Fluid Phase Equilibria 506 (2020) 112376 Contents lists available at ScienceDirect Fluid Phase Equilibria j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / fl u i d Density-based UNIFAC model for solubility prediction of solid solutes in supercritical fluids Yueqiang Zhao a, *, Weibing Wang b, Weiwei Liu a, Jing Zhu a, Xiaoqin Pei a a b Department of Chemical Engineering, Jiangsu Ocean University(originally Named Huaihai Institute of Technology), Lianyungang, 222005, Jiangsu, China Department of Applied Chemistry, Yuncheng University, Yuncheng, 044000, Shanxi, China a r t i c l e i n f o a b s t r a c t Article history: Received 19 June 2019 Received in revised form 27 October 2019 Accepted 28 October 2019 Available online 31 October 2019 A new theoretical expression of group interaction parameter as a function of fluid density was put forward in this work. The group interaction parameters between the supercritical solvents (carbon dioxide, ethane, fluoroform and chlorotrifluoromethane) and the functional groups of solutes at reference fluid density were presented, which exhibit good compatibility with the UNIFAC Vapor-Liquid Equilibrium (VLE) group interaction parameters. For the solid Solute þ C2H4 systems, satisfactory solubility data representation results can be obtained using the UNIFAC VLE group interaction parameters of “C]C”. This density-based UNIFAC model was examined for the solubility of 20 solid solutes in 5 supercritical solvents (51 systems, 1111 data points); the Average Relative Errors (ARE) in solubility prediction are 19.1%, 18.7%, 21.6%, 20.4% and 12.8% for CO2, C2H6, C2H4, CHF3 and CClF3 correspondingly. In the case of each UNIFAC group being one molecule, the density dependent expression of group interaction parameter derived in this model may be applicable to other activity coefficient models (UNIQUAC, Wilson, NRTL) containing binary interaction parameters for solubility prediction of solids in supercritical fluids. © 2019 Elsevier B.V. All rights reserved. Keywords: Solid solutes Supercritical fluid (SF) Solubility prediction 1. Introduction The Supercritical Fluid Extraction (SFE) technology has been widely used to separate and purify the valuable components in food and pharmaceutical processes [1]. Most of the valuable components in food and pharmaceutical processes are solids, such as caffeine, theobromine, artemisinin, naproxen, benzoic acid, and acridine; therefore the solubility of solid solutes in supercritical solvents is essential for the design and optimization of industrial SFE processes [2]. Until now, many methods have been developed to correlate or predict SFE solubility efficiently, such as equation of state (EOS) methods [3e8], solution models [9e15], and empirical equations [16e25]. Quiram et al. [5] developed a model using the density series virial EOS truncated at the third coefficient to predict the SFE solubility, which yields a linear relationship between the solid solute solubility and the supercritical fluid density. The EOS methods [3e8] consider the supercritical fluid as a compressed gas; in contrast, the solution models [9e15] look upon the supercritical fluid as an expanded liquid, where the activity coefficient of solute * Corresponding author. E-mail address: zyqlyg@jou.edu.cn (Y. Zhao). https://doi.org/10.1016/j.fluid.2019.112376 0378-3812/© 2019 Elsevier B.V. All rights reserved. depends on the supercritical fluid density. The commonly used activity coefficient models (UNIQUAC [26], UNIFAC [27,28], COSMORS [29], COSMO-SAC [30], NRTL-SAC [31], FH-HSP [32], etc) have been widely used in Vapor-Liquid Equilibrium (VLE), Liquid-Liquid Equilibrium (LLE) and Solid-liquid equilibrium (SLE) predictions. Since the activity coefficient of solute for these activity coefficient models [26e32] is a function of liquid mixture composition and temperature, which is nearly irrelative with the density of fluid mixtures; as a result, they can not be applied to SFE solubility predictions. The COSMO-VAC model [14] utilizes solvent vacancy to account for the variation of solute activity coefficient as a function of fluid density; this model is capable of predicting the solid solute solubility in liquid solvent and supercritical carbon dioxide at high pressures. Our previous work [15] developed an empirical model based on UNIFAC, in which the pressure influence on solute activity coefficient is taken into account by a power exponent for the reduced density; this model fails to predict the solubility minimum and solubility maximum behavior for SFE(i.e., the naphthaleneethylene system [33]). Most of the activity coefficient models (UNIQUAC [26], UNIFAC [27,28], COSMO-RS [29], COSMO-SAC [30], NRTL-SAC [31], FH-HSP [32], etc) are incapable of SFE solubility prediction, COSMO-VAC [14] and our previous model [15] fail to 2 Y. Zhao et al. / Fluid Phase Equilibria 506 (2020) 112376 predict the solubility extrema; therefore, there is a need for a new density-based UNIFAC model to be developed. In this work, we proposed a density-based UNIFAC model, where the activity coefficient of solute is a function of fluid density, composition and temperature. The model parameters were estimated, and the solubility prediction capability of this model was verified with successful results. Then the two mean intra-molecular potential energy terms are nearly identical [38]. ðaÞ ðbÞ < uintra;i > z < uintra;i > (5b) Substituting eqs (4) and (5b) into eq (3), we obtain ðbÞ ðaÞ U inter;i U inter;i V ðbÞ Ki ¼ ðaÞ ¼ ðaÞ exp RT V x ðbÞ xi 2. Theory (6) i 2.1. Solid-liquid equilibrium equation ðaÞ The solubility of solid solute in supercritical fluid (expanded liquid) is determined by the following equation [15,34]. lnðxi gi Þ ¼ ðP V S V L i i RT dp DHm RT 1 P0 T Tm (1) where DHm and Tm are enthalpy of fusion and the melting point of the solute correspondingly; V Si and V Li are the molar volume of solute in solid and liquid states correspondingly; T and P are system temperature and pressure correspondingly; R is the gas constant; P0 is the reference pressure, here P0 is selected as P0 ¼ 1atm. xi is the solute solubility (mole fraction); gi is the activity coefficient of solute in supercritical fluid (expanded liquid), which is calculated by the density-based UNIFAC model. Wang and Lin proposed to estimate the solid molar volume V Si by the co-volume term of Peng-Robinson (PR) EOS [8]. V Si . ¼ 0:0778RTc Pc (2) where Tc and Pc are the critical temperature and critical pressure, respectively; and the liquid molar volume V Li is determined from PR EOS. 2.2. Density-based UNIFAC model The statistical mechanical expression of equilibrium ratio is given as [35]. ðbÞ ðaÞ ðbÞ U p;i U p;i x V ðbÞ Ki ¼ ði aÞ ¼ ðaÞ exp RT V xi (3) whereV(a)and V(b)are the molar volume of mixture in fluid phases a ðaÞ and b correspondingly; xi is the mole fraction of species i in fluid ðaÞ phase a; U p;i is the mean molar potential energy of component i in ðaÞ ðaÞ ðaÞ fluid mixture, U p;i ¼ NA < ui > , NA is Avogadro's constant, < ui > is the ensemble average potential energy experienced by a molecule i in fluid mixture a. The average potential energy for one molecule < ui> contains mean intra-molecular potential energy and mean intermolecular potential energy [36,37]. < ui> ¼ < uintra,i > þ < uinter,i> ðVÞ ðLÞ ðaÞ ðaÞ ðbÞ Eii ¼ Eii RT ln V i Eii ðrÞ ¼ Eii ðr0 Þ RT lnðr0 = rÞ (7) (8a) or Eii ðPÞ ¼ Eii ðP0 Þ RT lnðr0 = rÞ (8b) where r0 is the reference density of pure liquid i at pressure P0, r is the fluid density at system pressure P. According to UNIFAC model [27], the group-interaction parameter is defined as aij ¼ Uij Ujj R (9) where Uij characterizes the interaction energy between groups i and j. Bruin and Prausnitz [40] suggested an estimation equation of Ujj Ujj ¼ cEjj (10) where c is a pre-factor, Ejj is the total molar cohesion energy of similar (like) groups i. The combination of eqs (9) and (10) gives aij ¼ Ejj Eij c R (11a) ðv;uÞ . ðu;uÞ ðv;uÞ c R ¼ Ejj Eij (11b) ðu;uÞ . ðu;uÞ ðu;uÞ c R ¼ Ell Ekl (11c) ðu;vÞ . ðv;vÞ ðu;vÞ c R ¼ Eii Eji (11d) (4) (5a) ðbÞ Vi where Eii is the total molar cohesion energy of similar (like) molecules (or the molar internal energy change on vaporization for ðoÞ ðoÞ ðoÞ pure liquid i) [39], Eii ¼ U inter;i ¼ NA < uinter;i > , < uinter;i > is the mean intermolecular potential energy deserved by one molecule of ðoÞ pure liquid i, U inter;i is the mean molar intermolecular potential energy of pure liquid i at system temperature and pressure. If fluid b is chosen as the reference state, then eq (7) becomes aij The Monte Carlo simulation results of Jorgensen et al. [36] showed that the value of mean molar intra-molecular potential energy for pure components (alkanes, alcohols, ethers, amides, ketones, amines, etc) generally makes little difference in going from liquid phase to vapor phase U intra;i zU intra;i where U inter;i is the mean molar intermolecular potential energy of species i in liquid mixture system a. >Equations (3) and (6) are the statistical mechanical expression ðaÞ ðbÞ of equilibrium ratio for fluid mixtures. For xi ¼ xi ¼ 1(pure fluid), we can obtain the relationship between the total molar cohesive energy of similar (like) molecules and the molar volume of pure fluid from equation (6) akl aji where the superscripts v and u represent solvent and solute ðu;uÞ correspondingly, Ejj is the total molar cohesive energy of similar ðv;uÞ (like) groups j from solute molecule, Eij represents the total Y. Zhao et al. / Fluid Phase Equilibria 506 (2020) 112376 molar cohesion energy between group i from solvent molecule and ðu;uÞ group j from solute molecule, akl is the interaction parameter between dissimilar (unlike) groups k and l from solute molecule, ðv;uÞ aij denotes the interaction parameter between group i from solvent molecule and group j from solute molecule. Equations (8a) and (8b) were developed for pure components, namely eqs (8a) and (8b) can be applied to groups from the same type of molecules ðv;vÞ ðv;vÞ ðu;vÞ (Ejj Ekl ). For groups form different type of molecules (Eji ), the ðu;vÞ relationship between Eji and fluid density r is unavailable; we ðu;vÞ presume that Eji follows a similar law of fluid density as that of ðv;vÞ pure solvent groupsEii . Considering the change in aij due to system pressure variation at constant temperature, we obtain 3 ðv;uÞ gases); for ideal gas (r ¼ 0), the value of DEij will be infinity, an empirical parameter c is introduced to resolve this issue ðs;tÞ c ¼ 4ij ðs;tÞ ½1 expð 60ar = r0 Þ 0 1 ¼ 4ij . ðb þ zÞ (14a) ð i ¼ j and s ¼ t Þ ðnotð i ¼ j and s ¼ t ÞÞ (14b) ðv;uÞ ðv;uÞ ðu;uÞ ðv;uÞ c R Daðv;uÞ ¼ aij ðPÞ aij ðP0 Þ ¼ DEjj DEij ij (12a) DEjjðu;uÞ ¼ Ejjðu;uÞ ðPÞ Ejjðu;uÞ ðP0 Þ (12b) where z is the compressibility factor of fluid, s and t denote solvent molecule or solute