In-Situ Raman Investigation of Polysulfide Formation in Li-S Cells

Gas diffusion layers (GDL) or Ni foam were used as substrate for the growth of CNTs by a CVD process. The detailed CNT coating procedure and the sulfur infiltration is described in Refs. 47–49. The rectangular electrodes had a size of approximately 0.9 cm². The polypropylene separator with about 2.0 cm² overlapped the electrodes.

The stainless steel test cell housing for the in-situ cell consists of a rectangular chamber with two windows and cylindrical parts at the ends. Stainless steel contacts with spring guides sealed the cell from both sides with two O-rings. The latter additionally prevented electrical contact, isolating the cell housing from the stainless steel contacts. Four microscope glass slides with 0.125 mm thickness were cut and put inside the cell housing and glued to the latter from outside. By this way the electrolyte could not get in contact with the glue. The four glass slides also isolated the electrode stack from the housing. The rectangular electrodes and separator cut by a special designed punch hole were placed in the chamber and contacted by stainless steel bars. A guided spring led to a defined pressure on the electrode stack. The test cell is shown in Fig. 2.

Zoom In Zoom Out Reset image size

All tests were conducted against metallic lithium (SA, 99.9% purity) as counter electrode. The lithium was scraped with a ceramic knife to remove surface layers and afterwards pressed through a calendar to obtain a homogeneous and reproducible surface. Commercial polypropylene separators with a thickness of around 25 μm and an average pore size of around 40 nm were used. As electrolyte 0.7 M lithium bis(trifluoromethanesulfonyl)imide (LiTFSI, SA, 99.95%) in DME:DIOX (2:1, v:v) (SA, anhydrous, 99.5% and 99.8% with 75 ppm butylated hydroxytoluene (BHT inhibitor) + 0.25 M/L LiNO₃ (ABCR 99.98%) was used. The assembly of the test cells was done in an argon-filled glove box (MBraun) with an O₂ and H₂O content below 1 ppm.

• The used cycle station was a Solartron ModuLab with "Cell Test" software. The Raman in-situ test cell was cycled between 1.5 and 3.0 V with a constant current of 0.56 mA cm⁻² (C/7.5). The sulfur load was 5.1 mg cm⁻².
• Raman measurements were made with a Horiba Jobin Yvon HR800. The in-situ test cell was fixed on a sample holder connected to the movable sample table which could be adjusted with 0.1 μm accuracy. The in-situ Raman spectra (Fig. 7) always show the Raman lines of exactly the same measurement point in the vicinity of the electrode surface such that Raman signals originating from species solvated in the electrolyte can be observed together with Raman signals originating from the electrode surface itself. Raman signals originating from the electrode surface include those bands due to fully oxidized S₈ and those bands due to insoluble products precipitating on the electrode, namely Li₂S and Li₂S₂. A red HeNe laser with λ₁ = 632.8 nm and 20 mW with 100x objective was used. Raman measurements of the high purity solvents (+ polysulfides), salts, electrolytes and sulfur were made in a cuvette with Teflon (PTFE) fiber sealed screw thread.

Zoom In Zoom Out Reset image size

The purities were: DME (SA, anhydrous, 99.5%), DIOX (SA, anhydrous 99.8% with 75 ppm BHT inhibitor), TEGDME (SA, 99%), tetrahydrofurane (THF, Roth, 99.9%) LiTFSI (SA, 99.95%), lithium trifluoromethanesulfonate (LiTFS, SA, 99.995%), Li₂S (Alfa Aesar, 99.9%, metals basis), sulfur (SA, >99.5%, purum p.a.), LiNO₃ (ABCR 98%; anhydrous, metals basis).

