Journal of Food Engineering 69 (2005) 217–223 www.elsevier.com/locate/jfoodeng
Supercritical carbon dioxide extraction of hazelnut oil ¨ zkal a, U. Salgın b, M.E. Yener S.G. O a
a,*
Department of Food Engineering, Middle East Technical University, 06531 Ankara, Turkey b Department of Chemical Engineering, Ankara University, 06100 Ankara, Turkey Received 17 March 2003; revised 29 April 2004; accepted 15 July 2004
Abstract Solubility of hazelnut oil in supercritical carbon dioxide (SC-CO2) was determined at 15–60 MPa, and 40–60 °C. The crossover pressure of hazelnut oil was between 15 and 30 MPa. The solubility increased with pressure, but increased with temperature above the crossover pressure. Hazelnut particles (1–2 mm) were extracted at 30–60 MPa, and 40–60 °C with SC-CO2 flow rate of 2 ml/min. Extraction occurred in two periods. The released oil on the surface of particles was extracted in the fast extraction period, and 39% of the initial oil was recovered at each condition. However, the duration of the fast extraction period decreased with increased pressure and temperature. The unreleased oil in the intact cells was extracted in the slow extraction period. The maximum recovery was 59% at 60 MPa and 60 °C, for 180 min of extraction. The fluid phase and solid phase mass transfer coefficients increased with increased pressure and temperature. Ó 2004 Elsevier Ltd. All rights reserved. Keywords: Hazelnut oil; Solubility; Extraction; Supercritical carbon dioxide; Mass transfer
1. Introduction Supercritical carbon dioxide (SC-CO2) extraction offers the advantages of using non-toxic, non-explosive and cost effective solvent. It enables extraction at low temperatures and, easy and complete removal of solvent from the final product. With SC-CO2, extraction of almond (Marrone, Poletto, Reverchon, & Stassi, 1998; ¨ nal & Pala, Passey & Gros-Louis, 1993), hazelnut (U 1996), peanut (Goodrum & Kilgo, 1987; Santerre, Goodrum, & Kee, 1994), pecan (Li, Bellmer, & Brusewitz, 1999; Zhang, Brusewitz, Maness, & Gasem, 1995), and pistachio nut (Palazogˇlu & Balaban) oils have been studied. These researches mostly focus on the reduction of the oil content of nuts. One of the main reasons is to produce a low calorie product by decreasing the oil con-
*
Corresponding author. Tel.: +90 312 2105630; fax: +90 312 2101270. E-mail address:
[email protected] (M.E. Yener). 0260-8774/$ - see front matter Ó 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.jfoodeng.2004.07.020
tent, as the consumption of oil in the daily diet is usually high. The second one is to prolong the shelf life of these products, since high oil content that consists of majorly unsaturated fatty acids, leads to fast deterioration due to oxidation. However, besides their contribution to human health, these oils have potential uses in designing new margarine and butter formulations to improve the physical properties (e.g. spreadability) of the product due to their lower viscosities compared to saturated fats. Hazelnut is the most produced nut in Turkey. It has a high nutritional value, containing, generally, 65% oil, 14% protein, and 16% carbohydrates. More than 90% of its oil consists of unsaturated fatty acids, especially oleic (C18:1, 80%) and linoleic (C18:2, 12%) acids. Although the extraction of hazelnut oil with a batch ¨ nal & Pala, 1996) has been studied, extraction system (U the extraction time is long (24 h), the pressure (20– 28 MPa) and temperature (40–45 °C) ranges are small. Conducting a successful supercritical fluid process requires knowledge of solute solubility. This information is essential prior to optimizing the extraction yield by
218
¨ zkal et al. / Journal of Food Engineering 69 (2005) 217–223 S.G. O
adjusting pressure, temperature, solvent flow rate and extraction time. Besides solubility the mass transfer behavior is needed in order to design or scale-up these processes. However, models which describe the solubility and extraction of nut oils, are scarce (Marrone et al., 1998). The objective of this study was to determine the solubility and mass transfer behavior of hazelnut oil in SC-CO2.
