4K-memristor analog-grade passive crossbar circuit

The superior density of passive analog-grade memristive crossbar circuits enables storing large neural network models directly on specialized neuromorphic chips to avoid costly off-chip communication. To ensure efficient use of such circuits in neuromorphic systems, memristor variations must be substantially lower than those of active memory devices. Here we report a 64 x 64 passive crossbar circuit with -99% functional nonvolatile metal-oxide memristors. The fabrication technology is based on a foundry-compatible process with etchdown patterning and a low-temperature budget. The achieved <26% coefficient of variance in memristor switching voltages is sufficient for programming a 4K-pixel gray-scale pattern with a <4% relative tuning error on average. Analog properties are also successfully verified via experimental demonstration of a 64 x 10 vector-by-matrix multiplication with an average 1% relative conductance import accuracy to model the MNIST image classification by ex-situ trained single-layer perceptron, and modeling of a large-scale multilayer perceptron classifier based on more advanced conductance tuning algorithm.

Automated summary

The superior density of passive analog-grade memristive crossbar circuits allows for storing large neural network models on specialized neuromorphic chips to avoid costly off-chip communication. A 64x64 passive crossbar circuit with high functional nonvolatile metal-oxide memristors was successfully fabricated with low variation in memristor switching voltages. Experimental verification of analog properties demonstrated accurate modeling for MNIST image classification and large-scale multilayer perceptron classifier based on advanced conductance tuning algorithm.