PHYSICS OF FLUIDS VOLUME 16, NUMBER 8 AUGUST 2004 R. Balasubramaniam a) National Center for Microgravity Research on Fluids and Combustion, NASA Glenn Research Center, Cleveland, Ohio 44135 R. Shankar Subramanian Clarkson University, Potsdam, New York 13699 Received 21 April 2004; accepted 10 May 2004; published online 6 July 2004 We analyze the velocity and temperature fields due to thermocapillary convection around a gas bubble that is stationary in a liquid. A linear temperature field is imposed in the undisturbed liquid. Our interest is in the effect of convective transport of momentum and energy on the velocity and temperature fields. We assume that buoyant convection is negligible. The relevant Reynolds and Marangoni numbers are assumed to be small compared with unity. When the Reynolds and Marangoni numbers are set equal to zero, the steady velocity field far from the bubble decays inversely with distance from it. We show that this behavior of the velocity field, coupled with the linear variation of the imposed temperature far away from the bubble, leads to an ill-posedness in the problem for the perturbation temperature field at steady state for small Marangoni numbers, when the Reynolds number is zero. We also consider the effect of convective transport of momentum for small Reynolds number, when the Marangoni number is zero. In this case, the velocity field far away from the bubble decays inversely as the distance from the bubble even in the presence of inertia. It follows that, for a fixed Prandtl number, when both the Marangoni and Reynolds number are nonzero, the behavior of the velocity field far from the bubble continues to yield an ill-posed problem for the steady perturbation temperature field. Therefore, inertia cannot relieve the ill-posedness of the perturbation problem for the steady temperature field. In order to understand the origin of this ill-posedness, we analyze a related problem involving the unsteady development of the temperature field in the liquid from the moment that a point force is applied to it, and held constant subsequently. We show that including the effect of convective transport of energy by a perturbation expansion leads to a temperature field in the liquid that is always time dependent. The temperature field never achieves a steady state. 2004 American Institute of Physics. DOI: 10.1063/1.1768091 I. INTRODUCTION Our goal in this paper is to examine the effects of inertia and convective transport of energy on the velocity and temperature fields induced by thermocapillarity around a stationary bubble. When a gas bubble or an immiscible liquid drop is present in a liquid in which a temperature gradient is imposed, motion is induced in the fluids due to the variation of surface tension that is caused by the temperature variation on the interface. If the bubble is free, it would move typically toward warm liquid. This motion is called thermocapillary migration, and is important in the reduced gravity environment of orbiting spacecraft for the processing of materials and other applications ee Subramanian and Balasubramaniam 1. Thermocapillary convection around stationary bubbles was first encountered in the pioneering experimental and theoretical study performed by Young, Goldstein, and Block. 2 These authors conducted experiments on small air bubbles in viscous silicone oils, held between the anvils of a machinist s a Author to whom correspondence should be addressed. Electronic mail: bala@grc.nasa.gov micrometer. The fluid was heated from below, and the Rayleigh number was less than the critical value needed to excite cellular convection in the liquid. Small bubbles were observed to move downward due to thermocapillarity. For large bubbles, the buoyancy force on the bubble overcame the effect of thermocapillarity and they were observed to move upward. Depending on the temperature gradient, bubbles of a certain size were observed to be nearly stationary. The theoretical description of the motion developed by Young et al., neglecting the effect of convective transport of momentum and energy, predicted that this critical size was independent of the viscosity of the liquid, and that it scaled linearly with the applied temperature