SPECIAL ISSUE PAPER 1315

Classical nucleation theory and its application

to condensing steam flow calculations
F Bakhtar1
, J B Young2, A J White2, and D A Simpson2
1
Department of Mechanical and Manufacturing Engineering, University of Birmingham, Birmingham, UK
2
Hopkinson Laboratory, Cambridge University Engineering Department, Cambridge, UK

The manuscript was received on 25 May 2004 and was accepted after revision for publication on 9 August 2004.

DOI: 10.1243/095440605X8379

Abstract: The paper discusses the classical theory of the homogeneous nucleation of water
droplets from supersaturated vapour and its application in predicting condensation in steam
nozzles. The first part consists of a review of classical nucleation theory, focusing on the
many modifications made to the original Becker –Döring theory and providing some new
insights into recent developments. It is concluded that the predictive accuracy required for
engineering calculations is not yet attainable with a theory derived from first principles. The
areas that require most attention relate to the properties of small molecular clusters and the
energy transfer processes in the non-isothermal theory. Experiments in converging – diverging
nozzles provide the best means for validation at the very high nucleation rates of interest, but
measurements of pressure distribution and the Sauter mean droplet radius are insufficient to
provide independent checks on the separate theories of nucleation and droplet growth. Never-
theless, a judicious choice for the nucleation rate equation, in combination with a standard
droplet growth model and a suitable equation of state for steam, can provide accurate predic-
tions over a wide range of conditions. The exception is at very low pressures where there is
evidence that the droplet growth rate in the nucleation zone is underestimated.

Keywords: nucleation theory, steam turbines, two-phase flow, non-equilibrium, wet steam

1 INTRODUCTION distribution of the droplets present and hence on

When the state path of an expanding pure vapour,
freed from foreign particles and ions and removed
from contact with solid surfaces, crosses the satur-
ation line, the vapour does not condense immediately
but instead becomes supersaturated. A super-
saturated vapour is in a metastable state and over
a period of time (which may be short or long depend-
ing on the degree of supersaturation) reversion to an
equilibrium state occurs via the formation and
growth of vast numbers of extremely small liquid
droplets. This process is termed homogeneous
nucleation, in contrast to heterogeneous nucleation,
which refers to condensation on any foreign particles
present. After nucleation, the flow behaviour of
the wet vapour is strongly dependent on the size

[5], Reiss [11], Dunning [12], Ford [13], and many
Corresponding author: Department of Mechanical Engineering
and Manufacturing, University of Birmingham, PO Box 363,
Edgbaston, Birmingham B15 2TT, UK.
the conditions under which nucleation occurred.
Understanding the nucleation process is thus central
to understanding the behaviour of wet steam.
Historically, theories of homogeneous nucleation
have developed along two lines. The classical
thermodynamic– kinetic approach is associated
with such celebrated names as Volmer and Weber
[1], Farkas [2], Becker and Döring [3], Zeldovich [4],
Frenkel [5], and others. A highly recommended
account concentrating on the underlying physics is
given by McDonald [6, 7] and a more rigorous
thermodynamic treatment can be found in the first
two chapters of the book by Abraham [8]. The
second line of approach is based on statistical
mechanics and originated from a desire to eradicate
the uncertainties that mar the classical theory. It is
linked with the names of Band [9], Bijl [10], Frenkel
others. The statistical mechanical theory has always
been controversial. Some of the earlier problems

C08404 # IMechE 2005 Proc. IMechE Vol. 219 Part C: J. Mechanical Engineering Science
1316 F Bakhtar, J B Young, A J White, and D A Simpson
are discussed in the two interesting collections
edited by Zettlemoyer [14, 15] and a recent careful
derivation is provided by Ford [13]. A number of
hybrid analytical theories have also been developed
(e.g. See Dillman and Meier [16]) and, with the avail-
ability of computers, a numerical line of attack has
been instigated using Monte Carlo methods (e.g.
Hale [17]) and molecular dynamics simulations
(e.g. Laasonen et al. [18]). Nevertheless, despite
a century of research by some of the great names
in physics, physical chemistry, and engineering,
a definitive theory of nucleation has yet to emerge.
The scientific (as opposed to engineering) litera-
ture on nucleation is vast. Much of this work con-
cerns chemical substances other than H2O at rates
of nucleation many orders of magnitude lower than
those observed in steam nozzles and turbines.
Indeed, the experimental equipment most often
used for obtaining data has been the thermal diffu-
sion cloud chamber and the piston expansion
chamber, and the degrees of supersaturation and
rates of nucleation achievable in these devices are
in no way comparable to those obtained during
rapid expansion in nozzles. Much of the literature
on nucleation is therefore not of direct relevance
to engineers and this paper will deal only with
those issues of importance for the homogeneous
nucleation of water droplets from pure steam in
high-speed flows.
The interest of the steam turbine community
stems from the desire to understand wetness pro-
blems in turbines. Laboratory studies of nucleation
and droplet growth have principally been carried
out in converging-diverging nozzles and many such
investigations have been reported. Originally, the
only data available were measurements of the
centre-line pressure distribution (e.g. see Yellot
et al. [19], Binnie and Woods [20], and Binnie and
Green [21]), and these were used to deduce the
limiting supersaturation in high-speed nucleating
flows. The pressure distributions were later used to
compare with theoretical solutions in which the
equations describing droplet formation and growth
were combined with the gas dynamic conservation
equations (e.g. see Oswatitsch [22], Wegener and
Pouring [23], Gyarmathy and Meyer [24], Hill [25],
Saltanov et al. [26], and Campbell and Bakhtar
[27]). Unfortunately, however, pressure distributions
alone were insufficient to resolve the uncertainties
associated with nucleation theory and in the 1960s
and 1970s optical techniques were developed for
measuring the droplet size (e.g. see Deych et al.
[28], Stein [29], Gyarmathy and Lesch [30], Petr
[31], Kròl [32], and Walters [33]). These measure-
ments provided additional data for comparison with
the theoretical solutions. Comparisons between a
representative selection of the droplet measurements
Proc. IMechE Vol. 219 Part C: J. Mechanical Engineering Science
in nozzles available at the time and theoretical
solutions were given by Young and Bakhtar [34] and
a review of all the nozzle experiments performed
before 1980 and a comparison with his theoretical
predictions is given by Young [35]. Comparisons
with data from high-pressure tests were reported
by Bakhtar and Zidi [36]. A new development by
Wyslouzil et al. [37] is the use of small-angle neutron
scattering to measure the size and structure of nano-
scale droplets. The technique is in its infancy but
holds the promise of providing more accurate infor-
mation about small clusters of molecules formed by
nucleation, which can be used to validate nucleation
theories.
The general conclusion from this work is that
qualitatively the classical theory describes nuclea-
tion in steam nozzles well but that quantitatively
the predictive capability is less good. Numerous
modifications of the theory have been proposed
but, unfortunately, a given set of measurements
can be matched by various combinations of models
so that considerable uncertainty still remains. Con-
sequently, the basis for all engineering calculations
is still the classical Becker – Döring theory. The
main steps in the derivation of this theory and
some of the proposed refinements will now be sum-
marized. A critical review of the recent developments
in the statistical mechanical approach is given by
Ford [38].
2 THE CLASSICAL THEORY
2.1 Thermodynamic aspects
The reversible work required to form a single spheri-
cal droplet of radius r from a supersaturated vapour
held at constant pressure p and temperature TG is
equal to the change in Gibbs free energy DG during
the process of formation. If the vapour behaves as a
perfect gas, the so-called classical expression for DG
as a function of r is
4
DG ¼ 4pr 2 s  pr 3 rL RTG ln S (1)
3
where s is the liquid surface tension, rL is the liquid
density, and R is the gas constant per unit mass.
S ¼ p/pS(TG) is the supersaturation ratio with
pS(TG) being the saturated vapour pressure at TG.
The derivation of equation (1) and its extension to
high pressure is given in Appendix 2.
For a supersaturated vapour, S . 1 and the second
term on the right-hand side of equation (1) is nega-
tive. The first term is always positive, however, and
for small r it dominates, as shown by the curves of
DG versus r in Fig. 1. It can be seen that for a given
value of S, DG increases with r up to a maximum
C08404 # IMechE 2005
 Classical nucleation theory and its application
Fig. 1 Schematic diagram showing the variation
of DG with r for superheated (S , 1) and
supersaturated vapour (S . 1)
value DG
occurring at the critical radius r
. A droplet
with r . r
can reduce the free energy of the system
by capturing molecules and has a tendency to
grow. The opposite is true for a droplet with r , r
.
A critical droplet is thus in unstable equilibrium
with the surrounding vapour.
Expressions for r
and DG
are obtained by differ-
entiating equation (1) with respect to r and equating
to zero. If s and rL depend only on temperature and
not on r, and if the vapour is assumed to behave as a
perfect gas, it is found that
2s
r
¼
rL RTG ln S
and
4pr
2 s 16ps 3
DG
¼ ¼ (3)
3 3(rL RTG ln S)2
A droplet of radius r comprising g molecules is
called a g-mer. If m is the mass of a molecule, the
droplet mass is 4pr 3 rL /3 ¼ gm and its surface area
is 4pr 2 ¼ Ag 2/3 , where A3 ¼ 36p(m/rL )2 . Substituting
into equation (1) gives DG in terms of g rather than r
DG As 2/3
¼ g  g ln S
kTG kTG
where k ¼ mR is Boltzmann’s constant. It is often
easier to work with equation (4) than equation (1).
Although supersaturation ratio S appears naturally
in nucleation theory, engineers often prefer to work
in terms of the supercooling DT because of its more
obvious physical significance. By definition, DT ¼
TS( p) 2 TG, where TS( p) is the saturated vapour
temperature at pressure p. S and DT are related
C08404 # IMechE 2005
1317
exactly via the saturated vapour properties but the
following useful approximate relationship can be
obtained from the Clausius – Clapeyron equation
hfg DT
ln S ffi (5)
RTS (p) TG
where hfg is the specific enthalpy of evaporation. DG

is the free energy barrier to the formation of
a droplet.
Equation (3) shows that as S increases DG

decreases but the barrier never disappears comple-
tely. Thus, for a supercooled vapour to condense it
must form droplets of radius r
, which is against its
natural tendency. The only path to the formation of
clusters of critical size is through fluctuations result-
ing from chance molecular collisions. The kinetic
part of the theory models this process.
2.2 Steady state cluster distribution in
a superheated vapour
In a superheated vapour, S , 1 and both terms on the
right-hand side of equation (4) are positive. Sustained
growth of liquid droplets to macroscopic size is pro-
hibited but, even so, small liquid-like clusters are con-
tinually being formed and disrupted by the molecular
collision processes. The identities of the individual
clusters change all the time but the size distribution
remains steady. By minimizing the free energy of
the total vapour – droplet system (e.g. see Chapter 4
of Abraham [8]) it can be shown that the
(2)
steady state size distribution of clusters follows the
Boltzmann law


