32 1

The initial motion of a gas bubble formed

in an inviscid liquid
Part 2. The three-dimensional bubble and the toroidal bubble

By J. K. WALTERS AND J. F. D A V I D S O N
Department of Chemical Engineering, Pembroke Street, Cambridge

(Received 19 November 1962 and in revised form 28 May 1963)

This paper deals with the initial motion of a gas bubble starting from rest in a
liquid in the form of a sphere. Part 1 (Walters & Davidson 1962) was concerned
with the similar problem of the initial motion of a two-dimensional bubble
starting from rest in the form of a cylinder.
Theory and experiments like those of Part 1are given for the present problem,
and yield qualitatively similar results, the three-dimensional bubble having an
initial acceleration equal to twice that of gravity, and distorting into the form of
a mushroom. This distortion ultimately causes break-up, but whereas the two-
diniensional bubble always detaches two -small bubbles a t its rear, the three-
dimensional bubble breaks up into a small spherical-cap bubble with a large
toroid below. A discussion of the toroidal bubble is given, and its relation to the
distorted sphere from which it is formed.
The initial-motion theory is extended to deal with the problem of the growing,
accelerating bubble, and leads to an expression for the volume of bubbles formed
continuously a t an orifice, and to a criterion for the gas flow-rate a t which
coalescence occurs between successive bubbles. These theoretical results are
compared with experimental data from the literature and from the authors’
experiments a t high gas flow-rates.

1. Introduction
The analysis here presented follows the same form as for the two-dimensional
case discussed in Part 1, which dealt with the initial motion of a cylindrical
bubble. The essential differences are ( a )the replacement of the trigonometrical
functions by Legendre polynomials on account of the axially symmetric form of
the velocity potential, and ( b ) the extension to the case of a growing bubble.
This latter theory is applied to the case of continuous bubble formation a t
constant gas flow-rates from a submerged orifice, and gives an expression for the
volume of bubbles as they detach. A criterion is also developed for the flow-rate
above which double-bubble formation may be expected to occur. Cine pictures
indicate that this is substantially correct, and also show that the back of the
bubble moves downwards a t the start, thus justifying the assumption of Davidson
& Xchiiler (1960) that the orifice is initially at the centre of the bubble.
21 Fluid Mech. 17
322 J . K . Walters and J . F . Davidson
The first part of the analysis also provides a way of calculating the circulation
associated with a toroidal bubble. Such bubbles were produced by injecting a
quantity of air into the bottom of a tank of water, and the observations show
that they move with the velocity of a vortex ring whose core has the same dimen-
sions as the bubble. The results also show that the circulation associated with the
vortex ring can be calculated from the above theory of distortion of a spherical
bubble.

2. The theory of the initial motion

For the case of an initially spherical bubble the velocity potential q5 is expressed
in terms of a series of Legendre polynomials P,) thus

Here a is the initial radius of the bubble, r and 8 are polar co-ordinates whose
origin moves with the bubble and has an upward velocity U at time t after the
start. The coefficients Pn are functions of time, and have to be adjusted so that the
pressure within the bubble shall be independent of 8. The first term in ( 1 )repre-
sents the potential due to a sphere, p1equalling Qua3,and the remaining terms
describe the subsequent distortion of the bubble. The pressure p in the liquid is
found from Bernoulli’s theorem, as in Part 1 )

where @/at is the partial derivative at a fixed point in space, q is the absolute
velocity and K is constant, since the pressure at infinity is presumed fixed.
Following the method of Lamb (1932))as in Part 1 , a#/at is calculated in terms
+
of = - U cos 8 and 8 = U sin 8 / r , the rates of change of r and 8 for a point fixed
in space, so that from ( l ) ,using recursion formulae for the P, (Magnus & Ober-
hettinger 1949, p. 5 0 ) )

Also, q2 = + (a$/a8)2/r2and again from ( 1 ) )using an expression for P, P,

given by Whittaker & Watson (1927, p. 331))

The pressurep, just outside the bubble is calculated, as in Part 1, by substituting

(3) and (4) into ( 2 ) with r = R,where R is the radius vector from the moving
origin of ( r ,8) to the surface of the bubble. R is then put equal to a(1 + c),
where 5, a dimensionless function of t , 8, and a ) describes the shape of the bubble.
Assuming 5 < 1 we get, from (2)-(4),
 The initial motion of a gas bubble formed in an inviscid liquid. Part 2 323
As in Past 1 , the radius a of the bubble is assumed constant and surface tension is
neglected, so that p s is equal to the pressure within the bubble. This must be
made independent of 0 and equating the coefficients of Pn to zero gives the
following equations for the pn:

