Axisymmetric Dynamics of A Bubble Near A Plane Wal

See discussions, stats, and author profiles for this publication at: https://www.researchgate.
net/publication/231890663
Axisymmetric dynamics of a bubble near a plane wall
Article in Journal of Fluid Mechanics · December 2009

DOI: 10.1017/S0022112009991340
CITATIONS READS
15 132
2 authors:
Cornelis Wilhelmus Maria van der Geld J. G. M. Kuerten

Eindhoven University of Technology Eindhoven University of Technology
160 PUBLICATIONS 1,816 CITATIONS 190 PUBLICATIONS 3,107 CITATIONS
SEE PROFILE SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Jets in bubbles View project
Contactless and Contactful Boiling for Controlled Evaporation View project
All content following this page was uploaded by Cornelis Wilhelmus Maria van der Geld on 24 March 2014.
The user has requested enhancement of the downloaded file.

J. Fluid Mech. (2009), vol. 640, pp. 265–303.
c Cambridge University Press 2009 265
doi:10.1017/S0022112009991340
Axisymmetric dynamics of a bubble near

a plane wall
C. W. M. V A N D E R G E L D† AND J. G. M. K U E R T E N
Department of Mechanical Engineering, Thermo Fluids Engineering Division, Eindhoven University of
Technology, P.O. Box 513, 5600 MB Eindhoven, The Netherlands
(Received 15 May 2008; revised 23 July 2009; accepted 23 July 2009; first published online
2 November 2009)
Explicit expressions for the added mass tensor of a bubble in strongly nonlinear
deformation and motion near a plane wall are presented. Time evolutions and
interconnections of added mass components are derived analytically and analysed.
Interface dynamics have been predicted with two methods, assuming that the flow is
irrotational, that the fluid is perfect and with the neglect of gravity. The assumptions
that gravity and viscosity are negligible are verified by investigating their effects
and by quantifying their impact in some cases of strong deformation, and criteria
are presented to specify the conditions of their validity. The two methods are an
analytical one and the boundary element method, and good agreement is found. It
is explained why a strongly deforming bubble is decelerated. The classical Rayleigh–
Plesset equation is extended with terms to account for arbitrary, axisymmetric
deformation and to account for the proximity of a wall. An expression for the
corresponding cycle frequency that is valid in the vicinity of the wall is derived. An
equation similar to the Rayleigh–Plesset equation is presented for the most important
anisotropic deformation mode. Well-known expressions for the angular frequencies
of some periodic solutions without a wall follow easily from the equations presented.
A periodically deforming bubble without initial velocity of the centroid and without
a dominating isotropic deformation component is eventually always driven towards
the wall. A simplified equation of motion of the centre of a deforming bubble is
presented. If desired, full deformation computations can be speeded up by selecting
an artificially low value of the polytropic constant Cp /Cv .
1. Introduction
Many submerged bodies or structures in ocean engineering and in process
equipment can be deforming strongly either near a similar body or near a wall.
In some of these situations the hypothesis of an ideal fluid (inviscid and irrotational)
can be used. If the deformation of a solid body is unaffected by gravity and if viscous
effects are negligible, the analysis of the present paper is applicable. The assumptions
that gravity and viscosity are negligible will be verified by theoretically investigating
their effects, by quantifying their impact in some cases of strong deformation and by
quantifying criteria for their validity. A boiling bubble, as another example, grows
typically from a radius of 0 to one of 0.5 mm in 1 ms. Even with forced convection,
practically no wake occurs and vorticity generated at the bubble–fluid interface is
confined to a thin layer at this surface. After detachment from the wall into a
† Email address for correspondence: c.w.m.v.d.geld@tue.nl

266 C. W. M. van der Geld and J. G. M. Kuerten
liquid at saturation temperature, the bubble wake is still undeveloped for some time,
while the bubble is moving and may be experiencing strong deformation. In this
isothermal situation thermocapillary stresses, the so-called Marangoni effect, do not
affect interface motion and heat transfer and phase change can be ignored. If the
velocity field that approaches the bubble is uniform, the ideal fluid hypothesis is
applicable. For oscillatory deformation, viscous damping will be seen to be negligible
if the initial bubble radius exceeds a certain threshold. It will be investigated whether
and to what extent shape deformations affect motion of the bubble away from the
wall. The rate at which vapour bubbles are transported to the centre of a pipe affects
the net heat transfer rate from the wall in nucleate flow boiling. In particular, at
conditions which are favourable for the generation of large vapour bubbles, that is, at
low system pressure and/or in microgravity, shape deformation and oscillation may
become important. Criteria will be given to indicate the conditions of validity of the
neglect of gravity and viscous effects in the analysis.
The dynamics of a non-deforming bubble is given by a six-dimensional set of
differential equations (Kirchhoff equations), as explained by Lamb (1932). The motion
of a spherical non-deforming bubble near a plane wall is governed by the dependencies
of the added mass components on a single geometrical parameter, which is the ratio of
distance to the wall to sphere diameter (Lamb 1932). When the bubble is isotropically
deforming, that is, a shrinking or an expanding sphere, other dependencies of the
added mass tensor are involved (Geld 2002). The more complex the deformation, the
more added mass coefficients and the more dependencies come into play. It will be
shown that a clear separation exists in the dependencies of the added mass coefficients
on shape deformation and on wall interaction. Only one parameter out of those that
determine the added mass coefficients accounts for the presence of the wall, all the
other parameters account for deformation.
There are good reasons to believe that the added mass tensor derived by inviscid
theory is also applicable to vorticity carrying flow. Several important contributions
towards an improved understanding of the added mass force have been made in the
past decade (Auton, Hunt & Prud’Homme 1988; Pelekasis & Tsamopoulos 1993;
Howe 1995; Magnaudet & Eames 2000; Mougin & Magnaudet 2002; Bagchi &
Balachandar 2003). Most of them deal with temporal accelerations of the relative
motion of fluid and a non-deformable body. The concept of added mass turns
out to be well defined also at moderate Reynolds numbers. Rigorous proof that
the added mass coefficients do not depend on the presence of vorticity exists for
the case of constant bubble volume (Howe 1995) and for some deformation cases
(Legendre, Borée & Magnaudet 1998) (see also the review by Geld 2009). Motivated
by these findings, and by the applications mentioned above, this paper is dedicated
to a study of the dynamics and the added mass tensor of a bubble that experiences
strong deformation and is moving near a plane wall in an ideal fluid. Interaction of
translational motion and deformation is highlighted.
The problem of a gas or vapour bubble moving suddenly away from a plane wall and
experiencing strong shape oscillations that affect its motion is a classical hydrodynamic
problem of impulsive motion that is essentially nonlinear. The nonlinearity is
introduced through the impermeable body and wall boundary conditions. Benjamin
& Ellis (1990) derived an expression for the drift velocity of an oscillating bubble in
the absence of a wall. When the dynamics of surface deformation modes is known,
it is possible to compute the drift velocity that results from second-order interactions
between surface deformation modes. Mei & Zhou (1991) studied nonlinear resonance
between two neighbouring shape modes and volume oscillations. They neglected
Axisymmetric dynamics of a bubble near a plane wall 267
coupling of the translational motion of the bubble with the dynamics of the shape
modes. Meiron (1989) computed the ellipsoidal shape of a bubble in steady rectilinear
motion in an infinite fluid. He assumed the pressure in the bubble to be constant
and did not study dynamics at all. Other authors (Longuet-Higgins 1989a,b,c;
Tsamopoulos & Brown 1983) performed a perturbation analysis to derive nonlinear
corrections of shape deformation of a bubble in an infinite liquid. Their findings
will be compared with results of the present study. Feng & Leal (1995) derived
coupled equations for the dynamics of the bubble centroid, volume oscillations and
two adjacent shape modes. The main application is an acoustically driven bubble in
the absence of a wall. Doinikov (2004) performed a similar analysis, but allowed for
all shape modes while removing limitations on natural frequencies. In both studies a
perturbation analysis was applied where the amplitudes of the shape modes and the
translational velocity of the bubble are small with respect to the initial radius and to
the time rate of change of the radius, respectively.
In the present study these limitations are discarded while the translational motion
of the centroid occurs in the vicinity of a plane wall. It is assumed that at all times the
bubble is star-shaped and axisymmetric, that is, can be described by a single function
of the longitudinal angle. Expressions for all components of the added mass tensor
are derived, and with the aid of the Euler–Lagrange formalism the coupled governing
equations are derived and an analytical solution obtained. The effect of the wall on
dynamics and added mass tensor of an initially spherical bubble is studied for two
generic cases:
(i) impulsive start of deformation of arbitrary shape mode, and
(ii) impulsive start of motion perpendicular to the wall.
It will be investigated whether the first initial condition, strong deformation close
to a wall, is sufficient to move bubbles away from the wall. Time histories of
added mass tensor coefficients are analysed, and the main agencies affecting the
motion of the centroid and affecting deformation are determined. For comparison,
the resulting bubble deformation and motion are alternatively computed with the
boundary element method (BEM). Further validation is obtained from comparison of
added mass tensor histories with analytical solutions of special cases, one such case
being given by results of a previous study (Geld 2002). In the latter study, added mass
coefficients for isotropic deformation were defined, whereas in the present study many
added mass coefficients are needed to account for more general shape deformation.
A new form of the well-known Rayleigh–Plesset equation is presented with terms
to account for shape deformation and the proximity of a plane wall. An equation
that governs anisotropic shape deformation, similar to the Rayleigh–Plesset equation,
is derived. Alternative derivations for the angular frequencies of periodic solutions
are presented. Frequency doubling of the fundamental radian oscillation mode will
be explained. The resonance case introduced by Longuet-Hggins (1989a) will be
analysed.
To facilitate reduction of the number of added mass coefficients required for
computational fluid dynamics (CFD) computations of interactions between multiple
bubbles and a wall it will be investigated which coefficients are dominant in some
typical cases. A simplification of the governing equations is sought that would not
essentially affect predictions of a complex system of deforming bubbles near a wall.
The problem definition is given in § 2.1. The analytical method is described in § 2.2,
and the BEM in § 2.3. Results for impulsive start of deformation near a plane wall are
described in § 3.5, and those for impulsive start from rest in § 3.6. Main conclusions
are summarized in § 4.
2. Methods
2.1. Problem statement
A deforming bubble moves near a plane solid boundary in an incompressible fluid.
Fluid motion is irrotational. Only cases will be considered in which purely radial
oscillations are unforced and viscous damping is negligible. According to Lamb
(1932, p. 641) and Longuet-Higgins (1989b, p. 552), this implies that the initial bubble
radius Rt = 0 is taken to satisfy
(Rt=0 (k − 2)k(k + 1))1/4 (4ν 2 ρL /σ )1/4
for any integer k exceeding 2. Here ν denotes the kinematic viscosity of the liquid
and σ the surface tension coefficient. Longuet-Higgins (1989b) made a small mistake
leading to an overestimate of the minimum bubble radius required. For water at
normal conditions it suffices that the bubble radius exceeds 0.02 mm. That this is an
excessively small value was already remarked for globes of liquid by Lamb (1932).
Note that the potential flow solution of an expanding sphere satisfies the full Navier–
Stokes equations. Thermal and radiation losses are negligible if viscous damping is
(Eller 1970; Prosperetti 1977). This will be quantified in § 3.5.5.
The bubble is star-shaped. For sake of brevity and clarity, axial symmetry is assumed
around an axis perpendicular to the wall. Its contour ∂Ω, the bubble liquid–vapour
interface, is described by a single planar C ∞ curve. The bubble is initially spherical
with radius Rt = 0 . Initially, at each point at the bubble interface either the velocity
(analytical approach of § 2.2) or the velocity potential (BEM approach of § 2.3) is
given. It is assumed that the bubble behaves polytropically, so that the pressure inside
the bubble pB and the bubble volume Vare related by pB Vγ = c, a constant. Here γ
is a constant for isentropic processes given by Cp /Cv , the ratio of the specific heats.
The initial bubble pressure pB,0 exceeds the homogeneous liquid pressure by 2σ/Rt = 0 .
If bubble dynamics are slow relative to the processes of heat and mass transfer, the
bubble interior may be assumed to remain isothermal (γ = 1). If bubble motion is
rapid, heat and vapour have insufficient time to diffuse across the bubble interior.
The thermal process is effectively adiabatic in that case (γ = 1.4).
2.2. Analytical approach
In this section the analytical approach for modelling three-dimensional axisymmetric
bubble deformation in a potential flow field is presented. It is based on the Euler–
Lagrange method that employs generalized coordinates that are now defined.
Consider a bubble that occupies a region Ω with boundary ∂Ω. A mirror bubble
is used to simulate a plane solid wall. Let the function R(θ, t) describe the bubble
interface. Here θ denotes the polar angle in a spherical coordinate system (r, θ, ψ) such
that θ = 0 for the point nearest to a plane infinite solid wall (see figure 1). Function R
is expanded in a series of Legendre polynomials Pn−1 . Indices i, j, l, m, n, p, q denote
non-zero positive integers. Index k is reserved for § 3.5. The second term b2 P1 of the
expansion
∞
R(θ, t) = bm (t) Pm−1 (cos θ) (2.1)
m=1
corresponds to translation of the bubble as a whole. It is always possible to select
the origin of the coordinate system x orig such that b2 is zero. Select the origin for
which b2 = 0, and let z/2 be the distance of x orig from the wall. Because of axial
symmetry it suffices to take z/2 as a generalized coordinate to specify x orig . Velocity
U , positive if away from the wall, is the corresponding generalized velocity. Parameter
z/2
r
θ
Figure 1. Schematic of bubble at distance z/2 from a plane wall, with polar angle θ and
velocity U that is positive if away from the wall.
b2 is obviously redundant and cannot be used as a generalized coordinate. All other

coefficients bm are generalized coordinates since they are independent and required
to describe arbitrary deformation. Also the corresponding agencies delivering work
are independent from one another, which will be employed in the derivation of
expressions for the generalized forces below (defined in (2.17)).
The plane wall is modelled with the aid of a mirror bubble. Because of mirror
symmetry, only a single set of coefficients {am } is needed to specify the velocity
potential φ that satisfies the Laplace equation φ = 0:

∞ ∞
−m m−1
φ= Pm−1 am (t) r + Γmq aq (t) r (2.2)
m=1 q=1
with {Γmq } a set of coefficients that account for the presence of the wall and that
depend on a single generalized coordinate (see Hobson 1955):

def q + m − 2 −m−q+1
Γmq = z .
m−1
At a given instant of time, the generalized coordinates,
(q1 , q2 , q3 , q4 , . . .) = (b1 , z/2, b3 , b4 , . . .),
as well as their corresponding generalized velocities,
(q̇1 , q̇2 , q̇3 , . . .) = (ḃ1 , U, ḃ3 , ḃ4 , . . .),
are supposedly known or can be computed from specific velocities at ∂Ω. Here, the
dot denotes derivative with respect to time. It will now be shown how the time-
dependent coefficients of the velocity potential {am } can be determined. Thereafter,
the generalized forces will be specified and Euler–Lagrange equations used to find a
closed solution for the unknowns
(q̈1 , q̈2 , q̈3 , q̈4 , . . .) = (b̈1 , U̇ , b̈3 , b̈4 , . . .).
The kinematic boundary condition expresses the fact that the normal component
of the velocity at a point of the bubble interface, x c = (R(θ, t), θ), equals the normal
component of the liquid velocity u there:
dx c
n· = n · u(x c , t), (2.3)
dt
with n the unit normal on ∂Ω that points into the fluid region. Since powers of R
occur in n · u = n · ∇φ|r = R (see (2.2)), it is convenient to define coefficients enml and
dij in the following way. The constants {enml } in the defining equation

Pn−1 Pm−1 = enml Pl−1 (2.4)
l=1
are structural coefficients, that is, constants that do not depend on time or other
def
parameters. Explicit expressions for the {ekml } are given in terms of ηm = (2m−1)!!/m!
by Gradsteyn & Ryzhyk (1980, § 8.915). These enml -coefficients can also be expressed
in terms of Clebsch–Gordan coefficients (Abramowitz & Stegun 1972). The sum
in (2.4) is finite. For i in Z, coefficients dij are defined by

∞
Ri = dij Pj −1 . (2.5)
j =1
The coefficients dij depend on the bm which depend on time.

The kinematic boundary condition (2.3) yields

∞ ∞
∂φ ∞
∂φ
1
ḃm Pm−1 R + U P1 bn Pn−1 − U P1 R = R
1
− 1
bm Pm−1 R (2.6)
m=1 n=1
∂r R m=1
∂θ R
with P1 = cos θ and P11 = − sin θ. The Pn1 are the associated Legendre functions of
the first order. Employing the summation rule (Gradsteyn & Ryzhyk 1980, § 8.915)
and
1 n (n + 1)
Pn1 (cos θ) = √ (Pn+1 (cos θ) − Pn−1 (cos θ)) , (2.7)
1 − cos2 θ 2 n + 1
it is easily shown that
[(j −2)/2]
n (n − 1)
1
Pn−1 Pj1−1 = − (Pn − Pn−2 ) (2 j − 4 l − 3) Pj −2 l−2 (2.8)
2n − 1 l=0
Here [j ] denotes the integer part of j . The operator

2m − 1 1
d (cos θ) Pm−1
2 −1
is now applied to the boundary condition (2.6). This yields

∞
∞
Gmn an = ḃm + U Fmn bn − U δm2 (2.9)
n=1 n=1
with Fmn and Gmn the following functions of the bi -coefficients:

n(n − 1)
Fmn = d(−1)i ωmni2 = d(−1)i (ei(n+1)m − ei(n−1)m ) (2.10)
i i
1 − 2n
and

Gmn = −n d(−n−1)i einm − d(−n−2)i bq ωmqin + Γ2n δm2
i i,q

+ (j + 1)Γ(j +2)n dj i ei(j +2)m − Γ2n d(−1)i bq ωmqi2
j,i i,q

− Γ(f +2)n d(f −1)i bq ωmqi(f +2) (2.11)
f,i,q
Here the set of structural coefficients {ωmnfj } is defined by

2m − 1
def
1
1
ωmnfj = dw Pm−1 (w)Pf −1 (w)Pn−1 (w)Pj1−1 (w). (2.12)
2 −1
With the aid of (2.8) coefficients ωmnfp can be expressed in a finite series of eijp -
coefficients, as the one on the right-hand side of (2.10).
It can be shown that the set of coefficients {Gmp }m,p corresponds to a Hilbert–
Schmidt operator G that possesses an inverse G−1 if one of the generalized velocities
is unequal to zero. For the case 0 = b3 = b4 = b5 = . . . , that is, the case of isotropic
deformation, a proof is given by Geld (2002); a general proof is a straightforward
but lengthy extension. Application of G−1 to the equivalent of (2.9), written in terms
of G, yields the coefficients aj of the velocity potential.
The velocity potential is now used to compute the kinetic energy in the fluid domain
Ω, that is, the half space without the bubble. This energy of the fluid T can be written
as the following quadratic function of the generalized velocities:
1 1
∞ ∞
1
T /{ρL · V} = αU 2 + U γm ḃm + ψij ḃi ḃj , (2.13)
2 2 m=1 2 i,j
where ρL denotes the mass density of the liquid. The coefficients α, γm and ψij only
depend on time and the generalized coordinates, and together constitute the added
mass tensor of the bubble.
Let G−1 −1
pm denote the element (G )pm of the inverse operator G for arbitrary non-zero
integers p and m. Straightforward computations yield

2
−1
ψml = − (2π/V) Gnl d(−n+2)j enj m + Γpn d(p+1)j epj m ,
2m − 1 n j p j

γm = −ψm2 + ψmq d(−1)i bj ωqj i2 − (2π/V) G−1
pm
q,i,j p

× d(−p+2)j Spj + Γnp d(n+1)j Snj ,
j n j

α = (2π/V) G−1
p2 − G−1
pm d(−1)i ωmni2 bn
p i,n,m

× d(−p+2)j Spj + Γmp d(m+1)j Smj , (2.14)
j m,j
where ωmni2 can be replaced by the explicit expression given in (2.10) and Skj is
defined by

def 2 2i(i + 1) 2(i − 1)(i − 2)

Snj = − enj 2 + d(−1)l enlp epj i · bi+1 − bi−1 .
3 l i,p
(2i + 1)(2i − 1) (2i − 3)(2i − 1)
(2.15)
To each generalized variable qj corresponds an Euler–Lagrange equation of the
form
d ∂T ∂T
− = Qj (2.16)
dt ∂ q̇j ∂qj
with Qj the generalized forces, in general also including conservative forces. The
generalized forces are in this case given by
∂V ∂AB
Qj = (pB − p∞ ) −σ . (2.17)
∂qj ∂qj
Here AB is the surface area of the bubble and p∞ is the pressure at infinity. Note
that drag and gravity are not considered in the present study. Equation (2.17) can
be derived by expressing dT /dt in terms of integrals over ∂Ω by substituting the
normal stress boundary condition in order to relate to the pressure inside the bubble,
by subsequent application of the surface divergence theorem to express the result
in dV/dt and dAB /dt, and by employing the fact that the agencies delivering work
and corresponding to the generalized coordinates are mutually independent (Geld
2002, 2009). Coefficients gj and hm are now introduced in order to facilitate the
determination of AB and its derivatives with respect to the bn . These new coefficients
depend on the bn and describe the dependence of the radial component of the normal
on the polar angle:

∞
1 ∞
nr = gj Pj −1 = hm Pm−1 . (2.18)
j =1
nr m=1

The area AB is given by 4π n d2n hn /(n − 1) and

∂AB 2
(2π) = epmn bp (hm + gm ) + 2 gl ω1mln bm .
∂bn 2n − 1 p l,m
It is easily seen that ∂V/∂bn = 4π d2n /(2n − 1). It is now straightforward to compute
the generalized forces Qj . Since the wall is fully wetted, the generalized force
corresponding to z/2, Q2 , is zero.
Equation (2.17) makes clear that two physical parameters control the dynamics of
a deforming object if gravity and viscous effects are negligible. One of them accounts
def
for capillary effects and often the Weber number We = ρL Rt = 0 U02 /σ is used as a
criterion to indicate its significance. The initial radius Rt = 0 in the definition of We
can be replaced by a volume-based mean radius if the initial shape is non-spherical,
of course. Sample computations in this paper will all start with a spherical shape.
The initial velocity U0 in the definition of the Weber number will be replaced by
another generalized velocity if U0 = 0. The second physical parameter in control of
the dynamics corresponds to the pressure p∞ . The initial value of the pressure in the
bubble pB is related to p∞ via the equation of state of the content of the bubble.
Consistent with the definition of the Weber number, a dimensionless pressure is
defined as p∞ /(ρL U02 ).
The Euler–Lagrange equations (2.16) yield a set of coupled differential equations
that are linear in the unknowns (b̈1 , U̇ , b̈3 , b̈4 , . . .), with proportionality coefficients
composed solely of constants and added mass tensor coefficients. The method of
reduction (Kantorovich & Krylov 1958) is applied to solve this system. To prevent
aliasing, series as those of (2.18) are truncated after 8N terms if the series of φ and R
are truncated after N terms. No iterative steps are required and integration in time is
the only numerical step involved. This numerical step is required to specify the values
of the generalized coordinates and velocities at a next instant of time. The derivatives
(b̈1 , U̇ , b̈3 , b̈4 , . . .) are integrated with the three-step Adams–Bashforth method after
start-up with Euler’s method. The new coordinates are determined with an adapted
three-step Adams–Bashforth algorithm that employs both the first-order derivatives
(ḃ1 , U, ḃ3 , ḃ4 , . . .) and the second-order derivatives (b̈1 , U̇ , b̈3 , b̈4 , . . .). The time step is
chosen sufficiently small, so that the method is both stable and accurate.
The importance of and connections between the added mass coefficients are
elucidated by introduction of a symmetric matrix A. It is defined as follows:
Aij = ψij = ψj i if both i and j are unequal to 2, A2j = Aj 2 = γj /2 if j = 2 and A22 = α.
The Euler–Lagrange equations then yield the matrix equation
A · b̈ = c, (2.19)
where b̈ represents the column vector (b̈1 , U̇ , b̈3 , b̈4 , . . .)T and c a column vector that
comprises generalized velocities and other parameters that are known at a certain
instant of time. The unknown generalized accelerations b̈ are determined from (2.19),
and subsequent integration in time yields the generalized velocities, and c, at the next
instant of time. The matrix A therefore controls the dynamics of large-amplitude
bubble deformation near a wall.
2.3. Boundary element method

In this section the numerical method used to validate the results of the analytical
approach is described. The BEM is chosen since it is very suitable for problems with
a deforming interface in an unbounded fluid. The BEM is based on the following.
If inside a region Ω with boundary ∂Ω the velocity potential satisfies the Laplace
equation φ = 0, the velocity potential in an arbitrary point x 0 ∈ ∂Ω obeys

∂φ ∂G
φ(x 0 ) = −2 G(x; x 0 ) (x) dSx + 2 φ(x) (x; x 0 ) dSx . (2.20)
∂Ω ∂n ∂Ω ∂n
Here G(x; x 0 ) = 1/(4π|x − x 0 |) is Green’s function of the three-dimensional Laplace
equation. In case the velocity potential on the boundary is known, (2.20) is an integral
equation for the unknown normal derivative of the velocity potential at the boundary.
We consider the same situation as in § 2.1 and 2.2. If the plane wall is present
the integrals in (2.20) extend over both the bubble and its mirror image. Due to the
axisymmetry, the integrals over the azimuthal angle can be performed analytically
(Pozrikidis 1997). To this end the boundary of the bubble is represented by
x = (x(s), r(s) cos ϕ, r(s) sin ϕ),
with s the arclength to represent the contour C in a polar plane. This leads to

1 ∂φ 1
φ(x0 , r0 ) = − r (x, r)I10 (x0 − x, r, r0 ) dl + rφ(x, r)
2π C ∂n 2π C
× {[(x0 − x)nx − rnr ]I30 (x0 − x, r, r0 ) + r0 nr I31 (x0 − x, r, r0 )} dl, (2.21)
where nx and nr are the components of n in the two directions in the polar plane.
Further, I10 , I30 and I31 are related to elliptic functions:
4
I10 (x0 − x, r, r0 ) = F (k̃),
((x − x0 )2 + (r + r0 )2 )1/2
4
I30 (x0 − x, r, r0 ) = J30 (k̃),
((x − x0 ) + (r + r0 )2 )3/2
2
4
I31 (x0 − x, r, r0 ) = J31 (k̃),
((x − x0 ) + (r + r0 )2 )3/2
2
4rr0 E(k̃)
k̃ = , J30 (k̃) = ,
(x − x0 )2 + (r + r0 )2 1 − k̃ 2

