Van der Waals-like global phase diagrams

Robert L. Scott Department of Chemistry and Biochemistry, University of California, L os Angeles, California 90095, USA Received 28th April 1999, Accepted 2nd July 1999

Since the rst global phase diagram (1970) based upon the van der Waals equation of state for binary mixtures, there have been many newer ones, most recently that obtained by Yelash and Kraska (Ber. Bunsen-Ges. Phys. Chem., 1998, 102 213) by combining the Carnahan and Starling hard-sphere equation of state (J. Chem. Phys., 1969, 51, 635) with the original van der Waals attraction. Of the various such equations the results for the original van der Waals equation seem to be the least satisfactory in comparison with the experimental one-component uid, and probably also for binary mixtures. However the Carnahan and Starling equation diverges only at a density much higher than close-packed. Sanchez (J. Chem. Phys., 1994, 101, 7003) has developed a new 4,3 Pade approximant that ts the virial coefficients of the hard sphere uid and which diverges almost exactly at the close-packed density. If this expression is used in a van der Waals-like equation of state, it produces the same kind of closed-loop diagrams that Yelash and Kraska found, although over a substantially smaller area of the global phase diagram. Extensions of global phase diagrams to mixtures of molecules of dierent size (even to polymer systems) seem promising. In particular a simple equation of state with the Flory mixing equation and the Prigogine entropy parameter (T he Molecular T heory of Solutions, North Holland, Amsterdam, 1957, ch. XVI) has led to a satisfactory account of tricritical points in hydrocarbon systems and should be investigated further.

Introduction

In the early years of this century van Laar1 used the van der Waals equation to calculate a number of binary-mixture phase diagrams, ones we now distinguish as being of three dierent kinds, depending upon the way critical lines appear in the pressuretemperature (p,T ) projection of the threedimensional pressuretemperaturecomposition diagram. However, he restricted himself to the geometric mean for the unlike attraction parameter [a \ (a a )1@2] and, presum12 11 22 ably because of the difficulty of hand calculations, apparently never determined the boundaries between the various types. For decades the van der Waals equation fell into disrepute, and it was not until 1968 that van Konynenburg and Scott2h 4 used the equation to determine the rst modern global phase diagram (shown later as Fig. 5). They found that the equation could account qualitatively for ve of the six major types of binary phase diagrams shown in Fig. 1, all but the one that shows a low-temperature closed loop (later called Type VI5). Since that time, many others have extended the concept, notably Griffiths and coworkers, who examined the threecomponent lattice gas6 and improved our understanding of the two-component van der Waals mixture7 by nding the shield region . However, all the known experimental systems of the closedloop type involve hydrogen bonding between the unlike components, and the standard explanation of the closed loop has involved the introduction of special orientational forces. Barker and Fock8 used the quasi-chemical model to calculate closed-loop diagrams, an approach extended recently by Hueda, Costas and Scott.9 Similar results have been obtained from other models in a number of studies.10 h13 Alternatively one may obtain the same result by introducing temperature Presented at the 1st Workshop on Global Phase Diagrams, (Bunsen-Kolloquium 77), Walberberg, Germany, 21st24th March 1999.

dependent equilibrium constants for complex formation,14 or even more simply by making the a parameter in the van der 12 Waals equation temperature-dependent ; however these last can hardly be called models. We restrict the term van der Waals-like equations of state to those formed by adding the usual van der Waals attraction term ([a/V ) to an equation of state for hard m spheres. Earlier work on non-van der Waals-like equations of state, e.g., the Lennard-Jones calculations of Boshkov,15 had

Fig. 1 Pressuretemperature projections of p, T , x phase diagrams for the six major types of binary mixture behavior. The solid lines are the vapor pressures of the two pure components, the dashed lines are three-phase (L L G) lines, and the dotted lines are critical lines. 2 [Reprinted by 1permission from ref. 22, Copyright, 1987, American Chemical Society.]

