Hacettepe Journal of Mathematics and Statistics

Volume 35 (2) (2006), 117 146

CONVECTIVE AND ABSOLUTE
INSTABILITIES IN THE INCOMPRESSIBLE
BOUNDARY LAYER ON A ROTATING-DISK
M. Turkyilmazoglu

Received 02 : 01 : 2003 : Accepted 19 : 09 : 2006

Abstract
The linear absolute/convective instability mechanisms of the incom-
pressible Von Karmans boundary layer ow over a rotating-disk are
revisited in the present paper in order to review and assemble the avail-
able results in the literature on the topic. For this purpose the linearized
system of stability equations of motion is rst treated numerically, by
employing a Spectral method based on Chebyshev collocation as well
as a fourth-order Runge-Kutta method in combination with a shooting
strategy. Inviscid/viscous stationary and travelling modes which lead
to both convective and absolute instability mechanisms were success-
fully reproduced and compare favorably with those obtained by previ-
ous investigators. The validation of the zero-frequency upper-branch
modes was also accomplished by the asymptotic expansion technique
of [17], which is later extended to cover the non-zero frequency distur-
bances. The importance of the present study lies in understanding the
roles of possible instability mechanisms on the laminar-turbulent tran-
sition phenomenon in the three-dimensional boundary layer ow over
a rotating-disk, as well as related aerodynamic bodies.
Keywords: Rotating-Disk Flow, Absolute Versus Convective Instability, Matched
Asymptotic Expansion, Laminar-Turbulent Transition.
2000 AMS Classication: 76 E15.

Mathematics Department, Hacettepe University, 06532 Beytepe, Ankara, Turkey.

E-mail : turkyilm@hotmail.com
118 M. Turkyilmazoglu
1. Introduction
Many types of instability mechanism may be operational during the physical process
of transition from laminar ow to turbulence in fully three dimensional boundary layer
ows. In this study we deal primarily with the stability of the three dimensional boundary
layer ow due to a rotating-disk. The stability properties of this ow are similar to those
due to the ow over a swept-back wing in the sense that both ows are subject to
inviscid crossow vortex instability induced by the inectional character of the mean
velocity prole. The present paper will review particularly the recent progress made
towards elucidating the roles of the instability mechanisms for the special case of the
three-dimensional boundary layer ow due to a rotating-disk.
The concept of absolute versus convective instability has been introduced by [6] and
[4] in plasma physics. A recent review of the topic relating to uid ows can be found
in [20]. Absolute/convective instability describes the behavior of the impulse response
(or Greens function) to a forcing of an unstable medium. If an impulsively generated
small amplitude disturbance wave grows exponentially in time at every xed position in
space, the laminar state is termed absolutely unstable. Examples of such ows are closed
ow systems such as the Taylor-Couette ow and Benard convection. However, if the
resulting wave packet is convected away from the source location and ultimately leaves
the ow undisturbed then the ow is said to be convectively unstable. Plane-Poiseuille
ow and mixing layer ow (mixing layers can also show absolute instability, see [19]) are
examples of convective instability.
The early works on the stability of rotating-disk ow enlightened many of the aspects
relating to the convective nature of the ow. The experimental work of [16] (hereafter
referred to as GSW) clearly demonstrated the presence of co-rotating stationary vortices,
which arose due to the amplication of zero-frequency waves. From the theoretical point
of view, the inviscid stability properties were rst investigated by GSW. Both travelling
waves and stationary waves were studied. Malik [26] calculated the neutral stability
curve for stationary disturbances and he additionally showed the existence of a viscous
stationary mode which corresponds to zero mean wall shear stress of the crossow velocity
prole. The high Reynolds number asymptotic structure of the viscous mode of [26] was
elucidated by [17]. By considering nonlinear eects and following the framework set up
by [17], [25] extended Halls asymptotic analysis and obtained an amplitude equation for
the growth of disturbances close to the neutral location. The nonlinear theory showed
that the instability is subcritical. [2] conducted a theoretical analysis of some aspects of
linear and nonlinear non-stationary vortices in the rotating-disk ow and, in particular,
investigated the properties of neutrally stable nonlinear modes. The stability curves for
travelling disturbances have been computed numerically by [1].
Instead of a pure review of the literature, we intend here to validate and compare
the available results by reproducing them using both numerical and analytical means.
The rst objective of the current work is thus to investigate numerically, the neu-
tral/spatial/temporal stability characteristics of disturbances evolving over a rotating-
disk boundary layer ow. For this reason, the linear disturbance equations are derived in
a rational manner using approximations which are self-consistent. When the non-parallel
terms are ignored with an additional assumption of r (the radial distance from the center
of the disk) being set to 1, the traditional sixth-order stability equations are arrived at.
These equations, which have been treated before by [26] and [1], are solved here using a
dierent technique based on Chebyshev collocation with a staggered grid structure. A
Runge-Kutta scheme was also employed for checking the self-consistency of our results.
The use of spatio-temporal linear stability analysis shows that, in line with the literature,
Convective and Absolute Instabilities 119
three families of eigenfunctions are possible in the rotating-disk ow. Monitoring the de-
velopment of the eigenvalues corresponding to these families shows branch interchanging
between distinct waves at certain Reynolds numbers.
A second objective is to study the non-stationary upper branch behavior of the dis-
turbances from an analytical point of view. For this purpose the asymptotic framework
of [17] has been extended to include non-zero frequency waves. An investigation of the
second-order eigenvalue problem, allowing viscous and non-stationary eects to appear,
reveals the appearance of a phase shift across the critical layer and a wall layer shift. The
matching of these shifts generates an eigenrelation for the non-stationary modes which
is akin to the one obtained by [17] and [7] for stationary modes.
Singularities arising in the dispersion relationship enable the nature of the instability
to be determined. The variation of the Reynolds number can cause such points to occur
in the ow eld and thus alter the behavior of a ow from a convectively unstable state
to an absolutely unstable regime. These points form whenever modes associated with
waves propagating in the opposite or the same directions coalesce. If the coalescing
branches originate from the waves propagating in opposite directions, the singularity
which causes resonance is said to be of pinch-type. For linearly unstable systems such
a direct resonance point separates an absolutely unstable region from a convectively
unstable region. Examples of ows demonstrating this phenomenon are near-wake ows
([5], [29] and [33]), swept-hiemenz ow ([32]) and mixing layer ows with backow ([19]).
On the other hand, if the two coalescing modes originate from waves propagating in
the same direction, then the corresponding singularity is of the double-pole type. When
these coalescing modes are nearly neutral, the damping rates are very small and thus
a resulting short-term algebraic growth for small times or short distances may carry
the whole system into the nonlinear stage long before the exponentially growing mode
does. This is the case addressed by, amongst others, [3], [21] and [31] for Plane-Poiseuille
ow and Blasius boundary layer ow. Both resonance cases were numerically examined
by [8, 9] when the disk is coated by a compliant material. The secondary instability
mechanism due to the absolutely unstable waves was also studied by [30].
The study of [23] has motivated us to reinvestigate the absolutely unstable viscous
regime for the rotating-disk boundary layer ow and to demonstrate whether this range
matches onto the absolutely unstable inviscid ow regime of [23] in the limit for high
Reynolds number. Lingwoods calculations were done using only a fourth-order Runge-
Kutta integrating method, but in this study we use a Spectral collocation technique.
In fact, the recent asymptotic study of [18] showed that the family of branch points
discovered by [23] is not the viscous continuation of the inviscid branch points, but is
fundamentally of a viscous long-wave character. Our numerical results well justify the
asymptotic results of [18].
The numerical simulation outcomes of [10, 11] pointed to the fact that absolute in-
stability is not an eventually dominated mode but that the transition seems to be as a
result of the convective nature of the rotating-disk boundary layer ow. There is then
the possibility that nonlinearity may be rst triggered by the direct spatial resonance
instability mechanism. A numerical task is therefore undertaken here to search for the
Reynolds numbers at which direct spatial resonance takes place. The results demonstrate
that the family I and II branches both of which originate in the same wave-number plane
form a direct spatial resonance at a Reynolds number of about 445, at which the ow is
still laminar.
This review consists of two interrelated parts organized as follows. The rst part is
devoted to convective instability and the second part to absolute and direct spatial reso-
nance instabilities. In 2 the general governing equations of the motion are given followed
120 M. Turkyilmazoglu
by the linear stability theory in 3-4. The linear stability results are then presented in
5-5.5. For the second part, in 6 a denition of absolute/convective instabilities based on
Briggs pinching criterion is presented and also the absolute instability results are given in
7. 8 involves the modal coalescence results. Further discussion and conclusions follow
in 9.
2. The basic equations
2.1. Governing equations of the ow. We consider the three dimensional boundary
layer ow of an incompressible uid on an innite disk which rotates about its axis with a
constant angular velocity . The Navier-Stokes equations are non-dimensionalized with
respect to a lengthscale L = r

e
, velocity scale U
c
= L, timescale L/U
c
and pressure scale
U
2
c
, where is the uid density. This leads to a global Reynolds number Re =
U
c
L

= R
2
,
where R is the Reynolds number based on the displacement thickness = (

)
1
2
. Thus,
relative to non-dimensional cylindrical polar coordinates (r, , z) which rotate with the
disk, the full time-dependent, unsteady Navier-Stokes equations governing the viscous
uid ow are the usual momentum and the continuity equations, and these are given as
follows:
(1)
u
t
+ u
u
r
+
v
r
u

+ w
u
z

v
2
r
2v r =
p
r
+
1
R
2
[
2
u
2
r
2
v

u
r
2
],
v
t
+ u
v
r
+
v
r
v

+ w
v
z
+
uv
r
+ 2u =
1
r
p

+
1
R
2
[
2
v +
2
r
2
u

v
r
2
],
w
t
+ u
w
r
+
v
r
w

+ w
w
z
=
p
z
+
1
R
2
[
2
w],
u
r
+
1
r
v

+
w
z
+
u
r
= 0.
The Laplacian operator in cylindrical coordinates is given as
(2)
2
=
_

