Automated Materials Discovery Platform Realized: Scanning Probe Microscopy of Combinatorial Libraries

TL;DR

This paper presents an automated scanning probe microscopy platform for high-throughput, composition-dependent characterization of ferroelectric properties in combinatorial materials libraries, integrating spectroscopy, photoluminescence, and Bayesian optimization.

Contribution

It introduces a fully automated SPM framework combined with multi-modal analysis and Bayesian optimization for accelerated materials discovery.

Findings

Identified the morphotropic phase boundary in SmBFO with enhanced ferroelectric response.

Discovered ferroelectric behavior at the phase-stability boundary in (Al,Sc,B)N.

Demonstrated autonomous exploration using Bayesian optimization to efficiently navigate material properties.

Abstract

Combinatorial materials libraries provide a powerful platform for mapping how physical properties evolve across binary and ternary cross-sections of multicomponent phase diagrams. While synthesis of such libraries has advanced since the 1960s and been accelerated by laboratory automation, their broader utility depends on rapid, quantitative measurements of composition-dependent structures and functionalities. Scanning probe microscopies (SPM), including piezoresponse force microscopy (PFM), offer unique potential for providing these functionally relevant, spatially resolved readouts. Here, we demonstrate a fully automated SPM framework for exploring ferroelectric properties across combinatorial libraries, focusing on binary Sm-doped BiFeO3 (SmBFO) and ternary Al$_{1-x-y}$Sc$_x$B$_y$N (Al,Sc,B)N systems. In SmBFO, automated exploration identifies the known morphotropic phase boundary with enhanced ferroelectric response and reveals a previously unreported double-peak fine structure. In the (Al,Sc,B)N library, ferroelectric behavior emerges at the phase-stability boundary, correlating with variations in morphology and defect concentration. By integrating automated SPM with wavelength-dispersive spectroscopy (WDS) and photoluminescence mapping, we resolve the composition-morphology-defect-property relationships underlying ferroelectric response and demonstrate a pathway toward a multi-tool, high-throughput characterization platform. Finally, we implement Gaussian-process-based single- and multi-objective Bayesian optimization to enable autonomous exploration, highlighting the Pareto front as a powerful framework for balancing competing physical rewards and accelerating data-driven physics discovery.