Power Consumption in Shaking Flasks

on Rotary Shaking Machines:

II. Nondimensional Description of
Specific Power Consumption and
Flow Regimes in Unbaffled Flasks at
Elevated Liquid Viscosity

Jochen Büchs,1 Ulrike Maier,1 Claudia Milbradt,1 Bernd Zoels2

1
Chair of Biochemical Engineering, Sammelbau Biologie, Aachen
University of Technology, D-52074 Aachen, Germany, telephone:
+49-241-805546; fax: +49-241-8888-265; e-mail:
buechs@biovt.rwth-aachen.de
2
ZET/ZT-A15, BASF AG, Ludwigshafen, Germany
Received 6 September 1999; accepted 14 February 2000

Abstract: This article is the second part of a series pre-
senting and modeling the hydrodynamics and specific
power consumption in shaking flasks on rotary (orbital)
shaking machines. In part I, a new method was intro-
duced that enables the accurate determination of the
specific power consumption in shaking flasks. The
method was first applied to investigate unbaffled flasks
with a nominal volume of ⱕ1 L at low viscosity. In part II,
the results for the specific power consumption of un-
baffled shaking flasks at elevated viscosities are investi-
gated after varying shaking frequency, flask size, filling
volume, and shaking diameter. The theory introduced in
part I is extended to liquids of elevated viscosities using
nondimensional equations. With these results, the spe-
cific power consumption in unbaffled shaking flasks can
now be fully described. For the first time, the phenom-
enon of the liquid being “out of phase” is observed and
described. This occurs at certain operating conditions
and is characterized by an increasing amount of liquid
not following the movement of the shaking table, thus
reducing the specific power consumption. This, of
course, has much relevance for practical work with mi-
crobial cultures. The phenomenon of being “out-of-
phase” is described in the form of a newly defined non-
dimensional phase number (Ph) in analogy to a partially
filled, rotating horizontal drum. The Ph can be used to
determine reasonable operating conditions for shaking
flask experiments when using viscous media, avoiding
unfavorable “out-of-phase” operation. © 2000 John Wiley
& Sons, Inc. Biotechnol Bioeng 68: 594–601, 2000.
Keywords: power consumption; shaking flasks, model-
ing; viscous media; operating conditions; filamentous
microorganisms
INTRODUCTION
Due to their simplicity shaking flasks are a major tool for
parallel studies in microbiological process development. In
Correspondence to: J. Büchs
larger companies working in the field of biotechnology, up
to several hundred-thousand experiments are carried out us-
ing shaking flasks annually. These bioreactors are well
suited to more or less routine tasks like medium evaluation
and strain selection, involving the need for many replicate
experiments. Nowadays, the design of agitated tank biore-
actors with a standard geometry no longer presents a big
problem. In contrast, shaking flasks, in which the majority
of biotechnological development takes place, are only in-
sufficiently described.
One of the important parameters in fermentation of aero-
bic microorganisms in shaking flasks and agitated tank bio-
reactors is the specific power consumption (per unit vol-
ume). In part I of this series (Büchs et al., 2000) a new
method was introduced that enables the accurate determi-
nation of the specific power consumption in a shaking flask
on rotary (orbital) shaking machines. The aim of part II is to
present algebraic equations based on simple, physically
sound, nondimensional models with which the specific
power consumption and reasonable operating conditions for
shaking flasks may be calculated for different flask sizes at
elevated viscosities for newtonian fluids (up to ␩ ⳱ 200
mPa ⭈ s). An investigation of the influence of baffles on
“out-of-phase” operating conditions will be presented in a
subsequent article of this series.
MODEL DEVELOPMENT
It has been shown that the specific power consumption in
shaking flasks at low viscosity may generally be described
by the modified power number (Ne⬘), which is a function of
the flask Reynolds number (Re), according to (Büchs et al.,
2000):
P
Ne⬘ = = C ⭈ f 共Re兲 (1)
␳ ⭈ n ⭈ d 4 ⭈ VL1 Ⲑ 3
3

© 2000 John Wiley & Sons, Inc.

