You are on page 1of 14

Chemical Engineering Science 65 (2010) 62966309

Contents lists available at ScienceDirect

Chemical Engineering Science


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

Mass or heat transfer from spheroidal gas bubbles rising through a stationary liquid
Bernardo Figueroa-Espinoza a, Dominique Legendre b,c,
a

Instituto de Ingeniera, Universidad Nacional Autonoma de Mexico, Campus Merida, 97100 Merida, Yucatan, Mexico Universite de Toulouse; INPT, UPS; IMFT (Institut de Mecanique des Fluides de Toulouse), Allee Camille Soula, F-31400 Toulouse, France c CNRS; IMFT; F-31400 Toulouse, France
b

a r t i c l e in f o
Article history: Received 26 May 2010 Received in revised form 10 September 2010 Accepted 11 September 2010 Available online 18 September 2010 Keywords: Bubble Fluid mechanics Heat/mass transfer Bubble deformation Multiphase ow Numerical simulation

a b s t r a c t
Mass transfer was studied for the case of a spheroidal bubble rising through a stationary liquid. A numerical code that solves the NavierStokes equations and the diffusionadvection equation for the concentration was used to characterize the transfer from the bubble to the surrounding liquid phase. Simulations were carried over systematically for Reynolds number ranging from 1 to 1000, Schmidt numbers from 1 to 500 and bubble aspect ratio from 1 to 3. It appears that the use of the equivalent diameter as the characteristic length is the more appropriate to describe the transfer. The effect of bubble aspect ratio on the Sherwood number has been analyzed. At rst order the extension of Boussinesq expression using the equivalent diameter can be used for practical purposes. The evolution of the correction factor that compares the Sherwood number to the one of a sphere with same equivalent Peclet number is presented and described using simple correlations. The implementation of these results into EulerEuler simulations of mass transfer is discussed. It appears that the modication of the interfacial area combined to the modication of the Sherwood number gives a signicant contribution to the interfacial source term in the equation of the concentration. Note that the results can also be considered for heat transfer and used for inviscid drops. & 2010 Elsevier Ltd. All rights reserved.

1. Introduction Mass and heat transfers are important from the chemical, environmental and industrial points of view. A great variety of processes such as fermentation or aerobic digestion imply the use of chemical and biological reactors, stirred vessels, bubble columns, etc. This work was motivated by a project related to mass transfer in waste water treatment plants, where biological reactors are aerated through the injection of small bubbles. Oxygen in gaseous phase coming from the bubbles is transferred through the gasliquid interface, allowing for the aerobic respiration of micro-organisms present in the liquid phase. The modeling of such facilities involves biological, chemical and physical subsystems and their interaction. In general, local and large scale hydrodynamic phenomena have to be studied separately given the impossibility of describing the physics for each bubble in a reactor. Not even the most powerful and modern computer could deal with the direct numerical simulation of a few number of bubbles rising in a (very) small aeration vessel (Koynov et al., 2005). A more practical viewpoint for numerical simulations is to model large scale physics separately, based on

Corresponding author.

E-mail addresses: dominique.legendre@imft.fr, legendre@imft.fr (D. Legendre). 0009-2509/$ - see front matter & 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.ces.2010.09.018

subgrid local models that feed the proper local behavior of the ow in terms of measurable quantities such as the gas/liquid volume fraction and mean bubble size. However, for these submodels to be precise and reliable, bubble dynamics, mass/heat transfer have to be studied at the scale of a single inclusion. Transfer (of mass, heat, momentum) depends on both the inclusion geometry (area, shape) and the ow hydrodynamics. From the point of view of the latter, this interdependence between hydrodynamics and topology can be understood in terms of the forces acting on the bubble: inertia, surface tension, buoyancy and viscous stress. With these elements, three dimensionless groups can be dened, for example the Reynolds, the Eotvos and the Morton number. Useful correlations of the bubble shape and velocities in terms of these quantities can be found in the work of Clift et al. (1978) as a function of the aforementioned dimensionless groups (see also Magnaudet and Eames, 2000 for recent progress on spherical bubbles hydrodynamics). Bubble shape for moderate deformation can be approximated as oblate spheroids. For signicant deformation (typically for an aspect ratio larger than 2.5 for air/water system) bubbles shape loses fore-and-aftsymmetry (Sanada et al., 2007; Zenit and Magnaudet, 2008). The regime of ow that concerns us is characterized by oblate ellipsoidal bubbles that rise following vertical rectilinear paths. The rise velocity of such bubbles has been studied theoretically by Moore (1965), who calculated the bubble aspect ratio and drag

B. Figueroa-Espinoza, D. Legendre / Chemical Engineering Science 65 (2010) 62966309

6297

force of slightly oblate ellipsoidal bubbles by solving the boundary layer between the bubble surface and the external (potential) ow. The effect of oblateness on the drag coefcient has also been recently revisited by Legendre (2007), who obtained the drag of a bubble from a simple correlation between the drag and the maximum vorticity produced on the bubble surface. The characteristic trajectories of these deformed rising bubbles was recently studied in detail by Ellingsen and Risso (2001), Sanada et al. (2007), Magnaudet and Mougin (2007), and Zenit and Magnaudet (2008). Within the context of mass/heat transfer, the case of spheroidal bubbles has been considered theoretically by Lochiel and Calderbank (1964), who developed analytical expressions for the transfer coefcients (Sherwood number) of solid and uid spheres and spheroids. These results are valid for the limiting cases of very small or very large Reynolds numbers.1 A general expression for the Sherwood number of an axisymmetric body (of arbitrary shape) was obtained. For the case of large Reynolds number, the transfer was derived using velocities coming from the potential ow. The resulting expression for the Sherwood number of an oblate spheroid (of major and minor semi-axis b and a) can be expressed as 2 Sheq w p Pe1=2 f w eq

Since Moores derivation provides a very good prediction for the drag for Reeq 4 50, its use for deriving the transfer is also expected to give a good correction to the Boussinesq solution. This will be discussed in the following. Note that Eq. (4) can be transformed in order to eliminate the divergence of the solution in the limit of small Reynolds numbers by considering that for Reeq b1,
1=2 122:89Re1=2 1=2 % 1 2:89Reeq 1=2 . eq In the opposite limit of small Reeq and Peeq numbers, the evolution of the Sherwood number of a spherical bubble is given by (Brenner, 1963)

1 Bre Sheq 2 Peeq 2

where w b=a is the aspect ratio (w Z1 in our study) and Peeq is the Peclet number based on the equivalent diameter. f w is the correction to the Sherwood number ShB of a spherical bubble of eq same Peclet number Peeq given by the potential ow solution (Boussinesq, 1905): 2 B Sheq p Pe1=2 eq

Theoretical and numerical studies have also been conducted to study transient transfer, transfer in unsteady and non uniform ows (Legendre and Magnaudet, 1999). A broad bibliographical review for momentum and mass/heat transfer with and without phase transition can be found in Michaelides (2003). As indicated before most of these studies have focused on spherical bubbles. The objective of this work is to analyze the effect of bubble aspect ratio on the transfer, in order to compare it with previous investigations and to obtain useful correlations in terms of the pertinent parameters. For this purpose, direct numerical simulations (DNS) solving the NavierStokes equations and the diffusionadvection equation for the concentration are performed on orthogonal boundary-tted grids.

