Advances in Space Research 34 (2004) 16271634 www.elsevier.

com/locate/asr

Dynamic free-surface deformations in axisymmetric liquid bridges

Bok-Cheol Sim
a b

a,*

, Woo-Seung Kim a, Abdelfattah Zebib

Department of Mechanical Engineering, Hanyang University, Seoul 133-791, Korea Department of Mechanical and Aerospace Engineering, Rutgers University, Piscataway, NJ 08854-8058, USA Received 20 March 2003; received in revised form 1 September 2004; accepted 11 September 2004

Abstract Thermocapillary convection in a dierentially heated liquid bridge is investigated in a two-dimensional numerical study. The deformable free surface is obtained as a solution of the coupled transport equations at xed unit Prandtl number and aspect ratio. Only steady convection can be realized in this axisymmetric computation with either non-deformable or deformable surfaces. Dynamic free-surface deformations do not induce transitions to oscillatory convection even at large Reynolds numbers. Free surfaces are convex near the cold wall due to large surface pressures at the stagnation-point. Near the hot wall they change from concave to convex with increasing Re due to large normal viscous stresses. Large Capillary numbers cause additional surface ripples there while having little eect on the interior. Heat loss from the free surface has dramatic inuence on the dynamics and must be included in theoretical models of thermocapillary instabilities. 2004 COSPAR. Published by Elsevier Ltd. All rights reserved.
Keywords: Axisymmetric liquid bridge; Thermocapillary convection; Dynamic surface deformation

1. Introduction Thermocapillary convection is a surface tension driven ow due to a temperature gradient along an interface. It is steady and axisymmetric when the driving temperature dierence in an open cylindrical cavity, or a liquid bridge, is suciently small (Sim and Zebib, 2002a,b). This steady convection undergoes transition to oscillatory ow (Sim and Zebib, 2002a,b) as the temperature dierence increases beyond a critical value. Because surface forces dominate body forces in outer space, understanding these transitions is important to material processing in microgravity. Marangoni eects in oating zone, Czochralski technique, and open boat-zone were summarized by Schwabe (1981). There are a good number of experimental studies (Preisser et al., 1983; Velten et al., 1991; Schwabe and

Corresponding author: Fax: +82 31 418 0153. E-mail address: sbcsim@naver.com (B.-C. Sim).

Frank, 1999) of surface tension driven convection in liquid bridges. The inuence of free surface shape as determined by the liquid volume was experimentally reported (Hu et al., 1994; Shevtsova et al., 1999; Sumner et al., 2001) and it was concluded that it signicantly inuenced the convection. Linear stability studies with curved, non-deformable interfaces as determined by bridge volume were performed (Chen and Hu, 1998; Kuhlmann et al., 2002). Flow-induced dynamic surface deformations in liquid bridges were investigated in the limit of small Capillary number (Kuhlman and Nienhuser, 2002) and had small inuence on the ow. There have been a number of computational studies of thermocapillary ows in liquid bridges which assumed non-deformable interfaces. In these models axisymmetric thermocapillary convection was always steady (Shevtsova and Legros, 1998). Three-dimensional numerical simulations with non-deformable and noncylindrical liquid bridges (Lappa et al., 2001, 2003) demonstrated transition to steady non-axisymmetric states with low Pr. Two and three-dimensional studies (Sim

0273-1177/$30 2004 COSPAR. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.asr.2004.09.003

1628

B.-C. Sim et al. / Advances in Space Research 34 (2004) 16271634

and Zebib, 2002a,b) of oscillatory convection with nondeformable curved surfaces in open cylinders and liquid bridges conrmed that axisymmetric states were steady. Critical conditions for transition to oscillatory azimuthal states were also established. A few numerical studies with deformable surfaces have been performed in two-dimensional rectangular cavities (Mundrane et al., 1995; Chen and Hwu, 1993; Hamed and Floryan, 2000). Mundrane and Zebib (1995) showed that small free surface deformations in the limit of small Capillary numbers did not induce transition from steady to oscillatory convection in a low Pr uid. In the present paper, we report on thermocapillary convection in liquid bridges with deformable surfaces via axisymmetric numerical simulations. The interface is determined as part of the complete solution. The Prandtl number is xed, while the inuence of heat loss, Capillary number, and Reynolds number on the dynamics is explored.

