Journal of Crystal Growth 246 (2002) 230 236
Surface polarity dependence of decomposition and growth of
GaNstudied using in situ gravimetric monitoring
Akinori Koukitu*, Miho Mayumi, Yoshinao Kumagai
Department of Applied Chemistry, Faculty of Technology, Tokyo University of Agriculture and Technology, Koganei,
Tokyo 184-8588, Japan
Abstract
The influence of lattice polarity of wurzite GaN(0 0 0 1) on its decomposition rate was investigated using an in situ
gravimetric monitoring (GM) method with freestanding GaN(0 0 0 1) substrates. At temperatures between 8001C
%
and 8501C, the decomposition rate of GaN(0 0 0 1) was faster than that of GaN(0 0 0 1). On the other hand, the
%
decomposition rate of GaN(0 0 0 1) was faster than that of GaN(0 0 0 1) at temperatures between 9001C and 9501C.
The relation between the decomposition rate and the H2 partial pressure (PH2) indicates that the rate-limiting reactions
are N(surface)+3H2(g)-NH3(g) at lower temperatures, but Ga(surface)+1H2(g)-GaH(g) at higher temperatures. In
2 2
addition, we used the GM method to measure the growth rate of GaN(0 0 0 1) on freestanding GaN(0 0 0 1) substrates.
%
In the low-temperature region, GaN(0 0 0 1) grew faster than GaN(0 0 0 1), whereas GaN(0 0 0 1) grew faster than
%
GaN(0 0 0 1) in the high-temperature region.
r 2002 Elsevier Science B.V. All rights reserved.
Keywords: A1. Etching; A1. Surface processes; A3. Vapor phase epitaxy; B1. Nitrides; B2. Semiconducting gallium compounds
1. Introduction VPE growth mechanism of III-nitride semicon-
ductors such as GaN is quite different from that
GaNand related compounds are very attractive of the other III V semiconductors such as GaAs.
materials for blue/green light-emitting diodes This is due to the difference between the growth
(LEDs), violet laser diodes (LDs), and high- reactions. In MOVPE growth, the reaction govern-
frequency electronic devices. From the view- ing GaN deposition is Ga(g)+NH3(g)-
point of understanding processes in vapor phase GaN(s)+3H2(g), whereas that for GaAs deposition
2
epitaxy (VPE) such as metalorganic vapor phase is Ga(g)+1As4(g)-GaAs(s) [1]. In HVPE growth,
4
epitaxy (MOVPE) and hydride vapor phase epitaxy the reactions are instead GaCl(g)+NH3(g)-
(HVPE), etching/decomposition and growth me- GaN(s)+HCl(g)+H2(g), where hydrogen is formed
chanisms of GaN are of great interest. The by the GaNdeposition, whereas GaAs deposition
has GaCl(g)+1As4(g)+1H2(g)-GaAs(s)+HCl(g).
4 2
Hydrogen is a reaction product in GaNMOVPE
and HVPE growth. Consequently, hydrogen is
*Corresponding author. Tel.: [phone]; fax: +81-
expected to play an important role in the GaN
423-86-3002.
E-mail address: [e-mail] (A. Koukitu). system.
0022-0248/02/$ - see front matter r 2002 Elsevier Science B.V. All rights reserved.
PII: S 0022- 0248( 02) 01746- 3
A. Koukitu et al. / Journal of Crystal Growth 246 (2002) 230 236 231
On the other hand, because wurtzite GaN is
polar, there are two surface structures along the
c-axis direction: the (0 0 0 1) Ga face, where a Ga
atom combines with three Natoms in bulk GaN,
%
and the (0 0 0 1) Nface, where a Natom combines
with three Ga atoms. A few groups have char-
acterized the polarity of GaNgrown on sapphire
(0 0 0 1) substrates, but the reaction between the
GaN surface and the vapor remains poorly
understood. For example, Kobayashi and Ko-
bayashi studied nitrogen desorption from GaN
surface using an in situ surface photoabsorption
(SPA) method [2]. Groh et al. studied GaN
decomposition by weighing a piece of GaN on
sapphire before and after annealing using an
analytical balance [3], whereas Rebey et al. studied
the same process using laser reflectometry [4].
