On the Thermal Decompositionof Trimethylgallium

A Molecular Beam Sampling Mass SpectroscopyStudy

Q. Chen and P. D. Dapkus
Department of Material Science and Department of Electrical Engineering, University of Southern California,
University Park, California 90089

ABSTRACT
Thermal decomposition of trimethylgaltium in H2, N2, and their admixtures is studied using molecular beam sampling
mass spectrometry. Methane and ethane are found as the primary volatile byproducts of decomposition. The relative
a m o u n t of CH4 increases and C~H6 decreases with an increase in Hz concentration relative to Ne. The activation energy for
the rupture of t h e first Ga-methyl bond is found to be EI~ = 64.6 kcal/mol. The activation energies for the rupture of the
second and third Ga-methyl bonds are estimated to be Elb = 52.6 kcal/mol and EI,~ = 54.1 kcal/mol. Our data reinforce a
radical m e c h a n i s m where TMGa decomposes by way of successively releasing its three methyl radicals.

Despite success in growing device quality III-V semi-
conductor materials by metallorganic chemical vapor dep-
osition (MOCVD), the understanding of the process is still
speculative in nature. Recent interest in atomic layer epi-
taxy (ALE) of GaAs using trimethylgallium (TMGa) and ar-
sine (ASH3) increases the need for accurate information on
the chemical properties of TMGa concerning its stability
in the gas phase and its interaction with the substrate sur-
face. Early work of J a c k o and Price (1) on TMGa thermal
decomposition in toluene showed that only two thirds of
methyl radicals were released from TMGa up to 650~ The
rate constants as a function of temperature for the rupture
of the first and second Ga-methyl bonds were determined
from the evolution of volatile methyl-containing hydrocar-
bon products on the assumption that at high temperatures
(>556~ the input TMGa quickly converted to DMGa*
(Ga(CH3)2) in a small fraction of the total contact time.
They derived, along with the activation energy for the rup-
ture of the second Ga-methyl bond, an extremely low pre-
exponential factor (8.7 • 107 sec -1) for a first-order uni-
molecular reaction that is inconsistent with theoretical ex-
pectations (1) and awaits a closer examination. TMGa ther-
mal decomposition was further studied by infrared
spectroscopy (IR) (2-6), mass spectroscopy (7-11), and la-
ser-induced fluorescence (LIF) (12). F r o m the decrease in
the mole fraction of TMG~i as a function of temperature,
DenBaars et al. were able to determine the activation
energy for the rupture o f the first Ga-methyl bond to be
58-62 kcal/mol. On the contrary, the FTIR data of Maz-
zarese et al. (6) gave an activation energy for this process as
low as 10 kcal/mol. This low value is most likely the mani-
festation of the heterogeneous reaction of TMGa mole-
cules on the stainless steel wall of the apparatus. Other in-
vestigations have been qualitative in general. TMGa
decomposition is generally found to occur appreciably
above about 420~ The major volatile low mass hydrocar-
bons are CH4 and C2H 6. The decomposition is more com-
plete than previously found in Ref. (1). By isotope labeling,
Larsen et al. (11) were able to identify the main source of
CH4 was from the abstraction of atomic H(D) from Hz(D 2)
by the m e t h y l radicals liberated during TMGa decom-
position. Various m e c h a n i s m s are proppsed for the initial
step(s), the detailed path(s), and the role of carrier gases in
the decomposition. However, a consistent picture along
with reliable kinetics data for TMGa gas phase thermal de-
composition are not available.
A m o n g the techniques used in the studies of TMGa gas
phase decomposition, mass spectrometry has the advan-
tages of being direct in species identification, highly sensi-
tive, and reasonably accurate in quantitative measure-
ment. There are doubts, however, whether the small
v o l u m e of sample introduced into the ionization chamber
through a conventional sampling assembly, such as a long
capillary tube, is still representative of the system under
investigation. In this paper, we describe a study of TMGa
thermal decomposition by utilizing a molecular beam
sampling mass spectrometer in conjunction with a plug
flow reactor. A differentially pumped, collision free nozzle
assembly, frequently used in chemical kinetics studies to
monitor reactive intermediates (13), has been employed in
our work. It is fast in response to the change in the gas vol-
u m e under investigation and able to provide high beam in-
tensity with low background contribution to the support-
ing v a c u u m system. The thermal decomposition behavior
of TMGa is examined from room temperature up to 830~
in H2, N2, and their controlled admixtures. The Arrhenius
parameters for the process of the breakage of the first Ga-
methyl bond is found directly from the change in the con-
centration of TMGa as a function of temperature. The acti-
vation energies for the rupture of the second and third Ga-
methyl bonds are also estimated by a two-parameter, least
square fit to the data. A m ech an i sm is proposed to explain
the data.
Experimental Procedure
Figure 1 shows the schematic diagram of the e x p e r i m e n t
setup. The quartz reactor is 5.4 cm id and 100 cm in length
while the length of the temperature uniform zone can be
varied from 67 to 33 cm by using different furnaces. The
large diameter geometry and central positioning of the
sampling assembly assures considerably less contribution
from the reactions at the surface. U n d e r typical operation
conditions, the average contact time is in the range of a few
seconds. The reactor ch am b er is p u m p e d by a mechanical
p u m p to reduce the pressure drop across the first nozzle
for stable molecular beam formation. The v o l u m e between
the two nozzles is turbo-pumped at a pressure of about
5 • 10 -5 Torr under gas load. The nozzles are made of
quartz tube corned at one end and with pin hole through
the thin wall at the tip. The first nozzle and the collimator
(the second nozzle) are critically aligned for concentricity.
When the longer furnace is used, the nozzles are immersed
in the hot zone. This allows a nearly in situ probing of the
gas mixture u n d er investigation. On the other hand, the
nozzle assembly can also be located outside the hot zone
w h e n the shorter furnace is used.
Palladium-purified Hz or scientific-grade Nz is used as
the carrier gas for TMGa. The TMGa bubbler is kept in an
ice bath. All gas flows are controlled by mass flow con-
trollers. Total flow is fine-adjusted by a metering valve at
Tomechanical
pump
ti:~ t~ r From
9, I] ~ < . ' : ^ " . . . . ."." ~ " " " . . ."---~......,^1
. ^ manifold
Fig. 1. Schematic of the molecular beam sampling massspectrometer
attached to a tube flow system.

