Neuromorphic Boolean Logic Implementation Using Ovonic Threshold Switch Devices

Boolean logic operations have played an indispensable role in digital information processing in the conventional computers as well as in the biological neural systems ^[1]. Recently, it was reported that the dendrites of pyramidal neuron in L2/3 layers in human cerebral cortex perform various Boolean logic functions including the linearly non-separable XOR as well as the linearly separable AND and OR functions ^[2]. In this work, inspired by this finding, we have investigated neuromorphic devices based on the Ovonic Threshold Switch (OTS) that can mimic those behaviors of the dendrites. By designing circuitries composed of OTS and a few passive electrical elements, we have successfully demonstrated the Boolean logic operations such as AND, OR, NOT, NOR, and XOR. To secure the biological plausibility, those logical operations are performed within the rate-coding framework — a neural processing model where information is conveyed through the rate of neuron firing. We have also demonstrated the behavior of the pyramidal cell in the cerebral cortex in the human brain can be reproduced well by using a circuit composed of the above neuromorphic Boolean operators.
To test the bio-inspired information processing, we have performed the edge-detection task of images using the XOR device. Using a single OTS-based electronic neuron, we constructed a three-terminal XOR operator. This circuit, consisting of two inputs and one output, was designed to perform logical operations by leveraging the voltage difference across both ends of a bidirectionally operating OTS neuron. Additionally, to enhance the operational reliability of the XOR operator, we conducted research on the electrode materials for the OTS neuron. Considering the impact of metal work function on the threshold voltage of the OTS neuron, we analyzed the operational characteristics using platinum and titanium nitride as electrode materials, thereby achieving high operational reliability. As a result, we have found that our XOR device can perform the edge-detection task very nicely, leading to the perfect accuracy. These findings demonstrate the versatility of OTS-based spiking Boolean logic circuits and highlight the potential of these devices not only to perform fundamental logical operations but also to address complex, biologically inspired tasks.

References
[1] G. E. Hinton, Nature Neuroscience, 2000, 3, 1170.
[2] A. Gidon et al., Science, 2020, 367, 83-87.