1186 J. Phys. Chem.

A 2000, 104, 1186-1195

Initial Reactions in Chemical Vapor Deposition of Ta2O5 from TaCl5 and H2O. An Ab
Initio Study

Magdalena Siodmiak,† Gernot Frenking,*,† and Anatoli Korkin*,‡

Fachbereich Chemie, Philipps-UniVersität Marburg, Hans-Meerwein-Strasse, D-35032 Marburg, Germany, and
Computational Technology Lab, Motorola Inc., Mesa, Arizona 85202
ReceiVed: July 29, 1999; In Final Form: NoVember 8, 1999

Quantum chemical calculations at the Hartree-Fock, B3LYP, MP2, and CCSD(T) levels of theory were
performed in order to study the mechanism of Ta2O5 chemical vapor deposition (CVD) from TaCl5 and H2O.
The geometries and vibrational frequencies of reactants, products and transition states of the reactions, which
are used for modeling the initial steps of the CVD process were calculated using effective core potentials for
Ta and Cl atoms. The reaction between TaCl5 and H2O proceeds via formation of a strongly bonded six-
coordinated tantalum complex (25 kcal/mol at CCSD(T)//B3LYP). The correlated methods show the transition
state energy to be close to the energy of the initial reagents. Formation of TaOCl3, in the unimolecular
decomposition of TaCl4OH and TaCl3(OH)2, has similar barriers of ∼20 kcal/mol. The catalytic effect of an
assisting water molecule opens a barrierless channel for TaOCl3 formation in the gas phase.

Introduction nism.14 Since in ALD the active components are introduced into
the reaction zone subsequently, it allows a homogeneous
Tantalum pentoxide thin films, a high-quality dielectric
distribution of reactants throughout the surface area. It also
material, have been intensively studied for different applications
separates certain deposition reaction steps, which provides films
in microelectronic industry, e.g., as capacitors for dynamic
with high precision thickness control and facilitates the study
random access memories, gate insulator for metal-oxide-
of the deposition mechanism. According to Aarik and co-
semiconductor devices and integrated optical devices1-8 The
workers14 formation of intermediate volatile TaOCl3 explains
high integration of an increasing number of semiconductor
etching, which occurs at high TaCl5 doses and increased
devices on chips of decreasing size leads to extremely high
temperature. The reactivity of TaCl5 toward H2O-treated
requirement of film quality. Among the different growth
tantalum oxide has been shown to be higher than the reactivity
techniques9 chemical vapor deposition (CVD)10 often provides
of H2O toward TaCl5-treated oxide surfaces.14
the best quality thin films with very low defect density.11 The
search for optimal precursors and operational conditions in a There is a growing number of theoretical ab initio studies of
CVD reactor, e.g., pressure, temperature, carrier gas, flow rate, molecules (clusters) and their reactions, which provide valuable
etc., can be accomplished either experimentally or via computer information about a CVD mechanism of semiconductor and
simulations. Experiments provide the most reliable results, but dielectric-type materials for known and potential applications
they are expensive and time and material consuming. Computer in the microelectronic industry (see, for example, ref 15 and
simulations are an efficient alternative, if an adequate and the literature cited therein). Besides fundamental qualitatiVe
reliable model is derived. Gas phase and surface reaction knowledge about the type of reactions which may occur in CVD
mechanism and kinetics are essential components for the processes, these studies provide quantitatiVe values of thermo-
construction of a computer model for CVD reactor simulation. chemical and kinetics parameters, which can be used in
Knowledge of nature, thermochemistry, and kinetics of critical semiempirical modeling of processes in CVD reactors. In this
reactions leading to film deposition and to byproduct formation paper we present a quantum chemical study of initial reactions
also helps in the search for obtaining high quality thin films in tantalum pentoxide deposition from the TaCl5/H2O system.16
with best chemical, electrical, structural, and mechanical proper- Although this study concerns only selected reactions of gas-
ties. phase molecules, we believe that it helps to understand the
Several precursors such as TaF5,12 TaCl5, Ta[N(CH3)2]5, Ta- molecular mechanism of Ta2O5 CVD. We predict that the
(OCH3)5, and Ta(OC2H5)513 have been applied in the Ta2O5 reaction between TaCl5 and H2O in the gas phase occurs without
CVD. Tantalum pentachloride has the advantage of avoiding a significant barrier. Assuming that fast vibrational energy
carbon contamination in growing films.13 It is also suitable for relaxation occurs on the surface of the growing Ta2O5 film, an
model experimental and theoretical studies of the CVD mech- intermediate complex of hexa-coordinated tantalum can be
anism since the number and type of intermediate species and formed between TaCl5 and hydroxyl groups on the surface. We
reactions involved are expected to be smaller compared to have used a variety of theoretical methods, i.e., HF, MP2,
tantalum compounds with alkylamino and alkoxi ligands. A B3LYP, and CCSD(T) in combination with effective core
study of atomic layer deposition (ALD) of Ta2O5 from TaCl5 potentials (ECP) and 6-31G* basis sets. According to the
and H2O revealed important details of the deposition mecha- literature data17 correlated methods in combination with ECP’s
can provide a reasonable (∼3-5 kcal/mol) accuracy for such
† Philipps-Universität Marburg reaction energies and barriers. We have also computed heats of
‡ Motorola Inc. formation, structures, and vibrational frequencies of the tantalum
10.1021/jp992673d CCC: $19.00 © 2000 American Chemical Society
Published on Web 01/25/2000
Chemical Vapor Deposition of Ta2O5 J. Phys. Chem. A, Vol. 104, No. 6, 2000 1187

Figure 1. Optimized geometries of (a) TaCl5, (b) TaOCl3, (c) TaClO2 at the B3LYP level. Values at HF followed by MP2 data are given in
parentheses. Bond lengths are given in Å, angles in degrees.

