Redox Mechanisms in Li and Mg Batteries Containing Poly(phenanthrene quinone)/Graphene Cathodes using Operando ATR‐IR Spectroscopy

Abstract The redox reaction mechanism of a poly(phenanthrene quinone)/graphene composite (PFQ/rGO) was investigated using operando attenuated total reflection infrared (ATR‐IR) spectroscopy during cycling of Li and Mg batteries. The reference phenanthrene quinone and the Li and Mg salts of the hydroquinone monomers were synthesized and their IR spectra were measured. Additionally, IR spectra were calculated using DFT. A comparison of all three spectra allowed us to accurately assign the C=O and C−O− vibration bands and confirm the redox mechanism of the quinone/Li salt of hydroquinone, with radical anion formation as the intermediate product. PFQ/rGO also showed exceptional performance in an Mg battery: A potential of 1.8 V versus Mg/Mg2+, maximum capacity of 186 mAh g−1 (335 Wh kg−1 of cathode material), and high capacity retention with only 8 % drop/100 cycles. Operando ATR‐IR spectroscopy was performed in a Mg/organic system, revealing an analogous redox mechanism to a Li/organic cell.


Introduction
From wireless remotes,m obile phones to wearable electronics, battery-powered personal devices have transformed the way we live. The developmento fl ow-cost, safer,m ore powerful, and more energy-dense battery technology will unlocki nnovation within the transporti ndustry and facilitateatransitiont o renewables by providing flexibility in the production and consumption of energy.M ost state-of-the-artL i-ion batteries employ inorganic cathodes. The cathode materials vary depending on the requirementso ft he application, and include lithium cobalt oxide,l ithium nickel manganese cobalto xide, lithium manganese oxide,l ithium iron phosphate, and nickel cobalt aluminumo xide. [1,2] Most of these materials contain toxic or relativelyr are elements (Ni, Co) and al ot of energy and high temperatures above 700 8Ca re required during ceramic synthesis. [3] Their theoretical capacities are limited to ap-proximately 200 mAh g À1 owing to the relativelyh igh atomic weighto ft ransitionm etals and the low number of exchanged electrons.C urrently,t he best material in terms of energy density is NMC 811, whichh as ac apacity and voltage of 200 mAh g À1 and 3.8 Vversus (vs.) Li/Li + ,respectively,which results in ag ravimetrice nergy density of approximately 760 Wh kg À1 of cathode material. [4] Mg batteries are becoming an increasingly attractive alternative to Li-based batteries. There are severalr easonsf or that, including lowerc ost, sustainability, an improved safety profile, and higher volumetric energy density.M oreover,l ithiumd eposits are located at some geopolitically sensitive locations and its extraction can have ab ig impact on the environment. On the other hand, Mg is 1000 times more abundant [5] and can be economically produced even from seawater.A sM gd oes not form dendrites, [6] it can be safely used as am etal anode with av ery high volumetric capacity 3832 Ah L À1 (Lih as ac apacity of 2062 Ah L À1 ). However,t here are also disadvantages such as passivation of the Mg anode,l ack of suitable electrolytes, [7] difficult insertion of the Mg 2 + ion into inorganic materials, and slow solid-state diffusion, which resultsi nt he lack of suitable cathodes and severely impedes the development of practical Mg batteries. [8] On the other hand, organic materials offer ab etter alternative to inorganic materials in terms of versatility and compatibility with different metal counter ions, [9,10] price, gravimetric energy density,a nd sustainability. [11][12][13][14][15] They can be produced from petrochemicals, biomaterials, organic waste, [3,13,[16][17][18] or even from CO 2 as as ource of carbon [19] under low-temperature synthesis conditions below 100 8C. Because of their low molecular mass and high number of exchanged electrons, they can reach high practical