Photoluminescence investigation of GaN films grown by ... chemical vapor deposition

Photoluminescence investigation of  GaN  films  grown  by ... chemical  vapor  deposition
of GaN films grown by metalorganic
chemical vapor deposition on (100) GaAs
C. H. Hong, D. Pavlidis, S. W. Brown,a) and S. C. Randa)
Solid State Electronics Laboratory, Department of Electronic Engineering and Computational Science,
The University of Michigan, Ann Arbor; Michigan 48109
-(Received 8 August 1994; accepted for publication 3 November 1994)
GaN films were grown on (100) GaAs substrates by metalorganic chemical vapor deposition and
were found to be of (200) cubic or (111) cubit/(0002) hexagonal phase. Their photoluminescence
characteristics remained invariant with material phase. We report assignment of band-edge
photoluminescence near 3!36 eV and 3.15-3.31 eV in apparently cubic GaN to intrinsic/bound
excitons and phonon-assisted, donor-acceptor pair recombination respectively, on the basis of
observed temperature and intensity dependences. A free exciton energy of 3.375 eV is deduced at
6.5 K. 0 1995 American Institute of Physics.
GaN is one of a family of wide band gap nitride III-V
semiconductors being actively investigated for applications
in the area of opto-electronics. Nitrides like AlN and GaN
exhibit direct gap emission and their ternary alloys are good
prospects for ultraviolet light emitting diodes and lasers,
electro-optic, piezo-electric, and acousto-optic modulators
and for negative electron affinity devices. In the past, numerous investigations have concentrated on samples with wurtzitic crystal structure, but more recently samples with the
zincblende structure have been prepared, principally by molecular beam epitaxy (MBE) and metalorganic chemical vapor deposition (MOCVD). Interest in cubic phase material
has been motivated by a desire to explore its predicted superiority for certain device applications and to integrate GaN
devices on readily available, high quality III-V semiconductor substrates which minimize lattice mismatch with the nitride crystal structure. With these considerations in mind, we
have studied the growth of GaN and reported the presence of
cubic (c-GaN) epitaxial films on both (100) (Ref. 1) and
(111) (Ref. 2) GaAs. This paper reports the optical properties
of GaN epitaxial films grown on (100) GaAs substrates at
temperatures below 55O”C, and our detailed studies of their
optical properties in order to establish the origin of the main
emission lines.
The hexagonal phase of GaN grows well on sapphire,
and both its structure and optical properties have been studied extensively in the past;3-7 However, the insulating character of sapphire substrates has hampered progress toward
large scale integration. Growth of cubic phase GaN films on
silicon,’ silicon carbide,g magnesium oxide,” and gallium
arsenide has been explored, with experiments on GaAs substrates having been reported only recently using MBE and
MOCVD techniques. ‘l-l3 While structural characteristics of
the material were carefully addressed in these earlier investigations, only limited aspects of luminescence properties of
c-GaN were reported. Strite’4 observed several cathodoluminescence emission peaks from c-GaN on GaAs in the vicin“)Division of Applied Physics, 1049 Randall Laboratory, The University of
Michigan, Ann Arbor, MI 48109-1120.’
J. Appl. Phys. 77 (4), 15 February 1995
ity of 3.2 eV, as well as a very intense, broad luminescence
band at midgap ascribed to defects. It was argued, chiefly by
analogy with work on hexagonal GaN,6’7 that several of the
emission lines in their spectra were due to electron-bound
hole and donor-acceptor (DA) pair recombination emission.
Okumura” also performed cathodoluminescence experiments, and observed a broad luminescence band centered at
325 nm, but there do not appear to be any prior reports of the
intrinsic photoluminescence spectrum. Here we report temperature and excitation intensity dependences of wellresolved features of the photoluminescence spectrum near
the band edge of (2OO)cubic or (11 l)cubic/(0002)hexagonal
phase GaN which permit assignments of peaks to free exciton emission, phonon-assisted exciton recombination, and
donor-acceptor (DA) pair processes.
GaN films were grown on (100) semi-insulating GaAs
substrates in a low pressure (60 Torr) MOCVD system containing using hydrogen (H2)carrier gas. Substrates were
chemically cleaned in a solution of H2S04:H202:H20
=5:1:1, mounted on a graphite susceptor and loaded into a
horizontal quartz reactor. Halogen lamps were used to heat
the substrate and temperature was monitored continuously
with a thermocouple. The substrates were first annealed in an
atmosphere of H, and ASH, at 650 “C for 5 min to remove
surface oxide. This step was followed by surface nitridation
in which substrates were exposed to 1 sl/min of ammonia
(NH,) for 10 min at a temperature of 600 “C and then lowered to the growth temperature of 530 “C. Trimethylgallium
(TMG) and NH, were supplied at a fixed V/III ratio of 3000.
