Lee, D.-S.; Rummler, B.: Eigenfunctions of the Stokes Operator.

III 399

ZAMM
Z. Angew. Math. Mech. 82 (2002) 6, 399––407

Lee, D.-S.; Rummler, B.

The Eigenfunctions of the Stokes Operator in Special Domains. III

We consider the eigenvalue problem of the Stokes operator in a bounded domain of R3 bounded by two concentrical
cylinders with homogeneous Dirichlet boundary conditions on the curved part of the boundary and periodical conditions
in along the cylinder axis (in x1 -direction). We deduce by separation the correspondent systems of ordinary differential
equations and solve them explicitly looking for solenoidal vector fields fulfilling the boundary conditions. The investiga-
tion of possible cases yields either the explicit eigenfunctions and eigenvalues or equations for the determination of the
eigenvalues and a general representation of the eigenfunctions. The completeness of the calculated system of eigenfunc-
tions in S can be proven analogously to ½9.

Key words: Stokes operator, eigenfunctions, concentrical cylinders

MSC (2000): 34A30, 34B24, 35Q30, 76D07

1. Preface

We give in this part of the series on Eigenfunctions of the Stokes Operator in Special Domains the results associated to
the numerical treatment of the so-called viscous Taylor-Couette flow. The spatial domain is the gap between two
concentrical cylinders. We take, for the boundary conditions, in addition to homogeneous Dirichlet boundary condi-
tions on the rigid part of the boundary, periodical boundary conditions in the direction of the axis of the cylinders.
This choice guarantees that the problems investigated here are well defined in the classical sense, too, and furthermore
that all eigenfunctions are in C 1 .
We note that the Stokes operator is a positive self-adjoint operator with compact inverse, which follows in a
usual way from the general theory of elliptical systems. The general theory of elliptical systems also provides the
existence of the countable system of Stokes eigenfunctions and the completeness of such a system in the subspace of
ðL2 ð:ÞÞ3 consisting of weak solenoidal vector fields which fulfil the generally required boundary conditions in the H 1=2
trace sense. The exact knowledge of the eigenfunctions of the Stokes operator is especially useful for the the numerical
treatment of the incompressible Navier-Stokes equations by Galerkin methods.
From the theoretical point of view the fields of applications are investigations to the existence of space-time
statistical solutions of the incompressible Navier-Stokes equation in unbounded domains ([6, 9, 11]). One needs a
detailed knowledge of the Stokes functions to prove the existence of such space-time statistical solutions with rota-
tional invariance properties in the measure sense [9].
We do not deal with the application of the present results to physical and technical problems here and we made
no attempts to give a survey of these lines of research. The reader interested in these topics will find such applications
e.g. in [1, 2, 6].
In part III of this series we investigate the eigenvalue-problem of the Stokes operator in a gap domain between
two concentrical cylinders with non-flux conditions on the curved part of the boundary and periodical conditions in the
direction of the axis of the cylinders.
In essential parts we succeed on the line of investigations used by the second author in [7].
The paper is organised as follows. The Stokes operator is defined in Section 2. For the understanding of the
results in Sections 3 and 4 it is not necessary to know the full theoretical background of Section 2 in detail. The reader
interested in applications of the results can take the Stokes operator as the product of the projection operator
p: ðL2 Þ3 7! ðL2 Þ3 onto solenoidal vector fields and the negative Laplacian. In Section 3 we give the system of ordinary
differential equations derived from the Stokes problem by using the method of separating the variables. In Section 4 we
solve such systems of ordinary differential equations for the possible cases. We obtain here the solenoidal eigenfunc-
tions of the Stokes operator by the consideration of the defined boundary conditions. We also give the eigenvalues
explicitly or equations for the determination of these eigenvalues.
In Section 5 we write down additionally the eigenfunctions of the Stokes operator in an open gap domain be-
tween two concentric circles.
We use here methods of the general theory of ordinary differential equations with case discussions to obtain our
results. Let us mention that it is advantageous to apply the present results in the numerical approximation to the
solution of the non-stationary incompressible Navier-Stokes equations by Galerkin method as real-valued cartesian
vector functions of cylindrical coordinates. For these purposes one can also employ representation formulas (cf. [3]) of
general Helmholtz decompositions.

# WILEY-VCH Verlag Berlin GmbH, 13086 Berlin, 2002 0044-2267/02/0606-0399 $ 17.50þ.50/0

 15214001, 2002, 6, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/1521-4001(200206)82:6<399::AID-ZAMM399>3.0.CO;2-6 by Curtin University Library, Wiley Online Library on [02/05/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
400 ZAMM
Z. Angew. Math. Mech. 82 (2002) 6

