Chemical vapor deposition of TiN from tetrakis(dimethylamido)titanium and ammonia: Kinetics and mechanistic studies of the gas-phase chemistry

Article abstract of DOI:10.1021/ja953468o

The gas-phase kinetics of the reaction of tetrakis(dimethylamido)titanium (Ti(NMe₂)₄) with NH₃ have been measured using a flow tube reactor and FTIR spectrometer. Ti(NMe₂)₄ reacts rapidly with NH₃ in a transamination reaction to form HNMe₂ as a direct product. The bimolecular rate constant for the reaction of Ti-(NMe₂)₄ with NH₃ at 24 °C is k = (1.2 ± 0.2) x 10^-16 cm³ molecules^-1 s^-1. A primary kinetic isotope effect of k(H)/k(D) = 2.6 ± 0.4 is observed with ND₃ indicating that cleavage of an N-H bond is the rate limiting step. Therefore the rate constant is assigned to the initial transamination reaction with NH₃. The temperature dependence of the rate constant gives activation parameters of log(A) = -10.0 ± 0.2 (ΔS⁺ = -19 cal/(mol K)) and E(a) = 8.1 ± 0.1 kcal/mol (ΔH⁺ = 6.9 ± 0.1 kcal/mol). When excess HNMe₂ is added to the gas flow, the reaction rate is strongly suppressed. This is evidence for a reversible initial transamination reaction: Ti(NMe₂)₄ + NH₃ ? (Me₂N)₃Ti-NH₂ + HNMe₂. The proposed mechanism for subsequent reaction is elimination of HNMe₂:(Me₂N)₃Ti-NH₂ → (Me₂N)₂-Ti=NH + HNMe₂. From the dependence of the observed rate constant on HNMe₂, the branching ratio is obtained for the above elimination reaction versus reaction with HNMe₂: (Me₂N)₃Ti-NH₂ + HNMe₂ → Ti(NMe₂)₄ + NH₃. The relevance of these results to the chemical vapor deposition of TiN by this chemistry is discussed. The gas-phase kinetics of the reaction of tetrakis(dimethylamido)titanium (Ti(NMe₂)₄) with NH₃ have been measured using a flow tube reactor and FTIR spectrometer. Ti(NMe₂)₄ reacts rapidly with NH₃ in a transamination reaction to form HNMe₂ as a direct product. The bimolecular rate constant for the reaction Ti-(NME₂)₄ with NH₃ at 24°C is k = (1.2±0.2) × 10^-16 cm³ molecules^-1 s^-1. A primary kinetic isotope effect of k_Hk_D = 2.6±0.4 is observed with ND₃ indicating that cleavage of an N-H bond is the rate limiting step. Therefore the rate constant is assigned to the initial transmission reaction with NH₃. The temperature dependence of the rate constant gives the activation parameters. When excess HNME₂ is added to the gas flow, the reaction rate is strongly suppressed. This is the evidence of a reversible initial transamination reaction. The proposed reaction for the subsequent reaction is given. From the dependence of the observed rate constant on HNMe₂, the branching ratio is obtained for the above elimination reaction versus reaction with HNMe₂. The relevance of these results to the chemical vapor deposition of TiN by this chemistry is discussed.

Full text of DOI:10.1021/ja953468o

J. Am. Chem. Soc. 1996, 118, 4975-4983
4975
Chemical Vapor Deposition of TiN from
Tetrakis(dimethylamido)titanium and Ammonia: Kinetics and
Mechanistic Studies of the Gas-Phase Chemistry
Bruce H. Weiller
Contribution from The Aerospace Corporation, Mechanics and Materials Technology Center,
P.O. Box 92957/M5-753, Los Angeles, California 90009-2957
ReceiVed October 16, 1995. ReVised Manuscript ReceiVed April 3, 1996X
Abstract: The gas-phase kinetics of the reaction of tetrakis(dimethylamido)titanium (Ti(NMe2)4) with NH3 have
been measured using a flow tube reactor and FTIR spectrometer. Ti(NMe2)4 reacts rapidly with NH3 in a
transamination reaction to form HNMe2 as a direct product. The bimolecular rate constant for the reaction of Ti-
(NMe2)4 with NH3 at 24 °C is k ) (1.2 ( 0.2) × 10-16 cm3 molecules-1 s-1. A primary kinetic isotope effect of
kH/kD ) 2.6 ( 0.4 is observed with ND3 indicating that cleavage of an N-H bond is the rate limiting step. Therefore
the rate constant is assigned to the initial transamination reaction with NH3. The temperature dependence of the rate
constant gives activation parameters of log(A ) ) -10.0 ( 0.2 (∆S† ) -19 cal/(mol K)) and Ea ) 8.1 ( 0.1
kcal/mol (∆H† ) 6.9 ( 0.1 kcal/mol). When excess HNMe2 is added to the gas flow, the reaction rate is strongly
suppressed. This is evidence for a reversible initial transamination reaction: Ti(NMe2)4 + NH3 a (Me2N)3Ti-NH2
+ HNMe2. The proposed mechanism for subsequent reaction is elimination of HNMe2: (Me2N)3Ti-NH2 f (Me2N)2-
TidNH + HNMe2. From the dependence of the observed rate constant on HNMe2, the branching ratio is obtained
for the above elimination reaction versus reaction with HNMe2: (Me2N)3Ti-NH2 + HNMe2 f Ti(NMe2)4 + NH3.
The relevance of these results to the chemical vapor deposition of TiN by this chemistry is discussed.
Introduction
superior in this regard but the temperature used to deposit TiN
must be below ∼350 °C to be compatible with mutilevel
metalization schemes.
Titanium nitride (TiN) has many useful properties including
high hardness, good electrical conductivity, high melting point,
and chemical inertness.1 The applications have included wear-
resistant coatings on machine tools and bearings, thermal control
coatings for windows,2 and erosion resistant coatings for
spacecraft plasma probes.3 However, an important new ap-
plication of TiN is as a diffusion barrier layer for metalization
in integrated circuits.
In the fabrication of integrated circuits, electrical contacts
are formed at interfaces between metal and semiconductor
layers. These interfaces are often not stable due to the diffusion
of the metal into the semiconductor and can lead to device
failure. A common solution to this problem is a diffusion
barrier, a thin sandwich layer that is electrically conductive and
blocks diffusion of the metal. TiN is one of the best diffusion
barrier materials for silicon integrated circuits due to slow
diffusion rates and good electrical conductivity.4
Future generations of integrated circuits will require much
higher integration densities with feature sizes of less than 350
nm.5 In order to achieve these goals, highly conformal coatings
of TiN are needed on the high-aspect-ratio nanostructures found
in integrated circuits. Although TiN is easily deposited by
sputtering, this line-of-site technique cannot provide the step
coverage required.6 Chemical vapor deposition (CVD) is
Two paths have been demonstrated for the CVD of TiN for
microelectronics: (a) titanium tetrachloride (TiCl4) and ammonia
(NH3)2 and (b) tetrakis(dialkylamido)titanium (Ti(NR2)4) and
NH3.7 Although the chloride reaction gives excellent step
coverage,8 the temperature requirement is too high for this
application and Cl contamination is a problem as well. The
step coverage is not as good from the amido reaction but it
gives TiN at the required temperature. Therefore, improving
the step coverage of this process is highly desirable and there
has been much recent effort toward this goal.9-12
CVD is a complex chemical process in which both gas-phase
and surface reactions play a role. It has been shown that gas-
phase chemistry is particularly important in this system. When
(6) The term step coverage is used to describe the conformality of a
coating over surface microstructures such as steps, trenches, or vias.
