Statistical mechanics of nucleation: a review
Transcript of Statistical mechanics of nucleation: a review
SPECIAL ISSUE PAPER
Statistical mechanics of nucleation a review
I J Ford
Department of Physics and Astronomy University College London Gower Street London WC1E 6BT UK
Abstract The principles of statistical mechanics have been used to develop a theory of thenucleation of a phase transition but a number of subtle questions remain and are highlighted in thisreview A central issue is the cluster definition the mathematical scheme which distinguishes amolecular cluster from a collection of separate molecules There is also the question of whetherthermodynamic transition state theory is suitable to describe the process or whether the developmentof kinetic rate equations describing cluster growth and decay is a better approach The classical theoryof nucleation is flawed but appears to provide useful estimates in some cases including waterPhenomenological extensions of the classical theory can improve matters However improvements inthe theory from microscopic considerations are not simple to apply require major computationaleffort and suffer from uncertainties due to lack of knowledge of the fundamental intermolecularinteractions Calculations of the nucleation rate of water droplets are especially difficult since thissubstance is notoriously difficult to model Nevertheless as capabilities improve accurate calculationsshould come within reach which will offer better understanding of the process for practicalapplications such as the transition from dry to wet steam
Keywords statistical mechanics nucleation
NOTATION
A parameter in the fit to experimental dataB parameter in the fit to experimental dataf mean force on the moleculef0 free energy per molecule in the metastable
phaseFethiTHORN free energy of an i-clusterf Langevin forceHethNTHORN system Hamiltoniani cluster size in moleculesi critical sizeI1 I2 I3 principal moments of inertia of clusterIm1 I
m2 I
m3 principal moments of inertia of the
moleculeJ nucleation ratek Boltzmann constantm molecular massM cluster massni population of i-clustersncei constrained equilibrium cluster
populationsnei equilibrium cluster populations in
subsaturated vapour
nsi i-cluster population in saturated vapournssi steady state cluster population in
supersaturated vapourns1 monomer population in saturated vapourpv vapour pressurepsv saturated vapour pressurePij transition probability in population
dynamicsq sticking coefficientq0 parameter in the DillmannndashMeier modelr radial position from the cluster centre of
massre escape radiusS supersaturationt timeT temperatureU0ethiTHORN cluster potential energy at restUm
0 molecular potential energy at restV system volumew molecular probability density in spaceZd Zeldovich factorZi i-cluster canonical partition functionzi modified i-cluster canonical partition
function
ai parameter in the DillmannndashMeiermodel
bi growth rate of an i-clusterThe MS was received on 15 July 2003 and was accepted after revision forpublication on 7 April 2004
883
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
g Langevin friction coefficientgkin cluster decay rate in the Langevin modelgi decay rate of an i-clusterDF change in free energyDW work of cluster formationDWcl classical work of cluster formationyi parameter in the DillmannndashMeier modelki parameter in the DillmannndashMeier model chemical potentials chemical potential in saturated vapourX grand canonical partition functionrl liquid phase densityrv vapour phase densityrsv saturated vapour densitys planar surface tensiont parameter in the DillmannndashMeier modelok cluster vibrational angular frequencyom
k molecular vibrational angular frequency
1 INTRODUCTION
One of the most remarkable properties of matter is itsability to take different physical forms for differentvalues of parameters such as temperature or pressureFor example a vapour will condense by the nucleationof droplets when its pressure is raised above thesupersaturated vapour pressure or when its temperatureis reduced below the dew point Such metamorphosesare fundamentally a consequence of the second law ofthermodynamics a principle that can be characterizedas a desire on the part of physical systems to minimizetheir free energy by any available means This caninclude the repositioning of molecules on a grand scalea phase change Phase transformations are the statisticalconsequences of the ability of large physical systemsand any environment with which they interact toexplore a huge range of microscopic configurationsThis dynamical behaviour underpinning the secondlaw remains only partially understood [1] but it seemsto be universal
A practical situation where a change in phase createsproblems is the behaviour of steam in turbines Thetemperature and pressure gradients are such that waterdroplets might nucleate with undesirable effects on theperformance of the machine including the erosion of theturbine blades On the other hand the steam must beexpanded as much as possible to extract the most workfrom it Some of the drawbacks can be avoided if thephase change can be controlled
The need for an understanding of the transition fromdry to wet steam in turbines has motivated this reviewbut this is just an example of a process which affectsmany aspects of the world The formation of droplets in
a turbine presents practical difficulties but withoutatmospheric clouds which form by a similar process [2]the global climate would be rather different from what itis at present [3] Aerosol formation in the atmosphereand dust in space [4] and in the case of condensedphases the freezing of liquids all proceed by the samebasic mechanism The principal idea is that there exists abottleneck in the transformation which is passedthrough only by fragments or molecular clusters ofthe new phase The bottleneck is narrowest when theclusters reach a so-called critical size and therefore theproperties of these critical clusters are central to thetheory of nucleation [5]
Experimental progress in understanding droplet for-mation in turbines has been significant [6 7] but at firstsight it would appear that progress on the theoreticalside has been less impressive The tool used most oftenin modelling studies [8] remains the classical nucleationtheory (CNT) derived many decades ago and based onthe most primitive ideas of the properties of the criticalcluster [9ndash13] The capillarity approximation isemployed whereby the critical cluster however smallis considered as a scaled-down macroscopic droplet ofthe condensed phase The model is a version oftransition state theory (TST) in which a realization ofthe system is identified which is in unstable thermo-dynamic equilibrium equally able to develop with timeinto the initial or the final phase How is it that thismodel is still employed when developments in micro-scopic modelling have proceeded so rapidly in recentyears Why has no better model been developed
One reason is pragmatism It is very surprising thatthe classical theory successfully predicts nucleation rateswithin a couple of orders of magnitude in the case ofwater perhaps the most important substance to under-stand for a temperature range of about 210ndash260K anda range of supersaturations Evidence for this has beencollected through various studies over the last 20 years[14ndash20] This level of agreement is good in the field ofnucleation theory However in these conditions thecritical cluster consists of only a few tens of moleculesand the capillarity approximation is questionable in theextreme
However if a broader view is taken and data for othermaterials are considered the failure of CNT is moreapparent The nucleation rate seems to have a depen-dence on supersaturation suggested by CNT but thetemperature dependence does not agree Often atemperature range where classical theory is reasonablyconsistent with experiment does not exist
Experimental data gathered during recent years canbe reviewed briefly Some early data on n-nonaneshowed deviations from CNT of four orders ofmagnitude either way [21] A study of several membersof the n-alkane series [22] confirmed the need forcorrection factors of again up to four orders ofmagnitude Early data for a range of alcohols showed
I J FORD884
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
deviations from CNT of over five orders of magnitude[23] In an effort to acquire definitive data for a singlesubstance the nucleation of n-pentanol was givenparticular attention by a number of research groups inthe late 1990s and early 2000s For temperatures in therange 260ndash290K most data lie three orders of magnitudeabove the CNT predictions [24] although some measure-ments lie between one and three orders of magnitudebelow [25] Further results for n-pentanol show goodagreement with the CNT at 273K [26] but withdeviations away from this temperature In the light ofthe reasonable success of CNT in explaining water datait is disappointing to find that hydrogen-bonded organicmaterials such as a series of glycols disagree verystrongly with the temperature dependence of thenucleation rate suggested by CNT [27] Similarly theorganic species dibutyl phthalate deviates by up to sixorders of magnitude from the model [28] The situationbecomes even more unsatisfactory for the nucleation oftwo or more species together Strong deviations frombinary CNT are found for the ethanolndashwater system [29]and for a range of n-nonanendashn-alcohol mixtures [30]However it is with metals that the failure of CNT ismost dramatic examples in the literature being caesium[31] mercury [32] and lithium [33] For caesium thecritical supersaturation (that required to drive nuclea-tion at a specified rate) is between two and ten timeslarger than the CNT predictions The difference betweenmeasured and predicted nucleation rates is astronom-ical Although in most materials the deviation is lesssevere than this it is certainly not the case that a smallcorrection to CNT is all that is needed to account for alldata
The usual CNT expression for the rate is
JCNT frac14 q2spm
1=2 rvrl
pvV
kT
6 exp 16ps3
3ethkTTHORN3r2l ethlnSTHORN2
eth1THORN
where q is a sticking or accommodation coefficient m isthe molecular mass k is Boltzmannrsquos constant s is thesurface tension of a planar interface between thecondensate saturated vapour pv is the vapour pressureand rl and rv are the densities of the bulk condensedphase and the vapour respectively in molecules per unitvolume S is the supersaturation defined by S frac14 pv=p
sv
where psv is the saturated vapour pressure at thetemperature T Note that J in this expression is thenumber of particles formed per second and it isproportional to the system volume V There are goodreasons to insert a factor of 1S into the prefactor of theabove expression but this is the often-quoted formCalculations are easy agreement is reasonable and sowhy should anyone bother to develop the theoryfurther
The problem in employing this formula for water isthat the laboratory data for water are taken at lowtemperatures and need to be extrapolated to hightemperatures for application to turbine conditions Asalready noted water seems to be a lucky case wheretheory and experiment actually agree for a particular setof conditions namely at a temperature of about 240Kand for a range of supersaturations Unfortunately adiscrepancy appears and grows as the temperature israised and at 260K the theory overpredicts the rate byan order of magnitude [18] The ratio of measurednucleation rate to CNT prediction has been fitted to asimple function by Wolk and Strey [18]
JexpethT STHORN frac14 JCNT exp Athorn B
T
eth2THORN
where A frac14 2756 and B frac14 6500K This formula fitsdata in the temperature range 220ndash260K but has alsobeen found to account for data at temperatures sometens of kelvins higher (B Wyslouzil personal com-munication 2003)
Nevertheless caution must be exercised in extrapolat-ing to temperatures around 370K for steam turbinesDevelopments in the theory are certainly needed It is achallenge to understand why CNTworks reasonably wellat 240K forwaterwhat terms are cancellingout andwhyHow canmeasured data be extrapolated with confidence
In this review the determination of the properties ofmolecular clusters using the tools of statistical mech-anics is again considered Similar accounts of the widerpicture have been given in various reviews and books[34 35] Starting from very general arguments attemptsare made to discern how a theory such as the CNTmight emerge The subtleties of defining a clusterprecisely in terms of the positions and momenta of itsconstituents are discussed It is noted that the transitionstate theory based on equilibrium cluster properties isequivalent to an approach based on the kinetic theory ofcluster decay at least in certain circumstances Somephenomenological models developed from CNT arereviewed and finally a perspective of future develop-ments in this field is given
By the very nature of a review the discussion is ratherbroad while at the same time some of the technicalitiesare outlined The intention is to give a reasonablycomplete picture of the current state of nucleationtheory at least from the point of view of the presentauthor and to show that significant progress has beenmade in the approximately 80 years of its history whilemore is still clearly necessary
2 TRANSITION STATE THEORY
A theory of nucleation based on statistical ideas isnecessary since direct simulation of the process from a
STATISTICAL MECHANICS OF NUCLEATION 885
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
microscopic point of view is very laborious and has beenrealized in only a very few example cases [36ndash39] Thetime scales involved make a nucleation event simply tooslow to follow directly for realistic conditions and withcurrent computational resources
The simplest statistical treatment of the nucleationprocess comes from TST within a framework of thetheory of free energy fluctuations According to thermo-dynamics a system with externally fixed temperatureand pressure (or volume) will adjust any internal prop-erties in order to minimize its Gibbs (or Helmholtz) freeenergy [40 41] Such adjustments will include changes inphase Nevertheless statistical fluctuations in systemproperties will naturally take place and it is useful tointerpret these as fluctuations in the free energy of thesystem away from the minimum Considering constant-volume conditions for the moment general arguments[42ndash45] suggest that the probability of a fluctuationDF in Helmholtz free energy is proportional toexpfrac12DF=ethkTTHORN
The physical fluctuations that are important in thenucleation process are those where a molecular clusterof new phase is formed temporarily from the originalphase The change in free energy accompanying theformation of a cluster of i molecules in the new phase iswritten DF frac14 DWethiTHORNFethiTHORN if0 where FethiTHORN is the freeenergy of the i cluster and f0 is the free energy permolecule in the old phase This is also called the lsquoworkof formationrsquo since it corresponds in thermodynamics tothe external work needed to insert an additional clusterreversibly into the system while maintaining the overall
temperature and pressure (or volume) Note that thisinsertion of a new cluster while maintaining suchconstraints is a theoretical operation since it would behard to achieve in practice
For a metastable initial phase the magnitude of thefree-energy fluctuation DF peaks at the so-called lsquocriticalsizersquo i This is illustrated in Fig 1 The perturbed statewith the additional critical cluster is therefore athermodynamic transition state since reversible changesin size of the extra cluster are accompanied byreductions in free energy It is presumed that this meansthat the new cluster has equal probabilities of growingor decaying and that the rate of nucleation is thereforerelated to the rate of formation of these critical clusters
The probability of forming the critical clusteraccording to this picture is proportional toexpfrac12DWethi THORN=ethkTTHORN but the proportionality constant(essentially a time scale) is not yet determined Tointroduce this simple kinetic theory [46] can be used tocalculate the collision rate bi between single molecules(monomers) and an i-cluster modelled as a sphere withbulk condensed phase molecular density rl
bi frac14qpv
eth2pmkTTHORN1=21thorn 1
i
1=2
1thorn i1=3 2
eth36pTHORN1=3
r2=3l
eth3THORN
where m is the molecular mass It is more usual toreplace the expression in square brackets with i2=3 thelimit for large i The simplest TST expression for the
Fig 1 Sketch of the cluster work of formation for a range of (constrained) cluster sizes The maximum in thecurve defines the critical size i and the critical work of formation DWethi THORN
I J FORD886
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
nucleation rate is then J bini where ni is the meannumber of critical clusters in the system
More sophisticated treatments of thermodynamicscan provide the time scale without recourse to the abovekinetic argument which is reasonably valid for gas-to-particle conversion but not for freezing for exampleAttempts based on linearized non-equilibrium thermo-dynamics [47 48] originate from the work of Onsager[49]
The next step is crucial and it is to relate thepopulation ni to the probability of a fluctuation and toan explicit model of the critical work of formation Theassumption in classical nucleation theory is that
ni frac14 n1 expDWclethi THORN
kT
eth4THORN
where DWclethiTHORN is the classical work of formation givenby
DWclethiTHORN frac14 4pR2s ikT lnS eth5THORN
and DWclethi THORN is the maximum of this function Thecluster is envisaged to be a sphere of radius R given by4pR3rl=3 frac14 i
Unfortunately some uncertainty accompanies thesesteps in the derivation The theory is flawed on severalcounts The work of formation contains a termassociated with the creation of a spherical interface(characterized by the bulk value of surface tension) anda change kT lnS in free energy per molecule due tothe change in phase from supersaturated vapour tocondensate Moreover the factor n1 appearing inequation (4) appears to have been introduced heuristi-cally Might it not instead be the monomer populationin the saturated vapour How does DWclethiTHORN comparewith the actual work of formation DWethiTHORN Is thedroplet really spherical with a sharp interface Have
any important factors been left out Furthermoreshould there be concern that equation (4) is in-consistent when i frac14 1 This has been a controversialissue [50ndash54]
3 KINETIC INTERPRETATION
The expression for the classical rate of nucleation givenin equation (1) emerges from the models in equation (4)and (5) when the thermodynamic transition stateanalysis is embellished by a kinetic interpretation [11]Instead of using the probability of fluctuations awayfrom the equilibrium state (the TST route) a set of rateequations is introduced to describe the dynamics ofcluster populations
The evolution of cluster populations is modelled usingthe following equations
dni
dtfrac14Xj
njPji niXj
Pij eth6THORN
where Pji is the mean rate at which a j cluster convertsinto an i cluster The transition processes are assumed tobe Markovian so that the rate coefficients depend onthe properties of the cluster and its environment but noton their history
The kinetic interpretation can be taken a step furtherby assuming that the only important transitions arethose involving the addition or loss of single moleculesto or from the cluster The scheme of transitionprocesses is illustrated in Fig 2 The only non-zerorate coefficients are then bi frac14 Piithorn1 for cluster growthand gi frac14 Pii1 for cluster decay The rate equations
Fig 2 The scheme of the birth and death of clusters addition of a monomer to a cluster will take it up thechain towards the right while the loss of a monomer will take it in the opposite direction Attachmentand detachment of dimers and larger clusters are usually ignored
STATISTICAL MECHANICS OF NUCLEATION 887
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
reduce to
dni
dtfrac14 bi1ni1 gini bini thorn githorn1nithorn1 eth7THORN
These are the BeckerndashDoring equations [11] and theyprovide a useful reinterpretation of the TST approachThere are two steady state solutions In the first thepopulations satisfy equation (7) and also the detailedbalance condition
bi1nei1 frac14 gin
ei eth8THORN
and the fnei g are taken to represent the cluster popula-tions in an equilibrium vapour at or below the saturatedvapour pressure For true thermal equilibrium themonomer population has to be less than or equal to itsvalue for the saturated vapour ns1 The cluster populationsin a saturated vapour are denoted fnsig Nevertheless asituation can be invented which is known as lsquoconstrainedequilibriumrsquo where the detailed balance condition is stillsatisfied but with n1 gt ns1 Since bi pvS while gi isindependent of vapour pressure equation (8) then definesso-called constrained equilibrium populations
ncei ampSinsi eth9THORN
for a supersaturated vapour It has been assumed thatS frac14 pv=p
svampn1=n
s1
The second steady state solution to the BeckerndashDoring equations is more realistic It is not meant torepresent thermal equilibrium The supersaturation isset to a value greater than unity and a boundarycondition of zero population at some cluster size imax isimposed These conditions will drive a steady productionrate of large clusters from the vapour The approach isfor mathematical convenience only and the same steadystate cluster populations nssi and nucleation rate wouldemerge if imax was sent off to infinity The detailedbalance condition (8) is replaced by J frac14 bi1n
ssi1 gin
ssi
defining the nucleation current J which can then becalculated analytically as shown in another paper in thisvolume [55] To a good approximation
J frac14 Zdbincei eth10THORN
which involves the constrained equilibrium population atthe critical size i the size with the lowest constrainedequilibrium population The condition imax 4 i musthold Zd is the so-called Zeldovich factor [12] givenapproximately by
Zd frac14 1
2pq2ethln ncei THORN
qi2
ifrac14i
1=2
eth11THORN
When equations (4) and (5) are used to specify ncei and theresult is substituted into equations (10) and (11) then theclassical nucleation rate given in equation (1) isrecovered as long as the large i limit of the attachment
rate coefficient in equation (3) is taken It is also possibleto derive an expression for J purely in terms of the ratecoefficients bi and gi This possibility is considered againin section 8
The main assumptions used in this derivation of theclassical theory principally the validity of equations (4)and (5) are now examined In order to explore this it isnecessary to see how the equilibrium cluster populationsnei in subsaturated vapours may be calculated usingstatistical mechanics
4 CLUSTER STATISTICAL MECHANICS
The statistical mechanical treatment of a vapour is bestdeveloped using a grand canonical ensemble [56ndash59] Thesystem is placed in contact with a particle reservoir at achemical potential and temperature T so that thenumber of molecules N in the system can change Thegrand partition function of the system X is then propor-tional to the integral of expf frac12HethNTHORN N=ethkTTHORNg overall possible configurations of molecular positions andmomenta whereHethNTHORN is the Hamiltonian of the system
Evaluating X exactly is exceedingly difficult and it ismuch simpler to proceed by considering the configura-tions of the system as arrangements of molecularclusters The vapour then is viewed as a mixture ofideal gases each composed of clusters of differentsizes The grand partition function of the system canbe reconstructed from partition functions for singlei-clusters In order to do this however a cluster must bedefined in a unique way in terms of molecular positionsand velocities An example of such a definition would beto require the separation between molecules in a clusterto be less than a maximum distance Rc [60] as illustratedin Fig 3
Fig 3 According to the Stillinger definition a molecule isconsidered part of a cluster if it lies closer than adistance Rc from a constituent molecule of the cluster
I J FORD888
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
At first sight the cluster definition is completelyarbitrary a means of categorizing and resumming allthe elementary configurations of the system Howeveraccurately representing the system as a mixture of idealgases requires the interactions between molecules lyingin different clusters to be minimized The clusterdefinition affects whether a particular interaction isregarded as intercluster or intracluster and so it plays acrucial role This will be discussed further in section 6Residual clusterndashcluster interactions can be taken intoaccount in the form of a virial series [61]
These ideas are now put into practice Define amodified canonical partition function zi for a clustercontaining i molecules This is simply the integral ofexpf frac12HcethiTHORN i=ethkTTHORNg over configurations involvingmolecules in the cluster alone with Hc the Hamiltonianof the i molecules comprising the cluster Using thepicture of a mixture of ideal gases the grand partitionfunction of the complete system can then be expressed as
XampXfnig
Yifrac141
znii
ni
eth12THORN
where the sum is over all possible cluster populationdistributions fnig This is demonstrated rigorously inappendix A of reference [57]
The next step is to determine the distribution fnei gthat dominates the sum in equation (12) The logarithmof the summand is extremized
qqni
Xifrac141
lnz
nii
ni
frac14 0 eth13THORN
which gives the following very simple expression for themost probable or equilibrium cluster size distributionfor the given conditions T and
nei frac14 ziethTTHORN frac14 ZethTTHORN ei=ethkTTHORN eth14THORN
which is equivalent to the so-called law of mass action
nei frac14 Zine1Z1
i
eth15THORN
where ZiethTTHORN frac14 expfrac12FethiTHORN=ethkTTHORN frac14 zi expfrac12 i=ethkTTHORN isthe canonical partition function for an i cluster and FethiTHORNis the i-cluster Helmholtz free energy This relationshipis often used as the starting point in models of clusterpopulations
The vapour pressure is now given by a sum of partialpressures
pvV frac14 kTXifrac141
nei frac14 kTXifrac141
ziethTTHORN eth16THORN
and this series will only converge if the equilibriumpopulations or the zi decrease sufficiently quickly forlarge i This limits the statistical mechanics approach to
the study of saturated or subsaturated vapours How-ever as in the preceding section the equilibriumpopulations nei in equation (14) (for S lt 1) can beextrapolated into constrained equilibrium populationsncei (with S gt 1) for use in equation (10) Saturatedvapour corresponds to a particular chemical potentialms and the saturated vapour cluster populations are nsi The constrained equilibrium populations are obtainedby inserting into equation (14) the chemical potential ofthe supersaturated vapour which is frac14 s thorn kT lnSgiving ncei frac14 Sinsi which is consistent with the resultobtained in equation (9) for dilute vapours
The constrained equilibrium populations are there-fore expressed as
ncei frac14 SiziethsTTHORN
frac14 expfrac12FethiTHORN is ikT lnS
kT
eth17THORN
and the correspondence with CNT can be made closerby introducing the n1 prefactor
ncei frac14 n1 exp frac12FethiTHORN is thorn kT ln n1 ikT lnS
kT
eth18THORN
The next task is to explore how the combinationFethiTHORN is thorn kT ln n1 might be approximated by theCNT result 4pR2s This can be explored most easilyfor solid rather than liquid clusters
5 SOLID CLUSTERS
The partition function of a solid cluster of i moleculescan be factorized into translational rotational andvibrational parts through a transformation to suitablerelative positions and momenta [62ndash64] Starting withthe vibrational degrees of freedom the atom positionsare defined with respect to their mean positions whichare determined by a fixed centre of mass and set oflattice vectors It is assumed that the potential energypart of the Hamiltonian is quadratic in the displace-ments from these mean positions Now the partitionfunction is worked out for configurations of the particlesrestricted to having zero total linear and total angularmomentum about the centre of mass This is writtenZvib corresponding to a product of partition functionsfor 3ji 6 independent oscillators (if there are j atomsper molecule and therefore a total of 3ji degrees offreedom if the molecule is not linear) each of which inthe classical limit takes the form kT=ethhoTHORN where o isthe angular frequency of the specific vibrational modeThe partition function has an additional factor ofexpfrac12U0ethiTHORN=ethkTTHORN where U0ethiTHORN is the system potentialenergy evaluated at the mean atom positions A
STATISTICAL MECHANICS OF NUCLEATION 889
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
different expression applies for low temperatures whenthe oscillator is not classically activated [65]
The rotational and translational degrees of freedomare next The previous restriction of the configurationsto zero angular momentum is lifted The approximate(classical) rotational partition function is
Zrot frac14 p1=28p2kTh2
3=2
ethI1I2I3THORN1=2 eth19THORN
where Ik are the principal moments of inertia about thecentre of mass for atoms fixed in their mean positions
Finally the translational free energy of the centre ofmass of the cluster is given by the partition function
Ztrans frac14 V2pMkT
h2
3=2
eth20THORN
where V is the system volume and M frac14 im is the clustermass The free energy of the cluster (plus kT ln n1 forconvenience) is therefore
FethiTHORN thorn kT ln n1
frac14 kT lnZtransZrotZvib
n1
frac14 U0ethiTHORN kT ln
eth2pMkTTHORN3=2
Srsvh3p1=2
8p2kTh2
3=2
6ethI1I2I3THORN1=2Y3ji6
kfrac141
kT
hok
eth21THORN
where rsv frac14 ns1=V is the molecular density of dilutesaturated vapour
By similar arguments the chemical potential of dilutesaturated vapour is
s frac14 Um0 kT ln
eth2pmkTTHORN3=2
rsvh3p1=2
8p2kTh2
3=2
6ethIm1 Im2 Im3 THORN1=2Y3j6
kfrac141
kT
homk
eth22THORN
where Um0 is the potential energy of a molecule when the
atoms are at their mean positions Imk are the threeprincipal moments of inertia of the molecule about thecentre of mass and om
k are the oscillatory frequencies ofmolecular internal vibrations
The classical work of formation DWclethiTHORN should nowcorrespond to the leading-order contribution to thedifference between equation (21) and i times equation(22) Calculations [54 64 66ndash68] demonstrate that theleading term is indeed roughly proportional to i2=3 thusjustifying the form of the classical model The suitabilityof the capillarity approximation depends on the extentto which the coefficient of this leading term correspondsnumerically to the planar surface free energy Howeverthe full statistical mechanical expression for the work of
formation contains additional terms and these will beimportant for small clusters and need to be taken intoaccount In order to evaluate them the variousstructural and energy properties of cluster are requiredall of which can be obtained from microscopic modelsand they can then be inserted into the formulae for thethermodynamic properties [equations (21) and (22)] theconstrained equilibrium populations [equation (18)] andultimately the nucleation rate [equation (10)] Followingthis route employs the almost universally held assump-tion that equilibrium properties such as free energies canbe used to describe non-equilibrium properties specifi-cally the cluster decay rate This assumption is re-examined in section 8
6 LIQUID DROPLETS AND THE CLUSTER
DEFINITION
Defining what was meant by a solid cluster in theprevious section was not an issue Within the assump-tions of the model (harmonic interactions) it wasimpossible for a molecule to escape from the clusterThe cluster definition in the case of liquids cannot bedismissed so easily Liquids are more volatile thansolids and the molecules in a cluster are much moremobile It is imperative now to characterize a clusterusing the molecular positions and velocities andperhaps some time scale and length scale too
The traditional microscopic method for defining acluster is to impose a geometric constraint on molecularpositions An example is to restrict molecules to a spherecentred on the centre of mass [69ndash71] or to impose amaximum allowable separation between the molecules(the Stillinger criterion [60]) More sophisticated treat-ments exist such as the n=v Stillinger cluster introducedby Reiss and co-workers [72ndash75] Here the confiningvolume v used to define the cluster is considered to be anadditional defining characteristic on the same level asthe cluster size n As time progresses a cluster changesits confining volume as the component molecules followtheir respective trajectories just as they change their sizewith the gain and loss of molecules The association of amolecular configuration with a precise confining volumeis complicated but tractable The dependence of the freeenergy on v determines the statistical dynamics of thiscluster volume A similar attempt to associate a clusterwith a specified volume within a grand canonicalensemble description was developed by Kusaka andco-workers [58 59 76] The simulation box was taken tobe the confining cluster volume Many of these ideasmay be viewed as a treatment of the cluster definition(eg the Stillinger radius) as a variational parameter inthe representation of the partition function Further
I J FORD890
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
refinement of this point of view using a treatment ofclusterndashcluster interactions to optimize the definingcluster volume has been presented [77] A review ofrecent progress in this area and the application of theseideas to nucleation theory has been provided by Sengeret al [78]
Such geometric constraints can be implemented buttheir realism is questionable Situations better regardedas close encounters between separate clusters would beclassed as snapshots of single larger clusters It shouldbe recalled that the cluster definition is an internalfeature of the model which is to be optimized to reduceclusterndashcluster interactions and to improve the repre-sentation of the grand partition function Misclassifyingconfigurations is fraught with consequences A secondissue to consider is that for convenience the clustersshould behave in a manner that can be modelled withBeckerndashDoring equations This means that the growthand decay rates should be free of memory effects ie thetransition processes should be Markovian The extent towhich this is satisfied (if indeed it is physically realistic)depends on the cluster definition
An intuitively reasonable choice of definition whichaddresses both these points is that a cluster shouldconsist of molecules that are bound energetically Thisshould minimize the clusterndashcluster contributions to thetotal energy According to such a definition if amolecule interacts strongly with a cluster then it islikely to be regarded as part of that cluster Also anenergetically bound cluster is likely to decay by thermalfluctuation in a stochastic Markovian fashion ratherthan by the separation of passing poorly boundfragments
Ideas have been proposed which require the cluster tobe energetically bound (in a loose sense since a cluster isintrinsically prone to decay) The time evolution of amolecular configuration and not simply the set ofinstantaneous positions of its component molecules isof central importance Hill [79] introduced a clusterdefinition requiring the total energies of pairs ofmolecules in a cluster to be negative (in their centre ofmass frame) The intention was to exclude situationswhere a molecule could leave the cluster within the timeit would take for a molecule to travel a typical lineardimension of the cluster Soto and Cordero [80] havedeveloped Hillrsquos criterion and Barrett [81] has per-formed Monte Carlo modelling of molecular configura-tions excluding those which decay in moleculardynamics within a given period A number of otherenergy-based criteria have been developed [82ndash85]
Harris and Ford [86] have recently proposed a newenergy-based cluster definition As with Barrettrsquosapproach it is based on the requirement that amolecular snapshot is energetically bound for a chosensubsequent period Evaporation according to thisscheme requires not only that a molecule should makea positive energy excursion but also that the subsequent
dynamics carry it far away from the cluster Onlyperhaps one in eight molecules manage to avoidrecapture once having satisfied the positive energycriterion [86] A snapshot of a cluster in this casesurrounded by a carrier gas is shown in Fig 4 Thehistory of the energy of a molecule in such a cluster isillustrated in Fig 5 three excursions into positive energyoccur in this sequence but only one leads to actualescape corresponding to the maintenance of positiveenergy for a significant time The instant of escape istaken to be the point at which the energy goes positivefor the third and final time
Cluster definitions implemented in moleculardynamics and Monte Carlo codes allow cluster freeenergies and hence the decay rate to be calculated Anenormous amount of simulation work has been carriedout and calculations can be laborious Many differentschemes and approximations have been made to easethe evaluation of these properties The principalconclusion from various studies using a range of clusterdefinitions and interaction models [49 57 58 70 85 87ndash92] seems to be that the work of formation of liquidcritical clusters does indeed contain a dominantcontribution proportional to i2=3 consistent with theclassical work of formation Hale and collaboratorspresent cluster free energies for water and other speciesin a way that makes this particularly clear indeed it issurprising how dominant the surface term can be foreven very small clusters [70 93 94] Neverthelessfurther contributions exist small clusters are notcharacterized just by scaled-down properties of largeclusters and it is impossible to escape the need todetermine the additional terms
In all the above discussion it has been assumed thatthe intermolecular interactions are accurately specifiedThis is a situation that is rarely possible Althoughfirst-principles quantum-mechanical models could be
Fig 4 Snapshot of a cluster of 50 argon atoms (brightspheres) bathed in a gas of helium (darker spheres)
STATISTICAL MECHANICS OF NUCLEATION 891
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
developed in practice empirical potentials are devel-oped on the basis of physical intuition containingparameters fitted to the measured properties of thesubstance in question [95] Usually these are bulkproperties and it is not always clear that the sameinteractions can be used for the very different circum-stances of a molecular cluster [96] In view of thesensitivity of nucleation rates to the thermodynamicproperties of clusters this is perhaps a crucial failingProgress can be made however by computing knowncluster properties where they are available and byrefining the potentials to suit Extraction of clusterproperties through use of the so-called nucleationtheorems (which relate derivatives of the nucleationrate to properties of the critical cluster [16 56 97ndash99])may provide just this sort of experimental information
A technique that has not yet been described is the useof density functional theory (DFT) to model theproperties of liquid droplets [100ndash104] This powerfulapplication of statistical mechanics promises to avoidlaborious computations at the microscopic level byinstead focusing on the density profile of moleculesmaking up the droplet This profile may be regarded asthe average distribution of molecules about the clustercentre of mass Although this might appear to involvecoarse-graining and approximation it is a well-foundedapproach and in principle can be exact In practical
implementations however a mean field approach istaken such that the constituent molecules move in aneffective potential well created by the other molecules inthe cluster [105] Also the interactions are assumed to becharacterized by strong repulsive cores and weaklyattractive tails
Considerable analytical work has been carried out[104 106] Various developments have been made todescribe droplets consisting of simple atoms as well aschain-like molecules [107] and polar molecules [108]The main drawback in these implementations is thatspecific interactions such as hydrogen bonds are hardto represent Certain details of the modelled systemare neglected [109 110] and the (implicit) clusterdefinition used is geometric and not energetic [111]On the other hand quite sophisticated systems havebeen modelled including multi-component dropletsThe method has been compared with otherapproaches [112 113] and certain features are betterdescribed particularly the approach to spinodalconditions where the nucleation barrier vanishesDFT calculations can be performed relatively easilyand they can provide a qualitative and perhapsquantitative insight into quite complicated clusteringproblems They provide an intermediate-level descrip-tion between the simple capillarity approximation anda microscopic description
Fig 5 The energy of a molecule in the centre-of-mass frame of a cluster occasionally becomes positive butthis does not necessarily lead to escape Only the third positive energy excursion in this sequenceindicated by an arrow results in particle escape A cluster definition must capture this behaviour
I J FORD892
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
7 PHENOMENOLOGICAL MODELS
The reader might be dismayed by the mathematicalcomplexity of the expressions for the work of formationof solid clusters in section 5 might be alarmed by theprocedural complexity of implementing a cluster defini-tion in section 6 may recoil from the difficulties incalculating free energies from molecular simulationmethods and ultimately may give up in view of theuncertainties in the underlying intermolecular interac-tions Considering these complications it is tempting totry to develop phenomenological models of the work offormation After all if the capillarity approximationworks reasonably well for water then surely a minorcorrection to account for small cluster effects ought tobe sufficient to improve matters
However after perhaps two decades of work nomodel has emerged offering broad success across arange of different substances It is perhaps asking toomuch to expect universal improvement CNT works as afirst approximation for water for a particular tempera-ture range but as discussed earlier it possesses thewrong temperature dependence so that predictionsaway from a temperature in the range 210ndash260K areuncertain A similar picture emerges for other sub-stances although in some cases the failure is even moresevere
The simplest phenomenological adjustment to equa-tion (4) the basis for CNT is to write
ncei frac14 n1 expfrac12DWclethiTHORN DWcleth1THORN
kT
eth23THORN
where DWclethiTHORN is given by equation (5) This amounts toa uniform downward shift of the classical free energybarrier illustrated in Fig 1 At least now the expressionis consistent at i frac14 1 The theory is known as internallyconsistent classical theory (ICCT) [51] Unfortunately itis not clear whether this constant shift in the work offormation should be a realistic representation of all thesubleading terms in the work of formation [52] It is anarbitrary modification and its success in accounting fordata is mixed For example nucleation rates for a rangeof alkanes [114] require further modification by factorsof between one and nine orders of magnitude to comeinto agreement with ICCT
It should be noted at this point that the correctionterm kT lnS which appears in DWcleth1THORN can beregarded as rigorous There are good reasons for thisbased on general considerations involving the nucleationtheorems [56] or more simply the principles of chemicalequilibrium represented by the law of mass action givenin equation (15) The correction converts the prefactorn1 in equation (4) into ns1 and results in the insertion of afactor 1=S in the nucleation rate in equation (1)
The Fisher droplet model is a slightly more sophis-ticated scheme [115] Here the constrained equilibrium
populations are modelled by
ncei frac14 ns1 exp yii2=3 thorn t ln i lnq0kT
psv
thorn i lnS
eth24THORN
The term proportional to yi may be viewed as a surfaceterm while the other two are corrections of a typesuggested by the free energy expressions developed insection 5 The parameters q0 yi and t are chosen toensure that the model correctly reproduces knownproperties of clusters and vapours including virialcoefficients and critical properties the theory is there-fore phenomenological
For large cluster sizes and for temperatures wellbelow the critical temperature the free energy offormation should tend towards that of a macroscopicliquid droplet yi should therefore approach A1s=ethkTTHORNfor large i where A1i
2=3 is the surface area 4pR2 of aspherical droplet containing i monomers at the bulkliquid density so that A1 frac14 eth36pTHORN1=3r2=3
l Dillmann andMeier [116] suggested writing
yi frac14kiA1skT
eth25THORN
with the Ansatz
ki frac14 1thorn a1i 1=3 thorn a2i 2=3 eth26THORN
The coefficients a1 and a2 can be found in terms of virialcoefficients of the vapour The form of this Ansatz wasmotivated by the curvature dependence of the bulksurface tension derived in classical thermodynamics byTolman [117]
The initial success of the DillmannndashMeier theory wasshort lived since modifications had to be made toovercome an internal inconsistency [118] Virial expan-sions for the vapour pressure and density were usedincorrectly to calculate the virial expansion of the ratioof the two Correcting this error [61] changed predictedrates by several orders of magnitude Nevertheless theapproach was very influential [119] In a furtherdevelopment Kalikmanov and van Dongen [120] seta2 frac14 0 (this term can be absorbed into q0 anyway) andchose a1 such that the experimental saturated vapourpressure psv is reproduced by the model The DillmannndashMeier model matched the first two virial coefficientsonly The Kalikmanovndashvan Dongen theory shouldtherefore approximate the (revised) DillmannndashMeiertheory at low temperatures but ought to provide abetter description at higher temperatures
Other phenomenological models have been proposedincluding the Hale scaling relation [121] This modelbased on CNT but employing simplified models for thesurface tension and liquid density based on scalingbehaviour near the critical point together with a rather
STATISTICAL MECHANICS OF NUCLEATION 893
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
more straightforward rate prefactor seems to be quitesuccessful in accounting for data for a wide range ofmaterials even metals [122] and two-component systems[123] Another phenomenological model is the diffuseinterface theory [124] which is based on a treatment ofthe phase interface using simplified density profiles ofthermodynamic properties Another general class ofmodel makes use of conditions at the vapourndashliquidspinodal to constrain the form of the model free energy[125ndash128]
The main drawback in phenomenological models isthe arbitrariness in the form of the proposed work offormation The parameter fitting in practice ensures thatthe populations of small clusters especially the mono-mer and dimer are well represented by the model butthe properties of larger clusters depend on the expres-sion assumed for the model and are therefore uncertainIntuition suggests that phenomenological modelling islikely to have mixed success and that microscopiccalculations of cluster properties do matter There mightbe better value in phenomenological models however ifthey were based securely on the results of microscopiccalculations
8 KINETICS OR THERMODYNAMICS
In section 3 kinetic equations representing the nuclea-tion process were described However the coefficients inthose equations were related to constrained equilibriumpopulations and hence to the equilibrium statisticalmechanics of clusters through the detailed balancecondition (8) Is it necessary to do this Could acompletely kinetic route be followed by evaluating themean growth and decay coefficients bi and gi directly bysome statistical approach The main problem is ofcourse the difficulty in performing such calculationsand this has restricted the development of such models[129ndash131] It is expected that the kinetic coefficients willcorrespond to those derived from the detailed balanceargument for near-equilibrium conditions but circum-stances can be imagined where non-equilibrium effectsare present which cannot be dealt with within standardstatistical mechanics
The framework for evaluating kinetic coefficients ofcluster decay can be briefly introduced [86 132] Amolecular dynamics simulation of a cluster can be set upand evolved according to the interactions until bychance an evaporation event occurs By extensivesimulation statistics for the mean evaporation rates gican be amassed in this way Similarly moleculardynamics simulations where a single molecule iscaptured by a cluster could be performed and thegrowth rates bi determined These will be functions ofcluster energy and so would fit best into a rate equationscheme involving both size and energy [133]
This is a laborious approach but statistical physicsoffers an approximate alternative representation of theescape and capture processes The basic tool is theLangevin equation The radial motion of a molecule inthe cluster is described by [134]
meurorr frac14 f ethrTHORN thornfethr _rr tTHORN eth27THORN
where r is the radial position m is the molecular mass fis the mean (time- and velocity-averaged) force on themolecule at position r and f is the stochastic forceThe latter is usually represented as a velocity-dependent friction term mg _rr where g is the so-calledfriction coefficient plus a fluctuating term ~ff ethr tTHORN withzero mean and correlation in time h ~ff ethr tTHORN~ff ethr t0THORNi frac14eth2gkTmTHORNdetht t0THORN where k is Boltzmannrsquos constant andwhere T has the characteristics of a temperature as willbe seen It is a standard manipulation [135] to convertthe Langevin description in the large friction limit intoa FokkerndashPlanck equation
qwqt
frac14 1
mg qethfwTHORN
qrthorn kT
q2wqr2
eth28THORN
which represents the evolution of wethr tTHORN the prob-ability density that the molecule should lie at the radialposition r It should be noted in passing that FokkerndashPlanck equations have often been used to describe thestochastic dynamics of a cluster in size space [48 136] Incontrast the use of the method described above refers tomotion of a molecule in real coordinate space relative tothe centre of mass of the cluster
The steady state equilibrium solution of equation (28)is
wethrTHORN exp FethrTHORNkT
eth29THORN
where FethrTHORN is called the potential of mean force relatedto the mean force on the molecule f throughf frac14 dF=dr It can now be seen how the parameter Tin the stochastic force plays the role of temperaturesince equation (29) looks like a Boltzmann distributionThis solution is not what is sought however The escapeproblem is characterized by a boundary conditionwethreTHORN frac14 0 where re is the radius at which a particleescapes from the system The resulting escape rate maybe shown to be [134]
gkin frac14 KethreT m gTHORN exp DFkT
eth30THORN
where the prefactor K depends on the parameters listedas well as the shape of the potential of mean force Theprincipal feature of equation (30) is the exponentialdependence on the depth DF of the potential of meanforce Therefore if the potential of mean force FethrTHORN can
I J FORD894
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
be evaluated by studying a molecular dynamics trajec-tory for example then the kinetic decay rate can bedetermined This could be extended to account for aradially dependent temperature and friction coefficientboth of which might be determined from simulation Asketch of a potential of mean force is given in Fig 6
Kinetically determined cluster growth and decayrates gkini and bkini respectively could provide a newroute for the determination of nucleation rates inde-pendent of assumptions about detailed balance and theuse of equilibrium statistical mechanics in a non-equilibrium situation
9 CONCLUSIONS
This review has been brief and in some ways partisan tothe personal views of the author It attempts to sketch acoherent picture of nucleation theory and how thevarious disparate strands relate to one another One ofthe fascinations of nucleation theory is the variety ofapproaches and the number of difficult issues that areencountered How well does free energy fluctuationtheory represent a real non-equilibrium process Howwell does the idea of detailed balance provide aframework for determining non-equilibrium decaycoefficients from equilibrium statistical mechanicsHow well do various methods perform in the evaluation
of cluster free energies Can cluster growth and decayrates be calculated directly without the need for such aframework How exactly are clusters to be defined
The theoretical problems are many but the issues canbe resolved to various degrees of satisfaction Thereremains of course the fundamental task of choosing anintermolecular potential that adequately represents thephysical interactions Without this all the finer points ofcluster definition and treatments of non-equilibriumprocesses are unimportant Modelling many problems inscience relies on the availability of an accurate micro-scopic description and so nucleation rate calculationsare not unique in this respect Most empirical molecularpotentials are fitted to the properties of bulk and surfaceproperties and in the very best examples attention isgiven to their transferability to more unusual environ-ments such as clusters In view of these capabilities andmodern theoretical developments in nucleation theoryit seems likely that successful microscopic models of thenucleation process based on statistical physics are atlast coming within reach
ACKNOWLEDGEMENTS
The present author is grateful to colleagues SarahHarris and Marco Arosio for providing some of thefigures
Fig 6 A molecule in a cluster may be considered to be confined within it by a potential of mean force FethrTHORN Itcan escape by reaching a radial position re driven by a phenomenological stochastic force
STATISTICAL MECHANICS OF NUCLEATION 895
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
REFERENCES
1 Gross D H E Microcanonical Thermodynamics 2001
(World Scientific Singapore)
2 Seinfeld J H and Pandis S N Atmospheric Chemistry
and Physics 1998 (John Wiley New York)
3 Climate Change 2001 Synthesis Report 2001 (Inter-
governmental Panel on Climate Change) (Cambridge
University Press Cambridge)
4 Kohler T M Gail H-P and Sedlmayr E MgO dust
nucleation in M-stars calculation of cluster properties
and nucleation rates Astronomy and Astrophysics 1997
320 553
5 Kashchiev D Nucleation Basic Theory with Applications
2000 (ButterworthndashHeinemann Oxford)
6 Bakhtar F Ebrahimi M and Webb R A On the
performance of a cascade of turbine rotor tip section
blading in nucleating steam Part 1 surface pressure
distributions Proc Instn Mech Engrs 1995 209(C2)
115ndash124
7 Bakhtar F Ebrahimi M and Bamkole B O On the
performance of a cascade of turbine rotor tip section
blading in nucleating steam Part 2 wake traverses Proc
Instn Mech Engrs 1995 209(C3) 169ndash178
8 Bakhtar F andMahpeykar M R On the performance of
a cascade of turbine rotor tip section blading in nucleating
steam Part 3 theoretical treatment Proc Instn Mech
Engrs Part C J Mechanical Engineering Science 1997
211(C3) 195ndash210
9 Volmer M and Weber A Z Physikalische Chemie 1926
119 277
10 Farkas L Z Physikalische Chemie (Leipzig) 1927 125
236
11 Becker R and Doring W Kinetische Behandlung der
Keinbildung in ubersattigten Dampfen Ann Physik
(Leipzig) 1935 24 719
12 Zeldovich J Soviet Physics JETP 1942 12 525
13 Frenkel J Kinetic Theory of Liquids 1946 (Oxford
University Press Oxford)
14 Wagner P E and Strey R Homogeneous nucleation
rates of water vapor measured in a two-piston expansion
chamber J Phys Chemistry 1981 85 2694
15 Miller R Anderson R J Kassner J L and Hagen
D E Homogeneous nucleation rate measurements for
water over a wide range of temperature and nucleation
rate J Chem Physics 1983 78 3204
16 Viisanen Y Strey R and Reiss H Homogeneous
nucleation rates for water J Chem Physics 1993 99 4680
17 Viisanen Y Strey R and Reiss H Erratum homo-
geneous nucleation rates for water J Chem Physics
2000 112 8205
18 Wolk J and Strey R Homogeneous nucleation of H2O
and D2O in comparison the isotope effect J Phys
Chemistry B 2001 105 11683
19 Heath C H Streletsky K and Wyslouzil B E H2Ondash
D2O condensation in a supersonic nozzle J Chem
Physics 2002 117 6176
20 Mikheev V B Irving P M Laulainen N S Barlow
S E and Pervukhin V V Laboratory measurements of
water nucleation using a laminar flow chamber J Chem
Physics 2002 116 10 772
21 Hung C H Krasnopoler M J and Katz J L
Condensation of a supersaturated vapor VIII The
homogeneous nucleation of n-nonane J Chem Physics
1989 90 1856
22 Rudek M M Fisk J A Chakarov V M and Katz
J L Condensation of a supersaturated vapor XII The
homogeneous nucleation of the n-alkanes J Chem
Physics 1996 105 4707
23 Strey R Wagner P E and Schmeling T Homogeneous
nucleation rates for n-alcohol vapors measured in a two-
piston expansion chamber J Chem Physics 1986 84
2325
24 Lihavainen H Viisanen Y and Kulmala M Homo-
geneous nucleation of n-pentanol in a laminar flow
diffusion chamber J Chem Physics 2001 114 10 031
25 Rudek M M Katz J L Vidensky I V Zdımal V and
Smolık J Homogeneous nucleation rates of n-pentanol
measured in an upward thermal diffusion cloud chamber
J Chem Physics 1999 111 3623
26 Grassmann A and Peters F Homogeneous nucleation
rates of n-propanol in nitrogen measured in a piston-
expansion tube J Chem Physics 2002 116 7617
27 Kane D and El-Shall S Condensation of supersaturated
vapors of hydrogen bonding molecules ethylene glycol
propylene glycol trimethyl glycol and glycerol J Chem
Physics 1996 105 7617
28 Hameri K and Kulmala M Homogeneous nucleation in
a laminar flow chamber the effect of temperature and
carrier gas pressure on dibutyl phthalate vapor nucleation
rate at high supersaturations J Chem Physics 1996 105
7696
29 Viisanen Y Strey R Laaksonen A and Kulmala M
Measurement of the molecular content of binary nuclei
II Use of the nucleation rate surface for waterndashethanol
J Chem Physics 1994 100 6062
30 Viisanen Y Wagner P E and Strey R Measurement of
the molecular content of binary nuclei IV Use of the
nucleation rate surfaces for the n-nonane-n-alcohol series
J Chem Physics 1998 108 4257
31 Fisk J A Rudek M M Katz J L Beiersdorf D and
Uchtmann H The homogeneous nucleation of cesium
vapor Atmos Res 1998 46 211
32 Uchtmann H Dettmer R Baranovskii S D and
Hensel F Photoinduced nucleation in supersaturated
mercury vapor J Chem Physics 1998 108 9775
33 Ferguson F T and Nuth J A Experimental studies of
the vapor phase nucleation of refractory compounds V
The condensation of lithium J Chem Physics 2000 113
4093
34 Laaksonen A Talanquer V and Oxtoby D W
Nucleation measurements theory and atmospheric
applications A Rev Phys Chemistry 1995 46 489
35 Debenedetti P G Metastable Liquids Concepts and
Principles 1996 (Princeton University Press Princeton
New Jersey)
36 Yasuoka K and Matsumoto M Molecular dynamics of
homogeneous nucleation in the vapor phase I Lennardndash
Jones fluid J Chem Physics 1998 109 8451
37 Yasuoka K and Matsumoto M Molecular dynamics of
homogeneous nucleation in the vapor phase II Water
J Chem Physics 1998 109 8463
I J FORD896
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
38 Wonczak S Strey R and Stauffer D Confirmation of
classical nucleation theory by Monte Carlo simulations in
the 3-dimensional Ising model at low temperature
J Chem Physics 2000 113 1976
39 Laasonen K Wonczak S Strey R and Laaksonen A
Molecular dynamics simulations of gasndashliquid nucleation
of LennardndashJones fluid J Chem Physics 2000 113 9741
40 Reiss H Methods of Thermodynamics 1996 (Dover
Publications New York)
41 Kondepudi D and Prigogine I Modern Thermodynamics
From Heat Engines to Dissipative Structures 1998 (John
Wiley Chichester West Sussex)
42 Smoluchowski M Ann Phys NY 1908 25 225
43 Einstein A Ann Phys NY 1910 338 1275
44 Hirschfelder J O Curtis C F and Bird R BMolecular
Theory of Gases and Liquids 1954 (John Wiley New
York)
45 Landau L D and Lifshitz E M Statistical Physics 1958
(Pergamon London)
46 Wu D T Nucleation theory Solid State Physicsmdash
Advances in Res and Applic 1997 50 37
47 Reguera D Rubı J M and Perez-Madrid A Fokkerndash
Planck equations for nucleation processes revisited
Physica A 1998 259 10
48 Schenter G K Kathmann S and Garrett B Variational
transition state theory of vapor phase nucleation
J Chem Physics 1999 109 7951
49 Onsager L Phys Rev 1931 37 405
50 Courtney W G J Chem Physics 1961 35 2249
51 Girshick S L and Chiu C-P Kinetic nucleation theory
a new expression for the rate of homogeneous nucleation
from an ideal supersaturated vapor J Chem Physics
1990 93 1273
52 Ford I J Barrett J C and Lazaridis M Uncertainties
in cluster energies in homogeneous nucleation theory
J Aerosol Sci 1993 24 581
53 Weakliem C L and Reiss H The factor 1S in the
classical theory of nucleation J Phys Chemistry 1994
98 6408
54 McClurg R B and Flagan R C Critical comparison of
droplet models in homogeneous nucleation theory
J Colloid Interface Sci 1998 201 194
55 Mashmoushy H Mahpeykar M R and Bakhtar F
Studies of nucleating and wet steam flows in two-dimen-
sional cascades Proc Instn Mech Engrs Part C J
Mechanical Engineering Science 2004 218(C8) 843ndash858
56 Ford I J Nucleation theorems the statistical mechanics
of molecular clusters and a revision of classical nuclea-
tion theory Phys Rev E 1997 56 5615
57 ten Wolde P R and Frenkel D Computer simulation
study of gasndashliquid nucleation in a LennardndashJones system
J Chem Physics 1998 109 9901
58 Kusaka I Wang Z G and Seinfeld J H Direct
evaluation of the equilibrium distribution of physical
clusters by a grand canonical Monte Carlo simulation
J Chem Physics 1998 108 3416
59 Kusaka I and Oxtoby D W Identifying physical
clusters in vapor phase nucleation J Chem Physics
1999 110 5249
60 Stillinger F H J Chem Physics 1963 38 1486
61 Ford I J Imperfect vaporndashgas mixtures and homo-
geneous nucleation J Aerosol Sci 1992 23 447
62 Lothe J and Pound G M J Chem Physics 1962 36
2080
63 Lee J K Barker J A and Abraham F F Theory and
Monte Carlo simulation of physical clusters in the
imperfect vapor J Chem Physics 1973 58 3166
64 Abraham F F Homogeneous Nucleation Theory 1974
(Academic Press New York)
65 Reif F Fundamentals of Statistical and Thermal Physics
1965 (McGraw-Hill New York)
66 Hoare M R and Pal P Adv Phys 1975 24 645
67 Hoare M R Adv Chem Physics 1979 40 49
68 Pal P and Hoare M R Thermodynamic properties
and homogeneous nucleation of molecular clusters of
nitrogen J Phys Chemistry 1987 91 2474
69 Reiss H Katz J L and Cohen E R Translationndash
rotation paradox in the theory of nucleation J Chem
Physics 1968 48 5553
70 Hale B N Monte Carlo calculations of effective surface
tension for small clusters Aust J Physics 1996 49 425
71 Kathmann S M Schenter G K and Garrett B C
Dynamical nucleation theory calculation of condensation
rate constants for small water clusters J Chem Physics
1999 111 4688
72 Reiss H Tabazadeh A and Talbot J Molecular theory
of vapor phase nucleation the physically consistent
cluster J Chem Physics 1990 92 1266
73 Ellerby H M Weakliem C L and Reiss H Toward a
molecular theory of vaporndashphase nucleation I Identifica-
tion of the average embryo J Chem Physics 1992 97
5766
74 Ellerby H M and Reiss H Toward a molecular theory
of vaporndashphase nucleation II Fundamental treatment of
the cluster distribution J Chem Physics 1992 97 5766
75 Weakliem C L and Reiss H Toward a molecular theory
of vaporndashphase nucleation III Thermodynamic properties
of argon clusters fromMonte Carlo simulations and modi-
fied liquid drop theory J Chem Physics 1994 101 2398
76 Kusaka I Wang Z G and Seinfeld J H Binary
nucleation of sulfuric acidndashwater Monte Carlo simula-
tion J Chem Physics 1998 108 6829
77 Oh K J and Zeng X C A small system ensemble
Monte Carlo simulation of supersaturated vapor
evaluation of barrier to nucleation J Chem Physics
2000 112 294
78 Senger B Schaaf P Corti D S Bowles R Voegel J
C and Reiss H A molecular theory of the homogeneous
nucleation rate I Formulation and fundamental issues
J Chem Physics 1999 110 6421
79 Hill T L Statistical Mechanics Principles and Selected
Applications 1956 (McGraw-Hill New York)
80 Soto R and Cordero P Nonideal gas of clusters at
equilibrium J Chem Physics 1999 110 7316
81 Barrett J C The significance of cluster lifetime in
nucleation theory J Chem Physics 2002 116 8856
82 Lushnikov A A and Kulmala M Dimers in nucleating
vapors Phys Rev E 1998 58 3157
83 Bahadur R and McClurg R B Comparison of cluster
definitions in homogeneous nucleation rate calculations
J Phys Chemistry 2001 105 11 893
84 Pugnaloni L A and Vericat F New criteria for cluster
identification in continuum systems J Chem Physics
2002 116 1097
STATISTICAL MECHANICS OF NUCLEATION 897
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
85 Chen B Siepmann J I Oh K J and Klein M L
Simulating vaporndashliquid nucleation of n-alkanes J Chem
Physics 2002 116 4317
86 Harris S A and Ford I J A dynamical definition of
quasi-bound molecular clusters J Chem Physics 2003
118 9216
87 Shreve A P Walton J R P B and Gubbins K E
Liquid drops of polar molecules J Chem Physics 1986
85 2178
88 ten Wolde P R Ruiz-Montero M J and Frenkel D
Numerical calculation of the rate of homogeneous gasndash
liquid nucleation in a LennardndashJones system J Chem
Physics 1998 109 9901
89 ten Wolde P R Oxtoby D W and Frenkel D Chain
formation in homogeneous gasndashliquid nucleation of polar
fluids J Chem Physics 1998 111 4762
90 Vehkamaki H and Ford I J Critical cluster size and
droplet nucleation rate from growth and decay simula-
tions of LennardndashJones clusters J Chem Physics 2000
112 4193
91 Schaaf P Senger B Voegel J C Bowles R K and
Reiss H Simulative determination of kinetic coefficients
for nucleation rates J Chem Physics 2001 114 8091
92 Chen B Siepmann J I Oh K J and Klein M L
Aggregation-volume-bias Monte Carlo simulations of
vaporndashliquid nucleation barriers for LennardndashJonesium
J Chem Physics 2001 115 10 903
93 Hale B N and DiMattio D J A Monte Carlo discrete
sum (MCDS) nucleation rate model for water In
Proceedings of the 15th International Conference on
Nucleation and Atmospheric Aerosols AIP Conference
Proceedings Vol 534 (Eds B N Hale and M Kulmala)
2000 pp 31ndash34
94 Kathmann S M and Hale B N Monte Carlo
simulations of small sulfuric acidndashwater clusters
J Phys Chemistry B 2001 105 11 719
95 Daura X Mark A E and van Gunsteren W F Para-
metrization of aliphatic CHn united atoms of GROMOS96
force field J Comput Chemistry 1998 19 535
96 Kathmann S M Schenter G K and Garrett B C
Understanding the sensitivity of nucleation kinetics a
case study on water J Chem Physics 2002 116 5046
97 Kashchiev D On the relation between nucleation work
nucleus size and nucleation rate J Chem Physics 1982
76 5098
98 Oxtoby D W and Kashchiev D A general relation
between the nucleation work and the size of the nucleus in
multicomponent nucleation J Chem Physics 1994 100
7665
99 Ford I J Thermodynamic properties of critical clusters
from measurements of vapourndashliquid homogeneous
nucleation rates J Chem Physics 1996 105 8324
100 Oxtoby D W and Evans R Nonclassical nucleation
theory for the gasndashliquid transition J Chem Physics
1988 89 7521
101 Zeng X C and Oxtoby D W Gasndashliquid nucleation in
LennardndashJones fluids J Chem Physics 1991 94 4472
102 Evans R Fundamentals of Inhomogeneous Fluids 1992
Ch 3 pp 85ndash175 (Marcel Dekker New York)
103 Talanquer V and Oxtoby D W Dynamical density
functional theory of gasndashliquid nucleation J Chem
Physics 1994 100 5190
104 Barrett J C First-order correction to classical nucleation
theory a density functional approach J Chem Physics
1999 111 5938
105 Barrett J C Cluster translation and growth in density
functional theories of homogeneous nucleation J Chem
Physics 1997 107 7989
106 Bykov T V and Zeng X C A patching model for
surface tension of spherical droplet and Tolman length J
Chem Physics 1999 111 10 602
107 Talanquer V and Oxtoby D W Nucleation in molecular
and dipolar fluids interaction site model J Chem
Physics 1995 103 3686
108 Seok C and Oxtoby D W Nucleation in n-alkanes a
density functional approach J Chem Physics 1998 109
7982
109 Reguera D Bowles R K Djikaev Y and Reiss H
Phase transitions in systems small enough to be clusters
J Chem Physics 2003 118 340
110 Khakshouri S and Ford I J J Chem Physics 2004
(submitted)
111 Ford I J (unpublished)
112 Nyquist R M Talanquer V and Oxtoby D W Density
functional theory of nucleation a semi-empirical
approach J Chem Physics 1995 103 1175
113 Talanquer V and Oxtoby D W Density functional
analysis of phenomenological theories of gasndashliquid
nucleation J Chem Physics 1995 99 2865
114 Katz J L Fisk J A and Chakarov V The accuracy of
nucleation theory In Proceedings of the 13th Interna-
tional Conference on Nucleation and Atmospheric Aerosols
(Eds N Fukuta and P E Wagner) 1992 pp 11ndash18
(Deepak Publishing Hampton Virginia)
115 Fisher M E The theory of condensation and the critical
point Physics 1967 3 255
116 Dillmann A and Meier G E A A refined droplet
approach to the problem of homogeneous nucleation
from the vapor phase J Chem Physics 1991 94 3872
117 Tolman R C J Chem Physics 1949 17 333
118 Ford I J Laaksonen A and Kulmala M Modification
of the DillmannndashMeier theory of homogeneous nuclea-
tion J Chem Physics 1993 99 764
119 Laaksonen A Ford I J and Kulmala M Revised
parametrization of the DillmannndashMeier theory of homo-
geneous nucleation Phys Rev E 1994 49 5517
120 Kalikmanov V I and van Dongen M E H Semipheno-
menological theory of homogeneous vaporndashliquid nuclea-
tion J Chem Physics 1995 103 4250
121 Hale B N Application of a scaled homogeneous
nucleation-rate formalism to experimental-data at
T 5Tc Phys Rev A 1986 33 4156
122 Martinez D M Ferguson F T Heist R H and Nuth
J A Application of scaled nucleation theory to metallic
vapor condensation J Chem Physics 2001 115 310
123 Hale B N and Wilemski G A scaled nucleation model
for ideal binary systems Chem Physics Lett 1999 305
263
124 Granasy L Diffuse interface theory for homogeneous
vapor condensation J Chem Physics 1996 104
5188
125 McGraw R and Laaksonen A Interfacial curvature free
energy the Kelvin relation and vaporndashliquid nucleation
rate J Chem Physics 1997 106 5284
I J FORD898
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
126 Talanquer V A new phenomenological approach to
gasndashliquid nucleation based on the scaling properties
of the critical nucleus J Chem Physics 1997 106
9957
127 Koga K and Zeng X C Thermodynamic expansion of
nucleation free-energy barrier and size of critical nucleus
near the vaporndashliquid coexistence J Chem Physics 1999
110 3466
128 Kusaka I On the scaling behavior of the free energetics of
nucleation J Chem Physics 2003 118 5510
129 Nowakowski B and Ruckenstein E A kinetic approach
to the theory of nucleation in gases J Chem Physics
1991 94 1397
130 Nowakowski B and Ruckenstein E Homogeneous
nucleation in gases a three-dimensional FokkerndashPlanck
equation for evaporation from clusters J Chem Physics
1991 94 8487
131 Wilcox C F and Bauer S H Estimation of homo-
geneous nucleation flux via a kinetic model J Chem
Physics 1991 94 8302
132 Hanggi P Talkner P and Borkovec M Reaction rate
theory fifty years after Kramers Rev Mod Physics 1990
62 251
133 Barrett J C Clement C F and Ford I J Energy
fluctuations in homogeneous nucleation theory for aero-
sols J Physics A 1993 26 529
134 Ford I J and Harris S A Molecular cluster decay
viewed as escape from a potential of mean force J Chem
Physics 2004 120 4428
135 Risken H The FokkerndashPlanck Equation Methods of
Solution and Applications 2nd edition 1989 (Springer-
Verlag Berlin)
136 Shizgal B and Barrett J C Time dependent nucleation
J Chem Physics 1989 91 6505
STATISTICAL MECHANICS OF NUCLEATION 899
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
g Langevin friction coefficientgkin cluster decay rate in the Langevin modelgi decay rate of an i-clusterDF change in free energyDW work of cluster formationDWcl classical work of cluster formationyi parameter in the DillmannndashMeier modelki parameter in the DillmannndashMeier model chemical potentials chemical potential in saturated vapourX grand canonical partition functionrl liquid phase densityrv vapour phase densityrsv saturated vapour densitys planar surface tensiont parameter in the DillmannndashMeier modelok cluster vibrational angular frequencyom
k molecular vibrational angular frequency
1 INTRODUCTION
One of the most remarkable properties of matter is itsability to take different physical forms for differentvalues of parameters such as temperature or pressureFor example a vapour will condense by the nucleationof droplets when its pressure is raised above thesupersaturated vapour pressure or when its temperatureis reduced below the dew point Such metamorphosesare fundamentally a consequence of the second law ofthermodynamics a principle that can be characterizedas a desire on the part of physical systems to minimizetheir free energy by any available means This caninclude the repositioning of molecules on a grand scalea phase change Phase transformations are the statisticalconsequences of the ability of large physical systemsand any environment with which they interact toexplore a huge range of microscopic configurationsThis dynamical behaviour underpinning the secondlaw remains only partially understood [1] but it seemsto be universal
A practical situation where a change in phase createsproblems is the behaviour of steam in turbines Thetemperature and pressure gradients are such that waterdroplets might nucleate with undesirable effects on theperformance of the machine including the erosion of theturbine blades On the other hand the steam must beexpanded as much as possible to extract the most workfrom it Some of the drawbacks can be avoided if thephase change can be controlled
The need for an understanding of the transition fromdry to wet steam in turbines has motivated this reviewbut this is just an example of a process which affectsmany aspects of the world The formation of droplets in
a turbine presents practical difficulties but withoutatmospheric clouds which form by a similar process [2]the global climate would be rather different from what itis at present [3] Aerosol formation in the atmosphereand dust in space [4] and in the case of condensedphases the freezing of liquids all proceed by the samebasic mechanism The principal idea is that there exists abottleneck in the transformation which is passedthrough only by fragments or molecular clusters ofthe new phase The bottleneck is narrowest when theclusters reach a so-called critical size and therefore theproperties of these critical clusters are central to thetheory of nucleation [5]
Experimental progress in understanding droplet for-mation in turbines has been significant [6 7] but at firstsight it would appear that progress on the theoreticalside has been less impressive The tool used most oftenin modelling studies [8] remains the classical nucleationtheory (CNT) derived many decades ago and based onthe most primitive ideas of the properties of the criticalcluster [9ndash13] The capillarity approximation isemployed whereby the critical cluster however smallis considered as a scaled-down macroscopic droplet ofthe condensed phase The model is a version oftransition state theory (TST) in which a realization ofthe system is identified which is in unstable thermo-dynamic equilibrium equally able to develop with timeinto the initial or the final phase How is it that thismodel is still employed when developments in micro-scopic modelling have proceeded so rapidly in recentyears Why has no better model been developed
One reason is pragmatism It is very surprising thatthe classical theory successfully predicts nucleation rateswithin a couple of orders of magnitude in the case ofwater perhaps the most important substance to under-stand for a temperature range of about 210ndash260K anda range of supersaturations Evidence for this has beencollected through various studies over the last 20 years[14ndash20] This level of agreement is good in the field ofnucleation theory However in these conditions thecritical cluster consists of only a few tens of moleculesand the capillarity approximation is questionable in theextreme
However if a broader view is taken and data for othermaterials are considered the failure of CNT is moreapparent The nucleation rate seems to have a depen-dence on supersaturation suggested by CNT but thetemperature dependence does not agree Often atemperature range where classical theory is reasonablyconsistent with experiment does not exist
Experimental data gathered during recent years canbe reviewed briefly Some early data on n-nonaneshowed deviations from CNT of four orders ofmagnitude either way [21] A study of several membersof the n-alkane series [22] confirmed the need forcorrection factors of again up to four orders ofmagnitude Early data for a range of alcohols showed
I J FORD884
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
deviations from CNT of over five orders of magnitude[23] In an effort to acquire definitive data for a singlesubstance the nucleation of n-pentanol was givenparticular attention by a number of research groups inthe late 1990s and early 2000s For temperatures in therange 260ndash290K most data lie three orders of magnitudeabove the CNT predictions [24] although some measure-ments lie between one and three orders of magnitudebelow [25] Further results for n-pentanol show goodagreement with the CNT at 273K [26] but withdeviations away from this temperature In the light ofthe reasonable success of CNT in explaining water datait is disappointing to find that hydrogen-bonded organicmaterials such as a series of glycols disagree verystrongly with the temperature dependence of thenucleation rate suggested by CNT [27] Similarly theorganic species dibutyl phthalate deviates by up to sixorders of magnitude from the model [28] The situationbecomes even more unsatisfactory for the nucleation oftwo or more species together Strong deviations frombinary CNT are found for the ethanolndashwater system [29]and for a range of n-nonanendashn-alcohol mixtures [30]However it is with metals that the failure of CNT ismost dramatic examples in the literature being caesium[31] mercury [32] and lithium [33] For caesium thecritical supersaturation (that required to drive nuclea-tion at a specified rate) is between two and ten timeslarger than the CNT predictions The difference betweenmeasured and predicted nucleation rates is astronom-ical Although in most materials the deviation is lesssevere than this it is certainly not the case that a smallcorrection to CNT is all that is needed to account for alldata
The usual CNT expression for the rate is
JCNT frac14 q2spm
1=2 rvrl
pvV
kT
6 exp 16ps3
3ethkTTHORN3r2l ethlnSTHORN2
eth1THORN
where q is a sticking or accommodation coefficient m isthe molecular mass k is Boltzmannrsquos constant s is thesurface tension of a planar interface between thecondensate saturated vapour pv is the vapour pressureand rl and rv are the densities of the bulk condensedphase and the vapour respectively in molecules per unitvolume S is the supersaturation defined by S frac14 pv=p
sv
where psv is the saturated vapour pressure at thetemperature T Note that J in this expression is thenumber of particles formed per second and it isproportional to the system volume V There are goodreasons to insert a factor of 1S into the prefactor of theabove expression but this is the often-quoted formCalculations are easy agreement is reasonable and sowhy should anyone bother to develop the theoryfurther
The problem in employing this formula for water isthat the laboratory data for water are taken at lowtemperatures and need to be extrapolated to hightemperatures for application to turbine conditions Asalready noted water seems to be a lucky case wheretheory and experiment actually agree for a particular setof conditions namely at a temperature of about 240Kand for a range of supersaturations Unfortunately adiscrepancy appears and grows as the temperature israised and at 260K the theory overpredicts the rate byan order of magnitude [18] The ratio of measurednucleation rate to CNT prediction has been fitted to asimple function by Wolk and Strey [18]
JexpethT STHORN frac14 JCNT exp Athorn B
T
eth2THORN
where A frac14 2756 and B frac14 6500K This formula fitsdata in the temperature range 220ndash260K but has alsobeen found to account for data at temperatures sometens of kelvins higher (B Wyslouzil personal com-munication 2003)
Nevertheless caution must be exercised in extrapolat-ing to temperatures around 370K for steam turbinesDevelopments in the theory are certainly needed It is achallenge to understand why CNTworks reasonably wellat 240K forwaterwhat terms are cancellingout andwhyHow canmeasured data be extrapolated with confidence