These are the equations corresponding to (6) and ( 7 )in Part 1 for the two-dimen-
sional bubble. In deriving (6)-(8)9second-order terms in ( 3 ) - ( 5 )were neglected;
the effects of these terms, and of similar approximations below, are discussed in
the Appendix to this paper. Proceeding as before by ignoring the last term in
each of (6)-( 8))the first approximation becomes
n = 1, 2ppla3 = U(1' = 2g t , (9)

The shape of the bubble is calculated by equating the two expressions,

( - a$lar),.=,
and ( a [ + U cos 0)) for the radial velocity at the bubble surface,
giving

the O ( 5 )terms from (1)having been retained, as in Part 1 . On substituting from

(9) and (10) into ( 1 1 ) we see that the summed terms become polynomials in
N = gtZ/a; by equating coefficients of powers o f t and using expressions for Pn
given by Jahnke & Emde (1945) we obtain a series solution, the first few terms of
which are
6 = - 0.750N2P2+ N3(O-93P3+ 0*3P1)- N4(1.3iP4+ 0*77P2+ 0.3). (12)
The Appendix to this paper gives a derivation of a second approximation for 5.
This result, ( 3 7 ) )shows that the first term in ( 1 2 )is correct, the N3term is changed
by a small amount. and the N4 term is appreciably changed; but ( 1 2 ) is adequate
for N < 0.5, and roughly correct for somewhat larger values of N .

3. The formation of the tongue of liquid

The tongue of liquid begins to form at the back of the bubble when the curva-
ture of the surface at that point changes sign. I n polar co-ordinates, the curva-
ture is
R2 + 2(dRld0)2- R d2Rld02
[A2+ (dRldO)2]g
3

and hence at the back of the bubble where 0 = rr and dRld0 = 0, the curvature
changes sign when d2RldO2 = R . Substituting from ( 1 2 ) into this equation, with
R = a( 1+ 5) and using the first three terms, then gives N = 0.36 as the point when
21-2
324 J . K . Walters and J . F . Davidson
the tongue should begin to form. This value of N corresponds to frame 6 in
figure 1, plate 1, and the tongue can be seen in the following frames as a dark
patch rising from the back of the bubble.

4. Experiments with a three-dimensional bubble

It was not easy to produce a stationary spherical bubble in a stagnant liquid
and several methods were tried, The most successful was to inflate a rubber
balloon almost to its breaking point and then to burst it with a pin in a closed
vessel completely filled with water, the rigidity of the vessel keeping the bubble at
I - 1 I I I I I I 1 I I I I 01
0 V I
0 0 .

"/." Bubble volumes (ml )

I- 0 /
;
A
*
v 110
0 200
0.4 A 240

0 280
x 1600
0 2200
0.2
+ 3000

I I 1 I I I I I I 1 I I
0 0.2 0.4 0.6 0.8 1.0 1-2 1-4
N* = t ( g / a ) +
FIGURE
3. The upward displacement o f a spherical bubble starting from rest.

a constant volume. The balloon was attached to a brass tube, inflated, and then
clamped to a portable framework. Lead weights were used to overcome the
buoyancy and hold the framework a t the bottom of an 18in. square tank about
3 ft. high made of Perspex $in. thick. The lid was then screwed down and the
vessel filled with water, all air being carefully expelled through a vent in the top
corner. The bubble was released by bursting the balloon with a pin attached to a
lever arm.
The subsequent motion of the bubble was observed photographically either
with a Fastax camera at about 1000 frames/sec or with a Path6 camera at
80 frames/sec, and two typical bubbles are shown in figure 1, plate 1 and figure 2,
plate 2. A millisecond time base was provided on the film for the Fastax camera
while a 200r.p.m. clock was included in the field of view for the Path6 camera
 The initial motion of a gas bubble formed in an inviscid liquid. Part 2 325
The film was analysed frame by frame by measuring the projected image of
the bubble. The centroid could not be determined precisely, but since the initial
distortion predicted by the first term in (12) is into an ellipsoid of revolution, it
seemed reasonable to take the average position of the front and back of the bubble
for comparison in figure 3. The vertical diameter was also measured, the tongue
of liquid projecting upwards from the rear being visible through the bubble, and
the results are presented in figure 4.

