Power electronics plays an important role in many energy conversion applications in modern society including consumer electronics, data centers, electric vehicles, and power grids, etc. The key components of power electronic circuits are power semiconductor devices including diodes and transistors, which determine the performance of power electronics circuits. Traditional power devices are based on the semiconductor silicon (Si), which have already reached the silicon's material limit. Gallium nitride (GaN) is a wide bandgap semiconductor with high electron mobility and high critical electric field. GaN-based power devices promise superior device performance over the Si-based counterpart. The primary design target of a unipolar power device is to achieve low on-resistance and high breakdown voltage. Although GaN high electron mobility transistor (HEMT) is commercially available in a voltage class from 15 V to 900 V, the performance of GaN devices is still far below the GaN material limit, due to several reasons: 1) To achieve the normally-off operation in a GaN HEMT, the density of two-dimensional electron gas (2DEG) channel cannot be too high; this limits the on-resistance reduction in the access region. 2) The gate capacitance of GaN HEMT is usually low so that the carrier concentration in the channel underneath the gate is relatively low, limiting the on-resistance reduction in the gated channel region. 3) The electric-field distribution in the drift region is not uniform, resulting in a limited breakdown voltage. We proposed to use the junction-based structure in GaN power devices to address the above problems and fully exploit GaN's material properties. The first part of this dissertation characterizes nickel oxide (NiO) as a p-type material to construct the junction-based GaN power devices. Although the homogenous p-GaN/n-GaN junction is preferred in many devices, the selective-area, p-GaN regrowth can lead to excessive leakage current; in comparison, the p-NiO/n-GaN junction is stable without leakage. This section describes the optimization of NiO deposition as well as the NiO characterization. Although acceptor in NiO is not generated by impurity doping, the acceptor concentration modulation is realized by tuning the O2 partial pressure during the sputtering process. Practical breakdown electric field is also characterized and confirmed to be higher than GaN. These results provide the design guidelines for NiO-GaN junction-based power devices. The second part of this dissertation demonstrates the 3D NiO-GaN junction gate to improve the GaN HEMT's on-resistance. The 3D junction gate structure enables a high carrier concentration under the gate region in the device on-state. Meanwhile, the strong depletion effect of the junction-based gate allows for a robust normally-off operation; as a result, the GaN wafer with a higher 2DEG concentration can be used to achieve both normally-off and low on-state resistance in HEMT devices. Simulation is also performed to project the performance space of trigate GaN junction HEMTs using the p-GaN instead of NiO. The third part of this dissertation presents the application of the p-GaN/n-GaN junction in the drift region of the multi-channel lateral devices to achieve the high breakdown voltage. Here p-GaN is grown in-situ with the multi-channel AlGaN/GaN structure, and there is no leakage problem. The structure is designed to achieve charge balance between the acceptor in p-GaN and the net donor in the multichannel AlGaN/GaN. This design enables a uniform electric field distribution and breakdown voltage over 10 kV. The fourth part of this dissertation presents the application of the p-NiO/n-GaN junction in vertical superjunction (SJ) devices. We show the design and simulation of this heterojunction structure in a SJ and confirm the uniform electric field and high breakdown voltage under the charge balance. Then the device fabrication is presented in detail, which mainly comprises the deep GaN trench etch, NiO self-aligned lift off, and photoresist trench planarization. The optimized device shows a trade-off between its drift region specific on-resistance versus breakdown that exceeds the 1D GaN's limit. The last part of this dissertation is exploring the design and fabrication of p-GaN/n-GaN based SJ devices. First, the challenges in p-GaN regrowth especially the introduction of interface impurities are discussed, followed by device simulation and modeling to optimize the SJ performance considering these interface impurities. The activation of regrown p-GaN in deep trenches is more difficult than planar p-GaN, and we present the characterization and physical model for the activation of the deep buried p-GaN. Last, the results of p-GaN filling regrowth and the acceptor concentration calibration in the lightly doped p-GaN are presented and discussed. In summary, our work combines experimental device fabrication and characterization, TCAD simulation, and device modeling to demonstrate the benefit of multi-dimensional, junction-based GaN power devices as compared to the traditional GaN power devices. The junction-based structure at gate region can provides stable normally-off operation and low on-resistance. When being