In Situ Reliability Investigation of All-Vanadium Redox Flow Batteries by a Stable Reference Electrode

It has been reported that the key parameter to stabilize the hydrogen evolution reaction at the DHE ²¹ is the adjustment of the current flow through the cell. Experiments started with adjusting the resistor in the circuit to tune the current flowing through the DHE. Figure 2 shows the charge-discharge curves of a full cell and its individual electrodes (cathode or anode vs DHE) with resistors of 3.3 and 1.0 MΩ. The selection of the resistors was aimed at approaching the optimal current (approximately 6 μA or less as reported ^21,25 ) through the Pt electrode of the DHE. As an example, a 9 V battery with 3.3 or 1 MΩ resistors should lead to a current of 2–9 μA assuming the contact resistance is insignificant.

Zoom In Zoom Out Reset image size

The accuracy of the DHE for both resistor-based systems (Figs. 2a and 2b) is validated since the potential difference between the cathode (vs DHE) and anode (vs DHE) is the same as the measured full cell potential (cathode vs anode). ²¹ In addition, when the resistance was decreased from 3.3 MΩ to 1.0 MΩ, the cathode or anode voltage (vs DHE) was increased by ca. 0.4 V that reflects the decrease of the DHE voltage by ca. 0.4 V. Therefore, the reported resistor effect on the reference DHE is validated in our experiment. Considering the cathode reaction (VO₂ ⁺ + 2H⁺ + e⁻ ↔ VO²⁺ +H₂O) of a vanadium redox flow battery (VRFB) with an equilibrium potential E° of 1.00 V vs NHE, the cathode voltage profile between 1–1.2 V vs DHE in Fig. 2b indicates the DHE potential approaches 0 V when using a 1.0 MΩ resistor. Similar cathode discharge potential has been reported by Choi in their DHE design. ²¹ The difference in Pt wire potential using the two resistors indicates the reactions occurring at the Pt wires are different. With 3.3 MΩ, the current passing through the Pt wire is relatively small and the vanadium redox reaction plays a dominant role which makes the Pt wire potential increase by 0.4 V. For the 1.0 MΩ case, the current passing through the Pt wire is much higher and the vanadium present at the vicinity of Pt wire might be consumed faster than normal. The hydrogen evolution reaction (HER) dominates the Pt wire potential which leads to a stable and reliable DHE reference electrode.

The design of the DHE was further investigated to improve the stability by examining the shape of the Pt wire. A curved Pt wire electrode was used to replace the straight Pt wire inserted into membranes as illustrated in Fig. 1b. The curved Pt wire was utilized to maximize the contact area between the Pt wire and the membrane and to further improve the stability of the DHE. As discussed earlier, the straight Pt wires in conjunction with a 1.0 MΩ resistor for the DHE has demonstrated a stable system in the initial DHE setup, (Fig. 2b). As shown in Fig. 3a, the voltage profile of the cathode or anode (vs DHE) continues to shift during cycling of the cell when a curved Pt wire with a 1.0 MΩ resistor is used. The initial discharge voltage of the cathode increases gradually from 0.88 V to 1.15 V vs DHE in the first four cycles which indicates that the voltage of DHE decreases continually and approaches 0 V where the hydrogen evolution reaction happens. Since the DHE (Pt wire) was inserted into the membranes, the shifting voltage of DHE reflects internal (electrolyte) environmental changes during the initial charging-discharging. This behavior is probably attributable to the depletion of vanadium ions in the vicinity of the Pt wire which causes a gradual increase in the proton transport properties of the Nafion matrix surrounding the DHE or decreased in the polarization at the Pt wire. ²¹

Zoom In Zoom Out Reset image size

To further develop a stable DHE using a curved Pt wire, a resistor with a lower resistance, 820 kΩ, was used. As shown in Fig. 3b, the cathode or anode discharge voltage profile (that started close to 1.2 V or −0.3 V vs DHE) stayed consistent during initial cycling which again indicates a stable DHE. As discussed earlier, the straight Pt wire based DHE was shown to have a stable HER with a 1.0 MΩ resistor whereas the curved Pt wire based DHE resulted in a stable HER with a slightly lower resistor value of 820 kΩ. The resistor value differences are likely attributed to an increased in the contact resistance between the curved (longer) Pt wire and the membrane. This results in a lower resistance needed to achieve an optimum current for the HER in the DHE circuit. The results demonstrate a stable DHE that can be achieved by optimized design and an adjustable resistor.