In this review the determination of the properties ofmolecular clusters using the tools of statistical mech-anics is again considered Similar accounts of the widerpicture have been given in various reviews and books[34 35] Starting from very general arguments attemptsare made to discern how a theory such as the CNTmight emerge The subtleties of defining a clusterprecisely in terms of the positions and momenta of itsconstituents are discussed It is noted that the transitionstate theory based on equilibrium cluster properties isequivalent to an approach based on the kinetic theory ofcluster decay at least in certain circumstances Somephenomenological models developed from CNT arereviewed and finally a perspective of future develop-ments in this field is given
By the very nature of a review the discussion is ratherbroad while at the same time some of the technicalitiesare outlined The intention is to give a reasonablycomplete picture of the current state of nucleationtheory at least from the point of view of the presentauthor and to show that significant progress has beenmade in the approximately 80 years of its history whilemore is still clearly necessary
2 TRANSITION STATE THEORY
A theory of nucleation based on statistical ideas isnecessary since direct simulation of the process from a
STATISTICAL MECHANICS OF NUCLEATION 885
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
microscopic point of view is very laborious and has beenrealized in only a very few example cases [36ndash39] Thetime scales involved make a nucleation event simply tooslow to follow directly for realistic conditions and withcurrent computational resources
The simplest statistical treatment of the nucleationprocess comes from TST within a framework of thetheory of free energy fluctuations According to thermo-dynamics a system with externally fixed temperatureand pressure (or volume) will adjust any internal prop-erties in order to minimize its Gibbs (or Helmholtz) freeenergy [40 41] Such adjustments will include changes inphase Nevertheless statistical fluctuations in systemproperties will naturally take place and it is useful tointerpret these as fluctuations in the free energy of thesystem away from the minimum Considering constant-volume conditions for the moment general arguments[42ndash45] suggest that the probability of a fluctuationDF in Helmholtz free energy is proportional toexpfrac12DF=ethkTTHORN
The physical fluctuations that are important in thenucleation process are those where a molecular clusterof new phase is formed temporarily from the originalphase The change in free energy accompanying theformation of a cluster of i molecules in the new phase iswritten DF frac14 DWethiTHORNFethiTHORN if0 where FethiTHORN is the freeenergy of the i cluster and f0 is the free energy permolecule in the old phase This is also called the lsquoworkof formationrsquo since it corresponds in thermodynamics tothe external work needed to insert an additional clusterreversibly into the system while maintaining the overall
temperature and pressure (or volume) Note that thisinsertion of a new cluster while maintaining suchconstraints is a theoretical operation since it would behard to achieve in practice
For a metastable initial phase the magnitude of thefree-energy fluctuation DF peaks at the so-called lsquocriticalsizersquo i This is illustrated in Fig 1 The perturbed statewith the additional critical cluster is therefore athermodynamic transition state since reversible changesin size of the extra cluster are accompanied byreductions in free energy It is presumed that this meansthat the new cluster has equal probabilities of growingor decaying and that the rate of nucleation is thereforerelated to the rate of formation of these critical clusters
The probability of forming the critical clusteraccording to this picture is proportional toexpfrac12DWethi THORN=ethkTTHORN but the proportionality constant(essentially a time scale) is not yet determined Tointroduce this simple kinetic theory [46] can be used tocalculate the collision rate bi between single molecules(monomers) and an i-cluster modelled as a sphere withbulk condensed phase molecular density rl
bi frac14qpv
eth2pmkTTHORN1=21thorn 1
i
1=2
1thorn i1=3 2
eth36pTHORN1=3
r2=3l
eth3THORN
where m is the molecular mass It is more usual toreplace the expression in square brackets with i2=3 thelimit for large i The simplest TST expression for the
Fig 1 Sketch of the cluster work of formation for a range of (constrained) cluster sizes The maximum in thecurve defines the critical size i and the critical work of formation DWethi THORN
I J FORD886
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
nucleation rate is then J bini where ni is the meannumber of critical clusters in the system
More sophisticated treatments of thermodynamicscan provide the time scale without recourse to the abovekinetic argument which is reasonably valid for gas-to-particle conversion but not for freezing for exampleAttempts based on linearized non-equilibrium thermo-dynamics [47 48] originate from the work of Onsager[49]
The next step is crucial and it is to relate thepopulation ni to the probability of a fluctuation and toan explicit model of the critical work of formation Theassumption in classical nucleation theory is that
ni frac14 n1 expDWclethi THORN
kT
eth4THORN
where DWclethiTHORN is the classical work of formation givenby
DWclethiTHORN frac14 4pR2s ikT lnS eth5THORN
and DWclethi THORN is the maximum of this function Thecluster is envisaged to be a sphere of radius R given by4pR3rl=3 frac14 i
Unfortunately some uncertainty accompanies thesesteps in the derivation The theory is flawed on severalcounts The work of formation contains a termassociated with the creation of a spherical interface(characterized by the bulk value of surface tension) anda change kT lnS in free energy per molecule due tothe change in phase from supersaturated vapour tocondensate Moreover the factor n1 appearing inequation (4) appears to have been introduced heuristi-cally Might it not instead be the monomer populationin the saturated vapour How does DWclethiTHORN comparewith the actual work of formation DWethiTHORN Is thedroplet really spherical with a sharp interface Have
any important factors been left out Furthermoreshould there be concern that equation (4) is in-consistent when i frac14 1 This has been a controversialissue [50ndash54]
3 KINETIC INTERPRETATION
The expression for the classical rate of nucleation givenin equation (1) emerges from the models in equation (4)and (5) when the thermodynamic transition stateanalysis is embellished by a kinetic interpretation [11]Instead of using the probability of fluctuations awayfrom the equilibrium state (the TST route) a set of rateequations is introduced to describe the dynamics ofcluster populations
The evolution of cluster populations is modelled usingthe following equations
dni
dtfrac14Xj
njPji niXj
Pij eth6THORN
where Pji is the mean rate at which a j cluster convertsinto an i cluster The transition processes are assumed tobe Markovian so that the rate coefficients depend onthe properties of the cluster and its environment but noton their history
The kinetic interpretation can be taken a step furtherby assuming that the only important transitions arethose involving the addition or loss of single moleculesto or from the cluster The scheme of transitionprocesses is illustrated in Fig 2 The only non-zerorate coefficients are then bi frac14 Piithorn1 for cluster growthand gi frac14 Pii1 for cluster decay The rate equations
Fig 2 The scheme of the birth and death of clusters addition of a monomer to a cluster will take it up thechain towards the right while the loss of a monomer will take it in the opposite direction Attachmentand detachment of dimers and larger clusters are usually ignored
STATISTICAL MECHANICS OF NUCLEATION 887
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
reduce to
dni
dtfrac14 bi1ni1 gini bini thorn githorn1nithorn1 eth7THORN
These are the BeckerndashDoring equations [11] and theyprovide a useful reinterpretation of the TST approachThere are two steady state solutions In the first thepopulations satisfy equation (7) and also the detailedbalance condition
bi1nei1 frac14 gin
ei eth8THORN
and the fnei g are taken to represent the cluster popula-tions in an equilibrium vapour at or below the saturatedvapour pressure For true thermal equilibrium themonomer population has to be less than or equal to itsvalue for the saturated vapour ns1 The cluster populationsin a saturated vapour are denoted fnsig Nevertheless asituation can be invented which is known as lsquoconstrainedequilibriumrsquo where the detailed balance condition is stillsatisfied but with n1 gt ns1 Since bi pvS while gi isindependent of vapour pressure equation (8) then definesso-called constrained equilibrium populations
ncei ampSinsi eth9THORN
for a supersaturated vapour It has been assumed thatS frac14 pv=p
svampn1=n
s1
The second steady state solution to the BeckerndashDoring equations is more realistic It is not meant torepresent thermal equilibrium The supersaturation isset to a value greater than unity and a boundarycondition of zero population at some cluster size imax isimposed These conditions will drive a steady productionrate of large clusters from the vapour The approach isfor mathematical convenience only and the same steadystate cluster populations nssi and nucleation rate wouldemerge if imax was sent off to infinity The detailedbalance condition (8) is replaced by J frac14 bi1n
ssi1 gin
ssi
defining the nucleation current J which can then becalculated analytically as shown in another paper in thisvolume [55] To a good approximation
J frac14 Zdbincei eth10THORN
which involves the constrained equilibrium population atthe critical size i the size with the lowest constrainedequilibrium population The condition imax 4 i musthold Zd is the so-called Zeldovich factor [12] givenapproximately by
Zd frac14 1
2pq2ethln ncei THORN
qi2
ifrac14i
1=2
eth11THORN
When equations (4) and (5) are used to specify ncei and theresult is substituted into equations (10) and (11) then theclassical nucleation rate given in equation (1) isrecovered as long as the large i limit of the attachment
rate coefficient in equation (3) is taken It is also possibleto derive an expression for J purely in terms of the ratecoefficients bi and gi This possibility is considered againin section 8
The main assumptions used in this derivation of theclassical theory principally the validity of equations (4)and (5) are now examined In order to explore this it isnecessary to see how the equilibrium cluster populationsnei in subsaturated vapours may be calculated usingstatistical mechanics
4 CLUSTER STATISTICAL MECHANICS
The statistical mechanical treatment of a vapour is bestdeveloped using a grand canonical ensemble [56ndash59] Thesystem is placed in contact with a particle reservoir at achemical potential and temperature T so that thenumber of molecules N in the system can change Thegrand partition function of the system X is then propor-tional to the integral of expf frac12HethNTHORN N=ethkTTHORNg overall possible configurations of molecular positions andmomenta whereHethNTHORN is the Hamiltonian of the system
Evaluating X exactly is exceedingly difficult and it ismuch simpler to proceed by considering the configura-tions of the system as arrangements of molecularclusters The vapour then is viewed as a mixture ofideal gases each composed of clusters of differentsizes The grand partition function of the system canbe reconstructed from partition functions for singlei-clusters In order to do this however a cluster must bedefined in a unique way in terms of molecular positionsand velocities An example of such a definition would beto require the separation between molecules in a clusterto be less than a maximum distance Rc [60] as illustratedin Fig 3
Fig 3 According to the Stillinger definition a molecule isconsidered part of a cluster if it lies closer than adistance Rc from a constituent molecule of the cluster
I J FORD888
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
At first sight the cluster definition is completelyarbitrary a means of categorizing and resumming allthe elementary configurations of the system Howeveraccurately representing the system as a mixture of idealgases requires the interactions between molecules lyingin different clusters to be minimized The clusterdefinition affects whether a particular interaction isregarded as intercluster or intracluster and so it plays acrucial role This will be discussed further in section 6Residual clusterndashcluster interactions can be taken intoaccount in the form of a virial series [61]
These ideas are now put into practice Define amodified canonical partition function zi for a clustercontaining i molecules This is simply the integral ofexpf frac12HcethiTHORN i=ethkTTHORNg over configurations involvingmolecules in the cluster alone with Hc the Hamiltonianof the i molecules comprising the cluster Using thepicture of a mixture of ideal gases the grand partitionfunction of the complete system can then be expressed as
XampXfnig
Yifrac141
znii
ni
eth12THORN
where the sum is over all possible cluster populationdistributions fnig This is demonstrated rigorously inappendix A of reference [57]
The next step is to determine the distribution fnei gthat dominates the sum in equation (12) The logarithmof the summand is extremized
qqni
Xifrac141
lnz
nii
ni
frac14 0 eth13THORN
which gives the following very simple expression for themost probable or equilibrium cluster size distributionfor the given conditions T and
nei frac14 ziethTTHORN frac14 ZethTTHORN ei=ethkTTHORN eth14THORN
which is equivalent to the so-called law of mass action
nei frac14 Zine1Z1
i
eth15THORN
where ZiethTTHORN frac14 expfrac12FethiTHORN=ethkTTHORN frac14 zi expfrac12 i=ethkTTHORN isthe canonical partition function for an i cluster and FethiTHORNis the i-cluster Helmholtz free energy This relationshipis often used as the starting point in models of clusterpopulations
The vapour pressure is now given by a sum of partialpressures
pvV frac14 kTXifrac141
nei frac14 kTXifrac141
ziethTTHORN eth16THORN
and this series will only converge if the equilibriumpopulations or the zi decrease sufficiently quickly forlarge i This limits the statistical mechanics approach to
the study of saturated or subsaturated vapours How-ever as in the preceding section the equilibriumpopulations nei in equation (14) (for S lt 1) can beextrapolated into constrained equilibrium populationsncei (with S gt 1) for use in equation (10) Saturatedvapour corresponds to a particular chemical potentialms and the saturated vapour cluster populations are nsi The constrained equilibrium populations are obtainedby inserting into equation (14) the chemical potential ofthe supersaturated vapour which is frac14 s thorn kT lnSgiving ncei frac14 Sinsi which is consistent with the resultobtained in equation (9) for dilute vapours
The constrained equilibrium populations are there-fore expressed as
ncei frac14 SiziethsTTHORN
frac14 expfrac12FethiTHORN is ikT lnS
kT
eth17THORN
and the correspondence with CNT can be made closerby introducing the n1 prefactor
ncei frac14 n1 exp frac12FethiTHORN is thorn kT ln n1 ikT lnS
kT
eth18THORN
The next task is to explore how the combinationFethiTHORN is thorn kT ln n1 might be approximated by theCNT result 4pR2s This can be explored most easilyfor solid rather than liquid clusters
5 SOLID CLUSTERS
The partition function of a solid cluster of i moleculescan be factorized into translational rotational andvibrational parts through a transformation to suitablerelative positions and momenta [62ndash64] Starting withthe vibrational degrees of freedom the atom positionsare defined with respect to their mean positions whichare determined by a fixed centre of mass and set oflattice vectors It is assumed that the potential energypart of the Hamiltonian is quadratic in the displace-ments from these mean positions Now the partitionfunction is worked out for configurations of the particlesrestricted to having zero total linear and total angularmomentum about the centre of mass This is writtenZvib corresponding to a product of partition functionsfor 3ji 6 independent oscillators (if there are j atomsper molecule and therefore a total of 3ji degrees offreedom if the molecule is not linear) each of which inthe classical limit takes the form kT=ethhoTHORN where o isthe angular frequency of the specific vibrational modeThe partition function has an additional factor ofexpfrac12U0ethiTHORN=ethkTTHORN where U0ethiTHORN is the system potentialenergy evaluated at the mean atom positions A
STATISTICAL MECHANICS OF NUCLEATION 889
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
different expression applies for low temperatures whenthe oscillator is not classically activated [65]
The rotational and translational degrees of freedomare next The previous restriction of the configurationsto zero angular momentum is lifted The approximate(classical) rotational partition function is
Zrot frac14 p1=28p2kTh2
3=2
ethI1I2I3THORN1=2 eth19THORN
where Ik are the principal moments of inertia about thecentre of mass for atoms fixed in their mean positions
Finally the translational free energy of the centre ofmass of the cluster is given by the partition function
Ztrans frac14 V2pMkT
h2
3=2
eth20THORN
where V is the system volume and M frac14 im is the clustermass The free energy of the cluster (plus kT ln n1 forconvenience) is therefore
FethiTHORN thorn kT ln n1
frac14 kT lnZtransZrotZvib
n1
frac14 U0ethiTHORN kT ln
eth2pMkTTHORN3=2
Srsvh3p1=2
8p2kTh2
3=2
6ethI1I2I3THORN1=2Y3ji6
kfrac141
kT
hok
eth21THORN
where rsv frac14 ns1=V is the molecular density of dilutesaturated vapour
By similar arguments the chemical potential of dilutesaturated vapour is
s frac14 Um0 kT ln
eth2pmkTTHORN3=2
rsvh3p1=2
8p2kTh2
3=2
6ethIm1 Im2 Im3 THORN1=2Y3j6
kfrac141
kT
homk
eth22THORN
where Um0 is the potential energy of a molecule when the
atoms are at their mean positions Imk are the threeprincipal moments of inertia of the molecule about thecentre of mass and om
k are the oscillatory frequencies ofmolecular internal vibrations
The classical work of formation DWclethiTHORN should nowcorrespond to the leading-order contribution to thedifference between equation (21) and i times equation(22) Calculations [54 64 66ndash68] demonstrate that theleading term is indeed roughly proportional to i2=3 thusjustifying the form of the classical model The suitabilityof the capillarity approximation depends on the extentto which the coefficient of this leading term correspondsnumerically to the planar surface free energy Howeverthe full statistical mechanical expression for the work of
formation contains additional terms and these will beimportant for small clusters and need to be taken intoaccount In order to evaluate them the variousstructural and energy properties of cluster are requiredall of which can be obtained from microscopic modelsand they can then be inserted into the formulae for thethermodynamic properties [equations (21) and (22)] theconstrained equilibrium populations [equation (18)] andultimately the nucleation rate [equation (10)] Followingthis route employs the almost universally held assump-tion that equilibrium properties such as free energies canbe used to describe non-equilibrium properties specifi-cally the cluster decay rate This assumption is re-examined in section 8
6 LIQUID DROPLETS AND THE CLUSTER
DEFINITION
Defining what was meant by a solid cluster in theprevious section was not an issue Within the assump-tions of the model (harmonic interactions) it wasimpossible for a molecule to escape from the clusterThe cluster definition in the case of liquids cannot bedismissed so easily Liquids are more volatile thansolids and the molecules in a cluster are much moremobile It is imperative now to characterize a clusterusing the molecular positions and velocities andperhaps some time scale and length scale too
The traditional microscopic method for defining acluster is to impose a geometric constraint on molecularpositions An example is to restrict molecules to a spherecentred on the centre of mass [69ndash71] or to impose amaximum allowable separation between the molecules(the Stillinger criterion [60]) More sophisticated treat-ments exist such as the n=v Stillinger cluster introducedby Reiss and co-workers [72ndash75] Here the confiningvolume v used to define the cluster is considered to be anadditional defining characteristic on the same level asthe cluster size n As time progresses a cluster changesits confining volume as the component molecules followtheir respective trajectories just as they change their sizewith the gain and loss of molecules The association of amolecular configuration with a precise confining volumeis complicated but tractable The dependence of the freeenergy on v determines the statistical dynamics of thiscluster volume A similar attempt to associate a clusterwith a specified volume within a grand canonicalensemble description was developed by Kusaka andco-workers [58 59 76] The simulation box was taken tobe the confining cluster volume Many of these ideasmay be viewed as a treatment of the cluster definition(eg the Stillinger radius) as a variational parameter inthe representation of the partition function Further
I J FORD890
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
refinement of this point of view using a treatment ofclusterndashcluster interactions to optimize the definingcluster volume has been presented [77] A review ofrecent progress in this area and the application of theseideas to nucleation theory has been provided by Sengeret al [78]
Such geometric constraints can be implemented buttheir realism is questionable Situations better regardedas close encounters between separate clusters would beclassed as snapshots of single larger clusters It shouldbe recalled that the cluster definition is an internalfeature of the model which is to be optimized to reduceclusterndashcluster interactions and to improve the repre-sentation of the grand partition function Misclassifyingconfigurations is fraught with consequences A secondissue to consider is that for convenience the clustersshould behave in a manner that can be modelled withBeckerndashDoring equations This means that the growthand decay rates should be free of memory effects ie thetransition processes should be Markovian The extent towhich this is satisfied (if indeed it is physically realistic)depends on the cluster definition
An intuitively reasonable choice of definition whichaddresses both these points is that a cluster shouldconsist of molecules that are bound energetically Thisshould minimize the clusterndashcluster contributions to thetotal energy According to such a definition if amolecule interacts strongly with a cluster then it islikely to be regarded as part of that cluster Also anenergetically bound cluster is likely to decay by thermalfluctuation in a stochastic Markovian fashion ratherthan by the separation of passing poorly boundfragments
Ideas have been proposed which require the cluster tobe energetically bound (in a loose sense since a cluster isintrinsically prone to decay) The time evolution of amolecular configuration and not simply the set ofinstantaneous positions of its component molecules isof central importance Hill [79] introduced a clusterdefinition requiring the total energies of pairs ofmolecules in a cluster to be negative (in their centre ofmass frame) The intention was to exclude situationswhere a molecule could leave the cluster within the timeit would take for a molecule to travel a typical lineardimension of the cluster Soto and Cordero [80] havedeveloped Hillrsquos criterion and Barrett [81] has per-formed Monte Carlo modelling of molecular configura-tions excluding those which decay in moleculardynamics within a given period A number of otherenergy-based criteria have been developed [82ndash85]
Harris and Ford [86] have recently proposed a newenergy-based cluster definition As with Barrettrsquosapproach it is based on the requirement that amolecular snapshot is energetically bound for a chosensubsequent period Evaporation according to thisscheme requires not only that a molecule should makea positive energy excursion but also that the subsequent
dynamics carry it far away from the cluster Onlyperhaps one in eight molecules manage to avoidrecapture once having satisfied the positive energycriterion [86] A snapshot of a cluster in this casesurrounded by a carrier gas is shown in Fig 4 Thehistory of the energy of a molecule in such a cluster isillustrated in Fig 5 three excursions into positive energyoccur in this sequence but only one leads to actualescape corresponding to the maintenance of positiveenergy for a significant time The instant of escape istaken to be the point at which the energy goes positivefor the third and final time
Cluster definitions implemented in moleculardynamics and Monte Carlo codes allow cluster freeenergies and hence the decay rate to be calculated Anenormous amount of simulation work has been carriedout and calculations can be laborious Many differentschemes and approximations have been made to easethe evaluation of these properties The principalconclusion from various studies using a range of clusterdefinitions and interaction models [49 57 58 70 85 87ndash92] seems to be that the work of formation of liquidcritical clusters does indeed contain a dominantcontribution proportional to i2=3 consistent with theclassical work of formation Hale and collaboratorspresent cluster free energies for water and other speciesin a way that makes this particularly clear indeed it issurprising how dominant the surface term can be foreven very small clusters [70 93 94] Neverthelessfurther contributions exist small clusters are notcharacterized just by scaled-down properties of largeclusters and it is impossible to escape the need todetermine the additional terms
In all the above discussion it has been assumed thatthe intermolecular interactions are accurately specifiedThis is a situation that is rarely possible Althoughfirst-principles quantum-mechanical models could be
Fig 4 Snapshot of a cluster of 50 argon atoms (brightspheres) bathed in a gas of helium (darker spheres)
STATISTICAL MECHANICS OF NUCLEATION 891
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
developed in practice empirical potentials are devel-oped on the basis of physical intuition containingparameters fitted to the measured properties of thesubstance in question [95] Usually these are bulkproperties and it is not always clear that the sameinteractions can be used for the very different circum-stances of a molecular cluster [96] In view of thesensitivity of nucleation rates to the thermodynamicproperties of clusters this is perhaps a crucial failingProgress can be made however by computing knowncluster properties where they are available and byrefining the potentials to suit Extraction of clusterproperties through use of the so-called nucleationtheorems (which relate derivatives of the nucleationrate to properties of the critical cluster [16 56 97ndash99])may provide just this sort of experimental information
A technique that has not yet been described is the useof density functional theory (DFT) to model theproperties of liquid droplets [100ndash104] This powerfulapplication of statistical mechanics promises to avoidlaborious computations at the microscopic level byinstead focusing on the density profile of moleculesmaking up the droplet This profile may be regarded asthe average distribution of molecules about the clustercentre of mass Although this might appear to involvecoarse-graining and approximation it is a well-foundedapproach and in principle can be exact In practical
implementations however a mean field approach istaken such that the constituent molecules move in aneffective potential well created by the other molecules inthe cluster [105] Also the interactions are assumed to becharacterized by strong repulsive cores and weaklyattractive tails
Considerable analytical work has been carried out[104 106] Various developments have been made todescribe droplets consisting of simple atoms as well aschain-like molecules [107] and polar molecules [108]The main drawback in these implementations is thatspecific interactions such as hydrogen bonds are hardto represent Certain details of the modelled systemare neglected [109 110] and the (implicit) clusterdefinition used is geometric and not energetic [111]On the other hand quite sophisticated systems havebeen modelled including multi-component dropletsThe method has been compared with otherapproaches [112 113] and certain features are betterdescribed particularly the approach to spinodalconditions where the nucleation barrier vanishesDFT calculations can be performed relatively easilyand they can provide a qualitative and perhapsquantitative insight into quite complicated clusteringproblems They provide an intermediate-level descrip-tion between the simple capillarity approximation anda microscopic description
Fig 5 The energy of a molecule in the centre-of-mass frame of a cluster occasionally becomes positive butthis does not necessarily lead to escape Only the third positive energy excursion in this sequenceindicated by an arrow results in particle escape A cluster definition must capture this behaviour
I J FORD892
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
7 PHENOMENOLOGICAL MODELS
The reader might be dismayed by the mathematicalcomplexity of the expressions for the work of formationof solid clusters in section 5 might be alarmed by theprocedural complexity of implementing a cluster defini-tion in section 6 may recoil from the difficulties incalculating free energies from molecular simulationmethods and ultimately may give up in view of theuncertainties in the underlying intermolecular interac-tions Considering these complications it is tempting totry to develop phenomenological models of the work offormation After all if the capillarity approximationworks reasonably well for water then surely a minorcorrection to account for small cluster effects ought tobe sufficient to improve matters
However after perhaps two decades of work nomodel has emerged offering broad success across arange of different substances It is perhaps asking toomuch to expect universal improvement CNT works as afirst approximation for water for a particular tempera-ture range but as discussed earlier it possesses thewrong temperature dependence so that predictionsaway from a temperature in the range 210ndash260K areuncertain A similar picture emerges for other sub-stances although in some cases the failure is even moresevere
The simplest phenomenological adjustment to equa-tion (4) the basis for CNT is to write
ncei frac14 n1 expfrac12DWclethiTHORN DWcleth1THORN
kT
eth23THORN
where DWclethiTHORN is given by equation (5) This amounts toa uniform downward shift of the classical free energybarrier illustrated in Fig 1 At least now the expressionis consistent at i frac14 1 The theory is known as internallyconsistent classical theory (ICCT) [51] Unfortunately itis not clear whether this constant shift in the work offormation should be a realistic representation of all thesubleading terms in the work of formation [52] It is anarbitrary modification and its success in accounting fordata is mixed For example nucleation rates for a rangeof alkanes [114] require further modification by factorsof between one and nine orders of magnitude to comeinto agreement with ICCT
It should be noted at this point that the correctionterm kT lnS which appears in DWcleth1THORN can beregarded as rigorous There are good reasons for thisbased on general considerations involving the nucleationtheorems [56] or more simply the principles of chemicalequilibrium represented by the law of mass action givenin equation (15) The correction converts the prefactorn1 in equation (4) into ns1 and results in the insertion of afactor 1=S in the nucleation rate in equation (1)
The Fisher droplet model is a slightly more sophis-ticated scheme [115] Here the constrained equilibrium
populations are modelled by
ncei frac14 ns1 exp yii2=3 thorn t ln i lnq0kT
psv
thorn i lnS
eth24THORN
The term proportional to yi may be viewed as a surfaceterm while the other two are corrections of a typesuggested by the free energy expressions developed insection 5 The parameters q0 yi and t are chosen toensure that the model correctly reproduces knownproperties of clusters and vapours including virialcoefficients and critical properties the theory is there-fore phenomenological
For large cluster sizes and for temperatures wellbelow the critical temperature the free energy offormation should tend towards that of a macroscopicliquid droplet yi should therefore approach A1s=ethkTTHORNfor large i where A1i
2=3 is the surface area 4pR2 of aspherical droplet containing i monomers at the bulkliquid density so that A1 frac14 eth36pTHORN1=3r2=3
l Dillmann andMeier [116] suggested writing
yi frac14kiA1skT
eth25THORN
with the Ansatz
ki frac14 1thorn a1i 1=3 thorn a2i 2=3 eth26THORN
The coefficients a1 and a2 can be found in terms of virialcoefficients of the vapour The form of this Ansatz wasmotivated by the curvature dependence of the bulksurface tension derived in classical thermodynamics byTolman [117]
The initial success of the DillmannndashMeier theory wasshort lived since modifications had to be made toovercome an internal inconsistency [118] Virial expan-sions for the vapour pressure and density were usedincorrectly to calculate the virial expansion of the ratioof the two Correcting this error [61] changed predictedrates by several orders of magnitude Nevertheless theapproach was very influential [119] In a furtherdevelopment Kalikmanov and van Dongen [120] seta2 frac14 0 (this term can be absorbed into q0 anyway) andchose a1 such that the experimental saturated vapourpressure psv is reproduced by the model The DillmannndashMeier model matched the first two virial coefficientsonly The Kalikmanovndashvan Dongen theory shouldtherefore approximate the (revised) DillmannndashMeiertheory at low temperatures but ought to provide abetter description at higher temperatures
Other phenomenological models have been proposedincluding the Hale scaling relation [121] This modelbased on CNT but employing simplified models for thesurface tension and liquid density based on scalingbehaviour near the critical point together with a rather
STATISTICAL MECHANICS OF NUCLEATION 893
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
more straightforward rate prefactor seems to be quitesuccessful in accounting for data for a wide range ofmaterials even metals [122] and two-component systems[123] Another phenomenological model is the diffuseinterface theory [124] which is based on a treatment ofthe phase interface using simplified density profiles ofthermodynamic properties Another general class ofmodel makes use of conditions at the vapourndashliquidspinodal to constrain the form of the model free energy[125ndash128]
The main drawback in phenomenological models isthe arbitrariness in the form of the proposed work offormation The parameter fitting in practice ensures thatthe populations of small clusters especially the mono-mer and dimer are well represented by the model butthe properties of larger clusters depend on the expres-sion assumed for the model and are therefore uncertainIntuition suggests that phenomenological modelling islikely to have mixed success and that microscopiccalculations of cluster properties do matter There mightbe better value in phenomenological models however ifthey were based securely on the results of microscopiccalculations
8 KINETICS OR THERMODYNAMICS
In section 3 kinetic equations representing the nuclea-tion process were described However the coefficients inthose equations were related to constrained equilibriumpopulations and hence to the equilibrium statisticalmechanics of clusters through the detailed balancecondition (8) Is it necessary to do this Could acompletely kinetic route be followed by evaluating themean growth and decay coefficients bi and gi directly bysome statistical approach The main problem is ofcourse the difficulty in performing such calculationsand this has restricted the development of such models[129ndash131] It is expected that the kinetic coefficients willcorrespond to those derived from the detailed balanceargument for near-equilibrium conditions but circum-stances can be imagined where non-equilibrium effectsare present which cannot be dealt with within standardstatistical mechanics
The framework for evaluating kinetic coefficients ofcluster decay can be briefly introduced [86 132] Amolecular dynamics simulation of a cluster can be set upand evolved according to the interactions until bychance an evaporation event occurs By extensivesimulation statistics for the mean evaporation rates gican be amassed in this way Similarly moleculardynamics simulations where a single molecule iscaptured by a cluster could be performed and thegrowth rates bi determined These will be functions ofcluster energy and so would fit best into a rate equationscheme involving both size and energy [133]
This is a laborious approach but statistical physicsoffers an approximate alternative representation of theescape and capture processes The basic tool is theLangevin equation The radial motion of a molecule inthe cluster is described by [134]
meurorr frac14 f ethrTHORN thornfethr _rr tTHORN eth27THORN
where r is the radial position m is the molecular mass fis the mean (time- and velocity-averaged) force on themolecule at position r and f is the stochastic forceThe latter is usually represented as a velocity-dependent friction term mg _rr where g is the so-calledfriction coefficient plus a fluctuating term ~ff ethr tTHORN withzero mean and correlation in time h ~ff ethr tTHORN~ff ethr t0THORNi frac14eth2gkTmTHORNdetht t0THORN where k is Boltzmannrsquos constant andwhere T has the characteristics of a temperature as willbe seen It is a standard manipulation [135] to convertthe Langevin description in the large friction limit intoa FokkerndashPlanck equation
qwqt
frac14 1
mg qethfwTHORN
qrthorn kT
q2wqr2
eth28THORN
which represents the evolution of wethr tTHORN the prob-ability density that the molecule should lie at the radialposition r It should be noted in passing that FokkerndashPlanck equations have often been used to describe thestochastic dynamics of a cluster in size space [48 136] Incontrast the use of the method described above refers tomotion of a molecule in real coordinate space relative tothe centre of mass of the cluster
The steady state equilibrium solution of equation (28)is
wethrTHORN exp FethrTHORNkT
eth29THORN
where FethrTHORN is called the potential of mean force relatedto the mean force on the molecule f throughf frac14 dF=dr It can now be seen how the parameter Tin the stochastic force plays the role of temperaturesince equation (29) looks like a Boltzmann distributionThis solution is not what is sought however The escapeproblem is characterized by a boundary conditionwethreTHORN frac14 0 where re is the radius at which a particleescapes from the system The resulting escape rate maybe shown to be [134]
gkin frac14 KethreT m gTHORN exp DFkT
eth30THORN
where the prefactor K depends on the parameters listedas well as the shape of the potential of mean force Theprincipal feature of equation (30) is the exponentialdependence on the depth DF of the potential of meanforce Therefore if the potential of mean force FethrTHORN can
I J FORD894
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
be evaluated by studying a molecular dynamics trajec-tory for example then the kinetic decay rate can bedetermined This could be extended to account for aradially dependent temperature and friction coefficientboth of which might be determined from simulation Asketch of a potential of mean force is given in Fig 6
Kinetically determined cluster growth and decayrates gkini and bkini respectively could provide a newroute for the determination of nucleation rates inde-pendent of assumptions about detailed balance and theuse of equilibrium statistical mechanics in a non-equilibrium situation
9 CONCLUSIONS
This review has been brief and in some ways partisan tothe personal views of the author It attempts to sketch acoherent picture of nucleation theory and how thevarious disparate strands relate to one another One ofthe fascinations of nucleation theory is the variety ofapproaches and the number of difficult issues that areencountered How well does free energy fluctuationtheory represent a real non-equilibrium process Howwell does the idea of detailed balance provide aframework for determining non-equilibrium decaycoefficients from equilibrium statistical mechanicsHow well do various methods perform in the evaluation
of cluster free energies Can cluster growth and decayrates be calculated directly without the need for such aframework How exactly are clusters to be defined
The theoretical problems are many but the issues canbe resolved to various degrees of satisfaction Thereremains of course the fundamental task of choosing anintermolecular potential that adequately represents thephysical interactions Without this all the finer points ofcluster definition and treatments of non-equilibriumprocesses are unimportant Modelling many problems inscience relies on the availability of an accurate micro-scopic description and so nucleation rate calculationsare not unique in this respect Most empirical molecularpotentials are fitted to the properties of bulk and surfaceproperties and in the very best examples attention isgiven to their transferability to more unusual environ-ments such as clusters In view of these capabilities andmodern theoretical developments in nucleation theoryit seems likely that successful microscopic models of thenucleation process based on statistical physics are atlast coming within reach
ACKNOWLEDGEMENTS
The present author is grateful to colleagues SarahHarris and Marco Arosio for providing some of thefigures
Fig 6 A molecule in a cluster may be considered to be confined within it by a potential of mean force FethrTHORN Itcan escape by reaching a radial position re driven by a phenomenological stochastic force
STATISTICAL MECHANICS OF NUCLEATION 895
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
REFERENCES
1 Gross D H E Microcanonical Thermodynamics 2001
(World Scientific Singapore)
2 Seinfeld J H and Pandis S N Atmospheric Chemistry
and Physics 1998 (John Wiley New York)
3 Climate Change 2001 Synthesis Report 2001 (Inter-
governmental Panel on Climate Change) (Cambridge
University Press Cambridge)
4 Kohler T M Gail H-P and Sedlmayr E MgO dust
nucleation in M-stars calculation of cluster properties
and nucleation rates Astronomy and Astrophysics 1997
320 553
5 Kashchiev D Nucleation Basic Theory with Applications
2000 (ButterworthndashHeinemann Oxford)
6 Bakhtar F Ebrahimi M and Webb R A On the
performance of a cascade of turbine rotor tip section
blading in nucleating steam Part 1 surface pressure
distributions Proc Instn Mech Engrs 1995 209(C2)
115ndash124
7 Bakhtar F Ebrahimi M and Bamkole B O On the
performance of a cascade of turbine rotor tip section
blading in nucleating steam Part 2 wake traverses Proc
Instn Mech Engrs 1995 209(C3) 169ndash178
8 Bakhtar F andMahpeykar M R On the performance of
a cascade of turbine rotor tip section blading in nucleating
steam Part 3 theoretical treatment Proc Instn Mech
Engrs Part C J Mechanical Engineering Science 1997
211(C3) 195ndash210
9 Volmer M and Weber A Z Physikalische Chemie 1926
119 277
10 Farkas L Z Physikalische Chemie (Leipzig) 1927 125
236
11 Becker R and Doring W Kinetische Behandlung der
Keinbildung in ubersattigten Dampfen Ann Physik
(Leipzig) 1935 24 719
12 Zeldovich J Soviet Physics JETP 1942 12 525
13 Frenkel J Kinetic Theory of Liquids 1946 (Oxford
University Press Oxford)
14 Wagner P E and Strey R Homogeneous nucleation
rates of water vapor measured in a two-piston expansion
chamber J Phys Chemistry 1981 85 2694
15 Miller R Anderson R J Kassner J L and Hagen
D E Homogeneous nucleation rate measurements for
water over a wide range of temperature and nucleation
rate J Chem Physics 1983 78 3204
16 Viisanen Y Strey R and Reiss H Homogeneous
nucleation rates for water J Chem Physics 1993 99 4680
17 Viisanen Y Strey R and Reiss H Erratum homo-
geneous nucleation rates for water J Chem Physics
2000 112 8205
18 Wolk J and Strey R Homogeneous nucleation of H2O
and D2O in comparison the isotope effect J Phys
Chemistry B 2001 105 11683
19 Heath C H Streletsky K and Wyslouzil B E H2Ondash
D2O condensation in a supersonic nozzle J Chem
Physics 2002 117 6176
20 Mikheev V B Irving P M Laulainen N S Barlow
S E and Pervukhin V V Laboratory measurements of
water nucleation using a laminar flow chamber J Chem
Physics 2002 116 10 772
21 Hung C H Krasnopoler M J and Katz J L
Condensation of a supersaturated vapor VIII The
homogeneous nucleation of n-nonane J Chem Physics
1989 90 1856
22 Rudek M M Fisk J A Chakarov V M and Katz
J L Condensation of a supersaturated vapor XII The
homogeneous nucleation of the n-alkanes J Chem
Physics 1996 105 4707
23 Strey R Wagner P E and Schmeling T Homogeneous
nucleation rates for n-alcohol vapors measured in a two-
piston expansion chamber J Chem Physics 1986 84
2325
24 Lihavainen H Viisanen Y and Kulmala M Homo-
geneous nucleation of n-pentanol in a laminar flow
diffusion chamber J Chem Physics 2001 114 10 031
25 Rudek M M Katz J L Vidensky I V Zdımal V and
Smolık J Homogeneous nucleation rates of n-pentanol
measured in an upward thermal diffusion cloud chamber
J Chem Physics 1999 111 3623
26 Grassmann A and Peters F Homogeneous nucleation
rates of n-propanol in nitrogen measured in a piston-
expansion tube J Chem Physics 2002 116 7617
27 Kane D and El-Shall S Condensation of supersaturated
vapors of hydrogen bonding molecules ethylene glycol
propylene glycol trimethyl glycol and glycerol J Chem
Physics 1996 105 7617
28 Hameri K and Kulmala M Homogeneous nucleation in
a laminar flow chamber the effect of temperature and
carrier gas pressure on dibutyl phthalate vapor nucleation
rate at high supersaturations J Chem Physics 1996 105
7696
29 Viisanen Y Strey R Laaksonen A and Kulmala M
Measurement of the molecular content of binary nuclei
II Use of the nucleation rate surface for waterndashethanol
J Chem Physics 1994 100 6062
30 Viisanen Y Wagner P E and Strey R Measurement of
the molecular content of binary nuclei IV Use of the
nucleation rate surfaces for the n-nonane-n-alcohol series
J Chem Physics 1998 108 4257
31 Fisk J A Rudek M M Katz J L Beiersdorf D and
Uchtmann H The homogeneous nucleation of cesium
vapor Atmos Res 1998 46 211
32 Uchtmann H Dettmer R Baranovskii S D and
Hensel F Photoinduced nucleation in supersaturated
mercury vapor J Chem Physics 1998 108 9775
33 Ferguson F T and Nuth J A Experimental studies of
the vapor phase nucleation of refractory compounds V
The condensation of lithium J Chem Physics 2000 113
4093
34 Laaksonen A Talanquer V and Oxtoby D W
Nucleation measurements theory and atmospheric
applications A Rev Phys Chemistry 1995 46 489
35 Debenedetti P G Metastable Liquids Concepts and
Principles 1996 (Princeton University Press Princeton
New Jersey)
36 Yasuoka K and Matsumoto M Molecular dynamics of
homogeneous nucleation in the vapor phase I Lennardndash
Jones fluid J Chem Physics 1998 109 8451
37 Yasuoka K and Matsumoto M Molecular dynamics of
homogeneous nucleation in the vapor phase II Water
J Chem Physics 1998 109 8463
I J FORD896
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
38 Wonczak S Strey R and Stauffer D Confirmation of
classical nucleation theory by Monte Carlo simulations in
the 3-dimensional Ising model at low temperature
J Chem Physics 2000 113 1976
39 Laasonen K Wonczak S Strey R and Laaksonen A
Molecular dynamics simulations of gasndashliquid nucleation
of LennardndashJones fluid J Chem Physics 2000 113 9741
40 Reiss H Methods of Thermodynamics 1996 (Dover
Publications New York)
41 Kondepudi D and Prigogine I Modern Thermodynamics
From Heat Engines to Dissipative Structures 1998 (John
Wiley Chichester West Sussex)
42 Smoluchowski M Ann Phys NY 1908 25 225
43 Einstein A Ann Phys NY 1910 338 1275
44 Hirschfelder J O Curtis C F and Bird R BMolecular
Theory of Gases and Liquids 1954 (John Wiley New
York)
45 Landau L D and Lifshitz E M Statistical Physics 1958
(Pergamon London)
46 Wu D T Nucleation theory Solid State Physicsmdash
Advances in Res and Applic 1997 50 37
47 Reguera D Rubı J M and Perez-Madrid A Fokkerndash
Planck equations for nucleation processes revisited
Physica A 1998 259 10
48 Schenter G K Kathmann S and Garrett B Variational
transition state theory of vapor phase nucleation
J Chem Physics 1999 109 7951
49 Onsager L Phys Rev 1931 37 405
50 Courtney W G J Chem Physics 1961 35 2249
51 Girshick S L and Chiu C-P Kinetic nucleation theory
a new expression for the rate of homogeneous nucleation
from an ideal supersaturated vapor J Chem Physics
1990 93 1273
52 Ford I J Barrett J C and Lazaridis M Uncertainties
in cluster energies in homogeneous nucleation theory
J Aerosol Sci 1993 24 581
53 Weakliem C L and Reiss H The factor 1S in the
classical theory of nucleation J Phys Chemistry 1994
98 6408
54 McClurg R B and Flagan R C Critical comparison of
droplet models in homogeneous nucleation theory
J Colloid Interface Sci 1998 201 194
55 Mashmoushy H Mahpeykar M R and Bakhtar F
Studies of nucleating and wet steam flows in two-dimen-
sional cascades Proc Instn Mech Engrs Part C J
Mechanical Engineering Science 2004 218(C8) 843ndash858
56 Ford I J Nucleation theorems the statistical mechanics
of molecular clusters and a revision of classical nuclea-
tion theory Phys Rev E 1997 56 5615
57 ten Wolde P R and Frenkel D Computer simulation
study of gasndashliquid nucleation in a LennardndashJones system
J Chem Physics 1998 109 9901
58 Kusaka I Wang Z G and Seinfeld J H Direct
evaluation of the equilibrium distribution of physical
clusters by a grand canonical Monte Carlo simulation
J Chem Physics 1998 108 3416
59 Kusaka I and Oxtoby D W Identifying physical
clusters in vapor phase nucleation J Chem Physics
1999 110 5249
60 Stillinger F H J Chem Physics 1963 38 1486
61 Ford I J Imperfect vaporndashgas mixtures and homo-
geneous nucleation J Aerosol Sci 1992 23 447
62 Lothe J and Pound G M J Chem Physics 1962 36
2080
63 Lee J K Barker J A and Abraham F F Theory and
Monte Carlo simulation of physical clusters in the
imperfect vapor J Chem Physics 1973 58 3166
64 Abraham F F Homogeneous Nucleation Theory 1974
(Academic Press New York)
65 Reif F Fundamentals of Statistical and Thermal Physics
1965 (McGraw-Hill New York)
66 Hoare M R and Pal P Adv Phys 1975 24 645
67 Hoare M R Adv Chem Physics 1979 40 49
68 Pal P and Hoare M R Thermodynamic properties
and homogeneous nucleation of molecular clusters of
nitrogen J Phys Chemistry 1987 91 2474
69 Reiss H Katz J L and Cohen E R Translationndash
rotation paradox in the theory of nucleation J Chem
Physics 1968 48 5553
70 Hale B N Monte Carlo calculations of effective surface
tension for small clusters Aust J Physics 1996 49 425
71 Kathmann S M Schenter G K and Garrett B C
Dynamical nucleation theory calculation of condensation
rate constants for small water clusters J Chem Physics
1999 111 4688
72 Reiss H Tabazadeh A and Talbot J Molecular theory
of vapor phase nucleation the physically consistent
cluster J Chem Physics 1990 92 1266
73 Ellerby H M Weakliem C L and Reiss H Toward a
molecular theory of vaporndashphase nucleation I Identifica-
tion of the average embryo J Chem Physics 1992 97
5766
74 Ellerby H M and Reiss H Toward a molecular theory
of vaporndashphase nucleation II Fundamental treatment of
the cluster distribution J Chem Physics 1992 97 5766
75 Weakliem C L and Reiss H Toward a molecular theory
of vaporndashphase nucleation III Thermodynamic properties
of argon clusters fromMonte Carlo simulations and modi-
fied liquid drop theory J Chem Physics 1994 101 2398
76 Kusaka I Wang Z G and Seinfeld J H Binary
nucleation of sulfuric acidndashwater Monte Carlo simula-
tion J Chem Physics 1998 108 6829
77 Oh K J and Zeng X C A small system ensemble
Monte Carlo simulation of supersaturated vapor
evaluation of barrier to nucleation J Chem Physics
2000 112 294
78 Senger B Schaaf P Corti D S Bowles R Voegel J
C and Reiss H A molecular theory of the homogeneous
nucleation rate I Formulation and fundamental issues
J Chem Physics 1999 110 6421
79 Hill T L Statistical Mechanics Principles and Selected
Applications 1956 (McGraw-Hill New York)
80 Soto R and Cordero P Nonideal gas of clusters at
equilibrium J Chem Physics 1999 110 7316
81 Barrett J C The significance of cluster lifetime in
nucleation theory J Chem Physics 2002 116 8856
82 Lushnikov A A and Kulmala M Dimers in nucleating
vapors Phys Rev E 1998 58 3157
83 Bahadur R and McClurg R B Comparison of cluster
definitions in homogeneous nucleation rate calculations
J Phys Chemistry 2001 105 11 893
84 Pugnaloni L A and Vericat F New criteria for cluster
identification in continuum systems J Chem Physics
2002 116 1097
STATISTICAL MECHANICS OF NUCLEATION 897
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
85 Chen B Siepmann J I Oh K J and Klein M L
Simulating vaporndashliquid nucleation of n-alkanes J Chem
Physics 2002 116 4317
86 Harris S A and Ford I J A dynamical definition of
quasi-bound molecular clusters J Chem Physics 2003
118 9216
87 Shreve A P Walton J R P B and Gubbins K E
Liquid drops of polar molecules J Chem Physics 1986
85 2178
88 ten Wolde P R Ruiz-Montero M J and Frenkel D
Numerical calculation of the rate of homogeneous gasndash
liquid nucleation in a LennardndashJones system J Chem
Physics 1998 109 9901
89 ten Wolde P R Oxtoby D W and Frenkel D Chain
formation in homogeneous gasndashliquid nucleation of polar
fluids J Chem Physics 1998 111 4762
90 Vehkamaki H and Ford I J Critical cluster size and
droplet nucleation rate from growth and decay simula-
tions of LennardndashJones clusters J Chem Physics 2000
112 4193
91 Schaaf P Senger B Voegel J C Bowles R K and
Reiss H Simulative determination of kinetic coefficients
for nucleation rates J Chem Physics 2001 114 8091
92 Chen B Siepmann J I Oh K J and Klein M L
Aggregation-volume-bias Monte Carlo simulations of
vaporndashliquid nucleation barriers for LennardndashJonesium
J Chem Physics 2001 115 10 903
93 Hale B N and DiMattio D J A Monte Carlo discrete
sum (MCDS) nucleation rate model for water In
Proceedings of the 15th International Conference on
Nucleation and Atmospheric Aerosols AIP Conference
Proceedings Vol 534 (Eds B N Hale and M Kulmala)
2000 pp 31ndash34
94 Kathmann S M and Hale B N Monte Carlo
simulations of small sulfuric acidndashwater clusters
J Phys Chemistry B 2001 105 11 719
95 Daura X Mark A E and van Gunsteren W F Para-
metrization of aliphatic CHn united atoms of GROMOS96
force field J Comput Chemistry 1998 19 535
96 Kathmann S M Schenter G K and Garrett B C
Understanding the sensitivity of nucleation kinetics a
case study on water J Chem Physics 2002 116 5046
97 Kashchiev D On the relation between nucleation work
nucleus size and nucleation rate J Chem Physics 1982
76 5098
98 Oxtoby D W and Kashchiev D A general relation
between the nucleation work and the size of the nucleus in
multicomponent nucleation J Chem Physics 1994 100
7665
99 Ford I J Thermodynamic properties of critical clusters
from measurements of vapourndashliquid homogeneous
nucleation rates J Chem Physics 1996 105 8324
100 Oxtoby D W and Evans R Nonclassical nucleation
theory for the gasndashliquid transition J Chem Physics
1988 89 7521
101 Zeng X C and Oxtoby D W Gasndashliquid nucleation in
LennardndashJones fluids J Chem Physics 1991 94 4472
102 Evans R Fundamentals of Inhomogeneous Fluids 1992
Ch 3 pp 85ndash175 (Marcel Dekker New York)
103 Talanquer V and Oxtoby D W Dynamical density
functional theory of gasndashliquid nucleation J Chem
Physics 1994 100 5190
104 Barrett J C First-order correction to classical nucleation
theory a density functional approach J Chem Physics
1999 111 5938
105 Barrett J C Cluster translation and growth in density
functional theories of homogeneous nucleation J Chem
Physics 1997 107 7989
106 Bykov T V and Zeng X C A patching model for
surface tension of spherical droplet and Tolman length J
Chem Physics 1999 111 10 602
107 Talanquer V and Oxtoby D W Nucleation in molecular
and dipolar fluids interaction site model J Chem
Physics 1995 103 3686
108 Seok C and Oxtoby D W Nucleation in n-alkanes a
density functional approach J Chem Physics 1998 109
7982
109 Reguera D Bowles R K Djikaev Y and Reiss H
Phase transitions in systems small enough to be clusters
J Chem Physics 2003 118 340
110 Khakshouri S and Ford I J J Chem Physics 2004
(submitted)
111 Ford I J (unpublished)
112 Nyquist R M Talanquer V and Oxtoby D W Density
functional theory of nucleation a semi-empirical
approach J Chem Physics 1995 103 1175
113 Talanquer V and Oxtoby D W Density functional
analysis of phenomenological theories of gasndashliquid
nucleation J Chem Physics 1995 99 2865
114 Katz J L Fisk J A and Chakarov V The accuracy of
nucleation theory In Proceedings of the 13th Interna-
tional Conference on Nucleation and Atmospheric Aerosols
(Eds N Fukuta and P E Wagner) 1992 pp 11ndash18
(Deepak Publishing Hampton Virginia)
115 Fisher M E The theory of condensation and the critical
point Physics 1967 3 255
116 Dillmann A and Meier G E A A refined droplet
approach to the problem of homogeneous nucleation
from the vapor phase J Chem Physics 1991 94 3872
117 Tolman R C J Chem Physics 1949 17 333
118 Ford I J Laaksonen A and Kulmala M Modification
of the DillmannndashMeier theory of homogeneous nuclea-
tion J Chem Physics 1993 99 764
119 Laaksonen A Ford I J and Kulmala M Revised
parametrization of the DillmannndashMeier theory of homo-
geneous nucleation Phys Rev E 1994 49 5517
120 Kalikmanov V I and van Dongen M E H Semipheno-
menological theory of homogeneous vaporndashliquid nuclea-
tion J Chem Physics 1995 103 4250
121 Hale B N Application of a scaled homogeneous
nucleation-rate formalism to experimental-data at
T 5Tc Phys Rev A 1986 33 4156
122 Martinez D M Ferguson F T Heist R H and Nuth
J A Application of scaled nucleation theory to metallic
vapor condensation J Chem Physics 2001 115 310
123 Hale B N and Wilemski G A scaled nucleation model
for ideal binary systems Chem Physics Lett 1999 305
263
124 Granasy L Diffuse interface theory for homogeneous
vapor condensation J Chem Physics 1996 104
5188
125 McGraw R and Laaksonen A Interfacial curvature free
energy the Kelvin relation and vaporndashliquid nucleation
rate J Chem Physics 1997 106 5284
I J FORD898
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
126 Talanquer V A new phenomenological approach to
gasndashliquid nucleation based on the scaling properties
of the critical nucleus J Chem Physics 1997 106
9957
127 Koga K and Zeng X C Thermodynamic expansion of
nucleation free-energy barrier and size of critical nucleus
near the vaporndashliquid coexistence J Chem Physics 1999
110 3466
128 Kusaka I On the scaling behavior of the free energetics of
nucleation J Chem Physics 2003 118 5510
129 Nowakowski B and Ruckenstein E A kinetic approach
to the theory of nucleation in gases J Chem Physics
1991 94 1397
130 Nowakowski B and Ruckenstein E Homogeneous
nucleation in gases a three-dimensional FokkerndashPlanck
equation for evaporation from clusters J Chem Physics
1991 94 8487
131 Wilcox C F and Bauer S H Estimation of homo-
geneous nucleation flux via a kinetic model J Chem
Physics 1991 94 8302
132 Hanggi P Talkner P and Borkovec M Reaction rate
theory fifty years after Kramers Rev Mod Physics 1990
62 251
133 Barrett J C Clement C F and Ford I J Energy
fluctuations in homogeneous nucleation theory for aero-
sols J Physics A 1993 26 529
134 Ford I J and Harris S A Molecular cluster decay
viewed as escape from a potential of mean force J Chem
Physics 2004 120 4428
135 Risken H The FokkerndashPlanck Equation Methods of
Solution and Applications 2nd edition 1989 (Springer-
Verlag Berlin)
136 Shizgal B and Barrett J C Time dependent nucleation
J Chem Physics 1989 91 6505
STATISTICAL MECHANICS OF NUCLEATION 899
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
deviations from CNT of over five orders of magnitude[23] In an effort to acquire definitive data for a singlesubstance the nucleation of n-pentanol was givenparticular attention by a number of research groups inthe late 1990s and early 2000s For temperatures in therange 260ndash290K most data lie three orders of magnitudeabove the CNT predictions [24] although some measure-ments lie between one and three orders of magnitudebelow [25] Further results for n-pentanol show goodagreement with the CNT at 273K [26] but withdeviations away from this temperature In the light ofthe reasonable success of CNT in explaining water datait is disappointing to find that hydrogen-bonded organicmaterials such as a series of glycols disagree verystrongly with the temperature dependence of thenucleation rate suggested by CNT [27] Similarly theorganic species dibutyl phthalate deviates by up to sixorders of magnitude from the model [28] The situationbecomes even more unsatisfactory for the nucleation oftwo or more species together Strong deviations frombinary CNT are found for the ethanolndashwater system [29]and for a range of n-nonanendashn-alcohol mixtures [30]However it is with metals that the failure of CNT ismost dramatic examples in the literature being caesium[31] mercury [32] and lithium [33] For caesium thecritical supersaturation (that required to drive nuclea-tion at a specified rate) is between two and ten timeslarger than the CNT predictions The difference betweenmeasured and predicted nucleation rates is astronom-ical Although in most materials the deviation is lesssevere than this it is certainly not the case that a smallcorrection to CNT is all that is needed to account for alldata
The usual CNT expression for the rate is
JCNT frac14 q2spm
1=2 rvrl
pvV
kT
6 exp 16ps3
3ethkTTHORN3r2l ethlnSTHORN2
eth1THORN
where q is a sticking or accommodation coefficient m isthe molecular mass k is Boltzmannrsquos constant s is thesurface tension of a planar interface between thecondensate saturated vapour pv is the vapour pressureand rl and rv are the densities of the bulk condensedphase and the vapour respectively in molecules per unitvolume S is the supersaturation defined by S frac14 pv=p
sv
where psv is the saturated vapour pressure at thetemperature T Note that J in this expression is thenumber of particles formed per second and it isproportional to the system volume V There are goodreasons to insert a factor of 1S into the prefactor of theabove expression but this is the often-quoted formCalculations are easy agreement is reasonable and sowhy should anyone bother to develop the theoryfurther
The problem in employing this formula for water isthat the laboratory data for water are taken at lowtemperatures and need to be extrapolated to hightemperatures for application to turbine conditions Asalready noted water seems to be a lucky case wheretheory and experiment actually agree for a particular setof conditions namely at a temperature of about 240Kand for a range of supersaturations Unfortunately adiscrepancy appears and grows as the temperature israised and at 260K the theory overpredicts the rate byan order of magnitude [18] The ratio of measurednucleation rate to CNT prediction has been fitted to asimple function by Wolk and Strey [18]
JexpethT STHORN frac14 JCNT exp Athorn B
T
eth2THORN
where A frac14 2756 and B frac14 6500K This formula fitsdata in the temperature range 220ndash260K but has alsobeen found to account for data at temperatures sometens of kelvins higher (B Wyslouzil personal com-munication 2003)
Nevertheless caution must be exercised in extrapolat-ing to temperatures around 370K for steam turbinesDevelopments in the theory are certainly needed It is achallenge to understand why CNTworks reasonably wellat 240K forwaterwhat terms are cancellingout andwhyHow canmeasured data be extrapolated with confidence
In this review the determination of the properties ofmolecular clusters using the tools of statistical mech-anics is again considered Similar accounts of the widerpicture have been given in various reviews and books[34 35] Starting from very general arguments attemptsare made to discern how a theory such as the CNTmight emerge The subtleties of defining a clusterprecisely in terms of the positions and momenta of itsconstituents are discussed It is noted that the transitionstate theory based on equilibrium cluster properties isequivalent to an approach based on the kinetic theory ofcluster decay at least in certain circumstances Somephenomenological models developed from CNT arereviewed and finally a perspective of future develop-ments in this field is given
By the very nature of a review the discussion is ratherbroad while at the same time some of the technicalitiesare outlined The intention is to give a reasonablycomplete picture of the current state of nucleationtheory at least from the point of view of the presentauthor and to show that significant progress has beenmade in the approximately 80 years of its history whilemore is still clearly necessary
2 TRANSITION STATE THEORY
A theory of nucleation based on statistical ideas isnecessary since direct simulation of the process from a
STATISTICAL MECHANICS OF NUCLEATION 885
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
microscopic point of view is very laborious and has beenrealized in only a very few example cases [36ndash39] Thetime scales involved make a nucleation event simply tooslow to follow directly for realistic conditions and withcurrent computational resources
The simplest statistical treatment of the nucleationprocess comes from TST within a framework of thetheory of free energy fluctuations According to thermo-dynamics a system with externally fixed temperatureand pressure (or volume) will adjust any internal prop-erties in order to minimize its Gibbs (or Helmholtz) freeenergy [40 41] Such adjustments will include changes inphase Nevertheless statistical fluctuations in systemproperties will naturally take place and it is useful tointerpret these as fluctuations in the free energy of thesystem away from the minimum Considering constant-volume conditions for the moment general arguments[42ndash45] suggest that the probability of a fluctuationDF in Helmholtz free energy is proportional toexpfrac12DF=ethkTTHORN
The physical fluctuations that are important in thenucleation process are those where a molecular clusterof new phase is formed temporarily from the originalphase The change in free energy accompanying theformation of a cluster of i molecules in the new phase iswritten DF frac14 DWethiTHORNFethiTHORN if0 where FethiTHORN is the freeenergy of the i cluster and f0 is the free energy permolecule in the old phase This is also called the lsquoworkof formationrsquo since it corresponds in thermodynamics tothe external work needed to insert an additional clusterreversibly into the system while maintaining the overall
temperature and pressure (or volume) Note that thisinsertion of a new cluster while maintaining suchconstraints is a theoretical operation since it would behard to achieve in practice
For a metastable initial phase the magnitude of thefree-energy fluctuation DF peaks at the so-called lsquocriticalsizersquo i This is illustrated in Fig 1 The perturbed statewith the additional critical cluster is therefore athermodynamic transition state since reversible changesin size of the extra cluster are accompanied byreductions in free energy It is presumed that this meansthat the new cluster has equal probabilities of growingor decaying and that the rate of nucleation is thereforerelated to the rate of formation of these critical clusters
The probability of forming the critical clusteraccording to this picture is proportional toexpfrac12DWethi THORN=ethkTTHORN but the proportionality constant(essentially a time scale) is not yet determined Tointroduce this simple kinetic theory [46] can be used tocalculate the collision rate bi between single molecules(monomers) and an i-cluster modelled as a sphere withbulk condensed phase molecular density rl
bi frac14qpv
eth2pmkTTHORN1=21thorn 1
i
1=2
1thorn i1=3 2
eth36pTHORN1=3
r2=3l
eth3THORN
where m is the molecular mass It is more usual toreplace the expression in square brackets with i2=3 thelimit for large i The simplest TST expression for the
Fig 1 Sketch of the cluster work of formation for a range of (constrained) cluster sizes The maximum in thecurve defines the critical size i and the critical work of formation DWethi THORN
I J FORD886
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
nucleation rate is then J bini where ni is the meannumber of critical clusters in the system
More sophisticated treatments of thermodynamicscan provide the time scale without recourse to the abovekinetic argument which is reasonably valid for gas-to-particle conversion but not for freezing for exampleAttempts based on linearized non-equilibrium thermo-dynamics [47 48] originate from the work of Onsager[49]
The next step is crucial and it is to relate thepopulation ni to the probability of a fluctuation and toan explicit model of the critical work of formation Theassumption in classical nucleation theory is that
ni frac14 n1 expDWclethi THORN
kT
eth4THORN
where DWclethiTHORN is the classical work of formation givenby
DWclethiTHORN frac14 4pR2s ikT lnS eth5THORN
and DWclethi THORN is the maximum of this function Thecluster is envisaged to be a sphere of radius R given by4pR3rl=3 frac14 i
Unfortunately some uncertainty accompanies thesesteps in the derivation The theory is flawed on severalcounts The work of formation contains a termassociated with the creation of a spherical interface(characterized by the bulk value of surface tension) anda change kT lnS in free energy per molecule due tothe change in phase from supersaturated vapour tocondensate Moreover the factor n1 appearing inequation (4) appears to have been introduced heuristi-cally Might it not instead be the monomer populationin the saturated vapour How does DWclethiTHORN comparewith the actual work of formation DWethiTHORN Is thedroplet really spherical with a sharp interface Have
any important factors been left out Furthermoreshould there be concern that equation (4) is in-consistent when i frac14 1 This has been a controversialissue [50ndash54]
3 KINETIC INTERPRETATION
The expression for the classical rate of nucleation givenin equation (1) emerges from the models in equation (4)and (5) when the thermodynamic transition stateanalysis is embellished by a kinetic interpretation [11]Instead of using the probability of fluctuations awayfrom the equilibrium state (the TST route) a set of rateequations is introduced to describe the dynamics ofcluster populations
The evolution of cluster populations is modelled usingthe following equations
dni
dtfrac14Xj
njPji niXj
Pij eth6THORN
where Pji is the mean rate at which a j cluster convertsinto an i cluster The transition processes are assumed tobe Markovian so that the rate coefficients depend onthe properties of the cluster and its environment but noton their history
The kinetic interpretation can be taken a step furtherby assuming that the only important transitions arethose involving the addition or loss of single moleculesto or from the cluster The scheme of transitionprocesses is illustrated in Fig 2 The only non-zerorate coefficients are then bi frac14 Piithorn1 for cluster growthand gi frac14 Pii1 for cluster decay The rate equations
Fig 2 The scheme of the birth and death of clusters addition of a monomer to a cluster will take it up thechain towards the right while the loss of a monomer will take it in the opposite direction Attachmentand detachment of dimers and larger clusters are usually ignored
STATISTICAL MECHANICS OF NUCLEATION 887
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
reduce to
dni
dtfrac14 bi1ni1 gini bini thorn githorn1nithorn1 eth7THORN
These are the BeckerndashDoring equations [11] and theyprovide a useful reinterpretation of the TST approachThere are two steady state solutions In the first thepopulations satisfy equation (7) and also the detailedbalance condition
bi1nei1 frac14 gin
ei eth8THORN
and the fnei g are taken to represent the cluster popula-tions in an equilibrium vapour at or below the saturatedvapour pressure For true thermal equilibrium themonomer population has to be less than or equal to itsvalue for the saturated vapour ns1 The cluster populationsin a saturated vapour are denoted fnsig Nevertheless asituation can be invented which is known as lsquoconstrainedequilibriumrsquo where the detailed balance condition is stillsatisfied but with n1 gt ns1 Since bi pvS while gi isindependent of vapour pressure equation (8) then definesso-called constrained equilibrium populations
ncei ampSinsi eth9THORN
for a supersaturated vapour It has been assumed thatS frac14 pv=p
svampn1=n
s1
The second steady state solution to the BeckerndashDoring equations is more realistic It is not meant torepresent thermal equilibrium The supersaturation isset to a value greater than unity and a boundarycondition of zero population at some cluster size imax isimposed These conditions will drive a steady productionrate of large clusters from the vapour The approach isfor mathematical convenience only and the same steadystate cluster populations nssi and nucleation rate wouldemerge if imax was sent off to infinity The detailedbalance condition (8) is replaced by J frac14 bi1n
ssi1 gin
ssi
defining the nucleation current J which can then becalculated analytically as shown in another paper in thisvolume [55] To a good approximation
J frac14 Zdbincei eth10THORN
which involves the constrained equilibrium population atthe critical size i the size with the lowest constrainedequilibrium population The condition imax 4 i musthold Zd is the so-called Zeldovich factor [12] givenapproximately by
Zd frac14 1
2pq2ethln ncei THORN
qi2
ifrac14i
1=2
eth11THORN
When equations (4) and (5) are used to specify ncei and theresult is substituted into equations (10) and (11) then theclassical nucleation rate given in equation (1) isrecovered as long as the large i limit of the attachment
rate coefficient in equation (3) is taken It is also possibleto derive an expression for J purely in terms of the ratecoefficients bi and gi This possibility is considered againin section 8
The main assumptions used in this derivation of theclassical theory principally the validity of equations (4)and (5) are now examined In order to explore this it isnecessary to see how the equilibrium cluster populationsnei in subsaturated vapours may be calculated usingstatistical mechanics
4 CLUSTER STATISTICAL MECHANICS
The statistical mechanical treatment of a vapour is bestdeveloped using a grand canonical ensemble [56ndash59] Thesystem is placed in contact with a particle reservoir at achemical potential and temperature T so that thenumber of molecules N in the system can change Thegrand partition function of the system X is then propor-tional to the integral of expf frac12HethNTHORN N=ethkTTHORNg overall possible configurations of molecular positions andmomenta whereHethNTHORN is the Hamiltonian of the system
Evaluating X exactly is exceedingly difficult and it ismuch simpler to proceed by considering the configura-tions of the system as arrangements of molecularclusters The vapour then is viewed as a mixture ofideal gases each composed of clusters of differentsizes The grand partition function of the system canbe reconstructed from partition functions for singlei-clusters In order to do this however a cluster must bedefined in a unique way in terms of molecular positionsand velocities An example of such a definition would beto require the separation between molecules in a clusterto be less than a maximum distance Rc [60] as illustratedin Fig 3
Fig 3 According to the Stillinger definition a molecule isconsidered part of a cluster if it lies closer than adistance Rc from a constituent molecule of the cluster
I J FORD888
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
At first sight the cluster definition is completelyarbitrary a means of categorizing and resumming allthe elementary configurations of the system Howeveraccurately representing the system as a mixture of idealgases requires the interactions between molecules lyingin different clusters to be minimized The clusterdefinition affects whether a particular interaction isregarded as intercluster or intracluster and so it plays acrucial role This will be discussed further in section 6Residual clusterndashcluster interactions can be taken intoaccount in the form of a virial series [61]
These ideas are now put into practice Define amodified canonical partition function zi for a clustercontaining i molecules This is simply the integral ofexpf frac12HcethiTHORN i=ethkTTHORNg over configurations involvingmolecules in the cluster alone with Hc the Hamiltonianof the i molecules comprising the cluster Using thepicture of a mixture of ideal gases the grand partitionfunction of the complete system can then be expressed as
XampXfnig
Yifrac141
znii
ni
eth12THORN
where the sum is over all possible cluster populationdistributions fnig This is demonstrated rigorously inappendix A of reference [57]
The next step is to determine the distribution fnei gthat dominates the sum in equation (12) The logarithmof the summand is extremized
qqni
Xifrac141
lnz
nii
ni
frac14 0 eth13THORN
which gives the following very simple expression for themost probable or equilibrium cluster size distributionfor the given conditions T and
nei frac14 ziethTTHORN frac14 ZethTTHORN ei=ethkTTHORN eth14THORN
which is equivalent to the so-called law of mass action
nei frac14 Zine1Z1
i
eth15THORN
where ZiethTTHORN frac14 expfrac12FethiTHORN=ethkTTHORN frac14 zi expfrac12 i=ethkTTHORN isthe canonical partition function for an i cluster and FethiTHORNis the i-cluster Helmholtz free energy This relationshipis often used as the starting point in models of clusterpopulations
The vapour pressure is now given by a sum of partialpressures
pvV frac14 kTXifrac141
nei frac14 kTXifrac141
ziethTTHORN eth16THORN
and this series will only converge if the equilibriumpopulations or the zi decrease sufficiently quickly forlarge i This limits the statistical mechanics approach to
the study of saturated or subsaturated vapours How-ever as in the preceding section the equilibriumpopulations nei in equation (14) (for S lt 1) can beextrapolated into constrained equilibrium populationsncei (with S gt 1) for use in equation (10) Saturatedvapour corresponds to a particular chemical potentialms and the saturated vapour cluster populations are nsi The constrained equilibrium populations are obtainedby inserting into equation (14) the chemical potential ofthe supersaturated vapour which is frac14 s thorn kT lnSgiving ncei frac14 Sinsi which is consistent with the resultobtained in equation (9) for dilute vapours
The constrained equilibrium populations are there-fore expressed as
ncei frac14 SiziethsTTHORN
frac14 expfrac12FethiTHORN is ikT lnS
kT
eth17THORN
and the correspondence with CNT can be made closerby introducing the n1 prefactor
ncei frac14 n1 exp frac12FethiTHORN is thorn kT ln n1 ikT lnS
kT
eth18THORN
The next task is to explore how the combinationFethiTHORN is thorn kT ln n1 might be approximated by theCNT result 4pR2s This can be explored most easilyfor solid rather than liquid clusters
5 SOLID CLUSTERS
The partition function of a solid cluster of i moleculescan be factorized into translational rotational andvibrational parts through a transformation to suitablerelative positions and momenta [62ndash64] Starting withthe vibrational degrees of freedom the atom positionsare defined with respect to their mean positions whichare determined by a fixed centre of mass and set oflattice vectors It is assumed that the potential energypart of the Hamiltonian is quadratic in the displace-ments from these mean positions Now the partitionfunction is worked out for configurations of the particlesrestricted to having zero total linear and total angularmomentum about the centre of mass This is writtenZvib corresponding to a product of partition functionsfor 3ji 6 independent oscillators (if there are j atomsper molecule and therefore a total of 3ji degrees offreedom if the molecule is not linear) each of which inthe classical limit takes the form kT=ethhoTHORN where o isthe angular frequency of the specific vibrational modeThe partition function has an additional factor ofexpfrac12U0ethiTHORN=ethkTTHORN where U0ethiTHORN is the system potentialenergy evaluated at the mean atom positions A
STATISTICAL MECHANICS OF NUCLEATION 889
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
different expression applies for low temperatures whenthe oscillator is not classically activated [65]
The rotational and translational degrees of freedomare next The previous restriction of the configurationsto zero angular momentum is lifted The approximate(classical) rotational partition function is
Zrot frac14 p1=28p2kTh2
3=2
ethI1I2I3THORN1=2 eth19THORN
where Ik are the principal moments of inertia about thecentre of mass for atoms fixed in their mean positions
Finally the translational free energy of the centre ofmass of the cluster is given by the partition function
Ztrans frac14 V2pMkT
h2
3=2
eth20THORN
where V is the system volume and M frac14 im is the clustermass The free energy of the cluster (plus kT ln n1 forconvenience) is therefore
FethiTHORN thorn kT ln n1
frac14 kT lnZtransZrotZvib
n1
frac14 U0ethiTHORN kT ln
eth2pMkTTHORN3=2
Srsvh3p1=2
8p2kTh2
3=2
6ethI1I2I3THORN1=2Y3ji6
kfrac141
kT
hok
eth21THORN
where rsv frac14 ns1=V is the molecular density of dilutesaturated vapour
By similar arguments the chemical potential of dilutesaturated vapour is
s frac14 Um0 kT ln
eth2pmkTTHORN3=2
