World Journal of Mechanics, 2012, 2, 77-83
doi:10.4236/wjm.2012.22009 Published Online April 2012 (http://www.SciRP.org/journal/wjm)
Time Constants of the Transition between Onset and
Decay Reynolds Numbers for the Appearance of
Taylor-Couette Flow
Takashi Watanabe
Graduate School of Information Science, Nagoya University, Nagoya, Japan
Received January 4, 2012; revised February 1, 2012; accepted February 20, 2012
ABSTRACT
We investigate the onset and the decay of Taylor-Couette flow in finite cylinders, and we report the estimated time
scales in the azimuthal section of the flow state transition between the super-critical state and the sub-critical state by
fitting the numerical result to the so lution of the Stuart-Landau equ ation. The inner cylind er rotates, and the outer cylin -
der and both end walls of the cylin ders are stationary. Near the end walls of the cylinders, the v alue of the time scale is
small. In the inner region, the radial velocity component has a large time scale near the center of the vortices, while the
axial velocity component has a large time scale between the vortices.
Keywords: Taylor-Couette Flow; Onset/Decay Flow; Time Scale; Stuart-Landau Equation
1. Introduction
Since the pioneering work of G. I. Taylor, Taylor-Cou-
ette flow has been investigated using theoretical, experi-
mental and numerical approaches [1]. In this paper, we
investigate the flows between the coaxial rotating inner
cylinder and the stationary outer cylinder, and the leng ths
of the cylinders are finite. The geometrical parameters
that determine this flow are the aspect ratio and the ra-
dius ratio of the two cylinders and the parameter govern-
ing the flow is the Reynolds number defined by the
circumferential velocity of the inner cylinder and the gap
width of the two cylinders.
Benjamin [2,3] showed, when the lengths of the cy-
linders are finite, that the bifurcation from Couette flow
to Taylor vortex flow is imperfect. Since this study,
many flow modes have been discovered, and bifurcation
diagrams in the space spanned in the Reynolds number
and the aspect ratio have been evaluated [4,5].
Theoretical work has determined the critical Reynolds
number for the appearance of Taylor vortex flow from
Couette flow [6]. The theoretical value is found only in
the flow between infinite cylinders. Recently, the onset
and decay of the flow in the range of the Reynolds num-
ber including the critical value were investigated by Ab-
shagen et al. [7]. In their numerical calculation, the basic
flow is that between the infinite cylinders, and the effect
of the end walls of the cylinder was considered by intro-
ducing the homotopy parameter. This means that the
exact effect of the solid stationary end wall is not esti-
mated. Manneville and Czarny [8] fully considered the
end walls and examined their effect. They found that,
while the decaying flow follows Landau-type develop-
ment, the onset flow departs from Landau-type develop-
ment as the aspect ratio increases. However, both Ab-
shagen et al. and Mann eville et al. investigated the time-
dependent flow at one point, or only in the one-dimen-
sional region whose axis is parallel to the cylinder axes.
Therefore, the flow development in the entire azimuthal
region is not clear.
In this study, we investigate the onset flow and decay
flow in the entire azimuthal region. It is currently diffi-
cult to correctly measure the velocity components in the
entire region experimentally. Therefore, a numerical ap-
proach is adopted here. That is, we assume that the flow
state transition caused by the change of the rotating speed
of the inner cylinder follows the Stuart-Landau equation,
and we determine the initial amplitude, final amplitude
and time scale by fitting the numerical solutions to the
exact solution. This method is helpful in visualizing the
flow conditions of this type of flow. In this paper, we
show in particular the result of the time scales at a radius
ratio of 0.667 .
Section 3 discusses the formulation and fitting method
used, and Section 4 presents the results obtained in this
study. Finally, some conclusions are given in Secion 5.

