食品和生物分离过程（英文）：32874_02

Chapter 2 Supercritical fluid extraction and its application in the food industry DAVID STEYTLER, School of Chemical Sciences, University of East Anglia, Norwich NR4 7TJ 2.1 INTRODUCTION Solvent extraction is one of the oldest methods of separation known and certainly dates back to prehistory. The science of solvent extraction has evolved accordingly over a long period of time and much progress has been made in the understanding of solvation and the properties of liquid mixtures used in extraction processes. The associated literature on phase behaviour is certainly extensive and, although representation of highly non-ideal mixtures is still problematic, many theoretical models have been successfully developed (Fredenslund, 1975; Hildebrand and Scott, 1950; Prausnitz et al., 1986). Extensive databanks of pure component properties have grown to support such models in order to predict solvent performance in process applications. Today, even with the introduction of new separation technologies, solvent extraction remains one of the most widespread techniques operating on an industrial scale. Hannay and Hogarth’s (1879) early observations of the dissolution of solutes in supercritical fluid (SCF) media introduced the possibility of a new solvent medium. However, it is only in recent years (since 1960) that commercial process applications of supercritical fluid extraction have been extensively examined. In the last decade many advances have been made in researching SCF extraction both in terms of fundamental aspects and commercial applications. In particular the high degree of selectivity and control over solubilities afforded by pressure (and temperature) variation has led to the introduction of many novel SCF extraction and fractionation processes. Of all possible gases, the benign properties (non-toxic, non-flammable) and accessible critical temperature of C02 have ensured its predominance as a safe SCF solvent for the food industry. The essential features of a modern solvent extraction process (using a liquid or SCF solvent medium) are illustrated schematically in Fig. 2.1. The material to be extracted is 18 D. Steytler Extractor Separator Condenser Fig 2 1 Schematic representation ot a solvent extraction process (S = solvent, M = material extracted, E = extract) placed in an extraction vessel (extractor) into which solvent is introduced under conditions (temperature, flow rate etc.) which optimise the dissolution of the desired components. The solvent stream is then passed to a separation vessel (separator) where conditions are set to selectively separate the solvent from the extracted components. The solvent is then condensed and recycled through the system. In conventional liquid extraction, solvents of low volatility are employed with vapour pressures less than one atmosphere. In the course of the extraction process the solvent exists as a liquid during the extraction stage and a gas when it is removed from the extract by distillation. Variations in pressure are small and do not significantly exceed the vapour pressure of the liquid at the extraction temperature. Although temperature variation gives some control over solubility, selective removal of components from a mixture is largely determined by the chemical nature of the solvent. Progressive fractionation can therefore only be achieved by a fortuitous response to temperature or by systematically changing the solvent or the composition of a mixed-solvent system. The initial aims of this chapter are to establish the basic principles involved in SCF extraction. Selected applications are later reviewed with reference to the underlying fun- damental properties that serve to differentiate the behaviour of SCFs from conventional liquid solvents. 2.2 THE SUPERCRITICAL FLUID STATE The P-T phase diagram for C02 showing all four physical states (solid, liquid, gas and SCF) is shown in Fig. 2.2. Below the freezing point solid COz ('dry ice') exists which melts on heating when the thermal energy of the molecules overcomes the lattice energy. The integrity of the liquid state so formed is maintained by relatively weak attractive intermolecular forces (van der Waals). The formation of a supercritical fluid state above the critical temperature (T, = 3 1 .OS"C) can be viewed as an analogous process in which the thermal energy of the molecules overcomes all attractive interactions maintaining the liquid state. Like a gas, the SCF state formed then occupies all available volume. Strictly the SCF state exists above both the critical temperature and pressure (T> T,; P > P,) Supercritical fluid extraction 19 I 600 - 1.1 11.0 - 400 - - - k3 e a -50 0 Tc 50 100 T (“C) Fig. 2.2. Pressure-temperature phase diagram for CO, showing isochores (g ~rn-~). though the latter condition is often relaxed in the technical literature. A substance above its critical temperature therefore behaves like a gas and always occupies all available volume as a single phase. However, unlike a gas, a SCF cannot be condensed to a coexisting liquid-gas state by application of pressure. Similarly when the critical pressure is exceeded it is possible to go from a SCF state to a compressed liquid condition by cooling, but a single-phase filling all available volume is always maintained. It should be appreciated that there are no phase boundaries delineating the SCF state and therefore no sharp changes in physical properties occur on entering this region. Transition to the SCF state from a gas or liquid is thus an ‘invisible’ process. However, if a coexisting liquid-gas mixture is heated at constant volume along the vapour pressure curve, the density of the liquid phase decreases while that of the gas phase increases, until at the critical point they become equal and the meniscus between them disappears. As this point is approached density fluctuations of microscopic dimensions give rise to a distinctive light-scattering phenomenon known as ‘critical opalescence’. Although the supercritical state offers a greater range of density, which in turn provides greater control over solubilities, the liquid state of compressed gases is often employed in extraction processes, particularly for separation of thermolabile components at low temperatures. In order to avoid restrictive and confusing nomenclature, it is convenient to use the term ‘near-critical liquid’ (NCL) to distinguish the state of a compressed gas just below T, from a ‘normal’ liquid at NTP, for which T < T,. The term ‘near critical fluid’ (NCF) will be used in this chapter to represent both SCF and NCL states of compressed-gas solvents. Many liquids commonly employed as solvents enter an SCF state on heating, but for most purposes the critical temperatures are too high to permit their use as SCF solvents 20 D. Steytler (e.g. Tc for hexane is 234°C). All substances with accessible critical temperatures are gases at NTP and representative examples for use in extraction processes are shown in Table 2.1. Being non-toxic, non-flammable, and chemically inert, C02 has obvious practical advantages over other potential gases for use in large-scale extraction processes under pressure. Table 2.1. Potential gases for near-critical fluid extraction Name Formula TC PC (“C) (bar) Carbon dioxide co2 31.1 73.8 Nitrous oxide N20 36.4 71.5 Ammonia NH3 132.4 111.3 Ethane C2H6 32.2 48.2 Prop an e C3H8 96.6 41.9 Freon I3 CClF3 28.9 38.7 Ethylene C2H4 9.2 49.7 2.2.1 Physical properties of NCF COz Density Isochores, representing constant density, are shown in Fig. 2.2 for COz in the NCL, gas and SCF regions of the P-T phase diagram. In the NCL phase, densities are typical of normal liquid solvents (900-1 100 kg m-3) and isothermal compressibility is relatively low. In contrast the SCF state includes a wide range of densities ranging from ‘gas-like’ values at low pressure (< 100 kg m-3) to ‘liquid-like’ values at elevated pressure. The region near the critical point is particularly interesting as it represents the region of highest compressibility. The capability of a solvent to solvate and dissolve a particular solute is directly related to the number of solvent molecules per unit volume. This is because the overall solvation energy is determined by the sum of the solute-solvent interactions occurring primarily within the first solvation shell. Density is therefore a key parameter in determining the effect of temperature and pressure on solubilities in NCF extraction. Indeed, solubility isotherms often exhibit a steep rise with pressure just above the critical point of the solvent where density is rapidly increasing with pressure. The ability to control solubilities through pressure is one of the main features that distinguish NCFs from liquid solvents. Moreover, the potential for differential control of solubilities in multicomponent systems (Johnston et al., 1987) can enable novel fractionation processes that would be impossible using conventional liquid extraction processes. A systematic assessment of the representation of density, and other thermodynamic properties, of C02 by various theoretical models has been made by IUPAC (Angus et al., 1976). This comprehensive treatise provides procedures based on equations of state which 1500 1000 v 5 F 500 0 T V) 30 20 = 60 LIQUID 100 -- SCF _______- - ---- ____----- - JJ150 ___--- GAssi I I I I 0 P, 200 400 600 800 250 2oo 150 - u) s 0 7 1 n 50 0 - GAS T (“C) 100 - 75 - 50 c -___ ________ ____ LIQUID I Ij I 25 I Supercritical fluid extraction 23 the value of the diffusion coefficient under these conditions (D = 4 x lo-* m2 s-l) with hexane at NTP (D = 4 x lo-' m2 s-'). These values are fairly representative and it is generally observed that self-diffusion coefficients for NCFs under typical extraction con- ditions are about an order of magnitude greater than in liquid solvents. Diffusion coeffi- cients of solutes in NCFs are generally enhanced to a similar extent (Section 2.3.3). Volatility (vapour pressure) In conventional extraction processes liquid solvents are recovered by distillation at elevated temperature (and/or reduced pressure) in which valuable volatile components of the extract can be lost. Near-critical fluids are highly volatile and can be completely removed and recycled at low temperatures during an extraction process. This has important implications for improving the quality of extracts, since: (1) (2) (3) Highly volatile components in the extract are retained. This is of particular significance in the extraction of flavours and fragrances. The extract is not subjected to thermal or chemical degradation (e.g. oxidation) at the elevated temperatures employed in distillation. The high volatility ensures 'complete' removal of solvent residues. Any legislative restrictions regarding residual solvent levels are thereby avoided. Chemical properties Of all NCFs, C02 is the safest medium for use in solvent extraction as it provides a non- flammable, non-oxidative environment. C02 does, however, undergo chemical reactions with water which often need to be considered when extracting food materials. One familiar set of reactions is the dissolution of C02 in water to produce carbonic acid: (2.1) The carbonic acid then dissociates and lowers the pH of the aqueous phase in contact with C02: K, C02+H20 4 H2C03 K* H2C03 + H20 e H30+ + HCOT (2.2) HCOC + H20 e H30+ + C0:- K3 (2.3) The pH of water is therefore primarily determined by the partial pressure of C02 with which it is in contact. Water in contact with atmospheric C02 has a pH of approximately 5.7 at 20°C and 3.8 when contacted with pure C02 at the same pressure. With increasing pressure the pH falls further, so that when in contact with liquid C02 at 100 bar the pH is about 3. This represents a fairly typical acidity for water in an NCF C02 extraction process. Since K2 S K3 the hydrogen ion concentration is primarily determined by the initial dissociation of carbonic acid. If the increased acidity is problematic it is possible to suppress the dissociation and buffer the coexisting aqueous phase by addition of 24 D. Steytler bicarbonate anion (Lovell, 1988). The pressure of C02 could, however, be used to control the pH of water in a unique fashion since no chemical residues (of acids or bases) remain. The potential applications of this technique have not been widely explored. A less familiar reaction of C02 with water is the formation of a solid hydrate below about 10°C: (2.4) C02 + 6H2O = C02.6H20 (g) (1) (SI This restricts the use of NCF CO, in the extraction of aqueous systems to temperatures up to 10°C higher than the freezing point of water. (Note: This depends on the type and concentration of the solute.) Biochemical properties At modest levels C02 is non-toxic and so represents a completely safe NCF solvent for food applications with no legislative restrictions governing its use. The only possible, but unlikely, physiological hazard involves asphyxiation by displacement of air following a considerable leak in a confined area. The combined effects of high hydrostatic pressure and low acidity in water-containing systems can be beneficially employed to prevent food spoilage by destroying bacteria (Kamihira et al., 1987; Taniguchi, 1987a). Rapid decompression of dissolved gas is sometimes used to expand and disrupt the cell structure of natural materials and could also be used as a means of sterilisation. Although SCF C02 can be an effective apolar medium for enzyme reactions (van Eijs et al., 1988; Steytler et al., 1991), it has also been used to selectively inactivate enzymes (Taniguchi, 1987b; Weber, 1980). In practice these techniques could be applied either in situ, during an extraction process, or as a separate unit operation. 