International Journal of Multiphase Flow 129 (2020) 103320

Contents lists available at ScienceDirect

International Journal of Multiphase Flow

journal homepage: www.elsevier.com/locate/ijmulflow

Path instability of a compressible air bubble rising in quiescent water

with consideration of variable thermophysical properties
Yuanwei Cao∗, Inés Mateos Canals, Rafael Macián-Juan
Institute of Nuclear Engineering, Department of Mechanical Engineering, Technische Universität München (TUM), Garching bei München 85748, Germany

a r t i c l e i n f o a b s t r a c t

Article history: A numerical study of a compressible air bubble rising in water with variable thermophysical properties
Received 16 January 2020 is conducted. The coupled level set and volume of fluid method is adopted to track the interface. Em-
Revised 18 March 2020
pirical correlations in terms of temperature for thermophysical properties and equation of state for air
Accepted 17 April 2020
are considered in the simulations. We simulate a compressible 3D static air bubble in water for the first
Available online 26 April 2020
time. The volume change of the bubble and the pressure difference between the inside and outside of
Keywords: the static bubble agree well with the theoretical solutions. We study the zigzag motion of a compressible
Bubble dynamics air bubble considering the temperature difference with respect to the surrounding water. The mechanism
Zigzag motion of the zigzag motion is clarified by analyzing the shape evolution and the wake structure. The periodic
Compressibility effect shedding of hairpin vortex with 2R mode is observed during the zigzag motion. There are two periods
Variable thermophysical properties of water of the bubble shape oscillation within one zigzag cycle. We obtain the relation between the oscillation
CLSVOF
frequency of the aspect ratio (fE ), the zigzag path (fp ) and vorticity component in X and Y direction accu-
mulated on the bubble surface (fv ) as fE = 2 f p = 2 fv .
© 2020 Elsevier Ltd. All rights reserved.

1. Introduction
Bubbles play an important role in a wide range of natural and
engineering applications, such as nuclear industry, steam gener-
ation in thermal plants, enhanced oil recovery, brewing indus-
try and so on (Anna, 2016; Zenit and Rodriguez, 2018). The bub-
ble topological changes and the rising path (Aboulhasanzadeh and
Tryggvason, 2014) could enhance the heat transfer during their ris-
ing motion. On the contrary, the heat transfer could in turn influ-
ence the bubble dynamics due to the dependence of water prop-
erties on the temperature (here we refer to the two-phase systems
with atmospheric pressure). Accounting for the variation of ther-
mophysical properties and the compressibility of air is crucial for
the investigation of bubble rising dynamics in consideration of the
temperature difference between the bubble and the surrounding
fluid. However, this research topic is barely covered in the liter-
ature (Majlesara et al., 2020). Therefore, the main objective of the
current study is to simulate a compressible air bubble rising in qui-
escent water with variable thermophysical properties and compare
the results among the cases with different thermal conditions.
The oscillatory motion of a bubble rising in quiescent liquid,
coupling the bubble shape, path and wake, has been widely stud-
∗
Corresponding author.
E-mail address: yuanwei.cao@tum.de (Y. Cao).
ied. Duineveld (1995) confirmed for the first time that the bub-
ble presented an oscillatory path in water when the bubble radius
is bigger than 0.91 mm by using hyper-purified water. In the nu-
merical simulation of Mougin and Magnaudet (2001), they found
that the path instability of the rising bubble resulted from the
wake instability behind the bubble. Later, Magnaudet and Mou-
gin (2007) studied the uniform flow past an ellipsoidal bubble
and identified the critical value of Re and aspect ratio (χ ) for the
appearance of wake instability. Cano-Lozano et al. (2016) identi-
fied different path of a bubble rising in water including rectiliear,
chaotic, spiral and zigzag. Zhang and Ni (2017) studied the ris-
ing bubble transiting from zigzag to spiral. They found the equal-
strength vortex pairs that shed twice during a period of zigzag mo-
tion. However, all the above mentioned works about the path in-
stability focus on the incompressible two phase flow systems with
constant physical properties.
Some researchers study the bubble dynamics with the vari-
ation of one property like surface tension (Tripathi et al.,
2015a; Balla et al., 2019) or viscosity (Potapov et al., 2006; Tri-
pathi et al., 2015b). A review can be found in the work of
Sahu (2017). The surface tension of water decreases almost lin-
early with increasing temperature, which is termed as “linear”
fluid (Tripathi et al., 2015a). The thermocapillary motion of bub-
bles in a “linear” fluid was first reported by Young et al. (1959).
Tripathi et al. (2015a) conducted a simulation of bubble ris-
ing by considering a linear and quadratic dependence of surface

https://doi.org/10.1016/j.ijmultiphaseflow.2020.103320
0301-9322/© 2020 Elsevier Ltd. All rights reserved.
2 Y. Cao, I. Mateos Canals and R. Macián-Juan / International Journal of Multiphase Flow 129 (2020) 103320

