Kinetics and mechanism of thermal decomposition of GaN

Article abstract of DOI:10.1016/S0040-6031(00)00558-X

A scheme of dissociative evaporation of GaN with the partial evolution of nitrogen in the form of free atoms has been invoked to interpret the decomposition mechanism of GaN in vacuum or inert gas atmosphere. A critical analysis of literature data and their comparison with theoretical calculations has shown that the main kinetic characteristics of decomposition, including the absolute decomposition rate and activation energy are in full agreement with the reaction: GaN(s) → Ga(g)+0.5N+0.25N₂. Condensation of the gallium vapor in the reaction zone and partial transport of condensation energy to GaN account for the features which are typical of many solid-state reactions and manifest themselves in the appearance of induction and acceleration periods in the course of the process. The low temperature decomposition of GaN in H₂ according to the equilibrium reaction GaN(s)+ 1.5H₂=Ga(g)+NH₃ is supported by a good agreement of experimental and calculated activation energies and by the strong inhibition effect of NH₃ on the GaN decomposition. As expected, condensation of Ga vapor in the Ga(l)/GaN(s) interface accelerates the reduction of GaN by H₂ several hundred times. (C) 2000 Elsevier Science B.V.

Full text of DOI:10.1016/S0040-6031(00)00558-X

Thermochimica Acta 360 (2000) 85±91
Kinetics and mechanism of thermal decomposition of GaN
Boris V. L'vov*
Department of Analytical Chemistry, St. Petersburg State Technical University, St. Petersburg 195251, Russia
Received 7 March 2000; received in revised form 3 May 2000; accepted 7 May 2000
Abstract
A scheme of dissociative evaporation of GaN with the partial evolution of nitrogen in the form of free atoms has been
invoked to interpret the decomposition mechanism of GaN in vacuum or inert gas atmosphere. A critical analysis of literature
data and their comparison with theoretical calculations has shown that the main kinetic characteristics of decomposition,
including the absolute decomposition rate and activation energy are in full agreement with the reaction:
GaN(s) ! Ga(g)‡0.5N‡0.25N2. Condensation of the gallium vapor in the reaction zone and partial transport of condensation
energy to GaN account for the features which are typical of many solid-state reactions and manifest themselves in the
appearance of induction and acceleration periods in the course of the process. The low temperature decomposition of GaN in
H2 according to the equilibrium reaction GaN(s)‡1.5H2ˆGa(g)‡NH3 is supported by a good agreement of experimental and
calculated activation energies and by the strong inhibition effect of NH3 on the GaN decomposition. As expected,
condensation of Ga vapor in the Ga(l)/GaN(s) interface accelerates the reduction of GaN by H2 several hundred times.
# 2000 Elsevier Science B.V. All rights reserved.
Keywords: Activation energy; Autocatalysis; Dissociative evaporation; Gallium nitride; Kinetics; Mechanism
1. Introduction
1965, Munir and Searcy [3] measured by a torsion-
effusion method the heat of activation for the reaction
2GaN(s) ! 2Ga(l)‡N2 and by a torsion-Langmuir
method the rate and heat of activation for the reaction
2GaN(s) ! 2Ga(g)‡N2. In 1965, Schoonmaker et al.
[4] determined the pressure over GaN and GaN‡Ga
mixture by weight loss effusion and torsion effusion
techniques. The results supported that the vaporization
coef®cient for the decomposition is much less than
unity. In contrast to that, the binary mixture (a direct
contact of GaN with liquid Ga) showed considerable
dissociation. The authors [4] proposed that a non-
reactant liquid Ga may participate as catalysts in the
vaporization process by dissolving GaN and disrupt-
ing its rigid wurtzite crystal structure.
Gallium nitride is a semiconductor with a large
direct band gap which makes this material interesting
in short-wavelength photonic devices for display and
storage applications. Its high thermal conductivity
also opens new routes in high-temperature/high-power
electronics. Knowledge of GaN decomposition con-
ditions and mechanism is vital for ®lm growth, pro-
cessing of devices and high-temperature operation.
