a non-porous solid (e.g. pervaporation membranes), a macroporous solid (e.g. micro?ltration membranes, membrane adsorbers) or a microporous solid (e.g. ultra?ltration membranes). Focus is given to work in the last 10 years in order to provide the reader with the current state of knowledge. We consider the already established protocols for preparing stimuli-responsive membranes and highlight some of the reported applications of these membranes. The way forward for responsive membranes is explored in terms of membrane development and potential applications in various ?elds. 1.2. Responsive mechanisms Building responsiveness into a membrane depends in part on whether the membrane has a porous or non-porous structure. Porous membranes generally are made responsive by grafting responsive polymer layers from the membrane external surface and, often, the pore walls. These functional polymers can be made to undergo changes in conformation in response to changes in the local environment, leading to reversible changes in the permeability and selectivity of the membranes. Non-porous membranes generally are made responsive by incorporating stimuli-responsive groups in the bulk of the membrane material. Conformational changes by these groups may lead to changes in the degree of swelling of the membrane barrier, hence triggering changes in the membrane permeability and selectivity. Of course, responsiveness in membrane systems is not limited to affecting a change in membrane barrier properties. Changes also may in?uence the ability of a membrane to bind and release a target compound, as needed, for example, to develop membrane adsorbers. Controlling the hydrophilic/hydrophobic behavior of a membrane surface using external stimuli can be used to reduce the level of membrane fouling and to design self-cleaning membrane surfaces. Stimuli-responsive membranes exploit the interplay among the pore structure and changes in the conformation/polarity/reactivity of responsive polymers or functional groups in the membrane bulk or on its surfaces. Such changes in specially tailored polymer systems have been used in many systems and devices to enable applications that demand reversibly switchable material properties. It follows that novel membranes can be designed using polymers/molecules that have been shown to undergo physicochemical changes in response to environmental cues. Responsiveness is known to occur as a two step process: (i) use of stimuli to trigger speci?c conformational transitions on a microscopic level and (ii) ampli?cation of these conformational 8 D. Wandera et al. / Journal of Membrane Science 357 (2010) 6–35 transitions into macroscopically measurable changes in membrane performance properties. Membrane stimuli-responsive properties can be explained based on phase transition mechanisms of the membrane materials (polymers) in controlled environments. Phase transitions may be induced by solvent quality, concentration or type of ions, temperature and other chemical or physical interactions. Polymer responsive mechanisms have been well explained in reviews by Luzinov et al. [1] and Minko [2]. Responsiveness generally refers to changes in polymer chain conformations. All polymers are sensitive to their immediate environments. They always respond to external stimuli to some extent by changing their conformation along the backbone, side chains, segments or end groups. Therefore, sophisticated membrane systems with responsive properties can be designed by variation of polymer chain length, chemical composition, architecture and topography. Most polymer responsive mechanisms are based on variations in surface energy, entropy of the polymers, and segmental interactions. Surface energy drives the surface responsive reorientation because, fundamentally, systems try to minimize the interfacial energy between the polymer surface and its immediate environment. To understand the impact of solvent quality on responsiveness, it is instructive to examine how polymer chains behave in solution. The root-mean-square end-to-end distance of a polymer chain is normally expressed as, r2 1/2 = ?(nCN )1/2 l where ? is the chain expansion factor, which is a measure of the effect of excluded volume; n is the number of freely jointed links in a hypothetical polymer chain of equal length, l; and CN is the characteristic ratio, which contains contributions from ?xed valence angles and restricted chain rotation [3]. Another way to express the above equation is by using the unperturbed (denoted by subscript 0) root-mean-square end-to-end distance: r2 1/2 = ? r2 1/2 0 The unperturbed dimensions are those of a real polymer chain in the absence of excluded volume effects, i.e., for ? = 1. In a poor solvent (? < 1), the dimensions of the polymer chain are smaller than those in the unperturbed state (? = 1). While in a good solvent (? > 1), where polymer–solvent interactions are stronger than polymer–polymer or solvent–solvent interactions, the dimensions of the polymer chain are larger than those in the unperturbed state (? = 1). So it can be said that polymers expand in good solvents and collapse in poor solvents. An example of this behavior is expansion and collapse of poly(N-isopropylacrylamide) (PNIPAAm) in water at different temperatures. At temperatures below the lower critical solution temperature (LCST), water is a good solvent, and PNIPAAm expands. When the temperature is increased above the LCST, water becomes a poor solvent, and PNIPAAm collapses. Changes in characteristic size between good and poor solvents are normally much more pronounced for surface-con?ned polymer chains than for polymer chains in solution. Thus, grafting PNIPAAm chains to a membrane surface imparts a temperature-responsiveness to that membrane. Grafting density is another parameter that affects the conformational responsiveness of polymer chains. At low chain grafting density, in the absence of strong interactions between the grafted polymer and the support surface, the response of the grafted chains to solvent quality is similar to that of the free polymer in solution. Yet, at