Sunteți pe pagina 1din 18

Colloids and Surfaces A: Physicochem. Eng.

Aspects 250 (2004) 367384

Hydrodynamic effects on the oscillations of supported bubbles:


implications for accurate measurements of surface properties
Ying-Chih Liao, Osman A. Basaran, Elias I. Franses
School of Chemical Engineering, Purdue University, 480 Stadium Mall Drive, West Lafayette, IN 47907-2100, USA

Received 21 January 2004; accepted 6 April 2004


Available online 26 October 2004

Abstract
Oscillating bubble techniques are commonly used to infer dynamic surface tensions (DST) from the measurements of the dynamic pressure
differences across an interface. In inferring DST from such measurements, it is assumed that hydrodynamic efects either are negligible or can
be approximated. In virtually all previous studies, the dynamic bubble shapes, which are set by the requisite balance of forces at the bubble
surface, are taken to be nearly spherical and assumed to be governed by the static YoungLaplace (YL) equation alone. To examine these
assumptions, the NavierStokes and continuity equations governing the flow of a liquid outside an axisymmetric bubble supported by a narrow
capillary are solved simultaneously by a rigorous finite element method. Cases of constant surface tension (pure liquids or solutions with very
fast adsorption) are tested to focus on understanding the effects of fluid motion on surface tension measurements. To test the capability of
the computational algorithm on describing the hydrodynamics, computed and experimental velocity profiles are compared and found to be
consistent, as are the apparent surface tensions. Parameters such as forcing frequency, forcing amplitude, chamber dimensions, and surface
tension and viscosity of the liquid are varied to find the limits where pulsating bubbles depart from spherical and where hydrodynamic
effects impact the determination of surface tension. In the commercial pulsating bubble surfactometer (PBS), the bubble shapes remain nearly
spherical for low pulsation or oscillation rates (100 Hz) with moderate volume amplitudes. Under these conditions, however, two major
hydrodynamic effects, due to the inertia of the bulk liquid, or to the liquid viscosity and the viscous forces acting on the chamber walls, are
found to be important and can cause large errors in the surface tension measurements. For measurements of low surface tensions (5 mN/m)
in a PBS, oscillating bubble methods that do not take into account shape deformations from the spherical are found to be inaccurate, and the
results from dynamic bubble shape analysis (solving the YL equation for a nonspherical axisymmetric surface) are shown to provide another
approach to obtain accurate results provided that hydrodynamic effects can be neglected. At higher frequencies (200 Hz), because of strong
convection around the surface of the bubble, the bubble shapes become highly deformed. Further extensions of this algorithm are needed, to
include the surfactant diffusion and adsorption effects by which the surface tension may change with time.
2004 Elsevier B.V. All rights reserved.

Keywords: Supported bubble; Hydrodynamics; Surface tension measurements; Oscillating bubble method for tensiometry ; Finite element method; Free surface
flow; YoungLaplace (YL) equation; Pulsating bubble surfactometer (PBS)

1. Introduction port can be studied by measurements of DST. DSTs un-


der constant area conditions can be measured with various
The dynamic surface tension (DST) is important in many methods [29]. Some of these methods, such as those in-
applications involving lung surfactants [118], drop and jet volving drops or bubbles, can be also used under pulsat-
breakup [1924], foams [25,26], and coatings [27,28]. The ing area conditions. Then, DSTs and dynamic surface rheo-
adsorbed surface density of surfactant and the DST at an logical properties can be inferred [30] from measuring dy-
interface are related to each other [7]. Thus, interfacial trans- namic shapes or dynamic pressure differences under area
pulsation.
One of the most important and popular techniques for mea-
Corresponding author. Tel.: +1 765 494 4078; fax: +1 765 494 0805. suring DST is the oscillating bubble method [218,29]. A
E-mail address: franses@ecn.purdue.edu (E.I. Franses). commonly used and commercially available instrument that

0927-7757/$ see front matter 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.colsurfa.2004.04.101
368 Y.-C. Liao et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 250 (2004) 367384

uses this method is the so-called pulsating bubble surfactome- the bubble, its instantaneous dimensions and shapes can also
ter, or PBS, which can be used under constant or pulsating be directly imaged by a camera for surface tension determi-
area conditions. When DST is measured with a PBS, it is nation by axisymmetric drop shape analysis (or ADSA) [40],
normally assumed that fluid mechanical effects can be either if the bubble is sufficiently deformed by gravity, as it may
totally neglected [2] or estimated from simplified conditions occur for low surface tensions [13,17].
[3,8]. A major reason that hydrodynamic effects have not rig- The DST is obtained by measuring the pressure jump at
orously been assessed to date in making such measurements the interface. A pressure transducer is used for measuring
is because doing so would require the numerical solution of a the pressure close to the bottom of the chamber, pL (Fig. 1),
complicated free surface flow [31]. Such a rigorous approach with reference to the atmospheric pressure p0 (Fig. 1). If one
would entail the simultaneous determination of the a priori assumes that hydrodynamic effects are negligible, or uses the
unknown transient profile of the bubble and the flow field quasistatic assumption, then the pressure jump across the
in the irregularly shaped and continuously changing domain gas/liquid interface follows from hydrostatics:
outside the bubble. Recently, however, the closely related
problem of oscillations of pendant drops has been solved with p = ptip p0 = pL gh p0 , (1)
the finite element method (FEM) [3235]. In these studies, where ptip is the dynamic pressure of the liquid at the
the FEM used is based on a technique of domain discretiza- gas/liquid interface, is the density of the liquid, g is the
tion known as the method of spines [36]. This method has gravitational acceleration constant, and h is the vertical dis-
been shown to be highly accurate [31] but can be cumber- tance between the bubble tip and the pressure sensor. If the
some to use [35]. More recently, Notz et al. [37] and Chen bubble is spherical and the quasistatic assumption holds, the
et al. [38] have shown that such problems can be analyzed DST can be obtained from the pressure jump at the interface
with equal accuracy using an FEM algorithm in which the and the radius using the YoungLaplace (YL) equation:
domain discretization is more conveniently achieved with el-
ri (t)p(t)
liptic mesh generation [39]. Therefore, the goal of this paper (t) = , (2)
is to analyze rigorously for the first time the fluid mechani- 2
cal effects in a PBS by using techniques which are similar to where ri (t) is the bubble radius at time t.
those described in [35] and improved upon in [37,38]. For practical surface tension measurements [2] in the fre-
The schematic for a commercially available PBS is shown quency range of 1 to 100 rpm, Eq. (2), which is based on the
in Fig. 1 [2]. A small amount (about 30 l) of a liquid solu- quasistatic and perfect sphere assumptions, is fairly accurate
tion or suspension is placed in the sample chamber (Fig. 2), [8,10]. Since the bubble is small compared to the capillary
which has a square cross section of 1.76 mm 1.76 mm and length (2.7 mm for water), the bubble usually remains spher-
a length of 1 cm. A small capillary, with an inner diameter ical for 20 mN/m (Fig. 3(a)). Under certain conditions
of 0.58 mm, just protrudes into the chamber from its top, as of DSTs, however, the bubble can be deformed by gravity,
shown in Fig. 2. The capillary is open to the atmosphere at as shown in Fig. 3(b) [3,4,13,17]. Then, the assumption of
its other end. The sample chamber sits on a gasket, which a spherical shape is clearly invalid. Moreover, the deforma-
has a movable metal rod in the center. A bubble is gener- tion of the bubble can also be a result from normal viscous
ated by moving the rod downward and remains continuously stresses, Marangoni stresses, or possibly surface rheological
attached to, or supported by, the capillary. Normally, the effects.
bubble radius is forced to oscillate between 0.4 and 0.55 mm The problem of a pulsating spherical bubble in contact
by moving the bottom rod sinusoidally in time at a constant with an infinite amount of liquid, i.e. with no wall effects,
frequency , which can vary from 1 to 100 rpm (0.017 to has been addressed by several authors, either with no
1.7 Hz). The bubble area also oscillates nearly sinusoidally. hydrodynamic effects [2], or with hydrodynamic effects ap-
The bubble radius is normally obtained by pre-calibration, proximated [6,41,42]. If these effects are important, then the
with the bubble assumed to be spherical. During pulsation of dynamic pressure in the liquid at the gas/liquid interface, ptip ,

Fig. 1. Schematic of a pulsating bubble surfactometer. For a detailed photograph and schematic of the sample chamber, see Fig. 2.
Y.-C. Liao et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 250 (2004) 367384 369