There are not many studies in the literature con-
cerning GaN decomposition though the ®rst report
was published by Johnson et al. more than 60 years
ago [1]. In 1964, Searcy [2] estimated a coef®cient for
congruent vaporization of GaN of the order 10 6. In
The decomposition of GaN in the presence of H2
was ®rst studied by Thurmond and Logan [5,6] in the
1970s. Products of the reaction are liquid gallium and
* Tel.: ‡7-812-552-7741; fax: ‡7-812-247-4384.
E-mail address: blvov@robotek.ru (B.V. L'vov)
0040-6031/00/$ ± see front matter # 2000 Elsevier Science B.V. All rights reserved.
PII: S 0 0 4 0 - 6 0 3 1 ( 0 0 ) 0 0 5 5 8 - X

86
B.V. L'vov / Thermochimica Acta 360 (2000) 85±91
ammonia gas. A comparison of the NH3 to H2 partial
pressure ratio at the exit of the reactor revealed a very
good agreement with thermodynamic calculations
which means that this reaction is close to equilibrium.
Jacob et al. [7] also measured the decomposition of
GaN in different gases and found a value of initial
decomposition temperature 9708C in Ar/N2 and
6008C in H2.
equation, Ea and A) has been described in a number of
previous publications [13±22]. Therefore, we are
going to present below only some ®nal relations
necessary for the calculations in this work.
In the case of a binary compound S decomposed in
vacuum into gaseous products A and B
Sꢀs† ! aAꢀg† ‡ bBꢀg†
(1)
the ¯ux of product A can be expressed through the
equivalent partial pressure PA (in atm) of this product
corresponding to the hypothetical equilibrium of reac-
tion (1) in the form
Ambacher et al. [8] have measured the ¯ux of N2
and activation energy for GaN decomposition in
vacuum by a quadropole mass spectrometer (QMS).
The enhanced GaN decomposition rate in H2 in the
presence of liquid gallium has been supported recently
by Koleske at al. [9] and Pisch and Schmid-Fetzer
[10]. Tanaka and Nakadaira [11] and Rebey et al. [12]
have investigated the GaN decomposition at various
temperatures in H2 and measured the activation energy
of this process.
The objective of this work is to discuss the mechan-
ism and the above-mentioned features of GaN decom-
position in vacuum, H2 and in the presence of liquid
gallium using a scheme of dissociative evaporation of
the reactant with simultaneous condensation of the
low-volatile product. This approach has been
employed earlier to explain the mechanism and
kinetics of the thermal decomposition of nitrates
[13±15], azides [16], carbonates [17], Li2SO4ÁH20
[18], Mg(OH)2 [19], Ag2O [20], HgO [21], and of a
number of other inorganic compounds [22±24],
including GaN. In the last case, the analysis was based
on the experimental results of only one work [3] and
it was concluded that the GaN decomposition in
vacuum occurs with the evolution of 40% of nitrogen
in the form of free atoms, i.e. by the reaction
GaN(s) ! Ga(g)‡0.4N‡0.3N2.
gMPA
JA ˆ
(2)
1=2
ꢀ2pMART†
whereMandMA arethemolarmassesofthereactantand
product A, respectively, g the coef®cient of conversion
from atmospheres to Pascals, and R is the gas constant.
A theoretical value of the partial pressure of product
A can be calculated from the equilibrium constant Kp
for reaction (1). In the absence of an excess of reaction
products in the reactor atmosphere, the situation cor-
responding to the equimolar evaporation mode, the
partial pressure PA can be expressed [23] as
ꢀ
a
F1=n
ꢁ
ꢀ
ꢁ
1=n
b=2n
KP
F
ꢀ
MA
PeA ˆ a
MB
b=2n
ꢁ
ꢀ
ꢁ
ꢀ
ꢁ
MA
DrS0T
nR
DrHT0
nRT
ˆ
exp
exp
(3)
MB
where
F ꢁ aa  bb
n ˆ a ‡ b
and
(4)
(5)
KP ˆ PaA Á PbB
(6)
Here DrHT0 and DrS0T are, respectively, the changes of
the enthalpy and entropy in process (1).
