You are on page 1of 11

Comput. Methods Appl. Mech. Engrg.

192 (2003) 2993–3003


www.elsevier.com/locate/cma

A steady-state solver for the simulation of Taylor vortices


in spherical annular flow
a,b
V.C. Loukopoulos , G.A. Katsiaris a, G.T. Karahalios a,b,*

a
Department of Physics, University of Patras, Patras 26500, Greece
b
Hellenic Open University, Sahturi 16, Patras 26222, Greece
Received 10 October 2002; received in revised form 24 March 2003

Abstract

A numerical technique is described by which the solution of the equations of steady axisymmetric flow of a viscous
incompressible fluid in any orthogonal system of coordinates can be obtained. In the present analysis the equations of
motion and the boundary conditions are approximated by finite differences that are of the second-order accuracy with
respect to the grid size. This technique is used for the numerical solution of the equations of steady motion in spherical
Couette flow, in order to obtain Taylor vortices symmetric and asymmetric at a prescribed medium and large annular
gap and given range of the Reynolds number of the flow. The computed results are compared to results available from
previous experimental and theoretical studies.
Ó 2003 Elsevier B.V. All rights reserved.
Keywords: Finite differences methods; Taylor vortices; Annular flow

1. Introduction

The present work consists of two parts. The first is the development of a finite difference technique of
obtaining an approximate numerical solution of the Navier–Stokes equations of an axisymmetric viscous
flow. The second is the application of this technique in order to obtain asymmetric Taylor vortices in
annular spherical flows. Regarding the first part, in such methods usually all derivatives are approximated
by central differences, this implying a second-order accuracy. Their efficiency relies on the rate of con-
vergence of the iterative procedures by which the relevant system of algebraic equations is solved to the
adopted criterion of accuracy. With regards to the second part, the usual way of obtaining Taylor vortices
in annular spherical flow is the following. A time-stepping code is adapted on the time-dependent form of
the Navier–Stokes equations of motion until the desired state is reached. The motivation for this study was
to obtain steady asymmetric Taylor vortices via the numerical integration of the equations of the steady
flow and not via the time-dependent flow. The efficiency of the method is monitored by the computed

*
Corresponding author. Address: Department of Physics, University of Patras, Patras 26500, Greece.
E-mail address: karaha@physics.upatras.gr (G.T. Karahalios).

0045-7825/03/$ - see front matter Ó 2003 Elsevier B.V. All rights reserved.
doi:10.1016/S0045-7825(03)00314-1
2994 V.C. Loukopoulos et al. / Comput. Methods Appl. Mech. Engrg. 192 (2003) 2993–3003

results that are in excellent agreement to similar results obtained before by other investigators via the time-
stepping code.
We start with the equations of steady axisymmetric flow in Cartesian coordinates x, y and z. Taking z as
the axis of symmetry so that the derivatives with respect to z are zero, the equations of motion in the x- and
y-direction are
ou ou 1 op
u þv ¼ þ mr2 u;
ox oy q ox

ov ov 1 op
u þv ¼ þ mr2 v;
ox oy q oy
together with the continuity equation
ou ov
þ ¼ 0:
ox oy
Introducing the stream function f , such that u ¼ ðof =oyÞ, v ¼ of =ox, the continuity equation is identi-
cally satisfied. Next, by cross-differentiating the equations of motion to rid off the pressure terms and using
the previous notation in suitable non-dimensional form, we deduce the equation of motion in terms of the
third component of the vorticity x ¼ ðn; n; fÞ, as follows:
 2 
o2 f of of o f of of
 Re þ Re þ ¼0 ð1Þ
ox2 oy ox oy 2 ox oy
in which x ¼ $  q is the vorticity, q ¼ ðu; v; wÞ is the fluid velocity and Re is the Reynolds number of the
flow. Accordingly if the partial derivatives of =ox and of =oy are replaced by the functions kðx; yÞ and gðx; yÞ
respectively and a square mesh of size h is used so that the f-derivatives are replaced by their finite difference
approximation, Eq. (1) takes the form
         