DG
ng ffi n1 exp  (6)
kTG
where ng and n1 are the numbers per unit volume of
g-mers and monomers respectively.
An expression very similar to equation (6) can also
be obtained from kinetic considerations. Thus,
in order for the cluster concentrations to remain
steady, the principle of detailed balance requires that
Cg ng ¼ Egþ1 ngþ1 (7)
(4)
where Cg is the rate at which a g-mer acquires
a molecule (the condensation rate) and Egþ1 is the
rate at which a (g þ 1)-mer loses a molecule (the
evaporation rate). In writing equation (7), it is
assumed that growth and evaporation occur by the
acquisition or loss of single molecules. Expressions
for Cg and Egþ1 are given below.
Although equations (6) and (7) undoubtedly
embody the essential physics of the vapour-cluster
equilibrium, it is important to appreciate that, as
Proc. IMechE Vol. 219 Part C: J. Mechanical Engineering Science
1318 F Bakhtar, J B Young, A J White, and D A Simpson
used in the classical theory, they are not quite
consistent. Thus, when DG from equation (1) is
substituted into equation (6) and the standard
expressions for Cg and Eg (equations (10) and (12)
below) are introduced into equation (7), the results
do not quite agree. This inconsistency, which is dis-
cussed later in the paper, has proved astonishingly
difficult to eradicate from the theory.
2.3 Derivation of the kinetic equation
In a supersaturated vapour the processes of cluster
formation and disruption occur in a similar way,
but if the clusters exceed the critical size they
encounter a favourable DG gradient and tend to
grow. The steady state described by equations (6)
and (7) is no longer applicable because the rates at
which the clusters grow and decay are no longer
balanced. To signify the difference, the symbol fg is
used to denote the concentration of g-mers under
these conditions. The net rate per unit volume Ig at
which g-mers grow to (g þ 1)-mers is termed the
nucleation current and can be expressed as
Ig ¼ Cg fg  Egþ1 fgþ1
The time rate of change of the concentration of
g-mers is then given by
@fg
¼ ½(Cg fg  Egþ1 fgþ1 )  (Cg1 fg1  Eg fg )
@t
@Ig
¼ (Ig  Ig1 ) ffi 
@g
To proceed further, expressions are needed for
Cg and Eg. Cg can be obtained from the kinetic
theory expression for the rate at which vapour mol-
ecules impact the droplet surface. Assuming that
clusters can be modelled as spherical droplets
down to the smallest sizes
rG v
G p
Cg ¼ qc Ag 2/3 ¼ qc Ag 2/3 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
4m 2pmkTG
where v
G is the mean speed of vapour molecules and
qc is the condensation coefficient, the fraction of
molecules incident on the surface that are absorbed
into the droplet. The second expression follows if
the vapour behaves as a perfect gas, but when deal-
ing with nucleation at high pressure it is simpler to
work in terms of density.
To evaluate Eg it is noted that, if a g-mer of
radius r and temperature TL (not necessarily equal
to TG) were at equilibrium in a supersaturated
environment, the vapour pressure would be given
Proc. IMechE Vol. 219 Part C: J. Mechanical Engineering Science
by the following expression (obtained by inverting
equation (2))


2s
p ¼ pS (TL ) exp (11)
rL RTL r
At this fictitious equilibrium condition, Eg would
equal the value of Cg given by equation (10) with
p from equation (11) and TG replaced by TL. It is
then argued that Eg depends only on TL and r, and
hence, whether the droplet is in a superheated or
supersaturated environment, at equilibrium or other-
wise, the evaporation rate is given by


2/3 pS (TL ) 2s
Eg ¼ qe Ag pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp (12)
2pmkTL rL RTL r
In equation (12), qe is the evaporation coefficient.
For a droplet at equilibrium qe must equal qc but, as
noted by Young [35], this relationship need not
necessarily hold when the droplet is growing or eva-
porating. Normally, however, it is assumed that
q e ¼ q c.
Equation (12) is a general expression for the evap-
(8) oration rate of a g-mer at an arbitrary temperature
TL. In the original classical theory, however, the
assumption was made that TL ¼ TG for all droplet
sizes. This isothermal assumption is difficult to jus-
tify a priori (in fact it cannot be justified in the case
of a pure vapour) but provides great simplification
as it is then unnecessary to consider the transfer of
energy to and from the clusters.
(9) The solution of equation (9) under isothermal
conditions can now proceed along either of two
lines. The more obvious starting point is to substitute
equations (10) and (12) directly. This method is
described by Ford [13] and has the advantage of
transparency but the disadvantage of complexity.
The original classical theory employed a more
elegant, if less obvious, approach, which leads to
the same result and will now be described. Instead
of substituting directly from equation (12), equation
(10) (7) is used to eliminate Egþ1 from equation (8) to give


Ig fg fgþ1 @ fg
¼  ffi (13)
Cg ng ng ngþ1 @g ng
The final expression is a differential approximation
to the discrete form. In equation (13), ng is defined
by equation (6) but with DG evaluated at supersatu-
rated conditions. It is important to appreciate that
ng does not represent an actual real distribution
and the substitution of equation (6) is simply a con-
venient mathematical trick for eliminating Egþ1.
Eliminating Ig from equations (9) and (13), the time
rate of change of the g-mer concentration is thus
C08404 # IMechE 2005
 Classical nucleation theory and its application 1319

approximated in differential form by time obtained by a numerical solution of equation

(14). In this example, the transient stage lasts about


@fg @Ig @ @ fg 0.1 ms whereas the active nucleation period in a
¼ ¼ Cg ng (14) nozzle flow lasts typically 10 – 50 ms. Based on such
@t @g @g @g ng
evidence, the transient stage is usually ignored in
Equation (14) with Cg given by equation (10) and ng condensing flow calculations, it being assumed that
given by equation (6) with DG from equation (1) is the steady-state nucleation rate is attained
the governing differential equation describing iso- instantaneously.
thermal nucleation.
The solution of equation (14) has been studied
under two separate sets of conditions. 2.5 The steady state solution
1. A transient stage where fg and Ig vary with both Taking the nucleation current as constant, equation
g and t. (13) may be integrated as follows
2. A steady-state stage where fg varies with g (but
not t) and Ig ¼ I, the same for all g. This is a ficti- ð1 ð1

dg fg
tious situation brought about by imagining that ICL ¼ d (15)
C
g¼1 g ng g¼1 n g
all large droplets are continually removed and
replaced by an equal mass of supersaturated
vapour. By this artifice, the system can exist in where ICL is the classical steady-state value of I.
a steady state with an unchanging cluster distri- To evaluate the integral on the right-hand side, it
bution fg and a nucleation current I that is con- is noted that fg/ng ! 1 for g ! 1 and fg/ng ! 0 for
stant along the ‘cluster chain’. g ! 1. Thus

"ð #1
1
2.4 The transient solution dg
ICL ¼ (16)
C
g¼1 g ng
Of the many investigations of the transient stage,
there is one by Kashchiev [39] in which he obtains
a semi-analytical solution of equation (14). The solu- This integral can be evaluated analytically with
tion is obtained in terms of a characteristic time t some small approximations. The result, derived in
and Kashchiev shows that if a given supersaturation Appendix 3, is
is suddenly imposed on a vapour initially in a satu-



rated condition, then for g , g
, Ig reaches 99 per r2G 2s 1=2 4pr
2 s
cent of its steady state value after about 5t. Calcu- ICL ¼ qc exp  (17)
rL pm3 3kTG
lations show this timescale to be short in comparison
with the overall time for active nucleation in nozzles.
For example, Fig. 2 shows the development of Ig with where r
is given by equation (2). This is the classical
expression for the rate of formation of critical
droplets per unit volume. Often, however, it is
more convenient to work in terms of the nuclea-
tion rate per unit mass of vapour J. Noting that
S ¼ p/pS(TG) ¼ rG/rGS(TG), where rGS is the satu-
rated vapour density at TG



rGS 2s 1=2 4pr
2 s
JCL ¼ qc S exp  (18)
rL pm3 3kTG

The variation of JCL with supercooling DT at

Fig. 2 Numerical calculation of transient nucleation of
H2O showing the variation of the nucleation
currents Jg with time: TG ¼ 303 K, S ¼ 4.7
various pressures is shown in Fig. 3. The calculations
were performed assuming the surface tension s to be
that of a flat surface and the condensation coefficient
qc to be unity. The nucleation rate is negligibly small,
even at moderate supersaturation ratios, but it rises
dramatically with an increase in DT. It can also be
seen that the value of DT required to achieve a
given value of JCL decreases as the pressure is raised.

