Ind. Eng. C h e m . Res.
1988,27, 727-735
KINETICS AND CATALYSIS
Thermal Hydrolysis of Vegetable Oils and Fats. 1. Reaction Kinetics
T. A. Patil, D. N. Butala, T. S. Raghunathan, and H. S. Shankar*
Department of Chemical Engineering, I n d i a n Institute of Technology, Bombay 400 076, India
A model is proposed to describe the liquid-liquid thermal hydrolysis of vegetable oils and animal
fats. Extensive data on hydrolysis equilibrium and rate have been obtained. Results of the present
and previous studies have been compared with model predictions. Uniformly excellent agreement
is indicated in all the cases.
The direct saponification of oil/fat in small-scale soap
manufacturing units poses many problems with regard to
recovery of byproduct glycerol. The hydrolysis of oils prior
to saponification provides possibilities of overcoming this
problem. Fatty acids and glycerol are valuable chemical
intermediates with a variety of end uses. Yet quantitative
information on the kinetics, thermodynamics, and engi-
neering aspects of this reaction is limited.
The hydrolysis reactions can be conducted thermally as
a liquid-liquid reaction or as a gas-liquid reaction using
superheated steam. It can also be conducted using lipolytic
enzymes at ambient conditions. All three modes of hy-
drolysis are being investigated, but we present here our
results on liquid-liquid thermal hydrolysis.
Commercially the reaction is carried out under the
conditions of 100-260 "C and 100-7000 kPa using 0.4-1.5
w/w initial water to oil charge with or without catalysts.
Feed stock flexibility, high extent of hydrolysis, minimal
color degradation, and high concentration of glycerol in
the aqueous phase are among the principal process re-
quirements. Thus, several alternative such as the batch,
the semicontinuous, and the continuous countercurrent
processes are in use.
In the batch Twitchell process, the acid-washed fat is
contacted with water at 100 "C and atmospheric pressure
using 25-50% w/w water and 0.75-1.25% by its own
weight Twitchell reagent in covered wooden vats (Mueller
and Holt, 1948). Oxidative discoloration and high steam
consumption constitute the major drawbacks of this pro-
cess. No direct information on the rate and equilibrium
characteristics of this process is reported.
In the batch autoclave process, the reaction is conducted
in the presence of liquid water and wet steam under
moderate conditions of 1000-3000 kPa and 180-230 "C;
agitation is provided by steam or a mechanical stirrer. The
degree of hydrolysis of about 90% is generally obtained
which may be increased to 96% at the cost of glycerol
dilution by employing a higher water-to-oil ratio (Lascaray,
1949).
In the continuous countercurrent process, the preheated
fat is fed at the bottom of a hydrolyzing column under
pressure while water is sprayed from the top of the reactor.
The descending stream of water encounters the rising fat
counter currently at around 260 "C and 4700 kPa to ac-
complish the necessary contact needed for the reaction.
No catalyst is generally used (Burrows, 1953).
Hydrolysis of oils is clearly a case of chemical reaction,
accompanied by mass transfer. In this heterogeneous
0888-5885/88/ 2627-0727$01.50f 0
727
system, water and glycerol can distribute between the
aqueous and the oil phases. Thus, both reaction and phase
equilibria are featured in this reaction. The dependence
of the equilibrium acid value on the initial water-oil ratio
is available (Lascaray, 1949,1952). These results suggest
that the degree of hydrolysis at equilibrium depends on
the initial water-to-oil ratio alone and is independent of
temperature.
Product glycerol is found to have an adverse effect on
the final degree of hydrolysis, and the empirical equation
X,= 1.0 - 0 . 8 ~ ~(0.8 ~ IX,I1.0) (1)
is given by Lascaray (1952) to quantify this effect. Here
X,is the overall conversion and y g eis the glycerol mass
fraction in the aqueous phase at equilibrium. Equation
1 shows clearly that a possible strategy for design would
be to extract glycerol from the fat phase to shift the
position of equilibrium in the favorable direction, thus the
importance of the continuous countercurrent scheme.
Mills and McClain (1949) studied the equilibria in the
hydrolysis of beef tallow and coconut oil at 235 and 250
"C in the range of 90-100% completeness using a novel
stirred batch reactor. A t equilibrium, the oil phase was
found to contain small quantities of diglycerides, mono-
glycerides, and glycerol, providing evidence for the fol-
lowing stepwise reaction:
I. triglyceride + water diglyceride + fatty acid
11. diglyceride + water e monoglyceride + fatty acid
111. monoglyceride + water 6
glycerol + fatty acid
(2)
The glycerol distribution coefficient m ( = y g / x ) for beef
tallow and coconut oil at 250 "C was found to \e 12 and
6, respectively. The reaction is said to occur in the oil
phase and that water is unimportant in establishing the
equilibria because it is present in large molecular pro-
portions. However, no data on the concentration of water