2. Theoretical
The method to be employed below consists in
comparing experimental data for the kinetic para-
meters with their theoretical values. The calculations
are based on the classical evaporation model of Hertz±
Langmuir, extended to the cases of dissociative eva-
poration of compounds. The scheme for the theoretical
calculation of the main kinetic parameters (the ¯ux of
the gaseous product J, the rate constant k, the product
partial pressure P and the parameters of the Arrhenius
As can be seen from Eq. (3), the calculated activa-
tion energy for reaction (1) can be written as
DrHT0
n
Eae ˆ
(7)
In order to take into account the partial transfer of
the energy released in the condensation of low-volatile
A product to the reactant, we introduced, as before
[18±21,24], into calculations of the enthalpy of

B.V. L'vov / Thermochimica Acta 360 (2000) 85±91
87
Table 1
attribute the anomalous thermal stability of GaN to the
speci®c features of its wurtzite crystalline structure
with very strong covalent bonds between each nitro-
gen atom and its four nearest gallium neighbors and,
as a result of this, to the partial evolution of nitrogen in
the process of thermal decomposition in the form of
free atoms. Actually, the distance of closest approach
of two nitrogen atoms in a lattice of GaN is about
Thermodynamic functions [27±29] used in the calculations
Species
State of
aggregation
DH10300
(kJ mol
S01300
1
)
1
(J mol 1 K
)
Ga
Ga
GaN
N2
l
37.8
302.4
58.9a
40.2
98.7
206.3
120.7a
236.8
183.8
174.2
261.9
g
s
g
g
g
g
N
497.8
38.4
Ê
3.2 A, whereas the N±N internuclear distance in the
Ê
gaseous N2 molecule is only 1.07 A. There are no
H2
NH3
21.5
structural units in the condensed phase which are
similar to gaseous N2 molecule. Therefore, it is dif®-
cult to expect that all nitrogen atoms of the solid GaN
evolve in the process of its decomposition in the form
of N2 molecules. By analogy with our analysis of the
mechanism of decomposition of azides [16], we can
represent the decomposition reaction as
a The entropy and heat capacity of ZnO (an isoelectronic solid)
were used in the evaluation of these functions.
decomposition reaction (1) an additional term
taDcHT0ꢀA†, where the coef®cient t corresponds to
the fraction of the condensation energy transferred to
the reactant. Thus, we can write
GaNꢀs† ! Gaꢀg† ‡ ꢀ1 i†N ‡ 0:5iN2
(9)
DrHT0 ˆ aDHT0ꢀA† ‡ bDHT0ꢀB† DHT0ꢀS†
where the interaction parameter i varies from 0 to l,
depending on the extent to which the nearest nitrogen
atoms interact with one another in the instant of
decomposition.
‡ taDcHT0ꢀA†
(8)
The most plausible of all conceivable mechanisms of
the energy transfer appears to be the thermal accom-
modation [25,26] or, in other words, direct transfer of
the energy at the reaction interface in collisions of the
low-volatile molecules with the reactant and product
surface. For equal temperatures of the solid phases,
one may expect equipartition of energy between the
two phases, i.e. tˆ0.5. Table 1 lists the values of the
thermodynamic functions [27±29] for all components
of the assumed reactions.
To select the appropriate value of interaction para-
meter (or a part of nitrogen evolved in a molecular
form), we used a comparison of experimental and
theoretically calculated kinetic parameters, namely,
the activation energies and absolute rates of decom-
position. Table 2 presents the results of this compar-
ison. Some comments to these data should be made. In
the case of Ea value obtained in [3], we used a
correction for the self-cooling effect at high tempera-
tures. The method of correction has been described
elsewhere [19]. Ambacher et al. [8] used two different
techniques for the determination of decomposition
rate. The ®rst one was based on direct measuring of
the N2 ¯ux from epitaxial GaN by mass spectrometry
in the temperature range from 1125 to 1270 K. The
3. Results and discussion
3.1. Decomposition of GaN in vacuum
In accordance with our previous interpretation of
the low coef®cient of GaN decomposition [22], we
Table 2
Activation energy and partial pressure of nitrogen (reduced to N2) for decomposition of GaN in vacuum at tˆ0
1
)
DT (K)
Ea (kJ mol
T (K)
P (atm)
Experimental
Calculated
Experimental
Calculated
8
8
1166±1428
1125±1270
1123±1223
313a [3]
379 [8]
288 [8]
354
354
354
1300
1200
1173
9.4Â10 [3]
5.0Â10
9
9
2.6Â10 [8]
3.2Â10
9
9
2.0Â10 [8]
1.4Â10
1
a After correction of the original value (305 kJ mol [3]) for the self-cooling effect.