1.0

0.8

0.6
.8
F.l

0.4
z
40 +

40
Bubble volumes
0
0.2

0 0.2 0.4 0.6 0.8 1.o 1.2

N* = t(g/a)*
FIGURE
4. The change in vertical diameter of a spherical bubble starting from rest.

5. Comparison between theory and experiment

The balloon burst by forming a circumferential split from the point of puncture,
after which the rubber peeled back along the air-water interface to leave the
surface of the bubble with a crinkly appearance rather like frosted glass. This
made it difficult to see the tongue of liquid that comes up from the bottom of the
bubble, but it is more easily seen in the projected cine film than in the prints in
figures 1 and 2. As in the two-dimensional case, the theory, (12), predicts the
formation of such a tongue of liquid, but with the change taking place rather
more rapidly. Davidson & Schuler (1960) show photographs of bubbles forming
at an orifice, the same type of tongue being produced just as the bubble is moving
away from the orifice.
The displacement of the bubble sg is compared with theory, (9), in figure 3.
It is not obvious that the movement of the centre of co-ordinates from (9) is the
same as that of the centre of gravity of the bubble; but in Part 1 it was shown, by
326 J . K . Walters and J. F . Davidson
using the second approximation, that the two centres do in fact move together
and the same is true of a three-dimensional bubble (see Appendix). The zero of
time was taken as the point when the pin punctured the balloon, but the with-
drawal of the rubber took about 2 or 3msec and this is the magnitude of the
difference between the theoretical line and the experimental points. There is
excellent agreement up to a value of N* of about 0.9 showing that the bubble does
in fact start to move with an acceleration of 2g.
The photographs in figures 1 and 2 show that the balloon is approximately
spherical at the start, but not precisely so. Two point support was tried but it
proved unsatisfactory as the balloon did not then burst in a reproducible manner.
Also, although the vessel was completely filled with water, the Perspex was
sufficiently elastic to allow a slight flexing of the walls when the excess pressure
was released by bursting the balloon. This caused about three initial pulsations
of the bubble with period about 9 msec, but the subsequent changes of shape are
in reasonable agreement with (12). The value for a was taken as the mean vertical
radius during the second pulsation, because of the slight overall expansion at the
start; these effects were almost negligible with the larger bubbles, but it is
interesting to note that the natural period of pulsation for a 200ml bubble in an
infinite volume of water can be shown (Walters 1962)to be about 9.5 msec.
Figure 4 shows observations of the vertical height D measured on the axis of
symmetry of the bubble, together with the theoretical line from (12) which is
D/2a = Q(2 + c,,+ 6 )= 1- 0.75N2- 2.38N4.
This equation is likely to be seriously in error when N > 0.5 owing to the approxi-
mations of the small deformation theory, discussed in detail in the Appendix.
However, it is interesting to note that (13) becomes zero when N* = 0.845 and
figures 3 and 4 indicate that a change in behaviour does occur in this region, and
in the subsequent motion a small spherical-cap bubble is formed, with a toroidal
bubble at the centre of a ring vortex in its wake. The circulation generated can
clearly be seen on the projected cine film, which also shows that most of the
volume is contained in the toroid which, however, soon breaks up. These changes
can be seen in the photographs in figure 2 and the formation of the tongue of
liquid in the earlier stages is seen in figure 1, frames 7 to 11. Theoretically this
tongue should begin to form when the curvature at the back of the bubble changes
sign and, as shown above, this occurs at N* = 0.6. This point corresponds approxi-
mately to frame 6 in figure 1 and the subsequent frames in fact show the tongue,
so the agreement is good. It is very difficult to locate the back of the bubble when
the tongue is beginning to form and this accounts for the lack of experimental
points around the ordinate value of 0.7 in figure 4.
The change from the irrotational motion in the initial stages to the fully
separated flow around the spherical-cap bubble takes place in the same manner
as in the two-dimensional case, and is caused by the tongue of liquid rising from
the rear. Once again the vorticity associated with the separated flow is produced
by the break-up of the bubble, rather than by viscous forces as in the case of the
wake behind a solid object. The difference between the two- and three-dimen-
sional bubbles is quantitative rather than qualitative; the volume of gas retained
 T h e initial motion of a gas bubble formed in an inviscid liquid. Part 2 327
by the spherical-cap bubble is much less in the three-dimensional case, and may
be zero as described in $ 10; and the detached toroid is of course a much larger
proportion of the original bubble than the two detached bubbles shown in figure 2
of Part 1 in the two-dimensional case.

