Carbon Felt Coated with Titanium Dioxide/Carbon Black Composite as Negative Electrode for Vanadium Redox Flow Battery

This investigation focuses on the effect of titanium dioxide (TiO 2 ) coatings of a carbon black (XC-72) negative electrode on the performance of a vanadium redox ﬂow battery (VRFB). TiO 2 , a hydrophilic material, was added to the carbon electrode to improve the wettability and reduce the electrical resistance of the electrode surface. The electrochemical performances of homemade TiO 2 , commercial TiO 2 , and carbon felt are investigated by using cyclic voltammetry and single-cell charge–discharge measure- ments. An electrode with 20 wt% of fabricated TiO 2 loading at a scan rate of 0.006 V s − 1 shows a speciﬁc capacitance ( C s,t ) of 186.2 F g − 1 , which is 55.5% and 12.2% higher than that of pure carbon electrode (119.7 F g − 1 ) and commercial TiO 2 (166.0 F g − 1 ), respectively. At current density of 200 mA cm − 2 , the energy storage efﬁciency ( η E = 65.4%) of the single cell with 20 wt% homemade TiO 2 /C-containing carbon felt negative electrode is 16.0% and 6.1% higher than that of the negative elec- trode with raw carbon felt ( η E = 56.4%) and of the negative electrode containing commercial TiO 2 /C ( η E = 61.6%), respectively. These results demonstrate the potential application of TiO 2 /C electrodes for high-efﬁciency were a containing and then ultrasonically agitated until the inks became homogeneous. A drop of the was spread on a holey copper grid for TEM observation. The size distribution of the anatase TiO 2 nanoparticles in TiO 2 /C electrode material was obtained by di- rectly measuring 300 randomly selected TiO 2 particles from the TEM images. The sessile drop method Angstro, USA) was used to measure the water contact angle of the TiO 2 /C composite electrodes. was onto one side of and then was ﬁve different and angle then

Vanadium redox flow battery (VRFB) has been proposed as a promising candidate for large scale energy storage applications, such as load-leveling applications. VRFB can store and stabilize intermittent electricity generated from wind turbine or photovoltaic. VRBF has several advantages over other types of batteries, such as excellent electrochemical reversibility, high roundtrip efficiency, flexible, and negligible cross-contamination between positive and negative electrolytes. [1][2][3][4][5] VRFB employs V(IV)/V(V) and V(II)/V(III) redox couples as positive and negative half-cells, respectively, with the standard open circuit cell potential approximate 1.26 V at 100% state of charge. 6 Carbon felt is a typical electrode material and it has wide operating electrode potential range, chemically stable, high surface area, and reasonable price. However, carbon felt electrode shows poor electrochemical activity. Therefore, much attention has been focus to electrode modification to enhance its electrochemical properties. [7][8][9][10][11][12][13][14] Several alternative electrode materials have been proposed to improve electrode performance, such as metal electrodeposition on carbon fibers. Various metal compounds have also been deposited on graphite fibers to improve the catalytic activity for vanadium redox couples and to enhance electrode stability in acidic vanadium solution. In the literature, metal compounds deposited on carbon felts include IrO 2,9 Ru(O 2 ), 10 and Ir, 11 whereas others include partial modification of the functional groups on the graphite surface. 12,13 The coating of metal nanoparticles on the fiber of carbon felt is a promising approach. Recently, transition metal oxides, e.g. TiO 2 , ZnO, have been studied extensively as a water adsorbent for improving the wettability of the catalyst layer, thus the performance of proton exchange membrane fuel cell 14,15 was enhanced. Besides, TiO 2 has satisfactory chemical stability, endurance, and lower production costs. 16 The property of TiO 2 is influenced by crystal structure, surface area, size distribution, porosity, bandgap, and surface hydroxyl density. 17,18 Titanium dioxide mixed with carbon black has been proven 19 to be a good electrode additive for VRFB application at a low current density of 20 mA cm −2 .The study of TiO 2 as an additive for the VRFB electrode at high current density is rarely discussed in literature. Our previous study 19 indicated that addition of TiO 2 powder on the VRFB electrode can * Electrochemical Society Fellow. ** Electrochemical Society Active Member. z E-mail: kanlinhsueh@hotmail.com; fsshieu@dragon.nchu.edu.tw enhance cell performance. Nevertheless, the current density tested in our previous study was far lower than that required for practical applications. Therefore, this study intends to explore the feasibility of adding TiO 2 on the electrode at much higher current density (200 mA cm −2 ) than 20 mA cm −2 .