rsvh3p1=2
8p2kTh2
3=2
6ethIm1 Im2 Im3 THORN1=2Y3j6
kfrac141
kT
homk
eth22THORN
where Um0 is the potential energy of a molecule when the
atoms are at their mean positions Imk are the threeprincipal moments of inertia of the molecule about thecentre of mass and om
k are the oscillatory frequencies ofmolecular internal vibrations
The classical work of formation DWclethiTHORN should nowcorrespond to the leading-order contribution to thedifference between equation (21) and i times equation(22) Calculations [54 64 66ndash68] demonstrate that theleading term is indeed roughly proportional to i2=3 thusjustifying the form of the classical model The suitabilityof the capillarity approximation depends on the extentto which the coefficient of this leading term correspondsnumerically to the planar surface free energy Howeverthe full statistical mechanical expression for the work of
formation contains additional terms and these will beimportant for small clusters and need to be taken intoaccount In order to evaluate them the variousstructural and energy properties of cluster are requiredall of which can be obtained from microscopic modelsand they can then be inserted into the formulae for thethermodynamic properties [equations (21) and (22)] theconstrained equilibrium populations [equation (18)] andultimately the nucleation rate [equation (10)] Followingthis route employs the almost universally held assump-tion that equilibrium properties such as free energies canbe used to describe non-equilibrium properties specifi-cally the cluster decay rate This assumption is re-examined in section 8
6 LIQUID DROPLETS AND THE CLUSTER
DEFINITION
Defining what was meant by a solid cluster in theprevious section was not an issue Within the assump-tions of the model (harmonic interactions) it wasimpossible for a molecule to escape from the clusterThe cluster definition in the case of liquids cannot bedismissed so easily Liquids are more volatile thansolids and the molecules in a cluster are much moremobile It is imperative now to characterize a clusterusing the molecular positions and velocities andperhaps some time scale and length scale too
The traditional microscopic method for defining acluster is to impose a geometric constraint on molecularpositions An example is to restrict molecules to a spherecentred on the centre of mass [69ndash71] or to impose amaximum allowable separation between the molecules(the Stillinger criterion [60]) More sophisticated treat-ments exist such as the n=v Stillinger cluster introducedby Reiss and co-workers [72ndash75] Here the confiningvolume v used to define the cluster is considered to be anadditional defining characteristic on the same level asthe cluster size n As time progresses a cluster changesits confining volume as the component molecules followtheir respective trajectories just as they change their sizewith the gain and loss of molecules The association of amolecular configuration with a precise confining volumeis complicated but tractable The dependence of the freeenergy on v determines the statistical dynamics of thiscluster volume A similar attempt to associate a clusterwith a specified volume within a grand canonicalensemble description was developed by Kusaka andco-workers [58 59 76] The simulation box was taken tobe the confining cluster volume Many of these ideasmay be viewed as a treatment of the cluster definition(eg the Stillinger radius) as a variational parameter inthe representation of the partition function Further
I J FORD890
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
refinement of this point of view using a treatment ofclusterndashcluster interactions to optimize the definingcluster volume has been presented [77] A review ofrecent progress in this area and the application of theseideas to nucleation theory has been provided by Sengeret al [78]
Such geometric constraints can be implemented buttheir realism is questionable Situations better regardedas close encounters between separate clusters would beclassed as snapshots of single larger clusters It shouldbe recalled that the cluster definition is an internalfeature of the model which is to be optimized to reduceclusterndashcluster interactions and to improve the repre-sentation of the grand partition function Misclassifyingconfigurations is fraught with consequences A secondissue to consider is that for convenience the clustersshould behave in a manner that can be modelled withBeckerndashDoring equations This means that the growthand decay rates should be free of memory effects ie thetransition processes should be Markovian The extent towhich this is satisfied (if indeed it is physically realistic)depends on the cluster definition
An intuitively reasonable choice of definition whichaddresses both these points is that a cluster shouldconsist of molecules that are bound energetically Thisshould minimize the clusterndashcluster contributions to thetotal energy According to such a definition if amolecule interacts strongly with a cluster then it islikely to be regarded as part of that cluster Also anenergetically bound cluster is likely to decay by thermalfluctuation in a stochastic Markovian fashion ratherthan by the separation of passing poorly boundfragments
Ideas have been proposed which require the cluster tobe energetically bound (in a loose sense since a cluster isintrinsically prone to decay) The time evolution of amolecular configuration and not simply the set ofinstantaneous positions of its component molecules isof central importance Hill [79] introduced a clusterdefinition requiring the total energies of pairs ofmolecules in a cluster to be negative (in their centre ofmass frame) The intention was to exclude situationswhere a molecule could leave the cluster within the timeit would take for a molecule to travel a typical lineardimension of the cluster Soto and Cordero [80] havedeveloped Hillrsquos criterion and Barrett [81] has per-formed Monte Carlo modelling of molecular configura-tions excluding those which decay in moleculardynamics within a given period A number of otherenergy-based criteria have been developed [82ndash85]
Harris and Ford [86] have recently proposed a newenergy-based cluster definition As with Barrettrsquosapproach it is based on the requirement that amolecular snapshot is energetically bound for a chosensubsequent period Evaporation according to thisscheme requires not only that a molecule should makea positive energy excursion but also that the subsequent
dynamics carry it far away from the cluster Onlyperhaps one in eight molecules manage to avoidrecapture once having satisfied the positive energycriterion [86] A snapshot of a cluster in this casesurrounded by a carrier gas is shown in Fig 4 Thehistory of the energy of a molecule in such a cluster isillustrated in Fig 5 three excursions into positive energyoccur in this sequence but only one leads to actualescape corresponding to the maintenance of positiveenergy for a significant time The instant of escape istaken to be the point at which the energy goes positivefor the third and final time
Cluster definitions implemented in moleculardynamics and Monte Carlo codes allow cluster freeenergies and hence the decay rate to be calculated Anenormous amount of simulation work has been carriedout and calculations can be laborious Many differentschemes and approximations have been made to easethe evaluation of these properties The principalconclusion from various studies using a range of clusterdefinitions and interaction models [49 57 58 70 85 87ndash92] seems to be that the work of formation of liquidcritical clusters does indeed contain a dominantcontribution proportional to i2=3 consistent with theclassical work of formation Hale and collaboratorspresent cluster free energies for water and other speciesin a way that makes this particularly clear indeed it issurprising how dominant the surface term can be foreven very small clusters [70 93 94] Neverthelessfurther contributions exist small clusters are notcharacterized just by scaled-down properties of largeclusters and it is impossible to escape the need todetermine the additional terms
In all the above discussion it has been assumed thatthe intermolecular interactions are accurately specifiedThis is a situation that is rarely possible Althoughfirst-principles quantum-mechanical models could be
Fig 4 Snapshot of a cluster of 50 argon atoms (brightspheres) bathed in a gas of helium (darker spheres)
STATISTICAL MECHANICS OF NUCLEATION 891
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
developed in practice empirical potentials are devel-oped on the basis of physical intuition containingparameters fitted to the measured properties of thesubstance in question [95] Usually these are bulkproperties and it is not always clear that the sameinteractions can be used for the very different circum-stances of a molecular cluster [96] In view of thesensitivity of nucleation rates to the thermodynamicproperties of clusters this is perhaps a crucial failingProgress can be made however by computing knowncluster properties where they are available and byrefining the potentials to suit Extraction of clusterproperties through use of the so-called nucleationtheorems (which relate derivatives of the nucleationrate to properties of the critical cluster [16 56 97ndash99])may provide just this sort of experimental information
A technique that has not yet been described is the useof density functional theory (DFT) to model theproperties of liquid droplets [100ndash104] This powerfulapplication of statistical mechanics promises to avoidlaborious computations at the microscopic level byinstead focusing on the density profile of moleculesmaking up the droplet This profile may be regarded asthe average distribution of molecules about the clustercentre of mass Although this might appear to involvecoarse-graining and approximation it is a well-foundedapproach and in principle can be exact In practical
implementations however a mean field approach istaken such that the constituent molecules move in aneffective potential well created by the other molecules inthe cluster [105] Also the interactions are assumed to becharacterized by strong repulsive cores and weaklyattractive tails
Considerable analytical work has been carried out[104 106] Various developments have been made todescribe droplets consisting of simple atoms as well aschain-like molecules [107] and polar molecules [108]The main drawback in these implementations is thatspecific interactions such as hydrogen bonds are hardto represent Certain details of the modelled systemare neglected [109 110] and the (implicit) clusterdefinition used is geometric and not energetic [111]On the other hand quite sophisticated systems havebeen modelled including multi-component dropletsThe method has been compared with otherapproaches [112 113] and certain features are betterdescribed particularly the approach to spinodalconditions where the nucleation barrier vanishesDFT calculations can be performed relatively easilyand they can provide a qualitative and perhapsquantitative insight into quite complicated clusteringproblems They provide an intermediate-level descrip-tion between the simple capillarity approximation anda microscopic description
Fig 5 The energy of a molecule in the centre-of-mass frame of a cluster occasionally becomes positive butthis does not necessarily lead to escape Only the third positive energy excursion in this sequenceindicated by an arrow results in particle escape A cluster definition must capture this behaviour
I J FORD892
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
7 PHENOMENOLOGICAL MODELS
The reader might be dismayed by the mathematicalcomplexity of the expressions for the work of formationof solid clusters in section 5 might be alarmed by theprocedural complexity of implementing a cluster defini-tion in section 6 may recoil from the difficulties incalculating free energies from molecular simulationmethods and ultimately may give up in view of theuncertainties in the underlying intermolecular interac-tions Considering these complications it is tempting totry to develop phenomenological models of the work offormation After all if the capillarity approximationworks reasonably well for water then surely a minorcorrection to account for small cluster effects ought tobe sufficient to improve matters
However after perhaps two decades of work nomodel has emerged offering broad success across arange of different substances It is perhaps asking toomuch to expect universal improvement CNT works as afirst approximation for water for a particular tempera-ture range but as discussed earlier it possesses thewrong temperature dependence so that predictionsaway from a temperature in the range 210ndash260K areuncertain A similar picture emerges for other sub-stances although in some cases the failure is even moresevere
The simplest phenomenological adjustment to equa-tion (4) the basis for CNT is to write
ncei frac14 n1 expfrac12DWclethiTHORN DWcleth1THORN
kT
eth23THORN
where DWclethiTHORN is given by equation (5) This amounts toa uniform downward shift of the classical free energybarrier illustrated in Fig 1 At least now the expressionis consistent at i frac14 1 The theory is known as internallyconsistent classical theory (ICCT) [51] Unfortunately itis not clear whether this constant shift in the work offormation should be a realistic representation of all thesubleading terms in the work of formation [52] It is anarbitrary modification and its success in accounting fordata is mixed For example nucleation rates for a rangeof alkanes [114] require further modification by factorsof between one and nine orders of magnitude to comeinto agreement with ICCT
It should be noted at this point that the correctionterm kT lnS which appears in DWcleth1THORN can beregarded as rigorous There are good reasons for thisbased on general considerations involving the nucleationtheorems [56] or more simply the principles of chemicalequilibrium represented by the law of mass action givenin equation (15) The correction converts the prefactorn1 in equation (4) into ns1 and results in the insertion of afactor 1=S in the nucleation rate in equation (1)
The Fisher droplet model is a slightly more sophis-ticated scheme [115] Here the constrained equilibrium
populations are modelled by
ncei frac14 ns1 exp yii2=3 thorn t ln i lnq0kT
psv
thorn i lnS
eth24THORN
The term proportional to yi may be viewed as a surfaceterm while the other two are corrections of a typesuggested by the free energy expressions developed insection 5 The parameters q0 yi and t are chosen toensure that the model correctly reproduces knownproperties of clusters and vapours including virialcoefficients and critical properties the theory is there-fore phenomenological
For large cluster sizes and for temperatures wellbelow the critical temperature the free energy offormation should tend towards that of a macroscopicliquid droplet yi should therefore approach A1s=ethkTTHORNfor large i where A1i
2=3 is the surface area 4pR2 of aspherical droplet containing i monomers at the bulkliquid density so that A1 frac14 eth36pTHORN1=3r2=3
l Dillmann andMeier [116] suggested writing
yi frac14kiA1skT
eth25THORN
with the Ansatz
ki frac14 1thorn a1i 1=3 thorn a2i 2=3 eth26THORN
The coefficients a1 and a2 can be found in terms of virialcoefficients of the vapour The form of this Ansatz wasmotivated by the curvature dependence of the bulksurface tension derived in classical thermodynamics byTolman [117]
The initial success of the DillmannndashMeier theory wasshort lived since modifications had to be made toovercome an internal inconsistency [118] Virial expan-sions for the vapour pressure and density were usedincorrectly to calculate the virial expansion of the ratioof the two Correcting this error [61] changed predictedrates by several orders of magnitude Nevertheless theapproach was very influential [119] In a furtherdevelopment Kalikmanov and van Dongen [120] seta2 frac14 0 (this term can be absorbed into q0 anyway) andchose a1 such that the experimental saturated vapourpressure psv is reproduced by the model The DillmannndashMeier model matched the first two virial coefficientsonly The Kalikmanovndashvan Dongen theory shouldtherefore approximate the (revised) DillmannndashMeiertheory at low temperatures but ought to provide abetter description at higher temperatures
Other phenomenological models have been proposedincluding the Hale scaling relation [121] This modelbased on CNT but employing simplified models for thesurface tension and liquid density based on scalingbehaviour near the critical point together with a rather
STATISTICAL MECHANICS OF NUCLEATION 893
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
more straightforward rate prefactor seems to be quitesuccessful in accounting for data for a wide range ofmaterials even metals [122] and two-component systems[123] Another phenomenological model is the diffuseinterface theory [124] which is based on a treatment ofthe phase interface using simplified density profiles ofthermodynamic properties Another general class ofmodel makes use of conditions at the vapourndashliquidspinodal to constrain the form of the model free energy[125ndash128]
The main drawback in phenomenological models isthe arbitrariness in the form of the proposed work offormation The parameter fitting in practice ensures thatthe populations of small clusters especially the mono-mer and dimer are well represented by the model butthe properties of larger clusters depend on the expres-sion assumed for the model and are therefore uncertainIntuition suggests that phenomenological modelling islikely to have mixed success and that microscopiccalculations of cluster properties do matter There mightbe better value in phenomenological models however ifthey were based securely on the results of microscopiccalculations
8 KINETICS OR THERMODYNAMICS
In section 3 kinetic equations representing the nuclea-tion process were described However the coefficients inthose equations were related to constrained equilibriumpopulations and hence to the equilibrium statisticalmechanics of clusters through the detailed balancecondition (8) Is it necessary to do this Could acompletely kinetic route be followed by evaluating themean growth and decay coefficients bi and gi directly bysome statistical approach The main problem is ofcourse the difficulty in performing such calculationsand this has restricted the development of such models[129ndash131] It is expected that the kinetic coefficients willcorrespond to those derived from the detailed balanceargument for near-equilibrium conditions but circum-stances can be imagined where non-equilibrium effectsare present which cannot be dealt with within standardstatistical mechanics
The framework for evaluating kinetic coefficients ofcluster decay can be briefly introduced [86 132] Amolecular dynamics simulation of a cluster can be set upand evolved according to the interactions until bychance an evaporation event occurs By extensivesimulation statistics for the mean evaporation rates gican be amassed in this way Similarly moleculardynamics simulations where a single molecule iscaptured by a cluster could be performed and thegrowth rates bi determined These will be functions ofcluster energy and so would fit best into a rate equationscheme involving both size and energy [133]
This is a laborious approach but statistical physicsoffers an approximate alternative representation of theescape and capture processes The basic tool is theLangevin equation The radial motion of a molecule inthe cluster is described by [134]
meurorr frac14 f ethrTHORN thornfethr _rr tTHORN eth27THORN
where r is the radial position m is the molecular mass fis the mean (time- and velocity-averaged) force on themolecule at position r and f is the stochastic forceThe latter is usually represented as a velocity-dependent friction term mg _rr where g is the so-calledfriction coefficient plus a fluctuating term ~ff ethr tTHORN withzero mean and correlation in time h ~ff ethr tTHORN~ff ethr t0THORNi frac14eth2gkTmTHORNdetht t0THORN where k is Boltzmannrsquos constant andwhere T has the characteristics of a temperature as willbe seen It is a standard manipulation [135] to convertthe Langevin description in the large friction limit intoa FokkerndashPlanck equation
qwqt
frac14 1
mg qethfwTHORN
qrthorn kT
q2wqr2
eth28THORN
which represents the evolution of wethr tTHORN the prob-ability density that the molecule should lie at the radialposition r It should be noted in passing that FokkerndashPlanck equations have often been used to describe thestochastic dynamics of a cluster in size space [48 136] Incontrast the use of the method described above refers tomotion of a molecule in real coordinate space relative tothe centre of mass of the cluster
The steady state equilibrium solution of equation (28)is
wethrTHORN exp FethrTHORNkT
eth29THORN
where FethrTHORN is called the potential of mean force relatedto the mean force on the molecule f throughf frac14 dF=dr It can now be seen how the parameter Tin the stochastic force plays the role of temperaturesince equation (29) looks like a Boltzmann distributionThis solution is not what is sought however The escapeproblem is characterized by a boundary conditionwethreTHORN frac14 0 where re is the radius at which a particleescapes from the system The resulting escape rate maybe shown to be [134]
gkin frac14 KethreT m gTHORN exp DFkT
eth30THORN
where the prefactor K depends on the parameters listedas well as the shape of the potential of mean force Theprincipal feature of equation (30) is the exponentialdependence on the depth DF of the potential of meanforce Therefore if the potential of mean force FethrTHORN can
I J FORD894
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
be evaluated by studying a molecular dynamics trajec-tory for example then the kinetic decay rate can bedetermined This could be extended to account for aradially dependent temperature and friction coefficientboth of which might be determined from simulation Asketch of a potential of mean force is given in Fig 6
Kinetically determined cluster growth and decayrates gkini and bkini respectively could provide a newroute for the determination of nucleation rates inde-pendent of assumptions about detailed balance and theuse of equilibrium statistical mechanics in a non-equilibrium situation
9 CONCLUSIONS
This review has been brief and in some ways partisan tothe personal views of the author It attempts to sketch acoherent picture of nucleation theory and how thevarious disparate strands relate to one another One ofthe fascinations of nucleation theory is the variety ofapproaches and the number of difficult issues that areencountered How well does free energy fluctuationtheory represent a real non-equilibrium process Howwell does the idea of detailed balance provide aframework for determining non-equilibrium decaycoefficients from equilibrium statistical mechanicsHow well do various methods perform in the evaluation
of cluster free energies Can cluster growth and decayrates be calculated directly without the need for such aframework How exactly are clusters to be defined
The theoretical problems are many but the issues canbe resolved to various degrees of satisfaction Thereremains of course the fundamental task of choosing anintermolecular potential that adequately represents thephysical interactions Without this all the finer points ofcluster definition and treatments of non-equilibriumprocesses are unimportant Modelling many problems inscience relies on the availability of an accurate micro-scopic description and so nucleation rate calculationsare not unique in this respect Most empirical molecularpotentials are fitted to the properties of bulk and surfaceproperties and in the very best examples attention isgiven to their transferability to more unusual environ-ments such as clusters In view of these capabilities andmodern theoretical developments in nucleation theoryit seems likely that successful microscopic models of thenucleation process based on statistical physics are atlast coming within reach
ACKNOWLEDGEMENTS
The present author is grateful to colleagues SarahHarris and Marco Arosio for providing some of thefigures
Fig 6 A molecule in a cluster may be considered to be confined within it by a potential of mean force FethrTHORN Itcan escape by reaching a radial position re driven by a phenomenological stochastic force
STATISTICAL MECHANICS OF NUCLEATION 895
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
REFERENCES
1 Gross D H E Microcanonical Thermodynamics 2001
(World Scientific Singapore)
2 Seinfeld J H and Pandis S N Atmospheric Chemistry
and Physics 1998 (John Wiley New York)
3 Climate Change 2001 Synthesis Report 2001 (Inter-
governmental Panel on Climate Change) (Cambridge
University Press Cambridge)
4 Kohler T M Gail H-P and Sedlmayr E MgO dust
nucleation in M-stars calculation of cluster properties
and nucleation rates Astronomy and Astrophysics 1997
320 553
5 Kashchiev D Nucleation Basic Theory with Applications
2000 (ButterworthndashHeinemann Oxford)
6 Bakhtar F Ebrahimi M and Webb R A On the
performance of a cascade of turbine rotor tip section
blading in nucleating steam Part 1 surface pressure
distributions Proc Instn Mech Engrs 1995 209(C2)
115ndash124
7 Bakhtar F Ebrahimi M and Bamkole B O On the
performance of a cascade of turbine rotor tip section
blading in nucleating steam Part 2 wake traverses Proc
Instn Mech Engrs 1995 209(C3) 169ndash178
8 Bakhtar F andMahpeykar M R On the performance of
a cascade of turbine rotor tip section blading in nucleating
steam Part 3 theoretical treatment Proc Instn Mech
Engrs Part C J Mechanical Engineering Science 1997
211(C3) 195ndash210
9 Volmer M and Weber A Z Physikalische Chemie 1926
119 277
10 Farkas L Z Physikalische Chemie (Leipzig) 1927 125
236
11 Becker R and Doring W Kinetische Behandlung der
Keinbildung in ubersattigten Dampfen Ann Physik
(Leipzig) 1935 24 719
12 Zeldovich J Soviet Physics JETP 1942 12 525
13 Frenkel J Kinetic Theory of Liquids 1946 (Oxford
University Press Oxford)
14 Wagner P E and Strey R Homogeneous nucleation
rates of water vapor measured in a two-piston expansion
chamber J Phys Chemistry 1981 85 2694
15 Miller R Anderson R J Kassner J L and Hagen
D E Homogeneous nucleation rate measurements for
water over a wide range of temperature and nucleation
rate J Chem Physics 1983 78 3204
16 Viisanen Y Strey R and Reiss H Homogeneous
nucleation rates for water J Chem Physics 1993 99 4680
17 Viisanen Y Strey R and Reiss H Erratum homo-
geneous nucleation rates for water J Chem Physics
2000 112 8205
18 Wolk J and Strey R Homogeneous nucleation of H2O
and D2O in comparison the isotope effect J Phys
Chemistry B 2001 105 11683
19 Heath C H Streletsky K and Wyslouzil B E H2Ondash
D2O condensation in a supersonic nozzle J Chem
Physics 2002 117 6176
20 Mikheev V B Irving P M Laulainen N S Barlow
S E and Pervukhin V V Laboratory measurements of
water nucleation using a laminar flow chamber J Chem
Physics 2002 116 10 772
21 Hung C H Krasnopoler M J and Katz J L
Condensation of a supersaturated vapor VIII The
homogeneous nucleation of n-nonane J Chem Physics
1989 90 1856
22 Rudek M M Fisk J A Chakarov V M and Katz
J L Condensation of a supersaturated vapor XII The
homogeneous nucleation of the n-alkanes J Chem
Physics 1996 105 4707
23 Strey R Wagner P E and Schmeling T Homogeneous
nucleation rates for n-alcohol vapors measured in a two-
piston expansion chamber J Chem Physics 1986 84
2325
24 Lihavainen H Viisanen Y and Kulmala M Homo-
geneous nucleation of n-pentanol in a laminar flow
diffusion chamber J Chem Physics 2001 114 10 031
25 Rudek M M Katz J L Vidensky I V Zdımal V and
Smolık J Homogeneous nucleation rates of n-pentanol
measured in an upward thermal diffusion cloud chamber
J Chem Physics 1999 111 3623
26 Grassmann A and Peters F Homogeneous nucleation
rates of n-propanol in nitrogen measured in a piston-
expansion tube J Chem Physics 2002 116 7617
27 Kane D and El-Shall S Condensation of supersaturated
vapors of hydrogen bonding molecules ethylene glycol
propylene glycol trimethyl glycol and glycerol J Chem
Physics 1996 105 7617
28 Hameri K and Kulmala M Homogeneous nucleation in
a laminar flow chamber the effect of temperature and
carrier gas pressure on dibutyl phthalate vapor nucleation
rate at high supersaturations J Chem Physics 1996 105
7696
29 Viisanen Y Strey R Laaksonen A and Kulmala M
Measurement of the molecular content of binary nuclei
II Use of the nucleation rate surface for waterndashethanol
J Chem Physics 1994 100 6062
30 Viisanen Y Wagner P E and Strey R Measurement of
the molecular content of binary nuclei IV Use of the
nucleation rate surfaces for the n-nonane-n-alcohol series
J Chem Physics 1998 108 4257
31 Fisk J A Rudek M M Katz J L Beiersdorf D and
Uchtmann H The homogeneous nucleation of cesium
vapor Atmos Res 1998 46 211
32 Uchtmann H Dettmer R Baranovskii S D and
Hensel F Photoinduced nucleation in supersaturated
mercury vapor J Chem Physics 1998 108 9775
33 Ferguson F T and Nuth J A Experimental studies of
the vapor phase nucleation of refractory compounds V
The condensation of lithium J Chem Physics 2000 113
4093
34 Laaksonen A Talanquer V and Oxtoby D W
Nucleation measurements theory and atmospheric
applications A Rev Phys Chemistry 1995 46 489
35 Debenedetti P G Metastable Liquids Concepts and
Principles 1996 (Princeton University Press Princeton
New Jersey)
36 Yasuoka K and Matsumoto M Molecular dynamics of
homogeneous nucleation in the vapor phase I Lennardndash
Jones fluid J Chem Physics 1998 109 8451
37 Yasuoka K and Matsumoto M Molecular dynamics of
homogeneous nucleation in the vapor phase II Water
J Chem Physics 1998 109 8463
I J FORD896
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
38 Wonczak S Strey R and Stauffer D Confirmation of
classical nucleation theory by Monte Carlo simulations in
the 3-dimensional Ising model at low temperature
J Chem Physics 2000 113 1976
39 Laasonen K Wonczak S Strey R and Laaksonen A
Molecular dynamics simulations of gasndashliquid nucleation
of LennardndashJones fluid J Chem Physics 2000 113 9741
40 Reiss H Methods of Thermodynamics 1996 (Dover
Publications New York)
41 Kondepudi D and Prigogine I Modern Thermodynamics
From Heat Engines to Dissipative Structures 1998 (John
Wiley Chichester West Sussex)
42 Smoluchowski M Ann Phys NY 1908 25 225
43 Einstein A Ann Phys NY 1910 338 1275
44 Hirschfelder J O Curtis C F and Bird R BMolecular
Theory of Gases and Liquids 1954 (John Wiley New
York)
45 Landau L D and Lifshitz E M Statistical Physics 1958
(Pergamon London)
46 Wu D T Nucleation theory Solid State Physicsmdash
Advances in Res and Applic 1997 50 37
47 Reguera D Rubı J M and Perez-Madrid A Fokkerndash
Planck equations for nucleation processes revisited
Physica A 1998 259 10
48 Schenter G K Kathmann S and Garrett B Variational
transition state theory of vapor phase nucleation
J Chem Physics 1999 109 7951
49 Onsager L Phys Rev 1931 37 405
50 Courtney W G J Chem Physics 1961 35 2249
51 Girshick S L and Chiu C-P Kinetic nucleation theory
a new expression for the rate of homogeneous nucleation
from an ideal supersaturated vapor J Chem Physics
1990 93 1273
52 Ford I J Barrett J C and Lazaridis M Uncertainties
in cluster energies in homogeneous nucleation theory
J Aerosol Sci 1993 24 581
53 Weakliem C L and Reiss H The factor 1S in the
classical theory of nucleation J Phys Chemistry 1994
98 6408
54 McClurg R B and Flagan R C Critical comparison of
droplet models in homogeneous nucleation theory
J Colloid Interface Sci 1998 201 194
55 Mashmoushy H Mahpeykar M R and Bakhtar F
Studies of nucleating and wet steam flows in two-dimen-
sional cascades Proc Instn Mech Engrs Part C J
Mechanical Engineering Science 2004 218(C8) 843ndash858
56 Ford I J Nucleation theorems the statistical mechanics
of molecular clusters and a revision of classical nuclea-
tion theory Phys Rev E 1997 56 5615
57 ten Wolde P R and Frenkel D Computer simulation
study of gasndashliquid nucleation in a LennardndashJones system
J Chem Physics 1998 109 9901
58 Kusaka I Wang Z G and Seinfeld J H Direct
evaluation of the equilibrium distribution of physical
clusters by a grand canonical Monte Carlo simulation
J Chem Physics 1998 108 3416
59 Kusaka I and Oxtoby D W Identifying physical
clusters in vapor phase nucleation J Chem Physics
1999 110 5249
60 Stillinger F H J Chem Physics 1963 38 1486
61 Ford I J Imperfect vaporndashgas mixtures and homo-
geneous nucleation J Aerosol Sci 1992 23 447
62 Lothe J and Pound G M J Chem Physics 1962 36
2080
63 Lee J K Barker J A and Abraham F F Theory and
Monte Carlo simulation of physical clusters in the
imperfect vapor J Chem Physics 1973 58 3166
64 Abraham F F Homogeneous Nucleation Theory 1974
(Academic Press New York)
65 Reif F Fundamentals of Statistical and Thermal Physics
1965 (McGraw-Hill New York)
66 Hoare M R and Pal P Adv Phys 1975 24 645
67 Hoare M R Adv Chem Physics 1979 40 49
68 Pal P and Hoare M R Thermodynamic properties
and homogeneous nucleation of molecular clusters of
nitrogen J Phys Chemistry 1987 91 2474
69 Reiss H Katz J L and Cohen E R Translationndash
rotation paradox in the theory of nucleation J Chem
Physics 1968 48 5553
70 Hale B N Monte Carlo calculations of effective surface
tension for small clusters Aust J Physics 1996 49 425
71 Kathmann S M Schenter G K and Garrett B C
Dynamical nucleation theory calculation of condensation
rate constants for small water clusters J Chem Physics
1999 111 4688
72 Reiss H Tabazadeh A and Talbot J Molecular theory
of vapor phase nucleation the physically consistent
cluster J Chem Physics 1990 92 1266
73 Ellerby H M Weakliem C L and Reiss H Toward a
molecular theory of vaporndashphase nucleation I Identifica-
tion of the average embryo J Chem Physics 1992 97
5766
74 Ellerby H M and Reiss H Toward a molecular theory
of vaporndashphase nucleation II Fundamental treatment of
the cluster distribution J Chem Physics 1992 97 5766
75 Weakliem C L and Reiss H Toward a molecular theory
of vaporndashphase nucleation III Thermodynamic properties
of argon clusters fromMonte Carlo simulations and modi-
fied liquid drop theory J Chem Physics 1994 101 2398
76 Kusaka I Wang Z G and Seinfeld J H Binary
nucleation of sulfuric acidndashwater Monte Carlo simula-
tion J Chem Physics 1998 108 6829
77 Oh K J and Zeng X C A small system ensemble
Monte Carlo simulation of supersaturated vapor
evaluation of barrier to nucleation J Chem Physics
2000 112 294
78 Senger B Schaaf P Corti D S Bowles R Voegel J
C and Reiss H A molecular theory of the homogeneous
nucleation rate I Formulation and fundamental issues
J Chem Physics 1999 110 6421
79 Hill T L Statistical Mechanics Principles and Selected
Applications 1956 (McGraw-Hill New York)
80 Soto R and Cordero P Nonideal gas of clusters at
equilibrium J Chem Physics 1999 110 7316
81 Barrett J C The significance of cluster lifetime in
nucleation theory J Chem Physics 2002 116 8856
82 Lushnikov A A and Kulmala M Dimers in nucleating
vapors Phys Rev E 1998 58 3157
83 Bahadur R and McClurg R B Comparison of cluster
definitions in homogeneous nucleation rate calculations
J Phys Chemistry 2001 105 11 893
84 Pugnaloni L A and Vericat F New criteria for cluster
identification in continuum systems J Chem Physics
2002 116 1097
STATISTICAL MECHANICS OF NUCLEATION 897
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
85 Chen B Siepmann J I Oh K J and Klein M L
Simulating vaporndashliquid nucleation of n-alkanes J Chem
Physics 2002 116 4317
86 Harris S A and Ford I J A dynamical definition of
quasi-bound molecular clusters J Chem Physics 2003
118 9216
87 Shreve A P Walton J R P B and Gubbins K E
Liquid drops of polar molecules J Chem Physics 1986
85 2178
88 ten Wolde P R Ruiz-Montero M J and Frenkel D
Numerical calculation of the rate of homogeneous gasndash
liquid nucleation in a LennardndashJones system J Chem
Physics 1998 109 9901
89 ten Wolde P R Oxtoby D W and Frenkel D Chain
formation in homogeneous gasndashliquid nucleation of polar
fluids J Chem Physics 1998 111 4762
90 Vehkamaki H and Ford I J Critical cluster size and
droplet nucleation rate from growth and decay simula-
tions of LennardndashJones clusters J Chem Physics 2000
112 4193
91 Schaaf P Senger B Voegel J C Bowles R K and
Reiss H Simulative determination of kinetic coefficients
for nucleation rates J Chem Physics 2001 114 8091
92 Chen B Siepmann J I Oh K J and Klein M L
Aggregation-volume-bias Monte Carlo simulations of
vaporndashliquid nucleation barriers for LennardndashJonesium
J Chem Physics 2001 115 10 903
93 Hale B N and DiMattio D J A Monte Carlo discrete
sum (MCDS) nucleation rate model for water In
Proceedings of the 15th International Conference on
Nucleation and Atmospheric Aerosols AIP Conference
Proceedings Vol 534 (Eds B N Hale and M Kulmala)
2000 pp 31ndash34
94 Kathmann S M and Hale B N Monte Carlo
simulations of small sulfuric acidndashwater clusters
J Phys Chemistry B 2001 105 11 719
95 Daura X Mark A E and van Gunsteren W F Para-
metrization of aliphatic CHn united atoms of GROMOS96
force field J Comput Chemistry 1998 19 535
96 Kathmann S M Schenter G K and Garrett B C
Understanding the sensitivity of nucleation kinetics a
case study on water J Chem Physics 2002 116 5046
97 Kashchiev D On the relation between nucleation work
nucleus size and nucleation rate J Chem Physics 1982
76 5098
98 Oxtoby D W and Kashchiev D A general relation
between the nucleation work and the size of the nucleus in
multicomponent nucleation J Chem Physics 1994 100
7665
99 Ford I J Thermodynamic properties of critical clusters
from measurements of vapourndashliquid homogeneous
nucleation rates J Chem Physics 1996 105 8324
100 Oxtoby D W and Evans R Nonclassical nucleation
theory for the gasndashliquid transition J Chem Physics
1988 89 7521
101 Zeng X C and Oxtoby D W Gasndashliquid nucleation in
LennardndashJones fluids J Chem Physics 1991 94 4472
102 Evans R Fundamentals of Inhomogeneous Fluids 1992
Ch 3 pp 85ndash175 (Marcel Dekker New York)
103 Talanquer V and Oxtoby D W Dynamical density
functional theory of gasndashliquid nucleation J Chem
Physics 1994 100 5190
104 Barrett J C First-order correction to classical nucleation
theory a density functional approach J Chem Physics
1999 111 5938
105 Barrett J C Cluster translation and growth in density
functional theories of homogeneous nucleation J Chem
Physics 1997 107 7989
106 Bykov T V and Zeng X C A patching model for
surface tension of spherical droplet and Tolman length J
Chem Physics 1999 111 10 602
107 Talanquer V and Oxtoby D W Nucleation in molecular
and dipolar fluids interaction site model J Chem
Physics 1995 103 3686
108 Seok C and Oxtoby D W Nucleation in n-alkanes a
density functional approach J Chem Physics 1998 109
7982
109 Reguera D Bowles R K Djikaev Y and Reiss H
Phase transitions in systems small enough to be clusters
J Chem Physics 2003 118 340
110 Khakshouri S and Ford I J J Chem Physics 2004
(submitted)
111 Ford I J (unpublished)
112 Nyquist R M Talanquer V and Oxtoby D W Density
functional theory of nucleation a semi-empirical
approach J Chem Physics 1995 103 1175
113 Talanquer V and Oxtoby D W Density functional
analysis of phenomenological theories of gasndashliquid
nucleation J Chem Physics 1995 99 2865
114 Katz J L Fisk J A and Chakarov V The accuracy of
nucleation theory In Proceedings of the 13th Interna-
tional Conference on Nucleation and Atmospheric Aerosols
(Eds N Fukuta and P E Wagner) 1992 pp 11ndash18
(Deepak Publishing Hampton Virginia)
115 Fisher M E The theory of condensation and the critical
point Physics 1967 3 255
116 Dillmann A and Meier G E A A refined droplet
approach to the problem of homogeneous nucleation
from the vapor phase J Chem Physics 1991 94 3872
117 Tolman R C J Chem Physics 1949 17 333
118 Ford I J Laaksonen A and Kulmala M Modification
of the DillmannndashMeier theory of homogeneous nuclea-
tion J Chem Physics 1993 99 764
119 Laaksonen A Ford I J and Kulmala M Revised
parametrization of the DillmannndashMeier theory of homo-
geneous nucleation Phys Rev E 1994 49 5517
120 Kalikmanov V I and van Dongen M E H Semipheno-
menological theory of homogeneous vaporndashliquid nuclea-
tion J Chem Physics 1995 103 4250
121 Hale B N Application of a scaled homogeneous
nucleation-rate formalism to experimental-data at
T 5Tc Phys Rev A 1986 33 4156
122 Martinez D M Ferguson F T Heist R H and Nuth
J A Application of scaled nucleation theory to metallic
vapor condensation J Chem Physics 2001 115 310
123 Hale B N and Wilemski G A scaled nucleation model
for ideal binary systems Chem Physics Lett 1999 305
263
124 Granasy L Diffuse interface theory for homogeneous
vapor condensation J Chem Physics 1996 104
5188
125 McGraw R and Laaksonen A Interfacial curvature free
energy the Kelvin relation and vaporndashliquid nucleation
rate J Chem Physics 1997 106 5284
I J FORD898
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
126 Talanquer V A new phenomenological approach to
gasndashliquid nucleation based on the scaling properties
of the critical nucleus J Chem Physics 1997 106
9957
127 Koga K and Zeng X C Thermodynamic expansion of
nucleation free-energy barrier and size of critical nucleus
near the vaporndashliquid coexistence J Chem Physics 1999
110 3466
128 Kusaka I On the scaling behavior of the free energetics of
nucleation J Chem Physics 2003 118 5510
129 Nowakowski B and Ruckenstein E A kinetic approach
to the theory of nucleation in gases J Chem Physics
1991 94 1397
130 Nowakowski B and Ruckenstein E Homogeneous
nucleation in gases a three-dimensional FokkerndashPlanck
equation for evaporation from clusters J Chem Physics
1991 94 8487
131 Wilcox C F and Bauer S H Estimation of homo-
geneous nucleation flux via a kinetic model J Chem
Physics 1991 94 8302
132 Hanggi P Talkner P and Borkovec M Reaction rate
theory fifty years after Kramers Rev Mod Physics 1990
62 251
133 Barrett J C Clement C F and Ford I J Energy
fluctuations in homogeneous nucleation theory for aero-
sols J Physics A 1993 26 529
134 Ford I J and Harris S A Molecular cluster decay
viewed as escape from a potential of mean force J Chem
Physics 2004 120 4428
135 Risken H The FokkerndashPlanck Equation Methods of
Solution and Applications 2nd edition 1989 (Springer-
Verlag Berlin)
136 Shizgal B and Barrett J C Time dependent nucleation
J Chem Physics 1989 91 6505
STATISTICAL MECHANICS OF NUCLEATION 899
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
microscopic point of view is very laborious and has beenrealized in only a very few example cases [36ndash39] Thetime scales involved make a nucleation event simply tooslow to follow directly for realistic conditions and withcurrent computational resources
The simplest statistical treatment of the nucleationprocess comes from TST within a framework of thetheory of free energy fluctuations According to thermo-dynamics a system with externally fixed temperatureand pressure (or volume) will adjust any internal prop-erties in order to minimize its Gibbs (or Helmholtz) freeenergy [40 41] Such adjustments will include changes inphase Nevertheless statistical fluctuations in systemproperties will naturally take place and it is useful tointerpret these as fluctuations in the free energy of thesystem away from the minimum Considering constant-volume conditions for the moment general arguments[42ndash45] suggest that the probability of a fluctuationDF in Helmholtz free energy is proportional toexpfrac12DF=ethkTTHORN
The physical fluctuations that are important in thenucleation process are those where a molecular clusterof new phase is formed temporarily from the originalphase The change in free energy accompanying theformation of a cluster of i molecules in the new phase iswritten DF frac14 DWethiTHORNFethiTHORN if0 where FethiTHORN is the freeenergy of the i cluster and f0 is the free energy permolecule in the old phase This is also called the lsquoworkof formationrsquo since it corresponds in thermodynamics tothe external work needed to insert an additional clusterreversibly into the system while maintaining the overall
temperature and pressure (or volume) Note that thisinsertion of a new cluster while maintaining suchconstraints is a theoretical operation since it would behard to achieve in practice
For a metastable initial phase the magnitude of thefree-energy fluctuation DF peaks at the so-called lsquocriticalsizersquo i This is illustrated in Fig 1 The perturbed statewith the additional critical cluster is therefore athermodynamic transition state since reversible changesin size of the extra cluster are accompanied byreductions in free energy It is presumed that this meansthat the new cluster has equal probabilities of growingor decaying and that the rate of nucleation is thereforerelated to the rate of formation of these critical clusters
The probability of forming the critical clusteraccording to this picture is proportional toexpfrac12DWethi THORN=ethkTTHORN but the proportionality constant(essentially a time scale) is not yet determined Tointroduce this simple kinetic theory [46] can be used tocalculate the collision rate bi between single molecules(monomers) and an i-cluster modelled as a sphere withbulk condensed phase molecular density rl
bi frac14qpv
eth2pmkTTHORN1=21thorn 1
i
1=2
1thorn i1=3 2
eth36pTHORN1=3
r2=3l
eth3THORN
where m is the molecular mass It is more usual toreplace the expression in square brackets with i2=3 thelimit for large i The simplest TST expression for the
Fig 1 Sketch of the cluster work of formation for a range of (constrained) cluster sizes The maximum in thecurve defines the critical size i and the critical work of formation DWethi THORN
I J FORD886
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
nucleation rate is then J bini where ni is the meannumber of critical clusters in the system
More sophisticated treatments of thermodynamicscan provide the time scale without recourse to the abovekinetic argument which is reasonably valid for gas-to-particle conversion but not for freezing for exampleAttempts based on linearized non-equilibrium thermo-dynamics [47 48] originate from the work of Onsager[49]
The next step is crucial and it is to relate thepopulation ni to the probability of a fluctuation and toan explicit model of the critical work of formation Theassumption in classical nucleation theory is that
ni frac14 n1 expDWclethi THORN
kT
eth4THORN
where DWclethiTHORN is the classical work of formation givenby
DWclethiTHORN frac14 4pR2s ikT lnS eth5THORN
and DWclethi THORN is the maximum of this function Thecluster is envisaged to be a sphere of radius R given by4pR3rl=3 frac14 i
Unfortunately some uncertainty accompanies thesesteps in the derivation The theory is flawed on severalcounts The work of formation contains a termassociated with the creation of a spherical interface(characterized by the bulk value of surface tension) anda change kT lnS in free energy per molecule due tothe change in phase from supersaturated vapour tocondensate Moreover the factor n1 appearing inequation (4) appears to have been introduced heuristi-cally Might it not instead be the monomer populationin the saturated vapour How does DWclethiTHORN comparewith the actual work of formation DWethiTHORN Is thedroplet really spherical with a sharp interface Have
any important factors been left out Furthermoreshould there be concern that equation (4) is in-consistent when i frac14 1 This has been a controversialissue [50ndash54]
3 KINETIC INTERPRETATION
The expression for the classical rate of nucleation givenin equation (1) emerges from the models in equation (4)and (5) when the thermodynamic transition stateanalysis is embellished by a kinetic interpretation [11]Instead of using the probability of fluctuations awayfrom the equilibrium state (the TST route) a set of rateequations is introduced to describe the dynamics ofcluster populations
The evolution of cluster populations is modelled usingthe following equations
dni
dtfrac14Xj
njPji niXj
Pij eth6THORN
where Pji is the mean rate at which a j cluster convertsinto an i cluster The transition processes are assumed tobe Markovian so that the rate coefficients depend onthe properties of the cluster and its environment but noton their history
The kinetic interpretation can be taken a step furtherby assuming that the only important transitions arethose involving the addition or loss of single moleculesto or from the cluster The scheme of transitionprocesses is illustrated in Fig 2 The only non-zerorate coefficients are then bi frac14 Piithorn1 for cluster growthand gi frac14 Pii1 for cluster decay The rate equations
Fig 2 The scheme of the birth and death of clusters addition of a monomer to a cluster will take it up thechain towards the right while the loss of a monomer will take it in the opposite direction Attachmentand detachment of dimers and larger clusters are usually ignored
STATISTICAL MECHANICS OF NUCLEATION 887
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
reduce to
dni
dtfrac14 bi1ni1 gini bini thorn githorn1nithorn1 eth7THORN
These are the BeckerndashDoring equations [11] and theyprovide a useful reinterpretation of the TST approachThere are two steady state solutions In the first thepopulations satisfy equation (7) and also the detailedbalance condition
bi1nei1 frac14 gin
ei eth8THORN
and the fnei g are taken to represent the cluster popula-tions in an equilibrium vapour at or below the saturatedvapour pressure For true thermal equilibrium themonomer population has to be less than or equal to itsvalue for the saturated vapour ns1 The cluster populationsin a saturated vapour are denoted fnsig Nevertheless asituation can be invented which is known as lsquoconstrainedequilibriumrsquo where the detailed balance condition is stillsatisfied but with n1 gt ns1 Since bi pvS while gi isindependent of vapour pressure equation (8) then definesso-called constrained equilibrium populations
ncei ampSinsi eth9THORN
for a supersaturated vapour It has been assumed thatS frac14 pv=p
svampn1=n
s1
The second steady state solution to the BeckerndashDoring equations is more realistic It is not meant torepresent thermal equilibrium The supersaturation isset to a value greater than unity and a boundarycondition of zero population at some cluster size imax isimposed These conditions will drive a steady productionrate of large clusters from the vapour The approach isfor mathematical convenience only and the same steadystate cluster populations nssi and nucleation rate wouldemerge if imax was sent off to infinity The detailedbalance condition (8) is replaced by J frac14 bi1n
ssi1 gin
ssi
defining the nucleation current J which can then becalculated analytically as shown in another paper in thisvolume [55] To a good approximation
J frac14 Zdbincei eth10THORN
which involves the constrained equilibrium population atthe critical size i the size with the lowest constrainedequilibrium population The condition imax 4 i musthold Zd is the so-called Zeldovich factor [12] givenapproximately by
Zd frac14 1
2pq2ethln ncei THORN
qi2
ifrac14i
1=2
eth11THORN
When equations (4) and (5) are used to specify ncei and theresult is substituted into equations (10) and (11) then theclassical nucleation rate given in equation (1) isrecovered as long as the large i limit of the attachment
rate coefficient in equation (3) is taken It is also possibleto derive an expression for J purely in terms of the ratecoefficients bi and gi This possibility is considered againin section 8
The main assumptions used in this derivation of theclassical theory principally the validity of equations (4)and (5) are now examined In order to explore this it isnecessary to see how the equilibrium cluster populationsnei in subsaturated vapours may be calculated usingstatistical mechanics
4 CLUSTER STATISTICAL MECHANICS
The statistical mechanical treatment of a vapour is bestdeveloped using a grand canonical ensemble [56ndash59] Thesystem is placed in contact with a particle reservoir at achemical potential and temperature T so that thenumber of molecules N in the system can change Thegrand partition function of the system X is then propor-tional to the integral of expf frac12HethNTHORN N=ethkTTHORNg overall possible configurations of molecular positions andmomenta whereHethNTHORN is the Hamiltonian of the system
Evaluating X exactly is exceedingly difficult and it ismuch simpler to proceed by considering the configura-tions of the system as arrangements of molecularclusters The vapour then is viewed as a mixture ofideal gases each composed of clusters of differentsizes The grand partition function of the system canbe reconstructed from partition functions for singlei-clusters In order to do this however a cluster must bedefined in a unique way in terms of molecular positionsand velocities An example of such a definition would beto require the separation between molecules in a clusterto be less than a maximum distance Rc [60] as illustratedin Fig 3
Fig 3 According to the Stillinger definition a molecule isconsidered part of a cluster if it lies closer than adistance Rc from a constituent molecule of the cluster
I J FORD888
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
At first sight the cluster definition is completelyarbitrary a means of categorizing and resumming allthe elementary configurations of the system Howeveraccurately representing the system as a mixture of idealgases requires the interactions between molecules lyingin different clusters to be minimized The clusterdefinition affects whether a particular interaction isregarded as intercluster or intracluster and so it plays acrucial role This will be discussed further in section 6Residual clusterndashcluster interactions can be taken intoaccount in the form of a virial series [61]
These ideas are now put into practice Define amodified canonical partition function zi for a clustercontaining i molecules This is simply the integral ofexpf frac12HcethiTHORN i=ethkTTHORNg over configurations involvingmolecules in the cluster alone with Hc the Hamiltonianof the i molecules comprising the cluster Using thepicture of a mixture of ideal gases the grand partitionfunction of the complete system can then be expressed as
XampXfnig
Yifrac141
znii
ni
eth12THORN
where the sum is over all possible cluster populationdistributions fnig This is demonstrated rigorously inappendix A of reference [57]
The next step is to determine the distribution fnei gthat dominates the sum in equation (12) The logarithmof the summand is extremized
qqni
Xifrac141
lnz
nii
ni
frac14 0 eth13THORN
which gives the following very simple expression for themost probable or equilibrium cluster size distributionfor the given conditions T and
nei frac14 ziethTTHORN frac14 ZethTTHORN ei=ethkTTHORN eth14THORN
which is equivalent to the so-called law of mass action
nei frac14 Zine1Z1
i
eth15THORN
where ZiethTTHORN frac14 expfrac12FethiTHORN=ethkTTHORN frac14 zi expfrac12 i=ethkTTHORN isthe canonical partition function for an i cluster and FethiTHORNis the i-cluster Helmholtz free energy This relationshipis often used as the starting point in models of clusterpopulations
The vapour pressure is now given by a sum of partialpressures
pvV frac14 kTXifrac141
nei frac14 kTXifrac141
ziethTTHORN eth16THORN
and this series will only converge if the equilibriumpopulations or the zi decrease sufficiently quickly forlarge i This limits the statistical mechanics approach to
the study of saturated or subsaturated vapours How-ever as in the preceding section the equilibriumpopulations nei in equation (14) (for S lt 1) can beextrapolated into constrained equilibrium populationsncei (with S gt 1) for use in equation (10) Saturatedvapour corresponds to a particular chemical potentialms and the saturated vapour cluster populations are nsi The constrained equilibrium populations are obtainedby inserting into equation (14) the chemical potential ofthe supersaturated vapour which is frac14 s thorn kT lnSgiving ncei frac14 Sinsi which is consistent with the resultobtained in equation (9) for dilute vapours
The constrained equilibrium populations are there-fore expressed as
ncei frac14 SiziethsTTHORN
frac14 expfrac12FethiTHORN is ikT lnS
kT
eth17THORN
and the correspondence with CNT can be made closerby introducing the n1 prefactor
ncei frac14 n1 exp frac12FethiTHORN is thorn kT ln n1 ikT lnS
kT
eth18THORN
The next task is to explore how the combinationFethiTHORN is thorn kT ln n1 might be approximated by theCNT result 4pR2s This can be explored most easilyfor solid rather than liquid clusters
5 SOLID CLUSTERS
The partition function of a solid cluster of i moleculescan be factorized into translational rotational andvibrational parts through a transformation to suitablerelative positions and momenta [62ndash64] Starting withthe vibrational degrees of freedom the atom positionsare defined with respect to their mean positions whichare determined by a fixed centre of mass and set oflattice vectors It is assumed that the potential energypart of the Hamiltonian is quadratic in the displace-ments from these mean positions Now the partitionfunction is worked out for configurations of the particlesrestricted to having zero total linear and total angularmomentum about the centre of mass This is writtenZvib corresponding to a product of partition functionsfor 3ji 6 independent oscillators (if there are j atomsper molecule and therefore a total of 3ji degrees offreedom if the molecule is not linear) each of which inthe classical limit takes the form kT=ethhoTHORN where o isthe angular frequency of the specific vibrational modeThe partition function has an additional factor ofexpfrac12U0ethiTHORN=ethkTTHORN where U0ethiTHORN is the system potentialenergy evaluated at the mean atom positions A
STATISTICAL MECHANICS OF NUCLEATION 889
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
different expression applies for low temperatures whenthe oscillator is not classically activated [65]
The rotational and translational degrees of freedomare next The previous restriction of the configurationsto zero angular momentum is lifted The approximate(classical) rotational partition function is
Zrot frac14 p1=28p2kTh2
3=2
ethI1I2I3THORN1=2 eth19THORN
where Ik are the principal moments of inertia about thecentre of mass for atoms fixed in their mean positions
Finally the translational free energy of the centre ofmass of the cluster is given by the partition function
Ztrans frac14 V2pMkT
h2
3=2
eth20THORN
where V is the system volume and M frac14 im is the clustermass The free energy of the cluster (plus kT ln n1 forconvenience) is therefore
FethiTHORN thorn kT ln n1
frac14 kT lnZtransZrotZvib
n1
frac14 U0ethiTHORN kT ln
eth2pMkTTHORN3=2
Srsvh3p1=2
8p2kTh2
3=2
6ethI1I2I3THORN1=2Y3ji6
kfrac141
kT
hok
eth21THORN
where rsv frac14 ns1=V is the molecular density of dilutesaturated vapour
By similar arguments the chemical potential of dilutesaturated vapour is
s frac14 Um0 kT ln
eth2pmkTTHORN3=2
rsvh3p1=2
8p2kTh2
3=2
6ethIm1 Im2 Im3 THORN1=2Y3j6
kfrac141
kT
homk
eth22THORN
where Um0 is the potential energy of a molecule when the
atoms are at their mean positions Imk are the threeprincipal moments of inertia of the molecule about thecentre of mass and om
k are the oscillatory frequencies ofmolecular internal vibrations
The classical work of formation DWclethiTHORN should nowcorrespond to the leading-order contribution to thedifference between equation (21) and i times equation(22) Calculations [54 64 66ndash68] demonstrate that theleading term is indeed roughly proportional to i2=3 thusjustifying the form of the classical model The suitabilityof the capillarity approximation depends on the extentto which the coefficient of this leading term correspondsnumerically to the planar surface free energy Howeverthe full statistical mechanical expression for the work of
formation contains additional terms and these will beimportant for small clusters and need to be taken intoaccount In order to evaluate them the variousstructural and energy properties of cluster are requiredall of which can be obtained from microscopic modelsand they can then be inserted into the formulae for thethermodynamic properties [equations (21) and (22)] theconstrained equilibrium populations [equation (18)] andultimately the nucleation rate [equation (10)] Followingthis route employs the almost universally held assump-tion that equilibrium properties such as free energies canbe used to describe non-equilibrium properties specifi-cally the cluster decay rate This assumption is re-examined in section 8
6 LIQUID DROPLETS AND THE CLUSTER
DEFINITION
Defining what was meant by a solid cluster in theprevious section was not an issue Within the assump-tions of the model (harmonic interactions) it wasimpossible for a molecule to escape from the clusterThe cluster definition in the case of liquids cannot bedismissed so easily Liquids are more volatile thansolids and the molecules in a cluster are much moremobile It is imperative now to characterize a clusterusing the molecular positions and velocities andperhaps some time scale and length scale too
The traditional microscopic method for defining acluster is to impose a geometric constraint on molecularpositions An example is to restrict molecules to a spherecentred on the centre of mass [69ndash71] or to impose amaximum allowable separation between the molecules(the Stillinger criterion [60]) More sophisticated treat-ments exist such as the n=v Stillinger cluster introducedby Reiss and co-workers [72ndash75] Here the confiningvolume v used to define the cluster is considered to be anadditional defining characteristic on the same level asthe cluster size n As time progresses a cluster changesits confining volume as the component molecules followtheir respective trajectories just as they change their sizewith the gain and loss of molecules The association of amolecular configuration with a precise confining volumeis complicated but tractable The dependence of the freeenergy on v determines the statistical dynamics of thiscluster volume A similar attempt to associate a clusterwith a specified volume within a grand canonicalensemble description was developed by Kusaka andco-workers [58 59 76] The simulation box was taken tobe the confining cluster volume Many of these ideasmay be viewed as a treatment of the cluster definition(eg the Stillinger radius) as a variational parameter inthe representation of the partition function Further
I J FORD890
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
refinement of this point of view using a treatment ofclusterndashcluster interactions to optimize the definingcluster volume has been presented [77] A review ofrecent progress in this area and the application of theseideas to nucleation theory has been provided by Sengeret al [78]
Such geometric constraints can be implemented buttheir realism is questionable Situations better regardedas close encounters between separate clusters would beclassed as snapshots of single larger clusters It shouldbe recalled that the cluster definition is an internalfeature of the model which is to be optimized to reduceclusterndashcluster interactions and to improve the repre-sentation of the grand partition function Misclassifyingconfigurations is fraught with consequences A secondissue to consider is that for convenience the clustersshould behave in a manner that can be modelled withBeckerndashDoring equations This means that the growthand decay rates should be free of memory effects ie thetransition processes should be Markovian The extent towhich this is satisfied (if indeed it is physically realistic)depends on the cluster definition
An intuitively reasonable choice of definition whichaddresses both these points is that a cluster shouldconsist of molecules that are bound energetically Thisshould minimize the clusterndashcluster contributions to thetotal energy According to such a definition if amolecule interacts strongly with a cluster then it islikely to be regarded as part of that cluster Also anenergetically bound cluster is likely to decay by thermalfluctuation in a stochastic Markovian fashion ratherthan by the separation of passing poorly boundfragments
Ideas have been proposed which require the cluster tobe energetically bound (in a loose sense since a cluster isintrinsically prone to decay) The time evolution of amolecular configuration and not simply the set ofinstantaneous positions of its component molecules isof central importance Hill [79] introduced a clusterdefinition requiring the total energies of pairs ofmolecules in a cluster to be negative (in their centre ofmass frame) The intention was to exclude situationswhere a molecule could leave the cluster within the timeit would take for a molecule to travel a typical lineardimension of the cluster Soto and Cordero [80] havedeveloped Hillrsquos criterion and Barrett [81] has per-formed Monte Carlo modelling of molecular configura-tions excluding those which decay in moleculardynamics within a given period A number of otherenergy-based criteria have been developed [82ndash85]
Harris and Ford [86] have recently proposed a newenergy-based cluster definition As with Barrettrsquosapproach it is based on the requirement that amolecular snapshot is energetically bound for a chosensubsequent period Evaporation according to thisscheme requires not only that a molecule should makea positive energy excursion but also that the subsequent
dynamics carry it far away from the cluster Onlyperhaps one in eight molecules manage to avoidrecapture once having satisfied the positive energycriterion [86] A snapshot of a cluster in this casesurrounded by a carrier gas is shown in Fig 4 Thehistory of the energy of a molecule in such a cluster isillustrated in Fig 5 three excursions into positive energyoccur in this sequence but only one leads to actualescape corresponding to the maintenance of positiveenergy for a significant time The instant of escape istaken to be the point at which the energy goes positivefor the third and final time
Cluster definitions implemented in moleculardynamics and Monte Carlo codes allow cluster freeenergies and hence the decay rate to be calculated Anenormous amount of simulation work has been carriedout and calculations can be laborious Many differentschemes and approximations have been made to easethe evaluation of these properties The principalconclusion from various studies using a range of clusterdefinitions and interaction models [49 57 58 70 85 87ndash92] seems to be that the work of formation of liquidcritical clusters does indeed contain a dominantcontribution proportional to i2=3 consistent with theclassical work of formation Hale and collaboratorspresent cluster free energies for water and other speciesin a way that makes this particularly clear indeed it issurprising how dominant the surface term can be foreven very small clusters [70 93 94] Neverthelessfurther contributions exist small clusters are notcharacterized just by scaled-down properties of largeclusters and it is impossible to escape the need todetermine the additional terms
In all the above discussion it has been assumed thatthe intermolecular interactions are accurately specifiedThis is a situation that is rarely possible Althoughfirst-principles quantum-mechanical models could be
Fig 4 Snapshot of a cluster of 50 argon atoms (brightspheres) bathed in a gas of helium (darker spheres)
STATISTICAL MECHANICS OF NUCLEATION 891
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
developed in practice empirical potentials are devel-oped on the basis of physical intuition containingparameters fitted to the measured properties of thesubstance in question [95] Usually these are bulkproperties and it is not always clear that the sameinteractions can be used for the very different circum-stances of a molecular cluster [96] In view of thesensitivity of nucleation rates to the thermodynamicproperties of clusters this is perhaps a crucial failingProgress can be made however by computing knowncluster properties where they are available and byrefining the potentials to suit Extraction of clusterproperties through use of the so-called nucleationtheorems (which relate derivatives of the nucleationrate to properties of the critical cluster [16 56 97ndash99])may provide just this sort of experimental information
A technique that has not yet been described is the useof density functional theory (DFT) to model theproperties of liquid droplets [100ndash104] This powerfulapplication of statistical mechanics promises to avoidlaborious computations at the microscopic level byinstead focusing on the density profile of moleculesmaking up the droplet This profile may be regarded asthe average distribution of molecules about the clustercentre of mass Although this might appear to involvecoarse-graining and approximation it is a well-foundedapproach and in principle can be exact In practical
implementations however a mean field approach istaken such that the constituent molecules move in aneffective potential well created by the other molecules inthe cluster [105] Also the interactions are assumed to becharacterized by strong repulsive cores and weaklyattractive tails
Considerable analytical work has been carried out[104 106] Various developments have been made todescribe droplets consisting of simple atoms as well aschain-like molecules [107] and polar molecules [108]The main drawback in these implementations is thatspecific interactions such as hydrogen bonds are hardto represent Certain details of the modelled systemare neglected [109 110] and the (implicit) clusterdefinition used is geometric and not energetic [111]On the other hand quite sophisticated systems havebeen modelled including multi-component dropletsThe method has been compared with otherapproaches [112 113] and certain features are betterdescribed particularly the approach to spinodalconditions where the nucleation barrier vanishesDFT calculations can be performed relatively easilyand they can provide a qualitative and perhapsquantitative insight into quite complicated clusteringproblems They provide an intermediate-level descrip-tion between the simple capillarity approximation anda microscopic description
Fig 5 The energy of a molecule in the centre-of-mass frame of a cluster occasionally becomes positive butthis does not necessarily lead to escape Only the third positive energy excursion in this sequenceindicated by an arrow results in particle escape A cluster definition must capture this behaviour
I J FORD892
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
7 PHENOMENOLOGICAL MODELS
The reader might be dismayed by the mathematicalcomplexity of the expressions for the work of formationof solid clusters in section 5 might be alarmed by theprocedural complexity of implementing a cluster defini-tion in section 6 may recoil from the difficulties incalculating free energies from molecular simulationmethods and ultimately may give up in view of theuncertainties in the underlying intermolecular interac-tions Considering these complications it is tempting totry to develop phenomenological models of the work offormation After all if the capillarity approximationworks reasonably well for water then surely a minorcorrection to account for small cluster effects ought tobe sufficient to improve matters
However after perhaps two decades of work nomodel has emerged offering broad success across arange of different substances It is perhaps asking toomuch to expect universal improvement CNT works as afirst approximation for water for a particular tempera-ture range but as discussed earlier it possesses thewrong temperature dependence so that predictionsaway from a temperature in the range 210ndash260K areuncertain A similar picture emerges for other sub-stances although in some cases the failure is even moresevere
The simplest phenomenological adjustment to equa-tion (4) the basis for CNT is to write
ncei frac14 n1 expfrac12DWclethiTHORN DWcleth1THORN
kT
eth23THORN
where DWclethiTHORN is given by equation (5) This amounts toa uniform downward shift of the classical free energybarrier illustrated in Fig 1 At least now the expressionis consistent at i frac14 1 The theory is known as internallyconsistent classical theory (ICCT) [51] Unfortunately itis not clear whether this constant shift in the work offormation should be a realistic representation of all thesubleading terms in the work of formation [52] It is anarbitrary modification and its success in accounting fordata is mixed For example nucleation rates for a rangeof alkanes [114] require further modification by factorsof between one and nine orders of magnitude to comeinto agreement with ICCT
It should be noted at this point that the correctionterm kT lnS which appears in DWcleth1THORN can beregarded as rigorous There are good reasons for thisbased on general considerations involving the nucleationtheorems [56] or more simply the principles of chemicalequilibrium represented by the law of mass action givenin equation (15) The correction converts the prefactorn1 in equation (4) into ns1 and results in the insertion of afactor 1=S in the nucleation rate in equation (1)
The Fisher droplet model is a slightly more sophis-ticated scheme [115] Here the constrained equilibrium
populations are modelled by
ncei frac14 ns1 exp yii2=3 thorn t ln i lnq0kT
psv
thorn i lnS
eth24THORN
The term proportional to yi may be viewed as a surfaceterm while the other two are corrections of a typesuggested by the free energy expressions developed insection 5 The parameters q0 yi and t are chosen toensure that the model correctly reproduces knownproperties of clusters and vapours including virialcoefficients and critical properties the theory is there-fore phenomenological
For large cluster sizes and for temperatures wellbelow the critical temperature the free energy offormation should tend towards that of a macroscopicliquid droplet yi should therefore approach A1s=ethkTTHORNfor large i where A1i
2=3 is the surface area 4pR2 of aspherical droplet containing i monomers at the bulkliquid density so that A1 frac14 eth36pTHORN1=3r2=3
l Dillmann andMeier [116] suggested writing
yi frac14kiA1skT
eth25THORN
with the Ansatz
ki frac14 1thorn a1i 1=3 thorn a2i 2=3 eth26THORN
The coefficients a1 and a2 can be found in terms of virialcoefficients of the vapour The form of this Ansatz wasmotivated by the curvature dependence of the bulksurface tension derived in classical thermodynamics byTolman [117]
The initial success of the DillmannndashMeier theory wasshort lived since modifications had to be made toovercome an internal inconsistency [118] Virial expan-sions for the vapour pressure and density were usedincorrectly to calculate the virial expansion of the ratioof the two Correcting this error [61] changed predictedrates by several orders of magnitude Nevertheless theapproach was very influential [119] In a furtherdevelopment Kalikmanov and van Dongen [120] seta2 frac14 0 (this term can be absorbed into q0 anyway) andchose a1 such that the experimental saturated vapourpressure psv is reproduced by the model The DillmannndashMeier model matched the first two virial coefficientsonly The Kalikmanovndashvan Dongen theory shouldtherefore approximate the (revised) DillmannndashMeiertheory at low temperatures but ought to provide abetter description at higher temperatures
Other phenomenological models have been proposedincluding the Hale scaling relation [121] This modelbased on CNT but employing simplified models for thesurface tension and liquid density based on scalingbehaviour near the critical point together with a rather
STATISTICAL MECHANICS OF NUCLEATION 893
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
more straightforward rate prefactor seems to be quitesuccessful in accounting for data for a wide range ofmaterials even metals [122] and two-component systems[123] Another phenomenological model is the diffuseinterface theory [124] which is based on a treatment ofthe phase interface using simplified density profiles ofthermodynamic properties Another general class ofmodel makes use of conditions at the vapourndashliquidspinodal to constrain the form of the model free energy[125ndash128]
The main drawback in phenomenological models isthe arbitrariness in the form of the proposed work offormation The parameter fitting in practice ensures thatthe populations of small clusters especially the mono-mer and dimer are well represented by the model butthe properties of larger clusters depend on the expres-sion assumed for the model and are therefore uncertainIntuition suggests that phenomenological modelling islikely to have mixed success and that microscopiccalculations of cluster properties do matter There mightbe better value in phenomenological models however ifthey were based securely on the results of microscopiccalculations
8 KINETICS OR THERMODYNAMICS
In section 3 kinetic equations representing the nuclea-tion process were described However the coefficients inthose equations were related to constrained equilibriumpopulations and hence to the equilibrium statisticalmechanics of clusters through the detailed balancecondition (8) Is it necessary to do this Could acompletely kinetic route be followed by evaluating themean growth and decay coefficients bi and gi directly bysome statistical approach The main problem is ofcourse the difficulty in performing such calculationsand this has restricted the development of such models[129ndash131] It is expected that the kinetic coefficients willcorrespond to those derived from the detailed balanceargument for near-equilibrium conditions but circum-stances can be imagined where non-equilibrium effectsare present which cannot be dealt with within standardstatistical mechanics
The framework for evaluating kinetic coefficients ofcluster decay can be briefly introduced [86 132] Amolecular dynamics simulation of a cluster can be set upand evolved according to the interactions until bychance an evaporation event occurs By extensivesimulation statistics for the mean evaporation rates gican be amassed in this way Similarly moleculardynamics simulations where a single molecule iscaptured by a cluster could be performed and thegrowth rates bi determined These will be functions ofcluster energy and so would fit best into a rate equationscheme involving both size and energy [133]
This is a laborious approach but statistical physicsoffers an approximate alternative representation of theescape and capture processes The basic tool is theLangevin equation The radial motion of a molecule inthe cluster is described by [134]
meurorr frac14 f ethrTHORN thornfethr _rr tTHORN eth27THORN
where r is the radial position m is the molecular mass fis the mean (time- and velocity-averaged) force on themolecule at position r and f is the stochastic forceThe latter is usually represented as a velocity-dependent friction term mg _rr where g is the so-calledfriction coefficient plus a fluctuating term ~ff ethr tTHORN withzero mean and correlation in time h ~ff ethr tTHORN~ff ethr t0THORNi frac14eth2gkTmTHORNdetht t0THORN where k is Boltzmannrsquos constant andwhere T has the characteristics of a temperature as willbe seen It is a standard manipulation [135] to convertthe Langevin description in the large friction limit intoa FokkerndashPlanck equation
qwqt
frac14 1
mg qethfwTHORN
qrthorn kT
q2wqr2
eth28THORN
which represents the evolution of wethr tTHORN the prob-ability density that the molecule should lie at the radialposition r It should be noted in passing that FokkerndashPlanck equations have often been used to describe thestochastic dynamics of a cluster in size space [48 136] Incontrast the use of the method described above refers tomotion of a molecule in real coordinate space relative tothe centre of mass of the cluster
The steady state equilibrium solution of equation (28)is
wethrTHORN exp FethrTHORNkT
eth29THORN
where FethrTHORN is called the potential of mean force relatedto the mean force on the molecule f throughf frac14 dF=dr It can now be seen how the parameter Tin the stochastic force plays the role of temperaturesince equation (29) looks like a Boltzmann distributionThis solution is not what is sought however The escapeproblem is characterized by a boundary conditionwethreTHORN frac14 0 where re is the radius at which a particleescapes from the system The resulting escape rate maybe shown to be [134]
gkin frac14 KethreT m gTHORN exp DFkT
eth30THORN
where the prefactor K depends on the parameters listedas well as the shape of the potential of mean force Theprincipal feature of equation (30) is the exponentialdependence on the depth DF of the potential of meanforce Therefore if the potential of mean force FethrTHORN can
I J FORD894
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
be evaluated by studying a molecular dynamics trajec-tory for example then the kinetic decay rate can bedetermined This could be extended to account for aradially dependent temperature and friction coefficientboth of which might be determined from simulation Asketch of a potential of mean force is given in Fig 6
Kinetically determined cluster growth and decayrates gkini and bkini respectively could provide a newroute for the determination of nucleation rates inde-pendent of assumptions about detailed balance and theuse of equilibrium statistical mechanics in a non-equilibrium situation
9 CONCLUSIONS
This review has been brief and in some ways partisan tothe personal views of the author It attempts to sketch acoherent picture of nucleation theory and how thevarious disparate strands relate to one another One ofthe fascinations of nucleation theory is the variety ofapproaches and the number of difficult issues that areencountered How well does free energy fluctuationtheory represent a real non-equilibrium process Howwell does the idea of detailed balance provide aframework for determining non-equilibrium decaycoefficients from equilibrium statistical mechanicsHow well do various methods perform in the evaluation