in the oil phase are reported.
Sturzenegger and Sturm (1951) have reported batch
hydrolysis data for some oils. These studies indicate an
autocatalytic behavior with a clear induction period. The
overall rate of reaction is also seen to be a strong function
of temperature. The reaction is found not to go to com-
pletion and that equilibrium acid value is a function of
0 1988 American Chemical Society
728 Ind. Eng. Chem. Res., Vol. 27, No. 5, 1988
water to oil ratio alone. The acid value-time data reveal
oscillations of 2-3 units around the equilibrium region.
Despite the above features, a pseudefirst-order irreversible
kinetic interpretation with rate constant
(3)
based on half-life is given. An activation energy E I R of
7000-8000 K is indicated.
Technically, a number of reactor designs are possible.
However, very limited studies are reported on the behavior
of hydrolysis reactors. Jeffreys et al. (1961) assumed
overall stoichiometry and pseudo-first-order irreversible
kinetics to model a countercurrent spray tower as a case
of simultaneous reaction in the oil phase and glycerol
transfer to the aqueous phase. Reaction conditions are not
specified, but results showed abnormal HTU values.
An approach to describe the performance of a spray
column with axial dispersion model is presented by Don-
ders et al. (1968) using overall reaction stoichiometry. A
value of the second-order rate constant for forward reac-
tion, k, = 0.009 m3/(kmol.min) and K, = 2, is used. A
simple equation is suggested to calculate the dispersion
coefficient from superficial velocity of the dispersed phase.
A mass-transfer coefficient of kL = 0.001 m/min and in-
terfacial area of 87 m2/m3is suggested. No experimental
data or evidence is reported.
Basu (1976) presented data on the splitting of rice bran
oil in a countercurrent column and compared the same
with conversion predicted from irreversible pseudo-first-
order kinetics; considerable disagreementwas found. Desai
et al. (1984) modeled the course of batch hydrolysis using
the following assumptions: (1)reaction takes place in the
oil phase as per eq 2 with the first reaction controlling; (2)
diglycerides and monoglycerides are stationary at their
equilibrium values; (3) the concentration of water in the
oil phase is linearly related to fatty acid concentration in
the oil phase; (4) the phase equilibria for glycerol hold
during hydrolysis; (5) elementary kinetic representation
applies so that the equilibrium constant for the reactions
can be given as
(4)
with the overall equilibrium constant given as
K, = K1K2K3= [C,3C,/C,CW3], (5)
The model applicability is tested against acid value-time
data of Sturzenegger and Sturm (1951) on beef tallow.
Excellent agreement is indicated. Agreement with respect
to other oils was poor.
Summing up, the available information on this reaction
is inadequate. The equivalence between acid and glycerol
formation, the concentration of water in the fat phase, and
the role of mass transfer relative to reaction are not well
established. No useful model exists to predict the reactor
performance under different design and operating con-
ditions, hence the motivation for the present work.
The objective of this study is therefore to develop a
suitable model to cover some of the above inadequacies.
In this paper, a model containing four equilibrium pa-
rameters and one rate parameter is developed to describe
the course of the liquid-liquid thermal hydrolysis. The
model is tested against present and previous data, and
uniformly good agreement is revealed in all cases.
Mathematical Model
The model for batch hydrolysis developed here uses the
following assumptions:
(a) The reaction takes place in the oil phase as per the
scheme shown in eq 2. Isomers of mono- and diglycerides
which are possible are not explicitly accounted.
(b) The mass and density of the phases remain un-
changed during the reaction so that
where i refers to component.
(c) All reactions are elementary. Equilibrium constants
for all the steps are the same and are temperature inde-
pendent.
(d) The reaction belongs to the very slow reaction re-
gime, and phase equilibria for glycerol exist at all times
so that
Yg = mxg (7)
where g refers to glycerol.
(e) Water transport to the oil phase is rapid, and its
concentration can be given as
(f)The first step of eq 2 is rate controlling, and the other
two steps are relatively rapid.
(g) Glycerides and fatty acids are insoluble in the
aqueous phase.
Let X1, X2,and X, be the extents of reactions with
respect to triglyceride in the system; since there is no
change in volume due to reaction, we have by stoichiometry
Mass balance for triglycerides gives
dCt/dt = k2CdCa - klCtCw (13)
The other two reactions being relatively rapid, we get
Cm - -k3Cw c,
_ -=- k5Cw
(14)
Cd k4Ca cm k6Ca
Since the solubility of water in fatty acids is large relative
to glyceride, eq 14 can be written approximately as
The above approximation leads to a simple solution of the
problem. Overall glycerol balance gives
 Ind. Eng. Chem. Res., Vol. 27, No. 5, 1988 729