gradient. These predictions were experimentally verified by Young et al. Thermocapillary flow around moving and stationary bubbles and drops is important in many applications in normal and reduced gravity. Kassemi et al. 3 state that three recent solidification experiments conducted in reduced gravity encountered unwanted effects from voids and bubbles in the melt. While free bubbles are expected to move due to the temperature gradient in the melt during solidification, bubbles and voids that adhere to the container will be stationary, which is one of the cases modeled by Kassemi et al. 1070-6631/2004/16(8)/3131/7/$22.00 3131 2004 American Institute of Physics
3132 Phys. Fluids, Vol. 16, No. 8, August 2004 R. Balasubramaniam and R. S. Subramanian An interesting flow around a stationary bubble was observed in the reduced gravity experiments in subcooled boiling performed by Straub et al. 4 These authors formed a stationary bubble at the tip of a heated thermistor immersed in the liquid Freon R11. The flow around the bubble was visualized by observing the system in an interferometer. A strong jet flow away from the bubble was clearly seen, which the authors attribute to thermocapillarity. Straub 5 suggests that thermocapillary flow arises in subcooled boiling due to the action of dissolved gases that are always present. Kim, Benton, and Wisniewski 6 also observed flows that are attributed to thermocapillarity in their boiling experiments. Prinz and Romero 7 mention that the antiparallel alignment of gravitational settling and thermocapillary motion that yields nearly stationary drops can be used in the processing of advanced alloys that are used in self-lubricating bearings. Visualization of the flow around stationary bubbles from space experiments is reported in Wozniak et al. 8 In these experiments, air bubbles were held stationary at the tip of a needle submerged in a silicone oil, which was subjected to a uniform temperature gradient. The thermocapillary flow was visualized by following the motion of tracer particles that were added to the liquid. The temperature field was visualized by interferometry. Results from experiments on bubbles attached to a wall under normal gravity have been reported in Kao and Kenning, 9 Raake et al., 10 and Kassemi and Rashidnia, 11 and in reduced gravity by Wozniak et al. 12 Available theoretical work on thermocapillary flow driven by a stationary bubble is limited to that presented by Young et al. for conditions when convective transport of momentum and energy are neglected altogether. Numerical solutions that include such effects are available for the case of hemispherical bubbles attached to a wall Larkin, 13 Kao and Kenning 9, but none for a bubble of a spherical shape in an unbounded liquid. Therefore, our objective in this work is to explore the influence of convective transport using a perturbation approach. We first show that inclusion of convective transport effects in the steady problem by a perturbation approach leads to ill-posedness of the problem for the temperature correction field. We then demonstrate that this is likely because of the lack of existence of a steady solution by solving a related unsteady problem. II. FORMULATION AND ANALYSIS OF THE STEADY STATE PROBLEM Consider the steady flow around a spherical gas bubble of radius R 0 that is present in a liquid of infinite extent. The liquid has a density, viscosity, thermal conductivity k, and thermal diffusivity. The viscosity and thermal conductivity of the gas are assumed negligible compared with the corresponding properties of the liquid. This decouples the transport problem in the liquid from that within the bubble. The rate of change of surface tension with temperature is denoted by T and is assumed to be a negative constant. All the physical properties of the liquid are assumed to be constant. A constant temperature gradient with magnitude G is assumed to be imposed in the undisturbed continuous phase. We assume that the bubble is not in motion. An example would be the case when the bubble is attached to the tip of a needle. When gravity is present, buoyant convection is neglected. A suitable reference velocity for the motion in the liquid may be obtained from the tangential stress balance at the bubble surface, where the jump in the tangential stress across the interface equals the thermocapillary stress at the interface. This velocity scale is v 0 TGR 0. 