Mechanism study on machine learning–based process optimization and ethanol-induced performance enhancement of Fe(Se,Te) memristors

In recent years, memristors have captured considerable interest owing to their distinct resistive switching (RS) capabilities and adjustable chemical composition, positioning them as promising elements for neuromorphic computing and nonvolatile memory systems. This study introduces a novel approach-applying machine learning (ML) to refine the fabrication process of Fe(Se,Te) thin films tailored for memristive applications. Notably, the devices derived from these films demonstrate reliable and enhanced RS behavior. Interestingly, when subjected to ethanol treatment, the memristive effect becomes more pronounced, hinting at a chemical modulation pathway. Experimental observations suggest that ethanol, under an external electric field, facilitates the creation of conductive filaments within the film, thereby increasing the resistance contrast between high and low states by a factor of approximately 2–7. This enhancement also leads to a substantial expansion of the memory window. By bridging data-driven modeling with experimental synthesis, our study not only advances material optimization strategies but also reveals a compelling new use for Fe(Se,Te) films-as sensitive detectors for alcohol-based compounds.