View Article Online / Journal Homepage / Table of Contents for this issue

Possible Mechanism for the Spontaneous Rupture of Thin,

Free Liquid Films
Published on 01 January 1966. Downloaded by University of Vermont on 19/11/2017 16:35:49.

BY A. VRIJ

vant Hoff Laboratory, Sterrenbos 19, Utrecht, The Netherlands

Received 2nd March, 1966

The stability of a free, thin liquid film against small, spontaneous thickness fluctuations is explored.
The film is unstable with respect to fluctuations with wavelengths larger than a critical wavelength
A, = [-2-.2y/(d2V/dh2)]4, where y is the interfacial tension and V(h) the free energy of interaction
as a function of the film thickness h. V(h) may include van der Waals attraction and double-layer
repulsion. The kinetics of the growing fluctuations is obtained by assuming a laminar liquid flow
between rigid fiLm surfaces at a constant viscosity. There are stable fluctuation-modes, which grow
exponentially with time, each with a characteristic time constant T, and modes with certain
wavelengths grow faster than all others (T = T~). If the van der Waals forces predominate A, and
T, are given by eqn. (4.2) and (4.3) respectively. For A = 10-14-10-12 erg, y = 30 dynelcm and
Iz = 100-lo00 A, A, ranges from 0.6-600 p and T, from a fraction of a second to several hours. The
life-time and critical thickness h, of an unstable film are also calculated ; they depend on the time
constant rrn and on the time of draining. The critical thickness is calculated for microscopic,
circular films and compared with measurements of Scheludko and Exerowa. For water and
aniline films, the calculated h, are 410 and 750 A respectively,whereas the experimental values are
270 and 410 A.

Studies on the stability of thin, free liquid films are of importance for an under-
standing of colloidal systems, such as foams and emulsions.1-4
The life-time of liquid films is determined by two processes, thinning and break-
ing. Thinning of a film occurs by draining of the liquid under the influence of
gravity and suction at the Plateau-borders. When the thickness is reduced to
about 1000 A, other forces influence the draining ; van der Waals attraction increases
the draining rate ; double layer repulsion decreases it. On further thinning-which
often occurs discontinuously-some films become metastable ; others, however,
collapse between 500 and 100 A. Metastability is reached when border suction,
van der Waals attraction and double-layer (or other) repulsion equilibrate ; in-
stability results when the attraction forces predominate.1
Rupturing of metastable films, if it takes place, seems to occur in an irregular
manner because of lack of control over external disturbances such as thermal shocks,
vibration, dust, etc.3 Spontaneous rupturing through the formation of a nucleus
(hole) requires a high activation energy if the film thickness is larger than lOOA
as shown by de Vries 5 ; accordingly, such a process becomes highly improbable.
Only for extremely thin films (bimolecular leaflets) may this mechanism become
important.6
Unstable films (often called transient films), always rupture spontaneously,
often at a characteristic critical thickness . At present, however, it is still un-
known why the critical thickness is usually several lOOA instead of a few 10A
necessary for the spontaneous formation of a hole.
Scheludko 7 proposed that, at its critical thickness, the film becomes unstable
with respect to small, spontaneous surface deformations, such that, in spite of an
23
 View Article Online

24 RUPTURE OF LIQUID FILMS

increase of the free energy through the increase of surface area, the total free energy
decreases because of the van der Waals forces. He associated the wavelength A
of a certain surface deformation with a critical thickness hc, as follows :
Published on 01 January 1966. Downloaded by University of Vermont on 19/11/2017 16:35:49.

