Ind. Eng. Chem. Res. 1993,32, 1471-1481
1471
Solubility of Naproxen in Supercritical Carbon Dioxide with and without Cosolvents Simon S. T. Ting, Stuart J. Macnaughton, David L. Tomasko,t and Neil R. Foster’ School of Chemical Engineering and Industrial Chemistry, University of New South Wales, P.O.Box 1, Kensington, N.S. W . 2033, Australia
The solubility of naproxen ((S)-6-methoxy-c~-methyl-2-naphthaleneacetic acid) in supercritical COz was determined at 313.1, 323.1, and 333.1 K. The influence of six polar cosolvents, ethyl acetate, acetone, methanol, ethanol, 1-propanol, and 2-propanol, was studied at concentrations of 1.75,3.5, and 5.25 mol % . The solubility enhancement with these cosolvents is considerable, and the cosolvent effect increases in the order ethyl acetate, acetone, methanol, ethanol, 2-propanol, 1-propanol. A nonlinear increase in solubility is observed with an increase in cosolvent concentration. The use of the Peng-Robinson and Soave-Redlich-Kwong equations of state to correlate these ternary systems requires the use of negative binary interaction parameters indicating strong interactions between naproxen and the cosolvents. The cosolvent effects cannot be explained by any one physical property of the cosolvents but appear to be influenced by hydrogen bonding ability as determined from solvatochromic parameters as well as the relative distance from the CO2-cosolvent binary critical point.
Introduction In recent years, a great deal of research has been carried out in the field of supercritical fluid (SCF) technology. The interest in using this technology for selective extraction or reaction is due to the superior properties that are inherent to this class of fluid, including the ability to vary solvent density and to effect a change in solvent properties by changing either the pressure or temperature. The viscosity of a SCF is much lower than a liquid, and diffusivitiescan vary between gaslike and liquidlikevalues. As a result, extraction processes can be carried out more rapidly. Another advantage of using SCF’s in separation processes is the relative ease of solvent recovery and the separation of the desirable product(s). For most high molecular weight, nonvolatile organic compounds, the solubility in SCFs is low requiring high temperatures and pressures for substantial loadings. Thus the capital cost for commercial-scale processes can be prohibitive, and this has been one of the major hindrances to the advance of SCF technology. Carbon dioxide is one of the most common gases used as a SCF mainly because it is an easy gas to handle, it is inert and nontoxic, it is nonflammable, and it has a convenient critical temperature. Although COz is the most common SCF being used, it does have limitations as a result of its lack of polarity and associated lack of capacity for specific solvent-solute interactions which would lead to high loading and/or selectivity for polar organic compounds. Pure SCFs exhibit polarization behavior that is primarily related to the density and SC CO2 at 308 K and 200 bar only has a solubilityparameter approachingthat of liquid isopentane (6.5 (cal/cm3)1/2);thus there is a great incentive to improve ita polarity. It has been found that the addition of a small amount of cosolvent to a SCF can have dramatic effects on its solvent power. In recent years, progress has been made toward understanding the interactionsinvolved in dilute supercritical mixtures. It has been shown that near the critical point of a SCF solution, the solvent molecules %luster” around
* To whom correspondence should be addressed.
Present address: Department of Chemical Engineering, The Ohio State University, Columbus, OH 43210-1180. t
the relatively large solute molecule to form a local density higher than the bulk density (Eckert et al., 1986;Kajimoto et al., 1988; Cochran and Lee, 1989; Debenedetti, 1987; Debenedetti et al., 1989; Petsche and Debenedetti, 1989, 1991;Brennecke et al., 1990;Morita and Kajimoto, 1990). When a cosolvent is added, the situation is further complicated by the differences in local and bulk compositions (Kim and Johnston, 1987a; Yonker and Smith, 1988). Frye et al. (1987) also indicated a change in composition of the cybotacticregion (aregionin the vicinity of the solute) in a cosolvent-modified SCF. The increase in solubility due to the addition of cosolvent is the result of additional interactions between the solute and the cosolvent. Considering the interactions possible, these cosolvent effects could be the result of several mechanisms. The addition of a cosolvent will generally increase the mixture density which will contribute to the overallsolubilityenhancement as will physical interactions like dipole-dipole, dipole-induced dipole, and induced dipole-induced dipole interactions. However, when using a polar cosolvent for polar solutes, the largest increase in solubility would be expected to be a result of specific chemical interactions like hydrogen bonding or charge transfer complex formation. There are relatively few reported studies of solid-SCF cosolvent solubility to date (Dobbs et al., 1986; Schmitt and Reid, 1986; Wong and Johnston, 1986; Van Alsten, 1986;Larson and King, 1986;Dobbs et al., 1987;Schaeffer et al., 1988; Lemert and Johnston, 1989,1991;Tavana et al., 1989; Cygnarowizc et al., 1990; Hollar and Ehrlich, 1990;Smith and Wormald, 1990;Gurdial et al., 1993;Ekart et al., 1992). Overall, only two SCFs have been used, ethane and COz-whereas a multitude of cosolvents have been used ranging from nonpolar gases to polar liquids. In this work, the flow technique coupled with gravimetric analysis was used to measure the solubility of naproxen, (S)-6-methoxy-a-methyl-2-naphthaleneacetic acid (a nonsteroidal antiinflammatory drug), in pure SC COz and also in various SC COz-cosolvent mixtures. The cosolvents chosen were all polar and could either exhibit selfassociation (alcohols) or not (ketone and ester). For all the cosolvents studied, three concentrations ranging from 1.75to 5.25 mol 5% were investigated at 333.1 K. This was to enable the study of the effect of concentration together
0888-588519312632-1471$04.00/0 0 1993 American Chemical Society
1472 Ind. Eng. Chem. Res., Vol. 32, No. 7, 1993 Table I. Source and Purity of Materials compound source purity naproxen carbon dioxide acetone ethyl acetate methanol ethanol 1-propanol 2-propanol
Sigma Chemicals 99+ % Liquid Air liquid withdraw grade, 99.8+ % BDH Aldrich BDH BDH BDH BDH
HiPerSolv grade, 99.8% by HPLC 99.9+ % by GLC HiPerSolv grade, 99.8% by GLC AnalaR grade, 99.7 % v/v HiPerSolv grade, 99.8% by GLC HiPerSolv grade, 99.8% by GLC
with the functionality of the cosolvent. Experiments were also carried out a t 323.1 and 318.1K with acetone cosolvent and 323.1 K with methanol, ethanol, and 2-propanol cosolvents. Materials The sources and purities of the various compounds used are given in Table I. These materials were used without further purification. Experimental Section Binary System. A schematic diagram of the equipment used is shown in Figure 1. The syringe pump used was an Isco Model 260D, with constant pressure operating capability, equipped with an external jacket for heating or cooling purposes. In the study of the solubility of naproxen in pure SC C02, the equilibrium cells consisted of two 6-in. by 0.5-in. 0.d. stainless steel tubes and a Jerguson sight gauge. For the cosolventstudiesa slight modification was made to the overallequipment setup used for solubility measurements in pure C02. Because of the anticipated higher solubility involved, the equilibrium cells were replaced with a 300-mL bomb half-filled with naproxen. The system temperature was monitored by a platinum resistance thermometer accurate to f O . l K, and the system pressure was measured by a Druck pressure transducer (Model TJE), with an accuracy of f 5 psi, located just after the sight gauge. The equilibrium cellsand sight gauge were placed in a water bath which was regulated to f O . l
K. The equilibrium cells were packed with naproxen, and each end was plugged with glass wool to prevent the fine naproxen powder from plugging the smaller 1/8-in. 0.d. interconnecting stainless steel tubing. Similarly, the sight gauge was three-quarters filled with naproxen and also plugged loosely with glass wool to prevent entrainment. The sight gauge provided a means of determining the physical state of the mixture (i.e., to detect potential melting of the solid). A 7-pm Nupro inline filter, F1, was placed after the pressure transducer to prevent any further entrainment of solid particles of naproxen. The pressure drop through the saturators was less than 0.5 bar. The method used in this study is similar to that used by Gurdial and Foster (1991). Initially, the system was purged with carbon dioxide at low pressure and then brought up to the required system pressure and temperature. After equilibrating for several hours, the system was purged with SC C02 equivalent to 520 cm3 or two syringes of liquid C02 at room temperature and system pressure which corresponded to at least three complete system volumes. For operation, the metering valve, V5 (Whitey 32RS4 with lubricant removed), was first closed and V4 was slowly opened and the system was allowed to equilibrate further at system pressure and temperature for 15-30 min. The experiment was then started by opening valve V5, which was heated by a 100-W lamp.
