EXTRACTION OF STEVIA GLYCOSIDES WITH CO2 + WATER, CO2 + ETHANOL,
AND CO2 + WATER + ETHANO

Abstract - Stevia leaves are an important source of natural sugar substitute. There are some restrictions on the use of stevia extract because of its distinctive aftertaste. Some authors attribute this to soluble material other than the stevia glycosides, even though it is well known that stevia glycosides have to some extent a bitter taste. Therefore, the purpose of this work was to develop a process to obtain stevia extract of a better quality. The proposed process includes two steps: i) Pretreatment of the leaves by SCFE; ii) Extraction of the stevia glycosides by SCFE using CO2 as solvent and water and/or ethanol as cosolvent.
The mean total yield for SCFE pretreatment was 3.0%. The yields for SCFE with cosolvent of stevia glycosides were below 0.50%, except at 120 bar, 16°C, and 9.5% (molar) of water. Under this condition, total yield was 3.4%. The quality of the glycosidic fraction with respect to its capacity as sweetener was better for the SCFE extract as compared to extract obtained by the conventional process. The overall extraction curves were well described by the Lack extended model.






  
INTRODUCTION


In 1900, the Paraguayan chemist Ovidio Rebaudi, after whom Bertoni named the plant, studied the major characteristics of stevia. He succeeded in isolating two types of substances: one extremely sweet and the other bitter, resembling a digestive appetizer. Of the two, it was the sweetening principle that attracted more attention at that time, as is still true today. The Stevia rebaudiana Bertoni contains a complex mixture of labdane diterpenes, triterpenes, stigmasterol, tannins, volatile oils, and eight diterpenenic glycosides: stevioside, steviobioside, dulcoside, and rebaudiosides A, B, C, D, and E. The most abundant substances are stevioside and rebaudioside A. Of the stevia glycosides rebaudioside A is the sweetest and the most stable, and it is less bitter than stevioside. Rebaudioside E is as sweet as stevioside, and rebaudioside D is as sweet as rebaudioside A, while the other glycosides are less sweet than stevioside (Cramer and Ikan, 1987).


A combined process involving a solid/liquid extraction step, followed by a liquid/liquid-purifying step, is traditionally used to extract the glycosides from stevia. However, the glycosidic extract has a pronounced bitter aftertaste that is responsible for many of the restrictions on the use of stevia as a sweetener. There are several hypotheses in regard to the source of the bitter aftertaste of stevia glycosides. Phillips (1987) described a European patent held by the Stevia Company, which attributes the bitter aftertaste to the presence of essential oils, tannins, and flavonoids. Soejarto et al. (1983) believed that the sesquiterpene lactones are responsible for the bitter aftertaste. Tsanava et al. (1991) suggested that caryophyllene and spathulenol contribute decisively to the aftertaste. Nevertheless, as pointed out by Phillips (1987), stevioside and rebaudioside A are partially responsible for the aftertaste, even though the contribution of rebaudioside A is significantly less than that of stevioside.


Tan et al. (1988) hold a Japanese patent for the production of stevia glycosides by supercritical fluid extraction (SCFE) with CO2 and a cosolvent. Methanol, ethanol, and acetone were used as cosolvents. The purification step is accomplished by adsorption. Kienle (1992) holds a similar patent in the USA. Pasquel et al. (1999) studied the SCFE of the nonglycoside fraction of stevia leaves. The following substances were identified in the extracts: spathulenol; decanoic acid; 8, 11, 14-ecosatrienoic acid; 2-methyl octadecane; pentacosane; octacosane; stigmasterol; b -sitosterol, a - and b -amyrine; lupeol; b -amyrin acetate; and pentacyclic triterpene. These substances represent 56% of the total extracts; therefore, 44% of the substances present in the extract still need to be identified.