TABLE 1: Computed and Experimental Vibrational B3LYP level and the structural parameters for the B3LYP
Frequencies (in cm-1) and IR Intensities (in % to the most optimized geometries.
active mode) for TaCl5
All calculations were performed using a quasi-relativistic
freq HF MP2 B3LYP expa effective core potential (ECP) for tantalum with a valence basis
ν1(a′1) 383 (0) 415 (0) 400 (4) 406 set (441/2111/21), which was derived from the [55/5/3] valence
ν2(a′1) 294 (0) 330 (0) 319 (0) 324 basis set of Hay and Wadt.27 Also, for chlorine, an ECP with a
ν3(a”2) 329 (100) 376 (100) 359 (100) 371 valence basis set (4/5)/[2s3p]28 extended by a d-type polarization
ν4(a”2) 161 (8) 176 (7) 166 (6) 155
ν5(e′) 371 (55) 413 (50) 399 (54) 402 function29 was used. For oxygen and hydrogen atoms we used
ν6(e′) 138 (4) 147 (3) 139 (3) 181 6-31G** basis sets.30 It has been demonstrated for a representa-
ν7(e′) 55 (1) 58 (3) 59 (2) 54 tive set of transition metal complexes31 that ECPs generated
ν8(e′′) 182 (0) 198 (0) 187 (0) 127 for ab initio methods can be applied in DFT-based calculations
a Reference 16 as well. To confirm the accuracy of the ECP in the reaction
mechanism study, we performed B3LYP calculations of tanta-
containing molecules TaCl5, TaOCl3, TaO2Cl, TaCl4OH, and lum oxochloride formation (see below) with a 6-31G* all-
TaCl3(OH)2. In this series of molecules only TaCl5 has been electron basis set30 for the chlorine atom. The difference in the
studied experimentally,18 while theoretical data are available for calculated reaction energies using both kinds of basis sets for
TaCl519 and TaOCl3.20,21 A comparison of our theoretical results chlorine atom is only 3 kcal/mol.
with literature data for these two molecules provides an estimate
of the accuracy of ab initio approaches applied in this study, Results and Discussion
while computational data obtained for unknown tantalum
Considering the mechanism of initial reactions leading to
compounds, e.g. TaO2Cl, TaCl4OH, and TaCl3(OH)2 should help
formation of tantalum pentoxide from tantalum pentachloride
in their experimental identification and characterization.
and water we have computed a series of tantalum-containing
molecules: TaCl5, TaOCl3, TaO2Cl, TaCl4OH, and TaCl3(OH)2.
Computational Methods
For TaCl4OH and TaCl3(OH)2 several geometric isomers can
Molecular geometries have been optimized using three be visualized based on different possible locations of Cl and
different theoretical methods: HF, MP2,22 and B3LYP23 as OH in a trigonal bipyramide (axial and equatorial) and H-
implemented in the Gaussian 98 program package.24 Analytical bonding pattern. Since a detailed conformational analysis of
harmonic frequencies have been computed at the B3LYP level tantalum hydroxychlorides is beyond the scope of our paper
of theory for all stationary structures and at the HF and MP2 we have considered only a few geometric isomers and confor-
levels for selected molecules. HF vibrational frequencies are mations, which apparently have low energies or are located on
scaled by 0.89.25 Single point coupled cluster CCSD(T)26 a minimum energy pathway toward a reaction transition state.
calculations have been done at the B3LYP optimized geometries. Conformational processes in those molecules are accompanied
All relative energies presented in the paper include zero-point by small energy changes and thus should be very fast.
energy (ZPE) corrections. If not specified otherwise relative Structure, Bonding, and Vibrational Frequencies of TaCl5,
energies discussed in the paper are given at the CCSD(T)// TaOCl3, and TaO2Cl. The calculated geometries of TaCl5,

TABLE 2: Computed Harmonic Vibrational Frequencies (in cm-1) and IR Intensities (in % to the most active mode) for
TaOCl3 and Experimental Vibrational Frequencies for VOCl3 and NbOCl3
freq HF MP2 B3LYP assignment VOCl3a NOCl3b
ν1(a1) 1010 (100) 949 (31) 1017 (100) TaO bond stretch 1043 997
ν2(a1) 372 (9) 405 (13) 398 (10) TaCl symmetric bond stretch 411 395
ν3(a1) 120 (2) 123 (2) 123 (2) ClTaCl symmetric bending 160 133
ν4(e) 377 (57) 420 (100) 410 (80) TaCl asymmetric bond stretch 508 448
ν5(e) 216 (4) 223 (2) 227 (3) OTaCl bending 246 225
ν6(e) 103 (1) 106 (1) 108 (1) ClTaCl asymmetric bending 124 106
a Reference 33. b Reference 31, 34.
1188 J. Phys. Chem. A, Vol. 104, No. 6, 2000 Siodmiak et al.

TABLE 3: Computed Vibrational Frequencies (in cm-1) and SCHEME 1

Intensities (in % to the most active mode) for TaO2Cl
freq HF MP2 B3LYP assignment
ν1(a′) 986 (43) 917 (23) 999 (42) TaO bond symmetric stretch
ν2(a′) 367 (25) 390 (61) 392 (34) TaCl stretch
ν3(a′) 293 (3) 316 (5) 329 (4) OTaO bending
ν4(a′) 102 (28) 155 (28) 156 (20) ClTa(O2) symmetric bending
ν5(a′′) 925 (100) 904 (100) 946 (100) TaO bond asymmetric stretch SCHEME 2
ν6(a′′) 153 (2) 176 (<1) 180 (<1) ClTa(O2) asymmetric bending

