Chemical Engineering Journal 124 (2006) 15

Short communication

Shear rate in stirred tank and bubble column bioreactors

J.A. Sanchez Perez a , E.M. Rodrguez Porcel a , J.L. Casas Lopez a, ,
J.M. Fernandez Sevilla a , Y. Chisti b
a Department of Chemical Engineering, University of Almera, E-04120 Almera, Spain
b Institute of Technology and Engineering, Massey University, Palmerston North 5320, New Zealand
Received 23 February 2006; received in revised form 7 July 2006; accepted 12 July 2006

Abstract
A rigorous theoretical analysis is used to show that for both Newtonian and non-Newtonian power law fluids agitated in stirred vessels, the
average shear rate in the fluid is a function of the rotational speed N of the impeller, as follows:
= constant N (laminar flow)

= constant N 3/2 (turbulent flow).

Only in turbulent flow, the proportionality constant in the above equation depends on the flow index and the consistency index of the power law
fluid. The above equations derived by theoretical reasoning are in excellent agreement with the long established empirical art.
In bubble columns, the average shear rate depends on the superficial gas velocity Ug , as follows:
= constant Ug1/(n+1)
where n is the flow index of the power law fluid. The proportionality constant in the above equation for bubble columns is a function of the flow
index, consistency index and the density of the liquid.
2006 Elsevier B.V. All rights reserved.

Keywords: Stirred tanks; Bubble columns; Bioreactors; Shear rate; Mixing; Agitation

1. Introduction shear rate with the rotational speed of the impeller [47], or with
power input that depends on impeller speed [8,9]. The equations
Stirred tanks and bubble columns are widely used as mixing in Table 1 were all obtained empirically. Here we show by purely
vessels and bioreactors. Average spatial fluid velocity gradients theoretical reasoning that the average shear rate in Newtonian
at the level of the eddies or the prevailing shear rate, is an impor- and non-Newtonian media in a stirred vessel is a function of
tant variable in bioreactors but is not easy to characterize. A only the rotational speed N (laminar flow), or N3/2 (turbulent
knowledge of shear rate is essential for at least two main reasons: flow). As demonstrated further, these theoretical outcomes are
(1) shear rate influences the average apparent viscosity of non- in excellent agreement with the well-established prior art.
Newtonian fluids and hence affects power absorption, mixing In bubble columns, the sole source of agitation is the pneu-
characteristics and mass transfer phenomena [1]; (2) microor- matic power input provided by isothermal expansion of the
ganisms, bioflocs and other suspended solids are susceptible to sparged gas. For such cases, the average shear rate is shown
damage that is dependent on the prevailing shear rate and asso- to exclusively depend on the superficial gas velocity and the
ciated shear stress [2,3]. rheological properties of the fluid in ways that are consistent
The main equations for estimating average shear rate and with other independent theoretical analyses [10,11].
the maximum shear rate max in the impeller zone of stirred tanks
are summarized in Table 1 [2,3]. Most authors have correlated
2. Theory

Corresponding author. Tel.: +34 950 015832; fax: +34 950 015484. The specific energy dissipation rate in a stirred tank is well
E-mail address: [email protected] (J.L. Casas Lopez). known to depend on the shear rate and the shear stress [12],

1385-8947/$ see front matter 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.cej.2006.07.002
2 J.A. Sanchez Perez et al. / Chemical Engineering Journal 124 (2006) 15

as follows:
Nomenclature P
= (1)
a gasliquid interfacial area per unit volume of liq- V
uid in bubble column (m1 ) where P is the power input and V is the volume of the fluid in
C constant in Eq. (9) the tank. Furthermore, for Newtonian fluids, the viscosity is
di diameter of the impeller (m) the ratio of shear stress and shear rate, i.e.
dT diameter of tank (m)
H height of fluid in tank (m) = , (2)

ki impeller constant
K consistency index (Pa sn ) therefore, Eq. (1) can be written as follows:
m constant in Eq. (24) P
M torque (N m) = = = 2 , (3)
V
n flow index
N agitation speed (s1 ) or
 