high grafting density, the response is weaker. The explanation is that high grafting density translates to a crowded layer of already highly stretched polymer chains. At moderate grafting densities, polymers in poor solvents form clusters on the surface to avoid unfavorable interactions with the solvents. In good solvents, the polymers in this moderate grafting density region swell and form homogenous layers of stretched, tethered chains. At these moderate grafting densities, the polymer chains demonstrate a pronounced response to solvent quality. The polymers used to prepare responsive membranes need not be neutral. Polyelectrolytes (PELs) have ionizable groups, and their interactions are determined in part by the degree of dissociation (f) of these ionizable groups. Due to their high f, strong PELs generally are insensitive to solution pH. However, at high salt concentration when the ionic strength of the solution approaches that inside the PEL, electrostatic screening results in conformational changes. Weak PELs respond to changes in external pH and ionic strength and may undergo abrupt changes in conformation in response to these external stimuli. Weakly basic PELs expand upon a decrease in pH, while weakly acidic PELs expand upon an increase in pH. At high ionic strength, weak PELs tend to collapse due to effective screening of like charges along the PEL. Photo-chromic units (azobenzene, spiropyran, diarylethene, viologen) undergo reversible photo-isomerization reactions on absorption of light. Reversible photo-isomerism leads to switching between two states of the photo-chromic moieties, hence leading to molecular changes in group polarity, charge, color, and size. These molecular changes can be ampli?ed into measurable macroscopic property changes. For example, membranes containing viologen groups have permeabilities that can be regulated reversibly by redox reactions. The viologen moieties have two different redox states [4]. On treatment with a reducing agent such as sodium hydrosul?te (Na2 S2 O4 ) solution, viologens undergo reversible reduction from the dicationic state to the radical cationic state. Normally viologens in the dicationic state are highly soluble in water, but their solubility decreases in the reduced radical cation state. Therefore, in viologen grafted membranes, when the grafted viologen is in its dicationic state, the polymer chain may be expelled by the charges on the side chains and extend more in the pores leading to low permeabilities. Whereas, when the grafted viologen polymer is changed to its cationic state, the hydrophobic radical chains may be in a more entangled or collapsed state leading to higher permeabilities. Finally, while many works cited employ one responsive mechanism, the literature contains examples of membranes modi?ed by mixed polymers or block copolymers, where each polymer responds to a different stimulus. Mixed polymer brushes and block copolymers may impart adaptive/switching properties due to reversible microphase segregation among the different functionalities in different environmental conditions. For example, the individual polymers may change their surface energetic states upon exposure to different solvents. By imposing combinations of two or more independent stimuli, such membranes exhibit more sophisticated permeability responses than membranes modi?ed by a single polymer type. 2. Design of responsive membranes Increasing demand has driven the development of so-called “smart” or “intelligent” membranes that respond to external stimuli in controlled and predictable ways. In this review, we elect to call these membranes “responsive”, rather than “smart” or “intelligent”, as the latter descriptors falsely suggest that the membranes have the capacity to make decisions. Several design and production practices have been suggested, and, generally, these can be placed into two categories: (a) synthesis of stimuli-responsive materials (polymers or copolymers) and processing of these materials into membranes and (b) modi?cation of existing membranes by various chemical/physical processes to incorporate stimuli-responsive D. Wandera et al. / Journal of Membrane Science 357 (2010) 6–35 9 polymers. In this section, we highlight different methods to prepare stimuli-responsive membranes. 2.1. Membrane processing Preparation of membranes from stimuli-responsive materials has been achieved using pure stimuli-responsive polymers and copolymers or by using these polymers as components of blends or as additives during membrane formation. 2.1.1. Radiation-based methods Radiation curing has been applied to develop stimuli-responsive membranes. In this method, a mixture of stimuli-sensitive and cross-linking monomers (and/or prepolymers) is coated on the surface of a porous ?lm and the coated layer is cured with UV irradiation. The coating formulation may also include chemical additives for controlled release applications. A variety of monomers may be used to prepare composite membranes with permeation/release pro?les that respond to changes in pH, temperature, ionic strength, etc. 2.1.2. Solvent casting Casting solutions of mixtures containing stimuli-responsive polymers or copolymers onto ?at surfaces has been used to develop composite membranes that respond to different stimuli. Membrane preparation involves dissolving stimuli-responsive polymers or copolymers in an appropriate solvent, casting the solutions obtained on ?at glass plates or laboratory dishes, and allowing the solvent to evaporate. The free-standing membranes formed are dried and crosslinked by annealing them. 2.1.3. Interpenetrating polymer networks (IPNs) Interpenetrating polymer networks (IPNs) have been used as responsive membranes. The high level of crosslinking in these membranes normally leads to responsive systems with good mechanical strength. A stimuli-responsive monomer is polymerized within a physically entangled copolymer in the presence of an initiator and a crosslinker to form the stimuli-responsive IPN membrane. 