Here K is a constant proportional to the Hamaker constant : A = 6nK, and y the

interfacial tension. The theoretical value of A could not be calculated, however.
Further, he proposed that the sudden formation of black film from a thicker one
occurs by a similar mechanism.
That surface corrugations-due to thermal motion-exist was shown by the
author 8 from the light scattering a liquid film exhibits when placed in an intense
light beam. From the intensity of the light scattering and its angular dependence,
information on surface forces, operating in the liquid film, can be obtained.
In this paper the stability of these surface corrugations and the kinetics of their
diminution and amplification are explored. In an unstable film some of these surface
corrugations may grow until they rupture the film. The mechanism is analogous
to that proposed by Cahn9 fcr the phase separation by spinodal decomposition in
solutions.
FREE ENERGY OF T H E C O R R U G A T E D FILM
Consider a free, liquid film (medium 2), with a mean thickness ho, between two
media 1 and 3. The upper interface (1-2) is in the X-Y plane ; the lower inter-
face (2-3) in a plane at a distance tzg beneath it. The free energy increase AP associ-
ated with u certain, small corrugation 212 in the interface (1-2) and a corrugation
223, in the interface (2-3) is given by 8

AF = j J { + m 1 2 1 a x ) 2 +(az,,/aYY +(az231ax>2+(az2,1aY)21+
+(d2~/dh2)0(Zi2 -z2,)23dxdY, (2.1)
where y is the interfacial tension and Y(h) the free energy of interaction per unit
area. The integral is taken over the surface of a square, a2. The first term in eqn.
(2.1) is the work associated with the increase of surface area owing to the corruga-
tions; the second term is due to interaction forces such as electrostatic repulsion
and van der Waals attraction.
Eqn. (2.1) may conveniently be separated as
AF = AF, +AF,,

where we have substituted 2 1 2 - z 2 3 = h-ho. Thus, the corrugations of the two

single surfaces, 212 and 223, are replaced by two normal corrugations given by
+
the linear combinations (212 223) and (212 - 223). The combination (212 + Z Z ~ ) ,
associated with AFl, depicts fluctuations in the bending of the film as a whole,
whereas the combination (212 -2 2 9 , associated with A F 2 , depicts fluctuations in
tht=film thickness.
AFl is always positive when y>O. AFz, however, may become negative if
(d2V/dh2)o<O and the second term in eqn. (2.4) predominates. This means that
the film is always stable with respect to fluctuations of the former type, but may
 View Article Online

A . VRIJ 25
become unstable for some fluctuations of the latter. Because we are only interested
in fluctuations that make the film unstable, we restrict the further discussion to
fluctuations in 212-223 = h-ho.
To specify such fluctuations further, it is convenient to consider the Fourier
components of h(x,y) :
Published on 01 January 1966. Downloaded by University of Vermont on 19/11/2017 16:35:49.

h - h, = c c Hpuexp[ip(px+ OY),
4-00

p=-a
+a,

@=-m
(2.5)

where p L==2nla. Since lz is real the (complex) Fourier coefficients are connected
in pairs by the relations :
H-p,-G- - Hp,;
* Hp,-u = HTp,G;H-.p,o= Hp*,-=.
Each fourier component contributes independently to AF2. For instance, the
component Hpo yields the contribution,
n 'HP,H:,[+yp2(p2 + +
02) $(d2Vjdh2)o].

tVlien d2V/dh2<0, this component will decrease the free energy if

wh.ere

A, is the wavelength of the critical fluctuation. A fluctuation with A<A, will

" fluctuate " around a (meta)stable equilibrium. A fluctuation with A>&, how-
ever, will grow in amplitude, making the film unstable. Such a fluctuation can
Elways be found if the dimensions of the film (e.g., a) are larger than&. Thus,
a >A,,
or with eqn. (2.7),
(d2V/dh2)0< - 27c2y/a2

GROWTH OF THICKNESS FLUCTUATIONS

To obtain the kinetics of the fluctuation process, the mechanism of liquid transport
through the film must be specified. Although the fluctuations may be created through
a spontaneous process caused by thermal motion, their average decrease or increase
in time may be described by macroscopic laws if A is large with respect to molecular
dimensions. Furthermore, eqn. (2.1) is valid only near equilibrium so that the
equation to be derived will only describe the kinetics of the initial stage of the growing
process.
For simplicity, we assume a laminar liquid flow parallel to the surfaces of the
film (of uniform thickness ho) ; this assumption is reasonable if A 9 ho and h - ho <ho.
It is further assumed that the liquid in the film has a constant viscosity up to the
interfaces, and that no slip occurs at the interfaces.10 Then we may write for the
liquid flux per unit length,ll
Q = -(h:/12~) grad AP, (3.1)
where pl is the viscosity of the liquid and AAP is the extra pressure on the film due to
the fluctuation. For AP, one may write 12
AP = .AP,,+A(dV/dh). (3 -2)
Mere AP is the capillary pressure counteracting this fluctuation whereas A(d V/dh)
is due to intermolecular forces in the film. For small fluctuations one may write
for eqn. (3.2),
AP = -+y(a2h/ax2 + a 2 h / a y 2 ) + (d2V/dh2),(h - ho). (3.3)
 View Article Online