C08404 # IMechE 2005 Proc. IMechE Vol. 219 Part C: J. Mechanical Engineering Science
1320 F Bakhtar, J B Young, A J White, and D A Simpson

correction has the effect of reducing the classical

value of J by a factor S. Its inclusion can also be
justified on the grounds that the classical form for
DG is inconsistent with the law of mass action
whereas the revised theory is and that equation (19)
satisfies Kashchiev’s first nucleation theorem
whereas equation (17) does not [13]. The correction
is now accepted by physicists as being fully justified
(e.g. see Reiss et al. [41]). It has been used by the
present authors for a considerable time [34].

3.2 The internally consistent classical theory

Fig. 3 Variation of the classical nucleation rate per
unit mass of vapour with supercooling at
various pressures
An examination of equation (19) shows that the
free energy change for g ¼ 1 (i.e. the formation of
monomers from monomers) is DG ¼ sA rather
than DG ¼ 0 as expected. Indeed, Fig. 4 shows that
from a selection of six different nucleation theories
only one plot of DG versus g actually passes through
the point g ¼ 1, DG ¼ 0. Physicists well-grounded
in statistical mechanics make little of this problem
because they maintain that the important result of
a theory is the prediction of the critical value DG
.
Equation (4) can be modified in a number of ways
to ensure consistency down to g ¼ 1. One possibility,
suggested by Girschik and Chu [42, 43], is to write

DG As 2=3
3 CORRECTIONS AND REFINEMENTS OF ¼ (g  1)  (g  1) ln S (21)
THE CLASSICAL THEORY kTG kTG

which is the one curve in Fig. 4 that passes through

It is beyond the scope of this paper to review all
g ¼ 1, DG ¼ 0. This results in the ICCT (internally
developments in the history of nucleation theory
and only those issues important for the nucleation
of water droplets in condensing flows of pure steam
are mentioned below.

3.1 Courtney’s correction

Courtney [40] noted that the effect of the partial
pressure of the clusters for g . 1 had been omitted
from the classical theory and that equation (4)
should really read

DG As 2/3
¼ g  (g  1) ln S (19)
kTG kTG

Adopting this equation, the expressions for r

and DG
are unchanged but the nucleation rate
becomes



rGS 2s 1=2 4pr
2 s
JCNY ¼ qc exp  (20)
rL pm3 3kTG Fig. 4 Variation of DG/(kTG) with cluster size g for
various nucleation theories: classical equation
where the suffix CNY denotes that Courtney’s (4), Courtney equation (19), ICCT equation (21),
correction is included in the expression. Courtney’s Reiss et al. [57], and Lothe and Pound [56]

Proc. IMechE Vol. 219 Part C: J. Mechanical Engineering Science C08404 # IMechE 2005
 Classical nucleation theory and its application
consistent classical theory) nucleation rate equation




rGS 2s 1=2 sA 4pr
2 s
JICCT ¼ qc exp exp 
rL pm3 kTG 3kTG
(22)
For water, the factor exp[sA/(kTG)] is about 103
at 300 K so the ICCT correction represents a substan-
tial modification. Its validity is doubtful, however,
because the correction is applied at the smallest
cluster size. Nevertheless, there is some unpublished
evidence by J. B. Young that equation (21) leads to a
reasonable estimate, not necessarily of J itself, but of
the steady-state cluster concentration in super-
heated vapour, as predicted by equation (6).
Using the association theory of imperfect vapours
as developed by Band [44], Saltz [45], and others, it is
possible to relate the size distribution of the smallest
clusters to the coefficients in a virial series. If the cal-
culations are based on equation (6) and the classical
expression (4), the resulting values of the second
virial coefficient B(TG) are too small by several
orders of magnitude. If, however, the ICCT expres-
sion (21) is used, a far more accurate result is
obtained. From the association theory of gases, it
can be shown that B(TG) (which is related to the
concentration of dimers) is given, according to the
ICCT theory, by
 
1 sA 2/3
B(TG ) ¼ exp  (2  1) (23)
rGS (TG ) kTG
Figure 5 shows a comparison between experimental
values of BrGS and those calculated from equation
(23) and also the classical theory, plotted as functions
of TG. However dubious the ICCT result, it has to
be conceded that an ability to predict B(TG) over
Fig. 5 Comparison of ICCT [equation (23)] and classical
prediction of the second virial coefficient for
water with data from the steam tables of
Keenan et al. [92]
C08404 # IMechE 2005
1321
a wide range of temperatures to within a factor of
2 – 3 just using surface tension data is impressive.
3.3 Variation of surface free energy with
droplet radius
Consideration of equation (18) for JCL with equation
(3) for DG
shows that the surface free energy of the
critical cluster s appears raised to the third power
in the exponential part of the expression. Small
changes in s clearly have a dramatic effect on the
nucleation rate.
At the very high supersaturation ratios achieved in
condensing nozzle flows, a critical droplet is com-
posed of 10 –50 molecules. It has been the subject
of much debate as to whether the surface free ener-
gies of such small clusters are well represented
by flat-film values of surface tension. Based on the
Gibbs theory of surface tension, liquid drop nuclea-
tion theory can be formally modified to include the
effect of a radius-dependent surface free energy
(Abraham [8], Chapter 2), but the problem comes
in specifying exactly how s varies with r. Several
theories have been proposed, e.g. by Tolman [46],
Koening [47], Kirkwood and Buff [48], Benson and
Shuttleworth [49], Plummer and Hale [50], Sinangoglu
[51], and Kegel [52], but there is no general agreement,
even on the sign of the variation. However, because
of the sensitivity of J to changes in s, it is unlikely
that a satisfactory nucleation theory will be forth-
coming until this problem is resolved.
3.4 The condensation coefficient
The condensation coefficient qc appears in the pre-
exponential factor in the nucleation rate equation
and so its effect is less drastic than that of the surface
free energy. Early experiments by Alty and MacKay
[53] indicated values around 0.02 – 0.03, but it is
now thought that these measurements were subject
to large systematic errors and the general consensus
is that qc takes values close to one (e.g. see Mills and
Seban [54] and Mozurkewitch [55]). Such values are
also consistent with results from Millikan’s classic oil
drop experiment, where it was estimated that about
90 per cent of incident molecules suffered diffuse,
as opposed to specular, reflection (implying that
they had accommodated completely to the surface
conditions before re-emission).
Although most workers take qc ¼ 1, it should be
noted that all experimental evidence has been
obtained with surfaces large on a molecular scale.
For very small clusters the value of qc is almost
certainly much less than one (consider, for example,
the extreme case of dimer formation from
monomer –monomer collisions). However, if the
Proc. IMechE Vol. 219 Part C: J. Mechanical Engineering Science
1322 F Bakhtar, J B Young, A J White, and D A Simpson
non-isothermal correction discussed below is
adopted, this has the fortuitous effect of making the
calculated nucleation rate almost insensitive to qc
for values in the range 0.1 – 1.0.
3.5 The translation –rotation paradox
Equation (1) for DG presupposes that droplets are
stationary in the bulk of the vapour despite the fact
that small clusters clearly behave as macromolecules
and are in continuous motion with translational and
rotational kinetic energy. Lothe and Pound [56]
attempted to correct DG for this omission and, in
doing so, stimulated the biggest controversy in the
history of nucleation theory. For water their analysis
indicated that the value of JCL should be increased
by about 17 orders of magnitude! Although this
obviously led to serious disagreement between
theory and experiment, it was found that agreement
could be restored simply by arbitrarily increasing the
surface free energy of the critical cluster by about
15 per cent.
The so-called translation–rotation paradox became
the main issue in nucleation theory for many years
and, in truth, has never been completely and satisfac-
torily resolved. The consensus of opinion at present
is that a correction, known as the replacement free
energy, is also required [57] and this largely cancels
the translation and rotation terms. Some of the
issues involved are discussed in the books edited by
Zettlemoyer [14, 15].
In passing, it is interesting to refer to Fig. 4, which
includes a curve showing the Lothe– Pound
expression for DG. Given that DG never becomes
positive, even for small clusters, it is difficult to see
why the theory was originally given such credence.
3.6 Configurational entropy corrections
An interesting study by Hale [58] on scaling the
nucleation rate equation for different substances
showed that, although the dependence of JCL on
the supersaturation ratio appeared to be correct, the
temperature dependence was less satisfactory. This
led to the introduction of a term proportional to
ln(g) in the DG expression, which (with the ICCT
correction) becomes
DG As 2 / 3
¼ (g  1)  (g  1) ln S þ tH ln g (24)
kTG kTG
where tH is a numerical factor. The extra term
(used much earlier by Fisher [59] in his theory of
the critical region) was justified as representing
configurational entropy contributions. According to
Proc. IMechE Vol. 219 Part C: J. Mechanical Engineering Science
Hale, the required value of tH was about 2.2 (in agree-
ment with Fisher’s value).
A ln(g) term was also used by Dillman and Meier
[16] in their semi-empirical method for improving
the equation for the nucleation rate. In order to pro-
vide more flexibility they also allowed the surface
tension to vary with droplet radius and attempted
to fix this variation empirically by requiring the
concentration of dimers to be consistent with experi-
mental values of the second virial coefficient
(in much the same way as described in section 3.2
above).
The Dillman and Meier theory initially appeared
to be extremely successful in predicting nucleation
rates in good agreement with experimental data,
not only for water but also for a number of other
substances that had previously proved troublesome.
Unfortunately, however, the theory has since been
shown by Laaksonen et al. [60] to include inconsis-
tencies which, when corrected, tend to spoil the
comparisons. Nevertheless, the Dillman– Meier
theory has had a beneficial effect in stimulating
theoretical work. For example, the ln(g) term has
now been justified by Reiss et al. [41], although
with a rather different coefficient of tH ffi 0.5.
3.7 Non-isothermal nucleation
As noted in section 2.2, it is assumed in the iso-
thermal classical theory that the vapour and droplet
temperatures (for all sizes of droplet) are equal,
TL ¼ TG. It is then unnecessary to consider the details
of how the energy released to a growing droplet by
the condensing vapour molecules finds its way
back to the vapour.
In the presence of a large quantity of inert gas (as
in diffusion cloud chambers or nozzle flows of
moist air), the isothermal approximation may be
justified because collisions between droplets and
non-condensible molecules provide the means for
energy transfer with only a small interphase temp-
erature difference. In flows of pure vapour, however,
a substantial temperature difference (TL 2 TG) is
required to drive the heat transfer. This affects the
nucleation rate because the evaporation rate, as
given by equation (12), depends on TL.
Despite its fundamental importance, remarkably
little work has been done on energy transfer in
nucleation. The problem was first considered by
Kantrowitz [61] who derived a simple correction
to the isothermal theory that seemed to be corrobo-
rated by the later and more elaborate analysis of
Feder et al. [62]. More recently, however, Wyslouzil
and Seinfeld [63] have re-examined the problem
with the conclusion that the rate-controlling step is
associated with energy transfer at the smallest cluster
C08404 # IMechE 2005
 Classical nucleation theory and its application
sizes. Unfortunately, it is this aspect of the nuclea-
tion process that is the most difficult to specify
accurately.
Kantrowitz’s non-isothermal theory has the effect
of multiplying the isothermal nucleation rate by
a factor c. Thus
JISO
JNISO ¼ cJISO ¼ (25)
1þn
where JNISO and JISO are the non-isothermal and
isothermal nucleation rates respectively. JISO can
include any correction thought to be appropriate.
As shown in Appendix 4, the quantity n is given by
the expression
rffiffiffiffiffiffiffiffiffi


