Evaluation of membranes for the novel vanadium bromine redox flow cell

Transcription

1 Journal of Membrane Science 279 (26) Evaluation of membranes for the novel vanadium bromine redox flow cell Helen Vafiadis, Maria Skyllas-Kazacos Centre for Electrochemistry and Mineral Processing, School of Chemical Engineering and Industrial Chemistry, University of New South Wales, Sydney 252, Australia Received 24 October 25; received in revised form 12 December 25; accepted 12 December 25 Available online 3 January 26 Abstract The novel vanadium bromide (V/Br) redox flow cell employs a vanadium(ii)/vanadium(iii) couple in the negative half-cell and a Br /Br 3 couple in the positive half-cell, using a mixture of vanadium bromide, hydrobromic acid and hydrochloric acid as the electrolyte in both half-cells. An ion exchange membrane separates the two V/Br half-cell electrolytes. In this project, small test cell cycling experiments were conducted to evaluate the performance of a range of ion exchange membranes in the V/Br redox flow cell. The features and charge/discharge results of the V/Br redox flow cell will also be discussed in this paper. 25 Elsevier B.V. All rights reserved. Keywords: Redox flow cells; Vanadium bromide redox cell; Vanadium electrolytes; Ion exchange membranes 1. Introduction Batteries and fuel cells are regarded as being among the most efficient devices for the conversion, storage and delivery of electrical energy [1]. A redox flow cell, also referred to as the regenerative redox fuel cell, is an energy conversion device that directly converts the chemical energy of soluble reactants into low voltage dc electricity. Energy is stored or released by means of a reversible electrochemical reaction between two redox couple solutions with different electrochemical potentials, physically separated by an ion exchange membrane. The vanadium redox flow cell, pioneered at the University of NSW, employs vanadium(ii)/vanadium(iii) and vanadium(v)/vanadium(iv) redox couples in sulphuric acid supporting electrolyte, in the negative and positive half-cells, respectively [2 5]. It possesses a number of features, which makes it superior to a number of other energy storage systems. The major advantage of the UNSW vanadium redox battery (VRB) is that it overcomes a major problem with other redox flow batteries (e.g. Regenessys and the Fe/Cr system) in which diffusion of ions through the membrane causes the cross-contamination of the positive and negative half-cells [5]. Cross-contaminated electrolytes must be replaced or regener- Corresponding author. Tel.: address: m.kazacos@unsw.edu.au (M. Skyllas-Kazacos). ated by external treatment [5]. In the VRB, this problem is overcome by the use of the same element, vanadium, on each side of the cell. If cross-mixing were to occur, the solution is readily regenerated upon charging. Occasional mixing of the two discharged solutions is used to regain any capacity loss caused by solution imbalance. Electrolyte service life is thus indefinite, eliminating problems of waste disposal. Despite its considerable advantages, the relatively low specific energy of the VRB, less than 25Whkg 1, currently limits its use in some stationary applications as well as in electric vehicles. The specific energy of a redox flow system is related to the concentration of redox ions in solution, the cell potential and to the number of electrons transferred during discharge per mole of active species. The concentration of vanadium in the vanadium redox flow battery is limited to M due to its solubility within the temperature ranges of 45 C [6]. In an attempt to increase the energy density of the system, other redox couples have been investigated. A vanadium chloride/polyhalide redox flow battery was recently described [7] employing the V(II)/V(III) redox couple in the negative and a Br /Br 2 Cl in the positive half-cell electrolyte. While this showed promising performance, the movement of ions across the membrane led to cross-contamination and water transfer problems. This was overcome by adding vanadium bromide to both half-cells [8,9]. The vanadium bromide (V/Br) redox flow cell thus employs a vanadium bromide solution in a mixture of hydrobromic and /$ see front matter 25 Elsevier B.V. All rights reserved. doi:1.116/j.memsci

2 H. Vafiadis, M. Skyllas-Kazacos / Journal of Membrane Science 279 (26) hydrochloric acid as its supporting electrolyte in each half-cell, with electron transfer taking place at carbon felt electrodes. As with the VRB, the V/Br cell employs the same elements in both half-cell solutions so there are no problems of crosscontamination. In this system however, the V(IV) ions on the positive side do not partake in the electrochemical reaction, i.e. the oxidation of V(IV) to V(V) during cell charging does not occur. Instead, the following overall reactions result [7]. At the negative electrode: V 3+ (aq) Charge + e V 2+ (aq) Discharge At the positive electrode: 2Br (aq) + Cl (aq) Charge Discharge ClBr 2(aq) + 2e The V/Br redox system thus retains all the advantages of the original vanadium redox flow cell, but the higher solubility of vanadium bromide means that it also has the potential to produce higher energy densities. The possibility of employing 3 4 M vanadium solutions and