Performance Evaluation of a Hydrogen-Vanadium Reversible Fuel Cell

Hydrogen-vanadium reversible fuel cells were tested using a Pt/C hydrogen electrode, carbon vanadium electrode and interdigitated ﬂow ﬁelds at both electrodes. Vanadium electrolyte ﬂow rate was varied to study its effect on mass transport performance. Two types of vanadium electrodes were explored, a single layer of high surface area carbon nanotube (CNT) electrode and three layers of nitric acid-treated carbon paper. Finally, four types of Naﬁon membranes were examined to determine the effect of membrane type and thickness on fuel cell charge and discharge performance. Higher performance was observed with higher vanadium ﬂow rate, thinner membranes and a CNT vanadium electrode. Peak power density of greater than 540 mW/cm 2 was obtained using a NR212 membrane and CNT vanadium electrode.

Hydrogen-vanadium fuel cells offer a feasible solution for storing electrical energy from the grid or directly from renewable energy sources such as wind and solar. 1 In a hydrogen-vanadium reversible fuel cell, the charge and discharge reactions are as follows: V anadium (+) Electrode : Overall Reaction : .99 V While charging, a hydrogen-vanadium fuel cell stores energy in the form of hydrogen and vanadium (V). During discharge, hydrogen is consumed at the negative electrode and vanadium (V) is reduced to vanadium (IV) at the positive electrode. Vanadium systems demonstrate no issues with membrane fouling or metal dendrite formation (e.g. iron and zinc electrode systems). 2,3 Additionally, vanadium solutions have low volatility, low corrosivity, and do not produce toxic vapors. 4,5 In particular, cells that utilize chlorine or bromine pose a significant safety concern due to their high vapor pressures and toxic properties. [6][7][8] Due to the relatively high cost of vanadium, the hydrogen-vanadium fuel cell is attractive over the all-vanadium system due to the 50% reduction in the amount of vanadium solution required. Since the cost of the vanadium electrolyte for the all-vanadium flow battery makes up roughly 40% of the total system cost, cutting the vanadium electrolyte requirement by half has a large impact on reducing the overall system cost. 1 Another benefit of a mixed gas/liquid electrolyte system (hydrogen gas at the negative electrode and liquid vanadium at the positive electrode) is the ease of separation if crossover occurs. Unfortunately, one of the major disadvantages of the hydrogen-vanadium system compared to the all-vanadium sys-tem is that a precious metal catalyst is required for HOR/HER at the hydrogen electrode.
Past research by Yufit et al. has shown the feasibility of the hydrogen-vanadium fuel cell and the importance of vanadium electrode wettability on fuel cell performance. 9 The wettability of the vanadium electrode is important for two reasons. First, only wetted area or area with access to the electrolyte is active, and second, a more wetted porous electrode allows vanadium electrolyte to more easily flow into and through the porous carbon electrode. The vanadium reaction is a one-phase (liquid) reaction involving solely aqueous ions and water. Yufit et al. achieved a peak performance of 114 mW/cm 2 using a vanadium flow rate of 200 mL/min and a catalytic active area of 25 cm 2 . Additionally, studies by Menictas et al. also revealed crossover of vanadium species during operation of the vanadium-oxygen fuel cell, sparking our interest in vanadium's effect on the hydrogen electrode catalyst. 10 Furthermore, Xie et al. examined vanadium permeability through different proton exchange membranes, including XL100, NR211, NR212, N115, and N117. His group found that certain processing and pretreatment conditions could significantly reduce the permeability of vanadium, but the changes were largely absent after repeated fuel cell cycling. 11 Houser et al. and other research groups have performed extensive optimization studies on the all-vanadium flow battery, but little work has been done thus far on the reversible hydrogen-vanadium fuel cell. 12,13 In this study, we examine the performance of a reversible hydrogen-vanadium fuel cell when using a Pt/C hydrogen electrode, a high surface area carbon electrode as the vanadium electrode, and interdigitated flow fields at both electrodes. First, we studied the effect of varying the vanadium electrolyte flow rate on fuel cell performance. Then, two different carbon electrodes were tested for the vanadium side to determine their effect on performance. In the first case, a carbon nanotube (CNT) vanadium electrode was tested. The high surface area CNT electrode was used in previous studies in a hydrogen-bromine fuel cell with remarkable performance. 14,15 In the second case, 3 layers of nitric acid-treated SGL 10AA were used as the vanadium electrode. Finally, we explored the effect of membrane type (extruded versus solution-cast) and thickness on fuel cell performance.