capacities of up to 600 mAh g À1 (Li-rhodiz- The redox reactionm echanism of ap oly(phenanthrene quinone)/graphene composite (PFQ/rGO) was investigated using operando attenuated total reflection infrared (ATR-IR) spectroscopy during cycling of Li and Mg batteries. The reference phenanthrene quinonea nd the Li and Mg salts of the hydroquinone monomers were synthesized and their IR spectra were measured. Additionally,I Rs pectra were calculated using DFT.A comparison of all three spectra allowed us to accuratelya ssign the C=Oa nd CÀO À vibration bands andc onfirm the redox mechanism of the quinone/Li salt of hydroquinone,w ith radical anion formation as the intermediate product. PFQ/rGOa lso showede xceptional performance in an Mg battery:Apotential of 1.8 Vv ersus Mg/Mg 2 + ,m aximum capacity of 186 mAh g À1 (335 Wh kg À1 of cathode material), and high capacity retention with only 8% drop/100 cycles. Operando ATR-IR spectroscopy was performed in aM g/organic system,r evealing an analogous redox mechanism to aL i/organic cell. onate), which could theoretically be increased up to 960 mAh g À1 (value for the C=Oc arbonyl group). This would result in at heoretical energy density of approximately 2000 Wh kg À1 of cathode for 2.0V vs. Li/Li + organic materials. The most commond rawback of organic materials is their good solubility in organic electrolytes, which results in afast capacity drop during cycling and self-discharge. [12,17,20] The mostpromising organic materials are actually redox-active polymers with very low solubility in electrolytes. Some promising examples are poly(imides), [21] poly(anthraquinonyl sulfide) (PAQS), [22] poly(2,5-dihydroxy-1,4-benzoquinonyl sulfide), [23,24] poly(vinylanthraquinone), [25] polymer-boundp yrene-4,5,9,10-tetraone, [26] poly(anthraquinone) (PAQ), [27] poly(benzenequinonyl sulfide), [28,29] and poly(diphenanthrene-quinone substitutedn orbornene). [30] Cathode or positive electrode materials based on anthraquinone, such as poly(anthraquinonyl sulfide) (PAQS) [22] and poly(anthraquinone) (PAQ) [27] have very stable electrochemistry,h igh Coulombic efficiency,g ood power performance (rate capability), and have been successfully exploited in both Li-and post Li-battery systems. [31][32][33] On the otherh and, their discharge potential is only approximately 2.2 Vv s. Li/Li + for discharge and correspondingly lower in aM gc ell.R ecently,w e developed an isomer of PAQ, poly(phenanthrenequinone) (PFQ), with ap otentialo f2 .6 Vv s. Li/Li + . [34] Unexpectedly,t his PFQ materialw as insoluble in organic solvents, which resulted in high cycling stabilityw ith only 9% drop in capacity after 500 cycles (50 mA g À1 or approximately 0.2 C). However,i nsolubility resulted in low material utilization.T herefore, ap orous composite material containing 21 wt %o fr educed graphene oxide (PFQ/rGO) was synthesized,w hich had better capacity utilization.R ecently,o ur [29,31,35] and other research groups [32,[36][37][38] have shown that redox-activeo rganic materials can be successfully used as Mg cathode materials. The advantage of organic materials in Mg battery systems can be attributed to their relativelyf lexible structure and swelling of organic materials inside the electrolyte, which allow better accessibility to the Mg ions. Additionally organic materials have lower redox-potentials that fit well inside the Mg electrolyte stability window.
Herein,w ec ompared the redoxr eactionm echanism of PFQ/ rGO in Li-and Mg-metal organic batteries. The redox mechanism in the Li battery was probed by collecting operando ATR-IR spectra during charge/discharge of the Li-PFQ/rGO battery. Then, DFT calculations and IR spectra of the chemically synthesized monomer molecules wereu sed to accuratelya ssign bands from the operando measurements and assist their interpretation. PFQ/rGOw as also used as an Mg battery cathode and its mechanism wasa nalyzed with the help of ATR-IR and DFT.

