Solid-State Electronics 47 (2003) 1081–1087
www.elsevier.com/locate/sse
GaN films annealed under high pressure
Francis Kelly a, Robert Chodelka a, Rajiv K. Singh a,*, S.J. Pearton a,
Mark Overberg a, James Fitz-Gerald b
b
a
Department of Materials Science and Engineering, University of Florida, Gainesville, FL, USA
Department of Materials Science and Engineering, University of Virginia, Charlottesville, VA, USA
Received 4 October 2002; received in revised form 29 November 2002; accepted 2 December 2002
Abstract
We present in this manuscript the results of a comparison study between MOCVD-grown GaN films annealed in a
novel high-pressure annealing furnace and those annealed in a conventional tube furnace under flowing argon gas at
similar temperatures for similar times. Atomic force microscopy, scanning electron microscopy, X-ray diffraction,
photoluminescence and cathodoluminescence were used to investigate the affects of the high-pressure anneals in
comparison to those which were annealed with the more conventional method. We demonstrate that high pressure
annealing has successfully prevented the degradation of GaN films at high temperatures.
Ó 2003 Published by Elsevier Science Ltd.
1. Introduction
The III-nitrides in general are presently the focus of a
considerable amount of effort in the research community
[1,2]. GaN in particular is an important material in the
fabrication of light-emitting and high-power/high-frequency devices such as LEDs [3–5], thyristors [6–8], laser
diodes [9–11], HBTs [12–14], and MESFETs [15–17].
GaN has a challenging set of materials properties that
often complicate annealing of films, not the least of
which is its propensity to dissociate into gallium metal
and nitrogen-containing gasses at temperatures in excess
of 1200 °C. Even at lower temperatures, there is a tendency for preferential loss of nitrogen from the surface.
Nitrogen vacancies are generally considered to be shallow donors and therefore they reduce the conductivity of
initially p-type GaN and may even lead to surface typeconversion. If films could be annealed at temperatures in
excess of 1200 °C without nitrogen loss, and without the
use of additional layers atop the film surface to prevent
dissociation, the activation of implanted layers for GaN-
*
Corresponding author.
E-mail address: rsing@mse.ufl.edu (R.K. Singh).
based devices would be improved significantly. In this
manuscript, we examine on the effectiveness of high
pressure annealing, and compare this with standard
annealing methods. It is well known [18–22] that a high
overpressure of nitrogen will be necessary to suppress
the dissociation process normally encountered when
annealing GaN, and in this case we have used a highpressure annealing furnace which can operate at sustained internal pressures of over 50 kbar for extended
time periods. By operating the annealing system at
pressures around 45 kbar, we are able to take advantage
of the increased thermodynamic stability of GaN under
high ambient pressure and anneal GaN films at temperatures which would otherwise result in complete loss
of the film. At very high ambient pressures (in excess of
20 kbar), it can be shown that the phase stability region
of GaN deviates from the values expected using the ideal
gas law as a model due to non-ideal thermodynamic
behavior of nitrogen gas at very high pressures, as
manifested in the activity of nitrogen derived by Karpinski et al. [18,19]. Thus the reaction: 2GaN(s) ! 2Ga
(1) þ N2 (g) will remain thermodynamically unfavored
and the GaN solid phase will remain stable at high
temperatures, as the Gibbs free energy required to form
diatomic nitrogen gas under such high pressures is
increased.
0038-1101/03/$ - see front matter Ó 2003 Published by Elsevier Science Ltd.
doi:10.1016/S0038-1101(02)00496-3
1082
F. Kelly et al. / Solid-State Electronics 47 (2003) 1081–1087
2. Experimental
The furnace used for the high-pressure anneals was
originally designed in Novosibirsk, Siberia, Russia for
the purposes of growing gem-quality synthetic diamonds. The machine, shown schematically in Fig. 1, consists of a large four-part steel outer shell inside of which
is a pair of hemispherical gaskets. Hydraulic fluid is
pumped into the shell and used to pressurize the system.