Experimental
Three different studies were completed on the hydrogen-vanadium reversible fuel cell. In the first study, vanadium electrolyte flow rate was varied to determine the effect on mass transport performance at high current densities. For the second study, two types of vanadium electrodes were tested. In the third study, different proton exchange membrane types and thicknesses were explored. Table I lists the key parameters used for each study. Vanadium electrolyte solution (500 mL) was prepared of 1 M vanadium(V) and 1 M vanadium(IV) by dissolving the appropriate amount of vanadyl sulfate (Sigma Aldrich) in 3 M sulfuric acid (H 2 SO 4 ) and then charging the fuel cell to 50% state of charge (SOC) to make vanadium(V). Excess electrolyte was used to ensure that the electrolyte concentration remained relatively constant during the polarization measurements. The membrane electrode assembly (MEA) was prepared by either one of two ways. In the first method, the MEA was prepared by hot pressing a proton exchange membrane onto a 3 cm by 3 cm Pt-coated GDL (0.45 mg/cm 2 Pt loading, 0.16 mg/cm 2 Nafion ionomer, SGL35BC from TVN Systems, Inc.). In the second method, the MEA was prepared by hot pressing a proton exchange membrane between a 3 cm by 3 cm Pt-coated GDL (0.45 mg/cm 2 Pt loading, 0.16 mg/cm 2 Nafion ionomer, SGL35BC from TVN Systems, Inc.) and a nitric acid-treated 3 cm by 3 cm GDL (SGL10AA from Ion Power, Inc.) that was coated with Nafion solution on one side to allow hot pressing. SGL35BC electrode thickness was ∼320 μm. The catalyst layer (loading of 0.45 mg/cm 2 Pt, 0.16 mg/cm 2 Nafion ionomer) was sprayed directly onto the microporous layer side of SGL35BC. TVN Systems, Inc. reported that the Pt catalyst used for the catalyst layer is commercially available (Tanaka) with Pt particle diameter of 1-2 nm. Commercially available membranes were purchased from Ion Power Inc. (DE, USA). MEAs were made using NR211 (∼25 um thick), NR212 (∼51 um thick), N115 (∼127 um thick), or N117 (∼183 um thick). Hot pressing was completed at 135 • C and 0.552 MPa (80 psi) for 5 min. Assembly of the fuel cell was carried out at 1.103 MPa (160 psi) using expanded polytetrafluoroethylene (PTFE) gaskets and interdigitated tantalum flow fields with 9.0 cm 2 flow area. The interdigitated flow field dimensions were 1.5 mm channel width, 1 mm channel depth and 2.5 mm shoulder width.
The vanadium electrode consisted of either a single CNT electrode or 3 layers of nitric acid-treated SGL10AA. The CNT electrode was prepared by growing nanotubes onto SGL10AA carbon paper using chemical vapor deposition as outlined in Reference 3. Nitric acid-treated SGL10AA was made using a 3-step process in order to improve electrode wettability. First, SGL10AA carbon paper was submerged in water while evacuating the air above the water using a vacuum pump at 0.03 MPa for 5 min. Then, the carbon paper was soaked in 2 M nitric acid for 24 h. Lastly, the carbon paper was thoroughly rinsed in deionized water and dried. 3 layers of SGL10AA were tested as the vanadium electrode because previous studies found that a single layer of SGL10AA did not have sufficient active area to provide high current density operation and when more than 3 layers were used the negative effect of increased transport distance from a thicker electrode became more dominant than the positive effect of having higher active surface area. 14,15 In our previous studies with CNT electrodes in the hydrogen-bromine fuel cell we found that a single layer of CNT electrode on the bromine side yielded similar to or better performance than 3 layers of SGL10AA. The CNTs, which were grown onto the GDL carbon fibers, increase the active surface area available for the vanadium reaction while maintaining the same electrode thickness and transport distance of a single layer of SGL10AA electrode (∼350 um). Increased electrode thickness (i.e. 3 layers of SGL10AA) leads to greater transport distance and fuel cell resistance, therefore making a single layer CNT electrode more attractive than using 3 layers of SGL10AA.