T. WATANABE 78
2. Formulation
2.1. Flow Filed and Governing Equation
The flow field consists of two coaxial cylinders. Figure
1 shows the diagrammatic figure of the flow field. The
length of the cylinders is finite and is denoted by L. The
end walls of the cylinders and the outer cylinder are fixed
and stationary, while the inner cylinder rotates. The radii
of the inner and outer cylinders are Ri and Ro, respec-
tively. The radius ratio
 is given as Ri/Ro, and the as-
pect ratio
 is defined as L/(Ro – Ri).
The origin of the cylindrical coordinate system (r,
 , z)
is at the center of one end wall. In the numerical formu-
lation, the reference length is the gap width of the cylin-
der (Ro – Ri), and the reference velocity is the circum-
ferencial velocity of the inner cylinder, and all of the
physical values are made dimensionless by these refe-
rence values. Let u, v and w be the velocity components
in the radial, azimuthal and axial direction, respectively,
and their vector form is represented by u. Then, the go-
verning equations are given as follows:
0 u (1)

1
Re
Tp
t


uu
uu (2)
Here, t is time, p is pressure, and Re is the Reynolds
number based on the gap width of the cylinders and the
circumferential velocity of the inner cylinder.
In the numerical calculation, a staggered grid is used.
This grid is useful in avoiding the singularity at the cor-
ner between the stationary wall and the moving wall. The
convection terms are discretized by the QUICK method,
and the central difference method is used for other spatial
derivatives. The time integral is performed according to
the fractional step method.
The number of grid points in the radial direction is 81.
The number of grid points in the axial direction is 81 fo r
unity aspect ratio, and the number for other aspect ratios
is determined by the proportionality to the aspect ratio.
Sufficient accuracy has been confirmed using these grid
points. The boundary condition at the solid wall is the
no-slip condition on the entire solid wall. In the initial
condition, the flow is at rest in the entire region.
2.2. Fitting for the Solution of the Stuart-Landau
Equation
The Stuart-Lan dau E quat i o n [ 7] .
3
03
0
d
d
A
A
A
t
A
  (3)
is the equation that governs the amplitude near the criti-
cal point. Here,

0 is the time scaling factor. The critical
Reynolds numb er is denoted by Re0, and
 is rep re sen ted
Figure 1. Test flow configur ation and c oordinate system.
by (Re - Re0)/Re0. In the present study, the critical Rey-
nolds number is given as the Reynolds number beyond
which Couette flow becomes Taylor vortex flow between
the infinite cylinders. The solution of this equation is
given by


/
2
2/
e
e1
tT
f
tT fi
A
AtAA 

(4)
where Ai is the initial amplitude, Af = 0 A

is the am-
plitude in the final state, and T is the time scale. This
solution represents the initial exponential growth and
then the slowing down effect that makes the amplitude
remains at a plateau.
From the numerical results, the least square method is
used to determine Ai, Af and T. Several formulations of
the least square method are considered. The method that
makes the derivations of the accumulation of the square
errors with respect to Ai, Af and T zeros indicates the
most stable solution process. In our calculation, we used
this method.
In the case of radius ratio 0.667, the critical Reynolds
number in the infinite length cylinders is 76.5 [6].
We assume that the sub-critical flow appears at Re =
75.5 and the super-critical flow emerges at Re = 77.5.
Once the well-developed sub-critical flow and super-
critical flow are established, the Reynolds number is
suddenly chang ed to initiate the sup er-critical flow or the
sub-critical flow, respectively. Then, the transition pro-
cess between the sub-critical state and the super-critical
state is estimated.
3. Results and Discussion
3.1. Fully Developed Flow at Sub-Critical and
Super-Critical Reynolds Numbers
Flows in the azimuthal section at the sub-critical Rey-
nolds number 75.5 and the super-critical Reynolds num-

