Agitation of Non-Newtonian Fluids

A. 8. METZNER and R. E. OTTO

University of Delaware, Newark, Delaware

Since the shear rate of a non-Newtonian fluid is of importance in fixing the rheological
or viscometric behavior of such a material, the present study has been conoerned with the
development of a general relationship between impeller speed and the shear rate of the
fluid. The resulting relationship was then used to interpret and correlate power-consumption data on three non-Newtonian fluids by use of a generalized form of the conventional
power-number-Reynolds-number plot for Newtonians.
Flat-bladed turbines from 2 to 8 in. in diameter were used exclusively. Tank diameters
ranged from 6 to 22 in. and power inputs from 0.5 to 176 hp./l,OOO gal. The study encompassed a 130-fold range of Reynolds numbers in the laminar and transition regions.
The results to date indicate that power requirements for the rapid mixing of non-Newtonian fluids are much greater than for comparable Newtonian materials.

The literature in the field of agitation

and mixing of fluids may be divided into
two general categories. First and most
extensive are the numerous papers dealing
with the dependence of power consumption upon geometric, kinematic, and
fluid-property parameters. The broad
fundamental and applications studies of
Mack, Miller, Oldshue, Rushton, and
various coworkers (5, 8, 15, 21, 62, for
example) are representative of work of
this nature. Studies of rates and quality
of mixing, and of the effect of agitator
geometric variables on these factors
comprise the second group. The com-

plexity of this problem has resulted in

little quantitative progress in this field
except in the mixing of particulate
solids, and the literature, therefore,
presents primarily qualitative guides as
opposed to the quantitative generalizations available for prediction of power
requirements. The most recent review by
Rushton (LO) discusses progress in this
field.
A major limitation of the prior art is
that it deals almost exclusively with
Newtonian fluids. The comparative simplicity of Newtonian systems has made
them the logical first choice for explora-

Fig. 1. Mixing system.

Fig. 2. Power-number-Reynolds-number
curve for Newtonian fluids in a baffled
tank (22).

tory work, but because of the importance

of non-Newtonian materials a study of
these systems is long overdue. The
purposes of the present investigation
therefore were as follows: (1) establishment of the quantitative relationships
between power consumption and the
geometric, kinematic, and fluid property
parameters for non-Newtonian fluids;
(2) qualitative study of rates and quality
of mixing in non-Newtonian systems to
shed some preliminary light on these
factors; and attaining (3) a quantitative
method of approach in classification of
non-Newtonian fluids and ascertaining
its relation t o mixing.
REVIEW OF PRIOR ART

In the study of the agitation or mixing of

fluids, the system which has received the
most attention consists of a single impeller
centered in a cylindrical tank, as shown in
Figure 1. The results of Newtonian powerconsumption studies are presented in terms
of dimensionless groups involving power
(Pgc/D5N3p)and a mixing Reynolds number
(D2Np/p)or modifications of these. Below a
Reynolds number of 300 the Froude number
(DNZ/g), which measures the variation in
flow due to changes in the free surface, was
not important (22). The effects of geometrical parameters other than the impeller
diameter [such as ( C / D ) , (TID),and
(BID)]were not important within the wide
ranges specified by previous workers.
The entire power-number-Reynolds-number
curve (Figure 2) has been divided into three
sections which are directly analogous t o the
familiar three regions of flow in a circular
pipe, i.e., the turbulent region in which
the power number (or friction factor) is
not greatly affected by Reynolds number,
the laminar region where power number and
friction factor are inversely proportional to
Reynolds number, and the intermediate or
transition region. However, unlike the case
of flow in a round tube, the transition from
laminar to turbulent flow does not occur
over the narrow rangc of Reynolds numbers
between 2,000 and 4,000 (6) but extends
over the large range from 10 to 1,000, as
shown in Figure 2.
To date only three papers have been
concerned with the agitation-power requirements of non-Newtonian fluids. Brown and
Petsiavas (2) presented a power-number
plot for a Bingham-plastic type* of nonNewtonian that makes use of the Binghamplastic Reynolds and Hedstrom numbers
(4, IS) t o correlate the data. Their viscometric data, taken with a Brookfield
viscometer, indicated that their fluids
deviated appreciably from Bingham-plastic
behavior, but for those few fluids which
closely approach the ideal Bingham plastic
this method of attack can be reworked into
*Appendix A presents and discusses the classical
types of non-Newtonian behavior.

Vol. 3, No. 1

A.1.Ch.E. Journal

Page 3

a rigorous and convenient design method.

Magnusson (9) reported a procedure for
calculating the apparent viscosity of nonNewtonian fluids in agitated tanks by
comparison with the power number curve
for a Newtonian fluid but presented no
method whereby such results might be
used for equipment design. Schultz-Grunow
(23) studied power requirements for the
agitation of slightly non-Newtonian fluids
in the laminar region and developed a correlation by means of dimensional analysis.
Since fluid density was not included in the
analysis, this correlation cannot be used
outside the laminar region and may not
accurately predict the end of the laminarflow region. These are rather serious limitations in view of the fact that little act,ual
mixing of non-Newtonian fluids appears t o
occur within the laminar region.
Summarizing, it may be concluded that
the prior art presents indications of several
different approaches t o the problem of
non-Newtonian power consumption in
agitated vessels. However, no useful equipment-design methods have been developed
and no over-all physical understanding of
the problem has as yet been presented.