2
r
2
+
1
r
2

2
+

2
z
2
+
1
r

r
_
.
In this analysis the uid is assumed to lie in the semi-innite space z 0. In the above
equations the curvature eects as well as the eects stemming from the Coriolis forces
are present. It is now well known that these terms have a strongly stabilizing impact in
the linear stability theory, as pointed out by [27] and [34].
2.2. The mean ow. The dimensionless mean-ow velocities and pressure are given
by Von K arm ans exact self-similar solution of the Navier-Stokes equations for steady
laminar ow. The boundary layer coordinate Z, which is of order O(1) is dened as
Z = zR, and the self-similar equations take the form
(3) (u
B
, v
B
, w
B
, p
B
) = (rF[Z], rG[Z],
1
R
H[Z],
1
R
2
P[Z]),
where the functions F, G, H and P satisfy the following ordinary dierential equations
(4)
F
2
(G + 1)
2
+ F

H F

= 0,
2F(G + 1) + G

H G

= 0,
P

+ H

H H

= 0,
2F + H

= 0.
Convective and Absolute Instabilities 121
Here primes denote derivative with respect to Z and the appropriate boundary conditions
are given as
(5)
F = G = H = 0 at Z = 0,
F = 0, G = 1, H = h

as Z .
The value of h

is a constant vertical velocity of the rotating uid in the far-eld above

the disk, and it has to be found numerically in the course of the solution of equations (4)
and (5).
3. Derivation of the linear stability equations
3.1. Viscous disturbance equations. We are here interested in perturbation so-
lutions of Von K arm ans self-similarity velocity proles (3). The instantaneous non-
dimensionalized velocity components imposed on the basic steady ow are u, v, w and
the pressure component p, and they can be expressed as
u[r, , z, t] = u
B
+ u

[r, , z, t].
Having linearized the equations for small perturbations, we nd that the linearized
Navier-Stokes operator has coecients independent of and hence the disturbances can
be decomposed into a normal mode form proportional to e
iR(t)
. Such an approxi-
mation leads the disturbances to be wave-like, separable in and t. Consequently, the
perturbations may be assumed to be of the form
u

=

f[r, Z]e
[iR(t)]
+ c.c.,
where and are respectively the wave-number in the azimuthal direction and the
scaled frequency of the wave propagating in the disturbance wave direction.
The separation in and t simplies the linear system of equations. However no
such simplication arises as far as the r dependence is concerned (except in the limit as
R ) and the full linearized partial dierential system has to be solved subject to
suitable initial conditions to determine the stability of the ow. Consider next the limit
R and introduce the scale X = Rr which is the appropriate scale on which the
disturbances develop. After allowing for the multiple-scale replacement of

r
by
R

X
+

r
,
and keeping only terms of O(1/R), the reduced system of equations stemming from these
approximations can be written in the following form
(6)
f

Hf

[iR(F + G ) +
2
+ F]f + 2(G + 1)g RF

h iRp = 0,
g

Hg

[iR(F + G ) +
2
+ F]g 2(G + 1)f RG

h iRp = 0,
h

Hh

[iR(F + G ) +
2
+ H

]h R p

= 0,
f + ig + h
z
= 0,
where
2
=
2
+
2
, = i +
1
R
. It should be pointed out that the non-parallel terms
which are of O(1/R) were also dropped from equations (6). The linear equation system
above is identical to the one used by previous investigators, see for instance [26] and [23].
The boundary conditions for this set of equations are f = g = h = 0 at the solid wall
(Z = 0). Considering the decaying property of the disturbances, the boundary conditions
to be imposed far away from the disk surface are derived from the asymptotic form of
equations (6).
122 M. Turkyilmazoglu
3.2. Inviscid Rayleigh equation. Further neglect of all the terms of order R
1
in (6)
leads to the well-known Rayleigh equation
(7) [(F + G)(D
2

2
) (D
2
F + D
2
G)]h = 0,
where,
2
=
2
+
2
and D =

Z
.
4. Solution Technique
4.1. Spectral treatment of the stability equations. The discretization technique
we use here is a Spectral collocation method using Chebyshev polynomials as basis func-
tions. A staggered grid is used only in the Z direction. The pressure is dened at the
cell centers, at half points, but other components at the cell faces. Since no pressure
points fall on the boundary, the critical boundary conditions for the pressure have there-
fore been avoided. Momentum equations are thus collocated at Gauss-Lobatto points
cos[k

N
], whereas the continuity equation is imposed at Gauss points cos[(k +
1
2
)

N
]. The
Chebyshev interpolation between these two dierent grids, from cell centers to cell faces
or otherwise is then employed as given in [28].
The computational physical plane has been mapped onto the Spectral space by means
of the linear transformation
(8) = 1 +
2
Z
max
Z.
Here Z
max
is the far-eld boundary of the ow, which throughout the calculations is set
to a nite value of 20, about 4 times the boundary layer thickness.
The resulting equations from the discretization ultimately can be assembled as a large
generalized matrix-eigenvalue problem of the form
(9) LU = 0,
where L is a (4N + 3) (4N + 3) full matrix and U = [f, g, h, p]
T
is the eigenfunction.
The incorporation of boundary conditions only necessitates some modications to the
rst and last row of the matrix L.
4.2. Newton-Raphson searching technique for the branch points. To determine
a branch point in the and complex planes, we have to x and R, and shoot
for simultaneous zeros of both the dispersion relation and the group velocity

. The
corrections to and which are denoted by and (which are complex in general)
can be obtained from the solution of the following linear equations
(10)
U
1
( = 1) +
U
1

( = 1) +
U
1

( = 1) = 0,

2
+

2

= 0.
Note that the variables are complex-valued, therefore they can be written in the matrix
form
(11) J = RHS,
where, J is a full 4 4 square matrix of coecients, RHS is a 4 1 known vector and
= [, ]
T
.
In computing the neutral branch points one has to consider the following equations
(12)
_
U
1
+
U
1

+
U
1

+ R
U
1
R
_
=1
= 0,

2
+

2

+ R

2

R
= 0.
Convective and Absolute Instabilities 123
Here, we x the azimuthal wave-number as real and the increment is now also real.
After one solution of the matrix equation (9) the quantity U
1
( = 1) is now known.
The corresponding group velocity can be calculated by two dierent approaches itemized
as follows
(i) Solve the matrix equation at successive
0
,
0
and
0
+ . Then, store
the corresponding frequencies as
0
,
1
and
1
. This set of eigenvalues is then
used to approximate the group velocity

as

1

1
2
and the second derivative
of the group velocity

2

2
as

1
2
0
+
1

2
. To compute successive derivatives
of the group velocity and the derivatives of the eigenfunctions required in the
equations (10) and (12), we can simply use the same procedure.
(ii) From the Taylor expansion we can approximate the group velocity as
(13)

=
U
1

U
1

( = 1).
However, one can estimate the derivatives of the eigenfunctions with respect to
and by dierentiating (9) to give
(14)
L
U

=
L

U,
L
U

=
L

U.
Having solved the matrix equation (11), the corrections are added to the previous eigen-
values to give a better approximation and reduce the discrepancies on the wall boundary
condition. Then, an iterative scheme is employed until the magnitude of U
1
( = 1)
and

are simultaneously less than some small prescribed value.

5. Linear stability results
5.1. Basic velocity proles. In the solution procedure, the equations (4) were reduced
to a system of rst order dierential equations and then solved by a fourth-order Runge-
Kutta scheme combined with a shooting procedure. The far-eld boundary condition
Z
max
= 20 was found to be satisfactory and the corresponding normal velocity h

was
then calculated up to 6 digits as 0.884423 from this method. The computed velocity
distribution for the basic ow on the rotating-disk is shown in Figure 1 (a). This similar-
ity solution has been used by the researchers cited herein, see for instance [23] and [15].
A combination of the radial and azimuthal velocities leads to a prole which is inec-
tional and which contributes to the occurrence of crossow instability in the rotating-disk
ow. A few examples of these proles are shown schematically in Figure 1 (b) at several
Reynolds number locations, for both stationary ( = 0) and a non-stationary ( = 7.9)
case.
From the literature, two types of instability waves are known to exist in the rotating-
disk boundary layer ow that cause either convective or absolute instabilities. The rst
one is the inviscid-type or the upper branch and the other one is the viscous-type or
the lower branch. We will investigate these two distinct branches in more detail in the
following sections.
5.2. Neutral waves.
5.2.1. Stationary and travelling waves. Neutral curves play a key role in the determi-
nation of the bounds of the stability. For comparison purposes, the neutral curves for
both stationary and non-stationary waves in the (R, ) and (R, ) planes are shown in
124 M. Turkyilmazoglu
Figures 1 (c-d) for a negative and several positive non-dimensional frequencies. The wave-
angle of a disturbance is dened as = arctan[

r
]. These curves are very similar to the
ones calculated previously by [26] and also [1]. In line with the ndings of the latter, it
is noticed that a minimum occurs on the lower branch for certain xed frequencies. The
existence of such second minima was also observed in the experiments of [12] and [13].
For the negative frequency = 5.0 in Figure 1 (c), there are no longer two minima.
Compared to zero-frequency disturbances, instability starts much earlier at a smaller
critical Reynolds number. The corresponding critical Reynolds number for a negative
frequency of = 5.0 is R = 284.910 and the associated eigenvalues are = 0.36310,
= 0.10710 and = 16.46