Table I. Specifications of Autoclave Used in the Present Three methods were tried for the determination of
Work glycerol content in aqueous phase viz. periodic acid oxi-
capacity 1L dation method, sodium metaperiodate oxidation method,
outer diameter of the vessel 0.111 m and specific gravity measurement (Perry and Chilton,
inner diameter of the vessel 0.08 m 1973). Standard samples of glycerol solution of 0.05-35%
depth 0.23 m w/w were prepaxed to evaluate the efficacy of the methods.
max pressure 50 000 kPa Excellent agreement was found between these over the
max temp 600 O C
stirrer type propeller range investigated. Therefore, the specific gravity method
stirring speed 150-1440 rpm which was simple and rapid was used in the present work
propeller diameter 0.05 m with frequent cross-checks against sodium metaperiodate
propeller width 0.017 m titrations. The error of the glycerol measurement by this
method was estimated as *0.026%.
Since mass and density are assumed constant when using The fat-phase glycerol analysis was carried out by
eq 6, we get shakiig a known amount of distilled water with a relatively
C, = Ca(AX3 + A ' ) (18) large fat-phase sample to extract the glycerol in water so
as to facilitate accurate measurement of aqueous glycerol
From eq 16 we get by the specific gravity method. The periodic acid oxidation
X3 = BX2 - B' (19) method was also used to confirm the measurements.
x2 = cx1- C' (20) Levels of monoglyceride were analyzed by the periodic
acid oxidation method. a-Monoglyceride could be mea-
Substituting for C,, Cd,C,, and C, in terms of X 1 in eq 13 sured by this method. Total monoglyceride was calculated
we get as per the procedure given by Mehlenbacher (1960). A
-P dX,/dt = XI2 - Q X 1 - R (21) measurement error of Zm f 0.0008 is indicated here.
Only clean oil free from mono- and diglyceride was used
The solution to eq 21 is given by in the experiments. In general, acid value (AV), saponi-
aP(1 - eNt) fication value (SV), aqueous glycerol fig), and mono-
x1= glyceride (Zm)were measured. A balance for fatty acid is
a - fieNt used to obtain the overall conversion in terms of measured
where a and fi are the roots of the right-hand side of eq quantities as shown below:
21. ( X i + Xz + X 3 ) / 3 = AV/SV - AVo/SVo (23)
Experimental Apparatus The quantities X 2 and X 3 could be expressed in terms
A laboratory autoclave of 1-L capacity made from of X 1 by using eq 19 and 20. The resulting equation in
stainless steel 316 mounted in a compact, electrically X I could be expressed in terms of elapsed time by using
heated furnace was used. The specifications of the auto- eq 22. Thus, we could relate the measured AV and SV
clave are summarized in Table I. The top flange is pro- values to time, and this relation involves the unknown
vided with a cooling coil having provision for cooling water parameters k,, m, K , al, and These parameters are
inlet and outlet. A stainless steel 316 rupture disc tested estimated via regression analysis of present and previous
at 500 atm and 600 "C is provided on the top flange. The data.
flange also contained a thermowell for temperature mea- As per the model, AV and SV data alone are necessary
surement and four ports. One of the ports is connected for the purpose of parameter estimation. In this work, 9,
to a pressure gauge for pressure measurement. The others and Zm are also measured to provide additional checks on
are connected via precision needle valves to sampling tubes the methods employed.
of different lengths for sampling from different points in The experimental data are obtained by analyzing at
the autoclave. A magnetically driven stirrer is housed at room conditions the samples from the reactor. The oil
the center. phase in the reactor contains fatty acids, glycerides, gly-
The stirrer speed was measured with a tachometer. A cerol, and water. During analysis at room conditions, water
precalibrated chromel-alumel thermocouple connected to and glycerol are removed. Thus, measured mass fractions
a digital multimeter indicated the liquid temperature in- are larger than at the reactor conditions. Hence, a cor-
side the reactor to within f0.5 "C. Temperature was rection is required to relate measured quantities to reactor
controlled by using controllers which regulate the power conditions. The magnitude of this correction increases
input to the electrical heaters. with temperature and is about 3% at 260 "C. This rela-
tionship can be found from a fatty matter balance as
Analytical Section 2, = q/(l - x, - X J (24)
Volumetric methods were used for determination of
fatty acid, saponifiable matter, monoglyceride, and gly- If model parameters are known, then this equation could
cerol. Water from the oil sample was removed by perco- be used to relate composition of the oil phase in the reactor
lating through anhydrous sodium sulfate prior to analysis. to the measured quantities at room conditions. The mass
All analytical procedures were adopted from Mehlenbacher fractions 2, and 5, are found as
(1960). x', = (AV)M,/561000 (25)
Free fatty acid content of the sample on a dry basis is
2, = (SV - AV)Mt/168300 (26)
determined via the acid value defined as the milligrams
of potassium hydroxide necessary to neutralize the free Since
acids in 1 g of sample. The saponifiable matter in the Lx, + G y , = G9,
sample is determined via determination of the saponifi-
cation value defined as the milligrams of potassium hy- we have by using eq 7
droxide required to saponify 1g of dry sample. In present mG/L
work, 1-butanol is used as the solvent and an accuracy of
*0.5 unit is estimated. yg= + mG/L
730 Ind. Eng. Chem. Res., Vol. 27, No. 5, 1988