2. Foundations

2.1 Function spaces

Let Tl be a gap domain between two concentrical cylinders defined by
Tl ¼ fx 2 R3 : x ¼ ðx1 ; x2 ; x3 Þ; jx1 j < l; s < x22 þ x23 < 1g ; with s > 0:
The boundary G of Tl consists of four parts G 1 ; G s ; G þl ; G l :
G 1; s ¼ fx 2 R3 : jx1 j  l; x22 þ x23 ¼ 1; sg ;
G l ¼ fx 2 R3 : x1 ¼ l; s < x22 þ x23 < 1g :
Let us denote the open gap between two concentrical circles in R2 by Kðs; 1Þ :
Kðs; 1Þ ¼ fx 2 R2 : x ¼ ðx2 ; x3 Þ; s < x22 þ x23 < 1g :
We will consider spaces of complex-valued vector fields uðxÞ ¼ ðu1 ðxÞ; u2 ðxÞ; u3 ðxÞÞ periodical in the x1 direction.
Definition 1: We denote by Sð¼ S0 Þ the closure of the set
P
V ¼ fvðxÞ 2 ðC 1 ðTl ÞÞ3 : r
v ¼ 0; vðxÞ ¼ vk1 ðx2 ; x3 Þ expðik1 x1 Þ : vk1 ðx2 ; x3 Þ 2 ðC01 ðK ðs; 1Þ ÞÞ3 g
k1 2 ðp=lÞ Z

in the norm of Hð¼ H0 Þ ¼ ðL2 ðTl ÞÞ3 . The space S1 arises from the closure of V in the norm of H1 ¼ ðW21 ðTl ÞÞ3 . The
spaces Sm for m ¼ 2; 3; 4; are defined through the relation Sm ¼ S1 \ Hm , where Hm ¼ ðW2m ðTl ÞÞ3 . The norms of the
spaces Sm agree for m ¼ 0; 1; 2; on Sm with the norms of Hm denoted by k:km and the scalar product on H is denoted
by ð:; :Þ.

2.2 Trace spaces and definition of the Stokes operator

Using the notations and investigations of [10] we denote by g the trace operator g : W21 ðTl Þ ! H 1=2 ðGÞ ; where
gu ¼ ujG 8u 2 C 1 ðTl Þ:
H 1=2 ðGÞ  L2 ðGÞ is the space of the general boundary values of W21 ðTl Þ.
g is a continuous mapping and the inequality
kgukL2 ðGÞ  const
kukW 1 ðTl Þ
2

is valid.
The normal component of boundary values of functions of S exists in the following sense: One sees slightly that
the space S is a subspace of the Hilbert-space E:
E ¼ fu 2 H : r
u 2 L2 ðTl Þg
endowed with the scalar product ð:; :ÞE ¼ ð:; :Þ þ ðr
:; r
:ÞL2 ðTl Þ .
We denote by n ¼ ðn1 ; n2 ; n3 Þ the outward directed unit normal on G.
Let H 1=2 ðGÞ be the dual space of H 1=2 ðGÞ with the duality relation being induced by the scalar product of
L2 ðGÞ. Then the trace operator gn is defined by
gn : E  S ! H 1=2 ðGÞ ;
where
gn u ¼ n
u 8u 2 ðC 1 ðTl ÞÞ3 :
We note that 8u 2 S and w 2 W21 ðTl Þ the general Gaussian Integral theorem is valid:
ðu; rwÞ þ ðr
u; wÞL2 ðTl Þ ¼ hgn u; gwi :

We will give here two theorems describing the properties of any of the function spaces defined above and proper-
ties of the Stokes operator, respectively. We note that the proofs of these theorems are similar to the proofs of corre-
sponding results shown in [6] and [9].
Theorem 1: Let S? be the orthogonal complement of S in H : H ¼ S?  S. Then the following identities are
valid:
S ¼ fu 2 H : r
u ¼ 0; gn ujG 1 ¼ gn ujG s ¼ 0; gn ujG þ þ gn ujG  ¼ 0g ;
S? ¼ frp : p 2 W21 ðTl Þ; gpjG þ  gpjG  ¼ 0g :

Definition 2: Let pl be the orthogonal projection of H onto S: pl : H ! S. We denote by Al the operator

Al ¼  pl 4 : S2 ! S, Let us call Al the Stokes operator.
We note that Al is a positive symmetric operator.
 15214001, 2002, 6, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/1521-4001(200206)82:6<399::AID-ZAMM399>3.0.CO;2-6 by Curtin University Library, Wiley Online Library on [02/05/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Lee, D.-S.; Rummler, B.: Eigenfunctions of the Stokes Operator. III 401

Theorem 2: The equation Al u ¼ f has for all f 2 S a unique solution u 2 S2 . For any f 2 Sm ; m ¼ 0; 1; 2; . . . ;
and l 1 there exists a number cm independent of l such that kukmþ2  cm kf km .
Remark 1: From Theorem 2 it follows that the operator Al is bounded below. Then there exists the inverse
operator A1
l since Al S2 ¼ S. Since the embedding operator of S2 in S is compact it follows that A1
l , considered as
Al : S ! S  S2 , is a compact operator. This means that Al is an operator with pure point sprectra and that the
1

system of eigenfunctions of Al is complete in S.

We note, that the completeness of the following explicitly calculated system of eigenfunctions can be proven
analogously as in [9], Sect. 4.3.