As there is co-existence of H⁺ and vanadium species in the electrolyte, the competition of HER and vanadium redox reaction occurs on the DHE. ²¹ Thus the acid (H⁺) concentration might affect the stability of DHE, which has not been reported in the open literature, however. For a stable DHE, the HER (Eq. i) continuously occurs in which the consistent supply of protons is required since protons are continuously consumed at the vicinity of the Pt wire. If the acid concentration is low, HER will compete with vanadium redox reaction, and a mixed potential might occur on the Pt wire, leading to an unstable and/or unreliable DHE potential.

$$\begin{eqnarray}&&2{{\rm{H}}}^{+}+2{{\rm{e}}}^{-}\to {{\rm{H}}}_{2}\,E^\circ =0.00\,{\rm{V}}\end{eqnarray} \tag{ i }$$

To investigate the effect of acid concentration on the DHE performance, a flow cell with a lower acid concentration (3.3 M H₂SO₄) based electrolyte was assembled with a DHE using the resistor of 820 kΩ. The voltage profiles of the full cell and the individual electrodes (vs DHE) are shown in Fig. S1a (available online at stacks.iop.org/JES/167/160541/mmedia). When the concentration of H₂SO₄ was decreased from 4 M to 3.3 M in the electrolyte, the cathode or anode voltage (vs DHE) decreased by ca. 0.35 V, indicating the DHE voltage was increased by ca. 0.35 V. In the previous 4 M H₂SO₄ based electrolyte, the DHE with a resistor of 820 kΩ demonstrated a stable reference electrode that approaches 0 V with continuous HER. Here the DHE voltage in the cell with 3.3 M H₂SO₄ based electrolyte shifts to ca. 0.35 V which is close to the equilibrium potential E° for the V (IV)/V (III) reaction (Eq. ii).

$$\begin{eqnarray}&&{{\rm{VO}}}^{2+}+2{{\rm{H}}}^{+}+{{\rm{e}}}^{-}\to {{\rm{V}}}^{3+}+{{\rm{H}}}_{2}{\rm{O}}\,E^\circ =0.34\,{\rm{V}}\end{eqnarray} \tag{ ii }$$

The results indicate that the (negative) Pt electrode in DHE might be dominated by V (IV)/V (III) reaction in a lower acid concentration of electrolyte which probably leads to the inaccuracy and instability of the DHE. Thus the acid (proton) concentration should be considered as a factor to influence the reaction occurring on the DHE. A higher proton concentration at the Pt wire of DHE enhances the chances of the HER happening which improves the accuracy and stability of the DHE.

To achieve a stable DHE in the cell with 3.3 M H₂SO₄, the resistor was adjusted to a lower value so that the HER could happen and be maintained on the Pt electrode at a higher current. As shown in Fig. S1b, the cathode (or anode) voltage increased and remained stable at 1.2 V (or −0.3 V) vs DHE when the value of the resistor was decreased to 511 kΩ, indicating the DHE approached 0 V with a HER. A lower resistor leads to a higher current in the DHE, where a faster consumption of vanadium species occurs at the vicinity of the Pt wire and the HER dominates the Pt wire potential. In addition, the low acid concentration decreases the membrane conductivity, which also requires resistor with a lower value to maintain constant HER at the Pt wire. A more systematical study on the acid (proton) effect on the DHE potential, such as splitting vanadium effect, pH effect due to proton concentration on the potential, and conductivity effect due to proton concentration will be a subject of ongoing investigation in our laboratory.

Figure 4 shows the long-term cycling performance of the all-vanadium RFB using 1.7 M vanadium, and 4.0 M H₂SO₄ based electrolyte and a Nafion N212 double membranes with the DHE inserted in the middle of the membranes.

