Metal organic vapour phase epitaxy of GaN and lateral overgrowth

Abstract

Gallium nitride (GaN) is an extremely promising wide band gap semiconductor material for optoelectronics and high temperature, high power electronics. Actually, GaN is probably the most important semiconductor since silicon. However, achievement of its full potential has still been limited by a dramatic lack of suitable GaN bulk single crystals. GaN has a high melting temperature and a very high decomposition pressure; therefore it cannot be grown using conventional methods used for GaAs or Si like Czochraslski or Bridgman growths.Since there is no GaN bulk single crystal commercially available, all technological development of GaN-based devices relies on heteroepitaxy. Most of the current device structures are grown on sapphire or 6H-SiC. However, since their lattice parameters and thermal expansion coefficients are not well-matched to GaN, the epitaxial growth generates huge densities of defects, with threading dislocations (TDs) being the most prevalent (109–1011 cm−2). As a comparison, homoepitaxially grown GaAs exhibits ∼102–104 dislocation cm−2, and homoepitaxial Si almost 0. Actually this large density of TDs in GaN drastically limits the performance and operating lifetime of nitride-based devices. Therefore, there is currently a tremendous technological effort to reduce these defects.Metal organic vapour phase epitaxy (MOVPE) is currently the most widely used technology. Actually, all optoelectronic commercial device structures are fabricated using MOVPE. In MOVPE, the most appropriate precursor for nitrogen is ammonia (NH3), whereas either trimethyl or triethylgallium may be used as a gallium source. MOVPE of GaN requires a high partial pressure of NH3, high growth temperatures (∼1000–1100°C) and a growth chamber specially designed to avoid premature reactions between the ammonia and gallium alkyls. Since sapphire (or 6H-SiC) and GaN are highly mismatched, direct growth of GaN is impossible. Therefore, the growth of GaN on any substrate first requires the deposition of a buffer layer, which, to some extent, accommodates the mismatch. Using appropriate nucleation layers allows a reduction of the dislocation density to the low 108 cm−2 range.Though laser diodes (LDs) were demonstrated in the late 1990s with such defect layers, the real breakthrough in laser technology was the dramatic improvement of the LD lifetime at the end of 1997, with the lifetime reaching 10 000 h. This was made possible by implementation of epitaxial lateral overgrowth (ELO) technology, which significantly reduces the dislocation density to below 107 cm−2.In ELO technology, parts of the highly dislocated starting GaN are masked with a dielectric mask, after which growth is restarted. At the beginning of the second growth step, deposition only occurs within the openings, with no deposition observed on the mask. This is referred to as selective area epitaxy (SAE). The TDs are prevented from propagating into the overlayer by the dielectric mask, whereas GaN grown above the opening (coherent growth) keeps the same TD density as the template, at least during the early stages of growth.Currently, two main ELO technologies exist: the simpler one involves a single growth step on striped openings. In this one-step-ELO (1S-ELO), growth in the opening remains in registry with the GaN template underneath (coherent part), whereas the GaN over the mask extends laterally (wings). This leads to two grades, namely highly dislocated GaN, above the openings, and low dislocation density GaN, above the masks. With this technique, devices have to be fabricated on the wings. Conversely, in the two-step-ELO (2S-ELO) process, the growth conditions of the first step are monitored to obtain triangular stripes. Inside these stripes, the TDs arising from the templates are bent by 90° when they encounter the inclined lateral facet. In the second step, the growth conditions are modified to achieve full coalescence. In this 2S-ELO technology, only the coalescence boundaries are defective. ELO technology produces high quality GaN, with TD densities in the mid 106 cm−2, line widths of the low temperature photoluminescence near band gap recombination peaks below 1 meV, and deep electron trap concentration below 1014 cm−3 (compared with mid 1015 cm−3 in standard GaN). Numerous modifications of the ELO process have been proposed either to avoid technological steps (maskless ELO) or to improve it (pendeoepitaxy, PE). To further reduce the TD density, multiple-step-ELO and pendeo have also been implemented.However, even ELO quality GaN is not good enough for the next generation of LDs. ELO samples do not yet offer a full surface suitable for laser technology. What is needed for LDs with at least 30 mW output power is high quality freestanding GaN with TDs close to or even below 106 cm−2. To reach this crystalline perfection, elaborate technologies are currently being implemented. They, at some stage, involve TD reduction mechanisms occurring in the ELO process.Self-supported GaN with at least ELO quality at an affordable cost is believed to be the next breakthrough in GaN technology.