Introduction

One of the key strategies to achieve carbon neutrality is the development of all-solid-state batteries. Conventional liquid batteries utilize flammable organic solvents as electrolytes, raising safety concerns such as leakage and resulting ignition. In contrast, the electrolyte in an all-solid-state battery is solid, effectively addressing the safety issues associated with liquid-based batteries. Moreover, all-solid-state batteries can outperform liquid batteries in practical applications, especially in the automotive and stationary sectors. They offer advantages such as a high lithium-ion transport ratio and high energy output1,2, enabling high voltage operation and enhanced capacity. Additionally, all-solid-state batteries can utilize high-capacity sulfur-based electrode materials3, which are incompatible with liquid electrolytes owing to their high reactivity with solvents4. As a result, extensive research is underway to explore and advance the potential of all-solid-state batteries in achieving these desirable features5,6.

One of the crucial components in the advancement of all-solid-state batteries is the solid electrolyte. Among various solid electrolytes, the sulfide-based systems7,8,9,10,11,12,13 exhibited high ionic conductivity. The Li2S–P2S5 system demonstrates different glassy and crystalline structures depending on the composition and heat treatment temperature. In the Li2S:P2S5 = 75:25 system, the glass exhibits a conductivity of ~ 10−4 S/cm, while the beta crystals and other structures in the high-temperature range have conductivity < 10−4 S/cm, similar to that of glass. In contrast, the Li2S:P2S5 = 70:30 system exhibits Li7P3S11 crystal with a remarkable ionic conductivity of 3.2 × 10 mS/cm14,15. However, this system contains P2S7 anions, so H2S generation is more likely to occur16. To address the H2S generation issue, researchers have developed solid electrolytes comparable to liquid electrolytes. An example of such a solid electrolyte is Li10GeP2S12, commonly known as LGPS, developed by Kanno, which exhibits a conductivity of 12 mS/cm17,18,19. Li9.54Si1.74P1.44S11.7Cl0.3 is another important solid electrolyte with a conductivity of 25 mS/cm20. These systems do not contain the P2S7 anions. However, these materials rely on the costly rare metal Ge and face challenges in terms of electrical stability. As a result, alternative materials using Sn or Si instead of Ge21,22 and designs incorporating O in place of S have also been explored23.

High ionic conductivity and the absence of rare metals are crucial for the industrial implementation of sulfide-based solid electrolytes. Herein, we have successfully developed a solid electrolyte material with high ionic conductivity by incorporating halogen elements into Li3PS4 glass, which serves as the base material24,25,26,27,28. In this study, we present a comprehensive analysis of the detailed structure of the developed material.

Experimental

Material synthesis

Li2S (> 99.9%, Idemitsu Kosan, Japan), P2S5 (> 99%, Sigma-Aldrich Japan, Japan), and LiBr (> 99.99%, Fujifilm Wako Pure Chemicals Corporation, Japan) were used as raw materials. These materials were carefully weighed and mixed in a glove box under an argon atmosphere with a dew point of − 80 °C. The aim was to obtain 10 g of each material, following the composition ratios shown in Table 1. The powders and ten zirconia balls with a diameter of 10 mm were added to a fully sealed alumina pot and placed in a planetary ball mill (Fritsch: Model No. P-7).

Table 1 Sample composition ratios and ionic conductivities.
Full size table

Initially, the planetary ball mill was operated at a low speed of 100 rpm for a few minutes to facilitate mixing. The rotation speed was then gradually increased to 370 rpm. Mechanical milling was conducted at 370 rpm for 20 h to synthesize the glass. The resulting glass was subjected to heat treatment at various temperatures to obtain glass ceramic, as listed in Table 1. The samples were treated at the desired temperature for 2 h under an argon atmosphere and then cooled to room temperature.

Characterization

This study examines the powder X-ray diffraction (XRD) measurements using a Smart Lab instrument (Rigaku, Japan). The diffraction angle (2θ) range of 10°–60° was scanned with a step size of 0.02° at room temperature (290–300 K). Before measurement, we meticulously positioned each sample in a glass holder with a smooth surface inside a glove box under an argon atmosphere, maintaining precise control over the dew point. The glass holder was then sealed using Kapton film tape.