DESCRIPTION AND CLASSIFICATION OF

NON-NEWTONIAN FLUIDS

In view of the previously reported

( I d , I S ) limitations of the common
rheological classifications of non-Newtonian behavior, the method of approach
suggested in this work is to define the
extent of non-Newtonian behavior by
the property n, as given by the power-law
equation:
7

I@>"

Newtonian fluids are defined as those

materials which in laminar flow exhibit
a linear relationship between the imposed
shear stress and the resulting shear rate;
i.e.. n has a constant value of unity and
R is equal to p/qc where p is termed the
viscosity of the Newtonian fluid. Since
the property n is a measure of the type
of fluid behavior, it has been termed the
$ow-behavior index of a fluid (13).
Similarly, K may be considered to be a
$uid-consistency index.
Non-Newtonian fluids are those materials for which the flow-behavior index
is not equal to unity although it is
frequently a constant, or nearly so, over
wide ranges of shear rates. The fact that
all conis not a true constant-over
ceivable ranges of shear rate-is
frequently irrelevant in view of the fact
that one needs a rheological equation
which correctly portrays only the fluid
behavior over the particular range of
shear rates which happen to be of interest
in a given engineering problem. The rigor
and utility with which a similar method
of approach has been applied to the
analogous problem of flow in pipes ( I d )
suggest its utility in the field of agitation.
As a matter of fact, for flow in pipes the

Page 4

property analogous t o n in Equation (1)

does not need to be a constant in order
for the analysis to be rigorous, but may
be allowed to vary with the shear rate
of the fluid. The problem of agitation of
fluids in vessels has been too complex to
enable a similar theoretical analysis t o
date, and one cannot yet say whether or
not the exponent n must be a constant
to make the approach a rigorous one.
From an engineering viewpoint it is
important to note that no fluids have
yet been found for which n changes
rapidly enough with shear rate to permit
an experimental analysis of this problem
or to reveal any discrepancies of data
attributable to this factor. Severs (25)
has suggested that studies using dilatant
fluids, for which n sometimes does change
rapidly with shear rate, would be
illuminating.
I n view of the mathematical simplicity
of Equation (1) it is remarkable that it
correlates rheological data as well as it
does. For example, the data of reference
3 are correlated a t least as perfectly by
Equation (1) as by the much more complex Eyring-Powell equation used by
Stevens et al. This statement of fact may
be readily verified by simply plotting
both the rheological data and the EyringPowell equation for each fluid (as given
by reference 3 ) on logarithmic coordinates
and comparing the fit obtained with that
obtainable by the simple straight line
which depicts Equation (1) on such
coordinates. The rheological data of the
fluids used in the present work (Figure 5 )
do not fall on a perfect straight line
over the entire range of shear rates for
which data are available, but for all
fluids except the 1.07, CMC the deviation from such a straight line is less than
the scattering of the data points. Even
for this CMC gel the data which are
within the range of average shear rates
used in the agitation studies fall on a
perfect straight line.

DEVELOPMENT OF THE CORRELATION

The flow of a fluid about a mixing

impeller is complex, but the controlling
factors of the drag on an impeller blade
may be examined by comparison with
simpler flow situations. Figure 3 shows
the laminar flow of a fluid around a
sphere. There is no separation of the
flow and viscous dissipation of energy
is the controlling factor. The total force
on the sphere is due to two effects:
(1) the shearing forces on the surface
of the object (of magnitudes indicated
approximately by arrows tangent to the
surface on Figure 3) and (2) the differences in pressure between the front and
the rear of the sphere. These pressure
differences in turn are due to the viscous
dissipation of energy in the fluid stream
and are indicated by the arrows perpendicular to the surface of sphere.

A.1.Ch.E. Journal

For a Newtonian fluid two thirds of the

drag is due to shear-stress forces on the
surface ( I 7 ) , the remainder being due to
differences in pressure caused by the
flow of material in the region of the
surface. For a flat plate of zero thickness
normal to the flow of fluid (Figure 4),
which may be assumed similar to a flatbladed turbine or paddle rotating in a
fluid, all the drag is due to differences in
pressure caused by the flow of the fluid
in the region of the object. as there is no
area along which the shear stress a t the
surface can exert a component causing a
net drag in the direction of the motion
of the fluid.* One may conclude from
this analogy that the study of such an
object moving through the fluid must
include the study of the fluid in the general
region of the object, since this determines
the viscous-energy dissipation and hence
the drag or power requirement.
In non-Newtonian technology, study
of the flow in this over-all region must
obviously include a consideration of the
shearing rates or shearing stresses. This
is a necessary consequence of the fact
that the response (viscosity) of these
fluids to an imposed stress is not a constant but depends on the magnitude of
either the shearing rate or shearing stress.
Another way of arriving a t the same
conclusion is to note that no simple
equation for the relationship between
shear stress and shear rate [such as
Equation (l)]has been found which will
correlate such data over all conceivable
ranges of these variables. Therefore, it is
necessary to know at least approximately
the ranges into which the shear rates about
an impeller may fall and the variables
which determine or control these shear rates.
Figure 5 shows the shear-stress-shearrate relationships for the materials used.
Evaluation of the ratio of shear stress to
shear rate (apparent viscosity) a t a
given point showed for the 1.25% Carbopol solution a variation in apparent
viscosity from 134,000 centipoises a t a
shear rate of 1.5 see.? t o 1,100centipoises
a t 1,000 see.-'. With such a large variation in apparent viscosity one could not
make a one-point measurement of the
apparent viscosity indiscriminately and
expect to obtain a correlation when the
range of shear rates in the viscometer was
different from that of the shear rates in
the system being studied.
The mathematical description of a
non-Newtonian material may be accomplished by the use of an equation [such
as Equation (l)]relating the shear stress
to shear rate or, equally well, by the
*This analogy of flow over a flat plate a s compared with flow past a mixing impeller is not completely exact as the postulated types of flow do not
include the radial components of velocity which are
parallel to the surface of the flat plate or impeller.
Such radial flow is necessary t.0 produce mixing.
This does not invalidate the argument, however,
because the drag due to radial flow does not produce
any forces perpendicular to the surface of the impeller and therefore does not directly affect the
torque felt b y the shaft and hence the power consumption of the system.