the excess bromide ions in the positive half-cell means that the energy densities can be effectively doubled. The ion exchange membrane has been identified as being the most critical component of a redox flow cell, defining its performance as well as its economic viability. It acts as a separator, preventing the mixing and direct chemical reaction of the oxidant and reductant electrolytes. In addition, it provides an ionic conduction path between the electrolytes [1]. The ideal membrane should be selective to the charge-carrying ions in order to provide high ionic conductance [11]. In the V/Br flow cell, the membrane must restrict the transfer of vanadium and polybromide ions from one half-cell to another without hindering the transport of the charge-carrying hydrogen ions. It must also be able to withstand attack from the highly oxidising polybromide ions and be relatively inexpensive. In this study, various types of experimentally and commercially available ion exchange membranes were assessed Fig. 1. Schematic diagram of the components of a half-cell. to establish their suitability for use in the V/Br flow cell. The initial behaviour of the ion exchange membranes during small-scale cell cycling experiments was investigated. Selected membranes were subsequently evaluated for properties of ion exchange capacity, conductivity, diffusivity, dimensional stability and water content. 2. Experimental The membranes evaluated in this study are listed in Table Cell performance Cell cycling was performed at a constant current density of 2 ma cm 2 using a Repower Battery Testing Unit (China). Unless otherwise specified, each cell was charged to 1.65 V and discharged to.25 V. The electrolyte solutions used in the cell consisted of 2 M V 3.5 (1 M V(III)+1M V(IV)) in a supporting electrolyte of 6.4 M HBr and 2 M HCl. Fig. 1 shows the overall setup of one half-cell. Each cell employed three, 5cm 5cm.3 cm graphite felt electrodes in each half-cell and copper plates covered by glassy carbon sheets as current collectors. The glassy carbon was used to ensure the copper does not come into contact with the electrolyte. Iwaki pumps (Japan) circulated the electrolyte through the cell via the flow frames with a 5 mm cavity that contained the carbon felt porous Table 1 Membranes studied Membrane Supplier Type Thickness (mm) SELEMION HSV ASAHI Glass Company, Japan Cation.12 SELEMION HSF ASAHI Glass Company, Japan Cation.12 Hipore TM Asahi Kasei, Japan Microporous separator.62 NEOSEPTA CM-1 Tokuyama Corp., Japan Cation.12 NEOSEPTA AM-1 Tokuyama Corp., Japan Anion.12 ABT-1 Australian Battery Tech. & Trading Cation.14 ABT-2 Australian Battery Tech. & Trading Cation.16 ABT-3 Australian Battery Tech. & Trading Cation.2 ABT-4 Australian Battery Tech. & Trading Cation.4 ABT-5 Australian Battery Tech. & Trading Cation.6 HZ cation LinAn Chemical, China Cation.42 HZ anion LinAn Chemical, China Anion.42 SZ Guangzhou Delong Technologies Pty Ltd, China Cation.13 Gore Select L1854 W.L Gore & Associates, USA Cation.3 Gore Select M4494 W.L Gore & Associates, USA Cation.4

3 396 H. Vafiadis, M. Skyllas-Kazacos / Journal of Membrane Science 279 (26) electrodes. Rubber gaskets were placed between cell components to prevent leakage. The membrane was placed between the two flow frames to separate each half-cell. Two PVC end plates were bolted together to compress the cell components. Clear polyethylene tubing was used to connect the solution reservoirs to the pumps and the cell. It must be noted that all membranes were pre-treated according to their manufacturing specifications. In addition, each membrane was soaked in the vanadium bromide electrolyte for a minimum of 2 h before cycling to allow equilibrium to be established. The performance of each membrane including chemical stability in the battery electrolyte, cycling behaviour, as well cycling efficiencies was evaluated. Efficiencies were calculated using the following equations: η c = η e = m 1 1 n 1 1 m 1 1 n (I j + I j+1 )(t j+1 t j ) 2 (I k + I k+1 )(t k+1 t k ) 2 (V j + V j+1 )(t j+1 t j ) 2 (V k + V k+1 )(t k+1 t k ) η e = η c η v (3) where η c is the coulombic efficiency, η v the voltage efficiency, η e the energy efficiency, I j and I k are the discharge and charge currents respectively, t j and t k are the discharge and charge times respectively, V j and V k are the discharge and charge voltages respectively, m the total number of data points in the discharge and n is the number of data points in the charge Membrane properties Ion exchange capacity (IEC) The IEC of the membranes was determined by the method described in Mohammadi and Skyllas-Kazacos [12] Conductivity The ionic conductivity of the membranes was determined by impedance spectroscopy over a frequency range of 1 to 1 6 Hz with AC amplitude of.2 V, using a Solartron 1255B FRA in tandem with a Solartron Electrochemical Interface SI1287. Measurements were conducted in the area resistance test cell shown in Fig. 2. The membrane was secured between two rubber gaskets containing a.985 cm diameter hole with silicon glue and then placed between two half-cells. The cell was held together with a G-clamp to prevent leakage. Fifty millilitres of a 2 M V 3.5 in 6.4 M HBr, 2 M HCl solution was placed in each half-cell. Two graphite discs, covered