 ␳ ⭈ n ⭈ d2
Re = (2)
␩
This form of equation can generally be used for operating
conditions at which the bulk of the liquid within the flask
circulates “in-phase” with the shaking table.
In contrast, at certain operating conditions (with the vis-
cosity being the main influencing factor), the bulk of the
liquid remains at the base of the flask and shows only very
little relative movement. These operating conditions are
termed as “out-of-phase” throughout this study.
To systematically determine these unsuitable out-of-
phase operating conditions an analogy to a partially filled,
rotating horizontal drum is used. The ratio of forces acting
in the plane perpendicular to the shaking flask axis is similar
to that present in a horizontal drum. Yet, the force present in
the shaking flask in the radial direction is the centrifugal
force induced by the translatory motion of the shaking table.
In the case of the shaking flask the gravitational force acts
in the axial direction. The flow regimes present in a hori-
zontal drum have been investigated in detail by Deiber and
Cerro (1976) as well as Semena and Khmelëv (1982). These
investigators assumed that the drum is infinitely long, so
that the side faces have no influence on the flow behavior.
They showed that the flow regime is described by a radial
Froude (Frr) number (radial/radial forces) and a liquid film
Reynolds (Ref) number:
共2 ⭈ ␲ ⭈ n兲 ⭈ d2
Frr = (3)
2⭈g
␳ ⭈ 共2 ⭈ ␲ ⭈ n兲 ⭈ f 2
Ref = (4)
␩
It should be noted that the liquid film Reynolds number
(Ref) is not identical to the flask Reynolds number (Re)
introduced in Eq. (2). As illustrated in Figure 1, the symbol
f represents the liquid film thickness for the hypothetical
case of a uniform distribution of the liquid inside of a hori-
zontal drum with the length h, according to:
␲
VL = ⭈ 共d 2 − d 2i 兲 ⭈ h (5)
4
Hence, the liquid film thickness ( f ) is given by:
共d − di兲
f= (6)
2
Two distinct flow regimes exist as shown in Figure 1 (sepa-
rated by boundary lines I and II or III). For high radial
Froude (Frr) or low liquid film Reynolds (Ref) numbers
(high viscosity, low filling volume) the fluid is almost com-
pletely distributed on the whole drum wall (regime A). For
low radial Froude or high liquid film Reynolds numbers
(low viscosity, large filling volume) only a very thin liquid
film is distributed on the drum wall, with the majority of the
liquid remaining in a groove at the base of the drum (regime
B). Between boundary lines II and III Semena and Khmelëv
BÜCHS ET AL.: POWER CONSUMPTION IN SHAKING FLASKS ON ROTARY SHAKING MACHINES. II
(1982) observed a bistable region. In a recent and more
detailed examination of horizontal drums of finite length it
was shown that the side faces influence the flow signifi-
cantly. This results in highly complicated hydrodynamical
interactions and many different flow regimes (Thoroddsen
and Mahadevan, 1997). It can be expected that the situation
in shaking flasks is just as complicated. Nevertheless, the
simplifying approaches of Deiber and Cerro (1976) and Se-
mena and Khmelëv (1982) are quite helpful and may serve
as a sufficient theoretical background for our ultimate goal
in the present work.
To transfer these conditions to shaking flasks and to un-
derstand the accelerations present, it is helpful to divide the
overall liquid motion into two partial movements: the rotary
movement of the liquid relative to the flask, and the trans-
latory movement of the complete shaking system (including
the individual flasks) relative to the ground. The liquid ac-
celeration relative to the flask for “in-phase” conditions is
simplified to being equal to (2 ⭈ ␲ ⭈ n)2 ⭈ d/2 (assuming d ⳱
constant). The radial acceleration of the translatory moving
system relative to the ground is equal to (2 ⭈ ␲ ⭈ n)2 ⭈ do/2.
Hence, the radial Froude (Frr) number (radial/radial forces)
in shaking flasks is equal to d/do. As this expression consists
of geometric dimensions only, it is defined as the geometry
ratio (GR):
d
GR = (7)
do
Our modeling approach based on the GR is reasonable only
for elevated shaking frequencies, at which nearly all liquid
is rotating on the flask walls. These are the conditions,
however, that are most commonly used in practical biotech-
nological development work. At low shaking frequencies,
the gravitational acceleration, which acts in the axial direc-
tion of the shaking flask, is predominant over the centrifugal
acceleration and the bulk of the fluid moves on the base of
Figure 1. Main flow conditions inside a horizontal rotating drum
(adapted from Semena and Khmelëv, 1982). (A) Complete distribution of
the liquid on the drum wall. (B) Region of entrained thin liquid layer.
Boundary line I: Frr ⳱ 2.2 ⭈ Re0.77
f ; boundary line II: Frr ⳱ 2.2 ⭈ Re0.13
f ;
boundary line III: Frr ⳱ 2.2 ⭈ Ref0.44.
595
the shaking flask. This resembles the case of the horizontal
drum of finite length, treated earlier, where the side walls
influence the hydrodynamics. This is not covered by our
model.
To distinguish between operating conditions (shaking fre-
quencies) that are or are not covered by our model, an axial
Froude (Fra) number (radial/axial forces) for shaking flasks
is defined as follows:
共2 ⭈ ␲ ⭈ n兲2 ⭈ do
Fra =
2⭈g
To determine the liquid film Reynolds number (Ref) for the
shaking flask in an analogous form to the horizontal drum,
the liquid film thickness (f) for the uniform distribution of
the liquid on the flask wall must be defined. As with the
drum geometry, the distribution of liquid is approximated
by a hollow liquid cylinder. We therefore use the maximum
inside flask diameter, d, to describe the outer diameter of the
hollow liquid cylinder; the inner diameter is expressed by di
(analogous to Fig. 1). Eq. (6) is also applied to calculate the
film thickness (f).
To determine di, the simplifying relation h ≅ VL1/3 is used
as a characteristic length scale for the height of the friction
area between the rotating liquid and the flask wall (height of
the assumed hollow liquid cylinder) as in the derivation of
the modified power number, Ne⬘ (Büchs et al., 2000). By
rearranging Eq. (5) to di and inserting this expression into
Eq. (6), the liquid film thickness (f) is described by:
d
f= ⭈
2
冉 冑 冉 冊冊
1− 1−
4 VL1 Ⲑ 3
␲ d
2
The liquid film Reynolds number (Ref) for a shaking flask is
now obtained by replacing f in Eq. (4) by Eq. (9):
冉 冑 冉 冊冊
2 2
␳ ⭈ 共2 ⭈ ␲ ⭈ n兲 d 2 4 VL1 Ⲑ 3
Ref = ⭈ 1− 1−
␩ 4 ␲ d
冉 冑 冉 冊冊 2 2
␲ 4 VL1 Ⲑ 3
= Re ⭈ 1− 1−
2 ␲ d
MATERIALS AND METHODS
Measurements (a total of 2143 operating points) were taken
using the measuring method described in part I of this series
(Büchs et al., 2000). The flask size (100 to 2000 mL, DIN
12380), the shaking diameter of the shaker drive (do ⳱ 2.5
and 5 cm), the viscosity (␩ ⳱ 0.8 to 200 mPa ⭈ s), the
shaking frequency (n ⳱ 80 to 380 min−1 L/min), and the
filling volume (4% to 20% of the nominal flask volume)
were varied. The viscosity was varied using aqueous solu-
tions of polyvinylpyrrolidone (Luviskol, BASF AG, Lud-
wigshafen, Germany) in concentrations of 3 to 300 g/L.
These solutions show nearly newtonian flow behavior.
Stock solutions of approximately 4, 17, 33, 60, 100, and 200
596 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 68, NO. 6, JUNE 20, 2000
mPa ⭈ s were prepared and used throughout all experiments.
To ensure near-constant viscosity, the solutions contained
0.5 M sodium chloride and were stored for at least 7 days
before use. Before each experiment the viscosity was
checked again and readjusted if necessary. Between each
experiment the flasks were thoroughly washed out with
deionized water and dried at 60°C. The measurements of
viscosity and power consumption were taken at a tempera-
ture of 37°C.
(8)
RESULTS AND DISCUSSION
“In-Phase” Operating Conditions
To present the results of the 2143 measuring points for
shaking flasks in a closed form, the modified Newton num-
ber (Ne⬘) is plotted over the flask Reynolds number (Re) in
Figure 2.
The “in-phase” operating conditions (1115 individual
points) for flask sizes of 100 to 2000 mL are represented by
the measuring points concentrated in the upper limit of the
diagram. This limit shows the same course as the power
number (Ne) for unbaffled agitated tanks in the transition
range between laminar and turbulent flow. When working
with liquids of waterlike viscosity in shaking flasks, the
Reynolds numbers indicate (Re > 104) that the flow regime
is only just turbulent (Büchs et al., 2000). Fluids of elevated
viscosity lie in the transition range to laminar flow (see
Fig. 2).
(9)
These “in-phase” points are fitted using a least-square-
error method resulting in the following equation:
Ne⬘ ⳱ 70 ⭈ Re−1 + 25 ⭈ Re−0.6 + 1.5 ⭈ Re−0.2 (11)
−1
This correlation consists of a laminar (Re ), a transition
(Re−0.6), and a turbulent term (Re−0.2) analogous to well-
known hydrodynamic problems (Schlichting, 1979). Using
the fitted curve Eq. (11) and the definition of the modified
Newton number (Ne⬘) according to Eq. (1), the specific
power consumption of favorable operating conditions (in-
(10) phase system) can now be determined in advance.
“Out-of-Phase” Operating Conditions
A large proportion of the measuring points (1028 points) in
Figure 2 lies in a region well below the fitted curve. These
points represent unfavorable operating conditions (“out-of-
phase” system).
To demonstrate “out-of-phase” operating conditions, an
example is given in Figure 3. The modified Newton number
(Ne⬘) is depicted over the flask Reynolds number (Re) for a
250-mL flask, a filling volume of 25 mL and a shaking
diameter of 2.5 cm for varying viscosities. Up to a viscosity
of 4 mPa ⭈ s, the measuring points are well represented by
the curve calculated by Eq. (11). Starting from a viscosity of
17 mPa ⭈ s the difference between the measured and the
calculated “in-phase” [Eq. (11)] modified Newton number
Figure 2. Closed presentation of all measuring points (2143) for unbaffled shaking flasks in form of the modified Newton number (Ne⬘) dependent on
the flask Reynolds number (Re) with variation of flask size, viscosity, shaking diameter, and shaking frequency. The lower boundary curve [Eq. (12)] is
used for the model development (see text). The final differentiation of “in-phase” (large symbols) and “out-of-phase” (small cross symbols) operating
conditions is determined by Eqs. (8), (10), and (14).

