You are on page 1of 54

Numerical simulation of spherical bubble collapse by a uniform bubble pressure approximation and detailed description of heat and

mass transfer with phase transition

Journal Pre-proof

Numerical simulation of spherical bubble collapse by a uniform


bubble pressure approximation and detailed description of heat and
mass transfer with phase transition

Jean Manuel Bermudez-Graterol, Mehrdad Nickaeen,


Romuald Skoda

PII: S0307-904X(21)00115-3
DOI: https://doi.org/10.1016/j.apm.2021.02.031
Reference: APM 13938

To appear in: Applied Mathematical Modelling

Received date: 24 September 2020


Revised date: 25 January 2021
Accepted date: 21 February 2021

Please cite this article as: Jean Manuel Bermudez-Graterol, Mehrdad Nickaeen, Romuald Skoda, Nu-
merical simulation of spherical bubble collapse by a uniform bubble pressure approximation and de-
tailed description of heat and mass transfer with phase transition, Applied Mathematical Modelling
(2021), doi: https://doi.org/10.1016/j.apm.2021.02.031

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition
of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of
record. This version will undergo additional copyediting, typesetting and review before it is published
in its final form, but we are providing this version to give early visibility of the article. Please note that,
during the production process, errors may be discovered which could affect the content, and all legal
disclaimers that apply to the journal pertain.

© 2021 Published by Elsevier Inc.


Highlights

• A computationally efficient model for detailed mass and heat transfer is


presented

• Even for bubble collapse, homobaricity and equilibrium assumptions re-


main valid
• Temporal and spatial resolution of transport processes reveal detailed flow
physics

• Bubble dynamics, mass and heat transfer occur in a time-shifted and non-
linear way

1
Numerical simulation of spherical bubble collapse by a
uniform bubble pressure approximation and detailed
description of heat and mass transfer with phase
transition

Jean Manuel Bermudez-Graterola,∗, Mehrdad Nickaeena , Romuald Skodaa


a Chair of Hydraulic Fluid Machinery, Ruhr University Bochum, Universittsstr. 150, 44801
Bochum, Germany

Abstract
A mathematical model for the simulation of the dynamics of spherical vapor-air
bubbles and its numerical implementation is presented. Heat and mass transfer
and phase transition in terms of evaporation and condensation as well as air
absorption and desorption are considered. Flow variables are discretized by a
mixed finite volume / finite difference scheme and solved either by a Crank-
Nicolson or Runge-Kutta scheme. Due to the assumption of homogeneous bub-
ble pressure (homobaricity), the solution of momentum equations is restricted
to the Rayleigh-Plesset equation which makes the model computationally ef-
ficient. The model is validated by measurement data for bubble growth and
applied to bubble collapse and rebound. By a comparison with Navier-Stokes
results from literature, the homobaricity assumption is shown to be appropriate
even in the last stage of bubble collapse. The relation of the local velocity and
temperature field with heat and mass transfer is discussed. Equilibrium bubble
interface conditions (liquid and gaseous side have the same temperature and
chemical potential) are compared to non-equilibrium conditions and are shown
to yield the same local velocity and temperature field for each stage of bubble
collapse and rebound.
Keywords: 1D CFD, bubble collapse, vapor-gas bubble, heat and mass
transfer, equilibrium bubble interface condition, homobaricity

1. Introduction

Cavitation and bubble dynamics are usually understood as formation, os-


cillation and collapse of vapor bubbles and occur due to rapid changes in the

∗ Correspondingauthor
Email address: Jean.BermudezGraterol@rub.de (Jean Manuel Bermudez-Graterol)

Preprint submitted to Applied Mathematical Modelling February 27, 2021


Nomenclature

3
4
pressure of the surrounding liquid. Cavitation may have serious complications
in technical applications such as generation of sound, vibration and cavitation
erosion. Cavitation erosion is associated with violent bubble collapse in the
proximity of solid walls [1]. In technical liquids, a considerable amount of non-
condensable gas, e.g. air is dissolved, so that bubbles consist of a mixture of
vapor and non-condensable gas. During the collapse, the air within the bub-
ble is compressed, experiences high temperature peaks and causes a subsequent
rebound. A particular type of cavitation, the presumably rather slow diffusion-
driven absorption and desorption of non-condensable gas in the liquid, often
referred to as air release, is also discerned as a serious issue in running e.g. hy-
draulic machinery and systems [2, 3, 4]. There seems to be a strong interaction
of vapor cavitation with air release [5, 6, 7] that comprises the transport of heat
and mass with phase transition, whose knowledge is still incomplete. The de-
tailed investigation of transport processes during bubble dynamics demands the
coupled consideration of nonlinear mass, momentum and heat transfer across
the bubble surface and the spatial and temporal distribution of flow variables
within the bubble and the surrounding liquid. Most of the previous studies on
bubble dynamics comprise more or less strong simplifications on the coupled
transport processes that are reviewed further below. In the present study, we
propose a model which strives to resolve the complete transport processes both
spatially and temporally.
The investigation of cavitation and air release is often reduced to the consid-
eration of single spherical bubble dynamics, a simplification that is also adopted
in the present study. Since the pioneering work by Rayleigh [8], who investigated
empty as well as gas filled bubbles in an incompressible liquid and neglected
surface tension and viscous effects, several extensions of the Rayleigh-equation
have been proposed to describe the motion of the bubble surface. An impor-
tant extension has been proposed in terms of the Rayleigh-Plesset equation by
the inclusion of surface tension and viscosity [9]. Further extensions consider
the bubble interior’s thermodynamics [10], a non-Newtonian fluid [11, 12], or
involve approximations for heat diffusion [13, 14] and liquid compressibility ef-
fects [15, 16, 17, 18]. All these and related studies did not consider detailed, i.e.
spatially and temporally resolved heat and mass transport. Due to the absence
of PDEs only nonlinear ODEs need to be solved, and several analytical approx-
imations and numerical solution methods have been presented, e.g. [19, 20], to
list only a few.
The most accurate spatial approximation of bubble dynamics problems con-
sists of a three-dimensional resolution of the bubble motion. There is a variety
of simulation studies available that involve three-dimensional effects of single-
bubble dynamics, e.g. bubble-wall interaction [21, 22, 23, 24] including the
complex flow in viscous wall boundary layers [25, 26], bubble collapse induced
shocks [27], interaction of single bubbles with shock waves [28], underwater ex-
plosions [29], nucleation [30], bubble-bubble-interaction by coalescence [31] or
due to shock waves [31, 32]. Although these studies comprise advanced nu-
merical techniques, to the best knowledge of the authors, only few 3D studies
such as [33] consider the synchronous mass and heat transfer together with the

3
transport of vapor as well as non-condensable gas, i.e. air.
Cavitating flow includes a tremendously high number of single bubbles.
A three-dimensional approach of numerous single bubbles, including the syn-
chronous mass and heat transfer within each bubble is unfeasible even with
recent computer capabilities. An interesting coupled ansatz is to embed the
motion of a multitude of single bubbles in a three dimensional Eulerian CFD
framework through a Lagrangian treatment of individual bubbles [34, 35, 36].
It is essential that the single bubble model is computationally efficient, due to
the large number of bubbles that are introduced into the Euler framework. Ac-
cordingly, the embedded bubble models in [34, 35, 36], are based on the pure
Rayleigh-Plesset equation and do not take into account detailed heat and mass
transport processes. Therefore, in this study we present a single bubble model
that on the one hand comprises detailed spatial transport processes, and on the
other hand is significantly faster than solving the Navier Stokes equations for
each single bubble. Based on this single bubble model, a multitude of single
bubbles can be incorporated into an Euler framework in subsequent studies at
feasible computational costs. Hence, we maintain the assumption of a spherical
bubble which also allows us to formulate the governing equations in spherical
coordinates.
Multiple approximations with the spatial transport of vapor [37] or non-
condensable gas [38, 39, 40] have been proposed. Sochard et al [41] studied
homogenous sonochemistry and have also considered gas-vapor inter-diffusion
within the bubble as well as the liquid-vapor phase transition across the in-
terface, but gas diffusion across the interface was neglected. Storey and Szeri
[42] proposed the solution of the Navier-Stokes equations with inter-diffusing
non-condensable gases in the bubble interior by a spectral collocation method
[43]. A similar method was used by Kamath and Prosperetti [44]. However, no
phase transition over the interphase and no vapor content in the bubble were
considered. On the other hand, the complete compressible Navier-Stokes equa-
tions with radial resolution of the concentration and temperature field within
and outside of a spherical single bubble have been solved [45, 46, 47]. Air dif-
fusion was modelled by Fick’s law, and the interphase condition for air was
considered by Henry’s law. Phase transition was modeled as a thermal non-
equilibrium process by the Hertz-Knudsen relation [18, 48]. An extended com-
pressible Rayleigh-Plesset equation considering mass transfer [18] determined
the time-dependent bubble interphase position. Jinbo et al [49] used a similar
approach as Matsumoto et al [45] but resolved the bubble interface with a level-
set [50] and ghost fluid [51] method. A similar method has been proposed by
Storey and Szeri [52] including chemical reactions. Yamamoto et al [53] carried
out an investigation based on molecular dynamics and proposed that a small
amount of non-condensable gas strongly affects the temperature field inside the
collapsing bubble. Thus, these studies [45, 46, 49, 52] constitute a complete
mathematical model for the simulation of the detailed heat and mass transport
processes for a spherical bubble, together with the solution of the Navier-Stokes
equations.
The numerical solution of the full Navier-Stokes equations comprises the

4
non-linear momentum conservation equations. This results in considerable nu-
merical effort that limits the number of single bubbles which potentially may
be incorporated into an Euler-Lagrange approach. An interesting simplification
is the determination of the bubble interior velocity field from the energy equa-
tion instead of the momentum equation. Kawashima and Kameda [54] used an
approach that originates from Nigmatulin et al [55] and assumes homobaric-
ity within the bubble, i.e. the bubble pressure is spatially homogeneous and
only dependent on time. They evaluated the spatial velocity distribution within
the bubble by an analytical expression that is derived from the energy equa-
tion. Additionally, an expression for the temporal evolution of the spatially
constant bubble pressure has been derived from the energy equation. A similar
approximation has been used by other authors [14, 56, 57, 58]. The homobaric-
ity simplification is justified if the bubble wall velocity is much lower than the
speed of sound within the bubble [55] and is assumed to remain valid for most
stages of bubble dynamics [42, 54, 55, 59], even for bubble collapse, which will
be investigated in detail in the present study.
In the present study, we adopt this homobaricity-based approach in combi-
nation with a non-equilibrium interface condition in terms of the Hertz-Knudsen
relation and thus circumvent the solution of the Navier-Stokes equations. Thereby,
our proposed model is detailed as well as computationally efficient. Due to its
lower computational cost, the model may allow the synchronous consideration
of a multitude of single bubbles within an Euler-Lagrange framework. It should
be pointed out that in this study, we focus on the single bubble modelling, and
its embedment into an Euler-Lagrange framework is left for future studies. Of
particular interest is the validity of the homobaricity assumption for different
bubble dynamics scenarios, so that we assess it also for bubble collapse. We
discuss the particular numerical implementation as well as the effects of heat
and mass transfer on bubble motion and temperature distribution around the
bubble in more detail than most previous studies. For a probable future expan-
sion of the bubble dynamics model from water towards multi-component fluid
mixtures, it might be much simpler to use an equilibrium interface condition,
that is commonly applied e.g. for droplet evaporation. Thus we assume a simple
equilibrium condition at the bubble interface rather than the common thermal
non-equilibrium condition. A thorough assessment of equilibrium on bubble
collapse scenarios has to our knowledge not been presented yet and opens the
opportunity to adopt complex fluid mixtures beyond a water-air system. The
validity and restriction of these simplifications is verified on different bubble
dynamics scenarios.
The paper is organized as follows: in the subsequent subsection, the math-
ematical formulation is presented. In section 3, we present a summary of the
numerical solution method with a certain emphasis on the implementation of
bubble interface boundary conditions. Further details on the mathematical for-
mulation and numerical implementation can be found in [60]. After the presen-
tation of the validation with literature measurement data on bubble growth in
section 4, bubble collapse and rebound results are presented in section 5, where
detailed spatial transport of heat and mass is discussed and the homobaricity

5
assumption together with an equilibrium vs. non-equilibrium phase transition
approach is evaluated.

2. Mathematical model formulation

2.1. Model outline, notations and assumptions


A single spherical bubble with initial radius R0 contains an ideal binary
mixture of vapor water and non-condensable gas, i.e. air. Air is treated as a
single pseudo-fluid. The bubble is surrounded with an infinite amount of liquid
water. Outside the bubble, air is dissolved and transported into the liquid.
Within the bubble (superscript β = G) and outside the bubble (superscript
β = L), mixture values (subscript m) are evaluated by a mass fraction weight
yαβ = ρβα /ρβm , where index α means either water (subscript α = H2 O) or air
(subscript α = Air), that adds to one:
β β
yH 2O
+ yAir =1 (1)

It is assumed that the bubble center is stationary with a moving bubble


wall. At the bubble wall, phase transition of liquid to vapor and vice versa
is considered. The momentum balance is not spatially resolved either in the
gaseous phase within the bubble or the liquid phase surrounding it. In the
bubble, we obtain the velocity field by the energy equation as summarized in
section 2.2.1. In the liquid, we solve the integral momentum balance in terms
of an extended Rayleigh-Plesset equation [18] which yields the time-dependent
bubble radius, i.e. the instantaneous location of the interface between gaseous
and liquid phase. The bubble interface conditions at the bubble wall are either
based on thermal and phase equilibrium or non-equilibrium as will be explicated
in subsection 2.3.2.
The assumptions on the mathematical model can be summarized as follows:

• The bubble is spherically symmetric, i.e. bubble wall motion and transport
processes are considered in radial direction only.
• Vapor and gas within the bubble are an ideal mixture of thermally and
calorically ideal gases.
• Air at the bubble wall obeys Henry’s law.

• Due to the small amount of air resolved in the liquid, liquid properties of
pure water are assumed.
• Homobaricity holds within the bubble, i.e. pressure is homogeneous in
radial direction.

• Fick’s law holds for mass diffusion, and mass diffusion due to pressure or
temperature gradients is neglected.

6
It should be noted that the assumption of an ideal gas holds only if the
pressure is well below the critical pressure. As well, Henry’s law also holds only
for low pressure. The pressure may however be large in particular at the final
stage of bubble collapse, as will be discussed in more detail in the result section
5.2.

2.2. Governing equations


The governing equations are provided in spherical coordinates. For both,
gaseous bubble interior and its surrounding liquid, the mass conservation of
mixture m and its components α reads:

∂ ρβm 1 ∂ 
+ 2 r2 ρβm uβ = 0 (2)
∂t r ∂r
  
∂ ρβm yαβ 1 ∂  1 ∂ ∂y β
+ 2 r2 ρβm yαβ uβ − 2 r2 ρβm Dβ α = 0 (3)
∂t r ∂r r ∂r ∂r
where α = H2 O or α = Air, and β = G or β = L. Note that since we solve
the governing equations on a moving computational grid, the absolute velocity
uβ will later be substituted by the velocity ũβ relative to the grid movement
as will be explicated in section 3.1. In Eq. 3, the binary diffusion coefficient
β β
fulfils the symmetry condition DH 2 O→Air
= DAir→H 2O
= Dβ . It is noteworthy
already here that we solve Eq. 3 in the numerical procedure and assume that
Eq. 2 is readily fulfilled when we sum up Eq. 3 for α = H2 O and α = Air, once
β β
yH 2O
and yAir are available. In fact, by summing up Eq. 3 we obtain Eq. 2,
assuming the symmetry of the binary diffusion coefficients.