All the conventional extraction processes described in the literature follow a similar methodology (Phillips, 1987). The stevia leaves are extracted with hot water or alcohols. In some cases, the leaves are pretreated with nonpolar solvents such as chloroform or hexane to remove the essential oils, lipids, chlorophyll, and other nonpolar substances. The extract is clarified by precipitation with salt or alkaline solutions. The extract is concentrated and redissolved in methanol for crystallization of the glycosides. The crystals are formed almost by pure stevioside.


Using the information just discussed, the objectives of the present work were to produce a stevia sweetener employing a two-step process: i) pretreatment of stevia leaves by SCFE with COc and ii) extraction of the stevia glycosides by SCFE using the following mixtures: CO2 + water, CO2 + ethanol, and CO2 + water + ethanol. Using information from Pasquel et al. (1999) pretreatment conditions were set at 200 bar and 30°C. The glycosides were obtained at 120 and 200 bar at 16, 30, and 45°C. The composition of the SCFE glycosidic extract was compared to the composition of stevia extract obtained by conventional low-pressure extraction.


 


MATERIAL AND METHODS


The Raw Material

Stevia leaves from the crop of 1995 were bought in Maringá (Paraná, Brazil). The solid material was cleaned, selected, packed in plastic bags, and stored at room temperature (20 to 32°C). The humidity of the raw material was determined using the toluene distillation method (Jacobs, 1973). The glycoside content was determined according to the phenol sulfur method for total carbohydrates (Alvarez et al., 1986).


Particles and Bed Characterization

The real density of the stevia particles was determined by picnometry with gas helium (Multivolume Picnometer 1305) at the Central Analítica, IQ – Unicamp. Apparent density was calculated from the mass used to fill the extraction cell. Bed porosity was defined using the real density of the particles and the apparent density of the bed. The mean diameter of the particles was evaluated using the methodology described by Corrêa (1994).


The Experimental Unit For the SCFE

The experimental unit used was that described by Pasquel et al. (1999) for the pretreatment of stevia leaves. A cosolvent pump was added to the system (Figure 1)




  

Experimental Procedure: SCFE

The mass of solid used varied from 69.10-3 to 82.10-3 kg. The triturated solid was packed inside the extraction cell (SS 316, with a length of 0.375 m and an inside diameter of 0.0283 m). The extraction cell was adapted to the SCFE unit and the heating and/or cooling system was turned on. Once the system reached a temperature of 30°C (approximately 3 hours), valves 2a, 2b, 2c, and 2h were opened. As soon as the system pressure reached 200 bar, valves 2j, 2m, and micrometering valve 15 were opened. The extracts were collected in 20mL glass flasks. An adsorption column containing Porapak Q (80 /100 mesh, Waters Associates Inc., USA) was adapted to prevent losses of volatile substances in the pretreatment step at the solvent outlet. The solvent flow rate was continuously monitored. Samples of the extract were collected every hour. Pretreatment was carried out at 200 bar, 30°C, and an average solvent flow rate of 4.82.10-5 kg/s for a period of 12 hours. The extraction cell containing the pretreated stevia leaves was stored in a domestic refrigerator.


For extraction of the glycosides, the extraction cell was readapted in the SCFE unit. The experimental procedure was similar to the one described above. Samples of the extract were collected every 30 minutes and the total extraction time was 12 hours. The experimental runs were conducted at 120 and 200 bar at 16, 30, and 45°C. The cosolvents used were 9.5% (molar) water, ethanol, or an equimolar mixture of water and ethanol. Because the experiments were very long (12 hours for the pretreatment, 12 hours for the glycoside extraction plus setup time), the experimental plan was a fractional factorial design and only one-third of the total was selected.


