Film deposition has been studied using an ultrahigh vacuum reactor that features a supersonic molecular beam. This approach allows for the precise control of precursor kinetic energy and molecular flux. In addition, a beam angle of incidence can be defined due to the high level of collimation. In a study of homoepitaxial deposition of silicon on Si (100), the beam angle of incidence was found to have a dramatic effects on film morphology. An increase in roughness and anisotropic feature formation were observed when beam angles of incidence exceeded 60° from normal. This agrees with recent simulation results as well as various experimental studies. The mechanism used to describe this reactivity-microstructure coupling is shadowing. At fixed angles of incidence, film feature size parallel to the incoming beam is a strong function of substrate temperature. Low energy electron diffraction indicates the deposited film near the surface is well ordered, even at the lowest deposition rates.; Polycrystalline silicon was also deposited using this technique. Polycrystalline material, preferring the columnar arrangement, was observed at substrate temperatures as low as 520°C. Film nucleation was shown to be a strong function of substrate temperature and beam kinetic energy. Furthermore, the nuclei density at fixed temperature suggests that beam kinetic energy can be used to tune grain size. Thin film transistors were fabricated using examples of the deposited films. Both threshold voltage and carrier mobility in these devices were shown to be comparable to material deposited by LPCVD. Building on these results, an atomic hydrogen source was used to Promote selective epitaxial growth. The addition of a continuous hydrogen flux considerably increases polycrystalline film incubation times. Selective epitaxial layers approaching 1000 Å in thickness were demonstrated. In a final study, reactor scale-up strategies were investigated. Polycrystalline silicon films were uniformly deposited on 3′′ wafers using a supersonic free-jet array. This source geometry was chosen based on a Monte Carlo simulation, whose accuracy was tested by both beam flux and film growth rate measurements. Experimentally, the flux distribution from the source compared well to model predictions.
展开▼
