Tseung-Yuen Tseng’s research while affiliated with National Yang Ming Chiao Tung University and other places

In recent time, the emergence of optoelectronic memristors has opened up new opportunities for the scientific community to realize their neurological functionalities of optoelectronic systems. Neuromorphic optoelectronic memristors (NOMs) can directly respond to optical pulses with possessing the desirable features of high bandwidth, zero latency, and low crosstalk. They can act as artificial ocular (vision) systems with their capability to integrate sensing, memory, and computing features, and effectively overcome the von Neumann bottleneck. In this review, recent developments in metal oxide semiconductors based NOMs are investigated, with an underscoring on their working principles and realization of neuro-synaptic functions. Attention is given to the synaptic weight modulation in optical–electrical synergistic mode and all optical modes. Their applications in neuromorphic computing systems such as 2D static image and pattern recognition, color recognition, and motion or movement detection are presented. Finally, the forward-looking outlooks are suggested to overcome the pending challenges that hinder the progress of emerging research area of NOMs.

Annealing improves the switching and synaptic performance of ITO/WOx/CuOx/ITO transparent devices. The device has low SET and RESET voltages, stable and robust AC endurance of up to 10⁶ cycles, and can retain the states for more than 10⁴ s. The device demonstrates synaptic capabilities by emulating neural functions under both electrical and light stimuli. The behaviors including long-term potentiation/depression, paired-pulse facilitation, spike-timing-dependent plasticity, and fully tunable relaxation time of short-term memory, mimicking the temporal dynamics of the biological neuron, are declared. A convolutional neural network operation is conducted by exploiting the synaptic functions of the device. The high accuracy of 96.67% with high noise tolerance (close to an ideal synapse) can be achieved. Material analyses are conducted, and switching/synaptic mechanisms are proposed to explain such phenomena.

Developing flexible and transparent memristors for emulating biological activities aligns with the growing demand for sustainable technologies in electronics. This paper presents the development and characterization of transparent memristors (transristors) on a flexible substrate, utilizing a structure of ITO/SnOx/HfOx/ITO/PEN. Hafnium oxide (HfOx) and tin oxide (SnOx) films are sequentially RF sputtered onto an indium doped tin oxide (ITO) bottom electrode, with polyethylene naphthalate serving as the flexible substrate. Then, an ITO top electrode is sputtered onto the SnOx layer using a shadow mask. Samples with varying thicknesses of HfOx and SnOx were prepared to optimize the device configuration. Electrical switching and synaptic characteristics of these samples were measured at room temperature, with a positive voltage applied to the top electrode and a negative voltage to the bottom electrode. This study identifies a configuration with 35 nm SnOx and 6 nm HfOx as the most effective, exhibiting excellent bipolar switching properties. Notably, it demonstrates low set/reset voltages of 1.3 and −1.6 V, with a compliance current of 100 μA. X-ray photoelectron spectroscopy was employed to assess the concentration of oxygen vacancies in the films. The device also shows the highest endurance up to 10⁴ cycles, long-term potentiation/depression characteristics over 350 cycles, a good nonlinearity value of 1.53 (potentiation)/1.46 (depression), and 100% pattern recognition accuracy at just 14 iterations. Multi-state resistive switching characteristics were also explored. Obtained characteristics reveal that the optimized device could serve as a flexible component in making artificial synapses.

Supercapacitors employing transition metal oxide electrodes exhibit larger specific capacities and energy densities. Performance enhancement of the transition metal oxide electrodes can be achieved by incorporation of carbonaceous materials, to form composite electrode. However, incorporation of carbonaceous materials during the synthesis process can alter the morphological properties of the transition metal oxides. Iron cobaltite (FeCo2O4) (FCO) nanosheets exhibit large specific surface area and pore volume, which enhances the loading and diffusion of ions within the electrode. Herein, we designed composite electrodes made up of FCO, reduced graphene oxide (rGO), and functionalized multi-walled carbon nanotubes (f-MWCNTs) while retaining the high specific surface area of the FCO nanosheets. At 3 A g⁻¹, the composite electrode exhibits specific capacity, Cs of 1091 C g⁻¹ as compared with 555 C g⁻¹ of the pristine FCO. Used in an asymmetric supercapacitor, the composite electrode demonstrates maximum energy density of 34 Wh kg⁻¹, maximum power density of 4479 W kg⁻¹, and 92% capacity retention after 5000 cycles. In contrast, the pristine FCO retains only 70% of its capacity after 3000 cycles.