, respectively. At = 4.0 the critical Reynolds number

and the eigenvalues in Figure 1 (c) are R = 297.630, = 0.40302, = 0.05601 and
= 7.96

for the upper branch and R = 172.820, = 0.15210, = 0.01202 and

= 4.48

for the lower branch, respectively. The same values for = 10.0 are R =
329.371, = 0.43001, = 0.03203 and = 4.20

for the upper branch and R = 68.709,

= 0.33601, = 0.12903 and = 20.98
o
for the lower branch, respectively. And
nally, the lowest critical Reynolds number for the lower branch occurs at frequency
= 7.9. The corresponding eigenvalues are R = 64.443, = 0.27597, = 0.10669 and
= 20.87
o
, respectively. When the frequency increases beyond 7.9 the lower critical
Reynolds number starts increasing too. The aforementioned critical values are almost
indistinguishable from those obtained by [1].
The wave-angle stability curves are depicted in Figure 1 (d). On the upper branch the
non-zero frequency waves approach the zero-frequency neutral curve at large Reynolds
numbers. Increasing frequency causes loops on the upper branch. However, these do not
correspond to a bifurcation. The resulting intersections of the curves is an artifact of
projecting the curves onto the plane. We can conclude that there exist a critical angle
below which all the waves are linearly damped. This critical angle is, according to our
calculations, 35.34

at a frequency 7.9. The same angle at frequency 10 is 35.04

. This
angle is given in [1] as 50.4

. From a consideration of the Rayleigh equation, for positive

frequencies we have found that basic velocity proles cease to have a point of inexion
for values of < 36.9

, and therefore there are no unstable eigenvalues for angles less

than this critical value. This value is consistent with our numerical computations cited
above for large Reynolds numbers. The wave-angle continuously increases on the lower
branch for increasing Reynolds numbers. It is known that the direction of zero wall mean
shear stress corresponds to a critical angle of about 39.64

, see [17]. At R = 10
6
the
corresponding wave-angle was found to be 39.216

.
5.3. Spatial and temporal waves. In this section we examine the spatial/temporal
instability characteristics of Von K arm ans basic velocity prole as shown in Figure 1. The
spatial amplication rate (
i
) contours are drawn for certain frequencies. Figure 2 (a-
c) show these contours in the (R, ), (R, ) and (R, ) planes for the stationary waves,
and Figures 2 (d-f) show the results for frequencies = 4. In good agreement with the
results of [1], all these contours show a linear variation as the growth rate increases in
magnitude. Beyond some critical amplication rate the second minimum disappears for
a xed frequency. From the stationary stability diagram Figure 2 (c) it can be seen that
the most unstable wave-angle varies slowly with Reynolds number on the upper branch.
We have also plotted the growth rate curves at some selected Reynolds numbers.
Figure 3 shows the spatial growth rate (
i
) curves at R = 1000 as a function of wave-
angle. On the lower branch the growing waves originate almost from the same place at
a wave-angle of about 11.45
o
, on the other hand, the upper branch amplifying waves
originate at a larger wave-angle as R increases. Travelling waves with a frequency of 10
Convective and Absolute Instabilities 125
are shown in Figure 3 (b). This graph suggests that travelling waves are less amplied
than the stationary ones. It should be remarked that as will be shown later the ow is
already absolutely unstable at R = 1000 and a true spatial growth rate for time-periodic
disturbances is unlikely to occur physically. Figures 3 (a-b) are thus shown only for
mathematical interest and comparison purposes. In Figures 3 (c-d), spatial calculations
are shown to determine the frequency of the most unstable wave of family 1 (see below)
modes at R = 500. It can be seen from Figure 3 (c) that unlike the stationary waves the
travelling waves are the most unstable. Moreover, the most unstable region is dominated
by the negative frequency region with the phase velocity

2
r
+
2
directed inwards.
Next, the evolution of dierent families of eigenfunctions is considered. The spatial
families of the stationary eigen-solutions originating at R = 400 are shown in Figures 4 (a-
f). As shown, the family 1 branches have neutral points on the upper and lower branches
of the neutral curve. Unlike the family 2, which includes only downstream propagating
damped waves, family 1 involves both damped and amplied downstream propagating
waves in a specic range of wave-numbers. The third family which seems to be only on
the lower plane has no neutral points at all and is characterized by large values of
i
,
corresponding to upstream propagating damped waves.
The behavior of the stationary eigenfamilies has also been captured in the work of [1].
A similar trend of non-stationary disturbances at = 7.9 is shown in Figures 4 (d-f),
at a Reynolds number of 400. The main dierence is that family 2 in this case also has
a neutral branch point at R = 400. This is a result of lower branch instability for the
positive frequencies. It should also be mentioned that we have investigated the real part
of the group velocity given by

. We found that for the most part, for family 3 up to

= 0.04, the real part remains negative, and the real part of the group velocity angle is
less than 90
o
. This indicates an inward energy transfer towards the center of the disk.
Consequently, the waves in this direction are strongly damped. However, for less than
0.04 the group velocity vector is directed outward and so the waves are amplied in this
direction.
Figures 3-4 point to another mechanism for instability. In particular, branch switching
in spatial stability calculations can be eectively used as an indication of qualitative
change in the nature of the instability from convective to absolute or vice versa. We shall
investigate this feature of the rotating-disk ow further in a later section.
5.4. Contour lines and cross-ow vortices. In this part, we display contour plots
and vector elds with cross-ow vortices for the frequencies = 0 and = 7.9 in
Figures 5 and 6. The contours contain the velocity components
F + G + Re[(f + g)e
[(ir+Rt)]
], (15)
H
R
+ Re[(h)e
[(ir+Rt)]
] (16)
for various Reynolds numbers. In each gure the portions (a) and (c) refer to (15) and
portions (b) and (d) refer to (16). Cats eye patterns, which are a characteristic feature
of the cross-ow instability, are clearly seen in these contour diagrams.
Vector plots are shown in Figures 6 (a-b) for the stationary cross-ow vortices and
Figures 6 (c-d) for the travelling waves. There are essentially two sets of vortices present.
The rst set of vortices are nearest to the wall as suggested by Stuart in GSW. Further-
more, they have formed only on the upper branches. The second set of vortex structures
are more visible on the upper branches of the neutral curves. Malik [26] also displayed
zero-frequency cross-ow vortices, but from his computations the near wall vortices were
126 M. Turkyilmazoglu
not captured. As one travels from the upper branch to the lower one, the location of the
vortex center also travels towards the wall.
5.5. Asymptotic analysis of the upper branch non-stationary modes. Hall [17]
computed the evolution of the upper branch modes considering only the zero-frequency
waves. In this section our intention is to extend his theory to include the cases where non-
zero frequency waves also exist. Following the work of [17] we dene a small parameter
= R
1/6
. On the upper branch we assume that the disturbances take the form
(U, V, W, P) = (u(z), v(z), w(z), p(z))e
i

3
(

r
dr+)
e

2
t
,
and we also expand the wave-numbers and frequency as
(, , ) = (
0
,
0
,
0
) + (
1
,
1
,
1
) + .
As in the study [17], we restrict our attention to neutral disturbances at a local position
r. In the following the regions where such waves develop are considered, and asymptotic
solutions are sought in each region. The analysis below also follows closely the work of
Cole [7].
5.5.1. Inviscid region = z
3
(z = O(R
1/2
)). The existence of this inviscid zone of
depth O(
3
) was shown by GSW. In this region u, v, w and p are expanded in the form
(u, v, w, p) = (u
0
, v
0
, w
0
, p
0
)() + (u
1
, v
1
, w
1
, p
1
)() + ,
Substituting these into Equations (1), and equating the terms of order of O(
3
), the
leading-order approximation leads to the inviscid Rayleigh equation, see [17]. Dening the
eective velocity prole

U
B
=
0
rF +
0
G and leading-order wave-number
2
0
=
2
0
+

2
0
r
2
,
the solution is restricted to satisfy

U
B
&

U

B
to vanish at a non zero =

, so that the
singularity in the Rayleigh equation is avoided. The eigenvalue problem was solved by
[17], and it was found that the unknowns
0
,
0
,
0
and

satisfy
(17)
0
= 1.16,

0
r

0
= 4.26,
0
= 1.1293,

0
r
= 0.2651,

= 1.46.
Proceeding to the next-order in the inviscid zone, the second-order equations come from
O(
2
) terms and eliminating u
1
, v
1
and p
1
we obtain the non-homogeneous Rayleigh
equation for w
1
in the form
(18)

U
B
_
w

1

2
0
w
1

B
w
1
= 2

U
B
w
0
+
0
r
_
F

U
B
_
w
0
+
0

U
B
w
0
,
where =
0

1
+

0

1
r
2
and =

1

0

0

2
0
. Due to the second term on the right-
hand side of (18), a Frobenius expansion reveals that a logarithmic singularity appears
belonging to w
1
at =

. This singularity manifests itself in the form
w

1

k
c

w
0
(

), >

, (19)
where k
c
is given in Appendix A. The singularity (19) can be removed by introducing
a critical layer at =

, and then the solution of (18) that satises w
1
() 0 is as
given in [17], except that an extra integral appears due to the non-zero frequency term
in equation (18).
Taking into consideration the viscous critical layer theory, if the path of the integration
is deformed into the complex plane near =

(i.e, continuation below the critical layer),
the following holds:
w
1
k
c
(

)
_
ln(

) i sgn(

B
|
=

w
0
(

), <

.
Convective and Absolute Instabilities 127
This and (19) give the well-known linear phase jump in the inviscid zone
(20) w

1
|

= i sgn(

B
(

))k
c
w
0
(

).
5.5.2. Viscous sublayer = z
4
(z = O(R
2/3
)). From the leading-order equations in
the inviscid zone we nd that
0, i(
0
u
0
+