2.3 PROPERTIES OF NCF SOLUTIONS 2.3.1 Solubilities in NCFs There has been much confusion in some of the literature concerning the solvent properties of NCF C02. An impression is often given that NCFs are universal solvents which can be ‘tuned’ to extract virtually any component of a mixture by selecting a suitable set of conditions of temperature and pressure. Statements to the effect that NCFs are ‘good’ solvents, implying that solute loadings are high, are also prevalent and highly misleading. Before examining the solvent properties of NCFs in detail, it is worth stating a few basic principles: (1) To be ‘supercritical’ intermolecular attractive interactions must be relatively weak compared with thermal energy. This necessitates an absence of all polar inter- actions, such as hydrogen bonding, and defines a medium of low dielectric constant. All NCFs are therefore essentially apolar solvents. The absence of strong attractive interactions between molecules means that solvation energies are generally low and solubilities in NCFs are thus often much lower than in liquid solvents. (2) Supercritical fluid extraction 25 NCFs can be highly discriminating and frequently offer a greater selectivity than liquid solvents. Any attempt at increasing solubilities by changing conditions or injecting entrainers (see Section 2.3.2) usually serves to reduce selectivity. Although some selectivity is sacrificed it is often preferable to operate at high pressures (and temperatures) to obtain sufficient solubility to make a process viable. The conditions cited for a specific process are often arrived at from an optimisation of these opposing effects of selectivity and solubility. However, the selectivity that is exploited in extraction processes is sometimes an intrinsic property of the NCF solvent and is not always dramatically changed by the conditions (e.g. in the selective extraction of triglycerides from phospholipids; Section 2.6.2). (3) General principles Effect of molecular structure Any pragmatic assessment of a solvent extraction process must examine what type of molecules are soluble and to what extent. With NCFs the molecular structure of the solute is of major importance as small changes in molecular weight and functional groups can affect solubility to a greater extent than with liquid solvents. In fact the viability of many simple separation processes using NCFs can be realised without recourse to extensive solubility data covering a wide range of conditions. Francis (1954) has pains- takingly measured the solubilities of 261 substances in liquid C02, and this pioneering work still acts as a useful guide to the relative solubilities of different classes of com- pounds in NCF C02. Dandge et al. (1985) have used this and other data to correlate the solubilities of different classes of chemical compounds with molecular structure. Some of the broad principles emerging from this work are given below: (1) Solubility is reduced by increasing polarity. A good illustration of this is to be found in the relative solubilities of ethanol and ethylene glycol in liquid C02. Whereas the former is completely miscible (M), increasing the overall polarity by introducing a second hydroxyl group reduces the solubility to 0.2%. Miscibility with liquid C02 can be recovered by methylation of the OH groups, which reduces the polarity of the molecule. HO 4 Ho # OH CH30 - OCH3 ethylene glycol ethylene glycol dimethyl ether ethanol (M) (0.2%) (M) (2) Solubility declines with increasing molecular weight and for any homologous series the solubility decreases rapidly beyond a given carbon number. This effect is illustrated in Fig. 2.5, which also serves to demonstrate the effect of polarity on solubility since the more polar alcohol has a lower carbon number 'cut off' than the parent alkane. 20 15 - c - 3 .- b 10 c% - .- D = - 5 0 - I I I - \ \ \ I \ CnH2nt2 Il,l, \, 20 - I I \ \ CnH2n+10H 0 5 10 15 h Supercritical fluid extraction 27 (5) Aromaticity decreases solubility. This is well demonstrated by the progressive decrease in solubility in the series decalin-tetralin-naphthalene as aromaticity is intro- duced into the molecule: a)aa decalin tetralin naphthalene (22%) (1 2%) (2%) A summary of some of the solubility characteristics of selected classes of compounds in liquid C02 is given below: (1) Substances with low molecular weight, and low or intermediate polarity are com- pletely miscible. (a) Aliphatic hydrocarbons (CnH2n+2) n < 12; (M) (note M = miscible) aromatic structures are less soluble but methyl and branched chain substi- tutions increase solubility. (b) Alcohols (CnH2,+1OH) n < 6; (M) further hydroxylation reduces solubility. (c) Carboxylic acids (C,,H2,+1COOH) n < 9; (M) (4 Esters (C,H2,,+1COOC,H2,+1) more soluble than parent acid if rn < n. (e) Aldehydes (C,,H2,,+1CHO) n < 8; (M) aromatic aldehydes are insoluble. Glycerides. The glycerides illustrate an interesting feature since increasing the extent of esterification of glycerol reduces the polarity but increases the molecular weight. The order of solubility reflects the delicate balance of these opposing effects: (f) monoglyceride < triglyceride < diglyceride (2) Macromolecules or highly polar molecules are essentially insoluble, e.g. salts, glycerol, sugars, proteins, starch. 28 D. Steytler (3) Surfactants. Recently there has been much interest in the formation and properties of reverse micelles and water-in-oil microemulsions in the NCF alkanes (ethane- butane) (Gale et al., 1987; Eastoe, 1990 (a, b)). Moreover, the related ‘Winsor 11’ systems display a clear dependence of droplet size on pressure which could be important in the selective separation of enzymes (McFann and Johnston, 1991). However, although some surfactants are soluble in NCF COz, and may well form reverse micelles therein (Consani and Smith, 1990), it is not an effective medium for stabilising microemulsions. Effect of temperature and pressure For liquid solvents with low compressibilities the pressure has very little influence on solubility. A simple explanation of the effect of pressure on solubility in NCFs can be made in terms of the number of solute-solvent interactions which depends upon the density of the solvent medium. The overall shapes of solubility isotherms therefore often closely resemble density isotherms of the pure solvent. At very high pressures, restraints of packing can adversely perturb the preferred molecular orientations required for opti- mum solvation, and solubilities can then begin to decrease with increasing pressure. As heats of solution are more often positive it is generally observed that solubilities in liquid solvents increase with temperature at constant pressure. However, with NCFs the situation is more complex since both density and temperature must be considered. A general statement governing the influence of these parameters is that ‘the solubility increases with increasing temperature at constant density’. This generality is more universally obeyed than the alternative statement in terms of temperature alone. To illustrate these effects the solubility of naphthalene is shown in Fig. 2.6(a) as a function of temperature and pressure. At constant temperature the solubility increases with pressure in accord with the simple picture of increasing solvation through increasing solvent density. Above about 150 bar the solubility increases with temperature as expected but at lower pressures this ‘normal’ trend is reversed and the solubility then declines with increasing temperature. This behaviour, which appears anomalous at first sight, can be explained in terms of the high thermal expansivity of the SCF in the lower pressure domain. In this highly expansive region the large drop in density on heating (at constant pressure) outweighs the thermal enhancement and the overall solubility declines. At higher pressures the thermal expansivity of the fluid is much reduced and the solubility then increases with temperature as in liquid solvents. Figure 2.6(b) shows how the solubility dependence can be simplified by replacing the pressure variable with density. 2.3.2 Theoretical models (equations of state (EOS)) Of all theoretical methods used for the prediction of solubilities in NCFs, the solution of phase equilibrium using equations of state has been most widely applied. The appeal of this approach lies in its simplicity, avoidance of intangible standard states and overall success in correlating the phase behaviour of a wide range of NCF mixtures. To illustrate the general principles involved in the EOS approach, a simple example involving the dissolution of a pure solid (Solute 2) in an NCF (Solvent 1) will be considered as represented schematically in Fig. 2.7(b). Assuming that the NCF does not Supercritical fluid extraction 29 A 500 - 500 - f 450- - & 400- ;;; 400- C al T=45’C i5 C c a, - 8 100- I 60 100 150 200 250 300 350 0 100 200 300 400 500 600 700 800 900 Density p (ghtre) (4 Pressure, P(bar) (b) Fig 2 6 Solubility 01 naphth‘ilene in NCF COz with (d) pressure and (b) density as variable (reproduced trom Brogle, 1982) Ts Ps TP (4 (b) Fig 2 7 Relationship between (a) sublimation of a pure solid and (b) dissolution In an NCF dissolve in the solid phase, the system comprises a pure solid phase, represented by (’), where x = mole fraction of solid in this phase. In this case x2 = 1. This is in equilibrium with an NCF solution represented by (”) with an unknown concentration of solid dis- solved in it, i.e. y;’= ?. The NCF phase is often referred to as the ‘gas’ phase and the symbol y, accordingly used for mole fraction of component 1. The following outlines the procedure for determining y;. The conditions for phase equilibrium are that pressure, temperature and fugacity of each component should be equal in both coexisting phases: 30 D. Steytler T' = T" = T p' = p" = p (2.5) (2.6) f.'