Fig. 2. (a) Photograph and (b) schematic showing cross-sectional view and dimensions of the PBS chamber.

is given by the following approximate equation [8,41,43],


which is valid for radial flow in an infinite amount of liquid
(no wall effects),
2(t) 4 4s
ptip = p0 r i 2 r i , (3)
ri (t) ri ri
where is the bulk viscosity of the liquid, r i dri /dt, and s
is the surface dilatational viscosity. The second term of Eq. (3)
is the pressure jump due to the surface tension as indicated
by the static YL equation. The third term is the normal vis-
cous stress exerted by the bulk liquid. The last term results
from the surface dilatational viscosity s , the precise mean-
ing and magnitude of which remain controversial [44,45]. It
is generally presumed that for pure liquids, s = 0.
If the wall effects are neglected, the liquid pressure pL
far from a bubble undergoing sphero-symmetric oscillations
can be obtained by integrating the r-component (in spherical
coordinates) of the NavierStokes equation as [8,41]
 
2
pL = ptip ri r i + 23 r i + gh, (4)

where r i d2 ri /dt 2 . The second term of Eq. (4) is due to the


inertia of the liquid. The last term is due to the gravitational
force. Combining Eqs. (1)(4), the apparent dynamic surface
tension exp as estimated from the experiment is
 
2 3 2 2s
exp = (t) + 2r i +
ri ri + ri r i +
r i . (5)
2 2 ri
Hence, the measured dynamic surface tension exp is dif-
ferent from the actual value (t) due to viscous forces, the
inertia of the liquid, and possibly the surface dilatational vis-
Fig. 3. Pictures of: (a) a spherical bubble ( = 50 mN/m from pressure jump
cosity. From comparing the order of magnitudes, Chang and
method) and (b) a deformed bubble ( 2 mN/m from shape analysis) in
dipalmitoylphospatidylcholine (DPPC) dispersions at 1 Hz. These results Franses [8] concluded that only the surface dilatational vis-
have been obtained using systems and experiments that are similar to those cosity could be generally important, but that it has a minor
previously published [17]. effect for the standard PBS operation at frequencies lower
370 Y.-C. Liao et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 250 (2004) 367384

than 100 rpm. Although with Eq. (5) one can obtain a rough dilute dispersion, measured by the PBS under constant area
estimate for the DST measurement error, one neglects the conditions, was 70 mN/m.
wall effects and assumes that the velocity only varies radially. The pulsating bubble surfactometer (PBS), purchased
In actual measurements, the bubble oscillates in a confined from Electronetics Company (Amherst, NY), was used for
liquid, and the wall effect can lead to significant flow circu- surface tension measurements. The measurement procedures
lation and viscous stresses. To analyze these effects, a more have been described in Section 1. The bubble shape was moni-
rigorous model is needed. tored continuously by using a CCD camera (Model MN401X,
To understand the causes of bubble deformation and eval- from ELMO Corp., Plainview, NY) and the images were dig-
uate the reliability of DST measurements under these cir- itized with a frame grabber (Model ALL-IN-WONDER 128
cumstances, a three-dimensional but axisymmetric computa- PRO, from ATI Technologies Inc., Marlborough, MA).
tional algorithm is developed to elucidate the fluid dynamics The velocity fields around a pulsating bubble in water were
in the liquid outside the bubble. In the computations, the obtained by particle image velocimetry (PIV) [47], with the
sample chamber is simulated as a long cylinder (with a circu- polystyrene particles used as tracers. The images of the dis-
lar rather than a square cross section used in typical sample persed particles and the bubble were recorded by the CCD
chambers) within which the bottom rod is driven to oscillate camera every 1/30 s during the bubble pulsation. The veloc-
sinusoidally in time. The bubble is in contact with an incom- ity profiles at different times were obtained by tracking the
pressible Newtonian liquid and is open to the atmosphere as particles displacements between pairs of consecutive images
shown in Fig. 1. In this article, the bubble will be taken to using a correlation method as described by Gui and cowork-
oscillate in pure liquids of various surface tensions, to focus ers [4850].
on developing an understanding of the effects on DST mea-
surements of the forcing frequency, forcing amplitude, and
the presence of having a solid wall at a finite distance from 3. Mathematical formulation and solution method
the bubble. Algorithms have also been developed to analyze 3.1. Problem statement
situations where DST varies with the presence of insoluble
or soluble surfactants [46]. A discussion of surfactant and The liquid surrounding the bubble is incompressible and
variable DST effects is set aside in this paper, to focus on the Newtonian with constant density , viscosity , and surface
effects of liquid motion on the measurements in the complex tension 0 . Since the bubble is open to the atmosphere through
geometry of a PBS instrument. Conditions for low surface a capillary tube with a radius R = 0.29 mm and the viscos-
tension fluids are presented and discussed, in order to eval- ity of air is quite small [8], the pressure in the gas phase
uate the effect of bubble deformation on the measurements is assumed to be uniform and constant (equal to p0 ) during
without added complications due to dynamic adsorption or the pulsation. To avoid the further complication of a three-
surface convection and diffusion of surfactants. dimensional problem with the actual square cross section of
The paper is organized as follows. First, experimental de- the container, the sample chamber is taken to be a circu-
tails are presented in Section 2. The computational algorithm lar cylinder. Moreover, both the bubble shapes and the flow
for solving the set of two-dimensional evolution equations around the bubble are assumed to be axisymmetric. Thus, it
is presented in Section 3. The code based on the algorithm is convenient to adopt a cylindrical coordinate system with
is tested and validated in Section 4. To show the reliabil- its origin located along the centerline of the chamber (cf.
ity of the computations, the predicted velocity profiles and Fig. 4(a)). The governing equations and boundary condi-
the pressure jumps across the air/water interface, and the ap- tions in the following section have been normalized by using
parent surface tensions from computations are compared to as reference the tube radius R and the capillary time scale
those from measurements in Section 5. Then, the effects of 
tc R3 /0 . The dimensionless time, spatial coordinates,
the forcing frequency, the forcing amplitude, the chamber velocity and pressure, are
dimensions, and the viscosity and surface tension of the liq-
uid are examined with detailed simulations. In Section 6, the t r z vtc p p0
t , r , z , v , p ,
conclusions are presented. tc R R R /tc
(6)
2. Materials and experimental methods where r and z are the radial and axial coordinates, respec-
tively, v is the velocity, and p is the liquid pressure. In Eq. (6)
The water used for all samples was first distilled and then and what follows, variables without tildes over them are the
passed through a Millipore four-stage cartridge system, re- dimensionless counterparts of those with tildes.
sulting in a water resistivity of 18 M cm at the exit port. The dimensionless continuity equation and equation of
Dispersions of 2.5 wt% polystyrene particles of 6 m diame- motion for an incompressible Newtonian liquid in a chamber,
ter were purchased from PolySciences Inc., Warrington, PA. the walls of which remain stationary, are
The particles dispersion was diluted with water to 0.06 wt%
before it was used. The equilibrium surface tension of the v = 0, (7)
Y.-C. Liao et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 250 (2004) 367384 371

Fig. 4. Schematics for the computational domain (left) and mesh (right) used for the FEM computations.

v where vrod is the velocity of the rod. The rod performs time
+ v v = Oh T Gez ,
t periodic oscillations in the axial direction, so that the bubble
volume varies from the initial value V0 to a maximum value
T pI + [v + (v)T ]. (8)
Vmax . The instantaneous position of the rod, with respect to

