R&D NOTE g@

Drag Coefficient-Reynolds Number Transition for

Gas Bubbles Rising Steadily in Viscous Fluids
Denis Rodrigue *
Department of Chemical Engineering (CERSIM), Laval University, Quebec, QC G l K 7P4,Canada

T
he motion of a single gas bubble in a viscous liquid is of great
importance in chemical engineering due to its direct implication Recently, the author (Rodrigue, 2001) developed a
in the design of contactors where gas and liquid phases must be generalized correlation for the motion of gas bubbles
intimately mixed. For this reason, bubble velocity and/or shape have rising steadily in uncontaminatedviscous Newtonian
been the focus of several reviews and books: Levich (1962), Clift et al. fluids of infinite extent. It is the purpose of this note
(1978), Grace and Waigeri (1986), Happel and Brenner (1991) and to show that this model can be written in an explicit
form for the drag coefficient as a function of the
Rodrigue (2001).
Reynolds and Morton numbers. This new correlation
Although an enormous amount of work has been performed on the is then used to predict the position of the minimum in
hydrodynamics of gas-liquid flows, they are still not completely a graphical representationof C, versus Re. It is shown
understood. It is known from previous studies that the deformability of that the model can predict quite nicely this hydrody-
the gas particle leads to different shapes and flow regimes. This namic transition for viscous fluids.
produced a variety of correlations describing the experimental observa-
tions and resulted in complex calculations and important discrepancies Cauteur (Rodrigue, 2001) a recemment etabli une
when evaluating say the terminal velocity or the drag coefficient for a correlation gkneralisee pour le deplacernent des
specific gas-liquid system. buttes de gaz en ascension en regime permanent dans
A standard way to represent the data is by the drag des fluides newtoniens visqueux non contarninPs
d'extension infinie. On se propose dans cette note de
coefficient-Reynolds number (CD-Re) curve. Since the boundary
montrer que ce modele peut @trePcrit explicitement
conditions are not exactly determined like for a solid sphere, no general sous la forme d'un coefficient de trainee en fonction
curve could be determined so far. One observation of interest would be des nombres de Reynolds et de Morton. Cette
the position of the minimum on a CD-Re curve which is a function of the nouvelle correlation sert ensuite a la prediction de la
physical properties of the gas-liquid system, namely the liquid density, position du minimum dans une representation
viscosity and surface tension. graphique de C, par rapport b Re. On montre que le
There are two objectives for this work. First, it is shown that the new modtile peut prPdire assez bien cette transition
model of Rodrigue (2001) can be written in an explicit relation for the hydrodynamique pour les fluides visqueux.
drag coefficient as a function of Reynolds and Morton numbers. Second,
the new model is compared to existing models and literature data to Keywords: bubble, drag coefficient, hydrodynamic
transition, generalized correlation.
predict the position of the minimum of C,-Re curves for a broad range
of Morton numbers. This criterion would be of great help to predict this
hydrodynamic transition, mainly for gas-liquid contactor design.

Previous Works
In the past, several correlations have been developed from theoretical
This is usually done by including the surface tension of
and experimental studies. A standard way to represent the motion of a
the gas-liquid system via a Weber, Bond, capillary or
gas bubble in a liquid medium is via a plot of the drag coefficient as a
Morton number.
function of the Reynolds number. In the general case of particle hydrody-
In the present literature, several correlations are
namics, this kind of representation has two major disadvantages.
available to predict the drag coefficient as a function of
The first problem is related to the fact that both the velocity and the
other dimensionless numbers. The most important
characteristic length (equivalent diameter) are in both dimensionless
ones have been reviewed recently by Rodrigue (2001).
numbers, leading to an impossibility of explicit determination from one
In most cases these correlations are broken down into
another. Nevertheless, some transformations can be done to solve this
several sub-equations with limited range of applicabil-
problem (Archimedes and Gallileo numbers for example).
ity and from which the transition between these sub-
The second problem is more difficult to solve. On a C,-Re plot, a
sections are frequently not exactly defined. This leads
single curve cannot be obtained for bubbles since a third parameter
to confusion and discrepancies when performing
must be included to take into account the deformation of the particle.
some calculations, especially for the prediction of the
hydrodynamic transition like the minimum point on a
C, versus Re plot. Fortunately, this is not a lost cause.
*E-mailaddress: [email protected]