  ov r vv rP r2 v; Re ot   oT r vT r2 T ; Ma ot

3 4

where v is the non-dimensional velocity vector, and P and T are the non-dimensional pressure and temperature. Re is the Reynolds number, Pr is the Prandtl number, and Ma is the Marangoni number dened by Re c DTH ; ml m Pr ; a Ma Pr Re; 5

where m, l, and a are kinematic viscosity, dynamic viscosity, and thermal diusivity respectively. The length, temperature, velocity, pressure, and time are normalized with respect to H, DT, cDT/l, cDT/H, and lH/ cDT, respectively. The velocity components in the r and z directions of a cylindrical coordinate system are u and v, respectively. The boundary conditions become u 0; u 0; u 0; v 0; T 0 v 0; ov 0; or T 1 oT 0 or at z 0; at z 1; at r 0: 6 7 8

2. Mathematical model The physical system considered is a liquid bridge with a deformable free surface as shown in Fig. 1. Aspect ratio (Ar = R/H) of 1, where R and H are, respectively, the radius and height of the liquid bridge, and unit Prandtl number are assumed. The upper and lower disks have non-dimensional temperatures Thot = 1 and Tcold = 0. The surface tension is assumed a linear function of temperature, r r0 cT T 0 ; 1 where c = or/oT, and subscript 0 represents a reference state. Neglecting body forces, the non-dimensional governing equations are as follows: r v 0; 2

The non-dimensionalized position of the free surface is described by a function r = g(t, z). Thermal, kinematic and tangential and normal stress balance boundary conditions at the interface are   1 oT oT g0 BiT ; 9 N or oz og g0 v; ot     ov ou ou ov 2g0 1 g02 or oz or oz   oT oT N g0 ; or oz  ! 2 ou ov ov ou g02 g0 P 2 oz or oz N or  00  1 CaT g 1 ; 2 CaN g N u 10

11

12

Fig. 1. Physical System.

where N = (1 + g 0 2)1/2, g 0 = og/oz and Ca is the Capillary number dene by cDT/r0. The Biot number in Eq. (9) is given by Bi = hH/k, where h is a heat transfer coecient to the surroundings at the cold wall temperature, and k is the thermal conductivity of the liquid. The tangential stress balance equation (11) denes the driving thermocapillary forces. The initial and boundary conditions for Eq. (12) considered here are:

B.-C. Sim et al. / Advances in Space Research 34 (2004) 16271634

1629

gt 0; z Ar; gt; z 0 Ar; gt; z 1 Ar: 13

At g 0; At g 1; At n 0;

T 0; u 0; v 0; T 1; u 0; v 0; oT ov 0; u 0; 0: on on

22 23 24

The liquid volume must satisfy the mass conservation, and its total volume should be constant Z 1 1 V 2 g2 dz 1; 14 Ar 0 where the liquid volume is normalized with respect to pR2H. Ca provides a measure of the surface deection in response to thermocapillary-induced stresses. If Ca = 0 (large surface tension), the free surface is at (cylindrical liquid bridge). This parameter is in the range of O(102)O(103) for Pr 6 4.4 (Kuhlman and Nienhuser, 2002).

At the interface (n = 1),   1 g02 oT oT g0 NBiT ; g og g on u

25

3. Numerical aspects In order to solve the problem with a deformable surface, the physical domain (t, r, z) is mapped into a rectangular computational domain (t, n, g), n r=gt; z; g z: 15 16

og g0 v 26 ot    0  ou 1 g02 ov g g03 ou 0 ov 2g 1 g02 on og on og g g oT ; 27 N og     2 ou ov 2g0 ov ou g0 P 2 g0 g on on og og N  00  1 CaT g 1 ; 28 CaN N2 g as in Eq. (12), P contains a free integration constant, c(t) g(t, z) and c(t) are determined by Eqs. (28), (13) and (14). A shooting method is used to nd c(t) at each time. The free-surface shape, g(t, z), is unknown and should be obtained as a solution to the coupled governing equations. The transformed governing Eqs. (17)(20) and boundary conditions Eqs. (22)(28) are solved by a nite volume method employing a SIMPLER algorithm. Non-uniform grids are constructed with ner meshes in the regions near the free surface and the upper and lower walls where boundary layers develop. All computations are started with g = Ar, v = 0 and conduction temperature distribution. A brief summary of the computational procedure is as follows: 1. Start with initial conditions for T, v, and g. 2. The rectangular computational domain is generated numerically. 3. Solve the transformed governing equations, Eqs. (17)(20), to nd T and v with the transformed boundary conditions, Eqs. (22)(27). 4. Calculate g and c with the normal stress balance and liquid volume equations, Eqs. (28), (13) and (14). 5. Steps (2)(4) are repeated at each time step until all conditions for T, v, and g are satised with the desired accuracy. 6. Return to step (1) for the next time step. Convergence criteria for iterations within a time step or a steady state are jsn + 1 snj < 109 and jsn + 1snj/ jsn + 1j < 104, where s is any variable (u, v, T, g) at all

