Membrane development for non-aqueous redox flow batteries

One of the largest challenges in the worldwide energy transformation to renewable energy is the fluctuating energy output that has to be balanced by large-scale energy storage. Redox flow batteries (RFBs) are one of the more promising approaches for such large-scale energy grids. Currently, vanadium based aqueous RFBs are the best performing, although the electrolysis of water intrinsically limits the cell potential. Also the temperature range and the redox partners are somewhat limited. For these reasons, non-aqueous RFBs were proposed and those are investigated within this thesis. Just like aqueous RFBs, non-aqueous RFBs consist of an electrolyte, which is now based on solvents, the redox active species, electrodes and a membrane, and mostly also supporting ions to improve conductivity. The often lower solubility of charged species in solvents compared to water is still a big challenge, which often necessitates the use of ligands for non-aqueous RFBs. This thesis focusses on two of the major components for non-aqueous RFBs, namely the redox-couples and membranes.

Known redox couples are either completely organic or metal-complexes. As organic redox couples often have radical intermediates, which lower the stability of the system. Thus, metal-complexes were investigated in this thesis. A copper acetonitrile complex with a bis(trifluoromethylsulfonyl)imide anion was shown to have a high solubility of 1.68 M at room temperature. Cyclic voltammetry showed that the Cu2+/Cu+ and Cu+/Cu redox couples result in a cell potential of 1.24 V. This system easily outperforms existing copper-based hybrid or aqueous systems in terms of cell potential. This results in a theoretical energy density of approximately 28 Wh L-1. With this system, a coulombic efficiency of 87 % and an energy efficiency of 44 % can be obtained at a current density of 5 mAcm-2.

As the cell potential of this all-copper system was only slightly higher than the electrochemical window of water, an RFB with different metals in each battery compartment was investigated to increase the possible cell potential. Zinc and cerium (ammonium cerium (IV) nitrate), being abundant, low-cost redox pairs, were chosen as active metals and an open cell potential of 1.86 V was obtained. Good reversibility and solubility were obtained with these. The coulombic efficiency was 94 %, while a voltage efficiency of 64 % was obtained at 2 mA cm-2.

From these two studies it was clearly evident that coulombic efficiency and voltage efficiency should be further improved and this was attempted through the development of new membranes. The organic solvent renders the chemical stability of membranes much more challenging. One way to improve chemical stability of polymers towards dissolution is through crosslinking. This approach was tested for poly (phenylene oxide) (PPO) based membranes. These were functionalized with bromomethyl groups and subsequently crosslinked with 4, 4'-bipyridine, which comes along with the generation of cationic pyridinium species which serve as anion exchange groups. At lower amounts of bipyridine (bipyridine / bromomethyl ratio=0.13 or 0.18), adequate stability at moderate ion capacity was observed, whereas high amounts of bipyridine (bipyridine / bromomethyl ratio=0.24), limited stability during operation of the non-aqueous copper RFB was observed. This might be explained by the less efficient crosslinking at higher bipyridine amounts. A RFB single cell with this crosslinked membrane yielded a coulombic efficiency of 89 %, a voltage efficiency of 61 % and an energy efficiency of 54 % at 7.5 mA cm-2. The results show that this membrane is promising for the application in RFBs.

Although anion exchange membranes are generally well suited for non-aqueous RFBs, non-charged porous membranes are more stable in organic solvents. Very recently it was shown that nanoporous membranes prepared by solvent treatment of more open structures are among the best performing membranes for aqueous RFBs. This concept of treating very open membranes with good pore interconnectivity with appropriate solvents to make pores shrink while conserving the high interconnectivity, was now applied for the first time in non-aqueous RFBs. The solvent treatment primarily shrinks the pores, while allows for efficient tuning of fluxes and separation performance. Using positron annihilation lifetime spectroscopy (PALS), it was shown that the pore size can efficiently be modified using different solvents (e.g. water or isopropanol) for polyvinylidene fluoride (PVDF) membranes. When subsequently applied as RFB membranes in an acetonitrile environment, however, there is barely any difference between membranes treated in these different solvents, as acetonitrile will re-swell the membranes, bringing the modified pore diameters back to the similar levels. This observation was confirmed by investigating acetonitrile pressure-drive permeation. They were proved to be solvent-treatment independent for PVDF, while water permeation was strongly influenced by the applied solvent treatment. Using a polysulfone (PSF) membrane, however, there is a clear difference in acetonitrile permeation. This was rationalized using Hansen solubility parameters: acetonitrile is a rather good solvent for PVDF, while it is a bad solvent for PSF. Thus, the changes that were induced by the solvent treatment in a PSF porous membrane withstand the subsequent acetonitrile treatment and proved to be a good method to tune the properties of PSF membranes for non-aqueous applications.