2.2.1. Within bubble


It is assumed that the specific heat and gas constants of vapor as well as air
are constant. Mixture values are evaluated by a weighted mean:

ΦG G G G G
m = yH2 O ΦH2 O + yAir ΦAir (4)

In Eq. 4, Φ may correspond to internal energy e, enthalpy h, or specific


heat cp or cv , with h = e + p/ρ. The specific gas constant is obtained by
<α = <Univ /Mα , and partial pressure is evaluated for a thermally ideal gas by
pG G G
α = ρα <α T , where α = H2 O or α = Air. Mixture pressure fulfils Dalton’s
law, p = pH2 O + pG
G G G
Air , so that also for p the ideal gas law holds:

p G = ρG
m <m T
G
(5)

It should be pointed out that albeit pG is spatially homogeneous within the


bubble (homobaricity), its contributions pG G
H2 O and pAir are usually not. Since
the gaseous phases are also calorically ideal, eα = cv,α T G and hG
G G G G
α = cp,α T , as
well as eG G
m = cv,m T
G
and hG G
m = cp,m T
G
hold.

7
The mixture energy conservation, neglecting viscous effects, reads:

∂ ρG G
m em 1 ∂  1 ∂ 
+ 2 r 2 ρG G G
m hm u =− 2 r2 q (6)
∂t r ∂r r ∂r
The heat flux is evaluated by:
!
G G G
∂T ∂yH2O
∂yAir
q= −λG
m − ρG
mD
G
hG
H2 O + hG
Air (7)
∂r ∂r ∂r

The first part of the right-hand side of Eq. 7 is the heat transport by
conduction, and the second part describes the heat transport by each of the
G
diffusing components. Employing the homobaricity assumption ∂p ∂r = 0, an
integro-differential expression for the velocity profile is obtained from energy
conservation, Eq. 6:
Z r 
r dpG 1 <m dpG
uG = − G + 2 G G(r) + G r2 dr (8)
3p dt r p 0 c p,m dt

with:
( " !#
G G
<m 1 ∂ 2 G G G
∂yH 2O
∂yAir
G(r) = G r ρm T D <H2 O + <Air
cp,m r2 ∂r ∂r ∂r
" !#
T G cG
v,m 1 ∂ 2 G G
∂yHG
2O
G
∂yAir
+ r ρm D <H2 O + <Air
<m r2 ∂r ∂r ∂r
!
G G G
G ∂T
∂yH2O
∂yAir
+ ρG
mD cG
v,H2 O + cGv,Air
∂r ∂r ∂r
 
1 ∂ ∂T G
+ 2 r 2 λG
m (9)
r ∂r ∂r

By evaluating Eq. 9 at the bubble wall r = R and re-arranging, an expression


for the temporal pressure evolution within the bubble is obtained:
Z R
2 G G
−R p u w + r2 G(r) dr
dpG 0
= Z R (10)
dt R3 <m 2
− G
r dr
3 0 cp,m

We provide a detailed derivation of equations 8 to 10 in [60]. Eqs. 6 to 10


have also been presented by Nigmatulin et al [55] and Kawashima et al [54] in
a similar form. It is noteworthy that in Eq. 7, Nigmatulin et al. [55] applied
the inner energy instead of enthalpy, see equation (1) in [55], while we assume
the latter more appropriate.
The velocity at the gaseous side of the bubble wall uG w is obtained by the

8
conservation of mass flux ṁ00 = ṁ00H2 O + ṁ00Air through the bubble wall ṁ00 =
 
ρG G
m,w Ṙ − uw which is assumed positive when flowing into the bubble. Thus:

ṁ00
uG
w = Ṙ − (11)
ρG
m,w

Together with the ideal gas assumption, Eqs 6 and 7 yield:


G G
 G

G ∂T G ∂T 1 ∂ 2 G ∂T
ρG c
m p,m = −ρ G G
c
m p,m u + r λ m
∂t ∂r r2 ∂r ∂r
!
G G G
G ∂T
∂yH2O
∂yAir dpG
+ ρG
mD cG
p,H2 O + cG
p,Air + (12)
∂r ∂r ∂r dt

With Eqs. 3 (β = G), 5, 8, 10 and 12 an equation set for the evaluation of


the gas field variables yαG , ρG G
m , u and T
G
as well as the time-dependent pressure
G
p is available.

2.2.2. Within liquid


L
Since the amount of dissolved air in terms of yAir is generally small, it is
assumed that the property change due to dissolved air can be neglected, so
that mixture and transport properties are evaluated for pure water, e.g. ρLm ≈
ρLH2 O := ρL and λLm ≈ λLH2 O := λL . Assuming cLp,m ≈ cLp,H2 O ≈ cLv,H2 O := cL ,
and neglecting viscous effects, the energy equation reads:
 L   
∂T ∂T L 1 ∂ ∂T L
ρ L cL + uL = 2 r2 λL (13)
∂t ∂r r ∂r ∂r

It should be noted that density variations in the liquid field around the bub-
ble are assumed to be small. In fact, preliminary simulations have revealed that
the results virtually do not change when ρL is either constant, or variable e.g. ac-
cording to IAPWS-IF97 standard [61]. Therefore, we prefer the incompressible
form of the energy equations here. It should be pointed out that homobaricity
is not assumed in the liquid phase, but the incompressible energy Eq. 13 does
not include the work done by the pressure force, according to Bird et al. [62].
The integral momentum balance is considered by a modified form of the
Rayleigh-Plesset equation that takes into account the liquid compressibility as
well as the mass flux ṁ00 across the bubble wall [18]:
! !
Ṙ ṁ00 3 2 4 ṁ00 4 Ṙ
RR̈ 1 − 2 L + L L + Ṙ 1 + −
a ρ a 2 3 ρL aL 3 aL
!  
m̈00 R Ṙ ṁ00 ṁ00 ṁ00
− L 1 − 2 L + L L − L Ṙ + L
ρ a ρ a ρ 2ρ
 L 
p∞ − pw L
R d L 
+ + L L p − pLw = 0 (14)
ρL ρ a dt ∞

9
Note that although we do not spatially resolve the momentum conservation
equation, it is integrally considered by the solution of Eq. 14. In contrast to
the energy Eq. 13, we will show that compressibility may be important in Eq.
14. Compressibility is accounted for by the liquid speed of sound aL , while the
incompressible form of the Rayleigh-Plesset equation is obtained by aL → ∞.
The terms dependent on mass flux ṁ00 seem to be important to couple the local
heat and mass transport to bubble dynamics. In the result section 5.2, we will
assess the effect of compressibility as well as the inclusion of mass transfer terms
on bubble collapse and rebound. Moreover, we assume that it is important to
retain viscous effects to take the proper damping of bubble wall motion into
account. Thus, the relation between bubble pressure pG and liquid pressure at
the bubble wall pLw is given by:
  2 
L G 2σH2 O 4µL ṁ00 (ṁ00 ) ρG m,w − ρ
L
pw = p − − Ṙ − L − (15)
R R ρ ρG
m,w ρ
L

The last term on the right hand side of Eq. 15 can be neglected as we verified
in preliminary tests. It should again be pointed out that although viscous effects
are considered in Eq. 15, we assume that they can be neglected in Eq. 13 to
obtain the spatial T L field.
By the same reasoning on mass conservation as applied to the velocity at
the gas side of the bubble wall in terms of Eq. 11, the liquid side velocity reads:

ṁ00
uLw = Ṙ − (16)
ρL
The liquid velocity field is simply obtained by the mass conservation through-
out the liquid domain:  
R2 ṁ00
uL = 2 Ṙ − L (17)
r ρ
Eq. 17 corresponds to a potential flow field so that pressure may be evaluated
by Bernoulli’s equation. With Eqs. 3 (β = L), 13, 14 and 17 an equation set
for the evaluation of the liquid field variables yαL , T L and uL as well as the
time-dependent bubble radius R is available.

2.3. Boundary conditions


2.3.1. At bubble center and in the liquid far field
The bubble center is assumed to be motionless, and all variables approach
the center with a vanishing gradient. Thus, Neumann boundary conditions are
formulated for the dependent variables:

∂ρG
m ∂yαG ∂uG ∂T G
= 0, = 0, = 0, =0 (18)
∂r ∂r ∂r ∂r
At the outer liquid boundary of the computational domain, referred to as
liquid far field (index ∞), uL is readily available by the kinematic condition Eq.

10
17 and does not demand any further boundary treatment. yαL and T L remain
at their initial value that is explicated in subsection 2.4 and are formulated as
Dirichlet conditions:
L L L
yα,∞ = yα,0 ; T∞ = T0L (19)
pL∞ that appears in the Rayleigh-Plesset Eq. 14 can be understood as driving
variable of the bubble motion and is explicitly prescribed as function of time.

2.3.2. At bubble wall


On the one hand, thermal and phase equilibrium is assumed at the bubble
wall. Thermal equilibrium means that the temperature at both sides of the
bubble wall is equal:
TwG = TwL (20)
Phase equilibrium means that liquid and vapor phase have the same chemical
potential, which can be expressed in terms of fugacity [63]. Assuming that the
vapor behaves like an ideal gas and the liquid like an ideal solution, the fugacity
and activity coefficient both equal one [64], which means that the vapor is in
saturation state, i.e. pSat G
H2 O = xH2 O,w p
G
. By using the mass fraction yH G
2 O,w
G
rather than mole fraction xH2 O,w , we obtain the phase equilibrium relation:
G
Mm,w
pSat G
H2 O = yH2 O,w pG (21)
MH2 O
G G G G
Mm is the gas mixture molar weight, 1/Mm = yH 2O
/MH2 O + yAir /MAir .
We refer to Eqs. 20 and 21 as equilibrium relations in the following.
Particularly for bubble collapse, considerable non-equilibrium effects may
be expected [9, 65]. Therefore, we apply non-equilibrium relations to assess
presumed limitations of the equilibrium relations. When the gas is not in equi-
librium, a thin vapor layer termed Knudsen layer forms near the interface. A
temperature gradient develops withing the Knudsen layer, so that the temper-
ature may deviate at the borders of the layer. Since the layer is thin, TwG − TwL
is generally small, even if the gradient is large. Details are described in the
textbook of Fujikawa [66]. Thus, optionally to Eq. 20, a temperature jump
relation is employed that has been obtained from the asymptotic analysis of the
Boltzmann equation [66, 67]:
r !
G L <H2 O TwL ṁ00H2 O
Tw = Tw 1 − C1 (22)
2 pSat
H2 O T L
w

We choose the constant C1 = 0.44675 for water, according to [54, 68]. Eq.
22 has to be considered as an approximation, since it is strictly valid for mono-
atomic molecules only. Optionally to the phase equilibrium relation Eq. 21, a
non-equilibrium relation given by a Hertz-Knudsen equation [69] is employed

11
that has been used in a similar form e.g. by [18, 45, 54, 55, 70]:
s  
2 γ 
00
ṁH2 O = √ pSat G
H2 O T L − pH2 O,w (23)
L
<H2 O Tw C2 γ + 2 π (1 − γ) w

In Eq. 23, pG H2 O,w corresponds to the partial vapor pressure at the bubble
wall, i.e. pG G G G
H2 O,w = yH2 O,w <H2 O ρm,w Tw , and the constant C2 is chosen as
C2 = 2.132 according to [54]. It should be noted that the accommodation
coefficient γ may show a significant scatter for evaporation and condensation
[69] and may vary for the different condensation stages during bubble collapse
[71]. Since γ decreases with increasing liquid pressure some authors [22, 72, 73]
chose γ in a range between 0.01 and 0.1, subject to the liquid pressure for single
bubble collapse simulations. Nevertheless, we use a unique value γ = 0.4 for
simplicity, which has also been used e.g. by [54, 70]. In following sections, we
refer to Eqs. 22 and 23 as non-equilibrium relations.
Thus, to assess the limitations of the equilibrium relations in terms of Eqs.
20 and 21, we optionally may use the non-equilibrium relations Eqs. 22 and 23,
particularly for bubble collapse scenarios presented in section 5. To complete the
set of boundary conditions at the bubble wall, further relations are employed.
First, heat balance reads:

L

L ∂T ∂T G
λw G
= λm,w + LH2 O ṁ00H2 O (24)
∂r w ∂r w

In contrast to Eq. 7, the heat transport by each of the diffusing components


has been neglected in Eq. 24 because it amounts to less than 1 % of LH2 O · ṁ00H2 O ,
which we have verified in preliminary simulations. The same simplification has
been made by [45, 54]. The mixture thermal conductivity λG m is defined in
section 2.5.2. The mass balance for species α on the liquid side of the bubble
wall reads, with α = H2 O or α = Air:
 
∂y L
−ṁ00α = ρL yα,w
L
uLw − DL α − yα,w
L
Ṙ (25)
∂r w

The mass flux of component α, ṁ00α , is given in the absolute frame of reference
in Eq. 25 and has two parts. The first part is the flux relative to the bubble
interface (first and second term in brackets on the right hand side of Eq. 25) and
the second part corresponds to the movement of the bubble interface (third term
in brackets). The relative mass flux of α is the sum of convective mass flux (first
term in brackets) and binary diffusion between liquid and dissolved air (second
term in brackets). After substituting Eq. 16 into Eq. 25 and re-arranging, we

12
obtain the mass balance equations for the liquid side:

∂yHL
2O
ṁ00H2 O (1 − L
yH 2 O,w
) − L
yH 2 O,w
ṁ00Air L
=ρ D L
(26)
∂r
L
w
∂y
ṁ00Air (1 − yAir,w
L L
) − yAir,w ṁ00H2 O = ρL DL Air (27)
∂r w

Accordingly, the mass balance equations for the gas side read:

∂yHG
00 G G 00 G G 2O
ṁH2 O (1 − yH2 O,w ) − yH2 O,w ṁAir = ρm,w D (28)
∂r
G
w
∂yAir
ṁ00Air (1 − yAir,w
G G
) − yAir,w ṁ00H2 O = ρG
m,w D
G
(29)
∂r w

We assume that the gas pressure level is moderate, which may be question-
able particularly during the last stages of collapse, but enables the assumption
that dissolved air is diluted in the liquid water and that Henry’s law holds.
This means that there is a simple linear relation between the concentration of
dissolved air and its partial pressure in the gas phase [74, 75]:

pG L
Air,w = xAir,w H (30)

The Henry coefficient H depends on T G as will be explicated in section 2.5.3.


By noting pG G
Air,w = xAir,w p
G
and utilizing the mass fraction rather than mole
L G
fraction, we obtain a direct relation between yAir,w and yAir,w :

L MAir <Air G G
yAir,w = L H <
p yAir,w (31)
Mm,w m,w

L
Mm,w is the mixture molar weight at the liquid side of the bubble wall,
L L L
1/Mm,w = yH 2 O,w
/MH2 O + yAir,w /MAir . Together with the ideal gas law Eq. 5,
G G G G G L
p = ρm,w <m Tw , and the complement Eq. 1, yH 2 O,w
+ yAir,w = 1 and yH 2 O,w
+
L
yAir,w = 1 we thus have obtained a complete boundary equation set to evaluate
the dependent variables at the bubble wall whose numerical implementation will
be explicated in section 3.3.

2.4. Initial conditions


The bubble is initialized in mechanical and thermal equilibrium in terms
of R0 , T0G = T0L = T0 and pL∞,0 . yH G
2O
G
and yAir as well as yHL
2O
L
and yAir are
prescribed homogeneously, and their values are uniquely given by Henry’s law,
Eq. 31 together with Eq. 1. The bubble pressure pG 0 is provided by Eq. 15
together with pLw,0 = pL∞,0 , Ṙ0 = 0, R̈0 = 0, ṁ000 = 0 and m̈000 = 0. It should be
L
noted that by Eq. 31, yAir,0 depends on pG0 and thus on R0 , what might deviate
from real situations where water is usually air-saturated at ambient pressure.

13
After initialization, the bubble motion is driven by the temporal evolution of
pL∞ that will be described for the particular test cases in sections 4 and 5.