Phys. Chem. Chem. Phys., 1999, 1, 4225 4231

4225

led to closed-loop behavior without explicit directional interactions, but all of these involved some temperature dependence of the attraction term. Very recently Yelash and Kraska16 have found that, if a van der Waals-like equation of state is formed with the Carnahan and Starling equation of state17 for hard spheres, the resulting global phase diagram shows a region of low-temperature closed-loop behavior (Types VI and VII). We propose in what follows to review some of the van der Waals-like equations of state, including a new one, and to consider the resulting global phase diagrams.

II

van der Waals-like equations of state

We consider as van der Waals-like any equation of state that has the following form : Z \ pV /RT \ f (y) [ y/T * (1) m where p is the pressure, V is the molar volume, R is the molar m gas constant, and T is the thermodynamic temperature. The variable y \ b/(4V ) is a dimensionless density, while m T * \ RT b/(4a) is a dimensionless temperature (not to be confused with T /T where T is the critical temperature) ; the c c parameters a and b are the usual van der Waals constants. The function f ( y) is just the hard-sphere compression factor. From eqn. (1) we can obtain the molar Helmholtz free energy A : m A (T , y)/RT \ m

f (y) dy/y [ y/T *

(2)

A simplifying feature of these equations is that contributions from energy and entropy are completely separate. The terms on the right are either temperature-independent (entropy) or inversely proportional to T (energy). The hard-sphere compression factor Z can be expanded in a virial series, in which the rst eight coefficients have been determined, the last several by van Rensburg :18 f (y) \ 1 ] 4y ] 10y2 ] 18.36477y3 ] 28.2245y4 ] 39.74y5 ] 53.54y6 ] (3)

We now consider four of the many possible closed-form hardsphere functions f (y) : van der W aals f (y) \ 1 \ 1 ] 4y ] 16y2 ] 64y3 ] 256y4 1 [ 4y ] 1024y5 ] 4096y6 ] Scott f (y) \ 1 ] 2y \ 1 ] 4y ] 8y2 ] 16y3 ] 32y4 1 [ 2y ] 64y5 ] 128y6 ] Carnahan and Starling f (y) \ 1 ] y ] y2 [ y3 \ 1 ] 4y ] 10y2 ] 18y3 ] 28y4 (1 [ y)3 ] 40y5 ] 54y6 ] Sanchezvan Rensburg f (y) \ Z 4,3 1 ] 1.024385y ] 1.104537y2 [ 0.4611472y3 [ 0.743038y4 \ 1 [ 2.975615y ] 3.007000y2 [ 1.097758y3 (7) The second ( Scott ) was rst proposed in a review chapter19 as an improvement on the original van der Waals 4226 Phys. Chem. Chem. Phys., 1999, 1, 4225 4231 (6) (5) (4)

