Technical Insight
Mechanical transfer with boron nitride
Unlike etching or laser-lift off, mechanical approaches for separating GaN epitaxial structures from their substrates can be quick, simple and scalable to large areas.
A mechanical approach for separating GaN devices from their sapphire substrates promises to replace the sacrificial etching and laser lift-off techniques used in today’s fabs, thanks to the recent development of BN films by engineers at NTT Basic Research Laboratories, Taiwan.
This team has demonstrated the strength of its approach by first depositing hexagonal BN films on sapphire substrates, and then growing various structures onto these templates , including: AlGaN/GaN structures with high electron mobility, and epitaxial layers for making multiple-quantum well LEDs. High-quality chips from 5 mm square to 2 cm square have been extracted from these wafers by mechanical lift-off and transferred to other substrates.
Corresponding author Yasuyuki Kobayashi believes that the BN-based mechanical transfer process could be popular with both LED and HEMT manufacturers: “We have already demonstrated that a very thin flexible GaN-based LED can be fabricated in a pair of laminate films, which may be of interest to LED manufacturers. Our technology also makes it feasible to transfer AlGaN/GaN HEMTs onto any materials having high thermal conductivity, which may be attractive for transistor manufacturers who have been suffering from heating problems.”
NTT has already filed several patent applications for its process, and it plans to license the technology to other chipmakers.
The Japanese lab started working on BN, a wide bandgap material that could be used for making devices operating in the deep ultraviolet spectral range, in 2005. Growth of single phase, hexagonal BN by MOCVD followed in 2007, and in 2008 this growth technology yielded epitaxial films of this wide bandgap semiconductor on sapphire.
“This success opened an avenue of one-step, damage-free release and transfer of a wide range of GaN-based devices,” claims Kobayashi. Epitaxial growth of BN by MOCVD is very challenging. “No lattice-matched substrate is available for BN, at least in affordable form,” says Kobayashi. He explained that specially designed substrate heating equipment is needed to achieve substrate temperatures of 1300 °C to 1500 °C, which are reported to be optimal for the growth of BN films.
To demonstrate the promise of their technology, the team deposited a BN layer on sapphire, before adding different structures. They found that inserting an AlN layer between the buffer and LED improved surface morphology – when GaN is grown directly onto hexagonal BN, it forms a rough, irregular island-shaped surface morphology, and is polycrystalline. Thanks to the addition of AlN, it is possible to form GaN films with a step-like flat surface and a root-mean-square roughness over a 5 µm by 5 µm area of just 0.69 nm, according to atomic force microscopy measurements.
Dark-field transmission electron microscopy reveals another benefit of the AlN layer: It acts as a dislocation filter, decreasing the density of threading dislocations in GaN. In this layer, the predominant type of defect is a mixed dislocation, which has a density of 8.6 x 109cm-2.
Deposition of a 25 nm-thick Al0.28Ga0.72N layer demonstrated the device quality of GaN. When removed from the substrate, characterization of the 2 cm square sample revealed a two-dimensional electron gas mobility of 1,100 cm2V-1s-1and a sheet carrier density of 1 x 1013 cm-2at room temperature.
The engineers from NTT have also fabricated an LED with a ten period multiple-quantum well and compared its performance to a conventional device, grown on a typical low-temperature AlN buffer layer. They found that electroluminescence intensities of the transferred LED at currents ranging from 10 mA to 50 mA were comparable to, or higher than, those produced by the control. The reason: Reflection from the backside contact indium.
Spectral width of the electroluminescence produced by both types of LEDs is similar, indicating that the active region maintains its quality during the transfer process.
A battery-powered LED prototype has also been fabricated by the team. A 2 mm square-release LED that is 3.4 µm-thick has been sandwiched between two commercially available laminate films featuring T-shaped Pd/Au electrodes. Application of indium tin oxide contact layers and a thermally activated sealing process creates a violet-blue emitting, flexible source.
Today’s targets for the team include an increase in the area of detachable devices, followed by improvements in the performance of conventional devices, such as LEDs and transistors, that can be transferred onto other materials.
Y. Kobayashi et al. Nature 484 223 (2012)
