The evolution of multicomponent systems at high pressures

Part II.¤ The Alder–Wainwright, high-density, gas–solid phase transition of the hard-sphere Ñuid
J. F. Kenney Joint Institute of the Physics of the Earth, Russian Academy of Sciences ; Gas Resources Corporation, 15333 JFK Boulevard, suite 160, Houston, T X 77032, USA. E-mail : JFK=alum.MIT .edu Received 3rd March 1999, Accepted 24th May 1999

The thermodynamic stability of the hard-sphere gas has been examined, using the formalism of scaled particle theory (SPT), and by applying explicitly the conditions of stability required by both the second and third laws of thermodynamics. The temperature and volume limits to the validity of SPT have also been examined. It is demonstrated that SPT predicts absolute limits to the stability of the Ñuid phase of the hard-sphere system, at all temperatures within its range of validity. Because SPT describes Ñuids equally well as dilute gases or dense liquids, the limits set upon the system stability by SPT must represent limits for the existence of the Ñuid phase and transition to the solid. The reduced density at the stability limits determined by SPT is shown to agree exactly with those of that estimated for the AlderÈWainwright, high-density gasÈsolid phase transition in a hard-sphere system, at a speciÐc temperature, and closely over a range of more than 1000 K. The temperature dependence of the gasÈsolid phase stability limits has been examined over the range 0.01 KÈ10 000 K. It is further shown that SPT describes correctly the variation of the entropy of a hard-core Ñuid at low temperatures, requiring its entropy to vanish as T ] 0 by undergoing a gasÈsolid phase transition at Ðnite temperature and all pressures.

1. Introduction
The high-pressure, supercritical, gasÈliquid phase transition was Ðrst reported speciÐcally for the real Ñuid normal octane, n-C H , in a previous article which will be referred to hence8 18 forth as I.1 That particular molecular species had been chosen for analysis, perhaps fortuitously, for reasons of an availability of data. In I, it was asserted that the high-pressure, Ðrst-order, gasÈliquid phase transition is a geometric property of Ñuids which must exist independently of the details of any attractive component of the intermolecular potential. This argument was supported in I by showing that the gasÈliquid phase transition would occur in a system composed of imaginary ““ isomers ÏÏ of n-octane which, while maintaining all other properties of that Ñuid, might possess a characteristic temperature, as deÐned by the formalism of the simpliÐed perturbed hard chain theory (SPHCT), of one millikelvin, T * \ 0.001 K. Such a characteristic temperature would describe a Ñuid e†ectively without any long-range, attractive van der Waals component to its intermolecular potential, and which would be a gas at all temperatures and low pressures. That ““ isomer ÏÏ of n-octane was shown to undergo the gasÈliquid phase transition at almost the same pressure as the real Ñuid. In order to demonstrate rigorously that the high-pressure, supercritical, Ðrst-order, gasÈliquid phase transition is a geometric property of Ñuids, in this article and a following one, the stability of hard-core Ñuids is explicitly investigated. Because the formalism of scaled particle theory (SPT) was applied extensively in I as the reference system for the joint partition function, the competence of that formalism to describe the stability of a hard-core system has been investigated thoroughly. In both this and the article to follow, the ability of SPT to predict phase transitions in a hard-core
¤ Part I : ref. 1.

system is examined by applying all the conditions of stability required by the three laws of thermodynamics. The purpose of this article is to demonstrate the ability of SPT to predict the AlderÈWainwright, high-density, gasÈsolid phase transition in a hard-sphere gas. There has been common misunderstanding that SPT only reproduces the pressureÈdensity relationship of a hard-sphere system from low densities until, and through, the AlderÈWainwright phase transition, but fails to predict or reproduce that transition. Such misunderstanding is corrected in the following sections where is explicitly described the prediction of the AlderÈ Wainwright transition by SPT when constrained by general stability conditions. The density at which the ÑuidÈsolid phase transition occurs in a hard-sphere system is shown to depend upon its temperature. Furthermore, the temperaturedependence of the ÑuidÈsolid phase transition predicted by SPT requires that the Ñuid undergoes such transition at all pressures as the temperature decreases to very low temperatures, precisely in keeping with general thermodynamic considerations. Additionally, the latent change of entropy of the ÑuidÈsolid transition predicted by SPT for hard sphere systems at low temperatures is shown to be consistent with experimental values of the melting entropy for volatile spherical molecules. Because SPT uses a factored partition function which is obtained from application of the BakerÈCampbellÈHausdor† expansion, it is a high-temperature formalism. Therefore, the low-temperature limits of the validity of SPT have been examined explicitly and are described below. The stabilities of the one-dimensional, ““ hard-rod ÏÏ, and two-dimensional, ““ hard-disk ÏÏ systems have also been investigated and are described in two Appendices. When the general stability conditions, as used with the three-dimensional, hardsphere Ñuid, are applied to the one-dimensional, ““ hard-rod ÏÏ system, scaled particle theory predicts no instability against a Phys. Chem. Chem. Phys., 1999, 1, 3277È3285 3277

ÑuidÈsolid phase transition of the AlderÈWainwright type. When those same stability conditions are applied to the twodimensional, ““ hard-disk ÏÏ system, it is shown to manifest a ÑuidÈsolid, ““ freezing ÏÏ transition of the AlderÈWainwright type at reduced densities somewhat greater than 0.8.

