Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

A polyoxometalate redox flow battery: functionality and upscale

A polyoxometalate redox flow battery: functionality and upscale While redox flow batteries carry a large potential for electricity storage, specifically for regenerative energies, the current technology-prone system—the all-vanadium redox flow battery—exhibits two major disadvantages: low energy and low power densities. Polyoxometalates have the potential to mitigate both effects. In this publication, 4– the operation of a polyoxometalate redox flow battery was demonstrated for the polyoxoanions [SiW O ] 12 40 9– (SiW ) in the anolyte and [PV O ] (PV ) in the catholyte. Emphasis was laid on comparing to which extent 12 14 42 14 an upscale from 25 to 1400 cm membrane area may impede efficiency and operational parameters. Results demonstrated that the operation of the large cell for close to 3 months did not diminish operation and the stability of polyoxometalates was unaltered. Keywords: energy storage; redox flow battery; polyoxometalates; upscale Current challenges for RFB chemistry are their power Introduction density, their energy density and their costs [1–4]. Redox flow batteries (RFBs) are one of the few options to According to Arenas et  al., the development of new RFB store energy from intermittent renewable-energy sources technology can be described as follows []. 5 As a first step, such as wind and solar electrochemically. The concept the fundamental electrochemistry is explored. Physical of the RFB has several advantages [1–3], such as the in- and chemical properties as well as important parameters, dependent scalability of power and energy content. The such as the equilibrium potential U and the electron- former is determined by the size of the power converter, transfer constant k of the involved species, are investi- whereas the latter is given by the energy density of the gated (Stage 1a). Following this, the cycling behaviour electrolyte and the size of the tanks. Furthermore, as op- and stability of the novel RFB electrolytes are assessed in posed to other battery types, the electrodes themselves H-cells or laboratory-scale flow cells (Stage 1b). The next are not redox-active and do not undergo conversion, inter - step then involves optimization of the electrochemical calation or alloying reactions, which often lead to degrad- power converter according to the chemical and phys- ation. Instead, the energy is stored in redox-active species ical properties of the RFB electrochemistry. Potential and that are dissolved in the electrolyte. Received: 15 January, 2019; Accepted: 19 June, 2019 © The Author(s) 2019. Published by Oxford University Press on behalf of National Institute of Clean-and-Low-Carbon Energy This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http:// creativecommons.org/licenses/by-nc/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, 278 provided the original work is properly cited. For commercial re-use, please contact journals.permissions@oup.com Downloaded from https://academic.oup.com/ce/article-abstract/3/4/278/5550272 by DeepDyve user on 10 December 2019 Friedl et al. | 279 2+ 3+ 2+ + current distribution, hydrodynamics, mass transport and transfer for the V/V and VO /VO redox reactions is −6 −1 cell geometry are optimized (Stage 2a), followed by the slow (k ≈ 10 cm s ) [7, 20], whereas electron transfer −2 −1 development of a pilot stack and a prototype RFB system for the POMs is facile (k ≈ 10 cm s ) [15, 21]. The con- (Stage 2b). The final stage according to Arenas et  al. is a sequence is that the charge-transfer resistance R for CT commercial implementation, which involves fabrication, the POMs is considerably lower than for the VRFB. This testing, maintenance, marketing and many other tasks leads to consequences for the design of the power con- outside the realm of electrochemical engineering (Stage 3). verter, as the power-converter design aims to reduce the The properties of the used redox shuttles and employed total resistance R, which is the sum of R , R and the tot CT diff solvents determine how the electron transfer and mass Ohmic losses R . The resistance of the cell determines Ohm transport proceed within the power converter. Therefore, the currents that can be dr I awn from a battery for a given ideally, the power converter should be optimized for the voltage efficiency η : specific redox shuttle. However, while many novel redox ΔU − R I tot η = electrochemistries were proposed for RFBs in recent years, ΔU + R I tot the literature focuses mostly on Stages 1a and 1b [ ]. A  6 re- with cell voltagΔ e U. In this study, we describe our findings cent literature review by our group showed that, of 24 of a novel flow battery electrochemistry for a laboratory- published redox electrochemistries for RFBs, only 4 were type cell of 25 cm and the scale-up to a commercial cell demonstrated with a power rating of more than 100 W of 1400 cm . Details on the basic electrochemistry of this [3]. These are the all-vanadium RFB (VRFB) [, 78], the zinc- novel RFB system can be found in a recent publication [15]. bromine cell [9 , 10], the iron-chromium cell [1112 , ] and the In this paper, we are describing the operational parameters bromine-polysulphide RFB [1 , 13, 14]. Of the remaining 20, 7 such as charge–discharge cycles, coulombic and energy effi- were tested in H-cell configurations only, all the others ex- ciencies, and stability of the electrolytes. Typical values such cept for 2 in cells with less than 1-W power output. So, there as R , R and R are determined using electrochemical im- Ohm CT diff is currently a knowledge gap between the laboratory Stage pedance spectroscopy (EIS) and compared for the two cells. 1b and the electrochemical technology Stage 2a. This is one of the reasons for showing how a scale-up of a novel electro- 2 2 chemistry from a 25-cm cell to a cell of 1400 cm may affect 1 Materials important performance parameters of the RFB. The ap- 1.1 Cells, sensors and parameters proach in our laboratory is to use polyoxometalates as redox shuttles. The redox electrochemistry described here is a Three electrode measurements were performed in polyoxometalate (POM) system, utilizing the polyoxoanions custom-built glass cells with a polished glassy carbon 4– 9– 2 [SiW O ] (SiW ) in the anolyte and [PV O ] (PV ) in the working electrode (surface area A  =  0.02  cm ), a gold wire 12 40 12 14 42 14 catholyte as nano-sized charge carriers [15]. (diameter d  =  0.5  mm) counter electrode and a mercury/ Due to their structural and chemical properties, these mercurous sulphate reference electrode in 1 M H SO (MSE, 2 4 metal-oxygen clusters have a number of advantages for 0.668 V vs. Standard Hydrogen Electrode [SHE]). Prior to the energy-storage applications [15]. First, due to their large measurements, the electrolyte was purged with nitrogen size, the interaction of the redox centres of the POM with and the cell was kept under nitrogen pressure during the solvent molecules is small. Therefore, the outer-sphere re- experiment. A  Bio-Logic SP-300 potentiostat was used for organization energy of the electron transfer is low, enabling control and data acquisition. fast kinetics and thus high power densities [1617 , ]. Also, The small flow cell used was a commercial cell (C-Tech the inner-sphere reorganization energy is low due to the 5x5, surface area A  =  25  cm ). Graphite felts (GFD, SGL added electrons often being delocalized [18], causing only Carbon) were used as electrodes and pre-treated at a minimal change in coordination upon reduction or oxi- 400°C for 24 h in a laboratory atmosphere. In the cell, the dation [19]. This behaviour adds to the fast kinetics and 4.6-mm-thick electrodes were compressed to 3.5  mm. As thus high power densities. Their nature as large anions the membrane, a cation-exchange membrane (FUMASEP— also prevents POMs from penetrating commercial cation- F-1075-PK) was used. During the experiment, the cell and exchange membranes [15]. Size exclusion and electrostatic the pump with tubing were kept in a polycarbonate box repulsion prohibit cross-over and mixing of the active spe- purged with nitrogen. The peristaltic pump could supply –1 cies. Furthermore, some POMs exhibit high solubility, e.g. flow rates of 12–150 ml min . Charge and discharge cycles –1 a concentration of 0.875 mol L can be obtained for SiW were measured using a Bio-Logic BCS-810 battery tester. [15]. With multiple redox-active centres per molecule, this EIS measurements were performed using the Bio-Logic enables a high energy density in the battery. Moreover, SP-300 (maximum supplied f = 7 MHz, measurement up to both SiW and PV are stable during the operation of the 200 kHz), as this could apply higher frequencies than the 12 14 battery. Losses in capacity only seem to stem from a para- battery tester (f up to 10 kHz). sitic reaction with residual oxygen, which could be avoided The large flow cell used was a commercial cell (J. with an improved airtight setup [15]. Schmalz GmbH, surface area A  =  1400  cm ), originally de- Comparing the asymmetric POM system with the signed for the VRFB chemistry. The same membrane as in VRFB, one main difference is apparent: the electron the small cell was used. The flow rate was adjusted via the Downloaded from https://academic.oup.com/ce/article-abstract/3/4/278/5550272 by DeepDyve user on 10 December 2019 280 | Clean Energy, 2019, Vol. 3, No. 4 pressure drop in the cell. Measured electrolyte flow and anolyte. PV was prepared following the synthesis for pressure drop were always proportional for both anolyte Na [H PV O ]∙28 H O described in ref. [22]. To prepare the 5 4 14 42 2 –1 and catholyte. Flow rates from 450 to 1872  ml min were catholyte, Na[H PV O ]∙28 H O was dissolved in DI water 5 4 14 42 2 determined. Both the charge–discharge cycles and EIS were with 1 M LiCl and pre-reduced by addition of hydrazine measured using a Bio-Logic BCS-815. In operando measure- (Sigma-Aldrich). This reduction was necessary, as other - ments of the pH values in the tanks were performed using wise both anolyte and catholyte would have been fully oxi- Unitrode pH sensors from Metrohm. For data acquisition of dized at the start of the battery operation, making cycling the pH values, a National Instruments cDAQ-9175 was used impossible. Cyclic voltammograms (CVs) of 1 mM PV (red in combination with a National Instruments Module 9205 line) and 1 mM SiW (blue line) are shown in Fig. 2a. PV 12 14 and a current loop converter CLC-01 from company pro- exhibits a multi-electron redox reaction in the range from viding electronic measurementF . ig. 1 shows a photograph 0.2 to 0.7 V vs. SHE. Experiments in a symmetric flow bat- of a 1400-cm cell with its periphery and containment. The tery (PV as both anolyte and catholyte) have shown that cell itself is fixed in the top part of an acrylic glass container, PV transfers at least seven electrons [15 SiW ]. shows two 14 12 θ,1 θ,2 the pumps are installed underneath and the electrolyte redox reactions at U  = 0.0 V vs. SHE and U   =  –0.21 SiW SiW 12 12 θ,3 tanks are to the left and right of the pumps. The electrolyte V vs. SHE. A  third redox wave centred on U   =  –0.37 V SiW tanks are continuously purged with nitrogen to avoid oxi- vs. SHE is a two-electron redox reaction as opposed to the dation of the reduced redox species by atmospheric oxygen. previous two, which are one-electron transfers. However, Tables 1 and 2 give an overview of the parameters used the third reaction is not used in the flow battery because in this work. These quantities are used to discuss and com- the potential at which it takes place leads to an irrevers- pare the performance of the two cells. ible dimerization of the POM onto the carbon electrodes [23]. The polyhedral