3. Introduction

We consider the eigenvalue problem of the operator pl D acting on a subspace of ðW22 ðTl ÞÞ3 with images in
fu 2 ðL2 ðTl ÞÞ3 : div u ¼ 0g.
Let Tl be an open bounded set given by
Tl ¼ fx 2 R3 : x ¼ ðx1 ; x2 ; x3 Þ; jx1 j < l; s < x22 þ x23 < 1g ; s > 0;
where the boundary of Tl consists of four parts G 1 ; G s ; G þl ; G l defined in Section 2.
We first notice that the investigations of the eigenvalue problem of the operator pl D consist in the following
task. We seek nontrivial vector functions u and a scalar function p, which are defined in Tl and satisfy the following
equations and boundary conditions:
Du þ lu ¼ rp ; ð1Þ
Dp ¼ 0 in Tl ; ð2Þ
r
u ¼ 0; ð3Þ
with uG 1 ¼ uG s ¼ 0 ; ð4Þ
where uðx1 ; x2 ; x3 Þ ¼ uðx1 þ 2l; x2 ; x3 Þ, pðx1 ; x2 ; x3 Þ ¼ pðx1 þ 2l; x2 ; x3 Þ :
Now we introduce the polar coordinates in the ðx2 ; x3 Þ-plane. These coordinates will be denoted by r and j.
This leads us to the new conditions:
uðx1 ; r; jÞ ¼ uðx1 ; r; j þ 2pÞ and pðx1 ; r; jÞ ¼ pðx1 ; r; j þ 2pÞ :
We denote the components of the velocity vector by ux1 ; ur , and uj .
The Laplacian in cylindrical coordinates is defined by
@2 @2 1 @ 1 @2
D¼ 2
þ 2þ þ 2 : ð5Þ
@x1 @r r @r r @j2
Taking this representation of the Laplacian we can write the eqs. (1)––(3) in cylindrical coordinates:
@p
Dux1 þ lux1 ¼ ;
@x1
2 @uj ur @p
Dur þ lur  2  2¼ ; ð6Þ
r @j r @r
2 @ur uj 1 @p
Duj þ luj þ 2  2 ¼ ;
r @j r r @j
Dp ¼ 0 ; ð7Þ
@ux1 @ur ur 1 @uj
þ þ þ ¼ 0: ð8Þ
@x1 @r r r @j
We look for solutions of the eigenvalue problem (6)––(8) satisfying the conditions explained above in the form
ukn ðx1 ; r; jÞ ¼ ukn ðrÞ expðikx1 þ injÞ ; ð9Þ
p
n 2 Z; k2 Z:
l
pkn ðx1 ; r; jÞ ¼ pkn ðrÞ expðikx1 þ injÞ ; ð10Þ
Under these assumptions we get from (6)––(8) the following system of ordinary differential equations, where the differ-
entiations with respect to r are denoted by 0 and 00 , respectively, and the factor expðikx1 þ injÞ is omitted:


00 1 0 2 n2
ukn; x1 ðrÞ þ u ðrÞ þ l  k  2 ukn; x1 ðrÞ ¼ ikpkn ðrÞ ; ð11Þ
r kn; x1 r


1 0 n2 þ 1 2in
u00kn; r ðrÞ þ ukn; r ðrÞ þ l  k2  ukn; r ðrÞ  2 ukn; j ðrÞ ¼ p0kn ðrÞ ; ð12Þ
r r2 r
 15214001, 2002, 6, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/1521-4001(200206)82:6<399::AID-ZAMM399>3.0.CO;2-6 by Curtin University Library, Wiley Online Library on [02/05/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
402 ZAMM
Z. Angew. Math. Mech. 82 (2002) 6


1 0 n2 þ 1 2in in
u00kn; j ðrÞ þ u ðrÞ þ l  k  2
ukn; j ðrÞ þ 2 ukn; r ðrÞ ¼ pkn ðrÞ ; ð13Þ
r kn; j r2 r r


1 0 n2
p00kn ðrÞ þ pkn ðrÞ  k2 þ 2 pkn ðrÞ ¼ 0 ; ð14Þ
r r
1 in
u0kn; r ðrÞ þ ukn; r ðrÞ þ ikukn; x1 ðrÞ þ ukn; j ðrÞ ¼ 0 : ð15Þ
r r
We obtain a step by step solvable system (11)––(13). Let us remark that eq. (15) is essential for the uncoupling of the
homogenous differential equation (12) and (13).