2. Background
The hard-core Ñuid has been extensively studied, especially since the discovery of the high-density, gasÈsolid phase transition by Alder and Wainwright in the hard-sphere system.2 A phase transition in a hard-sphere system had been suggested earlier in theoretical work by Kirkwood, who argued that an ordered conÐguration should appear in the system at a reduced density of approximately 0.67.3 Later, Born and Green developed the same conclusions.4 However, the investigations by Kirkwood (and also those by Born and Green) involved extrapolations of an assumed two-particle correlation function for dense Ñuids from its form in the thermodynamic regime of dilute gases. The calculations of Kirkwood (and Born) could not predict accurately even the fourth virial coefficient for simple Ñuids ; and such failure cast severe doubts about the validity of their predictions of a phase transition. The explicit solution by Alder and Wainwright of the equations of motion of assemblages of hard spheres in 1957, and their demonstration of the phase transition, removed all uncertainties about its existence.2 During the same period while Alder and Wainwright were conducting their computer experiments, Reiss et al.5h7 were developing the formalism of SPT for the analysis of hard-body systems which applied a radically di†erent perspective to statistical mechanics and employed a mathematical expression of the arcane branch of mathematics designated statistical geometry.5 The description of the thermodynamic properties of the hard-sphere gas by SPT represents one of the few exactly-solvable problems in quantum statistical mechanics. A particularly important aspect of SPT, which obtains from its property as an exact, geometric description, is that it generates an absolute solution. There are no adjustable parameters or additive constants to the solution to the problem of the hardcore gas given by SPT. This property of SPT is used directly below to predict, unequivocally, the AlderÈWainwright highdensity, gasÈsolid phase transition, both at high temperatures and as the temperature of the system approaches absolute zero.

with any single state of a given system. As such, these conditions are local, a property evidenced by their expression as di†erential inequalities. The second law constitutes a global constraint upon the stability of a system, as evidenced in its statement by De DonderÏs inequality : A(p, T , Mn N) \ [ ; lÕ kÕ [ 0 # Mi, o, / ½ (R 0 P)N. (2) j i, p i, p i, p, Õ In eqn. (2), the summations are taken over all components, i, phases, /, and chemical reactions, o, which involve possible initial (reagent, R) and Ðnal (product, P) states of the system ; the mathematical entity, A(p, T , Mn N), is the thermodynamic j affinity, so deÐned by De Donder.8 The non-local condition set forth by eqn. (2) places the constraint upon any possibility that a system in one state might evolve spontaneously into another such that its internal entropy must increase. While De DonderÏs inequality sets a necessary condition for stability, it sets the sufficient condition for phase separation only when the states involved refer unequivocally to separated phases. The third law also sets additionally a stringent condition for the physical existence of any state : The total entropy of the system must be positive.9h12 S P 0. (3)

The inequality (3) is, in fact, only a weaker corollary of the third law, and does not address the behavior of the entropy as temperature approaches absolute zero. Strictly, inequality (3) may be considered a consequence of BoltzmannÏs classical deÐnition, S \ k ln(W ), in which W represents the number of B conÐgurations which the system can assume, consistent with its thermodynamic constraints. Because W is a natural number, P1, the inequality (3) results. The third law states that entropy is an absolute entity, with no arbitrary additive constants, and always greater than zero. Unlike De DonderÏs inequality, a failure of NernstÏs theorem, inequality (3), for any phase of a system constitutes both a necessary and sufficient condition for a phase transition.

4. Scaled particle theory (SPT)
SPT was Ðrst developed by Reiss et al.,5h7,13h18 from arguments of statistical geometry. Carnahan and Starling19 later derived the SPT equation of state, augmented by the additional, negative term involving the cubic power of the reduced density, [g3/(1 [ g)3, by summing explicitly the series developed from the computed virial coefficients of the hard-sphere gas. SPT (with the CarnahanÈStarling term) generates the explicit canonical partition function : )(V , T ; N)SPT \ )IG)hvc \ 1 V N N ! j3 ] exp [

3. The thermodynamic conditions for stability
Each of the three laws of thermodynamics states di†erent conditions for the stability of a system. All must be satisÐed for the stability of a system. The Ðrst and second laws together develop the conditions for stability of a single state of a single-component, singlephase system which require that its isothermal compressibility be greater than zero, that its isometric speciÐc heat be greater than zero, that its isobaric speciÐc heat be greater than its isometric speciÐc heat, that the Ðrst derivative of its chemical potential with respect to pressure be positive and the second derivative be negative : n dp dT 1 dV [ 0; [ 0; [ t dT dV V dp t T V p t dp dT t [0 dT dV t V p o (1) dS dS \ C [ 0; T \C [C t T V p Vt dT dT V p t dk d2k t [ 0; ###BOT_TEXT### dp dp2 t T,n T,n p The conditions for stability set out in eqn. (1) are concerned

AB C A

g(4 [ 3g) (1 [ g)2

BD

N

. (4)

A B A BA B A BA B A B A B A B A B

In eqn. (1), the ““ thermal de Broglie wavelength,ÏÏ j \ h/J(2pmk T ), where h is PlanckÏs constant and k BoltzB B mannÏs constant ; T is the absolute temperature and m the molecular mass ; g is the ““ reduced-density ÏÏ parameter, g \ V */V \ n(p/6)p3/V \ o(p/6)p3, in which p is the diameter of the hard-spheres, n the molar abundance. The factored partition function, eqn. (4), constitutes implicit application of the BakerÈCampbellÈHausdor† lemma,20h24 Q \ trSexp([bH(p, q))T \ trSexp([b(T (p) ] X(q)))T \ trSexp([bT (p))TtrSexp([bX(q))T ] tr

TA

1[

b2 [T , X] ] É É É 2

BU

D trSexp([bT (p))TtrSexp([bX(q))T.

