In this thesis, a variety of topics related to high speed optoelectronic devices and measurement techniques using ultrafast optical pulses are presented.
Following a brief introduction, the second chapter describes a Q-switched semi-conductor laser using a multi-quantum well active layer both for gain and as an intracavity loss modulator. While Q-Switching does not produce as short a pulse as modelocking, it does offer the advantage of adjustability of the repetition rate making it attractive as a source for digital communication links. It is also found to be preferred to the similar approach of gain switching due to less demanding requirements on the rf modulation power level and waveform. Results include a pulse width of ~ 20 ps which is fairly independent of the repetition rate, and a limiting repetition rate of 3.2 GHz. The onset of an irregular pulse train which limits the maximum modulation frequency, is analyzed by a graphical approach.
The potential for optical interconnects has motivated a marriage between the two technologies of Si VLSI and GaAs optoelectronics. Direct integration by the growth of GaAs on Si had been impossible, but the MBE and MOCVD techniques now enable the growth of such layers and of a quality suitable for devices. The third chapter describes the operating characteristics of GaAs-on-Si lasers and photodiodes with particular attention to their high speed performance. Both the lasers and photodiodes show comparable high speed performance to similar structures fabricated on GaAs, with most of the shortcomings being in their dc characteristics.
In the fourth chapter, a novel approach to improving the resolution of photoconductive sampling is presented, called differential sampling. This technique obviates the need for carrier lifetime reduction usually used to improve temporal resolution, and is in principal only limited by a small (few ps) RC circuit time. An analysis of the minimum detectable signal voltage shows the technique does quite well compared with lifetime reduction techniques which also tend to reduce mobility and dark resistance. An experimental demonstration of this technique is presented in chapter five. Using a two gap sampler, accurate measurement (10 ps resolution) of a 60 ps pulse response from a photodiode is achieved using photoconductors with a recovery time of only 150 ps. Performance near the fundamental Johnson noise limit is also attained, though the minimum detectable signal is higher than predicted due to low response of the photoconductors (probably due to poor contacts).
Finally, in chapter six, the possibility of retrieving an impulse response from its autocorrelation is explored. The use of the logarithmic Hilbert transform for phase retrieval has been discounted in the literature since most such work is concerned with imaging problems for which it is not appropriate due to their symmetric nature. However, causality and the decay nature of transient phenomena make this technique very suitable for use with the impulse response of passive devices. Conditions for the validity of this technique for temporal problems are presented. Simulated retrieval of two functions with similar autocorrelations is demonstrated with sufficient clarity to distinguish them, as well as showing good agreement with the original. Practical limitations and aspects -- such as noise, finite time domain, etc. -- are also simulated and discussed.
