J Am Ceram SOC 73 [7] 2158-62 (19901

Kinetics of Titanium(lV] Chloride Oxidation

Sotiris E. Pratsinis,* Hebi Bai, and Pratim Biswast
Department of Chemical Engineering, Center for Aerosol Processes, University of Cincinnati,
Cincinnati, Ohio 45221-0171

Michael Frenklach
Department of Materials Science and Engineering, The Pennsylvania State University,
University Park, Pennsylvania 16802

Sebastian V, R. Mastrangelo
Chemicals and Pigments Department, E. I. du Pont de Nemours and Company,
Edgemoor, Delaware 19809

The oxidation rate of TiCI4 vapor was measured by FTIR reaction rate of TiC14 in their measurements. Taking this
spectroscopy between 700” and 1000°C in a furnace aerosol effect into account, they proposed a first-order reaction rate
reactor. The reaction rate was found to be first order with with respect to TiC14with activation energy 42 kJ/mol. These
respect to TiC14 and to follow the Arrhenius form. The ap- studies were confined to specific reactant ratios or did not
parent activation energy for the reaction was 88.8 f measure the complete reaction rate of TiC14 oxidation.
3.2 kJ/mol and the preexponential factor 8.26 X lo4 s-’. A This paper reports systematic measurements of the TiC14
kinetic mechanism for the oxidation of TiCI4 was proposed oxidation rate at various reactant ratios, reactor residence
that was consistent with the observed dependence of oxygen times, and temperatures in a furnace flow reactor. These data
concentration on the apparent rate constant. [Key words: ki- are used to propose a global rate expression for oxidation of
netics, titanium, chlorides, oxidation, titania.] TiC14 at various process conditions.

I. Introduction 11. Experimental Procedure

(1) Apparatus
M ORE than 2 x lo6 tons of titania are produced annually,
and used primarily in the pigment industry.’ This pow-
der is commonly made by oxidation of TiCI4 vapor in tubular
A schematic of the experimental apparatus is shown in
Fig. 1. Purified dried argon (Ar) carrier gas’ is bubbled
flow reactors by the so-called “chloride” process. Although through a fritted glass disk into a gas washing bottle
this is a mature technology and several patents on reactor de- (a) containing liquid TiCI4 (99.9%).’This bottle lies on a top-
sign, feed additives, and reactant mixing have been filed, the loading balancen that is used to monitor the entrainment of
fundamentals of the process are not well understood? Even TiCI4 vapor in Ar. The mixture of Ar-TiC14 vapor is diluted
basic chemical kinetics and the global reaction rate for the
oxidation of TiC14 are not available. It is of special interest
to understand the fundamentals of titania powder production
since it is the cheapest submicrometer-sized ceramic powder *Wright Brothers Inc, Cincinnati, OH.
made on an industrial scale. This can contribute to the de- *Aldrich Chemical Co., Milwaukee, WI.
‘Accuracy 0.01 mg, Fisher XT-100, Pittsburgh, PA.
velopment of large-scale processes for production of other
ceramic powders for manufacture of electronic or struc-
tural ceramics.
Antipov et ale3measured the conversion of TiCI4vapor in a
high-temperature reactor at 1:3 TiC14-to-02ratio and found
the apparent activation energy of the reaction as 76.6 kJ/mol.
Suyama and Kato4 measured the conversion of various halide

eP
vapors in excess 0 2 in horizontal and vertical furnace flow
reactors. They estimated the activation energy of the oxida-
tion of TiCI4 as 71 kJ/mol. Mahawili and Weinberg’ studied
the oxidation of TiC14 in a rotating plasma jet reactor and
found the reassociation of C1 radicals to affect the global
Of I
Ar
G. Messing-contributing editor

+
Manuscript No. 197902. Received December 14, 1989; approved April 6,
1990.
Supported by the National Science Foundation under Grant No. CTS-
8908197 and by E. I. du Pont de Nemours. Support for M. F. provided h the Treatment
Center for Particle Science and Engineering at The Pennsylvania &ate and
University. Exhaust
*To whom correspondence should be addressed.
‘Department of Civil and Environmental Engineering. Fig. 1. Schematic of the experimental apparatus.