structures shown in Fig. 2b and Fig. 2c reveal that both SiW and PV are of the Keggin structure, 12 14 1.2 Electrolytes and their electrochemistry with two additional V–O caps for the polyoxovanadate [24]. The electrochemistry and preparation of SiW and PV 12 14 were described in detail in ref. [15]. In short, SiW was 2 Results in a 25-cm cell bought from Sigma-Aldrich as tungstosilic acid (H SiW O ) 4 12 40 and dissolved in de-ionised (DI) water with 1 M LiCl. Diluted In order to obtain data from a polyoxometalate redox flow LiOH was used for pH adjustment to 1.8 to prepare the battery, a commercial cell from C-Tech with a membrane area of 25 cm was used. Heat-treated, compressed graphite felts (Sigracell GFD 4.6 EA) were used as electrodes as described earlier and a FUMASEP—F-1075-PK membrane was employed. The electrolytes comprised 80 mM SiW for the anolyte and 80 mM PV for the catholyte in 1 M LiCl each. In charge–dis- charge cycles, the 25-cm cell exhibited a coulombic efficiency 2 2 25cm 25cm η of 96% and an energy efficiency η of 64%. The the- C E theo,25cm oretic capacity was Q  = 214 mAh, of which 90% was dch reached. A study of the influence of the electrolyte concentra- tion on the diffusion resistance R was conducted. diff In Fig. 3, R is compared for the 25-cm cell with 80 mM diff SiW and 80 mM PV as electrolytes (blue data) and a setup 12 14 in which the anolyte was 600  mM SiW and the catholyte 300  mM PV (red data). To balance the charge, the PV was 14 14 pre-reduced by four electrons with hydrazine (instead of a pre-reduction by two electrons as for equal concentra- tions of POMs). The volume flow v has been recalculated to ˙ ˙ normalized mass flow c (see Table 2) by multiplying v with the used concentrations of catholyte. It can clearly be seen that the mass flow, that is the amount of unreacted active material that is brought into the cell per minute, determines .R diff 3 Results in a 1400-cm cell 3.1 Charge and discharge Fig. 1 Photograph of 1400-cm cell with periphery (tanks, pumps, An electrolyte solution with 80 mM SiW was prepared as tubing) and containment anolyte; a solution of 80 mM PV was used as catholyte. 14 Downloaded from https://academic.oup.com/ce/article-abstract/3/4/278/5550272 by DeepDyve user on 10 December 2019 Friedl et al. | 281 Table 1 Experimental parameters obtained or set in the measurements Experimental parameters Symbol Obtained from Surface area of cell A (cm ) Geometric area cell/membrane Charge time t (s) Charge–discharge experiment. Time until defined cut-off ch Discharge time t (s) voltage reached dch up Upper voltage limit U (V) Set voltage limits (cut-off voltages) low Lower voltage limit U (V) –2 Charge current density I (mA cm ) Set in galvanostatic charge–discharge experiment ch –2 Discharge current density I (mA cm ) dch Cell voltage ΔU(t) (V) Measured in galvanostatic charge–discharge experiment Anolyte pH value pH Measured in electrolyte tanks by pH sensors an Catholyte pH value pH cat ˙ –1 Volumetric flow rate V (ml min ) Set by pump control Pressure drop Δp (mbar) Measured by pumps in 1400-cm cell Ohmic resistance R (Ω cm ) Fit of impedance spectrum Ohm Charge-transfer resistance R (Ω cm ) Fit of impedance spectrum CT Diffusion resistance R (Ω cm ) Fit of impedance spectrum diff Anolyte volume V (L) Electrolyte preparation an Catholyte volume V (L) cat –1 Anolyte concentration c (mol L ) Electrolyte preparation an –1 Catholyte concentration c (mol L ) cat Transferred electrons of anolyte species n Defined by properties of molecule, n  = 2 an an Transferred electrons of catholyte species n Determined by pre-reduction with hydrazine cat PV [15] in combination with a change in pK [22]. During 14 a As Table 2 Derived parameters: F (Faraday constant) = 96 485 the charge of the battery, PV is oxidized, which releases mol 14 two electrons and two protons: Derived parameters Definition î ó 5− 5− V IV V − + (1) [H PV V O ] → H PV O + 2e + 2H theo 6 42 4 42 12 10 14 Theoretical capacity anolyte (mAh) Q = n · c · V · F an an an an theo Theoretical capacity catholyte (mAh)Q = n · c · V · F cat cat cat cat During discharge, the reverse process of Equation 1 takes dch Discharge capacity (mAh) ´ place: PV takes up protons from the catholyte and the pH Q = I dt dch dch increases. This effect is restricted to the catholyte because ´ ch Charge capacity (mAh) Q = I dt ch ch the majority of the cations that cross the cation-exchange dch membrane are Li ions. Fig. 4b compares the previously Coulombic efficiency (%) η = · 100 ch t th th dch shown 10 cycle to the 1000 cycle. The latter was re- U(t)dt Voltage efficiency (%) η = ´ · 100 V t ch U(t)dt corded 1508 h or 63 days after the former. As can be seen dch Energy efficiency (%) I U(t)dt dch in the graph, the general shape of the charge and discharge η = ´ · 100 ch I U(t)dt ch ˙ curves is not changed, exhibiting similarly high efficiencies Normalized volume flow ˙ v = 1000 1000 and η = 85.1%. The theoretical capacity –1 –2 η = 99.2% (ml min cm ) C E theo theo Q   = Q   =  6.4 Ah was not reached during cycling. In Normalized mass flow c˙ = v˙ · c cat an cat th –1 –2 the 10 cycle, a discharge capacity of 4.6 Ah was measured, (mol min cm ) which corresponds to 72% of the theoretical capacity. Two plateaus can easily be distinguished and can be assigned With n  = n  = 2 L and V  = V  = 1.5 L, the theoretical cap- to the first and second one-electron waves of SiW . The an cat an cat theo theo acity for both electrolytes was Q  = Q  = 6.4 Ah. Fig. 4a oxidation state of PV does not influence the potential that an cat shows cycle 10 as an example for the cycling behaviour much because of its multiple electron waves occurring at 2 –2 of the 1400-cm cell. At a current density of 4 mA cm , similar potentials. However, the plateau at higher voltages up the cell reaches its upper voltage cut-off U  = 1.4 V after is shorter than the one at low voltages, which can be ex- ~3000  s; the successive discharge takes another 3000  s plained by an insufficient pre-reduction of PV with hydra- low and stops at U   =  0 V.  During cycle 10, the coulombic zine or oxidation by residual oxygen, respectively. Since the efficiency was η = 99.13 % and the energy efficiency unit was not suited for applying a vacuum, it was impos- was measured to be η = 86.13%. The pH probes re- sible to completely empty the whole container of oxygen, veal that the concentration of hydronium ions in the so oxidation by air probably happened to a certain degree. anolyte (green line) does not change significantly during The long-term cycling behaviour of the 1400-cm cell one cycle, but that the pH in the catholyte drops during is shown in Fig. 5. The changes in pH of the anolyte and charge and increases during discharge. The explanation catholyte are given in Fig. 5a. The behaviour over 1400 for this is the proton-coupled electron transfer (PCET) of cycles is in line with the pH changes shown in Fig. 4a for Downloaded from https://academic.oup.com/ce/article-abstract/3/4/278/5550272 by DeepDyve user on 10 December 2019 282 | Clean Energy, 2019, Vol. 3, No. 4 0.8 1mM PV 0.6 1mM SiW 0.4 0.2 0.0 –0.2 –0.4 WE: Glassy Carbon –0.6 CE: Au RE: MSE in 1M H SO 2 4 –0.8 –1 Scan rate: 100 mV s –1.0 –0.4 –0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 Potential vs. SHE / V –1 Fig. 2 Cyclic voltammograms (CVs) and structure of the used POMs. (a) CVs of 1 mM SiW and 1 mM PV at a scan rate of 100 mV s. (b) Polyhedral 12 14 representation of SiW . (c) Polyhedral representation of PV. 12 14 –2 from 4 to 43 mA cm (cycles 185 to 240), the discharge cap- acity dropped to Q ≈ 3.2 Ah. Comparing Fig. 5a and Fig. 20 dch 25 cm c = 300 mM PV 600 mM SiW 14 12 5b, one can notice that the pH of the anolyte decreases 25 cm c = 80 mM SiW / PV 12 14 from approximately 1.5 to 1.0 when the current density is –2 15 increased from 4 to 43 mA cm . This observation can cur - rently not be explained. Fig. 5c shows the coulombic effi- ciency η (black data) and energy efficiency η (red data). C E For the 1400 cycles measured,η is at approximately 99%. During the cycles with higher current density (cycles 185 to 240), η increases to 100%. The coulombic efficiency η of C C 5 2 the PV –SiW system in the 1400-cm cell is higher than 14 12 VRFB it is for typical VRFB cells ( η ≈ 0.9 [1, 25, 26]). While the VRFB loses charge through cross-mixing of the electrolytes and by the parasitic hydrogen evolution reaction (HER) [27], –2 –2 –2 –2 –2 1.0×10 2.0×10 3.0×10 4.0×10 5.0×10 the POM system experiences no cross-over and the HER –1 –2 Normalised mass flow / (mol min cm ) has not been observed due to more positive potentials in the anolyte compartment. An energy efficiency η ≈ 86 % Fig. 3 Mass-transport resistance R over normalized mass flow for two diff –2 different concentrations of electrolyte measured in the 25-cm cell during the cycles at 4 mA cm and only η ≈ 40 % during –2 the cycles at 43 mA cm was achieved. An EIS study was conducted to determine which resistance, R , R or R , Ohm CT diff contributes most to the overvoltage and therefore the de- cycle 10; the hydronium concentration in the catholyte crease in η at higher rates. (green area) changes much more than that of the anolyte (blue area). The pH of the catholyte is stable; the pH of the anolyte increases slightly from 1.5 to 2 during 1400 cycles 3.2 EIS (pH data were not continuously recorded). Fig. 5b shows the capacity retention (red data) and the current density Impedance spectra were recorded at different volumetric 2 ˙ 2 (blue line) of the 1400-cm cell during 1400 cycles. The dis- flow rates V in the 1400-cm cell; Nyquist plots of these 1 1400 ˙ charge capacity dropped from Q  = 4.70 Ah to Q  = 3.95 spectra are shown in Fig. 6a. The volumetric flow rates V dch dch –3 –1 Ah, which equals an average capacity loss of 0.53 10 Ah/ (ml min ) were normalized to volumetric flow rates per –1 –2 cycle or 0.011% per cycle. The capacity loss of 0.75 Ah may surface area of the cell v (ml min cm ). To obtain R , R Ohm CT be due to slow ingress of atmospheric oxygen, taking up and R , semicircles were fitted to the spectra. The inter - diff protons while being reduced to water. This is supported cepts of these semicircles with the abscissa determine the by similar numbers of charges involved: 0.75 Ah is equal resistances. The first intercept at the highest frequencies is to 28  mmol of transferred electrons while the pH shift R , then R followed by R . As can be seen in Fig. 6a, es- Ohm CT diff from 1.5 to 2 mentioned earlier corresponds to 32 mmol of pecially in the inset Fig. 6b , the first semicircle (at high fre- used-up protons. When the current density was increased quencies) is relatively independent of the normalized flow Normalised R / Ω cm diff –2 Current density / mA cm Downloaded from https://academic.oup.com/ce/article-abstract/3/4/278/5550272 by DeepDyve user on 10 December 2019 Friedl et al. | 283 AB Cycle: 10 pH anolyte 1.4 2 1.4 2.8 2 Cell: 1400 cm Cell: 1400 cm pH catholyte Anolyte: 80 mM SiW Anolyte: 80 mM SiW 12 2.6 Voltage Catholyte: 80 mM PV 1.2 1.2 Catholyte: 80 mM PV 14 –2 –2 Current: 5.6 A / 4mA cm 2.4 Current: 5.6 A / 4mA cm 1.0 1.0 2.2 2.0 0.8 0.8 1.8 0.6 0.6 1.6 0.4 0.4 1.4 1.2 Cycle 10 0.2 0.2 Cycle 1000 1.0 0.0 0.0 0 1000 2000 3000 4000 5000 6000 01234 5 Time / s Capacity / Ah 2 th Fig. 4 Cycling behaviour of the 1400-cm cell. (a) Recorded observables of the 10 charge–discharge cycle over time. Measured pH values of anolyte (green line) and catholyte (blue line) are shown, as well as the cell voltage (red line). (b) Direct comparison of cycle 10 and cycle 1000. Cycle 1000 was th recorded 63 days after the 10 cycle. Cell: 1400 cm Anolyte: 80 mM SiW Catholyte: 80 mM PV 12 14 Cell: 1400 cm pH anolyte pH catholyte Anolyte: 80 mM SiW Catholyte: 80 mM PV 0 200 400 600 800 1000 1200 1400 –2 Current: 