2. Governing equations Let us consider an oblate spheroidal bubble moving through a liquid at a constant velocity U1 (Fig. 1). It is assumed that the bubble has reached a stationary state (from a reference frame moving with the bubble), so its shape does not change with time. The liquid of constant density r and kinematic viscosity n is therefore moving at velocity ~u at a given position ~u. The u x dimensionless variables are chosen in terms of the reference values: t tuU1 , L ~ u ~u u , U1 ~ x ~u x , L p pu
2 rU1

The results were compared with some previous experimental investigations, showing differences that can reach as much as 20% (Lochiel and Calderbank, 1964). Due to the potential ow assumption, this correlation (Eq. (1)) is valid for the limit Reeq -1, corresponding to asymptotic vanishing of the vorticity boundary layer (Moore, 1963, 1965). The effect of the wake recirculation was not considered in this particular case. This aspect will be investigated in Section 5.1. More recently, the solution was extended to oblate and prolate spheroids (Favelukis and Ly, 2005) and oblate spheroids at low Reynolds number (Favelukis, 2010). For moderate Reynolds number and high Schmidt numbers, Takemura and Yabe (1998) determined the oxygen concentration of millimetric (almost spherical) gas bubbles by means of precise partial pressure and (time-varying) diameter measurements. They compared previous expressions from the literature (Clift et al., 1978; Oellrich et al., 1973; Leclair and Hamielec, 1971; Winnikow, 1967) and proposed a semi-empirical relation that gives a better t of their experiments and numerical simulations: ( )1=2 2 2 1 TY Sheq p 1 2:5 Pe1=2 3 eq 2=3 3 1 0:09Reeq 3=4 p Among the expressions from the literature, we retain the analytical solution obtained by Winnikow (1967) which is based on the tangential velocity derived by Moore (1963). " #1=2 2 2:89 W 1=2 p 1 p Peeq 4 Sheq p Reeq

and c

cucu1 cus cu1

where tu, ~u, ~u, pu and cu are the dimensional time, velocity, u x position, pressure and concentration, respectively. cu1 is the gas concentration far from the interface (or the bulk concentration), cus the concentration at the interface in the liquid, U1 is the reference velocity and L is a characteristic length. In terms of a (Eulerian) frame of reference xed with the bubble, the liquid velocity eld is given by the NavierStokes equations in dimensionless form

r ~0 u

1 The denitions of the relevant dimensionless parameters like the Sherwood Sheq and Reynolds number Reeq are given in Section 2.

Fig. 1. Statement of the problem.

6298

B. Figueroa-Espinoza, D. Legendre / Chemical Engineering Science 65 (2010) 62966309

@~ u ~ r~ rp r t u u @t

u u where t Re1 r~ rT ~ is the viscous part of the stress tensor and Re U1 L=n is the Reynolds number. The advectiondiffusion equation is expressed as @c 1 2 r ~c u r c @t Pe where Pe LU 1 ReSc D 10 9

and the advectiondiffusion for the concentration (or temperature), for any orthogonal curvilinear coordinate system. The discretization method is nite volumes, which is well adapted to properties conservation. Precision is second order in time and space (RungeKutta/CrankNicolson schemes) and the code has been used to solve hydrodynamic and transfer problems in the past (see Magnaudet et al., 1995; Calmet and Magnaudet, 1997; Legendre and Magnaudet, 1998, 1999; Legendre et al., 1998) for a detailed presentation of the code). The RungeKutta/Crank Nicolson procedure guarantees numerical stability for time step p satisfying CFL numbers less than 3. 3.1. Numerical grid An orthogonal boundary-tted grid was used and a mesh renement is applied near the bubble surface to capture dynamic and mass boundary layers. The grids were generated by mapping a particular curvilinear coordinate system x, Z expressed in the cartesian representation (x,y) in such a way as to t the geometry of the problem. This curvilinear system is chosen in such a way as to t the geometry of the problem. Two different kinds of grids were tested: one grid based on the equipotentials and streamlines of the velocity eld of an uniform ow past an elliptical cylinder, as described by Legendre and Magnaudet (1998) and used in Legendre et al. (1998); Legendre and Magnaudet (1999) for heat transfer studies. This grid conguration will be henceforth denoted as type A grid. The total number of nodes is N Nb Nx NZ , where NZ is the number of nodes in the radial direction (x is the symmetry axis as shown in Fig. 2(a). The second type of grid (from now on denoted as type B), is based on the bijective transformation of curvilinear ellipsoidal coordinates to cartesian, and dened as (see Adoua et al., 2009): x kacosy p y k 1 a2 siny

is the Peclet number, c is the dimensionless concentration (or temperature) of a particular chemical species (for example oxygen) in the mixture, and Sc is the Schmidt number, dened as Sc

n
D

11

where D is the diffusion coefcient and n is the kinematic viscosity. Since the liquid viscosity is several orders of magnitude larger than that of the gas, the liquid at the interface experiences no shear at the bubble surface (slip condition). In the absence of phase change, the normal velocity is zero at the interface. Thus, the boundary conditions of the problem are ) ~~0 v n 12 at Z 0 ~ t ~ 0 n n where Z is the curvilinear coordinate that follows the outward normal ~ to the bubble surface. Note that these boundary n conditions are also satised for drops characterized by a viscosity much smaller than the viscosity of the surrounding liquid. For the concentration, the boundary conditions is ( 1 for Z 0 c 13 0 for Z-1 The effect of the bubble shape on the transfer is studied in terms of the Sherwood number: R ~ A Drcu n dAb 14 Sh b Dcu cu A
s 1

q p where the minor semi-axis a ka0 , b k 1 a2 and a0 1= w2 1 0 (a cte represents a family of concentric ellipses). Fig. 2(b) shows a portion of this grid (the center). The whole grid has a semi-circular shape far from the bubble ka b 1. The extent of the numerical domain in our case is R1 80b. The grid spacing increases monotonically with a growth rate between 0.8 and 1.2. The bubble surface is discretized into nx nodes, and only a half of the xy plane is considered because of symmetry. The total number of nodes is then nx ny. The surface of the bubble is consequently discretized onto nx elements. For the type B grid conguration, the velocity eld far from the bubble is uniform, entering the domain through the cells at the (left) boundary (i1 to ne, jny) and leaving it through the cells located to the right (ine +1 to nx, jny). ne is close to nx/2. For the 3D simulations reported in Section 5.3, the plane grid presented in Fig. 2 is rotated about the symmetry axis ex of the bubble. According to the tests done in previous studies, the grid is made of 64 uniformly distributed cells in the azimuthal direction (Magnaudet and Mougin, 2007; Adoua et al., 2009). 3.2. Preliminary tests Several parameters were tested for both types of numerical grids. Based on some previously reported tests with the code (Legendre and Magnaudet, 1998, 1999; Legendre et al., 1998, 2003; Adoua et al., 2009), a conguration with nx ny 80 was chosen for grids of type B. Once these parameters xed, the grid renement was imposed as a function of the minimum distance from the bubble surface to the rst grid line D. This distance is important since it gives the number of nodes ND within the

n where Ab is the bubble surface area and ~ is the unit normal to the interface. The local distribution of the transfer is presented using the local Sherwood number Shloc Lrcu ~=cus cu1 . As usual, n there is a thermal analogy for the Sherwood number: the Nusselt number. The results can be presented in terms of a length scale based on the bubble major axis (L2b) or based on the bubble equivalent diameter (or radius) L deq 2req 8pab 1=3 . For 3 the latter situation, the subscript eq will be added to the corresponding dimensionless number (Reeq, Sheq, etc). The absence of a subscript in Sh or Re (or any quantity involving L) denotes that L is taken to be the major axis length L2b. Note that the main dimensionless numbers used in this study are linked by the simple relations Sh w1=3 Sheq , Re w1=3 Reeq , Pe w1=3 Peeq 15
2

Moreover, for a spherical bubble w 1, deq 2b and the two normalizations are equivalent.