In Eqs. (7), and (8), the Ohnesorge number Oh / R0 the fixed origin, is
is a measure of the relative importance of the viscous forces V0
to the inertial forces, i.e., it is the reciprocal of the Reynolds zrod (t) = [cos(2t) 1], (12)
2W 2
number. The Bond number G gR2 /0 measures the im-
portance of the gravitational forces relative to the surface where (Vmax V0 )/V0 is the volume pulsation ampli-
tension forces, and ez is a unit vector in the z-direction par- tude, tc is the dimensionless frequency, and W is the
allel to gravity. T is the stress tensor and it too is measured dimensionless chamber radius (Fig. 4). The velocity of the
in units of /tc . rod is
Initially, the bubble has a volume V0 and is at equilibrium V0
vrod = 2 sin(2t)ez . (13)
with a stagnant liquid, i.e., W
Along the moving and deforming liquid/gas interface or the
v=0 at t = 0. (9)
bubble surface S(t), the kinematic and traction boundary con-
The initial bubble shape can be determined from the static ditions are applied:
YL equation along with a volume constraint [51], and the
n v = n vs , (14)
pressure in the liquid is determined by hydrostatics.
Because of axial symmetry, the following boundary con- Oh n T = (s n)n, (15)
ditions are used along the centerline of the chamber:
where n is the unit normal to S(t) pointing into the bubble, vs
er T ez = 0 and v er = 0 at r = 0, (10) is the velocity of points on the interface, and s (I nn)
is the surface gradient operator [52]. At the gas/liquid/solid
where er is the unit vector in the radial direction. On the solid contact line LS , the bubble is assumed to be pinned (i.e.
wall boundary, liquid neither penetrates nor slips on SW . This not moving) at the tip of the capillary,
adherence condition requires that
 r=1 at LS . (16)
vrod , at the rod,
v = vwall = (11) The model described above is complete and involves no
0, elsewhere on SW ,
other simplifying assumptions or approximations. The evolu-
372 Y.-C. Liao et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 250 (2004) 367384

tion in time of the bubble shapes, and the velocity and pres- S(t):
sure fields in the surrounding liquid are governed by nine 
dimensionless groups. Two key dimensionless groups, Oh i (v vs ) n dA = 0. (21)
and G, provide measures of the importance of the viscosity S(t)
and gravity. The chamber dimensions are specified by the
dimensionless chamber radius W and the height H. The con- 3.3. Mesh generation and time integration
nection between the capillary and the sample chamber is a
circular ring and is described by two parameters W1 and H1 An elliptic mesh generation technique [39] is appropri-
(Figs. 2(b) and 4(a)), which are not varied in the following ately combined with an algebraic mesh generation technique
calculations. In the bubble pulsation, the initial bubble size to generate the computational grid and account for the pres-
V0 , the forcing frequency , and the forcing amplitude af- ence of the moving boundaries. As shown in Fig. 4(b), the
fect substantially the dynamic shapes and the liquid pressure computational domain is divided into seven regions. Regions
distribution. 1 and 2, which are far from the free boundary S(t), are tessel-
lated (divided into elements) using algebraic mesh generation
3.2. FEM analysis [31,32,34,55]. In these two regions, the element size in the
r-direction is held fixed, and that in the z-direction is varied
The governing equations are solved numerically by a with time in proportion to the distance between the instanta-
method of lines using the Galerkin finite element method neous location of the rod and a fixed axial coordinate, which
[53] for spatial discretization and an adaptive finite difference divides the domain into one part where the algebraic mesh
method for the time integration [54]. Under mixed interpola- is used and another part where the elliptic mesh is used (cf.
tion, the Galerkin weighted residuals of the continuity Eq. (7) Eq. (11)). Regions 37 are tessellated using elliptic mesh gen-
and momentum Eq. (8) are eration [3739]. Since the problem is solved at moving mesh
 locations, the time derivative v/t in Eq. (18) is evaluated
Ri,C = i v dV = 0, i = 1, . . . , M, (17) by accounting for the movement of the mesh:
V (t)
v
= v x v (22)
 t
Ri,r where v denotes the time derivative at a mesh point and x is
Ri,M =
Ri,z the velocity of this mesh point.

The Galerkin weighted residuals (1721) and the mesh
v
= i ek + v v Oh T + Gez dV generation equations result in a large set of nonlinear
V (t) t differential-algebraic equations in time. The time derivatives
= 0, i = 1, . . . , N, (18) are discretized with either the backward difference method
or the trapezoidal rule over a time interval tp = tp tp1 .
where ek is either er or ez and Ri,r and Ri,z denote the r- The final system of nonlinear algebraic equations is solved by
momentum and z-momentum residuals. Here, M is the num- Newtons method. Integrals that arise in the residuals and the
ber of bilinear basis functions i used in representing the Jacobian matrix are evaluated by Gaussian quadrature [53].
pressure Initially, four backward difference time steps with a fixed
M time step size t = 102 are used to provide the necessary

p(r, z; t) = pi (t)i (r, z), (19) smoothing upon setting the rod into motion from rest [56]. In
i=1
the subsequent steps, a second-order AdamsBashforth pre-
dictor is used with the trapezoidal rule to evaluate the relative
where pi is the nodal values of the pressure, and N is the time truncation errors and determine time step sizes adap-
number of biquadratic finite element basis functions i used tively. A relative error of 0.1% per time step is prescribed in
in representing the velocity the computations [54]. The maximum time step size tmax
 N
is set to be Tc /20, where Tc 1/. Furthermore, following
vr (r, z; t)
v(r, z; t) = = vi (t)i (r, z) Chen and Tsamopoulos [57] and Wilkes and Basaran [33],
vz (r, z; t) whenever the rod reverses its direction, two backward differ-
i=1
 ence steps with fixed t = 102 are taken before reverting
N
vr,i (t) to the use of the trapezoidal rule.
= i (r, z), (20)
vz,i (t)
i=1
4. Code validation
where vr,i and vz,i are the nodal values of the radial velocity vr
and axial velocity vz , and vi are the nodal values of the veloc- The code implementing the FEM analysis has been writ-
ity. Galerkin weighted residuals of the kinematic boundary ten in FORTRAN. The calculations have been carried out on
condition Eq. (14) are used to determine the bubble shape a LINUX cluster with PIII 550 MHz processors at Purdue
Y.-C. Liao et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 250 (2004) 367384 373

University. The code has been validated in many aspects, in- experiments, while the bubble is growing, the velocity has a
cluding strict mesh refinement tests in which the number of maximum in the vicinity of the bubbles tip. Moreover, the
elements is systematically increased. Thus, the solutions re- liquid velocity is nearly zero near the top corner of the cham-
ported in this paper are insensitive to the further increase in ber. While the bubble is shrinking, the maximum velocity
the number of elements used. In the calculations, the volume also occurs at the bubbles tip, and there is little liquid motion
of the liquid is monitored and always found to be conserved near the top corner of the chamber. The computed velocity
within 0.01%. Moreover, the correctness of the code for solv- fields show similar qualitative patterns and agree well with
ing the NavierStokes equations with a moving interface has the experimental data within 5% or better. Since the predicted
been verified by analyzing the oscillations of a free bubble in a velocity fields are accurate, the pressure fields obtained from
slightly viscous liquid [58,59]. As shown in the Appendix A, the same computations should be reliable, as incompressibil-
the calculated oscillation periods for small-amplitude oscilla- ity is satisfied well in the calculations (cf. Section 4).
tions agree to within an error of 103 with the analytically ob- The computed bubble shapes and pressure contours for the
tained periods of oscillation of bubbles in an inviscid liquid. conditions of the commercial instrument are shown in Fig. 6,
More than 200 sample computations for pulsating bubbles where the standard parameters are used: V0 = 10.28, =
were carried out to evaluate the effects of each dimensionless 1.74, H = 34.5 and W = 3.03. For a bubble of the given size
group on the surface tension measurements. For each com- in pure water, the Bond number G and the Ohnesorge number
putation, the code solves simultaneously more than 20,000 Oh are small, 0.011 and 0.0069, respectively, indicating neg-
nonlinear algebraic equations during each of the 24 Newton ligible gravitational forces and viscous forces in the liquid
iterations required at each time step. It takes about 20 h of (except near the solid/liquid boundary and the gas/liquid in-
CPU time to simulate ten cycles of bubble oscillations. terface). Thus, the bubble shapes should remain nearly spher-
ical under these conditions. One measure of the bubble shape
5. Results and discussion is the shape factor
ra
5.1. Experimental results and tting to the model fs , (23)
ri
The computed and measured flow fields when a bubble is where ra is the maximum radial position of the liquid/gas
pulsating in pure water at 1 Hz are compared in Fig. 5. In the interface (Fig. 6) and ri is the volume-average radius of

Fig. 5. (a) Measured and (b) computed velocity fields for a pulsating bubble in pure water at 1 Hz at time (1) t = 0.15 s after the start of the pulsation, bubble
expanding, and (2) t = 0.75 s, bubble shrinking. In this and similar figures, each velocity vector shown indicates the velocity at its base point.
374 Y.-C. Liao et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 250 (2004) 367384