5.6 A / 4mA cm Cycles –2 43 mA cm Energy efficiency Coulombic efficiency –2 4 mA cm 0 2 10 0 200 400 600 800 1000 1200 1400 0 200 400 600 800 1000 1200 1400 Cycles Cycle number Fig. 5 Long-term behaviour of the 1400-cm cell. (a) pH of anolyte and catholyte over cycle number. (b) Capacity retention and current density. (c) Coulombic and energy efficiency. rate. The second semicircle (at lower frequencies) becomes polymer membrane is smaller than that of H [30]. The smaller with higher normalized flow rates. Determined charge-transfer resistance R is also independent of the CT values for R (red segment), R (blue segment) and R flow rate and, at R  = (0.382 ± 0.015) Ω cm , smaller than Ohm CT diff CT (green segment) are shown in Fig. 6c. The Ohmic resistance R . The explanation for this small value of R is that it is Ohm CT R represents Ohmic losses in the leads, the current col- inversely proportional to the electron-transfer constant k Ohm 0 lector, the contact resistance between the current collectors and k is large for the POMs [1531 , ]. The resistance invoked and the electrodes, the resistance of the solution and the by mass-transport limitations depends on v because the ionic resistance in the membrane [28]R . is independent flow rate governs at which rate fresh electrolyte is trans- Ohm of flow rate and with R   =  (2.806  ± 0.021) Ω cm higher ported to the electrodes, which can then undergo electron Ohm VRFB 2 –2 than typical values for the VRFB system, R ≈ 0.5 Ω cm transfer. Even at the highest flow rate (v  =  1.34  ml cm Ohm –1 −2 [29]. We assume that this high resistance stems mostly min ), R =  4.4 Ω cm and therefore diffusion presents diff from the ionic resistance of the membrane. Due to a lack a larger resistance than R and R combined. At lower Ohm CT of protons (pH  >  1), Li cations are used to enable charge flow rates, this problem is aggravated. There are two ways balance by their transport through the membrane and the to bring more unreacted electrolyte to the electrode in a diffusion coefficient for Li through a perfluorosulphonated given time: Current density pH of electrolytes –2 / mA cm pH Voltage / V Capacity / Ah Voltage / V Efficiency % Downloaded from https://academic.oup.com/ce/article-abstract/3/4/278/5550272 by DeepDyve user on 10 December 2019 284 | Clean Energy, 2019, Vol. 3, No. 4 A 3 2 –1 Cell: 1400 cm Normalised flow (ml (min cm ) ): R ohm Anolyte: 80 mM SiW 0.52 8.1 CT Catholyte: 80 mM PV 0.69 10 7.1 diff 0.99 1.24 5.9 5.3 1.34 8 4.8 4.4 0.4 0.3 0.41 0.38 0.37 0 0.2 0.39 0.37 0.1 2 0.38 0.0 2.8 2.8 2.8 2.8 2.8 2.8 –1 –0.1 2 34 5 2.6 2.8 3.0 3.2 0.4 0.6 0.8 1.0 1.2 1.4 Re (Z) / Ω cm 2 2 –1 Re (Z) / Ω cm Normalised flow / ml (min cm ) Fig. 6 Galvanostatic impedance data for the 1400-cm cell. (a) Nyquist plots for various normalized pump rates. (b) Detail of (a) showing the first semicircle. (c) Extracted resistances (R , R , R ) over the normalized flow. Measurements were made with no current bias at 36% SOC, which cor - ohm CT diff responds to the centre of the first one-electron redox wave of SiW. (i) Increase the concentration of the redox species to in- no vanadium signal, indicating that no PV has crossed crease the amount of reactant per pumped volume. over into the anolyte reservoir during the 1400 cycles. This This approach can be done without changing the result is in line with our earlier results that showed that power converter and has been shown earlier in this the negatively charged POMs do not cross cation-exchange work using the 25-cm cell (see Fig. 3). membranes [15]. CVs of both electrolytes were recorded (ii) Increase the pump rate and thereby the flow of elec- to check the state of health of the anolyte and to confirm trolyte. In order to increase the pump rate, the cell the result of the V NMR for the catholyte. Supplementary would have to be modified because, at the moment, Fig. 1b (see the online supplementary data) shows the CV the pressure drop within the cell is already at 900 mbar of the anolyte after 1400 cycles, which matches the typ- –2 –1 at 1.34 ml cm min , and the cell could most likely not ical CV of SiW (compare Fig. 2a). Similarly, the CV of the withstand higher pressure drops. Therefore, the cell anolyte shown in Supplementary Fig. 1c (see the online would have  to be converted to a design with a lower supplementary data) matches the CV of PV (compare Fig. impedance to flow, such as a flow-by design instead of 2a). In conclusion, the post-cycling analysis showed that a flow-through configuration [5, 32]. the POMs in the electrolyte were stable for 1400 cycles equalling 88 days of cycling. 3.3 Post-cycling analysis of electrolytes 4 Comparison of the two cells After 1400 cycles, the electrolytes used were extracted from the battery and investigated by V NMR and CV studies, the In order to compare the cycling performance of the two results of which are shown in the supporting information. cells under investigation, both the flow rate and the cap- Supplementary Fig. 1a (see the online supplementary data) acity are normalized, as shown in Fig. 7a. The 25-cm cell 51 25cm gives the V NMR spectra of the anolyte and the catholyte. reached a discharge capacity of Q   =  192 mAh, which dch theo,25cm The signal of the catholyte (red curve) shows the typical is 90% of its theoretical capacity of Q   =  214 mAh. dch 2 th fingerprint signal of PV [15, 22]. Two small additional For the 1400-cm cell, the 10 cycle that was described in peaks at a chemical shift of –505 p.p.m. and –525 p.p.m. can Fig. 4b is given for comparison. At a similar flow rate and 6– –2 be attributed to [VO ] , a polyoxovanadate that forms the same current density (4 mA cm), the coulombic ef- 10 28 1400cm from PV at pH  >2.3 but is in a pH-dependent dynamic ficiency of the larger cell is higher: η 100% versus 14 C 25cm equilibrium with PV [22]. We investigated the formation η  = 96%. It is assumed that the superior atmospheric 14 C 6– 2 of [V O ] and the time and pH dependency of this pro- containment of the 1400-cm cell keeps more oxygen from 10 28 cess earlier via V NMR and found that a pH value as low entering the tanks and therefore oxidation of reduced as 1.7 can be tolerated permanently and even lower values PV electrolyte is prevented. The difference in energy ef- 1400cm are acceptable for short periods of time [15]. All observed ficiency is larger, however, for the big cell: η   =  86% 25cm conversions are reversible and do not cause permanent as compared to η   =  64%. Therefore, the voltage effi- capacity loss. This indicates that PV in the catholyte ciency η of the 1400-cm cell must be higher than that was stable during the cycling and did not decompose to of the 25-cm cell (as η = η · η ). While the coulombic E V C single vanadium species. The anolyte (blue curve) shows efficiency is mainly given by the redox electrochemistry -Im (Z) / Ω cm -Im (Z) / Ω cm Norm. resistances / Ω cm Downloaded from https://academic.oup.com/ce/article-abstract/3/4/278/5550272 by DeepDyve user on 10 December 2019 Friedl et al. | 285 1.4 100 AB Anolyte: 80 mM SiW 25 cm cell: Catholyte: 80 mM PV 1.2 ohm η = 96%, η = 64% C E CT 1.0 10 R diff 0.8 η = 100%, η = 86% C E 0.6 1400 cm cell: ohm 0.4 CT 2 –2 –2 –1 0.2 R Cell: 25 cm ; 4 mA cm ; 1.08 ml cm min diff Anolyte: 80 mM SiW 2 –2 –2 –1 Cell: 1400 cm ; 4 mA cm ; 0.58 ml cm min Catholyte: 80 mM PV 0.0 0.1 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.5 1.0 4.0 4.5 5.0 5.5 6.0 6.5 2 –1 Normalised capacity Normalised flow / ml (min cm ) 2 2 2 Fig. 7 Comparison of a 25-cm cell and a 1400-cm cell with the same electrolyte. (a) A single charge–discharge cycle for the 1400-cm cell and the η η 25-cm cell with energy efficiency and coulombic efficiency given. The capacity is normalized to 1. (b) Resistances R , R and R for the 1400- E C Ohm CT diff 2 2 cm cell (larger circles) and the 25-cm cell (smaller squares) over the flow rate (normalized to membrane area). and the electrode materials, cell construction and its ma-the bigger cell performed better. This is indicated by the terials determine the voltage efficiency. To identify the higher energy efficiency during cycling, and confirmed by loss mechanism(s) that contribute to the difference in the lower resistances found in EIS studies. While R was Ohm voltage efficiency, we compare R , R and R for both similar in both cells, both R and R were considerably Ohm CT diff CT diff cells versus the normalized flow rate. Data points for the smaller in the 1400-cm cell. This suggests that the com- 1400-cm cell in Fig. 7b are the same as shown in Fig. 6b , mercial cell, even though it was designed to be used as a but here the resistances for the smaller cell are shown in VRFB, can yield a higher efficiency than the laboratory cell. addition. For both cells, R and R are independent of The most crucial parameter to improve is R (compare Ohm CT diff the flow rate, whereas R decreases with higher flow. R Fig. 6c). Looking at the equations for diffusion overvoltage diff Ohm (red data) is very similar for both cells, which is under - given by Vetter, an increase in the concentration of active standable, as this resistance should be determined mostly species should alleviate this problem [34]. As shown in Fig. by the membrane resistance and the same membrane is 3, R indeed depends heavily on mass flow. At high con- diff used in both cells. On average, the charge-transfer resist- centrations and for the highest measured rate R = 0.91 Ω diff ance is more than 12 times higher in the smaller cell, with cm . This demonstrates that increasing the concentration 2 2 25cm 2 1400cm 2 R  = 4.75 Ω cm versus R  = 0.382 Ω cm , and there- of electrolyte is a possible approach to reduce R . CT CT diff fore the charge transfer must be considerably faster in the 2 2 1400-cm cell than in the 25-cm cell. The difference in R CT values between the two cells cannot be due to the redox 5 Summary and conclusion electrochemistry, which is identical, and stems most likely from the used electrodes or their pre-treatment. The elec- In this study, we have shown that a recently developed trodes of the 25-cm cell, graphite felts (GFD, SGL Carbon), POM electrochemistry for flow batteries can work effect- were heated to 400°C for 24  h in laboratory atmosphere. ively in a redox flow battery. Data obtained for a 25-cm The electrodes in the 1400-cm cell were preconditioned in laboratory cell were quite satisfying, with a coulombic oxygen plasma. The exact parameters of this process are efficiency of 96%. The energy efficiency of 64% was con- unknown. When heat-treating the GFD 4.6 EA carbon felts siderably lower, probably due to specifications in the cell for 3 h at various temperatures in air, it was found that R design. It was also demonstrated that a large single cell CT reaches a minimum at around 600°C treatment tempera- of 1400  cm can be successfully employed. While the cell ture (Pfanschilling, unpublished work [33]). Whether this is was designed for the VRFB chemistry, it still shows good a long-term effect is yet to be investigated. performance for the investigated SiW–PV electro- 12 14 2 2 Also R (green data) is significantly larger in the 25-cm lytes. During 88  days, the 1400-cm cell was charged and diff cell than it is in the 1400-cm cell, varying with the flow discharged 1400 times. The initial discharge capacity of 1 1400 rate we determined factors from 4 to 7. The reason for this Q  = 4.70 Ah dropped to Q  = 3.95 Ah, which equals an dch dch –3 difference is unknown but we assume that the flow distri- average capacity loss of 0.53 10 Ah/cycle. During the cyc- –2 bution in the larger cell is more effective than the one in ling at 4 mA cm , a coulombic efficiency of η ≈ 99% and an the small cell. energy efficiency of η ≈ 86% were reached—both values The conclusion from the comparison of the 1400-cm