Zoom In Zoom Out Reset image size

As shown in Fig. 4a, the charge and discharge capacity continuously decreased from 1.203 and 1.135 Ah in the 1st cycle to 0.5934 and 0.5887 Ah in the 100th cycle respectively; discharge capacity retention is approximately 51.9%. The performance efficiencies (CE-Coulombic efficiency, VE-voltage efficiency and EE-energy efficiency) in Fig. 4b, however, exhibited much higher retention than the capacity, with the CE, VE and EE of 94.3, 90.8, and 85.6% in the 1st cycle and 99.2, 78.8, and 78.2% in the 100th cycle, respectively. The cell achieved a stable and even slightly increasing Coulombic efficiency of 98%–99% (except for the 1st cycle). And the EE, very close to the VE, achieved a high retention 91.3% for 100 cycles. Another phenomenon is that in the overall 100 cycles, both the capacity and the efficiencies remained nearly constant in the initial fifteen cycles. After 15 cycles, the capacity dropped significantly along with the VE (and EE). This indicated that the cell likely starts to degrade after the initial fifteen cycles.

The charge-discharge voltage profile over cycling in Fig. 4c further demonstrated that capacity fading was associated with increasing the overpotential of the cell. This is typically indicated by the continuously increasing charge voltage (in black) and decreasing discharge voltage (in red) as cycling proceeded. It has been reported that the increasing cell polarization with cycling possibly results from decreasing proton conductivity of the Nafion membrane over time and the increase in concentration polarization due to significant crossover. ^6,26

Utilizing the developed DHE reference electrode inserted between the two Nafion N212 membranes, the charge and discharge voltage profiles of the full cell and individual electrodes (vs DHE) during 100 cycles are shown in Figs. 5a–5b. As shown in the charge process of Fig. 5a, the voltage curves of the full cell continuously increase during cycling which indicates the overpotential continues to increase as discussed previously. The voltage curves of individual electrodes (cathode or anode vs DHE) present quite different behavior throughout the 100 cycles. The cathode curve continuously increases with cycling (∼0.15 V) which exhibits a similar trend of the full cell (∼0.17 V). The anode cycling curves, however, remain relatively consistent with cycling. The observed phenomenon of single electrodes indicates the performance degradation (overpotential increasing) during the charge process of the cell was dominated by the cathode side.

Zoom In Zoom Out Reset image size

For the discharge process shown in Fig. 5b, the voltage curves of the full cell shifted with cycling in the opposite direction of the charge process. Shifting of the voltage curves, however, still indicates an increase in overpotential with cycling. Unlike the charge process, the individual electrodes in the discharge process did not show consistent behavior throughout the whole 100 cycles. In the initial 45 cycles, the voltage curve of the full cell decreased gradually by 0.1 V during which time the anode curve increased by 0.1 V and the cathode curve was relatively constant. It can be deduced that the overpotential increase in the initial 45 cycles is mostly derived from the anode contribution. In the 46th−100th cycles, the voltage curve of the full cell further deceased by 0.1 V while the cathode curve decreased by 0.05 V and the anode curve increased by 0.05 V, indicating the performance fading of the full cell is from the degradation of both cathode and anode with a similar contribution. Thus, in the discharge process of the cell, the anode side dominates the initial ∼50 cycles of performance degradation whereas the cathode degradation is insignificant but plays a more significant role in later cycling (>∼50 cycles).

Prior art using DHE in RFBs suffered issues associated with potential drifting over long-term cycling. Previous work ²¹ found that the average potentials of cathode and anode (vs DHE) increased by 4.5 mV after 10 cycles and 76 mV after 50 cycles, this indicated a significant drift of the DHE potential to a lower potential with cycling that was associated with either a gradual reduction in proton transport for the Nafion matrix surrounding the DHE or an increase in the passivation at the Pt electrode. ²¹ The reported issue of the drifting of the DHE potential with long-term cycling was not observed in this work. Although the voltage curves of the cathode (vs DHE) continuously increase in Fig. 5a, the anode curve remained consistent over cycling, indicating the voltage shift was originally from degradation (overpotential increasing) in the cathode. The voltage curve of the full cell shows a matching trend to the cathode behavior. Moreover, the observed gradual decreasing and increasing of the voltage curves of the cathode and anode respectively in the discharge process (Fig. 5b) were associated with the decrease in the voltage curves of the full cell. The potential drifting associated with a DHE reference electrode will always lead to the potential increasing (or decreasing) of both cathode and anode (vs DHE) simultaneously. ²¹ A stable DHE was demonstrated for long term cycling of VRFBs in this work. This is the first study to report a stable DHE that can go through hundreds of cycles for a VRFB, and it is also the first study to disclose the individual electrode contribution to the performance degradation during long-term cycling of a VRFB by using a stable and reliable DHE.