2.5. Thermophysical properties


2.5.1. Liquid phase
Since we assume that property changes due to dissolved air can be neglected,
mixture properties are evaluated only for pure liquid water. ρL , λL , µL , aL
and cL are estimated by the IAPWS-IF97 standard [61] and are dependent on
temperature and pressure, i.e. T L and pL∞ . The binary diffusion coefficient
DL is evaluated based on the assumption of infinite dilution of air in water,
according to the Wilke-Chang estimation [76]:
p
L 7, 4 × 10−15 φMH2 O T L
D = (32)
µL (VAir 103 )0,6

with the association factor for the solvent, i.e. water, φ = 2.6 and the molar
volume of the solute, i.e. air at its normal boiling point, VAir = 0.03 m3 /kmol.

2.5.2. Gaseous phase


In contrast to the liquid where except from DL mixture properties are well
represented by the pure liquid water properties, in the bubble, a gas mixture of
water vapor and air needs to be considered. As already noted in section 2.2.1,
mixture internal energy eG G G
m , enthalpy hm as well as specific heat capacities cp,m
and cGv,m are linearly weighted by mass fraction and the corresponding vapor
and air values. For the evaluation of cG G
p,m and cv,m , the corresponding vapor
(α = H2 O) and air (α = Air) values are assumed to be constant for simplicity
and are obtained by:
κα <α
cG
p,α = (33)
(κα − 1)
with <α = <Univ /Mα , MH2 O = 18.015 and MAir = 28.97. κH2 O is evaluated
by the IAPWS-IF97 standard [61] at T G = 400 K and pG = 1.013 bar, and
κAir is evaluated by the NIST database [77] at the same thermodynamic state.
Accordingly:

cG G
v,α = cp,α − <α = (34)
(κα − 1)
The mixture gas constant <m is also evaluated by a linear weighted mean of
specific gas constants <α :
G G
< m = yH2O
<H2 O + yAir <Air (35)

Common approximations of λG m and D


G
are taken from [76] and briefly
G
summarized here. λm is evaluated by the Wassiljewa Equation [78]:

xG G
H2 O λH2 O xG G
Air λAir
λG
m = + G (36)
xG
H2 O
G
+ xAir AH2 O→Air G
xAir + xH2 O AAir→H2 O

14
It is worth noting that the mole fraction can readily be transformed to mass
fraction by xG G G
α = yα Mm /Mα . The coefficients AH2 O→Air and AAir→H2 O are
given by the Mason-Saxena modification [79], and e.g. AH2 O→Air reads:
h i2
1/2 1/4
1 + (λtr,H2 O /λtr,Air ) (MH2 O /MAir )
AH2 O→Air = 1/2
(37)
[8 (1 + MH2 O /MAir )]

Accordingly, AAir→H2 O is obtained by changing the subscripts H2 O and Air.


λtr,H2 O and λtr,Air are the monoatomic values of the thermal conductivity and
are estimated using critical properties [76]. λG
H2 O is evaluated depending on T
G
G G
and p by the IAPWS-IF97 standard [61], and λAir is obtained from [80], as
function of T G at pG = 1.013 bar.
The binary diffusion coefficient DG is obtained by the approximation given
by Fuller et al. [81], adjusted to SI units:
1.75 1/2
G 1.01116 · 10−2 T G [1/MH2 O + 1/MAir ]
D =  2 (38)
pG (ΣVH2 O )1/3 + (ΣVAir )1/3

where ΣVα is the atomic volume that is provided in [81] for water and air.

2.5.3. Bubble wall


Surface tension σH2 O , latent heat LH2 O and vapor saturation pressure pSat
H2 O
are obtained by the IAPWS-IF97 standard [61] in dependence on T G . We
estimate the Henry coefficient H of air in water by a simple mass fraction
weight of Oxygen and Nitrogen:

H = 0.24HO2 + 0.76HN2 (39)

HO2 and HN2 are approximated according to [74]:

1 ρL
Hα = cp (40)
Hα MH2 O

where α = O2 or α = N2 . Hαcp is the Henry solubility defined via concentration


[74] and depends on temperature:
  
cp cp d ln H 1 1
Hα = Hα,ref exp − (41)
d(1/T ) α TwL Tref
cp
Hα,ref is the Henry solubility at reference temperature Tref = 298.15 K,

d ln H
and corresponds to a slope of Hαcp versus temperature. The cor-
d(1/T ) α 
cp cp
responding values are [74]: HO = 1.3 · 10−5 mol/ m3 Pa , HN =
2 ,ref 2 ,ref
 d ln H d ln H
6.5 · 10−6 mol/ m3 Pa , = 1500 K and = 1300 K.
d(1/T ) O2 d(1/T ) N2

15
3. Simulation method

The governing equations constitute a set of coupled non-linear partial differ-


ential equations. In most studies, limited information on the numerical scheme
and solution procedure has been provided. Thus, we present our simulation
method in some detail.

3.1. Spatial discretization


A schematic of the discrete computational domain is illustrated in Figure
1. The bubble wall is located at r/R = 1. In preliminary simulations, an
extent of the liquid domain up to r/R = 3 has been found to be adequate. The
computational domain is discretized in radial direction by dimensionless grid
node locations r̂i = ri /R. The number of nodes on the gas and liquid part
of the computational domain is labeled nG and nL , respectively. We utilize a
staggered variable arrangement, i.e. the velocity is stored at node locations,
and all other variables are stored at cell center locations, where a cell is defined
to be within two neighbour nodes, and the cell center location is evaluated
as the arithmetic mean of neighbour node locations as labeled ĉ in Figure 1.
The staggered variable arrangement allows us to discretize the mass fraction
equations 3 by a finite volume scheme in a straightforward way as explained
below and thus mass conservation is ensured. Ghost cells are introduced to
store the boundary values at bubble center, bubble interface and in the liquid
far field. Note that in Figure 1, a homogeneous node distribution is shown
for clearness of the illustration, and for the simulation runs, a node clustering
is applied in the proximity of the bubble wall to adequately resolve spatial
gradients.

Bubble center Bubble wall (w) Liquid far field


Ghost Ghost
cell cell
Bubble domain
G
^c1 ^c2
G
^c3G ^cnG -1
G ^cnG ^cnG +1
G G

^rG0 ^rG2 ^rG3 ^rnG -2 ^


G rnGG-1 ^rnG +1G

^rnGG = 1
^rG1 = 0
Ghost Ghost
cell
Liquid domain cell
L
^c1 ^c2
L
^c3L cnLL-1
^ cnLL
^ ^cnL +1
L

^rL0 ^rL2 ^rL3 L


L
^rnL -2 ^rn -1 L ^rnL +1
L

rL1
^ =1 rnLL
^ =3

Figure 1: Discretized computational domain.

The bubble wall moves with the velocity Ṙ. Thus, we solve the governing
equations on a moving grid with the discrete grid velocity ugrid i = Ṙr̂i that
takes the value zero at the bubble center and Ṙ at the bubble wall. We use
an ALE method, which means that in the convective terms of any equation
the absolute velocity ui is substituted by the relative velocity ũi = ui − ugrid i
according to e.g. [82]. It is noteworthy that by the particular choice of ugrid
i , the

16
dimensionless grid node locations r̂i are stationary everywhere in the domain.
It should further be pointed out that the space conservation law [83] is fulfilled
as can be verified by a short calculation by paper and pencil, so that in the finite
volume framework of the mass fraction equations, mass conservation is fulfilled
even on the moving grid.
The mass fraction equations 3 are multiplied by r2 and integrated over the
computational cell as is illustrated as an example for the gas phase β = G:
Z riG
 Z riG  
2∂ ρG G
m yα ∂ 2 G G G
G
2 G G ∂yα
r dr = − r ρm yα ũ − r ρm D dr (42)
G
ri−1 ∂t G
ri−1 ∂r ∂r

The time derivative on the left hand side is assumed to be constant over the
computational cell, and after differentiation and integrating the right hand side
we obtain the semi-discrete form of equation 3 where the time discretization
will be discussed in section 3.2:
3 3 " #
riG − ri−1G
∂ ρG m yα
G

3 ∂t
j
  G 
G
 2 G G G G ∂yα
= ri ρm,i −yα,i ũi + Di
∂r i
"  G #
G
2 G G G G ∂yα
− ri−1 ρm,i−1 −yα,i−1 ũi−1 + Di−1 (43)
∂r i−1

Subscript j means  a cell center and i means a node value. While the time
derivative ∂ ρG G G
m yα /∂t and the velocity ũi are readily available at j and i,
respectively, ρG G G
m,i , yα,i and Di are obtained from cell locations j and j + 1 by
linear interpolation, corresponding to a central difference scheme, and the same
holds for node values at i − 1. The spatial derivativeis approximated  by central
differences as well, i.e. ∂yαG /∂r i ≈ 2 yα,j+1
G G
− yα,j G
/ ri+1 G
− ri−1 . The same
procedure is applied to the liquid mass fraction equation 3, for β = L.
Since we utilize the mass fraction equations in their conservative form, the
integration over the cell length is straightforward, and the finite volume dis-
cretization scheme is an obvious choice. For the other equations we prefer a
finite difference scheme. Here, we only present the semi-discrete form of equa-
tion 13, but the fully-discrete form of this and all other equations can be found
in [60].
 L  L
   L

∂T L ∂T 1 1 ∂ 2 L ∂T
= − ũ + L L 2 r λ (44)
∂t j ∂r j ρ c r ∂r ∂r j

For the interpolation from nodes i to cell centers j and for gradient ap-
proximations, again central differences are utilized. Finally, it should be noted
that all integration operations e.g. in equations 8 to 10 are performed by the
midpoint rule.

17
Summarizing, by a combined central finite volume / finite difference scheme
we have obtained a simple and robust second order accurate scheme which is
fully-conservative regarding mass fluxes. It is nevertheless important to evalu-
ate the magnitude of the Peclet number Pe in accordance to the local spatial
resolution to assess the stability of the central scheme. The Peclet number de-
scribes the ratio of convective to diffusive flux and is evaluated for the gaseous
and liquid phase in terms of Peβ :
 
ρβm ũβ riβ − ri−1
β

Peβ = (45)
µβm
where β = G or β = L. At high Peclet number, i.e. larger than the order of
one, the numerical solution may show artificial wiggles particularly for central
differencing schemes. Since ρLm = ρL is usually much larger than ρG m , keeping
PeL small is significantly more demanding than PeG . Moreover, due to the
particular grid kinematics, the highest values of the grid velocity and also of
the relative velocity ũL are present in the liquid far field. At the same time,
cells are allowed to become rather large at this location since the gradients of
the flow variable tend to zero at large distances from the bubble wall. In fact,
grid nodes are clustered preferably at the bubble wall to adequately resolve the
flow gradients. Node clustering is considerably more pronounced at the liquid
side due to much steeper gradients than within the bubble. As a consequence,
we cope with the fact that PeL reaches its largest values towards the liquid far
field. Summarizing, for any simulation result presented in section 4 and 5, PeG
amounts to less than 0.01 at any location within the bubble and at any time
instant. Regarding the liquid side, due to a very fine grid, PeL is lower than
10−5 close to the bubble wall, and may grow up to about 102 towards the liquid
far field for some cases and time instants. We consider this magnitude to be
still moderate, and by a careful inspection of the numerical solution field, we
have ensured that in fact no wiggles are present in any of the simulation results
presented in section 4 and 5.

3.2. Temporal discretization


The time integration is performed by a Crank-Nicolson scheme which is easy
to implement and accurate to second order [84]. For any dependent variable Φ
and a general partial differential equation ∂Φ/∂t = F Φ, r, t, ∂Φ/∂r, ∂ 2 Φ/∂r2 ,
it reads e.g. for cell j:
   
Φn+1
j − Φnj ∂Φ ∂ 2 Φ ∂Φ ∂ 2 Φ
= (1 − cnf) Fjn+1 Φ, r, t, , 2 + (cnf)Fjn Φ, r, t, , 2
∆t ∂r ∂r ∂r ∂r
(46)
with cnf = 1/2. The implicit part, i.e. the first term on the right hand side of
equation 46 demands an iterative procedure. To get an appropriate criterion to
assess convergence of the implicit loop and at the same time to avoid an effect of
round-off errors, we solve the non-dimensional form of the governing equations

18
by the choice of adequate reference values. Thus, the non-dimensional residua
of each equation are evaluated, and the implicit part of the solution is found to
be sufficiently converged when each residuum has dropped below 1 · 10−8 which
is a quite conservative criterion. The linear equation system is sparse due to the
compact central scheme stencil and the 1D problem and can easily be solved
by a simple Thomas algorithm. The time step that we utilize corresponds to a
maximum acoustic CFL number (evaluated with the liquid speed of sound) of
about 100.
For some cases, particularly for small bubbles and in the state immediately
before bubble collapse, extremely small time steps corresponding to maximum
CFL numbers in the order of 1 need to be employed to resolve the large temporal
gradients and fulfil the convergence criterion of the Crank-Nicolson scheme.
Thus, in some cases we alternatively employ an optimized four stage low storage
explicit RungeKutta time scheme [85] in combination with a CFL number of
about 1 which is accurate to third order.

3.3. Boundary conditions at the bubble wall


At the bubble wall (location w in Figure 1), 9 unknown boundary quantities
in terms of TwG , TwL , yH
L
2 O,w
L
, yAir,w G
, yH2 O,w
G
, yAir,w , ṁ00H2 O , ṁ00Air and ρG
m,w need to
be evaluated. Heat balance Eq. 24, mass balance Eqs. 26-29, Henry’s law Eq.
31 and the ideal gas law Eq. 5 together with Eqs. 20 and 21 are solved in case
of thermal and phase equilibrium. In the case of non-equilibrium, Eqs. 20 and
21 are substituted by Eqs. 22 and 23. These equations constitute a nonlinear
equation set that is solved by a Newton method for the unknown variables.
In the 9 boundary governing equations, gradients are discretized by a central
scheme where the variables in the cell centers adjacent to the bubble wall are
temporally frozen during the solution process. Once the 9 unknown boundary
quantities have been evaluated, an update of the ghost cell center values at
gas and liquid side is performed by a linear extrapolation from the respective
side, employing the just evaluated boundary values at location w. Subsequently,
the next iteration or time step for the Crank-Nicolson of Runge-Kutta scheme,
respectively, is performed.
Preliminary simulations have revealed that because of the small time steps
in the temporal solution procedure, only one Newton step needs to be performed
which is equivalent to a linearization of bubble wall boundary conditions. Thus,
an accurate and robust scheme has been obtained for the solution of the nonlin-
ear 9 equation set at the bubble wall that might easily be extended by further
equations in future studies e.g. on multi-component fluid mixtures.

3.4. Solution procedure


The solution procedure is for example presented for the Crank-Nicolson
scheme. In the first sweep, the explicit part in terms of the second term on
the right-hand-side of Eq. 46, and in a second sweep, the first term on the
right-hand-side of Eq. 46 is solved by an implicit loop. The solution scheme
is depicted in Figure 2. It is noteworthy that we consequently use a robust

19
summation method by Kahan [86] to avoid rounding errors during repeating
summation operations.
Initialize flow variables and grid

Solve explicit part, equations are analogous to implicit part


.
Solve R(t) and R(t) by Eq. 14
Solve pG(t) by Eq. 10
Evaluate bubble wall boundary conditions according to sec. 3.3
Implicit Loop

Solve T G(r,t) by Eq. 12 and T L (r,t) by Eq. 13


Time Loop

G
Solve ½m (r,t) by Eq. 5
G
Solve yAir (r,t) by Eq. 3 and yHGO(r,t) by Eq. 1
2

L
Solve yAir (r,t) by Eq. 3 and yHL O(r,t) by Eq. 1
2

Solve uG(r,t) by Eq. 8


Solve uL(r,t) by Eq. 17

If converged, proceed to next time step


Write results to file

Figure 2: Flow chart of sibub solution procedure for the Crank-Nicolson scheme.