The flow rate was normally maintained to within 10dm3/h C02 at ambient conditions. Consistent results could still be obtained when this flow rate was halved indicating that equilibrium was achieved. The solute which precipitated on expansion through valve V5 was collected in a 0.5-pm Nupro inline fiiter, F2. The total volume of CO2 at ambient conditions, after passing through a water saturator, was measured by a wet test meter (Type DM3A, Alexander Wright & Co.). At the end of each run, V4 was closed and the section between V4 and V5 was allowed to depressurize through V5. The valve V4 was located outside the constanttemperature water bath so that V5 together with the section of tubing connecting to V4 could be disconnected. As V4 was located outside the water bath, its temperature was controlled with a heating tape to that of the bath temperature. The solute collected in the valve and the connecting tubing was flushed with high-purity acetone (99% or better) into a Petri dish. The acetone was then evaporated until a constant mass was obtained. The Petri dish and the filter were then weighed and the mass difference was recorded. The reproducibility and uncertainty of the solubility data obtained using this method were within f 5 5%. Ternary Systems. To ensure that the solvent-cosolvent mixtures were supercritical at the chosen operating conditions, the critical locus for each system was determined for the concentration range of interest. A rigorous technique to determine the critical locus is via a vaporliquid equilibrium (VLE) experiment. However, the method proposed by Gurdial et al. (1991a,b) provides a quick technique for the determination of critical loci of binary mixtures. The critical loci for CO2-acetone, C02methanol, C02-l-propano1, and CO2-2-propanol have been established using this technique (Gurdial, 1991a,b). As no CO2-ethyl acetate critical locus data in the concentration range of interest were available, the critical locus was determined using the above method. The data obtained are shown in Figure 2. As expected with these dilute systems, the variations of critical temperatures and pressures are linear with composition. It can be seen that for the ethyl acetate402 system, a t 5.25 mol % ethyl acetate, the critical temperature and pressure are approximately 330 K and 97 bar, respectively, and thus all work done at this concentration was carried out above these conditions. To prepare the cosolvent mixtures, the barrel of the syringe pump was used as a mixing bomb. The syringe volume was calibrated with N2 at ambient temperature and at various pressures and was found to be 265 f 5 cm3. The maximum volume readout on the pump was found to be reliable and was close to the calibrated volume. The mixture was prepared by raising the head of the piston as far up as possible and then purgingwith COZ. The required amount of cosolvent was injected directly into the pump via T1 as shown in Figure 1. The barrel of the pump was then cooled by circulating chilled water through the water jacket. The three-way valve, V2, was switched to theliquid CO2 cylinder, and at the same time the piston was drawn down. The temperature was set at 274 K primarily because at this temperature and around 50-60 bar (Cot cylinder pressure) C02 exists only as a liquid. A secondary consideration was that, at this temperature, the density of liquid C02 is not too sensitive to small variations in pressure (f5 bar). With these parameters set, the required amount of COZto be added could be determined by setting the pressure. No account was made for excess volume of mixing. When the desired pressure (e.g., 52 bar) was
Ind. Eng. Chem. Res., Vol. 32, No. 7,1993 1473
T m w m Contd!ed W Wh
Figure 1. Flow apparatus for solubility measurements in pure and cosolvent modified SCF's.
H,C Figure a. Structure of naproxen.
i 75 0
1
2 3 4 MOLE % ETHYL ACETATE
5
6
70
Figure 2. Binary critical locus for the ethyl acetate402 system.
reached, V1 was closed and warm water (313.1-323.1 K) was circulated through the outside water jacket to provide thermal mixing of the solvent mixture. The mixture was allowed to equilibrate for about 15 min and then cooled and reheated to enhance the mixing process. The homogeneity of the COracetone mixture was checked by using an UV detector equipped with a high pressure flow cell. The solvent mixture was prepared as described above, and the whole syringe was pumped through the UV detector to check for the consistency of the baseline. Our results showed a reasonably stable baseline with respect to the volume of the solvent mixture paaaed through, indicatinga homogeneoussolventmixture along the length of the syringe. This was further confirmed by the reproduciblity of the solubility data. Prior to each change in cosolvent concentration and each change over to an entirely new cosolvent, the whole system was purged with at least two syringes of the cosolvent mixtures at the required conditions to ensure consistent results. The procedure was similar to that stated earlier. However, more care was taken to ensure any collected cosolvent was removed from the filter, F2. This was done by placing the filter and the Petri dish containing the solute, cosolvent, andacetone in avacuum oven. Although the solute was not analyzed, no change in appearance was observed after depressurization which implies that none of the cosolvents chemically reacted with naproxen. For these ternary systems, the reproducibilities were slightly better than with the pure studies because of the higher solubilities involved.
Choice of Cosolvents The choice of cosolvents used was based on availability in high purity, toxicology, and physical and chemical
characteristics. The functionality of the cosolvents was chosen such that they might interact differently with naproxen, whose structure is shown in Figure 3. As naproxen has an acid group, it was expected that interactions with cosolventa via hydrogen bonding might play an important role in solubility enhancement. Thus all the cosolvents chosen in this study have hydrogen bond accepting capability. Various workers have provided methods for cosolvent selection. Sun01et al. (1985) divided solvents into various classes accordingtotheir potentialtoformhydrogenbonds. These workers also listed the likelihood of hydrogen bond formation when two separate classesof solvent were mixed, andcosolventwaschosen basedon this. Walshet al. (1987, 1989) used a similar concept for choosing cosolvents for SCF systems. Tavana et al. (1989) used the ability of the cosolvent to reduce retention time of the solute in packed GC columns as a method for scanning potential cosolvents. In this work, the Kamlet-Taft solvatochromic solvent scale of acidity (a),basicity (B), and polarity/polarizability (r*) (Kamlet and Taft, 1976a,b; Kamlet et al., 1977,1983)was one tool used as a measure of hydrogen bonding capability. Perhaps a more quantitative measure of solvent power is the Hilderbrand solubility parameter (Hildebrand and Scott, 1950). This parameter can be partitioned into a dispersion (6d), polar (Q, and hydrogen bonding (6h) components (Hansen, 1967a,b, 1969; Hansen and Beerbower, 1971) which again provides a convenient tool for classifying solvent strength. The Kamlet-Taft a,8, and T* along with the Hansen sh for the ewolvents used are given in Table 11. The dipole moment for the various cosolvents are also included in Table 11. The dipole moment largely determines the orientation of a solvent around an organic solute molecule (Keesomforces)in the absence ofspecificsolutesolvent interactions. In turn, the dissolving power of a solvent also depends on the effectiveness of this electrostatic solvation. The polarizability a* of neighboring molecules is fundamental in accounting for the strength of both Debye and London forces between them. The a* values for all the cosolvents and naproxen were estimated and are listed in Table 11. However, the effectivenessof these attraction forces also dependson molecular size as suggested by Grant
1474 Ind. Eng. Chem. Res., Vol. 32, No. 7, 1993 Table XI. Solvatochromic and Solubility Parameters for All Compounds 6d (MPaW ~~
cosolvent acetone ethyl acetate methano1 ethanol 1-propanol 2-propanol naproxen
+a
0.71 0.55 0.60 0.54 0.52 0.48
aa
0.06 0.00 0.93 0.83 0.78 0.76
p 0.48 0.45 0.62 0.77
a*C(cm3 X loz6) 64.1 88.3 32.3 51.2 69.5 69.9 252.5 27.4
(D) 2.9 1.9
pb
1.7
1.7 1.7 1.7
0.95
coz
0
a*/D
u (cm3/mol)
8d
8,
bh
0.0525 0.0543 0.0485 0.0528 0.0559 0.0550 0.0850
74.0 98.5 40.7 58.5 75.2 76.8 178.3*
15.5 15.8 15.1 15.8 16.0 15.8 l0.U
10.4 5.3 12.3 8.8 6.8 6.1 6.98
7.0 7.2 22.3 19.4 17.4 16.4 20.oh
6taw 20.0 18.1 29.6 26.5 24.5 23.5 23.4e
0 Kamlet et al. (1983). b Reid et al. (1988). c Estimatedusingeq 2.132-3, Grant and Higuchi (1990). d Barton (1983). e Fedors group contribution method (Fedors, 1974). f Koenhen and Smolders (1975). 1 Group molar attraction constants (Hoy, 1970). 6h = (6t2 - 8d2-6p2)1/2 (Hansen, 1971).
