Investigation of the reaction mechanism of lithium sulfur batteries in different electrolyte systems by in situ Raman spectroscopy and in situ X-ray diffraction

Abstract

Lithium–sulfur batteries are of great interest owing to their high theoretical capacity of 1675 mA h g⁻¹ and low cost. Their discharge mechanism is complicated and it is still a controversial issue. In the present work, in situ Raman spectroscopy is employed to investigate the poly-sulfide species in the sulfur cathode and in the electrolyte during the cycling of Li–S batteries. The aim is to understand the discharge mechanism and the influence of the electrolyte on the dissolution of sulfur and poly-sulfides. S₈ⁿ⁻ is identified as the main species in the high voltage plateau of discharge together with cycloocta S₈, in the cell using 0.5 mol L⁻¹ LiTFSI–PY₁₃–FSI as the electrolyte. S₄²⁻, S₂²⁻ and S²⁻ are detected soon after the low voltage plateau is reached. A discharge mechanism in the PY₁₃–FSI is proposed based on the identified species which provides important information for improving and designing cathodes. Electrolytes of 0.5 mol L⁻¹ LiTFSI–PY₁₃–FSI and 1 mol L⁻¹ LiTFSI–DOL–DME are used in studying the dissolution of sulfur and poly-sulfides. The results demonstrate that the same poly-sulfide species are present in the two electrolytes. However, the rates of poly-sulfide formation and diffusion to the anode are slow in the ionic liquid compared to those in the ether-based electrolyte due to different ionic mobilities of various species in the two electrolytes. These differences are evidenced by the observation of poly-sulfide species in the DOL–DME from the very beginning of cell assembly even before starting the discharge whereas their appearances, in the ionic liquid, are delayed and only found at the end of the high voltage plateau. Notably, the soluble elemental sulfur is clearly observed in the ionic liquid electrolyte during the first discharge in the high voltage region, which is very different from the DOL–DME system where the elemental sulfur is quickly reduced to poly-sulfides due to self-discharge reactions. In addition, the elemental sulfur is also detected near the lithium anode in DOL–DME at the end of charge, for the first time to our knowledge, which suggests that the degradation of lithium metal is caused by the multiple reactions of the lithium metal surface with soluble poly-sulfides and/or elemental sulfur.

---

1. Introduction

Lithium–sulfur rechargeable batteries are one of the most promising candidates for various applications owing to their high theoretical energy density (2600 Wh kg⁻¹), high theoretical capacity (1675 mA h g⁻¹), good safety, natural abundance and low cost. Despite these advantages, the present performance of Li–-S batteries, such as in terms of rate capability, energy density and cyclability, is unable to meet the requirements for practical applications due to the following shortcomings:^1–6

(1) Low electrical conductivity of sulfur (10⁻¹⁸ S cm⁻¹ (ref. 7)); the insulating properties result in low rate capability, as well as low sulfur loading owing to the addition of large amounts of conductive materials.

(2) Dissolution of sulfur⁸ and poly-sulfides in the electrolyte that causes the loss of active material results in shuttle reactions of poly-sulfides between the cathode and anode, which contribute to the deterioration of the lithium anode. The shuttling also deposits quasi insulating Li₂S (∼10⁻¹³ S cm⁻¹ (ref. 9)) on the anode surface, which increases the cell resistance leading to rapid capacity fading.

(3) The large volume change during cycling induced by density differences between sulfur (2.03 g cm⁻¹ (ref. 10)) and lithium sulfide (1.63 g cm⁻¹ (ref. 10)) fractures the positive electrode and reduces its capacity.

(4) Depletion of the electrolyte over cycling due to irreversible side reactions.

To overcome these difficulties, recently, many efforts have been devoted to comprehending the mechanism of the charge–discharge process for designing and improving cathode materials for Li–S batteries. The high theoretical capacity of Li–S batteries is based on the electrochemical reduction of cycloocta S₈ to Li₂S, i.e., one S₈ ring hosts 16 Li⁺ ions at the end of discharge: this is a stepwise process involving different lithium poly-sulfides as intermediate phases. It is commonly believed that the long chain poly-sulfides, Li₂S_n (n = 8–6), are mainly responsible for the first discharge plateau at ∼2.4 V, whereas the short chain poly-sulfides Li₂S_n (n = 4–1) are the main species involved in the second plateau at ∼2.1 V.^11,12 However, the exact mechanism is still not totally understood due to the discrepancies in the intermediate poly-sulfide phases formed and their corresponding potentials, as reported by various groups. These differences in the observed data originate from the complexity of the electrochemical and chemical reactions involved, as well as from the sensitivity of the sulfur and poly-sulfides to the experimental conditions and their environment, such as the electrolyte.^13,14 Many studies were focused on the complex discharge/charge mechanism via various characterisation techniques, especially the in operando ones that provide electrode information as the battery is cycling. Several groups have studied the phase transformation of the sulfur cathode by in situ X-ray diffraction (XRD). Walus et al.¹⁵ observed the co-existence of α-sulfur and β-sulfur during charge and Li₂S during discharge. In addition, Cãnas et al.¹⁶ found that the sulfur phase obtained after charge was different from the initial sulfur structure, but no phase identification was reported. Recently, crystalline Li₂S₂ was identified for the first time under solvent-in-salt electrolyte conditions.¹⁷ In contrast to Walus' findings, Nelson and co-workers¹⁸ did not detect the formation of crystalline Li₂S during discharge; they also suggested that the recrystallization of sulfur at the end of the charge cycle depended on the preparation methods of the cathode (such as the carbon surface area, sulfur particle size, etc.). However, the non-crystalline nature of the intermediate phases in the Li–S battery makes their identification with XRD impossible. Other techniques, therefore, must be employed. Patel et al.¹⁹ recorded the Ultraviolet-Visible (UV/Vis) spectrum evolution of a Li–S battery during cycling; they identified that Li₂S₈ and Li₂S₆ existed at the high voltage plateau, while Li₂S₄ and Li₂S₃ were dominant at the low voltage one, and Li₂S₂ and Li₂S were the principal species when the voltage was lower than the low voltage plateau (∼2 V). Using a polymer electrolyte that partially reduces the dilution of the catholyte, the in operando measurements yielded UV-vis absorption bands that are well resolved, which improved our understanding of the spectra and identification of the various sulfur species present at different steps of the charge–discharge cycles. In particular, these measurements gave evidence that the charge and discharge mechanisms mostly differ by the formation of S₄²⁻ during discharge and S₆²⁻ during charge.²⁰ The poly-sulfides present during cycling were also characterised by in operando X-ray Absorption Spectroscopy (XAS).²¹ Cuisinier et al.^14,22 identified S₈/S₈²⁻ and S₆²⁻ species at the 2.4 V discharge plateau, whereas S₆²⁻ and S₄²⁻ were intermediate species at the 2.1 V plateau in the mixture of low dielectric solvents 1,3-dioxolane and 1,2-dimethoxyethane (DOL–DME). On the other hand, the S₃˙⁻ radical and S₄²⁻ were detected in high dielectric constant solvents including dimethylacetamide (DMA) and others as reviewed in ref. 13. Their results demonstrated that the intermediate poly-sulfides were closely related to the choice of the electrolyte and that of the solvent in particular. Wujcik and co-workers²³ observed that Li₂S₆, LiS₅ and S₈ were dominant from the open circuit voltage to 2.25 V; Li₂S, Li₂S_3/5 and Li₂S_6/8 were the main species from 2.25 to 2.02 V, whereas Li₂S, Li₂S₂ and Li₂S₅ increased as the discharge proceeded to 1.5 V. The in situ Raman spectroscopy technique is also effective in examining the transitional species during cycling.^24,25 Hagen et al.²⁶ have calculated the Raman band positions of various mono-anions and di-anions of poly-sulfides with Density Functional Theory (DFT). Combining the DFT calculation and in situ Raman measurements on a Li–S cell with several electrolytes, they proposed that there were polysulfide mono-anions next to di-anions in every state of charge. However, identifying the exact species in the experimental spectra is challenging due to the difficulty in de-convoluting the overlapped bands. With the same calculation and experimental technique, Hannauer and co-workers²⁷ indicated that S_n²⁻ species (n = 6, 8) peaked near the end of the ∼2.4 V plateau. They also observed that the intensity of the S₃˙⁻ band was highest just before the beginning of the 2.1 V plateau in the cell using TEGDME (glycol dimethyl ether)–DOL as the electrolyte. In addition, the S₃˙⁻ and S₂O₄²⁻, together with S₄²⁻ and S₄⁻ were observed by Wu et al.²⁸ in the cell using the same electrolyte. Other techniques, such as Fourier transform infrared spectroscopy (FTIR),²⁹ Electron Paramagnetic Resonance (EPR)¹² and High Performance Liquid Chromatography (HPLC)¹¹ were also used to detect and identify the intermediate species formed during cycling. Wild et al.³⁰ reviewed various analytical studies and physical models, and brought together recent advances in the mechanistic understanding of the Li–S battery. They suggested that this understanding was still hindered by the limitations of available techniques and the dependence of the mechanism on cell design. In addition the experimental analysis did not consider spatial aspects of the mechanism, as chemically much of it is envisaged in the electrolyte whilst electrochemical reactions occurred at surfaces.