6. The growing three-dimensional bubble

For an expanding sphere the velocity potential is a2a/r and so for a growing
bubble we take

where Po = a2a,and p1= QUa3 as before. The analysis is essentially the same as
that presented above in $ 2 but with the addition to each equation of terms
associated with Po. Equation ( 3 ) remains of the same form but with the sum-
mation beginning a t n = 0 rather than n = 1, but ( 4 ) has the additional terms

Substituting the new equations for a$/at and q2into Bernoulli's equation ( 2 )then
gives the pressure p s a t the surface of the bubble. This gives an equation like ( 5 )
but with the following terms added to the right-hand side

The pressure ps is again made independent of 19by equating the coefficients of

P, to zero, giving the following equations for the p,,
n = 1, P1-ga3-$Up2/az = 0,
n = 2, P 2 - 3 a p 2 1 a + ~ U P 1 - ~ U p , / a 2= 0,

Because the bubble is expanding, its radius a is not constant and will be given by

Gt = 4na3/3, so that G = 47ra2ci, (18)

where G is the volumetric flow rate of gas into the bubble. The simplest case to
consider is when G is constant, and substituting from ( 1 8 ) into (15)-(17), and
integrating, gives for the first approximation

n = 1 , 2PJl)/a3= u(1)= 94 (19)

( - l)n-l32n (n - l ) !( n ! ) 2 G
n' 2' "'= 8n(2n)!9.14...(5n-l)
~ (gt2)".

These are obtained by omitting the last term in each of (15), (16) and (17) and
using repeated integration n times with the integrating factor t-)("+l)each time.
328 J . K . Walters and J . P.Davidson
The shape of the bubble is calculated in the same way as before by equating
the two expressions, - a#/& and u [ + u( 1 + <) + U cos 8, for the radial velocity at
the bubble surface, giving

Substituting from (19) and (20) and solving as before gives

6= -O~173N2P2+N3(0~116P~+0~346Pl)
-N4(0*0863P4+ 0*0477P2+0.0181).
It is probable that the N4term is inaccurate because of error terms like those
discussed in the Appendix. Comparing (22) and ( 1 2) the general behaviour is seen
to be the same as for the non-expanding bubble, but the expansion term causes
the changes of shape to occur much more slowly. It is also interesting that (19)
indicates an upward acceleration of g in contrast to that of 3g for the non-
expanding bubble. This difference is explained by the simple theory of Davidson
& Schuler (1960) who assumed that the forming bubble is always spherical. On
this basis the rate of change of upward momentum is d($pVU)/dt, V being the
volume and p the liquid density. The bubble can therefore acquire momentum
on account of growth as well as on account of acceleration, so the latter must be
smaller for a growing bubble.
Since G has been assumed fixed, the above theory is applicable to the case of
continuous bubble formation from an orifice to which the gas is supplied a t a
constant volumetric flow-rate. Following the model put forward by Davidson &
Schuler (1960) we imagine the gas to be supplied from a fixed-point source within
the liquid, so that ideally the bubble will detach itself when i t has risen a distance
such that its back is level with the source, when

Here Cn is the shape factor at the back of the bubble where 8 = n. Substituting
from (19) and (22) gives an equation for N , the solution of which is N = 1.013.
Combining this with the first of (18) then gives for the bubble volume
V = 0*76g-%G8. (23)
This equation is of the same form as Davidson & Schuler’s (1960) result but with
a different numerical constant. Using their theory but with the effective inertia
of the bubble equal to $pV, it is easily shown that the constant is 1.138. It is quite
certain that the multiplying constant in (13) should be less than 1.138, because
the distortion of the bubble makes it detach earlier from the orifice than if it
had remained spherical. But the value of 0.76 is in error because
( a )the use of a finite number of terms in (22) leads to an overestimate of N a t
bubble detachment, and
( b ) the use of the first approximation in the derivation of (22) leads to errors
like those discussed in the Appendix, but less serious for the following reason:
in the present case we are dealing with the detachment of a bubble from an
 T h e initial motion of a gas bubble formed in an inviscid liquid. Part 2 329
orifice, and at the instant of detachment the radial distortion is of the same order
as the bubble radius; in the constant volume case discussed in the Appendix we
are interested in the point of bubble break-up, when the radial distortion is of
the same order as the bubble diameter.
The effect of these errors is hard to estimate, but ( 2 3 ) is probably correct to
10 yo.