Table 1x1. Solubility of Naproxen in P u r e SC COZ mole fraction naproxen x 106 press. (bar) 313.1 K 323.1 K 333.1 K 0.20
89.6 100.0 110.3 124.1 131.0 137.9 144.8 151.7 165.5 172.4 179.3 193.1
0.83
0.19 0.43
1.29
1.20
press. (bar) 89.6 96.5 110.3 124.1 137.9 151.7 165.5 179.3 193.1
0.70 1.08 1.77 1.56 2.33
1.72 2.08
2.32
2.43
2.91
Table IV. Solubility of Naproxen in SC COt with Acetone Cosolvent mole fraction naproxen x 106
2.71 3.18
a
0.005
8g 9
2I
0.002
0.001
0.0005
0.0003 0.0002
0.0001
0.67 1.49 2.67 3.91 5.05 6.09 5.75
333.1 K 3.5'
5.25'
323.1 K 3.5'
2.03 3.42 5.37 6.96 8.55 10.8
4.55 7.88 10.68 13.14 15.07 16.97
3.25 4.78 5.84 7.05 7.91 8.98
318.1 K 3.5' 2.45 3.17 4.66 5.22 6.05 6.79 7.18 7.63
Cosolvent composition in mol % (solute free).
Table V. Solubility of Naproxen in SC COa with Ethyl Acetate Cosolvent a t 333.1 K mole fraction naproxen x 106 press. (bar) 1.75' 3.5' 5-25" 110.3 0.64 2.10 5.32 1.32 3.43 7.38 124.1 2.36 5.27 9.64 137.9 3.26 6.55 11.60 151.7 4.15 7.18 13.11 165.5 179.3 5.26 9.75 14.33
0.003
z w
1.75'
a
6
10
12
14
16
18
20
DENSITY (MOUL)
Figure 4. Solubilityof naproxen in pure supercritical carbon dioxide. Solid line represents line of best fit.
and Higuchi (1990). A useful measure of the relative potential of these kind of interactions would be to divide the polarizability a* by the mean volume 0 of the molecule. The a*/D ratios for all the cosolvents and naproxen are listed in Table 11. The molar volume, u, was used instead of the mean molecular volume. The units used were such that this ratio remains a dimensionless entity. Acetone and ethyl acetate do not self-associate and are solely hydrogen bond acceptors. However, they vary greatly in terms of dipole moment, dielectric constant, molecular size, and critical properties. Alcohols on the other hand are able to be both hydrogen bond donors and acceptors. They also tend to self-associate even in SCF's (Fulton et al., 1991a,b). Because of the similarities in the chemical and physical behavior of compounds in a homologous series, their use as cosolventscould contribute to a further understanding of their contributions to solubility enhancement. The alcohols used are methanol, ethanol, 1-propanol, and 2-propanol. The reason for including 2-propanol was to study possible steric effects.
Cosolvent composition in mol % (solute free).
Results and Di8CU88iOn Pure Component Solubility. The solubilityof naproxen in pure SC C02 was obtained at 313.1,323.1, and 333.1 K and is shown in Table 111. As indicated by Figure 4, the logarithm of the experimental solubility data gave a good linear correlation with respect to pure COz density. This was expected as shown by various workers (Chrastil, 1982; Kumar and Johnston, 1988; Gurdial et al., 1989; Wells et al., 1990; Gurdial and Foster, 1991)and provides a check on the internal consistency of the data. Cosolvent Effect. The introduction of cosolvents resulted in a marked increase in solubility for all the cosolvents used in this study. These solubility data are given in Tables IV-IX. The shapes of the isotherms were similar to those obtained with pure C02, with each concentration offset by almost a constant distance from the previous one. The solubility isotherms are also linear on a log solubility-mixture density plot as represented in Figure 5 for the CO2-methanol system. Mixture densities were determined as described in the Effect of Density section. Thus the solubility behavior in SC COz-cosolvent mixtures is similar to that in pure CO2 under the conditions studied. Some workers have observed a significant shift of the crossover pressure (Gurdial, 1992;Dobbset al., 1986) when cosolvents were added, however this shift is not significant with naproxen and the cosolvents studied.
Ind. Eng. Chem. Res., Vol. 32, No. 7,1993 1475 Table VI. Solubility of Naproxen in SC COz with Methanol Cosolvent
1
0.1
1L-Z 0
-
3.5 moPh
mole fraction naproxen x 105 press. (bar) 110.3 124.1 137.9 151.7 165.5 179.3 193.1 0
1.75" 0.87 1.95 3.53 5.64 7.76 9.53
333.1 K 3.50 3.29 7.27 12.22 16.54 20.39 23.56 28.54
5.25' 8.11 15.63 22.99 30.25 35.90 41.61
323.1 K 3.5" 8.69 11.37 14.75 16.31 19.52 22.46 24.05
2
3
0.02 -
Ps
0.01
F
0.005
-
0.002
1
Pun
Z
w
0
0.001
Cosolvent composition in mol % (solute free). 0.0005
Table VII. Solubility of Naproxen in SC C02 with Ethanol Cosolvent mole fraction naproxen x 105 press. (bar) 110.3 124.1 137.9 151.7 165.5 179.3 a
1.75 mol%
1.75O 1.26 2.69 4.42 6.26 8.09 9.55
333.1 K 3.50 4.76 9.62 14.17 18.16 22.19 25.61
5.25" 12.78 21.47 29.87 36.43 42.58 47.78
323.1K 3.5' 8.13 11.42 14.21 15.24 16.82 19.18
a
ia
10
12 14 16 MIXTURE DENSITY (MOUL)
Figure 5. Solubility of naproxen in supercritical carbon dioxidemethanol mixtures at 333.1 K. Solid line represents line of best fit. 25
20
I-
*
ACETATE
Cosolvent composition in mol % (solute free).