Fig. 6. A pulsating bubble surrounded by water at 1 Hz in a standard PBS


chamber: (a) computed shapes at different times, as indicated, during one Fig. 7. Computed shapes for pulsating bubbles surrounded by water in a
period of oscillation; (b) selected pressure contours when t = 0.25; 100 standard PBS chamber at different times, as indicated, during one period of
dimensionless pressure units corresponds to 1.75 cm H2 O or 1.29 mm Hg. oscillation at a forcing frequency of: (a) 100 Hz, (b) 200 Hz, (c) 500 Hz, (d)
1000 Hz.
direction, and the pressure gradient dp/dz is constant at any
the pulsating bubble. When the bubble is spherical, fs = 1.
time instant (Fig. 6(b)). When the bubble pulsates at 1 Hz, the
When fs > 1 or fs < 1, the bubble is an oblate or a prolate
velocity profile in the central part of the chamber is parabolic
spheroid, respectively. The maximum and minimum values
(Fig. 8(a)). When > 1 Hz, the velocity profiles become ra-
of fs for a pulsating bubble surrounded by water in a stan-
dially more uniform and close to that of plug flow with a
dard PBS chamber at various frequencies are listed in Table 1.
thin boundary layer near the wall (Fig. 8(c)). This behavior
For 100 Hz, |fs 1| < 0.04, indicating that the bubble
indicates that inertial effects become more pronounced as the
shapes are close to spherical. Thus, at these conditions, the
forcing frequency increases, because as increases the time
assumption of a spherical bubble surface for the standard
scale for vorticity diffusion from the side wall of the cham-
PBS measurement is fair. For > 100 Hz, the bubbles can
ber becomes long compared to the period of oscillation of the
be deformed substantially (Fig. 7). For = 1000 Hz, in the
rod. The apparent surface tension is determined in the calcu-
first cycle when the rod reverses its direction (t = 0.5), the
lations similarly as in the measurements. First, the pressure
fast fluid motion near the centerline causes a dimple at the tip
sensor is assumed to be located in the center of the bottom
of the bubble, and the dimple gradually forms an overturn.
rod. After correcting for the hydrostatic pressure, the dimen-
When the second cycle begins, the rod changes its direction
again. The fluid around the bubble moves downward, except
Table 1
for the overturned portion. This flow pattern can eventually Shape factors for a pulsating bubble surrounded by water (Oh = 0.007,
lead to a satellite drop formation (not shown here). G = 0.011) in a standard PBS chamber
At low frequencies, 100 Hz, since the standard cham- (Hz) fs,max fs,min
ber is a long slender cylinder, far from both ends the flow is
1 1.003 1.000
unidirectional and the radial velocity is zero, except during 10 1.003 1.000
the short periods of time when the rod reverses direction (see 20 1.003 1.000
Section 5.5); an example is given in Fig. 8. The flow profile 50 1.007 1.000
in the central part of the chamber is quite similar to Poiseuille 100 1.037 0.969
flow in a circular tube: the pressure varies only along the z- 200 1.709 0.873
Y.-C. Liao et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 250 (2004) 367384 375

Fig. 8. Velocity fields in the central part of a standard PBS chamber for t = 0.25 at different forcing frequencies : (a) 1 Hz, (b) 10 Hz, and (c) 100 Hz. The
flow is essentially unidirectional. With increasing frequency, the axial velocity distribution becomes more radially uniform with a thin boundary layer near the
wall.

sionless calculated pressure jump across the interface, if the


hydrodynamic effects are neglected, is (cf. Eq. (1))

G
p(t) = prod + (zrod ztip ), (24)
Oh

where ztip is the axial position of the bubbles tip. From the
YL equation for a spherical interface, the apparent dimen-
sionless surface tension is (cf. Eq. (2))

ri (t)p(t)
app = Oh . (25)
2

For a bubble pulsating in water at = 1 Hz, the computed and


preset radii overlap (Fig. 9(a)). The radii measured from the
images, shown as square symbols, differ from the preset radii
by less than 5%. The computed pressure differences p dif-
fer by less than 5% from the measured values. Although the
computed apparent surface tension app 0 app is nearly
constant (0.5 mN/m), the measured surface tension shows
a larger amplitude of 4 mN/m. This experimental error is
due to the difference between the actual and the preset stan-
dard radii. This difference is caused by the movement of the
pressure sensor, and has been identified before [2]. While
the bubble is growing (Fig. 1), the increase of pL makes the
sensor to move toward the meter and causes some liquid to
leak out of the sample chamber because of the incomplete
seal. Therefore, the bubble becomes larger than the preset Fig. 9. Comparison between computations and experiments: (a) radii mea-
value and the measured pressure is larger than the value at sured from experimental pictures (), preset radii curve (solid line), and
the preset bubble size, resulting in a smaller apparent surface predictions from simulations (dashed line); (b) measured (solid line) ver-
sus predicted (dashed) pressure difference (p in Eq. (1)); (c) comparison
tension. While the bubble is contracting, the sensor moves of apparent surface tensions calculated from measured p and preset radii
away from the meter and some liquid below the rod is pushed curve (solid line) to the values calculated from predicted p and predicted
into the chamber, making the bubble smaller and resulting in radii (dashed line).
376 Y.-C. Liao et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 250 (2004) 367384

be safely neglected. If one ignores viscous forces and assumes


that all of the liquid in the chamber accelerates exactly as the
rod, the z-component of the NavierStokes equation becomes
vz dvrod V0 p
= 22 2 cos(2t) = Oh G.
t dt W z
(27)

Integrating this equation along the z-direction and assuming


that it holds for zrod < z < ztip , one obtains the following
expression for the dimensionless pressure difference between
the tip of the bubble and the rod
 
ztip zrod 2 V0
ptip prod = 2 cos(2t) G .
Oh W2
(28)

Fig. 10. The effect of forcing frequency on the dimensionless apparent


For a spherical bubble pulsating with a time dependent radius
surface tension for a bubble pulsating in water. Standard parameters are ri (t), the pressure at the bubble tip is approximately [8]
used: Oh = 6.92 103 , G = 0.012, V0 = 10.28, = 1.74, W = 3.03,
and H = 34.5. Values of used are indicated in the legend in the figure.
2 4ri
ptip . (29)
For 20 Hz, app starts from the unphysical value of 1.05 or more, and ri Oh ri
can range between 1.22 and 0.65 (8847 mN/m) due to hydrodynamic ef-
fects. Combining Eqs. (28) and (29) yields:
2 4ri ztip zrod
a larger apparent surface tension. This effect is more pro- prod =
nounced when the pressure pL varies over a wider range [2]. ri Oh ri Oh
 
2 V0
2 cos(2t) G . (30)
5.2. Effects of inertia on surface tension measurements W2
for low-viscosity uids
The apparent surface tension can be obtained by combining
Eqs. (25) and (30) as follows:
5.2.1. Effect of the forcing frequency on the apparent
surface tensions ri
app = [Oh prod + G(zrod ztip )]
When the bubble pulsates at = 1 Hz ( = 5.82 104 ), 2
app is virtually constant and equals to the theoretical value V0
of 1 (Fig. 10). For = 20 Hz, app starts with an unphysical = 1 + 2 Oh ri + ri 2 2 (ztip zrod ) cos(2t).
W
initial value of 1.04 and varies between 1.04 and 0.94. This
(31)
fluctuation amplitude increases with increasing frequency.
Thus, to have an accurate standard PBS measurement with a The apparent surface tension in Eq. (31) is comprised of three
maximum error less than 5%, one needs to use frequencies parts: the theoretical value of 1 (app = 0 ) without any hy-
of less than 20 Hz. The unphysical initial value and the fluc- drodynamic effects, the effects of the normal viscous stresses,
tuation of app are plainly artifacts due to the hydrodynamic and the effects of the inertial forces in the bulk liquid. The
effects. dimensionless interfacial velocity ri is of the same order as
. For conditions in which the spherical shape assumption
5.2.2. Approximate apparent surface tensions for holds, Oh and or ri are much smaller than unity; hence, the
low-viscosity uids normal stress effect on app is very small (104 for water).
Since the velocity field in the central part of the sample The second term in Eq. (31) is the same as the second (vis-
chamber is unidirectional (Fig. 8), the continuity equation cous normal stress) term in Eq. (5). The third term (hereafter
and the r-component of the NavierStokes equation lead to called inertial deviation) in Eq. (31) is different from the third
(inertia) term in Eq. (5). This is because Eq. (5) deals with a
vz p
=0 and = 0, (26) radial flow with a different geometry from that in the PBS.
z r The inertial deviation, is out of phase with the pulsating
indicating that the axial velocity does not vary along the z- radius ri , i.e. for a minimum (t = 0) radius or a maximum
direction (Fig. 8) and the pressure depends only on z and t (t = 0.5) radius, the apparent surface tension app from
(Fig. 6(b)). Moreover, for low-viscosity liquids, such as water, Eq. (31) shows instead a maximum or a minimum, respec-
the Ohnesorge number is quite small, indicating that viscous tively. The amplitude of the inertial deviation decreases as the
stresses away from the solid walls and the fluid interfaces can chamber radius W increases or as the forcing frequency , the
Y.-C. Liao et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 250 (2004) 367384 377