26 R U P T U R E O F LIQUID FILMS

To describe the dependence of h on time we need the law of conservation of

volume :
ahlat = -div Q. (3 04)
Combining eqn. (3.1), (3.3) and (3.4) yields the equation of motion :
Published on 01 January 1966. Downloaded by University of Vermont on 19/11/2017 16:35:49.

ah/& = (h;/12q)((d2 V/dh2)o(a2h/ax2 + a2h/ay2)-

+
+ ~ ( a ~ h i2a4h/a~2dy2
a~~ +
a4hpy4)}. (3.5)
This equation is formally equivalent to the equation of Cahn,g which describes
concentration fluctuations in solution as a function of tinie.
The solution of eqn. (3.5) is 9 :

with

and

Eqn. (3.6) is an extension of eqn. (2.5). It describes the thickness fluctuation as

the sum of independent stationary sinewaves, the amplitudes of which are exponential
functions of time. Thi: initial amplitudes and phases, represented by the (complex)
I$,&) are defined by the initial fluctuation at t = 0.
Substitution of the eqn. (2.7) and (2.8) into eqn. (3.8) shows that z<O when
A < & and that z>O when A>&. Thus, fluctuations of thc first kind (z<O) will
decrease exponentially with time and return to metastable equilibrium, whereas
 View Article Online

A . VRIJ 27
fluctuations of the second kind (z>O) will increase exponentially with time and move
a way from equilibrium.
z depends on k. It shows a sharp minimum at A, -- Ac23; see fig. 1. This
minimum appears because, when A is only slightly larger than Ac, the driving force,
and thus also the liquid flow speed, is small, whereas, when A is much larger than
Published on 01 January 1966. Downloaded by University of Vermont on 19/11/2017 16:35:49.

A, the time required to produce the fluctuation is large, because the liquid must be
transported over a larger distance. The minimum is equal to
, = (24yv/h;)(d2V/dh2)i2.
Z (3 -9)
Because the amplitudes of fluctuation modes have z in the exponent, the amplitude
of the modes with wavelength A m will grow faster than all others.
To visualize the thickness fluctuation (h-ho) of the film we refer to pictures
obtained by Cahn.9 He used a computor to calculate the Fourier sum with random
selections of directions, phases and amplitudes. One of these is redrawn and given
in fig. 2. It shows a pattern of interconnected hills and gullies with dimensions
of the order of A,. It seems reasonable to assume that the rupturing starts along
one of the gullies.

