Effect of the Electrolyte Solvent on Redox Processes in Mg–S Batteries

Jun 14, 2019 - Mg–S batteries are attractive for next-generation energy storage because of their high theoretical capacity and low cost. The foremos...
0 downloads 0 Views 3MB Size
Forum Article Cite This: Inorg. Chem. XXXX, XXX, XXX−XXX

pubs.acs.org/IC

Effect of the Electrolyte Solvent on Redox Processes in Mg−S Batteries Sarah C. Bevilacqua,† Kim H. Pham,‡ and Kimberly A. See*,† †

Division of Chemistry and Chemical Engineering, California Institute of Technology, Pasadena, California 91125, United States Department of Chemistry, Brown University, Providence, Rhode Island 02912, United States



Downloaded via UNIV OF SOUTHERN INDIANA on July 17, 2019 at 04:56:14 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

S Supporting Information *

ABSTRACT: Mg−S batteries are attractive for next-generation energy storage because of their high theoretical capacity and low cost. The foremost challenge in Mg−S batteries is designing electrolytes that support reversible electrochemistry at both electrodes. Here, we target a solution-mediated reduction pathway for the S8 cathode by tailoring the electrolyte solvent. Varying the solvent in Mg-based systems is complicated because of the active nature of the solvent in solvating Mg2+ and the complex dynamics of electrolyte−Mg interfaces. To understand the effect of the solvent on the S8 reduction processes in the Mg−S cell, the magnesium− aluminum chloride complex (MACC) electrolyte was prepared in different ethereal solvents. Reversible Mg electrodeposition is demonstrated in the MACC electrolyte in several solvent systems. The electrodeposition overpotentials and current densities are found to vary with the solvent, suggesting that the solvent plays a noninnocent role in the electrochemical processes at the Mg interface. Mg−S cells are prepared with the electrolytes to understand how the solvent affects the reduction of S8. A reductive wave is present in all linear-sweep voltammograms, and the peak potential varies with the solvent. The peak potential is approximately 0.8 V versus Mg/Mg2+, lower than the expected reduction potential of 1.7 V. We rule out passivation of the Mg anode as the cause for the low voltage peak potential, making processes at the S8 cathode the likely culprit. The ability to oxidize MgS with the MACC electrolyte is also examined, and we find that the oxidation current can be attributed to side reactions at the C−electrolyte interface.



INTRODUCTION The demand for less expensive and lighter energy storage compared to the ubiquitous Li-ion battery motivates research on new battery chemistries. Conversion reactions that utilize the available charge states of each atom in the active material promise theoretical capacities 1 order of magnitude greater than conventional mechanisms based on intercalation chemistry. For example, electrodeposition of Li metal offers theoretical capacities of 2061 mAh mL−1 and 3861 mAh g−1, while intercalation anodes like Li graphite are limited to 800 mAh mL−1 and 372 mAh g−1. Li-metal electrodes are arguably the holy grail anode, but safety issues caused by the formation of Li dendrites upon cycling prevents commercialization. Here, we target a Mg−S battery that utilizes Mg electrodepostion and stripping at the anode and reversible S8 reduction to MgS at the cathode. Mg−S batteries promise high energy densities of >1700 Wh kg−1 and >3200 Wh L−1,1 with a theoretical voltage of 1.77 V.2 Mg is a promising alternative to Li because Mg electrodeposits in a smoother morphology compared to Li at the same current densities,3 making Mg-metal anodes more attainable. The smooth morphology may be due to a difference in the self-diffusion coefficient of Mg compared to Li. The self© XXXX American Chemical Society

diffusion barrier in Mg is much lower compared to that of Li, allowing for structural rearrangement on the surface of the metal during electrodeposition.4 Smooth Mg electrodeposits have been reported and observed since 2001 by Aurbach and co-workers,5,6 and experimental results comparing Li and Mg electrodepositions head-to-head at the same current densities directly support the theoretical prediction.3 The role of high current densities imposed by the electrode geometry on the electrodeposit morphology is currently under investigation,7 highlighting the importance of determining Mg electrodeposition limitations prior to commercialization. Conversion of S8 to S2− is a promising cathode reaction to pair with the Mg-metal anode. S8 is a low-cost, highly abundant material that promises a high theoretical capacity of 1672 mAh g−1, assuming complete reduction to S2−.8 Intercalation of Mg2+ in a solid-state host lattice is another promising cathode chemistry; however, intercalation of the divalent cation is Special Issue: Celebrating the Year of the Periodic Table: Emerging Investigators in Inorganic Chemistry Received: March 27, 2019