1 The dimensionless parameters that are important are the Reynolds and Marangoni numbers. Definitions of these parameters that are based on the above velocity scale and the properties of the liquid are given below: Re v 0R 0, 2 Ma v 0R 0. 3 The capillary number Cav 0 /, where denotes the surface tension, is a parameter that influences the shape of the bubble. We assume Ca1, which permits us to neglect deformation from the spherical shape. A spherical polar coordinate system is used, with the origin located at the center of the bubble. The radial coordinate, scaled by R 0,isrand the polar coordinate measured from the direction of the temperature gradient is. The azimuthal coordinate is labeled. The temperature in the liquid is scaled by subtracting the temperature in the undisturbed continuous phase at the location of the bubble and dividing by the quantity GR 0 : T T*T 0, 4 GR 0 where T* is the physical temperature. The streamfunction is scaled by dividing the physical streamfunction by v 0 R 2 0. The scaled radial and tangential velocities are related to the dimensionless streamfunction via u 1 r 2, 1 v r1 2 1/2 r, 6 where scos. Our goal is to determine the effect of convective transport of momentum and energy on the velocity and temperature fields around the bubble. We assume that these effects are weak and attempt to determine the influence of convective transport by a perturbation analysis. Therefore, the Reynolds and Marangoni numbers will be assumed to be parameters whose values are small compared with unity. The governing equations for the streamfunction and temperature fields in the liquid are 5
Phys. Fluids, Vol. 16, No. 8, August 2004 3133 Re 1 r 2 Ma 1 r 2 1 r r E2 r E 4, T r T r 2 T, E2 2 r 2 E2 where E 2 ( 2 /r 2 )(1 2 )/r 2 ( 2 / 2 ), and s 1 2 r 2 1/r 2 /rr 2 /r/1 2 /. The boundary conditions are 1,s0, r 1 r 2 r 1,s1 2 T 1,s, T r 1,s0, 7 8 9 10 11 T rs as r, 12 1 r 2, 1 0 as r. 13 r1 2 1/2 r Equations 9 11 represent the kinematic, tangential stress, and adiabatic conditions, respectively, at the surface of the bubble. Far away from the bubble, the temperature gradient is assumed to be uniform and the liquid is expected to be quiescent, as described in Eqs. 12 and 13, respectively. A. Solution for Re, MaÄ0 When the Reynolds and Marangoni numbers are both set equal to zero, the effect of convective transport of momentum and energy can be neglected. We designate 0 and T 0 as the streamfunction and temperature fields in this limit. The solution is 0 1 4 r 1 r 1 2, 14 T 0 2 r 1 s. 15 2r The velocity field can be obtained from Eqs. 5, 6, and 14 and is seen to decay as 1/r far away from the bubble. B. The effect of convective transport of energy for ReÄ0 First, we explore the effect of convective transport of energy when the Reynolds number is zero. The Marangoni number is assumed to be a small parameter. The streamfunction and temperature fields are expanded in perturbation series as 0 Ma 1, TT 0 MaT 1, 16 17 where 0 and T 0 are the leading order fields given in Eqs. 14 and 15. The governing equations at O(Ma) are E 4 1 0, 2 T 1 u 0 " T 0 1 r 2 0 1 4 1 r 1 r 3 1 2r 4 1 T 0 r 0 T 0 r 6 2r s2 4 1 r 3 r 3 5 2r 4 3 2r 6. 18 19 The boundary conditions on 1 and T 1 can be obtained from Eqs. 9 to 13. One important condition is that T 1 /r 0 as r. The inhomogeneity (1/4r)(1 2 ) in Eq. 19 has a particular solution of the form T 1p 1 16 r32 20 that cannot be canceled by a homogeneous solution of Eq. 19. Therefore T 1 /r does not vanish as r and the solution for the temperature field is unacceptable as it violates the boundary condition. The inability of a higher order solution to satisfy the boundary condition as r is a symptom that the perturbation in Ma is singular Proudman and Pearson, 14 Acrivos and Taylor 15. Let us first consider this possibility. We need to investigate whether a scaling of the radial coordinate r can be determined when r is large i.e., in an outer region such that convective and conductive transport of energy in Eq. 8 are in balance, when