S T A B I L I T Y C O N D I T I O N S A N D TIME C O N S T A N T O F T H I C K N E S S
FLUCTUATIONS
To determine values for A m and zm, we need V(h) which is believed to contain
van der VCaals attraction and electrostatic repulsion terms. Graphs of V(h) repre-
senting the first two contributions have been discussed elsewhere.1 1 3 Schematic
plots of V(h) and also of d2V/dh2 are given in fig. 3(A, B).
The stability of the film against thickness fluctuations can now be read from
fig. 3B. According to eqn. (2.10), instability will occur if d2V/d1z2< - h 2 y / a 2 .
Large films (for instance, a = 1 cm) are practically unstable in the regions h<OP
and h > OQ. For smaller films the unstability region at h> OQ is reduced and for
the smallest ones it will disappear altogether. The instability region at h<OP,
however, will remain, although this does not imply that the film can reach such
thicknesses because of the potential barrier in V(h).
Let us digress on the important case where the van der Waals forces predominate
(at least in some region of h), and formulate V(h)as follows 1 (no retardation) :
V(h) = - A112nh2. (4.1)
A is the Hamaker constant. Using this equation to calculate d2V/dh2 and sub-
stituting the result into eqn. (2.7) yields
Ac = hg(4n3y/A)+. (4.2)
This equation is the equivalent of eqn. (1.1) of Scheludko. The numerical con-
stant, however, is different because of his approximate treatment.
A, as a function of ho is plotted in fig. 4 with y = 30 dynes/cm and with A =
10-14, 10-13 and 10-12 erg (plot 1, 2 and 3). Applying the stabiliiy condition
a>& (see eqn. (2.9)) it follows that films with n equal to about 0.1 mm should be
stable at least down to a thickness of about lOOOA. Much smaller films, for in-
stance, those separating oil globules in a creamed emulsion (say, a = IF), should
be stable even down to a thickness of about lOOA. For Zm one deduces from the
eqa. (3.9) and (4.1) :
, = 96n2yqh:A-2.
Z (4.3)
In fig. 5, Zm is plotted against ho with y = 30 dyneslcm ; = 0.01 poise and A =
10-14, 10-13 and 10-12 erg (plot 1, 2 and 3).
 View Article Online

28 R U P T U R E OF LIQUID FILMS
Published on 01 January 1966. Downloaded by University of Vermont on 19/11/2017 16:35:49.

FIG.3.--Schematical graphs of the free energy V(h)of interactionand its secondderivative,d2 Y(h)/dh2,
as a function of film thickness h. The stable and unstable regions are marked. y is the interfacial
tension and n the linear dimension of the film.

I I f
 View Article Online

A. VRIJ 29
To calculate the time needed for the fluctuations to increase to values about
equal to ha, i.e., the time needed to break the film, the initial values of the thickness

5
Published on 01 January 1966. Downloaded by University of Vermont on 19/11/2017 16:35:49.

10

lo"

Id
7
IQ

n
0 10
3
W

I 6'
-d
10

Id

l o1-0- ' I
30 100 300 1000

fluctuations have to be specified. In the draining film, i.e., a film in which lzo is a
function of t, the breaking time also depends on the draining rate. We consider
first the simpler case of a stationary film in which ho is constant with time.

B R E A K I N G TIME OF A S T A T I O N A R Y F I L M

The breaking time t b of an unstable film will depend on ho, zm and the initial
fluctuations of the film at t = 0 : i.e., the values of HpaH&(t = 0). The values of
these last quantities, however, are generally unknown and different or different
film samples. Therefore one may expect to have not just one value for t b , at a certain
ho, but a whole distribution of t b , because only one deep gully somewhere in the film
may lead to breaking. At present, we content ourselves with the calculation of an
approximate, mean vdue for t b , assuming that the initial fluctuations are generated
by thermal motion.
To this end let us suppose that the film will break when (h-h@ becomes of
the nrder nf h, where the;har means an averape over the film area. thus.

Substitution of eqn. (3.6) into the integral yields

hi C exp(2t~z)~,,,~~,(t
= 0).
P O
 View Article Online

30 R U P T U R E O F LIQUID F I L M S

This is a rough approximation because eqn. (3.6) was derived for fluctuations which
are small with respect to ho.
To obtain an approximate value of this sum we may take advantage of the fact
that for t$- only those terms closest to k = km, (k2 = p2(p2+02)), are important.
Therefore we replace HpaH&(t = 0) by their average value at km and use, instead
Published on 01 January 1966. Downloaded by University of Vermont on 19/11/2017 16:35:49.

of the full expression for z given by eqn. (3.8), its expansion around the minimum
z = zm:
117 = 1 / ~ , - ( 4 / k ~ ~ , J ( A k ) ~ , (5.3)
where Ak = k - k m , and replace the sum by an integral from Ak = - GO to + 00.
This yields
+a,

h i 2i (HH*),, exp(2tlzm) (2nkm/p2)exp -(8t/zmk?)(Ak)'d(Ak), (5 -4)

