02638762/05/$30.00+0.

00
# 2005 Institution of Chemical Engineers
Trans IChemE, Part A, October 2005
Chemical Engineering Research and Design, 83(A10): 1186 1195

www.icheme.org/journals
doi: 10.1205/cherd.04254

HYDRODYNAMICS CHARACTERIZATION OF ROTOR-STATOR

MIXER WITH VISCOUS FLUIDS
L. DOUCET, G. ASCANIO and P. A. TANGUY

URPEI, Department of Chemical Engineering, Ecole Polytechnique, Montreal, Canada

he hydrodynamics generated by a rotor-stator mixer has been characterized experimentally in terms of power draw and flow patterns with viscous Newtonian and
shear-thinning fluids. The power consumption was correlated to the impeller speed
and represented by means of the dimensionless power number versus Reynolds number
curve. The impeller shear rate constant, determined from a macroscopic balance as well as
with the Metzner Otto method, was found to be dependent on the fluid rheological properties.
An attempt was also made to determine the pseudo-cavern (well-mixed region surrounded by
a stagnant fluid) dimension as a function of the hydrodynamic parameters. Results show that
the shape and magnitude of the pseudo-caverns in the vicinity of the rotor-stator mixer are
central to the understanding of the relationship between the power draw and the ability of
the impeller to mix the fluid into the vessel.
Keywords: rotor-stator; power consumption; shear-thinning fluids; mixing behaviour;
pseudo-cavern.

INTRODUCTION
Rotor-stator mixers are extensively used in many industrial
applications, especially when high shear rates or high
energy dissipation rates are required. For instance, they
are used for their brute force ability to homogenize
highly viscous materials, incorporate powders into liquids
or in emulsification processes. Scientific literature on
rotor-stator mixers is very scarce and information about
their fundamental behaviour, including flow patterns, ability
to mix and/or power consumption is seldom found. There
is a significant lack of knowledge to design or predict the
performance of such impellers. The primary function of a
rotor-stator mixer is to impart shear to the medium. However, energy is also converted into stirring and recirculation
of the product in batch applications, or pumping when used
in-line. Since the energy dissipation rate is high and located
in a small volume close to the mixer, the ability to recirculate
may be limited by the viscosity of the product. Therefore, in
large volume tanks, additional impeller (turbines, anchor,
planetary mixers, and so on) promoting bulk motion is
often required, which results in high capital/operating
costs and also in long and tedious development time due
to the trial and error nature of the design and the uncertainty of the scale-up procedure.
In industry, rotor-stator mixers are commonly used with
non-Newtonian fluids. When the medium exhibits strong
shear-thinning properties, flow segregations are often


Correspondence to: Professor P. A. Tanguy, URPEI, Department of
Chemical Engineering, Ecole Polytechnique, Montreal, Canada.
E-mail: tanguy@urpei.polymtl.ca

observed with open impellers (Doucet et al., 2003).

Cavern effects were first described by Wichterle and
Wein (1975) as a well-mixed region around the impeller
surrounded by a stagnant region in the case of yield
stress fluids. Moving away from the well-mixed region
where turbulence often prevails, the shear rates decrease,
the viscosity increases and the fluid motion becomes
laminar and nearly stagnant close to the vessel wall.
Although the term cavern was coined for yield stress
fluids (considering such fluids existsee Macosko, 1994
and Watson, 2004), the same phenomenon can be observed
with shear-thinning fluids. In this work, we will use the
term pseudo-cavern to describe the cavern effect with
such fluids to avoid confusion. The cavern dimension is a
key process parameter to perform mixer scale-up and
several models have been proposed to predict the caverns
diameter and height. Wichterle and Wein (1975) first
determined the cavern size by moving the agitator toward
the liquid surface or vessel wall and then observing the
appearance of a dye that had been added into the cavern.
They proposed an empirical correlation to determine
cavern size:

2
rN 2n D2
dC

(1)
m
eD
where e is an experimentally determined constant function
of the impeller type.
Solomon et al. (1981) proposed a generalized model
based on torque balance, considering an idealized spherical
cavern centered on the impeller (Rushton turbine). Their
model assumes that the flow within the cavern is tangential,

1186

HYDRODYNAMICS OF ROTOR-STATOR MIXER WITH VISCOUS FLUIDS

as in an unbaffled tank. This assumption is valid until the
cavern reaches the baffles. The proposed model reads as


3

dC
4Po

Rey
D
p3

(2)

where Rey is the ratio of the inertial forces due to the fluid
motion to the fluid yield stress, namely
Rey

rN 2 D 2
ty

(3)