Ma1. Because uo(1/r), and T O(r) at large values of r, we see that u" TO(T/r 2 ) O(1/r). Also, 2 TO(T/r 2 )O(1/r). Thus, u" T 2 T for large r and Mau" T 2 T for large r. The implication is that convective transport effects are small compared with conduction everywhere when Ma1. Another way to cast this argument is to attempt a rescaling of the radial coordinate using Ma r. It can be seen that no value of can be found by requiring the convective transport term and the conduction term to balance each other. Therefore, an outer region where conductive and convective transport of energy are in balance does not appear to exist. Because we are unable to find an outer region where convective and conductive transport of energy are in balance, the particular solution for T 1 Eq. 20 forces us to conclude that the problem for the temperature field might be ill-posed. The inhomogeneity that leads to this particular solution can be traced to the interaction of the applied uniform temperature gradient and the 1/r decay of the leading order velocity field. We note that a similar behavior of the far-field temperature was encountered by Zhang, Subramanian, and Balasubramaniam 16 in the analysis of the motion of a drop in a vertical temperature gradient. The leading order velocity field decayed as 1/r for Re0. Even though an outer region for the temperature field with the scaling Mar was found to exist, the motion of the drop caused a first correction to
3134 Phys. Fluids, Vol. 16, No. 8, August 2004 R. Balasubramaniam and R. S. Subramanian the outer temperature field analogous to the temperature field T 1 ) to behave as ln for large. This unacceptable behavior was traced to the interaction of the 1/r decay of the velocity field and the applied temperature gradient. The remedy suggested by Zhang et al. was to include the effect of a small amount of convective transport of momentum as a perturbation. With the effect of inertia included in an Oseen type of analysis, they obtained the decay of the leading order velocity field to be O(1/r 2 ), which is shown to remove the ln behavior of the first correction to the outer temperature field. We therefore explore the effect of inertia next, to determine its influence on the behavior of the far-field temperature. C. The effect of convective transport of momentum for MaÄ0 In order to explore the effect of inertia, we first consider a case with Ma0. The velocity and temperature fields are then decoupled. For any value of the Reynolds number, the temperature field is that given in Eq. 15. The streamfunction is now expanded as follows: 0 Re 1, 21 where 0 is given in Eq. 14. The governing equation for 1 is E 4 1 1 r 2 2 r 2 0 E 2 0 0 r r s 0 1 2 r 1 0 r 3 4 1 r 5 1 r 3 s1 2. The boundary conditions are 1 1,s0, r 1 r 2 1 r 1,s0, 1 1 r 2, 1 1 r1 2 1/2 r E 2 0 22 23 24 0 as r. 25 Thus 1 satisfies homogeneous boundary conditions and is driven by the inhomogeneity in Eq. 22. A particular solution is 1p (1/32)r(1/r)(1 2 ), and a complete solution is 1 1 1609 1 r 2 1 32 r 1 r1 2. 26 Unlike the solution for the O(Ma) temperature field in the presence of convective transport of energy, we see that the streamfunction at O(Re), which includes the effect of convective transport of momentum, is well behaved for large r and is consistent with the boundary condition as r. Therefore, the regular perturbation solution with Re as the small parameter is successful up to this order. From Eqs. 14 and 26 we see that the velocity field decays as 1/r as r at O(1) as well as at O(Re). Therefore, inclusion of inertia has not altered the manner in which the velocity field decays far away from the bubble. Furthermore, in the case when Ma and Re are both nonzero, for a fixed Prandtl number, an expansion can be written in the Reynolds number Re. It is evident that using such an expansion will yield an equation for the first correction in the temperature field that is the same as Eq. 19, with the Prandtl number multiplying the inhomogeneity, even though Eq. 18 would be altered. The difficulty noted with respect to the behavior of the particular solution of Eq. 19, given in Eq. 20, will persist. Hence such a perturbation will fail. III. UNSTEADY DEVELOPMENT OF THE DISTURBANCE