In the present work, we employed in situ Raman spectroscopy to monitor the evolution of poly-sulfide species both on sulfur cathode surfaces and in electrolytes near the Li anode aiming at extending our knowledge in spatial aspects of the battery; in addition, we used in-situ XRD to follow the crystalline phase transformation in the cathode phase to elucidate the reaction mechanism during galvanostatic cycling. Ultimately, the information obtained will be useful for designing and improving Li–S batteries.

2. Experimental

2.1 Electrode preparation

Sulfur (purity > 99 998%, Sigma Aldrich) and Ketjen Black (AkzoNobel) were ground in a mortar. A slurry was prepared with a mixture of sulfur, Ketjen Black and PVDF binder (KF7305, Kureha) at a ratio of 60 : 30 : 10 in a SPEX mixer using N-methyl pyrrolidone (NMP, purity > 99.5%, Sigma Aldrich) as the solvent. The slurry was first coated on one side of an aluminium mesh via a doctor blade method, and then dried at 75 °C, followed by the coating and drying of the second side.

2.2 Electrochemical cell preparation and measurements

The in situ Raman cells were made from a 2030 coin cell with a 0.15 mm thick quartz plate as the window. The electrochemical cells consist of a sulfur electrode with aluminium mesh as the current collector, a separator (Celgard 3501 or Whatman glass microfiber filters Grade GF/C), a lithium counter electrode and an electrolyte. Two types of electrolytes were used: (a) 0.5 mol L⁻¹ LiTFSI (lithium bis-(trifluoromethylsulfonyl)imide) dissolved in PY₁₃ (N-methyl-N-propylpyrrolidinium)–FSI (bis(fluorosulfonyl)imide) and (b) 1 mol L⁻¹ LiTFSI dissolved in DOL–DME (DOL : DME = 1 : 1, v/v). The solid components were dried at 120 °C overnight under vacuum except for the cathode which was dried at 55 °C for 3 hours under low vacuum. The cells were then assembled in a helium-filled glovebox with two configurations. For studying the changes of the sulfur cathode during the charge–discharge, the electrode was placed in contact with the cell window (Fig. 1a), whereas for monitoring the species in the electrolyte on the lithium side, a small hole was punched at the center of the lithium disk which was in contact with the quartz window, see Fig. 1b.

Fig. 1 In situ Raman cell configurations for examining the changes in the (a) cathode and (b) electrolyte near Li metal.

The in situ XRD cells were made from a 2030 coin cell with a 0.25 mm thick beryllium disk as the window. 1 mol L⁻¹ LiTFSI dissolved in DOL–DME–2% LiNO₃ was used as the electrolyte.

The electrochemical measurements were carried out on a Biologic SP300 (BioLogic Science Instrument), controlled by EC-Lab (V10.19), between 2.8 V and 1.0 V at slow rates of C/48–C/60.

2.3 In situ Raman spectroscopy and X-ray diffraction measurements

The in situ Raman measurements were performed on a LabRam Aramis Spectrometer (Horiba, Jobin Yvon) controlled by LabSpec 5. A laser beam of wavelength 532.1 nm and energy 0.14–1.22 mW, with a spot size of ∼ 2.6 μm, was focused with a 10× objective. The measurements were carried out in the wavenumber range of 50–800 cm⁻¹ every 30 minutes with the data collection time varying between 600 and 45 seconds. The sulfur powder and cathode were also examined.

The in situ XRD measurements were performed on a D8 Advance Diffractometer (Bruker) with Cu-Kα radiation from 21–38° at a step size of 0.025° and 18 s per step. The sulfur powder, electrode and beryllium disk (Fig. S1, ESI†) were also characterised by X-ray diffraction from 10° to 80°.

