Mechanistic Study of MagnesiumâSulfur Batteries - Chemistry of

Article Cite This: Chem. Mater. 2017, 29, 9555-9564

pubs.acs.org/cm

Mechanistic Study of Magnesium−Sulfur Batteries Ana Robba,†,‡ Alen Vizintin,† Jan Bitenc,† Gregor Mali,† Iztok Arčon,§,∥ Matjaž Kavčič,§ Matjaž Ž itnik,§ Klemen Bučar,§ Giuliana Aquilanti,⊥ Charlotte Martineau-Corcos,#,∇ Anna Randon-Vitanova,○ and Robert Dominko*,†,‡ †

National Institute of Chemistry, Hajdrihova 19, 1000 Ljubljana, Slovenia Faculty of Chemistry and Chemical Technology, University of Ljubljana, Večna pot 113, 1000 Ljubljana, Slovenia § Institut Jožef Stefan, Jamova 39, 1000 Ljubljana, Slovenia ∥ University of Nova Gorica, Vipavska 13, 5000 Nova Gorica, Slovenia ⊥ Elettra-Sincrotrone Trieste S.C.p.A., s.s. 14 km 163.5, Basovizza, 34149 Trieste, Italy # Institut Lavoisier, Université de Versailles Saint-Quentin-en-Yvelines, 45 avenue des Etats-Unis, 78035 cedex Versailles, France ∇ CEMHTI-CNRS, 1d avenue de la Recherche Scientifique, 45071 Orléans, France ○ HONDA R&D Europe (Deutschland) GmbH, Carl-Legien-Strasse 30, 63073 Offenbach, Germany ‡

S Supporting Information *

ABSTRACT: Magnesium−sulfur batteries are considered as attractive energystorage devices due to the abundance of electrochemically active materials and high theoretical energy density. Here we report the mechanism of a Mg−S battery operation, which was studied in the presence of simple and commercially available salts dissolved in a mixture of glymes. The electrolyte offers high sulfur conversion into MgS in the first discharge with low polarization. The electrochemical conversion of sulfur with magnesium proceeds through two well-defined plateaus, which correspond to the equilibrium between sulfur and polysulfides (high-voltage plateau) and polysulfides and MgS (low-voltage plateau). As shown by XANES, RIXS (resonant inelastic X-ray scattering), and NMR studies, the end discharge phase involves MgS with Mg atoms in a tetrahedral environment resembling the wurtzite structure, while chemically synthesized MgS crystallizes in the rock-salt structure with octahedral coordination of magnesium.

■

INTRODUCTION Sulfur and magnesium are abundant elements in Earth’s crust and because of their low molecular weight are considered as attractive raw materials for sustainable energy storage in the future circular economy.1−6 Their difference in electronegativity7 is close to 1.7 V, and a combination of metallic magnesium anode and sulfur cathode offers theoretical energy densities of 1330 W h kg−1 and 2500 W h L−1 (calculated on the basis of the low voltage plateau at 1.4 V). Such high values are due to the ability of magnesium to provide two electrons during oxidation and due to the higher density of magnesium compared to that of lithium, which translates into an almost 2fold difference in the volumetric specific capacity (3832 mA h cm−3 versus 2062 mA h cm−3).2 Higher volumetric energy density of the magnesium metal anode compensates the low volumetric density of the sulfur cathode composite, and in the combination with nondendritic deposition of Mg metal this redox couple is highly interesting. However, coupling of a Mg anode and a sulfur cathode can be realized only in the presence of non-nucleophilic electrolytes, which has been first demonstrated by Muldoon and co-workers.1 By using hexamethyldisilazide magnesium chloride and aluminum chloride in THF, they were able to perform two discharge− © 2017 American Chemical Society

charge cycles. The cell exhibited high polarization, and the electrochemical activity of sulfur was confirmed by XPS. The short cycle life can be attributed to the high solubility of sulfur in THF. Later, Fichtner’s group2 applied magnesium bis(hexamethyldisilazide) [(HDMS)2Mg], AlCl3, and MgCl2 in diglyme and tetraglyme solvents with addition of an ionic liquid. Porous carbon (CMK-3)/sulfur was used as a cathode composite. The group’s focus was on the electrolyte modification, with the aim of improving the performance of the Mg−S battery, and by using a designated combination of solvents, they obtained a discharge curve with two plateaus. The capacity and the position of the high-voltage plateau were close to the thermodynamically expected values, while the lowvoltage plateau was ill-defined with an average voltage close to 1 V. On the basis of the ex situ XPS measurements at different points during the battery discharge, Fichtner’s group concluded that the conversion of sulfur in the presence of magnesium revealed inherently distinct battery chemistry from that of Li−S battery, due to the high kinetic barrier for MgS formation.2 The Received: September 18, 2017 Revised: October 18, 2017 Published: October 18, 2017 9555

DOI: 10.1021/acs.chemmater.7b03956 Chem. Mater. 2017, 29, 9555−9564

Article

Chemistry of Materials

K-edge X-ray absorption (XANES), resonant inelastic X-ray scattering (RIXS), and ex situ25 Mg solid-state NMR. A similar mechanism to that observed in Li−S batteries was confirmed, and the observed difference between the chemically synthesized and the electrochemically prepared MgS phase is discussed.

