3.1. Stability of solid electrolytes, nanosized versus micronsized
The electrochemical stability of solid electrolytes was recently shown to be significantly lower than initially thought, especially for sulfide based solid electrolytes, where consequential decomposition reactions have been shown to have large detrimental impact on the all-solid-state battery performance10, 43-44. To demonstrate the redox activity of the sulfur solid electrolyte material and the impact of particle size, several mixtures of sulfur solid electrolytes with carbon are galvanostatically cycled. As shown in Figure 1a and b, when nanosized Li3PS4 (nLPS) and Li6PS5Cl (nLPSC) are mixed with conductive carbon (referred to as nLPS-C and nLPSC-C respectively), they are readily oxidized at low oxidation potentials, <2 V vs. In (<2.62 V vs. Li/Li+), at a relatively low current density of 0.033 mA/cm2, in agreement with recent findings10, 25. This illustrates that the decomposition of nanosized solid electrolytes in cathodic mixtures is the prime reason for the short cycle life of current all-solid-state batteries utilizing sulfide solid electrolytes25. The combination of a Li2S cathode and the LPSC electrolyte has been intensively studied3-4, 29-30, where the harsh activation process of Li2S, caused by the low bulk conductivity of Li2S, and sluggish Li+ transport between the electrode and electrolyte represent another critical issue. A straightforward approach to achieve easier activation of Li2S and to improve the Li+ transport between the electrode and electrolyte is to reduce the particle size of both the Li2S and the LPSC (here referred to as nLi2S-nLPSC-C electrode). Accompanied by a large overpotential of ~800 mV, this results in a substantial capacity during first charge and discharge and upon subsequent cycling, as shown in Figure 1f and Figure S1, which is however hard to distinguish from the capacity of the electrolyte itself (because oxidation of LPSC is carried by the S2-/S redox).
The simplest strategy to reduce the contribution of the solid electrolyte to the capacity is to lower the ionic contact area through the use of micron sized solid electrolyte particles. To verify the smaller redox activity, micron sized LPSC (average diameter of 50 mm, as shown in Figure S2) was mixed with carbon, and charged (oxidized) to 3 V vs. In-Li (3.62 vs. Li/Li+) in the solid-state battery configuration, referred to as the mLPSC-C electrode. The result, shown in Figure 1c, demonstrates nearly no contribution of the LPSC solid electrolyte to the capacity, reflecting the smaller amount of decomposition reactions under the same current density. However, as expected, the small ionic contact area of micron sized LPSC compromises ion transport over the interface, leading to a very small capacity of the sulfur active material, even when nanosized Li2S is employed as shown in Figure 1d. Although nanosized Li2S displays a slightly larger capacity, the rapid voltage increases to the 3.5 V vs. In-Li (4.12 vs. Li/Li+) cut-off demonstrates that in both cases the sulfur cathode material is marginally activated. At the same time, decreasing the ionic contact area increases the internal resistance and thus the overpotentials experienced at the solid electrolyte, which will induce decomposition reactions and thus driving a self-amplifying resistance growth towards battery failure.
The mechanical mixing of halogen salts like LiI with Li2S and sulfide solid electrolytes is an often-applied strategy, to improve conductivity, although the exact mechanism has not been clarified29-31, 38-39. At present we take a different and more controlled approach to study the detailed impact of LiI on the Li-ion transport over the grain boundaries by NMR exchange experiments, and introduce LiI at the Li2S-LPSC interfaces. Rather than using the conventional ball-milling route, this is achieved by introducing LiI via solution, making use of the much better solubility of LiI in ethanol compared to Li2S (see supporting information Figure S3). The solution was then evaporated at 300 oC (Figure S3) to obtain a LiI-Li2S (1:3 molar ratio) composite, where LiI precipitates on the surface of Li2S as discussed below. This cathode was subsequently hand mixed with LPSC and C (referred to mLi2S(LiI)-mLPSC-C) to prepare the cathodic mixture and an all-solid-state mLi2S(LiI)-mLPSC-C|mLPSC|In-Li battery was assembled under 2 MPa pressure. This has a large impact on the electrochemical charging, as shown in Figure 1e, demonstrating that the introduction of LiI results in a very low sulfur redox activation plateau at 1.69 V vs. In (2.31 V vs. Li/Li+) of the micron sized Li2S combined with micron sized LPSC. The plateau is followed by a rapid increase in potential which reflects oxidation of the solid electrolyte and/or of LiI (to LiI3 which is known to occur at ~2.3 V vs. In)45-46. To identify the contribution of the LiI and/or LPSC oxidation, this measurement was repeated in a battery without Li2S, shown in Figure 1e, leading to charging at a higher voltage marking the oxidation of LPSC and/or LiI. In conclusion, deposition of LiI on Li2S via solution, and hand mixing this cathode with LPSC results in an extremely low activation (oxidation) potential for Li2S, suggesting that facile Li+ transport between the electrode and electrolyte is achieved even for a relatively small ionic contact area between the micronsized solid electrolyte and the electrode particles.