form ; it has no theoretical signicance, but has been used occasionally in dealing with mixtures and with global phase diagrams. The third ( C&S ) is a weighted average17 of the Percus Yevick pressure equation and the scaled particle equation (equivalent in three dimensions to the PercusYevick compressibility equation) and gives a remarkably close t of integers to the correct hard-sphere virial coefficients. The last ( SvR ) is a Pade approximant determined by Sanchez20 to yield on expansion the rst eight virial coefficients exactly. Sanchez has shown that the Z approximant 4,3 is fairly robust (distinctly better than his Z ). 3,4 Of these, the original van der Waals form is clearly the poorest representation of the real hard-sphere uid. Each of these expressions diverges when the denominator becomes zero. The rst two diverge at y \ 1/4 and y \ 1/2 respectively, well before the density of the close-packed crystal at y \ p/(3 \ J2) \ 0.740 480. Conversely, the Carnahan and Starling denominator has three degenerate roots at y \ 1, well beyond the close-packed density. The Sanchezvan Rensburg denominator looks very similar to that of Carnahan and Starling, but it has only one real root at y \ 0.739 514, surprisingly close to that of the close-packed crystal. Since it is not obvious that a simple combination of a hardsphere f (y) with the van der Waals attractive energy should be a good representation of a real uid, it is useful to see what these combination equations yield for the properties of the one-component uid. Table 1 shows the critical constants derived from eqn. (1). (For convenience and simplicity we label these van der Waals-like equations by the names of their hard-sphere contributors.) It should be noted how similar are the critical constants for the last three ; only the van der Waals values are substantially dierent. This is surely a result of the fact that, at these relatively low densities, the rst few virial coefficients are not very dierent. More serious is the failure of any of these van der Waalslike equations to yield a satisfactory value of the compression factor at the critical point of a simple uid (e.g., argon). Indeed it can be shown that no van der Waals-like equation (e.g., a hard-sphere set of all-positive virial coefficients combined with the van der Waals energy) can yield a value of Z lower than c 1/3. Figs. 2 and 3 show for each of these equations the reduced vapor pressure p/p and the reduced liquid density o/o \ y/y c c c as a function of reduced temperature T /T . Also shown are c the values for the RedlichKwong equation (a favorite of chemical engineers, but not a van der Waals-like equation) and the experimental values for argon. (It should be noted that no classical equation such as these can reproduce the experimental densities near the critical point ; experimentally the curve is nearly cubic, while analytic theories necessarily yield quadratic curves.) What is interesting is that, while the Carnahan and Starling equation yields vapor pressures close to the experimental (the van der Waals vapor pressure is way o), the situation is reversed for the liquid density ; the Scott equation gives better agreement than either the van der Waals or Carnahan and Starling equations, whose values are equally far from those for argon (in opposite directions). The values obtained using the Carnahan and Starling equation and those using the Sanchezvan Rensburg equation are virtually indistinguishable down to reduced temperatures T /T c of about 0.6, where the computed densities are about three times the critical density (i.e., y \ 0.4). Below this point, however, the liquid densities dier increasingly, as shown in Fig. 4. The dierences produced by the two dierent denominators appear most strikingly as one approaches zero temperature. The Carnahan and Starling uid not only ends at a much higher density (corresponding to y \ 1), but the degeneracy of the roots produces a zero slope (i.e., an innite coeffi-

Table 1 Critical constants for various f (y) y \ b/(4V ) c c van der Waals Scott Carnahan and Starling Sanchezvan Rensburg Experimental (argon) 0.083 333 33 0.133 974 59 0.130 443 88 0.130 100 45 RT b/(4a) c 0.074 074 07 0.098 076 21 0.094 328 70 0.094 209 84 p b2/(16a) c 0.002 314 81 0.004 809 47 0.004 416 81 0.004 400 42 Z \ (pV /RT ) c c 0.375 00 0.366 03 0.358 96 0.359 02 0.291

cient of thermal expansion !) Whether any of this will interest a pragmatic experimentalist, who knows that uids do not reach these temperatures and densities, is doubtful.

spheres (b \ b ), slightly revised to show the Griffiths 11 22 shield region in the center. The usual coordinates for these diagrams are the following dimensionless parameters : m \ (b 11 [ b )/(b ] b ) 22 11 22 (8)

III

Global phase diagrams

Fig. 5 shows the original (1968) global phase diagram deduced from the van der Waals equation for mixtures of equal-sized

Fig. 2 The reduced vapor pressure p/p as a function of reduced temperature T /T for various equations of cstate : VdWvan der Waals ; c SScott ; C&SCarnahan and Starling ; RKRedlichKwong.

Fig. 4 The reduced density o/o of the coexisting liquid shown over c to T /T \ 0 : S-vRSanchezvan the complete temperature range Rensburg ; other identications as in Fig. 2.c

Fig. 3 The reduced density o/o of the coexisting liquid as a function of reduced temperature T /T for cvarious equations of state : identicac tions as in Fig. 2.

Fig. 5 The global phase diagram for equal-sized molecules (m \ 0) from the van der Waals equation of state. Except for the addition of the Griffiths shield region in the upper center, this is the diagram originally shown in refs. 2 and 3. [Reprinted by permission from ref. 22, Copyright, 1987, American Chemical Society.]