molecule, a and b are adjustable parameters. As shown in eqs (13c) and (14a), in the limiting case of low fluid density (r/0), the pre-factor c approaches zero (c/0), therefore a low density compensation term should be added to rectify this residual error. The molecular dynamics simulation results of Lennard-Jone fluid by Nicolas et al. [41] demonstrated that the depth of the potential well is proportional to the critical temperature of component DEijðv;uÞ ¼ Eijðv;uÞ ðPÞ Eijðv;uÞ ðP0 Þ (12c) Tc;i εii ¼ kB 1:35 . The system pressure has little influence on the density of solid solute (hypothetical sub-cooled liquid), while it has a significant impact on fluid density; as a result, combining eqs. (8b), (12a), (12b) and (12c) yields DEjjðu;uÞ z0 (13a) DEðv;uÞ z RT lnðr0 = rÞ ij (13b) Daðv;uÞ zcT lnðr0 = rÞ ij (13c) The molar cohesive energy of like groups from solute molecules ðu;uÞ Ejj is nearly independent of system pressure (eq (13a)), the molar ðv;uÞ cohesive energy of like groups from different moleculesEjj is sensitive to fluid density (eq (13b)). As can be seen from eqs (11a)e(11d) and eqs (13a)-(13c), for like groups from the same type of molecules, the value of group-interaction parameter for this pair of groups is zero ðu;uÞ aii . ðu;uÞ ðu;uÞ c R¼0 ¼ Eii Eii . ðv;vÞ ðv;vÞ ðv;vÞ ajj ¼ Ejj Ejj c R¼0 (13d) (13e) And the change in group-interaction parameter due to solvent density variance is also zero Daiiðu;uÞ ¼ 0 (13f) Daðv;vÞ jj (13g) ¼0 ðv;uÞ . ðu;uÞ ðv;uÞ c R ¼ Ejj Ejj (13h) And the change in group-interaction parameter due to solvent density variance is also nonzero Daðv;uÞ zcT lnðr0 = rÞ jj where kB is Boltzmann's constant, then the energy parameter for unlike molecules is 0:5 εij Tc;i Tc;j ¼ kB 1:35 (15b) The combination of eqs (15a) and (15b) yields εij . εjj ¼ Tc;i Tc;j 0:5 (15c) Our previous work [38] illustrated that the energy parameter of dispersive interactions for like molecules is proportional to the total cavity surface area of molecule from conductor-like screening model εii fA0:5 COSMO;i kB (15d) Similarly, the total molar cohesive energy of like groups (k) from the same type of molecules (solvent) is proportional to the surface area parameter of group ðv;vÞ ðvÞ Ekk f Q k 0:5 (15e) Then the expression of the total molar cohesive energy between group k from solvent molecule and group l from solute molecule can be obtained ðv;uÞ Ekl ðvÞ ðuÞ f Qk Ql 0:25 (15f) and For like groups from different type of molecules, the value of group-interaction parameter for this pair of groups is not zero ajj (15a) (13i) Equation (13b) is valid for dense fluids (liquids and compressed ðv;uÞ ðEkl . ðu;uÞ Ell ðvÞ ðuÞ f Q k =Q l 0:25 (15g) From eqs (15c) and (15g), we have ðv;uÞ ðEkl . ðu;uÞ Ell ðv;vÞ ðu;uÞ f Tc Tc 0:5 ðvÞ ðuÞ 0:25 Q k =Q l (15h) As can be seen from eq (11b), as the fluid density approaches zero (r/0), the fluid mixture becomes ideal gas, and the total cohesion energy between group i from solvent molecule and group ðv;uÞ j from solute molecule will close in on zero (Eij /0); whereas the total cohesion energy of pure solid solute (hypothetical sub-cooled liquid) makes little change with the alteration of fluid density (eq 4 Y. Zhao et al. / Fluid Phase Equilibria 506 (2020) 112376 ðv;vÞ Solvent Cosolvent a CO2 C2H6 C2H4 CHF3 CClF3 acetone ethyl acetate ethanol 1-propanol 2-propanol methanol Tc (K) Pc (bar) u r0a (kg/m3) Ref. 304.25 305.35 282.40 299.25 301.95 508.2 523.2 513.9 536.8 508.3 512.6 73.8 48.8 50.3 49.5 39.2 46.6 38.3 61.4 51.7 47.6 80.9 0.225 0.098 0.089 0.272 0.172 0.318 0.362 0.635 0.623 0.665 0.556 771.20 380.36 440.84 819.00 924.00 790.55 900.63 789.30 803.60 786.30 791.30 [15,50] [15,50] [15,50] [50,57] [50,57] [15] [15] [15] [15] [15] [15] ðv;uÞ For r ¼ 0 and aij ðv;uÞ aij ðr ¼ 0Þ ðv;uÞ aij ðr ¼ r0 Þ ðv;uÞ ðv;uÞ aij Rk Qk Ref. CO2 C2H6 C2H4 CHF3 CClF3 C5HNa caffeine theobromine 1.0020 1.8022 0.9946b 1.6360 2.1721 2.3347 3.2873 2.9840 0.8800 1.6960 1.3448b 1.6060 2.1000 0.9930 2.6331 2.4396 [15] [15] [15] Sub-group of “Pyridine”. Estimated from experimental solubility data via eq (20). ðv;uÞ aij r¼0 (16c) ðr0 Þ > 0, eq (16a) can be rewritten as ðv;vÞ ðu;uÞ Tc ¼ 1 þ Tc For r ¼ 0 and aij group a 8 < 1 ðx > 0Þ SgnðxÞ ¼ 0 ðx ¼ 0Þ : 1 ðx < 0Þ Reference density of supercritical solvent at T ¼ 20 C and P ¼ 1 atm. Table 1b Group Volume and Surface area parameters. b ðu;uÞ molecule; T c andT c are the critical temperature of solvent and solute correspondingly; tanh(x) is the hyperbolic tangent function, tanh(x)¼ (exe-x)/(ex þ e-x); and Sgn(x) is the sign function Table 1a Physical properties of supercritical solvents. ðvÞ Qi . ðuÞ 0:25 Qj ðu;uÞ ¼ 1 T ðv;vÞ Tc c 0:5 ðvÞ Qi . ðuÞ 0:25 Qj As shown from eqs (16d) and (16e), Sgn(x) function ensures that ðv;uÞ ðr ¼ 0Þ > aij ðr ¼ r0 Þ. The attenuation factors of fluid density in eq (16a), expð xpfÞand tanh(x), ensure that as the fluid density increases the density compensation term (the right side of Eq (16a)) decreases rapidly. ðv;uÞ The resulting equation of aij is given below ðv;uÞ aij h ðv;uÞ ðv;uÞ ðrÞ ¼ aij ðr0 Þ 1 þ Sgn aij ðr0 Þ expð xpfÞ ðv;vÞ ðu;uÞ 0:5 ðvÞ . ðuÞ 0:25 i Qi Qj Tc Tc n . o ðs;tÞ þT lnðr0 =rÞ 4ij ½1 expð 60ar=r0 Þ ðb þ zÞ (17a) ðv;uÞ ðv;uÞ ðu;uÞ ðv;uÞ ðr/0Þ ðv;uÞ (16e) aij (13a)). As a result, the value of aij ðr /0Þ will become more ðv;uÞ positive, and the low density compensation term of aij can be approximated semi-empirically