= 11 f." (for all i) (2.7) If it is assumed that the NCF does not dissolve in the solid the last condition is simplified since only the fugacity of the solute (2) need be considered: fi = fi' (2.8) fi = Yip (2.9) For an ideal gas mixture the fugacity of each component is equal to the partial pressure: For general application to any system this relationship is modified to include a non- ideality term, the fugacity coefficient ($ i): fi = GiYiP (2.10) fi(T, P) = $;p2s(T) The fugacity of the pure solid phase(s) at the system T and P is given by (2.1 1) To obtain the fugacity at higher pressure it is necessary to introduce a correction term (the Poynting correction): fifi'(~, P) = $;pi(i,exp{Jls (V~IRT) dp} (2.12) Poynting correction where V2 is the molar volume of the pure solid 2. further simplified: If it is assumed that the solid is relatively incompressible, then equation (2.12) can be ~~(~,P,=QIP:(T)~~~{V,(P- P;(T))/RT} (2.13) The fugacity of the pure solid phase at the system temperature and pressure can therefore be obtained from sublimation pressure and molar volume data. Using the general form of equation (2.13) to express the fugacity of the solute in the NCF phase and applying the conditions for phase equilibrium (fi = f;), $; P; (TI exp{ V, ( P - P; (T))/RT} = $;Y;P (2.14) rearrangement then gives Y;' = (Pf(T)/P) ($;/$;')exP{[V,(P-P;(T))/RT]} (2.15) I I I Perfect Non- Poynting gas ideality correction Supercritical fluid extraction 3 1 Since the vapour in equilibrium with a pure solid phase is usually of low density and can be considered ideal ($.j' = l), equation (2.15) gives the mole fraction of solute in the NCF phase explicitly as a function of its sublimation pressure and molar volume. The only unknown quantity is $5, the fugacity coefficient of the solute in an NCF mixture. This can be expressed in terms of the volumetric properties of the mixture as given by equation (2.16), which can be derived from basic thermodynamics (Reid, 1987): ln$~=~,p{[~~ RT 0 2\12 T,p,n2 -~]p (2.16) Solution of equation (2.16) requires an equation of state (EOS) relating the pressure of a mixture to temperature, volume and composition: (2.17) Numerical methods can then be applied to solve the system of equations (2.15)-(2.17) for P = F(T, V, xi) Y5. One of the most familiar equations of state is that of van der Waals (1873): (2.18) RT a p=--- (V-b) V2 The equation essentially corrects the ideal gas equation (PV= RT) for molecular volume (b) and introduces a volume-dependent attractive term (a/V2). The constants for the pure components (ai and 0;) are obtained from the critical properties P, and Tc (values for the constants Fa and F, are given in Table 2.2). Table 2.2. Some equations of state Name Equation Fa Fb van der Waals p=--- 7/64 1 /8 RT a (V-b) v2 0.08664 0.4275 T 'I2 RT a Redlich-Kwong P = ~ - (V-b) T1l2V(V+b) 0.4275 F(w)* 0.08664 RT a Soave p=-- (V - 6) T'I2V(V + b) 0.4572 F(w)* 0.07780 RT a Peng-Robinson p=-- (V - b) V(V + 2b) - b2 * non-sphericity. F(w)* = [I + (0.3764 + 1.54226~- 0.26992~') (1 - T:/*)]' where (w) = the acentric factor representing 32 D. Steytler a=- F,R~T,~ (2.19) PC PC b=- Fb RTc (2.20) For mixtures, a and b can be evaluated from the pure components values (ai, bi) using mixing rules (equations (2.2 1)-(2.23)): a = cxi xj ajj a,. = (ai U,)*/~(I - k..) (2.21) (2.22) (2.23) ij I/ V 0 = c X; b; ij where kij = an adjustable interaction parameter. One important feature of the van der Waals equation is that it can be expressed in an alternative form that is cubic in volume and, given all other parameters, can be solved analytically for Vi andfi. In the quest to provide improved predictive models for NCF systems the van der Waals equation has provided the starting point for many related cubic equations of state. Two commonly used examples are the Redlich-Kwong (Redlich and Kwong, 1948) and Peng-Robinson (Peng and Robinson, 1976) equations (see Table 2.2) which both introduce a more sophisticated expression for the attraction term but retain the original form of the repulsion term. Both equations can easily be solved analytically or numerically for volume and fugacity coefficient and have been used in numerous correlations of solubility data in NCFs (McHugh and Krukonis, 1986; Prausnitz, 1965; Rizvi er ul., 1986). Many variations on the mixing rules (equations (2.21)-(2.23)) have also been proposed (Kwak and Mansoori, 1986; Deiters, 1982; Mathias and Copeman, 1983). A comparative study of the effectiveness of a variety of commonly used equations of state in representing phase equilibrium data in NCF mixtures has been made by Haselow et al. (1986). The simple EOS approach represents the solubilities of volatile, low molecular weight solutes in NCFs reasonably well, as shown by the fit obtained using the Peng-Robinson EOS to the data for naphthalene in SCF CO, shown in Fig. 2.8. Inspection of this data reveals how the components of equation (2.15) used in the EOS method contribute to define the solubility curve. In the low-pressure region the behaviour of the NCF is approximated to that of a perfect gas, for which the lowest pressure obtainable is that of the sublimation pressure of the pure solid. The low-pressure limit for the mole fraction of naphthalene in the mixture is simply given by the composition of the mixture at this limiting pressure. As the pressure is increased, yz initially declines in accord with the first ‘perfect gas’ term in equation (2.15). As the pressure is further increased y;’ begins to increase as the SCF progressively solvates the solute and $; decreases. Transfer of solute to the SCF phase is also enhanced by the effect of hydrostatic pressure through the Poynting correction term. Supercritical fluid extraction 33 lo-’ 10-2 ._ s .!l 10-3 ? i% 10-~ 2 P 10-5 c u - 0 E z 1o-6 0 50 100 150 200 250 Pressure (bar) Fig. 2.8. Representation of experimentally determined (0) solubility data for naphthalene in SCF C02 at 25OC by the Peng-Robinson equation of state (reproduced from Paulaitis et nl., 1983). The application of the EOS approach to liquid solutes is more involved since the NCF can dissolve in the liquid phase and the fugacities of both components in two coexisting phases must then be considered (Prausnitz and Benson, 1959). Relatively simple compu- ter methods are however available (McHugh and Krakonis, 1986) for an iterative solution of the condition of equal fugacities (equation (2.7)) of each component in both phases. Phase equilibrium data for liquids is usually represented at constant temperature as a function of pressure (or vice versa) as shown in Fig. 2.9 for the butanol/C02 system. The region enclosed by the loop represents a two-phase region in which a C02-rich ‘gas’ phase coexists with a butanol-rich ‘liquid phase’. At constant pressure the compositions of the coexisting ‘liquid’ and ‘gas’ phases are given by the points of intersection of a horizontal ‘tie line’ within the loop. The relative proportions of the coexisting phases can be obtained in the usual way using the ‘lever rule’. In this example, increasing the pressure increases the solubility of butanol in CO2 until at a critical pressure of approxi- mately 160 bar complete miscibility occurs. This represents the particularly simple behaviour of a liquid with low molecular weight, but as molecular weight and/or polarity is increased the phase behaviour becomes more complex. A rich variety of phase dia- grams have been observed and systematically classified for binary NCF mixtures (Schneider, 1970). One limiting factor which restricts application of EOS models for food-related applications is the lack of available data for the fundamental properties of pure components required for input. Another problem concerns the ambiguity of mixing rules and associated ‘adjustable parameter(s)’ (e.g. kc in equation (2.22)) which are often ‘floated’ in fitting experimental data. Although attempts have been made to define such parameters in terms of pure component properties, EOS models have been more - Redlich-Kwong prediction 4- -*"I*" 1 x 100 bar o 300bar 500bar X 3- x e OO Ox e& I li81 Y , . 2- -3 - u) I -5 . N 0 - F -4 40°C - o Benzene in COP X Phenol in C02 0 Naphthalene in COz A Caffeine in C02 7 -kA - I I I I I I Supercritical fluid extraction 37 the concentration of solute in the NCF solvent leaving the extraction vessel (which determines the number of cycles required to complete the extraction); the differential conditions of pressure and temperature between the extraction and separation vessels (which determine the energy consumption per cycle). In examining process efficiency, the extraction and separation stages will be treated (1) (2) separately. 2.4.1 Extraction stage To illustrate the differences between NCF and liquids as extraction media the example of leaching of porous particles will be used. This involves the dissolution of a solute, initially evenly distributed throughout the porous matrix within the particles that are themselves insoluble. There are widespread examples of leaching in the food industry, including extraction of seed oils, decaffeination of coffee and isolation of flavour components from plant materials. The rate of removal of solute will depend upon many factors which are common to any solvent, including: (1) (2) (3) (4) the amount of solute in the particle; its distribution within the matrix; the particle size and shape; the geometry of the porous network. To simplify the analysis it will be assumed that the solute to be removed is evenly distributed throughout a particle of regular shape (slab) as shown schematically in Fig. 2.12. Similar arguments apply for other geometries (e.g. sphere, cylinder). _I - j. I j. 0 (4 0 (b) E Fig. 2.12. Schematic representation of the extraction of solute from a slab by (a) the ‘shrinking core’ model and (b) the ‘free diffusion’ model. Mechanism of extraction The solubility of the material to be extracted is clearly a prime factor in determining the effectiveness of a solvent in an extraction process. However, when solubilities are low it can also determine the mechanism by which the extraction proceeds. Because solubilities in SCFs are often low, conditions are frequently incurred which serve to establish a 38 D. Steytler different mechanism (shrinking core) to that commonly found in liquid solvent extraction (free diffusion). The yree diflusion’ model When the solute in the pores is completely dissolved in the solvent it can undergo free translational diffusion in all directions within the confines of the pore network. This condition will be met when the solubility (S) significantly exceeds the concentration (0 of solute within the matrix expressed as the amount of solute per unit volume of solvent in the pores (Le. the pore volume): SBC (2.24) It is more convenient to rewrite this condition expressing C in terms of the overall concentration of solute in the particle (0 and the volume fraction of the pore space (q) s s- c/q (2.25) When inequality (2.25) applies, the extraction proceeds by a mechanism in which all of the solute molecules initially dissolve and undergo random translational diffusion, eventually reaching the surface of the particle where they enter the surrounding solvent. The diffusion equations defining the rate of extraction by this mechanism are given by Carman and Haul (1954) for a variety of particle shapes (sphere, cylinder, slab etc.). The fractional extraction of solute (M,/M,) from a infinite slab of half-thickness (L) into a fixed volume of solvent is given below (equation (2.26)): -- hf, - = 2a(l+a) exp(-Deffq:r/12) (2.26) where M, and M, = the mass of solute extracted at time t and infinity, a= the volume ratio of solvent to slab and the qn terms are the positive non-zero roots of M, 1 - 17=1 c 1 +a +ay tan qn = -aq, The extraction profile is given by the weighted sum of a series of exponential func- tions of time. The diffusion coefficient (0“‘) in equation (2.26) is the effective diffusion coefficient of the solute in the particle and represents a restricted diffusion process within the pore network. It is related to the bulk diffusion coefficient (Db) through the tortuosity factor, 7, and the porosity, cp, which accounts for the increased effective diffusion path imposed by the pore structure: D eff- -cp D b/ 7 (2.27) A comprehensive treatment of a wide variety of related diffusion problems relevant to solvent extraction is given by Crank (1989). The ‘shrinking core’ model A very different mechanism of extraction applies when the solubility of the solute is low and the condition of inequality (2.25) is reversed, i.e. s ‘=3 c/q (2.28) 0 II Ill IJ 40 D. Steytler Solubility The most important feature of the ‘shrinking core’ model is that the rate of extraction is determined not only by the diffusion coefficient but also by the solubility of the solute. Clearly when the solubility is low only a small portion of the solute can dissolve and participate in the diffusion process. Rates of extraction can therefore be very low when these conditions apply and this feature has been successfully exploited in the develop- ment of devices for controlled drug delivery. Generally solute loadings in SCFs are much lower than in