Big Blue ambition
IBM reseachers have grown GaN films on graphene to produce blue LEDs; wafer-scale manufacturing is next. Compound Semiconductor reports.
A blue LED breakthrough sees IBM getting very close to cheap GaN device manufacture
Earlier this year, IBM researcher, Jeehwan Kim, and colleagues unveiled a blue LED comprising GaN layers on epitaxial graphene. The GaN films had a low defect density and surface roughness on par with layers grown on conventional SiC and sapphire substrates while the performance of the LED matched that of a conventional GaN-on-sapphire device.
Kim's latest results form part of an IBM drive to produce wafer-scale, single crystalline sheets of graphene that can then be used as cheap, reusable substrates on which to grow single-crystalline III-nitride layers, and manufacture cheap GaN-based devices.
The plan is to lift off the semiconductor layers from a wafer-scale graphene substrate, and then transfer these layers to an arbitrary substrate. The remaining graphene substrate will be repeatedly re-used for further III-nitride film growth, giving way to extremely cheap GaN device manufacture. Kim and colleagues are getting close.
Materials growth
The main barrier to growing single crystalline films on graphene is the material's lack of so-called dangling bonds. As Kim explains: "The material has no broken bonds on its surface, there are no attraction forces and so it is very difficult to nucleate materials on top of graphene."
To overcome this, Kim and colleagues first grew graphene onto a SiC wafer containing nanometre-scale steps to act as nucleation sites for subsequent GaN epitaxy. For many researchers in the field, a next step has been to deposit an AlN buffer layer to a SiC substrate, followed by GaN growth. But Kim wanted to pursue a different route.
"If we used an AlN buffer layer on the graphene, growth would be easier but I didn't use it on purpose," he says. "To fabricate a vertical functional device, it is desirable to deposit layers onto a highly conductive layer, and AlN is not so ideal in this case."
Instead, Kim, and colleagues turned to van der Waals epitaxy. This alternative epitaxy mechanism circumvents the lattice matching requirements of conventional epitaxial growth as the substrate and overlying epitaxial layer only form weak van der Waals bonds, rather than the covalent bonds of conventional epitaxy.
And so, using carefully controlled growth kinetics in a MOCVD reactor, Kim and colleagues grew single-crystalline, 1 cm², GaN films onto the epitaxial graphene. As Kim highlights: "The GaN crystalline quality was comparable to that typically obtained with conventional AlN-buffer-assisted GaN epitaxy on SiC or sapphire substrates."
"This is totally different from conventional epitaxy growth, and you can substantially reduce dislocation-related defects in principle," he adds.
The next step was to release the GaN film and demonstrate exactly how re-usable the graphene-on-SiC substrate could be. To this end, the researchers deposited a layer of nickel onto the GaN film, to create enough strain within the semiconductor film to overcome the relatively weak GaN-graphene bonds. A thermal release tape was then added, with the GaN film being completely removed and transferred to a second substrate.
Schematic for growing and transferring single-crystalline thin films onto and from epitaxial graphene. [IBM]
Microscopy and spectroscopy analyses revealed the process to be a success. Uniform coverage of graphene was observed across the released interface, without any trace of graphene residue on the GaN film.
Crucially, Kim and colleagues went on to grow an LED stack on a recycled graphene-SiC substrate that had been previously used three times. Light emission peaked at a wavelength typical for III-nitride LEDs.
Kim is confident that this process will now scale as necessary. "I have demonstrated the [growth/transfer] process four times, but conceptually believe this could be repeated infinitely," he says.
Similarly, wafer-scale demonstration shouldn't pose a problem. "Right now our substrate size is 1cm by 1cm, limited by the size of our lab-scale MOCVD reactor," he adds. "But in principle, wafer-scale GaN release is possible. The graphene substrate is single crystalline, so if you perform the process at a manufacturing scale, it is going to work."
So where next for the IBM researchers? Working closely with chief collaborator, Can Bayram, Kim and colleagues have transferred GaN layers onto flexible plastic substrates and silicon wafers, a crucial step to low-cost wafer-scale fabrication of GaN devices. Kim is currently focused on fabricating more LED structures but has additional plans.
"I developed this technique to get an idea of how to grow any type of single crystalline material on graphene," he says. "I have demonstrated GaN and am now going to try other materials on graphene."