7. Multiple bubble formation

Coalescence of bubbles forming at an orifice has been reported frequently in the
literature at the larger flow rates, and by an extension of the above theory it is
possible to estimate the flow-rate G a t which this may be expected to occur.
We shall take as the criterion for this multiple bubble formation the condition
when the velocity v, of the rear surface of the bubble that has just left the orifice
is equal to the velocity vo of the front of the next bubble.
The velocity v, can be calculated from the above theory, but difficulty arises
because of divergence of the series for near N = 1. Because of this difficulty, the
theory of Davidson & Schuler is used to calculate the upward velocity of the
bubble at detachment, which is U = 1 * 1 3 8 g ~ Gusing ~, an effective inertia
of +pv.
To find the velocity vo of the front of the next bubble we assume that the
preceding bubble leaves a small nucleus of air at the orifice, the radius of which
is equal to that of the orifice, a,, so that
u0 = G/~;.U;,
then the condition vo 2 U for coalescence to occur gives as the criterion for
multiple bubble formation
G 3 28ghi. (24)

8. Experiments on continuous bubble formation

All previous workers have confined their attention to bubble formation at
small gas flow-rates, so the present work was performed in the range 600 ml/sec
to 10l/sec. Air from the main was metered through a rotameter and then led
through a 4in. diameter copper tube to the bottom of an open Perspex tank 18in.
square and 3ft. high which was filled with water. Three different orifices were
used, &in., gin. and lin. in diameter and the bubble formation was observed
photographically with the Path6 camera set at 80 frames/sec, a clock rotating at
200r.p.m. also being included in the field of view to give an accurate time scale.
The bubble volumes were determined by two methods, first, from the frequency
obtained by counting bubbles, and secondly from the mean dimensions of
several bubbles measured from the projected film. I n the latter case the hori-
zontal and vertical diameters of each bubble were measured and the volume
found by assuming the bubble to be an ellipsoid of revolution about the vertical
axis. The photographs in figure 6, plate 3, show considerable variation in bubble
shape, but the two methods are in good agreement, and the results are shown in
figure 5.
330 J. I<. Wulters and J. F. Duvidson
9. Discussion
Equation (23) for the bubble volume is of the same form as Davidson &
Schiiler (1960) obtained by equating the buoyancy force to the rate of change of
upward momentum, assuming the bubble remained spherical throughout its
formation. They took the value 1lp V /16 for the virtual mass of the bubble and
obtained 1-378 as the coefficient in (23). This value of the virtual mass is appro-
priate to a sphere moving perpendicularly near to a wall, and so the result is more

V = 1.138 GS/g+.

1.o 10 100 1000 10 000

G (mlsec-l)
FIGURE 5. Bubble volume as a function of gas flow-rate. Present work, air-water : 0 0 ,;t- in.
diameter tube; 0 m, 2 in. diameter tube; A A , 1in. diameter tube; bubble volume from
counting for open points, measurement for solid points. Calderbank (1956) air-water:
+, 0.265cm capillary, x , slots & to +in. wide. Van Krevelen & Hoftijzer (1950): T,
0.23 cm capillary, air-water; 1,H,-water. Davidson & h i c k (1956): V, 0.48 cm diameter
orifice, air-water. Davidson & Schiiler (1960): ---, 0.15 to 0.25 cm diameter orifices, air-
water.
applicable to bubbling from a hole in a plate. The virtual mass associated with a
sphere in an unbounded liquid is 4pPV and this gives a coefficient of 1.138.
Figure 5 shows that (23) with this coefficient gives an excellent fit over the whole
range of gas flow-rates, and it is of interest to note that van Krevelen & Hoftijzer
 The initial notion of a gas bubble formed in a n inviscid liquid. Part 2 331
(1950) also obtained this form of equation although their theoretical approach
was somewhat different. They obtained a coefficient of 1.72. It would thus seem
that the changes of shape predicted by the present theory are inhibited to such
an extent by the presence of the orifice that Davidson & Schuler's assumption,
that the bubble remains spherical, is well justified. The tongue of liquid a t the
back of the bubble does appear, however, after the bubble has left the orifice.
It can be clearly seen in the photographs of Davidson & Schuler, and good
pictures are also given by Helsby & Tuson (1955).
The criterion (24) for multiple bubble formation gives values of the gas flow
for the three orifices used of 282, 776 and 1600ml/sec, respectively, and figure 6,
plate 3, shows cine photographs of bubbles forming a t the 1in. orifice for four
flow-rates above the critical value. These indicate that there is a gradual transi-
tion rather than any sharp change in behaviour; a t low flows coalescence occurs
some distance above the orifice and as the flow is increased, the coalescence point
moves downwards. Good descriptions of coalescence are given by Helsby &
Tuson (1955) and also by Davidson & Amick (1956). These latter authors give
30-40ml/sec as the flow-rate for coalescence at an orifice of radius 0.16 cm, and
(24) gives 9ml/sec as the critical value. Davidson & Schuler's photographs with
G = 13.7 ml/sec and a, = 0.2 cm show coalescence occurring about two or three
bubble diameters above the orifice and (24) gives G = 16ml/sec for coalescence
at the orifice itself. Thus although there is no sharp change, (24) does give the
right order of magnitude for the transition over a wide range of gas flow-rates.
While it is not claimed to be a rigorous result, it may well be useful as a design
criterion for the upper limit of steady bubbling from a given orifice.

