TJMM 1 (2009), No.

1-2,

01-10

Greens Functions in Mathematical Physics

W ILHELM K ECS A BSTRACT. The determination of Green functions for some operators allows the effective writing of solutions to some boundary problems of mathematical physics.
2010 Mathematics Subject Classication. 34B27, 42A38. Key words and phrases. Greens functions, Fourier transform.

1. I NTRODUCTION The solution of many problems of mathematical physics is related to the construction of Greens function with the help of which the solutions of the boundaryvalue problems may be determined explicitly and presented in an integral form. Thus, the Green functions are closely connected to the fundamental solution of the linear differential operator corresponding to a given problem. In the case in which the fundamental solution is a distribution, we have to deal with Greens distributions. To illustrate the construction of a Green function as well as its connection to the fundamental solution of the operator corresponding to the given boundaryvalue problem, we shall consider the heat conduction equation for an innite bar, the generalized Poisson equation and the vibrating string equation. 2. H EAT CONDUCTION EQUATION We shall consider a cylindrical homogeneous and isotropic bar whose lateral surface is isolated from the rest of the medium. We take the symmetry axis of the bar as Ox-axis. We shall denote by u(x, t), x R, t 0 the temperature of the bar in the point x at the moment t, and by , c, k the density, the specic heat and the thermal conductivity. We use the Fourier-Newton law regarding the amount of heat owing in the unit time across the unit cross section of the bar, then, in the absence of internal sources of heat, the heat conduction equation is u(x, t) 2 u(x, t) = a2 , x R, t > 0, t x2 where the constant a2 has the expression a2 = k /c. We consider the parabolic equation (1) with the initial condition u(x, 0) = (x), C 0 (R). (2) (1)

The equation (1) with the condition (2) represents the boundary value problem of the heat conduction in an innite homogeneous bar.
1

Wilhelm Kecs

The boundary value problem has unique solution for the given function . The problem solution can be represented in the form u(x, t) = L((x)) = G(x, , t)( )d = (( ), G(x, , t)) , (3)

where the kernel of the integral operator L is G(x, , t) and is called Green function. To obtain the Green function, we will consider (x) = (x x0 ), where x0 R is a parameter. From (3), it results u(x, t) = ( ( x0 ), G(x, , t)) = ( ( ), G(x, + x0 , t)) = G(x, x0 , t), hence u(x, t) = G(x, x0 , t). (5) Consequently, the Green function G(x, x0 , t) represents the solution of the equation (1) with the condition u(x, 0) = (x x0 ). Hence, the Green function G(x, , t) satises the equation (1) G(x, , t) 2 G(x, , t) = t x2 and the initial condition G(x, , t)|t=0 = (x ).

(4)

(6)

(7)

(8)

Let E (x, t) D (R R) be the fundamental solution of the equation (1), hence E 2E a2 2 = (x, t) = (x) (t). (9) t x Applying the Fourier transform with respect to x to the equation (9) we obtain d (, t) = (t), E (, t) a2 (i)2 E dt (10)

(, t) = Fx [E (x, t)]. where E (, t) is the fundamental soluThe equation (11) shows that the distribution E tion of the operator d + a2 2 . (11) dt (, t), we consider the homogeneous equation To determine the distribution E dV (t) + a2 2 V (t) = 0, dt the general solution of which is V (t) = C ea
2

(12)

2 t

(13)

Imposing the condition V (0) = 0, we obtain C = 1 and consequently the fundamental solution of the operator (11) is V (t) = H (t)ea
2

2 t

(, t), =E

(14)

Greens Functions in Mathematical Physics

where H represents the Heaviside function { 0, x < 0, H (x) = 1, x 0. Applying the inverse Fourier transform, we yield ( ) [ ] x2 (t) (, t) = H exp 2 , x R, t R, E (x, t) = F1 E 4a t 2a t ( 2) 2 2 x because F1 [eb x ] = | b| exp 4b2 . Considering t as parameter, from (14), we obtain, for t > 0 ( ) 1 x2 Et (x) = exp 2 . 4a t 2a t