same research group later showed that the use of graphene oxide/sulfur composite can improve the battery stability compared to CMK-3/sulfur composite. However, the polarization after the first discharge significantly increases. High polarization in the low-voltage plateau was attributed to the formation of insoluble MgS2. Termination of the discharge, in their opinion, is caused by the formation of MgS.3 Interestingly, all works, where a magnesium bis(hexamethyldisilazide)-based electrolyte has been used, show high polarization during charging. The Manthiram group4 showed that the electrochemical stability can be significantly improved by introducing a carbon nanofiber coating between the cathode and the separator. They showed stable cycling with an initial capacity of 1200 mA h g−1, but without defined high- and low-voltage plateaus. Further improvement has been demonstrated by Zhang et al.,6 who used a boron-centered anion-based magnesium electrolyte which was prepared by one-step mixing of tris(2H-hexafluoroisopropyl) borate and MgF2 in 1,2-dimethoxyethane. Stable cycling with an average discharge capacity of 1080 mA h g−1 during the first 30 cycles was presented. All observed capacity was related to one voltage plateau at 1.1 V. No high-voltage plateau, as observed in other studies, could be seen in their measurements. The absence of a high-voltage plateau during discharge suggests a different mechanism of sulfur reduction, which could be induced by the difference in the electrolyte composition. On the other hand, use of the fluorinated alkoxyborate-based magnesium electrolytes in Mg− S batteries shows two distinct plateaus and decreasing polarization upon cycling.8 In Li−S batteries, two plateaus in the ether-based electrolyte solutions are typically related to the two different equilibriums. The high-voltage plateau corresponds to the solid−liquid equilibrium and the low-voltage plateau to a liquid−solid equilibrium with the formation of Li2S from short-chain polysulfides.9 Most studies show similar electrochemical behavior in Mg−S cells,2−4 where again two different equilibrium states can be expected, involving a reduction of sulfur through magnesium polysulfides to MgS2 or MgS. Formation of magnesium polysulfides has been determined by using XPS,1,2 which is a surface-based technique. However, it is still an open question whether this is also a bulk phenomenon and if precipitation of end discharge phases starts at the beginning of a low-voltage plateau. The electrolyte composition in Mg-based batteries plays an important role, since it defines the nature of the passivation layer on the anode and the conversion reaction mechanism on the cathodes based on sulfur or redox-active organic materials. Different generations of electrolytes have been proposed, but only few are suitable for magnesium−sulfur batteries. To simplify the electrolyte composition, our recent focus was on the mixture of ether solvents with different combinations of MgCl2 and Mg(TFSI)2. Using redox-active polymers in combination with the above-mentioned solvents and salts, we obtained excellent electrochemical stability with high capacity.10,11 Since redox-active polymers require non-nucleophilic electrolytes, similar electrolytes can also be used for Mg− S batteries. In this work our focus was on the elucidation of the bulk mechanism during Mg−S battery operation. All electrochemical characterizations were performed in the binary mixture of ether-based solvents by using commercially available salts. Reduction processes were studied by using in operando XRD, S

■

EXPERIMENTAL SECTION

Materials Preparation. Carbon/sulfur composite was prepared by ball-milling Ensaco 350G carbon from Imerys Graphite & Carbon (BET surface area 770 m2 g−1) and sulfur (Sigma-Aldrich, powder, 99.98%) for 30 min at 300 rpm in a mass ratio of 75:25, followed by heat treatment in argon atmosphere. The mixture was heated to 155 °C at a heating ramp of 0.2 °C min−1 and kept at 155 °C for 5 h. The mixture was cooled to room temperature at a rate of 0.5 °C min−1. The chemically synthesized MgS (MgSchem) was prepared by mixing Mg powder (Alfa Aesar, mesh 325, 99.8%) and sulfur (Sigma-Aldrich, powder, 99.98%) in a mass ratio of 1:1 inside an argon-filled glovebox. The mixture was transferred into a Swagelok reactor and placed in a quartz tube under argon atmosphere. The sample was heated to 500 °C at a heating ramp of 5 °C min−1 and kept at 500 °C for 8 h. The mixture was cooled to room temperature at a rate of 5 °C min−1. XANES and RIXS Measurements. The ex situ sulfur K-edge XANES spectra were measured at the XAFS beamline of Elettra synchrotron (Basovizza, Trieste)12 in fluorescence-detection mode. A Si(111) double-crystal monochromator was used with about 0.4 eV resolution at 2.5 keV. Higher-order harmonics were effectively eliminated by a double-flat silica mirror placed at a grazing angle of 8 mrad. The intensity of the monochromatic X-ray beam before the sample was measured by a 30 cm long ionization chamber detector, filled with a mixture of 30 mbar of N2 and 1970 mbar of He. The fluorescence signal was detected with a silicon drift detector. The absorption spectra were measured within the interval from −150 to 300 eV relative to the S K-edge (2472 eV). In the XANES region, equidistant energy steps of 0.2 eV were used with an integration time of 5 s per point. The exact energy calibration was established with an absorption measurement on native sulfur in transmission-detection mode. The maximum of the pre-edge peak was set to 2472.0 eV. In operando XANES and RIXS measurements were performed at the ID26 beamline of the European Synchrotron Radiation Facility (ESRF). Monochromatic X-rays with energy resolution of ∼0.35 eV at the energy of the sulfur K absorption edge were provided by a cryogenically cooled double-Si(111)-crystal monochromator. Two Si mirrors operating in total reflection were used to suppress higher harmonics. The total incoming photon flux was ∼5 × 1012 ph s−1. The incident photon beam was defocused to ∼1 × 2 mm spot size to disperse the photon flux over a larger target surface and to avoid beaminduced target radiation damage. RIXS spectra were collected by the tender X-ray emission spectrometer.13 A Si photodiode was installed in the vacuum chamber of the spectrometer to additionally record XANES spectra in fluorescence mode. Before performing the in operando measurements, we first collected ex situ XANES spectra and full K-L RIXS planes on precycled samples based on cathodes discharged to the different points of interest, a noncycled cathode immersed in electrolyte, and chemically synthesized MgS. The MgS standard was diluted with boron nitride in a mass ratio of 25:75 and subsequently pressed into pellets with a diameter of 9 mm. These ex situ samples were placed into a pouch cell with a 3.5 μm thick Mylar window foil. In operando measurements of the Mg−S battery were performed in the modified Swagelok cell originally designed for in operando XRD.14 The cell was mounted in the vacuum chamber of the spectrometer with the front window tilted at 45° relative to the incident beam. For the in operando RIXS measurements, two excitation energies were selected (Figure S1 in the SI). A sequence of RIXS and XANES spectra was recorded during the discharge process. Each measured point along the discharge curve consisted of a XANES spectrum (acquisition time = 30 s), followed by RIXS spectra recorded at both preselected energies (corresponding acquisition times were 45 and 20 s, respectively). This was followed by a nonresonant emission 9556