(7) (a) Fix, R. M.; Gordon, R. G.; Hoffman, D. M. Mater. Res. Soc.
Symp. Proc. 1990, 168, 357-362. (b) Fix, R. M.; Gordon, R. G.; Hoffman,
D. M. J. Am. Chem. Soc. 1990, 112, 7833-7835. (c) Fix, R. M.; Gordon,
R. G.; Hoffman, D. M. Chem. Mater. 1991, 3, 1138-1148.
(8) Yokoyama, N.; Hinode, K.; Homma, Y. J. Electrochem. Soc. 1991,
138, 190-195.
(9) (a) Ishihara, K.; Yamazaki, K.; Hamada, H.; Kamisako, K.; Tarui,
Y. Jpn. J. Appl. Phys. 1990, 29, 2103-2105. (b) Sandhu, G. S.; Meikle, S.
G.; Doan, T. T. Appl. Phys. Lett. 1993, 62, 240. (c) Intemann, A.; Koerner,
H.; Koch, F. J. Electrochem. Soc. 1993, 140, 3215-3222.
(10) (a) Katz, A.; Feingold, A.; Nakahara, S.; Pearton, S. J.; Lane, E.;
Geva, M.; Stevie, F. A.; Jones, K. J. Appl. Phys. 1992, 71, 993-1000. (b)
Sun, S. C.; Tsai, M. H. Thin Solid Films 1994, 253, 440-444.
(11) (a) Dubois, L. H.; Zegarski, B. R.; Girolami, G. S. J. Electrochem.
Soc. 1992, 139, 3603-3609. (b) Prybyla, J. A.; Chiang, C.-M.; Dubois, L.
H. J. Electrochem. Soc. 1993, 140, 2695. (c) Dubois, L. H. Polyhedron
1994, 13, 1329-1336.
(12) (a) Raaijmakers, I. J.; Vrtis, R. N.; Yang, J.; Ramaswami, S.;
Lagendijk, A.; Roberts, D. A.; Broadbent, E. K. Mater. Res. Soc. Symp.
Proc. 1992, 260, 99-105. (b) Raijmakers, I. J. Thin Solid Films 1994, 247,
85-93. (c) Raaijmakers, I. J.; Yang, J. Appl. Surf. Sci. 1993, 73, 31-41.
X Abstract published in AdVance ACS Abstracts, May 1, 1996.
(1) Toth, L. E. Refractory Materials; Academic Press: New York, 1971;
Vol. 7, Transition Metal Carbides and Nitrides.
(2) Kurtz, S. R.; Gordon, R. G. Thin Solid Films 1986, 140, 277-290
and references therein.
(3) (a) Wahlstrom, M. K.; Johansson, E.; Veszelei, E.; Bennich, P.;
Olsson, M.; Hogmark, S. Thin Solid Films 1992, 220, 315-320. (b)
Veszelei, M.; Veszelei, E. Thin Solid Films 1993, 236, 46-50.
(4) (a) Wittmer, M.; Melchor, H. Thin Solid Films 1982, 93, 397-405.
(b) Nowicki, R. S.; Nicolet, M. A. Thin Solid Films 1982, 96, 317-326.
(5) Solid State Technol. 1995, 38 (2), 42-50.
S0002-7863(95)03468-8 CCC: $12.00 © 1996 American Chemical Society

4976 J. Am. Chem. Soc., Vol. 118, No. 21, 1996
Weiller
Ti(NR2)4 is used alone to deposit TiN,7,13 the resulting material
is severely contaminated with carbon. In order to produce
low-carbon, low-resistivity TiN from this precursor, reac-
tion with NH3 is required.7,11 Gas-phase reaction between
Ti(NMe2)4 and NH3 occurs rapidly even at 25 °C to produce
HNMe2.11 This is consistent with studies of the solution
chemistry of these compounds that showed transamination reac-
tions between Ti(NR2)4 and HNR′2 occur readily to form HNR2
and Ti(NR2)4-n(NR′2)n as shown below.14
Ti(NR2)4 + nHNR′2 f Ti(NR2)4-n(NR′2)n + nHNR2
(1)
Isotopic labeling studies of the CVD system using 15NH3 and
ND3 have demonstrated the formation of Ti15N and the use of
ND3 gives DNMe2.11 These observations are strong support
for a gas-phase transamination reaction and also explain the
need for NH3 to produce low-carbon films. Reaction with NH3
effectively removes HNMe2 from the precursor thereby reducing
the carbon content of TiN presumably through the formation
of a reactive intermediate that decomposes on the surface.
Surface reaction in the absence of gas-phase reactions (<10-5
Torr) has been shown to give low-carbon TiN from Ti(NMe2)4
and NH3, apparently from a transamination reaction on the
surface.15 However, this is quite different from the high pressure
conditions (>10 Torr) found in the CVD process.
There is some evidence that, in addition to determining the
purity of TiN, the chemistry of the process also affects its step
coverage, resistivity, and morphology. It has been reported that
Ti(NEt2)4 gives higher quality TiN than Ti(NMe2)4.12 Indeed,
these workers found that the TiN produced with Ti(NMe2)4 was
unacceptable with high, unstable resistivity, poor step coverage,
and morphology. Apparently, the rate of reaction of Ti(NMe2)4
with NH3 is too fast and leads to the formation of intermediates
with high sticking coefficients or low surface mobility.16 The
reaction of Ti(NEt2)4 with NH3 is much slower and is the likely
reason for the improved film properties. This is particularly
intriguing because it indicates that, by controlling the reaction
rates, it may be possible to control the step coverage and other
properties of TiN.
The work described herein focuses on the gas-phase kinetics
of the reaction of Ti(NMe2)4 with NH3. We are particularly
interested in the relationship between gas-phase chemistry and
the properties of materials produced by CVD. The primary goal
of this work is to obtain a better fundamental understanding of
this reaction that is critical to the successful production of TiN
films by CVD. An additional goal is to obtain accurate kinetics
data that can be used in quantitative numerical models to aid
the design of optimized CVD reactors. Below data are presented
which satisfy both of these objectives, some of which have been
published in preliminary form.17 Accurate values for the rate
constant for reaction of Ti(NMe2)4 with NH3 are presented
including the room temperature isotope effect with ND3 as well
as the temperature dependence. The rate constant is not affected
Figure 1. Flow-tube reactor (FTR) interfaced to the FTIR spectrometer.
by mass transport effects, wall reactions, or total pressure.
However, the reaction rate with NH3 is inhibited by the addition
of excess HNMe2. The proposed mechanism is the reversible
initial transamination reaction, Ti(NMe2)4 + NH3 a (Me2N)3-
Ti-NH2 + HNMe2, followed by elimination of HNMe2 from
(Me2N)3Ti-NH2 to form (Me2N)2TidNH. From the depen-
dence of the observed rate constants on HNMe2, the branching
ratio for reaction of (Me2N)3Ti-NH2 with HNMe2 versus
elimination of HNMe2 is obtained. In addition to providing
important mechanistic insight, the use of amines may provide
a simple method to control the reaction of amido precursors
with NH3 to improve the properties of TiN and related nitride
materials.
Experimental Section
The experimental apparatus used to examine the gas-phase kinetics
is a flow tube reactor shown in Figure 1. It is a 1-m-long, 1.37-in.-i.d.
Teflon-coated stainless-steel tube equipped with a sliding injector port
that provides a variable distance from the focus of the IR beam. The
injector is a 1/4-in. tube ending in a Pyrex loop with many equally spaced
holes for gas injection counter-current to the main flow for good mixing.
