We present the most complete and accurate first-principles model of photoreflectance (PR) spectra from strained quantum wells currently in the field. We calculate the PR spectrum, [special characters omitted], by taking the difference between reflectivities as a function of quantum well absorption under electric fields representing the pump beam on and off. We calculate the band-to-band quantum well absorption using Fermi's Golden Rule, convolved with a Lorentzian lineshape. The Lorentzian broadened excitonic absorption includes variationally calculated binding energies. A 6 band Luttinger-Kohn k·p bandstructure calculation, including strain and valence-band mixing, provides dispersion relations, matrix elements, and wavefunctions. We compared the calculated PR spectrum for a 100 Å In0.18 Ga0.82As/GaAs single quantum well with the experimental spectrum with excellent agreement. Our results agreed very well with experimental photoluminescence and photoluminescence excitation spectroscopy data and with another group's empirical fit evaluation of the experimental PR spectrum from a similar sample. They identified one heavy-hole state in a particular energy region where we identified two closely spaced heavy-hole states. Our theoretical PR spectra compared favorably with the experimental spectra for two AlxGa 1-xAs/GaAs single quantum wells in the literature. Three other models exist for calculating the PR spectrum from quantum wells. The first PR model was a semi-empirical first derivative calculation. The second calculated the PR spectrum from an excitonic dielectric constant first derivative. Our model calculates both the excitonic and band-to-band absorption for two fields without assuming a derivative form. Our theoretical PR spectra matched these experimental spectra better than their theoretical models did. The third PR model used the band-to-band absorption derivative to calculate the spectrum for a superlattice, which our model was not designed to handle. The agreement between their experimental and theoretical spectra was poor. We determined sensitivity of the PR spectrum to various parameters. The excitonic contribution dominated, but the band-to-band contribution provided a significant background spectrum. The spectrum was fairly sensitive to quantum well width, composition, and conduction band offset. The spectrum was very sensitive to the high electric field (10 kV/cm). This first principles model provides a deeper understanding of the physics of photoreflectance spectroscopy.

Numerical simulation was used to analyze performance of Si$\sb{1-x}$Ge$\sb{x}$ pseudomorphic heterojunction bipolar transistors (PHBTs). Models for different material properties were developed and used in 2-D drift-diffusion-equations and 1-D hydrodynamic-equations simulation programs. N-p-N and P-n-P PHBTs were analyzed. For N-p-N PHBTs both metal- and poly-Si-emitter contacts were investigated. The figures of merit used are unity current gain frequency f$\sb{T}$ and maximum frequency of oscillation f$\sb{max}$. In all cases analyzed, Si$\sb{1-x}$Ge$\sb{x}$ offers significant advantages over equivalent Si devices: metal emitter N-p-N Si$\sb{1-x}$Ge$\sb{x}$ devices can have f$\sb{T}$ up to two to three times higher than Si; poly-Si emitter N-p-N Si$\sb{1-x}$Ge$\sb{x}$ up to 50% or more higher f$\sb{T}$; P-n-P metal-emitter Si$\sb{1-x}$Ge$\sb{x}$ devices shows similar improvements. f$\sb{max}$ is also improved in all devices, but not as much as f$\sb{T}$. Some devices are actually limited by their f$\sb{max}$ and not f$\sb{T}$ and to improve their performance, f$\sb{max}$ must be increased first by reducing parasitic resistances and capacitances.

A study of InGaAs channel heterostructure field effect transistors (HFETs) on GaAs substrates was undertaken utilizing the low pressure organometallic chemical vapor phase epitaxial (OMVPE) growth technique. Excellent quality HFET material properties were obtained for a split level donor structure, in which the Schottky gate was placed on an undoped AlGaAs layer grown on top of the doped AlGaAs donor layer. A one micron gate length fabrication process was developed to examine the device properties of these materials. A very strong correlation between material characterization results and device performance was observed in all cases. After demonstrating the consistency of the growth and fabrication processes using an $\rm In\sb{0.15}Ga\sb{0.85}As$ channel as a baseline, improvements to the device were undertaken. A delta doping technique was successfully developed and optimized using SIMS and Hall measurements to study the diffusion of the dopant spike. The sheet charge density and device transconductance increased for delta doped material. Increasing the channel indium content reduced the 2DEG mobility, but the expected improvements in transconductance and RF performance were observed. Critical layer thickness (CLT) issues were examined using $\rm In\sb{0.33}Ga\sb{0.67}As$ channel HFETs. Lightly dislocated material still exhibited superior device performance. An asymmetry in dislocation formation was observed, with dislocations forming preferentially in the (011) direction. Devices with a 50 A well width displayed a sharp drop in current in the (0-11) direction. The transconductance and RF properties were not as strongly affected. As the CLT was further exceeded the dislocation network became more symmetric and dense and device performance was severely degraded. A linear channel indium grading methodology was developed to delay the onset of misfit dislocations. Grading from 25-33% produced device properties commensurate with the ungraded 33% indium channel structure, without the asymmetry effects due to dislocation formation. Efforts at developing lattice constant engineered substrates were undertaken. Linear grading to 53% indium at a low growth temperature of 575$\sp\circ$C reduced the amount of three dimensional growth compared to other techniques.