4. Investigation of the possible cases

4.1 The case n ¼ k ¼ 0

We use the general solution of (14):
p00 ðrÞ ¼ c1 þ c2 ln r ; c1 ; c2 2 C ; ð16Þ
and receive the general solutions of the (11)––(13) corresponding system of ordinary differential equations by standard
methods as
pffiffiffi pffiffiffi
u00l; x1 ðrÞ ¼ ax1 J0 ð l rÞ þ bx1 Y0 ð l rÞ ;
pffiffiffi pffiffiffi c2
u00l; r ðrÞ ¼ ar J1 ð l rÞ þ br Y1 ð l rÞ þ ; a
; b
2 C : ð17Þ
pffiffiffi pffiffiffi l
r
u00l; j ðrÞ ¼ aj J1 ð l rÞ þ bj Y1 ð l rÞ :
The functions Jn ð:Þ and Yn ð:Þ used above denote Bessel’s and Weber’s functions of the index n, respectively (cf. [5]).
To fulfil the homogeneous boundary conditions on G 1 and G s and eq. (3) (resp. (8)) we have to seek nontrivial
solutions of a linear system of equations in ax1 , bx1 , ar , br , aj , bj , and c2 . Looking for the roots of the determinant of
coefficients we find that (3) provides ar ¼ br ¼ 0. Then it follows from the boundary conditions that c2 vanishes also.
We further interpret the determinant of coefficients as the product of sub-determinants and obtain as the first
class of eigenfunctions
00 pffiffiffi 1 pffiffiffi 0 Y ðpffiffilffi rÞ 11
J0 ð l rÞ ð 0
u00l ðx1 ; r; jÞ ¼ ax1 @@ 0 A  0 plffiffiffi sÞ @
J
0 AA ; ð18Þ
0 Y 0 ð l sÞ 0
pffiffiffi
written in cylindrical coordinates, where l runs through all of the roots of
 pffiffiffi pffiffiffi 
 J0 ð lÞ Y0 ðpffiffiffilÞ 
 p ffiffi
ffi
 J ð l sÞ Y ð l sÞ  ¼ 0 :
0 0

Here ax1 is determined by means of the relation

jjuknl ðx1 ; r; jÞjjL2 ¼ 1 ; ð19Þ
3
with the notation L2 ¼ ðL2 ðTl ÞÞ .
Taking a similar way we immediately get a second class of eigenfunctions of the form
00 1 pffiffiffi 0 11
0 0
J 1 ð l sÞ
u00l ðx1 ; r; jÞ ¼ aj @@ 0 ffiffiffi A 
p pffiffiffi @ 0 ffiffiffi AA ;
p ð20Þ
J1 ð l rÞ Y1 ð l sÞ Y ð l rÞ
1
pffiffiffi
written in cylindrical-coordinates, too. l runs here through all the roots of
 pffiffiffi pffiffiffi 
 J1 ð lÞ Y ð lÞ 

 J ðpffiffilffi sÞ Y ðpffiffilffi sÞ  ¼ 0
1
ð21Þ
1 1

and the numbers aj are determined with the help of the requirement (19).
Remark 2: This second class of eigenfunctions let us simultaneously obtain a class of eigenfunctions of the Stokes
operator acting on the open gap between two concentrical circles as a subset of R2 . Details are written down in Section 5.

4.2 The case n 6¼ 0, k ¼ 0

The general solution of (14) in the considered case is given by
p0n ðrÞ ¼ c1 rn þ c2 rn ; c1 ; c2 2 C ; ð22Þ
where we have omitted the factor expðinjÞ.
 15214001, 2002, 6, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/1521-4001(200206)82:6<399::AID-ZAMM399>3.0.CO;2-6 by Curtin University Library, Wiley Online Library on [02/05/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Lee, D.-S.; Rummler, B.: Eigenfunctions of the Stokes Operator. III 403

We obtain from the general problem (11)––(13) the following system of ordinary differential equations for the
general solution of the homogeneous system, again with the suppression of expðinjÞ on both sides of the equations:


1 0 n2
u000n; x1 ðrÞ þ u0n; x1 ðrÞ þ l  2 u0n; x1 ðrÞ ¼ 0 ;
r r

2

3 n 1
u000n; r ðrÞ þ 0
u0n; r ðrÞ þ l  u0n; r ðrÞ ¼ 0 ; ð23Þ
r r2


1 n2 þ 1 2in
u000n; j ðrÞ þ u00n; j ðrÞ þ l  2
u0n; j ðrÞ ¼  2 u0n; r ðrÞ ;
r r r

where the second and the third equation were uncoupled with the help of relation (15).
We get the general solution of the inhomogeneous system (11)––(13) through the step by step solution of system
(23) adding a special solution of the inhomogeneous equations
pffiffiffi pffiffiffi
u0nl; x1 ðrÞ ¼ ax1 Jn ð l rÞ þ bx1 Yn ð l rÞ ;
ar pffiffiffi br pffiffiffi c1 n n1 c2 n n1
u0nl; r ðrÞ ¼ Jn ð l rÞ þ Yn ð l rÞ þ r  r ;
r r l l
pffiffiffi pffiffiffi c1 in n1 c2 in n1 a
; b
2 C : ð24Þ
u0nl; j ðrÞ ¼ aj Jpnffiffiffiffiffiffiffiffi
2 þ1 ð l rÞ þ bj Jpffiffiffiffiffiffiffiffi
n2 þ1 ð l rÞ þ r þ r
pffiffiffi l l
i l pffiffiffi pffiffiffi pffiffiffi pffiffiffi
þ ðar ðJn1 ð l rÞ  Jnþ1 ð l rÞÞ þ br ðYn1 ð l rÞ  Ynþ1 ð l rÞÞÞ :
2n