March, 1957

ratio of its shear stress to shear rate as a

function of, say, shear rate. If one chooses
the latter, one has a simple comparison
with a Newtonian fluid because in the
limiting case, a Newtonian, the ratio
of shear stress to shear rate is not a
function of shear rate. Of course, the
choice is arbitrary. The latter was chosen
in this case to simplify the visualization
of the behavior of the non-Newtonian
fluid.
I n order to pinpoint the shear-rate
range in a mixing system, one may devise
a means of measuring the apparent
viscosity of the system and then determine its relation to the other variables of
the system. To define apparent viscosity
one may consider two identical sets of
mixing equipment, one of which contains
a Newtonian fluid and the other a nonNewtonian. If these fluids are agitated
in the laminar region, with the same
impeller speed used in each, and one
varies the viscosity of the Newtonian by
diluting it or thickening it so that the
power measured a t each impeller is the
same, then, because all variables are
identical, one may say that the average
viscosities are the same in both pieces of
equipment. Upon measuring the viscosity of the Newtonian fluid one knows
the apparent viscosity of the non-Newtonian existing under the given experimental conditions.
The preceding experiment defines the
apparent viscosity in the system. It is
necessary to determine its dependence on
the variables of the system. I n this paper
it is assumed that the fluid motion in the
general region of the impeller can always
be characterized by an average shear
rate which is linearly related to the rotational speed of the impeller, viz:

metric curve for the fluih (such as

Figure 5) to obtain the corresponding
average shear rate in the system. The
best value for the proportionality constant
k in Equation (2), obtained in this
manner, was 13. Consideration was given
to all fluids and T / D ratios studied in
the laminar region. It may be noted
that only data in the laminar region are
useful for this purpose as power data
outside the laminar region are insensitive
to viscosity.
RESULTS

The non-Newtonian materials tested

were two colloidal suspensions, CMC or
sodium carboxymethylcellulose (Hercules
Powder Company) and Carbopol 934
(Goodrich Chemical Company) and a
suspension of Bttasol clay (Attapulgus
Mineral and Chemical C n p a n y ) . As
shown in Figure 5, two con1 ntrations of
CMC and of Carbopol nd one of
Attasol were used. On 2 logarithmic
diagram such as Figure
the linear
shear-stress-shear-rate re1 ,ionship of
Newtonian fluids appears I a straight
45-deg. line (n = 1.00). 'I le deviation
from Newtonian behavior, ,herefore, is
indicated by the divergenci of the slope
from the value of unity. lttasol was
accordingly the most nc -Newtonian
(n = 0.24) of the fluids te ed. Because
all these materials were oj the pseudoplastic type the slopes of t h curves were
all less than unity. The 4ttasol-clay
suspensions of the concentra;ion studied
are frequently termed Bingham plastic
by the engineering literature and the

other materials are commonly, and

correctly, termed pseudoplastic.
The apparent viscosity of a given
material at a particular shear rate wm
determined by dividing the shear stress
by the particular shear rate. An important
point to note here is that the shear rates
in Figure 5 were not calculated by presuming a shear-stress-shear-rate relationship, nor were they calculated by
assuming that the shear rates were those
which a Newtonian fluid would exhibit
in the same viscometer. Both these
erroneous attacks, common in the literature, are not necessary at the present
state of the art. (Cf. Appendix C.) The
variation of the viscometric data with
temperature was shown to be negligible
for the small variations in room temperature encountered and were not always
recorded.
Figure 7 shows the power correlation
for all the materials tested. Of over 130
data points taken, all but ten lie within
15% of the curve drawn through the data.
The ranges of data were as follow:
Impeller diameters 2 to 8 in.
Tank diameters 6 to 22 in.
T/D
1.3 to 3.7 (laminar
region)
2.0 to 5.5 (transition
region)
Power input
0.5 to 176 hp./1,000
gal.
95 to 1,190 rev./min.
Speed
Apparent viscosity 7 to 180 poises
Reynolds number 2.0 to 270
Both baffled ( B = 0.1T) and unbaffled
systems were studied. The equipment
used is described in Appendix R ; the

TABLE4. MIXING-RATE
DATA

'

If this line of reasoning is followed, the

only necessary fluid properties are the
apparent viscosity and density, although
the latter would not be expected to be a
significant variable in the laminar region.
Until means are developed both for
measuring and averaging shear rates the
evaluation of k in Equation (2) must be
done indirectly.
In Figure 6 the power number for a
2.5% CMC solution is plotted as a
function of impeller speed in the system
previously described. Since the comparable Newtonian data for this system are
known (22), the viscosity which a Newtonian fluid would have under identical
conditions of speed and power consumption in the same equipment can be
obtained merely- by- referring to the
Newtonian power-number curve. According to the preceding arguments, this
must be identical to the apparent viscosity of the non-Newtonian. When the
apparent viscosity is thus Obtained'
reference may be made to the ViSCO-

Vol. 3, No. 1

Minimum

Minimum

h.p./1,000 gal.*

NRe+

TID

40.2
21.6
8.66

87.4
122.
94.0

2.95
2.95
2.95

50.0
74.0
16.8

49.0
42.5
75.8

1.97
2.00
1.97

32.4

1.48

27.7

9.8

1.33

Movement of fluid surface

3A
1.0% CMC
8
1.0% CMC

53.1
17.3

268.
81.7

2.95
3.00

15
5

Attasol
1.0% CMC

35.8
23.0

113.
94.0

1.97
2.00

17

Attasol

35.0

98.0

1.48

Median of all:

25-35

50-100

Run
Fluid
Movement at wall of tank
2
2.0% CMC
14
Attasol
3A
1.0% CMC

15

Carbopol
Carbopol
Attasol

17

Attasol

13

Carbopol

12

5.09

*Minimum values at which fluid movement could first be observed visually.

tRun taken in conjunction with run 10. See footnote ** in Tab!e 3 [footnote on page 61.