3. Results and discussion

3.1 Characterisation of sulfur powder

The sulfur powder was analyzed by X-ray diffraction prior to fabricating the electrode. All the diffraction peaks can be indexed to α-sulfur (PDF: 01-078-1888); the lattice parameters obtained from Rietveld refinement with a space group (S.G.) Fddd are: a = 10.46017(7), b = 12.86286(0), and c = 24.46326(0) (Fig. S2, ESI†). The crystallinity of the sulfur estimated with software DiffracEva (Bruker) is ∼76%, and its average crystal size is ∼60 nm as calculated by the Scherrer formula with the same software. The Raman spectrum of the powder shows a band at 79 cm⁻¹ related to the torsion mode in sulfur, bands at 147, 180, 213, 241 cm⁻¹ related to the bending mode and bands at 432 as well as 467 cm⁻¹ related to the stretching mode^31,32 (Fig. S3, ESI†).

3.2 Characterisation of the electrochemical properties of the sulfur electrode

Several cells using both 0.5 mol L⁻¹ LiTFSI–PY₁₃–FSI and 1 mol L⁻¹ LiTFSI–DOL–DME electrolytes were cycled at slow rates; the discharge capacities of the first cycle ranged from ∼700 to ∼1450 mA h g⁻¹, see Fig. 2. In general, cells using ionic liquids have low capacities, which are due to (1) the slower mobility of ionic species in more viscous liquids; (2) poor electrode-wetting properties; (3) the lower solubility of sulfur redox intermediates, which is positive for inhibiting parasitic reactions but negative in terms of kinetics. Despite the low capacities of certain cells, the discharge curves still show the typical features of the sulfur reduction reaction. The high and low voltage plateaus were observed at ∼2.2 V and ∼2.03 V for the cells using ionic liquid, and ∼2.3 V and ∼2.1 V for those using DOL–DME. The high voltage plateau is generally believed to be related to the formation of long chain poly-sulfides (Li₂S_n, n = 6–8), whereas the low voltage plateau is the result of the reduction of long chain poly-sulfides to the shorter ones.^11,19,22 The voltage plateaus of the cells using PY₁₃–FSI have lower values compared to those using DOL–DME, which suggests that it is more difficult to form poly-sulfide species in the ionic liquid owing to the low mobility of various species; therefore, more energy is required. The charge curves displayed only one voltage plateau with the mid values of ∼2.36 and ∼2.30 V for cells using 0.5 mol L⁻¹ LiTFSI–PY₁₃–FSI and 1 mol L⁻¹ LiTFSI–DOL–DME, respectively. The cell using DOL–DME with addition of LiNO₃ showed a high capacity of ∼1450 mA h g⁻¹ which was attributed to the formation of a passivation layer on the Li surface and suppression of the redox shuttle of lithium poly-sulfides.³³

Fig. 2 The capacities (first discharge cycle) obtained from the cells using PY₁₃–FSI and DOL–DME as electrolytes.

3.3 In situ Raman measurements

3.3.1 Development of Raman spectra collected from the sulfur cathode during cycling using 0.5 mol L⁻¹ LiTFSI in PY₁₃–FSI as the electrolyte. Raman spectra were collected from the sulfur electrode in which 0.5 mol L⁻¹ LiTFSI–PY₁₃–FSI was used as the electrolyte (Celgard 3501 as the separator) to reduce the amount of poly-sulfides dissolved in the liquid. Fig. 3a demonstrates the evolution of the Raman spectra of the first cycle. During the initial period of the discharge, all the bands appeared in the spectra are assigned to sulfur (S₈) except one at ∼730 cm⁻¹ ascribed to LiTFSI.^34,35

Fig. 3 A Li–S cell with 0.5 M LiTFSI–PY₁₃–FSI as the electrolyte cycled at a rate of C/48. (a) Evolution of Raman spectra of the sulfur electrode, and the inset is enlarged Raman spectra at two different voltages of discharge; (b) zoomed-in image of the green areas in (a), plotted every 3 spectra; (c) the polysulfide species detected via Raman with their corresponding potentials. Red: discharge; blue: charge. (d) The intensity ratio of different species as a function of electric potential.

The intensities of the sulfur bands decreased as the discharge proceeded and disappeared completely at the beginning of the low voltage plateau, ∼2.04 V (Fig. 3a), indicating that almost all the S₈ rings were opened to form long chain poly-sulfides. This ring opening was initially observed on the high voltage plateau, ∼2.17 V; it was characterised by the formation of a band at 398 cm⁻¹ which persisted up to ∼2.04 V (low voltage plateau) and was assigned to the S–S stretching mode in S₈ⁿ⁻ (n = 1, 2).^36,37 In this voltage range, the electrochemical (EC) reaction is:

S₈ + nLi⁺ne⁻ → Li_nS₈ (n = 1, 2) (EC-1)

S₆²⁻ was reported to be present in this voltage region by in situ UV-vis^19,20 and XANES²² using sulfolane,¹⁹ solid polyether-based polymer²⁰ and DOL–DME²² as electrolytes respectively; Hannauer et al.²⁷ reported the presence of S₈²⁻ and S₆²⁻ in the cell with TEGDME–DOL at discharge between 2.3 and 2.1 V. They also observed S₃˙⁻ which was believed to be the product of the disproportionation of S₆²⁻.¹¹ However, S₆²⁻ and S₃˙⁻ radicals were not detected in some other published studies which is attributed to the effect of solvents. The influences of solvents' properties, such as dielectric constant and donor number, on the sulfur redox reactions have been studied via using DOL–DME (low dielectric constant and low donor number) and dimethyl sulfoxide (DMSO; high dielectric constant and high donor number). The results demonstrated that the sulfur reduction mechanisms depend on the solvents employed; S₆²⁻ was not involved in the discharge in DOL–DME due to the fast reduction of S₄²⁻ from S₈²⁻ whereas S₆²⁻ as well as S₃˙⁻ participated in the reaction due to the disproportionation of S₈²⁻ to S₆²⁻ in DMSO which provided an additional pathway for sulfur reduction;¹³ the authors attributed the differences to the solvent's ability to solvate charged poly-sulfide anions.^13,38 The different mechanisms reported by various authors may arise from the different electrolytes used and non-identical cycling conditions; the overlapping of the characteristic bands in the spectra also makes the exact identification of the species difficult.^11,13,14,23

