Table of Contents:
Modern semiconductor technology is almost entirely based on Silicon. Its outstanding electronic properties and the process technology which has been refined for more than four decades have made this semiconductor an irreplaceable component of modern technology.
The indirect nature of its band gap however renders it rather unsuitable for the efficient emission of light. Optical gain is deemed unfeasible in this material. The monolithic integration of electronics and photonics on a single chip therefore needs to be accomplished by combining silicon with another material. Starting from these premises, two material systems that can be grown epitaxially on a silicon substrate are experimentally investigated with respect to their optical properties.
Quantum wells (qw) of Germanium were experimentally investigated by spectrally resolved white-light pump-probe-absorption spectroscopy at room temperature. The experiment was driven by an ultrafast Ti:Sapphire amplifier, thus achieving a temporal resolution better than 100fs while achieving very high excitation densities.
The technique yields absorption spectra after optical excitation as function of energy and time, thereby enabling the analysis of carrier scattering dynamics.
After exciting the sample via direct optical transitions where carriers are generated in gamma-like states, several scattering mechanisms start to redistribute the carriers.
Coulomb scattering broadens the carrier distribution while phonon scattering cools the carriers against the lattice. The non-polar bonding in Ge makes deformation potential scattering the dominant phonon scattering mechanism. At 300K, cooling takes place on a timescale of several 100fs in Ge. Holes in the valence band relax towards the gamma-like first heavy-hole state while electrons relax towards the L-point of the reciprocal space which is the global energy minimum of the conduction band in Ge. Carriers reside in these long-lived states for more than 10ns and recombine mostly non-radiatively. Phonon-assisted absorption further excites electrons from the L-valley into higher states, thereby creating an additional absorption channel after the excitation.
Shortly after a high-power excitation at a suitable energy, a sufficiently large electron population is formed in gamma-like states, thereby creating optical gain on a short timescale. The gain attains values of up to 8e-4/QW (comparable to direct-gap semiconductors) and exists for a few 10s of fs. This property may be exploited in order to realise an optical amplifier with this material system. Further tailoring of the material may allow for isolating the gamma-like properties of the band structure.
A second material class suitable for photonics on silicon is Ga(NAsP), which was grown as quantum wells on a silicon substrate matching the lattice constant of the substrate. The basic optical properties were determined using the variable stripe-length method. Here, the modal gain is measured by exciting a stripe-shaped region on the sample. Varying the stripe length leads to an exponential increase of the sample emission which is collected out of the facet at the fixed end of the stripe. A gain value of 80/cm was determined at room temperature. At 10K, laser activity was verified. Using the method of Hakki and Paoli, gain values in the range of 5/cm over a spectral range of 10meV were found under lasing conditions. The emission of the sample showed a clear mode structure up to 125K, verifying its laser activity.
In order to relate the results to those of established materials, a selection of comparable III/V semiconductors were measured in the same setups. The pump-probe measurements on (GaIn)As quantum wells exhibited a much more rapid scattering, which is attributed to the more efficient Fröhlich interaction of the polar semiconductor. In these material systems, quite similar optical gain values of 1e-3/QW were found with decay times of several 100ps. For (GaIn)(NAs), slightly higher values were determined. Using the variable stripe-length method, GaSb quantum wells with dot-like morphology were investigated. Apart from the rather high gain values of about 280/cm, these structures also exhibited a very broad spectral gain range, both of which renders them suitable for emitters in the infrared.
The variable stripe-length method itself was subjected to a thorough analysis with respect to its reliability. The analysis consisted of developing new numerical methods for evaluating the data as well as testing these for their stability against noise. Carrier depletion effects were investigated with a theoretical model. Combining the results, a precise set of criteria for the validity of evaluated gain spectra was established.