Experimental Procedure: Conventional Extraction

Stevia leaves subjected to the SCFE pretreatment and stevia leaves with no pretreatment were used. The method described by Alvarez and Couto (1984) and Goto (1997) was used. One liter of boiling water was added to fifty grams of stevia leaves. The infusion was kept at room temperature (25 to 30°C) for one hour. The aqueous extract was vacuum filtered. In a separation funnel the aqueous extract was mixed with isobutyl alcohol (Merck P.A., 99.99%) maintaining the 40:60 (v/v) proportion. The system was allowed to rest until complete phase separation was achieved. The butanolic extract was centrifuged at 3500 rpm (Solvall, RT 600D) for 15 minutes. The extract was heated up to 80°C, and percolated through a bed of activated carbon (1 g of activated carbon for every 100 mL of extract). The extract was concentrated in a rota-evaporator (Tecnal, TE 120) and allowed to rest for 24 hours to achieve crystallization of the glycosides. The crystals were washed with methanol (Merck P.A. 99.9%) and dried in an air-circulating oven. The crystallization mother liquor was concentrated and extracted with acetone (Merck P.A., 99.8%). The crystals were washed with anhydrous acetone and dried in an air-circulating oven.


Analysis of the Glycosidic Extracts

(a) Identification of the Glycosides

The preliminary identification of the stevioside in the extracts was made by thin-layer chromatography (TLC) using silica gel plates (Merck, lot PF254336) and the extracts were eluted with chloroform:methanol:water (30:20:4). The spots were developed by spraying with methanol:sulfuric acid (1:1) and heating to 110°C. The extracts were diluted with ethanol and hexane. The standard stevioside (95%, Steviafarma Industrial, Maringá, Paraná, Brazil) was diluted in the elution solvent. The chloroform, methanol, sulfuric acid, ethanol, and hexane were from Merck and were of chromatographic grade.
 

Identification of the glycosides was accomplished by HPLC (CG Instrumentos Científicos, Model CG-480C) with a UV detector (Jasco, model 970 UV) at a wavelength of 210 nm, using a NH2 Licrosorb column (5 m m, 220 x 4.6 mm, Technology –Techsphere). Acetonitrile and methanol (85:15) (Merck, HPLC grade) were used as mobile phase at a flow rate of 1.5 mL/min. Stevioside and rebaudioside A (95% and 85%, respectively, from Steviafarma Industrial, Maringá, Paraná, Brazil) were used as standard. The analyses were performed in the Chemical Engineering Department, UEM, Paraná.
 

(b) Quantification of the Glycosides

The aqueous extracts were concentrated and redissolved in water at known concentrations. The calibration curve was made using a commercial sample of stevioside (95% from Steviafarma Industrial, Maringá, Paraná, Brazil) diluted in water. Quantification of the glycosides was done using a spectrophotometer (Hitachi, model U-2000, double bean) at a wavelength of 210 nm (Nikolova-Damyanova et al., 1994).
 

Calculation of the Kinetic Parameters


The experimental data on mass of extract as a function of time of extraction were used to calculate the kinetic parameters for the constant extraction rate period (CER): the mass transfer rate, MCER; the duration of the CER period, tCER; the mass ratio of solute in the fluid phase at the bed outlet, YCER, and the yield (mass of extract/mass of feed), RCER. The methodology of Rodrigues et al. (2000) was used.
 
 

RESULTS AND DISCUSSION

 
The humidity of the stevia leaves was 7.0± 0.1%. The real density of the leaves was 1370± 30 kg/m3 and the bed apparent density was 400± 20 kg/m3. The average particle diameter was (0.98± 0.03).10-3 m. The mean bed porosity was 0.77± 0.01. The glycoside content of the leaves was determined as 5.0± 0.1% and was small compared to values found in the literature which vary from 7.8 to 14.5% (Phillips, 1987). This can be explained by the fact that besides the leaves the raw material also contained some stevia flowers.

 

SCFE Pre-Treatment of the Stevia Leaves

 
Figure 2 shows the overall extraction curves for the pretreatment step. Table 1 presents the kinetic parameters for the CER period. The average rate of mass transfer was MCER = (7.9± 0.7).10-8 kg/s, the average mass ratio in the fluid phase was YCER =(1.6± 0.2).10-3 kg-extract/kg-CO2, and the average CER and total yields were RCER = 1.7± 0.8% and RTOTAL = 3.0± 0.1%. About 63% of the total yield was obtained during the CER period.

