ANNUAL

REVIEWS Further
Quick links to online content
Ann. Rev. Fluid Mech. 1977. 9: 145-85
Copyriyht © 1977 by Annual Reviews Inc. All rights reserved

BUBBLE DYNAMICS =-=8100

AND CAVITATION
by Stanford University - Main Campus - Lane Medical Library on 10/19/12. For personal use only.
Annu. Rev. Fluid Mech. 1977.9:145-185. Downloaded from www.annualreviews.org

Milton S. Plesset
Department of Engineering Science, California Institute of Technology,
Pasadena, California 91125

Andrea Prosperetti
Istituto di Fisica, Universita degJi Studi, 20133 Milano, Italy

1 INTRODUCTION

The first analysis of a problem in cavitation and bubble dynamics was made by
Rayleigh (1917), who solved the problem of the collapse of an empty cavity in a
large mass of liquid. Rayleigh also considered in this same paper the problem of a
gas-filled cavity under the assumption that the gas undergoes isothermal com­
pression. His interest in these problems presumably arose from concern with
cavitation and cavitation damage. Wit h n eglect of surface tension and liquid viscosity
and with the assumption of liquid incompressibility, Rayleigh showed from the
momentum equation that the bubble boundary R (t) obeyed the relation

RR+W<)2 = p(R) - poo

p
, (1.1)

where p is the liquid density, Poo is the pressure in the liquid at a large distance
from the bubble, and peR) is the pressure in the liquid at the bubble boundary. For
this Rayleigh problem, peR) is also the pressure within the bubble. Incompressibility

= .
of the liquid means that the liquid velocity at a distance r from the bubble center is
R2
u(r. t) zR . (1.2)
r
The pressure in the liquid is readily found from the general Bernoulli equation to
be

= R 3 [. ( )
J
R 1 R .2
per, t) Poo + -,:- [p(R)- poo] +ZP-,:- R 1 - -,:- . (1.3)

While Rayleigh neglected surface tension and liquid viscosity and kept the pressure
Poo constant, his dynamical equation (1. 1 ) is easily extended to include these effects.

145
 146 PLESSET & PROSPERETTI

For a spherical bubble, viscosity affects only the boundary condition so that it
becomes
2a 4p .
p(R) = Pi - If - If R , (1.4)

where now Pi is the pressure ip the bubble and p(R). as before, is the pressure in
the liquid at the bubble boundarY. The surface-tension constant and the coefficient
of the liquid viscosity are a and p, respectively. By allowing Pro to be a function
by Stanford University - Main Campus - Lane Medical Library on 10/19/12. For personal use only.

of time, one can use Equation (1. 1 ) to describe the experimental observations on
cavitation-bubble growth and collapse in a liquid flow (Plesset 1949). Other effects
Annu. Rev. Fluid Mech. 1977.9:145-185. Downloaded from www.annualreviews.org

not considered by Rayleigh, such as the stability of the interface, the compressibility
of the liquid, the effect of energy flow into or out of the bubble, and the physical
conditions within the bubble, are described in the sections that follow.
We may write here a generalized Rayleigh equation for bubble dynamics, in view
of Equation (1.4), as

(1.5)

where the pr.essure in the gas at the bubble wall, Pi. may be a function of the time,
and the pressure at infinity, Pro, may also be a function of the time. We may also,
for future reference, define an equilibrium radius of a gas nucleus for given values
of Pi and Poco :
20-
Ro = -- · (1.6)
Pi- Pro

A bubble of this radius will clearly remain at rest if it is initially at rest, although
it should be noted that this equilibrium is an unstable one.
The discussion of bubble dynamics divides itself in a natural way on the one
hand into those situations in which the bubble interior consists for the most part of
permanent gas and on the other hand into those situations in which the bubble
interior is composed almost entirely of the vapor of the surrounding liquid. Vapor­
bubble dynamics can be often simpiified by a further subdivision into vapor-bubble
dynamics in a subcooled liquid and into vapor-bubble dynamics in a superheated
liquid (Piesset 1957). The subcooled-liquid case corresponds to that in which the
vapor density is so small that latent heat flow does not affect the motion which is
then controlled by the inertia of the liquid. In this sense the liquid may be said to
be "cold," and the liquid is usually described as a cavitating liquid. The superheated
liquid in a similar way may be described as one in which boiling phenomena occur.
In boiling, the vapor-bubble dynamics is controlled by the latent heat flow rather
than by the liquid inertia (Plesset 1969). Here the coupling of the energy equation
with the momentum equation is essential. The case of the gas bubble also requires
in most cases the simultaneous consideration of the momentum equation and the
energy equation. A large body of literature has developed on this topic, which is
considered in the following section.
 BUBBLE DYNAMICS AND CAVITATION 147

2 GAS BUBBLES

We consider in this section the case in which the medium filling the cavity is
essentially a permanent, noncondensable gas. We neglect all the effects associated
with the vapor of the liquid which necessarily is present in the bubble together with
the gas. Clearly this procedure is legitimate as long as the partial pressure of the
vapor is small compared with the gas pressure.
by Stanford University - Main Campus - Lane Medical Library on 10/19/12. For personal use only.

Small-Amplitude Oscillations
Annu. Rev. Fluid Mech. 1977.9:145-185. Downloaded from www.annualreviews.org

The small-amplitude (linearized) oscillations of a permanent gas bubble in a liquid

were first considered by Minnaert (1933), who showed that a substantial contribution
to the sound emitted by running water comes from the free radial pulsations of
entrained air bubbles. More recently the same type of problem has been considered
in greater detail by several investigators including Devin ( 1 959), Plesset & . Hsieh
( 1 960), Chapman & Plesset ( 197 1 ), and Prosperetti (1976a).
In many instances, the case of small-amplitude forced radial oscillations arises
when a bubble is immersed in a sound field of wavelength large compared with
the bubble radius. Such a sound field may be introduced as an approximation in
the Rayleigh equation by writing
(2. 1 )
where P ro i s the average ambient pressure, w i s the sound frequency, and e i s the
dimensionless amplitude of the pressure perturbation which is supposed to be small.
It is also assumed that the oscillations take place about the equilibrium radius Ro
defined by Equation ( 1 .6), so that one may write
R = Ro[l +x(t)], (2.2)
with x(t) a small quantity of order e. Finally, one must introduce a suitable
linearization of the internal pressure Pi(t). Denoting by a dot differentiation with
respect to time, we write

-i- x=
l l
ap' ap'
Pi(t) = Pi.eq + -;--'
UX
x(t) +
uX
x(t) +" ', (2.3)
x=O,x=o O,x=o
where apJaxlx=o,x=ox represents the component of the internal pressure pertur­
bation in phase with the driving force, and apJih Ix=o,x=ox the component out of
phase with it by n/2. Upon substitution of relations (2. 1), (2.2), and (2.3) into
Equation (1. 5), one finds an equation of the harmonic-oscillator form
(2.4)
where the constant (J. is defined by (J. = p00/pR6, and the damping constant f3 and
the effective natural frequency Wo are given formally by

(2.5 )
 148 PLESSET & PROSPERETTI

and
1 °Pi 2a
Wo2 = -
I
p Ox x�o,x�o
-
pR5'
(2.6)

Further progress is clearly dependent upon the specification of the quantItIes

appearing in Equation (2.3). A simple assumption is that the gas follows a polytropic

( )
law of compression with polytropic exponent"
by Stanford University - Main Campus - Lane Medical Library on 10/19/12. For personal use only.

Ro 3K
Pi = Pi,eq Ii . (2.7)
Annu. Rev. Fluid Mech. 1977.9:145-185. Downloaded from www.annualreviews.org

If this assumption were strictly valid, one would find that energy dissipation arises
only from liquid viscosity and compressibility (see below). With this assumption the
natural frequency is
P;,eq �.
w6 = 3" - (2.8)
pR6 pR5
If surface-tension effects are neglected, and if the pressure-volume relationship is
taken to be adiabatic so that" equals y, the ratio of the specific heats of the gas,
Equation (2.8) coincides with the result first obtained by Minnaert ( 1933).
Clearly, the accuracy of Equation (2.7) can be assessed only by considering the
complete set of linearized conservation equations of mass, momentum, and energy
both in the gas and in the liquid. An analysis of this type was first undertaken by
Pfriem (1 940; see also Devin 195 9), who made use of a somewhat artificial
Lagrangian formalism. An explicit treatment in terms of the conservation equations
of continuum mechanics has been given by PIesset & Hsieh ( 1960) and more
recently by Prosperetti ( l 976a). The case of the free oscillations has been analyzed
by Chapman & PIes set (1971).
The results of these studies are two-fold. In the first place, it is found that the
effective polytropic exponent" exhibits a strong dependence on the driving sound
frequency w. For the case y = 7/5 (diatomic gas) this dependence is illustrated in
Figure 1 in terms of the dimensionless frequency G 1 = MDgWfyRg Too and of another
dimensionless parameter G2 = R6w/Dg. Here Dg denotes the thermal diffusivity of
the gas, M its molecular weight, Rg the universal gas constant, and Too the absolute
temperature of the liquid at a distance from the bubble. The physical basis for the
behavior depicted in Figure 1 can be clarified by noting that essentially three length
scales are involved in the problem, namely, the bubble radius Ro, the wavelength
of sound in the gas )'g = 2n(yRg Too/M)I/2/W, and the thermal penetration depth in
the gas, Lth = (DgW)I/2. The thermal penetration depth in the liquid is so small that
in practice it can be taken to be zero with a negligible error. In terms of these
three fundamental length scales, it is seen that, essentially, G1 (Lth/)'g)2 and
1
�

Gz (Ro/Lth)2. If (G1G ) /2 (i.e. RO/Ag) is small, the pressure within the bubble is
2
�

spatially uniform, and one will observe an isothermal behavior," ;:::: 1 when Ro � Lth
(i.e. G2 is small); and one will observe an adiabatic behavior " � y when Ro � Lth
(i.e. Gz is large). In the first case, the oscillations are too slow to maintain an
appreciable temperature gradient in the bubble, whereas in the second case they are
 BUBBLE DYNAMICS A N D CAVITATION 1 49

1.4

1.3
y=7/5
K

1.2
by Stanford University - Main Campus - Lane Medical Library on 10/19/12. For personal use only.
Annu. Rev. Fluid Mech. 1977.9:145-185. Downloaded from www.annualreviews.org

1.1

I. JI--�

10

Figure 1 The effective polytropic exponent K for the small-amplitude forced oscillations
of a gas bubble containing a diatomic gas (y = 7/5). The numbers labeling the curves
denote different values of the dimensionless frequency G, = MDgw/yRgToo• The quantity
G2 is defined by (;2 = mR5/Dg (from Prosperetti 1976a).

so fast that most of the gas contained in the bubble is practically thermally insulated
from the liquid. However, as was first pointed out by PIes set (1964) and by Plesset
& Hsieh (1960), for frequencies so large that ,19 is of the order of Ro or smaller
[i.e. (G 1 G2) 1/2 is of order 1 or larger], pressure nonuniformities develop in the
bubble and a polytropic pressure-volume relationship loses its thermodynamic
meaning. As an associated effect, the polytropic exponent /{ takes on values outside
the range 1 < /{ < y, and may become even negative (see Prosperetti 1976a). The
interesting question of the thermodynamic behavior at high frequency of the bubble
as a whole, however, is still meaningful. As discussed by Plesset (1964) and
Prosperetti (1976a), one is led to the conclusion of an overall isothermal behavior,
caused by the establishment in the interior of the bubble of a temperature distri­
bution consisting of many standing waves of short wavelength. 1 In view of Equation
(2.g), the frequency dependence ofthe polytropic exponent reflects itselfin a frequency
dependence of the effective natural frequency, in the sense that the pole of the
oscillation amplitude determined from Equation (2.4) is a function of the driving
frequency. The frequency at which resonance oscillations take place does remain
well defined (see Prosperetti 1976a).
1 Clearly, at very high frequencies one cannot consider the wavelength of the sound in

the liquid large compared with the bubble radius. The bubble oscillations then will not be
purely radial, but the physical content of this statement remains nevertheless applicable.
 150 PLESSET & PROSPERETTI

The second important result obtained by the analysis of the complete (linearized)
fluid-mechanical problem concerns the damping of the radial oscillations. Figure 2
presents the results of Chapman & Plesset ( 1 97 1 ) for the logarithmic decrement
1\ = 2nfJ/wo of an air bubble in water. It is seen that, in the range 0.1 cm > Ro >
4 X 1 0 -4 cm, the thermal component, represented formally by the second term in
Equation (2.5), dominates the viscous and acoustic contributions to the energy
dissipation. The acoustic contribution can be computed by attributing a slight

( )
compressibility to the liquid, and the result is the addition of
by Stanford University - Main Campus - Lane Medical Library on 10/19/12. For personal use only.

1 wRo/c 1 WRo
f3acoustic = -,;w (2.9)
1 + (WRO/ C)2
Annu. Rev. Fluid Mech. 1977.9:145-185. Downloaded from www.annualreviews.org

c:::: 2: -c- OJ

to the damping coefficient of Equation (2.5). This quantity determines the fraction
of the work performed by the sound field on the bubble that is dissipated as sound
waves radiated into the liquid. Results for the case of forced oscillations of an air
bubble in water are shown in Figure 3 for two values of the equilibrium radius Ro
(the frequency corresponding to the resonant frequency of the bubble is indicated
by an opcn circlc in these figures). Except for cxtremely small bubbles for which
viscosity is very important, the low-frequency damping is dominated by thermal
effects, and the high-frequency damping by acoustic effects. The analytic expression
for the thermal damping constant is rather involved and is not given here; the reader
is referred to Prosperetti (1976a).

-<

UJ
...J
U
>-
u

D::
UJ
!l..

