Reynolds Number and Impeller Diameter Effects

on Instabilities in Stirred Vessels

C. Galletti and A. Paglianti
Laboratory of Process Equipment, Dept. of Chemical Engineering, Industrial Chemistry and Materials Science,
University of Pisa, Via Diotisalvi 2, Pisa I-56126, Italy

K. C. Lee and M. Yianneskis

Experimental and Computational Laboratory for the Analysis of Turbulence, Division of Engineering,
King’s College London, Strand, London WC2R 2LS, U. K.

DOI 10.1002/aic.10236
Published online in Wiley InterScience (www.interscience.wiley.com).

Flow instability phenomena in stirred vessels were studied with laser anemometry and
flow visualization. The effects of fluid density and viscosity, impeller Reynolds number
(Re), impeller design, diameter, and off-bottom clearance were investigated in order to
quantify the frequencies (f) of the macroinstabilities stemming from precessional motions.
The instabilities are characterized by two frequencies, one present at low Re, and another
at high Re values. For intermediate Re values, both frequencies were present. In all cases,
f was proportional to the impeller speed (N). The parameter f⬘ ⫽ f/N was shown to be
linearly related to the impeller diameter in the low Re range; f⬘ was not affected by
impeller clearance. At constant Re, a change in fluid density and viscosity did not affect
f’. The energy contained in the instability frequencies was shown to vary across the vessel.
The implications of the results mixing practice are discussed. © 2004 American Institute of
Chemical Engineers AIChE J, 50: 2050 –2063, 2004
Keywords: flow instabilities, precession, stirred vessel, fluid mixing, Reynolds number,
impeller diameter.

Introduction nova et al., 2000), introduce complexities that can affect mix-
The flow patterns encountered in stirred vessels for different ing and circulation times and turbulence levels and, thus, make
operational conditions and vessel/impeller geometries can crit- mixing process prediction and control a most complex task,
ically affect the fluid mixing processes in such apparatus, and, and gross over-simplifications are often necessary. For exam-
consequently, process performance, and, therefore, accurate ple, the r.m.s. levels could be broadened, leading to erroneous
characterization of such patterns is essential for optimization of estimation of the turbulence content (Yianneskis et al., 1987;
mixing process design (Tatterson, 1994). The flow fields are Roussinova et al., 2000).
three-dimensional (3-D), more often than not turbulent, and Although the trailing vortex structures and effects have been
their spatial and temporal features, such as the vortices trailing widely studied, knowledge of flow instabilities is far from
from the impeller blades (Yianneskis et al., 1987) and the complete. These instabilities can affect the mixing vessel per-
instabilities of the flow structures (Hasal et al., 2000, Roussi- formance in different ways, as pointed out by a number of
studies. Haam et al. (1992) observed temporal variations in the
local heat flux and heat-transfer coefficient at the wall of a
Correspondence concerning this article should be addressed to M. Yianneskis at stirred tank of up to 68%, which they attributed to the preces-
[email protected]
sion of an axial vortex structure moving relatively slowly in
© 2004 American Institute of Chemical Engineers comparison to the impeller rotational speed. Further evidence

2050 September 2004 Vol. 50, No. 9 AIChE Journal

is provided by Houcine et al. (1999), who observed jet inter- according to f ⫽ c1 ⫹ c2•N, where c1 ⫽ -0.009 to -0.04 and c2
mittence phenomena in the feedstream by means of a laser ⫽ 0.041– 0.05. In a later work (Bruha et al., 1996), they
induced fluorescence (LIF) technique. reported a linear dependence of the MI frequency on N (f’⫽
Different types of flow instability have been identified hav- 0.043– 0.048) for Re values above 9,000. No macroinstabilities
ing different timescales. One instability stems from variation in were noted for Re ⬍ 200, and an increase in f⬘was observed for
the impeller off-bottom clearance. Even when the flow is 200 ⬍ Re ⬍ 9,000. Montes et al. (1997) studied MIs induced
stable, changes in the impeller stream inclination (Yianneskis by a PBT, and observed different values for f⬘ depending on the
et al., 1987), or in the flow pattern (Jaworski et al., 1991, Kresta impeller Reynolds number: f⬘ ⫽ 0.09 for Re ⫽ 1,140 and f⬘ ⫽
and Wood, 1993) with variations in impeller clearance (C), 0.0575 for Re ⫽ 75,000. Hasal et al. (2000) observed for a PBT
have been reported for a Rushton turbine (RT) and a pitched- an f⬘value of 0.087 for Re ⫽ 750 and Re ⫽ 1,200, and a value
blade turbine (PBT), respectively. For some clearances, unsta- of 0.057 for Re ⫽ 75,000. In addition, they noticed that the
ble flow patterns occur, however. Rutherford et al. (1996) fraction of the total kinetic energy carried by the MIs (relative
studied a stirred tank equipped with two RTs, and found that magnitude) varied with the location inside the stirred vessel.
the flow pattern depended on both the clearance and the sep- Myers et al. (1997) used digital particle image velocimetry to
aration between the impellers, identifying three stable and investigate macroinstabilities in a stirred tank equipped with a
many unstable circulation patterns. In addition, flow variations PBT. They pointed out transients with a period ranging from 40
stemming from the clearance of a single RT were identified by to 300 impeller revolutions (f⬘ ⫽ 0.003– 0.025). Guillard et al.
Montante et al. (1999), who established that the conventional (2000) carried out LIF experiments on a stirred tank equipped
double-loop circulation pattern occurred when C/T ⬎ 0.20, but with two Rushton turbines, observing large time-scale oscilla-
the discharge flow became axial (single-loop) when the clear- tions of the concentration, induced by an interaction between
ance was decreased to C/T ⬍ 0.15. the flows from the impeller and a baffle. Roussinova et al.
A second kind of instability is related to a change in impeller (2000, 2001) performed experiments in two tank sizes, using
Reynolds number (Re). Nouri and Whitelaw (1990) found a various impeller types, clearances and working fluids. A value
transition due to Re variations in the flow pattern induced by a of f⬘⫽ 0.186 was established for the D/T ⫽ 0.5 PBTs studied.
PBT for both Newtonian and non-Newtonian fluids. For New- Recently, macroinstabilities were investigated by Nikiforaki
tonian fluids, the flow pattern transition occurred at about Re ⫽ et al. (2003), who used two different impellers (RT and PBT)
650. This value was confirmed by the power number measure- having the same diameter D/T ⫽ 0.33. The frequency of
ments carried out by Distelhoff et al. (1995). Hockey and Nouri macroinstabilities was found to be linearly related to the im-
(1996) measured the power number for a wide range of Re in peller speed with f⬘⫽ 0.015– 0.020. This value was indepen-
a tank equipped with a PBT. A sudden drop in the power dent of impeller clearance and design. In the aforementioned
number was found when the Re was increased to 1,200, and it article, experiments were carried out using only one fluid, that
was shown by means of flow visualization that the direction of is, water, and the impeller Reynolds number was Re ⬎ 20,000,
the discharge flow changed from radial to axial at this Re value. thus, the flow might be expected to be fully turbulent. This f⬘
They explained the value of 1,200, which is around twice that value is in agreement with those reported in many of the
previously established, by considering the difference in the aforementioned works, but the reasons for the wide range of f⬘
PBT size: since the tank and impeller used by Hockey and values reported for different configurations are manifold: dif-
Nouri (1996) were twice those used in the previous works ferent measurement locations and impeller/vessel geometries
(Nouri and Whitelaw, 1990, Distelhoff et al., 1995), the same have been considered in different studies, and there appears to
Vtip corresponded to the two Re numbers. Schäfer et al. (1998) be a substantial complexity of, and/or a number of possibly
observed by means of flow visualization the flow discharged by related flow phenomena.
a PBT to be directed axially at higher Re and radially at lower Re. The above brief review has shown that there are still many
The flowstream direction was unstable, varying from radial to unresolved questions in relation to MI phenomena in stirred
axial, for Re around 500. Computational fluid dynamics (CFD) tanks and, as MIs can strongly affect mixing processes, im-
results have also shown a change in the flow pattern stemming proved knowledge of the related phenomena could be em-
from a variation in Re. For example, Bakker et al. (2000) predicted ployed to improve mixing performance. The objective of this
that, for the flow discharged by a PBT, the impeller stream work was to shed light into such phenomena through extending
direction was radial for low Re, but axial for Re ⬍ 400. the investigation on macroinstabilities to different working
Finally, a third type of instability due to a precessional fluids, in order to establish the influence of physical properties,
motion of a vortex around the shaft is often detected as a large such as density and viscosity on the related phenomena, a
time-scale variation in the mean flow (termed macroinstabil- wider range of impeller Reynolds numbers, and for different
ity - MI). Yianneskis et al. (1987) observed the aforementioned impeller types and sizes. In this manner, this work aims to
vortex for a RT moving around the shaft with a period of 10 s quantify the effect of ␳, ␮, Re and D on the MIs, in order to
for an impeller rotational speed of N ⫽ 300 rpm (f/N ⫽ 0.02). attempt to explain the different results reported in the literature,
In the following text, a nondimensional MI frequency f⬘ ⫽ f/N and provide a more thorough understanding and improved
is used for convenience. Winardi and Nagase (1991) noticed characterization of mixing processes involved.
large time-scale fluctuations in a stirred vessel equipped with a
marine propeller. The work by Haam et al. (1992) cited earlier
also referred to this kind of instability. Bruha et al. (1995) used Stirred Vessel Configuration and Measurement
a device called a “tornadometer” to estimate the macroinsta- Techniques
bilities induced by a PBT, and they found a linear relation Measurements were made in a cylindrical stirred vessel of
between their frequency f and the impeller rotational speed N, diameter T ⫽ 0.294 m, equipped with four equally spaced