The observation region is a standard cross (NW-40) equipped with
purged windows, purged capacitance manometers, and a throttle valve
controller to maintain constant pressure. A tubular Teflon insert with
0.75-in.-diameter holes aids in separating the reactive flows from the
IR windows. Argon is used as the purge gas to reduce diffusion and
the build up of deposits on the windows. Both the flow tube and the
observation region are wrapped with heat tape for elevated temperature
operation when desired. The IR beam from the FTIR spectrometer
(Nicolet 800) is focused in the middle of the observation region using
a combination of flat and off-axis parabolic mirrors. A detector module
with focusing mirror is mounted on the opposite side of the flow tube.
The focusing optics and the detector module are enclosed in Plexiglas
boxes and the entire beam path is purged with dehumidified, CO2-free
air. Mass flow meters measure the separate flows of buffer (Ar or
He), carrier (Ar or He), and purge (Ar) gases, NH3, and HNMe2. Dilute
mixtures of NH3 (9.77%) and HNMe2 (16.6%) in He were used. The
flow meters were calibrated by monitoring the pressure rise in a known
volume as a function of time. The buffer gas, NH3, and HNMe2 flows
were mixed and fed into the side arm of the flow tube. A mixture of
Ti(NMe2)4 in buffer gas, generated from a glass bubbler at ambient
temperature, flowed through the sliding injector. A mechanical pump
(Sergeant Welch 1397) was equipped with a liquid nitrogen trap for
pumping. The Reynolds numbers for these experiments are typically
around 10, well within the laminar flow regime.18
(13) Sugiyama, K.; Pac, S.; Takahashi, Y.; Motojima, S. J. Electrochem.
Soc. 1975, 122, 1545-1549.
(14) (a) Bradley, D. C.; Gitlitz, M. H. J. Chem. Soc. (A) 1969, 980-
984. (b) Bradley, D. C.; Torrible, E. G. Can. J. Chem. 1963, 41, 134-138.
(c) Bradley, D. C.; Thomas, I. M. J. Chem. Soc. 1960, 3857-3861.
(15) Truong, C. M.; Chen, P. J.; Corneille, J. S.; Oh, W. S.; Goodman,
D. W. J. Phys. Chem. 1995, 99, 8831-8842.
(16) (a) Hsieh, J. J. J. Vac. Sci. Technol. A 1993, 11, 78-86. (b) Rey,
J. C.; Cheng, L.-Y.; McVettie, J. P.; Saraswat, K. C. J. Vac. Sci. Tech.
1991, A9, 1083.
(17) (a) Weiller, B. H. Mater. Res. Soc. Symp. Proc. 1993, 282, 605-
610. (b) Weiller, B. H.; Partido, B. V. Chem. Mater. 1994, 6, 260-261.
(c) Weiller, B. H. Mater. Res. Soc. Symp. Proc. 1994, 335, 159-164. (d)
Weiller, B. H. Mater. Res. Soc. Symp. Proc. 1994, 334, 379-384. (e)
Weiller, B. H. Chem. Mater. 1995, 7, 1609-1611.
For the O3 experiment, a mixture of He and O2 flowed into an ozone
generator and then into the sliding injector of the flow tube. (Warning:
Condensed ozone is an explosion hazard, do not use a liquid nitrogen
trap with ozone.) We routinely generated a mixture of 1.4% O3 in O2
and He. This was confirmed by IR absorption using the integrated
absorbance of the 1042-cm-1 band and the accepted integrated band
(18) Welty, J. R.; Wicks, C. E.; Wilson, R. E. Fundamentals of
Momentum, Heat and Mass Transfer; John Wiley & Sons: New York, 1984.

Chemical Vapor Deposition of TiN from Ti(NMe2)4 + NH3
J. Am. Chem. Soc., Vol. 118, No. 21, 1996 4977
Figure 2. Plot of ln(A/A0) vs time for the reaction of O3 with isobutene.
The squares and circles are for 1.15 and 2.32 Torr of isobutene,
respectively.
Figure 3. Plot of the observed decay constants (kobs) vs isobutene
partial pressure for the reaction of O3 with isobutene. The total pressure
is indicated in the legend.
intensity.19 Isobutene flowed into the side arm of the flow tube along
with the buffer gas.
After an extended number of runs with Ti(NMe2)4 and NH3, a tan
deposit formed on the walls of the flow tube reactor. A similar deposit
was noted in earlier work and an IR spectrum was reported.11a In order
to remove this deposit and obtain a clean Teflon surface, a fluorine
plasma was used to clean the flow tube ex situ. This procedure was
very effective in removing the deposit and resulted in either a clean
surface or a powdery white residue that could be wiped off easily.
Presumably this white substance is TiFx, which has been noted to
sublime readily at elevated temperatures.20
The following chemicals were used as received from the indicated
suppliers: Ti(NMe2)4 (Schumacher); Ti(NEt2)4 (Schumacher); Ar
(Spectra Gases, UHP grade); NH3 (Matheson, electronic grade), ND3
(Cambridge Isotope Labs), He (Spectra Gases, UHP grade), O2 (Airco),
isobutene, (Aldrich), NF3 (Matheson, electronic grade), and HNMe2
(Aldrich). The spectrometer was operated at 8-cm-1 resolution and
256 scans were averaged unless noted. For kinetics measurements
integrated intensities were used.
Results
Figure 4. IR spectra of the reaction of Ti(NMe2)4 with NH3 in the
static cell: (A) initial spectrum of Ti(NMe2)4 alone, (B) ∼7 s after
addition of NH3, (C) after complete reaction, and (D) reference spectrum
for HNMe2. The dashed lines indicate major features of Ti(NMe2)4 and
HNMe2.
A. Kinetics of a Known Reaction: O3 + Isobutene. The
decay of O3 was measured as a function of reaction time as
shown in Figure 2. The IR absorbance of O3 was monitored
by the ν3 band at 1042 cm-1 19
The isobutene pressure ranged
.
from 0.5 to 2.3 Torr and is in far excess over the [O3]. Therefore
pseudo-first-order conditions apply and we expect an exponential
decay of O3: [O3] ) [O3]0 exp(-kbi[C4H8]t), where kbi is the
bimolecular rate constant. Furthermore, we expect a linear
relationship between the logarithm of the IR absorbance and
the reaction time: ln(A/A0) ) -kobst, kobs ) kbi[C4H8]. Here A
is the integrated absorbance of the O3 band, A0 is the integrated
absorbance of the O3 band in the absence of isobutene (average
of initial and final values), kobs is the observed decay constant,
and kbi is the bimolecular rate constant. Figure 2 shows such
a linear dependence for two isobutene pressures. The slopes
of the lines in Figure 2 (kobs) should show a linear dependence
on isobutene partial pressure. Figure 3 shows this expected
result and provides the bimolecular rate constant, kbi ) (13.8
( 0.1) × 10-16 molecules-1 cm3 s-1 at 25 °C and 10 Torr total
pressure. Figure 4 also shows that when the total pressure is
changed to 5 Torr, there is no significant effect on the rate
constant.
reacted with NH3 in a static cell at room temperature. The top
spectrum shows Ti(NMe2)4 alone, the next spectrum is just after
(<7 s) the addition of 27.6 Torr of a mixture of 13.8% NH3 in
He, and the third spectrum is after no further changes were
observed. At the bottom is a reference spectrum for a sample
of dimethylamine. From this we can clearly see that dimeth-
ylamine is formed from the reaction of Ti(NMe2)4 with NH3 as
observed by Dubois et al.11 The reaction between Ti(NMe2)4
and NH3 is too fast to be measured in the static cell and the
flow tube reactor is needed for this.
