Chemical Engineering Science 62 (2007) 2109 2125 www.elsevier.

com/locate/ces

Numerical simulation of periodic bubble formation at a submerged orice with constant gas ow rate
D. Gerlach, N. Alleborn , V. Buwa, F. Durst
Lehrstuhl fr Strmungsmechanik (LSTM), Universitt Erlangen-Nrnberg, Cauerstr. 4, D-91058 Erlangen, Germany Received 22 December 2005; received in revised form 6 November 2006; accepted 22 December 2006 Available online 16 January 2007

Abstract Extensive numerical simulations were carried out to study the problem of bubble formation at submerged orices under constant inow conditions. A combined volume-of-uid and level-set method was applied to simulate the formation process, the detachment and the bubble rise above the orice in axisymmetric coordinates. On the one hand, the operating conditions of the formation process such as orice ow rate, orice radius and wettability of the orice plate were investigated for the working uids of air and water at 20 C. On the other hand, the inuence of the variation of uid properties (liquid density and viscosity, surface tension) was examined individually. In this frame, the present work focused on low and medium ow rate conditions, at which the formation takes place in a periodic manner, in contrast to aperiodic or double periodic modes. The results of the computations provide information on the inuence of various conditions on the bubble shapes, the bubble volume and the transition from a single to a double periodic formation process. The numerical results were extensively validated with experimental data available in the literature. 2007 Elsevier Ltd. All rights reserved.
Keywords: Bubble; Bubble formation; Period doubling; Orice; Volume-of-uid; Numerical analysis

1. Introduction The formation process of gas bubbles at submerged orices is of importance in many industrial applications of gasliquid contacting equipment. To design such systems, an accurate knowledge of the bubble size distribution produced under certain operating conditions is necessary. Different methods are known in the literature to supply the gas to the orice exit. These methods are distinguished typically by a non-dimensional capacitance number NC as in Tsuge (1986), which considers the degree of inuence of the gas chamber connected to the orice. For NC > 1 (intermediate or constant pressure regime) the gas chamber below the orice has an effect on the ow rate supplied to the bubble and thus also on the bubble volume formed, which is encountered, for example, in bubble columns using multihole/sieve plate sparger. For NC < 1 (constant ow regime) the ow rate through the orice is continuous, thus the formation period is directly given by the ratio of the detached
Corresponding author. Tel.: +49 9131 8529486; fax: +49 9131 8529503.

E-mail address: alleborn@lstm.uni-erlangen.de (N. Alleborn). 0009-2509/$ - see front matter doi:10.1016/j.ces.2006.12.061 2007 Elsevier Ltd. All rights reserved.

bubble volume and the ow rate. This regime is obtained by providing a large pressure drop between gas chamber and orice exit by means of a capillary or a porous section (Terasaka and Tsuge, 1993; Jamialahmadi et al., 2001; Zhang and Shoji, 2001). The objective of the present work is to simulate the bubble formation process through direct numerical simulation of the ow eld and the interface evolution during the formation and the initial rise of the bubble. The present paper considers the constant ow regime of bubble formation (NC < 1) in detail and compares the numerical results with experimental data. Many investigations of the basic case of the bubble formation at a single orice have been reported, as described in review articles by Kumar and Kuloor (1970), Clift et al. (1978), Tsuge (1986) and Kulkarni and Joshi (2005). Different theoretical approaches have been applied to study the problem of bubble formation: in the limit of very low gas ow rates (quasistatic limit), the bubble contour can be predicted analytically from the balance of pressure and capillary forces, as used by Siemens (1954) and Longuet-Higgins et al. (1991). Including the viscous and inertia effects at increasing ow rates, different theoretical models of varying complexity were developed.

2110

D. Gerlach et al. / Chemical Engineering Science 62 (2007) 2109 2125

Starting from the one-stage models of Davidson and Schler (1960a,b) for dynamic formation in viscous and inviscid liquids, improved models have been proposed by means of two-stage and non-spherical models (Tsuge, 1986). In these models the inuence of the inow conditions of the gas through the orice, i.e., constant gas or constant pressure regime, were intensively investigated. Boundary-integral methods were also successfully employed to examine the formation process, as by O uz and g Prosperetti (1993), Wong et al. (1998) and Higuera (2005). In the present approach, full numerical simulations were performed using a combination of the volume-of-uid and levelset methods (Sussman and Puckett, 2000; Son and Hur, 2002), where a single set of NavierStokes equations in an axisymmetric formulation was solved in the computational domain. The parameters studied were the orice ow rate and radius, the wettability of the orice material and the inuence of surface tension and the liquid density and viscosity. The constant orice ow rate was increased stepwise until a transition from single periodic (SP) to double periodic (DP) formation was detected, i.e., until the inuence of a bubble on the following one is such that the second bubble detaches earlier compared with the SP case. This phenomenon was also observed experimentally by Kyriakides et al. (1997), Zhang and Shoji (2001) and Tufaile and Sartorelli (2002). As a consequence, the trailing bubble of each pair is smaller than the leading bubble and two distinct but constant detachment periods exist which repeat regularly. In this DP regime, the two interacting bubbles may coalesce some distance above the orice or directly during formation (Zhang and Shoji, 2001), the latter case occurring at higher ow rates is not considered here. The present work focuses on the periodic bubble formation process, but also provides information about the conditions, under which the formation regime changes from a SP to a DP regime. Air and water at 20 C served as reference working uids for the computations. Based on their data, the uid properties were varied individually to study their inuence on the formation process. Less information is typically given in the literature about the orice material used, as discussed by Ponter and Surati (1997), which can make a comparison of different studies difcult. Most of the models available are based on the assumption that the bubble base coincides with the orice rim during formation, which can be an oversimplication depending on the orice material employed. Attempts have been made here to examine the inuence of the orice material by means of a static model for the contact line movement. Based on this background, the present study investigated bubble formation at submerged orices under exactly dened conditions and demonstrated that the present numerical approach can improve the understanding of the process of bubble formation. As an extension to previous studies dealing with the inuence of uid properties on the bubble formation process in the constant ow regime (Kumar and Kuloor, 1970; Terasaka and Tsuge, 1993; Jamialahmadi et al., 2001), the present work considers also the inuence of the properties in combination with bubble interactions (pairing) and the wettability of the orice plate. The present paper is arranged as follows. In the next section, the problem studied here is described and non-dimensional

numbers are used to estimate the importance of the forces inuencing the bubble formation process. A short description of the mathematical model and the numerical method is given. The results of a numerical parameter study are provided in Section 3 considering the inuence of the operating conditions and the uid properties. The phenomenon of period doubling of the bubble formation process, i.e., the transition from SP to DP formation, is discussed in Section 3.3. Finally, an existing correlation for the bubble volume based on non-dimensional quantities was used to validate the results with experiments. 2. Analysis 2.1. Problem description The formation and detachment of gas bubbles at single submerged orices is considered. Owing to the gas ow through the orice, the bubble volume increases continuously. When the lift forces exceed the retarding forces (viscous and capillary forces) the bubble detaches and rises upwards. For the low and medium ow rates involved in the present work, the process of growth and detachment is assumed to be axisymmetric, whereby the origin of the (r, z) coordinate system is placed at the center of the orice rim as indicated in Fig. 1. The ow rate through the orice of radius Ro is constant and thus independent of the pressure variations in the bubble. This is called the constant ow rate regime in contrast to the constant pressure regime (Clift et al., 1978). The properties of the liquid phase are the density l and the viscosity l , the respective properties of the gas phase being g and g . The surface tension between the two uids is assumed to be constant. A static contact angle model is applied to characterize the movement of the triple contact line of the gasliquidsolid components as described

Fig. 1. Computational domain and boundary conditions.