Results and Discussion
Li-PFQ/rGO battery Galvanostatic curveso ft he Li-PFQ/rGO battery are displayed in Figure 1a.P FQ/rGOh ad two chargep lateaus at 2.47 and 2.93 Vand two discharge plateaus at 2.78 and 2.26 Vv s. Li/Li + , which were more clearly visualized by obtaining ad Q/dE plot ( Figure S1). Twop lateaus and ar eversible capacity well above 130 mAh g À1 (130 mAh g À1 is the theoretical capacity for ao neelectron reaction) suggested that atwo-step redox reactionoccurred.I ne ach step, one electron should be exchanged. To confirm this hypothesis, we performed operando ATR-IR measurements during cycling of the Li-PFQ/rGO battery.F or this measurement,w eu sed as pecial spectro-electrochemical pouch cell with aS iw afer window,w hich was in contact with Ge ATRc rystal.F or this operando ATR-IR setup, the penetration depth was estimated to be 0.4-1.20 mmi nt he measurement region 1800-600 cm À1 . [39] The estimated thickness of our electrodes was approximately 150 mm, which means that only a small part of the electrode was probed. In the case of the insoluble PFQ polymer, in which the cathode active material does not dissolve, this is not ap roblem.Astrong IR signal from the Si window and electrolyte was subtracted according to the literature. [39][40][41] Subtraction visualizes only IR-active changes inside the cathode composite during cycling (Figure 1b). For the background, we used IR spectra of the dischargedc athode. An egative band in the operando spectra means ad ecrease, and ap ositive band means an increasei n the concentrationoft he corresponding species compared with the discharged state. During charging, the intensity of the carbonyl band C=Oa t1 678cm À1 increased, whereas the corresponding band of the CÀO À group located at 1377 cm À1 decreased. During discharging, the process was reversed, indicating ar eversible electrochemical redox reaction. These results indicated that the redox mechanism was reversible oxidation of the Li hydroquinone salt into quinone. These resultsw ere very analogous to operando ATR-IR measurement of ac omparable compound PAQS (C=Os tretching at 1670 cm À1 ,a nd 1650 cm À1 and CÀO À stretching at 1370 cm À1 ). [40] Besides these two bands, we observed many other positive/negative bands, which are analyzed in detail in the followings ection.
The assignment of the IR bands in the operando ATR-IR spectra was achieved by comparison with the theoretical spectra calculated by DFT and the IR spectrao ft he chemically synthesized monomer molecules. First, the accuracy of the DFT calculations was verified by calculating the FTIR spectra of the chemically synthesized monomer molecules 2,7-dibromo-9,10phenanthrenequinone( FQ), dilithium salt of 2,7-dibromo-9,10dihydroxy-phenanthrene (LiFQ), and the magnesium salt of 2,7-dibromo-9,10-dihydroxyphenanthrene (MgFQ). These calculated spectra were comparedw ith measured ones (Figure S2 a-c). The spectra matched surprisingly well, especially for the most intense bands such as the carbonyl C=Os tretching, C=Ca nd CÀO À stretching vibrations, ring vibrations and CÀHo ut-of-plane deformations. However,s ome bands were shifted.T he biggest mismatch was observed for the carbonyl band, for which the calculated frequency was 65 cm À1 higher than the measured value (FigureS2a). Inaccurate intensities of the bands were also calculated by DFT,e specially at lower frequencies;h igheri ntensities were observed by operando ATR measurements at lower frequencies.B ecauseo ft hese two reasons, we decided to zoom in on the calculated DFT spectra 3-6 times in the region below 1300 cm À1 to obtain more comparable resultswith operando ATRm easurements. Notably, the calculations revealed the expected complex nature of the bands that belong to the ring vibrations. Those bands were the combination of CÀC, C=C, CÀH, and CÀO À or C=Om odes that were strongly coupled. The appearance of such coupled vibrations can be an additional source of the observed differences between the calculated and measured spectra.
In the next step, DFT calculations were applied for LiFQ 3 and FQ 3 oligomers (Figure 2a), whichc onsist of three monomer units. Using larger oligomers as models would give us more accurate predictionsb ut would significantly extend the time required to perform the calculations. The subtracted DFT spectra werec alculated using the equation S DFT,L i = FQ 3 ÀLiFQ 3 .F or comparison, we also calculated the subtracted spectra for the chemically synthesized reference monomers S chem, Li = FQÀLiFQ (Figure 2b), whichw erev ery similart ot he operando ATR-IR spectra ( Figure 2c). For instance,t he carbonyl C=Os tretching band [42,43] at 1678 cm À1 from operando ATR-IR measurement was very close to the band observed for the chemically synthesized monomer at 1675 cm À1 .T he CÀO À stretching vibration at 1377 cm À1 was close to the predicted value of 1348 cm À1 and the measured value of 1383 cm À1 or the chemically synthe-  Table S1.
From the charge/dischargeg alvanostatic curves, two plateaus were easily distinguished, which was attributed to at woelectron reactioni nt wo separateo ne-electrons teps. Using the operando ATR-IR technique (Figure 3a), we detectedt he intermediate radicala nion LiPFQ* (Figure 3b). Twob ands at 1549 and 1484 cm À1 startedt oa ppear towards the middle of the charge/discharge cycle and were in good agreement with the bands at 1562 and 1487 cm À1 calculatedb yD FT (Figure 3c and 3d); they corresponded to the stretching vibration of the CPOa nd CPCb onds. The position of those bands was in good agreement with the data in the literature. [44] Other bands monotonically increased during charging and corresponded to CÀO À (1375 cm À1 ), ring vibrations (1254 cm À1 and 1048cm À1 ), and CÀHo ut-of-plane vibrations (808 cm À1 ).