 



 

 


The results show that kinetic parameter tCER has the largest dispersion while MCER and YCER have the smallest. These results corroborate the special attention that should be paid to the standardization of experimental techniques related to measurement of kinetic and equilibrium parameters for SCFE from a solid substratum. Not only does the feed material need to be kept under very strict control, but also, the experimental procedure has to be accurately specified if data from different sources are to be compared.

 
It is well known that a typical overall extraction curve for SCFE from a solid substratum has three distinctive steps (França et al., 1999): i) the constant extraction rate period (CER), characterized by the predominance of convective effects; ii) the falling or decreasing rate period (FER), for which both convective as well as diffusional effects are important; and iii) the diffusion-controlled rate period that is characterized by the diffusion of solvent and the solute/solvent mixture in the solid substratum. In Figure 2, these three periods can be easily identified; nevertheless, the asymptotic behavior expected for the third step was not achieved. This can be explained in terms of the composition of the extract obtained in the pretreatment (Pasquel et al., 1999).

 

SCFE of the Lycosides From Stevia Leaves
 

Ethanol, water, and an equimolar mixture of these two solvents were chosen as cosolvents. Ethanol was selected based on legal restrictions on the residual amount of organic solvents present in inputs for the food industry and also on information given by Kienle (1992) that attributes an increase in extraction yield to alcohol of up to 4 carbons used as cosolvents. Water was chosen due to its capability of solubilizing rebaudioside A, the sweetest stevia glycoside with the least aftertaste. Table 2 shows the kinetic parameters for SCFE of stevia glycosides. Because the fractional factorial design had three levels of temperature and cosolvent, and two levels of pressure, only six experiments were done; thus, no analysis of variance was performed. Figure 3 shows that the overall extraction curves for the SCFE of glycosides have the expected behavior. The yields were in general very low, except for the assay using water as cosolvent carried out at 120 bar and 16°C (Figure 4).




 

 



The results indicated a possible combined effect of temperature, pressure, and cosolvent. In fact, the solvation power of a solvent or a solvent mixture is directly related to mixture polarity as well as to intermolecular interactions of the following kinds: solvent/cosolvent, solvent/solute, and cosolvent/solute (McHugh and Krukonis, 1994). Water is highly polar compared to ethanol and has a larger dipole moment; therefore, its molecules should appreciably increase the CO2 polarity, resulting in an important increase in glycosides solubility. In addition, the increase in temperature negatively affects the attractive forces such as dipole-dipole between water and glycoside molecules (McHugh and Krukonis, 1994). This explains the behavior of the system at 120 bar and 16°C: The rate of mass transfer, MCER, was up to 35 times higher than that for the other experiments. A similar behavior was observed for YCER. Nevertheless, no pattern could be drawn for tCER. The yield (0.12%) for the assay at 16oC using ethanol as cosolvent and carried out at 65 bar was smaller than that for the other two experiments using ethanol as cosolvent, indicating the combined effect of temperature and pressure. Comparison of the yields for the experiments using water as cosolvent showed that regardless of whether it was used with or without ethanol, water definitively increased the solubility of the glycosides.


Yields for the conventional process were 2.60± 0.01% for pre-treated leaves and 2.10± 0.01% for stevia leaves with no pretreatment. These yields were appreciably smaller than those obtained for SCFE using water as cosolvent. The behavior of the pretreated leaves was as expected, since during pretreatment several substances were removed and therefore did not get the chance to compete with the glycosides for the solvent during glycoside extraction.


The Composition of the Extracts

 
The thin-layer chromatography showed that the pretreatment extracts had no glycosides. The SCFE glycosidic extracts were quantified by UV spectrophotometry as stevioside. Nevertheless, the HPLC analysis showed that the extracts were indeed a mixture of stevioside and rebaudioside A.