J. Electrochem. Soc., Vol. 138, No. 9, September 1991 9 The Electrochemical Society, Inc. 2821
Downloaded on 2014-11-09 to IP 129.173.72.87 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).
 2822 J. Electrochem. Soc., Vol. 138, No. 9, September 1991 9 The Electrochemical Society, Inc.

10"~ ,,,i,,,,i,,,, .... I .... ' .... I .... I .... l''''l''''l''''l'''~J i0.4 ......... , ......... , ......... , ......... , ......... , ...... -,r,
T = 24~ 3
T = 720 ~c
V i = 29 V
V i = 29 V
1
10~ ~ , 1 0 "5

g
10~
~ 10 "6

10 n _O10 7

..... .... .... . .... ......... . ....

10"
2o 40 60 80 100 20 40 60 80 100 120
(m/e) m/e

Fig. 2. (a, left) Typical spectrum of TMGa in nitrogen at room temperature. (b, right) typical spectrum of TMGa in nitrogen at 720~

the inlet of the reactor. The pressure of the reactor is meas- gas for some relevant runs. The multiple peak monitoring
ured by a capacitance barometer. The quadrupole mass and auto-ranging features of the computer data acquisition
spectrometer (Model 100-C, UTI) is interfaced to an HP are especially useful in this case to pick up a weak peak ad-
Vectra E5 PC with a S P E C T R A L I N K unit (UTI). The data jacent to a strong peak. The signal amplitudes for all of the
acquisition is thus made simple, quick, accurate, and re- Ga-containing ion species decreased with the same de-
producible. pendence as the temperature was increased. This allows us
Before each run, the reactor tube is cleaned with a mix- to reasonably speculate that the measured Ga-containing
ture of HF and HNO~ (1:9), thoroughly rinsed with DI species all originate from the u n d e c o m p o s e d parent mole-
water, and blown dry with N2. The tube was then heated to cule and that the signals from sampled intermediates are
200~ under v a c u u m for degassing. The molecular beam lower by at least two orders of magnitude. In other words,
sampling system is first tested for its behavior of beam in- the detection of any Ga-containing species as the products
tensity with respect to pressure and mole fraction. Trial of the thermal decomposition is masked by the high back-
runs were also done to calibrate the entire system with H2, ground contributed by the TMGa ionization pattern at
CH4, C2H6, N2, and TMGa in order to obtain a complete nominal ionization voltage of 29 V, (cf. Fig. 2a).
knowledge of the temperature-dependent mass selectivity When the peak intensities of remaining TMGa at differ-
of the sampler. Experiments on TMGa thermal decom- ent temperatures are plotted; Fig. 4 results. It is evident
position are generally carried out by monitoring multiple that appreciable decomposition does not occur until about
peak intensities at constant pressure and fixed TMGa-to- 430~ for TMGa in H2 and 480~ for TMGa in N2. TMGa
H2 ratio with varying temperature or at constant pressure molecules disappear almost completely from the gas
and temperature while varying the TMGa-to-Hz ratio. Peak stream above about 610~ The data as shown in Fig. 4 can
intensities are corrected for the temperature-dependent be analyzed to yield apparent activation energies charac-
mass selectivity and the thermal expansion of the gas. The teristic of the rate at which molecular TMGa disappears.
corrections are made as if the effluent gas is cooled to room The results are listed in Table I for several TMGa-to-H2
temperature. mole ratios, where a k n o w n amount of N2 is used as part or
the whole of the carrier gas.
Experimental Results
The i n v o l v em en t of H2 in the reaction under the condi-
The molecular beam sampling system showed a linear tions used in our e x p e r i m e n t can best be seen from the re-
d e p e n d e n c e of the beam intensity on total pressure over sult of the e x p e r i m e n t where variable admixture of TMGa/
the pressure range of operation (0.01-2 Torr) for a gas mix- H J N 2 is fed into the reactor at T = 600~ and P = 2 Torr.
ture of H2 saturated with TMGa maintained at 0~ when The intensity of the remaining TMGa dropped drastically
the m / e = 99 ion from TMGa is monitored. A slightly differ- when the dose of H2 is increased as shown in Fig. 5. The
ent test is also done for a H2/N2 mixture where a variable change in the relative amount of H2 in the input gas mix-
amount of H2 is introduced into a N2 flow but the total ture also changes the relative amount of CH4 and C~H6 in
pressure remains constant. The detected H2 signal the products as shown in Fig. 6. The advantage of plotting
( m / e = 2) changes linearly with the change in the H2 mole the abscissa in terms of [TMGa]d([TMGa]0 + [H2]0) in Fig. 6
fraction. This linear relationship between the beam inten- will b e c o m e apparent during the discussion, where the
sity and the v o l u m e density o f molecules was derived by [TMGa]0 and [H2]0 a r e t h e initial concentrations of TMGa
Kantrovitz and Grey (14) and refined by Parker e t a l . (15). and H2, respectively. We can define a quantity which is the
Typical mass spectra for TMGa in N2 at room tempera- sum of the moles of CH 4 and two times the moles of C~H6
ture (24~ and the temperature where considerable de- produced from the decomposition of each mole of the
composition occurs ( e . g . , 720~ are shown in Fig. 2a and b. input TMGa, i.e.
One task in using a mass spectrometer to obtain quantita-
tive information is to identify peaks that best represent the
molecular species of the gas phase sampled. Even at low 1O'S E " 1 " " 1 " " 1 ' " ' 1 ' " ' 1 ' " ' 1 ' "
ionization voltage (nominal 29 V measured from the ter- "I. (a)
minals in the instrument console) where further reduction
of the ionization voltage would result in drastic fall off in
the sensitivity for this particular gas analyzer, TMGa pro- 10 ~
duces only a very weak peak at the molecular ion position,
i.e., m / e = 114 and 116. Consequently, the Ga(CHa)2, ions at
m/e = 99 an 101 are monitored as a measure of the remain-
ing TMGa concentration. The low mass hydrocarbons
evolving from the reaction are measured by the changes in
I0~s
4
2