Article

Chemistry of Materials spectrum recorded at excitation energy 2.520 keV used for normalization purposes (acquisition time = 10 s) and another XANES spectrum recorded at the end to compare with the first, and to check for any possible chemical changes induced by the incident beam during measurement. For each measured point along the discharge curve, a fresh spot on the cathode was irradiated by moving the motorized target holder horizontally with the battery cell and the incident beam vertically, using the horizontal beamline mirror. The analysis of the sulfur K-edge XANES spectra was performed with the IFEFFIT program package ATHENA.15 Electrochemical Characterization. The electrodes were prepared by mixing the carbon/sulfur composite, the polymer binder polytetrafluoroethylene (PTFE), with multiwalled carbon nanotubes (Sigma-Aldrich, 98.0%) as conductive additive, in a mass ratio of 80:10:10 in 2-propanol solvent. The self-standing electrodes (0.79 cm2) were pressed onto a carbon-coated Al mesh collector (3.14 cm2). The electrodes were dried for 1 h at 50 °C. The sulfur loading on selfstanding electrodes was 0.75 mg cm−2. Laboratory pouch cells were assembled inside an argon-filled glovebox. The sulfur cathode was separated from the magnesium foil (Gallium Source, 99.95%, thickness 0.05 mm) anode with a glass fiber (Whatman GF/A) separator. The separator was wetted with 0.4 M Mg(TFSI)2 0.4 M MgCl2 in tetraglyme:1,3-dioxolane (TEGDME:DOL; 1:1 vol %) electrolyte. The electrolyte quantity was normalized to 100 μL mg−1 of sulfur. The batteries were cycled in a potential window from 0.1 to 2.5 V versus Mg/Mg2+ with a current density of C/60 (27.9 mA g−1) using a Bio-Logic VMP3 potentiostat/ galvanostat. The composite electrodes with carbon/sulfur ratio 1:1 were used for in operando XRD. The electrode surface was 2.00 cm2 with sulfur loading of 1.77 mg cm−2. The electrolyte quantity was normalized to 250 μL per cell. The cell was discharged to 0.1 V versus Mg/Mg2+ by using a Bio-Logic SP200 galvanostat/potentiostat at a current density of C/100 (16.7 mA g−1). The in operando XANES and RIXS measurements were performed in a modified Swagelok cell with a 6 μm Mylar foil plated with aluminum (500 Å) on the side facing the cathode.16 The cell was discharged to 0.1 V versus Mg/Mg2+ using a Bio-Logic SP200 galvanostat/potentiostat at a current density of C/60 (27.9 mA g−1). XRD measurement. The in operando XRD measurements were performed on a Siemens D5000 diffractometer with Cu Kα1 radiation (λ = 1.5406 Å) in the 2θ range 20−45° with the step of 0.04° per 11.5 s, by using a modified Swagelok cell with a 250 μm thick Be window.14 Powder X-ray diffraction measurement of synthesized MgS was performed in a glass capillary on a PANalytical X’pert PRO highresolution diffractometer with Cu Kα1 radiation (λ = 1.5406 Å) in the 2θ range 20−100° with a step of 0.033° per 3000 s. Solid-State NMR. The 25Mg MAS and static NMR spectra of MgSchem and cathodes, discharged to 0.1 V versus Mg/Mg2+, were recorded on an Avance III HD 750 Bruker NMR spectrometer equipped with a 4 mm MAS probe. The 25Mg Larmor frequency was 45.92 MHz. The samples were packed into zirconia rotors inside a glovebox under nitrogen atmosphere. The MgSchem sample was spun at 5 kHz. In the 25Mg MAS NMR measurement, a single 90° excitation pulse with duration of 20 μs was employed; the number of transients was 64, and repetition delay was 10 s. The 25Mg static NMR spectrum of the cathode material was recorded using a quadrupolar Carr-Purcell Meiboom-Gill (Q-CPMG)17 sequence with interpulse delay of 2 ms and 90° pulse of 8 μs. Repetition delay was set to 2 s, and about 100 000 transients were accumulated (measurement time was ∼2.5 days). Double-frequency sweep (DFS)18 was applied prior to CPMG to enhance sensitivity. The 25Mg chemical shifts were referenced to a saturated solution of MgCl2 at 0 ppm.

■

(Solvionic, 99.5%) salts in the ratio 1:1 was dissolved in the mixture of purified and distilled tetraglyme and dioxolane with low water content (less than 1 ppm). Such an electrolyte solution belongs to a class of non-nucleophilic electrolytes, which enables good reversibility for a metallic magnesium electrode. Stripping and deposition of the Mg metal anode in the selected electrolyte was tested in the Swagelok cell with a Pt working electrode and Mg counter and reference electrodes in the voltage window from −1 to 3.5 V versus Mg/Mg2+ (Figure S2). The stripping process proceeds through two different potentials, while deposition is shifted about 0.5 V to a lower potential, and it proceeds in the narrow voltage window. More importantly, the process is reversible and the oxidative stability is above 3 V, meaning that this simple non-nucleophilic electrolyte has a potential to be used in Mg−S batteries. Cycling of the Mg−S battery with the 0.4 M Mg(TFSI)2 and 0.4 M MgCl2 in TEGDME:DOL is shown in Figure 1. The first

Figure 1. Discharge−charge profiles of a Mg−S battery obtained in the 0.4 M Mg(TFSI)2 and 0.4 M MgCl2 in TEGDME:DOL electrolyte at the C/60 rate.