88
B.V. L'vov / Thermochimica Acta 360 (2000) 85±91
second technique (used to verify the data obtained by
the ®rst technique) consisted of measuring the GaN
®lm thickness by X-ray diffraction and determination
of the amount of GaN which was removed by ther-
mally induced decomposition at temperatures of 1123,
1173 and 1223 K.
The best agreement as a whole between the experi-
mental and calculated values corresponds to iˆ0.5
(50% of evolved nitrogen is in the form of N2 and
50%, in the form of free atoms) or the reaction
khin [14,15] in investigations of the decomposition
mechanism of some metal nitrates. Later on, the same
approach was used by L'vov [16] for the interpretation
of decomposition mechanism of azides. In the latter
case, the mass spectrometric data obtained by Walker
and co-workers [34±36] were used. The primary
condition for the right application of this technique,
which is necessary for preventing possible adsorption,
condensation or recombination of unstable species
evolved from a sample surface, is the elimination of
any collisions of these species with furnace and spec-
trometer walls on their way to the QMS ionizer. In the
experimental setup used by Ambacher et al. [8], the
furnace was set perpendicular to ionizer. Therefore,
any direct tracks of species from the sample to the
QMS were excluded. This explains the absence of any
appreciable m-s signals in the measurement of GaN
decomposition [8] not only for atomic nitrogen (as one
of the primary decomposition species) but even for
gallium atoms.
GaNꢀs† ! Gaꢀg†‡0:5N ‡ 0:25N2
(10)
Under this supposition, the discrepancy between the
equivalent partial pressures of nitrogen corresponded
to the ¯uxes of nitrogen at the mean temperatures of
experiments and equilibrium partial pressures for the
reaction implied is within a factor of two. The values
of activation energy compared are in a good agreement
as well, if we take into account that the lesser slope of
1
Arrhenius plot (Eaˆ288 kJ mol [8]) is determined
by only three points and, therefore, is the least reliable.
There are some other facts which support the
suggested scheme of GaN decomposition. Firstly, it
is supported by the absence of a strong depressive
effect of nitrogen atmosphere on the decomposition of
GaN. As reported in several works [10,12,30,31], the
difference in the decomposition rate of GaN in Ar (or
He) and N2 is rather small (in the range of factor of 2).
Note that in case of evolution of nitrogen in the
molecular form it would be higher of one order of
magnitude. Secondly, GaN cannot be prepared via a
direct syntheses of Ga(l/g) and N2 (under not very high
pressures). At the same time, it can be prepared by
reacting Ga with atomic N generated from N2 by
microwave discharge [32] or radio-frequency ®elds
[33]. The last argument in favor of the evolution of
atomic nitrogen is a strong adsorption of nitrogen
(about 90%) at the reactor walls observed by Amba-
cher et al. [8] in their vacuum experiments on the
decomposition of GaN. It may be suggested that free
N atoms evolved are chemisorbed on the surface of
reactor (quartz) or metal surface of QMS.
Let us turn now to the explanation of the catalytic
effect of liquid gallium on the decomposition of GaN.
The origin of these effect has been interpreted recently
in the framework of the mechanism of congruent
dissociative evaporation with simultaneous condensa-
tion of the low-volatility product at the reactant/pro-
duct interface [20,24,37]. The essence of this approach
is the transfer of one-half of the condensation energy
to the reactant and, as a result, increasing of its
decomposition rate. In the case of the decomposition
of GaN, it means that the enthalpy of reaction (10) is
reduced from 629.3 (tˆ0) to 488.0 kJ mol 1 (tˆ0.5)
and the equilibrium partial pressure of Ga at 1300 K
8
increased from 9.9Â10 up to 1.1Â10 4 atm. The
saturated vapor pressure of Ga at 1300 K is 9.7Â
10 6 atm. Therefore, in the presence (or after forma-
tion) of GaN(s)/Ga(l) interface, the decomposition of
GaN should occur with the simultaneous condensation
of Ga vapor and the initial temperature of GaN decom-
position should be reduced from 1300 to 1100 K.