Although they showed that hydrogen in the carrier
gas had a profound effect on the decomposition,
the decomposition rate is speculative.
A deeper understanding of the decomposition
and growth mechanism in VPE is relevant because
Fig. 1. Schematic of the GM system for GaNdecomposition.
it has implications for our understanding of
the chemistry in GaN VPE. In previous papers,
we have developed an in situ gravimetric monitor- (MOHVPE) on GaAs(1 1 1)A substrate [10]. One
ing (GM) system [5 9]. The use of this system is side of the GaN substrate was covered by a
attractive for investigating the decomposition and protective SiO2 mask. The substrate was sus-
growth mechanism of GaN because the system pended from the microbalance with a fused quartz
provides direct information on the decomposition fiber. The carrier gases used were H2 and He as an
and growth rates at the monolayer (ML) level in inert gas. NH3 was introduced over the GaN
real time. Here, we used the in situ GM system to substrate while heating the furnace to prevent
determine the polarity dependence of GaNdecom- GaN decomposition before measurements. The
position and growth. GaN decomposition rate under each fixed set of
conditions was measured after switching off the
NH3 flow. By monitoring the weight change of
2. Experimental procedure the GaN substrate, the decomposition rates of
%
both GaN(0 0 0 1) and GaN(0 0 0 1) faces were
Fig. 1 shows the in situ GM system used for measured in the temperature range from 7301Cto
monitoring the decomposition rate of GaN. The 9501C with PH2 ź 1 atm. In addition, we mon-
GM system, consisting of a vertical quartz reactor itored the dependence of decomposition rate on
and a recording microbalance, provides direct hydrogen partial pressure PH2 in the carrier gas by
information on the substrate weight change caused varying the ratio of H2 to H2+He.
by the decomposition of GaN from the surface Fig. 2 shows the in situ monitoring system for
under atmospheric pressure. This system has a measuring the growth rate of GaN by HVPE.
sensitivity of 0.004 mg under dynamical conditions. GaCl, which was obtained by reacting metallic Ga
We used a freestanding GaN(0 0 0 1) substrate with HCl at 7801C, flowed onto the substrate
(1.0 1.0 0.03 cm3) that was prepared by meta- surface using a suction pump controlled by a
lorganic hydrogen chloride vapor phase epitaxy computer. NH3 was used as a nitrogen source and
232 A. Koukitu et al. / Journal of Crystal Growth 246 (2002) 230 236
Fig. 2. Schematic of the GM system for GaNgrowth.
was introduced from a separate tube using air-
Temperature (°C)
operated valves. Also, the substrates used were
1000 900 800 700
freestanding GaN(0 0 0 1) substrates prepared by
MOHVPE on GaAs(1 1 1)A surfaces. One side of
101
the GaNsubstrate was covered by a SiO2 film.
GaN (0001) Ga
GaN (0001) N
3. Decomposition of GaN(0 0 0 1) surfaces
The decomposition rates of GaN(0 0 0 1) and
100
%
GaN(0 0 0 1) faces as a function of temperature
are shown in Fig. 3. The partial pressure of H2 in
the carrier gas was kept constant at 1 atm and the
decomposition temperature was varied from
7301C to 9501C. Regardless of the polarity, the
decomposition rates of both GaN(0 0 0 1) surfaces
-1
10
increase markedly with increasing temperature.
This result agrees with our previous study using an
epitaxial GaNfilm grown on sapphire (0 0 0 1) [8].
The data show that the dependence on tempera-
ture is different at low and high temperatures
0.8 0.9 1.0
(Fig. 3): from 8501C to 9501C, the activation
-1
% 1000/T(K )
energies for GaN(0 0 0 1) and GaN(0 0 0 1) are
242 and 259 kJ/mol, respectively; whereas from
%
Fig. 3. Decomposition rates of GaN(0 0 0 1) and GaN(0 0 0 1)
7301C to 8501C, the activation energies for as a function of temperature at PH2 ź 1 atm.