2.1.4. Phase inversion Traditional membrane preparation techniques such as the wet phase-inversion process have been utilized to fabricate stimuli-responsive membranes, again by using stimuli-responsive polymers in the membrane formulation. Solutions containing stimuli-responsive polymers or copolymers are cast on ?at surfaces and then immersed in an appropriate solvent such as water to enable membrane formation. 2.2. Surface modi?cation using stimuli-responsive functional polymers Successful membrane modi?cation must satisfy two conditions: (i) preservation of the useful properties of the base membranes and (ii) introduction of functional (responsive) moieties to the membranes. Two distinct surface-selective approaches are employed for membrane functionalization. The “grafting to” approach introduces preformed, end-functionalized small molecules or large macromolecules to the membrane surface. The “grafting from” technique is a heterogeneous, surface-initiated polymerization process whereby polymer chains grow from initiator sites on the membrane surface by monomer addition from solution. Advantages and disadvantages of these two approaches have been summarized in detail elsewhere [5]. One important distinction in consideration of membrane modi?cation by grafted polymers is that the polymer chain density achievable by the “grafting to” method depends on the chain molecular weight; whereas, grafting density and polymer molecular weight are independent design parameters in “grafting from” strategies. Grafting density is an important parameter, as it greatly affects the ?nal performance of the membranes. High grafting density can be important to shield the underlying membrane support from fouling agents on one hand, but, on the other hand, high grafting density limits the response to external stimuli since the chains are trapped in an extended con?guration. Having the ability to tailor grafting densities independently of polymer chain molecular weight provides ?exibility in membrane design. 2.2.1. Surface-initiated modi?cation (“grafting from”) Modi?cation by surface-initiated polymerization normally is done in two steps: the ?rst immobilizes an initiator precursor onto the membrane surface and the second initiates polymer growth by monomer addition to the immobilized initiator sites. Numerous methods are available for grafting from modi?cation, including photo-initiated grafting (UV and non-UV), redox-initiated grafting, plasma-initiated grafting, thermal grafting, and controlled radical grafting methods such as atom transfer radical polymerization (ATRP) and reversible addition-fragmentation chain transfer (RAFT) polymerization. In general, these grafting from methods differ by the mechanism used for radical generation. 2.2.1.1. Photo-initiated polymerization. Heterogeneous photoinitiated graft-polymerization has been used often to modify membranes with polymers that respond to changes in pH, temperature, ionic strength, light, etc. A photo-initiator such as benzophenone (BP) is coated onto the membranes by dipping them in a solution containing the photo-initiator. Polymerization from the immobilized initiator sites is carried out by soaking the membranes in monomer solutions and irradiating with UV light under an inert atmosphere. The membranes are weighed before and after polymerization to determine the degree of graft polymer modi?cation (DG): DG = mgr ? m0 × msp,A m0 m0 is the initial membrane mass, mgr is the mass after graft modi?cation, and msp,A is the speci?c or areal mass (mass/outer surface area). 2.2.1.2. Redox-initiated polymerization. Chemicals such as Fenton’s reagent (Fe2+ -H2 O2 ), persulfate salts, and cerium ammonium nitrate in nitric acid are used to produce free radicals from which graft polymerization is carried out in an inert atmosphere. This method can be used to graft different stimuli-responsive polymers onto membranes leading to membranes that can respond to pH, temperature, oxidoreduction, ionic strength, light, etc. 2.2.1.3. Radiation-induced polymerization. Radiation-induced graft polymerization can be used to modify polymeric membranes with stimuli-responsive polymers to develop novel responsive membranes. In one method, membranes are immersed in aqueous solutions of the monomer with various concentrations of hydrated copper(II) sulfate, bubbled with pure nitrogen, and irradiated with 60 Co -ray radiation. 2.2.1.4. Plasma-graft-?lling polymerization. Plasma-graft-?lling polymerization has been used to modify membranes and introduce stimuli-responsiveness to them. Membranes are irradiated with argon plasma to form initiator radicals, and then polymerization is carried out from these initiator sites. 10 D. Wandera et al. / Journal of Membrane Science 357 (2010) 6–35 2.2.1.5. Atom transfer radical polymerization. “Grafting from” using surface-initiated atom transfer radical polymerization (ATRP) is gaining popularity to modify membrane surfaces. It provides for the controlled growth of polymer chains from the membrane surface. One use has been the modi?cation of membranes with polymer nanolayers that respond to various external stimuli. Modi?cation is done in two steps: immobilization of an ATRP initiator onto the membrane surface and catalyst-activated polymerization from the immobilized initiator sites. 2.2.2. “Grafting to” method Membrane surface functionalization by grafting to modi?cation with responsive polymers can be done by physical adsorption or chemical grafting of pre-formed polymer chains. The latter method reacts functional groups of the membrane material with a reactive group on the polymer modi?er. The functional groups may be inherent to the membrane material or generated (photo)chemically. The result is a permanent immobilization of responsive macromolecules onto the membrane surface. 2.2.2.1. Physical adsorption—coating. A membrane is coated with a stimuli-responsive polymer by soaking it in a polymer solution and allowing it to dry. Annealing the coated membrane helps to strengthen the attachment of the polymer to the membrane surface. Stimuli-responsive groups (additives) can also be incorporated into the thin-?lm composite polymer coating that is applied to the membrane to impart switchable properties after modi?cation. 