Due to the imitation of the neural functionalities of the human brain via optical modulation of resistance states, photoelectric resistive random access memory (ReRAM) devices attract extensive attraction for synaptic electronics and in‐memory computing applications. In this work, a photoelectric synaptic ReRAM (PSR) of the structure of ITO/Zn2SnO4/Ga2O3/ITO/glass with a simple fabrication process is reported to imitate brain plasticity. Electrically induced long‐term potentiation/depression (LTP/D) behavior indicates the fulfillment of the fundamental requirement of artificial neuron devices. Classification of three‐channeled images corrupted with different levels (0.15–0.9) of Gaussian noise is achieved by simulating a convolutional neural network (CNN). The violet light (405 nm) illumination generates excitatory post synaptic current (EPSC), which is influenced by the persistent photoconductivity (PPC) effect after discontinuing the optical excitation. As an artificial neuron device, PSR is able to imitate some basic neural functions such as multi‐levels of photoelectric memory with linearly increasing trend, and learning‐forgetting‐relearning behavior. The same device also shows the emulation of visual persistency of optic nerve and skin‐damage warning. This device executes high‐pass filtering function and demonstrates its potential in the image‐sharpening process. These findings provide an avenue to develop oxide semiconductor‐based multifunctional synaptic devices for advanced in‐memory photoelectric systems.

Transparent memristor-based neuromorphic synapses are expected to be specialised devices for high-speed information transmission and processing. The synaptic linearity and potentiation/depression cycles are imperative issues for the application of memristors. This work explores a memristor for improving switching uniformity by introducing a thin HfOx interfacial layer as a diffusion-limiting layer sandwiched between WOx and ITO bottom electrodes. An optimized HfOx thickness not only provides the best switching properties but also shows superior synaptic properties. The optimized 15 nm thin WOx layer can retain the memristor's excellence in P/D linearity, a cycling stability of 494 epochs and image recognition up to 3 mm bending, making it suitable for flexible devices. The artificial synapse is capable of reversible short-term and long-term learning behaviors confirmed by spike-timing-dependent-plasticity (STDP) results. X-ray photoelectron spectroscopy confirms the device composition and provides the oxygen vacancy concentration at the WOx/HfOx interface to realize the switching mechanism. The thicknesses of the different layers are estimated from the high-resolution transmission electron microscopy observations. The fabricated device exhibits 92.2% transparency, as confirmed by the UV-Vis spectrum.

Photoelectric synaptic devices as a combination of electronic synapse and photodetector are considered as emerging bio-inspired device technologies. These devices have immense potential to conquer the bottleneck of von Neumann architecture based traditional computing systems. In this Letter, we propose an all-oxide based photoelectric neuro-synaptic resistive random access memory device with the structure of ITO/Ga2O3/ZnO/ITO/Glass, in which the conductance states are reversibly tuned by two different wavelengths (405/522 nm) of visible light spectrum. The strength of light pulse is altered to investigate the learning and forgetting phases of the photoelectric response of the device. A basic biomimetic function “learning-forgetting-rehearsal” behavior is imitated up to 20 cycles. Moreover, emulation of some typical synaptic functions such as associative learning and switching between short and long term plasticities indicate the wavelength awareness of the device. Based on the pure optically induced potentiation/depression characteristics, convolutional neural network simulation achieves an overall test accuracy of 82.5% for the classification of Zalando's article images. The noise tolerance capability of neural network is also examined by applying “salt and pepper” noise in high proportion (75%) to corrupt the images. This work may provide a promising step toward the development of transparent electronics in optogenetics-inspired neuromorphic computing.

Photo sensing capability of an artificial synaptic device make it more valuable for brain-inspired computing systems, which can conquer the von Neumann bottleneck. In this letter, Ba _0.7 Sr _0.3 TiO ₃ (BST) based single layered resistive random access memory (ReRAM) device is proposed with its four different photoelectric applications. First, bidirectional photoelectric response (BPR) is achieved via 405 and 633 nm wavelengths stimulation. Second, this novel BPR contributes to the emulation of light potentiation, and light depression behaviors. Third, an important neurological behavior “experience dependent plasticity (EDP)” is mimicked by frequency modulation of 405 nm light pulse. Fourth, ultra-high (>200%) indices of post-tetanic potentiation and depression (PTP/PTD) are achieved by amplifying the device conductance as a result of two different pulse train stimulations. A convolutional neural network is designed to classify Zalando’s article images with a test accuracy of 87.7%. These results provide an opportunity for the development of multi-wavelength dependent multifunctional artificial brain-inspired systems.