Fig. 11. Comparison of apparent surface tensions app from computations


Fig. 12. Dimensionless apparent surface tension app of an oscillating bubble
(solid lines) and approximated (dotted lines) values from Eq. (31) with the
at a frequency of 100 Hz for various forcing amplitudes = V/V0 ; other
normal stress term set to zero at: (a) 100, (b) 50, and (c) 20 Hz. Parame-
parameters used are the same as in Fig. 10. For the large amplitude = 1.74
ters used are the same as in Fig. 10. The approximated app values overlap
(solid line), app ranges from 0.3 to 2, indicating major hydrodynamic
with the computed values, indicating that the inertial forces dominate the
effects and a major failure of the quasistatic assumptions. For = 0.87
hydrodynamic effects.
(dashed line) and 0.435 (dotted line), app remains positive. The amplitude
of app decreases as decreases.
volume amplitude , or the chamber height H (ztip zrod )
decrease.
is long and narrow (W < 3 mm), the predicted variations in
Since H is much larger than 1 because a long slender cham-
app are quite close to the values approximated from Eq. (31).
ber is used, the inertial deviation is quite large: the maximum
If one uses a longer chamber, the fluctuation of app is larger
value of the inertial deviation for a bubble pulsating in water
because of increased inertial effects, as shown in Eq. (31).
at 20 Hz in a standard chamber is around 0.05. Moreover,
When W increases, the fluctuation of the computed app be-
Eq. (31) makes plain that at t = 0, or at the beginning of each
comes smaller, indicating that the wall effects are smaller. For
cycle, app starts from a value greater than one and increases
H = 1 cm, the maximum and minimum app reach plateaus
with frequency , as observed in Fig. 10. When Eq. (31) is
of 1.05 and 0.95, respectively, for W 6 mm. By contrast,
applied without the normal stress term for 20 Hz, the
Eq. (31) predicts values (solid lines in Fig. 13) that tend to
approximate and computed solutions are almost the same
1.0 as W grows without bound.
(Fig. 11), indicating that inertial forces, in accord with intu-
This discrepancy between the rigorous computational re-
ition, dominate when the liquid has a low viscosity.
sults and the approximate results from Eq. (31) can be ex-
plained as follows: as the chamber gets wider, the radial flows
5.2.3. Effect of the forcing amplitude around the bubble become more pronounced and the liquid
As the forcing amplitude increases, the dimensionless motion becomes close to that of a bubble pulsating in an in-
apparent surface tension app can deviate greatly from 1.0 finite amount of liquid (this case was also examined in the
(Fig. 12). Indeed, for the standard value of = 1.74, app can simulations, but is not shown here). Then, Eq. (31), which is
even be negative for = 100 Hz, as shown in Fig. 12. When derived based on a unidirectional flow assumption, is inaccu-
decreases, the deviation of app from 1.0 decreases sig- rate. For a pulsating bubble in an infinite expanse of liquid,
nificantly, indicating a major reduction of the hydrodynamic the apparent surface tension is given by Eq. (5) [8], or in
effects and of the artifacts in the surface tension measure- dimensionless form
ments. The deviation amplitude of app is nearly proportional  
to (app,max 1 = 0.9, 0.45, and 0.23 for = 1.74, 0.87, app = 1 + 2Oh ri + 21 ri2 ri + 23 ri ri2 . (32)
and 0.425). The linear dependence is consistent with the ap-
proximate equation, Eq. (31). Hence, hydrodynamic effects The maximum and minimum values of app from Eq. (32)
in aqueous solutions are minimal for very small forcing am- (dotted lines in Fig. 13) are the limits for no wall effects, or
plitudes, as would be expected. for W . Thus, as W (or W) rises, the predicted fluctua-
tion in app approaches the values from Eq. (32). The larger
5.2.4. Effects of chamber dimensions the value of H, the larger is the value of W needed to reach
The predicted maximum and minimum of app of a bubble the limit of an infinite amount of liquid. These results may
pulsating in water at 100 Hz in chambers of different radius provide useful guidelines for designing improved versions of
W and height H are shown in Fig. 13. For a chamber which supported bubble instruments.
378 Y.-C. Liao et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 250 (2004) 367384

Fig. 13. Effects of chamber radius W and height H on the dimensionless Fig. 14. (a) Comparison of apparent surface tensions app for a bubble pulsat-
apparent surface tension app of an oscillating bubble at a forcing frequency ing at 10 Hz for = 1100 cP, as indicated in the figure. The parameters used
of 100 Hz. Other parameters used are the same as in Fig. 10. The lines are the are Oh = 6.92 103 for 1 cP, G = 0.012, = 5.82 103 , V0 = 10.28,
maximum app,max or minimum app,min of dimensionless apparent surface = 1.74, H = 34.5, and W = 3.03. As the viscosity increases, app deviates
tension predicted from Eqs. (31) and (32), as indicated in the figure. The open more from 1; the maximum deviation does not occur at the maximum radius
and solid symbols are computed values of app,max and app,min , respectively, (t = 0.5). (b) The phase angle difference between app and the volume-
for different W and H, as indicated in the figure. For a narrow chamber averaged radius ri in figure (a). As the viscosity increases, for this particular
(W < 3 mm), the variation of the dimensionless apparent surface tension set of parameters used here, the phase angle difference decreases and reaches
app is large and follows the prediction of Eq. (31). As the chamber gets a plateau at a value of 82 at 200 cP.
wider, the variation gets smaller and gradually approaches the prediction of
Eq. (32). Moreover, the higher the H, the larger is the fluctuation. For details,
see text.
When bubbles oscillate in a high viscosity liquid ( =
100 cP), the velocity fields in the central part of in standard
5.3. Effects of viscosity PBS chamber are unidirectional (Fig. 15), similarly to those
in a low viscosity liquid (Fig. 8). In the latter, the axial ve-
The critical values of the forcing frequency c , below locity distribution changes from a parabolic profile into a
which the bubble shape factor fs remains 1 0.05 during nearly uniform profile as the forcing frequency increases. In
the pulsation, for different values of Ohnesorge number Oh, the former, the axial velocity distribution remains parabolic,
are listed in Table 2. For Oh < 0.692 ( < 100 cP), c is with magnitudes varying with frequency, indicating that the
0.0582 (c = 100 Hz), and it drops to 0.00582 (c = 10 Hz) momentum transport in these flows is quite efficient. These
for Oh = 6.92 ( = 1000 cP). This decrease in c shows parabolic flows can be approximated by the HagenPoiseuille
that the viscous forces become an important factor in DST equation. For mass conservation, the average velocity is the
measurement in high viscosity liquids, and they may even same as the rod velocity vrod :
cause the bubble to become nonspherical. With increasing
 
liquid viscosity, the deviation of app from 1.0 for a bubble r2
pulsating at = 0.00582 ( = 10 Hz) is more pronounced vz = 2 vrod 1 2 . (33)
W
(Fig. 14(a)). Unlike the inertial effects for low-viscosity liq-
uids (Section 5.2), for which app starts from more than 1.0
at the beginning of each cycle, for high viscosity liquids, app Table 2
starts from 1.0. Moreover, the app curves for viscous liquids Critical frequencies for having essentially spherical pulsating bubbles when
are not in phase with the pulsating bubble radius ri . The phase they are surrounded by liquids of different viscosity in a standard chambera
angle difference between the maxima of app (t) and ri (t) de- Oh (cP) c b c (Hz)
creases gradually from 180 for = 1 cP to 90 for 100 cP, 0.007 1 5.82 102 100
and reaches a plateau of 82 for 200 cP. This is the first 0.035 5 5.82 102 100
report of such a phenomenon, which is not due to DST ef- 0.173 25 5.82 102 100
fects. The phase angle shifting shows a transition from the 0.692 100 5.82 102 100
3.46 500 2.91 102 50
inertia-dominated region (phase angle difference equals to 6.92 1000 1.16 102 20
180 ) to a viscous-force-dominated region, as explained be- a Parameters used are G = 0.012, = 1.74, V = 10.28, H = 34.5, and
low. The value at the plateau, here 82 , may vary depending
0
W = 3.03.
on the set of conditions considered. b : the dimensionless frequency below which |f 1| < 0.05.
c s
Y.-C. Liao et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 250 (2004) 367384 379

Fig. 15. Velocity fields around a pulsating bubble in a high viscosity fluid ( = 100 cP) for t = 0.25 at: (a) 1 Hz, (b) 10 Hz, and (c) 100 Hz. The parameters
used are Oh = 0.692, G = 0.012, V0 =10.28, = 1.74, H = 34.5, and W = 3.03. The flow is essentially unidirectional in the central part of a standard PBS
chamber. The velocity profiles are parabolic with magnitudes varying with the forcing frequency.