tension on temperature without referring to path instability. density ρ , viscosity μ are based on the volume fraction α . The
Balla et al. (2019) simulated the bubble rising in “self-rewetting” physical properties are calculated in Section 2.4. ∇ · Uc α (1 − α )
fluid. The viscosity of water can also vary due to the presence is an anti-diffusion term utilised to sharpen the interface and the
of temperature variation. Tsamopoulos et al. (2008) studied the model for Uc is given by
buoyancy-driven rise of a bubble in a Newtonian or a viscoplastic ∇α
fluid assuming axial symmetry and steady flow. However, in reality, Uc = min(cα |U|, max |U| ) (5)
|∇α|
all the physical properties, including density, viscosity, surface ten-
sion and thermal conductivity, are functions of temperature, which Where cα is the compression coefficient,which is 1 in present
could affect the bubble dynamics and can not be ignored. study. In practice values of 1 ≤ cα ≤ 4 give good behaviour
Despite all studies on this subject, the simulations of a com- (Berberović et al., 2009; Deshpande et al., 2012) although for some
pressible air bubble rising in quiescent water with variable ther- cases it might be useful to use cα > 4. max |U| is the largest value
mophysical properties are still missing. Thus we conduct numer- of |U| anywhere in the domain whereas min (cα |U|, max |U|) re-
ical investigation on a compressible air bubble rising in quies- turns the minimum (cα |U|) field limited by maximum |U|.
cent water with variable thermophysical properties. For this pur- The advection Eq. (4) is solved using the multidimen-
pose, we implement the coupled Level Set and Volume of Fluid sional universal limiter with explicit solution (MULES) method
method (CLSVOF) (Albadawi et al., 2013) into the OpenFOAM solver (OpenCFD, 2012) improved based on flux corrected transport
compressibleInterFoam. Empirical correlations in terms of temper- method (FCT) (Zalesak, 1979), which ensures a sharp interface and
ature for thermophysical properties and equation of state for air limits the volume fraction to values between 0 and 1.
are added in the solver. The structure of the paper is organized as
follows: the numerical model employed in this work is introduced 2.2. Coupling VoF with LS method
in Section 2; isothermal and non-isothermal cases are used to val-
idate the numeircal method: single two-dimensional (2D) bubble The CLSVOF method in the work of Albadawi et al. (2013) is
rising, 3D skirted bubble rising, and non-isothermal static bubble, adopted here to improve surface tension implementation. After
which are described in Section 3; the mechanisms of the bubble solving Eq. (4) using MULES, a new level set field ϕ is introduced
zigzag motion in quiescent water in consideration of variable ther- by the iso-line contour α = 0.5 (Albadawi et al., 2013),
mophysical properties and the compressibility effect of the air bub- ϕ0 = (2αL − 1 ) ·  (6)
ble are covered in Section 4. Finally, Section 5 is devoted to provide
a summary of the main findings of this study. Where  is a small non-dimensional number whose value depends
on the mesh size x (non-dimensional),  = 0.75 x. Then the re-
2. Mathematical formulations initialization equation is given by:
 ∂ϕ
2.1. Governing equations = S(ϕ0 )(1 − |∇ϕ| )
∂τ (7)
ϕ ( x, 0 ) = ϕ0 ( x )
The present CLSVOF model is developed based on the default
OpenFOAM solver, compressibleInterFoam. The fundamental math- where τ is the artificial time step which is chosen as τ = 0.1 x
ematical equations consists of the conservation laws of mass, mo- and S(φ 0 ) is a sign function defined as,
mentum and energy as well as an advection equation. The gov- ϕ0
S ( ϕ0 ) = (8)
erning equations for two nonisothermal, compressible, immiscible |ϕ0 |
fluids including the conservation laws of mass, momentum, energy
To solve Eq. (7), only a few iterations (φ corr ) are required according
and advection equations are (Ma et al., 2016):
to Albadawi et al. (2013),
∂ρ
+ ∇ · (ρ U ) = 0 (1) φcorr =
∂t τ
(9)

∂ρ U where is the non-dimensional interface thickness calculated as

+ ∇ · (ρ UU ) − ∇ · (μ∇ U ) = Fσ − g · x∇ρ − ∇ pd (2) = 1.5 x.
∂t
2.3. Implementation of surface tension force
∂ρ T
+ ∇ · (ρ UT ) − ∇ · (μ∇ T )
∂t    Fσ in Eq. (2) is the surface tension term, which is treated as a
α 1−α ∂ρ k body force. For VoF method, the surface tension force is calculated
=− + + ∇ · ( ρ Uk ) + ∇ · ( U p ) (3)
Cv,water Cv,air ∂t as,
Fσ = σ κ (α )∇α (10)
∂α
+ U · ∇α + ∇ · Uc α (1 − α ) Where σ is the surface tension coefficient, κ is the interface cur-
∂t   vature calculated based on α ,
1 d ρ2 1 d ρ1
= α (1 − α ) − + α∇ · U (4) ∇α
ρ2 dt ρ1 dt κ (α ) = −∇ · (11)
|∇α|
Where U is the velocity vector, Fσ is the surface tension force,
g is the gravitational acceleration, t is the time and pd = p − ρ g · x For CLSVOF method, the surface tension force is calculated based
is the dynamic pressure used to avoid any sudden changes in the on the level set function, which can improve the accuracy of inter-
pressure at the boundaries for hydrostatic problem, k = |U|2 /2 is face curvature calculation,
the specific kinetic energy, T is the temperature, Cv, water and Cv, air ∇ϕ
κ (ϕ ) = −∇ · (12)
are the specific heat capacities at constant volume for water and |∇ϕ|
air phases, respectively. α is the volume fraction in each compu-
So the surface tension force can be calculated as,
tational cell, which can be used to distinguish between the two
fluids. The calculations of Fσ and the physical properties, like the Fσ = σ κ (ϕ )δ (ϕ )∇ϕ (13)
 Y. Cao, I. Mateos Canals and R. Macián-Juan / International Journal of Multiphase Flow 129 (2020) 103320 3