A

DOI: 10.1021/acs.inorgchem.9b00891 Inorg. Chem. XXXX, XXX, XXX−XXX

Forum Article

Inorganic Chemistry difficult and has only been demonstrated in a few systems such as the canonical Mo6S8 Chevrel phase9 and, more recently, TiS2.10 Examples of Mg2+ intercalation in oxides are rare.11 Research on Mg−S, however, represents an underexplored opportunity to reach high theoretical capacities using low-cost, Earth-abundant materials via redox mechanisms that are less dependent on divalent ion diffusion in the solid state. Significant challenges must be overcome before Mg−S batteries can be realized (Figure 1). The foremost challenge

Mg 2 + +

1 S8 + 2e− → MgS(s) 8

(1)

In a solvent-mediated mechanism, solubilized magnesium polysulfides (MgxSy) are formed as intermediates and subsequently reduced to solid-state discharge products like MgS: y x Mg 2 + + S8 + 2x e− → MgxSy (soln) (2) 8 (y − x)Mg 2 + + MgxSy

(soln)

+ 2(y − x)e− → y MgS(s) (3)

Solvent-mediated mechanisms are likely the only route to achieving high active material utilization because the diffusion of Mg2+ in MgS and S8 is very poor. Therefore, we expect that the best Mg−S electrochemistry will be achieved in solvent systems that support dissolution of MgxSy species. Here, we aim to systematically study the effect of the solvent on the Mg−S chemistry by developing a suite of electrolytes with the same supporting electrolyte in a variety of solvents. A promising example of a non-nucleophilic Mg electrolyte is the magnesium−aluminum chloride complex (MACC).15,16 The MACC electrolyte, first reported by Doe et al., consists of MgCl2 and AlCl3 and supports Mg electrodeposition and stripping at >99% Coulombic efficiency.15 Doe et al. reported Mg electrodeposition and stripping in both tetrahydrofuran (THF) and dimethoxyethane (DME) in their original paper,15 but the MACC electrolyte is most commonly studied in THF. In THF, the major complexes in solution are [Mg2(μ-Cl)3· 6THF]+ and AlCl4−, as confirmed by single-crystal X-ray diffraction16 and extensive solution-phase characterization.17 Because of active coordination of the solvent in the Mg complex, the speciation of Mg2+ in the MACC electrolyte likely changes with solvent. A different Mg complex [Mg2(μ-Cl)2]2+ was identified in DME, for example.18 The as-prepared MACC electrolyte in THF does not support reversible Mg electrodeposition and stripping and instead initially shows electrochemistry related to Al deposition.17,19 Upon continued cyclic voltammetry (CV) cycles, the shape of the voltammograms evolves, and stable Mg electrodeposition and stripping is eventually achieved at nearly 100% Coulombic efficiency.19 The electrolytic process used to activate the electrolyte is termed “conditioning”,19 and the mechanism involves the reduction of AlCl4− to Al(s), an increase in the [Mg2(μ-Cl)3· 6THF]+ concentration, and the formation of free Cl−.17 Interestingly, the addition of Mg powder to the MACC electrolyte was shown to increase the Coulombic efficiency and decrease the Mg deposition overpotential,20 which could be the chemical equivalent of the electrochemical conditioning process. Luo et al. suggest that the Mg powder removes deleterious species, improving the electrochemical performance of the electrolyte.20 The reactivity of S8 and the MACC electrolyte in THF was studied by Liu et al.16 An excess of S8 was added to the electrolyte, and no reactivity was observed, as indicated by 1H and 27Al NMR.16 While the complexes are suggested to be stable in contact with S8,16 there is no report of Mg−S electrochemistry in the MACC electrolyte. THF is an uncommon solvent for S8 electrochemistry, so the challenge remains to prepare MACC electrolytes in a variety of solvents and evaluate their behavior in a Mg−S battery. The solvent will not only affect the S8 electrochemistry, but it will also actively

Figure 1. Ideal Mg−S battery composed of a Mg-metal anode and a S8 cathode. The electrolyte governs both the ability to reversibly electrodeposit Mg at the anode and reversibly reduce S8 to MgS at the cathode. The challenges associated with the Mg−S system are highlighted.