rG R RTG hfg hfg 1 and droplet growth are intimately linked and it is dif-
n ¼ qc  (26)
ar 2p RTG RTG 2
where ar is the surface heat transfer coefficient of a
critical droplet of radius r
. At low pressure, r
is
always much smaller than the mean free path of
a vapour molecule and equation (26) then takes
the simpler form


2(g  1) hfg hfg 1
n ¼ qc  (27)
(g þ 1) RTG RTG 2
where g is the isentropic index of the vapour.
Typically, Kantrowitz’s correction reduces the nuclea-
tion rate for water by a factor of 50 – 100.
3.8 Choice of corrections
It should now be clear that there is no general
agreement on the precise equation for the nucleation
rate, even for such a well-researched substance
as water. Most investigators therefore accept the
status quo and tune the theory to fit the experimental
measurements by prefixing the expression for J
with an adjustable coefficient. Alternatively, as will
be shown below, the classical theory augmented
by the established corrections of Courtney and
Kantrowitz provides a fairly accurate representation
over a wide pressure range.
4 CONDENSATION IN CONVERGING
DIVERGING NOZZLES
As discussed earlier, experimental investigations of
limiting supersaturation and nucleation have been
pursued along two parallel lines. Pure scientific
studies have utilized piston expansion chambers
(e.g. see Viisanen et al. [64] and Peters and Meyer
[65]), thermal diffusion cloud chambers (e.g. see
C08404 # IMechE 2005
1323
Fisk and Katz [66]) and shock tubes (e.g. see Luijten
et al. [67]), while engineers have used converging –
diverging nozzles. The highest rates of cooling are
achieved in nozzle expansions and these provide
the best laboratory simulations of condensation in
steam turbines. The other techniques have only
been used to study nucleation in the presence of
large quantities of inert carrier gas and the measure-
ments are not directly applicable to nucleation in
pure steam flows.
4.1 Nucleation and droplet growth in
nozzle flows
In nozzle condensation, the processes of nucleation
ficult to separate the phenomena for individual
study. This can be illustrated by reference to Fig. 6,
which shows the results of a typical nucleation and
growth calculation in a hypothetical nozzle having
a constant rate of expansion, P ¼ 2500 s21 (the rate
of expansion is defined as P ¼ 2D(ln p)/Dt, where
D/Dt is the time rate of change following a fluid par-
ticle). The calculations have been initiated from a dry
saturated condition (DT ¼ 0) where the pressure and
temperature are 5.0 bar and 425 K respectively. It
should be noted that the scale on the abscissa is in
units of time rather than distance in order to empha-
size the temporal development of the flow.
Figure 6(a) shows the variations of supercooling
DT and wetness fraction Y through the expansion,
while the variations of the nucleation rate and drop-
let diameter are given in Fig. 6(b). As the expansion
proceeds, the supercooling rises rapidly although
the steam remains dry (Y ¼ 0) because the nuclea-
tion rate J is still negligible. However, as DT increases
so too does J and eventually the rate of heat release
due to condensation on the surfaces of the freshly
nucleated droplets is sufficient to prevent a further
increase in supercooling. DT then falls as Y rises
towards its equilibrium value.
The formation and growth of the droplets is shown
in detail in Fig. 6(b). The nucleation zone in the vicin-
ity of the maximum nucleation rate Jmax is defined
approximately by 0.01Jmax , J , Jmax and is short
both in time and space (about 20 ms in this example).
Outside this zone, nucleation can effectively be
ignored. Hence, integrating J with respect to time
through the nucleation zone (and ignoring aggl-
omeration, which can be shown to be negligible)
gives the number density of droplets in the down-
stream flow. However, although the number density
is directly related to the nucleation rate, it is the
condensation of vapour molecules on the surfaces
of these droplets that is mainly responsible for
the increase of liquid mass fraction: colloquially,
Proc. IMechE Vol. 219 Part C: J. Mechanical Engineering Science
1324 F Bakhtar, J B Young, A J White, and D A Simpson
Fig. 6 Variation of properties in a typical condensing
nozzle flow with a constant rate of expansion
2500 s21
‘nucleation supplies the number and droplet growth
supplies the mass’.
Figure 6(b) shows the growth of droplets formed at
different locations in the nucleation zone and illus-
trates why, far downstream of this zone, the droplet
size distribution has a wide spread. The curves
marked maximum and minimum refer to the
growth of those droplets formed at the beginning
and end of the nucleation zone. The droplets
nucleated earlier will have resided longer in a
Proc. IMechE Vol. 219 Part C: J. Mechanical Engineering Science
supercooled environment and grown to a larger
size. The dispersion (as opposed to the mean
diameter) is very difficult to measure. The neutron
scattering technique pioneered by Wyslouzil et al.
[37] may eventually be able to provide more accurate
information on the size distribution of these sub-
micrometre droplets.
4.2 Experimental data for stable subcritical
condensation in nozzle flows
Because of the high surface tension of water and high
expansion rates associated with nozzle flows, the
supercooling necessary to produce the high nuclea-
tion rates required for reversion to equilibrium of
steam in a nozzle expansion is substantial. For this
reason, in nozzle flows starting from saturated
or superheated states, it is necessary to expand the
steam to supersonic velocities before it supercools
sufficiently to condense. Thus, in investigations of
nucleating flows in steady-flow nozzles the nuclea-
tion and rapid condensation zones occur in the
diverging section where the flow is supersonic.
The release of the latent heat of condensation to
the supersonic stream tends to retard the flow and
cause the pressure to rise. When nucleation occurs
sufficiently far downstream of the throat, the Mach
number remains supersonic throughout the rever-
sion to equilibrium despite the heat release. This is
referred to as subcritical condensation in contrast
to supercritical condensation to be described
below. A typical example of subcritical condensation
is the experimental test case ‘nozzle B’ of Moore et al.
[68]. The inlet stagnation pressure and temperature
for this test were 0.250 bar and 358.1 K respectively,
corresponding to an inlet superheat of 20.0 K. The
experimental data are shown in Fig. 7 and consist
of the centre-line pressure distribution and the
Sauter mean droplet radius r32 far downstream of
the nucleation zone.
Figure 7 also includes some theoretical results
obtained using the two-dimensional, unstructured-
mesh, time-marching method for viscous wet-steam
flows described by Simpson and White [69]. The
code captures the two-dimensional flow structure
and the viscous boundary layers and is probably
the most accurate method currently available for
predicting nozzle flows with condensation.
The calculations of Fig. 7 were performed with four
different models of nucleation and droplet growth. In
this figure, JCL refers to the classical nucleation rate
(equation (18)), evaluated with the flat-film value of
surface tension and a condensation coefficient of
unity, JCNY includes Courtney’s correction for the
cluster partial pressures (equation (20)) and c refers
to Kantrowitz’s modification for non-isothermal
effects (equation (25)).
C08404 # IMechE 2005
 Classical nucleation theory and its application
Fig. 7 Experimental data and calculations for nozzle B
of Moore et al. [68]: Case 1, JCL and standard
droplet growth; Case 2, cJCL and standard
growth; Case 3, cJCNY and standard growth; Case
4, cJCNY and enhanced growth. An empirical
factor a is defined by qc/qe ¼ 1 þ a(TL 2 TG)/
TS(p) (see Young [35])
For Case 1, the classical nucleation rate JCL was
used, together with the standard droplet growth
model of Gyarmathy [70, 71], also described by
Young [35, 72]. In brief, for droplets much smaller
than the mean free path of a vapour molecule, the
growth rate is calculated using the well-known
Hertz– Knudsen theory. For droplets much larger
than the mean free path, the equations of
continuum heat transfer apply. For intermediate
sizes, Langmuir’s flux-matching theory provides a
physically based interpolation method.
As shown in Fig. 7, the nucleation and growth
models of Case 1 give poor agreement, both with
the pressure and droplet size measurements. Case 2
adds the non-isothermal correction of Kantrovitz
(cJCL), described in section 3.7. This decreases the
nucleation rate, thus shifting the condensation
downstream and giving a good match with the pres-
sure data. Unfortunately, however, it does little to
increase the droplet size. Case 3 adds Courtney’s cor-
rection (cJCNY). This shifts the condensation even
further downstream, destroys agreement with the
pressure distribution and only marginally increases
the droplet size.
This problem is characteristic of all very low pres-
sure nozzle tests and was first noted by Young [35].
He found that good agreement with both pressure
and droplet size measurements could be obtained
using cJCNY and the standard droplet growth
model, but only for those tests where the pressure
C08404 # IMechE 2005
1325
at the Wilson point was around 1 bar or higher. As
the Wilson point pressure fell, so the level of agree-
ment deteriorated. The Wilson point pressure in
Fig. 7 is about 0.1 bar, a very low value but nonethe-
less typical of nucleation in the large low-pressure
(LP) steam turbines used for electricity power
generation.
It should be noted that the Wilson point is best
defined as the point of maximum supercooling.
This does not quite correspond to the point of
maximum nucleation rate, nor to the point of mini-
mum pressure in a supersonic nozzle test. Although
the latter condition has often been used to define
the Wilson point, it is not to be recommended
because the pressure variation depends on the
nozzle shape and local flow behaviour. In subsonic
condensation, for example, reversion to equilibrium