aij (16d) ðr0 Þ < 0, eq (16a) becomes r ¼ r0 0:5 ðv;vÞ ðu;uÞ ðv;uÞ Tc 1 ¼ Sgn aij ðr0 Þ expð xpfÞ T c Similarity, we can obtain the expressions of akl ðu;vÞ and aji 0:5 ðu;uÞ aij ðr ¼ r0 Þ . ðvÞ ðuÞ 0:25 Qi Qj (16a) f ¼ tanhðpr = r0 Þ (16b) akl ðvÞ where Q i is the surface area parameter for group i from solvent ðu;vÞ aji ðu;uÞ ðPÞzakl ðu;vÞ ðrÞ ¼ aji ðP0 Þ (17b) ðr0 Þ½1 expð 60ar = r0 Þ (17c) where the interaction parameter of dissimilar (unlike) groups k and ðu;uÞ l from solute moleculeakl is nearly independent of system pressure P. In the limiting case of low fluid density (r/0), the Table 2 Physical properties of solid solutes. Solute M (g/mol) Tm (K) DHm (kJ/mol) Tc (K) Pc (bar) u Ref. naphthalene anthracene phenanthrene 2,6-Dimethylnaphthalene 2,3-Dimethylnaphthalene benzoic acid hexachloroethane biphenyl phenol p-chlorophenol 2,4-dichlorophenol fluorene triphenylmethane naproxen 2-naphthol 2-aminofluorene acridine pyrene caffeine theobromine 128.174 178.234 178.234 156.227 156.227 122.124 236.740 154.210 94.113 128.550 163.000 166.220 244.330 230.263 144.170 181.230 179.220 202.260 194.200 180.18 353.35 489.65 369.45 383.29 377.47 395.50 458.00 342.08 314.04 316.00 318.00 387.92 365.45 426.15 395.15 404.15 384.00 423.15 511.15 627.15 18.875 28.972 18.715 25.060 19.353 18.020 9.750 18.570 11.510 14.070 20.100 19.580 21.500 26.023\ 17.510 22.327 19.700 17.110 21.118 41.112 748.15 869.30 882.55 777.00 785.00 751.00 714.60 773.00 694.15 738.00 710.61 826.00 863.00 889.69 812.13 912.70 883.15 936.00 706.57 817.83 40.500 31.200 31.700 32.200 32.200 44.700 34.500 33.800 61.300 53.200 50.800 30.000 22.400 26.680 47.366 35.473 31.900 26.100 42.700 40.900 0.302000 0.353100 0.329900 0.420100 0.424000 0.602790 0.163000 0.402870 0.440000 0.484882 0.539112 0.404160 0.576000 0.966890 0.576533 0.665725 0.498000 0.507416 0.948000 0.725250 [15,51] [25,51] [15,51] [25,51] [25,51] [25,51] [15,51] [15,51] [25,51] [52] [52] [51,52] [51,52] [15,52] [25] [53] [25] [52] [38,59] [50,59] 1729.64 2719.65 714.41 113.72 2798.38 472.10 2157.25 ACH CH ACOH ACNH2 C5HN COOH 0.00 113.00 526.46 185.42 234.25 385.57 289.03 452.81 743.64 5075.79 2827.77 1657.66 1843.19 aij (K) CHF3 ACH COOH ACNH2 C5HN ACOH CHF3 ACH COOH ACNH2 C5HN ACOH 0.00 238.30 361.87 391.18 445.23 314.89 802.50 1915.59 2387.41 1510.64 2398.71 Table 6 Group Interaction Parameters at reference fluid density for Chlorotrifluoromethane. 1623.72 564.05 63.39 theobromine CH3CO ACCl C2H6 C2H6 ACH CH ACOH ACNH2 C5HN COOH aij (K) CClF3 ACH COOH ACOH ACNH2 C5HN CClF3 ACH COOH ACOH ACNH2 C5HN 0.00 287.39 424.53 365.19 471.85 533.32 893.96 2385.94 3195.4 3611.81 2193.17 interaction parameter between group j from solute molecule and ðu;vÞ group i from solvent molecule approaches zero (aji /0). For mixed solvent (supercritical solvent þ cosolvent) systems, the reference density of fluid mixture r0 is given by ðM xs þ Mc xc Þ Ms xs Mc xc þ r r r0 ¼ s s0 816.40 CH3O a ACOH aij (K) Table 5 Group Interaction Parameters at reference fluid density for Fluoroform. a 220.20 213.70 0.00 800.10a 1599.20a 16.05a 1835.00a 3266.43 200.70a 72.68 179.68 952.46 253.56 827.25 821.16 592.70 468.65 CO2 ACH CH2 COOH CH3O ACOH ACCl ACNH2 C5HN CH3COO CH3CO CH3OH OH theobromine caffeine UNIFAC group-interaction parameters from Ref. [15]. 1149.10 a COOH a CH2 a ACH CO2 aij (K) Table 3 Group Interaction Parameters at reference fluid density for Carbon Dioxide. 5 Table 4 Group Interaction Parameters at reference fluid density for Ethane. a ACNH2 C5HN CH3COO CH3OH OH caffeine Y. Zhao et al. / Fluid Phase Equilibria 506 (2020) 112376 (18) c0 where xs and xc are the mole fraction (solute free) of solvent and cosolvent correspondingly, Ms and Mc are the molecular weight of solvent and cosolvent correspondingly, rs0 and rc0 are the reference density (at T ¼ 20 C and P ¼ 1 atm) of solvent and cosolvent correspondingly. The group parameters Rk and Qk of supercritical solvents are obtained from the van der Waals group volume and surface areas Vwk and Awk given by Bondi [42]: Rk ¼ Vwk/15.17 (19a) Qk ¼ Awk /(2.5 109) (19b) ðv;uÞ The model parameters (aij ðr0 Þ, a, b, x) are estimated by minimizing the following objective function f ðv;uÞ aij ðr0 Þ; a; b; x 100% vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi u n . 2 u1 X ¼ ARE ¼ t ðxcalc xexpt 1 i i n i¼1 (20) where xexpt is the experimental value of solute solubility;xcalc is the i i calculated value of solute solubility via eq (1), where the activity 6 Y. Zhao et al. / Fluid Phase Equilibria 506 (2020) 112376 Table 7a Summary of the estimated model Parameters and Accuracy for solids solubility prediction in pure supercritical solvents a. Solvent No. of Solutes Nd a x b ARE Ref. CO2 C2H6 C2H4 CHF3 CClF3 18 8 6 7 6 453 237 143 72 68 0.638839 0.146707 0.344120 0.156616 0.902956 4.76109 1.51234 1.95355 3.83015 4.65904 1.023980 0.996624 0.600463 0.956236 1.746680 19.1% 18.7% 21.6% 20.4% 12.8% [7,49,54e57,59] [7,57] [7,55,58] [57] [57] a Detailed information of estimated model parameters and accuracy in solubility prediction for solid solutes in various supercritical solvents refers to the Supporting Material. Table 7b Solubility calculation results for Solid Solute in Supercritical Carbon Dioxide with Cosolvent. Solid Solute Solvent naproxen CO2 CO2 CO2 CO2 CO2 CO2 þ þ þ þ þ þ Ethyl Acetate Acetone Methanol Ethanol 1-Propanol 2-Propanol a x b Nd ARE Ref. 0.638839 4.761090 