Using the same assumptions, Elson et al. (1986) modified

equation (2) by considering the cavern shape as a right
circular cylinder of height hC yielding

3


dC
Po

Rey
D
p2 hC =dC 1=3

(4)

These two models are valid when D  dC  T. Elson

et al. (1986) found good agreements with cavern measurements for two non-Newtonian solutions and four impeller
sizes. They also found that in laminar flow, when Re  30,
the cavern diameter is constant and equal to the impeller
diameter.
Hirata and Aoshima (1996) investigated cavern shapes
and boundaries with laser Doppler anemometry. They
assumed that the cavern border was defined by the locus
of points where the local tangential mean velocity was
equal to 1% of the impeller tip speed. They carefully
described the cavern growth scheme versus the Reynolds
number. They confirmed the validity of the Elson et al.
model (1986) as long as dC  T, and proposed an extended
model based on viscous dissipation to deal with situations
when the caverns grow up to the vessel wall under baffled
conditions. Their model assumes that the representative
shear stress td in the mixing zone is proportional to the
average shear stress tav responsible for the dissipation
rate in a cavern, namely

td atav

(5)

1187

Although the power consumption and the knowledge of

the energy dissipation mechanisms in rotor-stator mixers
are essential for process design, they have not received
much attention in the literature. Bourne and Studer
(1992) found that the energy dissipation rate was uniform
within the annulus between the rotor and the stator. They
identified three types of shear: the planar shear which
occurs when fluid velocities change with respect to directions normal to the flow and is generated by the narrow
rotor-stator gap, the elongational shear due to the rapid
acceleration of the fluid, and the shear associated with the
smallest eddies in turbulent flow.
Myers et al. (2001) studied the power draw with a highshear homogenizer in the laminar, transitional and turbulent
regimes. They obtained power curves for the up-pumping
and down-pumping operating modes. In the turbulent
regime, the effect of off-bottom clearance and baffling
were studied. They also made an attempt to correlate a
single non-Newtonian experiment with the Newtonian
curve using a Metzner and Otto approach (1957).
Padron (2001) seems to be the first to have produced
consistent work on the power draw of rotor-stator mixers.
His thesis describes the influence of the rotor, stator and
tank geometry, and flow regime with Newtonian fluids
for rotational speed ranging from 25 s21 to 150 s21.
When establishing the power curves, he introduced several
definitions for the Reynolds numbers in order to determine
the best possible correlation. The first one is the common
Reynolds number for impellers in tanks, that is:
Re

rND2
ma

(8)

The second definition introduces the rotor-stator gap width

and the tip speed because the dissipation of power in the
gap could be related to the planar shear:
Re

rVtip dgap
ma

(9)

Another definition was also proposed, which uses for the

characteristic length the hydraulic radius of the stator
defined as the area of the slot divided by its wetted perimeter, namely

where a is a constant and

P
tav
m
VC a

1=2

Re
(6)

(10)

The definition of the power number

They finally obtained for a cylindrical cavern, assuming

td ty on the cavern boundary


3
hC dC
4a2 Po 2
Re

dC
D
p Re y

4rVtip Rh
ma

(7)

Recently, a new mathematical model has been introduced

by Amanullah et al. (1998) in order to predict torus-shaped
cavern diameters using axial flow impellers. The model best
applicable for extremely shear-thinning fluids needs the
knowledge of the fluid velocity at the cavern boundary, a
piece of information seldom known in practice.

Po

P
rN 3 D5

(11)

was also subject to several variations introducing the gap

width or the hydraulic radius as the characteristic length
instead of the classical rotor diameter.
Padron (2001) found that the transition regime occurred
at different Reynolds numbers depending on the rotor-stator
type. In the turbulent regime, he suggested that the main
source of power dissipation was not the viscous shear in
the stator gap, but rather the dissipation in the turbulent
jets discharged through the stator grid. It is also suggested

Trans IChemE, Part A, Chemical Engineering Research and Design, 2005, 83(A10): 11861195

1188

DOUCET et al.