The experimental data are in agreement with these
theoretical estimations. Munir and Searcy [3]
observed in their effusion experiments the appearance
of drops of elemental gallium in the Knudsen cell with
1.4 mm ori®ce, when the ratio of ori®ce area to upper
surface of sample was about 1/100 and the partial
pressure of Ga at 1300 K reached ca. 3Â10 5 atm, i.e.
three-fold of its saturated pressure. For the Knudsen
In connection with these mass spectrometry experi-
ments, a remark can be made. The most reliable and
obvious argument for, or against, the decomposition
mechanism could be obtained in a mass spectrometric
investigation of the primary products of decomposi-
tion. This technique was used by L'vov and Novichi-

B.V. L'vov / Thermochimica Acta 360 (2000) 85±91
89
cell with 3 mm ori®ce and about 1/30 ratio of ori®ce
area to upper surface of sample, the partial pressure of
Ga corresponded to ca. 1Â10 5 atm and no elemental
gallium remained in the cell.
In metal as well. This fact is also in agreement with the
theoretical model proposed and the experimental
observations of the catalytic effect of foreign impu-
rities on the decomposition of other solids, for exam-
ple, of metallic Ni on the decomposition of Ag2O [38].
A further proof of the proposed mechanism are the
results of investigation by Schoonmaker et al. [4] of
the catalytic effect of liquid gallium on GaN decom-
position by weight loss and torsion effusion techni-
ques. In agreement with the previous work [3], they
observed a black residue of ®nely divided gallium
covered the surface of the initially pale yellow nitride
in all runs (with different effusion cells), except those
with very large ori®ces (2.3 mm diameter). The next
comment is remarkable: ``As expected, the effusion
rate increased with increasing of temperature. How-
ever, another curious result was observed. In succes-
sive runs at the same temperature with the same
sample of initially pure, crystalline gallium nitride
the rate of effusion, calculated from weight loss,
increased from one run to the next''. In the framework
of our theory, the reason of this effect is rather
obvious. It is connected with increasing the degree
of formation of GaN(s)/Ga(l) reaction interface. This
increasing process should continue up to the moment
when the gallium ®lm covers the whole surface of
GaN and the decomposition reaction reaches a steady
state.
3.2. Decomposition of GaN in hydrogen
Most of the researchers explain the low temperature
decomposition of GaN in H2 by the reaction
GaNꢀs† ‡ 1:5H2 ˆ Gaꢀg† ‡ NH3
(11)
which is the reverse reaction to that used for a synth-
esis of GaN. This is supported by a good agreement of
experimental and calculated activation energies for
this reaction (Table 3). There are some other facts
which con®rm the equilibrium of this reaction. In the
experiments by Rebey et al. [12], at the initial decom-
position temperature 6008C (873 K) and 1 atm H2, the
equilibrium pressure of Ga vapor should be about
3Â10 8 atm. An addition to H2 a small amount ofNH3
(3%) inhibited the decomposition of GaN [12].
Indeed, in this case (at 873 K) the equilibrium pressure
of Ga was reduced to 4Â10 14 atm. Only at 1250 K,
its value should reach the same 3Â10 8 atm (as in the
absence of NH3 at 873 K). The same inhibition effect
of NH3 on the decomposition of GaN has been
observed by Furtado and Jacob [39].
Another important result of this work has been the
demonstration of GaN decomposition catalysis by the
liquid Ga metal added to GaN samples. No decom-
position of pure, crystalline GaN occurred after
180 min of heating at 1189 K in the open container.
After addition of a ball of liquid gallium into con-
tainer, the sample of GaN began to decompose and the
decomposition pressure slowly increased at constant
temperature (1189 K). The accelerating period (up to
the maximum equivalent pressure 1.4Â10 5 atm) was
about 50 min. In case of better contact between Ga
metal and GaN, the induction period is considerably
reduced. The above mentioned value of equivalent
pressure coincides with the calculated equilibrium
pressure for reaction (10) at the steady state condition
tˆ0.5. Therefore, in all instances (in the presence as
well as in the absence of Ga metal), the decomposition
of GaN in vacuum occurs in agreement with the
proposed scheme.