10-7!
the peak intensities at m / e = 16 and 30. They are identified 2
as mainly CH4 and C2H6 according to the spectra obtained
10-8 .i,,,.l.,, i,,,,i Ill, 10~ L J , h , , , h , , JI;,.L|,,,,I uJ
from 0.409% CH4 in H2 and 0.5% C2H6 in H2 standards, 5 10 15 20 25 30 5 10 15 20 25 30
Fig. 3a and b. The molecular ion of C2H6 is monitored in- (m/e) (m/e)
stead of the stronger peak at m / e = 28, because the ioniza-
tion pattern of TMGa at m / e = 28 would interfere with the Fig. 3. (a) The spectrum of 0.409% methane in hydrogen standard
measurement, and also because N2 will be used as carrier and (b) the spectrum of 0.5% ethane in hydrogen standard.
Downloaded on 2014-11-09 to IP 129.173.72.87 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).
 J. Electrochem. Soc., Vol. 138, No. 9, September 1991 9 The Electrochemical Society, Inc. 2823

1.2 1.2e-2

[] [].:....
9 [] t-
zi
[] r-
1.oe-2
9
.N 0.8 in nitrogen E 8.0e-3
G)

O
6.0e-3

0.4 J 4.0e-3
.o B
I- in hydrogen B G)
n
2.0e-3 []
B
El
[] [] []
D []
I
O,0 , I i o.oe+o i I n

300 400 500 600 -0.05 0.05 0.15

T(~ [n2] 0
[H~] o + [N 2]
Fig. 4. Detected TMGa vs. temperature, in hydrogen and in nitrogen.
Fig. 5. Remaining TMGa vs. input hydrogen mole fraction at 600~