is marked by a fall rather than a rise in pressure
and there is no minimum.
Young [35] postulated that agreement with both
pressure and droplet size data required a generally
higher droplet growth rate. The only way to achieve
this without violating physical principles appeared
to be to allow the ratio qc/qe (see equations (10)
and (12)) to exceed unity by a factor proportional
to TL 2 TG. Case 4 of Fig. 7 shows that such an
assumption does indeed result in good agreement
and this is also true for the other available low-
pressure nozzle tests. It must be stressed, however,
that the hypothesis is tenuous and was devised
solely to provide a physically acceptable explanation
of an awkward experimental fact.
At Wilson point pressures of around 1 bar, good
agreement with both pressure and droplet size
measurements can be obtained without artificially
enhancing the droplet growth rate. At these pres-
sures, deviations from perfect gas behaviour become
important and it is necessary to include in the
calculations a suitable equation of state for steam
such as the truncated virial equation developed by
Young [73]. An example of the level of agreement
that can then be achieved is provided by comparison
with the tests carried out by Petr [31] in three
nozzles with expansion rates of about 2000, 5000,
and 8000 s21.
The calculations were originally reported by
Young and Bakhtar [34] and were obtained using a
one-dimensional calculation procedure of fourth-
order accuracy and a virial equation of state truncated
after the second term. Droplet growth rates were calcu-
lated by simultaneous solution of the interphase heat
and mass transfer equations. All calculations were per-
formed using classical nucleation theory with the
refinements of Courtney and Kantrowitz. The effects
of varying the condensation coefficient and surface
tension (according to the Benson and Shuttleworth
[49] model, sg ¼ s1 (1  ð1/3Þ g 1/3 )Þ were also
Proc. IMechE Vol. 219 Part C: J. Mechanical Engineering Science
1326 F Bakhtar, J B Young, A J White, and D A Simpson
Fig. 8 Experimental data and calculations of droplet
size for the three nozzles of Petr [31] showing
the effect of varying the condensation coefficient
qc and surface tension sr (according to the
model of Benson and Shuttleworth [49])
investigated. Figure 8 shows a comparison between the
experimentally measured droplet sizes and the cal-
culations. At Wilson point pressures in the range
1.0 –1.5 bar, excellent agreement is achieved with
qc ¼ 1 and sg ¼ s1. However, for the two tests at
Proc. IMechE Vol. 219 Part C: J. Mechanical Engineering Science
pressures of 0.4 –0.5 bar, the agreement is beginning
to deteriorate, in keeping with the remarks above
about low-pressure condensation.
A word of caution is in order against placing exces-
sive reliance on the accuracy of droplet size data.
Firstly, the amount of data reported is actually very
small. Secondly, the chemicals used for water treat-
ment lead to the presence of impurities in the
steam and droplets formed by heterogeneous
nucleation round them can affect the light extinction
measurements with potential for errors. However,
without the benefit of an extensive new study
using modern techniques and past experience, it is
impossible to draw more precise conclusions.
It may also be added that, in turbine flows, the
droplets formed by heterogeneous nucleation are
relatively few in number and usually grow to a
sufficient size to deposit on the blade and casing
surfaces. They provide a mechanism for the transfer
of chemical species and water to the metallic sur-
faces, and cause erosion and corrosion in the phase
transition zone. However, condensation on these
droplets and on the blade and casing surfaces is
too small in magnitude to affect the behaviour
of the main flow. The dominant mechanism for
condensation in high-speed flows is therefore
homogeneous nucleation.
4.3 Supercritical condensation in nozzles
In supercritical condensation, reversion to equili-
brium occurs close enough to the throat for the
heat release to be sufficient to decelerate the flow
to the sonic velocity. A number of different operating
regimes have been identified, including a steady
state mode with an embedded adiabatic shock
wave and a periodically oscillating unsteady mode
in which the nucleation zone forms downstream
of the throat, propagates upstream, and finally dis-
appears in the subsonic section. Such unstable con-
ditions have been studied by Schmidt [74], Pouring
[75], Barschdorff [76], Gyarmathy [77], Yousif et al.
[78], Saltanov [79], Skillings and Jackson [80],
Moheban [81], Guha and Young [82], White and
Young [83], and Collignan and Laali [84]. A study of
the different unsteady modes of operation in moist
air (not pure steam) has been carried out by Adam
and Schnerr [85], who also discovered an unusual
type of flow that oscillated periodically in both the
streamwise and transverse directions. A numerical
solution for a similar asymmetric flow in a nozzle
with pure steam has recently been obtained by
Simpson and White [69], although this has not yet
been confirmed experimentally. A critical review of
the subject is given by Schnerr [86].
Figure 9 shows experimental data for oscillatory
condensation in a pure steam nozzle obtained by
C08404 # IMechE 2005
 Classical nucleation theory and its application
Fig. 9 Experimental data and calculations for the
periodically unsteady nozzle test (period T) of
Skillings et al. [87]: cJCL and enhanced droplet
growth
Skillings et al. [87] and is one of the few tests avail-
able that includes both droplet size and pressure
measurements. The inlet stagnation pressure and
temperature were 0.245 bar and 346.7 K respectively,
giving an inlet superheat of 9.0 K. It should be noted
that the far-downstream value of r32 actually varies
periodically but the optical equipment could only
record the time-averaged value. The computational
results were obtained using the two-dimensional
viscous code of Simpson and White [69]. In order
to predict the measured value of r32 at such low
pressure, it was necessary to use the enhanced
droplet growth theory. Reasonable agreement was
then obtained for the oscillation frequency and the
amplitude of the pressure fluctuations.
4.4 Nucleation at high pressure
Nuclear power stations using pressurized water
reactors have steam conditions at the inlet to the
high-pressure (HP) turbines that are close to dry
saturated. Unlike conventional plant, this means
that nucleation occurs also in the HP turbine
at comparatively high-pressure. It is therefore of
interest to investigate high-pressure condensation
in laboratory nozzles and a number of such studies
have been carried out, including those by Deich
et al. [88], Gyarmathy et al. [89], Valha and Ryley
[90], and Bakhtar and Zidi [91].
Theoretical solutions of nucleating flows are
dependent on the data adopted for the thermodyna-
mic properties of supercooled steam, particularly
the isobaric specific heat capacity and isentropic
index. Even at moderate and low pressure the
location of the nucleation zone in theoretical
solutions is sensitive to changes in the value of the
C08404 # IMechE 2005
1327
isentropic index and the discrepancies become
greater the higher the pressure.
Mutually consistent properties can be obtained by
adopting equations for the saturation line, cp at zero
pressure and a suitable p –v –T equation of state. Of
these, the first two are subject to essentially no
uncertainty and the problem therefore becomes
one of finding a suitable equation of state. The
equations used for calculating steam tables are only
valid for equilibrium states (e.g. see Keenan et al.
[92]); to calculate supercooled properties it is necess-
ary to extrapolate the relations into the metastable
region. This may be beyond their range of validity.
In an investigation by Bakhtar and Piran [93], it
was found that the virial equation developed by
Vukalovich [94] remains well behaved under extra-
polation. The important requirement is an accurate
representation of the changes in cluster concen-
trations near and below the saturation line, as it
is these that are responsible for the rapid changes
in property values. Vukalovich’s equation of state is
based on the association theory of gases and its result-
ing cluster distributions are realistic. The properties
of steam calculated from these equations are in excel-
lent agreement with the 1967 skeleton tables in
the superheated region and on the saturation line.
The corrections to these equations to allow for the
changes in cluster concentrations under nucleating
conditions were found to be negligibly small.
At high pressures, a further necessary refinement
concerns the droplet growth rates because these
must be calculated from the simultaneous solution
of the heat and mass transfer equations. Particularly
at high pressures, the droplets grow in size from
small to large in comparison with the mean free
path of a vapour molecule. The transfer of heat
between a small spherical surface and the surround-
ing atmosphere over the entire range of Knudsen
numbers presents no special problems, but the posi-
tion is different in the case of mass transfer. The
nucleated drops are initially small in comparison
with the mean free path and the mass transfer rate
can be calculated from collisionless kinetic theory.
The pattern changes as the droplets become larger
until eventually the transport processes fall within
the continuum regime. Mass is then transferred
entirely by diffusion and there is considerable uncer-
tainty about the coefficient of self-diffusion of steam.
The use of the values of the coefficient calculated
from existing theories lead to unrealistic results and
it was found necessary to develop an expression for
the coefficient of self-diffusion using a different
approach [95]. This was then used to develop an
expression for the mass transfer rate applicable to
all regimes.
Adopting these refinements, the treatment can
proceed as before. Using the expression for cJCNY
Proc. IMechE Vol. 219 Part C: J. Mechanical Engineering Science
1328 F Bakhtar, J B Young, A J White, and D A Simpson
