High-Energy-Density Lithium–Sulfur Battery Based on a Lithium Polysulfide Catholyte and Carbon Nanofiber Cathode

Abstract

Li–S batteries are promising large-scale energy storage systems but currently suffer from performance issues; a major reason is the dissolution of polysulfides in electrolytes. To this end, we report a high-energy-density Lithium–Sulfur (Li–S) battery that combines a catholyte and a sulfur-free carbon nanofiber (CNF) cathode. The cathode was synthesized by carbonizing binder-free polyacrylonitrile (PAN) nanofibers, affording a high surface area. In the catholyte, added polysulfides acted as both conductive Li salts and active materials. Investigating the electrochemical performance of this concept in both Swagelok- and pouch-type cells afforded energy densities exceeding 3 mAh cm⁻² at a discharge rate of 0.1 C. This combination could also be utilized in high-capacity pouch cells with capacities of up to 250 mAh g⁻¹. Both cell types exhibited good cycle performance. Adding LiNO₃ to the electrolyte suppressed the redox shuttle reactions. Moreover, the cathode being binder-free increased the energy density and simplified cathode fabrication. Characterizing the cathode before and after cycling revealed that deposition was reversible, and that cell reactions at least partially formed sulfur as the end product, resulting in high sulfur amounts in the cell. We expect our concept to greatly aid in the development of practically applicable Li–S cells.

1. Introduction

Though there have been significant strides in small-scale Li-battery technology for portable devices [1], the demand for large-scale energy storage systems has surged in recent years, driven by the need for renewable energy and electromobility. These applications require batteries with high capacities and long cycle lives, such as to provide electric vehicles with sufficient driving distance on a single charge. In this context, Lithium–Sulfur (Li–S) batteries exhibit immense promise owing to their high theoretical energy densities (1675 mAh g⁻¹, 2500 Wh kg⁻¹) that surpass those of current Li-ion batteries by 5–10 times (LiCoO₂, = 274 mAh g⁻¹, LiFePO₄ = 170 mAh g⁻¹, and LiNi_1/3Co_1/3Mn_1/3O₂ = 278 mAh g⁻¹) [2,3,4,5,6]. Our research aims to contribute to this field by presenting a high-energy-density Li–S battery concept that could revolutionize energy storage and electromobility.

The basic concept behind Li–S batteries has been known for some time; however, several performance issues—high self-discharge, low cycle life, and low charge/discharge efficiency—have hindered the realization of this technology in practice. The high capacities of Li–S batteries stem from the electrochemical conversion of pure S in the charged state to Li₂S in the discharged state. This transition occurs in a series of steps involving the formation of polysulfides (Li₂S_x, x = 1–8), of which those with higher molecular weights are highly soluble in commonly used electrolytes. The dissolution of the polysulfides in the electrolyte leads to the loss of active material and the occurrence of the shuttle phenomenon, resulting in capacity loss and unwanted side reactions that limit cyclability [7,8,9]. In addition, S has a low electrical conductivity (5 × 10⁻³⁰ S cm⁻¹ at 25 °C); this causes an additional conducting phase to be required for the cathode, which further decreases the practical energy of the Li–S cell [10,11].

Several strategies have been applied to address the problems of low cyclability and low practical energy density based on the development of new material concepts. For instance, S can be confined in carbon and polymer composites, such as in a nanoporous carbon matrix or conducting polymer coating, to prevent the dissolution of polysulfides and improve conductivity. This strategy has improved active material utilization, cycle life, and charge/discharge efficiency [12,13,14,15,16]. However, in general, the S loading is relatively low in these systems, and the process for producing porous carbon is slow; furthermore, there is an additional risk of sulfur loss at the high process temperatures used. Thus, there are limitations on the practical energy density that can be obtained when using this approach.

Recently, progress has been made by combining the S–C composite material approach with another approach of adding Li polysulfides to the electrolyte, known as the catholyte concept. Li–S cells based on the catholyte concept have exhibited improved cycle performances, which is attributed to the buffering effect of the catholyte preventing the dissolution of the cathode [17,18,19,20,21]. Furthermore, the added polysulfides also play a role in modifying the interphase formed on the anode, which improves the stability of the cell, and they are electrochemically active to some extent, which contributes to the cell capacity [20]. However, the overall energy densities of the cells are not improved because the addition of polysulfides increases the additional mass, and because the cathode is still a C–S composite that contains a binder and, in some cases, additional carbon black for improved conductivity.