of cluster free energies Can cluster growth and decayrates be calculated directly without the need for such aframework How exactly are clusters to be defined
The theoretical problems are many but the issues canbe resolved to various degrees of satisfaction Thereremains of course the fundamental task of choosing anintermolecular potential that adequately represents thephysical interactions Without this all the finer points ofcluster definition and treatments of non-equilibriumprocesses are unimportant Modelling many problems inscience relies on the availability of an accurate micro-scopic description and so nucleation rate calculationsare not unique in this respect Most empirical molecularpotentials are fitted to the properties of bulk and surfaceproperties and in the very best examples attention isgiven to their transferability to more unusual environ-ments such as clusters In view of these capabilities andmodern theoretical developments in nucleation theoryit seems likely that successful microscopic models of thenucleation process based on statistical physics are atlast coming within reach
ACKNOWLEDGEMENTS
The present author is grateful to colleagues SarahHarris and Marco Arosio for providing some of thefigures
Fig 6 A molecule in a cluster may be considered to be confined within it by a potential of mean force FethrTHORN Itcan escape by reaching a radial position re driven by a phenomenological stochastic force
STATISTICAL MECHANICS OF NUCLEATION 895
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
REFERENCES
1 Gross D H E Microcanonical Thermodynamics 2001
(World Scientific Singapore)
2 Seinfeld J H and Pandis S N Atmospheric Chemistry
and Physics 1998 (John Wiley New York)
3 Climate Change 2001 Synthesis Report 2001 (Inter-
governmental Panel on Climate Change) (Cambridge
University Press Cambridge)
4 Kohler T M Gail H-P and Sedlmayr E MgO dust
nucleation in M-stars calculation of cluster properties
and nucleation rates Astronomy and Astrophysics 1997
320 553
5 Kashchiev D Nucleation Basic Theory with Applications
2000 (ButterworthndashHeinemann Oxford)
6 Bakhtar F Ebrahimi M and Webb R A On the
performance of a cascade of turbine rotor tip section
blading in nucleating steam Part 1 surface pressure
distributions Proc Instn Mech Engrs 1995 209(C2)
115ndash124
7 Bakhtar F Ebrahimi M and Bamkole B O On the
performance of a cascade of turbine rotor tip section
blading in nucleating steam Part 2 wake traverses Proc
Instn Mech Engrs 1995 209(C3) 169ndash178
8 Bakhtar F andMahpeykar M R On the performance of
a cascade of turbine rotor tip section blading in nucleating
steam Part 3 theoretical treatment Proc Instn Mech
Engrs Part C J Mechanical Engineering Science 1997
211(C3) 195ndash210
9 Volmer M and Weber A Z Physikalische Chemie 1926
119 277
10 Farkas L Z Physikalische Chemie (Leipzig) 1927 125
236
11 Becker R and Doring W Kinetische Behandlung der
Keinbildung in ubersattigten Dampfen Ann Physik
(Leipzig) 1935 24 719
12 Zeldovich J Soviet Physics JETP 1942 12 525
13 Frenkel J Kinetic Theory of Liquids 1946 (Oxford
University Press Oxford)
14 Wagner P E and Strey R Homogeneous nucleation
rates of water vapor measured in a two-piston expansion
chamber J Phys Chemistry 1981 85 2694
15 Miller R Anderson R J Kassner J L and Hagen
D E Homogeneous nucleation rate measurements for
water over a wide range of temperature and nucleation
rate J Chem Physics 1983 78 3204
16 Viisanen Y Strey R and Reiss H Homogeneous
nucleation rates for water J Chem Physics 1993 99 4680
17 Viisanen Y Strey R and Reiss H Erratum homo-
geneous nucleation rates for water J Chem Physics
2000 112 8205
18 Wolk J and Strey R Homogeneous nucleation of H2O
and D2O in comparison the isotope effect J Phys
Chemistry B 2001 105 11683
19 Heath C H Streletsky K and Wyslouzil B E H2Ondash
D2O condensation in a supersonic nozzle J Chem
Physics 2002 117 6176
20 Mikheev V B Irving P M Laulainen N S Barlow
S E and Pervukhin V V Laboratory measurements of
water nucleation using a laminar flow chamber J Chem
Physics 2002 116 10 772
21 Hung C H Krasnopoler M J and Katz J L
Condensation of a supersaturated vapor VIII The
homogeneous nucleation of n-nonane J Chem Physics
1989 90 1856
22 Rudek M M Fisk J A Chakarov V M and Katz
J L Condensation of a supersaturated vapor XII The
homogeneous nucleation of the n-alkanes J Chem
Physics 1996 105 4707
23 Strey R Wagner P E and Schmeling T Homogeneous
nucleation rates for n-alcohol vapors measured in a two-
piston expansion chamber J Chem Physics 1986 84
2325
24 Lihavainen H Viisanen Y and Kulmala M Homo-
geneous nucleation of n-pentanol in a laminar flow
diffusion chamber J Chem Physics 2001 114 10 031
25 Rudek M M Katz J L Vidensky I V Zdımal V and
Smolık J Homogeneous nucleation rates of n-pentanol
measured in an upward thermal diffusion cloud chamber
J Chem Physics 1999 111 3623
26 Grassmann A and Peters F Homogeneous nucleation
rates of n-propanol in nitrogen measured in a piston-
expansion tube J Chem Physics 2002 116 7617
27 Kane D and El-Shall S Condensation of supersaturated
vapors of hydrogen bonding molecules ethylene glycol
propylene glycol trimethyl glycol and glycerol J Chem
Physics 1996 105 7617
28 Hameri K and Kulmala M Homogeneous nucleation in
a laminar flow chamber the effect of temperature and
carrier gas pressure on dibutyl phthalate vapor nucleation
rate at high supersaturations J Chem Physics 1996 105
7696
29 Viisanen Y Strey R Laaksonen A and Kulmala M
Measurement of the molecular content of binary nuclei
II Use of the nucleation rate surface for waterndashethanol
J Chem Physics 1994 100 6062
30 Viisanen Y Wagner P E and Strey R Measurement of
the molecular content of binary nuclei IV Use of the
nucleation rate surfaces for the n-nonane-n-alcohol series
J Chem Physics 1998 108 4257
31 Fisk J A Rudek M M Katz J L Beiersdorf D and
Uchtmann H The homogeneous nucleation of cesium
vapor Atmos Res 1998 46 211
32 Uchtmann H Dettmer R Baranovskii S D and
Hensel F Photoinduced nucleation in supersaturated
mercury vapor J Chem Physics 1998 108 9775
33 Ferguson F T and Nuth J A Experimental studies of
the vapor phase nucleation of refractory compounds V
The condensation of lithium J Chem Physics 2000 113
4093
34 Laaksonen A Talanquer V and Oxtoby D W
Nucleation measurements theory and atmospheric
applications A Rev Phys Chemistry 1995 46 489
35 Debenedetti P G Metastable Liquids Concepts and
Principles 1996 (Princeton University Press Princeton
New Jersey)
36 Yasuoka K and Matsumoto M Molecular dynamics of
homogeneous nucleation in the vapor phase I Lennardndash
Jones fluid J Chem Physics 1998 109 8451
37 Yasuoka K and Matsumoto M Molecular dynamics of
homogeneous nucleation in the vapor phase II Water
J Chem Physics 1998 109 8463
I J FORD896
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
38 Wonczak S Strey R and Stauffer D Confirmation of
classical nucleation theory by Monte Carlo simulations in
the 3-dimensional Ising model at low temperature
J Chem Physics 2000 113 1976
39 Laasonen K Wonczak S Strey R and Laaksonen A
Molecular dynamics simulations of gasndashliquid nucleation
of LennardndashJones fluid J Chem Physics 2000 113 9741
40 Reiss H Methods of Thermodynamics 1996 (Dover
Publications New York)
41 Kondepudi D and Prigogine I Modern Thermodynamics
From Heat Engines to Dissipative Structures 1998 (John
Wiley Chichester West Sussex)
42 Smoluchowski M Ann Phys NY 1908 25 225
43 Einstein A Ann Phys NY 1910 338 1275
44 Hirschfelder J O Curtis C F and Bird R BMolecular
Theory of Gases and Liquids 1954 (John Wiley New
York)
45 Landau L D and Lifshitz E M Statistical Physics 1958
(Pergamon London)
46 Wu D T Nucleation theory Solid State Physicsmdash
Advances in Res and Applic 1997 50 37
47 Reguera D Rubı J M and Perez-Madrid A Fokkerndash
Planck equations for nucleation processes revisited
Physica A 1998 259 10
48 Schenter G K Kathmann S and Garrett B Variational
transition state theory of vapor phase nucleation
J Chem Physics 1999 109 7951
49 Onsager L Phys Rev 1931 37 405
50 Courtney W G J Chem Physics 1961 35 2249
51 Girshick S L and Chiu C-P Kinetic nucleation theory
a new expression for the rate of homogeneous nucleation
from an ideal supersaturated vapor J Chem Physics
1990 93 1273
52 Ford I J Barrett J C and Lazaridis M Uncertainties
in cluster energies in homogeneous nucleation theory
J Aerosol Sci 1993 24 581
53 Weakliem C L and Reiss H The factor 1S in the
classical theory of nucleation J Phys Chemistry 1994
98 6408
54 McClurg R B and Flagan R C Critical comparison of
droplet models in homogeneous nucleation theory
J Colloid Interface Sci 1998 201 194
55 Mashmoushy H Mahpeykar M R and Bakhtar F
Studies of nucleating and wet steam flows in two-dimen-
sional cascades Proc Instn Mech Engrs Part C J
Mechanical Engineering Science 2004 218(C8) 843ndash858
56 Ford I J Nucleation theorems the statistical mechanics
of molecular clusters and a revision of classical nuclea-
tion theory Phys Rev E 1997 56 5615
57 ten Wolde P R and Frenkel D Computer simulation
study of gasndashliquid nucleation in a LennardndashJones system
J Chem Physics 1998 109 9901
58 Kusaka I Wang Z G and Seinfeld J H Direct
evaluation of the equilibrium distribution of physical
clusters by a grand canonical Monte Carlo simulation
J Chem Physics 1998 108 3416
59 Kusaka I and Oxtoby D W Identifying physical
clusters in vapor phase nucleation J Chem Physics
1999 110 5249
60 Stillinger F H J Chem Physics 1963 38 1486
61 Ford I J Imperfect vaporndashgas mixtures and homo-
geneous nucleation J Aerosol Sci 1992 23 447
62 Lothe J and Pound G M J Chem Physics 1962 36
2080
63 Lee J K Barker J A and Abraham F F Theory and
Monte Carlo simulation of physical clusters in the
imperfect vapor J Chem Physics 1973 58 3166
64 Abraham F F Homogeneous Nucleation Theory 1974
(Academic Press New York)
65 Reif F Fundamentals of Statistical and Thermal Physics
1965 (McGraw-Hill New York)
66 Hoare M R and Pal P Adv Phys 1975 24 645
67 Hoare M R Adv Chem Physics 1979 40 49
68 Pal P and Hoare M R Thermodynamic properties
and homogeneous nucleation of molecular clusters of
nitrogen J Phys Chemistry 1987 91 2474
69 Reiss H Katz J L and Cohen E R Translationndash
rotation paradox in the theory of nucleation J Chem
Physics 1968 48 5553
70 Hale B N Monte Carlo calculations of effective surface
tension for small clusters Aust J Physics 1996 49 425
71 Kathmann S M Schenter G K and Garrett B C
Dynamical nucleation theory calculation of condensation
rate constants for small water clusters J Chem Physics
1999 111 4688
72 Reiss H Tabazadeh A and Talbot J Molecular theory
of vapor phase nucleation the physically consistent
cluster J Chem Physics 1990 92 1266
73 Ellerby H M Weakliem C L and Reiss H Toward a
molecular theory of vaporndashphase nucleation I Identifica-
tion of the average embryo J Chem Physics 1992 97
5766
74 Ellerby H M and Reiss H Toward a molecular theory
of vaporndashphase nucleation II Fundamental treatment of
the cluster distribution J Chem Physics 1992 97 5766
75 Weakliem C L and Reiss H Toward a molecular theory
of vaporndashphase nucleation III Thermodynamic properties
of argon clusters fromMonte Carlo simulations and modi-
fied liquid drop theory J Chem Physics 1994 101 2398
76 Kusaka I Wang Z G and Seinfeld J H Binary
nucleation of sulfuric acidndashwater Monte Carlo simula-
tion J Chem Physics 1998 108 6829
77 Oh K J and Zeng X C A small system ensemble
Monte Carlo simulation of supersaturated vapor
evaluation of barrier to nucleation J Chem Physics
2000 112 294
78 Senger B Schaaf P Corti D S Bowles R Voegel J
C and Reiss H A molecular theory of the homogeneous
nucleation rate I Formulation and fundamental issues
J Chem Physics 1999 110 6421
79 Hill T L Statistical Mechanics Principles and Selected
Applications 1956 (McGraw-Hill New York)
80 Soto R and Cordero P Nonideal gas of clusters at
equilibrium J Chem Physics 1999 110 7316
81 Barrett J C The significance of cluster lifetime in
nucleation theory J Chem Physics 2002 116 8856
82 Lushnikov A A and Kulmala M Dimers in nucleating
vapors Phys Rev E 1998 58 3157
83 Bahadur R and McClurg R B Comparison of cluster
definitions in homogeneous nucleation rate calculations
J Phys Chemistry 2001 105 11 893
84 Pugnaloni L A and Vericat F New criteria for cluster
identification in continuum systems J Chem Physics
2002 116 1097
STATISTICAL MECHANICS OF NUCLEATION 897
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
85 Chen B Siepmann J I Oh K J and Klein M L
Simulating vaporndashliquid nucleation of n-alkanes J Chem
Physics 2002 116 4317
86 Harris S A and Ford I J A dynamical definition of
quasi-bound molecular clusters J Chem Physics 2003
118 9216
87 Shreve A P Walton J R P B and Gubbins K E
Liquid drops of polar molecules J Chem Physics 1986
85 2178
88 ten Wolde P R Ruiz-Montero M J and Frenkel D
Numerical calculation of the rate of homogeneous gasndash
liquid nucleation in a LennardndashJones system J Chem
Physics 1998 109 9901
89 ten Wolde P R Oxtoby D W and Frenkel D Chain
formation in homogeneous gasndashliquid nucleation of polar
fluids J Chem Physics 1998 111 4762
90 Vehkamaki H and Ford I J Critical cluster size and
droplet nucleation rate from growth and decay simula-
tions of LennardndashJones clusters J Chem Physics 2000
112 4193
91 Schaaf P Senger B Voegel J C Bowles R K and
Reiss H Simulative determination of kinetic coefficients
for nucleation rates J Chem Physics 2001 114 8091
92 Chen B Siepmann J I Oh K J and Klein M L
Aggregation-volume-bias Monte Carlo simulations of
vaporndashliquid nucleation barriers for LennardndashJonesium
J Chem Physics 2001 115 10 903
93 Hale B N and DiMattio D J A Monte Carlo discrete
sum (MCDS) nucleation rate model for water In
Proceedings of the 15th International Conference on
Nucleation and Atmospheric Aerosols AIP Conference
Proceedings Vol 534 (Eds B N Hale and M Kulmala)
2000 pp 31ndash34
94 Kathmann S M and Hale B N Monte Carlo
simulations of small sulfuric acidndashwater clusters
J Phys Chemistry B 2001 105 11 719
95 Daura X Mark A E and van Gunsteren W F Para-
metrization of aliphatic CHn united atoms of GROMOS96
force field J Comput Chemistry 1998 19 535
96 Kathmann S M Schenter G K and Garrett B C
Understanding the sensitivity of nucleation kinetics a
case study on water J Chem Physics 2002 116 5046
97 Kashchiev D On the relation between nucleation work
nucleus size and nucleation rate J Chem Physics 1982
76 5098
98 Oxtoby D W and Kashchiev D A general relation
between the nucleation work and the size of the nucleus in
multicomponent nucleation J Chem Physics 1994 100
7665
99 Ford I J Thermodynamic properties of critical clusters
from measurements of vapourndashliquid homogeneous
nucleation rates J Chem Physics 1996 105 8324
100 Oxtoby D W and Evans R Nonclassical nucleation
theory for the gasndashliquid transition J Chem Physics
1988 89 7521
101 Zeng X C and Oxtoby D W Gasndashliquid nucleation in
LennardndashJones fluids J Chem Physics 1991 94 4472
102 Evans R Fundamentals of Inhomogeneous Fluids 1992
Ch 3 pp 85ndash175 (Marcel Dekker New York)
103 Talanquer V and Oxtoby D W Dynamical density
functional theory of gasndashliquid nucleation J Chem
Physics 1994 100 5190
104 Barrett J C First-order correction to classical nucleation
theory a density functional approach J Chem Physics
1999 111 5938
105 Barrett J C Cluster translation and growth in density
functional theories of homogeneous nucleation J Chem
Physics 1997 107 7989
106 Bykov T V and Zeng X C A patching model for
surface tension of spherical droplet and Tolman length J
Chem Physics 1999 111 10 602
107 Talanquer V and Oxtoby D W Nucleation in molecular
and dipolar fluids interaction site model J Chem
Physics 1995 103 3686
108 Seok C and Oxtoby D W Nucleation in n-alkanes a
density functional approach J Chem Physics 1998 109
7982
109 Reguera D Bowles R K Djikaev Y and Reiss H
Phase transitions in systems small enough to be clusters
J Chem Physics 2003 118 340
110 Khakshouri S and Ford I J J Chem Physics 2004
(submitted)
111 Ford I J (unpublished)
112 Nyquist R M Talanquer V and Oxtoby D W Density
functional theory of nucleation a semi-empirical
approach J Chem Physics 1995 103 1175
113 Talanquer V and Oxtoby D W Density functional
analysis of phenomenological theories of gasndashliquid
nucleation J Chem Physics 1995 99 2865
114 Katz J L Fisk J A and Chakarov V The accuracy of
nucleation theory In Proceedings of the 13th Interna-
tional Conference on Nucleation and Atmospheric Aerosols
(Eds N Fukuta and P E Wagner) 1992 pp 11ndash18
(Deepak Publishing Hampton Virginia)
115 Fisher M E The theory of condensation and the critical
point Physics 1967 3 255
116 Dillmann A and Meier G E A A refined droplet
approach to the problem of homogeneous nucleation
from the vapor phase J Chem Physics 1991 94 3872
117 Tolman R C J Chem Physics 1949 17 333
118 Ford I J Laaksonen A and Kulmala M Modification
of the DillmannndashMeier theory of homogeneous nuclea-
tion J Chem Physics 1993 99 764
119 Laaksonen A Ford I J and Kulmala M Revised
parametrization of the DillmannndashMeier theory of homo-
geneous nucleation Phys Rev E 1994 49 5517
120 Kalikmanov V I and van Dongen M E H Semipheno-
menological theory of homogeneous vaporndashliquid nuclea-
tion J Chem Physics 1995 103 4250
121 Hale B N Application of a scaled homogeneous
nucleation-rate formalism to experimental-data at
T 5Tc Phys Rev A 1986 33 4156
122 Martinez D M Ferguson F T Heist R H and Nuth
J A Application of scaled nucleation theory to metallic
vapor condensation J Chem Physics 2001 115 310
123 Hale B N and Wilemski G A scaled nucleation model
for ideal binary systems Chem Physics Lett 1999 305
263
124 Granasy L Diffuse interface theory for homogeneous
vapor condensation J Chem Physics 1996 104
5188
125 McGraw R and Laaksonen A Interfacial curvature free
energy the Kelvin relation and vaporndashliquid nucleation
rate J Chem Physics 1997 106 5284
I J FORD898
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
126 Talanquer V A new phenomenological approach to
gasndashliquid nucleation based on the scaling properties
of the critical nucleus J Chem Physics 1997 106
9957
127 Koga K and Zeng X C Thermodynamic expansion of
nucleation free-energy barrier and size of critical nucleus
near the vaporndashliquid coexistence J Chem Physics 1999
110 3466
128 Kusaka I On the scaling behavior of the free energetics of
nucleation J Chem Physics 2003 118 5510
129 Nowakowski B and Ruckenstein E A kinetic approach
to the theory of nucleation in gases J Chem Physics
1991 94 1397
130 Nowakowski B and Ruckenstein E Homogeneous
nucleation in gases a three-dimensional FokkerndashPlanck
equation for evaporation from clusters J Chem Physics
1991 94 8487
131 Wilcox C F and Bauer S H Estimation of homo-
geneous nucleation flux via a kinetic model J Chem
Physics 1991 94 8302
132 Hanggi P Talkner P and Borkovec M Reaction rate
theory fifty years after Kramers Rev Mod Physics 1990
62 251
133 Barrett J C Clement C F and Ford I J Energy
fluctuations in homogeneous nucleation theory for aero-
sols J Physics A 1993 26 529
134 Ford I J and Harris S A Molecular cluster decay
viewed as escape from a potential of mean force J Chem
Physics 2004 120 4428
135 Risken H The FokkerndashPlanck Equation Methods of
Solution and Applications 2nd edition 1989 (Springer-
Verlag Berlin)
136 Shizgal B and Barrett J C Time dependent nucleation
J Chem Physics 1989 91 6505
STATISTICAL MECHANICS OF NUCLEATION 899
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
nucleation rate is then J bini where ni is the meannumber of critical clusters in the system
More sophisticated treatments of thermodynamicscan provide the time scale without recourse to the abovekinetic argument which is reasonably valid for gas-to-particle conversion but not for freezing for exampleAttempts based on linearized non-equilibrium thermo-dynamics [47 48] originate from the work of Onsager[49]
The next step is crucial and it is to relate thepopulation ni to the probability of a fluctuation and toan explicit model of the critical work of formation Theassumption in classical nucleation theory is that
ni frac14 n1 expDWclethi THORN
kT
eth4THORN
where DWclethiTHORN is the classical work of formation givenby
DWclethiTHORN frac14 4pR2s ikT lnS eth5THORN
and DWclethi THORN is the maximum of this function Thecluster is envisaged to be a sphere of radius R given by4pR3rl=3 frac14 i
Unfortunately some uncertainty accompanies thesesteps in the derivation The theory is flawed on severalcounts The work of formation contains a termassociated with the creation of a spherical interface(characterized by the bulk value of surface tension) anda change kT lnS in free energy per molecule due tothe change in phase from supersaturated vapour tocondensate Moreover the factor n1 appearing inequation (4) appears to have been introduced heuristi-cally Might it not instead be the monomer populationin the saturated vapour How does DWclethiTHORN comparewith the actual work of formation DWethiTHORN Is thedroplet really spherical with a sharp interface Have
any important factors been left out Furthermoreshould there be concern that equation (4) is in-consistent when i frac14 1 This has been a controversialissue [50ndash54]
3 KINETIC INTERPRETATION
The expression for the classical rate of nucleation givenin equation (1) emerges from the models in equation (4)and (5) when the thermodynamic transition stateanalysis is embellished by a kinetic interpretation [11]Instead of using the probability of fluctuations awayfrom the equilibrium state (the TST route) a set of rateequations is introduced to describe the dynamics ofcluster populations
The evolution of cluster populations is modelled usingthe following equations
dni
dtfrac14Xj
njPji niXj
Pij eth6THORN
where Pji is the mean rate at which a j cluster convertsinto an i cluster The transition processes are assumed tobe Markovian so that the rate coefficients depend onthe properties of the cluster and its environment but noton their history
The kinetic interpretation can be taken a step furtherby assuming that the only important transitions arethose involving the addition or loss of single moleculesto or from the cluster The scheme of transitionprocesses is illustrated in Fig 2 The only non-zerorate coefficients are then bi frac14 Piithorn1 for cluster growthand gi frac14 Pii1 for cluster decay The rate equations
Fig 2 The scheme of the birth and death of clusters addition of a monomer to a cluster will take it up thechain towards the right while the loss of a monomer will take it in the opposite direction Attachmentand detachment of dimers and larger clusters are usually ignored
STATISTICAL MECHANICS OF NUCLEATION 887
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
reduce to
dni
dtfrac14 bi1ni1 gini bini thorn githorn1nithorn1 eth7THORN
These are the BeckerndashDoring equations [11] and theyprovide a useful reinterpretation of the TST approachThere are two steady state solutions In the first thepopulations satisfy equation (7) and also the detailedbalance condition
bi1nei1 frac14 gin
ei eth8THORN
and the fnei g are taken to represent the cluster popula-tions in an equilibrium vapour at or below the saturatedvapour pressure For true thermal equilibrium themonomer population has to be less than or equal to itsvalue for the saturated vapour ns1 The cluster populationsin a saturated vapour are denoted fnsig Nevertheless asituation can be invented which is known as lsquoconstrainedequilibriumrsquo where the detailed balance condition is stillsatisfied but with n1 gt ns1 Since bi pvS while gi isindependent of vapour pressure equation (8) then definesso-called constrained equilibrium populations
ncei ampSinsi eth9THORN
for a supersaturated vapour It has been assumed thatS frac14 pv=p
svampn1=n
s1
The second steady state solution to the BeckerndashDoring equations is more realistic It is not meant torepresent thermal equilibrium The supersaturation isset to a value greater than unity and a boundarycondition of zero population at some cluster size imax isimposed These conditions will drive a steady productionrate of large clusters from the vapour The approach isfor mathematical convenience only and the same steadystate cluster populations nssi and nucleation rate wouldemerge if imax was sent off to infinity The detailedbalance condition (8) is replaced by J frac14 bi1n
ssi1 gin
ssi
defining the nucleation current J which can then becalculated analytically as shown in another paper in thisvolume [55] To a good approximation
J frac14 Zdbincei eth10THORN
which involves the constrained equilibrium population atthe critical size i the size with the lowest constrainedequilibrium population The condition imax 4 i musthold Zd is the so-called Zeldovich factor [12] givenapproximately by
Zd frac14 1
2pq2ethln ncei THORN
qi2
ifrac14i
1=2
eth11THORN
When equations (4) and (5) are used to specify ncei and theresult is substituted into equations (10) and (11) then theclassical nucleation rate given in equation (1) isrecovered as long as the large i limit of the attachment
rate coefficient in equation (3) is taken It is also possibleto derive an expression for J purely in terms of the ratecoefficients bi and gi This possibility is considered againin section 8
The main assumptions used in this derivation of theclassical theory principally the validity of equations (4)and (5) are now examined In order to explore this it isnecessary to see how the equilibrium cluster populationsnei in subsaturated vapours may be calculated usingstatistical mechanics
4 CLUSTER STATISTICAL MECHANICS
The statistical mechanical treatment of a vapour is bestdeveloped using a grand canonical ensemble [56ndash59] Thesystem is placed in contact with a particle reservoir at achemical potential and temperature T so that thenumber of molecules N in the system can change Thegrand partition function of the system X is then propor-tional to the integral of expf frac12HethNTHORN N=ethkTTHORNg overall possible configurations of molecular positions andmomenta whereHethNTHORN is the Hamiltonian of the system
Evaluating X exactly is exceedingly difficult and it ismuch simpler to proceed by considering the configura-tions of the system as arrangements of molecularclusters The vapour then is viewed as a mixture ofideal gases each composed of clusters of differentsizes The grand partition function of the system canbe reconstructed from partition functions for singlei-clusters In order to do this however a cluster must bedefined in a unique way in terms of molecular positionsand velocities An example of such a definition would beto require the separation between molecules in a clusterto be less than a maximum distance Rc [60] as illustratedin Fig 3
Fig 3 According to the Stillinger definition a molecule isconsidered part of a cluster if it lies closer than adistance Rc from a constituent molecule of the cluster
I J FORD888
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
At first sight the cluster definition is completelyarbitrary a means of categorizing and resumming allthe elementary configurations of the system Howeveraccurately representing the system as a mixture of idealgases requires the interactions between molecules lyingin different clusters to be minimized The clusterdefinition affects whether a particular interaction isregarded as intercluster or intracluster and so it plays acrucial role This will be discussed further in section 6Residual clusterndashcluster interactions can be taken intoaccount in the form of a virial series [61]
These ideas are now put into practice Define amodified canonical partition function zi for a clustercontaining i molecules This is simply the integral ofexpf frac12HcethiTHORN i=ethkTTHORNg over configurations involvingmolecules in the cluster alone with Hc the Hamiltonianof the i molecules comprising the cluster Using thepicture of a mixture of ideal gases the grand partitionfunction of the complete system can then be expressed as
XampXfnig
Yifrac141
znii
ni
eth12THORN
where the sum is over all possible cluster populationdistributions fnig This is demonstrated rigorously inappendix A of reference [57]
The next step is to determine the distribution fnei gthat dominates the sum in equation (12) The logarithmof the summand is extremized
qqni
Xifrac141
lnz
nii
ni
frac14 0 eth13THORN
which gives the following very simple expression for themost probable or equilibrium cluster size distributionfor the given conditions T and
nei frac14 ziethTTHORN frac14 ZethTTHORN ei=ethkTTHORN eth14THORN
which is equivalent to the so-called law of mass action
nei frac14 Zine1Z1
i
eth15THORN
where ZiethTTHORN frac14 expfrac12FethiTHORN=ethkTTHORN frac14 zi expfrac12 i=ethkTTHORN isthe canonical partition function for an i cluster and FethiTHORNis the i-cluster Helmholtz free energy This relationshipis often used as the starting point in models of clusterpopulations
The vapour pressure is now given by a sum of partialpressures
pvV frac14 kTXifrac141
nei frac14 kTXifrac141
ziethTTHORN eth16THORN
and this series will only converge if the equilibriumpopulations or the zi decrease sufficiently quickly forlarge i This limits the statistical mechanics approach to
the study of saturated or subsaturated vapours How-ever as in the preceding section the equilibriumpopulations nei in equation (14) (for S lt 1) can beextrapolated into constrained equilibrium populationsncei (with S gt 1) for use in equation (10) Saturatedvapour corresponds to a particular chemical potentialms and the saturated vapour cluster populations are nsi The constrained equilibrium populations are obtainedby inserting into equation (14) the chemical potential ofthe supersaturated vapour which is frac14 s thorn kT lnSgiving ncei frac14 Sinsi which is consistent with the resultobtained in equation (9) for dilute vapours
The constrained equilibrium populations are there-fore expressed as
ncei frac14 SiziethsTTHORN
frac14 expfrac12FethiTHORN is ikT lnS
kT
eth17THORN
and the correspondence with CNT can be made closerby introducing the n1 prefactor
ncei frac14 n1 exp frac12FethiTHORN is thorn kT ln n1 ikT lnS
kT
eth18THORN
The next task is to explore how the combinationFethiTHORN is thorn kT ln n1 might be approximated by theCNT result 4pR2s This can be explored most easilyfor solid rather than liquid clusters
5 SOLID CLUSTERS
The partition function of a solid cluster of i moleculescan be factorized into translational rotational andvibrational parts through a transformation to suitablerelative positions and momenta [62ndash64] Starting withthe vibrational degrees of freedom the atom positionsare defined with respect to their mean positions whichare determined by a fixed centre of mass and set oflattice vectors It is assumed that the potential energypart of the Hamiltonian is quadratic in the displace-ments from these mean positions Now the partitionfunction is worked out for configurations of the particlesrestricted to having zero total linear and total angularmomentum about the centre of mass This is writtenZvib corresponding to a product of partition functionsfor 3ji 6 independent oscillators (if there are j atomsper molecule and therefore a total of 3ji degrees offreedom if the molecule is not linear) each of which inthe classical limit takes the form kT=ethhoTHORN where o isthe angular frequency of the specific vibrational modeThe partition function has an additional factor ofexpfrac12U0ethiTHORN=ethkTTHORN where U0ethiTHORN is the system potentialenergy evaluated at the mean atom positions A
STATISTICAL MECHANICS OF NUCLEATION 889
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
different expression applies for low temperatures whenthe oscillator is not classically activated [65]
The rotational and translational degrees of freedomare next The previous restriction of the configurationsto zero angular momentum is lifted The approximate(classical) rotational partition function is
Zrot frac14 p1=28p2kTh2
3=2
ethI1I2I3THORN1=2 eth19THORN
where Ik are the principal moments of inertia about thecentre of mass for atoms fixed in their mean positions
Finally the translational free energy of the centre ofmass of the cluster is given by the partition function
Ztrans frac14 V2pMkT
h2
3=2
eth20THORN
where V is the system volume and M frac14 im is the clustermass The free energy of the cluster (plus kT ln n1 forconvenience) is therefore
FethiTHORN thorn kT ln n1
frac14 kT lnZtransZrotZvib
n1
frac14 U0ethiTHORN kT ln
eth2pMkTTHORN3=2
Srsvh3p1=2
8p2kTh2
3=2
6ethI1I2I3THORN1=2Y3ji6
kfrac141
kT
hok
eth21THORN
where rsv frac14 ns1=V is the molecular density of dilutesaturated vapour
By similar arguments the chemical potential of dilutesaturated vapour is
s frac14 Um0 kT ln
eth2pmkTTHORN3=2
rsvh3p1=2
8p2kTh2
3=2
6ethIm1 Im2 Im3 THORN1=2Y3j6
kfrac141
kT
homk
eth22THORN
where Um0 is the potential energy of a molecule when the
atoms are at their mean positions Imk are the threeprincipal moments of inertia of the molecule about thecentre of mass and om
k are the oscillatory frequencies ofmolecular internal vibrations
The classical work of formation DWclethiTHORN should nowcorrespond to the leading-order contribution to thedifference between equation (21) and i times equation(22) Calculations [54 64 66ndash68] demonstrate that theleading term is indeed roughly proportional to i2=3 thusjustifying the form of the classical model The suitabilityof the capillarity approximation depends on the extentto which the coefficient of this leading term correspondsnumerically to the planar surface free energy Howeverthe full statistical mechanical expression for the work of
formation contains additional terms and these will beimportant for small clusters and need to be taken intoaccount In order to evaluate them the variousstructural and energy properties of cluster are requiredall of which can be obtained from microscopic modelsand they can then be inserted into the formulae for thethermodynamic properties [equations (21) and (22)] theconstrained equilibrium populations [equation (18)] andultimately the nucleation rate [equation (10)] Followingthis route employs the almost universally held assump-tion that equilibrium properties such as free energies canbe used to describe non-equilibrium properties specifi-cally the cluster decay rate This assumption is re-examined in section 8
6 LIQUID DROPLETS AND THE CLUSTER
DEFINITION
Defining what was meant by a solid cluster in theprevious section was not an issue Within the assump-tions of the model (harmonic interactions) it wasimpossible for a molecule to escape from the clusterThe cluster definition in the case of liquids cannot bedismissed so easily Liquids are more volatile thansolids and the molecules in a cluster are much moremobile It is imperative now to characterize a clusterusing the molecular positions and velocities andperhaps some time scale and length scale too
The traditional microscopic method for defining acluster is to impose a geometric constraint on molecularpositions An example is to restrict molecules to a spherecentred on the centre of mass [69ndash71] or to impose amaximum allowable separation between the molecules(the Stillinger criterion [60]) More sophisticated treat-ments exist such as the n=v Stillinger cluster introducedby Reiss and co-workers [72ndash75] Here the confiningvolume v used to define the cluster is considered to be anadditional defining characteristic on the same level asthe cluster size n As time progresses a cluster changesits confining volume as the component molecules followtheir respective trajectories just as they change their sizewith the gain and loss of molecules The association of amolecular configuration with a precise confining volumeis complicated but tractable The dependence of the freeenergy on v determines the statistical dynamics of thiscluster volume A similar attempt to associate a clusterwith a specified volume within a grand canonicalensemble description was developed by Kusaka andco-workers [58 59 76] The simulation box was taken tobe the confining cluster volume Many of these ideasmay be viewed as a treatment of the cluster definition(eg the Stillinger radius) as a variational parameter inthe representation of the partition function Further
I J FORD890
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
refinement of this point of view using a treatment ofclusterndashcluster interactions to optimize the definingcluster volume has been presented [77] A review ofrecent progress in this area and the application of theseideas to nucleation theory has been provided by Sengeret al [78]
Such geometric constraints can be implemented buttheir realism is questionable Situations better regardedas close encounters between separate clusters would beclassed as snapshots of single larger clusters It shouldbe recalled that the cluster definition is an internalfeature of the model which is to be optimized to reduceclusterndashcluster interactions and to improve the repre-sentation of the grand partition function Misclassifyingconfigurations is fraught with consequences A secondissue to consider is that for convenience the clustersshould behave in a manner that can be modelled withBeckerndashDoring equations This means that the growthand decay rates should be free of memory effects ie thetransition processes should be Markovian The extent towhich this is satisfied (if indeed it is physically realistic)depends on the cluster definition
An intuitively reasonable choice of definition whichaddresses both these points is that a cluster shouldconsist of molecules that are bound energetically Thisshould minimize the clusterndashcluster contributions to thetotal energy According to such a definition if amolecule interacts strongly with a cluster then it islikely to be regarded as part of that cluster Also anenergetically bound cluster is likely to decay by thermalfluctuation in a stochastic Markovian fashion ratherthan by the separation of passing poorly boundfragments
Ideas have been proposed which require the cluster tobe energetically bound (in a loose sense since a cluster isintrinsically prone to decay) The time evolution of amolecular configuration and not simply the set ofinstantaneous positions of its component molecules isof central importance Hill [79] introduced a clusterdefinition requiring the total energies of pairs ofmolecules in a cluster to be negative (in their centre ofmass frame) The intention was to exclude situationswhere a molecule could leave the cluster within the timeit would take for a molecule to travel a typical lineardimension of the cluster Soto and Cordero [80] havedeveloped Hillrsquos criterion and Barrett [81] has per-formed Monte Carlo modelling of molecular configura-tions excluding those which decay in moleculardynamics within a given period A number of otherenergy-based criteria have been developed [82ndash85]
Harris and Ford [86] have recently proposed a newenergy-based cluster definition As with Barrettrsquosapproach it is based on the requirement that amolecular snapshot is energetically bound for a chosensubsequent period Evaporation according to thisscheme requires not only that a molecule should makea positive energy excursion but also that the subsequent
dynamics carry it far away from the cluster Onlyperhaps one in eight molecules manage to avoidrecapture once having satisfied the positive energycriterion [86] A snapshot of a cluster in this casesurrounded by a carrier gas is shown in Fig 4 Thehistory of the energy of a molecule in such a cluster isillustrated in Fig 5 three excursions into positive energyoccur in this sequence but only one leads to actualescape corresponding to the maintenance of positiveenergy for a significant time The instant of escape istaken to be the point at which the energy goes positivefor the third and final time
Cluster definitions implemented in moleculardynamics and Monte Carlo codes allow cluster freeenergies and hence the decay rate to be calculated Anenormous amount of simulation work has been carriedout and calculations can be laborious Many differentschemes and approximations have been made to easethe evaluation of these properties The principalconclusion from various studies using a range of clusterdefinitions and interaction models [49 57 58 70 85 87ndash92] seems to be that the work of formation of liquidcritical clusters does indeed contain a dominantcontribution proportional to i2=3 consistent with theclassical work of formation Hale and collaboratorspresent cluster free energies for water and other speciesin a way that makes this particularly clear indeed it issurprising how dominant the surface term can be foreven very small clusters [70 93 94] Neverthelessfurther contributions exist small clusters are notcharacterized just by scaled-down properties of largeclusters and it is impossible to escape the need todetermine the additional terms
In all the above discussion it has been assumed thatthe intermolecular interactions are accurately specifiedThis is a situation that is rarely possible Althoughfirst-principles quantum-mechanical models could be
Fig 4 Snapshot of a cluster of 50 argon atoms (brightspheres) bathed in a gas of helium (darker spheres)
STATISTICAL MECHANICS OF NUCLEATION 891
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
developed in practice empirical potentials are devel-oped on the basis of physical intuition containingparameters fitted to the measured properties of thesubstance in question [95] Usually these are bulkproperties and it is not always clear that the sameinteractions can be used for the very different circum-stances of a molecular cluster [96] In view of thesensitivity of nucleation rates to the thermodynamicproperties of clusters this is perhaps a crucial failingProgress can be made however by computing knowncluster properties where they are available and byrefining the potentials to suit Extraction of clusterproperties through use of the so-called nucleationtheorems (which relate derivatives of the nucleationrate to properties of the critical cluster [16 56 97ndash99])may provide just this sort of experimental information
A technique that has not yet been described is the useof density functional theory (DFT) to model theproperties of liquid droplets [100ndash104] This powerfulapplication of statistical mechanics promises to avoidlaborious computations at the microscopic level byinstead focusing on the density profile of moleculesmaking up the droplet This profile may be regarded asthe average distribution of molecules about the clustercentre of mass Although this might appear to involvecoarse-graining and approximation it is a well-foundedapproach and in principle can be exact In practical
implementations however a mean field approach istaken such that the constituent molecules move in aneffective potential well created by the other molecules inthe cluster [105] Also the interactions are assumed to becharacterized by strong repulsive cores and weaklyattractive tails
Considerable analytical work has been carried out[104 106] Various developments have been made todescribe droplets consisting of simple atoms as well aschain-like molecules [107] and polar molecules [108]The main drawback in these implementations is thatspecific interactions such as hydrogen bonds are hardto represent Certain details of the modelled systemare neglected [109 110] and the (implicit) clusterdefinition used is geometric and not energetic [111]On the other hand quite sophisticated systems havebeen modelled including multi-component dropletsThe method has been compared with otherapproaches [112 113] and certain features are betterdescribed particularly the approach to spinodalconditions where the nucleation barrier vanishesDFT calculations can be performed relatively easilyand they can provide a qualitative and perhapsquantitative insight into quite complicated clusteringproblems They provide an intermediate-level descrip-tion between the simple capillarity approximation anda microscopic description
Fig 5 The energy of a molecule in the centre-of-mass frame of a cluster occasionally becomes positive butthis does not necessarily lead to escape Only the third positive energy excursion in this sequenceindicated by an arrow results in particle escape A cluster definition must capture this behaviour
I J FORD892
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
7 PHENOMENOLOGICAL MODELS
The reader might be dismayed by the mathematicalcomplexity of the expressions for the work of formationof solid clusters in section 5 might be alarmed by theprocedural complexity of implementing a cluster defini-tion in section 6 may recoil from the difficulties incalculating free energies from molecular simulationmethods and ultimately may give up in view of theuncertainties in the underlying intermolecular interac-tions Considering these complications it is tempting totry to develop phenomenological models of the work offormation After all if the capillarity approximationworks reasonably well for water then surely a minorcorrection to account for small cluster effects ought tobe sufficient to improve matters
However after perhaps two decades of work nomodel has emerged offering broad success across arange of different substances It is perhaps asking toomuch to expect universal improvement CNT works as afirst approximation for water for a particular tempera-ture range but as discussed earlier it possesses thewrong temperature dependence so that predictionsaway from a temperature in the range 210ndash260K areuncertain A similar picture emerges for other sub-stances although in some cases the failure is even moresevere
The simplest phenomenological adjustment to equa-tion (4) the basis for CNT is to write
ncei frac14 n1 expfrac12DWclethiTHORN DWcleth1THORN
kT
eth23THORN
where DWclethiTHORN is given by equation (5) This amounts toa uniform downward shift of the classical free energybarrier illustrated in Fig 1 At least now the expressionis consistent at i frac14 1 The theory is known as internallyconsistent classical theory (ICCT) [51] Unfortunately itis not clear whether this constant shift in the work offormation should be a realistic representation of all thesubleading terms in the work of formation [52] It is anarbitrary modification and its success in accounting fordata is mixed For example nucleation rates for a rangeof alkanes [114] require further modification by factorsof between one and nine orders of magnitude to comeinto agreement with ICCT
It should be noted at this point that the correctionterm kT lnS which appears in DWcleth1THORN can beregarded as rigorous There are good reasons for thisbased on general considerations involving the nucleationtheorems [56] or more simply the principles of chemicalequilibrium represented by the law of mass action givenin equation (15) The correction converts the prefactorn1 in equation (4) into ns1 and results in the insertion of afactor 1=S in the nucleation rate in equation (1)
The Fisher droplet model is a slightly more sophis-ticated scheme [115] Here the constrained equilibrium
populations are modelled by
ncei frac14 ns1 exp yii2=3 thorn t ln i lnq0kT
psv
thorn i lnS
eth24THORN
The term proportional to yi may be viewed as a surfaceterm while the other two are corrections of a typesuggested by the free energy expressions developed insection 5 The parameters q0 yi and t are chosen toensure that the model correctly reproduces knownproperties of clusters and vapours including virialcoefficients and critical properties the theory is there-fore phenomenological
For large cluster sizes and for temperatures wellbelow the critical temperature the free energy offormation should tend towards that of a macroscopicliquid droplet yi should therefore approach A1s=ethkTTHORNfor large i where A1i
2=3 is the surface area 4pR2 of aspherical droplet containing i monomers at the bulkliquid density so that A1 frac14 eth36pTHORN1=3r2=3
l Dillmann andMeier [116] suggested writing
yi frac14kiA1skT
eth25THORN
with the Ansatz
ki frac14 1thorn a1i 1=3 thorn a2i 2=3 eth26THORN
The coefficients a1 and a2 can be found in terms of virialcoefficients of the vapour The form of this Ansatz wasmotivated by the curvature dependence of the bulksurface tension derived in classical thermodynamics byTolman [117]
The initial success of the DillmannndashMeier theory wasshort lived since modifications had to be made toovercome an internal inconsistency [118] Virial expan-sions for the vapour pressure and density were usedincorrectly to calculate the virial expansion of the ratioof the two Correcting this error [61] changed predictedrates by several orders of magnitude Nevertheless theapproach was very influential [119] In a furtherdevelopment Kalikmanov and van Dongen [120] seta2 frac14 0 (this term can be absorbed into q0 anyway) andchose a1 such that the experimental saturated vapourpressure psv is reproduced by the model The DillmannndashMeier model matched the first two virial coefficientsonly The Kalikmanovndashvan Dongen theory shouldtherefore approximate the (revised) DillmannndashMeiertheory at low temperatures but ought to provide abetter description at higher temperatures
Other phenomenological models have been proposedincluding the Hale scaling relation [121] This modelbased on CNT but employing simplified models for thesurface tension and liquid density based on scalingbehaviour near the critical point together with a rather
STATISTICAL MECHANICS OF NUCLEATION 893
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
more straightforward rate prefactor seems to be quitesuccessful in accounting for data for a wide range ofmaterials even metals [122] and two-component systems[123] Another phenomenological model is the diffuseinterface theory [124] which is based on a treatment ofthe phase interface using simplified density profiles ofthermodynamic properties Another general class ofmodel makes use of conditions at the vapourndashliquidspinodal to constrain the form of the model free energy[125ndash128]
The main drawback in phenomenological models isthe arbitrariness in the form of the proposed work offormation The parameter fitting in practice ensures thatthe populations of small clusters especially the mono-mer and dimer are well represented by the model butthe properties of larger clusters depend on the expres-sion assumed for the model and are therefore uncertainIntuition suggests that phenomenological modelling islikely to have mixed success and that microscopiccalculations of cluster properties do matter There mightbe better value in phenomenological models however ifthey were based securely on the results of microscopiccalculations
8 KINETICS OR THERMODYNAMICS
In section 3 kinetic equations representing the nuclea-tion process were described However the coefficients inthose equations were related to constrained equilibriumpopulations and hence to the equilibrium statisticalmechanics of clusters through the detailed balancecondition (8) Is it necessary to do this Could acompletely kinetic route be followed by evaluating themean growth and decay coefficients bi and gi directly bysome statistical approach The main problem is ofcourse the difficulty in performing such calculationsand this has restricted the development of such models[129ndash131] It is expected that the kinetic coefficients willcorrespond to those derived from the detailed balanceargument for near-equilibrium conditions but circum-stances can be imagined where non-equilibrium effectsare present which cannot be dealt with within standardstatistical mechanics
The framework for evaluating kinetic coefficients ofcluster decay can be briefly introduced [86 132] Amolecular dynamics simulation of a cluster can be set upand evolved according to the interactions until bychance an evaporation event occurs By extensivesimulation statistics for the mean evaporation rates gican be amassed in this way Similarly moleculardynamics simulations where a single molecule iscaptured by a cluster could be performed and thegrowth rates bi determined These will be functions ofcluster energy and so would fit best into a rate equationscheme involving both size and energy [133]
This is a laborious approach but statistical physicsoffers an approximate alternative representation of theescape and capture processes The basic tool is theLangevin equation The radial motion of a molecule inthe cluster is described by [134]
meurorr frac14 f ethrTHORN thornfethr _rr tTHORN eth27THORN
where r is the radial position m is the molecular mass fis the mean (time- and velocity-averaged) force on themolecule at position r and f is the stochastic forceThe latter is usually represented as a velocity-dependent friction term mg _rr where g is the so-calledfriction coefficient plus a fluctuating term ~ff ethr tTHORN withzero mean and correlation in time h ~ff ethr tTHORN~ff ethr t0THORNi frac14eth2gkTmTHORNdetht t0THORN where k is Boltzmannrsquos constant andwhere T has the characteristics of a temperature as willbe seen It is a standard manipulation [135] to convertthe Langevin description in the large friction limit intoa FokkerndashPlanck equation
qwqt
frac14 1
mg qethfwTHORN
qrthorn kT
q2wqr2
eth28THORN
which represents the evolution of wethr tTHORN the prob-ability density that the molecule should lie at the radialposition r It should be noted in passing that FokkerndashPlanck equations have often been used to describe thestochastic dynamics of a cluster in size space [48 136] Incontrast the use of the method described above refers tomotion of a molecule in real coordinate space relative tothe centre of mass of the cluster
The steady state equilibrium solution of equation (28)is
wethrTHORN exp FethrTHORNkT
eth29THORN
where FethrTHORN is called the potential of mean force relatedto the mean force on the molecule f throughf frac14 dF=dr It can now be seen how the parameter Tin the stochastic force plays the role of temperaturesince equation (29) looks like a Boltzmann distributionThis solution is not what is sought however The escapeproblem is characterized by a boundary conditionwethreTHORN frac14 0 where re is the radius at which a particleescapes from the system The resulting escape rate maybe shown to be [134]
gkin frac14 KethreT m gTHORN exp DFkT
eth30THORN
where the prefactor K depends on the parameters listedas well as the shape of the potential of mean force Theprincipal feature of equation (30) is the exponentialdependence on the depth DF of the potential of meanforce Therefore if the potential of mean force FethrTHORN can
I J FORD894
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
be evaluated by studying a molecular dynamics trajec-tory for example then the kinetic decay rate can bedetermined This could be extended to account for aradially dependent temperature and friction coefficientboth of which might be determined from simulation Asketch of a potential of mean force is given in Fig 6
Kinetically determined cluster growth and decayrates gkini and bkini respectively could provide a newroute for the determination of nucleation rates inde-pendent of assumptions about detailed balance and theuse of equilibrium statistical mechanics in a non-equilibrium situation
9 CONCLUSIONS
This review has been brief and in some ways partisan tothe personal views of the author It attempts to sketch acoherent picture of nucleation theory and how thevarious disparate strands relate to one another One ofthe fascinations of nucleation theory is the variety ofapproaches and the number of difficult issues that areencountered How well does free energy fluctuationtheory represent a real non-equilibrium process Howwell does the idea of detailed balance provide aframework for determining non-equilibrium decaycoefficients from equilibrium statistical mechanicsHow well do various methods perform in the evaluation
of cluster free energies Can cluster growth and decayrates be calculated directly without the need for such aframework How exactly are clusters to be defined
The theoretical problems are many but the issues canbe resolved to various degrees of satisfaction Thereremains of course the fundamental task of choosing anintermolecular potential that adequately represents thephysical interactions Without this all the finer points ofcluster definition and treatments of non-equilibriumprocesses are unimportant Modelling many problems inscience relies on the availability of an accurate micro-scopic description and so nucleation rate calculationsare not unique in this respect Most empirical molecularpotentials are fitted to the properties of bulk and surfaceproperties and in the very best examples attention isgiven to their transferability to more unusual environ-ments such as clusters In view of these capabilities andmodern theoretical developments in nucleation theoryit seems likely that successful microscopic models of thenucleation process based on statistical physics are atlast coming within reach
ACKNOWLEDGEMENTS
The present author is grateful to colleagues SarahHarris and Marco Arosio for providing some of thefigures
Fig 6 A molecule in a cluster may be considered to be confined within it by a potential of mean force FethrTHORN Itcan escape by reaching a radial position re driven by a phenomenological stochastic force
STATISTICAL MECHANICS OF NUCLEATION 895
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
REFERENCES
1 Gross D H E Microcanonical Thermodynamics 2001
(World Scientific Singapore)
2 Seinfeld J H and Pandis S N Atmospheric Chemistry
and Physics 1998 (John Wiley New York)
3 Climate Change 2001 Synthesis Report 2001 (Inter-
governmental Panel on Climate Change) (Cambridge
University Press Cambridge)
4 Kohler T M Gail H-P and Sedlmayr E MgO dust
nucleation in M-stars calculation of cluster properties
and nucleation rates Astronomy and Astrophysics 1997
320 553
5 Kashchiev D Nucleation Basic Theory with Applications
2000 (ButterworthndashHeinemann Oxford)
6 Bakhtar F Ebrahimi M and Webb R A On the
performance of a cascade of turbine rotor tip section
blading in nucleating steam Part 1 surface pressure
distributions Proc Instn Mech Engrs 1995 209(C2)
115ndash124
7 Bakhtar F Ebrahimi M and Bamkole B O On the
performance of a cascade of turbine rotor tip section
blading in nucleating steam Part 2 wake traverses Proc
Instn Mech Engrs 1995 209(C3) 169ndash178
8 Bakhtar F andMahpeykar M R On the performance of
a cascade of turbine rotor tip section blading in nucleating
steam Part 3 theoretical treatment Proc Instn Mech
Engrs Part C J Mechanical Engineering Science 1997
211(C3) 195ndash210
9 Volmer M and Weber A Z Physikalische Chemie 1926
119 277
10 Farkas L Z Physikalische Chemie (Leipzig) 1927 125
236
11 Becker R and Doring W Kinetische Behandlung der
Keinbildung in ubersattigten Dampfen Ann Physik
(Leipzig) 1935 24 719
12 Zeldovich J Soviet Physics JETP 1942 12 525
13 Frenkel J Kinetic Theory of Liquids 1946 (Oxford
University Press Oxford)
14 Wagner P E and Strey R Homogeneous nucleation
rates of water vapor measured in a two-piston expansion
chamber J Phys Chemistry 1981 85 2694
15 Miller R Anderson R J Kassner J L and Hagen
D E Homogeneous nucleation rate measurements for
water over a wide range of temperature and nucleation
rate J Chem Physics 1983 78 3204
16 Viisanen Y Strey R and Reiss H Homogeneous
nucleation rates for water J Chem Physics 1993 99 4680
17 Viisanen Y Strey R and Reiss H Erratum homo-
geneous nucleation rates for water J Chem Physics
2000 112 8205
18 Wolk J and Strey R Homogeneous nucleation of H2O
and D2O in comparison the isotope effect J Phys
Chemistry B 2001 105 11683
19 Heath C H Streletsky K and Wyslouzil B E H2Ondash
D2O condensation in a supersonic nozzle J Chem
Physics 2002 117 6176
20 Mikheev V B Irving P M Laulainen N S Barlow
S E and Pervukhin V V Laboratory measurements of
water nucleation using a laminar flow chamber J Chem
Physics 2002 116 10 772
21 Hung C H Krasnopoler M J and Katz J L
Condensation of a supersaturated vapor VIII The
homogeneous nucleation of n-nonane J Chem Physics
1989 90 1856
22 Rudek M M Fisk J A Chakarov V M and Katz
J L Condensation of a supersaturated vapor XII The
homogeneous nucleation of the n-alkanes J Chem
Physics 1996 105 4707
23 Strey R Wagner P E and Schmeling T Homogeneous
nucleation rates for n-alcohol vapors measured in a two-
piston expansion chamber J Chem Physics 1986 84
2325
24 Lihavainen H Viisanen Y and Kulmala M Homo-
geneous nucleation of n-pentanol in a laminar flow
diffusion chamber J Chem Physics 2001 114 10 031
25 Rudek M M Katz J L Vidensky I V Zdımal V and
Smolık J Homogeneous nucleation rates of n-pentanol
measured in an upward thermal diffusion cloud chamber
J Chem Physics 1999 111 3623
26 Grassmann A and Peters F Homogeneous nucleation
rates of n-propanol in nitrogen measured in a piston-
expansion tube J Chem Physics 2002 116 7617
27 Kane D and El-Shall S Condensation of supersaturated
vapors of hydrogen bonding molecules ethylene glycol
propylene glycol trimethyl glycol and glycerol J Chem
Physics 1996 105 7617
28 Hameri K and Kulmala M Homogeneous nucleation in
a laminar flow chamber the effect of temperature and
carrier gas pressure on dibutyl phthalate vapor nucleation
rate at high supersaturations J Chem Physics 1996 105
7696
29 Viisanen Y Strey R Laaksonen A and Kulmala M
Measurement of the molecular content of binary nuclei
II Use of the nucleation rate surface for waterndashethanol
J Chem Physics 1994 100 6062
30 Viisanen Y Wagner P E and Strey R Measurement of
the molecular content of binary nuclei IV Use of the
nucleation rate surfaces for the n-nonane-n-alcohol series
J Chem Physics 1998 108 4257
31 Fisk J A Rudek M M Katz J L Beiersdorf D and
Uchtmann H The homogeneous nucleation of cesium
vapor Atmos Res 1998 46 211
32 Uchtmann H Dettmer R Baranovskii S D and
Hensel F Photoinduced nucleation in supersaturated
mercury vapor J Chem Physics 1998 108 9775
33 Ferguson F T and Nuth J A Experimental studies of
the vapor phase nucleation of refractory compounds V
The condensation of lithium J Chem Physics 2000 113
4093
34 Laaksonen A Talanquer V and Oxtoby D W
Nucleation measurements theory and atmospheric
applications A Rev Phys Chemistry 1995 46 489
35 Debenedetti P G Metastable Liquids Concepts and
Principles 1996 (Princeton University Press Princeton
New Jersey)
36 Yasuoka K and Matsumoto M Molecular dynamics of
homogeneous nucleation in the vapor phase I Lennardndash
Jones fluid J Chem Physics 1998 109 8451
37 Yasuoka K and Matsumoto M Molecular dynamics of
homogeneous nucleation in the vapor phase II Water
J Chem Physics 1998 109 8463
I J FORD896
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
38 Wonczak S Strey R and Stauffer D Confirmation of
classical nucleation theory by Monte Carlo simulations in
the 3-dimensional Ising model at low temperature
J Chem Physics 2000 113 1976
39 Laasonen K Wonczak S Strey R and Laaksonen A
Molecular dynamics simulations of gasndashliquid nucleation
of LennardndashJones fluid J Chem Physics 2000 113 9741
40 Reiss H Methods of Thermodynamics 1996 (Dover
Publications New York)
41 Kondepudi D and Prigogine I Modern Thermodynamics
From Heat Engines to Dissipative Structures 1998 (John
Wiley Chichester West Sussex)
42 Smoluchowski M Ann Phys NY 1908 25 225
43 Einstein A Ann Phys NY 1910 338 1275
44 Hirschfelder J O Curtis C F and Bird R BMolecular
Theory of Gases and Liquids 1954 (John Wiley New
York)
45 Landau L D and Lifshitz E M Statistical Physics 1958
(Pergamon London)
46 Wu D T Nucleation theory Solid State Physicsmdash
Advances in Res and Applic 1997 50 37
47 Reguera D Rubı J M and Perez-Madrid A Fokkerndash
Planck equations for nucleation processes revisited
Physica A 1998 259 10
48 Schenter G K Kathmann S and Garrett B Variational
transition state theory of vapor phase nucleation
J Chem Physics 1999 109 7951
49 Onsager L Phys Rev 1931 37 405
50 Courtney W G J Chem Physics 1961 35 2249
51 Girshick S L and Chiu C-P Kinetic nucleation theory
a new expression for the rate of homogeneous nucleation
from an ideal supersaturated vapor J Chem Physics
1990 93 1273
52 Ford I J Barrett J C and Lazaridis M Uncertainties
in cluster energies in homogeneous nucleation theory
J Aerosol Sci 1993 24 581
53 Weakliem C L and Reiss H The factor 1S in the
classical theory of nucleation J Phys Chemistry 1994
98 6408
54 McClurg R B and Flagan R C Critical comparison of
droplet models in homogeneous nucleation theory
J Colloid Interface Sci 1998 201 194
55 Mashmoushy H Mahpeykar M R and Bakhtar F
Studies of nucleating and wet steam flows in two-dimen-
sional cascades Proc Instn Mech Engrs Part C J
Mechanical Engineering Science 2004 218(C8) 843ndash858
56 Ford I J Nucleation theorems the statistical mechanics
of molecular clusters and a revision of classical nuclea-
tion theory Phys Rev E 1997 56 5615
57 ten Wolde P R and Frenkel D Computer simulation
study of gasndashliquid nucleation in a LennardndashJones system
J Chem Physics 1998 109 9901
58 Kusaka I Wang Z G and Seinfeld J H Direct
evaluation of the equilibrium distribution of physical
clusters by a grand canonical Monte Carlo simulation
J Chem Physics 1998 108 3416
59 Kusaka I and Oxtoby D W Identifying physical
clusters in vapor phase nucleation J Chem Physics
1999 110 5249
60 Stillinger F H J Chem Physics 1963 38 1486
61 Ford I J Imperfect vaporndashgas mixtures and homo-
geneous nucleation J Aerosol Sci 1992 23 447
62 Lothe J and Pound G M J Chem Physics 1962 36
2080
63 Lee J K Barker J A and Abraham F F Theory and
Monte Carlo simulation of physical clusters in the
imperfect vapor J Chem Physics 1973 58 3166
64 Abraham F F Homogeneous Nucleation Theory 1974
(Academic Press New York)
65 Reif F Fundamentals of Statistical and Thermal Physics
1965 (McGraw-Hill New York)
66 Hoare M R and Pal P Adv Phys 1975 24 645
67 Hoare M R Adv Chem Physics 1979 40 49
68 Pal P and Hoare M R Thermodynamic properties
and homogeneous nucleation of molecular clusters of
nitrogen J Phys Chemistry 1987 91 2474
69 Reiss H Katz J L and Cohen E R Translationndash
rotation paradox in the theory of nucleation J Chem
Physics 1968 48 5553
70 Hale B N Monte Carlo calculations of effective surface
tension for small clusters Aust J Physics 1996 49 425
71 Kathmann S M Schenter G K and Garrett B C
Dynamical nucleation theory calculation of condensation
rate constants for small water clusters J Chem Physics
1999 111 4688
72 Reiss H Tabazadeh A and Talbot J Molecular theory
of vapor phase nucleation the physically consistent
cluster J Chem Physics 1990 92 1266
73 Ellerby H M Weakliem C L and Reiss H Toward a
molecular theory of vaporndashphase nucleation I Identifica-
tion of the average embryo J Chem Physics 1992 97
5766
74 Ellerby H M and Reiss H Toward a molecular theory
of vaporndashphase nucleation II Fundamental treatment of
the cluster distribution J Chem Physics 1992 97 5766
75 Weakliem C L and Reiss H Toward a molecular theory
of vaporndashphase nucleation III Thermodynamic properties
of argon clusters fromMonte Carlo simulations and modi-
fied liquid drop theory J Chem Physics 1994 101 2398
76 Kusaka I Wang Z G and Seinfeld J H Binary
nucleation of sulfuric acidndashwater Monte Carlo simula-
tion J Chem Physics 1998 108 6829
77 Oh K J and Zeng X C A small system ensemble
Monte Carlo simulation of supersaturated vapor
evaluation of barrier to nucleation J Chem Physics
2000 112 294
78 Senger B Schaaf P Corti D S Bowles R Voegel J
C and Reiss H A molecular theory of the homogeneous
nucleation rate I Formulation and fundamental issues
J Chem Physics 1999 110 6421
79 Hill T L Statistical Mechanics Principles and Selected
Applications 1956 (McGraw-Hill New York)
80 Soto R and Cordero P Nonideal gas of clusters at
equilibrium J Chem Physics 1999 110 7316
81 Barrett J C The significance of cluster lifetime in
nucleation theory J Chem Physics 2002 116 8856
82 Lushnikov A A and Kulmala M Dimers in nucleating
vapors Phys Rev E 1998 58 3157
83 Bahadur R and McClurg R B Comparison of cluster
definitions in homogeneous nucleation rate calculations
J Phys Chemistry 2001 105 11 893
84 Pugnaloni L A and Vericat F New criteria for cluster
identification in continuum systems J Chem Physics
2002 116 1097
STATISTICAL MECHANICS OF NUCLEATION 897
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
85 Chen B Siepmann J I Oh K J and Klein M L
Simulating vaporndashliquid nucleation of n-alkanes J Chem
Physics 2002 116 4317
86 Harris S A and Ford I J A dynamical definition of
quasi-bound molecular clusters J Chem Physics 2003
118 9216
87 Shreve A P Walton J R P B and Gubbins K E
Liquid drops of polar molecules J Chem Physics 1986
85 2178
88 ten Wolde P R Ruiz-Montero M J and Frenkel D
Numerical calculation of the rate of homogeneous gasndash
liquid nucleation in a LennardndashJones system J Chem
Physics 1998 109 9901
89 ten Wolde P R Oxtoby D W and Frenkel D Chain
formation in homogeneous gasndashliquid nucleation of polar
fluids J Chem Physics 1998 111 4762
90 Vehkamaki H and Ford I J Critical cluster size and
droplet nucleation rate from growth and decay simula-
tions of LennardndashJones clusters J Chem Physics 2000
112 4193
91 Schaaf P Senger B Voegel J C Bowles R K and
Reiss H Simulative determination of kinetic coefficients
for nucleation rates J Chem Physics 2001 114 8091
92 Chen B Siepmann J I Oh K J and Klein M L
Aggregation-volume-bias Monte Carlo simulations of
vaporndashliquid nucleation barriers for LennardndashJonesium
J Chem Physics 2001 115 10 903
93 Hale B N and DiMattio D J A Monte Carlo discrete
sum (MCDS) nucleation rate model for water In
Proceedings of the 15th International Conference on
Nucleation and Atmospheric Aerosols AIP Conference
Proceedings Vol 534 (Eds B N Hale and M Kulmala)
2000 pp 31ndash34
94 Kathmann S M and Hale B N Monte Carlo
simulations of small sulfuric acidndashwater clusters
J Phys Chemistry B 2001 105 11 719
95 Daura X Mark A E and van Gunsteren W F Para-
metrization of aliphatic CHn united atoms of GROMOS96
force field J Comput Chemistry 1998 19 535
96 Kathmann S M Schenter G K and Garrett B C
Understanding the sensitivity of nucleation kinetics a
case study on water J Chem Physics 2002 116 5046
97 Kashchiev D On the relation between nucleation work
nucleus size and nucleation rate J Chem Physics 1982
76 5098
98 Oxtoby D W and Kashchiev D A general relation
between the nucleation work and the size of the nucleus in
multicomponent nucleation J Chem Physics 1994 100
7665
99 Ford I J Thermodynamic properties of critical clusters
from measurements of vapourndashliquid homogeneous
nucleation rates J Chem Physics 1996 105 8324
100 Oxtoby D W and Evans R Nonclassical nucleation
theory for the gasndashliquid transition J Chem Physics
1988 89 7521
101 Zeng X C and Oxtoby D W Gasndashliquid nucleation in
LennardndashJones fluids J Chem Physics 1991 94 4472
102 Evans R Fundamentals of Inhomogeneous Fluids 1992
Ch 3 pp 85ndash175 (Marcel Dekker New York)
103 Talanquer V and Oxtoby D W Dynamical density
functional theory of gasndashliquid nucleation J Chem
Physics 1994 100 5190
104 Barrett J C First-order correction to classical nucleation
theory a density functional approach J Chem Physics
1999 111 5938
105 Barrett J C Cluster translation and growth in density
functional theories of homogeneous nucleation J Chem
Physics 1997 107 7989
106 Bykov T V and Zeng X C A patching model for
surface tension of spherical droplet and Tolman length J
Chem Physics 1999 111 10 602
107 Talanquer V and Oxtoby D W Nucleation in molecular
and dipolar fluids interaction site model J Chem
Physics 1995 103 3686
108 Seok C and Oxtoby D W Nucleation in n-alkanes a
density functional approach J Chem Physics 1998 109
7982
109 Reguera D Bowles R K Djikaev Y and Reiss H
Phase transitions in systems small enough to be clusters
J Chem Physics 2003 118 340
110 Khakshouri S and Ford I J J Chem Physics 2004
(submitted)
111 Ford I J (unpublished)
112 Nyquist R M Talanquer V and Oxtoby D W Density
functional theory of nucleation a semi-empirical
approach J Chem Physics 1995 103 1175
113 Talanquer V and Oxtoby D W Density functional
analysis of phenomenological theories of gasndashliquid
nucleation J Chem Physics 1995 99 2865
114 Katz J L Fisk J A and Chakarov V The accuracy of
nucleation theory In Proceedings of the 13th Interna-
tional Conference on Nucleation and Atmospheric Aerosols
(Eds N Fukuta and P E Wagner) 1992 pp 11ndash18
(Deepak Publishing Hampton Virginia)
115 Fisher M E The theory of condensation and the critical
point Physics 1967 3 255
116 Dillmann A and Meier G E A A refined droplet
approach to the problem of homogeneous nucleation
from the vapor phase J Chem Physics 1991 94 3872
117 Tolman R C J Chem Physics 1949 17 333
118 Ford I J Laaksonen A and Kulmala M Modification
of the DillmannndashMeier theory of homogeneous nuclea-
tion J Chem Physics 1993 99 764
119 Laaksonen A Ford I J and Kulmala M Revised
parametrization of the DillmannndashMeier theory of homo-
geneous nucleation Phys Rev E 1994 49 5517
120 Kalikmanov V I and van Dongen M E H Semipheno-
menological theory of homogeneous vaporndashliquid nuclea-
tion J Chem Physics 1995 103 4250
121 Hale B N Application of a scaled homogeneous
nucleation-rate formalism to experimental-data at
T 5Tc Phys Rev A 1986 33 4156
122 Martinez D M Ferguson F T Heist R H and Nuth
J A Application of scaled nucleation theory to metallic
vapor condensation J Chem Physics 2001 115 310
123 Hale B N and Wilemski G A scaled nucleation model
for ideal binary systems Chem Physics Lett 1999 305
263
124 Granasy L Diffuse interface theory for homogeneous
vapor condensation J Chem Physics 1996 104
5188
125 McGraw R and Laaksonen A Interfacial curvature free
energy the Kelvin relation and vaporndashliquid nucleation
rate J Chem Physics 1997 106 5284
I J FORD898
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
126 Talanquer V A new phenomenological approach to
gasndashliquid nucleation based on the scaling properties
of the critical nucleus J Chem Physics 1997 106
9957
127 Koga K and Zeng X C Thermodynamic expansion of
nucleation free-energy barrier and size of critical nucleus
near the vaporndashliquid coexistence J Chem Physics 1999
110 3466
128 Kusaka I On the scaling behavior of the free energetics of
nucleation J Chem Physics 2003 118 5510
129 Nowakowski B and Ruckenstein E A kinetic approach
to the theory of nucleation in gases J Chem Physics
1991 94 1397
130 Nowakowski B and Ruckenstein E Homogeneous
nucleation in gases a three-dimensional FokkerndashPlanck
equation for evaporation from clusters J Chem Physics
1991 94 8487
131 Wilcox C F and Bauer S H Estimation of homo-
geneous nucleation flux via a kinetic model J Chem
Physics 1991 94 8302
132 Hanggi P Talkner P and Borkovec M Reaction rate
theory fifty years after Kramers Rev Mod Physics 1990
62 251
133 Barrett J C Clement C F and Ford I J Energy
fluctuations in homogeneous nucleation theory for aero-
sols J Physics A 1993 26 529
134 Ford I J and Harris S A Molecular cluster decay
viewed as escape from a potential of mean force J Chem
Physics 2004 120 4428
135 Risken H The FokkerndashPlanck Equation Methods of
Solution and Applications 2nd edition 1989 (Springer-
Verlag Berlin)
136 Shizgal B and Barrett J C Time dependent nucleation
J Chem Physics 1989 91 6505
STATISTICAL MECHANICS OF NUCLEATION 899
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
reduce to
dni
dtfrac14 bi1ni1 gini bini thorn githorn1nithorn1 eth7THORN
These are the BeckerndashDoring equations [11] and theyprovide a useful reinterpretation of the TST approachThere are two steady state solutions In the first thepopulations satisfy equation (7) and also the detailedbalance condition
bi1nei1 frac14 gin
ei eth8THORN
and the fnei g are taken to represent the cluster popula-tions in an equilibrium vapour at or below the saturatedvapour pressure For true thermal equilibrium themonomer population has to be less than or equal to itsvalue for the saturated vapour ns1 The cluster populationsin a saturated vapour are denoted fnsig Nevertheless asituation can be invented which is known as lsquoconstrainedequilibriumrsquo where the detailed balance condition is stillsatisfied but with n1 gt ns1 Since bi pvS while gi isindependent of vapour pressure equation (8) then definesso-called constrained equilibrium populations
ncei ampSinsi eth9THORN
for a supersaturated vapour It has been assumed thatS frac14 pv=p
svampn1=n
s1
The second steady state solution to the BeckerndashDoring equations is more realistic It is not meant torepresent thermal equilibrium The supersaturation isset to a value greater than unity and a boundarycondition of zero population at some cluster size imax isimposed These conditions will drive a steady productionrate of large clusters from the vapour The approach isfor mathematical convenience only and the same steadystate cluster populations nssi and nucleation rate wouldemerge if imax was sent off to infinity The detailedbalance condition (8) is replaced by J frac14 bi1n
ssi1 gin
ssi
defining the nucleation current J which can then becalculated analytically as shown in another paper in thisvolume [55] To a good approximation
J frac14 Zdbincei eth10THORN
which involves the constrained equilibrium population atthe critical size i the size with the lowest constrainedequilibrium population The condition imax 4 i musthold Zd is the so-called Zeldovich factor [12] givenapproximately by
Zd frac14 1
2pq2ethln ncei THORN
qi2
ifrac14i
1=2
eth11THORN
When equations (4) and (5) are used to specify ncei and theresult is substituted into equations (10) and (11) then theclassical nucleation rate given in equation (1) isrecovered as long as the large i limit of the attachment
rate coefficient in equation (3) is taken It is also possibleto derive an expression for J purely in terms of the ratecoefficients bi and gi This possibility is considered againin section 8
The main assumptions used in this derivation of theclassical theory principally the validity of equations (4)and (5) are now examined In order to explore this it isnecessary to see how the equilibrium cluster populationsnei in subsaturated vapours may be calculated usingstatistical mechanics
4 CLUSTER STATISTICAL MECHANICS
The statistical mechanical treatment of a vapour is bestdeveloped using a grand canonical ensemble [56ndash59] Thesystem is placed in contact with a particle reservoir at achemical potential and temperature T so that thenumber of molecules N in the system can change Thegrand partition function of the system X is then propor-tional to the integral of expf frac12HethNTHORN N=ethkTTHORNg overall possible configurations of molecular positions andmomenta whereHethNTHORN is the Hamiltonian of the system
Evaluating X exactly is exceedingly difficult and it ismuch simpler to proceed by considering the configura-tions of the system as arrangements of molecularclusters The vapour then is viewed as a mixture ofideal gases each composed of clusters of differentsizes The grand partition function of the system canbe reconstructed from partition functions for singlei-clusters In order to do this however a cluster must bedefined in a unique way in terms of molecular positionsand velocities An example of such a definition would beto require the separation between molecules in a clusterto be less than a maximum distance Rc [60] as illustratedin Fig 3
Fig 3 According to the Stillinger definition a molecule isconsidered part of a cluster if it lies closer than adistance Rc from a constituent molecule of the cluster
I J FORD888
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
At first sight the cluster definition is completelyarbitrary a means of categorizing and resumming allthe elementary configurations of the system Howeveraccurately representing the system as a mixture of idealgases requires the interactions between molecules lyingin different clusters to be minimized The clusterdefinition affects whether a particular interaction isregarded as intercluster or intracluster and so it plays acrucial role This will be discussed further in section 6Residual clusterndashcluster interactions can be taken intoaccount in the form of a virial series [61]
These ideas are now put into practice Define amodified canonical partition function zi for a clustercontaining i molecules This is simply the integral ofexpf frac12HcethiTHORN i=ethkTTHORNg over configurations involvingmolecules in the cluster alone with Hc the Hamiltonianof the i molecules comprising the cluster Using thepicture of a mixture of ideal gases the grand partitionfunction of the complete system can then be expressed as
XampXfnig
Yifrac141
znii
ni
eth12THORN
where the sum is over all possible cluster populationdistributions fnig This is demonstrated rigorously inappendix A of reference [57]
The next step is to determine the distribution fnei gthat dominates the sum in equation (12) The logarithmof the summand is extremized
qqni
Xifrac141
lnz
nii
ni
frac14 0 eth13THORN
which gives the following very simple expression for themost probable or equilibrium cluster size distributionfor the given conditions T and
nei frac14 ziethTTHORN frac14 ZethTTHORN ei=ethkTTHORN eth14THORN
which is equivalent to the so-called law of mass action
nei frac14 Zine1Z1
i
eth15THORN
where ZiethTTHORN frac14 expfrac12FethiTHORN=ethkTTHORN frac14 zi expfrac12 i=ethkTTHORN isthe canonical partition function for an i cluster and FethiTHORNis the i-cluster Helmholtz free energy This relationshipis often used as the starting point in models of clusterpopulations
The vapour pressure is now given by a sum of partialpressures
pvV frac14 kTXifrac141
nei frac14 kTXifrac141
ziethTTHORN eth16THORN
and this series will only converge if the equilibriumpopulations or the zi decrease sufficiently quickly forlarge i This limits the statistical mechanics approach to