3. Numerical procedure The system of Eqs. (7)(9) was solved via direct numerical simulations using the JADIM code. This code can solve the unsteady 3D NavierStokes equations in terms of velocitypressure variables

B. Figueroa-Espinoza, D. Legendre / Chemical Engineering Science 65 (2010) 62966309

6299

1.01 1 0.99 0.98


Sc = 1 Sc = 100

0.97 0.96 0.95


Y

Sc = 500

0.94
... Nnodes (N, 1) Z X

0.4
(N+Nb, 1) N nodes ...

0.6

0.8

1.2

1.4

1.6

1.8

2 x 103

Fig. 3. Convergence of the Sherwood number ShD normalized by ShDmin (with Dmin 0:0003) vs. D for w 2 and Re 300. Grids of type B.

Magnaudet (1995) using the same code JADIM but with a different grid generation procedure. The type B conguration was chosen for the study because it appeared to be more stable at large Peclet numbers (with large Schmidt number) during the preliminary tests with ellipsoids of large aspect ratio. The reason of this is presumably the irregular shape of the node (N,1) for type A (see Fig. 2). The observed effect on the mean Sherwood number is a reduction less than 5% for Pe150 000 using type B compared to the value obtained using type A. Grid convergence is reported in Fig. 3 for w 2 and Re 300. Different Schmidt numbers were considered. The vertical axis represents the Sherwood number Sh D Sh=ShDmin , normalized with the corresponding value of the smallest grid size considered Dmin 0:0003 that corresponds to Dmin % dud =10b for Pe150 000. Note that Sh D 0:0005 is above 0.996 for Sc 500. We also observe that grid convergence is slower for large values of the Reynolds and Peclet numbers, where both the hydrodynamic and diffusive boundary layers are very thin.
X (nx, 1) ny nodes ...

...

(1, 1) ny nodes

4. Additional validations
Fig. 2. Curvilinear mesh based on elliptical coordinates transformation. (a) Type A; (b) Type B.

4.1. Bubble wake There is an important specic aspect concerning oblate spheroids that may have an effect on the transfer. In contrast with spherical bubbles, the wake is characterized by the appearance of a recirculation zone behind the bubble. This transition takes place at large Reynolds numbers and aspect ratios w. According to Magnaudet and Mougin (2007), the onset of the recirculation zone corresponds to the critical values wc 1:6 and Rec 130. The limits of appearance of this recirculation is, as a matter of fact, a function of both parameters (w and Re) as shown in Fig. 4, where the diagram Re vs. w is presented. The solid line to the left represents the appearance of a stationary wake and a recirculation zone, as calculated by Magnaudet and Mougin (2007). For large enough Reynolds numbers the bubble wake bifurcates into an unstationary 3D wake resulting in zigzag and helical paths. This situation is depicted in the zone labeled 3D wake in Fig. 4. The markers represent (most of) the simulations run during this study. Empty circles correspond to

1=2 diffusive boundary layer dud of order dud Odeq Peeq . According to our experience, a minimum of ve nodes is necessary. Table 1 2 2 presents the drag coefcient CD 2FD =preq rU1 where FD is the drag force and the Sherwood number Sh obtained for both types of grid and different mesh renements for a spherical bubble. The expression of Mei et al. (1994) for spherical bubbles, which is valid for all Reynolds numbers is also tabulated. Considering the maximum Peclet number Pe150 000 reported in Table 1 one has dud % 0:003deq . The results reported indicate that both type of grid give very close results for both the drag and Sherwood numbers. The results for spherical bubbles in terms of the drag given in Legendre and Magnaudet (1998) were obtained with type A grids. Some hydrodynamics aspects of spheroidal bubbles using type B grids were studied by Adoua et al. (2009) and Magnaudet and Mougin (2007). Very close results were found by Blanco and

6300

B. Figueroa-Espinoza, D. Legendre / Chemical Engineering Science 65 (2010) 62966309

2.5

Shloc Pe-1/2

1.5

0.5

0 0 20 40 60 80
Fig. 5. Local Sherwood number Shloc Pe 1/2 for spherical bubbles vs. tangential angle y for grid B. Potential solution (Favelukis and Ly, 2005), Pe 300, & Pe 104.

100

120

140

160

180

Fig. 4. Phase diagram of the recirculation and unsteady wake. from Magnaudet and Mougin (2007). Markers: this study. 3 absence of recirculation. axisymmetric recirculation, 3D unsteady wake.

the absence of recirculation. These bubbles should rise following a rectilinear path. The crosses correspond to bubbles where an axisymmetric recirculation zone appears. This recirculation does not break the symmetry of the ow and the trajectory is still rectilinear. This was veried through a zero lift force calculated from 3D simulations. The asterisks represent bubbles where a 3D wake develops, breaking symmetry and presenting path instability. The results obtained through simulations are in very good agreement with those reported in previous investigations (note that for w 2:5 and Re 43000, the wake is again axisymmetric). The gure also contains some representations of the typical streamline pattern for the different regimes. The transition to the stationary recirculation zone can be approximated as follows:

103
Sc = 1 Sc = 10 Sc = 100 Sc = 500

102 Sh ( = 1) 101 100 101 100 101 Re 102 103


Fig. 6. Sherwood number vs. Reynolds number for w 1. Legendre and Magnaudet (1999) same code but with grid A. Other symbols: this study, ShTY Takemura and B Yabe (1998), Sh 2Pe=p1=2 Boussinesq (1905), Eq. (4) from Winnikow (1967), 2 2 2 ShBre 2+Pe/2 Brenner (1963).

w1 Re % 0:0182lnRe3 0:3908lnRe2 2:5341lnRe 6:8255


16 And the onset of the unstationary wake can also be calculated using a similar expression:

w2 Re % 0:0286lnRe3 0:6846lnRe2 5:2016lnRe 14:9696


17 It is easy to obtain an approximation if one observes the same plot in a logarithmic scale. Both curves seem to t third order polynomials remarkably well. These approximations are depicted in Fig. 4 as dotted lines. Moreover, let us consider both test points at w 2:5, Re300 and w 2, Re1500; the results calculated from Eqs. (16) and (17) are consistent with the observed regimes. Note the absence of recirculation in the latter case. Concerning mass/heat transfer, the induced effect of the recirculation zone that forms at the rear of spheroidal bubbles will be further discussed in Section 5. 4.2. Local and total Sherwood number for spherical bubbles The local Sherwood number was also compared to theoretical expressions available from Boussinesq (1905), Ruckenstein (1959) and Favelukis and Ly (2005). Fig. 5 shows the local Sherwood number Shloc y as a function of the angle y in the tangential direction for Pe 300 (Re 300 and Sc 1) and Pe10 000 (Re100 and Sc 100). These values are chosen so that the

conditions of validity of the analytical potential solution (Re b 1 and Peb p 1) are satised. The results are normalized by the factor Pe=p. All the curves collapse (approximately) towards the theoretical values based on the potential ow assumption. Fig. 6 presents the Sherwood number as a function of Re for spherical bubbles. The markers correspond to numerical simulations for Sc1, 10, 100, 500 obtained with grid B. The numerical results are compared to the correlations (2)(5) listed in introduction. We also report simulations obtained for Sc1 using grid A by Legendre and Magnaudet (1999). Our simulations are in good agreement with the Takemura and Yabe correlation (3) that provides an accurate correction for Re4 1 to the Boussinesq solution. For example, consider Sc 500 and Re300 the characteristic values for a millimetric rising air bubble transferring oxygen in water. The Sherwood numbers given by Boussinesq solution (2) and Takemura and Yabe correlation (3) are ShB 437