TABLE 4: Results of the NBO Analysis at B3LYP: Bond

Polarity (%Ta), Relative Contribution of s, p, and d Orbitals
of Tantalum, and Atomic Charges (q)
bond %Ta %s (Ta) %p (Ta) %d (Ta) q(Ta) q(O) q(Cl)
TaOCl3 C3V
TaCl 18 27 1 72 1.66 -0.71 -0.32 SCHEME 3
TaO σ 21 18 1 81
TaO πa 23 0 1 99
TaO2Cl C2v (planar)
TaCl 8 68 10 23 2.23 -0.83 -0.56
TaO σ 15 4 12 84
TaO π⊥ 26 0 3 97
TaO π|| 11 5 26 69
TaO2Cl Cs (nonplanar) calculated value of 1.7 kcal/mol for the energy of pseudorotation
TaCl 14 50 2 48 2.06 -0.79 -0.47 barrier in TaCl5 computed at B3LYP coincides with an earlier
TaO σ 18 19 8 74
TaO π1 24 3 3 95 estimate at QCISD(T).19 It is also in good agreement with an
TaO π2 13 1 29 70 experimental value of 1.2 kcal/mol18 estimated using electron
a
diffraction data obtained at 100 °C.
There are two degenerated π orbitals in the C3V symmetric TaOCl3.
There are no experimenatal data available for the geometries
TABLE 5: Relative Energies of TaCl4OH and TaCl3(OH)2 and vibrational frequencies of TaOCl3. Our calculated molecular
Conformers (in kcal/mol) geometries (Figure 1b) can be compared only with the data
HF MP2 B3LYP CCSD(T) obtained from earlier theoretical studies at Hartree-Fock and
MP220 and at the DFT (local density approximation (LDA))21
TaCl4OH
OH-eq 0.0 0.0 0.0 0.0 levels of theory. In our and other calculations the ∠OTaCl angle
TS 1.9 1.9 2.0 1.9 is smaller than ∠ClTaCl in TaOCl3. Similar results were
OH-ax 1.2 0.9 1.1 0.9 obtained for the other group 5 oxychlorides, VOCl32132 (∠OVCl
TaCl3(OH)2 ) 107.6°, ∠ClVCl ) 111.3°) and NbOCl321,33 (∠ONbCl )
OH, OH-eq 0.0 0.0 0.0 0.0 106.0°, ∠ClNbCl ) 112.7°). This is in contrast to main-group
TS 2.8 2.9 2.8 3.1
OH-eq, OH-ax 1.3 0.4 0.8 0.9
15 oxochlorides, e.g., POCl3 (∠OPCl ) 114.8°, ∠ClPCl )
103.7°).34 The calculated harmonic frequencies for TaOCl3 are
TaOCl3, and TaO2Cl at B3LYP, HF, and MP2 are shown at given in Table 2 along with experimental frequencies for
Figure 1. The corresponding vibrational frequencies are pre- VOCl335 and NbOCl3.33,36 The frequency pattern is very similar
sented in Tables 1-3. Experimental data are available only for for all three molecules. Vibrational frequencies diminish with
TaCl5.18 The experimental value for the equatorial TaCl bond increasing atomic mass (ω ∼ M-1/2) and the reduced masses
length (2.227 Å) is shorter and for the axial bond it is longer of vibrations approach the masses of light ligand atoms, O or
(2.369 Å) than the corresponding calculated values. However, Cl, if the mass of the central heavy atom grows. According to
the computational values are within standard deviations esti- this we may simply extrapolate the TaOCl3 vibrational frequen-
mated for the experimental values (see Figure 1a). The B3LYP cies to be 10-20 cm-1 smaller than the corresponding values
method provides longer axial and equatorial bonds compared for NbOCl3.
to the other two computational approaches, HF and MP2. In an An NBO analysis for TaOCl3 at B3LYP is presented in Table
earlier ab initio HF pseudopotential study of TaCl5, values of 4. The Ta-Cl and Ta-O bonds are very polar. The σ and π
2.369 Å and 2.326 Å were reported for the axial and equatorial bonds have a maximum localization of only 26% at the tantalum
TaCl bonds, respectively.19 The B3LYP calculation of TaCl5 end. Thus, Ta always carries a large positive charge. The metal
with a 6-31G* basis set for chlorine atom gives a bond length s orbital contributes more to the Ta-Cl bonds than to the Ta-O
of 2.346 Å for the axial Ta-Cl bond, which is close to the σ bonds.
experimental value. The calculated equatorial Ta-Cl bond There are neither theoretical nor experimental data available
length (2.290 Å) has a larger deviation from experiment. for TaO2Cl. According to our computations the molecule is
The computed and experimentally assigned18 vibrational nonplanar with Ta-Cl and Ta-O bond lengths being larger
frequencies of TaCl5 are in good agreement except for ν6 and than in TaOCl3. The nonplanar structure is a result of a
ν8, where theory and experiment show opposite values. Similar compromise between ligand-ligand repulsion, which would lead
disagreement between theory and experiment in the symmetry to a planar structure, and the overlap between metal and ligand
assignment of these two modes has been pointed by Albright orbitals, which according to Landis37 favors angles smaller than
et al.19 The e′′ mode is formally symmetry forbidden, but 120° for any sdn hybridization. If the planar structure is imposed
coupling of vibrational modes may let the ν8 mode mix with in the geometry optimization, all bonds become longer (TaO
other vibrations. This may lead to an erroneous symmetry ) 1.737 Å, TaCl ) 2.361 Å) and the energy raises by 5.7 kcal/
assigment based on experimental IR and Raman intensities. Our mol at B3LYP.
Chemical Vapor Deposition of Ta2O5 J. Phys. Chem. A, Vol. 104, No. 6, 2000 1189

Figure 2. Optimized geometries of TaCl4OH and TaCl3(OH)2 conformers at the B3LYP level. Values at HF followed by MP2 data are given in
parentheses. Bond lengths are given in Å, angles in degrees.