Polysulfide reference solutions ("Li₂S₈", "Li₂S₆" and "Li₂S₄") were prepared by mixing Li₂S and S₈ powder in the particular stoichiometric proportions. The powder mixture was then added into an air sealed and solvent (TEGDME, DME, DIOX, THF, electrolyte) filled cuvette, which included a small magnet stirrer. After one night of stirring, Raman spectra were recorded (Fig. 5). The resulting solutions provide reference systems containing mixtures of polysulfides in equilibrium with the intended overall stoichiometry. Specifically, the reference solutions contain anionic polysulfide species with different chain lengths resulting from disproportionation reactions in equilibrium.³⁰ It is thus not possible to prepare solutions containing a specific polysulfide anion with a single, uniquely defined chain length. Reference Raman spectra attributable to polysulfide dianions (or monoanions) with a unique chain length are obtainable by theoretical means only. In the present work, theses reference spectra are obtained by density functional (DFT) calculations of the respective mono- and dianions.

Zoom In Zoom Out Reset image size

The interpretation of the experimental in-situ Raman spectra in terms of contributions attributable to individual reduced sulfur species in the electrolyte solution or due to solid-state precipitates at the electrode surface poses several difficulties: As reviewed in,³⁰ pronounced differences in Raman-spectra can be expected depending on the chain-length of the polyanions but additionally due to different rotational isomers. Experimental Raman data for individual polysulfide anions used for peak-assignment by comparison have often been obtained from crystalline samples. In these solid-state spectra, Raman signals originating from lattice vibrations are observed together with peaks due to the typical S-S stretching or bending and torsional modes of the anion. Additionally, the local symmetry around the anion as well as the polarization of S-S bonds induced by counterions in the lattice can decisively shift the Raman peaks and even lead to band splittings not observed in solution as shown by Jaroudi and coworkers.^43,50 In some cases, lattice effects even result in anion conformations which are unfavorable in solution. One example are the rotational isomers of S₅²⁻ which have local C_S (Na₂S₂) or C₂ (K₂S₂) symmetries, depending on the counterion in the lattice.³⁰ The latter isomers can be distinguished by their respective Raman fingerprints. It should be noted that polysulfide anions with a precisely defined chain length cannot be prepared in solution for reference measurements since disproportionation and redox-reactions result in a distribution of several polysulfide chain lengths. The resulting Raman spectra are thus not representative of a polysulfide anion with a unique chain length corresponding to the overall stochiometry.

In the present work, theoretical Raman spectra based on quantum-chemical calculations on reduced polysulfide di-anions S_n²⁻ as well as on the sulfur radical mono-anions S_n⁻ with n = 3–8 are used to establish an independent means to aid the assignment of the in-situ spectra. Since both Li₂S and Li₂S₂ are merely insoluble in the electrolyte used in the present study, the corresponding Raman spectra measured at the end of the charge cycle with short-chain sulfides (n≤2) are to be compared to the respective solid-state spectra obtained from crystalline samples. In both cases, lattice effects will influence the Raman signals as described above which is not the case for the solvated polysulfide species with n≥3.

Quantum-chemical calculations on sulfur-rich compounds including polysulfide di-anions and radical mono-anions were reviewed by Wong.⁵³ In solution, polysulfides predominantly form linear, unbranched chains. The symmetry of the latter varies between C₁, C₂ and C_s, depending on the S-S-S-S torsional angles along the chain, which are typically close to 110°. The order of the sign of the torsional angles along the chain defines the 'motif' of the anion which determines the Raman band positions, intensities and selection rules.³⁰ It is well documented that DFT-methods, especially those involving hybrid functionals yield reliable equilibrium structures and vibrational frequencies for sulfur-rich compounds, even though care must be taken when interpreting calculations with insufficiently large basis sets.

All our calculations were performed with the Gaussian03 program package.⁵¹ For each individual chain length from n = 3 to n = 8 (anions and radical anions respectively), we first generated starting structures for all diastereomeric conformations (motifs) resulting in distinct vibrational spectra based on symmetry considerations^†. The total number of diastereomeric conformers increases swiftly with the chain length: For the shorter chains such as S₃²⁻ and S₄²⁻ there is only one relevant conformer, S₅²⁻ has two diastereomeric conformers. S₆²⁻, S₇²⁻ and S₈²⁻ have 3, 6 and 10 conformers respectively. The initial molecular structures were fully optimized at the B3PW91/6–311G(2df,p) level of theory in vacuum, which also includes diffuse functions as found essential by Ref. 53. This level of theory was also used to calculate the vibrational wavenumbers, zero-point corrections to the energy and thermal energies including the Gibbs free energy. Following the initial refinement in vacuum, we modeled solvent effects on the vibrational spectra for THF within the framework of the polarizable continuum model (PCM) by further refinement using the B3PW91/6–311G(2df,p) + PCM (THF) level of theory as implemented in Gaussian03. Within the PCM model, solvation is simulated as an infinite polarizable continuum (the solvent) surrounding a molecular-shaped cavity (the solute). PCM produces the sum of the electronic energy of the solute and the Gibbs free energy of the continuum solvent.⁵²