Inside of the gaskets is a set of eight octant-shaped metal
dies, which, when assembled, forms a spherical pressuretransducer. Inside the spherical pressure transducer is a
octahedral cavity into which goes a second set of smaller
metal dies which in turn form a second pressure transducing assembly which has a small rectangular cavity at
its center, and therefore at the center of the spherical
furnace system. This small rectangular cavity houses a
core cell that is composed of as many as 20 individual
parts, including the sample to be annealed. The core cell
is assembled from ceramic, metal, and graphite components in the laboratory, loaded into the pressure
transducer for annealing and is destroyed upon recovery
of the samples.
The core cells were constructed and then placed in a
Forma Scientific vacuum oven at 140 °C for 24 h to
eliminate any moisture prior to being loaded in the highpressure system. The pressure system was then purged
with dry nitrogen (99.99 pure) for 120 s to force out as
much ambient air as possible before pressurization.
After the high-pressure system reached the desired
pressure, in this case 45 2 kbar, an AC current was
then applied to the graphite heating element inside the
cell. The slope of this line was calibrated and the power
set to the temperatures of interest in each case. The
power was ramped manually by adjusting the heater
voltage and the average ramp rate was kept around 10
°C/min.
To compare the effects of high pressure annealing
with that of standard furnace anneals, the furnaceannealed samples were cleaved from the wafer, cleaned in
methanol, allowed to dry in the laboratory, and loaded
into a digitally controlled Lindberg tube furnace with an
alumina sample tube. The air in the tube was first flushed out with argon and then the anneals were carried
out in the hot zone of the furnace under a 10 ml/min
continuous flow of argon for the same times and temperature as the high-pressure samples. The temperature
was ramped up to the desired temperature at a ramp
rate of 10 °C/min, controlled by the digital furnace
controller.
Each sample was mounted into a Phillips XÕPert highresolution X-ray diffractometer and X-ray diffraction
scans were carried out to detect any phases that may have
formed or disappeared as a result of the annealing process. All but one of the scans revealed a peak at 34.5°
which is associated to the well-known (0 0 0 2) reflection
of GaN. The sample annealed in flowing argon at 1300
°C, shown in showed no such GaN (0 0 0 2) peak. Auger
electron spectroscopy data performed on this sample
revealed only aluminum and oxygen, whereas similar
characterization of the other samples revealed gallium
and nitrogen.
Atomic force microscopy (AFM) was performed on
these samples using a Nanoscope III system. Three
representative areas were selected on each sample using
the deviceÕs optical system and those areas were probed
in AFM surface tapping mode. The Nanoscope III
software then analyzed the resulting topograph of the
Fig. 1. Cross-section view schematic of the high pressure transducer system used for high pressure annealing.
F. Kelly et al. / Solid-State Electronics 47 (2003) 1081–1087
surface and calculated a mean square roughness value
(RMS) for each probed area.
3. Results and discussion
Fig. 2 shows two AFM scans from two samples annealed at 1100 °C, one under high pressure, and the other
in the conventional tube furnace. The surface morphology from the former is clearly superior and indicates that
the high pressure approach is capable of suppressing loss
of nitrogen from the surface at this temperature. Average
RMS values for each sample were plotted in Fig. 3 as a
function of annealing temperature. Note the severe increase in surface roughness for anneals above 900 °C in
the tube furnace, compared to the much smaller changes
1083
recorded for the high pressure annealing. The AFM
comparison graph in Fig. 3 shows no data point for the
furnace-annealed sample at 1300 °C, as it was earlier
determined that that sample had no film remaining on it
at all, and, as such, the roughness measured would have
been from the bare substrate material alone. It can be
clearly seen that, for the 1 1 lm2 square areas probed
by the AFM, the samples annealed under high pressure
were generally of higher surface quality than those annealed in the conventional furnace. The high-pressureannealed samples all had RMS values of less than 10 nm,
and the samples annealed at higher than 1000 °C in the
furnace were more than 20 as rough as the high-pressure-annealed samples.
