Dithiolene Complexes of First‐Row Transition Metals for Symmetric Nonaqueous Redox Flow Batteries

Abstract Five metal complexes of the dithiolene ligand maleonitriledithiolate (mnt2−) with M=V, Fe, Co, Ni, Cu were studied as redox‐active materials for nonaqueous redox flow batteries (RFBs). All five complexes exhibit at least two redox processes, making them applicable to symmetric RFBs as single‐species electrolytes, that is, as both negolyte and posolyte. Charge–discharge cycling in a small‐scale RFB gave modest performances for [(tea)2Vmnt], [(tea)2Comnt], and [(tea)2Cumnt] whereas [(tea)Femnt] and [(tea)2Nimnt] (tea=tetraethylammonium) failed to hold any significant capacity, indicating poor stability. Independent negolyte‐ and posolyte‐only battery cycling of a single redox couple, as well as UV/Vis spectroscopy, showed that for [(tea)2Vmnt] the negolyte is stable whereas the posolyte is unstable over multiple charge–discharge cycles; for [(tea)2Comnt], [(tea)2Nimnt], and [(tea)2Cumnt], the negolyte suffers rapid capacity fading although the posolyte is more robust. Identifying a means to stabilize Vmnt 3−/2− as a negolyte, and Comnt 2−/1−, Nimnt 2−/1−, and Cumnt 2−/1− as posolytes could lead to their use in asymmetric RFBs.


Introduction
To meet rising global energy demands and to reduce fossil fuel consumption, renewables such as solar and wind are being increasingly implemented as energy sources. However, owing to their intermittentn ature,e fficient and cost-effective grid-scale energy storage is required before solara nd wind energy can achieve widespread implementation. [1,2] Ap romising candidate for grid-scale energy storage is the redox flow battery (RFB) technology,w hereby solutions of electroactivem aterials are pumped to/frome xternal tanks to the electrode interface for charging/discharging. [3][4][5] As energy is stored externally to the electrochemical reactor,t he capacity can be increased independently of the battery power.A tp resent,c ommercial RFBs utilize aqueous electrolyte solutions of inorganic metal salts, however, despite continualp rogress in powero utputs and efficiencies being made, the cell potentiali si nherently limitedb y the narrow (1.23 V) electrochemical window of water.I nstead, the developmento fn onaqueous RFBs, which use organic solvents with wide electrochemical windows,i sa nticipated to improve the voltage outputs. [6][7][8][9][10] Acetonitrile (MeCN) is an attractive solvent for nonaqueous RFBs, andi st he main solvento f choice here, owing to its wide ( % 5V)e lectrochemical window as well as low viscosity (0.34 vs. 0.89 MPa sfor water) and moderate dielectric constant(35.9 vs. 78.4 for water). [11] Metal-ligandc oordination complexes are good candidates for nonaqueous RFBelectrolytes as they can be stable in multiple oxidation states and have high solubility in organic solvents.F urthermore, careful choice of metal ion as well as modification of the ligand scaffold (e.g.,s olubilizing groups, denticity,d onorg roups) can allow for fine tuning of the desired properties for RFB applications. [12][13][14][15] Indeed, several metal coordination complexes have been tested as electrolytes for nonaqueous RFBs with cell potentials in excess of 1.23 V; these contain acetylacetonate, [12,[16][17][18][19][20][21] bipyridine, [13,15,[22][23][24][25] phenanthroline, [26,27] terpyridine-like, [14,15] trimetaphosphate, [28] and macrocyclic [29,30] ligands.