The implementation of the Runge-Kutta scheme is straightforward by setting


cnf = 1 in Eq. 46 and executing the loop 4 times. The computer program sibub
(= si ngle bubble dynamics) is coded in Fortran 90 and runs on any conventional
Linux workstation.

3.5. Conservation properties


We verify mass conservation by the evaluation of the bubble mass in two
ways: On the one hand, a spatial integration is applied at any evaluation time
instant teval :
Z G
4 X G h G 3 3 i
r=R n
mα,Spat = ρG
α dV ≈ π ρα,j ri+1 − riG (47)
r=0 3 i=1

with subscript α = H2 O or α = Air. On the other hand, the bubble mass is


tracked over time by a temporal integration of the mass flux into the bubble:
Z teval it
XEnd

mα,Temp = mα,0 + ṁ00α 4πR2 dt ≈ mα,0 + 4π ṁ00α R2 ∆t (48)


t=0 it=0

Time steps ∆t are small, so that the numerical integration of Eq. 48 is


performed by a simple first order explicit Euler scheme during simulation run-

20
time, which virtually corresponds to a summation from the initial time step
it = 0 to the current one, it = itEnd . It is particularly important that a robust
summation [86] is applied to avoid the amplification of rounding errors. By
monitoring mα,Spat and mα,Temp during each simulation run, we found that
their difference is not larger than about 1% for bubbles with R0 = 1000 µm and
decreases to below 0.01% for bubbles with R0 = 10 µm. Taking into account
some inaccuracies during numerical integration, this difference is considered
small, and it can be assumed mα,Spat = mα,Temp = mα .

4. Model validation on bubble growth

In preliminary simulations, the mathematical model and its implementation


have been tested on a multitude of simple examples to verify its plausibility.
For example, air diffusion on a quiescent bubble in saturated (nothing happens
but numerical noise), undersaturated (slow bubble shrinkage) and oversaturated
(slow bubble growth) water has been investigated. Further, the implementation
of the Rayleigh-Plesset Eq. 14 has been verified by comparing the results with
a different computer program solving the Rayleigh-Plesset equation. Here, we
present two realistic validation test cases where measurement data is available.

4.1. Bubble growth in superheated liquid


While the early stage of bubble growth is limited primarily by momentum
interaction between the liquid and the bubble, as the bubble grows the thermal
diffusion influence becomes more important until it dominates the growth rate
which is then bounded by the heat diffusion in the liquid [87, 88] and may be
characterized by a thermodynamic parameter [89]. For this thermally controlled
stage of bubble growth, by assuming of a thin thermal boundary layer in the
liquid surrounding the bubble, an approximate solution for the bubble growth
was obtained by Plesset and Zwick [87]. This solution was shown to agree very
well with the experimental results provided by Dergarabedian [90] for water with
low superheats at atmospheric pressure. The solution of the Rayleigh-Plesset
equation [8, 91] by the use of the Plesset-Zwick [87] approximation method
shows that a first critical time tc1 exists, above which the thermal term starts
to dominate the solution (thermally controlled region), and below that, the
Rayleigh-Plesset equation may be approximated by the linear Rayleigh equation
(inertially controlled region). tc1 will be discussed in more detail in section 4.2,
while for the test case considered here, we content ourselves with the result that
tc1 is very small and the bubble growth is thermally controlled.
Measurement data on bubble growth in superheated liquid was provided by
Dergarabedian [90]. In our previous study [92] we presented a simplified math-
ematical single-bubble model where heat transfer was assumed to be the only
driving mechanism for bubble growth. Here, initial and boundary conditions are
adopted from [92] for our sibub simulations and summarized in Table 1. It should
be pointed out that T0L , ∆TSH and R0 correspond immediately to the experi-
mental boundary conditions, and no input parameter fitting has been done in

21
the simulations. Since in sibub, we prefer to input a driving pressure rather than
a driving temperature, the superheat levels ∆TSH are reformulated to pressure
differences in terms of the tension ∆pTension by means of the Clausius-Clapeyron
relation, i.e. the temperature dependence of the vapor saturation pressure. In
the sibub simulations, the liquid pressure drops from its initial value pSat L
L
H2 O TL0
Sat
instantaneously to its final value p∞ , where the pressure difference pH2 O T L −p∞
0
L
corresponds to the nominally prescribed superheat level
∆TSH . p∞ is the atmo-
Sat
spheric pressure with a saturation temperature TH2 O pL = 373.15 K.

Table 1: Initial and boundary conditions for bubble growth in superheated water, adopted
from [92]
Symbol Unit Case 1 Case 2 Case 3 Case 4 Case 5
T0L K 374.6 375.3 376.3 377.7 378.5
∆TSH K 1.4 2.1 3.1 4.5 5.3
∆pTension Pa 5, 963 8, 901 13, 099 18, 975 22, 333
R0 µm 22.0 15.6 10.2 7.5 5.2

It is interesting to note that according to [93], the bubble evolution may be


dependent on the drop rate ∆pL∞ /∆tdrop . In preliminary simulations we have
in fact found that for small drop rates in the range of 104 Pa/s, the bubble
growth is affected by the drop rate, and for a drop rate > 106 Pa/s, this effect
vanished. The results that are presented below have been obtained by a drop
rate of 1010 Pa/s and are thus considered drop rate independent.
Results are obtained on a grid with nG = 100 and nL = 200 and a strong
node clustering towards the bubble wall. The non-equilibrium boundary condi-
tions according to Eqs. 22 and 23 are employed. In preliminary simulations with
a variation of the spatial resolution, grid independence of the results has been
assured and – as discussed in section 3.1 – the results are free from artificial wig-
gles due to sufficiently low Peclet numbers. In Figure 3, sibub simulation results
in terms of Ṙ are presented. Measurement data has been reprinted from [90].
In the experiment [90], the time uncertainty when the bubble starts to grow
amounts to about 1 ms, and the data shows a considerable scatter. In fact,
Dergarabedian [90] compared his measurement data with the bubble growth
approximation by Plesset and Zwick [87] theory, which we also adopt in this
study. Therefore, an approximate solution for bubble evolution based on [87] is
evaluated: r
∆TSH ρL cL 3 αL
Ṙ = (49)
LH2 O ρH2 O πt
 
with the vapor density ρH2 O = pSat L
H2 O T L / <H2 O T0 and the thermal diffusivity
  0

αL = λL / ρL cL . All properties in Eq. 49 are evaluated at T0L . Ṙ is plotted


in Figure 3, termed Plesset-Zwick. It should be pointed out that Eq. 49 only
holds for thermally controlled bubble growth, and is thus tailored for this test
case. In fact, we concluded that the sibub results are in good proximity to the
Plesset-Zwick solution. It should be noted that one main assumption leading to

22
Eq. 49 is that the thermal boundary layer is thin compared to the bubble size,
which is only approximately the case here and might be the reason for remain-
ing differences between the sibub and Plesset-Zwick solution. These remaining
differences are well within the scatter of measurement data, that lets us assume
the validity of sibub for thermally controlled bubble growth.

a) Case 1 b) Case 2 c) Case 3 d) Case 4 e) Case 5


0.20 ∆T = 1.4 K ∆T = 2.1 K ∆T = 3.1 K ∆T = 4.5 K ∆T = 5.3 K
Experiment, reprinted from [90]
sibub, simulation results
Plesset − Zwick, according to Eq. 49
Ṙ [m/s]

0.10

0.00
0.000 0.007 0.015 0.000 0.007 0.015 0.000 0.007 0.015 0.000 0.007 0.015 0.000 0.007 0.015
Time [s]

Figure 3: Ṙ vs. time for different superheat levels.

4.2. Inertia and heat-controlled bubble growth


The next example is adopted from [54] where single bubble experiments have
been performed. A single air bubble has been injected into a quiescent water
test tube. By a pressure reduction, bubble growth has been provoked, and by a
subsequent shock wave, bubble collapse has been initialized. In Figure 4a, the
measured test tube pressure and bubble radius R are depicted together with
our sibub simulation results. The same computational grid as before is utilized.
The measured test tube pressure is adopted as excitation pressure pL∞ for the
sibub simulations. The collapse occurs at t ≈ 22 ms. The simulation terminates
after the initial stage of collapse due to rather large time steps that corresponds
to CFL ∼ 100. We accept this termination during collapse here, since bubble
collapse will be discussed in detail in section 5.
Firstly, the simulation results have been obtained by the non-equilibrium
bubble wall conditions according to Eqs. 22 und 23 and show an excellent
agreement to measured R in the early bubble growth phase. Note that both
measured pL∞ and R have been reasonably well reproduced, with respect to
the published plots in [54]. This procedure contains some inaccuracies, which
may be associated with a potential overestimation of R for the late phase of
bubble growth. In Figure 4b, different pressure types at the bubble wall are
illustrated. After the excitation pressure pL∞ and the bubble wall pressure pLw
have dropped below the saturation pressure pSat
H2 O T L at t ≈ 10 ms, the bubble
w
L
starts to grow abruptly. In the measurement, p∞ has shown some oscillations
Sat
and at some instants protrudes above pH2 O T L , that can be seen in Figure 4b at
w
t ≈ 13 ms. It is interesting to note that due to the non-equilibrium assumption
at the bubble wall, the partial pressure of vapor in terms of pG H2 O does not

23
a) p∞L and bubble radius b) Pressure at bubble wall
×105 ×103
Close-up A
1.2 4.0 A 4 3.30

2.8
20 23
p∞L [Pa]

R [mm]
p [Pa]
Exp. [54]
0.6 non-equil. 2 3.15
equil. pwL
pHSat2 O | TwL
pHG2 O

2600 Pa
2900 Pa
p∞L
0.0 0 3.00 pG
0.0 12.5 25.0 0.0 12.5 25.0
Time [ms] Time [ms]
c) Bubble mass d) Bubble wall temperature
105
Close-up B
120 298.8
104
mα /mH2 O, 0 [−]

40
103 0 25
T [K]

298.2
102 B

101
mAir /mH2 O, 0 TwL
100 mH2 O /mH2 O, 0 297.6 TwG
0.0 12.5 25.0 0.0 12.5 25.0
Time [ms] Time [ms]

Figure 4: Temporal progression of a) excitation pressure pL ∞ and bubble radius R; b) different


pressure types at bubble wall; c) bubble mass and d) temperature at gas and liquid side of
bubble wall. If not otherwise stated, results have been obtained by non-equilibrium bubble wall
boundary condition. Local minima of pL ∞ in b) are marked by downward arrows. Measurement
data in terms of pL∞ and R have been reprinted from [54].


exactly correspond to the saturation pressure in terms of pSat L which would
H2 O Tw
be the case for equilibrium condition.
By the pressure progression, we can estimate the tension and discuss the bub-
ble growth regime in terms of the first critical time tc1 in more detail. According
to [89], tc1 is evaluated by tc1 = ∆pTension / ρL Σ2 with the thermodynamic pa-
2
L2 [pSat /(T L <H O )]
rameter Σ = H2 O LH√2 O L 30 L L2 . Σ is evaluated at temperature T0L , and for
T0 (ρ ) c λ
the tension, according to Figure 4b, we estimate ∆pTension = pSat L
H2 O T L − p∞ ≈
w
200 Pa. Thus, we obtain tc1 ≈ 4 ms which means that both, inertia and heat
transfer are important for bubble growth in the time range under investigation.

24
In Figure 4c, the vapor (mH2 O ) and air (mAir ) mass within the bubble are
illustrated, both related to the initial vapor mass mH2 O,0 . Albeit at the initial
state, the bubble contains about 60 times more air than vapor, at the largest
radius at t ≈ 21 ms, vapor has overshot air by almost three orders of magnitude
as a consequence of cavitation. Note the logarithmic scale of Figure 4c. Close-
up B in Figure 4c shows that air mass is increased by only about factor 2.5
during bubble growth as a consequence of the relatively slow diffusion process
of dissolved air in the liquid that limits the amount of air that is released into
the bubble. Vapor mass, on the other hand, increases in the same time interval
by more than four orders of magnitude, which illustrates that in fact cavitation
is accompanied by a phase change in terms of evaporation, while air content
within the bubble is only marginally changed.
In Figure 4d, the time-progression of the temperature on both sides of the
bubble wall is depicted and clearly demonstrates the temperature jump over the
bubble wall as a consequence of the non-equilibrium boundary conditions. As
discussed in section 2.3.2, this jump is small.
So far, simulation results obtained by the non-equilibrium boundary condi-
tions have been discussed. We have performed comparative simulations with
the equilibrium condition as well. The time-progression of R obtained by the
equilibrium condition has also been depicted in Figure 4a. It is interesting to
note that its maximum only deviates by about 1% from the non-equilibrium
counterpart, as is illustrated by Close-up A in Figure 4a. A more thorough
illustration of the differences between non-equilibrium and equilibrium condi-
tion is provided by Figure 5 where the spatial distribution of T G and T L at
selected time instants in the proximity of the bubble wall is shown. For the
non-equilibrium condition, the temperature jump over the bubble wall is again
clearly discernible, while the spatial progression of temperature is continuous
over the bubble wall for the equilibrium case. It is interesting to note that the
temperature distribution of T L is hardly affected by the interface condition,
which we trace back to the far field boundary condition where T L is fixed to
L
T∞ for both cases. Distinctive differences are rather reflected on the gas side by
T G that is by the temperature jump smaller for the non-equilibrium case than
for the equilibrium case.
So far, for the non-equilibrium results, an accommodation coefficient γ = 0.4
has been employed in Eq. 23. According to [89], the difference between non-
equilibrium and equilibrium boundary condition is small when γ is close to
one. To this effect, we assume that γ = 0.4 is in fact close to one. Therefore,
we repeated the simulations with γ = 0.01 which can be considered to be a
lower border of values used in the literature as discussed in section 2.3.2. The
results are not shown here in detail but only briefly summarized: For this sig-
nificantly reduced value of γ, the bubble radius peak in Figure 4a is halved and
thus considerably underpredicted. The pressure difference peak in Figure 4b,
pSat G G
H2 O T L − pH2 O,w increases considerably, essentially due to a drop of pH2 O,w .
w
Interestingly, the temperature difference TwL − TwG in Figs. 4d and Figure 5
only moderately changes. In the end, due to the significant underprediction

25
298.8
T0
TG
10 ms TL
T [K]
298.2

Bubble Wall
20 ms non-equil.
297.6 equil.

0.96 1.00 1.04


r/R [−]

Figure 5: Spatial distribution of field temperature within bubble (T G ) and liquid (T L ) at


selected time instants for both, non-equilibrium and equilibrium bubble wall boundary con-
dition.

of R with γ = 0.01, we conclude that γ = 0.4 is more appropriate for bubble


growth. It should again been pointed out that γ may show a significant scatter
for evaporation and condensation [69], as discussed in section 2.3.2. A system-
atic assessment of the γ effect is considered to be out of the scope of this study
and is left for future studies. In what follows, we retain γ = 0.4 also for bubble
collapse that is discussed in section 5.
It should be pointed out that Kawashima and Kameda [54] also presented
some simulation results with presumably the same mathematical model that
we use and non-equilibrium boundary conditions with γ = 0.4. We assume
differences in the property approximations and maybe more significantly, in
the numerical implementation that is only briefly summarized by [54]. We
see a considerable overestimation of R in our simulation results according to
Figure 4a, that is not discernible in the results presented in [54] and that -as
already discussed- may be associated with some inaccuracies of the measured pL∞
reproduction that servers as our simulation input. All further results beyond
the R progression are consistent with the ones documented in [54], so that
we can conclude that our model formulation and particularly our numerical
implementation are valid.
It is worth mentioning here that liquids may experience negative pressure
[94] until a rupture of seeds, in terms of small bubbles, may invoke cavitation
that raises the pressure to positive values. In preliminary simulations, we have
inputted an exaggerated drop of pL∞ , including tension in terms of negative
pressure peaks down to −1, 000 bar. It is interesting to note that albeit the
drop of pLw and thus the level of Ṙ correlates with the intensity of the tension,
pLw throughout remains positive.