Phys. Chem. Chem. Phys., 1999, 1, 4225 4231

4227

a a a a 22 [ 11 22 ] 11 (9) b2 b2 b2 b2 22 11 22 11 a 2a a a a 12 ] 11 22 ] 11 (10) j \ 22 [ b2 b b b2 b2 b2 22 11 22 11 22 11 The parameter m measures the dierence in size (molar volume) between the two species ; here we assume that the unlike parameter b \ (b ] b )/2 ; otherwise one needs a 12 11 22 fourth parameter. The parameter f measures the dierence in cohesive energy (or internal pressure ) between the two species ; for b \ b , it is simply the ratio of the dierence 11 22 between the (low-temperature) energies of vaporization and their sum. The third parameter introduces a property of the mixture, a , and j is a measure of the (low-temperature) 12 energy (of mixing). The scaling of the parameters is chosen so that even extreme dierences in the a s and b s can be accommodated on one graph ; moreover, it is only the relative values of the a s and b s that determine the parameters. If a \ (a a )1@2, i.e., if the unlike attraction is the geo12 11 22 metric mean of the like attractions, then a simple relation exists between the parameters j and f : f\

A A

BNA

BNA

(1 [ j)2 ] f2 \ 1

(11)
Fig. 6 Comparison of the global phase diagrams (m \ 0) obtained from the van der Waals equation (dashed line) and from the Scott equation (solid line). The diagrams are symmetrical around f \ 0, so only the positive f region is shown. The Griffiths shield region has been omitted (see Fig. 5). (Unpublished work.)

This describes the half-circle shown in Fig. 5. To use eqns. (1) and (2) for binary mixtures, one adopts a model in which the mixture is considered as a single uid with average values a and b : m m a \ x2a ] 2x x a ] x2a (12) m 1 11 1 2 12 2 22 (13) b \ x b ] x b \ b M1 ] x [2m/(1 [ m)]N m 1 11 2 22 11 2 Eqn. (1) can be used unchanged, but to eqn. (2) must be added a term for the entropy of mixing, traditionally the entropy of ideal mixing : A (T , y, x)/RT \ m

f (y) dy/y [ y/T * ] x ln x ] x ln x 1 1 2 2 (14)

(Note that these simple forms obscure the fact that, for mixtures, T * contains a , which is a function of composition, a m feature that must be recognized when taking derivatives with respect to mole fraction. Moreover, unless b \ b , b is 11 22 m also a function of composition, and b appears in both T * m and y.) Fig. 5 shows the dierent types of phase diagrams found in the original studies. The labels II-A and III-HA refer to the presence of azeotropes and heteroazeotropes, respectively, in Type II and Type III diagrams. Not distinguished in this gure, although found by van Konynenburg and Scott, is Type III , a subdivision of Type III with a pressure minimum m in the critical line. Of special interest is the small region labelled Type IV, in which, near the critical point of component 1, there is a small three-phase region resulting from a kink in the critical line ; while no examples of the type are known for molecules of equal size, the more general global phase diagram accounts for the Type IV behavior of the system methane ] n-hexane.21 Fig. 5 is a revision22 of the original diagram3,4 and includes the shield region around f \ 0, j \ 0.45, rst found theoretically by Griffiths and coworkers6,7 but never observed experimentally. Other choices for f (y), e.g., the Scott equation, yield a global phase diagram (Fig. 6) in which the boundaries between types are shifted quantitatively, but, until recently, no new qualitative features were found with any van der Waals-like equations of state. Thus the results reported by Yelash and Kraska16 for the Carnahan and Starling f (y) were more than a little surprising ; that they found Types VI and VII in a region 4228 Phys. Chem. Chem. Phys., 1999, 1, 4225 4231