1 1 1 1 1 1 1
 g0 h f 1 þ 1 þ k 0 h f 2 þ þ g0 h f 3 þ 1  k 0 h f 4 ¼ 2 1 þ f ð2Þ
Re 2 2 Re 2 2 Re 0
and terms of Oðh4 Þ have been neglected. However, the inherent error due to the finite difference approxi-
mation can be serious if kðx; yÞ and gðx; yÞ are either large or vary rapidly, except on the case of a very fine
mesh. This deficiency was overcome by Allen and Southwell [1] by the replacement of the previously re-
ferred functions in the neighborhood of 0 by their nodal values k0 and g0 and by the subsequent split of the
resulting equation into two, as follows:
o2 f of
 Reg0 ¼ Aðx; yÞRe; ð3Þ
ox2 ox

o2 f of
2
þ k0 ¼ Aðx; yÞ; ð4Þ
oy oy
where A is a function of the coordinates ðx; yÞ.
Accordingly Eqs. (3) and (4) were solved independently and each solution was used as expansion along
the line 301 and 402 respectively, Fig. 1. With this process the final approximate expression
g0 h½ðf1  f0 Þeg0 hRe þ f3  f0 ð1  ek0 h Þ ¼ f0 h½ðf2  f0 Þek0 h þ f4  f0 ð1  eg0 hRe Þ ð5Þ
was obtained, which was found to be a superior nodal approximation over (2), its superiority lying in the
approximation it makes to the first derivative terms in (1). The same algorithm (5) was used by Wood [2] for
reducing the fourth-order Eq. (1) to a lower-order system of equations in Cartesian coordinates. It was also
V.C. Loukopoulos et al. / Comput. Methods Appl. Mech. Engrg. 192 (2003) 2993–3003 2995

β = constant
2

0 1
3
h

4
α = constant

Fig. 1. Notation.

employed by Dennis and Hudson [3] for two-dimensional flow in Cartesian coordinates, in which case the
functions kðx; yÞ and gðx; yÞ represent velocity components and by Dennis [4] in the calculation of the steady
flow through a curved tube using cylindrical coordinates.
In the present paper the previous method is extended and can be applied to the study of a steady axi-
symmetric flow in any orthogonal system of coordinates. The description is developed in brief in Section 2,
where, initially, a relation analogous to (5) is obtained. However, the matrix associated with this as well as
with (5) is not necessarily diagonally dominant, a prerequisite for the convergence of the iterative scheme.
To overcome this obstacle, the velocity components and the exponentials appearing in a relation analogous
to (5) are expanded as a Taylor series, without loosing the order of accuracy of the truncation error. In this
way a set of linear algebraic Eqs. (13)–(15) is obtained that can be solved using an iterative method, such as
the relaxation methods are.
Next in Section 3 the above mentioned process is employed to investigate the spherical Couette flow, i.e.
the flow induced in a spherical shell by rotating the inner sphere and holding the outer at rest. This type of
flow is governed by the ratio r ¼ ðro  ri Þ=ri and the Reynolds number of the flow Re ¼ ri2 X=m, where ri and
ro are the radii of the inner and of the outer sphere respectively, X is the angular velocity of the rotating
sphere and is the coefficient of kinematic viscosity.
The properties of spherical Couette flow have been studied both experimentally [5–8] and theoretically
[9–14]. The motivation for the present work, was given by the fact that spherical annular flows are of
importance and consist fundamental problems in fluid dynamics and are used to investigate the motions in
planetary interiors. In this respect the dimensionless gap width between the spheres r ¼ 0:5 seems more
relevant to such motions and it fits to the geometry of certain planets of the solar system. On the other side
Taylor vortices are usually obtained numerically by starting with the Navier–Stokes equations of motion in
time-dependent form and adapting the time-stepping code until one reaches the desired state, at a given
dimensionless gap width. With regards to the dimensionless gap width, the flow regime is divided in narrow
gap with r < 0:12, medium gap with 0:12 6 r 6 0:24 and wide gap with r > 0:24. In the present study,
Taylor vortices are obtained for r ¼ 0:154, r ¼ 0:18 and r ¼ 0:336 using the previously described nu-
merical technique on the Navier–Stokes equations of motion for steady flow. The results are compared with
available experimental [5,6] and theoretical [10,11,14] results obtained by other investigators on the same
dimensionless gap width r, who worked numerically on the time-dependent form of the Navier–Stokes
equations of motion.
Conclusively, the contribution of the present work is twofold: (a) Extension of Allen and SouthwellÕs [1]
method to any orthogonal system of coordinates and conversion of the derived system of equations to
another that is solved by an iterative scheme rapidly convergent. (b) Employment of this technique to the
numerical simulation of Taylor vortices symmetric and asymmetric in annular spherical flow. To our
knowledge, asymmetric Taylor vortices have not been obtained so far by a steady-state solver. The method
2996 V.C. Loukopoulos et al. / Comput. Methods Appl. Mech. Engrg. 192 (2003) 2993–3003