larger than in the small laboratory cell. A post-cycling ana- 2 51 cell and the 25-cm cell using the same electrolytes is that lysis was performed on the electrolytes. Both V NMR and Voltage / V Normalised resistances / Ω cm Downloaded from https://academic.oup.com/ce/article-abstract/3/4/278/5550272 by DeepDyve user on 10 December 2019 286 | Clean Energy, 2019, Vol. 3, No. 4 cyclic voltammetry showed that the polyoxoanions in the or 80-µm thickness [36]. The 45-µm-thick membrane electrolyte were not damaged during the battery operation was found to be the optimum due to the combined and that the molecules had not crossed the membrane. effects of Ohmic losses and cross-over. We found that Using pH sensors in the electrolyte tanks, we could show SiW does not cross perfluorosulfonic acid (PFSA) that the pH value of the catholyte changes during cycling, membranes as thin as 40 µm due to size exclusion as expected for the PCET experienced by PV , but that the and electrostatic repulsion [15]. The minimum mem- average pH of the catholyte does not shift over 1400 cycles. brane thickness that still prevents cross-over should The average pH of the anolyte increases slightly during the be determined. experiment, from pH 1.5 to 2 during 1400 cycles. ○ Reduce the specific Ohmic resistance of the membrane. While the cell showed a high energy efficiency at low We assume that the high value for R stems from the Ohm current densities, this value dropped to η ≈ 40 % at 43 fact that Li cations are used for charge balance instead –2 mA cm . Using impedance spectroscopy, we found that of protons. The cell chemistry could either be trans- both the Ohmic resistance R and the charge-transfer ferred into a more acidic solvent, which allows the use Ohm resistance R were independent of the electrolyte of protons, or the membrane could be optimized for the CT flow rate, but that high flow rates decreased the mass- use of Li cations. A lower pH than currently employed transport resistance R . At the highest volumetric flow might lead to stability issues for PV [22]. diff 14 –2 –1 rate (1.34 ml cm min ), the measured resistances were • The geometry of the power converter and its flow design [ 5, as follows: R   =  2.78, R   =  0.39 and R   =  4.4 Ω cm Ohm CT diff 37]: Neglecting an effect of R , there are two limiting Ohm . Comparing the first electron transfer of SiW (k   =  4.2 12 0 cases for the type of rate control in a electrochemical –2 –1 2+ + –7 10 cm s ) and the VO /VO redox reaction (k  =  3 10 2 0 converter: charge-transfer control and mass-transport –1 cm s ), the electron transfer of the POM is more than five control [37]. In the former case, the rate of electron orders of magnitude faster than that of the latter [35 15], . transfer limits the current that can be drawn from This confirms the assumption from the introduction that, the cell; a larger overpotential or a larger surface area for redox couples with fast electron-transfer kinetics, the for the electrode can increase the rate. Under mass- total resistance R depends more on the cell param- total transport control, the supply of active species and its eters R and R than on the electrochemical param- diff Ohm removal after reaction determine the current; a limiting eter R . For higher concentrations, this will also remain CT current I ∝ v · c = c can be defined. Assuming cat/an true, with a larger influence from R . While the above Ohm that R is a measure for 1/I , the connection between I diff L L resistances were recorded at an electrolyte concentra- and c ˙ is empirically shown in Fig. 3. The proportionality tion of c   =  c   =  80  mM, we have shown in Fig. 3 that an cat depends on the geometry and typically turbulent flow –1 –2 R decreases with higher mass flows (mol min cm ) and diff allows higher I than laminar flow. As the POMs exhibit therefore concentrations. Simultaneously, R will de- CT fast electron transfer, and judging from Fig. 6c I needs , crease with higher concentrations, as it is inversely pro- to be increased to enhance the performance of the cell, portional to the concentration of redox species [ ]. The 7 which can be done in two ways: Ohmic drop R , which is governed by the membrane, Ohm ○ Increase the concentration c . This approach was cat/an will remain unchanged by a change in the concentration tested for the small cell with the result shown in Fig. of active material. 3. Clearly, a higher concentration of active species re- This indicates that, in order to increase the perform- duces R and increases I . diff L ance of the presented asymmetric POM RFB, the cell, spe- ○ Increase the rate at which fresh active material is sup- cifically parameters related to R and R need to be the Ohm diff plied to the electrodes. This can be done by increasing main focus for improvements. Components to work on are: the pump rate or by enhancing the spatial distribu- • Membrane/separator: The Ohmic drop R , which is tion of the mass flow in the cell to maximize electro- Ohm dominated by the membrane resistance, is likely to be lyte utilization. the highest resistance in a cell, with a higher concen- tration of active material, as R does not scale with Ohm the concentration of the redox electrolyte. To reduce Supplementary data R incurred from the membrane, two measures can Ohm Supplementary data is available at Clean Energy online. be taken: ○ Use a membrane thinner than the currently used FUMASEP—F-1075-PK (thickness 75 µm), as the ionic resistance of the membrane increases with thickness Acknowledgements of the membrane. For the VRFB, it was found that a This work was supported by a grant from Siemens AG. Additional sulphonated fluorinated poly(arylene ether) mem- support comes from the North East Centre of Energy Materials brane with 45-µm thickness enabled a higher power (NECEM) funded by EPSRC (EP/R021503/1) and Newcastle density than the same type of membrane with 28 University. Downloaded from https://academic.oup.com/ce/article-abstract/3/4/278/5550272 by DeepDyve user on 10 December 2019 Friedl et al. | 287 derivatives on glassy carbon electrodes. AnalChem 1999; References 71:4081–7. [1] Ponce  de  Leon  C, Friasferrer  A, Gonzalezgarcia  J, et  al. [20] Fink H, Friedl J, Stimming U. Composition of the electrode de- Redox flow cells for energy conversion. J Power Sources 2006; termines which half-cell’s rate constant is higher in a van- 160:716–32. adium flow battery. J Phys Chem C 2016; 120:15893–901. [2] Weber AZ, Mench MM, Meyers JP, et al. Redox flow batteries: a [21] Friedl  J, Al-Oweini  R, Herpich  M, et  al. Electrochemical review. J Appl Electrochem 2011; 41:1137–64. studies of tri-manganese substituted Keggin polyoxoanions. [3] Holland-Cunz  MV, Cording  F, Friedl  J, et  al. Redox flow bat- Electrochim Acta 2014; 141:357–66. teries—concepts and chemistries for cost-effective energy [22] Selling  A, Andersson  I, Pettersson  L, et  al. Multicomponent storage. Front Energy 2018; 12:198–224. polyanions. 47: The aqueous vanadophosphate system. Inorg [4] Zhang  M, Moore  M, Watson  JS, et  al. Capital cost sensitivity Chem 1994; 33:3141–50. analysis of an all-vanadium redox-flow battery. J Electrochem [23] Keita  B, Nadjo  L. New aspects of the electrochemistry of Soc 2012; 159:A1183–8. heteropolyacids.J Electr oanal Chem Interfacial Electrochem 1987; [5] Arenas LF, Ponce de León C, Walsh FC. Engineering aspects of 227:77–98. the design, construction and performance of modular redox [24] Keggin  JF. The structure and formula of 12-phosphotungstic flow batteries for energy storage. J Energy Storage 2017; 11:119–53. acid. Proc R Soc A Math Phys Eng Sci 1934; 144:75–100. [6] Noack  J, Roznyatovskaya  N, Herr  T, et  al. The chemistry of [25] Pezeshki  AM, Sacci  RL, Veith  GM, et  al. The cell-in-series redox-flow batteries. Angew Chemie Int Ed 2015; 54:9776–809. method: a technique for accelerated electrode degradation in [7] Friedl J, Stimming U. Determining electron transfer kinetics at redox flow batteries. J Electrochem Soc 2016; 163:A5202–A5210. porous electrodes. Electrochim Acta 2017; 227:235–45. [26] Skyllas-Kazacos  M, Grossmith  F. Efficient vanadium redox [8] Zhang H. Development and application of high performance flow cell. J Electrochem Soc 1987; 134:2950–4. VRB technology. In: IFBF 2017—International Flow Battery Forum, [27] Schweiss R, Pritzl A, Meiser C. Parasitic hydrogen evolution at Manchester, UK, 2017. different carbon fiber electrodes in vanadium redox flow bat- [9] Rabiul Islam FM, Al Mamun K, Amanullah MTO Smart Ener . gy teries. J Electrochem Soc 2016; 163:A2089–94. Grid Design for Island Countries. Basel, Switzerland: Springer, [28] Sun C-N, Delnick FM, Aaron DS, et al. Probing electrode losses in all-vanadium redox flow batteries with impedance spec- [10] Lim  HS, Lackner  AM, Knechtli  RC. Zinc-bromine secondary troscopy. ECS Electrochem Lett 2013; 2:A43–5. battery. J Electrochem Soc 1977; 124:1154–7. [29] Liu  QH, Grim  GM, Papandrew  AB, et  al. High performance [11] Johnson DA, Reid MA. Chemical and electrochemical behavior vanadium redox flow batteries with optimized electrode con- of the Cr(lll)/Cr(ll) half-cell in the iron-chromium redox en- figuration and membrane selection. J Electrochem Soc 2012; ergy system. J Electrochem Soc 1985; 132:1058–62. 159:A1246–52. [12] Nice AW. NASA Redox System Development Project StatusCleveland, [30] Samec Z, Trojanek A, Langmaier J. Diffusion coefficients of al- OH: Cleveland, OH. 1981. kali metal cations in Nafion® from ion-exchange measure- [13] Scamman DP, Reade GW, Roberts EPL. Numerical modelling of ments. J Electrochem Soc 1997; 144:4236. a bromide-polysulphide redox flow battery. Part 1: Modelling [31] Friedl J, Bauer CM, Rinaldi A, et al. Electron transfer kinetics of 2+ + approach and validation for a pilot-scale system. J Power the VO / VO —reaction on multi-walled carbon nanotubes. Sources 2009; 189:1220–30. Carbon N Y 2013; 63:228–39. [14] Morrisey P. Regenesys: a new energy storage technology. Int J [32] Zhou  XL, Zhao  TS, An  L, et  al. Critical transport issues for Ambient Energy 2000; 21:213–20. improving the performance of aqueous redox flow batteries. J [15] Friedl  J, Holland-Cunz  MV, Cording  F, et  al. Asymmetric Power Sources 2017; 339:1–12. polyoxometalate electrolytes for advanced redox flow bat- [33] Pfanschilling FL. Unpublished results. teries. Energy Environ Sci 2018; 11:3010–8. [34] Vetter KJ. Electrochemical Kinetics—Theoretical and Experimental [16] Laborda  E, Henstridge  MC, Batchelor-McAuley  C, et  al. Aspects, English edn. New York/London: Academic Press Inc., Asymmetric Marcus-Hush theory for voltammetry. Chem Soc Rev 2013; 42:4894–905. [35] Gattrell M, Park J, MacDougall B, et al. Study of the mechanism [17] Chen H-Y, Friedl J, Pan C-J, et al. In situ X-ray absorption near of the vanadium 4+/5+ redox reaction in acidic solutions. J edge structure studies and charge transfer kinetics of Na Electrochem Soc 2004; 151:A123. [V O ] electrodes. Phys Chem Chem Phys 2017; 19:3358–65. [36] Chen  D, Hickner  MA, Agar  E, et  al. Optimizing membrane 10 28 [18] Aparicio  PA, Poblet  JM, López  X. Tungsten redox waves in thickness for vanadium redox flow batteries. J Memb Sci 2013; [XMW O ] n- (X = P, Si, Al and M = W, Mo, V, Nb, Ti) Keggin 437:108–13. 11 40 compounds—effect of localised/delocalised charges. Eur J [37] Walsh  FC, Pletcher  D. Electrochemical engineering and cell Inorg Chem 2013; 2013:1910–6. design. In: Pletcher D, Tian Z-Q, Williams D (eds). Developments in Electrochemistry: Science Inspired by Martin Fleischmann. [19] Yang H-H, McCreery RL. Effects of surface monolayers on the electron-transfer kinetics and adsorption of methyl viologen Hoboken, NJ: John Wiley & Sons, 2014, 95–109. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Clean Energy Oxford University Press

Loading next page...
