Chemical Engineering Science, 1965, Vol. 20, pp. 1107-l 113. Pergamon Press Ltd., Oxford.

ord. Printed in Great Britain.

The kinetics of ilmenite beneficiation in a fltidised chlorinator

C. M. LAKSHMANAN,~ H. E. HOELSCHER$ and B. CHENNAKESAVAN~

(Received 7 September 1964; in revised form 12 January 1965)

Abstract-The beneficiation of ilmenite, the preferential removal of the iron oxides by selective
chlorination, has been studied in a fluidized bed reactor. Temperatures from 7OOC-95OC,inlet
mole ratios of carbon monoxide (the reducing agent used) to chlorine (the oxidizing agent used) from
l-2, and a wide range of values of the ratio of ,initial bed weight to gas-flow rate were investigated.
The volumetric average rate of solid and chlorine consumption were followed as a function of
elapsed time from the start of a run. Rate equations correlating the results for both rates are presented.
The ?ates being found functions of temverature but not strongly dependent upon other operating
conditions. -

amounts of reducing agent, carbon, occurred

THE BENEFICIATION of ilmenite by preferential
chlorination of the combined iron oxides yields a
starting material, a titanium rich residue, for the
production of pure titanium tetrachloride by
further chlorination of the residue. This two step
chlorination process is potentially of industrial
importance.
Although ilmenite can be considered as FeOTiO,,
most of the naturally occurring ilmenite contains a
higher percentage of iron as Fe,O, and a higher
percentage of TiOz than the theoretical value of
52 per cent. The higher ferric oxide content in
ilmenite has been attributed to the presence of
arizonite (Fe,0,.3TiO,) [9] but recent investiga-
tions [7, lo] have shown that the naturally occurring
ilmenite is a mixture of haematite, ilmenite and rutile.
Thermodynamic studies by WILSKA [14] and
PATEL and JERE [l l] on the chlorination of titanium
bearing minerals have shown the possibility
preferential chlorination of iron from ilmenite at
all temperatures. PATEL and JERE [l l] have also
indicated that among the two oxides of iron present
in ilmenite, Fe,O, is chlorinated in preference to
FeO. During experimental studies on the chlorina-
tion of Travancore ilmenite using carbon as a
reducing agent, MANOCHA [8] and KARKAR and
PATEL [3] observed that preferential chlorination
of the iron oxide in the presence of insufficient
readily. The industrial advantages of the prefer-
ential chlorination of iron in ilmenite have been
described in the patent literature [12], [13] and, as
already indicated, in the technical literature.
However, the data presented in the literature are
usually descriptive and very little information is
available on the kinetics of the reaction.
DORAISWAMY et al. [I] studied the preferential
chlorination of iron in ilmenite and assumed that,
in gas/solid reactions such as this, the form of the
rate equation is identical with that for gaseous
reactions catalysed by solid surfaces. They express
the rate equation in the standard Langmuir-
Hinshelwood form. The data were interpreted in a
manner like that normally used for solid catalysed
gaseous reactions. It was concluded that the rate of
reaction of chlorine based upon the weight of iron
present increased with time. Since the solid enters
of into the reaction with only one constituent of the
solid surface undergoing reaction this seems
unlikely. The iron oxide exposed on the solid
surface must decrease in extent with time and hence
this reactant becomes less available as time passes.
In the data of DORAISWAMY, a notable increase in
reaction rate with increase in time was seen only at
900C and an examination of the data indicates that
the reactor was operated at this temperature with
insufEcient quantity of chlorine. This did give

t Present address: Regional Research Laboratory, Sachivalaya Marga, Bhubaneswar, Orissa, India.
$ Present address: Schools of Engineering and Mines, Room 402 Engineering Hall, University of Pittsburgh. Pittsburgh.
Pennsylvania 15213.
$ Present address: Alagappa Chettiar College of Technology, University of Madras, Madras 25, India.

