A compliant substrate could allow high-performance optoelectronic devices to be grown on silicon substrates or indium phosphide-based devices to be grown on gallium arsenide (GaAs). "There are many III-V-material-based devices that we can’t make right now because of a lack of a lattice-matched substrate," explains Felix Ejeckam of Cornell University (Ithaca, NY). Ejeckam and others in Cornell professor Yu-Hwa Lo’s group may have solved that problem by using twist bonding to create a compliant universal substrate—a material on which high-quality films could be grown despite large lattice mismatches and differing crystal structures.
Researchers from Cornell University and Sandia National Laboratories (Albuquerque, NM) first reported growing high-quality materials on mismatched substrates at the IEEE LEOS Annual Meeting last November.1 A 10-µm GaAs film was bonded to a GaAs substrate with a deliberate misalignment between crystal axes. This created a dense two-dimensional array of screw dislocations at the boundary. Scattered screw dislocations have a long-range stress field associated with each one and can cause threading dislocations to form in material grown above the film. In the dense array of dislocations, however, the stress fields largely cancel out, leaving a uniform array of stretched bonds at the boundary between the substrate and film. "On this twisted film one can grow highly lattice-mismatched semiconductors," says Ejeckam, "because the twisted film is resting on extremely stretched, relatively weak bonds."
The researchers have grown several materials that usually exhibit threading dislocations on GaAs substrates. As a proof of principle, the researchers grew a 625-nm-thick layer of indium antimonide (InSb) on a 3-nm-thick GaAs film twist-bonded to a GaAs substrate at 45°. The researchers also grew a 300-nm In0.35Ga0.65As film grown on a 10-nm-thick GaAs film twist-bonded to the GaAs substrate at 9° misalignment. Gallium antimonide films also have been demonstrated. All three materials appear free of threading dislocations in transmission electron microscopy studies.
Device applications
John Bowers, director of the Multidisciplinary Optical Switching Technology Center at University of California, Santa Barbara (UCSB), says, "Twist bonding is potentially very important. It could allow high-performance devices to be made on silicon substrates or InSb devices on GaAs substrates." His group has begun work on twist bonding, which could allow vertical-cavity surface-emitting lasers and photodetectors to be integrated on silicon substrates.
Blue-emitting laser development could also benefit. Steve DenBaars, also of UCSB, explains that blue LEDs and (short-lived) lasers can be grown in gallium nitride; however, that material lacks a suitable substrate. Gallium nitride is currently grown on sapphire: although both are hexagonal crystals, the materials have a huge 18% lattice mismatch, which results in about 108 dislocations per cubic centimeter. Although GaN researchers are unsure what mechanism limits the lifetime of the blue-output lasers, DenBaars says, "Dislocations seem to affect lifetimes." A compliant substrate would dramatically reduce the number of dislocations.
Proving out
"This looks like a big breakthrough," says DenBaars. The work is still new, however. Scaling up to large areas could introduce problems, or the twisted bond might not prove strong and reliable at the high temperatures required to grow and process materials on the substrate. Mechanical tests at Cornell have shown no strength difference between aligned and twist-bonded films.
Bowers says, "It has not yet been proven, in the sense that no one has made a laser, photodetector, modulator, or transistor on a twist-bonded substrate and shown that it works as well as it would have on its native substrate." Ejeckam agrees, saying, "The ultimate test of the compliant substrate will be if one can make devices on them." Ejeckam has grown 10 InGaAs quantum wells on the GaAs compliant substrates that, according to x-ray studies, are free of defects.
Since November, the Cornell/Sandia group has produced other substrates at higher twist angles and even larger lattice mismatches. Lo says, "We`re currently trying to implement it with silicon and gallium nitride." The researchers will report further work at the Materials Research Society meeting in March (San Francisco, CA) and the 9th Annual International Conference on Indium Phosphide and Related Materials (Hyannis, MA) in May.
REFERENCE
1. F. E. Ejeckam et al., postdeadline paper PDP2.5 (1996).