D. Gerlach et al. / Chemical Engineering Science 62 (2007) 2109 2125

2111

in the next section. For the numerical solution of the problem, the continuity and NavierStokes equations for an incompressible Newtonian uid are solved with dened boundary conditions. In order to understand the inuencing forces acting on the bubble during formation, non-dimensional quantities are derived and their importance in the range of parameters considered here is discussed. At low gas ow rates the bubble volume at detachment VB can be estimated by the balance of the buoyancy and surface tension forces, i.e., 2 Ro VB (
l

g )g,

(1)

where g is the gravitational acceleration. The ratio of the two forces is characterized by the Bond number Bo =
2 l gR o

In addition to the Bond, capillary and Reynolds numbers, the wettability of the orice material is a further parameter of the present problem (Liow and Gray, 1988; Lin et al., 1994; Martn et al., 2006). This was characterized by the static contact angle s at the triple contact line at the orice plate. Following Gibbs inequality (Gibbs, 1906), the contact line stays at the orice rim as long as the contact angle is larger than s during formation, otherwise it spreads outwards and was assumed here to maintain in this case a constant contact angle equal to s . As the numerical results show, the wettability of the orice plate may have a signicant inuence on the bubble volume under the present conditions of low and medium gas ow rates. In the following section, the governing equations and the numerical method is described in the framework of the volumeof-uid method. 2.2. Numerical formulation For the numerical treatment of the two-phase problem by a volume-of-uid method, a void fraction function F is introduced as the fraction of the liquid inside a control volume (cell), with the void fraction taking the values pure gas cell, 0 1 pure liquid cell, F= (4) 0 < F < 1 two-phase cell including portions of the interface. Based on the void fraction, a single set of governing equations can be written for a two-phase ow consisting of the continuity and NavierStokes equations for an incompressible Newtonian uid, given by v = 0, (H ) (5) jv + vv = P + (H )g jt + [ (H )(v + (v)T )] + fsv , (6)

(2)

where the inuence of g was neglected, since g l here. Eq. (1) indicates that at low ow rates the bubble volume increases with the orice radius and the surface tension, but is inversely proportional to the density difference. The viscous force acting on the bubble can be approximated by the drag force due to the expansion of the bubble. Here, the velocity of the surrounding liquid can be estimated to be of the 2/3 order of Q/VB , resulting in a viscous force of the order of 1/3 l Q/VB , where Q is the orice ow rate. The ratio of viscous to capillary forces is described by the capillary number Ca =
lQ . 2 Ro

(3)

Thus the inuence of the viscous force compared with the capillary force is proportional to the viscosity and the orice ow rate. Finally, from the balance of viscous and buoyancy forces and by using Eq. (1), the relation Ca = 1/Bo1/3 can be found, which indicates when the ow regime can be considered to be dominated by viscous (Ca 1/Bo1/3 ) or hydrostatic (Ca 1/3 ) forces. 1/Bo The capillary and Bond numbers considered in the present work are 0 Ca 9.2 and 0.03 Bo 0.33. The Reynolds num ber Re= l Q/( l Ro ), usually employed in the literature (Wong et al., 1998; Tufaile and Sartorelli, 2002; Higuera, 2005) to characterize the orice ow rate, is in the range 0 Re 16, 633 in this study. Thus all three non-dimensional quantities are expected to be of importance for the present study, which focuses on the periodic formation in a waterair or similar systems. The smallest value of the term 1/Bo1/3 used in the present work is 1.45, whereby typical values of Ca are below 0.1, e.g. for the reference working uids (air and water) under the conditions Ro = 1 mm and Q = 100 ml/min, Ca = 0.023. Hence the analysis of the non-dimensional quantities shows that the cases of a periodic formation regime considered here are dominated by buoyancy and capillary forces, owing to the low viscosity and ow rates involved in most of the cases. However, Section 3.2.2 deals with the inuence of varying liquid viscosity leading to capillary numbers of up to 9.2, thus reaching a regime where viscous forces come into play.

where v is the velocity vector, t is the time, P is the pressure, g = (0, g) is the gravitational acceleration, fsv is the surface tension force and (H ) and (H ) are the density and dynamic viscosity, dened by (F ) = (F ) =
lH lH

+ +

g (1 H ), g (1 H ).

(7) (8)

The subscripts l and g denote the quantity with respect to the liquid and gas phase, respectively. The term fsv is the surface tension force per unit volume, which can be introduced in the momentum equation following the continuum surface force model of Brackbill et al. (1992) as fsv = (
)H (
), (9)

where is the (constant) surface tension, (

) is the local mean curvature of the interface,
is the level-set function and H (
) is a Heaviside function, dened as H (
) = 1 0 if
> 0, otherwise. (10)

2112

D. Gerlach et al. / Chemical Engineering Science 62 (2007) 2109 2125

The level-set function

is introduced in addition to F, because convenient formulae arise for the calculation of the local mean curvature (
) in the surface tension force term, Eq. (9).
is typically dened as a signed distance function from the interface (see, e.g., Sussman et al., 1994). Liquid regions are regions in which
> 0, whereas in gas regions
< 0. The interface is implicitly given by the
= 0 contour. The mean curvature in Eq. (9) is =
. |
| (11)

in z direction is prescribed as w(r) = 2wo 1 r R0

2

(14)

Since the uid type remains constant along particle paths, the void fraction F is passively advected by jF + v F = 0. jt (12)

Since the motion of the interface is computed as a combined volume-of-uid and level-set method, it is useful to solve a transport equation for the level-set function similar to Eq. (12) j
+ v
= 0. jt (13)

The numerical discretization of the above given equations is done as described in detail in a previous work by the authors (Gerlach et al., 2006). In this study, three different volume-of-uid methods were compared and validated by means of their application to surface tension dominant two-phase ow problems. The combined volume-of-uid and level-set method used here was one method under consideration. Hence, only a brief summary and some extensions are described in the following. The governing equations are discretized on a Cartesian and equidistant grid, where the ow variables are dened in a staggered manner (Harlow and Welch, 1965). For the bubble formation problem studied here, the code (Gerlach et al., 2006) was extended to allow computations in axisymmetric coordinates. The components to capture the interface, i.e., the front-capturing method, the interface reconstruction and the model to include surface tension forces are provided in Gerlach et al. (2006). The combined volume-of-uid and level-set method takes advantage of a mass conserving volume-of-uid discretization of the transport equation to advance the interface, and a smooth level-set function across the interface to determine the interface normal and the curvature (Eq. (11)), as suggested by Sussman and Puckett (2000) and Son and Hur (2002). For time integration of the continuity and momentum equations (Eqs. (5)(8)), a second-order Euler time stepping scheme was employed. A sketch of the computational domain is provided in Fig. 1. The origin of the axisymmetric (r,z) coordinate system is placed at the center of the orice rim. The orice radius is denoted by Ro . The boundary conditions at the centerline r = 0 and at the right boundary r = R (Fig. 1) are described as symmetry or slip boundary conditions, i.e., u=0 and jw/jr =jF /jr =j
/jr =0. Neumann conditions are dened at the outow (z = Z), i.e., the gradient normal to the boundary of all quantities is zero. At the inow, a parabolic prole of the velocity component

2 where wo = Q/ Ro is the averaged velocity at the orice. The Reynolds number of the gas ow in the pipe based on the gas properties is below the transition from the laminar to the turbulent regime. Therefore, the velocity prole of a laminar pipe ow as given above can be used as inow condition. Furthermore, at the inow it is u = 0 and F = 0. At the orice plate, no-slip and impermeability conditions (wall) are dened for the velocity eld (u = w = 0). Special treatments are necessary to capture the behavior of the triple contact line at the orice plate. Gibbs (1906) found, that if a three-phase contact line coincides with an edge on a solid surface, the contact angle , measured between the gasliquid interface and the horizontal plate through the liquid, can take values of s (180 ) + s , where s is the natural or static contact angle of the three-component system and is the angle of the solid wedge. This was conrmed theoretically by Dyson (1988) and was proved experimentally by Mason and co-workers (Oliver et al., 1977; Bayramli and Mason, 1978). For the present case ( =90 ), this means that the contact angle at the orice rim will not fall below the static contact angle s under static conditions. Instead, if < s , the bubble base (contact line) starts to spread outwards, i.e., the orice material essentially dewets. Furthermore, it was assumed in the present study that the moving contact line maintains a constant contact angle = s , which is independent of the velocity and the direction of the contact line movement. The static contact angle model described above was implemented in the computer code by xing the contact line for > s and dening a constant contact angle during contact line movement using the boundary condition j
/jz = cos s . The extensions of the computational domain R and Z (Fig. 1) are dened as R =5De /2 and Z =10De , where De is the diameter of an equivalent spherical bubble. The width was shown by Chen and Fan (2004) to be sufcient to keep the bubble formation process unaffected by the lateral boundaries. The height of the domain was chosen such that the bubble dynamics above the orice considered in this work are not inuenced by the outow boundary condition. The cell width was chosen to be 0.25 mm, which was checked by grid renement tests to be sufciently small and computationally feasible. The time step was dened such that the time step restrictions given in Gerlach et al. (2006) are fullled. A typical time step was 106 s.