High-energy X-ray scattering measurements using synchrotron radiation were carried out at BL04B2 of SPring-829,30. The X-ray system employed incident energy of 61.339 keV, consisting of four CdTe and two Ge detectors. To enable X-ray diffraction, we hermetically sealed all samples in 2.0-mmΦ borosilicate glass capillaries (WJM-Glas/Muller GmbH) inside a glove box under an argon atmosphere with precise dew point control. The required scattering intensities I(Q) were meticulously corrected for background, Compton scattering, polarization, and absorption. Subsequently, the structure factor S(Q) was calculated, and the reduced pair distribution function G(r) was obtained using the Fourier transform of S(Q).

Density functional theory structure optimization calculations were performed using BLYP/GGA with DMol3 from Material Studio31,32. Structural refinement was conducted on laboratory X-ray diffraction (lab-XRD) data using the Rietveld refinement method33. The calculation of Li conduction paths was accomplished using the SoftBV34,35,36. VESTA was used for the three-dimensional display of the paths37.

Results and discussion

Figure 1 depicts the heat flow behavior of glassy LiPS-Br annealed at 220 °C and the corresponding ionic conductivity at each annealing temperature. The XRD patterns shown in Fig. 2 reveal that within the temperature range of 190 °C–210 °C, a previously unidentified crystalline phase emerged as the primary source of ionic conductivity, surpassing 1 mS/cm. Conversely, when the temperature exceeded 220 °C, β-crystals became evident, considerably decreasing ionic conductivity. The newly observed unknown phase is presumed to possess a metastable structure, evident from its broad half-width and limited temperature range within a single phase.

Figure 1
figure 1

Heat flow of the glassy LiPS-Br annealed to 220 °C and the ionic conductivity at each annealing temperature.

Full size image
Figure 2
figure 2

XRD patterns at each annealing temperature.

Full size image

Figure 3 illustrates the experimental X-ray pair distribution function for the glass and the crystal after annealing at 200 °C, with the Li3PS4 glass serving as a reference. A comparison of the glass structures reveals a P–S correlation within the PS4 molecule at a distance of 2.0 Å, an S–S correlation within the same molecule at 3.4 Å, and a peak at ~ 4.0 Å originating from an S–S correlation between the molecules8. Adding Br does not significantly affect the glassy structure, as the peak positions and pair distribution functions (PDFs) above 5 Å remain almost identical before and after adding Br, despite variations in peak intensities caused by compositional differences. The area ratio of the P–S correlation located at 2.0 Å and the X-ray weight factor ratio was calculated to be 0.754 and 0.758, respectively. The difference in peak intensity can be attributed to compositional disparities, but there is no notable distinction in the P–S correlation or the presence of PS4 molecules within each glass structure. Adding Br increases the intensity of the peak located at 4.0 Å, indicating the presence of a correlation associated with Br. Comparing the Br added glass and crystal, no changes are observed in the P–S correlation within the PS4 molecule appeared at 2.0 Å or the S–S correlation within the same molecule located at 3.4 Å, confirming that no alterations have occurred within the PS4 molecule from glass to the crystal. However, the crystals exhibit an intensity increase in the peak observed at ~ 4.0 Å, corresponding to the S–S distance between PS4 molecules. Based on the observed correlation related to Br at this distance, it can be deduced that alternations occur in the correlation between PS4 molecules, between PS4 molecules and Br− anions, or among the Br− anions them can be inferred, given the correlation related to Br at this distance. Furthermore, a peak indicative of the ordering of the molecular arrangement is observed in the crystal above 5.0 Å.

Figure 3
figure 3

Comparison of Br-doped glass and glass ceramic with a pure Li3PS4 glass as a reference.

Full size image

Herein, we investigate the high ionic conductivity observed in Br-doped glass ceramic, which exhibit comparable performance to Li7P3S11, comprising PS4 and P2S7 molecules. Therefore, we determined the specific location of the Br− anion within the glass ceramic, which has an almost crystalline structure. The newly formed crystalline phase is considered metastable owing to its broad diffraction peak width, which poses challenges for applying Rietveld refinement without a hypothesis. To address this, we conducted XRD calculations for structurally stable configurations verified via self-consistent field calculations using density functional theory (see Fig. S1). Several initial structures were assumed, and insertion and substitution-type structural models were obtained based on the LGPS structure, as depicted in Fig. 4. As shown in the figure, the insertion type involves the presence of Br− ions within the gaps between PS4 molecules. Given that the PS4 molecule carries a charge of − 3 and the GeS4 molecule in LGPS structure is − 4, there are likely three or four Br− ions associated with each PS4 molecule. Br− ions tend to form clusters together with Li ions.

Figure 4
figure 4

Insertion type (a) and substitution-type (b) structural models of Br-doped glass ceramic.