An attractive and simple RFB system, which avoids crosscontamination of the two electrolyte solutionst hrough membrane crossover,e mploys as ingle species electrolyte in as ymmetric cell-that is, ab attery that uses only one species as both negolyte (electrolyte that is reduced on battery charging, that is, anolyte) and posolyte (oxidized electrolyte, that is, catholyte). In this approach, the battery does not suffer from irreversible capacity loss and self-discharge through the mixing of electrolytes, and insteadarebalancing procedure to restore the original negolyte/posolyte composition can be performed, as is done in aqueous all-vanadiumR FBs. [31] For this, the redoxactive species needs to have (at least) two redox processes and be stable across the three associated redox states. Dithiolene ligandsa re of particulari nterest here, as they are non-innocent when bound to transition metal ions leading to complexes with multiple redox events, through oxidation andr eductionc entered on either the metal or dithiolenel igand. [32][33][34][35] Recently,t he vanadium complex of the dithiolene ligand 1,2-dicyanoethylene-1,2-dithiolate (maleonitriledithiolate, Five metal complexeso ft he dithiolene ligand maleonitriledithiolate( mnt 2À )w ith M = V, Fe, Co, Ni, Cu weres tudied as redox-active materialsf or nonaqueousr edox flow batteries (RFBs). All five complexes exhibit at least two redox processes, making them applicable to symmetric RFBs as single-species electrolytes, that is, as both negolyte and posolyte.  [36] V mnt 2À undergoes two reversible metal-centered oneelectron reductionsa nd one reversiblel igand-centered oneelectron oxidation in MeCN solution. [35] The oxidation event was charged against the first reduction event in as tatic H-cell experiment,g iving a1 .09 Vc ell with 90 %c oulombic and 20 % voltaic efficiencies. [36] In this work, we extend the application of [(tea) 2

Synthesis
The widely used synthetic methodf or the mnt 2À ligand, first reported by Bähr and Schleitzeri n1 957, proceeds first by formation of the sodium cyanodithioformate (NaNCCS 2 ) intermediate from sodium cyanide and carbon disulfide, followed by dimerization/desulfurization to give Na 2 mnt (Scheme 1a). [37][38][39] However,i nt he interests of accessing Na 2 mnt from more environmentally benign and less toxic starting materials, an alternative methodr eported by Hoepping and co-workersw as followed (Scheme 1b). [40] Here, the intermediate NaNCCS 2 is prepared from chloroacetonitrile with NaOH and sulfur in DMF in yields identicalt ot he more hazardous route (69 %, this work;7 1%, [39] from NaCN + CS 2 ), despite not achieving the almost quantitative yields (88-97%)reported by Hoepping and co-workers. [40] The isolated intermediate was then dissolved in water and allowed to stand for 12 ht od imerize to Na 2 mnt, which was isolated by filtration to remove sulfur followed by evaporation of the filtratet og ive at anbrown solid, in quantitative yield from NaNCCS 2 .T he Na 2 mnt crude product was purified by first drying under high vacuum at 80 8Cf or several hours before recrystallization from EtOH/ Et 2 Ot og ive ab right-yellowm icroanalytically clean powder in moderate yield (49 %f rom NaNCCS 2 ,3 4% overall).

Electrochemical properties
Cyclic voltammetry (CV) of the five complexes in relevant conditions was performed on glassycarbon to assess their suitability as candidates forn onaqueous RFB electrolytes ( Figure 2, Ta ble 1). All complexes exhibit at least one oxidation and one reduction process, making them potential electrolytes for single-species flow batteries, that is, as both the posolyte and negolytes olutions. [(tea) 2 V mnt ]d isplays two reversible reduction and one reversible oxidation eventsa tÀ2.032 V, À0.849 V, and 0.230 V( vs. Fc/Fc + ). In aR FB, ap osolyte of [(tea) 2 V mnt ]a ccessingthe 0.230 Voxidation process could be charged against the À0.849 Vr eduction process in an egolyte of the same material to give ab attery of V cell = 1.08 V. To achieve even greater Scheme1.Synthetic pathways to the Na 2 mnt ligand:a )from NaCN and CS 2 in DMF [37][38][39] and b) from chloroacetonitrile, NaOH, and sulfur in DMF. [40]  potential, charging the 0.230 Vp rocess against a( pre-charged) negolyte reduction at À2.032 Vw ould yield V cell = 2.26 V. The bis-mnt complexes all exhibit one reduction and one oxidation process, allowingp otential use as single-species RFB electrolytes, with cell potentials of 1.12-1.92 V(Ta ble 1). Single-species RFBs operating at saturated concentrationso ft he dithiolene complexes would therefore have quite favorable theoretical maximum energy densitieso f7 .7-16 Wh L À1 ,w ith the exception of the poorly soluble [(tea)Fe mnt ]a t0 .6 Wh L À1 (Table 2). Each complex also exhibits at least one irreversible oxidation process at approximately 0.5-1 V( vs. Fc/Fc + ;s ee the Supporting Information, Figures S1-S5);h owever,i nt his work, the threshold potentials for charge-discharge battery cycling experiments are carefully chosen to avoid these irreversible oxidation processes.