1107
 C. M. LAKSJMANAN,
H. E. HOELSXIER,
and B. CHENNAKESAVAN
results similar to those obtained in a catalytic
reaction system.
DUNN [2] describes a process study rather than a
kinetic investigation in which titaniferrous ore is
beneficiated by chlorination, the rate being found
proportional to the weight of TiOz in the reactor.
In both the work of DUNNand that of DonAIswAMv,
the chlorinating gas used was a mixture of carbon
monoxide and chlorine.
EXPERIMENTALEQUIPMENTAND PROCEDURES
The object of this present study was to determine
the kinetics of the preferential chlorination of iron
oxide from ilmenite. The ilmenite was obtained
from Travancore beach sand. The fluidizing gas mix-
ture was chlorine and carbon monoxide. The range
of variables studied is shown in the following table :
Table 1
700C < T (temperature) < 950C
1 < r (ratio of CO to Clz in feed) < 2
9.02 cm/set < v (gas velocity) < 18.04cm/set
(measuredat 900C)
20 gm < WO(initial bed weight)< 75 gm
The apparatus used in the investigation is shown
schematically in Fig. 1. The fluidized bed reactor
consisted of a mullite tube 60 cm in length with an
( ) Product
u receiver
FIG. 1. Experimentalapparatus.
1108
inside dia. of 2.46 cm. A porcelain disk having 42
orifices each of 0.2 mm dia. served as the bed
support. The reactor is shown schematically in
Fig. 2. The preheating and reactor sections were
heated by two tubular furnaces of standard design.
The first had a capacity of 576 W and the second a
capacity of 760 W both being controlled by
suitable rheostats. The apparatus provided stand-
ard flow regulating and control devices, as indicated,
with suitable condensers and sample take-off lines
for analytical purposes.
Naturally occurring ilmenite freed from silica
sand particles by passing through a magnetic
separation was sieved and the -65 + 10 mesh
fraction was used in the reactor. The fluidization
characteristics of this material have been reported
elsewhere [5], [6]. The chemical composition of
this material is given in Table 2.
Carbon monoxide, chlorine, nitrogen, and carbon
dioxide were provided from standard supplies.
Table 2
Constituent % by Weight
TlOz 59.28
Fe0 10.84
Fez03 24.85
Rem. 4.95
(A1203, f&02, etc.)
Chlorine absorber assembly
 The kinetics of ilmenite beneficiation in a fluidised chlorinator

Thermocouple exit gas from the ferric chloride scrubber was sent
PVC gasket through the chlorine absorber for known periods of
L- 30 cm ___I
time to analyse for unreacted chlorine. Allowance
was made for the time lag required for the chlorine
b - to reach the absorber assembly from the gas inlet
75 cm
by noting the initial time of the chlorine absorption.
;
6cm
i FeCC, condeq After absorption for a particular time interval the
L 6 34 quickfit gas was diverted to a different set of absorbers for
joint
known time, thus the concentration of chlorine in
Asbestos / !