I-
Z
UJ
:2
UJ
D::
Irl
c

u
�
I
I-
iE
<[ 0.1
C>
9

EQUILIBRIUM RADIUS. Ro (em)

Figure 2 The logarithmic decrement for small-amplitude free oscillations of an air bubble
in water as a function of the equilibrium radius Ro (from Chapman & Plesset 1971).
 BUBBLE DYNAMICS AND CAVITATION 151
A s a final point, w e may remark that Plesset & Hsieh ( 1 960) have considered the
small-amplitude oscillations as an initial-value problem and have shown that the
solution does indeed approach the solution of the steady-state problem represented
by Equation (2.4) as t -+ w.
Considerable experimental work on gas-bubble oscillations has been reported in
the literature. The data for the natural frequency can be said to agree well with the
theory (Lauer 1 95 1 , Koger & Houghton 1968, Jensen 1 974), whereas a large scatter
appears in the data for the damping of the oscillations, as can be seen for example
by Stanford University - Main Campus - Lane Medical Library on 10/19/12. For personal use only.

in Figure 2 of Koger & Houghton ( 1968). While this situation may certainly be
ascribed to a large extent to the quality of the data (for discussions, see e.g. Devin
Annu. Rev. Fluid Mech. 1977.9:145-185. Downloaded from www.annualreviews.org

1959, Kapustina 1 970, van Wijngaarden 1 972), it appears that the situation has not
yet been completely clarified. Some references to the original works are contained
in the papers referenced above; in a recent work Ceschia & Iernetti (1974) made
use of the subharmonic threshold to measure the damping of the oscillations and
report acceptable agreement with theory for bubbles of different gases in water. An
exception is hydrogen, for which anomalous behavior is reported. A different use of
the subharmonic threshold, which appears capable of yielding very accurate data,
has recently been proposed by Prosperetti ( 1 976b).
Nonlinear Oscillations
With the assumption of polytropic behavior and with the neglect of thermal and
acoustic dissipation, the dynamical Equation ( 1 .5) takes the following form in an

]
oscillating pressure field:
.. 3 2 1 [ (Ro)3K 2(1 1 .
=p Pi.eq - P oo ( 1 - l] cos wt) - - 4 J1
•

RR + z(R) R , (2. 1 0)
R R R
where, in general, the dimensionless pressure amplitude I] is not necessarily small.
The numerical studies that can be found in the literature have shown the extreme
richness of this equation which appears to lie beyond the capabilities of the available
analytical techniques.
Among the earliest numerical work, the contribution ofNoltingk & Neppiras ( 1 950,
1951) should be mentioned. With the hypothesis of isothermal behavior (K = 1 ) and
with the neglect of viscous effects they were able to show the apparently explosive
behavior of the solutions of Equation (2. 1 0) which are sometimes found to consist
of a very rapid growth followed by a violent collapse to very small values of the
radius within a single period of the driving force. Basing themselves on these
numerical results, they deduced some conclusions regarding the mechanism and
the conditions for acoustic cavitation that have been used as a guide in much
subsequent research on this subject.
Noltingk and Neppiras's work was somewhat limited by the capabilities of the
computer at their disposal, but later Flynn (1964) and Borotnikova & Soloukin
(1964) published several radius-versus-time curves that illustrate the complexity of
the possible responses and their sensitivity to the parameters of the problem. Figure 4,
from Borotnikova & Solo ukin ( 1964), shows that the explosive behavior can be
observed after several oscillations, as had already been conjectured by Willard ( 1953),
 152 PLESSET & PROSPERETTI

both for bubbles driven below and above resonance. These examples give an indi­
cation of the importance of the initial conditions in the subsequent bubble motion,
an area in which little research has been conducted.
The most extensive numerical investigation of Equation (2. 10) is that undertaken
by Lauterborn ( 1968, 1 970a, 1970b, 1 976), who has collected an impressive amount
of results, still in part unpublished. Figure 5 from his 1 976 paper illustrates the
steady-state response XM = (Rmax - Ro)/Ro (where RlIlax is the maximum value of the
radius during the steady oscillations) as a function of the ratio w/wo for several
by Stanford University - Main Campus - Lane Medical Library on 10/19/12. For personal use only.

values of the dimensionless pressure amplitude 11. Here Wo is the resonant frequency
for the linearized oscillations given by Equation (2.8). These calculations refer to an
Annu. Rev. Fluid Mech. 1977.9:145-185. Downloaded from www.annualreviews.org

air bubble of radius Ro 10 em in water under a static pressure P

= - 3 1 bar. The
co =

complicated pattern of the resonances that gradually unfolds as the pressure

amplitude is increased indicates very clcarly the importancc of nonlinear effects for
large-amplitude bubble oscillations.
lo4 r-�-.--.-.-.-"rrr----'--'-'-'-"TTr---TT--'-'-'-rTrn

f3

w
(0)

Figure 3 The damping coefficient f3 for the small-amplitude forced oscillations of

 BUBBLE DYNAMICS AND CAVITATION 1 53
In Figure 5 the broken vertical lines represent the points at which the steady-state
response is observed to jump from some low value to a higher one. This behavior is
caused by an instability of the small-amplitude oscillation in a frequency region in
which the response curve is really multivalued (see Figure 6). This behavior was found
analytically by Prosperetti (1974, 1975, 1976b), who gave a description of small­
amplitude forced oscillations including some nonlinear elTects. Although his theory is
quantitatively accurate only for relatively small values of lJ[r/ ;:s 0.25( 1 + 2rr/R oP00)]
by Stanford University - Main Campus - Lane Medical Library on 10/19/12. For personal use only.

and can account for only a few of the resonances of the system, it gives an insight
into the complicatcd cffects of the initial conditions on the ensuing oscillatory
Annu. Rev. Fluid Mech. 1977.9:145-185. Downloaded from www.annualreviews.org

motion. An example of his results is shown in Figure 7, which depicts the behavior
of the first subharmonic amplitudes u(t), v(t) [with u(t) cos -iwt- v(t) sin -iwt. the sub­
harmonic component in the oscillation] in the (u, v) plane for three different sets of
initial conditions. The numbers along the curves represent time in units of wt, and
the dashed lines are the separatrices for the undamped forced oscillations (see

-3
Ro= I 0 em

f3

w
(b)

an air bubble in water as a function of the driving frequency (from Prosperetti J976a).
 1 54 PLESSET & PROSPERETTI

Prosperetti 1 975 for details). It is seen that it is not possible to express in a simple
way the relation between the initial conditions and the value of the amplitudes in
the steady-state oscillations (which is the point into which the curves spiral: the
origin for curves a and b, and a nonzero value for curve c) . One must expect that
a much more complicated pattern of this type would be applicable to all the
resonances of the complete Equation (2. 10), particularly for large-amplitude oscil­
lations for which a small change in the initial conditions can cause markedly
different transients and steady-state motions. Such a behavior is indeed reported in
by Stanford University - Main Campus - Lane Medical Library on 10/19/12. For personal use only.

the numerical studies by Lauterborn ( 1 976).

In addition to furnishing a suitable example to illustrate the importance of initial
Annu. Rev. Fluid Mech. 1977.9:145-185. Downloaded from www.annualreviews.org

conditions, subharmonic bubble oscillations are interesting per se in view of the

apparent connection between a signal at half the frequency of the sound field and
the development of cavitation in an acoustically irradiated liquid. A subharmonic
component in the spectrum of a liquid undergoing acoustic cavitation was first
reported by Esche ( 1952), and later by Negishi ( 1961). Subsequently De Santis et al
( 1967), Vaughan ( 1968), Neppiras ( 1969a,b), Mosse & Finch ( 1971), and Coakley
( 197 1) established an apparent causal connection between the onset of cavitation
and the appearance of the subharmonic signal. Neppiras ( 1969a,b), experimenting
with single bubbles, showed that they were indeed capable of emitting at a frequency
equal to one half of the driving frequency, and advanced the hypothesis that these

20
! !
R .- I I
I
II
,/
...... -
I
.,.... . !
i I I
! 1/ I , �I
.... ..
10 --- ---
, i
-
, i i
V: � Vl , 1'\
y'YII y' I
""

I I i I
5 10 15 20 2

(a)

p
�QYbtf#YRIfizt1
5 l d
r<:>"":/,<:>?,<2,Ac>oc/�S.A.5(R,,c>�9"'c,oc/">�9.�oc>ovo
T

(b)
Figure 4 Two examples of numerical results for the transient motion of a gas bubble in
an oscillating pressure field. For Figure 4a the conditions are tf = 1.5, w/wo = 0.154; for
Figure 4b tf = 5, w/wo = 1.54. The ordinate scale is the dimensionless radius R/Ro; the
abscissa scale is the dimensionless time wt (from Borotnikova & Soloukin 1964).
 BUBBLE DYNAMICS A N D CAVITATION 155

subharmonically oscillating bubbles would evolve into transient cavities that would
collapse and break up, and thus would produce the several phenomena associated
with acoustic cavitation. Such a h ypothesis encounters some difficulties which have
been summarized by Pros peretti ( 1975). This author has also proposed an alternative
explanation according to which the bubbles emitting at the subharmonic frequency
do not participate actively in the process of cavitation, but act merely as monitors
of its occurrence. Central to this explanation are the effects of the initial conditions
for the oscillations. Before leaving this subjcct mcntion should be made of a work
by Stanford University - Main Campus - Lane Medical Library on 10/19/12. For personal use only.

by Eller & Flynn ( 1 969), in which the threshold for the instability of the purely
harmonic motion in the subharmonic region is given. As discussed by Prosperetti
Annu. Rev. Fluid Mech. 1977.9:145-185. Downloaded from www.annualreviews.org

(1976b), this threshold does not coincide with the true subharmonic threshold except
in the immediate neighborhood of w = 2wo.
A certain number of results concerning free oscillations of gas bubbles [described
again by Equation (2. 10) with 1] = 0] are also available. For the particular case of
1\ = %, and of vanishing dissipative and surfa ce tension effects, Childs (1973) has
-

2.4

2.2

2.0

1 8

1.6

14
Rmax -Rq
Ro
1 2
2
T

101 :�
o8 f i 2

06 f
04 1
o
a
2[
01 0.4 0.5 0.6 0.8 0.9
015 02 0.3 0.7 1
-"L
Wo

Figure 5 Response curves for the steady oscillations of an air bubble of equilibrium
radius Ro = 10-3 em in water as a function of the ratio of the driving frequency w to the
natural frequency for small oscillations woo The fractions on the resonance peaks denote
the order of the resonance [i.e. min indicates that m cycles of the oscillations take place
during n cycles of the driving-pressure amplitude (from Lauterborn 1976)]. a is for IJ 0.4,
is for IJ = 0.5, c is for '1 = 0.6, d is for '1 = 0.7, is for '1 = 0.8. The dots and the arrows
=

b e

belong to curve e.
 1 56 PLESSET & PROSPERETTI

obtained an exact solution of Equation (2.10) in terms of elliptic integrals. Nonlinear

effects on the small-amplitude motion have been analyzed by Prosperetti ( 1 975),
who included surface tension and dissipation. Lauterborn (1968, 1 976) reported
extensive numerical results on large-amplitude free oscillations including viscous
dissipation.
It should be pointed out that in all the work on nonlinear oscillations mentioned
above no detailed analysis of the gas contained in the bubble was attempted. There-
by Stanford University - Main Campus - Lane Medical Library on 10/19/12. For personal use only.

{ stable
Annu. Rev. Fluid Mech. 1977.9:145-185. Downloaded from www.annualreviews.org

ANALYTICAL
unstable - - --
0.4
NUMERICAL -- - -- ---

0.3

XM ,

,
,
'\
0.2
,
,
\ '\
,
\
\
\
0.1 \
\
\
I
I
I
0 ---------------

1.95 2.00 2.05

J&
Wo

Figure 6 The response curve for the steady subharmonic oscillations of an air bubble of
equilibrium radius Ro = 10-3 em in water. The pressure amplitude is '1 = 0.3, and the
ambient pressure Pro = I bar. Only viscous damping is taken into account, and the polytropic
exponent is h' = 1.33. The numerical results shown have been obtained by Lauterborn ( 1974)
(from Prosperetti 1 974).
 BUBBLE DYNAMICS AND CAVITATION 1 57
fore, all the results described here have been derived with the assumption of a
polytropic relationship and sometimes with an assumption of an effective viscosity
to account in an approximate way for the effect of the thermal and acoustic
dissipation processes. Very recently Flynn ( 1975a) has derived a somewhat complex
formulation of the general problem of cavitation dynamics that includes a detailed
analysis of the gas behavior. This formulation has been applied by him to the study

V
by Stanford University - Main Campus - Lane Medical Library on 10/19/12. For personal use only.

./ ",... ------ ,
./ "-
0 .4 / "-
Annu. Rev. Fluid Mech. 1977.9:145-185. Downloaded from www.annualreviews.org

/ "-
"-
\ \
\ \
\ \
\ \
\ \
\ 190 \
\ \
c
\ \
I I
I
0.2 0.2 0.4 0.6 U
I
,
25
\ I
I
67
\ I
\ (
\ -0.2 /
\ i
\ / 10
/
\ / /
\ /
,
/
b /
, /
-0.4 "- /
"- /
....... /'
./
..... - --- --

Figure 7 Phase-plane behavior of the transient subharmonic amplitudes u, v [where

u(t) cos !wt- v(t) sin !wt is the subharmonic component of the bubble oscillations] for
different initial conditions. The driving frequency is w = 1.8wo, the pressure amplitude
IJ = 0.4, the ambient pressure P CD = 1 bar, and the polytropic exponent I( = 1.33. The
numbers on the curves denote time in units of wt. The steady oscillations will contain a
subharmonic component for the ease of the eurve labeled c, but not for curves a or b (from
Pros peretti 1975).
 1 58 PLESSET & PROSPERETTI

of large-amplitude free oscillations, and some interesting results on the energy

dissipation of the cavity have been obtained (Flynn 197 5b).
For large bubbles, such as those produced by underwater explosions, the most
significant damping mechanism is acoustic energy radiation in the liquid, and
therefore Equation (2.10) cannot be used. For this extreme case, Keller & Kolodner
( 1 956) have obtained another equation that accounts for liquid compressibility in
large-amplitude free oscillations. These results have recently been completed and
extended to the case of plane and cylindrical bubbles by Epstein & Keller ( 1 972).
by Stanford University - Main Campus - Lane Medical Library on 10/19/12. For personal use only.