0
r
v
0
) w

0
(0).
Therefore, to satisfy the zero-velocity requirement at the wall, a viscous sublayer is
required. Balancing the convection and diusion terms in the disturbance equations (1),
it is found that the thickness of this layer should be O(
4
). The basic velocity proles
are expanded as in [17]. The various quantities also expand as follows:
(u, v, w, p) = ( u
0
, v
0
, 0, 0)() + ( u
1
, v
1
, w
0
, p
0
)() +
2
( u
2
, v
2
, w
1
, p
1
)() + .
Substituting these into the Navier-Stokes Equations (1), and after some manipulations
as in [17], we nd the wall layer shift for large to be
w
0
w

0
(0)
_
+
A
i
(
0
)

_
,
with
0
= i

2
0
,
3
0
= i[
0
rF
0
+
0
G
0
] and =
_

0
A
i
()d. On matching with the
inviscid zone described earlier, we nd that w
1
must satisfy
(21) w
1
w

0
(0)
A

i
(
0
)

, 0.
We see from (18) that w
1
satises equations identical to w
0
apart from a non-homogeneous
part on the right-hand side. Therefore a certain solvability condition is required to ensure
that a solution exists. To proceed further this solvability condition can be obtained by
multiplying (18) with w
0
and integrating from wall to innity so that
(22)
_

0
w
0
Lw
1

U
B
d = w

0
(0)
2
A

i
(
0
)

isgn[

B
(

)]w
2
0
(

)k
c
,
where,
_
denotes the nite part of the integral, L =
d
2
dZ
2

2
0


U

U
B
and the use of
equations (17), (18), (20) and (21) is made. Furthermore, using equations (18) and (22)
we obtain the eigen-relation
(23) 2 I
1
+ r [I
2
+ a] +
0
[I
3
+ b] = w

0
(0)
2
A

i
(
0
)

,
where I
1
, I
2
, I
3
, a and b are given in Appendix A.
Now using the Reynolds number Re based on the boundary layer thickness and the
local azimuthal velocity of the disk, Re = rR
1/2
, the eective wave-number (
2
+

2
r
2
)
1
2
and the wave-angle in powers of Re are given by
(24)
= (
2
+

2
r
2
)
1
2
=
0
+

0
Re

1
3
+ ,
(tan )
1
=

0
r

0
+ Re

1
3
+ .
Note that if
0
= 0 the explicit expressions for the eective wave-number and the wave-
angle outlined in [17] are recovered from the eigen-relation (23). The extra term I
3
is due
to our consideration of non-zero frequency waves. In this case, for any non-zero
0
, the
equation (23) should be solved numerically, to determine the wave-number correction
and the wave-angle correction in (24).
Based on these asymptotic ndings, comparisons with the numerical calculations are
shown in Figure 7 for = 5, 0 and 10. In this gure the long curves show numerical
128 M. Turkyilmazoglu
results and shorter lines asymptotic ones. Notice that to make comparisons with the
numerics,
0
should be re-scaled by Re. It is seen that there is satisfactory agreement
between the asymptotic predictions and the numerical results. In order to increase the
agreement domain of numerical and asymptotic analysis for smaller Reynolds numbers,
more terms of the expansions in (24) need to be calculated, but this is beyond the scope
of the present study. Moreover, it is also possible to consider
0
as complex to further
investigate temporally growing waves. Furthermore, using the asymptotic relation of the
Airy function as |
0
| , the right-hand side of (23) can be replaced by
iw

0
(0)
2

3
0
,
which then gives explicit expressions for the wave-number and wave-angle corrections.
6. The absolute/convective instability of the rotating-disk ow
In this section we shall give a description of absolute/convective instabilities and
further investigate under what circumstances the three dimensional boundary layer ow
becomes absolutely/convectively unstable. In doing so, we shall use the ideas of [6] and
[4]. The specic example we consider is the three dimensional rotating-disk boundary
layer ow.
The complex dispersion relation which can be obtained through the application of
classical linear stability theory to a three dimensional parallel ow can be represented in
the form
(25) D(, ; , R) = 0.
Note that in three dimensional ows the relation (25) denes the local instability char-
acteristics due to the fact that parallel ow assumptions or local mean velocities have
been employed.
From now on we assume that the zero normal velocity condition on the wall is replaced
by a line forcing given in the form
(r r
f
)(t t
f
)e
iR
, (26)
where (r r
f
) and (t t
f
) are the Dirac delta functions at a non-dimensional radius
r
f
, and at a specic time t
f
(=0 formally), respectively. Based on the dispersion relation
(25) we can now form the impulse response or Greens function G(r, , t; , R) to the
point forcing (26) associated with the operator D(i

r
, , i

t
; , R) as a double integral
(27) G(r, , t; , R) =
1
4
2
e
iR
_
F
_
L

G
e
i(rt)
D(, ; , R)
dd.
The impulse response G above is dened as the instability wave eld generated by a
Dirac delta function in space and time, as posed by (26). A function

G in equation (27)
appears and is dependent on the normal coordinate, which essentially is the eigensolution
of the governing equations (6). Although the singularities caused by

G (generally the
branch cuts causing the continuous singularities) are important in the inversion integrals
in (27), we assume that the main contribution to the response (27) comes from the zeros
of D, which are the discrete eigenvalues obtained from (25). These eigenvalues provide
a mapping between the and planes dened by the F and L contours, respectively.
A rigorous mathematical criterion given by [4] and [6] can be applied to the dispersion
relation (25) to determine the nature of the instability. To do that, it is necessary to
perform a spatio-temporal stability analysis. The rst task is to compute analytically or
numerically the singularities
0
lying in the complex plane. Then the pairs of eigen-
values which contribute to a zero group velocity, i.e,

(
0
) = 0, at the corresponding
complex wave-number =
0
are identied. Note that this wave-number is associated
with the ray
r
t
= 0. Next, the amplication rate (also called the absolute growth rate)
Convective and Absolute Instabilities 129
at this point is given by
i
= Im(
0
), which characterizes the temporal evolution of the
disturbances. Finally, the criterion for the absolute or convective instability for any given
ow, as outlined by [6] and [4], simply takes the form
(28)
Im(
0
) > 0 = Absolute instability,
Im(
0
) < 0 = Convective instability.
Later, Briggs and Bers showed that the absolute/convective instability criterion given
above is not precise enough. What should be further taken into account is that the
two spatial branches associated with the second-order algebraic branch point
0
must lie
entirely in the upper and lower half planes, when the frequency contour L is located
above all the singularities of the dispersion relation (25). The method of continuation
is then used to deect the original L contour, shown in Figure 9 (a), to the L

contour
shown in Figure 9 (c). The ultimate purpose of this action is to lower the contour just
below the real -axis (to avoid absolute growth, of course). As the contour L in the
plane is lowered to L

, the corresponding branches in the plane will start moving

upwards/downwards. Therefore in order not to violate causality, the F contour in the
plane shown in Figure 9 (b) should also be deformed o the real -axis into F

. Of
course, the simultaneous deformation of the contours L and F into L

and F

(see
Figures 9 (e-f)) must terminate when L

touches () and F

becomes pinched between

the two spatial branches. When spatial branches pinch in the plane they constitute a
saddle point
s
and the mapping of F

pinches the curves in the plane constituting

a pinching frequency
p
. Pinching occurs precisely at the point (
s
,
p
), where the group
velocity

(
s
,
p
) is zero. Correspondingly a cusp appears at
p
in the locus of (),
a feature that can be used to detect the branch point, as demonstrated by [22]. In fact
the inversion in (27) can be made by enclosing the F

contour with semi-circles and

employing the residue theorem, assuming only the rst-order poles (higher-order residues
would be obtained by a limiting process of coalescing rst-order poles). A subsequent
integration in the plane leads to the following time-asymptotic response
(29) G
e
i[
s
(rr
f
)+R
p
R(tt
f
)]
_
2it
D