V mnt battery cycling experiments
In apreviousreport by Cappillino and co-workers, [36] [(tea) 2 V mnt ] was shown to have promising charge-discharge performance in non-flow H-cell experiments.H ere, we anticipated improved performance in af low cell, owing to the enhanced mass transport and decreased cell resistances arising from pumped electrolytesa nd smaller inter-electrode separations, respectively. Assessmento f[ (tea) 2 V mnt ]i nM eCN as as ymmetric electrolyte was performed in as mall-scale RFB with 2.08 cm 2 carbon paper electrodes, ap orous Celgard separator,a nd 10 mL of electrolyte in each half-cell (see the Supporting Information for af ull description). In total, 100 charge-discharge cycles were recordeda taconstant current density of AE 0.48 mA cm À2 with the threshold potential set to 1.5 Vf or charge cycles to avoid accessing the irreversible oxidationp rocess at approximately 1V(vs. Fc/Fc + ), and 0.3 Vf or discharge (Figure3a). On the initial charge, aplateau from approximately 1.0-1.2 Viso bserved, indicating that chargingi so ccurring about the expected cell potentialf or the one-electront ransfer process of 1.08 V, until a maximum capacity of 0.253 mA h, whichi s9 4% of the 0.268 mA ht heoretical capacity for ao ne-electronp rocess. Upon discharge,avoltage plateau from approximately 1.1-1.0 Vi so bserved, corresponding to ah igh voltaice fficiency (ratio of discharge to chargep otential) of9 5% for cycle 2 ( Figure 3b). The discharge capacity,0 .158 mA ho r5 9%,i sl ess than the charge capacity,c orrespondingt oalow coulombic efficiency (discharge to charge capacity ratio) of 62 %f or cycle 2, resulting in an energy efficiencyo f5 9%.T he charge-dischargeb ehavior is consistent for ten cycles,b ut by the 20th cycle as econd discharge plateau is observeda ta round 0.8 V, indicating that the battery compositionh as changed such that alternative redox processes are occurring. Although charge-discharge wasa chieved up to 100 cycles, good per- Table 1. Redox potentials (vs. Fc/Fc + ), as measured by CV at 100 mV s À1 on glassy carbono n1m m solutionso ft he complex in 0.1 m TBAPF 6 MeCN solution. The theoreticalcell potential, V cell = E 1/2 ox ÀE 1/2 red ,for asymmetric flow battery of the corresponding electrolyte solution is also given.
Complex [a] The larger cellp otentialf or ab attery employing the second reduction process of V mnt as the negolyte. formance is only observed to around2 0cycles.I ndeed, by operating in af low cell, enhanced performance is achieved in comparison to the previously reportedn on-flowing H-cell charge-discharge experiments. [36] Although the coulombic efficiency decreased from approximately 90 %t oa bout 60-70 %, the voltaice fficiency was markedlyi ncreased from around 20-25 %t o9 5%,r esulting in at ripling of the energy efficiency, from about 20 %t o6 0%,h ighlighting the importance of testing the proposed RFB electrolytes in flow conditions. Next, we explored the effects of changing the flow cell separator as well as electrolyte solvent on the battery cycling performance. Switching the Celgard for aF umapem F-930 cation exchange membrane (CEM) gave as lightly diminished performance with similar discharge capacities but energy efficiencies approximately 10 %l ower for F-930 than for Celgard( Figures S12 and S13).
We attribute the poorer performance of the appliedC EM to its smaller thickness (30 mm) and the undesirable physical properties of the membrane material in MeCN solvent. Presently,m embranes for use in organic solvents do not yet exist and those designed for aqueous electrolyte typically display excessive swelling and fragilityi ns ome organic solvents. In contrast, Celgard, being composed of polyethylene, demonstrates superb chemical stabilitya nd mechanical properties in aggressive electrolytes. Using propylene carbonate (PC) as the electrolyte solvent appearst oa gain give as imilar performance, with slightly better coulombic and slightly worse voltaic effi-ciencies,r esulting in energy efficiencies just below 50 %( Figures S15 and S16). We attribute the poorer battery performance to the lower conductivity and higher viscosity of PC, which results in higherc ell resistances. However,a gain the efficiencies obscure the overall poor performance of [(tea) 2 V mnt ]i n PC as the discharge capacities are in fact much lower than in MeCN at 28 %a nd 59 %, respectively,o nc ycle 2. The large capacity fade in PC solvento ccurs mostly owing to ap oor first cycle performance with charge-discharge capacities being 105 %/33 %a nd 43 %/28 %f or cycles 1a nd 2, respectively.
