A New Strategy for Optimizing Threshold Voltage of Artificial Neurons

Abstract

Over the years, memristors have received much attention as strong candidates in the field of neuromorphology and have presented a wide range of application scenarios. In order to ensure the effective imitation of neurons by memristors during electroforming, there is usually some variability in the selection of electrode materials at both ends, but this results in the threshold voltage of the device being asymmetric. However, electrical signals within the range of neuron threshold voltage can be used to carry and transmit information, the asymmetry of threshold voltage often leads to the error and leakage of stored information. In this work, we present a system consisting of an Au/TaOx/Ta memristor (MTSM), a diode, and two load resistors. We have developed artificial neurons on account of Au/TaOx/Ta that demonstrate outstanding power-on and power-off efficiency, with a conversion ratio above 10 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">5</sup> . Furthermore, these artificial neurons exhibit sustained stability over multiple cycle tests, indicating their potential for practical use. Furthermore, we found that the size of the load resistance can affect the switching of individual devices and thus control the threshold voltage range of artificial neurons. Therefore, this paper proposes a new method to optimize the threshold voltage of artificial neurons based on load resistance. This optimized strategy can be used to adjust the setting and resetting of the threshold voltage of neurons thus providing a potential platform to facilitate the development of electronic devices.