Other studies of compressibility effects in the motion of gas bubbles are discussed
below.
Annu. Rev. Fluid Mech. 1977.9:145-185. Downloaded from www.annualreviews.org

M ass-Diffusion Effects
The mass-diffusion processes taking place across the bubble-liquid interface play a
significant role in the behavior of gas bubbles because they may ultimately determine
the presence or absence of bubbles in a liquid. The key to these processes is
furnished by Henry's law, which establishes a connection between the partial
pressure of a gas acting on a liquid surface, Pg, and the equilibrium (or saturation)
concentration of gas in the liquid which we denote by Cs:
Cs =apo· (2. 11)
Here a is a constant characteristic of the particular gas-liquid combination and is
primarily a function of temperature. Equation (2. 1 1) is valid also at a nonplanc
interface.
If we consider first a situation in which the ambient pressure is fixed and equal
to p 00' then it is clear that unless the gas concentration c at the bubble surface
satisfies Equation (2. l 1) (where we now interpret Po as the gas pressure in the
bubble and neglect for simplicity its vapor content), the bubble will not be in
equilibrium, and it will either grow or shrink according to whether c > Cs or
c < Cs. The mathematical formulation for this process clearly consists of the diffusion

(r20C)
equation in the liquid
oc R 2 . Oc rx 0
+ R = (2. 12)
at -;z Or -;:z Or Or '
where rx is the coefficient of diffusion of the gas in the liquid, subject to the boundary
condition (2. 1 1) at the (moving) bubble surface. Strictly speaking, the pressure Pg in
(2. 1 1) should be determined from the dynamical Rayleigh equation. The problem
can be substantially simplified by observing that the bubble-wall velocities that are
induced by mass diffusion alone are usually quite small in view of the relatively
low value of (1..2 Indeed, a reasoning based on purely dimensional considerations
suggests that
. Coo - Cs
R R =rx , (2. 1 3)
Pg
---

2 For air-water at 20oe, for example, G< "" 2 x 10-5 cm2 sec-I An order-of-magnitude
estimate for this quantity is given by the familiar result due to Einstein, G< = 6npakBT
where kn is the Boltzmann constant, and a is an equivalent radius of the gas molecule.
 BUBBLE DYNAMICS AND CAVITATION 1 59
where Coo is the dissolved mass concentration at a large distance from the bubble
and pg is the density of the gas within the bubble. For typical values, one might
take (coo -- cs)/pg 10- 2, Ro 10- 3 cm ; Equation (2. 13) then gives an estimate of
� �

R 10-4 cm sec - 1 for the growth (or dissolution) velocity of an air bubble in water.
�

These very low values allow one to disregard the inertial effects associated with the
bubble motion and to neglect the convective term in Equation (2.12).
An approximate solution based on these simplifications was obtained some time
ago by Epstein & Plesset ( 1950), who included surface-tension effects. Their result,
by Stanford University - Main Campus - Lane Medical Library on 10/19/12. For personal use only.

with the neglect of the latter, takes the form

. c --c
Annu. Rev. Fluid Mech. 1977.9:145-185. Downloaded from www.annualreviews.org

RR a -:::--�{1+ R(nat)-1/2]. (2. 14)

Pg
=

The second term in this equation reflects the transient effect of the buildup of a
diffusion boundary layer adjacent to the bubble surface that quickly becomes large
compared with the bubble itself. Indeed, from (2. 13) we find that, asymptotically,
R � [2a(c",--cs)t/pg]1/2 from which R(nat)-lJ2 � [2(coo --cs)/npg]1/2 «: 1. The physi­
cal meaning of the result of Equation (2. 14) can be clarified by noting that if my
denotes the gas content in the bubble so that dmg/dt 4nR2pg dR/dt, elimination
=

of dR/dt formally reduces it to the expression for the heat flux at the surface of a
perfectly conducting sphere in an infinite medium (Carslaw & Jaeger 1959). Experi­
mental evidence in favor of Equation (2.14) was reported by Krieger et al (1967),
who proposed a method for the measurement of diffusion coefficients of gases in
liquids based on the observation of bubble-dissolution rates. A systematic pertur­
bation approach to the solution of the problem has been formulated by Duda &
Vrentas (1969a,b), and an interesting treatment of the case of more than one gas
has recently been given by Ward & Tucker (1975). Birkhoff et al (1958) havc obtaincd
a similarity solution for the complete problem including convection effects, and the
same solution was obtained by Scriven ( 1959). Szekely and co-workers (197 1, 1973)
have given a detailed analysis that includes dynamic, viscous, and surface-kinetics
effects.
The physical situation is quite different if the bubble is immersed in an oscillating
pressure field, and it is readily seen that in this case it may grow even in an under­
saturated solution. The mechanism giving rise to this effect may be explained in
physical terms in the following way. Suppose that the amplitude of the forced
oscillations is large enough so that the gas-liquid solution at the bubble surface
becomes undersaturated during the compression half-cycle and supersaturated
during the expansion half-cycle. Corresponding to these conditions there will be a
mass exchange of alternating direction between the bubble and the liquid. If the
interface were plane, the average flux over one oscillation would clearly be zero.
However, as a consequence of the spherical geometry, on the average the surface
area during the mass inflow is greater than during the mass outflow, so that a net
increase in the mass of gas contained in the bubble results. In view of its second­
order nature, this effect has been called "rectified mass diffusion" (Blake 1949). An
additional consequence of the spherical geometry, however, combines with the area
effect just mentioned, namely, the fact that during the expansion half-cycle the
th ick n ess of the diffusion layer adjacent to the bubble surface decreases, while it
 160 PLESSET & PROSPERETTI

increases during the compression half-cycle. As a result concentration gradients (and

hence mass-flow rates) during the mass inflow half-cycle are enhanced over those
prevailing during the mass outflow part of the oscillation. A quantitative analysis
of the convective effect shows it to be dominant, as was demonstrated by Hsieh &
Plesset (1961), who gave the first rigorous treatment of the rectified diffusion process
based on a linearized formulation of the problem. For the case in which the liquid
is saturated in the absence of the bubble, neglecting surface-tension effects and
assuming isothermal behavior, they find
by Stanford University - Main Campus - Lane Medical Library on 10/19/12. For personal use only.

RR="3"O:� (PmaxPo-Po)2,
.
2
Cs
(2.15)
Annu. Rev. Fluid Mech. 1977.9:145-185. Downloaded from www.annualreviews.org

pg
where Pmax and Po are the maximum and the average value of the bubble internal
pressure, respectively.
If P max is taken to be a constant, Equation (2. 1 5)
is readily integrated and for
large times predicts a tl/2 dependence of the bubble radius. The fact that in practice
bubbles are not seen to grow indefinitely, as this result would imply, is a consequence
of the instability of the spherical shape, which, as is discussed below, is an important
effect for relatively large bubbles (Hsieh & Plesset 1961). A significant feature of
Equation (2. 15) is that it shows bubble growth by rectified diffusion to be a very
slow process relative to the period of the sound fields commonly encountered. For
instance, for an air bubble in water at 20°C at one atmosphere, with max- 0)/ 0
0.25, Equation (2.1 5) predicts a doubling time that ranges from 1.1 x 1 06 sec for an
(P P P =
initial radius of 10- 1 cm to 1.1 x 102 sec for an initial radius of 10- 3 cm.
When the liquid is not saturated, the bubble will eventually disappear if the mass
flux caused by rectified mass diffusion is not sufficient to balance the loss of mass
required by Henry's law. The value, 'Ith. of the pressure amplitude at which the two
fluxes are equal corresponds to the threshold for the growth of the bubble by
rectified diffusion, and was estimated very simply by Strasberg (1961) by equating
Equation (2. 1 3) and Equation (2.15). His result, which includes surface-tension effects,
IS

Y/th = (z) 2 ( RoPo - )

3 1
I 1+ --
2a Coo 1/2
�

Cs
(2.16)

Here, and in the following, Cs is the saturation concentration at the average pressure
of the sound field only for bubbles driven so much below resonance that inertial
effects in their motion are insignificant. As was pointed out by Safar (1968), inertial
effects introduce a correction in Equation (2.16) which for isothermal behavior
becomes

= (1 Wo)( RoPo) ( 3RoPo)(1 RoPo)

2
+ � 1/ 1 +
� � -1/2
Coo +
Y/th H)I/2 _ « 1 _
, (2.17)
Cs

where Wo is given by Equation (2.8) with h" = 1.

A major approximation contained in the above results consists in the neglect of
the nonlinear effects associated with the bubble oscillations. This aspect of the
problem was analyzed by Eller & Flynn (1965), who made use of a method
 BUBBLE DYNAMICS AND CAVITATION 161

developed by Plesset & Zwick ( 1 952) to describe the growth of vapor bubbles. While
we return to this method in greater detail below, it may be mentioned here that
its applicability is limited to those situations in which appreciable concentration
gradients are present only in a layer adjacent to the bubble surface that is thin
compared with the bubble radius. Clearly, for the threshold problem under con­
sideration here, this condition takes the form Ro ;l> (a/w)I/2, and hence is usually met

(
in practice. Eller and Flynn's result is
)
by Stanford University - Main Campus - Lane Medical Library on 10/19/12. For personal use only.

Coo 2(J <R/Ro>

1+ 0 (2.18)
-Z- RoPo «R/Ro)4> = ,
Annu. Rev. Fluid Mech. 1977.9:145-185. Downloaded from www.annualreviews.org

where the brackets denote the average of the enclosed quantity over one period
of oscillation. The pressure amplitude of course enters in this equation through
the averages, which can be computed either numerically or by an analytic perturbation
scheme (Eller & Flynn 1 965; Eller 1 969, 1972, 1 975). For the case in which Coo = Cs
and rr/RoPo � 1, Eller gives the result
( (2)( )1/2(1 w2 )-1/2
30"
-- -
_

I]'h - 1 2 (2. 1 9)
- Wo RoPo S-2
Wo '

which is seen to coincide with Safar's modification of the Hsieh-Plesset theory

except for the last factor, which is introduced by the nonlinear effects in the second­
order approximation. A similar result has been derived from other techniques by
Skinner (1970).
Both Equations (2.17) and (2. 19) appear to predict a vanishing threshold at
resonance. This feature of course is a consequence of the neglect of damping effects
which have recently been analyzed by Eller ( 1 975). He found a very pronounced
minimum in the threshold curve at resonance and considered the effect of this
behavior on the evolution in time of a population of bubbles with given initial
radius.
Experimental observations of threshold values were first reported by Strasberg
(1959, 1961) at 25 k Hz, and more recently by Eller at 26.6 k Hz (1969) and at
11 k Hz (1972) and by Gould (1974) at 20 k Hz. It is found that the data given
by Strasberg at 25 k Hz and by Eller at 26.6 k Hz tend to fall somewhat below
the theoretical predictions. Although the discrepancy is not serious for bubble radii
greater than about 4 x 10-3 em, a definite disagreement appears to exist for smaller
radii. The data at 11 k Hz are in better agreement with theory, provided that the
bubbles are taken to oscillate adiabatically rather than isothermally (Eller 1972). No
data for bubbles smaller than 5 x 10-3 cm are reported, however. The data of Gould
(1974) are also generally in good agreement with Equation (2. 19) for a polytropic
form of the pressure-volume relation. The smallest bubble observed by him had a
radius of 5.5 x 1 0 -3 cm.
A rigorous analysis of nonlinear effects in the theory of bubble growth by
rectified diffusion is more complex than for the threshold calculation because now,
in addition to the oscillatory boundary layer of thickness (rJ./w)1/2, one must take
il1to account the layer in which the gas concentration is depleted because of the net
mass diffusion into the bubble. It is easy to convince oneself that the thickness of
 162 PLESSET & PROSPERETTI

this layer is of the order of the bubble radius or greater so that the applicability
of Eller and Flynn's method is not immediately evident. The problem has recently
been considered by Skinner ( 1972), who obtained a solution with the aid of multiple
time- and length-scale expansions. Surprisingly, his results are practically equal to
those previously derived by Eller ( 1969) by the same approach as that used for the
threshold condition by Eller & Flynn (1965). A comparison with data (Eller 1969),
however, shows that the bubble growth through resonance is much sharper than the
by Stanford University - Main Campus - Lane Medical Library on 10/19/12. For personal use only.

theory predicts. Other large discrepancies between theory and experiment for bubble
growth rates have also been reported by Eller ( 1 969), who observed rates twenty
Annu. Rev. Fluid Mech. 1977.9:145-185. Downloaded from www.annualreviews.org

times larger than the theoretical results. Recent experiments by Gould (1974) appear
'
to substantiate Eller's conject
streaming taking place in the vicinity of the bubble surface. He observed the
appearance of surface oscillations on the bubble and of a streaming motion in its
vicinity together with large increases in growth velocity above the theoretical values.
Growth rates measured in the absence of streaming, however, tend to be smaller
than the predicted values. Apparently, the connection between surface oscillations
and streaming, although experimentally well documented (Kolb & Nyborg 1956,
Elder 1959, Gould 1966), has not yet found a proper theoretical treatment. The
studies of Davidson ( 1971 ) and of Davidson & Riley (197 1) refer to the case in which
the streaming is produced by an oscillatory motion of the center of a spherical
bubble and is therefore not directly applicable to surface-oscillation-induced
streaming.

Acoustic Cavitation and Applications

In the terminology of Blake (1949; see also Flynn 1964) we may distinguish bctween
gaseous and vaporous acoustic cavitation. The first description refers to the
phenomena associated with bubbles that have predominantly a noncondensable
gas content; the second, to bubbles that have predominantly a vapor content. While
we have discussed several features of the dynamics of gaseous cavities above, and
we shall discuss the behavior of vapor cavities in the next section, it should be
realized that at present it is not yct possible to relate all of the specific features of
experimentally observed cavitation phenomena with the available experimental and
theoretical work on single gas or vapor bubbles. Although it is generally agreed
that cavitation damage, white noise, sonoluminescence, chemical reactions, and
other features of cavitation are associated with violent bubble motion, at present
not only a quantitative understanding of these phenomena is lacking but sometimes
even the physical mechanisms through which they take place are obscure.
Violent collapse, as can be seen in Figure 4, is certainly a possible behavior for
a gas bubble in an oscillating pressure field, and estimates of the temperatures and
pressures reached by the bubble content toward the end of the collapse are of the
order of several thousands of degrees Kelvin and of thousands of atmospheres,
respectively (see e.g. Flynn 1964; 1975a,b). Unfortunately, during the later stages of
the collapse the spherical shape of the bubble becomes highly unstable (see below),
and the duration of the collapse itself is so short that adequate theoretical or
 BUBBLE DYNAMICS AND CAVITATION 163

experimental investigations appear very difficult. These same considerations can

be applied also to vapor-bubble collapse in a highly subcooled liqu id.
No less obscure is the beginning of a vapor- or gas-bubble life. This problem is
essentially that of nucleation, and the uncertainties associated with it have a signifi­
cant bearing not only on acoustic cavitation but, perhaps more importantly, on
boiling heat transfer. Indeed, while it is observed that homogeneous-nucleation
theory based on thermodynamic fluctuations describes very well the behavior of
certain organic liquids, such as pentane, hexane, heptane, ether, and benzene (see
by Stanford University - Main Campus - Lane Medical Library on 10/19/12. For personal use only.

e.g. Cole 1974, Blander & Katz 1975), the same conclusion cannot be drawn for
such an important liquid as water for which the presence of impurities appears to
Annu. Rev. Fluid Mech. 1977.9:145-185. Downloaded from www.annualreviews.org

play a determinant role. We cannot enter here into any detail on these matters,
and we refer the reader to some pertinent recent papers (Apfel 1970, Gavrilov 1970,
HolI 1970, KelIer 1972). A useful summary of previous work is given by Flynn
(1964).
Once nucleation has taken place, a gas bubble will grow by rectified diffusion
over many periods of the driving pressure oscillations until it breaks up. The lifetime
of a vapor bubble, however, will in general be much shorter, not more than a few
acoustic cycles, and its growth will be determined more by dynamic efTects than by
a change in the mass content of the cavity. Some aspects of this prol:ess are dealt
with in the following section.
The description of cavitation events is made even more difficult by the apparent
importance of cooperative efTects of many bubbIes in several phases of the process
(Willard 1953). Space limitations prevent us from entering into any detail here. The
reader is referred to the works of WilIard (1953), Strasberg (1959), Noltingk (1962),
Flynn (1964), Coakley (1971), Hinsch, Bader & Lauterborn (1974), and Nyborg
(1974). Work on sonoluminescence is described in Negishi (1961), Taylor & Jarman
(1968, 1970), Margulis (1969), Saksena & Nyborg (1970), and Coakley & Sanders
(1973).
Finally, for some industrial applications of acoustic cavitation the reader is
referred to Neppiras (1965). A growing area of interest appears to be in biomedical
'applications (Rooney 1970, 1972; Nyborg 1974; Hill 1972).