2. Materials and methods 2.1. Materials Hazelnuts (‘‘Tombul’’ variety), grown in the Hendek region of Turkey, were obtained from Fiskobirlik (Giresun, Turkey). They were unshelled and packaged in vacuum-sealed packages and stored at 5 °C and 50% relative humidity. Before the experiments, hazelnut samples were ground into small sizes by using a kitchen type grinder (Arc¸elik, Turkey). Hazelnut particles were sorted by certified test sieves (Endecotts Ltd., London, England). Sieving was performed by a shaker (Octagon 200, Endecotts Ltd., London, England). Particles between 1.0 and 2.0 mm were used in the experiments. Further grinding was avoided because it caused a formation of paste like structure, which inhibited the sieving process. The samples contained 56% oil and 3% moisture. CO2 was purchased from Habasß (Turkey). 2.2. Solubility measurements and extraction 2.2.1. Extraction system Solubility measurements and extractions were done using Supercritical Fluid Extraction System (SFX System 2120, Isco Inc., Lincorn, NE). The system consists of an extractor (SFX 220) and two syringe pumps (Model 100DX). The pumps could pump up to 69 MPa (accuracy: 2% of full scale, repeatability: 1% of full scale within 48 h) with flow rates ranging between 0.1 ll/min and 50 ml/min. The volumetric flow rate is measured as liquid CO2 at the extractor pressure and at 10 °C, which is the cooling temperature of the pump. The temperature in the extractor chamber could be controlled up to 150 °C. The extractor is a 10 ml steel cartridge where SC-CO2 flows downward. The extract is passed through a coaxially heated adjustable restrictor and the solute is precipitated in test tubes. 2.2.2. Solubility measurements The extraction system allows the use of dynamic technique (Maxwell, 1996). Solubility measurements were done at 15, 30, 45 and 60 MPa, and each at 40, 50 and 60 °C by using 5 g of hazelnut sample. Solvent
flow rate was kept as low as 0.5 ml/min. Theoretically, this corresponds to 0.52 ± 0.02 g/min average mass flow rate (3% standard error) in each experiment, since the density of liquid CO2 changes between 960 and 1100 kg/m3 within the pressure range at 10 °C. It was determined that the oil concentration in the solvent was independent of the flow rate around 0.5 ml/min, indicating that the full saturation was achieved. The amount of precipitated oil was determined gravimetrically at definite time intervals and the solubility was calculated as mg oil/g CO2 from the slopes of the linear part of each extraction curve. 2.2.3. Extraction In the extractions 5 g of hazelnut sample was extracted with SC-CO2 at 30, 45, 60 MPa and each at 40, 50 and 60 °C. SC-CO2 flow rate was kept constant at 2 ml/min. Theoretically, this corresponds to 2.13 ± 0.07 g/min average mass flow rate (2% standard error), since the density of liquid CO2 changes between 1030 and 1100 kg/m3 within the pressure range at 10 °C. The instrument is capable of measuring total volume of CO2 used from the beginning of an extraction. It was observed that the flow rate changed about ±5% during the extractions. The hazelnut oil was precipitated in test tubes containing glass wool. The yield was determined gravimetrically at definite time intervals. Each extraction was continued until no significant amount of oil was extracted (250–300 min). The standard error in the yield (g oil/g sample) was less than 3% for three replicate extractions. The fatty acid compositions of hazelnut oils extracted with SC-CO2 were compared to that extracted with hexane. Three different hazelnut oil fractions were obtained during extractions at 30 MPa, and at 40, 50 and 60 °C in order to determine any composition change in hazelnut oil during SC-CO2 extraction. The time intervals were chosen to be the initial 30 min, between 70 and 120 min, and after 120 min of the extraction. The extracted oil was precipitated in test tubes containing hexane for the GC analysis. 2.3. Analytical methods The moisture content of hazelnut was determined by using AOAC Method 926.12 (AOAC, 1995). Total fat determination of hazelnut and extraction of oil with hexane were done with soxhlet (Nas, Go¨kalp, & ¨ nsal, 1992), where 5.0 g of samples were extracted U using n-hexane. Extracted oils were esterified using boron thrifluoride solution in methanol (AOAC, 1995) and their fatty acid composition were determined using a gas chromatograph (GC-14A, Shimadzu, Kyoto, Japan) equipped with a 30 m fused capillary column (0.25 mm inner diameter and 0.20 lm film thickness, SP2330, Supelco, Belle-