The TMG flow rate was maintained at 8.9 ~mole/min for all
experiments reported here. Typical growth rates were on the
order of 0.6 pm/h, and total thicknesses of GaN films grown
in this way were typically 0.35 pm. Analysis of x-ray diffraction spectra and transmission electron microscopy measurements shows a transformation from (111) cubic or (0002)
hexagonal to (200) cubic phase GaN as the growth temperature is raised from 530 to 600 “C. Details of the x-ray characterization are presented elsewhere.r6 The hexagonal phase
is found to dominate as the temperature is further increased
0021-8979/95/77(4)/l 705/5/$6.00
Q 1995 American Institute of Physics
Downloaded 21 Mar 2002 to Redistribution subject to AIP license or copyright, see
0.6 W/cm2
( cm-’ )
FIG. 1. The room temperature Raman spectrum of GaN (1,=.514.5 nm).
Inset: The absorption spectrum of a free-standing GaN firm of 2 pm thickness.
_ 60K
70 K
3.26 3.28 3.30 3.32 3.34 3.36 3.38 3.40
Energy ( eV )
above 650 “C while the cubic phase dominated in the range
of 570-650 “C. Post-growth annealing was performed for 20
min in an ammonia atmosphere at 670 “C.
For photoluminescence observations, a frequencydoubled picosecond DCM dye laser was used to provide
quasi-continuous excitation at 325 nm. Doubling was performed outside the dye laser cavity using an angle-tuned
LiIO, crystal at room temperature. Average powers on the
order of 1 mW were easily obtained at a cavity-dumped,
pulse repetition rate of 38 MHz and light from this source
was directed to samples mounted in an open-cycle liquid
helium cryostat which permitted temperature control in the
range 6-295 K. Luminescence from the sample was then
collected in a back-scattered geometry and focused onto the
entrance slits of a 1 m spectrometer (ISA THR-1000). Photoluminescence (PL) signals were detected with a
C3 1034A-02 GaAs photomultiplier in a photon-counting
mode, and in this manner spectra were recorded as functions
of both temperature and excitation intensity. All photoluminescence spectra presented in this paper correspond to films
grown at 530 “C which were of (111) cubic or (0002) hexagonal phase. Characterization of (200) cubic samples grown
at higher temperatures revealed essentially the same photoluminescence characteristics with no shift in their position
and thus no obvious difference of band gap.’The only remarkable difference was the intensity of the spectra which
increased for samples grown at higher temperatures. Although not possible to fully confirm at this stage, the results
obtained in this work seem to correspond to cubic GaN films.
Several samples were also etched using 10 ml NHbOH in
500 ml of 30% H202 to yield free standing GaN films .on
which absorption measurements were made.
J. Appl. Phys., Vol. 77, No. 4, 15 February 1995
FIG. 2. High resolution photoluminescence spectra of near band-edge emission in GaN (1,=325 nm) vs temperature.
In Fig. 1, the Raman spectrum of a typical GaN film
recorded using 488 nm excitation light is shown, together
with the absorption spectrum (inset). The Raman spectrum
showed strong LO and TO components at 91.4 and 68.9
meV, respectively, as expected.“:” The GaAs substrate was
chemically etched and the remaining GaN films showed a
distinct light yellow-brown coloration to the eye, and invariably showed a broad absorption tail (see inset of Fig. 1)
extending from the ultraviolet into the visible spectral region
which made a precise determination of the fundamental edge
difficult. PL spectra recorded between 3.26 and 3.40 eV at
low intensity for different temperatures clearly showed
strong near-edge features. These are shown in Fig. 2, where
only the near gap spectral region is plotted because mid-gap
emission reported at longer wavelengths (-550 nm) in earlier investigations of c-GaN (Ref. 14) was too weak to be
Two salient groups of lines appeared in the PL spectra.
First, an intense, structured feature at 3.366 eV dominated
the spectrum at 6.5 K, and broadened and shifted to lower
energies as the temperature was increased. Its position
shifted in a manner similar to that ascribed to the band edge
in earlier research on c-GaN,t4 decreasing by 5 meV between 6.4 and 70 K. In spectra recorded above 40 K, the
intensity of this peak diminished and a small thermally aceHong et al.