A.1.Ch.E. Journal

Page 5

detailed dimensions of the equipment

and the experimental data are tabulated
in Tables 1 to 4.*
DISCUSSION

OF RESULTS

Laminar Region

With the exception of 1.0% CMG and

the Attasol suspension, Table 3 and
Figure 7 show that all fluids were studied
in the laminar region in a t least one run.
The agreement between the data and
the conventional Newtonian correlating
curve is good.
Transition Region

In general, the gradual development of

turbulence and departure from laminar
behavior may be stated to be due to the
formation of eddies. Consideration of the
flow properties of the pseudoplastic

-\

- I-

Fig. 3. Flow of a Newtonian fluid about a

sphere.

fluids used in this work shows that their

apparent viscosity must increase with
increasing distance from the impeller, in
view of the lower velocities, hence shear
rates, of the fluids as they leave the vicinity of the impeller. A small volume of fluid
leaving the high-shear-rate region near
the impeller therefore encounters progressively more viscous material. This, in
turn, would cause a depressive effect on
the propagation of eddies. The last factor
would tend to decrease the rapidity with
which the power-number-Reynolds-number curve begins to diverge from the
45-deg. line of the laminar region as the
fluid becomes more non-Newtonian in the
direction of greater pseudoplasticity (n
approaching zero). The net result of this
behavior should be an extension of the
laminar region to Reynolds numbers
above 10 or to power numbers below 7.1,
the point at which Newtonian fluids
enter the transition region. Figure 7
shows that such a retardation was
indeed observed. It must be emphasized
that this extension of the laminar region
is not due to the particular form of the
Reynolds number chosen in this work.
The power number is used in its con~~

Tabular material has been deposited as document

5119 with the American Documentation Institute,
Photoduplication Service, Library of Congress,
Washington 25, D. C.. and may be obtained for
$1.25 for photoprints or 35-mm. microfilm.

Page 6

ventional orm, and since Newtonian

systems leave the laminar region a t a
power number of 7.1 no conceivable
change in Reynolds number could shift
the curve to coincide with the usual
Newtonian behavior, shown as a dashed
line in Figure 7. Furthermore, the data
indicate that baffling had no effect on
power consumption under the conditions
evaluated; hence this extension of the
laminar region was found to be common
to baffled as well as to unbaffled systems.
The more gradual transition from laminar
to turbulent flow for pseudoplastic
fluids has also been shown to occur in
flow through pipes ( l a ) , and Brown and
Petsiavas ( 2 ) noted the same behavior
in independent mixing studies.

and the other two papers (9,2.9) either

did not measure the rheological properties
of the fluid or worked with only slightly
non-Newtonian systems. One may, however, cite the work of Magnusson (9)to
extend the generality of the unique relationship between average shear rate and
rotational speed of the impeller, as
follows.
Equation (2) implies that scale-up of
a model should be carried out at constant
speed for a given non-Newtonian fluid,
so that the flow properties of the fluid
may be identical in the prototype and
the model. Figure 7 proves this for the
flat-bladed turbines of the present work,
and Magnusson's data confirm this for
two X-shaped paddles with diameters
of 2.8 and 6.2 in. It may be concluded
Validity of Assumptions
that since the same type of relationship
The correlation of Figure 7 confirms between shear rate a,nd impeller speed
the basic assumption made [Equation
(2)] regarding the relationship between
shear rate and impeller speed, at least
over the range of variables investigated
and in both the laminar and transition
regions. The conclusion that the average
shear rate, hence apparent viscosity, of
the non-Newtonian fluid depends only
on the rotational speed of the impeller
may not, a t first glance, appear to be
obvious. However, supporting evidence
mag also be obtained from Equation (10)
for-the shear rate a t the bob of a vis- Fig. 4. Flow of a Newtonian fluid about a
cometer in an infinite fluid. In this case
flat plate.
the rate of shear also depends only on
the rotational speed, and not on dimensions of the bob. In the laminar region holds for two greatly different types of
the analogy between a viscometer bob impellers, as well as for various different
and the impeller of a mixer is rather fluids, it appears to be of general utility.
close, since as a matter of fact, many No data are available, however, to concommercial viscometers employ bobs firm the constancy of k at a value of 13.0
which are geometrically similar to mixing for systems other than those discussed
impellers. In both the laminar and tran- here. In particular, k might be expected
sition regions investigated in this work to vary with the flow-behavior index n
the impellers generally behaved as though of the fluid, which was not varied widely
in an infinite fluid since changing the in the present studies. [Cf. Equation
distance between the tip of the impeller (lo).] It should be noted in passing that
and the wall of the tank (i.e., changing scale-up procedures cannot be carried
the T / D ratio) had no effect on power out in the usual sense of the term, howconsumption. As will be discussed later, ever, because the Reynolds numbers will
this is in agreement with extensive work not be identical in the prototype and
on Newtonian fluids. However, as the model when N , p, and p, are all held
T / D ratio decreases toward unity, i.e., constant.
as the diameter of the impeller approaches
that of the tank, a point must eventually Ranges of Variables
The power range encompassed in this
be reached where the shear rate, hence
power consumption, is a t least partly study was higher than the range which
dependent on either the T / D ratio is used industrially for Newtonian fluids
(clearance between impeller and tank (16), because pseudoplastic non-Newwall) or the peripheral speed of the tonian fluids are inherently more difficult
impeller or both. It appears plausible to agitate. Since the exact power requirethat a t T / D ratios close to unity the ments for different levels of mixing rate
shear rate a t the impeller may take a are not yet known for these materials, the
mathematical form similar to that for a power input was varied over an extremely
wide range to ensure applicability of the
bob-and-cup viscometer, Equation (6).
Recorrelation of data in previous results 60 industrial conditions.
While the range of Reynolds numbers
papers on non-Newtonian-fluid agitation
might appear desirable to extend the covered was only l3O-fold, it is the range
results of this work; however, sufficiently in which viscous, highly non-Newtonian
detailed tabulations of experimental data fluids are most likely to be agitated. With
were not available from one source ( 2 ) less viscous non-Newtonians the range