the study of saturated or subsaturated vapours How-ever as in the preceding section the equilibriumpopulations nei in equation (14) (for S lt 1) can beextrapolated into constrained equilibrium populationsncei (with S gt 1) for use in equation (10) Saturatedvapour corresponds to a particular chemical potentialms and the saturated vapour cluster populations are nsi The constrained equilibrium populations are obtainedby inserting into equation (14) the chemical potential ofthe supersaturated vapour which is frac14 s thorn kT lnSgiving ncei frac14 Sinsi which is consistent with the resultobtained in equation (9) for dilute vapours
The constrained equilibrium populations are there-fore expressed as
ncei frac14 SiziethsTTHORN
frac14 expfrac12FethiTHORN is ikT lnS
kT
eth17THORN
and the correspondence with CNT can be made closerby introducing the n1 prefactor
ncei frac14 n1 exp frac12FethiTHORN is thorn kT ln n1 ikT lnS
kT
eth18THORN
The next task is to explore how the combinationFethiTHORN is thorn kT ln n1 might be approximated by theCNT result 4pR2s This can be explored most easilyfor solid rather than liquid clusters
5 SOLID CLUSTERS
The partition function of a solid cluster of i moleculescan be factorized into translational rotational andvibrational parts through a transformation to suitablerelative positions and momenta [62ndash64] Starting withthe vibrational degrees of freedom the atom positionsare defined with respect to their mean positions whichare determined by a fixed centre of mass and set oflattice vectors It is assumed that the potential energypart of the Hamiltonian is quadratic in the displace-ments from these mean positions Now the partitionfunction is worked out for configurations of the particlesrestricted to having zero total linear and total angularmomentum about the centre of mass This is writtenZvib corresponding to a product of partition functionsfor 3ji 6 independent oscillators (if there are j atomsper molecule and therefore a total of 3ji degrees offreedom if the molecule is not linear) each of which inthe classical limit takes the form kT=ethhoTHORN where o isthe angular frequency of the specific vibrational modeThe partition function has an additional factor ofexpfrac12U0ethiTHORN=ethkTTHORN where U0ethiTHORN is the system potentialenergy evaluated at the mean atom positions A
STATISTICAL MECHANICS OF NUCLEATION 889
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
different expression applies for low temperatures whenthe oscillator is not classically activated [65]
The rotational and translational degrees of freedomare next The previous restriction of the configurationsto zero angular momentum is lifted The approximate(classical) rotational partition function is
Zrot frac14 p1=28p2kTh2
3=2
ethI1I2I3THORN1=2 eth19THORN
where Ik are the principal moments of inertia about thecentre of mass for atoms fixed in their mean positions
Finally the translational free energy of the centre ofmass of the cluster is given by the partition function
Ztrans frac14 V2pMkT
h2
3=2
eth20THORN
where V is the system volume and M frac14 im is the clustermass The free energy of the cluster (plus kT ln n1 forconvenience) is therefore
FethiTHORN thorn kT ln n1
frac14 kT lnZtransZrotZvib
n1
frac14 U0ethiTHORN kT ln
eth2pMkTTHORN3=2
Srsvh3p1=2
8p2kTh2
3=2
6ethI1I2I3THORN1=2Y3ji6
kfrac141
kT
hok
eth21THORN
where rsv frac14 ns1=V is the molecular density of dilutesaturated vapour
By similar arguments the chemical potential of dilutesaturated vapour is
s frac14 Um0 kT ln
eth2pmkTTHORN3=2
rsvh3p1=2
8p2kTh2
3=2
6ethIm1 Im2 Im3 THORN1=2Y3j6
kfrac141
kT
homk
eth22THORN
where Um0 is the potential energy of a molecule when the
atoms are at their mean positions Imk are the threeprincipal moments of inertia of the molecule about thecentre of mass and om
k are the oscillatory frequencies ofmolecular internal vibrations
The classical work of formation DWclethiTHORN should nowcorrespond to the leading-order contribution to thedifference between equation (21) and i times equation(22) Calculations [54 64 66ndash68] demonstrate that theleading term is indeed roughly proportional to i2=3 thusjustifying the form of the classical model The suitabilityof the capillarity approximation depends on the extentto which the coefficient of this leading term correspondsnumerically to the planar surface free energy Howeverthe full statistical mechanical expression for the work of
formation contains additional terms and these will beimportant for small clusters and need to be taken intoaccount In order to evaluate them the variousstructural and energy properties of cluster are requiredall of which can be obtained from microscopic modelsand they can then be inserted into the formulae for thethermodynamic properties [equations (21) and (22)] theconstrained equilibrium populations [equation (18)] andultimately the nucleation rate [equation (10)] Followingthis route employs the almost universally held assump-tion that equilibrium properties such as free energies canbe used to describe non-equilibrium properties specifi-cally the cluster decay rate This assumption is re-examined in section 8
6 LIQUID DROPLETS AND THE CLUSTER
DEFINITION
Defining what was meant by a solid cluster in theprevious section was not an issue Within the assump-tions of the model (harmonic interactions) it wasimpossible for a molecule to escape from the clusterThe cluster definition in the case of liquids cannot bedismissed so easily Liquids are more volatile thansolids and the molecules in a cluster are much moremobile It is imperative now to characterize a clusterusing the molecular positions and velocities andperhaps some time scale and length scale too
The traditional microscopic method for defining acluster is to impose a geometric constraint on molecularpositions An example is to restrict molecules to a spherecentred on the centre of mass [69ndash71] or to impose amaximum allowable separation between the molecules(the Stillinger criterion [60]) More sophisticated treat-ments exist such as the n=v Stillinger cluster introducedby Reiss and co-workers [72ndash75] Here the confiningvolume v used to define the cluster is considered to be anadditional defining characteristic on the same level asthe cluster size n As time progresses a cluster changesits confining volume as the component molecules followtheir respective trajectories just as they change their sizewith the gain and loss of molecules The association of amolecular configuration with a precise confining volumeis complicated but tractable The dependence of the freeenergy on v determines the statistical dynamics of thiscluster volume A similar attempt to associate a clusterwith a specified volume within a grand canonicalensemble description was developed by Kusaka andco-workers [58 59 76] The simulation box was taken tobe the confining cluster volume Many of these ideasmay be viewed as a treatment of the cluster definition(eg the Stillinger radius) as a variational parameter inthe representation of the partition function Further
I J FORD890
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
refinement of this point of view using a treatment ofclusterndashcluster interactions to optimize the definingcluster volume has been presented [77] A review ofrecent progress in this area and the application of theseideas to nucleation theory has been provided by Sengeret al [78]
Such geometric constraints can be implemented buttheir realism is questionable Situations better regardedas close encounters between separate clusters would beclassed as snapshots of single larger clusters It shouldbe recalled that the cluster definition is an internalfeature of the model which is to be optimized to reduceclusterndashcluster interactions and to improve the repre-sentation of the grand partition function Misclassifyingconfigurations is fraught with consequences A secondissue to consider is that for convenience the clustersshould behave in a manner that can be modelled withBeckerndashDoring equations This means that the growthand decay rates should be free of memory effects ie thetransition processes should be Markovian The extent towhich this is satisfied (if indeed it is physically realistic)depends on the cluster definition
An intuitively reasonable choice of definition whichaddresses both these points is that a cluster shouldconsist of molecules that are bound energetically Thisshould minimize the clusterndashcluster contributions to thetotal energy According to such a definition if amolecule interacts strongly with a cluster then it islikely to be regarded as part of that cluster Also anenergetically bound cluster is likely to decay by thermalfluctuation in a stochastic Markovian fashion ratherthan by the separation of passing poorly boundfragments
Ideas have been proposed which require the cluster tobe energetically bound (in a loose sense since a cluster isintrinsically prone to decay) The time evolution of amolecular configuration and not simply the set ofinstantaneous positions of its component molecules isof central importance Hill [79] introduced a clusterdefinition requiring the total energies of pairs ofmolecules in a cluster to be negative (in their centre ofmass frame) The intention was to exclude situationswhere a molecule could leave the cluster within the timeit would take for a molecule to travel a typical lineardimension of the cluster Soto and Cordero [80] havedeveloped Hillrsquos criterion and Barrett [81] has per-formed Monte Carlo modelling of molecular configura-tions excluding those which decay in moleculardynamics within a given period A number of otherenergy-based criteria have been developed [82ndash85]
Harris and Ford [86] have recently proposed a newenergy-based cluster definition As with Barrettrsquosapproach it is based on the requirement that amolecular snapshot is energetically bound for a chosensubsequent period Evaporation according to thisscheme requires not only that a molecule should makea positive energy excursion but also that the subsequent
dynamics carry it far away from the cluster Onlyperhaps one in eight molecules manage to avoidrecapture once having satisfied the positive energycriterion [86] A snapshot of a cluster in this casesurrounded by a carrier gas is shown in Fig 4 Thehistory of the energy of a molecule in such a cluster isillustrated in Fig 5 three excursions into positive energyoccur in this sequence but only one leads to actualescape corresponding to the maintenance of positiveenergy for a significant time The instant of escape istaken to be the point at which the energy goes positivefor the third and final time
Cluster definitions implemented in moleculardynamics and Monte Carlo codes allow cluster freeenergies and hence the decay rate to be calculated Anenormous amount of simulation work has been carriedout and calculations can be laborious Many differentschemes and approximations have been made to easethe evaluation of these properties The principalconclusion from various studies using a range of clusterdefinitions and interaction models [49 57 58 70 85 87ndash92] seems to be that the work of formation of liquidcritical clusters does indeed contain a dominantcontribution proportional to i2=3 consistent with theclassical work of formation Hale and collaboratorspresent cluster free energies for water and other speciesin a way that makes this particularly clear indeed it issurprising how dominant the surface term can be foreven very small clusters [70 93 94] Neverthelessfurther contributions exist small clusters are notcharacterized just by scaled-down properties of largeclusters and it is impossible to escape the need todetermine the additional terms
In all the above discussion it has been assumed thatthe intermolecular interactions are accurately specifiedThis is a situation that is rarely possible Althoughfirst-principles quantum-mechanical models could be
Fig 4 Snapshot of a cluster of 50 argon atoms (brightspheres) bathed in a gas of helium (darker spheres)
STATISTICAL MECHANICS OF NUCLEATION 891
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
developed in practice empirical potentials are devel-oped on the basis of physical intuition containingparameters fitted to the measured properties of thesubstance in question [95] Usually these are bulkproperties and it is not always clear that the sameinteractions can be used for the very different circum-stances of a molecular cluster [96] In view of thesensitivity of nucleation rates to the thermodynamicproperties of clusters this is perhaps a crucial failingProgress can be made however by computing knowncluster properties where they are available and byrefining the potentials to suit Extraction of clusterproperties through use of the so-called nucleationtheorems (which relate derivatives of the nucleationrate to properties of the critical cluster [16 56 97ndash99])may provide just this sort of experimental information
A technique that has not yet been described is the useof density functional theory (DFT) to model theproperties of liquid droplets [100ndash104] This powerfulapplication of statistical mechanics promises to avoidlaborious computations at the microscopic level byinstead focusing on the density profile of moleculesmaking up the droplet This profile may be regarded asthe average distribution of molecules about the clustercentre of mass Although this might appear to involvecoarse-graining and approximation it is a well-foundedapproach and in principle can be exact In practical
implementations however a mean field approach istaken such that the constituent molecules move in aneffective potential well created by the other molecules inthe cluster [105] Also the interactions are assumed to becharacterized by strong repulsive cores and weaklyattractive tails
Considerable analytical work has been carried out[104 106] Various developments have been made todescribe droplets consisting of simple atoms as well aschain-like molecules [107] and polar molecules [108]The main drawback in these implementations is thatspecific interactions such as hydrogen bonds are hardto represent Certain details of the modelled systemare neglected [109 110] and the (implicit) clusterdefinition used is geometric and not energetic [111]On the other hand quite sophisticated systems havebeen modelled including multi-component dropletsThe method has been compared with otherapproaches [112 113] and certain features are betterdescribed particularly the approach to spinodalconditions where the nucleation barrier vanishesDFT calculations can be performed relatively easilyand they can provide a qualitative and perhapsquantitative insight into quite complicated clusteringproblems They provide an intermediate-level descrip-tion between the simple capillarity approximation anda microscopic description
Fig 5 The energy of a molecule in the centre-of-mass frame of a cluster occasionally becomes positive butthis does not necessarily lead to escape Only the third positive energy excursion in this sequenceindicated by an arrow results in particle escape A cluster definition must capture this behaviour
I J FORD892
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
7 PHENOMENOLOGICAL MODELS
The reader might be dismayed by the mathematicalcomplexity of the expressions for the work of formationof solid clusters in section 5 might be alarmed by theprocedural complexity of implementing a cluster defini-tion in section 6 may recoil from the difficulties incalculating free energies from molecular simulationmethods and ultimately may give up in view of theuncertainties in the underlying intermolecular interac-tions Considering these complications it is tempting totry to develop phenomenological models of the work offormation After all if the capillarity approximationworks reasonably well for water then surely a minorcorrection to account for small cluster effects ought tobe sufficient to improve matters
However after perhaps two decades of work nomodel has emerged offering broad success across arange of different substances It is perhaps asking toomuch to expect universal improvement CNT works as afirst approximation for water for a particular tempera-ture range but as discussed earlier it possesses thewrong temperature dependence so that predictionsaway from a temperature in the range 210ndash260K areuncertain A similar picture emerges for other sub-stances although in some cases the failure is even moresevere
The simplest phenomenological adjustment to equa-tion (4) the basis for CNT is to write
ncei frac14 n1 expfrac12DWclethiTHORN DWcleth1THORN
kT
eth23THORN
where DWclethiTHORN is given by equation (5) This amounts toa uniform downward shift of the classical free energybarrier illustrated in Fig 1 At least now the expressionis consistent at i frac14 1 The theory is known as internallyconsistent classical theory (ICCT) [51] Unfortunately itis not clear whether this constant shift in the work offormation should be a realistic representation of all thesubleading terms in the work of formation [52] It is anarbitrary modification and its success in accounting fordata is mixed For example nucleation rates for a rangeof alkanes [114] require further modification by factorsof between one and nine orders of magnitude to comeinto agreement with ICCT
It should be noted at this point that the correctionterm kT lnS which appears in DWcleth1THORN can beregarded as rigorous There are good reasons for thisbased on general considerations involving the nucleationtheorems [56] or more simply the principles of chemicalequilibrium represented by the law of mass action givenin equation (15) The correction converts the prefactorn1 in equation (4) into ns1 and results in the insertion of afactor 1=S in the nucleation rate in equation (1)
The Fisher droplet model is a slightly more sophis-ticated scheme [115] Here the constrained equilibrium
populations are modelled by
ncei frac14 ns1 exp yii2=3 thorn t ln i lnq0kT
psv
thorn i lnS
eth24THORN
The term proportional to yi may be viewed as a surfaceterm while the other two are corrections of a typesuggested by the free energy expressions developed insection 5 The parameters q0 yi and t are chosen toensure that the model correctly reproduces knownproperties of clusters and vapours including virialcoefficients and critical properties the theory is there-fore phenomenological
For large cluster sizes and for temperatures wellbelow the critical temperature the free energy offormation should tend towards that of a macroscopicliquid droplet yi should therefore approach A1s=ethkTTHORNfor large i where A1i
2=3 is the surface area 4pR2 of aspherical droplet containing i monomers at the bulkliquid density so that A1 frac14 eth36pTHORN1=3r2=3
l Dillmann andMeier [116] suggested writing
yi frac14kiA1skT
eth25THORN
with the Ansatz
ki frac14 1thorn a1i 1=3 thorn a2i 2=3 eth26THORN
The coefficients a1 and a2 can be found in terms of virialcoefficients of the vapour The form of this Ansatz wasmotivated by the curvature dependence of the bulksurface tension derived in classical thermodynamics byTolman [117]
The initial success of the DillmannndashMeier theory wasshort lived since modifications had to be made toovercome an internal inconsistency [118] Virial expan-sions for the vapour pressure and density were usedincorrectly to calculate the virial expansion of the ratioof the two Correcting this error [61] changed predictedrates by several orders of magnitude Nevertheless theapproach was very influential [119] In a furtherdevelopment Kalikmanov and van Dongen [120] seta2 frac14 0 (this term can be absorbed into q0 anyway) andchose a1 such that the experimental saturated vapourpressure psv is reproduced by the model The DillmannndashMeier model matched the first two virial coefficientsonly The Kalikmanovndashvan Dongen theory shouldtherefore approximate the (revised) DillmannndashMeiertheory at low temperatures but ought to provide abetter description at higher temperatures
Other phenomenological models have been proposedincluding the Hale scaling relation [121] This modelbased on CNT but employing simplified models for thesurface tension and liquid density based on scalingbehaviour near the critical point together with a rather
STATISTICAL MECHANICS OF NUCLEATION 893
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
more straightforward rate prefactor seems to be quitesuccessful in accounting for data for a wide range ofmaterials even metals [122] and two-component systems[123] Another phenomenological model is the diffuseinterface theory [124] which is based on a treatment ofthe phase interface using simplified density profiles ofthermodynamic properties Another general class ofmodel makes use of conditions at the vapourndashliquidspinodal to constrain the form of the model free energy[125ndash128]
The main drawback in phenomenological models isthe arbitrariness in the form of the proposed work offormation The parameter fitting in practice ensures thatthe populations of small clusters especially the mono-mer and dimer are well represented by the model butthe properties of larger clusters depend on the expres-sion assumed for the model and are therefore uncertainIntuition suggests that phenomenological modelling islikely to have mixed success and that microscopiccalculations of cluster properties do matter There mightbe better value in phenomenological models however ifthey were based securely on the results of microscopiccalculations
8 KINETICS OR THERMODYNAMICS
In section 3 kinetic equations representing the nuclea-tion process were described However the coefficients inthose equations were related to constrained equilibriumpopulations and hence to the equilibrium statisticalmechanics of clusters through the detailed balancecondition (8) Is it necessary to do this Could acompletely kinetic route be followed by evaluating themean growth and decay coefficients bi and gi directly bysome statistical approach The main problem is ofcourse the difficulty in performing such calculationsand this has restricted the development of such models[129ndash131] It is expected that the kinetic coefficients willcorrespond to those derived from the detailed balanceargument for near-equilibrium conditions but circum-stances can be imagined where non-equilibrium effectsare present which cannot be dealt with within standardstatistical mechanics
The framework for evaluating kinetic coefficients ofcluster decay can be briefly introduced [86 132] Amolecular dynamics simulation of a cluster can be set upand evolved according to the interactions until bychance an evaporation event occurs By extensivesimulation statistics for the mean evaporation rates gican be amassed in this way Similarly moleculardynamics simulations where a single molecule iscaptured by a cluster could be performed and thegrowth rates bi determined These will be functions ofcluster energy and so would fit best into a rate equationscheme involving both size and energy [133]
This is a laborious approach but statistical physicsoffers an approximate alternative representation of theescape and capture processes The basic tool is theLangevin equation The radial motion of a molecule inthe cluster is described by [134]
meurorr frac14 f ethrTHORN thornfethr _rr tTHORN eth27THORN
where r is the radial position m is the molecular mass fis the mean (time- and velocity-averaged) force on themolecule at position r and f is the stochastic forceThe latter is usually represented as a velocity-dependent friction term mg _rr where g is the so-calledfriction coefficient plus a fluctuating term ~ff ethr tTHORN withzero mean and correlation in time h ~ff ethr tTHORN~ff ethr t0THORNi frac14eth2gkTmTHORNdetht t0THORN where k is Boltzmannrsquos constant andwhere T has the characteristics of a temperature as willbe seen It is a standard manipulation [135] to convertthe Langevin description in the large friction limit intoa FokkerndashPlanck equation
qwqt
frac14 1
mg qethfwTHORN
qrthorn kT
q2wqr2
eth28THORN
which represents the evolution of wethr tTHORN the prob-ability density that the molecule should lie at the radialposition r It should be noted in passing that FokkerndashPlanck equations have often been used to describe thestochastic dynamics of a cluster in size space [48 136] Incontrast the use of the method described above refers tomotion of a molecule in real coordinate space relative tothe centre of mass of the cluster
The steady state equilibrium solution of equation (28)is
wethrTHORN exp FethrTHORNkT
eth29THORN
where FethrTHORN is called the potential of mean force relatedto the mean force on the molecule f throughf frac14 dF=dr It can now be seen how the parameter Tin the stochastic force plays the role of temperaturesince equation (29) looks like a Boltzmann distributionThis solution is not what is sought however The escapeproblem is characterized by a boundary conditionwethreTHORN frac14 0 where re is the radius at which a particleescapes from the system The resulting escape rate maybe shown to be [134]
gkin frac14 KethreT m gTHORN exp DFkT
eth30THORN
where the prefactor K depends on the parameters listedas well as the shape of the potential of mean force Theprincipal feature of equation (30) is the exponentialdependence on the depth DF of the potential of meanforce Therefore if the potential of mean force FethrTHORN can
I J FORD894
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
be evaluated by studying a molecular dynamics trajec-tory for example then the kinetic decay rate can bedetermined This could be extended to account for aradially dependent temperature and friction coefficientboth of which might be determined from simulation Asketch of a potential of mean force is given in Fig 6
Kinetically determined cluster growth and decayrates gkini and bkini respectively could provide a newroute for the determination of nucleation rates inde-pendent of assumptions about detailed balance and theuse of equilibrium statistical mechanics in a non-equilibrium situation
9 CONCLUSIONS
This review has been brief and in some ways partisan tothe personal views of the author It attempts to sketch acoherent picture of nucleation theory and how thevarious disparate strands relate to one another One ofthe fascinations of nucleation theory is the variety ofapproaches and the number of difficult issues that areencountered How well does free energy fluctuationtheory represent a real non-equilibrium process Howwell does the idea of detailed balance provide aframework for determining non-equilibrium decaycoefficients from equilibrium statistical mechanicsHow well do various methods perform in the evaluation
of cluster free energies Can cluster growth and decayrates be calculated directly without the need for such aframework How exactly are clusters to be defined
The theoretical problems are many but the issues canbe resolved to various degrees of satisfaction Thereremains of course the fundamental task of choosing anintermolecular potential that adequately represents thephysical interactions Without this all the finer points ofcluster definition and treatments of non-equilibriumprocesses are unimportant Modelling many problems inscience relies on the availability of an accurate micro-scopic description and so nucleation rate calculationsare not unique in this respect Most empirical molecularpotentials are fitted to the properties of bulk and surfaceproperties and in the very best examples attention isgiven to their transferability to more unusual environ-ments such as clusters In view of these capabilities andmodern theoretical developments in nucleation theoryit seems likely that successful microscopic models of thenucleation process based on statistical physics are atlast coming within reach
ACKNOWLEDGEMENTS
The present author is grateful to colleagues SarahHarris and Marco Arosio for providing some of thefigures
Fig 6 A molecule in a cluster may be considered to be confined within it by a potential of mean force FethrTHORN Itcan escape by reaching a radial position re driven by a phenomenological stochastic force
STATISTICAL MECHANICS OF NUCLEATION 895
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
REFERENCES
1 Gross D H E Microcanonical Thermodynamics 2001
(World Scientific Singapore)
2 Seinfeld J H and Pandis S N Atmospheric Chemistry
and Physics 1998 (John Wiley New York)
3 Climate Change 2001 Synthesis Report 2001 (Inter-
governmental Panel on Climate Change) (Cambridge
University Press Cambridge)
4 Kohler T M Gail H-P and Sedlmayr E MgO dust
nucleation in M-stars calculation of cluster properties
and nucleation rates Astronomy and Astrophysics 1997
320 553
5 Kashchiev D Nucleation Basic Theory with Applications
2000 (ButterworthndashHeinemann Oxford)
6 Bakhtar F Ebrahimi M and Webb R A On the
performance of a cascade of turbine rotor tip section
blading in nucleating steam Part 1 surface pressure
distributions Proc Instn Mech Engrs 1995 209(C2)
115ndash124
7 Bakhtar F Ebrahimi M and Bamkole B O On the
performance of a cascade of turbine rotor tip section
blading in nucleating steam Part 2 wake traverses Proc
Instn Mech Engrs 1995 209(C3) 169ndash178
8 Bakhtar F andMahpeykar M R On the performance of
a cascade of turbine rotor tip section blading in nucleating
steam Part 3 theoretical treatment Proc Instn Mech
Engrs Part C J Mechanical Engineering Science 1997
211(C3) 195ndash210
9 Volmer M and Weber A Z Physikalische Chemie 1926
119 277
10 Farkas L Z Physikalische Chemie (Leipzig) 1927 125
236
11 Becker R and Doring W Kinetische Behandlung der
Keinbildung in ubersattigten Dampfen Ann Physik
(Leipzig) 1935 24 719
12 Zeldovich J Soviet Physics JETP 1942 12 525
13 Frenkel J Kinetic Theory of Liquids 1946 (Oxford
University Press Oxford)
14 Wagner P E and Strey R Homogeneous nucleation
rates of water vapor measured in a two-piston expansion
chamber J Phys Chemistry 1981 85 2694
15 Miller R Anderson R J Kassner J L and Hagen
D E Homogeneous nucleation rate measurements for
water over a wide range of temperature and nucleation
rate J Chem Physics 1983 78 3204
16 Viisanen Y Strey R and Reiss H Homogeneous
nucleation rates for water J Chem Physics 1993 99 4680
17 Viisanen Y Strey R and Reiss H Erratum homo-
geneous nucleation rates for water J Chem Physics
2000 112 8205
18 Wolk J and Strey R Homogeneous nucleation of H2O
and D2O in comparison the isotope effect J Phys
Chemistry B 2001 105 11683
19 Heath C H Streletsky K and Wyslouzil B E H2Ondash
D2O condensation in a supersonic nozzle J Chem
Physics 2002 117 6176
20 Mikheev V B Irving P M Laulainen N S Barlow
S E and Pervukhin V V Laboratory measurements of
water nucleation using a laminar flow chamber J Chem
Physics 2002 116 10 772
21 Hung C H Krasnopoler M J and Katz J L
Condensation of a supersaturated vapor VIII The
homogeneous nucleation of n-nonane J Chem Physics
1989 90 1856
22 Rudek M M Fisk J A Chakarov V M and Katz
J L Condensation of a supersaturated vapor XII The
homogeneous nucleation of the n-alkanes J Chem
Physics 1996 105 4707
23 Strey R Wagner P E and Schmeling T Homogeneous
nucleation rates for n-alcohol vapors measured in a two-
piston expansion chamber J Chem Physics 1986 84
2325
24 Lihavainen H Viisanen Y and Kulmala M Homo-
geneous nucleation of n-pentanol in a laminar flow
diffusion chamber J Chem Physics 2001 114 10 031
25 Rudek M M Katz J L Vidensky I V Zdımal V and
Smolık J Homogeneous nucleation rates of n-pentanol
measured in an upward thermal diffusion cloud chamber
J Chem Physics 1999 111 3623
26 Grassmann A and Peters F Homogeneous nucleation
rates of n-propanol in nitrogen measured in a piston-
expansion tube J Chem Physics 2002 116 7617
27 Kane D and El-Shall S Condensation of supersaturated
vapors of hydrogen bonding molecules ethylene glycol
propylene glycol trimethyl glycol and glycerol J Chem
Physics 1996 105 7617
28 Hameri K and Kulmala M Homogeneous nucleation in
a laminar flow chamber the effect of temperature and
carrier gas pressure on dibutyl phthalate vapor nucleation
rate at high supersaturations J Chem Physics 1996 105
7696
29 Viisanen Y Strey R Laaksonen A and Kulmala M
Measurement of the molecular content of binary nuclei
II Use of the nucleation rate surface for waterndashethanol
J Chem Physics 1994 100 6062
30 Viisanen Y Wagner P E and Strey R Measurement of
the molecular content of binary nuclei IV Use of the
nucleation rate surfaces for the n-nonane-n-alcohol series
J Chem Physics 1998 108 4257
31 Fisk J A Rudek M M Katz J L Beiersdorf D and
Uchtmann H The homogeneous nucleation of cesium
vapor Atmos Res 1998 46 211
32 Uchtmann H Dettmer R Baranovskii S D and
Hensel F Photoinduced nucleation in supersaturated
mercury vapor J Chem Physics 1998 108 9775
33 Ferguson F T and Nuth J A Experimental studies of
the vapor phase nucleation of refractory compounds V
The condensation of lithium J Chem Physics 2000 113
4093
34 Laaksonen A Talanquer V and Oxtoby D W
Nucleation measurements theory and atmospheric
applications A Rev Phys Chemistry 1995 46 489
35 Debenedetti P G Metastable Liquids Concepts and
Principles 1996 (Princeton University Press Princeton
New Jersey)
36 Yasuoka K and Matsumoto M Molecular dynamics of
homogeneous nucleation in the vapor phase I Lennardndash
Jones fluid J Chem Physics 1998 109 8451
37 Yasuoka K and Matsumoto M Molecular dynamics of
homogeneous nucleation in the vapor phase II Water
J Chem Physics 1998 109 8463
I J FORD896
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
38 Wonczak S Strey R and Stauffer D Confirmation of
classical nucleation theory by Monte Carlo simulations in
the 3-dimensional Ising model at low temperature
J Chem Physics 2000 113 1976
39 Laasonen K Wonczak S Strey R and Laaksonen A
Molecular dynamics simulations of gasndashliquid nucleation
of LennardndashJones fluid J Chem Physics 2000 113 9741
40 Reiss H Methods of Thermodynamics 1996 (Dover
Publications New York)
41 Kondepudi D and Prigogine I Modern Thermodynamics
From Heat Engines to Dissipative Structures 1998 (John
Wiley Chichester West Sussex)
42 Smoluchowski M Ann Phys NY 1908 25 225
43 Einstein A Ann Phys NY 1910 338 1275
44 Hirschfelder J O Curtis C F and Bird R BMolecular
Theory of Gases and Liquids 1954 (John Wiley New
York)
45 Landau L D and Lifshitz E M Statistical Physics 1958
(Pergamon London)
46 Wu D T Nucleation theory Solid State Physicsmdash
Advances in Res and Applic 1997 50 37
47 Reguera D Rubı J M and Perez-Madrid A Fokkerndash
Planck equations for nucleation processes revisited
Physica A 1998 259 10
48 Schenter G K Kathmann S and Garrett B Variational
transition state theory of vapor phase nucleation
J Chem Physics 1999 109 7951
49 Onsager L Phys Rev 1931 37 405
50 Courtney W G J Chem Physics 1961 35 2249
51 Girshick S L and Chiu C-P Kinetic nucleation theory
a new expression for the rate of homogeneous nucleation
from an ideal supersaturated vapor J Chem Physics
1990 93 1273
52 Ford I J Barrett J C and Lazaridis M Uncertainties
in cluster energies in homogeneous nucleation theory
J Aerosol Sci 1993 24 581
53 Weakliem C L and Reiss H The factor 1S in the
classical theory of nucleation J Phys Chemistry 1994
98 6408
54 McClurg R B and Flagan R C Critical comparison of
droplet models in homogeneous nucleation theory
J Colloid Interface Sci 1998 201 194
55 Mashmoushy H Mahpeykar M R and Bakhtar F
Studies of nucleating and wet steam flows in two-dimen-
sional cascades Proc Instn Mech Engrs Part C J
Mechanical Engineering Science 2004 218(C8) 843ndash858
56 Ford I J Nucleation theorems the statistical mechanics
of molecular clusters and a revision of classical nuclea-
tion theory Phys Rev E 1997 56 5615
57 ten Wolde P R and Frenkel D Computer simulation
study of gasndashliquid nucleation in a LennardndashJones system
J Chem Physics 1998 109 9901
58 Kusaka I Wang Z G and Seinfeld J H Direct
evaluation of the equilibrium distribution of physical
clusters by a grand canonical Monte Carlo simulation
J Chem Physics 1998 108 3416
59 Kusaka I and Oxtoby D W Identifying physical
clusters in vapor phase nucleation J Chem Physics
1999 110 5249
60 Stillinger F H J Chem Physics 1963 38 1486
61 Ford I J Imperfect vaporndashgas mixtures and homo-
geneous nucleation J Aerosol Sci 1992 23 447
62 Lothe J and Pound G M J Chem Physics 1962 36
2080
63 Lee J K Barker J A and Abraham F F Theory and
Monte Carlo simulation of physical clusters in the
imperfect vapor J Chem Physics 1973 58 3166
64 Abraham F F Homogeneous Nucleation Theory 1974
(Academic Press New York)
65 Reif F Fundamentals of Statistical and Thermal Physics
1965 (McGraw-Hill New York)
66 Hoare M R and Pal P Adv Phys 1975 24 645
67 Hoare M R Adv Chem Physics 1979 40 49
68 Pal P and Hoare M R Thermodynamic properties
and homogeneous nucleation of molecular clusters of
nitrogen J Phys Chemistry 1987 91 2474
69 Reiss H Katz J L and Cohen E R Translationndash
rotation paradox in the theory of nucleation J Chem
Physics 1968 48 5553
70 Hale B N Monte Carlo calculations of effective surface
tension for small clusters Aust J Physics 1996 49 425
71 Kathmann S M Schenter G K and Garrett B C
Dynamical nucleation theory calculation of condensation
rate constants for small water clusters J Chem Physics
1999 111 4688
72 Reiss H Tabazadeh A and Talbot J Molecular theory
of vapor phase nucleation the physically consistent
cluster J Chem Physics 1990 92 1266
73 Ellerby H M Weakliem C L and Reiss H Toward a
molecular theory of vaporndashphase nucleation I Identifica-
tion of the average embryo J Chem Physics 1992 97
5766
74 Ellerby H M and Reiss H Toward a molecular theory
of vaporndashphase nucleation II Fundamental treatment of
the cluster distribution J Chem Physics 1992 97 5766
75 Weakliem C L and Reiss H Toward a molecular theory
of vaporndashphase nucleation III Thermodynamic properties
of argon clusters fromMonte Carlo simulations and modi-
fied liquid drop theory J Chem Physics 1994 101 2398
76 Kusaka I Wang Z G and Seinfeld J H Binary
nucleation of sulfuric acidndashwater Monte Carlo simula-
tion J Chem Physics 1998 108 6829
77 Oh K J and Zeng X C A small system ensemble
Monte Carlo simulation of supersaturated vapor
evaluation of barrier to nucleation J Chem Physics
2000 112 294
78 Senger B Schaaf P Corti D S Bowles R Voegel J
C and Reiss H A molecular theory of the homogeneous
nucleation rate I Formulation and fundamental issues
J Chem Physics 1999 110 6421
79 Hill T L Statistical Mechanics Principles and Selected
Applications 1956 (McGraw-Hill New York)
80 Soto R and Cordero P Nonideal gas of clusters at
equilibrium J Chem Physics 1999 110 7316
81 Barrett J C The significance of cluster lifetime in
nucleation theory J Chem Physics 2002 116 8856
82 Lushnikov A A and Kulmala M Dimers in nucleating
vapors Phys Rev E 1998 58 3157
83 Bahadur R and McClurg R B Comparison of cluster
definitions in homogeneous nucleation rate calculations
J Phys Chemistry 2001 105 11 893
84 Pugnaloni L A and Vericat F New criteria for cluster
identification in continuum systems J Chem Physics
2002 116 1097
STATISTICAL MECHANICS OF NUCLEATION 897
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
85 Chen B Siepmann J I Oh K J and Klein M L
Simulating vaporndashliquid nucleation of n-alkanes J Chem
Physics 2002 116 4317
86 Harris S A and Ford I J A dynamical definition of
quasi-bound molecular clusters J Chem Physics 2003
118 9216
87 Shreve A P Walton J R P B and Gubbins K E
Liquid drops of polar molecules J Chem Physics 1986
85 2178
88 ten Wolde P R Ruiz-Montero M J and Frenkel D
Numerical calculation of the rate of homogeneous gasndash
liquid nucleation in a LennardndashJones system J Chem
Physics 1998 109 9901
89 ten Wolde P R Oxtoby D W and Frenkel D Chain
formation in homogeneous gasndashliquid nucleation of polar
fluids J Chem Physics 1998 111 4762
90 Vehkamaki H and Ford I J Critical cluster size and
droplet nucleation rate from growth and decay simula-
tions of LennardndashJones clusters J Chem Physics 2000
112 4193
91 Schaaf P Senger B Voegel J C Bowles R K and
Reiss H Simulative determination of kinetic coefficients
for nucleation rates J Chem Physics 2001 114 8091
92 Chen B Siepmann J I Oh K J and Klein M L
Aggregation-volume-bias Monte Carlo simulations of
vaporndashliquid nucleation barriers for LennardndashJonesium
J Chem Physics 2001 115 10 903
93 Hale B N and DiMattio D J A Monte Carlo discrete
sum (MCDS) nucleation rate model for water In
Proceedings of the 15th International Conference on
Nucleation and Atmospheric Aerosols AIP Conference
Proceedings Vol 534 (Eds B N Hale and M Kulmala)
2000 pp 31ndash34
94 Kathmann S M and Hale B N Monte Carlo
simulations of small sulfuric acidndashwater clusters
J Phys Chemistry B 2001 105 11 719
95 Daura X Mark A E and van Gunsteren W F Para-
metrization of aliphatic CHn united atoms of GROMOS96
force field J Comput Chemistry 1998 19 535
96 Kathmann S M Schenter G K and Garrett B C
Understanding the sensitivity of nucleation kinetics a
case study on water J Chem Physics 2002 116 5046
97 Kashchiev D On the relation between nucleation work
nucleus size and nucleation rate J Chem Physics 1982
76 5098
98 Oxtoby D W and Kashchiev D A general relation
between the nucleation work and the size of the nucleus in
multicomponent nucleation J Chem Physics 1994 100
7665
99 Ford I J Thermodynamic properties of critical clusters
from measurements of vapourndashliquid homogeneous
nucleation rates J Chem Physics 1996 105 8324
100 Oxtoby D W and Evans R Nonclassical nucleation
theory for the gasndashliquid transition J Chem Physics
1988 89 7521
101 Zeng X C and Oxtoby D W Gasndashliquid nucleation in
LennardndashJones fluids J Chem Physics 1991 94 4472
102 Evans R Fundamentals of Inhomogeneous Fluids 1992
Ch 3 pp 85ndash175 (Marcel Dekker New York)
103 Talanquer V and Oxtoby D W Dynamical density
functional theory of gasndashliquid nucleation J Chem
Physics 1994 100 5190
104 Barrett J C First-order correction to classical nucleation
theory a density functional approach J Chem Physics
1999 111 5938
105 Barrett J C Cluster translation and growth in density
functional theories of homogeneous nucleation J Chem
Physics 1997 107 7989
106 Bykov T V and Zeng X C A patching model for
surface tension of spherical droplet and Tolman length J
Chem Physics 1999 111 10 602
107 Talanquer V and Oxtoby D W Nucleation in molecular
and dipolar fluids interaction site model J Chem
Physics 1995 103 3686
108 Seok C and Oxtoby D W Nucleation in n-alkanes a
density functional approach J Chem Physics 1998 109
7982
109 Reguera D Bowles R K Djikaev Y and Reiss H
Phase transitions in systems small enough to be clusters
J Chem Physics 2003 118 340
110 Khakshouri S and Ford I J J Chem Physics 2004
(submitted)
111 Ford I J (unpublished)
112 Nyquist R M Talanquer V and Oxtoby D W Density
functional theory of nucleation a semi-empirical
approach J Chem Physics 1995 103 1175
113 Talanquer V and Oxtoby D W Density functional
analysis of phenomenological theories of gasndashliquid
nucleation J Chem Physics 1995 99 2865
114 Katz J L Fisk J A and Chakarov V The accuracy of
nucleation theory In Proceedings of the 13th Interna-
tional Conference on Nucleation and Atmospheric Aerosols
(Eds N Fukuta and P E Wagner) 1992 pp 11ndash18
(Deepak Publishing Hampton Virginia)
115 Fisher M E The theory of condensation and the critical
point Physics 1967 3 255
116 Dillmann A and Meier G E A A refined droplet
approach to the problem of homogeneous nucleation
from the vapor phase J Chem Physics 1991 94 3872
117 Tolman R C J Chem Physics 1949 17 333
118 Ford I J Laaksonen A and Kulmala M Modification
of the DillmannndashMeier theory of homogeneous nuclea-
tion J Chem Physics 1993 99 764
119 Laaksonen A Ford I J and Kulmala M Revised
parametrization of the DillmannndashMeier theory of homo-
geneous nucleation Phys Rev E 1994 49 5517
120 Kalikmanov V I and van Dongen M E H Semipheno-
menological theory of homogeneous vaporndashliquid nuclea-
tion J Chem Physics 1995 103 4250
121 Hale B N Application of a scaled homogeneous
nucleation-rate formalism to experimental-data at
T 5Tc Phys Rev A 1986 33 4156
122 Martinez D M Ferguson F T Heist R H and Nuth
J A Application of scaled nucleation theory to metallic
vapor condensation J Chem Physics 2001 115 310
123 Hale B N and Wilemski G A scaled nucleation model
for ideal binary systems Chem Physics Lett 1999 305
263
124 Granasy L Diffuse interface theory for homogeneous
vapor condensation J Chem Physics 1996 104
5188
125 McGraw R and Laaksonen A Interfacial curvature free
energy the Kelvin relation and vaporndashliquid nucleation
rate J Chem Physics 1997 106 5284
I J FORD898
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
126 Talanquer V A new phenomenological approach to
gasndashliquid nucleation based on the scaling properties
of the critical nucleus J Chem Physics 1997 106
9957
127 Koga K and Zeng X C Thermodynamic expansion of
nucleation free-energy barrier and size of critical nucleus
near the vaporndashliquid coexistence J Chem Physics 1999
110 3466
128 Kusaka I On the scaling behavior of the free energetics of
nucleation J Chem Physics 2003 118 5510
129 Nowakowski B and Ruckenstein E A kinetic approach
to the theory of nucleation in gases J Chem Physics
1991 94 1397
130 Nowakowski B and Ruckenstein E Homogeneous
nucleation in gases a three-dimensional FokkerndashPlanck
equation for evaporation from clusters J Chem Physics
1991 94 8487
131 Wilcox C F and Bauer S H Estimation of homo-
geneous nucleation flux via a kinetic model J Chem
Physics 1991 94 8302
132 Hanggi P Talkner P and Borkovec M Reaction rate
theory fifty years after Kramers Rev Mod Physics 1990
62 251
133 Barrett J C Clement C F and Ford I J Energy
fluctuations in homogeneous nucleation theory for aero-
sols J Physics A 1993 26 529
134 Ford I J and Harris S A Molecular cluster decay
viewed as escape from a potential of mean force J Chem
Physics 2004 120 4428
135 Risken H The FokkerndashPlanck Equation Methods of
Solution and Applications 2nd edition 1989 (Springer-
Verlag Berlin)
136 Shizgal B and Barrett J C Time dependent nucleation
J Chem Physics 1989 91 6505
STATISTICAL MECHANICS OF NUCLEATION 899
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
At first sight the cluster definition is completelyarbitrary a means of categorizing and resumming allthe elementary configurations of the system Howeveraccurately representing the system as a mixture of idealgases requires the interactions between molecules lyingin different clusters to be minimized The clusterdefinition affects whether a particular interaction isregarded as intercluster or intracluster and so it plays acrucial role This will be discussed further in section 6Residual clusterndashcluster interactions can be taken intoaccount in the form of a virial series [61]
These ideas are now put into practice Define amodified canonical partition function zi for a clustercontaining i molecules This is simply the integral ofexpf frac12HcethiTHORN i=ethkTTHORNg over configurations involvingmolecules in the cluster alone with Hc the Hamiltonianof the i molecules comprising the cluster Using thepicture of a mixture of ideal gases the grand partitionfunction of the complete system can then be expressed as
XampXfnig
Yifrac141
znii
ni
eth12THORN
where the sum is over all possible cluster populationdistributions fnig This is demonstrated rigorously inappendix A of reference [57]
The next step is to determine the distribution fnei gthat dominates the sum in equation (12) The logarithmof the summand is extremized
qqni
Xifrac141
lnz
nii
ni
frac14 0 eth13THORN
which gives the following very simple expression for themost probable or equilibrium cluster size distributionfor the given conditions T and
nei frac14 ziethTTHORN frac14 ZethTTHORN ei=ethkTTHORN eth14THORN
which is equivalent to the so-called law of mass action
nei frac14 Zine1Z1
i
eth15THORN
where ZiethTTHORN frac14 expfrac12FethiTHORN=ethkTTHORN frac14 zi expfrac12 i=ethkTTHORN isthe canonical partition function for an i cluster and FethiTHORNis the i-cluster Helmholtz free energy This relationshipis often used as the starting point in models of clusterpopulations
The vapour pressure is now given by a sum of partialpressures
pvV frac14 kTXifrac141
nei frac14 kTXifrac141
ziethTTHORN eth16THORN
and this series will only converge if the equilibriumpopulations or the zi decrease sufficiently quickly forlarge i This limits the statistical mechanics approach to
the study of saturated or subsaturated vapours How-ever as in the preceding section the equilibriumpopulations nei in equation (14) (for S lt 1) can beextrapolated into constrained equilibrium populationsncei (with S gt 1) for use in equation (10) Saturatedvapour corresponds to a particular chemical potentialms and the saturated vapour cluster populations are nsi The constrained equilibrium populations are obtainedby inserting into equation (14) the chemical potential ofthe supersaturated vapour which is frac14 s thorn kT lnSgiving ncei frac14 Sinsi which is consistent with the resultobtained in equation (9) for dilute vapours
The constrained equilibrium populations are there-fore expressed as
ncei frac14 SiziethsTTHORN
frac14 expfrac12FethiTHORN is ikT lnS
kT
eth17THORN
and the correspondence with CNT can be made closerby introducing the n1 prefactor
ncei frac14 n1 exp frac12FethiTHORN is thorn kT ln n1 ikT lnS
kT
eth18THORN
The next task is to explore how the combinationFethiTHORN is thorn kT ln n1 might be approximated by theCNT result 4pR2s This can be explored most easilyfor solid rather than liquid clusters
5 SOLID CLUSTERS
The partition function of a solid cluster of i moleculescan be factorized into translational rotational andvibrational parts through a transformation to suitablerelative positions and momenta [62ndash64] Starting withthe vibrational degrees of freedom the atom positionsare defined with respect to their mean positions whichare determined by a fixed centre of mass and set oflattice vectors It is assumed that the potential energypart of the Hamiltonian is quadratic in the displace-ments from these mean positions Now the partitionfunction is worked out for configurations of the particlesrestricted to having zero total linear and total angularmomentum about the centre of mass This is writtenZvib corresponding to a product of partition functionsfor 3ji 6 independent oscillators (if there are j atomsper molecule and therefore a total of 3ji degrees offreedom if the molecule is not linear) each of which inthe classical limit takes the form kT=ethhoTHORN where o isthe angular frequency of the specific vibrational modeThe partition function has an additional factor ofexpfrac12U0ethiTHORN=ethkTTHORN where U0ethiTHORN is the system potentialenergy evaluated at the mean atom positions A
STATISTICAL MECHANICS OF NUCLEATION 889
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
different expression applies for low temperatures whenthe oscillator is not classically activated [65]
The rotational and translational degrees of freedomare next The previous restriction of the configurationsto zero angular momentum is lifted The approximate(classical) rotational partition function is
Zrot frac14 p1=28p2kTh2
3=2
ethI1I2I3THORN1=2 eth19THORN
where Ik are the principal moments of inertia about thecentre of mass for atoms fixed in their mean positions
Finally the translational free energy of the centre ofmass of the cluster is given by the partition function
Ztrans frac14 V2pMkT
h2
3=2
eth20THORN
where V is the system volume and M frac14 im is the clustermass The free energy of the cluster (plus kT ln n1 forconvenience) is therefore
FethiTHORN thorn kT ln n1
frac14 kT lnZtransZrotZvib
n1
frac14 U0ethiTHORN kT ln
eth2pMkTTHORN3=2
Srsvh3p1=2
8p2kTh2
3=2
6ethI1I2I3THORN1=2Y3ji6
kfrac141
kT
hok
eth21THORN
where rsv frac14 ns1=V is the molecular density of dilutesaturated vapour
By similar arguments the chemical potential of dilutesaturated vapour is
s frac14 Um0 kT ln
eth2pmkTTHORN3=2
rsvh3p1=2
8p2kTh2
3=2
6ethIm1 Im2 Im3 THORN1=2Y3j6
kfrac141
kT
homk
eth22THORN
where Um0 is the potential energy of a molecule when the
atoms are at their mean positions Imk are the threeprincipal moments of inertia of the molecule about thecentre of mass and om
k are the oscillatory frequencies ofmolecular internal vibrations
The classical work of formation DWclethiTHORN should nowcorrespond to the leading-order contribution to thedifference between equation (21) and i times equation(22) Calculations [54 64 66ndash68] demonstrate that theleading term is indeed roughly proportional to i2=3 thusjustifying the form of the classical model The suitabilityof the capillarity approximation depends on the extentto which the coefficient of this leading term correspondsnumerically to the planar surface free energy Howeverthe full statistical mechanical expression for the work of
formation contains additional terms and these will beimportant for small clusters and need to be taken intoaccount In order to evaluate them the variousstructural and energy properties of cluster are requiredall of which can be obtained from microscopic modelsand they can then be inserted into the formulae for thethermodynamic properties [equations (21) and (22)] theconstrained equilibrium populations [equation (18)] andultimately the nucleation rate [equation (10)] Followingthis route employs the almost universally held assump-tion that equilibrium properties such as free energies canbe used to describe non-equilibrium properties specifi-cally the cluster decay rate This assumption is re-examined in section 8
6 LIQUID DROPLETS AND THE CLUSTER
DEFINITION
Defining what was meant by a solid cluster in theprevious section was not an issue Within the assump-tions of the model (harmonic interactions) it wasimpossible for a molecule to escape from the clusterThe cluster definition in the case of liquids cannot bedismissed so easily Liquids are more volatile thansolids and the molecules in a cluster are much moremobile It is imperative now to characterize a clusterusing the molecular positions and velocities andperhaps some time scale and length scale too
The traditional microscopic method for defining acluster is to impose a geometric constraint on molecularpositions An example is to restrict molecules to a spherecentred on the centre of mass [69ndash71] or to impose amaximum allowable separation between the molecules(the Stillinger criterion [60]) More sophisticated treat-ments exist such as the n=v Stillinger cluster introducedby Reiss and co-workers [72ndash75] Here the confiningvolume v used to define the cluster is considered to be anadditional defining characteristic on the same level asthe cluster size n As time progresses a cluster changesits confining volume as the component molecules followtheir respective trajectories just as they change their sizewith the gain and loss of molecules The association of amolecular configuration with a precise confining volumeis complicated but tractable The dependence of the freeenergy on v determines the statistical dynamics of thiscluster volume A similar attempt to associate a clusterwith a specified volume within a grand canonicalensemble description was developed by Kusaka andco-workers [58 59 76] The simulation box was taken tobe the confining cluster volume Many of these ideasmay be viewed as a treatment of the cluster definition(eg the Stillinger radius) as a variational parameter inthe representation of the partition function Further
I J FORD890
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
refinement of this point of view using a treatment ofclusterndashcluster interactions to optimize the definingcluster volume has been presented [77] A review ofrecent progress in this area and the application of theseideas to nucleation theory has been provided by Sengeret al [78]
Such geometric constraints can be implemented buttheir realism is questionable Situations better regardedas close encounters between separate clusters would beclassed as snapshots of single larger clusters It shouldbe recalled that the cluster definition is an internalfeature of the model which is to be optimized to reduceclusterndashcluster interactions and to improve the repre-sentation of the grand partition function Misclassifyingconfigurations is fraught with consequences A secondissue to consider is that for convenience the clustersshould behave in a manner that can be modelled withBeckerndashDoring equations This means that the growthand decay rates should be free of memory effects ie thetransition processes should be Markovian The extent towhich this is satisfied (if indeed it is physically realistic)depends on the cluster definition
An intuitively reasonable choice of definition whichaddresses both these points is that a cluster shouldconsist of molecules that are bound energetically Thisshould minimize the clusterndashcluster contributions to thetotal energy According to such a definition if amolecule interacts strongly with a cluster then it islikely to be regarded as part of that cluster Also anenergetically bound cluster is likely to decay by thermalfluctuation in a stochastic Markovian fashion ratherthan by the separation of passing poorly boundfragments
Ideas have been proposed which require the cluster tobe energetically bound (in a loose sense since a cluster isintrinsically prone to decay) The time evolution of amolecular configuration and not simply the set ofinstantaneous positions of its component molecules isof central importance Hill [79] introduced a clusterdefinition requiring the total energies of pairs ofmolecules in a cluster to be negative (in their centre ofmass frame) The intention was to exclude situationswhere a molecule could leave the cluster within the timeit would take for a molecule to travel a typical lineardimension of the cluster Soto and Cordero [80] havedeveloped Hillrsquos criterion and Barrett [81] has per-formed Monte Carlo modelling of molecular configura-tions excluding those which decay in moleculardynamics within a given period A number of otherenergy-based criteria have been developed [82ndash85]
Harris and Ford [86] have recently proposed a newenergy-based cluster definition As with Barrettrsquosapproach it is based on the requirement that amolecular snapshot is energetically bound for a chosensubsequent period Evaporation according to thisscheme requires not only that a molecule should makea positive energy excursion but also that the subsequent
dynamics carry it far away from the cluster Onlyperhaps one in eight molecules manage to avoidrecapture once having satisfied the positive energycriterion [86] A snapshot of a cluster in this casesurrounded by a carrier gas is shown in Fig 4 Thehistory of the energy of a molecule in such a cluster isillustrated in Fig 5 three excursions into positive energyoccur in this sequence but only one leads to actualescape corresponding to the maintenance of positiveenergy for a significant time The instant of escape istaken to be the point at which the energy goes positivefor the third and final time
Cluster definitions implemented in moleculardynamics and Monte Carlo codes allow cluster freeenergies and hence the decay rate to be calculated Anenormous amount of simulation work has been carriedout and calculations can be laborious Many differentschemes and approximations have been made to easethe evaluation of these properties The principalconclusion from various studies using a range of clusterdefinitions and interaction models [49 57 58 70 85 87ndash92] seems to be that the work of formation of liquidcritical clusters does indeed contain a dominantcontribution proportional to i2=3 consistent with theclassical work of formation Hale and collaboratorspresent cluster free energies for water and other speciesin a way that makes this particularly clear indeed it issurprising how dominant the surface term can be foreven very small clusters [70 93 94] Neverthelessfurther contributions exist small clusters are notcharacterized just by scaled-down properties of largeclusters and it is impossible to escape the need todetermine the additional terms
In all the above discussion it has been assumed thatthe intermolecular interactions are accurately specifiedThis is a situation that is rarely possible Althoughfirst-principles quantum-mechanical models could be
Fig 4 Snapshot of a cluster of 50 argon atoms (brightspheres) bathed in a gas of helium (darker spheres)
STATISTICAL MECHANICS OF NUCLEATION 891
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
developed in practice empirical potentials are devel-oped on the basis of physical intuition containingparameters fitted to the measured properties of thesubstance in question [95] Usually these are bulkproperties and it is not always clear that the sameinteractions can be used for the very different circum-stances of a molecular cluster [96] In view of thesensitivity of nucleation rates to the thermodynamicproperties of clusters this is perhaps a crucial failingProgress can be made however by computing knowncluster properties where they are available and byrefining the potentials to suit Extraction of clusterproperties through use of the so-called nucleationtheorems (which relate derivatives of the nucleationrate to properties of the critical cluster [16 56 97ndash99])may provide just this sort of experimental information
A technique that has not yet been described is the useof density functional theory (DFT) to model theproperties of liquid droplets [100ndash104] This powerfulapplication of statistical mechanics promises to avoidlaborious computations at the microscopic level byinstead focusing on the density profile of moleculesmaking up the droplet This profile may be regarded asthe average distribution of molecules about the clustercentre of mass Although this might appear to involvecoarse-graining and approximation it is a well-foundedapproach and in principle can be exact In practical
implementations however a mean field approach istaken such that the constituent molecules move in aneffective potential well created by the other molecules inthe cluster [105] Also the interactions are assumed to becharacterized by strong repulsive cores and weaklyattractive tails
Considerable analytical work has been carried out[104 106] Various developments have been made todescribe droplets consisting of simple atoms as well aschain-like molecules [107] and polar molecules [108]The main drawback in these implementations is thatspecific interactions such as hydrogen bonds are hardto represent Certain details of the modelled systemare neglected [109 110] and the (implicit) clusterdefinition used is geometric and not energetic [111]On the other hand quite sophisticated systems havebeen modelled including multi-component dropletsThe method has been compared with otherapproaches [112 113] and certain features are betterdescribed particularly the approach to spinodalconditions where the nucleation barrier vanishesDFT calculations can be performed relatively easilyand they can provide a qualitative and perhapsquantitative insight into quite complicated clusteringproblems They provide an intermediate-level descrip-tion between the simple capillarity approximation anda microscopic description
Fig 5 The energy of a molecule in the centre-of-mass frame of a cluster occasionally becomes positive butthis does not necessarily lead to escape Only the third positive energy excursion in this sequenceindicated by an arrow results in particle escape A cluster definition must capture this behaviour
I J FORD892
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
7 PHENOMENOLOGICAL MODELS
The reader might be dismayed by the mathematicalcomplexity of the expressions for the work of formationof solid clusters in section 5 might be alarmed by theprocedural complexity of implementing a cluster defini-tion in section 6 may recoil from the difficulties incalculating free energies from molecular simulationmethods and ultimately may give up in view of theuncertainties in the underlying intermolecular interac-tions Considering these complications it is tempting totry to develop phenomenological models of the work offormation After all if the capillarity approximationworks reasonably well for water then surely a minorcorrection to account for small cluster effects ought tobe sufficient to improve matters
However after perhaps two decades of work nomodel has emerged offering broad success across arange of different substances It is perhaps asking toomuch to expect universal improvement CNT works as afirst approximation for water for a particular tempera-ture range but as discussed earlier it possesses thewrong temperature dependence so that predictionsaway from a temperature in the range 210ndash260K areuncertain A similar picture emerges for other sub-stances although in some cases the failure is even moresevere
The simplest phenomenological adjustment to equa-tion (4) the basis for CNT is to write
ncei frac14 n1 expfrac12DWclethiTHORN DWcleth1THORN
kT
eth23THORN
where DWclethiTHORN is given by equation (5) This amounts toa uniform downward shift of the classical free energybarrier illustrated in Fig 1 At least now the expressionis consistent at i frac14 1 The theory is known as internallyconsistent classical theory (ICCT) [51] Unfortunately itis not clear whether this constant shift in the work offormation should be a realistic representation of all thesubleading terms in the work of formation [52] It is anarbitrary modification and its success in accounting fordata is mixed For example nucleation rates for a rangeof alkanes [114] require further modification by factorsof between one and nine orders of magnitude to comeinto agreement with ICCT
It should be noted at this point that the correctionterm kT lnS which appears in DWcleth1THORN can beregarded as rigorous There are good reasons for thisbased on general considerations involving the nucleationtheorems [56] or more simply the principles of chemicalequilibrium represented by the law of mass action givenin equation (15) The correction converts the prefactorn1 in equation (4) into ns1 and results in the insertion of afactor 1=S in the nucleation rate in equation (1)
The Fisher droplet model is a slightly more sophis-ticated scheme [115] Here the constrained equilibrium
populations are modelled by
ncei frac14 ns1 exp yii2=3 thorn t ln i lnq0kT
psv
thorn i lnS
eth24THORN
The term proportional to yi may be viewed as a surfaceterm while the other two are corrections of a typesuggested by the free energy expressions developed insection 5 The parameters q0 yi and t are chosen toensure that the model correctly reproduces knownproperties of clusters and vapours including virialcoefficients and critical properties the theory is there-fore phenomenological
For large cluster sizes and for temperatures wellbelow the critical temperature the free energy offormation should tend towards that of a macroscopicliquid droplet yi should therefore approach A1s=ethkTTHORNfor large i where A1i
2=3 is the surface area 4pR2 of aspherical droplet containing i monomers at the bulkliquid density so that A1 frac14 eth36pTHORN1=3r2=3
l Dillmann andMeier [116] suggested writing
yi frac14kiA1skT
eth25THORN
with the Ansatz
ki frac14 1thorn a1i 1=3 thorn a2i 2=3 eth26THORN
The coefficients a1 and a2 can be found in terms of virialcoefficients of the vapour The form of this Ansatz wasmotivated by the curvature dependence of the bulksurface tension derived in classical thermodynamics byTolman [117]
The initial success of the DillmannndashMeier theory wasshort lived since modifications had to be made toovercome an internal inconsistency [118] Virial expan-sions for the vapour pressure and density were usedincorrectly to calculate the virial expansion of the ratioof the two Correcting this error [61] changed predictedrates by several orders of magnitude Nevertheless theapproach was very influential [119] In a furtherdevelopment Kalikmanov and van Dongen [120] seta2 frac14 0 (this term can be absorbed into q0 anyway) andchose a1 such that the experimental saturated vapourpressure psv is reproduced by the model The DillmannndashMeier model matched the first two virial coefficientsonly The Kalikmanovndashvan Dongen theory shouldtherefore approximate the (revised) DillmannndashMeiertheory at low temperatures but ought to provide abetter description at higher temperatures
Other phenomenological models have been proposedincluding the Hale scaling relation [121] This modelbased on CNT but employing simplified models for thesurface tension and liquid density based on scalingbehaviour near the critical point together with a rather
STATISTICAL MECHANICS OF NUCLEATION 893
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
more straightforward rate prefactor seems to be quitesuccessful in accounting for data for a wide range ofmaterials even metals [122] and two-component systems[123] Another phenomenological model is the diffuseinterface theory [124] which is based on a treatment ofthe phase interface using simplified density profiles ofthermodynamic properties Another general class ofmodel makes use of conditions at the vapourndashliquidspinodal to constrain the form of the model free energy[125ndash128]
The main drawback in phenomenological models isthe arbitrariness in the form of the proposed work offormation The parameter fitting in practice ensures thatthe populations of small clusters especially the mono-mer and dimer are well represented by the model butthe properties of larger clusters depend on the expres-sion assumed for the model and are therefore uncertainIntuition suggests that phenomenological modelling islikely to have mixed success and that microscopiccalculations of cluster properties do matter There mightbe better value in phenomenological models however ifthey were based securely on the results of microscopiccalculations
8 KINETICS OR THERMODYNAMICS
In section 3 kinetic equations representing the nuclea-tion process were described However the coefficients inthose equations were related to constrained equilibriumpopulations and hence to the equilibrium statisticalmechanics of clusters through the detailed balancecondition (8) Is it necessary to do this Could acompletely kinetic route be followed by evaluating themean growth and decay coefficients bi and gi directly bysome statistical approach The main problem is ofcourse the difficulty in performing such calculationsand this has restricted the development of such models[129ndash131] It is expected that the kinetic coefficients willcorrespond to those derived from the detailed balanceargument for near-equilibrium conditions but circum-stances can be imagined where non-equilibrium effectsare present which cannot be dealt with within standardstatistical mechanics
The framework for evaluating kinetic coefficients ofcluster decay can be briefly introduced [86 132] Amolecular dynamics simulation of a cluster can be set upand evolved according to the interactions until bychance an evaporation event occurs By extensivesimulation statistics for the mean evaporation rates gican be amassed in this way Similarly moleculardynamics simulations where a single molecule iscaptured by a cluster could be performed and thegrowth rates bi determined These will be functions ofcluster energy and so would fit best into a rate equationscheme involving both size and energy [133]
This is a laborious approach but statistical physicsoffers an approximate alternative representation of theescape and capture processes The basic tool is theLangevin equation The radial motion of a molecule inthe cluster is described by [134]
meurorr frac14 f ethrTHORN thornfethr _rr tTHORN eth27THORN
where r is the radial position m is the molecular mass fis the mean (time- and velocity-averaged) force on themolecule at position r and f is the stochastic forceThe latter is usually represented as a velocity-dependent friction term mg _rr where g is the so-calledfriction coefficient plus a fluctuating term ~ff ethr tTHORN withzero mean and correlation in time h ~ff ethr tTHORN~ff ethr t0THORNi frac14eth2gkTmTHORNdetht t0THORN where k is Boltzmannrsquos constant andwhere T has the characteristics of a temperature as willbe seen It is a standard manipulation [135] to convertthe Langevin description in the large friction limit intoa FokkerndashPlanck equation
qwqt
frac14 1
mg qethfwTHORN
qrthorn kT
q2wqr2
eth28THORN
which represents the evolution of wethr tTHORN the prob-ability density that the molecule should lie at the radialposition r It should be noted in passing that FokkerndashPlanck equations have often been used to describe thestochastic dynamics of a cluster in size space [48 136] Incontrast the use of the method described above refers tomotion of a molecule in real coordinate space relative tothe centre of mass of the cluster
The steady state equilibrium solution of equation (28)is
wethrTHORN exp FethrTHORNkT
eth29THORN
where FethrTHORN is called the potential of mean force relatedto the mean force on the molecule f throughf frac14 dF=dr It can now be seen how the parameter Tin the stochastic force plays the role of temperaturesince equation (29) looks like a Boltzmann distributionThis solution is not what is sought however The escapeproblem is characterized by a boundary conditionwethreTHORN frac14 0 where re is the radius at which a particleescapes from the system The resulting escape rate maybe shown to be [134]
gkin frac14 KethreT m gTHORN exp DFkT
eth30THORN
where the prefactor K depends on the parameters listedas well as the shape of the potential of mean force Theprincipal feature of equation (30) is the exponentialdependence on the depth DF of the potential of meanforce Therefore if the potential of mean force FethrTHORN can
I J FORD894
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
be evaluated by studying a molecular dynamics trajec-tory for example then the kinetic decay rate can bedetermined This could be extended to account for aradially dependent temperature and friction coefficientboth of which might be determined from simulation Asketch of a potential of mean force is given in Fig 6
Kinetically determined cluster growth and decayrates gkini and bkini respectively could provide a newroute for the determination of nucleation rates inde-pendent of assumptions about detailed balance and theuse of equilibrium statistical mechanics in a non-equilibrium situation
9 CONCLUSIONS
This review has been brief and in some ways partisan tothe personal views of the author It attempts to sketch acoherent picture of nucleation theory and how thevarious disparate strands relate to one another One ofthe fascinations of nucleation theory is the variety ofapproaches and the number of difficult issues that areencountered How well does free energy fluctuationtheory represent a real non-equilibrium process Howwell does the idea of detailed balance provide aframework for determining non-equilibrium decaycoefficients from equilibrium statistical mechanicsHow well do various methods perform in the evaluation
of cluster free energies Can cluster growth and decayrates be calculated directly without the need for such aframework How exactly are clusters to be defined
The theoretical problems are many but the issues canbe resolved to various degrees of satisfaction Thereremains of course the fundamental task of choosing anintermolecular potential that adequately represents thephysical interactions Without this all the finer points ofcluster definition and treatments of non-equilibriumprocesses are unimportant Modelling many problems inscience relies on the availability of an accurate micro-scopic description and so nucleation rate calculationsare not unique in this respect Most empirical molecularpotentials are fitted to the properties of bulk and surfaceproperties and in the very best examples attention isgiven to their transferability to more unusual environ-ments such as clusters In view of these capabilities andmodern theoretical developments in nucleation theoryit seems likely that successful microscopic models of thenucleation process based on statistical physics are atlast coming within reach
ACKNOWLEDGEMENTS
The present author is grateful to colleagues SarahHarris and Marco Arosio for providing some of thefigures
Fig 6 A molecule in a cluster may be considered to be confined within it by a potential of mean force FethrTHORN Itcan escape by reaching a radial position re driven by a phenomenological stochastic force
STATISTICAL MECHANICS OF NUCLEATION 895
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
REFERENCES
1 Gross D H E Microcanonical Thermodynamics 2001
(World Scientific Singapore)
2 Seinfeld J H and Pandis S N Atmospheric Chemistry
and Physics 1998 (John Wiley New York)
3 Climate Change 2001 Synthesis Report 2001 (Inter-
governmental Panel on Climate Change) (Cambridge
University Press Cambridge)
4 Kohler T M Gail H-P and Sedlmayr E MgO dust
nucleation in M-stars calculation of cluster properties
and nucleation rates Astronomy and Astrophysics 1997
320 553
5 Kashchiev D Nucleation Basic Theory with Applications
2000 (ButterworthndashHeinemann Oxford)
6 Bakhtar F Ebrahimi M and Webb R A On the
performance of a cascade of turbine rotor tip section
blading in nucleating steam Part 1 surface pressure
distributions Proc Instn Mech Engrs 1995 209(C2)
115ndash124
7 Bakhtar F Ebrahimi M and Bamkole B O On the
performance of a cascade of turbine rotor tip section
blading in nucleating steam Part 2 wake traverses Proc
Instn Mech Engrs 1995 209(C3) 169ndash178
8 Bakhtar F andMahpeykar M R On the performance of
a cascade of turbine rotor tip section blading in nucleating
steam Part 3 theoretical treatment Proc Instn Mech
Engrs Part C J Mechanical Engineering Science 1997
211(C3) 195ndash210
9 Volmer M and Weber A Z Physikalische Chemie 1926
119 277
10 Farkas L Z Physikalische Chemie (Leipzig) 1927 125
236
11 Becker R and Doring W Kinetische Behandlung der
Keinbildung in ubersattigten Dampfen Ann Physik
(Leipzig) 1935 24 719
12 Zeldovich J Soviet Physics JETP 1942 12 525
13 Frenkel J Kinetic Theory of Liquids 1946 (Oxford
University Press Oxford)
14 Wagner P E and Strey R Homogeneous nucleation
rates of water vapor measured in a two-piston expansion
chamber J Phys Chemistry 1981 85 2694
15 Miller R Anderson R J Kassner J L and Hagen
D E Homogeneous nucleation rate measurements for
water over a wide range of temperature and nucleation
rate J Chem Physics 1983 78 3204
16 Viisanen Y Strey R and Reiss H Homogeneous
nucleation rates for water J Chem Physics 1993 99 4680
17 Viisanen Y Strey R and Reiss H Erratum homo-
geneous nucleation rates for water J Chem Physics
2000 112 8205
18 Wolk J and Strey R Homogeneous nucleation of H2O
and D2O in comparison the isotope effect J Phys
Chemistry B 2001 105 11683
19 Heath C H Streletsky K and Wyslouzil B E H2Ondash
D2O condensation in a supersonic nozzle J Chem
Physics 2002 117 6176
20 Mikheev V B Irving P M Laulainen N S Barlow
S E and Pervukhin V V Laboratory measurements of
water nucleation using a laminar flow chamber J Chem
Physics 2002 116 10 772
21 Hung C H Krasnopoler M J and Katz J L
Condensation of a supersaturated vapor VIII The
homogeneous nucleation of n-nonane J Chem Physics
1989 90 1856
22 Rudek M M Fisk J A Chakarov V M and Katz
J L Condensation of a supersaturated vapor XII The
homogeneous nucleation of the n-alkanes J Chem
Physics 1996 105 4707
23 Strey R Wagner P E and Schmeling T Homogeneous
nucleation rates for n-alcohol vapors measured in a two-
piston expansion chamber J Chem Physics 1986 84
2325
24 Lihavainen H Viisanen Y and Kulmala M Homo-
geneous nucleation of n-pentanol in a laminar flow
diffusion chamber J Chem Physics 2001 114 10 031
25 Rudek M M Katz J L Vidensky I V Zdımal V and
Smolık J Homogeneous nucleation rates of n-pentanol
measured in an upward thermal diffusion cloud chamber
J Chem Physics 1999 111 3623
26 Grassmann A and Peters F Homogeneous nucleation
rates of n-propanol in nitrogen measured in a piston-
expansion tube J Chem Physics 2002 116 7617
27 Kane D and El-Shall S Condensation of supersaturated
vapors of hydrogen bonding molecules ethylene glycol
propylene glycol trimethyl glycol and glycerol J Chem
Physics 1996 105 7617
28 Hameri K and Kulmala M Homogeneous nucleation in
a laminar flow chamber the effect of temperature and
carrier gas pressure on dibutyl phthalate vapor nucleation
rate at high supersaturations J Chem Physics 1996 105
7696
29 Viisanen Y Strey R Laaksonen A and Kulmala M
Measurement of the molecular content of binary nuclei
II Use of the nucleation rate surface for waterndashethanol
J Chem Physics 1994 100 6062
30 Viisanen Y Wagner P E and Strey R Measurement of
the molecular content of binary nuclei IV Use of the
nucleation rate surfaces for the n-nonane-n-alcohol series
J Chem Physics 1998 108 4257
31 Fisk J A Rudek M M Katz J L Beiersdorf D and
Uchtmann H The homogeneous nucleation of cesium
vapor Atmos Res 1998 46 211
32 Uchtmann H Dettmer R Baranovskii S D and
Hensel F Photoinduced nucleation in supersaturated
mercury vapor J Chem Physics 1998 108 9775
33 Ferguson F T and Nuth J A Experimental studies of
the vapor phase nucleation of refractory compounds V
The condensation of lithium J Chem Physics 2000 113
4093
34 Laaksonen A Talanquer V and Oxtoby D W
Nucleation measurements theory and atmospheric
applications A Rev Phys Chemistry 1995 46 489
35 Debenedetti P G Metastable Liquids Concepts and
Principles 1996 (Princeton University Press Princeton
New Jersey)
36 Yasuoka K and Matsumoto M Molecular dynamics of
homogeneous nucleation in the vapor phase I Lennardndash
Jones fluid J Chem Physics 1998 109 8451
37 Yasuoka K and Matsumoto M Molecular dynamics of
homogeneous nucleation in the vapor phase II Water
J Chem Physics 1998 109 8463
I J FORD896
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
38 Wonczak S Strey R and Stauffer D Confirmation of
classical nucleation theory by Monte Carlo simulations in
the 3-dimensional Ising model at low temperature
J Chem Physics 2000 113 1976
39 Laasonen K Wonczak S Strey R and Laaksonen A
Molecular dynamics simulations of gasndashliquid nucleation