The governing equations, transformed into the computational domain, are: 1 onu ov ov ng0 g 0; n on on og Re ou n og ou 1 onu2 ng0 ouv ouv ot g ot on ng on og g on 1 oP u r2 u; g on gn2 ! 17

18

ov n og ov 1 onuv ng0 ov2 ov2 Re ot g ot on ng on g on og 0 oP ng op r2 v; og g on

19 ! 20

oT n og oT 1 onuT ng0 ovT ovT Pr Re ot g ot on ng on og g on r2 T ;  !  1 1 o o g00 g g02 o 02 n r 2 ng n g n on on on g2

2

2ng0 o2 o2 2: g ogon og

21

The transformed boundary conditions become

1630

B.-C. Sim et al. / Advances in Space Research 34 (2004) 16271634 Table 1 Comparison of stream function minima (wmin 102) in a two-dimensional liquid bridge (Bi = 0.3, Ar = 1 and Ca = 0) Re 100 100 10 Pr 0.1 10 100 Sumner et al. (2001) 0.4221 0.4205 Wanschura et al. (1995) 1.02 0.425 Present results 1.031 0.4216 0.42

points and n is time marching or iteration level. In addition, time histories of velocities and temperature at the mid-point of the free surface were compared to determine suitable time steps. In order to examine grid dependence, stream function minima, surface velocity and temperature distributions are computed with various grids and Ca. Fig. 2 with Bi = 0, Re = 2500, and Ca = 0.05 provides one example of our grid renement studies. Because no data with deforming liquid bridges are available, the numerical code can only be validated against known results for a at surface (cylindrical liquid bridge) in Table 1. Our results with Ca = 0 are in good agreement with those from other studies (Sumner et al., 2001; Wanschura et al., 1995).

ble or deformable surfaces. This is consistent with studies of convection in other cylindrical geometries (Sim and Zebib, 2002a,b).

4. Result and discussion a. Eect of Re and Ca on thermocapillary convection with Bi = 0 We have investigated convection up to Re = 5000 with Pr = 1, Ar = 1 and various Ca (60.1), and have found no oscillatory axisymmetric states in liquid bridges with deformable free surfaces. Assuming nondeformable at interfaces with Ar = 1, the critical Re for transition to oscillatory states is about 2500 from linear theory (Wanschura et al., 1995) and three-dimensional numerical simulations (Sim and Zebib, 2002b). Since Ca is in the range of O(102)O(103) in most experiments, we conclude that dynamic free-surface deformations do not induce transition to oscillatory convection. Thus only azimuthal waves can generate oscillations in a liquid bridge with either non-deforma-

Fig. 3. Free surface deformations with Bi = 0, Ca = 0.05 and various Re. There are two peaks of the free surface at low Re, while three peaks develop at Re > 2000.

Fig. 2. Stream function minima, and surface velocity and temperature distributions with two grids (Bi = 0, Re = 2500 and Ca = 0.05). Grid independence almost achieved.

Fig. 4. Surface pressure distributions associated with Fig. 3. Surface curvature correlates with pressure at low Re.

B.-C. Sim et al. / Advances in Space Research 34 (2004) 16271634

1631

Fig. 3 shows free surfaces with Ca = 0.05 and various Re. These are convex near the lower cold wall, and change from concave to convex with increasing Re near the hot wall. At suciently low Re, the free surface is almost asymmetric about its mid point. It has two peaks and is elevated near the cold stagnation point where surface pressure achieves its maximum value. As Re increases at xed Ca, the free surface develops three peaks. The surface deformation is O(104),

and its maximum value is 1.2 103 with Ca = 0.1 and Re = 5000. Fig. 4 shows surface pressure distributions at the parameter values of Fig. 3. Curvature of the free surface, sign and magnitude, is determined by both the surface pressure and normal viscous stresses as shown in Eqs. (12) and (28). At Re = 1, viscous stresses are small, and the surface deformation shown in Fig. 3 correlates with surface pressure variation. With increasing Re, viscous stresses increase and together with the pressure determine the shape of the interface. Evidently, surface elevation near the hot corner at the largest Re is driven by viscous stresses. The eect of Ca on surface deformations is shown in Figs. 5 and 6. Surface elevations and depressions increase with increasing Ca, while its shape is independent

Fig. 5. Free surface deformation with Bi = 0, Re = 1 and various Ca. Two peaks appear at the free surface, and the surface is asymmetric about its mid point.