Table 11. Experimental Conditions of Present Work

- 0,A-Sturzonoggcr (1.01 (19511 -
temD 180-280 "C 260- - Model
water-to-oil ratio, G / L 0.0067-2.0 W / W
120-800
- -
oil Tollow
rPm 220- GIL- 0 7
glycerol in initial water charge, g, 0.0.171
oil coconut
density of oil 920 kg/m3
initial monoglyceride in oil 0.0
initial diglyceride in oil 0.0
inidial oil saponification value, SVo 253
initial oil-acid value, AVO 2.8-50

where m is at reaction temperature. The value of x g can

then be calculated from eq 7.
Procedure
The time required to reach equilibrium at different
temperatures was determined by trial. In equilibrium
experiments, known amounts of reactants were introduced
in the reactor and heated under stirring until equilibrium
was reached. The reaction mixture was quickly cooled to
room temperature by circulating chilled water through it.
Samples of oil and aqueous phases were then taken for
analysis.
In the kinetic experiments, reactants were heated and
stirred at constantly controlled temperature. Samples were
drawn intermittently for analysis. Aqueous-phaseglycerol
content is estimated from the end sample remaining in the
reactor after attaining equilibrium. In another set of runs,
reaction was halted after the desired time intervals; the
acid value of the fat phase and glycerol content in the
aqueous phase were measured. The rate experiments in
this work were confined to the temperature range 180-200
"C only since sufficient rate data at higher temperatures
were available in the literature. Experimental conditions
have been so chosen as to bring out the effect of reactant
ratio, feed composition, stirring speed on the equilibria,
and rate. The experimental conditions used in the present
work are summarized in Table 11.
Model Features and Parameter Estimates
The proposed model contains four equilibrium param-
eters and one rate parameter. Hence, extensive data at
equilibrium are required for its estimation. If these pa-
rameters were known, then one measurement, say acid
value, is necessary to estimate the composition of phases.
Thus, all the available acid value-time data in the litera-
ture could be used for the purpose of parameter estimation.
Butala (1984) estimated parameters of this model by
regression analysis of the acid value-time data of Stur-
zenegger and Sturm (1951). Arrhenius-type temperature
Time (mln)
Figure 1. Comparison of fit between model prediction and mea-
sured acid value for tallow fat.
dependences were assumed for the parameters. The re-
gression estimates of parameters for the different oils are
summarized in Table 111,while the Arrhenius temperature
coefficients are summarized in Table IV.
The value of m for beef tallow as shown in Table I1 is
in the neighborhood of the experimental value of Mills and
McClain (1949). Peanut oil shows a relatively high value
of m at all temperatures, the reasons for which could be
the higher level of unsaturation in the oil. The value of
m for coconut oil decreases from 63 at 190 "C to 4.5 at 280
"C, which indicates a exothermic heat of solution of 43.4
kJ/mol. The distribution coefficients for other oils also
show a similar trend.
The magnitude of the equilibrium constant K is esti-
mated to be 2.22 which is assumed temperature inde-
pendent in the model.
The concentration of water in the oil phase depends on
the concentration of triglyceride and fatty acid. The
contribution due to triglyceride alone can be estimated by
using a1 values and due to fatty acid alone from a2 values.
The contribution due to triglycerides during coconut oil
hydrolysis is estimated to vary between 0.0005 and 0.004
w/w and for peanut oil to vary between 0.002 and 0.01 w/w
over the temperature range 180-280 "C.The contribution
to water in the fat phase from fatty acid is seen to vary
between 0.025 and 0.06 w/w for coconut oil over the tem-
perature range 180-280 O C . The parameter & increases
with temperature, indicating an endothermic heat of so-
lution of 16.7 kJ/mol.
The agreement between model fitting and experimental
data of Sturzenegger and Sturm (1951) is displayed in

Table 111. Regression Estimates of Parameters of Model at Different Temperatures (Butala, 1984)
peanut oil a t tallow at coconut oil a t
225 "C 240 "C 260 O C 280 OC 225 "C 240 "C 260 "C 280°C 190°C 225 O C 240 "C 260 OC 280 "C
K 2.22 2.22 2.22 2.22 2.22 2.22 2.22 2.22 2.22 2.22 2.22 2.22 2.22
m 28 17 11.5 10 20 9 7 4.5 47 27 20 14 8
k," 0.040 0.053 0.075 0.093 0.036 0.055 0.092 0.133 0.016 0.040 0.046 0.064 0.10
62 0.45 0.55 0.60 0.70 0.45 0.55 0.65 0.75 0.326 0.450 0.50 0.55 0.70
6, 0.13 0.30 0.36 0.42 0.045 0.055 0.10 0.15 0.022 0.054 0.060 0.066 0.084
in m3/(kmol.min).

Table IV. Arrhenius Temperature Coefficients of Model Parameters (Butala, 1984)"

oil 62 m k,,kmol/(m3.min) K
coconut exp[4.17 - 2500/T] exp[-9.6 + 6470/T] exp[7.l - 5750/T] 2.22
peanut exp[3.83 - 2310/T] exp[-8.0 + 5680/T] exp[5.83 - 4505/T] 2.22
beef tallow exp[4.25 - 2500/T] exp[-10.25 + 6565/T] exp[10.34 - 6825/T] 2.22
T in degrees kelvin.
 Ind. Eng. Chem. Res., Vol. 27, No. 5, 1988 731

260

220
/003
0,A-Sturzeruggor el. a1 (1951)
- Model
o i l - Tallow
G/L- 0 7
300r-----7
2 60

20 40 80 o 120 160 200 i

Time (min) l i m e (min)

Figure 2. Comparison of fit between model prediction and mea- Figure 5. Comparison of fit between model prediction and mea-
sured acid value for tallow fat. sured acid value for coconut oil.

Model
t i
oil - Peanut
G / L - 0.5
220

o,2LkIY
0 10 20 30
Time (min)
40 50

Figure 6. Variation of predicted oil-phase mass fraction of di-

glyceride, monoglyceride, and glycerol during hydrolysis of tallow fat
at 280 "C, G f L = 0.7, and AVO= 5.5.