T. WATANABE 79
ber 77.5 are shown in Figure 2. In these panels, vectors
are two-dimensional vectors consisting of radial velocity
component u and axial velocity component w, and the
color contour represents the energy of the velocity given
by u2 + w2. The left side is the rotating cylinder and the
right side is the stationary outer cylinder.
When the aspect ratio is 4.0, vortices are found even at
Re = 75.5. This occurs because the effect of the Ekman
pumping driven by the Ekman vortices appears on the
stationary end walls, and the induced interior vortices
appear. At Re = 77.5, interior vortices appear more
clearly and we can count four vortices in total.
When the aspect ratio is 8.0, interior vortices do not
appear, except the two vortices near the end walls, in the
sub-critical state. However, the super-critical state in-
cludes interior vortices, and eight vortices appear.
In the flow for the aspect ratio 12.0, the sub-critical
state exhibits vortices near the end wall, while no clear
vortex appears in the interior section. The super-critical
state exhibits twelve vortices in the entire section.
3.2. Onset and Decay of Flows between
Sub-Critical and Super-Critical States
In Figure 3, the variations of the radial velocity compo-
nent in the middle region between cylinders are shown as
the profiles depending on the axial coordinate z. In this
case, time is measured from the point at which the Rey-
nolds number is changed. The panels include the flow
onset process from the sub-critical state to the super-
critical state and the decay process from the super-critical
state to the sub-critical state. In the onset process, the
well developed flow at Re = 75.5 is suddenly subjected
to the flow field with the inner cylinder rotating at the Re
= 77.5. Inversely, in the decay process, the Reynolds
number 77.5 is abruptly changed to 75.5. Then, the re-
stabilization is examined. In each panel, profiles are rep-
resented by black, blue and red lines as time progresses.
For example, the panels at
 = 8.0 show that the profiles
at times 0, 50 and 200 are represented by black, blue and
red lines, respectively.
When the aspect ratio is 8.0, the variation of the veloc-
ity component near the interior region of the cylinder
emerges, though this is not clear in Figure 2. At this as-
pect ratio, the development of flow almost terminates at t
= 200. In the profile of the onset p rocess, the values near
the end walls of the cylinders (z = 0.0, 8.0) change
sharply by t = 50 and then reach almost their final values
at t = 200, while this does not occur in the profile of the
interior regions. Similarly, the values at t = 50 near the
end walls in the decay process change to the final values
at t = 200. This means that the time scales for the onset
and decay processes are different at the end wall region
and at the interior region. We further discuss this differ-
ence in the next section.
The profile at  = 12.0 in the sub-critical state shows
the variation of the velocity component, even though no
Figure 2. Velocity vector and contour of the kinetic energy in the azimuthal section at the sub-critical Reynolds number 75.5
and the super-critical Reynolds number 77.5 at aspect ratio 4.0, 8.0 and 12.0.

T. WATANABE 80
Figure 3. Onset and decay of flow from the sub-critical state and the super-critical state, respectively. The profiles at the
three time points are shown. As time progresses, the profiles are denoted by black, blue and red lines.
obvious vortex is found in the interior region, as shown
in Figure 2. The effect of the change in the Reynolds
number first appears near the end walls of the cylinders,
and then the effect propagates into the interior region of
the flow field.
When the aspect ratio is 24.0, the value of the radial
velocity component in the sub-critical state remains small
at the inner region in the axial direction. On the other
hand, the effect of the Ekman pumping appears near the
end walls of the cylinder. As the Reynolds number in-
creases, the profile at t = 200 shows that the value be-
comes large from the end wall region to the inner region.
At t = 1000, the profile almost attains its final value. In
the decay process, the value in the interior region begins
to become small. In the final state at t = 1000, the value
of the profile is almost zero, except in the region that is
affected by the Ekman layer.
3.3. Variation of the Time Scale for Onset and
Decay in the Azimuthal Section
The transient process of the flow is assumed to follow
the Stuart-Landau equation, and the time scales at each
point in the azimuthal section are estimated.

T. WATANABE 81
The contours of the time scale of the radial velocity
component at
 = 8.0, 12.0 and 24.0 are shown in Figure
4. Please note the maximum value of the colo r bar of the
contour.
At aspect ratio 8.0, the time scale remains small near
the end walls of the cylinders: that is, the upper and
lower ends in the figure. The time scale is relatively large
at the center of vortices (please compare the positions at
large time scales with the positions of the center of the
velocity vector in Figure 2). This means that as the Rey-
nolds number changes, it takes a longer time for the ra-
dial velocity component to reach the plateau near the
vortex center where the radial velocity component is
small. From the legend of the color contour, it can be
said that the decay time is at least three times larger than
the onset time. This means that the onset and decay in the
range of the Reynolds number that includes the critical
value are not symmetric with respect to time.
The above results can also be found for the flow at  =
12.0 and 24.0. First, the time scale is large near the end
walls of the cylinders, regardless of the aspect ratio. Se-
cond, the time scales near the center of the vortices are
large compared with those of the other parts of the flow
region, such as areas near the boundaries of vortices.
Third, the time scale of the decay is larger than that of
the onset.
The contours of the time scale of the axial velocity
component are shown in Figure 5. When the flow is on-
set, the time scale is almost constant except near the end
walls of the cylinders, where the value is small. In the
decay process, the larger value appears at the inner re-
gion of the flow field. The large value appears near the
boundaries of the vortices. This is different from the time
scale profile of the radial velocity component which has
large near the center of vortices.
In the fitting method, we based our approach on the
solution given by Equation (4). In the calculation of the
real numbers, this equation assumes that
 is positive.
This means that the calculation using only real numbers
may not be applicable when  is negative and th e flow is
decay. On the other hand, our fitting method with real
number calculations exhibits well-converged result, even
when the flow is decay. Abshagen et al. [7] also used
Equation (4) as a model of the onset and decay flows and
showed good agreement with an experimental result.
Furthermore, no
 explicitly appears in Equation (4).
Therefore, at the first glance, Equation (4) is applicable
only the onset process, we can estimate the decay as well
Figure 4. Contour of the time constant of the radial velocity component, estimated from the solution of the Stuart-Landau
equation. In the color map , blue represents the value of 0.0 and red shows the value of Tmax.