Herein, we report a novel strategy for applying the catholyte concept to high-energy-density Li–S cells by using a S-free electrospun carbon nanofiber (CNF) cathode. In our catholyte, the added polysulfides acted as both the active material and conducting salt, as our electrolyte comprised only Li₂S₈ in an organic solvent (tetraethylene glycol dimethyl ether (TEGDME) and 1,3-dioxolane (DIOX), 3:7 w/w). Owing to the chosen solvent, many polysulfides could be dissolved, and the amount of S in the full cell was higher than that in typical S–C composite cathodes. Using CNFs as the cathode provided the advantages of a large surface area and binder-free system, which are also easy to prepare in larger formats [22]. We tested the performances of the catholyte/CNF Li–S cells and characterized the CNF cathode before and after cycling to reveal its role in the functionality of the cell using cyclic voltammetry (CV), field-emission scanning electron microscopy (FE-SEM), and X-ray diffraction (XRD).

2. Materials and Methods

The CNFs for the cathode were prepared by carbonizing electrospun polyacrylonitrile (PAN) nanofibers. To prepare the PAN nanofibers, a solution of 12 wt% PAN (Mw 150,000, Aldrich, St. Louis, MO, USA) in N,N-dimethylformamide (DMF) (Aldrich) was fed into a syringe pump (KD Scientific, Holliston, MA, USA, Model 210) at a constant flux of 0.1 mL min⁻¹ and, under an applied voltage of 20 kV, electrospun into nanofibers, which were collected on aluminum foil fixed on a grounded stainless-steel rotating drum. The PAN nanofibers were then annealed at 300 °C for 1 h in air, following which they were carbonized by heating from room temperature to 1000 °C (10 °C min⁻¹), then held at 1000 °C for 1 h under nitrogen flow (99.99%, 800 mL min⁻¹) [23,24].

Cathodes were prepared by first cutting the CNFs into disks with diameters of 10 mm for Swagelok cells (surface area = 0.7854 cm²) and rectangles with dimensions of 25 × 50 mm for bi-cell-type pouch cells (two cathodes, surface area = 12.5 cm²), followed by vacuum drying at 70 °C for 12 h. Li metal (thickness = 300 µm; Cyprus Foote Mineral Co., Williamsburg, VA, USA) was used as the anode. The reference carbon cathodes were prepared by preparing slurries of conductive carbon (Super-P or CNT) and PVdF at weight ratios of 3:1 in N-methyl pyrrolidone (NMP) (Aldrich) and casting on aluminum foil.

The catholyte, 0.5 m Li₂S₈ in TEDGME (Aldrich)/DIOX (Aldrich) (3:7 w/w), was prepared by mixing Li₂S (Aldrich) and S (Aldrich) in appropriate amounts in TEGDME, followed by the addition of DIOX. To investigate the possibility of preventing redox shuttle reactions caused by the polysulfides, 0.1 m LiNO₃ (Aldrich) was added to the catholyte; see further details below. The total amount of sulfur loading in catholyte is 3.82 mg/cm².

The cells were prepared by stacking the CNF cathode, separator (Celgard 2400 separator, Charlotte, NC, USA), and Li-metal anode in either Swagelok or pouch cells. All material preparations and cell assemblies were performed in an inert atmosphere (H₂O and O₂ levels < 10 ppm).

The morphology of the CNF cathode was investigated using FE-SEM (JEOL JSM 5600, Tokyo, Japan) at an accelerating potential of 15 kV. XRD (Bruker D8 Advance) was used to investigate the presence of reaction products on the CNF cathodes exposed to the electrolyte after charge/discharge cycling; the XRD experiments were performed at room temperature in a closed system using an A100-B37 holder (Bruker, Billerica, MA, USA) to avoid exposure to air.