Since the total Raman spectrum for an individual polysulfide chain with fixed chain-length represents a superposition of the Raman spectra of all individual motifs (conformers) present in solution, the Gibbs free energy of all contributing conformers at 298.15 K was evaluated. The total Raman spectrum for an individual polysulfide chain length is then obtained as Boltzmann-averaged superposition of the Raman spectra of all contributing conformers according to their respective free energy differences. Following Ref. 54, all DFT calculated peak positions are uniformly corrected by a scaling factor of 0.9573 for the B3PW91 functional. Fig. 3 and Fig. 4 show the resulting theoretical Raman spectra for the di-anions and radical mono-anions respectively. Raman lines are uniformly broadened by a Lorentzian lineshape with a finite width at half maximum (FWHM) of 10 cm⁻¹ to aid the visualization of the spectra. Individual Raman bands of all contributing conformers are indicated by vertical lines.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Considering the free energy differences between individual conformers (motifs) in THF solution, it was discovered that the fully stretched, helical chains (chainlike twisted structures) dominate the Raman spectra of the small chain lengths polysulfide di-anions S_n²⁻ with n≤6 at 298.15 K. For di-anions with longer chain lengths, such as S₇²⁻ and S₈²⁻, significant contributions from other conformers were found: In the case of S₇²⁻, conformations other than the fully stretched one contribute approximately 8.7% to the overall Raman spectrum while for S₈²⁻, such contributions amount to 22% of the final spectrum. The contribution of non-helical chains to the Raman-spectra is even more pronounced for the radical mono-anions S_n⁻, especially for the longer chain radical mono-anions. The fully stretched conformer becomes less stable compared to other conformers for S_n⁻ with n≥5 in THF solution. It is clearly visible that the solvent modeled within the PCM-framework has an effect on the Raman lines which can range from some slight shifts in peak positions (i.e. in S₈⁻) to the resolution of accidentally degenerated Raman lines.

For the polysulfide di-anions S_n²⁻, three groups of Raman-peaks related to S-S stretch vibrations can be distinguished in the region 350 cm⁻¹ to 500 cm⁻¹. These modes involve the central S-S linkage and terminal S-S linkages respectively (Fig. 3) with the terminal modes always shifted to higher wavenumbers. Interestingly, a systematic variation in Raman intensities was found for the symmetric S-S vibration belonging to the central S-S linkage which is predicted to be rather intense for chains with an even number of atoms. The corresponding peak positions for the central S-S vibration are 413 cm⁻¹ (S₄²⁻), 357 cm⁻¹ (S₆²⁻) and 362 cm⁻¹ (S₈²⁻) in THF solution. Symmetric and antisymmetric terminal S-S modes are observed at higher wavenumbers. For the short lengths radical mono-anions S_n⁻ (Fig. 4) we note a systematic shift of the symmetric and antisymmetric terminal S-S modes toward higher energies. Compared to the di-anions, the Raman lines of the radicals are observed at higher wavenumbers in the region above 500 cm⁻¹ in THF solution. The corresponding peak positions for the highest energy S-S vibrations above 500 cm⁻¹ are 558 cm⁻¹ (S₄⁻), 552 cm⁻¹ (S₅⁻) and 542 cm⁻¹ (S₆⁻).

An overview of Raman assignments for the various polysulfide anions and radicals is given in Table I. together with other Raman bands important for the interpretation of the Li-S in-situ data. The table includes Raman signals attributable to solvents commonly used as a basis for electrolytes and data for crystalline polysulfide samples. As discussed above, the novel in-situ technique developed in the present work records Raman signals originating from polysulfides, electrolyte and as well as signals due to solid state precipitates on the electrode (Li₂S, Li₂S₂). The values provided in Table I can be used to interpret the measured spectra and give a good overview concerning Raman measurements in Li-S cells.

