Transport, selectivity and antifouling in nanoscale membranes

Abstract

Desalination of seawater has been pursued since ancient times, first being recorded around 300 BC, largely to provide drinking water in ocean-going vessels. The primitive means used then depended on distilling seawater, which is an energy intensive process. In the late 1950s, reverse osmosis (RO) was demonstrated to be capable of desalinating seawater without vaporizing it. RO accounts for nearly 60% of all major desalination plants in operation today due to its lower energy consumption. The need for freshwater, however, has increased significantly in the last few decades, in towns and cities where populations have grown and water bodies diminished. Despite lower energy consumption, RO remains too energy-intensive to supply fresh water at city-scale cheaply, driving research into new RO membranes that can reduce energy consumption. A large number of new nanomaterials have emerged as a result, including carbon nanotubes (CNTs), boron nitride nanotubes (BNNTs), metal-organic framework (MOF) and zeolite membranes. These materials promise very high water permeance and salt rejection, which is expected to drive down desalination costs. Among these materials, CNTs have been considered the forerunner, promising orders of magnitude higher water permeance compared to traditional membranes. Simulations showed BNNTs could potentially provide higher permeance, but manufacturing difficulties prevented much experimental research on BNNTs. These have been recently overcome, and BNNTs once again appear as a promising membrane candidate. Molecular dynamics (MD) based simulation studies of BNNTs, however, present a few discrepancies. While some earlier studies showed permeance higher than CNTs, later studies showed CNTs had higher permeance than BNNTs. In this thesis, the analysis of the membrane transport resistance allowed the resolution of this discrepancy. Simulations studies were found to be largely conducted by modelling membranes having short nanotube lengths, where the flow resistance due to effects at the entrance and exit of the nanotubes dominates, hiding the behaviour inside the nanotubes. As the nanotube lengths are increased, the transport resistance within the nanotubes starts to increase, and in this regime, CNTs are found to have a lower transport resistance compared to BNNTs. The transport resistance within BNNTs is also found to be largely dependent on the partial charge chosen for the BN atoms in the simulation. Experimental evidence concurred with the higher resistance for BNNTs, supporting the hypothesis of a high partial charge on BN atoms, which is known to retard water flow. The high partial charge of BNNTs, while reducing water flow, provides BNNT membranes with the advantage of high selectivity. Traditional RO membranes are already operating near the thermodynamic limit of energy utilisation, indicating massive increases in permeability is unlikely to drive down energy use, and therefore costs. Instead, high selectivity is one avenue that could reduce the cost of post-treatment, where clean water from the RO membrane needs to undergo additional steps due to the inability of the membrane to selectively remove solutes. These include boron removal from desalinated seawater, and remineralization of drinking water to make up for essential minerals removed by the membrane. BNNT membranes in water have a charged surface, allowing the membrane to discriminate between cations and anions and repel anions. In this thesis, using both simulations and experiments, this charge-based rejection was shown to work alongside the size-based rejection of traditional membranes where particles are only rejected because they are too large for the membrane pores. As a result, BNNT membranes could be manufactured with larger pores, allowing up to 70% higher water flow rates for comparable rejection with reference to CNTs. The second avenue where cost reductions may be found lie in membrane cleaning. Membrane fouling and cleaning costs could nearly equal the energy costs and also result in reduced membrane life. The dominant method of membrane cleaning is through the use of chemicals, which requires the desalination plant to be put offline and creates a chemical-laden wastewater disposal problem. Chemical cleaning is therefore done infrequently, while foulant build-up on membranes is a continuous process. Physical methods of cleaning, such as by heating the membrane, appear promising and can be applied frequently without stopping the desalination process. Heating has been shown to act through biological means (by denaturing proteins) and physical means (causing shear flows and vapour expansion within pores), promising cleaning ability against a wide range of foulants. In this thesis, direct heating of membranes was found to be capable of removing nanoparticles from the membrane surface. Heating was shown to work through a mixture of bubble formation and increased Brownian motion of the nanoparticle to displace it away from the membrane surface. The large role played by Brownian forces, however, prevents a deterministic prediction of whether a particle will be removed or not when heat is applied to the surface. This work therefore deals with the three avenues of improvement in filtration and desalination membranes, i.e., permeability, selectivity and antifouling. This work resolves the question of permeability in BNNT membranes, established charge based rejection in BNNTs as the basis of a membrane capable of high selectivity and finally explores membrane heating as a promising candidate mechanism for the cleaning of membranes.