Investigations into the superionic glass phase of Li4PS4I for improving the stability of high-loading all-solid-state batteries†

In recent years, investigations into improving the performance of bulk-type solid-state batteries (SSBs) have attracted much attention. This is due, in part, to the fact that they offer an opportunity to outperform the present Li-ion battery technology in terms of energy density. Ni-rich Li1+x(Ni1−y−zCoyMnz)1−xO2 (NCM) and lithium-thiophosphate-based solid electrolytes appear to be a promising material combination for application at the cathode side. Here, we report about exploratory investigations into the 1.5Li2S/0.5P2S5/ LiI phase system and demonstrate that a glassy solid electrolyte has more than an order of magnitude higher room-temperature ionic conductivity than the crystalline counterpart, tetragonal Li4PS4I with the P4/nmm space group (∼1.3 versus ∼0.2 mS cm). In addition, preliminary results show that usage of the glassy 1.5Li2S–0.5P2S5–LiI in pellet stack SSB cells with an NCM622 (60% Ni content) cathode and a Li4Ti5O12 anode leads to enhanced capacity retention when compared to the frequently employed argyrodite Li6PS5Cl solid electrolyte. This indicates that, apart from interfacial instabilities, the stiffness (modulus) of the solid electrolyte and associated mechanical effects may also impact significantly the long-term performance. Moreover, SSB cells with the glassy 1.5Li2S–0.5P2S5–LiI and high-loading cathode (∼22 mgNCM622 cm) manufactured using a slurry-casting process are found to cycle stably for 200 cycles at C/5 rate and 45 °C, with areal capacities in excess of 3 mA h cm.


Introduction
Solid-state batteries (SSBs) are currently under intense investigation, primarily because of their prospect of delivering higher energy and power densities than state-of-the-art liquid-electrolyte-based Li-ion batteries. 1,2 Recent work on combining layered lithium metal oxide cathode active materials (CAMs) with lithium thiophosphate solid electrolytes has shown promising results regarding cell capacity and capacity retention. [3][4][5][6][7] This is due, in part, to the high ionic conductivity of both amorphous and crystalline thiophosphate-based solid electrolytes and their comparably low rigidity, allowing intimate physical contact with the CAM secondary particles. [8][9][10] The latter is of particular importance when using Ni-rich materials, such as Li 1+x (Ni 1−y−z Co y Mn z ) 1−x O 2 (NCM) with ≥60% Ni content, as they undergo large volumetric changes upon (de) lithiation. 6,[11][12][13][14][15][16] In general, volume change effects can be mitigated by introducing zero-or low-strain CAMs (at the expense of specific capacity though) 15,17 or somewhat accommodated by the solid electrolyte itself, with mechanically relatively soft lithium thiophosphates being the materials of choice at present.
In the quest for new, superionic solid electrolytes, a tetragonal Li 4 PS 4 I with P4/nmm space group synthesized by a solution-based method has been reported recently and shown to exhibit room-temperature Li-ion conductivities in the range of 6.4 × 10 −5 to 1.2 × 10 −4 S cm −1 , with activation energies for conduction of 0.37 to 0.43 eV. 18 Moreover, density functional theory (DFT) calculations pointed towards an exceptionally high ionic conductivity around 10 −1 S cm −1 . 19 Taken together, these reports and especially the prospects of achieving high ionic conductivity in the xLi 2 S/yP 2 S 5 /zLiI system (amorphous, crystalline or both) [18][19][20][21][22][23][24] motivated us to embark on an investigation into the preparation of the glass (ceramic) phase of Li 4 PS 4 I.