¨ zkal et al. / Journal of Food Engineering 69 (2005) 217–223 S.G. O
fonte, PA) and a flame ionization detector. Oven temperature was maintained at 190 °C and samples were injected using a split method. The fatty acid analyses were performed in two replicates. 2.4. Solubility and mass transfer behavior 2.4.1. Solubility behavior It is a common practice to model the solubility of solutes in SC-CO2 as a function of solvent density. One of the equations, which are commonly used for correlating ¨ stu¨ndagˇ & Tethe solubility behavior of oils (Gu¨c¸lu¨-U melli, 2000; Sovova´, Zarevu´cka, Vacek, & Stra´nsky´, 2001), is proposed by Chrastil (1982). This semi-empirical equation is in the form, a 2 c ¼ qa1 exp þ a3 ð1Þ T where c is solubility in kg/m3, q is the density of CO2 in kg/m3 and T is temperature in K. The improved form of this equation is proposed by del Valle and Aguilera (1988), which is in the form, a3 a4 a1 c ¼ q exp a2 þ þ 2 ð2Þ T T The units are same as in the Chrastil equation except q is in kg/dm3. The Adachi–Lu equation (Adachi & Lu, 1983), which contains five constants, is given as, a5 2 c ¼ qða1 þa2 qþa3 q Þ exp a4 þ ð3Þ T The original equation is modified so that the parameters have the same units as in the del Valle–Aguilera equation. Eqs. (1)–(3) were used to describe the solubility behavior of hazelnut oil. 2.4.2. Mass transfer behavior Mass transfer behavior, during SC-CO2 extraction is determined by either theoretical (Marrone et al., 1998; Sovova´, 1994) or kinetic models (Andrich et al., 2001; Sankar & Manohar, 1994). The theoretical models include the analytical or numerical solutions of the governing mass transfer equations. The kinetic model (Andrich et al., 2001) used in this study describes the unsteady state mass transfer as, dme ¼ kðms;t ms;t Þ dt
ð4Þ
where, me is the amount of oil extracted in grams at time t; k is the mass transfer coefficient in s1; ms,t is the amount of unextracted oil at time t; ms;t is the amount of unextracted oil at time t if equilibrium between two phases has been reached. Assuming, ms;t is negligible, because pure solvent continuously fed to the extractor, and ms,t is equal to the difference between oil initially present in the sample (ms,0) and oil extracted at time t, than Eq. (4) becomes,
dme ¼ kms;t ¼ kðms;0 me Þ dt After integration, it gives, ms;0 ln ¼ kt ms;0 me
219
ð5Þ
ð6Þ
Slope of the straight line passing through the origin of the axes t and ln(ms,0/ms,0 me) gives k and the amount of oil extracted at time t is, ð7Þ me ¼ ms;0 1 ekt The extraction yield in g oil/g sample was calculated by dividing the amount of oil extracted (Eq. (7)) by the amount of sample. 3. Results and discussion 3.1. Solubility measurements The solubility of hazelnut oil in SC-CO2 at each condition (Table 1) was determined from the slopes of the linear parts of extraction curves which were drown as g oil extracted vs. g CO2 used (Fig. 1) at the beginning of each extraction. Fig. 1 shows the effect of flow rate on the amount of oil extracted. The solubility values calculated by using SC-CO2 flow rate of 2 ml/min were 10% lower than the ones calculated by using 4 times lower flow rate of 0.5 ml/min. Therefore, SC-CO2 flow rate of 0.5 ml/min was considered to be low enough to assure saturation during solubility measurements. The solubility values were constant for the entire extraction time. The solubility of an oil in SC-CO2 changes during extraction only if its composition changes upon fractionation. The solubility of grape seed oil was reported to decrease during SC-CO2 extractions performed at 40 °C and at pressures up to 29 MPa. This is due to the high percentage of free fatty acids, monoand