Time (min)
oil in Figure 5. In all these figures, the equilibrium acid
Figure 3. Comparison of fit between model prediction and mea- value can be seen to be clearly insensitive to temperature.
sured acid value for peanut oil. The simulated variation of the concentration of the in-
termediates is shown in Figure 6. It is observed that
diglyceride, monoglyceride, and glycerol exist in the fat
Model phase but their levels are indeed small. For example, for
oil -Peanut MIOW at 280 "C and G / L = 0.7, Xd = 0.07, X , = 0.04, and
GIL- 0.5 x g = 0.02 are indicated at equilibrium.
220

Model Prediction and Present Data

AV
Time (min)
Figure 4. Comparison of fit between model prediction and mea-
sured acid value for peanut oil.
Figure 1which shows the variation in acid value against
time at 280 and 240 " C at GIL = 0.7 for tallow. Initially
the reaction is relatively slow due to the limitation in water
availability in the oil phase. This is the induction period
which can be seen to disappear at higher temperatures.
The time course of reaction reveals a clear maxima, and
the rate then falls markedly due to depletion of glycerides
and consequent accumulation of glycerol in the fat phase
as equilibrium is approached. Thus, the model shows most
of the distinct features of the reaction. Figure 2 shows the
fit for tallow at 260 and 225 "C at G / L = 0.7; the fit for
peanut oil is shown in Figures 3 and 4 and that for cocount
Special experiments have been conducted to bring out
the strengths and weaknesses of the present model. Table
V lists the comparison between equilibrium data for co-
conut oil hydrolysis and predictions using parameters of
Table 111. We note that the parameters of Table I11 were
based on literature data, and thus the comparison provides
an independent check on the model proposed.
In these equilibrium experiments, the effects of tem-
perature, water-to-oil ratio, and initial composition of
charge have been investigated. In all these experiments,
uniformly excellent agreement is seen between model
prediction and data.
Table VI summarizes the experimental level of x , at
equilibrium for coconut oil hydrolysis. The independent
predictions from Table I11 model parameters are also
shown. The agreement is satisfactory. The level of
monoglyceride is seen to lie in the range 0.01-0.04 w f w.
This low level of intermediates can be taken as a justifi-
cation for the hypothesis that reactions I1 and I11 are rapid
relative to reaction I.
A comparison between the measured course and pre-
dicted course of hydrolysis at 190 "C is shown in Figure
7 . The comparison between model prediction and data
to show the effect of initial water-to-oil ratio on the course
732 Ind. Eng. Chem. Res., Vol. 27, No. 5, 1988

-i
3001---!
2 60
O.Ol4C

AV
E O.O"t s'
-
E
0.004
- Model
0 - Experimentol
oil -Coconut

o,oo2\ 1 - 19oc
G/L-O.9

0 4 8 12 16 20 0
u 2 4 6
Time ( h r )
8 1 0
Time ( h r :
Figure 7. Comparison between model prediction and experimental Figure 10. Comparison between model prediction and experimental
data for coconut oil hydrolysis with feed at AVO= 3 and y@ = 0. values of levels of monoglycerides at 190 OC, G / L = 0.9, AVO= 3, and
ygo = 0.
300

AV

- Model i
0 Experimentol
oil -Coconut
G/L - 0.9
T - 19oc 1
Time ( h r ) wo i 1 '
lime ( h r )
6 Q iloo
Figure 8. Effect of initial water-to-oil ratio on the time course of
coconut oil hydrolysis at 190 OC, AVO= 3, and jlgo = 0. Figure 11. Comparison between model prediction and experimental
values of aqueous-phase glycerol concentration from termination
3 00 runs with feed at AVO= 3 and jl@ = 0.0.

260 dictions using the parameters of Table 111.

220
180
AV
I40
Time ( h r s )
Figure 9. Effect of initial composition of charge on the time course
of coconut oil hydrolysis at 190 "C.
of hydrolysis is displayed in Figure 8. The comparison
with respect to initial composition of charge is shown in
Figure 9 and that for the intermediate x , is shown in
Figure 10. Uniformally excellent agreement is seen in all
these comparisons.
In the hydrolysis literature, no data on Qg versus time
are reported. Runs were therefore carried out so that the
reaction could be frozen rapidly after desired time intervals
by rapid chilling of the reaction mixture. These experi-
ments were tedious but were nevertheless done to assess
the goodness of the model proposed. Figure 11 which
displays 9,-t data reveals very good agreement with pre-
Verification of Assumption and Parameter
Estimates
It is shown so far that there is good agreement between
prediction and data. It is now necessary to check the
goodness of parameter estimates and the applicability of
the assumptions. Carefully selected experiments were
therefore necessary.
Hydrolysis experiments were conducted at 190 and 260
"C using suitable amounts of water so that water, oil, and
other species formed in the reaction remained essentially
in a single phase during the entire course of hydrolysis.
The proportion of water to be added to neutral oil at
different temperatures to form a homogeneous phase was
judged by extrapolating the results of Hilder (1968). Acid
value, glycerol, monoglyceride, and saponification value
were measured. The value of kl was estimated from the
initial slope of the acid value-time curve, and the equi-
librium constanta K1,K,, and K3 were calculated using x,,
x g , and x g data at equilibrium.
The purpose of conducting the above homogeneous
hydrolysis experiments was to have an independent check
on the magnitude of k,,K1,K 2 ,and K3 and also test the
assumption of the equality Kl = K2 = K3. Table VI1 shows
the value of the rate constant k , obtained thus at 190 OC
to be 0.015 m3/(kmol-min),which is close to the regression
estimate of 0.016 m3/(kmol.min) shown in Table 111.
Equilibrium constants K1,K 2 ,and K 3 so determined are
 Ind. Eng. Chem. Res., Vol. 27, No. 5, 1988 733

Table V. Comparison between Model Prediction and Equilibrium Data for Coconut Oil Hydrolysis