Where δ (ϕ ) is the Dirac function used to limit the influence of sur- Table 1
Physical properties of the bubble and surrounding fluid for the test cases.
face tension to a narrow region around the interface, which is de-
fined as Test case ρ 1 (kg/m3 ) ρ2 μ1 (Pa · s) μ2 g(m/s2 ) σ (kg/s2 )
⎧
⎨0 if |ϕ| > 1 1000 100 10 1 0.98 24.5
δ (ϕ ) =  
⎩ 1 1 + cos π ϕ
(14)
if |ϕ| ≤
2
fd
2.4. Variation of physical properties St = (22)
UT
The density of air is correlated with pressure p and temperature Where R is the radius of the bubble, ρ 0 , μ0 and ρ i , μi are density
T by the following perfect gas equation of state (Ma et al., 2016): and viscosity of the continuous and dispersed phases, respectively.
f is the oscillation frequency of the path, d is the bubble diameter
p = ρair Rair T (15) and UT is the bubble terminal velocity (Cano-Lozano et al., 2016).
where Rair = 287J/(kg · K ) is the specific gas constant. In Open-
FOAM, the properties of water such as density, viscosity, thermal 3. Validation
conductivity, and surface tension coefficient, are treated as func-
tions of temperature T (OpenCFD, 2012): 3.1. 2D rising bubble
98.343885
ρwater = (16)
0.305421+(1− 647.13 ) Hysing et al. (2009) proposed two benchmark test cases, which
T 0.081

are TC1 and TC2, for quantitative validation and comparison of in-
 3670.6 terfacial flow codes. Because only the results of TC1 from all codes
μwater = exp −51.964 + + 5.7331 × ln T − 5.3495 × 10−29 × T 10 in (Hysing et al., 2009) agree very well with one another, we use
T
TC1 as the first validation case.
(17)
The initial configuration consists of a circular bubble of radius
r0 = 0.25 m centered at (0.5 m, 0.5 m) in a 1 m × 2 m rectan-
κwater = 1.815 × 10−9 × T 3 − 8.0065 × 10−6 gular domain. The physical properties of the bubble and surround-
ing fluid are listed in Table. 1. The no-slip boundary condition is
× T 2 + 0.0056903 × T − 0.4267 (18)
applied at the top and bottom boundaries while the free slip con-
ditions is used on the vertical walls. Zero gradient is applied for
 T (2.047× 647T.13 −3.554 )× 647T.13 +2.717
the pressure boundary conditions. The temperature is set to 293K
σwater = 0.18548 × 1 − (19)
and all the boundary conditions are set as zero gradient. The com-
647.13
putations were performed on regular hexahedron mesh with mesh
Here T is taken as the absolute temperature (Kelvin). Here we
size of h = 1/[160, 320]. The first-order Euler scheme was used for
assume the pressure of the two-phase systems to be atmospheric
time discretization with time step t = 1 × 10−4 s and maximum
pressure (101325 Pa). These values are very close to the experi-
Courant number Co = 0.05.
mental data of De-Yi et al. (1993) (the relative errors are within
Similar bubble shapes are observed compared to the results in
1%).
Hysing et al. (2009) shown in Fig. 1.
2.5. Dimensionless numbers Furthermore, for quantitative validation of the solver, the same
benchmark quantities in Hysing et al. (2009) are adopted, which
We employ the following non-dimensional parameters to de- are:
scribe the bubble rising problems in the present work: the Galilei • Centre of mass, position over time in the vertical axis:
number (Ga), Eötvös number (Eo), and Strouhal number (St).
√ xdS
ρ0 gR3/2 xcm = (xcm , ycm ) = α <0.5 (23)
Ga = (20)
μ0 α <0.5 dS