T. WATANABE 82
Figure 5. Contour of the time constant of the axial velocity component, estimated from the solution of the Stuart-Landau
equation. In the color map , blue represents the value of 0.0 and red shows the value of Tmax.
as onset by using the solution of Equation (3). Further
discussion is required on this point.
4. Conclusions
In Taylor-Couette flow between a rotating inner cylinder
and a stationary outer cylinder, the onset and decay
processes found by the change of the Reynolds number
in the range including the critical value for the infinite
cylinders are examined numerically. In addition, the time
scales of the onset and decay processes of the flow are
estimated in the entire region of the azimuthal flow sec-
tion.
When the aspect ratio is small, the Ekman vortices
formed on the end walls of the cylinder dominate the
entire flow. At the higher aspect ratio, the interior region
is independent from the effect of the end wall exists.
For the radial velocity component during the onset and
decay processes, the effect of the change in the Reynolds
number appears first near the end walls of the cylinder
and the effect gradually propagates into the interior re-
gion.
The numerical result is fitted to the solution of the Stu-
art-Landau equation, and the time scales of the onset and
decay of the radial and axial velocity components are
estimated. The radial velocity component has a small
time scale that appears near the end walls, and it has a
large time scale that emerges near the center of the vor-
tices. The axial velocity component has a large time scale
near the boundary of the vortices at the interior region of
the flow section. The time scale of the decay process is
larger than that of the onset process, regardless of the
radial and axial velo city components.
REFERENCES
[1] R. Tagg, “The Taylor Couette Problem” Nonlinear Sci-
ence Today, Vol. 4, 1994, pp. 1-25.
[2] T. B. Benjamin, “Bifurcation Phenomena in Steady Flows
of a Viscous Fluid I, Theory”, Proceedings of the Royal
Society A, Vol. 79, No. 1696, 1978, pp. 1-26.
doi:10.1098/rspa.1978.0028
[3] T. B. Benjamin, “Bifurcation Phenomena in Steady Flows
of a viscous Fluid II, Experiment”, Proceedings of the
Royal Socie ty A, Vol. 359, No. 1696, 1978, pp. 27-43.
doi:10.1098/rspa.1978.0030
[4] T. Watanabe, H. Furukawa and I. Nakamura, “Nonlinear
Development of Flow Patterns in an Annulus with decel-
erating Inner Cylinder”, Physics of Fluids, Vol. 14, No. 1,
2002, pp. 333-341.

T. WATANABE 83
[5] H. Furukawa, T. Watanabe, Y. Toya and I. Nakamura,
“Flow Pattern Exchange in the Taylor-Couette System
with a Very Small Aspect Ratio”, Physical Review E, Vol.
65, No. 3, 2002, Article ID: 036306.
doi:10.1103/PhysRevE.65.036306
[6] E. L. Koschmieder, “Bénard Cells and Taylor Vortices”,
Cambridge Univ. Press, Cambridge, 1993.
[7] J. Abshagen, O. Meincke, G. Pfister, K. A. Cliffe and T.
Mullin, “Transient Dynamics at the Onset of Taylor Vor-
tices”, Journal of Fluid Mechanics, Vol. 476, 2003, pp.
335-343. doi:10.1017/S002211200200321X
[8] P. Manneville and O. Czarny, “Aspect-Ratio Dependence
of Transition Taylor Vortices Close to Threshold”, Theo-
retical and Computational Fluid Dynamics, Vol. 23, No.
1, 2009, pp. 15-36. doi:10.1007/s00162-009-0093-x