applied to the drift region, the multidimensional junction structure can push the device specific on-resistance versus breakdown voltage trade-off near or even exceeding the material limit. These results will advance the performance and application spaces of GaN power devices. / Doctor of Philosophy / Power electronics plays an important role in many energy conversion applications in modern society including consumer electronics, data centers, electric vehicles, and power grids, etc. The key components of power electronic circuits are power semiconductor devices including diodes and transistors, which determine the performance of power electronics circuits. Traditional power devices are based on the semiconductor silicon (Si), which have already reached the silicon's material limit. Gallium nitride (GaN) is a wide bandgap semiconductor with high electron mobility and high critical electric field. GaN-based power devices promise superior device performance over the Si-based counterpart. Currently, GaN power devices performance is still far below its material limit due to several reasons: 1) To achieve normally-off operation, the carriers at gate region need to be fully depleted at zero bias. Due to a relatively limited depletion capability of the planar gate, the normally-off operation poses an upper limit on the channel carrier density, which increases the device on-resistance. 2) The electric field distribution is not uniform when the device is blocking off-state voltage, and the crowded electric field will cause the device premature breakdown. This work proposed to use multi-dimensional, p-n junction-based device structure to overcome the above challenges. The devices with diverse structures are fabricated, characterized, and compared with the commercially available devices. The multi-dimensional, junction-based gate structure provides strong electrostatic control to realize normally-off operation and allow for higher carrier concentration and lower on-resistance. The devices with multi-dimensional, junction-based drift region enables the uniform electric field distribution at the device off-state, allowing devices to block high voltage without compromising the on-state resistance. Examples of such devices investigated in this dissertation include the tri-gate junction transistors, reduced-surface-field (RESURF) diodes, and superjunction diodes. In summary, this work demonstrates the multi-dimensional, junction-based device structure to overcome the performance limitations of planar devices and fully exploit GaN's material benefits for power devices. The multi-dimensional, junction-based devices are experimentally fabricated and characterized, manifesting the superior performance over traditional GaN devices. This work will significantly boost the performance and application space of GaN power devices.

Doctor of Philosophy / Department of Physics / Hongxing Jiang / The optical and structural properties of Er-doped GaN/InGaN materials and devices synthesized by metal organic chemical vapor deposition (MOCVD) were investigated. Er-doped GaN via MOCVD emits a strong photoluminescence (PL) emission at 1.54 um using both above and below-bandgap excitation. In contrast to other growth methods, MOCVD-grown Er-doped GaN epilayers exhibit virtually no visible emission lines. A small thermal quenching effect, with only a 20% decrease in the integrated intensity of the 1.54 um PL emission, occurred between 10 and 300 K. The dominant bandedge emission of Er-doped GaN at 3.23 eV was observed at room temperature, which is red-shifted by 0.19 eV from the bandedge emission of undoped GaN. An activation energy of 191 meV was obtained from the thermal quenching of the integrated intensity of the 1.54 um emission line. It was observed that surface morphology and 1.54 um PL emission intensity was strongly dependent upon the Er/NH3 flow rate ratio and the growth temperature. XRD measurements showed that the crystalline ordering of the (002) plane was relatively unperturbed for the changing growth environment. Least-squares fitting of 1.54 um PL measurements from Er-doped GaN of different growth temperatures was utilized to determine a formation energy of 1.82 ± 0.1 eV for the Er-emitting centers. The crystalline quality and surface morphology of Er-doped InGaN (5% In fraction) was nearly identical to that of Er-doped GaN, yet the PL intensity of the 1.54 um emission from Er-doped InGaN (5% In fraction) was 16 x smaller than that of Er-doped GaN. The drop in PL intensity is attributed to the much lower growth temperature in conjunction with the high formation energy of the Er- emitting centers. Er-doped InGaN grown at fixed growth temperature with different growth pressures, NH3 flow rates, and Ga flow rates was also investigated, and showed that increased In fractions also resulted in a smaller 1.54 um PL intensity. Er-doped InGaN p-i-n diodes were synthesized and tested. The electroluminescence (EL) spectra under forward bias shows strong Er based emission in the infrared and visible region. The different emission lines from EL spectra in contrast to PL spectra implies different excitation methods for the Er based emission in the p-i-n diode than in the PL excited epilayer.