3 CAVITATION BUBBLES AND VAPOR BUBBLES

In this section we consider the analysis of some features of the bubbles that are
composed predominantly of vapor. As was suggested in a preceding section, these
bubbles also usually originate from small gas nuclei. However, if a bubble grows
to many times its initial size so rapidly that mass-diffusion effects are negligible, the
gas content plays an unimportant role in the dynamics except possibly near the end
of a violent collapse.
We are not concerned here with the problem of boiling heat transfer, which has
recently been considered by Rohsenow (1971). Rather, we limit our discussion to
the growth and collapse of vapor bubbles in unbounded liquids. In the presence of
boundaries, the free surface loses its spherical symmetry. Some of these boundary
effects are considered in the next section.
 1 64 PLESSET & PROSPERETTI

Preliminary Considerations
In the study of the dynamics of vapor bubbles it is helpful to make a distinction
between two limiting cases according to the importance of thermal effects. The
physical basis for this distinction rests on the combined effects of the strong
temperature dependence of the equilibrium vapor density and of the rate of change
of the equilibrium vapor pressure with temperature. To illustrate this point we may
consider the case of a bubble in water that grows to a radius R in a time t. The
by Stanford University - Main Campus - Lane Medical Library on 10/19/12. For personal use only.

amount of thermal energy required to fill the bubble with vapor in thermodynamic
equilibrium with the water is (4j3)nR3 Lp�(T). where L is the latent heat of
Annu. Rev. Fluid Mech. 1977.9:145-185. Downloaded from www.annualreviews.org

evaporation and p�(T) is the equilibrium vapor density at the water temperature
T. This energy is made available through a drop AT in the temperature of a
surrounding liquid layer of thickness of the order of the diffusion length given by
(Dt)1/2, where D = kjPCI is the thermal diffusivity in the liquid. The heat energy
supplied is, therefore, 4nR 2(Dlt)1 / 2pc[AT. When these two expressions are equated,
one obtains an estimate of the temperature drop AT as
� Rp�(T)L
AT -
- (3. 1 )
3 (D[t)1/2pC[
For water a t 1 5°e, with R = 0. 1 c m and t = 1 0 - 3 sec, Pv = 1.3 X 10 - 5 g cm - 3,
one finds A T 0.2°e. This temperature drop causes a decrease in the vapor pressure
�

of the order of 1 %, and hence has an insignificant effect on the growth of the
bubble. For water in the neighborhood of l OOoe, however, the equilibrium vapor
density is about 46 times its value at 1 5°e, and one finds AT ,.,,; l3oe, with a
corresponding decrease in vapor pressure of roughly 50%. Clearly, the bubble­
growth dynamics will be drastically altered with a longer growth time to the same
radius. Thermal effects rather than inertial effects now dominate the growth process.
We refer to the first case as "cavitation" bubbles and to the second one as "boiling"
or "vapor" bubbles. A similar difference in behavior will be observed in the collapse
process. For a cavitation bubble the internal pressure will remain practically constant
until the latest stages of the collapse, while for a boiling bubble a much greater
effect from the vapor pressure will be observed.
The Dynamics of a Cavitation Bubble
With the neglect of viscous effects and by use of the identity
� _1_. � (R3R2) = RR + 3R2
'2' , (3.2)
2 R2R dt
Equation (1 .5) can be integrated once if the ambient pressure is taken to be
independent of time and if, in the light of the above discussion, the internal pressure
P i = P v is also regarded as a constant. The result is

(3.3)

where the subscript zero denotes the initial conditions for the growth. If Pv > Poo'
 BUBBLE DYNAMICS A N D CAVITATION 165
it is seen that for R � Ro the velocity is approximately equal to its asymptotic value
Pv �
Poo fZ,
R =
G (3.4)

which can also be obtained by equating the kinetic energy of the flow to the work
performed by the pressure forces. It is shown below that Equation (3.4) gives also
the growth velocity of a boiling bubble at high superheats or low ambient pressures
in the early and intermediate stages. If Pv < Pro but an initial impulse is imparted
by Stanford University - Main Campus - Lane Medical Library on 10/19/12. For personal use only.

to the bubble wall, Equation (3.3) predicts that the bubble would reach a maximum
radius that, with the neglect of surface tension, is given by
Annu. Rev. Fluid Mech. 1977.9:145-185. Downloaded from www.annualreviews.org

R = -
[1 +1p R.5(poo - Pv) J JR o. (3.5)
Under the same assumptions one can obtain an expression similar to (3.3), valid
for the collapse of the bubble starting from some initial radius Ri. If the initial
velocity is taken to vanish, one has

(3. 6)

This equation would predict that the velocity approaches infinity as R - 3/2 as
R -+ 0, which is unacceptable. Therefore, one must conclude that the approximations
under which Equation (3.6) is obtained will eventually break down during the
collapse process. We return to this point below. Let us observe here that, in the
absence of surface-tension effects, one can compute from Equation (3.6) the time to

[(5/6) [
required for complete collapse of the cavity :
3np 1/2 ( P
)1/2
t o = f(1/3) 2(poo - v)
P J Ri :::e O.9 1 5
Pro - Pv
R
;,
(3.7)

which is the result of Rayleigh (19 1 7). An extension to the collapse of a closed cavity
of arbitrary shape has been given by Miles ( 1 966), who, for oblate and prolate
spheroids, finds a correction to (3.7) of the order of the fourth power of the
eccentricity (the quantity Ri is defined as the radius of a sphere of equivalent
volume for this problem).
Let us now return to the question of the dynamics of the bubble in the later
stages of the collapse. An important feature of this problem that should be kept in
mind in the following discussion is that the spherical configuration of the bubble
surface is unstable during the collapse, so that analyses based on the assumption
of spherical symmetry cannot be rigorously correct. Nevertheless, this assumption
furnishes an order-of-magnitude estimate of the several quantities involved, and it
gives a qualitative picture of the phenomenon. It is in this spirit that the following
considerations should be interpreted. The most obvious reason for the nonphysical
behavior of (3.6) for R � Ri is the neglect of liquid compressibility which becomes
important as soon as bubble-wall velocities become comparable with the speed of
sound in the liquid. A very successful modification of the Rayleigh equation that
takes into account liquid compressibility was obtained by Gilmore (1952 ; see also
Plesset 1 969) on the basis of the Kirkwood-Bethe ( 1 942) approximation. This
 166 PLESSET & PROSPERETTI

approximation consists in assuming that the quantity r(h + 1u2), with U the liquid
velocity and h the enthalpy, is propagated unaltered along the outgoing characteris­
tics, dr =" (u + c) dt ; that is,
D [ o [
1 2) + c -
- r(h + L:u 1 2)
J or r(h + L:u J = O. (3 .8)
Dt

In this equation c = c(p) denotes the speed of sound in the liquid, and D/Dt =

a/at + u% r is the material derivative. The derivatives with respect to r in the second
by Stanford University - Main Campus - Lane Medical Library on 10/19/12. For personal use only.

term of Equation (3.8) can be eliminated with the aid of the momentum equation,
Du/Dt "" - oh/or, and of the continuity equation,
Annu. Rev. Fluid Mech. 1977.9:145-185. Downloaded from www.annualreviews.org

ou 1 Dh 2u
-
or - c2 Dt -;:-

The result then leads to the following equation for the bubble wall
(1 1 . ) .
R RR. +i(R)2 1
. ( 1 .
3C R
) ( H 1 +
1 . ) (
R + 1 )
1 . R .
e R e H. (3.9)
C C
- - = -

H and C denote here the values of the quantities h and c at the bubble wall. If
surface ten si on
and viscosity are omitted from the boundary conditions and the
bubble internal pressure Pv is taken to be a constant, Equation (3.9) can be integrated
analytically once to obtain
ft.
10 g �
Ri
-
-
-2 J Ro
U ( U - C) d U
U 3 - 3 C U 2 + 2H U + 2HC '
(3. 10)

The bubble-wall velocity, R . has been written as U in this equation. The initial
conditions correspond to a bubble of radius Ri that at t 0 undergoes a step =

increase in the ambient pressure. The initial "release" velocity Ro for this situation
is easily computed from (3.9) and found to be given by R o � (pv P oo )/p", coo , where -

P oo and Coo are the values of the liquid density and sound speed at large distance
from the bubble. In the approximation I H I � C2, Equation (3. 10) gives the following
correction to Equation (3.6) :
( ) (1
-
Ri 3
= -
1 R 4
-- 1
)( 3 + - --
P ro .
R2
) , (3 . 1 1 )
R 3 C 2 Poo - Pv
from which we find R ex R 1/2 as R 0, in contrast to the incompressible approxi­
- -+

mation that was shown to give R ex R 3/ 2 . -

The accuracy of Equation (3.9) can be assessed through a comparison with the
full solution of the problem which can be obtained only numerically. Such an
analysis was conducted by Hickling & Plesset (1964), who considered both empty
cavities (i.e. essentially Poo Pv = constant) and cavities with a very small gaseous
-

content that undergo isentropic or isothermal compression. An example of their

results is shown in Figure 8, which shows that Equation (3.9) is accurate up to
bubble-wall Mach numbers of the order of 5. Their analysis included also the
details of the liquid flow, with the computation continued past the rebound point
of the gas bubble up to the formation of the shock wave in the liquid. They found
 BUBBLE DYNAMICS A N D CAVITATION 167

0: p. " I ATMOS
7 " 1.4
�
::I INCOMPRESSIILE -- - - ­
� 10
z KIRKWOOD - BE THE --
EXACT - - - -
Q
'"
::I
...J
...J
�
'"
...J
CD
by Stanford University - Main Campus - Lane Medical Library on 10/19/12. For personal use only.

CD
�
CD

:::>1<..>
Annu. Rev. Fluid Mech. 1977.9:145-185. Downloaded from www.annualreviews.org

1 02 -.
10 1 62 i6'
R / R ; . BUBBLE RADIUS

Figure 8 The bubble-wall Mach number as a function of the dimensionless radius R!Rj
for a gas bubble collapsing in a compressible liquid (from Hickling & Plesset 1 964).

that the peak pressure has an approximate l /r dependence during the rebound, with
pressures in the cavity of the order of tens of kilobars. Results of the same order
of magnitude were obtained by Flynn (1975b). While this estimate of the internal
pressure (and hence of the pressure distribution in the neighboring liquid) may be
off by as much as an order of magnitude, it appears likely that the l /r dependence
would not be greatly affected by a more accurate description of the latest stages of
the collapse. It may be of interest to observe that the Kirkwood-Bethe approximation
gives good results for collapse up to relatively large Mach numbers since the
collapse motion leads to an expansion wave in the liquid. For bubble growth at
very large velocities, the same accuracy could not be expected since this motion
would lead to the propagation of a shock wave into the liquid.
Following Benjamin (1 958), Jahsman ( 1968) has given an analytic treatment of
the collapse of a gas bubble in a compressible. liquid based on a perturbation
expansion in terms of a small parameter that is essentially the liquid compres·sibility.
His results show that the Kirkwood-Bethe assumption is verified for the zero and
first-order terms of the expansion, but breaks down in higher orders. In this
connection mention should be made also of a work by Hunter ( 1 960), who, guided
by a numerical investigation, developed a similarity solution for the collapse of an
empty cavity in the neighborhood of the collapse point in the region where c � coo ;
he found that in the latest stages of the collapse R ex R - 0.8. It should be remarked
that the early work on this subject, summarized by Cole (1948, see also Trilling
1 952), was motivated by interest in underwater explosions. Most of the later
contributions discussed above were undertaken to investigate the mechanism of
cavitation damage, since it was conjectured that the very strong pressure pulses
radiated by a collapsing cavity might be responsible for it The current under­
standing of cavitation damage, however, is quite different, as is discussed in the
following section.
In addition to the neglect of compressibility effects, a second obvious reason for
 1 68 PLESSET & PROSPERETTI

the failure of Equation (3.6) to model in an acceptable way the physical behavior
of a collapsing bubble is the neglect of the variation of Pv that is brought about
not only by the increasing importance of even small quantities of noncondensible
gas in the cavity as R -> 0 but also by the fact that condensation of the vapor
cannot keep up with the bubble-wall motion when its velocity becomes of the order
of the speed of sound in the vapor. Some remarks on this aspect of the problem
are made below.
A quantitative experimental confirmation of the collapse time predicted by
by Stanford University - Main Campus - Lane Medical Library on 10/19/12. For personal use only.