2.2.2.2. (Photo)chemical grafting. This involves grafting pre-formed polymer chains or hydrogels onto a membrane surface. In one strategy, attaching a photo-reactive group like azidophenyl to polymer chains enables the photoimmobilization of the polymer onto the membrane surface. Immobilizing stimuli-responsive polymer hydrogels or crosslinked polymer networks onto membrane surfaces has been used as a way of modifying membranes and making them responsive to speci?c external stimuli. 2.3. Gating membranes Molecular recognition gate membranes are highly sophisticated devices. Their response is triggered by molecular recognition in macromolecular structures within the membranes. The membranes can be developed using a stimuli-responsive polymer with a receptor (e.g., crown ether) attached to it. This receptor selectively captures a speci?c ion or chemical in its cavity and this selective binding event leads to changes in the properties of the stimuliresponsive polymer. Molecularly imprinted polymers (MIPs) also have been used in the fabrication of molecular-recognition gating membranes [6–15]. MIPs are synthetic materials that possess af?nity and selectivity towards certain target molecules (templates) due to special recognition sites formed during the preparation of the polymer matrix. MIPs are prepared by copolymerization of functional and cross-linking monomers in the presence of the template molecule. Subsequent removal of the template molecules leaves behind receptor sites that are complementary to the template in shape and position of the functional groups. Molecular memory is introduced into the polymer, which becomes capable of selectively rebinding the template molecule. Again, this selective binding event leads to changes in the mass transfer properties of the membrane. 2.4. Summary—design of responsive membranes Preparation of membranes from stimuli-responsive polymers, copolymers, and polymer-additive mixtures is an important approach in the design of responsive membranes. This approach enables fabrication of membranes with the desired mechanical properties, pore structure (porosity, pore size and pore-size distribution), barrier structure (symmetric versus asymmetric), and layer thickness(es). Membrane surface modi?cation also is important in the design of responsive membranes, as the optimal required membrane surface characteristics rarely are achieved from membrane forming polymers, copolymers or polymer-additive mixtures alone. Modi?cation imparts functionality that enhances membrane performance. By taking this approach, the useful properties of the base membrane are maintained, and responsive properties are introduced to the membrane surface. When modi?cation is done using controlled, surface-initiated polymerization strategies such as ATRP, polymer molecular architecture can be controlled precisely, allowing fundamental studies on the role that surface architecture plays on membrane responsiveness and performance. 3. Temperature-responsive membranes Over recent years, applications of temperature-responsive membranes as drug delivery systems, sensors, and solute separation systems have been investigated widely by many groups. PNIPAAm is among the polymers that is well known to respond to changes in temperature and has been applied broadly to develop temperature-responsive membranes. PNIPAAm is soluble in water at room temperature, but undergoes a phase separation at temperatures higher than its lower critical solution temperature (LCST), which is near 32 ? C. Above 32 ? C, the intrinsic af?nity of PNIPAAm chains for themselves is enhanced due to thermal dissociation of water molecules from the hydrated polymer chains. Hydrophobic interactions among isopropyl groups increase, and the polymer chains associate preferentially with each other, thus precipitating from aqueous solution. Such a phase transition alters membrane structure and barrier properties. Temperature-responsive polymers can be incorporated into the membrane bulk during membrane formation or as surfacemodifying agents following membrane formation. These general strategies are described separately. 3.1. Membrane formation—temperature-responsive membranes Temperature-responsive membranes have been prepared by solution casting of copolymers and polymer mixtures, and as interpenetrating polymer networks, core–shell microcapsules, and nanocomposites. 3.1.1. Copolymer systems Nonaka et al. [16] and Ogata et al. [17] reported on the synthesis of temperature-responsive poly(vinyl alcohol) (PVA)graft-PNIPAAm) membranes by evaporating dimethyl sulfoxide (DMSO) from a solution containing the copolymer. The copolymer was prepared by graft polymerization of NIPAAm onto PVA in DMSO using potassium peroxo-disulfate initiator. The permeation of lithium ions and methylene blue through the membranes was affected greatly by changing temperature below and above 33 ? C. The swelling ratio of the membranes in water increased gradually with decreasing temperature below 40 ? C, and a considerable increase in swelling occurred at the LCST. Polyethylene glycols (PEGs) with different molecular weights could be separated by sizeexclusion with the membranes by changing the temperature from 34 to 45 ? C. Temperature sensitivity of membranes has been exploited in pervaporation processes to separate liquid mixtures with close boiling points. Sun and Huang [18] synthesized temperaturesensitive PVA-graft-PNIPAAm membranes for pervaporation of D. Wandera et al. / Journal of Membrane Science 357 (2010) 6–35 11 LCST of PNIPAAm could be applied to the separation of macromolecules. Yamakawa et al. [22] studied the transport properties of ions through temperature-responsive membranes prepared from a polymer mixture of PVA, PNIPAAm, and poly(VA-co-2-acrylamido2-methylpropane sulfonic acid). The polymers were dissolved in DMSO and the solution was cast on glass plates and allowed to dry at 50 ? C, producing free-standing membranes. The membranes were then cross-linked either physically