with epoxy to produce a single side with constant surface area, were used as electrodes. One steel rod was inserted into each disc and joined with silver epoxy resin to provide electrical contact. The electrodes were held at a fixed distance apart and a constant depth of immersion by an electrode holder designed to fit over the cell. It must be noted that each membrane was soaked in a solution of 2 M V 3.5 in 6.4 M HBr, 2 M HCl solution for 24 h before (1) (2) Fig. 2. Conductivity test cell. testing. The resistance of the conductivity cell with and without the membrane was taken at the frequency where the phase shift was minimal. The area resistivity of the membrane was thus given by R = (r 1 r 2 ) A (4) where R is the area resistance of the membrane ( cm 2 ), A the membrane area (cm 2 ), r 1 the resistance of the cell with the membrane ( ) and r 2 is the resistance of the cell without the membrane ( ). It must be noted that the thickness of the membrane is not taken into account in the calculation of R Vanadium ion diffusion tests The diffusivity of the membrane was determined by the rate of diffusion of vanadium (IV) ions across the membrane using the static diffusion test outlined in Grossmith et al. [14]. A25cm 2 area of membrane was exposed to 5 ml of a 1 M V(IV) in 6.4 M HBr, 2 M HCl solution on one side of the cell and a 1 M ZnCl 2 in 6.4 M HBr solution on the other. The 1 M ZnCl 2 was used to equalise the ionic strength of the solutions and minimise the osmotic pressure effects on the rate of V(IV) diffusion. The Zn 2+ ions were found not to interfere with the V(IV) spectra. A wavelength of nm has been identified as corresponding to the absorbance of V(IV) ions in sulfuric acid [13]. Fig. 3 shows the absorbance of solutions containing different concentrations of V(IV) ions in the 6.4 M HBr, 2 M HCl mixtures until a peak absorbance of 1. The peak Fig. 3. Absorption spectra of (a).1 M V(IV), (b).2 M V(IV), (c).3 M V(IV) and (d).5 M V(IV) in 6.4 M HBr, 2 M HCl solution.

4 H. Vafiadis, M. Skyllas-Kazacos / Journal of Membrane Science 279 (26) In addition to the interruption test cell, the diffusion coefficient of the V(IV) ions in solution was calculated to establish whether the diffusion of ions through the membrane is slower than the diffusion of ions through the bulk. Cyclic voltammograms were thus performed over a range of scan rates using a graphite counter electrode and a saturated calomel electrode as the reference. Fig. 4. Plot of V(IV) concentration vs. peak absorbance taken at a wavelength of 748 nm. absorbance of the V(IV) ions was found to be 748 nm. Fig. 4 shows the linear plot of peak absorbance at 748 nm versus V(IV) concentration. The rate of V(IV) ion diffusion was determined by the rate of change in the absorbance according to the following derivation [14]: ln[absb 2absA] = ln[absb ] 2k s At/V A (5) where absb is the initial absorbance of solution B (the 1 M V(IV) solution), absa the absorbance of solution A (the 1 M Zn 2+ solution containing no V(IV) ions), A the area of the membrane exposed, t is time and V A is the volume of solution A. The mass transfer coefficient (k s ) of the V(IV) ions across the membranes was calculated from the slope of a plot of ln[absb 2absA] versus t. The linear plot produced has slope equal to 2k s A/V A from which k s can be determined. The diffusion coefficient, D was thus calculated using the following equation: D = k s dy (6) where y is the thickness of the membrane. It has been suggested that the behaviour of a system in which an ion exchange membrane separates two solutions is not exclusively determined by the processes occurring within the membrane [15]. The transfer of ions from one bulk solution to another may be predominantly controlled by either membrane diffusion (diffusion within the membrane) or film diffusion, from concentration gradients within the solution diffusion layer on either side of the membrane. In the bulk solutions, concentration levels may be kept constant by stirring, however, even violent agitation may be insufficient to overcome film diffusion [15]. To distinguish between membrane and film diffusion controls, an interruption test was conducted where two diffusion tests containing the same membrane were run simultaneously. Once an absorbance of approximately.5 was reached, the solutions contained in one cell were removed and were returned to the cell 3 s later. Absorbance measurements resumed until an absorbance of one (as previously). No change in the diffusion rate would indicate that the diffusion was predominantly controlled by membrane processes, whilst an increase in diffusion rate would imply film diffusion control Water content A sample of each membrane was immersed in distilled water for 24 h (membranes that were stored wet were rinsed with distilled water before being immersed). Each membrane was then removed, patted dry and placed on a piece of filter paper. The membrane weight was measured until a stable weight was established (i.e. until the surface of the membrane was dry). Each membrane was then dried under vacuum at 6 C for and weighed periodically until a constant weight had been obtained. The water content of the membrane is taken as the percent weight change before and after vacuum drying, i.e. weight percent of water on dry weight according to the following equation: W c = W wet W dry W dry (7) Dimensional stability To determine the dimensional stability of the ion exchange membranes, samples were immersed in solutions of distilled water, 2 M V 3.5 in 6.4 M HBr, 2 M HCl and in 6.4 M HBr, 2 M HCl for 15 days. Physical changes of length, width and thickness were determined using a digital caliper. 