As discharge proceeded, small bands at ∼453, ∼201 and ∼490 cm⁻¹ started to appear at the beginning of the low voltage plateau, ∼2.05 V, see Fig. 3b for clarity which shows plots every three spectra in the green areas of Fig. 3a. The 201 cm⁻¹ band was assigned to the bending mode of S₄²⁻, and the 453 and 490 cm⁻¹ bands were assigned to the stretching mode of S₄²⁻.^39,40 The presence of S₄²⁻ is widely observed with various analytical techniques despite the differences in the exact voltage intervals on the discharge curve. The S₄²⁻ species was detected on the voltage slope between high and low voltage plateaus in some studies,^11,28 whereas other investigators reported it on the low plateau.¹⁹ Again, the electrolytes used may play a role. The S₄²⁻ species was detected before the beginning of the low voltage plateau in the cells with TEGDME–DOL; the earlier polysulfide formation in this electrolyte than in the ionic liquid, in our case, is attributed to the higher solubility of poly-sulfides in the ether-based solvent with lower viscosity and faster reduction of S₈²⁻ to S₄²⁻ in it.¹³ Although the dissolution of the active material in the electrolyte is not desired in the battery, the poly-sulfides in the liquid phase may accelerate the electrochemical and chemical reactions. In addition, S₂²⁻ has its strongest band at ∼450 cm⁻¹ (stretching mode);^39,41,42 therefore, the 453 cm⁻¹ band observed may have a certain contribution of S₂²⁻ under these conditions. The small band at ∼360 cm⁻¹ is probably related to the characteristic T_2g phonon mode of Li₂S corresponding to the Li–S stretching mode.⁴³ Li₂S is thus formed as the end product of discharge, despite the chemical disproportionation of various poly-sulfides proceeding concurrently. The bands at 453 and 201 cm⁻¹ disappeared just before the end of discharge indicating that the S₄²⁻ and S₂²⁻ were consumed by the continuous formation of Li₂S. The major electrochemical and chemical (C) reactions at the low voltage plateau are proposed as follows:

Li₂S₈ + 2Li⁺ + 2e⁻ → 2Li₂S₄ (EC-2)

2Li₂S₈ → 2Li₂S₄ + S₈ (C-1)

Li₂S₄ + 2Li⁺ + 2e⁻ → 2Li₂S₂ (EC-3)

Li₂S₂ + 2Li⁺ + 2e⁻ → 2Li₂S (EC-4)

3Li₂S₂ → Li₂S₄ + 2Li₂S (C-2)

During the initial stage, the chemical reaction (C-1) was slow compared to (EC-1) and (EC-2), which was confirmed by the depletion of the S₈ as the voltage reached 2.04 V where S₈ⁿ⁻ and S₄²⁻ were the main species. Although S₂²⁻ has its strongest band at ∼450 cm⁻¹ close to the band of S₄²⁻(453 cm⁻¹), its amount was small at the beginning of the low voltage plateau, which was supported by the high intensity of the band at 201 cm⁻¹, characteristic of S₄²⁻. The ratio of I_{(201 cm⁻¹)}/I_{(453 cm⁻¹)} at this initial stage was ∼0.3, which continuously decreased as discharge proceeded, suggesting the increasing contribution of S₂²⁻ to the 453 cm⁻¹ peak (see Fig. 3a inset). As the discharge continued through the low voltage plateau, the main reactions are (EC-3), (EC-4) and (C-2), where disproportionation (C-2) was a slower process compared to the reduction reactions, as shown by the disappearance of S₄²⁻.

Just before the end of discharge, two more bands were observed, one at ∼392 cm⁻¹ near 1.9 V followed by another at ∼430 cm⁻¹ around 1.8 V; the former faded away before the end of discharge. Despite the S₈ⁿ⁻ band located near ∼392 cm⁻¹, the formation of S₈²⁻ from the disproportionation of S₄²⁻ or S₂²⁻ is energetically unfavorable based on the calculated results published.⁴⁴ At present, it is difficult to determine the species corresponding to the two bands (marked as “?” in Fig. 3a).

The formation of poly-sulfide species during charge was almost a reverse process to the one observed during discharge, in spite of only one voltage plateau in the charge curve as reported by some researchers.^27,45 The unidentified band at ∼430 cm⁻¹ disappeared first as charge began. The bands of S₄²⁻ and of S₂²⁻ were observed at 201 and 453 cm⁻¹ as soon as the voltage reached ∼2 V, and the characteristic band of Li₂S (∼360 cm⁻¹) vanished as the voltage reached the ∼2.2 V plateau via the oxidation from S²⁻ to S₂²⁻. In addition, S₈ⁿ⁻ was observed at ∼2.3 V and persisted up to ∼2.6 V where the sulfur ring, S₈, started to form and grew thereafter. The Raman investigation in the present study was not able to determine whether the cycloocta S₈ formed during charge is in the orthorhombic (α-S₈) or monoclinic (β-S₈) crystal structure because the fundamental vibration bands of the two types of sulfur are indistinguishable in our spectrum range. The differences between the two allotropes lie in the range of Raman shift <100 cm⁻¹.^46,47Fig. 3c summarizes the observed poly-sulfide species with their corresponding potential profiles, which show that the S₄²⁻ and S₂²⁻ are reversibly coupled with the charge/discharge process. Four bands which have the least overlap with the other bands are selected to represent S²⁻ (∼360 cm⁻¹), S₄²⁻ (∼200 cm⁻¹), S₈ⁿ⁻ (∼390 cm⁻¹) and S₈ (∼217 cm⁻¹) species. The intensity ratio of individual bands to the sum of the four bands is calculated as:

where i = S²⁻, S₄²⁻, S₈ⁿ⁻ and S₈. (The band at ∼200 cm⁻¹ is not the strongest of S₄²⁻, but it is well separated from the others, and therefore, when S₄²⁻ ∼0 obtained from the above equation does not necessarily mean no S₄²⁻, it could be a small amount). Fig. 3d shows the plots of the intensity ratio of the four species as a function of potential showing a relative amount of the poly-sulfide species at different stages of the cycling. The cycling curve in this figure becomes very noisy at the end of the charge process which is likely due to the bad contact inside the in situ cell.

All the mechanisms proposed in literature on Li–S batteries obtained via the in situ Raman technique, involved the formation of the S₃˙⁻ anion radical that is different from the present studies which is related to the experimental conditions and electrolytes involved. Table 1 shows the electrolyte used and the corresponding poly-sulfide species detected as well as the mechanisms proposed.

