Separation and Purification Technology 160 (2016) 106111

Contents lists available at ScienceDirect

Separation and Purification Technology

journal homepage: www.elsevier.com/locate/seppur

Deacidification of palm oil by solvent extraction

Cintia B. Gonalves a,, Christianne E.C. Rodrigues a, Elaine C. Marcon b, Antonio J.A. Meirelles b,
a
LES, Separation Engineering Laboratory, Department of Food Engineering (ZEA-FZEA), University of So Paulo (USP), P.O. Box 23, Zip Code 13635-900, Pirassununga, So Paulo, Brazil
b
EXTRAE, Department of Food Engineering (DEA-FEA), University of Campinas (UNICAMP), P.O. Box 6121, Zip Code 13083-862, Campinas, So Paulo, Brazil

a r t i c l e i n f o a b s t r a c t

Article history: The present work reports the influence of some process variables on the losses/transfer of fatty com-
Received 28 September 2015 pounds during the deacidification of palm oil by liquidliquid extraction. The response surface method-
Received in revised form 12 January 2016 ology (RSM) was used to analyze the effect of the mass ratio of oil to solvent and the water content in the
Accepted 13 January 2016
solvent, aiming to minimize the losses of neutral oil and maximize the transfer of free fatty acids plus
Available online 14 January 2016
carotenoids preservation. By using appropriate conditions observed in RSM analysis (mass ratio of oil
to solvent equal to 0.74 and water content around 6 mass%), the deacidification of palm oil by continuous
Keywords:
liquidliquid extraction was performed in a perforated rotating disc contactor (PRDC). The experimental
Palm oil
Liquidliquid extraction
results indicate that it is possible to obtain refined palm oil with a free acidity lower than 0.1 mass% by
Deacidification continuous liquidliquid extraction.
Carotenoids 2016 Elsevier B.V. All rights reserved.
PRDC

1. Introduction
The rapid expansion in world production of palm oil over the
last decades has attracted the attention of the oils and fat industry.
Crude palm oil is extracted from the fresh mesocarp of the palm
fruit and contains a small amount of undesirable components
and impurities, such as mesocarp fibers, free fatty acids (FFA),
phospholipids, trace metals, oxidation products, and odoriferous
substances. Thus, as a consequence, palm oil is normally refined
to obtain a stable product, which can be used for direct consump-
tion or for formulation of edible products [1].
Two methods are available for refining crude palm oil: chemical
and physical refining. They differ basically in the manner in which
the free fatty acids are removed (deacidification step). Chemical
refining, which involves neutralization of FFA with a solution of
sodium hydroxide, is not recommended for oils with high acidity,
such as palm oil, because it causes high losses of neutral oil by
saponification and emulsification. Because of its efficiency and
simple effluent treatment, physical refining is the major processing
route for removing FFA from palm oil. However, the drastic condi-
tions in which this process is carried out (temperature in the range
240260 C and pressure in the range 13 mmHg) led to the com-
plete destruction of carotenoids [2] and to a significant reduction
of tocopherols [3], both important components that confer to palm
Corresponding authors.
E-mail addresses: cintiabg@usp.br (C.B. Gonalves), tomze@unicamp.br
(A.J.A. Meirelles).
oil a high nutritional value, and promote several benefits to human
health [4]. Nevertheless, some studies indicate that the use of
alternative technologies [5] or optimized conditions in the deodor-
ization step of refining [6] can lead to retention of significant
amounts of these important nutraceutical compounds.
The deacidification of oils by liquidliquid extraction using an
appropriate solvent, such as ethanol, can be an alternative tech-
nique for refining palm oil. As this process is carried out at room
temperature and atmospheric pressure, less energy is consumed
and the oil is submitted to softer treatments, potentially preserving
the nutraceutical compounds. This technique is based on the dif-
ference of solubility of FFA and triacylglycerols in an appropriate
solvent [7]. Previous works indicate the decrease of FFA content
in the oil submitted to solvent extraction, as well as, the losses of
neutral oil and nutraceutical compounds during this process are
also reported [715]. Literature also provides liquidliquid equilib-
rium data for systems containing triacylglycerols (TAG), free fatty
acids (FFA) and short chain alcohols, which are essential to plan-
ning and developing liquidliquid extraction process [1626]. In
our prior works [20,21], liquidliquid equilibrium data of systems
containing palm oil, fatty acids (palmitic and oleic acids), aqueous
ethanol and nutraceutical compounds were measured and
correlated by thermodynamic models.
The present work reports the influence of process variables on
the losses of neutral oil, the transfer of free fatty acids and the
preservation of carotenoids during the deacidification of palm oil
by liquidliquid extraction. The response surface methodology
(RMS) was used to analyze the effect of process variables, such as