8 260 2.0 0.0650 0.0674 257.8 257.0

9 260 1.8 0.0711 0.0750 254.5 255.5
10 260 1.5 0.0843 0.0850 253.4 253.2
11 260 1.3 0.0962 0.1030 251.5 251.5
12 260 1.0 0.1220 0.1287 248.0 248.0
13 260 0.8 0.1433 0.1534 234.7 235.4
14 260 0.6 0.1770 0.1990 221.4 222.4
15 260 0.4 0.2460 0.2610 212.0 211.0
20 260 0.250 0.329 0.3330 184.5 185.5
25 240 1.8 0.0720 0.0750 256.0 256.0
26 240 1.5 0.0850 0.0887 252.8 253.0
27 240 0.96 0.1294 0.1363 249.2 249.3
28 240 0.8 0.1485 0.1547 239.4 238.0
29 240 0.6 0.1840 0.1960 226.4 226.6
30 240 0.4 0.2500 0.2610 210.5 210.9
31 240 0.25 0.3110 0.2820 170.6 170.3
33 220 2.0 0.0670 0.0670 256.5 256.8
34 220 1.5 0.0864 0.0890 255.0 255.4
35 220 1.0 0.1240 0.1310 246.1 246.7
36 220 0.8 0.1480 0.1560 237.4 238.1
37 220 0.6 0.1882 0.1976 229.0 229.8
38 220 0.4 0.2526 0.2580 209.9 210.9
39 220 0.25 0.3400 0.3158 183.8 184.9
41 200 0.9 0.1345 0.1290 241.8 240.9
42 190 2.0 0.0650 0.0656 249.8 250.9
43 190 0.9 0.1330 0.1430 238.3 241.0
45 190 0.825 0.2300 0.2120 232.6 230.2
46 190 0.764 0.3230 0.3330 216.8 222.0
47 190 0.4 0.2730 0.2620 219.1 214.8
48 190 0.0107 NA 0.0023° NA 22.0
49 190 0.0067 NA 0.0009' NA 43.0
50 180 1.0 0.1220 0.1220 249.2 250.3
'Homogeneous hydrolysis. NA = not available.

Table VI. Monoglyceride Levels for Hydrolysis of Coconut In the present work, attempts were made to measure the
Oil at Equilibrium solubility of water in coconut fatty acids alone at high
run T,"C GIL (xm)cxptl (xm),,red
temperatures. In this two-component system, fatty acids
8 260 2.0 0.010 0.024
were found to dissolve upto x, = 0.2 w/w at 240 "C, con-
12 260 1.0 0.017 0.042 firming the results of Mills and McClain (1949). However,
15 260 0.4 0.035 0.078 the batch hydrolysis model predicts only xw = 0.03-0.04
41 200 0.9 0.013 0.014 w/w in the fat phase at 240 "C. The reason for this dif-
42 190 2.0 0.004 0.005 ference in solubility is not clear at this stage. The above
43 190 0.9 0.012 0.011 results also compare favorably with a value of x , =
47 190 0.4 0.017 0.022
59 180 1.0 0.010 0.008
0.01-0.017 w/w at about the same temperature from
countercurrent column experiments of Jeffreys et al.
also seen to be nearly equal and in the vicinity of the (1961).
regression value shown in Table IV. The system has six components (C = 6): two liquid and
Table VI11 summarises the comparison between model one vapor phase (P= 3) and three independent reactions
prediction and present data for xg, yg, and m over the (R = 3). Thus, the phase rule gives (C - R ) 2 - P = 2+
temperature range 190-260 "C. Here equilibrium samples as the number of degrees of freedom. The net heat of
were drawn from both fat and aqueous phases. The results reaction is small so that temperature as one degree of
show good agreement between experimental m and re- freedom has negligible effect. Hence, the observable degree
gression values of Table 111. of freedom is one. This feature explains why the liquid-

Table VII. Estimate of Rate and Equilibrium Constants from Homogeneous Hydrolysis of Coconut Oil (k,in
m3/(kmol min))
T , "C GIL ki K1 K.7 KQ K.
190 0.0067 0.015 2.8 2.0 2.52 14.8
280 0.015 NA 2.48 2.22 2.05 11.3