is general and applies to axisymmetric flows with any body forces (i.e. convective flows with temperature
variations, MHD flows etc.).

2. Equations of motion and analysis

For motion symmetrical about an axis let q


¼ ðu
; v
; w
Þ be the velocity of the fluid particle in any
orthogonal system of coordinates (a
; b
; c
). In the above notation a
and b
are orthogonal coordinates in
a meridian plane and c
is the azimuthal angle u. All quantities are independent of u.
The equation of continuity
o


o

ðh2 h3 u Þ þ
ðh
1 h
3 v
Þ ¼ 0 ð6Þ
oa ob
is identically satisfied by the following definition of the stream function, which is
1 of
1 of

h
3 u
¼ ; h
3 v
¼  : ð7Þ
h
2 ob
h
1 oa

In the above notation h


1 , h
2 , h
3 are scale factors and h3 is the distance from the axis of revolution.
Introducing the vorticity x
¼ ðn
; n
; f
Þ and the angular velocity v
¼ h
3 w
, the equations of motion are [15]
2v
oðv
; h
3 Þ 1 oðf
; h
3 f
Þ 2h
3 f
oðf
; h
3 Þ
2


þ

¼ vD ðh3 f Þ;
h
1 h
2 h
2

3 oða ; b Þ h
1 h
2 h
3 oða
; b Þ h
1 h
2 h
2

3 oða ; b Þ
1 oðf
; v
Þ
¼ vD
2 v
;
h1 h2 h3 oða
; b
Þ


where
 
 


1 o h2 of
o h1 of
1
f ¼




þ



¼ 
D
2 f

h1 h2 oa h3 h1 oa ob h2 h3 ob h3
and
 
 


2h
3 o h2 o o h1 o
D ¼

þ
:
h1 h2 oa
h
1 h
3 oa
ob h
2 h
3 ob

Next, setting
1 2
f
¼ L3 Xf ; v
¼ L2 Xv; h
1 ¼ h1 ; h
2 ¼ Lh2 ; D ; a
¼ La; b
¼ b;
h
3 ¼ Lh3 ; D
2 ¼
L2
where L is a characteristic length and is a characteristic angular velocity, the non-dimensional Navier–
Stokes equations for steady axisymmetric flow of a viscous incompressible fluid can be expressed as follows:
J ¼ D2 f ; ð8Þ

