Oxygen Vacancy Engineering via High-Ionization-Potential Doping for Neuromorphic Visual Reservoir Computing

In this work, achieving a critical balance between defect activity and lattice stability is demonstrated in Sn-doped oxides, leveraging Sn’s high ionization potential. Optimizing the Sn concentration to 8.6 at. % enhances the distribution and concentration of oxygen vacancies, significantly improving the performance of indium zinc tin oxide (IZTO) neuromorphic devices. The devices exhibit superior bias stability and enhanced memory characteristics under optical pulses, with a 26% increase in photocurrent retention. Uniform oxygen vacancies enable persistent photoconductivity-driven facilitation and memory transitions, mimicking neuromorphic preprocessing functions such as high-pass filtering for visual edge enhancement. Furthermore, mapping the conductive state evolution to a reservoir computing framework achieves 87% accuracy in clothing category classification, demonstrating robustness under noise. This work establishes high-ionization-potential element doping as a strategic approach to stabilize optoelectronic neuromorphic devices and advance neuromorphic computing-hardware integration.