that affects every aspect of the cell from bulk reactivity to interfacial processes is the development of a suitable electrolyte, which we focus on here. Few electrolytes are compatible with both Mg and S8, and none are ideal. Organomagnesium-based electrolytes, commonly used in Mg−Mo6S8 batteries, are likely too nucleophilic and will react with electrophilic S8. Efforts have therefore focused on the development of non-nucleophilic electrolytes, such as the Hauser base hexamethyldisilazane magnesium chloride (HMDSMgCl) salt used in the first report of the Mg−S system.12 Since then, various supporting electrolyte and solvent combinations have been pursued, but the mechanistic role of the electrolyte on the redox mechanisms remains unknown. The Mg−S system is plagued by reduction processes occurring at much lower voltages than the theoretical 1.77 V and the inability to oxidize the reduction product during charge. We know from the Li−S system that the electrolyte plays a pivotal role in directing the redox reactions at the S8 cathode because the reaction pathways go through a solvent-mediated mechanism.13,14 Although the redox mechanisms for Mg−S batteries are still debated, there are presumably two possible pathways: a solidstate conversion pathway and a solution-mediated pathway. In a solid-state conversion mechanism, Mg2+ and S8 form MgS directly: B

DOI: 10.1021/acs.inorgchem.9b00891 Inorg. Chem. XXXX, XXX, XXX−XXX

Forum Article

Inorganic Chemistry modify the speciation of Mg2+ in solution. We target etherbased solvent systems that show promising reversibility for Li− S systems and evaluate Mg electrodeposition and stripping processes, S8 reduction, and MgS oxidation.



coated Al produced the lowest background current and is thus used as the current collector henceforth. We prepared the S electrodes on Ccoated Al, and upon disassembly, we observed no corrosion on the Ccoated Al. The S8 electrodes were prepared in air. The electrode slurry was prepared by adding 50 wt % S8 (99.998%, Sigma-Aldrich), 40 wt % Super P conductive C (VWR), and 10 wt % poly(vinylidene fluoride) (PVDF, MTI Corp.) to 1-methyl-2-pyrrolidinone (NMP; SigmaAldrich). A typical slurry contained 0.200 g of S8, 0.160 g of Super P, 0.040 g of PVDF, and 4 mL of NMP. The slurry was mixed in a Thinky centrifugal mixer for 30 min. Control cathodes without S8 were prepared in the same manner with 60 wt % Super P and 40 wt % PVDF. All electrodes were cast with a micrometer adjustable film applicator (MTI Corp.) set to a thickness of 20 μm on C-coated Al (MTI Corp.). The casts were dried overnight at room temperature and then at 60 °C for 8 h. MgS electrodes were made in the same manner as the S8 electrodes but without exposure to air. The electrodes were prepared with 50 wt % MgS, 40 wt % Super P, and 10 wt % PVDF in the glovebox with cyclopentanone (Sigma-Aldrich) instead of NMP (casting with NMP resulted in white precipitates on the dried electrodes). MgS electrodes were mixed in the Thinky centrifugal mixer in a sealed container filled with Ar. The electrodes were cast with a micrometer adjustable film applicator (MTI Corp.) set to a thickness of 20 μm on C-coated Al (MTI Corp.) in the glovebox. The electrodes were dried in a vacuum oven inside the glovebox at 80 °C overnight. Two-Electrode Cell Preparation. Mg−S cells were evaluated in a two-electrode geometry. The electrochemical cell was designed to prevent contact between the electrolyte and metal components to reduce the influence of corrosion on the measured electrochemistry. Details and schematics of the cell geometry can be found in the SI. Briefly, the cell is sealed to the top of the cathode, preventing electrolyte contact with the back of the cathode and current collector. Contact to the Mg electrode is physically separated from the electrolyte, preventing electrolyte contact with the anode current collector. The anode and cathode are separated by a Whatman glass fiber separator (GF/D) previously dried under vacuum at 150 °C overnight. The electrolyte volume was controlled at 0.3 mL for every cell. Variable-temperature electrochemical experiments were performed in a Nylon 1/2-in. Swagelok cell. The temperature was controlled by placing the cell in a Thermo Scientific Heratherm gravity convection oven. Electrochemical Methods. The electrochemistry was controlled by a VMP-3 BioLogic potentiostat or a BCS-805 battery cycler. Mg electrodeposition and stripping curves were obtained in a vial cell geometry with a Pt wire working electrode and a Mg foil counter electrode prepared in the same manner as the electrodes used for conditioning. The CV curves were measured between −1.2 and 3.0 V (vs Mg/Mg2+) at 5 mV s−1. S8 reduction was measured with LSV in the two-electrode cell described above by sweeping negative of the open-circuit voltage (OCV) at 0.05 mV s−1, unless otherwise noted. The current was normalized to the geometric surface area of the electrode. In all cases, the S8 areal loading was between 1.5 and 2.5 mg cm−2. Galvanostatic discharge experiments of the Mg−S cells were performed in the two-electrode cell described above. The nominal C rate was calculated based on the total mass of S8 in the electrode assuming full reduction to S2−. The cell was discharged to 0.05 V (vs Mg/Mg2+). MgS oxidation was probed by LSV in the two-electrode cell described above. The MgS casts were used as working electrodes versus a Mg-metal counter/reference electrode. The cell was swept positive of the OCV to 2.75 V (vs Mg/Mg2+) at 0.05 mV s−1. Spectroscopic Characterization. 27Al NMR spectroscopy was performed on a 400 MHz Varian instrument in 5 mm screw-cap NMR tubes. An external standard, 1 M Al(NO3)3 in D2O, was measured simultaneously in a coaxial, sealed capillary. The standard was referenced to 0 ppm for each sample. The 27Al NMR was measured with a 90° pulse width of 7.12 μs, an acquisition time of 0.2 s, and a relaxation delay of 0.01 s.