TABLE 6: Zero Kelvin Heats of Initial Reactions of TaCl5 Cl bond. In the nonplanar TaO2Cl molecule the Ta AO has
Hydrolysis and Ta2O5 Deposition (in kcal mol)a almost exact sd hybridization in the TaCl bond. There is a
reaction HF MP2 B3LYP CCSD(T) significant difference between the charge distribution in the
(1) TaCl5 + H2O f -8.0 -7.6 -8.6 -7.3 planar and nonplanar form of TaO2Cl. In the nonplanar form
TaCl4OH + HCl the Ta atom is less positive (+2.06) than in the planar form
(2) TaCl4OH + H2O f -8.0 -6.8 -7.4 -6.8 (+2.23). Thus, bending of the molecule increases the covalent
TaCl3(OH)2 + HCl contribution to the chemical bonding in TaO2Cl.
(3) TaCl4OH f 11.7 0.1 9.2 8.0 Energies and Structures of TaCl4OH and TaCl3(OH)2.
TaOCl3 + HCl
(4) TaCl3(OH)2 f 19.7 6.9 16.6 14.9
Tantalum hydroxytetrachloride [TaCl4OH] and tantalum dihy-
TaOCl3 + H2O droxytrichloride [TaCl3(OH)2] are intermediate species in the
(5) TaOCl3 + H2O f 90.0 76.6 78.7 82.0 hydrolysis of tantalum pentachloride yielding tantalum pentox-
TaO2Cl + 2HCl ide. Two geometric isomers can be visualized for TaCl4OH,
a Lower energy conformers of TaCl OH and TaCl (OH) have been with equatorial and axial positions of the hydroxyl group,
4 3 2
used used in computation of heats of reactions 1-4. respectively. The two isomers apparently can be converted into
each other via a square pyramidal transition state (Berry
TABLE 7: Estimated Heats of Formation and Ta-O Bond pseudorotation), Scheme 1.
Dissociation Energies in Tantalum-Containing Molecules We located both geometric isomers and the energetically low
TaCl4OH TaCl3(OH)2 TaOCl3 TaO2Cl lying transition state shown in Scheme 1 (see Figure 2 a-c
∆H°f [kcal/mol] -225.8 -268.3 -195.7 -127.3 and Table 5). The isomer with an equatorial position of the OH
ETaO [kcal/mol] 126.4 126.1 220.5 179.8 group (Figure 2a) is more stable. All methods predict that the
OH-ax isomer (see Figure 2c) is ca. 1 kcal/mol higher in energy.
In accord with the bond elongation in TaO2Cl, the corre- The barrier for interconversion of OH-eq into OH-ax is ∼2 kcal/
sponding Ta-O and Ta-Cl bond stretch vibrations have lower mol. In all three structures the OH group forms an H-bond with
frequencies compared to TaOCl3 (see Table 3). The predicted the nearest chlorine atom. The average OH‚‚‚Cl intramolecular
IR spectrum of TaO2Cl has four intense bands: two modes H-bond distance is 2.932 Å at B3LYP, which can be compared
around 1000 cm-1, which correspond to symmetric and asym- with 2.56 (2.57) Å intermolecular H-bond in (HCl)2 dimer
metric Ta-O bond stretching, one mode with a frequency 370 computed at the B3LYP (MP2) level with the 6-31+G** basis
cm-1 corresponding to Ta-Cl bond stretching mixed with set.38 According to Del Bene et al.38 the B3LYP method
∠OTaO bond angle bending. Finally, the vibration at about 100 somewhat overestimates binding energies of H-bonds and gives
cm-1 corresponds to the ∠TaOCl bond angle bending. shorter than experimental bond distances, particularly for small
The NBO analysis (see Table 4) shows a similar Ta basis sets.
hybridization character in the σ(Ta-O) and Ta-Cl bonds of For TaCl3(OH)2, the product of the second step of TaCl5
TaOCl3, while in TaO2Cl the Ta d-orbitals dominate in the σ- hydrolysis, we can visualize three isomers with respect to the
(Ta-O) bonds and the s-orbital mainly contributes to the Ta- position of OH and Cl in the trigonal bipyramide (Scheme 2).
1190 J. Phys. Chem. A, Vol. 104, No. 6, 2000 Siodmiak et al.

We have computed all three isomers at the B3LYP level. As

expected from the results of the TaCl4OH computations, stability
of isomers with axial OH groups is lower. Thus the relative
energy of the least stable isomer with two axial OH groups
(OH,OH-ax) is 1.8 kcal/mol above the energy of the most stable
isomer with two equatorial OH groups (OH,OH-eq). We have
computed only the first two geometric isomers in Scheme 2
and the transition state between them at all four levels of theory
(see Figure 2d-f and Table 5). The barrier for pseudorotation
is ∼3 kcal/mol for TaCl3(OH)2, which is 1 kcal/mol higher than
the barrier in TaCl4(OH). According to the results derived for
TaCl4OH, two patterns in the H-bond should be favorable
(Scheme 3).
We have found at B3LYP that the trans H-bond conformation
is less favorable than the cis conformation by 0.7 kcal/mol. Since
an extensive conformational analysis is beyond of the scope of
our paper we have not considered the trans H-bond conforma-
tion.
Thermochemistry of Gas-Phase Reactions in the System
TaCl5/H2O. Hydrolysis of TaCl5 and deposition of Ta2O5
include a series of elementary reactions of substitution of
chlorine by oxygen in the coordination sphere of tantalum. We
have computed some initial reactions between TaCl5 and H2O,
which may occur in the tantalum pentoxide CVD process in
the gas phase or on the growing Ta2O5 surface terminated by
Cl and OH groups. Zero Kelvin heats of reactions are presented
in Table 6. The first and second steps of TaCl5 hydrolysis have
similar reaction energies at all levels of theory. Since the
coordination of tantalum remains the same on the left and right
sides in eqs 1 and 2, correlation effects probably compensate
in the reaction energies of the two reactions. Such a compensa-
tion effect is not observed in the last three reactions, where the
coordination number of Ta becomes lower in the reaction course
and strong correlation effects may be expected. Note that even
in the nonisodesmic eqs 3-5 in Table 6 the B3LYP reaction
energies are close to CCSD(T) values within a 3 kcal/mol range,
while MP2 values differ by about 8 kcal/mol from CCSD(T).
Figure 3. Geometries of TaCl5-H2O complex and TaCl5-H2O
transition state optimized at the B3LYP level. Values at HF followed While the substitution of Ta-Cl bonds by Ta-OH in
by MP2 data are given in parentheses. Bond lengths are in Å, angles reactions 1 and 2 is exothermic, the formation of a TadO double
are given in degrees. bond is moderately endothermic in reactions 3 and 4 and

Figure 4. Calculated reaction path for the hydrolysis of TaCl5 at CCSD(T). Relative energies at HF, MP2, and B3LYP are given in parentheses.
Chemical Vapor Deposition of Ta2O5 J. Phys. Chem. A, Vol. 104, No. 6, 2000 1191