2v oðv; h3 Þ 1 oðf ; J Þ 2J oðf ; h3 Þ


Re þ Re  Re ¼ D2 J ; ð9Þ
h1 h2 h23 oða; bÞ h1 h2 h3 oða; bÞ h1 h2 h23 oða; bÞ

1 oðf ; vÞ
Re ¼ D2 v; ð10Þ
h1 h2 h3 oða; bÞ
where h3 w ¼ Rev, Re ¼ L2 X=m and J ¼ h3 f is the vorticity function.
V.C. Loukopoulos et al. / Comput. Methods Appl. Mech. Engrg. 192 (2003) 2993–3003 2997

Our aim is to construct an algorithm by which an approximate numerical solution of the preceding
system of Eqs. (8)–(10) can be obtained. We thus approximate each one of them in finite differences using a
uniform grid in which the grid lines are parallel to the coordinate axes at each point ða; bÞ. If the grid size is
h in the a-direction and ‘ in the b-direction, all quantities at a typical set of grid points ða0 ; b0 Þ, ða0 þ h; b0 Þ,
ða0 ; b0 þ ‘Þ, ða0  h; b0 Þ, ða0 ; b0  ‘Þ––a classical five points discretization for the Laplace operator ), are
denoted by the subscripts 0, 1, 2, 3 and 4 respectively (Fig. 1).
Using the notation (7), Eq. (9) is split into two equations that are written as
    2 
o2 J h1 h3 o h2 oJ oJ 2h1 oh3 2h1 oh3
þ þ ðReh1 uÞ þ Re v þ u J ¼ h21 X þ Aða; bÞ; ð11Þ
oa2 h2 oa h1 h3 oa oa h2 h3 ob h3 oa
  
o2 J h2 h3 o h1 oJ oJ h22
þ þ ðReh 2 vÞ ¼  Aða; bÞ; ð12Þ
ob2 h1 ob h2 h3 ob ob h21
where
 
2v ov oh3 ov oh3
X ¼ Re 
h1 h2 h23 oa ob ob oa
and Aða; bÞ is an auxiliary unknown function. Working like in [2–4] we eliminate Aða; bÞ and we deduce a
relation analogous to (5). This relation is an algebraic equation with a number of unknowns, in which the
coefficients of the unknowns are exponential functions of the grid size. The presence of these terms makes
the matrix of the coefficients of the unknowns not necessarily diagonally dominant. This deficiency is
bypassed by expanding the velocity components and the exponentials as a Taylor series with the prede-
termined order of accuracy. In this way we deduce an algebraic system of linear equations in which the
matrix of the coefficients of the unknowns is diagonally dominant and the order of accuracy is kept. A
detailed description of the successive steps of the numerical technique is given in [16].
After lengthy manipulations Eq. (9) takes the form
a1 J1 þ a2 J2 þ a3 J3 þ a4 J4 þ a0 J0 ¼ G0 : ð13Þ
Working on Eq. (10) in a similar way like in (9) we deduce:
n1 v1 þ n2 v2 þ n3 v3 þ n4 v4 þ n0 v0 ¼ 0: ð14Þ
Finally, Eq. (8) gives
m1 f1 þ m2 f2 þ m3 f3 þ m4 f4 þ m0 f0 ¼ h2 J0 : ð15Þ
The analytic expressions of the coefficients of the unknowns in Eqs. (13)–(15) are presented in Appendix A.
In Eqs. (13)–(15) the error terms of order equal to or bigger than the order of the error terms that are
inherent in these equations have been omitted. In addition these equations hold at all grid points of the
mess for which amin < a0 < amax and bmin < b0 < bmax .
The matrices associated with Eqs. (13)–(15) are all diagonally dominant. The solution of these equations
is obtained by employing a relaxation method at all internal points of the annular region subject to
boundary conditions that are either known a priori or are recalculated during the process at the end of each
cycle. The iterative sequence terminates when all quantities have converged to limits.

