# Compact, reconfigurable, and scalable photonic neurons by modulation-and-weighting microring resonators

## TL;DR

This paper introduces a scalable photonic neuron using microring resonators that can perform both modulation and weighting, enabling efficient and flexible neuromorphic computing.

## Contribution

The novel architecture combines modulation and weighting in a single microring resonator, enabling compact, reconfigurable photonic neurons with low power and high compute density.

## Key findings

- A 10-microring resonator array achieved a 3×3 convolution with <5% error and high-frequency financial time-series prediction.
- Each modulation-weighting element occupies 80×45 μm² and consumes 0.186 mW, achieving 4.67 TOPS/s/mm² compute density.
- The on-chip tuning efficiency reached approximately 105 TOPs/W, comparable to state-of-the-art implementations.

## Abstract

Neuromorphic photonics promises sub-nanosecond latency, ultrawide bandwidth, and high parallelism, but practical scalability is constrained by fabrication tolerances, spectral alignment, and tuning energy. Here, we present a large-scale, compact, and reconfigurable photonic neuron in which each microring performs modulation and weighting simultaneously. By exploiting both carrier and thermal tuning within a single device, this architecture reduces footprint, relaxes spectral alignment requirements to just two optical components, and yields a steep transfer response that lowers tuning energy. The proposed neuron supports multiple operating configurations, allowing its dynamical behavior to be adapted to different computational tasks. In particular, a short electrical feedback path enables recurrent operation, providing tunable short- and long-term memory for temporal processing. Using a 10-microring resonator array, we demonstrate both spatial and temporal computing, including a 3\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\times$$\end{document}×3 convolution for image processing with an error of <5% and high-frequency financial time-series prediction. Each modulation-weighting element occupies 80\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\times$$\end{document}×45  \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\upmu$$\end{document}μm\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$^2$$\end{document}2 and consumes an average of 0.186 mW, corresponding to a compute density of 4.67 TOPS/s/mm\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$^2$$\end{document}2. Excluding electronic power, the on-chip tuning efficiency reaches approximately 105 TOPs/W, which is comparable to state-of-the-art implementations. These results indicate that modulation-and-weighting microring resonator banks provide a scalable building block for large-scale neuromorphic photonic systems, offering a favorable combination of compact footprint, low power consumption, and functional flexibility.

The online version contains supplementary material available at 10.1186/s43593-026-00122-3.