Data-Driven Scanning Probe Microscopy for Understanding Carrier Dynamics in Lead Halide Perovskite Semiconductors

Abstract

Scanning probe microscopy (SPM) plays a pivotal role in advancing our understanding ofsemiconductor materials by enabling nanoscale characterization of structural and functional properties. Dynamic SPM techniques provide insight into transient charge carrier dynamics in semiconducting materials. Time-resolved electrostatic force microscopy (trEFM) is one such dynamic SPM method which measures nanoscale surface potential dynamics with high spatial and temporal resolution, providing information about key properties such as local photovoltaic quantum efficiency, electronic carrier recombination, and ion motion. The multi-dimensional datasets obtained from trEFM—combining both spatial andtemporal information—are ripe opportunities for the application of data science and machine learning (ML) techniques. The application of ML methods, such as neural networks, can significantly enhance the extraction of meaningful dynamics from noisy experimental data, allowing for a more accurate understanding of the spatial heterogeneity of electronic and ionic carrier phenomena. In this work, we examine lead halide perovskites with trEFM. Lead halide perovskites arean exciting class of semiconducting materials with applications in photovoltaics, light emission, and radiation detection. Their ability to achieve high efficiencies, combined with solution- processability and defect tolerance, has led to rapid advancements in perovskite-based optoelectronic devices. The performance of these materials is influenced by the intricate interplay between their crystallographic and morphological features, which often span from nanometers to micrometers in size. Using trEFM, we can gain detailed insights into how local changes in surface potential relate to the structural features of perovskite films, ultimately providing a deeper understanding of their optoelectronic behavior. We find that the surface recombination velocity, as modified by defect passivation, and defect-enabled ion motion are the primary factors that influence the spatial heterogeneity we observe in polycrystalline lead halide perovskite. We first present two ML models that enable advanced signal processing for trEFM dynamics, exemplifying the natural harmony between large-data dynamic SPM methods and machine learning. We then explore the factors (surface recombination velocity and ion motion) that govern the trEFM dynamics we observe. Overall, this work demonstrates how combining advanced SPM techniques with ML can significantly enhance our ability to study and optimize next-generation semiconductor materials.