A.1.Ch.E. Journal

March, 1957

must be a function of (d2N2-np)/gcK

and n. The function of n h a n n o t be
evaluated for mixing work at present but
the assumption that it is the same as in
the pipe-flow case, i.e. power number is a
Effect of Geometric Variables
function of [(D2N2-np)/gcK]
[8(n/6n+ 2 ) n ]
The position of the impeller did not or ( D W - * p ) / y , yields a Reynolds number
critically affect power consumption when which is numerically proportional to the
vortices were not formed. Vortices could (D2Np)/pa reported here. Further, more
be formed, however, if the impeller were extensive data are needed to determine
placed closer to the surface than specified whether use of the generalized Reynolds
number in mixing work is of widespread
in Appendix B.
No consistent effects due to baffling or generality or merely a coincidence in the
T / D ratios greater than 2 were found, case of the present data.
which is in agreement with the work of
Rushton et al. (22).Hirsekorn and Miller Quality of Mixing
The type of flow or the quality of
(6) report that in the laminar region
T / D can be made much smaller than the mixing is, of course, the end result of the
value of 2 reported by Rushton. Runs choice of a given agitator. However, as
13 and 17, with T / D ratios of 1.33 and mentioned before, the logical procedure
1.45 respectively, indicate that these for studying agitation is first to determine

may be extended to higher Reynolds

numbers, but usually the same highly
non-Newtonian character is not observed
in such less viscous systems.

Although decreasing the T / D ratio to

nearly unity would almost certainly
reduce the Reynolds numbers, and hence
the power required for complete movement of the fluid, it appears questionable
whether this power reduction would be
great enough to bring the power consumption at a given mixing rate down to
the level of Newtonian fluids. Most of
the data of Table 4 do, however, show an
appreciable decrease in power required
to mix a given fluid completely (as
defined by movement at the tank wall)
as T / D is decreased. Such complete
fluid turnover does not, on the other
hand, imply higher shear rates, as these
are probably affected adversely by use
of larger impellers (lower T / D ratios)
and lower rotational speeds, as shown by
Equation ( 2 ) . Thus it is more difficult
to obtain simultaneously both high rates

100

lL

v)

10

[L

z 5
(L

v)

05

100

1000

10,000

Fig. 5. Flow curves of fluids used.

conclusions are valid also for nonNewtonian materials in both the laminar
and transition regions. However, the
number of data points available is too
small to support such a conclusion
firmly.

the amount of power dissipation and

then to turn to the more complex problem
of mixing patterns and rates. The
measures of the rates of mixing studied
to date (Table 4) are, therefore, qualitative. Nevertheless, several important
conclusions may be drawn at this time
Similarity to Pipe Flow
concerning the comparative rates of
An interesting similarity appears to mixing of Newtonian and non-Newtonian
exist between the means of correlating fluids at a given power input per unit
pipe-flow data and mixing-power-con- volume of fluid.
Hirsekorn and Miller (5) found that
sumption data which suggests an alternate means of correlation for the latter. particles could be suspended in viscous
For a non-Newtonian fluid which obeys Newtonian fluids during agitation within
Equation (1) it may be readily shown the laminar region (NRabelow 10). The
that the friction factor for pipe flow is a maximum power input required to
function of (DnVz72-.)/gcK.8(n/6n 2)., suspend the particles was 6.0 hp./1,000
which is a special form of the generalized gal. of fluid. While the data of Table 4
number (DVz-p)/y proposed for pipe- are very irregular in the sense that
flow work ( I d ) . One will notice that these Reynolds numbers and power requiregroups both degenerate to the usual ments for movement of all the fluid vary
Reynolds number for Newtonian fluids widely, in only one case out of thirteen
( K = p/g, and n = 1.00). Application was reasonably complete fluid movement
of dimensional analysis to the mixing attained in the laminar region and, again
problem (with the only important length in only one case, was any fluid movement
dimension assumed to be the impeller a t the tank wall noticeable a t power
diameter) shows that the power number inputs as low as 6.0 hp./l,OOO gal.