that the Reynolds number used to correlate power data for

rotor-stator mixers should follow the conventional
definition [equation (8)].
Calabrese et al. (2002) performed accurate flow
measurements in an in-line rotor-stator using laser Doppler
anemometry and compared their results with CFD simulations. They reported that the flow pattern in the stator
slots and volutes is extremely complex and unbalanced.
Each stator slot has a different circulation and re-entrainment pattern. The data revealed that the flow field is
highly three-dimensional. The simulations indicate that
the flow in the rotor-stator gap is not a simple shear flow
but more a classical turbulent shear flow. As a result, the
use of a shear rate definition based on tip speed and gap
width may not be useful. Their work also indicates that,
in the turbulent regime, the energy dissipation rate is controlled by fluid impingement on stator slot surfaces or by
breakdown of the jets at the outlet of the stator slots.
In the literature, there is almost no data available on
rotor-stator hydrodynamics with viscous fluids and the
influence of the stator on flow patterns, cavern growths
and power consumption is not documented. The purpose
of the present work is to describe experiments performed
with one rotor-stator geometry with Newtonian and nonNewtonian fluids, in the laminar and transition regimes.
Power data and pseudo-caverns phenomena will be
discussed with respect to fluid rheological properties and
compared with established results available for the classical
open impellers.
EXPERIMENTAL SETUP AND PROCEDURE
The experimental setup is shown in Figure 1 and the
rotor-stator impeller in Figure 2. The experimental apparatus is composed of a Rayneri Turbotest mixer, retrofitted
with a shaft-mounted torque transducer (Vibrac TQ-200
with a 0.5% f.s. precision) and a speed encoder. We have
estimated the uncertainty on the torque measurements of
5% maximum. The rotational speed can be set between
0.5 s21 and 25 s21. The tank is a cylindrical unbaffled
polycarbonate vessel of 17 liters with the following

Figure 2. Rotor-stator impeller used.

dimensions: T 0.295 m, D/T 0.305, H/T  0.85. In

the experiments, h/H was kept constant to 1/3.
The rotor-stator impeller provided by the French manufacturer VMI is composed of a four-blades impeller rotating in a slotted stator. It has the following characteristics:
dStator 0.090 m
hStator 0.025 m
Number of slots on the stator 72
Width of slots 2 mm
Height of slots 22 mm
Number of circular orifices on the upper part of the
stator 8;
dRotor D 0.085 m
Height of rotor blades 0.022 m;
Rotor-stator gap dGap 1.5 mm.
Viscous Newtonian and non-Newtonian fluids were used in
the experiments.
Glucose solutions were used as Newtonian fluids. Their
viscosity ranged from 0.5 Pa s to 29 Pa s, and their density
from 1326 kg m23 to 1380 kg m23. Xanthan gum (Keltrol
RD, CP Kelco) and sodium carboxymethyl cellulose
(grade 30000, Cekolthereafter called CMC) solutions
were used as non-Newtonian fluids. For each fluid, the viscous behaviour was measured at room temperature with a
Bohlin CVO rheometer. The rheological properties of
these fluids are reported in Table 1.
The power law model
n1
ma mg_ av

(12)

Table 1. Rheological properties of the non-Newtonian test fluids.

Fluid

wt%

ty [Pa]

m [Pa sn]

n []

Xanthan

0.20
0.30
0.35
0.50
1.0
1.0
2.0

4
9
10
20

0.65
0.9
1.2
2.2
4
5
55

0.34
0.3
0.3
0.26
0.26
0.4
0.24

CMC
Figure 1. Experimental setup.

Trans IChemE, Part A, Chemical Engineering Research and Design, 2005, 83(A10): 11861195

HYDRODYNAMICS OF ROTOR-STATOR MIXER WITH VISCOUS FLUIDS

1189

was used to represent the behaviour of shear-thinning fluids

and the Herschel-Bulkey model
n
t ty mg_ av

(13)

that of the yield stress fluids.

Flow experiments were carried out with the above fluids
in a range of operability conditions corresponding to laminar
and transition regimes (speed up to 25 s21). In these experiments, flow patterns were recorded using a Sony DCR-PC
101 digital video camera and a Nikon Coolpix 900 digital
camera. Lamps were placed on the top and behind the
tank to enhance the image quality.
The power draw was also calculated by measuring
the torque produced by the impeller in rotation. Because
the shaft guiding system induces a residual torque, the
torque value must therefore be corrected by subtracting
the torque due to the friction (measured without the fluid)
from the measured torque, namely:
MC MM  MR

(14)

The power dissipated in the tank was calculated by:

P 2pNMC

(15)