Figure 5 shows the IR spectra obtained when Ti(NMe2)4 is
reacted with NH3 in the flow tube reactor. The top spectrum
(A) is in the absence of NH3. The relatively intense NC2
symmetric stretch at 950 cm-1 is a good signature for Ti-
(NMe2)4,21 and we use it to monitor the number density of Ti-
(NMe2)4. The middle four spectra result from the addition of
0.245 Torr of NH3 and increasing the distance between the
injector and observation region from 10 to 80 cm at constant
NH3 pressure. The flow rate was 170 sccm giving the reaction
times listed in the caption. In spectra B through E, we see NH3
bands at 968 and 932 cm-1 (ν2) and at 1628 cm-1 (ν4).22 These
B. Room Temperature Kinetics of Ti(NMe2)4 + NH3.
Figure 4 shows the IR spectra obtained when Ti(NMe2)4 is
(19) Pugh, L. A.; Rao, K. N. Intensities from Infrared Spectra. In
Molecular Spectroscopy: Modern Research; Academic: New York, 1976;
Vol. II, p 165f.
(20) Cotton, F. A.; Wilkinson, G. AdVanced Inorganic Chemistry, 4th
ed.; Wiley: New York, 1980; p 692.
(21) van der Vis, M. G. M.; Konings, R. J. M.; Oskam, A.; Walter, R.
J. Mol. Struct. 1994, 323, 93-101.

4978 J. Am. Chem. Soc., Vol. 118, No. 21, 1996
Weiller
Figure 7. Plot of the observed decay constants (kobs) vs NH3 (9) and
ND3 (b) pressure. He buffer gas was used.
discussed above for O3, a plot of the logarithm of the IR
absorbance vs the reaction time should be linear: ln(A/A0) )
-kobst, kobs ) k1[NH3], where A is now the integrated absorbance
of the Ti(NMe2)4 band. In a time-dependent experiment, A0
would be the absorbance at zero time. We cannot obtain A0
this way since at zero time (distance) the injector would block
the IR beam. Instead we substitute the NH3 flow with buffer
gas and measure the Ti(NMe2)4 absorbance. This is equivalent
since we have shown that in the absence of NH3, there is no
dependence of the Ti(NMe2)4 absorbance on the position of the
injector over the length of the flow tube (see below). A0 is
recorded before and after each kinetic run to confirm that there
has been no change in the Ti(NMe2)4 concentration due to
bubbler fluctuations. The experiment was repeated for a series
of NH3 pressures. The data were fit to straight lines using a
weighted least-squares routine and the slopes of these lines give
the observed decay constants (kobs) at each NH3 pressure. When
the observed decay constants (kobs) are plotted against NH3
pressure (Figure 7), we find a linear dependence with a zero
intercept as expected for pseudo-first-order conditions. The
slope of this plot gives the bimolecular rate constant k1 ) (1.2
Figure 5. IR spectra for the reaction of Ti(NMe2)4 with NH3 in the
flow tube reactor. The top spectrum (A) is prior to NH3 addition and
the bottom spectrum (F) is for a sample of HNMe2. Spectra B through
E result from the addition of 0.245 Torr of NH3 to the sample in (A)
and the following reaction times: (b) 0.69, (c) 1.37, (d) 2.75, (e) 5.49
s.
( 0.2) × 10-16 molecules-1 cm3 s-1
.
C. Effects of Buffer Gas, Total Pressure, and Flow
Velocity. The effects of buffer gas, total pressure and flow
velocity on the reaction kinetics were examined. In order to
test for the possibility that the observed rates are affected by
the mixing rates, the effect of buffer gas on these measurements
was studied. The calculated diffusional mixing time at 25 °C
is about three times longer in Ar than in He (see below). If the
mixing time was significant, this would be reflected in a smaller
observed rate constant in Ar. When the measurements were
repeated in Ar buffer gas, the observed bimolecular rate constant
is well within experimental error of the He value. The effect
of total pressure was also investigated. When the total pressure
was varied from 10 to 20 Torr with the linear flow velocity
held constant at 24 cm/s, there is no significant difference in
the observed rate constants. Furthermore, when the linear flow
velocity was changed by a factor of 2 from 24 to 49 cm s-1 at
constant pressure (10 Torr), no change in the decay constants
was observed.
D. Kinetics of Ti(NMe2)4 + ND3. In order to gain some
insight into the reaction mechanism, the kinetics of Ti(NMe2)4
with ND3 were examined. Since the transamination reaction
involves cleavage of an N-H bond, we would expect a
significant kinetic isotope effect with ND3 if this is the rate-
limiting step. When ND3 is used instead of NH3, good pseudo-
first-order kinetics are observed as shown in Figure 8. When
the decay constants are plotted against ND3 pressure, the rate
constant is reduced substantially from the NH3 value. Figure
Figure 6. Plot of ln(A/A0) vs time for the reaction of Ti(NMe2)4 with
NH3. The NH3 pressures are 0.090 (b), 0.179 (2), and 0.357 (9) Torr.
spectra clearly show the disappearance of Ti(NMe2)4 with
increasing reaction time. The bottom spectrum is for a sample
of HNMe2 at 8-cm-1 resolution and by comparison we can
clearly see the formation of HNMe2. It should be noted that
these results are not limited by mixing rates (see below) and
therefore HNMe2 is a direct product from the reaction of NH3
with Ti(NMe2)4.
Figure 6 shows that when the integrated intensity of the 950
cm-1 band is plotted vs time on a semilog plot, a linear
dependence is observed as expected for a pseudo-first-order
reaction. From the available vapor pressure data,23 the estimated
Ti(NMe2)4 partial pressure is ∼0.01 Torr and is much less than
the NH3 pressure (0.1-0.4 Torr). Therefore a simple expo-
nential decay is expected: [Ti(NMe2)4] ) [Ti(NMe2)4]0 exp(-
k1[NH3]t), where k1 is the bimolecular rate constant. As
(22) Herzberg, G. Infrared and Raman Spectra of Polyatomic Molecules,
1st ed.; Van Nostrand Reinhold Company: New York, 1945; p 295.
(23) (a) The vapor pressure curve for Ti(NMe2)4 is log(Ptorr) ) 8.60 -
2850/T, or at 25 °C it is P ) 0.11 Torr: Roberts, D. The Schumacher
Corporation. Personal communication. (b) Similar data can be found in:
Intemann, A.; Koerner, H.; Koch, F. J. Electrochem. Soc. 1993, 140, 3215-
3222.

Chemical Vapor Deposition of TiN from Ti(NMe2)4 + NH3
J. Am. Chem. Soc., Vol. 118, No. 21, 1996 4979
Figure 8. Plot of ln(A/A0) vs time for the reaction of Ti(NMe2)4 with
ND3 using He buffer gas. The ND3 pressures are 0.0915 (9) and 0.364
(b) Torr.
Figure 10. Effect of HNMe2 on the reaction rate of Ti(NMe2)4 with
NH3. The slope is kobs and the open points have added HNMe2 (0.198)
and the solid points do not.
-1
Figure 11. The dependence of kobs (b) and (kobs
)
(0) on HNMe2
pressure. The NH3 pressure was held constant at 0.46 Torr.
conditions, the concentration of Ti(NMe2)4 clearly increases.
At the highest HNMe2 pressure used, there is about a 3-fold
increase in [Ti(NMe2)4] or an extent of reaction of only 25%.