Table VIII. Comparison between Model Prediction and Experimental Values of Glycerol Distribution Coefficient at
Equilibrium for Coconut Oil Hydrolysis at Different Temperatures
run T , "C (YJprd (YJexptl (xg)prd (Xg)exptl (m)pred (mlexptl
55 260 0.1310 NA 0.0094 NA 14 NA
54 240 0.1344 0.1297 0.0067 0.0063 20 20.6
53 220 0.1356 0.1317 0.0045 0.0043 30 30.6
52 200 0.1346 0.1330 0.0033 0.0028 40 47.5
51 190 0.1329 0.1338 0.0028 0.0021 47 63.0
ONA = not available.
734 Ind. Eng. Chem. Res., Vol. 27, No. 5, 1988
'O0I 1
63
0 2 4 6 8 IO
Time ( h r )
Figure 12. Effect of stirring speed on the time course of coconut
oil hydrolysis at 190 O C , G / L = 0.9, AVO= 3, and QgO = 0.
liquid equilibria in this reaction depend on water-to-oil
ratio alone.
The effect of external mass-transfer resistance on the
overall rate of reaction can be understood from runs 56,
57, and 58 in which stirrer speed was varied. The exper-
imental results are shown as acid value-time in Figure 12.
The plot based on the model which ignores external
mass-transfer resistance is also shown. The close agree-
ment indicates the role of external mass transfer to be
small a t 190 O C as per model assumption.
The conventional means of discerning the controlling
mechanism compares the kinetic parameter and mass-
transfer parameter. In the present study, KLa = 3.0 min-'
>> d,k,C,, = 0.09 min-l suggests that the overall process
belongs to a kinetically governed situation. In these cal-
culations, KL = 6 X lo3 m/min and a = 500 m2/m3from
Sarkar et al. (1980), and kl = 0.1 m3/(kmol.min) at 280 "C,
d, = 0.35, and Cto = 1.337 kmol/m3 are used.
A criterion, diffusion time tD,much smaller than reaction
time, t R , is given by Astarita (1967) for the applicability
of the very slow reaction hypothesis. Values of tR = 1.17
min and tD= 0.0167 min are estimated using available data
for D A and KL, indicating the overall process to be a case
of very slow reaction. The activation energy for the re-
action velocity constant k l for all the oils is generally in
excess of 40 kJ/mol as shown in Table IV which suggests
the overall process to be kinetically controlled.
The concentration of water and glycerol in the oil phase
is small. During the course of reaction, water is transferred
to the fat phase and glycerol produced in the fat phase are
extracted into aqueous phase. Due to these two transfer
processes in opposite directions, changes in density and
mass of both phases and estimated to be less than 2 % ,thus
the justification for the constancy of mass and density of
the two liquid phases during reaction.
In the present work and in the work of Sturzenegger and
Sturm (1951),oscillations of 2-3 units in acid value have
been observed around the equilibrium region. These os-
cillations were generally 1-1.5% of the measured value.
This effect was found to be more pronounced above 240
"C and G / L above 0.5. Such behavior is also seen in
industrial batch autoclaves. This feature though small in
magnitude is not explained by the present model. This
aspect is under investigation.
Summing up, we observe that the model describes the
course of hydrolysis very well. The assumptions used are
seen to be reasonable. In nonlinear regression, parametric
insensitivities are likely. In the present work, these
problems have been overcome by checking the regression
estimates with experimental values.
Conclusions
A model which contains four equilibrium parameters
and one rate parameter is developed to describe the
thermal hydrolysis of vegetable oils. The model is shown
to exhibit most of the kinetic and thermodynamic features
and also to give an excellent description of the present and
previous hydrolysis data.
The liquid-liquid thermal hydrolysis of oils could thus
be viewed as a three-step reaction taking place in the oil
phase with negligible change in the density of phases. The
triglyceride hydrolysis is the controlling step, and the
overall reaction belongs to very slow reaction regime. Mass
transfer of glycerol and water across the phases is fast
relative to reaction.
The concentration of water in the oil phase is only 2-4%
w/w even though the oil phase may contain predominantly
fatty acids and in which water solubilities can be expected
to be high. The reasons for the marked reduction in water
"solubility" in the oil phase are not clear at this stage.
Nomenclature
A = l / ( m G / L + 1)
A' = (xpo + y p o G / L o ) M t / x t f l g (+
l mG/L)
AV = acid value, mg of KOH/g of fat
B = K382/(A + K362)
B' = (-4'- K ~ & C , O / C ~ O ) + / ( AK362)
C = K282/(1 + K262 - B )
C' = [B' + (Cmo/Cto)- Kz&Cdo/Cto]/(1+ K282 - B )
C, = concentration of species i, kmol/m3
D=l+C+BC
D'= (Cao/Cm)- (C'+ BC'+ B ' )
DA = diffusion coefficient of the species in the liquid system,
mz/s
E=l-C
G = mass of aqueous phase, kg
k , = equilibrium constant for the ith reaction step, i = 1-3
K , = overall reaction equilibrium constant
KL = overall mass-transfer coefficient, m/min
k , = specific reaction rate constant in the fat phase, m3/
(kmol-min)
L = mass of fat phase, kg
A4,= molecular weight of species i
m = equilibrium distributioncoefficient for glycerol on weight
basis
N = (a- p ) / P
n = 1, 3, or 5 for forward reactions and 2 , 4 , or 6 for reverse
reactions
P = l/[k18,C,(D - 61/82 + D E / ( k & ) ) ]
Q = (D - 2(61/62) - c,o/Cto - C,c,E/(Ct&z&))/(D - 6 1 / 6 2 +
DEI (K262))
R = (6,/6z + Ca,/C,)/(D - 61/62 + D E / ( K z ~ z ) )
SV = saponification value, mg of KOH/g of fat
T = temperature, " C
t = time, min
tR = 1/(4klCto)= reaction time, min
tD = 4D/aKL2= diffusion time, min
x , = mass fraction of species i in the fat phase
y , = mass fraction of species i in the aqueous phase
Greek Symbols
(Y = [ Q + ( Q 2 + 4R)lI2]/2, as defined in eq 22
/3 = [ Q - ( Q 2 + 4R)'I2]/2, as defined in eq 22
6, = as defined in eq 8
d, = fat phase holdup per unit reactor volume
