Separation of d-limonene from supercritical CO 2 by means of membranes

J. of Supercritical Fluids 34 (2005) 143 147 Separation of d-limonene from supercritical CO 2 by means of membranes Luiz Henrique Castelan Carlson, Ariovaldo Bolzan, Ricardo Antônio Francisco Machado Universidade Federal de Santa Catarina, Depto. de Engenharia Química, Caixa Postal 476, 88040-900 Florianópolis, SC, Brazil Abstract The extraction of essential oil from vegetable matrices with supercritical fluids is a promising technology which is still searching for an economical method to recycle the dense carbon dioxide (CO 2 ). The use of the reverse osmosis membrane separation process can be an alternative to avoid the intense depressurization step which is necessary for the recovering of the extracts. The membranes exhibit satisfactory permeability for supercritical CO 2 filtration, with very good resistance under the severe experimental operating conditions. In this work, four commercial reverse osmosis and nanofiltration membranes were applied to perform the separation of limonene and supercritical carbon dioxide. The tests were conducted under a pressure of 12 MPa, a transmembrane pressure of 0.5 MPa and a temperature of 40 C. Pure CO 2 flux, limonene + CO 2 flux and the limonene retention factor were measured. The highest limonene retention factor was above 0.94. 2004 Published by Elsevier B.V. Keywords: d-limonene; Membrane separation; Reverse osmosis; Solvent recovering; Supercritical extraction. 1. Introduction The use of supercritical CO 2 for the extraction of natural products is still considered a new process on an industrial scale [1]. This process employs unproblematic, innocuous fluids as solvent and yields solvent-free products in a thermally gentle manner. Simple fraction of the products is possible by a variation of pressure and/or temperature. Nowadays, this process has become particularly familiar in the food, coffee and tobacco industries [2]. Processes involving extraction with supercritical fluids typically require a pressurization step, a heating or cooling step, an extraction step, and a subsequent separation and solvent regeneration step [3]. Although CO 2 is comparatively innocuous in small quantities and is relatively cheap, it cannot be discharged in unlimited quantities and it must be largely recovered in a well-designed process [4]. The costs of recompression of gaseous CO 2 to liquid or supercritical CO 2 is high, Corresponding author. Tel.: +55 48 33319554; fax: +55 48 3319687. E-mail address: carlson@enq.ufsc.br (L.H.C. Carlson). as a powerful compression equipment and often a refrigeration step prior to compression are required. The association of a membrane to the supercritical fluid extraction process could avoid the intense depressurization step, reducing the recompression costs [5]. Semenova et al. [6] studied the separation of supercritical CO 2 and ethanol mixtures with an asymmetric polyimide membrane, and a separation factor (α ethanol/co2 ) of 8.7 was obtained. The separation factor is defined by Eq. (1): α ethanol/co2 = y ethanol/y CO2 x ethanol /x CO2 (1) where y i and x i are the mole fraction of the component i (i = ethanol or CO 2 ) in the permeate and in the feed. For the separation of supercritical CO 2 and iso-octane mixtures, a separation factor (α iso-octane/co2 ) of 12.8 was obtained [7]. Sarrade et al. [8] characterized the behavior of organomineral nanofiltration membranes in supercritical CO 2. Permeability variations were investigated as a function of the temperature and pressure. The membranes exhibit satisfactory permeability for supercritical CO 2 filtration, with very good re- 0896-8446/$ see front matter 2004 Published by Elsevier B.V. doi:10.1016/j.supflu.2004.11.007