with qc ¼ 1 and sg ¼ s1 exactly as for low pressure,
good agreement has been obtained with pressure
and droplet size measurements for inlet stagnation
pressures as high as 100 bar [36]. An example is
shown in Fig. 10. It may be added that the measured
droplet sizes in this study were subject to some error
due to the presence of impurities in the steam.
4.5 Simplified methods for predicting
nucleation and droplet growth
All of the calculations discussed above were car-
ried out using numerical integration of the droplet
growth equations through the nucleation zone, a
procedure that requires a highly developed compu-
ter code. To avoid this problem, two simpler pro-
cedures have been reported, a tabular method by
Bakhtar and Webb [96] and an analytical method
by Huang and Young [97].
The tabular method provides pre-computed data
on droplet and vapour behaviour after the sudden
imposition of a given degree of supercooling on
the steam. For each given pressure and imposed
supercooling the data give the time required to
form given amounts of water and the resulting
mean droplet size. This method can be used for
estimating the nucleating behaviour of steam
Fig. 10 Comparison of theoretical and experimental
results at high pressure (from Bakhtar and
Zidi [36])
Proc. IMechE Vol. 219 Part C: J. Mechanical Engineering Science
Fig. 11 Variation of the Wilson point supercooling
DTw with expansion rate P for various Wilson
point pressures pw. The comparison is between
the analytical method of Huang and Young
[97] (solid lines) and a multi-droplet-group
numerical integration procedure (black circles).
It is assumed that P is constant through the
nucleation zone up to the Wilson point
without the need for elaborate calculations and is
particularly useful for estimating reversion beha-
viour of steam in turbines.
More recently Huang and Young [97] have
developed a method in which the Wilson Point
supercooling DTw is calculated from an analytic alge-
braic equation. Once DTw is known, it is possible to
obtain analytically the total number of droplets
formed in the nucleation zone, the far-downstream
Sauter mean radius and other properties of interest,
including an approximation to the droplet size distri-
bution. Figure 11 illustrates the excellent accuracy of
the method by comparing some analytical estimates
of DTw at different Wilson point pressures and
expansion rates, with calculations using the full
multi-droplet-group numerical integration approach
(as used to obtain Fig. 6). Finally, it should be noted
that, although the calculations of Fig. 11 have
assumed a constant expansion rate through the
nucleation zone, the theory is not restricted by
such an approximation.
5 CONCLUSIONS
The paper has reviewed the classical theory of the
homogeneous nucleation of water droplets from
supersaturated vapour, focusing on the major modi-
fications that have been proposed since publication
of the original Becker – Döring theory. Despite a
huge research effort, many uncertainties remain
C08404 # IMechE 2005
 Classical nucleation theory and its application
and it is concluded that the predictive accuracy
required for engineering calculations is not yet
achievable with a theory derived from first principles.
The areas that require most attention relate to the
properties of small molecular clusters and the
energy transfer processes in the non-isothermal
theory.
Experiments in converging –diverging nozzles
still provide the best means of validating the
theory at the very high nucleation rates of interest.
However, measurements of nozzle pressure distri-
bution and mean droplet diameter are insufficient
to provide completely independent checks on the
separate theories of nucleation and droplet growth.
At present, the most successful form of the nuclea-
tion rate equation is the original Becker –Döring
equation with the Courtney and Kantrowitz modifi-
cations, a condensation coefficient of one and an
unmodified flat-film value of surface tension. This
prescription, in combination with a standard
droplet growth model and a suitable equation of
state for steam, can provide accurate predictions
of nozzle pressure distributions and Sauter mean
droplet radii over a wide range of conditions. The
exception appears to be at very low pressures,
where there is evidence that the droplet growth rate
in the nucleation zone is underestimated. Because
nucleation in LP steam turbines occurs at similar
pressures, this is a problem of considerable practical
importance.
Finally, attention has been drawn to two
approaches, one tabular and one analytical, for esti-
mating the limiting supercooling and droplet size
without undertaking extensive numerical calcu-
lations. Such methods should find application in
reducing the heavy computational burden of nuclea-
ting flow calculations and should also prove useful in
exploring the relationship of nucleation and droplet
growth in the crucial region leading up to the
Wilson point.
ACKNOWLEDGEMENTS
The fourth author, DAS, is grateful for the support of
an Engineering and Physical Sciences Research
Council (EPSRC) research studentship. He also
wishes to thank his PhD supervisor, the third
author, AJW, who in turn acknowledges his PhD
supervisor, the second author, JBY, who is grateful
to his own PhD supervisor, the first author, FB. The
first author has considerable pleasure in acknowled-
ging the contribution of all his research students to
the progress of the subject and wishes to place on
record how much he has enjoyed working with
every one of them.
C08404 # IMechE 2005
1329
REFERENCES
1 Volmer, M. and Weber, A. Keimbildung in Ubersattigten
Gebilden. Z. Phys. Chem., 1926, 119, 277 – 301.
2 Farkas, L. Keimbildungsgeschwindigkeit in übersättigten
Dampfen. Z. Phys. Chem. (Leipzig), 1927, A125, 236–242.
3 Becker, R. and Döring, W. Kinetische behandlung der
Keimbildung in ubersattingten damfen. Annln der
Physik, 1935, 24, 719 –752.
4 Zeldovich, J. Theory of the formation of a new phase.
J. Expl. Theoret. Physics (USSR), 1942, 12, 525.
5 Frenkel, J. Kinetic theory of liquids, 1955 (Dover,
New York).
6 McDonald, J. E. Homogeneous nucleation of vapour
condensation. I – Thermodynamic aspects. Am. J.
Physics, 1962, 30, 870 – 877.
7 McDonald, J. E. Homogeneous nucleation of vapour
condensation. II – Kinetic aspects. Am. J. Physics,
1963, 31, 31 – 41.
8 Abraham, F. F. Homogeneous nucleation theory, 1974
(Academic Press, New York).
9 Band, W. Dissociation treatment of condensing
systems. J. Chem. Phys., 1939, 7, 324 – 326 and 927 – 931.
10 Bijl, A. Discontinuities in the energy and specific
heats. PhD Dissertation, University of Leiden, The
Netherlands, 1938.
11 Reiss, H. Treatment of droplike clusters by means of
the classical phase integral in nucleation theory.
J. Stat. Physics, 1970, 2, 83 – 104.
12 Dunning, W. J. General and theoretical introduction.
In Nucleation (Ed. A.C. Zettlemoyer), 1969 (Marcel
Dekker, New York).
13 Ford, I. J. Nucleation theorems, the statistical mecha-
nics of molecular clusters and a revision of classical
nucleation theory. Phys. Rev. E, 1997, 56, 5615 – 5629.
14 Zettlemoyer, A. C. (Ed.) Nucleation, 1969 (Marcel
Dekker, New York).
15 Zettlemoyer, A. C. (Ed.) Nucleation phenomena, 1977
(Elsevier, Oxford).
16 Dillman, A. and Meier, G. E. A. A refined droplet approach
to the problem of homogeneous nucleation from the
vapour phase. J. Chem. Phys., 1991, 94, 3872–3884.
17 Hale, B. N. Monte Carlo calculations of effective
surface tension for small clusters. Aust. J. Physics,
1996 49, 425 – 434.
18 Laasonen, K., Wonczak, S., Strey, R., and Laaksonen, A.
Molecular dynamics simulations of gas–liquid nuclea-
tion of Lennard–Jones fluid. J. Chem. Phys., 2000, 113,
9742–9747.
19 Yellott, J. I., Holland, C. K., and Hoboken, N. J. The
condensation of flowing steam. Trans. ASME, 1937,
59, 171– 183.
20 Binnie, A. M. and Woods, M. W. Pressure distribution
in a convergent – divergent steam nozzle. Proc. Instn
Mech. Engrs, 1938, 138, 229 – 266.
21 Binnie, A. M. and Green, J. R. An electrical detector of
condensation in high velocity steam. Proc. R. Soc., 1943,
A181, 134– 154.
22 Oswatitsch, K. Kondensationsercheinungen in
Überschalldüsen. Z. Agnew. Math. Mechanics, 1942,
22, 1– 14.
Proc. IMechE Vol. 219 Part C: J. Mechanical Engineering Science
1330 F Bakhtar, J B Young, A J White, and D A Simpson
23 Wegener, P. P. and Pouring, A. A. Experiments on con-
densation of water vapour by homogeneous nucleation
in nozzles. Phys. Fluids, 1964, 7, 352 – 361.
24 Gyarmathy, G. and Meyer, H. Spontaneous con-
densation phenomena – Parts 1 and 2. In VDI
Forschungsheft, 1965 Vol. 508, (VDI-Verlag, Dusseldorf).
25 Hill, P. G. Condensation of water vapour during super-
sonic expansion in nozzles. J. Fluid Mechanics, 1966,
25, 593 – 620.
26 Saltanov, G. A. Seleznev, L. I., and Tsiklauri, G. V.
Generation and growth of condensed phase in high-
velocity flows. Int. J. Heat and Mass Transfer, 1973,
16, 1577 – 1587.
27 Campbell, B. A. and Bakhtar, F. Condensation
phenomena in high speed flows of steam. Proc. Instn
Mech. Engrs, 1970 – 71, 185(25/71), 395 – 405.
28 Deych, M. Ye., Kurshakov, A. V., Saltanov, G. A., and
Yatcheni, I. A. A study of the structure of two phase
flow behind a condensation shock in supersonic
nozzles. Heat Transfer Sov. Res., 1969, 1, 95 –105.
29 Stein, G. D. Angular and wavelength dependence of
light scattered from a cloud of particles formed by
homogeneous nucleation. J. Chem. Phys., 1969, 51,
938 – 942.
30 Gyarmathy, G. and Lesch, F. Fog droplet observations
in Laval nozzles and in an experimental turbine. Proc.
Instn Mech. Engrs, 1969 – 70, 184, (3G), 88 – 95.
31 Petr, V. Measurement of an average size and number of
droplets during spontaneous condensation of super-
saturated steam. Proc. Instn Mech. Engrs, 1969 – 70,
184 (3G), 73 – 79.
32 Kròl, T. Results of optical measurements of diameters
of drops formed due to condensation of steam in a de
Laval nozzle (in Polish). Trans. Inst. Fluid Flow Machi-
nery (PIMP, Poland), 1971, 57, 19.
33 Walters, P. T. Optical measurements of water droplets
in wet steam. In Wet steam 4, 1973, pp. 32 – 40 (Insti-
tution of Mechanical Engineers, Bury St Edmunds
and London).