(5)

3278

Phys. Chem. Chem. Phys., 1999, 1, 3277È3285

(In eqn. (5), the Hamiltonian, H, has been expressed conventionally in terms of its respective kinetic and potential energy operators, T (MpN) and )(MqN) of its system momenta MpN, and spatial MqN coordinates ; and b \ 1/k T .) Neglect of the third B factor on the right side of eqn. (5) constitutes a hightemperature approximation. Thus, from the form of its partition function, SPT is a formalism valid only at high temperatures ; and its low temperature limits are investigated explicitly in Section 6. The SPT canonical partition function generates the Helmholtz free energy : g(4 [ 3g) V ]1] \ FIG ] Fhc ; FSPT \ Nk T [ ln B (1 [ g)2 Nj3 (6) of which the isothermal volume derivative gives the system pressure : 1 1 2g(2 [ g) \ pIG ] phc. ] p \ Nk T B V V (1 [ g)3

C A B C

D

D

(7)
Fig. 1 Entropy of a hard-sphere gas at 300 K as a function of the reduced-density parameter, g.

5. The absolute entropy of a hard-sphere system, and the Alder–Wainwright, high-density, gas–solid phase transition
The hard-core contribution to the partition function as described by SPT is purely entropic. This property results intrinsically from the arguments from geometric statistics which underlie SPT and are indi†erent to temperature. Therefore, the only contribution to the internal energy of a hardsphere Ñuid is obtained from the ideal gas factor in its partition function, such that USPT \ 3/2Nk T . When the B general thermodynamic equation for the Helmholtz free energy, F \ U [ T S, is applied, the entropy for a hard-core gas is given explicitly by the CarnahanÈStarling extension of SPT as : g(4 [ 3g) 5 V [ \ SIG ] Shc. (8) ] ln SSPT \ Nk B 2 (1 [ g)2 nN j3 A If the last term in eqn. (8) is neglected, it becomes SakurÈ Tetrode equation, to which the SPT equation for entropy reduces in circumstances of dilute densities. The two leading terms in the entropy, the ideal gas contribution, SIG, are always positive, and dominate that function in the region of low and moderate densities. The third term in the entropy, the hard-core contribution, Shc, is always negative and, most importantly, possesses a second-order singularity in the reduced density. Therefore, at sufficiently high densities, the entropy of the hard-core Ñuid must change sign and become negative. Such behavior by the system would constitute a glaring violation of the third law of thermodynamics. Therefore, the density at which the entropy of the hard-core Ñuid changes sign represents an absolute limit for the Ñuid, which cannot exist at that, or higher, densities. This limit determines the AlderÈWainwright, high-density, gasÈsolid phase transition. Scaled particle theory describes equally the dilute gas and the dense liquid states. Therefore, the limit determined by SPT for the existence of the system so described must set the boundary for a solid phase. This property of the entropy of the hardcore Ñuid is shown graphically in Fig. 1 for a hard-sphere gas, at the temperature 300 K, as a function of its reduced density, g. The calculated values are represented graphically in Fig. 1 where are shown also the respective contributions of the ideal gas and hard-core entropies. Also in Fig. 1 is drawn the vertical line which represents the reduced density for the gasÈsolid phase transition as estimated by Alder and Wainwright using the Ðxed value for the reduced-density parameter,

A

A

B

B

gAW \ 0.67114.25 The position of the gasÈsolid phase transition estimated by the Ðxed value of the reduced-density parameter, gAW, is seen to lie very closely, within a few percent, to that determined by the vanishing of the SPT entropy. In fact and as may be observed in Fig. 1, the estimate of the AlderÈWainwright transition using the Ðxed value, gAW would allow the Ñuid, at least at the temperature 300 K, to possess a negative entropy within the range of pressures between that determined by the sign change of the entropy and gAW. The inadequacies of the local stability conditions set out in eqn. (1) to predict the AlderÈWainwright transition may be noted by calculation of the isothermal compressibility, [V (dp/dV ) , the isometric thermal variance of the pressure, T (dp/dT ) , the isobaric thermal variance of the volume, V (dV /dT ) , and the isobaric speciÐc heat, C , at the density at p p which the system entropy changes sign. All these thermodynamic variables satisfy the local conditions of stability in the Ñuid phase at the AlderÈWainwright transition. A less widely used determinant of a phase transition is the isobar inÑection, (dCp/dp) \ 0, the importance of which for T investigating the phase stability of Ñuids has been described by Deiters and de Reuck26 and by Bernal,27 who pointed out the experimental demonstration of its signiÐcance by Jones and Walker.28 When the fourth Clausius equation, (dCp/dp) \ [T (d2V /dT 2) , is applied, the isobar inÑection T p can be determined from the second isobaric temperature derivative of the volume. SPT gives for the second isobaric temperature derivative of the volume :

A B
]

V d2V SPT \[ T2 dT 2 p (1 [ g)g2(1 ] g ] g2 [ g3)(3 ] 9g [ 6g2 [ 4g3 ] 5g4 [ g5) . (1 ] 4g ] 4g2 [ 4g3 ] g4)3 (9)

The isobar inÑection is determined by the roots of the polynomial of eighth degree in the numerator of eqn. (9). Aside from its obvious roots at zero and unity, eqn. (9) admits no real roots for values of the reduced-density parameter, g, Phys. Chem. Chem. Phys., 1999, 1, 3277È3285 3279

Fig. 2 The volume and temperature limits of the validity of scaled particle theory.

greater than zero or less than one ; of its four real roots, two are negative and the other two greater than one. The values of the second isobaric temperature derivative of the volume, calculated at the density at which the entropy vanishes for the hard-sphere system, do not vanish.