by annealing them at 50 ? C for 20 min or chemically by immersing them in an aqueous solution of glutaraldehyde, HCl and NaCl at 25 ? C for 24 h. Permeation experiments in dialysis systems consisting of the membranes and mixed solutions of KCl and CaCl2 showed that the permeation of Ca2+ through the membranes was controlled by temperature in two distinct ways: downhill transport (from high to low concentration) occurred at temperatures below the LCST of PNIPAAm, and uphill transport (from low to high concentration) occurred at temperatures above the LCST. These membranes may ?nd use in self-regulating systems that adjust the concentration of speci?c solutes in response to temperature changes. Fig. 1. Change in reduced osmotic ?ux through anion exchange membranes with temperature for ( ) anion-exchange membrane without NIPAAm and ( ) anionexchange membrane with NIPAAm. Reproduced with permission from [20]; Copyright (2000) Elsevier. ethanol-water mixtures. The membranes were prepared by grafting NIPAAm onto PVA using hydrogen peroxide-ferrous ion as initiator. A maximum in pervaporation selectivity (?PV ) was observed for the membranes as the operating temperature was raised from 20 to 40 ? C. Maximum value of ?PV was observed at 30–32 ? C, equal to the LCST of PNIPAAm in water, indicating that the membrane separation performance was affected greatly by the grafted PNIPAAm chains. Further evidence that the PNIPAAm was responsible for the increase in ?PV was given by the observation that the ?PV values of unmodi?ed PVA membranes decreased monotonically as temperature increased. Also, the permeate ?ux increased with increasing temperature due to swelling of the polymer network and, thus, increased diffusional mobility of the permeating species. Sata et al. [19,20] prepared temperature-responsive anionexchange membranes by copolymerization of glycidyl methacrylate (GMA), ethylene glycol dimethacrylate (EGDMA), and NIPAAm using benzoyl peroxide and reaction of the epoxy groups with trimethylamine to introduce ammonium groups. The osmotic ?ux through anion-exchange membranes prepared without NIPAAm generally increased with increasing temperature due to swelling of the polymer matrix. However, for anion-exchange membranes containing NIPAAm, the ?ux decreased with increasing temperature from 25 ? C due to de-swelling and showed a constant value from 32 ? C until about 40 ? C as shown in Fig. 1. Ying et al. [21] prepared temperature-responsive micro?ltration (MF) membranes from poly(vinylidene ?uoride) (PVDF)-graftPNIPAAm copolymers by phase inversion from aqueous solution at 27 ? C. The copolymers were synthesized by thermally induced graft copolymerization of NIPAAm with ozone-pretreated PVDF. Xray photoelectron spectroscopy (XPS) analyses of the membranes revealed substantial surface enrichment by NIPAAm. Increasing the concentration of NIPAAm in the copolymers yielded membranes with higher degrees of temperature-responsive swelling in aqueous solutions. The transmembrane permeability of model drugs, calcein and ?uorescein isothiocyanate-dextran, in phosphate buffer solutions exhibited strong and reversible dependence on permeate temperature in the temperature range 4–55 ? C, with the largest change in permeability occurring from 27 to 32 ? C. The impact of temperature on the pore diameters below the 3.1.2. Interpenetrating network membranes Several groups have developed temperature-sensitive interpenetrating network (IPN) membrane systems that have been applied in various applications. Aoki et al. [23] synthesized IPN membranes from poly(acrylic acid) (PAA) and poly(N,N-dimethylacrylamide) (PDMAAm). These IPNs showed reversible and pulsatile solute release, re?ecting the “on” state at higher temperatures and the “off” state at lower temperatures. Gutowska et al. [24] prepared temperature-responsive semi-IPNs composed of PNIPAAm and linear poly(ether(urethane-urea)) (Biomer) by UV-initiated solution polymerization. The swelling ratios of the semi-IPNs decreased with increasing temperature. Muniz and Geuskens [25] studied the in?uence of temperature on the permeability of semi-IPNs based on cross-linked poly(acrylamide) (PAAm) with PNIPAAm chains entangled in the network. The permeability of Orange II dye through the semi-IPNs increased with increasing temperature, and a transition was observed at the LCST of PNIPAAm (32 ? C). Above the LCST, the permeability increased markedly because the PNIPAAm chains collapsed. Guilherme et al. [26] synthesized sandwich-like temperatureresponsive membranes of IPN hydrogels that can be used in separation processes. NIPAAm and methylene-bis-acrylamide (MBAAm) cross-linker were co-polymerized inside previously synthesized cross-linked PAAm by UV photopolymerization using periodate as sensitizer. SEM images showed that the hydrogel membranes comprised three layers, with an internal layer fully enveloped by two external layers that had similar morphologies. The images also showed signi?cant differences in the morphologies of the internal and external layers. Warming the hydrogel above the LCST of PNIPAAm showed that the PNIPAAm network was present mainly in the internal layer. This layer contracted, while the external layers expanded and remained highly porous. Unlike the work by Muniz and Geuskens [25], the permeability of Orange II dye through the layered membranes decreased signi?cantly as the temperature was increased above the LCST of PNIPAAm (Fig. 2). It was observed that the permeability decreased by 52% as the temperature increased from 25 to 40 ? C. This difference in performance between the two studies likely is due to differences in membrane structure. When PNIPAAm chains are well distributed throughout the IPN network, their collapse above LCST creates a more open ?ow structure. However, formation of a separate PNIPAAm-rich layer within the membrane structure leads to decreased ?ow above LCST due to collapse of that layer into a dense ?lm. 12 D. Wandera et al. / Journal of Membrane Science 357 (2010) 6–35 occurred primarily within the water-?lled regions in spaces delineated by polymer chains. Higher solute release rates thus occurred from PNIPAAm-grafted microcapsules at low temperatures, where the polymer ?lm was well hydrated. 