Equation (3.7) was given by Lauterborn ( 1 972a), who realized a very close physical
approximation to the Rayleigh empty-bubble model by focusing a giant pulse of a
Annu. Rev. Fluid Mech. 1977.9:145-185. Downloaded from www.annualreviews.org

Q-switched ruby laser on an interior point of a liquid mass (Lauterborn 1 972b). He

reported an experimental value of to 297 X 1 0 - 6 sec to be compared with a
=

theoretical value of 300 x 10- 6 sec for a 0.378-cm bubble in water. It should be
remarked that the final stage of the collapse is so rapid that Equation (3.7) is
negligibly affected by the liquid compressibility, the behavior of the bubble content,
or the deviation from spherical shape discussed below.
The first investigation of the behavior of a cavitation bubble in a time-dependent
pressure field was carried out some time ago by the integration of the Rayleigh
equation with constant Pi for a bubble entrained in a liquid flowing past a submerged
object (PIesset 1 949). For the ambient pressure Poo (t) he made use of the experi­
mentally determined pressure distribution along the object in the absence of the
bubble at the point occupied by the bubble at time t, and he obtained a very good
agreement with the data. Another interesting study of cavitation-bubble dynamics
in a time-varying pressure field has been made by Hsieh ( 1970), who obtained
approximate analytic expressions for the maximum radius that a cavity would
attain in a viscous liquid subject to a transient pressure pulse. From his results he
concluded that the empirical relation P'h OC /10.2, obtained by Bull ( 1956) with a
stress-wave technique, between liquid viscosity and threshold pressure for cavitation
inception appears to be due more to a coincidental distribution of nuclei in the
different liquids tested than to other more basic features of the cavitation process.
The Growth of Vapor Bubbles
As mentioned above, a vapor bubble will designate a bubble in the dynamics of
which thermal effects play a dominant role. I t is supposed that the nucleus from
which the bubble will eventually grow is a small spherical cavity of radius Ro in a
liquid at uniform temperature Too . At equilibrium, the internal pressure in the
nucleus will be the equilibrium vapor pressure corresponding to the liquid tempera­
ture, p�(Too), and the equilibrium radius is
20"
Ro = �.1�
Pv ( Too) - Poo
•

where as usual Poo denotes the ambient pressure, which is here taken to be constant.
The pressure Poo corresponds to a well-determined equilibrium temperature, a
boiling temperature, which we denote by Tb. Equation (3. 1 2) implies Too > Tb ; the
difference � T = Too - Tb is termed the liquid superheat, This simple model for the
 BUBBLE DYNAMICS AND CAVITATION 169
vapor nucleus is certainly highly idealized, but it can be shown that initial conditions
do not affect significantly the growth of the bubble.
The process of bubble growth in superheated liquids can readily be described
in physical terms as follows. When the equilibrium situation depicted by (3. 1 2) is
disturbed, the bubble starts to grow very slowly under the restraining effect of
surface tension. If the initial superheat is sufficiently large, the growth velocity will
eventually reach the asymptotic value for a cavitation bubble, as given by (3.4),
before the rate of vapor inflow (which is proportional to R 2) is so large as to produce
by Stanford University - Main Campus - Lane Medical Library on 10/19/12. For personal use only.

a substantial cooling of the surrounding liquid. At this point, both inertial and
thermal effects limit the growth rate. The growth rate then begins to decrease making
Annu. Rev. Fluid Mech. 1977.9:145-185. Downloaded from www.annualreviews.org

inertial effects less and less important until the radius has grown so large that the
growth process is limited only by the rate at which heat can be supplied to the
bubble wall. The velocity of growth during this asymptotic stage is readily estimated
by noting that the bubble internal pressure will have decreased nearly to Pc m so that
the bubble surface temperature is Tb. The heat flow into the bubble from the liquid
is then approximately 4nR 2k/I'l T/(D/ t)112, where D/ and k/ are the liquid thermal
diffusivity and conductivity, respectively, and t is the time from inception of the
growth. This heat flux will be balanced by the absorption of the latent heat
necessary to vaporize the liquid into the bubble, which is given by 4nR2 Lp�(Tb)R
where L is the latent heat and P�(Tb) is the equilibrium vapor density corresponding

()
to the boiling temperature. Equating the above quantities, we find
. 3 1 /2 k/ T - Tb
R- . (3. 1 3)
00

- -; Lp�(Tb) (D/t) 11 2

The factor (3/n) 112 has been inserted in (3. 1 3) so as to give the result that is
obtained in the asymptotic analysis of Piesset & Zwick ( 1954). Integration of
Equ ation (3. 13) leads to R ex t 1 / 2 , a widely used result in bubble-growth theory. It
should be emphasized that Equation (3. 13) expresses only an asymptotic relation­
ship that is valid only for times large enough so that the growth velocity predicted
by it is smaller than the inertia-controlled value (3.4). Indeed, this remark enables
one to make a distinction between a cavitation and a boiling bubble since the
'
characteristic growth rate for cavitation-bubble growth is (3.4) and the characteristic
growth rate for the boiling-bubble growth is (3. 1 3) (in this connection, see also
Brennen 1 973).
The complete mathematical formulation of the problem of spherical-bubble
growth is similar to the one indicated above for the mass-diffusion problem inasmuch
as one has a partial differential equation for the energy (or temperature) in the
liquid,

(3. 14)

that is coupled to the Rayleigh equation. For the time being we assume that the
internal pressure in the bubble is given by the equilibrium vapor pressure corre­
sponding to the bubble surface temperature as determined by (3. 14). Some comments
 1 70 PLESSET & PROSPERETTI

on nonequilibrium effects are made below. The initial and boundary conditions for
Equation (3. 14) are
T(r, 0) = Too , (3. 1 5a)
T(r, t ) -> Too as r -> 00 , (3. 1 5b)
aT d 4
- 4rrR 2 k[ = L - b'np�R 3) , at I' = R (t), (3. 15c)
or
-

dt
by Stanford University - Main Campus - Lane Medical Library on 10/19/12. For personal use only.

where the last equation is an expression of the heat balance at the bubble boundary.
A general solution to the problem posed by (3. 1 4) and (3. 1 5) was obtained by
Annu. Rev. Fluid Mech. 1977.9:145-185. Downloaded from www.annualreviews.org

Plesset & Zwick ( 1 952) under the assumption that appreciable temperature gradients
are established only in a thin layer surrounding the cavity radius. Their result for
the bubble surface temperature Ts ( t) is

(-;) I t ft 2 aT
R (0) 8r ( R . 0)
Ts ( t) = Too -
DI 1 / 2
[ 1 /2
dO, (3. 1 6)
o R4(Je) d),
e J
where o T(R, t)!or is an arbitrarily specified temperature gradient at I' = R (t). Upon
substitution of (3. 1 5c) for this quantity and with neglect of the temperature depen­
dence of L, P V ' kl• and with the approximation of the pressure-temperature relation
by a straight line, Plesset & Zwick ( 1 954) obtained the following formulation of

( ) [ I"
the problem :
d 2p dP 2 dp
J
2a
p 7/ 3 _2 + lp4/ 3 = 3 I - II
_ (U - V)- 1 / 2 _ (v) dv - - , (3. 1 7)
du du 0 dv p 1/ 3
where p = R 3jR g is the normalized volume and u is a new dimensionless time scale
defined as

u(t) = C(4
Ro
It0
R4(O) dO. (3. 1 8)

The constants C( and J.J are defined by

(3. 1 9)

From (3. 1 7) and (3. 1 8) they then proceeded to obtain various asymptotic expansions
for the radius-time dependence, the leading term of which, for large times, is given
by (3. 1 3) (see also Plesset & Zwick 1 955). The experimental data of Dergarabedian
( 1953, 1 960) obtained for low superheats (up to about 6nC) in water, carbon tetra­
chloride, benzene, and other organic liquids agree very well with the results of
Plesset and Zwick, but for these cases the asymptotic stage is reached so fast that
essentially only the asymptotic results (3. 1 3) could be tested with the data. The same
can be said of the work of Niino et al ( 1 973), who obtained 'superheats of up to
 BUBBLE DYNAMICS AND CAVITATION 171
14°C. A close experimental simulation with high superheats of the theoretical
conditions to which the complete Plesset-Zwick theory applies is rather difficult
because large thermal gradients in the liquid can be easily generated, and further
because observation of the bubble in the inertia-controlled phase of the growth,
when the radius is small and the growth velocity large, is difficult. The best support
of the theory (at least as a mathematical approximation to a more complete set of
equations) can be found in comparing its predictions with the results of some
extensive numerical computations per for med by Daile Donne & Ferranti ( 1 975),
by Stanford University - Main Campus - Lane Medical Library on 10/19/12. For personal use only.

who solved the complete energy equation, Equation (3. 1 4), simultaneously with the
Rayleigh equation. Their investigation was primarily motivated by doubts about
Annu. Rev. Fluid Mech. 1977.9:145-185. Downloaded from www.annualreviews.org

the validity of the thin-thermal-layer assumption for liquids of high thermal con­
ductivity such as liquid sodium. A detailed comparison of the Plesset-Zwick theory
with Daile Donne and Ferranti's results is available elsewhere (Plesset & Prosperetti
1 976, 1977). In summary it can be said that the agreement with Equation (3. 1 5) is
extremely good up to liquid-sodium superheats of 300°C provided that the liquid
and vapor properties are evaluated at the boiling temperature as indicated in
Equations (3. \ 9). An example of this comparison is given in Figure 9. The physical
reason for this result is simply that at very high superheats inertial effects play a

p 1 aIm
1 0-4 em
=

R.=

6T = 1 3 3 . I ·C

R, cm

, sec

Figure 9 Comparison of the results computed with the aid o f the Plesset-Zwick formu­
lation [Equation (3. 1 7), full line] with those obtained by Daile Donne & Ferranti ( 1 975)
by means of numerical integration of the complete equation (3.14) (open circles) for the
growth of a vapor bubble in su perheated sodium (from Plesset & Pros peretti 1 976).
 1 72 PLESSET & PROSPERETTI

dominant role essentially until the bubble surface has cooled to the boiling
ternperature.
Plesset & Prosperetti ( 1 976) have also shown that for p � 1 Equation (3. 1 7)
admits a scaling such that a universal bubble-growth law for any liquid and any
superheat is obtained in the form
S = S(r)
where
by Stanford University - Main Campus - Lane Medical Library on 10/19/12. For personal use only.

S = p2R/Ro, (3.20)
Annu. Rev. Fluid Mech. 1977.9:145-185. Downloaded from www.annualreviews.org

Several examples of growth velocities R(t) scaled according to (3.20) (i.e. graphs
of dS/dt versus t) for different values of the parameter Jl are shown in Figure 10.
The dots on the curves mark the point where R/Ro = 10 ; it is seen that approxi­
mately from this point on, all the curves fall on a single line which demonstrates
the validity of the scaling law (3.20). Similar results are obtained for the radius. In
the dimensionless variable of Equation (3.20) the inertia-controlled growth velocity
(3.4) has the constant value dS/dt (2/3)1/ 2 , and the thermally controlled one given
=

by (3. 1 3 ) is dS/dt = n - I(3tr 1/ 2 . This curve is also shown in Figure 10, and it is
apparent that it describes a substantial part of the growth only for relatively small
subcooling (large values of Jl). It is also shown by Plesset & Prosperetti (1976)
that the analytic expression for bubble growth obtained by Mikic, Rohsenow &
Griffith (1970) can be written in terms of the scaled variables S, t provided that the
pressure-temperature relationship is approximated not by a tangent but by a chord
(see Theofanous & Patel 1976). This expression is
2
S = 2 (i)3/2[(h2t + 1)3/2 _ {1n2 t) 3/2 - 1] , (3 . 2 1 )
n
and its derivative is also plotted in Figure 10. It is seen that Equation (3.2 1 ) is a
good approximation except in the region where inertial and thermal effects are of
comparable importance.
In addition to the studies mentioned, other attempts have been made at deducing
a scaling law for vapor-bubble growth. We note here those of Birkhoff et al ( 1958).
and of Scriven ( 1959), who showed that with the assumption R = fJtl/Z, the energy
equation (3. 14) admits a similarity variable that reduces it to an ordinary differential
equation that can be solved in closed form. Insertion of the result into the boundary
condition (3.1 5c) determines then the constant fJ as a function of the several para­
meters of the problem. For low superheats, it is found that the value of fJ coincides
with the one obtained upon integration of (3. 1 3). The limitations of this approach
should be clear from the discussion given here. In particular, it is seen from Figure 10
that a tl/2 dependence of the radius cannot account for dynamical effects, so that
the applicability of such an analysis must be restricted to low superheats. In the
problem of mass diffusion, however, which has the same mathematical structure,
but for which inertial effects are in general much less pronounced, the similarity
approach is quite appropriate. A very useful feature of it for the application to the
mass-diffusion problem is that the diffusion equation is solved exactly without any
assumption concerning the thickness of the boundary layer.
 BUBBLE DYNAMICS AND CA VITAnON 1 73

Vapor-Bubble Collapse
The theoretical modeling of the collapse of vapor bubbles under conditions in which
thermal effects play a significant role is more difficult than the analysis of their
growth, and no entirely satisfactory theoretical results are available. The principal
difficulty here is that, unlike the growth case, the thickness of the liquid layer in
which substantial temperature gradients are present cannot in general be taken to
be small compared with the bubble radius for all times. An approach that makes
by Stanford University - Main Campus - Lane Medical Library on 10/19/12. For personal use only.

use of the Piesset-Zwick result (3. 1 6) for the surface temperature has been followed
by Florschuetz & Chao (1965), but the applicability of that expression to the collapse
Annu. Rev. Fluid Mech. 1977.9:145-185. Downloaded from www.annualreviews.org

case is open to doubt, especially for those cases in which large thermal effects result
in small collapse velocitics. Among thc notable fcatures of the rcsults of Florschuetz
and Chao is a nonmonotonic decrease in the bubble radius when both inertial and
thermal effects are important in the collapse. The experimental evidence in favor of
this point appears to be rather unclear for the collapse of spherical bubbles in
unbounded liquids (LevenspieI 1959, Akiyama 1965, Hewitt & Parker 1 968), possibly
because shape deformations during the coll ap se introduce errors in th e preci se
determination of the radius. It may be mentioned, however, that no such behavior
was observed by Gunther ( 1 9 5 1 ) for the collapse of hemispherical bubbles on a
heated solid wall. Florschuetz and Chao also reported experiments performed with
the aid of a free-fall tower to minimize the effects of translational motion and initial
buoyancy-induced deformations. They observed that the shape of all of the bubbles
deviated from the spherical during the collapse. Whilc for small subcoolings
(� Tj ::5 1 3 °C) they appeared to oscillate in the lowest mode between prolate and
oblate spheroidal configurations, for larger subcoolings much more marked devi­
ations occurred with the formation of jets and large deformations beginning early
in the collapse process.

