1262 The Journal of Physical Chemistry, Vol. 83, No. 10, 1979
Harbison et
(42) F. J. Miiiero, A. Laferriere, and P. V. Chetirkin, J. Phys. Chem., 81, 1737 (1977). (43) G.Akerlof and G. Kegeies, J. Am. Chem. SOC.,61,1027 (1939). (44) 0. Akerlof and P. Bender, J. Am. Chem. SOC.,63, 1085 (1941). (45) K. S. Pitzer and G. Mayorga, J. Phys. Chem., 77, 2300 (1973). (46) H. S. Harned and 8. B. Owen, "The Physical Chemistry of Electrolyte Solutions", Reinhold, New York, 1958. (47) R. A. Robinsonand R. H. Stokes, "Electrolyte Solutions", Butterworths, London, 1959. (48) L. G. Sillen and A. E. Martell, "Stability Constants of Metal-Ion Complexes", The Chemical Society, London, 1964.
al.
(49) F. J. Miliero, G. K. Ward, and P. V. Chetirkin, J. Acoust. SOC.Am., 61, 1492 (1977). ( 5 0 ) F. J. Miilero in "Water and Aqueous Solutions", R. A. Horne, Ed.,
Wlley-Interscience, New York, 1972. (51) H. S.Harned and B. B. Owen, ref 46,p 405. (52) R. LinderstromLangand C. F. Jacobsen, C . R. Trav. Lab. Cai-fsberg, Ser. Chim., 24, l(1941). (53) D. A. Lown, H. R. Thirsk, and Lord Wynnedones, Trans. Faraday SOC.,64,2073 (1968). (54)J. G. Mathieson and B. E. Conway, J. Solotion Chem.,3, 455 (1974). (55) J. E. Desnoyers and P. R. Philip, Can. J. Chem., 50, 1094 (1972).
Solute Infinite-Dilution Partition Coefficients with Mixtures of Squalane and Dinonyl Phthalate Solvents at 30.0 OC M. W. P. Harblson,? R. J. Laub," D. E. Martlre,?J. H. Purneli,t and P. S. Willlamst Departments of Chemistry, Georgetown Universky, Washington, D.C. 20057, The Ohio State University, Columbus, Ohio 432 IO, and University College of Swansea, Swansea, Wales SA2 BPP, United Kingdom (Received February 16, 1979) Publication costs assisted by the National Science Foundatlon
Using gas-liquid chromatography (GLC), activity and partition coefficients at 30.0 "C were obtained for aliphatic, alicyclic, and aromatic solutes with 12 mixtures of squalane + bis(3,5,5-trimethylhexyl)phthalate (hereafter called dinonyl phthalate) over the mole fraction range 0-1. The random error in the partition coefficients is estimated to be less than 1.0%. The average difference between the GLC results and those determined by a static technique is found to be f0.5%, thus establishing (for the first time) the validity and attainable accuracy of the GLC method for thermodynamic studies of ternary solutions. Theoretical expressions derived from an extension of a conventional (Tompa) solution model are next employed to describe the variation of the solute partition coefficient ( P R ) with the volume fraction ($1 of dinonyl phthalate, and are found to fit the data to within experimental error. However, although the fitted solution parameters are physically reasonable and internally self-consistent, independent tests of the solution model prove it to be inconclusive in the present instance. The data are then examined in light of the diachoric solutions model, wherein P R is said to vary linearly with $ over the entire solvent composition range. Deviations from linearity of up to 9% are observed for the systems at hand and so, it cannot here be claimed that the diachoric solutions hypothesis applies unless it is postulated that other concurrent solution phenomena (such as solvent dimerization) obtain. The success of the diachoric solutions equation in terms of analytical applications (i.e., GLC separations),which is based upon prediction of relatiue partition coefficients, can, however, be rationalized in terms of the current data and is shown to be accurate to &2.5% for the systems herein examined. It is concluded that, in view of the importance of mixed solvent systems in thermodynamics, in chromatography, and in spectroscopy, in light of the apparent conformity to diachoric behavior of hundreds of mixtures of a remarkable range of solute and mixed solvent types, and because of the still-unresolvednature of the situation, additional appropriate experiments are called for.
