Congratulations to our colleague MSc. Nguyen Bao Ngoc and other student, for their recent publication entitled "Interface-engineered Cu−Cr−O:Mg/CrOₓ heterostructures for artificial synapses and neuromorphic – deep learning systems" in the journal "Journal of Alloys and Compounds", which was a collaboration with our colleagues in the Vietnamese - German Universesity.
Mg-doped delafossite (Cu−Cr−O:Mg) is primarily recognized as a p-type transparent conductive oxide due to its hole-dominated transport, while amorphous metal oxides are widely utilized in memristive devices to achieve analog resistive switching via oxygen vacancy migration. However, single-layer amorphous chromium oxide (CrOx) typically lacks intrinsic memristive effects, exhibiting instead a non-pinched I–V hysteresis loop characteristic of non-linear capacitive polarization or ferroelectric-related charging–discharging behavior. In this study, we propose a multilayer heterostructure that integrates p-type Cu−Cr−O:Mg with amorphous CrOₓ to overcome these limitations and engineer stable resistive switching. The optimal Cu−Cr−O:Mg/CrOx/Cu−Cr−O:Mg stacked trilayer device exhibits stable endurance over 1000 cycles, multi-level retention exceeding 104 seconds, and low drift at low voltages. Furthermore, it successfully emulates key synaptic functions, including long-term potentiation/ depression and the transition from short-term to long-term memory. When implementing memristor-quantized conductance in a ResNet-20 neural network architecture, the 50 ms pulse-width condition model achieves up to 89% classification accuracy on the CIFAR-10 dataset. These findings suggest that combining delafossite semiconductors with amorphous oxides is a promising strategy for developing high-performance artificial synapses for next-generation neuromorphic computing.
For more details, please visit:
https://doi.org/10.1016/j.jallcom.2026.189600
Post a Comment