6. The temperature and volume limitations of scaled particle theory
The pressures and reduced volumes of the gasÈsolid phase transition for a hard-sphere Ñuid have been calculated at temperatures from 0.01 to 10 000 K. The hard-sphere Ñuid considered is characterized by a hard-core covolume of 20 cm3 mol~1 and a mass of 35 u ; these parameters are typical of light gases such as methane or nitrogen. For comparison, the principal results of the investigation have been carried out also for a gas with covolume and mass Ðve times greater. SPT uses a factored partition function, eqn. (4), in which the

contribution of the kinetic energy of the Ñuid has been obtained by integrating out the MaxwellÈBoltzmann distribution of kinetic energies of an ideal gas. Therefore, there exist restrictions, of both volume and temperature, upon the validity of the SPT formalism. SpeciÐcally, the SPT partition function is restricted such that values of the system volume used satisfy the inequality : V A Nj3. In short, the SPT formalism is valid only for such volumes for which the average interparticle separation must be much greater than the particle dimensions. In Fig. 2, the calculated values of the volume of the hardsphere gas for which the entropy vanishes, V (T ; S \ 0), are shown on a double-logarithmic scale through the temperature range 0.01È10 000 K, together with the values of its ““ kinetic ÏÏ volume, V (T ) \ Nj3. The temperature at which V \ V is kin kin approximately 0.16 K, and represents an absolute limit for the validity of the SPT formalism. At such low temperatures, not only has the system volume become very small, in order to provide a zero for the entropy equation, eqn. (8), but the thermal de Broglie wavelength has become relatively much larger because of the square root of the temperature in its denominator. The value of temperature at which V \ V is kin marked in Fig. 2 by a broad dashed stripe, approximately to divide visually the regions in which the SPT formalism is valid and invalid. Visual inspection in Fig. 2 of the traces of V (S \ 0) and V persuades that a reasonable limit of conkin Ðdence for SPT could be placed at about the temperatures 5È10 K, for at 10 K, V (S \ 0) B 100V . kin The pressures at which the entropy of the Ñuid vanishes, p(S \ 0), are shown in Fig. 3 on a double-logarithmic scale for the range of temperatures 0.1È10 000 K. On Fig. 3, are shown also the values of the pressure for which V \ V , which are kin marked with a broad dashed stripe, again to indicate the pressure limits of validity of the SPT formalism. The two regions of validity and invalidity, respectively, are indicated in Fig. 3. Shown also on Fig. 3 are the pressures estimated for the phase transition by the Ðxed value of the reduced volume of the AlderÈWainwright transition given by the value of the reduced-density parameter gAW \ 0.67114. As may be noted in Fig. 3, the predictions of the gasÈsolid transition pressures from these two calculations give very similar values over a range of temperatures extending three orders of magnitude. At high temperatures, the estimate obtained from the Ðxed value of gAW underestimates the transition pressure. At low temperatures, the estimates of the transition pressure are plainly too small, and certainly incorrect ; for at temperatures less than approximately 500 K, a phase transition which might occur at the pressures estimated by gAW would allow the Ñuid to exist with negative entropy. This error is shown for the lower temperatures and shows clearly the errors developed by the estimate of a Ðxed, temperature-independent, value for the reduced-density parameter at the gasÈliquid phase transition. For temperatures less than 100 K, the assumption that the gasÈsolid phase transition should occur at gAW \ 0.67114 would allow for the Ñuid system to exist in a state of negative entropy over ranges of almost an order of magnitude in pressure.

7. The high-pressure, gas–solid Alder–Wainwright transition and its temperature dependence
In Fig. 4, are shown the values of the reduced-density parameter at the transition point where the entropy vanishes, g(S \ 0), as a function of temperature over the range 10È 10 000 K. Shown also on Fig. 4 is the trace of the Ðxed value of the reduced-density parameter estimated for the transition point by Alder and Wainwright, gAW \ 0.67114. Although the ordinate has been drawn with a very expanded, linear scale in Fig. 4, the value of the reduced-density parameter at the transition point determined by the vanishing of the entropy di†ers

Fig. 3 The pressure and temperature dependence limits of the validity of scaled particle theory, and the pressures of the AlderÈ Wainwright transition, as functions of temperature, p(S \ 0 ; T ).

3280

Phys. Chem. Chem. Phys., 1999, 1, 3277È3285

Fig. 4 Reduced-density parameter, g, at the transition point determined by S \ 0, as a function of temperature.

a gasÈsolid phase transition as its temperature approaches absolute zero. In Fig. 5, the pressures of the gasÈsolid phase transition may be observed as a function of temperature over the low range 0.1È15 K on a linear scale. While taking cognizance of the temperature limitations of the SPT formalism, speciÐcally that the details of the description of the system are unreliable at temperatures less than approximately 1È5 K, the values of the pressure at which the entropy of the Ñuid becomes zero are nonetheless seen in Fig. 5 to decrease rapidly with temperature and to approach zero as T ] 0. Between the temperatures 5È15 K, the values of the pressure at which the Ñuid system becomes absolutely unstable against transition to the solid phase decrease by approximately half an order of magnitude, while the temperature decreases by only 10 degrees. Thus, p(S \ 0) ] 0 as T ] 0, and faster than does the temperature itself. The shape of the curve which represents the pressure of the AlderÈWainwright transition at low temperature argues strongly that, as the temperature of the hard-sphere gas approaches absolute zero, the Ñuid becomes increasingly unstable at any pressure. The pressure determined by gAW is also drawn on Fig. 5, and is easily seen to represent unphysical behavior at low temperatures.

by that estimated at 0.67114 by only a few percent : at 100 K, gS/0 \ 0.63206, 5.8% lower than gAW ; at 1000 K, gS/0 \ 0.68227, 1.6% higher.