http://dx.doi.org/10.1016/j.seppur.2016.01.016
1383-5866/ 2016 Elsevier B.V. All rights reserved.
 C.B. Gonalves et al. / Separation and Purification Technology 160 (2016) 106111
mass ratio of oil to solvent and water content in the solvent, aiming
to minimize the losses of neutral oil and maximize the transfer of
free fatty acids plus carotenoids preservation. By using appropriate
conditions observed in RSM analysis, the deacidification of palm oil
by continuous liquidliquid extraction was performed in a perfo-
rated rotating disc contactor (PRDC).
2. Material and methods
In this study, two different samples of bleached palm oil (kindly
supplied by the Agropalma, Par, Brazil) were used. Both oils were
analyzed by gas chromatography of fatty acid methyl esters, accord-
ing to the official method Ce 162 [27], given that the samples were
prepared in the form of fatty acid methyl esters according to the
official method Ce 266 [27]. An HP 5890 gas chromatograph with
a flame ionization detector and an integrator was used under the
following experimental conditions: capillary fused silica column
of cyanopropylsiloxane (60 m
0.25 lm
0.32 mm), hydrogen as
the carrier gas at a rate of 2.5 ml min1, an injection temperature
of 548.2 K, a column temperature of 448.2498.2 K (1.3 K min1),
and a detection temperature of 578.2 K. The fatty acid methyl esters
were identified by comparison with the retention times of NU
CHECK Inc. standards (Elysian, IL, USA) and the quantification was
accomplished by internal normalization.
Fatty acid composition of the bleached palm oil used in the
liquidliquid equilibrium experiments (for composing the
experimental design) has already been reported by Gonalves
and Meirelles [20]. Such sample presented 3.88 0.01 mass% of
free fatty acids, determined by titration (official method 2201 of
IUPAC) [28] with an automatic burette (Metrohm, model Dosimat
215, Herisan, Switzerland) and 255 0.01 mg kg1 of total carote-
noids, determined by spectrophotometry (model Lambda 40, Per-
kin Elmer, Waltham, MA, USA) at 450 nm, according to the
method developed by the Palm Oil Research Institute of Malaysia
(Porim Test Methods) [29], using hexane (Merck, Germany) as sol-
vent and b-carotene 99% (Sigma) as standard. The solvents used
were anhydrous ethanol (from Merck, Germany), with purity
greater than 99.5%, and alcoholic solutions containing 1.91 0.03,
5.76 0.02, 10.00 0.05 and 12.41 0.01 mass% of water, prepared
by the addition of deionized water (Milli-Q, Millipore) to anhy-
drous ethanol. The water concentration in the solvent was deter-
mined by Karl Fisher titration, according to the official method
Ca 2355 [27].
For the PRDC experiments, a bleached palm oil (BPO) containing
4.23 0.01 mass% of free fatty acids and 225 0.01 mg kg1 of
total carotenes was utilized. In addition, such sample was charac-
terized in relation to mono-, di-, triacylglycerols and polymerized
triacylglycerols by gel permeation HPLC according to the official
method Cd 22-91 [27] (HPLC system Perkin Elmer model 250,
refractive index detector Sicon Analytic, columns Jordi Gel DVB
300 mm
7.8 mm id, 0.01 and 0.05 lm, mobile phase tetrahydro-
furane, sample solution 1% (w/v) in tetrahydrofurane); peroxide
value, according to the official method Cd 8b-90 [27]; iodine and
saponification values calculated from fatty acid composition, fol-
lowing AOCS methods Cd 1c-85 and Cd 3a-94, respectively [27];
total tocopherols (tocopherols and tocotrienols) by HPLC (AOCS
official method Ce 8-89 [27]; HPLC system Perkin Elmer model
250, fluorescence detector Shimadzu RF-10 AXL with excitation
wavelength at 290 nm and emission wavelength at 330 nm, col-
umn Merck Li Chrosorb Si 60, 5 lm, 250 mm
4 mm id, mobile
phase isopropanol in hexane 1:99 (v/v)); total carotenoids,
according to Porim Test Methods [29]; and Lovibond color read
in a Lovibond Tintometer model E in a 5.2500 cell and expressed
in units of yellow (Y), red (R) and blue (B). For the experiments
in the PRDC, neutral ethanol (containing 5.8 mass% of water), food
107
grade, purchased from Usina Ester (Cosmpolis, Brazil) was used as
solvent.
Physical properties of the oils samples at 45 C were also mea-
sured. The density measurements were performed using a DMA 58
Density Meter (Anton Paar, Graz, Austria). The dynamic viscosity
data were obtained by an AMV 200 Viscometer (Anton Paar, Graz,
Austria).
In order to compare the characteristics of oils submitted to dif-
ferent steps of refining process. the analysis described above were
also performed in a crude palm oil (CPO), in a refined palm oil
(RPO), both supplied by Agropalma (Par, Brazil), and in the palm
oil deacidified by liquidliquid extraction (RPO-LLE) obtained in
this work.
2.1. Response surface methodology
Liquidliquid equilibrium experiments were accomplished by
mixing bleached palm oil with ethanolic solvents (water content
in solvent varying from 0 to 12 mass%) at different oil to solvent
mass ratios (O/S varying from 0.36 to 2.77). The components were
weighed on an analytical balance (Adam, model A250L, Milton
Keynes, United Kingdom), accurate to 0.0001 g, and placed in
polypropylene tubes (15 ml, Corning Inc., Lowell/MA, USA). The
tubes were vigorously stirred for at least 15 min and left to rest
for 24 h in a thermostatic bath (Cole-Parmer, model 12101-05,
Chicago, IL) at 45 0.1 C, temperature in which the palm oil
samples were completely in liquid state.
After phase equilibrium is established, samples of both phases
were taken and analyzed. The concentration of free fatty acids
was determined by titration according to IUPAC official method
2201 [28], using an automatic burette (Metrohm, model Dosimat
715, Herisan, Switzerland). The total solvent concentration was
determined by evaporation at 65.0 C in a vacuum oven (Napco,
model 5831, New York, USA), with an inner absolute pressure
equal to 126 mmHg. The water concentration was determined by
Karl Fischer titration, according to AOCS official method Ca 23-55
[27], with a KF Titrino (Metrohm, model 701, Herisan, Switzer-
land). The quantification of total carotenoids was determined by
spectroscopy at 450 nm, according to Porim Test Methods [29]
[28]. The neutral oil (or triacylglycerol) concentration was deter-
mined by difference. All measures were performed at least in
triplicate.
The response surface methodology was used to investigate the
effect of the oil to solvent mass ratio (O/S) and the water content
in solvent (%Water) on the carotenoids maintenance (CaroteneOil),
losses of neutral oil (LNO) and on the free fatty acid transfer (TFFA)
during one equilibrium stage of the deacidification process by liq-
uidliquid extraction.
The TFFA and the LNO were calculated by Eq. (1).
mAP  wAP
T FFA or LNO % 100  i
1
mOil  wOil
i
where m is mass, w is mass fraction, AP is alcohol phase and i is free
fatty acid or neutral oil, given that mOil is the initial mass of palm oil
and mAP is calculated by mass balance.
The carotenoids maintenance was expressed as the respective
content remaining in oil (Caroteneoil, in mg kg1) after one stage
of equilibrium (Eq. (2)).
1
 1
CaroteneOil mg kg 1  wOP OP
solv  Carotene mg kg 2
where solv is the solvent (ethanol plus water) and OP is oil phase.
The experimental set was planned to obtain a quadratic model,
consisting of 22 trials plus a star configuration with three
repetitions in central point [30,31]. Response surfaces were built
using the quadratic model for the statistically significant variables.
108 C.B. Gonalves et al. / Separation and Purification Technology 160 (2016) 106111