[(tea) 2 V mnt ]e xhibits as econd reversible reduction process, V mnt ,w hich if charged as an egolyte against ap osolyte V mnt 2À/1À process, gives aR FB of V cell = 2.26 V, which would have al arger theoretical energy density of 16 Wh L À1 .T oa ccess the second reduction process in an RFB, first ab attery of V mnt 2À in each half-cell was charged to 1.5 V, to give V mnt 3À negolyte and V mnt 1À posolyte solutions. The V mnt 1À posolyte was discarded and replaced with fresh starting material, giving V mnt 3À negolyte and V mnt 2À posolyte initial solutions, which were then subjected to charge-discharge cycling up to 2.7 V. Performing this experiment in both MeCN and PC gave initial charging curves with plateaus at 2.2-2.5 Va round the expected potential of 2.26 V; however, essentially zero capacityw as observedu pon discharge and subsequentc harge cycles (Figures S14 and S17).
[(tea) 2 V mnt ]w as furtheri nvestigated as ar edox-active material for nonaqueous RFBs by assessing the performance as an egolyte anda sap osolyte separately by independents ingle redox couple cycling. Flow cell experiments were performed for both the negolyte and posolyte for 1mm [(tea) 2 V mnt ]i n 0.1 m TBAPF 6 (tetrabutylammonium hexafluorophosphate) MeCN solution with Celgard separator.T he negolyte experiment, whereby the V mnt 3À/2À redox couple is charged/discharged, reveals good performance up to 100 cycles ( Figure 4a)w ith low overpotentials ( % 0.1 V) and discharge capacity fading from 63 %oncycle 1to36% for cycle 100 (Figure 4b, blue data). The posolyte experiment, cycling of V mnt 2À/1À showedp oorer performance compared with the negolyte,w ith the capacity fading much faster until almostz ero capacityr emained by cycle 30 (Figure 4b,r ed data). These independentb attery cycling data for the negolyte and posolyte solutionso f[ (tea) 2 V mnt ]s how that in the symmetric RFBt he posolyte, that is, the V mnt 2À/1À redoxp rocess, is less stable and is mostly responsible for the significant capacity fade observed by cycle 50 (Figure 3).
For each of the M = V, Co, Ni, Cu complexes, the posolyte and negolyte solutions weree xtracted from the symmetric RFB charge-discharge experiments after the first charge cycle and measured by UV/Vis spectroscopy to observe their stability over time. For [(tea) 2 V mnt ], the charged negolyte solution gives ad istinct spectrum from the uncharged electrolyte, most notably with the absence of the peaks at 308 and 580 nm, and appears to be stable with minimal change in the spectrum over 18 h( Figure 5). The posolyte V mnt 1À also gives ad istinct spectrum immediately after the initial charge cycle, witht he loss of the 308 nm peak and an increase in intensity and shift to lower wavelength of the 580 nm peak ( Figure S35). However, over 25 minthe spectrumevolves, most notably with the reap- pearance of the peak at 308 nm, to give an almost identical spectrum to that of the initial uncharged V mnt 2À electrolyte ( Figure 5). These data are in agreement with the symmetric single-redox couple flow cell data (Figure 4), with the V mnt 3À negolyteb eing more stable than the V mnt 1À posolyte. Furthermore, the spectrar eveal that the posolyte solutioni ss elf-dischargingt ot he initial V mnt 2À dianion, and that this process occurs over as hort time frame outside of the battery cycling environment. The mechanism for the self-discharge of the V mnt 1À speciesi su nclear;h owever,i ti se vident that ar educing agent( possibly trace water)m ust be present in the electrolyte to chemically reduce V mnt 1À .
Co mnt battery cyclinge xperiments [(tea) 2 Co mnt ]c harge-discharge cycling revealed voltagep lateaus centered around the expected V cell of 1.68 Va nd excellent voltaic efficiencies of 96 %f or the first ten cycles ( Figure 6). Coulombic and energy efficiencies were 63-69 % and 61-66 %o ver the first ten cycles, with discharge capacity fading from 56 %t o3 7%.D espite good performancei nitially, the discharge capacity faded steadily to 14 %o nc ycle 50, and to essentially zero capacity( 6%)b yt he 100th cycle ( Figure 6). In addition, an unexpected second plateau near the 2.2 V threshold was observed on charge,w hich became more prominent with cycling, indicating that the battery chemistry evolvedw ith increasing cycle number.    In the independent Co mnt 3À/2À negolyte single redox couple battery cycling, there is steady, almost complete, capacity fade over 50 cycles from 38 %to1% (Figures S27 and S29). The posolyte in this case, that is, the Co mnt 2À/1À singleredox couple, appears more stable to multiple charge-dischargec ycles,w ith almostz ero overpotential, and ac apacity fade from 75 %t o 54 %o ver 50 cycles (Figures S28 and S29).