CV was performed at a constant sweep rate of 0.1 mV s⁻¹ over the voltage range of 1.5–2.8 V using an automatic galvanostatic charge–discharge WBCS3000 battery cycler (WonA Tech. Co., Seoul, Republic of Korea). Current densities of 127, 245, 635, and 2540 μA/cm² were used for Swagelok cells (corresponding to the rates 0.02, 0.04, 0.1, and 0.4 C, respectively), and 628 μA/cm² (0.1 C) was used for the pouch cells. All the cycling experiments were performed at room temperature.

3. Results and Discussion

Figure 1 shows the morphologies of the PAN nanofibers after carbonization. The standard process for converting polymeric PAN nanofibers into CNFs involves oxidative stabilization in an air atmosphere and subsequent carbonization. Oxidative stabilization is essential for increasing the carbon yield and preventing the formation of hollow-core fibers during the subsequent carbonization step. Then, carbonization at high temperatures eliminates unnecessary impurities. As shown in Figure 1a, fabrication via electrospinning resulted in well-constructed nanofiber structures. Figure 1b shows that the nanofiber structure was preserved well upon carbonization, while the nanofiber surface became smoother.

Figure 2a shows the discharge/charge behavior of the Li/catholyte/CNF cell. The first discharge curves of these cells with S-free cathodes showed only one plateau because there was no elemental S in the cells before discharge [25]. The cycling performance of the CNF/catholyte cell was stable, with an energy density of approximately 2.5 mAh/cm².

To investigate the effects of different carbon materials on cell performance, we compared the performance of the CNF-based cell with those of two other sulfur-free cathodes based on carbon nanotubes (CNTs) and Super-P. The profiles of the discharge curves were highly similar for the three materials, but the cells with the CNT- and Super-P-based cathodes delivered much lower capacities. Notably, the charge cycles of all three cells ended with very flat plateaus, which was attributed to the presence of a pronounced redox shuttle process. This was expected because the active material in each cell completely dissolved in the electrolyte, resulting in complete conversion to sulfur during charging, which is most likely not achieved. This can also be inferred from the discharge behavior, as there was little to no sign of the first plateau at around 2.4 V during discharge.

A common method for addressing redox shuttle reactions in Li–S cells is to add a small amount of LiNO₃ to the electrolyte to form a passivating layer on the Li metal anode [26,27]. The mechanism of the action of LiNO₃ and the effect of various concentrations of LiNO₃ were previously studied [28,29]. Figure 3 shows the CV results and cycling behavior of the Li/(catholyte + LiNO₃)/CNF cell at different current densities (discharge/charge rates). In the first cycle of the CV experiment, only one reduction peak at 2.0 V was observed, which is expected because the catholyte cell started discharging from Li₂S₈ [26]. Two separate reduction peaks were observed at 2.3 and 2.0 V in the subsequent cycles. This is clear evidence that, during the first charge cycle, the reaction proceeded all the way to S and that the redox shuttle reactions were efficiently suppressed.

Suppression of the redox shuttle and the resulting conversion to elemental S resulted in a higher discharge capacity. Figure 4b shows the discharge/charge behavior of Li/(catholyte + LiNO₃)/CNF cells. Although the rates were considerably higher (245 and 635 µA cm⁻² for discharge and charge, respectively, compared to 127 µA cm⁻² in Figure 3), a higher energy density of 3.1 mAh cm⁻² was observed, and a relatively stable capacity profile was obtained, with 84% of the capacity being retained after 30 cycles. At a higher current density (2540 µA cm⁻² or 0.4 C), the cell exhibited an initial energy density of 1.8 mAh cm⁻² and retained 60% of the capacity after 100 cycles. This excellent performance was also preserved when employing a high-energy-density pouch cell. Figure 4 shows the cycling behavior of a Li/(catholyte + LiNO₃)/CNF pouch cell; it delivered a discharge capacity of 230 mAh in the second cycle when the full potential was utilized after conversion to S during the first charge, and the electrode retained 80% of its capacity after 10 cycles.