Wide-bandgap (WBG) semiconductor materials such as silicon carbide (SiC) and gallium-nitride (GaN) allow higher voltage ratings, lower on-state voltage drops, higher switching frequencies, and higher maximum temperatures. All these advantages make them an attractive choice when high-power density and high-efficiency converters are targeted. Two different gate-driver designs for SiC power devices are presented. First, a dual-function gate-driver for a power module populated with SiC junction field-effect transistors that finds a trade-off between fast switching speeds and a low oscillative performance has been presented and experimentally verified. Second, a gate-driver for SiC metal-oxide semiconductor field-effect transistors with a short-circuit protection scheme that is able to protect the converter against short-circuit conditions without compromising the switching performance during normal operation is presented and experimentally validated. The benefits and issues of using parallel-connection as the design strategy for high-efficiency and high-power converters have been presented. In order to evaluate parallel connection, a 312 kVA three-phase SiC inverter with an efficiency of 99.3 % has been designed, built, and experimentally verified. If parallel connection is chosen as design direction, an undesired trade-off between reliability and efficiency is introduced. A reliability analysis has been performed, which has shown that the gate-source voltage stress determines the reliability of the entire system. Decreasing the positive gate-source voltage could increase the reliability without significantly affecting the efficiency. If high-temperature applications are considered, relatively little attention has been paid to passive components for harsh environments. This thesis also addresses high-temperature operation. The high-temperature performance of two different designs of inductors have been tested up to 600_C. Finally, a GaN power field-effect transistor was characterized down to cryogenic temperatures. An 85 % reduction of the on-state resistance was measured at −195_C. Finally, an experimental evaluation of a 1 kW singlephase inverter at low temperatures was performed. A 33 % reduction in losses compared to room temperature was achieved at rated power. / <p>QC 20160922</p>

Experimental investigations of the optical properties of GaN nanostructured light emitting diode (LED) arrays are presented. Microphotoluminescence spectroscopy with pulsed and continuous wave lasers was used to probe the carrier dynamics and emission mechanisms of nanorod LED arrays fabricated by a top down etching method. Results show a possible reduction in internal electric field as nanorod diameter decreases. Localisation effects were also observed, affecting the spectral shape of the nanorod emission. Under two-photon excitation, quantum dot-like sharp spectral peaks in the PL spectra are found to exist in abundance amongst all the nanorod samples. The optical properties of these localised states, which are shown to be associated with the nanorod free-surfaces, are characterised using non-linear and time resolved spectroscopy. An investigation into spatially resolved single nanorods was also carried out. Single nanorods were isolated, and characterised using pulsed lasers. The etching is shown to increase the carrier decay life-time at extended intervals over several hundred ns. The temporal evolution and excitation power density dependence of the quantum dot-like states are also presented for the first time. The long lived localised states are thought to arise from surface effects, in particular Fermi-surface pinning, causing localisation and spatial separation of carriers. Additional work on nano-pyramid array LEDs, with quantum wells on semi-polar surfaces is also presented. Optical properties using micro-photoluminescence are compared to cathodoluminescence studies. An uneven distribution of emission wavelengths across the pyramid facet is thought to lead to an emission mechanism involving carriers transferring between multiple spatially localised states. Finally, experimental techniques and fabrication methods for future work are documented in detail.

Neste trabalho estudamos as propriedades físicas das impurezas de elementos de terra rara (TR) nos cristais de GaN, nas estruturas cristalinas zincblenda e wurtzita, e de ZnO. Para tal, consideramos as impurezas de Eu, Gd, Tb, Dy, Ho, Er e Tm substitucionais no sítio do cátion (Ga ou Zn), pois esta posição é mais estável do que no sítio do ânion (N ou O). As investigações foram efetuadas através de simulações computacionais baseadas em métodos de primeiros princípios dentro do formalismo da teoria do funcional da densidade e utilizando o método FP-LAPW (Full Potential Linear Augmented Plane Waves), implementado no código computacional WIEN2k, dentro do esquema de supercélula, com relaxações atômicas tratadas de modo apropriado. No tratamento dos estados de valência 3d (Ga ou Zn) e 4f dos átomos de TR foi introduzida uma correção local de Hubbard, para levar em conta efeitos de alta correlação eletrônica. Inicialmente, estudamos as propriedades eletrônicas e estruturais dos elementos de TR em suas fases metálicas. Estes resultados serviram para validar a metodologia de obtenção dos valores dos parâmetros U de Hubbard de modo autoconsistente, pois para este grupo de materiais existem resultados experimentais. Comparando a localização e o desdobramento entre os estados ocupados e desocupados relacionados aos orbitais 4f, em relação ao nível de Fermi dos sistemas, obteve-se uma ótima concordância. Nossos resultados para os sistemas dopados, utilizando a correção do potencial U de Hubbard, mostram que a suas descrições estão adequadas. Somente com a introdução da correção é possível localizar corretamente, em relação ao topo da banda de valência, os estados ocupados e desocupados relacionados aos orbitais 4f. Estes resultados estão de acordo com esquemas propostos na literatura, mostrando que o estado de oxidação das impurezas de TR é trivalente e que, em geral, em ambos sistemas cristalinos, não há a introdução de níveis de energia na região do gap