Full size image

Figure S1a represents the structure without Br, while Fig. S1b shows the addition of one Br atom to both the (0,0,0) and (0.5,0.5,0.5) sites. The circle in the in the figure represents the experimental XRD, while the blue line represents the calculated XRD based on the structural model. One Br added to the model results in an approximate LiBr ratio (mol%) of 7.7. Considering a more realistic scenario, we added approximately two Br ions, resulting in a LiBr ratio of ~ 14.3 mol%. Additionally, we investigated the substitution-type structural model using the same approach. Four S ions at positions (0.00000, 0.30099, 0.79168), (0.00000, 0.69901, 0.79168), (0.20443, 0.50000, 0.59727), and (0.79557, 0.50000, 0.59727) on the P1 site were replaced with one to four Br− ions, as shown in Fig. S1c–e. As depicted in Fig. S1f,g, we examined structural models for insertion and substitution types. The results indicated that the substitution-type with four Br atoms at the P1 site closely matched the experimental data. The PDF G(r) for these two models was calculated and compared to validate the insertion and substitution types in terms of local structure. The experimental data are represented by black lines, the calculated insertion type by light green solid lines, and the substitution-type by blue solid lines (see Fig. 5). It is evident from these figures that the peak reproducibility is higher for the substitution-type compared to the insertion type. Furthermore, the substitution-type structural model accurately reproduced the characteristic peak appeared at ~ 4.0 Å, confirming its presence in the system with Br. Therefore, it is highly likely that Br forms clusters at the P1 site.

Figure 5
figure 5

Comparison of G(r) calculations and experimental data for insertion type (a) and substitution-type (b) structural models of Br-doped glass ceramic.

Full size image

Based on our consideration, we chose the substitution-type structural model and replaced the P1 site with a Br cluster. Structural optimization was then carried out using Rietveld analysis. The results revealed that the occupancy of Br is 0.15 ~ 0.25, suggesting that 15 ~ 25% of PS4 molecules are replaced by the Br cluster. Figure 6 shows that incorporating an 8% Br cluster in the structural model accurately reproduces the XRD pattern. Detailed structural parameters can be found in Table S1. Notably, the atomic displacement of P in the Br cluster differs significantly from the unsubstituted structural model, indicating substantial anion fluctuations within the structure compared to the pure unsubstituted model.

Figure 6
figure 6

Rietveld refinement results using a substituted structural model.

Full size image

Figure 7a illustrates the Li-ion conduction pathway investigated by depicting the ΔE = 0.80 eV in the Br cluster-substituted structural model. The presence of Br clusters hinders the pathway for Li-ion conduction, while the pathway emerges in regions without Br. Upon examining the overall structural model, it is evident that the paths do not form in the c-axis direction but rather in the b-axis direction. The calculations for β-Li3PS4, which exhibits low conductivity, yield similar results as depicted in Fig. 7b, indicating the absence of such pathways. In contrast, the structural model of LGPS with high ionic conductivity displays a formation of a three-dimensional pathway, as depicted in Fig. 7c. The presence of Br is considered a contributing factor to the enhanced ionic conductivity through halogen addition. However, the limited formation of high-conductivity pathways along the c-axis in the vicinity of Br likely accounts for the disparity in ionic conductivity compared to LGPS.

Figure 7
figure 7

Visualization of Li-ion conduction paths using the Bond Valence Sum method. (a) Br-substituted LiPS in this work, (b) β-Li3PS4, and (c) LGPS systems.

Full size image

Conclusion

Br-doped Li3PS4 exhibits a minimal presence of P2S7 molecules that effectively suppresses the generation of H2S. Instead, it forms an LGPS-like structure and demonstrates high ion conductivity. This characteristic makes it a promising material for solid electrolytes in all-solid-state batteries, striking a balance between desirable properties.

The local structural characterization conducted through high-energy XRD and PDF analysis confirms that the molecular framework structure of the precursor glass remains unaltered. Upon converting to glass ceramic and subsequent crystallization, PS4 sites randomly incorporate Br atoms. Furthermore, using the Bond valence sum method to visualize Li-ion conduction pathways in the structural model, one can observe that Br facilitates the formation of ionic conduction paths resembling those found in LGPS. However, the formation of pathways along the c-axis direction is insufficient owing to structural geometry limitations.

This study emphasizes the importance of structure control in glass ceramic as it can pave the way for developing novel glass/glass–ceramic solid electrolyte materials in future research.