For this equation, the corresponding pressure gradient in the Eq. (36) and ri is about 82 , which is the same as the plateau
central part of the chamber is value for viscosity 200 cP in Fig. 14(b). The effects of
, , W, and H on the viscous deviation are similar to those
dp vrod G
= 8 2 , (34) discussed in Section 5.2.
dz W Oh When the approximate Eq. (36) is applied to a bub-
The pressure difference between the bubble tip and the rod ble pulsating at 10 Hz, the approximate and computed
can be obtained by integrating the above equation: app are almost the same for Oh = 0.692 ( = 100 cP)
  (Fig. 16), indicating that the pressure drop due to the viscous
vrod G
ptip prod = 8 2 + (ztip zrod ). (35) forces dominates the hydrodynamic effects. For Oh = 0.173
W Oh
From Eqs. (25), (29), and (35), the apparent surface tension
is
V0
app = 1 + 2 Oh ri + 4 Oh ri (ztip zrod ) sin(2t).
W4
(36)

As mentioned in Section 5.2.2, the first term is the theoreti-


cal value without hydrodynamic effects, and the second term
results from the normal viscous stress, which is of the order
of 0.01 for Oh = 0.7 (100 cP) and  0.06 (100 Hz). The
last term in Eq. (36), unlike that in Eq. (31), results from the
pressure drop caused by viscous forces acting on the cham-
ber walls, and is called the viscous deviation term hereafter.
Since the chamber is slender, the term ztip zrod H is large
and the viscous deviation results in a larger variation for app
than the variation resulting from the normal stress close to
the bubble. The amplitude of the viscous deviation decreases Fig. 16. Comparison of apparent surface tensions app from computations
with decreasing Ohnesorge number Oh, forcing frequency (solid lines) and approximated values (dashed lines) from Eq. (36). (a) For
100 cP, the approximate app values overlap with the computed ones, indi-
, volume amplitude , and chamber height H, or with in-
cating that the viscous deviation dominates the hydrodynamic effects. (b)
creasing chamber radius W. When the viscous deviation is For 25 cP, the approximate app fits the computed app fairly well but with a
the only significant hydrodynamic effect, for this particular small phase angle difference, indicating that both viscous and inertial effects
set of parameters used, the phase angle between app from are important.
380 Y.-C. Liao et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 250 (2004) 367384

Fig. 17. Effect of surface tension on the deformation of pulsating bubbles


at 1 Hz in simulations of the PBS chamber at different time stages. Standard
operating parameters are used: = 1.74, V0 = 10.28, H = 34.5, and W =
3.03. The time difference between two adjacent computed shapes is 0.05 s.
The apparent surface tension from calculations is shown in the figure as an
average with a fluctuation. The surface tensions calculated from the dynamic
shapes using the YL equation for nonspherical surfaces are the same as the
input surface tensions 0 .

Fig. 18. Comparison of dynamic shapes (solid lines) and static shapes
( = 25 cP), although app from Eq. (36) and the computed (dashed lines) for 0 = 5 mN/m (Fig. 17). The static shapes are obtained by
solving the YL equation for static bubbles having the same volume as the
values are similar, they show a slight phase lag. Hence, dynamic shapes. The dynamic and static shapes agree well, and they cannot
hydrodynamic effects other than simply the viscous devia- be distinguished in the figure, indicating that the surface tension measure-
tion are significant. For liquids of intermediate viscosities ments from the ADSA analysis can be accurate so long as hydrodynamic
(0.05 < Oh < 0.2), the fluctuation in app results from both effects are negligible.
inertial and viscous forces, and should be obtained from the
rigorous numerical solutions. highly deformed oblate spheroidal shapes and are slightly
overturned (the bubble shape has a vertex higher than the
5.4. Effects of surface tension contact line) at the edge of the capillary at the onset of the
pulsations. The computation actually had to be stopped at
The time evolution of bubble shapes under standard PBS t = 0.35 because the strong overturning caused the bubble
conditions at 1 Hz is shown in Fig. 17 for four values of to touch the chamber wall. In the actual experiments, such
the surface tension 0 . For 0 = 72 mN/m, the Bond Num- touching of the wall by the bubble surface may cause a
ber is small, G = 0.01, and the dynamic shapes are close contact line to move or jump, and the measurement would
to spherical (|fs 1| < 0.01). The apparent surface ten- be expected to fail. To prevent such situations, one should
sion app , as computed from the pressure difference p use a smaller forcing amplitude or a smaller capillary
and the volume-averaged radius ri , is the same as the input radius R.
value of 72 mN/m with a variation of about 0.5 mN/m. For For the case of 0 = 5 mN/m in Fig. 17, for which gravity
0 = 20 mN/m, the dynamic shapes start to deform slightly can deform the bubble even with no flow, the dynamic shapes
into oblate spheroids due to the larger gravitational effect at different time stages agree well with the corresponding
(G = 0.04), which causes the pressure jump at the tip of the static shapes (Fig. 18), which are obtained from solving the
bubble to be smaller than that of a spherical shape. The aver- YL equation for a static bubble of the same volume. Thus,
age computed app is 19.3 mN/m, or about 3% smaller than when one uses axisymmetric drop shape analysis (ADSA)
the input value of 20 mN/m. The deformation of the dynamic [40], for low surface tension (G > 0.1) measurements, the
bubble shape is more severe when 0 = 5 mN/m. The fluc- accuracy is higher than using the pressure jump method for
tuation of app can be as large as 10% of the actual value, a spherical interface. These results apply, of course, when
and the tension measurement error is 20%. For still lower the forcing frequency is small and the bubble shape is not
surface tension, 0 = 2 mN/m, the bubble shapes start from distorted by the fluid motion (cf. Fig. 7).
Y.-C. Liao et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 250 (2004) 367384 381

Fig. 19. Streamlines at different times in the first oscillation cycle for a pulsating bubble in water at = 0.00582 or = 10 Hz. Parameters used are Oh =
0.00692, G = 0.012, = 1.74, V0 = 10.28, H = 34.5, and W = 3.03. The eddy near the corner of the chamber at t < 0.5 grows when the bubble increases
in size, and at t = 0.5 occupies the whole chamber. When the rod reverses its direction, the eddy vanishes; after the passage of certain time, the eddy forms
again but circulates in a different direction. This pattern is repeated in each cycle.

Fig. 20. Streamlines at t = 0.5 for bubbles pulsating at 10 Hz in liquids of increasing Ohnesorge number or bulk viscosity. Parameters used (except Oh as
indicated in the figure) are the same as those in Fig. 19. The vorticity is stronger in higher-viscosity liquids and can cause the formation of several counter
rotating eddies.
382 Y.-C. Liao et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 250 (2004) 367384