Downloaded 21 Mar 2002 to Redistribution subject to AIP license or copyright, see
Energy ( eV )
Energy ( eV )
FIG. 3. Photoluminescence spectra vs temp&ture at high excitation intensity (P=38 W/cm 2 and 1,,=325 nm). Inset: Logarithmic plot of the 3.31 eV
emission line and satellite intensities at 77 K.
FIG. 4. Photoluminescence spectra of GaN vs excitation intensity, at a fixed
temperature of 6.5 K (1,,=325 mn). Magnification factors for the three
cnrves are Xl, X5, X40 (top to bottom).
vated peak positioned 10 meV higher in energy, which also
tracked the band edge, became evident.
A second group of lines appeared at energies below 3.32
eV. In this region, an intense feature accompanied by satellites was seen. To improve signal-to-noise ratio for the
weaker features, spectra were recorded at elevated excitation
intensities and are shown in Fig. 3. This resulted in a considerable loss of resolution. However traces recorded this
way (inset: Fig. 3) served to rather clearly establish the relationship between the electronic origin at 3.310 eV and its
three evenly spaced satellites at lower energies. At the lowest
temperature, the satellites had transition energies of 3.239,
3.164, and 3.095 eV, and as temperature was raised the dominant shift in their positions was again in close accord with
the shift in the absorption band edge. Based on the Raman
splittings given above, which are very similar to those of
wurtzitic GaN,18.the three low energy satellites can be assigned as phonon-assisted replicas of the recombination line
at 3.310 eV corresponding to the emission of one, two, and
three TO phonons, respectively. (The peak at 3.366 eV in
Fig. 3 corresponds to the first group of lines described earlier, greatly broadened and saturated under these conditions.)
To provide further information on the origin of luminescence features, we recorded PL spectra for various excitation
intensities at low temperature. These results are presented in
Fig. 4. As excitation intensity was increased, very different
behavior was observed for the two sets of lines distinguished
above. Fist, structure on the low energy wing of the dominant spectral peak at 3.366 eV was resolved into two additional peaks whose intensity ratio with respect to the main
peak was greatly altered at high intensity. This was evidently
due to a saturation effect experienced by the main peak
which did not affect the subsidiary peaks, suggesting that the
lines in the wing were of different origin. All three lines had
the common feature however of showing no shift with intensity, consistent with bound exciton emission. By contrast, the
main peak at 3.310 eV of the second group showed a significant shift to higher energy as a function of incident intensity,
in a direction opposite to that of the band edge variation with
The shift of the 3.310 eV line and its satellites to higher
energy as excitation intensity is increased is an important
signature of donor-acceptor pair recombination.” This .effect
is explained by the saturation of distant donor-acceptor pairs
with slow decay rates at high intensity and a relative increase
in the proportion of near pair decays which occur both faster
and at shorter wavelengths. Consequently we assign the peak
at 3.310 eV to a DA pair recombination band origin and its
three satellites to phonon-assisted replicas of the pair recombination process. However we were unable to determine the
donor or acceptor species responsible for this emission band.
Our interpretation of the features grouped at higher energy is as follows. The shift of the dominant line at 3.366 eV
with temperature tracks expected behavior of the band
edge.r4 However its temperature dependence distinguishes it
from the highest energy feature at 3.376 eV and its intensity
independence is different from that of the two lower energy
features in this group of lines (seen most clearly at 3.358 and
3.352 eV in the top curve of Fig. 4). Hence we -conclude
there are at least three distinct types of emission evident in
our PL spectra between 3.35 and 3.38 eV. The dominant line
is attributed to a single bound exciton line for reasons further
discussed below, although it could also arise from holes
(electrons) combining with donor-(acceptor-) bound electrons (holes). The two features in the low energy wing of this
line are attributed to an exciton bound to a different center.
The small feature in the high energy wing of the dominant
line is evidently also of excitonic origin but warrants closer
J. Appl. Phys., Vol. 77, No. 4, 15 February 1995
Hong eta/.
Downloaded 21 Mar 2002 to Redistribution subject to AIP license or copyright, see
scrutiny, since by virtue of its short wavelength it is a candidate for the free exciton emission line.
The shortest wavelength emission from our samples occurred at an energy only 10 meV higher than the dominant
PL line. Its splitting with respect to the main line is therefore
quite different from either the LO or the TO phonon frequency, making it unlikely that this line is phonon-related.