B. Figueroa-Espinoza, D. Legendre / Chemical Engineering Science 65 (2010) 62966309

6301

and ShTY 394, respectively. The corresponding difference between the two values is around 10%. The value given by our simulation is Sh400 corresponding to 8% and 1.5% difference with ShB and ShTY, respectively. As expected the solution derived by Winnikow (1967) provides a very good description of the transfer for Re4 50 and the agreement is for example less than 1% for Sc500 and Re300. In the limit of small Reynolds numbers our simulations are in good agreement with the solution (5). Note however that relation (3) has to be used sensibly since it does not converge to Sh2 when Re-0 and Pe-0.

5. Results Simulations were carried over systematically for Reynolds number Re ranging from 1 to 1000, Schmidt numbers Sc from 1 to 500 and bubble shape w from 1 to 3. Section 5.1 rst presents the results concerning the local distribution of the transfer at the bubble surface. The mean Sherwood number is then given in Section 5.2 and a simple correlation that ts the numerical results is proposed. Finally, some unsteady effects are discussed in Section 5.3. Simulations reported in Sections 5.1 and 5.2 are carried out using the axisymmetric grid. In order to minimize the cost of the simulations, the ow eld is rst calculated and a steady state is reached for each Reynolds number considered. Then the concentration equation is solved for different Schmidt numbers and the corresponding Sherwood number is calculated. 5.1. Local transfer Figs. 7(a)(d) show the local Sherwood number at the bubble surface for different w and Peclet numbers. These results correspond to stationary state. It is clear that as the aspect ratio

increases, the difference with the theoretical potential prediction becomes more important. The gure also shows that this discrepancy is not to be very sensitive to variations on the Pe number (for Pe Z 100). Another phenomenon becomes evident from these gures: let us consider subgure (c): after attaining a maximum at 901, the local transfer decreases to zero at about 1251, and then starts increasing up to the rear of the bubble at 1801. This is the effect of the recirculation zone that develops in the bubble wake and already discussed (see Fig. 4). Obviously the potential theory does not take into account this type of wake. However the differences between potential theory and direct simulations are important along the curve long before the recirculation appears. The reason for this is vorticity production at bubble surface; it was shown by Magnaudet and Mougin (2007) that the curvature variations at the equator y 903 causes an increase on the vorticity production. For a given Reynolds number the increase of the aspect ratio results in the increase of the vorticity omax deq =U1 $ w8=3 . As a consequence the discrepancies with the potential ow solution increases with the bubble aspect ratio as observed in Fig. 7. This effect is shown in Fig. 8 that presents the tangential velocity at the bubble surface for different aspect ratios: (a) w 1, (b) w 1:2, (c) w 1:5, (d) w 2. The tangential velocity from potential ow theory (continuous line) and simulations for Re300 (dashed line) and 500 (dash-dot) are compared. Subgure (a) (spherical bubbles) shows that as the Reynolds number increases, the difference (lled dots) decreases, and can be explained in terms of the boundary layer analysis (Moore, 1963), represented as dotted lines. However, this approximation diverges near the rear stagnation point. Subgure (b) contains the same representation for a slightly deformed bubble with aspect ratio w 1:2. The general shape is not the same anymore, even if the tangential velocity is close to the spherical bubble case. The

= 1.2 2.5 2 Shloc () 1.5 1 0.5 0 0 50 =2 2.5 2 Shloc () 1.5 1 0.5 0 0 50 100 150
onset of recirculation

= 1.5 2.5 2 Shloc () 1.5 1 0.5 0

100

150

50

100

150

= 2.5 2.5 2 Shloc () 1.5 1 0.5 0 0 50 100 150

Fig. 7. Local Sherwood number vs. y. 2 2 2 Pe 104 (Re 100 and Sc 100),  Pe 103 (Re 100 and Sc 10), Pe 9 104 (Re 300 and Sc 300) Potential solution (Favelukis and Ly, 2005) Note: the sketched bubbles are approximative and may not have the indicated aspect ratio.

6302

B. Figueroa-Espinoza, D. Legendre / Chemical Engineering Science 65 (2010) 62966309

=1 2 2

= 1.2

1.5

1.5

u 0 50 = 1.5 100 150

0.5

0.5

0 0 50 =2 2 100 150

1.5

1.5 u 0 50 100 150

0.5

0.5

0 0 50 100 150

Fig. 8. Tangential velocity uy vs. y at the bubble surface. tangential velocity utheo from theory,2 usim Re 500, usim Re 300,  utheo usim for Re 300, boundary y y y y theta layer approximation spherical bubbles (Moore, 1963) for Re 300, utheo usim for w 1 and Re 500, 3 Moore approx. for Re 500. y theta

difference between the potential ow and simulated velocities is less sensitive to the Reynolds number, and now it has two crests and a local minimum at y 903 . These gures clearly show that the difference between the potential ow predictions and the simulations increases with the bubble aspect ratio as explained above. This evidences a more complex boundary layer, responsible of important differences that nally turn into negative velocities, the onset of the recirculation zone, as observed in subgure (d). One other particular effect of the bubble shape is the sharp evolution of the transfer near the equator y 03 . Fig. 7 clearly reveals that the transition between the front part and the rear part of the bubble becomes sharper when increasing the aspect ratio . The consequence is that most of the transfer is done on the front side of the bubble. We also observe the development of a peak of transfer at the equator corresponding to the peak of tangential transport as shown by the velocity distribution in Fig. 8.

the concentration boundary layer evolves as dud Odeq Pe1=2 as eq discussed in the previous section. Considering now the effect of the aspect ratio, Fig. 9 indicates that it can have a signicant effect. For example, considering the case Re100 and Sc 500, Shw 1 219 and Shw 3 244 corresponding to an increase of 11%. On the other hand, considering the case Re 10 and Sc 100 we have Shw 1 28:4 and Shw 3 23:7 showing a decrease of the interfacial transfer (about 16%) when increasing the aspect ratio. When using the equivalent diameter as the characteristic length scale, we observe that for small Sc, as well as Reeq 4 100, the data tend to align on the same curve corresponding to a simple extension of the Boussinesq solution. This indicates that the use of L deq as a characteristic length scale is more appropriate than L2b to describe the evolution of the Sherwood number in the limit Reeq b 1 where the Sherwood number is found to evolve as 2 Sheq $ p Pe1=2 eq 18