2.3. Validation For the validation of the computer code, it is referred to a previous work (Gerlach et al., 2006), in which the present front-capturing method was tested against two other methods by applying them to a variety of two-phase problems. The combined volume-of-uid and level-set method was shown there to perform well with low computational costs compared to the other methods considered by Gerlach et al. (2006). The

D. Gerlach et al. / Chemical Engineering Science 62 (2007) 2109 2125

2113

46 ms

46 ms

0.03 0.025 0.02 0.015 0.01 0.005 0 -0.01 -0.005 0 0.005

0.03 0.025 0.02 0.015 0.01 0.005 0 0.01 -0.01 -0.005 0 0.005

0.03 0.025 0.02 0.015 0.01 0.005 0 0.01 -0.01 -0.005 0 0.005 0.01

0.03 0.025 0.02 0.015 0.01 0.005 0 -0.01 -0.005 0 0.005

0.03 0.025 0.02 0.015 0.01 0.005 0 0.01 -0.01 -0.005 0 0.005 0.01

46 ms

46 ms

Fig. 2. Comparison between numerically computed bubble shapes with experiments of Zhang and Shoji (2001), Q = 100 ml/min, Do = 2 mm.

bubble formation at an underwater orice under quasi-static conditions was compared with an analytical solution based on the YoungLaplace equation (Gerlach et al., 2006). The bubble shapes during formation were found to agree excellently. Tomar et al. (2005) applied the computer code to the problem of lm boiling including heat transfer and interface mass transfer. A direct comparison between bubble shapes of the numerical simulation and experiments of Zhang and Shoji (2001) for Q = 100 ml/min and Do = 2 mm has been provided in Fig. 2.1 The bubble shapes show very good qualitative agreement, the agreement of the total formation time T is excellent. 3. Results 3.1. Inuence of operating conditions In the following, the effect of the air ow rate through the orice Q, the orice radius Ro and the orice material s (wettability) on the bubble formation process is investigated. In these investigations, the uid properties were kept constant
1 Reprinted from Chemical Engineering Science, 56, Zhang, L., Shoji,

Table 1 Properties of the reference uid: airwater (aw) system at 20 C

Property
l,aw g,aw l,aw g,aw aw

Unit kg/m3 kg/m3 Pa s Pa s N/m

Value 998.12 1.188 1.002 103 1.824 105 72.8 103

at reference values of an airwater (index aw) system at 20 C as given in Table 1. Owing to limitations on numerical efforts, orice radii only of Ro = 1 and 1.5 mm were studied. Q was increased stepwise from 1 ml/min by 50 ml/min until a DP formation process was observed. s was varied between 50 (wetting liquid) and 110 (non-wetting liquid). Additionally, the case of a totally wetting liquid ( s = 0 ) was examined, for which the triple contact line of the bubble is pinned at the orice rim independent of the other parameters. 3.1.1. Orice material The inuence of the wettability ( s ) of the orice material on the bubble formation process is shown in Fig. 3, where

M., Aperiodic bubble formation from a submerged orice, 5371-5381, Copyright (2001), with permission from Elsevier.

2114

D. Gerlach et al. / Chemical Engineering Science 62 (2007) 2109 2125

t/T=0 0.015 0.015

t/T=1/3 0.015

t/T=2/3

0.01 z [m] z [m]

0.01 z [m] 0.005

0.01

0.005

0.005

0 -0.005 0 r [m] 0.005

0 -0.005 0 r [m] 0.005

0 -0.005 0 r [m] 0.005

0.015

0.015

0.015

0.01 z [m] z [m]

0.01 z [m] 0.005

0.01

0.005

0.005

0 -0.005 0 r [m] 0.005

0 -0.005 0 r [m] 0.005

0 -0.005 0 r [m] 0.005

0.015

0.015

0.015

0.01 z [m] z [m]

0.01 z [m] 0.005

0.01

0.005

0.005

0 -0.005 0 r [m] 0.005

0 -0.005 0 r [m] 0.005

0 -0.005 0 r [m] 0.005

Fig. 3. Three instants of bubble formation for the case (a) s = 0 (bubble pinned at the orice), (b) s = 70 and (c) Ro = 1 mm. The time period between the pictures is (a) T /3 = 0.014 s, (b) T /3 = 0.018 s and (c) T /3 = 0.042 s.

s = 110 with Q = 100 ml/min,

sequences of bubble shapes are provided for s = 0 (pinned bubble base or contact line), s = 70 and s = 110 . The ow rate and the orice radius are xed at Q=100 ml/min and Ro = 1 mm. The time period between the snapshots of a sequence is T /3 with T being the detachment time, i.e., T /3 = 0.014, 0.018 and 0.042 s for s = 0 , 70 and 110 , respectively. As can be seen in the snapshots, the contact angle at the bubble base has large values at the detachment of the bubble. During the formation, decreases and leads to a bubble base

spreading along the orice plate if s (Figs. 3(b) and (c)), in accordance with the static model of the contact line movement, as described in Section 2.2. In contrast, if s is smaller than the smallest value of reached during formation, the bubble base remains at the orice and the bubble volume at detachment is therefore governed by Ro , as demonstrated exemplarily for the limiting case of s = 0 in Fig. 3(a). As the analytical study of Gerlach et al. (2005) showed for the slow formation case, the minimal value of reached during formation decreases with

D. Gerlach et al. / Chemical Engineering Science 62 (2007) 2109 2125

2115

300
Q=1ml/min Q=10ml/min Q=50ml/min Q=100ml/min Q=150ml/min Q=200ml/min Q=250ml/min anal.,Fritz(1935)

DP

Table 2 Comparison of the bubble volume for different static contact angles s at an orice of Ro = 1 mm under low ow rates of Q = 1 and 10 ml/min with analytical data
s

( )

VB (mm3 ) VOF (ml/min) 1 10 39 39 45 79 139 Analytical data (Q 0) Gerlach et al. (2005) 35 35 33 69 126 Fritz (1935) 0 12 32 69 126

200 VB [mm3]

DP
0 50 70 90 110

100

35 35 39 69 127

50 s []

100

Fig. 4. Inuence of the wetting conditions of the orice plate ( s ) on the bubble volume depending on Q for Ro = 1 mm.

decrease in Ro . Consequently, orice plates with increasing s cause spreading earlier for a given Ro and can lead to formation processes which are dominated by s rather than by Ro . Furthermore, the maximal bubble base diameters DB reached during formation also increase with increase in s . This results in an increasing capillary force acting downwards, which to a rst approximation is DB , and thus leads to larger bubble volumes VB . From the bubble shapes and the detachment times given, the strong increase in bubble volume or detachment time between s =0 and 110 (by a factor of 3) becomes apparent. Similar phenomena were observed experimentally by Lin et al. (1994). In Fig. 4, the simulation results for different s are summarized in a VB s plot. Additionally, a curve representing the low ow limit data of Fritz (1935) is added. His data are based on the assumption that the bubble formation occurs with a constant contact angle = s , which is true in his case of the growth of vapor bubbles on a solid surface without an orice. Agreement can be seen for the lowest ow rate in the present study (1 ml/min) in the cases s = 90 and 110 , where the formation is dominated by s . Strong deviations from the s -dominated results of Fritz occur for smaller values of s , since the contact line is pinned over a longer time. Finally, for s = 50 and 0 , the bubble volume is independent of s , because the bubble remains pinned at the orice at any time. The two curves Q = 200 and 250 ml/min need further explanation. The branching of these curves indicate the transition from a SP to a DP formation regime. It can be seen that two bubble volumes exist under constant operating conditions. The rst bubble of each pair is typically larger than the second, because the rst one forces the second to detach earlier than in the SP regime. Only the periodic and the rst DP results are shown here, i.e., DP detachments were found for an orice with