strongly endothermic in reaction 5, which certainly makes

tantalum oxychlorides, particularly TaO2Cl, difficult for ex-
perimental observation. The B3LYP calculation of the energy
of the reaction TaCl5 + H2O f TaCl3O + 2HCl (sum of the
energies of reactions 1 and 3 in Table 6) using a 6-31G* basis
set for chlorine gives 3.6 kcal/mol which is in reasonable
agreement with 0.6 kcal/mol obtained in the calculations using
an ECP for Cl.
The calculated reaction energies presented in Table 6 allow
to estimate heats of formation of all tantalum-containing
molecules in Table 6, based on ∆H°f values available for TaCl5
(-182.8 kcal/mol), H2O (-57.8 kcal/mol), and HCl (-22.1 kcal/
mol)39 (see Table 7). For the estimation of the heats of formation
we used the most reliable CCSD(T) values for the reaction
energies. Using these data and known heats of formation39 for
gaseous Ta (186.9 kcal/mol), Cl (29.0 kcal/mol), O (59.6 kcal/
mol), and OH (9.3 kcal/mol) we can also estimate the Ta-O
bond dissociation energies using a simple additivity approach
and the assumption that the Ta-Cl bonds have the same
(average) energy of 102.9 kcal/mol in all tantalum-containing
molecules, which can be estimated from the heat of atomization
of TaCl5 (Table 7). It is seen that the Ta-O single bond energies
are similar in TaCl4OH and in TaCl3(OH)2, which can be
(tentatively) used for validating an additivity approach in
calculating bond energies in other TaClx(OH)5-x molecules. Note
that the TadO double bond in TaOCl3 has a higher bond energy
than in TaO2Cl, indicating a partial triple bond character in C3V
symmetric oxotrichloride. The computed TaO bond energies can
also be compared with Ta-C (67 kcal/mol), TadC (126 kcal/
mol), and Ta-I (62.9 kcal/mol) avarage bond energies.40
Probably the ionic character makes the TaO bond much stronger
than the less polar (more covalent) TaC and TaI bonds. This is
particularly displayed in the Ta-O single bond, which is
predicted to have similar bond energy as the TadC double bond.
Assuming that the further hydrolysis steps have similar (∼
-7 kcal/mol) reaction energies as for reactions 1 and 2 in Table
6, we may use the difference in the estimated heats of formation
for TaCl4OH and TaCl3(OH)2 as an increment to calculate ∆H°f
for completely hydrolyzed Ta(OH)5. This gives a value of ∆H°f
) -395.8 kcal/mol, which can be used further to estimate the Figure 5. Geometries of the TaOCl3-H2O transition state and the
TaOCl3-H2O complex optimized at B3LYP. Values at HF followed
heat of reaction for conversion of hypothetical gaseous Ta(OH)5 by MP2 data are given in parentheses. Bond lengths are in Å, angles
into solid Ta2O5: are given in degrees.

2Ta(OH)5 f Ta2O5 + 5H2O ∆H°r ) 13.6 kcal/mol of TaCl5 hydrolysis (eq 1 in Table 6) and for TaCl3(OH)2
dehydration and for TaCl4OH dehydrochlorination (eqs 1, 4,
A calculation based on tabulated heats of formation of gaseous and 3, respectively, in Table 6).
TaCl5 and H2O and solid Ta2O5 (-489.0 kcal/mol)39 gives a Mechanism of Hydrolysis of TaCl5. High valent transition
negative (exothermic) value of the reaction heat for the metal compounds easily form complexes with electron donors
formation of Ta2O5: increasing the coordination number of the central atom. Ac-
cording to our calculated results TaCl5 forms a strong donor-
2TaCl5 + 5H2O f Ta2O5 + 10HCl acceptor complex with H2O (see Figure 3a and Figure 4). The
∆H°r ) -55.4 kcal/mol stabilization energy of the complex varies within a 6 kcal/mol
range between 20 kcal/mol value at HF and 26 kcal/mol at MP2,
Entropy effects favor formation of tantalum pentoxide for both with the B3LYP value being close to the HF value and MP2
reactions, particularly at higher temperatures applied in CVD. matching CCSD(T). The Ta-O distance in the TaCl5-H2O
Although thermodynamic calculations are important in esti- complex is about 0.5 Å longer than in tantalum hydroxychlo-
mating the “chemical driving force”, energy balance and relative rides: TaCl4OH and TaCl3(OH)2 (Figure 2). The calculated
stability of the intermediates, and products of deposition, in most 2.315 Å at B3LYP can be compared with the experimental Ta-
cases CVD processes occur at conditions which are far from OH2 distance in Cp*Ta(OH2)(p-tert-butylcalix[4]arene), which
thermodynamic equilibrium. The film structure and composition is equal to 2.188 Å.41
are governed by other factors: chemical kinetics, surface and The transition state energy in the reaction of TaCl5 hydrolysis
gas diffusion, and mass transfer. To reveal important details of is lower than the energy of the separated TaCl5 and H2O
potential energy surfaces for the types of reactions which may molecules at all but the HF level. Overestimation of the reaction
occur in tantalum pentachloride CVD, we have computed barrier is typical for the Hartree-Fock method, while B3LYP
transition states and intermediate complexes for the first step and CCSD(T) give similar barriers within a reasonable 3 kcal/
1192 J. Phys. Chem. A, Vol. 104, No. 6, 2000 Siodmiak et al.

Figure 6. Calculated reaction path for the dehydration of TaCl3(OH)2 at CCSD(T) without (top) and with (bottom) water assistance. Relative
energies at HF, MP2, and B3LYP are given in parentheses. Relative energies for the water-assisted reaction are calculated with respect to separated
TaCl3(OH)2 and H2O molecules.

Figure 7. B3LYP optimized geometries of TaCl3(OH)2, transition state and TaCl2O-H2O complexes with assisting water molecule. Values at HF
followed by MP2 data are given in parentheses. Bond lengths are in Å, angles are given in degrees.

mol range. Thus our calculations predict that in the gas phase than in the TaCl5-H2O complex, while the Ta-O distance is
TaCl5 hydrolysis should occur without a barrier. However, intermediate between the values in the complex and in TaCl4-
considering the same reaction of TaCl5 adsorbed on the surface OH.
(or similar reaction for a TaClx surface fragment), we must take We have not been able to locate a TaCl4OH-HCl complex
into account the surface as a part of the reacting system. The with six-coordinated tantalum on a “product site” similar to the
energy released in formation of TaCl5-H2O complex (∼26 kcal/ TaCl5-H2O complex on the “reactant site”. At each level the
mol) dissipates in surface and bulk film vibrational modes. Thus optimization did collapse to a hydrogen bonded form of TaCl4-
the barrier for the reaction on the surface is not equal to the OH‚‚‚ClH without formation of a donor-acceptor bond between
energy difference between separated TaCl5/H2O and the transi- Ta and Cl.
tion state but the one between TaCl5-H2O complex and TS We calculated next the structures and energies along the path
(∼20 kcal/mol). of hydrolysis of tantalum hydroxytetrachloride. TaCl4OH forms
The Ta-Cl distance in the transition state (see Figure 3b), a strong complex with water (-21.6 kcal/mol at B3LYP), in
which corresponds to a leaving Cl atom, is about 0.5 Å longer which H2O is positioned at the backside of the OH group. The
Chemical Vapor Deposition of Ta2O5 J. Phys. Chem. A, Vol. 104, No. 6, 2000 1193

Figure 8. Calculated reaction path for the dehydrochlorination of TaCl4OH at CCSD(T): (a) direct decomposition (b) water assisted reaction.
Relative energies at HF, MP2, and B3LYP are given in parentheses.