I' = 6.9 X 1 0-5

1 .0

dS
dt'

0.5

·r

Figure 10 Dimensionless growth velocity versus dimensionless time for vapor-bubble

growth, for several values of the parameter 1-', which shows the scaled growth behavior.
The open circles denote the points where the bubble radius equals ten times its initial
value. The dashed line is the Mikic- Rohsenow-Griffith (1 970) result (3.21), and the line labeled
"thermal" is the asymptotic result of Plesset & Zwick ( 1 954), Equation (3. 1 3) (from Plesset
& Prosperetti 1976).
 174 PLESSET & PROSPERETTI

Bubble collapses at high subcoolings (up to 60°C) in water have been obtained
by Board & Kimpton ( 1974), whose data however appear to be affected by strong
wall effects. Very recently Delmas & Angelino ( 1976) have studied the collapse of
rising bubbles with subcoolings L1 T; between TC and 42°C in water. They reported
two different collapse modes, one for large bubbles and high subcoolings, for which
inertial effects are important, the other for smaller bubbles and lower subcoolings.
The bubbles belonging to the first class appear to occupy a well-defined region
in the (R;. L1 T;)-plane. They exhibit large collapse velocities and rapid « 1 .4 X
by Stanford University - Main Campus - Lane Medical Library on 10/19/12. For personal use only.

10-4 sec) fragmentation upon attainment of the minimum radius. The initial
conditions giving rise to the second collapse mode are less predictable. The bubbles
Annu. Rev. Fluid Mech. 1977.9:145-185. Downloaded from www.annualreviews.org

present an irregular surface and selected condensation sites. Fragmentation takes

place earlier than in the other case and the smaller bubbles to which it gives rise
continue to condense. These results are interesting, but it is difficult to estimate the
effect of the velocity of rise of the bubbles on which the authors did not make
any comment.
Other Topics in the Dynamics of Vapor Bubbles
In recent years research in the acoustic cavitation process in cryogenic liquids (see
e.g. Masse & Finch 197 1 , Neppiras & F inch 1972, Akulichev 1974) and in the
development of ultrasonic bubble chambers (Akulichev et al 1970 ; Brown et a1 1970 ;
Hilke 1973 ; Shestakov & Tkachev 1973, 1 975) have prompted the theoretical
investigation of the behavior of vapor bubbles in oscillating pressure fields (Finch
& Neppiras 1 973, Wang 1 973). In these studies, the energy equation in the liquid
and the Rayleigh equation have been solved under the assumption of small-amplitude
oscillations, allowing for the evaporation-condensation coupling expressed by
Equation (3. 1 5c). It has been found that when the liquid temperature is sufficiently
elose to the saturation temperature, there are two values of the bubble radius that
give rise to resonant oscillations for a fixed driving frequency. One of them
corresponds to that found for a permanent gas bubble, while the other one is
smaller, approximately by an order of magnitude for the case of water at one
atmosphere and at frequencies of the order of 104 Hz. T h e resonance peaks are
broader than for gas bubbles, and they merge to give a large oscillation amplitude
for values of the radius spanning one or two orders of magnitude. In addition to
this effect of the evaporation of the liquid, Wang (1973) has also considered the
simultaneous diffusion of a permanent gas into the bubble.
In these studies, the thermodynamic state of the vapor in the bubble has been
assumed to correspond to conditions of thermodynamic equilibrium at the instan­
taneous bubble-wall temperature. In a certain sense, this assumption corresponds
to the validity of Henry's law, Equation (2. 1 1), in the case of rectified diffusion of
mass into the bubble. Therefore, one may expect a rectified heat-transfer effect into
the bubble by means of which the average vapor content of the bubble increases
with time. Such a process has indeed been demonstrated theoretically (Wang 1 974),
but no experimental verification of the theory is available. It should be noticed
that this effect would be of first (rather than second) order in the oscillation
amplitude if thermodynamic nonequilibrium effects for the vapor were accounted
for. Whether this increased order of magnitude would give rise to a contribution
 BUBBLE DYNAMICS AND CA VITA nON 1 75

bigger than the rectified effect depends strongly on the accommodation coefficients
for evaporation and condensation. This aspect of the problem has not as yet been
investigated.
More generally, the question of effects associated with a state of thermodynamic
nonequilibrium between liquid and vapor is of course an important one in the
dynamics of vapor bubbles. An estimate of the difference between the equilibrium
vapor density p � and the actual vapor density Pv is readily obtained by a mass
balance over the bubble
by Stanford University - Main Campus - Lane Medical Library on 10/19/12. For personal use only.

(3.22)
Annu. Rev. Fluid Mech. 1977.9:145-185. Downloaded from www.annualreviews.org

where J is the net mass flux at the bubble wall which, on the basis of the elementary

(R )
kinetic theory of gases, can be written as
T 1/2
2:M
= (J. (p�- Pv) ' (3.2 3)
J

Here Rg is the gas constant, M is the molecular weight of the vapor, and 0( is the
accommodation coefficient for evaporation, taken to be equal to that for conden­
sation. Continuity of temperature across the bubble surface has also been assumed
in (3.23). Substitution of (3.23) into (3.22) and neglect of the term involving dPv /dt
gives then (plesset 1 964) :

P�- P v R
- 0((R T/2nM)1/2 + R " (3.24)
P� g
In view of the fact that the sound speed in the vapor is given by c = (yRgT/M)I/2,
this equation shows that the nonequilibrium correction is of the order of the Mach
number of the bubble-wall motion in the vapor whenever rJ. � 1, and therefore
negligible except perhaps near the end of violent collapses. Much more significant
effects may be expected, however, for small values of 0(. Nonequilibrium analyses
of vapor-bubble growth have been given by Bornhorst & Hatsopoulos ( 1 967) and
by Theofanous et al ( 1 969a). As one might expect, very small values of 0( give rise
to decreased growth rates, but the results corresponding to 0( = 1 are in very good
agreement with the equilibrium behavior. Theofanous et al ( 1 969b) have treated in
a similar way the case of the collapse of vapor bubbles, although their assumption
of a thin thermal layer may be questionable in this case for the reasons already
stated.
In addition to the neglect of thermodynamic equilibrium, the formulation of
vapor-bubble dynamics given here has been simplified in other respects. For instance,
the equation of continuity at the bubble surface gives

(3.25)
(where u, and u" are the liquid and vapor velocities at the liquid-vapor interface,
respectively), from which
.
R= �� --
PI
PI- Pv
UI +
Pv
PI - Pv
Uv ' ( 3 . 2 6)
 1 76 PL I'SSET & PROSPERETTI

·
Clearly, setting Ul R" .: lls :was implicitly done in the derivation of the Rayleigh
==

equation, is legitimate�Mly if Pv � Pl' While this relation is normally satisfied by

liquids at temperature; sufficiently below the critical point, its applicability in some
special situations (si:!cb as cryogenic liquids) should not be automatically taken for
'
granted. Similarly, tn� condition of continuity of normal stresses, Equation ( 1.4),
(
and of conservation o energy at the interface, Equation (3. 1 5c), in principle should
contain other terms of order Pv/Pl' For a more detailed exposition of these matters
the reader is referred to Scriven (1 959) and Hsieh (1965). The latter author gives also
by Stanford University - Main Campus - Lane Medical Library on 10/19/12. For personal use only.

a usefttl unified treatment of several mathematical aspects of the whole subject of

bubble dynamics.
Annu. Rev. Fluid Mech. 1977.9:145-185. Downloaded from www.annualreviews.org

4 THE DYNAMICS OF NONSPHERICAL BUBBLES

It is obvious that as soon as the property of spherical symmetry is lost, the analysis
of the various aspects of bubble dynamics becomes exceedingly complex both from
the theoretical and the experimental standpoint (Hsieh 1972b). This situation is
unfortunate, in view of the practical importance of the effects associated with
deviations from the spherical shape and in view of the multitude of factors that
promote such deviations. Among these one may mention the inherent dynamical
instability of contracting bubbles, the proximity of solid boundaries or of free
surfaces, and buoyancy effects. Such important phenomena as bubble breakup or
coalescence and cavitation damage are dominated by these effects which are also
found to increase heat and mass-transfer rates.
The first problem that is encountered in the dynamics of nonspherical bubbles is
that of the specification of the bubble shape. An obvious possibility, which is used
in practically all the work described below, is the use of an expansion in terms of
spherical harmonics Y:;'((J, cp),

S(r, (J, cp, t) = r- R (t)- L anm (t) Y:;'((J, cp), (4.1)

n .m
where S (r, e, qJ, t) = 0 is the equation of the bubble surface, R(t) is the instantaneous
average radius, and anm (t) is the amplitude of the spherical harmonic component
of order n and degree m. It should be emphasized, however, that (4. 1 ) is not the only
possible choice, nor is it the most convenient for all problems.
A basic result in the dynamics of nonspherical bubbles has been obtained by
Plesset ( 1954), who solved the equation of motion of an incompressible, inviscid,
unbounded fluid with a free surface given by (4. 1 ) in the small-amplitude approxi­
mation I an," I/R � 1. In this approximation the equations for R and the anm's are
uncoupled, the first one being just the Rayleigh equation. The equations for the
anm's do not contain the index m (which accordingly will be dropped in the following)
and are found to be
(4.2)
If the surface-tension term is neglected, this equation reduces to one also given by
Birkhoff ( 1 954). It is easily seen that n = 1 corresponds to a translation of the bubble
 BUBBLE DYNAMICS AND CAVITATION 1 77
center. In this case Equation (4.2) gives just R3a1 = constant, which expresses the
constancy of the total liquid momentum which is proportional to the bubble volume
times its translational velocity. If all the other a's vanish this happens of course to
be im exact result independent of the small-amplitude approximation. This situation,
however, is of little interest to us here,
. and we shall restrict our considerations to
the case n � 2 in the following.
The introduction of viscous effects complicates the matter substantially, giving to
Equation (5.2) an integro-differential structure (Prosperetti 1976c) except in the case
by Stanford University - Main Campus - Lane Medical Library on 10/19/12. For personal use only.

where the viscous-diffusion length is small compared with the bubble radius, for
which it becomes
Annu. Rev. Fluid Mech. 1977.9:145-185. Downloaded from www.annualreviews.org

an + [3R/R + 2(n + 2)(2n + l) v,/R2Jan

+ (n - 1)[(n + l)(n + 2)0"/PIR3 + 2(n + 2)vI R/R 3 - R/RJan = 0, (4.3)

VI denoting the liquid
kinematic viscosity. Both Plesset (1954) and Prosperetti (1 976c)
have also considered the dynamical effects of the fluid contained in the bubble. The
expressions given above are the limiting forms obtained when its density and viscosity
vanish.
The Stability of Spherical Growth and Collapse
Even with the neglect of the surface-tension term (which has always a stabilizing
effect, but which is important only for very small bubble radii), the qualitative
behavior of the solution of Equation (4.2) is not readily understood a priori. Indeed,
the stability characteristics do not depend only on the acceleration of the interface,
as in the familiar plane case of the Rayleigh-Taylor instability, but also on its
velocity. This situation is clearly a consequence of the geometry, since the divergence
of the streamlines during bubble growth has a stabilizing effect, while the reverse
occurs during collapse. In the limiting case of very large acceleration and of small
velocity, such as take place for instance during the early stages of the growth of an
underwater-explosion bubble, Equation (4.2) predicts an instability of the spherical
shape, which is indeed known to exist (Cole 1948, illustration facing p. 247). Another
example of the occurrence of this situation is furnished by the large-amplitude radial
pulsations of a permanent gas bubble, for which the radial acceleration becomes
very large and positive near the position of minimum radius, with the consequent
possibility of large deformations and breakup. A similar situation might be expected
to take place also in the early stages of the growth of a cavitation or boiling bubble,
but here the duration of this stage of large acceleration and small velocity is so
short that the instability does not have the time to develop.
The quantitative investigation of the stability for a bubble expanding or collapsing
under a fixed pressure difference P = Pi - p", (cavitation bubble) has been carried
out by PIesset & Mitchell (1956), who were able to obtain closed-form solutions of
Equation (4.2). The qualitative features of their results can be readily understood
with the aid of Equations (3.4) and (3.6). For the growth case we have 1{2 c::: (2/3)P/p.
from which one readily obtains an -+ constant as R -+ OCJ . Hence, although on the
basis of the plane Rayleigh-Taylor case one would expect an unstable behavior
 1 78 PLESSET & PROSPERETTI

during the growth, in practice no significant deviations from the spherical shape
take place because the acceleration is large only for a small fraction of the process,
and its destabilizing influence is very effectively counterbalanced by the stretching
of the bubble surface produced by the divergence of the streamlines. It may also
be remarked that, in spite of the factor (n - 1) multiplying the radial acceleration
in Equation (4.2), high growth rates for instabilities of large order n do not take place
because of viscous effects [cf Equation (4.3)J .
For the study of the collapse of a cavitation bubble it is expedient to set bn
by Stanford University - Main Campus - Lane Medical Library on 10/19/12. For personal use only.