Ro = 1 mm for s < 90 with Q = 250 ml/min, for s < 70 withQ = 200 ml/min and for all values of s considered here, when Q = 300 ml/min (not shown). Although locating the transition points accurately was not the aim of this work, the step in the volume ux Q brackets the transition. These results indicate that for Q > 300 ml/min also an aperiodic formation occurs. As reported in the literature, further transitions will occur with increasing ow rate, i.e., transitions from DP to triple periodic formation and nally to chaotic detachment periods (Zhang and Shoji, 2001). However, for all cases shown the bubble volume increases with increasing orice ow rate. Furthermore, VB increases strongly with s , when the formation is s dominated. For s > 100 (non-wetting liquid) VB is roughly three times larger than for s < 50 (wetting liquid). This suggests that the inuence of the orice material should be taken into account to predict accurately the bubble detachment volume under the present conditions. 3.1.2. Orice ow rate First, the bubble formation at low ow rates near static conditions was investigated. In Table 2, the bubble volume after detachment is given for two ow rates Q = 1 and 10 ml/min with s being a parameter. The results are compared with those of the analytical studies of Fritz (1935) and Gerlach et al. (2005). In these analytical studies, static bubble contours are computed as a result of the balance of pressure and capillary forces (YoungLaplace equation) and boundary conditions. They are applicable at very low ow rates (quasi-static conditions), where viscous forces are negligible. In Fritz (1935), the solutions of the YoungLaplace equation were calculated for gas bubbles on solid surfaces maintaining different values of s . It was assumed that the bubble volume increases by means of evaporation. Also, based on the balance of pressure and capillary forces, Gerlach et al. (2005) presented a model which accounts for the Gibbs condition at the orice rim and the inuence of a possible contact line movement away from the orice rim on the bubble formation. The treatment of the orice was the same as described in Section 2.2. The bubble volumes of Gerlach et al. (2005) given in Table 2 are the volumes above the bubble neck and for the low ow limit, i.e., for Q 0. As is known from the literature and will be seen below, the bubble

2116

D. Gerlach et al. / Chemical Engineering Science 62 (2007) 2109 2125

Table 3 Comparison of the bubble volume for different static contact angles s at an orice of Ro = 1.5 mm under low ow rates of Q = 1 and 10 ml/min with analytical data
s

volume is essentially constant at very low gas ow rates as a result of the buoyancy and capillary force balance. Changes of Q under such conditions result in a change in the bubble formation frequency, but with VB = constant. Since the solution of the YoungLaplace equation provides bubble contours only up to the last static/equilibrium stage, the dynamic necking until detachment is not included in the analytical approach. However, from Wong et al. (1998), the time for the neck to pinch off from a radius equal to the orice radius can be estimated by tn = 7.6
l,aw Ro aw

( )

VB (mm3 ) VOF (ml/min) 1 10 57 57 61 77 137 Analytical data (Q 0) Gerlach et al. (2005) 50 50 50 69 126 Fritz (1935) 0 12 32 69 126

= 0.1 ms

(15)

for the airwater case at an orice with Ro =1 mm, which under the present conditions is negligibly small. Hence the additional gas volume tn Q supplied during pinch off has no signicant inuence on VB for the cases considered in Table 2. Excellent agreement of the simulation results with the analytical data of Gerlach et al. (2005) can be seen in the case Q = 1 ml/min, which was shown in Gerlach et al. (2006) to be valid also for the bubble shapes. Further comparisons with Fritz (1935) reveal that two formation modes can be identied under the present conditions. For s = 0 and 50 the bubble stays attached to the orice, so that the results agree with static models, which use a pinned bubble base as a boundary condition (Longuet-Higgins et al., 1991; Gerlach et al., 2005). On the other hand, for s = 90 and 110 the formation processes are governed by the spreading with a constant contact angle and thus agree with Fritz (1935) and Gerlach et al. (2005). Only the intermediate value of s = 70 deviates signicantly. The reason is that the analytical approach predicts a s -dominated formation, which is opposite to the simulation, where the growth is dominated by the orice radius. Another interesting issue is to compare the results with the static force balance equation (1) of capillary and buoyancy forces. From this equation, a bubble volume of 47 mm3 is obtained for the uid properties in Table 1 and Ro =1 mm. Eq. (1) is based on the assumption of a bubble forming at the rim of the orice with radius Ro . Compared with the data in Table 2 for static conditions or a ow rate of 1 ml/min, the volume differs by about 33%. From Table 2, it can also be seen that already at Q = 10 ml/min noticeable deviations exist between simulation and the analytical results indicating the transition from static to dynamic bubble formation, where the validity of the analytical approach ends. Similar results to those described above were observed for the computations using an orice with Ro = 1.5 mm. Those results are summarized in Table 3, again for low ow rates. The inuence of Q on the detachment time T (Fig. 5(a)) and on the detached bubble volume VB (Fig. 5(b)) can be seen to be signicant. As described in the previous section, calculations were performed with increasing orice ow rates until DP formation was detected, i.e., for each curve s = constant in Fig. 5 the highest ow rate represents the last SP formation observed. As the distributions in Fig. 5(a) show for Ro = 1 mm, the transition to DP formation is delayed for increasing values of s . It can be assumed that this is a consequence of the fact that the bubble size increases with s , so that the bubble formation time

0 50 70 90 110

51 51 52 67 126

is increased and the interaction between successive bubbles is reduced. Also shown in Fig. 5 are estimations for the critical bubble volume VB,c and time period Tc , at which a transition to DP formation may be expected based on simplied assumptions (see Section 3.3). Results of O uz and Prosperetti (1993) using the same radius g (Ro = 1 mm) and also constant ow conditions are included in Fig. 5(b) for validation. O uz and Prosperetti (1993) investig gated numerically the bubble formation and detachment from a needle. Although a difference in the bubble volume between the detachment from a needle compared to an orice can be expected, one can compare their results with those for the case s = 0 , i.e., in the case where the bubble is pinned at the orice. Indeed agreement can be found with the experiments of O uz and Prosperetti (1993) as shown in Fig. 5(b). For a g comparison of the point of transition (SPDP), two publications with comparable conguration are available. The transition to a DP formation regime was found by Zhang and Shoji (2001) to occur at about 250 ml/min. Kyriakides et al. (1997) reported for a Ro = 1 mm nozzle that the transition takes place at Q = 228 ml/min for water. Both experimental results of the transition to a DP formation (SP DP) are comparable to the present data in Fig. 5. An important feature of the data in Fig. 5 is that for very low ow rates or static conditions, as has been shown above to be valid for 1 ml/min, the bubble volume is almost inde pendent of Q. As a consequence, T decreases strongly with increasing Q in this range (Fig. 5(a)). The reason is that at low ow rates capillary and buoyancy forces govern the bubble volume at detachment. An increase in the orice ow rate increases the bubble formation frequency but leaves the bubble volume constant. For higher ow rates an opposite behavior can be observed: the detachment frequency is almost independent of the ow rate, but the bubble volume increases sig nicantly with Q. For liquids of low viscosities and for high ow rates, the bubble volume can be predicted by the onestage model of Davidson and Schler (1960a) for an inviscid liquid by VB = 1.378 Q6/5 . g 3/5 (16)

D. Gerlach et al. / Chemical Engineering Science 62 (2007) 2109 2125

2117

101

5E-07 4E-07 3E-07 X

s=0 s=50 s=70 s=90 s=110 Tc

2E-07 VB [m3]

s=0 s=50 s=70 s=90 s=110 Oguz & Prosperetti (1993) Davidson & Schuler (1960a) VB,c

100 T [s]

1E-07 X

10-1

SP DP 0 50 100 150 200 250 Q [ml/min] 300 350 400 10 0 10 1

X X SP DP 10 2

Q [ml/min]

Fig. 5. Inuence of the orice ow rate Q on the bubble formation time T (a) and volume VB (b) for Ro = 1 mm. For each curve s = constant, the highest ow rate represents the last periodic formation process found using a stepping of 50 ml/min. A critical bubble volume VB,c and time Tc can be estimated (see Section 3.3).