To further explain why the newly developed DHE is much more accurate and stable than the prior art, the main factors to influence the DHE should be taken into consideration, which turned out to be the DHE design, acid (proton) concentration, and the cell size. Compared with the prior art of DHE, the improvement in this work for the DHE design optimization is the contact enhancement between DHE (Pt wires) and the membranes, such as using different (curved) shape of Pt electrode to increase its contact area with the membranes, as well as introducing a lamination process combined with Nafion solution soaking (shown in Experimental). Moreover, a highly concentrated acid (4 M H₂SO₄) environment is another main reason to improve the stability of DHE as discussed above, whereas a relatively lower concentration of acid (e.g. 2–3 M) was mostly used in the previous studies. ²¹ In addition, the effect of cell (electrode) size, which was not discussed in detail here, was probably hardly to be ignored. Our recent study revealed that DHE might be more stable in a relatively smaller cell (e.g. 10 cm² for the electrode area in this work) than that in a bigger cell (e.g. 49 cm²), possibly due to the non-uniform current distribution. ²⁷ The stability of DHE in a scaled cell will be a subject of ongoing study in our laboratory.

Figure 6a–6b shows the selected overpotentials as a function of cycle numbers for the full cell and individual electrodes (vs DHE) at the end of charge and discharge. The relative proportion of the overpotential for individual electrodes was also plotted in Figs. 6 c–6d. The overpotential (ΔV) was calculated from the difference between the voltage at the operation current (50 mA cm⁻²) and the OCV at the end of charge (1.63 V) or discharge (0.5 V). In general, the overpotential of the full cell gradually increased over cycling at the end of charge or discharge. Relatively higher overpotentials were observed at the end of discharge.

Zoom In Zoom Out Reset image size

At the end of charge in Figs. 6a, 6c, the overpotential increased with cycling in both the cathode and anode, whereas the overpotential of the anode was much higher than that of the cathode, indicating the anode reaction limited the charge capacity. ²¹ After approximately 70 cycles, the overpotential of the anode reached its highest point and then reached a plateau; the overpotential of the cathode gradually increased over cycling. At the end of discharge in Figs. 6b, 6d, the overpotential of the anode increased while the overpotential of the cathode slightly decreased until the 25th cycle where the anode and the cathode reached the highest and lowest overpotentials respectively. Beyond 25 cycles, the overpotentials had opposite trends: the anode overpotential decreased and the cathode overpotential increased until they almost reached a similar overpotential near 100 cycles.

The results indicate that the anode reaction limited the charge capacity of the cell throughout 100 cycles, and it also limits the discharge capacity in the initial 50 cycles. The cathode reaction more strongly influenced the discharge capacity (31%–49% ΔV in Fig. 6d) than in the charge capacity (20%–39% ΔV in Fig. 6c). In addition, the cathode overpotential contributed a more significant role in the capacity limitation as the cycles proceeded (especially after ∼50 cycles). For example, the cathode showed almost the same influence as the anode on discharge capacity limitation at the 93rd cycle. It has been reported that the anode reaction of a VRFB limits both the charge and discharge capacity in the initial cycles, ²¹ which agrees well with our result. This phenomenon is probably associated with the (vanadium ion based) electrolyte crossover issue during cycling that leads to the imbalanced vanadium active species and the asymmetrical valence of vanadium ions in positive and negative electrolytes. ⁶ This work was first to disclose the role of individual electrodes in limiting the charge or discharge capacity over long-term cycling. The results were achieved by using a stable DHE reference electrode and in situ monitoring of the cathode and anode throughout 100 cycles of cell charging-discharging.