9. The entropy of the gas–solid phase transformation estimated by SPT
It has been shown that a hard-sphere system in the Ñuid state, as described by SPT, is characterized by limits determined by the stability condition imposed by the third law of thermodynamics. That such limit determines a ÑuidÈsolid separation is easily recognized. The limit set by the change of sign of the system entropy occurs at densities always greater by approximately 30% than that of the close-packed conÐguration. Certainly the hard-sphere system exists at such densities, as well as at greater until that of the close-packed limit. Because SPT describes the Ñuid system equally in the dilute gas or dense liquid circumstances, the stability limit must be that for the separation of the system in its ÑuidÈsolid phases. Because SPT predicts a solidÈÑuid phase transition for a hard-sphere gas, it is of interest to calculate the latent heat of transformation at its sublimation or melting point. Because the latent heat of transition represents the enthalpic di†erence between two phases, the entropy of transformation cannot be calculated presently for the hard-sphere system because an exact thermodynamic description is not yet available for the hard-sphere solid. Here follows an approximate calculation of the limiting value of the SPT entropy of melting at low temperatures. As a hard-sphere system approaches absolute zero, its latent heat of transition for the solidÈÑuid phase transition, *L sf, approaches a constant : *L (T )sf \ *Hsf \ (U(T )fluidU(T )solid) ] psf(V (T )fluid [ V (T )solid), (10)

8. The temperature dependence predicted by SPT for the Alder–Wainwright transition at low temperatures
A consequence of the third law requires that the speciÐc heat, however measured, must vanish at T ] 0 at least as fast as T (1`d) for d [ 0. However, as has been noted in Section 4, the speciÐc heat for a hard-core Ñuid described by SPT remains at all temperatures simply that of an ideal gas : CSPT \ 3/2Nk . V B By such property, SPT describes a Ñuid phase which fails to manifest correct physical behavior at very low temperatures ; a fact which suggests that SPT might indicate the presence of a low-temperature gasÈsolid phase transition. SPT does indeed strongly indicate that the Ñuid undergoes

in which psf represents the pressure of the solidÈÑuid phase transition. As shown in Fig. 5 of the previous section, the pressure of the solidÈÑuid phase transition, psf, goes rapidly to zero at low temperatures. Therefore, although there exists a volume di†erence between the two phases, the second term in parentheses on the right side of eqn. (10) decreases rapidly as the system temperature approaches absolute zero. The Ðrst term in parentheses on the right side of eqn. (10) does not vanish as the system temperature approaches absolute zero. The internal energy of the gas phase vanishes linearly with temperature,
Fig. 5 Low temperature behavior of the AlderÈWainwright transition predicted by S \ 0.

U(T )fluid \ 3 Nk T , (11) 2 B while that of the solid may be assumed to have the general Phys. Chem. Chem. Phys., 1999, 1, 3277È3285 3281

form U(T )solid D 3 Nk T 2 B

C A B

2p4 T 3 ] É É É O(exp([T D/T )) , 5 TD

D

(12)

in which T D \ +wD/k represents the e†ective Debye temB perature ; and wD \ kDc where kD represents the geometric, s maximal, Debye wave number and c the velocity of sound. s Thus the limit of the entropic change in passing from the solid to gas phase is : lim (*Ssf)SPT \

C

ÈÈÈÕ 3 Nk 2 B T?0 B 2.98 cal mol~1 K~1.

*L (T )sf T sf

D

(13)

This interesting property has not been previously reported in the extensive literature dealing with SPT. The entropy of fusion at the melting point for almost all metals lies in the range 1.7È2.6 cal mol~1 K~1.29 That the entropy of melting for real metals should be less than that given by the SPT limit, eqn. (13), should be expected, for the limit to which eqn. (13) refers is that for which the temperature vanishes. Real metals fuse at high temperatures, at which the calculated entropy would have subtracted from eqn. (12) contributions for both the binding energy of the metal and also its volume change at melting. The typically high melting temperature of most metals indicates substantial binding energy, and thereby a signiÐcant contribution to the change in enthalpy from the term *Usl/T . An example of a real material composed of volatile molecules of almost spherical shape, the nonassociating hydrocarbon tetramethylmethane (2-2-dimethylpropane), C H , melts at 250.2 K and possesses a latent heat 5 12 *Hsl \ 778 cal mol~1, such that its entropy of transformation *Hsl/T sl \ 3.0 cal mol~1 K~1.30 Thus, SPT predicts a limiting value for the transformation entropy which is close to the measured values of real materials.

10. Discussion
The purpose of this short paper has been to demonstrate that SPT accurately predicts the high-density, gasÈsolid AlderÈ Wainwright transition in a hard-sphere system. The qualitative features of the mathematical computations described above remove all doubts on that question. Furthermore, these results show that SPT additionally describes interesting features of the hard-sphere system not previously been reported : the temperature dependence of the reduced-density parameter at the gasÈsolid phase transition, g(T ) ; and the requireS/0 ment which SPT appears to place upon the hard-core system, as the system temperature approaches absolute zero, to condense into the solid phase at Ðnite temperatures and pressures. The temperature dependence of the density at which the hard-sphere system undergoes a ÑuidÈsolid phase transition, g(T ) , should be expected. The phase rule of Gibbs in no S/0 way depends upon any detail of the intermolecular potential, and holds true for a hard-sphere system as any other. A single-component, hard-sphere Ñuid in a single phase must be described by two independent, intensive variables, and at its binary-phase transition, one. Although it has been shown that the Ñuid phase of a hardcore system, as described by SPT, is characterized by limits determined by the stability condition imposed by the third law of thermodynamics, the detailed quantities calculated above cannot be considered to be strictly accurate, for such were calculated by considering only the entropy of the gas state. An exact calculation of the pressures of the AlderÈ Wainwright transition must include, in addition to the entropy of the gas phase, that of the solid phase of the hardcore system. However, the thermodynamics of the hard-sphere solid have not yet been worked out, and it is therefore not possible to calculate exactly either the entropy of the hard3282 Phys. Chem. Chem. Phys., 1999, 1, 3277È3285