discharge is hampered by high initial polarization due to the passive film typically present on the magnesium metal in TFSIcontaining electrolytes.19 When the Mg metal surface is activated, an increase in the potential occurs instantly. Furthermore, during the first discharge, two well-defined plateaus are observed, similar to the electrochemical reduction of sulfur in the Li−S battery.20 The cell exhibits high electrochemical activity, since close to 80% of sulfur is converted into sulfide (the capacity obtained in the first discharge was 1320 mA h g−1). The oxidation process (charging) proceeds through the apparent single plateau with a well-defined cutoff potential. The measured charge capacity during the first oxidation, as well as during the following cycles, is slightly lower compared to the capacity during the discharge process. This is in contrast to what is typically observed in Li−S batteries, where, because of the “shuttle” phenomena, overcharge can significantly reduce the Coulombic efficiency. Nevertheless, the capacity fading is fast, and the battery possesses only half capacity after only four cycles. Recalculation of the electrochemical potentials from Li−S batteries shows that reduction and oxidation potentials are shifted approximately 150 mV because of increased polarization. Namely, sulfide formation (low-voltage plateau) is expected at approximately 1.35 V, while in this study it could be observed at 1.2 V; the oxidation process should occur at 1.7 V, while in

RESULTS AND DISCUSSION

A mixture of simple and commercially available magnesium salts was used for the preparation of the electrolyte solution in the binary mixture of ether-based solvents. Typically, a combination of MgCl2 (Alfa Aesar, 99.99%) and Mg(TFSI)2 9557


Article


components at preselected excitation energies.16 Two excitation energies were selected, as explained in the Experimental Section: one at the sulfur pre-edge resonance to enhance sensitivity for the polysulfide detection, and the second one at slightly higher energy to facilitate the detection of MgS formation. The actual spectral shape for the Mg polysulfides (MgSx) at pre-edge resonant excitation energy was determined using the ex situ sample of the cathode that had been stopped at the end of the first plateau where the shoulder in XANES and high-energy resolution fluorescence detected (HERFD-XAS) spectrum is most pronounced (Figure S1). The polysulfide signal is enhanced because of resonant excitation and can therefore be followed with high sensitivity and accuracy along the discharge process. Fits of the normalized in operando RIXS spectra recorded for the Mg−S battery at the lower excitation energy during the discharge are shown in Figure 3a. The measured spectra can be successfully described as linear combinations of the spectrum of a polysulfide component and of the spectrum of pure sulfur. The intensity of each separate component can be extracted precisely and followed through the discharge, as shown in Figure 3a. For the detection of MgS, we used slightly higher (about 2 eV) excitation energy. In this case, the obtained spectra are composed of two characteristic Kα1,2 doublets. While pure sulfur and polysulfides yield the same characteristic line for this particular excitation energy, the MgS doublet is shifted toward lower emission energy by approximately 0.5 eV, as it was determined by using the reference spectrum measured on the MgSchem sample. This was used to resolve the MgS component and to follow its intensity through the discharge (Figure 3b). While in XANES analyses the sulfate signal from the electrolyte dominates the spectra, the latter is heavily suppressed in the RIXS approach by using excitation well below the sulfate resonance. In this case, the electrolyte signal does not interfere with the sulfur cathode signal, which is of main interest here. Discussion of the results obtained by RIXS is presented together with the XANES results. Sulfur K-edge XANES spectra were measured in operando during the discharge of a Mg−S battery on the back side of the cathode (Figure S4), and ex situ on the electrodes taken from the batteries stopped at different states of discharge, on the side of the electrode facing the separator (Figure S3, spectra I−III). In both cases, the investigation depth is about 10 μm. The spectra exhibit a very similar edge profile and pre-edge resonances as those recorded on the cathode of a more common Li−S battery (Figure S3, spectra I and II; and Figure S4). The XANES spectrum of the cathode immersed into the electrolyte is composed of two major components; the first one is the signal of pure sulfur from the cathode with a characteristic sulfur resonance at around 2472 eV, and the second is a strong resonance at around 2480 eV, which corresponds to the sulfate signal from the TFSI anion. For the samples obtained during discharge at the high-voltage plateau (Figure S3, spectrum II; and Figure S4), a pre-edge shoulder of the main sulfur resonance is clearly observed. In direct analogy with the in operando studies performed commonly on Li−S battery systems, 21−28 this pre-edge resonance is attributed to polysulfides produced electrochemically during the discharge process. In the XANES spectrum recorded on the fully discharged cathode, the pre-edge shoulder and also the main sulfur resonance are diminished, and another spectral component appears, filling the region between the sulfur and

this setup it was found at 1.85 V. To our knowledge this result represents one of the highest experimentally measured potentials for sulfide precipitation in a Mg−S battery. Even greater improvement was achieved in the case of oxidation where a clear end of the oxidation process can be observed with the lowest polarization through all presented cycles. The difference between previous literature reports and the results of this study can be due to the different electrolyte compositions. Trying to identify the potential mechanism of sulfur reduction, we tested the Mg−S battery in the in operando XRD cell. The cell was slowly discharged at a rate of C/100, and each diffractogram corresponds to the theoretical change of Δx = 0.02 in the MgxS composition. Figure 2 shows the

Figure 2. Electrochemical discharge curve at the C/100 rate (a) and selected set of XRD diffractograms (b), measured during the compositional change marked by dotted lines. Vertical lines show the position of diffraction peaks for sulfur and MgS with rock-salt structure.

electrochemical curve and the selected diffractograms measured at the beginning of the reduction process, during the highvoltage plateau (first five diffractograms), in the middle of the low-voltage plateau (diffractogram 6), and at the end of the low-voltage plateau (diffractograms 7 and 8). Sulfur reduction proceeds and completes almost entirely during the high-voltage plateau, since peak intensities decrease and almost disappear by the end of the diffractogram 3, which coincides with the start of the transition from the high-voltage to the low-voltage plateau. Diffractograms obtained in the middle and at the end of the low-voltage plateau show no diffraction peaks of sulfur. Even though the discharge capacity is close to 1000 mA h g−1, no other crystalline phases are present at the end of discharge. In the previous reports, where the mechanism was studied with XPS on ex situ samples, it was concluded that at the end of discharge MgS2 and MgS were present in the cathode composite.1,2 The absence of diffraction peaks of MgS crystallized in the rock-salt structure does not necessarily imply that there is no MgS at the end of discharge procedure; it is possible that MgS is present in the form of nanoparticles or in an amorphous form, neither of which could be detected by XRD. To obtain more information about the reduction of sulfur in a Mg−S battery in the bulk and to detect formation of the end discharge phase, we performed ex situ and in operando RIXS and XANES measurements. The methods are complementary to one another and can be measured simultaneously during one experiment. RIXS measurements enable detection of various 9558