ρ0 gR2 where x represents the position, x = (x, y ) and dS is a surface

Eo = (21) infinitesimal.
σ

Fig. 1. The bubble shape evolution of test case at different time instants (h = 1/160).
4 Y. Cao, I. Mateos Canals and R. Macián-Juan / International Journal of Multiphase Flow 129 (2020) 103320
Fig. 2. Quantitative validation of (a) circularity, (b) center of mass, (c) mean rise velocity.
Table 2
The physical properties of skirted bubble in the work of Sharaf et al. (2017).
phases Density (kg/m3 ) Viscosity (mPa s)
liquid 1254 967.8
gas 1 0.0151
• Mean rise velocity (MRV):
UdS
u = α <0.5
α <0.5 dS
where U is the instant velocity at a particular point
• Circularity, defined as the ratio of the perimeter of an area-
equivalent circle by the actual perimeter of the bubble:
π deq
c=
 dx
where deq denotes the diameter of are-equivalent circle,  is a
curve defined by the interface between water and air in a two-
dimensional case.
The results obtained by present simulation are also contrasted
with those acquired by other three CFD methods appeared in
Hysing et al. (2009), which are Transport Phenomena in 2D (TP2D)
(Turek, 1999), Free-Surface Library of Finite Element (FreeLIFE)
(Parolini, 2004) and Mathematics and object-oriented Numerics in
MagDeburg (MooNMD) (John and Matthies, 2004). All the bench-
mark quantities obtained by present solver are smaller than that
by other methods from Fig. 2. Similar findings were reported in
the work of Klostermann et al. (2013) and Yamamoto et al. (2017).
But the relative errors are marginal (0.36%, 0.97% and 2.70%for cen-
ter of mass, mean rise velocity, circularity respectively). Based on
the aforementioned results, the accuracy of the present solver for
2D bubble rising is confirmed.
3.2. 3D rising bubble
We use the skirted bubble from the experimental work of
Sharaf et al. (2017) for 3D validation. The physical properties are
shown in Table. 2. The size of the fluid domain is 6d × 6d × 15d,
where d is the bubble diameter. Structured hexahedral mesh is
adopted for the present simulation. Slip boundary conditions are
applied for all the vertical walls (Cao and Macián-Juan, 2020). The
velocity and pressure of the top wall are set as pressureInletOut-
letVelocity and totalPresusre. The pressure of the bottom wall is set
as zeroGradient. The temperature inside the flow domain is 293K
and zeroGradient is applied for all the boundary conditions. The
mesh size is d/[10, 20, 40]. The time step is set to x/10s, where
x is the dimensionless mesh size. The adaptive time step com-
putation is controlled with Courant number (Co = 0.1). Ztip is the
Surface tension (kg/s2 ) Eo Ga
0.0624 73.28 10.86
bubble tip position normalized by the bubble radius and t∗ is the
dimensionless time normalized with R/g, where R is the bubble
(24) radius and g is the gravity acceleration.
We find an excellent agreement on the time evolution of a
skirted bubble and the corresponding numerical simulation shown
in Fig. 3(a). From the temporal variation of Ztip in Fig. 3(b), the re-
sults of mesh size d/20 and d/40 agree well with the experimental
value. Thus the accuracy of the solver for 3D isothermal bubble
rising is validated.
(25)
3.3. Static bubble
In the work of Francois et al. (2006), the authors simulated a
static bubble and concluded that a consistent coupling between
surface tension and pressure is necessary for obtaining a dis-
crete force balance. The CLSVOF method could reduce the spu-
rious currents due to the accurate calculation of interfae curva-
ture. Deshpande et al. (2012), Albadawi et al. (2013), Liu and
Palm (2016). Albadawi et al. (2013) compared the results of 2D
static bubble obtained from CLSVOF to that of VoF method. In the
present work, a 3D static air bubble at equilibrium in a zero gravity
field is revisited to validate our solver. But the difference is that we
consider temperature difference between the air bubble and water.
The initial bubble diameter is Dini = 0.01 m positioned at the
center of the cubic domain. The size of the cubic domain is
2D × 2D × 2Dm3 filled with water. Three cases will be simulated:
1) the temperature of water and air bubble is the same (isother-
mal); 2) the temperature of the air bubble is 293K while the wa-
ter is at 363K, which means the bubble will expand (cold bub-
ble); 3) the temperature of the air bubble is 363K while the wa-
ter is at 293K, which means the bubble will contract (hot bub-
ble). The viscosity of air is set as constant, 1.84 × 10−5 Pa · s.
We can get other properties, such as air density, water properties
from Eqs. (15) and (16)–(19). The boundary conditions are listed
in Table 3. The pressure value of the top wall is set to the atmo-
spheric pressure (101325 Pa). We allow the fluid in and out at the
top wall. So the bubble is allowed to expand and contract. The
mesh resolution is 20 and 40 cells per bubble diameter and the
time step are set to t = 1 × 10−5 s, which satisfies the stability
condition (Deshpande et al., 2012) given by
ρ x3
t< (26)
σ
 Y. Cao, I. Mateos Canals and R. Macián-Juan / International Journal of Multiphase Flow 129 (2020) 103320 5

Fig. 3. (a) Comparison of the bubble shape evolution between experimental (Sharaf et al., 2017) and present numerical results with different mesh sizes. (b) Comparison of
the temporal variation of Ztip obtained from different mesh sizes and the experimental results.

Table 3
The setting of boundary conditions for the static bubble.

boundary volume fraction pressure velocity temperature

bottom wall zeroGradient fixedFluxPressure fixedValue (0 0 0) zeroGradient

top wall inletOutlet totalPressure pressureInletOutletVelocity zeroGradient
side walls zeroGradient zeroGradient slip zeroGradient

Table 4
The errors of the pressure difference for bubble diameter 0.01 m at t = 0.2 s.
a
Mesh size Cases Rtheo (m) Ptheo (Pa) P0 (Pa) E0 (%) Ptotal (Pa) Etotal (%)

a 0.005 29.09 29 0.31 27.55 5.29

d/20 b 0.0053697 22.6645 22 2.93 23.12 2
c 0.004655 31.2627 31 0.84 29.95 4.29
a 0.005 29.09 29 0.31 28.931 0.55
d/40 b 0.0053697 22.6645 22 2.93 21.952 3.14
c 0.004655 31.2627 31 0.84 30.91 1.13
a
means the theoretical bubble radius calculated from Eq. (28)

To validate the solver with variable thermophysical properties,
we compare the simulation results of the pressure difference be-
tween the inside and outside of the bubble to the theoretical ones.
The pressure difference can be calculated theoretically from the
Young-Laplace equation (Brochard-Wyart and Quéré, 2004):
2σ
Ptheo =
R and < 3.14% for x = d/40. We can find that the relative error
where R is the radius of the bubble. In consideration of the com-
pressibility effect, the bubble radius is not constant. Therefore, the
perfect gas equation of state (28) should be considered as well.
pV = nRair T
where V is the bubble volume, n is the number of moles. Follow-
ing Albadawi’s work (Albadawi et al., 2013), two numerical approx-
imation of the pressure difference were calculated by the following
equations:
P0 = P0 − P∞
Ptotal = Pin − Pout
where
• P0 and P∞ are the value of pressure at the bubble center and at
the wall boundary, respectively.
• Pin and Pout are the averaged value of the pressure inside and
outside of the bubble, respectively.
The theoretical and numerical pressure difference as well as the
corresponded relative error at t = 0.2 s are shown in Table 4. The
errors E0 for all the cases with different mesh sizes are the same
(27) ( < 2.5%). The relative errors of Etotal are < 5.3% for x = d/20
Etotal becomes smaller with the mesh refinement. But the accuracy
of the results with mesh size d/20 is also acceptable.
In order to analyse how the bubble behaves under different
(28) combinations of air and water temperature, the bubble volume, the
position of its centre of mass, and sphericity of mesh resolution
d/40 have been plotted over time in Fig. 4. The bubble volume is
non-dimensionalized with the initial bubble volume. The vertical
position of its centre of mass is non-dimensionalized by its initial
vertical position. We can observe the obvious volume change for
(29) the non-isothermal bubbles. The relative errors of volume change
are 0.037% and 0.32% for cold and hot bubbles, respectively. All the
non-isothermal bubbles moved down away from the initial posi-
(30) tion and then converged to a point located lower than the initial
position. This is due to the bubble volume change and the setting
of the boundary conditions (top wall). We also checked the dis-
placement in other two directions for both bubble, which are be-
low the machine precision. Actually, the displacement is 1.33% and
6 Y. Cao, I. Mateos Canals and R. Macián-Juan / International Journal of Multiphase Flow 129 (2020) 103320