The reported energy density, more than 3 mAh/cm², is an excellent performance compared with other results reported in the literature, which are typically in the range of 1–1.9 mAh cm⁻² [19,30,31]. However, a recent study reported a similar concept to the one presented here; the study reported a very high energy density of 2.9 mAh cm⁻², but a very low actual S utilization of approximately 650 mAh g⁻¹ [32]. This shows that to obtain a high-energy density cell, high S utilization, which is typically used as a benchmark of performance in the literature, is not sufficient. Therefore, the excellent performance of the CNF cathode can be directly attributed to the fact that it consists entirely of conducting material without a binder.

To understand the functionality of the CNF cathode and the mechanisms involved in the catholyte cell, we characterized the morphology of the fiber network before and after cycling. Figure 5a shows the SEM image of the as-prepared CNF cathode. The network comprised bead-free fibers with an average diameter of 450 nm. It can be seen in Figure 5b that after exposure to the catholyte (0.5 m Li₂S₈ in TEGDME/DIOX + 0.1 m LiNO₃) but before discharge, the material was deposited on the surfaces of the fibers, but their shapes and dimensions were not altered. After discharge, as shown in Figure 5c, the amount of deposited material was significantly higher, with the deposit almost completely covering the fibers. However, after the subsequent charge cycle, very few deposits remained (Figure 5d). This shows that the fiber network played an active role during charge/discharge cycling and that the deposition process was reversible and did not degrade the CNF.

Further insight into the mechanism was obtained by analyzing the nature of the deposits through XRD experiments. Figure 6 shows the XRD patterns of the as-prepared CNF, CNF exposed to the catholyte, and cathode after discharging and charging. The XRD patterns of Li₂S and S are also shown for comparison. No sharp peaks appeared in the XRD pattern of the as-prepared CNF material; rather, broad features characteristic of a material with low crystallinity or a fully amorphous material were observed. Similarly, no sharp peaks were observed for the CNF exposed to the catholyte, indicating that the deposit on the CNF surface was amorphous; however, the XRD pattern after exposure to the catholyte must be considered, as the residues of the electrolyte most likely contributed because the experiments were performed without drying the cathodes. In contrast, the XRD patterns of the cathode after charging and discharging showed sharp peaks, indicating the presence of a crystalline material on the CNF surface. The XRD pattern for the cathode after discharging was identified against the pattern for Li₂S, which is the expected end-product; the peaks were rather broad, indicating the presence of a disordered structure or a small-sized crystallite of Li₂S on the CNF surface. Meanwhile, after charging, the expected end product is S if the reaction is fully reversible. In the XRD pattern of the CNF cathode after charging, peaks corresponding to S were observed; however, the overall pattern was not a full match. This shows that the reaction in the cell was at least partially successful in reaching S, despite starting with Li₂S₈ in the initial cycle [30,31].

From the results, it is clear that this catholyte concept, wherein lithium polysulfides act as both the conducting salt and active material, can be used for high-energy-density Li–S batteries. The key to this mechanism is the combination of the high surface area of the CNF cathode and the absence of non-conducting components in the cathode. The cathode structure and morphology were observed by comparing the different carbon structures (CNT and Super-P). Another aspect of the concept presented herein is that it is more environmentally friendly than traditional electrolyte concepts; because we do not use salts other than Li₂S₈ or LiNO₃ in our Li–S cell, it is fluorine-free, unlike standard concepts based on fluorinated salts such as LiTFSI or LiCFSO₃. This also decreases the sensitivity to degradation as the hydrolysis of fluorine salts, which creates HF, does not occur.

4. Conclusions

We demonstrated a novel Li–S cell concept based on a Li-salt-free catholyte and self-standing CNF cathode. The fabricated cells exhibited good performances in terms of capacity and cyclability. We then showed that the addition of LiNO₃ to the electrolyte efficiently suppressed the redox shuttle mechanism, even in this system in which all active materials are dissolved. The CNF cathode is a critical component of this concept because it has a large surface area and an open structure. The role of the cathode surface was revealed by identifying that the end products Li₂S and S were obtained after discharging and charging, respectively. The CNF cathode has the advantage of being binder-free, which increases the practical energy densities of the cells. Furthermore, the absence of a binder considerably simplified the cathode preparation process, which provides excellent potential for upscaling.

Although the concept presented here exhibited excellent performance, there is room for further improvement: the CNF cathode surface and the network morphology can be optimized, and the catholyte can be further optimized to improve the reversibility of the system.