the strength of connection between neurons, called synaptic plasticity. [3][4][5] Two-terminal synaptic devices such as memristors are typically composed of a metal-insulator-metal structure that exploits phase changes, for example. Despite their simple structure, two-terminal synaptic devices can perform simultaneous transmission and learning processes. [6][7] On the other hand, three-terminal devices have separate terminals for training and transmission processes and the performance of complex functions. [7][8] Research on synaptic devices has mainly focused on electrical stimulation. Still, synaptic changes due to light stimulation are also significant and the demand for light-sensitive synaptic devices is expected to increase. [9] Practically, optical approaches exhibit higher spatiotemporal resolutions than electrical stimulation. [10] Photonic-based neuromorphic systems can enable multimodal plasticity with increased speed, lower power consumption, and higher efficiency. [11][12][13] Among various photo-synaptic devices, configurations that trap/de-trap photo-induced carriers via interfacial or bulk trap sites have shown good synaptic properties. [14] However, low-dimensional materials are difficult to mass-produce. In the case of transistors, pinch-off characteristics do not appear, making it difficult to ensure device stability during operation. In our previous study, optically sensitive oxide materials were used as semiconductors.The deep traps at the interfaces between the channel layer and dielectric play an essential role in capturing carriers, resulting in synaptic behaviors. [15][16][17] Atomic layer deposition (ALD) can be used to grow transistor channels and dielectric materials. ALD enables mass production of photo-synaptic devices, complementary metal-oxide-semiconductor (CMOS) compatibility, durability, and stability. In synaptic devices, it is essential that a large number of states can be implemented. Recently, a synaptic device with 265 analog states with excellent linearity during the synaptic potentiation process induced by ultra-violet (UV) light pulses was demonstrated. [9] Complex sensory synaptic devices using optical and electrical stimulation in parallel are also being actively investigated. [14,18] Moreover, for synaptic devices to have an accurate prediction ability, an excellent linear relationship must exist between the number of stimuli and the current during the learning process. This can be achieved by a high on/ off current ratio during the learning process.