It should be noted that the published values for the "S₂²⁻" peaks in Table I refer to Raman signals obtained from crystalline Li₂S₂ in which the Raman bands arise from lattice vibrations.¹⁵ The difference between the calculated Raman peak positions and the experimental results reported are thus mainly due to solid-state effects for this species. As Li₂S₂ and Li₂S are insoluble in the electrolyte used, all comparisons should be based on the crystalline Raman band positions for Li₂S₂ and Li₂S respectively. In solution or in vacuum, the latter ion has no Raman spectrum since there is no vibrational mode. Published values indicate peak positions at 174 cm⁻¹ and 514 cm⁻¹ for crystalline Li₂S₂¹⁵ and at 378 cm⁻¹ for crystalline Li₂S,¹⁶ but the Li₂S Raman peak intensity is rather low and can almost not be identified in a discharged sulfur cathode.¹⁶ Nevertheless it can be questioned whether the examined species in ref. 15 is pure Li₂S₂ for it can be assumed that there are additional or maybe even different polysulfide species at the electrode. Yeon et al. also observed a peak above 500 cm⁻¹ in discharged state¹⁶ which also can be seen in the in-situ spectra in Fig. 7. This Raman line is assigned to S₃⁻, perfectly fitting to our DFT calculations of this species.

Since Li₂S and Li₂S₂ are insoluble in the electrolyte, the corresponding experimental solid-state spectra must be taken as reference for the Raman peak assignment. Table I includes calculated Raman peak positions for S₂²⁻ both in vacuum (560 cm⁻¹) and in THF solution (428 cm⁻¹) only for completeness.

Comparing the Raman lines of the calculated polysulfide mono- and dianions (S_n⁻, S_n²⁻ with n>2) with literature results of alkali metal or alkaline earth metal polysulfides one can see the impact of the coordinating cation and as well the solvent. The calculated Raman lines of S₄²⁻ are at 462, 445, 413, 212, 168 and 89 cm⁻¹ in THF. For aqueous Na₂S₄, peaks are reported at 484, 446, 410, 256, 194 and 144 cm⁻¹.⁴² Polarization in a solvent such as THF can lead to shifts in Raman band positions as evidenced by comparing the calculated vacuum results to the THF case (see Fig. 3, 4). Typically, the Raman-line in THF is found at slightly higher wavenumbers due to polarization effects in the solvent which tend to increase S-S bond strength. Additionally, the relative peak intensities are also affected by the solvent.

As mentioned above, the electrolyte can have a dramatic effect on the reaction mechanism in Li-S cells which influences the distribution of polysulfide chains in solution. Comparing different electrolytes, the overall Raman spectra are thus different due to different amounts of individual polysulfides or different sulfur and polysulfide solubility.

The effect of the solvent on polysulfide Raman peak positions – considering one individual chain length – is not so pronounced. Comparing the calculated THF and vacuum spectra in Fig. 3, we find that these spectra are overall very similar. There are variations in intensity and also some shifts in wavenumbers which we estimate to be below 15–20 cm⁻¹ at most. The notable exception is S₂²⁻ as discussed above. Here, the discrepancy is larger but Li₂S₂ is not solvable in the used electrolyte, so comparisons should be made with the solid-state spectra. For the calculated wavenumbers, we note that one of the important parameters for the PCM model is the dielectric constant of the solvent, which does not vary to a large degree for different ether based solvents. We thus conclude that the solvent effect on peak positions is only minor compared to other uncertainties and calculations of Raman-spectra. Therefore THF can be used as a good reference for ether-based electrolytes regarding the calculations. This could also be confirmed by experimental data of polysulfides in various ether based solvents (Fig. 5). There a characteristic polysulfide induced peak at around 535 cm⁻¹ could be identified with different intensities depending on the solvent. Therefore one can suppose that the effect of the specific ether used on the polysulfide Raman line positions is comparably low and taken into account within the calculations.

