The Green's Functions for the Rectangle Obtained by the Finite Fourier Transformations Author(s): A. W.

Jacobson Reviewed work(s): Source: Proceedings of the American Mathematical Society, Vol. 1, No. 5 (Oct., 1950), pp. 682686 Published by: American Mathematical Society Stable URL: http://www.jstor.org/stable/2032301 . Accessed: 15/10/2012 04:44
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

American Mathematical Society is collaborating with JSTOR to digitize, preserve and extend access to Proceedings of the American Mathematical Society.

http://www.jstor.org

THE GREEN'S FUNCTIONS FOR THE RECTANGLE OBTAINED BY THE FINITE FOURIER TRANSFORMATIONS
A. W. JACOBSON

1. Introduction. When the finite Fourier transformation method is employed, the solutions of very general boundary value problems (in contrast to initial value and mixed problems) can be expressed in terms of special functions.1 In this paper we propose to derive, in terms of a special function to be introduced, a new formula for the Green's function for the Laplace equation Au = 0 in a rectangle such that u vanishes on the boundary. 2. The function B12(x, u). In this section we introduce a key function and obtain its inverse Fourier cosine transform. If j r j <1, then log (I+reiO)= - Z,=l (-r)Pei1'/v. Also if R2=(1+r cos 0)2 +r2 sin2 0 and tan cb=r sin 0/(1 +r cos 0), then log Re'
(1)
=

log (1 + rei0)
??_

(-r)

_eiP

1=1

Equating the real part of (1), we get

log R = log [1 + 2r cos 0 + r2]1/2

(2)
1=1V

I( 1)'r"cos vG

When r| < 1, the series (2) is a Fourier cosine series. The Fourier cosine transforms, defined by the operation C{ F(x) } = f*oF(x) cos nx dx=f,(n) (n0=, 1, 2, . . - ), are except for the factor r/2 the coefficients of a Fourier cosine series. Hence, in view of (2), we get
{ I1 + 2r
(n

cos+

= 1, 2, *)

=O
It can be shown that (-1)

(n= 0).

fe(n) = C{ F('7r-x)}. Using this fact and

Received by the editors April 18, 1949 and, in revised form, June 27, 1949. 1 A. W. Jacobson, A generalizedconvolution thefinite Fouriertransformations, for Thesis (Microfilmed),University of Michigan, 1948. Also see Solutionto steadystate temperature problemswith the aid of a generalized fourier convolution,Quarterly of Applied Mathematics, vol. 7 (1949).

682

THE GREEN'S FUNCTION

683

putting r = e', 0= x, we obtain when z 0 the following formula: (3)

Clog

ex ) 2(cosh z- cos x)J

7rSre-n

n =0

(n= 1,2, ..)

(n= 0).

Let bl(n, u) denote the function c 1 cosh nu = (4) {eI[(2P+1)yo-u]n n sllnh nyo po n

e-[(2P+1)lo+u]n}.

Also let B12(x, u) denote the inverse cosine transform C-1 {bi(n, u) }, of the function b1(n, u). Wheni - u _ 0, we have according to formula (3) 00 1 B12(x, U) = - log II ? e2 (5)T
(2+1) yo

4 [cosh (2vyo+yo-u) -cos x] [cosh (2pyo+yo+u) -cos x] 3. A boundary value problem. Let it be required to determine the temperature V(x, y) in a rectangle R when the source function H(x, y) is a prescribed, sectionally continuous function of x and y and such that V vanishes on the boundary. That is, let V(x, y) be the solution of the following problem: (A) V2x + Vyy = H(x, y) in R, V(x, +0) = V(x, yo- 0) = 0 V(+O, y) = V(7r-0, y) =0

(O < x < 7),

(O < y < yo).

The Fourier sine transformation, defined by the operation S{ F(x) I * ), in particular, yields when =fofF(x) sin nxdx =f8(n) (n =1, 2, applied to the second derivative of F(x) the formula S {F"(x) }
=
-

n2f,(n)

+ n [F(O) -

(-1) F(7r)].

Applying the sine transformation with respect to x to the equations in (A), we obtain in view of the last condition the following transformed problem:
d2v

(Al)

dy2

n2v(n, y) = =

h(n, y),

v(n, 0) where k(n, y)

v(n, yo) = 0, S{H(x, y) }.