Fig. 7. Surface temperature and velocity distributions with Bi = 0, Re = 2500 and various Ca. They are almost independent of Ca (60.1).

Fig. 6. Similar to Fig. 7 but with Re = 2500. Four peaks appear at the free surface, and the surface deformation becomes larger with increasing Ca.

Fig. 8. (a) Isotherms and (b) streamlines with Bi = 0, Re = 2500 and Ca = 0.05. The gures show typical two-dimensional thermocapillary convection.

1632

B.-C. Sim et al. / Advances in Space Research 34 (2004) 16271634

Fig. 9. Surface (a) temperature and (b) velocity distributions with Bi = 0, Ca = 0.05 and various Re.

of Ca at a xed Re. At Re = 1 the free surface has two peaks and a reection point near its center as shown in Fig. 5. Four peaks are observed on the free surface at a high Re as shown in Fig. 6. Fig. 7 shows stream function minima, surface temperatures and velocities at various Ca and Re = 2500. They are independent of Ca. Thus dynamic free-surface deformations with Ca 6 0.1 do not inuence the convection in the liquid bridge. Fig. 8 shows isotherms and streamlines with Ca = 0.05 and Re = 2500. Surface temperature and velocity distributions with Ca = 0.05 and various Re are shown in Fig. 9. Both gures show a typical axisymmetric thermocapillary ow in the liquid bridge. Therefore, detailed discussions are omitted here and can be found in other studies with a at surface.

b. Eect of free surface heat loss Fig. 10 shows the variations of free surface shape with Bi at xed Ca = 0.05 and Re = 2500. The corresponding surface temperature and velocity distributions are shown in Fig. 11, and surface pressure in Fig. 12. Heat loss from the free surface inuences the surface temperature gradients and hence the driving thermocapillary stress. Fig. 11 shows a dramatic increase in the temperature gradient near the hot corner with increasing Bi similar to that caused by the increase in Re shown in Fig. 9(a). Surface deformation also follows a trend similar to the change with Re shown in Fig. 3. Surface velocity, however, increases with increasing Bi unlike the dimensionless surface velocities of Fig. 9(b) which

Fig. 10. Eect of Bi on surface deformation with Ca = 0.05 and Re = 2500. Surface deformation becomes larger with increasing Bi due mainly to increased normal viscous stresses.

Fig. 11. Surface temperature and velocity distributions associated with Fig. 10.

B.-C. Sim et al. / Advances in Space Research 34 (2004) 16271634

1633

increasing Re at xed Ca. At low Re, surface deformations are determined by surface pressure variation. Two peaks are observed at low Re and Ca, and increase to four with increasing Re. Normal viscous stresses become important with increasing Re in particular near the hot corner where consecutive free surface elevation and depression occurs. Surface deformations are larger with increasing Ca and is O(104 ) with Ca 6 0.1. Thermocapillary convection in the liquid bridge interior is insensitive to variations in Ca. Heat loss from the free surface substantially inuences the dynamics. Surface temperature gradients near the hot corner, velocities, size of elevations and depressions, all increase with increasing Bi. It must be included in theoretical studies of thermocapillary instabilities.

Fig. 12. Surface pressure associated with Fig. 10. It is almost independent of heat loss.

Acknowledgements This work was supported by the Brain Korea 21 Project in 2002. We gratefully acknowledge the support of BK 21 and computer resources from the Rutgers Computational Grid composed of a Distributed Linux PC Cluster on which all computations were performed.

Table 2 Stream function minima (wmin 103) with Bi at a xed Re = 2500 Bi 0 1 2 5 Ca = 0 3.52 4.33 4.79 5.44 Ca = 0.05 3.52 4.32 4.78 5.42