When sufficient HNMe2 is added to the gas stream, the reaction
of Ti(NMe2)4 with NH3 is substantially inhibited over this time
scale.
The decay rates of Ti(NMe2)4 with and without added HNMe2
are shown in Figure 10. The NH3 pressure was 0.46 Torr and
the data are plotted as ln(A/A0) vs time. Pseudo-first-order
conditions apply for both data sets and the semilog plots are
Figure 9. IR spectra of the reaction of Ti(NMe2)4 with NH3 as a
function of added HNMe2. The NH3 was held constant at 0.46 Torr
and the HNMe2 pressure was varied from 0 (bottom) to 0.298 (top)
Torr. Features are marked with A for HNMe2 and B for Ti(NMe2)4.
Each spectrum is offset by 1.5 × 10-3 absorbance units.
7 shows this plot along with the corresponding NH3 data. The
ND3 rate constant is kd ) (4.5 ( 0.5) × 10-17 molecules-1
cm3 s-1 giving an isotope effect ratio of kH/kD ) 2.4 ( 0.4.
This indicates a primary isotope effect and that H-atom transfer
is involved in the rate-limiting step of this reaction. This is
consistent with labeling studies that showed the amine proton
in the product HNMe2 originates from NH3.11 Below we discuss
the mechanistic implications of this result.
E. Effect of Dimethylamine on the Kinetics. The IR
spectra when Ti(NMe2)4 is reacted with NH3 in the presence of
HNMe2 are shown in Figure 9. The reaction time was 0.60 s
as determined by the injector position (30 cm) and the flow
velocity (49.7 cm s-1). The partial pressure of HNMe2 was
varied (0, 0.051, 0.098, 0.148, 0.198, 0.248, and 0.297 Torr)
while the total and NH3 flows were held constant (NH3 ) 0.46
Torr). The bands at 1156 and 730 cm-1 (A) are both for HNMe2
(CH3 rock and NH bend, respectively)24 and the band at 950
cm-1 (B) is the NC2 stretch for Ti(NMe2)4. Reference spectra
for NH3 acquired under similar experimental conditions were
used to subtract the NH3 bands. In the initial spectrum the extent
of reaction is about 75% as measured by the amount of Ti-
(NMe2)4 remaining. When the partial pressure of HNMe2 is
increased at constant NH3 pressure (0.46 Torr) and flow
linear. The rate constant (k1) is calculated from the slope (kobs
)
and the NH3 pressure where kobs ) k1(NH3). In the absence of
HNMe2, the rate constant is the same within experimental error
as that observed earlier, k1 ) (1.4 ( 0.2) × 10-16 cm3
molecules-1 s-1. In the presence of 0.198 Torr of HNMe2, the
rate constant is reduced significantly to k1 ) (0.43 ( 0.03) ×
10-16 cm3 molecules-1 s-1. In an independent confirmation of
this result, we have also examined the reaction rates of
Ti(NMe2)4 with NH3 in a static cell. Again we find that the
addition of HNMe2 significantly reduces the reaction rates.
Using data from Figure 9, kobs was calculated from ln(A/A0)
and the reaction time. The resulting values for kobs are plotted
in Figure 11 showing that kobs is inversely dependent on the
HNMe2 pressure. Also plotted is (kobs)
-1, showing a linear
dependence on HNMe2 pressure. Figure 12 shows that the
dependence of kobs on NH3 pressure in the presence of 0.297
Torr of HNMe2 is linear.
F. Temperature Dependence. We have obtained data on
the temperature dependence of the rate constant for the removal
of Ti(NMe2)4 by ammonia. To accomplish this, the bubbler
was operated at room temperature and the flow tube was
wrapped with heat tape and heated with temperature controllers.
We have already demonstrated that the dependence of kobs on
(24) Gamer, G.; Wolff, H. Spectrochim. Acta 1973, 29A, 129.

4980 J. Am. Chem. Soc., Vol. 118, No. 21, 1996
Weiller
Figure 12. Dependence of kobs on NH3 in the presence of HNMe2.
The HNMe2 pressure is 0.297 Torr.
Figure 14. Temperature dependence of the rate constant (k1) for Ti-
(NMe2)4 + NH3.
important in atmospheric chemistry and it has been measured
numerous times over the years by several different techniques,25
(2) the rate constant is close in magnitude to the value for Ti-
(NMe2)4 + NH3, and (3) the IR absorption bands of isobutene
are sufficiently removed from the O3 band at 1042 cm-1 to avoid
overlap.
As Figures 2 and 3 show, we find the rate constant is (13.8
( 0.1) × 10-18 molecules-1 cm3 s-1 at 25 °C and 10 Torr total
pressure. This is in excellent agreement with the best literature
value for this reaction of (13.6 ( 0.2) × 10-18 molecules-1
cm3 s-1 25
This is particularly significant since very different
.
conditions and techniques were used previouslysa static sample
at a total pressure of 760 Torr and ex situ measurement of the
O3 concentration. The fact that we are able to reproduce this
rate constant is important. A serious concern in flow tube
studies is the rate of mixing of the two reagents to produce a
homogeneous mixture. For very fast reactions this can be rate
limiting and can prevent measurement of the reaction rate
constant. The fact that there is no dependence on total pressure
is also significant since the diffusional mixing time differs by
a factor of 2 between these two pressures. This observation
coupled with the good agreement with the literature clearly
demonstrates that mixing is not rate limiting under these
conditions and confirms the reliability of our kinetics data.
The flow tube kinetics data in Figure 5 show the direct, rapid
reaction of Ti(NMe2)4 with NH3 to produce HNMe2. We have
shown that our kinetics are not determined by mixing rates
through several sets of results. First, the rate constants are not
affected by changing the buffer gas from He to Ar which would
increase the diffusional mixing time by a factor of 3.2 from
0.07 and 0.2 s in He and Ar, respectively, at 25 °C.26 The
diffusion constant in He was estimated to be 8.9 cm2 s-1 at
298 K and 10 Torr from a collision radius for Ti(NMe2)4 of 4.5
Å and the mixing times were calculated as described else-
where.27 This calculation is an upper bound and the actual
mixing times will be significantly shorter when turbulence and
off-axis flow velocity are considered. The shortest observation
time is 0.05 s and mixing should not be significant for these
measurements. Second, when the total pressure is varied from
10 to 20 Torr, which would increase the mixing time by a factor
of 2, there is no effect on the observed reaction rates. Finally,
we observe a significant isotope effect which clearly shows that
the rates cannot be determined by mass transport. Therefore
mixing is not rate limiting under these experimental conditions
and the kinetics are chemically controlled. This conclusion is
Figure 13. Temperature dependence of the Ti(NMe2)4 removal rates.
The NH3 densities are listed in Table 1.
Table 1. Kinetics Data for Ti(NMe2)4 + NH3
T (°C)
kobs (s-1
)
NH3 (1015 cm-3
)
k (10-16 cm3 s-1
)
25.0
50.0
75.0
0.894 ( 0.03
2.31 ( 0.1
5.54 ( 0.3
11.0 ( 1.1
7.43
6.87
6.38
5.95
1.20 ( 0.04
3.4 ( 0.1
8.7 ( 0.5
100.0
18.5 ( 2.0
NH3 is first order. Therefore, we determined the observed decay
constant at a single NH3 pressure (0.230 Torr) at each of four
temperatures, 25, 50, 75, and 100 °C. Figure 13 shows the
decay plots for Ti(NMe2)4 at each of these temperatures. The
slopes (kobs) of these plots are shown in Table 1 along with the
NH3 densities and the calculated bimolecular rate constants.