Vol. 3, No. 1

100
N

SHEAR RATE, SEC.-

A.1.Ch.E. Journal

Fig. 6. Power number vs. impeller speed

for 2.5% CMC.

of fluid turnover and high shear rates in

these non-Newtonian systems without
high power requirements.
Until really definitive data become
available, adoption of the median values
of power input and Reynolds numbers
given in Table 4 is suggested as the
criterion below which mixing will frequently be ineffective in these nonNewtonian systems. Some confirmation
of these high power-input levels may be
obtained from.industria1 practice: in one
installation the mixing of non-Newtonian
fluids is in the range of 5 to 50 hp./1,000
gal. in full-scale equipment, while pilot
plant work (indicative of future operations) ranges from 50 to 160 hp./1,000
gal. This industrial experience also
indicates that to date no enormously
improved agitator or system designs
which might be particularly suitable for
the non-Newtonian fluids described here
have been developed. Fundamentally,
this problem arises because the apparent
viscosity of a pseudoplastic fluid increases with distance from the impeller;
hence the fluib tends to set up, or
Page 7

as well as non-Newtonian fluids requires

quantitative study.
4. Extension of the present work to a
greater variety of non-Newtonian fluids,
including dilatant materials, should be of
considerable interest to prove conclusively
whether or not the value of k [Equation (2)]
depends on the flow behavior index of the
fluid and whether the generalized Reynolds
number is truly applicable to correlation of
power-consumption data for all fluids.

100
80
60
40

ACKNOWLEDGMENl

8 10

20

40

w eo

IM)

eoo

44)

DN~
-

r.

Fig. 7. Power-number-Reynolds-number curve for non-Newtonian fluids; all points in

the crowded regions not shown.

remain motionless, under conditions

where a Kewtonian fluid is mixed relatively completely. Therefore the problem
is somewhat alleviated by use of multiple
impellers inside a single tank and by use
of very low T / D ratios in the industrial
examples cited.

2. From viscometric data (shear stress vs.

The authors would like to thank J. Y.

Oldshue and the Mixing Equipment Company for their assistance in designing
equipment and J. H. Rushton and J. Y.
Oldshue for reviewing the manuscript. The
Attapulgus Clay and Mineral Company
and Continental Diamond Fibre Company
are to be thanked.for donating test materials
and equipment, respectively. This work
was sponsored by the Office of Ordnance
Research, U. S. Army.
NOTATION

shear rate) for the fluid in question, la

is
NoTn:-since
the final correlation is
calculated at the above-average shear rate.
3. The Reynolds number (DzNp)/pais then based on dimensionless groups, any concalculated and the corresponding power sistent set of units may be used. The units
given in the following table refer to those
number read from Figure 7.
used by the authors in the data tables unless
As the procedure is empirical, extrap- specific units are given.
olation of the variables beyond the
Design Procedure
ranges covered cannot be recommended. B = width of baffles, ft.
C = height of impeller off bottom of
Within the laminar region (NRe < 20
tank, ft.
when n ranges between about 0.25 and CONCLUSIONS AND SUMMARY
D
=
impeller
diameter or bob diam0.45, and N R e < 10 when n = 1.00)
1. The assumption that average fluid
eter, ft.
conditions are sufficiently well defined for shear rates are related only to impeller
= gravitational acceleration, ft./
the mixing equation, which is somewhat speed has led to an understanding and g
sec.2
analogous t o Poiseuilles law for friction correlation of the power requirements for
gc
=
conversion
factor, (lb. mass)(ft.)/
in a round tube, to be written as sug- agitation of non-Newtonian fluids. The
(lb. force)(sec.z)
gested by Rushton and Oldshue (2f). quantitative relationships, which are ap= height of viscometer bob, ft.
This equation is the same for Newtonian plicable to both Newtonian and non- h
K = fluid property in Equation (l),
and non-Newtonian fluids:
Newtonian fluids over the ranges of
(lb. force) (sec.m/sq. ft.)
N, = 71/NR,
(3) variables investigated, represent a simple k = proportionality constant, dimengeneralization of the well-known Tesults
sionless
Substitution of the definitions of the from Newtonian systems.
L = length of pipe or capillary, ft.
Reynolds numbers used in this work gives
2. The laminar region may extend N = rotational speed, rev./sec.
to higher Reynolds numbers in pseudo- N , = power number, dimensionless, Pgc
71cl,
DzNz
P=
(4) plastic fluids than in Newtonian systems.
/D5N3p
9.
3. Preliminary qualitative observa- N R e = Reynolds number, dimensionless,
It has been suggested that these tions indicate that more power is required
taken as ( D z N p ) / p a
equations are particularly useful for for the rapid mixing of highly non- n
= flow-behavior index, dimensiondesign purposes. The constant given was Newtonian systems than for Newtonian
less
developed for hlixco turbines with six fluids.
flat blades and is dependent on the
n = d In (ApD/4L)
particular type of impeller used. As RECOMMENDATIONS FOR FURTHER WORK
d In (8&/7rD3)
would be expected from the analogous
1. Although data were taken on 8-in.
problem of flow in tubes (f2), highly impellers, the limitations of the present ,tr
= -d In (2tl7rD;h)
non-Newtonian fluids with a flow-be- equipment did not allow high rotational
d In ( 4 ~ N / l l/sz)
havior index n near zero show a smaller speeds for these larger systems. Therefore,
change of power with impeller speed in order to approach plant-scale conditions P = power, (ft.)(lb. force)/sec.
than do Newtonian fluids (n = 1.00), more closely one must extend the data to p
= pressure, lb. force/sq. ft.
larger and more powerful mixing systems. Q = flow rate, cu. ft./sec.
since pa decreases as N increases.
2. In some commercial installations the T
Under any flow Conditions the recom= radius, ft.
mended procedure for estimating power T / D ratio may be smaller than the values S = scale reading, Ib. force
consumption once the type, size, and studied in the present work. Limited data s
= D,/Di
indicate that the power consumption may
speed of the impeller have all been not be significantly affected by such changes, T = tank diameter, ft.
fixed may be reviewed as follows:
= torque, ft.(lb. force)
but this influence of small values of T / D t
= point velocity, ft./sec.
1. Knowing N , one may evaluate kN should be investigated further for both u
V = volumetric average velocity, ft./
[Equation (2)] to determine the average Newtonian and non-Newtonian fluids.
3. The quality of mixing of Newtonian
see.
shear rate in the system.
Page 8

A.1.Ch.E. Journal

March, 1957

A
pa

p
7

= generalized viscosity coefficient,

Y = (qcK)/g.(6y
4- 2/n)., 1b.
mass/(ft.) (see.%-n )
= difference of
= apparent viscosity, lb. mass/(ft.)
(see.)
= density, Ib. mass/cu. ft.
= shear stress, Ib. force/sq. ft.