2158
July 1990 Communications of the American Ceramic Society 2159

with a controlled amount of additional Ar, combined with O2

and introduced into an aerosol reactor (b). The reactor is an
1/8-in.-I.D., 30-in.4ong alumina tube (99.8% of A1203)** that
is externally heated by a horizontal Lindberg furnace (c). The
- 1000

reactor effluents are rapidly mixed and cooled with Ar at

room temperature in a perforated quartz dilutor (d). The re-
-
U

actor effluents are passed through a filter (g) to remove the w

D:
product titania particles and then enter into a gas flow cell of 5
*
the FTIR spectrometer (h) for measurement. The gas exiting 500
the FTIR cell is neutralized with a sodium hydroxide solu- a
tion before exhausting it through the laboratory hood: Mass EE
flow controllers (e) (MKS 1259B) followed by rotameters (f)
are used for precise flow control. Traps are used in between
the mass flow controller and gas washing bottle to prevent
backflow of TiCI4 through the mass flow controller.
-30 -20 -10 0 10 20 30
(2) Measurements
AXIAL POSITION ALONG REACTOR TUBE (cm)
The concentration of TiC1, vapor is measured by FTIR
spectroscopy in a 15-cm-long, flow gas cell7 that has been Fig. 3. Temperature profile for four flow rates through the reactor
made with KBr windows to allow measurements at 498 cm-', at 1000°C.
the strongest absorbance frequency of TiCl,.' The analysis is
carried out using an infrared spectrometer.++For each spec- temperature. Simple heat-transfer calculations indicated that
trum 16 scans are taken at a speed of 20 kHz and resolution >96% of the reaction occurs in this isothermal zone. This
of 2 cm-'. A calibration curve (Fig. 2) for TiCI4 is developed zone defines the effective reactor length (30 cm), resulting in
and used to relate absorbance to actual compound concentra- an effective reactor volume of 2.38 -C 0.12 cm3.
tion. This curve was developed by recording the weight of the The conversion of TiCl, vapor was measured at furnace
TiCl,-containing bottle before and after each experiment temperatures ranging from 700" to 1000°C and residence
carried out at constant Ar flow through the bottle. The con- times ranging from 0.27 to 1.27 s in the isothermal zone of
centration of TiC14 in the Ar stream leaving the bottle was the reactor. The reactant TiCI, : O2 ratios ranged from 1:1 to
determined by dividing the net weight loss of liquid TiC1, 1:40 in excess of Ar (greater than about 99% by volume).
by the product of the duration of the experiment and the
Ar flow rate.
The temperature is not uniform along the axis of the 111. Results
reactor. Detailed temperature measurements by a copper- The overall reaction of TiCI4 and O2 is described by the
constantan (K type) thermocouple along the reactor axis in the equation
absence of reactant gases have been made to determine the
temperature distribution in the reactor as a function of Ar TiCI4 + O2 + T i 0 2 + 2C12
flow rate.6 Figure 3 shows typical axial temperature profiles
Although the detailed mechanism of this reaction is not
in the reactor at various Ar flow rates at a set temperature of
known, it is assumed that the kinetics of TiC1, oxidation are
1000°C. Similar axial temperature profiles have been obtained
first order with respect to TiCI,, implying that there is a slow
at all furnace set temperatures. At low gas flow rates (up to
step involving one TiC1, molecule. Further it is assumed that
3 L/min), a nearly isothermal zone exists between 15 and
the reaction rate of the latter step is described by the Arrhe-
45 cm from the reactor entrance. In this zone, the tempera-
nius law. This approach has been successfully used in meas-
ture remains constant and nearly identical to the furnace set
urement and modeling halide o~idation','~as well as silane
decomposition." The ratio of the outlet to inlet TiC1, concen-
**Coors Ceramics Inc., Golden, CO.
tration, x = Co/Ci, can be related to the reactor residence
"Digilab FTS-40, BIORAD, Cambridge, MA. time, t, through the apparent Arrhenius activation energy E
and preexponential factor, A:9
In [- (In x ) / t ] = -E/RT + In A (1)
I where T is the reactor temperature and R is the gas constant.
I The TiCI4 conversion was measured at various reactant
ratios, reactor residence times, and temperatures. At each
ratio the validity of Eq. (1) was investigated by fitting it
through the respective TiCI4 conversions at all t and T and
examining the coefficient of determination ( r 2 )of the regres-
sion. Figure 4 shows the Arrhenius plot for the 1:5 TiC14-to-
O2 ratio at various residence times. It is clearly seen that
Eq. (1) nicely fits these data in the employed temperature
range. The coefficient of determination, r2 = 0.93, provides
further support for the validity of the assumptions made to
develop Eq. (1). Similar results were obtained for all reactant
ratios (Table I). All coefficients of determination are statisti-
cally significant at the 95% confidence level. The measured
average activation energy is about 15% to 20% higher than
the activation energies measured in past flow reactor stud-
0
I ies3z4This is common variability for determination of global
I I I I -r-- chemical kinetics.'" In addition, the present data provide a
complete expression for the oxidation rate of TiC1,. It should
be noted, however, that our activation energy is twice as large
Fig. 2. Calibration curve of TiC14 concentration with FTIR. as the one obtained in a recent plasma reactor study.5 This
2160 Communications of the American Ceramic Society Vol. 73, No. 7