[CH4] [C2H6] (20). An electron impact study also gave a value of

PRD(CH~) - - - +2- -
[TMGa]0 [TMGa]0 166.8 kcal/mol for this quantity (21).
The completion of the decomposition is also tested by
This quantity is then a measure of the n u m b e r of moles of putting a few small pieces of Si wafers and quartz plates in
methyl radicals released for each mole of TMGa con- the uniform hot zone of the reactor for several runs. They
sumed, ignoring m u c h smaller contributions from other are taken out for evaluation. Droplets of diameters up to
low hydrocarbons. This is a true measure of the overall de- about 6 ~m can be seen on the surface after decomposition
composition progress, irrespective of the details of how in either H2 or N2. The overall film on the quartz plates is
these products are formed. The quantity PRD(CH3) is plot- electrically conducting. The deposit dissolves slowly in
ted in Fig. 7 as a function of reactor temperature. In princi- HC1 with little residue. On the contrary, the deposit on the
ple, as detailed in the Appendix, the apparent Arrhenius entrance section of the reactor wall, where the rising
parameters for successive release of the three methyl radi- shoulder of the temperature profile occurs, leaves more
cals can be found independently by a multiple parameter residue even after the etch in t:9 mixture of HF and HNO3.
least squares curve fit. The functional behavior of the Ar- This is especially true after decomposition in N~. These ob-
rhenius expressions result in the rate constant being far servations suggest that the decomposition is complete
more sensitive to the changes in the activation energy than with the production of Ga atoms under most conditions
to the changes in the pre-exponential factor. To avoid the used here, that certain reactions involving CHz are more
inaccuracies, we choose to use the rate constant expres- pronounced in N2 ambient (cf. reaction [5] below), and that
sion found from the temperature dependence of the re- polymeric deposits tend to form in or diffuse to the low-
maining reactant (TMGa) as the apparent rate constant for temperature region.
the breakage of the first Ga-methyl bond, k~. We also esti-
mate the Arrhenius pre-exponential factors for the uni- Discussion
molecular reactions of the rupture of the second and third To explain our data, we have proposed the following
Ga-methyl bonds, Alb and A~, to be 4 • 10 '4 and minimal reaction scheme specifically for TMGa decom-
1 • 1014s 1, respectively, by adopting a statistical ther- position in H2 with relatively long contact time so that
m o d y n a m i c s approach as given, for example, by Benson steady-state conditions are likely to be established
(16). The two-parameter least squares fits to the data as TMGa --->DMGa* + CH~ [la]
plotted in Fig. 7 give the activation energies for the reac-
tions of the rupture of the second and the third Ga-methyl DMGa* --> MMGa* + CH* [lb]
bonds as Elb = 52.6 kcal/mol and E~c = 54.1 kcal/mol, re-
spectively. The sum of the squared errors between the fit- MMGa* --~ Ga + CH~ [lc]
ted expression and the experimental data has single sharp
m i n i m u m with respect to Elb and E~c, and the values of E~b
1.0
and E1c are not very sensitive to the changes in A~b and A~c.
It is important to note that the sum of the three activation ]] m [CH4]/3[TMG]
energies thus determined i n d e p e n d e n t ly is 171.3 kcal/mol, 9 [C2H6]/1.5[TMG]
in ag reem en t with the 172.4 kcal/mol thermochemical 0.8
[]
m e a s u r e m e n t of Long (17), the value of 175.7 kcal/mol
summarized by Tuck (18), the value of 183.9 kcal/mol that
is derived according to the compilation of Pilcher and 0.6 rn

Skinner (19), and the value of 176.1 kcal/mol that is calcu-

lated from the compilation of Tirtowidjojo and Pollard 9
B
[]
r~
0.4 m
[]
[]
Table I. Apparent first-order rate constants. m