3.2. Electrode and LiI coating synthesis and characterization
To understand the role of LiI in the activation of Li2S, a detailed structural investigation was performed. Both LiI and Li2S have a cubic structure indexed to the Fd-3m space group. Three Li2S-LiI composites were prepared via dissolution and precipitation where Li2S:LiI molar ratios of 9:1, 3:1, and 1:1 were added to ethanol, followed by evaporation of the solution at 300 °C. Li2S and LiI were also individually dissolved and precipitated from ethanol via evaporation for comparison. X-ray patterns of pristine Li2S and LiI precipitated Li2S and LiI and the three Li2S-LiI composites are provided in Figure 2a and S4. From a cursory inspection it can be observed that the peaks corresponding to the precipitated Li2S are much broader than those of the parent Li2S, while the peak width of precipitated LiI is comparable to the parent LiI. This indicates that on precipitation smaller primary crystallites of Li2S are obtained. In the three composite mixtures, both the Li2S and LiI phases could be indexed, albeit with shifts in peak positions of the Li2S component indicating changes in lattice parameters of this phase. Rietveld refinement was further performed of all the patterns depicted in Figure S5, and the lattice parameters obtained from the refinement are given in Figure S4. It can be seen that the lattice parameter of LiI (6.025 Å) remains unchanged from that of the pristine material. On the other hand, with increasing amounts of LiI in the composite the lattice parameter of Li2S keeps increasing from 5.701 Å (pristine) to 5.750 Å (1Li2S:1LiI), which could be due to the much smaller average crystallite size of 9.18 nm (1Li2S:1LiI) compared to 162.92 nm (pristine) as shown in Figure S6 or by incorporation of the I in the Li2S lattice.
Additional SEM and TEM measurements are performed to study the morphology of the pristine Li2S and the Li2S-LiI mixture. As shown in Figure 2c, the prepared mixture consists of a microstructure comprising of micron sized secondary particles with a relatively uniform particle size of around 5 mm similar to pristine Li2S in Figure 2b. TEM is used to study the morphology at smaller length scales (100 nm). As seen from the TEM image and energy spectrum (Figure 2d), the EDS mapping of the particle surface shows uniform S and I distribution, indicating a mixture on the nanoscale was obtained with this precipitation method, and the LiI was uniformly distributed over the surface structure of Li2S particles. To further verify the structure of the Li2S-LiI material, XPS depth profiling was performed as shown in Figure 2e and f. The S 2p XPS signal is relatively low until a depth of ∼100 nm, and vice versa the I 3d is relatively high to approximately the same depth (selected window diameter is as small as 14 mm to locate only a few particles). Therefore, the present precipitation method results in micron sized secondary cathode particles, referred to as mLi2S(LiI), that exist of agglomerates of LiI coated nanosized primary Li2S particles, where the micron sized agglomerates are coated by a relatively thick LiI layer at some positions accumulating to large domains of LiI (as observed with XRD).