Np Power number 1 P 1/2
P power input (W) = . (4)
V
q constant in Eq. (20)
Re impeller Reynolds number Eq. (4) applies to laminar, turbulent and transitional flows.
V volume of fluid (m3 ) For non-Newtonian fluids obeying the power law [1], we
Vs volume swept by the impeller (m3 ) have:
w constant in Eq. (23) = K n (5)
W width of impeller blade (m)
where K is the consistency index and n is the flow behavior
Greek symbols index of the fluid. Because the apparent viscosity a is given as
energy input per unit mass (W kg1 ) follows [4]:
average shear rate (s1 )
max maximum shear rate (s1 ) a = = K n1 , (6)

viscosity (Pa s)
a apparent viscosity (Pa s) for non-Newtonian media the equation corresponding to Eq. (3)
the number pi becomes the following:
density of fluid (kg m3 ) P
shear stress (Pa) = a 2 (7)
V
constant in Eq. (29)
and, therefore,
 
1 P 1/(n+1)
= . (8)
KV
Eq. (8) applies to both laminar and turbulent flow regimes.
For agitation under laminar flow (Re 10), the Power number
Table 1
(Np ) and the agitator Reynolds number (Re) are related [13,14]
Correlations for shear rate in stirred tanks as follows:
C
Equation Reference Np = (9)
Re
= ki N Metzner and Otto [4]
 d 0.3 d where the constant C depends on the geometry of the tank and
= 4.2N i i
Bowen [6]
 4n dTn/(n1)
W
the impeller [13,14]. Substituting the following definitions [1]
= ki 3n+1 N Calderbank and Moo-Young
[5] of the Power number and Reynolds number,
  0.42 0.55 P
= 0.367

P
V
V
Vs Np Hoffmann et al. [8] Np = (10)
 0.5 N 3 di5
= P
Va Henzler and Kauling [10]
cited by Candia and Deckwer Ndi2
[9] Re = (11)
 0.3 d
max = 9.7N ddTi i
Bowen [6]
 W0.5 in Eq. (9), we obtain the following equation:
max = 3.3N di
1.5 Robertson and Ulbrecht [7] 
 1/(1+n)
N 2n di2 P
max = N(1 + 5.3n)1/n Wichterle et al. cited in
K =C . (12)
Robertson and Ulbrecht [7] N 3 di5 Ndi2
 J.A. Sanchez Perez et al. / Chemical Engineering Journal 124 (2006) 15 3

In Eqs. (10)(12), is the density of the fluid, N the rotational where w is a constant for a Newtonian fluid because the Power
speed of the agitator and di is the diameter of the impeller. number and viscosity are constants.
The power input of course depends on the torque M on the For turbulent flow in non-Newtonian fluids, replacing the vis-
impeller, as follows: cosity term in Eq. (23) by Eq. (5) and further rearrangement can
be used to obtain the following equation for average shear rate:
P = 2MN. (13)
 1/(1+n)
For Newtonian fluids, the substitution of Eq. (13) in Eq. (12) 4Np di2
= N 3/(1+n) = mN 3/(1+n) (24)
leads to the following equation: 33 K
Cdi3 where m is a constant. Eq. (24) is quite general and applies also
M= N. (14)
2 to Newtonian media (n = 1, K = ).
Eq. (14) applies to laminar flow.
In turbulent flow, the Power number is constant [13,14], there- 2.1.1. Extension to bubble columns
fore, from Eqs. (10) and (13) we have the following:
P 2NM In bubble columns, the power input per unit volume of liquid
Np = 5
= (15) is related with the superficial gas velocity Ug , as follows:
3
N di N 3 di5
or P
= gUg (25)
Np di5 2 V
M= N . (16)
2 Substitution of the above in Eq. (8) leads to the following expres-
From Eq. (13) for a stirred tank of standard geometry [1], sion:
i.e. H = dT = 3 di , the specific power input can be expressed as  1/(n+1)
1
follows: = gUg . (26)
K
P 2NM 8
= = 3 3 NM. (17) Eq. (26) is identical to the equation that was obtained for bub-
V /4 (3di )3 3 di
ble columns by Henzler and Kauling [10] through dimensional
Substitution of Eq. (14) in Eq. (17) gives the equation: analysis. Henzler and Kauling [10] related the average shear rate
in bubble columns to the energy input per unit mass (i.e. ), as
P 8 Cdi3 4C 2
= 3 3N N= N . (18) follows:
V 3 di 2 33  1/(n+1)
Hence, in laminar flow, the specific power input is a function of = (27)
K
N2 . Eqs. (3) and (18) imply that
where
P 4C 2
= 2 = N (19)
V 33 = gUg . (28)
or
  An equation similar to Eq. (26) was derived theoretically by
4C 1/2 Kawase and Kumagai [11]. This equation included a proportion-
= N = qN (20)
33 ality constant as follows:
where q is a constant. Eq. (20) is valid for both Newtonian and  1/(n+1)
K
non-Newtonian fluids. = 2/n 1/(n+1) . (29)
For turbulent flow (Re > 104 ) in standard stirred tanks, Eqs.
(16) and (17) lead to the following equation: If in Eq. (29) the value of is estimated as proposed by Kawase
P 8 Np di5 4Np di2 3 and Moo-Young [15] and Eq. (28) is substituted for the specific
= 3 3N N2 = N . (21) energy input, the following equation is obtained:
V 3 di 2 33
 1/(n+1)
Thus, in turbulent flow, the specific power input is a function 0.63 1/(n+1) 1/(n+1) K
= (10.3n ) (Ug g) . (30)
of N3 . Furthermore, from Eqs. (3) and (21), we can deduce the
following: 1/(n+1)
Except for the multiplier (10.3n0.63 ) , Eq. (30) is identi-
P 4Np di2 3 cal to Eqs. (26) and (27).
= 2 = N , (22)
V 33
or 3. Discussion
 1/2
4Np di2
= N 3/2 = wN 3/2 (23) The theoretically derived Eqs. (20) and (23) suggest that
33
in Newtonian media, the average shear rate depends on N
4 J.A. Sanchez Perez et al. / Chemical Engineering Journal 124 (2006) 15