EXPERIMENTAL SECTION

Electrolyte Preparation. The MACC electrolytes were prepared in an Ar-filled glovebox according to Barile et al.19 Tetrahydrofuran (THF; Fisher Scientific), dimethoxyethane (DME; VWR), and 1,3dioxolane (DOL; VWR) were dried on a solvent purification system and brought into the glovebox without exposure to air. Tetraglyme (G4) was purchased anhydrous from Sigma-Aldrich. All solvents were dried with activated 3 Å molecular sieves and stored in the glovebox. All solvents contained 10−20 ppm water, as measured by Karl Fischer titration. Anhydrous AlCl3 and MgCl2 were purchased from SigmaAldrich and Fisher Scientific, respectively, and opened in the glovebox. All MACC electrolytes maintain a 2:1 mole ratio of MgCl2 to AlCl3. The concentrations of MgCl2 and AlCl3 in the 1xMACC electrolyte are 60 and 30 mM, respectively. The concentrations of MgCl2 and AlCl3 in the 5xMACC electrolyte are 300 and 150 mM, respectively. MACC in THF was prepared by cooling 2.5 mL of THF with a Peltier plate (approximately 0 °C). AlCl3 (0.0200 g for 1xMACC and 0.1000 g for 5xMACC) was cooled similarly in a separate vial. Cold THF was added dropwise to cold AlCl3 to yield a clear, colorless solution upon agitation. MgCl2 (0.0286 g for 1xMACC and 0.1428 g for 5xMACC) was added to 2.5 mL of room temperature THF and stirred for approximately 10 min to yield a suspension. The AlCl3 solution was added dropwise to the MgCl2 suspension, and the resulting mixture was stirred overnight at room temperature to afford a clear, colorless solution. The MACC in the G4 electrolyte was prepared identically to the THF analogue. The electrolytes with binary solvent mixtures were prepared similarly to the THF analogue. For binary mixtures containing G4, G4 (2.5 mL) was cooled and added dropwise to cold AlCl3. MgCl2 was added to the second solvent in the binary mixture, either DOL or THF, and stirred. The AlCl3 solution was then added dropwise to MgCl2 in the DOL or THF suspension. The electrolytes were stirred overnight to yield a clear, colorless solution. The DME/DOL electrolyte was prepared similarly to G4/DOL, but the DME solvent was cooled and used to prepare the AlCl3 solution. The MACC electrolyte was conditioned in the glovebox with a 0.5mm-diameter Pt wire (Sigma-Aldrich) working electrode and a Mg foil (Solution Materials LLC and MTI Corp.) counter/reference electrode. The Pt was cleaned in concentrated HNO3 for at least 24 h and heated in a H2 flame prior to use. The Mg foil was mechanically cleaned with a razor blade in the glovebox. The cell was cycled from −1.2 to 3.0 V (vs Mg/Mg2+) at 5 mV s−1 until the electrolyte was conditioned or until the CV curves overlaid from cycle-to-cycle with >80% Coulombic efficiency. The electrochemistry was controlled with a VMP-3 BioLogic potentiostat. MgS Preparation. MgS was prepared from Mg-metal powder and S8 (Acros Organics, 99.5+%) without further purification. Mg and S8 were combined in a 1:1 mole ratio and ground using a mortar and pestle in the glovebox. The reactants were pressed into pellets and sealed in a vitreous silica ampule under vacuum (3 V (vs Mg/Mg2+) measured on Pt electrodes. The anodic stability associated with each solvent is listed in Table 1 and was estimated at the inflection point of the positive sweep run to higher cutoff voltages (see the SI for the CV curves). Interestingly, the conditioning process is also affected by the solvent. When prepared in THF, the deposition and stripping curves while conditioning the MACC electrolyte initially show low Coulombic efficiency, which increases to approximately 100% over many cycles. When the MACC electrolyte is made in other solvents, the Coulombic efficiency is initially greater than 100% and proceeds to drop to 98%. The Coulombic efficiency is defined as the cumulative charge passed during Mg stripping divided by the cumulative charge passed during Mg deposition. Therefore, Coulombic efficiencies over 100% suggest oxidative decomposition processes during early conditioning cycles contributing to excess anodic current. Tetraglyme (G4), a longer-chain linear ether, was also evaluated as a possible solvent for the MACC electrolyte. Although MgCl2 and AlCl3 are remarkably soluble in G4, yielding clear solutions almost immediately upon addition of the salts, the electrolyte does not support efficient Mg electrodeposition and stripping. Figure 2c shows the best CV curve obtained with the MACC electrolyte in G4. The shape of the CV curve in G4 is similar to those obtained in D