At high current density such as 200 mA cm −2 , numerous side reactions or problems may be encountered. In long-term VRFB operation, gas evolution may occur on the electrodes during the battery charging period. Gas evolutions (both hydrogen and oxygen evolution reactions) result in the loss of energy efficiency and storage capacity. 20 This study aims to evaluate the effects of hygroscopic titanium dioxide (TiO 2 ) nanoparticle deposited onto a carbon black negative electrode on the performance of a vanadium redox flow battery (VRFB) at high current density. Charge-discharge measurement was conducted under ambient conditions. Electrode material containing various TiO 2 amounts was loaded onto the carbon black (XC-72) composite. This study aims to find the preparation condition of the electrode containing TiO 2 and the electrochemical performance of the electrode for VRFB.

Experimental
The experimental conditions, procedure, and instruments used in this study are described in detail in the following sections.
Preparation of anatase TiO 2 particles and TiO 2 /C composite electrode.-A two-step sol-gel process (hydrolysis, condensation, and calcination) in basic medium was used to synthesize anatase TiO 2 particles from the precursor, tetra-n-butyl titanate (TnBT, Alfa Aesar). The hydrolysis process was carried out in an ice-water bath. The TnBT (0.05 mol) dissolved in 20 mL ethylene glycol (EG, Aldrich), was added to the mixture of water and ethylene glycol (V water /V EG = 1/1, V mix = 20 mL). This mixture was magnetically stirred to yield a 40 mL solution. Subsequently, the solution was thermostatic ally dehydrated in an open vessel. The solution temperature was gradually increased to 103 • C for 6 h to remove the solvent by evaporation. It was then dried at 120 • C for 10 h to remove the by-product, butyl alcohol. The product was washed by centrifugation/re-dispersion cycles with ethanol and water to remove residual EG. The precipitate was then dried in a vacuum oven at 300 • C overnight. The TiO 2 /C composite electrodes were prepared by mixing TiO 2 particles and carbon black (Vulcan XC-72). The TiO 2 /C composite electrodes were prepared by mixing TiO 2 particles and carbon black (Vulcan XC-72). The mixing was carried out using the thermal-reflux method. First, well-mixed suspension of carbon black was obtained by adding 0.8 g of carbon black and 30 mL of ethanol under ultrasonication for 4 h. Afterward, 0.2 g of TiO 2 and 20 mL of deionized water were added to the carbon black ink under ultrasonication for another 4 h. A 1.0 M NaOH solution was gradually added to the mixture until the pH reached 10 to 11. Finally, the mixed ink (TiO 2 /C ink) was constantly stirred under thermal reflux at 80 • C for 4 h. The 20 wt% TiO2/C powder was filtered and dried overnight at 90 • C in a vacuum oven. The weight ratios of the homemade TiO 2 /XC-72 (HM-TOC-20) and commercial-TiO 2 /carbon (COM-TOC-20) were both 20 wt%.
Material characterization analysis.-The crystal structure of the carbon black (Vulcan XC-72), commercial TiO 2 (Biocare, USA), and homemade-TiO 2 (HM-TiO 2 ) were analyzed by X-ray diffraction (XRD, Rigaku D/MAX 2500) with Cu K α radiation (λ=1.5418 Å).The XRD diffraction spectra were recorded in the range from 20 • to 70 • (2θ) at a scan rate of 2 • min −1 and the result was compared with the JCPDS data files. High-resolution field-emission transmission electron microscopy (HR-TEM, JEOL Model JEM-2100F) operated at 200 kV was used to examine the size, morphology, and distribution of the TiO 2 /C composition material. The TiO 2 /C sample for TEM analysis was prepared by the following procedure and condition. The commercial-TiO 2 and homemade-TiO 2 inks were placed in a vial containing ethanol and then ultrasonically agitated until the inks became homogeneous. A drop of the inks was spread on a holey copper grid for TEM observation. The size distribution of the anatase TiO 2 nanoparticles in TiO 2 /C electrode material was obtained by directly measuring 300 randomly selected TiO 2 particles from the TEM images. The sessile drop method (First Ten Angstro, FTA 2000, USA) was used to measure the water contact angle of the TiO 2 /C composite electrodes. TiO 2 /C ink was sprayed onto one side of commercial hydrophobic carbon felt, and then dried at 60 • C for 30 min. Each sample was measured five times at different locations, and the average water contact angle was then calculated.