and N3/2 in laminar (Re < 10) and turbulent (Re > 104 ) flows, 4. Conclusions
respectively. This theoretically derived dependence is of course
in complete agreement with the well-established empirical Theoretical reasoning presented here leads to equations for
equation of Metzner and Otto [4] (Table 1), thus revealing a correlating the average shear rate in stirred vessels operated with
previously unknown theoretical foundation for that equation. various types of fluids in laminar and turbulent flow regimes.
Other empirical correlations that suggest a direct dependence The theoretically derived Eqs. (20), (23) and (24) are over-
between shear rate and N are those due to Bowen [6] and whelmingly consistent with well-known independent empirical
Calderbank and Moo-Young [5], as shown in Table 1. Clearly, observations, thus lending credence to the theoretical reason-
the empirical evidence is overwhelmingly consistent with the ing used and rationalizing the prior empirical observations. In
theory. both Newtonian and power law fluids, the average shear rate in
In turbulent flow, the empirical evidence shows that the shear laminar flow is confirmed to depend on the impeller rotational
rate depends on N1.5 (Table 1; [7]) and this is consistent with the speed N. In turbulent flow in both Newtonian and non-Newtonian
theoretically derived Eq. (23). As a further empirical evidence media, the average shear rate is shown to depend on N3/1+n . In
for Eq. (23), Kelly and Gigas [17] correlated values of the aver- bubble columns, the average shear rate is related with the superfi-
age impeller shear rate obtained through computational fluid cial aeration velocity and the rheological properties of the fluid in
dynamic modelling, with the rotational speed of the impeller, as accordance with Eq. (26). This mechanistic relationship agrees
follows: with other independent observations.

