hide
Free keywords:
-
Abstract:
The lithium thiophosphate material class provides promising candidates for solid state electrolytes in lithium ion batteries due to high lithium ion conductivities and low material cost.
Theoretical ab-initio studies probing lithium ion conductivity are constrained in system size and simulated time scales. This limits the transferability of their results to real-world materials in the structure class. Those are characterized by complex thiophosphate microchemistry and grain boundaries influencing the material performance. A method with reduced computational cost, which nevertheless reproduces the material’s complex chemistry, would hence be highly desirable. We present the development of a near-universal atomistic potential for the LPS
material class employing Gaussian process regression. The atomistic potential can describe likewise crystal and glassy materials and different P-S connectivities (PmSn). Furthermore, we apply the ML potential with the aim to probe lithium ion conductivity. The materials studied are crystals (modifications of Li3PS4 and Li7P3S11), glasses of the x Li2S-(100 - x)P2S5 type (x = 67, 70 and 75) and glass-ceramic interfaces.
The obtained material properties for likewise crystals and glasses show a good agreement with results from theory and experiment. For the amorphous materials, the effects of thiophosphate microchemistry and Li2S content on lithium ion conductivity were explicitly investigated. No influence was found for the former, but ion conductivity increases alongside the latter one.