arXiv:1201.6059v1 [physics.

class-ph] 29 Jan 2012

Greens functions for Neumann boundary conditions

Jerrold Franklin Department of Physics Temple University, Philadelphia, PA 19122-6082 January 31, 2012

Abstract Greens functions for Neumann boundary conditions have been considered in Math Physics and Electromagnetism textbooks, but special constraints and other properties required for Neumann boundary conditions have generally not been noticed or treated correctly. In this paper, we derive an appropriate Neumann Greens function with these constraints and properties incorporated.

Introduction

The Greens function method for solving Sturm Liouville problems of the form1 Lr (r) = [p(r)(r)(r)] + q (r)(r) = (r) (1) is described in many textbooks. We list some of the more recent texts that treat Greens functions in references [1-4], and review the usual textbook treatment below. The motivation for using a Greens function, G(r, r ), is that it satises homogeneous boundary conditions that makes it easier to solve for than the
Internet address: Jerry.F@TEMPLE.EDU For simplicity, we will take q (r) = 0. Our general conclusions also apply to the case with q (r), but some of the equations would be more complicated.
1

original problem with inhomogeneous boundary conditions. Application of Greens theorem [(r )Lr G(r, r)G(r, r )Lr (r )]d = dS [(r )p(r) G(r, r)G(r, r )p(r ) (r )] (2) provides the solution to the original problem if G(r, r ) satises

L r G( r , r ) = ( r r ) . In that case, Eq (2) leads to (r ) = G(r, r)(r )]d +

(3)

dS [(r)p(r ) G(r, r) G(r, r )p(r ) (r )].

(4) The two simplest boundary conditions for which the Greens function method is applicable are the Dirichlet boundary condition for which the solution (rS ) is given on all bounding surfaces, and the Neumann boundary condition for which its normal derivative n (rS ) is given. For the Dirichlet boundary condition, the Greens function satises the homogeneous boundary condition GD ( r , r ) = 0 (5) for r on all bounding surfaces. This reduces Eq. (4) to (r ) = GD (r, r )p(r)(r )]d + dS [(r )p(r) GD (r, r)], (6)

which is the solution to the Dirichlet problem once GD (r, r) is known. The Dirichlet Greens function is the solution to Eq. (3) satisfying the homogeneous boundary condition in Eq. (5). Textbooks generally treat the Dirichlet case as above, but do much less with the Greens function for the Neumann boundary condition, and what is said about the Neumann case often has mistakes of omission and commission. First of all, the Neumann boundary condition for the solution (r) must satisfy the constraint dS p(r)(r) = d, (7)

which follows from applying the divergence theorem to Eq. (1). Most texts do not mention this important constraint on the Neumann boundary condition.2
2

This constraint is noted on p. 66 of Ref. [3].

There are cases where the boundary condition is Neumann on some surfaces and Dirichlet on others. In those cases, the normal derivative of on the Dirichlet surfaces will adjust to satisfy the constraint surface integral. But for pure Neumann boundary conditions, the constraint must be satised or no solution exists. In the following sections, we will assume pure Neumann boundary conditions for which constraint equation (7) holds. We treat the one dimensional Neumann Greens function in Section 2, and then the three dimensional case in Section 3.

1D Neumann Greens function

We rst review how the Dirac delta function arises when a function f (x), dened in the nite range 0 x L, is expanded in orthonormal Dirichlet eigenfunctions un (x) of a Sturm-Liouville operator Lx . The eigenfunctions satisfy the dierential equation Lx un (x) = d dun p(x) = n un (x). dx dx (8)

with homogeneous Dirichlet boundary conditions un (0) = 0, un (L) = 0. (9)

The function f (x) can be expanded in the Dirichlet eigenfunctions as

f (x) =
n=1

bn un (x),

(10)

with the expansion coecients bn given by

L

bn =

u n (x)f (x)dx.

(11)

If Eq. (11) for the expansion coecients is substituted into Eq. (10) for f (x), and the sum executed before the integral, we get
L

f (x) =

dx
n=1

u n (x )un (x)f (x ).

(12)

From the denition of the Dirac delta function by its sifting property,
L

f (x) =

dx (x x )f (x ) 3

(13)

for x and x in the range [0, L], we see that the delta function can be represented by a sum over Dirichlet eigenfunctions as

(x x ) =
n=1

u n (x )un (x).