J-a,
where (HH*)om is written for H,,H,*,(t = 0 ; k = km). Furthermore, ( H H * ) m is
replaced by a mean value, to be obtained with the help of the principle of equi-
partition that states that the mean free energy of each fluctuation-mode, given by
eqn. (2.6), should be equal to kT'2; thus
HpbHp)li6= a-2[$yk2 +f(d2V/dh2)0]-1(kT/2). (5.5)
Application of this equation, at k = km, however, makes HPaHZq< O ; this is im-
possible : indeed, eqn. (5.5) becomes invalid at k < k,. To circumvent this difficulty
we assume that the interaction forces are " turned on " at t = 0, and drop the term
(d2Y/d@)o. Then ( H H * ) h becomes
(HH*),, = 2kT/ykia2. (5.6)
Substituting eqn. (5.6) into eqn. (5.4) and performing the integration then yields
h i E! (kT/2ynf)[exp(2 tlz,) Jl(2t/zm)*. (5.7)
Solving this equation for t yields an approximate value for t b . One finds that t b
is a small multiple of ,z :
tb = r m f ( b , y ) (5.8)
For instance,franges from 4.5 to 6-9for ho = 100 to 1000 A, using y = 30 dyneslcm.
After a time approximately equal to t b , the unstable film either ruptures or abruptly
decreases its thickness to form a thinner metastable film, depending on the shape
of V(h). This mechanism may explain the sudden formation of black film from a
much thicker one, which is often observed.

LIFE T I M E A N D CRITICAL THICKNESS O F A D R A I N I N G F I L M

The life time tl of a draining film will not only depend on the time of breaking
lb,but also on the time of draining. When the film is still thick the draining usually
proceeds rapidly and the thickness fluctuations grow slowly or not at all. When
the film becomes thinner, however, the rate of draining slows down, whereas the
stable fluctuations (if present) grow faster and faster until a critical thickness h, is
reached at which one of the fluctuations grows so fast that the film breaks.
The quantitative treatment of this process is complex because ho depends on t
and, in turn, Ac, A, and rm on ho. Where ho decreases with increasing t, there is
a corresponding decrease in A,, A m and zm. To take this into account a greater
number of terms should be included in the sum (5.2) than for the analysis of a
stationary film, where ho was constant and only the fastest growing fluctuation, at
that ho, had to be included. Therefore, a simplified, graphical treatment is pro-
posed which may be applied to films where the dependence of ho on t is relatively
 View Article Online

A . VRIJ 31
simple and known, e.g., the small circular films investigated by Scheludko 7.10. 12.14-16
and others.17-19
In fig. 6, the breaking time t b and the lime of draining t are plotted, schematically,
as a function of ho-within that range of ho for which modes with A m < a exist-
for an unstable film (fig. 6A) and for an unstable film thinning to a metastable one
Published on 01 January 1966. Downloaded by University of Vermont on 19/11/2017 16:35:49.

(fig. 6B). Also, t + t b is plotted as a function of ho (dashed curve). fb is obtained

from eqn. (5.8) and (3.9). For unstable films, eqn. (4.3) for zm may apply
1

FIG.6.-Schematical plot of the time t of draining and the time tb of breaking as a function of film
thickness ho. Part A : only van der Waals forces ; part B : both van der Waals forces and double-
layer forces. tl is the lifetime of the film,lz, and he the critical and equilibrium thicknesses ; h,, is
the thickness at the minimum in t + f b .

and, consequently t b will increase continuously with ho (see fig. 6A). For an un-
stable film thinning to an equilibrium film with thickness he, V(h) may be of the
form displayed in fig. 3A, with he near the minimum and the transient unstable
region at ho>OQ.
Now the life time of a draining film with a thickness ho is always smaller than
+
the corresponding value of t t b . For smaller ho, also t+ t b becomes smaller until
a minimum is reached at ho = hjn. This minimum value of t + t b is taken equal to
the life-time of the film; thus,
vtl = ( t +tb)min. (6.1)
is obtained with the relation
dtldh + dtb/dh = 0. (6.2)
The critical thickness h, corresponds to the thickness on the drainage line where
t = t l (see fig. 6). According to this procedure it should be considered a minimum
value. In fig. 6A rupturing takes place at h,; in fig. 6B the thickness suddenly
changes from h, to he, the equilibrium thickness.

