# Excited State Opto‐Ionic Reservoir Computing in Hybrid Perovskite Electrochemically‐Gated Luminescent Cells

## TL;DR

A new computing system uses light and ionic changes in perovskite crystals to perform complex calculations efficiently.

## Contribution

A novel neuromorphic computing platform using opto-ionic modulation in perovskite microcrystals for high-dimensional computation.

## Key findings

- The system achieves 800 pJ per node-operation energy efficiency with 106 nodes per cm².
- It robustly distinguishes 4-bit pulse sequences with 87% mean accuracy.
- Excited-state dynamics in perovskites are shown to support nanoscale, high-speed computing.

## Abstract

We introduce a neuromorphic reservoir computing concept that leverages the complex interplay between electronic and ionic states in lead halide perovskites to run algorithms by harnessing opto‐ionic modulation of photoexcited state populations. The system leverages the heterogeneous material microstructure and ultrafast spatio‐temporal electronic state dynamics in perovskite microcrystals to generate a high‐dimensional internal state space reservoir within the charge carrier populations. This reservoir exhibits complex, nonlinear, and adaptive behavior. The computation output is read directly from the photogenerated luminescence using diffraction‐limited resolution with 106 nodes per cm2 and energy of 800 pJ per node‐operation. The system performs robustly in distinguishing 4‐bit pulse sequences with a mean accuracy of 87%, showcasing its potential for neuromorphic computing tasks. Our work reveals excited‐state dynamics as a platform for exploring nanoscale computing with photoactive materials, also at high speeds using ultrafast photophysics, with large potential for the development of next‐generation neuromorphic technologies.

A neuromorphic computing system exploiting opto‐ionic modulation in lead halide perovskite microcrystals demonstrates high‐dimensional reservoir dynamics with diffraction‐limited node resolution. Leveraging ultrafast excited‐state interactions, it achieves efficient computation (800 pJ/node‐operation), robustly distinguishing 4‐bit pulse sequences, paving the way for next‐generation nanoscale, high‐speed neuromorphic platforms based on photophysics.