AIChE Journal September 2004 Vol. 50, No. 9 2051

 Table 1. Geometrical Details of the Impellers Studied analysis to confirm that the resolution did not affect the fre-
RT RT PBT RT quencies reported.
Flow visualization experiments were also carried out using a
D [mm] 98 119.5 136 195
D/T [⫺] 0.33 0.41 0.46 0.66
light sheet obtained by a combination of lenses, including a
H b /T [⫺] 0.20 0.20 0.20 cylindrical lens. For the visualization tests, the fluid was seeded
W b /D [⫺] 0.25 0.25 0.25 with polystyrene particles of 800 ␮m mean diameter and the
W/D [⫺] 0.264 resulting images were recorded by a video camera.
t b /D [⫺] 0.016 0.018 0.016 0.018

Results
LDA Experiments
baffles of width B ⫽ T/10. The tank was placed inside a trough
in order to minimize refraction effects, due to the vessel cur- Impeller Reynolds Number Effects. First, radial velocity
vature. The liquid level was equal to the tank diameter (H ⫽ T). data were taken in the stirred tank equipped with the D/T ⫽
Since, as pointed out by Nikiforaki et al. (2003), the presence 0.33 Rushton turbine set at C/T ⫽ 0.5. The measurement
or absence of a lid has little effect on the macroinstabilities, a volume was located close to the free surface (z/T ⫽ 0.90 –
free surface configuration was used. The stirred tank was 0.95), at two different radial distances from the vessel axis
equipped with various impellers. Three RTs of different diam- (r/T ⫽ 0.12 and r/T ⫽ 0.17), and in a plane located midway
eter (D/T ⫽ 0.33, 0.41 and 0.66), and a 4-blade 45° PBT of between two neighboring baffles (␪ ⫽ 45°). No significant
D/T ⫽ 0.46 were used. The geometrical details of the four differences between the results acquired in the two locations
impellers are listed in Table 1. were observed which indicated the macroinstability phenome-
The D/T ⫽ 0.33 Rushton turbine was set at three different non was present in both locations in the vessel. The variation of
clearances (C/T ⫽ 0.15, 0.33, and 0.5). Two clearances were the intensity of the macroinstability across the entire vessel is
used for the D/T ⫽ 0.41 Rushton turbine (C/T⫽ 0.15 and 0.5), discussed later in this section.
and the PBT (C/T ⫽ 0.33 and 0.5). The largest Rushton turbine As already pointed out by Nikiforaki et al. (2003), the
was studied only at C/T ⫽ 0.5. The impeller rotational speed macroinstability frequency was found to be linearly related to
was varied from 60 –350 rpm. the impeller rotational speed. However, an important and hith-
Water and aqueous sucrose solutions with different sucrose erto never reported difference in the proportionality constant
content were used as working fluids to study the flows in a wide arose when considering different impeller Reynolds numbers.
range of impeller Reynolds numbers. The viscosity of the fluids In Figure 1 the variation of the nondimensional macroinstabil-
was measured with a rotational rheometer (ARES Rheometer ity frequency f⬘ is reported against Re for different working
by Rheometric Scientific). The ranges of Re investigated for fluids. The graph may be divided into three regions.
each configuration are listed in Table 2. ● Region 1 (400 ⬍ Re ⬍ 6,300), for which the frequency
Measurements were taken in different locations inside the spectra exhibited a single peak linearly related to the impeller
tank, covering a region from r/T ⫽ 0.12 to 0.46, and from z/T⫽ speed with a value of f⬘ ⫽ 0.106.
0.25 to 0.95. Most of the measurements were taken in the ● Region 2 (6,300 ⬍ Re ⬍ 13,600), for which the frequency
vertical plane located midway two neighboring baffles (␪ ⫽ spectra exhibited the presence of two different peaks, linearly
45°), but a few measurements were also taken for 0° ⱕ ␪ ⱕ 90° related to the impeller speed with f⬘ values of 0.106 and 0.015,
for the PBT set at C/T ⫽ 0.33. All measurement location respectively.
coordinates (r/T, z/T, and ␪) are specified in the figure captions, ● Region 3 (13,600 ⬍ Re ⬍54,400), for which the fre-
and discussed in the relevant text where appropriate. quency spectra exhibited a single peak with f⬘ ⫽ 0.015, in
Measurements were taken with a single-component LDA agreement with the findings of Nikiforaki et al. (2003).
system. A 514 nm wavelength beam was supplied by a 3W The three regions identified above may be considered to
Argon Ion laser, and split by a diffraction grating unit, which correspond to the laminar, transitional and turbulent regimes,
also provided frequency shifting. A Dantec BSA was used to respectively. However, such a classification must be made with
process the Doppler signals. A large number of samples extreme care, as it is well established that the extent of turbu-
(300,000 –700,000) was recorded for each run. The data rate lent flow in a stirred vessel varies with not only Re, but also
varied from 0.4 –1.5 kHz. The errors in the mean velocity and with the location in the vessel (Bittorf and Kresta, 2000): even
r.m.s. values reported were estimated to be 1–2% and 5–10%, if the flow near the impeller is turbulent, it may become
respectively, through extensive comparisons with earlier LDA transitional and even laminar with increasing distance from the
data (for example, Lee and Yianneskis, 1998, and others), and
an analysis of the individual error sources in the manner Table 2. Ranges of Impeller Reynolds Number
suggested by Durst et al. (1981). Velocity data were analyzed Investigated for Each Configuration
using a FFT technique performed through subroutines available
Impeller D/T [⫺] C/T [⫺] Re [⫺]
in the Matlab software package. The randomness of particle
arrival intervals made necessary the resampling of velocity- RT 0.33 0.15 500–9,800
time data in order to obtain data evenly distributed in time, over 0.33 9,100–54,400
0.5 500–54,400
which the FFT could be calculated. Data were interpolated and 0.41 0.15 30–1,700
resampled using the same amount of samples taken during the 0.5 130–53,500
LDA acquisition, thus, the resampling frequency corresponded 0.66 0.5 2,000–6,200
to the mean data rate. Three different frequency resolutions PBT 0.46 0.33 400–47,700
0.5 400–34,700
(␦f ⫽ 0.0075, 0.01, and 0.015 Hz) were tested in the spectral