The software Statistica (Statsoft, v. 8.0) was used to analyze the
results by non-linear multiple regression.
2.2. Deacidification in continuous equipment
Palm oil deacidification were performed in a jacketed glass per-
forated rotating disc contactor (PRDC), which is a continuous
equipment that consists of a column equipped with a central rotat-
ing shaft carrying equally spaced perforated discs (total of 33),
whose dimensions (in mm) are as follows: column inside diameter,
50; disc diameter, 47; column height, 1300; extraction zone height,
1000; distance between adjacent discs, 25. The flow free area in the
discs was 20%, containing holes of 3 mm diameter. A schematic
drawing of the PRDC can be found in our previous work [12].
The experiments were accomplished at 45 0.1 C and local
atmospheric pressure (727 mmHg), given that the column temper-
ature was controlled by a thermostatic bath (Cole-Parmer, Model
12101-15, Chicago, IL) connected to the column jacket. The equip-
ment was filled with food-grade ethanol containing
5.80 0.14 mass% of water (continuous phase) into the bottom of
the column and its flow rate was maintained at a constant value
equal to 24.85 g min1. The rotor was started and the rotating
speed was fixed at 150 rpm, checked by a digital tachometer
1726 (Ametek, Largo, FL). Subsequently, the bleached oil (dispersed
phase) was fed to the top of the column with a constant flow rate
equal to 18.39 g/min, resulting in an oil to solvent mass ratio equal
to 0.74. Both feed streams (oil and ethanol) were pumped into the
and the main fatty acids present in bleached palm oil, palmitic
and oleic acids, can be approximately replaced by an equivalent
pseudo-fatty acid for equilibrium calculations, as already shown
by Gonalves et al. [17], Rodrigues et al. [18,19] and Gonalves
and Meirelles [20]. Furthermore, this approach allows a first esti-
mation of the number of ideal stages required for the deacidifica-
tion process, information that can be helpful in the evaluation of
this refining technology.
3. Results
Table 1 presents all combinations of the studied variables in the
statistical analysis and the respective responses for the experimen-
tal design studied.
The statistical analysis of the experimental results allowed for-
mulating models representing the free fatty acid transfer (TFFA), the
losses of neutral oil (LNO) and carotene content in oil (Caroteneoil),
given by Eqs. (5) to (7), respectively.
T FFA % 97:38  51:76  O=S 7:85  O=S2  1:58  %Water
0:84  %Water  O=S 5
LNO % 27:40  16:73  O=S 2:80  O=S2  3:34  %Water
0:12  %Water2 0:83  O=S  %Water 6
Oil 1
Carotene mg  kg 159:07 4:72  %Water 7
In Eqs. (5)(7), to obtain the values of T , L and CaroteneOil,
FFA NO