0.2
(Ink(sec-')=lnA-R~).
j | I I I | = | |
0.0
[TMGa] 0.0 0.2 0.4 0.6 0.8 ,0

[H2] 0.30% 1.71% 8.03% oo [TMGa] o

In A 31.3 31.7 31.7 37.9 [TMGa]o + [tt2] o

(kcal / Fig. 6. Methane and ethane formation vs. TMGa mole fraction at
Ea \-~oI/ 50.9 51.5 54.8 64.6
600~
Downloaded on 2014-11-09 to IP 129.173.72.87 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).
 2824 J. Electrochem. Soc., Vol. 138, No. 9, September 1991 9 The Electrochemical Society, Inc.
i !
9 '
3.0 9 In hydrogen
OO0,
9 B
2.5 " In nitrogen
r []
-$
2.0
B
1.5
_e
r great to make this a favorable path. The molecular orbital
0- 1.0
ILl
0.5 9 la
. 9 . , . ~ . a ~". . . , ~-"
i i
0.0
300 400 500
T (~
Fig 7. Total low mass hydrocarbons in equivalence to the methyls in
each mole of input T M G a
H2 + CH*3 ~ CH4 + H
H + CH*3 ~ CH4
CH3 + CH*3 ~ C2H6
(CH3)xGa + CH*3 ---) (CH3)=-IGaCH*2 + CH4
I low mass hydrocarbon formation during TMGa thermal
--~ ((CH3)~._IGaCH2), (x = 3, 2, 1)
(CH3)=Ga + H--* (CH~)=._,Ga* + CH4 (x = 3, 2, 1) [6]
The decomposition in N= (nonreacting carrier gas) can also
be discussed by ignoring the reactions originating from
the presence of H 2. The reaction [5] is included to account
for the methane formed when TMGa decomposes in N2.
More reactions may be added to account for small amount
of C2H4 and higher mass hydrocarbons ( m / e = 40-44) seen
in the spectra, but they are ignored here because of their
low relative concentrations. This set of reactions has been
proposed previously by us (22) recognizing both the partic-
ipation of hydrogen in subsequent reactions and reaction
[5], which leads to incomplete decomposition. A similar re-
action scheme without reaction [5] can be found in
Ref. (11).
The inclusion of methyl radicals in the intermediates de-
ma nd s a few words of explanation. Butler and co-workers
reported the observation of free CH~ from TMGa thermal
decomposition over a hot susceptor using infrared-diode
laser spectroscopy (5). Although the direct observation of
free methyl radicals by mass spectroscopy was claimed by
the publications of Eltenton (23) from a thermally decom-
posed m i x t u re of tetramethyllead and ethane and Lossing
and Tickner (24) from dimethylmercury, respectively, re-
cent application of the technique is rare. We attempted to
detect the net CH~ signal over the parent background of
TMGa, CH4, and C2H6 using a very low ionization voltage,
nominal 18 V and the lowest reasonable value for this gas
analyzer. After analyzing the ionization efficiency curves
ofTMGa, CH~, and C2H6 obtained with this mass spectrom-
eter, we find that the appearance potential (AP) of CH3 ~ ion
from molecular TMGa is about 12.7 eV, greater by only a
small margin, 2.9 eV, than the ionization potential (IP) of
free CH~ (9.83 eV (25)). This is to be compared to the cor-
responding differences (AP(X) - IP(CH~)) for X = CH4 of
4.6 eV, and X = C2H6 of 4.4 eV. Consequently, there will al-
ways be appreciable contribution to the m / e = 15 signal
from the ionization cracking pattern of molecular TMGa
and, at an appropriate low ionization voltage, only the con-
tribution from TMGa is important compared to the contri-
bution from CH4 and C2H6 generated during the thermal
decomposition. By operating the mass spectrometer at tow
ionization voltage we are able to obtain a net m / e = 15 sig-
nal after subtracting the contribution from the remaining
molecular TMGa. The net m / e = 15 signal starts to appear
Downloaded on 2014-11-09 to IP 129.173.72.87 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).
at the temperature at which decomposition occurs appre-
B~BBm
ciably. It increases with T at first and then decreases after
passing a m a x i m u m , resembling the behavior of CH~ ob-
served by Butler e t a l . (5). Methyl radicals also manifest
themselves in the formation of such observed low mass
hydrocarbon products as CH4 and C2H6. These products
are not likely to be formed directly from TMGa molecules
through bond rearrangement. The additional energy in
distorting the simple, near-planar structure of TMGa is too
calculation of Tsuda e t al. (26) also showed process [la] as
the initial step in TMGa decomposition.
The temperature at which the decomposition becomes
appreciable is related to the absolute sensitivity of the de-
tection techniques and the contact time. In Fig. 4, we com-
pare the decomposition in H2 and N 2 for the same contact
|
time. TMGa decomposition in H2 c o m m e n c e s at lower
600
temperatures than in N2- This shift of the curve to lower
temperatures in Fig. 4 is occasionally disputed as the ef-
fect of the difference in thermal and hydrodynamic prop-
erties of the carrier gases (8). The change in relative
amount of CH4 vs. CzH6 generated for different TMGa/N2/
H2 mixtures as shown in Fig. 6 is, however, a strong indica-
tion of the i n v o l v em en t of H2 in the reactions. This con-
[2] firms the results of Larsen e t al. (11). Here, the data are pre-
senl:ed in terms of [TMGa]0/([TMGa]0 + [H210) since an
[3] analysis by taking only the Eq. [ta]-[4] assumes a neat form
[4] w h e n the [CH4]/3[TMGa]0 and [C2H6]/1.5[TMGa]0 are ex-
pressed as functions of [TMGa]0/([TMGa]0 + [Ha]0). Such
analysis shows that Eq. [la]-[4] describe the trend of the
decomposition in presence of Hz. The direct reaction be-
[5] tween H2 TMGa is ruled out by energy considerations.
However, reaction [2] has a low activation energy around
10 kcal/mol (27-29). It will co m p et e with reactions [3] and
[5] for the supply of CH~ and generate atomic H. The kinet-
ics of attack on (CH3)=Ga by atomic H, reaction [6], are not
known. It is considered as an important path in the de-
struction of atomic H because the activation energy for
radical-molecule reaction is generally low.
At low concentration of H2, the recombination of methyl
radicals to form C2H6 dominates. A small a m o u n t of CH4
forms from the abstraction of H by CH* from Ga contain-
ing hydrocarbons, most likely Ga(CH3)*, i.e., x = 1 in reac-
tion [5] because at this later stage of TMGa decomposition,
the probability that a gallium-containing species will col-
lide with methyl radicals is the highest. A reasonable car-
bon balance supports this point. Since we did not observe
the accumulation of either Ga(CH2)* or CH* species in the
gas phase, we have to assume that the Ga(CHe)* is not vola-
tile and the G a - C H 2 bond is stable. This may account for
the dark deposit on the reactor wall at the entrance of the
hot zone, especially after TMGa decomposition in N2. This
deposit dissolves partially in a mixture of H F and HNO3
with films of polymer-resembling residue. At high H2 con-
centration, reactions [2], [3], and [6] will dominate and CH4
is then the primary low mass hydrocarbon species.
Frequently, a direct test of the effect of carrier gas on
properties of the grown epilayers by conventional
MOCVD hydride growth shows little dependence on car-
rier gas (30, 31). This occurs because AsH3 is a m o r e effec-
tive source of atomic H than H= itself and facilitates the re-
moval of CHs from the surface. It may also be pointed out
that the decomposition behavior discussed here is ob-
tained with moderate contact time. It is expected that with
very short contact time the reactions between interme-
diates and carrier gas will be diminished. The effect of car-
rier gas on TMGa thermal decomposition becomes less
perceptible.
The attack on TMGa by atomic H could explain the vari-
ation of the apparent activation energies with TMGa to H2
ratios. In the e x t r e m e of TMGa thermal decomposition in
N2, only reactions [la], [lb], [lc], [4], and [5] are to be con-
sidered. The apparent rate constant is obviously k~ and
the measured apparent activation energy can be taken as
the activation energy for reaction [la], i.e., E , , = 64.6 kcal/
mol. The determination of first-order rate parameter from
temperature d e p e n d e n c e of the reaction concentrations
 J. Electrochem. Soc., Vol. 138, No. 9, September 1991 9 The Electrochemical Society, Inc. 2825