2052 September 2004 Vol. 50, No. 9 AIChE Journal

 a moving-window-average technique, by calculating each av-
erage over 1,000 samples, and it shows visibly the cyclic
variation.
Figure 3a shows the frequency spectrum obtained for the
same working fluid, but using an impeller speed of N ⫽ 60 rpm
in order to achieve the first (low Re) flow regime (Re ⫽ 5,400).
The peak frequency is f ⫽ 0.108 Hz (f⬘⫽ 0.108), thus, the
period is 9.2 s. This period is indicated in the corresponding
instantaneous velocity variation shown in Figure 3b. The mov-
ing-window-average technique to obtain the black continuous
line was performed by calculating each average over 500
samples. In view of the low Re value for this case, it should be
noted that the velocity fluctuations about the MI mean velocity
variation appear large because the local velocities are small (of
the order of 0.1 m/s), and the ordinate scale has been consid-
erably expanded in relation to that of Figure 2.
Figure 1. Nondimensional macroinstability frequency as A characteristic frequency spectrum acquired in the second
a function of impeller Reynolds number for the (intermediate Re) regime (N ⫽ 120 rpm, Re ⫽ 10,800) is
D/T ⴝ 0.33 Rushton turbine set at C/T ⴝ 0.5, shown in Figure 4a. Two peaks of f ⫽ 0.030 Hz (t ⫽ 33 s) and
and using different working fluids. f ⫽ 0.210 Hz (t ⫽ 4.8 s) can be clearly distinguished. The first
Data obtained at r/T ⫽ 0.12– 0.29, z/T ⫽ 0.3– 0.95, ␪ ⫽ 45°. peak corresponds to a nondimensional frequency value of f⬘ ⫽
0.015, identical to that noted earlier for the high Re region. The
blades. Consequently, in order to avoid confusion, the charac- second peak corresponds to f⬘ ⫽ 0.105, which is the same as
terization low, intermediate, and high Re regions is preferred, that observed for the low Re region. The instantaneous velocity
and used in the text later to denote the three regions identified variation is illustrated in Figure 4b. The black continuous line
earlier. The scatter in the data stems mainly from the frequency was obtained by means of a moving-window-average tech-
resolution employed, and is discussed further later. nique, by calculating each average over 2,000 samples. In this
Figure 2a illustrates a characteristic frequency spectrum manner, a cyclic fluctuation of a period of about 33 s can be
obtained in region 3. identified. In addition, a more detailed observation points out
Measurements were taken with an aqueous sucrose solution the presence of a short period fluctuation superimposed to the
with a sucrose content of 20 wt. %, using an impeller rotational one described earlier. Part of the time series of Figure 4b is
speed of N ⫽ 300 rpm: the Re was 27,000. The peak frequency enlarged in Figure 4c, where the instantaneous velocity is
is f ⫽ 0.075 Hz (f⬘ ⫽ 0.015), which corresponds to a period of shown over a period of only 10 s: the black continuous line was
13.3 s. The cyclic variation of the macroinstabilities are illus- obtained by means of the window-average technique over a
trated in Figure 2b, where the instantaneous velocity is plotted lower number of samples (200). This enabled the detection of
vs. time, and the aforementioned 13.3 s period is clearly in the short period cyclic fluctuation (4.8 s) related to the high
evidence. The black continuous line was obtained by means of frequency peak. The velocity data were analyzed in a similar

Figure 2. Rushton turbine, D/T ⴝ 0.33, C/T ⴝ 0.5, N ⴝ 300 rpm, Re ⴝ 27,000, r/T ⴝ 0.12, z/T ⴝ 0.90, ␪ ⴝ 45°, fⴕ ⴝ 0.015:
(a) frequency spectrum, (b) time series.

AIChE Journal September 2004 Vol. 50, No. 9 2053

Figure 3. Rushton turbine, D/T ⴝ 0.33, C/T ⴝ0.5, N ⴝ 60 rpm, Re ⴝ 5,400, r/T ⴝ 0.12, z/T ⴝ 0.95, ␪ ⴝ 45°, fⴕ ⴝ 0.108:
(a) frequency spectrum, (b) time series.

manner over the entire range of impeller Reynolds numbers and 9), but they also show that neither C nor impeller design
studied. The intermediate Re state was described as the one in affect the f⬘ values obtained.
which two peaks could be observed, and as previously men- Effect of Impeller-to-Tank Diameter Ratio. Velocity data
tioned, corresponded to Re ⫽6,200 –13,600. Frequency spectra were also taken with a larger Rushton turbine (D/T ⫽ 0.41) in
acquired in the intermediate Re region show that the relative order to establish the influence of the impeller size on the
amplitude of the two peaks varies with Re. The magnitude of macroinstability frequency. The impeller was initially set at
the f⬘ ⫽ 0.108 peak decreases with increasing Re, and this peak C/T ⫽ 0.5, and in Figure 6 the variation of f⬘ with Re obtained
disappears for Re ⫽ 13,600. The two peaks have almost the for three of the different working fluids is presented. As already
same amplitude at Re⫽ 10,800, as indicated in Figure 4a. observed for the D/T ⫽ 0.33 Rushton turbine, the plot can be
Effect of Fluid Properties. As stated previously, both water divided into three regions, corresponding to the low, interme-
and different aqueous sucrose solutions were used, enabling an diate and high Re regions. The intermediate Re region extends
increase in the kinematic viscosity by about 40 times (from from Re ⫽ 13,000 to 20,000, so it is found at higher Re values
8.8 䡠 10⫺7 to 3.4 䡠 10⫺ 5 m2/s). As can be seen in Figure 1, the with respect to the smaller (D/T ⫽ 0.33) Rushton turbine.
f⬘ results obtained for different working fluids overlap when The nondimensional macroinstability frequency was f⬘ ⫽
plotted in terms of Re, and this shows conclusively that the 0.018 for the high Re region. This value does not differ sig-
macroinstabilities are affected by Re, but not solely by a nificantly from the one obtained with the smaller Rushton
change in fluid properties, such as density and viscosity. turbine. Moreover, the frequency resolution must be taken into
Effect of Impeller Off-Bottom Clearance. Measurements account, since the finest resolution used in this work, that is,
were also performed with the Rushton turbine placed at off- ␦f ⫽ 0.0075 Hz, is around 50% of the frequency of some of the
bottom clearances of C/T⫽ 0.15, 0.33 and 0.5. The first con- high Re region peaks.
figuration, as shown by Montante et al. (1999), leads to a single In contrast, the nondimensional macroinstability frequency
circulation loop, with the discharge flow directed axially. Fig- obtained for the low Re region was f⬘ ⫽ 0.169, thus higher than
ure 5 shows the macroinstability frequency values obtained the result gained for the D/T ⫽ 0.33 Rushton turbine. This
against the impeller rotational speed for the 3 clearances. Some finding indicates an influence of the impeller-to-tank diameter
of the data can be fitted by a line with a steeper slope (f⬘ ⫽ ratio on the macroinstability frequency for the low Re region.
0.109), and these are the data acquired in the first (low Re) It can be seen from Figure 6 that, as already mentioned in the
region. The remaining data can be fitted by a line with a slope previous section, a variation in fluid properties alone does not
of f⬘ ⫽ 0.015, and those were taken in the third (high Re) affect f⬘, but it is Re that is the characteristic parameter.
regime. It can be noted that data for the three different clear- Experiments were also performed with the D/T ⫽ 0.41
ances both in the low and high Re regions overlap, indicating Rushton impeller placed at C/T ⫽ 0.15 (data not shown here
therefore, the clearance does not affect the macroinstability for brevity, but contained in Figure 9) and they yielded iden-
frequency. As far as the high Re region is concerned, this tical f⬘ values to those obtained with the higher clearance, both
finding agrees with the work of Nikiforaki et al. (2003). Results for the low and the high Re regions. This result confirms that
for the C/T ⫽ 0.33 are found only in the f⬘ ⫽ 0.015 group, since impeller clearance does not affect the macroinstability fre-
they were taken only in the third (high Re) region. The results quency, as also noted for the smaller RT.
obtained for different clearances for the D/T ⫽ 0.41 Rushton For a better understanding of the dependence of macroinsta-
turbine, and the PBT are discussed later in this work, in the bilities on impeller size, measurements were also taken with a
context of impeller diameter and design effects (Figures 6, 7, larger Rushton turbine (D/T ⫽ 0.66) in the low Re region. In

2054 September 2004 Vol. 50, No. 9 AIChE Journal

 Figure 4. Rushton turbine, D/T ⴝ 0.33, C/T ⴝ 0.5, N ⴝ
120 rpm, Re ⴝ 10,800, r/T ⴝ0.12, z/T ⴝ 0.93, ␪
ⴝ 45°, fⴕ ⴝ 0.015 and fⴕ ⴝ 0.105: (a) frequency
spectrum, (b) 100 s time series, (c) 10 s time
series.

