|Abstract||In this thesis the epitaxy of semipolar GaN on patterned sapphire substrates was systematically investigated. Adjusting the growth conditions, the crystal and surface quality of different crystal orientations were optimized. Additionally, a subsequent detailed investigation of the doping behavior depending on the growth direction should help to answer the question, whether semipolar (In)GaN could help to overcome the low quantum efficiencies of the LEDs emitting in the green region.
A new process from scratch to pattern the sapphire substrates was developed. By replacing nickel by photoresist acting as an etching mask, the structuring process was dramatically simplified and accelerated. Due to the nearly infinite possibilities to change the shape of the photoresist stripes via lithography and the systematical
improvement of the dry etching conditions, the sidewalls of the sapphire trenches were optimized to be more c-plane-like. Compared to the previous method, (11-22)-GaN forms a much smoother surface. The surface roughness (RMS) could be reduced from 100 nm to 10 nm in an area of 70 um x 70 um, measured by AFM. Based on (11-22)-GaN, the epitaxy of (10-11)- and (20-21)-GaN was established. In order to avoid parasitic growth on the non-c-plane sapphire facets, a much higher reactor temperature was necessary to overcome the apparently lower selectivity during the nucleation compared to (11-22)-GaN. However, an increased reactor temperature pushes the growth perpendicular to the c-direction - the GaN touches the sapphire stripes from the top and new defects penetrating to the surface were generated. The growth conditions had to be adjusted to keep the main growth rate in c-direction and to improve the selectivity at the beginning at once in order to improve the crystal quality significantly. A certain amount of defects close to the interface between sapphire and GaN could be stopped by an in situ deposited SiN interlayer. Its deposition time and the thickness of the GaN buffer layer were optimized systematically. Whereas (11-22)- and (10-11)-GaN developed a coalesced, planar semipolar surface, (20-21)-GaN formed a very rough surface dominated by the more stable (0001)-, (10-11)- and (10-10)-facets. A suitable coalescence of the GaN stripes via MOVPE was not possible. Several publications from other research groups confirm the completly different growth behavior of (20-21)-GaN in comparison to the other directions. A very fast coalescence of the MOVPE-GaN stripes was achieved only by HVPE overgrowth. A strong correlation between the crystal quality of MOVPE- and
HVPE-GaN was observed. An improved crystal quality of the (20-21)-MOVPE-GaN led to smoother surfaces of HVPE-GaN. However, due to the limited availability of the HVPE-reactor during this thesis, a systematic investigation of the deposition of InGaN/GaN quantum wells on (20-21)-GaN was not possible. Furthermore the doping behavior of oxygen, silicon and magnesium in c-, (11-22)- and (10-11)-GaN depending on the growth conditions was investigated. SIMS measurements confirm a much higher oxygen content in semipolar than in c-GaN. Using standard growth conditions, (11-22)- and (10-11)-GaN showed an oxygen concentration in the range of 2 x 10^18 cm³ whereas in c-GaN grown at the same conditions no oxygen above the detection limit of 10^17 cm³ was measured. Fortunately, the
incorporation efficiency of oxygen in (11-22)-GaN strongly depends on the growth temperature. Lowering it from 1050 °C to 900 °C, the parasitic doping level could be reduced to 10^17 cm³. However, the oxygen incorporation efficiency in c- and (10-11)-GaN seems to be almost temperature independent. Mg doping of semipolar and polar GaN seems to be completly different as well. Whereas SIMS measurements show a Mg concentration of 10^19 cm³ in c-plane GaN, (10-11)-GaN contains just the half of doping atoms, using the same Mg flow. In contrast, the Mg incorporation
efficiency on (11-22)-oriented surfaces strongly depends on the growth temperature as well. At 1050 °C, a concentration of just 10^18 cm³ was achieved. Reducing the reactor temperature to 900 °C, the concentration increased to 5 x 10^18 cm³. Due to the temperature dependence of the oxygen and magnesium incorporation, a suitable p-conductivity at low growth temperatures was achieved. Unfortunately, the constantly high parasitic doping level of (10-11)-GaN resulted in a n-conductivity even at high Mg concentrations. Next, InGaN/(In)GaN quantum well structures showing an emission wavelength of about 500 nm were deposited on the coalesced planar GaN surfaces ((0001), (11-22) and (10-11)). On c-plane GaN, due to the quantum confined Stark effect, a reduced In content was necessary to achieve the same emission wavelength. The internal quantum efficiencies were determined by temperature dependent photoluminescence measurements. Whereas (0001)- and (11-22)-InGaN/GaN quantum wells showed an IQE between 20 % and 22 %, an IQE of almost 40 % on (10-11)-GaN was achieved. TEM investigations indicate strong phase separations in the InGaN quantum wells. Phases with different indium contents are developed and form, independent on crystal orientation, locally an inhomogeneous indium distribution. The idea of adding
indium into the barriers to reduce the strain and therefore the defect density in the quantum well structures was systematically investigated. Though the absolute intensities in PL measurements could be increased significantly by using InGaN/InGaN structures, temperature dependent PL measurements showed an IQE of just 16 %. Space-resolved cathodoluminescence shows an inhomogeneous indium distribution on a scale of several 100 nm. It seems, that only the light outcoupling efficiency could be improved, whereas the internal quantum efficiency was reduced. Investigations on full LED structures on (11-22)-MOVPE-GaN confirmed a reduced efficiency of the InGaN/InGaN-structures compared to the InGaN/GaN quantum well structures. Whereas InGaN/GaN-LEDs showed an optical power of 140 uW applying an electrical current of 20 mA, just 33 uW were achieved by using InGaN/InGaN structures. Due to the inhomogeneous indium distribution, the electrical current could flow between the islands containing more indium and therefore does not contribute completely to the radiative recombination. Similar results were achieved on (11-22)-HVPE-GaN. Temperature dependent PL measurements showed an increased IQE of 27 %. However, electroluminescence measurements resulted in an optical power of just 16 uW. Via EBIC measurements no suitable space-charge region and therefore no functionable p-n-junction could be measured. Currently, detailed studies about
the crystal quality of semipolar HVPE-GaN and the origin of the significantly reduced efficiency in EL measurements are done by Marian Caliebe in this institute. Unfortunately, due to the high parasitic oxygen doping level in (10-11)-GaN, it was not possible to achieve a suitable p-conductivity and therefore a working LED structure. Other research groups could theoretically and experimentally show the high potential of semi- and nonpolar GaN to reduce the efficiency droop of c-oriented LEDs emitting in the green region especially at high current densities. The techniques presented in this thesis to produce large scale, in principle arbitrarily oriented GaN on patterned sapphire substrates could help to overcome the problem of limited sample sizes of bulk GaN. To improve the efficiency of the LEDs grown on these planar surfaces to catch up to LEDs bases on bulk GaN, mainly three points have to be
investigated more detailed and optimized:
- Crystal quality near the nucleation: The reactor conditions at the beginning of the growth have to be optimized in order to achieve a lower defect density while keeping the selectivity of GaN constant.
- Defects penetrating to the surface: An in situ deposited SiN interlayer stops defects only to a limited extent. Normally, the +c-wing of the GaN stripes shows a higher defect density compared to the -c-wing. By overgrowth the +c-wing by the nearby -c-wing, the total defect density close to the surface could be reduced significantly.
- Parasitic background doping: Semipolar GaN incorporates much more parasitic oxygen compared to standard c-plane GaN. Therefore, the purity of the material sources like TMGa, TEGa etc. has to be improved to be able to
achieve a suitable p-conductivity especially on (10-11)-GaN.||dc.description.abstract