In an analysis of the great majority of quantitatively useful GLC-based data allowing evaluation of the infinite-dilution partition coefficient (KORm)of a solute component (A) distributed between a binary liquid mixture (B + C) and the gas phase, Laub, Purnell, and cow o r k e r ~ found ~ - ~ that, with few exceptions and for a wide variety of system types, the results were described by the linear relation KOR(M,=
$B@R(~,
+ $c@R(~)
(1)
to within experimental error (which in a few cases reached &lo%),where $i represents the volume fraction of solvent component i and P R G ) (i = B or C) pertains to A in either pure solvent. Solutes of almost all types, and solvent mixtures ranging from those where only van der Waals 'Department of Chemistry, Georgetown University. *Author to whom correspondence should be addressed a t the Department of Chemistry, The Ohio State University. t Department of Chemistry, University College of Swansea. 0022-365417912083-1262$0 1 .OOlO
interaction is expected to those where there is spectroscopic evidence of complexation were found to conform to eq 1. Laub and PurnelP proposed that systems which are described by eq 1 be termed diachoric. Noteworthy at this point is that expressions equivalent to eq 1have, in the past, been employed empirically with success in analytical GLC, beginning with Primavesi6 who showed in 1959 that the solute specific retention volume (vOgcM,)varied linearly with the weight fraction (wi)of solvent component i, viz. =
u.rBvog(~)i"Cvom
(2)
for mixtures of triisobutylene with silver nitrate-ethylene glycol. Another version of eq 1in terms of capacity factors h' [= ( t -~t A ) / t A ] presented by Hildebrand and Reilley6 in 1964 (3) k ' ( ~=) W B ~ ' ( B ) + W C ~ ' ( C ) was found to describe solute elution behavior with mixtures of the solvents, silicone oil with polyethylene glycol (mol 0 1979 American Chemical Society
The Journal of Physical Chemistry, Vol. 83, No. 10, 1979
Partition Coefficients of Miixtures
wt = 400). Equations :2 and 3 have, in the immediate past,
been verified and applied by Klein and Widdecke7and by Laub, Purnell, and c o - w o r k e r ~and , ~ ~ reviewed ~ by Laub and W e l l i n g t ~ n . ~ In a contemporary study based upon conventional models of solutionslO Martire and co-~orkersll-~~ attempted to reconcile thermodynamic and spectroscopic measurements related to mixed-solvent systems and to rationalize diachoric solution behavior. Their approach leads to a general equation which predicts some curvature in plots of KORc against 4c (or of vOgcM,or k'(M) against w M ) . Depengng upon the values of the molecular parameters of the systems in question, this curvature may or may not be significant relative to experimental error (see later). In order to explore the situation further, Laub, Martire, and Purne1114J5recently considered, both experimentally and theoretically, solute-pure-solvent and solute-binary-solvent mixtures of 12-alkanes. It was shown,14first, that activity and partition coefficients of the binary systems could be predicted frorn the Janini-Martirelo modification of Prigogine's treatment. Secondly, an extension15of this study indicated that partition coefficients of n-alkane solutes with binary n-alkane solvents could be calculated from a relation of the form (cf. eq 17) @ R ( ~ ) = $B
In
v'4
@R(,,
+ 4c In @ R ( ~ ) + - x c ( ~ ) ~ B (4) ~c VC
where Vi is the molar volume of species i, xc(s is the additive/solveint F1or:y-type interaction parameter (see later) and wherie the "additive" was taken to be the solvent component of lower molecular weight. Clearly, eq 4 predicts that plots of .KoR(M,vs. 4c will be curved, although the extent of curvature, as noted above, may be slight. For the systems examined (n-pentane through n-octane solutes; n-octadecane-n-hexatriacontane mixed solvents) the deviation from eq 1 was predicted1&to be less than 1% a t 4cls = 0.5. In fact, the linear regression correlation coefficients for each set of experimental data (two end points; seven mixtures) were in excess of 0.9995, and no distinction could be drawn between eq 1 and 4. The only other comprehensive study of ternary n-alkane mixtures (with one component at infinite dilution) known to us is that reported in 1970 by Hayduk and ChenglGfor ethane with n-hiexane-1%-hexadecane at 25 "C. Their data, when plotted as KoR(M, vs. r$c6 (cf. Figure 6.17 of ref 4), deviate negatively from linearity by ca. -17% at 4% = 0.6.17 Recent studiesla-zl of systems other than n-alkanes indicate substantial deviation from eq 1, calling into question its generality even as an approximate description. Accordingly, in view o f the importance of mixed-solvent systems in thermodynamics and spectroscopy, and in view of the still unresolved nature of the situation, additional appropriate experiments were called for. To this end, GLC studies of solute infinite-dilution activity and partition coefficients of aliphatic, alicyclic, and aromatic solutes with squalane (SQ) and dinonyl phthalate (DNP) at 30.00 "C were undertaken and are herein reported. These systems were selected, first, in order to compare GLC data with those extrapolated from precise static measurements of low solute concentration in mixtures of SQ and DNP.18 The static data (n-hexane solute) are noteworthy in that there is indication of deviation from the linearity predicted by eq 1 of up to ca. 7%. In addition, this comparison also permits a comprehensive test of the GLC method of obtaining thermodynamic solution data for binary mixtures of solvents comprising other than n-alkanes and, for the first time, ternary ]mixtures. The work thus extends that previously reported by Laub, Purnell, Williams,
1263
Harbison, and Martire,zz who showed that GLC-derived activity and partition coefficient data agreed to within il% with those obtained via staticz3methods for neat SQ and DNP solvents. Secondly, intimate mixtures of several phases have been used for what has proved to be a completely successful procedure for the prediction of retention (hence, separation) via various forms of eq Simply, separation factors ( a values) are calculated from data pertaining to each of the pure solvents for each solute pair from which, in accordance with the relation 1.318,g124-28
the optimum solvent mixture for the separation of all solutes can be deduced. Further, relative retentions with intimately and mechanically mixed SQ and DNP have commonly proved to be identical. These findings must be reconciled with those studies noted above in which eq 1 was not obeyed.