10. The toroidal bubble

During the preliminary experiments designed to produce the spherical bubble
described in 9 4, ring-shaped or toroidal bubbles were sometimes observed.
However, a more reproducible method of forming toroidal bubbles was found
to be the injection of a single pulse of air through a small tube inserted into the
bottom of a large Perspex tank filled with water. The air was previously con-
tained in a small vessel and let into the tank by quickly opening and closing a
valve, different injection times and vessel pressures producing different bubble
volumes. On leaving the orifice the bubble was moving with a large velocity, but
after it had travelled a short distance the front surface slowed down, and was
penetrated by the tongue of liquid from the back, thus forming a toroidal bubble.
The plane of the ring was perpendicular to the direction of motion, and as the
bubble moved steadily upwards its velocity decreased and the ring diameter
increased. Cine photographs of each bubble were taken with the Path6 camera
a t about 80 frameslsec, and figure 7, plate 4, shows three pictures taken a t about
a sec intervals. The lower half of each frame shows the elevation through the side
of the tank, whilst the top portions show the plan view obtained by reflexion in a
mirror supported a t 45" above the tank. I n the third picture the bubble is about
to break the surface, which is clearly seen erupting some distance ahead of the
bubble. The filmed record of each toroidal bubble was analysed frame by frame,
332 J. K . Waiters and J. F. Dauidson
enabling the ring dimensions and displacement us time curves to be obtained for
each bubble. The gradient of these curves then gave the bubble velocity at each
instant, and the bubble volume was calculated from the ring dimensions.
Taking the bubble to be the core of a ring vortex, the velocity u is given by

where d is the ring diameter, h the core diameter, and I? the associated circulation.
Figure 8 shows ud/[ln ( 8 d j h )- 41 plotted us time for three bubbles and indicates
I I I 1 I I
120 - -
0
a
a n a n o n o O o n

100 -
I 0 0 a
0
-
0 0 0
,ik
80 - A A -
A A A A A A A A
‘’ A A A A A A A A A A A A A
60 - -
0
O O O O 0 0 O 0 0

40 -
o o 0
O O -

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7

Time (sec)
FIGURE
8. Circulation ws time for three horoidal bubbles.
Bubble volumes: 0 , 21 ml; A , 43 ml; , 88 ml.

that r remains substantially constant as the bubble rises. Thus it is the injection
of the bubble and its subsequent changes of shape that cause the growth of the
circulation. Once the toroid has been formed the results in figure 8 show that
there is no further increase in I?. We may obtain an estimate of the order of
magnitude of this circulation by assuming that the bubble is initially spherical
and that it moves according to the a,nalysisin § 2. Referring to figure 9, J”q.ds
from A to E is calculated by integrating round the path ABCDE. We then
assume that the bubble turns into a toroid by the liquid at E making contact
with the liquid at A, and that J”q.ds, defined above, is equal to the circulation
associated with the resulting toroid. Hence

where $E is the velocity potential at the rear of the bubble (8 =).; and $xi that
at the front (8 = 0 ) . Substituting from (1) putting r = a(1 + 5) then gives
 The initial motion of u gas bubble formed in an inviscid liquid. Part 2 333
and further substitutions from (9), (10) and (12) give, keeping terms up to N 4 ,
r =0.98(~~~)t(i+2.2~2+5.8~4). (26)
The value of I' associated with the toroidal bubble is given by inserting the value
of N at which the toroid is formed, when the tongue of liquid from the base of the

k-.,
I \
I \

I /
/
I /
I /
/
P-
FIGURE
9. Accelerating bubble on the point of forming
into a toroidal bubble.