2
D

1/2

s
,
p
.
We should also bear in mind that there might be cases where the two spatial branches
in the plane of a branch point
p
may be located, for high enough Im(), in the same
half plane. Such pairs (
s
,
p
) do not satisfy the so called pinching criterion, and as
will be discussed later, they cause only an algebraic growth, not an absolute growth rate.
7. Absolute instability results
Our procedure for computing branch points was rst veried by reproducing the results
of [19] for the mixing layer problem, see Figure 8. The results shown here were obtained
by solving the inviscid Rayleigh equation for the normal mode given by Equation (7), as
well as the viscous sixth-order system of equations of order given by Equations (6). In
this section we present these results separately and later show that the high-Reynolds
number viscous modes match onto the inviscid Rayleigh modes.
7.1. Inviscid results. We computed the absolute instability range of the inviscid Ray-
leigh equation as shown in Figure 10 (a), which is the same as the one displayed in [23].
In the interval shown in the gure, the complex , complex and the wave-angle
parameters which constitute the absolutely unstable regime of the inviscid-ow over a
rotating-disk boundary layer are identied. In compliance with the ndings of [23], it is
130 M. Turkyilmazoglu
seen from the gure that for long wavelengths, near the vicinity of the origin, the ow is
neutrally absolutely unstable.
The range of absolute instability which is given by the positive sign of
i
ceases near
the azimuthal wave-number = 0.265. Here, the corresponding eigenvalues are
r
=
-0.070, = 0.34 - i 0.058 and = 38.12
o
, respectively. These points are the upper limits
of the region of absolute instability.
Briggs criterion as described earlier requires a branch point singularity between two
or more spatial branches of the dispersion relation, of which, at least two must lie in
the distinct halves of the complex plane when
i
is suciently large and positive. To
demonstrate this feature, we have plotted the two spatial branches in Figure 11 for which
the branches merge in the distinct half- planes, for suciently large positive values of
Im() = constant, when traces a horizontal line in the plane. Figure 11 (b) shows the
progression of the two spatial branches in the plane at = 0.1. The corresponding
i
s
to each gure are 0.045, 0.021, 0.012 (pinching point) and 0.011, respectively. As Im() is
lowered down towards 0.012 from the large values of Im(), there occurs a pinching-point
in the plane at =
p
= 0.023 + i0.012, and a corresponding saddle-point in the
plane at =
s
= 0.24 i0.0680 as shown in the third part of the gure. Since the sign
of
i
is positive, this point gives an absolute growth rate in the ow. The last portion of
the graph shows branch interchanging after pinching has taken place.
7.2. Viscous results. The viscous absolute instability boundaries were calculated in
[23] and [15]. The pinching phenomenon is shown in Figure 11 (a) for R = 524 and
= 0.1. At large positive values of
i
, the branches shown lie in distinct halves of the
plane. At the pinching point,
i
is 0.15, and so this point exhibits absolute instability.
The branches drawn in the gure are for
i
= 5.34, 2.15, 0.15, and 0, respectively. The
corresponding wave-number is 0.20 i0.112, and the wave-angle = 26.6
o
. Note that
the qualitative behavior of the spatial branches is very similar to that for the inviscid
case in Figure 11 (b).
Figure 11 (b) shows the parameter regions of neutral absolute instability in the (R, ),
(R,
r
), (R,
r
), (R,
i
) and (R, ) planes. Inside the curves the imaginary part of
the frequency is positive and thus the particular ow there is absolutely unstable.
Outside, the ow becomes convectively unstable. The critical Reynolds number for the
ow to undergo absolute instability was found to be R = 507.40, and the corresponding
eigenvalues are = 0.135, = (0.217, 0.122), = 17.72 and = 31.84
o
, respectively.
Note that Lingwood (1995) calculated the critical Reynolds number of absolute instability
as 510 and later corrected this value in [24] to 507.30, which is very close to the value
obtained here. We believe that the discrepancy between these two results is due to
the dierent numerical methods employed. Experiments reveal that transition occurs
at approximately R = 513. Below this critical value,
i
can be reduced to zero in the
calculation of the integral given in (27), before the occurrence of pinching, indicating
that the ow is only convectively unstable. As Healey [18] recently demonstrated from
his asymptotic theory of the upper branch neutral modes, the pinch-point displayed in
Figure 11 (b) does not asymptote towards the inviscid neutral curve of Figure 10 (a).
Healey [18] concludes that the branch points of Figure 10 (a) rather have a viscous long-
wave origin.
To make a comparison with the inviscid absolutely unstable region of Figure 10 (a) and
its viscous origin, we have obtained the solutions at a suciently large Reynolds number,
namely R = 15000. The viscous equations in this case result in an absolutely unstable
region as shown in Figure 12 (a). The upper limit in both cases is almost identical,
yielding a value of azimuthal wave-number = 0.265. In Figure 12 (b) the behavior of
Convective and Absolute Instabilities 131
the eigenfunction of the normal velocity component (h) is shown again at R = 15000.
The signicance of the results displayed in Figure 12 is due to its demonstration that the
inviscid branch points have a viscous correspondence.
8. Direct spatial resonance results
The Benney [3] resonance mechanism with modal coalescence can also be observed for
three dimensional disturbances. In fact, the existence of several modes in the rotating-
disk ow, of which at least two have opposite wave energies (family 1 and family 2)
signies modal coalescence. As mentioned in the introduction, when two coalescing modes
originate from waves propagating in the same direction as in the case of a convectively
unstable ow, then the corresponding branch point is not a pinch-point, but only a
double-pole of the dispersion relation. According to linear theory, such disturbances
normally will decay ultimately. However, the short-term algebraic growth associated with
such a double-pole may be decisive and the corresponding potentially large amplitudes
may initiate nonlinearity and so carry the whole system into the nonlinear stages. This
may be particularly important provided that the coalescing modes are linearly neutral.
Consequently, the corresponding local response to the damping rates given as O(
1

i
) can
be so large that it may initiate the nonlinear stages before the exponentially growing
modes, which is implied by the recent numerical simulation results of [10, 11]). Such
resonance cases between coalescing modes occurring in the same plane has been studied
by [21]. He investigated the direct resonance in a plane Poiseuille ow and in a Blasius
boundary layer ow, which are typical examples of convectively unstable ows.
Figure 13 (b) shows such neutral branch point curves, as a function of Reynolds num-
ber. It can be seen in the fourth gure that in the vicinity of a Reynolds number of 445
there occurs direct resonance, in fact it starts at this Reynolds number, where
i
be-
comes almost zero. In order to support this, in Figure 13 (a) we plot two spatial branches
both emerging in the upper half plane. Here, the particular set of parameters are =
0.0387, = (0.184, 0.0) and = (3.25,0), respectively. This point is clearly a bifurcation
point with a real from which several branches emerge. An indication of the achievable
amplitude amplication at this resonance point can be evaluated as the inverse of
i
,
for which
i
is practically zero. Therefore it is suggested that, unlike plane Poiseuille
ow or Blasius boundary layer ow, direct resonance could be physically relevant in the
rotating-disk boundary layer ow. At this Reynolds number of 445 the ow is still in
the laminar region. It should also be mentioned that we did not encounter any other
branch point apart from the above point which seems to be the only point leading to
direct spatial resonance.
9. Conclusions
The linear stability features of the rotating-disk boundary layer ow have been re-
examined. The stability parameters of the stationary as well as non-stationary waves
have been computed and stability diagrams have been produced using a Spectral method
based on a Chebyshev collocation approximation as well as a fourth-order Runge-Kutta
integration technique. The results obtained within this study are totally in line with the
previously found results cited in the references herein. The ndings are briey summa-
rized below.
9.1. Convective instability. It has been found that for positive frequencies there ex-
ists two minimum points on the neutral stability curves. One of these is related to the
inviscid-type instability occurring on the upper branch and the other one is located on
the lower branch associated with the viscous-type instability. For increasing positive
132 M. Turkyilmazoglu
frequencies, the lower-branch curve is more discernible. The lowest critical Reynolds
number on this branch was found to be 64.443 and it occurs at the frequency = 7.9.
The lowest critical Reynolds number on the upper-branch however, occurs at a negative
frequency.
The upper-branch neutral stability of three dimensional disturbances imposed on the
Von K arm ans boundary layer prole has been investigated asymptotically in particular
for the non-stationary cross-ow disturbances. It has been found that at very large
Reynolds numbers, the upper branch for all waves tends asymptotically to a nite value.
The asymptotic analysis of [17] has been extended to include the non-zero frequency
waves for the upper branch instability modes. Travelling waves with negative frequencies
as well as positive frequencies approach one branch of the zero-frequency neutral curve
corresponding to an angle of about 10.6

. The growth rate for a wave with negative

wave-angle has been found to be small compared with that for the waves with positive
wave-angle.
9.2. Absolute instability. We have also investigated the absolute stability character-
istics of the three dimensional boundary-layer ow due to a rotating-disk. The singular-
ities in the dispersion relation have been studied and several results have been obtained.
First, the coalescing modes of spatial and temporal instability waves originate from waves
propagating in opposite directions. This type of branch point singularity is known to be
related to the absolute instability in a specic range of the parameter space. Therefore,
we can conclude that, above a certain critical Reynolds number, which has been found
to be 507.40 in this study, the ow over the rotating-disk becomes absolutely unstable,
causing the disturbances at a xed radial point in space to grow to large amplitudes in
time. Otherwise, below the critical point the ow is convectively unstable, showing that
growing disturbances travel downstream and eventually leave the ow undisturbed.
The second main result from this work is that, the coalescing modes which originate in
the same wave-number plane form a second-order singularity. This kind of singularity
has been found to occur in the laminar regime of the ow at about R = 445. The
existence of such singularities which lead to the direct spatial resonance of the modes
may cause a locally algebraic growth and consequently the initiation of nonlinearity.
Finally the work presented here may be extended in many ways. For example, one
could compare the neutral stability and growth rates for both the stationary and non-
stationary cross-ow vortices with the asymptotic theory of [14] in the case of high
Reynolds numbers (R >> 1). As we have mentioned before, the analysis presented here
is a local analysis in which the non-parallel eects were omitted. How non-parallelism
will aect the absolute instability regime determined from this investigation requires
further work and moreover requires the numerical treatment of the full linearized stability
equations outlined in this report. Therefore, a more global analysis which makes use of
non-parallel eects should also be considered.
Convective and Absolute Instabilities 133
Figure 1. Von-Karmans Velocity Distribution in the Rotating-disk Flow
(a) Von-Karmans self-similarity solutions corresponding to the basic eld of the
rotating-disk ow; cross-ow velocity (F), azimuthal velocity (G) and normal velocity
to the disk (H),
(b) Shapes of the combined velocity (F +G) for selected Reynolds numbers. Left:
Stationary ( = 0) cross-ow velocity proles starting from the upper branch; (..) R =
15000 and () R = 286.045, to the lower branch; (--) R = 453.755 and ( ) R = 15000.
Right: Non-stationary proles of = 7.9 for the same conditions,
(c) Neutral stability curves in the (R, ) plane, and
(d) In the (R, ) plane; (..) = 5, (--) = 0, ( ) = 4, () = 7.9 and ()
= 10.
(a) (b)
10
0
10
1
10
2
10
3
10
4
10
5
10
6
10
7
R
10
-7
10
-6
10
-5
10
-4
10
-3
10
-2
10
-1
10
0

