High-throughput materials discovery via combinatorial library and ML-enabled autonomous scanning probe microscope

Combinatorial libraries, coupled with rapidly advancing AI, offer a promising pathway for high-throughput materials discovery across a range of critical applications. However, their broader deployment is hindered by the lack of fast, reliable methods for mapping composition–property relationships. Scanning probe microscopies (SPM), including piezoresponse force microscopy (PFM), hold significant potential for providing quantitative and functionally relevant readouts from combinatorial libraries. Here, we demonstrate the use of machine learning-enabled, fully automated SPM for high-throughput characterization of such libraries. Case studies include PFM-based exploration of a ferroelectric ternary library (Al,Sc,B)N, a thickness-gradient HfZrO₃ (HZO) library, and topographic mapping of a high-entropy alloy library. We show that automated SPM accelerates the discovery process by one to two orders of magnitude compared to conventional grid-based searches, while simultaneously offering insight into growth dynamics via topographic imaging and functional properties through spectroscopic imaging. Furthermore, the investigation of the HZO thickness library highlights a unique opportunity to disentangle genuine ferroelectric piezoresponse from electrochemically induced PFM signals by correlating topography and piezoresponse variations with composition and writing voltage on the same sample.