V+(rnl)+
F I G ~10.
E Circulation wus (volume)*for toroidal bubbles.
334 J . K . Walters and J . F . Davidson
bubble makes contact with the top. Although a t this stage the deviations of the
bubble shape from sphericity are clearly not small, and the approximations of
Q 2 are therefore invalid, it is worth while to use the small deformation theory
to get an order of magnitude for r. Accordingly we solve (13) with D = 0,
giving N = 0.715 and substitute this value into (26), giving

r = Q~+V+. (27)
This is clearly only a rough approximation, since the behaviour of the three
terms in (26) suggests that they are part of a divergent series, and this is in
accordance with the fact that the deviations from sphericity are large. Never-
theless, (27) is in reasonable agreement with the experimental data shown in
figure 10. Each point on figure 10 was derived from a plot like those in figure 8,
the value of r for each bubble thus being a mean value derived from a number of
measurements of u, d , and h.

J. K. W. wishes to acknowledge the financial assistance of a Research Student-

ship from the Department of Scientific and Industrial Research throughout the
duration of this work.

Appendix
I n Part 1 there was a discussion of the magnitude of terms neglected in deriving
the first approximation pg), and a second approximation was derived.
However, further consideration shows that only the first of these, pi2),is truly a
second approximation, and the higher terms, pi2),pi2),etc., given by (9) in Part 1,
are little more accurate than the first approximation. The discussion below gives,
for the spherical bubble, a more detailed analysis of the error terms, and a second
approximation correct as far as pi2)is derived. This shows that the first approxi-
mation is valid for N < 1, but seriously in error when N + 1; of course all the
infinite series involved in the solution diverge in the region of N = 1, so that for
practical purposes the second approximation is not of much more use than the
first. But the second approximation does give some idea of the maximum value of
N for which the first approximation is reasonably accurate, and this maximum
value is probably not much more than 0-5.

Second approximation, for the non-expanding bubble

I n Part 2 , s 2, which gives the theory of distortion of a bubble of constant volume,
terms in ( 4 ) and (5) were neglected in order to derive approximate results, (9),
(10) and (12), giving the distortion of the bubble as a function of time. These
approximate results will now be used to estimate the order of magnitude of the
neglected terms, which are then used to derive the second approximation.
In ( 4 ) ,the first of the O ( p i )terms, giving the error in &q2is

A1(&g2)= (@/tcs)(l-5+ 1-71P2+ 1*29P4), (28)

putting R N a a t the surface of the bubble.

 The initial motion of a gas bubble formed in an inviscid liquid. Part 2 335
I n (5) there are errors derived from (28), and from putting R = a instead of
R = a(1 + 6) as it should. The latter approximation leads to the following errors
in ( 5 ) , derived from ( 2 ) ,

A(gr cos 8) = gacPl. (31)

I n deriving ( 2 9 )and (30),terms like l/Rnwere replaced by ( l / a n )( 1- n6) so that
the error terms ( 2 9 ) and (30)represent the first of a series of terms involving c2,
c3, etc.
Using the first terms of the series in ( 2 9 ) and (30)we get

Approximate expressions for these errors are now obtained by substituting for
/I1, p2 and 5 from ( 9 ) ,( l o ) ,and (12),giving, after simpIification, from (28), (31),
(32)and (33),
- A(gr cos 8 )
A(a$/at)- A1(Qq2)- A2(&q2)
= (a2/t2) +
( - 2*25N4 0*9N3Pp, +
- S*65N4P2 1*35N3P3- 4*04N4P4).(34)

