Fluid Phase Equilibria, 81 (1992) 307-320 307

Elsevier Science Publishers B.V., Amsterdam

Phase equilibrium study for the separation and

fractionation of fatty oil components using supercritical
carbon dioxide

Ragunath Bharath, Hiroshi Inomata, Tadafumi Adschiri and Kunio Arai

Department of Biochemistry and Chemical Engineering, Tohoku University, Sendai 980
(Japan)
(Received July 30, 1991; accepted in final form May 18, 1992)

ABSTRACT

Bharath, R., Inomata, H., Adschiri, T. and Arai, K., 1992. Phase equilibrium study for the
separation and fractionation of fatty oil components using supercritical carbon dioxide.
Fluid Phase Equilibria, 81: 307-320.

Vapor-liquid equilibria were measured for the binary systems oleic acid-carbon dioxide
and triolein-carbon dioxide and for the ternary system oleic acid-triolein-carbon dioxide
with the aim of separating the fatty acids and triglycerides. Experimental results showed that
oleic acid can be selectively extracted using supercritical carbon dioxide as a solvent.
Supercritical carbon dioxide was also considered for the fractionation of mixtures of
triglycerides. High pressure vapor-liquid equilibria were measured for the systems palm
kernel oil-carbon dioxide and sesame oil-carbon dioxide. Supercritical carbon dioxide was
found to fractionate these triglyceride mixtures on the basis of the total carbon number of
the constituent fatty acids of the triglyceride.

INTRODUCTION

Fatty acids and triglycerides are important raw materials in many pro-
cesses in the food, pharmaceutical and detergent industries. The pharmaceu-
tical industry, for example, demands a high degree of purity of some
polyunsaturated fatty acids, and the food industry requires concentration of
the middle-range triglycerides to obtain a product with the desired melting
character. Thus, separation and fractionation processes are an essential part
of the oil industry. At present, two conventional processes predominate:
vacuum distillation and solvent extraction. However, the isolation of sub-
stances from natural products remains problematic, especially with ther-
mally labile natural materials such as polyunsaturated fatty acids and their

Correspondence to: K. Arai, Department of Biochemistry and Chemical Engineering,

Tohoku University, Sendai 980, Japan,

0378-3812/92/$05.00 0 1992 - Elsevier Science Publishers B.V. All rights reserved

308 R. Bharath et al. 1 Fluid Phase Equilibria 81 (1992) 307-320

glycerides. Vacuum distillation requires relatively high temperatures; this

results in decomposition or degradation of these substances. Some of the
disadvantages of solvent extraction are the tendency for the solvent to
remain in the product after extraction and the continually shrinking list of
safe solvents.
For these reasons, much effort has centered on the development of
separation and fractionation methods that are clean, efficient and free of
thermal degradation of substances. One promising state-of-the-art process is
supercritical fluid extraction. Some excellent monographs (McHugh and
Krukonis, 1986; Stahl et al., 1987) on this subject have appeared in recent
years. Despite all the attention given to the application of supercritical
carbon dioxide in the fatty oil industry, fundamental data on phase equi-
libria involving fatty oils and carbon dioxide are still lacking.
Most of the work related to separation of fatty acids and triglycerides
deals only with the solubility of the triglyceride or of the fatty acid, that is,
the vapor-phase composition only. Chrastil (1982) measured the solubility
of certain fatty acids and triglycerides in supercritical carbon dioxide within
the pressure range 8-25 MPa and the temperature range 40-80C. Bam-
berger et al. (1988) measured the solubility of triglycerides and fatty acids in
carbon dioxide. They reported that the solubilities of trilaurin and .palmitic
acid are about the same, although their molecular weights differ by a factor
of about 2.5. They attributed this to the higher polarity of free fatty acids
compared with triglycerides. Brunnetti et al. ( 1989) also measured the
solubility of fatty acids and triglycerides in supercritical carbon dioxide at
20 and 30 MPa and in the temperature range 40-60C.
However, at high pressures, a large amount of carbon dioxide dissolves in
the liquid phase causing a change in its thermodynamic properties. Informa-
tion concerning the liquid phase is essential for designing the supercritical
fluid separation or fractionation processes. In the present work, the two
aims of the vapor-liquid equilibria measurements were firstly to separate
the fatty acid and triglyceride and secondly to fractionate mixtures of
triglycerides. For the separation, high pressure vapor-liquid equilibria were
measured for the binary systems oleic acid-carbon dioxide and triolein-
carbon dioxide and for the single ternary system oleic acid-triolein-carbon
dioxide. High pressure vapor-liquid equilibria were also measured for the
systems palm kernel oil-carbon dioxide and sesame oil-carbon dioxide;
this is related to the second aim.