just below the f \ 0 line contradicted some of the usual ideas about low-temperature closed loops. However, given the eccentric properties of the Carnahan and Starling equation at high densities, it seemed useful to repeat the calculations with a well-behaved hard-sphere f (y), i.e., the previously outlined Sanchezvan Rensburg Pade approximant. The resulting phase boundaries for it are compared with those for the Carnahan and Starling equation in Fig. 7. One still nds Type VI and VII behavior, although in a distinctly smaller region. Thus it is clear that the unusual behavior that Yelash and Kraska found is not just a result of going to densities higher than the close-packed or of the degeneracy of the Carnahan and Starling denominator. However, it should be emphasized that, in these regions of the global phase diagram, where the densities are very high (close to the point of divergence of the equation of state), the results are extremely sensitive to the exact form of the denominator of f (y). [After this manuscript was prepared for the Walberberg Workshop, a paper by Yelash, Kraska and Deiters23 proposed two new hard-sphere equations which, although simpler

Fig. 7 A portion of the global phase diagram showing closed-loop regions obtained using the Carnahan and Starling f (y) (dashed lines) and the Sanchezvan Rensburg f (y) (solid lines). Note the exaggerated scale of the ordinate j.

and less accurate than the SvR form, have a factor (3 [ 4y) in the denominator that produces a divergence at y \ 0.75, close to that of the close-packed crystal. When combined with the van der Waals attraction term, both produce closed loops, and the better of the two (a biquadratic equation rather than a cubic) yields a global phase diagram very similar to that shown in Fig. 7.] The closed loops are found mainly at low temperatures and high densities, so their signicance for experiments on real uid mixtures is still debatable. For example, no experimental evidence for closed loops in nonpolar systems is likely to be found until one can get a uid system to a temperature below the maximum pressure of the Type VI critical line. In our calculations using the Sanchezvan Rensburg f (y), we never found the maximum above a reduced temperature T /T of c 0.25 in the Type VI region ; however, in the Type VII region it is found at higher reduced temperatures, ending above T /T \ 0.5 at the cusp between the V VII boundary and the c m VIIV boundary, what Yelash and Kraska call the double critical end cusp (DCEC). If there are any real systems that fall in this very narrow region of f and j, and if this van der Waals-like equation is appropriate, this temperature is high enough that experimentalists might nd a closed loop in a binary mixture of nonpolar components. Technical details about some of the calculations will be found in the Appendix.

Fig. 8 The van der Waals global phase diagram for systems satisfying the geometric mean eqn. (11). The solid line is the locus of tricritical points. The dashed line is the boundary between Types IV and III. The dotted lines show boundaries of azeotropic subtypes. Shown on the diagram are the f,m coordinates of various n-alkane systems. Diagram corrected from the original diagram in ref. 3. [Reproduced by permission from ref. 24.]

IV

Dierences in size (b > b ) 22 11

There have been only a few examples of van der Waals-like global phase diagrams for systems of molecules of dierent size (m [ 0), most of them coming from our laboratory.3,4,24,25 (Some studies of the eect of size dierences have been carried out for non-van der Waals-like equations of state, e.g., with the RedlichKwong equation of state by Deiters and Pegg.26) In order to extend global phase diagrams to such systems, in particular to systems of nonpolar polymers, additional assumptions or restrictions are advisable : 1. Most molecules are more linear than globular, so the assumption that b is simply the arithmetic mean of b and 12 11 b [eqn. (13)] seems reasonable. 22 2. For chain molecules, the Flory mixing equation is probably preferable to the ideal mixing equation. Ideal entropy of mixing is, of course, correct for the dilute gas mixture, while the Flory entropy is surely preferable for the dense liquid mixture. Which is preferable at densities corresponding to the critical density of a one-component uid is unknown and is likely to remain unknown unless someone can carry out a careful simulation. Most of the critical lines we have calculated are at densities greater than this, so we prefer to use the Flory equation throughout. 3. A corresponding states treatment of molecules of dierent length requires the introduction of a third parameter c (rst introduced by Prigogine27) that allows for the entropy dierences produced by a chain molecule with internal degrees of freedom. A full global phase diagram with three independent parameters f, j, and m would be three-dimensional, so one usually assumes the geometric mean for a , i.e., eqn. (10). With this 12 restriction, an early calculation3,22 produced the j, m diagram shown in Fig. 8. Although this gure was based upon ideal entropy and the van der Waals f (y), it nevertheless accounted qualitatively for the distribution of mixtures of n-alkanes between Types II, IV, and III. A later calculation (unpublished), using the Flory entropy of mixing, shifts the boundaries slightly, but leaves the general picture unchanged. Wei and Scott24 examined the question of whether large dierences in size, i.e., large values of m, could shift the Grifths shield region over to the geometric-mean line. They used mainly the ideal entropy and found that, at best, the shield