This relation indicates that in the limit of high ow rates, T = VB /Q is only weakly dependent on Q. In the literature, these two regimes are named constant volume and constant frequency formation (Clift et al., 1978). The data presented here are in the constant volume regime and in the transition to the constant frequency case. Since the present work is restricted to SP formation, no predictions are given for higher ow rates. The present results in Fig. 5(b) indicate that they are converging towards the curve based on Eq. (16) for increasing gas ow rates. The inuence of s can be seen to decrease with increase in Q. For the case Q = 100 ml/min, Ro = 1.5 mm and s = 0 , Fig. 6(a) gives an impression of the velocity eld around a forming and rising bubble by means of a vector and streamline plot. Here the frame of reference is stationary. In Fig. 6(b) a comparison is provided between the stationary reference frame (right half) and a reference frame (left half), which moves with the constant rise velocity of the detached bubbles in positive z-direction. The parameters used in this case are those of a highly viscous liquid as will be discussed in Section 3.2.2 ( l = 0.15 Pa s, l = aw , = aw , Q = 100 ml/min, Ro = 1 mm and = 70 ). In the moving frame of reference the streamlines s resemble patterns known from creeping ow theory (Batchelor, 1985). For low viscosity liquids, like water (Fig. 6(a)), the bubbles above the orice are wobbling, in accordance with the bubble regime map of Clift et al. (1978), a representation in a moving reference frame is therefore not pursued here. 3.1.3. Orice diameter Similarly to Table 2 for Ro = 1 mm, the bubble volumes formed at an orice with Ro =1.5 mm under very low ow rates are summarized in Table 3. The static results of Fritz (1935) are independent of Ro , thus remaining constant compared with

Table 2. Additionally, data based on the method used by Gerlach et al. (2005) were calculated for Ro = 1.5 mm. The rst observation from a comparison of the results for Ro = 1 and 1.5 mm is that the bubble volume is increased with a larger orice radius. This enhancement is almost in agreement with the linear dependence of Eq. (1). Second, it is interesting that in the case s = 70 , Ro = 1.5 mm the numerically computed bubble volume agrees with the cases s = 0 and 50 , in which the bubble forms at the orice rim. Comparing these results with the data in Table 2, the conclusion agrees with Gerlach et al. (2005) that the tendency of the bubble base to spread outwards and to result in a s -dominated formation as described by Fritz (1935) increases with decrease in orice radius. During formation, smaller orices reach smaller values of , thus favoring spreading. In Fig. 7, T vs. Q and VB vs. Q are plotted for Ro = 1.5 mm, as in Fig. 5 for the smaller orice. SP bubble formations were found for higher ow rates compared with the Ro = 1 mm case. This is in agreement with the experimental data of Kyriakides et al. (1997), who observed the transition to occur for Ro = 1.5 mm at around Q = 262 ml/min compared with Q = 228 ml/min for Ro = 1 mm. As for the smaller orice Ro = 1 mm, the transition from SP to DP formation happens at lower ow rates for smaller values of s . A closer look at the s = 90 and 110 curves for both orice radii show that they agree for each s value up to Q = 150 ml/min, which demonstrate that these cases are s dominated and thus inde pendent of the orice radius. For Q > 150 ml/min the bubbles produced for s =90 and 110 at the orice with Ro =1.5 mm are larger than for Ro = 1 mm. In contrast, for small values of s (0 and 50 ), the bubble volume agrees for a given orice radius and for most of the ow rates, but increases with the orice radius.

2118

D. Gerlach et al. / Chemical Engineering Science 62 (2007) 2109 2125

b
0.03

moving ref. frame

fixed ref. frame

a
0.02

0.025

0.02

z [m]

0.015

0.015

z [m]

0.01

0.01

0.005

0.005

0 -0.01 -0.005 0 r [m] 0.005 0.01

0 -0.01 -0.005 0 r [m] 0.005 0.01

Fig. 6. (a) Streamline (left half) and vector (right half) plot for the case Q = 100 ml/min, Ro = 1.5 mm and s = 0 . (b) Streamline plot for a highly viscous liquid case ( = 0.15 Pa s, Q = 100 ml/min, Ro = 1 mm, s = 70 , coordinate system moves in z-direction with the constant rise velocity of the detached l bubbles (left half), stationary coordinate system (right half).

a
101

b
5E-07 4E-07 3E-07
s=0 s=0 s=50 s=70 s=90 s=110 VB,c

s=90 s=110 Tc

VB [m3]

100 T [s]

s=50 s=70

2E-07

1E-07

10-1

SP

SP DP

DP 0 50 100 150 200 Q [ml/min] 250 300 350 400 100 101 Q [ml/min] 102

Fig. 7. Inuence of the orice ow rate Q on the bubble formation time T (a) and volume VB (b) for Ro = 1.5 mm. For each curve s = constant, the highest ow rate represents the last periodic formation process found using a stepping of 50 ml/min. A critical bubble volume VB,c and time Tc can be estimated (see Section 3.3).

D. Gerlach et al. / Chemical Engineering Science 62 (2007) 2109 2125

2119

a
0.45 0.4 0.35 0.3 T [s] 0.25 0.2 0.15 0.1 0.05 0 0.25 SP DP 0.5 0.75 400 600

l/g,aw 800 1000 1200

b
400 600

l/g,aw 800 1000 1200

s=0 s=50 s=70 s=90 Tc,Q=100ml/min

101

anal.,Q=1ml/min Q=10ml/min Q=50ml/min Q=100ml/min Q=200ml/min Tc,Q=200ml/min

T [s]

100

10-1 SP DP 1 l* 1.25 1.5 0.25 0.5 0.75 l* 1 1.25 1.5

Fig. 8. Inuence of varying liquid density = l / l,aw (or l / g,aw ) on the bubble formation period for Ro = 1 mm depending on the static contact angle l s with Q = 100 ml/min (a) and the orice ow rate Q with s = 70 (b). A critical time period Tc for the transition from single periodic (SP) to double periodic (DP) formation can be estimated (Section 3.3).

3.2. Inuence of uid properties The inuence of the uid properties on the bubble formation process was studied based on a reference set of parameters. The reference case is dened by the uid properties in Table 1, Q = 100 ml/min, Ro = 1 mm and s = 70 . In the following, the uid properties and operating conditions are taken to be the reference values as given above, if not stated otherwise. In Sections 3.2.13.2.3 the effects of the liquid density are considl , liquid viscosity l and the surface tension ered. In each section, s and Q are varied as additional parameters. The gas density and viscosity were not varied in the present work, since their inuence is assumed to be small (see Table 1: g l, g l ). 3.2.1. Inuence of the liquid density When the liquid density is increased and all other properties remain constant, the buoyancy force increases for a given bubble volume, whereas the surface tension force can be assumed to stay constant. In this case, the bubble volume is expected to decrease for increasing liquid density in a rst estimation, which is indicated by the simple balance of Eq. (1). This tendency is conrmed in Fig. 8(a), where the inuence of the nondimensional liquid density = l / l,aw and the liquidgas l density ratio on the detachment time period is depicted. A 1/ l behavior as suggested by Eq. (1) can be observed. At given Q a smaller T implies smaller VB . At small l , the static contact angle s has a strong inuence on T, whereas for increasing l the effect almost disappears. In Fig. 8(b), the inuence of varying liquid density is shown for different values of Q. The analytical solution, as calculated by Gerlach et al. (2005) for static formation, is added for a