3.3. Li-ion conductivity and role of LiI in the Li-ion transport mechanism
To investigate the impact of the LiI coating on the conductivity, impedance spectroscopy and 6Li solid NMR spectroscopy experiments are conducted. The temperature dependence of the ionic conductivities for the pristine Li2S and LiI as well as pellets of the 3:1 Li2S-LiI composite are presented in Figure 3a. The conductivity of all the materials follow an Arrhenius law, resulting in activation energies of 0.235, 0.107 and 0.212 eV for Li2S, LiI and the Li2S-LiI mixture, respectively. The room-temperature conductivity of the Li2S-LiI mixture (6.72 × 10−9 S/cm at 25 °C) is between that of Li2S (7.51 × 10−11 S/cm at 25 °C) and LiI (0.97 × 10−7 S/cm at 25 °C), indicating that the LiI in the Li2S agglomerates enhances the overall conductivity of the cathode material. To investigate the role of LiI as interphase material between the Li2S electrode and LPSC solid electrolyte, (2D) 6Li–6Li exchange (2D-EXSY) solid state NMR experiments are performed. These experiments can provide selective and non-invasive quantification of the spontaneous Li+ diffusion, charge transfer, over the solid–solid electrolyte–electrode interface (between two phases) in realistic solid state cathode mixtures, as previously reported4, 24. The one-dimensional (1D) 6Li magic angle spinning (MAS) NMR spectra of the micron sized Li2S-LPSC cathode mixture, shown in Figure 3b, displays two resonances with chemical shifts of 2.31 and 1.29 ppm, representing Li in Li2S and in LPSC respectively. Compared to Li2S, the larger screening of Li+ in the LPSC results in the upfield 6Li chemical shift position. The difference in chemical shift between Li in Li2S and LPSC that allows to distinguish both species, makes it possible to conduct the 2D exchange experiments. In the 2D exchange spectrum, Figure 3c, both Li+ environments observed in the 1D spectra are clearly observed, where the more LPSC narrow resonance is due to the higher mobility of Li+ in the solid electrolyte. 2D exchange NMR effectively measures the spectrum of the 6Li ions at t=0 s, then waits a mixing time Tmix, and subsequently measure the spectrum of the same ions again at t=Tmix. Li+ diffusion over the grain boundaries between the two chemical Li environments (Li2S and in LPSC) should result in off-diagonal cross-peaks, positioned in the dotted boxes in Figure 3c. The intensity of these cross-peaks reflects the amount of Li+ exchange, which is expected to increase when the diffusion time (Tmix) and temperature are increased4. The absence of off-diagonal intensity, even for the maximum Tmix and temperature (Tmix= 20s, 373K) indicates that the Li+ exchange (flux) over the solid–solid interface between LPSC and Li2S (without LiI coating), is too small to be observed, reflecting sluggish Li+ mobility across the interface with the solid electrolyte. This rationalizes the observation in Figure 1d, that these mixtures do not facilitate activation of Li2S. As expected, the addition of LiI to the cathodic mixture, mLi2S(LiI)-mLPSC, results in the appearance of the Li resonance at -4.56 ppm associated with LiI, in the 1D 6Li NMR spectrum (Figure 3d). The impact of the LiI on the spontaneous Li+ charge transfer, between Li2S and LPSC is dramatic, as can be observed in Figure 3e-i. At short mixing time, Tmix = 10 ms, no appreciable cross-peak intensity is observed in the 2D EXSY spectrum (Figure 3g). However, increasing the mixing time, Tmix, to 10 s, and raising the temperature to 373 K, results in a strong cross-peak intensity (Figure 3h and i), which is a measure of the Li+ exchange between Li2S and LPSC. The evolution of the normalized cross-peak intensity as a function of Tmix measured at a range of temperatures and a Tmix range of 10 ms−10 s is provided in Figure 3e. Exchange between the Li2S and LPSC phases was quantified by fitting the evolution of the cross-peak intensity as a function of Tmix to a diffusion model derived from Fick’s law, described elsewhere24. From the fit, the diffusion coefficient (D) as a function of temperature can be obtained, which in this case pertains to Li-ion transport across the Li2S-LPSC interface. The diffusion coefficients as a function of temperature obtained from the fit are given in Figure 3f. The data for Li2S-LPSC diffusion can be fit to an Arrhenius law, yielding an activation energy of 0.107 eV for the charge transfer. This activation energy equals that of LiI as measured by impedance of the pure phase, suggesting that LiI is responsible for lowering the interfacial barrier between electrode and electrolyte. The value for the activation energy and observed Li+ exchange is comparable to that obtained for interfacial diffusion between nanosized Li2S and nanosized (average particle size ~ 100 nm) LPSC argyrodite (0.10-0.13 eV)3-4. The remarkable conclusion is that, is despite the small ionic contact area of the present micron sized LPSC (average particle size ~ 50 μm) in the cathodic mixtures, the interface transport is improved to such an extent that it matches that of nanostructured mixtures having a much large ionic contact area. The two orders of magnitude difference in diameter between the LPSC in the cathodic mixtures, suggests that the LiI improves the Li+ diffusion over the interface with 4 orders of magnitude.