3.1.4. Nanocomposite membranes Csetneki et al. [29] explored the possibility of using composite gel membranes to regulate permeability in response to external temperature change. The membranes contained ordered nanochannels that acted as reversible permeability valves. The channels were designed to contain an ordered array of stimuliresponsive core–shell beads that change size in response to external stimuli. This structure was achieved by preparing magnetite (Fe3 O4 ) nanoparticles within polystyrene latex using a seed polymerization process and coating the surfaces of the magnetic polystyrene (MPS) beads with PNIPAAm. NIPAAm monomer, MBAAm cross-linker and potassium persulfate initiator were mixed with the MPS at 65 ? C under a nitrogen atmosphere for 30 min to form MPS-PNIPAAm beads. A uniform magnetic ?eld was then applied to form arrays of the MPS-PNIPAAm beads, and the channel array structures were ?xed in place by polymerization with PVA. Solutions of bovine serum albumin (BSA) were used to test the temperature-responsive performance of the composite membranes. With increasing solution temperature, the permeability increased. The ‘on’ permeation value was approximately one order of magnitude larger than the ‘off’ permeation value. Fig. 4 illustrates the permeation mechanism through these nanochannel containing membranes. Similar permeation patterns through the membranes were obtained for other permeating species, including methylene blue and ribo?avin. Zhou et al. [30] reported on an in situ method to prepare thermo-sensitive polyurethane (TSPU)/silicon dioxide (SiO2 ) nanocomposite membranes by the wet inversion process for use in water vapor permeation. The water vapor permeability of the membranes depended on the size of the SiO2 particles. When SiO2 particles with an average diameter of 100 nm were well dispersed in the PU matrix, the nanocomposite membranes showed lower water vapor permeability at low temperatures and higher water vapor permeability at high temperature, relative to pure TSPU. At low temperatures (below the phase transition temperature of the soft segment), the nano-SiO2 particles enhance the crystallinity of the soft segment in the polymer and this reduces the free volume available for water transport. When the SiO2 content was higher than 5.0 wt%, the SiO2 particles agglomerated throughout the polymer matrix, and this decreased water vapor permeation. 3.2. Membrane modi?cation—temperature-responsive membranes As described in detail in Section 2, membrane surface modi?cation with responsive polymers can be done by physical adsorption or chemical grafting of pre-formed polymer chains (grafting to modi?cation) or by surface-initiated polymerization (grafting from modi?cation). Numerous methods are available for grafting from modi?cation, including radiation grafting (UV and non-UV), redoxinitiated grafting, plasma-initiated grafting, thermal grafting, and controlled radical grafting methods such as ATRP and RAFT polymerization. Again, these grafting from methods differ by the mechanism used for radical generation. 3.2.1. Grafting to modi?cation Yoshida and co-workers [31–33] have studied the temperaturecontrolled transport properties of ion track membranes. The membranes were prepared by chemically grafting PNIPAAm hydrogel onto single/multi-pore ion track membranes of poly(ethylene terephthalate) (PET). It was discovered that, below 31 ? C, the hydro- Fig. 2. Permeability of orange II dye through (2.5–2.5) sandwiched-like IPN hydrogel membranes and (2.5–0) pure PAAm hydrogel membranes as a function of temperature. Reproduced with permission from [26]; Copyright (2006) Elsevier. 3.1.3. Membrane microcapsules Chu et al. [27,28] prepared temperature-responsive core–shell microcapsules with a porous outer membrane and PNIPAAm gates. Interfacial polymerization was used to prepare polyamide microcapsules, and plasma polymerization was used to graft PNIPAAm into the pores of the microcapsule walls. The PNIPAAm-grafted microcapsules showed reversible and reproducible temperatureresponsive release of solute (sodium chloride or vitamin B12 ) that had been loaded into the inner space of the microcapsules. As illustrated in Fig. 3, at low graft yields, the release rate was higher at temperatures above the LCST than at temperatures below the LCST due to pore opening controlled by the PNIPAAm gates. In contrast, at high graft yields, the release rate was lower at temperatures above the LCST than at temperatures below the LCST. At high graft yields, the pores were fully blocked and a PNIPAAm layer covered the entire capsule surface. However, the grafted PNIPAAm was still highly hydrophilic and water-soluble below LCST, and dramatically became hydrophobic and insoluble in water above LCST. Since the solutes used were water soluble, diffusion through the membranes Fig. 3. A schematic illustration of the temperature-responsive release mechanism of core–shell microcapsules with a porous membrane and temperature-responsive PNIPAAm gates. Reproduced with permission from [27]; Copyright (2001) Elsevier. D. Wandera et al. / Journal of Membrane Science 357 (2010) 6–35 13 Fig. 4. Schematic representation of the permeation mechanism through the channels of MPS-PNIPAAm smart nanocomposite membranes: (a) ‘off’ state below LCST and (b) ‘on’ state above LCST. Reproduced with permission from [29]; Copyright (2006) American Chemical Society. gel took up water rapidly, which led to enormous swelling and pore blocking. The opening and closing of pores was cycled repeatedly many times over long periods by varying the temperature above and below 31 ? C [31]. The mass ?ow of various molecules with molecular weights ranging from 18 to 69,000 g/mol through the membranes could be controlled thermally, and the degree of control increased with increasing molecular weight of the permeating molecule [32]. The permeation rates of orange G, methylene blue and BSA through