144 L.H.C. Carlson et al. / J. of Supercritical Fluids 34 (2005) 143 147 sistance under the severe experimental operating conditions. The membrane was applied to the fractionation of triglycerides in supercritical medium [9]. Sartorelli et al. [10] used two different inorganic membranes to the separation of low volatile compound from supercritical CO 2. Retention indexes between 80 and 90% were obtained. Spricigo et al. [5] applied a cellulose acetate reverse osmosis membrane to perform the separation of nutmeg essential oil and dense carbon dioxide. The effects of feed stream essential oil concentration, temperature and transmembrane pressure on essential oil retention and CO 2 permeability were investigated. The average essential oil retention was 96.4% and it was not affected significantly by any of the process variables. The membrane presented good CO 2 permeability and resisted well to the severe pressure conditions applied. The regeneration of supercritical CO 2 from mixtures containing caffeine by a commercial nanofiltration membrane, denoted as MD5, having a thin layer of ZrO 2 TiO 2 with an average pore size of 3 nm on the substrate carbon (Tech-Sep Co.) was studied by Tan et al. [11]. The experimental data showed that a caffeine rejection as high as 100% in the first 6 h and a supercritical carbon dioxide permeation flux of 0.024 mol m 2 s 1 could be obtained at 34.8 C and 7.95 MPa. d-limonene is a very effective, biodegradable solvent and degreaser, occurring in nature as the main component of citrus peel oil (as high as 95%). Due to its high solvency, attractive citrus odor, versatility, and GRAS rating ( Generally Recognized As Safe ) from the US FDA, d-limonene can be used safely and effectively in a wide range of products and applications. The objective of the present work was to study the separation of d-limonene in solution with supercritical CO 2 by means of different commercial reverse osmosis membranes. 2. Experimental 2.1. Equipment The experimental unit used in this work is represented schematically in Fig. 1. The essential oil was load in a stainless steel jacketed extractor (10) of 1000 cm 3 and internal diameter of 6 cm. Inside of the membrane unit (13), a membrane was placed over a sintered stainless steel plate disc support and sealing was made with poly(tetrafluoroethylene) rings. The membrane filtration area was of 139 cm 2 and crossflow regime was applied. The working pressure in the extractor (10) and in the membrane unit (13) was monitored by a pressure transducer (12) (Model RTP12/BE53R, AEP, Italy) and controlled through a pneumatic needle control valve (7) (Model 807, Badger Meter, USA). The pressure in the jacketed surge tank (6) was maintained close to the working pressure by a gas booster (3) (Model DLE 15-1, MAXPRO Technologies, Germany). The temperature of the system was kept constant by a thermostatic water bath (23) and (24). Carbon dioxide (White Martins, Brazil) of 99.95% purity was used. In the membrane permeate side the pressure was controlled by a backpressure regulator (20) (Model 26-1724-24, TESCOM, USA). The retentate CO 2 flow was controlled by a metering valve (16) (Swagelok Model SS-4MGD USA). Retentate essential oil samples were collected at a pressure of 2.0 MPa, maintained by a forward pressure regulator valve (14) (MTR Model 200-70 Brazil), in a jacketed separator (15) and the CO 2 flow was measured by a rotameter (18). Permeate essential oil samples were collected at atmospheric pressure in a separator (21) and the CO 2 flow was measured by a flow meter (22) (Model PV005LPM0CC, Key Instruments, USA). Fig. 1. Experimental unit: (1) CO 2 cylinder; (3) gas booster; (6) surge tank; (10) extractor; (13) membrane unit; (15, 21) separators; (5, 12, 19) pressure transducers; (7) control valve; (20) back pressure regulator; (14) forward pressure regulator; (22, 18) flow meter; (23, 24) thermostatic water baths; (26) computer; (2, 4, 8, 9, 16, 17, 25) metering valves; (11) on off ball valve.