Fig. 4. Evolution of (a) bubble volume, (b) vertical position, and (c) sphericity for 10 mm bubble.

Fig. 5. Velocity vectors around and inside 10 mm bubble at t = 0.01 s (top row) and t = 0.2 s (bottom row), a) case a, b) case b, c) case c.

3.33% for cold and hot bubbles respectively. As observed in Fig. 4,
the sphericity of the bubbles, after an initial drop in the first in-
stants, recovers a little and reaches a steady value very close to 1
(less than 0.15% away from a perfect sphericity). Therefore, we can
conclude that the present solver provides accurate predictions for
non-isothermal two phase flow systems.
The implementation of the CSF model in the momentum equa-
tion generates spurious currents in the vicinity of the interface
(Renardy and Renardy, 2002). Here we the analyze influence of
heat transfer on the spurious currents around the bubble shown
in Fig. 5. The maximum velocity magnitudes appeared around the
interface between air and water (colours closer to red) with an or-
der of 10−1 . We can observe the volume change in Fig. 5(b) and
(c). The cold and hot bubbles in Fig. 5(b) and (c) move down due
to the volume change of the bubble and the boundary conditions.
The inflow and outflow are only allowed on the top wall, which
can be observed in Fig. 5(b) and (c). The expansion of the cold
bubble leads to the outflow on the top wall and the contraction
of the hot bubble results in the inflow on the top wall. With the
development of the velocity field inside the domain, the bubbles
move down.
4. Results and discussions
4.1. Mesh independence study
Before studying the oscillatory bubble dynamics, we conduct
mesh independence study. We adopt a cuboid flow domain with
6d × 6d × 10d. The bubble size and the boundary conditions
are the same as those in Section 3.3. The computations are per-
formed on a structured hexahedral mesh with the size of x =
d/[10, 20, 30, 40]. The time step is set as x/10 and the Courant
number is set to 0.5. In the following sections, all variables are
made dimensionless using the initial bubble diameter d and the
gravitational time d/g as characteristic length and time scales,
respectively. The velocity scale is gd.
We simulate the isothermal case (case a) in Section 3.3 for the
mesh independence study. The time history of the bubble rising
 Y. Cao, I. Mateos Canals and R. Macián-Juan / International Journal of Multiphase Flow 129 (2020) 103320
Fig. 6. The time history of the bubble rising velocity with different mesh resolution.
velocity with different mesh resolution is shown in Fig. 6. We can
see that the results of mesh size d/20 are converged. Therefore, the
mesh size x = d/20 is chosen for the following simulation to save
Fig. 7. Different paths of the bubbles.
7
computational time. Actually, in the work of Zun et al. (2012) and
Rek (2019), they suggested that there should be at least 15 cells
per bubble diameter, which proves the present mesh resolution
reasonable and computationally economic.
4.2. Overview of the simulations
In this section, we present the simulation results of a compress-
ible air bubble rising in quiescent water with variable thermophys-
ical properties. The 3D domain has a horizontal cross section of
6d × 6d and is 60d high in order to allow the bubble to reach a
steady state. The initial bubble diameter simulated in this section
is 10 mm. Four cases are investigated: 1) isothermal water and air
(293K); 2) hot air (363K) in cold water (293K); 3) isothermal wa-
ter and air (363K); 4) cold air (293K) in hot water (363K). Other
flow conditions are the same for these four cases. We initiate all
the simulations with a spherical bubble shape with a diameter of
10 mm. The bubble center initially stands 3R above the bottom
wall. The boundary conditions are the same as listed in Table 3.
The viscosity of air is set to 1.84 × 10−5 Pa · s. We can get other
properties, such as air density from Eq. (15) and water properties
from (16)–(19).
8 Y. Cao, I. Mateos Canals and R. Macián-Juan / International Journal of Multiphase Flow 129 (2020) 103320
Fig. 8. Transient results of aspect ration E, and displacement of bubble in X and Y direction for the four cases.
Fig. 9. Temporal evolution of the bubble shapes at different time.
Different trajectory paths are observed due to the variation of
water properties and compressible effect of air (Fig. 7). We observe
three stages in case 1 and 2, straight path, aperiodic zigzag and pe-
riodic zigzag motion (planar zigzag in Cano-Lozano et al. (2016)).
But in case 3 and 4, the trajectory can be divided into four
stages, straight path, spiral (chaotic spiral for case 4), aperiodic
zigzag and periodic zigzag motion (flattened spiraling in Cano-
Lozano et al. (2016)). The mechanism for the different paths will
be analyzed in the following sections.
4.3. Shape evolution
We employ the aspect ration, E recommended by Charin et al.
(2019) to evaluate the effect of bubble shape evolution on the path
instability, which is defined by:
Lz
E=
Lx Ly
The bubble rises due to buoyancy force initially following a straight
vertical trajectory. The distances of the straight path are about
8.4d, 8.8d, 8.0d and 7.9d for case 1 to 4 respectively (8). Dur-