Article


Figure 3. (a) Sequence of normalized in operando sulfur RIXS spectra at the excitation energy of the polysulfide pre-edge resonance recorded at different times through the discharge. A linear combination fit is used to decompose the measured spectra into MgSx (blue) and sulfur (red) components, which are followed through the discharge. In the last three plotted spectra, recorded toward the end of the discharge, another weak spectral component (green) has appeared at the low-energy side. This is the MgS signal that was assigned on the basis of the reference spectrum measured on a MgS standard. (b) Sequence of four in operando sulfur RIXS spectra at the second excitation energy recorded at different times toward the end of the discharge. A linear combination fit using reference spectra of sulfur and MgS standards is used to separate the MgS signal (green) from the rest of the sulfur signal (red).

The obtained reference MgS spectrum (Mgelectrochem) is very similar, but not identical, to the XANES spectrum measured on crystalline MgSchem chemically synthesized in our laboratory (Figure S3a). All in operando XANES spectra measured during the discharge can be completely described as linear combinations of the four reference spectra. The quality of the linear combination fits (LCFs) is demonstrated in Figure 4. Using linear combination fits, we obtained relative amounts of four sulfur-containing compounds (sulfur, MgSx, MgSelectrochem, and electrolyte) in the cathode during the first discharge. The relative amounts of three sulfur compounds (pure sulfur, MgSx, and MgSelectrochem) are presented in Figure 5, and compared with the absolute yields determined from RIXS measurements. While RIXS spectra provide normalized absolute yields of each separate component, XANES spectra yield relative amounts (concentrations). Most importantly, the general evolution of separate sulfur components during the discharge determined from both sets of spectra are consistent with each other providing a clean, reliable picture of sulfur electrochemistry

the electrolyte main peaks (Figure S3, spectrum III; and Figure S4). This component is attributed to MgS. Principle component analysis (PCA),15 of the entire set of in operando XANES spectra shows that a set of four principal components completely describes all XANES spectra. Natural candidates for the four spectral components with a physical meaning present in the cathode are sulfur, sulfate groups in the electrolyte, Mg polysulfides (MgSx), and magnesium sulfide (MgS; Figure S3). The reference XANES spectra of sulfur and electrolyte were measured on pure compounds, properly diluted in carbon black to match the conditions in the battery. We were not able to record reference XANES spectra of MgSx because of lack of reference standard compounds. Although the synthesis of MgS was successful, slight discrepancies in spectra were observed, indicating that MgSchem was not a proper reference for XANES measurements. Consequently, the reference spectra of MgSx and MgS were extracted directly from the set of spectra measured on the cathode of the Mg−S battery, as described in the Supporting Information. 9559


Article


Figure 4. Linear combination fit of the sulfur K-edge XANES spectra measured in four intermediate states during discharge at nominal compositions: as prepared (a), at capacity of 195 mA h g−1 (b), at capacity of 290 mA h g−1 (c), and at capacity of 820 mA h g−1 (d). Spheres, experiment; line, best fit with linear combination of the four reference XANES profiles (sulfur, MgSx, MgS, and electrolyte), plotted below. The uncertainty of each component in the linear combination fit is ±1% or lower.

within the Mg−S battery. These results represent the composition of approximately 10 μm in depth of the back side of the cathode. However, we can assume that this is representative of the whole bulk volume, since the spectra measured on ex situ samples on the side of the electrode facing the separator reveal the presence of the same compounds in comparable amounts to those found in the in operando experiment. Thus, this set of measurements provides an insight into the bulk mechanism of sulfur conversion. Conversion of sulfur to sulfide follows a similar reaction mechanism to the recently described mechanism in a Li−S battery.16 Here we need to emphasize that the mixture of solvents in the electrolytes used in this study was the same as that used in our recent report on Li−S batteries.26 During the high-voltage plateau, sulfur is electrochemically converted into polysulfides, since with a linear decrease in sulfur ratio, a linear increase of polysulfide ratio in the cathode composite was detected. Both methods, RIXS and XANES, show unambiguous consumption of sulfur and formation of polysulfides. This part of the electrochemical curve can be considered as a solid−liquid equilibrium. As previously discussed, the length of the polysulfides cannot be determined because of the lack of suitable reference compounds and no possibility to perform

sulfur K-edge EXAFS analysis, since that part of the absorption spectrum is hampered by the signals of sulfate and chloride anions present in the electrolyte. Nevertheless, the experiments do show that sulfur reduction into polysulfides is accomplished by the end of the high-voltage plateau, and the relative amount of sulfur in the composite remains constant by the end of discharge. Similar to Li−S batteries, in Mg−S batteries MgS precipitation starts at the beginning of the low-voltage plateau. Again, this plateau corresponds to the equilibrium between polysulfides and the solid MgS phase. Generally, it can be expected that the sulfide ratio in the cathode should increase linearly with the depth of discharge, but because of inhomogeneity of the cathode used in the experiment and the small area exposed to the radiation beam, we observed inhomogeneity in the composition, which was reflected in a sudden increase of sulfide component at the beginning of the low-voltage plateau. This sudden increase corresponds to the measurement at the edge of the electrode where we can expect better electronic contact, which can facilitate the electrochemical conversion of polysulfides into sulfide. Similar behavior was observed by using other battery materials.29 In the second half of the low-voltage plateau, the relative amount of polysulfides is decreasing linearly, while the relative amount 9560