(15)

(16)

(17)

We remark that Et (x), t > 0 is an element of the test space S (R) of indenitely differentiable functions which, for |x| , approach zero together with all their 1 derivative of any order, faster than any power of |x| . It satises the relation u2 E (x)dx = 1, because R e du = . R t By direct calculation it is veried that Et is the solution of the equation (1) and satises the initial condition Et (x)|t=+0 = (x). Indeed, taking into account (14), we have [ ] (, t) = 1 = Fx [ (x)], lim Fx [E (x, t)] = Fx lim Et (x) = lim E
t+0 t+0 t+0

(18)

(19) (20)

hence
t+0

lim Et (x) = (x).

( ) (x)2 exp 4a2 t , 2a t

Taking account of (7), (8) and (17), we obtain G(x, , t) = Et (x ) =

t > 0,

(21)

E (x ) = H (t)G(x, , t). These relations show the dependence between the Green function and the fundamental solution of the heat conduction equation in a homogeneous innitely bar. Consequently, the solution of the boundary value problem (1), (2) is given by the convolution with respect to x R u(x, t) = Et (x) (x) = Et (x )( )d = G(x, , t)( )d, (22)

where is considered to be a bounded function and lim u(x, t) = (x) in all

t+0

continuity points of the function . All the results can be generalized to Rn . Thus, the heat conduction equation in n R is u(x, t) = a2 u(x, t), t > 0, a > 0, x Rn , (23) t

4 2 x2 1 2 x2 n

Wilhelm Kecs

where = + ... + represents the Laplace operator. The initial condition is u(x, 0) = (x), C 0 (Rn ).
n

(24)

The Green function G(x, , t), x, R , t > 0 for the boundary value problem (23), (24) represents the solution of the equation (23) with the initial condition G(x, , t)|t=+0 = (x ).
n

(25)

Let Et (x), x R , t > 0, be the solution of the equation (23) with the initial condition Et (x)|t=+0 = (x). (26) Then, the Green function G(x, , t) is given by the relation G(x, , t) = Et (x ). The boundary value problem (23), (24) is u(x, t) = Et (x) (x) = Et (x )( )d =
Rn t+0

(27)

G(x, , t)( )d

Rn

(28)

where is considered a bounded function and lim u(x, t) = (x) in all continuity points of the function . We remark that the solution Et (x) of the equation (23) with the condition (26) can be obtained applying the Fourier transform with respect to the variable x Rn . Thus, applying the Fourier transform to the equation Et (x) = a2 Et (x) t and the condition (26), we obtain d d 2 Et () + a2 E Et () t ( ) = 0, dt dt (29)

= 1,
t=+0

(30)

t () = F [Et (x)] (), Rn . where E The solution of the problem (30) is the function ( ) t () = exp a2 2 t , t > 0. E According to the formula ) ( 2 [ ( )] ( )n 2 2 F exp b x = exp 2 , b > 0, b 4b we obtain [ ] [ ] t () (x) = F1 exp(a2 2 t) Et (x) = F1 E ) ( 2 1 x = . exp 2 4a t (2a t)n

(31)

(32)

(33)

Hence, the Green function corresponding to the boundary value problem of the heat conduction in Rn is ) ( 2 1 x . (34) G(x, , t) = Et (x ) = exp 4a2 t (2a t)n