1 2 − k̃ 2
J31 (k̃) = 2 −2F (k̃) + E(k̃)
k̃ 1 − k̃ 2
and F (k̃) and E(k̃) are the complete elliptic integrals of the first and second kind
(Gradsteyn & Ryzhyk 1980).
The numerical method starts from a given shape of the bubble and a given velocity
potential at the bubble surface. The tangential velocity components follow easily from
∂φ/∂s. Next, (2.21) is used to solve for ∂φ/∂n at the bubble surface. For this purpose,
collocation points are chosen on the contour of the bubble surface to which (2.21)
is applied and the shape of the contour is represented by cubic B-splines (Toose
et al. 1996). The singularity in the second integral in (2.21) is removed by a quadratic
transformation (Telles 1987) and the integrals are evaluated numerically using a high-
order Gauss quadrature. The resulting system of equations is solved with a direct
method.
In this way, both velocity components at the bubble surface are determined and
the bubble shape and velocity potential at its surface at a next instant of time follow
by integration of the kinematic and dynamic boundary conditions. The kinematic
boundary condition is given by equation (2.3) where x c is now the position of a
collocation point. The dynamic boundary condition follows from Bernoulli’s equation
and can be written as
dφ 1 σ pB − p∞ μ ∂un
= |∇φ|2 + C− +2 + g̃x2 . (2.22)
dt 2 ρL ρL ρL ∂n
In this equation d/dt is the material derivative ∂/∂t + ∇φ · ∇, C is the local curvature
def
of the surface, μ = νρL the dynamic viscosity of the liquid and un = u · n. The local
distance to the wall is x2 and g̃ denotes gravitational acceleration. The term ∂un /∂n is
evaluated from the known velocity potential, its normal derivative and φ = 0, which
leads to
∂un ∂ 2φ r ∂φ x ∂φ
=− 2 − − x r − r x − ,
∂n ∂s r ∂s r ∂n
where primes denote derivatives with respect to the parameter s.
Equation (2.22) shows that two new dimensionless parameters can be defined if
gravity and viscosity are accounted for. In the same way, and with similar annotations,
as the Weber number was introduced, the Bond and the Reynolds numbers are
def def
defined: Bo = g̃R02 ρL /σ and Re = R0 U0 ρL /μ. From these definitions it is obvious
that viscous effects are negligible if Re is high and that the effect of gravity can
be ignored if Bo small. Examples will be given to show what values in practice are
required to justify the neglect of gravity and viscous effects.
Equations (2.3) and (2.22) are integrated in time with the standard fourth-order
Runge–Kutta method, which implies that (2.21) has to be solved four times in
each time step. All geometric quantities, such as the bubble volume and the local
curvature, are computed with the use of the spline representation and a high-order
Gauss quadrature. The time step is chosen sufficiently small, so that the method is
both stable and accurate. In cases with large deformation clustering of collocation
points is prevented by redistributing them after every time step in such a way that
they remain approximately equidistant. For the test cases shown in § § 3.1–3.5 this
redistribution was not applied.
3. Results
In this section, added mass predictions of the analytical method are presented, as
well as deformation history predictions of both the analytical and BEM methods.
Four comparisons with known solutions will be made. One is oscillatory motion
governed by the Rayleigh–Plesset equation (§ 3.1). Another known solution is the
added mass tensor of a stationary non-deforming ellipsoid without a wall (§ 3.2). A
third category of known solutions are the anisotropic shape oscillations without a
wall computed by Lamb (1932, pp. 117–475) (§ 3.3). Finally, in the presence of a plane
wall a validation case is offered by the more recently found solutions for spherical
bubble expansion near a plane wall (Geld 2002).
After these comparisons, more complicated motion and deformation cases are
analysed. Two generic cases will be considered. The first case is that of a bubble
whose centre is initially at rest. Motion of the centre is induced by the deformation
that is set in at time zero by prescribing non-zero deformation velocities. The second
case is that of an initially spherical bubble whose centre is given a velocity away
from the wall. The second case is of practical relevance in boiling (see § 1). In
both generic cases, the deformation and added mass coefficients are studied. The
controlling dimensionless parameter (Weber number) will be varied sufficiently to
cover all interesting phenomena. In all cases considered fluid viscosity is neglected
and the fluid surrounding the bubble is assumed to be perfect, as argued in § 2.1. The
process that the content of the bubble experiences is taken to be a polytropic one.
Computations shown in figures in this paper are for γ = 1.4.
All results presented have been validated by changing the number of collocation
points (BEM) or expansion coefficients (analytical method) and by changing the
time step. To analyse the number of generalized coordinates required to describe the
deforming shape, the following shape error is defined:
π
π
def 1
Err = (R − R∞ ) R sin(ϑ) dϑ
2 2
R 2 sin(ϑ) dϑ. (3.1)
Rt=0 0 0
This error is the average over the entire bubble surface of the difference between
the computed radius R(ϑ) and the radius computed with a large (in principle ∞)
number of coefficients, R∞ (ϑ). Figure 2 shows the dependency of the shape error on
the number of shape coefficients N at various times for a deformation test case that
is fully discussed in § 3.5. From figure 2 it follows that the shape error varies roughly
proportionally to N −6 at all times. When the error is about 0.01 % maximum at all
times the number of coefficients is considered to be sufficient. The value of |bN | then
usually differs five orders of magnitude with the value of |b3 |. The gap between the
bubble and the wall merely increases between time zero and time 0.341. At times
10–2
Error in shape prediction, Err

10–4
10–6 0.341
0.681
1.020
1.361
1.701
10–8
10 15 20 25
Number of shape coefficients, N
Figure 2. Average shape error dependence on number of shape coefficients N for various
dimensionless times, −db3 /dt|t=0 t/Rt=0 , of a deformation history that will be depicted in
figure 4. The error is averaged over the entire bubble surface and given relative to the initial
bubble radius.
later than 0.4, this gap has got its minimal value during a cycle at least once, and
the truncation error is relatively high when the gap has this minimal value. For this
reason the shape errors of figure 2 are closer to each other at later times, while the
error at time 0.341 is smallest. In § 3.5.2 results of the expansion method and BEM
will be compared for one specific test case involving a bubble deforming close to a
wall.
3.1. No wall, isotropically deforming sphere
The case of an expanding or shrinking spherical bubble without a wall and without
motion of its centroid is the simplest case which can be calculated. The BEM
computation starts with the initial velocity potential generated by a monopole at the
bubble centre, that is, φ = c̃/|x|. The time evolution of the radius of the bubble is
described by the Rayleigh–Plesset equation (Brennen 1995, p. 101). If the strength
of the monopole c̃ is sufficiently small the change in radius is so small that the
Rayleigh–Plesset equation can be linearized with solution
R(t) = Rt=0 (1 + β sin(ωRP t)),
where Rt = 0 is the initial bubble radius and

3γp∞ 2(3γ − 1)σ
ωRP = 2
+ 3
(3.2)
ρL Rt=0 ρL Rt=0
is the natural frequency corresponding to isotropic deformation of the bubble at
pressure p∞ . A derivation of (3.2), and of modifications to account for the proximity
of a wall, will be given in § 3.4 and § 3.5.7. More often than not the pressure term
of (3.2) dominates the frequency ωRP . A special case of ωRP will be considered in
§ 3.5.5. Usually the time 2π/ωRP is the smallest physical time scale involved, even if
deformation is anisotropic and with large amplitude, and this time determines the
smallest integration time step needed in time integration, the only numerical step in
the analytical method of § 2.2. Amplitude β is related to the monopole strength by
0.5 Present work

Lamb (1932)
Added mass coefficient, α

0.4
0.3
0.2
0.1
0 2 4 6 8
Ratio of axes, a/b
Figure 3. Comparison of added mass coefficients of an oblate ellipsoid given by Lamb

(1932) and the analytical method of § 2.2.
β = − c̃/ωRP Rt3= 0 . The analytical method starts with

ḃ1 t=0 = Rt=0 β ωRP
and a spherical shape, which yields the same monopole strength. Both the analytical
results and the numerical result are in good agreement with the solution of the
linearized Rayleigh–Plesset equation. For the analytical method this is no surprise
since the Euler–Lagrange equation for b1 , that is, the first row-equation of (2.19),
for this case reduces to the classical Rayleigh–Plesset equation. Extensions of the
Rayleigh–Plesset equation will be presented in § 3.4 and § 3.5.4. Even after hundreds
of oscillations no difference in phase or amplitude can be seen between the analytical
and numerical solution.
3.2. No wall, non-deforming ellipsoid
The added mass coefficient corresponding to motion of a prolate non-deforming
ellipsoid in the direction of its longest axis in an unbounded fluid α (see (2.13)) was
analytically computed by Lamb (1932, p. 153). Let the longest axis have length a, and
the shortest axis have length b. The numeric eccentricity e is given by 1 − (b/a)2 = e2 .
The solution of Lamb is α = α0 /(2 − α0 ) with

2 −3 1
α0 = 2(1 − e )e ln((1 + e)/(1 − e)) − e .
2
This solution, as well as some values that were tabulated by Lamb, are given in
figure 3. To apply the method of § 2.2, the contour of the ellipsoid, given by

R(ϑ) = b/ {1 − e2 cos2 (ϑ)},
is expanded in the series given by (2.1). The more the eccentricity deviates from 1,
the more coefficients bm are required to describe the contour properly, and because
of reflectional symmetry 0 = b2 = b4 = b6 = . . . . The coefficients gj of the normal
(see (2.18)) have a similar property. Values of added mass α obtained from (2.14)
represented by discrete dots in figure 3 are the same as corresponding values tabulated
by Lamb.
3.3. Anisotropic, small deformation not very close to a wall
If a gas–liquid interface is axial-symmetric, it is conveniently described by the
time-dependent coefficients bj of expansion (2.1): R = ∞ j = 1 j Pj −1 (cos(θ)) (see
b
§ 2.2). Consider an initially spherical bubble. Let k be 3, 4, 5, . . . , and let
excitation of mode k, or bk -excitation, be defined as the initial condition given
by b1 = Rt = 0 , 0 = b2 = b3 = b4 = . . . , U = 0, ḃj = 0 for j = k, and ḃk = 0. Counter k is
taken to exceed 2 since excitation of coefficient b1 is a special case (see § 3.1 and § 3.4)
and since coefficient b2 corresponds to motion of the interface as a whole (see § 2.2).
In the absence of a wall no preferred direction of motion exists, and coefficient b2 ,
selected to be zero, simply disappears. In the presence of a wall, motion perpendicular
to the wall can be induced. The distance of the centre of the bubble to the wall, z/2,
is the generalized coordinate that replaces b2 , and the initial value of z, z0 , may affect
excitation of mode k.
Excitation of mode k in the absence of a wall leads to an oscillatory change of this
bk -coefficient at a well-known frequency (Lamb 1932) to be derived in an alternative
way below. This frequency is constant because viscous dissipation is neglected. Even
in the presence of a plane wall, expressions of the analytical approach, § 2.2, permit
to derive the governing frequency in the following way.
def def
Define f˜ = bk /b1 , g = b1 /z and write
R = b1 (1 + f˜Pk−1 ) = Rt=0 (1 + β̃(t)){1 + f˜(t)Pk−1 (cos(ϑ))}. (3.3)
Before linearizing the kth row-equation of equation (2.19) it is divided by (∂VB /∂bk )/f˜
for which the following approximation is used:
∂VB ˜ 8π 2
f ≈ b . (3.4)
∂bk 2k − 1 1
Specify small deformation by requiring f˜2 1 and β̃ 2 1 at all times. Specify ‘not
very close to a wall’ by requiring that terms with g and g 2 are retained while terms
in g 3 and higher orders of g are negligible. More specifically, let |g 4 | = |(b1 /z)4 | 1,
but let it be possible that the orders of magnitude of g 2 and |f˜| are the same. Note
that |g| 6 0.5 everywhere, making the condition |g 4 | 1 satisfied at all places not in
the vicinity of the wall. It can now be derived that

∂AB ∂VB
Rt=0 ≈ 2 + 12 (k + 1)(k − 2) (3.5)
∂bk ∂bk
and that
3 (2k − 2)! 2k−1
ψkk ≈ 1+ g ≈ 3 /(k(2k − 1)) . (3.6)
k(2k − 1) {(k − 1)!}2
Equations (3.4)–(3.6) hold also for k = 1 and are used to approximate terms in the
linearized kth row-equation of (2.19) with c containing the right-hand side of (2.17).
Since p∞ = pB,0 − 2σ/Rt = 0 and
pB − pB,0 ≈ −3 γ β̃ pB,0 ,
the linearized equation (terms in f˜2 , g 2 , fg, U are negligible) can be written as
σ
b̈k ≈ −k(k + 1)(k − 2) 3
bk . (3.7)
ρL Rt=0
It is noted that for any k > 3, the term (2k − 2)!g 2k−1 /{(k − 1)!}2 of (3.6) that is
neglected in (3.7) is less than 0.06 if g 6 0.4 (note that g = 1/2 corresponds to a sphere
touching the wall). With the value 0.06 the governing frequency derived from (3.7)
is reduced by about 3 % due to changes in ψkk . More significant changes will be
encountered in § 3.5.1. All initial bubble positions in § 3.5 will be selected in the close
proximity of the wall in order to enable a study of the dependence of the cycle
frequency on the distance to the wall.
The period of oscillation Tk and the initial deformation speed ḃkt = 0 are used to
define a dimensionless ‘natural frequency’ ωk by means of