34 Young, J. B. and Bakhtar, F. A comparison between
theoretical calculations and experimental measure-
ments of droplet sizes in nucleating steam flows.
Trans. Inst. Fluid Flow Machinery (PIMP, Poland),
1976, 70– 72, 259 –271.
35 Young, J. B. The spontaneous condensation of steam in
supersonic nozzles. PhysicoChemical Hydrodynamics
(PCH), 1982, 3, 57– 82.
36 Bakhtar, F. and Zidi, K. Nucleation phenomena in
flowing high-pressure steam. Part 2: theoretical analy-
sis. Proc. Instn Mech. Engrs, Part A: J. Power and
Energy, 1990, 204, 233 – 242.
37 Wyslouzil, B. E., Wilemski, G., Cheung, J. L., Strey, R.,
and Barker, J. Doppler shift anisotropy in small
angle neutron scattering. Phys. Rev. E, 1999, 60,
4330 – 4335.
38 Ford, I. J. Statistical mechanics of water droplet nuclea-
tion. Proc. Instn Mech. Engrs, Part C: J. Mechanical
Engineering Science, 2004, 218(C8), 883 –899.
39 Kashchiev, D. Solution of the non-steady state problem
in nucleation kinetics. Surf. Sci., 1969, 14, 209 – 220.
40 Courtney, W. G. Remarks on homogeneous nucleation.
J. Chem. Phys., 1961, 35, 2249 – 2250.
Proc. IMechE Vol. 219 Part C: J. Mechanical Engineering Science
41 Reiss, H., Kegel, W. K., and Katz, J. L. Resolution of the
problems of replacement free energy, 1/S, and internal
consistency by consideration of the length scale for
mixing entropy. Phys. Rev. Lett., 1997, 78, 4506 – 4509.
42 Girschik, S. L. and Chu, C.-P. Kinetic nucleation
theory: a new expression for the rate of homogeneous
nucleation from an ideal supersaturated vapour.
J. Chem. Phys., 1990, 93, 1273 – 1277.
43 Girschik, S. L. and Chu, C.-P. Comment. J. Chem. Phys.,
1991, 94, 826 – 827.
44 Band, W. An introduction to quantum statistics,
1955, Sec. 6.4 (Van Nostrand-Reinhold, Princeton,
New Jersey).
45 Saltz, D. Using the non-interacting cluster theory to
predict the properties of real vapour. J. Chem. Phys.,
1994, 101, 6038 – 6051.
46 Tolman, R. C. The effect of drop size on surface
tension. J. Chem. Phys., 1949, 17, 333 – 337.
47 Koening, F. O. On the thermodynamic relation between
surface and curvature. J. Chem. Phys., 1950, 18, 449.
48 Kirkwood, J. G., and Buff, E. P. The statistical mechanical
theory of surface tension. J. Chem. Phys., 1949, 17, 338.
49 Benson, G. C., and Shuttleworth, R. The surface energy
of small nuclei. J. Chem. Phys., 1951, 19, 130 – 131.
50 Plummer, P. L. M. and Hale, B. N. Molecular models
for prenucleation water clusters. J. Chem. Phys., 1972,
56, 4329 – 4434.
51 Sinangoglu, O. Microscopic surface tension down
to molecular dimensions and microthermodynamic
surface areas of molecules or clusters. J. Chem. Phys.,
1981, 75, 463 – 468.
52 Kegel, W. On the variation of the interfacial tension
with cluster size in connection to homogeneous
nucleation from the vapour phase. J. Chem. Phys.,
1995, 102, 1094 – 1095.
53 Alty, T. and MacKay, C. A. Accommodation coefficient
and the evaporation coefficient of water. Proc. R. Soc.,
1935, A14, 104.
54 Mills, A. F. and Seban, R. A. The condensation
coefficient of water. Int. J. Heat and Mass Transfer,
1967, 10, 1815 – 1827.
55 Mozurkewitch, M. Aerosol growth and the con-
densation coefficient for water: a review. Aerosol Sci.
Technol., 1986, 5, 223 –236.
56 Lothe, J. and Pound, G. M. Reconsiderations of nuclea-
tion theory. J. Chem. Phys., 1962, 36, 2080 – 2085.
57 Reiss, H., Katz, J. L., and Cohen, E. R. Translation –
rotation paradox in the theory of nucleation. J. Chem.
Phys., 1968, 48, 5553 –5560.
58 Hale, B. N. The scaling of nucleation rates. Metall.
Trans., 1992, 23A, 1863 – 1868.
59 Fisher, M. E. The theory of condensation and the
critical point. Physics, 1967, 3, 255 – 283.
60 Laaksonen, A., Ford, I. J., and Kulmala, M. Revised
parametrisation of the Dillman– Meier theory of
homogeneous nucleation. Phys. Rev. E, 1994, 49,
5517 – 5524.
61 Kantrowitz, A. Nucleation in very rapid vapour expan-
sions. J. Chem. Phys., 1951, 19, 1097 – 1100.
62 Feder, J., Russell, K. C., Lothe, J., and Pound, G. M.
Homogeneous nucleation and growth of droplets in
vapours. Adv. Physics, 1966, 15, 111 – 178.
C08404 # IMechE 2005
 Classical nucleation theory and its application
63 Wyslouzil, B. E. and Seinfeld, J. H. Nonisothermal homo-
geneous nucleation. J. Chem. Phys., 1992, 97, 2661–2669.
64 Viisanen, Y., Strey, R., and Reiss, H. Homogeneous
nucleation rates for water. J. Chem. Phys., 1993, 99,
4680 – 4692.
65 Peters, F. and Meyer, K. A. J. Measurement and
interpretation of growth of monodispersed water drop-
lets suspended in pure vapour. Int. J. Heat and Mass
Transfer, 1995, 38, 3285 – 3293.
66 Fisk, J. A. and Katz, J. L. Condensation of supersatu-
rated vapours. Pressure and non-ideal gas effects.
J. Chem. Phys., 1996, 104, 8649 – 8656.
67 Luijten, C. C. M., Baas, O. D. E., and van Dongen, M. E. H.
Homogeneous nucleation rates for n-pentanol from
expansion wave tube experiments. J. Chem. Phys., 1997,
106, 4152–4156.
68 Moore, M. J., Walters, P. T., Crane, R. I., and Davidson,
B. J. Predicting the fog drop size in wet steam turbines.
In Wet steam 4, 1973, paper C37/73, pp. 101 – 109 (Insti-
tution of Mechanical Engineers, Bury St Edmunds and
London).
69 Simpson, D. A. and White, A. J. Viscous and unsteady
flow calculations of condensing steam in nozzles.
Int. J. Heat and Fluid Flow, 2004, 26, 71 – 79.
70 Gyarmathy, G. On the growth rate of small liquid drop-
lets in a supersaturated atmosphere (in German).
Z. Angew. Math. Physics, 1963, 14, 280 – 293.
71 Gyarmathy, G. The spherical droplet in gaseous
carrier streams: review and synthesis. In Handbook of
chemistry and physics, 62nd edition, 1982, Multiphase
science and technology, Vol. 1, pp. 99 – 279 (McGraw-
Hill, New York).
72 Young, J. B. The condensation and evaporation of liquid
droplets in a pure vapour at arbitrary Knudsen number.
Int. J. Heat and Mass Transfer, 1991, 34, 1649– 1661.
73 Young, J. B. An equation of state for steam for turbo-
machinery and other flow calculations. Trans. ASME,
J. Engng Gas Turbines and Power, 1988, 110, 1– 7.
74 Schmidt, B. Beobachtungen über das Verhalten
der durch Wasserdampfkondensation ausgelösten
Störungen in einer Überschall-Windkaneldüse. PhD
Thesis, Universität Karlsruhe (TH), Germany, 1962.
75 Pouring, A. P. Thermal choking and condensation in
nozzles. Phys. Fluids, 1965, 8, 1802 –1810.
76 Barschdorff, D. Verlauf der Zustandsgrössen und
gas-dynamische Zuammenhänge bei der spontanen
Kondensation reinen Wasserdampfes in Lavaldüsen.
Forsch. Ing.-Wes., 1971, 37, 146 – 157.
77 Gyarmathy, G. Condensation shock diagrams for
steam flow. Forsch. Ing.-Wes., 1963, 29, 105.
78 Yousif, F. H., Campbell, B. A., and Bakhtar, F. Instabi-
lity in condensing flow of steam. Proc. Instn Mech.
Engrs, 1972, 186(37), 439 – 448.
79 Saltanov, G. A. Non-stationary flows of super-heated
and condensing steam (in Russian). Trans. Inst. Fluid
Flow Machinery (PIMP, Poland), 1976, 70– 72, 433 – 453.
80 Skillings, S. A. and Jackson, R. A robust time-
marching solver for one-dimensional nucleating steam
flows. Int. J. Heat and Fluid Flow, 1987, 8, 139 –144.
81 Moheban, M. A fast time-marching method for non-
equilibrium wet steam flows in turbines. PhD Thesis,
Engineering Department, Cambridge University, 1983.
C08404 # IMechE 2005
1331
82 Guha, A. and Young, J. B. Time-marching prediction
of unsteady condensation phenomena due to supercriti-
cal heat addition. In IMechE European Conference on
Turbomachinery, 1991, paper C423/057, pp. 167–177.
83 White, A. J. and Young, J. B. A time-marching method
for the prediction of two-dimensional unsteady flows of
condensing steam. Am. Inst. Aeronaut. Astronaut.,
J. Propulsion and Power, 1993, 9, 579 – 587.
84 Collignan, B. and Laali, A. R. Modelling of unsteady
steam flow with condensation: theoretical approach and
experimentation in a nuclear power plant. In Proceedings
of the Tenth International Conference on Steam and gas
turbines for power and co-generation plants, Karlovy
Vary, Czech Republic, 1994, pp. 37–44.
85 Adam, S. and Schnerr, G. H. Instabilities and bifur-
cation of non-equilibrium two-phase flow. J. Fluid
Mechanics, 1997, 348, 1 – 28.
86 Schnerr, G. H. Instationary flows.
87 Skillings, S. A., Moore, M. J., Walters, P. T., and Jackson, R.
A reconsideration of wetness loss in LP steam turbines.
In Proceedings of the BNES (British Nuclear Energy
Society) Conference on Technology of turbine plant operat-
ing with wet steam, London, 1988, pp. 171–177.
88 Deich, M. E., Tsiklauri, G. V., Shanin, V. K., and
Danilin, V. S. Investigation of flow of wet steam in
nozzles. High Temp., 1972, 10(1), 102.
89 Gyarmathy, G., Burkhard, H. P., Lesch, F., and
Siegenthaler, A. Spontaneous condensation of steam in
high pressure: first experimental results. In Proceedings
of IMechE Conference on Heat and fluid flow in steam
and gas turbine plant, 1973, paper C66/73, Vol. 3, pp.
182–186.
90 Valha, J. and Ryley, D. J. Optical studies of nucleation
in high pressure steam. In Proceedings of the ASME
Conference on Condensation in high speed flows
(Ed. A. A. Pouring), 1977, pp. 27– 42.
91 Bakhtar, F. and Zidi, K. Nucleation phenomena in
flowing high pressure steam: experimental results.
Proc. Instn Mech. Engrs, Part A: J. Power and Energy,
1989, 203, 195 – 200.
92 Keenan, J. H., Keyes, F. G., Hill, P. G., and Moore, J. G.
Steam tables, 1969 (John Wiley, Chichester).
93 Bakhtar, F. and Piran, M. Thermodynamic properties
of supercooled steam. Int. J. Heat and Fluid Flow,
1979, 1, 53 – 62.
94 Vukalovich, A. Thermodynamic properties of water and
steam, 1958, 6th edition (Mashgis, Moscow).
95 Bakhtar, F. and Zidi, K. On the self diffusion of water
vapour. Proc. Instn Mech. Engrs, Part C: J. Mechanical
Engineering Science, 1985, 199(C2), 159 – 164.
96 Bakhtar, F. and Webb, R. A. Tabular method of esti-
mating nucleating steam flows. Int. J. Heat and Fluid
Flow, 1983, 4, 217 – 228.
97 Huang, L. and Young, J. B. An analytical solution for
the Wilson point in homogeneously nucleating flows.
Proc. R. Soc., 1996, A452, 1459 – 1473.
APPENDIX 1
Notation
A Ag 2/3 ¼ 4pr 2 so that A3 ¼ 36p(m/rL )2
Proc. IMechE Vol. 219 Part C: J. Mechanical Engineering Science
1332 F Bakhtar, J B Young, A J White, and D A Simpson