 
/lp/oxford-university-press/a-polyoxometalate-redox-flow-battery-functionality-and-upscale-vd8rQEfVoK
Publisher
Oxford University Press
Copyright
© The Author(s) 2019. Published by Oxford University Press on behalf of National Institute of Clean-and-Low-Carbon Energy
ISSN
2515-4230
eISSN
2515-396X
DOI
10.1093/ce/zkz019
Publisher site
See Article on Publisher Site

Abstract

While redox flow batteries carry a large potential for electricity storage, specifically for regenerative energies, the current technology-prone system—the all-vanadium redox flow battery—exhibits two major disadvantages: low energy and low power densities. Polyoxometalates have the potential to mitigate both effects. In this publication, 4– the operation of a polyoxometalate redox flow battery was demonstrated for the polyoxoanions [SiW O ] 12 40 9– (SiW ) in the anolyte and [PV O ] (PV ) in the catholyte. Emphasis was laid on comparing to which extent 12 14 42 14 an upscale from 25 to 1400 cm membrane area may impede efficiency and operational parameters. Results demonstrated that the operation of the large cell for close to 3 months did not diminish operation and the stability of polyoxometalates was unaltered. Keywords: energy storage; redox flow battery; polyoxometalates; upscale Current challenges for RFB chemistry are their power Introduction density, their energy density and their costs [1–4]. Redox flow batteries (RFBs) are one of the few options to According to Arenas et  al., the development of new RFB store energy from intermittent renewable-energy sources technology can be described as follows []. 5 As a first step, such as wind and solar electrochemically. The concept the fundamental electrochemistry is explored. Physical of the RFB has several advantages [1–3], such as the in- and chemical properties as well as important parameters, dependent scalability of power and energy content. The such as the equilibrium potential U and the electron- former is determined by the size of the power converter, transfer constant k of the involved species, are investi- whereas the latter is given by the energy density of the gated (Stage 1a). Following this, the cycling behaviour electrolyte and the size of the tanks. Furthermore, as op- and stability of the novel RFB electrolytes are assessed in posed to other battery types, the electrodes themselves H-cells or laboratory-scale flow cells (Stage 1b). The next are not redox-active and do not undergo conversion, inter - step then involves optimization of the electrochemical calation or alloying reactions, which often lead to degrad- power converter according to the chemical and phys- ation. Instead, the energy is stored in redox-active species ical properties of the RFB electrochemistry. Potential and that are dissolved in the electrolyte. Received: 15 January, 2019; Accepted: 19 June, 2019 © The Author(s) 2019. Published by Oxford University Press on behalf of National Institute of Clean-and-Low-Carbon Energy This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http:// creativecommons.org/licenses/by-nc/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, 278 provided the original work is properly cited. For commercial re-use, please contact journals.permissions@oup.com Downloaded from https://academic.oup.com/ce/article-abstract/3/4/278/5550272 by DeepDyve user on 10 December 2019 Friedl et al. | 279 2+ 3+ 2+ + current distribution, hydrodynamics, mass transport and transfer for the V/V and VO /VO redox reactions is −6 −1 cell geometry are optimized (Stage 2a), followed by the slow (k ≈ 10 cm s ) [7, 20], whereas electron transfer −2 −1 development of a pilot stack and a prototype RFB system for the POMs is facile (k ≈ 10 cm s ) [15, 21]. The con- (Stage 2b). The final stage according to Arenas et  al. is a sequence is that the charge-transfer resistance R for CT commercial implementation, which involves fabrication, the POMs is considerably lower than for the VRFB. This testing, maintenance, marketing and many other tasks leads to consequences for the design of the power con- outside the realm of electrochemical engineering (Stage 3). verter, as the power-converter design aims to reduce the The properties of the used redox shuttles and employed total resistance R, which is the sum of R , R and the tot CT diff solvents determine how the electron transfer and mass Ohmic losses R . The resistance of the cell determines Ohm transport proceed within the power converter. Therefore, the currents that can be dr I awn from a battery for a given ideally, the power converter should be optimized for the voltage efficiency η : specific redox shuttle. However, while many novel redox ΔU − R I tot η = electrochemistries were proposed for RFBs in recent years, ΔU + R I tot the literature focuses mostly on Stages 1a and 1b [ ]. A  6 re- with cell voltagΔ e U. In this study, we describe our findings cent literature review by our group showed that, of 24 of a novel flow battery electrochemistry for a laboratory- published redox electrochemistries for RFBs, only 4 were type cell of 25 cm and the scale-up to a commercial cell demonstrated with a power rating of more than 100 W of 1400 cm . Details on the basic electrochemistry of this [3]. These are the all-vanadium RFB (VRFB) [, 78], the zinc- novel RFB system can be found in a recent publication [15]. bromine cell [9 , 10], the iron-chromium cell [1112 , ] and the In this paper, we are describing the operational parameters bromine-polysulphide RFB [1 , 13, 14]. Of the remaining 20, 7 such as charge–discharge cycles, coulombic and energy effi- were tested in H-cell configurations only, all the others ex- ciencies, and stability of the electrolytes. Typical values such cept for 2 in cells with less than 1-W power output. So, there as R , R and R are determined using electrochemical im- Ohm CT diff is currently a knowledge gap between the laboratory Stage pedance spectroscopy (EIS) and compared for the two cells. 1b and the electrochemical technology Stage 2a. This is one of the reasons for showing how a scale-up of a novel electro- 2 2 chemistry from a 25-cm cell to a cell of 1400 cm may affect 1 Materials important performance parameters of the RFB. The ap- 1.1 Cells, sensors and parameters proach in our laboratory is to use polyoxometalates as redox shuttles. The redox electrochemistry described here is a Three electrode measurements were performed in polyoxometalate (POM) system, utilizing the polyoxoanions custom-built glass cells with a polished glassy carbon 4– 9– 2 [SiW O ] (SiW ) in the anolyte and [PV O ] (PV ) in the working electrode (surface area A  =  0.02  cm ), a gold wire 12 40 12 14 42 14 catholyte as nano-sized charge carriers [15]. (diameter d  =  0.5  mm) counter electrode and a mercury/ Due to their structural and chemical properties, these mercurous sulphate reference electrode in 1 M H SO (MSE, 2 4 metal-oxygen clusters have a number of advantages for 0.668 V vs. Standard Hydrogen Electrode [SHE]). Prior to the energy-storage applications [15]. First, due to their large measurements, the electrolyte was purged with nitrogen size, the interaction of the redox centres of the POM with and the cell was kept under nitrogen pressure during the solvent molecules is small. Therefore, the outer-sphere re- experiment. A  Bio-Logic SP-300 potentiostat was used for organization energy of the electron transfer is low, enabling control and data acquisition. fast kinetics and thus high power densities [1617 , ]. Also, The small flow cell used was a commercial cell (C-Tech the inner-sphere reorganization energy is low due to the 5x5, surface area A  =  25  cm ). Graphite felts (GFD, SGL added electrons often being delocalized [18], causing only Carbon) were used as electrodes and pre-treated at a minimal change in coordination upon reduction or oxi- 400°C for 24 h in a laboratory atmosphere. In the cell, the dation [19]. This behaviour adds to the fast kinetics and 4.6-mm-thick electrodes were compressed to 3.5  mm. As thus high power densities. Their nature as large anions the membrane, a cation-exchange membrane (FUMASEP— also prevents POMs from penetrating commercial cation- F-1075-PK) was used. During the experiment, the cell and exchange membranes [15]. Size exclusion and electrostatic the pump with tubing were kept in a polycarbonate box repulsion prohibit cross-over and mixing of the active spe- purged with nitrogen. The peristaltic pump could supply –1 cies. Furthermore, some POMs exhibit high solubility, e.g. flow rates of 12–150 ml min . Charge and discharge cycles –1 a concentration of 0.875 mol L can be obtained for SiW were measured using a Bio-Logic BCS-810 battery tester. [15]. With multiple redox-active centres per molecule, this EIS measurements were performed using the Bio-Logic enables a high energy density in the battery. Moreover, SP-300 (maximum supplied f = 7 MHz, measurement up to both SiW and PV are stable during the operation of the 200 kHz), as this could apply higher frequencies than the 12 14 battery. Losses in capacity only seem to stem from a para- battery tester (f up to 10 kHz). sitic reaction with residual oxygen, which could be avoided The large flow cell used was a commercial cell (J. with an improved airtight setup [15]. Schmalz GmbH, surface area A  =  1400  cm ), originally de- Comparing the asymmetric POM system with the signed for the VRFB chemistry. The same membrane as in VRFB, one main difference is apparent: the electron the small cell was used. The flow rate was adjusted via the Downloaded from https://academic.oup.com/ce/article-abstract/3/4/278/5550272 by DeepDyve user on 10 December 2019 280 | Clean Energy, 2019, Vol. 3, No. 4 pressure drop in the cell. Measured electrolyte flow and anolyte. PV was prepared following the synthesis for pressure drop were always proportional for both anolyte Na [H PV O ]∙28 H O described in ref. [22]. To prepare the 5 4 14 42 2 –1 and catholyte. Flow rates from 450 to 1872  ml min were catholyte, Na[H PV O ]∙28 H O was dissolved in DI water 5 4 14 42 2 determined. Both the charge–discharge cycles and EIS were with 1 M LiCl and pre-reduced by addition of hydrazine measured using a Bio-Logic BCS-815. In operando measure- (Sigma-Aldrich). This reduction was necessary, as other - ments of the pH values in the tanks were performed using wise both anolyte and catholyte would have been fully oxi- Unitrode pH sensors from Metrohm. For data acquisition of dized at the start of the battery operation, making cycling the pH values, a National Instruments cDAQ-9175 was used impossible. Cyclic voltammograms (CVs) of 1 mM PV (red in combination with a National Instruments Module 9205 line) and 1 mM SiW (blue line) are shown in Fig. 2a. PV 12 14 and a current loop converter CLC-01 from company pro- exhibits a multi-electron redox reaction in the range from viding electronic measurementF . ig. 1 shows a photograph 0.2 to 0.7 V vs. SHE. Experiments in a symmetric flow bat- of a 1400-cm cell with its periphery and containment. The tery (PV as both anolyte and catholyte) have shown that cell itself is fixed in the top part of an acrylic glass container, PV transfers at least seven electrons [15 SiW ]. shows two 14 12 θ,1 θ,2 the pumps are installed underneath and the electrolyte redox reactions at U  = 0.0 V vs. SHE and U   =  –0.21 SiW SiW 12 12 θ,3 tanks are to the left and right of the pumps. The electrolyte V vs. SHE. A  third redox wave centred on U   =  –0.37 V SiW tanks are continuously purged with nitrogen to avoid oxi- vs. SHE is a two-electron redox reaction as opposed to the dation of the reduced redox species by atmospheric oxygen. previous two, which are one-electron transfers. However, Tables 1 and 2 give an overview of the parameters used the third reaction is not used in the flow battery because in this work. These quantities are used to discuss and com- the potential at