Figure 7, f results are plotted against the impeller rotational

speed for the three Rushton turbines. For each impeller, data
taken with different clearances are shown by means of open or
f⬘ ⫽ a 䡠 冉冊
D
T
⫹b (1)

filled symbols. Since, as shown in the previous section, the

impeller off-bottom clearance does not affect the MI fre- The constant values are a ⫽ 0.83 and b⫽ -0.17. The corre-
quency, data taken with different C/T, but with the same D/T, lation coefficient is 0.994. Therefore, as far as the low Re
have been interpolated by means of a single regression line, region and the RT data are concerned, the D/T ratio does affect
which is also illustrated in Figure 7. A set of data (shown by the frequency of the macroinstability phenomena. The single
squares) acquired with the PBT in the low Re region is also D/T measurement with the PBT can also be described through
shown: these data were obtained for two different off-bottom Eq. 1, but the conclusion reached for the RT cannot be con-
clearances (C/T ⫽ 0.33 and 0.5), but they can be fitted well by firmed for this impeller without further measurements.
a single straight line. Thus, the clearance does not affect the A similar analysis could not be carried out for the high Re
frequency of the macroinstabilities induced by a PBT, as al- region in a manner as accurate as for the low Re one, because
ready observed for the RT. Moreover, it can be noted from the peak frequency values were lower, ranging from around
Figure 7 that the slope of the lines increases with the impeller- 2–17 times the finest resolution used in the FFT analysis (␦f ⫽
to-tank diameter ratio. This is illustrated in Figure 8, where the 0.0075 Hz). A linear dependence of the nondimensional mac-
nondimensional macroinstability frequency is plotted against roinstability frequency on the impeller to tank diameter ratio
D/T. The data can be fitted by a straight line given by an was observed, and could be described by f ⬘ ⫽ 0.068(D/T) ⫺
equation of the form 0.007, however, the accuracy due to the frequency resolution

AIChE Journal September 2004 Vol. 50, No. 9 2055

Figure 5. Macroinstability frequency as a function of im-
peller rotational speed for the D/T ⴝ 0.33 Figure 7. Macroinstability frequency as a function of im-
Rushton turbine set a three different clear- peller rotational speed for the different impel-
ances (C/T ⴝ 0.15, C/T ⴝ 0.33 and C/T ⴝ 0.5). lers in the low Re region.
Data obtained at r/T ⫽ 0.12– 0.29, z/T ⫽0.3– 0.95, ␪ ⫽ 45°. Data obtained at r/T ⫽ 0.12– 0.46, z/T ⫽ 0.25– 0.95, ␪ ⫽ 45°.

was not sufficient to draw firm conclusions on this variation. In Figure 9, nondimensional macroinstability frequency data
Moreover, results obtained for the two Rushton turbines (D/ are plotted against impeller Reynolds number for all configu-
T ⫽ 0.33 and D/T ⫽ 0.41), and the PBT were all in the range rations investigated in this work, comprising different impeller
f⬘ ⫽ 0.012– 0.028, thus, they cover a relatively small f interval. types, sizes, and clearances. For some configurations, the ex-
It can be also observed from Figure 8 that macroinstability periments covered all three low, intermediate, and high Re
frequency results acquired for the D/T ⫽ 0.46 PBT fit well with regions. In this analysis, the existence of a dominant peak in the
those obtained with the three Rushton turbines for the low Re frequency spectrum was determined from the ratio of the power
region. As regards to the high Re region, f⬘ values obtained of the particular peak to those of other peaks in the low
with the PBT are essentially identical to those obtained with the frequency region of the spectrum.
Rushton turbines. Therefore, both low and high Re region data Two parameters that could be of interest for an improved
show that impeller design does not affect the macroinstability understanding of the transition from the low to the high Re
frequency. This finding is in agreement with the work of region macroinstability frequency are the range (⌬Re) over
Nikiforaki et al. (2003) who observed nondimensional macro- which the intermediate Re region takes place for the different
instability frequencies f⬘ ⫽ 0.011– 0.022 for both a Rushton cases studied, as well as the highest Re value for which the low
turbine and a 6-bladed PBT with the same D/T ratio, and Re region macroinstability peak is present (Remax,low), and the
operating in the high Re region. This work confirms this finding lowest Re value for which the high Re region peak is found
and shows that it also applies to the low Re region. (Remin,high). An exact quantification of such values would ne-
cessitate an even more extensive data set than the current one,

Figure 6. Nondimensional macroinstability frequency as

a function of impeller Reynolds number for the Figure 8. Nondimensional macroinstability frequency as
D/T ⴝ 0.41 Rushton turbine set at C/Tⴝ 0.5 a function of impeller to tank diameter ratio.
using different working fluids. Rushton turbine and PBT. Data obtained at r/T ⫽ 0.12– 0.46,
Data obtained at r/T ⫽ 0.12– 0.37, z/T ⫽ 0.3– 0.95, ␪ ⫽ 45°. z/T ⫽ 0.25– 0.95, ␪ ⫽ 45°.

2056 September 2004 Vol. 50, No. 9 AIChE Journal

Figure 9. Nondimensional macroinstability frequency as
a function of impeller Reynolds number for Figure 10. Flow visualization images taken every 1 s.
different configurations (impeller design, D/T, Rushton turbine, D/T ⴝ 0.33, C/T ⴝ 0.5, N ⴝ
C/T). 100 rpm, Re ⴝ 2,000.
Data obtained at r/T ⫽0.12– 0.46, z/T ⫽ 0.25– 0.95, ␪ ⫽ 45°.

(dv/T ⫽ 0.10). The latter vortex is weaker and it is moving

but indicative average values can be extracted from these data. closer and toward the vessel axis, decaying as it approaches the
For the D/T ⫽ 0.33 RT, Remax,low is on average around 13,000, shaft (Figure 10 (4)).
and Remin,high 6,000, although with the D/T ⫽ 0.41 RT the The frequency of the macroinstabilities could be approxi-
corresponding values are around 20,000 and 13,000; in both mately estimated by considering the path of the two vortices
cases, thus, the ⌬Re range is approximately 7,000. With the over a period of time, by
PBT the intermediate Re region is encountered at much lower
Re values, with Remax,low being on average around 2,000 and
Remin,high 500, so that ⌬Re is only around 1,500. It is empha-
sized, however, that the above values comprise only indicative
f⫽2䡠
␯␯
2 䡠 ␲ 䡠 r␯
⫽2䡠
2 䡠 ␲ 䡠 r␯ 冉冊
共 ␪ ␯ 䡠 r ␯ /t ␯兲 1 ␪ ␯
⫽ 䡠
t␯ ␲
(2)

averages for the cases studied.

In the above equation, vv is the vortex velocity, and ␪v is the
angle (in radians) covered by the vortices during the period tv
Flow visualization experiments (3 s). The coefficient 2 takes into account the presence of two
Flow visualization was performed to acquire a better under- vortices moving around the shaft at the same time. For exam-
standing of the macroinstability phenomenon, and especially to ple, the circumferential angle travelled by a vortex in 3 s can be
understand better the difference between macroinstability fre- approximately estimated from Figure 10 as ␪v ⫽ 2␲/3. This
quencies for the low and the high Re regions. The laser sheet value leads to a macroinstability frequency of f ⫽ 0.22 Hz,
was directed horizontally, and placed close to the free surface thus, f⬘ ⫽ 0.13, which does not differ much from the expected
to visually observe the macroinstability phenomena near the value (f⬘ ⫽ 0.108).
surface of the liquid. Experiments were carried out for the Figure 11 shows a similar flow visualization experiment for
D/T ⫽ 0.33, and the D/T ⫽ 0.41 Rushton turbines.
Long-term observations of the flow recordings showed that,
at a fixed region near the shaft, the number of vortices crossing
the region corresponded closely to the f⬘ values reported ear-
lier. This was the case for both the low and the high Re regions,
and is illustrated in Figures 10 and 11.
Figure 10 shows images obtained in the stirred vessel with
the D/T ⫽ 0.33 Rushton turbine set at C/T ⫽ 0.5, and using an
aqueous sucrose solution with a sucrose content of 40 wt. %.
The impeller was rotating clockwise with a rotational speed of
100 rpm. The impeller Reynolds number was 2,000; conse-
quently, the low Re flow region was examined. The four video
images shown were grabbed from the recorded tape at 1 s
intervals, covering, thus, a total period of 3 s. In Figure 10 (1)
two vortices can be observed. The vortex indicated by the
green arrow is quite intense, whereas the second one, on the left
side, and indicated by a red arrow, is weaker. The former
vortex is developing and moving around the shaft, the distance Figure 11. Flow visualization images taken every 1 s.
of the vortex centre from the vessel axis being about 60 mm Rushton turbine, D/T ⴝ 0.41, C/T ⴝ 0.5, N ⴝ
(rv/T ⫽ 0.20) and its mean diameter around dv ⫽ 30 mm 125 rpm, Re ⴝ 3,700.