Decomposition rate (
µ
m/h)
A. Koukitu et al. / Journal of Crystal Growth 246 (2002) 230 236 233
%
GaN(0 0 0 1) and GaN(0 0 0 1) are 143 and 191 kJ/ ture region), respectively. On the other hand, the
3
%
mol, respectively. These results indicate that the values of n for GaN(0 0 0 1) and GaN(0 0 0 1) are
2
rate-limiting reactions for decomposition on both from 8001C to 8501C, and from 8001C to 8751C
surfaces change with temperature. (hereafter the low-temperature region), respec-
Furthermore, the decomposition rate depends tively. Thus, the decomposition rates of
%
on the polarity of the GaNsurface. Below about GaN(0 0 0 1) and GaN(0 0 0 1) are proportional
8201C, the decomposition rate of GaN(0 0 0 1) is to P1=2 in the high-temperature region, whereas the
H2
%
faster than that of GaN(0 0 0 1). Conversely, the decomposition rates of both surfaces are propor-
%
decomposition rate of GaN(0 0 0 1) is faster than tional to P3=2 in the low-temperature region. These
H2
that of GaN(0 0 0 1) at temperatures ranging from results indicate that the rate-limiting reactions for
about 8501C to 9501C. We argue below that these decomposition of both surfaces are
differences in decomposition rate for the two 1
GaðsurfaceÞ þ H2ðgÞ
2
different lattice polarities is due to the different
-GaHðgÞ ðhigh-temperature regionÞ; ð1Þ
bonding configuration of the GaNsurfaces.
To better understand the surface reaction of H2 and
on GaN, we measured the decomposition rates of 3
NðsurfaceÞ þ H2ðgÞ
2
GaN for PH2 values ranging from 0 to 1. Fig. 4
-NH3ðgÞ ðlow-temperature regionÞ: ð2Þ
shows the PH2 dependence of decomposition rates
%
for GaN(0 0 0 1) and GaN(0 0 0 1) in the tempera- Consequently, the rate-limiting reactions for
ture range from 8001C to 9501C. The decomposi- the decomposition of both GaN(0 0 0 1) and
% %
tion rates of both GaN(0 0 0 1) and GaN(0 0 0 1) GaN(0 0 0 1) likely shift from reaction (1) to
increased with increasing PH2 at a given tempera- reaction (2) with increasing temperature.
ture. Therefore, it is clear that H2 plays an As described above, decomposition of GaN is
important role in the decomposition of GaN. limited by the Ga surface on the GaNsubstrate in
To make this more specific, we fit the data to the the high-temperature region. The topmost Ga
rate equation for GaN decomposition r ź kPn ; atoms on the GaN(0 0 0 1) surface each combine
H2
where r; k; and n are the decomposition rates, rate with three Natoms (three back-bonds) in the bulk,
%
constants, and order of reaction, respectively. In whereas Ga atoms on the GaN(0 0 0 1) surface
Fig. 4, the values of n are drawn at each each form one back-bond with only one Natom in
temperature. The values of n for GaN(0 0 0 1) the bulk. This is the likely reason why the
1
% %
and GaN(0 0 0 1) are from 8751C to 9501C, and decomposition rate of GaN(0 0 0 1) is faster than
2
from 9251C to 9501C (hereafter the high-tempera- that of GaN(0 0 0 1) in the high-temperature
GaN (0001) GaN (0001)
5
n=1/2
5
4
4
n=1/2
n=1/2
3
3
n=1/2
n=0.7
n=1/2
2
2
n=1.0
n=3/2
n=3/2
1 n=3/2
1
n=3/2
n=3/2
950
950
1.0
1.0
900 0.8
900 0.8
0.6
0.6
850 0.4
850 0.4
0.2
0.2
800 0.0
800 0.0
(b)
(a)
%
Fig. 4. PH2 influence on decomposition rate for: (a) GaN(0 0 0 1) and (b) GaN(0 0 0 1) in the temperature range from 8001C to 9501C.