Specifically, we present a comparative study following the changes in the ionic conductivity and crystal structure during mechanochemical synthesis (ball-milling-assisted amorphization of a precursor mixture stoichiometrically equivalent to 1.5Li 2 S-0.5P 2 S 5 -LiI or Li 4 PS 4 I) and post annealing at 155-250°C. The results indicate that, in agreement with the literature on analogous solid electrolytes, the glass phase possesses the highest ionic conductivity and small amounts of crystalline side products or impurities have a relatively strong negative effect on the transport properties (ion dynamics). Moreover, 1.5Li 2 S-0.5P 2 S 5 -LiI-based SSBs with an LiNbO 3coated NCM622 (60% Ni) cathode and a Li 4 Ti 5 O 12 (LTO) anode are shown to outperform (in terms of long-term cycling performance) identical cells with an argyrodite Li 6 PS 5 Cl solid electrolyte and to cycle stably even when using a high-loading, slurry-cast cathode.

Materials and synthesis
For the synthesis of the 1.5Li 2 S-0.5P 2 S 5 -LiI solid electrolyte, 1.5 g of stoichiometric amounts of Li 2 S (Sigma-Aldrich; 99.9%), P 2 S 5 (Sigma-Aldrich; 99%) and LiI (99.99%; Alfa Aesar) were mixed in a planetary ball-mill (Fritsch) under an argon atmosphere for up to 12 h at 450 rpm using a 70 mL zirconia jar loaded with twenty 10 mm diameter zirconia balls. For subsequent heat-treatment, 250 mg of powder were pressed under an argon atmosphere into 10 mm diameter pellets. The pellets were then vacuum-sealed (1.0 × 10 −3 mbar) in quartz ampoules and annealed in a box furnace (Nabertherm) for 12 h at temperatures up to 250°C. The heating rate was 5°C min −1 . The quartz ampoules were dried under dynamic vacuum for 15 min at 650°C prior to usage.
For the synthesis of the Li 6 PS 5 Cl solid electrolyte, 5 g of a mixture of Li 2 S (10 mol% deficiency), P 2 S 5 and LiCl (99+%; Alfa Aesar) was milled under an argon atmosphere, first for 1 h at 250 rpm and then for 20 h at 450 rpm using a 250 mL zirconia jar containing 10 mm diameter zirconia balls with a 30 : 1 ball-to-powder ratio. LiCl was dried in a vacuum for 12 h at 300°C prior to usage. After milling, the powder was heated in a vacuum for 5 h at 300°C. The room-temperature ionic conductivity of the resulting Li 6 PS 5 Cl solid electrolyte was ∼2.0 mS cm −1 (∼7.4 mS cm −1 at 45°C).
Small-sized NCM622 with d 50 = 2.9 μm and d 90 = 6.0 μm was received from BASF SE. 25 A 1 wt% LiNbO 3 coating was applied to the CAM particles prior to usage. 4,26 Characterization The phase composition was studied by powder X-ray diffraction (PXRD) in borosilicate glass capillaries (Hilgenberg) with an inner diameter of 0.48 mm and a wall thickness of 0.01 mm using a STADI P diffractometer (STOE) equipped with a Cu-Kα 1 radiation source.
Rietveld analysis was performed using FullProf software. The Thompson-Cox-Hastings pseudo-Voigt function was used as a profile function and the scale factor, background coefficients (Chebyshev function with 24 parameters), reflection shape parameters, lattice parameters, atomic coordinates (for Li 4 PS 4 I) and zero-shift were refined. Isotropic displacement factors were fixed in the refinement analysis to avoid divergence. 18 Electrochemical impedance spectroscopy (EIS) was performed on an SP-200 potentiostat (BioLogic). For the conductivity measurements, 150 mg of solid electrolyte powder were compressed at 440 MPa in a custom-built setup. A constant pressure of 250 MPa was maintained during the experiment. Impedance spectra were collected in the range of 100 mHz to 7.0 MHz with an AC voltage amplitude of 10 mV. The experimental data were fitted with EC-Lab software (BioLogic) using (R 1 Q 1 )Q 2 and (R 1 Q 1 )(R 2 Q 2 )(R 3 Q 3 )Q 4 equivalent circuits for the solid electrolyte conductivity and SSB measurements, respectively.
Differential scanning calorimetry (DSC) was conducted under an argon atmosphere on samples sealed in aluminum crucibles using a NETZSCH DSC 204 F1 Phoenix. The heating rate was 5°C min −1 .