TEMPERATURE FIELD DUE TO A POINT MOMENTUM SOURCE One possible reason for the ill-posedness of the perturbation problem for the steady temperature field is that the temperature and velocity fields around the bubble may be inherently unsteady in an unbounded medium, and a steady state solution does not exist. In the remainder of this work, we demonstrate that this is a strong possibility. While it is straightforward to pose the unsteady problem, analytical solution has proven to be difficult. Therefore, we have chosen to show here that a steady solution does not exist for the temperature field in a closely related problem involving steady state flow in an unbounded medium. This is the problem of a steady point source of momentum located at the position occupied by the center of the bubble. The point force is assumed to act in the direction opposite to the temperature gradient in the undisturbed liquid. At the surface of the bubble, the shear stress communicated to the liquid because of the surface tension gradient adds momentum to the liquid. Thus, the bubble may be viewed as a source of momentum that drives motion in the liquid. At large distances from the bubble, the influence of the finite size of the bubble should become small, and as r the solution for the present problem should approach that for a point source of momentum. A. Steady flow from a point momentum source First, we discuss the steady velocity field from a point momentum source. For an arbitrary Reynolds number, an exact solution of the steady Navier Stokes equation for a point momentum source was found independently by Landau ee Batchelor 17 and Squire. 18 The solution is *2r* 12 1c, 27 where * and r* are the physical streamfunction and the radial coordinate, respectively, and is the kinematic viscosity of the liquid. The constant c is related to the nondimensional force F FF*/(2) exerted on the fluid at the origin by the following expression:
Phys. Fluids, Vol. 16, No. 8, August 2004 3135 32 1c c 3 ln c2c 2c 41c2 81cF. 28 For small Reynolds number, Batchelor shows that F8/c 1. Expanding Eq. 28 for large c, weget c 8 17 1 F 120 FOF3. 29 Using Eq. 29 in Eq. 27 and expanding for small F yields * 1 4 r*f12 1 32 r*f2 s1 2 OF 3. 30 Scaling the streamfunction by v 0 R 2 0, we see that the first term in the solution for the point momentum source is consistent with the far-field leading order Stokes flow result for the streamfunction in the flow caused by the bubble Eq. 14 when F is identified with Rev 0 R 0 /. With this choice, the dimensionless streamfunction is 1 4 r12 Re 32 rs12 ORe 2. 31 For large r, we see that the expression for 1 for the stationary bubble Eq. 26 far from the bubble is consistent with the corresponding contribution to the solution for the streamfunction for the point momentum source in Eq. 31. From Eq. 27 it can be inferred that the velocity field behaves as 1/r for any value of the Reynolds number for the point momentum source. Because the stationary bubble should appear as such a source as r, it must be true that the velocity field far from the bubble also should decay as 1/r for any value of the Reynolds number. B. Unsteady temperature field In the following development, we shall use just the leading term in the expansion in the Reynolds number given in Eq. 31. This corresponds to Stokes flow due to a point source. 19 We assume that a point source of momentum is introduced at the origin at time t*0 and held constant thereafter. In the limit Re 0, the steady solution is given in physical variables as * F* 8 r*12. 32 For convenience we use physical variables in this analysis. The governing equation for the temperature field in the liquid may be written as * T* t* 1 r* 2 The boundary conditions are T* * T* r* r* 2 T*. 33 T* T 0 Gr*s as r*, 34 T*t*,0,s is bounded. The initial condition is T*0,r*,sT 0 Gr*s. 35 36 Let where T*T 0 Gr*sGt*f,s, r* 2t*. The governing equation for f is 2 f f Pe 4 2 1 2 f 2 21 2 2 f 12 f 1 2, f 37 38 39 where PeF*/(2) is a Peclet number. The boundary conditions that f (,s) must satisfy are f 0 as, 40 f 0,s is bounded, f,1 is bounded. We expand f in a perturbation series as follows: 41 42 f Pe 4 f 0oPe. 43 The equation for f 0 is 2 f 0 f 0 2 12 1 2 f 2 0 1 2 f 0. 44 The boundary conditions on f 0 can be obtained from Eqs. 41 and 42. The solution is f 0 f 0,0 P 0 f 0,2 P 2, f 0,0 4 3 4 3 1 1 e2 2 erf, f 0,2 3 4 9 1 2 2 4 e 2 1 3 3 43 4 erf 3 2 3 e 2. 45 46 47 From the solution given above, the temperature field T* may be expanded for large t* as