Comparison of the chromatogram for the SCFE glycosidic extract using water as cosolvent with the glycosidic extract obtained by conventional extraction from pretreated stevia leaves (200 bar, 30°C) clearly showed the larger amount of rebaudioside A obtained by SCFE. For the assay at 120 bar, 16°C and 9.5% (molar) of water, the relationship between stevioside/rebaudioside A was approximately 3 to 1 (Table 3). This result corroborates the fact that the CO2+water mixture is capable of extracting larger amounts of rebaudioside A than the conventional process using water and organic solvents. Considering that rebaudioside A is about 50% sweeter than stevioside and has less aftertaste, the glycosidic extract obtained is a much better sweetener than the stevia extract produced by the conventional process.





  
Modeling the Mass Transfer for the Pretreatment Step
 

Among the various models presented in the literature to describe SCFE from a solid substratum, the extended Lack model developed by Sovová (1994) was chosen. The model was selected based on the shape of the overall extraction curves that shows a predominance of the constant extraction rate period. As discussed before, during the CER period as much as 63% of the extract was obtained (Table 1). The model developed by Sovová (1994) assumes that the solvent flows axially at a constant superficial velocity through a fixed bed of cylindrical shape. The solvent is pure at the entrance of the extractor; temperature and pressure are kept constant throughout the fixed bed. In addition, the bed is homogeneous in terms of solute distribution as well as particle size. All these conditions were fulfilled by the SCFE from the bed of stevia. Therefore, the mass balance for a bed element is given by

Solid Phase:
 
                    (1)

Fluid phase:

                             (2)

where X and Y are the solute mass ratio for the solid and fluid phases, respectively; t is the time (s); e is the bed porosity; U is the superficial solvent velocity (m/s); rs and rf are the solid and fluid phase densities respectively, and J(X,Y) is the rate of interfacial mass transfer. Usually, for the SCFE process the fluid phase is a diluted solution; therefore, the solvent density, , can adequately be used as rf.

The solution to the system of Error! Reference source not found. and (2 given by Sovová (1994) is described by the following system of equations:
 

For the CER period, t < tCER

             (3)


For the falling extraction rate period (FER)

                 (4)

For the diffusion-controlled rate period, t ³ tFER

                     (5)


with the following restrictions:

                     (6)


                    (7)



                (8)


                    (9)


                    (10)



where mext is the mass of extract (kg), N is the mass of inert solid (kg), Y* is the operational solubility (kg/kg), kYa is the fluid-phase mass transfer coefficient (s-1), kXa is the solid-phase mass transfer coefficient (s-1), tFER is the duration of the falling extraction rate period (s), X0 is the initial solute mass ratio in the solid phase, and Xk is the solute mass ratio for the unruptured cells in the solid phase.

To use the model it is necessary to know the model parameters. Several authors have obtained the parameters by directly fitting the overall extraction curves to the model equations. This procedure is very helpful if experimental data, including operational solubility data, are available. A database for SCFE from a solid substratum is still being built; therefore, the required data can be found in the literature for only a few systems. We will now discuss a slightly different approach to evaluating model parameters, based on the theory of similarity used over the years in other areas of process engineering. The methodology used data from the literature and just one overall extraction curve to evaluate the parameters required to test the model proposed by Sovová (1994). kYa and Y* were calculated using the kinetic and operational solubility data measured by Monteiro (1999) for the ginger/CO2 system in the same equipment used to obtain the data for the present work. The model parameters were evaluated under the assumption that there was a close similarity between the system of interest and a reference system for which experimental data were available. Therefore, the systems should be similar hydrodynamically as well as thermodynamically. The hydrodynamic similarity was guaranteed since both systems were studied in the same equipment. The thermodynamic similarity was a bit more difficult to achieve. Nevertheless, the amount of CO2 soluble material in ginger (3%) is approximately equal to the CO2 soluble material in stevia leaves. Thus, the hypothesis of thermodynamic similarity was assumed to be valid. Therefore, the fluid-phase mass transfer coefficient for the stevia/CO2 system was assumed to be equal to that of the ginger/CO2 system. The ginger/CO2 fluid-phase mass transfer coefficient was calculated using the following definition:

                 (11)

where S is the cross-sectional bed area (m2), H the bed length (m),  the solvent density (kg/m3), and D Y the average logarithmic mean for the solute mass ratio in the fluid phase (Ferreira et al., 1999):

                 (12)

Yin and Yout are the solute mass ratio at the bed entrance and exit, respectively. The data of Monteiro (1999) measured at 200 bar and 30°C were used. For the kinetic experiments, the solvent flow rate ranged from 4.82.10-5 to 4.98.10-5 kg/s. while for the operational equilibrium experiments the solvent flow rate was 1.67.10-5 kg/s. Bed porosity varied from 0.76 to 0.74 and mean particle diameter from 0.98.10-3 to 1.02.10-3 m. The ginger/CO2 mass transfer coefficient was calculated as KYa 3.019.10-4 s-1. Using the mean value for MCER (Table 1), the operational solubility of the stevia extract in CO2 was estimated as Y* = 2.27.10-3 kg/kg.

 The solid-phase mass transfer coefficient was evaluated using the following definition:

                 (13)

where

                 (14)

X* is the mass ratio of solute to the inert solid in equilibrium with Y*. XP is the mass ratio of easily accessible solute in the solid phase, and Xk is the mass ratio of solute inside the cells in the solid phase, as defined by Sovová (1994). The initial solute content was estimated from the highest yield for the pretreatment. Table 4 presents the values of the model parameters used to simulate the overall extraction curve shown in Figure 5. The model systematically overestimates the overall extraction curves for the CER period. In spite of that, the overall yield of the process is only marginally different from the experimental value. Therefore, the procedure described here can be used to get first-hand information for projects of SCFE systems. Pasquel (1999) estimated the solubility of the stevia extract in CO2, using the Sherwood number calculated by Monteiro (1999). Using the fluid-phase mass transfer coefficient of Monteiro (1999) he fitted the overall extraction curve to the model discussed by Sovová (1994). His results were better those the ones presented here mainly due to the fact that his solid-phase mass transfer was obtained by fitting the experimental data to the model equations.

 




 

  
CONCLUSIONS

 
The mean total yield for SCFE pretreatment of stevia leaves at 200 bar and 30°C was 3.0% (m/m). About 63% of this total was obtained during the CER period. Conversely, yields for SCFE with cosolvent of stevia glycosides were below 0.50%, except at 120 bar, 16°C, and 9.5% (molar) water. For this condition, the total yield was 3.4± 0.3% and as much as 70% of the total glycosidic fraction was obtained during the CER period. The yields for the conventional process were approximately equal regardless of whether untreated or pretreated leaves were used. The quality of the glycosidic fraction with respect to its capacity as a sweetener was better for the SCFE in terms of the relative amount of stevioside and rebaudioside A. The overall extraction curves were well described by the Lack extended model using operational solubility and the mass transfer coefficient estimated with data from the literature.

 

ACKNOWLEDGEMENTS

 
This work was financed by FAPESP under grant 95/05262-3. A. Pasquel thanks FAPESP for the Ph. D. assistantship (95/03390-4). The authors also thank Silvania R. M. Moreschi and DEQ/UEM for their assistance in the glycoside analysis and Steviafarma Industrial for the stevioside and rebaudioside standards.

  

REFERENCES

 
Alvarez, M.; Couto, A.C.C. Processo de fracionamento dos componentes das folhas da Stevia rebaudiana (Bert.) Bertoni (Fundação Universidade Estadual de Maringá), Patent, Brazil, protocol No. 84 02 752 (1984).

Alvarez, M.; Kusumoto, I.T.; Arraes, L.S. Correlação entre o teor de carboidratos totais e o teor de glicosídeos diterpênicos nas folhas secas de Stevia rebaudiana (Bert.) Bertoni, Arquivos Biológicos Tecnológicos, 5, 19 (1986).

Corrêa, N.C.F. Estudo da cinética de extração do óleo da semente de maracujá com CO2 supercítico. M.S. thesis, Universidade Federal do Pará, Belém, PA, Brazil (1994).