10
0
10
1
10
2
10
3
10
4
10
5
10
6
10
7
10
-7
10
-6
10
-5
10
-4
10
-3
10
-2
10
-1
10
0
10
0
10
1
10
2
10
3
10
4
10
5
10
6
10
7
10
-7
10
-6
10
-5
10
-4
10
-3
10
-2
10
-1
10
0
10
0
10
1
10
2
10
3
10
4
10
5
10
6
10
7
10
-7
10
-6
10
-5
10
-4
10
-3
10
-2
10
-1
10
0
10
0
10
1
10
2
10
3
10
4
10
5
10
6
10
7
10
-7
10
-6
10
-5
10
-4
10
-3
10
-2
10
-1
10
0
10
0
10
1
10
2
10
3
10
4
10
5
10
6
10
7
10
-7
10
-6
10
-5
10
-4
10
-3
10
-2
10
-1
10
0
10
0
10
1
10
2
10
3
10
4
10
5
10
6
10
7
10
-7
10
-6
10
-5
10
-4
10
-3
10
-2
10
-1
10
0
10
0
10
1
10
2
10
3
10
4
10
5
10
6
10
7
10
-7
10
-6
10
-5
10
-4
10
-3
10
-2
10
-1
10
0
10
0
10
1
10
2
10
3
10
4
10
5
10
6
10
7
10
-7
10
-6
10
-5
10
-4
10
-3
10
-2
10
-1
10
0
10
0
10
1
10
2
10
3
10
4
10
5
10
6
10
7
10
-7
10
-6
10
-5
10
-4
10
-3
10
-2
10
-1
10
0
10
0
10
1
10
2
10
3
10
4
10
5
10
6
10
7
10
-7
10
-6
10
-5
10
-4
10
-3
10
-2
10
-1
10
0
10
0
10
1
10
2
10
3
10
4
10
5
10
6
10
7
10
-7
10
-6
10
-5
10
-4
10
-3
10
-2
10
-1
10
0
10
0
10
1
10
2
10
3
10
4
10
5
10
6
10
7
10
-7
10
-6
10
-5
10
-4
10
-3
10
-2
10
-1
10
0
10
0
10
1
10
2
10
3
10
4
10
5
10
6
10
7
10
-7
10
-6
10
-5
10
-4
10
-3
10
-2
10
-1
10
0
10
0
10
1
10
2
10
3
10
4
10
5
10
6
10
7
10
-7
10
-6
10
-5
10
-4
10
-3
10
-2
10
-1
10
0
(c)
0 100 200 300 400 500 600 700 800 900 1000 1100
R
-38
-34.2
-30.4
-26.6
-22.8
-19
-15.2
-11.4
-7.6
-3.8
0
3.8
7.6
11.4
15.2
19
22.8
26.6
30.4
34.2
38
0 100 200 300 400 500 600 700 800 900 1000 1100
-38
-34.2
-30.4
-26.6
-22.8
-19
-15.2
-11.4
-7.6
-3.8
0
3.8
7.6
11.4
15.2
19
22.8
26.6
30.4
34.2
38
0 100 200 300 400 500 600 700 800 900 1000 1100
-38
-34.2
-30.4
-26.6
-22.8
-19
-15.2
-11.4
-7.6
-3.8
0
3.8
7.6
11.4
15.2
19
22.8
26.6
30.4
34.2
38
0 100 200 300 400 500 600 700 800 900 1000 1100
-38
-34.2
-30.4
-26.6
-22.8
-19
-15.2
-11.4
-7.6
-3.8
0
3.8
7.6
11.4
15.2
19
22.8
26.6
30.4
34.2
38
0 100 200 300 400 500 600 700 800 900 1000 1100
-38
-34.2
-30.4
-26.6
-22.8
-19
-15.2
-11.4
-7.6
-3.8
0
3.8
7.6
11.4
15.2
19
22.8
26.6
30.4
34.2
38

(d)
134 M. Turkyilmazoglu
Figure 2. Stability Curves in the (R, ), (R, ) and (R, ) Planes for = 0
and = 4 Waves
(a-c): Stability curves for = 0 waves,
(d-f): Stability curves for = 4 waves.
0 100 200 300 400 500 600 700 800 900 1000
R
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
0

-
i
=0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
(a)
0 100 200 300 400 500 600 700 800 900 1000
R
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
0.2

-
i
=0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
(b)
0 100 200 300 400 500 600 700 800 900 1000
R
0
2.5
5
7.5
10
12.5
15
17.5
20
22.5
25
27.5

-
i
=0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
(c)
0 100 200 300 400 500 600 700 800 900 1000
R
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
0

0 100 200 300 400 500 600 700 800 900 1000
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
0
0 100 200 300 400 500 600 700 800 900 1000
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
0
0.005
0.01
0.03
0.05
0.07
0.09
0.11
-
i
=0
(d)
0 100 200 300 400 500 600 700 800 900 1000
R
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
0.2
0 100 200 300 400 500 600 700 800 900 1000
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
0.2
0 100 200 300 400 500 600 700 800 900 1000
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
0.2

0.005
0.01
0.03
0.05
0.07
0.09
0.11
-
i
=0
(e)
0 100 200 300 400 500 600 700 800 900 1000
R
0
2
4
6
8
10
12
14
16
18
0 100 200 300 400 500 600 700 800 900 1000
0
2
4
6
8
10
12
14
16
18
0 100 200 300 400 500 600 700 800 900 1000
0
2
4
6
8
10
12
14
16
18
0 100 200 300 400 500 600 700 800 900 1000
0
2
4
6
8
10
12
14
16
18

-
i
=0
.005
.01
.03
.05 .07 .09 .11
(f)
Convective and Absolute Instabilities 135
Figure 3. Growth Rate Curves for Zero-frequency and Frequency-10 Waves
(a-b) Distribution of the spatial amplication rate of zero-frequency, family 1 normal
modes as a function of wave-angle at a Reynolds number of 1000 for zero and frequency
10 modes, respectively,
(c-d) Maximum spatial amplication rate
i
, radial wave-number
r
, azimuthal
wave-number and wave-angle are shown against frequency at a Reynolds number of
500.
10 13.2 16.4 19.6 22.8 26

0
0.04
0.08
0.12
0.16
0.2
-

i
10 13.2 16.4 19.6 22.8 26
0
0.04
0.08
0.12
0.16
0.2
600
REY=1000
(a)
-35 -25.6 -16.2 -6.8 2.6 12

0
0.03
0.06
0.09
0.12
0.15
-35 -25.6 -16.2 -6.8 2.6 12
0
0.03
0.06
0.09
0.12
0.15
-35 -25.6 -16.2 -6.8 2.6 12
0
0.03
0.06
0.09
0.12
0.15
-

i
450
600
REY=1000
(b)
-24 -12 0 12 24

0
0.1
0.2
0.3
0.4
-24 -12 0 12 24
0
0.1
0.2
0.3
0.4
-24 -12 0 12 24
0
0.1
0.2
0.3
0.4

-
i
(c)
-24 -12 0 12 24

0
8
16
24
32

(d)
136 M. Turkyilmazoglu
Figure 4. Spatial Branches of Zero-Frequency Waves at R = 400, 1000 and
15000
Emerging branches of three families of stationary waves at R = 400 for:
(a-c) Zero frequency waves,
(d-f) 7.9 frequency waves.
0 0.3 0.6 0.9
wavenumber (
r
)
-0.2
0
0.2
0.4
0.6
g
r
o
w
t
h
r
a
t
e

(
-

i
)

0 0.3 0.6 0.9

wavenumber (
r
)
-0.2
0
0.2
0.4
0.6
g
r
o
w
t
h
r
a
t
e

(
-

i
)

0 0.3 0.6 0.9

wavenumber (
r
)
-0.2
0
0.2
0.4
0.6
g
r
o
w
t
h
r
a
t
e

(
-

i
)

REY=400
1
2
3
(a)
0 0.04 0.08 0.12 0.16 0.2
wavenumber ()
-0.2
0
0.2
0.4
0.6
g
r
o
w
t
h
r
a
t
e

(
-

i
)

0 0.04 0.08 0.12 0.16 0.2

wavenumber ()
-0.2
0
0.2
0.4
0.6
g
r
o
w
t
h
r
a
t
e

(
-

i
)

0 0.04 0.08 0.12 0.16 0.2

wavenumber ()
-0.2
0
0.2
0.4
0.6
g
r
o
w
t
h
r
a
t
e

(
-

i
)

REY=400
1
2
3
(b)
0 0.04 0.08 0.12 0.16 0.2
wavenumber ()
0
20
40
60
80
100
w
a
v
e

a
n
g
l
e

(

0 0.04 0.08 0.12 0.16 0.2

wavenumber ()
0
20
40
60
80
100
w
a
v
e

a
n
g
l
e

(

0 0.04 0.08 0.12 0.16 0.2

wavenumber ()
0
20
40
60
80
100
w
a
v
e

a
n
g
l
e

(

REY=400
1
2
3
(c)
0 0.3 0.6 0.9
wavenumber (
r
)
-0.2
0
0.2
0.4
0.6
g
r
o
w
t
h
r
a
t
e

(
-

i
)

0 0.3 0.6 0.9

wavenumber (
r
)
-0.2
0
0.2
0.4
0.6
g
r
o
w
t
h
r
a
t
e

(
-

i
)

0 0.3 0.6 0.9

wavenumber (
r
)
-0.2
0
0.2
0.4
0.6
g
r
o
w
t
h
r
a
t
e

(
-

i
)

REY=400
1
2
3
(d)
0 0.04 0.08 0.12 0.16 0.2
wavenumber ()
-0.2
0
0.2
0.4
0.6
g
r
o
w
t
h
r
a
t
e

(
-

i
)

0 0.04 0.08 0.12 0.16 0.2

wavenumber ()
-0.2
0
0.2
0.4
0.6
g
r
o
w
t
h
r
a
t
e

(
-

i
)

0 0.04 0.08 0.12 0.16 0.2

wavenumber ()
-0.2
0
0.2
0.4
0.6
g
r
o
w
t
h
r
a
t
e

(
-

i
)