region could only occur with very large j values associated with large m values of opposite sign, describing this situation as a mixture of Teon and mercury . A few inadequate calculations were made with the Flory mixing entropy, but this research needs to be pursued with it as well as with better f (y) functions. Further research on mixtures of chain molecules (or indeed of any nonglobular molecules) should probably replace eqns. (1) and (2) with the more general ones that Pegg et al.25 used : Z \ pV /RT \ 1 ] c[ f (y) [ 1] [ y/T * m A (T , y, x)/RT \ ln y ] c m

(15)

[ f (y) dy [ 1]/y [ y/T *

] x ln(x b /b ) ] x ln(x b /b ) (16) 1 1 11 m 2 2 22 m Here c appears in a form suggested by Beret and Prausnitz.28 They evaluated the parameter to t experiment, but, at least for n-alkanes, it can be expressed simply in terms of the number of carbon atoms n : c(n \ 1) \ 1 ; c(n [ 1) \ 1 ] n/3 (17)

Fig. 9 Upper and lower critical-end-point lines in the tricritical region of mixtures of ethane with n-alkanes. (Here n is the number of 2 carbon atoms in the higher hydrocarbon or the average value in the ternary quasi-binary mixture.) The two lines converge to the tricritical point and the region between the two is where three phases can occur. The solid lines are calculated from eqn. (16). The dashed lines show a selection of experimental points for binary and quasibinary mixtures.29,30 [Reproduced by permission from ref. 25.]

Phys. Chem. Chem. Phys., 1999, 1, 4225 4231

4229

Then for the mixture c is simply the mole fraction average of m c and c 1 2 c \x c ]x c (18) m 1 1 2 2 Pegg et al.25 used this equation [with the van der Waals f (y) and no adjustable parameters] with remarkable success to interpret the behavior of mixtures of n-alkanes in the vicinity of tricritical points. Fig. 9 shows the experimental29,30 and theoretical upper and lower critical-end-point lines for binary and quasi-binary mixtures of ethane with mixtures of higher hydrocarbons from C to C . Eqn. (16) predicts the tri16 20 critical point to be at the higher hydrocarbon n \ 16.48 while 2 the experimental value occurs at 17.61. With the Scott equation, one gets n \ 17.3 (unpublished). What one would get 2 with other (better ?) choices of f (y) is unknown, and such games should probably not be encouraged. How successfully this approach can be generalized to treat dierences of size in other kinds of chain molecules remains to be seen.