F
P
0
”w 0.35 -
T ’
’P
’
2
z
D:
El 0.10-
3 d __--
8-’

E
u?
D:
Lr -0.15 -
0 o 720 “C
s 740 “C
A 760 OC

/
-1I I I I I t I I I I , I I I I , 1
I
-0.40
-4.70 -4.20
I I
-3.70
I
-3.20
0.86 0.91 0.96 1.01 LOG. OXYGEN CONCENTRATION
T ( OK ) Fig. 5. Reaction rate constant of Tic14 oxidation as a function of
O2 concentration.
Fig. 4. Arrhenius plot of the oxidation rate of TiC14.
IV. Discussion
difference can be attributed to the high degree of ionization The oxidation of Tic14 and the formation of titania powder
that characterizes plasmas. in the present system can be understood in terms of the fol-
Interestingly enough Eq. (1) fits the experimental data even lowing sequence of chemical reactions. The process is ini-
at the equimolar reactant ratio, though the apparent activa- tiated by the temperature- and pressure-dependent thermal
tion energy is 19% lower than the average of the value ob- decomposition of TiCI4
tained in excess oxygen. The dependence of the rate of
reaction on the oxygen concentration at three temperatures TiCI4 + M + TiCI3 + C1 + M (2)
was further investigated by examining the effect of the latter followed by the low-activation-energy abstraction reaction
on the global reaction rate constant, k = A exp(-E/RT).
Specifically, the initial TiCI4concentration was maintained at TiCI4 + C1 -+ TiC13 + C12 (3)
2.5 x lo-’ mol/L while the oxygen concentration varied from
2.5 x to 1.1 x mol/L corresponding to variation of The TiC13 radicals undergo similar decomposition and ab-
O2 excess from 1 to 40. Figure 5 shows a log-log plot of k vs straction reactions
O2 concentration at 720”, 740”, and 760°C. The slopes of the TiC13 + M + C1 + M
+ TiCI2 (4)
curves are the apparent order of the reaction with respect to
02.These curves are qualitatively similar for all tempera- TiC13 + C1 + Tic]? + C12 (5)
tures. At the lower oxygen concentrations (log [Oz] < -3.7), forming Ti&, and so on. The radicals TiCI, produced in this
the slope is nearly zero, indicating that the reaction rate manner, havingx equal to 2, 1, and 0, can abstract a C1 atom
shows little or no O2 concentration dependence until it
from TiC14:
reaches a 10-fold excess. At higher oxygen concentrations, the
slope increases to about 0.5, indicating that the reaction rate TiCI4 + TiCI, -+ TiCI3 + TiCI,,, (6)
becomes half order with respect to oxygen. This may also ex-
plain the lower coefficient of determination obtained for the thus providing additional channels for the “activation” of
1:20 ratio in Table I. These data can be used to propose a TiCI4, or undergo radical disproportionation reactions
global rate for Tic14 oxidation in the form TiCI, + TiCI, -+ TiCl,-, + TiCI,,, (7)
-d[TiCl,]/dt = (k + k ’ [ 0 2’”)] [TiCI4] wherey = 3,2,1.
The radicals TICI, ( x = 3,2, 1,O) formed in reactions (2)to
where (7) are then oxidized by O2via
k = 8.26 x lo4 exp(-E/RT) s-l TiCI, +0 2 -+ TiO2Cl,-, + nC1 (8)
k‘ = 1.4 x 10’ exp( -E/RT) (L/mol)’”. s-l 01