Regarding eq. (3) (resp. (8)) we found that aj ¼ bj ¼ 0. For this reason we have to seek non-trivial solutions of
a linear system of equations in ax1 , bx1 , ar , br , c1 , and c2 to fulfil the homogeneous boundary conditions on G 1 and G s .
Working with sub-determinants of the determinant of coefficients we get a first class of eigenfunctions in the form of
00 pffiffiffi 1 pffiffiffi 0 Y ðpffiffilffi rÞ 11
Jn ð l rÞ J ð l sÞ n
n
u0nl ðx1 ; r; jÞ ¼ ax1 expðinjÞ@@ 0 A pffiffiffi @ 0 AA ; ð25Þ
0 Y n ð l sÞ 0
pffiffiffi
written in cylindrical coordinates, where l runs through all of the roots of
 pffiffiffi pffiffiffi 
 Jn ð lÞ Yn ðpffiffiffilÞ 

 J ð ffiffilffi sÞ
p
Yn ð l sÞ 
¼0
n

and the numbers ax1 are determined by the requirement (19).

From the investigations above it follows that we now have only to regard the case where aj ¼ bj
¼ ax1 ¼ bx1 ¼ 0.
From the calculation of the corresponding sub-determinant we receive the incidental equation for the eigenvalues

8n pffiffiffi pffiffiffi pffiffiffi pffiffiffi

¼ s n1 ðJnþ1 ð l sÞ Yn1 ð lÞ  Jn1 ð lÞ Ynþ1 ð l sÞÞ
pls 2 pffiffiffi pffiffiffi pffiffiffi pffiffiffi
þ s n1 ðJnþ1 ð lÞ Yn1 ð l sÞ  Jn1 ð l sÞ Ynþ1 ð lÞÞ ; ð26Þ

where l runs through all the roots of (26). We can write the solutions of the linear sub-system following from the
homogeneous boundary conditions on G 1 and G s for every root l of (26) in the form br ¼ ar
BðlÞ, c1 ¼ ar
CðlÞ, and
c2 ¼ ar
C~ðlÞ. In this way we receive a second class of eigenfunctions:
8 2 0 1 0 13
>
<n 0 0
6 n1 B C B C7
u0nl ðx1 ; r; jÞ ¼ ar expðinjÞ 4Cr @ 1 A þ C~rn1 @ 1 A5
>
:l
i i
0 1 0 19
0 0 >
>
B pffiffiffi C B pffiffiffi C>
=
B Jn ð l rÞ=r C B Yn ð l rÞ=r C
þB pffiffiffi C þ BB pffiffiffi C :
@i l pffiffiffi pffiffiffi A @i l pffiffiffi pffiffiffi A>
>
ðJn1 ð l rÞ  Jnþ1 ð l rÞÞ ðYn1 ð l rÞ  Ynþ1 ð l rÞÞ >
;
2n 2n

The values of ar are determined with the help of requirement (19).

 15214001, 2002, 6, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/1521-4001(200206)82:6<399::AID-ZAMM399>3.0.CO;2-6 by Curtin University Library, Wiley Online Library on [02/05/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
404 ZAMM
Z. Angew. Math. Mech. 82 (2002) 6

Remark 3: We also obtain simultaneous the eigenfunctions of the Stokes operator acting on the open gap
between two concentrical circles as a subset of R2 . These functions are written down in detail in Section 5.

4.3 The case k 6¼ 0, n ¼ 0

In the main points we follow the way of investigations used in the previous subsections. Leaving out the factor
expðikx1 Þ we get the general solution of (14) by
pk0 ðrÞ ¼ c1 I0 ðjkj rÞ þ c2 K0 ðjkj rÞ ; c1 ; c2 2 C : ð27Þ
The functions In ð:Þ and Kn ð:Þ denote Bessel’s functions of imaginary argument and Mac-Donald’s functions of the
index n 2 Z, respectively.
From the general problem (11)––(13) follows the system of ordinary differential equations for the general solution
of the homogeneous system, corresponding the case considered here; with the suppression of expðikx1 Þ on both sides of
the equations:
1 0
u00k0; x1 ðrÞ þ u ðrÞ þ ðl  k2 Þ uk0; x1 ðrÞ ¼ 0 ;
r k0; x1