26
5. Detailed transport processes at bubble collapse and rebound

5.1. Case description


In section 4, we have demonstrated the validity of the mathematical model by
means of measurement data, where we have restricted our validation to bubble
growth so far.
Bubble collapse is more challenging than bubble growth due to the rapid
temporal change of flow variables and the extreme thermodynamic conditions
in terms of high bubble pressure and temperature. Thus, for the collapse phase
of bubble dynamics, it is interesting to discuss the capabilities and limitations
of the inherent omission of the momentum equation in our mathematical model,
as well as further simplifications, such as the assumption of an ideal gas and the
validity of Henry’s Law. As a reference, the simulation results from the study by
Matsumoto and Takemura [45] are consulted where a full Navier-Stokes model
was utilized. If not otherwise mentioned, we utilize the non-equilibrium bubble
interface condition in terms of Eqs. 22 and 23 since most studies emphasize
the non-equilibrium character of the collapse by the use of a Hertz-Knudsen
relation e.g. [45, 54, 55]. We nevertheless aim at an assessment of the effect
of the equilibrium bubble interface condition in terms of Eqs. 20 and 21. It
should be noted that [45] already provided a detailed discussion of the transport
processes at single bubbles of different sizes. In our study, we supplement this
discussion with additional details e.g. on heat transfer results as well as with
an assessment of the homobaricity and equilibrium assumptions.
We adopt the test case by [45] where pL∞ is specified by a pressure jump
from 104 to 105 Pa to initiate the bubble collapse. A broad range of R0 , from
10 µm to 1000 µm is covered. The initial temperature equals T0 = 293 K. We
performed a grid study with up to nG = 400 and nL = 800. As for the test cases
presented in section 4, a strong clustering of grid nodes in the proximity of the
bubble wall is applied. Grid independence is assumed when no visual difference
to the next finer grid results is discernible in any result plot. It has been found
that as for the test cases presented in section 4, nG = 100 and nL = 200 yield
grid independent results for most stages of bubble dynamics. In order to resolve
particularly the strong gradients during collapse stage, we choose nevertheless
nG = 200 and nL = 400 for any results presented in the following, and can
for sure assume that the results are grid independent. For R0 = 1000 µm, it
was found by a preliminary time step study, that a time step corresponding to
a CFL number of up to CFL ≈ 300 together with the implicit Crank-Nicolson
scheme, is appropriate to obtain a sufficient resolution of any collapse stage. For
smaller radii, the implicit scheme fails to converge even with a significant time
step reduction. Thus, CFL ≈ 0.5 is applied for R0 = 100 µm and R0 = 10 µm
together with the explicit Runge-Kutta scheme.

5.2. Preliminary assessment of the mathematical model


G
In Figure 6, the temporal
q progression of R and p is depicted. Time is non-
dimensionalized by τ = R0 ρref /pref , with ρref = 997 kg/m3 and pLref = 105 Pa,
L L L

27
where the latter corresponds to the liquid pressure pL∞ after its abrupt rise.
In Figure 6a and b, results obtained by the full equation set that has been
solved to obtain the results presented so far in section 4 are depicted, named
full transport model in what follows. For the results depicted in Figure 6c
and d, any transport processes are omitted, what that means is that none of
the partial differential equations that were presented in section 2 are solved,
and the solution is restricted to the solution of the Rayleigh-Plesset Eq. 14.
Isothermal bubble content is assumed, which is equivalent to an infinitely fast
heat transfer, and this simplified model is named isothermal model. For Gthe
isothermal model, the bubble pressure is evaluated by pG = pSat L +p ,
H2 O T Air
0
3
where pG G
Air = pAir,0 ·(R0 /R) . For both, full transport and isothermal model, the
bubble performs an oscillating motion with an alternating collapse and rebound
phase, and its radius attenuates due to damping processes. For the isothermal
model, viscous damping due to the third term on the right-hand side of Eq.
15 is the only damping mechanism, while for the full transport model thermal
damping also contributes to the abatement. As a consequence, the bubble
radius attenuates faster, and pressure peaks during collapse are lower for the
full transport model. It is particularly interesting, that for the isothermal model,
the oscillation period continuously raises with decreasing value of R0 , while for
the full transport model, it takes a minimum at R0 = 100 µm. This non-linearity
will be investigated in detail further below. For the time being, we note the
considerable effect of transport processes on alternating bubble collapse and
rebound. In what follows, we proceed with the full transport model.
In Figure 7, for example for R0 = 1000 µm, the effect of neglecting com-
pressibility (named Incompressible RP-eq.) or mass transfer (named Omission
of mass flux ) in the Rayleigh-Plesset Eq. 14 is considered, by either setting
aL → ∞ or ṁ00 = 0. The case Compressible RP-eq. corresponds to the full
transport model results already discussed by Figure 6a and b. Neglecting com-
pressibility shows a significant effect from the first rebound on, and is associated
with a lower radius and peak pressure attenuation as well as with a longer os-
cillation period. Neglecting the mass transfer shows the same trend, albeit less
pronounced. It should be pointed out that it is essentially the omission of the
vapor mass flux by setting ṁ00H2 O = 0 that affects the bubble dynamics, while
omitting ṁ00Air has virtually no effect. This is in line with the fact that the
initial air content of the bubble exceeds its initial vapor content by one order of
magnitude, and moreover, ṁ00H2 O is one order of magnitude larger than ṁ00Air as
will be discussed further below. Thus, a large amount of air is virtually trapped
in the bubble. It is of utmost importance to note that we have implemented the
condition ṁ00 = 0 by substituting the interface boundary conditions in terms of
Eq. 23 and Eq. 27 by the interface boundary conditions ṁ00H2 O = 0 as well as
ṁ00Air = 0, this means that mass transfer is omitted not only in the Rayleigh-
Plesset Eq. 14, but also in any transport equation, for example in Eqs. 16,
17, 22, 24 and 29. It is interesting to note that if we set ṁ00 = 0 only in the
Rayleigh-Plesset Eq. 14 and 15 but maintain it in all other equations, virtually
no effect on bubble dynamics can be observed. It can be concluded that mass

28
a) Bubble radius - Full transport b) Bubble pressure - Full transport
108
1.0 R0 = 1000 µm
107 R0 = 100 µm
R0 = 10 µm
R/R0 [−]

p G [Pa]
0.5 106

105
0.0
104
c) Bubble radius - Isothermal d) Bubble pressure - Isothermal
108
1.0
107
R/R0 [−]

p G [Pa]

0.5 106

105
0.0
104
0 5 10 0 5 10
t/τ [−] t/τ [−]

Figure 6: Temporal progression of bubble radius (a,c) and bubble pressure (b,d), with the
consideration of all transport processes (a,b) or the omission of any transport processes and
assumption of isothermal bubble content (c,d).

transfer affects the motion of the bubble wall rather by the changed bubble
content in terms of pG than by the terms including ṁ00 in Eqs. 14 and 15, and
that these terms are of minor importance for alternating bubble collapse and
rebound. It should be added that we have verified this conclusion for smaller
initial radii R0 = 10 µm and R0 = 100 µm, although these results are not shown
here. In what follows, the compressible Rayleigh-Plesset equation is utilized,
and mass transfer terms are retained anywhere.
In Figure 8, our results named sibub are compared with reference results from
[45], who utilized a Navier-Stoles method, so that homobaricity is not assumed.
The Navier-Stokes results have been carefully extracted from [45] and binarized.
Since no information on the spatial distribution of pG has been provided in [45]
we restrict our comparison to R and the vapor content mH2 O /mH2 O,0 of the
bubble. Regarding R, only minor differences are discernible. Although the
deviation of mH2 O /mH2 O,0 is more pronounced than the deviation of R, the
trends are clearly reproduced by our sibub simulations. It should be pointed
out that remaining differences may be influenced by slightly different initial and
boundary conditions, since particular details of the Navier-Stokes simulation
e.g. on initial dissolved gas content or on property evaluation are not given in
[45]. Overall, we conclude that the agreement to the Navier-Stokes solution is

29
a) Bubble radius b) Bubble pressure
108
1.0 Compressible RP-Eq.
Incompressible RP-Eq.
107 Omission of mass flux
R/R0 [−]

p G [Pa]
0.5 106

105
0.0
104
0 5 10 0 5 10
t/τ [−] t/τ [−]

Figure 7: Temporal progression of a) bubble radius and b) bubble pressure for R0 = 1000 µm,
with the compressible and incompressible (aL → ∞) form of the Rayleigh-Plesset equation
(RP-Eq.), in terms of equation 14. Furthermore, the compressible RP-Eq. with the omission
of mass flux, i.e. ṁ00 = 0 is illustrated.

remarkable, and it can be assumed that the spatial variation of pG is small, and
thus that the homobaricity assumption is valid even during collapse. We have
verified that |Ṙ| < 0.1aG is fulfilled at any stage of bubble dynamics and for
any value of R0 which is in line with the condition |Ṙ|  aG for homobaricity
[55].

a) Bubble radius b) Bubble radius c) Vapor mass


(Close-up A)
1.0 R0 = 100 µm 0.41 1.0
mH2 O /mH2 O, 0 [−]
m

µm
µm

1000
=1
R0 =

R0 =
R/R0 [−]

R/R0 [−]

R0
1000

100

R0 =
m

0.5 0.28 0.5


R0 =

0 µm
10 µ

100
R0 =

R0 = 1000 µm
µm
R0 =

R0 = 100 µm
R0 =

A R0 = 10 µm
10 µ

sibub
100

Navier-
0.0 0.15 0.0
µm

Stokes [45]

0 1 2 0.95 1.05 1.15 0 1 2


t/τ [−] t/τ [−] t/τ [−]

Figure 8: Bubble radius (a,b) and vapor mass (c), obtained by sibub or a Navier-Stokes
solution [45]. Navier-Stokes results have been reprinted from [45].

Although it can be assumed that the homobaricity assumption is justified,


the assumption that both vapor and air behave like ideal gases, may be chal-
lenged. Ideal gas behavior may be assumed if the pressure is significantly smaller
than the critical pressure, or the temperature is significantly higher than the
critical temperature. The critical pressure amounts to about 3.8 · 106 Pa and
2.2 · 107 Pa for air and water, respectively. At the final stage of collapse, pG

30
rises to almost 107 Pa, so that particularly for air, the ideal gas behavior should
be challenged. The critical temperature amounts to about 133 K and 647 K for
air and water, respectively. Thus, during rebound when the temperature is low,
ideal gas behavior may be assumed for air while it may be challenged for water.
Moreover, the validity of Henry’s law is questionable specifically for high bubble
pressure during collapse. We nevertheless retain these assumptions in this study
due to their simplicity and, not least, because the same assumptions have been
made by the reference Navier-Stokes solution [45] and accept that the absolute
peak values need to be taken with care.
By this preliminary assessment of the mathematical model, the importance
of compressibility as well as heat and mass transfer on alternating bubble col-
lapse and rebound has been demonstrated. In the following, compressibility and
the transport processes of the full transport model are therefore retained.

5.3. Effect of thermal and mass diffusion on bubble dynamics


In Figure 9a, the temporal progression of R is depicted. The bubble per-
forms an oscillating motion with alternating collapse and rebound phases. Its
radius attenuates due to viscous and thermal damping and approaches a new
equilibrium state, which occurs faster as R0 is reduced. For t/τ = 20, R/R0
equals 0.50, 0.45 and 0.56 for R0 = 1000 µm, 100 µm and 10 µm, respectively.

a) Bubble Radius
1.0
R/R0 [−]

0.5

0.0
R0 = 1000 µm
b) Vapor and air mass R0 = 100 µm
R0 = 10 µm
1.0 1.0000
mH2 O /mH2 O, 0 [−]

mAir /mAir, 0 [−]

Air
0.5 Vapor 0.9975

0.0 0.9950
0 5 10 15 20
t/τ [−]

Figure 9: Temporal progression of bubble radius as well as vapor and air mass within the
bubble.

In Figure 9b, the temporal progression of the bubble content is depicted. The
initial vapor mass mH2 O,0 corresponds to 7 · 10−11 , 7 · 10−14 and 7 · 10−17 kg, and

31
the initial air mass mAir,0 to 4 · 10−10 , 5 · 10−13 and 1 · 10−15 kg, for R0 = 1000,
100 and 10 µm respectively. As discussed in section 2.4, bubbles have been
initialized in equilibrium, which has also been done by [45]. Thus both initial
air and vapor mass correlate with R0 . On the other hand, bubbles contain
significantly more air than vapor for any initial radius. According to Figure 9b,
the bubbles loose vapor as well as air in the course of time, while the former is
much more pronounced due to evaporation and re-condensation during collapse
and rebound, respectively, this is also demonstrated by the mass flux portions
in terms ṁ00H2 O and ṁ00Air which are depicted in Figure 10 for the first three
oscillation cycles.

a) Vapor mass flux b) Air mass flux


0.2 0.02
ṁ 00H2 O [kg/(s m 2 )]

ṁ 00Air [kg/(s m 2 )]

0.0 0.00
0.2 0.02
R0 = 1000 µm
R0 = 100 µm
0.6 0.06 R0 = 10 µm
0 3 6 0 3 6
t/τ [−] t/τ [−]

Figure 10: Temporal progression of a) vapor and b) air mass flux.

The negative mass flux peaks indicate flux out of the bubble, and in fact
are more developed by one order of magnitude for the vapor flux than the air
flux. Thus, air mass is essentially constant during bubble oscillation which
can be traced back to the fact that air de- and absorption is limited by a
slow diffusion of dissolved air in the liquid, compared to evaporation and re-
condensation of vapor. It should be remembered that according to section 2.4,
L
yAir,0 is evaluated by Henry’s law and thus depends on pG 0 . The smaller R0 is,
G
the larger p0 is for the same initial liquid pressure p∞,0 = 104 Pa, and therefore,
L
L
yAir,0 L
also varies with R0 . In particular, yAir,0 equals 1.8 · 10−6 , 2.1 · 10−6 and
−6
5.1·10 for R0 = 1000, 100 and 10 µm respectively. In preliminary simulations,
L
we have initialized yAir,0 by atmospheric pressure pG 5
0 = 10 Pa, which yields
yAir,0 = 2.3 · 10 for any R0 . Interestingly, the time evolution of ṁ00H2 O and
L −5

ṁ00Air are virtually not affected by the different initialization.


It is interesting to note that the vapor mass abates in a kind of non-linear
way, which means that final vapor mass does not arrange in order of the initial
bubble radius. To be more precise, vapor mass mH2 O /mH2 O,0 at t/τ = 20
corresponds to 0.16, 0.09 and 0.18, for R0 = 1000, 100 and 10 µm, respectively,
and vapor mass loss is thus more pronounced for R0 = 100 µm, which is also
reflected by the negative vapor mass flux peaks in Figure 10a which are also more

32
pronounced for R0 = 100 µm. This non-linear behavior of bubble vapor content
is associated with the duration of collapse and rebound as will be discussed
next.

a) Bubble Radius
1.0 A
R/R0 [−]

Collapse Rebound
0.5 1.0 0.35

Close-up A

Close-up B
B 0.7 0.15
0.0 1.2 2.2 0.9 1.1

107 b) Bubble pressure

106
p G [Pa]

105

104
c) Bubble mean temperature
1400 R0 = 1000 µm
300
Close-up C

R0 = 100 µm
R0 = 10 µm
T G [K]

800
250
1.2 2.2

200 C
0.0 1.5 3.0 4.5 6.0
t/τ [−]

Figure 11: Temporal progression of radius, bubble pressure and bubble mean temperature.