Article


Figure 5. Absolute yields and relative amounts of separate sulfur compounds within the battery cathode during the discharge determined from in operando RIXS (left) and XANES (right) measurements, respectively. The corresponding voltage diagram of the Mg−S battery cell is added at the top. Spectra were collected at different points of the cathode to diminish radiation damage. Error bars of the measured points are mainly due to inhomogeneity of the cathode composition. Relative errors were around 6%, determined by measuring several points across the surface of the cathode stopped at the end of the discharge.

of sulfide is increasing linearly. At approximately 0.8 V versus Mg/Mg2+, the conversion of polysulfides into sulfide slows down, probably because of clogging of pores and surface available for MgS precipitation. On the basis of in operando measurements, we elucidated the electrochemical conversion of sulfur into sulfide through

polysulfide formation. The remaining questions concern the length of the polysulfides and the structure of the precipitated MgS. It can be expected that polysulfides detected during battery discharge are typically mixtures of polysulfides of different chain lengths, for which equilibrium depends on the rate of discharge and the type of electrolyte. In contrast, 9561


Article

Chemistry of Materials formation of MgS should be independent of the electrolyte composition. Whereas in operando XRD suggested the absence of crystalline MgS phase at the end of discharge, we were able to detect MgS by RIXS and XANES. The two analyses also suggested that the electrochemically obtained MgSelectrochem has a different structure from the chemically synthesized MgSchem. The difference between XANES spectra for MgSchem and MgSelectrochem is shown in Figure S3. The structure of MgSchem was evaluated by the Rietveld analysis (Figure 6), which

Figure 7. The 25Mg MAS NMR spectrum of MgSchem and 25Mg static Q-CPMG spectrum of MgSelectrochem. For the latter sample, direct Fourier transformation of the time-domain train of echoes yields a spikelet spectrum, whereas Fourier transformation of the sum of these echoes gives rise to a “true” smooth spectrum.

actually slightly distorted, leading to weak electric quadrupolar interaction (with the magnitude of the quadrupolar coupling constant below 100 kHz). The electrochemically prepared MgSelectrochem exhibits a substantially different 25Mg NMR spectrum. This spectrum, recorded under static conditions, shows a relatively broad contribution between ∼20 and ∼80 ppm, with a maximum at ∼70 ppm. The maximum position clearly shows that Mg in this sample is tetrahedrally coordinated.31 Because of the relatively poor quality of the spectrum (poor signal-to-noise ratio even though the spectrum was scanned for ∼2.5 days), one cannot distinguish whether broadening of the NMR signal is due to electric quadrupolar interaction or chemical shift anisotropy. Nevertheless, it is clear that the tetrahedral MgS4 environment in this sample is more similar to the tetrahedral environment within the MgS−wurtzite form than to the tetrahedral environment of the MgS−zincblende form. In the latter, magnitudes of both electric quadrupolar interaction and chemical shift anisotropy are expected to be zero, whereas in the former they are expected to differ from zero.31 The difference in the local structures of the two MgS phases as observed by NMR also explains the difference in the XANES spectra of the end discharge phase (MgSelectrochem) and of MgSchem. Knowing the local environment in the MgSelectrochem material enables us to calculate the expected volumetric changes during sulfur conversion into MgS. The expected volumetric change during full electrochemical conversion of sulfur into MgS would be 71% if we consider that MgS is in the wurtzite crystallographic structure. This volumetric change can be reduced to 33% if one can precipitate MgS in the rock-salt crystal structure. At this stage of research, it is not clear whether this can be done effectively by changing any of the components in the battery environment.

Figure 6. Rietveld refinement of the MgSchem, synthesized by thermal treatment at 500 °C.

showed the presence of two very similar cubic rock-salt phases: the first with cell parameter a = 5.197 ± 0.001 Å and the second with cell parameter a = 5.189 ± 0.001 Å. Their intensity ratio was 3:1. Impurities of MgO were also observed, originating from starting materials. To inspect the structure of the electrochemically prepared MgSelectrochem, which exhibits no X-ray diffraction peaks, 25Mg solid-state NMR spectroscopy was employed. NMR spectroscopy offers an insight into the local structure of materials, even if the latter do not exhibit long-range order (i.e., even if they are not crystalline). The 25Mg NMR spectrum of MgSelectrochem is compared to the spectrum of the MgSchem in Figure 7. The spectrum of MgSchem, recorded at MAS conditions, exhibits a single sharp contribution at −1.1 ppm. The observed isotropic chemical shift is characteristic for octahedrally coordinated Mg species and thus confirms that the material adopts the rock-salt crystalline form, as observed by the Rietveld analysis shown in Figure 6.30,31 This form is the most stable crystalline form of MgS, more stable than the zincblende and wurtzite forms, and the only crystalline form in which Mg is octahedrally coordinated by S. In the MgS−rock-salt structure, 25Mg nuclei should experience zero chemical shift anisotropy and zero electric quadrupolar interaction. The spinning sidebands, which can be seen in the 25Mg MAS NMR spectrum of MgSchem in Figure 7, suggest that in this material MgS6 octahedra are

■

CONCLUSIONS The electrolyte solution prepared from MgCl2 and Mg(TFSI)2 salts dissolved in a binary mixture of ether solvents has been used for the preparation of the Mg−S battery, for which a 9562



■

detailed investigation of the discharge mechanism was conducted. Discharge of a Mg−S battery proceeds through well-defined plateaus, and around 80% of sulfur is converted into MgS during the first discharge. During charge, a single plateau with well-defined cutoff voltage was obtained. In operando XRD experiments show the disappearance of sulfur diffraction peaks during the high-voltage plateau, due to conversion of sulfur into polysulfides, which was confirmed by in operando RIXS and XANES analysis. The low-voltage plateau during discharge corresponds to the equilibrium between polysulfides and precipitated MgS. The latter could not be detected by XRD, but it was undoubtedly found by deconvolution of RIXS and XANES spectra. Finally, we showed that the electrochemically precipitated MgS (MgSelectrochem) has a different local structure from the synthesized MgSchem. Differences observed in NMR spectra suggest that magnesium is octahedrally coordinated in MgS chem , whereas, in MgSelectrochem, magnesium is tetrahedrally coordinated. The local geometry of the observed MgS4 tetrahedra resembles the geometry of the tetrahedra within the crystalline wurtzite form of MgS. The similarity of the structures of the precipitated MgS and MgS−wurtzite suggests the possibility of high volumetric changes during cycling, which could be one of the reasons for the fast capacity fading observed in this work.