The Canadian Journal of Chemical Engineering,Volume 79, February 2001 119

Peebles and Carber (1 953) performed a series of experiments and
on the motion of air bubbles in various Newtonian fluids. From
their results, they decomposed the CD-Re curves into four
regions and proposed a correlation for each. The minimum
represents the intersection between their regions 2 and 3, given Some refinements of Equations (8) to (12) were performed
by: successively by Moore (1 965) and by El Sawi (1 974). The results
were shown to fit better the experimental data a t high Reynolds
C, = 1~ . ~ R c O . ~ * (1) number without modifying significantly the value of the
minimum.
and Later, Miyahara and Takahashi (1 985) proposed two correla-
tions (before and after the minimum) based on a series of
C, = 0.0275M Re4 (2) experimental data taken from the literature. The minimum
represents the intersection between their correlations given by:
respectively. Here the standard definitions are used for the
dimensionless numbers: the drag coefficient (C), the Reynolds
number (Re) and the Morton number (4:
and

(3)

In Equations (1 3) and (1 4) C,‘and Re’are corrected parameters

(4) for the deformation of the bubble defined by:

This gives a transition Reynolds number (Re), and a Re’= Re):(/

minimum value for the drag coefficient (C,) as:

Re, = 4.029M-0.214 (6) where (D/A) is the ratio of the sphere equivalent diameter over
the major axis and is a function of the Tadaki number (Ta = Re
, ,C = 7.249M0.’46 (7) M0.23).For the purpose of calculations, they used the relations
proposed by Tadaki and Maeda (1 961) as:
Moore (1959) did some approximate calculations for the
motion of a gas bubble rising at high Re and took into account
the deformation via a Weber number (We = pU2D/o ) and a ):( = aTab
x
deformation parameter (ratio of the transverse and longitudi-
nal axes). He obtained the drag coefficient as a function of the
Reynolds and Morton numbers: where

a = 1.1 b = -0.1 76 2<Ta<6 (18b)

where
a = 1.36 b = -0.28 6 < Ta < 16.5 (18c)

a = 0.62 b=O 16.5 < Ta (1 8d)

and
From Equations (1 5) and (16), the values of Re, and C, are:

(10) Re, =15.8(!)M-0.136

A _[15.“M(o’23ko’’36) I’
l-b (1 9)

From these equations, it was shown that a minimum value (20)

CD, = 13( Re
),
for the Weber number is 2.302 (when the computations are
made with greater accuracy than in the original paper). This
produces a transition at: The difficulty of using these equations comes from the fact
that the value of (D/A) is a function of the parameters a, b and
Re, = 3.415M-’l5 (11) Re, thus leading to an iterative process.

120 The Canadian Journal of Chemical Engineering, Volume 79, February 2001
In the next section, a new model predicting the transition is Further transformations can be made to obtain an explicit
developed from an explicit correlation of CD as a function of relation:
both Reand M.

Model
Recently, Rodrigue (2001) proposed a new generalized correla-
tion for the motion of a gas bubble in a viscous Newtonian 16
fluid. Based on a dimensional analysis and using experimental CD =Re
data encompassing a very broad range of physical properties, a
relation was given in terms of the velocity number (6and the
flow number (0 as:
1
-F with
V= 12
[1+ 0.018 FI3l4

where
8 = (0.01 8) -
3(:r3 Re8/3M"3

v= I%[" 113 Using Equation (24), the first derivative dCddRe is set to zero
in order to obtain the position of the minimum (Re,,) as:

Rem = (0.01 g)-9/831/82-7/2~-1/8 (26)

Substitution of Re by Re, back into Equation (24) gives the

minimum value of the drag coefficient (CD,,,)
as:

After some algebraic manipulations using Equations (3) to (S), ,C = 4.862M1l8 (27)
Equation (21) can be written in an implicit form for the drag
coefficient as: Results
The different correlations presented in the previous sections are
C D = - 16[ 1+0.018
Re
(i,
-C Re2 , " " M ~ ~ 9 r 4
now compared with experimental data. In order to do so, the
literature was searched to obtain data for gas bubbles rising in
Newtonian fluids in which the transition was observed. Table 1

Table 1. Experimental data taken from the literature.