Mg-PFQ/rGO battery
The electrochemical performance of the PFQ/rGO composite material in an Mg battery was evaluated using 0.6 m Mg(TFSI) 2 -2MgCl 2 in dimethoxyethane (DME)e lectrolyte at 0.5 Cc urrent density.T he initial capacity was only 65 mAh g À1 ,b ut it quickly increased to 130 mAh g À1 after only ten cycles (Figure 4a). The subsequenti ncrease to 186 mAh g À1 (72 %o ft he theoretical capacity C theo = 260 mAh g À1 )w as very gradual. Between 140-400 cycles,t he capacitys lowly decreased to 146 mAh g À1 (8 % drop/100 cycles). To the best of our knowledge,t his is one of the best results for an Mg battery in terms of dischargec apacity andc ycle stability.T he Coulombic efficiency in the first 14 cycles was above 100 %o wing to large polarizationi nt he first cycles, which prematurely finishes the discharge. Consequently,s ome of the initial capacity (pristine material was in a chargeds tate), was retained and the discharge capacity was larger than the chargec apacity in the previous charge. The efficiency with cycling dropped to 94 %a fter 250 cycles and later increased back to 96 %u ntil the 400 th cycle. After 400 cycles, sporadicC oulombic efficiency drops were observed and the capacityd rop was more prominent. The most likely reason for this was degradationo ft he Mg anode. As shown in the galvanostatic curves, the polarization decreased until 100-200cycles (Figure 4b). At the 100 th cycle, two discharge plateausa t2 .19 and 1.71 Vv s. Mg/Mg 2 + ,a nd two charge plateausa t2 .25 and 2.55 Vv s. Mg/Mg 2 + were observed and more clearly visualized with ad Q/dE plot ( Figure S3). Compared with the results of the Li system,t he equilibrium potential was only approximately 420 mV lower in the Mg system and the expected value was approximately 700 mV (E (Li/Li +) ÀE (Mg/Mg2 +) = 3.040 VÀ2.372 V = 668 mV). Ah igher potential in the Mg battery system could be explained by an energetically more favorable coordination of the Mg 2 + ion with two CÀO À in the ortho position compared with two Li + ions. [35,45] We also tested the C-rate performance of the Mg-PFQ/rGO battery from 0.5 Ct o5 0Crate (Figure 4c). Owing to as low activation, the capacity was stabilized after 32 cycles (at 2C). However,w ec ould stillc ompare the capacities relative to 0.5 Ca tt he end of the cycling (183 mAh g À1 , 100 cycles). At ac urrent density of 5C,t he capacities decreasedt oa pproximately one thirda nd at > 20 Cn oc apacity was obtained. The Coulombic efficiency during the rate capability tests decreases with higher C-rate from approximately 99 %t o8 3%,w hich is rather unusual-in most cases the efficiency improves at high rates because there is less time for side reactions. In our case, there may be ap roblem with prematurec ut-off at high rates, at whichp olarizationb ecomes significant.M ore detailed information about the C-rate experiments was obtained from galvanostatic curves (Figure 4d). The polarization increased with higherC -rate,w hich was expected. The capacity drop was mainly attributed to the chargec ut-off potential-the charging plateau at 5C and higher rates was finished too early at 2.8 Vv s. Mg/Mg 2 + .T herefore, the capacity at high rates could be improved by setting the cut-off voltage to higherv alues, but this is limited by the electrolyte stability window,w hich is approximately 2.8 Vv s. Mg/Mg 2 + for highsurface-area electrodes. The low capacity in the early cycles can also be explained in the same manner:T he material could not be fully charged in the first cycles until the polarization gradually drops to allow full charging of the material.
The operando ATR-IR spectra during cycling of the Mg-PFQ/ rGO battery were obtained in as imilar manner as for the Li battery system. During charging,t he carbonyl C=Ob and at 1680 cm À1 increased and the CÀO À band at 1392 cm À1 decreased, although both werem uch smaller comparedw ith those in the Li system owing to al ower capacity utilization (Figure 5a). However,w ep roposed ar edoxr eaction mechanism in which the Mg hydroquinone salt MgPFQ was reversibly oxidizedi nto quinone PFQ via ar adicala nion MgPFQ* (Figure 5b).
We used the same strategy as for the Li system to assign the bands (Figure 6a and 6b). The characteristic bands for C=O, C=C, CÀO À ,a nd CÀHo ut-of-ring vibrationsw ere at similar positions as in the Li system (Table S1). On the other hand, the ring vibrations from 1350 to 1000 cm À1 in the Mg system were quite different compared with the Li-system.T he same pattern was observedi nt his region in the spectra obtained by DFT calculations and for the chemically synthesized monomers: two bands (1350cm À1 and 1330 cm À1 ), followed by three bands (1183 cm À1 ,1 135 cm À1 and 1069 cm À1 ), and an egative band at 1050 cm À1 .
Twop lateaus were also observed in the Mg system during charging/discharging, suggesting that at wo-electron redox reaction has two separatep rocesses with an intermediate relatively stable radical anion (Figure 5b). Operando ATR-IR spectroscopyw as used to confirm this hypothesis (Figure 5a). Again, two characteristic bands at 1544a nd 1485 cm À1 were observed, corresponding to CPOa nd CPCv ibrations of an anion radical. However, al arge difference between the measured and calculatedv alues (1648 cm À1 and 1587 cm À1 )w as observed (Figure 7). One of the reasons could be because we were not able to use 1/2 Mg 2 + ion per monomer unit in the DFT calculations;. the MgCl + ion was used instead. The use of MgCl + was justifiedb yt he fact that in Mg battery systems with bis(trifluoromethane)sulfonimide magnesium salt or Mg(TFSI) 2 and MgCl 2 -based electrolytes, the cathode material is exchanging both Mg 2 + and MgCl + . [39,46] All other bands were    also present in the charging spectra (Figure6c, Ta ble S1) and they were slightly shifted.