h)
/
m/h)
m
µ
(
µ
(
Rate
Rate
n
n
Decompositio
Decompositio
Tempe
Tempe
(atm)
(atm)
rature(
rature(
Pressure
Pressure
°
Partial
Partial
C)
°
C)
ogen
Hydrogen
Hydr
234 A. Koukitu et al. / Journal of Crystal Growth 246 (2002) 230 236
region. On the other hand, the limiting surface is As described above, the temperature of max-
the N surface on the GaN substrate in the low- imum growth rate depends on the partial pressures
temperature region. The decomposition rate of of the source materials. In the in situ monitoring
%
GaN(0 0 0 1) is faster than that of GaN(0 0 0 1) system for GaNgrowth, the input GaCl pressure
because each Natom on the GaN(0 0 0 1) surface used was low compared with typical values used
combines with only one Ga atom in the bulk, for GaNHVPE because this allows more precise
%
whereas each Natom on GaN 0 0 1) forms back- measurements in the low-growth rate region. The
(0
%
bonds to three Ga atoms in the bulk. Thus, the growth rates of GaN(0 0 0 1) and GaN(0 0 0 1)
different bonding configuration of the GaN surfaces increase with increasing temperature in
surfaces can explain the difference of the decom- the low-temperature region up to 7001C, then they
position rates between GaN(0 0 0 1) and decrease with an increase of temperature (Fig. 5).
%
GaN(0 0 0 1). The activation energies for GaNgrowth obtained
H2 is generally used as a carrier gas in MOVPE from the slopes in the low-temperature region are
and HVPE growth. Additionally, H2 is formed as 68 kJ/mol for GaN(0 0 0 1) and 64 kJ/mol for
%
a by-product of the reaction that forms GaN. GaN(0 0 0 1).
%
Therefore, the reaction between H2 gas and the The growth rate of the GaN(0 0 0 1) surface is
GaN surface should cause some GaN decom- faster than that of the GaN(0 0 0 1) surface in the
position even during growth. For example, in the low-temperature region, which is the reverse of the
high-temperature region, the net growth rate of relation in the high-temperature region. Although
GaN(0 0 0 1) should be faster than that of the growth rate dependence on the temperature
%
GaN(0 0 0 1) because the net growth rate in VPE was shifted to low temperature due to the low
is determined by the growth rate minus the input partial pressure of GaCl used in this work,
decomposition rate. This growth rate difference
agrees with the data on MOVPE growth of
GaNreported by Sumiya et al. [11] and Rouviere
Temperature (°C)
et al. [12].
800 750 700 650 600 550
10
GaN(0001)Ga
4. Growth on GaN(0 0 0 1) surfaces
_
GaN(0001)N
The kinetics of HVPE is considerably more
complex. In the low-temperature region, an in-
crease in GaCl partial pressure actually results in a
decrease in the growth rate because GaCl hinders
desorption of HCl from the surface. On the other
hand, the growth rate has a maximum at an
intermediate temperature due to competing tem-
perature trends from thermodynamics and surface
kinetics. In particular, because growth is generally
exothermic, the growth rate should decrease with PGaCl : 1x10-4 atm
1
increasing temperature. In fact, this decrease in the
PNH3: 8x10-2 atm
high-temperature region is approximately equal to
that predicted from thermodynamics. But in the
low-temperature region, growth is limited by sur-
0.95 1.00 1.05 1.10 1.15 1.20 1.25
face kinetics such that the growth rate increases
1000/T(K-1)
with increasing temperature. Consequently, the
temperature of maximum growth rate depends on
%
Fig. 5. Growth rates of GaNon GaN(0 0 0 1) and (0 0 0 1) as a
the amount of GaCl or NH3 in the vapor. function of temperature.
Growth Rate (
µ
m/h)
A. Koukitu et al. / Journal of Crystal Growth 246 (2002) 230 236 235
%
than that of GaN(0 0 0 1). On the other hand, in
750°C
P : 1x10-4atm
GaCl
700°C the high-temperature region, the decomposition
%
5
rate of GaN(0 0 0 1) was faster than that of
GaN(0 0 0 1). The decomposition rates of both
surfaces increased rapidly with increasing tem-
4
perature. Dependence of the GaNdecomposition
rate on PH2 showed that the decomposition rates
GaN(0001) Ga
for both polarities were proportional to P3=2 and
H2
3
P1=2 in the low- and the high-temperature regions,
H2
GaN(0001) N
respectively. Based on these results, the relations
between the lattice polarity and the decomposition
2
of GaNcan be explained clearly by considering the
rate-limiting reactions and the bonding configura-
400 600 800
tions on GaNsurfaces.