# Reference Liquid T ("c) M @em) CDm

1 Jamialahmadiet al. (1 994) H,O/Glycerin ?' 1.83 x 61-1 37 0.66

2 Kubota et al. (1 967) Distilled H,O 20 2.56 x 10-1 184-581 0.68
47% H,0/53% Glycerin 20 5.51 x 10-8 93-1 98 1.04
34% H20/66%Glycerin 20 1.49 x 10-6 43-1 22 1.47
Ethyl acetate m 3.28 x 10-1 303-702 0.56
Ethanol 14 2.77 x lW9 194-227 0.33
3 Maxworthy et al. (1 996) 0.2% H,0/99.8% Glycerin 40 1.93 x 1&1 14-18 3.1 2
20% H,0/80% Glycerin 20 3.77 x 10-4 16-27 1.69
40% H,0/60% Glycerin 20 2.17~10-~ 68-1 14 0.69
60% H20/40%Glycerin 20 5.75 x lW9 119-203 0.37
70% H,0/30% Glycerin 20 1.1 1 x 10-9 21 3-279 0.29
80% H20/20%Glycerin 20 2.61 x 1V'O 270-364 0.23
90% H,0/1 0% Glycerin 20 8.03 x 10-l' 324-441 0.20
H2O 20 2.72 x 1Wl1 230-521 0.1 5
H2O 35 7.71 x lV1* 590664 0.1 3
4 OBrien and Gosline (1 935) Livestock 20 1.46 x 1W2 23-1 18 2.76
5 Peebles and Garber (1 953) 27% Ethyl acetate/73% Cottonseed oil 7' 6.54 x 10-6 30-54 1.16
41% Ethyl acetate/59%Cottonseed oil ?l 3.38 x lW7 53-1 05 0.95
6 Tadaki and Maeda (1961) lsomyl alcohol 12.5 8.07 x lW7 60-1 11 0.87
25% H20/75%Glycerin 15.7 2.53 x lo4 27-35 1.60
7 Tsuge and Hibino (1971) 50% H,0/50% Glycerin 23.2 2.32 x 10-8 5 7-1 76 1.14

'Not given in the article. I

I

The Canadian Journal of Chemical Engineering,Volume 79, February 2001 121

gives in an alphabetical order these studies which include 21 In Table 2,the data of Table 1 are analyzed to report for the
different gas-liquid systems taken from seven studies and minimum drag coefficient measured (C),, and the range of
encompassing a wide range of Morton numbers (1.93x lo-' < Reynolds numbers (Re,,,) for the two closest data points
M < 7.71 x 10-l2). available on each side of .,C, This is necessary due to the fact
In Figure 1, some experimental data taken from Table 1 and that experimental data points are incremental measurements
representative of the whole range of Morton numbers are and the chance to obtain exactly the minimum point is rather
compared with Equation (25). It can be seen that the new low. From these results, the next section will compare qualita-
correlation is in good agreement with the drag curves, tively each correlation over the whole range of Morton
especially for Morton numbers greater that about 1 O-8. numbers.

Discussion
1 000 In general, none of the four correlations presented in this paper
can predict exactly the transition (value and position) over the
whole range of Morton numbers. A closer look a t Table 2
100 reveals two different trends for fluids having a Morton number
less or greater than about 1 O-8. Nevertheless, each model could
predict more or less the values of C,,,, and Re,,,. For the sake of
comparison, good agreement between experiments and
0" 10 models is taken as: Re,,, predicted within the experimental range
observed or a deviation less than 20% for ,,C ,, (see Table 2).
In general, the model of Peebles and Carber (1 953)predicts
1 the data very well for M > 1 O-8. For M < 1 0-8, it overpredicts
both Re,,, and .,C
,,,
Moore (1959) pointed out that his calculations were valid
0.1 '
0.01
I ....l'''

0.1
' ' .'.'..I
1
' ' .,''"'
10
' ' """'
100
' ' """'' ' -J
1000 10000
only for Morton numbers less than 1 0 7 . Effectively, Equations
(1 1) and (1 2) predict Re,,, and ,,C,, with high precision in this
Re interval. But as the Morton number increases, the degree of
Figure 1. Comparison for typical drag coefficient-Reynolds number
underprediction for Re,,, and overprediction for ,,C,, increases
curves - Equation (25).The experimental data are taken from the
literature (see Table 1 for the original reference): W: M = 1.93 x 1 O-l, 0:
also.
M = 3.77 x lo4, +:
M = 8.07 x lo-', 0:M = 2.77x m: M = Equations (19)and (20) of Miyahara and Takahashi (1985)
3.28 x predicts very well Re,,, for M < while an overprediction is

Table 2. Models comparison for C, and Re,,.