Fuel cell testing was completed using a hydrogen pressure of 0.136 MPa (5 psig) and vanadium electrolyte flow rate of 5, 6, or 12 mL/min (Equivalence of 1.55, 1.86, and 3.72 A/cm 2 , respectively, for the vanadium (V) concentration used, 1.735M at an OCV of 1.09V). All testing was completed at room temperature (∼23 • C). The start-up procedure included cycling the reversible fuel cells in charge (1.3 V) and discharge modes (0.6 V) every 10 min for over 12 h. Prior to collecting all discharge and charge polarization curves, the vanadium electrolyte was charged to ∼90% SOC. Multiple polarization curves were collected to ensure adequate membrane hydration and repeatable results.
Electrochemical impedance spectroscopy (EIS) was conducted on the hydrogen-vanadium fuel cell after operation in order to measure the total resistance of the fuel cell. EIS was also conducted on the fuel cell without the membrane in order to measure the electronic resistance of the fuel cell. Figure 1 shows the assembly layout for the reversible hydrogen-vanadium fuel cell.

Results and Discussion
Due to charging the vanadium electrolyte solution to ∼90% SOC prior to collecting polarization curves, the open circuit potential (OCV) for the hydrogen-vanadium reversible fuel cell was approximately 1.09 V (as predicted by the Nernst equation). All polarization curves have been normalized using the OCV (i.e. cell voltage minus OCV) in order to more easily compare all cases.   shows the polarization and discharge power density curves for the hydrogen-vanadium fuel cell using various vanadium flow rates (Study 1). We observe better mass transport performance at higher current densities as we increase the vanadium flow rate from 5 mL/min to 6 mL/min (1.55 and 1.86 A/cm 2 , respectively, at the concentration used). During discharge at 5 mL/min, we start to observe a mass transport effect at around 100 mA/cm 2 and limiting current at 230 mA/cm 2 . During discharge at 6 mL/min, we start to observe a mass transport effect at around 250 mA/cm 2 and limiting current above 350 mA/cm 2 . This significant increase in performance after a 20% increase in flow stoichiometry is attributed to better transport of active materials to and removal of products from the porous electrodes because of deeper penetration of the electrolyte into the electrodes due to using interdigitated flow fields. [16][17][18][19] When increasing the vanadium flow rate to 12 mL/min (3.72 A/cm 2 equivalence), we observe a more subdued increase in performance at higher current densities. Since the performance at 6 mL/min and 12 mL/min are similar, a vanadium flow rate of 12 mL/min is used for all subsequent studies to eliminate the mass transport effect caused by low vanadium flow rate. We did not observe further mass transport improvement when increasing the vanadium flow rate above 12 mL/min. Previous hydrogen-vanadium studies completed by Yufit et al. 9 with serpentine flow fields required vanadium flow rates above 150 mL/min (with 25 cm 2 active area and serpentine flow fields) in order to minimize mass transport limitations. Figure 3 compares the fuel cell performance of a CNT vanadium electrode to the one made from 3 layers of nitric acid-treated SGL10AA (Study 2). We observe improved performance for the CNT electrode at higher current densities. The limiting currents for the CNT electrode and the 3 layer SGL10AA electrode were 0.7 A/cm 2 and 0.55 A/cm 2 , respectively. The mass transport limitation observed in Figure  3 during charge for V_SGL10AA_NR212 arises due to the increased electrode thickness when using 3 layers of SGL10AA (∼1050 μm) versus 1 layer of CNT electrode (∼350 μm). From previous studies with the hydrogen-bromine system, a single layer of this CNT electrode was found to have active surface area equivalent to greater than 10 layers of conventional carbon electrode with an additional benefit of reduced thickness and therefore shorter transport distance from the flow channels. Therefore, the increased performance can be attributed to the higher active surface area available to the CNT electrode. This leads to lower activation loss, faster transport, and lower ohmic resistance. The ohmic resistance of the electrolyte near the active area is reduced since the local current density per active site is lower when the total current is distributed over a larger area. Figure 4 compares the fuel cell performance when using various types and thicknesses of Nafion membranes (Study 3). Identical positive (3 layers of SGL10AA) and negative (SGL35BC) electrodes, flow fields, hydrogen pressure, and electrolyte flow rate (12 mL/min) were used for each cell. We observe