The high partial pressure of Ga vapor in the decom-
position of GaN in H2 results in its condensation and
formation of droplets of liquid metal on the surface of
sample. Koleske et al. [9] studied this process at
different pressures of H2 (10±700 Torr) and two dif-
ferent temperatures: 1174 and 1265 K. For pressures
less than 40 Torr no Ga droplets were observed. This is
in agreement with the theoretical calculations. At
1265 K and 40 Torr H2, the equilibrium partial pres-
sure of Ga for reaction (11) is smaller than its saturated
Table 3
Activation energy for the decomposition of GaN in H2
1
)
Conditions
DT (K)
Ea (kJ mol
Experimental
155 [11]
Calculated
163
1
7 l min H2
1085±1186
931±1070
flow, 0.1 atm
1
3 l min H2
180 [12]
163
Lastly, Schoonmaker et al. [4] have shown that the
decomposition of GaN is catalyzed not only by Ga, but
flow, 1 atm

90
B.V. L'vov / Thermochimica Acta 360 (2000) 85±91
pressure (3.8Â10 6 and 4.9Â10 6 atm, respectively).
At 76 Torr of H2, when Ga droplets were observed, the
equilibrium partial pressure of Ga (6.1Â10 6 atm) is
24% greater than the saturated vapor pressure. The
remarkable accord between the results of these equi-
librium calculations and observations supports the
validity of thermodynamic functions for GaN pre-
sented in Table 1.
the vaporization coef®cient for decomposition of GaN
10
in vacuum is as low as 1Â10 8 at 1300 K and 4Â10
at 1000 K. It can be expected that this feature (the
partial evolution of diatomic gases in the form of free
atoms) will be observed not only for Al and In nitrides,
but for some other O±, S± or Se±containing com-
pounds with a wurtzite structure. Our preliminary
estimations [22] support this conclusion.
The presence of liquid gallium catalyses the reac-
tion (11) in the same way as it does in the vacuum
decomposition of GaN. It explains `a dramatic
increase in the GaN decomposition rate correlated
with the increased Ga surface accumulation' observed
by Koleske et al. [9]. The maximum rate of GaN
decomposition in H2 and in the presence of liquid
gallium should correspond to tˆ0.5 condition. At
1265 K and 76 Torr H2, as an example, the equilibrium
partial pressure of Ga in reaction (11) should be
increased to 3.3Â10 3 atm, which is several hundred
times higher than that in the absence of liquid gallium
(6.1Â10 6 atm). The equivalent partial pressure of Ga
that corresponded to the decomposition of GaN at
these conditions (1265 K and 76 Torr H2) was equal to
1.3Âl0 4 atm [9]. This is 25 times lower than
expected but it should be taken into account that
the degree of covering of GaN surface by liquid Ga
in this experiment was very small.
Acknowledgements
The author is grateful to Professor F.F. Grekov for
helpful discussions in the process of the manuscript
preparation.
References
[1] W. Johnson, J. Parsons, M. Crew, J. Phys. Chem. 36 (1932)
2651±2654.
[2] A.W. Searcy, High Temperature Technology, Butterworths,
London, 1964.
[3] Z.A. Munir, A.W. Searcy, J. Chem. Phys. 42 (1965) 4223±
4228.
[4] R.C. Schoonmaker, A. Buhl, J. Lemley, J. Phys. Chem. 69
(1965) 3455±3460.
[5] C.D. Thurmond, R.A. Logan, J. Electrochem. Soc. 119 (1972)
622±626.
[6] R.A. Logan, C.D. Thurmond, J. Electrochem. Soc. 119 (1972)
1727±1735.
4. Conclusions
[7] G. Jacob, R. Madar, J. Hallais, Mater. Res. Bull. 11 (1976)
445±450.
As it follows from the above discussion, all the
peculiarities of thermal decomposition of GaN in
vacuum or in an atmosphere of hydrogen can be
quantitatively interpreted and predicted in the frame-
work of the scheme of dissociative evaporation of the
reactant with simultaneous condensation of the low-
volatile product. The reason for the high thermal
stability of GaN in vacuum or in inert gases is in its
wurtzite crystalline structure with strong covalent
bonds between each nitrogen atom and its four nearest
gallium neighbors and the long distances between
nitrogen atoms in the lattice. As a result of these,
the partial evolution of nitrogen in the process of
thermal decomposition is in the form of free atoms.