G
In Figure 11, the temporal progression of R, pG and T is depicted for the
initial phase of bubble oscillation, i.e. up to t/τ = 6. The spatial mean tem-
G
perature T is evaluated by a volume-weighted average. Collapse temperature
and to a smaller extent also collapse pressure are lower for smaller R0 . What
is more, a smaller initial bubble radius is associated with a stronger damping
G
of R, pG and T oscillations which can be traced back to different magnitudes
of viscous and thermal damping terms. Viscous damping is important for very
small bubbles, and also the role of thermal damping increases for smaller bub-
bles [89], which is in line with our observations in Figure 11. While the damping
properties clearly correlate with R0 , it is interesting to note that the bubble col-

33
lapse duration shows a nonlinear behaviour: the duration of the first collapse
and rebound is evaluated by the minimum and maximum temporal value of R
and is summarized in Table 2. The intermediate R0 = 100 µm bubble shows the
shortest collapse and rebound duration, in agreement with the Navier-Stokes
results by [45].

Table 2: Different phases of bubble dynamics during first collapse and rebound, in terms of
∆t/τ .
R0 [µm]
Reference 1000 100 10
Collapse Duration Fig. 11a 1.02 1.01 1.06
Rebound Duration 0.95 0.90 0.92
G
Duration until T maximum Fig. 11c 1.02 1.02 1.05
G
Duration from T maximum to 0.86 0.53 0.27
its minimum
Duration of condensation period Fig. 13 & 14 1.25 1.15 1.14
Duration of evaporation period 0.87 0.92 0.92

It can also be seen from Figure 11 that pG synchronizes with bubble motion,
which means that when R has its minimum at the end of the collapse phase, pG
G
gets its maximum and vice versa. Also T gets its maximum essentially when R
G
has its minimum, which is discernible in Table 2, where the duration until the T
maximum virtually corresponds to the collapse duration. However, particularly
G
for smaller bubbles, the duration from T maximum to its minimum is shorter
than the collapse duration which is also summarized in Table 2. A temporal
G
offset occurs between minimum of T (marked in Close-up C in Figure 11c)
and maximum of R (marked in Close-up A in Figure 11a). This offset is also
illustrated in Close-up C in Figure 11c by horizontal arrows and gets obviously
larger with a smaller value of R0 .
G
The offset between minimum T and maximum R can be traced back to
heat transfer over the bubble wall [45]. In Figure 12, for the largest and the
smallest value of R0 , the time progression of volume work rate Ẇ , heat flux rate
Q̇ and total energy rate Ė are depicted which are defined as:
dVB
Ẇ = −pLw (50)
dt
 L

L ∂T 00
Q̇ = AB λw − LH2 O ṁH2 O (51)
∂r w
Ė = Ẇ + Q̇ (52)

with the bubble volume VB = 4/3πR3 and the bubble surface AB = 4πR2 . It
should be pointed out that Eq. 52 is an approximation of energy conservation
that is based on the first principle of thermodynamics for a closed system.
However, the bubble cannot be considered as a closed system as there is a

34
a) Volume work
30 0.003
Ẇ [J/s]

Ẇ [J/s]
0 0.000

Close-up C R0 = 1000 µm
0.5 0.00025
30 0.0 0.00000
R0 = 10 µm 0.003
b) Heat flux
1.5 0.00075
1 1.0 1.2 0.0005
0 0.0000
Q̇ [J/s]

Q̇ [J/s]
C
2 0.0010
Close-up B
Close-up A 0.1 0.0001
4 10 0.0005 0.0020
0 0.0000 0.0 0.0000
c) Total energy 0.1 0.0001
30 30 0.0015 1.7 2.1 0.0015
1.0 1.4

B
Ė [J/s]

Ė [J/s]
0 0.0000

30 A 0.0015
0.0 1.5 3.0 4.5 6.0
t/τ [−]

Figure 12: Rate of volume work Ẇ , heat flux Q̇ and total energy Ė.

mass flux through the interface. While the heat flux by phase transition is
considered by the second term in brackets on the right hand side of Eq. 51,
the heat transport by mass diffusion is neglected. Since heat transport by mass
diffusion is small compared to phase transition (note our discussion on Eq. 24),
we consider this approximation to be appropriate.
During collapse, the bubble receives work from the liquid, and heat is trans-
ferred from the bubble into the surrounding liquid, corresponding to Ẇ > 0 and
Q̇ < 0. Ẇ changes rapidly during the final stage of collapse, and its highest
gradient coincides with the minimum peak of R. At the final collapse stage,
Ẇ changes its sign, and the bubble starts to supply work to the liquid. Q̇ gets
positive somewhat later during the rebound phase, due to the wall-adjacent
temperature gradients within the bubble that change their sign delayed in time.
G
Thus, Q̇ changes its sign with a time lag to Ẇ . In the rebound phase, T takes

35
its minimum when Ė gets zero, which is in line with energy conservation. Par-
ticularly, Ė gets zero before R reaches its maximum. This observation is valid
for any initial bubble radius and corresponds to the offset between R maximum
G
and T minimum, as discussed in Figure 11.
In Figure 13a, the vapor saturation pressure pSat H2 O and partial vapor pressure
G
pH2 O,w are depicted for R0 = 1000 µm. Their difference constitutes the driving
pressure difference of the vapor mass flux ṁ00H2 O according to Eq. 23. In the col-
lapse phase, pSat G 00
H2 O T L − pH2 O,w < 0, and thus ṁH2 O < 0, which corresponds to
w
vapor condensation. Thus, bubble collapse is clearly associated with condensa-
tion. In Figure 13b, the temperature on the vapor and liquid side of the bubble
wall in terms of TwG and TwL is depicted. According to Eq. 22, the temperature
difference is immediately associated to ṁ00H2 O , and condensation corresponds to
TwL − TwG < 0 which is clearly reflected by the temperature progression in Figure
13b. It is interesting to note that the condensation process lasts well into the
rebound phase. Not earlier than at t/τ = 1.25 (marked by an arrow in Figure
13), both, pSat G L G 00
H2 O T L − pH2 O,w and Tw − Tw change their sign, thus ṁH2 O gets
w
positive and evaporation starts. At t/τ = 2.13 (also marked by an arrow in Fig-
ure 13), the sign of ṁ00H2 O again changes and again the next condensation
cycle
starts. The alternating pressure and temperature differences pSat L − pG
H2 O T H2 O,w
w
and TwL − TwG are even more pronounced for smaller bubbles which is illustrated
in Figure 14 for example for R0 = 10 µm. Compared to R0 = 1000 µm, the
intersection points slightly move to earlier instants in terms of t/τ = 1.14 and
2.06, respectively. The duration of the condensation and evaporation phase is
also summarized in Table 2. Interestingly, with decreasing R0 , the condensa-
tion duration abates, while the evaporation phase slightly increases. This shows
a different trend as the collapse and rebound duration and demonstrates that
bubble dynamics, mass transfer and heat transfer occur in a time-shifted and
non-linear way.
Heat and mass transfer are closely related to the temperature and mass
fraction gradients within and outside the bubble. By evaluating the location
where 99% of the respective far field value is obtained, the liquid boundary
layer thickness for water mass fraction δyHL O and temperature δT L is obtained
2
(not shown here) and reveals that δT L is about three times larger than δyHL O .
2
For R0 = 10 µm, δ even reaches the order of R in the last stage of bubble
collapse. It is interesting to note that in their pioneering work on heat transfer
assessment, [13] assumed R  δT which is clearly not the case here.
In Figure 15 and Figure 16, for distinct time instants between t/τ = 0 and
t/τ = 1 (collapse phase) as well as between t/τ = 1 and t/τ = 2 (rebound
phase), radial profiles on the gaseous and liquid side of the bubble wall are
depicted for R0 = 1000 µm and R0 = 10 µm. As discussed above, during bubble
collapse, condensation in terms of pSat G
H2 O T L −pH2 O,w < 0 occurs, and vapor mass
G
w
G
fraction at the bubble wall yH 2O w
, i.e. yH 2O
at r/R = 1, decreases according to
Figure 15a and Figure 16a. For R0 = 1000 µm, condensation lasts to t/τ = 1.25

36
a) Saturation and partial vapor pressure at the bubble wall for R0 = 1000µm
7000 A
Close-up A Close-up B pHSat2 O | TwL
7000 2600
pHG2 O, w
2.13
p [Pa]

4500
2000 2300
1.25 0.9 1.1 1.2 2.4
B
2000
b) Bubble wall temperature for R0 = 1000µm
310 Close-up C TwG
295
TwL
2.13
T [K]

300
1.25 293
C 1.2 2.4

290
0.0 1.5 3.0 4.5 6.0
t/τ [−]

Figure 13: Saturation and vapor pressure (a) and temperature (b) at bubble wall for R0 =
1000 µm.

according
to Table 2, but even after t/τ = 1.0, i.e. before evaporation starts,
G
yH O
2 w
rises again according to Figure 15c, Close-up B. The liquid mass fraction
yHL rises also steeply after t/τ = 1.0, according to Figure 15d. Thus, the
2O w
dynamics of wall mass fraction synchronizes with the collapse duration rather
than condensation and evaporation which are associated with the spatial wall
gradients of mass fraction. On the liquid side, steep gradients occur essentially
in the wall proximity as seen in Figure 15b and d. On the gaseous side during
collapse, the boundary layer does not quite reach the bubble center. In the
G
subsequent rebound phase, while at the wall yH 2O w
rises, in the inner bubble
G
region between about r/R = 0.2 ... 0.8, yH2 O continues to decrease.
For R0 = 10 µm (Figure 16), the water mass fraction on the liquid side
yHL gets its minimum at t/τ = 1.1, as can be seen in Figure 16d, which
2 O w
essentially corresponds to the end of the collapse phase according to Table 2.
Afterwards in the rebound phase, yH L rises again. Regarding the gaseous
2O w
G
boundary layer, Figure 16a and c, the vapor mass fraction yH 2O
decreases over
the entire bubble radius rather homogeneously in the collapse phase, mean-
ing that the boundary layer extends to the bubble center. In spite of a slight
G
overshoot in the bubble center proximity in the initial phase of rebound, yH 2O
also increases essentially homogeneously during rebound. Thus, for R0 = 10 µm
there is a fundamental difference of inner-bubble boundary layer dynamics com-
pared to R0 = 1000 µm that is traced back to different diffusion and dynamic

37
a) Saturation and partial vapor pressure at the bubble wall for R0 = 10µm
4000 pHSat2 O | TwL
pHG2 O, w
p [Pa]

3000 1.14
2.06

2000
b) Bubble wall temperature for R0 = 10µm
305.0 TwG
TwL
T [K]

297.5 1.14
2.06

290.0
0.0 1.5 3.0 4.5 6.0
t/τ [−]

Figure 14: Saturation and vapor pressure (a) and temperature (b) at bubble wall for R0 =
10 µm.

time scales according to [45]. These time scales are defined as:

R2
τdif = (53)
DG
R
τdyn = (54)

In the course of collapse, τdyn gets very small, while τdif takes a peak value
due to a strong temperature-dependent decrease of DG . Particularly for the
large bubble R0 = 1000 µm, the ratio τdyn /τdif takes very small values down
to 2 · 10−4 in the late stage of collapse and early stage of rebound. Since
τdif  τdyn , there is not enough time for the diffusive vapor transport towards
the bubble center, and vapor is trapped around the center. For the small bubble
R0 = 10 µm, the ratio τdyn /τdif takes values always above 5 · 10−2 even in the
late stage of the collapse so that τdif is much closer to τdyn , leading to a more
homogeneous mass fraction distribution within the bubble.
Regarding the spatial temperature distribution, the variation of the liquid
temperature is within a range of 10 K and thus generally low. According to [55],
the heat transfer is controlled by the thermal resistance of the bubble and not
of the liquid, and therefore we focus our discussion on the bubble temperature
distribution in terms of T G . The radial distribution of T G is depicted in Figure
17 and Figure 18 for R0 = 1000 µm and R0 = 10 µm, respectively. According

38
a) Vapor mass fraction b) Liquid water mass fraction
R0 = 1000µm, Collapse phase R0 = 1000µm, Collapse phase
0.2 1.00000
t/τ = 0.0 A Non-equilibrium
Equilibrium

t/τ = 0.9
t/τ

.0
=1
yHG2 O [−]

yHL 2 O [−]
= 1.0 Close-up A

t/τ
0.1 0.99995 1.00000 t/τ = 0.0

Time Time = 0.8

946
t/τ = 0.9
t/τ

yHL2 O → 0.99

943
Non-equilibrium 0.99998

yHL2 O → 0.99
0.0 Equilibrium 0.99990 1.000 1.001
0.0 0.5 1.0 1.000 1.003 1.006
r/R [−] r/R [−]
c) Vapor mass fraction d) Liquid water mass fraction
R0 = 1000µm, Rebound phase R0 = 1000µm, Rebound phase
0.2 Non-equilibrium 1.00000
Equilibrium Time
t/τ = 1.0 C Non-equilibrium
t/τ = Equilibrium
yHG2 O [−]

yHL 2 O [−]

Close-up B 2.0 2.0 Close-up C


0.1 0.1
Time 0.99995 1.00000 t/τ = 2.0
t/τ = 1.1

t/τ =
1.0 1.3 Time
1.2
1.2

1.1
t/τ = 1.0

1.1 t/τ =
t/τ

0.0
0.0 B 0.99990
0.99998
0.95 1.00 1.000 1.001
0.0 0.5 1.0 1.000 1.003 1.006
r/R [−] r/R [−]

Figure 15: Radial distribution of water mass fraction for R0 = 1000 µm within bubble (a,c)
and outside bubble (b,d), for collapse phase (a,b) and rebound phase (c,d).

to [45], when the initial radius is large, the motion approaches that of an adi-
abatic bubble, and when the radius is small, the motion approaches that of an
isothermal one, reflected by the T G distribution. While for R0 = 1000 µm the
T G boundary layer is restricted to the proximity of the bubble wall and the tem-
perature in the inner bubble region is essentially homogeneous, for R0 = 10 µm,
there is a spatial temperature distribution that extends over the entire bubble
which is clearly associated with a lower peak temperature than for the large
bubble. It is worthy to recap that according to Figure 12 Close-up C, the heat
flux changes its sign at t/τ ≈ 1.1 for both initial bubble sizes, meaning that from
this instant on, heat starts to flow from the liquid into the bubble, correspond-
G
ing to a wall temperature gradient ∂T ∂r > 0. Particularly for R0 = 1000 µm,
w
G
the mean temperature T is however yet significantly higher than the liquid
temperature, an observation that has been also made by [55], and could make

39
a) Vapor mass fraction b) Liquid water mass fraction
R0 = 10µm, Collapse phase R0 = 10µm, Collapse phase
0.08 Non-equilibrium 1.00000
Equilibrium
t/τ = 0.0
yHG2 O [−]

yHL 2 O [−]

9 .
t/τ = 0
0.04 0.99996

Time

t/τ = 1.0
t/τ = 1.0
Non-equilibrium
0.00 Time 0.99992 Equilibrium

0.0 0.5 1.0 1.00 1.02 1.04


r/R [−] r/R [−]
c) Vapor mass fraction d) Liquid water mass fraction
R0 = 10µm, Rebound phase R0 = 10µm, Rebound phase
0.08 0.015 t/τ = 1.1 1.5 1.00000 Time t/τ = 2.0
Non-equilibrium
Close-up A

1.2 1.4 Equilibrium . 1


t/τ = 1
1.2
Time
yHG2 O [−]

yHL 2 O [−]

0.005 t/τ = 1.3


Close-up B
0.04 0.99996
1.3

0.0 0.4 t/τ = 2.0


0.99995
1.2
1.0
t/τ = 1.0 Time = 1.1
t/τ
t/τ = 1.0

1.3
1.0
0.00 t/τ = 1.3 A 0.99992 B 0.99975
Time 1.00 1.01
0.0 0.5 1.0 1.00 1.02 1.04
r/R [−] r/R [−]

Figure 16: Radial distribution of water mass fraction for R0 = 10 µm within bubble (a,c) and
outside bubble (b,d), for collapse phase (a,b) and rebound phase (c,d).

believe that a heat flux is going out of the bubble. Thus, the consideration of
G
T only for an assessment of heat flux might be completely misleading.