and TiCI, + 0 2 + TiOCl,-, + OC1, (9)

TiOCI, + O2 -+ TiO2Cl,-, + OC1, (10)
E = 88.8 +: 3.2 kJ/mol.
where n = 1,. . . , x and m = 0,1,2. The TiO,CI, oxides can
readily coalesce, because the excess of energy accumulated in
Table I. Activation Energy ( E ) and Preexponential Factor the products of binary collisions
(A) of the Arrhenius Rate Constant for TiCI4 Oxidation at TiOkCI, + TiOkCI, (TiOk)zCl,,,-, + nC1
-+ (11)
Various TiCI4:O2Ratios along with the Coefficients of
Determination (r’) of the Regression Analysis is removed by breaking Ti-CI or 0-C1 bonds where
Ratio E (kJ/mol) A (C’) r2 k = O,l, 2. This means that reactions (ll),which initiate the
1:20 95.9* 4.7 1.40 x 10’ 0.87
coagulation process, should be fast, having near-collisional
rates. Subsequent coagulation reactions
1:lO 84.4 1.7 2.53 x lo4 0.97
1:5 101.2* 3.5 1.59 X 10’ 0.93 (TiOk),CI,+ (TiOk)jCl,-+ (TiOk)t+,Clz+y-,,
+ nC1 (12)
1:l 73.7 %2.8 6.10 x 103 0.90
88.8 * 3.2
will be equally fast. As coagulation reactions (11) progress
Average 8.26 x 104
with the increasing i and j , C1 atoms are being completely
July 1990 Communications of the American Ceramic Society 2161
removed from the cluster structure and further oxidation 0 2 + R -+ RO + 0 (22)
occurs through Eq. (lo), thus leading to the classical coagu-
lation kinetics lead to the increased production of very reactive 0 atoms and
OC1 radicals. Reaching concentrations comparable to or even
(Ti02)i + (Ti02)j + (TiOz),+, (13) larger than that of C1, the abstraction reactions of these active
describing the formation of titania particles4 chain carriers
Although the reaction mechanism presented above can be TiCl, + OC1, -+ TiCl,-l + OCl,,,
supported by the current advances in understanding of sev- ( x = 1,2,3,4; y = 91) (23)
eral related system^^'^^'^-^^ the critical thermochemical data
are not available to attempt a detailed kinetic simulation. begin to compete with the abstractions by C1 (i.e., with reac-
Nonetheless, the basic kinetic features of titania powder for- tions (2) and (4)). The overall reaction
mation in TiC14 oxidation can be obtained by making several
simplifying yet realistic approximations. First, let us consider TiCI4 + 0 3 TiCI3 + OC1 (V)
the following system of reactions that represent the general represents the principal kinetic effects of reactions (23).
features of the detailed reaction mechanism discussed above: Adding step (V) to steps (I) to (IV) and repeating the analysis
TiC14 2 R + C1 (1) in the limit of (19), we obtain
TiCI4 + C1 3 R + C12 (11) k = ki + kz[CI] + k5[O] (24)
R + O2 5 P + C1 Additional dependence on [O] may result from possible oxi-
dative attacks of 0 atoms on titanium chlorides, similarly to
P+P!%T
reactions (8) to (10). Since the concentration of 0 atoms de-
Reactions (I) and (11) in this overall scheme represent elemen- pends on the concentration of O2molecules (e.g., due to reac-
tary reactions (2) to (7) producing the TiCl, radical pool, tions (21) and (22)), Eq. (24) explains the dependence of the
lumped now into species R; overall reaction (111) summarizes overall rate coefficient on the oxygen concentration observed
oxidation reactions (8) to (lo), which form P, the oxyhalide experimentally at the large [02]-to-[TiC1,] ratios. If we as-
precursors for coagulation; and overall reaction (IV) ex- sume that 0 atoms are in partial equilibrium with 0 2which
,
presses the entire coagulation process, reactions (11) to (13) is a reasonable assumption for the conditions of the large O2