3. Taylor vortices in annular spherical flow

In this section we apply the previous analysis in the solution of the equations of motion of a given in-
compressible viscous fluid contained between two concentric spheres having radii ri ¼ a=k (k > 1) the inner
2998 V.C. Loukopoulos et al. / Comput. Methods Appl. Mech. Engrg. 192 (2003) 2993–3003

er

r
ψ

α/k

Fig. 2. Spherical annulus.

and ro ¼ a the outer. The inner sphere is rotating with angular velocity X about a diameter and the outer is
at rest, Fig. 2. The motion is described by Eqs. (8)–(10) adapted to spherical coordinates ðr; w; uÞ and the
equations are presented in Appendix B.
The boundary conditions of the flow are

of
f ¼ ¼0 on both boundaries; ð16Þ
ow

v ¼ sin2 w at r ¼ 1 ðinner boundaryÞ;


ð17Þ
v ¼ 0 at r ¼ 1 þ r ðouter boundaryÞ;

for the angular velocity function and

3f ð1 þ h; wÞ J ð1 þ h; wÞ
J ð1; wÞ ¼   at r ¼ 1;
h2 2
ð18Þ
3f ð1 þ r  h; wÞ J ð1 þ r  h; wÞ
J ð1 þ r; wÞ ¼   at r ¼ 1 þ r;
h2 2

for the vorticity, where r ¼ 1 and r ¼ 1 þ r are the non-dimensional radii of the inner and of the outer
sphere. The boundary conditions (18) are of Oðh2 Þ.
The formation of Taylor vortices depends on the dimensionless gap width r ¼ ðro  ri Þ=ri and on the
Reynolds number of the flow Re ¼ ri2 X=m where m is the coefficient of the kinematic viscosity.
In order to obtain Taylor vortices at r ¼ 0:154 we scan like in [12] through a range of Reynolds numbers
0 < Re < 2600. The 0-vortex mode exists in the range Re < 803 and Re > 1858. The 1-vortex mode exists in
the range 804 < Re < 915 and 1761 < Re < 1857 and the 2-vortex mode exists in the range 916 < Re <
1760. Also, in the range 1761 < Re < 1857 the 1-vortex mode is asymmetric with respect to the equator. In
V.C. Loukopoulos et al. / Comput. Methods Appl. Mech. Engrg. 192 (2003) 2993–3003 2999

Fig. 3. (a) Stream function pattern of the 0-vortex flow without pinches (r ¼ 0:154, Re ¼ 780). (b) Stream function pattern of the 0-
vortex flow with pinches (r ¼ 0:154, Re ¼ 800). (c) Stream function pattern of the 1-vortex flow with pinches (r ¼ 0:154, Re ¼ 825). (d)
Stream function pattern of the 2-vortex flow without pinches (r ¼ 0:154, Re ¼ 1750). (e) Stream function pattern of the 1-vortex
asymmetric flow without pinches (r ¼ 0:154, Re ¼ 1765).

Fig. 3 Taylor vortices are shown for r ¼ 0:154. In all our calculations the typical truncation used was
21 points in the radial direction r ¼ ½1; 1 þ r times 181 points in the meridional direction w ¼ ½0; p . In
3000 V.C. Loukopoulos et al. / Comput. Methods Appl. Mech. Engrg. 192 (2003) 2993–3003

addition all calculations where run at higher resolution to ensure that 21  181 is sufficient for the con-
vergence of the iterative process to the adopted criterion of accuracy, which is