Subscripts

i
0

= bob
= cup

wall
wall

LITERATURE CITED

1. Alves, G. E., D. F. Boucher, and R. L.

Pigford, Chem. Eng. Progr., 48, 385
(1952).
2. Brown, G. A., and D. N. Petsiavas,
paper presented a t New York A.1.Ch.E.
Meeting (December, 1954).
3. Christiansen, E. B., N. W. Ryan, and
W. E. Stevens, A.1.Ch.E. Journal, 1,
544 (1955).
4. Hedstrom, B. 0. A,, Ind. Eng. Chem.,
44, 651 (1952).
5. Hirsekorn, F. S., and S. A. Miller,
Chem. Eng. Progr., 49, 459 (1953).
6. Hunsaker, J. C., and B. G. Rightmire,
Engineering Applications of Fluid
Mechanics, McGraw-Hill Book Company, Inc., New York (1947).
7. Krieger, I. M., and S. H. Maron, J .
A p p l . Phys., 25, 72 (1954).
8. Mack, D. E., and V. W. Uhl, Chem.
Eng., 54, 119 (1947).
9. Magnusson, Karl, Iva. (Sweden),23, 56
(1952).
10. Matthews, T. A., 11, private communication (April 29, 1954).
11. Metzner, A . B., Chenz. Eng. Progr., 50,
27 (1954).
and J. C. Reed, A.Z.Ch.E.
12. --,
Journal, 1, 434 (1955).
13. Metzner, A. B., in Advances in
Chemical Engineering, Vol. I, Academic Press, Inc., New York (1956).
14. Mooney, Melvin, J. Rheol., 2, 210
(1931).
15. OConnell, F. P., and D. E. Mack,
Chem. Eng. Progr., 46, 358 (1950).
16. Oldshue, J. Y., grivate communication
(1954).
17. Otto. R. E.. Ph.D. thesis. Univ. Delaware: Sew&k (1957).

18. Reed, J. C., M.Ch.E. thesis, Univ.

Dekawdre, Newark (1954).
19. Reiner, M., Deformation and Flow,
H. K. Lewis and Compmy, London
(1949).
20. Rushton, J. H., Ind. Eng. Chem., 47,
582 (1955).
21. -- and J. Y. Oldshue, Chem Eng.
Progr., 49, 161 and 267 (1953).
22. Rushton, J. H., E. W. Costich, and
H. J. Everett, Chem. Eng. Progr., 46,
395 and 467 (1950).
23. Schultz-Grunow, F., Chem. Zng. Tech.,
26, 18 (1954).
24. Severs, E. T., and J. M. Austin, Ind.
Eng. Chem., 46, 2369 (1954).
, private communication (1956).
25. I

rheologists. These definitions are based on

the relationship between the shearing stress
imposed on a fluid, 7, and the resulting
shear rate, duldr, as measured in various
kinds of viscometers. Newtunian fluids
show the familiar straight-line relationship,
the slope of the line being defined as the
viscosity of the fluid. A Bingham-plastic
fluid is defined as one which also shows a
linear relationJhip between shear stress and
shear rate, but the relationship does not pass
through the origin. Pseudoplastic and dilatant fluids do not show a linear relationship.
Figure 9 shows the equivalent logarithmic
plot.
For the Newtonian fluid the ratio of
shear stress divided by shear rate is a
constant and defined as the viscosity of
the fluid. For the non-Newtonian, the
equivalent, or apparent, viscosity is not
a constant. Instead the ratio of shear stress
divided by shear rate changes with shear
rate. For Bingham-plastic and pseudoplastic
materials this ratio decreases with increasing values of shear rate; for dilatant
materials it increases.
There are two other kinds of nonNewtonian behavior, termed thixotropy and
rheopezy. The apparent viscosities of
thixotropic and rheopectic fluids depend
on the time of shear as well as on rate of
shear. These two kinds of behavior have
been too complex to study and, since they
are of less frequent industrial importance,
will not be discussed here. Industrially,
pseudoplastic behavior is probably more
important than the other types of nonNewtonian behavior combined.
All the foregoing definitions are restricted
t o materials which do not exhibit elastic recovery or viscoelasticity. That is to say,
once they have been sheared there is no
tendency for the fluid to return to its original
shape or configuration. The necessity of this
assumption may prove to be a more serious
limitation of the present work than the
assumed absence of thixotropy and rheopexy but cannot be dealt with until the
engineering problems of design for pseudoplastic, Bingham-plastic, and dilatant behavior hsve been well-developed.