1.023983 18 33 26 24 15 22 6.87% 9.75% 8.95% 7.97% 12.11% 7.01% [49] [49] [49] [49] [49] [49] Table 8 Comparison of solubility prediction accuracy between this model and PR EOS method. T ( C) System Naphthalene þ CO2 2,3-Dimethylnaphthalene þ CO2 Naphthalene þ C2H6 Phenanthrene þ C2H4 a b 35 55 35 45 55 20 25 35 35 45 55 45 55 65 Nd 9 16 5 5 5 7 7 6 15 12 14 5 5 5 This model PR EOS parameters ARE k12 a ¼ 0.638839 5.66% 19.81% 9.93% 6.53% 17.53% 26.00% 20.35% 23.67% 20.29% 18.90% 30.55% 14.78% 9.37% 13.66% 0.0980 0.0992 0.0989 0.1042 0.1078 0.0427 0.0457 0.0386 0.0448 0.0446 0.0448 0.0459 0.0355 0.0318 x ¼ 4.761090 b ¼ 1.023983 a ¼ 0.146707 x ¼ 1.512340 b ¼ 0.996624 a ¼ 0.344120 x ¼ 1.953550 b ¼ 0.600463 Ref. ARE a a a a a a a a a a a b b b 13.48% 28.46% 18.50% 30.46% 22.57% 27.77% 21.31% 36.60% 37.21% 38.54% 44.07% 55.85% 49.12% 42.77% [54] [54] [55] [55] [55] [7] [7] [7] [57] [57] [57] [55] [55] [55] Values of k12 from Ref. [51]. Values of k12 from Ref. [55]. coefficient of solid solute in supercritical fluid is computed by the original UNIFAC model equations [27] plus the density dependent expressions of group-interaction parameter (eq (17a), (17b) and (17c)). The fitting quality of this model to the experimental solubility data is measured in terms of the average relative error (ARE) defined in eq (20). 3. Results and discussions Fig. 1. Solubility of naphthalene in carbon dioxide at T ¼ 328.15K The reported density calculation accuracy of Hang EOS [43] for carbon dioxide is within 1% around the critical point and within 0.1%e0.2% outside the critical region; and the Hang EOS [43] is more reliable than other equations of state (BWR [44], Martin-Hou [45] and Bender [46]) in the PeV-T property predictions of carbon dioxide. The work of Peng and Robinson [47] showed that the PR EOS [48] is reliable in predicting the critical properties of pure substances and multi-component systems. Based on the reasons mentioned above, we used the Hang EOS [43] to calculate the fluid density of pure carbon dioxide, and utilized the PR EOS to compute the fluid density of other pure supercritical solvents. The density of mixed solvents (supercritical solvent þ cosolvent) was calculated by PR EOS [48] with the binary interaction parameters from Y. Zhao et al. / Fluid Phase Equilibria 506 (2020) 112376 7 Fig. 2. Solubility of 2,3-dimethylnaphthalene in carbon dioxide. literature [49]. The physical properties of the supercritical solvents (carbon dioxide, ethane, ethylene, fluoroform and chlorotrifluoromethane) and cosolvents (ethyl acetate, acetone, methanol, ethanol, 1-propanol, 2-propanol) are shown in Table 1a, where the reference density r0 is chosen as the density of saturated liquid at normal temperatures (20 C); the group volume and surface area parameters of the supercritical solvents and some groups are listed in Table 1b; the physical properties of solid solutes [49e59] are given in Table 2; the regressed group interaction parameters at ðv;uÞ reference density aij ðr0 Þfor supercritical solvents (CO2, C2H6, CHF3 and CClF3) are given in Tables 3e6. The estimated model parameter (a, b, x) and the average relative error (ARE) in solubility calculation are illustrated in Table 7a, 7b and 8 (Tables S1eS5 in Supporting Material), and some solubility calculation results compared with the experimental solubility data [49,54,55,59] are depicted in Figs. 1e7; from which we can observe that this model describes the solubility behavior of solid solute in pure supercritical solvents and mixed solvents (supercritical solvent þ cosolvent) successfully. For system temperature T ¼ 333.1 K and pressure P ¼ 179.3 bar with cosolvent composition xs ¼ 1.75% (solute free Fig. 3. Solubility of 2,6-dimethylnaphthalene in ethylene. mole fraction), the solubilities of naproxen (mole fraction naproxen 105) in supercritical carbon dioxide with cosolvents are 6.09, 5.26, 9.53, 9.55, 11.20 and 10.84 for acetone, ethyl acetate, methanol, ethanol, 1-propanol and 2-propanol correspondingly [49]; which shows that the best cosolvent for naproxen is 1propanol. We can see from the accuracy of solubility calculation (Table 7a) that the estimated group interaction parameters at reference density for supercritical solvents (CO2, C2H6, CHF3 and CClF3) exhibit good compatibility with the UNIFAC VLE group interaction parameters. For the solid SoluteeC2H4 systems, where the entire ethylene molecule was treated as one functional group and the UNIFAC VLE group interaction parameters of “C]C” [27,28] were used here, satisfactory solubility prediction results was obtained for these systems (Table 7a, Table S3 in Supporting Material). This density based UNIFAC model has been examined for the solubility of 20 solid solutes in 5 supercritical solvents (CO2, C2H6, C2H4, CHF3 and CClF3) over a pressure range of 42.3e490.4 bar, a temperature range of 20e70 C and a wide solubility range of 106101 (51 systems, 1111 data points), the average relative errors in predicted solubility are 19.1%, 18.7%, 21.6%, 20.4% and 12.8% for CO2, C2H6, C2H4, CHF3 and CClF3 correspondingly; which demonstrates that this density-based UNIFAC model is simple and reliable for practical application. For CaffeineCO2 and TheobromineCO2 systems, both caffeine and theobromine molecules contain two adjacent rings [59], and the structures of each ring for caffeine and theobromine are different form that of