One potential explanation for its origin is band-to-band recombination, but the 10 meV interval between this feature
and the one next lower in energy is much smaller than the
exciton binding energy, presumed to be very close to the
effective mass value of 30 meV estimated for h-GaN.6 Also,
it could in principle result from the B exciton terminating on
a light hole level. In h-GaN, the G7-Gs valence band splitting
is only 6 meV however,7 rendering this possibility remote.
An alternative explanation would be that 10 meV represents
a donor (or acceptor) binding energy and that the emission is
due to free carrier-bound carrier recombination. In this case
the difference between the location of the conduction band
edge and this emission feature would furnish the offset of the
donor (acceptor) level from the conduction (valence) band.
We note however, that recent experimental estimates of the
band gap”*22 as well as the absorption data of Fig. 1 suggest
an altogether different interpretation.
The most direct determination of the gap to date”” relied
on an absorption measurement and furnished a value of
E,=3.30?0.02 eV compared to 3.4120.02 eV for the hexagonal phase at room temperature. The difference between
these two values is DE,=O.ll
eV, in excellent agreement
with the theoretical estimate of 0.10 eV provided by
Bloom.13 Reflectivity measurements2t are also in accord with
Powell’s experiment, giving a gap energy of roughly 3.35
eV. Earlier cathodoluminescence resultst3’14 suggested much
higher values, but were based on unconfirmed spectral analysis.
If we calculate a low temperature free exciton energy for
c-G+ by simply subtracting the experimental DE, from the
experimental free exciton energy in h-GaN,% we obtain
E,,,=3.37t0.02 eV for c-GaN. This corresponds very
closely with our highest energy feature (3.372 eV) at liquid
helium temperature. Hence we think it most likely that the
3.372 eV feature in our PL spectra corresponds to the free
exciton line, which is thermally activated from a bound exciton or other recombination center. Evidence for thermal
activation of the exciton level from the upper level of the
transition at 3.366 eV comes from plotting the intensity of
this feature relative to that of the 3.366 eV line versus temperature. When this is done using the data of Fig. 2, a fit to a
Boltzmann distribution yields a value of 102 meV for the
activation energy, which corresponds precisely to the splitting between the two features. Moreover the splitting between these peaks remains constant as temperature increases,
as one would expect. Hence the most consistent interpretation that we can offer for the origin of the highest energy
feature is that it corresponds to the free exciton. By making
use of the same value of exciton binding energy for c-GaN
as that estimated for h-GaN,4 we can infer a band gap of
E,=3.376 eV+0.030 eV=(3.406%0.015) eV for c-GaN at
6.5 K from our PL data.
J. Appl. Phys., Vol. 77, No. 4, 15 February 1995
Comparing these results with earlier cathodoluminescence spectra in MBE material, we note that the luminescence peak reported by Okumura at 325 nm (Ref. 15) for
c-GaN/(lOO)GaAs does not correspond well with the features reported here, whereas the CL spectrum for c-GaN/
(1ll)GaAs in the same paper may be comprised of unresolved peaks from all the spectral features identified here.
Luminescence at longer wavelengths in spectra of Ref. 15 is
virtually absent, just as in our MOCVD samples. Strite’2 on
the other hand resolved many peaks in cathodoluminescence,
all quite different from ours, and reported a very strong band
between 530 and 710 nm which they attributed to emission
from midgap impurity levels. In view of the intense point
defect emission observed by them in the visible spectral region, it is perhaps not surprising that defect-related emission
at shorter wavelengths was also quite different from that reported here. It seems likely that the weak excitonic and DA
pair spectral features identified here were either masked in
the latter work by more intense emission of different origin
or that electron-hole interactions were screened by ionized
impurities which reduced the probability of exciton formation.
In summary, we have grown (200) cubic and (111)
cubit/(0002) hexagonal GaN films on (100) GaAs substrates
and have observed and analyzed their near band-edge optical
spectra. The observed features remained invariant with material phase except for the spectrum intensity which became
stronger for samples grown at higher temperatures. Excitation intensity dependence, together with the observed Raman
spectrum of GaN, have permitted us to identify ,a prominent
DA pair recombination line at 3.3 10 eV accompanied by TO
phonon replicas. Features at 3.366, 3.358, and 3.352 eV were
attributed to distinct bound exciton lines, and from their temperature dependences the free exciton was identified and its
energy deduced to be 3.376=0.003 eV at liquid helium temperature.