(14)

For example, for the simple case L x u n = u n = n u n ,

un =

2/L sin(nx/L),

n =

n L

(15)

the delta function is represented by

(x x ) = (2/L)
n=1

sin(nx /L) sin(nx/L)

(16)

A Dirichlet Greens function that satises the dierential equation Lx GD (x, x ) = (x x ), (17)

and satises the homogeneous Dirichlet boundary conditions in the variable x can be formed from the Dirichlet eigenfunctions as GD (x, x ) =
u n (x )un (x) . n n=1

(18)

Acting on this Greens function with the Sturm Liouville operator Lx removes the denominator in Eq. (18) leaving the delta function of Eq. (17), showing that this is the appropriate Dirichlet Greens function. The above straightforward derivation for homogeneous Dirichlet boundary conditions is given in most texts, but a corresponding derivation for homogeneous Neumann boundary conditions is generally absent. Neumann eigenfunctions, vn (x), satisfy the same dierential equation (8) as the Dirichlet eigenfunctions, but have the boundary conditions
vn (0) = 0, vn (L) = 0.

(19)

The expansion in Neumann eigenfunctions has a constant eigenfunction corresponding to a zero eigenvalue, so the expansion is given by

f (x) = a0 +
n=1

an vn (x),

(20)

The Neumann expansion coecients an are given by the integrals

L

an = and

vn (x)f (x)dx,

n 1,

(21)

1 L f (x)dx =< f >, (22) L 0 where < f > represents the average value of the function f (x) over the interval [0, L]. Now putting Eqs. (21) and (22) into the expansion Eq. (20) results in a0 =
L

f (x) =

1 u + dx n (x )un (x) f (x ), L n=1

(23)

so the representation of the delta function in terms of Neumann eigenfunctions is 1 (x x ) = + v (x )vn (x). (24) L n=1 n The additional constant term 1/L is not generally recognized in textbooks. For the simple case of Lx vn = vn = n with the eigenfunctions satisfying the homogeneous Neumann boundary conditions of Eq. (19) the delta function is represented by 1 2 (x x ) = + cos(nx /L) cos(nx/L). L L n=1

(25)

A Neumanns Green function can be formed using Neumann eigenfunctions of the operator Lx as the sum GN (x, x ) =
vn (x )vn (x) . n n=1

(26)

This Greens function satises the homogeneous Neumann boundary conditions (27) x GN (x, x )|(x =0) = 0, x GN (x, x )|(x =L) = 0. Because of the constant term 1/L in Eq. (25), The operation on GN (x, x ) by Lx is (28) Lx GN (x, x ) = (x x ) 1/L, 5

so the Neumann Greens function satises a dierent dierential equation than the Dirichlet Greens function. We now use the Greens function GN (x, x ) to nd the solution of the dierential equation Lx f (x) = d df p(x) = (x), dx dx (29)

with the inhomogeneous Neumann boundary conditions

f (0) = f0 , f (L) = fL .

(30)

The boundary values must satisfy the constraint

p(L)fL p(0)f0 =

(x)dx,

(31)

which follows from a rst integral of Eq. (29). Greens theorem in one dimension (or integration by parts) for this differential equation leads to
L 0 +GN (x, 0)p(0)f0 . [f (x )Lx GN (x, x )GN (x, x )Lx f (x )]dx = GN (x, L)p(L)fL

(32) We have used the boundary conditions (27) to eliminate terms containing x G(x, x ) at the endpoints. Then, using Eq. (28), we get
L

f (x) =< f > +

(33) which constitutes the solution to the Sturm-Liouville problem for Neumann boundary conditions. The constant < f >, the average value of f (x), arises from the term 1/L in Eq. (28). Actually, any constant can be added to the solution f (x) since the solution of the Neumann problem is only unique up to an additive constant. The constant term in Eq. (33) will always be the average value of the solution because the variable terms have zero average value. Although we have used an expansion in eigenfunctions to give an heuristic derivation of the Greens function for Neumann boundary conditions, any function satisfying the dening equation (28) with the boundary conditions 6

GN (x, x )(x )dx GN (x, L)p(L)fL + GN (x, 0)p(0)f0 ,

of Eq. (27) will be a suitable Greens function GN (x.x ). For example, a Greens function for the problem f (x) = x, is given by x2 +x 2L x2 x x L : GN 2 (x, x ) = + x . 2L 0 x x : GN 1 (x, x ) =
f (0) = f0 , f (L) = fL = f0 + L2 /2,

(34)

(35)

2 The term x2 /2L in GN 1 and GN 2 provides the 1/L term in x GN (x, x )., and also satises the homogeneous Neumann boundary condition at x = 0. The x term in GN 2 satises the homogeneous Neumann boundary condition at x = L, and also provides a unit step in GN at x = x so that the next 2 derivative will give the delta function in x GN (x, x ). Finally, the term x in GN 1 makes GN (x, x ) continuous at x = x. Putting this Greens function into Eq. (33) gives the solution to Eq. (34). We note that the Greens function in Eq. (35) is not symmetric with respect to x and x . To show that a Neumann Greens function need not be symmetric, we repeat the usual proof of symmetry here. Applying Greens theorem to two Greens functions, G(x1 , x ) and G(x2 , x ), of a Sturm-Liouville operator L gives L 0