cyclohexane, benzene and pyridine; as a result, the whole molecule is treated as one UNIFAC group for the two solid solutes; here, the UNIFAC model is reduced to the UNIQUAC method; this means that the similar forms of equations (17a), (17b) and (17c) may be applied to other activity coefficient models (UNIQUAC [26], Wilson [60], NRTL [61]) containing binary interaction parameters for SFE solubility predictions. A comparison of solubility prediction accuracy between the original UNIFAC, density based UNIFAC and PR EOS method is given in Fig. 1 and Table 8; which shows that both the proposed model and PR EOS can describe solubility extrema (solubility minimum and solubility maximum), and the original UNIFAC model gives far worse solubility prediction results than density based UNIFAC and PR EOS. In addition, the binary interaction parameter k12 for EOS methods is temperature and solute-solvent system dependent; 8 Y. Zhao et al. / Fluid Phase Equilibria 506 (2020) 112376 Fig. 4. Solubility of Naproxen in supercritical Carbon Dioxide. with Ethyl Acetate Cosolvent at T ¼ 333.1K. Fig. 5. Solubility of Naproxen in supercritical Carbon Dioxide. with Ethanol Cosolvent at T ¼ 333.1K. while the parameters of this model (a, b, x) are irrelative to system temperature, this may be due to that these parameters (a, b, x) were estimated via experimental data from different system temperatures. For some solute-solvent systems (Table 7a and 7b, Tables S1eS5 in Supporting Material), the values of model parameters (a, b, x) seem to be universal constants. The origin of the prediction error of this model may come from experimental error in solubility data or from the uncertainty in solute property values (DHm, Tc, Pc, u) [5]. For example, the uncertainty (or reproducibility) of the experimental solubility data from the same literature was reported [48] within ±5%; the agreement of solubility results for the same solutesolvent system between different reference sources was reported [57] within ±10%. 4. Conclusions A new density-based UNIFAC model for SFE Solubility prediction was presented in this work, in which the group interaction parameter is a function of supercritical fluid density. The group interaction parameters for supercritical solvents (carbon dioxide, ethane, fluoroform and chlorotrifluoromethane) at reference fluid density were estimated, which manifest good compatibility with the UNIFAC VLE group interaction parameters. For the solid SoluteeC2H4 systems, satisfactory solubility prediction results can be obtained using the UNIFAC VLE group interaction parameters of “C]C”. This new model has been verified for the solubility of 20 solid solutes in five superficial solvents (CO2, C2H6, C2H4, CHF3 and CClF3) over a pressure range of 42.3e490.4 bar, a temperature range of 20e70 C and a wide solubility range of 106101 (51 systems, 1111 data points) with successful results; which indicates that this model is simple and reliable for practical application. This densitybased UNIFAC model is comparable to the PR EOS method with regard to the accuracy of SFE solubility prediction. For some solutesolvent systems, the values of model parameters (a, b, x) Y. Zhao et al. / Fluid Phase Equilibria 506 (2020) 112376 9 Fig. 6. Solubility of Caffeine in supercritical Carbon Dioxide at T ¼ 333.15 K. Fig. 7. Solubility of Theobromine in supercritical Carbon Dioxide. seem to be universal constants. When each UNIFAC group is one molecule, the density dependent expression of group interaction parameter derived in this model may be applicable to other activity coefficient models (UNIQUAC, Wilson, NRTL) containing binary interaction parameters, which can expand the application range of these activity coefficient models from liquid phase to supercritical regime. This model is reliable for SFE solubility prediction of solid solute in mixed solvents. Acknowledgment We gratefully acknowledge financial Support by a Project Funded by the Priority Academic Program Development of Jiangsu Higher Education Institutions, the Public science and technology research funds projects of ocean (201505023). List of symbols Declaration of competingf interest Does the appearance of financial or other “material” interests by authors of scientific papers present a more serious bias than other competing interests so as to require a signed Conflict of Interest (COI) statement? NO, this paper does not require a signed Conflict of Interest (COI) statement. ACH ðv;uÞ aij ARE Eij aromatic carbon group interaction parameter between group i from solvent molecule and group j from solute molecule, K average relative error in solubility prediction defined via eq(20) total molar cohesive energy of dissimilar (unlike) groups i and j, J/mol 10 Ejj DHm kij M Nd P P0 Pc Qk R Rk T Tc Tm ðoÞ U inter;i V z Y. Zhao et al. / Fluid Phase Equilibria 506 (2020) 112376 total molar cohesive energy of similar (like) groups i, J/ mol enthalpy of fusion for solid solute, kJ/mol binary interaction parameter for equation of state molecular weight, g/mol number of experimental data points pressure, bar reference pressure, bar critical pressure, bar surface area parameter for group k gas constant volume parameter for group k system temperature, K critical temperature, K melting point of solid solute, K mean molar intermolecular potential energy of pure liquid i at system temperature and pressure, J/mol molar