5.2. Total transfer The total transfer expressed using the Sherwood number is reported in Fig. 9 and the two possible normalizations are compared. Subgure (a) presents Sh vs. PeReSc while subgure (b) presents Sheq vs. Peeq ReeqSc. The net effect of the bubble aspect ratio can be seen from the arrows. A simple extension of p B B Boussinesqs solution, respectively Sh 2 Pe=p and Sheq 2 p Peeq =p is also reported in both gures. It is shown that the general evolution remains Sh $ Pe1=2 (resp. Sheq $ Pe1=2 ) showing eq that both the enhancement of interfacial vorticity production and the wake structure have not a signicant effect at rst order on the behavior of the transfer. The main reason is that most of the transfer is done on the front part of the bubble (see Fig. 7) where

An unexpected results is that the Sherwood number is very close to the Sherwood number of a spherical bubble with the same Peclet number based on the equivalent diameter, i.e. f w Sheq w,Reeq ,Sc=Sheq w 1,Reeq ,Sc % 1. Using a front tracking method, Darmana et al. (2006) report the Sherwood number for an ellipsoidal bubble for Re108, Sc 1 and w 1=0:49. They observe a minor effect of the deformation on the transfer in agreement with our simulation for a very close case (Re100, Sc 1 and w 2). As shown in Fig. 9b there is a different behavior at large but nite Peeq O102 oPeeq o O104 with large Sc number because it corresponds to Re o O10. The potential approximation cannot be used at such Reynolds numbers to

B. Figueroa-Espinoza, D. Legendre / Chemical Engineering Science 65 (2010) 62966309

6303

10

3
Sc = 1 Sc = 10 Sc = 100 Sc = 300 Sc = 500

=3 =1

102 Sh

transfer increases again. Fig. 10 clearly shows that the agreement with the potential solution increases with the Reynolds number but remains limited to small aspect ratio. A signicant difference is observed for large aspect ratio due to the restrictive validity of potential ow for deformed bubbles as explained above. Considering the range of value of interest in some common gasliquid systems, a simple correlation can be used to describe the effect of bubble aspect ratio for 500 oRe o 1000 and Sc 4 100: f w Sheq w,Reeq ,Sc a bw cw2 dw3 Sheq w 1,Reeq ,Sc 19

=1

101

=3

with a 0.524, b 0.88, c 0.49 and d 0.086. This relation is reported in Fig. 10 for Re 300, Re500 and Re 1000.

100 101

5.3. Some unsteady effects


100 101 102 Pe 103 104 105 106

103
Sc = 1 Sc = 10 Sc = 100 Sc = 300 Sc = 500

102 Sheq
=1

101

=3

100 101

100

101

102 Peeq

103

104

105

106

Fig. 9. Sherwood number for different aspect ratio w (see arrows). (a) Sh vs. Pe B (normalization based on L 2b). Sh 2Pe=p1=2 . (b) Sheq vs. Peeq (normalB isation based on L deq), Sheq 2Peeq =p1=2 .

describe the transfer on the frontal part of the bubble since the boundary layer is diffused. A detailed inspection of the effect of the bubble aspect ratio can be conducted using Sheq. The correction factor f w Sheq w=Sheq w 1 is shown in Fig. 10. The solution obtained by Lochiel and Calderbank (1964) under the assumption of potential ow is also reported for comparison. Every case presented in Fig. 10 can be described using a simple third order polynomial represented with dotted lines. For Re10 we observe the most signicant effect of the aspect ratio on the Sherwood number. The Sherwood number decreases almost linearly with the aspect ratio and can be described as f w 10:13w1. For larger Reynolds number (Re 4100), Fig. 10 conrms that the effect of the bubble aspect ratio is moderate when considering the Sherwood number Sheq. The effect is less than 10%. The correction factor f w presents some small evolutions that can be explained by considering the bubble interfacial velocity and the wake structure as commented above (see also Fig. 7). First due to the deformation the tangential velocity increases so does the transfer until w % 1:2. Then due to the increase of both the surface curvature and the tangential velocity, the interfacial vorticity increases and it induces a reduction of the ow rate near the interface and consequently the tangential transport and mass transfer are reduced. And nally, for 2 o w o 2:5 due to the positive effect of the wake, the

In this section we consider two unsteady evolutions for the mass transfer. The rst one is due to the transient evolution when a bubble is suddenly introduced in a liquid of constant concentration. The second one is the consequence of the wake destabilization. Simulations reported in this section are carried out by simultaneously solving the NavierStokes equations and the concentration equation. We rst consider the transfer for a bubble suddenly introduced at its terminal velocity in a liquid of constant concentration cu1 . Fig. 11 reports the Sherwood number as a function of dimensionless time for two aspect ratios, w 1:2 and w 2, respectively. The time is normalized by the advective time b=U1 . It can be observed from the gure that the characteristic time for establishing the transfer is tu $ Ob=U1 . For w 2 we can observe that the solution is not really stabilized since it continues to slightly decrease until convergence to the steady value for tu $ 200b=U1 . An additional effect of the recirculation was evidenced when analyzing the temporal evolution of the mean transfer. Apparently there is a time-dependent evolution of the transfer that gives a boss on the Sherwood evolution in time, as seen near point (c) of Fig. 11. We have tested by changing the grid renement and the time step that this evolution is grid and time independent. The answer of this particular behavior is given by examining the concentration eld in Fig. 12 and the corresponding local Sherwood distribution at the bubble surface reported in Fig. 13. The labels a to f make reference to different characteristic times observed in Fig. 11. In time a corresponding to tu $ Ob=U1 the concentration has formed a thin boundary and start to be advected in the bubble wake. After time a, the main changes of the local transfer are located in the rear part of the bubble 1303 r y r1803 . The evolution of the concentration eld in this region is thus responsible of the maximum observed between times b and c. The boss corresponds to the time when the transfer at the rear of the bubble is maximum. Before time c, the rear of the bubble has a low concentration, and the recirculation (which is visible through the streamlines pattern to the right) is not yet developed completely. The concentration gradient is sharp and consequently the transfer is large, as seen in Fig. 13. The maximum is observed when the recirculation is completely developed and before the concentration diffuses in the recirculation zone. After time c, due to the concentration diffusion, the gradient on the rear part of the surface decreases (see the shaded region behind the bubble in 12c), and the process tends towards its nal state in f, where concentration is almost homogeneous in the stationary wake. The ux transferred in the wake has been reduced (see Fig. 13) compared to its values at tu $ OU1 =b. This transitory time to the nal steady state is thus much larger than the transport time b=U1 . It corresponds to the time necessary for

6304

B. Figueroa-Espinoza, D. Legendre / Chemical Engineering Science 65 (2010) 62966309

1.2
Re=10

1.2
Re=100

1.1 1 0.9 0.8 0.7 1.2


Re=300

1.1 1 0.9 0.8 0.7 1.2


Re=500

1.5

2.5

1.5

2.5

1.1 1 0.9 0.8 0.7 1.2


Re=1000

1.1 1 0.9 0.8 0.7

1.5

2.5

1.5

2.5

1.1 1 0.9 0.8 0.7


Sc = 1 Sc = 10 Sc = 100 Sc = 300 Sc = 500

1.5

2.5

Fig. 10. Effect of bubble aspect ratio on the Sherwood number Sheq presented using f w vs. w. : Lochiel and Calderbank (1964). third order polynomial tting. f w 10:13w1 for Re 10 and relation (19) for Re 300, 500, 1000.