Electrochemical characterization techniques: cyclic voltammetry and charge-discharge tests.-The electrochemical characteristics of the TiO 2 /C composite electrodes were evaluated by cyclic voltammetry (CV), CHI Instruments (Model 614B, USA) with a three-electrode cell. The TiO 2 /C composite ink (1.25 mg TiO 2 /C + 50 μL of 5 wt% Nafion solution) was coated on the working electrode for electrochemical measurements. The working electrode was a glassy-carbon electrode with a geometric surface area of 0.196 cm 2 . Pure Pt wire was used as the counter electrode, whereas a saturated Ag/AgCl was used as the reference electrode. During CV, the electrode potential was first scanned toward the positive direction. The potential scanning rate varied from 0.002 V · s −1 to 0.006 V · s −1 in a 1.0 M H 2 SO 4 (Aldrich, 98.5%) solution.
A single cell was used for the battery charging/discharging experiment. Nafion 117 (DuPont), a proton-exchange membrane, was used as separator. Commercial raw carbon felt, with dimensions of 5.0 × 5.0 × 0.6 cm 3 (Carbon Felt-CF060A01836, CeTech Co., Ltd.), served as the base of the positive and negative active electrodes, as shown in Fig. 1. The Nafion 117 membrane [(7.0 × 7.0) cm 2 ] was pretreated with 5 wt% H 2 O 2 (Aldrich), deionized water, 1.0 M H 2 SO 4 (98.5%; Aldrich), and deionized water at 80 • C for 1 h. Each graphite plate (Beam Associate Co., Ltd.) was filled with an equal size of raw carbon felt (fillister dimensions of 5.0 × 5.0 × 0.6 cm 3 ). A constant-current charge-discharge test was performed using a singlecell testing system (Beam Associate Co., Ltd.). The positive electrode was a raw carbon felt with an area of 25 cm 2 (5.0 cm × 5.0 cm), while the negative electrode was TiO 2 /C-modified carbon felt.
The negative electrode was prepared by following procedure: a mixture of TiO 2 /C composite electrode ink was prepared by adding TiO 2 /C nanoparticles, 5 wt% Nafion solution (DuPont, USA), and It is also evaluated to understand the effects of vanadium ion concentration on the charge-discharge performance of a single cell at 200 mA cm −2 . The electrolyte composition is listed in Table II.

Results and Discussion
Material analysis.- Figure 2 , and (204) crystalline planes, respectively, which correspond to typical anatase titanium dioxide phase. The broad peak near 2θ = 25 • is attributed to carbon black. Figure 3a shows that the HM-TiO 2 nanoparticles were well-dispersed on the carbon black (Vulcan XC-72) surface. The histogram of the samples is inserted in Figs. 3a and 3c. The average particle sizes of HM-TiO 2 and commercial-TiO 2 were approximately 9.1 ± 2.0 nm and 16.1 ± 1.2 nm, respectively. A typical high-resolution TEM image, as shown in Fig. 3b and Fig. 3d, reveals the highly crystallized TiO 2 nanocrystal with a lattice spacing of 0.35 nm, which corresponds to the (101) plane of anatase TiO 2.
The hydrophilic characteristic of the TiO 2 /C composite electrode was investigated by a sessile drop method. Figure 4 shows the water contact angle images of carbon black (XC-72, CB), commercial TiO 2 , and HM-TiO 2 , which are 132 • , 88 • , and 52 • , respectively. The contact angle of the electrode layer containing TiO 2 nanoparticles is  significantly lower than that of the electrode without TiO 2 . Therefore, the wettability of the nanoparticle electrode layer containing TiO 2 is better than that of the electrode without TiO 2 .