sphere solid or its melting entropy at general temperatures. As has been pointed out in the previous section, the contribution of the low-temperature limit of the melting entropy is approximately 3 cal mol~1 K~1, and this value is of the same order as, and very closely approximates, the entire entropy of melting in simple, real materials. Therefore, in order to examine how inclusion of the entropy of melting might alter the details of the AlderÈWainwright transition, as determined by SPT, the pressures for that transition have been calculated using the condition S(p, T ) ] (*S)sf \ 0. In Fig. 6, are shown the pressures of the AlderÈWainwright transition as determined by the vanishing of the system entropy as calculated both from the Ñuid phase alone, and also with inclusion of the contribution of the entropy of melting, as estimated from its low-temperature limit. The data represented in Fig. 6, show clearly that the inclusion of the entropy of melting has no e†ect upon the qualitative features of the AlderÈWainwright transition and produces only very small quantitative change above the temperature of approximately 100 K. Furthermore, the low-temperature behavior of the AlderÈWainwright transition as predicted by SPT, characterized by the vanishing of the Ñuid phase at all pressures as the temperature goes to zero, remains. The magnitude of the calculated reduceddensity parameter at the density of the AlderÈWainwright transition as predicted by SPT, g(S ] (*S)sf \ 0), deserves particular attention : In Fig. 6, it may be observed that the reduced-density parameter at the AlderÈWainwright transition is essentially indistinguishable from the Ðxed value, gAW \ 0.67114, estimated by Alder and Wainwright, over a range of temperatures extending almost three orders of magnitude. Because any density at which the system entropy vanishes determines a point of absolute instability for that phase, the densities of the AlderÈWainwright transition calculated by SPT should be considered the solid-end points of the gasÈ solid coexistence line for the phase transition, at each temperature. The positions of the gas-end of the gasÈsolid coexistence line can be determined only with an exact solution of the statistical mechanics of the hard-sphere solid, which does not exist. The description of the high-density AlderÈWainwright transition given above by SPT is not restricted to artiÐcial hardsphere gases ; it holds equally true in all its qualitative features for real Ñuids. The mathematical description of the entropy of every real Ñuid system must possess a component which accounts for the entropy of exclusion, as determined by SPT.

Fig. 6 Pressure of the AlderÈWainwright transition, determined with inclusion of the entropy of transformation, (*S)gf B 3/2k . B

The entropy of exclusion is obtained from the impenetrable property of real molecules, and is a consequence of the mathematics of statistical geometry, and is always negative. Furthermore, the negative term which describes the entropy of exclusion of a Ñuid will possess always the second-order singularity determined by SPT, Shc(g) D [h(g)/(1 [ g)2, where h(g) represents any non-singular function of g, usually a polynomial. Even when the complications of a non-spherical shape of the hard-core have been taken into account, as has been done by, for example, Boublik et al.,31 Chapman et al.,32 Nezbeda et al.,33 Vimalchand et al.,34 and Zhou et al.,35 the second-order singularity in the reduced density remains. The same fact holds also when the temperature dependence of the hard core is introduced, as has been done by, for example, Deiters et al.36,37 The entropy of every real gas system is positive at low and modest pressures at all temperatures, of course, for if not, the gas state would always be unstable against transition into the solid state and could not exist. However, as stated above, the total entropy of every real gas system possesses also, at all temperatures, a negative component, the entropy of exclusion attributable to the universal property of impenetrability of its molecules as described by SPT. The (negative) component, the entropy of exclusion, by virtue of its second-order singularity, is unbounded at highdensities. Therefore, at some sufficiently high density, the Ñuid phase of every real gas must become absolutely unstable. That instability is the famous AlderÈWainwright transition. The AlderÈWainwright transition is not an artifact of the machinecalculated simulations with which Alder and Wainwright Ðrst determined the high-density gasÈsolid phase transition ; it is a real phenomenon which occurs in all real Ñuids. Because there are very few models in quantum statistical mechanics which admit exact solutions, the constraints of the third law have not often been applied for stability analysis. Therefore, and in order to show that such analysis does not predict a spurious instability or phase transition, in two Appendices which follow, the identical analysis as has been used for the three-dimensional hard-sphere gas is applied to, respectively, the one-dimensional hard-body, or ““ hard rod ÏÏ, system, and the two-dimensional, or ““ hard-disk ÏÏ, system. In those Appendices is shown that, the constraints of the third law, when applied to SPT and the one- and two-dimensional hard-core systems, do not predict a spurious instability for the Ðrst case, and do predict one-such, and its accompanying Ðrstorder, ÑuidÈsolid phase transition, in the second case, exactly as demonstrated by molecular dynamics computer simulations.38h40 Notwithstanding its use of the modern formalism of SPT, this prediction of the AlderÈWainwright, ÑuidÈsolid phase transition, above, is very much in the spirit of the analysis of the gasÈliquid phase transition by van der Waals. The traditional prediction of the low-temperature gasÈliquid phase transition obtained by the van der Waals equation of state does not strictly describe any liquid conÐguration. When the van der Waals equation of state is used to describe a gas, its isothermal compressibility is observed to change sign at a certain (critical) temperature ; and for all lower temperatures, the gasÈliquid phase coexistence line is inferred a fortiori, using the (appropriately named) ““ Maxwell construction ÏÏ. Similarly, when SPT is used to describe a hard-sphere gas, its entropy is observed to change sign at a certain density ; and for all higher densities, the solid state is inferred a fortiori.

given by the Tonks equation :41 pSPT(1vdim) \

Nk T 1 1 g B \ Nk T ] B L (1 [ g) L L (1 [ g)

C

D
(14)

\ pIG(1vdim) ] phc(1vdim)

where g \ Np/L , in which p represents the length of the hard rod and L the system dimension (““ volume ÏÏ). The Helmholtz free energy of the one-dimensional hard-rod system is given explicitly as : L ] 1 [ ln(1 [ g) FSPT(1vdim) \ [ Nk T ln B Nj \ FIG(1vdim) ] Fhc(1vdim).