Cramer, B.; Ikan, R. Progress in the chemistry and properties of rebaudiosides. In: Grenby, T. H. (Ed.), Developments in Sweeteners 3, Elsevier, New York, p.45 (1987).

Ferreira, S.R.S.; Nikolov, Z.; Doraiswamy, L.K.; Meireles, M.A.A.; Petenate, A.J. SCFE of Black Pepper (Piper nigrun) Essential Oil. J. Supercritical Fluids 14, No. 3, 235 (1999).

França, L.F.; Reber G., Meireles M.A.A.; Machado, N.T.; Brunner, G. Supercritical Extraction of Carotenoids and Lipids from Buriti (Mauritia flexuosa), A Fruit From the Amazon Region. J. Supercritical Fluids, 14, No. 3, 247 (1999).

Goto, A. Estudo da influência do rebaudiosídeo A na solubilidade do esteviosídeo e análise sensorial dos produtos obtidos, M.S. thesis, Universidade Estadual de Maringá, Maringá, PR, Brazil (1997).

Jacobs, M.B. The chemical analysis of foods and food products, 3rd Ed., Robert Krieger Publishing Co, New York (1973).

Kienle, U. Method of making a natural sweetener based on Stevia rebaudiana, and use thereof. Int. Cl.5 A61K 35/78. U.S. Cl. 424/195.1, 586690 Sep. 24, 1990; May 12, 1992. Chemical abstracts No. 111, 22442z.

McHugh, M.; Krukonis, V. Supercritical fluid extraction, 2d Ed., Butterworth-Heinemann, Boston (1994).

Monteiro, A.R. Extração do óleo essencial/oleoresina de gengibre (Zingiber officinale Roscoe) com CO2 supercrítico: uma avaliação do pré-tratamento e das variáveis de processo, Ph.D. diss., Universidade Estadual de Campinas, Campinas, SP, Brazil (1999).

Nikolova-Damyanova, B.; Bankova, V.; Popov, S. Separation and quantification of stevioside and rebaudioside A in plant extracts by normal-phase high performance liquid chromatography and thin-layer chromatography: A comparison, Phytochemical Analysis 5, 81 (1994).

Pasquel, A. Extração dos Glicosídeos da Estévia com Fluidos Pressurizados, Ph.D. diss., Universidade Estadual de Campinas, Campinas, SP, Brazil (1999).

Pasquel, A.; Meireles, M.A.A.; Marques, M.O.M. Stevia (Stevia rebaudiana Bertoni) leaves pretreatment with pressurized CO2: an evaluation of the extract composition, Proceedings of the 6th Meeting on Supercritical Fluids: Chemistry and Materials, 501 (1999).

Phillips, K.C. Stevia: steps in developing a new sweetener. In: T. H. Grenby (Ed.), Developments in Sweeteners 3, Elsevier, New York, p. 1 (1987).

Rodrigues, V.M.; Meireles, M.A.A.; Marques, M.O.M. Determination of the Solubility of Clove Essential Oil in ScCO2: a standardization of the dynamic method. Proceedings of the 5th International Symposium on Supercritical Fluids (CDROM) (2000).

Soejarto, D.D.; Compadre, C.M.; Medon, P.J.; Kamath, S.K.; Kinghorn, A.D. Potential sweetening agents of plant origin-II: Field search for sweet-tasting Stevia species, Econ. Bot. 37, 71 (1983).

Sovová, H. Rate of the vegetable oil extraction with supercritical CO2: I.- Modeling of extraction curves, Chemical Engineering Science 49, 409 (1994).

Tan, S.; Shibuta, Y.; Tanaka, O. Isolation of sweetener from Stevia rebaudiana. Jpn. Kokai 63, 177, 764 (1988).

Tsanava, V.P.; Sardzhveladze, G.P.; Kharebava, L.G. Effect of technological procedures on the composition of volatile substances in Stevia rebaudiana, Subtrop. Kul't 3 64. Apud: Chemical Abstracts 116: 82387g (1991).

Source: Brazilian Journal of Chemical Engineering