Input V/III Ratio
From the preliminary in situ monitoring of the
Fig. 6. Growth rates of GaNas a function of input V/III ratio.
growth rates, it is found that the growth rate on
%
the GaN(0 0 0 1) surface is faster than that on
GaN(0 0 0 1) in the low-temperature region, but
because of the precise in situ monitoring, the the reverse relation holds in the high-temperature
relation of growth rates with the GaNpolarities is region. The relation of growth rates with the GaN
very close to the nature indicated by the results of polarities is very close to that predicted from
the GaNdecomposition. considering the GaNdecomposition.
To help clarify the difference of growth rates
between the two polarities, we show the growth
rates as a function of the input V/III ratio in Acknowledgements
Fig. 6. For these measurements, the input GaCl
pressure was kept constant at 1 10 4 atm and the The authors would like to express their sincere
input NH3 pressure was varied from 4 10 2 to thanks to Y. Matsuo for preparation of our GaN
8 10 2 atm. All the growth rates increase with an samples. This work was partly supported by the
increase of the input V/III ratio at 7001C (i.e., in Foundation for Promotion Material Science and
the low-temperature region in this work) and Technology of Japan (MST) and by the Grant-in-
7501C (high-temperature region in this work). The Aid for Scientific Research from the Ministry
dependence of the growth rate on the lattice of Education, Culture, Sports, Science and
polarity is significantly clarified here. In the low- Technology.
temperature region, the growth rate on the
%
GaN(0 0 0 1) substrate is faster than that on the
GaN(0 0 0 1) substrate, whereas the growth rate on References
the GaN(0 0 0 1) is faster than that on the
% [1] A. Koukitu, T. Taki, N. Takahashi, H. Seki, J. Crystal
GaN(0 0 0 1) in the high-temperature region.
Growth 197 (1999) 99.
[2] Y. Kobayashi, N. Kobayashi, J. Crystal Growth 189/190
(1998) 301.
5. Conclusions
[3] R. Groh, G. Gerey, L. Bartha, J.I. Pankove, Phys. Stat.
Sol. A 26 (1974) 353.
[4] A. Rebey, T. Boufaden, B.E. Jani, J. Crystal Growth 203
The dependence of GaNdecomposition on the
(1999) 12.
lattice polarity has been investigated by an in situ
[5] A. Koukitu, H. Ikeda, H. Yasutake, H. Seki, Jpn. J. Appl.
GM method using a freestanding GaN(0 0 0 1)
Phys. 30 (1991) L1847.
substrate. In the low-temperature region, the
[6] A. Koukitu, Y. Kumagai, T. Taki, H. Seki, Jpn. J. Appl.
decomposition rate of GaN(0 0 0 1) was faster Phys. 38 (1999) 4980.
µ
m/h)
Growth Rate (
236 A. Koukitu et al. / Journal of Crystal Growth 246 (2002) 230 236
[7] Y. Kumagai, M. Mayumi, A. Koukitu, H. Seki, Appl. [11] M. Sumiya, M. Tanaka, K. Ohtsuka, T. Ohnishi,
Surf. Sci. 159/160 (2000) 427. I. Ohkubo, M. Yoshimoto, H. Koinuma, M. Kawasaki,
[8] M. Mayumi, F. Satoh, Y. Kumagai, K. Takemoto, S. Fuke, Appl. Phys. Lett. 75 (1999) 674.
A. Koukitu, Jpn. J. Appl. Phys. 39 (2000) L707. [12] J.L. Rouviere, M. Arlery, R. Niebuhr, K.H. Bachem,
[9] M. Mayumi, F. Satoh, Y. Kumagai, T. Takemoto, O. Briot, MRS Internet J. Nitride Semicond. Res. 1
A. Koukitu, Phys. Stat. Sol. B 228 (2001) 537. (1996) 33.
[10] Y. Kumagai, H. Murakami, A. Koukitu, K. Takemoto,
H. Seki, Jpn. J. Appl. Phys. 39 (2000) L703.