80 78 76 74 e 72 a b c d f

the concentration to ll the recirculation zone. This is done by diffusion across the toroidal stream lines. The corresponding time scale is then tuw L2 =D based on Lw, the characteristic size of the w stationary wake behind the bubble. tuw is thus linked to the advective time scale by the relation:  2 Lw b tuw Pe 20 U1 b Considering the case reported in Figs. 1113 for Pe3000, Lw can be estimated from Fig. 12. if we consider Lw as the equivalent diameter of the recirculation zone volume, one has Lw % b=4 and relation (20) gives tuw % 190b=U1 in agreement with the time evolution reported in Fig. 11. When the steady state is reached, the transfer at the rear of the bubble is controlled by the diffusion across the recirculation zone that exchanges with the surrounding liquid. Considering the induced effect on the mean Sherwood effect (the difference between the value at times b and f), it appears to be relatively small so that at rst order the transfer can be considered to be established after tu $ OU1 =b. A detailed inspection of this additional transient behavior necessary to ll the recirculation zone at the bubble concentration need the characterization of Lw. This is out of the scope of this study.

Sh

70 68 66 64 62 60 101

100

101

102

t, U/b
Fig. 11. Transient evolution of the Sherwood number vs. dimensionless time t U1 tu=b for Re 300 and Sc 10. w 1:2, w 2. Labels af refers to Fig. 12.

B. Figueroa-Espinoza, D. Legendre / Chemical Engineering Science 65 (2010) 62966309

6305

0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.05

0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.05

0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.05

0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.05

0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.05

0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.05

Fig. 12. Concentration eld and streamlines for Re 300, Sc 10 and w 2 for the instants marked in Fig. 11. (a) t 2.5, (b) t 7.5, (c) t 20, (d) t 50, (e) t 75, t 100.

The second unsteady effect that we would like to comment is relative to the wake destabilization. Let us consider the region of the Rew plane where the wake of a spheroidal bubble looses its

symmetry, causing a lift force to deviate the bubble from the rectilinear path, as seen in the region labeled 3D wake in Fig. 4. Under these conditions the recirculation zone is destabilized,

6306

B. Figueroa-Espinoza, D. Legendre / Chemical Engineering Science 65 (2010) 62966309

affecting also the mass transfer coefcient: a 3D simulation of a single bubble corresponding to Re300, Sc 10 and w 3 was carried out. The resulting instantaneous Sherwood number Sh(t) as a function of time t tuU1 =b is depicted in Fig. 14, as well as the corresponding instantaneous lift and drag coefcients CL(t) and CD(t). The lift and drag forces reported in Fig. 14 are ~ respectively the ~y - and ~x components of the total force Fu e e ~ acting on the bubble. Fu is calculated by integration of the total shear stress on the bubble surface at each time step. The dimensionless lift and drag coefcients CL and CD are obtained by dividing the corresponding components of the force by pb2 rU1 =2.

2.5

The axisymmetric ow prevails for a long time in this simulation because there were no external perturbations imposed to the ow. The instability arises anyway, since there is always numerical noise to trigger its growth. One can observe the appearance of an oscillating lift force t % 200 accompanied by an increase of the mean drag. The effect on the transfer is clearly reected. The breakup of the stationary wake causes a decrease in the Sherwood number and the time required to reach the permanent regime is % 50U1 =b. The mean value changes from Sh66.8 to Sh 61.5 and the amplitude of oscillation is 0.61, 1% of the mean value. In this case the mean value has been changed by 8% compared to the value obtained by forcing axial symmetry of the ow. The oscillations in the transfer are relatively small. This indicates that axisymmetric simulation can provide an acceptable estimation of the transfer when the bubble wake is destabilized. The reason is that the total transfer is mainly controlled by the transfer on the front part of the bubble not affected by the wake destabilization. A further inspection for such ow conditions has to be done in order to generalize this result but this was out of the scope of the present work.

Shloc Pe-1/2

1.5

6. Discussion
1

0.5

20

40

60

80

100

120

140

160

180

Fig. 13. Time evolution of the local Sherwood number Shloc Pe1=2 for Re 300, Sc 10 and w 2. Labels ad correspond to the time indicated in Fig. 11. (a) FFt 2.5, (b) t 7.5, (c) t 20, (d) t 50, (e) 2 2 t 75, B t 100.

The results discussed above are presented using the Sherwood number. It is the normalized total mass transferred at the bubble surface. Even if Sh could be seen as the mean transfer at the bubble surface, it does not really express the mass or heat exchange potential of a given bubble since surface of exchange increases as the bubble deforms. A direct application of the results presented above is their use to describe mass transfer in bubbly situations. The effect of bubble deformation for model transfer has been presented in many studies. See for example Nedeltchev et al. (2007) where the Higbie penetration theory is corrected by a correction factor attributed to bubble deformation and expressed as a function of the Eotvos number in order to match satisfactorily experiments.

0.8 0.6 0.4 0.2 0 0.2 0 80 50 100 150 200 250 300 350
CL CD

70

Sh

60

50 0 50 100 150 t U/b


Fig. 14. Evolution in time of the Sherwood number and the drag and lift coefcients for w 3, Re 300 and Sc 10. 3 2D Axisymmetric simulation.

200

250

300

350

B. Figueroa-Espinoza, D. Legendre / Chemical Engineering Science 65 (2010) 62966309

6307

We consider a gas void fraction aG . Vb being the bubble volume, the interfacial area is aI aG Ab =Vb . For spherical bubble, aI 6aG =deq . For a spheroidal bubble of aspect ratio w, the interfacial area is

aI aG
with

6 fa w deq " p!# 1 w w2 1 pLn p 4w1=3 w2 1 w w2 1

21

fa w

w2=3
2

22

The ux of concentration FL transferred per unit volume in the liquid is

FL aI kL cus cu1 aI

Sheq D cus cu1 deq

23

where cu1 is here the bulk concentration and kL Sheq D=deq is the mass transfer coefcient. FL describes the interfacial transfer of concentration per unit volume and is the classical source term in the Eulerian two-uid model concentration equation for the liquid (Fayolle et al., 2007). The interfacial transfer for mobile interface (i.e. clean bubble) is usually modeled considering the penetration theory proposed by Higbie (1935). If the contact or exposure time is chosen as deq/Vr where Vr is the relative velocity, the mass transfer coefcient is kL 2DV r =pdeq 1=2 and is B equivalent to the Boussinesq solution Sh 2Peeq =p1=2 with Peeq deq Vr =D. As shown above, this expression for the transfer is a satisfactory estimation of the transfer for oblate ellipsoidal bubbles if Reeq 4 O100. Indeed a correction f w can be introduced but it is less than 10% for Reeq 4O100 as discussed in the previous section. Finally, when bubbles deform, the interfacial transfer needs to be corrected by both the surface correction fa and the Sherwood correction f w as