which it takes place leads to an irrevers- pare the performance of the two cells. ible dimerization of the POM onto the carbon electrodes [23]. The polyhedral structures shown in Fig. 2b and Fig. 2c reveal that both SiW and PV are of the Keggin structure, 12 14 1.2 Electrolytes and their electrochemistry with two additional V–O caps for the polyoxovanadate [24]. The electrochemistry and preparation of SiW and PV 12 14 were described in detail in ref. [15]. In short, SiW was 2 Results in a 25-cm cell bought from Sigma-Aldrich as tungstosilic acid (H SiW O ) 4 12 40 and dissolved in de-ionised (DI) water with 1 M LiCl. Diluted In order to obtain data from a polyoxometalate redox flow LiOH was used for pH adjustment to 1.8 to prepare the battery, a commercial cell from C-Tech with a membrane area of 25 cm was used. Heat-treated, compressed graphite felts (Sigracell GFD 4.6 EA) were used as electrodes as described earlier and a FUMASEP—F-1075-PK membrane was employed. The electrolytes comprised 80 mM SiW for the anolyte and 80 mM PV for the catholyte in 1 M LiCl each. In charge–dis- charge cycles, the 25-cm cell exhibited a coulombic efficiency 2 2 25cm 25cm η of 96% and an energy efficiency η of 64%. The the- C E theo,25cm oretic capacity was Q  = 214 mAh, of which 90% was dch reached. A study of the influence of the electrolyte concentra- tion on the diffusion resistance R was conducted. diff In Fig. 3, R is compared for the 25-cm cell with 80 mM diff SiW and 80 mM PV as electrolytes (blue data) and a setup 12 14 in which the anolyte was 600  mM SiW and the catholyte 300  mM PV (red data). To balance the charge, the PV was 14 14 pre-reduced by four electrons with hydrazine (instead of a pre-reduction by two electrons as for equal concentra- tions of POMs). The volume flow v has been recalculated to ˙ ˙ normalized mass flow c (see Table 2) by multiplying v with the used concentrations of catholyte. It can clearly be seen that the mass flow, that is the amount of unreacted active material that is brought into the cell per minute, determines .R diff 3 Results in a 1400-cm cell 3.1 Charge and discharge Fig. 1 Photograph of 1400-cm cell with periphery (tanks, pumps, An electrolyte solution with 80 mM SiW was prepared as tubing) and containment anolyte; a solution of 80 mM PV was used as catholyte. 14 Downloaded from https://academic.oup.com/ce/article-abstract/3/4/278/5550272 by DeepDyve user on 10 December 2019 Friedl et al. | 281 Table 1 Experimental parameters obtained or set in the measurements Experimental parameters Symbol Obtained from Surface area of cell A (cm ) Geometric area cell/membrane Charge time t (s) Charge–discharge experiment. Time until defined cut-off ch Discharge time t (s) voltage reached dch up Upper voltage limit U (V) Set voltage limits (cut-off voltages) low Lower voltage limit U (V) –2 Charge current density I (mA cm ) Set in galvanostatic charge–discharge experiment ch –2 Discharge current density I (mA cm ) dch Cell voltage ΔU(t) (V) Measured in galvanostatic charge–discharge experiment Anolyte pH value pH Measured in electrolyte tanks by pH sensors an Catholyte pH value pH cat ˙ –1 Volumetric flow rate V (ml min ) Set by pump control Pressure drop Δp (mbar) Measured by pumps in 1400-cm cell Ohmic resistance R (Ω cm ) Fit of impedance spectrum Ohm Charge-transfer resistance R (Ω cm ) Fit of impedance spectrum CT Diffusion resistance R (Ω cm ) Fit of impedance spectrum diff Anolyte volume V (L) Electrolyte preparation an Catholyte volume V (L) cat –1 Anolyte concentration c (mol L ) Electrolyte preparation an –1 Catholyte concentration c (mol L ) cat Transferred electrons of anolyte species n Defined by properties of molecule, n  = 2 an an Transferred electrons of catholyte species n Determined by pre-reduction with hydrazine cat PV [15] in combination with a change in pK [22]. During 14 a As Table 2 Derived parameters: F (Faraday constant) = 96 485 the charge of the battery, PV is oxidized, which releases mol 14 two electrons and two protons: Derived parameters Definition î ó 5− 5− V IV V − + (1) [H PV V O ] → H PV O + 2e + 2H theo 6 42 4 42 12 10 14 Theoretical capacity anolyte (mAh) Q = n · c · V · F an an an an theo Theoretical capacity catholyte (mAh)Q = n · c · V · F cat cat cat cat During discharge, the reverse process of Equation 1 takes dch Discharge capacity (mAh) ´ place: PV takes up protons from the catholyte and the pH Q = I dt dch dch increases. This effect is restricted to the catholyte because ´ ch Charge capacity (mAh) Q = I dt ch ch the majority of the cations that cross the cation-exchange dch membrane are Li ions. Fig. 4b compares the previously Coulombic efficiency (%) η = · 100 ch t th th dch shown 10 cycle to the 1000 cycle. The latter was re- U(t)dt Voltage efficiency (%) η = ´ · 100 V t ch U(t)dt corded 1508 h or 63 days after the former. As can be seen dch Energy efficiency (%) I U(t)dt dch in the graph, the general shape of the charge and discharge η = ´ · 100 ch I U(t)dt ch ˙ curves is not changed, exhibiting similarly high efficiencies Normalized volume flow ˙ v = 1000 1000 and η = 85.1%. The theoretical capacity –1 –2 η = 99.2% (ml min cm ) C E theo theo Q   = Q   =  6.4 Ah was not reached during cycling. In Normalized mass flow c˙ = v˙ · c cat an cat th –1 –2 the 10 cycle, a discharge capacity of 4.6 Ah was measured, (mol min cm ) which corresponds to 72% of the theoretical capacity. Two plateaus can easily be distinguished and can be assigned With n  = n  = 2 L and V  = V  = 1.5 L, the theoretical cap- to the first and second one-electron waves of SiW . The an cat an cat theo theo acity for both electrolytes was Q  = Q  = 6.4 Ah. Fig. 4a oxidation state of PV does not influence the potential that an cat shows cycle 10 as an example for the cycling behaviour much because of its multiple electron waves occurring at 2 –2 of the 1400-cm cell. At a current density of 4 mA cm , similar potentials. However, the plateau at higher voltages up the cell reaches its upper voltage cut-off U  = 1.4 V after is shorter than the one at low voltages, which can be ex- ~3000  s; the successive discharge takes another 3000  s plained by an insufficient pre-reduction of PV with hydra- low and stops at U   =  0 V.  During cycle 10, the coulombic zine or oxidation by residual oxygen, respectively. Since the efficiency was η = 99.13 % and the energy efficiency unit was not suited for applying a vacuum, it was impos- was measured to be η = 86.13%. The pH probes re- sible to completely empty the whole container of oxygen, veal that the concentration of hydronium ions in the so oxidation by air probably happened to a certain degree. anolyte (green line) does not change significantly during The long-term cycling behaviour of the 1400-cm cell one cycle, but that the pH in the catholyte drops during is shown in Fig. 5. The changes in pH of the anolyte and charge and increases during discharge. The explanation catholyte are given in Fig. 5a. The behaviour over 1400 for this is the proton-coupled electron transfer (PCET) of cycles is in line with the pH changes shown in Fig. 4a for Downloaded from https://academic.oup.com/ce/article-abstract/3/4/278/5550272 by DeepDyve user on 10 December 2019 282 | Clean Energy, 2019, Vol. 3, No. 4 0.8 1mM PV 0.6 1mM SiW 0.4 0.2 0.0 –0.2 –0.4 WE: Glassy Carbon –0.6 CE: Au RE: MSE in 1M H SO 2 4 –0.8 –1 Scan rate: 100 mV s –1.0 –0.4 –0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 Potential vs. SHE / V –1 Fig. 2 Cyclic voltammograms (CVs) and structure of the used POMs. (a) CVs of 1 mM SiW and 1 mM PV at a scan rate of 100 mV s. (b) Polyhedral 12 14 representation of SiW . (c) Polyhedral representation of PV. 12 14 –2 from 4 to 43 mA cm (cycles 185 to 240), the discharge cap- acity dropped to Q ≈ 3.2 Ah. Comparing Fig. 5a and Fig. 20 dch 25 cm c = 300 mM PV 600 mM SiW 14 12 5b, one can notice that the pH of the anolyte decreases 25 cm c = 80 mM SiW / PV 12 14 from approximately 1.5 to 1.0 when the current density is –2 15 increased from 4 to 43 mA cm . This observation can cur - rently not be explained. Fig. 5c shows the coulombic effi- ciency η (black data) and energy efficiency η (red data). C E For the 1400 cycles measured,η is at approximately 99%. During the cycles with higher current density (cycles 185 to 240), η increases to 100%. The coulombic efficiency η of C C 5 2 the PV –SiW system in the 1400-cm cell is higher than 14 12 VRFB it is for typical VRFB cells ( η ≈ 0.9 [1, 25, 26]). While the VRFB loses charge through cross-mixing of the electrolytes and by the parasitic hydrogen evolution reaction (HER) [27], –2 –2 –2 –2 –2 1.0×10 2.0×10 3.0×10 4.0×10 5.0×10 the POM system experiences no cross-over and the HER –1 –2 Normalised mass flow / (mol min cm ) has not been observed due to more positive potentials in the anolyte compartment. An energy efficiency η ≈ 86 % Fig. 3 Mass-transport resistance R over normalized mass flow for two diff –2 different concentrations of electrolyte measured in the 25-cm cell during the cycles at 4 mA cm and only η ≈ 40 % during –2 the cycles at 43 mA cm was achieved. An EIS study was conducted to determine which resistance, R , R or R , Ohm CT diff contributes most to the overvoltage and therefore the de- cycle 10; the hydronium concentration in the catholyte crease in η at higher rates. (green area) changes much more than that of the anolyte (blue area). The pH of the catholyte is stable; the pH of the anolyte increases slightly from 1.5 to 2 during 1400 cycles 3.2 EIS (pH data were not continuously recorded). Fig. 5b shows the capacity retention (red data) and the current density Impedance spectra were recorded at different volumetric 2 ˙ 2 (blue line) of the 1400-cm cell during 1400 cycles. The dis- flow rates V in the 1400-cm cell; Nyquist plots of these 1 1400 ˙ charge capacity dropped from Q  = 4.70 Ah to Q  = 3.95 spectra are shown in Fig. 6a. The volumetric flow rates V dch dch –3 –1 Ah, which equals an average capacity loss of 0.53 10 Ah/ (ml min ) were normalized to volumetric flow rates per –1 –2 cycle or 0.011% per cycle. The capacity loss of 0.75 Ah may surface area of the cell v (ml min cm ). To obtain R , R Ohm CT be due to slow ingress of atmospheric oxygen, taking up and R , semicircles were fitted to the spectra. The inter - diff protons while being reduced to water. This is supported cepts of these semicircles with the abscissa determine the by similar numbers of charges involved: 0.75 Ah is equal resistances. The first intercept at the highest frequencies is to 28  mmol of transferred electrons while the pH shift R , then R followed by R . As can be seen in Fig. 6a, es- Ohm CT diff from 1.5 to 2 mentioned earlier corresponds to 32 mmol of pecially in the inset Fig. 6b , the first semicircle (at high fre- used-up protons. When the current density was increased quencies) is relatively independent of the normalized flow Normalised R / Ω cm diff –2 Current density / mA cm Downloaded from https://academic.oup.com/ce/article-abstract/3/4/278/5550272 by DeepDyve user on 10 December 2019 Friedl et al. | 283 AB Cycle: 10 pH anolyte 1.4 2 1.4 2.8 2 Cell: 1400 cm Cell: 1400 cm pH catholyte Anolyte: 80 mM SiW Anolyte: 80 mM SiW 12 2.6 Voltage Catholyte: 80 mM PV 1.2 1.2 Catholyte: 80 mM PV 14 –2 –2 Current: 5.6 A / 4mA cm 2.4 Current: 5.6 A / 4mA cm 1.0 1.0 2.2 2.0 0.8 0.8 1.8 0.6 0.6 1.6 0.4 0.4 1.4 1.2 Cycle 10 0.2 0.2 Cycle 1000 1.0 0.0 0.0 0 1000 2000 3000 4000 5000 6000 01234 5 Time / s Capacity / Ah 2 th Fig. 4 Cycling behaviour of the 1400-cm cell. (a) Recorded observables of the 10 charge–discharge cycle over time. Measured pH values of anolyte (green line) and catholyte (blue line) are shown, as well as the cell voltage (red line). (b) Direct comparison of cycle 10 and cycle 1000. Cycle 1000 was th recorded 63 days after the 10 cycle. Cell: 1400 cm Anolyte: 80 mM SiW Catholyte: 80 mM PV 12 14 Cell: 