5.5. Recirculating eddies and streamlines For a pulsating bubble in the standard PBS chamber
to remain nearly spherical, which is one essential assump-
For a pulsating bubble in water at 10 Hz, a recirculation tion of the PBS measurements, the forcing frequency has
zone in the liquid is predicted to develop near the corner of the to be smaller than 0.06 (or < 100 Hz) for Oh 0.7 (or
chamber. The eddy grows when the bubble expands (Fig. 19), 100 cP). The apparent surface tension for a liquid with a
and can occupy the whole chamber at t = 0.5 at which in- constant surface tension from the pulsating bubble surfac-
stant the rod velocity is zero. Although the circulation appears tometer could still vary with time, even when the bubble
to dominate the whole chamber volume, the magnitude of is nearly spherical, because of significant hydrodynamic ef-
the maximum velocity is small ( 2 103 ). Hence, these fects.
circulations affect only slightly the surface tension measure- For pulsating bubbles with nearly spherical shapes in low-
ments. When the rod starts moving in the opposite direction, viscosity liquids (Oh < 0.05), such as water, hydrodynamic
the eddy does not arise for t ranging from 0.55 to 0.7. When effects cause errors in surface tension measurements, mostly
the eddy reappears, the liquid circulation is in the opposite because of the inertial forces in the bulk liquid. The inertial
direction. The eddy grows in size as the rod slows down and deviation, app 1, is out of phase with the volume-averaged
ultimately disappears as the rod returns to its starting position radius ri . At the beginning of each pulsation cycle, app starts
(t = 1.0). This pattern is repeated in each cycle. with an unphysical value of greater than 1.0. The inertial de-
For higher viscosity, the circulation pattern at t = 0.5 viation term is proportional to the forcing amplitude , the
causes several eddies to form faster (Fig. 20). The pressure square of the forcing frequency , and the chamber height H,
fields are affected more significantly the higher the viscosity and inversely proportional to the square of the chamber radius
is. For Oh = 6.92 ( = 1000 cP), the eddies near the bubble W. For the commercially available experimental setup with
surface seem to cause large normal stresses which might de- a preset forcing amplitude of = 1.74, frequencies lower
form the bubble. Thus, the effects of the eddies on the surface than 10 Hz are predicted to give accurate measurements. For
tension measurements can generally be significant, unless the accurate measurements at higher frequencies (>10 Hz), to
liquid viscosity is low. diminish inertial effects, one needs to either reduce the forc-
ing amplitude or use a wider chamber. For W > 21 (or
W > 6 mm), app varies as app of a pulsating bubble in an
6. Conclusions infinite amount of liquid, as has been modeled before [8].
For nearly spherical pulsating bubbles in high viscosity
The problem of a bubble that is supported by a small capil- liquids, errors in app are due mostly to the pressure drop
lary and oscillates in a Newtonian liquid with a constant sur- arising from the viscous shear stresses exerted by the liquid
face tension is solved for the first time with the complete set on the chamber walls. The error due to the viscous forces in
of hydrodynamic equations by using a rigorous finite element app , or the viscous deviation term, for liquids in a standard
method. The only significant simplification in the computa- PBS sample chamber, has a phase angle difference of 82 ,
tions is the use of an axisymmetric container with a circular unlike the phase angle difference of 180 of the inertial devi-
rather than a square cross section in the actual experimental ation term. The viscous deviation term is proportional to , ,
conditions. The model contains nine dimensionless groups. and H, and inversely proportional to W 4 . The phase angle dif-
The Ohnesorge number Oh and Bond number G provide ference between app and ri decreases gradually from 180 to
measures for the importance of viscosity and gravity, respec- 82 with increasing Oh, or viscosity . For Oh > 0.7, the vis-
tively. The chamber is a long cylinder with a dimensionless cous deviation dominates the hydrodynamic effects for app .
radius W and a height H. The chamber is open to the atmo- For intermediate viscosities (0.05 < Oh < 0.7), the fluctua-
sphere with a capillary. The connection between the capillary tion of app results from both inertial and viscous forces and
and the sample chamber is a circular ring and is described by has to be evaluated numerically.
two parameters W1 and H1 , which are kept constant in the For the commercial PBS, which has a maximum frequency
calculations as in the standard sample chamber. Three op- of 1.7 Hz, the dynamic bubble shapes remain nearly spherical
erational parameters, the initial bubble size V0 , the forcing for liquids with surface tension 0 greater than 20 mN/m. For
frequency , and the forcing amplitude , are needed for low frequencies and nearly spherical bubble shapes, errors
describing the bubble pulsation. The computed flow fields in PBS measurements are less than 5% for 0 20 mN/m.
agree fairly well with the measured flow fields from particle At lower surface tensions (0 5 mN/m), pulsating bubbles
image velocimetry. Moreover, the calculated pressure jump can and will be deformed by gravity. Thus, the apparent sur-
p and apparent surface tension are fairly close to the mea- face tension from PBS, which is based on the static YL
sured values. The consistency between the computations and equation for spherical bubbles, is less accurate. With ax-
the experiments indicates the correctness and accuracy of the isymmetric drop shape analysis (ADSA) one will be able
current model and the FEM algorithm used, which are used to to determine accurately the surface tension for 0 5 mN/m
evaluate the measurement errors over wide parameter ranges, when the hydrodynamic effects can be safely neglected. For
and also the relevance of the computations to actual measure- 0 2 mN/m, the bubble deforms more substantially and
ments. could touch the chamber walls during the pulsation. Then
Y.-C. Liao et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 250 (2004) 367384 383

the measurement is expected to fail unless a smaller volume


amplitude or a smaller capillary radius is used.
The hydrodynamic effects on the surface tension measure-
ments with the pulsating bubble surfactometer for liquids of
constant surface tension are elucidated thoroughly. The rig-
orous finite element algorithm used here is found to describe
accurately the hydrodynamics with a complex geometry, a de-
formed bubble shape, and a moving interface. The algorithm
can be modified to include time-varying dynamic surface ten-
sions (DST) due to surfactant adsorption and surface con-
vection and diffusion. Then DSTs can be determined more
reliably than done previously by a better accounting of the
hydrodynamic effects. Before any conclusions can be drawn
for surface rheological properties of aqueous surfactant or
protein solutions by extracting DST data from pulsating bub-
ble experiments, the hydrodynamic effects must be fully ac-
counted for as in this article. This problem is addressed in
Fig. A.1. Schematic of a slightly deformed free bubble pulsating in a low-
[46]. viscosity liquid. The dashed line is a sphere of the same volume as the bubble
before the deformation.

Acknowledgements where is the polar angle. The frequencies of infinitesimal-


amplitude oscillation .  1 have been determined by
This research was supported in part by grants from the Rayleigh [58] (see also [59]) using a linearized analysis. For
National Science Foundation (Grant CTS 0135317) and the a bubble oscillating in an inviscid liquid, the linearized fre-
National Institutes of Health (Grant HL 54641-02) to Elias quencies of oscillation n0 are given by
I. Franses, and grants from the Basic Energy Sciences Pro-
gram of the US Department of Energy (Grant DE-FG02- n0 = [(n2 1)(n + 2)]1/2 , n = 2, 3, . . . (38)
96ER14641) and the Purdue Research Foundation to Osman The values of n0 and the periods of oscillation T 2/n0
A. Basaran. The authors would also like to thank Dr. Lichuan for n = 2, 3, and 4 are listed in Table A.1.
Gui and Professor Steve Wereley for discussions and help in Here, to test the code, the problem of a slightly deformed
the PIV measurements. We offer special thanks to Dr. Goran (. = 0.01) bubble in a low-viscosity fluid, Oh = 103 , is
Enhorning, and to Electronetics Corporation (Now General solved numerically. In the calculations, the infinite liquid do-
Transco Inc.), for the design of the commercial PBS instru- main has been truncated at a finite radius r = Rout (Fig. A.1)
ment, which allowed us and many others to obtain many valu- and the liquid far from the bubble surface, i.e. for r Rout ,
able data and inspired us to develop this and related models. is taken to be at rest or v = 0. The position of the bubble tip
We also thank Drs. Pat Notz and Alvin Chen whose pioneer- r( = 0, t), which executes sinusoidal oscillations, is plotted
ing uses of elliptic mesh generation in studies of oscillating against time, and the period of oscillation T is obtained by
and breaking drops were instrumental in expediting the de- the difference between successive times for which r( = 0, t)
velopment of the present algorithm. equals unity.
The resulting T for different modes of oscillation n are
listed in Table A.1. For n = 3, elements of the same size
Appendix A. Code validation for a bubble (N = 15; Nr = 5, 20 and 45 for Rout = 2, 5, 10; where N
undergoing small-amplitude oscillations in a
low-viscosity liquid Table A.1
Oscillation time periods of deformed bubbles with a small deformation of
An important accuracy check for the algorithm is pro- . = 0.01 in a low-viscosity fluid of Oh = 103
vided by analyzing the small-amplitude oscillations of a free n n0 T 2/n0 Rout N a Nr b Tcalculated
bubble immersed in a low-viscosity liquid. The computed 2 3.464 1.814 5 15 20 1.813
predictions are then compared to theoretical analyses of lin-
3 6.325 0.993 2 15 5 1.000
earized oscillations of a bubble in an inviscid liquid of infi- 2 30 10 1.000
nite expanse [58,59]. Initially, the bubble is perturbed from a 5 15 20 0.993
spherical shape (Fig. A.1) and the perturbation is described 5 30 40 0.994
by Legendre polynomials of order n Pn [cos()] with an am- 10 15 45 0.994
plitude ., i.e., the initial shape in spherical coordinates is 4 9.487 0.662 5 15 20 0.660
a N : number of elements in direction.
r = 1 + .Pn [cos()] (37) b Nr : number of elements in r direction.
384 Y.-C. Liao et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 250 (2004) 367384