Table VIII. Solubility of Naproxen in SC COz with l-Propanol Coeolvent at 333.1 K press. (bar)
110.3 124.1 137.9 151.7 179.3 0
mole fraction naproxen x 105 1.75" 3.50 5.25O
8.66 14.34 19.30 23.18 31.58
2.23 3.86 5.88 7.35 11.20
25.17 34.65 42.34 50.40 61.82
Cosolvent composition in mol % (solute free).
Table IX. Solubility of Naproxen in SC CO2 with 2-Proeanol Cosolvent mole fraction naproxen x 105 press. (bar) 110.3 124.1 137.9 151.7 166.5 179.3 0
1.75O 1.37 3.20 5.32 7.22 8.91 10.84
333.1K 3.50 7.01 11.80 16.83 21.67
5.25O 19.71 28.09 36.33 43.82
28.11
56.60
323.1 K 3.50 9.54 13.80 15.85 18.31 21.64 22.00
Cosolvent composition in mol % (solute free).
In order to illustrate the enhancement as the result of the introduction of cosolvent more clearly, the cosoluent effect is defined as the ratio of the solubility obtained with cosolvent to that obtained without cosolvent. The cosolvent effects as a function of cosolvent composition on a solute-free basis at 333.1 K and 179.3 bar for all the cosolventsystems are shown in Figure 6. Ekart et al. (1992) observed that the cosolvent effect for most of the systems they studied varied almost linearly with cosolvent compositions. They studied the cosolvent effects of a wide selection of cosolvents on a variety of organic compounds in SC ethane using SCF chromatography. However, as can be seen in Figure 6, the naproxen solubility varied nonlinearly with composition and the cosolvent effect increases more rapidly at higher concentration. This may be indicative of higher order interactions between the solute and the cosolvent.
0'
0
'
'
1
I
'
'
I
2 3 4 MOLE PERCENT COSOLVENT
'
I
5
'
'
6
Figure 6. Cosolvent effect as a function of cosolvent concentration at 333.1 K and 179.3 bar.
Effect of Density. The addition of a cosolvent generally increases the bulk density of the fluid mixture which would contribute to solubility enhancement. A large variation in density would be anticipated close to the critical point where the isothermal compressibility is largest. However, at pressures and temperatures further away from this region, where the fluid is less compressible, the increase in bulk density is not expected to be very significant and should be within a few percent (0-3% for P > 180bar) for the cosolventconcentration range between 1 and 5 mol %. The magnitude of the density contribution to the cosolvent effect was estimated using the Peng-Robinson equation of state (PR EOS). In order to obtain a reasonable estimate of the mixture density, the following procedure was used. First, the ratio of the calculated mixture density to the calculated pure SC COz density was obtained using the PR EOS. This ratio was then multiplied by the actual COz density (Wells, 1991) to give the estimated density. This procedure will help the density curves fit the shape of pure COz isotherms. The binary interaction parameters necessary were obtained by fitting the EOS to binary vapor-liquid equilibrium data as described later. This procedure gives binary cosolvent-C0z density estimates accurate to within 20 % and will in general overpredict the correct value based on the limited density data available (Dobbs et al., 1987; Tilly, 1992). On the basis of the calculated mixture densities, the density contribution to the overall enhancement was estimated by determining the increase in naproxen solubility in pure SC COz at the same temperature and density as the Cot-cosolvent mixture. The contribution of the bulk density increase to the overall cosolvent effect for the methanol system at 333.1 K is shown in Figure 7. A t
1476 Ind. Eng. Chem. Res., Vol. 32, No. 7, 1993
17s 3.5 1.21
1 ,I15 125
MOLE PERCENT COSOLVENT
120
130
140
is0 160 tm PRESSURE (BAR)
BO
190
Figure 7. Contribution to the cosalvent effect from the increased bulkdensity ass result of cmlvent addition. System shown is carbon dioxidemethanol at 333.1 K.
Figure 8. Cmlvent effect as a function of pressure at 333.1 K and a cosolvent concentration of 5.25 mol %.
1.75% cosolvent, the density increase is responsible for 3&70% of the total cosolvent effect at all pressures. This is clearly significant and shows that the density increase must be considered when interpreting these effects. It can be seen that at higher cosolvent concentrations these contributions do not seem as significant when compared to the total solubility enhancement observed. This result is consistent both with an operating condition removed from the high compressibility region and with strong specificinteractions between the cosolvents and naproxen. These observations suggest that accurate density data are needed for SCF-cosolvent mixturesso that density effects may he more clearly separated from specific interactions in systems where both are present. Effect of Pressure. The addition of cosolvent to a binary SCF-solid system can cause the phase behavior to change in three significant ways (Lemert and Johnston, 1989). The cosolvent lowers the melting temperature of the solid at a given pressure. This can be as much as 70 'C as shown by the 2-naphthol-methanol-COz system. The shape of the temperature-pressure projection of the solid-liquid-gas melting point curve can also change. Finally, the presence of cosolvent lowers the upper critical end point (UCEP) pressure significantly. A t a critical end point (LCEP or UCEP), the solubility, y ~ is, very sensitive to pressure and ayZ/aP is large (Kim et al., 1985; Lemert and Johnston, 1990) when compared to that obtained in the pure SCF. Although aydt3P values for all the cosolvent systems are higher than those in pure COz, there were no abrupt rises in solubility (usually observed near a CEP) for the range of temperatures and pressures studied, indicating that CEPs were not approached in this study. The effect of pressure on the cosolvent effect changes with cosolvent concentration (see Figure 7). For the system COrmethanol-naproxen the total cosolvent effect is essentially constant (within experimentaluncertainty) a t 1.757% methanol. It goes through a clear maximum at 138 bar for 3.5 % methanol and then monotonically decreases with pressure for 5.25% methanol. All the cosolvent systems exhibited a decreasing cosolvent effect with increasing pressure at 5.25% cosolvent as shown in Figure 8. Various workers have investigated the nature of the solute-cosolvent interactions under these conditions (Kim and Johnston, 1987a,b;Yonker and Smith, 1988,1989).It was shown that the region within the first few solvent shells around the solute molecule is enriched with the cosolvent and the local concentration can be several times that of the bulk composition. The significance of this local ordering of the cosolvent molecules decreases with
increasing pressure, and at high enough pressures, the concentration of the cosolvent around the solute will ultimately approach the bulk concentration (Yonker and Smith, 1988). An important point to realize is that while the localcomposition enhancementdecreases,the absolute local concentration of cosolvent around the solute will alwaysincreaaewithincreasingpressuredueto the increase indensity. In the presentcase, itappeara thatthecosolvent effect at low cosolvent concentrations depends predominantly on the absolute concentration of cosolvent around the soluteas no distinct pressure dependence was observed. As the cosolvent concentration is increased, the effect of local composition enhancement becomes apparent. Since the local composition enhancement is maximized in the region of high compressibility, it seems plausible that the decrease in the difference between bulk and local cosolvent concentration with pressure would lead to the observed decrease in cosolvent effect with pressure. This could be confirmed with information on the compressibility of the cosolvent mixtures. Effect of Mixture Critical Point. As an aid to the interpretation of the effect of alcohol chain length and nonspecificinteractions, we compare the relative distance of the operating conditions from the critical state of each binaryC0rcosolvent system. This comparison is possible because the critical locus data for all the COz-cosolvent binary mixtures used in this study are available. The critical temperatures of the respective Con-cosolvent systems as a function of composition (Gurdial et al., 1991a,b) are shown along with the operating condition in Figures 9 and 10. For a given alcohol composition, the trend in critical temperature of the systems is similar to that of the observed cosolvent effect. Thus it can be seen that the system closer to the binary critical point of the solvent mixture also has a larger cosolvent effect. This ohervation is also true for critical pressures. It is probable that any specific interactions involved in the alcohol systems are similar in all cases because of the similarity of the chemical nature. The trend observed for cosolvent effect could then be explained by the larger isothermal compressibility and solutepartial molar volume for system closer to the binary critical point. The ethyl acetate and acetone cosolvent systems exhibit different behavior. The critical locus behavior of ethyl acetate is similar to that of 1-propanol and acetone is similar to ethanol, but the cosolvent effects were generally much smaller for ethyl acetate and acetone than for the alcohols. Even though there is a substantial difference in the critical temperatures, the cosolvent effects are quite similar. Furthermore, the experimental conditions are
Ind. Eng. Chem. Res., Vol. 32, No. 7,1993 1477 350
340
330
I
I-PROPANOL
t
PPROPANOL 0
ETHANOL
t
m
/
MIX-IANOL
0
Y
OPERATING CONDmON
#320
310
Inn “WW
I
0
2
I
I
0
4 6 MOLE % COSOLVENT
Figure 9. Mixture critical temperature as a function of composition for alcohol cosolventa. Operating condition (5.25%, 333.1K)shown. Solid lime representa line of best fit. ACETATE
340
t
0
I
1
2 3 4 5 MOLE % COSOLVENT
6
U
ACETONE
0 OPERATNG CONDITION
7
Figure 10. Mixture critical temperatures as a function of composition for acetone and ethyl acetate cosolventa. Operating conditions (5.25%,333.1 K)shown. Solid line represents line of best fit.