Nevertheless, the electrolyte composition does have an effect on the overall Raman spectra resulting from all species within solution plus the solid state precipitates. This is because the electrolyte solvents influence the reaction mechanisms and the distribution between different polysulfide chain lengths due to different polysulfide and sulfur solubility. Additionally, the electrolyte solvents also contribute with their own Raman lines to the overall spectrum.

The reaction mechanism in Li-S cells seems to be dependent on the electrolyte solvents² as discussed above. Consequently the solvents should have a high impact on polysulfide creation and decomposition. Therefore various measurements of typical Li-S electrolyte solvents with polysulfides were made, to allow a more precise interpretation of the Li-S reaction mechanism. Fig. 5f shows the corresponding Raman measurements of various standards including solvents like THF, TEGDME, DME, DIOX and salts like LiTFSI, LiTFS and LiNO₃ which are frequently used in Li-S electrolytes (compare Fig. 1). In the Raman spectrum of LiTFSI in DME:DIOX (+LiNO₃) the DIOX peaks at 947 cm⁻¹ and 967 cm⁻¹ and the DME peaks at 364 cm⁻¹, 820 cm⁻¹ and 847 cm⁻¹ are clearly visible. LiTFSI peaks in the electrolyte can be identified at 278 cm⁻¹, 749 cm⁻¹ and between 300–400 cm⁻¹. Interestingly LiNO₃ peaks cannot be detected in the electrolyte with the exception of a possible peak at 740 cm⁻¹, which could have been formed by LiTFSI alone or together with LiNO₃. Nevertheless the sharp LiNO₃ peaks at 238 cm⁻¹ and 1071 cm⁻¹ are clearly not visible in the electrolyte, despite the high LiNO₃ concentration of 0.25 M/L.

The LiTFS in TEGDME:DIOX shows again the typical DIOX peaks and indicates TEGDME peaks between 800–900 cm⁻¹. The LiTFS Raman lines seem to have been moved to smaller wave numbers (320 → 314 cm⁻¹, 778 → 758 cm⁻¹ and possibly 1077 → 1044 cm⁻¹) showing again the strong impact of solvation.

To get an impression of characteristic Raman polysulfide lines, a stoichiometric powder mixture of Li₂S and S₈ to obtain "Li₂S₈, Li₂S₆ and Li₂S₄" was added to various solvents or electrolytes. The polysulfide concentration was between 5 mM L⁻¹ and 50 mM L⁻¹ to allow a complete dissolution of the solid powders. Different polysulfide compositions led to different solvent colors with yellow for Li₂S₄, orange for Li₂S₆ and red for Li₂S₈. It was not expected to obtain polysulfides with exact Li₂S₈, Li₂S₆ and Li₂S₄ chain length but the different amount of sulfur should at least influence the equilibrium of the various polysulfides, which are likely to be formed in the solvents. The corresponding Raman spectra in dependency of the solvents are shown in Fig. 5a–5e. In all polysulfide/solvent spectra the Raman peaks of the base solvent/electrolyte are clearly visible. Additionally all spectra showed a peak between 520–550 cm⁻¹ which can be attributed to polysulfides, likely through S-S vibration, since it was not visible in the pure solvents. Nevertheless this peak could not clearly be identified, because there were several mono-anion induced peaks and the S₂²⁻ peak nearby. Nevertheless, it seems to be implausible to obtain a S₂²⁻ peak in solutions with sulfur excess (stoichiometric Li₂S₈) without electrochemical reduction. Most likely the observed peak is created by S₃⁻, which would therefore have been created in all solvents even with excess of sulfur.