5.4. Assessment of equilibrium bubble interface condition


The discussion in section 5.3 has been based on results that have been ob-
tained by the non-equilibrium bubble interface condition in terms of Eqs. 22
and 23. In section 4.2 we have already revealed that the differences to results
obtained by the equilibrium condition in terms of Eqs. 20 and 21 are minor
for the case of bubble growth. It is interesting to assess the impact of the wall
interface condition also for collapse and rebound, so that we have repeated the
simulations by employing the equilibrium condition. Only selected results are
discussed in detail, since the main conclusion is that differences from the non-
equilibrium results are again small. For example, in the plots in Figure 13a and
Figure 14a, the pSat
H2 O T L progression of the non-equilibrium and equilibrium re-
w

40
a) Bubble temperature b) Bubble temperature
R0 = 1000µm, Collapse phase R0 = 1000µm, Rebound phase
1400 340 t/τ = 0.5 t/τ = 1.0 Time 1400 300

1.4
1.8 t/τ

1.6
= 1.0
t/τ = 0.4

2.0
Close-up A

Close-up B
t/τ

=
t/τ = 0.3

=1

t/τ
.9
T G [K]

T G [K]
290 250
800 0.6 1.0 800 0.6 1.0 t/τ
t/τ = 0.9 = 1.1
Non-equilibrium Non-equilibrium
Equilibrium t/τ = 0.8 Equilibrium t/τ = 1
t/τ = 1.3 .2

200 t/τ = 0.0


A 200 t/τ = 2.0
Time B
0.0 0.5 1.0 0.0 0.5 1.0
r/R [−] r/R [−]

Figure 17: Radial distribution of bubble temperature T G for R0 = 1000 µm in collapse phase
(a) and rebound phase (b).

a) Bubble temperature b) Bubble temperature


R0 = 10µm, Collapse phase R0 = 10µm, Rebound phase
600 t/τ = 1.0 Time Close-up A 600 Close-up B
310 t/τ = 1.0 300 2.0 t/τ
=1
.2
t/τ = 1
Time

.
1
1.3
Time

1.5
T G [K]

T G [K]

290 200 1.4


400 0.9 1.0
400
Non-equilibrium 0.0 1.0
Equilibrium
2.0 t/τ = 1.2
t/τ = 0.0 A 1 .3
Non-equilibrium 1.5 B
200 Equilibrium 200 1.4

0.0 0.5 1.0 0.0 0.5 1.0


r/R [−] r/R [−]

Figure 18: Radial distribution of bubble temperature T G for R0 = 10 µm in collapse phase


(a) and rebound phase (b).

lation can visually not been told apart (the latter is not shown in Figure 13a and
Figure 14a for clearness of the illustration). In fact, differences are below 0.3%
even during the latest stage of collapse. This result is remarkable, because for
the non-equilibrium model the pressure difference at the wall pSat G
H2 O T L − pH2 O,w
w
can be considered as a driving force for the mass transfer and in this stage, the
pressure difference is particularly pronounced. For the equilibrium model, on
Sat
the other hand, pG H2 O,w equals per definition the saturation pressure pH2 O T L .
w

41
In the temporal progression of mass fraction profiles in Figure 15 and Figure
16 the equilibrium results are depicted. It is again interesting to note that they
hardly deviate from the non-equilibrium results. Some more pronounced dif-
ferences are discernible for the small bubble R0 = 10 µm that are however still
minor.
The wall temperature TwL from the equilibrium model (not shown in Fig-
ure 13b and Figure 14b for clearness of the illustration) per definition equals
TwG and can again visually not been told apart from non-equilibrium TwL . Re-
maining minor differences are on the same low level as differences of pG H2 O,w ,
i.e. below 0.3%. The radial temperature distribution in the liquid domain (not
shown in the preceding section) is essentially the same for the equilibrium and
the non-equilibrium case a similar finding has already been discussed in section
4.2, Figure 5. Regarding the bubble interior, we have again included the radial
T G profiles obtained by the equilibrium model in Figure 17 and Figure 18. As
for the mass fraction, it is interesting to note that the equilibrium results can
hardly be distinguished from the non-equilibrium results. As already noted in
section 4.2, according to Brennen [89], the difference between non-equilibrium
and equilibrium boundary condition is small when the accommodation coeffi-
cient γ is close to one, so that the similarity of equilibrium and non-equilibrium
results is certainly favored by the rather large value of γ = 0.4. On the other
hand, this result is in line with [55] who pointed out that at moderate velocities
of the bubble surface, the phase transition
virtually follows a quasi-equilibrium
scheme in terms of TwG ≈ TwL ≈ THSat L , which means that there is no delay of
2 O p
w

phase transition kinematics. We have evaluated the maximum of Ṙ during the


final stage of collapse to be in the range of 20 to 30 m/s, subject to R0 which
is obviously low enough on the one hand to justify the homobaricity assump-
tion discussed above, and on the other hand to keep the equilibrium assumption
valid. It should be recapped that we have applied a pressure jump in terms of
pL∞ from 104 to 105 Pa. By increasing the pressure jump, it can be expected
that the equilibrium results increasingly depart from the non-equilibrium re-
sults, and it would be interesting to figure out the validity and limitations of
the equilibrium assumption for higher pressure jumps.
In preliminary simulations, we have applied a pressure jump up to 8 · 105 Pa.
The maximum of Ṙ and of pG rise to about 450 m/s and 7, 000 bar, respectively,
so that for sure the homobaricity, ideal gas and Henry’s law assumptions are
seriously violated. In spite of this limitation, it is interesting to note that the
percentage differences between the equilibrium and non-equilibrium solutions
are still in the same small range as for the jump to 1 · 105 Pa. Of course, due to
the questionable assumptions of homobaricity, ideal gas and Henry’s law, this
investigation is preliminary and needs to be verified by e.g. a Navier-Stokes
method and by a real gas model. Regarding the former, Eqs. 8, 10 and 17 have
to be substituted by the spatially resolved momentum balance, which needs to
be solved together with the mass and energy balance either by a coupled or by
a segregated solution algorithm. Moreover, it should be contemplated lowering
γ for such a violent collapse, what has been suggested by [22, 72, 73]. This is

42
however out of the scope of this study and will be investigated in subsequent
studies. Here, we content ourselves with the conclusion that for the applied
boundary conditions in terms of the pL∞ jump, together with the assumption of
homobaricity, ideal gas, Henry’s law and an accommodation coefficient γ = 0.4,
the equilibrium model is essentially equivalent to the non-equilibrium model.

6. Conclusions

A mathematical model based on the homobaricity assumption has been pre-


sented, so that the numerical solution of momentum equations can be omitted.
The numerical scheme and corresponding computer program implementation
have been presented in detail, unlike most available studies. After a validation
on bubble growth, the collapse of differently sized spherical bubbles has been
studied, and details of local transport processes have been discussed. While
the main test case has been adopted from [45] who already discussed detailed
transport phenomena, the following conclusions can be drawn which go beyond
the available publications:
• Bubble collapse and subsequent rebound are accompanied by phase tran-
sition in terms of condensation and evaporation, while the amount of non-
condensable gas within the bubble only marginally changes by ab- and
desorption.
• Due to the good agreement with the Navier-Stokes results of [45] the
validity of the homobaricity assumption even for the late stage of bubble
collapse is demonstrated for |Ṙ| < 0.1 aG .
• Due to strong internal temperature variation, the use of the mean bub-
ble temperature for an assessment of heat transfer may be completely
misleading.
• The bubble motion does not correlate with evaporation and condensation
in a straightforward way. For a variation of the initial bubble size, collapse
and rebound duration do not correlate with condensation and evaporation
phase duration.
• For the pL∞ jump level in the range of 105 Pa, an equilibrium interface
condition at the bubble wall is appropriate even during the last stage of
bubble collapse and essentially the same mass fraction and temperature
profiles within and outside the bubble are obtained as with use of the
well-established non-equilibrium condition in terms of a Hertz-Knudsen
relationship, with an accommodation coefficient γ = 0.4.
Certainly, this last conclusion on the validity of the equilibrium interface
condition will be further explored for higher jump levels of pL∞ in a subsequent
study, together with a Navier-Stokes method, a real gas model and a lower
value of the accommodation coefficient γ. The equilibrium condition is of our
particular interest since in future studies, we intend to extend the mathematical

43
model beyond the water-air system, e.g. to n-Alkane mixtures. Mixing rules
are often based on equilibrium assumptions and may easily be adopted e.g.
from the well-established field of multi-component droplet evaporation that will
allow a straightforward development of a multi-component bubble dynamics and
cavitation model.
Regarding the homobaricity assumption, it can be concluded that it seems
to remain valid in a wide range of bubble dynamics scenarios. Due to the
omission of the spatially resolved momentum equation, it is computationally
less expensive than the Navier-Stokes equations and may open the opportunity
to embed a large number of single bubbles in a 3D Euler-Lagrange framework, in
combination with detailed transport processes for each single Lagrange bubble
which will be the subject of our subsequent studies.

Acknowledgment

The preparation of a first version of a sibub computer program manual by


Simon Paepenmller and Yannick Dren is kindly acknowledged. The authors
gratefully acknowledge the financial support by the German Research Founda-
tion (DFG), Grant No. 355240670. The authors also would like to thank the
editor and the anonymous reviewers for their valuable comments and suggestions
to improve the quality of the paper.

References

[1] K.-H. Kim, G. Chahine, J.-P. Franc, A. Karimi. (Eds.), Advanced experi-
mental and numerical techniques for cavitation erosion prediction, Springer,
2014.
[2] K. Schmitz, H. Murrenhoff, Modelling of the influence of entrained and
dissolved air on the performance of an oil-hydraulic capacity, Int. J. Fluid
Power 16 (3) (2015) 175–183. doi:10.1080/14399776.2015.1110094.

[3] U. Iben, A. Makhnov, A. Schmidt, Numerical study of the effects of dis-


solved gas release in cavitating flow, AIP Conf. Proc. 2027 (1) (2018)
030128. doi:10.1063/1.5065222.
[4] U. Iben, A. Makhnov, A. Schmidt, Numerical study of dissolved gas release
induced by cavitation in a high speed channel flow, J. Phys.: Conf. Ser.
1400 (2019) 077037. doi:10.1088/1742-6596/1400/7/077037.
[5] U. Iben, F. Wolf, H.-A. Freudigmann, J. Fröhlich, W. Heller, Optical mea-
surements of gas bubbles in oil behind a cavitating micro-orifice flow, Exp.
Fluids 56 (6) (2015) 114. doi:10.1007/s00348-015-1979-6.

[6] H.-A. Freudigmann, A. Drr, U. Iben, P. F. Pelz, Modeling of Cavitation-


Induced Air Release Phenomena in Micro-Orifice Flows, J. Fluids Eng.
139 (11), 111301. doi:10.1115/1.4037048.

44
[7] K. Kowalski, S. Pollak, R. Skoda, J. Hussong, Experimental Study on
Cavitation-Induced Air Release in Orifice Flows, J. Fluids Eng. 140 (6),
061201. doi:10.1115/1.4038730.
[8] L. Rayleigh, VIII. On the pressure developed in a liquid during the collapse
of a spherical cavity, Philos. Mag. 34 (200) (1917) 94–98. doi:10.1080/
14786440808635681.
[9] M. S. Plesset, A. Prosperetti, Bubble dynamics and cavitation, Annu. Rev.
Fluid Mech. 9 (1) (1977) 145–185. doi:10.1146/annurev.fl.09.010177.
001045.

[10] W. Soh, On the thermodynamic process of a pulsating vapor bubble, Appl.


Math. Model. 18 (12) (1994) 685 – 690. doi:10.1016/0307-904X(94)
90394-8.
[11] A. Aliabadi, A. Taklifi, The effect of magnetic field on dynamics of gas
bubbles in visco-elastic fluids, Appl. Math. Model. 36 (6) (2012) 2567 –
2577. doi:10.1016/j.apm.2011.09.040.
[12] D. Albernaz, F. Cunha, Unsteady motion of a spherical bubble in a complex
fluid: Mathematical modelling and simulation, Appl. Math. Model. 37 (20)
(2013) 8972 – 8984. doi:10.1016/j.apm.2013.03.065.
[13] M. S. Plesset, S. A. Zwick, A nonsteady heat diffusion problem with
spherical symmetry, J. Appl. Phys. 23 (1) (1952) 95–98. doi:10.1063/
1.1701985.
[14] W. K. Soh, A. A. Karimi, On the calculation of heat transfer in a
pulsating bubble, Appl. Math. Model. 20 (9) (1996) 638 – 645. doi:
10.1016/0307-904X(96)00044-3.

[15] A. Lezzi, A. Prosperetti, Bubble dynamics in a compressible liquid. Part


2. Second-order theory, J. Fluid Mech. 185 (1987) 289321. doi:10.1017/
S0022112087003185.
[16] L. Trilling, The collapse and rebound of a gas bubble, J. Appl. Phys. 23 (1)
(1952) 14–17. doi:10.1063/1.1701962.

[17] J. B. Keller, M. Miksis, Bubble oscillations of large amplitude, J. Acoust.


Soc. Am. 68 (2) (1980) 628–633. doi:10.1121/1.384720.
[18] S. Fujikawa, T. Akamatsu, Effects of the non-equilibrium condensation
of vapour on the pressure wave produced by the collapse of a bub-
ble in a liquid, J. Fluid Mech. 97 (3) (1980) 481–512. doi:10.1017/
S0022112080002662.
[19] H. Alehossein, Z. Qin, Numerical analysis of Rayleigh-Plesset equation for
cavitating water jets, Int. J. Numer. Methods Eng. 72 (7) (2007) 780–807.
doi:10.1002/nme.2032.

45
[20] N. A. Kudryashov, D. I. Sinelshchikov, Analytical solutions for problems of
bubble dynamics, Phys. Lett. A 379 (8) (2015) 798 – 802. doi:10.1016/
j.physleta.2014.12.049.
[21] S. Mller, P. Helluy, J. Ballmann, Numerical simulation of a single bubble
by compressible two-phase fluids, Int. J. Numer. Methods Fluids 62 (6)
(2010) 591–631. doi:10.1002/fld.2033.
[22] E. Lauer, X. Hu, S. Hickel, N. Adams, Numerical modelling and investiga-
tion of symmetric and asymmetric cavitation bubble dynamics, Comput.
Fluids 69 (2012) 1 – 19. doi:10.1016/j.compfluid.2012.07.020.

[23] M. Koch, C. Lechner, F. Reuter, K. Khler, R. Mettin, W. Lauterborn,


Numerical modeling of laser generated cavitation bubbles with the finite
volume and volume of fluid method, using OpenFOAM, Comput. Fluids
126 (2016) 71 – 90. doi:10.1016/j.compfluid.2015.11.008.
[24] S. Joshi, J. P. Franc, G. Ghigliotti, M. Fivel, Bubble collapse induced cavi-
tation erosion: Plastic strain and energy dissipation investigations, J. Mech.
Phys. Solids 134 (2020) 103749. doi:10.1016/j.jmps.2019.103749.
[25] Q. Zeng, S. R. Gonzalez-Avila, R. Dijkink, P. Koukouvinis, M. Gavaises,
C.-D. Ohl, Wall shear stress from jetting cavitation bubbles, J. Fluid Mech.
846 (2018) 341355. doi:10.1017/jfm.2018.286.