ing this period, aspect ratios of all cases are oscillating, which is
caused by wobbling behavior of the bubbles. But the bubble shape
is axisymmetric displayed Fig. 9 at t = 0.3 s. The aspect ratio of
case 2(4) is a bit bigger(smaller) than that of case 1(3) due to the
decrease(increase) of the bubble volume resulting from the com-
pressibility effect of the air. In the aperiodic zigzag motion, the
bubbles lose their symmetry (Fig. 9 at t = 0.7 s). There are cer-
tain inclination of the bubble with respect to its path. During the
aperiodic motion, aspect ratios of all cases increase to around 0.6
and oscillate for 3 to 4 cycles (Fig. 8). In the periodic zigzag mo-
tion, the frequency of oscillation of the periodic zigzag motion is
about 5 Hz. The St number is about 0.2 using Eq. (22) after check-
ing the terminal bubble velocity. However, the frequency of the as-
pect ratio oscillation is about 10 Hz, twice that of the zigzag mo-
tion. We can find that when the bubble passes the furthermost po-
sitions, the aspect ratio decreases to the minimum value. In con-
trast, when the bubble passes the average position, the aspect ra-
tio reaches the maximum value. The reason can be clarified as fol-
lows. At the start of the zigzag loop, there is no inclination of the
bubble with respect to its path. Within the zigzag loop, the bub-
ble begins to tilt toward the direction of the path, achieving the
maximum inclination in the middle of the half cycle of the zigzag
(Cano-Lozano et al., 2016). After that, the inclination of the bubble
reduces until it vanishes again at the end of the present cycle of
zigzag. So there are two periods within one zigzag cycle. When the
bubble is in the extremity of zigzag, the bubble has no inclination,
corresponding to the maximal lateral excursion and minimum E.
In the middle of the half cycle of zigzag period, the bubble reaches
the maximum inclination and has the maximum E.
4.4. Wake structure
As the bubble rises, the axisymmetric toroidal rings develops
and attaches to the bubble (t = 0.3 s in Fig. 10). The flow around
the bubble separates from the body at some point because of its
excessive curvature and the vortices are not formed symmetrically
around the bubble (t = 0.6 s in Fig. 10). The asymmetric vortices
change the pressure distribution along the surface and different lift
(31) forces develop on each side of the bubble leading to zigzag motion
of the bubble. For case 1 and 2, we can find that the periodic hair-
pin vortex shedding in the periodic zigzag motion. The wake struc-
tures of both case have 2R mode, meaning double-sided periodic
sequences of vortex rings(Horowitz and Williamson, 2010; Cano-
Lozano et al., 2016). For case 3 and 4, the vortex attached to the
bubble is twisted and spiral corresponding to the spiral motion of
the bubble in Fig. 7(c) and (d). Different from the finding in Cano-
Lozano et al. (2016), the haipin is present in our work. At t = 1.2 s
and t = 2 s (Fig. 10(c) and (d)), the periodic shedding of hairpin
vortex with appearance of 2R mode can be clearly observed.
 Y. Cao, I. Mateos Canals and R. Macián-Juan / International Journal of Multiphase Flow 129 (2020) 103320
Fig. 10. The vortex structures revealed by λ2 criterion (with λ2 = 20) at different time for different cases. The iso-vortices are color coded with z .
4.5. Vortex dynamics
In this section, we calculate the vorticity component in X and
Y directions and the norm of streamwise vorticity accumulated on
the bubble surface, that is τ(X,Y ) = s ω(X,Y ) dS and |τZ | = s |ωZ |dS
(Zhang and Ni, 2017), where S is the bubble interface and ωX,Y is
the vorticity component in X and Y direction.
For case 1 and 2 in the stage of straight path, the positive
and negative vortices in X and Y direction can balance each other
(Fig. 11). In the aperiodic zigzag motion of the bubble, the oscilla-
tions of τ X,Y,Z and |τ Z | are aperiodic as well. In the stage of periodic
zigzag motion starting from about 1 s, the periodic oscillations of
τ X and τ Y are observed and the frequency is about 5 Hz, which
is the same to that of zigzag motion as shown in Fig. 8. We find
that the frequency of the oscillation of |τ Z | is about 10 Hz (twice
of the zigzag motion as well as τ X and τ Y ), which is the same to
the conclusion in Zhang and Ni (2017). In the paper of Zhang and
Ni (2017), they think that when the accumulation of |τ Z | on the
bubble surface reaches a critical value, the vortex shedding can
9
be triggered. In our paper, the critical values are different for the
two cases and the value of case 2 is smaller, which is caused by
the decrease of bubble volume due to the compressibility effect. In
the work of Magnaudet and Mougin (2007), they guess the critical
value is probably related to the interface curvature, which can be
proved in the present work. From Fig. 8, the frequency and further-
most position of E are the same as that of |τ Z |.
For case 3 and 4, in the stage of spiral motion (0.2s-0.5s), the
frequencies of the oscillation of τ X and τ Y are quite large and am-
plitude is increasing, which is quite different from the aperiodic