liquid solvents, for which it is possible to choose from a wide range of solvent polarity in order to optimise solubility. The ‘shrinking core’ mechanism is there- fore more likely to prevail in NCF solvents and is often responsible for the slow extraction rates associated with NCF leaching processes. Diffusion coefficient It is frequently reported that the higher diffusion coefficients in SCFs give rise to a more rapid extraction than could be achieved using liquid solvents. This is most certainly the case when the ‘free diffusion’ model applies. Unfortunately the low solubility in SCFs often serves to switch the mechanism to a ‘shrinking core’ model in which the enhanced diffusion in the SCF is offset by the diminished solubility. Therefore, although having other advantages, SCF extraction can sometimes be very slow in comparison with extrac- tion processes using liquid solvents. Adsorption The extent by which a solute molecule in solution partitions to a surface is determined by the differential energy of interaction with the surface (adsorption energy) and solvent (solvation energy). Because solvation energies in SCFs are generally lower than in liquid solvents, partition coefficients to surfaces are often enhanced and adsorption effects are more pronounced. This has significant implications for extraction since adsorption serves both to retard the important rate-determining diffusion processes in leaching and also to reduce the maximum equilibrium concentration of solute that can be attained in the extraction vessel. In the leaching process the simplest case of rapid adsorption/desorption of solute onto the internal pore surface can be accounted for by a linear isotherm. In this model the diffusion coefficient is replaced by a modified diffusion coefficient (0“‘) inversely re- lated to the adsorption coefficient (Kads) (Crank, 1989): Deff = D /( Kads -k l) (2.30) where concentration of adsorbed solute concentration of freely diffusing solute Kads = When Kads > 1 this will clearly reduce the rate of extraction. In a fixed-bed extraction process, in the absence of adsorption, the maximum attain- able concentration of solute in the solvent stream (Cm”) is given by the solubility. This Supercritical fluid extraction 41 determines the minimum amount of solvent required to complete the extraction. In the presence of selective partitioning to the particle bed Cmax is similarly attenuated: cmax = S/(K,,, + 1) (2.31) The implications for extraction efficiency are twofold: (1) (2) More solvent will be required to complete the extraction, Le. the process must operate through more cycles. Because the level of the solute in the SCF solvent is lower, a greater pressure differential is required to affect its precipitation in the separation stage of the process. Both of these factors serve to increase the energy consumption of the process. The role of water The addition of entrainers to promote solubility in NCFs has been mentioned previously. In the extraction of plant materials it is often found that the addition of water is essential in order to achieve a realistic extraction rate. The role of water in enhancing NCF extraction is quite distinct from that of entrainers, since it does not usually increase solubilities in the SCF. This is not surprising, since the solubility of water in C02 is very low (< 0.5% w/w; Wiebe and Gaddy, 1941). One way in which water affects the rate of extraction is through rehydrating and swelling the internal cellular structure of dried plant material. This can affect the extraction rate in two opposing ways. Firstly increasing the particle size will also increase the distance of diffusion (e.g. L in equations (2.26), (2.29)) within the pore network which will retard the leaching process. A compensating factor, however, is provided by the expansion of the internal structure which can shorten the diffusion path to the surface by opening channels and removing restrictions. Hydrating the plant material can therefore enhance the overall extraction rate. In some cases water can play a crucial role in determining not only the rate but also the mechanism by which the extraction proceeds. Such behaviour is observed in the decaffeination of tea (or coffee) with SCF C02. Dry tea contains approximately 3% w/w caffeine, for which the solubility in SCF C02 is between 0.05 and 0.20%. Assuming a pore volume fraction of 0.2 it is clear that the criteria of equation (2.28) are met and in the absence of other factors the extraction of caffeine from dry tea will follow a ‘shrink- ing core’ mechanism. The extraction rate is then expected to be strongly solubility- limited. The solubility of caffeine in water, however, is significantly higher (2% w/w at 25°C to 10% w/w at 6OOC) than in C02. If the tea is completely hydrated, much water is absorbed and the leaves swell to approximately three times their dry volume. All of the caffeine can then dissolve in the water within the internal cellular structure and extraction can proceed by the ‘free diffusion’ model. These predictions are borne out by experiment (Steytler, 1988) as shown in Fig. 2.14, which represents the extraction of caffeine from the same quantity of wet and dry tea into a fixed volume of SCF C02. From the limiting extent of extraction it is clear that caffeine is strongly adsorbed onto the tea and this preferential association persists even when water is added. The selectivity in the extrac- tion of caffeine from wet tea takes place at the water/C02 interface at the 42 D. Steytler 12 10 ----------d Wet tea - ga s 24 .G 6 e 0 +- 2 0 0 50 100 150 200 Time (min) Fig. 2.14. Extraction of caffeine from dry and humidified tea (containing 50% water) into a fixed volume of supercritical C0,,(T=,4ODC, P = 300 bar). The dashed line represents a best fit of the ‘free diffusion’ model (eq. (2.26)) to the wet tea data. surface of the leaf. Within the leaf many other water-soluble components may be dis- solved and freely diffusing. 2.4.2 Separation stage Since the solubility of all solutes in SCFs declines with decreasing pressure, the obvious way to separate the SCF at the separation stage is by decompression. This can, however, make heavy demands on energy input to the process, since the SCF then has to be recompressed on each cycle through the system. One solution to this problem has been to preferentially partition the solute to a coexisting solid or liquid phase in the separation vessel maintained at the same pressure as the extractor. This technique is particularly useful for processes in which the extracted components are undesirable contaminants present at low levels. Active carbon can then be employed to selectively adsorb the dissolved components from the CO2 stream in an ‘isobaric’ process. In processes for the decaffeination of coffee, caffeine is often removed in this way by adsorption onto active carbon or by partitioning to a coexisting aqueous phase. Unfortunately most low molecular weight organic solvents are miscible with CO, (Francis, 1954) and cannot be used for selective partitioning. A feature of the separation stage, which is currently receiving much attention, is crystallisation. It is generally observed that the rapid decompression accompanying the separation of crystalline solids produces smaller crystals. This has formed the basis of a process for commutation of materials that would otherwise be difficult to handle by conventional milling techniques (Krukonis, 1988). Unlike temperature, pressure variation offers a rapid, isotropic response which can be used to control crystal size (Tavana and Randolph, 1989) and morphology (Ohgaki et al., 1990). One process of relevance to the food industry concerns the application of repeated decompression cycles to induce net- work growth in aspartame crystals (Steytler et al., 1990) as shown in Fig. 2.15. This Supercritical fluid extraction 43 Fig. 2.15. Electron micrographs showing aspartame crystals after repeated decompression cycles in SCF CO, at 40°C. process differs from other techniques of SCF nucleation since it involves repeated dissolution/deposition of material to induce overall particle growth. The product is a free-flowing powder without the ‘dust’ problems associated with the fine crystalline structure of the starting material. 44 D. Steytler 2.5 EQUIPMENT AND EXPERIMENTAL TECHNIQUES USED IN NCF EXTRACTION AND FRACTIONATION Although there is some degree of overlap, distinction will be made between equipment that has been designed for use in either fractionation or extraction processes. Details of high-pressure equipment design and engineering are not included here but can be obtained from standard references (Tongue, 1959; Tsiklis, 1968). A description of large- scale commercial-scale plant is also omitted since accurate information of plant operating in the commercial sector is often difficult to obtain and highly process specific. Pertinent design features of large-scale SCF extraction equipment are available in specialist reviews (Bohm et al., 1990; Korner, 1988; Marentis and Vance, 1990; Eggers and Tschiersch, 1980), conference proceedings (Erlangen, 1984) and directly from manufacturers. 2.5.1 Extraction Pilot plants with recirculation A typical pilot plant configuration is shown schematically in Fig. 2.16. Although this type of equipment is designed for testing applications on a relatively small scale (< 10 kg) it includes most of the important features of a commercial-scale plant. The essential components of the plant are an extraction vessel, separation vessel, condenser and pump. The C02 is stored at its vapour pressure, in the condenser, as an NCL, and is pumped into the extraction vessel by a high-pressure liquid pump. The state of the C02 in the extraction/separation stages (SCF/NCL or gas) is determined by the temperature and pressure of the vessels. The temperature is controlled by a thermostatted recirculating fluid flowing through ‘heating jackets’ surrounding the vessels and the pressure is main- tained by pressure relief valves. Conditions of temperature, pressure and flow rate are continuously monitored and logged at strategic points throughout the system. In an extraction process the material to be extracted is placed in the extraction vessel, which is purged with gaseous C02 to remove all air from the system. The extraction is then started by pumping the liquid C02 through a heat exchanger into the extraction vessel. The flow rate, which is determined by the pump, is usually set to allow sufficient residence time in the extraction vessel for equilibrium solubility to be attained. The solution then passes to the separation vessel, where conditions are set to minimise the solubility of the extracted components. This often involves decompression to a low- density SCF state (T> T,) or to a state on or below the liquid-gas coexistence line (T< T,). Gaseous C02 then passes to the cooled condenser, where it is condensed and stored as a liquid. Because pressure vessels incorporate thick metal walls, visual monitoring of the extraction process presents some problems. This is particularly pronounced at the separa- tion stage, where it is necessary to monitor the amount and quality of extract obtained during the extraction. If the extract is a liquid it can be periodically removed from the separation vessel by a drain valve, but if a solid it is necessary to halt the extraction to remove the deposits by washing. Non-intrusive monitoring of the extraction can be achieved by spectroscopy (UV/Vis/NIR), if high pressure windows are incorporated in Liquid alternative - separator Pressure i.e, #toptical cell!r Pressure y control control valve a valve inlet EXTRACTOR SEPARATOR VESSEL VESSEL Max. pressure Max. pressure 700 bar 200 bar Temperature Temperature +5to t80"C co2 SUBCRITICAL CONDENSER - A CG B Supercritical fluid extraction 47 Cascades of separation vessels Fractionation of a mixture can be effected on a ‘time-controlled’ basis, using the pilot plant described above, by collecting fractions at constant temperature and pressure during the course of the extraction. This is because the more soluble components of the mixture are preferentially extracted in the early stages of the extraction. However, because the solubilities of less soluble