CA B

D

(15)

The hard-core terms in the free energy, and the canonical partition function, of a hard-rod system in one dimension are purely entropic, as for the three-dimensional case ; and the internal energy, U, of the hard-rod system is exactly 1/2Nk T . B The entropy of the hard-rod system, given explicitly by S \ (U [ F)/T , is : 3 L ] ln(1 [ g) . ] ln SSPT(1vdim) \ Nk B 2 Nj

C

A B

D

(16)

The kinetic volume, V , of a D-dimensional system may be kin written as (V )Dvdim \ (Nj3)D@3. The condition of validity for kin the hard-core Ñuid in any dimension, D, can be written as (Nj3)D@3/LD \ V /V @ 1. kin If the one-dimensional hard-rod system were compressed to the length at which the entropy, as given by eqn. (16), would vanish, its length would then have to satisfy the equation : 3 V kin \ exp (1 [ g) B 4.482(1 [ g). (17) 2 V S/0 However, both eqn. (17) and the condition of validity cannot be satisÐed simultaneously. Because the limits of the reduceddensity parameter are 0 O g O 1, the value of the system length at which the expression for the entropy, as given by eqn. (16), vanishes, L , clearly lies outside the limits of S/0 validity of the equation of state, and of the hard-rod partition function from which it was obtained. (The mathematically valid but physically meaningless values of, say, g \ 0.99999 . . . , are not considered, for such must be interpreted simply as the system compressed to close-packing, not as a phase transition. For all ÑuidÈsolid phase transitions, the change in density at the temperature of transformation is approximately 10È15%. Therefore, when considering eqn. (17), it would be reasonable to restrict the reduced-density parameter such that g O D0.9.) Therefore, the stability condition that the entropy of the system must always be positive does not predict any phase transition for the one-dimensional hard-core system at any volume within the limits of validity of its description. Application of the constraints of the third law to the one-dimensional system does not involve any prediction of instability or phase transition for the one-dimensional hard-rod gas, and is consistent with other stability analyses.

AB

Appendix B Application of the constraints of the third law to the two-dimensional, hard-disk system :
SPT develops the exact solution also for the equation of state of the two-dimensional hard-body system, or ““ hard-disk ÏÏ system : pSPT(2vdim) \ 1 g(2 [ g) Nk T 1 B ] \ Nk T B A A (1 [ g)2 A(1 [ g)2

Appendix A Application of the constraints of the third law to the one-dimensional hard-rod system
The statistical mechanical problem of a one-dimensional hardbody gas admits an exact solution, and no phase transition. The equation of state of a one-dimensional hard-body Ñuid is

C

D
(18) 3283

\ pIG(2vdim) ] phc(2vdim) Phys. Chem. Chem. Phys., 1999, 1, 3277È3285

where g \ opp2/4, in which p represents the diameter of the hard disk and A the system area (““ volume ÏÏ). The Helmholtz free energy of the two-dimensional hard-disk system is given explicitly as : V g [1] \ ln(1 [ g) ; FSPT(2vdim) \ Nk T [ ln B V (1 [ g) kin (19) and the entropy of the hard-disk system is therefore given explicitly by S \ (U [ F)/T as : g V [ ] ln(1 [ g) . (20) SSPT(2vdim) \ Nk 2 ] ln B (1 [ g) V kin If the two-dimensional hard-disk system were compressed to the area at which the entropy, as given by eqn. (20), would vanish, its area would then have to satisfy the equation : g V kin \ y(g) \ (1 [ g)exp 2 [ (1 [ g) V S/0 g B 7.39(1 [ g)exp [ . (1 [ g)

C A B A B

D

C

D

A

B

A

B

(21)

Fig. 8 Area of the two-dimensional hard-disk system at the point S \ 0, and its kinetic area, A \ (Nj3)2@3, as functions of temkin pereature.

At low densities, eqn. (21) and the condition of validity obviously cannot be satisÐed simultaneously. However, the limits of the reduced-density parameter, or reduced density, are 0 O g O 1, and the value of the function y(g) diminishes rapidly for values of its argument greater than approximately 6. A plot of y(g) is shown in Fig. 7 for values of the argument between 10~4È0.95, in which can be seen that the requirements of inequality of Fig. 7 are easily satisÐed for reduced densities greater than 6. The line V /V \ 1, which marks the kin absolute limits of validity of the SPT formalism is also shown as a broad, dashed gray band on Fig. 7. The region within which the gasÈsolid phase transition occurs is to the right of the trace of y(g) and (well) below the line A /A \ 1. The limits of the validity of the SPT formalkin ism in two dimensions are shown graphically also in Fig. 8 for a two-dimensional gas of particles characterized by a mass of 35 u and a molar covolume of 20 cm3 mol~1, which values are typical for a light gas. In Fig. 9 is shown graphically the value of the reduced-density parameter at its values for which the system entropy of the Ñuid phase vanishes, g(S \ 0). Similarly as a real, three-dimensional system, the transition value of the
Fig. 9 Reduced-density parameter at the transition point, g(S \ 0), of the two-dimensional hard-disk system as a function of temperature.