1 1
u00k0; r ðrÞ þ 0 2
uk0; r ðrÞ þ l  k  2 uk0; r ðrÞ ¼ 0 ; ð28Þ
r r


1 1
u00k0; j ðrÞ þ u0k0; j ðrÞ þ l  k2  2 uk0; j ðrÞ ¼ 0 :
r r

By adding a special solution of the inhomogeneous equations to the general solution of system (28) we get the
general solution of the inhomogeneous system
c1 ik c2 ik
uk0l; x1 ðrÞ ¼ ax1 J0 ðmrÞ þ bx1 Y0 ðmrÞ þ I0 ðjkj rÞ þ K0 ðjkj rÞ ;
l l
c1 jkj c1 jkj
uk0l; r ðrÞ ¼ ar J1 ðmrÞ þ br Y1 ðmrÞ þ I1 ðjkj rÞ  K1 ðjkj rÞ ; a
; b
2 C : ð29Þ
l l
uk0l; j ðrÞ ¼ aj J1 ðmrÞ þ bj Y1 ðmrÞ ;
where we use the abbreviation
pffiffiffiffiffiffiffiffiffiffiffiffiffi
m :¼ l  k2 : ð30Þ
Firstly, we are able to regard only the sub-determinant to aj ; bj to fulfil the homogeneous boundary conditions
on G 1 and G s for the j-component of our solution, since the j-component does not appear explicitly in the eq. (3)
(resp. (8)). (The eq. (3) is always satisfied by a vector function with only non vanishing j-component independent of
j.)
So we get the first class of eigenfunctions by
00 1 0 11
0 0
J 1 ðmsÞ
uk0l ðx1 ; r; jÞ ¼ aj expðikx1 Þ@@ 0 A  @ 0 AA ; ð31Þ
Y1 ðmsÞ
J1 ðmrÞ Y1 ðmrÞ

written in cylindrical coordinates, where l runs through all of the roots of

 
 J1 ðmÞ Y1 ðmÞ 

 J1 ðmsÞ Y1 ðmsÞ  ¼ 0 :

Now we study as the second case the solution with aj ¼ bj ¼ 0. It follows from relation (3), that the vector field
is solenoidal, if the equations
pffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffi
ikax1 þ l  k2 ar ¼ 0 and ikbx1 þ l  k2 br ¼ 0 ð32Þ

hold. For this reason we have to seek non-trivial solutions of a linear system of equations in ar , br , c1 , and c2 to fulfil
the homogeneous boundary conditions on G 1 and G s . From the calculation of the corresponding sub-determinant we
receive the following equation for the eigenvalues:
 
 J0 ðmÞ Y0 ðmÞ I0 ðjkjÞ K0 ðjkjÞ 

 J0 ðmsÞ Y0 ðmsÞ I0 ðjkj sÞ K0 ðjkj sÞ 
 ¼0 ð33Þ
 J2 ðmÞ Y2 ðmÞ I2 ðjkjÞ K2 ðjkjÞ 

 J2 ðmsÞ Y2 ðmsÞ I2 ðjkj sÞ K2 ðjkj sÞ 

where l runs through all the roots of (33).

 15214001, 2002, 6, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/1521-4001(200206)82:6<399::AID-ZAMM399>3.0.CO;2-6 by Curtin University Library, Wiley Online Library on [02/05/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Lee, D.-S.; Rummler, B.: Eigenfunctions of the Stokes Operator. III 405

In this way we obtain a second kind of eigenfunctions given by

0 0 1 0 1 0 1 0 11
im im ikI ðjkj rÞ ikK ðjkj rÞ
J ðmrÞ Y ðmrÞ 0 0
B Bk 0 C Bk 0 C c1 c2 C
uk0l ðx1 ; r; jÞ ¼ expðikx1 Þ@ar@ J1 ðmrÞ A þ br @ Y1 ðmrÞ A þ @ jkjI1 ðjkj rÞA þ @jkj K1 ðjkj rÞAA; ð34Þ
l l
0 0 0 0

which is written in cylindrical coordinates, too. The coefficients c1 ; c2 , and br are easily determinable as linear functions
of ar from (34). The nullity of the matrix for the satisfaction of the boundary conditions on G 1 and G s is strictly one if
l is a root of (33). (One has further to take into account (32) and aj ¼ bj ¼ 0.) Then the resolution of three arbitrary
chosen equations of the four equations from the boundary conditions results in the above mentioned coefficients which
depend linearly on the parameter ar .
The values of ar are determined with the help of requirement (19).

4.4 The case k 6¼ 0, n 6¼ 0

Omitting the factor expðiðkx1 þ njÞÞ we begin again with the general solution of the ordinary differential equation
(14). The general solution of this equation is given by
pkn ðrÞ ¼ c1 In ðjkj rÞ þ c2 Kn ðjkj rÞ ; c1 ; c2 2 R : ð35Þ

From (11)––(13) we obtain the system of ordinary differential equations for the general solution of the homogeneous
system suppressing again expðiðkx1 þ njÞÞ on both sides of the equations:


00 1 0 2 n2
ukn; x1 ðrÞ þ u ðrÞ þ l  k  2 ukn; x1 ðrÞ ¼ 0 ;
r kn; x1 r

2

00 3 0 2 n 1 2ik
ukn; r ðrÞ þ u ðrÞ þ l  k  ukn; r ðrÞ ¼  ukn; x1 ðrÞ ; ð36Þ
r kn; r r2 r


1 0 n2 þ 1 2in
u00kn; j ðrÞ þ ukn; j ðrÞ þ l  k2  ukn; j ðrÞ ¼  2 ukn; r ðrÞ :
r r2 r