AIChE Journal September 2004 Vol. 50, No. 9 2057

the D/T ⫽ 0.41 Rushton turbine operating with the same fluid. difference between the vortex center locations between the two
The impeller speed was N ⫽125 rpm and, hence, the low Re flow regimes was not significant, and it clearly was not large
region was observed (Re ⫽ 3,700). The images were again enough to explain the difference, that is, the ratio of 7, in the
captured from the recorded video tape at 1 s intervals. As also macroinstability frequencies. The main observation that can be
observed for the smaller impeller, there are two vortices mov- made from the flow visualization is that the low Re region is
ing around the shaft at the same time. The mean diameter of the characterized by more than one vortex moving around the shaft
vortices is approximately 40 mm (dv/T ⫽ 0.14), and the dis- at the same time. These vortices are quite small and they last a
tance from the vessel axis of about 60 mm (rv/T ⫽ 0.20). The few seconds before decaying near the shaft. In contrast, in the
first video image shows two vortices, one partly obscured by high Re region there is a single big vortex, which has a longer
the shaft (indicated by the green arrow) and the other one on lifetime and moves more slowly around the shaft. Even if the
the right (red arrow). The vortices move around the shaft and slightly larger radius at which the high Re case vortex is located
complete almost half a revolution in 3 s. The vortex pointed by is taken into consideration, this would only account for a
the red arrow decays in Figure 11(4). Applying Eq. 2, a frequency around 1.35 times lower than the low Re case. If a
frequency f ⫽ 0.33 Hz, and a nondimensional frequency f⬘ ⫽ combined effect of the number of vortices and vortex center
0.16 is obtained (the value expected from the LDA data is f⬘ ⫽ location (a factor of approximately 2.4) is considered, 7/2.4 ⫽
0.169). The flow visualization showed that, when the impeller 2.9, and this factor of 2.9 cannot be accounted for without
to tank diameter ratio is varied, no significant differences of the further analysis and/or experimentation, but its probable ori-
vortex locations and sizes could be observed. Nevertheless, it is gins are considered further in the Discussion section.
difficult to provide an exact value of those parameters, since
the vortices change position while moving around the shaft.
As far as the high Re region is concerned, the flow visual- Variation of the intensity of macroinstabilities with
ization was performed in the vessel filled with an aqueous location in the vessel
sucrose solution with a sucrose content of 20 wt % and stirred As established by previous works carried out by different
by the smaller Rushton turbine (D/T ⫽ 0.33) placed at C/T ⫽ researchers and with different techniques (LDA, PIV), macro-
0.5. The impeller speed was 200 rpm, leading to an impeller instabilities are not a local phenomenon and exert an influence
Reynolds number Re ⫽ 17,600 that corresponds to the high Re over a large part of the stirred tank.
region for this configuration. The recording showed the pres- In this work, an area near the top of the vessel, that is, at the
ence of only a big vortex, which forms a whirlpool-type flow z/T ⫽ 0.93 level, was selected to carry out further velocity
into the liquid. The laser sheet crosses the vortex and produces measurements with the D/T ⫽ 0.33 Rushton turbine set at
a large bright area, resulting in a lower quality images than for C/T ⫽ 0.5, for both the high and the low Re regions. Velocity
the earlier cases: such poor quality prevents from illustrating data were taken at different radial distances from the vessel
the images taken in the high Re region in this article. The axis, at spatial intervals of 10 mm.
vortex has a mean diameter of about 80 mm (dv/T ⫽ 0.27), The recorded data were studied by means of frequency
thus, it is much larger than the vortices observed for the low Re analysis and the resulting spectra showed clear macroinstability
region. Its distance from the vessel axis is 70 mm (rv/T ⫽ 0.24), peaks for radial positions close to the shaft, up to about 0.25T
consequently, it does not differ significantly from that for the and 0.29T, for the low and the high Re regions, respectively.
low Re region. Moreover, this vortex lasts many seconds and it When moving the LDA measurement volume further nearer the
moves relatively slowly. From the images, it was derived that walls, the frequency spectra became more spread, and they did
the vortex moves slowly and covers a circumferential angle of not exhibit clear peaks. In Figure 12, the instantaneous velocity
about ␲/2 in 6 s. For a similar analysis as provided earlier, Eq. is plotted against time for different radial positions for the high
2 must be written in the form of Eq. 3 for the high Re region, Re region (Re ⫽ 17,600). The black continuous line was
because only a single vortex should be taken into account. obtained by means of the moving-window-average technique,
by performing each average over 1,000 samples. Figures 12a

f⫽
1
2 䡠 t␯
䡠 冉冊
␪␯
␲
(3)
and 12b (r/T ⫽ 0.12 and 0.19, respectively) show a pronounced
fluctuation. When moving toward the walls, the oscillations are
less pronounced. However, the power spectrum of the data of
Figure 12c (r/T ⫽ 0.25) also exhibited a clear peak. Finally, the
When inserting tv ⫽ 6 s and ␪v ⫽ ␲/2 in the above equation, recording of Figure 12d (r/T ⫽ 0.36) is uniform, and its
the resulting macroinstability frequency is f ⫽ 0.042 Hz and f⬘ spectrum did not show any macroinstability peak. This is in
⫽ 0.013. good agreement with the flow visualization observations,
The previous calculations have been made in an approximate which revealed the presence of vortices moving around the
manner, however, they facilitate a quantitative explanation of shaft at distances from the vessel axis of about rv/T ⫽ 0.20 and
the difference between the macroinstability frequencies ob- rv/T ⫽ 0.23, for the low and the high Re regions, respectively.
served for the low and the high Re regions. For example, the As far as experiments carried out with the impeller placed at
D/T ⫽ 0.33 Rushton turbine exhibits a macroinstability fre- the lowest clearance (C/T ⫽ 0.15) are concerned, the frequency
quency for the low Re region which is 0.108/0.015 ⫽ 7 times spectra of the velocity data recorded near the free surface did
higher than that obtained for the high Re one. Possible reasons not allow the identification of a clear peak. This may be
for this difference between the two Re ranges is that the explained by considering that the impeller was located too far
vortices move faster, or they travel a shorter path (that is, at from the surface to produce sufficiently strong macroinstabili-
smaller radii) in the same time, or that there is more than one ties in the top part of the vessel. In addition, the free surface
vortex present. However, the flow visualization showed that the appeared smooth and no vortices could be observed. However,