The samples were then scanned in the Phillips XÕPert
high resolution diffractometer and all but one of the
Fig. 2. A pair of AFM scans taken from (a) the sample annealed at 1100 °C under high pressure and (b) the sample annealed at 1100
°C in flowing argon. Note the vertical scales are different for the two scans.
1084
F. Kelly et al. / Solid-State Electronics 47 (2003) 1081–1087
Fig. 3. Comparison graph of film roughness versus anneal temperature. The dashed line represents the samples annealed in flowing
argon, the solid line represents those annealed under high pressure.
Fig. 4. Rocking curves for the high-pressure annealed GaN films.
scans revealed a peak at 34.5° which is associated to
the well-known (0 0 0 2) reflection of GaN. The sample
annealed in flowing argon at 1300 °C showed no such
GaN (0 0 0 2) peak. Auger electron spectroscopy data
performed on this sample revealed only aluminum and
oxygen, whereas similar characterization of the other
samples revealed gallium and nitrogen. High resolution
x 2h X-ray scans were also recorded for each sample to investigate any changes in lattice parameter of
the GaN films upon annealing. The resulting curves
taken from the high-pressure annealed samples are
shown superimposed in one plot in Fig. 4, and the
corresponding curves for the samples annealed in
flowing argon in Fig. 5. These data were then analyzed
in the PhillipÕs XÕPert software and a full width at half
maximum (FWHM) was extracted for each rocking
curve. These FWHM values are compiled and plotted
versus temperature in Fig. 6. As shown by the rocking
curve data, the furnace-annealed samples seem to be of
higher quality than the high-pressure-annealed samples
at low temperatures, but the slopes of the two trend
lines indicate that the FWHM at temperatures above
F. Kelly et al. / Solid-State Electronics 47 (2003) 1081–1087
1085
Fig. 5. Rocking curves taken from the samples annealed in flowing argon.
Fig. 6. Comparison plot of FWHM versus temperature for the two annealing processes. The solid line represents the high-pressure
process, and the dashed line represents the flowing argon process.
1300 °C will tend to be lower for the high-pressure
samples.
Cathodoluminescence data taken from the sample
annealed under high-pressure at 1300 °C is shown in
Fig. 7. The band edge emission at 3.4 eV can clearly be
seen, as can a yellow emission feature near 2.2 eV.
Given the proximity of the sample to the graphite
heating element in the high-pressure system, this yellow
peak is most likely due to the diffusion of carbon impurities into the film during the annealing process, which
would, to some extent, account for the increase in
FWHM of the high-pressure sample high resolution
x 2h curves as well. Substitutional carbon defects
on the nitrogen sublattice would not likely lead to a
large increase in FWHM of the x 2h X-ray scans, as
the atomic radii of nitrogen and carbon are fairly close
to each other. Substitutional carbon on the gallium sublattice, however, should lead to increased
FWHM of those scans, since the gallium atom is much
larger than the carbon atom. This is supported by the
1086
F. Kelly et al. / Solid-State Electronics 47 (2003) 1081–1087
Fig. 7. Cathodoluminescence spectrum of the sample annealed under high pressure at 1300 °C.
Fig. 8. Photoluminescence spectrum taken at 50 K of the sample annealed under high pressure at 1100 °C.
photoluminescence spectra, as there are prominent band
edge and yellow emission signals from the sample annealed at 1100 °C under high pressure, as can be seen in
Fig. 8. The furnace annealed samples, which were not
annealed is such close proximity to any carbon source
do not show the increased FWHM in their X-ray scans.
4. Conclusions
While the data presented herein clearly show that
the high pressure process has been successful in suppressing the tendency of GaN to dissociate upon hightemperature annealing, other factors which degrade the
GaN films in other ways are also observed to occur.