column by peristaltic pumps (Cole Parmer, Chicago, IL). After a
waiting time of 120 min for attaining the steady state, samples of
the outlet streams, extract and raffinate, were taken during the fol-
lowing 120 min in order to quantify the concentrations of solvent
[27], fatty acids [28] and carotenoids [29]. This procedure was
repeated using the raffinate stream of the prior experiment and
fresh solvent as feeds until the free acidity in refined oil was less
than 0.3 mass%, totalizing three experiments (or steps) in the PRDC
column. Each processing step is a countercurrent contact while the
global process, i.e., the set of three steps, is a crosscurrent configu-
ration, once fresh solvent is introduced in each one.
The operational and free fatty acids equilibrium concentrations
in the raffinate stream allowed calculating the number of ideal
equilibrium stages, N, required in each step (Eq. (3)). This approach
is valid when the operating and equilibrium lines are both straight
over a given concentration range, and when just one compound (in
this case, FFA) is transferred from one phase to another. For this, it
is necessary to use streams concentrations in an acid free-basis
[11,32].
h   i
w0RFFA  w0RFFA = w0F FFA  w0EFFA
log
N h   i
log w0EFFA  w0RFFA = w0F FFA  w0RFFA
the independent variables %Water and O/S, must be replaced by
the real values presented in Table 1.
Table 2 shows the analysis of variance (ANOVA) for the
responses at 95.0% of confidence.
As can be observed in Table 2, the responses TFFA and LNO pre-
sented high correlation coefficients, and the F-test shows that the
respective models are reliable since the calculated F values are at
least 12 times greater than the values obtained from Box et al.
[30]. However, it is important to emphasize that Eq. (6) does not
provided a good prediction when very low values of LNO are consid-
ered. In relation to the response CaroteneOil the correlation coeffi-
cient has not been very high, the F value is 6 times greater than
the tabled one [30] at 95% of confidence.
Figs. 13 present the contour curves by the models obtained in
Eqs. (5)(7), representing the influence of the oil to solvent mass
ratio (O/S) and of the water content in solvent (% Water) on the
responses studied.
As can be observed in Fig. 1, lower O/S mass ratios and lower
water contents in the solvent provide a better transference of the
free fatty acids to the alcoholic phase. However, this effect is more
pronounced in the case of the variable O/S mass ratio.
3 Fig. 2 shows that the losses of neutral oil are minimized with
the increase of the water content in the solvent, but the reduction

In Eq. (3), wF FFA , wEFFA and wRFFA represent the concentration of

fatty acids in the feed, extract and raffinate streams, respectively; Table 1
the superscripts 0 and indicate acid-free basis and equilibrium Experimental design: 22 + star configuration + central points.

concentration, correspondingly [33]. The equilibrium curve used Coded variables Real Variables Responses
in the calculations was obtained from equilibrium data reported O/S %Water O/S %Water TFFA (%) LNO (%) CaroteneOil (mgkg1)
in Gonalves and Meirelles [20] for systems containing bleached
+1 +1 2 10.00 25.46 0.38 189.43
palm oil, free fatty acids, ethanol and water, and it presents a cor- +1 1 2 1.91 24.96 2.49 158.65
relation coefficient of 0.99. This line is given in Eq. (4). 1 +1 0.5 10.00 63.75 2.69 203.24
1 1 0.5 1.91 74.09 14.92 175.35
w0EFFA 1:1627  w0RFFA 4 0 0 1 5.76 51.00 3.07 195.23
0 0 1 5.76 49.93 2.92 186.07
It should be noted that the approach mentioned above requires 0 0 1 5.76 49.85 2.94 184.53
a one-component mass transfer system. For this reason its use in 1.41 0 0.36 5.76 68.26 7.93 198.71
the present case is a first approximation, since other fatty com- 0 1.41 1 0 51.42 13.24 157.92
+1.41 0 2.77 5.76 19.05 0.43 179.59
pounds classes are also transferred during the process. Neverthe-
0 +1.41 1 12.41 43.12 0.78 228.11
less, the fatty acids are the major components to be transferred
 C.B. Gonalves et al. / Separation and Purification Technology 160 (2016) 106111 109

Table 2
Analysis of variance (ANOVA).

Source of variation TFFA LNO CaroteneOil

a b
SS MS DF F SS MS DF F SS MS DF Fc
Regression 3211.0 802.75 4 255.11 51.02 5 3115.5 3115.5 1
Residual 51.91 8.65 6 92.80 4.00 0.80 5 63.78 896.01 99.56 9 31.29
Total 3262.9 10 259.11 10 4011.5 10
Correlation coefficient 0.988 0.999 0.793

SS = sum of squares; MS = mean square; DF = degrees of freedom;

Fa0.95;4;6 = 4.53; Fb0.95;5;5 = 5.05; Fc0.95;1;9 = 5.12.