Conclusions
PFQ/rGOa ctive material is an ext generation cathode material with extraordinaryc ycling stabilityi nb oth Li-and Mg-based organic batteries. In aM gs ystem, the material displays high capacityu tilization of 186 mAh g À1 and as mall capacity fade 0.08 %p er cyclew ith an average dischargep otentialo f1 .8 V vs. Mg/Mg 2 + .T he rate capability of Mg-PFQ/rGOw as moderate, which was mainly attributed to relatively high polarization of the Mg metal anode.S tarting with the Li electrolyte, the electrochemical mechanism of PFQ/rGOw as investigated through operando ATR-IR. Operando ATR-IR spectra were complemented with the synthesis of model compounds and DFT calculations, which confirmed the quinone/hydroquinone salt electrochemical mechanism.T he formation of ar adical anion was identified as an intermediate step through the appearance of new bands at 1549 and 1484 cm À1 , which were the most intense at the middle of the half cycle and assigned with the help of DFT calculations. Interestingly,b ands of the radical anion were also observed in the Mg system,w hich might indicate ac omplex electrochemical mechanism with the possibility of MgCl + or other complex monovalenti ons serving as counterions. Another option is that Mg 2 + ions interact with two neighboring PFQr adicali ons. Nevertheless, the PFQ/rGOc athode showedp romisingp erformance as ac athode in an Mg battery,w hich opens ap ath towards its application in other multivalent battery systems (e.g.,A l, Zn, Ca). Its high cycling stabilitym ight lead to practical multivalent organic batteries.

ATR-IR spectra
Infrared spectra of the reference monomers FQ, LiFQ, and MgFQ were measured inside ag lovebox under an inert Argon atmosphere on an IR spectrophotometer Bruker Alpha using Germanium ATRc rystal. The measuring range was 4000-600 cm À1 .3 2c onsecutive scans were measured with aresolution 4cm À1 .

Operando ATR-IR spectroscopy
The ATR-IR measurements were performed on aB ruker Vertex 80 equipped with aS pecac Silver Gate Ge crystal ATRa nd al iquid-nitrogen-cooled mercury cadmium telluride (MCT) detector.T he spectra were collected in absorbance mode with 64 scans at ar esolution of 4cm À1 in the range from 4000 cm À1 to 600 cm À1 .T he IR spectra were collected in operando mode with as eries of repetitive scans every 2min during galvanostatic cycling of the batteries with ac urrent density of 0.2 Co rC /5 rate for the Li battery system and 0.5 Cf or the Mg battery system. The battery was assembled in as pectro-electrochemical cell with aS iw afer window. [39,40] The at-ChemSusChem 2020, 13,2328 -2336 www.chemsuschem.org 2020 The Authors. Publishedb yWiley-VCH Verlag GmbH &Co. KGaA, Weinheim mospheric compensation was performed on the IR spectra in OPUS version 7.8 software. The ATRd ifference spectra were obtained by subtracting the discharge spectrum from the obtained IR spectrum at aspecific point of the discharge/charge.