The addition of a special solution of the inhomogeneous equations to the general solution of system (36) yields
the general solution of this system (11)––(13):
c1 ik c2 ik
uknl; x1 ðrÞ ¼ ax1 Jn ðmrÞ þ bx1 Yn ðmrÞ þ In ðjkj rÞ þ Kn ðjkj rÞ ;
l l
ar br ikax1 ikbx1
uknl; r ðrÞ ¼ Jn ðmrÞ þ Yn ðmrÞ þ Jn1 ðmrÞ þ Yn1 ðmrÞ ;
r r m m
c1 jkj c2 jkj
þ ðIn1 ðjkj rÞ þ Inþ1 ðjkj rÞÞ  ðKn1 ðjkj rÞ þ Knþ1 ðjkj rÞÞ ;
2l 2l
a
; b
2 C ; ð37Þ
kax1 kbx1
uknl; j ðrÞ ¼ aj Jpnffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffi
2 þ1 ðmrÞ þ bj J n2 þ1 ðmrÞ  Jn1 ðmrÞ  Yn1 ðmrÞ
m m
c1 ijkj c2 ijkj
þ ðIn1 ðjkj rÞ  Inþ1 ðjkj rÞÞ  ðKn1 ðjkj rÞ  Knþ1 ðjkj rÞÞ
2l 2l
ar im br im
þ ðJn1 ðmrÞ  Jnþ1 ðmrÞÞ þ ðYn1 ðmrÞ  Ynþ1 ðmrÞÞ ;
2n 2n
where we use the abbreviation m (cf. (30)).
It follows from the relation (3) that the vector field is solenoidal if aj ¼ 0 and bj ¼ 0 .
So the investigation of the fulfilment of the required homogeneous boundary conditions on G 1 and G s yields six
equations for the coefficients ax1 , bx1 , ar , br , c1 , and c2 . From the calculation of the corresponding determinant we
receive the following equation for the eigenvalues:
 
 mJn ðmÞ mYn ðmÞ 0 0 2jkj In ðjkjÞ 2jkj Kn ðjkjÞ 

 mJn ðmsÞ mYn ðmsÞ 0 0 2jkj In ðjkj sÞ 2jkj Kn ðjkj sÞ 

 Jn1 ðmÞ Y ðmÞ J ðmÞ Y ðmÞ In1 ðjkjÞ Kn1 ðjkjÞ 
 n1 n1 n1
 Jn1 ðmsÞ Yn1 ðmsÞ Jn1 ðmsÞ Yn1 ðmsÞ I ðjkj sÞ K n1 ðjkj sÞ 
 ¼ 0: ð38Þ
 n1
 0 0 Jnþ1 ðmÞ Ynþ1 ðmÞ Inþ1 ðjkjÞ Knþ1 ðjkjÞ  

 0 0 Jnþ1 ðmsÞ Ynþ1 ðmsÞ Inþ1 ðjkj sÞ Knþ1 ðjkj sÞ 

l runs here through all of the roots of eq. ((38). The number of roots of this equations is infinite and the nullity on
every root of (38) is one.
 15214001, 2002, 6, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/1521-4001(200206)82:6<399::AID-ZAMM399>3.0.CO;2-6 by Curtin University Library, Wiley Online Library on [02/05/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
406 ZAMM
Z. Angew. Math. Mech. 82 (2002) 6

From the general solution (37) it follows that the possible eigenfunctions have the form
0 0 1 0 1
Jn ðmrÞ Yn ðmrÞ
B B ik C B ik C
uknl ðx1 ; r; jÞ ¼ expðikx1 þ injÞ @ax1 @ m Jn1 ðmrÞ A þ bx1 @ m Yn1 ðmrÞ A
m k J ðmrÞ m k Y ðmrÞ
n1 n1
0 1 0 1
0 0
B Jn ðmrÞ=r C B Yn ðmrÞ=r C
þ ar @ A þ br @ A
im im
2n ðJn1 ðmrÞ  J nþ1 ðmrÞÞ 2n ðY n1 ðmrÞ  Y n1 ðmrÞÞ
0 0 1 0 111
sgnðkÞ iIn ðjkj rÞ sgnðkÞ iKn ðjkj rÞ
jkj B B 1 ðI ðjkj rÞ þ I ðjkj rÞÞ C B1 CCC
þ @c1 @ 2 n1 nþ1 A  c2 @ 2 ðKn1 ðjkj rÞ þ Knþ1 ðjkj rÞÞ AAA
l i ðI ðjkj rÞ  I ðjkj rÞÞ i ðK ðjkj rÞ  K ðjkj rÞÞ
2 n1 nþ1 2 n1 nþ1

written in cylindrical coordinates, too.

The coefficients ax1 , bx1 , c1 ; c2 , and br are easily determinable as linear functions of ar from (37). The nullity of
the matrix for the satisfaction of the boundary conditions on G 1 and G s is strictly one if l is a root of (38). (One has
further to take into account that aj ¼ bj ¼ 0.) Then the resolution of five arbitrary chosen equations of the six
equations from the boundary conditions results in the above mentioned coefficients which depend linearly on the
parameter ar .
Finally, the values of ar are determined with the help of requirement (19).