Experimental Section Materials. Bis(3,5,5-trimethylhexyl)phthalate was obtained from BDH Ltd. Chromatographic analysis (13 ft. X 0.25411. glass column; 10% HI-EFF 8 BP; 240 "C) of this material showed that it contained no more than 70% of the main isomer. NMR analysisz9indicated that the foreign material consisted of positional isomers of the DNP ester groups; GLC data also indicated the presence of about 9% of more highly branched alkyl esters and a little over 20% less highly branched isomers. In all, 11isomers are present in detectable amounts. The presence of these isomers should in no way affect the validity of eq 1, however, since, if the mixture is diachoric, thenz5 n
(5) i.e., KOR will still be linear in 4i. If, on the other hand, nondiacfkc behavior is confirmed the matter cannot be ignored. Purification was not attempted since, in all published studies to date, DNP has been employed as supplied. Squalane (2,6,10,15,19,23-hexamethyltetracosane) was obtained from Applied Science and was used as received. GLC analysis of the material indicated purity in excess of 99%. The stationary liquids and their mixtures were deposited on Chromosorb G (60/80 mesh, acid-washed, and silanized with dimethylchlorosilane) solid support from solution in dichloromethane. The volatile solvent was removed by rotary evaporation and the resultant free-flowing powders were packed by aspiration into 0.25-in. 0.d. coiled stainless steel tubes 4 f t in length. The weight percent of the liquid phase employed ranged from 7 to lo%, and was determined accurately from the weight loss of a t least three separately ashed samples (ca. 1 g each) of coated support.2z~30 A small (0.0021 g/g ashed support) correction was made for the weight loss of bare support. In all, 12 different columns were employed at 30.00 "C, corresponding to the following volume fractions of DNP (dc) in the solvent mixture: 0.000, 0.075, 0.148, 0.215, 0.272, 0.325,0.452, 0.558, 0.711,0.820, 0.907, 1,000. The solutes studied are listed in Table I. Properties of the Pure Materials. Table I gives the molar volumes of the pure liquid components, and the saturation vapor pressures and second virial coefficients of the solutes, all pertaining to 30.00 "C. The molar volumes of SQ and DNP were determined from density
1264
Harbison et al.
The Journal of Physical Chemistry, Vol. 83, No. 10, 1979
TABLE I: Properties of the Pure Components at 30.00 ’C
Partition coefficients were evaluated via the standard relations pertaining to infinite dilution
solutes (A) n-pentane n-hexane n-heptane n-octane cyclohexane methylcyclohexane benzene toluene
-BAA/
V/cm3 mol-’
mmHg
cm3 mol-’
117.03 132.53 148.39 164.48 109.40 129.06 89.95 107.44
614.80 187.10 58.38 18.44 121.73 58.67 119.34 36.66
1146 1808 2717 3964 1612 2401 1412 2508
PQA/
(7) where Vs is the total volume of stationary liquid in the column and V,, the fully corrected (“net”) retention volume, is the product of the corrected flow rate (FC),the carrier compressibility correction factor (j), and the solute = tR retention time adjusted for column dead space ( t ’ ~ - tA). Since the carrier was helium a t inlet pressures of