Figure 14 shows an Arrhenius plot of the log of the rate
constants against inverse temperature. The slope and intercept
of this plot give values of the activation energy Ea and A factor
as follows: Ea ) 8.1 ( 0.1 kcal/mol and log(A ) ) -10.0 (
0.2 The error bars are the statistical result from the fitting
procedure. While this temperature range is less than the limit
to the experimental apparatus (∼200 °C), it was not possible to
go to higher temperatures given the signal-to-noise ratio of the
data. Higher temperatures would require better time resolution
(higher flow velocity) that would result in greater dilution by
buffer gas. In any event, the excellent precision of the data
allows reliable extrapolations to higher temperatures.
Discussion
In order to confirm that the flow tube reactor produces reliable
kinetics data, we have measured the rate constant for the reaction
of O3 with isobutene. This reaction was selected for several
reasons: (1) a reliable value is available since this reaction is
(25) Japar, S. M.; Wu, C. H.; Niki, H. J. Phys. Chem. 1974, 78, 2318.
(26) Hirschfelder, J. O.; Curtiss, C. F.; Bird, R. B. Molecular Theory of
Gases and Liquids; Wiley: New York, 1954.
(27) Keyser, L. F. J. Phys. Chem. 1984, 88, 4750-4758.

Chemical Vapor Deposition of TiN from Ti(NMe2)4 + NH3
J. Am. Chem. Soc., Vol. 118, No. 21, 1996 4981
also supported by our ability to reproduce the literature rate
constant for the reaction of O3 with isobutene.
Scheme 1. Proposed Mechanism of the Reaction between
Ti(NMe2)4 and NH3
There is no evidence for surface reactions of Ti(NMe2)4 with
the Teflon-coated walls of the flow tube reactor. The measured
rate constant, (1.2 ( 0.2 ) × 10-16 molecules-1 cm3 s-1, is
relatively small and a large number of collisions between Ti-
(NMe2)4 and the walls of the flow tube reactor will occur prior
to reaction with NH3. Therefore the removal of Ti(NMe2)4
along the length of the flow tube in the absence of NH3 was
examined. The IR intensity of the Ti(NMe2)4 bands is invariant
to the position of the injector over the length of the tube, and
consequently the reaction between Ti(NMe2)4 alone and the
Teflon-coated walls of the reactor is negligible. Furthermore,
if there was a significant component of wall reaction, one would
expect a non-zero intercept for ln(A/A0) against time. Figures
6, 8, and 10 show this is not the case, and therefore we conclude
that wall reactions do not contribute significantly to the
measured rate constant.28 They may occur but the gas-phase
reaction is much faster.
The pressure dependence of the rate constant for the reaction
of Ti(NMe2)4 with NH3 was also measured for two purposes.
First the existence of a pressure dependence would provide
mechanistic clues about the reaction. Second, modeling efforts
include both low-pressure (LPCVD)29 and atmospheric pressure
(APCVD) systems,30 and therefore it is important to determine
if the rate constant determined at low pressure would be valid
at higher total pressures. When data like that in Figure 7 are
obtained at both 10.0 and 20.0 Torr, there is no significant
difference in the slope of the two fits. Therefore we conclude
that there is no pressure dependence of the rate constant over
this pressure range. A pressure dependence might be observed
if collisional quenching of a vibrationally excited intermediate
is important in the reaction mechanism. However, this is not
likely given the relatively large size of the molecules involved
and that a second product is formed that can carry excess energy
away. This is supported by the observed independence of the
rate constant on buffer gas since Ar is a much more efficient
quencher than He. Given the lack of a pressure dependence in
this range, it is very unlikely that one would be observed at
greater total pressures.
(NMe2)4 will not survive intact to reach the surface of the
growing film.
Figures 7 and 8 show the kinetics of the reaction of Ti(NMe2)4
with ND3 and that an isotope effect of kH/kD ) 2.4 ( 0.4 is
observed. Using transition state theory, the calculated maximum
isotope effect for cleavage of an N-H bond in NH3 is kH/kD )
9 at 298 K.22,31 The experimental result is consistent with a
primary isotope effect due to the zero point energy difference.
Therefore N-H bond cleavage is the rate-limiting step. Tran-
samination is consistent with this observation and the probable
reaction mechanism is presented in Scheme 1. Initially NH3
could form a weak intermediate adduct with the Ti center in
Ti(NMe2)4. Unlike those formed with TiCl4,32 this adduct is
probably not very stable given the steric bulk of amido ligands
and the reduced acidity of the Ti center. H-atom transfer
between coordinated NH3 and an amido ligand probably occurs
via a four-center transition state followed by elimination of
HNMe2.33 Once the first amido ligand is exchanged, the steric
congestion is reduced and subsequent reactions should be faster.
Therefore we propose that, in the absence of added HNMe2,
the initial transamination reaction is rate limiting. After the
initial transamination reaction, elimination reactions could
compete with further transaminations. Calculations have found
these reactions to be facile as well34 and we will discuss their
potential role below.
The fact that we have successfully measured the gas-phase
rate constant for the reaction of Ti(NMe2)4 with NH3 is
significant. This is apparently the first measurement of the rate
constant for a transamination reaction of any metal amide in
either gas phase or solution. Previous workers observed the
time-dependent removal of Ti(NMe2)4 by reaction with NH3;
however, the rates were determined by mass transport in that
study.11 The calculated gas kinetic rate constant for this reaction
is 5 × 10-10 cm3 molecules-1 s-1 and the rate constant
represents a small (2 × 10-7) reaction probability. However,
it is larger than expected considering the low temperature (24
°C) and that both Ti(NMe2)4 and NH3 are closed-shell, stable
molecules. Our result shows that Ti(NMe2)4 and NH3 react
rapidly even at room temperature. For example, in 1 Torr of
NH3 the lifetime (1/e) of Ti(NMe2)4 is only ∼1.0 s at room
temperature. Clearly under CVD process conditions of several
Torr of NH3 and temperatures between 100 and 400 °C, Ti-
The inhibition of the reaction between Ti(NMe2)4 and NH3
by excess HNMe2 as shown by Figures 9 through 11 has
important mechanistic implications. All of the data are con-
sistent with a reversible reaction between Ti(NMe2)4 and NH3.
This is not surprising when one considers the thermochemistry
of transamination reactions. The Ti-NR2 bond energies are
independent of R to first approximation35 and the calculated
bond energy for Ti-NH2 is close the measured values for Ti-
NR2.36 Since the entropies of the free amines are similar, ∆S
∼ 0, the free energy change for the transamination reactions is
also approximately zero, ∆G ∼ 0. This is consistent with the
early studies of transamination reactions of Ti(NR2)4 with
numerous dialkylamines in solution.14 Others have examined
(31) Moore, J. W.; Pearson, R. G. Kinetics and Mechanism; John Wiley
and Sons: New York, 1981.
(32) Saeki, Y.; Masuzaki, R.; Yajima, A.; Akiyama, M. Bull. Chem. Soc.
Jpn. 1982, 55, 3193-3196.
(33) Chisholm, M. H.; Rothwell, I. P. In ComprehensiVe Coordination
Chemistry; Wilkinson, G., Ed.; Pergamon Press: New York, 1987.
(34) Cundari, T. R.; Gordon, M. S. J. Am. Chem. Soc. 1993, 115, 4210-
4217.
(35) (a) Lappert, M. F.; Patil, D. S.; Pedley, J. B. J. Chem. Soc. 1975,
830-831. (b) Baev, A. K.; Mikahailov, V. E. Russ. J. Phys. Chem. 1989,
63, 949-955.