closest to the binary ethyl acetate-C02 critical point but acetone gives a slightly larger cosolvent effect. Presuming that proximity to the mixture critical point does in fact lead to higher cosolvent effects, this observation suggests a cancellation of effects in these two systems. For example, if the dipole moment contributes significantly to the cosolvent effect, then acetone, having a very much larger dipole moment than ethyl acetate, would be a better cosolvent. However, being closer to the binary critical point for the naproxen-ethyl acetate-C02 system compensates for the difference in dipole moment thus producing a similar cosolvent effect. Effect of Specific Interactions. It has been shown by various workers (Walsh et al., 1987,1989; Walsh and Donohue, 1989; Grant and Higuchi, 1990; Lemert and Johnston, 1991; Ekart et al., 1992) that one of the most important factors in improving solubility of polar solutes is to increase or enhance the specific interactions possible between the solute and the solvent or cosolvent. One of the most important specific interactions is hydrogen bonding. Walsh et al. (1987) have proposed that the cosolvent effect is the result of chemical association via hydrogen bonding, and they were able to rationalize the results of Van Alsten et al. (1984) and Schmitt and Reid (1986)based on this concept. Another specific interaction that is expected to cause a significant cosolvent effect is the formation of charge-transfer complexes. This was demonstrated by Lemert and Johnston (1991) for the solubility of hydroquinone in the COz-tri-n-butyl phosphate cosolvent system.
As naproxen is an organic acid, cosolvents would be expected to interact to a certain extent through hydrogen bonding and the solubility data should reflect this. Both acetone and ethyl acetate are only hydrogen bond acceptors as indicated by the molecular structure and the KamletTaft CY parameters (0.06 and 0, respectively). The p values for both ethyl acetate and acetone are similar indicating that both of these compounds have a similar tendency for accepting hydrogen bonds. The Hansen solubility parameters (ah) of these two compounds also reflect this. On the basis of this observation and assuming that hydrogen bonding would be a major contributor to positive cosolvent effects, then these two cosolvents should provide similar results. This was in fact observed for these two cosolvents with a maximum cosolvent effect around 11.2 (a solubility increase of about 579%) at 124.1 bar and 333.1 K with 5.25 mol ?6 cosolvent. Since both cosolvents vary significantly in their structure, dipole moments, and polarizabilities, a similar cosolvent effect could be explained in many ways. As discussed above, rationalization based on thermodynamic state and polarity implies an approximate cancellation of equal effects. A more consistent explanation would seem to be the similar hydrogen bond accepting ability of both indicating that specificcosolventsolute interactions dominate in this system. The alcohols used in this study produced a much larger enhancement than both acetone and ethyl acetate. As alcohols are amphiprotic they are both proton donors and acceptors as indicated by the Kamlet-Taft (Y and 0 parameters. This capability has the potential of increasing the number of ways that they can interact specifically with naproxen compared with acetone or ethyl acetate. In addition, the Hansen hydrogen bonding solubility parameters bh for the alcohols are larger than those for either acetone or ethyl acetate. Thus it is expected that the alcohols would produce larger cosolvent effects. The observed cosolvent effect for the alcohols increased in the order methanol, ethanol, 2-propanol and then followed by l-propanol which produced a cosolvent effect of almost 50 at 124.1 bar and 333.1 K (5.25 mol % cosolvent). The 8h values for the alcohols given in Table I1 decrease with increasing alcohol chain length as do the Kamlet-Taft CY parameters. However, the p parameter increases with increasing alcohol chain length. It appears that the ability of the alcohols to accept a hydrogen bond plays an important role in the solvation of naproxen. The fact that l-propanol and 2-propanol give quite different cosolvent effects is remarkable considering their similarity. One might interpret this as a steric effect whereby the active site for interaction is somewhat shielded in 2-propanol relative to l-propanol. However, in order to account for the difference in cosolvent effect, one would expect greater differences in the solvatochromic parameters of these two components. It appears likely that a slight steric effect combined with the effect of relative proximity to the mixture critical point to produce the observed differences. Data Treatment. Cubic equations of state have been used extensively to correlate solubility of solids in supercritical solvents. Both the Soave-Redlich-Kwong (SRK) (Soave, 1972) and the Peng-Robinson (PR) EOS’s (Peng and Robinson, 1976) were used in this study to correlate both the binary and ternary systems presented above. The classical mixing rules with one binary interaction parameter for the cross energy parameter were used. The required properties for carbon dioxide, naproxen, and the respective cosolventsare given in Table X. Except for the melting point, none of the required properties of
1478 Ind. Eng. Chem. Res., Vol. 32,No. 7, 1993 Table X. Physical Properties of All Compounds
P v (bar x 108)
e
compound
MW
carbon dioxide ethyl acetate acetone methanol ethanol l-propanol 2-propanol naproxen
44.01 88.11 58.08 32.04 46.07 60.1 60.1 230.3
Tc (K) 304.1 523.2 508.2 512.6 513.9 536.8 508.3 807c
Pc (bar) 73.8 38.3 46.6 80.9 61.4 51.7 47.6 24.2d
w
313.1 K
318.1 K
323.1 K
333.1 K
0.225" 0.362b 0.318" 0.556' 0.635" 0.623b 0.665b 0.904e
3.33
6.37
11.9
38.8
0 Smith and Van Ness (1975). b Reid et al. (1988). c Lyman et al. (1982);Lyderson-Miller method. d Lyman et al. (1982);Lyderaen's method. Lyman et al. (1982);Edmister's method. f Lyman et al. (1982); modified Wateon correlation.