For the Newtonian fluids, runs were performed with and

without the stator cage in order to check the stator effect.
Non-Newtonian trials were carried out with the rotorstator configuration only. In each run, the torque and the
pseudo-cavern size were measured.
Several methods have been proposed to investigate
cavern phenomena experimentally including X-ray (Elson
et al., 1986) and laser Doppler anemometry (Hirata et al.,
1994). In this study we used a flow visualization technique
based on a fast acid/base reaction described by Brennan
and Lehrer (1976) and more recently by Arratia et al.
(2004). The principle is the following: a tracer solution is
prepared with 0.08% purple Bromocresol as aqueous indicator. 100 mL of this solution is added to the tank and
thoroughly mixed. Subsequently, addition of small quantity
of 3M NaOH turns the bulk solution to a uniform purple
colour. Conversely, addition of 3M HCl turns the solution
to yellow. Conveniently, purple caverns are more visible
in yellow bulk. Preliminary trials were performed in
order to check that the method did not alter the rheological
properties of the fluids studied.
Each experiment was started with a fresh solution of the
fluid tested loaded with purple Bromocresol, the rotational
speed was set at the desired value and the solution containing NaOH was injected directly inside the stator cage with
a pipette through one of the eight suction holes on the top
of the stator cage (see Figure 2). It was observed that
the pseudo-cavern rapidly grows until it reaches its final
size. The experiments were recorded and allowed to
follow the development of this flow segregation when
increasing the speed.

Figure 3. Pseudo-cavern phenomenon.

of the blades. A similar phenomenon has been observed

in this work with the rotor-stator investigated irrespective
of the fluid rheology.
A rotor-stator is designed as an axial pump and a dispersing device. In the model used, the liquid is pumped in the
stator cage from the top circular orifices and discharges in
the radial direction through the stator slots. In practice, we
observed that depending on the rotor speed, only part of the
flux may flow through the stator while the remaining exits
beneath the stator. This creates a pseudo-cavern that has
a right circular cylinder shape. Figure 3 illustrates the
phenomenon in the case of one of the glucose solutions.
In our experiments, the typical pseudo-cavern height to
diameter ratio was in the range 0.4 , hc =dc , 1:0.
We show in Figure 4 the evolution of the pseudo-cavern
shape with the Reynolds number [as defined in equation
(8)] for the Newtonian solutions.
At low Reynolds numbers, the shape is that of a right
circular cylinder. As the Reynolds number increases,
the lower part of the well-mixed region grows faster than
the upper part reaching the walls near the bottom of the
vessel. At this stage, the pseudo-cavern has a kind of
pine tree shape. When the Reynolds number is further
increased, the energy imparted by the impeller to the
fluid is sufficiently high to push it through the stator
slots. The flow reaches the walls and the pseudo-cavern
becomes a cylinder with dc T, which will eventually
occupy the whole vessel volume at higher rotor speed.
In our experiments, the tank was considered to be fully
agitated at Reynolds numbers above 400. Although the

RESULTS AND DISCUSSION

Pseudo-Cavern Phenomena
As seen previously, flow caverns are formed with open
impellers rotating in non-Newtonian shear-thinning fluids
and correspond to the well-mixed region in the vicinity

Figure 4. Evolution of cavern shapes with the Newtonian fluids.

Trans IChemE, Part A, Chemical Engineering Research and Design, 2005, 83(A10): 11861195

1190

DOUCET et al.

rotor-stator vendors practice is to scale-up based on rotor

tip speed, which is the usual criterion in emulsification
and solid dispersion, the above results show that the flow
patterns are identical at the same Reynolds number,
meaning that an alternate mixer scale-up parameter is to
maintain the Reynolds number constant to ensure an
adequate flow circulation in the vessel.
We observed a different pseudo-cavern phenomenon
with non-Newtonian fluids. At the same Reynolds number,
the pseudo-cavern size and shape is influenced by the
apparent viscosity near the impeller. Figure 5 presents the
caverns obtained with xanthan gum 0.5 wt% and CMC
2 wt% at various rotation speeds. Both fluids have a quite
similar shear-thinning index but the CMC solution exhibits
a much higher consistency index (see Table 1). Two
different Reynolds numbers have been computed for each
of the cases presented, one denoted Ren that corresponds
to the Reynolds number of a power law fluid assuming
that g_ av N (Rieger and Novak, 1973), namely
Ren

rN 2n D2
m

(16)

the other one being Rey [see equation (3)].