(28) Under conditions where a large amount of deposit has accumulated
on the walls of the flow tube reactor, we find that the rate constants are
significantly higher. When the walls are cleaned as described in the
Experimental Section, the rate constants return to the quoted value. Prior
to each run, the rate constant was measured under control conditions to
confirm the lack of wall effects.
(29) Cale, T. S.; Chaara, M. B.; Raupp, G. B.; Raijmakers, I. J. Thin
Solid Films 1993, 236, 294-300.
(30) Toprac, A. J.; Wang, S.-Q.; Musher, J.; Gordon, R. G. Mater. Res.
Soc. Symp. Proc. 1994, 236.
(36) Allendorf, M. D.; Janssen, C. L.; Colvin, M. E.; Melius, C. F.;
Nielson, I. M. B.; Osterheld, T.; Ho, P. In Process Control, Diagnostics,
and Modeling in Semiconductor Manufacturing; The Electrochemical
Society: Pennington, NJ, 1995, in press.

4982 J. Am. Chem. Soc., Vol. 118, No. 21, 1996
Weiller
Ti(NMe2)4 + ButNH2 f 1/2[(Me2N)2Ti(NBut)]2 (7)
the reactions of metal amides with NH3 and observed the
irreversible formation of insoluble products.37 This is not
inconsistent with our results since we propose that only the initial
transamination reaction to form (Me2N)3Ti-NH2 is reversible,
not that the overall reaction is at equilibrium. The subsequent,
product-forming reactions should be irreversible.
Possible mechanisms for the reaction are shown below. The
first step in the proposed mechanism is reversible transamination
to form the transient intermediate (Me2N)3Ti-NH2 (eq 2). The
subsequent reactions are of particular interest since they control
which species reach the surface. Two reasonable possibilities
are the following: (a) reaction with NH3 in a second transami-
nation step to form (Me2N)2Ti(NH2)2 (eq 3) or (b) elimination
of HNMe2 to form an imide (eq 4):
Ta(NMe2)5 + ButNH2 f (Me2N)3TadNBut
(8)
Both of these reactions probably proceed by an initial tran-
samination followed by elimination of HNMe2. Experiments
attempting to directly confirm the mechanism are in progress.
This mechanism is consistent with all of our experimental
data. There are three sets of data to consider: (1) the
dependence of kobs on NH3 (Figure 7), (2) the dependence of
kobs on HNMe2 (Figure 11), and (3) the dependence of kobs on
-1
NH3 with added HNMe2 (Figure 12). The plot of (kobs
)
on
HNMe2 in Figure 11 is linear with slope ) 5.86 ( 0.16 Torr-1
s and intercept ) 0.44 ( 0.03 s. The dependence of (kobs
-1
)
on HNMe2 predicted by eq 6 is shown below:
Ti(NMe2)4 + NH3 { k } (Me2N)3Ti-NH2 + HNMe2 (2)
1
k-1
k-1[HNMe2]
k1k3[NH3]
1
kobs
1
)
+
(9)
k2
k1[NH3]
(Me2N)3Ti-NH2 + NH3
98 (Me2N)2Ti(NH2)2 + HNMe2
(3)
The product of the intercept and [NH3] gives k1 ) (1.5 ( 0.2)
× 10-16 molecules-1 cm3 s-1 which is in good agreement with
the value determined from the dependence of kobs on NH3
k3
(Me2N)3Ti-NH2
98 (Me2N)2TidNH + HNMe2 (4)
(Figure 7): k1 ) (1.2 ( 0.2) × 10-16 molecules-1 cm3 s-1
.
From the slope of eq 9, k1, and [NH3], the calculated value for
(k3/k-1) ) 3.1 × 1015 molecules cm-3. This quantity enables
us to calculate the expected slope for Figure 12 from eq 6, 3.0
The steady state approximation on (Me2N)3Ti-NH2 for the
two mechanisms, eqs 2 and 3 and 2-4, gives expressions for
38
kobs
:
× 10-17 molecules-1 cm3 s-1
. This is consistent with the
observed value, 4.3 × 10-17 molecules-1 cm3 s-1. A relevant
quantity to consider is the ratio of first-order rate constants,
k′-1/k3, where k′-1 is the pseudo-first-order rate constant, k′-1
) k-1[HNMe2]. At the HNMe2 pressure used in Figure 12
(0.297 Torr), k′-1/k3 ) 3.0, indicating that reaction of (Me2N)3-
Ti-NH2 with HNMe2 is three times as fast as elimination at
this [HNMe2]. This supports the proposed mechanism of pre-
equilibrium followed by elimination.
k1k2[NH3]2
kobs
)
(5)
(6)
k-1[HNMe2] + k2[NH3]
k1k3[NH3]
kobs
)
k-1[HNMe2] + k3
Both expressions give the observed linear dependence of kobs
on NH3 when [HNMe2] ) 0, kobs ) k1[NH3], and the derived
bimolecular rate constant represents the forward transamination
Finally we consider the temperature dependence results in
light of this mechanism. In the absence of added HNMe2, kobs
) k1[NH3] and therefore the bimolecular rate constants are for
the initial transamination reaction. Therefore the activation
energy Ea ) 8.1 ( 0.1 kcal/mol and pre-exponential factor log-
(A ) ) -10.0 ( 0.2 are for this reaction. The calculated
activation enthalpy is ∆H† ) 6.9 ( 0.1 kcal/mol.31 Unfortu-
nately, literature searches have not uncovered any kinetics data
for similar reactions in solution or gas phase for comparison.
However, it is useful to consider the magnitude of the pre-
exponential factor for consistency with this reaction. The A
factor is in good agreement with that expected for a bimolecular
metathesis reaction involving polyatomic reactants.42 A more
intuitive interpretation of the A factor is the activation
entropy: ∆S† ) -19 cal/(mol K) for a standard state of 1 atm.
This is very reasonable for the proposed bimolecular reaction.
It is instructive to compare the activation energy to that
derived from CVD studies. The only relevant temperature
dependent data on this system comes from one study of TiN
deposition from Ti(NMe2)4 and NH3 at low pressure in a rotating
disk reactor (RDR).43 Most workers have studied this system
in standard reactors at higher pressures and have found the
deposition rate to be mass transport limited over a wide
temperature range with a typical activation energy of 3.5 kcal/
mol.10 However, in the low-pressure RDR study, both transport-
limited and kinetically-limited regimes for the deposition rate
were found. The kinetically-limited regime below 300 °C gives
-1
reaction (k1). The observed linear dependence of (kobs
)
on
HNMe2 is also predicted by both expressions. The distinguish-
ing feature is the order with respect to NH3 in the presence of
excess HNMe2: (5) is second order and (6) is first order. The
dependence of kobs on NH3 with added HNMe2 (0.297 Torr)
was linear with a zero intercept as shown in Figure 12. While
this is not conclusive, we favor mechanism 2-4 for several
additional reasons. (Me2N)2TidNH could easily give com-
plexes with bridging imido groups similar to those previously
postulated as intermediates in this system.11 A similar imide
has been isolated as an intermediate in TiN CVD from TiCl4
and NH339 and there are several other related examples as well.40
Furthermore, a similar reaction sequence has been observed
when Ti(NMe2)4 or Ta(NMe)5 reacts with ButNH2:41
(37) (a) Brown, G. M.; Maya, L. J. Am. Ceram. Soc. 1988, 71, 78-82.
(b) Seyferth, D.; Mignani, G. Mater. Sci. Lett. 1988, 7, 487-488.
(38) Here we have assumed that all subsequent reactions are fast and
therefore kinetically unimportant.