6, = as defined in eq 8
pf = density of the fat phase, kg/m3
Subcripts and Superscripts
a = fatty acid
d = diglyceride
e = equilibrium state
 Ind. Eng. C h e m . R e s . 1988,27, 735-739
f = fat phase
g = glycerol
m = monoglyceride
t = triglyceride
w = water
-
0 = initial condition
= measured value at room conditions
Literature Cited
Astarita, G. Mass Transfer with Chemical Reaction; Elsevier: Am-
sterdam, 1967.
Basu, P. K. Chem. Age India 1976,27(10), 871.
Burrows, K. Trans. Znst. Chem. Eng. 1953,31(10), 250.
Butala, D. N. M.Tech. Dissertation, Department of Chemical En-
gineering, Indian Institute of Technology, Bombay, India, 1984.
Desai, S. M.; Raghunathan, T. S.; Shankar, H. S. Frontiers in
Chemical Reaction Engineering;Wiley Eastern: New York, 1984;
Vol. 1, p 253.
Thermal Hydrolysis of Vegetable Oils and Fats. 2. Hydrolysis in
Continuous Stirred Tank Reactor
T. A. Patil, T. S. Raghunathan, and H. S. Shankar*
Department of Chemical Engineering, Indian Institute of Technology, Bombay 400 076, India
The liquid-liquid thermal hydrolysis of coconut oil is studied experimentally on a laboratory-scale
continuous stirred tank reador over the range 225 O C , 3000 kPa to 260 O C , 5500 kPa. Good agreement
is indicated between the model prediction and data.
Hydrolysis as a liquid-liquid reaction has been practiced
commercially for a long time. Batch autoclaves are com-
monly used in small-scale operations, while continuous
countercurrent columns are employed in large-scale op-
erations. Batch operation involves high specific energy
consumption and idle time. The continuous countercur-
rent spray towers require very high initial investment. In
India both batch and continuous countercurrent operations
are in use.
The hydrolysis can be brought about over a batch of oil
by a spray of high-pressure water. This scheme could have
somewhat poor productivity in relation to continuous
countercurrent operations. No work on semicontinuous
hydrolysis has been reported so far.
The reaction can also be conducted in a continuous
stirred tank reactor. This scheme offers advantages with
respect to energy integration and productivity in addition
to cost advantages in terms of investment. In this scheme,
a mixture of water and oil is pumped to a reactor. In this
paper, we examine the behavior of this configuration
theoretically by using the model developed earlier (Patil
et al., 1988). The model predictions are then tested ex-
perimentally.
Experimental Apparatus
A laboratory autoclave of 1-L capacity made from 316
stainless steel mounted in a compact, electrically heated
furnace was used. The specifications of the autoclave are
given elsewhere (Patil et al., 1988). The top flange houses
a cooling coil with provision for cooling the water inlet and
outlet. A 316 stainless steel rupture disc tested at 50 000
kPa and 600 "C is also provided. The flange contains a
thermowell for temperature measurement and four ports.
One of the ports is connected to a pressure gauge for
pressure measurement. The others are connected via
0888-5885/ 8812627-0135$01.50/0
735
Donders, A. J. M.; Wijffels, J. B.; Reitema, K. Proceedings of the
Fourth European Symposium on Chemical Reaction Engineering,
Brussels, Sept 9-11, 1968.
Hilder, M. H. J. Am. Oil Chem. SOC.1968,45 703.
Jeffreys, G. V.; Jenson, V. G.; Miles, P. R. Trans. Znst. Chem. Eng.
1961, 39, 389.
Lascaray, L. Znd. Eng. Chem. 1949, 47, 486.
Lascaray, L. J. Am. Oil Chem. SOC.1952,29, 362.
Mehlenbacher, V. L. Analysis of Fats and Oils; Gerrad: Chapaign,
IL, 1960.
Mills, V.; McClain, H. K. Znd. Eng. Chem. 1949, 47, 1982.
Mueller, H. H.; Holt, E. K. J. Am. Oil Chem. SOC.1948, 25, 305.
Perry, R. H.; Chilton, C.H. Chemical Engineers Handbook, 4th ed.;
McGraw-Hill: Kogakusha, Tokyo, 1973; p 3.87.
Sarkar, S.; Mumford, C. J.; Phillips, C. R. Znd. Eng. Chem. Process
Des. Dev. 1980, 19, 672.
Sturzenegger, A.; Sturm, H. Znd. Eng. Chem. 1951, 43(2), 510.
Received for review November 19, 1985
Revised manuscript received August 14, 1987
Accepted November 30, 1987
precision needle valves to sampling tubes of variable length
for sampling from inside the vessel. A magnetically driven
stirrer is housed at the center.
Schematic representation of the arrangement is shown
in Figure 1. A mixture of a predetermined quantity of
coconut oil and water is prepared in a storage tank and
fed to reactor via a dosing pump supplied by Jagdish
Engineering Works, Bombay, capable of operating up to
20000 kPa and having adjustable stroke length with a
maximum capacity of 2 L/h. The high-pressure dosing
pump and reactor are connected by 316 stainless steel tube
on which a check valve is mounted. This is provided in
addition to a built-in nonreturn valve on the pump. The
reactor outlet is connected to a cooler where cooling water
circulates followed by a 316 stainless steel back-pressure
regulating valve to maintain pressure inside the reactor
during the experiment.
Experimental Procedure
Water and oil are mixed in the desired proportion and
stirred well to obtain a uniform mixture, and this was
continuously pumped to the reactor by the high-pressure
dosing pump. It was found that no segregation of wateroil
emulsion occurred in the feed line, and also it was verified
that the back flow of the reactants was completely pre-
vented by providing the additional nonreturn valve on the
feed line. Runs were conducted at chosen residence times
until steady state was reached, which was determined by
trial to vary between 1.5 and 2 h. Cooled product emerging
out of the reactor was collected, maintaining constant
pressure inside the vessel by prior setting of the back-
pressure regulating valve.
The mixing pattern inside reactor was examined; for this
purpose, a step input of a tracer solution containing freshly
prepared sodium hydroxide was used.
0 1988 American Chemical Society