It can be seen in Figure 5 that the pseudo-cavern growth
phenomenon is more complicated than for the Newtonian
fluids. Three steps can be identified. At low Reynolds
number, the radial flow is unable to go through the slots
and it develops as a swirling cavity below the stator
head. When the Reynolds number is increased, the radial
flow starts discharging through the slots. The pseudocavern grows with a roughly constant hc/dc ratio, until
reaching the vessel walls. However, a different behaviour
can be noticed with the CMC solution (a fluid with a
much higher apparent viscosity) as the swirling cavity
regime persists until the well-mixed region reaches the
walls. It seems that the combined effect of yield stress
and high consistency index impedes the radial discharge
even at rotational speed as high as 5 s21.
The results of Figure 5 show that the flow patterns do not
scale-up with Ren. Considering the presence of a yield
stress in the definition of the Reynolds number as in Rey
seems, however, more relevant. Indeed, for the four fluids
exhibiting a yield stress (see Table 1), the swirling cavity
regime is maintained up to Rey 10, and when Rey . 10
the radial flow discharges through the slots. When the
pseudo-cavern has reached the walls and starts to expand
vertically in the vessel, its size seems to be similar for Rey
around 100 and the tank is fully agitated for Rey . 800.
These new results seem to be promising and will need to
be further refined in a subsequent work.
The above observations are of high practical use as they
highlight the possible creation of cavities around the mixing
head at low Re, which is not always possible to notice when
the fluid is opaque or the vessel is non-transparent.
As mentioned in the literature review, models have been
developed to predict the volumes of caverns generated by
open impellers with yield stress fluids. It might be interesting to know whether these models are applicable to the case
of rotor-stators. We show in Figure 6 a comparison of the
predicted cavern volumes with the experimental measurements for two of the xanthan and CMC solutions versus
Rey. The comparison was made with two models: the

Figure 5. Evolution of cavern shapes with non-Newtonian fluids.

Trans IChemE, Part A, Chemical Engineering Research and Design, 2005, 83(A10): 11861195

HYDRODYNAMICS OF ROTOR-STATOR MIXER WITH VISCOUS FLUIDS

1191

Figure 7. Newtonian power curve for rotor-stator impeller.

Figure 6. Comparison of experimental cavern volume with the published

correlations: (a) xanthan at 1.0 wt%, and (b) CMC at 2.0 wt%.

torque balance based approach proposed by Elson et al.

(1986) in which we introduced the dimensionless ratio
hC =dC determined experimentally, and the Hirata and
Aoshima (1996) model. For the latter, several values of
the adjustable constant a were tested as, according to the
own authors remarks, this value depends on the cavern
regime and the baffled conditions and possibly the vessel
geometry. The best fit (a 0.18 for xanthan and a 0.13
for CMC) is presented here.
One can readily see that none of the models is able to
represent accurately the cavern growth, although the results
are acceptable for the xanthan solution. The predicted
volumes depart drastically from the experimental measurements as soon as the pseudo-cavern has grown up to the
walls (Rey above 100), especially with the model proposed
by Elson et al. (1986). For both fluids the model that best
fits the experimental results is the one from Hirata and
Aoshima (1996). On the other hand and because of the
adjustable constant a, this model is unable to predict
a priori the cavern volume. Similar results were obtained
with the other fluids. Both models were developed for
Rushton turbine and expectedly are not suitable with
rotor-stator impellers.

power constant Kp Po  Re is 314 and the power number

in the turbulent regime is constant and equal to 3. These
values are in the same range as those reported by Padron
(2001) in the case of a rotor-stator provided with a slotted
stator head.
To determine the influence of the stator on the power
draw, power consumption has been measured for the
rotor alone after removing the stator (Figure 8). It appears
that the stator head has a significant influence on the power
draw in the laminar regime as the KP value turns to be
almost three times lower (KP 92.7) than for the complete
rotor-stator configuration. On the other hand, the turbulent
power number value remains the same for both configuration at Po 3. This result is in agreement with the conclusions given by Calabrese et al. (2002) who postulated
that the power draw in the laminar regime is mainly due
to the stator head, and in the turbulent regime is due to
the jet discharged by the rotor. It can be also observed
that the transition regime starts at Re 10 for the stator
alone and at Re 100 for the rotor-stator.
It is possible to rework the above results in order to
establish some kind of average rate of deformation in the
vessel. Indeed, using a macroscopic approach for the
power consumption, it is straightforward to show that
P ma g_ 2av V

where ma is the medium viscosity. Rearranging equation

(17) and assuming that the power is dissipated in the

Power Draw
Newtonian fluids
Figure 7 shows the classical power curve obtained with
Newtonian fluids. Re and Po have been calculated using
equations (8) and (11), respectively. It can be seen that
the curve slope in the laminar region is close to the theoretical value of 21. For the rotor-stator investigated, the

(17)

Figure 8. Newtonian power curve for rotor impeller.

Trans IChemE, Part A, Chemical Engineering Research and Design, 2005, 83(A10): 11861195

1192

DOUCET et al.

cavern region, we can use V VC and then obtain the

average shear rate in the cavern:
s
P=VC
g_ av
(18)
ma
This average shear rate is therefore directly related to the
energy dissipation rate in the vessel.
Now assuming that the average shear rate around the
impeller can be linearly correlated to the impeller speed
as postulated by Oldshue (1983), one can have:

g_ av CS N

(19)

where CS is a mixer constant.