TEST O F E Q U A T I O N S O N D A T A O F M I C R O S C O P I C FILMS
Scheludko and Exerowa 7 , 1 * , 14-16 have investigated the draining and rupturing
of circular microscopic aniline and water films in air. Some surfactant had been
added to prevent local thinning (Gibbs-Marangoni effect) in the earlier stages of
drainage and to allow the films to drain gradually and regularly with the opposing
 View Article Online

32 R U P T U R E OF LIQUID FILMS

interfaces nearly parallel until breaking occurs at a thickness of a few lOOA. Ac-
cording to Scheludko 12 the drainage cf the circular free films is governed by the
equation
dhi2/dt = o ~ P , (7.1)
where P is the pressure applied on the outside of the film m d
Published on 01 January 1966. Downloaded by University of Vermont on 19/11/2017 16:35:49.

a = 4/3yr2. (7.2)
r is the radius of the film and q the viscosity of the liquid. (For a further discussion
of these equations, see Frankel and Mysels 20 and Platikanov.21)
P is equal to
P = P()-II,-rl,,. (7.3)
PO is the capillary suction at the Plateau border, lIw is the contribution of the van
der Wads forces, given by
II, = -dV/dh = - A / 6 ~ h & (7.4)
and IIel is the electrostatic repulsion between the film surfaces. n,l may be neglected
if double layers are absent or if the electrolyte content is sufficiently high.
A useful approximate analytical expression for h, may be obtained if PO and
I I e 1 are neglected with respect to n,, which is a reasonable approximation for small
ho and high salt concentrations. It applies only to the last stage of the draining
process. Then one may write
dh,/dt 21 -A/9nqr2. (7.51
Further, dlbldh may be obtained from eqn. (5.8) and (4.3). Neglecting the dependence
off on ho and using (7.5) and (6.2) one obtains :
h, -N O.222(Ar2/fy)*. (74
According to this equation, h, does not depend on q . A more general analysis,
including the term PO,does not change this conclusion perceptibly. Experiments
on films with varying y would show if this conclusion is correct.
It is also of interest to calculate the approximate value of Ac at h,. Substituting
eqn. (7.6) into eqn. (4.2) yields
A,(h=h,)~0*55rf-'. (7.7)
Takingf = 7, one finds that A, N 0.21r.
Scheludko and Exerova measured the ho(t) curve for aniline films stabilized
with 0-5 % dodecylalcohol and obtained A N 7 x 10-12 erg (Po = 430 dynes cm-2 and
r = 10-2 cm). The experimentally obtained critical thickness of aniline films
stabilized with decylalcohol was about 420 A to within a small variation. Assuming
that the draining rates of films stabilized with dodecyl- and decyl-alcohol are equal,
h, is calculated to be 750 using the graphical procedure and y = 39.4 dyneslcm
and y = 0.044 poise. Eqn. (7.6) gives the more approximate value 890 A df = 7).
From the ho(t) curve for films of 0.1 M aqueous KCl solutions, stabilized with
5 x 10-4 % saponin the same authors obtained A N 10-12 erg (PO= 730 dynes cm-2 ;
r = 10-2 crn). Critical thicknesses of films containing 0.1 M KCl and stabilized
with propionic and butyric acid were also measured. They found hc" 270 A. The
distribution of values of hc, however, was broader than for aniline. Assuming that
the drainage of these films is the same for all stabilizers, and carrying out the same
procedure as for aniline with y = 65 dynelcm and q = 0.01 poise, we calculate
?I, = 410 A. Eqn. (7.6) yields h, = 485 A (f= 6.5).
The analysis shows that the experimental values of h, for aniline and water are
about 1.8 and 1.5 times smaller than the theoretical values. This result seems
reasonable considering the number of approximations that has to be made. The
 View Article Online

