Paper detail

Hyperspectral nanoscale mapping of hybrid perovskite photophysics at the single grain level

Hybrid organic-inorganic perovskites have drawn significant interest for applications in optoelectronics over the last few years. Despite rapid progress in understanding the photophysics of perovskite, there remains a need for improved understanding of the effect of microstructure on perovskite photophysical processes. Here, we combine unsupervised machine learning and cathodoluminescence microscopy of a prototypical hybrid perovskite film to decode photophysical processes that are otherwise lost with conventional Gaussian image processing. Hyperspectral maps are decoded with non-negative matrix factorization, revealing components relating to primary band-edge emission, photon recycling, and defect emission. A blind-spectral non-negative matrix factorization procedure provides additional understanding of changes in an intermediate perovskite phase under electron beam exposure and illustrates how traditional Gaussian techniques may hide relevant emission features that are critical to the development of environmentally robust perovskite devices