components may be very low, this mode of operation can be highly inefficient in terms of time. It is therefore preferable to gradually increase the ‘solvent power’ of the NCF by progressively changing the conditions of temperature and/or pressure in the extraction vessel. One such procedure involves progressively increasing the extraction pressure in stages such that the more soluble components are removed first. To achieve good separation between stages it would be necessary to continue the extraction at each pressure until equilibrium was approached and the rate of removal of components began to level off. The separation vessel would then have to be emptied, which could involve halting the extraction. This would clearly give rise to a very protracted process. A simple solution to this problem is to introduce the fractionation at the separation stage of the process by using a cascade of separation vessels as shown in Fig. 2.18 Here the sample is extracted under extreme conditions (high pressure/temperature) at which all extractable components are removed in the extractor. The CO2 stream then passes to the first separation vessel where the conditions P,, TI are set to precipitate the first fraction of least soluble components. The output stream from the first separator is then passed to a second vessel at lower pressure P2, where the second fraction precipitates. The fractionation thereby takes place down a ‘cascade’ of separation vessels set at pro- gressively decreasing pressures. Because separation is much more rapid than extraction the process is not rate-limited. Care must, however, be taken to avoid carrying-over of precipitated material between separation vessels. Separator 1 Separator 2 Separator 3 P1, T1 P2, T2 P3, T3 Fig 2 18. Separation stage configuration tor NCF fractionation using a ‘cascade’ of separation vessels (E = extract) Zosel’s ‘hot finger’ fractionation column The anomalous effect of temperature at low pressures in SCFs has been introduced previously (Section 2.3.1) in which increasing the temperature results in a drop in solubil- ity. If the temperature rise is sufficiently high, this behaviour can also be extended 48 D. Steytler to fluids at higher pressures. Zosel (1978) has exploited this effect in an elegantly simple fractionation apparatus (Fig. 2.19) in which fractionation and extraction take place in a single high-pressure vessel. The mixture to be fractionated is placed in the bottom of the vessel and a ‘hot finger’ probe at the top serves to establish a temperature gradient throughout the fluid in the column. Ignoring the small pressure gradient required to maintain flow, the vessel can be considered to be at constant pressure. The density of the fluid and solubility of compo- nents are therefore uniquely determined by temperature. Solvation in the high-tempera- ture (low-density) region in the vicinity of the probe will be at a minimum and only the most soluble components will remain in solution and be able to pass through to the separation vessel. Less soluble components will precipitate and remain under reflux in the lower portion of the column. Fractions can be removed from the column in order of decreasing solubility by progressively lowering the temperature of the probe and/or increasing the pressure. The potential of the apparatus has been demonstrated in the fractionation of fish oils into over 50 discrete fractions. Fig. 2.19. An NCF fractionation column using a ‘hot finger’ to control selectivity (Zosel, 1978). 2.6 APPLICATIONS The last decade has seen a rapid growth in research activity in the area of SCF extraction. Although many glowing reviews and hundreds of publications have appeared covering a wide variety of applications, there are surprisingly few processes and plant operating on a commercial scale. It is possible that many of the early publications overstressed the potential advantages of SCFs without addressing the limitations. It is certainly the case that many reports of applications do not differentiate between the extraction/identification of trace amounts of components and realistic quantities on which a process could be Supercritical fluid extraction 49 based. Without the necessary data it is often difficult or impossible to quantify the efficiency of a reported SCF extraction process in terms of mass transfer and throughput of C02. Against this background it is not surprising that many popular misconceptions abound regarding SCFs. SCF extraction is an expensive process and, although offering many advantages (Table 2.3), should not be used simply because it is a ‘novel’ technique. Unless use is being made of its unique features there is no rationale for its implementation if a cheaper separation process can meet the requirements of the separation as effectively. The one clear advantage that C02 does offer for food applications is its lack of toxicity. In the current climate of growing consumer concern regarding food safety this feature will undoubtedly promote its use in the food industry. The following review of applications includes examples of processes currently operat- ing on a commercial scale and applications that could soon be implemented. Examples of applications showing novel features that serve to illustrate the potential of SCF extraction are also included. Table 2.3. Criteria determining implementation of SCF CO2 extraction Criteria Advantages Disadvantages Economic Can save energy High capital cost Cheap Stable market Safety Non-toxic* High pressure Non-flammable Physical properties Enhanced control through T Low solubilities and P (fractionation*) High vapour pressure enables separation at low Low viscosity can provide enhanced mass transfer Chemical properties Non-oxidative pH effects and hydrate Environment* formation when water T* present * = implication for improved quality. 2.6.1 Decaffeination of coffee and tea Over the past decade there has been a growing consumer aversion to the levels of stimulants in beverages and there is now a large market in decaffeinated products. Although there are some notable differences, the conventional solvent and NCF extraction processes for coffee and tea share many common features. 50 D. Steytler In the production of coffee ‘green’ beans are roasted to generate the coffee oils that later impart flavour to the infusion. To avoid co-extraction of flavour components in the decaffeination process green beans are therefore generally extracted. Moist beans are used since it is found that dry beans do not allow effective extraction. The crucial role of water in the process has not been unambiguously established but is thought to ‘free’ the caffeine from adsorption on the surface, reduce complexation with other molecules and reduce the tortuosity factor by swelling the cellular structure. In the conventional decaffeination process, organic solvents such as methylene chloride or ethyl acetate are employed to reduce the level of caffeine from approximately 1% w/w to 0.06%. Heightened awareness of the potentially harmful effects of residual levels of these solvents has provided some impetus to examine alternative safer solvents such as NCF C02. Extensive kinetic studies of the infusion of coffee beans into water (Spiro et al., 1984) and organic solvents (Bischel, 1979) have been reported. Judging by the scale of its implementation, decaffeination of coffee is one of the most successful commercial applications of C02 extraction with large plants operating in West Germany (Bremerhaven, 27.3 million kg per year) and the USA (Dallas). At first sight C02 extraction would not appear to be promising since the solubility of caffeine in NCF CO, is relatively low (< 0.2% w/w; Ebelling and Franck, 1984), but this is offset by the high added value of the process. Moreover, C02 provides a very selective solvent for decaffeination which does not remove as many of the desirable flavour-precursor compo- nents as alternative organic solvents (e.g. ethyl acetate). Selectivity for caffeine is probably greatest in NCL C02, though solubility is low (approx. 0.05%). Increasing pressure and temperature serves to increase the solubility of caffeine but reduces the separation factor. In some processes for production of de- caffeinated instant coffee, flavour components that are co-extracted are later separated and added back at the drying stage (Roselius et al., 1974). It is often stated that the water added in decaffeination serves to increase the solubility of caffeine in NCF C02. This hypothesis is not borne out by experimental evidence, since increasing the levels of water to saturation does not significantly affect the solubility of caffeine in SCF C02 (Moulson, 1988). A comprehensive review of the patent literature on decaffeination (and other NCF extraction processes) is given in an excellent treatise on NCF extraction by McHugh and Krukonis (1986). Most of the methods employed share common features. In a typical process (Zosel, 1974), wet green beans are contacted with SCF C02 in an extraction vessel and the extracted caffeine is removed by partitioning to a coexisting aqueous phase. Other methods of separation of caffeine from the C02 stream involve adsorption onto active carbon (Roselius el ul., 1979). The latter approach is less desirable since the caffeine, which can be sold as a by-product, is destroyed by burning during the reactivation of the carbon adsorbant. Some patents (Prasad et ul., 1981; Margolis and Chiovini, 198 I) describe techniques for the extraction of aqueous roasted coffee extracts in a continuous counter-current extraction process. One of the more novel claims in the patent literature involves mixing the beans with active carbon and carrying out the extraction and separation in one vessel (Zosel, 1981). This not only saves on capital costs, but also promotes rapid mass transfer. The activated carbon (containing extracted caffeine) is later removed by sieving. Supercritical fluid extraction 51 Decaffeination illustrates well the adverse effects of solute adsorption since it is observed that considerably more C02 is required to effect decaffeination than would be expected on solubility grounds alone. This is because caffeine adsorbs onto the beans and prevents equilibrium solubility being reached in the extraction vessel (McHugh and Krukonis, 1986). Decaffeination of coffee represents one of the most widely patented applications of NCF extraction with innumerable variations and permutations claimed. In contrast the NCF decaffeination of tea has been less well documented (Vitzthum and Hubert, 1979). This may well be due to the more delicate flavour profile of tea, which is more susceptible to damage during the extraction process. The higher levels of caffeine in tea (3% compared to 1% for coffee) may also be a contributing factor. 