Figure 9 shows the variation of the average rate of deformation with the rotor speed for the Newtonian solutions in
the full range of Reynolds number investigated. It can be
observed that the average shear rate is a linear function
of the mixer speed. Two trend lines are obtained, depending
on whether the pseudo-cavern border has reached the vessel
wall.
In Figure 10, the range of values of the shear constant,
CS, is given versus the Reynolds number. The two
pseudo-cavern regimes correspond to two different values
of CS, as evidenced by the two data clouds shown on
the figure. On the other hand, the CS values obtained for
the single rotor turbine seem to give an almost constant
value around 4, lower than the value obtained by Oldshue
(1983) for Rushton turbine. With the complete rotorstator configuration it then becomes clear that if one
wants to simply correlate the average shear rate with the
impeller speed, it is needed to know a priori the cavern
regime under which it falls. For instance, with this rotorstator when the pseudo-cavern has grown up to the walls,
the CS value is approximately 7.5.
It is also interesting to note that for a same Reynolds
number range (between 10 and 60) both cavern regimes
may overlap. This is apparently depending on the medium
viscosity.

Figure 10. Shear constant CS as a function of the Reynolds number

(Newtonian solutions).

Non-Newtonian fluids
Standard power curves can also be plotted for nonNewtonian fluids. The most common approach is that of

Rieger and Novak (1973), which considers that the shear

rate is equal to the rotational speed of the rotor, i.e.,
g_ av N. The Reynolds number is therefore calculated
using equation (16).
To get a convenient insight for scale-up problems,
non-Newtonian power curves can be shifted on the
Newtonian power curve. For this purpose, the knowledge
of the apparent viscosity is needed, which requires the
determination of the average shear rate in the vessel.
This can be readily obtained with the classical Metzner
and Otto (1957) concept which assumes the shear rate
value g_ av KS  N directly proportional to the rotating
speed. We show in Figure 11 the KS values obtained for
each of the fluids in the laminar regime, which are plotted
versus the rotational speed.
One can readily see that the KS values depend on the
fluid tested. For the xanthan solutions KS is clearly independent on the rotational speed, however the independency is
less obvious for CMC solutions. Moreover the range of
KS values is really wide, from 0.6 to 70. Myers et al.
(2001) reported a KS value of 71 for a Greerco homogenizer, but this value was obtained for a single set of data
with CMC. Moreover, the Metzner and Otto approach
was performed in the transition regime; therefore, we
would not consider this reported value as representative.
KS could be considered to be dependent of the shearthinning index for strongly shear thinning-fluids like in

Figure 9. Average shear rate as a function of the impeller speed (Newtonian

solutions).

Figure 11. KS from Metzner and Otto versus rotational speed for nonNewtonian fluids.

Trans IChemE, Part A, Chemical Engineering Research and Design, 2005, 83(A10): 11861195

HYDRODYNAMICS OF ROTOR-STATOR MIXER WITH VISCOUS FLUIDS

the case of helical ribbon impellers (Brito-de la Fuente,
1992). In our particular case n ranges from 0.26 to 0.34
and from 0.24 to 0.40 for xanthan and CMC solutions,
respectively. Multiple regression was then performed on
the rheological parameters and showed that KS would
depend on the consistency index, a property that has
never been observed before. The viscoelastic behaviour
of concentrated non-Newtonian solutions might be an
other explanation. This assumption will need to be further
refined in a subsequent work.
As for Newtonian fluids, a macroscopic approach can
also be used in order to characterize the power draw
knowing that the particular flow pattern induced by the
rotor-stator may govern the energy dissipation rate in the
vessel. The average shear rates are then calculated based
on equations (20) and (21) for power law or yield stress
fluids, namely.



P=V 1=n1
g_ av
m
n1
P=V ty g_ av mg_ av

(20)
(21)

Shear rates from the non-linear equation (21) are obtained

using a Newton algorithm.
Figure 12 shows the calculated shear rates as a function
of the rotational speed of the impeller for the xanthan gum
solutions. The circled region represents the regime where
the pseudo-cavern did not expand to the vessel walls. It
also corresponds to the region where the uncertainty on
the torque is the highest, so for the calculated shear rates.
As expected, it appears that the shear rates decrease with
increasing shear-thinning.
We have plotted in Figure 13 the shear constant CS as a
function of the rotational speed for xanthan solutions. The
circled region represents the regime before the pseudocavern reaches the walls. It almost corresponds to the
results obtained with Newtonian fluids. As the rotational
speed increases, the shear constant values tend to gather
around a constant value. With this approach, the shear
constant increases when the well-mixed region occupies
the vessel volume.
A hybrid definition of the dimensionless numbers that
uses the pseudo-cavern diameter instead of the impeller

