Li-S Batteries with Li₂S Cathodes and Si/C Anodes

Commercially available Li₂S powder (Sigma Aldrich) was used as-received to prepare Li₂S cathodes. The Li₂S particle size ranged from few hundreds of nanometers to more than 20 micrometer, as shown by the SEM image in Fig. 1a; particle size analysis by laser scattering (not shown) indicated a median diameter of ≈10 μm (volume-averaged distribution). The ink for the electrode coating was prepared by mechanically mixing the Li₂S powder with commercial Vulcan XC72 carbon (≈30 nm primary particles and ≈200–300 nm primary agglomerates) using a mortar and pestle, followed by dispersion in a solution of N-Methyl-2-pyrrolidone (NMP) with dissolved polyvinylidene fluoride (PVDF) for two hours using a high speed magnetic stirrer. The solids content of the ink was approximately 150 mg (Li₂S+C+PVDF) per milliliter of NMP, with a solids composition of 60%wt Li₂S, 30%wt C, and 10%wt PVDF. The ink was coated onto an 18 μm thick aluminum foil using a 300 μm gap doctor blade. Because of the high reactivity of Li₂S with moisture, all the processing steps were carried out inside an Ar-filled glove box (MBRAUN; <1 ppm H₂O & <1 ppm O₂). After drying inside the glove box, 10 mm diameter electrodes were punched out and further dried at 80°C under dynamic vacuum in a Büchi oven for 2 hours. For all the measurements, the Li₂S loading in the cathode was maintained at ≈2.0 ±0.1 mg_Li2S/cm² (≡ 2.3 ±0.1 mAh/cm² theoretical capacity). Morphological characterization of the cathode electrodes was carried out using SEM-EDS (JEOL, JSM 6000 equipped with EDS) using an air-tight specimen holder to transfer samples from the glove box to the SEM chamber.

Zoom In Zoom Out Reset image size

The Si anodes used in this study contained 20%wt nano-silicon particles, 60%wt graphite, 12%wt conductive carbon-black, and 8%wt Na-CMC binder, supported on a Cu-foil current collector (Si anodes were provided by Wacker Chemie AG). The areal weight of the anode electrodes was 2.2 mg/cm² (including Si, graphite, carbon-black, and binder). Assuming specific capacities of 3580 mAh/g_Si,^32,33 372 mAh/g_graphite, and 150 mAh/g_carbon-black,³⁴ the theoretical capacity of the Si/C anodes equates to 2.1 mAh/cm². Anodes were punched out with a diameter of 10 mm (i.e., 0.785 cm²).

All electrochemical testing was conducted in a Swagelok type cell using DOL:DME (1:1 v/v) with 1 M LiTFSI and 0.5 M LiNO₃ electrolyte; the amount of electrolyte was normalized to ≈40 μl/cm² for Li/Li₂S or Li/Si half-cell measurements, and to ≈35 μl/cm² for Si/Li₂S full-cell measurements. Three layers of Celgard C480 were used as a separator. Most experiments were conducted in a two-electrode configuration, except in few specifically mentioned instances when a three-electrode configuration with a metallic lithium reference electrode was used. The cells were cycled galvanostatically (BioLogic VMP3 potentiostat) at different currents (rates) between 1.7 and 3.0 V_Li for Li/Li₂S half-cells, between 0.02 and 1.5 V_Li for Li/Si half-cells, and between 1.3 and 2.6 V_cell for Si/Li₂S full-cells. Before cycling the cells, the Li₂S cathodes were initially activated by charging them up to 4.0 V_Li for half-cells (Li/Li₂S) and 3.8 V_cell for the full-cells (Si/Li₂S). All cells were cycled in a climatic chamber at 25°C.

A specially designed cell was used for the in-situ X-ray diffraction measurements, in which the aluminum current collector of the cathode electrode served as X-ray window. X-ray diffraction (XRD) patterns were recorded using a STOE Stadi P diffractometer equipped with a linear position-sensitive detector (Dectris Mythen 1 K) and a Mo K_α source (50 kV of tube voltage and 20 mA current) in a Bragg-Brentano configuration. Each diffraction pattern was measured in three ranges with a step size of 15° (2θ = 4–18°, ω = 6.5° fixed). The 2θ range was chosen so that the reflections from the aluminum window do not interfere with the reflections of the Li₂S. The exposure time for each range was 180s, equating to an acquisition time of 9 minutes for a complete in-situ diffractogram.