1400 cm pH anolyte pH catholyte Anolyte: 80 mM SiW Catholyte: 80 mM PV 0 200 400 600 800 1000 1200 1400 –2 Current: 5.6 A / 4mA cm Cycles –2 43 mA cm Energy efficiency Coulombic efficiency –2 4 mA cm 0 2 10 0 200 400 600 800 1000 1200 1400 0 200 400 600 800 1000 1200 1400 Cycles Cycle number Fig. 5 Long-term behaviour of the 1400-cm cell. (a) pH of anolyte and catholyte over cycle number. (b) Capacity retention and current density. (c) Coulombic and energy efficiency. rate. The second semicircle (at lower frequencies) becomes polymer membrane is smaller than that of H [30]. The smaller with higher normalized flow rates. Determined charge-transfer resistance R is also independent of the CT values for R (red segment), R (blue segment) and R flow rate and, at R  = (0.382 ± 0.015) Ω cm , smaller than Ohm CT diff CT (green segment) are shown in Fig. 6c. The Ohmic resistance R . The explanation for this small value of R is that it is Ohm CT R represents Ohmic losses in the leads, the current col- inversely proportional to the electron-transfer constant k Ohm 0 lector, the contact resistance between the current collectors and k is large for the POMs [1531 , ]. The resistance invoked and the electrodes, the resistance of the solution and the by mass-transport limitations depends on v because the ionic resistance in the membrane [28]R . is independent flow rate governs at which rate fresh electrolyte is trans- Ohm of flow rate and with R   =  (2.806  ± 0.021) Ω cm higher ported to the electrodes, which can then undergo electron Ohm VRFB 2 –2 than typical values for the VRFB system, R ≈ 0.5 Ω cm transfer. Even at the highest flow rate (v  =  1.34  ml cm Ohm –1 −2 [29]. We assume that this high resistance stems mostly min ), R =  4.4 Ω cm and therefore diffusion presents diff from the ionic resistance of the membrane. Due to a lack a larger resistance than R and R combined. At lower Ohm CT of protons (pH  >  1), Li cations are used to enable charge flow rates, this problem is aggravated. There are two ways balance by their transport through the membrane and the to bring more unreacted electrolyte to the electrode in a diffusion coefficient for Li through a perfluorosulphonated given time: Current density pH of electrolytes –2 / mA cm pH Voltage / V Capacity / Ah Voltage / V Efficiency % Downloaded from https://academic.oup.com/ce/article-abstract/3/4/278/5550272 by DeepDyve user on 10 December 2019 284 | Clean Energy, 2019, Vol. 3, No. 4 A 3 2 –1 Cell: 1400 cm Normalised flow (ml (min cm ) ): R ohm Anolyte: 80 mM SiW 0.52 8.1 CT Catholyte: 80 mM PV 0.69 10 7.1 diff 0.99 1.24 5.9 5.3 1.34 8 4.8 4.4 0.4 0.3 0.41 0.38 0.37 0 0.2 0.39 0.37 0.1 2 0.38 0.0 2.8 2.8 2.8 2.8 2.8 2.8 –1 –0.1 2 34 5 2.6 2.8 3.0 3.2 0.4 0.6 0.8 1.0 1.2 1.4 Re (Z) / Ω cm 2 2 –1 Re (Z) / Ω cm Normalised flow / ml (min cm ) Fig. 6 Galvanostatic impedance data for the 1400-cm cell. (a) Nyquist plots for various normalized pump rates. (b) Detail of (a) showing the first semicircle. (c) Extracted resistances (R , R , R ) over the normalized flow. Measurements were made with no current bias at 36% SOC, which cor - ohm CT diff responds to the centre of the first one-electron redox wave of SiW. (i) Increase the concentration of the redox species to in- no vanadium signal, indicating that no PV has crossed crease the amount of reactant per pumped volume. over into the anolyte reservoir during the 1400 cycles. This This approach can be done without changing the result is in line with our earlier results that showed that power converter and has been shown earlier in this the negatively charged POMs do not cross cation-exchange work using the 25-cm cell (see Fig. 3). membranes [15]. CVs of both electrolytes were recorded (ii) Increase the pump rate and thereby the flow of elec- to check the state of health of the anolyte and to confirm trolyte. In order to increase the pump rate, the cell the result of the V NMR for the catholyte. Supplementary would have to be modified because, at the moment, Fig. 1b (see the online supplementary data) shows the CV the pressure drop within the cell is already at 900 mbar of the anolyte after 1400 cycles, which matches the typ- –2 –1 at 1.34 ml cm min , and the cell could most likely not ical CV of SiW (compare Fig. 2a). Similarly, the CV of the withstand higher pressure drops. Therefore, the cell anolyte shown in Supplementary Fig. 1c (see the online would have  to be converted to a design with a lower supplementary data) matches the CV of PV (compare Fig. impedance to flow, such as a flow-by design instead of 2a). In conclusion, the post-cycling analysis showed that a flow-through configuration [5, 32]. the POMs in the electrolyte were stable for 1400 cycles equalling 88 days of cycling. 3.3 Post-cycling analysis of electrolytes 4 Comparison of the two cells After 1400 cycles, the electrolytes used were extracted from the battery and investigated by V NMR and CV studies, the In order to compare the cycling performance of the two results of which are shown in the supporting information. cells under investigation, both the flow rate and the cap- Supplementary Fig. 1a (see the online supplementary data) acity are normalized, as shown in Fig. 7a. The 25-cm cell 51 25cm gives the V NMR spectra of the anolyte and the catholyte. reached a discharge capacity of Q   =  192 mAh, which dch theo,25cm The signal of the catholyte (red curve) shows the typical is 90% of its theoretical capacity of Q   =  214 mAh. dch 2 th fingerprint signal of PV [15, 22]. Two small additional For the 1400-cm cell, the 10 cycle that was described in peaks at a chemical shift of –505 p.p.m. and –525 p.p.m. can Fig. 4b is given for comparison. At a similar flow rate and 6– –2 be attributed to [VO ] , a polyoxovanadate that forms the same current density (4 mA cm), the coulombic ef- 10 28 1400cm from PV at pH  >2.3 but is in a pH-dependent dynamic ficiency of the larger cell is higher: η 100% versus 14 C 25cm equilibrium with PV [22]. We investigated the formation η  = 96%. It is assumed that the superior atmospheric 14 C 6– 2 of [V O ] and the time and pH dependency of this pro- containment of the 1400-cm cell keeps more oxygen from 10 28 cess earlier via V NMR and found that a pH value as low entering the tanks and therefore oxidation of reduced as 1.7 can be tolerated permanently and even lower values PV electrolyte is prevented. The difference in energy ef- 1400cm are acceptable for short periods of time [15]. All observed ficiency is larger, however, for the big cell: η   =  86% 25cm conversions are reversible and do not cause permanent as compared to η   =  64%. Therefore, the voltage effi- capacity loss. This indicates that PV in the catholyte ciency η of the 1400-cm cell must be higher than that was stable during the cycling and did not decompose to of the 25-cm cell (as η = η · η ). While the coulombic E V C single vanadium species. The anolyte (blue curve) shows efficiency is mainly given by the redox electrochemistry -Im (Z) / Ω cm -Im (Z) / Ω cm Norm. resistances / Ω cm Downloaded from https://academic.oup.com/ce/article-abstract/3/4/278/5550272 by DeepDyve user on 10 December 2019 Friedl et al. | 285 1.4 100 AB Anolyte: 80 mM SiW 25 cm cell: Catholyte: 80 mM PV 1.2 ohm η = 96%, η = 64% C E CT 1.0 10 R diff 0.8 η = 100%, η = 86% C E 0.6 1400 cm cell: ohm 0.4 CT 2 –2 –2 –1 0.2 R Cell: 25 cm ; 4 mA cm ; 1.08 ml cm min diff Anolyte: 80 mM SiW 2 –2 –2 –1 Cell: 1400 cm ; 4 mA cm ; 0.58 ml cm min Catholyte: 80 mM PV 0.0 0.1 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.5 1.0 4.0 4.5 5.0 5.5 6.0 6.5 2 –1 Normalised capacity Normalised flow / ml (min cm ) 2 2 2 Fig. 7 Comparison of a 25-cm cell and a 1400-cm cell with the same electrolyte. (a) A single charge–discharge cycle for the 1400-cm cell and the η η 25-cm cell with energy efficiency and coulombic efficiency given. The capacity is normalized to 1. (b) Resistances R , R and R for the 1400- E C Ohm CT diff 2 2 cm cell (larger circles) and the 25-cm cell (smaller squares) over the flow rate (normalized to membrane area). and the electrode materials, cell construction and its ma-the bigger cell performed better. This is indicated by the terials determine the voltage efficiency. To identify the higher energy efficiency during cycling, and confirmed by loss mechanism(s) that contribute to the difference in the lower resistances found in EIS studies. While R was Ohm voltage efficiency, we compare R , R and R for both similar in both cells, both R and R were considerably Ohm CT diff CT diff cells versus the normalized flow rate. Data points for the smaller in the 1400-cm cell. This suggests that the com- 1400-cm cell in Fig. 7b are the same as shown in Fig. 6b , mercial cell, even though it was designed to be used as a but here the resistances for the smaller cell are shown in VRFB, can yield a higher efficiency than the laboratory cell. addition. For both cells, R and R are independent of The most crucial parameter to improve is R (compare Ohm CT diff the flow rate, whereas R decreases with higher flow. R Fig. 6c). Looking at the equations for diffusion overvoltage diff Ohm (red data) is very similar for both cells, which is under - given by Vetter, an increase in the concentration of active standable, as this resistance should be determined mostly species should alleviate this problem [34]. As shown in Fig. by the membrane resistance and the same membrane is 3, R indeed depends heavily on mass flow. At high con- diff used in both cells. On average, the charge-transfer resist- centrations and for the highest measured rate R = 0.91 Ω diff ance is more than 12 times higher in the smaller cell, with cm . This demonstrates that increasing the concentration 2 2 25cm 2 1400cm 2 R  = 4.75 Ω cm versus R  = 0.382 Ω cm , and there- of electrolyte is a possible approach to reduce R . CT CT diff fore the charge transfer must be considerably faster in the 2 2 1400-cm cell than in the 25-cm cell. The difference in R CT values between the two cells cannot be due to the redox 5 Summary and conclusion electrochemistry, which is identical, and stems most likely from the used electrodes or their pre-treatment. The elec- In this study, we have shown that a recently developed trodes of the 25-cm cell, graphite felts (GFD, SGL Carbon), POM electrochemistry for flow batteries can work effect- were heated to 400°C for 24  h in laboratory atmosphere. ively in a redox flow battery. Data obtained for a 25-cm The electrodes in the 1400-cm cell were preconditioned in laboratory cell were quite satisfying, with a coulombic oxygen plasma. The exact parameters of this process are efficiency of 96%. The energy efficiency of 64% was con- unknown. When heat-treating the GFD 4.6 EA carbon felts siderably lower, probably due to specifications in the cell for 3 h at various temperatures in air, it was found that R design. It was also demonstrated that a large single cell CT reaches a minimum at around 600°C treatment tempera- of 1400  cm can be successfully employed. While the cell ture (Pfanschilling, unpublished work [33]). Whether this is was designed for the VRFB chemistry, it still shows good a long-term effect is yet to be investigated. performance for the investigated SiW–PV electro- 12 14 2 2 Also R (green data) is significantly larger in the 25-cm lytes. During 88  days, the 1400-cm cell was charged and diff cell than it is in the 1400-cm cell, varying with the flow discharged 1400 times. The initial discharge capacity of 1 1400 rate we determined factors from 4 to 7. The reason for this Q  = 4.70 Ah dropped to Q  = 3.95 Ah, which equals an dch dch –3 difference is unknown but we assume that the flow distri- average capacity loss of 0.53 10 Ah/cycle. During the cyc- –2 bution in the larger cell is more effective than the one in ling at 4 mA cm , a coulombic efficiency of η ≈ 99% and an the small cell. energy efficiency of η ≈ 