2.6.2 Seed oil extraction Extraction and processing of seed oils is a large-scale commercial operation with high throughput. The type of oilseeds processed depends almost entirely on regional agricultural policies; in the USA soya oil is by far the largest commodity, whereas in Canada rapeseed oil is more common. However, seed oils all contain the same basic triglyceride units, though the distribution of individual carboxylic acids in the triglycerides is a unique feature of each oil which imparts individual characteristics. The overall solubility of seed oils in NCF C02 does not appear to vary much and basic principles and conditions established for the extraction of one oil translate reasonably well to another. In the conventional process for oilseed extraction the pretreated seeds are extracted using hexane. This also removes phospholipids (lecithin) which, although beneficial to health, present physical problems when the oil is used for cooking. In the refining process a degumming stage is therefore required to remove the phospholipids from the oil which is then bleached (to reduce colour) and deodorised. Most of the pioneering work on the SCF C02 extraction of seed oils has been carried out in the USA and West Germany. The solubility of soya oil in SCF C02 has been measured at high pressure (Friedrich et al., 1982; Stahl et al., 1980, 1983a, 1984) and there has been some success in fitting solubility data using simple EOS models (Klein and Schultz, 1989). The most important feature of Friedrich’s measurements was that at high pressures and temperatures (800 bar, 70°C) the oil became completely miscible with C02. This suggested the possibility of an efficient high-pressure SCF extraction process (Friedrich and Pryde, 1984). Under these extreme conditions the rate of extraction from soya flake was found to be rapid, and nearly complete extraction could be achieved in 20 min. Moreover, since the solubility falls markedly at lower pressures, most of the oil could be separated from the C02 stream without having to undergo complete decompres- sion. Other investigators of the rate of NCF extraction of oilseeds have worked at lower pressures (e.g. King et al., 1987; Brunner, 1984). Probably the most significant feature of the SCF extraction of oilseeds is that phos- pholipids are not co-extracted (Friedrich and Pryde, 1984). This eliminates the need for chemical degumming of the oil. If the phospholipids are required as a separate commod- ity they can be extracted from the seeds in a secondary extraction with a solvent 52 D. Steytler such as hexane. Oil extracted with SCF C02 is also often reported as having a lighter colour. Although showing many advantages over conventional processing techniques (List and Friedrich, 1985) NCF extraction is not currently economically viable for large-scale oilseed extraction. This is due to the low bulk value of seed oils, high plant cost and inconvenience of batch processing large quantities of solid materials under high-pressure conditions. As a result there has been much interest and speculation concerning the development of systems for continuously feeding solids into high-pressure vessels. Although a successful design was originally implemented for transferring coal slurries at low pressure (Blisset et al., 1979) there appears to have been little success in producing a working system for operating under the higher pressures involved in SCF extraction, though many designs have been proposed (Stahl and Stadler, 1984). Implementation of SCF extraction of bulk commodity seed oils will probably have to await the development of continuous processing plant or the tightening of legislation governing the use of petrochemical solvents. SCF extraction is, however, appropriate for smaller-scale processes involving high-value oils such as evening primrose (Tolboe et al., 1988) and wheatgerm (Christianson et al., 1984) for which there is a growing market in the health sector. 2.6.3 Purification of lecithin Lecithin is an important emulsifying agent in the food and pharmaceutical industries. Crude lecithin as produced in the degumming of seed oils is a darkly coloured, highly viscous substance with a 'grease-like' consistency. After extraction with acetone, which removes associated oil and other components, the purified lecithin is obtained as a pale yellow powder with a high phospholipid content (95%). The lecithin residue contains significant levels of residual solvent which are removed at high temperatures at which further degradation of the product can occur. The large differential in solubility between triglyceride oils and phospholipid suggests the feasibility of using NCF CO2 to purify lecithin. The main problem here is the highly viscous state of lecithin, which presents enormous difficulties in its transfer into extraction vessels, and achieving good solvent contact therein. Stahl and Stadler (1984) have proposed a process for the continuous extraction of lecithin which involves high-pressure extrusion of the crude lecithin into an extraction vessel through a small nozzle. Another solution to this problem, incorporating some ingenious features, has been reported by Peter et al. (1989). The possibility of using NCFs to reduce the viscosity of liquids has been previously mentioned (Section 2.2. I ). However, because C02 is not very soluble in lecithin it is not possible to use SCF C02 to reduce its viscosity. Instead, NCL propane was employed by Peter to thin the lecithin which was then pumped into an extraction vessel containing SCF C02 at 55"C, 80 bar, where most of the lecithin was selectively precipitated. By employing temperatures at which the lecithin can be removed as a liquid the plant could be made to operate continuously and at low pressures. The paper presents a model example of an SCF process development study including phase equilibrium measurements, plant design and cost analysis. Supercritical fluid extraction 53 2.6.4 Lowering cholesterol levels in foods Although the correlation between dietary intake and levels of cholesterol in the blood is not universally accepted, initial suspicions have led to strong consumer aversion to high- cholesterol foods. This has been compounded by advertising campaigns claiming low levels of cholesterol, even in products that would not be expected to contain it. On solubility grounds there would appear to be a good chance of selectively extracting cholesterol from oils and fats using NCF CO2. Chrastll (1982) has measured and collated the solubilities of a variety of food components in NCF CO, and correlated his data using a simple ‘mass action’ model. Even from the limited pressure range used in this study it is clear that the solubility of cholesterol is significantly greater than that of triglyceride oils. Krukonis (1988) has tested the feasibility of removing cholesterol from butter, egg yolk and beef tallow by measuring the partition coefficients of the individual components in SCF CO,. Selectivity at 60°C and 150 bar was found to be greatest for egg yolk (12.2) and least for butter (3.4). This trend is in accord with the relative solubilities of the compone:it oils in SCF CO2. In these trials 90% removal of the cholesterol from butter was reported with an overall yield of 70% of cholesterol-reduced product. Distribution coefficients of cholesterol between milk fat and SCF CO2 reported by Bradley (1989) support the viability of selective separation. Studies on the effect of extraction conditions upon the composition of SCF CO2 extracted egg yolk powder (Froning et al., 1990) are in broad agreement with available solubility data. 2.6.5 Fractionation of high-value oils and fats Butterfat Milk fat contains fatty acids representing a wide range of molecular weight and unsaturation. The acid residues are distributed in a ‘random’ fashion in the constituent triglyceride oils. Conventional fractionation techniques (e.g. cry stallisation - see Chapter 8) are unable to effectively concentrate a short chain acid fraction (C4-Cl0) since these acids are combined in the triglyceride structure with higher molecular weight acids (C14, C16, CIS) which have a controlling influence on the separation. The fractionation of butterfat is of considerable interest in the dairy industry for manipulation of physical properties (e.g. spreadability, mouthfeel) and functionality in other milk products and foodstuffs. Short-chain fatty acids are also of physiological importance since they are more rapidly assimilated in the body. Kaufmann el al. (1 982) have examined the ‘time-controlled’ fractionation of butterfat in an SCF extraction process at 80°C, 200 bar and the two fractions taken showed clearly different compositions of component oils. The first fraction was found to be 81% enriched in the short-chain (C4-Clo) acids with a simultaneous 44% reduction in the oleic acid content. Accordingly the second fraction showed a higher proportion of longer-chain acids and contained 99% of the C24-C44 triglycerides. More recently Kanare et al. (1989) have reported the composition of extracts taken using a ‘pressure-controlled’ SCF extraction process in which the pressure was progres- sively increased from 100 to 400 bar at 5OoC. Analysis revealed a similar pattern to 54 D. Steytler Kaufmann with the short-chain acids being selectively extracted in the early stages of the extraction. Both cholesterol and lactones were also preferentially removed in the initial extracts. Fish oils Certain fish oils, which are rich in the highly unsaturated eicosapentaenoic (EPA, C20:5) and docosahexaenoic (DHA, C22:6) acids, are now believed to be of great dietary value. There is therefore considerable interest in the isolation of fractions of fish oils enriched in these component acids. To effect a realistic separation by any technique the acids in the oil must firstly be removed from the triglyceride structure by hydrolysis and esterified, usually as the ethyl ester. The conventional fractionation process then involves distilla- tion techniques which are limited in terms of both thermal degradation of the oils and the separation factors obtainable. One of the first reports of the application of SCF extraction in the fractionation of fish oils was by Eisenbach (1984) who demonstrated enrichment of the component acids of cod liver oil by carbon number. Further separation of acids differing only by number and position of double bonds could not be achieved. However, the technique was successful in producing a C20 fraction containing about five times the concentration of EPA of the starting material. Nilsson et al. (1988) solved the problem of limited separation of EPA in a combined SCF extraction/urea crystallisation process in which less saturated acids (e.g. C20:0, C20: 1) can be selectively removed. In this combined process, fractions containing 96% pure EPA were obtained from menhaden (large herrings) oil. In the course of this work distribution coefficients of more than twenty individual acid components were determined in SCF C02. The feasibility of this fractionation process has recently been established in a series of pilot scale tests by Krukonis (1988). 2.6.6 Extraction of flavours and fragrances Flavours and fragrances are conventionally isolated from botanical sources either as an absolute, using solvent extraction, or by steam distillation. The main drawbacks to these methods are thermal degradation (e.g. of sesquiterpenes), loss of volatile ‘top notes’ and indiscriminate separation of high molecular weight components. It was realised at an early stage in the development of NCF extraction that these problems could be largely overcome by using C02. The main advantages in using NCF C02 for the isolation of flavours and fragrances are: (1) The extraction and separation can be carried out at low temperature in an inert environment, thereby avoiding thermal damage and chemical degradation. (2) The extract has improved solubility in formulations since less terpenes are ex- tracted. (3) The high vapour pressure of C02 enables it to be removed without losses in the highly volatile ‘top notes’. (4) Undesirable components such as proteins, waxes, sugars, chlorophyll are not ex- tracted. Supercritical fluid extraction 55 Most of the early work on the use of liquid C02 for flavour extraction was carried out in the Soviet Union in the 1960s and reported in Russian. Since then the variety of flavours and fragrances examined has grown enormously and now represents probably the largest class of researched applications for CO2 extraction in the food industry. Of all examples, that of hop oil extraction deserves special mention since it is one of the few applications regularly operating on a large commercial scale. In hop oils it is the soft resins consisting of alpha acids (humulones) and beta acids (lupulones) that are the important components for flavour production. In the brewing process alpha acids are isomerised to give beer its characteristic taste. Conventional hop extraction processes use either hexane or methylene chloride and, although residual levels of these solvents are within allowed limits, CO, offers many of the above advantages ((1)-(4)) and, being non- toxic, has a far higher level of acceptability. A process for hop extraction using SCF C02 has been patented by Vitzthum et al. (1972) and a complementary process using NCL C02 was later developed by Laws et al. (1980). The latter process uses NCL C02 at 7-10°C, where it is claimed that a solubility maximum exists. The hop extract is rich in the desirable acids and essential oils and contains no ‘hard’ resins. The process was implemented on a commercial scale at Reigate (UK). Similar plants are operating in West Germany