Scanning electron microscopy (SEM) imaging was performed using a LEO-1530 microscope (Carl Zeiss AG) equipped with a field emission source at 10 kV. Samples for cross-sectional SEM were prepared by breaking them in half.

Electrode preparation and cell testing
The cathode composite for binder-free SSB cells was prepared by mixing NCM622 CAM with 1.5Li 2 S-0.5P 2 S 5 -LiI or Li 6 PS 5 Cl solid electrolyte (7 : 3 weight ratio) and 1 wt% Super C65 carbon black (Timcal) in a planetary ball-mill under an argon atmosphere for 30 min at 140 rpm using 10 mm diameter zirconia balls. 27 The anode composite was prepared by mixing carbon-coated LTO (NEI Corp.), Super C65 carbon black and solid electrolyte in a 3 : 1 : 6 weight ratio. Both LTO and Super C65 carbon black were dried in a vacuum for 12 h at 300°C prior to usage.
A custom-built setup with two stainless steel dies and 10 mm diameter PEEK sleeve was used for the SSB cell assembly. 100 mg of solid electrolyte were compressed first at 125 MPa, followed by 11 mg of cathode composite and 65 mg of anode composite on each side. The pellet stack was then compressed at 440 MPa. Charge/discharge testing was performed at 55 MPa and at 45°C and different rates from 1C to C/5 (1C = 180 mA g NCM622 −1 ) in the voltage range between  28 with a weight ratio of 2.9 : 6.9 : 0.1 : 0.1. The preparation process involved a series of mixing steps. The first step was a dry mixing of the NCM622, 1.5Li 2 S-0.5P 2 S 5 -LiI and carbon black at 600 rpm for 2 min in a Thinky mixer. The resulting powder was then wetted dropwise with the solvent (o-xylene from Sigma-Aldrich; anhydrous, 97%). The amount of o-xylene added was calculated to achieve a solid content of ∼64 wt% in the final slurry. The second step was a wet mixing at 600 rpm for 2 min, after which the binder solution (4 wt% Oppanol® in o-xylene) was added. This mixture was mixed twice at 2000 rpm for 6 min. The slurry obtained was then coated onto an Al foil current collector using a mini tape casting coater from MTI. Finally, the film was dried at room temperature in a twostep drying process, firstly until no visible wet spots remained and finally in a vacuum. All chemicals and processing steps were handled in an argon environment. The same custombuilt setup was used and the sequence of assembly was identical. The slurry-cast cathode was punched into a circular geometry (9 mm diameter) and placed on top of the compressed solid electrolyte. The cells were cycled under a stack pressure of 55 MPa and at C/5 rate and 45°C in the same voltage range of 1. 35

Results and discussion
First, a mixture of Li 2 S, P 2 S 5 and LiI, stoichiometrically equivalent to Li 4 PS 4 I (1.5Li 2 S-0.5P 2 S 5 -LiI), was subjected to ball milling and the evolution of the crystal structure was followed ex situ by PXRD. Fig. 1a shows PXRD patterns collected after different milling times from 7 to 12 h. Reflections of both Li 2 S (Fm3m space group) and LiI (Fm3m space group) were still clearly visible after 7 h. In the further course of milling, they gradually vanished, thereby indicating that progressive amorphization occurred.
In order to probe the effect that ball milling has on the electrical (ionic) conductivity, EIS measurements at 25°C were conducted on compressed pellets. The spectra obtained revealed both a partial semicircle and a capacitive tail (see Fig. S1 of the ESI †). With increasing milling time, the semicircle could hardly be resolved anymore, demonstrating increasing conductivity. Fitting of the experimental data was performed using an (R 1 Q 1 )Q 2 equivalent circuit. The conductivity was calculated from the value of R 1 and found to increase from ∼0.75 mS cm −1 for 7 h to ∼1.3 mS cm −1 (∼3.1 mS cm −1 at 45°C) for 10 h (Fig. 1b). Although the PXRD pattern of the latter sample showed still some minor precursor reflections, longer ball milling did not lead to any further increases in conductivity.