bubbly ows. It has two contributions, one is due to the increase of the surface area and the second is due to the increase of the Sherwood number. Finding the optimal diameter to maximize the quantity of matter transferred has an important impact in industry for chemical, petrochemical, biochemical, and environmental applications. When using bubbles, there is a competition between small bubbles that produce a large area for transfer and a longer resident time and larger bubbles with a larger interfacial exchange due to a larger relative velocity. Using results presented above, the dependency of FL with the bubble diameter deq can be expressed and the impact of bubble deformation can be commented. For Re 4 100, the transfer can be described using the Boussinesq solution so that kL % 2DV r =deq p1=2 . The transfer depends on the relative velocity. In most of applications, the relative velocity is imposed by buoyancy. It can be estimated by using the drag solution for spheroidal bubble proposed by Moore (1965): Vr % d2 g=36nGw where g is the gravity and Gw, function eq given in Moore (1965), increases with bubble aspect ratio. Typically, G(1)1, G(2) 2.77 and G(3)5.49. Using these expressions the ux transferred per unit volume is  1=2 2 f wf w g FL w p aG a 1=2 Sc1=2 cus cu1 25 deq p Gw For given gas void fraction aG and physical properties, FL depends on bubbles diameter and their shape. The bubble deformation induces a contribution fF w fa wf wGw1=2 . The correction fF w is reported in Fig. 15 and is shown to decrease with the bubble aspect ratio. Since w increases with the bubble size, the optimal value of FL w is obtained by reducing the bubble size. Since the resident time of the bubbles in such a system decreases with the bubble size, the total mass transfer is improved when reducing bubble size. Other parameters have not been considered in this rough analysis: both gas solubility and bubble size depend on the ambient pressure, and can change with liquid depth. Small air bubbles can also transfer all their oxygen before leaving the system (see Motarjemi and Jameson, 1978). The main purpose here was to comment on the effect of the bubble aspect ratio.

FL w fa wf wFL w 1

24

The corresponding corrections are reported in Fig. 15. We can observe that fa w gives the dominant correction to FL . The total effect cannot be neglected since for w 2 the correction is more than 10% and is about 35% for w 3. This indicates that bubble deformation has to be considered when modeling mass transfer in
1.4 1.3 1.2 1.1 1 0.9 0.8 0.7 0.6 0.5

7. Conclusion The effect of bubble deformation on mass or heat transfer was estimated using numerical simulations of single axisymmetric (oblate) spheroidal bubbles. Our results are consistent with previous investigations from the hydrodynamical and mass/heat transfer viewpoints. Available theoretical and experimental results for the mean and local Sherwood (or Nusselt) numbers were compared to our simulations for spherical bubbles. The local Sherwood numbers of deformed bubbles showed signicant differences with theoretical predictions which are based on potential ow calculations. Even if there is a zero shear boundary condition at the bubble surface, the boundary layer in this case w 4 1 has a more complicated structure than the one corresponding to a spherical bubble. The appearance of a standing eddy behind the bubble has two effects: the recirculation tends to increase the transfer for short times, decreasing as the standing eddy concentration homogenizes, so the transient has a long duration. It appears that the use of the equivalent diameter as the characteristic length is the more appropriate to describe the transfer. The effect of bubble aspect ratio on the Sherwood number has been analyzed. At rst order the extension of Boussinesq expression using the equivalent diameter can be used for practical purposes. An additional effect of bubble aspect ratio has been found to correct this correlation by a factor of about

1.2

1.4

1.6

1.8

2.2

2.4

2.6

2.8

Fig. 15. Effect of the aspect ratio on the interfacial transfer FL . fa w, f w, correction factor fa wf w, F F F total contribution fa wf w=Gw1=2 .

6308

B. Figueroa-Espinoza, D. Legendre / Chemical Engineering Science 65 (2010) 62966309

510%. The evolution of the correction factor that compares the Sherwood number to the one of a sphere with same equivalent Peclet number has been presented and described using simple correlations. The implementation of these results in EulerEuler simulations of mass transfer has been discussed. It appears that the modication of the interfacial area combined with the modication of the Sherwood number gives a signicant contribution to the interfacial source term in the equation of the concentration. Additional effect in bubble ows may have also some relevant inuence on the transfer. Bubble wake interactions are known to generate induced turbulence and additional dissipation (Riboux et al., 2010). The consequence is a decrease of the bubble terminal velocity when the void fraction increases as reported by several experimental studies. The corresponding induced effects (liquid agitation and/or turbulence, bubbles interaction, shape oscillation, etc.) on the mass transfer is still not understood and needs some further investigations to be characterized. For such studies, the numerical strategy (adaptative grid) used in this study is not adapted and volume of uid method, front tracking method or level set method are more appropriate to consider several bubbles (Radl et al., 2008).

Nomenclature Ab a b c bubble surface area minor semi-axis of the ellipsoid major semi-axis of the ellipsoid dimensionless concentration (or temperature) of a particular chemical species in the mixture drag coefcient lift coefcient chemical species concentration far from the interface or bulk concentration chemical species concentration at the interface diffusion coefcient equivalent diameter unit vector along the x-direction unit vector along the y-direction correction factor Sheq w=Sheq w 1 total dimensional force acting on the bubble FD ~u:~x drag force F e FL ~u:~y lift force contribution along the F e ~y - direction e computational index for the numerical grid computational index for the numerical grid characteristic length scale recirculation characteristic length number of nodes in curvilinear reference frame in the Z direction (grid type A behind the bubble) number of nodes in curvilinear reference frame in the Z direction (grid type A in front of the bubble) number of nodes in the bubble surface for grid type A number of nodes behind the bubble surface for grid type A number of nodes in the bubble surface for grid type B number of nodes before and after the bubble for grid type B

CD CL cu1 cus D deq 2req ~x e ~y e f w ~u F FD FL i j L Lw NZ

number of nodes to dene the exit of the numerical grid type B Pe Peclet number based on Lb Peeq Peclet number based on Ldeq pu the dimensional pressure Re Reynolds number based on Lb Reeq Reynolds number based on Ldeq req equivalent radius R1 extent of the numerical domain Rec critical Reynolds number Sc Schmidt number Sh Sherwood number based on Lb Sheq Sherwood number based on Ldeq Shloc local Sherwood number Sh* normalized Sherwood number (for convergence testing) tu dimensional time t t U1 tu=b dimensionless time tw characteristic time of the recirculation transitory U1 velocity of the ow far from the bubble 2 We deq rU1 =s Weber number w w b=a aspect ratio n liquid cinematic viscosity ~u dimensional position vector x ~ ~ ~u=b dimensionless position vector x x x ~u dimensional velocity u ~ ~ ~u=U1 dimensionless velocity u u u r liquid density t viscous part of the stress tensor Z curvilinear coordinate that follows the outward normal to the bubble surface dimensional diffusive boundary layer dud curvilinear coordinate that follows the bubble x surface p a0 a0 1= w2 1, geometric parameter for the ellipsoidal coordinate transformation D dimensional distance (normalized by b) from the nearest numerical node to the bubble wc critical aspect ratio on the onset of the recirculation angle describing the bubble surface y

ne

Acknowledgements This work was granted the support of the National Research Agency of France (Agence Nationale de la Recherche), Project O2STAR, Reference: ANR-07-ECOT-007-01. D.L. would like to thank A. Cockx and colleagues from the CNRS Federation FERMaT, the LISBP and the Cemagref dAntony for making this collaboration possible. B.F. thanks Annaig Pedrono and the IMFT personnel for their help with the code and their suggestions during his stay in Toulouse.

Nb Nx nx ny

Appendix A See Table 1.