improved performance as membrane thickness decreased and a peak power density of more than 300 mW/cm 2 with NR211. Thinner membranes lead to lower cell ohmic resistance or faster hydronium ions transport across the membrane during charge and discharge. Figure 5 compares the discharge polarization curves before and after iR correction for various membrane types and thicknesses. The fuel cell resistance measured using EIS for NR211, NR212, N115, and N117 were 0.04, 0.068, 0.104, and 0.166 ohms, respectively. Full iRcorrected discharge polarization curves remove all ohmic losses in the fuel cell, including the ohmic resistance of the membrane. This allows us to directly compare the kinetic and mass transport effect for the various membrane types and thicknesses. First, with the ohmic resistance of the membranes removed, we observe similar kinetic performance in the low overpotential region for the two groups of membranes, the extruded N115/N117 membranes and the solution-cast NR211/NR212 membranes. Next, we are able to identify the current density in which the mass transport effect starts to negatively impact the discharge polarization curves. For the extruded N115/N117 membranes, the mass transport effect starts at 0.18 A/cm 2 . However, mass transport doesn't start to negatively impact the solution-cast NR211/NR212 membranes  until greater than 0.5 A/cm 2 . Overall, we observe increased performance when using solution-cast membranes, vice extruded, due to the improved mass transport effect. Solution-cast membranes (NR211 & NR212) are known to have a more hydrophilic surface (due to a higher concentration of sulfonate ion groups at the membrane's surface) than extruded membranes (N115 & N117). 20 The increased hydrophilicity of solution-cast membranes leads to increased ionic conduction near the membranes surface, therefore improving ionic and mass transport. Figure 6 graphs the polarization and power density curves for the reversible hydrogen-vanadium fuel cell with a CNT vanadium electrode and NR212 membrane. Full iR-corrected and electronic iRcorrected polarization and power density curves are shown. Graphing the full and electronic iR-corrected curves allows us to visually analyze the current and power densities we expect to obtain if we overcame all ohmic losses in the fuel cell and compare these to the case of removing only the ohmic losses from electronic connections. When using improved current collectors in a commercial fuel cell, we expect very low ohmic losses attributed from electronic connections. Our fuel cells have not been designed for low electronic loss. The full iR-corrected curves provide us with the maximum achievable performance possible if all the ohmic losses due to the electrodes, electronic connections and membrane are overcome. After removing the iR loss, we can also see the kinetic loss effect and when mass transport loss Figure 6. Polarization and discharge power density curves of the reversible hydrogen-vanadium fuel cell using a CNT vanadium electrode and NR212 membrane before and after IR correction.
begins. During discharge, we start to observe a mass transport effect at around 500 mA/cm 2 and limiting current at 700 mA/cm 2 . However, during charge, we don't observe a mass transport effect within the overpotential or current density range evaluated. Based on the polarization curves in Figure 6, we observe very high kinetics on both charge and discharge, but mass transport losses take over during discharge as the overpotential is increased. Our studies show a 3-4 times increase in peak power density over the previous hydrogen-vanadium studies conducted by Yufit, et al. 9 The performance enhancement is attributed to using interdigitated vice serpentine flow fields, thinner proton exchange membranes, and advanced CNT electrodes.

Summary
We conducted a performance study on a reversible hydrogenvanadium fuel cell using interdigitated flow fields at both electrodes. Two different types of vanadium electrodes (CNT and nitric acid-treated SGL10AA) were explored, as well as various types of proton exchange membranes. The hydrogen-vanadium fuel cell shows promising performance by achieving peak power greater than 540 mW/cm 2 .
Further investigation is needed to determine whether electrospun nanofiber composite membranes made of perfluorosulfonic acid (PFSA) fibers and inert fibers to control membrane swelling will be suitable for the hydrogen-vanadium fuel cell. Electrospun nanofiber composite membranes offer the potential of lower vanadium crossover and lower membrane resistance. Therefore, these membranes may lead to increased fuel cell performance and cycling efficiency.