1. Isothermal expansion (or compression) of the

Cg molecular rate of condensation on a g-mer
vapour from p to pS(TG)
Eg molecular rate of evaporation from a g-mer ð pS(T )
fg concentration of g-mers in supersaturated 4 3 G 1
DG1 ¼ pr rL dp
vapour 3 p r G
g number of molecules in a cluster  
G Gibbs function 4 p
¼  pr 3 rL RTG ln (28)
hfg specific enthalpy of evaporation 3 pS (TG )
I nucleation rate per unit volume
J nucleation rate per unit mass of vapour The second expression assumes perfect gas beha-
k Boltzmann’s constant viour. Real gas behaviour for high-pressure steam
m mass of one molecule can be modelled using a truncated virial series,
ng concentration of g-mers in superheated giving


vapour 4 r
p pressure DG1 ¼  pr 3 rL RTG ln G þ 2B(rGS  rG )
3 rGS
pS saturated vapour pressure 
3 4
q c, q e condensation and evaporation coefficients þ C(rGS  rG ) þ D(rGS  rG )
2 2 3 3
(29)
2 3
respectively
r droplet radius where rGS ¼ rGS(TG) and B, C, and D are the
r32 Sauter mean
Ð droplet
Ð second, third, and fourth virial coefficients
radius ¼ nr r 3 dr/ nr r 2 dr respectively.
R gas constant per unit mass 2. Condensation of vapour to bulk liquid at constant
S supersaturation ratio ¼ p/pS(TG) temperature and pressure pS(TG) gives
t time
T temperature DG2 ¼ 0 (30)
DT supercooling ¼ TS( p) 2 TG
Y wetness fraction
3. Isothermal compression of the liquid to pressure
a heat transfer coefficient p gives
g isentropic exponent of the vapour
P rate of expansion ¼ 2D(ln p)/Dt 4
DG3 ¼ pr 3 ½p  pS (TG ) (31)
r density 3
s surface tension
t characteristic timescale for transient 4. Formation of a droplet of radius r and surface area
nucleation 4pr 2 from bulk liquid gives
c non-isothermal correction factor to
nucleation rate DG4 ¼ 4pr 2 s (32)
where s is the surface free energy per unit area
(the surface tension). Equation (32) is the capillar-
Subscripts ity approximation and has been the source of
much controversy. The overall free energy
g g-mer change is DG ¼ DG1 þ DG2 þ DG3 þ DG4. Neglect-
G gas (vapour) ing DG3, which is small, gives equation (1) in the
L liquid main text if equation (28) is used and a more
r droplet of radius r complicated expression if equation (29) is used.
S saturated


critical size

APPENDIX 3

APPENDIX 2 Evaluation of the steady state nucleation rate

Derivation of the classical expression for DG Starting with equation (16), ng is substituted from
equation (6) to give
The change in Gibbs free energy during the for-
"ð #1
mation of a liquid droplet of radius r in a supersatu- 1
1
rated vapour held at constant p and T can be broken ICL ¼ dg (33)
g¼1 Cg n1 exp½DG=(kTG )
down into four steps as follows.

Proc. IMechE Vol. 219 Part C: J. Mechanical Engineering Science C08404 # IMechE 2005
 Classical nucleation theory and its application 1333

Cg is proportional to g 2/3, which is a slowly varying where u is related to the condensation heat release
function of g compared to the rapidly changing expo- via the energy balance
nential term that passes through a sharp maximum
at g ¼ g
. Cg is therefore approximated by Cg
. DG is 4 3
expanded as a Taylor series about g
as pr rL hfg Ig ¼ 4pr 2 ar ufg (39)
3


@DG where ar is the surface heat transfer coefficient for a
DG ¼ DG
þ (g  g
)
@g
droplet of radius r. Substituting equations (38) and
2
2  (39) into equation (37) to eliminate Egþ1(TG þ u)
(g  g
) @ DG
þ þ... (34) yields
2 @g 2

" rffiffiffiffiffiffiffiffiffi
#
Truncating after the second-order term and noting rG R RTG hfg hfg 1
Ig 1 þ qc 
that (@DG/@g)
¼ 0 gives ar 2p RTG RTG 2

 ¼ Cg (TG )fg  Egþ1 (TG )fgþ1 (40)
DG

ICL ¼ Cg
n1 Z exp  (35)
kTG
Defining a simplifying symbol n by
where Z is the so-called Zeldovich factor, given by rffiffiffiffiffiffiffiffiffi

rG R RTG hfg hfg 1
ð1 
2   n ¼ qc  (41)
1 @ DG (g  g
)2 ar 2p RTG RTG 2
¼ exp  dg
Z g¼1 @g 2
2kTG
 1=2 then allows equation (40) to be written as
(@2 DG/@g 2 )

ffi (36)
2pkTG
Ig (1 þ n) ¼ Cg (TG )fg  Egþ1 (TG )fgþ1 (42)
The second expression is obtained by extending
the lower limit of the integral to g ¼ 21. Evalua- The analysis now proceeds as before, yielding
ting (@2 DGg /@g 2 )
from equation (4) and substi-
tuting into equation (35) gives equation (17) in the INISO JNISO 1
¼ ¼ ¼c (43)
main text. IISO JISO 1þn

APPENDIX 4 At high pressure, the diameter of a critical droplet

may be comparable to the mean free path of a
vapour molecule and the expression used for ar
Non-isothermal nucleation theory
should allow for this possibility (see Bakhtar and
Kantrowitz [61] assumed that the droplet – vapour Zidi [36]). This also allows consistency between the
temperature difference u ¼ TL 2 TG was independent nucleation rate and droplet growth equations. At
of the droplet radius. The nucleation current is then low pressure, however, heat transfer from a critical
given by cluster will be in the free molecule regime and a suit-
able expression for ar can be shown to be
Ig ¼ Cg (TG )fg  Egþ1 (TG þ u)fgþ1 (37)
 
p gþ1
From equation (13) and the Clausius – Clapeyron ar ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi R (44)
2pRTG 2(g  1)
equation, it can be shown that

 which is independent of droplet radius. Substituting
u hfg equation (44) into equation (41) gives equation (27)
Eg (TG þ u) ffi 1 þ  0:5 Eg (TG ) (38)
TG RTG in the main text.

C08404 # IMechE 2005 Proc. IMechE Vol. 219 Part C: J. Mechanical Engineering Science