Electrochemical characterization of TiO 2 /C composite material.- Figure 5 shows the cyclic voltammograms of Vulcan XC-72, COM-TOC-20, and HM-TOC-20 composite electrodes in 1.0 M H 2 SO 4 solution at a scan rate of 0.006 V s −1 . At −0.4 V against Ag/AgCl, the TOC composite electrodes exhibit a higher double-layer charging current than XC-72. The oxidation-reduction current within the potential range of −0.4 V to −0.5 V was due to hydrogen adsorption/evolution reaction. The hydrogen reduction peak was not clearly observed in this cyclic voltammogram. Therefore only the oxidation current was used to evaluate the electrochemical surface area or activity of the electrodes. Results suggest that the composite electrodes have higher specific capacitance and higher electrochemical active area than that of the XC-72 electrode.
In the potential region between −0.5 and −0.4 V, the specific capacitance of the composite electrodes is a combination of double-layer capacitance and faradaic pseudo-capacitance attributed to hydrogen adsorption/desorption. 21,22 The current detected in the potential range from −0.4 to −0.3 V is mainly attributed to double-layer chargingdischarging current. The specific capacitance (C s,t ) of the composite electrodes can be roughly calculated from the current measured in this potential region, which was calculated by using from the following equation: 23,24 where I a and I c are the currents of anodic current and cathodic current at −0.40 V, respectively; W is the mass of the composites (in grams); and dV/dt is the scan rate (in volts per second). Figures 6a to 6c show the CVs of XC-72, COM-TOC-20, and HM-TOC-20 at different scanning rates. Table I lists the values of specific capacitance (C s,t ) calculated from Eq. 1 and the TiO 2 loading of each sample. In general, C s,t increases with the increase in TiO 2 loading. The C s,t value of pure XC-72 and HM-TOC-20 are 119.7 and 186.2 F g −1 , respectively. Addition of TiO 2 in the electrode reduced the contact angle between water and electrode, as shown in Fig. 4. This finding suggests that addition of TiO 2 increases the effective electrode area (hydrophilic area inside electrode), which is also evidenced by the results of specific capacity measurement. With the same amount of TiO 2 added on the electrode, the C s,t of COM-TOC-20 was 166.0 F g −1 , which was less    Pores on the Vulcan XC-72 surface are generally hydrophobic and are inaccessible for aqueous solution and ion transport. Therefore, hydrophobic pores are both inaccessible for ions and electrochemical inactive. 25 The introduction of hydrophilic TiO 2 produces faradaic pseudo-capacitance. Vulcan XC-72 has a high specific area and excellent conductivity. Addition of hydrophilic TiO 2 further improves the hydrophilic property and the double-layer capacitance of the mixture of TiO 2 and Vulcan XC-72. Figure 7 presents the relationship between the specific capacitance and the potential scanning rate of different composite electrodes. The figure shows that the composite electrode has higher specific capacitance than that of Vulcan XC-72. The specific capacitance of a given sample is independent of the scanning rate. The optimal TiO 2 loading is approximately 20 wt% in this study. Charge-discharge test.-Our previous study 19 indicates that H 2 evolution occurs more easily in raw carbon electrode than in acidicor thermal-treated carbon electrode. The CV at potential < −0.5 V against Ag/AgCl, as shown on Fig. 8, confirmed that H 2 evolution easily occurred in the raw carbon felt electrode in the V 2+ /V 3+ redox reaction. After adding TiO 2 , H 2 evolution reaction rate is suppressed. Addition of TiO 2 enhances the hydrophilic ability of carbon felt and the electrochemically active area of composite electrode is increased, as indicated by the hydrogen adsorption/desorption peaks in Fig. 8. As a result, the hydrophilic TiO 2 was added to the negative electrode in subsequent experiments to prevent H 2 evolution.
