Overview

The First International Symposium on Gallium Nitride and Related Materials was held in conjunction with the Fall MRS Meeting Nov. 27 - Dec. 1, 1995. The Symposium AAA was the largest of the entire MRS Meeting having had a first day attendance of over 400. The work presented at the conference shows steady improvement in the material quality, device performance and fundamental understanding.

GaN LEDs

Hewlett Packard

Dr. George Craford of HP led off the session with a general discussion of LED applications. He predicted that the GaN LED market would hit $2B by 2005.

Nichia Chemical

Shuji Nakamura of Nichia Chemical presented an update of their industry leading InGaN LED production. By incorporating increasing amounts of In into a nominally pseudomorphic QW, GaN-based LEDs were extended all the way out to 600 nm. Included in the presentation was a demonstrator comprised of roughly 20 GaN-based LEDs emitting from in a continuous series violet to the orange-red. By replacing a thick co-doped InGaN active region with a pseudomorphic InGaN QW, Nichia can now market blue (2.5 Cd brightness, 450 nm center wavelength, 5 mW output power, 9.1% power efficiency) and green (12 Cd, 520 nm, 3 mW, 6.3%) LEDs. Combined with Candela class red AlGaAs or AlInGaP LEDs, blue and green GaN LEDs enable large area full color LED displays for both indoor and outdoor applications. A photograph of a 10 x 15 meter display installed in S. Korea was shown as an example.

Cree Research

Several papers from Cree Research discussed their GaN on SiC LED technology. Cree is still struggling to develop a suitable conductive buffer layer technology which would enable them to fully capitalize on their conductive SiC approach. At present, conductive buffer layers cause cracks while buffer layers optimized to avoid cracking present a large barrier to carrier injection from the SiC substrate. A fully vertical structure was realized by using a shorting ring to connect the lower cladding layer to the SiC substrate. In addition to requiring a mesa etch process to create sidewalls for the evaporated metal ring, the shorting ring reduces light output by 40%. This may also contribute to a reduced yield, which was reported to be 50% in the Q&A session. Cree Research also uses a GaN:Mg active region which also limits efficiency. Nevertheless, their devices are clearly candela class blue LEDs emitting 1.7 mW at 430 nm center wavelength with a 3% power efficiency at 20 mA forward bias. Degradation studies at 30 mA showed 80% light output after 1000 hours and 75% after 2500 hours.

Toyoda-Gosei

The final commercial LED technology presented was from Toyoda-Gosei. Their p-n junction devices radiate 2.5 Cd (450 nm, 3.6 mW, 5.1%) via a donor-acceptor transition which was not revealed.

Stimulated Emission in GaN

Introduction

Numerous groups reported stimulated emission in all types of heterostructures and geometries. It was clear that many different groups were closing in on laser operation. Therefore, the announcement by Nichia Chemical of a GaN laser diode shortly after the MRS Meeting adjourned was not a major surprise to conference attendees.

Meijo University

Prof. Amano represented the Nagoya group in Boston. They reported the low optical pumping threshold values for room temperature stimulated emission of 1 MW/cm2 for MOCVD (MBE regrown) structures on sapphire substrates and 2 MW/cm2 for SiC substrates. Cleaved GaN/SiC facets were shown by AFM to have a roughness of 2 monolayers. P-n diodes grown on SiC substrates using 20% Al AlGaN had a 3.5 V turn on voltage.

Lincoln Laboratories

Richard Molnar reported the results obtained on Lincoln Lab ECR-MBE on HVPE buffer layer on sapphire structures. HVPE was shown to still be capable of high growth rates (70 um/hr) and excellent crystal quality, although not necessarily 2D growth. An advantage is that if the HVPE GaN buffer layer is grown thick enough, then proper alignment can cause the sapphire substrate to break on a GaN cleavage plane. Stimulated emission was observed at a threshold of 0.5 MW/cm2 at room temperature.

Prospects for a GaN Laser

NCSU/Schetzina

Prof. Schetzina pointed out that unacceptability of today’s p-type ohmic contacts for GaN laser structures. He extrapolated state of the art contact resistivities to the small areas and high currents needed for laser operation to show that a stripe laser would require 40 V bias at threshold current. P-type contacts do not limit today’s LEDs because the contact areas are much larger, i.e. current densities much lower.

Electronics

APA Optics/Khan

Dr. Asif Khan described 150 x 0.25 μm HFETs with a 90 mS/mm transconductance, 3-4 V pinch-off, good saturation, and ft and fmax of 36 and 71 GHz, respectively. The gm for 1 μm gate length was 125 mS/mm. They also reported preliminary power devices with PAE of 20 - 30% at 10 GHz and 10V.

NRL/Doverspike

Dr. Kathy Doverspike presented the FET work at NRL. Their approach is to grow a highly insulating undoped GaN buffer layer on which the conductive channel is deposited. AlGaN/GaN FETs of 1 um channel lengths had transconductances of 23, 10 and 8 mS/mm at room temperature, 400 C and 500 C respectively. At 500 C, the device no longer pinched-off.

Photodetectors

Honeywell/Goldenberg

Lateral MSM n-GaN (or equivalently, ungated FET) structures were described. An optimized sheet resistance of 10^5 Ohm/square was reported as the figure of merit which is obtained by trading off channel thickness and background carrier concentration.

Alloys

Given the relative quality that has been achieved in GaN, investigators are now turning their attention to the ternaries, particularly those containing In. InGaN is of particular interest since it serves as the active layer in blue emitting devices. Several studies noted the difficulty in growing good quality material with moderate In concentrations.