Greens Functions in Mathematical Physics

The formulae (28), (34) generalize those obtained for the case n = 1. Finally, we observe that Et (x) S (Rn ), t > 0, and we have Rn Et (x)dx = 1. Let E (x, t) D (Rn+1 ) be the fundamental solution of the equation (23), hence of the operator /t a2 of heat conduction in Rn . Proceeding as in the case n = 1, we can establish the relation E (x, t) = H (t)Et (x). Thus, we can write E (x, t) a2 E (x, t) = (x, t) = (x) (t). t 3. G ENERALIZED P OISSON EQUATION The generalized Poisson equation is u(x, y, z ) k 2 u(x, y, z ) = f (x, y, z ), (37) where is the Laplace operator, k = const, and f is a distribution from D (R3 ). For k = 0 we obtain the Poisson equation u(x, y, z ) = f (x, y, z ). (38) By denition, the Green function corresponding to the equation (37) represents the solution of the equation u(x, y, z ) k 2 u(x, y, z ) = (x , y , z ), (39) where , , are parameters. Let E (x, y, z ) D (R3 ) be the fundamental solution of the operator k 2 ; hence, we have E (x, y, z ) k 2 E (x, y, z ) = (x, y, z ). (40) The Green function G(x, y, z ; , , ) satises the equation G(x, y, z ; , , ) k 2 G(x, y, z ; , , ) = (x , y , z ) and, consequently, the dependence between G and E is G(x, y, z ; , , ) = E (x , y , z ). (42) This relation shows that the Green function is obtained by a translation of the distribution E (x, y, z ) to the point (, , ). Applying the Fourier transform to the equation (40) in R3 , we obtain (, , ) k 2 E (, , ) = 1 (2 + 2 + 2 )E (43) (, , ) and , , are real variables. where F [E (x, y, z )] = E From (43) it results 1 (, , ) = . (44) E k 2 + 2 + 2 + 2 We shall denote by Fx [ ], Fy [ ], Fz [ ] the Fourier transform in R, with respect to the variables x, y, z , respectively; we have [ ] ( ) 1 Fx exp(kr) = 2K0 2 + k 2 y 2 + z 2 , (45) r where K 0 represents the modied Bessel function of second kind and zero order and r = x2 + y 2 + z 2 is the radius vector. (41) (35)

(36)

Wilhelm Kecs

Also, we can write

) ( [ ( )] exp |z | 2 + 2 + k 2 2 + k 2 y 2 + z 2 = Fy K0 , 2 + 2 + k 2 ) ( exp |z | 2 + 2 + k 2 2 = . Fz 2 + 2 + 2 + k 2 2 + 2 + k 2

(46)

(47)

Because F[ ] represents the Fourier transform in R3 , with respect to all variables, we have [ ] 1 4 F exp(kr) = 2 , (48) r + 2 + 2 + k2 wherefrom [ ] 1 1 1 F = exp(kr), (49) 2 2 2 2 + + +k 4r r being the radius vector. Taking into account (44), we obtain [ ] (, , ) = 1 exp(kr); E (x, y, z ) = F1 E (50) 4r in accordance with the relation (42), the Green function corresponding to the generalized Poisson equation has the expression 1 G(x, y, z ; , , ) = exp(k), (51) 4 where = (x )2 + (y )2 + (z )2 . (52) Particularly, for k = 0, we obtain the Green function corresponding to the equation (38) in the form 1 G(x, y, z ; , , ) = E1 (x , y , z ) = , (53) 4 where E1 (x, y, z ) = 1/4r represents the fundamental solution of the Laplace operator in R3 , hence E1 (x, y, z ) = (x, y, z ). (54) With the help of the Green function (51), the solution of the generalized Poisson equation is u(x, y, z ) = E (x, y, z ) f (x, y, z ), (55) where f D (R3 ) is a distribution with compact support. If f is a distribution of function type for which the convolution product E (x, y, z ) f (x, y, z ) exists, then, developing the convolution product (55), we obtain u(x, y, z ) = E (x , y , z )f (, , )d d d R3 = G(x, y, z ; , , )f (, , )d d d = L (f (x, y, z )) , (56)
R3

where the kernel of the integral operator L is the Green function.

