Giulio Barletta

Perovskite solar cells achieve remarkable power conversion efficiencies, yet operational stability remains a major barrier to large‐scale deployment. Reliable and rapid assessment of device state of health is therefore essential. Conventional electrical diagnostics, such as illuminated current–voltage (J–V) sweeps, provide accurate performance metrics but are time‐consuming and do not resolve spatially localized degradation, motivating non‐invasive imaging‐based alternatives. Here, a deep‐learning framework is introduced to estimate PSC efficiency retention, R _PCE = PCE _t /PCE ₀ , directly from multimodal luminescence imaging acquired during device aging. Each sample combines electroluminescence (EL), open‐circuit photoluminescence (PL _oc ), and short‐circuit photoluminescence (PL _sc ) at an aged state with device‐specific reference images at t  = 0, enabling learning of degradation‐relevant spatial changes. LumPerNet, a compact convolutional neural network, is benchmarked against a spatially homogenized control in which each luminescence channel is replaced by its spatial average while retaining the same learning framework and leakage‐aware protocol. The comparison indicates that global luminescence evolution contains most of the predictive signal, while spatial information provides a secondary contribution to robustness. These results establish spatially resolved luminescence imaging as a practical route for accelerated stability testing and non‐invasive degradation monitoring in perovskite photovoltaics.

We assume that a sufficiently large database is available, where a physical property of interest and a number of associated ruling primitive variables or observables are stored. We introduce and test two machine learning approaches to discover possible groups or combinations of primitive variables, regardless of data origin, being it numerical or experimental: the first approach is based on regression models, whereas the second on classification models. The variable group (here referred to as the new effective good variable) can be considered as successfully found when the physical property of interest is characterized by the following effective invariant behavior: in the first method, invariance of the group implies invariance of the property up to a given accuracy; in the other method, upon partition of the physical property values into two or more classes, invariance of the group implies invariance of the class. For the sake of illustration, the two methods are successfully applied to two popular empirical correlations describing the convective heat transfer phenomenon and to the Newton’s law of universal gravitation.