packing the exit gases was obtained.
,o At the end of a run, the reactant gas mixture was
Reactor_ shut off and nitrogen gas was introduced to purge
section all reactant gases and iron chloride in the reactor.
Bed support- Thickness 2.5 mm
The system was then allowed to cool. The condenser
was removed carefully and unreacted ilmenite in the
Mullite tube reactor was collected and weighed. Ferric chloride
i.d. 2.46 cm
o.d. 3.02 cm 0.2 mm (42 holes) was collected by washing the condenser, receiver
\ and the scrubber with dilute hydrochloric acid and
this solution was analysed to estimate the chlorine
Porcelain bed support content [4]. Unreacted chlorine absorbed in the
Gas preheating
section __
carbon tetrachloride and potassium iodide solution
i
was estimated by standard iodometric titration
Asbestos H /Cement techniques. A chlorine mass balance was made to
packing S check for loss of chlorine due to side reactions or
physical losses and an over-all mass balance on the
GI inlet solid was made by checking the weight of the charge
FIG. 2. Details of reactor construction.
against the combined weight and residue and iron
oxide reacted.
Details of the experimentation are reported
elsewhere [5]. An experimental run was started RESULTSAND DISCUSSION:GENERAL
by assembling the apparatus and starting the fur- All experimental data and details on analytical
nace. The system was equilibrated at a temperature techniques are available elsewhere [5]. The key
approximately 20C below the desired operating component in the system used for calculating the
temperature. The bed was kept in a fluidized state gas phase reaction rate was the chlorine, whereas
during this pre-run period using nitrogen. The gas the disappearance of Fe,03 was followed to
composition desired for the run was prepared and characterize the solid reaction rate. Overall material
mixed but continuously vented prior to the start balances on the chlorine always accounted for more
of the run. When the system was in readiness, the than 95 per cent of that fed. That this accounting
nitrogen was turned off and the pre-mixed heated was incomplete is probably due
run gases were sent to the reactor at the same time (1) to the many side reactions that occur in this
the temperature was increased to the desired run complex system
temperature. The temperature of the reactor during (2) to the fact that iron chloride obtained as the
any one run was held constant to within + 5C. product of the reaction was always estimated
Ferric chloride formed during the course of the as ferric chloride, and
reaction was condensed and collected in the product (3) to physical losses.
receiver. Traces of ferric chloride carried away Iron is present in the ilmenite both as ferrous and
by the exit gases were collected in the scrubber ferric oxides and no method is available for
containing dilute hydrochloric acid solution. The calculating the amount of the individual oxides

1109
 C. M. LAKSHMANAN,H. E. HOELSCHER
and B. CHENNAKESAVAN

reacted from the experimental data. All of the iron RESULTS AND DISCUSSION: REACTION OF SOLID
present in the ilmenite was assumed to be Fe,O,.
The rate of solid consumption was followed by
Chemical analysis shows the presence of 10.84
calculating W (the weight of solid remaining in the
per cent Fe0 and 24.85 per cent Fe,O,. If all the
system at elapsed time 0 from the start of the run)
iron present were calculated as Fe,O, for this sys-
using the measured outlet concentrations and a
tem the error introduced could then be as high as
material balance. An analysis based on the assump-
2.8 per cent. Considering the amount of chlorine
tion that the rate of weight loss from the solid is
required to convert the ferrous oxide to ferric
proportional to the exposed surface area (after
chloride versus the amount required to convert
some initial time period during which the solid
ferric oxide into ferric chloride, an error in the
surface reaction is zeroth order with respect to
chlorine consumption of approximately 3 per cent
exposed surface) yields:
could be made by assuming the iron in the ore to be
all in the higher oxidation state.
=A-B8 (1)
The process was followed by the analytical
techniques described above. To check the accuracy where W,, = initial weight of bed,
and reproducibility of this technique, replicate W = weight of bed at time 9
runs were made at 900C using 50 g of ilmenite at a
feed gas flow rate of 1000 cm3 (N.T.P.) per min for Plots of (W/ Wo)/3 vs. 8 yielded lines which were
different time durations, that is, 5, 10, 15, 20, 25, nearly straight as required by this equation, all of
and 30 min. The amount of chlorine in the exit which yielded an intercept between 0.99 and I.00
gas and the weight of unreacted solid in each considering the accuracy of the data, this value is
experiment were obtained. The results of these taken as unity [A = l] suggesting that an equation
experiments are shown in table 3. The amount of of the form:
chlorine unaccounted for averaged 5 per cent and
the error in the material balance on the solids was [&(&)13]
sC=BB (2)
found to be less than plus or minus 2 per cent.
The results show a high degree of reproducibility, should be tried. Figure 3 shows one such result for

Table 3. Batch experimental data?