■

REFERENCES

(1) Kim, H. S.; Arthur, T. S.; Allred, G. D.; Zajicek, J.; Newman, J. G.; Rodnyansky, A. E.; Oliver, A. G.; Boggess, W. C.; Muldoon, J. Structure and Compatibility of a Magnesium Electrolyte with a Sulphur Cathode. Nat. Commun. 2011, 2, 427. (2) Zhao-Karger, Z.; Zhao, X.; Wang, D.; Diemant, T.; Behm, R. J.; Fichtner, M. Performance Improvement of Magnesium Sulfur Batteries with Modified Non-Nucleophilic Electrolytes. Adv. Energy Mater. 2015, 5 (3), 1401155. (3) Vinayan, B. P.; Zhao-Karger, Z.; Diemant, T.; Chakravadhanula, V. S. K.; Schwarzburger, N. I.; Cambaz, M. A.; Behm, R. J.; Kuebel, C.; Fichtner, M. Performance Study of Magnesium-Sulfur Battery Using Graphene Based Sulfur Composite as Cathode Electrode and NonNucleophilic Mg Electrolyte. Nanoscale 2016, 8, 3296−9906. (4) Yu, X.; Manthiram, A. Performance Enhancement and Mechanistic Studies of Magnesium−Sulfur Cells with an Advanced Cathode Structure. ACS Energy Lett. 2016, 1 (2), 431−437. (5) Li, W.; Cheng, S.; Wang, J.; Qiu, Y.; Zheng, Z.; Lin, H.; Nanda, S.; Ma, Q.; Xu, Y.; Ye, F.; Liu, M.; Zhou, L.; Zhang, Y. Synthesis, Crystal Structure, and Electrochemical Properties of a Simple Magnesium Electrolyte for Magnesium/sulfur Batteries. Angew. Chem., Int. Ed. 2016, 55 (22), 6406−6410. (6) Zhang, Z.; Cui, Z.; Qiao, L.; Guan, J.; Xu, H.; Wang, X.; Hu, P.; Du, H.; Li, S.; Zhou, X.; Dong, S.; Liu, Z.; Cui, G.; Chen, L. Novel Design Concepts of Efficient Mg-Ion Electrolytes toward HighPerformance Magnesium-Selenium and Magnesium-Sulfur Batteries. Adv. Energy Mater. 2017, 7 (11), 1602055. (7) Zu, C.-X.; Li, H. Thermodynamic Analysis on Energy Densities of Batteries. Energy Environ. Sci. 2011, 4 (8), 2614−2624. (8) Zhao-Karger, Z.; Gil Bardaji, M. E.; Fuhr, O.; Fichtner, M. A New Class of Non-Corrosive, Highly Efficient Electrolytes for Rechargeable Magnesium Batteries. J. Mater. Chem. A 2017, 5 (22), 10815−10820. (9) Zhang, T.; Marinescu, M.; O’Neill, L.; Wild, M.; Offer, G. Modeling the Voltage Loss Mechanisms in Lithium-Sulfur Cells: The Importance of Electrolyte Resistance and Precipitation Kinetics. Phys. Chem. Chem. Phys. 2015, 17 (35), 22581−22586. (10) Bitenc, J.; Pirnat, K.; Bančič, T.; Gaberšcě k, M.; Genorio, B.; Vitanova, A. R.; Dominko, R. Anthraquinone-Based Polymer as Cathode in Rechargeable Magnesium Batteries. ChemSusChem 2015, 8 (24), 4128−4132. (11) Bitenc, J.; Pirnat, K.; Mali, G.; Novosel, B.; Randon Vitanova, A.; Dominko, R. Poly(hydroquinoyl-Benzoquinonyl Sulfide) as an Active Material in Mg and Li Organic Batteries. Electrochem. Commun. 2016, 69, 1−5. (12) Aquilanti, G.; Giorgetti, M.; Dominko, R.; Stievano, L.; Arčon, I.; Novello, N.; Olivi, L. Operando Characterization of Batteries Using X-Ray Absorption Spectroscopy: Advances at the Beamline XAFS at Synchrotron Elettra. J. Phys. D: Appl. Phys. 2017, 50 (7), 074001− 074011. (13) Kavčič, M.; Budnar, M.; Mühleisen, A.; Gasser, F.; Ž itnik, M.; Bučar, K.; Bohinc, R. Design and Performance of a Versatile CurvedCrystal Spectrometer for High-Resolution Spectroscopy in the Tender X-Ray Range. Rev. Sci. Instrum. 2012, 83 (3), 33113. (14) Leriche, J. B.; Hamelet, S.; Shu, J.; Morcrette, M.; Masquelier, C.; Ouvrard, G.; Zerrouki, M.; Soudan, P.; Belin, S.; Elkaïm, E.; Baudelet, F. An Electrochemical Cell for Operando Study of Lithium Batteries Using Synchrotron Radiation. J. Electrochem. Soc. 2010, 157 (5), A606−A610. (15) Ravel, B.; Newville, M. ATHENA, ARTEMIS, HEPHAESTUS: Data Analysis for X-Ray Absorption Spectroscopy Using IFEFFIT. J. Synchrotron Radiat. 2005, 12 (4), 537−541. (16) Kavčič, M.; Bučar, K.; Petric, M.; Ž itnik, M.; Arčon, I.; Dominko, R.; Vizintin, A. Operando Resonant Inelastic X-Ray Scattering: An Appropriate Tool to Characterize Sulfur in Li−S Batteries. J. Phys. Chem. C 2016, 120 (43), 24568−24576. (17) Larsen, F. H.; Skibsted, J.; Jakobsen, H. J.; Nielsen, N. C. SolidState QCPMG NMR of Low-?? Quadrupolar Metal Nuclei in Natural Abundance. J. Am. Chem. Soc. 2000, 122 (29), 7080−7086.