dos materiais. Nosso estudo da correção do potencial U de Hubbard, obtido autoconsistemente, na teoria do funcional da densidade, mostra que seus valores não são universais, dependendo do estado de carga e do ambiente em que o átomo está inserido. Mais ainda, mostra que o procedimento adotado é totalmente apropriado para descrever a correlação eletrônica dos elétrons 4f. / In this work we studied the physical properties of rare earth (RE) impurities in GaN, in the zincblend and the wurtzite crystal structures, and in ZnO. Eu, Gd, Tb, Dy, Ho, Er and Tm substitutional impurities in the cation site (Ga or Zn) were considered, since this position is more stable than anion site (N or O). The investigations were carried out by computational simulations using ab initio methods, based on the density functional theory and the FP-LAPW (Full Potential Linear Augmented Plane Waves) method, as implemented in the WIEN2k code, considering the supercell approach and atomic relaxations. The 3d-Ga or 3d-Zn and the 4f-RE valence states were treated with the introduction of on-site Hubbard correction, in order to correctly describe the strongly correlated electrons. First, the electronic and structural properties of RE metallic systems were investigated and the results were compared with available experimental data, showing a good agreement. Those results helped to identify the appropriate procedure to compute the Hubbard U potential. This procedure should also provide a reliable description about the electronic properties of RE elements as impurities in semiconductors. Then, we have computed the properties of substitutional RE impurities in the cation site using the same methodology and procedures. The Hubbard U potential was necessary to correctly describe the position of the 4f occupied and unoccupied states, related to the valence band top. These results are in agreement with proposed electronic properties of RE doped GaN and ZnO found in the literature. The RE impurities oxidation states are in general trivalent in both crystal systems and introduce no energy levels in the gap region. Our Hubbard U potential correction, obtained in a self-consistent way, depends on the considered element and the neighborhood, and it is not an universal parameter. Moreover, our investigation shows that the adopted procedure is totally appropriate to describe the electronic correlation of the 4f electrons.

Im Rahmen dieser Arbeit werden Untersuchungen der Spinrelaxation in epitaktischen GaN-Schichten mit unterschiedlichen Donatorkonzentrationen und Versetzungsdichten mit Hilfe spin- und zeitaufgelöster differentieller Reflektanzspektroskopie präsentiert. Dabei wurden die optischen Anregungsbedingungen sehr sorgfältig gewählt. Neben der genauen Abstimmung der Anregungsenergie, unterstützt durch die Modellierung der differentiellen Reflektanz, wurden insbesondere spektral schmale Laserpulse verwendet. Diese erlauben eine selektive Anregung der untersuchten Übergänge. Es wurden Spinlebensdauern von 30 bis 170~ps bei tiefen Temperaturen für das freie A-Exziton bestimmt. In der Nähe des Metall-Isolator-Übergangs ließ sich eine langsamere Spinrelaxation als für schwächer dotierte Proben nachweisen. Die längsten beobachteten Spinrelaxationszeiten zeigen freistehende GaN-Schichten hoher Materialqualität mit sehr geringen Versetzungsdichten. In der Literatur besteht eine kategorische Unterteilung der Ergebnisse in lange elektronische Spinlebensdauern bis in den Nanosekundenbereich, erhalten mit Kerr-Messungen, und extrem kurze exzitonische Spinrelaxation in Reflektanz-Experimenten im (Sub-)Pikosenkundenbereich. Dieses Bild wird hier nicht bestätigt. Die beobachteten Spinrelaxationszeiten liegen eineinhalb bis zweieinhalb Größenordnungen über Ergebnissen, von denen bisher mit der hier verwendeten Methode berichtet wurde. Es wird gezeigt, dass die Beobachtungen extrem kurzer Spinrelaxationszeiten an anderer Stelle eine Folge der optischen Anregungsbedingungen sind. Die Verwendung sehr kurzer und damit spektral breiter Laserpulse, die eine selektive Exziton-Anregung verbieten, führt zu einem deutlich anderen zeitlichen Verhalten und stark verfälschten Ergebnissen. Diese Beobachtung löst den scheinbaren Widerspruch zwischen den beiden Ergebnisgruppen in der Literatur auf und bildet die Grundlage für weiterführende Untersuchungen. / In this work, an investigation of spin relaxation in GaN epitaxial layers with different doping concentrations and dislocation densities is presented. The measurements were carried out by the means of spin- and time-resolved differential reflectance spectroscopy. The conditions of optical excitation were chosen with special care. In particular, spectrally narrow laser pulses were used to achieve selective excitation of the examined transitions in addition to the precise adjustment of the excitation energy, supported by the modeling of the differential reflectance. The spin relaxation times obtained for the free A exciton at low temperatures are in the range of 30 to 170 ps. In the proximity of the metal insulator transition, a slower spin relaxation was observed than for lower doping concentrations. The longest spin relaxation times were found in high quality, free-standing GaN layers with very low dislocation densities. Existing results in the literature can be strictly grouped into long electronic spin lifetimes of up to a few nanoseconds, obtained with Kerr rotation, and extremely short spin relaxation in the (sub)picosecond range, measured with reflectance experiments. This picture cannot be confirmed here. The spin relaxation times observed here lie 1.5 to 2.5 orders of magnitude above the values previously reported using the same experimental method. It is shown that the instances of extremely fast spin relaxation are caused by the properties of the optical excitation. The use of ultra-short and thus spectrally broad laser pulses, which prohibits the selective excitation of excitons, leads to a significantly different temporal behavior and strongly distorted results. This finding elucidates the apparent conflict between the two groups of results and forms the basis for further investigations.