[26] P. Koukouvinis, G. Strotos, Q. Zeng, S. R. Gonzalez-Avila, A. Theodor-


akakos, M. Gavaises, C.-D. Ohl, Parametric investigations of the induced
shear stress by a laser-generated bubble, Langmuir 34 (22) (2018) 6428–
6442. doi:10.1021/acs.langmuir.8b01274.
[27] S. Nagrath, K. Jansen, R. T. Lahey, I. Akhatov, Hydrodynamic simulation
of air bubble implosion using a level set approach, J. Comput. Phys. 215 (1)
(2006) 98 – 132. doi:10.1016/j.jcp.2005.10.020.
[28] E. Johnsen, T. Colonius, Implementation of WENO schemes in compress-
ible multicomponent flow problems, J. Comput. Phys. 219 (2) (2006) 715
– 732. doi:10.1016/j.jcp.2006.04.018.

[29] N. V. Petrov, A. A. Schmidt, Multiphase phenomena in underwater ex-


plosion, Exp. Therm. Fluid Sci. 60 (2015) 367 – 373. doi:10.1016/j.
expthermflusci.2014.05.008.
[30] N. Petrov, A. Schmidt, Effect of a bubble nucleation model on cavitating
flow structure in rarefaction wave, Shock Waves 27 (4) (2017) 635–639.
doi:10.1007/s00193-016-0699-z.
[31] A. Smolianski, H. Haario, P. Luukka, Numerical study of dynamics of single
bubbles and bubble swarms, Appl. Math. Model. 32 (5) (2008) 641 – 659.
doi:10.1016/j.apm.2007.01.004.

46
[32] D. Fuster, T. Colonius, Modelling bubble clusters in compressible liquids,
J. Fluid Mech. 688 (2011) 352389. doi:10.1017/jfm.2011.380.
[33] A. Zein, M. Hantke, G. Warnecke, On the modeling and simulation of
a laser-induced cavitation bubble, Int. J. Numer. Methods Fluids 73 (2)
(2013) 172–203. doi:10.1002/fld.3796.
[34] S. Yakubov, B. Cankurt, T. Maquil, P. Schiller, M. Abdel-Maksoud,
T. Rung, Advanced Lagrangian Approaches to Cavitation Modelling in
Marine Applications, Springer Netherlands, Dordrecht, 2013, pp. 217–234.
doi:10.1007/978-94-007-6143-8_13.
[35] E. Ghahramani, M. H. Arabnejad, R. E. Bensow, A comparative study
between numerical methods in simulation of cavitating bubbles, Int. J.
Multiph. Flow 111 (2019) 339 – 359. doi:10.1016/j.ijmultiphaseflow.
2018.10.010.
[36] A. Peters, O. el Moctar, Numerical assessment of cavitation-induced erosion
using a multi-scale Euler-Lagrange method, J. Fluid Mech. 894 (2020) A19.
doi:10.1017/jfm.2020.273.
[37] R. I. Nigmatulin, N. S. Khabeev, Dynamics of vapor bubbles, Fluid Dyn.
10 (3) (1975) 415–421. doi:10.1007/BF01015265.
[38] A. Arefmanesh, S. G. Advani, E. E. Michaelides, An accurate numerical
solution for mass diffusion-induced bubble growth in viscous liquids con-
taining limited dissolved gas, Int. J. Heat Mass Transf. 35 (7) (1992) 1711
– 1722. doi:10.1016/0017-9310(92)90141-E.
[39] A. Naji Meidani, M. Hasan, Mathematical and physical modelling of bubble
growth due to ultrasound, Appl. Math. Model. 28 (4) (2004) 333 – 351.
doi:10.1016/j.apm.2003.10.001.
[40] K. J. Vachaparambil, K. E. Einarsrud, Numerical simulation of bubble
growth in a supersaturated solution, Appl. Math. Model. 81 (2020) 690 –
710. doi:10.1016/j.apm.2020.01.017.
[41] S. Sochard, A.-M. Wilhelm, H. Delmas, Gas-vapour bubble dynamics and
homogeneous sonochemistry, Chem. Eng. Sci. 53 (2) (1998) 239 – 254.
doi:10.1016/S0009-2509(97)85744-2.
[42] B. D. Storey, A. J. Szeri, Mixture segregation within sonoluminescence bub-
bles, J. Fluid Mech. 396 (1999) 203221. doi:10.1017/S0022112099005984.
[43] J. P. Boyd, Orthogonal rational functions on a semi-infinite interval, J.
Comput. Phys. 70 (1) (1987) 63 – 88. doi:10.1016/0021-9991(87)
90002-7.
[44] V. Kamath, A. Prosperetti, Numerical integration methods in gasbubble
dynamics, J. Acoust. Soc. Am. 85 (4) (1989) 1538–1548. doi:10.1121/1.
397356.

47
[45] Y. Matsumoto, F. Takemura, Influence of internal phenomena on gas
bubble motion: Effects of thermal diffusion, phase change on the gas-
liquid interface and mass diffusion between vapor and noncondensable
gas in the collapsing phase, JSME Int. J. Ser. B 37 (2) (1994) 288–296.
doi:10.1299/jsmeb.37.288.

[46] F. Takemura, Y. Matsumoto, Influence of internal phenomena on gas bub-


ble motion: Effects of transport phenomena and mist formation inside bub-
ble in the expanding phase, JSME Int. J. Ser. B 37 (4) (1994) 736–745.
doi:10.1299/jsmeb.37.736.
[47] Y. Matsumoto, S. Yoshizawa, Behaviour of a bubble cluster in an ul-
trasound field, Int. J. Numer. Methods Fluids 47 (67) (2005) 591–601.
doi:10.1002/fld.833.
[48] K. Aoki, Y. Sone, Gas flows around the condensed phase with strong evap-
oration or condensation — fluid dynamic equation and its boundary con-
dition on the interface and their application —, in: R. Gatignol, Soub-
baramayer (Eds.), Advances in Kinetic Theory and Continuum Mechanics,
Springer Berlin Heidelberg, Berlin, Heidelberg, 1991, pp. 43–54.
[49] Y. Jinbo, K. Kobayashi, M. Watanabe, H. Takahira, Numerical simula-
tion of bubble collapse and the transfer of vapor and noncondensable gas
through the bubble interface using the ghost fluid method, J. Phys.: Conf.
Ser. 656 (2015) 012021. doi:10.1088/1742-6596/656/1/012021.
[50] M. Sussman, P. Smereka, S. Osher, A level set approach for computing
solutions to incompressible two-phase flow, J. Comput. Phys. 114 (1) (1994)
146 – 159. doi:10.1006/jcph.1994.1155.
[51] R. P. Fedkiw, T. Aslam, B. Merriman, S. Osher, A Non-oscillatory Eulerian
Approach to Interfaces in Multimaterial Flows (the Ghost Fluid Method),
J. Comput. Phys. 152 (2) (1999) 457 – 492. doi:10.1006/jcph.1999.6236.
[52] B. D. Storey, A. J. Szeri, Water vapour, sonoluminescence and sonochem-
istry, Proc. R. Soc. A 456 (1999) (2000) 1685–1709. doi:10.1098/rspa.
2000.0582.

[53] K. Yamamoto, K. Kobayashi, M. Watanabe, H. Fujii, M. Kon, H. Takahira,


Influence of a small amount of noncondensable gas on shock wave generation
inside a collapsing vapor bubble, Phys. Rev. Fluids 4 (2019) 063603. doi:
10.1103/PhysRevFluids.4.063603.
[54] H. Kawashima, M. Kameda, Dynamics of a spherical vapor/gas bubble in
varying pressure fields, J. Fluid Sci. Technol. 3 (8) (2008) 943–955.
[55] R. Nigmatulin, N. Khabeev, F. Nagiev, Dynamics, heat and mass transfer
of vapour-gas bubbles in a liquid, Int. J. Heat Mass Transf. 24 (6) (1981)
1033 – 1044. doi:10.1016/0017-9310(81)90134-4.

48
[56] A. Prosperetti, The thermal behaviour of oscillating gas bubbles, J. Fluid
Mech. 222 (1991) 587616. doi:10.1017/S0022112091001234.
[57] C. F. Delale, Ş. Pasinlioğlu, First iterative solution of the thermal behaviour
of acoustic cavitation bubbles in the uniform pressure approximation, J.
Phys.: Conf. Ser. 656 (2015) 012016. doi:10.1088/1742-6596/656/1/
012016.
[58] A. Ali, S. B. Ake, A new formulation and analysis of a collapsing bubble
with different content in a liquid induced during acoustic cavitation, Int. J.
Numer. Methods Heat Fluid Flow 26 (6) (2016) 1729–1746. doi:10.1108/
HFF-02-2015-0044.

[59] H. Lin, B. D. Storey, A. J. Szeri, Inertially driven inhomogeneities in vio-


lently collapsing bubbles: the validity of the Rayleigh-Plesset equation, J.
Fluid Mech. 452 (2002) 145162. doi:10.1017/S0022112001006693.
[60] M. Nickaeen, Spatially resolved simulations of the non-equilibrium cavita-
tion bubble dynamics including vapor and air transport, Doctoral Thesis,
Ruhr-Universität Bochum, Universitätsbibliothek (2020). doi:10.13154/
294-7027.
[61] W. Wagner, H.-J. Kretzschmar, International steam tables : properties of
water and steam ; based on the industrial formulation IAPWS-IF97, Berlin
Springer, 2008.

[62] R. Bird, W. Stewart, E. Lightfoot, Transport Phenomena, Wiley, 2006.


[63] G. N. Lewis, The law of physico-chemical change, Proc. Am. Acad. Arts
Sci. 37 (3) (1901) 49–69.
[64] M. D. Koretsky, Engineering and chemical thermodynamics, Hoboken, NJ
Wiley 2013, 2013.
[65] I. Akhatov, O. Lindau, A. Topolnikov, R. Mettin, N. Vakhitova, W. Lauter-
born, Collapse and rebound of a laser-induced cavitation bubble, Phys.
Fluids 13 (10) (2001) 2805–2819. doi:10.1063/1.1401810.

[66] S. Fujikawa, T. Yano, M. Watanabe, Vapor-liquid interfaces, bubbles and


droplets: fundamentals and applications, Springer Science I& Business Me-
dia, 2011.
[67] Y. Sone, Molecular gas dynamics: theory, techniques, and applications,
Springer Science I& Business Media, 2007.

[68] M. Kotani, T. Tsuzuyama, Y. Fujii, S. Fujikawa, Nonequilibrium vapor


condensation in shock tube, JSME Int. J. Ser. B 41 (2) (1998) 436–440.
doi:10.1299/jsmeb.41.436.

49
[69] A. H. Persad, C. A. Ward, Expressions for the evaporation and conden-
sation coefficients in the Hertz-Knudsen relation, Chemical Rev. 116 (14)
(2016) 7727–7767.
[70] S. Fujikawa, M. Maerefat, A study of the molecular mechanism of vapour
condensation, JSME Int. J. Ser. 2, Fluids Eng., Heat Transf., Power,
Combust., Thermophys. Prop. 33 (4) (1990) 634–641. doi:10.1299/
jsmeb1988.33.4_634.
[71] K. Kobayashi, T. Nagayama, M. Watanabe, H. Fujii, M. Kon, Molecu-
lar gas dynamics analysis on condensation coefficient of vapour during
gasvapour bubble collapse, J. Fluid Mech. 856 (2018) 10451063. doi:
10.1017/jfm.2018.722.
[72] E. Lauer, X. Y. Hu, S. Hickel, N. A. Adams, Numerical investigation of
collapsing cavity arrays, Phys. Fluids 24 (5) (2012) 052104. doi:10.1063/
1.4719142.
[73] R. Marek, J. Straub, Analysis of the evaporation coefficient and the con-
densation coefficient of water, Int. J. Heat Mass Transf. 44 (1) (2001) 39 –
53. doi:10.1016/S0017-9310(00)00086-7.
[74] R. Sander, Compilation of Henry’s law constants (version 4.0) for water
as solvent, Atmos. Chem. Phys. 15 (8) (2015) 4399–4981. doi:10.5194/
acp-15-4399-2015.

[75] J. M. Prausnitz, R. N. Lichtenthaler, E. G. de Azevedo, Molecular thermo-


dynamics of fluid-phase equilibria, Pearson Education, 1998.
[76] B. E. Poling, J. P. O’Connell, J. M. Prausnitz, The properties of gases and
liquids, New York McGraw-Hill, 2001.

[77] NIST Chemistry Webbook. NIST Standard Reference Database Number


69. (2018). doi:10.18434/T4D303.
[78] A. Wassiljewa, Wrmeleitung in Gasgemischen, Phys. Z. 5 (22) (1904) 737–
742.

[79] E. A. Mason, S. C. Saxena, Approximate formula for the thermal con-


ductivity of gas mixtures, Phys. Fluids 1 (5) (1958) 361–369. doi:
10.1063/1.1724352.
[80] H. Recknagel, E.-R. Schramek, Taschenbuch fr Heizung und Klimatech-
nik : einschlielich Warmwasser- und Kltetechnik, Mnchen Oldenbourg-
Industrieverl, 2003.
[81] E. N. Fuller, K. Ensley, J. C. Giddings, Diffusion of halogenated hydrocar-
bons in helium. The effect of structure on collision cross sections, J. Phys.
Chem. 73 (11) (1969) 3679–3685. doi:10.1021/j100845a020.

50
[82] C. Hirt, A. Amsden, J. Cook, An Arbitrary LagrangianEulerian Computing
Method for All Flow Speeds, J. Comput. Phys. 135 (2) (1997) 203 – 216.
doi:10.1006/jcph.1997.5702.
[83] I. Demirdi, M. Peri, Space conservation law in finite volume calculations
of fluid flow, Int. J. Numer. Methods Fluids 8 (9) (1988) 1037–1050. doi:
10.1002/fld.1650080906.
[84] J. Crank, P. Nicolson, A practical method for numerical evaluation of solu-
tions of partial differential equations of the heat-conduction type, Advanc.
Comput. Math. 6 (1) (1996) 207–226. doi:10.1007/BF02127704.

[85] M.-H. Lallemand, Dissipative properties of Runge-Kutta schemes with up-


wind spatial approximation for the Euler equations., Research Report RR-
1173, INRIA (1990).
[86] W. Kahan, Pracniques: Further remarks on reducing truncation errors,
Commun. ACM 8 (1) (1965) 40. doi:10.1145/363707.363723.

[87] M. S. Plesset, S. A. Zwick, The growth of vapor bubbles in superheated


liquids, J. Appl. Phys. 25 (4) (1954) 493–500. doi:10.1063/1.1721668.
[88] H. K. Forster, N. Zuber, Growth of a vapor bubble in a superheated liquid,
J. Appl. Phys. 25 (4) (1954) 474–478. doi:10.1063/1.1721664.

[89] C. E. Brennen, Cavitation and bubble dynamics, The Oxford engineering


science series: 44, New York [u.a.] Oxford Univ. Press 1995, 1995.
[90] P. Dergarabedian, The rate of growth of vapor bubbles in superheated
water., ASME J. Appl. Mech. 20 (1953) 537–545.
[91] M. Plesset, The dynamics of cavitating bubbles., ASME J. Appl. Mech. 16
(1949) 277–282.
[92] M. Nickaeen, T. Jaskolka, S. Mottyll, R. Skoda, A study on thermally
controlled bubble growth in a superheated liquid with a thermal non-
equilibrium cavitation model based on energy balances on a fixed fluid
mass, WIT Trans. Eng. Sci. 89 (2015) 387 – 398. doi:10.2495/MPF150331.

[93] M. Alamgir, J. H. Lienhard, Correlation of Pressure Undershoot During


Hot-Water Depressurization, J. Heat Transfer 103 (1) (1981) 52–55. doi:
10.1115/1.3244429.
[94] F. Wolf, Gaskavitation in Kohlenwasserstoffen: Einfluss von ruhender, str-
mender und kavitierender Flssigkeit, Doctoral Thesis, Technische Univer-
sität Dresden (2015).

51

You might also like