UV/Vis spectrao f[ (tea) 2 Co mnt ] ( Figures S36-S38) indicate that the charged Co mnt 3À negolyte solutioni su nstable, giving an almosti denticals pectrum to the initial Co mnt 2À solution after only af ew minutes. The posolyte Co mnt 1À speciesi sm uch more robust in solution, giving ad istinct UV/Vis spectrum, whichi s almostu nchanged over 22 h. This is in line with the symmetric single redox couple battery cycling data, for which the posolyte-onlyC o mnt 2À/1À battery significantly outperformed the negolyte-only Co mnt 3À/2À system.

Ni mnt batterycycling experiments
Battery cycling of [(tea) 2 Ni mnt ]r eached only small capacities of 23-32 %o nc harge for the first five cycles and only small voltage plateaus around the expected V cell of 1.92 Vw ere observed ( Figure S19). Upon discharge, the cell potentials teadily decreased, with no plateau,t ot he lower threshold potentialo f 1V with < 15 %c apacity.R epeatingt he experiment in PC solvent resulted in ag reat improvement in battery cycling performance (FiguresS20 and S21), with the initial charge cycle showingalong plateau around 1.92 Vu pt oacapacity of 97 %. The first cycle resulted in al arge capacity fade, with a discharge capacity of 41 %. The independentr edox couple RFB experimentso f [(tea) 2 Ni mnt ]i nM eCN are perhaps the most insightful, with the negolyte retaining almost no capacity and the posolyte showing av ery robust performance with respectt ol ong-term cycling ( Figure 7). The Ni mnt 3À/2À negolyte system shows ap oor initial discharge capacity of 16 %, which rapidly fades to 4% after just ten cycles (Figure 7b,b lue data), indicating that [(tea) 2 Ni mnt ]i su nstable as the negative electrolyte, and would account for the very poor battery cycling performance of the MeCN symmetric RFB, which displayed almost zero discharge capacityo nt he first cycle ( Figure S19). In contrast, the Ni mnt 2À/1À posolyte system is very stable to charge-discharge cycling with ac apacity fade from 66 %t o5 1% over 100 cycles (Figure 7). Despite [(tea) 2 Ni mnt ]b eing shown to be ineffective as as ingle-speciese lectrolyte in MeCN for symmetric RFBs, the Ni mnt 2À/1À redox couple appears to be verys table over multiple charge-discharge cycles,s oc ouldb eu tilized as ap osolyte material in an asymmetric RFB.
UV/Vis spectra of charged electrolyte solutionsof[ (tea) 2 Ni mnt ] (Figures S39-S41) indicatet hat the charged Ni mnt 3À negolyte solutioni sv ery unstable,g iving an almost identical spectrum to the starting electrolyte Ni mnt 2À solution after only af ew minutes. The reduced species Ni mnt 3À in MeCN and1 ,2-dimethoxyethane solutions have previously been observed to be air-sensitive and unstable in solution [44] -indicatingt hat in our battery system,e ven trace amountso fo xygen may be causing rapid discharge of the Ni mnt 3À species to the startingN i mnt 2À state resulting in very small capacity retention for the symmetric cell. Despite this, the posolyte Ni mnt 1À species is far more stable in solution, giving ad istinct UV/Vis spectrum, which showsl ittle change over 22 h.

Cu mnt battery cycling experiments
[(tea) 2 Cu mnt ]w as tested under 100 charge-dischargec ycles at AE 0.48 mA cm À2 constantc urrent density in MeCN with ae ither aC elgards eparator or Fumapem F-930 cation exchange membrane ( Figures S22-S25).T he initial chargec ycle with Celgard reachedacapacity of 0.358 mA h( 134 %), suggesting that selfdischargeo ccurred during charging. The charge-discharge cycles exhibit voltage profiles with plateauss lightly above and below V cell = 1.12 Va t1 .1-1.3 Va nd 1.1-0.85 Vo nc harge and discharge, respectively,g iving ac onsistentv oltaic efficiency of 85 %o ver 100 cycles (Figures S22a nd S23). The coulombic efficiencya nd hence alsoe nergy efficiency remainc onsistenta t 60-68 %a nd 51-58 %, respectively,o verc ycles 2-100. The overall energy efficiency of this system is comparable to [(tea) 2 V mnt ]a nd [(tea) 2 Co mnt ]u nder the same conditions, however,t he capacity retention of [(tea) 2 Cu mnt ]i sf ar superior,w ith the discharge capacity only fading from 65 %t o5 3% from cycle 1t o5 0, and retaining 43 %a tt he 100th cycle. Moreover, the excellent solubility of [(tea) 2 Cu mnt ]i nM eCN, 0.91 m,m ake For [(tea) 2 Cu mnt ], as is seen with the other bis-mnt complexes [(tea) 2 Co mnt ]a nd [(tea) 2 Ni mnt ], the negolyte is less stable with a poor discharge capacity of 23 %a nd steady fade over 100 cycles to 9% (Figures S31 andS 33). Despite this, the [(tea) 2 Cu mnt ]p osolyte single redox couple battery,t hat is, cycling of Cu mnt 2À/1À shows am ore stable performance (Figures S32 and S33), with discharge capacities approximately twice that of the negolyte over the first ten cycles before fading to as imilar capacity by cycle 100.