3136 Phys. Fluids, Vol. 16, No. 8, August 2004 R. Balasubramaniam and R. S. Subramanian 2 r* T*T 0 Gr*sGt*Pe 3 6t* r* 24t* r*2 18t* O 1 t* 2P 2 r*2 45t* O 1 t* 2 OPe 2. 48 From Eq. 48 we see that for large t*, the perturbation from the undisturbed temperature field is proportional to t*. Therefore, the temperature field is always unsteady in an unbounded system, and a steady state is never achieved. IV. DISCUSSION We have determined that when the Reynolds and Marangoni numbers are zero, the velocity field caused by a stationary bubble due to thermocapillarity decays as the inverse of the distance from the bubble, far away from it. When the effect of convective transport of energy is considered as a small perturbation Ma1, we see that at steady state we are unable to satisfy the boundary condition on the perturbation temperature field far away from the bubble. Thus the problem for the steady temperature field becomes ill-posed. We tried to relieve the ill-posedness heuristically by supposing that an outer region might exist far away from the bubble where convective and conductive transport of energy are of the same order of magnitude. When the velocity field decays as 1/r, however, we are unable to find such an outer region. We then explored the effect of inertia as a small perturbation Re1 to determine its influence on the behavior of the velocity field far away from the bubble. This was motivated by the finding of Zhang et al. 16 that for a drop moving in a temperature gradient, a physically unacceptable far-field behavior of the perturbation temperature is predicted when the Reynolds number is zero for which the velocity field decays as 1/r), and the convective effects are included as a small perturbation in the Marangoni number. When inertial effects are included as a small perturbation in the Reynolds number, Zhang et al. show that the velocity field decays as O(1/r 2 ), and relieves the ill-posedness of the temperature field problem. We find that for the stationary bubble, the velocity field decays as O(1/r) far away from the bubble even in the presence of inertia. By comparing the velocity field far from the bubble with that generated by a point source of momentum at steady state, we showed that the stationary bubble, viewed from far away, is equivalent to a point source of momentum, for which it can be established that the velocity field varies as 1/r regardless of the Reynolds number. Therefore, the effect of inertia cannot relieve the ill-posedness of the problem for the steady temperature field for the thermocapillary flow around a stationary bubble. Our hypothesis is that the effect of convective transport of energy renders the temperature field inherently unsteady around the stationary bubble in an unbounded liquid. To explore this possibility, we have analyzed here the simpler problem of the evolution of the unsteady temperature field in a fluid that has a uniform temperature gradient in the undisturbed state, when the fluid is subjected to a point force that causes a steady Stokes flow. The point force is applied at time zero, and thereafter is held constant. In the analysis, the effect of convective transport of energy is treated as a small perturbation Pe1. We find that the perturbation to the undisturbed temperature field in this problem is always unsteady, and never achieves a steady state. Because the steady flow field far from the stationary bubble approaches that due to a steady point source of momentum, this is compelling evidence that the forever-evolving temperature field induced by the effect of convective transport of energy is the reason for the ill-posedness of the problem for the steady perturbation temperature field caused by a stationary bubble. Our analysis does not include the effect of free convection, and is therefore applicable only when the buoyancy force can be neglected, such as in the reduced gravity environment aboard spacecraft. It is