(Ne⬘) increases continually, indicating the onset of “out-of-

phase” operating conditions.
The photographs shown in Figure 4 show the liquid dis-
tribution in the flasks for three representative operating con-
ditions. At a viscosity of 1 mPa ⭈ s, the bulk of the liquid is
at all times driven in the direction of the centrifugal accel-
eration inside the flask (Fig. 4a). At increasing viscosities
the bulk of the liquid shows an increasing delay (of up to
90°) relative to the direction of the centrifugal acceleration
(Fig. 4b). At a viscosity of 35 mPa ⭈ s (Fig. 4c), the liquid is
completely “out-of-phase”: the majority of the liquid re-
mains at the base of the flask and the maximum liquid
height is significantly reduced. With this only a minor frac-
tion of the liquid is actually rotating on the flask wall and
the total momentum transfer (i.e., power consumption) is
thereby greatly reduced. A comparable phenomenon is well
known when operating magnetic stirrers. With these devices Figure 3. Modified Newton number (Ne⬘) over the flask Reynolds num-
the magnetic forces normally assure that the magnetic bar ber (Re) for a 250-mL flask with a shaking diameter of 2.5 cm and a filling
rotates “in-phase” with the magnetic drive. Yet, if the vis- volume of 25 mL for varying viscosities.

BÜCHS ET AL.: POWER CONSUMPTION IN SHAKING FLASKS ON ROTARY SHAKING MACHINES. II 597
 cous forces of the liquid exceed the magnetic forces, the
stirrer bar will jump out of position and solely vibrate with-
out rotating. This results in near-zero power consumption.
As the complex flow conditions in the shaking flask are
somewhat simplified in our modeling approach, the in-
phase points do not lie perfectly on the fitted curve [Eq.
(11)] of Figure 2 but are scattered within a band. To account
for these scattered points, a lower boundary curve is intro-
duced in Figure 2, according to Eq. (12) (for derivation see
Appendix).