Greens Functions in Mathematical Physics 3

7 2

Remark 1. . The fundamental solution E (x, y, z ) D (R ) of the operator + k can be obtained using the derivative rule of the homogeneous functions having at the origin, x = 0, a singular point. Then, according with [21], p.126, the function h(x) = exp(ikr)r2n , x Rn {0}, r = x , k C, n 3, is locally integrable, and the functions xj r1n , xj rn are homogeneous functions. It is shown the relation h(x) h(x) + k 2 h(x) = ik (n 3) (n 2)S1 (x), (57) r / where S1 = 2 n/2 (n/2) represents the area of the unit sphere from Rn . Particularly, for n = 3, S1 = 4 and considering k = ik1 , we obtain the relation
2 h1 (x, y, z ) k1 h1 (x, y, z ) = 4 (x, y, z ),

(58) (59)

where h1 (x, y, z ) = From (58), it results

1 exp (k1 r) , r = x2 + y 2 + z 2 . r

( ) ( ) h1 h1 2 k1 = (x, y, z ); 4 4 h1 (x, y, z ) 1 = exp (k1 r) . 4 4r

(60)

2 is hence, the fundamental solution of the operator k1

E (x, y, z ) =

(61)

4. G REEN S FUNCTION FOR THE VIBRATING STRING For small vibrations of an homogeneous string, in the absence of the external forces, the motion equation of hyperbolic type is
2 2 u(x, t) 2 u(x, t) a = 0, (62) t2 x2 where the constant a2 has the expression a2 = T /, where T is the tension in the string, and is the density of the string per unit of length. In the case of an innite string the Cauchy problem consists in the determination of the function u(x, t) C 2 (R R+ ) which satises the equation (62), as well as the initial conditions u(x, t) u(x, t)|t=0 = (x), = (x), (63) t t=0

where , C 0 (R). We remark that the vibrating string (62) as well as the initial condition (63) can be considered in the distributions space. The solution of the Cauchy problem will be thus the distribution u(x, t) D (R) depending on the parameter t 0; , are distributions from D (R). The Green function corresponding to the Cauchy problem (62) and (63) for the vibrating string is G(x, , t) C 2 (R R+ ), x R, t 0, where R is a parameter, satisfying the conditions 2G 2G = 0, x R, t > 0, t2 x2 (64)

Wilhelm Kecs

G(x, , t)|t=+0 = 0,

G(x, , t) t

= (x ).
t=+0

(65)

Let Et (x) = E (x, t) C 2 (R R+ ) be the solution of the equation (62), hence

2 2 2 E ( x ) a Et (x) = 0, t t2 x2 which satises the conditions = (x). Et (x) Et (x)|t=+0 = 0, t t=+0

(66)

(67)

In this way, between the functions G(x, , t) and Et we have g (x, , t) = Et (x ). (68) With the help of Greens function, the solution of the Cauchy problem (62) and (63) for the vibrating string can be write in the form (Et (x) (x)) u(x, t) = Et (x) (x) + t = Et (x ) ( )d + Et (x )( )d. t R R Taking into account (68), we have u(x, t) = G(x, , t) ( )d + G(x, , t)( )d. t R R

(69)

(70)

We shall denote by P(D) = P (t , x ) the operator corresponding to the vibrating string, namely 2 2 P(D) = 2 a2 2 . (71) t x Because 2 2 Et (x) 2 Et (x) P(D)Et (x) = a = 0, t2 x2 on the basis of differentiation rule of the convolution product we obtain [P(D)Et (x) (x)] = 0. (72) t From here it results that the two terms of the solution (69), Et (x) (x) and (Et (x) (x))/t are solutions of the vibrating string equation (62). Taking into account the conditions (67) we obtain P(D)u(x, t) = P(D)Et (x) (x) + u(x, t)|t=+0 = Et (x)|t=+0 (x) + Et (x) (x) t t=+0 = 0 (x) + (x) (x) = (x), 2 Et (x) (x) t2 t=+0 t=+0 = (x) (x) + 0 (x) = (x), (x) + = 0.
t=+0

(73)

u(x, t) t because

=
t=+0

Et (x) t

(74)