does not involve any assumption on reaction details in pro- d[TMGa]

ducing products. Neither does it require an absolute cali- - - - k,a[TMGa] [A-1]
bration of the monitoring system. This type of measure- dt
ment and analysis is easier and probably more reliable d[DMGa]
than that from the temperature d e p e n d e n c e of the prod- - - - kl,[TMGa] - k,b[DMGa] [A-2]
ucts. The reported values for the E~ are 59.5 kcal/mol (1) dt
and 58-62 kcal/rnol (4). A molecular orbital calculation (26)
resulted in 64.9 kcaYmol. The agreement is reasonable, d[MMGa]
In Ref. (1), the rate of the reaction [lb], k,6, was assumed dt - klb[DMGa] - klc[MMGa] [A-3]
to be m u c h slower than that of the reaction [ 1a] in the high-
temperature range such that all input TMGa is converted d[CH3]
to DMGa in a small fraction of the contact time. A low acti- - kla[TMGa] + k~b[DMGa] + kk.[MMGa] [A-4]
vation energy for reaction [lb] was d e d u c e d from the data dt
analysis using this assumption. The result that E,b is much
I m p l e m e n t i n g the integrals and noting that [TMGa] =
lower than Ela would generally mean that the rate of the re- [TMGa]0 and [DMGa] = [MMGa] = [CH3] = o at t = 0, where
action [lb] should be m u c h faster than that of reaction the t is the total contact line, we obtain
[la]--contradicting the original assumption--unless reac-
tions [la] and [lb] proceed by a totally different mecha- [CH3]
nism. Not only is this self inconsistency noteworthy, but - - - Y1 + Y2 + Y3 [A-5]
also the absence of Ga(CH3)*2 and Ga(CH3)* species from [TMGa]
various diagnostic techniques would suggest that they are where
in relatively low concentrations. An equilibrium analysis
of MOCVD (20) using the available kinetic data, however, Y, = 1 - e -klat [A-6]
predicts Ga(CH3)* at more than a quarter of the CH4 con-
centration. This high level of Ga(CH~)* has never been ob- Y2 - k,~ (1-e kla
-k'~t) - - ( 1 - e k,bt) [A-7]
served experimentally. Our values of E~,, E~b, and E~r are k~b - k~ k,b - kl~
consistent with the picture that TMGa decomposes by re-
leasing successively its three methyl radicals along the and
whole length of the hot zone and the intermediate species
Ga(CH3)* 2 and Ga(CH3)* do not survive for long in the envi- Y3 - klbklc
(1 -- e -klat)
r o n m e n t that has caused the rupture of the first Ga-methyl ( k l b -- k,a)(k,c - lkla)
bond. The stabilities of Ga(CH3)*2 and Ga(CH3)* are differ-
ent only by a small margin in favor of the last Ga-methyl klaklb
bond. Finally, the presence of atomic Ga in the final prod- (1 - e -klct)
ucts also points out the complete breakdown of TMGa (klb - kla)(klc - k,a )
above about 610~
(1 -- e klbt)
Conclusions (klb -- kl~)(klc - klb)
An intensive study has been carried out of thermal de-
composition of TMGa in H2, N2, and their admixtures by klaklb
~{ (1 -- e -kld) [A-8]
using molecular beam sampling mass spectrometry. Meth- (klb -- kla)(klc - klb)
ane and ethane are found as the primary volatile hydrocar-
bon products. At high enough temperatures, the relative account for the equivalent methyls from reactions [la],
amounts of CH4 and C~H6 formed in the gas phase are af- [lb], and [lc], correspondingly.
fected by the presence of H2. The amount of CH4 increases
the C2H6 decreases as the mole fraction of Ha increases ion The estimation o f the pre-exponential f a c t o r s . - - W i t h i n
the frame of the transition state theory of chemical kinet-
a TMGa/HJN2 mixture. The measured apparent activation ics, the unimolecular Arrhenius pre-exponential parame-
energies from the temperature d e p e n d e n c e of the remain- ter is related to the entropy change frc~m the reacting mole-
ing reactant differ for decompositions occurring in H2 and cules to their transition states, hS$. A statistical thermo-
N2. This difference is explained qualitatively by the parti- dynamics approach can be used to estimate the entropy
cipation of H2 in the reaction specifically by the attack of change. We consider only the rotational and vibrational
TMGa by atomic H. The apparent activation energy ob- contributions. Actually, the vibrational contribution is al-
tained from the t em p e r a tu r e dependence of remaining re- ready small. The structure of TMGa is planar and those of
actant for TMGa in N2 is taken as the activation energy for DMGa* and MMGa* are taken to be linear. Their corres-
the process of rupturing the first Ga-methyl bond, i.e., ponding transition states are taken to more resemble those
EI~ = 64.6 kcal/mol. The activation energies for the reac- of the expected products. The energized bond is assumed
tions of breaking the second and the third Ga-methyl bond to elongate to about 2.8 times the original length (16). The
vibration modes concerning the Ga--C bonds are, accord-
are also estimated independently to be E~b = 52.6 kcal/mol
ing to Tuck (18), ]}Ga-Me.stretch= 521 cm -l (sym), 570 cm -I
and E,r = 54.1 kcal/mol from the temperature dependence
(asym), VGa_C3.deform = 163cm -~, and P G a - M e . r o c k = 612, 727,
of the volatile hydrocarbon products. Our data reinforce a 769 cm -1.
radical m e c h a n i s m where TMGa decomposes by way of re-
leasing successively its three methyl radicals. The rupture REFERENCES
of the first Ga-methyl bond is, by far, the rate-limiting step. 1. M. G. Jacko and S. J. W. Price, Can. J. Chem., 141, 1560
(1963).
Acknowledgments 2. M. R. Leys and H. Veenvliet, J. Cryst. Growth, 55, 145
(1981).
This work was partially supported by the Office of Naval 3. J. Nishizawa and T. Kurabayashi, This Journal, 132,
Research, Grant No. N00014-89-J-1229, and the Solar 413 (1983).
Energy Research Institute, Subcontract No. XB-5-05009-3. 4. S. P. DenBaars, B. Y. Maa, P. D. Dapkus, A. D. Danner,
and H. C. Lee, J. Cryst. Growth, 77, 188 (1986).
Manuscript submitted April 27, 1990; revised manu- 5. J. E. Butler, N. Bottka, R. S. Sillmon, and D. K. Gas-
script received April 5, 1991. kill, ibid., 77, 163 (1986).
6. D. Mazzarese, A. Tripathi, and W. C. Conner, J. Elec-
APPENDIX A tron. Mater., 18(3), 369 (1986).
7. M. Yoshida, H. Watanabee, and A. Uesugi, This Jour-
Activation Energies from Products nat, 132, 677 (1985).
The f o r m u l a t i o n . - - T h e three methyls in a TMGa mole- 8. P. W. Lee, T. R. Omstead, D. R. Mckenna, and K.F.
cule, if released, will s o m e h o w form CH4, C2H6, and C2H4. Jensen, J. Cryst. Growth, 85, 165 (1987).
They can readily be detected. Therefore, we consider reac- 9. N. Putz, H. Heinecke, M. Hegen, and P. Balk, J. Cryst.
t i o n s [la], [lb], and [lc] only. The rate equations are Growth, 74, 292 (1986).
Downloaded on 2014-11-09 to IP 129.173.72.87 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).
 2826 J. Electrochem. Soc., Vol. 138, No. 9, September 199t 9 The Electrochemical Society, Inc.