volume, m3/mol compressibility factor Greek Letters g activity coefficient r0 reference density of supercritical solvent, or the density of corresponding pure liquid solvent at normal temperatures, kg/m3 r the fluid density at system temperature and pressure, kg/m3 u acentric factor Subscripts i, j, k, l group index Superscripts Calc calculated expt experimental u solute v solvent Appendix A. Supplementary data Supplementary data to this article can be found online at https://doi.org/10.1016/j.fluid.2019.112376. References ~ ez, Sub-and supercritical fluid extraction of [1] M. Herrero, A. Cifuentes, E. Iban functional ingredients from different natural sources: plants, food-byproducts, algae and microalgae: a review, Food Chem. 98 (2006) 136e148. [2] J.M. Prausnitz, R.N. Lichtenthaler, E.G. de Azevedo, Molecular Thermodynamics of Fluid-phase Equilibria, third ed., Prentice-Hall, London, 1999. [3] P. Coimbra, C.M.M. Duarte, H.C. de Sousa, Cubic equation-of-state correlation of the solubility of some anti-inflammatory drugs in supercritical carbon dioxide, Fluid Phase Equilib. 239 (2006) 188e199. [4] M. Shamsipur, J. Fasihi, A. Khanchi, Y. Yamini, A. Valinezhad, H. Sharghi, Solubilities of some 9,10-anthraquinone derivatives in supercritical carbon dioxide:a cubic equation of state correlation, J. Supercrit. Fluids 47 (2008) 154e160. [5] D.J. Quiram, J.P. O'Conell, H.D. Cochran, The solubility of solids in compressed gases, J. Supercrit. Fluids 7 (1994) 159e164. [6] K.P. Johnston, C.A. Eckert, An analytical Carnahan-Starling-van der Waals model for solubility of hydrocarbon solids in supercritical fluids, AIChE J. 27 (1981) 773e779. [7] K.P. Johnston, D.H. Ziger, C.A. Eckert, Solubilities of hydrocarbon solids in supercritical fluids. The augmented van der Waals treatment, Ind. Eng. Chem. Fundam. 21 (1982) 191e197. [8] L.H. Wang, S.T. Lin, A predictive method for the solubility of drug in supercritical carbon dioxide, J. Supercrit. Fluids 85 (2014) 81e88. [9] M.E. Mackay, M.E. Paulaitis, Solid solubilities of heavy hydrocarbons in supercritical solvents, Ind. Eng. Chem. Fundam. 18 (1979) 149e153. [10] H.D. Cochran, L.L. Lee, D.M. Pfund, Application of the KirkwoodBuff theory of solutions to dilute supercritical mixtures, Fluid Phase Equilib. 34 (1987) 219e234. [11] H.Y. Shin, K. Matsumoto, H. Higashi, Y. Iwai, Y. Arai, Development of a solution model to correlate solubilities of inorganic compounds in water vapor under high temperatures and pressures, J. Supercrit. Fluids 21 (2001) 105e110. [12] J.S. Cheng, Y.P. Tang, Y.P. Chen, Calculation of solid solubility of complex molecules in supercritical carbon dioxide using a solution model approach, Mol. Simul. 29 (2003) 749e754. [13] C.S. Su, Y.P. Chen, Correlation for the solubilities of pharmaceutical compounds in supercritical carbon dioxide, Fluid Phase Equilib. 254 (2007) 167e173. [14] Y. Shimoyama, Y. Iwai, Development of activity coefficient model based on COSMO method for prediction of solubilities of solid solutes in supercritical carbon dioxide, J. Supercrit. Fluids 50 (2009) 210e217. [15] Y. Zhao, W. Liu, Z. Wu, Solubility model of solid solute in supercritical fluid solvent based on UNIFAC, Ind. Eng. Chem. Res. 49 (2010) 5952e5957. [16] J. Chrastil, Solubility of solids and liquids in supercritical gases, J. Phys. Chem. 86 (1982) 3016e3021. [17] Y. Adachi, B.C.Y. Lu, Supercritical fluid extraction with carbon dioxide and ethylene, Fluid Phase Equilib. 14 (1983) 147e156. [18] D.H. Ziger, C.A. Eckert, Correlation and prediction of solidsupercritical fluid phase equilibria, Ind. Eng. Chem. Process Des. Dev. 22 (1983) 582e588. [19] M. Hojjati, Y. Yamini, M. Khajeh, A. Vetanara, Solubility of some statin drugs in supercritical carbon dioxide and representing the solute solubility data with several density-based correlations, J. Supercrit. Fluids 41 (2007) 187e194. [20] Y. Shimoyama, M. Sonoda, K. Miyazaki, H. Higashi, Y. Iwai, Y. Arai, Measurement of solubilities for rhodium complexes and phosphine ligand in supercritical carbon dioxide, J. Supercrit. Fluids 44 (2008) 266e272. vez, Evaluation of density-basedmodels for [21] D.L. Sparks, R. Hernandez, L.A. Este the solubility of solids in supercritical carbon dioxide and formulation of a new model, Chem. Eng. Sci. 63 (2008) 4292e4301. [22] M. Sauceau, J.J. Letourneau, D. Richon, J. Fages, Enhanced densitybasedmodels for solid compound solubilities in supercritical carbondioxide with cosolvents, Fluid Phase Equilib. 208 (2003) 99e113. [23] S. Li, G.S. Varadarajan, S. Hartland, Solubilities of theobromine and caffeine in supercritical carbon dioxide: correlation with density-based models, Fluid Phase Equilib. 68 (1991) 263e280. [24] G.S. Gurdial, P.A. Wells, N.R. Foster, R.P. Chaplin, The role of polarity in correlations of solid supercritical fluid phase systems, J. Supercrit. Fluids 2 (1989) 85e89. [25] S.E. Guigard, W.H. Stiver, A density-dependent solute solubility parameter for correlating solubilities in supercritical fluids, Ind. Eng. Chem. Res. 37 (1998) 3786e3792. [26] D.S. Abrams, J.M. Prausnitz, Statistical thermodynamics of liquid mixtures: a new expression for the excess Gibbs energy of partly or completely miscible systems, AIChE J. 21 (1975) 116e128. [27] A. Fredenslund, R.L. Jones, J.M. Prausnitz, Group-Contribution estimation of activity coefficients in non-ideal liquid mixtures, AIChE J. 21 (1975) 1086e1099. [28] H.K. Hansen, P. Rasmussen, A. Fredenslund, M. Schiller, J. Gmehling, Vaporliquid equilibria by UNIFAC group contribution. 