A peak at around 480 cm⁻¹ close to the S₈ standard peak could be seen in polysulfide- THF and DIOX but not in DME, TEGDME and electrolyte. Close to this peak are several di-anion Raman lines like S₅²⁻, S₇²⁻ and S₈²⁻ but also the S₈ peak at 472 cm⁻¹. Another peak at 219 cm⁻¹ is also very clearly visible in polysulfide-THF and -DIOX but only very little in DME, TEGDME and the electrolyte. Again the S₈ peak and several di-anion peaks like S₅²⁻ and S₈²⁻ are close to this position but no mono-anion peaks were calculated for this area. This could lead to the following conclusions: Either the sulfur S₈ did not react with the Li₂S in THF and DIOX, leading to the S₈ Raman signals, or the creation of di-anion polysulfides seems to be somehow inhibited in TEGDME and DME. Most likely the poor S₈ dissolution in DIOX brings forth S₈ Raman signals.¹⁶

The fact that the DME:DIOX electrolyte shows merely no peaks at around 219 cm⁻¹ and 480 cm⁻¹ is remarkable and leads to the suggestion that DME could have a dominating impact over DIOX. Nevertheless, the electrolyte, DME and also TEGDME show a low intensity peak at 219 cm⁻¹ which seems to be present especially if "Li₂S₈" or generally more sulfur was added. This could mean that no more polysulfides were created and that the small excess of S₈ led to a Raman line, corresponding to S₈.

To get an impression how the Raman lines of a binder free CNT-S electrode are affected by addition of the electrolyte and a Li metal anode, the in-situ test cell was assembled with a bare CNT-S electrode with Ni foam current collector (Fig. 6). Subsequently electrolyte and a Li metal anode were added in order to examine possible impacts on the Raman lines.

Zoom In Zoom Out Reset image size

Measurements of the bare CNT-S electrode showed typical S₈ peaks at 152, 219 and 473 cm⁻¹. Additionally there were broad CNT induced peaks at 1330 cm⁻¹ and 1586 cm⁻¹.^55,56 With addition of electrolyte the S₈ peaks shifted to slightly higher wave numbers (154, 220, 439, 474 cm⁻¹) and peaks induced by the electrolyte were visible. With further addition of a lithium metal anode there was a further small shift of the lower wave number S₈ peaks to 155 and 222 cm⁻¹. The electrolyte induced peaks somehow lost some intensity and could not be identified in several positions any more (e.g. 743, 853, 877, 944 cm⁻¹ et cetera).

As a conclusion, the influence of electrolyte and lithium metal on S₈ Raman peaks is rather low and within the range of a few cm⁻¹. Nevertheless, a lithium counter electrode with separator somehow removes some electrolyte induced peaks and starts to create polysulfides through sulfur-electrolyte dissolution and by polysulfide shuttle self-discharge. An explanation for the missing electrolyte induced peaks could be a reaction of the lithium salt with the lithium metal anode, leading to a solid electrolyte interface (SEI) or a reaction with the polysulfides. Nevertheless, the concentration of LiTFSI should have been high enough to leave some additional salt, which should have been identified by Raman.

A carbon structure electrode, coated with CNT and a sulfur load of 5.1 mg cm⁻¹ was assembled vs. a Li metal anode. A polypropylene separator and an excess of 200 μL electrolyte were added. Raman spectra were taken at various cell voltages during charge and discharge. The cell voltage profile with indicated measurement points including information about the reaction mechanism out of the literature^1,57 are shown in Fig. 7a. The discharge spectra are demonstrated in Fig. 7b and the charge spectra in Fig. 7c. The measurement range was limited between 100 and 600 cm⁻¹ in order to shorten the measurement time and to reduce the possibility of laser induced effects on sensible substances. This measurement range is sufficient to discuss polysulfide creation and the reaction mechanism. An interpretation of electrolyte decomposition products would only be possible if the measurement range is increased to much higher wave numbers.¹⁵ This was, however, not the intention of this work.

The open circuit voltage (OCV) before discharge (Fig. 7b) shows typical S₈ peaks at around 150, 220 and 470 cm⁻¹. But compared to previously shown S₈ peaks those peaks were broad, indicating the possible existence of polysulfide di-anions (S₇²⁻, S₈²⁻) and mono-anions (S₅⁻, S₈⁻). After some time of discharge (2.33 V) these peaks were almost completely reduced. Nevertheless, this does not need to be the case for every electrode. Especially with very high sulfur loads (> 10 mg cm⁻¹) it is possible to obtain S₈ peaks even after the end of the first discharge (not shown here). This should be due to bad electrical contact of the isolating sulfur to the electrode carbon network. But even those isolating sulfur spots were reduced in the second discharge, likely by solid-liquid reactions indicating that a completely homogenous sulfur distribution is not necessary to obtain an electrode with high sulfur utilization. More important seems to be a sufficient conductive electrode surface and easy accessibility of the electrode.⁴⁹