Fig. 1. Contour curves of TFFA (%) expressed as function of real variables for mass
ratio oil to solvent (O/S) and water content in solvent (%Water). Fig. 3. Contour curves of CaroteneOil (mg kg1) expressed as function of real
variables for mass ratio oil to solvent (O/S) and water content in solvent (%Water).

remaining in the oil after one stage of equilibrium in the deacidifi-

Fig. 2. Contour curves of LNO (%) expressed as function of real variables for mass
ratio oil to solvent (O/S) and water content in solvent (%Water).
of the mass ratio O/S only exerts a more significant influence on the
this response when solvents with lower water contents are used.
In Fig. 3, it can be observed that the water content in the solvent
is the main effect on the response CaroteneOil. As higher the water
concentration in the solvent, larger the carotene concentration
cation process. However, even if solvents with low concentrations
of water are used, it is possible to retain 65 mass% of the total of
carotenoids present in bleached oil, i.e., all the range of variables
studied provides a good preservation of the carotenoids in oil. On
the other hand, for TFFA and LNO such variables exert a significant
and opposing influence, as can be observed in Figs. 1 and 2. Thus,
it is important to specify an appropriate region in which it is
possible to obtain a good transference of the FFA without great loss
of neutral oil.
As the loss of neutral oil has a significant effect on the total cost
of refining process, it is important to establish an acceptable
maximum limit. Many suppliers of physical refining systems offers
loss warranties based on the amount of fatty acids in the feed. They
usually claim a minimum loss ranging between 0.2% and 0.4% plus
1.11.2 times the FFA content in the feed [34]. Applying these
limits to the bleached palm oil (with 4.23 mass% of FFA), which
was used in the perforated rotating disc contactor experiments, it
can be considered that a loss of neutral oil less than 4.855.48%
would be acceptable for liquidliquid extraction.
Analyzing Fig. 2, it can be observed that several combinations of
mass ratio of oil to solvent and water content in solvent turn
possible the deacidification of palm oil with losses of neutral oil
less than the stipulated value, in one stage of equilibrium. Consid-
ering Fig. 1, it can be seen that high values of TFFA (>50%) were
obtained for mass ratios of oil to solvent less than 1.0. Applying
this restriction in Fig. 2, the range of water concentration in the
solvent is also limited for values higher than 4%. However, it is
not reasonable to choose high values of water concentration
110 C.B. Gonalves et al. / Separation and Purification Technology 160 (2016) 106111

Table 3 Table 5
Experimental free fatty acid transfer (TFFA) and neutral oil loss (LNO). Chemical and physical characteristics of palm oils.

