Process optimization on kesterite-based ceramics for enhancing their thermoelectric performances assisted by active machine learning approach: A tool for metal-sulfide ceramics development

The thermal process parameters are crucial in metal-sulfides ceramics as they affect significantly the resulting physico-chemical properties. In the present work, we investigated the sintering effect in the kesterite Cu 2.125 Zn 0.875 SnS 4 on its structural, microstructural, and thermoelectric (TE) properties to highlight the non-negligible contribution of the thermal process often ignored in metal-sulfide ceramics. For this purpose, we developed an approach combining data science with the conventional material experiment/theory approach which can be used as a tool to shortcut the time-consuming steps of TE material optimization. We confirmed that the optimization and control of the densification process is critical in unravelling the highest potential on metal sulfide TE ceramics with a non-negligible increase of its z T up to 60 %. We propose a scientific tool, the synergic combination of active machine learning with conventional chemistry/theory approaches, to either identify the most proficient sintering process as well as the process to avoid the degradation of the metal-sulfide ceramic properties and thus in a shorten number of experiments. This approach can be extended not only to other metal-sulfide ceramics for thermoelectricity but also to other research fields.