2058 September 2004 Vol. 50, No. 9 AIChE Journal

 transition in the flow pattern occurs, leading to the observations
of different frequency peaks. It would be interesting in future
work to provide evidence of the flow pattern transition, and
determine if there is a mechanism through which it affects the
macroinstabilities. For Re ⫽ 500, only the frequency peak
characteristic of the low Re region (f⬘ ⫽0.189) was noted. For
Re ⫽ 2,000, only a low-frequency peak (f⬘ ⫽ 0.034) typical of
the high Re region was observed. For Re ⫽ 1,000, both peaks
could be distinguished in the frequency spectrum. Velocity
data were taken in a rectangular grid at vertical levels of z/T ⫽
0.25– 0.95, and radial distances of r/T ⫽ 0.12– 0.43, in 28
locations. The frequency spectra were analyzed to estimate the
ratio of the amplitude of the macroinstability peak, and the
background frequency content, to give an indication of its
distribution. This is useful in order to establish the regions
where macroinstabilities manifest themselves strongly, and can
be easily detected.
Figure 13a shows, by means of a contour plot, the peak (aLP)
Figure 12. Time series acquired at z/T ⴝ 0.93, ␪ ⴝ 45°.
to background (aB) amplitude ratio for the low Re region. The
Rushton turbine, D/T ⫽ 0.33, C/T ⫽ 0.5, N ⫽ 200 rpm,
Re ⫽ 17,600: (a) r/T ⫽ 0.12, (b) r/T ⫽ 0.19, (c) r/T ⫽ 0.25, macroinstability peaks are more pronounced in the impeller
(d) r/T ⫽ 0.36. region. Below the impeller, the peaks are less pronounced, but
they can still be distinguished. Finally, above the impeller, the
peaks are clearly defined for radial locations up to about r/T ⫽
as the measurement volume was moved closer to the impeller, 0.2, but less so nearer the vessel walls.
that is, at the level z/T ⫽ 0.30, frequency spectra characterized Before considering the intermediate Re region results in
by well-defined peaks were obtained. Figure 13b, it is instructive to examine Figure 13c, where the
As far as the PBT is concerned, since clearances of C/T ⫽ peak amplitude (aHP) to background (aB) ratio is illustrated for
0.33 and C/T ⫽ 0.5 were studied, peaks were identified in the the high Re region. It can be observed that the macroinstability
frequency spectra of the data recorded near the free surface. frequency peaks are well defined near the free surface, espe-
However, the spectra were more spread compared to the ones cially close to the shaft, while they are smaller near the shaft
obtained with the Rushton turbines, as previously noted by and the impeller. However, the MI peak amplitude is signifi-
Nikiforaki et al. (2003), who observed many harmonics in cant in the vessel bulk, even near the walls.
addition to the main peak. In that work, this was considered to The dissimilarity between the contour plots obtained in low
be due to the breaking up and reformation of the precessing and high Re regions indicates a different behavior of the
vortex at irregular time intervals. macroinstabilities. In the low Re region, the phenomenon
In this work, measurements were made in a vertical plane seems to be more pronounced near the impeller, and weaker
containing the vessel axis with the vessel stirred by a PBT set toward the free surface, strongly affecting a small region above
at C/T ⫽ 0.5. The working fluid was an aqueous sucrose the impeller around r/T ⫽ 0.2. In the high Re region, the
solution with sucrose content of 57 wt. %. Velocity measure- macroinstability amplitude is less dominant near the impeller,
ments were performed using impeller rotational speeds of 50, extending to the walls and strongly affecting the free surface
100, and 200 rpm, corresponding to impeller Reynolds num- flow. This finding agrees with the flow visualization, which
bers of 500, 1,000 and 2,000, respectively. These values cor- established the presence of a big vortex moving around the
respond to the low, intermediate, and high Re regions for the shaft near the surface for the high Re region, whereas smaller
PBT, as indicated by the results shown earlier in Figure 9. vortices were observed locally for the low Re region.
Clearly, the value of Re ⫽ 2,000 is much lower than that of Figure 13b shows data at an impeller rotational speed of N ⫽
20,000, which is normally considered as that above which 100 rpm, which corresponds to the intermediate Re state, where
turbulent flow is present in the vessel. Therefore, it should not two peaks in the frequency spectra were observed. The colored
be expected that the flow across the entire vessel is turbulent at contour plot represents the ratio between the amplitude of the
Re ⫽ 2,000, and for this reason the PBT results were checked low- frequency peak (aHP, characteristic of the high Re region)
and repeated very carefully. The same peaks were found, which and the background (aB). The contour lines refer to the low Re
may indicate that at least at some regions in the vessel the region; they show the values of the ratio between the high-
transition to turbulence occurs at relatively low Re with the frequency peak (aLP, characteristic of the low Re region) and
PBT. This may be partly related to, or triggered by, the change the background (aB). It can be noted that the high Re case peak
in the flow pattern due to the Re number. As cited in the is stronger close to the shaft and the free surface, while the low
introduction, many studies (Nouri and Whitelaw, 1990; Re peak is more concentrated in the impeller region. In con-
Hockey and Nouri, 1996; Bakker et al., 2000) observed the clusion, the relative importance of the two peaks depends on
flow discharged by a PBT to become radial when the impeller the location inside the vessel. Near the surface, and in the large
Reynolds number was decreased. The values of Re at which the part of the vessel bulk, the macroinstability phenomena are
transition occurred varied between different works, however, similar to those observed for the high Re region, whereas above
they were in the range of Re ⫽ 400 –1,200. Since these values and close the impeller, the low Re peak is dominant.
are close to those of this work, it may be assumed that a As mentioned in the introduction, the macroinstabilities in

AIChE Journal September 2004 Vol. 50, No. 9 2059

Figure 13. Contour plot of macroinstability peak magnitude to background ratio, for the PBT set at C/T ⴝ 0.5 using
an aqueous sucrose solution (57 wt. %) as working fluid, for: (a) low Re region (N ⴝ 50 rpm, Re ⴝ 500); (b)
intermediate Re region (N ⴝ 100 rpm, Re ⴝ 1,000); (c) high Re region (N ⴝ 200 rpm, Re ⴝ 2,000).
Data obtained at r/T ⫽ 0.12– 0.42, z/T ⫽ 0.25– 0.95, ␪⫽ 45°.

stirred vessels equipped with a PBT have been widely inves- low frequency peak amplitude) was estimated for each location
tigated in literature (Bruha et al., 1995, 1996; Montes et al., and illustrated through the contour plot shown in Figure 16.
1997; Hasal et al., 2000; Myers et al., 1997; Roussinova et al., The region between two baffles could not be completely cov-
2000, 2001; Nikiforaki et al., 2003), but the spread of the ered, because of lack of optical access. Values above 1, that is,
reported f⬘ values is significant. As also pointed out in the colors from yellow to red, indicate that the f⬘ ⫽ 0.18 peak is the
introduction, this may be due to different flow phenomena, dominant one. It can be noted that, although the low frequency
measurement locations, and/or approaches. peak is the dominant one over a large part of the investigated
In this work, a further possible explanation of differences in region, there is a significant region closer to the wall, where the
f⬘ was identified. To study this, the PBT was set at C/T ⫽ 0.33 high frequency peak amplitude is dominant. A possible reason
in the vessel, which was filled with an aqueous sucrose solution is that the baffles cause the formation (shedding) of vortices
of 40 wt. %. The impeller speed was N ⫽ 200 rpm, resulting in behind them, leading to higher frequency phenomena. This
a Re ⫽ 11,000, which, according to the considerations made could explain the high frequency peaks noted by many authors.
earlier for the PBT, should correspond to the high Re region. For example, Roussinova et al. (2001) reported a peak of f’ ⫽
The impeller was rotating clockwise, as viewed from above a 0.18 with a 45° PBT having D/T ⫽ 0.5, and they took mea-
horizontal plane located at z/T ⫽0.25, was selected to carry out surements upstream of baffles for z/T ⫽ 0.40 – 0.80.
the velocity measurements, and, because of the internal sym-
metry of the vessel, just one-quarter of this plane (that is, an
area described by an angle of 90°) was investigated. Velocity Discussion
data were recorded in a measurement grid obtained by varying The presence of macroinstabilities in stirred vessels may
the location in steps of radial distance of 10 mm, and of polar have beneficial implications for mixing process operation and
angle of 15°. The shape of the frequency spectra associated to efficiency, as such flow motions can enhance mixing through
velocity data was analyzed for each location. mean-flow variations. For example, the associated low-fre-
For many locations, the frequency spectra showed only the quency, high-amplitude oscillatory motions in regions of low
low frequency peak of f⬘ ⫽ 0.02, characteristic of the high Re turbulence in a vessel, have the capability of transporting
region, as the spectrum illustrated in Figure 14a. However, in substances fed to a mixing process over relatively long dis-
a few locations a second peak with f⬘ ⫽ 0.18 was observed. In tances (Larsson et al., 1996). In addition, MIs could have
Figure 14b a frequency spectrum with this latter peak is shown. similar effects to those reported for laminar mixing in stirred
In many locations, the frequency spectra exhibited both peaks. tanks by Murakami et al. (1980), and Nomura et al. (1997):
This is the case shown in Figure 15a. In addition, if a logarith- additional raising and lowering of a rotating impeller and/or the
mic-logarithmic representation is used as in Figure 15b and a reversal of the rotation of an impeller produced unsteady mean
single characteristic frequency is sought, the amplitude of the flow motions that either destroyed segregated regions or pre-
high frequency peak might be deemed as more clearly defined venting them from forming, and could produce desired mixing
in relation to the lower frequency peak. times with energy savings of up to 90% in comparison to
For a better understanding of the phenomenon, the amplitude normal impeller operation.
ratio of the two peaks (aHF, the high frequency over aHP, the A thorough understanding of the parameters and mecha-