References decrease with increasing Re. Furthermore, Fig. 12 conrms that surface pressure is almost independent of variations in heat loss. Thus, it is the normal viscous stresses which respond to changes in Bi and drives the surface deections of Fig. 10. Table 2 gives values of stream function minima at various Bi and Ca. At xed Bi, Ca does not inuence the convection which is consistent with Fig. 7. Convection is sensitive to variations in Bi at xed Ca. It is evident that free surface heat loss is more inuential than dynamic surface deformations for Ca 6 0.1. It must be included in theoretical models to achieve better agreement with experiments (Sim and Zebib, 2002c).
Chen, J., Hwu, F. Oscillatory thermocapillary ow in a rectangular cavity. Int. J. Heat Mass Transfer 36, 37433749, 1993. Chen, Q., Hu, W. Inuence of liquid bridge volume on instability of oating half zone convection. Int. J. Heat Mass Transfer 41, 825 837, 1998. Hamed, M., Floryan, J. Marangoni convection. Part 1. A cavity with dierentially heated sidewalls. J. Fluid Mech. 405, 79110, 2000. Hu, W., Shu, J., Zhou, R., Tang, Z. Inuence of liquid bridge volume an the onset of oscillation in oating zone convection. I. Experiments. J. Cryst. Growth 142, 379384, 1994. Kuhlman, H., Nienhuser, C., Rath, H., Yoda, S. Inuence of the volume of liquid on the onset of three-dimensional ow in thermocapillary liquid bridges. Adv. Space Res. 29, 639644, 2002. Kuhlman, H., Nienhuser, C. Dynamic free-surface deformations in thermocapillary liquid bridges. Fluid Dynamics Res. 31, 103127, 2002. Lappa, M., Savino, R., Monti, R. Three-dimensional numerical simulation of Marangoni instabilities in non-cylindrical liquid bridges in microgravity. Int. J. Heat Mass Transfer 44, 19832003, 2001. Lappa, M., Yasushiro, S., Imaishi, N. 3D numerical simulation of on ground Marangoni ow instabilities in liquid bridges of low Prandtl number uid. Int. J. Num. Meth. Heat Fluid Flow 13, 309 340, 2003. Mundrane, M., Xu, J., Zebib, A. Thermocapillary convection in a rectangular cavity with a deformable interface. Adv. Space Res. 16, 4153, 1995. Mundrane, M., Zebib, A. Low Prandtl number Marangoni convetion with a deformable interface. J. Thermophys. Heat Transfer 9, 795 797, 1995.

5. Conclusions Steady axisymmetric thermocapillary convection in a liquid bridge heated from the upper wall is computed. Two-dimensional simulations with either non-deformable or deformable surface surfaces predict steady convection even at very high Re and Ca. Thus, dynamic deformations do not induce transitions to oscillatory convection. Free surfaces are convex at the cold corner. At the hot corner they change from concave to convex with

1634

B.-C. Sim et al. / Advances in Space Research 34 (2004) 16271634 Sim, B.-C., Zebib, A. Thermocapillary convection in liquid bridges with undeformable curved surfaces. J. Thermophys. Heat Transfer 16, 553561, 2002b. Sim, B.-C., Zebib, A. Eect of free surface heat loss and rotation on transition to oscillatoiy thermocapillary convection. Phys. Fluids 14, 225231, 2002c. Sumner, L., Neitzel, G., Fontaine, J.-P., DellAversana, R. Oscillatory thermocapillary convection in liquid bridges with highly deformed free surfaces: Experiments and energy-stability analysis. Phys. Fluids 13, 107120, 2001. Velten, R., Schwabe, D., Scharmann, A. The periodic instability of thermocapillary convection in cylindrical liquid bridges. Phys. Fluids 3, 267279, 1991. Wanschura, M., Shevtsova, V., Kuhlmann, H., Rath, H. Convective instability mechanisms in thermocapillary liquid bridges. Phys. Fluids 7, 912925, 1995.

Preisser, F., Schwabe, D., Scharrnann, A. Steady and oscillatory thermocapillary convection in liquid columns with free cylindrical surface. J. Fluid Mech. 126, 545567, 1983. Schwabe, D. Marangoni eects in crystal growth melts. PCH Phys. Chem. Hydrodynam. 2, 263280, 1981. Schwabe, D., Frank, S. Experiments on the transition to chaotic thermocapillary ow in oating zones under microgravity. Adv. Space Res. 24, 13911396, 1999. Shevtsova, V., Legros, J. Oscillatory convective motion in deformed liquid bridges. Phys. Fluids 10, 16211634, 1998. Shevtsova, V., Mojahed, M., Legros, J. The loss of stability in ground based experiments in liquid bridges. Acta Astronaut. 44, 625634, 1999. Sim, B.-C., Zebib, A. Thermocapillary convection with undeformable faces in open cylinders. Int. J. Heat Mass Transfer 45, 49834994, 2002a.