the membranes were controlled thermally, with the BSA permeation rate in the membrane with open pores being 35 times higher than that in the membrane with closed pores [33]. Park et al. [34] investigated the permeation rate of water and tryptophan through porous polycarbonate (PC) membranes that had been modi?ed with immobilized PNIPAAm. Photoimmobilization was carried out by attaching a photoreactive azidophenyl group onto the PNIPAAm and casting it on the porous membranes. When small amounts of polymer were immobilized, the membrane pores were covered only partially, yielding porous membranes. When large amounts of polymer were immobilized on the membranes, the pores were covered completely, yielding non-porous membranes. Water permeation through the modi?ed porous membranes changed at different temperatures; higher ?ux rates were observed above the LCST of PNIPAAm as the polymer chains contracted and the pores opened. Water permeation through the modi?ed non-porous membranes was not observed at any temperature. Tryptophan permeation through the modi?ed porous membranes decreased below the LCST of PNIPAAm; whereas, that through the modi?ed non-porous membranes increased below the LCST. This result was attributed to the fact that the non-porous (gel like) membranes swell below the LCST, thereby enhancing the rate of diffusion of tryptophan and collapsed (densi?ed) above the LCST, reducing the rate of diffusion. 3.2.2. UV photografting Peng and Cheng [35] investigated the effect of grafting yields on the temperature-responsive permeability of porous PNIPAAm-g-PE membranes. The membranes were prepared by grafting PNIPAAm onto porous PE MF membranes by UV irradiation. They observed that temperature changes affected diffusional permeation in two distinct ways depending on the grafting yield: permeability of vitamin B12 increased with increasing temperature in low graft yield membranes and decreased with increasing temperature in high graft yield membranes. For low graft yield membranes, permeability was controlled by pore-grafted PNIPAAm, and the expanded conformation of the grafted polymer below the LCST gave rise to reduced effective pore size in comparison with the collapsed state above LCST. As the grafting yield increased, polymer ?lled the pores and also formed a con?uent layer of grafted polymer at the membrane–solution interface. In higher graft yield membranes, permeability was controlled by this polymer surface layer. With increasing temperature, the layer densi?ed and became more resistant to diffusion, resulting in decreased permeability. Liang et al. [36] developed temperature-sensitive membranes by grafting PNIPAAm on the surface of hydrophilic polypropylene (PP) MF membranes by UV photopolymerization. The membranes showed reversible separation behavior by changing the external temperature. Below the LCST of PNIPAAm, the separation properties were characteristic of ultra?ltration (UF) membranes, while, above its LCST, they were characteristic of MF membranes. With changes in temperature, there was reversible swelling and shrinking of the PNIPAAm modifying layer, which caused the membrane pores to shrink or expand. The ?ux through the membranes varied from 1200 to 10,000 L m?2 h?1 for temperature changes from 22 ? C to 40 ? C. It was discovered that for solutions of dextran with molecular weights between 6.3 kDa and 2000 kDa, a marked change in separation performance occurred by adjusting the temperature. Yang and Yang [37] developed temperature-responsive membranes regulated by pore-covering polymer brushes by photografting NIPAAm onto PET MF track membranes with BP as initiator. ATR-FTIR and SEM con?rmed that the PNIPAAm brushes were grafted only on the membrane surfaces but not inside the pores. The modi?ed membranes responded to changes in temperature. Pore sizes and water ?ux varied as the PNIPAAm brushes swelled and collapsed with temperature variation. Wu et al. [38] used rapid ‘bulk’ surface photopolymerization to modify nylon MF membranes with poly(N,N-diethylacrylamide) (PDEAAm), using BP as an initiator. ATR-FTIR spectra con?rmed the successful grafting of PDEAAm on the membrane surfaces. The water ?ow rate through the membranes increased sharply with increasing temperature within the range 30–35 ? C, near the LCST of PDEAAm. Changes in surface morphology were observed with changes in temperature, as shown by the AFM images in Fig. 5. 14 D. Wandera et al. / Journal of Membrane Science 357 (2010) 6–35 Fig. 5. AFM images of dried PDEAAm modi?ed nylon membranes: (A) above the LCST and (B) below the LCST. Reproduced with permission from [38]; Copyright (2006) Elsevier. The PDEAAm chains expanded below the LCST of PDEAAm and collapsed onto the membrane surface above the LCST. Vertommen et al. [39] demonstrated the reversible on/offswitching of BSA permeation through temperature-responsive composite membranes. The membranes were prepared by UVphotografting of PNIPAAm onto PET MF membranes using BP as the photoinitiator and N,N -methylenebisacrylamide (MBAAm) as the crosslinking agent. They observed that above the LCST of the PNIPAAm hydrogel (on-state), the collapsed grafted layer appears to only partially cover the membrane pores, allowing BSA permeation through the uncovered pores. Provided the grafting density is high enough, the swollen PNIPAAm covers the membrane pores completely below LCST (off-state) thus preventing BSA permeation. Negligible permeation was measured in the off-state. The authors proposed an on-demand release mechanism that is based on switching the membrane surface coverage, rather than the effective pore size, as discovered earlier by Peng and Cheng [35]. 