Feed gasflow rate: 1000 cm3 at N.T.P. per min Temperature = 900C
Gas composition (CO/C12 ratio) = 15 Initial wt. of bed = 50 g
Inlet Cl2 flow rate = 400 cn?/min

Time Wt. of solid VE % Cl2 Vol.of Cl2 wt. of % Solid % Fe203

min. remaining (cm) unaccounted reacted Fez03 unaccounted reacted
0 vo- VE reacted
(cm) (9)

5 45.09 182 1622 9.8 1818 4.31 -1.20 23.5

5 45.07 176 1698 6.3 1824 4.34 -1.20 23.6
10 42.68 965 2855 4.5 3035 7.20 -0.24 39.2
10 42.66 1067 2835 2.6 2938 6.97 -1.14 38.0
15 39.33 1799 3828 6.2 4201 9.99 -1.36 54.5
20 38.12 3318 4311 4.6 4682 11.11 -1.54 60.5
20 38.18 2985 4580 5.5 5015 11.89 +0.14 64.7
25 35.73 4123 5332 5.4 5877 13.95 -064 76.0
25 37.13 4352 5330 3.2 5648 13.41 +1*89 73.1
30 34.70 5580 5741 5.6 6420 15.28 -o&I 83.2
30 35.66 5714 5763 4.4 6286 14.93 +1.18 81.4
t Gas volumes are expressed at N.T.P.
VE = Total volume of chlorine in exit gas, cm3
V9r = Total volume of chlorine in product, cm3
vo = Total volume of chlorine fed, cm3

1110
 The kinetics of ilmenite beneficiation in a fluidised chlorinator

40
r particularly, W,-JF,, (ratio of initial bed rate to
30
gas-flow rate) and through the range of velocities
t varied at constant initial value of (WJF,), that is,
20
Sample plot of f=BB
900C data r=l,l.5$8
for 9.05 < v < 18.01, B is found proportional to
u-o.33.
t /
(3) Analysis of 21 data points at r(ratio of CO to
Cl, in feed) = l-5( Y. = O-4), T = 900C and
F = 1000 cm3/min, and of an additional 21 at the
same value of Y. and T but holding W, constant at
50 while varying F, from 600 cm3/min to 1200
cm3/min, yielded a measure of the effect of Wo/Fo,
on B with the result:

(3)
I I I I I III
b-IO 020 0.30 0.50 I.0 15 202530
8
These will be discussed in turn.
FIG. 3. Sample plot of solid reaction rate data.
The fact that, through the available operating
(Correlation-Equation 2) range, the rates measured were not found to be
functions of inlet gas phase concentrations (for
illustration. The constant n was found to be inde- either solid or chlorine consumption rates-see
pendent of all operating variables listed in Table 1 the following section) indicates that the surface was
through the entire range of each studied in this always supplied with sufficient of both the reducing
experiment and a constant for all data (142 individ- agent (CO) and the oxidizing agent (Cl,) so that
ual data points) at a value of diffusion of neither provided a limitation upon the
n = 0.75 surface attack rates. This leads one to suggest that
Table 4 summarizes the results for the slope B. further work at both lower and higher ratios is
A plot of 1nB versus l/T (an Arrhenius plot, see required.
Fig. 4) indicates two ranges of operation, the differ-
ence being considered significant considering the
accuracy of the result reported on Table 4.
Arrhenius plot
Table 4. LN B vs $

Temp. B N* rt

iF0 0.0107 14 OGOO6

900 0.0093 46 0.0003
850 0.0082 6 0@002
800 0~0070 15 0%)03
700 0.0057 14 0.0002

* N = number of data points at each temperature

t y = r.m.s. deviation from B through n data points

Analysis of the effect of other variables on B

yielded the following conclusions : -5.5L I I I
8 9 IO II
(1) For values of Y0 (inlet mole fraction of Clz) I/T, OK x104
varying from 0.5 to 0.333 (the inlet chlorine mole
fraction), there is no observable effect of Y, on the FIG. 4. Arrenius plot of constant in correlation
equation for solid reaction rate data. Temperature
parameter B; coefficient: Region 1 (High temperature) E/R = 5130,
(2) For all other variables constant, most Region 2 (Low temperature) E/R = 2640.

