Technical Insight
Crystalline GaN grows on an amorphous platform
Tiny tunnels enable the growth of single-crystalline GaN on amorphous substrates
Crystalline GaN is created on SiO2 via growth in a tunnel formed on the substrate (see top). GaN grows laterally, on the side of AlN, which is shown in red. Confinement in two orthogonal directions leads to the formation of a large, single GaN crystal, shown in blue. The AlN stripe and the tunnel are formed through growth, patterning, SiO2 deposition and etching steps (see bottom).
Yale University engineers are claiming to have produced the first crystalline, planar GaN with dimensions of tens of microns on an amorphous substrate.
This accomplishment follows the recent success by engineers at Samsung, who grew nearly crystalline GaN on fused silica.
“The limitation of [Samsung’s] work is that the form of GaN is necessarily pyramid-shaped, with dimensions of a few micrometres,” explains lead-author Benjamin Leung. The size of these crystals, which are not orientated to each other, imposes severe restrictions on both device geometries and non-standard device processing.
Leung and his co-workers form their single-crystal GaN on SiO2/silicon (100) templates by a relatively complex, patented approach that involves the deposition of a textured AlN film and the growth of GaN in a tunnel. The team claims that it can routinely produce single-crystal GaN as large as 10 μm by 20 μm. They believe that this size is big enough for making some devices , including GaN FETs.
“Contemporary heterogeneous integration – for example in silicon photonics – does not emphasize planar wafer bonding or heteroepitaxy, but rather adapting to a new paradigm of preparing device islands or chiplets,” argues Leung. This shift is driving interest in single crystals with dimensions that match the requirements of the device.
“We see tremendous opportunity to realise traditional and novel device structures,” claims Leung, who believes that his team’s technology occupies the middle ground between a top-down approach and a bottom up one. “[It combines] the precise positioning and dimension controls of lithographic procedures with the flexibility and parallelism of a bottom up process.”
The team’s process for forming GaN begins with deposition of a textured AlN film by magnetron sputtering. This film, which is fibrously textured, reduces the degree of freedom in the material from three – randomly ordered – to just one. It features a random in-plane orientation, but includes orientational distribution along the c-axis.
GaN deposited on this AlN film retains this characteristic, and its remaining degree of freedom is eliminated with a method that is refereed to as evolutionary selection. The result is crystalline GaN.
To form this, the AlN-coated wafer is processed into AlN stripes, before a blanket of SiO2 is added and then partially removed to expose one end of the AlN stripes (see Figure). These stripes are etched back so a little material remains, creating a tunnel. GaN is then grown in this tiny cavern, along a direction perpendicular to the surface normal, which is also perpendicular to the growth axis of the initial AlN that provides a textured seed.
The engineers from Yale have formed GaN in a tunnel that is 4 μm wide, 0.65 μm high and 20 μm long. GaN is polycrystalline when it nucleates on GaN, according to electron backscatter diffraction, but it is transformed to a uniform single-crystalline grain over a distance of 4 μm, thanks to evolutionary selection. Plan-view, high-resolution, transmission electron microscopy confirms that GaN is a single crystal, and reveals a hexagonal six-fold arrangement of the wurtzite lattice structure.
Vertical confinement is not optional for forming crystalline GaN. When the researchers have omitted this, polycrystalline GaN is formed along the entire length of the channel.
Leung says that the process that they have developed could be applied to other materials. “Unfortunately, AlN does not grow selectively in an MOCVD process, due to its high sticking coefficient, and it will randomly nucleate on the amorphous substrate.” But many other materials can grow selectively, and should form single crystals with Yale’s process, including InN, GaAs, InP, InAs, ZnO, silicon and germanium.
The team’s goals including extending its technology to other material systems, and demonstrating devices based on this platform.
Leung et al. Adv. Mater. 25 1285 (2013)