Peebles and Carber Moore Miyahara and Rodrigue

Experimental values (1 953) (1 959) Takahashi (1 985) (2001)

M (Re,) 'Dm Rem 'Dm Rem CD, Rern 'Drn 'Drn

1.93x lo-' 14-1 8 3.1 2 5.7 5.70 4.70 10.7 14 3.1 1 11 3.96
1.46x 1 0-2 23-1 18 2.76 10 3.92 8.0 6.39 21 2.1 9 16 2.87
3.77 x 10-4 16-27 1.69 22 2.30 17 3.07 38 1.33 25 1 A1
2.53 x lo4 27-35 1.60 24 2.1 7 18 2.84 40 1.25 26 1.73
6.54x 10" 30-54 1.16 52 1.28 37 1.37 70 0.79 41 1.09
1.49x 43-1 22 1.47 71 1.03 50 1.02 88 0.66 50 0.91
8.07x 10-7 60-1 1 1 0.87 81 0.94 57 0.90 96 0.61 54 0.84
3.38x 10-7 53-1 05 0.95 98 0.83 67 0.76 109 0.55 60 0.75
2.17x 10-7 68-1 14 0.69 107 0.78 73 0.69 117 0.52 64 0.71
13 3 x 10-7 61-1 37 0.66 111 0.76 76 0.67 120 0.51 65 0.70
5.51 x 10-8 93-1 98 1.04 144 0.64 97 0.53 144 0.44 75 0.60
2.32 x 5 7-1 76 1.14 173 0.56 115 0.44 163 0.39 84 0.54
5.75 x 10-9 1 19-203 0.37 234 0.46 152 0.33 201 0.33 100 0.45
2.77 x 10-9 194-257 0.33 273 0.41 176 0.29 225 0.30 109 0.41
1.11 x 10-9 21 3-279 0.29 332 0.36 21 1 0.24 258 0.27 123 0.37
2.61 x 10-lo 270-364 0.23 453 0.29 282 0.1 8 318 0.23 147 0.31
8.03 x 1 0-1 324-441 0.20 583 0.25 357 0.14 373 0.21 170 0.27
3.28 x 1 0-1 303-702 0.56 706 0.22 427 0.1 2 421 0.19 190 0.24
2.72 x 230-521 0.1 5 735 0.21 443 0.11 432 0.19 195 0.23
2.56x lo-' 184-581 0.68 744 0.21 448 0.1 1 436 0.18 196 0.23
7.71 x 590-664 0.1 3 962 0.18 570 0.09 51 3 0.1 6 228 0.20

122 The Canadian Journal of Chemical Engineering, Volume 79, February 2001
CD, drag coefficient a t the minimum
Table 3. Parameters in Equation (28). D sphere equivalent diameter, (m)
f function defined by Equation (1 0)
Model K n F flow number, [F = g(D8p5/ak4)1/3]
9 gravitational acceleration, (m/s2)
Peebles and Garber (1953) 29.21 -0.068 K parameter in Equation (28)
Moore (1 959) 50.78 0 M Morton number (M = gp4/po3)
Miyahara and Takahashi (1985) n parameter in Equation (28)
Ta < 2 29.75 -0.0408 Re Reynolds number, (Re = pUD/p)
2<Ta<6 22.01 -0.0549 Re ' corrected Reynolds number, (Re' = Re/(D/A))
6 < Ta < 16.5 20.68 -0.061 4 Re, Reynolds number at the minimum
16.5 < Ta 18.45 -0.0408 Ta Tadaki number (Ta = Re M 0.23)
Rodrigue (2001) 45.26 0 U bubble velocity (m/s)
V velocity number [V = U(@p2/op)1/31
We Weber number (We = pU2D/o)