In der vorliegenden Arbeit werden verschiedene Aspekte des Einflusses von Lokalisation, Unordnung und Polarisationsfeldern auf Elektron-Loch Zustände in Quantengräben (QWs von engl. quantum wells) aus Halbleitern mit großer Bandlücke theoretisch untersucht. Unter Verwendung eines Schwerpunktseparationsansatzes wird das Verhalten von QW Exzitonen und Biexzitonen im Grenzfall schwacher Lokalisation beschrieben. Es zeigt sich, daß die Lokalisationsenergie des Biexzitons mehr als doppelt so groß ist wie die des Exzitons. Dies wird verursacht durch ein universelles Gesetz der Lokalisation in schwachen zwei-dimensionalen Potentialen, welches lediglich durch das "Potentialvolumen" und die Masse des lokalisierten Teilchens bestimmt wird. Ein einfaches Modell des QW Biexzitons wird entwickelt, dessen Ergebnisse gut mit jenen übereinstimmen, die man mit Hilfe eines aufwendigeren numerischen Modells erhält. Der Grenzfall starker Lokalisation von QW Exzitonen und höheren Exzitonenkomplexen wird mittels einer Dichtefunktionalrechnung untersucht. Es wird gezeigt, daß Zustände bis mindestens zum X4 in den nm-großen Potentialminima lokalisieren können, die durch Phasenseparation in (In,Ga)N/GaN QWs enstehen. Es wird das Übergangsspektrum des sukzessiven Zerfalls eines lokalisierten X4 berechnet. Auf Grundlage der selbstkonsistenten Lösung von Poisson- und Schrödinger-Gleichung wird der Einfluß des Probendesigns von (In,Ga)N/GaN QW-Strukturen auf den makroskopischen Verlauf des Polarisationsfeldes in Wachstumsrichtung und somit auf optische Übergangsenergie und Oszillatorstärke systematisch untersucht. Besondere Bedeutung kommt dabei der Abschirmung der Felder durch Raumladungszonen zu. Es wird gezeigt, daß die Position des QW bezüglich einer ausgedehnten Oberflächen-Verarmungszone - die in n-dotierten, Ga-polarisierten Proben auftritt - erheblichen Einfluß auf Übergangsenergie und Oszillatorstärke hat. Durch die räumliche Variation der Polarisationsfeldstärke in dieser Verarmungszone kann das optische Übergangsspektrum eines Mehrfach-QW Schultern oder mehrere Maxima aufweisen. Indium Oberflächen-Segregation ruft eine Blauverschiebung der Übergangsenergie hervor, die bis zu einem Drittel der vom Polarisationsfeld verursachten quantum confined Stark-Verschiebung kompensiert. Diese Blauverschiebung wird von einer Verringerung des Elektron-Loch Überlapps begleitet. Die Polarisationsfelder in (In,Ga)N/GaN Mehrfach-QWs verschmieren das stufenförmige Einteilchen-Absorptionsspektrum. Durch die Aufhebung der näherungsweisen Diagonalität von Inter-Subband Übergängen und durch die Miniband-Dispersion in höheren, gekoppelten Zuständen haben diese Felder, neben dem Beitrag von Potentialfluktuationen, einen entscheidenden Einfluß auf die Form des Absorptionsspektrums. Ein in der Literatur diskutierter Mechanismus, der allein durch Polarisationsfelder eine Verbreiterung optischer Spektren hervorruft, kann nicht bestätigt werden. Unter Annahme einer unkorrelierten Zusammensetzung von (In,Ga)N und einer lateral korrelierten Grenzflächenrauhigkeit von einer Monolage in jeder Grenzfläche zeigt die spektrale Breite des Exzitonen-Schwerpunktpotentials eine Verschmälerung mit zunehmendem Feld. Diese wird verursacht durch das Eindringen der Teilchen in die binären Barrieren und durch ein vergrößertes Exzitonenvolumen. Im Fall einer langreichweitigen Grenzfächenrauhigkeit findet man eine Aufspaltung des Spektrums in einzelne Linien. / In this thesis, various aspects of the influence of localization, disorder, and polarization fields on electron-hole states in wide-gap semiconductor quantum wells (QWs) are investigated theoretically. A theoretical treatment of quantum well exciton and biexciton states in the limit of weak localization is presented, using a center-of-mass separation ansatz. It shows that the localization energy of the biexciton is more than twice as large as that of the exciton due to the universal behaviour of localization in weak two-dimensional potentials which is ruled only by the potential "volume" and the mass of the localized particle. A useful simple model of the QW biexciton wavefunction is developed which provides good agreement with the results obtained with an extensive numerical solution. The limit of strong localization of QW excitons and higher exciton complexes is investigated with a density functional calculation. It is demonstrated that states at least up to X4 may localize in nm-scale potential boxes caused by indium phase separation in (In,Ga)N/GaN QWs. The transition spectrum of the successive recombination of a localized X4 is calculated. A systematic investigation of the influence of the sample design of (In,Ga)N/GaN QW structures on optical transition energy and oscillator strength reveals the importance of space charge layers with regard to screening of polarization fields along the QW-axis. Based on a self-consistent solution of the Schrödinger-Poisson equations, the overall situation of the macroscopic spontaneous and piezoelectric polarization fields is discussed in dependence on various substantial sample parameters. It is found that the position of a QW in the sample with respect to an extended surface depletion layer - which is shown to exist in n-type Ga-face grown material - severely affects transition energy and electron-hole overlap. Due to the spatial variation of the