at supercritical conditions (Miinchsmunster). Although the higher temperature of the SCF process facilitates a more rapid extraction, some selectivity is lost and more chlorophyll is also co-extracted. Hop extraction is a highly seasonal process and can be carried out only over a limited period. Most commercial hop extraction plants are therefore also used for other applica- tions out of the hop season, e.g. in the extraction of flavours, herbs and spices, and are also often offered for hire on a ‘contract’ basis. Judging by the number of publications appearing over the last decade there must be few botanical plants providing flavour and fragrance extracts that have not been subjected to C02 extraction. Most tests show an improved quality extract resembling more closely the botanical source for which quantitative evidence is often presented in the form of chromatograms. However, specific information regarding efficiency and mass transfer is often lacking. As in the case of hop extraction liquid C02 is found to give a higher quality extract, but this is probably offset by a less efficient extraction. A selection of relevant literature is given in Table 2.4. NCF extraction has also been used to extract flavours and aromas from a variety of other sources including fruit juices (Schultz and Randall, 1970), wine (Jolly, 1981), tea (Vitzthum et al., 1975a), cocoa (Vitzthum et al., 1975b) and tobacco (Luganskaya et al., 1967). Inexperience of high-pressure technology, lack of design data and high capital costs have all contributed to limit the application of SCF extraction in the food industry. However, selected applications making use of the unique properties of SCFs have been, and are being, profitably exploited. It is noteworthy that the products of such processes, although sometimes more expensive, have found a place in the market by virtue of their improved quality. This reflects the current trends of the consumer towards purer and more natural processed foods. It seems likely that these consumer demands, combined with increasing legislative restrictions, will dictate greater implementation of SCF extrac- tion technology in the future. 56 D. Steytler Table 2.4. Examples of SCF extraction processes Plant Reference Almonds Calame and Steiner (1982) Anise Stahl and Gerard (1982) Basil Pekhov et al. (1975) Bergamot* Black pepper* camomile* Caraway Cardamon* Meerov et al. (1971) Cinnamon Clove* Coriander* Volodicheva and Lybarskii (1974) Cumin Gangadhara and Mukhopadhyay (1988) Fennel Volodicheva and Lybarskii (1974) Ginger* Hop* Moyler (1986) Horseradish Juniper* Lemon* Lime* Lovage* Mace* Marjoram* Bouclier and Koller (1986) Nutmeg* Orange* Paprika* Coenen and Hagen (1983) Parsley* Patchouli* Peppermint Shaftan et al. (1973) Pimento* Moyler (1986) Red pepper Rosemary* Sage* Sandalwood Thyme* Bestmann et al. (1985) Vanilla* Moyler (1986), Sankar and Manohar (1988), Calame and Steiner (1982), Vitzthum and Hubert (197 1) Stahl and Schilz (1976), Weust et nl. (1981) Stahl and Hubert (1976), Stahl and Gerard (1982) Stahl and Gerard (1982), Tateo and Gerard (1989) Stahl and Chizzini (1982), Moyler (1986), Gopalakrishnan et 01. (1 990) Stahl and Schilz (1976), Pellerin (1988), Sankar and Manohar (1988), Moyler (1986), Chen et (11. (1986) Stahl et 01. (1 983a or b) Calame and Steiner (1982), Temelli et nl. (1988) Hubert and Vitzthum (1978), Moyler (1986) Hubert and Vitzthum (1978) Pellerin (1988), Tateo and Fellin (1988) Stahl and Shilz (1976), Pellerin (1988) Sankar and Manohar (I 988) Shuetz (1981), Vitzthum and Hubert (1971) * Commercially available extracts. Supercritical fluid extraction 57 Symbols used in text a b C = concentration D = diffusion coefficient Deff = effective diffusion coefficient f =fugacity k = binary interaction parameter K = equilibrium constant L = half thickness of slab M, = mass of solute extracted at time t M, = mass of solute extracted at infinite time P, = critical pressure R = gas constant S = solubility Tc = critical temperature V = molar volume x y 77 = viscosity z = tortuosity factor $ = fugacity coefficient q = constant in attractive term of EOS = constant in repulsive term of EOS = mole fraction (in liquid or solid phase) = mole fraction (in gas or NCF phase) = volume fraction of pores REFERENCES Angus, S., Armstrong, B. and de Reuk, K. M. (1976) International Thermodynamic Tables of the Fluid State, Vol. 3 (CO,), Pergamon Press, Oxford. Bestmann, H. J., Erler, J. and Vostrowsky, 0. (1985) Extraktion von Thymian mit Flussig CO, im Labormastab, 2. Lebensm. Unters Forsch. 180, 49 I. Bischel, B. (1979) Diffusion phenomena during the decaffeination of coffee beans, Food Chem., 5, 53. Blisset, L. A., Friggens, G. R. and McGee, J. P. (1979) Apparatus and method for feeding coal into a coal gasifier. US patent No. 4,153,427. Bohm, F., Heinisch, Peter, S. and Widner, E. (1990) Design, Construction and Operation of a Multipurpose Plant for Commercial Supercritical Gas Extraction, ACS Sympo- sium Series 406 (Supercritical Science and Technology), p. 499. Bouclier, N. and Koller, W. D. (1986) Influence of temperature on odour and volatile ingredients of comminuted marjoram and its CO, extract. In The Shelf Life of Foods and Beverages, Proc. 4th Intern. Flavour Conference, Rhodes, Greece, 23-26 July (ed. G. Charalambous) Elsevier. Bradley, R. L. (1989) Removal of cholesterol from milk fat using supercritical carbon dioxide, J. Dairy Sci., 72 2834. 58 D. Steytler Brandani, V., Re, G. del, Giacomo, G. di and Mucciante, V. (1990) Phase equilibria of essential oil components and supercritical carbon dioxide, Fluid Phase Equilibria, 59, 135. Brogle, H. (1982), C02 in Solvent extraction - its properties and applications, Chemistry and Industry, 12, 385. Brunner, G. (1984) Mass transfer from solid material in gas extraction, Ber. Bunsenges. Phys. Chem., 88, 887. Calame, J. P. and Steiner, R. (1982) Carbon dioxide in the flavour and perfumery indus- try, Chern. and Irzd., 12, 399. Carman, P. C. and Haul, R. A. W. (1954) Measurement of diffusion coefficients, Proc. R. SOC. A, 222, 109. Chen, C.-C., Kuo, M.-C., Wu, C.-M. and Ho, C.-T. (1986) Pungent compounds of ginger (Zirzgiber oflcinale Roscoe) extracted by liquid carbon dioxide, J. Agric. Food Chenz., 34, 411. Chen, S.-H. (1983) A rough-hard-sphere theory for diffusion in supercritical carbon dioxide, Chern. Eng. Sci., 38, 655. Chrastll, J. (1982) Solubility of solids and liquids in supercritical gases, .I. Phys. Chern., 86, 1982. Christianson, D. D., Friedrich, J. P., List, G. R., Warner, K., Bagley, E. B., Stringfellow, A. C. and Inglett, G. E. (1984) Supercritical fluid extraction of dry-milled corn germ with carbon dioxide, J. Food Science, 49, 229. Coenen, H. and Hagen, R. (1983) Naturlicher Farbstoff aus Paprika durch Gasextraktion, Gordiarz, 83, 164. Consani, K. A. and Smith, R. D. (1990) Observations on the solubility of surfactants and related molecules in carbon dioxide at 5OoC, J. Supercrit. Fluids, 3, 51. Crank, J. (I 989) The Mathernatics ofDifSusion, Oxford University Press, Oxford. ' Dandge, D. K., Heller, J. P., and Wilson, K. (1985) Structure solubility correlations: organic compounds and dense carbon dioxide binary systems, Ind. Eng. Chem., 24, 163. de Azevedo, E. G., Matos, H. A., da Ponte, M. N. and Simoes, P. C. (1988) Phase equilibrium of systems containing limonene, cineole and SCF C02, Proc. Intern. Syrnp. Supercritical Fluids (ed. Perrut, M.), SociCtC FranGaise de Chimie, Vol. 1, p. 135. Dieters, U. (1982) A new semiempirical equation of state for fluids - 111, Application to phase equilibria in binary mixtures, Chenz. Eng. Sci., 37, 855. Eastoe, J., Robinson, B. H. and Steytler, D. C. (1990a) Influence of pressure and tem- perature on microemulsion stability, J. Chern. SOC. Faraday, 86, 51 1. Eastoe, J., Young, W. K., Robinson, B. H. and Steytler, D. C. (1990b) Scattering studies of microemulsions in low-density alkanes, J. Chern. SOC. Faraday, 86, 2883. Ebelling, H. and Franck, E. U. (1984) Spectroscopic determination of caffeine solubility in supercritical carbon dioxide, Ber. Bunsenges. Phys. Chern., 88, 862. Eggers, R. and Tschiersch, R. (1980) Development and design of plants for high pressure extraction of natural products. In Extraction with Supercritical Gases, (eds. Schneider, G. M., Stahl, E. and Wilke, G.), Verlag Chemie, p. 165. Supercritical fluid extraction 59 Eisenbach, W. (1984) Supercritical fluid extraction: a film demonstration, Ber. Bunsenges. Phys. Chem., 88, 882. Erlangen (1984) Intern. Symp. High Pressure Chemical Engineering, Erlangen, 1984. Verein Deutscher Ingenieure VDI Gesellschaft Verfahrentechnik und Chemieingienieurwesen. Feist, R. and Schneider, G. M. (1982) Determination of binary diffusion coefficients of benzene, phenol, naphthalene and caffeine in supercritical C02 between 308 and 333 K in the pressure range 80 to 160 bar with supercritical fluid chromatography (SFC), Sep. Sci. and Technol., 17,261. Francis, A. W. (1954) Ternary systems of liquid carbon dioxide, J. Phys. Chem., 58, 1099. Fredenslund, A,, Jones, R. L. and Prausnitz, J. M. (1975) Group-contribution estimation of activity coefficients in non-ideal liquid mixtures, AIChE J., 21, 1086. Friedrich, J. P. and Pryde, E. H. (1984) Supercritical C02 extraction of lipid-bearing materials and characterisation of the products, J. Am. Oil Chem. SOC., 61, 223. Friedrich, J. P., List, G. R. and Heaking, A. J. (1982) Petroleum-free extraction of oil from soybeans with supercritical CO,, J. Am. Oil Chem. SOC., 59, 288. Froning, G. W., Wehling, R. L., Cuppett, S. L., Peirce, M. M., Neimen, L. and Siekman, D. K. (1990) Extraction of cholesterol and other lipids from dried egg yolk using supercritical carbon dioxide, J. Food Sci., 55, 95. Gale, R. W., Fulton, J. L. and Smith, R. D. (1987) Organized molecular assemblies in the gas phase; reverse micelles and microemulsions in supercritical fluids, J. Am. Chem. SOC., 109, 920. Gangadhara Rao, V. S. and Mukhopadhyay, M. (1988) Mass transfer for supercritical fluid extraction of spices, Proc. Intern. Symp. Supercritical Fluids, Perrut, M. (ed.), SociCtC FranGaise de Chimie, Vol. 2, p. 643. Gopalakrishnan, N., Shanti, P. P. V. and Narayanan, C. S. (1990) Composition of clove (Syzygiunz aromaticum) bud oil extracted using carbon dioxide, J. Sci. Food Agric., 50, 111. Hannay, J. B. and Hogarth, J. (1879) On the solubility of solids in gases, Proc. R. SOC. London, 29, 324. Haselow, J. S., Han, S. J., Greenkorn, R. A. and Chao, K. C. (1986) Equations of state for supercritical extraction, ACS Symp. Series 300, 156. Higuchi, T. (1963) Mechanism of sustained action medication. Theoretical analysis of rate of release of drugs dispersed in solid matrices, J. Pharm. Sci., 52, 1145. Higuchi, W. I. and Roseman, T. J. (1970) Release of medroxyprogesterone acetate from a silicone polymer, J. Pharm. Sci., 59, 353. Hildebrand, J. H. and Scott, R. L. (1950) The SolubiliQ of Non-electrolytes, 2nd edn., Prentice-Hall, Englewood Cliffs, N.J. Hubert, P. and Vitzthum, 0. G. (1978) Liquid extraction of hops, spices and tobacco with supercritical gases, Angew. Chenz. Intl. Ed., 17, 7 10. Jay, A. J. and Austen, A. (1990) A spectrophotometric study of solvation in supercritical fluids. Presented at the 2nd Nottingharn-Leeds Discussion Forum and Exhibition on Supercritical Fluids, University of Leeds. 