For [(tea) 2 Cu mnt ], both the posolyte andn egolyte solutions give distinct spectra to the initial electrolyte and are stable for at least af ew minutes ( Figures S42-S44). However,b oth chargede lectrolytes appear to self-discharge over the course of 24 ht og ive spectra resembling that of the initial Cu mnt 2À electrolyte-this is unsurprising given that both the posolyte and negolyte have similar dischargec apacities after 100 cycles in the independents ingle redox couple battery cycling experiments.U nlike the M = V, Co, Ni complexes, which each had one of either the posolyte or negolyte observed by UV/Vis to rapidly self-discharge, both posolyte and negolyte solutionso f [(tea) 2 Cu mnt ]a re initially stable, allowing 100 charge-discharge cycles in the symmetric RFB with low capacity fade.

Comparison of M mnt batterycycling performances
In as ymmetric full-cell RFB, [(tea) 2 V mnt ]a nd [(tea) 2 Co mnt ] showedm odest performances, with the initial ten cycles displaying high voltaice fficiencieso f9 5-96 %e ach;h owever, long-term cycling was not possible with significantd ischarge capacityf ade to 20 %a nd 14 %, respectively,b yc ycle 50 (Figure 8a). [(tea) 2 Cu mnt ]d isplays ac omparable performance over ten charge-discharge cycles, however,i ts hows superior capacity retention with as maller capacity fade to 43 %o ver 100 cycles (Figure 8a), and is the best performing symmetric RFB studied here. For [(tea) 2 Ni mnt ], which displayed the largest cell potential for asimple one-electron disproportionation symmetric electrolyte, almost zero discharge capacity could be achievedi nM eCN ( Figure 8a); however,aswitch to propylene carbonate solvent did allow for charge-discharge cycling, albeit with poor performance. Battery cycling of [(tea)Fe mnt ]r eturned almost zero capacity upon discharge in the first cycle, and then failed to hold any significant capacity upon recharging ( Figure S18), possibly arising from the poor reversible behavior of the redox process at À0.689 V( vs. Fc/Fc + ). In addition, the plateau voltage at approximately 0.9-1.2 Vi sw ell below the expected V cell of 1.42 V, indicating that alternative redox processes are occurring during the charge cycle.
Independent single redox couple flow cell experiments revealed that for [(tea) 2 V mnt ]t he negolyte V mnt 3À/2À is more robust than the posolyte V mnt 2À/1À with greaterc apacity retention ach-ieved for the negolyte-only experiment (Figure 8b andc ), despite both processes being reversible on aC Vt imescale. Ap re- vious study on the V mnt redoxs eries, using X-ray absorption spectroscopy andD FT calculations, determined that oxidation of V mnt 2À = [V IV (mnt 3 6À )] 2À is al igand-centered process to the diradical [V IV (mnt 3 5À C)] À = V mnt 1À species andi nvolves ac hange in geometry from distorted octahedral to trigonal prismatic, whereas reduction to V mnt 3À is metal-centered, that is, V IV to V III . [35] Although no conclusions on the stability of the redox products were made in the above report,a nd both redox processes are reversible by CV, [35] two different processes (ligandvs. metal-centered redox) appear to give RFB electrolytes of contrasting stability.