In relation to equilibrium
[8] O. Ambacher, M.S. Brandt, R. Dimitrov, T. Metzger, M.
Stutzmann, R.A. Fisher, A. Miehr, A. Bergmaier, G.
Dollinger, J. Vac. Sci. Technol. B 14 (1996) 3532±3542.
[9] D.D. Koleske, A.E. Wickenden, R.L. Henry, M.E. Twigg,
J.C. Culbertson, R.J. Gorman, Appl. Phys. Lett. 73 (1998)
2018±2020.
[10] A. Pisch, R. Schmid-Fetzer, J. Crystal Growth 187 (1998)
329±332.
[11] H. Tanaka, A. Nakadaira, J. Crystal Growth 189/190 (1998)
730±733.
[12] A. Rebey, T. Boufaden, B. El Jani, J. Crystal Growth 203
(1999) 12±17.
[13] B.V. L'vov, Mikrochim. Acta (Wien) II (1991) 299±308.
[14] B.V. L'vov, A.V. Novichikhin, Spectrochim. Acta Part B 50
(1995) 1427±1448.
[15] B.V. L'vov, A.V. Novichikhin, Spectrochim. Acta Part B 50
(1995) 1459±1468.
[16] B.V. L'vov, Thermochim. Acta 291 (1997) 179±185.
[17] B.V. L'vov, Thermochim. Acta 303 (1997) 161±170.
[18] B.V. L'vov, Thermochim. Acta 315 (1998) 145±157.
2GaNꢀs† ˆ 2Gaꢀl† ‡ N2
(12)

B.V. L'vov / Thermochimica Acta 360 (2000) 85±91
91
[19] B.V. L'vov, A.V. Novichikhin, A.O. Dyakov, Thermochim.
Acta 315 (1998) 135±143.
[29] L.V. Gurvich, I.V. Veits, V.A. Medvedev, et al., Thermo-
dynamic Properties of Individual Substances, Nauka, Mos-
cow, 1978±1982 (in Russian).
[20] B.V. L'vov, Thermochim. Acta 333 (1999) 13±19.
[21] B.V. L'vov, Thermochim. Acta 333 (1999) 21±26.
[22] B.V. L'vov, A.V. Novichikhin, Thermochim. Acta 290 (1997)
239±251.
[30] R.J. Sime, J.L. Margrave, J. Phys. Chem. 60 (1956) 810±
811.
[31] M. Furtado, G. Jacob, J. Crystal Growth 64 (1983) 257±
267.
[23] B.V. L'vov, Spectrochim. Acta Part B 52 (1997) 1±23.
[24] B.V. L'vov, Spectrochim. Acta Part B 53 (1998) 809±820.
[25] J.P. Hirth, G.M. Pound, Condensation and Evaporation Ð
Nucleation and Growth Kinetics, Pergamon Press, Oxford,
1963.
[32] B.B. Kosicki, D. Kahng, J. Vac. Sci. Technol. 6 (1969)
593±596.
[33] H.J. Hovel, J.J. Cuomo, Appl. Phys. Lett. 20 (1972) 71±75.
[34] R.F. Walker, N. Gane, F.P. Bouden, Proc. Roy. Soc. 294
(1966) 417±436.
[26] B. McCarroll, G. Ehrlich, Condensation and energy transfer
on crystals, in: Condensation and Evaporation of Solids,
Gordon and Breach, New York, 1965, pp. 521±538.
[27] V.P. Glushko (Ed.), Thermodynamic Constants of Substances,
Akad. Nauk SSSR, Moscow, 1962±1982 (in Russian).
[28] V.A. Ryabin, M.A. Ostroumov, T.F. Svit, Thermodynamic
Characteristics of Substances, Handbook, Khimiya, Lenin-
grad, 1977 (in Russian).
[35] R.F. Walker, J. Phys. Chem. Solids 29 (1968) 985±1000.
[36] R.F. Walker, Trans. Faraday Soc. 65 (1969) 3324±3327.
[37] B.V. L'vov, Thermochim. Acta, in press.
[38] M.M. Dubinin, O. Kadlec, V. Ponec, Kinet. Kataliz 8 (1967)
292±298.
[39] M. Furtado, G. Jacob, J. Crystal Growth 64 (1983) 257±
267.