where F is the numerator of the selected f (y), and G@ is the 0 0 rst derivative (with respect to y) of the denominator, both evaluated at y . Eqn. (19) is linear in w2 (quadratic in x ) and 0 0 0 is easily solved ; the resultant w (or x ) is the same for all 0 0 values of f. This w can then be used to obtain p for any 0 0 particular value of f. y2(f [ w ) 0 p /p \ 0 (zero j only) (20) 0 c p*w (1 [ f) c 0 where p is the critical pressure of component 1 and p* is c c p b2/(16a) taken from Table 1. Evidently p changes sign at c 0 f \ w \ (1 [ 2x ), and the character of the phase diagram 0 0 will change there. For the Carnahan and Starling f (y) with its degenerate denominator, G@ is zero, so w \ 0, x \ 1/2, and p turns out 0 0 0 0 to be positive innity for all values of f. For all the other f (y), with a nondegenerate denominator, p can be positive or 0 negative, but is normally nite. This termination is important in recognizing dierent types of phase diagrams, because for any positive value of j, the critical line pulls away from the T \ 0 axis and p goes to 0 positive innity at a T corresponding to the critical solution 0 temperature of a close-packed simple mixture at x \ 1 (i.e., 0 2 one gets only Type II, III, or IV phase diagrams). Conversely, for any negative value of j, the critical line ends at T \ 0, x \ 0 and a negative pressure p of nite magnitude that is 0 0 independent of the values of f and j (as long as j is negative) and a function only of the choice of equation of state : p /p \ y2/p* (negative j only) (21) 0 c 0 c [It may be noted that eqn. (21) is the limiting value of eqn. (20) for f \ 0 and is necessarily at a more negative pressure than that given by eqn. (20) for any nonzero f.] With the van der Waals or Scott f (y), the IVIII and m III III boundaries end asymptotically on the j \ 0 line at m T \ 0, p \ 0. From eqns. (19) and (20) one nds that the 0 IVIII boundary ends at f \ w \ 1 [ 2x \ 1/J2 \ 0.7071 m 0 0 with the van der Waals f (y) and at f \ w \ 1 [ 2x \ 0 0 1/J3 \ 0.5774 with the Scott f (y). The asymptotic termination of these lines makes Type V impossible and, in addition, m precludes Types VI and VII for these equations of state. With the Sanchezvan Rensburg f (y) and j \ 0, p changes 0 sign at x \ 0.333 533, f \ 0.332 934 and the phase diagram 0 changes from Type I to Type II ; moreover, in contrast to the van der Waals or Scott diagrams, this is the asymptotic end of a line that extends into the region of negative j values, marking the boundary between Types I and VI. It is interesting to note that this line separating Types I and II, III and V, etc., which for all van der Waals-like equations of state coincides with the j \ 0 line, can be dierent for other equations of state. Thus Deiters and Pegg,26 who call this the zero-Kelvin endpoint curve , found that for the Redlich Kwong equation, it did not coincide with the j \ 0 line, but was found below that line (except at f \ 0).

Some general comments

Anyone working in the eld of global phase diagrams is surely aware of two serious problems : 1. No simple equation of state can ever be expected to yield a quantitative t to experimental results. We are lucky when we get a reasonably semi-quantitative understanding. 2. The known experimental results for binary mixtures of nonpolar nonelectrolytes cover only a small area of the global phase diagram (at or above the geometric mean for the unlike attraction). Unfortunately this probably means that some of the interesting phase diagrams deduced theoretically will never be found experimentally, at least not for nonpolar systems. Indeed it is likely that most, if not all, of the nonpolar systems classied as I or V are really II or IV, with the lowtemperature upper critical solution temperature (UCST) hidden below the solid line. This should be the case whenever the low-temperature enthalpy of mixing is known to be positive. Dickinson et al.31 used the van der Waals and Scott equations of state to show that the atness of the freezing curve of the system methane ] n-hexane can easily be interpreted in terms of metastable liquidliquid phase separation 5 to 15 K below.

Acknowledgements
I wish to thank all of my collaborators over the years who have contributed to our work on global phase diagrams, in particular (but not exclusively) Peter van Konynenburg, Ian L. Pegg, and my colleague Charles M. Knobler.