1111
 C. M. LAKSHMANAN,
H. E. HOELSCHER
and B. CHENNAKE~AVAN

Both the second and third result listed above seem
at first to suggest that a diffusional component is
of some consequence to the reaction. They are,
however, more likely due to the influence of the bed
expansion ratio on conversion efficiency, the bed
being more dilute at higher velocities yielding a
lower conversion efficiency. It would seem (from
the third result) that the effect of bed expansion
ratio is of greater importance to conversion
efficiency than is residence time.
Although it is possible from the available data to
obtain the constants in the proportionality between
B and v-o.33 and between f/0o75and ( Wo/Fo)-12,
this is not considered desirable at present due to the
limited range of the operating variables covered in
this work. Further exploration of these variables is
necessary.
RESULTSANDDISCUSSION
: GASPHASEREACTION
The volumetric average rate of chlorine consump-
tion in the bed was calculated from the known inlet
and the measured outlet concentrations of chlorine
as a function of the independent operating variables
and elapsed time from the start of the run. Plots
of the fraction conversion vs. time factor, W/F, on
log-probability paper (with the conversion on the
probability scale and time factor on the log-scale)
gave excellent straight lines through the range of
the variables studied. This is perhaps to be expected
since it was previously determined and reported that
the residence time distribution in the bed was
essentially Gaussian [5], [6]. These plots were
useful for interpolation but proved unnecessarily
complex for correlation purposes.
It was found, empirically, that the volumetric
average chlorine consumption rate, R (moles/min/g
of solid present), and the fraction of solid con-
sumed could be related to the elapsed time
since the start of the run by an equation of the form:
correlation, a measure of the goodness-of-fit to the
data.
The value of l/0, for the 850, SOO,and 700C
data is probably best taken as zero within the limit
of accuracy of the data while the values at the two
higher temperatures are probably correctly non-
zero since the straight line plots of R/f versus 8
must approach the origin with zero slope. However,
the value of B. is that value beyond which the
correlation is not useful, possibly a measure of the
time required at each temperature for the generation
of a significant pore structure due to iron removal
from the solid. As expected, the plots ( W/Wo)3
versus 0 likewise tend to curve at long times,
undoubtedly for the same reason.
Equation (4) for the gas phase rate and equation
(2) for the solids disappearance rate completely
describe the system. That the correlation is suitable
within the accuracy of the data is evident from the
results of the statistical analysis reported.
CONCLUSIONS
The kinetics of the fluidized bed beneficiation of
ilmenite has been studied and results for both the
solid and gas phase reaction rates reported. The
correlations presented in this paper for the rate of
removal of iron from the solid and the consumption
of chlorine in the bed indicate that the process is
controlled by the availability of solid surface and ;is

(4)

wherein, f = 1 - (W/W,) = the fraction of solid

consumed when the average rate is R at tome 0.
A sample plot of this form is shown as Fig. 5. The
constants in this equation are shown in Table 5
along with the number of individual data points FIG. 5. Sample plot of gas phase reaction rate data.
used in the correlation and the coefficient of (Correlation-Equation 4)

1112
 The kinetics of ilmenite beneficiation in a fluidised chlorinator

not a function of the carbon monoxide/chlorine NOTATION

ratio in the inlet over the range investigated (i.e.
Constants in equations (1) and (4)
from 1 to 2) nor is it strongly a function of operating Constant in equation (1)
conditions other than temperature. Further work to [1 - ( W/ W0)1/3]as defined in equation (2)
determine thelowerlimit of carbon monoxide concen- [I - W/ WO] as detined in equation (4)
Volumetric feed gas flow rate, cm3 (N.T.P.)/min
tration in the feed gases for which these conclusions Constant in equation (2)
remain valid and the effect of bed expansion ratio Number of data points in Tables 4 and 5
on conversion efficiency is obviously required. Mole ratio of CO to Cls in feed gases
R Volumetric average rate of chlorine consumption,
Table 5. Constants in the correlation equation moles/min/g
R Gas constant