Table 1 Summary of mechanisms of Li–S batteries proposed based on in situ Raman studies

Authors	Electrolyte used	Species detected	Mechanism proposed
Hagen et al.^26	0.7 M LiTFSI in DME–DOL (2 : 1) + 75 ppm BHT + 0.25 M/L LiNO3	OCV:
		S8 and possibly S7^2−, S8^n− (n = 1, 2), S5^−
		Discharge:
		• S3˙^−
		• Likely to have Sn^2− (n = 2–7)
		Charge:
		Almost the reverse of discharge
Hannauer et al.^27	1 M LiTFSI in TEGDME–DOL (1 : 1)	Discharge:	Discharge:
		S8, S8^2−/S6^2−, S3˙^−
		Charge:	• S8 → S8^2−/S6^2− → S3˙^− + possibly other short chain PS → S^2−
		S3˙^−, S8^2−/S6^2−	• Charge is a reverse process of discharge
Wu et al.^28	1 M LiTFSI in TEGDME–DOL (1 : 1)	Discharge:	1^st order reaction model
		S8, S8^2−, S4^2−, S4^−, S3˙^−, Sn^2− (n = 4–8), S2O4^2−
		Charge:	• Short chain poly-sulfides are formed directly from S8
		S4^2−, S4^−, S3˙^−, S8, S8^2−, Sn^2− (n = 4–8), S2O4^2−	• Charge and discharge are quasi-reversible and have similar rate constants
Yeon et al.^25	1 M LiTFSI in	Discharge:	Discharge in TEGDME & TEGDME–DOL:
	• TEGDME–DOL (4 : 6)		• S8 + 2e^− → S8^2−
	• TEGDME	S8, Sn^2− (n = 4–8), S^2−, S3˙^−	• S8^2− + ne^− → (8/n) Sn^2− (n = 4–8)
	• DOL		• 2S4^2− → 2S3˙^− + S2^2−
		Charge:	• S2^2− + 2e^− → 2S^2−
			Discharge in DOL:
		S^2−, S3˙^−, Sn^2− (n = 4–8), S8	• S8 + 2e^− → S8^2−
			• S8^2− + ne^− → (8/n)Sn^2− (n = 4–8)

---

3.3.2 Development of Raman spectra from the electrolytes near the anode during cycling using 0.5 mol L⁻¹ LiTFSI in PY₁₃–FSI as the electrolyte. The poly-sulfide species in the electrolyte on the lithium metal side of the cell were also investigated. Cells with the configuration in Fig. 1b were prepared using the electrolytes (a) 0.5 mol L⁻¹ LiTFSI in PY₁₃–FSI and (b) 1 mol L⁻¹ LiTFSI in DOL–DME. We used Whatman filter paper as separator in order to provide abundant electrolyte. The spectra were collected beside the lithium anode.

The Raman spectra from the cell using 0.5 mol L⁻¹ LiTFSI PY₁₃–FSI electrolyte are shown in Fig. 4a. All the bands in the initial spectrum were assigned to the electrolyte⁴⁸ (Fig. S4, ESI†) indicating that no sulfur/poly-sulfide had diffused to the anode side before cycling began. As the discharge voltage decreased to ∼2.19 V, sulfur bands were observed at 149, 217 and 471 cm⁻¹ and existed up to ∼2.09 V. This is the first time to report the solubility of elemental sulfur in the ionic liquid. Soon after detection of the sulfur bands, the initial broad band centred at ∼400 cm⁻¹ (it can be de-convoluted into two peaks centered at ∼395 and 404 cm⁻¹) increased in intensity at 2.13 V, especially the part at ∼395 cm⁻¹ which is attributed to the formation of S₈ⁿ⁻. The band at ∼433 cm⁻¹ from the cell became enhanced and broadened, which can be de-convoluted into two peaks at 433 and 448 cm⁻¹. The intensity of the 448 cm⁻¹ peak increased as discharge proceeded. In addition, a band at 198 cm⁻¹ was observed as the 448 cm⁻¹ peak grew stronger; this pair of bands is associated with the S₄²⁻. The presence of high order poly-sulfides at the end of discharge indicates that the conversion of the active material is not complete with the utilization of active sulfur being inefficient. Towards the end of charge (V > 2.37 V), the intensities of the signals decreased and the spectra became noisy; it is difficult to discern any bands related to the cathode material. Throughout cycling, two types of poly-sulfide anions, S₈²⁻ and S₄²⁻, were observed and at the end of charge, no elemental sulfur was detected. The species observed at the different stages of the cycling are labelled in Fig. 4b and their intensity ratios are plotted versus potential in Fig. 4c.

Fig. 4 A Li–S cell with 0.5 M LiTFSI–PY₁₃–FSI as the electrolyte cycled at a rate of C/48. (a) Evolution of Raman spectra collected from the electrolyte on the lithium side of the cell; E: electrolyte; (b) the poly-sulfide species detected via Raman spectroscopy with their corresponding potentials, red: discharge; blue: charge; (c) intensity ratios of poly-sulfides vs. potential.

3.3.3 Polysulfide ion mechanism evolution in the Li–S battery using 0.5 mol L⁻¹ LiTFSI in PY₁₃–FSI as the electrolyte. Considering the plots in Fig. 3d and 4c it is possible to describe the evolution of polysulfide species using 1 mol L⁻¹ LiTFSI in PY₁₃–FSI as the electrolyte as shown in Fig. 5.

Fig. 5 Schematic of the reduction process combining the results related to the cathode and lithium metal side.

Obviously at ∼2.20 V, the elemental sulfur is the main species at the cathode side and it is also detected at the same voltage near the lithium metal. During discharge, S₈ is converted to S₈ⁿ⁻ at the cathode. On the lithium side, the amount of S₈ⁿ⁻ suddenly increases at 2.15 V which is caused by two factors: (i) reduction of S₈ to S₈ⁿ⁻ in contact with lithium metal and (ii) the diffusion of S₈ⁿ⁻ from the positive electrode. On further discharge, the S₈ was exhausted at ∼2.05 V on both sulfur and lithium sides. Approaching the end of the S₈ reduction near the lithium metal surface, S₄²⁻ was observed due to subsequent reduction of S₈ⁿ⁻ to smaller poly-sulfides. Disulfides are only detected at the positive electrode below 2.05 V while S₄²⁻ is consumed. It is important to note that below 2.05 V the variation of the relative amounts of the species is small on the lithium metal side.

