Experimental discovery of structure–property relationships in ferroelectric materials via active learning

Emergent functionalities of structural and topological defects in ferroelectric materials underpin an extremely broad spectrum applications ranging from domain wall electronics to high dielectric electromechanical responses. Many these have been discovered quantified via local scanning probe microscopy methods. However, the search has until now based on either trial error, or using auxiliary information such as topography structure identify potential objects interest basis intuition operator pre-existing hypotheses, with subsequent manual exploration. Here we report development implementation a machine learning framework that actively discovers relationships between polarization-switching characteristics encoded hysteresis loop. The loops their scalar descriptors nucleation bias, coercive bias loop area (or more complex functionals shape) corresponding uncertainties are used guide discovery automated piezoresponse force spectroscopy experiments. As such, this approach combines power methods learn correlative high-dimensional data, well human-based physics insights into acquisition function. For materials, workflow demonstrates path sampling points on- off-field largely different, indicating dominated by different mechanisms. proposed is universal can be applied range modern imaging other modalities electron chemical imaging.