L.H.C. Carlson et al. / J. of Supercritical Fluids 34 (2005) 143 147 145 The permeate flow (22), the system pressures (5, 12, and 19) and the control valve (7) were monitored and controlled by a computer software developed by the authors. 2.2. Membrane and d-limonene One nanofiltration membrane (Model Sepa CF Thin Film HL, Osmonics, USA) and three different reverse osmosis membranes (Models Sepa CF: thin film SG; polyamide AK; cellulose acetate CE; Osmonics, USA) were used in this work. Those membranes were designed for around 97% rejection of NaCl. Natural d-limonene (purity 98%) was supplied by Duas Rodas Industrial (Brazil). It is a terpene hydrocarbon with a molecular weight of 136.23 g mol 1.In this study, d-limonene was used because of its little size, weight and easy extraction with supercritical CO 2, been a good method to test the membranes retention factor. 2.3. Procedures To perform the tests of pure CO 2 flux through the membranes, the valves (8) and (11) were closed to isolate the extractor (10) from the system. Valve (9) was slowly opened to promote a slow pressurization rate of 0.1 MPa min 1 on the retentate side, not to allow a transmembrane pressure higher than 0.5 MPa. A working pressure of 12 MPa, regulated by a pneumatic control valve (7), a transmembrane pressure of 0.5 MPa and a temperature of 40 C were used for all the experiments. The pressure of the permeate side was regulated to 11.5 MPa with back pressure regulator (20), and the corresponding CO 2 fluxes were measured. The retention tests were performed by loading the extractor (10) with 200 g of d-limonene. The valve (9) was closed and the valve (8) was opened to pressurized the extractor to 12 MPa, then the valve (11) was opened to allow the CO 2 flow through the extractor and reach the membrane. The cross-flow was applied by regulating a CO 2 flow of 1 L min 1 (STP) with the valve (16) at the retentate side. The d-limonene retentate and permeate fluxes were measured at time intervals by evaluation of the mass deposited in the separators (15) and (21), respectively. The separator (15) operated with a pressure of 5 MPa and a temperature of 36 C. Real retention factor ( ) calculations were made as indicated in the Eq. (2): = 1 ( Cp C r ) where C p is the permeate d-limonene concentration and C r is the retentate d-limonene concentration, an experimental error of 5% was founded. The working pressure and temperature were chosen based on a previous work by Spricigo et al. [5]. 3. Results and discussion Fig. 2 presents the permeate flux of pure supercritical CO 2 through the membranes at an upstream pressure of 12 MPa, a transmembrane pressure of 0.5 MPa and a temperature of 40 C. The CO 2 flux value has not changed with time and there were no need to wait a long time to reaches a stationary value as expected [5]. This may be due to the fact the pressurization was done with an existing transmembrane pressure, promoting the membrane compaction from the beginning of the pressurization. The HL membrane gives the highest CO 2 flux value (31.3 kg h 1 m 2 ), followed by the SG (12.2 kg h 1 m 2 ) and AK (10.56 kg h 1 m 2 ) membrane. The HL membrane, which is a nanofiltration membrane, was already expected to present the highest CO 2 flux. Those values are in accordance with the ones obtained by Spricigo et al. [5]. The CE membrane showed a near zero CO 2 flux and could not be measured. This may be the result of a membrane compaction while pressurizing. Thus, no retention experiments were done with the CE membrane. (2) Fig. 2. Pure CO 2 flux with a transmembrane pressure of 0.5 MPa (working pressure = 12 MPa and temperature = 40 C) for the HL, CE, AK and SG membranes.

146 L.H.C. Carlson et al. / J. of Supercritical Fluids 34 (2005) 143 147 Fig. 3. d-limonene real retention factor for the three different membranes tested with a transmembrane pressure of 0.5 MPa (working pressure = 12 MPa and temperature = 40 C). Retention experiments were performed by maintaining a constant retentate flow of 1 L min 1 (STP) with an upstream pressure of 12 MPa, a transmembrane pressure of 0.5 MPa and a temperature of 40 C. The maximum d-limonene feed stream concentration was 30% and slowly dropped with the extraction time. Fig. 3 shows the limonene retention factors obtained for the tested membranes. The SG membrane presented the highest limonene retention factor (over 0.8). The AK membrane retention factor dropped to zero after 80 min and the HL membrane reached a stabilized value of 0.3 after 75 min. The permeate CO 2 flux (Fig. 4) decreased for all the tested membranes during the retention experiments. The SG membrane CO 2 flux fell to zero after 120 min, demonstrating that a high retention factor was followed by an irreversible clogging of the membrane and the occurrence concentration polarization phenomenon. Even after a cleaning step with pure supercritical CO 2 in the same operational conditions the CO 2 flux remained at a zero value. The AK and HL membrane CO 2 flux fell until 40 min and gently increased again to values near to that obtained with pure CO 2. After a cleaning step the AK and HL membranes have the initial CO 2 flux reestablished. A second test was performed with the SG membrane expecting a high limonene retention factor avoiding the membrane clogging. Attempt was made not allowing the permeate CO 2 flux to decrease too much and periods of a CO 2 + limonene and pure CO 2 flow were alternated to accomplish this. Fig. 5 presents the permeate CO 2 flux through the SG membrane and the limonene retention factor for the second test, where the straight lines divides the periods of time where CO 2 + limonene and pure CO 2 flow were alternated. A high retention factor (over 0.94) could be achieved, the membrane clogging could be avoided and the CO 2 flux fell to 1.9 kg h 1 m 2 and have not changed after a cleaning step. Fig. 4. Permeate CO 2 flux for limonene retention tests with a transmembrane pressure of 0.5 MPa (working pressure = 12 MPa and temperature = 40 C).