e- $01
!t= ,40e-2760/RT ' !-
T Temperature, C
f [ Gas velocity, cm/set
kF Weight of solid present at time eg
R = Chlorine conversion rate averaged over the
Wo Initial bed weight, g
reactor, moles/min/g of solid
Yo Inlet mole fraction of chlorine
-f%l_!!$ fraction of solid consumed r.m.s. deviation, see Table 4
! Coefficient of correlation--&e reference, Table 5)
T = Temperature ye Elapsed time from the start of the run
00 Constant defined in equation (4)
TC Ao l/80 Y* Nt
950 0.0602 0.0150 0.9842 16 Acknowledgement-One of the authors (C. M. LAKSH-
900 00553 O~OlOfl 0.9854 109 MANAN) was supported during the progress of the research
850 0.0574 0.0035 0.9848 6 activities from which this paper is in part derived by fellow-
800 0.0634 -0~0019 0.9808 16 ships from the University Grants Commission and the
700 0.0676 -0*0035 0.9901 17 Council of Scientific and Industrial Research of India. The
authors gratefully acknowledge this support. The authors
y* = Coefficient of correlation, a measure of the goodness also acknowledge the interest and support of Dr. G. S.
of fit LADDHA of the University of Madras, India, throughout
= 1 for perfect fit the work. Mr. DONALDDECKER assisted in preparation of
= 0 for zero correlation the manuscript and Miss DOROTHV SHAEFFERand Mr.
Ref: A. J. DUNCAN, Quality Control and Industrial Statistics, GOPAL SUBRAMANIAN were of much help in the calculations
p. 661. RevisedEdition (1959)Richard D.Irwin Inc. Illinois leading to the final correlations. For this assistance also, the
N number of data points used. authors express their thanks.

REFERENCES
DORAISWAMY L. K., BLJAWATH. C. and KUNTE M. V., Chem. Engrs. Progr. 1959 55 80.
DUNN W. E. Jr., Trans. Met. Sot. AIME 1960 218 6.
KARKARD. P. and PATELC. C., J. Sci. Znd. Res., 1958 17B 367.
KOLTHOFFI. M. and BELCHERRl, Volumetric A&ysis Interscience publishers, New York, Vol. 111 p. 342 1957.
LAKSHMANAN C. M., Studies in Applied Chemical Kinetics PhD Thesis, from the Univ. of Madras, Madras, India, 1964.
LAKSHMANAN C. M., HOELXHER H. E., and CHENNAKESAVOR B. Fluidization Studies in an Operating Fluidized Bed
Reactor, paper presented at symposium on Ruidization, Jan. 1964, Karagpur, India and to be published in the proceedings
of that symposium.
LYND L. E., SIOURDSONH., NORTH C. H., and ANDERSONW. E. Mining Engng. 1954 6 817.
MANOCHAR., Indian Inst. Metals 1953 7 95
MILLER R., Amer. Min. 1945 30 65.
OV~RHOLTJ. L., VAUX G. and RODD J. L., Amer. Min. 1950 35 117.
PATELC. C. and JERE G. V., Trans. Met. Sot. AZME, 1960 218 219
SAUREFABRIK SCHWEI~ERHALL, Swiss Patent 325,452 Dec. 31 1957 C.A. 52 11581 (1958).
TITAN Co. Inc., Brit. Patent 715,257 Sept. 8, 1954. C.A. 49 3769 1955.
WUSKA, S., SUOMENKEMISTILEHTI, 29-A, 220 1956. C.A. 51480 1957.
Zusammenfassung-In einem Wirbelbettreaktor wurde die Aufbereitung von Ilmenit, d.h. die bevor-
zugte Beseitigung der Eisenoxide durch selektive Chlorierung, untersucht. Die Versuche wurden bei
Temperaturen zwischen 700 und 95OC, bei Molverhaltnissen von Kohlenoxid (reduzierendes Gas)
zu Chlor (oxidierendes Gas) zwischen 1 und 2, und bei zahlreichen Wertendes Quotienten Bettgewicht
zu Gasgeschwindigkeit durchgeftihrt. Die Abnabme des Ilmenit-Gewichtes und der Verbrauch von
Chlor wurden w&rend der Reaktionszeit verfolgt; die Ergebnisse beider Messungen wurden in
Geschwindigkeitsgleichungen korreliert. Die Reaktionsgeschwindigkeit ist abtingig von der Tem-
peratur, Sindert sich jedoch nur wenig mit den anderen Betriebsbeclmgungen.

1113