[G(x1 , x )Lx G(x2 , x ) G(x2 , x )Lx G(x1 , x )]

x =L

= [G(x1 , x )p(x )x G(x2 , x ) G(x2 , x )p(x )x G(x1 , x )]x =0. (36) The right hand side of Eq. (36) vanishes for either Dirichlet or Neumann homogeneous boundary conditions. For a Dirichlet Greens function, with Lx GD (x, x ) = (x x ), Eq. (36) reduces to GD (x1 , x2 ) GD (x2 , x1 ) = 0, (37)

so a Dirichlet Greens function must be symmetric. However, for a Neumann Greens function, Lx GD (x, x ) = 1/L + (x x ), and Eq. (36) reduces to GN (x1 , x2 ) GN (x2 , x1 ) = 1 L
L 0

[GN (x1 , x ) GN (x2 , x )]dx , 7

(38)

so a Neumann Greens function is not required to be symmetric. We do know however, from the eigenfunction expansion in Eq. (26) that any real Neumann Greens function can be made symmetric. For instance, adding the term x2 /2L to the Greens function in Eq. (35) will make that Neumann Greens function symmetric without changing any of its actions. There is no need to do this however, since the non-symmetric Greens function is simpler, and either form will solve the original dierential equation.

3D Neumanns Green function

In this section, we extend the one dimensional results of the previous section to three dimensions. In three dimensions, we seek the solution of the dierential equation Lr (r) = [p(r)(r)] = (r) (39)

with the inhomogeneous Neumann boundary conditions on any bounding surface n ( r ) = f ( r S ) , (40) where n is the outward normal vector to the surface and f (rS ) is an almost arbitrary function specied on all surfaces. A solution to Eq. (39) exists only if the boundary conditions satisfy the constraint dS p(r)(r) = (r)d. (41)

This constraint follows by applying the divergence theorem to Eq. (39). The Neumann Greens function for this problem satises the dierential equation (42) Lr GN (r, r ) = (r r ) 1/V, with the homogeneous boundary condition n G N ( r , r ) = 0 (43)

on all surfaces. This Greens function automatically satises the constraint dS p(r ) GN (r, r) = 0, which follows by applying the divergence theorem to Eq. (42). 8 (44)

The extra term 1/V in Eq. (42) is the 3D equivalent of the term 1/L in Eq. (28), and arises due to the constant term in any expansion using Neumann eigenfunctions. The 1/V goes to zero for an innite volume (the so called exterior problem), but is important if the volume considered is nite (the interior problem). The solution to Eq. (39) is given by (r ) =< > + G(r, r)(r )]d dS G(r, r)p(r ) (r), (45)

which follows from Greens theorem, and Eqs. (42) and (43) for GN (r, r ). As in the 1D case, < > is an arbitrary constant that equals the average value of in the volume. It can be chosen to be zero to simplify the equation.

Summary

To summarize, we have found seven essential dierences between the Neumann Greens function treated here and the Dirichlet Greens function which is generally treated in Math Physics or Electromagnetism texts for the solution of a partial dierential equation of the form Lr (r) = [p(r)(r)](r) = (r). We list again the dierent Neumann properties below: (1) The Neumann boundary condition for the solution (r) is n ( r ) = f ( r S ) , (2) with the constraint dS p(r)(r) = required for any solution to exist. (3) The Neumann Greens function satises the dierential equation Lr GN (r, r ) = (r r ) 1/V, (49) (r)d, (48) (47) (46)

(4) with the homogeneous boundary condition n G N ( r , r ) = 0 on all surfaces. (5) The solution to Eq. (46) is given by (r ) =< > + G(r, r)(r )]d dS G(r, r )p(x ) (r ), (51) (50)

(6) where < > is an arbitrary constant that equals the average value of (r) in the volume. (7) The Neumann Greens function is not necessarily symmetric, but can always be made symmetric by adding a function of r to GN (r, r ). Adding any function of r to a Neumann Greens function does not change its actions. Of these seven dierences, only number (1) is generally mentioned in Math Physics or EM textbooks. Some books[1, 2, 3] give the homogeneous boundary condition in Eq. (50), but dont mention that it is inconsistent with omitting the 1/V term in Eq. (49). Other books[4] recognize this by making the Greens function boundary condition inhomogeneous, but this is not a satisfactory remedy. A Greens function with inhomogeneous boundary conditions can be just as hard to nd as the original solution. For instance, constructing a Greens function by an expansion in eigenfunctions is dicult without homogeneous boundary conditions.

References
[1] Arfken G B and Weber H J 2005 Mathematical Methods for Physicists 6th edn (Boston: Elsevier) [2] Butkov E 1968 Mathematical Physics (Reading, Mass.: Addison-Wesley) [3] Franklin J 2005 Classical Electromagnetism (San Francisco: AddisonWesley) [4] Jackson J D 1999 Classical Electrodynamics 3rd edn (John Wiley & Sons, New York) 10