Ta-O bond length in TaCl4OH-H2O is 2.289 Å at B3LYP, that another isomer with one axial and one equatorial OH group
which is 0.026 Å shorter than in TaCl5-H2O. The hydrolysis (Figure 2f) is the direct product of the reverse reaction. We
of TaCl4OH proceeds similar to the hydrolysis of TaCl5 with have found computationally a transition state between two
transition state energy close to the energies of separated TaCl4- TaCl3(OH)2 isomers (Figure 2e), which is discussed above.
OH and H2O molecules: -2.4 kcal/mol at B3LYP. Since the transition state between the TaCl3(OH)2 isomers has
Mechanism of Dehydration of Tantalum TaCl3(OH)2. The a much lower energy than the transition state for H2O elimina-
termination of tantalum pentoxide film surface by the OH group tion, Berry pseudorotation does not play a significant role in
is most plausible under atmospheric condition or exposure to the reaction of TaCl3(OH)2 dehydration and can be ignored in
water under deposition condition, which include H2O as a most of the kinetics modeling studies.
precursor (oxygen source). According to Aarik and co-workers14 Also TaOCl3 forms a strong complex with water with a
the film etching, which occurs at higher TaCl5 concentrations stabilization energy at ∼20 kcal/mol. The water molecule is
and elevated temperature, can be explained by the formation positioned trans to the Ta-Cl bond. The relative energy of
of intermediate volatile tantalum oxychlorides, e.g., TaOCl3. TaOCl3-H2O with respect to TaCl3(OH)2 depends substantially
Formation of TaO2Cl is unfavorable because of a very high on the computational method. MP2 shows the largest deviation
endothermicity for its formation (see Table 6). from CCSD(T).
We optimized the stationary structures, minima, and transition Since in a TaCl5/H2O f Ta2O5 CVD process water vapor is
states, for TaOCl3 formation from TaCl3(OH)2 (Figure 2d-f used as a precursor, we have also considered the dehydration
and Figure 5). The most stable isomer of TaCl3(OH)2 (both of TaCl3(OH)2 assisted by one water molecule. The catalytic
OH- groups in equatorial positions, Figure 2d) was used as effect of external H2O is provided by increasing polarity
the starting point of the study of dehydration of tantalum (microsolvation) and decreasing the strain in the transition state
dihydroxytrichloride. However, the direct minimum energy path for dehydration. The energy of two H-bonds in the intermediate
from the transition state (see Figure 6) to the reactant shows TaCl3(OH)2-H2O complex (see Figures 6 and 7a) is 9.6 kcal/
1194 J. Phys. Chem. A, Vol. 104, No. 6, 2000 Siodmiak et al.

bond is elongated by ∼0.5 Å. Compared to the equilibrium

positions in HCl (1.295 Å) and in TaCl4OH (0.968 Å) the
corresponding ClH and OH bond lengths are elongated by 0.354
and 0.283 Å, respectively.
Water assistance significantly reduces the reaction barrier for
HCl elimination from TaCl4OH. However, in contrast to the
TaCl3(OH)2/H2O system, where the transition state structure
resembles the complex of TaCl3(OH)O- with H3O+ (Figure 7b),
the TS for water assisted HCl elimination is product-like (see
Figure 9b) and closely corresponds to the H-bonded complex
of TaCl3O, H2O, and HCl. We have only located the TS for
H2O assisted HCl elimination at the HF level. The dominating
component in the imaginary frequency vector corresponds to
the proton transfer between H2O and HCl, while the corre-
sponding vector in the TS for water assisted H2O elimination
from TaCl3(OH)2 shows synchronous motion of two protons.
Neither transition state nor the subsequent H-bonded complex
of TaCl3O/H2O/HCl has been located at the correlated levels,
B3LYP and MP2 (Figure 8b). We believe that the water assisted
decomposition of TaCl4OH occurs without a barrier, when
correlation energy is included. This is also (indirectly) confirmed
by a very low single point CCSD(T) energy (-10 kcal/mol)
computed at the HF optimized structure of the TS (see Figure
9b).

Summary and Conclusion

The results of our theoretical investigation can be summarized
as follows:
1. The calculated geometries and vibrational frequencies of
the high-valent tantalum compounds which are predicted at the
HF, MP2, and B3LYP levels of theory are in good agreement
with experimental results of known species. The reaction
energies and complex formation energies which are obtained
at the CCSD(T) and B3LYP levels do not exhibit large
deviations from each other, while the MP2 energies appear to
be less reliable.
2. The experimentally yet unknown compounds TaOCl3 and
TaO2Cl have been characterized theoretically. The formation
of TaO2Cl from TaCl5 and TaOCl3 is a thermodynamically
Figure 9. B3LYP optimized geometries of the transition states for disfavored process.
elimination of HCl from TaCl4OH in (a) direct decomposition (b) water 3. The calculations predict that TaCl5 hydrolysis in the gas
assisted reaction. Values at HF followed by MP2 data are given in phase should take place with a small or no barrier at all. The
parentheses; *values only at the HF level. Bond lengths are in Å, angles first step of the reaction is the formation of strongly bonded
are given in degrees. water complex, which eliminates HCl with a substantial barrier
with respect to the complex.
mol at CCSD(T). The transition state in the H2O assisted 4. Water loss from TaCl3(OH)2 and HCl loss from TaCl4OH
dehydration of TaCl3(OH)2 (Figure 7b) qualitatively corresponds have been studied as model reactions for etching of Ta2O5 films
to the complex of negatively charged TaCl3(OH)O- with H3O+ at high TaCl5 concentrations at elevated temperatures. Both
cation. The actual NBO charge transfer between these two reactions have large barriers for the unimolecular pathway. The
fragments is 0.69 electrons in NBO analysis. The barrier for activation barrier becomes much lower through the catalytic
the water-assisted hydrogen transfer to TaCl3O-H2O complex assistance of a water molecule, which opens a barriereless
is now reduced from 22.6 kcal/mol to 4 kcal/mol at CCSD(T). reaction channel for reactions in the gas phase. TaOCl3 forms
Such phenomenon of microsolvation in the gas phase has already a strong water complex like TaCl5 does with a similar complex
been reported.42 formation energy. The role of the assisting water molecule in
Mechanism of HCl Loss of TaCl4OH. Another reaction the etching process is emphasized.
which may lead to the formation of TaOCl3 is dehydrochlori-
nation of TaCl4OH (reaction 3 in Table 6). This reaction presents Acknowledgment. This work was supported by the Deutsche
a simplified model of HCl chemical desorption from the Ta2O5 Forschungsgemeinschaft (Graduiertenkolleg Metallorganische
surface. We have studied a unimolecular (direct) and bimolecular Chemie). We acknowledge excellent service and generous
(water assisted) decomposition of TaCl4OH into TaOCl3 and allotment of computer time of the Computational Technology
H2O. The energy diagrams for both reactions are shown on Lab at Semiconductor Products Sector, Motorola, HRZ Marburg
Figure 8. The barriers for a unimolecular decomposition of and HHLRZ Darmstadt.
TaCl4OH with HCl elimination (20.4 kcal/mol) and TaCl3(OH)2
with H2O elimination (23.5 kcal/mol) are similar. In the References and Notes
transition state for the HCl elimination (see Figure 9a) the TaCl (1) Yun, J. H.; Rhee, S. W. Thin Solid Films 1997, 292, 324.
Chemical Vapor Deposition of Ta2O5 J. Phys. Chem. A, Vol. 104, No. 6, 2000 1195