zigzag. In the transition phase from spiral to periodic zigzag, the
oscillation of τ X and τ Y is irregular. When τ X and τ Y reach a cer-
tain value, the bubble enters into the zigzag motion, same to the
flattened spiraling regime in Cano-Lozano et al. (2016). We can see
that the amplitude is larger than that of case 1 and 2 but the fre-
quency is almost the same, which is 5 Hz. Periodic oscillation of
|τ Z | is found and the frequency is 10 Hz twice that of zigzag mo-
tion, which is the same to that of case 1 and 2 but with larger
values.
10 Y. Cao, I. Mateos Canals and R. Macián-Juan / International Journal of Multiphase Flow 129 (2020) 103320
Fig. 11. The time history of the vorticity component in X and Y direction (τ (X,Y) ) and the magnitude of vorticity component in Z direction (|τ Z |) accumulated on bubble
surface for all cases.
5. Conclusion
In this work, we investigate the path instability of compress-
ible air bubbles rising in quiescent water considering the variation
of water properties. The CLSVOF method is adopted to track the
interface. The equation of state for ideal gas was considered. The
thermophysical properties are considered as function of tempera-
ture.
The solver is validated by a 2D and 3D bubble rising in quies-
cent liquid with published results. For the first time, we simulate
a compressible static 3D air bubble in water with variable thermo-
physical properties. The volume change of the 3D bubble and the
pressure difference between the inside and outside of the bubble
agree well with the theoretical solutions. The cold and hot bubble
move down due to the bubble volume change and the setting of
the boundary conditions.
We simulate the path instability of a compressible air bubble
rising in quiescent water with variable thermophysical properties
for the first time. Four cases are investigated: 1) isothermal water
and air (293K); 2) hot air (363K) in cold water (293K); 3) isother-
mal water and air (363K); 4) cold air (293K) in hot water (363K).
All the cases present zigzag motion but with distinctions. We ob-
serve three stages of trajectory in case 1 and 2, straight path, ape-
riodic zigzag and periodic zigzag motion. In case 3 and 4, the tra-
jectory can be divided into four stages, straight path, spiral (chaotic
spiral in case 4), aperiodic zigzag and periodic zigzag motion. The
wake structures of all the cases are dominated by a 2R mode. The
frequency of the aspect ratio oscillation is about 10 Hz, twice that
of the zigzag motion. The St number is about 0.2 for the periodic
zigzag motion. When the bubble passes the furthermost positions,
the aspect ratio decreases to the minimum value. In contrast, when
the bubble passes the average position, the aspect ratio reaches the
maximum value. The frequency of the periodic oscillations of τ X
and τ Y is 5 Hz, which is the same to that of zigzag motion. The
frequency of the oscillation of |τ Z | is twice of the zigzag motion as
well as τ X and τ Y , but the same as that of the aspect ratio.
Declaration of Competing Interest
The authors declare that they have no known competing finan-
cial interests or personal relationships that could have appeared to
influence the work reported in this paper.
CRediT authorship contribution statement
Yuanwei Cao: Conceptualization, Methodology, Software, For-
mal analysis, Investigation, Data curation, Writing - original draft,
Writing - review & editing, Visualization, Project administra-
tion. Inés Mateos Canals: Validation, Data curation, Visualization.
Rafael Macián-Juan: Resources, Funding acquisition.
Acknowlgedgments
Yuanwei Cao is financially supported by China Scholarship
Council (No. 201606060149). The authors gratefully acknowledge
the Gauss Centre for Supercomputing e.V. (www.gauss-centre.eu)
for funding this project by providing computing time on the
GCS Supercomputer SuperMUC at Leibniz Supercomputing Centre
(www.lrz.de).
Supplementary material
Supplementary material associated with this article can be
found, in the online version, at 10.1016/j.ijmultiphaseflow.2020.
103320.
References
Aboulhasanzadeh, B., Tryggvason, G., 2014. Effect of bubble interactions on mass
transfer in bubbly flow. Int. J. Heat Mass Transf. 79, 390–396.
Albadawi, A., Donoghue, D., Robinson, A., Murray, D., Delauré, Y., 2013. Influence of
surface tension implementation in volume of fluid and coupled volume of fluid
with level set methods for bubble growth and detachment. Int. J. Multiphase
Flow 53, 11–28.
Anna, S.L., 2016. Droplets and bubbles in microfluidic devices. Annu. Rev. Fluid
Mech. 48, 285–309.
Balla, M., Tripathi, M.K., Sahu, K.C., Karapetsas, G., Matar, O.K., 2019. Non-isothermal
bubble rise dynamics in a self-rewetting fluid: three-dimensional effects. J. Fluid
Mech. 858, 689–713.
Berberović, E., van Hinsberg, N.P., Jakirlić, S., Roisman, I.V., Tropea, C., 2009. Drop
impact onto a liquid layer of finite thickness: dynamics of the cavity evolution.
Phys. Rev. E 79 (3), 36306.
Brochard-Wyart, F., Quéré, D., 2004. Capillarity and Wetting Phenomena: Drops,
Bubbles, Pearls, Waves. Springer.
Cano-Lozano, J.C., Martinez-Bazan, C., Magnaudet, J., Tchoufag, J., 2016. Paths and