of LennardndashJones fluid J Chem Physics 2000 113 9741
40 Reiss H Methods of Thermodynamics 1996 (Dover
Publications New York)
41 Kondepudi D and Prigogine I Modern Thermodynamics
From Heat Engines to Dissipative Structures 1998 (John
Wiley Chichester West Sussex)
42 Smoluchowski M Ann Phys NY 1908 25 225
43 Einstein A Ann Phys NY 1910 338 1275
44 Hirschfelder J O Curtis C F and Bird R BMolecular
Theory of Gases and Liquids 1954 (John Wiley New
York)
45 Landau L D and Lifshitz E M Statistical Physics 1958
(Pergamon London)
46 Wu D T Nucleation theory Solid State Physicsmdash
Advances in Res and Applic 1997 50 37
47 Reguera D Rubı J M and Perez-Madrid A Fokkerndash
Planck equations for nucleation processes revisited
Physica A 1998 259 10
48 Schenter G K Kathmann S and Garrett B Variational
transition state theory of vapor phase nucleation
J Chem Physics 1999 109 7951
49 Onsager L Phys Rev 1931 37 405
50 Courtney W G J Chem Physics 1961 35 2249
51 Girshick S L and Chiu C-P Kinetic nucleation theory
a new expression for the rate of homogeneous nucleation
from an ideal supersaturated vapor J Chem Physics
1990 93 1273
52 Ford I J Barrett J C and Lazaridis M Uncertainties
in cluster energies in homogeneous nucleation theory
J Aerosol Sci 1993 24 581
53 Weakliem C L and Reiss H The factor 1S in the
classical theory of nucleation J Phys Chemistry 1994
98 6408
54 McClurg R B and Flagan R C Critical comparison of
droplet models in homogeneous nucleation theory
J Colloid Interface Sci 1998 201 194
55 Mashmoushy H Mahpeykar M R and Bakhtar F
Studies of nucleating and wet steam flows in two-dimen-
sional cascades Proc Instn Mech Engrs Part C J
Mechanical Engineering Science 2004 218(C8) 843ndash858
56 Ford I J Nucleation theorems the statistical mechanics
of molecular clusters and a revision of classical nuclea-
tion theory Phys Rev E 1997 56 5615
57 ten Wolde P R and Frenkel D Computer simulation
study of gasndashliquid nucleation in a LennardndashJones system
J Chem Physics 1998 109 9901
58 Kusaka I Wang Z G and Seinfeld J H Direct
evaluation of the equilibrium distribution of physical
clusters by a grand canonical Monte Carlo simulation
J Chem Physics 1998 108 3416
59 Kusaka I and Oxtoby D W Identifying physical
clusters in vapor phase nucleation J Chem Physics
1999 110 5249
60 Stillinger F H J Chem Physics 1963 38 1486
61 Ford I J Imperfect vaporndashgas mixtures and homo-
geneous nucleation J Aerosol Sci 1992 23 447
62 Lothe J and Pound G M J Chem Physics 1962 36
2080
63 Lee J K Barker J A and Abraham F F Theory and
Monte Carlo simulation of physical clusters in the
imperfect vapor J Chem Physics 1973 58 3166
64 Abraham F F Homogeneous Nucleation Theory 1974
(Academic Press New York)
65 Reif F Fundamentals of Statistical and Thermal Physics
1965 (McGraw-Hill New York)
66 Hoare M R and Pal P Adv Phys 1975 24 645
67 Hoare M R Adv Chem Physics 1979 40 49
68 Pal P and Hoare M R Thermodynamic properties
and homogeneous nucleation of molecular clusters of
nitrogen J Phys Chemistry 1987 91 2474
69 Reiss H Katz J L and Cohen E R Translationndash
rotation paradox in the theory of nucleation J Chem
Physics 1968 48 5553
70 Hale B N Monte Carlo calculations of effective surface
tension for small clusters Aust J Physics 1996 49 425
71 Kathmann S M Schenter G K and Garrett B C
Dynamical nucleation theory calculation of condensation
rate constants for small water clusters J Chem Physics
1999 111 4688
72 Reiss H Tabazadeh A and Talbot J Molecular theory
of vapor phase nucleation the physically consistent
cluster J Chem Physics 1990 92 1266
73 Ellerby H M Weakliem C L and Reiss H Toward a
molecular theory of vaporndashphase nucleation I Identifica-
tion of the average embryo J Chem Physics 1992 97
5766
74 Ellerby H M and Reiss H Toward a molecular theory
of vaporndashphase nucleation II Fundamental treatment of
the cluster distribution J Chem Physics 1992 97 5766
75 Weakliem C L and Reiss H Toward a molecular theory
of vaporndashphase nucleation III Thermodynamic properties
of argon clusters fromMonte Carlo simulations and modi-
fied liquid drop theory J Chem Physics 1994 101 2398
76 Kusaka I Wang Z G and Seinfeld J H Binary
nucleation of sulfuric acidndashwater Monte Carlo simula-
tion J Chem Physics 1998 108 6829
77 Oh K J and Zeng X C A small system ensemble
Monte Carlo simulation of supersaturated vapor
evaluation of barrier to nucleation J Chem Physics
2000 112 294
78 Senger B Schaaf P Corti D S Bowles R Voegel J
C and Reiss H A molecular theory of the homogeneous
nucleation rate I Formulation and fundamental issues
J Chem Physics 1999 110 6421
79 Hill T L Statistical Mechanics Principles and Selected
Applications 1956 (McGraw-Hill New York)
80 Soto R and Cordero P Nonideal gas of clusters at
equilibrium J Chem Physics 1999 110 7316
81 Barrett J C The significance of cluster lifetime in
nucleation theory J Chem Physics 2002 116 8856
82 Lushnikov A A and Kulmala M Dimers in nucleating
vapors Phys Rev E 1998 58 3157
83 Bahadur R and McClurg R B Comparison of cluster
definitions in homogeneous nucleation rate calculations
J Phys Chemistry 2001 105 11 893
84 Pugnaloni L A and Vericat F New criteria for cluster
identification in continuum systems J Chem Physics
2002 116 1097
STATISTICAL MECHANICS OF NUCLEATION 897
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
85 Chen B Siepmann J I Oh K J and Klein M L
Simulating vaporndashliquid nucleation of n-alkanes J Chem
Physics 2002 116 4317
86 Harris S A and Ford I J A dynamical definition of
quasi-bound molecular clusters J Chem Physics 2003
118 9216
87 Shreve A P Walton J R P B and Gubbins K E
Liquid drops of polar molecules J Chem Physics 1986
85 2178
88 ten Wolde P R Ruiz-Montero M J and Frenkel D
Numerical calculation of the rate of homogeneous gasndash
liquid nucleation in a LennardndashJones system J Chem
Physics 1998 109 9901
89 ten Wolde P R Oxtoby D W and Frenkel D Chain
formation in homogeneous gasndashliquid nucleation of polar
fluids J Chem Physics 1998 111 4762
90 Vehkamaki H and Ford I J Critical cluster size and
droplet nucleation rate from growth and decay simula-
tions of LennardndashJones clusters J Chem Physics 2000
112 4193
91 Schaaf P Senger B Voegel J C Bowles R K and
Reiss H Simulative determination of kinetic coefficients
for nucleation rates J Chem Physics 2001 114 8091
92 Chen B Siepmann J I Oh K J and Klein M L
Aggregation-volume-bias Monte Carlo simulations of
vaporndashliquid nucleation barriers for LennardndashJonesium
J Chem Physics 2001 115 10 903
93 Hale B N and DiMattio D J A Monte Carlo discrete
sum (MCDS) nucleation rate model for water In
Proceedings of the 15th International Conference on
Nucleation and Atmospheric Aerosols AIP Conference
Proceedings Vol 534 (Eds B N Hale and M Kulmala)
2000 pp 31ndash34
94 Kathmann S M and Hale B N Monte Carlo
simulations of small sulfuric acidndashwater clusters
J Phys Chemistry B 2001 105 11 719
95 Daura X Mark A E and van Gunsteren W F Para-
metrization of aliphatic CHn united atoms of GROMOS96
force field J Comput Chemistry 1998 19 535
96 Kathmann S M Schenter G K and Garrett B C
Understanding the sensitivity of nucleation kinetics a
case study on water J Chem Physics 2002 116 5046
97 Kashchiev D On the relation between nucleation work
nucleus size and nucleation rate J Chem Physics 1982
76 5098
98 Oxtoby D W and Kashchiev D A general relation
between the nucleation work and the size of the nucleus in
multicomponent nucleation J Chem Physics 1994 100
7665
99 Ford I J Thermodynamic properties of critical clusters
from measurements of vapourndashliquid homogeneous
nucleation rates J Chem Physics 1996 105 8324
100 Oxtoby D W and Evans R Nonclassical nucleation
theory for the gasndashliquid transition J Chem Physics
1988 89 7521
101 Zeng X C and Oxtoby D W Gasndashliquid nucleation in
LennardndashJones fluids J Chem Physics 1991 94 4472
102 Evans R Fundamentals of Inhomogeneous Fluids 1992
Ch 3 pp 85ndash175 (Marcel Dekker New York)
103 Talanquer V and Oxtoby D W Dynamical density
functional theory of gasndashliquid nucleation J Chem
Physics 1994 100 5190
104 Barrett J C First-order correction to classical nucleation
theory a density functional approach J Chem Physics
1999 111 5938
105 Barrett J C Cluster translation and growth in density
functional theories of homogeneous nucleation J Chem
Physics 1997 107 7989
106 Bykov T V and Zeng X C A patching model for
surface tension of spherical droplet and Tolman length J
Chem Physics 1999 111 10 602
107 Talanquer V and Oxtoby D W Nucleation in molecular
and dipolar fluids interaction site model J Chem
Physics 1995 103 3686
108 Seok C and Oxtoby D W Nucleation in n-alkanes a
density functional approach J Chem Physics 1998 109
7982
109 Reguera D Bowles R K Djikaev Y and Reiss H
Phase transitions in systems small enough to be clusters
J Chem Physics 2003 118 340
110 Khakshouri S and Ford I J J Chem Physics 2004
(submitted)
111 Ford I J (unpublished)
112 Nyquist R M Talanquer V and Oxtoby D W Density
functional theory of nucleation a semi-empirical
approach J Chem Physics 1995 103 1175
113 Talanquer V and Oxtoby D W Density functional
analysis of phenomenological theories of gasndashliquid
nucleation J Chem Physics 1995 99 2865
114 Katz J L Fisk J A and Chakarov V The accuracy of
nucleation theory In Proceedings of the 13th Interna-
tional Conference on Nucleation and Atmospheric Aerosols
(Eds N Fukuta and P E Wagner) 1992 pp 11ndash18
(Deepak Publishing Hampton Virginia)
115 Fisher M E The theory of condensation and the critical
point Physics 1967 3 255
116 Dillmann A and Meier G E A A refined droplet
approach to the problem of homogeneous nucleation
from the vapor phase J Chem Physics 1991 94 3872
117 Tolman R C J Chem Physics 1949 17 333
118 Ford I J Laaksonen A and Kulmala M Modification
of the DillmannndashMeier theory of homogeneous nuclea-
tion J Chem Physics 1993 99 764
119 Laaksonen A Ford I J and Kulmala M Revised
parametrization of the DillmannndashMeier theory of homo-
geneous nucleation Phys Rev E 1994 49 5517
120 Kalikmanov V I and van Dongen M E H Semipheno-
menological theory of homogeneous vaporndashliquid nuclea-
tion J Chem Physics 1995 103 4250
121 Hale B N Application of a scaled homogeneous
nucleation-rate formalism to experimental-data at
T 5Tc Phys Rev A 1986 33 4156
122 Martinez D M Ferguson F T Heist R H and Nuth
J A Application of scaled nucleation theory to metallic
vapor condensation J Chem Physics 2001 115 310
123 Hale B N and Wilemski G A scaled nucleation model
for ideal binary systems Chem Physics Lett 1999 305
263
124 Granasy L Diffuse interface theory for homogeneous
vapor condensation J Chem Physics 1996 104
5188
125 McGraw R and Laaksonen A Interfacial curvature free
energy the Kelvin relation and vaporndashliquid nucleation
rate J Chem Physics 1997 106 5284
I J FORD898
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
126 Talanquer V A new phenomenological approach to
gasndashliquid nucleation based on the scaling properties
of the critical nucleus J Chem Physics 1997 106
9957
127 Koga K and Zeng X C Thermodynamic expansion of
nucleation free-energy barrier and size of critical nucleus
near the vaporndashliquid coexistence J Chem Physics 1999
110 3466
128 Kusaka I On the scaling behavior of the free energetics of
nucleation J Chem Physics 2003 118 5510
129 Nowakowski B and Ruckenstein E A kinetic approach
to the theory of nucleation in gases J Chem Physics
1991 94 1397
130 Nowakowski B and Ruckenstein E Homogeneous
nucleation in gases a three-dimensional FokkerndashPlanck
equation for evaporation from clusters J Chem Physics
1991 94 8487
131 Wilcox C F and Bauer S H Estimation of homo-
geneous nucleation flux via a kinetic model J Chem
Physics 1991 94 8302
132 Hanggi P Talkner P and Borkovec M Reaction rate
theory fifty years after Kramers Rev Mod Physics 1990
62 251
133 Barrett J C Clement C F and Ford I J Energy
fluctuations in homogeneous nucleation theory for aero-
sols J Physics A 1993 26 529
134 Ford I J and Harris S A Molecular cluster decay
viewed as escape from a potential of mean force J Chem
Physics 2004 120 4428
135 Risken H The FokkerndashPlanck Equation Methods of
Solution and Applications 2nd edition 1989 (Springer-
Verlag Berlin)
136 Shizgal B and Barrett J C Time dependent nucleation
J Chem Physics 1989 91 6505
STATISTICAL MECHANICS OF NUCLEATION 899
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
different expression applies for low temperatures whenthe oscillator is not classically activated [65]
The rotational and translational degrees of freedomare next The previous restriction of the configurationsto zero angular momentum is lifted The approximate(classical) rotational partition function is
Zrot frac14 p1=28p2kTh2
3=2
ethI1I2I3THORN1=2 eth19THORN
where Ik are the principal moments of inertia about thecentre of mass for atoms fixed in their mean positions
Finally the translational free energy of the centre ofmass of the cluster is given by the partition function
Ztrans frac14 V2pMkT
h2
3=2
eth20THORN
where V is the system volume and M frac14 im is the clustermass The free energy of the cluster (plus kT ln n1 forconvenience) is therefore
FethiTHORN thorn kT ln n1
frac14 kT lnZtransZrotZvib
n1
frac14 U0ethiTHORN kT ln
eth2pMkTTHORN3=2
Srsvh3p1=2
8p2kTh2
3=2
6ethI1I2I3THORN1=2Y3ji6
kfrac141
kT
hok
eth21THORN
where rsv frac14 ns1=V is the molecular density of dilutesaturated vapour
By similar arguments the chemical potential of dilutesaturated vapour is
s frac14 Um0 kT ln
eth2pmkTTHORN3=2
rsvh3p1=2
8p2kTh2
3=2
6ethIm1 Im2 Im3 THORN1=2Y3j6
kfrac141
kT
homk
eth22THORN
where Um0 is the potential energy of a molecule when the
atoms are at their mean positions Imk are the threeprincipal moments of inertia of the molecule about thecentre of mass and om
k are the oscillatory frequencies ofmolecular internal vibrations
The classical work of formation DWclethiTHORN should nowcorrespond to the leading-order contribution to thedifference between equation (21) and i times equation(22) Calculations [54 64 66ndash68] demonstrate that theleading term is indeed roughly proportional to i2=3 thusjustifying the form of the classical model The suitabilityof the capillarity approximation depends on the extentto which the coefficient of this leading term correspondsnumerically to the planar surface free energy Howeverthe full statistical mechanical expression for the work of
formation contains additional terms and these will beimportant for small clusters and need to be taken intoaccount In order to evaluate them the variousstructural and energy properties of cluster are requiredall of which can be obtained from microscopic modelsand they can then be inserted into the formulae for thethermodynamic properties [equations (21) and (22)] theconstrained equilibrium populations [equation (18)] andultimately the nucleation rate [equation (10)] Followingthis route employs the almost universally held assump-tion that equilibrium properties such as free energies canbe used to describe non-equilibrium properties specifi-cally the cluster decay rate This assumption is re-examined in section 8
6 LIQUID DROPLETS AND THE CLUSTER
DEFINITION
Defining what was meant by a solid cluster in theprevious section was not an issue Within the assump-tions of the model (harmonic interactions) it wasimpossible for a molecule to escape from the clusterThe cluster definition in the case of liquids cannot bedismissed so easily Liquids are more volatile thansolids and the molecules in a cluster are much moremobile It is imperative now to characterize a clusterusing the molecular positions and velocities andperhaps some time scale and length scale too
The traditional microscopic method for defining acluster is to impose a geometric constraint on molecularpositions An example is to restrict molecules to a spherecentred on the centre of mass [69ndash71] or to impose amaximum allowable separation between the molecules(the Stillinger criterion [60]) More sophisticated treat-ments exist such as the n=v Stillinger cluster introducedby Reiss and co-workers [72ndash75] Here the confiningvolume v used to define the cluster is considered to be anadditional defining characteristic on the same level asthe cluster size n As time progresses a cluster changesits confining volume as the component molecules followtheir respective trajectories just as they change their sizewith the gain and loss of molecules The association of amolecular configuration with a precise confining volumeis complicated but tractable The dependence of the freeenergy on v determines the statistical dynamics of thiscluster volume A similar attempt to associate a clusterwith a specified volume within a grand canonicalensemble description was developed by Kusaka andco-workers [58 59 76] The simulation box was taken tobe the confining cluster volume Many of these ideasmay be viewed as a treatment of the cluster definition(eg the Stillinger radius) as a variational parameter inthe representation of the partition function Further
I J FORD890
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
refinement of this point of view using a treatment ofclusterndashcluster interactions to optimize the definingcluster volume has been presented [77] A review ofrecent progress in this area and the application of theseideas to nucleation theory has been provided by Sengeret al [78]
Such geometric constraints can be implemented buttheir realism is questionable Situations better regardedas close encounters between separate clusters would beclassed as snapshots of single larger clusters It shouldbe recalled that the cluster definition is an internalfeature of the model which is to be optimized to reduceclusterndashcluster interactions and to improve the repre-sentation of the grand partition function Misclassifyingconfigurations is fraught with consequences A secondissue to consider is that for convenience the clustersshould behave in a manner that can be modelled withBeckerndashDoring equations This means that the growthand decay rates should be free of memory effects ie thetransition processes should be Markovian The extent towhich this is satisfied (if indeed it is physically realistic)depends on the cluster definition
An intuitively reasonable choice of definition whichaddresses both these points is that a cluster shouldconsist of molecules that are bound energetically Thisshould minimize the clusterndashcluster contributions to thetotal energy According to such a definition if amolecule interacts strongly with a cluster then it islikely to be regarded as part of that cluster Also anenergetically bound cluster is likely to decay by thermalfluctuation in a stochastic Markovian fashion ratherthan by the separation of passing poorly boundfragments
Ideas have been proposed which require the cluster tobe energetically bound (in a loose sense since a cluster isintrinsically prone to decay) The time evolution of amolecular configuration and not simply the set ofinstantaneous positions of its component molecules isof central importance Hill [79] introduced a clusterdefinition requiring the total energies of pairs ofmolecules in a cluster to be negative (in their centre ofmass frame) The intention was to exclude situationswhere a molecule could leave the cluster within the timeit would take for a molecule to travel a typical lineardimension of the cluster Soto and Cordero [80] havedeveloped Hillrsquos criterion and Barrett [81] has per-formed Monte Carlo modelling of molecular configura-tions excluding those which decay in moleculardynamics within a given period A number of otherenergy-based criteria have been developed [82ndash85]
Harris and Ford [86] have recently proposed a newenergy-based cluster definition As with Barrettrsquosapproach it is based on the requirement that amolecular snapshot is energetically bound for a chosensubsequent period Evaporation according to thisscheme requires not only that a molecule should makea positive energy excursion but also that the subsequent
dynamics carry it far away from the cluster Onlyperhaps one in eight molecules manage to avoidrecapture once having satisfied the positive energycriterion [86] A snapshot of a cluster in this casesurrounded by a carrier gas is shown in Fig 4 Thehistory of the energy of a molecule in such a cluster isillustrated in Fig 5 three excursions into positive energyoccur in this sequence but only one leads to actualescape corresponding to the maintenance of positiveenergy for a significant time The instant of escape istaken to be the point at which the energy goes positivefor the third and final time
Cluster definitions implemented in moleculardynamics and Monte Carlo codes allow cluster freeenergies and hence the decay rate to be calculated Anenormous amount of simulation work has been carriedout and calculations can be laborious Many differentschemes and approximations have been made to easethe evaluation of these properties The principalconclusion from various studies using a range of clusterdefinitions and interaction models [49 57 58 70 85 87ndash92] seems to be that the work of formation of liquidcritical clusters does indeed contain a dominantcontribution proportional to i2=3 consistent with theclassical work of formation Hale and collaboratorspresent cluster free energies for water and other speciesin a way that makes this particularly clear indeed it issurprising how dominant the surface term can be foreven very small clusters [70 93 94] Neverthelessfurther contributions exist small clusters are notcharacterized just by scaled-down properties of largeclusters and it is impossible to escape the need todetermine the additional terms
In all the above discussion it has been assumed thatthe intermolecular interactions are accurately specifiedThis is a situation that is rarely possible Althoughfirst-principles quantum-mechanical models could be
Fig 4 Snapshot of a cluster of 50 argon atoms (brightspheres) bathed in a gas of helium (darker spheres)
STATISTICAL MECHANICS OF NUCLEATION 891
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
developed in practice empirical potentials are devel-oped on the basis of physical intuition containingparameters fitted to the measured properties of thesubstance in question [95] Usually these are bulkproperties and it is not always clear that the sameinteractions can be used for the very different circum-stances of a molecular cluster [96] In view of thesensitivity of nucleation rates to the thermodynamicproperties of clusters this is perhaps a crucial failingProgress can be made however by computing knowncluster properties where they are available and byrefining the potentials to suit Extraction of clusterproperties through use of the so-called nucleationtheorems (which relate derivatives of the nucleationrate to properties of the critical cluster [16 56 97ndash99])may provide just this sort of experimental information
A technique that has not yet been described is the useof density functional theory (DFT) to model theproperties of liquid droplets [100ndash104] This powerfulapplication of statistical mechanics promises to avoidlaborious computations at the microscopic level byinstead focusing on the density profile of moleculesmaking up the droplet This profile may be regarded asthe average distribution of molecules about the clustercentre of mass Although this might appear to involvecoarse-graining and approximation it is a well-foundedapproach and in principle can be exact In practical
implementations however a mean field approach istaken such that the constituent molecules move in aneffective potential well created by the other molecules inthe cluster [105] Also the interactions are assumed to becharacterized by strong repulsive cores and weaklyattractive tails
Considerable analytical work has been carried out[104 106] Various developments have been made todescribe droplets consisting of simple atoms as well aschain-like molecules [107] and polar molecules [108]The main drawback in these implementations is thatspecific interactions such as hydrogen bonds are hardto represent Certain details of the modelled systemare neglected [109 110] and the (implicit) clusterdefinition used is geometric and not energetic [111]On the other hand quite sophisticated systems havebeen modelled including multi-component dropletsThe method has been compared with otherapproaches [112 113] and certain features are betterdescribed particularly the approach to spinodalconditions where the nucleation barrier vanishesDFT calculations can be performed relatively easilyand they can provide a qualitative and perhapsquantitative insight into quite complicated clusteringproblems They provide an intermediate-level descrip-tion between the simple capillarity approximation anda microscopic description
Fig 5 The energy of a molecule in the centre-of-mass frame of a cluster occasionally becomes positive butthis does not necessarily lead to escape Only the third positive energy excursion in this sequenceindicated by an arrow results in particle escape A cluster definition must capture this behaviour
I J FORD892
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
7 PHENOMENOLOGICAL MODELS
The reader might be dismayed by the mathematicalcomplexity of the expressions for the work of formationof solid clusters in section 5 might be alarmed by theprocedural complexity of implementing a cluster defini-tion in section 6 may recoil from the difficulties incalculating free energies from molecular simulationmethods and ultimately may give up in view of theuncertainties in the underlying intermolecular interac-tions Considering these complications it is tempting totry to develop phenomenological models of the work offormation After all if the capillarity approximationworks reasonably well for water then surely a minorcorrection to account for small cluster effects ought tobe sufficient to improve matters
However after perhaps two decades of work nomodel has emerged offering broad success across arange of different substances It is perhaps asking toomuch to expect universal improvement CNT works as afirst approximation for water for a particular tempera-ture range but as discussed earlier it possesses thewrong temperature dependence so that predictionsaway from a temperature in the range 210ndash260K areuncertain A similar picture emerges for other sub-stances although in some cases the failure is even moresevere
The simplest phenomenological adjustment to equa-tion (4) the basis for CNT is to write
ncei frac14 n1 expfrac12DWclethiTHORN DWcleth1THORN
kT
eth23THORN
where DWclethiTHORN is given by equation (5) This amounts toa uniform downward shift of the classical free energybarrier illustrated in Fig 1 At least now the expressionis consistent at i frac14 1 The theory is known as internallyconsistent classical theory (ICCT) [51] Unfortunately itis not clear whether this constant shift in the work offormation should be a realistic representation of all thesubleading terms in the work of formation [52] It is anarbitrary modification and its success in accounting fordata is mixed For example nucleation rates for a rangeof alkanes [114] require further modification by factorsof between one and nine orders of magnitude to comeinto agreement with ICCT
It should be noted at this point that the correctionterm kT lnS which appears in DWcleth1THORN can beregarded as rigorous There are good reasons for thisbased on general considerations involving the nucleationtheorems [56] or more simply the principles of chemicalequilibrium represented by the law of mass action givenin equation (15) The correction converts the prefactorn1 in equation (4) into ns1 and results in the insertion of afactor 1=S in the nucleation rate in equation (1)
The Fisher droplet model is a slightly more sophis-ticated scheme [115] Here the constrained equilibrium
populations are modelled by
ncei frac14 ns1 exp yii2=3 thorn t ln i lnq0kT
psv
thorn i lnS
eth24THORN
The term proportional to yi may be viewed as a surfaceterm while the other two are corrections of a typesuggested by the free energy expressions developed insection 5 The parameters q0 yi and t are chosen toensure that the model correctly reproduces knownproperties of clusters and vapours including virialcoefficients and critical properties the theory is there-fore phenomenological
For large cluster sizes and for temperatures wellbelow the critical temperature the free energy offormation should tend towards that of a macroscopicliquid droplet yi should therefore approach A1s=ethkTTHORNfor large i where A1i
2=3 is the surface area 4pR2 of aspherical droplet containing i monomers at the bulkliquid density so that A1 frac14 eth36pTHORN1=3r2=3
l Dillmann andMeier [116] suggested writing
yi frac14kiA1skT
eth25THORN
with the Ansatz
ki frac14 1thorn a1i 1=3 thorn a2i 2=3 eth26THORN
The coefficients a1 and a2 can be found in terms of virialcoefficients of the vapour The form of this Ansatz wasmotivated by the curvature dependence of the bulksurface tension derived in classical thermodynamics byTolman [117]
The initial success of the DillmannndashMeier theory wasshort lived since modifications had to be made toovercome an internal inconsistency [118] Virial expan-sions for the vapour pressure and density were usedincorrectly to calculate the virial expansion of the ratioof the two Correcting this error [61] changed predictedrates by several orders of magnitude Nevertheless theapproach was very influential [119] In a furtherdevelopment Kalikmanov and van Dongen [120] seta2 frac14 0 (this term can be absorbed into q0 anyway) andchose a1 such that the experimental saturated vapourpressure psv is reproduced by the model The DillmannndashMeier model matched the first two virial coefficientsonly The Kalikmanovndashvan Dongen theory shouldtherefore approximate the (revised) DillmannndashMeiertheory at low temperatures but ought to provide abetter description at higher temperatures
Other phenomenological models have been proposedincluding the Hale scaling relation [121] This modelbased on CNT but employing simplified models for thesurface tension and liquid density based on scalingbehaviour near the critical point together with a rather
STATISTICAL MECHANICS OF NUCLEATION 893
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
more straightforward rate prefactor seems to be quitesuccessful in accounting for data for a wide range ofmaterials even metals [122] and two-component systems[123] Another phenomenological model is the diffuseinterface theory [124] which is based on a treatment ofthe phase interface using simplified density profiles ofthermodynamic properties Another general class ofmodel makes use of conditions at the vapourndashliquidspinodal to constrain the form of the model free energy[125ndash128]
The main drawback in phenomenological models isthe arbitrariness in the form of the proposed work offormation The parameter fitting in practice ensures thatthe populations of small clusters especially the mono-mer and dimer are well represented by the model butthe properties of larger clusters depend on the expres-sion assumed for the model and are therefore uncertainIntuition suggests that phenomenological modelling islikely to have mixed success and that microscopiccalculations of cluster properties do matter There mightbe better value in phenomenological models however ifthey were based securely on the results of microscopiccalculations
8 KINETICS OR THERMODYNAMICS
In section 3 kinetic equations representing the nuclea-tion process were described However the coefficients inthose equations were related to constrained equilibriumpopulations and hence to the equilibrium statisticalmechanics of clusters through the detailed balancecondition (8) Is it necessary to do this Could acompletely kinetic route be followed by evaluating themean growth and decay coefficients bi and gi directly bysome statistical approach The main problem is ofcourse the difficulty in performing such calculationsand this has restricted the development of such models[129ndash131] It is expected that the kinetic coefficients willcorrespond to those derived from the detailed balanceargument for near-equilibrium conditions but circum-stances can be imagined where non-equilibrium effectsare present which cannot be dealt with within standardstatistical mechanics
The framework for evaluating kinetic coefficients ofcluster decay can be briefly introduced [86 132] Amolecular dynamics simulation of a cluster can be set upand evolved according to the interactions until bychance an evaporation event occurs By extensivesimulation statistics for the mean evaporation rates gican be amassed in this way Similarly moleculardynamics simulations where a single molecule iscaptured by a cluster could be performed and thegrowth rates bi determined These will be functions ofcluster energy and so would fit best into a rate equationscheme involving both size and energy [133]
This is a laborious approach but statistical physicsoffers an approximate alternative representation of theescape and capture processes The basic tool is theLangevin equation The radial motion of a molecule inthe cluster is described by [134]
meurorr frac14 f ethrTHORN thornfethr _rr tTHORN eth27THORN
where r is the radial position m is the molecular mass fis the mean (time- and velocity-averaged) force on themolecule at position r and f is the stochastic forceThe latter is usually represented as a velocity-dependent friction term mg _rr where g is the so-calledfriction coefficient plus a fluctuating term ~ff ethr tTHORN withzero mean and correlation in time h ~ff ethr tTHORN~ff ethr t0THORNi frac14eth2gkTmTHORNdetht t0THORN where k is Boltzmannrsquos constant andwhere T has the characteristics of a temperature as willbe seen It is a standard manipulation [135] to convertthe Langevin description in the large friction limit intoa FokkerndashPlanck equation
qwqt
frac14 1
mg qethfwTHORN
qrthorn kT
q2wqr2
eth28THORN
which represents the evolution of wethr tTHORN the prob-ability density that the molecule should lie at the radialposition r It should be noted in passing that FokkerndashPlanck equations have often been used to describe thestochastic dynamics of a cluster in size space [48 136] Incontrast the use of the method described above refers tomotion of a molecule in real coordinate space relative tothe centre of mass of the cluster
The steady state equilibrium solution of equation (28)is
wethrTHORN exp FethrTHORNkT
eth29THORN
where FethrTHORN is called the potential of mean force relatedto the mean force on the molecule f throughf frac14 dF=dr It can now be seen how the parameter Tin the stochastic force plays the role of temperaturesince equation (29) looks like a Boltzmann distributionThis solution is not what is sought however The escapeproblem is characterized by a boundary conditionwethreTHORN frac14 0 where re is the radius at which a particleescapes from the system The resulting escape rate maybe shown to be [134]
gkin frac14 KethreT m gTHORN exp DFkT
eth30THORN
where the prefactor K depends on the parameters listedas well as the shape of the potential of mean force Theprincipal feature of equation (30) is the exponentialdependence on the depth DF of the potential of meanforce Therefore if the potential of mean force FethrTHORN can
I J FORD894
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
be evaluated by studying a molecular dynamics trajec-tory for example then the kinetic decay rate can bedetermined This could be extended to account for aradially dependent temperature and friction coefficientboth of which might be determined from simulation Asketch of a potential of mean force is given in Fig 6
Kinetically determined cluster growth and decayrates gkini and bkini respectively could provide a newroute for the determination of nucleation rates inde-pendent of assumptions about detailed balance and theuse of equilibrium statistical mechanics in a non-equilibrium situation
9 CONCLUSIONS
This review has been brief and in some ways partisan tothe personal views of the author It attempts to sketch acoherent picture of nucleation theory and how thevarious disparate strands relate to one another One ofthe fascinations of nucleation theory is the variety ofapproaches and the number of difficult issues that areencountered How well does free energy fluctuationtheory represent a real non-equilibrium process Howwell does the idea of detailed balance provide aframework for determining non-equilibrium decaycoefficients from equilibrium statistical mechanicsHow well do various methods perform in the evaluation
of cluster free energies Can cluster growth and decayrates be calculated directly without the need for such aframework How exactly are clusters to be defined
The theoretical problems are many but the issues canbe resolved to various degrees of satisfaction Thereremains of course the fundamental task of choosing anintermolecular potential that adequately represents thephysical interactions Without this all the finer points ofcluster definition and treatments of non-equilibriumprocesses are unimportant Modelling many problems inscience relies on the availability of an accurate micro-scopic description and so nucleation rate calculationsare not unique in this respect Most empirical molecularpotentials are fitted to the properties of bulk and surfaceproperties and in the very best examples attention isgiven to their transferability to more unusual environ-ments such as clusters In view of these capabilities andmodern theoretical developments in nucleation theoryit seems likely that successful microscopic models of thenucleation process based on statistical physics are atlast coming within reach
ACKNOWLEDGEMENTS
The present author is grateful to colleagues SarahHarris and Marco Arosio for providing some of thefigures
Fig 6 A molecule in a cluster may be considered to be confined within it by a potential of mean force FethrTHORN Itcan escape by reaching a radial position re driven by a phenomenological stochastic force
STATISTICAL MECHANICS OF NUCLEATION 895
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
REFERENCES
1 Gross D H E Microcanonical Thermodynamics 2001
(World Scientific Singapore)
2 Seinfeld J H and Pandis S N Atmospheric Chemistry
and Physics 1998 (John Wiley New York)
3 Climate Change 2001 Synthesis Report 2001 (Inter-
governmental Panel on Climate Change) (Cambridge
University Press Cambridge)
4 Kohler T M Gail H-P and Sedlmayr E MgO dust
nucleation in M-stars calculation of cluster properties
and nucleation rates Astronomy and Astrophysics 1997
320 553
5 Kashchiev D Nucleation Basic Theory with Applications
2000 (ButterworthndashHeinemann Oxford)
6 Bakhtar F Ebrahimi M and Webb R A On the
performance of a cascade of turbine rotor tip section
blading in nucleating steam Part 1 surface pressure
distributions Proc Instn Mech Engrs 1995 209(C2)
115ndash124
7 Bakhtar F Ebrahimi M and Bamkole B O On the
performance of a cascade of turbine rotor tip section
blading in nucleating steam Part 2 wake traverses Proc
Instn Mech Engrs 1995 209(C3) 169ndash178
8 Bakhtar F andMahpeykar M R On the performance of
a cascade of turbine rotor tip section blading in nucleating
steam Part 3 theoretical treatment Proc Instn Mech
Engrs Part C J Mechanical Engineering Science 1997
211(C3) 195ndash210
9 Volmer M and Weber A Z Physikalische Chemie 1926
119 277
10 Farkas L Z Physikalische Chemie (Leipzig) 1927 125
236
11 Becker R and Doring W Kinetische Behandlung der
Keinbildung in ubersattigten Dampfen Ann Physik
(Leipzig) 1935 24 719
12 Zeldovich J Soviet Physics JETP 1942 12 525
13 Frenkel J Kinetic Theory of Liquids 1946 (Oxford
University Press Oxford)
14 Wagner P E and Strey R Homogeneous nucleation
rates of water vapor measured in a two-piston expansion
chamber J Phys Chemistry 1981 85 2694
15 Miller R Anderson R J Kassner J L and Hagen
D E Homogeneous nucleation rate measurements for
water over a wide range of temperature and nucleation
rate J Chem Physics 1983 78 3204
16 Viisanen Y Strey R and Reiss H Homogeneous
nucleation rates for water J Chem Physics 1993 99 4680
17 Viisanen Y Strey R and Reiss H Erratum homo-
geneous nucleation rates for water J Chem Physics
2000 112 8205
18 Wolk J and Strey R Homogeneous nucleation of H2O
and D2O in comparison the isotope effect J Phys
Chemistry B 2001 105 11683
19 Heath C H Streletsky K and Wyslouzil B E H2Ondash
D2O condensation in a supersonic nozzle J Chem
Physics 2002 117 6176
20 Mikheev V B Irving P M Laulainen N S Barlow
S E and Pervukhin V V Laboratory measurements of
water nucleation using a laminar flow chamber J Chem
Physics 2002 116 10 772
21 Hung C H Krasnopoler M J and Katz J L
Condensation of a supersaturated vapor VIII The
homogeneous nucleation of n-nonane J Chem Physics
1989 90 1856
22 Rudek M M Fisk J A Chakarov V M and Katz
J L Condensation of a supersaturated vapor XII The
homogeneous nucleation of the n-alkanes J Chem
Physics 1996 105 4707
23 Strey R Wagner P E and Schmeling T Homogeneous
nucleation rates for n-alcohol vapors measured in a two-
piston expansion chamber J Chem Physics 1986 84
2325
24 Lihavainen H Viisanen Y and Kulmala M Homo-
geneous nucleation of n-pentanol in a laminar flow
diffusion chamber J Chem Physics 2001 114 10 031
25 Rudek M M Katz J L Vidensky I V Zdımal V and
Smolık J Homogeneous nucleation rates of n-pentanol
measured in an upward thermal diffusion cloud chamber
J Chem Physics 1999 111 3623
26 Grassmann A and Peters F Homogeneous nucleation
rates of n-propanol in nitrogen measured in a piston-
expansion tube J Chem Physics 2002 116 7617
27 Kane D and El-Shall S Condensation of supersaturated
vapors of hydrogen bonding molecules ethylene glycol
propylene glycol trimethyl glycol and glycerol J Chem
Physics 1996 105 7617
28 Hameri K and Kulmala M Homogeneous nucleation in
a laminar flow chamber the effect of temperature and
carrier gas pressure on dibutyl phthalate vapor nucleation
rate at high supersaturations J Chem Physics 1996 105
7696
29 Viisanen Y Strey R Laaksonen A and Kulmala M
Measurement of the molecular content of binary nuclei
II Use of the nucleation rate surface for waterndashethanol
J Chem Physics 1994 100 6062
30 Viisanen Y Wagner P E and Strey R Measurement of
the molecular content of binary nuclei IV Use of the
nucleation rate surfaces for the n-nonane-n-alcohol series
J Chem Physics 1998 108 4257
31 Fisk J A Rudek M M Katz J L Beiersdorf D and
Uchtmann H The homogeneous nucleation of cesium
vapor Atmos Res 1998 46 211
32 Uchtmann H Dettmer R Baranovskii S D and
Hensel F Photoinduced nucleation in supersaturated
mercury vapor J Chem Physics 1998 108 9775
33 Ferguson F T and Nuth J A Experimental studies of
the vapor phase nucleation of refractory compounds V
The condensation of lithium J Chem Physics 2000 113
4093
34 Laaksonen A Talanquer V and Oxtoby D W
Nucleation measurements theory and atmospheric
applications A Rev Phys Chemistry 1995 46 489
35 Debenedetti P G Metastable Liquids Concepts and
Principles 1996 (Princeton University Press Princeton
New Jersey)
36 Yasuoka K and Matsumoto M Molecular dynamics of
homogeneous nucleation in the vapor phase I Lennardndash
Jones fluid J Chem Physics 1998 109 8451
37 Yasuoka K and Matsumoto M Molecular dynamics of
homogeneous nucleation in the vapor phase II Water
J Chem Physics 1998 109 8463
I J FORD896
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
38 Wonczak S Strey R and Stauffer D Confirmation of
classical nucleation theory by Monte Carlo simulations in
the 3-dimensional Ising model at low temperature
J Chem Physics 2000 113 1976
39 Laasonen K Wonczak S Strey R and Laaksonen A
Molecular dynamics simulations of gasndashliquid nucleation
of LennardndashJones fluid J Chem Physics 2000 113 9741
40 Reiss H Methods of Thermodynamics 1996 (Dover
Publications New York)
41 Kondepudi D and Prigogine I Modern Thermodynamics
From Heat Engines to Dissipative Structures 1998 (John
Wiley Chichester West Sussex)
42 Smoluchowski M Ann Phys NY 1908 25 225
43 Einstein A Ann Phys NY 1910 338 1275
44 Hirschfelder J O Curtis C F and Bird R BMolecular
Theory of Gases and Liquids 1954 (John Wiley New
York)
45 Landau L D and Lifshitz E M Statistical Physics 1958
(Pergamon London)
46 Wu D T Nucleation theory Solid State Physicsmdash
Advances in Res and Applic 1997 50 37
47 Reguera D Rubı J M and Perez-Madrid A Fokkerndash
Planck equations for nucleation processes revisited
Physica A 1998 259 10
48 Schenter G K Kathmann S and Garrett B Variational
transition state theory of vapor phase nucleation
J Chem Physics 1999 109 7951
49 Onsager L Phys Rev 1931 37 405
50 Courtney W G J Chem Physics 1961 35 2249
51 Girshick S L and Chiu C-P Kinetic nucleation theory
a new expression for the rate of homogeneous nucleation
from an ideal supersaturated vapor J Chem Physics
1990 93 1273
52 Ford I J Barrett J C and Lazaridis M Uncertainties
in cluster energies in homogeneous nucleation theory
J Aerosol Sci 1993 24 581
53 Weakliem C L and Reiss H The factor 1S in the
classical theory of nucleation J Phys Chemistry 1994
98 6408
54 McClurg R B and Flagan R C Critical comparison of
droplet models in homogeneous nucleation theory
J Colloid Interface Sci 1998 201 194
55 Mashmoushy H Mahpeykar M R and Bakhtar F
Studies of nucleating and wet steam flows in two-dimen-
sional cascades Proc Instn Mech Engrs Part C J
Mechanical Engineering Science 2004 218(C8) 843ndash858
56 Ford I J Nucleation theorems the statistical mechanics
of molecular clusters and a revision of classical nuclea-
tion theory Phys Rev E 1997 56 5615
57 ten Wolde P R and Frenkel D Computer simulation
study of gasndashliquid nucleation in a LennardndashJones system
J Chem Physics 1998 109 9901
58 Kusaka I Wang Z G and Seinfeld J H Direct
evaluation of the equilibrium distribution of physical
clusters by a grand canonical Monte Carlo simulation
J Chem Physics 1998 108 3416
59 Kusaka I and Oxtoby D W Identifying physical
clusters in vapor phase nucleation J Chem Physics
1999 110 5249
60 Stillinger F H J Chem Physics 1963 38 1486
61 Ford I J Imperfect vaporndashgas mixtures and homo-
geneous nucleation J Aerosol Sci 1992 23 447
62 Lothe J and Pound G M J Chem Physics 1962 36
2080
63 Lee J K Barker J A and Abraham F F Theory and
Monte Carlo simulation of physical clusters in the
imperfect vapor J Chem Physics 1973 58 3166
64 Abraham F F Homogeneous Nucleation Theory 1974
(Academic Press New York)
65 Reif F Fundamentals of Statistical and Thermal Physics
1965 (McGraw-Hill New York)
66 Hoare M R and Pal P Adv Phys 1975 24 645
67 Hoare M R Adv Chem Physics 1979 40 49
68 Pal P and Hoare M R Thermodynamic properties
and homogeneous nucleation of molecular clusters of
nitrogen J Phys Chemistry 1987 91 2474
69 Reiss H Katz J L and Cohen E R Translationndash
rotation paradox in the theory of nucleation J Chem
Physics 1968 48 5553
70 Hale B N Monte Carlo calculations of effective surface
tension for small clusters Aust J Physics 1996 49 425
71 Kathmann S M Schenter G K and Garrett B C
Dynamical nucleation theory calculation of condensation
rate constants for small water clusters J Chem Physics
1999 111 4688
72 Reiss H Tabazadeh A and Talbot J Molecular theory
of vapor phase nucleation the physically consistent
cluster J Chem Physics 1990 92 1266
73 Ellerby H M Weakliem C L and Reiss H Toward a
molecular theory of vaporndashphase nucleation I Identifica-
tion of the average embryo J Chem Physics 1992 97
5766
74 Ellerby H M and Reiss H Toward a molecular theory
of vaporndashphase nucleation II Fundamental treatment of
the cluster distribution J Chem Physics 1992 97 5766
75 Weakliem C L and Reiss H Toward a molecular theory
of vaporndashphase nucleation III Thermodynamic properties
of argon clusters fromMonte Carlo simulations and modi-
fied liquid drop theory J Chem Physics 1994 101 2398
76 Kusaka I Wang Z G and Seinfeld J H Binary
nucleation of sulfuric acidndashwater Monte Carlo simula-
tion J Chem Physics 1998 108 6829
77 Oh K J and Zeng X C A small system ensemble
Monte Carlo simulation of supersaturated vapor
evaluation of barrier to nucleation J Chem Physics
2000 112 294
78 Senger B Schaaf P Corti D S Bowles R Voegel J
C and Reiss H A molecular theory of the homogeneous
nucleation rate I Formulation and fundamental issues
J Chem Physics 1999 110 6421
79 Hill T L Statistical Mechanics Principles and Selected
Applications 1956 (McGraw-Hill New York)
80 Soto R and Cordero P Nonideal gas of clusters at
equilibrium J Chem Physics 1999 110 7316
81 Barrett J C The significance of cluster lifetime in
nucleation theory J Chem Physics 2002 116 8856
82 Lushnikov A A and Kulmala M Dimers in nucleating
vapors Phys Rev E 1998 58 3157
83 Bahadur R and McClurg R B Comparison of cluster
definitions in homogeneous nucleation rate calculations
J Phys Chemistry 2001 105 11 893
84 Pugnaloni L A and Vericat F New criteria for cluster
identification in continuum systems J Chem Physics
2002 116 1097
STATISTICAL MECHANICS OF NUCLEATION 897
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
85 Chen B Siepmann J I Oh K J and Klein M L
Simulating vaporndashliquid nucleation of n-alkanes J Chem
Physics 2002 116 4317
86 Harris S A and Ford I J A dynamical definition of
quasi-bound molecular clusters J Chem Physics 2003
118 9216
87 Shreve A P Walton J R P B and Gubbins K E
Liquid drops of polar molecules J Chem Physics 1986
85 2178
88 ten Wolde P R Ruiz-Montero M J and Frenkel D
Numerical calculation of the rate of homogeneous gasndash
liquid nucleation in a LennardndashJones system J Chem
Physics 1998 109 9901
89 ten Wolde P R Oxtoby D W and Frenkel D Chain
formation in homogeneous gasndashliquid nucleation of polar
fluids J Chem Physics 1998 111 4762
90 Vehkamaki H and Ford I J Critical cluster size and
droplet nucleation rate from growth and decay simula-
tions of LennardndashJones clusters J Chem Physics 2000
112 4193
91 Schaaf P Senger B Voegel J C Bowles R K and
Reiss H Simulative determination of kinetic coefficients
for nucleation rates J Chem Physics 2001 114 8091
92 Chen B Siepmann J I Oh K J and Klein M L
Aggregation-volume-bias Monte Carlo simulations of
vaporndashliquid nucleation barriers for LennardndashJonesium
J Chem Physics 2001 115 10 903
93 Hale B N and DiMattio D J A Monte Carlo discrete
sum (MCDS) nucleation rate model for water In
Proceedings of the 15th International Conference on
Nucleation and Atmospheric Aerosols AIP Conference
Proceedings Vol 534 (Eds B N Hale and M Kulmala)
2000 pp 31ndash34
94 Kathmann S M and Hale B N Monte Carlo
simulations of small sulfuric acidndashwater clusters
J Phys Chemistry B 2001 105 11 719
95 Daura X Mark A E and van Gunsteren W F Para-
metrization of aliphatic CHn united atoms of GROMOS96
force field J Comput Chemistry 1998 19 535
96 Kathmann S M Schenter G K and Garrett B C
Understanding the sensitivity of nucleation kinetics a
case study on water J Chem Physics 2002 116 5046
97 Kashchiev D On the relation between nucleation work
nucleus size and nucleation rate J Chem Physics 1982
76 5098
98 Oxtoby D W and Kashchiev D A general relation
between the nucleation work and the size of the nucleus in
multicomponent nucleation J Chem Physics 1994 100
7665
99 Ford I J Thermodynamic properties of critical clusters
from measurements of vapourndashliquid homogeneous
nucleation rates J Chem Physics 1996 105 8324
100 Oxtoby D W and Evans R Nonclassical nucleation
theory for the gasndashliquid transition J Chem Physics
1988 89 7521
101 Zeng X C and Oxtoby D W Gasndashliquid nucleation in
LennardndashJones fluids J Chem Physics 1991 94 4472
102 Evans R Fundamentals of Inhomogeneous Fluids 1992
Ch 3 pp 85ndash175 (Marcel Dekker New York)
103 Talanquer V and Oxtoby D W Dynamical density
functional theory of gasndashliquid nucleation J Chem
Physics 1994 100 5190
104 Barrett J C First-order correction to classical nucleation
theory a density functional approach J Chem Physics
1999 111 5938
105 Barrett J C Cluster translation and growth in density
functional theories of homogeneous nucleation J Chem
Physics 1997 107 7989
106 Bykov T V and Zeng X C A patching model for
surface tension of spherical droplet and Tolman length J
Chem Physics 1999 111 10 602
107 Talanquer V and Oxtoby D W Nucleation in molecular
and dipolar fluids interaction site model J Chem
Physics 1995 103 3686
108 Seok C and Oxtoby D W Nucleation in n-alkanes a
density functional approach J Chem Physics 1998 109
7982
109 Reguera D Bowles R K Djikaev Y and Reiss H
Phase transitions in systems small enough to be clusters
J Chem Physics 2003 118 340
110 Khakshouri S and Ford I J J Chem Physics 2004
(submitted)
111 Ford I J (unpublished)
112 Nyquist R M Talanquer V and Oxtoby D W Density
functional theory of nucleation a semi-empirical
approach J Chem Physics 1995 103 1175
113 Talanquer V and Oxtoby D W Density functional
analysis of phenomenological theories of gasndashliquid
nucleation J Chem Physics 1995 99 2865
114 Katz J L Fisk J A and Chakarov V The accuracy of
nucleation theory In Proceedings of the 13th Interna-
tional Conference on Nucleation and Atmospheric Aerosols
(Eds N Fukuta and P E Wagner) 1992 pp 11ndash18
(Deepak Publishing Hampton Virginia)
115 Fisher M E The theory of condensation and the critical
point Physics 1967 3 255
116 Dillmann A and Meier G E A A refined droplet
approach to the problem of homogeneous nucleation
from the vapor phase J Chem Physics 1991 94 3872
117 Tolman R C J Chem Physics 1949 17 333
118 Ford I J Laaksonen A and Kulmala M Modification
of the DillmannndashMeier theory of homogeneous nuclea-
tion J Chem Physics 1993 99 764
119 Laaksonen A Ford I J and Kulmala M Revised
parametrization of the DillmannndashMeier theory of homo-
geneous nucleation Phys Rev E 1994 49 5517
120 Kalikmanov V I and van Dongen M E H Semipheno-
menological theory of homogeneous vaporndashliquid nuclea-
tion J Chem Physics 1995 103 4250
121 Hale B N Application of a scaled homogeneous
nucleation-rate formalism to experimental-data at
T 5Tc Phys Rev A 1986 33 4156
122 Martinez D M Ferguson F T Heist R H and Nuth
J A Application of scaled nucleation theory to metallic
vapor condensation J Chem Physics 2001 115 310
123 Hale B N and Wilemski G A scaled nucleation model
for ideal binary systems Chem Physics Lett 1999 305
263
124 Granasy L Diffuse interface theory for homogeneous
vapor condensation J Chem Physics 1996 104
5188
125 McGraw R and Laaksonen A Interfacial curvature free
energy the Kelvin relation and vaporndashliquid nucleation
rate J Chem Physics 1997 106 5284
I J FORD898
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
126 Talanquer V A new phenomenological approach to
gasndashliquid nucleation based on the scaling properties
of the critical nucleus J Chem Physics 1997 106
9957
127 Koga K and Zeng X C Thermodynamic expansion of
nucleation free-energy barrier and size of critical nucleus
near the vaporndashliquid coexistence J Chem Physics 1999
110 3466
128 Kusaka I On the scaling behavior of the free energetics of
nucleation J Chem Physics 2003 118 5510
129 Nowakowski B and Ruckenstein E A kinetic approach
to the theory of nucleation in gases J Chem Physics
1991 94 1397
130 Nowakowski B and Ruckenstein E Homogeneous
nucleation in gases a three-dimensional FokkerndashPlanck
equation for evaporation from clusters J Chem Physics
1991 94 8487
131 Wilcox C F and Bauer S H Estimation of homo-
geneous nucleation flux via a kinetic model J Chem
Physics 1991 94 8302
132 Hanggi P Talkner P and Borkovec M Reaction rate
theory fifty years after Kramers Rev Mod Physics 1990
62 251
133 Barrett J C Clement C F and Ford I J Energy
fluctuations in homogeneous nucleation theory for aero-
sols J Physics A 1993 26 529
134 Ford I J and Harris S A Molecular cluster decay
viewed as escape from a potential of mean force J Chem
Physics 2004 120 4428
135 Risken H The FokkerndashPlanck Equation Methods of
Solution and Applications 2nd edition 1989 (Springer-
Verlag Berlin)
136 Shizgal B and Barrett J C Time dependent nucleation
J Chem Physics 1989 91 6505
STATISTICAL MECHANICS OF NUCLEATION 899
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
refinement of this point of view using a treatment ofclusterndashcluster interactions to optimize the definingcluster volume has been presented [77] A review ofrecent progress in this area and the application of theseideas to nucleation theory has been provided by Sengeret al [78]
Such geometric constraints can be implemented buttheir realism is questionable Situations better regardedas close encounters between separate clusters would beclassed as snapshots of single larger clusters It shouldbe recalled that the cluster definition is an internalfeature of the model which is to be optimized to reduceclusterndashcluster interactions and to improve the repre-sentation of the grand partition function Misclassifyingconfigurations is fraught with consequences A secondissue to consider is that for convenience the clustersshould behave in a manner that can be modelled withBeckerndashDoring equations This means that the growthand decay rates should be free of memory effects ie thetransition processes should be Markovian The extent towhich this is satisfied (if indeed it is physically realistic)depends on the cluster definition
An intuitively reasonable choice of definition whichaddresses both these points is that a cluster shouldconsist of molecules that are bound energetically Thisshould minimize the clusterndashcluster contributions to thetotal energy According to such a definition if amolecule interacts strongly with a cluster then it islikely to be regarded as part of that cluster Also anenergetically bound cluster is likely to decay by thermalfluctuation in a stochastic Markovian fashion ratherthan by the separation of passing poorly boundfragments
Ideas have been proposed which require the cluster tobe energetically bound (in a loose sense since a cluster isintrinsically prone to decay) The time evolution of amolecular configuration and not simply the set ofinstantaneous positions of its component molecules isof central importance Hill [79] introduced a clusterdefinition requiring the total energies of pairs ofmolecules in a cluster to be negative (in their centre ofmass frame) The intention was to exclude situationswhere a molecule could leave the cluster within the timeit would take for a molecule to travel a typical lineardimension of the cluster Soto and Cordero [80] havedeveloped Hillrsquos criterion and Barrett [81] has per-formed Monte Carlo modelling of molecular configura-tions excluding those which decay in moleculardynamics within a given period A number of otherenergy-based criteria have been developed [82ndash85]
Harris and Ford [86] have recently proposed a newenergy-based cluster definition As with Barrettrsquosapproach it is based on the requirement that amolecular snapshot is energetically bound for a chosensubsequent period Evaporation according to thisscheme requires not only that a molecule should makea positive energy excursion but also that the subsequent
dynamics carry it far away from the cluster Onlyperhaps one in eight molecules manage to avoidrecapture once having satisfied the positive energycriterion [86] A snapshot of a cluster in this casesurrounded by a carrier gas is shown in Fig 4 Thehistory of the energy of a molecule in such a cluster isillustrated in Fig 5 three excursions into positive energyoccur in this sequence but only one leads to actualescape corresponding to the maintenance of positiveenergy for a significant time The instant of escape istaken to be the point at which the energy goes positivefor the third and final time
Cluster definitions implemented in moleculardynamics and Monte Carlo codes allow cluster freeenergies and hence the decay rate to be calculated Anenormous amount of simulation work has been carriedout and calculations can be laborious Many differentschemes and approximations have been made to easethe evaluation of these properties The principalconclusion from various studies using a range of clusterdefinitions and interaction models [49 57 58 70 85 87ndash92] seems to be that the work of formation of liquidcritical clusters does indeed contain a dominantcontribution proportional to i2=3 consistent with theclassical work of formation Hale and collaboratorspresent cluster free energies for water and other speciesin a way that makes this particularly clear indeed it issurprising how dominant the surface term can be foreven very small clusters [70 93 94] Neverthelessfurther contributions exist small clusters are notcharacterized just by scaled-down properties of largeclusters and it is impossible to escape the need todetermine the additional terms
In all the above discussion it has been assumed thatthe intermolecular interactions are accurately specifiedThis is a situation that is rarely possible Althoughfirst-principles quantum-mechanical models could be
Fig 4 Snapshot of a cluster of 50 argon atoms (brightspheres) bathed in a gas of helium (darker spheres)
STATISTICAL MECHANICS OF NUCLEATION 891
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
developed in practice empirical potentials are devel-oped on the basis of physical intuition containingparameters fitted to the measured properties of thesubstance in question [95] Usually these are bulkproperties and it is not always clear that the sameinteractions can be used for the very different circum-stances of a molecular cluster [96] In view of thesensitivity of nucleation rates to the thermodynamicproperties of clusters this is perhaps a crucial failingProgress can be made however by computing knowncluster properties where they are available and byrefining the potentials to suit Extraction of clusterproperties through use of the so-called nucleationtheorems (which relate derivatives of the nucleationrate to properties of the critical cluster [16 56 97ndash99])may provide just this sort of experimental information
A technique that has not yet been described is the useof density functional theory (DFT) to model theproperties of liquid droplets [100ndash104] This powerfulapplication of statistical mechanics promises to avoidlaborious computations at the microscopic level byinstead focusing on the density profile of moleculesmaking up the droplet This profile may be regarded asthe average distribution of molecules about the clustercentre of mass Although this might appear to involvecoarse-graining and approximation it is a well-foundedapproach and in principle can be exact In practical
implementations however a mean field approach istaken such that the constituent molecules move in aneffective potential well created by the other molecules inthe cluster [105] Also the interactions are assumed to becharacterized by strong repulsive cores and weaklyattractive tails
Considerable analytical work has been carried out[104 106] Various developments have been made todescribe droplets consisting of simple atoms as well aschain-like molecules [107] and polar molecules [108]The main drawback in these implementations is thatspecific interactions such as hydrogen bonds are hardto represent Certain details of the modelled systemare neglected [109 110] and the (implicit) clusterdefinition used is geometric and not energetic [111]On the other hand quite sophisticated systems havebeen modelled including multi-component dropletsThe method has been compared with otherapproaches [112 113] and certain features are betterdescribed particularly the approach to spinodalconditions where the nucleation barrier vanishesDFT calculations can be performed relatively easilyand they can provide a qualitative and perhapsquantitative insight into quite complicated clusteringproblems They provide an intermediate-level descrip-tion between the simple capillarity approximation anda microscopic description
Fig 5 The energy of a molecule in the centre-of-mass frame of a cluster occasionally becomes positive butthis does not necessarily lead to escape Only the third positive energy excursion in this sequenceindicated by an arrow results in particle escape A cluster definition must capture this behaviour
I J FORD892
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
7 PHENOMENOLOGICAL MODELS
The reader might be dismayed by the mathematicalcomplexity of the expressions for the work of formationof solid clusters in section 5 might be alarmed by theprocedural complexity of implementing a cluster defini-tion in section 6 may recoil from the difficulties incalculating free energies from molecular simulationmethods and ultimately may give up in view of theuncertainties in the underlying intermolecular interac-tions Considering these complications it is tempting totry to develop phenomenological models of the work offormation After all if the capillarity approximationworks reasonably well for water then surely a minorcorrection to account for small cluster effects ought tobe sufficient to improve matters
However after perhaps two decades of work nomodel has emerged offering broad success across arange of different substances It is perhaps asking toomuch to expect universal improvement CNT works as afirst approximation for water for a particular tempera-ture range but as discussed earlier it possesses thewrong temperature dependence so that predictionsaway from a temperature in the range 210ndash260K areuncertain A similar picture emerges for other sub-stances although in some cases the failure is even moresevere
The simplest phenomenological adjustment to equa-tion (4) the basis for CNT is to write
ncei frac14 n1 expfrac12DWclethiTHORN DWcleth1THORN
kT
eth23THORN
where DWclethiTHORN is given by equation (5) This amounts toa uniform downward shift of the classical free energybarrier illustrated in Fig 1 At least now the expressionis consistent at i frac14 1 The theory is known as internallyconsistent classical theory (ICCT) [51] Unfortunately itis not clear whether this constant shift in the work offormation should be a realistic representation of all thesubleading terms in the work of formation [52] It is anarbitrary modification and its success in accounting fordata is mixed For example nucleation rates for a rangeof alkanes [114] require further modification by factorsof between one and nine orders of magnitude to comeinto agreement with ICCT
It should be noted at this point that the correctionterm kT lnS which appears in DWcleth1THORN can beregarded as rigorous There are good reasons for thisbased on general considerations involving the nucleationtheorems [56] or more simply the principles of chemicalequilibrium represented by the law of mass action givenin equation (15) The correction converts the prefactorn1 in equation (4) into ns1 and results in the insertion of afactor 1=S in the nucleation rate in equation (1)
The Fisher droplet model is a slightly more sophis-ticated scheme [115] Here the constrained equilibrium
populations are modelled by
ncei frac14 ns1 exp yii2=3 thorn t ln i lnq0kT
psv
thorn i lnS
eth24THORN
The term proportional to yi may be viewed as a surfaceterm while the other two are corrections of a typesuggested by the free energy expressions developed insection 5 The parameters q0 yi and t are chosen toensure that the model correctly reproduces knownproperties of clusters and vapours including virialcoefficients and critical properties the theory is there-fore phenomenological
For large cluster sizes and for temperatures wellbelow the critical temperature the free energy offormation should tend towards that of a macroscopicliquid droplet yi should therefore approach A1s=ethkTTHORNfor large i where A1i
2=3 is the surface area 4pR2 of aspherical droplet containing i monomers at the bulkliquid density so that A1 frac14 eth36pTHORN1=3r2=3
l Dillmann andMeier [116] suggested writing
yi frac14kiA1skT
eth25THORN
with the Ansatz
ki frac14 1thorn a1i 1=3 thorn a2i 2=3 eth26THORN
The coefficients a1 and a2 can be found in terms of virialcoefficients of the vapour The form of this Ansatz wasmotivated by the curvature dependence of the bulksurface tension derived in classical thermodynamics byTolman [117]
The initial success of the DillmannndashMeier theory wasshort lived since modifications had to be made toovercome an internal inconsistency [118] Virial expan-sions for the vapour pressure and density were usedincorrectly to calculate the virial expansion of the ratioof the two Correcting this error [61] changed predictedrates by several orders of magnitude Nevertheless theapproach was very influential [119] In a furtherdevelopment Kalikmanov and van Dongen [120] seta2 frac14 0 (this term can be absorbed into q0 anyway) andchose a1 such that the experimental saturated vapourpressure psv is reproduced by the model The DillmannndashMeier model matched the first two virial coefficientsonly The Kalikmanovndashvan Dongen theory shouldtherefore approximate the (revised) DillmannndashMeiertheory at low temperatures but ought to provide abetter description at higher temperatures
Other phenomenological models have been proposedincluding the Hale scaling relation [121] This modelbased on CNT but employing simplified models for thesurface tension and liquid density based on scalingbehaviour near the critical point together with a rather
STATISTICAL MECHANICS OF NUCLEATION 893
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
more straightforward rate prefactor seems to be quitesuccessful in accounting for data for a wide range ofmaterials even metals [122] and two-component systems[123] Another phenomenological model is the diffuseinterface theory [124] which is based on a treatment ofthe phase interface using simplified density profiles ofthermodynamic properties Another general class ofmodel makes use of conditions at the vapourndashliquidspinodal to constrain the form of the model free energy[125ndash128]
The main drawback in phenomenological models isthe arbitrariness in the form of the proposed work offormation The parameter fitting in practice ensures thatthe populations of small clusters especially the mono-mer and dimer are well represented by the model butthe properties of larger clusters depend on the expres-sion assumed for the model and are therefore uncertainIntuition suggests that phenomenological modelling islikely to have mixed success and that microscopiccalculations of cluster properties do matter There mightbe better value in phenomenological models however ifthey were based securely on the results of microscopiccalculations
8 KINETICS OR THERMODYNAMICS
In section 3 kinetic equations representing the nuclea-tion process were described However the coefficients inthose equations were related to constrained equilibriumpopulations and hence to the equilibrium statisticalmechanics of clusters through the detailed balancecondition (8) Is it necessary to do this Could acompletely kinetic route be followed by evaluating themean growth and decay coefficients bi and gi directly bysome statistical approach The main problem is ofcourse the difficulty in performing such calculationsand this has restricted the development of such models[129ndash131] It is expected that the kinetic coefficients willcorrespond to those derived from the detailed balanceargument for near-equilibrium conditions but circum-stances can be imagined where non-equilibrium effectsare present which cannot be dealt with within standardstatistical mechanics
The framework for evaluating kinetic coefficients ofcluster decay can be briefly introduced [86 132] Amolecular dynamics simulation of a cluster can be set upand evolved according to the interactions until bychance an evaporation event occurs By extensivesimulation statistics for the mean evaporation rates gican be amassed in this way Similarly moleculardynamics simulations where a single molecule iscaptured by a cluster could be performed and thegrowth rates bi determined These will be functions ofcluster energy and so would fit best into a rate equationscheme involving both size and energy [133]
This is a laborious approach but statistical physicsoffers an approximate alternative representation of theescape and capture processes The basic tool is theLangevin equation The radial motion of a molecule inthe cluster is described by [134]
meurorr frac14 f ethrTHORN thornfethr _rr tTHORN eth27THORN
where r is the radial position m is the molecular mass fis the mean (time- and velocity-averaged) force on themolecule at position r and f is the stochastic forceThe latter is usually represented as a velocity-dependent friction term mg _rr where g is the so-calledfriction coefficient plus a fluctuating term ~ff ethr tTHORN withzero mean and correlation in time h ~ff ethr tTHORN~ff ethr t0THORNi frac14eth2gkTmTHORNdetht t0THORN where k is Boltzmannrsquos constant andwhere T has the characteristics of a temperature as willbe seen It is a standard manipulation [135] to convertthe Langevin description in the large friction limit intoa FokkerndashPlanck equation
qwqt
frac14 1
mg qethfwTHORN
qrthorn kT
q2wqr2
eth28THORN
which represents the evolution of wethr tTHORN the prob-ability density that the molecule should lie at the radialposition r It should be noted in passing that FokkerndashPlanck equations have often been used to describe thestochastic dynamics of a cluster in size space [48 136] Incontrast the use of the method described above refers tomotion of a molecule in real coordinate space relative tothe centre of mass of the cluster
The steady state equilibrium solution of equation (28)is
wethrTHORN exp FethrTHORNkT
eth29THORN
where FethrTHORN is called the potential of mean force relatedto the mean force on the molecule f throughf frac14 dF=dr It can now be seen how the parameter Tin the stochastic force plays the role of temperaturesince equation (29) looks like a Boltzmann distributionThis solution is not what is sought however The escapeproblem is characterized by a boundary conditionwethreTHORN frac14 0 where re is the radius at which a particleescapes from the system The resulting escape rate maybe shown to be [134]
gkin frac14 KethreT m gTHORN exp DFkT
eth30THORN
where the prefactor K depends on the parameters listedas well as the shape of the potential of mean force Theprincipal feature of equation (30) is the exponentialdependence on the depth DF of the potential of meanforce Therefore if the potential of mean force FethrTHORN can
I J FORD894
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
be evaluated by studying a molecular dynamics trajec-tory for example then the kinetic decay rate can bedetermined This could be extended to account for aradially dependent temperature and friction coefficientboth of which might be determined from simulation Asketch of a potential of mean force is given in Fig 6
Kinetically determined cluster growth and decayrates gkini and bkini respectively could provide a newroute for the determination of nucleation rates inde-pendent of assumptions about detailed balance and theuse of equilibrium statistical mechanics in a non-equilibrium situation
9 CONCLUSIONS
This review has been brief and in some ways partisan tothe personal views of the author It attempts to sketch acoherent picture of nucleation theory and how thevarious disparate strands relate to one another One ofthe fascinations of nucleation theory is the variety ofapproaches and the number of difficult issues that areencountered How well does free energy fluctuationtheory represent a real non-equilibrium process Howwell does the idea of detailed balance provide aframework for determining non-equilibrium decaycoefficients from equilibrium statistical mechanicsHow well do various methods perform in the evaluation
of cluster free energies Can cluster growth and decayrates be calculated directly without the need for such aframework How exactly are clusters to be defined
The theoretical problems are many but the issues canbe resolved to various degrees of satisfaction Thereremains of course the fundamental task of choosing anintermolecular potential that adequately represents thephysical interactions Without this all the finer points ofcluster definition and treatments of non-equilibriumprocesses are unimportant Modelling many problems inscience relies on the availability of an accurate micro-scopic description and so nucleation rate calculationsare not unique in this respect Most empirical molecularpotentials are fitted to the properties of bulk and surfaceproperties and in the very best examples attention isgiven to their transferability to more unusual environ-ments such as clusters In view of these capabilities andmodern theoretical developments in nucleation theoryit seems likely that successful microscopic models of thenucleation process based on statistical physics are atlast coming within reach
ACKNOWLEDGEMENTS
The present author is grateful to colleagues SarahHarris and Marco Arosio for providing some of thefigures
Fig 6 A molecule in a cluster may be considered to be confined within it by a potential of mean force FethrTHORN Itcan escape by reaching a radial position re driven by a phenomenological stochastic force
STATISTICAL MECHANICS OF NUCLEATION 895
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
REFERENCES
1 Gross D H E Microcanonical Thermodynamics 2001
(World Scientific Singapore)
2 Seinfeld J H and Pandis S N Atmospheric Chemistry
and Physics 1998 (John Wiley New York)
3 Climate Change 2001 Synthesis Report 2001 (Inter-
governmental Panel on Climate Change) (Cambridge
University Press Cambridge)
4 Kohler T M Gail H-P and Sedlmayr E MgO dust
nucleation in M-stars calculation of cluster properties
and nucleation rates Astronomy and Astrophysics 1997
320 553
5 Kashchiev D Nucleation Basic Theory with Applications
2000 (ButterworthndashHeinemann Oxford)
6 Bakhtar F Ebrahimi M and Webb R A On the
performance of a cascade of turbine rotor tip section
blading in nucleating steam Part 1 surface pressure
distributions Proc Instn Mech Engrs 1995 209(C2)
115ndash124
7 Bakhtar F Ebrahimi M and Bamkole B O On the
performance of a cascade of turbine rotor tip section
blading in nucleating steam Part 2 wake traverses Proc
Instn Mech Engrs 1995 209(C3) 169ndash178
8 Bakhtar F andMahpeykar M R On the performance of
a cascade of turbine rotor tip section blading in nucleating
steam Part 3 theoretical treatment Proc Instn Mech
Engrs Part C J Mechanical Engineering Science 1997
211(C3) 195ndash210
9 Volmer M and Weber A Z Physikalische Chemie 1926
119 277
10 Farkas L Z Physikalische Chemie (Leipzig) 1927 125
236
11 Becker R and Doring W Kinetische Behandlung der
Keinbildung in ubersattigten Dampfen Ann Physik
(Leipzig) 1935 24 719
12 Zeldovich J Soviet Physics JETP 1942 12 525
13 Frenkel J Kinetic Theory of Liquids 1946 (Oxford
University Press Oxford)
14 Wagner P E and Strey R Homogeneous nucleation
rates of water vapor measured in a two-piston expansion
chamber J Phys Chemistry 1981 85 2694
15 Miller R Anderson R J Kassner J L and Hagen
D E Homogeneous nucleation rate measurements for
water over a wide range of temperature and nucleation
rate J Chem Physics 1983 78 3204
16 Viisanen Y Strey R and Reiss H Homogeneous
nucleation rates for water J Chem Physics 1993 99 4680
17 Viisanen Y Strey R and Reiss H Erratum homo-
geneous nucleation rates for water J Chem Physics
2000 112 8205
18 Wolk J and Strey R Homogeneous nucleation of H2O
and D2O in comparison the isotope effect J Phys
Chemistry B 2001 105 11683
19 Heath C H Streletsky K and Wyslouzil B E H2Ondash
D2O condensation in a supersonic nozzle J Chem
Physics 2002 117 6176
20 Mikheev V B Irving P M Laulainen N S Barlow
S E and Pervukhin V V Laboratory measurements of
water nucleation using a laminar flow chamber J Chem
Physics 2002 116 10 772
21 Hung C H Krasnopoler M J and Katz J L
Condensation of a supersaturated vapor VIII The
homogeneous nucleation of n-nonane J Chem Physics
1989 90 1856
22 Rudek M M Fisk J A Chakarov V M and Katz
J L Condensation of a supersaturated vapor XII The
homogeneous nucleation of the n-alkanes J Chem
Physics 1996 105 4707
23 Strey R Wagner P E and Schmeling T Homogeneous
nucleation rates for n-alcohol vapors measured in a two-
piston expansion chamber J Chem Physics 1986 84
2325
24 Lihavainen H Viisanen Y and Kulmala M Homo-
geneous nucleation of n-pentanol in a laminar flow
diffusion chamber J Chem Physics 2001 114 10 031
25 Rudek M M Katz J L Vidensky I V Zdımal V and
Smolık J Homogeneous nucleation rates of n-pentanol
measured in an upward thermal diffusion cloud chamber
J Chem Physics 1999 111 3623
26 Grassmann A and Peters F Homogeneous nucleation
rates of n-propanol in nitrogen measured in a piston-
expansion tube J Chem Physics 2002 116 7617
27 Kane D and El-Shall S Condensation of supersaturated
vapors of hydrogen bonding molecules ethylene glycol
propylene glycol trimethyl glycol and glycerol J Chem
Physics 1996 105 7617
28 Hameri K and Kulmala M Homogeneous nucleation in
a laminar flow chamber the effect of temperature and
carrier gas pressure on dibutyl phthalate vapor nucleation
rate at high supersaturations J Chem Physics 1996 105
7696
29 Viisanen Y Strey R Laaksonen A and Kulmala M
Measurement of the molecular content of binary nuclei
II Use of the nucleation rate surface for waterndashethanol
J Chem Physics 1994 100 6062
30 Viisanen Y Wagner P E and Strey R Measurement of
the molecular content of binary nuclei IV Use of the
nucleation rate surfaces for the n-nonane-n-alcohol series
J Chem Physics 1998 108 4257
31 Fisk J A Rudek M M Katz J L Beiersdorf D and
Uchtmann H The homogeneous nucleation of cesium
vapor Atmos Res 1998 46 211
32 Uchtmann H Dettmer R Baranovskii S D and
Hensel F Photoinduced nucleation in supersaturated
mercury vapor J Chem Physics 1998 108 9775
33 Ferguson F T and Nuth J A Experimental studies of
the vapor phase nucleation of refractory compounds V
The condensation of lithium J Chem Physics 2000 113
4093
34 Laaksonen A Talanquer V and Oxtoby D W
Nucleation measurements theory and atmospheric
applications A Rev Phys Chemistry 1995 46 489
35 Debenedetti P G Metastable Liquids Concepts and