684

A. W. JACOBSON

[October

The solution of the transformed problem, in terms of the Green's function g(n, y, ,u), is2 (6) where
g(n, y,

v(n, y) = f

g(n, y, A)h(n, 14)d/2,

A) =

sinh n, sinh n(yo

--

y)
GI < Y)

n sinh nyo sinh ny sinh n(yo n sinh nyo

-

A)

or g(n, y, cosh n(yo-y--p)

,u) =.-.(JA

cosh n(yo-y

+ p)

2n sinh nyo

2n sinh nyo

y),

and the same expression with y and ,u interchanged when u>y. According to formula (4) this can be written g(n, y, ,u) = 2-1b1(n, yo - y - A) - 2-1b(n,
yo
-

y + ,u),

(,u ? y).

The inverse cosine transform of bi(n, u) is the function B12(x, u). Hence, if C>1Ig(n, y, ,) } =G(x, y, ,), we have
(7)
G(x, y, ,u) = 2-1B12(x, YO

A)

2-'B12(x,

YO-

y + ,u).

The solution (6) of the transformed problem can then be written

Yo

(8)

SI V(x, y)} = f

C {G(x,

y, A)}H(x, I

A)} d.

The product of the two transforms in (8) can be expressed in terms of the sine transform of the convolution of the two functions. For if P(x) is even and Q(x) is an odd sectionally continuous function, and = if Q(x + 2w7r) Q(x), then C{P(x) }S{Q(x)

}=

2-1S{P(x)*Q(x)

},

where the convolution is defined thus:'

P(x) * Q(x) = f
Hr

P(x

X)Q(X)dX.

Hence the solution

2

(8) becomes,

upon inverse transformation,

E. L. Ince, Ordinarydifferentialequations,Dover, 1944, p. 258. in V. Churchill, Modernoperationalmathematics engineering,McGraw-Hill, 1944, p. 274.

3 R.

1950]

THE GREEN'S FUNCTION

1 rYO

685

V(x, or

y) =
-

2o

G(x, y, ,u)*H(x, ,)dy,

V(x, y) =

1 rYo r T 2 JoJ

II

G(x-X,

y, /A)H(X, A)dXdl/2.

Since H(X, ,) is extended as an odd function of X, we get

1 I r voeT -,G(x [G(x - X, y,O) V(x, y) = -I

2 J0J

+X, y, At)]H(X,,u)dAd/2.

y, ,u)-G(x+X, y, ,u)]. Then the Let r(x, y, X, A)=2-1[G(x-X, function r is the Green's function for the boundary value problem (A). In terms of our function B12(x, u), formula (7), it can be written P(x, y, X, )
=4-1
[B2(-

X, YO+

y
y -)

- B2(x-

X, yO-y Y+

+ ,u)
/)]

B12(x

X, YO -

+ B12(x + X, Yo -

4. Relation between the function B12and the Weierstrass's sigma function. The Green's function K for Au =0 in the rectangle R: O<x<7r, O<y<yo, with u=O on the boundary is in terms of the Weierstrass's sigma function c: (10) where z = x + iy, =X +4,
=X-iM,

K(x, y, X, IA)

= -

l o(zlog 2-7r a-(z1 9

1, CO C2)0(Z ) CO CO2) a(Z 1,

+ +

w,CO 2)
CO2)

, C1,

C1=

r,

iY.4

The expression for the Green's function r for the same problem, obtained in the preceding section, is given in terms of the function B12by formula (9). The two functions, K and r, are equivalent. Comparing the two expressions above, it can be shown' that the function B12 is related to the sigma function. For the various arguments of
B12

the following relations hold:

B12(X-

X, yO -

y-

-I-log

2
7r

I o(z -

+-[(X _r(y+)2] -)2

+C

4R. Courantand D. Hilbert, Methoden der mathematischen Physik I, Berlin, 1931,

p. 335.
5

A. W. Jacobson, A generalized convolution for the finite Fourier transformations,

Thesis, University of Michigan, 1948, pp. 76-84.

686

A. W. JACOBSON

2
B2(-X, Yo y + u) =
-

-log
ir

I (z +
-

1)|
A)2] + Cl

+-[(X

)2 -(y

B12(X+ XI YO

2
y
= 7r

log I oi(z +
+ X%)2- (Y + Au)2] + Co

7r
+-[(X

and
B12(x + A,
-y

U) =

-log
2r

Ia(z [(X + X)2

-

+ 7r

(y- _

)2] +

C,

where the constants q and C are defined as follows:

7r 1r 0

12 and C = - log 2
WY 7U 1EI

sinh2 vyo C

log

II (1 -

2vo).

WAYNE UNIVERSITY