3. Results and discussion 3.1. Cell performance A summary of the cycling performance of the membranes tested is given in Table 2. In general, the membranes that cycled successfully were cation exchange membranes. Very few anion exchange membranes were however examined and thus the effect of exchange functionality is not yet apparent. As the charge-carrying ions are the hydrogen ions, it was expected that cation exchange membranes would give the lowest resistance losses during battery cycling. During cycling, a preferential transfer of solution from the positive to the negative half-cell was observed with most membranes, the rate of which was dependent on the type of membrane. As a result, the half-cell with the lower volume was charged/discharged first, causing a decrease in the capacity of the system. To recover the capacity of the system, the re-balancing of the solution levels was required periodically. During the cycling of the VRB, volumetric transfer occurred in the opposite direction, as transfer from the negative to the positive half-cells was found to occur with cation exchange membranes, while transfer from the positive to the negative was observed with anion exchange membranes [16 18]. As seen in Table 2, a number of membranes, namely the SELEMION HSV, NEOSEPTA CM-1 and AM-1, HZ cation and anion and the SZ, showed a very high resistance during

5 398 H. Vafiadis, M. Skyllas-Kazacos / Journal of Membrane Science 279 (26) Table 2 Summary of results obtained from the cell cycling experiments conducted Membrane No. of cycles Average coulombic efficiency Average voltage efficiency Average energy efficiency Cycling behaviour HSV 9 61% 39% 24% Rapid capacity fade. Membrane surface blistered. Colour change from brown to light brown HSF 14 84% 6% 5% Low resistance, large transfer of solution from + to side. Surface blistered, opaque texture lost with browny/red staining on + side, blue/green on side CM1 N/A N/A N/A High resistance, no cell cycling possible. No physical changes evident AM1 N/A N/A N/A High resistance, no cell cycling possible. No physical changes evident ABT1 N/A N/A N/A Cell charged for triple theoretical time and did not reach upper limit Half the solution crossed over from the + to side. Two outer layers detached from inner ion exchange layer ABT2 4 38% 7% 27% Two outer layers detached from inner ion exchange layer ABT % 72% 56% Solution transfer from + to side required solution remixing. Unusual discharge curve. No physical changes evident ABT4 2 (.5 A) 63% 64% 36% Low coulombic efficiency. No physical changes evident ABT5 3 45% 66% 3% Low coulombic efficiency. Capacity decreased by factor of 1 over first 1 cycles. No physical changes evident HZ cation 2 (.5 A) 83% 43% 36% High resistance, upper limit increased to 1.75 V in attempt to improve efficiency. Membrane attacked by electrolyte, causing it to soften and tear. Colour change from blue to green HZ anion 4 (.5 A) 9% 43% 36% High resistance, upper limit increased to 1.75 V in attempt to improve efficiency. Membrane attacked by electrolyte, causing it to soften and tear. Colour change from yellow to green on negative side, brown on positive SZ 42 91% 44% 39% High resistance, solution transfer from + to side required remixing every 1 cycles. Slight colour loss black to red L % 62% 53% Solution transfer from + to side required solution remixing. When transfer significant spike appears on charge. Membrane blistered M % 77% 66% Solution transfer from + to side required solution remixing. Membrane blistered Hipore TM 5 6% 63% 38% Initially, frequent solution transfer from + to side requiring remixing. Charge times greater than theoretical 5 8% 7% 56% cycling, characterised by voltage efficiencies less than 5%. The resistance of the NEOSEPTA was so high at a current of 1 A that cycling was unsuccessful. The conductivity of a membrane is determined by factors such as membrane thickness, the concentration of fixed ionic groups, the degree of cross-linking, the size and valence of the counter ions [15]. Although all mentioned membranes are of different composition, a relationship between the resistance and membrane thickness was observed, as the thicker membranes (range mm) showed the higher resistances. The Hipore TM microporous separator was also subjected to cell cycling experiments. It has been cycled for over 1 cycles and is still continuing. Over the first 5 cycles, an average coulombic efficiency of approximately 6% and average voltage of 63%. Preferential crossover of solution from the positive to the negative half-cell was occurring more often than other ion exchange membranes tested. As a result, the cell required more frequent solution re-balancing. The charge times recorded were consistently over a third greater than theoretical, indicating that the vanadium ions of different oxidation states were diffusing through. This low ion selectivity