DSC was used to explore the glassy 1.5Li 2 S-0.5P 2 S 5 -LiI regarding phase transitions (Fig. 2a). Both a sharp peak at ∼160°C and a relatively broad and less intense peak centered at 225°C were observed. In order to gain insight into the phase composition(s) around these transitions, the sample after ball milling for 12 h was annealed in vacuum-sealed quartz ampoules for 12 h at different temperatures from 155 to 250°C and then PXRD patterns were recorded (Fig. 2b). At 155°C, just below the first transition region, minor reflections of LiI were visible. Upon increasing the temperature to 175°C, new reflections appeared, which can be assigned to Li 6 PS 5 I (F43m space group) and Li 4 PS 4 I (P4/nmm space group). Higher  temperatures led to more intense and narrower reflections of the tetragonal Li 4 PS 4 I phase due to progressive crystallization. Rietveld refinement of the PXRD pattern for the sample annealed at 250°C yielded lattice parameters similar to those reported for Li 4 PS 4 I synthesized by a solution-based (wet chemistry) method (see Fig. S2 of the ESI †). 18 The presence of crystalline Li 6 PS 5 I and LiI side products was estimated at ∼16 and ∼2 wt%, respectively.
Next, EIS measurements were performed again to follow the evolution of room-temperature ionic conductivity in the annealed samples (see Fig. S3 of the ESI †). The calculated values are displayed in Fig. 2a. Annealing at 155°C already led to a drop in conductivity from ∼1.3 to ∼0.8 mS cm −1 . The conductivity decreased further with increasing temperature, reaching ∼0.1 mS cm −1 after 250°C. We note that the ionic conductivity of the solution-based Li 4 PS 4 I was about the same. 18 Taken together, thermally induced crystallization of tetragonal Li 4 PS 4 I (and some other impurity phases) from the glassy 1.5Li 2 S-0.5P 2 S 5 -LiI was found to negatively affect the conductivity. This suggests that for the thiophosphate iodide material studied in the present work, disorder is beneficial to the Li-ion transport properties, in agreement with literature reports. 20,24 Because the glassy 1.5Li 2 S-0.5P 2 S 5 -LiI shows superior ionic conductivity over crystalline Li 4 PS 4 I, its performance in bulktype ( pellet stack) SSBs was analyzed in some detail. The cells comprised LiNbO 3 -coated NCM622, LTO, 1.5Li 2 S-0.5P 2 S 5 -LiI and Super C65 carbon black. For comparison reasons, identical cells with an argyrodite Li 6 PS 5 Cl solid electrolyte were simultaneously assembled and tested. Long-term cycling was performed at a rate of C/5 and 45°C. The areal loading was ∼10 mg NCM622 cm −2 . As shown in Fig. 3a, initial specific charge and discharge capacities of 201 and 172 mA h g NCM622 −1 were achieved with the 1.5Li 2 S-0.5P 2 S 5 -LiI, corresponding to a coulombic efficiency of ∼86%. In the second cycle, the cells delivered a specific capacity of 172 mA h g NCM622 −1 (∼1.7 mA h cm −2 ). Using the Li 6 PS 5 Cl solid electrolyte (Fig. 3b), initial specific charge and discharge capacities of 201 and 180 mA h g NCM622 −1 were achieved. The difference in first-cycle coulombic efficiency (∼90% versus ∼86%) is indicative of more pronounced side reactions, probably solid electrolyte degradation, in the cells with the glassy 1.5Li 2 S-0.5P 2 S 5 -LiI. Some (minor) capacity fading was apparent during the first five cycles at C/5 rate (Fig. 3c). Upon increasing the C-rate to C/2, the specific discharge capacity dropped to 155 and 145 mA h g NCM622 −1 for the Li 6 PS 5 Cl-and 1.5Li 2 S-0.5P 2 S 5 -LiIbased cells, respectively. This result can be explained by the difference in ionic conductivity, among others. Note that the electrical conductivity strongly affects the charge/discharge kinetics and this is particularly obvious when increasing the C-rate. 29 As expected, the discrepancy between the specific discharge capacities was even larger at 1C rate, with 130 vs. 105 mA h g NCM622 −1 . In the further course of cycling at a rate of C/5, the cells showed similar fading, with specific capacities of ∼127 mA h g NCM622 −1 (∼1.3 mA h cm −2 ) after 100 cycles.