field strength in this surface depletion layer, the optical transition spectrum of a Ga-face grown multiple-QW can display shoulders or even a multiple-peak structure. Indium surface segregation results in a blueshift of the transition energy compensating up to one third of the quantum confined Stark shift produced by the polarization field. This blueshift is accompanied by a decrease of the electron-hole overlap. Polarization fields in (In,Ga)N/GaN multiple-QWs result in a smoothing of the step-like single-particle absorption spectrum. Apart from the contribution of compositional fluctuations, the fields have significant influence on the shape of the spectrum via the abrogation of the nearly diagonality of inter-subband transitions and via the mini-band dispersion of higher coupled states in case of a periodic structure. A line broadening-mechanism due to polarization fields in (In,Ga)N/GaN QWs, as sometimes discussed in literature, could not be confirmed. Assuming uncorrelated (In,Ga)N alloy and in-plane-correlated interface roughness of one monolayer in each interface, the calculation of the spectral width of the QW exciton center-of-mass potential yields a narrowing with increasing average field. This is a result of the penetration of the carriers into the barriers and of an increasing exciton volume. In case of a long-range interface roughness, a splitting of the spectrum into individual lines can be predicted.

Diese Arbeit beschreibt die optische Charakterisierung mittels Photolumineszenzspektroskopie (PL) von Halbleiter-Nanodrähten (ND) im allgemeinen und einzelnen GaN-ND und GaN-ND-Ensembles im speziellen. ND werden oftmals als vielversprechende Bausteine zukünftiger, kleinster Bauele- mente bezeichnet. Diese Vision beruht insbesondere auf einigen attraktiven Eigenheiten, die ND im allgemeinen zugeschrieben werden. Im ersten Teil dieser Arbeit werden exemplarisch einige dieser Eigenschaften näher untersucht. So wird anhand von temperaturabhängigen PL-Messungen an Au- und selbstinduzierten GaAs/(Al,Ga)As-ND der Einfluss des Keimmaterials auf die PL der ND untersucht. Weiterhin werden die optischen Eigenschaften von ZnO-ND untersucht, die auf Si-, Saphir- und ZnO-Substraten gewachsen wurden. Die optische Charakterisierung von GaN-ND nimmt den Hauptteil dieser Arbeit ein. Die detaillierte Untersuchung einzelner GaN-ND und von GaN-ND-Ensembles zeigt die Relevanz des großen Oberflächen-zu-Volumen-Verhältnisses und dass jeder ND ganz eigene optische Eigenschaften aufweist. Die unerwartet starke Verbreiterung des strahlenden Übergangs donatorgebundener Exzitonen wird durch das vermehrte Auftreten von Oberflächendonatoren erklärt, deren statistische Relevanz durch PL-Messungen an einzelnen ausgestreuten und freistehenden GaN-ND nachgewiesen werden kann. Weiterhin wird der Einfluss elektrischer Felder auf die optischen Eigenschaften von GaN-ND ermittelt. Die Ein- und Auskopplung von Licht mit GaN ND wird mithilfe von Reflektanz- und Ramanmessungen bestimmt. Die zentralen Ergebnisse dieser Arbeit motivieren die Einführung eines Modells, dass die typischerweise nichtexponentielle Rekombinationsdynamik in ND-Ensemblen erklärt. Es basiert auf einer Verteilung der Rekombinationsraten. Vorläufige Ergebnisse dieses Modells beschreiben das nichtexponentielle Rekombinationdynamik in GaN ND-Ensemblen zufriedenstellend und erlauben eine Abschätzung ihrer internen Quanteneffizienz. / This thesis presents a detailed investigation of the optical properties of semiconductor nanowires (NWs) in general and single GaN NWs and GaN NW ensembles in particular by photoluminescence (PL) spectroscopy. NWs are often considered as potential building blocks for future nanometer-scaled devices. This vision is based on several attractive features that are generally ascribed to NWs. In the first part of the thesis, some of these features are examined using semiconductor NWs of different materials. On the basis of the temperature-dependent PL of Au- and self-assisted GaAs/(Al,Ga)As core-shell NWs, the influence of foreign catalyst particles on the optical properties of NWs is investigated. The effect of the substrate choice is studied by comparing the PL of ZnO NWs grown on Si, Sapphire, and ZnO substrates. The major part of this thesis discusses the optical properties of GaN NWs. The investigation of the PL of single GaN NWs and GaN NW ensembles reveals the significance of their large surface-to-volume ratio and that each NW exhibits its own individual recombination behavior. An unexpected broadening of the donor-bound exciton transition is explained by the abundant presence of surface donors in NWs. The existence and statistical relevance of these surface donors is confirmed by PL experiments of single GaN NWs which are either dispersed or free-standing. Furthermore, the influence of electric fields on the optical properties of GaN NWs is investigated and the coupling of light with GaN NWs is studied by reflectance and Raman measurements. The central results of this thesis motivate the introduction of a model that explains the typically observed nonexponential recombination dynamics in NW ensembles. It is based on a distribution of recombination rates. Preliminary simulations using this model describe the nonexponential decay of GaN NW ensembles satisfactorily and allow for an estimation of their internal quantum efficiency.