EXPERIMENTAL

Apparatus and procedure

High pressure vapor-liquid equilibrium data were obtained using a

static-type apparatus with a recirculating system, as shown in Fig. 1. Details
R. Bharath et al. 1 Fluid Phase Equilibria 81 (1992) 307-320 309

Equilibrium Cell
Air Bath
Magnetic Stirrer
Magnetic Pump
Piston Cylinder
Motor
Pressure Controller
COzBomb
Back Pressure Regulator
Diaphragm Pump
Trap
Sampling Cylinder
MV: Melering Valve

Fig. 1. Experimental apparatus.

of this system have been described elsewhere (Inomata et al., 1988). This
equipment consists of an equilibrium cell with a tempered glass window, a
recycling system for the vapor phase, a feed system and a sampling system.
A sample of solute was fed into the evacuated equilibrium cell and
degassed. Then, carbon dioxide which was pressurized with a diaphragm
pump (Orlita MHS-8/30) was also fed into the equilibrium cell. The system
was agitated with a magnetic pump and a magnetic stirrer for about 6 h for
the binary systems and 12 h in the case of multicomponent systems to attain
equilibrium.
After the system had reached equilibrium, the agitation and recycling was
stopped. The pressure sensor of the variable-volume PID-controlled piston
cylinder was finely adjusted to the equilibrium pressure. The equilibrated
samples of vapor and liquid phases were removed from the cell by decom-
pression through a metering valve, keeping the system pressure constant.
This was achieved by automatically advancing the piston into the cylinder at

TABLE 1
Phase equilibrium systems and conditions

System Temp. (K) Pressure (MPa) Feed

(wt.% triolein)

Binary system
CO1 -oleic acid 313,333,353 10.0-30.0
CO1 -triolein 313,333 15.0-31.0
CO,-palm kernel oil 313,323,333,353 6.0-32.3
COz -sesame oil 313,333,353 10.0-33.0

Ternary system
CO2 - oleic acid- triolein 313,333 20,25, 30 20,50,70
310 R. Bharath et al. /Fluid Phase Equilibria 81 (1992) 307-320

the same rate at which the sample was withdrawn through the metering
valve. The pressure change in the cell was 0.1 MPa during sampling of the
liquid phase and 0.2 MPa during sampling of the vapor phase.
Owing to decompression, the solute precipitated and was retained in a
tared trap while the carbon dioxide was collected as dry ice in a tared cylinder
which was immersed in a liquid nitrogen bath. To remove some of the solute
still trapped in the metering valve, about 200 cm3 of hexane was pumped
through the metering valve and the solute-solvent mixture was collected in
the same trap. This was carried out a couple of times until all the solute was
recovered. Then the hexane was completely removed from the solute using
an evaporator. The vapor and liquid phase compositions were determined
gravimetrically for the binary systems. For the ternary systems, the solute
mixture compositions were determined by gas chromatography.
The error in measurement occurs mainly during the sampling procedure.
The maximum overall error in determining the equilibrium composition was
estimated to be 5%. The equilibrium temperature was measured with a
platinum resistance thermometer accurate to 0.5 K. The system pressure was
measured with a Bourdon-type gauge (Nagano Keiki Ltd.) whose accuracy
was checked with a dead weight tester to be 0.02 MPa.

Systems and conditions

Table 1 shows all the phase equilibrium systems measured and their range
of measurements.