However, distinct differences regarding the capacity retention were noticeable at higher cycle numbers. While the Li 6 PS 5 Cl-based cells showed virtually linear fading, decelerated capacity decay was observed for the 1.5Li 2 S-0.5P 2 S 5 -LiI-based cells, with 82 vs. 105 mA h g NCM622 −1 (∼1.0 mA h cm −2 ) after 200 cycles, corresponding to ∼46 and ∼61% capacity retention, respectively. The improved capacity retention was also somewhat reflected in the coulombic efficiency (see Fig. S4 of the ESI †). In fact, the 1.5Li 2 S-0.5P 2 S 5 -LiI-based cells were found to surpass 99% in the third cycle and it stabilized above 99.5% after 50 cycles (99.8% after 200 cycles). In contrast, for the cells using the Li 6 PS 5 Cl solid electrolyte, the rate of increase of the coulombic efficiency in the initial cycles was slower and even after 200 cycles, it was still below 99.5%. These observations apparently suggest that the glassy 1.5Li 2 S-0.5P 2 S 5 -LiI has a better electrochemical stability than Li 6 PS 5 Cl. However, considering the lower first-cycle coulombic efficiency and its evolution during the initial cycles, it rather seems that a thicker and probably more robust decomposition layer formed at the interfaces with the NCM622 and carbon black additive. 27,30  In order to gain more insight into the factors at play leading to the differences in capacity retention, EIS measurements at 45°C were conducted on the cells after 200 cycles. As can be seen from the Nyquist plots of the electrochemical impedance in Fig. 4a and b, the measured spectra showed a depressed semicircle and a Warburg tail. Curve fitting was performed assuming an (R 1 Q 1 )(R 2 Q 2 )(R 3 Q 3 )Q 4 equivalent circuit, where R 1 is the solid electrolyte bulk (separator) resistance, R 2 is the solid electrolyte grain boundary resistance and R 3 represents primarily the cathode interfacial resistance. 27,31 The calculated values of R 1 are ∼39 and ∼69 Ω for the Li 6 PS 5 Cland 1.5Li 2 S-0.5P 2 S 5 -LiI-based cells, respectively. The larger value in the case of the thiophosphate iodide solid electrolyte can be explained by its lower ionic conductivity. Moreover, solid electrolyte grain boundary and cathode interfacial resistances of ∼7 and ∼32 Ω, respectively, were determined for the latter cells. In comparison, the respective resistances were found to be ∼18 and ∼5 Ω when using the argyrodite solid electrolyte. The higher cathode interfacial resistance in the 1.5Li 2 S-0.5P 2 S 5 -LiI-based cells agrees with the hypothesis that a more pronounced decomposition layer is formed in the initial cycle(s). The fact that the solid electrolyte grain boundary resistance was higher for Li 6 PS 5 Cl compared to the 1.5Li 2 S-0.5P 2 S 5 -LiI might hint at particle fracture (note that mechanical degradation is not necessarily limited to the CAM/ solid electrolyte interface) considering that the solid electrolyte undergoes some deformation upon electrochemical cycling; however, this requires further study.
Apart from (electro)chemical instability issues, mechanical degradation because of volume expansion/contraction of the CAM particles has been recognized to play a key role in the performance of especially pelletized SSB cells. 12,31 For NCM CAMs, the absolute relative volume changes upon charging increase with increasing Ni content. [11][12][13][14][15] Hence, the use of lower stiffness solid electrolytes (higher elasticity)-provided that solid electrolyte degradation can be inhibited effectively, for example, by applying a protective surface coating to the CAM particles-appears to be beneficial to accommodate for local volume changes, thereby reducing the likelihood of mechanical failure.