REY=400
1
2
3
(e)
0 0.04 0.08 0.12 0.16 0.2
wavenumber ()
0
20
40
60
80
100
w
a
v
e

a
n
g
l
e

(

0 0.04 0.08 0.12 0.16 0.2

wavenumber ()
0
20
40
60
80
100
w
a
v
e

a
n
g
l
e

(

0 0.04 0.08 0.12 0.16 0.2

wavenumber ()
0
20
40
60
80
100
w
a
v
e

a
n
g
l
e

(

REY=400
1
2
3
(f)
Convective and Absolute Instabilities 137
Figure 5. Isosurfaces for = 0 Waves
(a-b) Contour-lines for = 0 at the critical eigenvalues R = 286.045, = 0.38407,
= 0.07745, and
(c-d) Contour-lines for = 7.9 at the critical eigenvalues R = 316.601, = 0.42108,
= 0.03933.
0 1.796 3.591 5.387 7.183 8.979 10.77 12.57
r+ R
0
4
8
12
16
20
z
(a)
0 1.796 3.591 5.387 7.183 8.979 10.77 12.57
r+ R
0
4
8
12
16
20
z
(b)
0 1.796 3.591 5.387 7.183 8.979 10.77 12.57
r+ R
0
4
8
12
16
20
z
(c)
0 1.796 3.591 5.387 7.183 8.979 10.77 12.57
r+ R
0
4
8
12
16
20
z
(d)
138 M. Turkyilmazoglu
Figure 6. Stationary Cross-ow Vortices in the Rotating Boundary Layer
Flow
(a-b) Stationary cross-ow vortex structure is at the critical eigenvalues (a) R =
286.045, = 0.38407, = 0.07745; (b) R = 453.755, and
(c-d) Travelling cross-ow vortex structure ( = 7.9) is at the critical eigenvalues (c)
R = 316.601, = 0.42108, = 0.03933, (d) R = 64.443, = 0.27597, = 0.10669.
0 2.095 4.19 6.285 8.38 10.48 12.57
r+ R
0
5
10
z
(a)
0 2.095 4.19 6.285 8.38 10.48 12.57
r+ R
0
10
20
z
(b)
0 2.095 4.19 6.285 8.38 10.48 12.57
r+ R
0
5
10
z
(c)
0 2.095 4.19 6.285 8.38 10.48 12.57
r+ R
0
10
20
z
(d)
Convective and Absolute Instabilities 139
Figure 7. Upper Branch Modes from Asymptotic Approach
A comparison of the numerical and asymptotic calculations of the stationary and
non-stationary -- ( = 5) and = 10 waves, in the (R, ) and (R, ) planes. In the
gures the long curves show numerical results and the shorter lines asymptotic ones.
10
0
10
1
10
2
10
3
10
4
10
5
10
6
10
7
R
10
-2
10
-1.41
10
-0.81
10
-0.22

10
0
10
1
10
2
10
3
10
4
10
5
10
6
10
7
R
10
-2
10
-1.41
10
-0.81
10
-0.22

10
0
10
1
10
2
10
3
10
4
10
5
10
6
10
7
R
10
-2
10
-1.41
10
-0.81
10
-0.22

10
0
10
1
10
2
10
3
10
4
10
5
10
6
10
7
R
10
-2
10
-1.41
10
-0.81
10
-0.22

10
0
10
1
10
2
10
3
10
4
10
5
10
6
10
7
R
10
-2
10
-1.41
10
-0.81
10
-0.22

10
0
10
1
10
2
10
3
10
4
10
5
10
6
10
7
R
10
-2
10
-1.41
10
-0.81
10
-0.22

(a)
10
1
10
2
10
3
10
4
10
5
10
6
10
7
R
-10
-3
4
11
18
25

10
1
10
2
10
3
10
4
10
5
10
6
10
7
R
-10
-3
4
11
18
25

10
1
10
2
10
3
10
4
10
5
10
6
10
7
R
-10
-3
4
11
18
25

10
1
10
2
10
3
10
4
10
5
10
6
10
7
R
-10
-3
4
11
18
25

10
1
10
2
10
3
10
4
10
5
10
6
10
7
R
-10
-3
4
11
18
25

10
1
10
2
10
3
10
4
10
5
10
6
10
7
R
-10
-3
4
11
18
25

(b)
Figure 8. Absolute Instability Range in the Mixing Layer
Absolutely and convectively unstable range and branch points in the mixing layer
problem are demonstrated.
0 0.05 0.1 0.15 0.2 0.25

r
-0.1
-0.05
0
0.05
0.1
0.15
0.2
0.25

increasing R
(a)
0 0.5 1 1.5 2 2.5
R
-0.3
-0.19
-0.08
0.03
0.14
0.25

0 0.5 1 1.5 2 2.5

R
-0.3
-0.19
-0.08
0.03
0.14
0.25

0 0.5 1 1.5 2 2.5

R
-0.3
-0.19
-0.08
0.03
0.14
0.25

0 0.5 1 1.5 2 2.5

R
-0.3
-0.19
-0.08
0.03
0.14
0.25

i
(b)
140 M. Turkyilmazoglu
Figure 9. A Sketch of the Instability Branches and the Impulse Response
Sketches of the typical temporal (a), (c), (e) and spatial (b), (d), (f) branches are
demonstrated. The physical impulse response is shown corresponding to
(g) convectively unstable ow, and
(h) absolutely unstable ow.
L
L
L
F
F
F
t<0 t>0
t<0
t>0
t<0 t>0

r
i
r
r

i
r
r
i
i
i
r

( )
( ) F
( ) F
F
( ) L
L ( )
L
L
( )
( )
L
L
( )
( )
r>r
f
r<r
f
r>r
f
r<r
f
r>r f
r<r
f
i
(a) (b)
(c) (d)
(e)
(f)
t
r
t
r

i
i
i
i
>0
<0
>0
<0
(g) (h)
Convective and Absolute Instabilities 141
Figure 10. Absolute Instability Range and Pinching Phenomenon in the
Inviscid Rotating-disk Flow.
(a) Plot showing the locus of the branch-points (, ) against . Variation of wave-
angle is also given. Data is taken from the solution of the inviscid Rayleigh equation
(7) and claries the range where the rotating-disk ow becomes inviscidly absolutely
unstable,
(b) Demonstrating the progression of the two spatial branches in the plane at = 0.1
as traces a horizontal line, in the inviscid rotating-disk ow. Figures are respectively for

i
= 0.045,
i
= 0.021,
i
= 0.012 (pinching point),
i
= 0.011. Pinching takes place in
the third portion and clearly demonstrates a true pinching in the sense that Briggs-Bers
criterion is satised. The direction of the arrows indicates increasing frequency.
0 0.05 0.1 0.15 0.2 0.25 0.3

-0.08
-0.03
0.02

0 0.05 0.1 0.15 0.2 0.25 0.3

-0.08
-0.03
0.02

r
0 0.05 0.1 0.15 0.2 0.25 0.3

-0.2
0.1
0.4

0 0.05 0.1 0.15 0.2 0.25 0.3

-0.2
0.1
0.4

i
0 0.05 0.1 0.15 0.2 0.25 0.3

10
40

0 0.05 0.1 0.15 0.2 0.25 0.3

10
40

(a)
0 0.1 0.2 0.3 0.4 0.5

r
-0.5
0.5

0 0.1 0.2 0.3 0.4 0.5

r
-0.5
0.5

0 0.1 0.2 0.3 0.4 0.5

r
-0.5
0.5

0 0.1 0.2 0.3 0.4 0.5

r
-0.5
0.5

0 0.1 0.2 0.3 0.4 0.5

r
-0.5
0.5

0 0.1 0.2 0.3 0.4 0.5

r
-0.5
0.5

0 0.1 0.2 0.3 0.4 0.5

r
-0.5
0.5

0 0.1 0.2 0.3 0.4 0.5

r
-0.5
0.5

(b)
142 M. Turkyilmazoglu
Figure 11. Pinching Phenomenon and Neutral Absolute Instability Curves
in the Viscous Rotating-disk Flow
(a) Pinching phenomena in the viscous rotating-disk boundary layer ow. Branch
point at = 0.1, = (0.20, 0.112), = (16.69, 0.15) and R = 524 (third portion).
Graphs are for
i
= 5.34, 2.15, 0.15, 0, respectively. Pinching requirements are satised.
The direction of the arrows indicates increasing frequency,
(b) Neutral absolute instability curves dening the region of absolute instability in
the viscous rotating-disk boundary-layer ow are shown in the (R, ), (R,
r
), (R,
r
),
(R,
i
) and (R, ) planes, respectively. For all the curves, inside the loop signies absolute
instability, while outside we have convective stability/instability.
0 0.1 0.2 0.3 0.4 0.5 0.6

r
-0.4
0.4

0 0.1 0.2 0.3 0.4 0.5 0.6

r
-0.4
0.4

0 0.1 0.2 0.3 0.4 0.5 0.6

r
-0.4
0.4

0 0.1 0.2 0.3 0.4 0.5 0.6

r
-0.4
0.4

0 0.1 0.2 0.3 0.4 0.5 0.6

r
-0.4
0.4

0 0.1 0.2 0.3 0.4 0.5 0.6

r
-0.4
0.4

0 0.1 0.2 0.3 0.4 0.5 0.6

r
-0.4
0.4

0 0.1 0.2 0.3 0.4 0.5 0.6

r
-0.4
0.4

0 0.1 0.2 0.3 0.4 0.5 0.6

r
-0.4
0.4

0 0.1 0.2 0.3 0.4 0.5 0.6

r
-0.4
0.4

0 0.1 0.2 0.3 0.4 0.5 0.6

r
-0.4
0.4

0 0.1 0.2 0.3 0.4 0.5 0.6

r
-0.4
0.4

0 0.1 0.2 0.3 0.4 0.5 0.6

r
-0.4
0.4

0 0.1 0.2 0.3 0.4 0.5 0.6

r
-0.4
0.4

0 0.1 0.2 0.3 0.4 0.5 0.6

r
-0.4
0.4

0 0.1 0.2 0.3 0.4 0.5 0.6

r
-0.4
0.4

(a)
500 600 700 800 900 1000
0.04
0.25

500 600 700 800 900 1000

-87
0

500 600 700 800 900 1000

0.13
0.24

500 600 700 800 900 1000

-0.22
-0.08

500 600 700 800 900 1000

REYNOLDS NUMBER
15
54

(b)
Convective and Absolute Instabilities 143
Figure 12. Absolute Instability Range and Eigenfunction h at R = 15000
(a) Locus of the branch points (, ) are given as varies. Variation in is also given.
Data is taken from the solution of the full sixth-order viscous equations (6) for R = 15000
and reveals the range where the rotating-disk ow becomes viscously absolutely unstable,
(b) Eigenfunctions of the normal velocity component (h) are shown in the viscous
rotating-disk boundary ow layer at R = 15000.
0 0.05 0.1 0.15 0.2 0.25 0.3