Transfer/loss Step 1 Step 2 Step 3 Whole process Characteristic CPO BPO RPO-LLE RPO Codex [35]
N = 2.5 N = 1.0 N = 1.0
FFA (mass%) 4.14 4.23 0.14 0.14 <0.3 mass%
TFFA (%) 81.24 70.71 26.41 98.95 Carotenoids (mg kg1) 647.3 224.5 184.6 ndf 5002000g
LNO (%) 4.95 4.02 1.49 10.67 Tocopherols (mg kg1) 691.6 718.1 218.8 322.3 1501500
DAGa + MAGb (mass%) 6.97 8.14 0.81 8.46 nah
IV c 54.1 55.1 55.0 53.6 5055
SVd (mg KOH/g oil) 197.7 198.0 197.4 197.7 190209
(>7 mass%, for example), once more equilibrium stages would be PVe (mEq kg1) 6.76 11.49 7.48 1.08 <10
necessary to obtain a palm oil containing less than 0.3 mass% of Lovibond color Red (R) 29 11 10 4 nah
FFA. This statement will be corroborated ahead through Yellow (Y) 21 20 20 20 nah
Density (kg m3) at 45 C 899.6 896.6 896.32 900.85 891899
calculations.
Viscosity (mPa s) at 45 C 37.16 30.29 30.96 40.01 nah
Values within this selected range (mass ratio O/S = 0.74 and
a
water content in solvent = 5.8 mass%) were used to accomplish Diacylglicerol.
b
Monoacylglicerol.
the experiments in the perforated rotating disc contactor (PRDC). c
Iodine value.
Concerning the experimental conditions of mass ratio O/S (0.74) d
Saponification value.
and water content in solvent (5.8 mass%) in Eqs. (5) and (6), it e
Peroxide value.
was possible to obtain, for one equilibrium stage, a TFFA equal to f
Not detected.
g
57.81% and a LNO equal to 4.78%. Three experimental steps in the h
In unbleached palm oil.
Not available.
PRDC were necessary to obtain refined oil with the final acidity
required by the Codex Alimentarius [35] for refined vegetable oils.
Using the fatty acid concentrations obtained after each experi-
mental step, Eq. (3) provides the following number of ideal equilib- in solvent, higher the number of equilibrium stages), the calcula-
rium stages: step 1 N = 2.5; step 2 N = 1.0; step 3 N = 1.0. tions described above were also performed using the equilibrium
Table 3 presents the experimental results of TFFA and LNO for each data for the system palm oil, fatty acids and 12.41 mass% aqueous
processing step and for the whole process. ethanol taken from Gonalves and Meirelles [20]. It resulted in a
As observed in Table 3 most part of FFA content in BPO was number of ideal stages equal to 32, much higher than 7.5.
transferred in the first step, but the following two steps were nec- In order to evaluate the impact of the deacidification by liquid
essary to attain the required final acidity. The loss of neutral oil liquid extraction on the quality of refined oil, several analyses were
was also large in the first step; the reason for the observed behav- accomplished in the bleached palm oil used in the experiments
ior relies on the higher solubility of the neutral oil in the alcoholic (BPO) and in the refined palm oil deacidified by liquidliquid
phase when FFA concentration in this phase is high. In fact, the extraction (RPO-LLE). Such analyses were also performed in further
equilibrium data published in our previous work [20] indicate that two different palm oil samples: crude palm oil (CPO) and industri-
neutral oil has a limited solubility in aqueous ethanol, whose value ally refined palm oil (RPO).
is enhanced by the presence of FFAs. The LNO is also significant in Tables 4 and 5 show, respectively, the fatty acid composition
the other two steps, since fresh solvent was used in the last two and the physicalchemical properties of CPO, BPO, RPO and
ones. RPO-LLE.
If it were possible to operate the whole process in a countercur- As can be observed in Table 4, the fatty acid composition of the
rent way, with fresh solvent being fed just once, the LNO would be palm oil are not affected by liquidliquid extraction, presenting not
lower, since the neutral oil solubility limit in the alcoholic phase significant differences in comparison with the palm oil industrially
would be attained. In this case, the LNO value would not be higher refined, the bleached palm oil used in the experiments and the
than 4.95%, a value near the estimated by the response surface crude palm oil.
methodology. Concerning the number of ideal stages in a counter- Table 5 shows that liquidliquid extraction process allowed the
current configuration for the whole process, it can be estimated by deacidification, promoting the attainment of a refined palm oil
Eq. (3) using FFA concentrations in the feed stream of first step and containing a free acidity of 0.14 mass% in solvent free-basis, and
in the output stream of the third step. This calculation results in a maintaining a significant level of carotenoids.
number of equilibrium stages equal to 7.5, higher than the sum of As also presented in Table 5, liquidliquid extraction reduced
stages estimated for each experimental step. With the purpose of the peroxide value (PV) from 11.49 (measured in BPO) to 7.48
confirming the previous statement (as higher the water content (measured in RPO-LLE), but in comparison with the traditionally

Table 4
Fatty acid composition of Crude Palm Oil (CPO), Bleached Palm Oil (BPO), Refined Palm Oil (RPO), and Refined Palm Oil deacidified by LiquidLiquid Extraction (RPO-LLE).

