Electrical Characterization of ZnO thin films grown by molecular beam epitaxy

Vorschau

Inhaltsverzeichnis, Datei (52 KB)
Leseprobe, Datei (210 KB)

For the electronic and optoelectronic device realization a precise control of the electrical properties in the utilized material is a very important issue. Doping profiles in realized p-njunctions influence the functionality of the devices. The morphological and crystal properties of a device material directly influence the electrical ones. Dislocations present in a region of p-n-junctions can short circuit them leading to malfunctions. Too rough surfaces during epitaxial growth could lead to inhomogeneities in a single or multiple quantum wells and superlattices.

The main goal of the present work was to provide the basis for a reliable p-type doping of ZnO grown by molecular beam epitaxy. Firstly, the well established heteroepitaxial growth on c-sapphire substrates has been employed. Based on the theoretical and experimental works, suggesting nitrogen to be the impurity that builds the most shallow acceptor level in ZnO comparing to other group-V elements, it has been implied as a dopant. To generate reactive nitrogen atoms an rf-plasma source has been utilized in the MBE process. The resulting samples have been characterized by such methods as AFM, XRD, TEM, PL spectroscopy, temperature domain Hall measurements (TDHM) and ECV-profiling. First results of TDHM have shown that even in undoped samples the temperature dependencies of the electron mobility and carrier concentration have regions which are difficult to interpret. It is necessary to fit them with theoretical curves in order to extract the correct values. This task has proven to be very difficult. The complicated character of the dependencies has been explained in terms of the multilayer conduction model dividing a layer in thin interfacial region with mobility and carrier concentration μ1 and n1 respectivly and bulk region with a higher mobility μ2 and lower carrier concentration n2. The electrical transport in the bulk region has been modeled in terms of the general scattering theory in polar semiconductors. Such scattering mechanisms as scattering on polar-optical phonons, piezoelectric phonons, acoustic deformation potential, strain induced fields, dislocations, ionized and neutral impurities have been taken into account. Two cases have been considered to model transport in the interfacial region: 1) transport takes place in the conduction band of a highly doped degenerate semiconductor; 2) transport takes place in the impurity band formed by intermediate concentration of impurities and in conduction band in parallel. In the second case transport at the interface in conduction band has been neglected in the region of the low temperatures due to the impurities freeze-out and carrier concentration has been taken temperature independent like in the first case. To investigate experimentally the transport character in these two regions independently a mobility-spectrum analysis has been conducted. Theoretical results utilizing the two models have been compared with experimentally extracted mobility and carrier concentration in the interfacial region. It has been concluded that the concentration of donors in the layers is not high enough for the impurity band to merge with the conduction band and the second model is more consistent. The theoretically acquired donor concentration profiles have been compared with ECV-profiles. The agreement is very good. Simulations have revealed a shallow donor state with the ionization energy of approximately 45 meV . In the literature, this donor state in ZnO is attributed to hydrogen. However, due to the high diffusion mobility of hydrogen in ZnO, an annealing process would obviously decrease the carrier concentration in the samples which has not been the case. It has been suggested that the main donor centers are the electrically active crystal point defects generated by dislocations. Layers doped with nitrogen have been grown at very low temperatures (≈ 200°C) and at temperatures ranging from 400°C to 500°C, which are optimal for the epitaxial growth of ZnO. The samples grown at low temperatures are single crystalline with mosaic structure. In both cases, the introduction of the dopant increased the carrier concentration. This has been accounted for a bad crystal quality resulting in the inhomogeneous incorporation of nitrogen and for high background donor concentration due to the high dislocations densities. Additionally, the incorporation of acceptor centers shifts the Fermi-level increasing the formation probability of the compensating point defects. The analysis of TDHM showed an inconsistency of the one donor level model in the case of nitrogen doped samples. This fact and the decrease in the carrier concentration after annealing at 800°C for 30 minutes in ambient air can be explained by nitrogen forming donor-like defect complexes. In an attempt to improve the crystal quality of the heteroepitaxial layers, 15 periods of a ZnO/Zn0.6Mg0.4O superlattice structure have been inserted between the conventional double HT-MgO/LT-ZnO buffer and a main HT-ZnO layer. TDHM has revealed a very high mobility close to the values measured in a bulk ZnO for the temperature range of 20 – 300 K. However, TEM investigations of the samples have not indicated any decrease in the dislocation density comparing with the similar samples without a superlattice. Such a high mobility has been attributed to an electron transport in the superlattice structure. Heteroepitaxial growth of high quality ZnO-layers has proven to be challenging leaving the homoepitaxial growth as the only possibility to obtain the epitaxial layers with the best structural and electrical properties. The hydrothermally grown bulk ZnO substrates from two supplying companies, CrysTec and TokyoDenpa, have been employed for homoepitaxy. The substrates from CrysTec have not been epi-ready. Although AFM images reveal very flat surface, this has been damaged by the process of the chemomechanical polishing. This damaged layer must be removed. This has been achieved by the thermal annealing for 3 hours at 1050°C in ambient air. The thermally treated surfaces resulted in atomically flat terraces. XRD measurements have indicated an improvement of the crystal quality after annealing. The resistivity of the bulk substrates decreased after the thermal treatment due to out-diffusion of the compensating Li atoms letting Al, Ga and In atoms to contribute to conduction. After the longer annealing processes the etch-pits have been discovered on O-polar faces. The same features could be achieved by the chemical etching in a nitric acid on Zn-polar faces. The density of the threading dislocations on both polar faces for both types of substrates calculated by the etch-pit density investigation is about 105 1/cm2. Further the thermally treated substrates with atomically flat terraces have been utilized for homoepitaxy. The differences in growth kinetics during the molecular beam epitaxy on such substrates with the improved surface quality depending on their polarity have been investigated by RHEED measurements. The growth on a Zn-polar face has a 3D-character independently on a supplier. Morphologies of the resulting O- and Zn-polar layers have shown to be different. This has been explained by the presence of dangling bonds on Opolar face and thus, shorter diffusion time of the impinging Zn atoms on the surface. XRD and TEM measurements have shown a perfect crystal quality of the overgrown layers. The PL spectra of homoepitaxial layers are governed by the donor impurities diffused from the substrates. Considering the SIMS measurements of homoepitaxial layers found in the literature it has been concluded that the diffusion of donors in the layers grown on Zn-polar faces takes less effect then for the O-polar films. This conclusion has enforced the utilization of Zn-polar substrates supplied by CrysTec for the experiments with nitrogen doping of ZnO because of their affordable price. The electrical properties measured by ECV-profiling in series of homoepitaxial layers with varied growth parameters have shown an increase of the carrier concentration with the nitrogen incorporation. In addition, it has also been shown that the resulting electrical properties near the interface are governed mostly by the initial properties of the substrates. With increasing thickness of the layers carrier concentration saturated to the values of around 1016 1/cm3. The recent successful realization of the p-type MgZnO layers on TokyoDenpa substrates by researchers from Japan suggests switching to the p-type doped alloys because the above discussed results indicate that p-type doping with nitrogen of a pure ZnO is very difficult or even impossible. This is due to a rather fundamental reason: the formation of the compensating donor centers with the incorporation of acceptor atoms. As the first step in the future works, it is obvious to try to reproduce the results of the ZnMgO p-type doping with nitrogen employing growth on ZnO substrates.