25th Anniversary Plenary Lecture II

Chair: Hideo Hosono, Tokyo Institute of Technology

Monday 11:45–12:30

Silicon vs. the Rest

John Robertson

Engineering Dept, Cambridge University, Cambridge CB2 1PZ, UK

There have been many periods in the development of amorphous semiconductors. The first period started with the chalogenides such as STAG glasses and a-Se, due to the importance of Xerox machines and fundamentals. Many important concepts such as tail states, metal-insulator transitions and localisation were explored, plus the very important idea of negative U.

The amorphous group IV semiconductors became important in the 1970s, structurally, but electronically they suffered from a high density of gap states. Their impact blossomed once the defects were passivated by hydrogen and the facile silane PECVD growth process arrived. This led to doping, photovoltaics and TFTs and multi-billion dollar businesses of AMLCDs and photovoltaics. But in the 1980s and early 1990s there was a focus on hydrogen-based instability mechanisms, perhaps at the expense of problem solving.

Recently, it has become clear that the classical materials a-Si:H (and a-Se) could be surpassed for specific applications by others. For TFTs, amorphous oxide semiconductors have much higher carrier mobilities than a-Si, reversing the standard idea that amorphous means low mobility, and making them a preferred candidate for future TFTs. AOSs suffer from some instabilities; however these are not yet very well understood and the literature suffers from a somewhat chaotic, unscientific approach.

At the same time, in the chalcogenide field, GeSbTe alloys have been extremely important as phase change materials for information storage (optical and electrical), based on the similar energies of their amorphous and crystalline phases, different bonding, and rapid transitions between the two phases. The different bonding arises from the destruction of resonant bonding in p states by disorder. The rapid transition arises from "fragile" structural liquids and a breakdown of the Einstein relation between viscosity and diffusion, where diffusion involves not bond breaking, but a shuffling motion.

Finally, HfO2 has surpassed SiO2 as the dielectric of choice in MOSFETs due to its high dielectric constant. Despite all its innate disadvantages in terms of oxygen defects, HfO2 has been successfully implemented in VLSI MOSFETs since 2007.

Keywords: silicon, amorphous oxide semiconductors, TFTs, phase change memories, HfO2, gate insulators