(2) McKinley, K. A.; Sandler, N. P. Thin Solid Films 1996, 290-291, Ochterski, J.; Petersson, G. A.; Ayala, P. Y.; Cui, Q.; Morokuma, K.; Malick,
440. D. K.; Rabuck, A. D.; Raghavachari, K.; Foresman, J. B.; Cioslowski, J.;
(3) Rausch, N.; Burte, N. P. Microelectron. J. 1993, 421. Ortiz, J. V.; Stefanov, B. B.; Liu, G.; Liashenko, A.; Piskorz, P.; Komaromi,
(4) Treichel, H.; Mitwalsky, A.; Sandler, N. P.; Tribula, D.; Kern, W.; I.; Gomperts, R.; Martin, R. L.; Fox, D. J.; Keith, T.; Al-Laham, M. A.;
Lane, A. P. AdV. Mater. Opt. Electron. 1992, 1, 299. Peng, C. Y.; Nanayakkara, A.; Gonzalez, C.; Challacombe, M.; Gill, P. M.
(5) Tanimoto, S.; Matsui, M.; Aoyagi, M.; Kamisako, K.; Kuroiwa, W.; Johnson, B.; Chen, W.; Wong, M. W.; Andres, J. L.; Gonzalez, C.;
K.; Tarui, Y. Jpn. J. Appl. Phys. 1991, 30, L330. Head-Gordon, M.; Replogle, E. S.; Pople, J. A. Gaussian 98, Revision A.3;
(6) Zaima, S.; Furuta, T.; Yasuda, Y.; Lida, M. J. Electrochem. Soc. Gaussian, Inc.: Pittsburgh, PA, 1998.
1990, 137, 1297. (25) Hehre, W. J.; Radom, L.; Pople, J. A.; Schleyer, P. v. R. Ab Initio
(7) Nishioka, Y.; Shinriki, H.; Mukai, K. J. Electrochem. Soc. 1987, Molecular Orbital Theory; Wiley: New York, 1986.
134, 410. (26) (a) Cizek, J.; J. Chem. Phys. 1966, 45, 4256. (b) Pople, J. A.;
(8) Ingrey, S. J.; Westwood, Y. C.; Cheng, Y. C.; Wei, J. Appl. Opt. Krishnan, R.; Schlegel, H. B.; Binkley, J. S. Int. J. Quantum Chem. 1978,
1975, 14, 2194. 14, 545. (c) Bartlett, R. J.; Purvis, G. D. Int. J. Quantum Chem. 1978, 14,
(9) (a) Wold, A.; Bellavance, D. In PreparatiVe methods in solid-state 561. (d) Purvis, G. D.; Bartlett, R. J. J. Chem. Phys. 1982, 76, 1910. (e)
chemistry; Hagenmuller, P., Ed.; Academic Press: New York, 1972. (b) Raghavachari, K.; Trucks, G. W.; Pople, J. A.; Head-Gordon, M. Chem.
Müller-Buschbaum, Hk. Angew. Chem., Int. Ed. Engl. 1981, 20, 22. Phys. Lett. 1989, 157, 479. (f) Bartlett, R. J.; Watts, J. D.; Kucharski, S.
(10) Schäfer, H. Chemical transport reactions; Academic Press: New A.; Noga, J. Chem. Phys. Lett. 1990, 165, 513.
York, 1964. (27) Hay, P. J.; Wadt, W. R. J. Chem. Phys. 1985, 82, 299.
(11) Suntola, T. In Handbook of Crystal Growth 3, Thin Films and (28) Bergner, A.; Dolg, M.; Kuechle; W., Stoll, H.; Preuss, H. Mol.
Epitaxy, Part B. Growth Mechanics and Dynamics, Atomic Layer Epitaxy; Phys. 1993, 80, 1413.
Elsevier: Amsterdam, 1994. (29) Huzinaga, S. Gaussian basis sets for molecular calculations;
(12) Devine, R. A. B.; Chaneliere, C.; Autran, J. L.; Balland, B.; Paillet, Elsevier: Amsterdam, Oxford, New York, Tokyo, 1984.
P.; Leray, J. L. Microelectron. Eng. 1997, 36, 61. (30) (a) Ditchfield, R.; Hehre, W. J.; Pople, J. A. J. Chem. Phys. 1971,
(13) Koyama, H.; Tanimoto, S.; Kuroiwa, K.; Tarui, Y. Jpn. J. Appl. 54, 724. (b) Hariharan, P. C.; Pople, J. A. Theor. Chim. Acta 1973, 28,
Phys. 1994, 33, 6291. 213. (c) Francl, M. M.; Pietro, W. J.; Hehre, W. J.; Binkley, J. S.; Gordon,
(14) (a) Aarik, J.; Aidla, A.; Kukli, K.; Uustare, T. J. Cryst. Growth M. S.; DeFrees, D. J.; Pople, J. A. J. Chem. Phys. 1982, 77, 3654.
1994, 75, 180. (b) Kukli, K.; Aarik, J.; Aidla, A.; Kohan, O.; Uustare, T.; (31) Russo, T. V.; Martin, R. L.; Hay, P. J. J. Phys. Chem. 1995, 99,
Sammelselg, V. Thin Solid Films 1995, 260, 135. (c) Simon, H.; Aarik, J. 17085.
J. Phys. IV 1995, 5, C5-277. (d) Aarik, J.; Kukli, K.; Aidla, A.; Pung, L. (32) (a) Spectroscopic Properties of Inorganic and Organometallic