Figure 13. Shear constant versus rotational speed for xanthan solutions.

diameter was also investigated, which yields for the

Reynolds and Power numbers, respectively:

rNDdC
ma
P
Po
3
rN D2 dC3
Re

(22)
(23)

Those definitions emphasize the size of the pseudo-cavern

in the values of the dimensionless numbers and represent
more accurately the energy dissipation in the vessel.
In the Reynolds number, the tip speed (Vtip pND) was
used as the characteristic velocity as suggested by Padron
(2001). For the characteristic dimensions, two choices
were considered. The first definition used the rotor-stator
gap width, dgap, as the characteristic length to emphasize
the dissipation of power in the gap due to viscous shear.
The Reynolds number is then:
Re(gap)

rVtip dgap
ma

(24)

The second definition used the hydraulic radius, Rh, as the

characteristic length. The hydraulic radius was defined as
the area of slots divided by its wetted perimeter. Calabrese
et al. (2002) reported that in the turbulent regime, the
energy dissipation rate is controlled by fluid impingement
on the stator slot surfaces or by the breakdown of the jets
exiting from the slots. Therefore, this definition may be
appropriate as soon as most of the liquid is pumped through
the stator slots. With this choice, the Reynolds number
reads as:
Re(hydraulic)

Figure 12. Shear rates from macroscopic approach versus rotational speed
for xanthan solutions.

1193

4rVtip Rh
ma

(25)

Finally, Rey was also considered to scale power curves.

Figure 14 summarizes the power curves obtained for
the xanthan and CMC solutions. The Newtonian power
curve as well as the power curves obtained with the
Rieger and Novak (1973), Metzner and Otto (1957), macroscopic and hybrid [only in Figure 14(b)] approaches are
reported.

Trans IChemE, Part A, Chemical Engineering Research and Design, 2005, 83(A10): 11861195

1194

DOUCET et al.
approaches did not provide a good shift of the nonNewtonian curves on the Newtonian one. The alternate
definitions of the dimensionless numbers did not give
significant results confirming Calabrese et al. (2002) and
Padron (2001) the previous conclusion of these puzzling
results will need to be further investigated to provide a
better understanding of the behaviour of non-Newtonian
fluids in rotor-stator geometries, with CFD likely to play a
major role.
CONCLUSIONS
It has been shown that caverns are almost an unavoidable
phenomenon when mixing viscous and/or non-Newtonian
fluids with rotor-stator impellers. The experiments were
performed in batch configuration under laminar and
intermediate regimes corresponding to industrial operating ranges (i.e., when preparing personal care emulsion
cream). From a practical standpoint, it is expected to
predict caverns size and power consumption during a
conversion process in order to control the overall mixing
in a tank.
The pseudo-caverns and their growth are described in
Newtonian and non-Newtonian fluids. It appears that the
cavern size and shape do scale with the Reynolds
number. Pseudo-caverns observed in Newtonian fluids
scale with the classical definition of Reynolds number
[see equation (8)] while caverns formed within nonNewtonian fluids scale with Rey [see equation (3)]. Several
regimes were described and allow the designer to predict
the approximate cavern shape when performing scale-up.
Correlations taken from literature (developed for Rushton
turbine) to predict cavern diameter and height were
unable to fit our experimental data due to the particular
mixing behaviour of the rotor-stator impeller.
Several approaches were tested in order to establish
whether it was possible to develop a power master curve.
The Metzner and Otto method was found applicable.
However, it was impossible to determine a unique value of
the Metzner Otto constant, limiting the practical interest
of the approach.
NOMENCLATURE

Figure 14. Non-Newtonian power curves: (a) xanthan at 0.2 wt%,

(b) xanthan at 0.5 wt%, (c) CMC at 1 wt%, and (d) CMC at 2 wt%.

The Metzner and Otto (1957) approach seems to be the

only one that reasonably correlates the power draw data.
However, as seen before, KS cannot be predicted a priori
and an average value cannot be used due to the large variability observed, which limits considerably the interest
of the method. Both the macroscopic and the hybrid

CS
D
dSubscript
e
f.s.
h
hSubscript
H
KP
KS
m
n
MC
MM
MR
N
P
Po
Re
Ren
Rey
Rh
T

shear constant
impeller diameter, m
diameter, m
constant in equation (1)
fullscale
height of the impeller into the vessel, m
height, m
liquid level height, m
power constant
MetznerOtto constant
consistency index, Pa.sn
shear-thinning index
corrected torque, N.m
measured torque, N.m
residual torque, N.m
impeller rotational speed, s21
power drawn by the impeller, W
Power number
Reynolds number
Reynolds number for a power law fluid [see equation (14)]
Reynolds number [see equation (3)]
hydraulic radius, m
tank diameter, m