Figure 2a shows the initial five charge/discharge cycles of the Si anode in a Li/Si half-cell at a rate of C/4.5 (referenced to the theoretical Si/C anode areal capacity of 2.1 mAh/cm², and corresponding to a geometric current density of 0.46 mA/cm²). It can be observed that the very first lithiation capacity (red dashed line in Fig. 2a) reaches the calculated theoretical capacity (see Experimental section), following a potential profile which is characteristic of crystalline silicon.³³ After the first lithiation and delithaiation cycle, the specific lithiation and delithiation capacity decreases to a constant value of ≈1.8 mAh/cm², corresponding to ≈85% of the theoretical areal capacity of the composite anode. The higher capacity observed in the first lithiation is attributed to SEI formation on the Si surface. The potential profile of the subsequent lithiation cycles is different from the first one, reflecting the typical lithiation profile of amorphized silicon.³³ As shown in Figure 2b, the 2^nd cycle lithiation/delithiation capacities for these Si/graphite composite anodes is nearly independent of rate between C/18 and C/4.5. Overall, from these results it can be concluded that the reversible capacity of the Si anodes in our study is ≈1.8 mAh/cm² (within the first four cycles after the initial formation) and does not change significantly within the measured range of C-rates.

Zoom In Zoom Out Reset image size

Similar to the initial high-potential activation required during charging of Li₂O₂ pre-filled cathodes in the case of Li-O₂ batteries,³⁵ we found that the initial charging of Li₂S cathodes also requires activation at higher potentials. This observation was reported also by Yang et al.,²⁶ who showed that a ≈1 V overpotential is required during the first charge of a Li/Li₂S half-cell in comparison to charging a discharged Li/S cell, which they ascribed to the initial barrier of forming soluble polysulfides in the electrolyte, which in turn serve as an electron shuttle between the electronically conductive carbon matrix and the Li₂S particles. Consistent with this hypothesis, Meini et al.³⁶ found that the high potential required in the first charging step of Li₂S cathodes can be eliminated in the presence of redox-active additives. Figure 3a shows the initial charging of a Li₂S cathode against a Li anode (Li/Li₂S half-cell) at rates of C/10 (blue line) and 1 C (red line), with the voltage initially increasing to potentials of ≈3.4 V_Li and ≈3.9 V_Li, respectively, followed by a sharp decrease to a lower voltage plateau region. At the rate of C/10 (blue line), two charging plateaus at ≈3.2 and at ≈3.5 V_Li are observed until the charging capacity approaches its theoretical value of 1165 mAh/g_Li2S, at which point the voltage increases to ≈4.2 V_Li, where the current is sustained by the continuous oxidation of the DOL/DME electrolyte. On the other hand, at a rate of 1 C, the charging voltage strongly fluctuates and the charging process continues indefinitely between 2.5 and 3.5 V_Li, which we ascribe to the intermittent formation of lithium dendrites at this rather high current density of 2.3 mA/cm² (dendrite formation was shown to occur by Aurbach et al.³⁷ at >1 mA/cm²). This is evidenced by SEM/EDS analysis of the middle separator removed from the Li/Li₂S cell after 1C charging (s. Fig. 4a): the separator was severely penetrated by lithium dendrites with a spot-size of 50 μm or more, and EDS mapping shows that these spots have a high oxygen content, indicating lithium oxides, hydroxides, or carbonates formed during air exposure of the separator (the separator was intentionally exposed to air to oxidize metallic lithium dendrites). These results confirm that the high current density during 1 C charging (2.3 mA/cm²) leads to severe dendrite formation if metallic lithium used as anode.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

As one would expect, lithium dendrite formation at high C-rate can be avoided during initial charging of a Li₂S cathode in Si/Li₂S full-cells, as evidenced by the smooth charging profile at 1 C in full-cells shown in Figure 3b (red line), where the potential is referenced against a metallic lithium reference electrode (using a three-electrode cell). Again, the capacity obtained as the potential reaches 4 V_Li is close to the theoretical capacity of the Li₂S cathode.