Using the developed stable DHE, the electrochemical impedance spectroscopy (EIS) of the full cell and individual electrodes (cathode or anode vs DHE) was also measured during long-term cycling using the stable DHE. The impedance distribution across the whole flow cell can be dissected and the contributions from positive and negative electrodes can be decoupled. Shown in Fig. 7 are typical Nyquist plots at the end of charge (a–b) or discharge (c–d) with different cycles. The theoretical value (Z_cathode + Z_anode) shown in Figs. 7a and 7c, the sum of measured cathode and anode impedance, was in a good agreement with the experimental result of the full cell; this behavior demonstrates the good accuracy of the DHE in the EIS measurement. ²² The measured full cell, cathode and anode spectra demonstrated different features during cycling. At high frequencies in the full cell and anode spectra, there is an inductance resulting in negative –Z_IM, whereas the cathode exhibits a high frequency arc which indicates a different process, e.g. membrane-electrode interface resistance, etc. The impedance measurements consist of a semicircle at higher frequencies followed by flat and straight lines at lower frequencies. The semicircle is indicative of the charge transfer process; the intercept at higher frequency at the real axis represents the Ohmic resistance (R_oh). The diameter of the semicircle represents the charge transfer resistance (R_CT). ⁴ The frequency range for the semicircle is also different for the full cell, the anode and the cathode. For the full cell and the anode, the semicircle appeared at high frequency range whereas the semicircle appears at medium to low frequency range for the cathode. The difference in frequency range indicates the difference in the double layer capacitance or constant phase element (CPE) at the electrode/electrolyte interface accompanying the charge transfer process. The high frequency range of the semicircle for the full cell and anode indicates small effective double layer capacitance at the electrode/electrolyte interface while the low frequency range in cathode indicates a large effective double layer capacitance for the cathode electrode/electrolyte interface. This behavior suggests that the electrode/electrolyte interface might be different for the cathode and anode. The straight line at the lowest frequency range is caused by a Warburg resistance due to the electroactive species transport. The features between the straight line and semicircle are likely due to other processes, e.g. non-ideal charge transfer process (R_CT perhaps performs like a CPE) and/or other processes accompanied with charge transfer resistance such as absorption etc. ²⁶

Zoom In Zoom Out Reset image size

The ohmic and charge-transfer resistances of the full cell and individual electrodes were plotted as a function of cycle number in Fig. 8 using Zview software. As shown in Figs. 8a, 8d, the R_oh of the full cell increased significantly with cycling for both the charge and discharge process, which was derived from the continuous increase of the R_oh in both the cathode and anode where much larger ohmic resistance was observed in the anode than the cathode, and the cathode showed the higher increase rate than the cathode after 50 cycles. The R_CT of the full cell, however, remained stable in the initial cycles and then decreased gradually over cycling, as shown in Figs. 8b, 8e. In particular, much higher R_CT was observed at the end of discharge than charge. Here the individual electrodes showed quite different behaviors: the anode started to decrease after initial cycles that was consistent with the full cell while the cathode tended to the opposite direction.

Zoom In Zoom Out Reset image size

The relative proportions of R_oh and R_CT for individual electrodes were further plotted in Figs. 8c, 8f. At the end of charge, the total R_oh (combining cathode and anode) played a dominate and gradually increasing role (from 79% to 92%) in the total resistance while the R_CT was insignificant and kept decreasing (from 21% to 8%) throughout 100 cycles. In comparison, the R_CT dominated at the end of discharge during initial cycling and then slowly decreased (from 64% to 33%) until 70 cycles followed by slight increase (to 35%) near 100 cycles. The R_oh contribution increased as cycling increased (from 36% to 67%) until 70 cycles followed by slight decrease (to 65%) in the end.

It can be concluded that cell impedance is greater at the negative side at the end of charge and discharge process, which suggests that the rate limiting step in a VRFB is at the anode side. These results are in agreement with previously published results (particular in the initial 50 cycles). ¹⁹ In general, the impedance of the positive electrode increased relatively slow before 50 cycles, but had a greater contribution to the performance degradation over long-term cycling. The results are in good agreement with the overpotential trend shown in Fig. 6.