DOI: 10.1021/acs.inorgchem.9b00891 Inorg. Chem. XXXX, XXX, XXX−XXX

Forum Article

Inorganic Chemistry Table 1. Mg Electrodeposition and Stripping Metrics of MACC Electrolytes precursor salt 2:1 2:1 2:1 2:1 2:1 2:1 2:1 2:1 2:1 2:1 2:1 2:1 2:1

MgCl2/AlCl3 MgCl2/AlCl3 MgCl2/AlCl3 MgCl2/AlCl3 MgCl2/AlCl3 MgCl2/AlCl3 + 5 mg mL−1 Mg(s) MgCl2/AlCl3+ 5 mg mL−1 Mg(s) MgCl2/AlCl3 + 5 mg mL−1 Mg(s) MgCl2/AlCl3 MgCl2/AlCl3 MgCl2/AlCl3 MgCl2/AlCl3 MgCl2/AlCl3

[Mg] (M)

solventa

ηdep(mV)

CE (%)

anodic stability (V vs Mg/Mg2+)

0.06 0.25 0.06 0.06 0.06 0.04 0.04 0.04 0.06 0.06 0.06 0.06 0.06

THF DME DME THF DME DME G2 THF THF DME DME/DOL G4/THF G4/DOL

930 200 150 290 200 225 330 220 150 150 450 400 350

34 >99 92 90 95 >99 94 >99 >99 >99 93 90 95

not reported 3.1 >3 3.4 3.4 3.3 3.4 3.5 3.0 3.1 3.3 3.3 3.5

ref 22 15 19 16 18 20 20 20 this this this this this

work work work work work

a

THF = tetrahydrofuran, DME = dimethoxyethane, G2 = digylme, DOL = 1,3-dioxolane, and G4 = tetraglyme.

Figure 3. Spectroscopic characterization of 0.3 M MgCl2 and 0.15 M AlCl3 (5xMACC) in different solvent systems. 27Al NMR spectra of 0.15 M AlCl3, as-prepared 5xMACC, and conditioned 5xMACC in (a) DME, (b) G4/DOL, (c) DME/DOL, and (d) G4/THF. The resonances are referenced to an internal coaxial Al(NO3)3 standard in D2O (0 ppm). Raman spectra of the neat solvent, as-prepared 5xMACC, and conditioned 5xMACC in (e) DME, (f) G4/DOL, (g) DME/DOL, and (h) G4/THF. Modes i and ii are assigned to AlCl4−.