Conversely,f or the square-planar bis-mnt complexes (M = Co, Ni, Cu), it was the posolyte that showedg reater capacity retention over multiple charge-dischargec ycles of the single redox couple, whereas the negolyte in each case was unstable (Figure 8b and c). Ap reviouss tudy,i nT HF solution, found that metal-centered reduction of Co mnt 2À gives the air-sensitive unstable Co mnt 3À trianion,w hereas ligand-centered oxidation to the Co mnt 1À monoanion is more robust. [45] Similarly,f or Ni mnt 2À , reduction is metal-centered and the Ni mnt 3À product has been observed to be unstablei na ir,r everting back to the dianion. [44] Trace amounts of oxygen present in the solvent, despite working under glovebox conditions, could well be responsible for the rapid capacity fade of the negolytes of the Co mnt and Ni mnt RFBs. Another previous dithiolene study observed that weak protic acids oxidizet he basic and strongly reducingC o mnt 3À and Ni mnt 3À species in THF solution,e ither directly to M mnt 2À ,o r throughp rotonation to [M(H)(mnt) 2 ] 2À hydrides followed by subsequentd ecomposition to M mnt 2À andH 2 ; [46] this could be another possible mechanism for the self-discharge of Co mnt 3À and Ni mnt 3À negolytes observed by UV/Vis in the presents tudy. More recently,s ulfur K-edge X-ray absorption spectroscopy has been used to determine Ni mnt 2À oxidationt oN i mnt 1À is ligandcentered, [34] and resultsi navery stable Ni mnt 2À/1À posolyte system in the present study.A gain, we note ac ontrastb etween ligand-versus metal-centered redoxp rocesses and the stabilityo ft he resulting RFB electrolytes;h owever,c onversely to six-coordinate tris-mnt V mnt where metal-centered reduction gives as table negolyte, for four-coordinate bis-mnt Co mnt and Ni mnt it is ligand-centered oxidation that results in the more stable posolytes. The posolyte of Cu mnt displays poorer capacity retention than the posolytes of Co mnt and Ni mnt ( Figure 8c); interestingly,o xidation of Cu mnt 2À to Cu mnt 1À is am etal-centered Cu II to Cu III process, [34] in contrastt ol igand-centered oxidation of Co mnt 2À and Ni mnt 2À . The best performing electrolytes are the [(tea) 2 V mnt ] 3À/2À negolyte (Figure 8b), and the [(tea) 2 Co mnt ] 2À/1À and [(tea) 2 -Ni mnt ] 2À/1À posolytes (Figure8c), which retain approximately 40-50 %c apacity after 100 charge-discharge cycles and are therefore promising foruse in asymmetricRFBs.

Conclusions
Five transition-metal complexes of the dithiolene ligand mnt have been assessed fora pplication as redox-active materials for single-species electrolytes in symmetric nonaqueous RFBs. The V, Co, Ni, and Cu complexes all exhibit at least two revers-  2 Cu mnt ] were able to be charged/discharged for up to 100 cycles with high voltaice fficiencies. However,c apacity retentiono ver multiple cyclesp roved challenging, with 43 %r etention being the best achievedo ver 100 cycles for [(tea) 2 Cu mnt ].
Analyzing the negolytea nd posolyte solutions separately,b y independents ingle redox couple "0 V" battery cycling, revealed that in each case one oxidation state wasf ar more robust for long-term cycling. This was especiallyi nsightful for [(tea) 2 Ni mnt ] , which couldn ot be charged/discharged in as ym-metricM eCN RFB;t he negolyte suffers immediate capacity fade, whereas the posolyte is robust for 100 charge-discharge cycles. Monitoring of the UV/Vis spectra of freshly charged negolyte and posolyte solutions of each complex over time agreedw ith the observations of posolyte/negolyte stability from the single redox couple RFB experiments, and indicated that unstabler edox states were self-discharging to the starting dianionic state. The self-discharging mechanism results in a cell imbalance,i nherently limiting the capacity of the symmetric RFBs. Future work will target derivatized dithiolene ligands to increase the stability of the oxidized and reduced states of the complexes.
Although the dithiolene complexes studied here are not suitable as materials for single-speciess ymmetric RFBs, longerterm stabilityw as observed for [(tea) 2 V mnt ] 3À/2À negolyte and [(tea) 2 Co mnt ] 2À/1À and [(tea) 2 Ni mnt ] 2À/1À posolyte solutions and are therefore applicable as single electrolytes in asymmetric RFBs. The development of ion-selective membranes that are suitable for use with nonaqueous electrolytes is at presenta limitation in the field of nonaqueous RFBs and as such hinders the use of the studied dithiolene metal complexes in asymmetric battery designs.