2060 September 2004 Vol. 50, No. 9 AIChE Journal

Figure 14. Frequency spectra of velocity data acquired with a PBT set at C/T ⴝ 0.33, using an aqueous sucrose
solution (40 wt. %) as working fluid.
N ⫽ 200 rpm, Re ⫽ 11,000, z/T ⫽ 0.25: (a) r/T ⫽ 0.31, ␪ ⫽ 75°, (b) r/T ⫽0.41, ␪⫽ 45°.

nisms responsible for the generation and control of MIs could, swirling flow system of considerable complexity. A recent
therefore, aid the improvement of mixing process operation. study on simpler systems, precessional radial jets induced by
The manifestation of the macroinstabilities has been indicated rotating bodies on thin layers (Vladimirov et al., 2001), re-
by this, as well as previous works (Yianneskis et al., 1987; ported the jet precessional speed to be linearly related to the
Haam et al., 1992; Nikiforaki et al., 2003), as a precessional propeller rotational speed. In that work a large variety of
motion of one or more vortices around the shaft. rotating bodies was investigated (propellers having different
Vortex precession in swirling flows stems from a flow- number of blades and disks), and it was established that the jet
induced precessional velocity, and the frequency of the preces- precessional speed is not affected by the rotating body design.
sion is linearly related to the swirl intensity, while under certain This agrees with the findings of this work, where by means of
flow conditions more than one precessing vortices may be experiments carried out with both a RT and a PBT, it was
present (Alekseenko et al., 1999). A stirred vessel comprises a shown that impeller design does not affect the macroinstability

Figure 15. Frequency spectrum of velocity data acquired with a PBT set at C/T ⴝ 0.33, using an aqueous sucrose
solution (40 wt %) as working fluid. N ⴝ 200 rpm, Re ⴝ 11,000, z/T ⴝ 0.25, r/T ⴝ 0.26, ␪ ⴝ 15°: (a) linear plot,
(b) logarithmic - logarithmic plot.

AIChE Journal September 2004 Vol. 50, No. 9 2061

 relation to low Re flow, this may provide some justification for
the unaccounted for difference in the low and high Re region
frequencies, and should be investigated further. The observed
change in f⬘ with Re range, and in particular the influence that
the extent of low, intermediate, and high Re flows in stirred
vessels has on this change, constitute important research ques-
tions that should be addressed in future work. Their signifi-
cance for process prediction can be considerable because of the
important effect turbulence can have on micromixing in par-
ticular.

Conclusions
In this work macroinstabilities have been studied using dif-
ferent working fluids, impeller types, diameters, and clear-
ances. The already established linear dependence of the mac-
roinstability frequency on the impeller rotational speed f/N
⬇0.02 was confirmed, however, it was found that macroinsta-
bilities exhibited a different behavior when considering the
low, intermediate, and high Re regions. This leads to different
Figure 16. Contour plot of ratio of high frequency peak values of the proportionality constant. Specifically, it was
to high Re peak magnitudes for the PBT set at found that in the low Re region macroinstabilities occurred
C/T ⴝ 0.33, using an aqueous sucrose solu- with a nondimensional frequency about 7 times greater than
tion (40 wt %) as working fluid. that observed for the high Re region, for the case of the D/T ⫽
Data obtained at z/T ⫽ 0.25, r/T ⫽ 0.12– 0.46, ␪⫽ 0 –90°. 0.33 Rushton turbine. For the high Re region, flow visualiza-
tion established the presence of one big vortex that moves
slowly around the shaft and lasts many seconds. For the low Re
frequency. Moreover, Vladimirov et al. (2001) found that the region, two small vortices moving relatively faster around the
proportionality constant between the jet precessional speed, shaft and lasting just a few seconds before decaying near the
and the rotating body rotational speed was dependent on the shaft, were observed. When using different working fluids, no
ratio between the rotating body diameter and the layer diame- differences of the macroinstability frequencies were observed,
ter, when the other geometrical parameters are constant. This indicating that the phenomenon is not affected by the fluid
dependence was only found for rotating body diameters larger properties alone.
than 0.1 times the thin layer diameter; this is the case of this When different impeller off-bottom clearances were used, no
work, as the above ratio may be replaced with the impeller to significant variation in the macroinstability frequency was ob-
tank diameter ratio. These findings provide strong support for served, therefore, the clearance should not exert any effect on the
the conclusion that macroinstabilities in stirred vessels stem macroinstability, both for the low and the high Re regions. This
predominantly, if not exclusively, from the precessional motion finding was obtained both for the Rushton turbine and the PBT.
of a vortex (or a number of vortices) around the shaft. In contrast, a difference in the frequency was found when the
However, the reasons for the difference in the f⬘ values found impeller diameter was varied. This result was well established
for low and high Re are not clear. Guo et al. (2001) studied for the low Re region, and a linear dependence of the nondi-
turbulent swirling flow in a pipe downstream of a sudden expan- mensional macroinstability frequency on the impeller to tank
sion, and found that a precessing vortex was present for low swirl diameter ratio was established f⬘ ⫽ a 䡠 (D/T) ⫹ b. A similar
numbers (S) and an additional, around 10 times higher in fre- analysis could not be carried out as accurately for the high Re
quency, oscillation, was superimposed to the lower-frequency regime, because the peak frequencies were only a few times
precession for S ⬎ 0.13. The low-order precession almost stopped greater than the finest resolution used in the frequency analysis,
when S ⫽ 0.17, and reversed direction for S ⬎ 0.22, while the preventing the formulation of a reliable relationship between f⬘
high-order one remained in the same direction. They suggested the and D/T. However, a similar relation to that above is indicated
latter might have been associated with vortex shedding. Although by the results, with a far smaller dependence of f⬘ on D/T. In
vortex shedding from baffle tips may affect the frequencies de- addition, it was established that impeller design does not affect
tected in this work, the effect of increasing N (and correspondingly the macroinstability frequencies both for the low and the high
S) is the opposite to that of Guo et al. (2001), and the low Re flows Re regions.
are essentially laminar. The data presented provide an improved characterization and
There is no direct evidence that the MI phenomenon is understanding of macroinstability phenomena in stirred ves-
coupled with the flow turbulence; an estimation of the integral sels. The mechanism of transition from the low to the high Re
turbulence time scale based on the direct measurements by Lee macroinstability frequency is not, however, fully understood
and Yianneskis (1998), indicates that the MI period is around and requires further study. In view of the potential benefits of
50 times higher and, therefore, the power contained in the MI MIs for the improvement of mixing efficiency in stirred vessels
peaks in the spectrum may be attributed to coherent fluctua- mentioned in the previous section, more exact knowledge of
tions due to the precession. However, if in the high Re region the mechanisms involved in their generation and control could
the precessing vortex fluctuates more about its mean position in supplement the optimization of mixing processes.