The next step is to equate to zero the coefficients of P,,

in (5), so as to make the
pressure within the bubble independent of 8, and including the terms from (34),
and the last terms in each of (6)-( 8), we get
pi2)- ga3 - 9 U(1)@)/5a2+ 0-9N3a41t2= 0,
&2) + #u(2)/3$2)
- I$u(l)p$1)- 8.65N4aSIt2 =
(ii) 0,
(35)
1$2) + g ~ ( l ) / 3 $ 1 1-35NWIt2
)+ =
(iii) 0,
/@ ++EU(1)/p
I - 4.04N4a'/t2 = 0. (iv)
I n (35iii), it was sufficiently accurate to omit the term up4and to use U(1)#)
instead of U(2)pb2).Similarly in (35iv) it was possible to omit the term UP, though
it was necessary to use U1)/3i2) rather than U(l)p$'),because p3 changes more than
p2 in the second approximation.
Integrating (35)gives the following expressions for pi2),etc.,
pi2)= Nu4(1 - 0*9N2)/t, pi2' = - N2a5(1 - 2.54N )/
t'} (36)
,&$2) = 0.45N3a6 It, Pi2)= 0.246N4a7/t.
Using the above expressions for pi2',etc., in ( 1 1 )gives a second approximation
for 6,
6 = - 0.75N2P2+N3(0*75P3+ 0.3PJ - N4(0*746P4-0*24P2+0*3). (37)
336 J . K . Walters and J . F . Davidson
Comparison with (12) shows that there is a small change in the N3 term, and
there are considerable changes in the N 4 term. Examination of the derivation
shows that a third approximation would change only the N 4 term in (37),but
not the N 2 and N 3 terms. From (37)it is easy to get an expression for the central
height D of the bubble,
D/2a = 1 - 0.75N2- 0.806N4. (38)
Although the third term of (38)is very different from the corresponding term in
(13),the difference in the estimated value of D is small for N < 0.5; of course the
value of N to give D = 0 is considerably different ( N = 0-715 from (13) and
0.861 from (38)),but for these values of N the whole method of solution breaks
down, because the deformation of the bubble is of the same order as its radius.

Motion of the centroid

If s is the distance moved by the centre of co-ordinates, and sffthe distance moved
by the centroid,
lo=
s - 8 = - ysinI9cosOdI9
ff

(cf. ( 1 5 )in Part 1 of this paper), and substituting from (37) gives sff- s = 0.3N3a.
By integrating the first of (36),we get s = Na(1- 0.3N2),so that sg = N a ; this
means that the centroid has an acceleration of exactly 2g.

REFERENCES
CALDERBANK, P. H. 1956 Trans. Inst. Chem. Engrs, 34, 79.
DAVIDSON,L. & AMICK,E. H. 1956 A.I.Ch.E. J . 2 , 337.
DAVIDSON,J. F. & SCHULER, B. 0. G. 1960 Trans. Inst. Chem. Engrs, 38, 335.
HELSBY,F. W. & TUSON,K. R. 1955 Research, 8, 270.
JAHNKE,E. & EMDE,F. 1945 Tables of Functions, 4th ed., p. 108. New York: Dover.
LAMB,SIR HORACE 1932 Hydrodynamics, pp. 239, 241. Cambridge University Press.
MAGNUS,W. & OBERHETTINGER, F. 1949 Special Functions of Mathematical Physics, p. 50.
New York: Chelsea Publishing Co.
VANKREVELEN, D. W. & HOFTIJZER, P. J. 1950 Chem. Engng Progr. 46, 29.
WALTERS,J. K. 1962 Bubble motion and leakage from sieve trays. Ph.D. dissertation,
University of Cambridge.
WALTERS,J. K. & DAVIDSON, J. F. 1962 J . Fluid Mech. 12, 408.
WHITTAKER, E. T. C" WATSON, G. N. 1927 Modern Analysis, 4th ed., p. 331. Cambridge
University Press.

Errata to Part 1
The third term in equation (13)should read
- 0*0612N4(co~
48 + ...
and the second reference should read
DAVIES,R. M. & TAYLOR,
SIR GEOFFREY1950 Proc. Roy. SOC.A, 200, 375.
Journal of Fluid Mechunics, Vol. 17, part 3 Plate 1

%.
0
0
0
m

0
00
4
9
P

m
E5
+j
.-
a

WALTERS AND DAVIDSON (>’acing p . 336)

Journal of Fluid Mechunics, Vol. 17, part 3 Plate 2

.+
42

0
OD
Journal of Fluid Mechanics, Vol. 17, part 3

-
h

8
-
9,

m
Q
2
Journal of Fluid Mechanics, Vol. 17, part 3 Plate 4

WALTERS AND 1 ~ 4 ~ 1 ~ ~ 0 ~