Figure 4. Liquid distribution in shaking flasks for a 250-mL flask with a

Ne⬘trans =
冉 70

Re ⭈ 10c Ⲑ
公1 + 1 Ⲑ m2
+
冉Re ⭈ 10 公 cⲐ
25
1+1 Ⲑ m2 冊 0.6

冊
shaking diameter of 2.5 cm, a filling volume of 25 mL, and a shaking
frequency of 200 L/min for: (a) a viscosity of 1 mPa ⭈ s, Ph ⳱ 2.61; (b) a
+
1.5
⭈ 10−c Ⲑ
公m2+1 (12)
viscosity of 75 mPa ⭈ s, Ph ⳱ 0.83; and (c) a viscosity of 135 mPa ⭈ s, Ph
⳱ 0.59. First row: photographs taken in the direction of the centrifugal
acceleration. Second row: photographs taken perpendicular to the centrifu-
冉Re ⭈ 10 公
cⲐ 1+1 Ⲑ m2冊0.2

gal acceleration. The arrows on the left side of the photographs show the
direction of the centrifugal acceleration. Every measuring point that lies above the lower boundary
curve according to Eq. (12) is used in the following model
development (Fig. 5) as an “in-phase” operating point. It

Figure 5. Separation of “in-phase” (large symbols) from “out-of-phase” (small filled circle symbols) operating conditions in dependency of the geometry
ratio (GR) and the liquid film Reynolds number (Ref).