for all i and j . excess, we obtain
Assuming steady state for P, the rate of powder formation,
r, is given as k = k, + k&1] + k5K#202]"2 (25)
where K O is the equilibrium constant of O2 = 0 + 0. The
r = k4[PI2 = k3[R][02] (14) obtained one-half power dependence on the concentration
and assuming steady state for R , Eq. (14) takes the form of O2in Eq. (25) is in complete accord with the present experi-
mental findings.
Examination of Eq. (20), the expression for the overall rate
coefficient in the oxygen-independent regime, reveals that
the apparent activation energy can have values from about
where k-2 is the rate coefficient of the reverse reaction (11). 350 kJ/mol, the TiCI3-CI bond dissociation energy,I5 to about
Equation (14) defines the overall rate coefficient of titania 110 kJ/mol, a lower bound for the activation energy of the C1
powder formation and, under the assumptions made, the abstraction. The larger values will be attained when the ther-
overall rate coeffecient of TiC14 decomposition mal decomposition, reaction (2), dominates and the lower
ones when the C1 abstraction, reaction (3), dominates. Since
the former reaction has a higher activation energy than the
latter, the apparent activation energy of TiCI4 decomposition
At very low O2 concentrations, i.e., when should increase with the increase in reaction temperature.
This may explain the fact that the apparent activation energy
determined in this work for the TiCI4 decomposition is much
lower than the TiC13-C1 bond energy as compared to the
smaller differences reported' for oxidation of SiCI4, GeC14,
and POC13. Our experiments were carried out at lower tem-
peratures, 700" to 1000"C, than the experiments performed on
these other 1100" to 1300°C. Hence, assuming that
In this limit, the overall rate coefficient becomes dependent all these comparable processes are governed by similar ele-
on the oxygen concentration, which may explain the experi- mentary chemical reactions-thermal decomposition of the
mentally observed bending of the curves in Fig. 5 at the low- precursor molecule followed by the atomic/radical abstrac-
est concentrations of 0 2 .When inequality (17) is reversed, tion-the apparent activation energy relative to the dissocia-
i.e., when tion energy should be lower in our case, as was indeed
observed. The fact that the experimental value of 88 kJ/mol
k-dClzl Q k3[021 (19) is lower than the lower-bound estimate indicates the contri-
Eq. (16) has its limit bution of other low-activation-energy reactions that convert
TiC14into titanium chloride radicals, like reactions (6) and (7)
and possibly TiCI4 + O 2 -+ TiC13 + C102. Very recently
which explains the independence of the overall rate coeffi- spectroscopic studies of TiC14 oxidation at high temperatures
cient on the oxygen concentration observed experimentally (llOOo to 1500°C) at equimolar reactant ratios have shown that
for the moderate 02-to-TiC14concentration ratios. Ti0Cl2is in great abundance at the early reaction stages, indi-
At a large excess of O2 over TiCI4, C1-0 chemistry begins cating that Eq. (9) may be one of the most important elemen-
to contribute to the overall reaction progress. At these condi- tary reactions of the mechanism.16 The role of the many
tions, reactions conceivable secondary reactions can be established quanti-
tatively by detailed kinetic modeling when the required
0 2 + c1- oc1 + 0 (21) thermodynamic and kinetic data become available.
2162 Communications of the American Ceramic Society Vol. 73, No. 7