unconditioned MACC electrolytes in THF. Dissimilar to THF, the current density continually decreases upon extended conditioning and Mg electrodeposition and stripping behavior is not observed. Although neat G4 does not support Mg deposition and stripping, we evaluated mixtures of G4 with cyclic ethers. Figure 2d shows the CV curves obtained in the MACC electrolyte with G4/DOL and G4/THF solvents at 1:1 (v/v) ratios. While reversible Mg deposition is observed in mixtures of G4/DOL and G4/THF, the current densities are lower than that in THF and comparable to that of 1xMACC in DME/ DOL. 1xMACC was insoluble in other possible solvent systems including neat DOL and 1:1 THF/DOL. Although the MACC electrolyte in DOL yielded cloudy solutions, we attempted to condition it. The initial CV curves showed cathodic currents at >0 V and features that look like

irreversible Mg plating but at extremely high overpotentials (approximately 1.5−2 V). Eventually, the current drops to nearly 0 mA, suggesting that the electrolyte cannot be conditioned. We aimed to study the speciation in the MACC electrolyte to determine how the active species change in response to solvent systems that do not contain THF. To increase the signal from the active complex, higher-concentration electrolytes were prepared and characterized. Electrolytes at 0.3 M MgCl2 and 0.15 M AlCl3, termed 5xMACC, yield clear, colorless solutions in DME, DME/DOL, G4/DOL, and G4/ THF. The deposition and stripping CV curve obtained in the 5xMACC electrolytes can be found in the SI. The 5xMACC electrolytes were characterized with 27Al NMR and Raman spectroscopy. Parts a−d of Figures 3 show the 27Al NMR of the electrolytes as-prepared and conditioned E

DOI: 10.1021/acs.inorgchem.9b00891 Inorg. Chem. XXXX, XXX, XXX−XXX

Forum Article

Inorganic Chemistry

The LSV curves of the cathodic sweep with a S8/C composite cathode and a Mg counter/reference electrode are shown in Figure 4. All further voltages are referenced to the

in addition to the AlCl3 component solutions. Speciation of the AlCl3 solutions varies depending on the solvent with contributions from a broad resonance at 64 ppm assigned to neutral AlCl3·2THF with some autoionized AlCl3 to form the 103 ppm resonance associated with AlCl4− and the 25 ppm resonance associated with AlCl2·4THF+.23−25 Upon the addition of MgCl2 to make as-prepared MACC, the Al speciation in all solvents shifts nearly completely to a sharp resonance at 103 ppm assigned to AlCl4−.24 The 27Al NMR of the G4/THF electrolyte maintains a small resonance at 90 ppm that accounts for 6% of the Al speciation. We cannot assign the 90 ppm resonance directly, but the high chemical shift suggests that it is an anionic complex. After conditioning, the area of the peak at 103 ppm decreases, which supports our observation of irreversible Al deposition in CV curves early in the conditioning process and is consistent with the conditioning process of the MACC electrolyte in THF.17 In the G4/THF solution, the 90 ppm resonance also decreases to 0.5% of the Al signal. In the as-prepared and conditioned MACC solutions across all of the solvents, the major AlCl4− species must be charged-balanced with a cationic complex that is not observed in the 27Al NMR. We therefore suggest that the Mg complex in all cases is cationic. To probe the Mg speciation, the Raman spectrum of each electrolyte was measured. Parts e−h of Figure 3 show Raman spectra of the 5xMACC electrolyte in DME, G4/DOL, DME/ DOL, and G4/THF, respectively, before and after conditioning compared with the neat solvents. Only two modes are observed in the as-prepared or conditioned solutions that are not solvent modes, and the modes are labeled i and ii in Figure 3. The spectra are normalized to solvent modes, which remained unchanged after addition of the salts. In the Raman spectra of MACC before and after conditioning, mode i (180 cm−1) and mode ii (349 cm−1) are observed in all solvent systems. Modes i and ii are both fundamental modes of tetrahedral AlCl4− and are assigned to the F2 bending and A1 symmetric stretching modes, respectively.26,27 Interestingly, we do not observe any modes related to Mg complexes. The Raman mode at 210 cm−1 that is observed in the Raman of MACC in THF is absent. The mode at 210 cm−1 is assigned to the breathing mode in the C3v-symmetric [Mg2(μ-Cl)3· 6THF]+, which has solvent ligands.28 Upon a change in the solvent, then, it stands to reason that the complex adopts a different coordination shell, but we are unable to characterize it by Raman. Regardless, we observe efficient Mg deposition and stripping behavior in the different solvent systems despite apparent speciation changes due to the solvent. Characterization of the active complex would require extensive research and will be the subject of future studies. Solvent Effects on S8 Reduction. The different solvent systems that support reversible Mg deposition and stripping in the MACC electrolytes were used in Mg−S cells to understand how the solvent affects reduction of S8 at the cathode. Assuming that Mg−S cells discharge via a solution-mediated pathway, we expect that the S8 reduction mechanism will be affected by the solvent. S8 reduction was evaluated in several solvent systems with the MACC electrolyte and a Mg counter electrode. A primary concern with using Cl−-containing electrolytes is corrosion. Therefore, we designed an electrochemical cell that largely prevents electrolyte contact with the metal components. Details on the cell geometry can be found in the SI. All of the subsequent electrochemistry was measured in the new cell geometry unless otherwise noted.