86% were reached—both values The conclusion from the comparison of the 1400-cm larger than in the small laboratory cell. A post-cycling ana- 2 51 cell and the 25-cm cell using the same electrolytes is that lysis was performed on the electrolytes. Both V NMR and Voltage / V Normalised resistances / Ω cm Downloaded from https://academic.oup.com/ce/article-abstract/3/4/278/5550272 by DeepDyve user on 10 December 2019 286 | Clean Energy, 2019, Vol. 3, No. 4 cyclic voltammetry showed that the polyoxoanions in the or 80-µm thickness [36]. The 45-µm-thick membrane electrolyte were not damaged during the battery operation was found to be the optimum due to the combined and that the molecules had not crossed the membrane. effects of Ohmic losses and cross-over. We found that Using pH sensors in the electrolyte tanks, we could show SiW does not cross perfluorosulfonic acid (PFSA) that the pH value of the catholyte changes during cycling, membranes as thin as 40 µm due to size exclusion as expected for the PCET experienced by PV , but that the and electrostatic repulsion [15]. The minimum mem- average pH of the catholyte does not shift over 1400 cycles. brane thickness that still prevents cross-over should The average pH of the anolyte increases slightly during the be determined. experiment, from pH 1.5 to 2 during 1400 cycles. ○ Reduce the specific Ohmic resistance of the membrane. While the cell showed a high energy efficiency at low We assume that the high value for R stems from the Ohm current densities, this value dropped to η ≈ 40 % at 43 fact that Li cations are used for charge balance instead –2 mA cm . Using impedance spectroscopy, we found that of protons. The cell chemistry could either be trans- both the Ohmic resistance R and the charge-transfer ferred into a more acidic solvent, which allows the use Ohm resistance R were independent of the electrolyte of protons, or the membrane could be optimized for the CT flow rate, but that high flow rates decreased the mass- use of Li cations. A lower pH than currently employed transport resistance R . At the highest volumetric flow might lead to stability issues for PV [22]. diff 14 –2 –1 rate (1.34 ml cm min ), the measured resistances were • The geometry of the power converter and its flow design [ 5, as follows: R   =  2.78, R   =  0.39 and R   =  4.4 Ω cm Ohm CT diff 37]: Neglecting an effect of R , there are two limiting Ohm . Comparing the first electron transfer of SiW (k   =  4.2 12 0 cases for the type of rate control in a electrochemical –2 –1 2+ + –7 10 cm s ) and the VO /VO redox reaction (k  =  3 10 2 0 converter: charge-transfer control and mass-transport –1 cm s ), the electron transfer of the POM is more than five control [37]. In the former case, the rate of electron orders of magnitude faster than that of the latter [35 15], . transfer limits the current that can be drawn from This confirms the assumption from the introduction that, the cell; a larger overpotential or a larger surface area for redox couples with fast electron-transfer kinetics, the for the electrode can increase the rate. Under mass- total resistance R depends more on the cell param- total transport control, the supply of active species and its eters R and R than on the electrochemical param- diff Ohm removal after reaction determine the current; a limiting eter R . For higher concentrations, this will also remain CT current I ∝ v · c = c can be defined. Assuming cat/an true, with a larger influence from R . While the above Ohm that R is a measure for 1/I , the connection between I diff L L resistances were recorded at an electrolyte concentra- and c ˙ is empirically shown in Fig. 3. The proportionality tion of c   =  c   =  80  mM, we have shown in Fig. 3 that an cat depends on the geometry and typically turbulent flow –1 –2 R decreases with higher mass flows (mol min cm ) and diff allows higher I than laminar flow. As the POMs exhibit therefore concentrations. Simultaneously, R will de- CT fast electron transfer, and judging from Fig. 6c I needs , crease with higher concentrations, as it is inversely pro- to be increased to enhance the performance of the cell, portional to the concentration of redox species [ ]. The 7 which can be done in two ways: Ohmic drop R , which is governed by the membrane, Ohm ○ Increase the concentration c . This approach was cat/an will remain unchanged by a change in the concentration tested for the small cell with the result shown in Fig. of active material. 3. Clearly, a higher concentration of active species re- This indicates that, in order to increase the perform- duces R and increases I . diff L ance of the presented asymmetric POM RFB, the cell, spe- ○ Increase the rate at which fresh active material is sup- cifically parameters related to R and R need to be the Ohm diff plied to the electrodes. This can be done by increasing main focus for improvements. Components to work on are: the pump rate or by enhancing the spatial distribu- • Membrane/separator: The Ohmic drop R , which is tion of the mass flow in the cell to maximize electro- Ohm dominated by the membrane resistance, is likely to be lyte utilization. the highest resistance in a cell, with a higher concen- tration of active material, as R does not scale with Ohm the concentration of the redox electrolyte. To reduce Supplementary data R incurred from the membrane, two measures can Ohm Supplementary data is available at Clean Energy online. be taken: ○ Use a membrane thinner than the currently used FUMASEP—F-1075-PK (thickness 75 µm), as the ionic resistance of the membrane increases with thickness Acknowledgements of the membrane. For the VRFB, it was found that a This work was supported by a grant from Siemens AG. Additional sulphonated fluorinated poly(arylene ether) mem- support comes from the North East Centre of Energy Materials brane with 45-µm thickness enabled a higher power (NECEM) funded by EPSRC (EP/R021503/1) and Newcastle density than the same type of membrane with 28 University. Downloaded from https://academic.oup.com/ce/article-abstract/3/4/278/5550272 by DeepDyve user on 10 December 2019 Friedl et al. | 287 derivatives on glassy carbon electrodes. AnalChem 1999; References 71:4081–7. [1] Ponce  de  Leon  C, Friasferrer  A, Gonzalezgarcia  J, et  al. [20] Fink H, Friedl J, Stimming U. Composition of the electrode de- Redox flow cells for energy conversion. J Power Sources 2006; termines which half-cell’s rate constant is higher in a van- 160:716–32. adium flow battery. J Phys Chem C 2016; 120:15893–901. [2] Weber AZ, Mench MM, Meyers JP, et al. Redox flow batteries: a [21] Friedl  J, Al-Oweini  R, Herpich  M, et  al. Electrochemical review. J Appl Electrochem 2011; 41:1137–64. studies of tri-manganese substituted Keggin polyoxoanions. [3] Holland-Cunz  MV, Cording  F, Friedl  J, et  al. Redox flow bat- Electrochim Acta 2014; 141:357–66. teries—concepts and chemistries for cost-effective energy [22] Selling  A, Andersson  I, Pettersson  L, et  al. Multicomponent storage. Front Energy 2018; 12:198–224. polyanions. 47: The aqueous vanadophosphate system. Inorg [4] Zhang  M, Moore  M, Watson  JS, et  al. Capital cost sensitivity Chem 1994; 33:3141–50. analysis of an all-vanadium redox-flow battery. J Electrochem [23] Keita  B, Nadjo  L. New aspects of the electrochemistry of Soc 2012; 159:A1183–8. heteropolyacids.J Electr oanal Chem Interfacial Electrochem 1987; [5] Arenas LF, Ponce de León C, Walsh FC. Engineering aspects of 227:77–98. the design, construction and performance of modular redox [24] Keggin  JF. The structure and formula of 12-phosphotungstic flow batteries for energy storage. J Energy Storage 2017; 11:119–53. acid. Proc R Soc A Math Phys Eng Sci 1934; 144:75–100. [6] Noack  J, Roznyatovskaya  N, Herr  T, et  al. The chemistry of [25] Pezeshki  AM, Sacci  RL, Veith  GM, et  al. The cell-in-series redox-flow batteries. Angew Chemie Int Ed 2015; 54:9776–809. method: a technique for accelerated electrode degradation in [7] Friedl J, Stimming U. Determining electron transfer kinetics at redox flow batteries. J Electrochem Soc 2016; 163:A5202–A5210. porous electrodes. Electrochim Acta 2017; 227:235–45. [26] Skyllas-Kazacos  M, Grossmith  F. Efficient vanadium redox [8] Zhang H. Development and application of high performance flow cell. J Electrochem Soc 1987; 134:2950–4. VRB technology. In: IFBF 2017—International Flow Battery Forum, [27] Schweiss R, Pritzl A, Meiser C. Parasitic hydrogen evolution at Manchester, UK, 2017. different carbon fiber electrodes in vanadium redox flow bat- [9] Rabiul Islam FM, Al Mamun K, Amanullah MTO Smart Ener . gy teries. J Electrochem Soc 2016; 163:A2089–94. Grid Design for Island Countries. Basel, Switzerland: Springer, [28] Sun C-N, Delnick FM, Aaron DS, et al. Probing electrode losses in all-vanadium redox flow batteries with impedance spec- [10] Lim  HS, Lackner  AM, Knechtli  RC. Zinc-bromine secondary troscopy. ECS Electrochem Lett 2013; 2:A43–5. battery. J Electrochem Soc 1977; 124:1154–7. [29] Liu  QH, Grim  GM, Papandrew  AB, et  al. High performance [11] Johnson DA, Reid MA. Chemical and electrochemical behavior vanadium redox flow batteries with optimized electrode con- of the Cr(lll)/Cr(ll) half-cell in the iron-chromium redox en- figuration and membrane selection. J Electrochem Soc 2012; ergy system. J Electrochem Soc 1985; 132:1058–62. 159:A1246–52. [12] Nice AW. NASA Redox System Development Project StatusCleveland, [30] Samec Z, Trojanek A, Langmaier J. Diffusion coefficients of al- OH: Cleveland, OH. 1981. kali metal cations in Nafion® from ion-exchange measure- [13] Scamman DP, Reade GW, Roberts EPL. Numerical modelling of ments. J Electrochem Soc 1997; 144:4236. a bromide-polysulphide redox flow battery. Part 1: Modelling [31] Friedl J, Bauer CM, Rinaldi A, et al. Electron transfer kinetics of 2+ + approach and validation for a pilot-scale system. J Power the VO / VO —reaction on multi-walled carbon nanotubes. Sources 2009; 189:1220–30. Carbon N Y 2013; 63:228–39. [14] Morrisey P. Regenesys: a new energy storage technology. Int J [32] Zhou  XL, Zhao  TS, An  L, et  al. Critical transport issues for Ambient Energy 2000; 21:213–20. improving the performance of aqueous redox flow batteries. J [15] Friedl  J, Holland-Cunz  MV, Cording  F, et  al. Asymmetric Power Sources 2017; 339:1–12. polyoxometalate electrolytes for advanced redox flow bat- [33] Pfanschilling FL. Unpublished results. teries. Energy Environ Sci 2018; 11:3010–8. [34] Vetter KJ. Electrochemical Kinetics—Theoretical and Experimental [16] Laborda  E, Henstridge  MC, Batchelor-McAuley  C, et  al. Aspects, English edn. New York/London: Academic Press Inc., Asymmetric Marcus-Hush theory for voltammetry. Chem Soc Rev 2013; 42:4894–905. [35] Gattrell M, Park J, MacDougall B, et al. Study of the mechanism [17] Chen H-Y, Friedl J, Pan C-J, et al. In situ X-ray absorption near of the vanadium 4+/5+ redox reaction in acidic solutions. J edge structure studies and charge transfer kinetics of Na Electrochem Soc 2004; 151:A123. [V O ] electrodes. Phys Chem Chem Phys 2017; 19:3358–65. [36] Chen  D, Hickner  MA, Agar  E, et  al. Optimizing membrane 10 28 [18] Aparicio  PA, Poblet  JM, López  X. Tungsten redox waves in thickness for vanadium redox flow batteries. J Memb Sci 2013; [XMW O ] n- (X = P, Si, Al and M = W, Mo, V, Nb, Ti) Keggin 437:108–13. 11 40 compounds—effect of localised/delocalised charges. Eur J [37] Walsh  FC, Pletcher  D. Electrochemical engineering and cell Inorg Chem 2013; 2013:1910–6. design. In: Pletcher D, Tian Z-Q, Williams D (eds). Developments in Electrochemistry: Science Inspired by Martin Fleischmann. [19] Yang H-H, McCreery RL. Effects of surface monolayers on the electron-transfer kinetics and adsorption of methyl viologen Hoboken, NJ: John Wiley & Sons, 2014, 95–109.

Journal

Clean EnergyOxford University Press

Published: Dec 6, 2019

References