(39) (a) Winter, C. H.; Sheridan, P. H.; Lewkebandra, T. S.; Heeg, M.
J.; Proscia, J. W. J. Am. Chem. Soc. 1992, 114, 1095-1097. (b)
Lewkebandra, T. S.; Sheridan, P. H.; Heeg, M. J.; Rheingold, A. L.; Winter,
C. H. Inorg. Chem. 1994, 33, 5879-5889.
(40) (a) Hill, J. E.; Profilet, R. D.; Fanwick, P. E.; Rothwell, I. P. Angew.
Chem., Int. Ed. Engl. 1990, 29, 664-665. (b) Roesky, H. W.; Voelker, H.;
Witt, M.; Noltemeyer, M. Angew. Chem., Int. Ed. Engl. 1990, 29, 669-
670. (c) Cummins, C. C.; Schaller, C. P.; Van Duyne, G. D.; Wolczanski,
P. T.; Chan, A. W. E.; Hoffman, R. J. Am. Chem. Soc. 1991, 113, 2985-
2994.
(41) (a) Nugent, W. A.; Haymore, B. L. Coord. Chem. ReV. 1980, 31,
123-175. (b) Nugent, W. L.; Harlow, R. L. J. Chem. Soc., Chem. Commun.
1978, 579.
(42) Benson, S. W. Thermochemical Kinetics; Wiley: New York, 1976.
(43) Hillman, J. T.; Rice, M. J., Jr.; Studiner, D. W.; Foster, R. F.;
Fiordalice, R. W. Proc. VMIC Conf. 1992, 246-252.

Chemical Vapor Deposition of TiN from Ti(NMe2)4 + NH3
J. Am. Chem. Soc., Vol. 118, No. 21, 1996 4983
an activation energy of 7.5 kcal/mol. This is in excellent
agreement with our data and suggests that the transamination
reaction is rate limiting in the kinetic regime in the CVD system.
with the proposed rate-limiting step of transamination. The
precision of the activation parameters is high and allows
calculation of the rate constant over a wide range of temperatures
for modeling this system. The activation energy is in excellent
agreement with the activation energy for the deposition of TiN
from Ti(NMe2)4 and NH3 under kinetically controlled condi-
tions: Ea ) 7.4 kcal/mol. This is consistent with transamination
as the rate-limiting step in the deposition of TiN.
Conclusions
The reaction between Ti(NMe2)4 and NH3 is rapid even at
room temperature. When Ti(NMe2)4 is mixed with less than 5
Torr of NH3 in a closed gas cell, the reaction is complete in
less than 5 s. The only gas-phase product that could be
identified conclusively is HNMe2 and this is consistent with a
transamination reaction between Ti(NMe2)4 and NH3. This
explains the importance of NH3 in the formation of low-carbon
TiN films; carbon is removed from the starting material through
the formation of HNMe2.
We have used a flow tube reactor to measure the kinetics of
the gas-phase reaction between Ti(NMe2)4 and NH3. In the flow
tube reactor, the IR spectra show that Ti(NMe2)4 is completely
reacted in 5 s in the presence of 0.1 Torr of NH3 with the
formation of HNMe2 as a direct product. The dependence of
the decay rates on NH3 pressure gives the rate constant at 24
°C: k ) (1.2 ( 0.2) × 10-16 molecules-1 cm3 s-1. The use of
He or Ar buffer gas has no effect on the value of the rate
constant demonstrating that the kinetics are not determined by
mass transport. This rate constant is relatively large and
confirms quantitatively the rapid reaction between Ti(NMe2)4
and NH3. For example, the rate constant indicates that in the
presence of 1 Torr of NH3, the lifetime of Ti(NMe2)4 is ∼1 s
at room temperature. Clearly under the CVD process conditions
of high NH3 pressures and temperatures, extensive gas reactions
do occur and Ti(NMe2)4 will be converted into a new species
before it reaches the surface of the growing film. In order to
understand this and other CVD processes, it is critical to obtain
data on the rates of these gas-phase reactions.
The rate constant is independent of total pressure and shows
that the rate constant is valid for modeling systems under a range
of pressures. In addition, wall reactions do not contribute
significantly to the observed reaction rates. When ND3 is used
instead of NH3, a normal isotope effect is observed showing
that an N-H bond is broken in the rate-limiting step of reaction.
Given this result and the observation of dimethylamine as the
direct product, we propose that the transamination reaction
between Ti(NMe2)4 and NH3 is the rate-limiting step of the
reaction. The rate constant is assigned to the transamination
of the first amido ligand. After transamination, elimination to
form an imido intermediate is most consistent with the kinetic
data and with prior studies. Unfortunately the data do not allow
for more detailed identification of the intermediates responsible
for film growth. Direct in situ spectroscopic studies will be
required for this and are planned.
As discussed in the introduction, the properties of TiN
deposited from Ti(NMe2)4 and NH3 are not acceptable and Ti-
(NEt2)4 gives much better films. The reaction rate of Ti(NEt2)4
with NH3 is much slower than for Ti(NMe2)444 and this appears
to explain the difference in film properties. One of the most
important of these is step coverage, and Ti(NEt2)4 is much better
in this regard than Ti(NMe2)4. Step coverage is thought to be
predominantly controlled by the sticking coefficient of the
molecule reacting with the surface. Molecules with low sticking
coefficients can survive numerous collisions with the side walls
of a feature to react with the bottom and hence will give better
step coverage. The large rate constant for Ti(NMe2)4 means
that, under CVD conditions, it will be completely reacted and
may be converted into an intermediate that has a large sticking
coefficient. In the case of Ti(NEt2)4, the rate constant is much
smaller, the extent of reaction is much smaller, and intermediates
with smaller sticking coefficients could be formed.
One of the most significant results of this work is the
observation that a transamination reaction between a metal
amide and NH3 is reversible, apparently for the first time. This
has important implications for the chemistry of amido com-
pounds as well as for the CVD of TiN from them. For example,
the use of amines may provide a simple method for controlling
the extent of gas-phase reaction and therefore the properties of
TiN. In addition, this result will provide an important consid-
eration in understanding and controlling the chemistry of other
CVD processes since amido precursors are useful for the CVD
of many nitride materials.45 Currently we are exploring the
generality of this effect for the gas-phase reactions of other metal
amido complexes with NH3 as well as the effect added amine
has on the properties of TiN deposited.
Acknowledgment. This work was supported by Sematech
and by The Aerospace Sponsored Research Program. Thanks
are due Drs. B. Brady, R. Heidner, R. Martin, and M. Allendorf
for critical comments.
JA953468O
(44) Preliminary data provide an estimate of the relative rate constants
for Ti(NMe2)4 and Ti(NEt2)4 with NH3 at 85 °C to be kMe/kEt ) 155; Weiller,
B. H. To be published.
(45) (a) Gordon, R. G.; Hoffman, D. M.; Riaz, U. Mater. Res. Soc. Symp.
Proc. 1992, 242, 445-450. (b) Gordon, R. G.; Hoffman, D. M.; Riaz, U.
Chem. Mater. 1992, 4, 68-71. (c) Gordon, R. G.; Riaz, U.; Hoffman, D.
M. J. Mater. Res. 1992, 7, 1679-1684. (d) Hoffman, D. M. Polyhedron
1994, 13, 1169-1179.
The temperature dependence of the rate constant was
determined and provides activation parameters for the rate
constant: log(A ) ) -10.0 ( 0.2; Ea ) 8.1 ( 0.1 kcal/mol.
The activation entropy, ∆S† ) -19 cal/(mol K), is consistent