low ow rate of 1 ml/min, which was validated for = 1 in l Table 2. As indicated by the branching of the Q = 200 ml/min distribution, the bubble formation regime was found to change from a SP to a DP regime by increasing from 1 to 1.25 at l this ow rate. A detailed examination of the data in Fig. 8(b) reveals that the inuence of l on T decreases for increasing Q in accordance with Eq. (1), if this equation is transformed to T = 2 Ro /( l g Q). As an example of Fig. 8(b), T decreases in 1.5 for Q=1 ml/min by 93%, whereas for the range 0.25 l = 100 ml/min it is only 77%. Also for the periodic results for Q the case Q = 200 ml/min, the same tendency can be observed. In Fig. 9, the inuence of l is visualized through snapshots of the bubble shapes for the cases = 0.25, 1 and l 1.5. Again, large differences can be seen clearly by changing = 1 (b) either to 1.5 (c) or to 0.25 (a). In the rst l case VB differs slightly (12%), whereas in the second case a large increase in the bubble volume by a factor of about 4 is observable. 3.2.2. Inuence of the liquid viscosity In a review by Kumar and Kuloor (1970), contradictory results reported in the literature concerning the inuence of the liquid viscosity on bubble formation process were discussed based on their theoretical two-stage model. Their conclusions were that the inuence of the liquid viscosity is large at higher ow rates and viscosities and also large when the surface tension effect is small or small orices are used, which can also be concluded from the capillary number (Eq. (3)). Consequently, at low ow rates the inuence of viscosity is negligible, since the only forces are the capillary and the buoyancy force. Based on these observations and their theoretical models, Kumar and Kuloor claried existing contradictions in the literature.

2120

D. Gerlach et al. / Chemical Engineering Science 62 (2007) 2109 2125

0.03

0.03

0.03

0.02 z [m] z [m]

0.02 z [m] 0.01

0.02

0.01

0.01

0 -0.01 0 r [m] 0.01

0 -0.01 0 r [m] 0.01

0 -0.01 0 r [m] 0.01

Fig. 9. Inuence of varying liquid density on the bubble contour at detachment: = (a) 0.25, (b) 1 and (c) 1.5. Q = 100 ml/min, Ro = 1 mm and l

s = 70 .

0.13 0.12 0.11 0.1 0.09

b
s=0 s=50 s=70 s=90 Terasaka & Tsuge (1993) Tc,Q=100ml/min

100

0.08 0.07 0.06 0.05 0.04 0.03 10-1 100 l* 101

anal.,Q=1 ml/min Q=10 ml/min Q=50ml/min Q=100 ml/min Q=200 ml/min Eq.(34),200 ml/min Tc,Q=100 ml/min Tc,Q=200 ml/min

T [s]

T [s] 10-1

DP SP 102 10-2 10-1 100

DP

101 l*

102

Fig. 10. Inuence of varying liquid viscosity = l / l,aw on the bubble formation period for Ro = 1 mm depending on the static contact angle s with l Q = 100 ml/min (a) and the orice ow rate Q with s = 70 (b). A critical time period Tc for the transition from single periodic (SP) to double periodic (DP) formation can be estimated (Section 3.3).

The present results are in accordance with the conclusions drawn by Kumar and Kuloor (1970). In Fig. 10, the inuence of varying liquid viscosity = l / l,aw is shown for different l values of s (a) and Q (b). The formation time can clearly be seen to increase with increase in viscosity (Fig. 10(a)). However, the increase in T is very small (a factor of 1.7 in the case s = 70 ), although l was increased by a factor of 2000 from 0.1 103 to 0.2 Pa s. It is difcult to identify such small tendencies experimentally for the low and medium ow rates considered here. A large increase in the viscosity of the liquid may lead to small changes in the surface tension, which in turn

may have an inuence on the detachment frequency, as will be shown in the next section. This may be the reason for contradictory results as summarized by Kumar and Kuloor (1970). From the balance of the viscous and capillary forces (Ca), one can estimate that under the conditions of Fig. 10(a) the viscous force becomes dominant over the inuence of the capillary force for > 44, which is in agreement with the data shown. l For validation, one data point extracted from the experimental results of Terasaka and Tsuge (1993) is also shown in Fig. 10(a) for the case = 109, Ro = 1 mm and Q 100 ml/min using l a stainless-steel orice. Very good agreement for this highly

D. Gerlach et al. / Chemical Engineering Science 62 (2007) 2109 2125

2121

a
0.03

b
0.03

c
0.03

0.02 z [m] z [m]

0.02 z [m] 0.01

0.02

0.01

0.01

0 -0.01 0 r [m] 0.01

0 -0.01 0 r [m] 0.01

0 -0.01 0 r [m] 0.01

Fig. 11. Inuence of varying liquid viscosity on the bubble contour at detachment: = (a) 0.1, (b) 1 and (c) 150. Q = 100 ml/min, Ro = 1 mm and l

s = 70 .

viscous case and the numerical results of s = 0 , 50 can be seen. In Fig. 10(b) the inuence of varying viscosity is shown for different ow rates. For the low ow rate limit (Q = 1 ml/min), where viscosity has no effect on the formation, the analytical solution is plotted, which is essentially constant for varying . In addition, results for 100 and 200 ml/min are provided. l Also, the inuence of viscosity on T is small under the present conditions, as described above, an increase from = 10 to 50 l is sufcient to stabilize the formation process from a DP to a periodic regime. A comparison of T of periodic formations at Q = 100 and 200 ml/min shows that the increase of T in case of the higher ow rate is 6% larger than that for the lower, i.e., the importance of viscosity increases with increasing ow rate. Also depicted in Fig. 10 is the correlation of Davidson and Schler (1960b) for the limit of highly viscous liquids and high ow rates: VB = 6.484
lQ lg

11(c)) can be seen to be signicantly elongated in the vertical direction compared with the two other cases of lower viscosity ((a) and (b)). A reason could be that the pinch off time is directly proportional to the liquid viscosity according to Eq. (15). After the upward forces at the bubble exceed the downward forces, the bubble remains attached during the slowed necking and becomes elongated. This process resembles the second stage (detachment stage) of the two-stage models often used in theoretical approaches (Tsuge, 1986). Also, the detached bubbles show noticeable deviations for different liquid viscosities. These are in agreement with the bubble shape regime map of Clift et al. (1978). An increase in means for a freely rising l bubble a change in the shape towards a sphere. Almost no difference can be seen between Fig. 11(a) and (b) demonstrating the fact that T is almost constant for < 1 or, in other words, l for Ca 1. 3.2.3. Inuence of the surface tension Based on the reference uid of Table 1, the surface tension = / was varied in the range 0.41 aw 2.75 for different = 100 ml/min (Fig. 12(a)) and static contact angles s with Q for different ow rates with s = 70 (Fig. 12(b)). The surface tension can be seen to have a signicant inuence on the bubble detachment period under the conditions of the present study. As indicated by Eq. (1), the bubble volume increases with increasing surface tension, e.g. in the case s = 90 the detachment time T increases by a factor of about four in the range of surface tensions considered (Fig. 12(a)). In this gure it can also be seen that the effect is more pronounced for larger values of s . It is interesting that the detachment mechanism changes in the three cases s = 0 , 50 and 70 for < 0.69 from a periodic to a DP regime. This is illustrated in Fig. 13, where the case of a DP detachment for = 0.41 (a) and a periodic case at high surface tension = 2.75 (c) are shown together with the periodic reference case (b).

3/4

(17)

The highest ow rate shown in Fig. 10(b) (200 ml/min) can be seen to converge towards the correlation, but is still too small to agree exactly. The issues discussed above can further be conrmed by a comparison of bubble shapes during formation for different viscosities as provided in Fig. 11, where contours just before detachment are given in the cases = 0.1 (a), 1 (b) and 150 (c) l (Q = 100 ml/min, Ro = 1 mm and s = 70 ). It can already be seen from the bubble shapes that the difference in bubble volume or detachment period is very small, although differs by l a factor of 1500 between the snapshots in Fig. 11(a) and (c). However, signicant differences in the detached bubbles between Fig. 11(c) and the other two can be observed concerning their shapes. The attached bubble in a very viscous liquid (Fig.