In order to confirm that Li₂S is oxidized completely during the first activation cycle and to determine whether the lower and upper plateaus at the C/10 charging profile in the Li/Li₂S half-cell (blue line in Fig. 3a) are indeed related to Li₂S charging, in-situ XRD analysis was carried out with a Li₂S cathode and a lithium metal counter electrode. Figure 5a shows the charging potential profile (vs. lithium) for a Li₂S cathode at C/10 in the in-situ XRD cell, which is in excellent agreement with the charging profile obtained with a standard Swagelok cell (see Fig. 3a, blue line). The labels placed along the charging curve (letters a - i in Fig. 5a) mark the collection of the XRD diffractograms shown in Fig 5b (note that the diffractograms were acquired over a time of 9 minutes during cell charging). As expected, sharp peaks corresponding to the (111) and (200) diffractions of Li₂S are observed at the beginning of the charging process (a in Fig. 5b), then gradually decrease during the first charging plateau at ≈3.2 V_Li (b – d in Fig. 5b), and remain visible throughout the second charging plateau of ≈3.5 V_Li (e and f in Fig. 5b). The Li₂S diffraction peaks only disappear after the second charging plateau, i.e., after having reached 85–90% of the theoretical Li₂S charge capacity (g – i in Fig. 5b). This may be compared to the study by Cañas et al.,³⁸ who observed the disappearance of Li₂S diffractions during the charge of a discharged S cathode at already ≈50% of the charging capacity, which is most likely due to the much smaller Li₂S particles formed during S cathode discharge compared to the ≈20 μm large Li₂S particles in our electrodes.

Zoom In Zoom Out Reset image size

The appearance and growth of new diffraction peaks (marked by * in Fig. 5b) can be observed at already ≈60% of charging (d in Fig. 5b), developing into sharp diffraction peaks after ≈90% of charging (g in Fig. 5b), which suggests the formation of large crystallites. These peaks must be related to the formation of crystalline sulfur, even though their positions do not exactly match the PDP database, indicating the formation of a new crystalline phase of sulfur. A similar observation was reported by Cañas et al.,³⁸ who reported sulfur reflections after ≈90% of charge of a previously discharged sulfur cathode, with diffraction patterns which were also different from the PDP database. Furthermore, Nelson et al.³⁹ also observed the formation of crystalline sulfur after ≈80% charging of a discharged sulfur cathode. These findings, however, contradict the in-situ XRD study by Yang et al.,²⁶ where no crystalline sulfur is observed even after complete charging of their Li₂S cathode; the origin of this discrepancy is unfortunately not clear.

In summary, the analysis of Figure 5 suggests that the charging process in the first plateau at ≈3.2 V_Li, which extends to ≈60% of the charging, can be assigned to the transformation of Li₂S to lithium polysulfides and possibly small amounts of sulfur, as the potential is sufficiently high to oxidize polysulfides to sulfur. After the second charging plateau (g in Fig. 5), all Li₂S diffraction peaks have vanished and sharp diffraction peaks related to crystalline sulfur appear (at around 2θ = 12.8°), indicating the formation of large sulfur crystallites in the charged Li₂S cathode.

The formation of large crystalline sulfur domains after charge of the Li₂S cathode was verified by ex-situ SEM/EDS analysis of the Li₂S cathode after charging to a cutoff potential of 4V_Li at C/10, i.e., after nearly 100% of the theoretical charge was obtained. For this investigation, the cell was disassembled inside a glove box, washed with DME to remove any residual polysulfides from the cathode, and then the electrode transferred into ambient air in order to convert potentially remaining Li₂S into Li₂CO₃, LiOH, and Li₂O; subsequently, the sample was transferred into the SEM. Strikingly, as shown in Fig. 6a, the surface of the cathode was covered with large sheets of sulfur, as confirmed by the EDS elemental mapping: the sulfur map matches the SEM image (compare Figs. 6a and 6c), while no common patterns are observed between the oxygen map and the SEM image (compare Figs. 6a and 6d), which after air exposure could only be the case for elemental sulfur. The sulfur sheets extend over several hundreds of micrometers across the cathode surface, whereby imprints from the separator can be observed on the sulfur sheets. The cross-section of the cathode was examined after bending of the electrode, as shown in Fig. 6b. Empty pockets inside the cathode electrode can be clearly seen, the sizes of which are consistent with the initial size of the Li₂S powder used to prepare the Li₂S cathode, which is consistent with the absence of Li₂S diffraction peaks after charging (s. Fig. 5).

Zoom In Zoom Out Reset image size

Analogous to our study, the formation of large crystalline sulfur domains at the cathode/separator interface upon charging of a discharged sulfur electrode was also observed by Cañas et al.,³⁸ albeit with different shapes/morphology. We believe that the accumulation of sulfur at the cathode/separator interface during the last stages of the charging process is most likely due to the oxidation of dissolved polysulfides (solubilities of 0.1 to 1 mol/l in the electrolyte) contained inside the separator (rather than inside the cathode) during the final stages of charging: under this condition, the high cathode potential will cause the rapid and highly localized oxidation of polysulfides which are stored inside the separator region once they reach the cathode/separator interface, thereby forming a deposit at this very interface.