diglycerides (Sovova´ et al., 2001) in grape seed oil, which are more soluble in SC-CO2 than triglycerides. Initially the solubility of grape seed oil in SC-CO2 was higher since these fractions were rich in free fatty acids. The solubility of triglycerides depends on the type of fatty acids attached to the glyceride chain. Palm kernel oil which is rich in lauric (C12:0, 48%) and myristic (C14:0, 15.6%) acids was fractionated during extraction performed at 34.5 MPa and 70 °C. The percentage of short chain and saturated fatty acid (C8:0, C10:0, C12:0) decreased with time, while the percentage of long chain and/or unsaturated fatty acids (C16:0, C18:0, C18:1, C18:2) increased with time (Hassan, Ab Rahman, Ibrahim, & Mohd. Omar, 2000). The composition of myristic acid (C14:0) remained unchanged. This fractionation was diminished at higher pressures. Fractions of hazelnut oil were collected at 30 MPa, to detect any fractionation which was possible only at
¨ zkal et al. / Journal of Food Engineering 69 (2005) 217–223 S.G. O
220
Table 1 Solubility and extraction yield of hazelnut oil CO2 density (kg/m3)
Extraction yield (g oil/g sample)
Oil recovery (% oil)
1.5 1.4 1.0
– – –
– – –
922 883 841
7.0 8.0 8.3
0.20 0.26 0.29
36 46 52
40 50 60
985 951 921
13.5 16.8 18.3
0.25 0.29 0.32
45 52 57
40 50 60
1032 1007 981
16.0 19.3 28.8
0.28 0.32 0.33
50 57 59
Pressure (MPa)
Temperature (°C)
15
40 50 60
787 705 603
30
40 50 60
45
60
Solubility (mg/g)
0.40
30
Solubility (mg/g)
0.35
Oil extracted (g)
0.30 0.25 0.20 0.15 0.10
20
10
0 0
0.05
20
40
60
Pressure (MPa) 0.00 0
10
20
30
40
50
60
CO used (g) 2
Fig. 1. Effect of flow rate on the amount of oil extracted at 30 MPa and 60 °C. (d) 0.5 ml/min, (n) 2.0 ml/min.
low pressures. Three different fractions obtained in the first 30 min, between 70 and 120 min, and after 120 min of each extraction. The fatty acid compositions of these fractions were not significantly different than the oil extracted with hexane (Table 2) indicating that hazelnut oil was not fractionated during extractions. The change in the solubility of hazelnut oil in SC-CO2 with temperature and pressure is shown in Fig. 2. It was observed that the solubility increased with pressure and
Fig. 2. Solubility of hazelnut oil in SC-CO2. (d) 40 °C, (j) 50 °C, (m) 60 °C, (––) Adachi–Lu equation.
temperature above 30 MPa. However, at 15 MPa the solubility showed a slight decrease with temperature (Table 1). This is consistent with the crossover phenomena generally observed for oils (King & Bott, 1995). The solubilities of oils in SC-CO2 increase both with the density of SC-CO2 and the volatility of fatty acids. The crossover phenomenon is due to the competing effects of reduction in density of SC-CO2 and increase in the fatty acids volatility, which accompany the temperature rise. Therefore, it could be concluded that the crossover pressure for hazelnut oil was in between 15 and 30 MPa (Fig. 1). This pressure is low, compared to 35 MPa for peanut
Table 2 Fatty acid composition of hazelnut oils extracted with SC-CO2 at 30 MPa and with hexane Temperature (°C)
40 50 60 Hexane
Fatty acids (%)a Palmitic (C16:0)
Palmitoleic (C16:1)
Stearic (C18:0)
Oleic (C18:1)
Linoleic (C18:2)
5.99 ± 0.11 5.86 ± 0.22 5.91 ± 0.25 5.56
0.15 ± 0.034 0.15 ± 0.035 0.05 ± 0.002 0.18
2.17 ± 0.07 2.14 ± 0.07 2.14 ± 0.07 2.22
79.34 ± 0.35 79.62 ± 0.30 79.38 ± 0.31 80.13
11.45 ± 0.26 11.37 ± 0.17 11.44 ± 0.18 11.08
a Each fatty acid composition reported were the average of compositions of the hazelnut oil fractions obtained between three different time intervals during SC-CO2 extraction (initial 30 min, between 70 and 120 min, and after 120 min).
¨ zkal et al. / Journal of Food Engineering 69 (2005) 217–223 S.G. O
3.2. Extraction The process parameters that affected the extraction of hazelnut oil were temperature and pressure of SC-CO2, and the extraction time. The moisture content and the particle size of the hazelnut samples were equivalently important. It was reported (Snyder, Friedrich, & Christianson, 1984) that nearly theoretical oil yields were obtained from ground or thinly flaked soybeans (
