Home > Research Content
Recently, a research team led by Professor Chen Feng from the School of Physics at Shandong University developed an all-optical thin-film lithium niobate memristor. The findings were published in Nature Photonics under the title “All-optical thin-film lithium niobate memristor.” Postdoctoral researcher Gao Xu is the first author, and Professors Tan Yang and Chen Feng are the corresponding authors. Shandong University is the sole affiliated institution.
Photonic memristors are key components for neuromorphic optical computing and in-memory computing. However, integrating them with optical waveguides typically requires additional functional materials, increasing fabrication complexity and cost. Thin-film lithium niobate (TFLN) is an established photonic platform, but it inherently lacks non-volatile optical memory. Enabling memory within the platform itself remains an important challenge.
To address this challenge, the researchers used a complementary metal–oxide–semiconductor (CMOS)-compatible oxygen-ion implantation process to introduce memristive functionality directly into TFLN. Implantation at 500 keV produces spatially separated regions of radiation damage and oxygen doping. Under 450 nm illumination, electrons released by oxygen dopants are trapped at lattice defects, forming spatially separated polarons with different charges. The resulting internal electric field changes the refractive index through lithium niobate’s electro-optic effect, writing the memory state. Illumination at 1,064 nm promotes electron–hole recombination, removing the field and restoring the initial refractive index to erase the state.
Figure: Oxygen-ion implantation of a TFLN MRR.
The devices achieve eight non-volatile memory states, corresponding to 3-bit storage, with switching energies of hundreds of femtojoules per bit and submicrosecond optical response. They retain their states for more than 100 hours and withstand one million write–erase cycles. The team also demonstrated a monolithically integrated 10 × 28 memristor array on a single TFLN chip.
By enabling optical memory without heterogeneous integration, this work extends the functionality of TFLN and provides a promising route towards scalable, programmable photonic circuits and future optical neural networks.
The research was supported by the National Natural Science Foundation of China, the Key Research and Development Program of Shandong Province, and the Open Research Fund of the State Key Laboratory of Nuclear Physics and Technology at Peking University. The team also received support from the shared instrumentation platform of the School of Physics at Shandong University.