ωk = 2πRt=0 Tk ḃkt=0 = ω̃k Rt=0 ḃkt=0 ,
def
(3.8)
with ω̃k the radian frequency. The solution of (3.7) yields an angular frequency named
ωk,Lamb given by
2
ωk,Lamb = (k − 2) · k · (k + 1)/ We. (3.9)
Here the initial Weber number We is defined by We = Rt=0 ρL ḃk2t=0 /σ . The expression
for ωk derived by Lamb (1932, equation (12), pp. 117 and 475, where n = k − 1 and
Sn = Pn ) can be cast in the same form as (3.9). The above is therefore an alternative
derivation of Lamb’s radianfrequency. It is noted that the dimensional radian
frequency is given ω̃k,Lamb = k(k − 2)(k + 1)σ/(ρL Rt3= 0 ). A nonlinearized version
of (3.7) for the case k = 3 will be given in § 3.5.
Examples of non-isotropic bubble deformation without a wall, that is, for z = z0 = ∞,
have been modelled with the models of § 2. The example of b4 -excitation is generated
by choosing the initial velocity potential equal to a quadrupole potential with strength
a4 (BEM) or by selecting
ḃ4t=0 = −4a4 Rt=0
in the analytical method. This leads to an oscillating shape in which the coefficient
of the third Legendre polynomial b4 varies sinusoidally if the amplitude is small. The
shape histories from both calculation methods are indistinguishable and the calculated
angular frequency agrees with (3.9). A further comparison of the two methods will
be given in § 3.5.2.
3.4. Isotropically deforming spheres near but not very close to a wall
In the presence of a wall, added mass coefficients other than ψ11 come into play if a
spherical bubble is made to deform isotropically (expand or shrink). The added mass
tensor of a spherical bubble near a wall was derived in previous work (Geld 2002).
Use was made of an operator that was defined by the prescription
i − 1 (i + j − 2)!
G̃ij = −δij + (b1 /z)i+j −1 ,
i (i − 1)!(j − 1)!
where b1 = R, the radius of the spherical bubble (see (2.1)). From Gmp = b1−p−1 mG̃mp
and G−1 p+1 −1
pq = b1 G̃pq /q it is easily shown that (2.14) in this case reduce to
3
α = −1 − G̃−1 ,
2 22
γ1 = 6G̃−1
21 ,
∞
ψ11 = 3 − 3 G̃−1 k
k1 (R/z) . (3.10)
k=1
This is exactly the result given by Geld (2002), where tr(β) = ψ11 and ψ3 = − γ1 , by
definition. It is noted that Geld (2002, § 3) gives a rapidly converging series expansion,
as well as a good approximation, for the added mass coefficients of this case with the
aid of the term 1/(1 − (R/z)3 ).
In this case of an oscillating spherical bubble near a plane wall, an extended
Rayleigh–Plesset equation is derived from the Euler–Lagrange equation (2.16) for b1 ,
that is, the first row-equation of (2.19), in the following way. Since in this case R = b1
it follows that

∂VB ∞
1
2
= 4π b1 + bj = 4πb12 ,
2
(3.11a)
∂b1 j =3
2j − 1

∂AB ∞
1
= 4π b1 (h1 + g1 ) + bk (hk + gk ) = 8πb1 . (3.11b)
∂b1 k=3
2k − 1
The first row-equation of (2.19) reads
1 1 1 2 σ 1 ∂b3 ψ11
ψ11 b1 b̈1 = − γ1 b1 U̇ + (pB −p∞ ) − − ḃ12 b1−2 1
3 6 ρL b1 ρL 6 ∂b1
3
2 ∂ψ11 1 ∂b α 1 ∂γ1
− U ḃ1 b1 + U 2 b1−2 1 − U 2 b1 . (3.12)
3 ∂z 6 ∂b1 3 ∂z
Not very close to a wall, see the definition below (3.4) in § 3.3, γ1 − 3g 2 , α 12 ,
ψ11 3(1 + g) and it follows that

3 1 1 pB − p∞ 2σ
b1 b̈1 (1+g) + ḃ12 +2g − U 2 (1 − 8g 3 ) − b1 U̇ g 2 − 2U Ṙ g 2 = − .
2 4 2 ρL b1 ρL
(3.13)
This is an extended Rayleigh–Plesset equation for a sphere near a plane wall. Other
extensions will be derived below, (3.16) and (3.17). An equation similar to (3.13)
for b1 is (3.15) for b3 that will be derived in § 3.5.1. The linearizing of (3.13) shows
that the governing frequency is now in first-order approximation given by (1 +
Rt=0 /zt=0 )−1/2 ωRP because of the g-dependency of ψ11 (see (3.6)). Full deformation
computations with both methods of § 2 show a good agreement with this frequency
for many oscillation cycles further away from the wall, for example at zt=0 = 30.Rt = 0 ,
def
where g0 = Rt = 0 /zt = 0 0.03. Note that even at this distance the phase shift induced
by the correction (1 + g0 )−1/2 is considerable; without the correction the agreement
would already be lost after one oscillation cycle.
Close to the wall, at g0 = 0.45 for example, the frequency turns out not to be given
by (1 + g0 )−1/2 ωRP anymore. Although in this case deformation becomes considerable,
and coefficients b3 and b4 soon become as large as b1 − 1, deformation is not the
cause of the frequency shift. This is easily proven by a simulation with the analytical
method in which all coefficients except b1 are frozen to zero. The additional frequency
change turns out to be a consequence of the induced motion U of the centre of
the bubble. Velocity U is directed away from the wall, so an isotropically deforming
bubble propels itself away from the wall. In § 3.5.7, a way to predict U will be
presented and used to derive an improved approximation formula for the decreased
frequency. Equation (3.13) turns out to be accurate also very close to the wall as long
as Ug 4 -terms are negligible.
Time = 0 Time = 0.20439 Time = 0.40877
1 1 1
0 0 0
–1 –1 –1
0 1 2 0 1 2 0 1 2
Time = 0.61316 Time = 0.81755 Time = 1.0219
1 1 1
0 0 0
–1 –1 –1
0 1 2 0 1 2 0 1 2
Time = 1.2263 Time = 1.4307 Time = 1.6351
1 1 1
0 0 0
–1 –1 –1
0 1 2 0 1 2 0 1 2
Figure 4. Bubble shape at equidistant time instants for We = 0.12. The wall is indicated by
a set of straight parallel lines. Times are in units of −Rt=0 /(db3 /dt)|t=0 , distances in units
of Rt=0 . Initial deformation is according to what is defined as b3 -excitation in the text, with
γ = 1.4 and U |t=0 = 0.
3.5. Bubble motion induced by anisotropic large-amplitude deformation

3.5.1. An equation governing deformation of the b3 -mode near but not very
close to a wall
Let the bubble initially be a sphere with radius Rt = 0 that starts to deform as (see
(3.3))

R = Rt=0 (1 + β̃) 1 + f˜3 P2 , (3.14)
with β̃ = (b1 − Rt=0 )/Rt=0 and f˜3 = b3 /b1 such that β̃ 2 1 and f˜23 1 at all times.
Large amplitude implies shape deformations as depicted in figure 4. For ease of
reference f˜3 will be written as f˜ in the remainder of this section. In a similar way as
the extended Rayleigh–Plesset equation (3.13) has been derived, the 3th row-equation
of (2.19) can be reduced. Not very close to a wall (see the definition under (3.4) in
§ 3.3), this will yield an equation that governs b3 -changes. Using approximations for
derivatives like, for example,
−2
∂G−1 3 2 2 3 4
21
g b1 1 − f˜ 1 − f˜ ,
∂b3 10 5 7
the following approximations for derivatives of added mass coefficients are derived:
−1
1 ∂γ1 V 9 −1 2 31 ˜ 3˜
b1 g 1 − f 1− f ,
V ∂b3 5 105 5

1 ∂γ3 V 6 15 33
− b1−1 g 2 1 + f˜ + g 2 ,
V ∂b3 5 42 28
−2
1 ∂ψ12 V 9 −1 2 3
b1 g 1 − f˜ ,
V ∂b3 10 5
−2
1 ∂ψ32 V 3 −1 2 2˜ 3 ˜2 33 2 3˜
b1 g 1 + f − f + g 1− f .
V ∂b3 5 7 35 14 5
Terms with partial derivatives with respect to distance z, like ∂α/∂z, have turned out
to be contributing only in close proximity of the wall. Not very close to the wall these
terms are negligible and the following equation, dominating the time evolution of the
anisotropic deformation component b3 can be derived:

1 4˜ 1 2 199 ˜ 1 3˜ 1 2 16 ˜
1 + f b1 b̈3 ≈ − ḃ3 1 + f − ḃ1 ḃ3 1 + f − U 1 + f
6 7 21 70 2 7 4 35

3 32 ˜ 1 15 ˜ 33
2
+ U ḃ1 g 1 + f − U ḃ3 g 1 + f + g
2 2
4 105 2 42 28

1 1 2σ 201 ˜
+ (pB − pB,0 )f˜ − (2Rt=0 /b1 − 1)f˜ 1 − f . (3.15)
ρL ρL Rt=0 392
The terms comprising the dimensionless distance squared g 2 are terms containing the
velocity of the centroid U , and terms with U only contribute if the induced velocity
U is non-negligible as compared to the imposed velocity ḃ3 . To assess the relative
importance of terms in (3.15), simulations in § 3.5.2 and further will start at zero
centre velocity, that is, Ut=0 = 0, whereas simulations in § 3.6 will have a non-zero
initial centre velocity. In § 3.5.2 and further, initial velocity ḃkt=0 will be taken such
that We = Rt=0 ρL ḃk2t=0 /σ = 0.12. Since (3.15) indicates no effect of the wall if U = 0
and the bubble position is not in the vicinity of the wall, the initial position will be
selected such that gt=0 = Rt=0 /zt=0 = 1/2.2, that is, only 10 % below the maximum
value. Before discussing the results, in § 3.5.2 and further, some other considerations
regarding (3.15) will be made.
Note that pB − pB,0 ≈ −3γ β̃ pB,0 and that (3.15) contains all dominant terms
up to second-order perturbation quantities, that is, terms proportional to
f˜, β̃ f˜, f˜2 , g 2 f˜, g 2 , g 4 . The linearizing of this equation leads to the oscillation
frequency ω3,Lamb as discussed in § 3.3. The (pB − pB,0 ) term is proportional to β̃ f˜
and therefore of second order, while the surface tension term is of first order. This is
the reason why all bk -excitations (k > 3) are dominated by surface tension, while the
isotropic Rayleigh–Plesset oscillatory motion (ωRP ) in § 3.1 was seen to be dominated
by pressure.
The occurrence of the U 2 term on the right-hand side of (3.15) implies that if far
away from any boundary a motion of the centroid in certain direction ez is set in,
the bubble interface described by R(ϑ) starts deforming according to P2 (cos ϑ) =
1/2(3 cos2 (ϑ) − 1), with ϑ the polar angle to ez . At the same time, also isotropic
deformations start (see (3.13)). In the U ḃ1 g 2 term and the U ḃ3 g 2 term of (3.15) the
distance to the wall appears, and these terms therefore only contribute in the presence
of a wall. As noted in the above, these contributions will be investigated in § 3.6.
(a) We = 0.12; γ = 1.4
0.15
70*(b1–Rt=0)
0.10 b3
b4
0.05 b5
–0.05
–0.10
–0.15
–0.20
0 0.5 1.0 1.5
Dimensionless time, –db3/dtt=0 t/Rt=0
(b) We = 0.12; γ = 1.4

0.04
70*(b1–Rt=0)
0.03 b3
0.02 b4
b5
0.01
–0.01
–0.02
–0.03
–0.04
0 0.5 1.0 1.5
Figure 5. Histories of shape coefficients bk for b3 -excitation (a), and b5 -excitation (b).
γ = 1.4 and We = 0.12. The result for γ = 0.1 is indistinguishable.
3.5.2. Effect of the wall on the angular frequency of bk -excitation