A single cell with four different electrode materials was analyzed to determine the effect of TiO 2 /C-modified felt on charge-discharge performance. Raw felt was used as positive electrode for all cells tested. The negative electrode of cell A used raw felt, whereas COM-TOC-20 and HM-TOC-20-modified felt served as the negative electrodes for cells B and C, respectively. The thickness of all electrodes was 6 mm. We investigated the effect of electrolyte composition on the chargedischarge curves, as given in Fig. 9a. Electrolytes with three different compositions were tested, and their compositions were listed in Table II. The cutoff voltage during charging was 1.9 V, and the cutoff voltage during discharge was 0.8 V. Both the charging and discharging times were increased as the vanadium concentration in the solution increased, as shown in Fig. 9a. The voltage drop when current switched from charge to discharge was decreased as the vanadium concentration increased, and the theoretical capacities calculated from the vanadium concentration were listed in Table II. Both capacities measured from Fig. 8a and from the theoretical calculation are listed in Table II. As an example, the theoretical and experimental capacities of solution 3 were 5226 mAh and 3976 mAh, respectively. The 3976 mAh was calculated from Fig. 9a, 5000 mA × 2863 s × 3600 −1 s = 3976 mAh. Experimental capacities depend  linearly on vanadium concentration. The ratio of the experimental capacity to the theoretical capacity is also listed on the same Table II. The concentrated solution (solution 3) had the highest ratio among three solutions tested. This result may be attributed the fact that the cell was not fully charged. To avoid possible water electrolysis during the charge period, the upper limit of cell voltage during charge was set at 1.9 V. At fixed current density and electrolyte flow rate, high   concentration over-potential is expected for cells containing low vanadium concentration. At 1.9 V, the state of charge (SOC) of the cells containing low vanadium concentration is lower than that containing high vanadium concentration. Figure 9b represents the charge-discharge curves of cells at a constant current density of 200 mA cm −2 . An abrupt cell voltage drop was detected when the current was reversed from charge to discharge. Cell internal resistance R (in ohm cm −2 ) can be estimated from the cell voltage drop during current switching from charge to discharge (Fig. 9b). The IR drop of the different electrodes follows the order: cell C (0.56 V) < cell B (0.60 V) < cell A (0.64 V), as shown in Fig. 9b (curves A to C at 200 mA cm −2 ). Internal resistance values calculated from Eq. 2 26 are in the order of cell C (1.4 cm −2 ) < cell B (1.5 cm −2 ) < cell A (1.6 cm −2 ), as shown in Fig. 9b (curves A to C at 200 mA cm −2 ).
where I is the current value, V C is the average charging voltage, V D is the average discharging voltage, and A is the geometrical area of the electrode. The electrode active area among cells A, B, and C was in the order of cell C > cell B > cell A, as given by the capacitance measurement in Table I. The "apparent current density" during charging-discharging cycle was at 200 mA cm −2 . Considering the actual electrode area, the "actual current densities" applied on each cell was different and in the order of cells A > cell B > cell C. Hence, the "apparent internal resistance" is in the order of cell A > cell B > cell C. Figures 10a to 10c present the charge-discharge cycling curves of cells A to C at a constant charge-discharge current density of 200 mA cm −2 . Table III lists the calculated coulomb, voltage, and energy efficiencies of these cells. The efficiency of each cell with added TiO 2 follows the same order: cell C > cell B > cell A. This observation is consistent with the specific capacitance results, where the C s,t value of each sample follows the order HM-TOC-20 > COM-TOC-20 > XC-72. Wetting of negative electrode and small particle size (9.1 nm versus 16.1 nm) of HM-TOC-20 results in better coulomb, voltage, and overall efficiency than that of the raw felt and COM-TOC-20 negative electrodes.
The results demonstrate that VRFB performance can be improved by the addition of TiO 2 nanoparticles onto the negative electrode layer. In summary, the cell performance with a negative electrode layer containing TiO 2 nanoparticles is primarily determined by a competition mechanism between the positive effect (wettability enhancement of the cathode active area) and negative effect (TiO 2 nanoparticle aggregation and increased intrinsic resistance). The highest coulomb, voltage, and energy efficiency of cell C with added HM-TOC-20 were 91.5%, 73.1%, and 66.8%, respectively.

Conclusions
Spherical TiO 2 nanoparticles of average size 9.1 ± 2.0 nm were successfully prepared by using a sol-gel method. The homemade TiO 2 exhibited excellent hydrophilicity, by which the water contact angle of the 20 wt% HM-TiO 2 nanoparticles (52 • ) is smaller than that of the 20 wt% commercial TiO 2 nanoparticles (88 • ) and the carbon black XC-72 (132 • ). The single cell where negative electrode containing hydrophilic TiO 2 nanoparticles had larger electrochemical active area, lower internal resistance, higher specific capacitance, and energy storage efficiency than the single cell where negative electrode containing commercial TiO 2 nanoparticles or without TiO 2 nanoparticles. A 90% coulomb efficiency, 73% voltage efficiency, and 65.3% energy efficiency could be achieved at 200 mA cm −2 with electrolyte containing 3 M V(V)/V(III), 2.0 M H 2 SO 4 .