L.H.C. Carlson et al. / J. of Supercritical Fluids 34 (2005) 143 147 147 Fig. 5. Time-dependence of CO 2 flux and limonene retention factor with a transmembrane pressure of 0.5 MPa (working pressure = 12 MPa and temperature = 40 C) for the SG membrane. 4. Conclusions The SG membrane proved to be the best choice for the separation of supercritical CO 2 and limonene mixture. A limonene retention factor as high as 0.94 could be achieved for a high feed concentration of limonene (30%), but it presented low fluxes of CO 2. More experiments should be done with a lower limonene concentration and a higher retentate CO 2 flux to promote a higher permeate CO 2 flux and avoid the concentration polarization phenomenon. The CE membrane could not be tested because it presented almost no pure CO 2 flux. The AK membrane showed no limonene retention and the HL presented a low limonene retention factor of 0.3 and a good permeate CO 2 flux. The tested membranes presented, with the exception of the CE membrane, a good chemical resistance to the limonene and a good mechanical resistance to the high-pressure environment. The results obtained with the SG membrane showed that the coupling of this membrane with the supercritical CO 2 extraction can be a potential application to lower the energy requirement to the solvent recycling. With this membrane, almost 70% of the solvent (permeate CO 2 ) could be recycled with only 0.5 MPa of pressurization (from 11.5 to 12.0 MPa) while the other 30% of the solvent (retentate CO 2 ) should be pressurized from 5 to 12 MPa, where phase change will occur and more energy is spent to reach the extraction pressure. Acknowledgements The authors wish to thank to CNPq and UFSC (Brazil) for their technical and financial support to this work. References [1] L.H.C. Carlson, R.A.F. Machado, C.B. Spricigo, L.K. Pereira, A. Bolzan, Extraction of lemongrass essential oil with dense carbon dioxide, J. Supercrit. Fluids 21 (2001) 33. [2] U. Sievers, R. Eggers, Heat recovery in supercritical extraction process with separation at subcritical pressure, Chem. Eng. Process. 35 (1996) 239. [3] R.L. Smith, H. Inomata, M. Kanno, K. Arai, Energy analysis of supercritical carbon dioxide extraction processes, J. Supercrit. Fluids 15 (1999) 145. [4] E. Lack, H. Seidlitz, Commercial scale decaffeination of coffee and tea using supercritical CO 2, in: M.B. King, T.R. Bott (Eds.), Extraction of Natural Products Using Near-Critical Solvents, Chapman & Hall, Bishopbriggs/Glasgow, 1993, p. 101. [5] C.B. Spricigo, L.H.C. Carlson, A. Bolzan, R.A.F. Machado, J.C.C. Petrus, Separation of nutmeg essential oil and dense CO 2 with a cellulose acetate reverse osmosis membrane, J. Membr. Sci. 188 (2001) 173. [6] S.I. Semenova, H. Ohya, T. Higashijima, Y. Negishi, Separation of supercritical CO 2 and ethanol mixtures with an asymmetric polyimide membrane, J. Membr. Sci. 74 (1992) 131. [7] H. Ohya, T. Higashijima, Y. Tsuchiya, H. Tokunaga, Y. Negishi, Separation of supercritical CO 2 and iso-octane mixtures with an asymmetric polyimide membrane, J. Membr. Sci. 84 (1993) 185. [8] S. Sarrade, G.M. Rios, M. Carlés, Nanofiltration membrane behaviour in a supercritical medium, J. Membr. Sci. 114 (1996) 81. [9] S. Sarrade, G.M. Rios, M. Carlés, Supercritical CO 2 extraction coupled with nanofiltration separation. Application to natural products, Sep. Purif. Technol. 14 (1998) 19. [10] L. Sartorelli, G. Brunner, Membrane separation of extracts from supercritical carbon dioxide, in: Proceedings of the 5th International Symposium on Supercritical Fluids, Atlanta, USA, 2000. [11] C. Tan, Y. Chiu, Regeneration of supercritical carbon dioxide by membrane at near critical conditions, J. Supercrit. Fluids 21 (2001) 81.