The effect of the wall on deformation and motion of a bubble starting from
position given by gt=0 = 1/2.2 and starting with Ut=0 = 0 has been investigated for
k = 3, 4, 5, 6, 7. The shape history for k = 3 is shown in figure 4. The corresponding
history of bk -coefficients is shown in figure 5(a). In these figures, time is made
dimensionless using −ḃ3t=0 /Rt=0 , where ḃ3t=0 is selected to be negative in order to have
an initial induced velocity of the centroid away from the wall. The dimensionless
times when b3 = 0 are multiples of π/ω = 1/2|ḃk,t=0 |T /Rt=0 where ω is ω3 (see (3.8)).
The dimensionless frequency is initially about 10 % less than ω3,Lamb , which is 10 (see
figure 5a: 2π/ω ≈ 0.7 initially).
The frequency for b5 -mode excitation is √ in figure 5(b) observed to be also about
10 % less than ω5,Lamb , which equals 5 30. Closing in to the wall reduces the
frequency of bk -excitation (k > 2), similar to the reduction of frequency of isotropically
deforming spheres (see § 3.4). In his derivation of (3.9), Lamb took b1 to be a constant
and independent of the equation of state of the bubble content, pB Vγ = c. The
accuracy of this simplification will be studied in § 3.5.3 and § 3.5.5.
In general, coordinates bm with m close to k experience with bk -excitation a periodic
motion with non-negligible amplitude. Comparison of figures 5(a) and 5(b) shows
that the ratio of max(b3 ) to max(b5 ) is at least 20 % larger for b3 -excitation than the
ratio of max(b5 ) to max(b3 ) for b5 -excitation; the maximum is here taken over several
cycles of periodic motion. This coupling difference is due to the added mass tensor
that is most conveniently described with the matrix A of (2.19). This coupling will be
further discussed in § 3.5.3.
For the b3 -excitation case shown in figure 4 and figure 5(a), differences between the
expansion method and those of the BEM have been studied in the following way. In
both methods the time step has been chosen so small that further reduction of the
time step results in negligible differences. In order to obtain the same initial condition
in both methods the initial velocity potential of the expansion method has been used
as initial condition in the BEM. In the expansion method 21 coefficients are used,
whereas the number of collocation points used in the BEM is varied between 64
and 128. Figures 6(a) and 6(b) show the differences between both methods in two
expansion coefficients, b3 and b7 , as functions of time. These results show that the
error due to the finite number of coefficients in the expansion method is negligibly
small. Errors in the BEM decrease more than quadratically with the element size. For
128 collocation points the relative error in b3 is of the order of 0.01 %, whereas the
relative error in b7 is smaller than 1 %.
As compared to the case where no wall is present, not only the frequency changes,
but also a velocity of the centrepoint U is induced that is oscillating, see figure 7.
The cycle-averaged value of U has a tendency to become negative eventually. This
implies that despite the fact that initially the distance to the wall increases, the shape
oscillations eventually drive the bubble towards the wall. This tendency is observed
for excitation of any mode k > 2. Shape deformation and oscillation couple to volume
oscillations, which cause the bubble to be attracted by its image in the wall. This will
be further studied in § 3.5.6 and § 3.5.7.
The polytropic constant γ has usually values exceeding 1.0, and γ equals 1.0 for a
constant bubble temperature. In this study, a typical value of an isentropic process
with γ given by the ratio of the specific heats (γ = 1.4) is used, but for comparison
also a value of 0.1 has been employed. By taking the exponent γ to be 0.1 the coupling
between volume and pressure is obviously suppressed. However, this does not affect
motion and velocities of the centrepoint. Figures 5 and 7 are the same for γ = 0.1,
and similar observations are made for the histories of the added mass coefficients.
An effect of γ is merely found for the accelerations, and in particular those of b1 ,
as will be discussed in § 3.5.5 and § 3.5.7. Since the computation time depends on the
smallest time scales involved in the process, see § 3.1, and since the smallest time scale
decreases with increasing γ , it is possible to reduce computation time of deformation
histories by taking an artificially low value of γ . Only when coupling of volume and
pressure are of interest this reduction of computation time is impossible.
3.5.3. Coupling of motion and of deformation modes

The coupling of deformation and motion of the centre-point is made by added mass
coefficients, as the analytical method clarifies and is most clearly shown by (2.19).
In derived equations such as (3.15), added mass coefficients are expressed in terms
of generalized coordinates. Since these coordinates depend on time, so do the added
mass coefficients (see figures 8 and 9). Initial conditions are the same as in § 3.5.2:
(a) (× 10–4)
2
1
b3,BEM – b3,expansion
–1
Ncoll = 64
Ncoll = 96
Ncoll = 128
–2
0 0.2 0.4 0.6 0.8 1.0
Dimensionless time, –db3 / dtt = 0 t /Rt = 0
(b) (× 10–4)
2
b7,BEM – b7,expansion
–1
Ncoll = 64
–2
Ncoll = 96
Ncoll = 128
0 0.2 0.4 0.6 0.8 1.0
Dimensionless time, –db3 / dtt = 0 t /Rt = 0
Figure 6. Differences in the value of b3 (a) and b7 (b) between a calculation with the
expansion method and with BEM for b3 -excitation. γ = 1.4 and We = 0.12.
zt=0 = 2.2Rt=0 and We = 0.12. The value of the familiar added mass coefficient α is
0.5 for a sphere far away from the wall, of course, but about 0.7 and oscillating for
a strongly deforming bubble in the vicinity of the wall (see figure 8). The amplitude
of oscillation is larger for b3 than for b5 -excitation (figure 8). Figure 9 shows some
of the ψ1j coefficients, j ranging from 3 to 5 and higher; these coefficients are
clearly not constant in time. Coefficient ψ11 , on the contrary, is nearly constant in the
corresponding cases, at a value close to 4.5. The value of ψ33 is oscillating around
0.25 with amplitude of about 0.04, and values of ψii , i > 3, are about constant in
time, positive and less than 0.2 and decreasing with increasing value of i. The cycle-
averaged values of these ψij parameters increase when the bubble gets very close to
the wall. Apart from the amplitude of oscillation, time histories of the γ -added mass
coefficients are similar (figure 8). The same holds for the time histories of ψ13 and ψ14
We = 0.12; γ = 1.4
0.025
0.020
0.015
0.010
0.005
–0.005
(h–ht=0)/Rt=0
–0.010
U/Rt=0 in s–1
–0.015
0 0.5 1.0 1.5 2.0 2.5 3.0
Figure 7. Trajectory and history of velocities of the centre point for b5 -excitation and
γ = 1.4. The result for γ = 0.1 is exactly the same.
in, respectively, b3 and b4 -excitation, as comparison of figures 9(a) and 9(b) shows. In
both cases the value of ψ13 is oscillating around 0.25, but with different amplitude.
As a consequence, matrix A has a diagonal (ψii and α) that is about constant in
time in the deformation cases of this section. Values at the first row and first column
(ψ1j = ψj 1 ) can be significant and are not constant, but fluctuate with minimum values
close to zero (see figure 9). Also elements close to the diagonal are contributing at
some times, but remaining values of A are usually negligible. There are times when
matrix A is roughly a diagonal matrix and there are times when A is nearly zero
anywhere outside the diagonal and outside the first column and row. This observation
holds for excitation of any mode k > 2.
If the matrix A would always be a diagonal matrix, the equations of the generalized
accelerations would be uncoupled. The importance of the first row and first column
(corresponding to b1 and z) shows that there is always coupling, for any bk -excitation
and for any generalized coordinate, with the b1 parameter of spherical expansion
or contraction. Because of this coupling, generalized coordinate b1 is found to
experience two superposed oscillations (see the typical result of figure 10). In histories
of generalized coordinates shown in figure 5 the high, ωRP -like, frequency component
cannot be distinguished. For this reason, figure 10 shows the history of the acceleration
b̈1 . Far away from the wall, the high frequency is the frequency ωRP given by (3.2).
In the example of b5 -excitation these oscillations are superposed on a frequency that
is twice the natural frequency of b5 , ω5 (see (3.8) and (3.9)). The reason of this
low-frequency doubling of ω5 in b1 , which is also clearly observed in figure 5, is
investigated in § 3.5.5 with the aid of a Rayleigh–Plesset equation that is extended
with terms to account for deformation.
3.5.4. A Rayleigh–Plesset equation with deformation

With only spherical deformation, the first row-equation of (2.19) would yield the
extended Rayleigh–Plesset equation (3.13) (see § 3.1). In the case of full deformation,
the first row-equation of (2.19) is found to encompass not only b̈1 but all other
generalized accelerations. This Euler–Lagrange equation, a further extension of the
(a) We = 0.12; γ = 1.4
0.8
0.6 γ1
α
0.4 γ3
0.2 γ4
–0.2
–0.4
–0.6
–0.8
0 0.5 1.0 1.5 2.0 2.5 3.0
(b) We = 0.12; γ = 1.4

0.8
0.6 γ1
α
0.4
γ3
0.2 γ4
–0.2
–0.4
–0.6
–0.8
0 0.5 1.0 1.5 2.0 2.5 3.0
Figure 8. Histories of added mass coefficients α and γj for b3 -excitation (a), and
b5 -excitation (b). We = 0.12 and γ = 1.4. The plain solid line corresponds to γ5 .
Rayleigh–Plesset equation, can be written as:

∞
2ψ11 Vb̈1 + γ1 VU̇ + 2V ψk1 b̈k
k=3

∂ψ11 ∂ψ11 V
∞
∂ψ11 V ∂αV
= 2Q1 /ρL − ḃ12 − 2ḃ1 2U V + ḃk + U2
∂b1 ∂z k=3
∂bk ∂b1

∞
∂γk V
∞
∂ψmk V ∞
∂ψk1
+U ḃk + ḃk ḃm − 4U V ḃk
k=3
∂b1
k,m=3
∂b1
k=3
∂z
∂γ1 ∞
∂ψm1 V ∞
∂γ1 V
− 2U 2 V −2 ḃk ḃm −U ḃk . (3.16)
∂z k,m=3
∂bk
k=3
∂bk
The generalized force Q1 is given by (2.17) with derivatives given by the equalities on
the left-hand side of (3.11), for example. Note that explicit analytical expressions exist
(a) We = 0.12; γ = 1.4
0.7
0.6 ψ13
0.5 ψ14
ψ15
0.4
ψ16
0.3
0.2
0.1
0
–0.1
–0.2
0 0.5 1.0 1.5
(b) We = 0.12; γ = 1.4
0.30
0.25 ψ13
ψ14
0.20
ψ15
0.15 ψ16
0.10
0.05
0
–0.05
–0.10
0 0.5 1.0 1.5
Figure 9. Histories of added mass coefficients ψ1j for b3 -excitation (a), and b4 -excitation (b).
We = 0.12 and γ = 1.4. The plain solid line corresponds to ψ17 .
We = 0.12; γ = 1.4
1.0
d2b1/dt 2/((db3/dtt =0)2 /Rt =0)
0.5
–0.5
–1.0
0 0.1 0.2 0.3 0.4 0.5
Figure 10. History of acceleration of the first generalized coordinate b1 corresponding

to isotropic deformation for b3 -excitation. We = 0.12 and γ = 1.4. The frequency of the
high-frequency component increases with increasing γ .
for all derivatives in (3.16) (see § 2.2). The classical Rayleigh–Plesset equation was used
in many applications, viz. boiling and cavitation (Brennen 1995), Bjerknes attraction
(Prosperetti 1984) and generation of sound (Versluis et al. 2000). No equivalent or
extended Rayleigh–Plesset equation to account for non-spherical deformation close to
a wall has been found in the literature, despite its importance in the aforementioned
applications.
The simplification of (3.16) for b3 -excitation is analysed away from the wall in
§ 3.5.5, leading to (3.20).
3.5.5. Frequency-doubling of b1 component, away from the wall

In case of b3 -excitation as defined in § 3.3, each sum ∞ k=3 in (3.16) reduces to the
case k = 3 only. With f˜ = b3 /b1 the volume is then given by V= 43 πb13 (1 + 35 f˜2 + 352 f˜3 ).
This yields away from the wall and for U ḃ3 and U̇ b̈3 :
2ψ11 b̈1 + 2ψ31 b̈3

pB − p∞ 2 σ 1 ∂ψ33 V 1 ∂ψ31 V
≈6 b1 + b32 /5 b1−3 − 12 b1−2 + ḃ32 − 2ḃ32 . (3.17)
ρL ρL V ∂b1 V ∂b3
The following approximations are derived for the partial derivatives of added mass
coefficients remaining in (3.17):
−1
1 ∂ψ33 V 3 −1 3˜ 3 ˜2 2 ˜3
b1 1 + f 1+ f + f , (3.18a)
V ∂b1 5 7 5 35
−1
1 ∂ψ31 V 1 3 2
b1−1 1 + f˜ 1 + f˜2 + f˜3 , (3.18b)
V ∂b3 7 5 35
while the added mass coefficients in (3.17) away from the wall can be approximated
as follows:
−1
3 2
ψ11 3 1 + f˜2 + f˜3 , (3.19a)
5 35
−1
4 3 2
ψ31 = ψ13 f˜ 1 + f˜ 1 + f˜ + f˜ 2 3
. (3.19b)
7 5 35
The b1 equation (3.17) can with (3.18b) and (3.19b) be written as

1 3˜ pB − p∞ ˜ σ −1 3
b1 b̈1 + b3 b̈3 1 + f (1 + f /5) − 2 b1
2
1 + f˜2
3 7 ρL ρL 5

2 ˜3 7 1 ˜
+ f + ḃ32 − − f , (3.20)
35 30 105
where the first term is reminiscent of the first term in (3.13) and no g-terms appear
because of the assumption made to be away from the wall.
Away from the wall, coefficient b3 in a b3 -excitation turns out to be well described
by
b3 = −βRt=0 sin(ω̃3 t) (3.21)
with ω̃3 corresponding to ω3,Lamb , that is, ω̃3 = ω3,Lamb |ḃ3t=0 |/Rt=0 , and β a small
parameter. For constant σ , ρL and Rt=0 , We ∝ ḃ02 , ω3,Lamb ∝ 1/|ḃ3 |t=0 , ω̃3 is constant
and β = 1/2|ḃ3 |t=0 . Typical values are We = 0.12 and β = 0.1 (see also figure 5). Let
it now be assumed that b1 can be expanded in the following series:
b1 = b1,0 + εb1,1 + ε2 b1,2 + · · ·
with b1,0 = Rt=0 and with ε a small parameter, ε 1. If will follow from equation (3.20)
that ε is indeed a small parameter if β is. With
2 3γp∞ 2(3γ − 1)σ
ωRP = 2
+ 3
,
ρL b1,0 ρL b1,0
see also (3.2), it can now be shown that the generalized force, Q1 , in first order
approximation is given by

b1 − b1,0 1 2
Q1 − 2 4
+ β sin (ω̃3 t) 4π b1,0 2
ωRP ρL . (3.22)
b1,0 5
The b1 equation (3.20) can with (3.22) be written as