The activation and subsequent cycling of Si/Li₂S full-cells was investigated using Si/C anodes and Li₂S cathodes of theoretical capacities of ≈2.1 and ≈2.3 mAh/cm², respectively. In order to monitor the charge/discharge behavior of each electrode, the Si/Li₂S full-cells were assembled with a metallic lithium reference electrode in a three-electrode cell. The questions to be addressed were: i) whether the polysulfide shuttling currents would be low enough to allow for a complete charging of the Si/C anodes; ii) whether the high voltages at 1 C activation (≈3.3/≈3.5 V_Li in the 1^st/2^nd charging plateau; red line in Fig. 3b) compared to C/10 activation (≈2.5/≈3.1 V_Li in the 1^st/2^nd charging plateau; blue line in Fig. 3b) would affect the subsequent cycling stability, as activation at lower potentials was suggested to improve cycle-life;³⁶ and, iii) whether the long-term cycling stability of Si/Li₂S full-cells would be comparable with that of Li/Li₂S half-cells.

The Si/Li₂S full-cells were subjected to an initial activation from OCV (≈0.3 V_cell) to a cell voltage of 3.8 V_cell at two different rates, viz., C/10 (Fig. 7a) and 1 C (Fig. 7b), followed by continuous cycling between 1.3 and 2.8 V_cell at C/5 rate. Figures 7a and 7b show the potential variation of the Li₂S cathodes vs. Li reference (curve 1, dashed blue line), the cell potential of the Si anodes vs. the Li₂S cathodes (curve 2, black line), and the potential of the Si anodes vs. Li reference (curve 3, dashed red line). Examining the first activation cycle at both C/10 and 1 C (initial charging panels in Figs. 7a and 7b, respectively), it is clear that the potential of the Si anodes is gradually lowered to <0.5 V_Li, where lithium ion intercalation will occur. The fact that the Si anode potential decreases faster to below 0.5 V_Li at 1 C (at ≈120 mAh/g_Li2S; dotted red line in Fig. 7b) compared to charging at C/10 (at ≈320 mAh/g_Li2S; dashed red line in Fig. 7a) is ascribed to the polysulfide shuttling current, which consumes a larger fraction of the charging current at the lower C-rate. At the end of the activation cycle, the Si anode potentials reach ≈0.1 V_Li, which according to Fig. 2b would correspond to a ≈50% charge of the Si anode (≈1 mAh/cm²). The missing charge compared to the expected ≈2.1 mAh/cm² (s. above) must be caused by the irreversible lithium loss for SEI formation. This is consistent with the only ≈50% theoretical cell capacity obtained in the second discharge/charge cycle at C/5 shown in Fig. 7: ≈550 mAh/g_Li2S corresponding to ≈1.1 mAh/cm² at C/10 and ≈500 mAh/g_Li2S corresponding to ≈1.0 mAh/cm². If compared with the Si anode half-cell data at the same current density (red lines in Fig. 2b), a charge/discharge capacity of ≈1.0–1.1 mAh/cm² is consistent with the observed Si anode potentials in the 2^nd cycle, ranging between ≈0.12 and ≈0.5 V_Li (s. C/5 segments in Fig. 7). These results clearly indicate that the capacity of the cell is limited by the Li₂S cathode due to a large irreversible lithium loss during the first charging cycle, while only about ≈50% of the anode's capacity is used. Consequently, we expect that the Si/Li₂S full-cell performance will improve by applying an excess cathode capacity of Q_Li2S/Q_Si between 1.5/1 and 2/1, which we will explore in future experiments.

Zoom In Zoom Out Reset image size

The Li₂S cathode potential profiles during activation (blue dashed lines in Fig. 7) are essentially identical to those shown in Fig. 3b, except that potential oscillations are observed toward the end of the C/10 charging curve (dashed blue line in Fig. 7a). Indeed, we have frequently observed these potential oscillations during charging at low C-rates, and Fig. 7a clearly shows that it is caused by the Li₂S cathode and not the anode electrode. The fact that these oscillations appear in the second charging plateau, where we had observed the onset of the formation of crystalline sulfur phases (see Fig. 5) suggests that they are related to the formation of current-blocking sulfur domains at the separator/cathode interface (see Fig. 6). The absence of these oscillations at the higher rate of 1 C would then suggest that the formation of large sulfur sheets might be suppressed at higher rates, which could be explained by fact that the cathode potential in that case is far above the potential required for the complete reduction of polysulfides to sulfur, so that the polysulfide concentration in the separator will remain low (see above discussed hypothesis for the sulfur sheet formation at the separator/cathode interface).