Voltammetry
Voltammetry experiments were performed in anaerobic 1mm solutions of each complex, with 0.1 m TBAPF 6 supporting electrolyte, in HPLC grade acetonitrile. All voltammetry experiments were performed under an inert N 2 atmosphere;s olutions were fully purged before use and aN 2 headspace was maintained throughout the experiments. Cyclic voltammetry (CV) was performed by using as tandard 20 mL 3-electrode glass cell (BASi)c onsisting of ap latinum wire auxiliary electrode, Ag/AgPF 6 quasi-reference (Ag wire in a glass fritted tube of 0.1 m TBAPF 6 in MeCN), and ag lassy carbon (GC) disk working electrode (3.0 mm diameter,B ASi, Alvatek, UK). Redox-couple reversibility and diffusionc oefficients, calculated by Randles-Sevcik analysis, were assessed by variation of the scan rate. Rotating disk electrode (RDE) studies were performed by using a6 0mLR RDE-3A apparatus (ALS Co.,L td) with a5mm diameter GC working electrode at rotation rates in the range 300-3000 rpm. Electrochemical rate constants were derived from the RDE data by Koutecký-Levich analysis. Working electrodes were polished before use with two grades of diamond slurries (3 mm and 0.25 mm, Buehler) and alumina suspension (0.05 mm, Buehler) prior to sonication in deionized water,a cetone rinsing, and air drying. Redox potentials were reported against the ferrocene/ferrocenium ion redox couple as an internal standard, except for variable scan rate studies, which are reported against the Ag + /Ag quasi reference (see the Supporting Information). Measurements were recorded by using aP C-controlled Emstat (PalmSens) with a resolution of 1mV.

Flow batterycharge-dischargee xperiments
Galvanostatic battery experiments were performed by using ac onventional zero-gap flow-cell manufactured in house;t he "Gen 2 flow-cell" was reproduced from ar eported method [47,48] (see the Supporting Information for further details). Experiments were conducted by using af low-through flow field (FTFF), 1mmc arbon paper electrodes (Technical Fibre Products Ltd.,p olyvinyl alcohol binder,2 .08 cm 2 active area) and either aC elgard membrane (Celgard 2500 Microporous Membrane, 25 mmt hickness) or F-930 cation exchange membrane (fumapem F-930, FuMA-Tech GmbH, 30 mmt hickness). Battery experiments were conducted with 10 mL half-cell solutions (20 mL total volume) of 1mm redox material in 0.1 m TBAPF 6 (TCI chemicals) in either MeCN (99.9 %, extra dry,o ver molecular sieves, AcroSeal,A CROS Organics)o rp ropylene carbonate (99.5 %, anhydrous, AcroSeal,A CROS Organics)a ta flow-rate of 10 mL min À1 by use of aM asterflex L/S peristaltic pump (Cole-Parmer). Experiments were conducted within aN 2 glovebox (Saffron Scientific Ltd. or MBRAUN), which was maintained with oxygen and water levels at am aximum of 1ppm. Charge cycles were performed at constant current density until the defined upper and lower potential thresholds were reached. The same magnitude of current was used upon both the charge and discharge. Charge-discharge cycling was controlled by either an Autolab (Metrohm AG) or Compactstat (Ivium Technologies) potentiostat. For the single redox couple "0 V" experiments, the following example procedure was performed;t oe xamine the negolyte, a symmetric flow cell with M mnt 2À initial starting electrolyte in each half-cell was charged at constant current to access the M mnt 3À as the negolyte and M mnt À as the posolyte. The posolyte was then replaced with fresh M mnt 2À electrolyte, before battery cycling at constant current between upper/lower potential thresholds just above/below 0V .T oe xamine the posolyte, the M mnt 3À negolyte is instead replaced with fresh M mnt 2À electrolyte after the initial electrolysis.
UV/Vis spectroscopy UV/Vis spectra were recorded with an Agilent Cary 60 spectrophotometer.S pectra of the as-synthesized materials were recorded for 50 mm solutions in MeCN. Spectra of charged electrolytes were recorded from solutions prepared as follows:i naN 2 -filled glovebox, an initial charge cycle was first performed on as olution of 1mm complex in 0.1 m TBAPF 6 MeCN solution in each half-cell. Next, 50 mm solutions of each charged electrolyte were prepared by extracting 250 mLf rom the battery electrolytes and diluting to 5mL (5 mm TBAPF 6 supporting salt concentration). The solutions were transferred to sealed quartz cuvettes (Starna Scientific, 1cmp ath length), removed from the glovebox, and the UV/Vis spectra were recorded immediately.T he time of t = 0p resented in the results represents approximately 5min after the initial charge cycle of the flow cell was completed.