(RdR)3/2an. so that Equation (4.2) takes the form

bn - [(3/4) (J�/R) 2 + ( n + !)i�/R - (n - l ) ( n + 1 ) (n + 2) a/p,R3]bn = O. (4.4)

Annu. Rev. Fluid Mech. 1977.9:145-185. Downloaded from www.annualreviews.org

Now, from Equation (3.6) we have during collapse that R2 ::: - (2/3) (P/Pl)(RjR)3
from which the coefficient of bn in this equation is seen to have the form - nc2R 5 ,
where c is a constant. An approximation of the W.K.B. type gives then
-
-4 (
an ::: R 1 / exp ± icn 1 / 2 f )
R - 5 / 2 dt ,

-
whence it is seen that an increases proportionally to R 1 /4 and oscillates with
increasing frequency as R 0 (Piesset & Mitchell 1956 ; Birkhoff 1954, 1956). The
-->

decisive role played by the sink like converging nature of the flow in the development
of this instability is apparent from the second term of Equation (4.2).
Clearly, Equation (4.4) implies the breakup of the bubble when the amplitude
of the oscillations becomes of the order of R. The results of the linearized approxi­
mation, however, cannot be expected ·to be quantitatively valid up to this point. A
fully nonlinear numerical calculation of the shape of an empty bubble during the
collapse has been carried out by Chapman & Plesset (1972). They have obtained
local bubble-wall velocities of the order of hundreds of meters per second at the
points where the breakup of the bubble takes place. They have also shown that the
linearized result is surprisingly accurate for most of the duration of the collapsc.
The greatest error in the description of the bubble shape is caused by the growth
of harmonic components of different orders from those initially present, which is
induced by the nonlinear couplings and which cannot be predicted on the basis of
the linear theory.
The instability of the spherical shape of collapsing vapor bubbles at high sub­
coolings is well documented in the literature (see e.g. Florschuetz & Chao 1965,
Figure 1 1 ; and Lauterborn 1974). However, no reduction of the data in terms of
spherical harmonic components has yet been attempted.
Bubble Col/apse in the Vicinity of a Rigid Boundary
We have seen that, in principle, deviations from the spherical shape for a bubble in
an unbounded liquid can take place only through the amplification of preexisting
small perturbations. No such initial perturbation, however, is necessary for the
occurrence of deformations of a bubble near a boundary, because the asymmetry
of the flow that the boundary itself introduces is sufficient to give rise to highly
distorted bubble shapes. It is possible to get an insight into these effects by observing
 BUBBLE DYNAMICS AND CAVITATION 1 79

that, with the neglect of viscosity, the potential problem can be solved by the
method of images that replaces the rigid boundary by an image bubble equal to
the real one and located symmetrically to it with respect to the boundary. From
this observation one is led to expect that the portion of the bubble farther from
the wall acquires a greater velocity than the one near the wall, because there the
velocity of the flow induced by the image bubble adds to the collapse velocity,
while in the second case it decreases the collapse velocity. This asymmetry leads
to the formation of a high-velocity jet directed toward the wall, while the overall
by Stanford University - Main Campus - Lane Medical Library on 10/19/12. For personal use only.

characteristics of the sink like flow of the image attracts the bubble toward the wall
(Lauterborn & Bolle 1975).
Annu. Rev. Fluid Mech. 1977.9:145-185. Downloaded from www.annualreviews.org

That jet formation during bubble collapse could be responsible for cavitation
damage was suggested as early as 1 944 by Kornfeld & Suvorov, but their conjecture
was not explored further, and the accepted explanation remained that of the large
pressures associated with the latest stages of bubble collapse that had originally been
put forward by Rayleigh (1917). The first experimental demonstration of jet
formation was obtained by Naude & Ellis ( 1961), and definitive conclusions about
the mechanism of cavitation damage was later reached by Benjamin & Ellis ( 1966).
This latter paper contains also an extensive summary of the history of the subject
and of related work. Additional considerations on the magnitude of the stresses
induced by jet impingement can be found in PIesset & Chapman ( 1971) and in
Plesset (1 974). Direct measurement of jet velocities is rather difficult. Operating at
low ambient pressure, Benjamin & Ellis (1966) obtained velocities of 10 m sec- 1,
which scale to velocities an order of magnitude greater at an ambient pressure of
one atmosphere. The value of 1 20 m sec - 1 is reported by Kling & Hammitt (1972)
and Lauterborn & Bolle (1975), and is consistent with the numerical results of Plesset
& Chapman (1971).
As might be expected, an analytical treatment of the problem of bubble
collapse in the vicinity of a rigid boundary is very difficult. An early perturbation
approach is that of Rattray (1951), who was successful in predicting the possibility
of jet formation and also in predicting the elongation of the bubble in a direction
perpendicular to the boundary in the early stages of the collapse. The first complete
theoretical analysis of the collapse of an empty cavity in the neighborhood of a
rigid wall was obtained numerically by Plesset & Chapman (1971), whose results
have been experimentally confirmed by Lauterborn & Bolle ( 1975). Figure 1 1, from
the work of these authors, shows a comparison of the theoretical results (continuous
lines) with the experimental ones (open circles). The comparison is quite satisfactory,
particularly in view of the difficulty in establishing a correspondence between
theoretical and experimental initial conditions. The case of a bubble collapsing
attached to a wall has also been worked out numerically by Plesset & Chapman
( 1971).

Surface Oscillations
Before considering some of the phenomena associated with the interplay between
radial motion and surface deformations of permanent gas bubbles, we discuss the
shape oscillations of a bubble of fixed radius Ro. The governing equation for this
 180 PLESSET & PROSPERETTI

o
by Stanford University - Main Campus - Lane Medical Library on 10/19/12. For personal use only.
Annu. Rev. Fluid Mech. 1977.9:145-185. Downloaded from www.annualreviews.org

Figure 1 1 Comparison of the experimental results of Lauterborn & Bolle ( 1975, open
circles) with the theoretical ones of Plesset & Chapman (1971) for the collapse of a
cavitation bubble in the vicinity of a rigid bound ary (from Lauterborn & Bolle 1975).
Bubble shapes corresponding to later values of time than those shown here are given by
P Ies set & Cha pman (1971).

situation is readily obtained from (4.2) or (4.3), and one finds the following expression
for the natural frequency of the nth mode
w5n = (n - l) (n + l ) (n + 2)0"/R gPI' (4.5)
The familiar dispersion relation for capillary waves on a flat liquid surface is readily
obtained from this equation by observing that the wavelength A is given by ). =
2nRo/n and by letting Ro and n tend to infinity with A held fixed. For small viscosity,
one obtains from (4.3) the viscous damping constant fl. of the oscillations as
fln = (n + 2) (2n + l ) vdR 6 (cf Lamb 1932, p. 475, p. 641).
An understanding of the general behavior of the surface deformations for a bubble
undergoing radial pulsations can be obtained in the limit of small-amplitude
oscillations by writing R = Ro(1 +b si n wt) and then retaining only linear terms in
b (Hsieh & Plesset 1 96 1 ; Hsieh 1 972a, 1974a). Equation (4.4) takes then the form
of a Mathieu equation,
 BUBBLE DYNAMICS AND CAVITATION 181

On+ (w5n + Cln sin wt)bn = 0, (4.6)

with Cln = [(n + 1j2)w2 - 3w5n]b and w5n given by (4.5). It is well known from the
theory of the Mathieu equation that, depending on the magnitudes of the constants
IXn and w6n, the solutions of (4.6) may have the form of modulated oscillations the
amplitude of which grows exponentially with time. Although Equation (4.5) itself is
not valid unless I bn I q Ro, this behavior does indicate that parametric excitation
of the instability of the spherical shape is possible, and describes the early stages
by Stanford University - Main Campus - Lane Medical Library on 10/19/12. For personal use only.

of its development. It may be observed that the most easily excited surface
oscillation is the one co.rresponding to n = 2, the threshold of which has a minimum
in the vicinity of w = 2wo. The acoustic emission from a bubble executing surface
Annu. Rev. Fluid Mech. 1977.9:145-185. Downloaded from www.annualreviews.org

oscillations in this mode would contain a sub harmonic component, and indeed this
possibility for the origin of the subharmonic component in acoustically cavitating
liquids has also been considered. However, the observed intensity of the signal is
incompatible with this hypothesis because, as pointed out by Strasberg ( 1956),
bubbles executing shape oscillations are inefficient sound sources because of the
rapid decrease of the velocity potential with distance from the bubble center.
Approximations to Equation (4.4) of a higher order in b than (4.6) have been
carried out by Eller & Crum (1970) in an attempt to explain the observed
instability of the position of a pulsating bubble in <l; sound field first reported by
Benjamin and Strasberg (see Benjamin 1964). Although the connection between the
onset of surface distortions and the observed erratic translatory motion of the
bubble is not obvious, the instability thresholds computed from (4.4) do exhibit a
fair agreement with the experimental data for the onset of the bubble translatory
motion. More direct experimental observations on the onset of surface oscillations
are not very readily obtained in low-viscosity liquids for several reasons. Gould
( 1974) reported some results that have a large scatter but show agreement with
theory in the general trend. Storm ( 1974) has studied large bubbles trapped in a
gel, but the interpretation of his observations is somewhat obscured by the complex
rheological characteristics of the suspending medium. ResultS' of studies with large
bubbles trapped on a plate have also been reported (Gould 1966, Hawkins 1 965,
Blue 1967), but here a difficulty is encountered in accounting for the presence of
the rigid boundary in the theory (in this connection see also Strasberg 1953).
Numerical results for stability thresholds for freely oscillating gas bubbles have
been obtained by Strube (1971), who solved simultaneously the Rayleigh equation
for a freely pulsating gas bubble and Equation (4.4). He presents results on maximum
bubble deformations (which frequently appear shortly after the instant of minimum
radius) and gives a chart of the stability boundaries. No work of this type is
available for forced oscillations. Experiments on the distortion of a translating gas
bubble subject to a pressure step were reported by Smulders & Van Leeuwen
( 1974) and compared with a theoretical analysis by Hermans (1973). Several results
drawn from the theory are confirmed on a qualitative basis although a quantitative
comparison was not attempted by the investigators.
As the last topic of this article, we mention a variational approach to nonspherical
bubble dynamics formulated by Hsieh ( 1974b). The difficulties encountered in the
accurate description of the shape of nonspherical bubbles are such that there
 1 82 PLESSET & PROSPERETTI

certainly is a strong case for attempting a variational attack. The results obtained
so far, however, are rather limited, and it appears that additional research is
necessary before the variational formulation can fulfill its promises of effectiveness
and relative simplicity.

5 CONCLUDING REMARKS

It is clear that, in view of the very large number of papers on the subject, even
by Stanford University - Main Campus - Lane Medical Library on 10/19/12. For personal use only.

fhe present list of references is quite incomplete. Further, in view of the fact that the
present article is the first one on this subject to appear in the Annual Review of
Annu. Rev. Fluid Mech. 1977.9:145-185. Downloaded from www.annualreviews.org

Fluid Mechanics, wc felt that a relatively detailed exposition of the most important
established results was in order even at the expense of omitting some more recent
developments such as the dynamics of bubbles in non-Newtonian liquids. We have
also tried to cover the fundamental physical aspects of the problems with little or no
mention of practical developments. Thus we have been very concise in our coverage
of the literature on bubble dynamics in connection with boiling heat transfer. We
have also given only a summary of the processes and results for acoustic cavitation
and flow cavitation. Finally, we have neglected entirely the phenomena related to
the translatory motion of bubbles that have recently been considered elsewhere by
Harper (1 972).

Literature Cited
Akiyama, M. 1 96S. Bull. Jpn. Soc. Mech. Eng. Borotnikova, M. I., Solollkin, R. 1. 1 964.
8 : 683-94 SOD. Phys. A coust. 1 0 : 28-32
Akulichev, V. A. 1 974. Sov. Phys. Acoust. Brennen, C. 1973. J. Fluids Eng. 95 : 533-41
20 : 94-95 Brown, R. C. A., H arigel, G., Hilke, H . J.
Akulichev, V. A. et al. 1 970. Sov. Phys. 1 970. Nucl. lnstrum. Methods 82 : 327-30
A coust. 1 5 : 439-43 Bull, T. H. 1 956. Br. J. Appl. Phys. 7 : 4 1 6--
Apfel, R. E. 1 970. J. Acoust. Soc. Am. 48 : 18
1 1 79-86 Carslaw, M . S., Jaeger, J . C . 1 959. Conduction
Benjamin, T. B. 1 958. Symp. Naval Hydro­ of Heat in Solids, 2nd ed . Oxford : Claren­
dyn .. 2nd. Washington. D.C.. pp. 207-33 don
Benjamin, T. B. 1964. Cavitation in Real Ceschia, M., Iernetti, G. 1 974. J. Acoust.
ed. R. Davies, pp. 164-80. New
Liquids, Soc. Am. 5 6 : 369-73
Yark : Elsevier Chapman, R. B., Pies set, M. S. 1971. J. Basic
Benjamin, T B., Ellis, A. T. 1 966. Phi/os. Eng. 93 : 373-76
Trans. R. Soc. London Ser. A 260 : 221-40 Chapman, R. 8., Plesset, M. S. 1 972. J. Basic
Birkholf, G. 1 954. Q. Appl. Math. 1 2 : 306--9 Eng. 94 : 1 42-45
Birkholf, G. 1 956. Q. Appl. Math. 1 3 : 4 5 1 -53 Childs, D. R. 1973. Int. J. Non-Linear Mech.
Birkholf, G., Margulies, R. S., Horning, W. A. 8 : 371-79
1 : 201-4
1 958. Phys. Fluids Coakley, W. T 1 9 7 1 . J. Acoust. Soc. Am. 49 :
Blake, F. G. 1949. Harvard Univ. Acoust. 792-801
Res. Lab., Tech. Mem. No. 12 Coakley, W. T, Sanders, M. F. 1 973. J.
Blander, M., Katz, J. L. 1 975. AIChE J. Sound Vib. 28 : 73-85
2 1 : 8 33-48 Cole, R. 1 974. Adv. Heat Transfer 1 0 :
Blue, J. E. 1 967. J. A coust. Soc. Am. 41 : 86-- 1 66
369-72 Cole, R. H. 1 948. Underwater Explosions.
Board, S. J., Kimpton, A. D. 1 974. Chem. Princeton, NJ : Princeton Univ. Press.
Eng. Sci. 29 : 363-71 Reprinted by Dover Publications, New
Bornhorst, W. J. , Hatsopolllos, G. N. 1967. York
J. Appl. Mech. 34 : 847-53 Daile Donne, M., Ferranti, M. P. 1 975. lilt.
 BUBBLE DYNAMICS AND CAVITATION 1 83

J. Heat Mass Transfer 18 : 477-93 59-129

Davidson, B. 1. 1 9 7 1 .
J. Sound Vib. 17: Hermans, W. A. 1973. PhD thesis. Tech.
26 1-70 Univ. Eindhoven, Netherlands
Davidson, B. J., Riley, N. 1 9 7 1 . J. Sound Hewitt, H. c., Parker, J. D. 1 968. J. Heat
Vib. 1 5 : 2 1 7-33 Transfer 90 : 22-26
Delmas, H., Angelino, H. 1976. Int. J. Heat Hickling, R., Plesset, M. S. 1 964. Phys.
M ass Transfer 1 9 : 1 1 8-21 Fluids 7 : 7-14
Dergarabedian, P. 1 953. J. Appl. Mecho 75 : Hilke, H. 1 973. Ultrasonics 1 1 : 51-52
537-45 Hill, C. R. 1972. J. A coust. Soc. Am. 52 :
Dergarabedian, P. 1960. J. Fluid Mech. 9 : 667-72
by Stanford University - Main Campus - Lane Medical Library on 10/19/12. For personal use only.

39-48 Hinsch, K., Bader, F., Lauterborn, W. 1 974.

