Thin Solid Films.

18 (1973) 91-98
0 Elsevier Sequoia S.A., Lausanne-Prihted in Switzerland 91

ON THE ATOMISTIC THEORY OF NUCLEATION RATE

S. STOYANOV
Institute of Physical Chemistry, Bulgarian Academy of Sciences, SoJia 13 (Bulgaria)
(Received March 7, 1973; accepted June 5, 1973)

The atomistic nucleation theory has been extended and generalized. A new
criterion has been proposed for a comparison of the theory with experimental
results. The data of Walton, Rhodin and Rollins as well as the data of Poppa
have been analysed.

1. INTRODUCTION

Walton’-3 has obtained expressions for the nucleation rate of deposits from
the vapour by using statistical mechanics and kinetic theory. He has defined the
nucleation rate as the rate at which clusters of critical size become supercritical.
In deriving expressions for the nucleation rate an equilibrium size distribution is
assumed to exist. Another assumption is that none of the clusters which are one
atom larger than critical can decay. The errors caused by these assumptions
cannot be estimated in the framework of the Walton theory. In addition, the
possibilities for verification of Walton’s model are strongly restricted, especially
as far as the temperature dependence of the nucleation rate is concerned. This
restriction is due to the lack of directly obtained values of some parameters, such
as the dissociation energy of the critical nucleus and the activation energies of
desorption and surface diffusion.
The purpose of this paper is to build up a nucleation rate theory free from
the assumptions mentioned above, and to ascertain an additional criterion for a
comparison with the experimental data.

2. NUCLEATION RATE

Let us consider the formation of a solid on a surface due to condensation from

a vapour. The phase transformation occurs through the adsorption of gas atoms
on the surface. The adsorbed atoms are assumed to form a two-dimensional
lattice gas in which interaction is limited to nearest neighbours. Clusters of two
or more atoms and single adsorbed atoms coexist on the surface (Fig. 1). These
clusters grow individually and decay because of the addition and detachment
of single atoms, with the result that some of them reach macroscopic dimensions.
The clusters are assumed to follow a definite sequence of atom configurations
during their growth (Fig. 2).
92 S. STOYANOV

., ..,
~..~.,.~..~...~..~

Fig. I. Adatoms and flat clusters on the substrate surface

Fig. 2. A sequence of cluster growth

Let the frequencies of single atom addition and detachment from a cluster
consisting of i atoms be denoted by o+i and 0-i respectively. These frequencies
may be written approximately as

o+~ = cli - v exp

NO (1)

where E; is the dissociation energy of a cluster of size i, I?& is the activation energy
for surface diffusion, v the atom vibrational frequency, N, the number of adsorp-
tion sites per square centimetre of the surface and N, the adatom concentration.
The exact number of ways of formation of a cluster of size i+ 1 by joining an
adatom to a cluster of size i is signified by cli. We have denoted by pi the number
of ways of single atom detachment from a cluster of size i. Since to every birth
process i+i+ 1 there corresponds a death process if l+i (Fig. 3) the following
equality connects CQand pi:

Equations (1) are not valid for clusters of size 5. 6,... and new expressions must be
written for w+~, CO+~,._.as well as for o_~, o_~,.... However. this turned out to
be unnecessary for our aim here.
ATOMISTIC THEORY OF NUCLEATION RATE 93

If the adatom concentration is held constant a kinetic steady state occurs after
a short induction period. The steady state rate of nucleation is given by the well-
known expression4

(3)

Fig. 3. Different possibilities for atom detachment and addition to the clusters.

where I is the size at which the clusters irreversibly leave the system.
We do not need any definition of a critical nucleus here because the steady
state rate43 5

ri = O+i_l Ni_ 1-O-i Ni = I

does not depend on i (Ni is the steady state concentration of clusters of size i).
Therefore we can consider I as the rate at which the largest invisible clusters
become visible. Defined in this way I is an experimentally observable quantity.
Substituting eqns. (1) and (2) into eqn. (3) we obtain

I=o+,Nl

\
Ii=
~ Mi
1
(4)

where

i= 1,2,3

W-2 0-3...O-i
Mi= i = 4, 5,...
w+2 o+3-.-“+i

Taking into account that E, = 0 and CQ= p2 one obtains M, = 1. The adatom
concentration may be written approximately as

where R is the incidence rate, z is the mean adatom stay time on the substrate and
Ed is the desorption energy. Substituting the last equation into eqn. (4) gives
94 S. STOYANOV

(5)

with

i= 1, 2, 3

O-2 O_J...W-i
Mi= i = 4, 5,...
O+2 o+3...w+i

It is easy to see that when any Mi is much larger than the others eqn. (5)
becomes the Walton equation for a critical nucleus of size i* = i. For example, for

M,>iM,
i=2

one obtains

(6)

which is identical to the Walton equation for i' = 1. All terms Mi increase with
a temperature rise or a decrease in incidence rate and both M2 and M, are the
first to become comparable with and then larger than M,. Taking into account
the relations
E, 5 2E, M,Zl@ E4 N 2E3

it can be concluded that M3 is the largest term in the denominator of eqn. (5)
in a given interval of experimental conditions. In this interval the nucleation
rate is given approximately by

I = cIJ R(&~exp(E3+4~~-Es”) (7)

which is the Walton equation for i* = 3.