Symbol Fatty acid CX:Ya Mb (g mol1) CPO BPO RPO-LLE RPO

mol% mass% mol% mass% mol% mass% mol% mass%
L Lauric C12:0 200.32 0.00 0.00 0.34 0.25 0.00 0.00 0.00 0.00
M Myristic C14:0 228.38 1.09 0.92 1.25 1.06 1.03 0.87 1.10 0.93
P Palmitic C16:0 256.43 44.16 41.89 42.73 40.53 42.93 40.66 44.41 42.13
Po Palmitoleic C16:1 254.42 0.18 0.17 0.33 0.31 0.18 0.17 0.16 0.15
S Stearic C18:0 284.49 4.76 5.01 4.71 4.96 4.83 5.08 4.74 4.99
O Oleic C18:1 282.47 39.00 40.75 39.71 41.49 40.28 42.03 39.00 40.76
Li Linoleic C18:2 280.45 10.01 10.38 10.11 10.49 9.94 10.30 9.94 10.31
Le Linolenic C18:3 278.43 0.30 0.31 0.31 0.32 0.27 0.28 0.17 0.18
A Arachidic C20:0 312.54 0.35 0.41 0.35 0.41 0.37 0.43 0.35 0.40
Ga Gadoleic C20:1 310.52 0.14 0.16 0.16 0.18 0.16 0.18 0.13 0.15
a
In CX:Y, X = number of carbons, Y = number of double bonds.
b
M = molar mass.
 C.B. Gonalves et al. / Separation and Purification Technology 160 (2016) 106111
refined palm oil (RPO), such value is high. Analyzing the Lovibond
color results, it can be observed that yellow factor (Y) remained the
same along traditional refining process and after liquidliquid
extraction. On the other hand, the red factor (R) decreased after
traditional deacidification steps, probably due to reduction of car-
otenoids content, but in the case of palm oil processed by liquid
liquid extraction the red factors has almost the same value mea-
sured for the bleached oil. In relation to mono-acylglycerols
(MAG) and di-acylglycerols (DAG), the results show that LLE pro-
motes a considerable reduction of these compounds. It is impor-
tant to note that great part of losses of neutral oil shown in
Table 3 can be a consequence of the reduction of MAG and DAG
levels. This result is positive, once these partial acylglycerols may
cause foaming and bitter taste in oil [36]. In addition, the gel per-
meation HPLC analysis showed that polymeric acylglycerols were
not detected in any samples studied. The results also show that
the iodine and saponification values did not suffer significant
changes after liquidliquid extraction, indicating that RPO-LLE
maintains the same characteristics of the RPO.
4. Conclusions
The response surface methodology analysis allowed the selec-
tion of better process conditions that maximize the FFA transfer,
minimize the loss of neutral oil and preserve the carotenoids.
The deacidification of palm oil by solvent extraction using some
selected conditions obtained in RSM were accomplished success-
fully, permitting the attainment of a refined oil (% FFA < 0.3 mass%)
with high concentration of carotenoids, so maintaining its
nutritional value.
Acknowledgements
The authors wish to acknowledge FAPESP (99/12033-1,
00/01685-7; 01/13733-9, 12/23203-1, 14/09446-4), CAPES, CNPq
(308024/2013-3) and FINEP for the financial support.
References
[1] Y. Barison, Palm oil, in: F. Shahidi (Ed.), Baileys Industrial Oil and Fat Products,
vol. 2, Wiley Interscience, New Jersey, 2005, pp. 333429.
[2] A. Szydowska-Czerniak, K. Trokowski, G. Karlovits, E. Szyk, Effect of refining
processes on antioxidant capacity, total contents of phenolics and carotenoids
in palm oils, Food Chem. 129 (2011) 11871192, http://dx.doi.org/10.1016/
j.foodchem.2011.05.101.
[3] T.M. Deboni, M.S. Cuevas, P.M. Neto, R.V. Mota, F.S. Damasceno, L.H.M. Silva,
et al., Deacidification of soybean oil by ion exchange, Food Bioprocess Technol.
(2012), http://dx.doi.org/10.1007/s11947-012-1022-3.
[4] M.S. Cuevas, T.M. Deboni, P. Mielke Neto, F.S. Damasceno, R.V. Mota, L.H.M. Da
Silva, et al., Using a strong anion-exchange resin to deacidify red palm oil, J.
Am. Oil Chem. Soc. 90 (2013) 15891597, http://dx.doi.org/10.1007/s11746-
013-2305-1.
[5] H.L.N. Lau, Y.M. Choo, A.N. Ma, C.H. Chuah, Production of refined carotene-rich
palm oil from palm mesocarp (Elaeis guineensis) using supercritical carbon
dioxide, J. Food Lip. 14 (2007) 396410. Journal of Food Lipids 14 (2007) 396
410.
[6] V. Martincic, J. Golob, W. Greyt, R. Verh, S. Knez, V.V. Hoed, L. FeleZilnik, K.
Potocnik, A. Rizner Hra, J.V. Ayala, Optimization of industrial-scale
deodorization of high-oleic sunflower oil via response surface methodology,
Eur. J. Lipid Sci. Technol. 110 (2008) 245253.
[7] C. Thomopoulos, Mthod de Desacidification des Huiles par Solvant Slectif,
Rev. Fran. des Corps Gras. 18 (1971) 143150.
[8] A.C. Bhattacharyya, S. Majumdar, D.K. Bhattacharyya, Refining of high FFA rice
bran oil by isopropanol extraction and alkali neutralization, Olagineaux 42
(1987) 431433.
111
[9] K.J. Shah, T.K. Venkatesan, Aqueous isopropyl alcohol for extraction of free
fatty acids from oils, J. Am. Oil Chem. Soc. 66 (1989) 783787.
[10] V. Kale, S.P.R. Katikaneni, M. Cheryan, Deacidifying rice bran oil by solvent