Carbon from the heating element is most likely the
cause of the increased yellow emission in the films and
also the increase in the FWHM of the X-ray rocking
curves.
Acknowledgements
We would like to thank Mr. Eric Lambers of the
Major Analytical Instrumentation Center (M.A.I.C.) at
the University of Florida performing Auger spectroscopy of the samples, and Dr. Valentin Craciun for all of
his help with the Phillips XÕPert system. We would also
like to thank Dr. Reza Abbaschian of the University of
Florida as well as Dr. Alexander Novikov and Dr.
Nikolay Patrin of the Gemesis Corporation for their
expertise with the high pressure system. The work at UF
is partially supported by NSF DMR 01001438.
References
[1] Pearton SJ, Ren F, Zhang AP, Lee KP. Mater Sci Eng R
2000;30:55–212.
[2] Pearton SJ, Zolper JC, Shul RJ, Ren F. Appl Phys Rev
1999;86:1–78.
F. Kelly et al. / Solid-State Electronics 47 (2003) 1081–1087
[3] Szweda R. III–VÕs Rev 1997;10:20–5.
[4] Harle V, Hahn B, Lugauer H-J, Bader S, Bruderl G, Baur
J, et al. Phys Stat Sol 2000;180:5–13.
[5] Oda O, Inoue T, Seki Y, Kainosho K, Yaegashi S,
Wakahara A, et al. Phys Stat Sol 2000;180:51–8.
[6] Bandic ZZ, Bridger PM, Piquette EC, McGill TC. SolidState Electron 2000;44:221–8.
[7] Bandic ZZ, Piquette EC, Bridger PM, Beach RA, Kuetch
TF, McGill TC. Solid-State Electron 1998;42:2289–94.
[8] Cao XA, Pearton SJ, Donovan SM, Abernathy CR, Ren F,
Zolper JC, et al. Mater Sci Eng B 1999;59:362–5.
[9] Bour DP, Kneissl M, Van De Walle GG, Northrup J,
Romano LT, Teepe M, et al. Phys Stat Sol 2000;180:
139.
[10] Bader S, Hahn B, Lugauer H-J, Lell A, Weimar A, Bruderl
G, et al. Phys Stat Sol 2000;180:177.
[11] Ambe C, Takeuchi T, Katoh H, Isomura K, Satoh T,
Mizumoto R, et al. Mater Sci Eng B 1999;59:382–5.
[12] Asbeck PM, Yu ET, Lau SS, Sun W, Dang X, Shi C. SolidState Electron 2000;44:211–9.
1087
[13] Ren F, Han J, Hickman R, Van Hove JM, Chow PP,
Klaassen JJ, et al. Solid-State Electron 2000;44:239–44.
[14] McCarthy LS, Smorchkova IP, Xing H, Kozodoy P, Fini
P, Limb J, et al. IEEE Trans Electron Dev 2001;48:543–51.
[15] Lee C, Lu W, Piner E, Adesida I. Solid-State Electron
2002;46:743–6.
[16] Boudart B, Gaquiere C, Guhel Y, de Jaeger JC, Poisson
MA. Elecron Lett 2001;37:1–2.
[17] Polyakov VM, Schwierz F. IEEE Trans Electron Dev
2001;48:512–6.
[18] Van Vechten JA. Phys Rev B 1973;7:1479–507.
[19] Porowski S. Mater Sci Eng B 1997;44:407–13.
[20] Cao XA, LaRoche JR, Ren F, Pearton SJ, Lothian JR,
Singh RK, Wilson RG, Guo HJ, Pennycook SJ. SolidState Electronics 1999;43:1235–8.
[21] Hong J, Lee JW, Vartuli CB, MacKensie JD, Donovan
SM, Abernathy CR, Crockett RV, Pearton SJ, Zolper JC,
Ren F. Solid State Electronics 1997;41:681–94.
[22] Karpinski J, Jun J, Porowski S. J Crystal Growth 1984;
66:1–10.