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.chemmater.7b03956. Experimental details for XANES and RIXS measurements, in operando sulfur k-edge XANES spectra and reference XANES spectra with sulfur K-edge spectra measured on cathodes stopped at different characteristic points, cyclic voltammogram of stripping and deposition of Mg metal anode, and calculation of volumetric changes due to formation of different crystallographic structures of MgS (PDF)

■

Article

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Phone: +38614760362. Fax: +38614760300. ORCID

Robert Dominko: 0000-0002-6673-4459 Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS The work has been partly supported by the Slovenian Research Agency (Grants P2-0393 and P1-0112) and by Honda R&D Europe. Access to NMR equipment and to synchrotron radiation facilities of Elettra (beamline XAFS, CERIC Project 20162028), and ESRF beamline ID26 (Project CH-5124) is acknowledged. We would like to thank Luca Olivi of Elettra for expert advice on XAFS beamline operation. The excellent assistance of the ESRF ID26 beamline staff (L. Amidani, P. Glatzel) in preparation of the synchrotron experiment is acknowledged. Financial support from the TGIR-RMN-THC Fr3050 CNRS for conducting the research is gratefully acknowledged. 9563


Article

Chemistry of Materials (18) Kentgens, A. P. M.; Verhagen, R. Advantages of Double Frequency Sweeps in Static, MAS and MQMAS NMR of Spin I = 3/2 Nuclei. Chem. Phys. Lett. 1999, 300 (3−4), 435−443. (19) Tutusaus, O.; Mohtadi, R.; Singh, N.; Arthur, T. S.; Mizuno, F. Study of Electrochemical Phenomena Observed at the Mg Metal/ Electrolyte Interface. ACS Energy Lett. 2017, 2 (1), 224−229. (20) Yamin, H.; Peled, E. Electrochemistry of a Nonaqueous Lithium/sulfur Cell. J. Power Sources 1983, 9 (3), 281−287. (21) Gao, J.; Lowe, M. A.; Kiya, Y.; Abruña, H. D. Effects of Liquid Electrolytes on the Charge−Discharge Performance of Rechargeable Lithium/Sulfur Batteries: Electrochemical and in-Situ X-Ray Absorption Spectroscopic Studies. J. Phys. Chem. C 2011, 115 (50), 25132− 25137. (22) Cuisinier, M.; Cabelguen, P. E.; Evers, S.; He, G.; Kolbeck, M.; Garsuch, A.; Bolin, T.; Balasubramanian, M.; Nazar, L. F. Sulfur Speciation in Li-S Batteries Determined by Operando X-Ray Absorption Spectroscopy. J. Phys. Chem. Lett. 2013, 4 (19), 3227− 3232. (23) Patel, M. U. M.; Arčon, I.; Aquilanti, G.; Stievano, L.; Mali, G.; Dominko, R. X-Ray Absorption Near-Edge Structure and Nuclear Magnetic Resonance Study of the Lithium-Sulfur Battery and Its Components. ChemPhysChem 2014, 15 (5), 894−904. (24) Lowe, M. A.; Gao, J.; Abruña, H. D. Mechanistic Insights into Operational Lithium−sulfur Batteries by in Situ X-Ray Diffraction and Absorption Spectroscopy. RSC Adv. 2014, 4 (35), 18347−18353. (25) Cuisinier, M.; Hart, C.; Balasubramanian, M.; Garsuch, A.; Nazar, L. F. Radical or Not Radical: Revisiting Lithium-Sulfur Electrochemistry in Nonaqueous Electrolytes. Adv. Energy Mater. 2015, 5 (16), 1401801. (26) Dominko, R.; Patel, M. U. M.; Lapornik, V.; Vizintin, A.; Koželj, M.; N. Tušar, N.; Arčon, I.; Stievano, L.; Aquilanti, G. Analytical Detection of Polysulfides in the Presence of Adsorption Additives by Operando X-Ray Absorption Spectroscopy. J. Phys. Chem. C 2015, 119 (33), 19001−19010. (27) Gorlin, Y.; Siebel, A.; Piana, M.; Huthwelker, T.; Jha, H.; Monsch, G.; Kraus, F.; Gasteiger, H. A.; Tromp, M. Operando Characterization of Intermediates Produced in a Lithium-Sulfur Battery. J. Electrochem. Soc. 2015, 162 (7), A1146−A1155. (28) Gorlin, Y.; Patel, M. U. M.; Freiberg, A.; He, Q.; Piana, M.; Tromp, M.; Gasteiger, H. A. Understanding the Charging Mechanism of Lithium-Sulfur Batteries Using Spatially Resolved Operando X-Ray Absorption Spectroscopy. J. Electrochem. Soc. 2016, 163 (6), A930− A939. (29) Ouvrard, G.; Zerrouki, M.; Soudan, P.; Lestriez, B.; Masquelier, C.; Morcrette, M.; Hamelet, S.; Belin, S.; Flank, A. M.; Baudelet, F. Heterogeneous Behaviour of the Lithium Battery Composite Electrode LiFePO4. J. Power Sources 2013, 229, 16−21. (30) Pallister, P. J.; Moudrakovski, I. L.; Ripmeester, J. A. Mg-25 Ultra-High Field Solid State NMR Spectroscopy and First Principles Calculations of Magnesium Compounds. Phys. Chem. Chem. Phys. 2009, 11 (48), 11487−11500. (31) Mali, G. Ab Initio Crystal Structure Prediction of Magnesium (Poly)sulfides and Calculation of Their NMR Parameters. Acta Crystallogr., Sect. C: Struct. Chem. 2017, 73 (3), 229−233.

9564


Mechanistic Study of MagnesiumâSulfur Batteries - Chemistry of

Recommend Documents

Mechanistic Study of MagnesiumâSulfur Batteries - Chemistry of