1 b1 − b1,0 1 2 2
b1 b̈1 + b3 b̈3 − + β sin (ω̃3 t) b1−2 b1,0
4 2
ωRP
3 b1,0 5

7 1
+ ḃ32 − + b3 /b1 . (3.23)
30 105
It follows that b1,1 is given by
−2
εb1,1 = ÃRt=0 {cos(2ω̃3 t) − 1} + BRt=0 ωRP (cos(ωRP t) − 1) (3.24)
with 2 2
1 2 ωRP − 3ω̃23 1 2 2 ωRP + 4ω̃23
Ã = β and B=
β ω̃3 2 . (3.25)
10 2
ωRP − 4ω̃23 20 ωRP − 4ω̃23
Parameter ε turns out to be proportional to β 2 , so small if ω̃3 ωRP , which is
normally the case. Since b1 occurs in the governing equation of b3 (see (3.15)), the
proportionality of ε with β 2 explains why (3.21) gives such a good description of b3 .
Coefficient Ã being a constant, solution (3.24) clearly shows the frequency doubling
that is manifest in b1 histories of the numerical solutions of bk -excitations. It is
noted that Prosperetti (1977) showed that for oscillatory motion described by (3.21)
radiation and thermal losses are proportional to β 4 , so negligible with the present
assumptions.
It is easy to rewrite (3.24) into

2 1
2 2 1
εb1,1 /Rt=0 = β ωRP − 3ω̃3 2
ωRP − 4ω̃3 2 sin
2
(2ω̃3 + ωRP )t
10 2

1 1 1 ω̃23
× sin (ωRP − 2ω̃3 )t + β 2 1+ 2
(cos(ωRP t) − 1), (3.26)
2 10 2 ωRP
which has an amplitude that becomes large if ωRP /2 approaches ω̃3 . This is a nice
extension of the work of Longuet-Higgins (1989a), who tried to explain the source of
underwater noise at sea (see also Longuet-Higgins 1989b,c). Longuet-Higgins showed
that each oscillation with amplitude β at the frequency of a linear fundamental
mode produces oscillations that behave at a distance just as those of what he calls
the breathing mode, which is the fundamental radial mode with radian frequency
ωRP , and with an amplitude proportional to β 2 . This is the frequency doubling
observed in the above. Longuet-Higgins merely derived the component with cos(2ω̃3 t)
in solving his equation (6.23) with a velocity potential given by ZRt=0 sin(2ω̃3 t)/r.
It can, however, easily be shown that ZRt=0 sin(2ω̃3 t)/r + QRt=0 sin(2ωRP t) is also a
family of solutions of his equation (6.23), for any value of Q. Longuet-Higgins took
Q to be zero. In his approach, three parameters are perturbed simultaneously. In
our approach, two of his parameters (velocity potential and pressure) are expressed
in terms of the other, which is interface deformation as described by the generalized
coordinates.
If ω̃3 is close to ωRP /2, the amplitude of the oscillatory component with ω̃3 in (3.26)
reaches a maximum in the typical time π/(ωRP − 2ω̃3 ). However, way before that
time ε is found not to be small anymore. The coupled equations for b1 and b3 need
2
to be solved together in that case. In many practical cases, ωRP exceeds ω̃23 by far,
while relative volume changes are less than 0.001. In these cases part of the solution
of Tsamopoulos & Brown (1983) is recovered. These authors also found an isotropic
component with amplitude (b3 /Rt=0 )2 /5, in their equation (56a) combined with their
definition 12. This part of the frequency doubling is apparently not connected to
volume changes, since Tsamopoulos and Brown did not employ an equation of state
and assumed the bubble to have a constant volume. Further comparison with the
results of these authors is not only hampered by their using a different equation of
state, but mostly by their assumption of a fully periodic motion and therefore not
specifying the initial shape precisely. The decrease in frequency with amplitude, which
they and other authors found in the absence of a wall, is recovered in our approach
with a realistic equation of state.
Summarizing, in the extended Rayleigh–Plesset equation (3.17) terms with ḃ32 are
significant whenever ḃ3 is not close to zero. Shape expansion coefficient ḃ3 is oscillating
around zero at about ω3,Lamb (figure 5a), and a quadratic term yields a frequency
doubling. Since velocity U is also oscillating at about the natural frequency ω3,Lamb
(figure 7), the terms with U 2 and U ḃ3 in (3.16) add to the frequency-doubling
observed in b1 . The importance of a term with U 2 can be judged from an analysis of
the contributions to the kinetic energy T because T contains the term αU 2 . Since the
history of the kinetic energy budget also reveals the coupling of shape deformation
to motion of the centrepoint, it is studied in § 3.5.6 before in § 3.5.7 the second
row-equation of (2.19) is analysed.
3.5.6. Energy transfer

In figure 11, the history of kinetic energy T , (see (2.13)) is compared with its initial
value, T0 , and with the potential energy change history of σ (AB − AB,0 ), where AB,0
is the initial bubble surface area. Integration of
dT ∂V dAB
= (pB − p∞ ) q̇j − σ (3.27)
dt j
∂q j dt
easily shows that T + σ Ab + p∞ V− pB,0 Vγ0 1−γ

1
V1−γ is constant. The pressure-related
energy part contributes to the energy budget of the high-frequency oscillations of
§ § 3.1, 3.3 and 3.4, for example, but turns out to be negligible in the results of
figure 11. The volume oscillations induced by shape deformation cause the bubble to
move to the wall, but at a longer time scale.
Comparison of figure 11 with figure 5(a) shows that kinetic energy T is oscillating
at twice the natural frequency ω3,Lamb with non-constant amplitude. Kinetic energy is
never zero, but as in any undamped spring-mass system there is continuous transfer
of kinetic energy to potential energy and vice versa. In the present deformation case
all potential energy is in the enlargement of the area of the gas–liquid interface
We = 0.12; γ = 1.4
1.0
(T – T0)
σ(A – A0)
Energy/(Vb ρL (db3/dt|t =0)2/4)

(T/T0)
0.5
–0.5
0 0.2 0.4 0.6 0.8 1.0
Figure 11. Histories of energy contributions for b3 -excitation. The kinetic energy has initial
value T0 and Vb = V. The virtual mass 1/2ρL V and the initial deformation velocity have
been used for scaling. We = 0.12 and γ = 1.4. The plain solid line is the sum of T − T0 and
σ (AB − AB,0 ).
(see figure 11). At initial time zero, U = 0 and only the ψ33 term is contributing to T
(see (2.13)). Since ḃ3 < 0, the γ3 term proportional to both ḃ3 and U on the right-hand
side of (2.13) starts decreasing the value of T at subsequent times. In the first crest
of potential energy, at around 0.17 in dimensionless time units, ḃ3 = 0 and the small
value of T corresponds to the small value of U 2 at that time. At the subsequent
maximum value of T , at time 0.33, ḃ3 is positive and possesses a maximum value
(figure 5a) while U is close to its maximum (positive) value. Nearly all kinetic energy is
recovered, while the remainder is stored in σ (AB − AB,0 ) (see figure 11). One period of
oscillation of T later, both U and ḃ3 are negative and possess minimum values. Each
time the main deformation amplitude reaches its maximum value (ḃ3 = 0), kinetic
energy in the fluid (T ) reaches a minimum value that is fully determined by the
kinetic energy of the motion of the centre point of the bubble (αU 2 ). The governing
equation of this motion is studied in the next § 3.5.7.
3.5.7. Governing equation for motion of the centroid
The motion of the centrepoint would be fully determined by the second row-
equation of the matrix equation in A, (2.19), if the motion of all generalized
coordinates would be uncoupled. Since the coupling of U with coordinates other
than b3 and b4 is weak, see the low values of the coupling constants γ4 , γ5 , . . . in
figure 8, and since accelerations b̈1 are found to be small, the second row-equation
of (2.19) is expected to govern the motion of the centrepoint. Its main contributions
will now be analysed.
The corresponding Euler–Lagrange equation can be given the form

1 d 1 d ∞ ∞
∂ψij
∞
∂γm
{2αU V} + V γm ḃm − 2 ḃi ḃj + −2U ḃm 0, (3.28)
V dt V dt m i,j
∂z m
∂z
where the approximate sign is to account for the neglect of the term − 2U 2 ∂α/∂z
on the left-hand side of this equation. This term is found to be negligible for
deformation-driven bubble motion at all times. The histories of three remaining terms
0.5
0.4
0.3
Acceleration/((db5/dt|t=0)2/Rt=0)
0.2
0.1
0
–0.1
–0.2
–0.3
–0.4
–0.5
0 0.1 0.2 0.3 0.4 0.5
Figure 12. Histories of dimensionless accelerations for b5 -excitation. We

= 0.12 and γ = 1.4.
∞
(1/V)(d/dt){2U αV}, open circles denote (1/V)(d/dt){V m ḃm γm }, solid dots
Triangels denote
denote − 2 ij ḃi ḃj (∂ψij /∂z). The main contributions to the second Lagrange–Hamilton
equation are shown. The plain solid line is the sum of the three contributions.
are for b5 -excitation plotted in figure 12. The high-frequency oscillation component
related to ωRP is clearly visible in the generalized accelerations shown. The fourth
term on the left-hand side of (3.28) can be neglected in first-order approximation (see
the solid line in figure 12). Integration of (3.28) then yields
t ∞
∂ψij
U (αVU )|t=0 /(αV) + V ḃi ḃj dt /(αV)
0 i,j
∂z

∞

∞
+ (γ1 ḃ1 )t=0 + (γj ḃj )t=0 − γ1 ḃ1 − γj ḃj (2α), (3.29)
j =3 j =3
where the terms are evaluated at time t unless denoted otherwise. The volume ratio
Vt=0 /Vin the third term on the right-hand side is neglected since this ratio is about
10−4 % maximum in all simulations of § 3.5. In all cases of the present section,
def
U0 = U |t=0 = 0, and the first term on the right-hand side of (3.29) is zero. It is
retained here for ease of reference.
t The integral in the second term on the right-hand
side is approximately given by 0 V ∞ 2 ∂ψjj
j ḃj ∂z dt. Because of the negative sign of each
derivative ∂ψjj /∂z this second term on the right-hand side is always negative. Also
in the simulation of figure 12 this term appears to be negative always. This negative
sign explains the following general trend that has been observed in all simulations.
Finite-amplitude excitation of any mode bk in the presence of a wall always leads
to motion of the bubble towards the wall (see figure 7). If the sign of the initial
deformation velocity is taken to be positive, initially induced velocities U are negative
and the bubble merely reaches the wall in a shorter time span. This tendency is a
long-term trend that follows from the second term on the right-hand side of (3.29). If
this term were left out, the resulting error in U would be about 10 % after a quarter
of a cycle of ω5 .
The third term on the right-hand side of (3.29) clearly explains why velocity U has
an oscillatory component at about the natural frequency ωk,Lamb of ḃk (see figures 7
and 5b). It turns out that for the deformation simulations ofbk -excitations of § 3.5 it
is possible to replace the sum ∞ j =3 in this third term by the
k+1
j =k−1 without affecting
deformation predictions much.
Equation (3.29) allows the following derivation of the cycle frequency of isotropic
oscillation near a wall (see also § 3.4). Since ḃ12 terms do not contribute in (3.13), U̇ on
the left-hand side of this equation can be approximated with −γ1 b̈1 /(2α) (see (3.29)).
With the already used approximation −3g 2 for γ1 in (3.13), the following equation
for the cycle frequency ω is then easily derived:

3
ω2 = ωRP
2
1 + g0 − g04 (2α) . (3.30)
2
The value of α is approximately 0.7 close to a wall, and this yields the following
3 4 −1/2
approximation of ω: (1 + g0 − 2.8 g0 ) ωRP . Full deformation computations with both
methods of § 2 show a good agreement with this frequency for many oscillation cycles
close to the wall.
3 4 −1/2
The frequency (1+g0 − 2.8 g0 ) ωRP is somewhat reduced as compared to ωRP . The
physical explanation of the decrease in oscillation frequency of isotropic expansion
when the distance of the bubble to the wall is reduced is the increase of added mass
(ψ11 ). Liquid in between bubble and wall needs to be directed away, that is, needs
to be accelerated in a direction parallel to the wall. The fluid that is pushed away
from the half of the sphere that faces the wall needs to find its way through a surface
area that at distance r from the centre of the sphere is considerably smaller than
2πr 2 . Thus, acceleration is required that leads to more kinetic energy which means
increased added mass (see the defining equation (2.13)). The higher the mass of a
pendulum, the lower its frequency.
A bubble that is close to a wall can only get away from this wall either if its
deformation has a strong isotropic component or if its centroid possesses a substantial
positive velocity U . The latter case is therefore investigated in § 3.6.
3.6. Bubble deformation induced by motion perpendicular to a plane wall
Another generic test case is that of an initially spherical bubble set into motion
away from the wall at arbitrary Weber number. The corresponding Weber number is
defined as
def
We = ρL U02 Rt=0 /σ,
where U0 is the initial velocity of the bubble. If the Weber number is small, surface
tension tends to prevent large deformation of the bubble. As shown in § 3.5.7, an
initially high positive value of the velocity of the centroid is the only way for a
strongly anisotropically deforming bubble to escape from the influence of the wall
and get at a distance from the wall.
3.6.1. Effect of Weber number
In the first example, We is taken to be 0.12, which corresponds to an air bubble in
water at room temperature with radius of 1 mm and initial velocity of 0.1 m s−1 . The
pressure at infinity equals atmospheric pressure. The initial distance of the bubble
centroid to the wall, z0 /2, is again chosen to be 1.1Rt=0 . The initial pressure inside
the bubble is chosen such that the bubble would be in equilibrium if U0 would be
zero. The BEM starts from an initial velocity potential generated by a dipole with
strength 1/2U0 Rt=0 at the bubble centre plus a dipole with the same strength and
4.0
3.5
Interface position, x/Rt = 0

3.0
2.5
2.0
1.5
1.0
0.5
0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6

Dimensionless time, U0 t/Rt = 0
Figure 13. Positions of the leading edge (circle), centroid (triangle) and trailing edge (square
box); solid lines: We = 1.2; dashed lines: We = 0.12.
opposite direction at the centre of the bubble mirror image. The analytical method
starts, equivalently, from a spherical shape and from a given velocity potential with
coefficients am that correspond to two dipoles, each with strength as used in the BEM.
Then the corresponding velocities {ḃm } and U are computed by solving (2.9), with
a2 = 1/2U0 Rt=0 , 0 = a1 = a3 = a4 = a5 = . . . . Although it is somewhat more work in
the analytical approach, this initialization greatly facilitates the BEM computation.
Clearly, the initial condition is a mixed one of both deformation and velocity of the
centre U , although U is by far the most important generalized velocity.
In figure 13 the time evolution of the positions of the leading edge, centroid and
trailing edge is shown by dashed lines. It can be seen that the velocity of the centroid
is practically constant and that the deformation of the bubble is small. As a further
illustration, the bubble shape at various times is shown in figure 14.
In the next example the only difference with the previous example is that the
Weber number is increased to We = 1.2. This implies that the deformation restrictive
influence of surface tension is less. The time evolution of the positions of leading
edge, bubble centre of mass and trailing edge are shown in figure 13 by solid lines.
By comparing the two cases two phenomena are noticeable. First, the higher Weber
number corresponds to larger deformations, which can be deduced from the variations
in distance between leading and trailing edge, as well as from the shape history of
figure 15. Second, the velocity of the bubble centroid gets smaller for the higher
Weber number. This phenomenon is explained below.
3.6.2. The reason why velocity decreases with increasing deformation

Figure 16 shows that in the deformation case corresponding to We = 1.2 at least
about twenty generalized coordinates (b1 , b3 , b4 , . . .) are required to describe the shape
at all times. This figure shows both the distance to the wall and the dimensionless
time of passing. The changes of the amplitudes {bi } at times later than those shown
in figures 16 and 15 are comparable to those seen in figure 16. Figures 17 and 18
show the histories of the dominant added mass coefficients, α, γj and ψ1m , for the
case We = 1.2, as well as the corresponding distances from the wall. At large distance
from the wall the value of α of a sphere is 0.5. Without deformation, the value of
α would be monotonically decreasing with increasing distance from the wall (Geld
Time = 0 Time = 0.20451 Time = 0.40902
1 1 1
0 0 0
–1 –1 –1
0 1 2 3 0 1 2 3 0 1 2 3
Time = 0.61353 Time = 0.81804 Time = 1.0225
1 1 1
0 0 0
–1 –1 –1
0 1 2 3 0 1 2 3 0 1 2 3
Time = 1.2271 Time = 1.4316 Time = 1.6361
1 1 1
0 0 0
–1 –1 –1
0 1 2 3 0 1 2 3 0 1 2 3
Figure 14. Bubble shape at equidistant time instants for We = 0.12. The wall is indicated by
a set of straight parallel lines. Times are in units of Rt=0 /U0 , distances in units of Rt=0 .
2002). Figure 17 shows an increase of α with increasing distance to the wall that
therefore can only be caused by deformation. Values of ψ11 are in the range 4–4.5,
and those of ψ33 in the range 0.2–0.25, as in § 3.5.3. The importance of off-diagonal
terms in the matrix A is similar to that in § 3.5.3.
It will now be shown that velocity U decreases with increasing deformation mainly
due to the increased value of the well-known mass coefficient α. To explain the
influence of deformation on translation, the Euler–Lagrange equation corresponding
to generalized coordinate z, that is, the second row-equation of (2.19), is examined. It
can be written in the following form:

1 d ∂α 1 d ∞
{2U αV} − 2U 2 + V ḃm γm
V dt ∂z V dt m

∞
∂γm ∂ψij
+ − 2U ḃm −2 ḃi ḃj = 0. (3.31)
m
∂z ij
∂z
The histories of the dominant terms of (3.31) are shown in figure 19. This figure
shows that the terms with partial derivatives with respect to z in (3.31) are only
contributing in the vicinity of the wall, that is, at distances z/Rt = 0 < 3.5. For U0 -
induced deformation at We = 1.2, the last term on the left-hand side of (3.31) turns
Time = 0 Time = 0.20451 Time = 0.40902
1 1 1
0 0 0
–1 –1 –1
0 1 2 3 0 1 2 3 0 1 2 3
Time = 0.61353 Time = 0.81804 Time = 1.0225
1 1 1
0 0 0
–1 –1 –1
0 1 2 3 0 1 2 3 0 1 2 3
Time = 1.2271 Time = 1.4316 Time = 1.6361
1 1 1
0 0 0
–1 –1 –1
0 1 2 3 0 1 2 3 0 1 2 3
Figure 15. Bubble shape at equidistant time instants for We = 1.2. The wall is indicated by a
set of straight parallel lines. Times are in units of Rt=0 /U0 , distances in units of Rt=0 .
Dimensionless distance to the wall, z/(2Rt=0)
0.20 1.37 1.61 1.79 1.94 2.09 2.28

0.15
0.10
0.05
0
–0.05
–0.10
10* (b1 – Rt=0)
–0.15
b3
–0.20 b4
–0.25 b5
0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6

Dimensionless time, U0 t/Rt=0
Figure 16. Histories of shape coefficients bk (k > 2) and of 10(b1 − 1) for an initially spherical
bubble suddenly set into motion at We = 1.2. See figure 15 for the corresponding shape history.
Also, the corresponding distances of the centroid to the wall, z/2, are shown on top. The plain
solid line corresponds to b6 .
1.0 1.37 1.61 1.79 1.94 2.09 2.28

0.8
0.6
0.4
0.2
0
–0.2 γ1
–0.4 α
γ3
–0.6 γ4
–0.8
0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8
Figure 17. Histories of added mass coefficients α and γm for an initially spherical bubble
suddenly set into motion at We = 1.2. See figure 15 for the corresponding shape history. Also,
the corresponding distances of the centroid to the wall, z/2, are shown. The plain solid line
corresponds to γ5 .
0.6 1.37 1.61 1.79 1.94 2.09 2.28
0.4
ψ11/8
2ψ33
0.2
ψ13
ψ14
0
ψ15
–0.2
–0.4
0 0.5 1.0 1.5 2.0
Figure 18. Histories of added mass coefficients ψ1j (j = 3, 4, 5), ψ11 /8 and 2ψ33 for an initially
spherical bubble suddenly set into motion at We = 1.2. See figure 15 for the corresponding
shape history. The plain solid line corresponds to ψ16 .
out to be negligible, while for deformation-induced motion (§ 3.5) this term could not
be neglected because the bubble could not get away from the wall.
Further away from the wall, velocity U is therefore given by (3.29) that in the
present case can be written as

∞
U (αVU )|t=0 / (αV) + −γ1 ḃ1 − γj ḃj (2α) (3.32)
j =3
with the first term on the right-hand side comprising the value of αVU at time zero
when all generalized velocities are zero except U = U0 . Because of the corresponding
values of γm , the deformation velocities ḃm corresponding to the lowest values of m
yield the dominant deformation contributions in (3.32) at later times. The second
(a) (× 10–6) Dimensionless distance to the wall, z/(2Rt=0)
3 1.37 1.61 1.79 1.94 2.09 2.28
2 We = 1.2
Acceleration/(U02 /Rt = 0)
1
0
–1
–2 d/dt(2UαVb)/Vb
–2U2∂α/∂z
–3
d/dt(Σγj dbj/dt Vb)/Vb
–4
–2U Σbj∂γj/∂z
–5
0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6

(b) (× 10–6)
2 1.37 1.65 1.95 2.23 2.53 2.82

Acceleration/(U02/Rt=0)
–1
d/dt(2UαVb)/Vb
–2 –2U2∂α/∂z
d/dt(Σγj dbj/dt Vb)/Vb
We = 1.2
–3 –2U Σbj∂γj/∂z
0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6

Figure 19. Histories of dimensionless accelerations for motion-induced bubble deformation

at We = 1.2 (a) and We = 0.12 (b); Vb = V. The main contributions to the second
Lagrange–Hamilton equation are shown. Note that the initial scaled accelerations are the
same for both Weber numbers.

term on the right-hand side, − ∞ m ḃm γm /(2α) is at maximum 15 % of U |t = 0 for
all strong deformations shown in figure 15, while U changes by more than 50 %
at the same time. As a result, the velocity U is roughly inversely proportional to
added mass α, with 30 % deviations due to the ḃm γm -terms of (3.32). There is a
direct coupling between bubble shape and motion of the centroid (U ) since the
familiar added mass coefficient α depends on shape because α depends on the bm -
coefficients. This dependency on shape is dominant. However, some specific (per unit
of mass) momentum of the motion of the centroid αVU is transformed to specific
deformation momentum at the interface, the Vḃm γm terms. For example, γ1 is negative
and asymptotically goes to a value of −0.2 with increasing distance to the wall (see
figure 17), implying that isotropic expansion (ḃ1 > 0) increases the momentum 2αVU
anywhere (see (3.32)). Expansion tends to propel the bubble centre away from the wall.
The high-frequency oscillatory changes of b1 (see (3.13) and § 3.5.5) therefore cause
similar oscillatory changes in U . These high-frequency oscillations in U are in phase
with the isotropic shape oscillations corresponding to b1 . As in the generic example
5
Interface position, x/Rt=0

3
1 No buoyancy, no drag
Buoyancy, no drag
No boyancy, drag
0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
Figure 20. Comparison of the effects of gravity and Levich drag on long-term interfacial
deformation for We = ρL R0 U02 /σ = 1.2 and p∞ /(ρL U02 ) = 8571. The result with gravity and
without drag is for Bo = g̃R02 ρL /σ = 0.01 and the results for drag without gravity is for
Re = R0 U0 ρL /μ = 926. For a fluid with ρL = 800 kg m−3 , μ = 1.55 · 10−4 kg.m s−1 and
σ = 0.07 N m−1 and for g̃ = 9.8 m s−12 these values correspond for example to a bubble with
initial radius R0 = 0.328 mm with initial velocity U0 = 0.545 m s−1 . Positions of the leading edge,
centroid and trailing edge are shown without the symbols of figure 13.
of § 3.5, these are the high frequencies that can also be observed in the accelerations,
see figure 19. The apparent mirror symmetry for z/Rt=0 > 3.5 in figure 19 of the two
terms dtd {2U αV}/V and dtd {V ∞ m ḃm γm }/V reflects the constancy of the left-hand
side of (3.32). The remaining changes in each of the two terms on the left-hand side
of (3.32) are mainly due tothe non-isotropic deformation revealed by, for example,
figure 15. For the term V ∞ m ḃm γm this is obvious (see also figures 16 and 17). For
the term 2αVU this follows from the changes in α in figure 17. The added mass
coefficient α increases with increasing degree of deformation (see above). The stronger
the deformation, the slower the translational motion of the bubble. This trend was
already concluded from figure 13, in the above, and is now explained.
3.6.3. The effects of gravity and viscosity
The results shown up to now have been obtained for Bond number equal to
zero and Reynolds number infinitely large, that is, with the neglect of gravity and
viscosity. To investigate the effect of gravity and viscosity, two test cases have been
investigated with finite values of the Bond and the Reynolds numbers. In each of
these test cases, three conditions have been prescribed: one without gravity and
without drag, one with gravity and without drag, one without gravity and with
drag. Two high Weber numbers (1.2 and 2.07) have been applied in order not to
have more than one agency (surface tension) controlling the bubble shape. The total
runtime has been made more than two times that of figure 13, while the same
length scales as plotted in this figure have been employed. The results are shown
in figure 20 and in figure 21. The corresponding dimensionless numbers are given
in the captions, for figure 21 for example We = ρL R0 U02 /σ = 2.07, p∞ /(ρL U02 ) = 355,
Bo = g̃R02 ρL /σ = 0.0473 or zero and Re = R0 U0 ρL /μ = 709 or infinitely large. The
results make clear why the runtime has been chosen so large: only at the latest
times gravity and viscosity may introduce small deviations in shape. For a fluid with
ρL = 965 kg m−3 , μ = 2.92 · 10−4 kg.m s−1 and σ = 0.05 N m−1 and for g̃ = 9.8 m s−2
5
Interface position, x/Rt=0