Materials

The liquefied carbon dioxide used was of relatively high purity (99.9%)
from Nippon Sanso Products. Oleic acid of purity more than 90% was
supplied by KAO Oil and Fat Industries Ltd. and triolein of 70% purity was
obtained from Wako Chemicals. Gas chromatographic analysis showed that

TABLE 2
Triglyceride composition (wt.%) of palm kernel oil and sesame oil

TG carbon no. 30 32 34 36 38 40 42 44 46 48 50 52 54

Palm kernel oil 7.0 9.5 15.4 29.6 17.8 6.6 8.9 5.2
Sesame oil In this range

TABLE 3
Fatty acid distribution (wt.%) of palm kernel oil

Fatty acid ClO-0 c12-0 c14-0 C16-0 C18-0 C18-1 C18-2

Distribution 2.3 55.7 16.2 7.6 1.4 14.9 1.9
R. Bharath et al. / Fluid Phase Equilibria 81 (1992) 307-320 311

TABLE 4
Fatty acid distribution (wt.%) of sesame oil

Fatty acid C16-0 Cl80 Cl81 C18-2 Cl83

Distribution 9.0 4.1 40.4 44.9 0.8

the impurities in the triolein sample were mainly oleic acid, monolein and
diolein. Triolein was purified by supercritical carbon dioxide extraction
prior to the experiment in order to remove impurities that are soluble in
supercritical carbon dioxide.
Table 2 shows the triglyceride composition of the two vegetable oils used
in the experiment. Palm kernel oil is a mixture of triglycerides, with the total
number of carbon atoms in the three constituent fatty acids varying from 30
to 44; in sesame oil the variation is from 48 to 54. However, we were not
able to determine the exact triglyceride composition of sesame oil owing to
degradation of the thermally labile triglycerides during the gas chromato-
graphic analysis.
Tables 3 and 4 show the fatty acid distribution in the oils used in this work.

TABLE 5
Vapor-liquid equilibrium data for the CO,-oleic acid system

P (MPa) T (K) Weight fraction of CO*

Liquid Vapor Difference (%) a

(this work) (this work)
Brunetti Chrastil

10.17 313.15 0.3820 -

15.39 313.15 0.4238 0.9915 98.4%
20.02 313.15 0.4352 0.9852 63% 112.0%
24.99 313.15 0.4335 0.9782
30.00 313.15 0.4336 0.9670 58%

10.73 333.15 0.2309 -

15.15 333.15 0.2886 0.9927 33.0%
20.40 333.15 0.3353 0.9896 60%
24.92 333.15 0.3664 0.9818 79.8%
30.02 333.15 0.4002 0.9716 61%

11.06 353.15 0.1851 -

16.34 353.15 0.2634 0.9937
20.95 353.15 0.3155 0.9932
25.70 353.15 0.3537 0.9855
29.34 353.15 0.3826 0.9744

a Difference (%) = ( Ythisworlr- Yrir)/ Y,, x 100, where Y = g solute/g carbon dioxide.
312 R. Bharath et al. 1 Fluid Phase Equilibria 81 (1992) 307-320

TABLE 6
Vapor-liquid equilibrium data for the CO,-triolein system

P (MPa) T(R) Weight fraction of CO*

Liquid Vapor Difference (%) a

(this work) (this work)
Brunetti Chrastil

15.34 313.15 0.3063 0.9964 24.6%

20.10 313.15 0.3227 0.9936 34.0% 69.3%
24.82 313.15 0.3519 0.9910 82.6%
29.08 313.15 0.3453 0.9901 55.0%

15.40 333.15 0.2679 0.9963

22.50 333.15 0.3146 0.9959 2.5%
24.99 333.15 0.3398 0.9945
31 .oo 333.15 0.3569 0.9870 27.7%

a Difference (%) = ( Ythiswork- Y,,,)/Y,,, x 100, where Y = g solute/g carbon dioxide.