Table XI. Sources of VLE Data and Regressed Binary Interaction Parameters CO2-cosolvent methyl acetate acetone methanol ethanol l-propanol 2-propanol
sources used Ohgaki and Katayama, 1975 Katayama et al., 1975 Ohgaki and Katayama, 1975; Semenova et al., 1979 Suzuki et al., 1990: Jennings et al., 1991 Suzuki et al.; 1990' Radosz, 1986
naproxen are available in the open literature and thus have to be estimated. The estimation methods used here were obtained from those presented by Lyman et al. (1982). Correlating solubility data in this study involved an optimization process whereby a kij is obtained for each binary pair; therefore, for a ternary system, three binary interaction parameters would result, klz,k13, and k23. For the COz-naproxen binary system, the klz was obtained by minimizing the following objective function at each temperature:
ku(333.1 K) SRK PR -0.0759 -0.0645 0.0081 0.0137 0.0695 0.0749 0.0827 0.0861 0.0780 0.0796 0.0881 0.0929
temp (K) 298.15,313.15 298.15,313.15 298.15 323.15.348.15.373.15 313.4,333.4; 314.5,325.2 313.4,333.4 316.7,334.9,354.5,394.5 0.005 0.003
-
0.002
-
Bx g
om1
$
0.0005
2z
-l
8
/
0.0003
-
0.0002
-
0
/
SRK
.
313.1 K - . 0
313.1 K 323.1 K 333.1 K 333.1K l
n nom
I
I
1
I
,
I
,
I
,
_.-_-I
BO
Then for the COz-cosolvent-naproxen ternary systems, the klz obtained above was used and the k13 was obtained independently from the correlation of the respective COZcosolvent VLE data using the required EOS. The sources of the VLE data and the corresponding k13 at 333.1 K for both the SRK and PR equations are given in Table XI. Where no VLE data exists at 333.1 K, the k13 was either interpolated or extrapolated from the available data assuming a linear relationship. In this study the COZethyl acetate system was not available and it was approximated by the COz-methyl acetate system. The case where k13 = 0 was also considered. At 333.1 K, the solubility of naproxen in each COZcosolvent system is available for three cosolvent concentrations: 1.75,3.5,and 5.25mol % on a solute-free basis. As binary interaction parameters are not functions of composition, the correlation was obtained by minimizing the following objective function:
The number of points at each concentration is given by Nj. From this optimization process, a k z 3 was obtained for each COz-cosolvent-naproxen system at 333.1 K. Correlation Results. The correlations for the solubility of naproxen in pure supercritical COZat 313.1,323.1, and 333.1 K using the SRK EOS are shown in Figure 11. The PR EOS gave similar results and hence they are not shown. The comparisons between the two EOS's are given
1w
120
1 4
160
180
200
PRESSURE (BAR)
Figure 11. SRK equationof state correlationof naproxen solubility in pure SC C02.
in Table XI1 and shows them to be almost identical for correlating binary solubility data. The klz obtained for this system appears to be onlyslightly sensitiveto relatively small temperature differences, which might prove to be useful for correlating solubility data at other temperatures. As discussed above, the k13 for the ternary systems was obtained independently by correlating the respective COT cosolvent system. It may also be reasonable to set this parameter to zero as the size differences between these components are relatively small. Both of these options for k13 were considered in this study. For k 1 ~= 0, the difference in performance between these two EOS's is quite significant as shown in Figure 12 for the methanol system at 333.1 K. The SRK EOS performed better for the alcohol systems while the PR EOS performed better for the ethyl acetate and acetone systems. It appears that the PR EOS is more sensitive to the 1223 values for the alcohol systems and hence gives poorer correlations. The introduction of the respective k13'~ did not necessarily improve the correlations or alter the relative performance of the two EOS's. It should be noted that the currently favored method of obtaining the binary interaction parameters separately may not necessarily be representative of the interactions occurring in the mixture. The differences between the two options for k13 are shown in Table XIII. The absolute value of k23 for the ethyl acetate and acetone systems is smaller than that of the alcoholsystems.
Ind. Eng. Chem. Res., Vol. 32, No. 7,1993 1479
i t
" 0.05 O'l
Ly
0.02
E
0.01
2
$ 1 ae
0.005
=
0.002
solubility enhancement. The cosolvent effect was also found to vary nonlinearly with the composition of the cosolvent and decrease with increasing pressure. The cosolvent effect increases in the following order: ethyl acetate, acetone, methanol, ethanol, 2-propanol, 1-propanol. It was difficult to explain this trend qualitatively with any one physical property of the cosolvent. However, it was found that the Kamlet-Taft B parameter for the cosolvents follows the trend observed for the cosolvent effect for each system. This may provide an indication that specific interaction via hydrogen bonding plays an important role in solubility enhancement. The difference between 2-propanol and 1-propanol can be explained by a combination of steric effects and proximity to the mixture critical point. Further understanding of the cosolvent effect can be realized when the relative distance to the COz-cosolvent binary critical point is taken into account. The solubility data presented were correlated with the SRK and PR EOS's. The required physical properties of naproxen were estimated using established estimation methods. For the binary systems, both EOS's gave similar results. The best correlations for the ternary systems were obtained by using the SRK EOS. It was found that the introduction of k13, obtained independently from correlating the respective VLE data, did not necessarilyimprove the correlation. For design purposes, it would be justifiably convenient to set k13 to zero. The large and negative k23 values indicate very strong cosolvent-solute interactions. Unfortunately, these parameters exhibited no trend. These results suggest the possibility of chemical association between the various cosolvents and naproxen and that a chemical-physical equation of state may be an appropriate model. Further work along these lines is progressing.
F
I
temp (K) 313.1 323.1 333.1
biz 0.240 0.239 0.244
I
SRK % AARD 15.1 13.2 9.2
1
I
kiz 0.223 0.223 0.229
,
.
PR % AARD 13.9 13.7 9.4
Table XIII. Effect of 4 s on Correlation of Supercritical COdosolvent-Naproxen Ternary Systems at 333.1 K cosolvent ethyl acetate acetone methanol ethanol 1-propanol 2-propanol average ethyl acetate acetone methanol ethanol 1-propanol 2-propanol average
k13 = 0 kzs %AARD SRK EOS 18.4 0.057 12.3 -0.034 13.5 -0.434 9.1 -0.376 12.8 -0.373 9.2 -0.382 12.6 -0.086 -0.196 -0.607 -0,533 -0.501 -0.538
PR EOS 9.8 8.3 16.3 16.3 21.1 18.5 15.1
kis # 0 kz3 %AARD 0.075 -0.034 -0.441 -0.389 -0.392 -0.440
23.3 11.8 12.0 10.1 17.0 13.2 14.6
-0.086 -0.203 -0.620 -0.559 -0.541 -0.538
14.8 7.6 16.2 18.7 25.0 23.1 17.6
The values for the alcohol systems are large and negative. When km is less then zero, the attractive interaction is stronger than that approximated by the geometric mean. The unusually large and negative k 2 3 ' ~obtained for the alcohol systems indicate that strong interactions exist between naproxen and the alcohols. These interactions are smaller in the ethyl acetate and acetone systems as indicated by the magnitude of the respective k23's. The absence of a general trend that corresponds with the trend in cosolvent effects observed for these systems makes assigning a physical meaning to the k23's difficult. Thus, at beat, they can only be considered as adjustable parameters.