reduced-density parameter depends weakly upon temperature. At 100 K, gT \ 0.793 ; at 1000 K, gT \ 0.890 K. For reference, the value of the reduced-density parameter at close-packing, gcp \ 2/J3 B 1.15 has also been plotted in Fig. 9. Therefore, as may be noted in Fig. 9, when the constraints of the third law of thermodynamics are applied, SPT unambiguously establishes an absolute instability of the Ñuid phase in a two-dimensional system at temperatures greater than 10 K for values of the reduced-density parameter greater than approximately 0.80È0.88, depending upon the temperature. Such ÑuidÈsolid phase transition in two-dimensional harddisk systems, at exactly these densities, has already been demonstrated (most recently) by Mitus et al.,38 and (two decades ago) by Hoover et al.39

References
Fig. 7 Ratio of kinetic to system area (volume) of a two-dimensional hard-disk system at the point S \ 0, as a function of reduced density. 1 J. F. Kenney, Fluid Phase Equilib., 1998, 148, 21. 2 B. J. Alder and T. E. Wainwright, J. Chem. Phys., 1957, 27, 1208. 3 J. G. Kirkwood, J. Chem. Phys., 1935, 3, 300. 4 M. Born and H. S. Green, Proc. R. Soc. L ondon, 1946, 188, 10.

3284

Phys. Chem. Chem. Phys., 1999, 1, 3277È3285

5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

H. Reiss, H. L. Frisch and J. L. Lebowitz, J. Chem. Phys., 1959, 31, 369. H. Reiss, H. L. Frisch, E. Gelfand and J. L. Lebowitz, J. Chem. Phys., 1960, 32, 119. H. Reiss, ““Scaled particle theory in the statistical thermodynamics of ÑuidsÏÏ, in Adv. Chem. Phys., ed. I. Prigogine, Interscience, New York, 1965, vol. 9. T. De Donder, L ÏAffinity, Paris, 1936. I. Prigogine and R. Defay, Chemical T hermodynamics, Longmans, London, 1954. P. Glansdor† and I. Prigogine, T hermodynamic T heory of Structure, Stability, and Fluctuations, John Wiley, New York, 1971. I. Prigogine, Introduction to the T hermodynamics of Irreversible Processes, John Wiley, New York, 1967. I. Prigogine and C. George, Proc. Nat. Acad. Sci., 1983, 80, 4590. H. Reiss, H. L. Frisch and J. L. Lebowitz, ““Mixtures of hard spheresÏÏ, in T he Equilibrium T heory of Classical Fluids, ed. H. L. Frisch and J. L. Lebowitz, W. A. Benjamin, New York, 1964. H. Reiss, J. Chem. Phys., 1967, 47, 186. H. Reiss and D. M. Tully-Smith, J. Chem. Phys., 1971, 55, 1674. H. Reiss and R. V. Casberg, J. Chem. Phys., 1974, 61, 1107. H. Reiss, ““Scaled particle theory of hard sphere Ñuids to 1976ÏÏ, in Statistical Mechanics and Statistical Methods in T heory and Application, ed. U. Landman, Plenum, London, 1977. D. M. Tully-Smith and H. Reiss, J. Chem. Phys., 1970, 33, 4015. N. F. Carnahan and K. E. Starling, J. Chem. Phys., 1969, 51, 635. H. F. Baker, Proc. L ondon Math. Soc., 1902, 34, 347. J. E. Campbell, Proc. L ondon Math. Soc., 1898, 29, 14. H. F. Baker, Proc. L ondon Math. Soc., 1903, 2, 293. H. F. Baker, Proc. L ondon Math. Soc., 1904, 3, 24. F. Hausdor†, Ber. V erh. Saechs. Akad. W iss. L eipzig, MathNaturwiss., 1906, 58, 19.

25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41

B. J. Alder, W. G. Hoover and D. A. Young, J. Chem. Phys., 1968, 49, 3688. U. K. Deiters and K. M. de Reuck, International Union of Pure and Applied Chemistry, 1997. J. D. Bernal, Nature, 1959, 183, 141. G. O. Jones and P. A. Walker, Proc. Phys. Soc. B., 1953, 69, 1348. R. H. Fowler and E. A. Guggenheim, Statistical Mechanics, Cambridge University Press, Cambridge, 1939. J. Timmermans and L. De†et, L e Polymorphisme des Composes Organiques, Gauthier-Villars, Paris, 1939. T. Boublik, C. K. Viga and M. Diaz-Pena, J. Chem. Phys., 1990, 93, 730. W. G. Chapman, K. E. Gubbins, G. Jackson and M. Radosz, Ind. Eng. Chem. Res., 1990, 29, 1709. I. Nezbeda, Mol. Phys., 1997, 90, 661. P. Vimalchand, A. Thomas, I. G. Economou and M. D. Donohue, Fluid Phase Equilib, 1992, 73, 39. Y. Zhou, S. W. Smith and C. Hall, Mol. Phys., 1995, 85, 1157. U. K. Deiters and S. L. Randzio, Fluid Phase Equilib, 1995, 103, 199. U. K. Deiters, Mol. Phys, 1992, 96, 539. A. C. Mitus, H. Weber and D. Marx, Phys. Rev., 1997, E55, 6855. W. G. Hoover, N. E. Hoover and K. Hansen, J. Chem. Phys., 1979, 70, 1837. E. Helfand, H. L. Frisch and J. L. Lebowitz, J. Chem. Phys., 1961, 34, 1037. L. Tonks, Phys. Rev., 1936, 50, 955.

Paper 9/01700C

Phys. Chem. Chem. Phys., 1999, 1, 3277È3285

3285