and Nr are the number of elements in the and r direc- [22] R.V. Cratser, O.K. Matar, D.T. Papageorgiou, Phys. Fluids 14 (2002)
tions) are used to find the value of Rout for approximating 1364.
Rout . The calculated time period Tcalculated is found [23] M.E. Timmermans, J.R. Lister, J. Fluid Mech. 459 (2002) 289.
[24] V.X. Nguyen, K.J. Stebe, Phys. Rev. Lett. 88 (2002) 164501.
to differ by less than 103 (error criterion for the calcula- [25] L. Brown, G. Narsimhan, P.C. Wankat, Biotechnol. Bioeng. 36 (1990)
tion) for Rout = 5 and 10. Hence, Rout = 5 is adequate for 947.
approximating Rout . When the element size is dou- [26] A. Pinazo, L. Perez, M. Infante, E.I. Franses, Colloids Surf. A 189
bled (N = 30 and Nr = 40) for Rout = 5, Tcalculated varies (2001) 225.
by less than 103 . Thus, the original number of elements [27] J.E. Valentini, W.R. Thomas, P. Sevenhuysen, T.S. Jiang, H.O. Lee, L.
Yi, S.-C. Yen, Ind. Eng. Chem. Res. 30 (1991) 453.
with Rout = 5 is sufficient for accurate computation for the [28] P.R. Schunk, L.E. Scriven, in: P.M. Schweizer, S.F. Kistler (Eds.), Liq-
deformed bubble problem, and is used for perturbed shapes uid Film Coating: Scientific Principles and Their Technological Impli-
with n = 2 and 4. Comparing Tcalculated for n = 2, 3, and 4 cations, Chapman & Hall, New York, 1996.
with the theoretical values of T, one finds an error less than [29] E.I. Franses, O.A. Basaran, C.-H. Chang, Curr. Opin. Colloid & Inter-
103 . Thus, the code does satisfy the test and provides addi- face Sci. 1 (1996) 296.
[30] R. Miller, R. Wustneck, J. Kragel, G. Kretzschmar, Colloids Surf. A
tional confidence in the accuracy of the computations. 111 (1996) 75.
[31] E.D. Wilkes, S.D. Phillips, O.A. Basaran, Phys. Fluids 11 (1999) 3577.
[32] O.A. Basaran, D.W. DePaoli, Phys. Fluids 6 (1994) 2923.
References [33] E.D. Wilkes, O.A. Basaran, Phys. Fluids 9 (1997) 1512.
[34] E.D. Wilkes, O.A. Basaran, J. Fluid Mech. 393 (1999) 333.
[1] R.H. Notter, P.E. Morrow, Ann. Biomed. Eng. 3 (1975) 119. [35] E.D. Wilkes, O.A. Basaran, J. Colloid Interface Sci. 242 (2001) 180.
[2] G. Enhorning, J. Appl. Physiol.: Respir. Environ. Exerc. Physiol. 43 [36] S.F. Kistler, L.E. Scriven, Int. J. Numer. Methods Fluids 4 (1984) 207.
(1977) 198. [37] P.K. Notz, A.U. Chen, O.A. Basaran, Phys. Fluids 13 (2001) 549.
[3] S.B. Hall, M.S. Bermel, Y.T. Ko, H.J. Palmer, G. Enhorning, R.H. [38] A.U. Chen, P.K. Notz, O.A. Basaran, Phys. Rev. Lett. 88 (2002) 174501.
Notter, J. Appl. Physiol. 75 (1993) 468. [39] K.N. Christodoulou, L.E. Scriven, J. Comp. Phys. 99 (1992) 39.
[4] R.H. Notter, Lung Surfactants: Basic Science and Clinical Appilca- [40] O.I. del Rio, A.W. Neumann, J. Colloid Interface Sci. 196 (1997) 136.
tions, Marcel Dekker Inc., Basel, Switzerland, 2000. [41] K.S. Avramidis, T.S. Jiang, J. Colloid Interface Sci. 147 (1991) 262.
[5] S.Y. Park, C.-H. Chang, D.J. Ahn, E.I. Franses, Langmuir 9 (1993) [42] R.L. Kao, D.A. Edwards, D.T. Wasan, E. Chen, J. Colloid Interface
3640. Sci. 148 (1992) 247.
[6] C.-H. Chang, E.I. Franses, Chem. Eng. Sci. 49 (1994) 313. [43] L.E. Scriven, Chem. Eng. Sci. 12 (1960) 98.
[7] C.-H. Chang, E.I. Franses, Colloids Surf. A. 100 (1995) 1. [44] K.S. Avramidis, J. Colloid Interface Sci. 155 (1993) 516.
[8] C.-H. Chang, E.I. Franses, J. Colloid Interface Sci. 164 (1994) 107. [45] D.A. Edwards, R.L. Kao, D.T. Wasan, J. Colloid Interface Sci. (1993)
[9] K.A. Coltharp, E.I. Franses, Colloids Surf. A 108 (1996) 225. 518.
[10] C.-H. Chang, K.A. Coltharp, S.Y. Park, E.I. Franses, Colloids Surf. A. [46] Y.-C. Liao, Adsorption Dynamics and Fluid Mechanics of Surfactant
114 (1996) 185. Solutions in Free Surface Flows. Ph.D. thesis, Purdue University, W.
[11] S.Y. Park, S.C. Peck, C.-H. Chang, E.I. Franses, in: D. Shah (Ed.), Lafayette, IN, 2004.
Dynamic Properties of Interfaces and Association Structures, American [47] M. Raffel, C. Willert, J. Kompenhans, Particle Image Velocimetry,
Oil Chemists Society Press, Champaign, IL, 1996, pp. 122. Springer-Verlag, New York, 1998.
[12] E.I. Franses, C.-H. Chang, J.B. Chung, K.A. Coltharp, S.Y. Park, D.J. [48] L. Gui, W. Merzkirch, Exp. Fluids 29 (2000) 30.
Ahn, in: D. Shah (Ed.), Micelles, Microemulsions, and Monolayers: [49] R. Lindken, L. Gui, W. Merzkirch, Chem. Eng. Technol. 22 (1999) 202.
Science and Technology, M. Dekker, New York, 1998, pp. 417435. [50] S.T. Wereley, L. Gui, Exp. Fluids 34 (2003) 42.
[13] X. Wen, K.C. McGinnis, E.I. Franses, Colloids Surf. A, 143 (1998) [51] F.M. Orr, R.A. Brown, L.E. Scriven, J. Colloid Interface Sci. 60 (1977)
373. 137.
[14] S.Y. Park, R.E. Hannemann, E.I. Franses, Colloids Surf. B 15 (1999) [52] W.M. Deen, Analysis of Transport Phenomena, Oxford University
325. Press, New York, 1998.
[15] S.Y. Park, J.E. Baatz, R.E. Hannemann, E.I. Franses, in: K. Mittal, P. [53] G. Strang, G.J. Fix, An Analysis of the Finite Element Method,
Kumar (Eds.), Emulsions, Foams, and Thin Films, Special Volume in Prentice-Hall Inc., Englewood Cliffs, NJ, 1973.
Honor of Prof. D.T. Wasan, M. Dekker, New York, 2000, pp. 329348. [54] P.M. Gresho, R.L. Lee, R.C. Sani, in: C. Taylor, K. Morgan (Eds.),
[16] X. Wen, E.I. Franses, Langmuir 17 (2001) 3194. Recent Advances Numerical Methods in Fluids, vol. 1, Pineridge,
[17] A. Pinazo, X. Wen, Y.-C. Liao, A.J. Prosser, E.I. Franses, Langmuir 18 Swansea, 1979, pp. 2779.
(2002) 8888. [55] X. Zhang, R.S. Padgett, O.A. Basaran, J. Fluid Mech. 329 (1996) 207.
[18] S.J. McClellan, E.I. Franses, Colloids Surf. B 30 (2003) 1. [56] M. Luskin, R. Rannacher, Applicable Anal. 14 (1982) 117.
[19] H.A. Stone, L.G. Leal, J. Fluid Mech. 220 (1990) 161. [57] T.-Y. Chen, J. Tsamopoulos, J. Fluid Mech. 255 (1993) 373.
[20] B. Ambravaneswaran, O.A. Basaran, Phys. Fluids 11 (1999) 997. [58] J.W.S. Rayleigh, Proc. R. Soc. Lond. 29 (1979) 71.
[21] O.A. Basaran, AIChE J. 48 (2002) 1842. [59] J.A. Tsamopoulos, R.A. Brown, J. Fluid Mech. 127 (1983) 519.

S-ar putea să vă placă și