= 64.3N. (31) Acknowledgments

The regression coefficient for the above equation was 0.96 [17];
This research was supported by the Ministerio de Educacion
however, for the same data (Fig. 10a of reference [17]), we
y Ciencia (CTQ2004-04454) and Junta de Andaluca, PAI III
obtained the following correlation:
(CVI 263).
= 33.1N 1.4 . (32)
References
Eq. (32) correlated the data with a regression coefficient of
0.999, or substantially better than Eq. (31) proposed by Kelly [1] J.M. Coulson, J.F. Richardson, Chemical Engineering, vol. 1, fourth ed.,
and Gigas [17]. The exponent on N-term in Eq. (32) is quite Pergamon, Oxford, 1990, pp. 227263.
[2] Y. Chisti, Shear sensitivity, in: M.C. Flickinger, S.W. Drew (Eds.),
close to the theoretically derived value of 1.5. Eqs. (31) and Encyclopedia of Bioprocess Technology: Fermentation, Biocataly-
(32) are for an A315 axial flow hydrofoil impeller (LIGHTNIN sis, and Bioseparation, vol. 5, Wiley, New York, 1999, pp. 2379
Mixers, Rochester, NY, USA) operated in the transitional flow 2406.
regime. [3] Y. Chisti, Hydrodynamic damage to animal cells, Crit. Rev. Biotechnol. 21
As further evidence in support of the analysis presented here, (2001) 67110.
[4] A.B. Metzner, R.E. Otto, Agitation of non-Newtonian fluids, AIChE J 3
for non-Newtonian power law fluids, the empirical equation (1957) 310.
cited by Robertson and Ulbrecht [7] (Table 1) reveals that the [5] P.H. Calderbank, M.B. Moo-Young, The prediction of power consumption
shear rate is proportional to N3/1+n , or in exact agreement with in the agitation of non-Newtonian fluids, Chem. Eng. Res. Des. 37 (1959)
the theoretical Eq. (24). In non-Newtonian media, as expected 2633.
[6] R. Bowen, Unraveling the mysteries of shear-sensitive mixing systems,
[7,16], the shear rate depends on the rheological properties of the
Chem. Eng. 9 (June) (1986) 5563.
fluid in addition to depending on the rotational speed. Clearly, [7] B. Robertson, J.J. Ulbrecht, Measurement of shear rate on an agitator in a
empirical evidence from three independent sources [7,Wichterle fermentation broth, in: C.S. Ho, J.Y. Oldshue (Eds.), Biotechnology Pro-
et al. cited in 7, 17] supports the theoretical Eqs. (23) and cesses: Scale-up and Mixing, American Institute of Chemical Engineers,
(24). New York, 1987, pp. 7281.
[8] J. Hoffmann, K. Buescher, D.C. Hempel, Determination of maximum
In bubble columns, the mechanistically derived Eq. (26)
shearstress in stirred vessels, Chemie Ingenieur Technik 67 (1995)
shows that the average shear rate depends on the superficial 210214, in German.
aeration velocity, the rheological properties of the power law [9] Candia J-LF, W.-D. Deckwer, Xanthan gum, in: M.C. Flickinger, S.W. Drew
fluid and its density. This mechanistically derived relation- (Eds.), Encyclopedia of Bioprocess Technology: Fermentation, Biocataly-
ship is supported by independent analyses [10,11]. Although sis, and Bioseparation, vol. 5, Wiley, New York, 1999, pp. 26952711.
many empirical correlations for shear rate in bubble columns [10] H.J. Henzler, J. Kauling, Scale-up of mass transfer in highly viscous liquids,
in: Fifth European Conference on Mixing, Wurzburg, Germany, Cranfield,
have disregarded any relationship between shear rate and prop- BHRA, 1985, pp. 303312, paper 30.
erties of the fluid (i.e. K, n and ) (see Chisti [2] for a [11] Y. Kawase, T. Kumagai, Apparent viscosity for non-Newtonian fermenta-
review), this approach has been questioned in the literature tion media in bioreactors, Bioprocess Eng. 7 (1991) 2528.
[2,16,18]. [12] B. Metz, N.W.F. Kossen, J.C. van Suijdam, The rheology of mold suspen-
In summary, the agreement of the theory discussed here with sions, Adv. Biochem. Eng. 11 (1979) 103156.
[13] J.H. Rushton, E.W. Costich, H.J. Everett, Power characteristics of mixing
several independent empirical observations in different flow impellers. I, Chem. Eng. Prog. 46 (1950) 395404.
regimes and types of fluids, strongly supports the theoretical [14] J.H. Rushton, E.W. Costich, H.J. Everett, Power characteristics of mixing
Eqs. (20), (23), (24) and (26). impellers. II, Chem. Eng. Prog. 46 (1950) 467476.
 J.A. Sanchez Perez et al. / Chemical Engineering Journal 124 (2006) 15 5

[15] Y. Kawase, M. Moo-Young, Solid turbulent fluid heat and mass-transfera [17] W. Kelly, B. Gigas, Using CFD to predict the behavior of power law fluids
unified model based on the energy-dissipation rate concept, Chem. Eng. J. near axial-flow impellers operation gin the transitional flow regime, Chem.
36 (1987) 3140. Eng. Sci. 58 (2003) 21412152.
[16] Y. Chisti, M. Moo-Young, On the calculation of shear rate and apparent [18] Y. Chisti, Airlift Bioreactors, Elsevier, London, 1989.
viscosity in airlift and bubble column bioreactors, Biotechnol. Bioeng. 34
(1989) 13911392.