Conclusion The solubility behavior of naproxen in pure SC C02 and in a range of polar cosolvent-SC C02 mixtures was studied. The experimental conditions ranged from 100to 200 bar and the temperature ranged from 313.1 to 333.1 K. It was found that all the cosolventa used produced
Acknowledgment Partial financial support for this work was provided by the Australian Research Council through Grant No. A89030392. D.L.T. gratefully acknowledges financial support from the National Science Foundation through Grant No. INT-9203312. Literature Cited Barton, A. Handbook of Solubility Parameters and Other Cohesive Parameters; CRC Press: Boca Raton, FL, 1983. Brennecke, J. F.; Tomasko, D. L.; Peshkin, J.; Eckert, C. A. Fluorescence Spectroscopy Studies of Dilute Supercritical Solutions. Znd. Eng. Chem. Res. 1990, 29, 1682. Chrastil, J. Solubility of Solids and Liquids in Supercritical Gases. J. Phys. Chem. 1982,86,3016. Cochran, H. D.; Lee, L. L. SolvationStructure in Supercritical Fluid Mixtures Based on MolecularDistribution Functions. ACS Symp. Ser. 1989,406, 27. Cygnarowicz, M. L.; Maxwell, R. J.; Seider, W. D. Equilibrium Solubilities of @-carotenein Supercritical Carbon Dioxide. Fluid Phase Equilib. 1990, 59, 57. Debenedetti, P. G. Clustering in Dilute, Binary Supercritical Mixtures: AFluctuation Analysis. Chem. Eng. Sci. 1987,42 (9),2203. Debenedetti, P. G.; Petsche, I. B.; Mohamed, R. S. Clustering in Supercritical Mixtures: Theory, Applications and Simulations. Fluid Phase Equilib. 1989,52, 347. Dobbs, J. M.; Wong, J. M.; Johnston, K. P. Nonpolar Cosolvents for Solubility Enhancement in Supercritical Fluid Carbon Dioxide. J. Chem. Eng. Data 1986,31, 303. Dobbs,J. M.;Wong,J. M.;Lahiere,R. J.;Johnston, K. P. Modification of Supercritical Fluid Phase Behavior Using Polar Cosolvents. Ind. Eng. Chem. Res. 1987,26, 56. Eckert, C. A.; Ziger, D. H.; Johnston, K. P.; Kim, S. Solute Partial Molar Volumes in Supercritical Fluids. J. Phys. Chem. 1986,90, 2738.
1480 Ind. Eng. Chem. Res., Vol. 32, No. 7, 1993 Ekart, M. P.; Bennett, K. L.; Ekart, S. M.; Liotta, C. L.; Eckert, C. A. The Influence of Specific Interactions on Cosolvent Effects in SuDercritical Fluids Solutions. Submitted for publication in AZChE J . 1992. Fedors, R. F. A Method for Estimating Both the Solubility Parameter and Molar Volumes of Liauids. Polym. Eng. Sci. 1974,14, 147. Frye, S. L.; Yonker, C. R.; Kaikwarf, D. R.; Smith, R. D. Application of Solvatochromic Probes to Supercritical and Mixed Fluid Solvents. ACS Symp. Ser. 1987,329, 29. Fulton, J. L.; Yee, G. G.; Smith, R. D. Hydrogen Bonding of Methyl Alcohol-d in Supercritical Carbon Dioxide and Supercritical Ethane Solutions. J . Am. Chem. SOC.1991a, 113,8327. Fulton, J. L.; Yee, G. G.; Smith, R. D. Hydrogen Bonding of Simple Alcohols in Supercritical Fluids: An FT-IR Study. Presented at the Annual Meeting of the AIChE, Los Angeles, CA, 1991b. Grant, D. J. W.; Higuchi, T. Solubility Behaviour of Organic Compounds. In Techniques of Chemistry; Weissberger, A., Ed.; Wiley: New York, 1990; Vol. xxi. Gurdial, G. S. Solubility Behaviour of Organic Compounds in Supercritical Carbon Dioxide. Ph.D. Thesis, University of New South Wales, 1992. Gurdial, G. S.;Foster, N. R. Solubility of o-HydroxybenzoicAcid in Supercritical Carbon Dioxide. Znd. Eng. Chem. Res. 1991, 30, 575. Gurdial, G. S.; Wells, P. A.; Foster, N. R. The Role of Polarity on Correlations of Solid-Supercritical Fluid Phase Systems. J. Supercrit. Fluids 1989, 2, 85. Gurdial, G. S.; Foster, N. R.; Yun, S. L. J. The Critical Loci of Binary Polar and Non-Polar Organic Compounds-CO2 Systems at Low Solute Concentrations. Presented at the Second International Conference on Supercritical Fluids, Boston, MA, 1991a. Gurdial, G. S.; Foster, N. R.; Yun, S. L. J.; Ting, S. S. T. P-T-x Diagrams in the Critical Region of Binary COZ-Polar and NonPolar Organic Systems at Low Solute Concentrations. Presented at the Nineteenth Australasian Chemical Engineering Conference (CHEMECA) '91, Newcastle, New South Wales, 1991b. Gurdial, G. S.; Macnaughton, S. J.; Tomasko, D. L.; Foster, N. R. Influence of Chemical Modifiers on the Solubility of 0- and m-Hydroxybenzoic Acid in Supercritical COz. Znd. Eng. Chem. Res. 1993, accompanying paper in this issue. Hansen, C. M. The Three Dimensional Solubility Parameter-Key to Paint Component Affinities. I. Solvents,Plasticizers,Polymers, and Resins. J . Paint Technol. 1967a, 39, 104. Hansen, C. M. The Three Dimensional Solubility Parameter-Key to Paint Component Affinities. 11. Dyes, Emulsifiers, Mutual Solubility and Compatibility, and Pigments. J . Paint Technol. 1967b, 39, 505. Hansen, C. M. The Universality of Solubility Parameter. Znd. Eng. Chem. Prod. Res. Deu. 1969,8, 2. Hansen, C. M.; Beerbower,A. SolubilityParameters. In Encyclopedia ofChemica1 Technology, 2nd ed.; Wiley: New York, 1971;Suppl. VOl. Hildebrand, J. H.; Scott, R. L. The Solubility of Non-Electrolytes, 3rd ed.; Reinhold Dover, NY, 1950. Holler, W. E.; Ehrlich, P. Solubility of Naphthalene in Mixtures of Carbon Dioxide and Ethane. J . Chem. Eng. Datu 1990,35,271. Hoy, K. L. New Values of the Solubility Parameters from Vapour Pressure Data. J . Paint Technol. 1970,42, 76. Jennings, D. W.; Lee, R.-J.; Teja, A. S. Vapor-Liquid Equilibria in the Carbon Dioxide + Ethanol and the Carbon Dioxide + 1-Butanol Systems. J . Chem. Eng. Data 1991,36, 303. Kajimoto, 0.;Futakami, M.; Kobayashi, T.; Yamasaki, K. ChargeTransfer-State Formation in Supercritical Fluids: (N,N-Dimethy1amino)benzonitrile in CFaH. J . Phys. Chem. 1988,92,1347. Kamlet, M. J.; Taft, R. W. The SolvatochromicComparison Method. 1. The 8-Scale of Solvent Hydrogen-Bond Acceptor (HBA) Basicities. J . Am. Chem. SOC.1976a, 98, 377. Kamlet, M. J.; Taft, R. W. The SolvatochromicComparison Method. 2. The a-Scale of Solvent Hydrogen-Bond Donor (HBD) Acidities J. Am. Chem. SOC.1976b, 98, 2886. Kamlet, M. J.; Abboud, J. L.; Taft, R. W. The Solvatochromic Comparison Method. 6. The r*Scale of Solvent Polarities. J . Am. Chem. SOC.1977,99,6027. Kamlet,M. J.;Abboud, J.-L. M.;Abraham,M.H.;Taft,R. W.Linear Solvation Energy Relationships. a*,a, and 8, and Some Methods for Simplifyingthe Generalized SolvatochromicEquation. J. Org. Chem. 1983,48, 2877. Katayama, T.; Ohgaki, K.; Maekawa, G.; Goto, M.; Nagano, T. Isothermal Vapor-Liquid Equilibria of Carbon Dioxide and