2062 September 2004 Vol. 50, No. 9 AIChE Journal

Notation a Turbine-Agitated Tank using a Fluorescence Technique,” Exp. Fluids,
28, 225 (2000).
a, b ⫽ constants, Eq. 1 Guo, B., T. A. G. Langrish, and D. F. Fletcher, “Simulation of Turbulent
aB ⫽ amplitude of the background in the frequency spectrum, m2/s Swirl Flow in an Axisymmetric Sudden Expansion,” AIAA J., 39, 96
aHF ⫽ amplitude of the high frequency peak in the frequency spec- (2001).
trum, m2/s Haam, S., R. S. Brodkey, and J. B. Fasano, “Local Heat-Transfer in a
aLP ⫽ amplitude of the low Re peak in the frequency spectrum, m2/s Mixing Vessel using Heat-Flux Sensors,” Ind. Eng. Chem. Res., 31,1384
aHP ⫽ amplitude of the high Re peak in the frequency spectrum, m2/s (1992).
B ⫽ baffle width, m Hasal, P., J. L. Montes, H. C. Boisson, and I. Fort, “Macro-Instabilities of
C ⫽ impeller off-bottom clearance, m Velocity Field in Stirred Vessel: Detection and Analysis,” Chem. Eng.
c1, c2 ⫽ constants Sci., 55, 391 (2000).
D ⫽ impeller diameter, m Hockey, R. M., and M. Nouri, “Turbulent Flow in a Baffled Vessel Stirred
dv ⫽ precessing vortex mean diameter, m by a 60° Pitched Blade Impeller,” Chem. Eng. Sci., 51, 4405 (1996).
f ⫽ macroinstability frequency, Hz Houcine, I., E. Plasari, R. David, and J. Villermaux, “Feedstream Jet
f⬘ ⫽ nondimensional macroinstability frequency, f⬘ ⫽ f/N Intermittency Phenomenon in a Continuous Stirred Tank Reactor,”
H ⫽ liquid height, m Chem. Eng. J., 72, 19 (1999).
Hb ⫽ blade height for the Rushton turbine, m Jaworski, Z., A. W. Nienow, E. Koutsakos, E., K. Dyster, and W. Bujalski,
N ⫽ impeller rotational speed, 1/s “An LDA Study of Turbulent Flow in a Baffled Vessel Agitated by a
r ⫽ radial co-ordinate measured from the vessel axis, m Pitched Blade Turbine,” Chem. Eng. Res. Des., 69, 313 (1991).
Re ⫽ impeller Reynolds number, Re ⫽ ␳ 䡠 N 䡠 D/␮ Kresta, S. M., and P. E. Wood, “The Mean Flow Field Producted by a 45°
Remax,low ⫽ highest impeller Reynolds number at which the low Re peak Pitched Blade Turbine: Changes in the Circulation Pattern due to Off
is present in the frequency spectrum of velocity data Bottom Clearance,” Can. J. Chem. Eng., 71, 42 (1993).
Remin,high ⫽ lowest impeller Reynolds number at which the high Re peak Larsson, G., Tornkvist, M., Stahl Wernersson, E., Trägårdh, K., Noorman,
is present in the frequency spectrum of velocity data H., and S.-O. Enfos, “Substrate Gradients in Bioreactors: Origin and
rv ⫽ distance from the precessing vortex centre to the vessel axis, Consequences,” Bioprocess. Eng., 14, 281 (1996).
m Lee, K. C., and M. Yianneskis, “Turbulence Properties of the Impeller
S ⫽ swirl number Stream of a Rushton Turbine,” AIChE J., 44, 13 (1998).
T ⫽ tank inner diameter, m Montante, G., K. C. Lee, A. Brucato, and M. Yianneskis, “Double- to
tb ⫽ disk and blade thickness, m Single- Loop Flow Pattern Transition in Stirred Vessels,” Can. J. Chem.
tv ⫽ period of time in which the precessing vortex transcribes the Eng., 77, 649 (1999).
angle ␪v, s Montes, J. L., H. C. Boisson, I. Fort, and M. Jahoda, “Velocity Field
vv ⫽ precessing vortex velocity, m/s Macro-Instabilities in an Axially Agitated Mixing Vessel,” Chem. Eng.
W ⫽ blade width for the PBT, m J., 67, 139 (1997).
Wb ⫽ blade width for the Rushton turbine, m Murakami, Y., T. Hirose, T. Yamato, H. Fujiwara, and M. Ohshima,
z ⫽ axial co-ordinate measured from the vessel bottom, m “Improvement in Mixing of High Viscosity Liquid by Additional Up-
and-Down Motion of a Rotating Impeller,” J. Chem. Eng. Jpn., 13, 318
Greek letters (1980).
Myers, K. J., R. W. Ward, and A. Bakker, “A Digital Particle Image
␦f ⫽ frequency resolution, Hz Velocimetry Investigation of Flow Field Instabilities of Axial-Flow
⌬Re ⫽ range of impeller Reynolds numbers for which the low and Impellers,” J. Fluid. Eng., 119, 623 (1997).
high Re peaks are both present in the frequency spectrum of Nikiforaki, L., G. Montante, K. C. Lee, and M. Yianneskis, “On the Origin,
velocity data Frequency and Magnitude of Macroinstabilities of the Flows in Stirred
␪ ⫽ tangential co-ordinate of the angle measured from a baffle in Vessels,” Chem. Eng. Sci. 58, 2937 (2003).
clockwise direction, degrees Nomura, T., T. Uchida, and K. Takahashi, “Enhancement of Mixing by
␪v ⫽ angle covered by the precessing vortex during the interval of Unsteady Agitation of an Impeller in an Agitated Vessel,” J. Chem. Eng.
time tv, radians Jpn., 30, 875 (1997).
␮ ⫽ dynamic viscosity, Pa 䡠 s Nouri, J. M., and J. H. Whitelaw, “Flow Characteristics of Stirred Reactors
␳ ⫽ density, kg/m3 with Newtonian and Non-Newtonian Fluids,” AIChE J., 36, 627 (1990).
Roussinova, V. T., B. Grgic, and S. M. Kresta, “Study of Macro-Instabil-
ities in Stirred Tanks using a Velocity Decomposition Technique,”
Literature Cited Chem. Eng. Res. Des., 78, 1040 (2000).
Alekseenko, S. V., P. A. Kuibin, V. L. Okulov, and S. I. Shtork, “Helical Roussinova, V. T., R. Weetman, and S. M. Kresta, “Large Eddy Simulation
Vortices in Swirl Flow,” J. Fluid Mech., 382, 195 (1999). of Macro-Instabilities in a Stirred Tank with Experimental Validation at
Bakker, A., R. D. La Roche, M. Wang, and R. Calabrese, “Sliding Mesh Two Scales,” NAMF 2001, (2001).
Simulation of Laminar Flow in Stirred Reactors,” Published in The Rutherford, K., K. C. Lee, S. M. S. Mahmoudi, and M. Yianneskis,
online CFM Book at http://www.bakker.org.cfm (2000). “Hydrodynamic Characteristics of Dual Rushton Impeller Stirred Ves-
Bittorf, K., and S. Kresta, “Limits of Fully Turbulent Flow in a Stirred sels,” AIChE J., 42, 332 (1996).
Tank,” Proc. 10th European Conf. on Mixing, Delft, The Netherlands Schäfer, M., M. Yianneskis, P. Wächter, and F. Durst, “Trailing Vortices
(2000). Around a 45° Pitched-Blade Impeller,” AIChE J., 44, 1233 (1998).
Bruha, O., I. Fort, and P. Smolka, “Phenomenon of Turbulent Macro- Tatterson, G. B., Scaleup and Design of Industrial Mixing Processes,
Instabilities in Agitated Systems,” Collect. Czech. Chem. C., 60, 85 McGraw-Hill, New York (1994).
(1995). Vladimirov, V. A., V. I. Yudovich,M. Y. Zhukov, and P. V. Denissenko,
Bruha, O., I. Fort, P. Smolka, and M. Jahoda, “Experimental Study of “Asymmetric Flows Induced by a Rotating Body in a Thin Layer,”
Turbulent Macroinstabilities in an Agitated System with Axial High- HIMSA Report, University of Hull, UK (2001)
Speed Impeller and with Radial Baffles,” Collect. Czech. Chem. C., 61, Winardi, S., and Y. Nagase, “Unstable Phenomenon of Flow in a Mixing
856 (1996). Vessel with a Marine Propeller,” J. Chem. Eng. Jpn., 24, 243 (1991).
Distelhoff, M. F. W., J. Laker, A. J. Marquis, and J. Nouri, “The Appli- Yianneskis, M., Z. Popiolek, and J. H. Whitelaw, “An Experimental Study
cation of a Strain Gauge Technique to the Measurement of the Power of the Steady and Unsteady Flow Characteristics of Stirred Reactors,” J.
Characteristics of 5 Impellers,” Exp. Fluids., 20, 56 (1995). Fluid Mech., 175, 537 (1987).
Durst, F., Melling, A., and J. H. Whitelaw, Principles and Practice of
Laser-Doppler Anemometry, Academic Press, London (1981).
Guillard, F., Trägårdh, C., and L. Fuchs, “A Study of Turbulent Mixing in Manuscript received Apr. 16, 2003, and revision received Dec. 18, 2003.

AIChE Journal September 2004 Vol. 50, No. 9 2063