extraction and membrane technology, J. Am. Oil Chem. Soc. 76 (1999) 723
727.
[11] C.G. Pina, A.J.A. Meirelles, Deacidification of corn oil by solvent extraction in a
perforated rotating disc column, J. Am. Oil Chem. Soc. 77 (2000) 553559.
[12] C.E.C. Rodrigues, C.B. Gonalves, E.C. Marcon, E.A.C. Batista, A.J.A. Meirelles,
Deacidification of rice bran oil by liquidliquid extraction using a renewable
solvent, Sep. Purif. Technol. 132 (2014) 8492, http://dx.doi.org/10.1016/j.
seppur.2014.05.009.
[13] S. Kim, C. Kim, H. Cheigh, S. Yoon, Effect of caustic refining, solvent refining and
steam refining on the deacidification and color of rice bran oil, J. Am. Oil Chem.
Soc. 62 (1985) 14921495.
[14] S. Turkay, H. Civelekoglu, Deacidification of sulfur olive oil. I. Single-stage
liquidliquid extraction of miscella with ethyl alcohol, J. Am. Oil Chem. Soc. 68
(1991) 8386.
[15] S. Turkay, H. Civelekoglu, Deacidification of sulfur olive oil. II. Multi-stage
liquidliquid extraction of miscella with ethyl alcohol, J. Am. Oil Chem. Soc. 68
(1991) 818821.
[16] E. Batista, S. Monnerat, L. Stragevitch, A.J.A. Meirelles, Liquid liquid
equilibrium for systems of canola oil, oleic acid, and short-chain alcohols, J.
Chem. Eng. Data. (1999) 13601364.
[17] C.B. Gonalves, E. Batista, A.J.A. Meirelles, Liquid liquid equilibrium data for
the system corn oil + oleic acid + ethanol + water at 298.15 K, J. Chem. Eng.
Data (2002) 416420.
[18] C.E.C. Rodrigues, R. Antoniassi, A.J.A. Meirelles, Equilibrium data for the system
rice bran oil + fatty acids + ethanol + water at 298.2 K, J. Chem. Eng. Data
(2003) 367373.
[19] C.E.C. Rodrigues, P.A.P. Filho, A.J.A. Meirelles, Phase equilibrium for the system
rice bran oil + fatty acids + ethanol + water + g-oryzanol + tocols, Fluid Phase
Equilib. 216 (2004) 271283, http://dx.doi.org/10.1016/j.fluid.2003.11.008.
[20] C.B. Gonalves, A.J.A. Meirelles, Liquidliquid equilibrium data for the system
palm oil + fatty acids + ethanol + water at 318.2 K, Fluid Phase Equilib. 221
(2004) 139150, http://dx.doi.org/10.1016/j.fluid.2004.05.002.
[21] C.B. Gonalves, P.A. Pessa Filho, A.J.A. Meirelles, Partition of nutraceutical
compounds in deacidification of palm oil by solvent extraction, J. Food Eng. 81
(2007) 2126, http://dx.doi.org/10.1016/j.jfoodeng.2006.10.003.
[22] M.S. Cuevas, C.E.C. Rodrigues, G.B. Gomes, A.J.A. Meirelles, Vegetable oils
deacidification by solvent extraction: liquid liquid equilibrium data for
systems containing sunflower seed oil at 298.2 K, J. Chem. Eng. Data (2010)
38593862.
[23] M. Mohsen-Nia, H. Modarress, H.R. Nabavi, Measuring and modeling liquid
liquid equilibria for a soybean oil, oleic acid, ethanol, and water system, J. Am.
Oil Chem. Soc. 85 (2008) 973978.
[24] M.G. Granero, C.E.C. Rodrigues, C.B. Gonalves, Liquidliquid equilibrium data
for the system lard + oleic acid + ethanol + water at 318.2 K: cholesterol
distribution coefficients, J. Chem. Eng. Data (2012) 210.
[25] C.E.C. Rodrigues, A. Filipini, A.J.A. Meirelles, Phase equilibrium for systems
composed by high unsaturated vegetable oils + linoleic acid + ethanol + water
at 298.2 K, J. Chem. Eng. Data (2006) 1521.
[26] C.E.C. Rodrigues, F.A. Silva, A. Marsaioli Jr, A.J.A. Meirelles, Deacidification of
Brazil nut and macadamia nut oils by solvent extraction: liquidliquid
equilibrium data at 298.2 K, J. Chem. Eng. Data (2005) 517523.
[27] AOCS, fifth ed., in: D. Firestone (Ed.), Official Methods and Recommended
Practices of the American Oil Chemists Society, AOCS Press, Champaign, 1997.
[28] IUPAC, seventh ed., in: C. Paquot (Ed.), Standard Methods for the Analysis of
Oils, Fats and Derivatives, Blackwell, Oxford, 1987.
[29] Porim Test Methods Caroten Content, Palm Oil Research Institute of Malaysia,
1990.
[30] G.E.P. Box, W.G. Hunter, J.S. Hunter, Statistic for Experimenters An
Introduction to Design, Data Analysis and Model Building, John Wiley &
Sons, New York, 1978.
[31] A.I. Khuri, J.A. Cornell, Response Surface-Design and Analysis, ASQC Quality
Press, New York, 1987.
[32] R.E. Treybal, Mass Transfer Operations, third ed., McGraw-Hill, New York,
1980.
[33] W.L. McCabe, J.C. Smith, Unit Operations of Chemical Engineering, third ed.,
McGraw-Hill Book, New York, 1976.
[34] W. De Greyt, M. Kellens, Deodorization, in: F. Shahidi (Ed.), Baileys Industrial
Oil and Fat Products, vol. 5, Wiley Interscience, New Jersey, 2005, pp. 341383.
[35] Codex Alimentarius fats, oils and related products joint FAO/ WHO food
and Agriculture Organization of the United Nations and World Health
Organization, Rome, 1993, vol. 8, pp. 2932.
[36] M. Stuchlk, S. Zk, Lipid-based vehicle for oral drug delivery, Biomed. Pap. 15
(2001) 1726.