| Popis: |
This thesis is an investigation into both the fundamental and experimental aspects of using semiconductor lasers to generate extremely short (100's of fs) and very high repetition frequency (> 50 GHz) optical pulses. The pulses are produced through modelocking, a technique of forcing a laser to operate in a number of optical modes simultaneously and to hold a constant phase relationship between these modes. Both the shortest and highest repetition rate pulses have been obtained from passive modelocking. An inherently nonlinear technique which does not use any active external timing source. Two structures, ridge-waveguide stripe lasers and liquid phase epitaxy (LPE) regrown lasers, were used to directly generate picosecond width pulses. Using cross-correlation techniques, pulse shape and phase measurements are made. Linear dispersion compensation is shown to achieve nearly a factor of 20 in pulse compression. Stable pulses down to 260 fs are generated. Showing that exitonic effects are not essential in these devices, wavelength tunability was combined with dispersion compensation to create the first broadly tunable subpicosecond semiconductor source. The device is found to give tunability ranges and mode-locked spectral widths that are comparable to the best results achieved in dye lasers in terms of fractions of the operating gain spectral width. Results for different regimes in the tuning range are examined, and pulses directly from the laser are found to have about a 2 to 1 fall-time to rise-time ratio. A significant nonlinear chirp is found only when the laser is tuned to the short wavelength side of its tuning range and was determined to cause long tails in the autocorrelations of compressed pulses. Additionally, spread-resistant pulses are described and experimentally analyzed. The case of high-repetition-rate modelocking, which more likely involves about 5 modes instead of 5000 modes, is examined. Approximations in the leading theory of passive modelocking are shown to be inadequate in this case. A steady-state model for high-repetition-rate modelocking is developed including phase effects and is tailored to parameters of semiconductor lasers. Self-consistent solutions show that a lower threshold gain can exist for a supermode than for single mode operation. Predictions of the laser's behavior upon modifying key material, geometric, and bias parameters are made. Experimental results show that through adjustment of the gain current, "chirp-controlled" modelocking is obtained with operation in any of the three chirp regimes (up-chirped, chirp-free, or down-chirped). This pulse chirp and resulting broadening are due to the algebraic addition of opposite-signed chirps from saturation of the absorber and gain sections. Theoretical modelling from the supermode analysis also traverses the same chirp regimes when the photon intensity is increased. |