In order to examine whether any microstructural changes occurred in the cathode layer in both the Li 6 PS 5 Cl-and 1.5Li 2 S-0.5P 2 S 5 -LiI-based cells, cross-sectional SEM images were collected before and after cycling. The spherical NCM622 secondary particles were clearly visible in both cases ( Fig. 5a-d). However, as somewhat expected, more intimate physical contact between the solid electrolyte and the CAM was achieved when using the glassy 1.5Li 2 S-0.5P 2 S 5 -LiI. In addition, the cathode structure/morphology was reasonably well retained after 200 cycles (see also low-magnification SEM images in Fig. S5 of the ESI †). Subtle differences between the cathode layers can probably be attributed to some extent to the different mechanical properties of the solid electrolytes used. The greater mechanical softness of 1.5Li 2 S-0.5P 2 S 5 -LiI seems to help maintain sufficiently good physical contact between the individual components and mitigate void formation. This conclusion is also in agreement with the lower Young's modulus typically observed for amorphous (or glassy) thiophosphate-based solid electrolytes when compared to their crystalline counterparts. [32][33][34][35] Moreover, we note that both NCM622 secondary particle fracture and electrode cracking were not apparent from the imaging data.
The glassy solid electrolyte was also tested in more practical SSB cells. To this end, cathodes of similar composition, but containing 1 wt% polyisobutylene binder, were manufactured using a slurry-casting process. 3,28,36 The areal loading was ∼22 mg NCM622 cm −2 . Specific charge and discharge capacities of 193 and 162 mA h g NCM622 −1 (∼3.6 mA h cm −2 ) were achieved in the initial cycle (Fig. 6a). Interestingly, there was only a minor increase in charge/discharge overpotential (by ∼50 mV), despite the much higher CAM loading and the presence of a polymer binder. In addition, the first-cycle coulombic efficiency was only slightly lower compared to that of cells using the binder-free NCM622 cathode (∼83% versus ∼86%). However, it stabilized above 99.5% within the first five cycles and reached ∼100% after 25 cycles (see Fig. S6 of the ESI †), thus indicating high reversibility. As can be seen from Fig. 6b Finally, SEM and energy dispersive X-ray (EDX) spectroscopy confirmed that the cathode of thickness >100 μm was uniform over the entire cross section, without signs of delamination, major particle displacement or cracking (Fig. 7a-c). Hence, this result provides evidence that mechanical degradation can be mitigated further through implementation of a polymer binder.

Conclusions
In summary, the 1.5Li 2 S/0.5P 2 S 5 /LiI solid electrolyte phase system was characterized with respect to structure and electrical conductivity using PXRD, DSC and EIS. Ball milling-   assisted amorphization of the precursor mixture resulted in increases in room-temperature ionic conductivity up to ∼1.3 mS cm −1 . Subsequent annealing of the glassy 1.5Li 2 S-0.5P 2 S 5 -LiI at temperatures above 150°C decreased the ionic conductivity considerably due to the formation of crystalline LiI, Li 6 PS 5 I and Li 4 PS 4 I. This result therefore suggests that for lithium-thiophosphate-iodide-based solid electrolytes, high conductivity can be achieved by introducing (long-range) structural disorder (i.e., in amorphous or glassy phases), in contrast to chlorine-and bromine-containing lithium thiophosphates. Moreover, pellet stack SSB cells with an NCM622 cathode and a LTO anode using the glassy 1.5Li 2 S-0.5P 2 S 5 -LiI solid electrolyte were found to outperform argyrodite Li 6 PS 5 Cl-based cells in terms of long-term performance. In addition, highloading, slurry-cast cathodes were able to cycle stably (capacity fading of only ∼0.065% per cycle on average at C/5 rate and 45°C) in the same cell configuration. This seems to be due to the 1.5Li 2 S-0.5P 2 S 5 -LiI, which not only allows intimate contact with the NCM622 secondary particles but also improves the microstructural integrity upon electrochemical cycling. In a broader context, the experimental data indicate that the mechanical properties of solid electrolytes must be considered more carefully in developing long-life, bulk-type SSBs.

Conflicts of interest
There are no conflicts to declare.