2122

D. Gerlach et al. / Chemical Engineering Science 62 (2007) 2109 2125

0.3 0.25 0.2 0.15 0.1

101

100 T [s]

anal.,Q=1 ml/min Q=10 ml/min Q=50 ml/min Q=100 ml/min Q=200 ml/min Tc,Q=100 ml/min Tc,Q=200ml/min

T [s] 0.05 SP DP
s=0 s=50 s=70 s=90 Tc,Q=100ml/min

10-1

DP 3 0.5 1 1.5 *

DP 2 2.5 3

0.5

1.5 *

2.5

Fig. 12. Inuence of varying surface tension = / aw on the bubble formation period depending on the static contact angle s (a) and the orice ow rate Q (b). A critical time period Tc for the transition from single periodic (SP) to double periodic (DP) formation can be estimated (Section 3.3).

a
0.03

b
0.03

c
0.03

0.02 z [m] z [m]

0.02 z [m] 0.01

0.02

0.01

0.01

0 -0.01 0 r [m] 0.01

0 -0.01 0 r [m] 0.01

0 -0.01 0 r [m] 0.01

Fig. 13. Inuence of varying surface tension on the bubble contour at detachment: (a) = 0.41 with double periodic detachment, (b) = 1 and (c) = 2.75 with periodic detachment. Q = 100 ml/min, Ro = 1 mm and s = 70 .

The inuence of the gas ow rate is provided in Fig. 12(b). A closer look at the periodic data in this gure shows that the inuence of surface tension decreases signicantly with increase in ow rate, since the importance of the viscous effects increases compared with the surface tension effects indicated by capillary numbers of up to 0.45 in Fig. 12. At very low gas ow rates, the bubble volume is entirely given by the balance of surface tension and buoyancy force. With increasing ow rate, the inuence of surface tension decreases and nally disappears, as demonstrated, for example, by Davidson and Schler (1960a). Their theoretical and experimental work under high gas ow rates revealed no inuence of surface tension. It can

also be seen that an increasing surface tension has a stabilizing inuence on the formation process, i.e., can change the formation from DP to SP (Fig. 12), because an increase in surface tension results in an increase in bubble volume, which in turn reduces the interaction between successive bubbles. 3.3. Transition from SP to DP formation Although the present work focuses on periodic bubble formation, DP formations were also observed. Two examples of bubble contours for this case of formation are provided in Figs. 13(a) and 14. In the case of DP formation, the detachment

D. Gerlach et al. / Chemical Engineering Science 62 (2007) 2109 2125

2123

0.045 0.04 0.035

Tc [s]

0.07 0.065 0.06 0.055 0.05 0.045 0.04 0.035 0.03 0.025 0.02 0 100 200 300 Q [ml/min] 400 500 double periodic formation single periodic formation

0.03 z [m] 0.025 0.02 0.015 0.01 0.005 0 -0.01 0 r [m] 0.01

Fig. 15. Critical formation period Tc depending on the orice ow rate based on Eq. (18) and the reference uid (Table 1).

Fig. 14. Double periodic formation for the case Q=200 ml/min, Ro =1.5 mm and s = 0 .

trailing bubble disappears. A height of 1.7De was found to give good results for the present numerical data. Equating T2 = T1,rise = 1.7De uT (18)

and rise of a bubble have an inuence on the following bubble in such a way that the trailing bubble detaches earlier. As a result, bubbles rise in pairs and can coalesce at some distance above the orice or even directly during formation, depending on the ow rate, whereby the leading bubble has typically a larger volume VB,1 than the trailing bubble VB,2 . The experimental results of Zhang and Shoji (2001) and Tufaile and Sartorelli (2002) showed that period doubling causes a branch ing of the VB vs. Q dependence and that an increase in Q results in an increasing difference VB,1 VB,2 . Furthermore, for high enough ow rates, further transitions or branches can occur to triple periodic formation or nally to a chaotic one. The present results for the transitions described in the previous sections show the same features as found in experiments. Direct comparisons of the point of transition between numerical simulations and experiments (Kyriakides et al., 1997; Zhang and Shoji, 2001) made in Sections 3.1.2 and 3.1.3 showed qualitative agreement. For the results reported here, the transition from SP to DP formation can be estimated based on simplied assumptions. It can be assumed that the leading or detached bubble inuences the forming bubble, if the time of formation T2 = VB,2 /Q is of the order of the time T1,rise , which the leading bubble needs to rise a certain distance until its wake does not inuence the formation of the trailing bubble. The terminal rise velocity uT of an ellipsoidal bubble of equivalent diameter De > 1.3 mm in a pure liquid system can be approximated from Clift et al. (1978) by uT = (2.14 / l De + 0.505gD e )0.5 . The leading bubble is assumed to rise a certain height, until the wake effect on the

and using De = 2(3VB,1 /(4 ))1/3 , a critical bubble volume VB,c = VB,1 = VB,2 (or critical formation period Tc ) can be calculated, which is equal to both bubble volumes at the point of transition. For the airwater system (Table 1), the distribution is given in Fig. 15 in the Tc Q space. If the bubble formation takes longer than the critical time, i.e., T > Tc , then SP formation can be expected, since the trailing bubble has moved far enough and thus has no inuence on the following bubble. In the other case, for T < Tc , period doubling occurs. The critical quantities VB,c and Tc according to Eq. (18) were added already in Figs. 5, 7, 8, 10 and 12. Good agreement between the theoretical estimations and the numerically determined region of period doubling was found for the present data. 3.4. Representation of the results in non-dimensional quantities In order to validate the present results with literature data additionally to single data points as in the previous sections, the data are compared with a general correlation equation in nondimensional quantities. Jamialahmadi et al. (2001) proposed a non-dimensional correlation based on their comprehensive experiments using the functional relationship VB =F (Fr, Ga, Bo), where the Froude and Galileo numbers are dened in terms 5 3 of Ro by Fr = Q2 /Ro g and Ga = 2 Ro g/ 2 . With denil tions of the Bo, Fr and Ga numbers as used in the present work, the correlation of Jamialahmadi et al. (2001) takes the

2124

D. Gerlach et al. / Chemical Engineering Science 62 (2007) 2109 2125

following form: 4 VB = 3 3 Ro Fr0.36 1.119 + 1.406 0.39 + 0.469Fr0.51 . 1.08 Bo Ga (19)

Jamialahmadi et al. (2001) reported an absolute mean average error of 3.2% between their experiments and the predictions of Eq. (19) for the equivalent bubble diameter of the detached bubble, which is claimed to out-perform all other existing correlations. The corresponding value for the present data for periodic bubble formation and s = 0 is 3.3% (or 9.4% for VB ), where 54 data points were generated and compared. This comparison demonstrates the excellent accuracy of the present approach. The largest errors in the bubble diameter between correlation and numerical simulations of around 6% exist for the low ow limit cases with Q = 1 ml/min, where the analytical results of quasi-static formation already hold (see Section 3.1.2). 4. Conclusions The results of a comprehensive numerical study of bubble formation at single submerged orices and under constant inow conditions were presented. In the low ow rate limit, the numerical results were used together with analytical data to study the bubble formation under such conditions in detail. The inuence of various operating conditions on the formation was examined with increasing ow rate until a double periodic (DP) formation was detected. For the variations of the uid properties, it was observed that the bubble volume increases with decreasing liquid density as well as with increasing viscosity and surface tension. With increasing ow rate, the degree of this inuence decreases for density and surface tension, but increases for viscosity. The inuence of the material of the orice plate was incorporated into the numerical method using a static model for the contact line movement. The numerical results show a strong effect of the contact angle on the bubble formation process. Generally, an increase in the static contact angle promotes the bubble base to spread outwards, which leads in turn to larger bubble volumes. This can result in formation processes which are governed by the contact angle instead of the orice radius. A systematic experimental study of the inuence of the wetting properties of the orice material except for the quasi-static case, however, is not yet available in the literature. In addition to the simulations of periodic bubble formation, where successive bubbles form after regular time intervals, DP formations were also detected, where every second bubble detaches earlier compared with the single periodic (SP) case due to the wake effect of the previous bubble. The comparison of the point of transition between both regimes with available experimental data showed reasonable agreement. Based on simplied assumptions, it was possible for the present results to estimate a critical bubble volume below which the transition from SP to DP formation occurs. Here, it was assumed that two successive bubbles interact when the formation time is shorter than the time needed by the previous bubble to rise out of range.