3
1 No buoyancy, no drag
Buoyancy, no drag
No boyancy, drag
0 0.5 1.0 1.5 2.0 2.5 3.0 3.5

Figure 21. Comparison of the effects of gravity and Levich drag on long-term interfacial
deformation for We = ρL R0 U02 /σ = 2.07 and p∞ /(ρL U02 ) = 355. The result with gravity and
without drag is for Bo = g̃R02 ρL /σ = 0.0473 and the results for drag without gravity is for
Re = R0 U0 ρL /μ = 709. Positions of the leading edge, centroid and trailing edge are shown
without the symbols of figure 13.
the dimensionless numbers of figure 21 correspond to a bubble with initial radius

R0 = 0.463 mm with initial velocity U0 = 0.463 m s−1 , for example. Another example of
practical fluid and flow conditions corresponding to the values of the dimensionless
numbers chosen is given in the captions of figure 20.
The results show that if the Reynolds number exceeds 1000 and if the Bond number
is less than 0.01 the effects of gravity and viscosity can be ignored at times less than
3R0 /U0 , even if the Weber number is as large as 2.
4. Conclusions
The added mass tensor of a bubble moving and strongly deforming in
incompressible fluid in the vicinity of a plane wall has been computed. The dynamics
of axisymmetric bubble deformation at a distance from a plane solid wall have
been computed with this added mass tensor, in which most coefficients relate to
shape deformation. These dynamics have alternatively been computed with the BEM.
No artificial viscosity was needed in the BEM to obtain stability. Full deformation
computations can be accelerated by selecting an artificially low value of the polytropic
constant Cp /Cv . Good agreement between results of the two methods has been found.
Solutions of cases available in the literature are correctly reproduced. The explicit
expressions for the added mass tensor could for isotropic (spherical) deformation be
reduced to results known from a previous study. The coupled dynamics of motion
and deformation is most conveniently described by a tensor with matrix A. This
symmetric matrix has relatively high values at the diagonal and the neighbouring
off-diagonal places, as well as in the first row and column (corresponding to the
fundamental radial oscillation mode and b1 ). The coupling with the motion of the
centroid is via the terms on the second row (and second column).
The Rayleigh–Plesset equation, governing isotropic deformation, has been extended
with terms to account for arbitrary axisymmetric deformation and to account for the
proximity of a plane wall. A governing equation for the important deformation mode
connected to the second Legendre polynomial, P2 , has been presented and simplified
in order to obtain an equation for this mode that is similar to the Rayleigh–Plesset
equation. The angular frequencies of periodic solutions of the linearized governing
equations in the absence of a wall, like ωRP for the isotropic mode, follow easily from
the equations presented. Because of the presence of an U 2 term in the governing
equation, motion of the bubble centre induces anisotropic deformation even at large
distances from the wall. A periodically and anisotropically deforming bubble without
initial velocity of the centroid is eventually always driven towards the wall because
of the fact that shape deformation and oscillation couple to volume oscillations.
The angular frequencies of periodic solutions are decreased if the distance to the
wall is decreased. A strong coupling of anisotropic deformation modes with isotropic
deformation exists. The frequency doubling of the relatively low frequencies of the
anisotropic modes in the isotropic mode is explained. The frequency of this isotropic
mode is even very close to the wall accurately given by

3
2
ωRP 1 + R/z − (R/z) 4
(2α) .
2 t=0
The resonance case introduced by Longuet-Higgins (1989a) is shown to have

non-singular behaviour when ωRP is equal to the twice the radian frequency of a
fundamental deformation mode. The larger the amplitude of deformation components,
the more the motion of the bubble centre is slowed down. It has been shown that the
main reason for this is the increase of the ‘classical’ added mass coefficient α with
increased deformation.
It has been shown in what cases, and how, the equation that dominates motion of
the bubble centroid in many cases can be simplified. This simplified equation may
facilitate prediction of motion of a deforming bubble in cfd programmes. When
initially a bubble is set into motion away from a plane wall, the simplified equation of
motion can further away from the wall be integrated, to yield the following equation
for the velocity U of the centre:

∞
U (αVU )|t=0 / (αV) + −γ1 ḃ1 − γj ḃj (2α)
j =3
with the second term on the right-hand side about 15 % of the first term at maximum,
even for cases of strong deformation. Since only the added mass coefficients γj need
to be evaluated, the effect of strong deformation on motion of a bubble not too close
to a plane wall is relatively easily computed with the above equation. This result
holds when gravity and drag are negligible, but it has been shown that these agencies
may be neglected in many practical cases of strong deformation. Criteria to specify
the conditions of validity of the neglect of gravity and drag have been presented.
This work was financially supported by the EC-project AD-700-2.
REFERENCES
Abramowitz, M. & Stegun, I. A. 1972 Handbook of Mathematical Functions. Dover.
Auton, T. R., Hunt, J. C. R. & Prud’Homme, M. 1988 The force exerted on a body in inviscid,
unsteady non-uniform rotational flow. J. Fluid Mech. 197, 241–257.
Bagchi, P. & Balachandar, S. 2003 Inertial and viscous forces on a rigid sphere in straining flows
at moderate Reynolds numbers. J. Fluid Mech. 481, 105–148.
Benjamin, T. B. & Ellis, A. T. 1990 Self-propulsion of asymmetrically vibrating bubbles. J. Fluid
Mech. 212, 65–80.
Brennen, C. E. 1995 Cavitation and Bubble Dynamics. Oxford University Press.
Doinikov, A. A. 2004 Translational motion of a bubble undergoing shape oscillations. J. Fluid
Mech. 501, 1–24.
Eller, A. I. 1970 Damping constants of pulsating bubbles. J. Acoust. Soc. Am. 47, 1469–1470.
Feng, Z. C. & Leal, L. G. 1995 Translational instability of a bubble undergoing shape oscillations.
Phys. Fluids 7, 1325–1336.
van der Geld, C. W. M. 2002 On the motion of a spherical bubble deforming near a plane wall.
J. Engng Math. 42, 91–118.
van der Geld, C. W. M. 2009 The dynamics of a boiling bubble before and after detachment. Heat
Mass Transfer 45 (1), 831–846.
Gradsteyn, I. S. & Ryzhyk, I. M. 1980 Table of Integrals, Series and Products. Academic Press.
Hobson, E. W. 1955 The Theory of Spherical and Ellipsoidal Harmonics. Cambridge University Press.
Howe, M. S. 1995 On the force and moment of a body in an incompressible fluid, with application
to rigid bodies and bubbles at low and high Reynolds numbers. Q. J. Mech. Appl. Math 48,
401–426.
Kantorovich, L. V. & Krylov, V. L. 1958 Approximate Methods of Higher Analysis. Interscience.
Lamb, H. 1932 Hydrodynamics, 6th edn. Cambridge University Press.
Legendre, D., Borée, J. & Magnaudet, J. 1998 Thermal and dynamic evolution of a spherical
bubble moving steadily in a superheated or subcooled liquid. Phys. Fluids 10 (6), 1256–1272.
Longuet-Higgins, M. S. 1989a Monopole emission of sound by asymmetric bubble oscillations.
Part 1. Normal modes. J. Fluid Mech. 201, 525–541.
Longuet-Higgins, M. S. 1989b Monopole emission of sound by asymmetric bubble oscillations.
Part 2. An initial value problem. J. Fluid Mech. 201, 543–565.
Longuet-Higgins, M. S. 1989c Some integral theorems relating to the oscillations of bubbles.
J. Fluid Mech. 204, 159–166.
Magnaudet, J. & Eames, I. 2000 The motion of high-Reynolds-number bubbles in inhomogeneous
flows. Annu. Rev. Fluid Mech. 32, 659–708.
Mei, C. C. & Zhou, X. 1991 Parametric resonance of a spherical bubble. J. Fluid Mech. 229, 29–50.
Meiron, D. I. 1989 On the stability of gas bubbles rising in an inviscid fluid. J. Fluid Mech. 198,
101–114.
Mougin, G. & Magnaudet, J. 2002 The generalized kirchhoff equations and their application to
the interaction between a rigid body and an arbitrary time-dependent viscous flow. Intl J.
Multiph. Flow 28, 1837–1851.
Pelekasis, N. A. & Tsamopoulos, J. A. 1993 Bjerknes forces between two bubbles. Part 1. Response
to a step change in pressure. J. Fluid Mech. 254, 467–499.
Pozrikidis, C. 1997 Introduction to Theoretical and Computational Fluid Mechanics. Oxford
University Press.
Prosperetti, A. 1977 Thermal effects and damping mechanisms in the forced radial oscillations of
gas bubbles in fluids. J. Acoust. Soc. Am. 61, 17–27.
Prosperetti, A. 1984 Bubble phenomena in sound fields: part two. Ultrasonics 22, 115–123.
Telles, J. C. F. 1987 A self-adaptive coordinate transformation for efficient numerical evaluation
of general boundary element integrals. Intl J. Numer. Meth. Engng 24, 959–973.
Toose, E. M., van den Ende, H. T. M., Geurts, B. J., Kuerten, J. G. M. & Zandbergen, P. J.
1996 Axisymmetric non-Newtonian drops treated with a boundary integral method. J. Engng
Math. 30, 131–150.
Tsamopoulos, J. A. & Brown, R. A. 1983 Nonlinear oscillations of inviscid drops and bubbles.
J. Fluid Mech. 127, 519–537.
Versluis, M., Schmitz, B., von der Heydt, A. & Lohse, D. 2000 How snapping shrimp snap:
through cavitating bubbles. Science 289, 2114–2117.
View publication stats

Axisymmetric Dynamics of A Bubble Near A Plane Wal

Uploaded by

Copyright:

Available Formats

Axisymmetric Dynamics of A Bubble Near A Plane Wal

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Axisymmetric Dynamics of A Bubble Near A Plane Wal

Uploaded by

Copyright:

Available Formats

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

Axisymmetric dynamics of a bubble near a plane wall

Article in Journal of Fluid Mechanics · December 2009

Cornelis Wilhelmus Maria van der Geld J. G. M. Kuerten

SEE PROFILE SEE PROFILE

Jets in bubbles View project

Contactless and Contactful Boiling for Controlled Evaporation View project

The user has requested enhancement of the downloaded file.

Axisymmetric dynamics of a bubble near

† Email address for correspondence: c.w.m.v.d.geld@tue.nl

b2 is obviously redundant and cannot be used as a generalized coordinate. All other

The coeﬃcients dij depend on the bm which depend on time.

Here [j ] denotes the integer part of j . The operator

is now applied to the boundary condition (2.6). This yields

with Fmn and Gmn the following functions of the bi -coeﬃcients:

Here the set of structural coeﬃcients {ωmnfj } is deﬁned by

def 2 2i(i + 1) 2(i − 1)(i − 2)

2.3. Boundary element method

Error in shape prediction, Err

0.5 Present work

Added mass coefficient, α

Figure 3. Comparison of added mass coeﬃcients of an oblate ellipsoid given by Lamb

β = − c̃/ωRP Rt3= 0 . The analytical method starts with

The ﬁrst row-equation of (2.19) reads

Time = 0.61316 Time = 0.81755 Time = 1.0219

Time = 1.2263 Time = 1.4307 Time = 1.6351

3.5. Bubble motion induced by anisotropic large-amplitude deformation

(b) We = 0.12; γ = 1.4

3.5.2. Eﬀect of the wall on the angular frequency of bk -excitation

3.5.3. Coupling of motion and of deformation modes

3.5.4. A Rayleigh–Plesset equation with deformation

(b) We = 0.12; γ = 1.4

Rayleigh–Plesset equation, can be written as:

Figure 10. History of acceleration of the ﬁrst generalized coordinate b1 corresponding

3.5.5. Frequency-doubling of b1 component, away from the wall

2ψ11 b̈1 + 2ψ31 b̈3

3.5.6. Energy transfer

easily shows that T + σ Ab + p∞ V− pB,0 Vγ0 1−γ

Energy/(Vb ρL (db3/dt|t =0)2/4)

Figure 12. Histories of dimensionless accelerations for b5 -excitation. We

Interface position, x/Rt = 0

0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6

3.6.2. The reason why velocity decreases with increasing deformation

Time = 0.61353 Time = 0.81804 Time = 1.0225

Time = 1.2271 Time = 1.4316 Time = 1.6361

Time = 0.61353 Time = 0.81804 Time = 1.0225

Time = 1.2271 Time = 1.4316 Time = 1.6361

0.20 1.37 1.61 1.79 1.94 2.09 2.28

0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6

1.0 1.37 1.61 1.79 1.94 2.09 2.28

Dimensionless distance to the wall, z/(2Rt=0)

0.6 1.37 1.61 1.79 1.94 2.09 2.28

(a) (× 10–6) Dimensionless distance to the wall, z/(2Rt=0)

3 1.37 1.61 1.79 1.94 2.09 2.28

Dimensionless distance to the wall, z/(2Rt=0)

2 1.37 1.65 1.95 2.23 2.53 2.82

0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6

Figure 19. Histories of dimensionless accelerations for motion-induced bubble deformation

Interface position, x/Rt=0