The authors wish to thank J. Singh for useful discussions, N. Draidia for assistance with sample preparation, and
M. Yoder for his enthusiasm and encouragement of this research. Two of us (S.C.R. and S.W.B.) wish to acknowledge
partial support from the National Science Foundation Science and Technology Center for Ultrafast Optical Science
(STC PHY 8920108). Research on GaN growth was funded
by the Office of Naval Research under contract NOO14-9251552.
‘D. Pavlidis, C. H. Hong, and K. Wang, Proceedings of Workshop on
Compound Semiconductors and Integrated Cckts (WOCSDICE), Cork,
Ireland, May 1994, paper 1.2.
‘C H. Hong, K. Wang, and D. Pavlidis, Proceedings of 36th Electronic
Materials Conference (EMC), Boulder, CO, June 1994, paper Q5.
“J. I. Pankove, S. Bloom, and G. Harbeke, RCA Rev. 36, 163 (1975).
4V. S. Vavilov, S. I. Makarov, M. V. Chukichev, and I. F. Chetverikova,
Sov. Phys. Semicond. 13, 1259 (1979).
‘H. G. Grimmeis and B. Monemar, J. Appl. Phys. 41, 4054 (1970); 0.
Lagerstedt and B. Monemar, J. Appl. Phys. 45, 2266 (1974).
Hong et al.
Downloaded 21 Mar 2002 to Redistribution subject to AIP license or copyright, see
‘R. Dingle and M. llegems, Solid State Commun. 9, 175 (1971).
‘R. Dingle, D. D. Sell, S. E. Stokowski, and M. Ilegems, Phys. Rev. B 4,
1211 (1971).
*T Lei, T. D. Moustakas, R. J. Graham, Y. He, and S. J. Berkowitz, J. Appl,
Phys. 17,4933 (1992).
yhl. J. Paisley, Z. Sitar, J. B. Posthill, and R. F. David, J. Vat. Sci. Technol.
B 7, 701 (1989).
‘OR C. Powell, G. A. Tomasch. Y. W. Kim, J. A. Thorston, and J. E. Greene,
Proc. MRS Symp. 162, 525 (1992).
“S Miyoshi, K. Onabe, N. Ohkouchi, H. Yaguchi, R. Ito, S. Fukatsu, and Y.
Shiraki, J. Cryst. Growth 124,439 (1992).
“M. Mizuta, S. Fujieda, Y. Matsumoto, and T. Kawamura, Jpn. I. Appl.
Phys. 25, L945 (1986).
13S. Fujieda and Y. Matsumoto, Jpn. J. Appl. Phys. 30, L1665 (1991).
14S. Strite, J. Ruan, 2. Li, A. Salvador, H. Chen, D. J. Smith, W. J. Choyke,
and H. Morkoc, J. Vat. Sci. Technol. B 9,’1924 (1991).
“H. Okumura, S. Misawa, and S. Yoshida, Appl. Phys. Lett. 59, 1058
“C. H. Hong, K. Wang, and D. Pavlidis, Proceedings of 21st International
J. Appl. Phys., Vol. 77, No. 4, 15 February 1995
Symposium on Compound Semiconductors, San Diego, CA, Sept.. 1994.
17S.Murugkar, R. Merlin, T. Lei, and T. D. Moustakas, Bull. Am. Phys. Sot.
37, 556 (1992), abstract M23-7.
Is.5. W . Brown, S. C. Rand, C.-H. Hong, and D. Pavlidis, Proceedings of the
Materials Research Society, San Francisco, CA, April 4-8, 1994, paper D
“V. Lemos, C. A. Arguello, and R. C. C. Leite, Solid State Comtnun. 11,
1351 (1972); A. Cingolani, M. Ferrara, M. Lugara, and G. Scamarcio,
Solid State Commun. 58, 823 (1986).
“D. G. Thomas, J. J. Hopfield, and W. M. Augustyniak, Phys. Rev. A 140,
202 (1965).
alH. Okumura, S. Yoshida, and T. Okabisa, Appl. Phys. I.&t. 64, 2997
“R. Powell, G. Tomasch, Y. Kim, J. Thornton, and J. Greene, Mater. Res.
Sot. Symp. Proc. 162, 525 (1990).
a3S. Bloom, G. Harbeke, E. Meier, and I. B. Ortenburger, Phys. Status Solidi
B 66, 161 (1974).
s4B. Monemar, Phys. Rev. B 10, 676 (1974).
Hong et a/.
Downloaded 21 Mar 2002 to Redistribution subject to AIP license or copyright, see
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Related manuals

Download PDF