Self-supervised deep learning for tracking degradation of perovskite light-emitting diodes with multispectral imaging

Emerging functional materials such as halide perovskites are intrinsically unstable, causing long-term instability in optoelectronic devices made from these materials. This leads to difficulty in capturing useful information on device degradation through time-consuming optical characterization in their operating environments. Despite these challenges, understanding the degradation mechanism is crucial for advancing the technology towards commercialization. Here we present a self-supervised machine learning model that utilizes a multi-channel correlation and blind denoising to recover images without high-quality references, enabling fast and low-dose measurements. We perform operando luminescence mapping of various emerging optoelectronic semiconductors, including organic and halide perovskite photovoltaic and light-emitting devices. By tracking the spatially resolved degradation in electroluminescence of mixed-halide perovskite blue-light-emitting diodes, we discovered that lateral ion migration (perpendicular to the external electric field) during device operation triggers the formation of chloride-rich defective regions that emit poorly—a mechanism that would not be resolvable with conventional imaging approaches. Halide perovskites are promising materials for light-emitting devices, given their narrowband emission and solution processability. However, detailed information on device degradation during operation is required to improve their stability, and this is challenging to obtain. Ji et al. propose a self-supervised deep learning method to capture multi-dimensional images of such devices in their operating regime faster than allowed by conventional imaging techniques.