3.3.4 Development of Raman spectra from the electrolytes near the anode during cycling using 1 mol L⁻¹ LiTFSI in DME–DOL as the electrolyte. The cell with 1 mol L⁻¹ LiTFSI in DOL–DME was also studied to understand the effect of electrolytes on the presence and dissolution of poly-sulfides. The cell was first discharged to 1.6 V, followed by a charge programmed to 2.8 V which was not reached at the first cycle because the time limit set in the program was too short; but the cell did reach 2.8 V during the second cycle. Fig. 6a and b show the Raman spectra obtained from the electrolyte near the Li metal anode during the first two cycles with the band assignments, note that the band at 466 cm⁻¹ is assigned to Whatman filter paper (see Fig. S5, ESI†). Fig. 6a clearly indicates that cycloocta S₈ converted to S₈ⁿ⁻ quickly after the cell was made and diffused to the anode side even before cycling began. This confirms the problem of rapid self-discharge in this widely used liquid electrolyte and shows a clear difference from the ionic liquid where the elemental S₈ band began to appear in the middle of the high voltage plateau and survived until the end of the plateau. After discharge began, the S₈ⁿ⁻ band grew slightly at first and then faded away. Similar to the case of S₈ⁿ⁻, bands ascribed to S₄²⁻ were observed before discharge began and persisted through the high voltage plateau. As discharge went on, all the bands became undetectable from the beginning of the 2.1 V plateau to the end of discharge. This may be due to the poor solubility of short chain poly-sulfides (Li₂S₂ or Li₂S) in the electrolyte, as well as the rising fluorescence contributing to the high background of the spectra. The charge step was started as soon as discharge reached 1.6 V; no Raman bands were observed from 1.6 to 2.26 V (∼middle of the charge plateau). After that, the S₄²⁻ and S₈ⁿ⁻ bands were evident and grew. At the end of charge, S₈ⁿ⁻ and S₄²⁻ existed in the electrolyte with S₈ⁿ⁻ as the dominant species.

Fig. 6 A Li–S cell with 1 M LiTFSI–DOL–DME as the electrolyte cycled at C/60. The Raman spectra were collected from the electrolyte on the lithium side. Evolution of Raman spectra during (a) first cycle; _*:separator; (b) second cycle; dotted lines represent more spectra; (c) the poly-sulfide species detected via Raman and their corresponding potentials. Red: discharge; blue: charge; green: constant voltage; (d) relative intensities of polysulfides vs. potential.

In order to verify whether other reduction products can be detected in the electrolyte at the end of discharge, the second cycle was discharged to 1 V and kept at this voltage for 5 hours before charge was started. Similar to the first cycle, no Raman bands were observed between the low voltage plateau (2.1 V) of the discharge step and the first half of the charge plateau (2.3 V). Subsequently, S₄²⁻ and S₈ⁿ⁻ coexisted before the voltage reached ∼2.4 V. Starting at ∼2.4 V, the intensities of the S₄²⁻ and S₈ⁿ⁻ bands declined quickly, accompanied by fast formation of S₈, suggesting the electrochemical reactions:

2Li₂S₄ → Li₂S₈ + 2Li⁺ + 2e⁻ (EC-5)

Li₂S₈ → S₈ + 2Li⁺ + 2e⁻ (EC-6)

Only S₈ was seen at the end of charge and its amount was almost unchanged during the 5 hour hold at 2.8 V, see Fig. 6b. The detection of elemental sulfur near the lithium metal at the end of charge (Fig. S6, ESI†) indicates that it plays an important role in lowering battery capacity and cyclability because (1) the sulfur can deposit on the lithium anode and change its surface morphology and conductivity; (2) the elemental sulfur corrodes the anode by reacting with lithium metal and forming thermodynamically preferred phase Li₂S₈;⁴⁴ (3) the soluble Li₂S₈ formed reacts with lithium or disproportionates into other poly-sulfides and then reacts with lithium, which corrodes lithium further and forms resistive Li₂S on the surface;^49,50 (4) the soluble polysulfides formed enhance the shuttle reaction. All these effects increase the cell impedance and thus reduce the capacity and shorten the cycle life. The observation of elemental sulfur and polysulfides in the electrolyte besides lithium confirms most directly that the deterioration of the lithium anode and shuttle reactions are the important mechanisms for the capacity fade of the battery. The cycling curve and the polysulfide species detected in this cell are plotted in Fig. 6c and the variation of intensity ratios of the species versus potential is shown in Fig. 6d.

The poly-sulfide species detected in the two electrolytes near lithium metal are the same, i.e., S₈ⁿ⁻ and S₄²⁻ due to the low donor number and low dielectric constant of both electrolytes,^13,38,51 but they were observed at different stages of cycling indicating the differences in the kinetics of poly-sulfide formation and diffusion in different electrolytes (Fig. S7, ESI†). Examining Fig. 4 and 6, it is clear that for the cells containing DOL–DME, the reactions of S₈ → S₈²⁻ and S₈²⁻ → S₄²⁻ as well as the migration of S₈²⁻ and S₄²⁻ to the lithium side happened as soon as the cells were assembled and the S₈²⁻ and S₄²⁻ were observed before cycling began (though in Fig. 6d, the intensity ratio of S₄²⁻ is 0 at 2.35 V, because at this moment, no S₄²⁻ band at ∼200 cm⁻¹ was seen, but other S₄²⁻ bands were detected as shown in Fig. 6a). In contrast, in the cell using LiTFSI–PY₁₃–FSI, no sulfur related species were observed until discharge to ∼2.20 V; while V < 2.20 V, the movement of S₈ species to the lithium side was seen, followed by the ring opening and the breakup of the long chain. Comparing Fig. 4c and 6d, the intensity ratios of S₈ⁿ⁻ in both cells were similar at 2.1 V; but at 2.05 V, the intensity ratio of S₈ⁿ⁻ was close to its value at 2.1 V in the ionic liquid cell whereas it is greatly reduced in the ether-based electrolyte accompanied by a quick increase of S₄²⁻. This slow kinetics of poly-sulfide formation in the ionic liquid cell is mainly due to the higher viscosity and lower poly-sulfide solubility in it in comparison to DOL–DME,^52,53 which is beneficial to minimise the loss of active material and reduce the shuttle reaction. But this advantage is overridden by the low ionic mobility in the ionic liquid owing to its high viscosity. Our cyclability tests confirmed this by showing that the cells using PY₁₃–FSI have lower capacity and faster capacity fade (Fig. S8, ESI†) than those using DOL–DME.