Les nitrures d’éléments III, à savoir GaN, InN, AlN et leurs alliages, forment une famille de matériaux semi-conducteurs dont les propriétés sont particulièrement intéressantes pour la réalisation de diodes électroluminescentes (LEDs). Leur intérêt réside en particulier dans leur bande interdite qui est directe et qui couvre une large bande spectrale de l’infrarouge (0,65eV pour InN) à l’ultraviolet (6,2eV pour AlN). En raison de l’absence de substrats accordés en maille avec ces matériaux, les couches minces hétéroépitaxiées de nitrure sont généralement touchées par des problèmes de qualité cristalline. Grâce au phénomène de relaxation des contraintes en surface, les nanofils offrent une solution prometteuse pour résoudre ce problème. Ils combinent de nombreux autres avantages : en comparaison des couches minces, l’efficacité quantique interne des LEDs peut être améliorée (surface effective plus importante permettant de diminuer l’effet Auger à courant injecté identique, absence de champ de polarisation en utilisant les facettes non polaires des nanofils) et l’extraction des photons est facilitée par l’effet guide d’onde des nanofils. Cependant, une des difficultés est de parvenir à contrôler la synthèse de ces nano-objets pour garantir une homogénéité des propriétés structurales d’un fil à l’autre et au sein d’un même fil. Dans ce contexte, mon travail de thèse a consisté à étudier d’un point de vue expérimental et théorique l’impact des inhomogénéités structurales sur les propriétés optoélectroniques de dispositifs à nanofils de type LED. J’ai pu mettre en évidence et modéliser un effet de concentration du courant dans les régions riches en indium lorsque les courants injectés sont modérés. Pour de forts courants, le courant se concentre à proximité du contact sur la coquille dopée p. Théoriquement, j’ai montré que la dérive des porteurs de charge dans les puits quantiques et leur diffusion unipolaire et ambipolaire en présence d’un gradient de compositions des puits étaient négligeables. Par ailleurs, je me suis également intéressé à l’interprétation des caractéristiques courant-tension. A l’aide d’un modèle simple, j’ai également identifié la présence de courant de fuite par effet tunnel dans des structures présentant une densité importante de défaut. Dans une seconde partie de ma thèse, je me suis également intéressé à la caractérisation de nanofils à structure coeur-coquille par la technique de courant induit par faisceau d’électrons (Electron Beam Induced Current). La dépendance des cartographies EBIC en fonction de la tension appliquée et de l’énergie du faisceau incident a été modélisée. Ce travail m’a notamment amené à proposer une nouvelle méthode de caractérisation permettant de cartographier les résistivités du coeur et de la coquille des nanofils. / III-nitrides i.e. GaN, InN, AlN and their alloys are semiconductors of choice to fabricate optoelectronic devices such as Light Emitting Diodes (LEDs). One of their most interesting features relies in their direct band gap that covers a very wide spectral range, from infrared (0.65 eV for InN) to UV (6.2 eV for AlN). However, the lack of lattice-matched substrate has been responsible for strong crystalline quality issues in heteroepitaxial thin films. Thanks to stress relaxation at their surface, nanowires provide a smart solution to this problem. Besides, they have a few more assets. In comparison to thin films, nanowires can improve the internal quantum efficiency of LEDs because of their higher effective surface that leads to lowered current densities and thus mitigated Auger effect. The internal quantum efficiency also benefits from the possibility to grow the active region on non polar facets, thus getting rid of the detrimental high internal polarization-induced electric field in quantum wells. Furthermore, the photon extraction efficiency is enhanced by the guiding effect of nanowires. However, despite all this promising advantages, one of the main challenges remains the control of structural homogeneity from wire to wire but also inside single wires.In this context, my work has consisted into studying from an experimental and theoretical point of view the consequences of these structural inhomogeneities on the optoelectronic properties of nanowire based LED devices. I have shown that the current tends to gather into indium-rich regions for moderate bias. At higher bias, the dominant current path though the junction is generally located under the p-contact on the nanowire shell. I have theoretically demonstrated that the unipolar and ambipolar diffusion of carriers as well as their drift induced by a composition gradient inside the quantum wells is not significant in the devices I have studied. Moreover, I took also an interest in the detailed analyze of I-V curves. Thanks to a simple model, I have identified the presence of leakage current related to defect- and phonon-assisted tunneling effect. In the second part of my work, I have focused onto the characterization of core-shell wires using the Electron Beam Induced Current technique. The bias-dependant and acceleration voltage-dependant EBIC maps has been explained with a theoretical model based on equivalent circuits. This study leads me to suggest a new experimental method that can be used to map the nanowire core and shell resistivity.