These classincations are discussed in

more detail elsewhere (I, 4, 11, 13, 19, for
example). Many engineering publications
have been concerned with fluids which were
believed to be of the Bingham-plastic type.
At the beginning of the present experimental program extensive determinations
of the shear-stress-shear-rate relationships
of fluids which have been claimed to be
Bingham plastics led to the conclusion that
true Bingham plltstics probably exist only
very rarely, if at all. Except for a few
specially prepared materials this type of
behavior broke down over shear-rate ranges
greater than about 1:lOO.
B. Experimental Apparatus and Procedure

The mixing equipment used in this work

is shown in Figures 10 and 11. It consisted
of a x-hp. variable-speed motor, four
Mixco Standard flat-bladed turbines (Mixing Equipment Go.), and four cylindrical
tanks. To measure torque accurately, the
motor was mounted on a large ball-bearing
ring fixed in a cast-aluminum plate; the
motor rotated freely with very little friction.
The reaction torque developed by the
motor in driving the turbine was taken
from the motor by a torque ring attached
to the motor hesd and transferred by a
small, essentially frictionless pulley to a
dynamometer scale. The turbine shaft was
fitted inside a special hollow shaft provided
on the motor by the manufacturer (Mixco)
so that the shaft height was variable.
Turbine speeds were measured by an
electric tachometer geared to the motor
shaft.
Four cylindrical flat-bottomed tanks 6,
8.2, 11.6, and 22 in. in diameter were used
in this work. The smallest was a beaker,
the next two were Pyrex tanks, and the final
one was a 55-gal. drum. All tanks except
the smallest one were fitted with four
removable baffles with a width of one tenth
of the tznk diameter.
The impellers which were used had
diameters of 2, 4, 6, and 8 in. The ratios of
the width and length of the impeller blades
relative to their diameter were 1:5 and 1:4
respectivcly. With the smallest flat-bladed
turbine the torque readings were too low

BINGHAM
PLASTIC

1001

v)

/
A
PSEUDOPLASTIC

K
I

APPENDIX
A. Definitions of Non-Newtonian Behavior

Figure 5 illustrates on arithmetic coordinates the classical definitions used by

Vol. 3, No. 1

SHEAR RATE

$!-

Fig. 8. Fluid characteristics (arithmetic

scale).

A.1.Ch.E. Journal

10

SHEAR RATE

Fig. 9. Fluid characteristics (logarithmic

scale).

Page 9

for the dynemometer scale; hence a torque

table was improvised from a dead-weight
tester connected through a system of
pulleys to a fmaller spring scale.
I n order not to vury a large number of
geometric varizbles a t on-e, the fluid
level in the tanks was maintained a t one

tank diameter. The impeller was placed a t

one impeller diameter off the bottom of
the tank unless this placement brought the
impeller withln one impeller diameter of
the fluid surface. In this case the impeller
was centered halfway between the liquid
surface and the tank bottom.
Rheological properties were measured
both on a Stormer rotational viscometer
and with a capillary-tube viscometer
(Figure 1I). The capillary-tube viscometer
was an instrument used for the extrusion of
plastics (24).
The fluid to be tested was placed in a
stainless steel pressure chamber and extruded under pressure from a capillary tube
mounted in the bottom of the chamber.
The flow rate of the fluid was measured by
use of a stop watch and a weighing balance.
The extrusion pressure was supplied by a
nitrogen cylinder and was measured on
laboratory test gauges. During the experimental runs the pressure chamber was
surrounded by water to control the temperature.
The Stormer viscometer was modified in
that a smooth-walled cylinder was used
for the cup. End effects on the bob were
corrected by calculation of the equivalent
bob height from data taken with National
Bureau of Standards calibrated oils. This
equivalent bob height was about 25%
greater than the actual height. The fact
that the data (Figure 5 and Table 2) from
both instruments coincide within experimental error supports the adequacy of this
procedure for accounting for end effects in
the rotational viscometer as well as for
the absence of any similar problems with
the capillary-tube viscometer.
Detailed equipment dimensions are given
in Table 1.
C. Viscometry

Fig. 10. Mixing equipment.

General equations (which do not require

the presumption of a shear-stress-shear-rate
relationship) have been reported in the
literature; these were used exclusively t o
interpret the viscometric measurements.

For the rotational viscometer, Krieger

and Maron ( 7 ) have succeeded in writing
an infinite series which converges rapidly
for cup-to-bob diameter ratios of less than
1.2:l. Their equations are
Shear stress a t the bob:
7. =

2
P

t
D,h

Shear rate a t the bob:

_-du
dr -

1 47rN
- 1,s [l

+ h($

- 1)

n r f is evaluated a t the shear stress calculated by means of Equation (5). The

instrument constants h.1 and k z are

k,

c2s2-2(1 + $Ins);

s2 - 1
k 2 -----I
ns

(7)

6s2

I n any one run n is the slope of a

logarithmic plot of the torqne plotted
against rotational speed, as all other factors
in n are constant.
Rabinowitschs solution of the motion of
fluids flowing through a pipe or capillary
tube gives upon rearrangement (18)
7 = -APD
- -

(8)

4L

--[-I

du
32Q 3n
dr - 7rD3

+1

(9)

Again n must be evaluated a t the corresponding shear stress given in Equation (8).
For a given capillary tube n is found from
a logarithmic plot of pressure drop vs. flow
rate.
Another general equation given
by
Krieger and Maron ( 7 ) is of use in viscometric equipment approximating a long
cylinder rotating in an infinite fluid:

Brookfield viscometers (Brookfield Engineering Laboratories) may fall into this

category, depending on the shapes of the
bobs user The shear stress is given by
Equation ( 5 ) .One cannot use the Brookfield
conversion factors to determine the apparent viscosity directly because the formulas
supplied with the instrument are for
Newtonian fluids (n = 1.00). However,
since a Brookfield has several speeds n
can be obtained from a logarithmic plot of
scale reading vs. rotational speed, and the
shear rates calculated by means of Equation
(10).
For a power-function non-Newtonian
[Equation (I)]:

n
Fig. 11. Impellers and viscoinetric equipment.

Page 10

A.1.Ch.E. Journal

riff.

(11)

Presented at A.I.Ch.E. Detroit meeting

March, 1957