598 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 68, NO. 6, JUNE 20, 2000
should be noted that the conditions depicted in Figure 2 as
being “in-phase” (with large symbols) do not correspond to
this definition, but are the results of Eq. (14) (see later). The
measuring points taken for the lowest viscosity liquid (wa-
ter, Re > 104) are always “in-phase,” independent of the
operating conditions.
To predict “in-phase” conditions, the model of a partially
filled rotating horizontal drum (described earlier) has been
applied to shaking flasks [Eq. (7) and Eq. (10)]. To elimi-
nate measuring points of low shaking frequencies, where the
majority of the liquid is in contact with the flask base and
where our model is invalid, only measuring points with
axial Froude numbers (Fra) of >0.4 [Eq. (8)] are considered
further.
The “in-phase” and “out-of-phase” operating points re-
sulting from Eqs. (8) and (12) are shown in Figure 5, using
the geometry ratio (GR), Eq. (7), and the film Reynolds
number (Ref), Eq. (10). The small filled-circle symbols
show “out-of-phase” operating points. All other symbols
represent in-phase operating points, at which the liquid
moves synchronously with the rotating movement of the
drive. Figure 5 shows that in the shaking flasks evaluated
there is a clear boundary between the two hydrodynamic
regimes, with only a small percentage of measuring points
(8.9%) not in agreement. Using a trial-and-error method in
a least-squares sense, the boundary curve between “in-
phase” and “out-of-phase” points is given by:
d will suffer from inadequate substrate supply to the cells.
GR = = 0.79 + 2.38 ⭈ log10共Ref兲 (13)
d0
It is quite remarkable that the slopes of the boundary curve
in the lower left and upper right-hand corner of Figure 5
(shaking flasks) are nearly equivalent to boundary lines I
and II of Figure 1 (rotating drum), respectively. The de-
duced hydrodynamic regimes in shaking flasks developed in
analogy to the basic problem of the rotating horizontal drum
can be considered legitimate.
To obtain an easy-to-use algebraic criterion to determine
the boundary up to which the liquid remains “in-phase”, we
introduce a nondimensional phase number (Ph). By multi-
plying Eq. (13) by a factor of 1.26 and rearranging, the
following equation results:
d0
Ph = ⭈ 共1 + 3 ⭈ log10共Ref兲兲 ⬎ 1.26 (14)
d
Eq. (14) states that operating points at which Ph > 1.26 are
“in-phase”. These “in-phase” operating conditions are de-
fined by Fra > 0.4 [see Eq. (8)], the liquid film Reynolds
number (Ref) [Eq. (10)], and the phase number Ph > 1.26
[Eq. (14)]. These criteria are used to differentiate between
“in-phase” and “out-of-phase” operating conditions in Fig-
ure 2. The calculated differentiation agrees reasonably well
for practical application with the differentiation represented
by the lower boundary curve [Eq. (12)]. It should also be
noted that the phase numbers for the operating conditions
shown in Figure 4a–c are given in the corresponding legend.
BÜCHS ET AL.: POWER CONSUMPTION IN SHAKING FLASKS ON ROTARY SHAKING MACHINES. II
The following general tendencies can be deduced from Eqs.
(8), (10), and (14). The flow in a shaking flask tends to be
“in-phase” at:
1. High shaking frequency (n).
2. Low viscosity (␩).
3. Large filling volume (VL).
4. Large shaking diameters (do).
5. Small maximum inside shaking flask diameters (d).
CONCLUSION
For the first time it has become possible to use algebraic,
physically based, nondimensional equations [Eq. (11) and
Eq. (14)] to determine sensible operating conditions for un-
baffled shaking flasks and to estimate the achievable power
consumption. When working with shaking flasks of ⱕ1 L
nominal volume, “out-of-phase” operating conditions are
only observed at elevated liquid viscosities. While in culture
experiments with bacteria and yeasts of single-cell morphol-
ogy this should not present problems, caution should be
exerted with filamentous microorganisms. In this case, the
phenomenon of “out-of-phase” operating conditions may be
of great practical relevance. Working under such conditions
will significantly reduce the oxygen transfer and mixing
intensity (detailed investigations are currently being con-
ducted in our laboratory). If a screening program with fila-
mentous microorganisms is unintentionally performed using
“out-of-phase” operating conditions, the whole screening
Improved strains will hardly be recognizable. In addition,
these operating conditions will make it very likely that
screening would be led in a entirely wrong direction. Strains
may attract attention by increased product titers that are in
reality not due to increased specific productivity but are the
result of an altered morphology. These changes in morphol-
ogy of the strains (shorter hyphae with fewer branches or
even single-cell growth) may lead to lower viscosities, re-
sulting in a switch to “in-phase” operating conditions and
therefore sufficient substrate supply of the cells, higher
metabolic activity and product titers. These seemingly “im-
proved” strains will probably not have increased specific
productivities. The benefit of such screening procedures for
a production process in a stirred-tank fermentor, assuring
sufficient substrate supply, is quite questionable. The same
applies to media development using “out-of-phase” operat-
ing conditions. Those media compositions that lead to
higher viscosities by polymers or by enhanced filamentous
growth will result in impaired substrate supply to the cells
and lower product titers. Possibly, the same media would
turn out to be superior at “in-phase” operating conditions
with a sufficient supply of substrates. In at least one indus-
trial case there is very strong evidence that the problems just
described have been encountered (unpublished results). It
can be expected that quite similar problems exist in many
other laboratories.
To determine sensible operating conditions for viscous
fermentation broths, which show non-newtonian behavior,
599
information about the level of the average shear rate must be The form of Eq. (11) is logarithmized:

冋 册
available. The influence of the operating conditions on the
average shear rate in shaking flasks is currently being in- 70 25 1.5
y = log x
+ 0.6x
+ (A1)
vestigated in our laboratory. In addition, the appearance of 10 10 100.2x
“out-of-phase” operating conditions at low viscosities (wa-
ter) in baffled shaking flasks and in large unbaffled shaking This curve is differentiated to determine the local slope:
flasks (ⱖ2 L nominal volume) will be reported on in a later
publication. 70 25 1.5
x
+ 0.6 ⭈ 0.6x
+ 0.2 ⭈
dy 10 10 100.2x
The experiments presented were performed at the life science =− =m (A2)
dx 70 25 1.5
laboratory of the BASF AG, Ludwigshafen, Germany. The au- + +
thors thank B. Mayer and the staff members of the biological 10x 100.6x 100.2x
pilot plant (ZHV/TB) for their technical assistance. We also
thank Schott Glaswerke, Mainz, Germany, for kindly supplying To determine the x- and y-translation along the normal to
the shaking flasks for the photographs.
the local slope, the following geometrical relations (Fig.
A1) are applied in each point of the fitted curve [Eq. (11)]:
NOMENCLATURE
m = aⲐb and c2 = a2 + b2 (A3)
A friction area (m2)
C, C1, C2, C3, C4, C5 constants (−)
resulting in:
c transition factor, in this case 0.1 (−)
di inner diameter of the liquid film inside a cylin-
c
der (m)
b= (A4)
d maximum inside shaking flask diameter, drum
diameter (m)
公m2 + 1
do shaking diameter (orbital shaker) (m)
f liquid film thickness inside a horizontal drum or The coordinates x* and y* (see Fig. A1) of the lower bound-
a shaking flask regime assuming total distribu- ary curve are thus related by:
tion of the liquid, Eq. (6) (m)
f(Re) function depending on flow (−) x* = x − a and y* = y − b

冋 册
Fra axial Froude number, Eq. (8) (−)
Frr radial Froude number, Eq. (3) (−) 70 25 1.5
g gravitational acceleration (m/s2) y* = log + + −b
GR geometry ratio for shaking flasks, Eq. (7) (−) 10共x*+a兲
共10 兲
共x*+a兲 0.6
共10 共x*+a兲 0.2
兲
h drum length, characteristic height of friction (A5)
area (m)
m gradient of the upper limiting curve at each Transforming Eq. (A5) to the linear state results in:

冉
point (−)
n shaking frequency, number of revolutions (s−1) 70 25
Ne⬘trans = +
冉Re ⭈ 10 公 冊
Ne Newton number (−)
Ne⬘ modified Newton number, Eq. (1) (−)
Re ⭈ 10c Ⲑ
公 1+1 Ⲑ m2 cⲐ 1+1 Ⲑ m2
0.6

冊
Ne⬘trans modified lower boundary Newton number, Eq.
(12) (−)
+
1.5
⭈ 10−c Ⲑ
公m2+1
冉Re ⭈ 10 公 冊
P power (W)
0.2
Ph phase number, Eq. (14) (−) cⲐ 1+1 Ⲑ m2
r drum radius (m)
Re flask Reynolds number, Eq. (2) (−) (A6)
Ref liquid film Reynolds number, Eq. (4) (horizon-
tal drum) and Eq. (10) (flask) (−)
VL filling volume (−)

Greek symbols

␳ density of liquid (kg/m3)

␩ dynamic viscosity of fluid (mPa ⭈ s)

APPENDIX

Derivation of Lower Boundary Curve

The lower boundary curve was designed in the logarithmic
scale by connecting points, which were moved by a specific
distance along the normal to the local slope of the fitted Figure A1. Geometric relations applied to determine the lower boundary
curve [Eq. (11)]. This method is now explained. curve.

600 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 68, NO. 6, JUNE 20, 2000
The width of the band is defined by the distance between the measurement in unbaffled flasks at low liquid viscosity. Biotechnol
upper and lower boundary curve. To account for the major- Bioeng 68:589–593.
Deiber JA, Cerro RL. 1976. Viscous flow with a free surface inside a
ity of “in-phase” operating points, the distance along the horizontal rotating drum. I. Hydrodynamics. Ind Eng Chem Fund
normal to the local slope (m) was chosen to be c ⳱ 0.1. 15:102–110.
Schlichting H. 1979. Boundary layer theory, 7th ed. New York: McGraw
Hill.
Semena MG, Khmelëv YA. 1982. Hydrodynamic regimes of a liquid in a
References smooth-walled rotating heat pipe. 1. J Eng Phys 43:1235–1242.
Thoroddsen ST, Mahadevan L. 1997. Experimental study of coating flows
Büchs J, Maier U, Milbradt C, Zoels B. 2000. Power consumption in in a partially-filled horizontally rotating cylinder. Exper Fluids 23:
shaking flasks on rotary shaking machines: I. Power consumption 1–13.

BÜCHS ET AL.: POWER CONSUMPTION IN SHAKING FLASKS ON ROTARY SHAKING MACHINES. II 601