Interface engineering in optoelectronic hafnium oxide-based heterojunction synaptic device for neuromorphic computing.

With the exponential growth of information data, there is an urgent demand for emerging computing architectures that integrate memory and computation, represented by neuromorphic computing. Against this backdrop, this study proposes a memristor based on the Sr:HfO2 and TiO2 heterojunction. The doping of Sr can increase the switching ratio of the device, while the introduction of a TiO2 interlayer enhances its photoresponse. Building on this, the device brilliantly simulates a series of biological synaptic plasticity mechanisms, such as short-term plasticity, long-term plasticity, paired-pulse depression, and spike-timing-dependent plasticity. Additionally, the device exhibits excellent optoelectronic response characteristics. Under the combined action of optical pulses and electrical signals, it simulates plasticity dependent on the width, power, and number of optical pulses. By integrating a convolutional neural network for image recognition, the accuracy rates reached 97 % on the MNIST and 82.5 % on the Fashion-MNIST datasets. Therefore, the device demonstrates outstanding optoelectronic synaptic properties, low power consumption, and non-volatility. This work provides a potential pathway for the actualization of next-generation highly efficient brain-inspired chips.