Figure 4. LSV curves of two-electrode cells with S8/SuperP/PVDF working electrodes and Mg metal counter/reference electrodes. The electrolyte is conditioned 1xMACC prepared in (a) THF, (b) DME, (c) DME/DOL, (d) G4/DOL, (e) G4/THF, and (f) G4 solvents. The control cells with a SuperP/PVDF working electrode are shown for comparison. The dashed, vertical line indicates the voltage of the reduction wave measured with 1xMACC in the THF electrolyte. All cells were swept negative from the OCV at 0.05 mV s−1.

Mg counter/reference electrode, which we expect to be near the thermodynamic potential of Mg/Mg2+. We first evaluate S8 reduction in 1xMACC in THF to benchmark the MACC system. The cell was swept negative of the OCV at 0.05 mV s−1. As shown in Figure 4a, a reductive wave is present in the LSV curve at 0.54 V, which is lower than the theoretical reduction potential of S8 to MgS (1.77 V). The peak current densities are low (−0.05 mA cm−2 = 12 mA g−1), which is reflected by the low capacities obtained during galvanostatic discharge (vide infra). The low reduction potential and low current densities could suggest that the reduction processes are unrelated to the targeted S8 reduction. To determine if sidereactions are the cause for the low capacity and potential, a cell was assembled identically to the Mg-S cell but without S8 in the cathode. The LSV of the control cell without S8 is shown in Figure 4a. Without S8 present, no reductive wave is observed in the LSV providing evidence that the wave in the Mg-S cell is due to S8 reduction processes. To determine if S8 loading is limiting the electrochemical availability of S8, causing low current densities, we prepared cathodes with lower S8 loadings down to 10% S8. Indeed, lower S8-content electrodes increase the measured peak current to around 400 mA g−1, suggesting that the electrochemical availability of S8 plays a significant role in determining the measured capacity, similar to that observed in the Li−S literature. LSV curves of S reduction using electrodes with 10% S8, 30% S8, and 50% S8 can be found in the SI. Because the S8 cathode composition is not a main focus of this study, we simply maintain the same S8 loading for all subsequent experiments at 50%. The cathode composition, F