was expected due to the porous nature of the separator and its lack of ion exchange functionality. At cycle number 5, the cell was pulled apart and reassembled with fresh solution. No physical changes were observed for the separator. During the subsequent 5 cycles, the performance of the cell improved, as average coulombic and voltage efficiencies recorded increased to 8% and 7%, respectively. It should also be mentioned that the solution levels did not require re-balancing during the last 5 cycles. At this stage, it is not clear what caused the sudden change in behaviour of the cell, but further studies are underway to optimise the separator properties. The most promising ion exchange membranes tested in this study were the ABT3, L1854, M4494 membranes and the results obtained thus warrant further investigation. Fig. 5 shows Fig. 5. A typical set of curves for a small-scale cell using the ABT3 membrane. Charge/discharge current of 1 A with 35 ml of solution in each half-cell.

6 H. Vafiadis, M. Skyllas-Kazacos / Journal of Membrane Science 279 (26) Fig. 6. Rate of V(IV) ion diffusion of the ABT3 membrane in a static test cell and an interruption test cell. a typical charge/discharge curve taken from the cycling results of the ABT3 membrane. Average coulombic and voltage efficiencies for the ABT3 membrane are 8% and 72%, while corresponding values for the L1854 membrane are 86% and 62%, respectively. Although the M4494 membrane has not been subjected to the same number of cycles as the ABT3 and L1854 membranes, its average coulombic and voltage efficiencies are 86% and 78%, respectively. At this point it must be noted that the values cited in this section represent the results obtained from small-scale tests. They do not take into account any resistance losses from, for example, contact resistances, which can be eliminated or greatly reduced with improved cell design. It should also be mentioned that the 5 mm half-cell cavity used in the membrane tests cell, is greater than what would be employed in an optimised cell design. Considerable improvements in voltage efficiency, and thus the overall energy efficiency, can be expected with a reduced half-cell cavity thickness (e.g. 2 mm) and a single carbon felt electrode (3 mm thickness). In an attempt to understand the behaviour of the membranes during cycling, selected samples were analysed to determine their properties Membrane properties Fig. 6 compares the rate of diffusion of the V(IV) ions through the ABT3 membrane in a static test cell and an interruption test cell. The interruption of the diffusion test cell did not affect the rate of V(IV) ion diffusion through the membrane. It can thus be assumed that the diffusion processes are predominantly controlled by membrane diffusion. To verify this, the diffusion coefficient of the V(IV) ions in the vanadium bromide electrolyte was determined using cyclic voltammetry. The cyclic voltammograms of the 1 M V(IV) solution at different scan rates and switching potentials of 1. and.1 V are shown in Fig. 7. The reverse peaks of V(IV)/V(II) were used in the analysis. Fig. 8 shows a linear plot of the peak current versus the square root of scan rate, indicating that the reaction is diffusion and not membrane controlled. The linear slope obtained was used in the equation for irreversible reactions shown in the following equation [19]: i p = ( ) n (αn a ) 1/2 AC D 1/2 v 1/2 (8) Fig. 7. Cyclic voltammograms of 1 M V(IV) at scan rates of (a).5 V s 1, (b).1 V s 1, (c).2 V s 1, (d).3 V s 1, (e).4 V s 1. where i p is the peak current (A), n the number of electrons processes, α the transfer coefficient, A the electrode area (cm 2 ), C o the bulk concentration (mol cm 3 ), D the diffusion coefficient (cm 2 s 1 ) and ν is the scan rate (V s 1 ). Assuming a α value of.5, from the slope of the linear plot in Fig. 8, a V(IV) ion diffusion coefficient of cm 2 min 1 was obtained. Compared to the values listed in Table 3, it is evident that the rate of diffusion of V(IV) ions through the solution is three to four times greater than through the ion exchange membrane. The IEC refers to the concentration of fixed charges in the exchanger, analogous to the concentration of an electrolyte solution [2]. It is a measure of how many counter ions can be taken up by the membrane. An ideal ion exchange membrane would have a high IEC to allow the passage of the charge-carrying H + ions and thus providing a low resistance. Factors leading to a high IEC include a low cross-linking density, flexible crosslinks rather than rigid, large number of functional groups and a maximum number of active donor atoms per active functional group [11]. Although a high IEC may be desirable, a low cross-linking density renders the membrane susceptible to the effects of swelling. When immersed in an aqueous solution, ion exchange membranes swell, absorbing water and ions. The degree of swelling of the membrane significantly influences the rate of ion diffusion as well as the ion selectivity. Generally, the higher the degree of cross-linking, the higher the selectivity. Table 3 shows the properties of the various ion exchange membranes evaluated. The changes of length, width and thick- Fig. 8. Plot of peak current (i p ) vs. the square root of the scan rate (ν 1/2 ).