-1250
-540
170

0 0.05 0.1 0.15 0.2 0.25 0.3

-1250
-540
170

r
0 0.05 0.1 0.15 0.2 0.25 0.3

-0.2
0.1
0.4

0 0.05 0.1 0.15 0.2 0.25 0.3

-0.2
0.1
0.4

i
0 0.05 0.1 0.15 0.2 0.25 0.3

10
40

(a) (b)
144 M. Turkyilmazoglu
Figure 13. Pinching Phenomena and Neutral Branch Points in the Viscous
Rotating-disk Flow
(a) The progression of the two spatial branches at R = 445 and = 0.0387, in
the plane is given in the viscous rotating-disk boundary layer ow. Graphs are for

i
= 5, 2, 0.25, 0 and -0.02, respectively. Branch point is at = (0.184,0), = (3.25,0)
(fourth portion). A direct spatial resonance occurs between the two spatial branches since
the corresponding and are neutral. The direction of the arrows indicates increasing
frequency,
(b) Neutral branch points of two merging branches from the same plane are shown.
0.1 0.15 0.2 0.25 0.3 0.35 0.4

r
0
0.2

0.1 0.15 0.2 0.25 0.3 0.35 0.4

r
0
0.2

0.1 0.15 0.2 0.25 0.3 0.35 0.4

r
0
0.1

0.1 0.15 0.2 0.25 0.3 0.35 0.4

r
0
0.1

0.1 0.15 0.2 0.25 0.3 0.35 0.4

r
-0.06
0.06

0.1 0.15 0.2 0.25 0.3 0.35 0.4

r
-0.06
0.06

0.1 0.15 0.2 0.25 0.3 0.35 0.4

r
-0.06
0.06

0.1 0.15 0.2 0.25 0.3 0.35 0.4

r
-0.06
0.06

0.1 0.15 0.2 0.25 0.3 0.35 0.4

r
-0.06
0.06

0.1 0.15 0.2 0.25 0.3 0.35 0.4

r
-0.06
0.06

(a)
400 430 460 490 520 550
0
0.1

400 430 460 490 520 550

-5
15

400 430 460 490 520 550

0.15
0.22

400 430 460 490 520 550

-0.01
0.01

400 430 460 490 520 550

REYNOLDS NUMBER
0
20

(b)
Convective and Absolute Instabilities 145
Appendix A
The quantities arising from the upper branch asymptotic analysis in 1.5.5. are:
k
c
=
_

0
r
1

B
[
F

F

U

B

U

B
] +
0

U
2
B
_
=

.
I
1
=
_

0
w
2
0
d,
I
2
=
0
_

0
1

U
B
[F

U
B
]w
2
0
d,
I
3
=
_

0

U
2
B
w
2
0
d,
a = i
0
w
2
0
(

)sgn[

B
(

)]
_
1
|

B
|
[F

B
]
_
=

,
b = iw
2
0
(

)sgn[

B
(

)]
_
1

B
1
|

B
|

B
_
=

.
References
[1] Balakumar, P. and Malik, M. R. Travelling disturbances in rotating-disk ow, Theoret. Com-
put. Fluid Dyn. 2, 125137, 1990.
[2] Bassom, A. P. and Gajjar, J. S. B. Non-stationary cross-ow vortices in a three-dimensional
boundary layer, Proc. Roy. Soc. London Ser. A 417, 179212, 1988.
[3] Benney, D. J. and Gustavsson, L. H. A new mechanism for linear and non-linear hydrody-
namic instability, Stud. Appl. Math. 64, 185209, 1981.
[4] Bers, A. Linear waves and instabilities, Physique des Plasmas., 117225, 1975.
[5] Betchov, R. and Criminale, W. O. Spatial instability of the inviscid jet and wake, Phys.
Fluids. 9, 359362, 1966.
[6] Briggs, R. J. Electron-Stream Interaction With Plasmas (MIT Press, 1964).
[7] Cole, J. W. Hdrodynamic stability of compressible ows, PhD thesis (University of Exeter,
1995).
[8] Cooper, A. J. and Carpenter, P. W. The stability of rotating-disk boundary-layer ow over
a compliant wall. part i. type i and ii instabilities, J. Fluid Mech. 350, 231259, 1997.
[9] Cooper, A. J. and Carpenter, P. W. The stability of rotating-disk boundary-layer ow over
a compliant wall. part ii. absolute instability, J. Fluid Mech. 350, 261270, 1997.
[10] Davies, C. and Carpenter, P. W. A novel velocity-vorticity formulation of the navier-stokes
equations with application to boundary layer disturbance evolution, J. Comput. Phys. 172,
119165, 2001.
[11] Davies, C. and Carpenter, P. W. Global behaviour corresponding to the absolute instability
of the rotating-disc boundary layer, J. Fluid Mech. 486, 287329, 2003.
[12] Faller, A. J. and Kaylor, R. E. A numerical study of the instability of the laminar ekman
boundary layer, J. Atm. Sci. 23, 466480, 1966.
[13] Federov, B. I., Plavnik, G. Z., Prokhorov, I. V. and Zhukhovitskii, L. G. Transitional ow
conditions on a rotating-disk, J. Eng. Phys. 31, 14481453, 1976.
[14] Gajjar, J. S. B. Nonlinear evolution of a 1st mode oblique wave in a compressible boundary-
layer. Part 1. Heated cooled walls, IMA Journal of Applied Mathematics 53, 221248, 1994.
[15] Jasmine, H. A. and Gajjar, J. S. B. Convective and absolute instability in the incompressible
boundary layer on a rotating disk in the presence of a uniform magnetic eld, Journal of
Engineering Mathematics 52, 337353, 2005.
[16] Gregory, N., Stuart, J. T. and Walker, W. S. On the stability of three-dimensional boundary
layers with applications to the ow due to a rotating-disk, Philos. Trans. R. Soc. London
Ser. A 248, 155199, 1955.
[17] Hall, P. An asymptotic investigation of the stationary modes of instability of the boundary
layer on a rotating-disk, Proc. Roy. Soc. London Ser. A 406, 93106, 1986.
146 M. Turkyilmazoglu
[18] Healey, J. J. On the relation between inviscid and viscous absolute instability of the rotating-
disk boundary layer, J. Fluid Mech. 511, 179199, 2004.
[19] Huerre, P. and Monkewitz, P. A. Absolute and convective instabilities in free shear layers,
J. Fluid Mech. 159, 151168, 1985.
[20] Huerre, P. and Monkewitz, P. A. Local and global instabilities in spatially developing ows,
Ann. Rev. Fluid Mech. 22, 473537, 1990.
[21] Koch, W. Direct resonance in orr-sommerfeld equation, Acta Mech. 58, 1129, 1986.
[22] Kupfer, K., Bers, A. and Ram, A. K. The cusp map in the complex-frequency plane for
absolute instabilities, Phys. Fluids. 30, 30753082, 1987.
[23] Lingwood, R. J. Absolute instability of the boundary layer on a rotating-disk, J. Fluid Mech.
299, 1733, 1995.
[24] Lingwood, R. J. On the application of the briggs and steepest-descent method to a boundary-
layer ow, Stud. Appl. Math. 98, 213254, 1997.
[25] Mackerrel, S. O. A nonlinear asymptotic investigation of the stationary modes of instability
of the 3-dimensional boundary layer on a rotating-disk, Proc. Roy. Soc. London Ser. A 413,
497513, 1987.
[26] Malik, M. R. The neutral curve for stationary disturbances in rotating-disk ow, J. Fluid
Mech. 164, 275287, 1986.
[27] Malik, M. R. and Poll, D. I. A. Eect of curvature on three-dimensional boundary layer
stability, AIAA Journal 23, 13621369, 1985.
[28] Malik, M. R., Zang, T. A. and Hussaini, M. Y. A spectral collocation method for the navier-
stokes equations, J. Comput. Phys. 61, 6488, 1984.
[29] Monkewitz, P. A. The absolute and convective nature of instability in two-dimensional wakes
at low reynolds numbers, Phys. Fluids. 31, 9991006, 1988.
[30] Pier, B. Finite-amplitude crossow vortices, secondary instability and transition in the
rotating-disk boundary layer, J. Fluid Mech. 487, 315343, 2003.
[31] Shanthini, R. Degeneracies of the temporal orr-sommerfeld eigenmodes in plane poiseuille
ow, J. Fluid Mech. 201, 1334, 1989.
[32] Turkyilmazoglu, M. and Gajjar, J. S. B. On the absolute instability of the attachment-line
and swept-hiemenz boundary layers, Theoret. Comput. Fluid Dyn. 13, 5775, 1999.
[33] Turkyilmazoglu, M., Gajjar, J. S. B. and Ruban, A. I. The absolute instability of thin wakes
in an incompressible/compressible uid, Theoret. Comput. Fluid Dyn. 13, 91114, 1999.
[34] Wilkinson, S. P. and Malik, M. R. Stability experiments in the ow over a rotating-disk,
AIAA Journal 23, 588595, 1985.