4. In situ X-ray diffraction measurements

The phases formed during cycling were monitored by in situ XRD. Fig. 7a shows the plots of the cell potential as a function of time, and Fig. 7b shows the evolution of XRD spectra of a cell using 1 mol L⁻¹ LiTFSI–DOL–DME as the electrolyte, with addition of 2 vol% LiNO₃ to reduce the shuttling effect.^33,54 The pristine cathode contained α-sulfur (PDF: 01-078-1888) having an orthorhombic crystal lattice with a space group Fddd (Fig. S2, ESI†). The discharge started at 2.43 V, and α-sulfur was undetectable at 2.34 V approaching the end of the high voltage plateau suggesting that the sulfur structure became non-crystalline quickly. A broad peak centred at 2θ ∼ 26.9° started to form in the second half of the low voltage plateau and grew thereafter; it reached the maximum when the cell was held at a constant voltage of 1.5 V. This broad peak is attributed to Li₂S; its broadness indicated that the crystallinity of the phase was poor. The average crystal size estimated with the Scherrer formula was ∼10 nm. The Li₂S peak was almost unchanged until the charge voltage reached ∼2.3 V at the middle of the charge plateau; it gradually diminished and completely vanished at ∼2.36 V. Only one small peak was evident at the end of the first charge, whereas more peaks were observed at the end of the second and third charge cycles. Although some peaks appeared at 24.02° and 35.02° it is uncertain whether β-sulfur (PDF: 04-007-2070) was formed at this moment considering that other high intensity peaks were not detected in the spectrum. The peaks formed at the end of the 3rd charge and the following constant voltage period were indexed to β-sulfur with a monoclinic structure (S.G. P2₁/c), which was also observed by Walus et al.¹⁵ The β-sulfur is reported to be stable at temperatures > 95 °C (ref. 55) whereas it is obtained in the Li–S cell at room temperature. It is possible that the nucleation of β-sulfur from the long chain poly-sulfides dissolved in the electrolyte is more energetically preferred than the α-phase; as a result, β-sulfur is formed first as a metastable phase, as schematically shown in Fig. 8. This phase identification is complementary to our Raman experimental results.

Fig. 7 A Li–S cell using 1 M LiTFSI–DOL–DME–2% LiNO₃ as the electrolyte cycled at a rate of C/40. (a) Discharge–charge curve; (b) evolution of XRD spectra showing the formation of β-sulfur at the end of charge. Red: discharge; blue: charge; green: constant voltage.

Fig. 8 A schematic diagram illustrating the reaction energies in the formation of β and α sulfur.

5. Conclusions

The poly-sulfide species formed at different voltages during galvanostatic cycling have been identified. For the cell using 0.5 mol L⁻¹ LiTFSi–PY₁₃–FSI electrolyte, the S₈ ring and S₈ⁿ⁻ coexisted on the first reduction plateau, whereas S₄²⁻, S₂²⁻and S²⁻ were formed on the second plateau at almost the same potential. At the end of discharge, S²⁻ and other unknown species were observed, with the former as the end product of the reduction and the latter probably related to the disproportionation reaction. A discharge mechanism is proposed for the cell using the ionic liquid which provides fundamental information on the discharge process which is an important step for the amelioration of Li–S batteries. The evolution of polysulfide species during charge is the reverse process to discharge despite the differences of their corresponding potentials. The end product of charge is cycloocta S₈ identified with a monoclinic structure via XRD. By examining the poly-sulfide species in the electrolyte phase near lithium in both PY₁₃–FSI and DOL–DME, we found that the species in the two electrolytes are the same, but the kinetics of sulfur ring opening and breakup into short chain poly-sulfides are slower in 0.5 mol L⁻¹ LiTFSI–PY1₃–FSI than in 1 mol L⁻¹ LiTFSI–DOL–DME, which is beneficial for reducing the shuttling effect. On the other hand, the low mobility of polysulfides in the liquid electrolyte lowers the utilization of active sulfur in the battery, preventing the participation of soluble polysulfide in electrochemical reactions in the cathode and reducing the cyclability. The detection of S₈ in the electrolyte near the lithium electrode confirms that the parasitic reactions on the Li metal anode involve the direct self-discharge reduction of elemental sulfur, as well as soluble polysulfide species, which are related to lithium corrosion and the loss of active material in the Li–S battery.

Conflict of interest

The authors declare no competing financial interest.