2 Et (x) t2

Greens Functions in Mathematical Physics

Indeed, from the equation 2 Et (x) 2 Et (x) a2 = 0, 2 t x2 by passing to the limit, we obtain [ ] 2 2 2 Et (x) 2 2 = a lim E ( x ) = a (0) = 0. lim t t+0 t2 x2 t+0 x2 (75)

(76)

We shall determine now the expression of the function Et (x), x R, t > 0. Applying the Fourier transform with respect to the variable x R to the equation (75) and the conditions (67), we obtain
d2 dt2 Et ()

t () = 0, + a2 2 E = 0,
d dt

t () E

t=+0

t () E

(77) = 1.
t=+0

The solution of this Cauchy problem is t () = sin(at) . E a Applying the inverse Fourier transform, we obtain [ ] { 1 1 2a , |x| at, Et (x) = Fx Et () = 0, |x| > at. (78)

(79)

Indeed, we can write at 1 1 at Fx [Et (x)] = eix Et (x)dx = eix dx = eix at 2a at 2ix R ) 1 ( ita 1 = e eita = sin(at). (80) 2ia a Taking into account (68), we shall obtain for Greens function corresponding to the Cauchy problem of the vibrating spring the expression { 1 { 1 2a , |x | at = 2a , [x at, x + at], G(x, , t) = Et (x ) = 0, |x | > at 0, / [x at, x + at]. (81) Thus, (70) becomes x+at x+at 1 1 u(x, t) = ( )d + ( )d. (82) 2a xat 2a t xat Applying the Leibniz formula of differentiation of integrals depending on a parameter we obtain x+at 1 (x + at) + (x at) + ( )d, (83) u(x, t) = 2 2a xat which represents the dAlemberts formula. We remark that the formula (69) take place in the case when the equation (62) is considered in the distributions space D (R), where u(x, t) is a distribution from D (R), t > 0 being the parameter, and , from the initial conditions (63) being distributions with compact support.

10

Wilhelm Kecs

Let E (x, t) D (R R) be the fundamental solution of the operator (71), hence

2 2 E (x, t) 2 E (x, t) a = (x, t) = (x) (t). t2 x2 The expression of E (x, t) is t < 0, 0, 1 1 , 0 |x| at, H (at |x|) = E (x, t) = 2a 2a 0, |x| > at 0,

(84)

or

E (x, t) = H (t)Et (x). (85) Indeed, applying the Fourier transform with respect to the variable x R, we obtain d2 (, t) = (t), E (, t) + a2 2 E (86) dt2 (, t) = Fx [E (x, t)]. where E (, t) is the fundamental solution of This relation/ shows that the distribution E 2 2 2 2 the operator d dt + a . But, the fundamental solution of this operator is the distribution (, t) = H (t) sin (at) . E (87) a Applying the inverse Fourier transform we obtain [ ] 1 1 E (x, t) = F E ( , t ) = H (at |x|) , (88) x 2a this because R R) F [H (R |x|)] = R eix dx = 2 sin( , { H (R |x|) = From (85) and (68), it results E (x , t) = H (t)Et (x ) = H (t)G(x, , t), (90) fact which shows the dependence between the fundamental solution E (x, t) of the vibrating spring equation (62) and the Greens function corresponding to Cauchy problem (62), (63). R EFERENCES
[1] Schilov, G.E.,Generalized Functions and Partial Differential Equations, Gordon and Breach Science Publishers, 1968 [2] Kecs, W.W., Teoria distribut , iilor s , i aplicat , ii, Editura Academiei Romne, Bucharest, 2003 a, Bucures [3] Nowacki, W., Dinamica sistemelor elastice, Editura Tehnic , ti, 1969 U NIVERSITY OF P ETROS , ANI D EPARTMENT OF M ATHEMATICS , II N O 20, 332006, P ETROS U NIVERSITAT , ANI , R OMNIA

0, R < |x| , 1, |x| R.

(89)