!Conclusions
I (1979).
6P. Biswas, X. Li, and S. E. Pratsinis, “Optical Waveguide Perform Fabri-
The reaction rate of TiCI4 oxidation is first order with re- cation: Silica Formation in a High-Temperature Aerosol Reactor,” J. Appl.
spect to TiCI4 and nearly zero order for O2 up to a 10-fold Phys., 65, 2445-50 (1989).
’D.L. Wood, K . L. Walker, J. B. MacChesney, J . R . Simpson, and
oxygen excess. The apparent activation energy calculated R. Csencsits, “Germanium Chemistry in the MCVD Process for Optical
from the data is 88.8 t 3.2 kJ/mol and the preexponential Fiber Fabrication,” J. Lightwave Technol., LT-5, 277-85 (1987).
factor is 8.26 x lo4 s-l at 700” to 1000°C. In the range of 8K. Nakamoto, Infrared and Raman Spectra of Inorganic and Coordina-
higher O2 concentration, the reaction rate was found to be tion Compounds. Wiley, New York, 1978.
9W. G. French, L. J. Pace, and V. A. Foertmeyer, “Chemical Kinetics of
half order with respect to 02.A detailed chemical kinetic the Reaction of SKI4, GeCI4, POCL,, and BCI, with Oxygen,” 1. Phys.
mechanism that accounts for the measured activation energy Chem., 82, 2191-95 (1978).
and dependence on reactant concentrations was proposed. laD. R. Powers “Kinetics of SiCI, Oxidation,”J. Am. Ceram. Soc., 61 [7-81
295-97 (1978).
“M. E. Coltrin, R. J. Kee, and G. H. Evans, ‘AMathematical Model of
References the Fluid Mechanics and Gas-Phase Chemistry in Rotating Disk Chemical
IT. J. Wiseman, “Inorganic White Pigments”; Ch. 4 in Characterization of Vapor Deposition Reactor,” J. Elenrochem. Soc., 136,819-29 (1989).
Powder Surfaces. Edited by G. D. Parfitt and K. S.W. Sing. Wiley, New 12S.W. Benson, “The Mechanism of Inhibition of Knock by Lead Addi-
York, 1976. tives, a Chain Debranching Reaction,” J. Phys. Chem., 92, 1531-33 (1988).
2(a) E. J. Mezey, “Pigments and Reinforcing Agents”; pp. 423-51 in Vapor W.W. Dean, M. Frenklach, and J. Phillips, ”The Catalytic Etching of
Deposition. Edited by C. F. Powell, J. H. Oxley, and J. M. Blocher, Jr. Wiley, Platinum Foils and Thin Films During Hydrogen Oxidation,”f. Phys. Chem.,
New York, 1966. (b) H. K. Bowen, “Basic Research Needs on High Tempera- 92, 5731-38 (1988).
ture Ceramics for Energy Applications,” Muter. Sci. Eng., 44, 1-56 (1980). 14M.Frenklach, “Kinetics of Powder Nucleation”; presented at the Third
31.V. Antipov, B. G . Koshunov, and L. M. Gofman, “Kinetics of t h e International Conference on Ceramic Powder Processing Science, San
Reaction of Titanium Tetrachloride with Oxygen,” J. Appl. Chem. USSR Diego, CA, Feh. 4-7, 1990.
(Engl. Transl.) 40, 11-15 (1967). I5M.W. Chase, Jr., C. A. Davis, J. R. Downey, Jr., D. J. Frurip, R.A.
4Y. Suyama and A. Kato, “Ti02 Produced by Vapor-Phase Oxygenolysis of McDonald, and A. N. Syverud, “JANAF Thermochemical Tables, 3d ed.,
TiCI4,”1. Am. Ceram. Soc., 59 [3-41 146-49 (1976). Part 1,”f. Phys. Chem. Ref: Data, 14, Suppl. l(1985).
sI. Mahawili and F. J. Weinberg, ‘RStudy of Titanium Tetrachloride Oxi- I6D,L. Hildenbrand and K. H. Lau, SRI International, personal commu-
dation in a Rotating Arc Plasma Jet,” AfChE Symp. Ser., 75 [186] 11-24 nication, 1990. 0