Singh and Moustakas (Boston University) for example reported evidence of phase separation for alloys with XIn = 0.37 grown at 750°ree;C. Boutros and Bedair (NCSU) also examined InGaN as well as AlGaInN, both grown at relatively high growth temperatures. The InGaN in this study exhibited a FWHM of ~450 arcsec. These authors proposed a model involving the formation of In droplets at the growth surface to explain their results and those of the previously mentioned speakers. This model may begin to explain why the material quality of the In-containing alloys lags substantially behind that obtained in AlN and GaN. It is not yet known whether this separation is due to segregation or immiscibility.

Doping

Introduction

In addition to further development of Mg doping in GaN, two other p-type dopant elements were shown at the symposium to offer promise. Several groups are studying ion implantation for GaN doping.

Howard Univ/Spencer.

Perhaps the most surprising result of the entire week was the report from Prof. Mike Spencer of conductive AlN when large C concentrations are incorporated. Undoped AlN had a normal resistivity of 108 Ohm cm. When 4 atomic % of C was incorporated, the resistivity dropped to 1 Ohm cm. At 8 a%, the resistivity is only 0.1 Ohm cm. An activation energy of the p-type carriers of 33 meV was measured. Maximum hole concentrations were in the mid 1018 cm-3 range, which is an extremely low activation given the high C incorporation and low activation energy. Efforts were made to identify filamentary conduction paths, but to date, the material seems to be AlN, even at such high C incorporation. A p-AlN:C/n-SiC LED emitted a characteristic SiC spectrum which was further proof that the carriers in AlN was p-type. Conductive AlN may find applications as a buffer layer technology for GaN/SiC LEDs. No good understanding of this phenomenon is yet available.

Sandia/Zolper

Dr. J. C. Zolper presented data on the electrical activation of implanted impurities in GaN. The rapid thermal annealing activation threshold for implanted Si was (1050 C, 10s), with increased activation reported for (1100 C, 10 s). Mg implantation was investigated with the best results obtained when P was co-implanted. Ca was also identified as a p-type dopant. Hole concentrations up to 1-2 x 1017 cm-3 were achieved both with and without P co-implantation.

IBM/Strite

A group at IBM Zurich reported attempts to optically activate ion implanted Zn. Their best results were obtained by using SiNx caps for 1100 C, 10s RTA conditions, but RT PL was still a factor of 40 less intense than epitaxial GaN:Zn.

Crystal Growth

NRL/Doverspike

Dr. Kathy Doverspike described the MOCVD growth of GaN/sapphire for FETs. The dislocation density was shown to be growth condition dependent, and could be optimized at 10^9 cm-2. Strained layer superlattices had negligible effect on threading dislocation propagation. Further optimization of the growth conditions permitted a highly resistive (10^10 Ohm cm) undoped GaN buffer layer technology to be realized which is highly suitable for FETs.

Substrates

Introduction

Many new oxide substrates were discussed, including various aluminates, gallates and spinels, but the most exciting substrate was bulk GaN single crystals grown under high pressure at Unipress.

Unipress/Suski

Dr. Tadeusz Suski represented the Unipress (formerly Polish Academy of Science) group in Boston. The Unipress group has been producing bulk GaN single crystals for several years under conditions ranging from 10^3 - 10^4 bar at 1300 - 1500 C. These conditions are necessary in order to produce significant N solubility in Ga. Typical bulk material is highly conductive (n=10^19 cm-3, mobility = 120 cm2/Vs) and defect free as determined by TEM. Bulk GaN substrates have enabled some of the physical parameters of GaN such as optical transition energies, lattice constants etc. to be determined with very high precision due to the absence of heteroepitaxial strain. Measurements under high pressure revealed that the N vacancy is the donor responsible for the n-type conductivity. The yellow emission was also shown to be shallow donor related, suggesting that it stems from N vacancies also. A 40% Al mole fraction was sufficient to produce insulating AlGaN.

Processing

Introduction

Several groups reported reactive ion etching of GaN. Rough sidewalls were observed to varying degrees by all authors. It was claimed that the roughening is produced by etching of the SiNx mask, and that RIE of GaN with improved masks may produce the smooth sidewalls critical to avoid diffraction loss in lasers.

Florida/Pearton

Prof. Steve Pearton presented RIE etch results on GaN and AlN. Nitrides have basically the same chemistry as arsenides, but etch more slowly due to their stronger bond energies. GaN etched roughly 5x slower than GaAs. No thermal activation was observed up to 170 C. The GaN etched stoichiometrically in a Cl2/CH4/H2/Ar mixture. Higher ion energies do contribute to faster etch rates, but ion damage becomes a problem. Sidewall roughness on a 100 nm length scale was attributed to etching of the SiNx mask.

Univ. New Mexico

Mr. J. Brown of UNM presented a recipe for n-type ohmic contacts to GaN. A base dip is recommended, followed by a pre-anneal of 900 C, 30 s, perhaps to liberate H. Ti/Al contacts are preferred because the Ti prevents balling of the Al metal during post-annealing. A Cr cap was also useful for this purpose.

NCSU/Davis

Ms. Laura Smith described Al contacts to GaN. They were stable up to 450 C and produced resistances of 10-4 Ohm cm2. At 550 C, the contact degrades by 1-2 orders of magnitude due to AlN formation. TiN contacts to GaN were much more resistant to annealing and produced 1.5 x 10-5 Ohm cm2 contact resistances. Ti alone does not work well. P-type Au/GaN:Mg contacts suffered from poor adhesion and reacted at the interface when annealed.

Cree/Dmitriev

Dr. Vladimir Dmitriev used Pd for p-type contacts to GaN.

Future Meetings

At the close of the Symposium in Boston it was announced that the second meeting in this series will be held in Japan in Oct. or Nov. 1997. These meetings will occur biannually on a rotational basis between the United States, Japan and Europe.