Trans IChemE, Part A, Chemical Engineering Research and Design, 2005, 83(A10): 11861195

HYDRODYNAMICS OF ROTOR-STATOR MIXER WITH VISCOUS FLUIDS

VSubscript
Vtip

volume, m3
tip speed of the impeller, m s21

Greek symbols
a
constant
g_
shear rate, s21
g_ av
average shear rate, s21
dgap
rotor-stator gap width, m
m
Newtonian viscosity, Pa s
ma
apparent viscosity, Pa s
r
density of the liquid, kg m23
tav
average shear stress in a cavern, Pa
td
representative shear stress in the dissipating zone, Pa
ty
yield stress, Pa
Subscripts
C
rotor
stator

cavern
rotor
stator

REFERENCES
Amanullah, A., Hjorth, S.A. and Nienow, A.W., 1998, New mathematical
model to predict cavern diameters in highly shear thinning, power law
liquids using axial flow impellers, Chem Eng Sci, 53(3): 455469.
Arratia, P.E., Lacombe, J.P., Shinbrot, T. and Muzzio, F.J., 2004,
Segregated regions in continuous laminar stirred tank reactors, Chem
Eng Sci, 59: 1481 1490.
Bourne, J.R., Studer, M., 1992, Fast reactions in rotor-stator mixers of
different size, Chem Eng and Processing, 31(5): 285 296.
Brennan D.J. and Lehrer, I.H., 1976, Impeller mixing in vessels, experimental studies on the influence of some parameters and formulation
of a general mixing time equation, Trans IChemE, 54: 139152.
Brito-de la Fuente, E., 1992, Mixing of rheological complex fluids with
helical ribbon and helical ribbon screw impellers, PhD Thesis, Laval
University, Quebec, Canada.
Calabrese, R.V., Francis, M.K., Kevala, K.R., Mishra, V.P., Padron, G.A.
and Phongikaroon, S., 2002, Fluid dynamics and emulsification in high

1195

shear mixers, 3rd World Congress on Emulsions, Theme II, Conf. 03,
Lyon, France.
Doucet, L., Ascanio, G. and Tanguy, P.A., 2003, Rotor-stator study in
batch geometry: power consumption, mixing behaviour, caverns, 53rd
Canadian Chemical Engineering Conference, Hamilton, Canada.
Elson, T.P., Cheesman, D.J. and Nienow, A.W., 1986, X-ray studies of
cavern sizes and mixing performance with fluids possessing a yield
stress, Chem Eng Sci, 41: 25552562.
Hirata, Y., Nienow, A.W. and Moore, I.P.T., 1994, Estimation of cavern
sizes in shear-thinning plastic fluid agitated by a Rushton turbine
based on LDA measurements, J Chem Eng Japan, 27(2): 235 237.
Hirata, Y. and Aoshima, Y., 1996, Formation and growth of cavern in yield
stress fluids agitated under baffled and non-baffled conditions, Trans
IChemE, Part A, Chem Eng Res Des, 74: 438 444.
Macosko, C.W., 1994, Rheology: Principles, Measurements and Applications (VCH Publishers, Inc., New York, USA).
Metzner, A.B. and Otto, R.E., 1957, Agitation of non-Newtonian fluids,
AIChE J, 3: 3 10.
Myers, K.J., Reeder, M.F. and Ryan, D., 2001, Power draw of a high-shear
homogenizer, Canadian J Chem Eng, 79(1): 94 99
Oldshue, J., 1983, Fluid Mixing Technology (McGraw Hill, New York,
USA).
Padron, G.A., 2001, Measurement and comparison of power draw in batch
rotor-stator mixers, MSc Thesis, University of Maryland, College Park,
MD.
Rieger, F. and Novak, V., 1973, Power consumption of agitators in highly
viscous non-Newtonian liquids, Trans IChemE, 51: 105111.
Solomon, J., Elson, T.P., Nienow, A.W. and Pace, G.W., 1981, Cavern
sizes in agitated fluids with a yield stress, Chem Eng Commun, 11:
143164.
Watson, J.H., 2004, The diabolical case of the recurring yield stress, Appl
Rheol, 14(1): 40 45.
Wichterle, K. and Wein, O., 1975, Agitation of concentrated suspensions.
CHISA75, Prague, Czechoslovakia, Paper B4.6.

ACKNOWLEDGEMENTS
The financial assistance of NSERC and Total S.A. is gratefully
acknowledged.

Trans IChemE, Part A, Chemical Engineering Research and Design, 2005, 83(A10): 11861195