With further discharge until the end of discharge a peak between 520–550 cm⁻¹ emerges, reaching its highest intensity at around 2.10 V. The peak position is the same as in Fig. 5 at which the polysulfides were created by mixing Li₂S and S₈. Again the interpretation of this important peak position is difficult, since there are many mono-anion peaks (e.g. S₃⁻, S₆⁻) and the possible Li₂S₂ solid-state peak of the precipitate nearby. Yeon et al.¹⁶ assigned this peak to S₃⁻, created through decomposition of S₄²⁻³⁸ with S₂²⁻ as additional product. This hypothesis can be confirmed by our Raman in-situ measurements, since S₄²⁻ vibrations (445 and 462 cm⁻¹) are very likely to be found in the very broad Raman peaks in this area. Peaks induced due to crystalline Li₂S₂ cannot be identified directly at 514 cm⁻¹ but very likely at 174 cm⁻¹,¹⁵ especially after continuous discharge.

Between 2.29 V and the OCV voltage after discharge two very broad peaks emerge among 340 and 420 cm⁻¹ and 420 and 480 cm⁻¹, reaching their intensity maximum at 2.10 V. These broad peaks are likely be induced by several coexisting polysulfides which should have a medium to short chain length according to the reaction mechanism theory. Regarding the polysulfide calculations in Table I the peak between 340 and 420 cm⁻¹ is likely to be created by a mixture of S₆²⁻, S₇²⁻, S₈²⁻. With further discharge the S₆²⁻, S₈²⁻ induced shoulder between 350 and 370 cm⁻¹ loses its intensity, indicating lower concentrations of these polysulfides. The broad peak between 420 and 480 cm⁻¹ is likely to be created by shorter chain polysulfides like S₃²⁻, S₄²⁻, S₅²⁻ and S₄⁻.

Two further peaks at 130–160 cm⁻¹ and 170–220 cm⁻¹ emerge between 2.19 V and 1.51 V. According to the DFT-results, the peak with the lower wave number can be attributed to polysulfide mono-anions (S₅⁻, S₇⁻, S₈⁻). This is unlikely or at least surprising because in this state of discharge polysulfides with lower chain length were expected. The other peak contains the Raman lines of S₂²⁻ (Li₂S₂) and S₃²⁻, especially at the end of discharge which perfectly fits to the theory of short chain polysulfides in this state of charge. Nevertheless, the peak maxima at 200 cm⁻¹ does not perfectly fit to a specific calculated polysulfide but could have been created together through S₃²⁻, S₄²⁻, S₅²⁻.

During charging we almost see the reverse process of polysulfides being created and consumed with Raman line positions already described above. The peak between 520–550 cm⁻¹ loses a lot of its intensity during charge, indicating that short chain polysulfides (e.g. S₃⁻) are likely responsible for its creation. The first sulfur peaks (likely together with long chain polysulfides e.g. S₅²⁻, S₈²⁻ and S₅⁻, S₈⁻) appear at 2.53 V, when the cell voltage increases rapidly to the end of charge voltage. These results should prove that there can be S₈ creation in Li-S cells during charge, although some publication denied this.⁵⁸ Nevertheless, we could not discover the three S₈ related peaks in every in-situ measurement with every CNT-S electrode, although we examined the electrodes at several measurement points. Therefore there seem to be certain electrode hotspots at which the accumulation of elemental sulfur is favored. Probably those electrode spots have a good electrolyte accessibility, high surface and low resistance. After end of charge and some additional minutes of relaxation another measurement of the electrode was made. The intensity of the three S₈ peaks was a little lower and the peak shoulders (especially at 215 cm⁻¹) seem to be more significant. This could be due to the beginning of a self-discharge being fast at the high level plateau with long chain polysulfides. The self-discharge leads to the decomposition of S₈ to long chain polysulfides and to medium chain polysulfides, created out of the former after some time.