Principles 1996 (Princeton University Press Princeton
New Jersey)
36 Yasuoka K and Matsumoto M Molecular dynamics of
homogeneous nucleation in the vapor phase I Lennardndash
Jones fluid J Chem Physics 1998 109 8451
37 Yasuoka K and Matsumoto M Molecular dynamics of
homogeneous nucleation in the vapor phase II Water
J Chem Physics 1998 109 8463
I J FORD896
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
38 Wonczak S Strey R and Stauffer D Confirmation of
classical nucleation theory by Monte Carlo simulations in
the 3-dimensional Ising model at low temperature
J Chem Physics 2000 113 1976
39 Laasonen K Wonczak S Strey R and Laaksonen A
Molecular dynamics simulations of gasndashliquid nucleation
of LennardndashJones fluid J Chem Physics 2000 113 9741
40 Reiss H Methods of Thermodynamics 1996 (Dover
Publications New York)
41 Kondepudi D and Prigogine I Modern Thermodynamics
From Heat Engines to Dissipative Structures 1998 (John
Wiley Chichester West Sussex)
42 Smoluchowski M Ann Phys NY 1908 25 225
43 Einstein A Ann Phys NY 1910 338 1275
44 Hirschfelder J O Curtis C F and Bird R BMolecular
Theory of Gases and Liquids 1954 (John Wiley New
York)
45 Landau L D and Lifshitz E M Statistical Physics 1958
(Pergamon London)
46 Wu D T Nucleation theory Solid State Physicsmdash
Advances in Res and Applic 1997 50 37
47 Reguera D Rubı J M and Perez-Madrid A Fokkerndash
Planck equations for nucleation processes revisited
Physica A 1998 259 10
48 Schenter G K Kathmann S and Garrett B Variational
transition state theory of vapor phase nucleation
J Chem Physics 1999 109 7951
49 Onsager L Phys Rev 1931 37 405
50 Courtney W G J Chem Physics 1961 35 2249
51 Girshick S L and Chiu C-P Kinetic nucleation theory
a new expression for the rate of homogeneous nucleation
from an ideal supersaturated vapor J Chem Physics
1990 93 1273
52 Ford I J Barrett J C and Lazaridis M Uncertainties
in cluster energies in homogeneous nucleation theory
J Aerosol Sci 1993 24 581
53 Weakliem C L and Reiss H The factor 1S in the
classical theory of nucleation J Phys Chemistry 1994
98 6408
54 McClurg R B and Flagan R C Critical comparison of
droplet models in homogeneous nucleation theory
J Colloid Interface Sci 1998 201 194
55 Mashmoushy H Mahpeykar M R and Bakhtar F
Studies of nucleating and wet steam flows in two-dimen-
sional cascades Proc Instn Mech Engrs Part C J
Mechanical Engineering Science 2004 218(C8) 843ndash858
56 Ford I J Nucleation theorems the statistical mechanics
of molecular clusters and a revision of classical nuclea-
tion theory Phys Rev E 1997 56 5615
57 ten Wolde P R and Frenkel D Computer simulation
study of gasndashliquid nucleation in a LennardndashJones system
J Chem Physics 1998 109 9901
58 Kusaka I Wang Z G and Seinfeld J H Direct
evaluation of the equilibrium distribution of physical
clusters by a grand canonical Monte Carlo simulation
J Chem Physics 1998 108 3416
59 Kusaka I and Oxtoby D W Identifying physical
clusters in vapor phase nucleation J Chem Physics
1999 110 5249
60 Stillinger F H J Chem Physics 1963 38 1486
61 Ford I J Imperfect vaporndashgas mixtures and homo-
geneous nucleation J Aerosol Sci 1992 23 447
62 Lothe J and Pound G M J Chem Physics 1962 36
2080
63 Lee J K Barker J A and Abraham F F Theory and
Monte Carlo simulation of physical clusters in the
imperfect vapor J Chem Physics 1973 58 3166
64 Abraham F F Homogeneous Nucleation Theory 1974
(Academic Press New York)
65 Reif F Fundamentals of Statistical and Thermal Physics
1965 (McGraw-Hill New York)
66 Hoare M R and Pal P Adv Phys 1975 24 645
67 Hoare M R Adv Chem Physics 1979 40 49
68 Pal P and Hoare M R Thermodynamic properties
and homogeneous nucleation of molecular clusters of
nitrogen J Phys Chemistry 1987 91 2474
69 Reiss H Katz J L and Cohen E R Translationndash
rotation paradox in the theory of nucleation J Chem
Physics 1968 48 5553
70 Hale B N Monte Carlo calculations of effective surface
tension for small clusters Aust J Physics 1996 49 425
71 Kathmann S M Schenter G K and Garrett B C
Dynamical nucleation theory calculation of condensation
rate constants for small water clusters J Chem Physics
1999 111 4688
72 Reiss H Tabazadeh A and Talbot J Molecular theory
of vapor phase nucleation the physically consistent
cluster J Chem Physics 1990 92 1266
73 Ellerby H M Weakliem C L and Reiss H Toward a
molecular theory of vaporndashphase nucleation I Identifica-
tion of the average embryo J Chem Physics 1992 97
5766
74 Ellerby H M and Reiss H Toward a molecular theory
of vaporndashphase nucleation II Fundamental treatment of
the cluster distribution J Chem Physics 1992 97 5766
75 Weakliem C L and Reiss H Toward a molecular theory
of vaporndashphase nucleation III Thermodynamic properties
of argon clusters fromMonte Carlo simulations and modi-
fied liquid drop theory J Chem Physics 1994 101 2398
76 Kusaka I Wang Z G and Seinfeld J H Binary
nucleation of sulfuric acidndashwater Monte Carlo simula-
tion J Chem Physics 1998 108 6829
77 Oh K J and Zeng X C A small system ensemble
Monte Carlo simulation of supersaturated vapor
evaluation of barrier to nucleation J Chem Physics
2000 112 294
78 Senger B Schaaf P Corti D S Bowles R Voegel J
C and Reiss H A molecular theory of the homogeneous
nucleation rate I Formulation and fundamental issues
J Chem Physics 1999 110 6421
79 Hill T L Statistical Mechanics Principles and Selected
Applications 1956 (McGraw-Hill New York)
80 Soto R and Cordero P Nonideal gas of clusters at
equilibrium J Chem Physics 1999 110 7316
81 Barrett J C The significance of cluster lifetime in
nucleation theory J Chem Physics 2002 116 8856
82 Lushnikov A A and Kulmala M Dimers in nucleating
vapors Phys Rev E 1998 58 3157
83 Bahadur R and McClurg R B Comparison of cluster
definitions in homogeneous nucleation rate calculations
J Phys Chemistry 2001 105 11 893
84 Pugnaloni L A and Vericat F New criteria for cluster
identification in continuum systems J Chem Physics
2002 116 1097
STATISTICAL MECHANICS OF NUCLEATION 897
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
85 Chen B Siepmann J I Oh K J and Klein M L
Simulating vaporndashliquid nucleation of n-alkanes J Chem
Physics 2002 116 4317
86 Harris S A and Ford I J A dynamical definition of
quasi-bound molecular clusters J Chem Physics 2003
118 9216
87 Shreve A P Walton J R P B and Gubbins K E
Liquid drops of polar molecules J Chem Physics 1986
85 2178
88 ten Wolde P R Ruiz-Montero M J and Frenkel D
Numerical calculation of the rate of homogeneous gasndash
liquid nucleation in a LennardndashJones system J Chem
Physics 1998 109 9901
89 ten Wolde P R Oxtoby D W and Frenkel D Chain
formation in homogeneous gasndashliquid nucleation of polar
fluids J Chem Physics 1998 111 4762
90 Vehkamaki H and Ford I J Critical cluster size and
droplet nucleation rate from growth and decay simula-
tions of LennardndashJones clusters J Chem Physics 2000
112 4193
91 Schaaf P Senger B Voegel J C Bowles R K and
Reiss H Simulative determination of kinetic coefficients
for nucleation rates J Chem Physics 2001 114 8091
92 Chen B Siepmann J I Oh K J and Klein M L
Aggregation-volume-bias Monte Carlo simulations of
vaporndashliquid nucleation barriers for LennardndashJonesium
J Chem Physics 2001 115 10 903
93 Hale B N and DiMattio D J A Monte Carlo discrete
sum (MCDS) nucleation rate model for water In
Proceedings of the 15th International Conference on
Nucleation and Atmospheric Aerosols AIP Conference
Proceedings Vol 534 (Eds B N Hale and M Kulmala)
2000 pp 31ndash34
94 Kathmann S M and Hale B N Monte Carlo
simulations of small sulfuric acidndashwater clusters
J Phys Chemistry B 2001 105 11 719
95 Daura X Mark A E and van Gunsteren W F Para-
metrization of aliphatic CHn united atoms of GROMOS96
force field J Comput Chemistry 1998 19 535
96 Kathmann S M Schenter G K and Garrett B C
Understanding the sensitivity of nucleation kinetics a
case study on water J Chem Physics 2002 116 5046
97 Kashchiev D On the relation between nucleation work
nucleus size and nucleation rate J Chem Physics 1982
76 5098
98 Oxtoby D W and Kashchiev D A general relation
between the nucleation work and the size of the nucleus in
multicomponent nucleation J Chem Physics 1994 100
7665
99 Ford I J Thermodynamic properties of critical clusters
from measurements of vapourndashliquid homogeneous
nucleation rates J Chem Physics 1996 105 8324
100 Oxtoby D W and Evans R Nonclassical nucleation
theory for the gasndashliquid transition J Chem Physics
1988 89 7521
101 Zeng X C and Oxtoby D W Gasndashliquid nucleation in
LennardndashJones fluids J Chem Physics 1991 94 4472
102 Evans R Fundamentals of Inhomogeneous Fluids 1992
Ch 3 pp 85ndash175 (Marcel Dekker New York)
103 Talanquer V and Oxtoby D W Dynamical density
functional theory of gasndashliquid nucleation J Chem
Physics 1994 100 5190
104 Barrett J C First-order correction to classical nucleation
theory a density functional approach J Chem Physics
1999 111 5938
105 Barrett J C Cluster translation and growth in density
functional theories of homogeneous nucleation J Chem
Physics 1997 107 7989
106 Bykov T V and Zeng X C A patching model for
surface tension of spherical droplet and Tolman length J
Chem Physics 1999 111 10 602
107 Talanquer V and Oxtoby D W Nucleation in molecular
and dipolar fluids interaction site model J Chem
Physics 1995 103 3686
108 Seok C and Oxtoby D W Nucleation in n-alkanes a
density functional approach J Chem Physics 1998 109
7982
109 Reguera D Bowles R K Djikaev Y and Reiss H
Phase transitions in systems small enough to be clusters
J Chem Physics 2003 118 340
110 Khakshouri S and Ford I J J Chem Physics 2004
(submitted)
111 Ford I J (unpublished)
112 Nyquist R M Talanquer V and Oxtoby D W Density
functional theory of nucleation a semi-empirical
approach J Chem Physics 1995 103 1175
113 Talanquer V and Oxtoby D W Density functional
analysis of phenomenological theories of gasndashliquid
nucleation J Chem Physics 1995 99 2865
114 Katz J L Fisk J A and Chakarov V The accuracy of
nucleation theory In Proceedings of the 13th Interna-
tional Conference on Nucleation and Atmospheric Aerosols
(Eds N Fukuta and P E Wagner) 1992 pp 11ndash18
(Deepak Publishing Hampton Virginia)
115 Fisher M E The theory of condensation and the critical
point Physics 1967 3 255
116 Dillmann A and Meier G E A A refined droplet
approach to the problem of homogeneous nucleation
from the vapor phase J Chem Physics 1991 94 3872
117 Tolman R C J Chem Physics 1949 17 333
118 Ford I J Laaksonen A and Kulmala M Modification
of the DillmannndashMeier theory of homogeneous nuclea-
tion J Chem Physics 1993 99 764
119 Laaksonen A Ford I J and Kulmala M Revised
parametrization of the DillmannndashMeier theory of homo-
geneous nucleation Phys Rev E 1994 49 5517
120 Kalikmanov V I and van Dongen M E H Semipheno-
menological theory of homogeneous vaporndashliquid nuclea-
tion J Chem Physics 1995 103 4250
121 Hale B N Application of a scaled homogeneous
nucleation-rate formalism to experimental-data at
T 5Tc Phys Rev A 1986 33 4156
122 Martinez D M Ferguson F T Heist R H and Nuth
J A Application of scaled nucleation theory to metallic
vapor condensation J Chem Physics 2001 115 310
123 Hale B N and Wilemski G A scaled nucleation model
for ideal binary systems Chem Physics Lett 1999 305
263
124 Granasy L Diffuse interface theory for homogeneous
vapor condensation J Chem Physics 1996 104
5188
125 McGraw R and Laaksonen A Interfacial curvature free
energy the Kelvin relation and vaporndashliquid nucleation
rate J Chem Physics 1997 106 5284
I J FORD898
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
126 Talanquer V A new phenomenological approach to
gasndashliquid nucleation based on the scaling properties
of the critical nucleus J Chem Physics 1997 106
9957
127 Koga K and Zeng X C Thermodynamic expansion of
nucleation free-energy barrier and size of critical nucleus
near the vaporndashliquid coexistence J Chem Physics 1999
110 3466
128 Kusaka I On the scaling behavior of the free energetics of
nucleation J Chem Physics 2003 118 5510
129 Nowakowski B and Ruckenstein E A kinetic approach
to the theory of nucleation in gases J Chem Physics
1991 94 1397
130 Nowakowski B and Ruckenstein E Homogeneous
nucleation in gases a three-dimensional FokkerndashPlanck
equation for evaporation from clusters J Chem Physics
1991 94 8487
131 Wilcox C F and Bauer S H Estimation of homo-
geneous nucleation flux via a kinetic model J Chem
Physics 1991 94 8302
132 Hanggi P Talkner P and Borkovec M Reaction rate
theory fifty years after Kramers Rev Mod Physics 1990
62 251
133 Barrett J C Clement C F and Ford I J Energy
fluctuations in homogeneous nucleation theory for aero-
sols J Physics A 1993 26 529
134 Ford I J and Harris S A Molecular cluster decay
viewed as escape from a potential of mean force J Chem
Physics 2004 120 4428
135 Risken H The FokkerndashPlanck Equation Methods of
Solution and Applications 2nd edition 1989 (Springer-
Verlag Berlin)
136 Shizgal B and Barrett J C Time dependent nucleation
J Chem Physics 1989 91 6505
STATISTICAL MECHANICS OF NUCLEATION 899
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
developed in practice empirical potentials are devel-oped on the basis of physical intuition containingparameters fitted to the measured properties of thesubstance in question [95] Usually these are bulkproperties and it is not always clear that the sameinteractions can be used for the very different circum-stances of a molecular cluster [96] In view of thesensitivity of nucleation rates to the thermodynamicproperties of clusters this is perhaps a crucial failingProgress can be made however by computing knowncluster properties where they are available and byrefining the potentials to suit Extraction of clusterproperties through use of the so-called nucleationtheorems (which relate derivatives of the nucleationrate to properties of the critical cluster [16 56 97ndash99])may provide just this sort of experimental information
A technique that has not yet been described is the useof density functional theory (DFT) to model theproperties of liquid droplets [100ndash104] This powerfulapplication of statistical mechanics promises to avoidlaborious computations at the microscopic level byinstead focusing on the density profile of moleculesmaking up the droplet This profile may be regarded asthe average distribution of molecules about the clustercentre of mass Although this might appear to involvecoarse-graining and approximation it is a well-foundedapproach and in principle can be exact In practical
implementations however a mean field approach istaken such that the constituent molecules move in aneffective potential well created by the other molecules inthe cluster [105] Also the interactions are assumed to becharacterized by strong repulsive cores and weaklyattractive tails
Considerable analytical work has been carried out[104 106] Various developments have been made todescribe droplets consisting of simple atoms as well aschain-like molecules [107] and polar molecules [108]The main drawback in these implementations is thatspecific interactions such as hydrogen bonds are hardto represent Certain details of the modelled systemare neglected [109 110] and the (implicit) clusterdefinition used is geometric and not energetic [111]On the other hand quite sophisticated systems havebeen modelled including multi-component dropletsThe method has been compared with otherapproaches [112 113] and certain features are betterdescribed particularly the approach to spinodalconditions where the nucleation barrier vanishesDFT calculations can be performed relatively easilyand they can provide a qualitative and perhapsquantitative insight into quite complicated clusteringproblems They provide an intermediate-level descrip-tion between the simple capillarity approximation anda microscopic description
Fig 5 The energy of a molecule in the centre-of-mass frame of a cluster occasionally becomes positive butthis does not necessarily lead to escape Only the third positive energy excursion in this sequenceindicated by an arrow results in particle escape A cluster definition must capture this behaviour
I J FORD892
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
7 PHENOMENOLOGICAL MODELS
The reader might be dismayed by the mathematicalcomplexity of the expressions for the work of formationof solid clusters in section 5 might be alarmed by theprocedural complexity of implementing a cluster defini-tion in section 6 may recoil from the difficulties incalculating free energies from molecular simulationmethods and ultimately may give up in view of theuncertainties in the underlying intermolecular interac-tions Considering these complications it is tempting totry to develop phenomenological models of the work offormation After all if the capillarity approximationworks reasonably well for water then surely a minorcorrection to account for small cluster effects ought tobe sufficient to improve matters
However after perhaps two decades of work nomodel has emerged offering broad success across arange of different substances It is perhaps asking toomuch to expect universal improvement CNT works as afirst approximation for water for a particular tempera-ture range but as discussed earlier it possesses thewrong temperature dependence so that predictionsaway from a temperature in the range 210ndash260K areuncertain A similar picture emerges for other sub-stances although in some cases the failure is even moresevere
The simplest phenomenological adjustment to equa-tion (4) the basis for CNT is to write
ncei frac14 n1 expfrac12DWclethiTHORN DWcleth1THORN
kT
eth23THORN
where DWclethiTHORN is given by equation (5) This amounts toa uniform downward shift of the classical free energybarrier illustrated in Fig 1 At least now the expressionis consistent at i frac14 1 The theory is known as internallyconsistent classical theory (ICCT) [51] Unfortunately itis not clear whether this constant shift in the work offormation should be a realistic representation of all thesubleading terms in the work of formation [52] It is anarbitrary modification and its success in accounting fordata is mixed For example nucleation rates for a rangeof alkanes [114] require further modification by factorsof between one and nine orders of magnitude to comeinto agreement with ICCT
It should be noted at this point that the correctionterm kT lnS which appears in DWcleth1THORN can beregarded as rigorous There are good reasons for thisbased on general considerations involving the nucleationtheorems [56] or more simply the principles of chemicalequilibrium represented by the law of mass action givenin equation (15) The correction converts the prefactorn1 in equation (4) into ns1 and results in the insertion of afactor 1=S in the nucleation rate in equation (1)
The Fisher droplet model is a slightly more sophis-ticated scheme [115] Here the constrained equilibrium
populations are modelled by
ncei frac14 ns1 exp yii2=3 thorn t ln i lnq0kT
psv
thorn i lnS
eth24THORN
The term proportional to yi may be viewed as a surfaceterm while the other two are corrections of a typesuggested by the free energy expressions developed insection 5 The parameters q0 yi and t are chosen toensure that the model correctly reproduces knownproperties of clusters and vapours including virialcoefficients and critical properties the theory is there-fore phenomenological
For large cluster sizes and for temperatures wellbelow the critical temperature the free energy offormation should tend towards that of a macroscopicliquid droplet yi should therefore approach A1s=ethkTTHORNfor large i where A1i
2=3 is the surface area 4pR2 of aspherical droplet containing i monomers at the bulkliquid density so that A1 frac14 eth36pTHORN1=3r2=3
l Dillmann andMeier [116] suggested writing
yi frac14kiA1skT
eth25THORN
with the Ansatz
ki frac14 1thorn a1i 1=3 thorn a2i 2=3 eth26THORN
The coefficients a1 and a2 can be found in terms of virialcoefficients of the vapour The form of this Ansatz wasmotivated by the curvature dependence of the bulksurface tension derived in classical thermodynamics byTolman [117]
The initial success of the DillmannndashMeier theory wasshort lived since modifications had to be made toovercome an internal inconsistency [118] Virial expan-sions for the vapour pressure and density were usedincorrectly to calculate the virial expansion of the ratioof the two Correcting this error [61] changed predictedrates by several orders of magnitude Nevertheless theapproach was very influential [119] In a furtherdevelopment Kalikmanov and van Dongen [120] seta2 frac14 0 (this term can be absorbed into q0 anyway) andchose a1 such that the experimental saturated vapourpressure psv is reproduced by the model The DillmannndashMeier model matched the first two virial coefficientsonly The Kalikmanovndashvan Dongen theory shouldtherefore approximate the (revised) DillmannndashMeiertheory at low temperatures but ought to provide abetter description at higher temperatures
Other phenomenological models have been proposedincluding the Hale scaling relation [121] This modelbased on CNT but employing simplified models for thesurface tension and liquid density based on scalingbehaviour near the critical point together with a rather
STATISTICAL MECHANICS OF NUCLEATION 893
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
more straightforward rate prefactor seems to be quitesuccessful in accounting for data for a wide range ofmaterials even metals [122] and two-component systems[123] Another phenomenological model is the diffuseinterface theory [124] which is based on a treatment ofthe phase interface using simplified density profiles ofthermodynamic properties Another general class ofmodel makes use of conditions at the vapourndashliquidspinodal to constrain the form of the model free energy[125ndash128]
The main drawback in phenomenological models isthe arbitrariness in the form of the proposed work offormation The parameter fitting in practice ensures thatthe populations of small clusters especially the mono-mer and dimer are well represented by the model butthe properties of larger clusters depend on the expres-sion assumed for the model and are therefore uncertainIntuition suggests that phenomenological modelling islikely to have mixed success and that microscopiccalculations of cluster properties do matter There mightbe better value in phenomenological models however ifthey were based securely on the results of microscopiccalculations
8 KINETICS OR THERMODYNAMICS
In section 3 kinetic equations representing the nuclea-tion process were described However the coefficients inthose equations were related to constrained equilibriumpopulations and hence to the equilibrium statisticalmechanics of clusters through the detailed balancecondition (8) Is it necessary to do this Could acompletely kinetic route be followed by evaluating themean growth and decay coefficients bi and gi directly bysome statistical approach The main problem is ofcourse the difficulty in performing such calculationsand this has restricted the development of such models[129ndash131] It is expected that the kinetic coefficients willcorrespond to those derived from the detailed balanceargument for near-equilibrium conditions but circum-stances can be imagined where non-equilibrium effectsare present which cannot be dealt with within standardstatistical mechanics
The framework for evaluating kinetic coefficients ofcluster decay can be briefly introduced [86 132] Amolecular dynamics simulation of a cluster can be set upand evolved according to the interactions until bychance an evaporation event occurs By extensivesimulation statistics for the mean evaporation rates gican be amassed in this way Similarly moleculardynamics simulations where a single molecule iscaptured by a cluster could be performed and thegrowth rates bi determined These will be functions ofcluster energy and so would fit best into a rate equationscheme involving both size and energy [133]
This is a laborious approach but statistical physicsoffers an approximate alternative representation of theescape and capture processes The basic tool is theLangevin equation The radial motion of a molecule inthe cluster is described by [134]
meurorr frac14 f ethrTHORN thornfethr _rr tTHORN eth27THORN
where r is the radial position m is the molecular mass fis the mean (time- and velocity-averaged) force on themolecule at position r and f is the stochastic forceThe latter is usually represented as a velocity-dependent friction term mg _rr where g is the so-calledfriction coefficient plus a fluctuating term ~ff ethr tTHORN withzero mean and correlation in time h ~ff ethr tTHORN~ff ethr t0THORNi frac14eth2gkTmTHORNdetht t0THORN where k is Boltzmannrsquos constant andwhere T has the characteristics of a temperature as willbe seen It is a standard manipulation [135] to convertthe Langevin description in the large friction limit intoa FokkerndashPlanck equation
qwqt
frac14 1
mg qethfwTHORN
qrthorn kT
q2wqr2
eth28THORN
which represents the evolution of wethr tTHORN the prob-ability density that the molecule should lie at the radialposition r It should be noted in passing that FokkerndashPlanck equations have often been used to describe thestochastic dynamics of a cluster in size space [48 136] Incontrast the use of the method described above refers tomotion of a molecule in real coordinate space relative tothe centre of mass of the cluster
The steady state equilibrium solution of equation (28)is
wethrTHORN exp FethrTHORNkT
eth29THORN
where FethrTHORN is called the potential of mean force relatedto the mean force on the molecule f throughf frac14 dF=dr It can now be seen how the parameter Tin the stochastic force plays the role of temperaturesince equation (29) looks like a Boltzmann distributionThis solution is not what is sought however The escapeproblem is characterized by a boundary conditionwethreTHORN frac14 0 where re is the radius at which a particleescapes from the system The resulting escape rate maybe shown to be [134]
gkin frac14 KethreT m gTHORN exp DFkT
eth30THORN
where the prefactor K depends on the parameters listedas well as the shape of the potential of mean force Theprincipal feature of equation (30) is the exponentialdependence on the depth DF of the potential of meanforce Therefore if the potential of mean force FethrTHORN can
I J FORD894
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
be evaluated by studying a molecular dynamics trajec-tory for example then the kinetic decay rate can bedetermined This could be extended to account for aradially dependent temperature and friction coefficientboth of which might be determined from simulation Asketch of a potential of mean force is given in Fig 6
Kinetically determined cluster growth and decayrates gkini and bkini respectively could provide a newroute for the determination of nucleation rates inde-pendent of assumptions about detailed balance and theuse of equilibrium statistical mechanics in a non-equilibrium situation
9 CONCLUSIONS
This review has been brief and in some ways partisan tothe personal views of the author It attempts to sketch acoherent picture of nucleation theory and how thevarious disparate strands relate to one another One ofthe fascinations of nucleation theory is the variety ofapproaches and the number of difficult issues that areencountered How well does free energy fluctuationtheory represent a real non-equilibrium process Howwell does the idea of detailed balance provide aframework for determining non-equilibrium decaycoefficients from equilibrium statistical mechanicsHow well do various methods perform in the evaluation
of cluster free energies Can cluster growth and decayrates be calculated directly without the need for such aframework How exactly are clusters to be defined
The theoretical problems are many but the issues canbe resolved to various degrees of satisfaction Thereremains of course the fundamental task of choosing anintermolecular potential that adequately represents thephysical interactions Without this all the finer points ofcluster definition and treatments of non-equilibriumprocesses are unimportant Modelling many problems inscience relies on the availability of an accurate micro-scopic description and so nucleation rate calculationsare not unique in this respect Most empirical molecularpotentials are fitted to the properties of bulk and surfaceproperties and in the very best examples attention isgiven to their transferability to more unusual environ-ments such as clusters In view of these capabilities andmodern theoretical developments in nucleation theoryit seems likely that successful microscopic models of thenucleation process based on statistical physics are atlast coming within reach
ACKNOWLEDGEMENTS
The present author is grateful to colleagues SarahHarris and Marco Arosio for providing some of thefigures
Fig 6 A molecule in a cluster may be considered to be confined within it by a potential of mean force FethrTHORN Itcan escape by reaching a radial position re driven by a phenomenological stochastic force
STATISTICAL MECHANICS OF NUCLEATION 895
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
REFERENCES
1 Gross D H E Microcanonical Thermodynamics 2001
(World Scientific Singapore)
2 Seinfeld J H and Pandis S N Atmospheric Chemistry
and Physics 1998 (John Wiley New York)
3 Climate Change 2001 Synthesis Report 2001 (Inter-
governmental Panel on Climate Change) (Cambridge
University Press Cambridge)
4 Kohler T M Gail H-P and Sedlmayr E MgO dust
nucleation in M-stars calculation of cluster properties
and nucleation rates Astronomy and Astrophysics 1997
320 553
5 Kashchiev D Nucleation Basic Theory with Applications
2000 (ButterworthndashHeinemann Oxford)
6 Bakhtar F Ebrahimi M and Webb R A On the
performance of a cascade of turbine rotor tip section
blading in nucleating steam Part 1 surface pressure
distributions Proc Instn Mech Engrs 1995 209(C2)
115ndash124
7 Bakhtar F Ebrahimi M and Bamkole B O On the
performance of a cascade of turbine rotor tip section
blading in nucleating steam Part 2 wake traverses Proc
Instn Mech Engrs 1995 209(C3) 169ndash178
8 Bakhtar F andMahpeykar M R On the performance of
a cascade of turbine rotor tip section blading in nucleating
steam Part 3 theoretical treatment Proc Instn Mech
Engrs Part C J Mechanical Engineering Science 1997
211(C3) 195ndash210
9 Volmer M and Weber A Z Physikalische Chemie 1926
119 277
10 Farkas L Z Physikalische Chemie (Leipzig) 1927 125
236
11 Becker R and Doring W Kinetische Behandlung der
Keinbildung in ubersattigten Dampfen Ann Physik
(Leipzig) 1935 24 719
12 Zeldovich J Soviet Physics JETP 1942 12 525
13 Frenkel J Kinetic Theory of Liquids 1946 (Oxford
University Press Oxford)
14 Wagner P E and Strey R Homogeneous nucleation
rates of water vapor measured in a two-piston expansion
chamber J Phys Chemistry 1981 85 2694
15 Miller R Anderson R J Kassner J L and Hagen
D E Homogeneous nucleation rate measurements for
water over a wide range of temperature and nucleation
rate J Chem Physics 1983 78 3204
16 Viisanen Y Strey R and Reiss H Homogeneous
nucleation rates for water J Chem Physics 1993 99 4680
17 Viisanen Y Strey R and Reiss H Erratum homo-
geneous nucleation rates for water J Chem Physics
2000 112 8205
18 Wolk J and Strey R Homogeneous nucleation of H2O
and D2O in comparison the isotope effect J Phys
Chemistry B 2001 105 11683
19 Heath C H Streletsky K and Wyslouzil B E H2Ondash
D2O condensation in a supersonic nozzle J Chem
Physics 2002 117 6176
20 Mikheev V B Irving P M Laulainen N S Barlow
S E and Pervukhin V V Laboratory measurements of
water nucleation using a laminar flow chamber J Chem
Physics 2002 116 10 772
21 Hung C H Krasnopoler M J and Katz J L
Condensation of a supersaturated vapor VIII The
homogeneous nucleation of n-nonane J Chem Physics
1989 90 1856
22 Rudek M M Fisk J A Chakarov V M and Katz
J L Condensation of a supersaturated vapor XII The
homogeneous nucleation of the n-alkanes J Chem
Physics 1996 105 4707
23 Strey R Wagner P E and Schmeling T Homogeneous
nucleation rates for n-alcohol vapors measured in a two-
piston expansion chamber J Chem Physics 1986 84
2325
24 Lihavainen H Viisanen Y and Kulmala M Homo-
geneous nucleation of n-pentanol in a laminar flow
diffusion chamber J Chem Physics 2001 114 10 031
25 Rudek M M Katz J L Vidensky I V Zdımal V and
Smolık J Homogeneous nucleation rates of n-pentanol
measured in an upward thermal diffusion cloud chamber
J Chem Physics 1999 111 3623
26 Grassmann A and Peters F Homogeneous nucleation
rates of n-propanol in nitrogen measured in a piston-
expansion tube J Chem Physics 2002 116 7617
27 Kane D and El-Shall S Condensation of supersaturated
vapors of hydrogen bonding molecules ethylene glycol
propylene glycol trimethyl glycol and glycerol J Chem
Physics 1996 105 7617
28 Hameri K and Kulmala M Homogeneous nucleation in
a laminar flow chamber the effect of temperature and
carrier gas pressure on dibutyl phthalate vapor nucleation
rate at high supersaturations J Chem Physics 1996 105
7696
29 Viisanen Y Strey R Laaksonen A and Kulmala M
Measurement of the molecular content of binary nuclei
II Use of the nucleation rate surface for waterndashethanol
J Chem Physics 1994 100 6062
30 Viisanen Y Wagner P E and Strey R Measurement of
the molecular content of binary nuclei IV Use of the
nucleation rate surfaces for the n-nonane-n-alcohol series
J Chem Physics 1998 108 4257
31 Fisk J A Rudek M M Katz J L Beiersdorf D and
Uchtmann H The homogeneous nucleation of cesium
vapor Atmos Res 1998 46 211
32 Uchtmann H Dettmer R Baranovskii S D and
Hensel F Photoinduced nucleation in supersaturated
mercury vapor J Chem Physics 1998 108 9775
33 Ferguson F T and Nuth J A Experimental studies of
the vapor phase nucleation of refractory compounds V
The condensation of lithium J Chem Physics 2000 113
4093
34 Laaksonen A Talanquer V and Oxtoby D W
Nucleation measurements theory and atmospheric
applications A Rev Phys Chemistry 1995 46 489
35 Debenedetti P G Metastable Liquids Concepts and
Principles 1996 (Princeton University Press Princeton
New Jersey)
36 Yasuoka K and Matsumoto M Molecular dynamics of
homogeneous nucleation in the vapor phase I Lennardndash
Jones fluid J Chem Physics 1998 109 8451
37 Yasuoka K and Matsumoto M Molecular dynamics of
homogeneous nucleation in the vapor phase II Water
J Chem Physics 1998 109 8463
I J FORD896
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
38 Wonczak S Strey R and Stauffer D Confirmation of
classical nucleation theory by Monte Carlo simulations in
the 3-dimensional Ising model at low temperature
J Chem Physics 2000 113 1976
39 Laasonen K Wonczak S Strey R and Laaksonen A
Molecular dynamics simulations of gasndashliquid nucleation
of LennardndashJones fluid J Chem Physics 2000 113 9741
40 Reiss H Methods of Thermodynamics 1996 (Dover
Publications New York)
41 Kondepudi D and Prigogine I Modern Thermodynamics
From Heat Engines to Dissipative Structures 1998 (John
Wiley Chichester West Sussex)
42 Smoluchowski M Ann Phys NY 1908 25 225
43 Einstein A Ann Phys NY 1910 338 1275
44 Hirschfelder J O Curtis C F and Bird R BMolecular
Theory of Gases and Liquids 1954 (John Wiley New
York)
45 Landau L D and Lifshitz E M Statistical Physics 1958
(Pergamon London)
46 Wu D T Nucleation theory Solid State Physicsmdash
Advances in Res and Applic 1997 50 37
47 Reguera D Rubı J M and Perez-Madrid A Fokkerndash
Planck equations for nucleation processes revisited
Physica A 1998 259 10
48 Schenter G K Kathmann S and Garrett B Variational
transition state theory of vapor phase nucleation
J Chem Physics 1999 109 7951
49 Onsager L Phys Rev 1931 37 405
50 Courtney W G J Chem Physics 1961 35 2249
51 Girshick S L and Chiu C-P Kinetic nucleation theory
a new expression for the rate of homogeneous nucleation
from an ideal supersaturated vapor J Chem Physics
1990 93 1273
52 Ford I J Barrett J C and Lazaridis M Uncertainties
in cluster energies in homogeneous nucleation theory
J Aerosol Sci 1993 24 581
53 Weakliem C L and Reiss H The factor 1S in the
classical theory of nucleation J Phys Chemistry 1994
98 6408
54 McClurg R B and Flagan R C Critical comparison of
droplet models in homogeneous nucleation theory
J Colloid Interface Sci 1998 201 194
55 Mashmoushy H Mahpeykar M R and Bakhtar F
Studies of nucleating and wet steam flows in two-dimen-
sional cascades Proc Instn Mech Engrs Part C J
Mechanical Engineering Science 2004 218(C8) 843ndash858
56 Ford I J Nucleation theorems the statistical mechanics
of molecular clusters and a revision of classical nuclea-
tion theory Phys Rev E 1997 56 5615
57 ten Wolde P R and Frenkel D Computer simulation
study of gasndashliquid nucleation in a LennardndashJones system
J Chem Physics 1998 109 9901
58 Kusaka I Wang Z G and Seinfeld J H Direct
evaluation of the equilibrium distribution of physical
clusters by a grand canonical Monte Carlo simulation
J Chem Physics 1998 108 3416
59 Kusaka I and Oxtoby D W Identifying physical
clusters in vapor phase nucleation J Chem Physics
1999 110 5249
60 Stillinger F H J Chem Physics 1963 38 1486
61 Ford I J Imperfect vaporndashgas mixtures and homo-
geneous nucleation J Aerosol Sci 1992 23 447
62 Lothe J and Pound G M J Chem Physics 1962 36
2080
63 Lee J K Barker J A and Abraham F F Theory and
Monte Carlo simulation of physical clusters in the
imperfect vapor J Chem Physics 1973 58 3166
64 Abraham F F Homogeneous Nucleation Theory 1974
(Academic Press New York)
65 Reif F Fundamentals of Statistical and Thermal Physics
1965 (McGraw-Hill New York)
66 Hoare M R and Pal P Adv Phys 1975 24 645
67 Hoare M R Adv Chem Physics 1979 40 49
68 Pal P and Hoare M R Thermodynamic properties
and homogeneous nucleation of molecular clusters of
nitrogen J Phys Chemistry 1987 91 2474
69 Reiss H Katz J L and Cohen E R Translationndash
rotation paradox in the theory of nucleation J Chem
Physics 1968 48 5553
70 Hale B N Monte Carlo calculations of effective surface
tension for small clusters Aust J Physics 1996 49 425
71 Kathmann S M Schenter G K and Garrett B C
Dynamical nucleation theory calculation of condensation
rate constants for small water clusters J Chem Physics
1999 111 4688
72 Reiss H Tabazadeh A and Talbot J Molecular theory
of vapor phase nucleation the physically consistent
cluster J Chem Physics 1990 92 1266
73 Ellerby H M Weakliem C L and Reiss H Toward a
molecular theory of vaporndashphase nucleation I Identifica-
tion of the average embryo J Chem Physics 1992 97
5766
74 Ellerby H M and Reiss H Toward a molecular theory
of vaporndashphase nucleation II Fundamental treatment of
the cluster distribution J Chem Physics 1992 97 5766
75 Weakliem C L and Reiss H Toward a molecular theory
of vaporndashphase nucleation III Thermodynamic properties
of argon clusters fromMonte Carlo simulations and modi-
fied liquid drop theory J Chem Physics 1994 101 2398
76 Kusaka I Wang Z G and Seinfeld J H Binary
nucleation of sulfuric acidndashwater Monte Carlo simula-
tion J Chem Physics 1998 108 6829
77 Oh K J and Zeng X C A small system ensemble
Monte Carlo simulation of supersaturated vapor
evaluation of barrier to nucleation J Chem Physics
2000 112 294
78 Senger B Schaaf P Corti D S Bowles R Voegel J
C and Reiss H A molecular theory of the homogeneous
nucleation rate I Formulation and fundamental issues
J Chem Physics 1999 110 6421
79 Hill T L Statistical Mechanics Principles and Selected
Applications 1956 (McGraw-Hill New York)
80 Soto R and Cordero P Nonideal gas of clusters at
equilibrium J Chem Physics 1999 110 7316
81 Barrett J C The significance of cluster lifetime in
nucleation theory J Chem Physics 2002 116 8856
82 Lushnikov A A and Kulmala M Dimers in nucleating
vapors Phys Rev E 1998 58 3157
83 Bahadur R and McClurg R B Comparison of cluster
definitions in homogeneous nucleation rate calculations
J Phys Chemistry 2001 105 11 893
84 Pugnaloni L A and Vericat F New criteria for cluster
identification in continuum systems J Chem Physics
2002 116 1097
STATISTICAL MECHANICS OF NUCLEATION 897
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
85 Chen B Siepmann J I Oh K J and Klein M L
Simulating vaporndashliquid nucleation of n-alkanes J Chem
Physics 2002 116 4317
86 Harris S A and Ford I J A dynamical definition of
quasi-bound molecular clusters J Chem Physics 2003
118 9216
87 Shreve A P Walton J R P B and Gubbins K E
Liquid drops of polar molecules J Chem Physics 1986
85 2178
88 ten Wolde P R Ruiz-Montero M J and Frenkel D
Numerical calculation of the rate of homogeneous gasndash
liquid nucleation in a LennardndashJones system J Chem
Physics 1998 109 9901
89 ten Wolde P R Oxtoby D W and Frenkel D Chain
formation in homogeneous gasndashliquid nucleation of polar
fluids J Chem Physics 1998 111 4762
90 Vehkamaki H and Ford I J Critical cluster size and
droplet nucleation rate from growth and decay simula-
tions of LennardndashJones clusters J Chem Physics 2000
112 4193
91 Schaaf P Senger B Voegel J C Bowles R K and
Reiss H Simulative determination of kinetic coefficients
for nucleation rates J Chem Physics 2001 114 8091
92 Chen B Siepmann J I Oh K J and Klein M L
Aggregation-volume-bias Monte Carlo simulations of
vaporndashliquid nucleation barriers for LennardndashJonesium
J Chem Physics 2001 115 10 903
93 Hale B N and DiMattio D J A Monte Carlo discrete
sum (MCDS) nucleation rate model for water In
Proceedings of the 15th International Conference on
Nucleation and Atmospheric Aerosols AIP Conference
Proceedings Vol 534 (Eds B N Hale and M Kulmala)
2000 pp 31ndash34
94 Kathmann S M and Hale B N Monte Carlo
simulations of small sulfuric acidndashwater clusters
J Phys Chemistry B 2001 105 11 719
95 Daura X Mark A E and van Gunsteren W F Para-
metrization of aliphatic CHn united atoms of GROMOS96
force field J Comput Chemistry 1998 19 535
96 Kathmann S M Schenter G K and Garrett B C
Understanding the sensitivity of nucleation kinetics a
case study on water J Chem Physics 2002 116 5046
97 Kashchiev D On the relation between nucleation work
nucleus size and nucleation rate J Chem Physics 1982
76 5098
98 Oxtoby D W and Kashchiev D A general relation
between the nucleation work and the size of the nucleus in
multicomponent nucleation J Chem Physics 1994 100
7665
99 Ford I J Thermodynamic properties of critical clusters
from measurements of vapourndashliquid homogeneous
nucleation rates J Chem Physics 1996 105 8324
100 Oxtoby D W and Evans R Nonclassical nucleation
theory for the gasndashliquid transition J Chem Physics
1988 89 7521
101 Zeng X C and Oxtoby D W Gasndashliquid nucleation in
LennardndashJones fluids J Chem Physics 1991 94 4472
102 Evans R Fundamentals of Inhomogeneous Fluids 1992
Ch 3 pp 85ndash175 (Marcel Dekker New York)
103 Talanquer V and Oxtoby D W Dynamical density
functional theory of gasndashliquid nucleation J Chem
Physics 1994 100 5190
104 Barrett J C First-order correction to classical nucleation
theory a density functional approach J Chem Physics
1999 111 5938
105 Barrett J C Cluster translation and growth in density
functional theories of homogeneous nucleation J Chem
Physics 1997 107 7989
106 Bykov T V and Zeng X C A patching model for
surface tension of spherical droplet and Tolman length J
Chem Physics 1999 111 10 602
107 Talanquer V and Oxtoby D W Nucleation in molecular
and dipolar fluids interaction site model J Chem
Physics 1995 103 3686
108 Seok C and Oxtoby D W Nucleation in n-alkanes a
density functional approach J Chem Physics 1998 109
7982
109 Reguera D Bowles R K Djikaev Y and Reiss H
Phase transitions in systems small enough to be clusters
J Chem Physics 2003 118 340
110 Khakshouri S and Ford I J J Chem Physics 2004
(submitted)
111 Ford I J (unpublished)
112 Nyquist R M Talanquer V and Oxtoby D W Density
functional theory of nucleation a semi-empirical
approach J Chem Physics 1995 103 1175
113 Talanquer V and Oxtoby D W Density functional
analysis of phenomenological theories of gasndashliquid
nucleation J Chem Physics 1995 99 2865
114 Katz J L Fisk J A and Chakarov V The accuracy of
nucleation theory In Proceedings of the 13th Interna-
tional Conference on Nucleation and Atmospheric Aerosols
(Eds N Fukuta and P E Wagner) 1992 pp 11ndash18
(Deepak Publishing Hampton Virginia)
115 Fisher M E The theory of condensation and the critical
point Physics 1967 3 255
116 Dillmann A and Meier G E A A refined droplet
approach to the problem of homogeneous nucleation
from the vapor phase J Chem Physics 1991 94 3872
117 Tolman R C J Chem Physics 1949 17 333
118 Ford I J Laaksonen A and Kulmala M Modification
of the DillmannndashMeier theory of homogeneous nuclea-
tion J Chem Physics 1993 99 764
119 Laaksonen A Ford I J and Kulmala M Revised
parametrization of the DillmannndashMeier theory of homo-
geneous nucleation Phys Rev E 1994 49 5517
120 Kalikmanov V I and van Dongen M E H Semipheno-
menological theory of homogeneous vaporndashliquid nuclea-
tion J Chem Physics 1995 103 4250
121 Hale B N Application of a scaled homogeneous
nucleation-rate formalism to experimental-data at
T 5Tc Phys Rev A 1986 33 4156
122 Martinez D M Ferguson F T Heist R H and Nuth
J A Application of scaled nucleation theory to metallic
vapor condensation J Chem Physics 2001 115 310
123 Hale B N and Wilemski G A scaled nucleation model
for ideal binary systems Chem Physics Lett 1999 305
263
124 Granasy L Diffuse interface theory for homogeneous
vapor condensation J Chem Physics 1996 104
5188
125 McGraw R and Laaksonen A Interfacial curvature free
energy the Kelvin relation and vaporndashliquid nucleation
rate J Chem Physics 1997 106 5284
I J FORD898
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
126 Talanquer V A new phenomenological approach to
gasndashliquid nucleation based on the scaling properties
of the critical nucleus J Chem Physics 1997 106
9957
127 Koga K and Zeng X C Thermodynamic expansion of
nucleation free-energy barrier and size of critical nucleus
near the vaporndashliquid coexistence J Chem Physics 1999
110 3466
128 Kusaka I On the scaling behavior of the free energetics of
nucleation J Chem Physics 2003 118 5510
129 Nowakowski B and Ruckenstein E A kinetic approach
to the theory of nucleation in gases J Chem Physics
1991 94 1397
130 Nowakowski B and Ruckenstein E Homogeneous
nucleation in gases a three-dimensional FokkerndashPlanck
equation for evaporation from clusters J Chem Physics
1991 94 8487
131 Wilcox C F and Bauer S H Estimation of homo-
geneous nucleation flux via a kinetic model J Chem
Physics 1991 94 8302
132 Hanggi P Talkner P and Borkovec M Reaction rate
theory fifty years after Kramers Rev Mod Physics 1990
62 251
133 Barrett J C Clement C F and Ford I J Energy
fluctuations in homogeneous nucleation theory for aero-
sols J Physics A 1993 26 529
134 Ford I J and Harris S A Molecular cluster decay
viewed as escape from a potential of mean force J Chem
Physics 2004 120 4428
135 Risken H The FokkerndashPlanck Equation Methods of
Solution and Applications 2nd edition 1989 (Springer-
Verlag Berlin)
136 Shizgal B and Barrett J C Time dependent nucleation
J Chem Physics 1989 91 6505
STATISTICAL MECHANICS OF NUCLEATION 899
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
7 PHENOMENOLOGICAL MODELS
The reader might be dismayed by the mathematicalcomplexity of the expressions for the work of formationof solid clusters in section 5 might be alarmed by theprocedural complexity of implementing a cluster defini-tion in section 6 may recoil from the difficulties incalculating free energies from molecular simulationmethods and ultimately may give up in view of theuncertainties in the underlying intermolecular interac-tions Considering these complications it is tempting totry to develop phenomenological models of the work offormation After all if the capillarity approximationworks reasonably well for water then surely a minorcorrection to account for small cluster effects ought tobe sufficient to improve matters
However after perhaps two decades of work nomodel has emerged offering broad success across arange of different substances It is perhaps asking toomuch to expect universal improvement CNT works as afirst approximation for water for a particular tempera-ture range but as discussed earlier it possesses thewrong temperature dependence so that predictionsaway from a temperature in the range 210ndash260K areuncertain A similar picture emerges for other sub-stances although in some cases the failure is even moresevere
The simplest phenomenological adjustment to equa-tion (4) the basis for CNT is to write
ncei frac14 n1 expfrac12DWclethiTHORN DWcleth1THORN
kT
eth23THORN
where DWclethiTHORN is given by equation (5) This amounts toa uniform downward shift of the classical free energybarrier illustrated in Fig 1 At least now the expressionis consistent at i frac14 1 The theory is known as internallyconsistent classical theory (ICCT) [51] Unfortunately itis not clear whether this constant shift in the work offormation should be a realistic representation of all thesubleading terms in the work of formation [52] It is anarbitrary modification and its success in accounting fordata is mixed For example nucleation rates for a rangeof alkanes [114] require further modification by factorsof between one and nine orders of magnitude to comeinto agreement with ICCT
It should be noted at this point that the correctionterm kT lnS which appears in DWcleth1THORN can beregarded as rigorous There are good reasons for thisbased on general considerations involving the nucleationtheorems [56] or more simply the principles of chemicalequilibrium represented by the law of mass action givenin equation (15) The correction converts the prefactorn1 in equation (4) into ns1 and results in the insertion of afactor 1=S in the nucleation rate in equation (1)
The Fisher droplet model is a slightly more sophis-ticated scheme [115] Here the constrained equilibrium
populations are modelled by
ncei frac14 ns1 exp yii2=3 thorn t ln i lnq0kT
psv
thorn i lnS
eth24THORN
The term proportional to yi may be viewed as a surfaceterm while the other two are corrections of a typesuggested by the free energy expressions developed insection 5 The parameters q0 yi and t are chosen toensure that the model correctly reproduces knownproperties of clusters and vapours including virialcoefficients and critical properties the theory is there-fore phenomenological
For large cluster sizes and for temperatures wellbelow the critical temperature the free energy offormation should tend towards that of a macroscopicliquid droplet yi should therefore approach A1s=ethkTTHORNfor large i where A1i
2=3 is the surface area 4pR2 of aspherical droplet containing i monomers at the bulkliquid density so that A1 frac14 eth36pTHORN1=3r2=3
l Dillmann andMeier [116] suggested writing
yi frac14kiA1skT
eth25THORN
with the Ansatz
ki frac14 1thorn a1i 1=3 thorn a2i 2=3 eth26THORN
The coefficients a1 and a2 can be found in terms of virialcoefficients of the vapour The form of this Ansatz wasmotivated by the curvature dependence of the bulksurface tension derived in classical thermodynamics byTolman [117]
The initial success of the DillmannndashMeier theory wasshort lived since modifications had to be made toovercome an internal inconsistency [118] Virial expan-sions for the vapour pressure and density were usedincorrectly to calculate the virial expansion of the ratioof the two Correcting this error [61] changed predictedrates by several orders of magnitude Nevertheless theapproach was very influential [119] In a furtherdevelopment Kalikmanov and van Dongen [120] seta2 frac14 0 (this term can be absorbed into q0 anyway) andchose a1 such that the experimental saturated vapourpressure psv is reproduced by the model The DillmannndashMeier model matched the first two virial coefficientsonly The Kalikmanovndashvan Dongen theory shouldtherefore approximate the (revised) DillmannndashMeiertheory at low temperatures but ought to provide abetter description at higher temperatures
Other phenomenological models have been proposedincluding the Hale scaling relation [121] This modelbased on CNT but employing simplified models for thesurface tension and liquid density based on scalingbehaviour near the critical point together with a rather
STATISTICAL MECHANICS OF NUCLEATION 893
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
more straightforward rate prefactor seems to be quitesuccessful in accounting for data for a wide range ofmaterials even metals [122] and two-component systems[123] Another phenomenological model is the diffuseinterface theory [124] which is based on a treatment ofthe phase interface using simplified density profiles ofthermodynamic properties Another general class ofmodel makes use of conditions at the vapourndashliquidspinodal to constrain the form of the model free energy[125ndash128]
The main drawback in phenomenological models isthe arbitrariness in the form of the proposed work offormation The parameter fitting in practice ensures thatthe populations of small clusters especially the mono-mer and dimer are well represented by the model butthe properties of larger clusters depend on the expres-sion assumed for the model and are therefore uncertainIntuition suggests that phenomenological modelling islikely to have mixed success and that microscopiccalculations of cluster properties do matter There mightbe better value in phenomenological models however ifthey were based securely on the results of microscopiccalculations
8 KINETICS OR THERMODYNAMICS
In section 3 kinetic equations representing the nuclea-tion process were described However the coefficients inthose equations were related to constrained equilibriumpopulations and hence to the equilibrium statisticalmechanics of clusters through the detailed balancecondition (8) Is it necessary to do this Could acompletely kinetic route be followed by evaluating themean growth and decay coefficients bi and gi directly bysome statistical approach The main problem is ofcourse the difficulty in performing such calculationsand this has restricted the development of such models[129ndash131] It is expected that the kinetic coefficients willcorrespond to those derived from the detailed balanceargument for near-equilibrium conditions but circum-stances can be imagined where non-equilibrium effectsare present which cannot be dealt with within standardstatistical mechanics
The framework for evaluating kinetic coefficients ofcluster decay can be briefly introduced [86 132] Amolecular dynamics simulation of a cluster can be set upand evolved according to the interactions until bychance an evaporation event occurs By extensivesimulation statistics for the mean evaporation rates gican be amassed in this way Similarly moleculardynamics simulations where a single molecule iscaptured by a cluster could be performed and thegrowth rates bi determined These will be functions ofcluster energy and so would fit best into a rate equationscheme involving both size and energy [133]
This is a laborious approach but statistical physicsoffers an approximate alternative representation of theescape and capture processes The basic tool is theLangevin equation The radial motion of a molecule inthe cluster is described by [134]
meurorr frac14 f ethrTHORN thornfethr _rr tTHORN eth27THORN
where r is the radial position m is the molecular mass fis the mean (time- and velocity-averaged) force on themolecule at position r and f is the stochastic forceThe latter is usually represented as a velocity-dependent friction term mg _rr where g is the so-calledfriction coefficient plus a fluctuating term ~ff ethr tTHORN withzero mean and correlation in time h ~ff ethr tTHORN~ff ethr t0THORNi frac14eth2gkTmTHORNdetht t0THORN where k is Boltzmannrsquos constant andwhere T has the characteristics of a temperature as willbe seen It is a standard manipulation [135] to convertthe Langevin description in the large friction limit intoa FokkerndashPlanck equation
qwqt
frac14 1
mg qethfwTHORN
qrthorn kT
q2wqr2
eth28THORN
which represents the evolution of wethr tTHORN the prob-ability density that the molecule should lie at the radialposition r It should be noted in passing that FokkerndashPlanck equations have often been used to describe thestochastic dynamics of a cluster in size space [48 136] Incontrast the use of the method described above refers tomotion of a molecule in real coordinate space relative tothe centre of mass of the cluster
The steady state equilibrium solution of equation (28)is
wethrTHORN exp FethrTHORNkT
eth29THORN
where FethrTHORN is called the potential of mean force relatedto the mean force on the molecule f throughf frac14 dF=dr It can now be seen how the parameter Tin the stochastic force plays the role of temperaturesince equation (29) looks like a Boltzmann distributionThis solution is not what is sought however The escapeproblem is characterized by a boundary conditionwethreTHORN frac14 0 where re is the radius at which a particleescapes from the system The resulting escape rate maybe shown to be [134]
gkin frac14 KethreT m gTHORN exp DFkT
eth30THORN
where the prefactor K depends on the parameters listedas well as the shape of the potential of mean force Theprincipal feature of equation (30) is the exponentialdependence on the depth DF of the potential of meanforce Therefore if the potential of mean force FethrTHORN can
I J FORD894
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
be evaluated by studying a molecular dynamics trajec-tory for example then the kinetic decay rate can bedetermined This could be extended to account for aradially dependent temperature and friction coefficientboth of which might be determined from simulation Asketch of a potential of mean force is given in Fig 6
Kinetically determined cluster growth and decayrates gkini and bkini respectively could provide a newroute for the determination of nucleation rates inde-pendent of assumptions about detailed balance and theuse of equilibrium statistical mechanics in a non-equilibrium situation
9 CONCLUSIONS
This review has been brief and in some ways partisan tothe personal views of the author It attempts to sketch acoherent picture of nucleation theory and how thevarious disparate strands relate to one another One ofthe fascinations of nucleation theory is the variety ofapproaches and the number of difficult issues that areencountered How well does free energy fluctuationtheory represent a real non-equilibrium process Howwell does the idea of detailed balance provide aframework for determining non-equilibrium decaycoefficients from equilibrium statistical mechanicsHow well do various methods perform in the evaluation
of cluster free energies Can cluster growth and decayrates be calculated directly without the need for such aframework How exactly are clusters to be defined
The theoretical problems are many but the issues canbe resolved to various degrees of satisfaction Thereremains of course the fundamental task of choosing anintermolecular potential that adequately represents thephysical interactions Without this all the finer points ofcluster definition and treatments of non-equilibriumprocesses are unimportant Modelling many problems inscience relies on the availability of an accurate micro-scopic description and so nucleation rate calculationsare not unique in this respect Most empirical molecularpotentials are fitted to the properties of bulk and surfaceproperties and in the very best examples attention isgiven to their transferability to more unusual environ-ments such as clusters In view of these capabilities andmodern theoretical developments in nucleation theoryit seems likely that successful microscopic models of thenucleation process based on statistical physics are atlast coming within reach
ACKNOWLEDGEMENTS
The present author is grateful to colleagues SarahHarris and Marco Arosio for providing some of thefigures
Fig 6 A molecule in a cluster may be considered to be confined within it by a potential of mean force FethrTHORN Itcan escape by reaching a radial position re driven by a phenomenological stochastic force
STATISTICAL MECHANICS OF NUCLEATION 895
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
REFERENCES
1 Gross D H E Microcanonical Thermodynamics 2001
(World Scientific Singapore)
2 Seinfeld J H and Pandis S N Atmospheric Chemistry
and Physics 1998 (John Wiley New York)
3 Climate Change 2001 Synthesis Report 2001 (Inter-
governmental Panel on Climate Change) (Cambridge
University Press Cambridge)
4 Kohler T M Gail H-P and Sedlmayr E MgO dust
nucleation in M-stars calculation of cluster properties
and nucleation rates Astronomy and Astrophysics 1997
320 553
5 Kashchiev D Nucleation Basic Theory with Applications
2000 (ButterworthndashHeinemann Oxford)
6 Bakhtar F Ebrahimi M and Webb R A On the
performance of a cascade of turbine rotor tip section
blading in nucleating steam Part 1 surface pressure
distributions Proc Instn Mech Engrs 1995 209(C2)
115ndash124
7 Bakhtar F Ebrahimi M and Bamkole B O On the
performance of a cascade of turbine rotor tip section
blading in nucleating steam Part 2 wake traverses Proc
Instn Mech Engrs 1995 209(C3) 169ndash178
8 Bakhtar F andMahpeykar M R On the performance of
a cascade of turbine rotor tip section blading in nucleating
steam Part 3 theoretical treatment Proc Instn Mech
Engrs Part C J Mechanical Engineering Science 1997
211(C3) 195ndash210
9 Volmer M and Weber A Z Physikalische Chemie 1926
119 277
10 Farkas L Z Physikalische Chemie (Leipzig) 1927 125
236
11 Becker R and Doring W Kinetische Behandlung der
Keinbildung in ubersattigten Dampfen Ann Physik
(Leipzig) 1935 24 719
12 Zeldovich J Soviet Physics JETP 1942 12 525
13 Frenkel J Kinetic Theory of Liquids 1946 (Oxford
University Press Oxford)
14 Wagner P E and Strey R Homogeneous nucleation
rates of water vapor measured in a two-piston expansion
chamber J Phys Chemistry 1981 85 2694
15 Miller R Anderson R J Kassner J L and Hagen
D E Homogeneous nucleation rate measurements for
water over a wide range of temperature and nucleation
rate J Chem Physics 1983 78 3204
16 Viisanen Y Strey R and Reiss H Homogeneous
nucleation rates for water J Chem Physics 1993 99 4680
17 Viisanen Y Strey R and Reiss H Erratum homo-
geneous nucleation rates for water J Chem Physics
2000 112 8205
18 Wolk J and Strey R Homogeneous nucleation of H2O
and D2O in comparison the isotope effect J Phys
Chemistry B 2001 105 11683
19 Heath C H Streletsky K and Wyslouzil B E H2Ondash
D2O condensation in a supersonic nozzle J Chem
Physics 2002 117 6176
20 Mikheev V B Irving P M Laulainen N S Barlow
S E and Pervukhin V V Laboratory measurements of
water nucleation using a laminar flow chamber J Chem
Physics 2002 116 10 772
21 Hung C H Krasnopoler M J and Katz J L
Condensation of a supersaturated vapor VIII The
homogeneous nucleation of n-nonane J Chem Physics
1989 90 1856
22 Rudek M M Fisk J A Chakarov V M and Katz
J L Condensation of a supersaturated vapor XII The
homogeneous nucleation of the n-alkanes J Chem
Physics 1996 105 4707
23 Strey R Wagner P E and Schmeling T Homogeneous
nucleation rates for n-alcohol vapors measured in a two-
piston expansion chamber J Chem Physics 1986 84
2325
24 Lihavainen H Viisanen Y and Kulmala M Homo-
geneous nucleation of n-pentanol in a laminar flow
diffusion chamber J Chem Physics 2001 114 10 031
25 Rudek M M Katz J L Vidensky I V Zdımal V and
Smolık J Homogeneous nucleation rates of n-pentanol
measured in an upward thermal diffusion cloud chamber
J Chem Physics 1999 111 3623
26 Grassmann A and Peters F Homogeneous nucleation
rates of n-propanol in nitrogen measured in a piston-
expansion tube J Chem Physics 2002 116 7617
27 Kane D and El-Shall S Condensation of supersaturated
vapors of hydrogen bonding molecules ethylene glycol
propylene glycol trimethyl glycol and glycerol J Chem
Physics 1996 105 7617
28 Hameri K and Kulmala M Homogeneous nucleation in
a laminar flow chamber the effect of temperature and
carrier gas pressure on dibutyl phthalate vapor nucleation
rate at high supersaturations J Chem Physics 1996 105
7696
29 Viisanen Y Strey R Laaksonen A and Kulmala M
Measurement of the molecular content of binary nuclei
II Use of the nucleation rate surface for waterndashethanol
J Chem Physics 1994 100 6062
30 Viisanen Y Wagner P E and Strey R Measurement of
the molecular content of binary nuclei IV Use of the
nucleation rate surfaces for the n-nonane-n-alcohol series
J Chem Physics 1998 108 4257
31 Fisk J A Rudek M M Katz J L Beiersdorf D and
Uchtmann H The homogeneous nucleation of cesium
vapor Atmos Res 1998 46 211
32 Uchtmann H Dettmer R Baranovskii S D and
Hensel F Photoinduced nucleation in supersaturated
mercury vapor J Chem Physics 1998 108 9775
33 Ferguson F T and Nuth J A Experimental studies of
the vapor phase nucleation of refractory compounds V
The condensation of lithium J Chem Physics 2000 113
4093
34 Laaksonen A Talanquer V and Oxtoby D W
Nucleation measurements theory and atmospheric
applications A Rev Phys Chemistry 1995 46 489
35 Debenedetti P G Metastable Liquids Concepts and
Principles 1996 (Princeton University Press Princeton
New Jersey)
36 Yasuoka K and Matsumoto M Molecular dynamics of
homogeneous nucleation in the vapor phase I Lennardndash
Jones fluid J Chem Physics 1998 109 8451
37 Yasuoka K and Matsumoto M Molecular dynamics of
homogeneous nucleation in the vapor phase II Water
J Chem Physics 1998 109 8463
I J FORD896
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
38 Wonczak S Strey R and Stauffer D Confirmation of
classical nucleation theory by Monte Carlo simulations in
the 3-dimensional Ising model at low temperature
J Chem Physics 2000 113 1976
39 Laasonen K Wonczak S Strey R and Laaksonen A
Molecular dynamics simulations of gasndashliquid nucleation
of LennardndashJones fluid J Chem Physics 2000 113 9741
40 Reiss H Methods of Thermodynamics 1996 (Dover
Publications New York)
41 Kondepudi D and Prigogine I Modern Thermodynamics
From Heat Engines to Dissipative Structures 1998 (John
Wiley Chichester West Sussex)
42 Smoluchowski M Ann Phys NY 1908 25 225
43 Einstein A Ann Phys NY 1910 338 1275
44 Hirschfelder J O Curtis C F and Bird R BMolecular
Theory of Gases and Liquids 1954 (John Wiley New
York)
45 Landau L D and Lifshitz E M Statistical Physics 1958
(Pergamon London)
46 Wu D T Nucleation theory Solid State Physicsmdash
Advances in Res and Applic 1997 50 37
47 Reguera D Rubı J M and Perez-Madrid A Fokkerndash
Planck equations for nucleation processes revisited
Physica A 1998 259 10
48 Schenter G K Kathmann S and Garrett B Variational
transition state theory of vapor phase nucleation
J Chem Physics 1999 109 7951
49 Onsager L Phys Rev 1931 37 405
50 Courtney W G J Chem Physics 1961 35 2249
51 Girshick S L and Chiu C-P Kinetic nucleation theory
a new expression for the rate of homogeneous nucleation
from an ideal supersaturated vapor J Chem Physics
1990 93 1273
52 Ford I J Barrett J C and Lazaridis M Uncertainties
in cluster energies in homogeneous nucleation theory
J Aerosol Sci 1993 24 581
53 Weakliem C L and Reiss H The factor 1S in the
classical theory of nucleation J Phys Chemistry 1994
98 6408
54 McClurg R B and Flagan R C Critical comparison of
droplet models in homogeneous nucleation theory
J Colloid Interface Sci 1998 201 194
55 Mashmoushy H Mahpeykar M R and Bakhtar F
Studies of nucleating and wet steam flows in two-dimen-
sional cascades Proc Instn Mech Engrs Part C J
Mechanical Engineering Science 2004 218(C8) 843ndash858
56 Ford I J Nucleation theorems the statistical mechanics
of molecular clusters and a revision of classical nuclea-
tion theory Phys Rev E 1997 56 5615
57 ten Wolde P R and Frenkel D Computer simulation
study of gasndashliquid nucleation in a LennardndashJones system
J Chem Physics 1998 109 9901
58 Kusaka I Wang Z G and Seinfeld J H Direct
evaluation of the equilibrium distribution of physical
clusters by a grand canonical Monte Carlo simulation
J Chem Physics 1998 108 3416
59 Kusaka I and Oxtoby D W Identifying physical
clusters in vapor phase nucleation J Chem Physics
1999 110 5249
60 Stillinger F H J Chem Physics 1963 38 1486
61 Ford I J Imperfect vaporndashgas mixtures and homo-
geneous nucleation J Aerosol Sci 1992 23 447
62 Lothe J and Pound G M J Chem Physics 1962 36
2080
63 Lee J K Barker J A and Abraham F F Theory and
Monte Carlo simulation of physical clusters in the
imperfect vapor J Chem Physics 1973 58 3166
64 Abraham F F Homogeneous Nucleation Theory 1974
(Academic Press New York)
65 Reif F Fundamentals of Statistical and Thermal Physics
1965 (McGraw-Hill New York)
66 Hoare M R and Pal P Adv Phys 1975 24 645
67 Hoare M R Adv Chem Physics 1979 40 49
68 Pal P and Hoare M R Thermodynamic properties
and homogeneous nucleation of molecular clusters of
nitrogen J Phys Chemistry 1987 91 2474
69 Reiss H Katz J L and Cohen E R Translationndash
rotation paradox in the theory of nucleation J Chem
Physics 1968 48 5553
70 Hale B N Monte Carlo calculations of effective surface
tension for small clusters Aust J Physics 1996 49 425
71 Kathmann S M Schenter G K and Garrett B C
Dynamical nucleation theory calculation of condensation
rate constants for small water clusters J Chem Physics
1999 111 4688
72 Reiss H Tabazadeh A and Talbot J Molecular theory
of vapor phase nucleation the physically consistent
cluster J Chem Physics 1990 92 1266
73 Ellerby H M Weakliem C L and Reiss H Toward a
molecular theory of vaporndashphase nucleation I Identifica-
tion of the average embryo J Chem Physics 1992 97
5766
74 Ellerby H M and Reiss H Toward a molecular theory
of vaporndashphase nucleation II Fundamental treatment of
the cluster distribution J Chem Physics 1992 97 5766
75 Weakliem C L and Reiss H Toward a molecular theory
of vaporndashphase nucleation III Thermodynamic properties
of argon clusters fromMonte Carlo simulations and modi-
fied liquid drop theory J Chem Physics 1994 101 2398
76 Kusaka I Wang Z G and Seinfeld J H Binary
nucleation of sulfuric acidndashwater Monte Carlo simula-
tion J Chem Physics 1998 108 6829
77 Oh K J and Zeng X C A small system ensemble
Monte Carlo simulation of supersaturated vapor
evaluation of barrier to nucleation J Chem Physics
2000 112 294
78 Senger B Schaaf P Corti D S Bowles R Voegel J
C and Reiss H A molecular theory of the homogeneous
nucleation rate I Formulation and fundamental issues
J Chem Physics 1999 110 6421
79 Hill T L Statistical Mechanics Principles and Selected
Applications 1956 (McGraw-Hill New York)
80 Soto R and Cordero P Nonideal gas of clusters at
equilibrium J Chem Physics 1999 110 7316
81 Barrett J C The significance of cluster lifetime in
nucleation theory J Chem Physics 2002 116 8856
82 Lushnikov A A and Kulmala M Dimers in nucleating
vapors Phys Rev E 1998 58 3157
83 Bahadur R and McClurg R B Comparison of cluster
definitions in homogeneous nucleation rate calculations
J Phys Chemistry 2001 105 11 893
84 Pugnaloni L A and Vericat F New criteria for cluster
identification in continuum systems J Chem Physics
2002 116 1097
STATISTICAL MECHANICS OF NUCLEATION 897
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
85 Chen B Siepmann J I Oh K J and Klein M L
Simulating vaporndashliquid nucleation of n-alkanes J Chem
Physics 2002 116 4317
86 Harris S A and Ford I J A dynamical definition of
quasi-bound molecular clusters J Chem Physics 2003
118 9216
87 Shreve A P Walton J R P B and Gubbins K E
Liquid drops of polar molecules J Chem Physics 1986
85 2178
88 ten Wolde P R Ruiz-Montero M J and Frenkel D
Numerical calculation of the rate of homogeneous gasndash
liquid nucleation in a LennardndashJones system J Chem
Physics 1998 109 9901
89 ten Wolde P R Oxtoby D W and Frenkel D Chain
formation in homogeneous gasndashliquid nucleation of polar
fluids J Chem Physics 1998 111 4762
90 Vehkamaki H and Ford I J Critical cluster size and
droplet nucleation rate from growth and decay simula-
tions of LennardndashJones clusters J Chem Physics 2000
112 4193
91 Schaaf P Senger B Voegel J C Bowles R K and
Reiss H Simulative determination of kinetic coefficients
for nucleation rates J Chem Physics 2001 114 8091
92 Chen B Siepmann J I Oh K J and Klein M L
Aggregation-volume-bias Monte Carlo simulations of
vaporndashliquid nucleation barriers for LennardndashJonesium
J Chem Physics 2001 115 10 903
93 Hale B N and DiMattio D J A Monte Carlo discrete
sum (MCDS) nucleation rate model for water In
Proceedings of the 15th International Conference on
Nucleation and Atmospheric Aerosols AIP Conference
Proceedings Vol 534 (Eds B N Hale and M Kulmala)
2000 pp 31ndash34
94 Kathmann S M and Hale B N Monte Carlo
simulations of small sulfuric acidndashwater clusters
J Phys Chemistry B 2001 105 11 719
95 Daura X Mark A E and van Gunsteren W F Para-
metrization of aliphatic CHn united atoms of GROMOS96
force field J Comput Chemistry 1998 19 535
96 Kathmann S M Schenter G K and Garrett B C
Understanding the sensitivity of nucleation kinetics a
case study on water J Chem Physics 2002 116 5046
97 Kashchiev D On the relation between nucleation work
nucleus size and nucleation rate J Chem Physics 1982
76 5098
98 Oxtoby D W and Kashchiev D A general relation
between the nucleation work and the size of the nucleus in
multicomponent nucleation J Chem Physics 1994 100
7665
99 Ford I J Thermodynamic properties of critical clusters
from measurements of vapourndashliquid homogeneous
nucleation rates J Chem Physics 1996 105 8324
100 Oxtoby D W and Evans R Nonclassical nucleation
theory for the gasndashliquid transition J Chem Physics
1988 89 7521
101 Zeng X C and Oxtoby D W Gasndashliquid nucleation in
LennardndashJones fluids J Chem Physics 1991 94 4472
102 Evans R Fundamentals of Inhomogeneous Fluids 1992
Ch 3 pp 85ndash175 (Marcel Dekker New York)
103 Talanquer V and Oxtoby D W Dynamical density
functional theory of gasndashliquid nucleation J Chem
Physics 1994 100 5190
104 Barrett J C First-order correction to classical nucleation
theory a density functional approach J Chem Physics
1999 111 5938
105 Barrett J C Cluster translation and growth in density
functional theories of homogeneous nucleation J Chem
Physics 1997 107 7989
106 Bykov T V and Zeng X C A patching model for
surface tension of spherical droplet and Tolman length J
Chem Physics 1999 111 10 602
107 Talanquer V and Oxtoby D W Nucleation in molecular
and dipolar fluids interaction site model J Chem
Physics 1995 103 3686
108 Seok C and Oxtoby D W Nucleation in n-alkanes a
density functional approach J Chem Physics 1998 109
7982
109 Reguera D Bowles R K Djikaev Y and Reiss H
Phase transitions in systems small enough to be clusters
J Chem Physics 2003 118 340
110 Khakshouri S and Ford I J J Chem Physics 2004
(submitted)
111 Ford I J (unpublished)
112 Nyquist R M Talanquer V and Oxtoby D W Density
functional theory of nucleation a semi-empirical
approach J Chem Physics 1995 103 1175
113 Talanquer V and Oxtoby D W Density functional
analysis of phenomenological theories of gasndashliquid
nucleation J Chem Physics 1995 99 2865
114 Katz J L Fisk J A and Chakarov V The accuracy of
nucleation theory In Proceedings of the 13th Interna-
tional Conference on Nucleation and Atmospheric Aerosols
(Eds N Fukuta and P E Wagner) 1992 pp 11ndash18
(Deepak Publishing Hampton Virginia)
115 Fisher M E The theory of condensation and the critical
point Physics 1967 3 255
116 Dillmann A and Meier G E A A refined droplet
approach to the problem of homogeneous nucleation
from the vapor phase J Chem Physics 1991 94 3872
117 Tolman R C J Chem Physics 1949 17 333
118 Ford I J Laaksonen A and Kulmala M Modification
of the DillmannndashMeier theory of homogeneous nuclea-
tion J Chem Physics 1993 99 764
119 Laaksonen A Ford I J and Kulmala M Revised
parametrization of the DillmannndashMeier theory of homo-
geneous nucleation Phys Rev E 1994 49 5517
120 Kalikmanov V I and van Dongen M E H Semipheno-
menological theory of homogeneous vaporndashliquid nuclea-
tion J Chem Physics 1995 103 4250
121 Hale B N Application of a scaled homogeneous
nucleation-rate formalism to experimental-data at
T 5Tc Phys Rev A 1986 33 4156
122 Martinez D M Ferguson F T Heist R H and Nuth
J A Application of scaled nucleation theory to metallic
vapor condensation J Chem Physics 2001 115 310
123 Hale B N and Wilemski G A scaled nucleation model
for ideal binary systems Chem Physics Lett 1999 305
263
124 Granasy L Diffuse interface theory for homogeneous
vapor condensation J Chem Physics 1996 104
5188
125 McGraw R and Laaksonen A Interfacial curvature free
energy the Kelvin relation and vaporndashliquid nucleation
rate J Chem Physics 1997 106 5284
I J FORD898
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
126 Talanquer V A new phenomenological approach to
gasndashliquid nucleation based on the scaling properties
of the critical nucleus J Chem Physics 1997 106
9957
127 Koga K and Zeng X C Thermodynamic expansion of
nucleation free-energy barrier and size of critical nucleus
near the vaporndashliquid coexistence J Chem Physics 1999
110 3466
128 Kusaka I On the scaling behavior of the free energetics of
nucleation J Chem Physics 2003 118 5510
129 Nowakowski B and Ruckenstein E A kinetic approach
to the theory of nucleation in gases J Chem Physics
1991 94 1397
130 Nowakowski B and Ruckenstein E Homogeneous
nucleation in gases a three-dimensional FokkerndashPlanck
equation for evaporation from clusters J Chem Physics
1991 94 8487
131 Wilcox C F and Bauer S H Estimation of homo-
geneous nucleation flux via a kinetic model J Chem
Physics 1991 94 8302
132 Hanggi P Talkner P and Borkovec M Reaction rate
theory fifty years after Kramers Rev Mod Physics 1990
62 251
133 Barrett J C Clement C F and Ford I J Energy
fluctuations in homogeneous nucleation theory for aero-
sols J Physics A 1993 26 529
134 Ford I J and Harris S A Molecular cluster decay
viewed as escape from a potential of mean force J Chem
Physics 2004 120 4428
135 Risken H The FokkerndashPlanck Equation Methods of
Solution and Applications 2nd edition 1989 (Springer-
Verlag Berlin)
136 Shizgal B and Barrett J C Time dependent nucleation
J Chem Physics 1989 91 6505
STATISTICAL MECHANICS OF NUCLEATION 899
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
more straightforward rate prefactor seems to be quitesuccessful in accounting for data for a wide range ofmaterials even metals [122] and two-component systems[123] Another phenomenological model is the diffuseinterface theory [124] which is based on a treatment ofthe phase interface using simplified density profiles ofthermodynamic properties Another general class ofmodel makes use of conditions at the vapourndashliquidspinodal to constrain the form of the model free energy[125ndash128]
The main drawback in phenomenological models isthe arbitrariness in the form of the proposed work offormation The parameter fitting in practice ensures thatthe populations of small clusters especially the mono-mer and dimer are well represented by the model butthe properties of larger clusters depend on the expres-sion assumed for the model and are therefore uncertainIntuition suggests that phenomenological modelling islikely to have mixed success and that microscopiccalculations of cluster properties do matter There mightbe better value in phenomenological models however ifthey were based securely on the results of microscopiccalculations
8 KINETICS OR THERMODYNAMICS
In section 3 kinetic equations representing the nuclea-tion process were described However the coefficients inthose equations were related to constrained equilibriumpopulations and hence to the equilibrium statisticalmechanics of clusters through the detailed balancecondition (8) Is it necessary to do this Could acompletely kinetic route be followed by evaluating themean growth and decay coefficients bi and gi directly bysome statistical approach The main problem is ofcourse the difficulty in performing such calculationsand this has restricted the development of such models[129ndash131] It is expected that the kinetic coefficients willcorrespond to those derived from the detailed balanceargument for near-equilibrium conditions but circum-stances can be imagined where non-equilibrium effectsare present which cannot be dealt with within standardstatistical mechanics
The framework for evaluating kinetic coefficients ofcluster decay can be briefly introduced [86 132] Amolecular dynamics simulation of a cluster can be set upand evolved according to the interactions until bychance an evaporation event occurs By extensivesimulation statistics for the mean evaporation rates gican be amassed in this way Similarly moleculardynamics simulations where a single molecule iscaptured by a cluster could be performed and thegrowth rates bi determined These will be functions ofcluster energy and so would fit best into a rate equationscheme involving both size and energy [133]
This is a laborious approach but statistical physicsoffers an approximate alternative representation of theescape and capture processes The basic tool is theLangevin equation The radial motion of a molecule inthe cluster is described by [134]
meurorr frac14 f ethrTHORN thornfethr _rr tTHORN eth27THORN
where r is the radial position m is the molecular mass fis the mean (time- and velocity-averaged) force on themolecule at position r and f is the stochastic forceThe latter is usually represented as a velocity-dependent friction term mg _rr where g is the so-calledfriction coefficient plus a fluctuating term ~ff ethr tTHORN withzero mean and correlation in time h ~ff ethr tTHORN~ff ethr t0THORNi frac14eth2gkTmTHORNdetht t0THORN where k is Boltzmannrsquos constant andwhere T has the characteristics of a temperature as willbe seen It is a standard manipulation [135] to convertthe Langevin description in the large friction limit intoa FokkerndashPlanck equation
qwqt
frac14 1
mg qethfwTHORN
qrthorn kT
q2wqr2
eth28THORN
which represents the evolution of wethr tTHORN the prob-ability density that the molecule should lie at the radialposition r It should be noted in passing that FokkerndashPlanck equations have often been used to describe thestochastic dynamics of a cluster in size space [48 136] Incontrast the use of the method described above refers tomotion of a molecule in real coordinate space relative tothe centre of mass of the cluster