To understand the role of the LiI in the diffusion, Figure 4a focuses on the exchange of Li in Li2S and LPSC with LiI, hence the three phase Li-ion exchange in the 2D EXSY measurements shown in Figure 3. Figure 4a displays clear exchange of Li+ between LiI and both Li2S and LPSC, reflecting the equilibrium exchange of Li+ between the three phases in the cathodic mixture. This represents a unique view into the charge transport between the coating and the electrode and electrolyte phases in a solid state battery. By measuring and fitting the exchange intensities as a function mixing time and temperature, Figure 4c-f, similar to the evaluation of the direct exchange between Li2S and LPSC, the diffusion coefficients and activation energy over both the LPSC-LiI and LiI-Li2S interfaces is quantified. To the best of our knowledge, this is the first quantification of the local ion diffusion between a coating and its facing solid phases, providing insight in the impact of a coating on the Li-ion transport. The high diffusivity and very low activation energies for charge transfer from Li2S to LiI and LPSC to LiI, 0.142 eV and 0.117 eV respectively, are similar to the overall charge transfer between Li2S and LPSC. This indicates that LiI facilitates the charge transfer and thus functions as the bridge between electrode and electrolyte as summarized schematically in Figure 4b. Apparently, the ductile LiI40 creates grain boundaries between both the electrolyte and electrode material that do not pose an additional barrier for Li-ion diffusion, and the diffusivity of LiI itself dictates the diffusivity between electrode and electrolyte.
3.4. Electrochemical performance
To test the efficacy of the Li2S-LiI cathode in combination with the micron sized LPSC, mLi2S(LiI)-mLPSC-C|mLPSC|In-Li, all-solid-state batteries were assembled, during which only a mild pressure, 2MPa, was applied (see methods section for assembly details). The battery performance is shown in Figure 5. Although the battery can be activated during charge to over 900 mAh/g capacity as shown in Figure 1e, discharge leads to large overpotentials. The increase of the oxidation potential towards 900 mAh/g exceeds the oxidation potential of LiI (2.3 V vs. In-Li) and that of the LPSC electrolyte (2.1 V vs. In-Li), which will result in poorly conducting species near the interfaces that increase the impedance. To prevent this, the battery is charged to specific capacities (fixed charge capacity) as shown in Figure 5a, followed by discharging to a fixed potential (0.8 V vs. In-Li) to achieve complete discharge. After the battery is initially charged to 600 mAh/g directly, the battery was cycled up to 50 cycles at different currents with an average Coulombic efficiency higher than 97.8% (Figure 5b). The 1st, 25th and 50th charge and discharge curves of the mLi2S(LiI)-mLPSC-C|mLPSC|In-Li battery cycled at 0.132 mA/cm2, shown in Figure 5c, demonstrate an ultra-low activation potential of 1.7 V vs. In-Li, amounting an overpotential of only ~100 mV. This is the lowest overpotential reported (Figure 5d and Table S1) to date and the consequence of the facile Li-ion transport induced by the LiI coating as observed by the exchange NMR experiments. Post-mortem XRD analysis after different states of charge, and NMR analysis after 50 cycles (ending in the charged state) of the cycled mLi2S(LiI)-mLPSC-C active materials can be found in Figure S7. After the first charge to 600 mAh/g and even after the 50th charge to 600 mAh/g, only the Li2S is oxidized, towards an amorphous structure, and the micron sized LPSC solid electrolyte remains intact as no decomposition products are observed47. The long cycling stability of the battery with a higher mass loading, 6.4 mg/cm2, is demonstrated in Figure 5e. After 200 cycles, the Coulombic efficiency maintains values exceeding 99.9% resulting in an average Coulombic efficiency of 99.6%. Most notable is that this is achieved in combination with micron sized electrolyte particles, a small ionic contact area, and that further optimization can be expected upon variation of the amount of LiI, the applied pressure, charge capacity and the cathodic mixture.