7 4 H. Vafiadis, M. Skyllas-Kazacos / Journal of Membrane Science 279 (26) Table 3 Properties of the membranes evaluated Membrane IEC (mmol g 1 ) Resistivity ( cm 2 ) V(IV) diffusivity (1 7 cm 2 /min) Water content (%) Gore L Gore M ABT ABT ABT SZ Hipore TM Table 4 Percentage changes in membrane length after 1 day and 15 days of soaking in each solution Membrane Distilled water (%) 2 M V 3.5 in 2 M HCl M HBr (%) 2 M HCl M HBr (%) 1 day 15 days 1 day 15 days 1 day 15 days Gore M Gore L ABT ABT ABT SZ Hipore Table 5 Percentage changes in membrane width after 1 day and 15 days of soaking in each solution Membrane Distilled water (%) 2 M V 3.5 in 2 M HCl M HBr (%) 2 M HCl M HBr (%) 1 day 15 days 1 day 15 days 1 day 15 days Gore M Gore L ABT ABT ABT SZ Hipore ness of the membranes during tests of dimensional stability are shown in Tables 4 6. All membranes, except the Hipore TM, showed the most significant dimensional changes of length and width in the distilled water. This result was expected, as the concentration of ion exchange groups within the membrane is higher than the distilled water, producing a marked difference in osmotic pressure. Minimal degrees of swelling were expected in the more concentrated solutions, as the osmotic pressure difference between the external solution and the membrane is decreased. Most membranes exhibited lowest levels of swelling in the 2 M V 3.5 electrolyte. The difference observed between the final and initial dimensions indicates that the membranes generally swell or shrink by a certain degree then slowly stabilize and come to equilibrium with the solution. Changes in membrane Table 6 Changes in membrane thickness after 1 day and 15 days of soaking in each solution Membrane Distilled water (mm) 2 M V 3.5 in 2 M HCl M HBr (mm) 2 M HCl M HBr (mm) 1 day 15 days 1 day 15 days 1 day 15 days Gore M Gore L ABT3 ABT ABT SZ Hipore

8 H. Vafiadis, M. Skyllas-Kazacos / Journal of Membrane Science 279 (26) thickness (Table 6) appear to be more pronounced in the acidic solutions of 2 M HCl, 6.4 M HBr and 2 M V 3.5. Changes detected for most membranes were however very small and at the limits of the measuring apparatus. Although changes in thickness have been reported, they are merely for comparison as it is difficult to draw valid conclusions. From the relationships described above, it was assumed that the ion exchange membranes with high IEC would have a high rate of vanadium ion diffusion (due to their decreased selectivity), a low area resistivity and high degree of swelling. This general trend was not observed with all ion exchange membranes. Ion exchange membranes ABT3, ABT4 and ABT5 differ only by the number of functional groups on the surface and by thickness. The IEC of the ABT3 was almost double, whilst measured resistivity was significantly lower than the ABT4 and ABT5 membranes (Table 3). This is consistent with the trend outlined above, however the V(IV) diffusivity of the ABT3 membrane is not, being an order of magnitude lower than the ABT4 and ABT5. The behaviour of the ABT3, ABT4 and ABT5 membranes during the small-scale cycling experiments, outlined in Table 2, support the results of the IEC, conductivity and diffusivity tests. Compared to the ABT3 membrane, the ABT4 and ABT5 membranes produced lower voltage efficiencies, and the higher rates of solution crossover resulted in lower coulombic efficiencies. The swelling behaviour of the ABT range of membranes is consistent with the properties listed in Table 3. The ABT3 membrane, exhibited one of the highest degrees of swelling in distilled water (>4%). In the concentrated acid solutions however, the membrane length was less than the original value. This can be attributed to the membrane s high IEC and low resistivity. No change in ABT3 thickness was observed in any solution. The thickness of the ABT4 increased only in the acidic solutions, whilst the ABT5 increased both in the distilled water and the acid solutions. The SZ membrane showed a high IEC, high rate of V(IV) diffusion and a high area resistance. These properties are consistent with the low voltage efficiency (44%) and high rate of solution crossover observed during the cycling of the membrane (see Section 3.1). The SZ membrane showed high dimensional changes, approximately 3%, which did not differ greatly in the distilled water or the acids. The Gore L1854 and