5. The eigenfunctions of the Stokes operator on an open gap domain

between two concentric circles

We have studied in Section 4 the cases with k ¼ 0, where the eigenfunctions are independent of the argument in the
x1 -direction. Simultaneously with the searched functions in the three-dimensional case, a simple restriction on the two-
dimensional problem provides the eigenfunctions of the Stokes operator on an open gap domain between two con-
centric circles, with homogeneous Dirichlet boundary on the inner periphery and the outer periphery of the gap do-
main.
Avoiding a repeat of the derivation of the eigenfunctions, we regard r and j as polar coordinates in the plane.
Then the eigenfunctions in the case n ¼ 0 are given by

 pffiffiffi
!
0
pffiffiffi J 1 ð l sÞ 0
pffiffiffi
u00l ðr; jÞ ¼ aj  pffiffiffi ; ð39Þ
J1 ð l rÞ Y1 ð l sÞ Y1 ð l rÞ
pffiffiffi
where l runs through all the roots of (21). The coefficients aj are determined with the use of relation (19), where
L2 :¼ ðL2 ðKðs; 1Þ ÞÞ2 .
If n 6¼ 0, we get the following solutions (cf. sub-section 4.2):
( 


n n1 1 ~ n1 1
u0nl ðr; jÞ ¼ ar expðinjÞ Cr þ Cr
l i i
0 pffiffiffi 1 0 pffiffiffi 19
Jn ð l rÞ=r Yn ð l rÞ=r >
=
B pffiffiffi C B pffiffiffi C
þ@ i l pffiffiffi pffiffiffi A þ B@ i l pffiffiffi pffiffiffi A ;
ðJn1 ð l rÞ  Jnþ1 ð l rÞÞ ðYn1 ð l rÞ  Ynþ1 ð l rÞÞ >
;
2n 2n
where l runs through all the roots of (26). We use here again representation br ¼ ar
BðlÞ, c1 ¼ ar
CðlÞ, and
c2 ¼ ar
C~ðlÞ of sub-section 4.2.
The values of ar are determined with the help of requirement (19), where L2 :¼ ðL2 ðKðs; 1Þ ÞÞ2 .

Acknowledgement

The main part of this paper was acquired during the visit of the first author to Magdeburg (446 KOR-112/2/00).

References

1 Batcho, P. F.; Karniadakis, G. E.: Generalized Stokes eigenfunctions: A new trial basis for the solution of incompressible
Navier-Stokes equations. J. Comp. Phys. 115 (1994), 121––146.
2 Boberg, L.; Brosa, U.: Onset of turbulence in a pipe. Z. Naturforsch. 43a (1988), 697––726.
 15214001, 2002, 6, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/1521-4001(200206)82:6<399::AID-ZAMM399>3.0.CO;2-6 by Curtin University Library, Wiley Online Library on [02/05/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Lee, D.-S.; Rummler, B.: Eigenfunctions of the Stokes Operator. III 407

3 Brosa, U.: Zur L€ osung von Randwertproblemen mit Vektorfeldern. Habilitationsschrift, Marburg 1985.
4 Brosa, U.: Linear analysis of the currents in a pipe. Z. Naturforsch. 41a (1986), 1141––1153.
5 Jahnke, E.; Emde, F.; Lo € sch, F.: Tafeln h€
oherer Funktionen. Teubner, Leipzig 1960.
6 Rummler, B.: Partiell-translationshomogene statistische L€ osungen der Navier-Stokes-Gleichung im Rohr und die Berechnung der
ebenen Couette-Str€omung. Dissertation, Leipzig 1984.
7 Rummler, B.: The eigenfunctions of the Stokes operator in special domains. I. ZAMM 77 (1997) 8, 619––627.
8 Rummler, B.: The eigenfunctions of the Stokes operator in special domains. II. ZAMM 77 (1997) 9, 669––675.
9 Rummler, B.: Zur L€ osung der instation€aren inkompressiblen Navier-Stokesschen Gleichungen in speziellen Gebieten. Habilita-
tionsschrift, Magdeburg 1999.
10 Temam, R.: Navier-Stokes equations, theory and numerical analysis. 3rd rev.ed. North-Holland, Amsterdam 1984.
11 Vi
sik, M. J.; Fursikov, A. V.: Mathematische Probleme der statistischen Hydromechanik. Geest & Portig, Leipzig 1986.

Received September 5, 2000, revised March 15, 2001, accepted September 17, 2001

Addresses: Prof. Dr. Doo-Sung Lee, Konkuk University, Dept. of Mathematics (Education), College of Education, 1, Hwayang-
Dong, Kwangjin-Gu, Seoul, Korea, email: [email protected]; PD. Dr. Bernd Rummler, Dept. of Mathematics,
Otto-von-Guericke-Universit€
at Magdeburg, PSF 4120, D-39016 Magdeburg, Germany,
email: [email protected]