Characterization And Modeling Of New High-frequency Field Effect Devices

As the further requirements of modern electronic systems, such as higher performance, low cost, minimization, and so on, new high-frequency field effect devices based on wide band-gap semiconductor materials silicon carbide (SiC) and gallium nitride (GaN), and zero band-gap semiconductor material Graphene have been widely attracted. Modeling of semiconductor devices is an indispensable technique for device characterization and circuits design. However, because of the the new devices'higher performance, different device structure and operation principles compared with devices based on silicon (Si) and gallium arsenide (GaAs), it is difficuit to characterize the new devices correctly by using the available models. As a result, characterization and modeling of new high-frequency devices is in urgent need to pave the way in the applications of moden integrated circuits. In this paper, some key problems in devices based on SiC, GaN and Graphene have been focused.1. Large-signal nonlinear modeling of 4H-SiC metal-semiconductor field effect transistors (MESFET). Due to the effects of self-heating and traps in 4H-SiC MESFET, which results in absent of accurate and useful larger-signal model, the bottom-up technique from small-signal model to large-singal model has been used. A nonlinear equivalent circuit model, including self-heating and trap effects, has been developed by adding modifications in traditional direct current (DC) current-voltage (I-V) model and nonlinear capacitance model. The design-kit based on native process line in commercial microwave simulation software Aglient advanced design system (ADS) has been fully developed. During these work, a wide-band (0-20 GHz) and accurate small-signal equivalent circuit model has been established based on an improved small-signal topology and a novel parameter-extraction method. More over, a black-box nonlinear modeling method based on support vector machine (SVR) has been proposed.Compared with methods by using artificial neural network (ANN), the proposed technique can not only overcome problems like over-fitting,requirement of large samples, time consuming, and so on, but also have good physical meaning of transistor. 2. Characterization and modeling of high-frequency noise in GaN high electron mobility transistors (HEMT). Due to the absent of accurate high-frequency noise model for GaN HEMT, which has great potential in the low-noise applications, the high-frequency noise modeling method, noise mechanism, and the device structure dependent noise performance have been studied by noise parameters model, the minimum noise figure (Fmin) model, and numerical physical based model, respectively. And a noise parameters model based on support vector regression (SVR), accurate Fmin model, and the effect of Al mole fraction on noise performance have been established. More over, the high frequency noise performance of AlGaN/GaN/AlGaN double heterojunction HEMTs is analyzed based on the unique property of strong polarization inâ…¢-N compounds,and the results show that better high-frequency noise performance can be achieved. During these works, a source parasitic impedance model is proposed based on Boltzmann transport equation, and the analysis of its effect on high frequency noise performance shows that the dispersion of Fmin is actually nonlinear due to the dispersion of parasitic impedance.3. Characterization of Graphene resonant channel transistor(RCT). Due to the large effect of parasitic parameters in globe gate FET structure Graphene mechanical resonator, which makes it difficult to readout and analysis of mechanical signal, a novel direct RF measurement method based on local back-gate Graphene RCT is proposed by combining the operation principles of Graphene FET and mechanical resonator. And the fabrication technique of local back-gate RCT and the theory of direct RF readout method are studied. Compared with the mixing scheme used in globe back-gate FET structure Graphene mechanical resonator, the readout time by using the proposed method is two orders of magnitude less and both of the amplitude and the phase of the RCT can be obtained. More over, a new chemical-free fabrication method for Graphene RCT is proposed, which is promising for the study of intrinsic property of Graphene nanoelectromechanical systems (NEMS).