conceivable that if the effect of free convection was included, then the ill-posedness of the problem for the steady temperature field might disappear. The role of other physical effects not considered in the analysis, for example, temperature dependent physical properties, also needs to be explored in relieving the ill-posedness of the steady temperature field. Of course, an outer boundary, no matter how far away, should relieve the ill-posedness by restricting the domain to finite values of the radial coordinate r, as we have shown in Ref. 20; in that paper, we analyzed the Stokes problem of a bubble held stationary within a spherical container and reported the solution for the leading order streamfunction and temperature fields and the first temperature perturbation field solution, using the Marangoni number as a perturbation parameter. Young et al. did not need to consider such complications in their analysis for a stationary gas bubble to interpret their experimental results because their theory is only valid for ReMa0. The theory has no difficulty in this limit. The problem arises only when we attempt to incorporate nonzero convective transport of energy. ACKNOWLEDGMENT We have benefited from discussions with Professor M. Lawrence Glasser, Department of Physics, Clarkson University, regarding the unsteady problem for a stationary bubble. 1 R. S. Subramanian and R. Balasubramaniam, The Motion of Bubbles and Drops in Reduced Gravity Cambridge University Press, Cambridge, 2001. 2 N. O. Young, J. S. Goldstein, and M. J. Block, The motion of bubbles in a vertical temperature gradient, J. Fluid Mech. 6, 350 1959. 3 M. Kassemi, S. Barsi, M. Kaforey, and D. Matthiesen, Effect of void location on segregation patterns in microgravity solidification, J. Cryst. Growth 225, 5162001. 4 J. Straub, G. Picker, J. Winter, and M. Zell, Effective cooling of electronic components by boiling phase transition in microgravity, Acta Astronaut. 40, 1191997. 5 J. Straub, Origin and effect of thermocapillary convection in subcooled boiling: observations and conclusions from experiments performed at microgravity, Ann. N.Y. Acad. Sci. 974, 348 2002. 6 J. Kim, J. F. Benton, and D. Wisniewski, Pool boiling heat transfer on small heaters: effect of gravity and subcooling, Int. J. Heat Mass Transfer 45, 3919 2002.
Phys. Fluids, Vol. 16, No. 8, August 2004 3137 7 B. Prinz, A. Romero, and L. Ratke, Casting process for hypermonotectic alloys under terrestrial conditions, J. Mater. Sci. 30, 4715 1995. 8 G. Wozniak, R. Balasubramaniam, P. H. Hadland, and R. S. Subramanian, Temperature fields in a liquid due to the thermocapillary motion of bubbles and drops, Exp. Fluids 31, 842001. 9 Y. S. Kao and D. B. R. Kenning, Thermocapillary flow near a hemispherical bubble on a heated wall, J. Fluid Mech. 53, 715 1972. 10 D. Raake, J. Siekmann, and Ch-H. Chun, Temperature and velocity fields due to surface tension driven flow, Exp. Fluids 7, 164 1989. 11 M. Kassemi and N. Rashidnia, Steady and oscillatory thermocapillary convection generated by a bubble, Phys. Fluids 12, 3133 2000. 12 G. Wozniak, K. Wozniak, and H. Bergelt, On the influence of buoyancy on the surface tension driven flow around a bubble on a heated wall, Exp. Fluids 21, 1811996. 13 B. K. Larkin, Thermocapillary flow around hemispherical bubble, AIChE J. 16, 1011970. 14 I. Proudman and J. R. A. Pearson, Expansions at small Reynolds number for the flow past a sphere and a circular cylinder, J. Fluid Mech. 2, 237 1957. 15 A. Acrivos and T. D. Taylor, Heat and mass transfer from single spheres in Stokes flow, Phys. Fluids 5, 387 1962. 16 L. Zhang, R. S. Subramanian, and R. Balasubramaniam, Motion of a drop in a vertical temperature gradient at small Marangoni number the critical role of inertia, J. Fluid Mech. 448, 197 2001. 17 G. K. Batchelor, An Introduction to Fluid Dynamics Cambridge University Press, Cambridge, 1967. 18 H. B. Squire, The round laminar jet, Q. J. Mech. Appl. Math. 4, 321 1951. 19 J. Happel and H. Brenner, Low Reynolds Number Hydrodynamics Martinus Nijhoff, Dordrecht, 1986. 20 R. Balasubramaniam and R. S. Subramanian, Thermocapillary convection in a spherical container due to a stationary bubble, Adv. Space Res. 32, 137 2003.