wakes of deformable nearly spheroidal rising bubbles close to the transition to
path instability. Phys. Rev. Fluids 1 (5), 53604.
Cao, Y., Macián-Juan, R., 2020. Numerical investigation of central breakup of large
bubble induced by liquid jet. Phys. Fluids 32 (3), 033302.
Charin, A.H., Lage, P.L., Silva, L.F.L., Tuković, Ž., Jasak, H., 2019. On the dynamic be-
havior of rising droplets. Int. J. Multiphase Flow 110, 165–178.
 Y. Cao, I. Mateos Canals and R. Macián-Juan / International Journal of Multiphase Flow 129 (2020) 103320
De-Yi, S., Bu-Xuan, W., Yang, W., Yui, Q., 1993. Study on liquid laminar free convec-
tion with consideration of variable thermophysical properties. Int. J. Heat Mass
Transf. 36 (14), 3411–3419.
Deshpande, S.S., Anumolu, L., Trujillo, M.F., 2012. Evaluating the performance of the
two-phase flow solver interfoam. Comput. Sci. Discov. 5 (1), 014016.
Duineveld, P., 1995. The rise velocity and shape of bubbles in pure water at high
Reynolds number. J. Fluid Mech. 292, 325–332.
Francois, M.M., Cummins, S.J., Dendy, E.D., Kothe, D.B., Sicilian, J.M., Williams, M.W.,
2006. A balanced-force algorithm for continuous and sharp interfacial surface
tension models within a volume tracking framework. J. Comput. Phys. 213 (1),
141–173.
Horowitz, M., Williamson, C., 2010. The effect of Reynolds number on the dynamics
and wakes of freely rising and falling spheres. J. Fluid Mech. 651, 251–294.
Hysing, S.-R., Turek, S., Kuzmin, D., Parolini, N., Burman, E., Ganesan, S., Tobiska, L.,
2009. Quantitative benchmark computations of two-dimensional bubble dy-
namics. Int. J. Numer. Methods Fluids 60 (11), 1259–1288.
John, V., Matthies, G., 2004. MooNMD–a program package based on mapped finite
element methods. Comput. Vis. Sci. 6 (2–3), 163–170.
Klostermann, J., Schaake, K., Schwarze, R., 2013. Numerical simulation of a single
rising bubble by VOF with surface compression. Int. J. Numer. Methods Fluids
71 (8), 960–982.
Liu, Q., Palm, B., 2016. Numerical study of bubbles rising and merging during con-
vective boiling in micro-channels. Appl. Therm. Eng. 99, 1141–1151.
Ma, Z., Causon, D., Qian, L., Mingham, C., Ferrer, P.M., 2016. Numerical investigation
of air enclosed wave impacts in a depressurised tank. Ocean Eng. 123, 15–27.
Magnaudet, J., Mougin, G., 2007. Wake instability of a fixed spheroidal bubble. J.
Fluid Mech. 572, 311–337.
Majlesara, M., Abouali, O., Kamali, R., Ardekani, M.N., Brandt, L., 2020. Numerical
study of hot and cold spheroidal particles in a viscous fluid. Int. J. Heat Mass
Transf. 149, 119206.
Mougin, G., Magnaudet, J., 2001. Path instability of a rising bubble. Phys. Rev. Lett.
88 (1), 14502.
OpenCFD, O., 2012. The Open Source CFD Toolbox–User’s Guide, vol. 1. OpenCFD
Ltd., United Kingdom.
Parolini, N., 2004. Computational Fluid Dynamics for Naval Engineering Problems.
Technical Report. EPFL.
11
Potapov, A., Spivak, R., Lavrenteva, O.M., Nir, A., 2006. Motion and deformation of
drops in Bingham fluid. Ind. Eng. Chem. Res. 45 (21), 6985–6995.
Rek, Z., 2019. Using a dynamic and constant mesh in numerical simulation of the
free-rising bubble. Fluids 4 (1), 38.
Renardy, Y., Renardy, M., 2002. PROST: a parabolic reconstruction of surface tension
for the volume-of-fluid method. J. Comput. Phys. 183 (2), 400–421.
Sahu, K.C., 2017. A review on rising bubble dynamics in viscosity-stratified fluids.
Sādhanā 42 (4), 575–583.
Sharaf, D., Premlata, A., Tripathi, M.K., Karri, B., Sahu, K.C., 2017. Shapes and paths
of an air bubble rising in quiescent liquids. Phys. Fluids 29 (12), 122104.
Tripathi, M.K., Sahu, K., Karapetsas, G., Sefiane, K., Matar, O., 2015. Non-isothermal
bubble rise: non-monotonic dependence of surface tension on temperature. J.
Fluid Mech. 763, 82–108.
Tripathi, M.K., Sahu, K.C., Karapetsas, G., Matar, O.K., 2015. Bubble rise dynamics in
a viscoplastic material. J. Nonnewton. Fluid Mech. 222, 217–226.
Tsamopoulos, J., Dimakopoulos, Y., Chatzidai, N., Karapetsas, G., Pavlidis, M., 2008.
Steady bubble rise and deformation in newtonian and viscoplastic fluids and
conditions for bubble entrapment. J. Fluid Mech. 601, 123–164.
Turek, S., 1999. Efficient Solvers for Incompressible Flow Problems: An Algorithmic
and Computational Approache, vol. 6. Springer Science & Business Media.
Yamamoto, T., Okano, Y., Dost, S., 2017. Validation of the S-CLSVOF method with
the density-scaled balanced continuum surface force model in multiphase sys-
tems coupled with thermocapillary flows. Int. J. Numer. Methods Fluids 83 (3),
223–244.
Young, N., Goldstein, J., Block, M.J., 1959. The motion of bubbles in a vertical tem-
perature gradient. J. Fluid Mech. 6 (3), 350–356.
Zalesak, S.T., 1979. Fully multidimensional flux-corrected transport algorithms for
fluids. J. Comput. Phys. 31 (3), 335–362.
Zenit, R., Rodriguez, J.R., 2018. The fluid mechanics of bubbly drinks. Phys. Today 71
(11), 44–50.
Zhang, J., Ni, M.-J., 2017. What happens to the vortex structures when the rising
bubble transits from zigzag to spiral? J. Fluid Mech. 828, 353–373.
Zun, I., Perpar, M., Gregorc, J., Hayashi, K., Tomiyama, A., 2012. Mixing of thermally
stratified water layer by a free rising wobbling air bubble. Chem. Eng. Sci. 72,
155–171.