60 D. Steytler Jay, A. J., Knights, M. K. and Steytler, D. C. (1991) Spectrophotometric study of food colours in near-critical C02, J. Supercrit. Fluids, 4, 131. Johnston, K. P., Reid, N. K., Barry, S. E. and Holcomb, T. R. (1987) Separation of isomers using retrograde cry stallisation from supercritical fluids, Znd. Eng. Chem. Res.,, 26, 1987. Johnston, K. P., Sunwook, K. and Combes, J. (1990) Spectroscopic determination of solvent strength and structure in supercritical fluid mixtures: a review, ACS Symp. Series 406 (Supercritical Science and Technology), p. 52. Jolly, D. R. P. (1981), Wine flavour extraction with liquid carbon dioxide, Process Biochemistry, 16 (5), 36. Kamihira, M., Taniguchi, M. and Kobayashi, T. (1987) Sterilisation of microorganisms with supercritical carbon dioxide, Agric. Biol. Chem., 51, 407. Kanare, V., Antila, V., Harvala, T. and Komppa, V. (1989) Extraction of milk fat with supercritical carbon dioxide, Milchwissenschaf, 44, 407. Kaufmann, W., Bienorth, G., Frede, E., Merk, W., Precht, D. and Timmen, H. (1982) Fractionation of butterfat by extraction with supercritical C02, Milchwissenscsharft, 37, 92. King, M., Kassim, K. and Bott, T. R. (1983a) Mass transfer into near-critical extractants, Fluid Phase Equilibria, 10, 249. King, M. B., Alderson, D. A., Fallah, F. H., Kassim, D. M., Kassim, K. M., Sheldon, J. R. and Mahmud, R. S. (1983b) Some vapour liquid and vapour/solid measurements of relevance for supercritical extraction operations and their correlation. In Chemical Erzgineerirzg at Supercritical Fluid Conditions, Paulaitis, M. E., Penninger, J. M. L. and Gray, R. D. (eds.), Ann Arbor, MI. p. 31. King, M., Bott, T. R., Barr, M. J., Mamhud, R. S. and Saunders, N. (1987) Equilibrium and rate data for the extraction of lipids using compressed carbon dioxide, Sep. Sci. and Tecrzol., 22, 1103. Klein, T. and Schultz, S. (1989) Phase equilibria in mixtures of glycerides and carbon dioxide and applications of continuous thermodynamics to mixtures and rapeseed oil and carbon dioxide, Fluid Phase Equilibria, 50, 79. Korner, J. P. (1988) New developments in the design and construction of industrial-size SCGE plants, Proc. Intern. Symp. Supercritical Fluids, Perrut, M. (ed.), SociCtC Franiaise de Chimie, Vol. 2, p. 633. Krukonis, V. J. (1988) Supercritical fluid processing: current research and operations, Proc. Intern. Synzp. Supercritical Fluids, Perrut, M. (ed.), Socittt Franiaise de Chimie, Vol. 2, p. 541. Kwak, T. Y. and Mansoori, G. A. (1986) Van der Waals mixing rules for cubic equations of state. Applications for supercritical fluid extraction modelling, Chem. Eizg. Sci., 41, 1303. Laws, D. R. J., Bath, A. and Wheldon, A. G. (1980) Hop extraction with carbon dioxide. US Patent 4,2 18,49 1. List, G. R. and Friedrich, J. P. (1985) Processing characteristics and oxidative stability of soybean oil extracted with supercritical carbon dioxide at 50°C and 8000 psi, J. Am. Oil Chem Soc., 62, 82. Supercritical fluid extraction 61 Lovell, D. (1988) Private communication. Luganskaya, L. N., Krasnokutskaya, E. B. and Yasinskaya, L. B. (1967) Use of tobacco dust extract for flavouring tobacco, Tabac (Moscow), 28, 30. McFann, G. J. and Johnston, K. (1991) Phase behaviour of AOT microemulsions in compressible liquids, J. Phys. Chem., 95,4889. McHugh, M. A. and Krukonis, V. J. (1986) In Supercritical Fluid Extraction Principles and Practice, Butterworths, London. Marentis, R. T. and Vance, S. W. (1990) Selection of components for commercial super- critical food processing plants, ACS Symp. Series 406 (Supercritical Science and Tech- nology), p. 525. Margolis, G. and Chiovini, J. (1981) Decaffeination process. US Patent 4,251,559. Mathias, P. M. and Copeman, T. W. (1983) Extension of the Peng-Robinson equation of state to complex mixtures: evaluation of the various forms of the local composition concept, Fluid Phase Equilibria, 13, 91. Meerov, Ya, S., Popova, S. A. and Ponomarenk, I. Ya. (1971) Quality of cardamon oil obtained by various methods, Pischchevoi Promyshennosti, 5, 203. Moulson, P. M. (1988) Private communication. Nilsson, W. B., Gauglitz, E. J., Hudson, J. K., Stout, V. K. and Spinelli, J. (1988) Fractionation of menhaden oil ethyl esters using supercritical C02, J. Am. Oil Chem. Soc., 55, 109. Ohgaki, K., Kobayashi, H., Kataysma, T. and Hirokawa, N. (1990) Whisker formation from jet of supercritical fluid solution, J. Supercritical Fluids, 3, 103. Paulaitis, M. E., Krukonis, V. J., Kurnick, T. and Reid, R. C. (1983) Supercritical fluid extraction, Rev. Chem. Eng., 1, 179. Pekhov, A. V., Kasyanov, G. I. and Klimova, E. S. (1975) Carbon dioxide extraction of Ocimutn Gratissimum, Maslo-Zhir. Prom-st., 8, 23. Pellerin, P. (1988) Aromatic natural raw materials extraction by liquid or supercritical C02, Proc. Intern. Symp. Supercritical Fluids, Perrut, M. (ed.), Socittt FranGaise de Chimie, Vol. 2, p. 677. Peng, D. Y. and Robinson, D. B. (1976) A new two-constant equation of state, Ind. Eng. Chem. Fundam., 15, 59. Peter, S., Schneider, M., Weidner, E. and Zeigeltz, R. (1989) The separation of lecithin and soya oil in a countercurrent column by near critical fluid extraction, Chem. Eng. Technol., 10, 37. Prasad, R., Gottesman, M. and Scarella, R. A. (1981) Decaffeination of aqueous extracts. US Patent 4,246,29 1. Prausnitz, J. M. (1965) Solubility of solids in dense gases. US National Bureau of Stand- ards Technical Note 3 16, US Printing Office, Washington, D.C. 20402. Prausnitz, J. M. and Benson, P. R. (1959) Solubility of liquids in compressed hydrogen, nitrogen and carbon dioxide, AIChE, 5, 161. Prausnitz, J. M., Litchenthaler, R. N. and de Azevedo, E. G. (1986) Molecular Thermody- namics of Fluid Phase Equilibria, 2nd end, Prentice-Hall, Englewood Cliffs, NJ. Redlich, 0. and Kwong, J. N. S. (1948) On the thermodynamics of solutions. An equa- tion of state. Fugacities of gaseous solutions, Chem. Rev., 44, 223. 62 D. Steytler Reid, R., Prausnitz, J. M. and Poling, B. E. (1987) In The Properties of Gases and Liquids, 4th edn, McGraw-Hill, New York. Rizvi, S. S. H., Benado, A. L., Zollweg, J. A. and Daniels, J. A. (1986) Supercritical fluid extraction: fundamental principles and modelling methods, Food Technol., June, 55. Roselius, L., Vitzthum, 0. and Hubert, P. (1974) Method for the production of caffeine- free coffee extract. US Patent 3, 843, 824. Roselius, L., Kurzhals, H.-A,, Sylla, K. F. and Hubert, P. (1979) Process for extracting stimulants from coffee, US Patent 4,168,324. Sankar, K. U. (1989) Studies on the physicochemical characteristics of volatile oil from pepper (Piper Nigr~m) extracted by supercritical carbon dioxide, J. Sci. Food Agric., 48, 483. Sankar, K. U. and Manohar, B. (1988) Extraction of essential oils using supercritical carbon dioxide, Proc. Intern. Symp. Supercritical Fluids, Perrut, M. (ed.), SociCtC Franpiise de Chimie, 2, p. 807. Scheutz, E., Vollbrecht, H. R., Sander, K., Sand, T. and Muehlnickel, P. (1 98 1) Extrac- tion of flavouring from the vanilla pods. European Patent No. 75, 134. Schneider, G. M. (1970) Phase equilibria in fluid mixtures at high pressure, Adv. Chein. PI7ps., 17, 1. Schultz, W. G. and Randall (1970) Liquid carbon dioxide for selective aroma extraction, Food Technol., 24, 94. Schults, W. G., Schultz, T. H., Carlson, R. A. and Hudson, J. S. (1974) Pilot plant extraction with liquid carbon dioxide, Food Technol., 28, 32. Shaftan, E. A., Laminina, N. V. and Senich, V. Ya (1973) Determination of menthol in essential oil and a carbon dioxide extract of peppermint, Maslo-Zhir Prom., 11, 28. Shmitt, W. J. and Reid, R. C. (1986) The use of entrainers in modifying the solubility of phenanhtrene and benzoic acid in supercritical carbon dioxide and ethane, Fluid Phase Equil., 32, 77. Spiro, M., Selwood, R. M. and Page, C. (1984) The kinetics and mechanism of caffeine infusion from coffee, J. Sci. Food Agric., 35, 915-925. Stahl, E. and Gerard, D. (1982) High pressure extraction of natural substances with supercritical and liquefied gases. 9. Methods for obtaining essential oils under mild conditions, Parfiein. Koaner., 63, 117-124. Stahl, E. and Gerard, D. (1983) Pnrfiem. Kosnzet., 65, 237. Stahl, E. and Schilz, W. (1976) Extraction with supercritical gases in direct coupling with thin layer chromatography. Applications to natural product chemistry, Chem lng. Tech., 48, 773. Stahl, E. and Stadler, R. (1984) Systeme zum ein oder Austrag von Schuttgutern in oder BLIS Druckraumen, Chein. Ing. Tech., 56, 755. Stahl, E., Schilz, W., Schutz, E. and Willing, E. (1980) A quick method for the microanalytical evaluation of the dissolving power of supercritical gases. In Extraction with supercritical gases, Schneider, G. M., Stahl, E. and Wilke, G. (eds.), Verlag Chemie. Stahl, E., Quirin, K. W. and Gerard, D. (1983a) Solubilities of soybean oil, jojoba oil and cuticular wax in dense carbon dioxide, Fette Siefen Anstrichmittel, 85, 458. Supercritical fluid extraction 63 Stahl, E., Quirin, K. W. and Blagrove, R. J. (1983b) Extraction of seed oils with super- critical carbon dioxide, J. Agric. Food Chem., 32, 938. Stahl, E., Quirin, K. W., Glatz, A,, Gerard, D. and Rau, G. (1984) New developments in the field of high pressure extraction of natural products with dense gases, Ber. Bunsenges. Phys. Chem., 88, 900. Steytler, D. C. (1988) Presented at Leatherhead Food RA Training Course and Sympo- sium T167, Advances in Separation Technology for the Food Industry. Steytler, D. C., Moulson, P. M., Clark, S. A. and Parker, M. L. (1990) Network recrystallisation in a supercritical fluid, J. Phys. Chem, 94, 8748. Steytler, D. C., Moulson, P. M. and Reynolds, J. (1991) Biotransformations in near- critical carbon dioxide, Enzyme Microb. Technol., 13, 193. Sun, C. K. J. and Chen, S. H. (1985) Diffusion of benzene, toluene, naphthalene and phenanthrene in supercritical dense 2,3-dimethylbutane, AlChE J., 31, 1904. Taniguchi, M., Suzuki, H., Sato, M. and Kobayashi, T. (1987a) Sterilisation of plasma powder by treatment with supercritical carbon dioxide, Agric,. Bid. Chenz., 51, 3425. Taniguchi, M., Kamihira, M. and Kobayashi, T. (1987b) Effect of treatment with super- critical carbon dioxide on enzyme activity, Agric. Biol. Chenz., 51, 593. Tateo, F. and Chizzini, F. (1989) The composition and quality of supercritical CO, extracted cinnamon, J. Ess. Oil Res., 1, 165. Tateo, F. and Fellin, M. (1988) Production of Rosemary Oleoresin using supercritical carbon dioxide, Perfumer and Flavorist, 13, 27. Tavana, A. and Randolph, A. D. (1989) Manipulating solids CSD in a supercritical fluid crystalliser: C02-benzoic acid, AZChE J., 35, 1625. Temelli, F., Chen, C. S. and Braddock, R. J. (1988) Supercritical fluid extraction in citrus oil processing, Food Technol., June, 145. Tolboe, O., Hansen, I. R. and Shukla, V. K. S. (1988) Extraction of oil from evening primrose seed, Proc. Intern. Symp. Supercritical Fluids, Perrut, M. (ed.), SociCtC FranGaise de Chimie, Vol. 2, p. 685. Tongue, H. (1959) The Design and Construction of High Pressure Chemical Plant, Chapman and Hall, London. Tsiklis, D. S. (1968) Handbook of Techniques in High-pressure Research arid Engineer- ing, Plenum Press, New York. van der Waals, J. D. ( 1873) Over de Cuntinuitet can der Gas en Vloeistoftoes Tana. PhD Dissertation. University of Leiden, Holland. van Eijs, A. M. M., de Long, J. P. L., Doddema, H. J. and Lindeboom, D. R. (1988) Trans-esterification in supercritical C02, Proc. Intern. Sy7zp. Supercriticul Fluids, Perrut, M. (ed.), SociCtC Franpise de Chimie, Vol. 2, p. 933. van Leer, R. A. and Paillaitis, M. E. (1980) Solubilities of phenol and chlorinated phenols in supercritical carbon dioxide, J. Chem. Eng. Data, 25, 257. Vitzthum, 0. G. and Hubert, P. (1971) Process for liquid C02 extraction of flavours. West German Patent No. 21,27,611. Vitzthum, 0. G. and Hubert, P. (1979) Method for the manufacture of csffeine-free black tea. US Patent No. 4,167,589. Vitzthum, 0. G., Hubert, P. and Stirl, W. (1972) Hop extracts. Canadian Patent 987,250. 64 D. Steytler Vitzthum, 0. G., Werkhoff, P and Hubert, P. (1975a) New volatile constituents of black tea aroma, J. Agric. Food Chem., 23, 999. Vitzthum, 0. G., Werkhoff, P. and Hubert, P. (1975b) Volatile constituents of roasted cocoa. Basic fraction, J. Food Sci., 40, 91 1. Volodicheva, L. F. and Lybarskii, L. N. (1974) Liquid C02 extraction of coriander, caraway seed, aniseed and fennel, Vses. Zaochny Inst. Pishevoi. Promyshlennosti, 9, 11. Weber, J. K P. (1980) Effect of supercritical carbon dioxide on proteins, Z. Lebensm. Unters. Forsch, 171, 95. Weust, R., Pfeiffer, H. and van der Mei, H. (1981) Plant extracts with sensory properties. European Patent No. 62,893. Wiebe, R. and Gaddy, V. L. (1941) Vapour phase composition of carbon dioxide - water mixtures at various temperatures and pressures to 700 atmospheres. J. Am. Chem. SOC., 63, 475. Zosel, K. (1974) Process for recovering caffeine. US Patent No. 3,806,619. Zosel, K. (1978) Separation with supercritical gases: practical applications, Angewente Zosel, K. (1981) Process for the decaffeination of coffee. US Patent No. 4,247,570. Chenzie (Int. Ed.), 17, 702.