RESULTS AND DISCUSSION

Separation of fatty acids and triglycerides

High pressure vapor-liquid equilibria for the binary systems oleic acid-car-
bon dioxide and triolein-carbon dioxide were measured and the results are
presented in Tables 5 and 6 respectively. These tables also include the difference
between the solubility reported by Chrastil(l982) and Brunetti et al. (1989)
and the results obtained in this work for these two systems. The values reported
by these authors are much higher than those obtained in this work. We think
that the higher values are due to the solubilization of low molecular weight
components present in the reagents. Although the triolein we obtained was

10 000

0 0.5 1.0 095 1.0

Wt. fraction CO2

Fig. 2. Vapor-liquid equilibrium for the CO,-oleic acid system.

R. Bharath et al. 1 Fluid Phase Equilibria 81 (1992) 307-320 313

0 a5 1.0 0.95 10
Wt. fraction CO2
Fig. 3. Vapor-liquid equilibrium for the CO,-triolein system.

TABLE 7a
Vapor-liquid equilibrium data for the CO,-oleic acid-Triolein system (T = 313.15 K)

P (MPa) Component Weight fraction CO*

Liquid Vapor

20.0 co2 0.341 0.979

Oleic acid 0.609 0.021
Triolein 0.050 _
25.0 CO* 0.364 0.981
Oleic acid 0.581 0.019
Triolein 0.055 -
30.0 CO, 0.379 0.971
Oleic acid 0.574 0.029
Triolein 0.046 _

20.0 co* 0.329 0.989

Oleic acid 0.333 0.110
Triolein 0.337 -
25.0 CO, 0.354 0.984
Oleic acid 0.380 0.012
Triolein 0.265 0.004
30.0 CO, 0.370 0.980
Oleic acid 0.394 0.016
Triolein 0.235 0.004

20.0 co* 0.339 0.971

Oleic acid 0.199 0.029
Triolein 0.462 0.004
25.0 CO, 0.355 0.986
Oleic acid 0.158 0.010
Triolein 0.487 0.004
30.0 CO, 0.375 0.982
Oleic acid 0.191 0.010
Triolein 0.433 0.008
314 R. Bharath et al. 1 Fluid Phase Equilibria 81 (1992) 307-320

TABLE 7b
Vapor-liquid equilibrium data for the CO,-oleic acid-triolein system (T = 333.15 K)

P (MPa) Component Weight fraction CO,

Liquid Vapor

20.0 CO, 0.327 0.989

Oleic acid 0.647 0.011
Triolein 0.025 -
25.0 COZ 0.357 0.973
Oleic acid 0.594 0.027
Triolein 0.049 _
30.0 CO, 0.383 0.975
Oleic acid 0.595 0.025
Triolein 0.022 _

20.0 COZ 0.310 0.985

Oleic acid 0.335 0.014
Triolein 0.355 -
25.0 CO* 0.348 0.987
Oleic acid 0.431 0.009
Triolein 0.220 0.003
30.0 CO, 0.369 0.978
Oleic acid 0.314 0.020
Triolein 0.317 0.001

20.0 CO* 0.311 0.994

Oleic acid 0.150 0.005
Triolein 0.539 0.001
25.0 CO* 0.342 0.988
Oleic acid 0.196 0.010
Triolein 0.462 0.002
30.0 CO* 0.375 0.982
Oleic acid 0.191 0.010
Triolein 0.433 0.008

of only 70% purity, supercritical carbon dioxide was used to remove the low
molecular weight components such as oleic acid prior to the experiment. The
purity of the oleic acid used in this work was more than 90%, compared with
the 68% purity of the oleic acid of Brunetti et al. (1989). Chrastil (1982),
however, did not report the purity of the reagents used in his measurements.
Figures 2 and 3 show the vapor-liquid equilibria for the oleic acid-car-
bon dioxide and triolein-carbon dioxide systems respectively. For all the
isotherms, as the pressure increases, the solubility of the solute in the vapor
phase and that of carbon dioxide in the liquid phase increases. The
two-phase region expands as the temperature increases in the range of
measurement. When we compare Figs. 2 and 3 at a constant pressure and
R. Bharath et al. / Fluid Phase Equilibria 81 (1992) 307-320 315