A . VRIJ 33
discrepancy may be due to additional stabilizing mechanisms not included in the
theory. Any film elasticity will damp fluctuations in h, especially if they tend to
become larger. Consequently, the thickness at which the film breaks is smaller
than that calculated here because then the van der Waals force outweigh those
arising from the Gibbs-Marangoni effect. Reducing the constant A of aniline by
Published on 01 January 1966. Downloaded by University of Vermont on 19/11/2017 16:35:49.

a factor of 8 and that of water by a factor of 3 would also remove the discrepancy.
An 8-fold difference for aniline is outside the limit of experimental error, although
the reported value of A seems high with respect to that of water. A 3-fold difference
for water, however, is still in the limit of accuracy of which this value is known.
The influence of the electrolyte concentration on the breaking process is marked.
Scheludko 7 found a continuous draining to the equilibrium thickness with KCl
concentrations lower than 10-2 M and a formation of black holes or rupturing at
concentration of about 10-1 M. These facts may be explained qualitatively from
fig. 6B : decreasing the salt concentration decreases 1 -d2l/ldh2[ and thus increases
i b . In the time corresponding to the minimum vdue of t + t b the film has already
drained continuously to a thickness close to the equilibrium thickness he, which
means that no sudden thinning to the equilibrium thickness occurs. There are
indications, however, that the stability of the film against black spot formation or
rupturing cannot be explained from the shape of the d2V/dh2 curve only.22
The theory given above has been applied to the relatively simple, small, circular
film. It should be, however, also a valuable starting point for the consideration
of more complicated systems such as macroscopic films in a foam, very small films
in a creamed emulsion, and films on a solid support.

The author thanks Prof. Dr. J. Th. G. Overbeek for his encouragement and advice,
and Dr. J. W. Vanderhoff for his assistance in the final preparation of the paper.
1 Overbeek, J. Physic. Chem., 1960,64, 1178.
2 Lyklema, Rec. trau. chim., 1962, 81, 890.
3 Kitchener, Recent Progress in Surface Science (AcademicPress, New York, 1964), vol. 1, chap. 2.
4 Mysels, J. Physic. Chem., 1964, 68, 3441.
5 de Vries, Rec. trau. chim., 1958, 77, 392.
6Derjaguin and Gutop, Koll. Zhur., 1962,24,431.
7 Scheludko, Proc. K. Akad. Wetensch. B, 1962, 65, 87.
8 Vrij, J. Colloid Sci., 1964, 19, 1.
9 Cahn, J. Chem. Physics, 1965,42,93.
10 Scheludko, Proc. K. Akad. Wetensch B, 1962, 65, 76.
11 Mysels, Shinoda and Frankel, Soap Films (Pergamon Press, London, 1959), 27.
12 Scheludko, Kolloid-Z., 1963, 191, 52.
13 Verweyand Overbeek, Theoryof the Stabilityof Lyophobic Colloids (Elsevier, Amsterdam, 1948),
p. 106.
14 Scheludko, Kolloid-Z., 1957, 155, 39.
15 Scheludko and Exerowa, Kolldd-Z., 1959, 165, 148.
16 Scheludko and Exerowa, Kolloid-Z., 1960, 168,24.
I7 Platikanov and Manev, Proc. 4th Int. Congr. Surface Actiuity (Brussels, 1964), preprint B/16.
18 Sonntag and Klare, KolZoid-Z., 1964, 195, 35.
19 Sonntag, Proc. 4th Int. Congr. Surface Activity (Brussels, 19641, preprint B/Vl 14.
20 Frankel and Mysels, J. Physic. Chem., 1962, 66, 190.
21 Platikanov, J. Physic. Chem., 1964, 68, 3619.
22 Scheludko, Exerowa and Platikanov, Koll. Zhur., 1963, 25, 606.

Note added in proof. Eqn. (7.9,obtained by neglecting Po and 1Ta with respect to IT, is a
poor approximation in the range of ho in which it is applied. Neglection of IIw and Eel with re-
spect to Po gives a better approximation. Then eqn. (7.6) becomes hcrtL0-268 (A2$/Pofy)f and
cqn. (7.7) h,(h =-. IrJ N O*80(ArsyJ/P$f4)i1r.