10. M. Mashita, S. Horiguchi, M. Shimizu, K. Kamon, M.
Mihara, and M. Ishii, ibid., 77, 194 (1986).
11. C. A. Larsen, N.I. Buchan, and G.B. Stringfellow,
Appl. Phys. Lett., 52(6), 480 (1988). This work is re-
viewed and a more complete compilation of the
work of this group is presented in "Organometallic
Vapor Phase Epitaxy," by G. B. Stringfellow, Aca-
demic Press, San Diego CA (1989).
12. H. Suzuki, K. Mori, M. Kawasaki, and H. Sato, J. Appl.
Phys., 64(1), 371 (1988).
13. M. A. A. C]yne and W. S. Nip, in "Reactive Interme-
diates in Gas Phase: Generation and Monitoring,"
D.W. Setser, Editor, p. 38, Academic Press, New
York (1979).
14. A. Kantrowitz and J. Grey, Rev. Sci. Instrum. 22(5), 328
(1951).
15. H. Parker, A. R. Kuhthau, R. Zapata, and J. E. Scott,
Jr., in "Rarefield Gas Dynamics," F.M. Devienne,
Editor, p. 69, Pergamon Press, New York (1958).
16. S.W. Benson, "Thermochemical Kinetics: Methods for
Estimation of Thermochemical Data and Rate Pa-
rameters," 2nd ed., J o h n Wiley and Sons, New York
(1976).
17. L. H. Long, Pure Appl. Chem., 2, 61 (1961).
18. D. G. Tuck, in "Comprehensive Organometallic Chem-
istry," Vol. 1, G. Wilkinson, F. G. A. Stone, and E. W.
Abel, Editors, p. 683, Pergamon Press, New York
(1982).
19. G. Pilcher and H. A. Skinner in "Chemistry of Metal-
Carbon Bond," F. R. Hartley and S. Patai, Editors,
Chap. 2, p. 43, J. Wiley, New York (1982).
20. M. Tirtowidjojo and R. Pollard, J. Cryst. Growth, 77,
200 (1986).
21. F. Glockling and R.G. Stratford, J. Chem. Soc., (A),
1761 (1971).
22. (a) P. D. Dapkus, S. P. DenBarrs, Q. Chen, and B. Y.
Maa, in "Mechanisms of Reactions of Organometal-
lic Compounds with Surfaces," D. J. Cole-Hamilton
and J.O. Williams, Editors, p. 257, NATO ASI
Series B: Physics Vol. 198, P l e n u m Press, New York
(1989). (b) Proceedings of NATO Advanced Research
Workshop on the Mechanisms of Reactions of Or-
ganometallic Compounds with Surfaces, J u n e 22-24,
1988, in St. Andrews, Fife, Scotland.
23. G. C. Eltenton, J. Chem. Phys., 15(7), 455 (1947).
24. F. P. Lossing and A. W. Tickner, J. Chem. Phys., 20, 907
(1952).
25. "Handbook of Chemistry and Physics," 68th ed., R. C.
Weast, Editor, CRC Press, Boca Raton, FL (1987-
1988).
26. M. Tsuda, S. Oikawa, M. Morishita, and M. Mashita,
Jpn. J. Appl. Phys., 26, L564 (1987).
27. G. L. Pratt, "Gas Kinetics," p. 159, J o h n Wiley and
Sons, New York (1969).
28. P. C. Kobrinsky and P. D. Pacey, Can. J. Chem., 52,
3655 (1974).
29. G. L. Pratt and D. Roger, J. Chem. Soc., Faraday
Trans. 1, 12, 2769 (1976).
30. T. F. Kuech and E. Veuhoff, J. Cryst. Growth, 68, 148
(1984).
31. G. Arens, H. Heineke, N. Putz, L. Luth, and P. Balk, J.
Cryst. Growth, 76, 305 (1986).