DOI: 10.1021/acs.inorgchem.9b00891 Inorg. Chem. XXXX, XXX, XXX−XXX

Forum Article

Inorganic Chemistry however, will be a very crucial subject of future research and will play a vital role in directing the performance. Figure 4b shows an LSV curve of S8 reduction in 1xMACC in DME. The reductive wave is shifted to a slightly higher potential than that observed in 1xMACC in THF with a similar current (normalized to the mass of S8). The addition of DOL in a 1:1 ratio to DME shifts the reductive wave to 0.79 V (Figure 4c). Parts d and e of Figure 4 show the LSV curves of S8 reduction in 1xMACC in G4 mixtures with DOL and THF, respectively. With neat G4 as the solvent, the reductive wave is largely absent (Figure 4f). We note that the counter-electrode reactions are nonideal in 1xMACC in G4, suggesting that G4 is not a viable solvent for Mg−S electrochemistry. The S8 reduction wave is observed upon the addition of THF and DOL to the G4 solvent system at 0.49 and 0.68 V, respectively. In all cases, the wave is absent in control cells again suggesting the cathodic processes are related to S8 reduction (Figure 4). The shifts in the peak potential as a function of the solvent suggest that the solvent plays a role in defining the reduction pathways at the S8 electrode. Electrolyte solvent systems without THF support higher-voltage S8 reduction. S8 reduction occurs at the highest potential in electrolytes containing DOL. The reference potential could also shift as a function of the solvent; however, we expect these changes to be minor because the Mg deposition and stripping CV curves do not show oxidation overpotentials in any of the electrolytes. Three-electrode cells would be ideal to study the S8 reduction potential. However, the development of a stable reference electrode and cell geometry is required. The LSV curves were reversed to determine the reversibility of the cathodic processes. A very broad anodic wave with current in great excess of the cathodic wave is observed in THF and G4/THF, suggesting that the anodic processes are dominated by side reactions or decomposition processes. Interestingly, the DME and G4/DOL electrolytes show no anodic current upon reversal of the scan, suggesting that the reduction is irreversible. The full CV curves can be found in the SI. To determine the effect of the electrolyte concentration on S8 reduction, Mg−S cells were prepared with 5xMACC in THF. Figure 5 shows the LSV curves of S8 reduction in 1xMACC and 5xMACC electrolytes. Note that the LSV curves with 1xMACC electrolytes are replicated from Figure 4 to facilitate a direct comparison. The current normalized to S8 mass is greatly increased in the 5xMACC electrolyte. The position of the wave, however, is not dependent on the electrolyte concentration and remains at a low voltage (0.6 V). A similar increase in the current is observed in the DME, G4/ DOL, and DME/DOL solvent systems with higher-concentration MACC. The voltage shifts to slightly lower potentials with 5xMACC in all cases. Because higher concentrations of the electrolyte do not cause a positive voltage shift in the S8 reduction wave, we suggest that neither the IR drop in the electrolyte nor the activity of Mg2+ in solution causes the high overpotential (low voltage) for S8 reduction. The greater activity of Mg2+ does increase the current associated with S8 reduction, however, highlighting the need to carefully evaluate the electrolyte compositions for Mg−S chemistry. The effect of solvent and electrolyte concentration on the galvanostatic reduction of S8 in a Mg−S cell were also evaluated. Figure 6a shows the discharge curves at C/50 with the 1xMACC electrolyte in various solvents. The profiles show a single plateau between 0.5 and 0.75 V, depending on the

Figure 5. LSV curves of two-electrode cells with S8/SuperP/PVDF working electrodes and Mg metal counter/reference electrodes. The electrolyte is conditioned MACC prepared in (a) THF, (b) DME, (c) G4/DOL, and (d) DME/DOL solvents at both 1x and 5x concentrations. The 1xMACC LSV curves are replicated from Figure 4. Higher current densities are observed in the more concentrated electrolytes, but the location of the reduction wave is largely unaffected. The dashed, vertical line indicates the voltage of the reduction wave measured with 1xMACC in the THF electrolyte. All cells were swept negative from the OCV at 0.05 mV s−1.

Figure 6. Galvanostatic discharge profiles of two-electrode cells with S8/SuperP/PVDF working electrodes and Mg-metal counter/ reference electrodes at C/50. The electrolytes are (a) 1xMACC and (b) 5xMACC in the indicated solvents.

solvent, consistent with previous results in electrolytes with HMDSMgCl + AlCl3 in THF,12 Mg(TFSI)2 in 1:1 DME/ G2,29 and [Mg(THF)6][AlCl4]2 in ionic liquid,30 which show galvanostatic reduction plateaus at 0.72, 0.2, and 0.8 V, G

DOI: 10.1021/acs.inorgchem.9b00891 Inorg. Chem. XXXX, XXX, XXX−XXX

Forum Article

Inorganic Chemistry

replicate cells was 0.019 ± 0.008 V, which cannot account for the large voltage disparity in the Mg−S cell, although the opencircuit potential is slightly larger than that of a symmetric “pristine” Mg cell: 0.006 ± 0.004 V. We also evaluated the overpotential associated with the oxidation of Mg in Mg−Mg symmetric cells. The oxidation potential was measured at 0.01 mA cm−2 using a “pristine” Mg electrode compared to the Mg anode from a discharged Mg−S cell. Three replicate experiments were performed for both geometries, and the data can be found in the SI. We observe that the oxidation potential of “pristine” Mg is similar to that of the Mg used as a counter electrode in Mg−S cells ranging from approximately 0.01 to 0.25 V with variations from cell to cell. Although we cannot directly compare the oxidation potential of Mg in a galvanostatic experiment to the potential of the Mg counter electrode in an LSV, the oxidation potential measured in the galvanostatic experiment gives us a rough idea of where the voltage may lie. Because the oxidation overpotential is