Optimize Before You Synthesize-Enhancing the Ionic Conductivity of Li₇SiPS₈ Using Bayesian Optimization

Tetragonal ${\mathrm{Li}}_7{\mathrm{SiPS}}_8$ is a superionic solid electrolyte, yet its Li ion conductivity suffers from the presence of an amorphous side phase. Attempts to optimize the ionic conductivity, however, are incremental and hence time-consuming, because the relationship between synthesis conditions and electrolyte performance is largely unknown. In this work, we employ Bayesian optimization (BO) as an efficient design-of-experiment approach to increase the ionic conductivity of the ${\mathrm{Li}}_7{\mathrm{SiPS}}_8$ system. Our data-driven workflow reproducibly yields ${\mathrm{Li}}_7{\mathrm{SiPS}}_8$ with ionic conductivities exceeding 7 mS  ${\mathrm{cm}}^{-1}$ at room temperature, an increase by up to $350\%$ compared to previously reported routes. Simultaneously, the optimized solid-state synthesis lowered the synthesis temperature by 100 K ( $20\%$ ) and shortened the reaction time by 76 h ( $76\%$ ), delivering a more energy-efficient and, hence, sustainable process. To probe the origin of the increased conductivity, we examined six representative samples by quantitative Rietveld refinements, synchrotron x-ray powder diffraction, pair distribution function analysis, solid-state and pulsed-field-gradient NMR, electron microscopy, and Raman spectroscopy. We demonstrate that BO can help navigate the complex synthesis parameter space, thereby accelerating the development of high-performance sulfide electrolytes for next-generation batteries.