Search

Artificial synapses for neuromorphic computing have been increasingly highlighted, owing to their capacity to emulate brain activity. In particular, solid-state electrolyte-gated electrodes have garnered significant attention because they enable the simultaneous achievement of outstanding synaptic characteristics and mass productivity by adjusting proton migration. However, the inevitable interface traps restrict the protons at the channel–electrolyte interface, resulting in the deterioration of synaptic characteristics. Herein, we propose a solid-state electrolyte-based artificial synaptic device with magnesium oxide (MgO) to achieve outstanding synaptic characteristics in humanlike mechanisms by reducing the interface trap density via dangling bond passivation. In addition, the feasibility of utilizing MgO as a proton reservoir, capable of supplying protons stably and maintaining the proton–electron coupling effect, is demonstrated. With the proton reservoir layer, a significantly greater number of conductance weight states, as well as long-term plasticity over 200 s, is achieved at a low operating power (250 fJ). Furthermore, a pattern recognition simulation is performed based on the synaptic characteristics of the proposed synaptic device, yielding a high pattern recognition accuracy of 94.03%. These results imply the potential for advancing high-performance neuromorphic computing systems.