M4494 membranes produced the lowest values of IEC and V(IV) ion diffusivity, whilst retaining the lowest values of area resistance. Small dimensional changes were observed in the acidic solutions, however the M4494 membrane showed the greatest dimensional changes in length and width in distilled water (6.5%). Changes were consistently larger than the L1854 membrane, possibly due to the greater IEC (and thus higher diffusivity) of Gore M Very small thickness changes were detected for both membranes. Although the Hipore TM microporous separator has no ion exchange functionality, an IEC value of 1.14 mmol g 1 was obtained. This can be attributed to the separators ability to uptake solution into its pores, which can then readily diffuse into another solution (as seen in the cell cycling results). The water content calculated for the Hipore TM was 62% (Table 3), an order of magnitude greater than most other membranes. The highly porous structure allowed a large uptake of water into its pores during the soaking, which in turn was easily removed during the drying. It was thus established that the methods for calculating the IEC and water content of the ion exchange membranes are not as suitable for evaluating porous separators and thus cannot be directly compared. The dimensional changes of the Hipore TM microporous separator were expected to be significant as the ability of the separator to uptake solution was demonstrated in the IEC and water content tests. Interestingly, the separator showed a slight decrease in length and width, however showed a.3 mm increase in thickness. Table 3 lists the water content values calculated for the ion exchange membranes. No direct relation between membrane behaviour and water content was found. All ABT membranes contained similar percentages of water, however their properties and behaviour were quite different. It is however interesting to note that the Gore M4494 membrane had higher IEC and water content than the L Chemical stability A major limiting factor in the lifetime of the V/Br system is the chemical stability of the ion exchange membrane in the highly oxidising electrolyte. Membranes such as the SELEMION HSF, HZ cation and HZ anion showed good initial coulombic efficiencies (84%, 83% and 9%, respectively), however their life in the cell was limited due to their poor chemical stability. The SELEMION HSF membrane blistered within 14 cycles and lost its opaque colour and shinny texture. The HZ cation and anion exchange membranes were broken down by the electrolyte and became very fragile after 2 and 4 cycles, respectively. Other membranes such as the HSV, ABT1 and ABT2 did not show good cycling properties or chemical stability. Although their immersion time in the battery electrolyte was limited (less than 1 cycles), it was sufficient to demonstrate their lack of chemical resistance. The HSV membrane blistered and its dark brown colour faded to a light brown. Both the ABT1 and ABT2 membranes delaminated, as the backing material used to increase the mechanical stability of the membranes detached from the ion exchange layer. The ion exchange layer was too thin and fragile to act as a separator and thus a large volumetric transfer of solution from the positive to the negative side occurred. The ABT4 and ABT5 membranes were characterised by relatively low coulombic efficiencies (63% and 45%) and higher rate of solution crossover. No physical changes after cycling were however observed. During long term cycling, however, the coulombic efficiency of the ABT3 membrane increased, while the voltage efficiency steadily decreased. This is thought to be the fouling of the membrane, leading to the blockage of the pores and reduced transfer rates of the redox ions across the membrane. Although the L1854 membrane lasted at least 16 cycles, when removed from the cell it was found to have blistered. It is not known at what stage the blistering began, or how much longer the membrane would last in this condition. Long term stability