Methanol Carbon Dioxide Systems at High Pressures. J . Chem. Eng. Jpn. 1975,8,89. Kim, S.; Johnston, K. P. Clustering in Supercritical Fluid Mixtures. AZChE J . 19878, 33 (lo), 1603. Kim, S.;Johnston, K. P. Molecular Interactions in Dilute Supercritical Fluids Mixtures. Znd. Eng. Chem. Res. 1987b, 26,1206. Kim, S.; Wong, J. M.; Johnston, K. P. Theory of the Pressure Effect in Dense Gas Extraction. In Supercritical Fluid Technology; Penninger, J. M., Radosz, M., McHugh, M. A., Krukonis, V. J., Eds.; Elsevier Science: Amsterdam, 1985. Koenhen, D. M.; Smolders, C. A. The Determination of Solubility Parameters of Solvents and Polymers by Means of Correlations with Other Physical Quantities. J . Appl. Polym. Sci. 1975, 19, 1163. Kumar, S. K.; Johnston, K. P. Modelling the Solubility of Solids in Supercritical Fluids with Density as the Independent Variable. J . Supercrit. Fluids 1988, 1 , 15. Larson, K. A.; King, M. L. Evaluation of SupercriticalFluid Extraction in the Pharmaceutical Industry. Biotechnol. Prog. 1986,2, 73. Lemert, R. M.; Johnston, K. P. Solid-Liquid-Gas Equilibria in Multicomponent Supercritical Fluid Systems. Fluid Phase Equilib. 1989, 45,265. Lemert, R. M.; Johnston, K. P. Solubilities and Selectivities in Supercritical Fluid Mixtures Near Critical End Points. Fluid Phase Equilib. 1990, 59, 31. Lemert, R. M.; Johnston, K. P. Chemical Complexing Agents for Enhanced Solubilities in Supercritical Fluid Carbon Dioxide. Znd. Eng. Chem. Res. 1991,30, 1222. Lyman, W. J.; Reehl, W. F.; Rosenblatt, D. H.Handbook of Chemical Property Estimation Methods; McGraw-Hilk New York, NY, 1982. Morita, A.; Kajimoto, 0. Solute-Solvent Interaction in Nonpolar Supercritical Fluid A Clustering Model and Size Distribution. J. Phys. Chem. 1990,94,6420. Ohgaki, K.; Katayama, T. Isothermal Vapor-Liquid Equilibria for System Ethyl Ether-Carbon Dioxide and Methyl Acetate CarbonDioxide at High Pressures. J . Chem. Eng. Data 1976, 20, 264. Peng, D. Y.; Robinson, D. B. A New Two Constants Equation of State. Znd. Eng. Chem. Fundam. 1976,15, 59. Petsche, I. B.; Debenedetti, P. G. Solute-Solvent Interactions in Infinitely Dilute Supercritical Mixtures: A Molecular Dynamics Investigation. J . Phys. Chem. 1989,93, 7075. Petsche, I. B.; Debenedetti, P. G. Influence of Solute-Solvent Asymmetry Upon the Behaviour of Dilute Supercritical Mixtures. J. Phys. Chem. 1991,95, 386. Radosz, M. Vapor-Liquid Equilibrium for 2-Propanol and Carbon Dioxide. J. Chem. Eng. Data 1986, 31, 43. Reid, R. C.; Prausnitz, J. M.; Poling, B. E. The Properties of Gases & Liquids, 3rd ed.; McGraw-Hill International Editions: Singapore, 1988. Schaeffer, S. T.; Zalkow, L. H.; Teja, A. S.Solubility of Monocrotaline in Supercritical Carbon Dioxide and Carbon Dioxide-Ethanol Mixtures. Fluid Phase Equilib. 1988,43, 45. Schmitt, W. J.; Reid, R. C. The Use of Entrainers in Modifying the Solubility of Phenanthrene and Benzoic Acid in Supercritical Carbon Dioxide and Ethane. Fluid Phase Equilib. 1986,32,77. Semenova, A. I.; Emel'yanova, E. A,; Taimmerman, S. S.; Tsiklis, D. S. Phase Equilibria in the Methanol-CarbonDioxideSystem. Rum. J . Phys. Chem. 1979,53,1428. Smith, J. M.; Van Ness, H. C. Introduction to Chemical Engineering Thermodynamics, 3rd ed.; McGraw-Hill International Editions: Singapore, 1975. Smith, G. R.; Wormald, C. J. Solubilities of Naphthalene in (C02 + CzHe) and (COS+ CaHs) up to 333 K and 17.7 MPa. Fluid Phase Equilib. 1990, 57, 205. Soave, G. Equilibrium Constants from a Modified Redlich-Kwong Equation of State. Chem. Eng. Sci. 1972,27,1197. Sunol, A. K.; Hagh, B.; Chen, S. Entrainer Selection in Supercritical Extraction. In Supercritical Fluid Technology;Penninger, J . M., Radosz, M., McHugh, M. A., Krukonis, V. J., Eds.; Elsevier Science: Amsterdam, 1985. Suzuki, K.; Sue, H.; Itou, M.; Smith, R. L.; Inomata, H.; Arai, K.; Saito, S. Isothermal Vapor-Liquid Equilibrium Data for Binary Systems at High Pressures: Carbon Dioxide-Methanol, Carbon Dioxide-Ethanol, Carbon Dioxide-1-Propanol, Methane-Ethanol, Methane-1-Propanol, Ethane-Ethanol, Ethane-1-Propanol. J . Chem. Eng. Datu, 1990,35,63. Tavana, A.; Chang, J.; Randolph, A. D.; Rodriguez, N. Scanning of Cosolvents for Supercritical Fluids Solubilieation of Organics. AZChE J . 1989,35,645.
Ind. Eng. Chem. Res., Vol. 32, No. 7, 1993 1481 Tilly, K. D. Viscosities of Supercritical Fluids. Ph.D. Thesis, University of New South Wales, 1992. Van Alsten, J. G. Structural and Functional Effects in Solutions with Pure and Entrainer-Doped Supercritical Solvents. Ph.D. Thesis, University of Illinois at Urbana-Champaign, 1986. Van Alsten, J. G.; Hansen, P. C.; Eckert, C. A. Supercritical Enhancement Factors for Nonpolar and Polar Systems. Presented at the Annual Meeting of the AIChE, San Francisco, CA, 19W, paper 84a. Walsh, J. M.; Donohue, M. D. Hydrogen Bonding in Entrainer Cosolvent Mixtures: A Parametric Analysis. Fluid Phase Equilib. 1989, 52, 397. Walsh, J. M.; Ikonomou, G. D.; Donohue, M. D. Supercritical Phase Behaviour: The Entrainer Effect. Fluid Phase Equilib. 1987,33, 296. Walsh, J. M.; GreenField, M. L.; Ikonomou, G. D.; Donohue, M. D. Hydrogen-BondingCompetition in Entrainer CosolventMixtures. Chem. Eng. Commun. 1989,86,125.
Wells, P. A. Diffusion in Supercritical Fluids. Ph.D. Thesis, University of New South Wales, 1991. Wells, P. A,; Chaplin, R. P.; Foster, N. R. Solubility of Phenylacetic Acid and Vanillan in Supercritical Carbon Dioxide. J.Supercrit. Fluids 1990, 3, 8. Wong,J. M.; Johnston, K. P. Solubilisationof Biomoleculesin Carbon Dioxide Based Supercritical Fluids. BiotechnoL Prog. 1986,2,29. Yonker, C. R.; Smith, R. D. Solvatochromic Behaviour of Binary Supercritical Fluids: The Carbon Dioxidel2-Propanol System. J. Phys. Chem. 1988,92, 2374. Yonker, C. R.; Smith, R. D. Thermochromic Shifts in Supercritical Fluids. J. Phys. Chem. 1989,93, 1261. Received for review October 22, 1992 Revised manuscript received February 16, 1993 Accepted February 23, 1993