The long-term cycling performance of Si/Li₂S full-cells is shown in Fig. 8a (showing the average and the standard deviation of two cells for each test sequence), exploring the effect of C-rate during activation and during long-term cycling on capacity retention: i) initial activation at 1 C followed by cycling at C/5; ii) initial activation as well as cycling at 1 C; and, iii) initial activation at C/10 followed by C/5 cycling. It can be seen that the cells that were activated at C/10 followed by C/5 cycling show somewhat higher capacities (red line in Fig. 8a) compared to the cells which were activated at 1 C and cycled at either C/5 or 1 C (black and gray lines in Fig. 8a). While the difference is small, it might be related to the much lower charging voltages if the initial activation step (i.e., the very first charging cycle) is done at C/10 instead of at 1 C (see Fig. 7), which might prevent the formation of electrolyte decomposition products at the initially high voltages that could compromise cycle-life. The higher coulombic efficiency of the cells cycled at 1 C (gray lines in top panel of Fig. 8a) compared to the cells cycled at C/5, independent of the activation procedure (red and black lines in Fig. 8a), is indicative of significant polysulfide-shuttle currents, which play a lesser role at the higher C-rate. After 70 cycles, the cells activated at C/10 and cycled at C/5 have a remaining capacity of 280 mAh/g_Li2S, (≡400 mAh/g_S). To our knowledge, our study is the first to examine the long-term stability of Si/Li₂S full-cells based on Li₂S powder based cathodes and Si based anodes. The only related study is the work by Yang et al.,⁴⁰ who cycled Si anodes with Li₂S cathodes obtained by lithiation of a sulfur cathode with n-butyllithium, observing mych higher capacity fading (to ≈250 mAh/g_Li2S after only 20 cycles at a similar rate of C/3).

Zoom In Zoom Out Reset image size

On the other hand, Si/S full-cell data have been obtained combining electrochemically pre-lithiated Si anodes with S cathodes. For the latter configuration, Elazari et al.²⁹ demonstrated ≈380 mAh/g_S after 60 cycles at C/5, and Brückner et al.³¹ showed ≈400 mAh/g_S after 1400 cycles at C/2 (≈700 mAh/g_S after 70 cycles at C/2). They also showed increased capacity fading with increasing C-rate for Li/S half-cells, while high C-rates of a large number of cycles could be maintained with Si/S full-cells based on S cathodes and pre-lithiated Si anodes, which they ascribed to both a larger irreversible electrolyte loss at the lithium anode and/or lithium dendrite formation. Without providing any proof, their hypothesis would suggest that the rate capability of Si/S or Si/Li₂S full-cells would be higher than that of Li/S or Li/Li₂S half-cell. That this is indeed the case is shown by our Si/Li₂S full-cell tests, which demonstrate the same capacity retention of 350 mAh/g_S (≡250 mAh/g_Li2S) after 70 cycles, independent of whether cells were cycled at 1 C or C/5 (gray and black lines in Fig. 8a, respectively). Thus, we can provide for the first time an unambiguous proof that the poor rate capability in Li/S or Li/Li₂S half-cells is caused by the lithium anode and that the kinetics of the sulfur cathode is much faster than previously assumed based on half-cell data.

Furthermore, the fact that the capacity retention of Si/Li₂S full-cells is the same at C/5 and 1 C demonstrates that capacity fading is not related to total time, but to the number of cycles, which points toward irreversible lithium loss, originating from SEI expansion/contraction during each cycle. This is indeed supported by a comparison of the capacity retention between Si/Li₂S full-cells and Li/Li₂S half-cells at C/5 (s. Fig. 8b): while the initial capacities are quite similar for half-cells and full-cells, the capacity retention of half-cells is superior (≈550 vs. ≈400 mAh/g_S after 70 cycles), which supports the above hypothesized active lithium loss into the SEI. Here it may be noted that the capacity retention of our Li/Li₂S half-cells compares reasonably well with that shown previously by Yang et al.²⁶ (≈650 mAh/g_S after 50 cycles at C/10) and Cai et al.⁴¹ (≈550 mAh/g_S after 45 cycles at C/5).