Greek Symbols
0 parameter defined by Equation (25b)
obtained otherwise. A very good agreement is also obtained for viscosity of the liquid, (Pa.s)
p
,C when M < 10-9, while a minimum of 20% underprediction p density of the liquid, (kg/m3)
is observed otherwise. (T surface tension of the liquid, (N/m)
The new model of Equations (26) and (27) predicts with high x deformation parameter
accuracy (~10%)all data for M > 1O-8. For M < 104, the model
underpredicts Re, with an overprediction of C., References
Another interesting observation came up from these calcula- Clift, R., J.R. Grace and M.E. Weber, "Bubbles, Drops and Particles",
tions. It was found that the product of the drag coefficient times Academic Press, New York, NY (1 978).
the Reynolds number a t the minimum (C,, Re,) is almost El Sawi, M.,"Distorted Gas Bubbles at large Reynolds Number", J.Fluid
independent of the Morton number. For each correlation, an Mech. 62, 163-1 83 (1 974).
equation like: Grace, J.R. and T. Waigeri, "Properties and Characteristics of Drops and
Bubbles", in "Encyclopedia of Fluid Mechanics", Vol. 3, N.P.
Cheremisinoff, Ed., Ch. 3, Gulf Publishing Company, Houston, TX
(1 986), pp. 43-57.
Happel, J.and H. Brenner, "Low Reynolds Number Hydrodynamics", Sth
can be obtained with a value of n close to zero. The constants ed., Kluwer Academic Publisher, Dordrecht, The Netherlands (1 991).
K and n are given in Table 3 for each correlation. Care must be Jamialahmadi, M., C. Branch and H. Muller-Steinhager, "Terminal
given for the correlation of Miyahara and Takahashi (1985) Bubble Rise Velocity in Liquids", Trans. IChemE 72, 1 19-1 22 (1 994).
where the parameters (7 and b are functions of both Re and M, Kubota, M., T. Akehata and T. Shirai, "The Behavior of Single Air Bubbles
thus leading to the different values of K and n as reported in in Liquids of Small Viscosity", Kagaku Kogaku 31, 1074-1 080 (1 967).
Table 3. This observation will be studied in more details in a Levich, V.G., "Physicochemical Hydrodynamics", Prentice-Hall,
future report. Englewood Cliffs, NJ (1 962).
Maxworthy, T., C. Gnann, M. Kurten and F. Durst, "Experiments on the
Rise of Air Bubbles in Clean Viscous Liquids", J. Fluid Mech. 321,
Conclusions 421-441 (1996).
It was shown that the generalized equation of Rodrigue (2001) Miyahara, T. and T. Takahashi, "Drag Coefficient of a Single Bubble
for single gas bubbles rising steadily in Newtonian fluids can be Rising Through a Quiescent Liquid", Int. Chem. Eng. 25, 146-148
written in an explicit form for the drag coefficient as a function (1985).
of the Reynolds and Morton numbers. Using data taken from Moore, D.W., "The Rise of a Gas Bubble in a Viscous Liquid", J. Fluid
the literature, the value and the position of the drag coefficient Mech. 6,113-1 30 (1 959).
Moore, D.W., "The Velocity of Rise of Distorted Gas Bubbles in a Liquid
a t the minimum (hydrodynamic transition) were compared
of Small Viscosity", J.Fluid Mech. 23, 749-766 (1965).
with different existing correlations. It can be concluded that the O'Brien, M.P. and J.E. Gosline, "Velocity of Large Bubbles in Vertical
correlation presented herein is in very good agreement for Tubes", Ind. Eng. Chem. 27, 1436-1 440 (1 935).
highly viscous fluids ( M > while the correlation of Moore Peebles, F.N. and H.J.Garber, "Studies on the Motion of Gas Bubbles in
(1959) is the best for fluids of low viscosity ( M < 1O-8). Overall, Liquids", Chem. Eng. Prog. 49, 88-97 (1953).
the new model presented in this report seems the best for the Rodrigue, D., "A Generalized Correlation for Bubble Motion", accepted
whole range of Morton numbers. AlChE J.47, 39-44 (2001).
Tadaki, T. and 5. Maeda, "On the Shape and Velocity of Single Air
Bubbles Rising in Various Liquids", Kagaku Kogaku 25, 254-264
Acknowledgements (1961).
The author wishes to acknowledge the financial support from the Tsuge, H. and 5.1. Hibino, "The Motion of Single Gas Bubbles Rising in
National Sciences and Engineering Research Council of Canada (NSERC). Various Liquids", Kagaku Kogaku 35, 65-71 (1 971).

Nomenclature
a,b parameters in Equation (1 7)
A major axis of the bubble, (m).
C, drag coefficient (C, = 4gD/3U2) Manuscript received March 28, 2000; revised manuscript received
Cb corrected drag coefficient [C,' = C, (D/A)2] August 7, 2000; accepted for publication October 13, 2000.

The Canadian Journal of Chemical Engineering,Volurne 79, February 2001 123