The results demonstrate that the present numerical approach is a powerful tool for studying the highly dynamic process of bubble formation and detachment. Extensive validations of the computed bubble volumes with experimental results conrmed this for the case when the bubble base is attached to the orice rim. Notation Bo Ca De F Fr fsv g Ga H NC P Q r R Re Ro t T u uT v VB w wo z Z Bond number capillary number diameter of an equivalent spherical bubble void fraction function Froude number surface tension force per unit volume gravitational acceleration Galileo number Heaviside function capacitance number pressure ow rate radial component of cylindrical polar coordinates extension of the computational domain in rdirection Reynolds number orice radius time formation time period velocity component in r-direction terminal rise velocity velocity vector bubble volume velocity component in z-direction averaged orice velocity vertical component of cylindrical polar coordinates extension of the computational domain in zdirection

Greek letters angle of solid wedge local mean curvature dynamic viscosity density surface tension level set function contact angle static contact angle properties of an airwater system at 20 C critical quantity for transition from SP to DP gas property liquid property

Subscripts aw c g l

Superscripts non-dimensional quantity

D. Gerlach et al. / Chemical Engineering Science 62 (2007) 2109 2125

2125

Abbreviations DP SP double periodic formation single periodic formation

Acknowledgments The support of one of the authors (V. Buwa) by a fellowship of the Alexander von Humboldt Foundation is gratefully acknowledged. The authors would like also to thank G. Biswas and G. Tomar for many useful discussions. References
Batchelor, G.K., 1985. An Introduction to Fluid Mechanics. Cambridge University Press, Cambridge. Bayramli, E., Mason, S.G., 1978. Liquid spreading: edge effect for zero contact angle. Journal of Colloid and Interface Science 66 (1), 200202. Brackbill, J.U., Kothe, D.B., Zemach, C., 1992. A continuum method for modeling surface tension. Journal of Computational Physics 100, 335354. Chen, C., Fan, L.-S., 2004. Discrete simulations of gasliquid bubble columns and gasliquidsolid uidized beds. A.I.Ch.E. Journal 50 (2), 288301. Clift, R., Grace, J.R., Weber, M.E., 1978. Bubbles, Drops, and Particles. Academic Press, New York. Davidson, J.F., Schler, B.O.G., 1960a. Bubble formation at an orice in a inviscous liquid. Transactions of the Institute of Chemical Engineers 38, 335342. Davidson, J.F., Schler, B.O.G., 1960b. Bubble formation at an orice in a viscous liquid. Transactions of the Institute of Chemical Engineers 38, 144154. Dyson, D.C., 1988. Contact line stability at edges: comments on Gibbss inequalities. Physics of Fluids 31 (2), 229232. Fritz, W., 1935. Berechnung des Maximalvolumens von Dampfblasen. Physikalische Zeitschrift 36, 379384. Gerlach, D., Biswas, G., Durst, F., Kolobaric, V., 2005. Quasi-static bubble formation on submerged orices. International Journal of Heat and Mass Transfer 48, 425438. Gerlach, D., Tomar, G., Biswas, G., Durst, F., 2006. Comparison of volumeof-uid methods for computing surface tension dominant two-phase ows. International Journal of Heat and Mass Transfer 49, 740754. Gibbs, J.W., 1906. Scientic Papers, vol. 1, p. 326 (Dover Reprint, New York). Harlow, F.H., Welch, J.E., 1965. Numerical calculation of time-dependent viscous incompressible ow of uid with free surface. Physics of Fluids 8 (12), 21822189. Higuera, F.J., 2005. Injection and coalescence of bubbles in a very viscous liquid. Journal of Fluid Mechanics 530, 369378. Jamialahmadi, M., Zehtaban, M.R., Mller-Steinhagen, H., Sarra, A., Smith, J.M., 2001. Study of bubble formation under constant ow conditions. Chemical Engineering Research and Design 79 (A5), 523532.

Kulkarni, A.A., Joshi, J.B., 2005. Bubble formation and bubble rise velocity in gasliquid systems: a review. Industrial Engineering Chemistry Research 44, 58735931. Kumar, R., Kuloor, N.R., 1970. The formation of bubbles and drops. Advances in Chemical Engineering 8, 256368. Kyriakides, N.K., Kastrinakis, E.G., Nychas, S.G., Goulas, A., 1997. Bubbling from nozzles submerged in water: transition between bubbling regimes. Canadian Journal of Chemical Engineering 75, 684691. Lin, J.N., Banerji, S.K., Yasuda, H., 1994. Role of interfacial tension in the formation and the detachment of air bubbles. 1. A single hole on a horizontal plane immersed in water. Langmuir 10, 936942. Liow, J.-L., Gray, N.B., 1988. A model of bubble growth in wetting and non-wetting liquids. Chemical Engineering Science 43, 31293139. Longuet-Higgins, M.S., Kerman, B.R., Lunde, K., 1991. The release of air bubbles from an underwater nozzle. Journal of Fluid Mechanics 230, 365390. Martn, M., Montes, F.J., Galn, M.A., 2006. Numerical calculation of shapes and detachment times of bubbles generated from a sieve plate. Chemical Engineering Science 61, 363369. O uz, H.N., Prosperetti, A., 1993. Dynamics of bubble growth and detachment g from a needle. Journal of Fluid Mechanics 257, 111145. Oliver, J.F., Huh, C., Mason, S.G., 1977. Resistance to spreading of liquids by sharp edges. Journal of Colloid and Interface Science 59 (3), 568581. Ponter, A.B., Surati, A.I., 1997. Bubble emission from submerged oricesa critical review. Chemical Engineering Technology 20, 8589. Siemens, W., 1954. Gasblasen in Flssigkeiten, Teil I: Entstehung von Gasblasen an nach oben gerichteten kreisfrmigen Dsen. Chem.-Ing.Techn. 26, 479496. Son, G., Hur, N., 2002. A coupled level-set and volume-of-uid method for the buoyancy-driven motion of uid particles. Numerical Heat Transfer B 42 (6), 523542. Sussman, M., Puckett, E.G., 2000. A coupled level set and volume-of-uid method for computing 3D and axisymmetric incompressible two-phase ows. Journal of Computational Physics 162, 301337. Sussman, M., Smereka, P., Osher, S., 1994. A level set approach for computing solutions to incompressible two-phase ow. Journal of Computational Physics 114, 146159. Terasaka, K., Tsuge, H., 1993. Bubble formation under constant-ow conditions. Chemical Engineering Science 48 (19), 34173422. Tomar, G., Biswas, G., Sharma, A., Agrawal, A., 2005. Numerical simulation of bubble growth in lm boiling using CLSVOF method. Physics of Fluids 17 (1), 112103. Tsuge, H., 1986. Hydrodynamics of bubble formation from submerged orices. Encyclopedia of Fluid Mechanics 3, 191232. Tufaile, A., Sartorelli, J.C., 2002. Bubble and spherical air shell formation dynamics. Physical Review E 66, 056204. Wong, H., Rumschitzki, D., Maldarelli, C., 1998. Theory and experiment on the low-Reynolds-number expansion and contraction of a bubble pinned at a submerged tube tip. Journal of Fluid Mechanics 356, 93124. Zhang, L., Shoji, M., 2001. Aperiodic bubble formation from a submerged orice. Chemical Engineering Science 56, 53715381.