temperature, it can be seen that the solubility of triolein is lower that that
of oleic acid and that the carbon dioxide dissolves less in the liquid phase in
the case of the carbon dioxide-triolein system.
In order to determine the selectivity of supercritical carbon dioxide,
vapor-liquid equilibria for the ternary system oleic acid (FA) triolein
(TG) -carbon dioxide (CO*) were measured at two different temperatures
and the results are shown in Tables 7a and 7b. The ternary diagrams are
shown in Figs. 4(a) and 4(b). Three different feed compositions were used.
The solid circles in the figures are the binary system results. The distribution
coefficients of each component (K1 value = JJ~/x,) were calculated and are
shown in Fig. 5.
The distribution coefficient of both oleic acid and triolein varies not only
with pressure but also with the feed composition. As pressure increases, the

P=20MPa
T=40C

LFA TG

FA TG

FA TG

Fig. 4(a)
316 R. Bharath et al. / Fluid Phase Equilibria 81 (i99.?) 307-320

FA TG

f \
TG
c:
Fig. 4. Vapor-liquid equilibrium for the CO*-oleic acid-triolein system: (a) T = 313 K;
(b) T = 333 K.

K value of both oleic acid and triolein increases, but the K value of oleic
acid is ten times larger than that of triolein; this indicates the selective
solubil~ation of oleic acid by supercritical carbon dioxide even at high
pressures. Triolein was detected in the vapor phase only for the 70 wt.%
triolein feed.

Fractionation of triglyceride mixtures

High pressure vapor-liquid equilibria for the systems palm kernel oil-
carbon dioxide and sesame oil-carbon dioxide were measured and the
results are presented in Tables 8 and 9 and in Figs. 6 and 7 respectively.
The vapor- and liquid-phase triglyceride compositions for the palm kernel
oil-carbon dioxide system were determined by gas chromatographic analysis
R. Bharath et al. / Fluid Phase Equilibria 81 (1992) 307-320 317

Fig. 5. Distribution coefficient of oleic acid and triolein.

30

Q
%
x 20
Is
t
g!
a. 10

0 a5 1.0 oe 1.0
wtfrwtion co2
Fig. 6. Vapor-liquid equilibrium for the CO,-palm kernel oil system.

at 333 K. Owing to difficulties in peak analysis, and hence to a decrease in

accuracy, the vapor- and liquid-phase triglyceride components were divided
into three groups based on the total number of carbon atoms in the three
fatty acid radicals, and the distribution coefficient (K value) for each group
was calculated. We assume that sesame oil is a single group owing to its
narrow triglyceride distribution.
The K value plots are shown in Fig. 8. As the pressure increases, the
difference in li; values between the groups becomes smaller. Sesame oil is a
318 R. Bharath et al. 1 Fluid Phase Equilibria 81 (1992) 307-320

TABLE 8
Vapor-liquid equilibrium data for the CO*--palm kernel oil system

P (MPa) T(K) Weight fraction of CO*

Liquid Vapor

10.38 313.15 0.3117

t 5.08 313.15 0.3478 0.9850
20.40 313.15 0.3790 0.9824
24.64 313.15 0.4070 0.9747
30.12 313.15 0.4304 0.9621
32.33 313.15 0.4423 0.9563

15.20 323.15 0.3276 0.9865

20.10 323.15 0.3647 0.9884
24.82 323.15 0.4671 0.9770
30.04 323.15 0.4355 0.9626
34.55 323.15 0.4666 0.9498

6.28 333.15 0.1347

15.71 333.15 0.3099 0.9948
19.52 333.15 0.3462 0.9905
26.74 333.15 0.405 1 0.9791
30.30 333.15 0.4291 0.9658
31.87 333.15 0.4428 0.9625

10.28 353.15 0.2377 -

15.08 353.15 0.2625
20.49 353.15 0.3197 0.9993
24.19 353.15 0.3534 0.9909
30.05 353.15 0.4085 0.9761
31.08 353.15 0.4203 0.9690

r
0 4oc
0 60
0 80 7)

:
0 0.5 1,o 0.95 FO
Wt.fmcUonCO2

Fig. 7. Vapor-liquid equilibrium for the CO*--sesame oil system.