References

1. N. Jayaprakash, J. Shen, S. S. Moganty, A. Corona and L. A. Archer, Angew. Chem., 2011, 123, 6026.
2. R. Xu, J. Lu and K. Amine, Adv. Energy Mater., 2015, 5, 1500408.
3. M. A. Pope and I. A. Aksay, Adv. Energy Mater., 2015, 5, 1500124.
4. S. E. Cheon, K. S. Ko, J. H. Cho, S. W. Kim, E. Y. Chin and H. T. Kim, J. Electrochem. Soc., 2003, 150(6), A800.
5. C. Li, H. Zhang, L. Otaegui, G. Singh, M. Armand and L. M. Rodriguez-Martinez, J. Power Sources, 2016, 326, 1.
6. A. Manthiram, Y. Fu and Y. S. Su, Acc. Chem. Res., 2013, 46(5), 1125.
7. J. A. Dean, Lange's Handbook of Chemtry, McGRAW-HILL, New York, 15th edn, 1999.
8. P. P. R. M. L. Harks, C. B. Robledo, T. W. Verhallen, P. H. L. Notten and F. M. Mulder, Adv. Energy Mater., 2016, 1601635.
9. K. Zhang, L. Wang, Z. Hu, F. Cheng and J. Chen, Sci. Rep., 2014, 4, 6467.
10. D.-W. Wang, Q. Zeng, G. Zhou, L. Yin, F. Li, H.-M. Cheng, I. R. Gentle, G. Qing and M. Lu, J. Mater. Chem. A, 2013, 1, 9382.
11. C. Barchasz, F. Molton, C. Duboc, J.-C. Leprêtre, S. Patoux and F. Alloin, Anal. Chem., 2012, 84, 3973.
12. Q. Wang, J. Zheng, E. Walter, H. Pan, D. Lv, P. Zuo, H. Chen, Z. D. Deng, B. Y. Liaw, X. Yu, X. Yang, J.-G. Zhang, J. Liu and J. Xiao, J. Electrochem. Soc., 2015, 162(3), A474.
13. Q. Zou and Y.-C. Lu, J. Phys. Chem. Lett., 2016, 7, 1518.
14. M. Cuisinier, C. Hart, M. Balasubramanian, A. Garsuch and L. F. Nazar, Adv. Energy Mater., 2015, 5, 1401801.
15. S. Walus, C. Barchasz, R. Bouchet, J.-C. Leprêtre, J.-F. Colin, J.-F. Martin, E. Elkaïm, C. Baehtz and F. Alloin, Adv. Energy Mater., 2015, 5, 1500165.
16. N. A. Cañas, S. Wolf, N. Wagner and K. A. Friedrich, J. Power Sources, 2013, 226, 313.
17. A. Paolella, W. Zhu, H. Marceau, C.-S. Kim, Z. Feng, D. Liu, C. Gagnon, J. Trottier, A. Guerfi, P. Hovington, A. Vijh, G. P. Demopoulos, M. Armand and K. Zaghib, J. Power Sources, 2016, 325, 641.
18. J. Nelson, S. Misra, Y. Yang, A. Jackson, Y. Liu, H. Wang, H. Dai, J. C. Andrews, Y. Cui and M. F. Toney, J. Am. Chem. Soc., 2012, 134, 6337.
19. M. U. M. Patel, R. Demir-Cakan, M. Morcrette, J.-M. Tarascon, M. Gaberscek and R. Dominko, ChemSusChem, 2013, 6, 1177.
20. H. Marceau, C. S. Kim, A. Paolella, S. Ladouceur, M. Lagacé, M. Chaker, A. Vijh, A. Guerfi, C. M. Julien, A. Mauger, M. Armand, P. Hovington and K. Zaghib, J. Power Sources, 2016, 319, 247.
21. Y. Gorlin, A. Siebel, M. Piana, T. Huthwelker, H. Jha, G. Monsch, F. Kraus, H. A. Gasteiger and M. Tromp, J. Electrochem. Soc., 2015, 162, A1146.
22. M. Cuisinier, P.-E. Cabelguen, S. Evers, G. He, M. Kolbeck, A. Garsuch, T. Bolin, M. Balasubramanian and L. F. Nazar, J. Phys. Chem. Lett., 2013, 4, 3227.
23. K. H. Wujcik, T. A. Pascal, C. D. Pemmaraju, D. Devaux, W. C. Stolte, N. P. Balsara and D. Prendergast, Adv. Energy Mater., 2015, 5, 1500285.
24. H. Yao, K. Yan, W. Li, G. Zheng, D. Kong, Z. W. Seh, V. K. Narasimhan, Z. Liang and Y. Cui, Energy Environ. Sci., 2014, 7, 3381.
25. J.-T. Yeon, J.-Y. Jang, J.-G. Han, J. Cho, K. T. Lee and N.-S. Choi, J. Electrochem. Soc., 2012, 159, A1308.
26. M. Hagen, P. Schiffels, M. Hammer, S. Dorfler, J. Tubke, M. J. Hoffmann, H. Althues and S. Kaskel, J. Electrochem. Soc., 2013, 160, A1205.
27. J. Hannauer, J. Scheers, J. Fullenwarth, B. Fraisse, L. Stievano and P. Johansson, ChemPhysChem, 2015, 16, 2755.
28. H.-L. Wu, L. A. Huff and A. A. Gewirth, ACS Appl. Mater. Interfaces, 2015, 7, 1709.
29. Y. Diao, K. Xie, S. Xiong and X. Hong, J. Electrochem. Soc., 2012, 159, A1816.
30. M. Wild, L. O'Neill, T. Zhang, R. Purkayastha, G. Minton, M. Marinescu and G. J. Offer, Energy Environ. Sci., 2015, 8, 3477.
31. N. N. Greenwood, Spectroscopic Properties of Inorganic and Organometallic Compounds, The Chemical Society, Burlington House, London, 1975, vol. 8, p. 245.
32. G. Gautier and M. Debeau, Spectrochim. Acta, Part A, 1974, 30, 1193.
33. S. S. Zhang, Electrochim. Acta, 2012, 70, 344.
34. L. Suo, Z. Fang, Y.-S. Hu and L. Chen, Chin. Phys. B, 2016, 25, 016101.
35. J.-C. Lassègues, J. Grondin, C. Aupetit and P. Johansson, J. Phys. Chem. A, 2009, 113, 305.
36. F. P. Daly and C. W. Brown, J. Phys. Chem., 1975, 79, 350.
37. F. P. Daly and C. W. Brown, J. Phys. Chem., 1973, 77, 1859.
38. Y. C. Lu, Q. He and H. A. Gasteiger, J. Phys. Chem. C, 2014, 118, 5733.
39. G. J. Janz, J. R. Downey Jr, E. Roduner, G. J. Wasilczyk, J. W. Coutts and A. Eluard, Inorg. Chem., 1976, 15, 1759.
40. O. El Jaroudi, E. Picquenard, A. Demortier, J.-P. Lelieur and J. Corset, Inorg. Chem., 2000, 39, 2593.
41. O. El Jaroudi, E. Picquenard, A. Demortier, J.-P. Lelieur and J. Corse, Inorg. Chem., 1999, 38, 2394.
42. O. El Jaroudi, E. Picquenard, N. Gobeltz, A. Demortier and J. Corset, Inorg. Chem., 1999, 38, 2917.
43. K. Zhang, L. Wang, Z. Hu, F. Cheng and J. Chen, Sci. Rep., 4, 6467,  DOI:10.1038/srep06467.
44. R. S. Assary, L. A. Curtiss and J. S. Moore, J. Phys. Chem. C, 2014, 118, 11545.
45. Z. Yuan, H.-J. Peng, T.-Z. Hou, J.-Q. Huang, C.-M. Chen, D.-W. Wang, X.-B. Cheng, F. Wei and Q. Zhang, Nano Lett., 2016, 16, 519.
46. Z. Černosek, J. Holubova, E. Černoskova and A. Ruzicka, J. Non-oxide Glasses, 2009, 1, 38.
47. G. Gautier and M. Debeau, Spectrochim. Acta, Part A, 1976, 32, 1007.
48. L. Suo, O. Borodin, T. Gao, M. Olguin, J. Ho, X. Fan, C. Luo, C. Wang and K. Xu, Science, 2015, 350, 938.
49. J. Zheng, D. Lv, M. Gu, C. Wang, J.-G. Zhang, J. Liu and J. Xiao, J. Electrochem. Soc., 2013, 160, A2288.
50. H. Kim, J. T. Lee and G. Yushin, J. Power Sources, 2013, 226, 256.
51. E. Izgorodina, M. Forsyth and D. Macfariane, Phys. Chem. Chem. Phys., 2009, 11(14), 2452.
52. Q. Zhou, W. A. Henderson, G. B. Appetecchi, M. Montanino and S. Passerini, J. Phys. Chem. B, 2008, 112, 13577.
53. L. Suo, Y.-S. Hu, H. Li, M. Armand and L. Chen, Nat. Commun., 2013, 4, 1481.
54. X. Liang, Z. Wen, Y. Liu, M. Wu, J. Jin, H. Zhang and X. Wu, J. Power Sources, 2011, 196, 9839.
55. B. Meyer, Chem. Rev., 1976, 76, 367.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6se00104a

---