De Santis, P., Sette, D., Wanderlingh, F. In Finite-Amplitude Wave Effects in Fluids,
1 967. J. A coust. Soc. Am. 42 : 5 1 4- 1 6 ed. L. Bjv'rnv', pp. 240-44. Guilford : I.P . c.
Devin, C . 1 959. J. A co ust. Soc. Am. 3 1 : Sci. Techno!. Press
Annu. Rev. Fluid Mech. 1977.9:145-185. Downloaded from www.annualreviews.org

1 654-67 Holl, 1. W. 1970. J. Basic Eng. 92 : 681-88

Duda, J. L., Vrentas, 1. S. 1 969a. Chem. Eng. Howkins, S. D. 1965. J. Acoust. Soc. Am. 3 7 :
Sci. 24 : 46 1-70 504-8
Duda, J. L., Vrentas, J. S. 1969b. AIChE J. Hsieh, D. Y. 1 965. J. Basic ling. 87 : 991-
1 5 : 3 5 1-56 1005
Elder, S. A. 1959. J. Acoust. Soc. Am. 3 1 : Hsieh, D. Y. 1970. J. Basic Eng. 92 : 8 1 5- 1 8
54-64 Hsieh, D. Y. I 972a. J. A coust. Soc. Am. 52 :
Eller, A. 1 969. J. Acoust. Soc. Am. 46 : 1246-- 151
50 H sieh, D. Y . I 972b. J. Bas ic Eng. 94 : 655-65
Eller, A. 1972. J. A coust. Soc. Am. 52 : 1447- Hsieh, D. Y. 1974a. J. A coust. Soc. Am. 56 :
49 392-93
Eller, A. 1 975. J. A coust. Soc. Am. 57 ; 1374- Hsieh, D. Y. 1974b. In Finite-Amplitude
78 Wave Effects in Fluids. ed. L. Bjv'rn9l, pp.
Eller, A., Crum, L. A. 1 970. J. A coust. Soc. 220-26. Guilford : I.P.c. Sci. Techno!.
Am. 47 : 762-67 Press
Eller, A., Flynn, H. G. 1 965. J. A coust. Soc. Hsieh, D. Y., Plesset, M. S. 196 1 . J. Acoust.
Am. 37 : 493-503 Soc. Am. 33 : 206- 1 5
Eller, A., Flynn, H. G. 1969. J. Acoust. Soc. Hunter, C . 1960. J. Fluid M ech . 8 : 241-63
Am. 4 6 : 722-27 Jahsman, W. E. 1968. J. Appl. Mech. 35 :
Epstein, D., Keller, 1. B. 1972. J. A coust. Soc. 579-87
Am. 52 : 975-80 Jensen, F. B. 1974. J. Fluids Eng. 96 : 389-93
Epstein, P., Pies set, M. S. 1950. J. Chem. Kapustina, O. A. 1970. Sov. Phys. A coust.
Phys. 1 8 : 1 505-9 1 5 : 427-38
Esche, R. 1952. A custica 2 : (AB)208-1 8 Keller, A. 1972. J. Basic Eng. 94 : 9 1 7-25
Finch, R. D . , Neppiras, E. A. 1973. J. A coust. Keller, J. B., Kolodner, 1. 1. 1956. J. App/.
Soc. Am. 53 : 1402-10 Phys. 22 : 1 1 52-61
Florschuetz, L. W., Chao, B. T. 1 9 65 . J. Kirkwood, 1. G., Bethe, H. A. 1942. OSRD
Heat Transfer 87 : 209-20 Rep. No. 588
Flynn, H. G. 1964. Physics of Acoustic Cavi­ Kling, C. L., Hammitt, F. G. 1972. J. Basic
tation in Liquids. Physical Acoustics, ed. Eng. 94 : 825-33
W. P. Mason, Vol. I B : 58 : 172. New York : Koger, H. G., Houghton, G. 1968. J. A coust.
Academic Soc. Am. 43 : 571-75
Flynn, H. G. 1975a. J. A coust. Soc. Am. 57 : Kolb, J., Nyborg, W. L. 1956. J. A coust. Soc.
1 379-96 Am. 28 : 1237-42
Flynn, H. G. 1975b. J. A coust. Soc. Am. 58 : Kornfeld, M., Suvorov, L. 1 944. J. Appl.
1 1 60-70 Phys. 1 5 : 495-506
Gavrilov, L. R. 1970. Sov. Phys. A coust. 1 5 : Krieger, I. M., Mulholland, G. W., Dickey,
285-95 C. S. 1967. J. Phys. Chem. 7 1 : 1 123-29
Gilmore, F. R. 1952. Calif. Inst. Technol. Lamb, H. 1932. Hydrodynamics. Cambridge :
Eng. Div. Rep. 26-4, Pasadena, Calif. The University Press. Reprinted 1945 by
Gould, R. K. 1966. J. A coust. Soc. Am. 40 : Dover Publications, New York
2 1 9-25 Lauer, H. 1 9 5 1 . A custica 1 : AB 1 2-24
Gould, R. K. 1974. J. A coust. Soc. Am. 5 6 : Lauterborn, W. 1968. Acustica 20 : 1 4-20
1 740-46 Lauterborn, W. 1970a. Acustica 22 : 238-39
Gunther, F. C. 1 9 5 1 . Trans. ASME 73 : 1 1 5- Lauterborn, W. 1 970b. Acustica 23 : 73-81
23 Lauterborri, W. 1 972a. X eme Congres Inter-
Harper, J. F. 1 972. Adv. Appl. M ech o 1 2 : national de Cinematographie V ltra-Rapide.
 1 84 PLESSET & PROSPERETTI

pp. 206-9. Paris : Assoc. Nat!. Rech. 203-9. Guilford : I.P.c. Sci. Techno!'
Technique Press
Lauterborn, W. 1972b. Appl. Phys. Lett. 2 1 : Plesset, M. S., Chapman, R. B. 1 9 7 1 . J. Fluid
27-29 M echo 47 : 283-90
Lauterborn, W. 1 974. In Finite-Amplitude Pies set, M. S., Hsieh, D. Y. 1 960. Phys. Fluids
Wave Effects in Fluids. ed. L. Bj¢rn¢, pp. 3 : 882-92
1 95-202. Guilford : I.P.c. Sci. Techno!' Plesset, M. S., Mitchell, T. P. 1956. Q. Appl.
Press Math. 1 3 : 41 9-30
Lauterborn, W. 1 976. J. Acoust. Soc. Am. Plesset, M. S., Prosperetti, A. 1976. J. Fluid
59 : 283-93 M echo In press
Lauterborn, w., Bolle, H. 1 975. J. Fluid P!esset, M. S., Prosperetti, A. 1 977. Int. J.
by Stanford University - Main Campus - Lane Medical Library on 10/19/12. For personal use only.

Mech. 72 : 391-99 Heat Mass Transfer. In press

Levenspiel, O. 1959. Ind. Eng. Chem. 51 : Plesset, M. S., Zwick, S. A. 1952. J. Appl.
Annu. Rev. Fluid Mech. 1977.9:145-185. Downloaded from www.annualreviews.org

787-90 Phys. 23 : 95-98

Margulis, M. A. 1 9 69 . Sov. Phys. A coust. 1 5 : Plesset, M. S., Zwick, S. A. 1954. J. Appl.
1 35-51 Phys. 25 : 493-500
Mikic, B. B., Rohsenow, W. M., Griffith , P. Plesset, M. S., Zwick, S. A. 1955. J. Math.
1970. Int. J. Heat M ass Transfer 1 3 : 657- Phys. 33 : 308-30
66 Prosperetti, A. 1974. l. A coust. Soc. Am. 56 :
Miles, 1. W. 1966. J. Fluid M echo 25 : 743-60 878-85
Minnaert, M. 1933. Phi/os. Mag. 1 6 : 23 5-48 Prosperetti, A. 1975. J. Acoust. Soc. Am. 57 :
M osse, A., Finch, R. D. 1971 . J. Acoust. 8 10-2 1
Soc. Am. 49 : 1 56-65 Prosperetti, A. 1976a. Thermal effects and
Naude, C. F., Ellis, A. T. 196 1 . J. Basic Eng. damping mechanisms in the forced radial
83 : 648-56 oscillations of gas bubbles in liquids. J.
Negishi, K. 1 96 1 . l. Phys. Soc. lpn. 1 6 : 1450- Acoust. Soc. Am. In press
65 Prosperetti, A. 1976b. Application of the
Neppiras, E. A. 1965. Ultrasonics 3 : 9-1 7 subharmonic threshold to the measure­
Neppiras, E . A . 1969a. J . Sound Vib. 10 : 176- ment of the damping of oscillating gas
86 bubbles. 1. A coust. Soc. Am. In press
Neppiras, E. A. 1 969b. J. Acoust. Soc. Am. Pros peretti, A. 1976c. Q. Appl. Math. In press
46 : 587-60 1 Rattray, M. 1 9 5 1 . Perturbation effects in
Neppiras, E. A., Finch, R. D. 1972. J. Acoust. cavitation bubble dynamics. PhD thesis.
Soc. Am. 52 : 335-43 Calif. Inst. Techno!., Pasadena, Calif.
Niino, M., Toda, S., Egusa, T. 1973. Heat Rayleigh, Lord. 1 9 1 7 . Phi/os. Mag. 34 : 94-
Transfer lpn. Res. 2 (4) : 26-36 98
Noitingk, B. E. 1 9 62. Encyclopedia ofPhysics. Rohsenow, W. M. 1 9 7 1 . Ann. Rev. Fluid
ed. S. Flugge, Vo!. XI/2. Acoustics II : Mech. 3 : 2 1 1-36
258-86 Rooney, 1. A. 1 970. Science 169 : 869-7 1
Noltingk, B. E., Neppiras, E. A. 1950. Proc. Rooney, 1. A. 1972. J. A coust. Soc. Am. 52 :
Phys. Soc. London. Sect. B 63 : 647-85 1718-24
Noltingk, B. E., Neppiras, E. A. 1 9 5 1 . Proc. Safar, M . H . 1 9 68 . J. A coust. Soc. Am. 43 :
Phys. Soc. London. Sect. B 64 : 103 2 38 1 1 88-89
Saksena, T. K., Nyborg, W. L 1970. J.
-

Nyborg, W . L. 1974. In Finite-Amplitude

Wave Effects in Liquids. ed. L. Bj¢rn¢, pp. Chem. Phys. 53 : 1722-34
245-51 . Guilford : I.P.c. Sci. Techno!. Scriven, L. E. 1959. Chem. Eng. Sci. 1 0 : 1 - 1 3
Press Shestakov, V. D . , Tkachev, L. G. 1973. Sov.
Pfriem, H. 1 940. Akust. Zh. 5 : 202-1 2 Phys. Acoust. 1 9 : 169-72
PIes set, M . S . 1 949. J. App/. M echo 1 6 : 277- Shestakov, V. D., Tkachev, L. G. 1975. Int.
82 J. Heat Mass Transfer 1 8 : 685-87
Plesset, M. S. 1954. J. App/. Phys. 25 : 96-98 Skinner, L. A. 1970. J. A coust. Soc. Am. 47 :
Plesset, M. S. 1 9 5 7. Symp. Naval Hydrodyn.. 327-3 1
1st. Washington D.C.. 1 956. pp. 297-323 Skinner, L. A. 1972. J. A coust. Soc. Am. 5 1 :
Piesset, M. S. 1 964. Bubble dynamics, Chap. 378-82
I. In Cavitation in Real Liquids. ed. R. Smulders, P. T., Van Leeuwen, H. 1 . W.
Davies. Amsterdam : Elsevier 1 974. In Finite-Amplitude Wave Effects in
Plesset, M . S. 1 969. Topics in Ocean Fluids. ed. L. Bj�rn¢, pp. 227-33. Guilford :
Engineering. ed. C. I. Bretschneider, Vo!. LP.C. Sci. Techno!. Press
1, pp. 8 5-95. Houston : Gulf Pub!. Co. Storm, D. L. 1974. In Finite-Amplitude Wave
Plesset, M. S. 1974. In Finite-Amplitude Effects in FlUids, ed. L. Bj¢rn¢, pp. 234-
Wave Effects in Fluids. ed. L. Bj¢rn¢, pp. 39. Guilford : LP.C. Sci. Techno!. Press
 . BUBBLE DYNAMICS A N D CAVITATION 185

Strasberg, M . 1 953. J. Acoust. Soc. Am. 25 : 885-97

536-37 Theofanus, T. G., Biasi, L., Isbin, H. S.,
Strasberg, M . 1956. J. Acoust. Soc. Am. 28 : Fauske, H. K. 1 969b. Paper presented at
20--26 Natl. Heat Transfer Can!. 1 1 th. Min­
Strasberg, M. 1959. J. Acoust. Soc. Am. 3 1 : neapolis. August 1969
1 63-76 Theofanus, T. G., Patel, P. D. 1976. Int. J.
Strasberg, M. 1 961. J. Acoust. Soc. Am. 33 : Heat Mass Transfer 19(4) : 425-29
359 Trilling, L. 1 952. J. Appl. Phys. 23 : 14-17
Strube, H . W. 1971. Acustica 25 : 289-303 Van Wijngaarden, L. 1 972. Ann. Rev. Fluid
Szekely, 1., Fang, S. D. 1973. Chern. Eng. M ech. 4 : 369-96
Vaughan, P. W. 1968. J. Sound Vib. 7 : 236-
by Stanford University - Main Campus - Lane Medical Library on 10/19/12. For personal use only.

Sci. 28 : 2 127-40
Szekely, 1., Martins, G. P. 1971. Chern. Eng. 46
Sci. 26 : 147-60 Wang, T. 1973. J. Acoust. Soc. Am. 56 : 1 1 3 1-
Taylor, K. J., Jarman , P. D. 1 968. Br. J.
Annu. Rev. Fluid Mech. 1977.9:145-185. Downloaded from www.annualreviews.org

43
Appl. Phys. (J. Phys. D) 1 : 653-55 Wang, T. 1 974. Phys. Fluids 1 7 : 1 121-26
Taylor, K. 1., Jarman, P. D. 1 970. Aust. J. Ward, C. A., Tucker, A. S. 1975. J. Appl.
Phys. 23 : 3 1 9-34 Phys. 46 : 233-38
Theofanus, T. G., Biasi, L., Isbin, H. S., Willard, G. W. 1 953. J. Acoust. Soc. Am.
Fauske, H. K. 1969a. Chern. Eng. Sci. 24 : 25 : 669-86