Photonic memristors are a key building block for neuromorphic optical computing, but integrating them with waveguide platforms has usually required complex, expensive heterogeneous fabrication. A team at Shandong University (Xu Gao, Yang Tan, Feng Chen and colleagues) now shows that thin-film lithium niobate (TFLN), a mainstream photonic platform, can itself be turned into a memristive material.
The trick is CMOS-compatible oxygen-ion implantation. It creates spatially separated oxygen-doped and defect-rich layers in the crystal, which enables all-optically controlled refractive-index modulation. In other words, the state of the material (its refractive index, read out through the shift of a microring resonator's transmission spectrum) depends on its illumination history, the defining property of a memristor — and the underlying photorefractive mechanism is a property of the matter itself rather than of an added device.
Reported figures: 3-bit multilevel non-volatile memory, switching energy of hundreds of femtojoules per bit, sub-microsecond optical response and long data retention over extended operation.
Why it matters: wafer-scale TFLN plus ion implantation offers a homogeneous, scalable and programmable material platform for integrated photonic circuits, with potential for large-scale photonic neural networks.
Paper: Gao, X. et al., "All-optical thin-film lithium niobate memristor", Nature Photonics (published 6 November 2026), https://doi.org/10.1038/s41566-026-02048-1
Source page: https://www.nature.com/articles/s41566-026-02048-1
No comments:
Post a Comment