Ellipsometric Thickness Measurement of Combined a-Si and Ti

Films on c-Si
J. C. Jans
Philips Research Laboratories, 5600 JA Eindhoven, The Netherlands

C. S. Yoo
Taiwan Semiconductor Manufacturing Company, Science Based Industrial Park, Hsin-Chu, Taiwan, China

ABSTRACT
Experimental and theoretical results are reported on the use of monochromatic ellipsometry to measure the thickness
of amorphous silicon (a-Si) and titanium (Ti) overlayers and of combined a-Si/Ti overlayers on crystalline silicon (c-Si) sub-
strates. Accurate thickness monitoring of these overlayer structures is very important for the formation of titanium sili-
cide straps to be used as an interconnect in integrated circuit technology. The presence of an intermediate Ti film between
the a-Si and the c-Si is shown to considerably enhance the ellipsometric sensitivity to variations in a-Si thickness.

As an integrated circuit technology step, a titanium sili-
cide (TiSi2) interconnect strap is formed after the deposi-
tion of amorphous silicon (a-Si) on titanium (Ti) via appro-
priate sample treatment. To achieve this the a-Si layer is
first etched to form the local interconnect pattern and sub-
sequently it is annealed in a nitrogen a m b i e n t to simulta-
neously form titanium silicide (in a reaction of a-Si with Ti)
and titanium nitride (TIN). This is followed by selective
etching of the TiN. In this process, the thickness control of
the a-Si layer is of critical importance, since too thick a
layer results in silicide overgrown with a-Si, while too thin
a layer results in silicon-suckout from the substrate. To
avoid these problems the thickness of the films should be
periodically checked. This can be done with x-ray fluo-
rescence spectroscopy (XRF), which in turn is calibrated
with Rutherford backscattering spectroscopy (RBS). A
major disadvantage of this procedure is the fact that it is a
rather time-consuming off-line check.
In this paper we show that monochromatic ellipsometry
can provide an attractive alternative as it is noncontact,
nondestructive, and can be used in-line. However, with
commercially available monochromatic ellipsometers
(1, 2), interpretation can be a problem as most software op-
Downloaded on 2014-11-09 to IP 129.173.72.87 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).
erates as a "black box." No checks at all are possible on the
accuracy or even on the correctness of the results ob-
tained, while the program code is usually also inaccessible.
For this reason, ellipsometric analysis was carried out with
home-made analysis software basically very similar to that
already illustrated by McCrackin (3).
The principles of ellipsometry are extensively discussed
in literature (4-8). Upon oblique reflection from a planar in-
terface, both phase and amplitude of an incident, linearly
polarized collimated beam of light are altered. Ellip-
sometry measures the relative changes in phase and ampli-
tude u p o n reflection of p- and s-polarized light, for which
the electric field vectors are in and normal to the plane of
incidence, respectively. A complex reflection ratio p (4) can
be defined which relates ellipsometric parameters ~ and h
to an optical model for the measured specimen using the
complex Fresnel reflection coefficients ~ and ~
P = ~,/~5 = (tr, Hr~l) exp i(sp - ~) = tan ~ exp (ih) [1]
Here tan 9 refers to the relative amplitude attenuation,
while A gives the relative phase change for the p- and s-
polarized light waves.