Device design and transport issues in nitride and ferroelectric heterostructure devices.

Abstract

The silicon-silicon dioxide heterostructure has sustained microelectronics for over 30 years. However, as the device channel length reaches 50 nm, this heterostructure will run into serious problems such as gate tunneling leakage, doping fluctuation and increasing junction resistance. This opens opportunities for other material systems. in this dissertation we explore electronic properties and device potential of polar semiconductors. They are nitride compounds (e.g. GaN, AlGaN) and ferroelectric materials (e.g. BST). These materials have the property that there is a net displacement of the cation and anion sublattice leading to a polarization in the material. With proper device design its polar charge can be exploited and under right conditions be used to replace dopants. Additionally one can use heterostructure physics to produce high performance electronic and optoelectronic devices. Studies on charge control and mobility in GaN/AlGaN HEMTs grown by MBE and MOCVD are presented. The sheet charge density in the device channel is found to be controlled by spontaneous polarization, piezoelectric effect and barrier thickness, and sheet charge as high as 10<super>13</super> cm<super> -1</super> can be introduced without doping. Through mobility study we see that interface disorder of the order of one monolayer or more reduces the mobility in the channel considerably. We also see that for low lying electronic states localization occurs in samples with high Al content in the barrier. This is because of the high carrier mass and the very high interface fields that arise. Two-dimensional transport in AlGaN/GaN is examined against AlGaAs/GaAs by using an ensemble Monte Carlo approach. We find that for small bias conditions, GaAs based devices show a shorter transit time. However, when the field increases the GaN channel has a shorter transit time. This difference can be traced to the velocity field relations in the two materials. Our results show that for low power applications GaAs based devices have superior high frequency performance while at large bias values AlGaN/GaN devices have superior performance which has been demonstrated in the experiments. For the BST MOSFET, results on a one-dimensional charge control model and estimation of gate tunneling probability are presented. The modulation charge, tunneling probability and the band profile are examined as a function of spontaneous charges, and applied voltage. In comparison to the conventional MOS the new structure has almost five times more induced-charge density and the tunneling probability is greatly suppressed.