R. Bharath et al. 1 Fluid Phase Equilibria 81 (1992) 307-320 319

TABLE 9
Vapor-liquid equilibrium data for the CO,-sesame oil system

P (MPa) T (K) Weight, fraction of CO,

Liquid Vapor

10.02 313.15 0.2400

15.05 313.15 0.2786
20.58 313.15 0.3021 0.9937
23.31 313.15 0.3173 0.9831
26.05 313.15 0.3262 0.9930
29.56 313.15 0.3335 0.9888
33.05 313.15 0.3519 0.9885

9.80 333.15 0.1547

15.74 333.15 0.2636
21.15 333.15 0.2990 0.9875
24.37 333.15 0.3169 0.9941
26.87 333.15 0.3204 0.9904
28.97 333.15 0.3406 0.9914
31.90 333.15 0.3562 0.9885

9.83 353.15 0.1547

15.25 353.15 0.2249 -
21.04 353.15 0.2720 0.9878
25.66 353.15 0.3042 0.9945
30.00 353.15 0.3329 0.9944

I T=60C I

,63315 20 25 30

Pressure(MPa)
Fig. 8. Distribution coefficient for triglyceride groups.
320 R. Bharath et al. /Fluid Phase Equilibria 81 (1992) 307-320

mixture of triglycerides of carbon number more than 48, and thus its K
value is the lowest. Supercritical carbon dioxide tends to dissolve the
triglycerides selectively on the basis of their carbon numbers. The short-
chain triglycerides have a higher K value at lower pressures. This result
shows that triglyceride mixtures can be fractionated by using supercritical
carbon dioxide at a temperature of 60C and a pressure of around 20 MPa.

CONCLUSIONS

The vapor-liquid equilibria were measured with two aims in mind. The
first was the separation of fatty acids and triglycerides. In this case, the
binary vapor-liquid equilibria were measured for the systems oleic acid-
carbon dioxide and triolein-carbon dioxide. In order to determine the
selectivity of supercritical carbon dioxide, one ternary system, oleic acid- tri-
olein-carbon dioxide, was also measured. It was found that oleic acid is
selectively soluble in supercritical carbon dioxide.
The second aim was to fractionate mixtures of triglyceride. To do this,
vapor-liquid equilibria were measured for two vegetable oils: palm kernel
oil and sesame oil. The short-chain triglyceride group showed a higher K
value at lower pressures, while as the pressure increased, the difference in K
values between the groups became smaller. This implies that supercritical
carbon dioxide solvent is a viable medium for the fractionation of mixtures
of triglycerides.

ACKNOWLEDGMENT

Financial support from Toray foundation and the Japanese National

Food Research Institute are gratefully acknowledged.

REFERENCES

Bamberger, T., Erickson, J. C., Cooney, C. L. and Kumar S. K., 1988. Measurement and
model prediction of solubilities of pure fatty acids, pure triglycerides and mixtures of
triglycerides in supercritical carbon dioxide. J. Chem. Eng. Data, 33: 327-333.
Brunetti, L., Daghetta, A., Fedeli, E., Kikic, I. and Zanderighi, L., 1989. Deacidification of
olive oils by supercritical carbon dioxide. J. Am. Oil Chem. Sot., 66: 209-217.
Chrastil, J., 1982. Solubility of solids and liquids in supercritical gases. J. Phys. Chem., 86:
3016-3021.
Inomata, H., Kondo, T., Hirohama, S., Arai, K., Suzuki, Y. and Konno, M., 1989. Vapour-
liquid equilibria for binary mixtures of carbon dioxide and fatty acid methyl esters. Fluid
Phase Equilibria, 46: 41-52.
McHugh, M., Krukonis, V. J., 1986. Supercritical Fluid Extraction: Principle and Practice.
Buttetworths, Boston, MA.
Stahl, E., Quirin, K. W. and Gerard, D., 1987. Dense Gases for Extraction and Refining.
Springer Verlag, Berlin.