Search

Bio-inspired visual systems require a critical component to perceive optical stimuli and convert them into spiking signals. In this study, a compact photosensitive spiking neuron is demonstrated using a single band-modulation device called zero subthreshold swing and zero impact ionization FET (Z2-FET). This device is fully compatible with the standard CMOS process and features integrate-and-fire (IF) neural behavior. The firing threshold and spiking frequency of the Z2-FET neuron can be modulated by incident light with various intensities and wavelengths, similar to the photoresponse observed in biological neurons. TCAD simulations validate the working principle, originating from the deep depletion effect and photo-induced reduction of energy barriers. Inspired by biological neurons with adjustable spike encoding functionality, the Z2-FET neuron achieves controllable photoresponsive frequency when the input current, top-gate, and back-gate voltages are tuned. With its excellent scaling capability, low energy consumption, and tunable spiking sensitivity, the photosensitive Z2-FET neuron holds great potential in multifunctional neuromorphic applications, such as intelligent visual perception and neural computing.

The $\text{Z}^{{2}}$ -FET is a compact device fabricated using fully depleted silicon-on-insulator (FD-SOI) technology, and it operates with band-modulation mechanism. Due to its sharp-switching and hysteresis characteristics, the $\text{Z}^{{2}}$ -FET has shown promising applications in one-transistor dynamic random access memory (1T-DRAM) and artificial spiking neuron. In this article, we develop a novel compact $\text{Z}^{{2}}$ -FET model using an artificial neural network (ANN) approach. In light of its unique gate-controlled hysteresis behavior, our model innovatively treats the $\text{Z}^{{2}}$ -FET as a hybrid-controlled voltage source, significantly simplifying the modeling process and ensuring good accuracy in circuit simulations. To capture the transient behavior of the $\text{Z}^{{2}}$ -FET, a supplementary ANN (SA) model is established based on the unique gate charge storage effect and feedback mechanism of the $\text{Z}^{{2}}$ -FET. Using this model, the SPICE simulation results agree well with TCAD simulations, demonstrating its accuracy and reliability. Furthermore, we propose a methodology for modeling laboratory-manufactured $\text{Z}^{{2}}$ -FET devices based on measured data, which is crucial for the practical application of the developed model.

A novel implementation for electrostatic discharge (ESD) protection is demonstrated, which involves two structures named MOSFET in parallel with diode (MOS-DIO) and MOSFET in parallel with silicon-controlled rectifier (MOS-SCR). The fabricated structures exhibit enhanced robustness enabled through the double silicon-on-insulator (DSOI) technology. The implementation combines a MOSFET in the top silicon film with a diode or a silicon-controlled rectifier (SCR) located in the middle silicon layer. Transmission line pulse (TLP) experiments show that compared with the traditional single nMOS structure, both DSOI MOS-DIO and MOS-SCR exhibit superior current discharge capability with an increase in failure current of up to 39% under the same footprint. The leakage of the proposed structures is also reduced to below 0.2 pA/ $\mu \text{m}$ thanks to the double buried oxide isolation. In addition, both triggering and holding voltages can be flexibly tuned via the positioning of the two shunt devices. Technology computer-aided design (TCAD) simulations were carried out for MOS-SCR to interpret the physical mechanism and predict the performance under different device parameters, including gate length, stacked modes, and ambient temperature.