9 42 H. Vafiadis, M. Skyllas-Kazacos / Journal of Membrane Science 279 (26) testing will review these questions. The M4494 membrane was also found to have blistered at some stage during the 3 cycles completed thus far, however the effect of this on the cycling behaviour has not yet been apparent. The cell testing is still continuing for this membrane. 4. Conclusion Most of the ion exchange membranes tested to date are not suitable for use in a V/Br flow cell. A high resistance was shown by a large number of membranes, some too high to allow any cell cycling to occur with the set voltage limits. Other membranes have shown a low chemical stability in the highly oxidising electrolyte. The membranes that have shown positive results in the V/Br system are the ABT3, L1854 and M4494 membranes. Over 16 cycles, the ABT3 membrane has shown average coulombic and voltage efficiencies of 8% and 64%, whilst the L1854 membrane, 86% and 62%, respectively. Considerable improvements in voltage efficiencies can be achieved by the optimisation of cell design. The stability of the Hipore TM microporous separator in the cell electrolyte and the improved cycling behaviour is promising. Studies will endeavour to optimise its selectivity and IEC. Further investigations are also needed to understand the unusual cycling behaviour seen with the discharge of the ABT3 membrane and the blistering of the L1854 and M4494 membranes. These investigations are currently being conducted in parallel to the screening of further membrane materials as candidates for the V/Br redox flow cell. Acknowledgments The authors are grateful to W.L Gore and Associates for the samples of M4494; to the Australian Battery and Technology Trading Company for the ABT membrane samples; to Guangzhou Delong Technologies Pty Ltd for the SZ membrane samples and to Mitsubishi Australia for the Hipore TM microporous separator. References [1] A. Landgrebe, F. McLarnon, Atmospheric impacts of fuel cell and battery technology, in: Proceedings of the Symposium on Fuel Cells, San Francisco, California, [2] E. Sum, M. Skyllas Kazacos, A study of the V(II)/V(III) redox couple for redox flow cell applications, J. Power Sources 15 (1985) [3] E. Sum, M. Rychcik, M. Skyllas Kazacos, Investigation of the V(V)/V(IV) system for use in the positive half-cell of a redox battery, J. Power Sources 16 (1985) [4] M. Skyllas-Kazacos, M. Rychcik, R.G. Robins, A.G. Fane, M.A. Green, New all-vanadium redox flow cell, J. Electrochem. Soc. 133 (1986) [5] M. Skyllas Kazacos, T.G. Robbins, US 4,786,567, [6] M. Skyllas Kazacos, C. Menictas, M. Kazacos, Thermal stability of concentrated V(V) electrolytes in the vanadium redox cell, J. Electrochem. Soc. 143 (1996) L86 L88. [7] M. Skyllas-Kazacos, Novel vanadium chloride/polyhalide redox flow battery, J. Power Sources 124 (23) [8] M. Skyllas Kazacos, Vanadium/polyhalide redox flow battery, WO 3/19714 A1, 23. [9] M. Skyllas Kazacos, M. Kazacos, A. Mousa, Metal halide redox flow battery, WO 3/92138 A2, 23. [1] D.G. Oei, Permeation of vanadium cations through anionic and cationic membranes, J. Appl. Electrochem. 15 (1985) [11] S.C. Chieng, M. Kazacos, M. Skyllas-Kazacos, Preparation and evaluation of composite membrane for vanadium redox battery applications, J. Power Sources 39 (1992) [12] T. Mohammadi, M. Skyllas Kazacos, Modification of anion-exchange membranes for vanadium redox flow battery applications, J. Power Sources 63 (1996) [13] E. Wiedemann, A. Heintz, R.N. Lichtenthaler, Transport properties of vanadium ions in cation exchange membranes: determination of diffusion coefficients using a dialysis cell, J. Membr. Sci. 141 (1998) [14] F. Grossmith, P. Llewellyn, A. Fane, M. Skyllas-Kazacos, Evaluation of membranes for all-vanadium redox cell, in stationary energy storage, load leveling, and remote applications, in: Proceedings of the Electrochemical Society, Honolulu, 1988, pp [15] F.G. Helfferich (Ed.), Diffusion Across Membranes, McGraw-Hill Book Company, Inc., New York, 1962, pp [16] S.C. Chieng, Membrane Processes and Membrane Modification for Redox Flow Battery Applications, PhD, Chemical Engineering, University of New South Wales, Sydney, [17] T. Mohammadi, S.C. Chieng, M. Skyllas Kazacos, Water transport study across commercial ion exchange membranes in the vanadium redox flow battery, J. Membr. Sci. 133 (1997) [18] T. Sukkar, Membrane Studies for Vanadium Redox Flow Cell Applications, PhD, School of Chemical Engineering and Industrial Chemistry, University of New South Wales, Sydney, 22. [19] A.J. Bard, L.R. Faulkner, Electrochemical Methods, Fundamentals and Applications, 2nd ed., John Wiley & Sons Inc., NJ, 21. [2] R. Paterson, An Introduction to Ion Exchange, Heyden and Son, Ltd., London, 197.