Appl. Surf. Sci. 1996, 103, 331. Compounds; The Royal Society of Chemistry, Garden City Press Limited:
(15) (a) Doren, D. J. In AdVances in Chemical Physics, Vol. XCV; London, 1984; Vol. 16, p 762. (b) Dobs, K. D.; Hehre, W. J. J. Comput.
Progogine, I., Rice, S. A., Eds.; Wiley: New York, 1996; p 1. (b) Simka, Chem. 1987, 8, 861. (c) Benson, M. T.; Cundari, T. R.; Lim, S. J.; Nguyen,
H.; Hierlemann, M.; Ulz, M.; Jensen, K. F. J. Electrochem. Soc. 1996, 143, H. D.; Pierce-Beaver, K. J. Am. Chem. Soc. 1994, 116, 3966. (d) Sosa, C.;
2646. (c) Hierlemann, M.; Werner, C.; Spitzer, A. J. Vac. Sci. Technol. B. Andzelm, J.; Elkins, B. C.; Wimmer, E.; Dobbs, K. D.; Dixon, D. A. J.
1997, 15, 935. (d) Ho, P.; Colvin, M. E.; Melius, C. E. J. Phys. Chem. Phys. Chem. 1992, 96, 6630. (e) Russo, T. V.; Martin, R. L.; Hay, P. J. J.
1997, 101, 9470. (e) Harris, S. J.; Kiefer, J.; Zhang Q.; Schoene, A.; Lee, Chem. Phys. 1995, 102, 9315.
K.-W. J. Electrochem. Soc. 1998, 145, 3203. (f) Allendirf, M. D.; (33) Rosenkilde, C.; Voyiatzis, G.; Jensen, V. R.; Ystens, M.; Østvold,
Teyssandlier, F. J. Electrochem. Soc. 1998, 145, 2167. (g) Korkin, A. A.; T. Inorg. Chem. 1995, 34, 4360.
Cole, J. V.; Sengupta, D.; Adams, J. B. J. Electrochem. Soc. 1999, 146, (34) (a) Structure Data of Free Polyatomic Molecules; Springer: Berlin,
4203. 1987. (b) Modern Organic Chemistry; Marcel Dekker: New York, 1973;
(16) The preliminary results were presented at E-MRS 1999 Spring p 421.
Meeting in Strasbourg, 1-4 June 1999. (35) Figueira, R. R.; Fournier, L. L.; Varetti, E. L. Spectrochim. Acta
(17) Frenking, G.; Antes, I.; Böhme, M.; Dapprich, S.; Ehlers, A. W.; Part A 1982, 38, 965.
Jonas, V.; Neuhaus A.; Otto, M.; Stegmann, R.; Veldkamp, A.; Vyboish- (36) Beattie, I. R.; Livingstone, K. M. S.; Reynolds, D. J.; Ozin, G. A.
chikov, S. F. ReV. Computat. Chem. 1996, 8, 63 and references cited there. J. Chem. Soc. A. 1970, 1210.
(18) Ischenko, A. A.; Strand, T. G.; Demidov, A. V.; Spiridonov, V. P. (37) Landis, C. R.; Cleveland, T.; Firman, T. K J. Am. Chem. Soc. 1995,
J. Mol. Struct. 1978, 43, 227. 117, 1859.
(19) Kang, S. K.; Tang, H.; Albright, T. A. J. Am. Chem. Soc. 1993, (38) Del Bene, J. E.; Person, W. B.; Szczepaniak, K. J. Phys. Chem.
115, 1971. 1995, 99, 9, 10705.
(20) Benson, M. T.; Cundari, T. R.; Li, Yu; Strohecker, L. A. Int. J. (39) Chase, M. W. NIST-JANAF Thermochemical Tables, Fourth
Quantum Chem.: Quantum Chem. Symp. 1994, 28, 181. Edition, J. Phys. Chem. Ref. Data, Monograph 9, 1998.
(21) Deng, L.; Ziegler, T. Organometallics 1996, 15, 3011. (40) Lubin, L.; Liting, Li, Tobin, J. M. J. Am. Chem. Soc. 1997, 119,
(22) (a) Møller, C.; Plesset, M. S. Phys. ReV. 1934, 46, 618. (b) Binkley, 8574.
J. S.; Pople, J. A. Int. J. Quantum Chem. 1975, 9S, 229. (41) Acho, J. A.; Doerrer, L. H.; Lipppard, S. J. Inorg. Chem. 1995,
(23) Becke, A. D. J. Chem. Phys. 1993, 98, 5648. 34, 2542.
(24) Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Secures, G. E.; Robb, (42) (a) Zielinski, J. J.; Poirier, R. A.; Peterson, M. R.; Csizmadia, I.
M. A.; Cheeseman, J. R.; Zakrzewski, V. G.; Montgomery, J. A. Jr.; G. J. Comput. Chem. 1983, 4, 419. (b) Nguyen, M. T.; Ha, T.-K. J. Am.
Stratmann, R. E.; Burant, J. C.; Dapprich, S.; Millam, J. M.; Daniels, A. Chem. Soc. 1984, 106, 599. (c) Nguyen, M. T.; Hegarty, A. F.; J. Am.
D.; Kudin, K. N.; Strain, M. C.; Farkas, O.; Tomasi, J.; Barone, V.; Cossi, Chem. Soc. 1984, 106, 1552. (d) McKee, M. L. J. Phys. Chem. 1993, 97,
M.; Cammi, R.; Mennucci, B.; Pomelli, C.; Adamo, C.; Clifford, S.; 13608.