cs ðr; wÞ

max 1  sþ1 6 5  105 :
c ðr; wÞ

In the previous notation stands for anyone of the variables J , f and v, in Eqs. (13)–(15) and s is the number
of iterations. Fig. 3a depicts the zero vortex mode at Re ¼ 780. The figure is a typical two-dimensional
projection of the flow on the ðr; wÞ plane at a fixed azimuthal angle u. Counter-clockwise circulation is
shown as solid line, clockwise circulation is shown as dashed line. Owing to the Ekman pumping at the
poles, fluid is thrown outward along the rotating inner sphere and forms inward radial jets at the poles and
outward radial jets at the equator. The flow is reflection-symmetric about the equator. As the Reynolds
number is increased, the flow develops a pinching of streamlines. This is illustrated in Fig. 3b at Re ¼ 800.
The flow is again reflection-symmetric about the equator and is characterized by a stagnation point within
the contours and a strong outward flow at the equator. The pinch is located about one gap width from the
equator, in agreement with [12]. An example of the 1-vortex with pinches (supercritical) flow at Re ¼ 825 is
shown in the lines of constant streamfunction of Fig. 3c. The flow is axisymmetric and reflection-symmetric
about the equator. The Taylor vortices are near the equator, their diameter is approximately equal to one
gap width and their rotation causes an inward weak radial jet at the equator. Further increase in the
Reynolds number Re ¼ 1750 results in 2-vortex flow, Fig. 3d. The Taylor vortices each have a diameter
nearly to the gap width and the equator is an outflow boundary. Finally, in Fig. 3e we have a stable,

Fig. 4. (a) Stream function pattern of the 0-vortex flow without pinches (r ¼ 0:18, Re ¼ 600). (b) Stream function pattern of the
0-vortex flow with pinches (r ¼ 0:18, Re ¼ 650). (c) Stream function pattern of the 1-vortex flow with pinches (r ¼ 0:18, Re ¼ 700).
(d) Stream function pattern of the 2-vortex flow without pinches (r ¼ 0:18, Re ¼ 900).
V.C. Loukopoulos et al. / Comput. Methods Appl. Mech. Engrg. 192 (2003) 2993–3003 3001

Fig. 5. (a) Stream function pattern of the 1-vortex symmetric flow (r ¼ 0:336, Re ¼ 728). (b) Stream function pattern of the 1-vortex
asymmetric flow (r ¼ 0:336, Re ¼ 896).

equatorially asymmetric state for Re ¼ 1765. The results are in good agreement with these of B€ uhler [12]
and Mamun and Tuckerman [13].
In order to obtain Taylor vortices at r ¼ 0:18 we scan like in [11] through the range of Reynolds num-
bers 0 < Re < 2000. The 0-vortex mode exists in the range Re < 635 and Re > 1285. The 1-vortex mode
exists in the range 662 < Re < 793 and 1209 < Re < 1284 and the 2-vortex mode exists in the range
714 < Re < 1208. In Fig. 4 Taylor vortices are shown for r ¼ 0:18. For this value of the dimensionless gap
width we were not able to produce asymmetric vortices, in agreement with [11].
In order to obtain Taylor vortices at a wide gap width (r ¼ 0:336) we scan like in [14] through the range
of Reynolds numbers 415 < Re < 2040. The 1-vortex mode exists in the range 719 6 Re 6 880 and the
asymmetric 1-vortex mode exists in the range 880 < Re 6 896. In Fig. 5 Taylor vortices are shown for
r ¼ 0:336.
From our numerical scanning of the range of existence of Taylor vortices at a given gap width it is
concluded that increasing in the dimensionless gap width up to about r ¼ 0:35 makes the Taylor vortices to
appear at smaller Reynolds numbers and reduces the range of existence of Taylor vortices, in agreement
with [8].

4. Conclusions

The present numerical technique is an extension of previous work on the solution of the Navier–Stokes
equations of steady flow in Cartesian and cylindrical coordinates, to any orthogonal system of coordinates.
This technique is employed to the numerically simulation of Taylor vortices in spherical Couette flow. Our
main interest is focused on Taylor vortices at a dimensionless gap width r ¼ 0:154, r ¼ 0:18 and r ¼ 0:336.
The numerical results are in good agreement with the adapted time-dependent form used by Marcus and
Tuckerman [11], B€ uhler [12] and Hollerbach [14]. In particular we have obtained Taylor vortices both
symmetric and asymmetric at r ¼ 0:154 and r ¼ 0:336 that are in agreement to [11,12,14] and symmetric
vortices at r ¼ 0:18, that are in agreement to [11,13].
3002 V.C. Loukopoulos et al. / Comput. Methods Appl. Mech. Engrg. 192 (2003) 2993–3003