DFT calculations
DFT calculations were performed using the Gaussian 09 software package. [49] The M06-2X hybrid functional [50] and 6-31 + G(d,p) basis sets were used to do geometry optimization and calculation of the vibrational frequencies and IR intensities. To model the polymer PFQ and its reduced forms in the Li and Mg battery system, a trimer consisting of three monomer units that end with am ethyl group was used (FQ 3 ,L iFQ 3 ,L iFQ 3 *, MgFQ 3 ,a nd MgFQ 3 *) in all cases ( Figure S4). The models ensured suitable accuracy of the calculation while keeping the computational cost reasonable. Vibrational modes of methyl groups that are included in the model owing to the finite size of the trimer,a re not included in the analysis and plotted spectra. To account for the effects of the surrounding environment, ad ielectric constant of 7.4 was used, which is a common value for ether-based electrolytes. Discrepancies between the theoretical and experimental spectra arise owing to approximations of the theoretical approach, such as the treatment of the electronic Hamiltonian, finite basis set, and the harmonic approximation. This leads to overestimation of the calculated frequencies by approximately 5%.T herefore, we applied as caling factor of 0.95, which allows easier visual comparison between the theoretical and experimental values.

NMR spectra
NMR spectra were collected using a3 00 MHz Varian Unity Inova. Deuterated solvents CDCl 3 or [D 6 ]DMSO were used and tetramethyl silane was used as astandard.

Assembly of Li batteries
Electrodes were prepared by mixing 60 mg of composite PFQ/rGO, 30 mg of carbon black (Printex XE2), and 10 mg of polytetrafluoroethylene (PTFE) (60 wt %w ater dispersion, Aldrich) and 0.5 mL of isopropyl alcohol (IPA). All these ingredients were ball milled in 12 mL stainless steel grinding jars (10 mm Øb alls) with ap lanetary ball mill (Retsch PM100) at 300 rpm for 30 min in an air atmosphere. The obtained slurry was kneaded with am ortar and pestle to obtain ac ompact black gum. The gum was rolled between two pieces of an onadhesive paper with ar oller to obtain an electrode film of approximately 5 5cms ize. An aluminum mesh (100 mesh size) was deposited on top of this film and the film was rolled again to glue the electrode composite and the mesh together,a nd then dried in an air atmosphere. Afterward, electrode discs with a diameter of 1.2 cm were cut and pressed with al oad of 1t on and further dried at 80 8Ci nv acuum for 1day.T he average loading on the electrode was 2.5 AE 0.4 mg of active material per cm 2 .B attery cells were assembled in an argon-filled glovebox (water and oxygen levels < 1ppm). Swagelok-type battery cells were assembled using the above-mentioned electrodes, a1 3mmg lass fiber separator (Whatman GF/A), and freshly rolled lithium (12 mm diameter,A ldrich). 1 m Bis(trifluoromethane)sulfonimide lithium salt (Aldrich) in am ixture of dry 1,3-dioxolane/dimethoxyethane was used an electrolyte (1 m LiTFSI/DOL + DME).

Assemblyo fM gb atteries
An electrode composite gum for Mg tests was prepared in the same way as that for Li. Then, the gum was rolled in between two sheets of nonadhesive paper and afterwards electrode discs with 1.2 cm diameter were cut out to give self-standing electrodes. The obtained electrodes were subsequently dried at 80 8Ci nv acuum for 1day and transferred into an argon-filled glove box. An average loading on electrode was 2.5 AE 0.4 mg of active material per cm 2 .M gc ells were assembled in Swagelok type battery cells using graphite disc as cathode current collector,1 3mmg lass fiber separator (Whatman GF/A) and freshly brushed Mg foil (Gallium Sources, 99.95 %, 0.05 mm). 0.6 m Mg(TFSI) 2 -2MgCl 2 in DME was used as electrolyte.

Electrochemical measurements
Ap otentiostat/galvanostat VMP3 (Bio-Logic, France) was used at room temperature (25 8C) to perform the electrochemical measurements. Batteries were cycled between 1.5-3.5 Vv s. Li/Li + at ac urrent density of approximately 0.2 Co rC /5. Mg batteries were cycled from 0.8-2.8 Vv s. Mg/Mg 2 + at different C-rates from 0.5 to 50 C.