According to Walton’-3 eqn. (6) is dominant for

whereas for

eqn. (7) will hold good ; the condition

M,=l
ATOMISTIC THEORY OF NUCLEATION RATE! 95

determines the transition temperature q. According to our treatment eqns. (6)

and (7) are valid at M, a 1 and Mz B 1 respectively. When both M, and M3 are
comparable with M, the nucleation rate I is given by

R
a1 R- ew
I=- v No
(8)
M,+M,+M,

Thus eqn. (5) smooths out the corners of the nucleation rate plot (Fig. 4).
Let us consider the cast when the cluster growth follows the sequence shown
in Fig. 5. The frequencies O+i and 0-i are now given by eqn. (1) for i = 1, 2,..., 7
and the nucleation rate is described by eqn. (5). The general expression (5) can
be reduced to eqn. (6) when the temperature is sufficiently low or the value of R
is sufficiently high, so that

MI>iMi
i=2

The term M, increases with rise in temperature (decrease in R) and its value
becomes comparable with M, in a given interval of experimental conditions. Then
the nucleation rate is given by
R
a1 R- ev
I=
v No
(9)
M,+M2

which differs from the Walton expressions.

At higher temperatures (lower atom arrival rates) M2 turns out to be the
largest term in the denominator and eqn. (5) can be rewritten as

Fig. 4. Schematic temperature dependence of the nucleation rate according to Walton (broken
line) and to eqn. (5) (full line).

0 0 00
00 00 00
00 oooo”ooooo
Ooooo
Fig. 5. Sequence of growth of flat clusters.
96 S. STOYANOV

which is identical with the Walton equation for i’ = 2.

If the binding energy I$ is an additive quantity a temperature T2 will exist at
which

Mi = M3 i = 4, 5, 6

and therefore

cL)+i= W-i i = 3, 4, 5, 6

In the Walton theory 5 is a transition temperature from i” = 2 to i” = 6 and the

nucleation rate at this temperature is given either by eqn. (10) or by

(11)

According to eqn. (5) in the neighbourhood of G the nucleation rate is given by

R
I=a,R-- exp Mi
vNo

and at G the rate I is four times smaller than the value predicted by both eqn. (10)
and eqn. (11). When

Mew 1 Mi
i= l,i#6

eqn. (5) turns into eqn. (11) for i” = 6.

Thus we have obtained the basic formulae of the atomistic theory, starting
from the classical Becker-Doting expression. Using this approach we have
established that the Walton equations are not valid near the transitions, i.e.
when two or more Mi are comparable. New expressions describing the nucleation
rate under such conditions have been derived.

3. ANALYSIS OF EXPERIMENTAL DATA

Walton et ~1.~ have investigated the nucleation of silver on sodium chloride.

They found that the nucleation rate is given by eqn. (6) at R = 10’3cm-2sec-’
and with

1.9<$<2.05

According to our considerations it can be concluded that under these conditions

all A$ (i = 2, 3,...) are much smaller than Ml = 1. The nucleation rate at R =
6 x 1013cm-2sec-’ and with
ATOMISTIC THEORY OF NUCLEATION RATE 97

has been found to satisfy eqn. (7) i.e. under these conditions M3 is much larger
than all other terms in the denominator of eqn. (5). In the framework of the
Walton treatment the agreement between the theory and experimental data is
good and the values

Ed = 0.4 eV E, = 2.1 eV
were obtained from the slope of nucleation rate plots. Our treatment, however,
suggests an additional possibility for comparison with experiment, namely, we
can estimate the values of M, at R, = 10’3cm-2sec-’ and 103/r1 = 1.95 (where
it has been established that i’ = 1) and at R2 = 6 x 10’3cm-2sec-’ and 103/T2 =
I .X5 (where i’ = 3). After simple calculations we obtain

M 3
(R

M, (R2, T,) =[$ $) 22.7x lO-29

This result shows that no agreement exists between the exact expression (5) and
the data of Walton et al. An agreement would exist if M,(R,,Tl)were at least
two orders of magnitude smaller than M3(R2, G).
As another example let us consider the experimental results given by Poppa7
for Bi deposits on amorphous carbon substrates. The nucleation rate at R =
1.7 x 1Ol4 falls by a factor of 10 between T' = 573 “K and T' = 590"K. Poppa
assumed that below T' a single atom is a critical nucleus, whereas above T' the
critical nucleus is of size z* = 3. This assumption permitted him to determine the
slope of the high temperature branch, although the nucleation rate has not been
measured at temperatures higher than T" = 590“K. Since eqn. (5) shows that
no corners exist on the nucleation rate plot the real slope at T" will be greater than
that determined by Poppa (Fig. 6(a)). Because of the lack of data at higher tem-
peratures it is not possible to decide which of eqns. (7) (Fig. 6(b)) and (8) (Fig. 6(c))
will describe the nucleation rate at T" = 590“K.

4. CONCLUSIONS

The work described in this paper leads to the following conclusions.

(1) The concept of a critical nucleus is not necessary for calculation of the
nucleation rate.
(2) When the frequency of single atom addition to a cluster of size i is
comparable with the frequency of detachment from this cluster the simple Walton
equations do not apply. The general equation (5) always applies and it smooths
out the corners of the nucleation rate plot.
(3) The data of Walton et al. do not satisfy the additional criterion derived
in this paper.
 S. STOYANOV

Fig. 6. Interpretation of the experimental results for the deposition of Bi on C substrates.

REFERENCES

1 D. Walton, J. Chem. Phys., 37 (1962) 2182.

2 D. Walton. in A. C. Zettlemoyer (ed.). Nucleation, Marcel Dekker, New York, 1969, p. 379.
3 T. N. Rhodin and D. Walton. Pror. Conf. on Single Crystal Films, Pmnsvlcania. May. 1963.
Pergamon Press, New York, 1964, p. 3 1.
4 R. Becker and W. Diiring, Ann. Phys., 24 (1935) 719.
5 V. Halpern, Brit. J. Appl. Phys., 18 (1967) 163.
6 D. Walton, T. N. Rhodin and R. W. Rollins, J. Chem. Phys., 38 (1963) 2698.
7 H. Poppa, J. Appl. Phys., 38 (1967) 3883.