5. Revision and extension, Ind. Eng. Chem. Res. 30 (1991) 2352e2355. [29] A. Klamt, V. Jonas, T. Bürger, J.C.W. Lohrenz, Refinement and parametrization of COSMO-RS, J. Phys. Chem. A 102 (1998) 5074e5085. [30] S.T. Lin, S.I. Sandler, A priori phase equilibrium prediction from a segment contribution solvation model, Ind. Eng. Chem. Res. 41 (2002) 899e913. [31] C.C. Chen, Y. Song, Solubility modeling with a nonrandom two-liquid segment activity coefficient model, Ind. Eng. Chem. Res. 43 (2004) 8354e8362. [32] T. Lindvig, M.L. Michelsen, G.M. Kontogeorgis, A floryehuggins model based on the hansen solubility parameters, Fluid Phase Equilib. 203 (2002) 247e260. [33] R.T. Kurnik, R.C. Reid, Solubility extrema in solid-fluid equilibria, AIChE J. 27 (1981) 861e863. [34] I. Kikic, M. Lora, A. Bertucco, A thermodynamic analysis of three-phase equilibria in binary and ternary systems for applications in rapid expansion of a supercritical solution (RESS), particles from gas-saturated solutions (PGSS), and supercritical antisolvent (SAS), Ind. Eng. Chem. Res. 36 (1997) 5507e5515. [35] Y. Zhao, Z. Wu, W. Liu, Statistical mechanical theory of fluid mixtures, Phys. A Stat. Mech. Appl. 393 (2014) 62e75. [36] W.L. Jorgensen, D.S. Maxwell, J. Tirado-Rives, Development and testing of the OPLS all-atom force field on conformational energetics and properties of organic liquids, J. Am. Chem. Soc. 118 (1996) 11225e11236. [37] D. Frenkel, B. Smit, Understanding Molecular Simulation: from Algorithms to Applications, Academic, San Diego, 2002. [38] Y. Zhao, Z. Wu, W. Liu, X. Pei, A new theoretical model for predicting the solubility of solid solutes in different solvents, Fluid Phase Equilib. 412 (2016) 123e134. [39] Y. Zhao, W. Liu, J. Zhu, X. Pei, Solvent similarity analysisfrom qualitative to quantitative, Chem. Eng. Sci. 184 (2018) 149e160. [40] S. Bruin, J.M. Prausnitz, One-parameter equation for excess Gibbs energy of strongly nonideal liquid mixtures, Ind. Eng. Chem. Process Des. Dev. 10 (1971) 562e572. [41] J.J. Nicolas, K.E. Gubbins, W.B. Streett, D.J. Tildesley, Equation of state for the Lennard-Jones fluid, Mol. Phys. 37 (1979) 1429e1454. [42] A. Bondi, Physical Properties of Molecular Crystals, Liquids and Glasses, John Wiley& Sons, New York, 1968. Y. Zhao et al. / Fluid Phase Equilibria 506 (2020) 112376 [43] F.H. Huang, M.H. Li, L.L. Lee, K.E. Starling, F.T.H. Chung, An accurate equation of state for carbon dioxide, J. Chem. Eng. Jpn. 18 (1985) 490e496. [44] M. Benedict, G.B. Webb, L.C. Rubin, An empirical equation for thermodynamic properties of light hydrocarbons and their mixtures I. Methane, ethane, propane and n-butane, J. Chem. Phys. 8 (1940) 334e345. [45] J.J. Martin, Y.C. Hou, Development of an equation of state for gases, AIChE J. 1 (1955) 142e151. [46] E. Bender, Equation of state exactly representing the phase behaviors of pure substances, proc. 5th symposium on thermophysical properties, Am. Soc. Mech. Eng. (1970) 227e235. [47] D.Y. Peng, D.B. Robinson, A rigorous method for predicting the critical properties of multi-component systems from an equation of state, AIChE J. 23 (1977) 137e144. [48] D.Y. Peng, D.B. Robinson, A new two-constant equation of state, Ind. Eng. Chem. Fundam. 15 (1976) 59e64. [49] S.S.T. Ting, S.J. Macnaughtn, D.L. Tomasko, N.R. Foster, Solubility of naproxen in supercritical carbon dioxide with and without cosolvents, Ind. Eng. Chem. Res. 32 (1993) 1471e1481. [50] B.E. Poling, J.M. Prausnitz, J.P. O'Connell, The Properties of Gases and Liquids, fifth ed., McGraw Hill, New York, 2001. ndez, L.A. Este vez, Calculation of interaction pa[51] A.C. Caballero, L.N. Herna rameters for binary Solid-SCF Equilibria using several EOS and Mixing rules, J. Supercrit. Fluids 5 (1992) 283e295. 11 [52] P.S. Ma, Experimental Properties Handbook of Organic Compoundssection Including Carbon, Hydrogen, Oxygen, Halogen, Chemical Industry Press, Beijing, 2006. [53] J.S. Chickos, C.M. Braton, D.G. Hesse, J.F. Liebman, Estimating entropies and enthalpies of fusion of organic compounds, J. Org. Chem. 56 (1991) 927e938. [54] M. Mchugh, M.E. Paulaitis, Solid solubilities of naphthalene and biphenyl in supercritical carbon dioxide, J. Chem. Eng. Data 25 (1980) 326e329. [55] R.T. Kurnik, S.J. Holla, R.C. Reid, Solubility of solids in supercritical carbon dioxide and ethylene, J. Chem. Eng. Data 26 (1981) 47e51. [56] R.A. Van Leer, M.E. Paulaitis, Solubilities of phenol and chlorinated phenols in supercritical carbon dioxide, J. Chem. Eng. Data 25 (1980) 257e259. [57] W.J. Schmitt, R.C. Reid, Solubility of monofunctional organic solids in chemically diverse supercritical fluids, J. Chem. Eng. Data 31 (1986) 204e212. [58] G.A.M. Diepen, F.E.C. Scheffer, The solubility of naphthalene in supercritical ethylene, J. Phys. Chem. 57 (1953) 575e577. [59] S. Li, G.S. Varadarajan, S. Hartland, Solubilities of theobromine and caffeine in supercritical carbon dioxide: correlation with density-based models, Fluid Phase Equilib. 68 (1991) 263e280. [60] G.M. Wilson, Vapor-liquid equilibrium. XI. A new expression for the excess free energy of mixing, J. Am. Chem. Soc. 86 (1964) 127e130. [61] H. Renon, J.M. Prausnitz, Local compositions in thermodynamic excess functions for liquid mixtures, AIChE J. 14 (1968) 135e144.