Afterglow-Mediated UV-to-Visible Temporal Preprocessing with Hybrid Optoelectronic Synaptic Transistor Readout for Neuromorphic Recognition

Abstract Ultraviolet signals in practical sensing scenarios often appear as weak perturbations, short pulses, intermittent exposures, or noisy spatial patterns. Conventional instantaneous readout can detect UV inputs at a given moment, but is limited in retaining the exposure history and filtering temporally uncorrelated noise. Here, we propose an afterglow-mediated UV-to-visible-to-current neuromorphic sensing strategy that integrates optical temporal preprocessing with a hybrid optoelectronic synaptic transistor readout. A SrAl2O4:Eu2+, Dy3+ long-afterglow layer embedded in a polydimethylsiloxane matrix was used to construct an all-optical afterglow synaptic unit. Under 365 nm UV excitation, the device produces a persistent green afterglow centered at approximately 515 nm and exhibits synapse-like temporal decay and pulse-dependent accumulation. Transient or intermittent UV inputs are thereby converted into green exposure-memory maps containing spatial and temporal information. The CdSe/ZnS quantum dot/PDVT-10 hybrid optoelectronic synaptic transistor was further fabricated to read out the wavelength-matched afterglow signal. Under 515 nm modulation, the transistor converts the afterglow into tunable synaptic current states with optical potentiation and electrical depression. On the basis of the extracted afterglow dynamics and calibrated optoelectronic response, a device-informed artificial neural network model was constructed. Compared with direct recognition of instantaneous UV patterns, the afterglow-preprocessed route improves the recognition accuracy from 76.5% to 97.0% and maintains higher robustness under increased noise density and varied UV pulse numbers. This work provides an interpretable material-device system route for nonvisible-light neuromorphic sensing.