laplace transform method in boundary value problems for the helmholtz equation
TRANSCRIPT
MECHANICS RESEARCH COMMUNICATIONS Vol. 7(3),159-164, 1980. Printed in the USA. 0093-6413/80/030159-06502.00/0 Copyright (c) Pergamon Press Ltd.
LAPLACE TRANSFORM METHOD IN BOUNDARY VALUE PROBLEMS FOR THE HEL~OLTZ EQUATION
R. Souchet Laboratoire de M~canique Th~orique Universit~ de Poitiers, France
(Received and accepted for print 15 February 1980)
Abstract
The purpose of this present note is to obtain boundary integral equation formulations of boundary value problems for the two-dimensional Helmholtz equation. The integral equations are derived using the Laplace transform method.
Introduction
Many texts present the derivation of boundary integral equations on the ba-
sis of the elementary solution for the partial differential equation con-
nected with the physical problem II, 2, 3 I . However this method provides
two real disadvantages :
| - The required elementary solutions depend on the partial differen-
tial operators.
2 - The integral equations involve singular kernels, thereby hampering
numerical analysis.
In the present paper, we propose an idea, originally posed by Picone 141
and Amerio 151, to eliminate the above difficulties. A formulation of boun-
dary value problems using the Laplace transform is shown to lead to bounda-
ry integral equations. No attempt has been made to include general partial
differential equations. Our object is to present the method on a particular
case, namely the two-dimensional Helmholtz equation. Guided by this example,
we expect that other boundary value problems, e.g. in elasticity, will be
solved in such a way.
Moreover, it is of interest to test the method, using a computational pro-
cedure. This was made on a Neumann problem for the Laplace operator 161.
159
160 R. SOUCHET
The numerical results are simplified for two reasons :
I - The kernels of integral equations do not depend in essential way
on the differential operator.
2 - The integral equations involve regular kernels. So the actual me-
thod does not possess the above disadvantages of the singular integral
equations.
~erio bound.ayy formula
Let 12 denote a bounded domain in ~2 with a piece-wise smooth boundary ~.
Let (xl, x 2) be cartesian coordinates and let (xl, x2) be the outward unit
normal on ~. We consider the Helmholtz equation
- ~u + k2u = f , k ~ 0 (I)
where f represents a given, continuous function. In what follows, we shall
made use of the second Green formula
(u .&v - v.Au) dXldX 2 = (u "~-~n - v . ds . (2) ~ ~ dn
We should therefore require that any solution u of equation (I) is availa-
ble for formula (2). Let E denote this class of solutions.
Let u be a solution of class E. It is well known 141 that u can be charac-
terized by its Laplace transform
I xl + 2) u(x ,x2) dxldx 2 u ( p l , P 2 ) = e x p (Pl P2 x " 1 (3) ~
I t s h o u l d b e n o t e d t h a t p l , P 2 a r e two i n d e p e n d a n t c o m p l e x v a r i a b l e s . I n o r -
d e r t o d e t e r m i n e u , we f o l l o w a p r o c e d u r e due t o P i c o n e 14[ . The w e i g h t i n g
f u n c t i o n v i n f o r m u l a (2 ) i s a s s u m e d t o b e t h e L a p l a c e k e r n e l
L ( P l , p 2) ( X l , X 2) = e x p ( P l X l + P2X2 ) (4 )
H e n c e we f i n d
(p l 2 2 u • I - + P2 - k2) (PI'P2) = j f" L(PI'P2) dXldX2
I d L ( P l ' P 2 ) I an - u . dn ds + - - . L ( P l , p 2) ds (5 ) ~ j ~ dn
k 2 ~ . 2 2 _ # O , i t i s c l e a r t h a t we c a n f i n d u zn t e r m s o f t h e b o u n - When Pl + P2 du
d a r y v a l u e s u and ~-~. N e x t , when we h a v e
2 _ k 2 Pl 2 + P2 = O (6)
i t f o l l o w s t h a t a s o l u t i o n u o f c l a s s E s a t i s f i e s t h e i d e n t i t y
LAPLACE TRANSFORMS FOR HELMHOLTZ EQUATION 161
J d L ( P I ' P 2 ) f du I u . dn ds - d--~ " L ( P l ' P 2 ) d s - f . L ( P l , p 2) dXldX 2 = 0 (7) ~ ~ ~
This i m p o r t a n t i d e n t i t y was d e r i v e d w i t h o u t undue e f f o r t . However, Amerio
151 p o i n t s ou t the need f o r a supp l emen ta ry c o n d i t i o n in o r d e r to g u a ran -
t e e un iquenes s : the domain ~ must be s imply c o n n e c t e d . Thus we have the
f o l l o w i n g theorem
Theorem 1 - A f u n c t i o n u o f c l a s s E i s s o l u t i o n of e q u a t i o n (1) in a s im-
p ly c o n n e c t e d bounded domain ~, i f and on ly i f the boundary v a l u e s u and
du 2 P22 k 2" d-~ s a t i s f y the Amerio boundary fo rmula (7) when Pl + =
We c onc lude t h a t e q u a t i o n (7) p r o v i d e s i n t e g r a l e q u a t i o n s f o r boundary
v a l u e p rob lems . Thus, g i v e n u on 2£ ( D i r i c h l e t p r o b l e m ) , the fo rm u la becomes du
an i n t e g r a l e q u a t i o n f o r ~nn " N e v e r t h e l e s s , e q u a t i o n (7) i s no t a boundary
i n t e g r a l e q u a t i o n i n the usua l s e n s e .