Ces travaux de thèse portent sur les phénomènes d’amplification du signal de diffusion Raman par effet de surface et par effet de pointe. Des réseaux de motifs métalliques de taille nanométrique arrangés spatialement ont été fabriqués par la méthode de transfert Langmuir-Blodgett et par lithographie à faisceau d’électrons. De telles structures de géométries contrôlées déposées à la surface de lamelles de microscope ont été développées afin d’amplifier le signal Raman de molécules adsorbées par effet SERS (Surface Enhanced Raman Spectroscopy). Ces nanostructures triangulaires en or de taille proche de la longueur d’onde ont des bandes de résonance plasmon dans le domaine spectral visible. En utilisant une source de laser appropriée dans ce domaine spectral, les facteurs d’amplification Raman d’une couche mono-moléculaire d’un dérivé azobenzène sont de plusieurs ordres de grandeur, et ce pour les deux techniques de nano-lithographie employées. Afin de compléter ces premiers résultats, des réseaux de fils d’or avec de grands facteurs de forme ont été fabriqués. Ces derniers montrent des résonances plasmons multipolaires et des facteurs d’amplification de l’ordre de 105. Les techniques de microscopie en champ proche ont également été développées afin de localiser précisément l’exaltation Raman et d’accroitre la résolution spatiale de mesures Raman. Des pointes métalliques en or de taille nanométrique ont ainsi permis d’amplifier localement le signal de diffusion de molécules placées à leur proximité par effet TERS (Tip Enhanced Raman Spectroscopy). Les développements logiciels et mécaniques entre un microscope confocal Raman et un microscope à force atomique ont été implémentés afin de contrôler simultanément les deux instruments. Ce montage expérimental a été appliqué à l’étude de nanofils semi-conducteurs de nitrure de gallium permettant de suivre leur signal vibrationnel avec une résolution spatiale inférieure à 200 nm. / This thesis work focuses on Raman scattering enhancements by metallic nanostructures. In the first part of this work, arrays of metallic patterns with nanometer dimensions were fabricated by the Langmuir-Blodgett deposition technique and electron-beam lithography. Such structures made of gold were fabricated onto microscope slides with the goal to enhance the Raman signal through SERS effect (Surface Enhanced Raman Spectroscopy). These patterns formed by an assembly of triangular nanostructures with sizes of hundreds of nanometers, exhibit plasmon resonance bands in the visible spectral region. By using an appropriate excitation laser source with respect to the plasmon frequency, Raman enhancement factors of a monolayer were found to be of several order of magnitude for both Langmuir-Blodgett and electron-beam lithography platforms. To further complement these results, gold wires arrays with large aspect ratio made by electron-beam lithography showed multipolar plasmon resonances with enhancement factors up to 105. In the second part of this thesis, near-field Raman microscopy has been developed with the aim to localize precisely the Raman enhancement and improve spatial resolution of Raman measurements. Atomic force microscopy gold tips have been used to locally enhance scattering signal of molecules in close proximity to the tip opening new opportunities. This approach known as TERS (Tip Enhanced Raman Spectroscopy) is of significant interest to probe nanomaterials, nanostructures or monolayers. Software and mechanical developments have been made between a confocal Raman microscope and an atomic force microscope to control simultaneously both instruments. This experimental setup was used to characterize gallium nitride semi-conductors nanowires with spatial resolution better than 200 nm.