3.2.3. Radiation curing Kaetsu et al. [40,41] found that responsive membranes for drug permeation and release could be obtained by coating a mixture of stimuli-sensitive monomers and UV curable cross-linkable prepolymers on the surfaces of porous ?lms and drug-containing ?lms and curing the coating with UV irradiation. The porous membranes were prepared by irradiating polymeric ?lms such as PET and PC and etching them with ion beams. Mixtures of an electrolyte monomer (AA), a crosslinker such as tetraethyleneglycol dimethacrylate, and a UV curable prepolymer were then coated on the surface of the porous membranes by spraying or dipping. The coated membranes were irradiated with UV light. For immobilization of enzymes, aqueous solutions of the enzymes were added to the coating mixture. Membranes were coated with the enzyme solution and cured similarly by irradiation. Fig. 6 shows how both the drug permeation and drug release membranes were used. Nakayama et al. [42] developed membranes for pH- and temperature-responsive drug release by coating and radiation curing of polymer–drug composite ?lms with PEL (AA) or NIPAAm (NIPAAm)-containing mixtures. Fig. 7 shows how the membranes were prepared and how they were applied as stimuli-responsive systems. Mixtures of a matrix monomer such as 2-hydroxyethyl methacrylate (HEMA) and a model drug such as methylene blue were cast into molds consisting of two reinforced glass plates and polymerized by UV irradiation into drug containing membranes. The membranes were then coated by dipping or spraying them with stimuli-responsive hydrogels obtained from NIPAAm or AA, a crosslinker such as PEGDMA and a UV curable cross-linking prepolymer. Finally, the coated membranes were placed on a conveyer and irradiated continuously with UV light. Lequieu et al. [43] developed PET track-etched membranes with thermo-adjustable porosity and separation properties by surface photochemical immobilization of poly(N-vinylcaprolactam) (PVCL). Photo-reactive azidophenyl was ?rst incorporated into the PVCL chains, and then the chains were cast on PET membranes followed by irradiation with UV light. The water permeation through the modi?ed membranes increased drastically when the cloud point (Tcp ) (27 ? C) of the grafted PVCL chains was reached. Below Tcp , the immobilized polymer chains are in a swollen state, which decreases the average effective pore diameter. Above Tcp , the polymer chains collapse, resulting in more open pores and increased water permeability. The permeability of a mixture of dextran molecules also was affected strongly by the temperature. Grafted PVCL chains act as valves, regulating the barrier properties of the membranes with changes in temperature. Geismann et al. [44] reported on the photograft copolymerization of a temperature-responsive monomer (NIPAAm) from PET track-etched membranes. In order to facilitate surface-selective grafting, two variations of the photoinitiator precoating were selected: adsorption of BP to unmodi?ed PET surfaces (with moderate surface carboxyl density) and adsorption of BP carboxylic acid to aminated PET surface, prepared as described previously. Utilizing the acid–base complexation to enhance adsorption of photoinitiator to the membrane surface led to higher graft copolymerization ef?ciencies. Highly pronounced changes in permeability as a function of temperature were measured for membranes modi?ed with PNIPAAm. 3.2.4. Radiation-induced grafting (non-UV) Shtanko et al. [45] prepared track-etched membranes with controllable permeability by the radiation-induced graft polymerization of NIPAAm onto PET and PP track-etched membranes. AFM topography images of pristine PP membranes showed a smooth surface with pore diameters of about 900 nm; while the images of the modi?ed membranes in water at room temperature showed a swollen polymer structure on the surface, and the pore diameters had dropped to about 300 nm. On warming the water to 50 ? C, the polymer structure collapsed, and the modi?ed membrane pore diameters increased to about 600 nm. The changes in pore sizes with temperature also affected water permeability through the membranes. Flux through the modi?ed membranes increased as the temperature increased with the most signi?cant increase between 32 and 34 ? C, consistent with the LCST of PNIPAAm. Shtanko et al. [46] investigated the properties of polymeric track-etched membranes modi?ed by grafting poly(2-methyl5-vinylpyridine) (PMVP) and PNIPAAm. PET membranes were modi?ed by radiation-induced graft polymerization of MVP and NIPAAm to create hydrophilic and temperature-responsive copoly- D. Wandera et al. / Journal of Membrane Science 357 (2010) 6–35 15 Fig. 6. Application mechanisms of drug permeation and drug release using “intelligent” membranes. Reproduced with permeation from [41]; Copyright (2001) Elsevier. mers on the membrane surfaces. Conductometric measurements showed that the conductivity across the modi?ed membranes increased substantially in the temperature range 32–34 ? C on heating and decreased drastically in the same temperature range on cooling. Water ?ltration through the modi?ed membranes revealed a substantial increase in permeability on warming to temperatures between 30 and 34 ? C, indicating that the pore sizes of these membranes could be controlled by changing temperature. This conclusion was supported by AFM and SEM images of the membrane surfaces at different temperatures. Lin et al. [47] prepared temperature-responsive membranes by 60 Co -ray radiation-induced grafting of NIPAAm onto brominated poly(2,6-dimethyl-1,4-phenylene oxide) (BPPO) membranes. The water ?ux of these BPPO-g-PNIPAAm membranes changed instantly upon changes in environmental temperature, and changes were most pronounced at 32 ? C, the LCST of PNIPAAm. They later prepared positively charged, temperature-responsive membranes by quaternary amination of the BPPO-g-PNIPAAm membranes [48]. The amination process was carried out by immersing the BPPO-g-PNIPAAm membranes in triethylamine solution. Residual bromine atoms on the BPPO membranes reacted with triethylamine [(C2 H5 )3 N] to form –(C2 H5 )3 N+ Br? groups on the membrane surface. The permeability coef?cients of the ionic model drug, sodium salicylate, through the membranes at different temperatures increased markedly after amination. The authors concluded that the temperature sensitivity of the membranes was enhanced by amination. 3.2.5. Plasma-initiated grafting Choi et al. [49] developed a temperature-r