Appendix A

For Eq. (13) we have:


Reu0 h2 oh1 Re2 h21 u20 h2 ReS1 h1 u0 h2
a1 ¼ 2  þ  þ S1 h  Reh1 u0 h;
 2 oa 4 2 
h2 Rev0 ‘2 oh2 Re2 h22 v20 ‘2 ReS2 h2 v0 ‘2
a2 ¼ k2 12 2  þ  þ S2 ‘  Reh2 v0 ‘ ;
h2 2 ob 4 2
Reu0 h2 oh1 Re2 h21 u20 h2 ReS1 h1 u0 h2
a3 ¼2 þ   S1 h þ Reh1 u0 h;
 2 oa 4 2 
h2 Rev0 ‘2 oh2 Re2 h22 v20 ‘2 ReS2 h2 v0 ‘2
a4 ¼ k2 12 2  þ   S2 ‘ þ Reh2 v0 ‘ ;
h 2 ob 4 2
2 2 2 2 2
 
Re h1 u0 h oh1 u0 h21 h3 o h2
a0 ¼ 4þ  Reu0 h2 þ Reð2h2 Þ
2 oa 2h2 oa h1 h3
 2 
2h v
1 0 oh 3 2h u
1 0 oh 3 h2
þ Reð2h2 Þ þ þ 4k2 12
h2 h3 ob h3 oa h2
 2 2 2 2 2
  2 
Re h2 v0 h oh 2 v 0 h h 3 o h1 h1
þ  Rev0 h2 þ Reð2h2 Þ 2 ;
2 ob 2h1 ob h2 h3 h22
where
   
2v0 ov oh3 ov oh3 h h1 h3 o h2
G0 ¼ 2h2 h21 Re  ; k¼ ; S1 ¼ and
h1 h2 h23 oa ob ob oa ‘ h2 oa h1 h3
 
h2 h3 o h1
S2 ¼ :
h1 ob h2 h3
For Eq. (14):
Reu0 h2 oh1 Re2 h21 u20 h2 ReS1 h1 u0 h2
n1 ¼ 2  þ  þ S1 h  Reh1 u0 h;
 2 oa 4 2 
h2 Rev0 ‘2 oh2 Re2 h22 v20 ‘2 ReS2 h2 v0 ‘2
n2 ¼ k2 12 2  þ  þ S2 ‘  Reh2 v0 ‘ ;
h2 2 ob 4 2
Reu0 h2 oh1 Re2 h21 u20 h2 ReS1 h1 u0 h2
n3 ¼2 þ   S1 h þ Reh1 u0 h;
 2 oa 4 2 
h2 Rev0 ‘2 oh2 Re2 h22 v20 ‘2 ReS2 h2 v0 ‘2
n4 ¼ k2 12 2  þ   S2 ‘ þ Reh2 v0 ‘ ;
h 2 ob 4 2
2 2 2 2 2
 
Re h1 u0 h oh1 u0 h21 h3 o h2
n0 ¼ 4þ  Reu0 h2 þ Reð2h2 Þ
2 oa 2h2 oa h1 h3
2
 2 2 2 2   2 
h Re h v h oh 2 v 0 h2 h3 o h1 h1
þ 4k2 12 þ 2 0
 Rev0 h2 þ Reð2h2 Þ 2 :
h2 2 ob 2h1 ob h2 h3 h22
For Eq. (15):
 
1 h3 o h2 h
m1 ¼ 2 þ ;
h1 h1 h2 oa h1 h3 2
 
k2 h3 o h1 ‘k2
m2 ¼ 2 þ ;
h2 h1 h2 ob h2 h3 2
 
1 h3 o h2 h
m3 ¼ 2  ;
h1 h1 h2 oa h1 h3 2
V.C. Loukopoulos et al. / Comput. Methods Appl. Mech. Engrg. 192 (2003) 2993–3003 3003