The steady state equilibrium solution of equation (28)is
wethrTHORN exp FethrTHORNkT
eth29THORN
where FethrTHORN is called the potential of mean force relatedto the mean force on the molecule f throughf frac14 dF=dr It can now be seen how the parameter Tin the stochastic force plays the role of temperaturesince equation (29) looks like a Boltzmann distributionThis solution is not what is sought however The escapeproblem is characterized by a boundary conditionwethreTHORN frac14 0 where re is the radius at which a particleescapes from the system The resulting escape rate maybe shown to be [134]
gkin frac14 KethreT m gTHORN exp DFkT
eth30THORN
where the prefactor K depends on the parameters listedas well as the shape of the potential of mean force Theprincipal feature of equation (30) is the exponentialdependence on the depth DF of the potential of meanforce Therefore if the potential of mean force FethrTHORN can
I J FORD894
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
be evaluated by studying a molecular dynamics trajec-tory for example then the kinetic decay rate can bedetermined This could be extended to account for aradially dependent temperature and friction coefficientboth of which might be determined from simulation Asketch of a potential of mean force is given in Fig 6
Kinetically determined cluster growth and decayrates gkini and bkini respectively could provide a newroute for the determination of nucleation rates inde-pendent of assumptions about detailed balance and theuse of equilibrium statistical mechanics in a non-equilibrium situation
9 CONCLUSIONS
This review has been brief and in some ways partisan tothe personal views of the author It attempts to sketch acoherent picture of nucleation theory and how thevarious disparate strands relate to one another One ofthe fascinations of nucleation theory is the variety ofapproaches and the number of difficult issues that areencountered How well does free energy fluctuationtheory represent a real non-equilibrium process Howwell does the idea of detailed balance provide aframework for determining non-equilibrium decaycoefficients from equilibrium statistical mechanicsHow well do various methods perform in the evaluation
of cluster free energies Can cluster growth and decayrates be calculated directly without the need for such aframework How exactly are clusters to be defined
The theoretical problems are many but the issues canbe resolved to various degrees of satisfaction Thereremains of course the fundamental task of choosing anintermolecular potential that adequately represents thephysical interactions Without this all the finer points ofcluster definition and treatments of non-equilibriumprocesses are unimportant Modelling many problems inscience relies on the availability of an accurate micro-scopic description and so nucleation rate calculationsare not unique in this respect Most empirical molecularpotentials are fitted to the properties of bulk and surfaceproperties and in the very best examples attention isgiven to their transferability to more unusual environ-ments such as clusters In view of these capabilities andmodern theoretical developments in nucleation theoryit seems likely that successful microscopic models of thenucleation process based on statistical physics are atlast coming within reach
ACKNOWLEDGEMENTS
The present author is grateful to colleagues SarahHarris and Marco Arosio for providing some of thefigures
Fig 6 A molecule in a cluster may be considered to be confined within it by a potential of mean force FethrTHORN Itcan escape by reaching a radial position re driven by a phenomenological stochastic force
STATISTICAL MECHANICS OF NUCLEATION 895
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
REFERENCES
1 Gross D H E Microcanonical Thermodynamics 2001
(World Scientific Singapore)
2 Seinfeld J H and Pandis S N Atmospheric Chemistry
and Physics 1998 (John Wiley New York)
3 Climate Change 2001 Synthesis Report 2001 (Inter-
governmental Panel on Climate Change) (Cambridge
University Press Cambridge)
4 Kohler T M Gail H-P and Sedlmayr E MgO dust
nucleation in M-stars calculation of cluster properties
and nucleation rates Astronomy and Astrophysics 1997
320 553
5 Kashchiev D Nucleation Basic Theory with Applications
2000 (ButterworthndashHeinemann Oxford)
6 Bakhtar F Ebrahimi M and Webb R A On the
performance of a cascade of turbine rotor tip section
blading in nucleating steam Part 1 surface pressure
distributions Proc Instn Mech Engrs 1995 209(C2)
115ndash124
7 Bakhtar F Ebrahimi M and Bamkole B O On the
performance of a cascade of turbine rotor tip section
blading in nucleating steam Part 2 wake traverses Proc
Instn Mech Engrs 1995 209(C3) 169ndash178
8 Bakhtar F andMahpeykar M R On the performance of
a cascade of turbine rotor tip section blading in nucleating
steam Part 3 theoretical treatment Proc Instn Mech
Engrs Part C J Mechanical Engineering Science 1997
211(C3) 195ndash210
9 Volmer M and Weber A Z Physikalische Chemie 1926
119 277
10 Farkas L Z Physikalische Chemie (Leipzig) 1927 125
236
11 Becker R and Doring W Kinetische Behandlung der
Keinbildung in ubersattigten Dampfen Ann Physik
(Leipzig) 1935 24 719
12 Zeldovich J Soviet Physics JETP 1942 12 525
13 Frenkel J Kinetic Theory of Liquids 1946 (Oxford
University Press Oxford)
14 Wagner P E and Strey R Homogeneous nucleation
rates of water vapor measured in a two-piston expansion
chamber J Phys Chemistry 1981 85 2694
15 Miller R Anderson R J Kassner J L and Hagen
D E Homogeneous nucleation rate measurements for
water over a wide range of temperature and nucleation
rate J Chem Physics 1983 78 3204
16 Viisanen Y Strey R and Reiss H Homogeneous
nucleation rates for water J Chem Physics 1993 99 4680
17 Viisanen Y Strey R and Reiss H Erratum homo-
geneous nucleation rates for water J Chem Physics
2000 112 8205
18 Wolk J and Strey R Homogeneous nucleation of H2O
and D2O in comparison the isotope effect J Phys
Chemistry B 2001 105 11683
19 Heath C H Streletsky K and Wyslouzil B E H2Ondash
D2O condensation in a supersonic nozzle J Chem
Physics 2002 117 6176
20 Mikheev V B Irving P M Laulainen N S Barlow
S E and Pervukhin V V Laboratory measurements of
water nucleation using a laminar flow chamber J Chem
Physics 2002 116 10 772
21 Hung C H Krasnopoler M J and Katz J L
Condensation of a supersaturated vapor VIII The
homogeneous nucleation of n-nonane J Chem Physics
1989 90 1856
22 Rudek M M Fisk J A Chakarov V M and Katz
J L Condensation of a supersaturated vapor XII The
homogeneous nucleation of the n-alkanes J Chem
Physics 1996 105 4707
23 Strey R Wagner P E and Schmeling T Homogeneous
nucleation rates for n-alcohol vapors measured in a two-
piston expansion chamber J Chem Physics 1986 84
2325
24 Lihavainen H Viisanen Y and Kulmala M Homo-
geneous nucleation of n-pentanol in a laminar flow
diffusion chamber J Chem Physics 2001 114 10 031
25 Rudek M M Katz J L Vidensky I V Zdımal V and
Smolık J Homogeneous nucleation rates of n-pentanol
measured in an upward thermal diffusion cloud chamber
J Chem Physics 1999 111 3623
26 Grassmann A and Peters F Homogeneous nucleation
rates of n-propanol in nitrogen measured in a piston-
expansion tube J Chem Physics 2002 116 7617
27 Kane D and El-Shall S Condensation of supersaturated
vapors of hydrogen bonding molecules ethylene glycol
propylene glycol trimethyl glycol and glycerol J Chem
Physics 1996 105 7617
28 Hameri K and Kulmala M Homogeneous nucleation in
a laminar flow chamber the effect of temperature and
carrier gas pressure on dibutyl phthalate vapor nucleation
rate at high supersaturations J Chem Physics 1996 105
7696
29 Viisanen Y Strey R Laaksonen A and Kulmala M
Measurement of the molecular content of binary nuclei
II Use of the nucleation rate surface for waterndashethanol
J Chem Physics 1994 100 6062
30 Viisanen Y Wagner P E and Strey R Measurement of
the molecular content of binary nuclei IV Use of the
nucleation rate surfaces for the n-nonane-n-alcohol series
J Chem Physics 1998 108 4257
31 Fisk J A Rudek M M Katz J L Beiersdorf D and
Uchtmann H The homogeneous nucleation of cesium
vapor Atmos Res 1998 46 211
32 Uchtmann H Dettmer R Baranovskii S D and
Hensel F Photoinduced nucleation in supersaturated
mercury vapor J Chem Physics 1998 108 9775
33 Ferguson F T and Nuth J A Experimental studies of
the vapor phase nucleation of refractory compounds V
The condensation of lithium J Chem Physics 2000 113
4093
34 Laaksonen A Talanquer V and Oxtoby D W
Nucleation measurements theory and atmospheric
applications A Rev Phys Chemistry 1995 46 489
35 Debenedetti P G Metastable Liquids Concepts and
Principles 1996 (Princeton University Press Princeton
New Jersey)
36 Yasuoka K and Matsumoto M Molecular dynamics of
homogeneous nucleation in the vapor phase I Lennardndash
Jones fluid J Chem Physics 1998 109 8451
37 Yasuoka K and Matsumoto M Molecular dynamics of
homogeneous nucleation in the vapor phase II Water
J Chem Physics 1998 109 8463
I J FORD896
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
38 Wonczak S Strey R and Stauffer D Confirmation of
classical nucleation theory by Monte Carlo simulations in
the 3-dimensional Ising model at low temperature
J Chem Physics 2000 113 1976
39 Laasonen K Wonczak S Strey R and Laaksonen A
Molecular dynamics simulations of gasndashliquid nucleation
of LennardndashJones fluid J Chem Physics 2000 113 9741
40 Reiss H Methods of Thermodynamics 1996 (Dover
Publications New York)
41 Kondepudi D and Prigogine I Modern Thermodynamics
From Heat Engines to Dissipative Structures 1998 (John
Wiley Chichester West Sussex)
42 Smoluchowski M Ann Phys NY 1908 25 225
43 Einstein A Ann Phys NY 1910 338 1275
44 Hirschfelder J O Curtis C F and Bird R BMolecular
Theory of Gases and Liquids 1954 (John Wiley New
York)
45 Landau L D and Lifshitz E M Statistical Physics 1958
(Pergamon London)
46 Wu D T Nucleation theory Solid State Physicsmdash
Advances in Res and Applic 1997 50 37
47 Reguera D Rubı J M and Perez-Madrid A Fokkerndash
Planck equations for nucleation processes revisited
Physica A 1998 259 10
48 Schenter G K Kathmann S and Garrett B Variational
transition state theory of vapor phase nucleation
J Chem Physics 1999 109 7951
49 Onsager L Phys Rev 1931 37 405
50 Courtney W G J Chem Physics 1961 35 2249
51 Girshick S L and Chiu C-P Kinetic nucleation theory
a new expression for the rate of homogeneous nucleation
from an ideal supersaturated vapor J Chem Physics
1990 93 1273
52 Ford I J Barrett J C and Lazaridis M Uncertainties
in cluster energies in homogeneous nucleation theory
J Aerosol Sci 1993 24 581
53 Weakliem C L and Reiss H The factor 1S in the
classical theory of nucleation J Phys Chemistry 1994
98 6408
54 McClurg R B and Flagan R C Critical comparison of
droplet models in homogeneous nucleation theory
J Colloid Interface Sci 1998 201 194
55 Mashmoushy H Mahpeykar M R and Bakhtar F
Studies of nucleating and wet steam flows in two-dimen-
sional cascades Proc Instn Mech Engrs Part C J
Mechanical Engineering Science 2004 218(C8) 843ndash858
56 Ford I J Nucleation theorems the statistical mechanics
of molecular clusters and a revision of classical nuclea-
tion theory Phys Rev E 1997 56 5615
57 ten Wolde P R and Frenkel D Computer simulation
study of gasndashliquid nucleation in a LennardndashJones system
J Chem Physics 1998 109 9901
58 Kusaka I Wang Z G and Seinfeld J H Direct
evaluation of the equilibrium distribution of physical
clusters by a grand canonical Monte Carlo simulation
J Chem Physics 1998 108 3416
59 Kusaka I and Oxtoby D W Identifying physical
clusters in vapor phase nucleation J Chem Physics
1999 110 5249
60 Stillinger F H J Chem Physics 1963 38 1486
61 Ford I J Imperfect vaporndashgas mixtures and homo-
geneous nucleation J Aerosol Sci 1992 23 447
62 Lothe J and Pound G M J Chem Physics 1962 36
2080
63 Lee J K Barker J A and Abraham F F Theory and
Monte Carlo simulation of physical clusters in the
imperfect vapor J Chem Physics 1973 58 3166
64 Abraham F F Homogeneous Nucleation Theory 1974
(Academic Press New York)
65 Reif F Fundamentals of Statistical and Thermal Physics
1965 (McGraw-Hill New York)
66 Hoare M R and Pal P Adv Phys 1975 24 645
67 Hoare M R Adv Chem Physics 1979 40 49
68 Pal P and Hoare M R Thermodynamic properties
and homogeneous nucleation of molecular clusters of
nitrogen J Phys Chemistry 1987 91 2474
69 Reiss H Katz J L and Cohen E R Translationndash
rotation paradox in the theory of nucleation J Chem
Physics 1968 48 5553
70 Hale B N Monte Carlo calculations of effective surface
tension for small clusters Aust J Physics 1996 49 425
71 Kathmann S M Schenter G K and Garrett B C
Dynamical nucleation theory calculation of condensation
rate constants for small water clusters J Chem Physics
1999 111 4688
72 Reiss H Tabazadeh A and Talbot J Molecular theory
of vapor phase nucleation the physically consistent
cluster J Chem Physics 1990 92 1266
73 Ellerby H M Weakliem C L and Reiss H Toward a
molecular theory of vaporndashphase nucleation I Identifica-
tion of the average embryo J Chem Physics 1992 97
5766
74 Ellerby H M and Reiss H Toward a molecular theory
of vaporndashphase nucleation II Fundamental treatment of
the cluster distribution J Chem Physics 1992 97 5766
75 Weakliem C L and Reiss H Toward a molecular theory
of vaporndashphase nucleation III Thermodynamic properties
of argon clusters fromMonte Carlo simulations and modi-
fied liquid drop theory J Chem Physics 1994 101 2398
76 Kusaka I Wang Z G and Seinfeld J H Binary
nucleation of sulfuric acidndashwater Monte Carlo simula-
tion J Chem Physics 1998 108 6829
77 Oh K J and Zeng X C A small system ensemble
Monte Carlo simulation of supersaturated vapor
evaluation of barrier to nucleation J Chem Physics
2000 112 294
78 Senger B Schaaf P Corti D S Bowles R Voegel J
C and Reiss H A molecular theory of the homogeneous
nucleation rate I Formulation and fundamental issues
J Chem Physics 1999 110 6421
79 Hill T L Statistical Mechanics Principles and Selected
Applications 1956 (McGraw-Hill New York)
80 Soto R and Cordero P Nonideal gas of clusters at
equilibrium J Chem Physics 1999 110 7316
81 Barrett J C The significance of cluster lifetime in
nucleation theory J Chem Physics 2002 116 8856
82 Lushnikov A A and Kulmala M Dimers in nucleating
vapors Phys Rev E 1998 58 3157
83 Bahadur R and McClurg R B Comparison of cluster
definitions in homogeneous nucleation rate calculations
J Phys Chemistry 2001 105 11 893
84 Pugnaloni L A and Vericat F New criteria for cluster
identification in continuum systems J Chem Physics
2002 116 1097
STATISTICAL MECHANICS OF NUCLEATION 897
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
85 Chen B Siepmann J I Oh K J and Klein M L
Simulating vaporndashliquid nucleation of n-alkanes J Chem
Physics 2002 116 4317
86 Harris S A and Ford I J A dynamical definition of
quasi-bound molecular clusters J Chem Physics 2003
118 9216
87 Shreve A P Walton J R P B and Gubbins K E
Liquid drops of polar molecules J Chem Physics 1986
85 2178
88 ten Wolde P R Ruiz-Montero M J and Frenkel D
Numerical calculation of the rate of homogeneous gasndash
liquid nucleation in a LennardndashJones system J Chem
Physics 1998 109 9901
89 ten Wolde P R Oxtoby D W and Frenkel D Chain
formation in homogeneous gasndashliquid nucleation of polar
fluids J Chem Physics 1998 111 4762
90 Vehkamaki H and Ford I J Critical cluster size and
droplet nucleation rate from growth and decay simula-
tions of LennardndashJones clusters J Chem Physics 2000
112 4193
91 Schaaf P Senger B Voegel J C Bowles R K and
Reiss H Simulative determination of kinetic coefficients
for nucleation rates J Chem Physics 2001 114 8091
92 Chen B Siepmann J I Oh K J and Klein M L
Aggregation-volume-bias Monte Carlo simulations of
vaporndashliquid nucleation barriers for LennardndashJonesium
J Chem Physics 2001 115 10 903
93 Hale B N and DiMattio D J A Monte Carlo discrete
sum (MCDS) nucleation rate model for water In
Proceedings of the 15th International Conference on
Nucleation and Atmospheric Aerosols AIP Conference
Proceedings Vol 534 (Eds B N Hale and M Kulmala)
2000 pp 31ndash34
94 Kathmann S M and Hale B N Monte Carlo
simulations of small sulfuric acidndashwater clusters
J Phys Chemistry B 2001 105 11 719
95 Daura X Mark A E and van Gunsteren W F Para-
metrization of aliphatic CHn united atoms of GROMOS96
force field J Comput Chemistry 1998 19 535
96 Kathmann S M Schenter G K and Garrett B C
Understanding the sensitivity of nucleation kinetics a
case study on water J Chem Physics 2002 116 5046
97 Kashchiev D On the relation between nucleation work
nucleus size and nucleation rate J Chem Physics 1982
76 5098
98 Oxtoby D W and Kashchiev D A general relation
between the nucleation work and the size of the nucleus in
multicomponent nucleation J Chem Physics 1994 100
7665
99 Ford I J Thermodynamic properties of critical clusters
from measurements of vapourndashliquid homogeneous
nucleation rates J Chem Physics 1996 105 8324
100 Oxtoby D W and Evans R Nonclassical nucleation
theory for the gasndashliquid transition J Chem Physics
1988 89 7521
101 Zeng X C and Oxtoby D W Gasndashliquid nucleation in
LennardndashJones fluids J Chem Physics 1991 94 4472
102 Evans R Fundamentals of Inhomogeneous Fluids 1992
Ch 3 pp 85ndash175 (Marcel Dekker New York)
103 Talanquer V and Oxtoby D W Dynamical density
functional theory of gasndashliquid nucleation J Chem
Physics 1994 100 5190
104 Barrett J C First-order correction to classical nucleation
theory a density functional approach J Chem Physics
1999 111 5938
105 Barrett J C Cluster translation and growth in density
functional theories of homogeneous nucleation J Chem
Physics 1997 107 7989
106 Bykov T V and Zeng X C A patching model for
surface tension of spherical droplet and Tolman length J
Chem Physics 1999 111 10 602
107 Talanquer V and Oxtoby D W Nucleation in molecular
and dipolar fluids interaction site model J Chem
Physics 1995 103 3686
108 Seok C and Oxtoby D W Nucleation in n-alkanes a
density functional approach J Chem Physics 1998 109
7982
109 Reguera D Bowles R K Djikaev Y and Reiss H
Phase transitions in systems small enough to be clusters
J Chem Physics 2003 118 340
110 Khakshouri S and Ford I J J Chem Physics 2004
(submitted)
111 Ford I J (unpublished)
112 Nyquist R M Talanquer V and Oxtoby D W Density
functional theory of nucleation a semi-empirical
approach J Chem Physics 1995 103 1175
113 Talanquer V and Oxtoby D W Density functional
analysis of phenomenological theories of gasndashliquid
nucleation J Chem Physics 1995 99 2865
114 Katz J L Fisk J A and Chakarov V The accuracy of
nucleation theory In Proceedings of the 13th Interna-
tional Conference on Nucleation and Atmospheric Aerosols
(Eds N Fukuta and P E Wagner) 1992 pp 11ndash18
(Deepak Publishing Hampton Virginia)
115 Fisher M E The theory of condensation and the critical
point Physics 1967 3 255
116 Dillmann A and Meier G E A A refined droplet
approach to the problem of homogeneous nucleation
from the vapor phase J Chem Physics 1991 94 3872
117 Tolman R C J Chem Physics 1949 17 333
118 Ford I J Laaksonen A and Kulmala M Modification
of the DillmannndashMeier theory of homogeneous nuclea-
tion J Chem Physics 1993 99 764
119 Laaksonen A Ford I J and Kulmala M Revised
parametrization of the DillmannndashMeier theory of homo-
geneous nucleation Phys Rev E 1994 49 5517
120 Kalikmanov V I and van Dongen M E H Semipheno-
menological theory of homogeneous vaporndashliquid nuclea-
tion J Chem Physics 1995 103 4250
121 Hale B N Application of a scaled homogeneous
nucleation-rate formalism to experimental-data at
T 5Tc Phys Rev A 1986 33 4156
122 Martinez D M Ferguson F T Heist R H and Nuth
J A Application of scaled nucleation theory to metallic
vapor condensation J Chem Physics 2001 115 310
123 Hale B N and Wilemski G A scaled nucleation model
for ideal binary systems Chem Physics Lett 1999 305
263
124 Granasy L Diffuse interface theory for homogeneous
vapor condensation J Chem Physics 1996 104
5188
125 McGraw R and Laaksonen A Interfacial curvature free
energy the Kelvin relation and vaporndashliquid nucleation
rate J Chem Physics 1997 106 5284
I J FORD898
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
126 Talanquer V A new phenomenological approach to
gasndashliquid nucleation based on the scaling properties
of the critical nucleus J Chem Physics 1997 106
9957
127 Koga K and Zeng X C Thermodynamic expansion of
nucleation free-energy barrier and size of critical nucleus
near the vaporndashliquid coexistence J Chem Physics 1999
110 3466
128 Kusaka I On the scaling behavior of the free energetics of
nucleation J Chem Physics 2003 118 5510
129 Nowakowski B and Ruckenstein E A kinetic approach
to the theory of nucleation in gases J Chem Physics
1991 94 1397
130 Nowakowski B and Ruckenstein E Homogeneous
nucleation in gases a three-dimensional FokkerndashPlanck
equation for evaporation from clusters J Chem Physics
1991 94 8487
131 Wilcox C F and Bauer S H Estimation of homo-
geneous nucleation flux via a kinetic model J Chem
Physics 1991 94 8302
132 Hanggi P Talkner P and Borkovec M Reaction rate
theory fifty years after Kramers Rev Mod Physics 1990
62 251
133 Barrett J C Clement C F and Ford I J Energy
fluctuations in homogeneous nucleation theory for aero-
sols J Physics A 1993 26 529
134 Ford I J and Harris S A Molecular cluster decay
viewed as escape from a potential of mean force J Chem
Physics 2004 120 4428
135 Risken H The FokkerndashPlanck Equation Methods of
Solution and Applications 2nd edition 1989 (Springer-
Verlag Berlin)
136 Shizgal B and Barrett J C Time dependent nucleation
J Chem Physics 1989 91 6505
STATISTICAL MECHANICS OF NUCLEATION 899
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
be evaluated by studying a molecular dynamics trajec-tory for example then the kinetic decay rate can bedetermined This could be extended to account for aradially dependent temperature and friction coefficientboth of which might be determined from simulation Asketch of a potential of mean force is given in Fig 6
Kinetically determined cluster growth and decayrates gkini and bkini respectively could provide a newroute for the determination of nucleation rates inde-pendent of assumptions about detailed balance and theuse of equilibrium statistical mechanics in a non-equilibrium situation
9 CONCLUSIONS
This review has been brief and in some ways partisan tothe personal views of the author It attempts to sketch acoherent picture of nucleation theory and how thevarious disparate strands relate to one another One ofthe fascinations of nucleation theory is the variety ofapproaches and the number of difficult issues that areencountered How well does free energy fluctuationtheory represent a real non-equilibrium process Howwell does the idea of detailed balance provide aframework for determining non-equilibrium decaycoefficients from equilibrium statistical mechanicsHow well do various methods perform in the evaluation
of cluster free energies Can cluster growth and decayrates be calculated directly without the need for such aframework How exactly are clusters to be defined
The theoretical problems are many but the issues canbe resolved to various degrees of satisfaction Thereremains of course the fundamental task of choosing anintermolecular potential that adequately represents thephysical interactions Without this all the finer points ofcluster definition and treatments of non-equilibriumprocesses are unimportant Modelling many problems inscience relies on the availability of an accurate micro-scopic description and so nucleation rate calculationsare not unique in this respect Most empirical molecularpotentials are fitted to the properties of bulk and surfaceproperties and in the very best examples attention isgiven to their transferability to more unusual environ-ments such as clusters In view of these capabilities andmodern theoretical developments in nucleation theoryit seems likely that successful microscopic models of thenucleation process based on statistical physics are atlast coming within reach
ACKNOWLEDGEMENTS
The present author is grateful to colleagues SarahHarris and Marco Arosio for providing some of thefigures
Fig 6 A molecule in a cluster may be considered to be confined within it by a potential of mean force FethrTHORN Itcan escape by reaching a radial position re driven by a phenomenological stochastic force
STATISTICAL MECHANICS OF NUCLEATION 895
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
REFERENCES
1 Gross D H E Microcanonical Thermodynamics 2001
(World Scientific Singapore)
2 Seinfeld J H and Pandis S N Atmospheric Chemistry
and Physics 1998 (John Wiley New York)
3 Climate Change 2001 Synthesis Report 2001 (Inter-
governmental Panel on Climate Change) (Cambridge
University Press Cambridge)
4 Kohler T M Gail H-P and Sedlmayr E MgO dust
nucleation in M-stars calculation of cluster properties
and nucleation rates Astronomy and Astrophysics 1997
320 553
5 Kashchiev D Nucleation Basic Theory with Applications
2000 (ButterworthndashHeinemann Oxford)
6 Bakhtar F Ebrahimi M and Webb R A On the
performance of a cascade of turbine rotor tip section
blading in nucleating steam Part 1 surface pressure
distributions Proc Instn Mech Engrs 1995 209(C2)
115ndash124
7 Bakhtar F Ebrahimi M and Bamkole B O On the
performance of a cascade of turbine rotor tip section
blading in nucleating steam Part 2 wake traverses Proc
Instn Mech Engrs 1995 209(C3) 169ndash178
8 Bakhtar F andMahpeykar M R On the performance of
a cascade of turbine rotor tip section blading in nucleating
steam Part 3 theoretical treatment Proc Instn Mech
Engrs Part C J Mechanical Engineering Science 1997
211(C3) 195ndash210
9 Volmer M and Weber A Z Physikalische Chemie 1926
119 277
10 Farkas L Z Physikalische Chemie (Leipzig) 1927 125
236
11 Becker R and Doring W Kinetische Behandlung der
Keinbildung in ubersattigten Dampfen Ann Physik
(Leipzig) 1935 24 719
12 Zeldovich J Soviet Physics JETP 1942 12 525
13 Frenkel J Kinetic Theory of Liquids 1946 (Oxford
University Press Oxford)
14 Wagner P E and Strey R Homogeneous nucleation
rates of water vapor measured in a two-piston expansion
chamber J Phys Chemistry 1981 85 2694
15 Miller R Anderson R J Kassner J L and Hagen
D E Homogeneous nucleation rate measurements for
water over a wide range of temperature and nucleation
rate J Chem Physics 1983 78 3204
16 Viisanen Y Strey R and Reiss H Homogeneous
nucleation rates for water J Chem Physics 1993 99 4680
17 Viisanen Y Strey R and Reiss H Erratum homo-
geneous nucleation rates for water J Chem Physics
2000 112 8205
18 Wolk J and Strey R Homogeneous nucleation of H2O
and D2O in comparison the isotope effect J Phys
Chemistry B 2001 105 11683
19 Heath C H Streletsky K and Wyslouzil B E H2Ondash
D2O condensation in a supersonic nozzle J Chem
Physics 2002 117 6176
20 Mikheev V B Irving P M Laulainen N S Barlow
S E and Pervukhin V V Laboratory measurements of
water nucleation using a laminar flow chamber J Chem
Physics 2002 116 10 772
21 Hung C H Krasnopoler M J and Katz J L
Condensation of a supersaturated vapor VIII The
homogeneous nucleation of n-nonane J Chem Physics
1989 90 1856
22 Rudek M M Fisk J A Chakarov V M and Katz
J L Condensation of a supersaturated vapor XII The
homogeneous nucleation of the n-alkanes J Chem
Physics 1996 105 4707
23 Strey R Wagner P E and Schmeling T Homogeneous
nucleation rates for n-alcohol vapors measured in a two-
piston expansion chamber J Chem Physics 1986 84
2325
24 Lihavainen H Viisanen Y and Kulmala M Homo-
geneous nucleation of n-pentanol in a laminar flow
diffusion chamber J Chem Physics 2001 114 10 031
25 Rudek M M Katz J L Vidensky I V Zdımal V and
Smolık J Homogeneous nucleation rates of n-pentanol
measured in an upward thermal diffusion cloud chamber
J Chem Physics 1999 111 3623
26 Grassmann A and Peters F Homogeneous nucleation
rates of n-propanol in nitrogen measured in a piston-
expansion tube J Chem Physics 2002 116 7617
27 Kane D and El-Shall S Condensation of supersaturated
vapors of hydrogen bonding molecules ethylene glycol
propylene glycol trimethyl glycol and glycerol J Chem
Physics 1996 105 7617
28 Hameri K and Kulmala M Homogeneous nucleation in
a laminar flow chamber the effect of temperature and
carrier gas pressure on dibutyl phthalate vapor nucleation
rate at high supersaturations J Chem Physics 1996 105
7696
29 Viisanen Y Strey R Laaksonen A and Kulmala M
Measurement of the molecular content of binary nuclei
II Use of the nucleation rate surface for waterndashethanol
J Chem Physics 1994 100 6062
30 Viisanen Y Wagner P E and Strey R Measurement of
the molecular content of binary nuclei IV Use of the
nucleation rate surfaces for the n-nonane-n-alcohol series
J Chem Physics 1998 108 4257
31 Fisk J A Rudek M M Katz J L Beiersdorf D and
Uchtmann H The homogeneous nucleation of cesium
vapor Atmos Res 1998 46 211
32 Uchtmann H Dettmer R Baranovskii S D and
Hensel F Photoinduced nucleation in supersaturated
mercury vapor J Chem Physics 1998 108 9775
33 Ferguson F T and Nuth J A Experimental studies of
the vapor phase nucleation of refractory compounds V
The condensation of lithium J Chem Physics 2000 113
4093
34 Laaksonen A Talanquer V and Oxtoby D W
Nucleation measurements theory and atmospheric
applications A Rev Phys Chemistry 1995 46 489
35 Debenedetti P G Metastable Liquids Concepts and
Principles 1996 (Princeton University Press Princeton
New Jersey)
36 Yasuoka K and Matsumoto M Molecular dynamics of
homogeneous nucleation in the vapor phase I Lennardndash
Jones fluid J Chem Physics 1998 109 8451
37 Yasuoka K and Matsumoto M Molecular dynamics of
homogeneous nucleation in the vapor phase II Water
J Chem Physics 1998 109 8463
I J FORD896
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
38 Wonczak S Strey R and Stauffer D Confirmation of
classical nucleation theory by Monte Carlo simulations in
the 3-dimensional Ising model at low temperature
J Chem Physics 2000 113 1976
39 Laasonen K Wonczak S Strey R and Laaksonen A
Molecular dynamics simulations of gasndashliquid nucleation
of LennardndashJones fluid J Chem Physics 2000 113 9741
40 Reiss H Methods of Thermodynamics 1996 (Dover
Publications New York)
41 Kondepudi D and Prigogine I Modern Thermodynamics
From Heat Engines to Dissipative Structures 1998 (John
Wiley Chichester West Sussex)
42 Smoluchowski M Ann Phys NY 1908 25 225
43 Einstein A Ann Phys NY 1910 338 1275
44 Hirschfelder J O Curtis C F and Bird R BMolecular
Theory of Gases and Liquids 1954 (John Wiley New
York)
45 Landau L D and Lifshitz E M Statistical Physics 1958
(Pergamon London)
46 Wu D T Nucleation theory Solid State Physicsmdash
Advances in Res and Applic 1997 50 37
47 Reguera D Rubı J M and Perez-Madrid A Fokkerndash
Planck equations for nucleation processes revisited
Physica A 1998 259 10
48 Schenter G K Kathmann S and Garrett B Variational
transition state theory of vapor phase nucleation
J Chem Physics 1999 109 7951
49 Onsager L Phys Rev 1931 37 405
50 Courtney W G J Chem Physics 1961 35 2249
51 Girshick S L and Chiu C-P Kinetic nucleation theory
a new expression for the rate of homogeneous nucleation
from an ideal supersaturated vapor J Chem Physics
1990 93 1273
52 Ford I J Barrett J C and Lazaridis M Uncertainties
in cluster energies in homogeneous nucleation theory
J Aerosol Sci 1993 24 581
53 Weakliem C L and Reiss H The factor 1S in the
classical theory of nucleation J Phys Chemistry 1994
98 6408
54 McClurg R B and Flagan R C Critical comparison of
droplet models in homogeneous nucleation theory
J Colloid Interface Sci 1998 201 194
55 Mashmoushy H Mahpeykar M R and Bakhtar F
Studies of nucleating and wet steam flows in two-dimen-
sional cascades Proc Instn Mech Engrs Part C J
Mechanical Engineering Science 2004 218(C8) 843ndash858
56 Ford I J Nucleation theorems the statistical mechanics
of molecular clusters and a revision of classical nuclea-
tion theory Phys Rev E 1997 56 5615
57 ten Wolde P R and Frenkel D Computer simulation
study of gasndashliquid nucleation in a LennardndashJones system
J Chem Physics 1998 109 9901
58 Kusaka I Wang Z G and Seinfeld J H Direct
evaluation of the equilibrium distribution of physical
clusters by a grand canonical Monte Carlo simulation
J Chem Physics 1998 108 3416
59 Kusaka I and Oxtoby D W Identifying physical
clusters in vapor phase nucleation J Chem Physics
1999 110 5249
60 Stillinger F H J Chem Physics 1963 38 1486
61 Ford I J Imperfect vaporndashgas mixtures and homo-
geneous nucleation J Aerosol Sci 1992 23 447
62 Lothe J and Pound G M J Chem Physics 1962 36
2080
63 Lee J K Barker J A and Abraham F F Theory and
Monte Carlo simulation of physical clusters in the
imperfect vapor J Chem Physics 1973 58 3166
64 Abraham F F Homogeneous Nucleation Theory 1974
(Academic Press New York)
65 Reif F Fundamentals of Statistical and Thermal Physics
1965 (McGraw-Hill New York)
66 Hoare M R and Pal P Adv Phys 1975 24 645
67 Hoare M R Adv Chem Physics 1979 40 49
68 Pal P and Hoare M R Thermodynamic properties
and homogeneous nucleation of molecular clusters of
nitrogen J Phys Chemistry 1987 91 2474
69 Reiss H Katz J L and Cohen E R Translationndash
rotation paradox in the theory of nucleation J Chem
Physics 1968 48 5553
70 Hale B N Monte Carlo calculations of effective surface
tension for small clusters Aust J Physics 1996 49 425
71 Kathmann S M Schenter G K and Garrett B C
Dynamical nucleation theory calculation of condensation
rate constants for small water clusters J Chem Physics
1999 111 4688
72 Reiss H Tabazadeh A and Talbot J Molecular theory
of vapor phase nucleation the physically consistent
cluster J Chem Physics 1990 92 1266
73 Ellerby H M Weakliem C L and Reiss H Toward a
molecular theory of vaporndashphase nucleation I Identifica-
tion of the average embryo J Chem Physics 1992 97
5766
74 Ellerby H M and Reiss H Toward a molecular theory
of vaporndashphase nucleation II Fundamental treatment of
the cluster distribution J Chem Physics 1992 97 5766
75 Weakliem C L and Reiss H Toward a molecular theory
of vaporndashphase nucleation III Thermodynamic properties
of argon clusters fromMonte Carlo simulations and modi-
fied liquid drop theory J Chem Physics 1994 101 2398
76 Kusaka I Wang Z G and Seinfeld J H Binary
nucleation of sulfuric acidndashwater Monte Carlo simula-
tion J Chem Physics 1998 108 6829
77 Oh K J and Zeng X C A small system ensemble
Monte Carlo simulation of supersaturated vapor
evaluation of barrier to nucleation J Chem Physics
2000 112 294
78 Senger B Schaaf P Corti D S Bowles R Voegel J
C and Reiss H A molecular theory of the homogeneous
nucleation rate I Formulation and fundamental issues
J Chem Physics 1999 110 6421
79 Hill T L Statistical Mechanics Principles and Selected
Applications 1956 (McGraw-Hill New York)
80 Soto R and Cordero P Nonideal gas of clusters at
equilibrium J Chem Physics 1999 110 7316
81 Barrett J C The significance of cluster lifetime in
nucleation theory J Chem Physics 2002 116 8856
82 Lushnikov A A and Kulmala M Dimers in nucleating
vapors Phys Rev E 1998 58 3157
83 Bahadur R and McClurg R B Comparison of cluster
definitions in homogeneous nucleation rate calculations
J Phys Chemistry 2001 105 11 893
84 Pugnaloni L A and Vericat F New criteria for cluster
identification in continuum systems J Chem Physics
2002 116 1097
STATISTICAL MECHANICS OF NUCLEATION 897
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
85 Chen B Siepmann J I Oh K J and Klein M L
Simulating vaporndashliquid nucleation of n-alkanes J Chem
Physics 2002 116 4317
86 Harris S A and Ford I J A dynamical definition of
quasi-bound molecular clusters J Chem Physics 2003
118 9216
87 Shreve A P Walton J R P B and Gubbins K E
Liquid drops of polar molecules J Chem Physics 1986
85 2178
88 ten Wolde P R Ruiz-Montero M J and Frenkel D
Numerical calculation of the rate of homogeneous gasndash
liquid nucleation in a LennardndashJones system J Chem
Physics 1998 109 9901
89 ten Wolde P R Oxtoby D W and Frenkel D Chain
formation in homogeneous gasndashliquid nucleation of polar
fluids J Chem Physics 1998 111 4762
90 Vehkamaki H and Ford I J Critical cluster size and
droplet nucleation rate from growth and decay simula-
tions of LennardndashJones clusters J Chem Physics 2000
112 4193
91 Schaaf P Senger B Voegel J C Bowles R K and
Reiss H Simulative determination of kinetic coefficients
for nucleation rates J Chem Physics 2001 114 8091
92 Chen B Siepmann J I Oh K J and Klein M L
Aggregation-volume-bias Monte Carlo simulations of
vaporndashliquid nucleation barriers for LennardndashJonesium
J Chem Physics 2001 115 10 903
93 Hale B N and DiMattio D J A Monte Carlo discrete
sum (MCDS) nucleation rate model for water In
Proceedings of the 15th International Conference on
Nucleation and Atmospheric Aerosols AIP Conference
Proceedings Vol 534 (Eds B N Hale and M Kulmala)
2000 pp 31ndash34
94 Kathmann S M and Hale B N Monte Carlo
simulations of small sulfuric acidndashwater clusters
J Phys Chemistry B 2001 105 11 719
95 Daura X Mark A E and van Gunsteren W F Para-
metrization of aliphatic CHn united atoms of GROMOS96
force field J Comput Chemistry 1998 19 535
96 Kathmann S M Schenter G K and Garrett B C
Understanding the sensitivity of nucleation kinetics a
case study on water J Chem Physics 2002 116 5046
97 Kashchiev D On the relation between nucleation work
nucleus size and nucleation rate J Chem Physics 1982
76 5098
98 Oxtoby D W and Kashchiev D A general relation
between the nucleation work and the size of the nucleus in
multicomponent nucleation J Chem Physics 1994 100
7665
99 Ford I J Thermodynamic properties of critical clusters
from measurements of vapourndashliquid homogeneous
nucleation rates J Chem Physics 1996 105 8324
100 Oxtoby D W and Evans R Nonclassical nucleation
theory for the gasndashliquid transition J Chem Physics
1988 89 7521
101 Zeng X C and Oxtoby D W Gasndashliquid nucleation in
LennardndashJones fluids J Chem Physics 1991 94 4472
102 Evans R Fundamentals of Inhomogeneous Fluids 1992
Ch 3 pp 85ndash175 (Marcel Dekker New York)
103 Talanquer V and Oxtoby D W Dynamical density
functional theory of gasndashliquid nucleation J Chem
Physics 1994 100 5190
104 Barrett J C First-order correction to classical nucleation
theory a density functional approach J Chem Physics
1999 111 5938
105 Barrett J C Cluster translation and growth in density
functional theories of homogeneous nucleation J Chem
Physics 1997 107 7989
106 Bykov T V and Zeng X C A patching model for
surface tension of spherical droplet and Tolman length J
Chem Physics 1999 111 10 602
107 Talanquer V and Oxtoby D W Nucleation in molecular
and dipolar fluids interaction site model J Chem
Physics 1995 103 3686
108 Seok C and Oxtoby D W Nucleation in n-alkanes a
density functional approach J Chem Physics 1998 109
7982
109 Reguera D Bowles R K Djikaev Y and Reiss H
Phase transitions in systems small enough to be clusters
J Chem Physics 2003 118 340
110 Khakshouri S and Ford I J J Chem Physics 2004
(submitted)
111 Ford I J (unpublished)
112 Nyquist R M Talanquer V and Oxtoby D W Density
functional theory of nucleation a semi-empirical
approach J Chem Physics 1995 103 1175
113 Talanquer V and Oxtoby D W Density functional
analysis of phenomenological theories of gasndashliquid
nucleation J Chem Physics 1995 99 2865
114 Katz J L Fisk J A and Chakarov V The accuracy of
nucleation theory In Proceedings of the 13th Interna-
tional Conference on Nucleation and Atmospheric Aerosols
(Eds N Fukuta and P E Wagner) 1992 pp 11ndash18
(Deepak Publishing Hampton Virginia)
115 Fisher M E The theory of condensation and the critical
point Physics 1967 3 255
116 Dillmann A and Meier G E A A refined droplet
approach to the problem of homogeneous nucleation
from the vapor phase J Chem Physics 1991 94 3872
117 Tolman R C J Chem Physics 1949 17 333
118 Ford I J Laaksonen A and Kulmala M Modification
of the DillmannndashMeier theory of homogeneous nuclea-
tion J Chem Physics 1993 99 764
119 Laaksonen A Ford I J and Kulmala M Revised
parametrization of the DillmannndashMeier theory of homo-
geneous nucleation Phys Rev E 1994 49 5517
120 Kalikmanov V I and van Dongen M E H Semipheno-
menological theory of homogeneous vaporndashliquid nuclea-
tion J Chem Physics 1995 103 4250
121 Hale B N Application of a scaled homogeneous
nucleation-rate formalism to experimental-data at
T 5Tc Phys Rev A 1986 33 4156
122 Martinez D M Ferguson F T Heist R H and Nuth
J A Application of scaled nucleation theory to metallic
vapor condensation J Chem Physics 2001 115 310
123 Hale B N and Wilemski G A scaled nucleation model
for ideal binary systems Chem Physics Lett 1999 305
263
124 Granasy L Diffuse interface theory for homogeneous
vapor condensation J Chem Physics 1996 104
5188
125 McGraw R and Laaksonen A Interfacial curvature free
energy the Kelvin relation and vaporndashliquid nucleation
rate J Chem Physics 1997 106 5284
I J FORD898
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
126 Talanquer V A new phenomenological approach to
gasndashliquid nucleation based on the scaling properties
of the critical nucleus J Chem Physics 1997 106
9957
127 Koga K and Zeng X C Thermodynamic expansion of
nucleation free-energy barrier and size of critical nucleus
near the vaporndashliquid coexistence J Chem Physics 1999
110 3466
128 Kusaka I On the scaling behavior of the free energetics of
nucleation J Chem Physics 2003 118 5510
129 Nowakowski B and Ruckenstein E A kinetic approach
to the theory of nucleation in gases J Chem Physics
1991 94 1397
130 Nowakowski B and Ruckenstein E Homogeneous
nucleation in gases a three-dimensional FokkerndashPlanck
equation for evaporation from clusters J Chem Physics
1991 94 8487
131 Wilcox C F and Bauer S H Estimation of homo-
geneous nucleation flux via a kinetic model J Chem
Physics 1991 94 8302
132 Hanggi P Talkner P and Borkovec M Reaction rate
theory fifty years after Kramers Rev Mod Physics 1990
62 251
133 Barrett J C Clement C F and Ford I J Energy
fluctuations in homogeneous nucleation theory for aero-
sols J Physics A 1993 26 529
134 Ford I J and Harris S A Molecular cluster decay
viewed as escape from a potential of mean force J Chem
Physics 2004 120 4428
135 Risken H The FokkerndashPlanck Equation Methods of
Solution and Applications 2nd edition 1989 (Springer-
Verlag Berlin)
136 Shizgal B and Barrett J C Time dependent nucleation
J Chem Physics 1989 91 6505
STATISTICAL MECHANICS OF NUCLEATION 899
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
REFERENCES
1 Gross D H E Microcanonical Thermodynamics 2001
(World Scientific Singapore)
2 Seinfeld J H and Pandis S N Atmospheric Chemistry
and Physics 1998 (John Wiley New York)
3 Climate Change 2001 Synthesis Report 2001 (Inter-
governmental Panel on Climate Change) (Cambridge
University Press Cambridge)
4 Kohler T M Gail H-P and Sedlmayr E MgO dust
nucleation in M-stars calculation of cluster properties
and nucleation rates Astronomy and Astrophysics 1997
320 553
5 Kashchiev D Nucleation Basic Theory with Applications
2000 (ButterworthndashHeinemann Oxford)
6 Bakhtar F Ebrahimi M and Webb R A On the
performance of a cascade of turbine rotor tip section
blading in nucleating steam Part 1 surface pressure
distributions Proc Instn Mech Engrs 1995 209(C2)
115ndash124
7 Bakhtar F Ebrahimi M and Bamkole B O On the
performance of a cascade of turbine rotor tip section
blading in nucleating steam Part 2 wake traverses Proc
Instn Mech Engrs 1995 209(C3) 169ndash178
8 Bakhtar F andMahpeykar M R On the performance of
a cascade of turbine rotor tip section blading in nucleating
steam Part 3 theoretical treatment Proc Instn Mech
Engrs Part C J Mechanical Engineering Science 1997
211(C3) 195ndash210
9 Volmer M and Weber A Z Physikalische Chemie 1926
119 277
10 Farkas L Z Physikalische Chemie (Leipzig) 1927 125
236
11 Becker R and Doring W Kinetische Behandlung der
Keinbildung in ubersattigten Dampfen Ann Physik
(Leipzig) 1935 24 719
12 Zeldovich J Soviet Physics JETP 1942 12 525
13 Frenkel J Kinetic Theory of Liquids 1946 (Oxford
University Press Oxford)
14 Wagner P E and Strey R Homogeneous nucleation
rates of water vapor measured in a two-piston expansion
chamber J Phys Chemistry 1981 85 2694
15 Miller R Anderson R J Kassner J L and Hagen
D E Homogeneous nucleation rate measurements for
water over a wide range of temperature and nucleation
rate J Chem Physics 1983 78 3204
16 Viisanen Y Strey R and Reiss H Homogeneous
nucleation rates for water J Chem Physics 1993 99 4680
17 Viisanen Y Strey R and Reiss H Erratum homo-
geneous nucleation rates for water J Chem Physics
2000 112 8205
18 Wolk J and Strey R Homogeneous nucleation of H2O
and D2O in comparison the isotope effect J Phys
Chemistry B 2001 105 11683
19 Heath C H Streletsky K and Wyslouzil B E H2Ondash
D2O condensation in a supersonic nozzle J Chem
Physics 2002 117 6176
20 Mikheev V B Irving P M Laulainen N S Barlow
S E and Pervukhin V V Laboratory measurements of
water nucleation using a laminar flow chamber J Chem
Physics 2002 116 10 772
21 Hung C H Krasnopoler M J and Katz J L
Condensation of a supersaturated vapor VIII The
homogeneous nucleation of n-nonane J Chem Physics
1989 90 1856
22 Rudek M M Fisk J A Chakarov V M and Katz
J L Condensation of a supersaturated vapor XII The
homogeneous nucleation of the n-alkanes J Chem
Physics 1996 105 4707
23 Strey R Wagner P E and Schmeling T Homogeneous
nucleation rates for n-alcohol vapors measured in a two-
piston expansion chamber J Chem Physics 1986 84
2325
24 Lihavainen H Viisanen Y and Kulmala M Homo-
geneous nucleation of n-pentanol in a laminar flow
diffusion chamber J Chem Physics 2001 114 10 031
25 Rudek M M Katz J L Vidensky I V Zdımal V and
Smolık J Homogeneous nucleation rates of n-pentanol
measured in an upward thermal diffusion cloud chamber
J Chem Physics 1999 111 3623
26 Grassmann A and Peters F Homogeneous nucleation
rates of n-propanol in nitrogen measured in a piston-
expansion tube J Chem Physics 2002 116 7617
27 Kane D and El-Shall S Condensation of supersaturated
vapors of hydrogen bonding molecules ethylene glycol
propylene glycol trimethyl glycol and glycerol J Chem
Physics 1996 105 7617
28 Hameri K and Kulmala M Homogeneous nucleation in
a laminar flow chamber the effect of temperature and
carrier gas pressure on dibutyl phthalate vapor nucleation
rate at high supersaturations J Chem Physics 1996 105
7696
29 Viisanen Y Strey R Laaksonen A and Kulmala M
Measurement of the molecular content of binary nuclei
II Use of the nucleation rate surface for waterndashethanol
J Chem Physics 1994 100 6062
30 Viisanen Y Wagner P E and Strey R Measurement of
the molecular content of binary nuclei IV Use of the
nucleation rate surfaces for the n-nonane-n-alcohol series
J Chem Physics 1998 108 4257
31 Fisk J A Rudek M M Katz J L Beiersdorf D and
Uchtmann H The homogeneous nucleation of cesium
vapor Atmos Res 1998 46 211
32 Uchtmann H Dettmer R Baranovskii S D and
Hensel F Photoinduced nucleation in supersaturated
mercury vapor J Chem Physics 1998 108 9775
33 Ferguson F T and Nuth J A Experimental studies of
the vapor phase nucleation of refractory compounds V
The condensation of lithium J Chem Physics 2000 113
4093
34 Laaksonen A Talanquer V and Oxtoby D W
Nucleation measurements theory and atmospheric
applications A Rev Phys Chemistry 1995 46 489
35 Debenedetti P G Metastable Liquids Concepts and
Principles 1996 (Princeton University Press Princeton
New Jersey)
36 Yasuoka K and Matsumoto M Molecular dynamics of
homogeneous nucleation in the vapor phase I Lennardndash
Jones fluid J Chem Physics 1998 109 8451
37 Yasuoka K and Matsumoto M Molecular dynamics of
homogeneous nucleation in the vapor phase II Water
J Chem Physics 1998 109 8463
I J FORD896
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
38 Wonczak S Strey R and Stauffer D Confirmation of
classical nucleation theory by Monte Carlo simulations in
the 3-dimensional Ising model at low temperature
J Chem Physics 2000 113 1976
39 Laasonen K Wonczak S Strey R and Laaksonen A
Molecular dynamics simulations of gasndashliquid nucleation
of LennardndashJones fluid J Chem Physics 2000 113 9741
40 Reiss H Methods of Thermodynamics 1996 (Dover
Publications New York)
41 Kondepudi D and Prigogine I Modern Thermodynamics
From Heat Engines to Dissipative Structures 1998 (John
Wiley Chichester West Sussex)
42 Smoluchowski M Ann Phys NY 1908 25 225
43 Einstein A Ann Phys NY 1910 338 1275
44 Hirschfelder J O Curtis C F and Bird R BMolecular
Theory of Gases and Liquids 1954 (John Wiley New
York)
45 Landau L D and Lifshitz E M Statistical Physics 1958
(Pergamon London)
46 Wu D T Nucleation theory Solid State Physicsmdash
Advances in Res and Applic 1997 50 37
47 Reguera D Rubı J M and Perez-Madrid A Fokkerndash
Planck equations for nucleation processes revisited
Physica A 1998 259 10
48 Schenter G K Kathmann S and Garrett B Variational
transition state theory of vapor phase nucleation
J Chem Physics 1999 109 7951
49 Onsager L Phys Rev 1931 37 405
50 Courtney W G J Chem Physics 1961 35 2249
51 Girshick S L and Chiu C-P Kinetic nucleation theory
a new expression for the rate of homogeneous nucleation
from an ideal supersaturated vapor J Chem Physics
1990 93 1273
52 Ford I J Barrett J C and Lazaridis M Uncertainties
in cluster energies in homogeneous nucleation theory
J Aerosol Sci 1993 24 581
53 Weakliem C L and Reiss H The factor 1S in the
classical theory of nucleation J Phys Chemistry 1994
98 6408
54 McClurg R B and Flagan R C Critical comparison of
droplet models in homogeneous nucleation theory
J Colloid Interface Sci 1998 201 194
55 Mashmoushy H Mahpeykar M R and Bakhtar F
Studies of nucleating and wet steam flows in two-dimen-
sional cascades Proc Instn Mech Engrs Part C J
Mechanical Engineering Science 2004 218(C8) 843ndash858
56 Ford I J Nucleation theorems the statistical mechanics
of molecular clusters and a revision of classical nuclea-
tion theory Phys Rev E 1997 56 5615
57 ten Wolde P R and Frenkel D Computer simulation
study of gasndashliquid nucleation in a LennardndashJones system
J Chem Physics 1998 109 9901
58 Kusaka I Wang Z G and Seinfeld J H Direct
evaluation of the equilibrium distribution of physical
clusters by a grand canonical Monte Carlo simulation
J Chem Physics 1998 108 3416
59 Kusaka I and Oxtoby D W Identifying physical
clusters in vapor phase nucleation J Chem Physics
1999 110 5249
60 Stillinger F H J Chem Physics 1963 38 1486
61 Ford I J Imperfect vaporndashgas mixtures and homo-
geneous nucleation J Aerosol Sci 1992 23 447
62 Lothe J and Pound G M J Chem Physics 1962 36
2080
63 Lee J K Barker J A and Abraham F F Theory and
Monte Carlo simulation of physical clusters in the
imperfect vapor J Chem Physics 1973 58 3166
64 Abraham F F Homogeneous Nucleation Theory 1974
(Academic Press New York)
65 Reif F Fundamentals of Statistical and Thermal Physics
1965 (McGraw-Hill New York)
66 Hoare M R and Pal P Adv Phys 1975 24 645
67 Hoare M R Adv Chem Physics 1979 40 49
68 Pal P and Hoare M R Thermodynamic properties
and homogeneous nucleation of molecular clusters of
nitrogen J Phys Chemistry 1987 91 2474
69 Reiss H Katz J L and Cohen E R Translationndash
rotation paradox in the theory of nucleation J Chem
Physics 1968 48 5553
70 Hale B N Monte Carlo calculations of effective surface
tension for small clusters Aust J Physics 1996 49 425
71 Kathmann S M Schenter G K and Garrett B C
Dynamical nucleation theory calculation of condensation
rate constants for small water clusters J Chem Physics
1999 111 4688
72 Reiss H Tabazadeh A and Talbot J Molecular theory
of vapor phase nucleation the physically consistent
cluster J Chem Physics 1990 92 1266
73 Ellerby H M Weakliem C L and Reiss H Toward a
molecular theory of vaporndashphase nucleation I Identifica-
tion of the average embryo J Chem Physics 1992 97
5766
74 Ellerby H M and Reiss H Toward a molecular theory
of vaporndashphase nucleation II Fundamental treatment of
the cluster distribution J Chem Physics 1992 97 5766
75 Weakliem C L and Reiss H Toward a molecular theory
of vaporndashphase nucleation III Thermodynamic properties
of argon clusters fromMonte Carlo simulations and modi-
fied liquid drop theory J Chem Physics 1994 101 2398
76 Kusaka I Wang Z G and Seinfeld J H Binary
nucleation of sulfuric acidndashwater Monte Carlo simula-
tion J Chem Physics 1998 108 6829
77 Oh K J and Zeng X C A small system ensemble
Monte Carlo simulation of supersaturated vapor
evaluation of barrier to nucleation J Chem Physics
2000 112 294
78 Senger B Schaaf P Corti D S Bowles R Voegel J
C and Reiss H A molecular theory of the homogeneous
nucleation rate I Formulation and fundamental issues
J Chem Physics 1999 110 6421
79 Hill T L Statistical Mechanics Principles and Selected
Applications 1956 (McGraw-Hill New York)
80 Soto R and Cordero P Nonideal gas of clusters at
equilibrium J Chem Physics 1999 110 7316
81 Barrett J C The significance of cluster lifetime in
nucleation theory J Chem Physics 2002 116 8856
82 Lushnikov A A and Kulmala M Dimers in nucleating
vapors Phys Rev E 1998 58 3157
83 Bahadur R and McClurg R B Comparison of cluster
definitions in homogeneous nucleation rate calculations
J Phys Chemistry 2001 105 11 893
84 Pugnaloni L A and Vericat F New criteria for cluster
identification in continuum systems J Chem Physics
2002 116 1097
STATISTICAL MECHANICS OF NUCLEATION 897
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
85 Chen B Siepmann J I Oh K J and Klein M L
Simulating vaporndashliquid nucleation of n-alkanes J Chem
Physics 2002 116 4317
86 Harris S A and Ford I J A dynamical definition of
quasi-bound molecular clusters J Chem Physics 2003
118 9216
87 Shreve A P Walton J R P B and Gubbins K E
Liquid drops of polar molecules J Chem Physics 1986
85 2178
88 ten Wolde P R Ruiz-Montero M J and Frenkel D
Numerical calculation of the rate of homogeneous gasndash
liquid nucleation in a LennardndashJones system J Chem
Physics 1998 109 9901
89 ten Wolde P R Oxtoby D W and Frenkel D Chain
formation in homogeneous gasndashliquid nucleation of polar
fluids J Chem Physics 1998 111 4762
90 Vehkamaki H and Ford I J Critical cluster size and
droplet nucleation rate from growth and decay simula-
tions of LennardndashJones clusters J Chem Physics 2000
112 4193
91 Schaaf P Senger B Voegel J C Bowles R K and
Reiss H Simulative determination of kinetic coefficients
for nucleation rates J Chem Physics 2001 114 8091
92 Chen B Siepmann J I Oh K J and Klein M L
Aggregation-volume-bias Monte Carlo simulations of
vaporndashliquid nucleation barriers for LennardndashJonesium
J Chem Physics 2001 115 10 903
93 Hale B N and DiMattio D J A Monte Carlo discrete
sum (MCDS) nucleation rate model for water In
Proceedings of the 15th International Conference on
Nucleation and Atmospheric Aerosols AIP Conference
Proceedings Vol 534 (Eds B N Hale and M Kulmala)
2000 pp 31ndash34
94 Kathmann S M and Hale B N Monte Carlo
simulations of small sulfuric acidndashwater clusters
J Phys Chemistry B 2001 105 11 719
95 Daura X Mark A E and van Gunsteren W F Para-
metrization of aliphatic CHn united atoms of GROMOS96
force field J Comput Chemistry 1998 19 535
96 Kathmann S M Schenter G K and Garrett B C
Understanding the sensitivity of nucleation kinetics a
case study on water J Chem Physics 2002 116 5046
97 Kashchiev D On the relation between nucleation work
nucleus size and nucleation rate J Chem Physics 1982
76 5098
98 Oxtoby D W and Kashchiev D A general relation
between the nucleation work and the size of the nucleus in
multicomponent nucleation J Chem Physics 1994 100
7665
99 Ford I J Thermodynamic properties of critical clusters
from measurements of vapourndashliquid homogeneous
nucleation rates J Chem Physics 1996 105 8324
100 Oxtoby D W and Evans R Nonclassical nucleation
theory for the gasndashliquid transition J Chem Physics
1988 89 7521
101 Zeng X C and Oxtoby D W Gasndashliquid nucleation in
LennardndashJones fluids J Chem Physics 1991 94 4472
102 Evans R Fundamentals of Inhomogeneous Fluids 1992
Ch 3 pp 85ndash175 (Marcel Dekker New York)
103 Talanquer V and Oxtoby D W Dynamical density
functional theory of gasndashliquid nucleation J Chem
Physics 1994 100 5190
104 Barrett J C First-order correction to classical nucleation
theory a density functional approach J Chem Physics
1999 111 5938
105 Barrett J C Cluster translation and growth in density
functional theories of homogeneous nucleation J Chem
Physics 1997 107 7989
106 Bykov T V and Zeng X C A patching model for
surface tension of spherical droplet and Tolman length J
Chem Physics 1999 111 10 602
107 Talanquer V and Oxtoby D W Nucleation in molecular
and dipolar fluids interaction site model J Chem
Physics 1995 103 3686
108 Seok C and Oxtoby D W Nucleation in n-alkanes a
density functional approach J Chem Physics 1998 109
7982
109 Reguera D Bowles R K Djikaev Y and Reiss H
Phase transitions in systems small enough to be clusters
J Chem Physics 2003 118 340
110 Khakshouri S and Ford I J J Chem Physics 2004
(submitted)
111 Ford I J (unpublished)
112 Nyquist R M Talanquer V and Oxtoby D W Density
functional theory of nucleation a semi-empirical
approach J Chem Physics 1995 103 1175
113 Talanquer V and Oxtoby D W Density functional
analysis of phenomenological theories of gasndashliquid
nucleation J Chem Physics 1995 99 2865
114 Katz J L Fisk J A and Chakarov V The accuracy of
nucleation theory In Proceedings of the 13th Interna-
tional Conference on Nucleation and Atmospheric Aerosols
(Eds N Fukuta and P E Wagner) 1992 pp 11ndash18
(Deepak Publishing Hampton Virginia)
115 Fisher M E The theory of condensation and the critical
point Physics 1967 3 255
116 Dillmann A and Meier G E A A refined droplet
approach to the problem of homogeneous nucleation
from the vapor phase J Chem Physics 1991 94 3872
117 Tolman R C J Chem Physics 1949 17 333
118 Ford I J Laaksonen A and Kulmala M Modification
of the DillmannndashMeier theory of homogeneous nuclea-
tion J Chem Physics 1993 99 764
119 Laaksonen A Ford I J and Kulmala M Revised
parametrization of the DillmannndashMeier theory of homo-
geneous nucleation Phys Rev E 1994 49 5517
120 Kalikmanov V I and van Dongen M E H Semipheno-
menological theory of homogeneous vaporndashliquid nuclea-
tion J Chem Physics 1995 103 4250
121 Hale B N Application of a scaled homogeneous
nucleation-rate formalism to experimental-data at
T 5Tc Phys Rev A 1986 33 4156
122 Martinez D M Ferguson F T Heist R H and Nuth
J A Application of scaled nucleation theory to metallic
vapor condensation J Chem Physics 2001 115 310
123 Hale B N and Wilemski G A scaled nucleation model
for ideal binary systems Chem Physics Lett 1999 305
263
124 Granasy L Diffuse interface theory for homogeneous
vapor condensation J Chem Physics 1996 104
5188
125 McGraw R and Laaksonen A Interfacial curvature free
energy the Kelvin relation and vaporndashliquid nucleation
rate J Chem Physics 1997 106 5284
I J FORD898
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
126 Talanquer V A new phenomenological approach to
gasndashliquid nucleation based on the scaling properties
of the critical nucleus J Chem Physics 1997 106
9957
127 Koga K and Zeng X C Thermodynamic expansion of
nucleation free-energy barrier and size of critical nucleus
near the vaporndashliquid coexistence J Chem Physics 1999
110 3466
128 Kusaka I On the scaling behavior of the free energetics of
nucleation J Chem Physics 2003 118 5510
129 Nowakowski B and Ruckenstein E A kinetic approach
to the theory of nucleation in gases J Chem Physics
1991 94 1397
130 Nowakowski B and Ruckenstein E Homogeneous
nucleation in gases a three-dimensional FokkerndashPlanck
equation for evaporation from clusters J Chem Physics
1991 94 8487
131 Wilcox C F and Bauer S H Estimation of homo-
geneous nucleation flux via a kinetic model J Chem
Physics 1991 94 8302
132 Hanggi P Talkner P and Borkovec M Reaction rate
theory fifty years after Kramers Rev Mod Physics 1990
62 251
133 Barrett J C Clement C F and Ford I J Energy
fluctuations in homogeneous nucleation theory for aero-
sols J Physics A 1993 26 529
134 Ford I J and Harris S A Molecular cluster decay
viewed as escape from a potential of mean force J Chem
Physics 2004 120 4428
135 Risken H The FokkerndashPlanck Equation Methods of
Solution and Applications 2nd edition 1989 (Springer-
Verlag Berlin)
136 Shizgal B and Barrett J C Time dependent nucleation
J Chem Physics 1989 91 6505
STATISTICAL MECHANICS OF NUCLEATION 899
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
38 Wonczak S Strey R and Stauffer D Confirmation of
classical nucleation theory by Monte Carlo simulations in
the 3-dimensional Ising model at low temperature
J Chem Physics 2000 113 1976
39 Laasonen K Wonczak S Strey R and Laaksonen A
Molecular dynamics simulations of gasndashliquid nucleation
of LennardndashJones fluid J Chem Physics 2000 113 9741
40 Reiss H Methods of Thermodynamics 1996 (Dover
Publications New York)
41 Kondepudi D and Prigogine I Modern Thermodynamics
From Heat Engines to Dissipative Structures 1998 (John
Wiley Chichester West Sussex)
42 Smoluchowski M Ann Phys NY 1908 25 225
43 Einstein A Ann Phys NY 1910 338 1275
44 Hirschfelder J O Curtis C F and Bird R BMolecular
Theory of Gases and Liquids 1954 (John Wiley New
York)
45 Landau L D and Lifshitz E M Statistical Physics 1958
(Pergamon London)
46 Wu D T Nucleation theory Solid State Physicsmdash
Advances in Res and Applic 1997 50 37
47 Reguera D Rubı J M and Perez-Madrid A Fokkerndash
Planck equations for nucleation processes revisited
Physica A 1998 259 10
48 Schenter G K Kathmann S and Garrett B Variational
transition state theory of vapor phase nucleation
J Chem Physics 1999 109 7951
49 Onsager L Phys Rev 1931 37 405
50 Courtney W G J Chem Physics 1961 35 2249
51 Girshick S L and Chiu C-P Kinetic nucleation theory
a new expression for the rate of homogeneous nucleation
from an ideal supersaturated vapor J Chem Physics
1990 93 1273
52 Ford I J Barrett J C and Lazaridis M Uncertainties
in cluster energies in homogeneous nucleation theory
J Aerosol Sci 1993 24 581
53 Weakliem C L and Reiss H The factor 1S in the
classical theory of nucleation J Phys Chemistry 1994
98 6408
54 McClurg R B and Flagan R C Critical comparison of
droplet models in homogeneous nucleation theory
J Colloid Interface Sci 1998 201 194
55 Mashmoushy H Mahpeykar M R and Bakhtar F
Studies of nucleating and wet steam flows in two-dimen-
sional cascades Proc Instn Mech Engrs Part C J
Mechanical Engineering Science 2004 218(C8) 843ndash858
56 Ford I J Nucleation theorems the statistical mechanics
of molecular clusters and a revision of classical nuclea-
tion theory Phys Rev E 1997 56 5615
57 ten Wolde P R and Frenkel D Computer simulation
study of gasndashliquid nucleation in a LennardndashJones system
J Chem Physics 1998 109 9901
58 Kusaka I Wang Z G and Seinfeld J H Direct
evaluation of the equilibrium distribution of physical
clusters by a grand canonical Monte Carlo simulation
J Chem Physics 1998 108 3416
59 Kusaka I and Oxtoby D W Identifying physical
clusters in vapor phase nucleation J Chem Physics
1999 110 5249
60 Stillinger F H J Chem Physics 1963 38 1486
61 Ford I J Imperfect vaporndashgas mixtures and homo-
geneous nucleation J Aerosol Sci 1992 23 447
62 Lothe J and Pound G M J Chem Physics 1962 36
2080
63 Lee J K Barker J A and Abraham F F Theory and
Monte Carlo simulation of physical clusters in the
imperfect vapor J Chem Physics 1973 58 3166
64 Abraham F F Homogeneous Nucleation Theory 1974
(Academic Press New York)
65 Reif F Fundamentals of Statistical and Thermal Physics
1965 (McGraw-Hill New York)
66 Hoare M R and Pal P Adv Phys 1975 24 645
67 Hoare M R Adv Chem Physics 1979 40 49
68 Pal P and Hoare M R Thermodynamic properties
and homogeneous nucleation of molecular clusters of
nitrogen J Phys Chemistry 1987 91 2474
69 Reiss H Katz J L and Cohen E R Translationndash
rotation paradox in the theory of nucleation J Chem
Physics 1968 48 5553
70 Hale B N Monte Carlo calculations of effective surface
tension for small clusters Aust J Physics 1996 49 425
71 Kathmann S M Schenter G K and Garrett B C
Dynamical nucleation theory calculation of condensation
rate constants for small water clusters J Chem Physics
1999 111 4688
72 Reiss H Tabazadeh A and Talbot J Molecular theory
of vapor phase nucleation the physically consistent
cluster J Chem Physics 1990 92 1266
73 Ellerby H M Weakliem C L and Reiss H Toward a
molecular theory of vaporndashphase nucleation I Identifica-
tion of the average embryo J Chem Physics 1992 97
5766
74 Ellerby H M and Reiss H Toward a molecular theory
of vaporndashphase nucleation II Fundamental treatment of
the cluster distribution J Chem Physics 1992 97 5766
75 Weakliem C L and Reiss H Toward a molecular theory
of vaporndashphase nucleation III Thermodynamic properties
of argon clusters fromMonte Carlo simulations and modi-
fied liquid drop theory J Chem Physics 1994 101 2398
76 Kusaka I Wang Z G and Seinfeld J H Binary
nucleation of sulfuric acidndashwater Monte Carlo simula-
tion J Chem Physics 1998 108 6829
77 Oh K J and Zeng X C A small system ensemble
Monte Carlo simulation of supersaturated vapor
evaluation of barrier to nucleation J Chem Physics
2000 112 294
78 Senger B Schaaf P Corti D S Bowles R Voegel J
C and Reiss H A molecular theory of the homogeneous
nucleation rate I Formulation and fundamental issues
J Chem Physics 1999 110 6421
79 Hill T L Statistical Mechanics Principles and Selected
Applications 1956 (McGraw-Hill New York)
80 Soto R and Cordero P Nonideal gas of clusters at
equilibrium J Chem Physics 1999 110 7316
81 Barrett J C The significance of cluster lifetime in
nucleation theory J Chem Physics 2002 116 8856
82 Lushnikov A A and Kulmala M Dimers in nucleating
vapors Phys Rev E 1998 58 3157
83 Bahadur R and McClurg R B Comparison of cluster
definitions in homogeneous nucleation rate calculations
J Phys Chemistry 2001 105 11 893
84 Pugnaloni L A and Vericat F New criteria for cluster
identification in continuum systems J Chem Physics
2002 116 1097
STATISTICAL MECHANICS OF NUCLEATION 897
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
85 Chen B Siepmann J I Oh K J and Klein M L
Simulating vaporndashliquid nucleation of n-alkanes J Chem
Physics 2002 116 4317
86 Harris S A and Ford I J A dynamical definition of
quasi-bound molecular clusters J Chem Physics 2003
118 9216
87 Shreve A P Walton J R P B and Gubbins K E
Liquid drops of polar molecules J Chem Physics 1986
85 2178
88 ten Wolde P R Ruiz-Montero M J and Frenkel D
Numerical calculation of the rate of homogeneous gasndash
liquid nucleation in a LennardndashJones system J Chem
Physics 1998 109 9901
89 ten Wolde P R Oxtoby D W and Frenkel D Chain
formation in homogeneous gasndashliquid nucleation of polar
fluids J Chem Physics 1998 111 4762
90 Vehkamaki H and Ford I J Critical cluster size and
droplet nucleation rate from growth and decay simula-
tions of LennardndashJones clusters J Chem Physics 2000
112 4193
91 Schaaf P Senger B Voegel J C Bowles R K and
Reiss H Simulative determination of kinetic coefficients
for nucleation rates J Chem Physics 2001 114 8091
92 Chen B Siepmann J I Oh K J and Klein M L
Aggregation-volume-bias Monte Carlo simulations of
vaporndashliquid nucleation barriers for LennardndashJonesium
J Chem Physics 2001 115 10 903
93 Hale B N and DiMattio D J A Monte Carlo discrete
sum (MCDS) nucleation rate model for water In
Proceedings of the 15th International Conference on
Nucleation and Atmospheric Aerosols AIP Conference
Proceedings Vol 534 (Eds B N Hale and M Kulmala)
2000 pp 31ndash34
94 Kathmann S M and Hale B N Monte Carlo
simulations of small sulfuric acidndashwater clusters
J Phys Chemistry B 2001 105 11 719
95 Daura X Mark A E and van Gunsteren W F Para-
metrization of aliphatic CHn united atoms of GROMOS96
force field J Comput Chemistry 1998 19 535
96 Kathmann S M Schenter G K and Garrett B C
Understanding the sensitivity of nucleation kinetics a
case study on water J Chem Physics 2002 116 5046
97 Kashchiev D On the relation between nucleation work
nucleus size and nucleation rate J Chem Physics 1982
76 5098
98 Oxtoby D W and Kashchiev D A general relation
between the nucleation work and the size of the nucleus in
multicomponent nucleation J Chem Physics 1994 100
7665
99 Ford I J Thermodynamic properties of critical clusters
from measurements of vapourndashliquid homogeneous
nucleation rates J Chem Physics 1996 105 8324
100 Oxtoby D W and Evans R Nonclassical nucleation
theory for the gasndashliquid transition J Chem Physics
1988 89 7521
101 Zeng X C and Oxtoby D W Gasndashliquid nucleation in
LennardndashJones fluids J Chem Physics 1991 94 4472
102 Evans R Fundamentals of Inhomogeneous Fluids 1992
Ch 3 pp 85ndash175 (Marcel Dekker New York)
103 Talanquer V and Oxtoby D W Dynamical density
functional theory of gasndashliquid nucleation J Chem
Physics 1994 100 5190
104 Barrett J C First-order correction to classical nucleation
theory a density functional approach J Chem Physics
1999 111 5938
105 Barrett J C Cluster translation and growth in density
functional theories of homogeneous nucleation J Chem
Physics 1997 107 7989
106 Bykov T V and Zeng X C A patching model for
surface tension of spherical droplet and Tolman length J
Chem Physics 1999 111 10 602
107 Talanquer V and Oxtoby D W Nucleation in molecular
and dipolar fluids interaction site model J Chem
Physics 1995 103 3686
108 Seok C and Oxtoby D W Nucleation in n-alkanes a
density functional approach J Chem Physics 1998 109
7982
109 Reguera D Bowles R K Djikaev Y and Reiss H
Phase transitions in systems small enough to be clusters
J Chem Physics 2003 118 340
110 Khakshouri S and Ford I J J Chem Physics 2004
(submitted)
111 Ford I J (unpublished)
112 Nyquist R M Talanquer V and Oxtoby D W Density
functional theory of nucleation a semi-empirical
approach J Chem Physics 1995 103 1175
113 Talanquer V and Oxtoby D W Density functional
analysis of phenomenological theories of gasndashliquid
nucleation J Chem Physics 1995 99 2865
114 Katz J L Fisk J A and Chakarov V The accuracy of
nucleation theory In Proceedings of the 13th Interna-
tional Conference on Nucleation and Atmospheric Aerosols
(Eds N Fukuta and P E Wagner) 1992 pp 11ndash18
(Deepak Publishing Hampton Virginia)
115 Fisher M E The theory of condensation and the critical
point Physics 1967 3 255
116 Dillmann A and Meier G E A A refined droplet
approach to the problem of homogeneous nucleation
from the vapor phase J Chem Physics 1991 94 3872
117 Tolman R C J Chem Physics 1949 17 333
118 Ford I J Laaksonen A and Kulmala M Modification
of the DillmannndashMeier theory of homogeneous nuclea-
tion J Chem Physics 1993 99 764
119 Laaksonen A Ford I J and Kulmala M Revised
parametrization of the DillmannndashMeier theory of homo-
geneous nucleation Phys Rev E 1994 49 5517
120 Kalikmanov V I and van Dongen M E H Semipheno-
menological theory of homogeneous vaporndashliquid nuclea-
tion J Chem Physics 1995 103 4250
121 Hale B N Application of a scaled homogeneous
nucleation-rate formalism to experimental-data at
T 5Tc Phys Rev A 1986 33 4156
122 Martinez D M Ferguson F T Heist R H and Nuth
J A Application of scaled nucleation theory to metallic
vapor condensation J Chem Physics 2001 115 310
123 Hale B N and Wilemski G A scaled nucleation model
for ideal binary systems Chem Physics Lett 1999 305
263
124 Granasy L Diffuse interface theory for homogeneous
vapor condensation J Chem Physics 1996 104
5188
125 McGraw R and Laaksonen A Interfacial curvature free
energy the Kelvin relation and vaporndashliquid nucleation
rate J Chem Physics 1997 106 5284
I J FORD898
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
126 Talanquer V A new phenomenological approach to
gasndashliquid nucleation based on the scaling properties
of the critical nucleus J Chem Physics 1997 106
9957
127 Koga K and Zeng X C Thermodynamic expansion of
nucleation free-energy barrier and size of critical nucleus
near the vaporndashliquid coexistence J Chem Physics 1999
110 3466
128 Kusaka I On the scaling behavior of the free energetics of
nucleation J Chem Physics 2003 118 5510
129 Nowakowski B and Ruckenstein E A kinetic approach
to the theory of nucleation in gases J Chem Physics
1991 94 1397
130 Nowakowski B and Ruckenstein E Homogeneous
nucleation in gases a three-dimensional FokkerndashPlanck
equation for evaporation from clusters J Chem Physics
1991 94 8487
131 Wilcox C F and Bauer S H Estimation of homo-
geneous nucleation flux via a kinetic model J Chem
Physics 1991 94 8302
132 Hanggi P Talkner P and Borkovec M Reaction rate
theory fifty years after Kramers Rev Mod Physics 1990
62 251
133 Barrett J C Clement C F and Ford I J Energy
fluctuations in homogeneous nucleation theory for aero-
sols J Physics A 1993 26 529
134 Ford I J and Harris S A Molecular cluster decay
viewed as escape from a potential of mean force J Chem
Physics 2004 120 4428
135 Risken H The FokkerndashPlanck Equation Methods of
Solution and Applications 2nd edition 1989 (Springer-
Verlag Berlin)
136 Shizgal B and Barrett J C Time dependent nucleation
J Chem Physics 1989 91 6505
STATISTICAL MECHANICS OF NUCLEATION 899
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
85 Chen B Siepmann J I Oh K J and Klein M L
Simulating vaporndashliquid nucleation of n-alkanes J Chem
Physics 2002 116 4317
86 Harris S A and Ford I J A dynamical definition of
quasi-bound molecular clusters J Chem Physics 2003
118 9216
87 Shreve A P Walton J R P B and Gubbins K E
Liquid drops of polar molecules J Chem Physics 1986
85 2178
88 ten Wolde P R Ruiz-Montero M J and Frenkel D
Numerical calculation of the rate of homogeneous gasndash
liquid nucleation in a LennardndashJones system J Chem
Physics 1998 109 9901
89 ten Wolde P R Oxtoby D W and Frenkel D Chain
formation in homogeneous gasndashliquid nucleation of polar
fluids J Chem Physics 1998 111 4762
90 Vehkamaki H and Ford I J Critical cluster size and
droplet nucleation rate from growth and decay simula-
tions of LennardndashJones clusters J Chem Physics 2000
112 4193
91 Schaaf P Senger B Voegel J C Bowles R K and
Reiss H Simulative determination of kinetic coefficients
for nucleation rates J Chem Physics 2001 114 8091
92 Chen B Siepmann J I Oh K J and Klein M L
Aggregation-volume-bias Monte Carlo simulations of
vaporndashliquid nucleation barriers for LennardndashJonesium
J Chem Physics 2001 115 10 903
93 Hale B N and DiMattio D J A Monte Carlo discrete
sum (MCDS) nucleation rate model for water In
Proceedings of the 15th International Conference on
Nucleation and Atmospheric Aerosols AIP Conference
Proceedings Vol 534 (Eds B N Hale and M Kulmala)
2000 pp 31ndash34
94 Kathmann S M and Hale B N Monte Carlo
simulations of small sulfuric acidndashwater clusters
J Phys Chemistry B 2001 105 11 719
95 Daura X Mark A E and van Gunsteren W F Para-
metrization of aliphatic CHn united atoms of GROMOS96
force field J Comput Chemistry 1998 19 535
96 Kathmann S M Schenter G K and Garrett B C
Understanding the sensitivity of nucleation kinetics a
case study on water J Chem Physics 2002 116 5046
97 Kashchiev D On the relation between nucleation work
nucleus size and nucleation rate J Chem Physics 1982
76 5098
98 Oxtoby D W and Kashchiev D A general relation
between the nucleation work and the size of the nucleus in
multicomponent nucleation J Chem Physics 1994 100
7665
99 Ford I J Thermodynamic properties of critical clusters
from measurements of vapourndashliquid homogeneous
nucleation rates J Chem Physics 1996 105 8324
100 Oxtoby D W and Evans R Nonclassical nucleation
theory for the gasndashliquid transition J Chem Physics
1988 89 7521
101 Zeng X C and Oxtoby D W Gasndashliquid nucleation in
LennardndashJones fluids J Chem Physics 1991 94 4472
102 Evans R Fundamentals of Inhomogeneous Fluids 1992
Ch 3 pp 85ndash175 (Marcel Dekker New York)
103 Talanquer V and Oxtoby D W Dynamical density
functional theory of gasndashliquid nucleation J Chem
Physics 1994 100 5190
104 Barrett J C First-order correction to classical nucleation
theory a density functional approach J Chem Physics
1999 111 5938
105 Barrett J C Cluster translation and growth in density
functional theories of homogeneous nucleation J Chem
Physics 1997 107 7989
106 Bykov T V and Zeng X C A patching model for
surface tension of spherical droplet and Tolman length J
Chem Physics 1999 111 10 602
107 Talanquer V and Oxtoby D W Nucleation in molecular
and dipolar fluids interaction site model J Chem
Physics 1995 103 3686
108 Seok C and Oxtoby D W Nucleation in n-alkanes a
density functional approach J Chem Physics 1998 109
7982
109 Reguera D Bowles R K Djikaev Y and Reiss H
Phase transitions in systems small enough to be clusters
J Chem Physics 2003 118 340
110 Khakshouri S and Ford I J J Chem Physics 2004
(submitted)
111 Ford I J (unpublished)
112 Nyquist R M Talanquer V and Oxtoby D W Density
functional theory of nucleation a semi-empirical
approach J Chem Physics 1995 103 1175
113 Talanquer V and Oxtoby D W Density functional
analysis of phenomenological theories of gasndashliquid
nucleation J Chem Physics 1995 99 2865
114 Katz J L Fisk J A and Chakarov V The accuracy of
nucleation theory In Proceedings of the 13th Interna-
tional Conference on Nucleation and Atmospheric Aerosols
(Eds N Fukuta and P E Wagner) 1992 pp 11ndash18
(Deepak Publishing Hampton Virginia)
115 Fisher M E The theory of condensation and the critical
point Physics 1967 3 255
116 Dillmann A and Meier G E A A refined droplet
approach to the problem of homogeneous nucleation
from the vapor phase J Chem Physics 1991 94 3872
117 Tolman R C J Chem Physics 1949 17 333
118 Ford I J Laaksonen A and Kulmala M Modification
of the DillmannndashMeier theory of homogeneous nuclea-
tion J Chem Physics 1993 99 764
119 Laaksonen A Ford I J and Kulmala M Revised
parametrization of the DillmannndashMeier theory of homo-
geneous nucleation Phys Rev E 1994 49 5517
120 Kalikmanov V I and van Dongen M E H Semipheno-
menological theory of homogeneous vaporndashliquid nuclea-
tion J Chem Physics 1995 103 4250
121 Hale B N Application of a scaled homogeneous
nucleation-rate formalism to experimental-data at
T 5Tc Phys Rev A 1986 33 4156
122 Martinez D M Ferguson F T Heist R H and Nuth
J A Application of scaled nucleation theory to metallic
vapor condensation J Chem Physics 2001 115 310
123 Hale B N and Wilemski G A scaled nucleation model
for ideal binary systems Chem Physics Lett 1999 305
263
124 Granasy L Diffuse interface theory for homogeneous
vapor condensation J Chem Physics 1996 104
5188
125 McGraw R and Laaksonen A Interfacial curvature free
energy the Kelvin relation and vaporndashliquid nucleation
rate J Chem Physics 1997 106 5284
I J FORD898
Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science C12903 IMechE 2004
126 Talanquer V A new phenomenological approach to
gasndashliquid nucleation based on the scaling properties
of the critical nucleus J Chem Physics 1997 106
9957
127 Koga K and Zeng X C Thermodynamic expansion of
nucleation free-energy barrier and size of critical nucleus
near the vaporndashliquid coexistence J Chem Physics 1999
110 3466
128 Kusaka I On the scaling behavior of the free energetics of
nucleation J Chem Physics 2003 118 5510
129 Nowakowski B and Ruckenstein E A kinetic approach
to the theory of nucleation in gases J Chem Physics
1991 94 1397
130 Nowakowski B and Ruckenstein E Homogeneous
nucleation in gases a three-dimensional FokkerndashPlanck
equation for evaporation from clusters J Chem Physics
1991 94 8487
131 Wilcox C F and Bauer S H Estimation of homo-
geneous nucleation flux via a kinetic model J Chem
Physics 1991 94 8302
132 Hanggi P Talkner P and Borkovec M Reaction rate
theory fifty years after Kramers Rev Mod Physics 1990
62 251
133 Barrett J C Clement C F and Ford I J Energy
fluctuations in homogeneous nucleation theory for aero-
sols J Physics A 1993 26 529
134 Ford I J and Harris S A Molecular cluster decay
viewed as escape from a potential of mean force J Chem
Physics 2004 120 4428
135 Risken H The FokkerndashPlanck Equation Methods of
Solution and Applications 2nd edition 1989 (Springer-
Verlag Berlin)
136 Shizgal B and Barrett J C Time dependent nucleation
J Chem Physics 1989 91 6505
STATISTICAL MECHANICS OF NUCLEATION 899
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science
126 Talanquer V A new phenomenological approach to
gasndashliquid nucleation based on the scaling properties
of the critical nucleus J Chem Physics 1997 106
9957
127 Koga K and Zeng X C Thermodynamic expansion of
nucleation free-energy barrier and size of critical nucleus
near the vaporndashliquid coexistence J Chem Physics 1999
110 3466
128 Kusaka I On the scaling behavior of the free energetics of
nucleation J Chem Physics 2003 118 5510
129 Nowakowski B and Ruckenstein E A kinetic approach
to the theory of nucleation in gases J Chem Physics
1991 94 1397
130 Nowakowski B and Ruckenstein E Homogeneous
nucleation in gases a three-dimensional FokkerndashPlanck
equation for evaporation from clusters J Chem Physics
1991 94 8487
131 Wilcox C F and Bauer S H Estimation of homo-
geneous nucleation flux via a kinetic model J Chem
Physics 1991 94 8302
132 Hanggi P Talkner P and Borkovec M Reaction rate
theory fifty years after Kramers Rev Mod Physics 1990
62 251
133 Barrett J C Clement C F and Ford I J Energy
fluctuations in homogeneous nucleation theory for aero-
sols J Physics A 1993 26 529
134 Ford I J and Harris S A Molecular cluster decay
viewed as escape from a potential of mean force J Chem
Physics 2004 120 4428
135 Risken H The FokkerndashPlanck Equation Methods of
Solution and Applications 2nd edition 1989 (Springer-
Verlag Berlin)
136 Shizgal B and Barrett J C Time dependent nucleation
J Chem Physics 1989 91 6505
STATISTICAL MECHANICS OF NUCLEATION 899
C12903 IMechE 2004 Proc Instn Mech Engrs Vol 218 Part C J Mechanical Engineering Science