Appendix
A general picture of the character of a global phase diagram can be obtained by examining phase diagrams along the line j \ 0. This line necessarily separates Types II, IV, III , and m III (occurring only with positive j values) from Types I and Vand maybe VI, VII, and V (occurring only with negam tive j values). The tricritical line necessarily crosses the j \ 0 line, but others may not. An important feature of the j \ 0 line is the termination of the critical line at the maximum y (i.e., at the divergence of 0 the denominator). This occurs necessarily at T \ 0, but the pressure p at this limit depends upon the parameter f and the 0 choice of the equation of state. In particular one can derive an analytic expression for the limiting mole fraction x for any 0 f (y) : If one writes w \ 1 [ 2x , then 0 0 F w2 ] y G@ (1 [ w2) \ 0 (19) 0 0 0 0 0 4230 Phys. Chem. Chem. Phys., 1999, 1, 4225 4231

References
1 2 3 4 5 6 7 8 J. J. van Laar, Proc. Sect. Sci., K. Ned. Acad. W et., 1904 05, 7, 517, 636, 646 ; 1905 06, 8, 33 ; 1906 07, 9, 226. P. H. van Konynenburg, PhD Dissertation, UCLA, 1968. R. L. Scott and P. H. van Konynenburg, Discuss. Faraday Soc., 1970, 49, 87. P. H. van Konynenburg and R. L. Scott, Philos. T rans. R. Soc. L ondon, 1990, 298, 495 (includes a major fraction of ref. 2). P. Clancy, K. E. Gubbins and C. G. Gray, Faraday Discuss. Chem. Soc., 1978, 66, 116. D. Furman, S. Dattagupta and R. B. Griffiths, Phys. Rev. B : Solid State, 1977, 15, 441. D. Furman and R. B. Griffiths, Phys. Rev. B : Solid State, 1978, 17, 1139. J. A. Barker and W. Fock, Discuss. Faraday Soc., 1953, 15, 188.

9 10 11 12 13 14 15 16 17 18 19 20 21 22

Y. Hueda, M. E. Costas and R. L. Scott, J. Phys. Chem. B, 1997, 101, 8676. G. R. Anderson and J. C. Wheeler, J. Chem. Phys., 1978, 69, 2082. J. S. Walker and C. A. Vause, J. Chem. Phys., 1983, 79, 2660. L. R. Corrales and J. C. Wheeler, J. Chem. Phys., 1989, 91, 7097. G. Jackson, Mol. Phys., 1991, 72, 1365. V. Talanquer, J. Chem. Phys., 1992, 96, 5408. L. Z. Boshkov, Dokl. Akad. Nauk SSSR, 1987, 294, 901. L. V. Yelash and T. Kraska, Ber. Bunsen-Ges. Phys. Chem., 1998, 102, 213. N. F. Carnahan and K. E. Starling, J. Chem. Phys., 1969, 51, 635. J. van Rensburg, J. Phys. A, 1993, 26, 4805. R. L. Scott, in Physical Chemistry, An Advanced T reatise, ed. D. Henderson, Academic Press, New York, 1971, vol. 8A, ch. 1, p. 1. I. Sanchez, J. Chem. Phys., 1994, 101, 7003. A. J. Davenport and J. S. Rowlinson, T rans. Faraday Soc., 1963, 59, 78. R. L. Scott, Acc. Chem. Res., 1987, 20, 97.

23 24 25 26 27 28 29 30 31

L. V. Yelash, T. Kraska and U. K. Deiters, J. Chem. Phys., 1999, 110, 3079. I. C. Wei and R. L. Scott, J. Stat. Phys., 1988, 52, 1315. I. L. Pegg, C. M. Knobler and R. L. Scott, J. Chem. Phys., 1990, 92, 5442. U. K. Deiters and I. L. Pegg, J. Chem. Phys., 1989, 90, 6632. I. Prigogine, T he Molecular T heory of Solutions, North Holland, Amsterdam, 1957, ch. XVI. S. Beret and J. M. Prausnitz, AIChE J., 1975, 21, 1123. M. C. Goh, J. Specovius, R. L. Scott and C. M. Knobler, J. Chem. Phys., 1987, 86, 4120. M. C. Goh, R. L. Scott and C. M. Knobler, J. Chem. Phys., 1988, 89, 2281. E. Dickinson, C. M. Knobler and R. L. Scott, J. Chem. Soc., Faraday T rans. 1, 1973, 69, 2179.

Paper 9/03400E

Phys. Chem. Chem. Phys., 1999, 1, 4225 4231

4231