B. Figueroa-Espinoza, D. Legendre / Chemical Engineering Science 65 (2010) 62966309

6309

Table 1 Preliminary grid tests. Re 1 1 1 1 1 1 300 300 300 300 300 300 1 1 1 1 1 1 300 300 300 300 300 300 Sc 1 1 1 500 500 500 1 1 1 500 500 500 1 1 1 500 500 500 1 1 1 500 500 500 Grid type A A A A A A A A A A A A B B B B B B B B B B B B Nd 4 25 4 25 4 25 8 10 13 8 13 18 1 2 5 44 50 4 50 14 19 24 13 18 23 1 2 5

d=R
0.002 0.001 0.0005 0.002 0.001 0.0005 0.002 0.001 0.0005 0.002 0.001 0.0005 0.002 0.001 0.0005 0.002 0.001 0.0005 0.002 0.001 0.0005 0.002 0.001 0.0005

CD 17.649 17.666 17.633 17.682 17.638 18.7991 0.141 0.141 0.141 0.141 0.141 0.141 17.649 17.666 17.685 17.649 17.667 17.686 0.140 0.140 0.140 0.140 0.140 0.140

CD (Mei et al., 1994)

Sh 2.363 2.369 2.369 16.973 16.950 17.788 19.349 19.351 19.353 395.8 399.8 400.5 2.364 2.364 2.365 16.93 16.94 16.95 19.37 19.34 19.34 393.6 399.9 401.8

17.575

17.575

0.139

0.139

17.575

17.575

0.139

0.139

References
Adoua, R., Legendre, D., Magnaudet, J., 2009. Reversal of the lift force on an oblate bubble in a weakly viscous linear shear ow. J. Fluid Mech. 628, 2341. Blanco, A., Magnaudet, J., 1995. The structure of the axisymetric high-Reynolds number ow around an ellipsoidal bubble of xed shape. Phys. Fluids 7 (6), 12651274. Boussinesq, J., 1905. Calcul du pouvoir refroidissant des courants uides. J. Math. Pure Appl. 6, 285332. Brenner, H., 1963. Forced convection heat and mass transfer at small Peclet numbers from a particle of arbitrary shape. Chem. Eng. Sci. 18, 109122. Calmet, I., Magnaudet, J., 1997. Large-eddy simulation of high-Schmidt number mass transfer in a turbulent channel ow. Phys. Fluids 9 (2), 438455. Clift, R., Grace, J.R., Weber, M.E., 1978. Bubbles, Drops and Particules. Academic Press, New York. Darmana, D., Deen, N.G., Kuipers, K.A.M., 2006. Detailed 3D modeling of mass transfer processes in two-phase ows with dynamic interfaces. Chem. Eng. Technol. 29, 10271033. Ellingsen, K., Risso, F., 2001. On the rise of an ellipsoidal bubble in water: oscillatory paths and liquid-induced velocity. J. Fluid Mech. 440, 235268. Favelukis, M., 2010. Mass transfer around oblate spheroidal drops at low Reynolds numbers. Chem. Eng. Sci. 65, 38083813. Favelukis, M., Ly, C.H., 2005. Unsteady mass transfer around spheroidal drops in potential ow. Chem. Eng. Sci. 60, 70117021. Fayolle, Y., Cockx, A., Gillot, S., Roustan, M., Heduit, A., 2007. Oxygen transfer prediction in aeration tank using CFD. Chem. Eng. Sci. 62, 71637171. Higbie, R., 1935. The rate of absorption of a pure gas into a still liquid during short periods of exposure. Trans. A.I.Ch.E. 31, 365389. Koynov, A., Khinast, J.G., Tryggvason, G., 2005. Mass transfer and chemical reactions in bubble swarms with dynamic interfaces. A.I.Ch.E. J. 10, 27862800. Leclair, B.P., Hamielec, A.E., 1971. Viscous ow through particle assemblages at intermediate reynolds numbersa cell model for transport in bubble swarms. Can. J. Chem. Eng. 49, 713720. Legendre, D., 2007. On the relation between the drag and the vorticity produced on a clean bubble. Phys. Fluids 19, 018102. Legendre, D., Magnaudet, J., 1998. The lift force on a spherical body in a viscous linear shear ow. J. Fluid Mech. 368, 81126. Legendre, D., Magnaudet, J., 1999. Effet de lacceleration dun ecoulement sur le  transfert thermique ou massique a la surface dune bulle spherique. C. R. ACad. Sci. Paris Srie IIb 327, 6370. Legendre, D., Magnaudet, J., Boree, J., 1998. Thermal and dynamic evolution of a spherical bubble moving steadily in a superheated or subcooled liquid. Phys. Fluids 10, 12561272.

Legendre, D., Magnaudet, J., Mougin, G., 2003. Hydrodynamic interactions between two spherical bubbles rising side by side in a viscous liquid. J. Fluid Mech. 497, 133166. Lochiel, A.C., Calderbank, P.H., 1964. Mass transfer in the continuous phase around axisymmetric bodies of revolution. Chem. Eng. Sci. 19, 471484. Magnaudet, J., Eames, I., 2000. The motion of high-Reynolds-number bubbles in inhomogeneous ows. Ann. Rev. Fluid Mech. 32, 659708. Magnaudet, J., Rivero, M., Fabre, J., 1995. Accelerated ows past a rigid sphere or a spherical bubble. Part 1. Steady strainning ow. J. Fluid Mech. 284, 97135. Magnaudet, J., Mougin, G., 2007. Wake instability of a xed spheroidal bubble. J. Fluid Mech. 572, 311337. Mei, R., Klausner, J.F., Lawrence, C.J., 1994. A note on the history force on a spherical bubble at nite reynolds number. Phys. Fluids A 6, 418420. Michaelides, E.E., 2003. Hydrodynamic force and heat/mass transfer from particles, bubbles, and dropsthe freeman sholar lecture. J. Fluids Eng. 125, 209238. Moore, D.W., 1963. The boundary layer on a spherical gas bubble. J. Fluid Mech. 16, 161176. Moore, D.W., 1965. The velocity of rise of distorted gas bubble in a liquid of small viscosity. J. Fluid Mech. 23, 749766. Motarjemi, M., Jameson, J., 1978. Mass transfer from very small bubblesthe optimum bubble size for aeration. Chem. Eng. Sci. 33, 14151423. Nedeltchev, S., Jordan, U., Schumpe, A., 2007. Correction of the penetration theory based on mass-transfer data from bubble columns operated in the homogeneous regime under high pressure. Chem. Eng. Sci. 62, 62636273. Oellrich, L., Schmidt-Traub, H., Brauer, H., 1973. Theoretische berechnung des stofftransports in der umgebung einer einzelblase. Chem. Eng. Sci. 28, 711721. Radl, S., Koynov, A., Tryggvason, G., Khinast, J.G., 2008. DNS-based prediction of the selectivity of fast multiphase reactions: hydrogenation of nitroarenes. Chem. Eng. Sci. 63, 32793291. Riboux, G., Risso, R., Legendre, D., 2010. Experimental characterization of the agitation generated by bubbles rising at high Reynolds number. J. Fluid Mech. 643, 509539. Ruckenstein, E., 1959. On heat transfer between vapour bubbles in motion and the boiling liquid from which they are generated. Chem. Eng. Sci. 10, 2230. Sanada, T., Shirota, M., Watanabe, M., 2007. Bubble wake visualization by using photochromic dye. Chem. Eng. Sci. 62, 72647273. Takemura, F., Yabe, A., 1998. Gas dissolution process of spherical rising bubbles. Chem. Eng. Sci. 53 (15), 26912699. Winnikow, S., 1967. Letters to the editor. Chem. Eng. Sci. 22, 477. Zenit, R., Magnaudet, J., 2008. Path instability of rising spheroidal air bubbles: a shape-controlled process. Phys. Fluids 20, 061702.

You might also like