  2
k2 h3 o h1 ‘k
m4 ¼ 2  ;
h2 h1 h2 ob h2 h3 2
 
2 2k2
m0 ¼  2 þ 2 :
h1 h2

Appendix B

For spherical polar coordinates ðr; w; uÞ we take a as r and b as w where 1 6 r 6 1 þ r, 0 6 w 6 p and


0 6 u 6 2p. Then h1 ¼ 1, h2 ¼ r, h3 ¼ r sin w. Accordingly
1 of 1 of Rev o2 cos w o 1 o2
u¼ ; v¼ ; w¼ ; D2 ¼  þ
2
r sin w ow r sin w or r sin w or2 r2 sin w ow r2 ow2
and Eqs. (9) and (10) take the form
    
Re oðf ; J Þ 2J of of Re2v ov ov
 þ r cos w  sin w  r cos w  sin w ¼ D2 J
r2 sin w oðr; wÞ r sin h or ow r3 sin2 w or ow
and
Re oðf ; vÞ
 ¼ D2 v:
r2 sin w oðr; wÞ

References

[1] D.N. Allen, R. Southwell, Relaxation methods applied to determine the motion in two dimensions of a viscous fluid past a fixed
cylinder, Quar. J. Mech. Appl. Math. 8 (1955) 129–145.
[2] L.C. Wood, A note on the numerical solution of fourth order differential equations, Aerospace Quart. 5 (1954) 176–184.
[3] S.C.R. Dennis, J.D. Houdson, in: Proc. of the 1st International Conference on Numerical Methods in laminar and turbulent flow,
London, Pentech Press, 1978, 69–80.
[4] S.C.R. Dennis, Finite differences associated with second-order differential equations, Quart. J. Mech. Appl. Math. 13 (1960) 487–
507.
[5] O. Sawatzki, J. Zieper, Das Stromfeld im Spalt Zwishen Zwei Konzentrischen Kugelfl€achen, von denen die innere rotiert, Acta
Mech. 9 (1970) 13–35.
[6] M. Wimmer, Experiments on a viscous fluid flow between concentric rotating spheres, J. Fluid Mech. 78 (1976) 317–335.
[7] C. Egbers, H.G. Rath, The existence of Taylor vortices and wide-gap instabilities in spherical Couette flow, Acta Mech. 111 (1995)
125–140.
[8] M. Liu, C. Blohm, C. Egbers, P. Wulf, H.J. Rath, Taylor vortices in wide spherical shells, Phys. Rev. Lett. 77 (2) (1996) 286–289.
[9] F. Bartels, Taylor vortices between two concentric rotating spheres, J. Fluid Mech. 119 (1982) 1–25.
[10] G. Schrauf, The first instability in spherical Taylor–Couette flow, J. Fluid Mech. 166 (1986) 287–303.
[11] P.S. Marcus, L.S. Tuckerman, Simulation of flow between concentric rotating spheres, Part 1. Steady states, J. Fluid Mech. 185
(1987) 1–30.
[12] K. B€uhler, Symmetric and asymmetric Taylor vortex flow in spherical gaps, Acta Mech. 81 (1990) 3–38.
[13] C.K. Mamun, L.S. Tuckerman, Asymmetry and Hopf bifurcation in spherical Couette flow, Phys. Fluids 7 (1) (1995) 80–91.
[14] R. Hollerbach, Time-dependent Taylor vortices in wide-gap spherical Couette flow, Phys. Rev. Lett. 81 (15) (1998) 3132–3135.
[15] S. Goldstein, Modern Developments in Fluid Dynamics, Vol. 1, Dover, New York, 1965, 114–115.
[16] V.C. Loukopoulos, in: Proc. of the 4th GRACM Congress on Computational Mechanics, Patras, Greece, LFME, Vol. 1, 2002,
88–95.

You might also like