Boundary integral equations
2 P2 2 k 2 For simplicity, we shall assume that k # O. It follows that Pl + =
if and only if
Pl = ~I k( ~p + P) ' P2 = ~ k( - p) , p # 0 (8)
where p is a complex variable• Hence the Amerio boundary formula reduces
to
k [ (! -- k (i z + p ~)} ds ~ J u. n + p n). exp {7 p
- ~ P
_ [ du k 1 ~ ~-~n" exp {7 (~ z + p ~)} as
f k (~ + ~)} = 0 - f . exp {7 p Z P dXldX 2 (9)
(z = x! + i x2, ~ = x I - i x 2 ; n = n I + i n2, ~ = n! - i n2).
Next we wish to introduce the following function of the complex variable p
(p # O)
f k 1 -- _ du} k (.~ + ~)} as F(p) = {~ u(~ n + p n) dn " exp {~ P Z P ~
f k (! -- • z + p z)} dXldX 2 (IO) - f exp {7 p
162 R. SOUCHET
Now we consider the boundary ~ as a finite number of arcs C, so that the
integrand in (;0) possesses the required properties on each C in order to
have an analytic function. Hence the function F is analytic throughout
the complex plane except at the isolated singularity p = 0. Thus the
Amerio boundary formula takes the reduced form
F ( p ) = 0 , p ~ ~ 2 \ { 0 } . ( 1 1 )
Now we attempt to find a boundary integral equation. We derive from (11)
Re F(p) = O , p ~ ~, (O ¢ ~)" (12)
So, the truth of (12) is a necessary condition for the truth of (ll). But
it is also a sufficient condition. Indeed, if we take O # ~ u ~, ReF is
a harmonic function in ~, such that (12).
Hence, we have a Dirichlet problem ; therefore
Re F(p) = O , p ~ ~. (13)
~:~
ff
-~ ~:~
a ) F i r s t c a s e : 0 ~ S~ ~ SS2 ; ~ i s a s i m p l y - c o n n e c t e d r e g i o n .
It follows that
F(p) = i K , p ~ ~ (14)
where K is a real constant. Then
F(p) = iK , p ~ C\{O} (15)
using the principle of analytic continuation. But we observe that F(p) is a
real number provided that p = I (in fact IPl = I) ; thus K = O.
In the case 0 E ~, the proof must be modified. We emphasize that
Im F(p) = O , IPl = I , (16)
LAPLACE TRANSFORMS FOR HELMHOLTZ EQUATION 163
so that
d__ Re F(p) = 0 Ipl = i. (17) dn '
If ~ coincides with the unit circle, from the two conditions (12), (16),
(i.e. F(p) = 0 , IPl = I), all the coefficients of the Laurent series of
F are zero. Thus F = O.
If ~ is an arbitrary contour, with the two conditions (12), (17), i.e.
d Re F(p) = 0 , p ~ ~ and ~n Re F(p) = 0 , IPl = I ,
we have a mixed problem for Re F in each region ~' limited by ~ and the
circle IPl = ]" Hence
Re F(p) = 0 , P 6 ~' , (18)
and clearly
Re F(p) = 0 , IPl = I . (19) ~,
~
b) An example of the second case : O ~ ~ ; ~ is a simply connected region.
Hence all the coefficients of the Laurent series of F are~ zero. Thus again
F = O.
Let us now introduce the following kernels
k (! k (_; z + p 7)}} (20) = n + p ~) exp {-~ P R(p,x) -~ Re { P
k (± z + p 7)}} (21) S(p,x) = Re {exp {~ P
where x = (x 1,x 2). Then we find the following basic result.
Theorem 2 - A function u of class E is solution of equation (I) in a simply du
connected domain ~ if an only if the boundary values u and ~n satisfy the
boundary integral formula
R(p,x) . u(x) ds- S(p,x).~-~n (x) ds - S(p,x).f(x) dx= O (22)
x ~ ~ x ~ ~ x ~!~
164 R. SOUCHET
where p is a (complex) variable specifying points on the boundary $~.
We observe two important facts : the kernels (20), (21) are regular when
p and x describe ~ and were trivially derived from the Laplace kernel
exp (plx! + P2X2 ). Since the formula (22) involves the boundary values u du
and ~ , we conclude that formula (22) provides boundary integral equations
for boundary value problems. Thus, given u on ~ , the formula (22) becomes du
a boundary integral equation for the basic unknown ~ . We would like to
repeat that (22) is sufficient for a simply connected domain.
Finally, one should note the following remarks :
; - In theorem 2, we can replace the condition p d ~ by p 6 ~, where
~ is a domain having the same properties that ~ 161.
2 - For the Laplace equation - &u = f , a similar boundary integral
equation 161 can be derived from the Amerio boundary formula (7).
3 - If the domain ~ is multiply connected, we must consider simply con-
nected sub-regions : the boundary integral equation (22) is used for each
sub-domain and the continuity of the functions and their normal derivatives
is assigned along common boundaries.
Conclusion
The formulation discussed in the present paper has proved successful with a
torsion problem of Saint-Venant involving computational procedure 161. The
absence of any singularities in the integrand simplifies the numerical treat-
ment, since no analytical approximation need be made to the integral over
some neighborhood of a singularity. It should also be pointed out that the
actual method can be extended to others boundary value problems, e.g. in
classical elasticity.
References
I! M.A. Jaswon, Proc. Roy. Soc. A, 275, 23 (!963)
12 F.T. Rizzo, Quart. Appl. Math., 25, 83 (|967)
13 J.Co Lachat and J.O. Watson, Intern. J. for Num. Meth. in Engin., IO, 991, (1976)
14 M. Picone, Rend. Acc. Lincei, 2, 365, (1947)
15 L. Amerio , Rend. Ist. Lombardo di Sci. e Lettere, 78, 79, (!944)
16 R. Souchet and D. Robinaud, to be communicated at the Second Int. Semo on Recent Advances in Boundary Element Methods, Southampton, England, (1980).