Programmable silicon photonic neuron adds directional control for neuromorphic computing

A new silicon photonic neuron can steer optical spikes in different directions and tune their timing, a step that could make brain-inspired computing systems more flexible. The work, published in Opto-Electronic Science, uses a reconfigurable microring design to control spiking, feedback and signal integration without changing the device geometry.

Why it matters: - Neuromorphic systems need components that can mimic how neurons decide when to fire and how they interact with neighbors. - Directional control in photonic neurons could help future optical networks manage feedback, synchronization and inhibition with more precision. - The device could support heterogeneous photonic neural networks, where individual nodes are tuned for different roles.

What happened: - Researchers developed a programmable silicon photonic neuron that controls the direction and timing of optical spikes. - The work was published in Opto-Electronic Science on Sept. 1, 2026. - The neuron uses engineered non-Hermitian photonics to create direction-dependent responses in a compact silicon device. - The design centers on the Dynamically Reconfigurable Unified Microresonator, or DRUM.

The details: - DRUM lets electrical control adjust coupling between optical modes traveling in opposite directions. - The device can be set to remain quiet, become excitable or operate in a more sensitive spiking state. - No physical change to the device geometry is needed to switch between operating modes. - The programmable behavior affects firing threshold, integration time and refractory period. - Changing optical coupling conditions alters when the device starts spiking and how soon it can respond again. - The neuron can perform temporal integration by accumulating energy from successive optical pulses until it reaches spike threshold. - The timing of the spike can be programmed by changing the coupling phase. - Back-action becomes part of the device’s operation instead of being treated only as interference. - The architecture gives adjustable control over backward-propagating signals. - Simulations of two connected photonic nodes showed inhibition and synchronization. - The network behavior has not yet been experimentally demonstrated.

Between the lines: - The design tackles a core limitation of conventional microresonators, which tend to respond to forward and reverse light in closely linked ways. - Separating directionality, feedback and back-action gives designers more freedom to shape how photonic neurons communicate. - The result points toward optical computing hardware that is more adaptable than fixed-function resonator designs.

What's next: - The next step is to move from single-device behavior to experimentally verified networked photonic neurons. - Future systems could use different thresholds, timing characteristics and feedback responses for different nodes. - That could make it easier to build larger neuromorphic photonic networks that combine specialized components.