]>

Appendix 7A: Two- and Three-Dimensional Green’s Functions

Appendix 7A: Two- and Three-Dimensional Green’s Functions

7A.1Introduction

We discussed Green’s function briefly (Section 7.2). The one-dimensional Green’s function of the Laplace equation, with Dirichlet boundary conditions, is the solution of the differential equation

(7A.1a)d2Gdx2=δ(xx),0<x<L,
d2Gdx2=δ(xx),0<x<L,

subject to the boundary conditions

(7A.1b)G=0,x=0,
G=0,x=0,
(7A.1c)G=0,x=L,
G=0,x=L,

and is shown to be

(7A.2a)G=G1=xLxL,0<x<x,
G=G1=xLxL,0<x<x,
(7A.2b)G=G2=x(Lx)L,x<x<L.
G=G2=x(Lx)L,x<x<L.

For an arbitrary input f(x), the response y(x) can be obtained from a superposition integral

(7A.3)y(x)=L0f(x)G(x,x)dx.
y(x)=0Lf(x)G(x,x)dx.

The differential equation for y is

(7A.4a)d2ydx2=f(x),0<x<L,
d2ydx2=f(x),0<x<L,

and the boundary conditions are

(7A.4b)y=0,x=0,
y=0,x=0,
(7A.4c)y=0,x=L.
y=0,x=L.

Equation 7A.3 for y is the solution to Equation 7A.4, where G is given by Equation 7A.2. Equation 7A.3 may be written more explicitly, by considering the solution in the two domains 0 ≤ xx′ and x′ ≤ xL:

(7A.5a)y(x)=L0x(Lx)Lf(x)dx,0xx,
y(x)=0Lx(Lx)Lf(x)dx,0xx,
(7A.5b)y(x)=L0x(Lx)Lf(x)dx,xxL.
y(x)=0Lx(Lx)Lf(x)dx,xxL.

Note: Some books define the Green’s function using a positive impulse for the input strength:

(7A.6)d2Gdx2=δ(xx).
d2Gdx2=δ(xx).

Then the solution will be of opposite sign to Equations 7A.2a and 7A.2b:

(7A.7a)G=G1=xxLL,0<x<x,
G=G1=xxLL,0<x<x,
(7A.7b)G=G2=xxLL,x<x<L.
G=G2=xxLL,x<x<L.

We also considered the one-dimensional Helmholtz equation, with Dirichlet boundary conditions,

(7A.8a)d2Gdx2+k2G=δ(xx),
d2Gdx2+k2G=δ(xx),
(7A.8b)G=0,x=0,
G=0,x=0,
(7A.8c)G=0,x=L,
G=0,x=L,
(7A.9a)G=G1=1ksinkxsink(Lx)sinkL,0<x<x,
G=G1=1ksinkxsink(Lx)sinkL,0<x<x,
(7A.9b)G=G2=1ksinkxsink(Lx)sinkL,x<x<L.
G=G2=1ksinkxsink(Lx)sinkL,x<x<L.

7A.2Alternate Form: Infinite Series

The procedure used in Section 7.2 is to solve the homogeneous equation in the two domains excluding the point x = x′ where the impulse is applied. Using the boundary conditions, and the source conditions, the undetermined constants were determined. The result is Green’s function in closed form. An alternative form will now be determined using eigenfunction expansion.

Let us solve Equation 7A.6 using this technique. The solution will be written as an infinite series

(7A.10)G(x,x)=n=1an(x)sinnπxL.
G(x,x)=n=1an(x)sinnπxL.

Note that each of the terms on the RHS of Equation 7A.10 satisfies the boundary conditions of G = 0 at x = 0 or L. From Equation 7A.10, we obtain

(7A.11)d2Gdx2=(nπL)2an(x)sinnπxL.
d2Gdx2=(nπL)2an(x)sinnπxL.

By substituting Equation 7A.11 into Equation 7A.6, multiplying by sin mπx/L, and then integrating from 0 to L, we obtain

(7A.12)(nπL)2an(x)L0sinnπxLsinmπxLdx=L0δ(xx)sinmπxLdx.
(nπL)2an(x)0LsinnπxLsinmπxLdx=0Lδ(xx)sinmπxLdx.

From the orthogonality property,

(7A.13)L0sinnπxLsinmπxLdx=0,mn=L2,m=n.
0LsinnπxLsinmπxLdx==0,mnL2,m=n.

Thus, from Equation 7A.12, we get

(mπL)2am(x)L2=sinmπxL,am(x)=2Lm2π2sinmπxL.
(mπL)2am(x)L2am(x)==sinmπxL,2Lm2π2sinmπxL.

Thus, we have

(7A.14)G(x,x)=2Lπ2n=11n2sinnπxLsinnπxL.
G(x,x)=2Lπ2n=11n2sinnπxLsinnπxL.

Equation 7A.14 is an alternative form to Equations 7A.7a and 7A.7b, although Equation 7A.14 appears as an infinite series. The solution of

(7A.15)d2ydx2=f(x)
d2ydx2=f(x)

can be written as

(7A.16)y(x)=2Lπ2n=11n2sinnπxLL0sinnπxLf(x)dx.
y(x)=2Lπ2n=11n2sinnπxL0LsinnπxLf(x)dx.

The solution converges due to the term 1/n2 in Equation 7A.16.

7A.3Sturm–Liouville Operator

A generalization of the series method for a one-dimensional problem with a general second-order differential equation is formulated in terms of a Sturm–Liouville operator L:

(7A.17a)[L+λr(x)]y=f(x),

where

(7A.17b)L={ddx[p(x)ddx]q(x)}.

Let ψn be a complete set of orthonormal eigenfunctions for the L operator, that is,

(7A.18)[L+λnr(x)]Ψn(x)=0

subject to the same boundary condition as the original problem of Equation 7A.17a. If G(x, x′) is Green’s function

(7A.19)[L+λr(x)]G(x,x)=δ(xx)

subject to the same boundary condition as the original problem, then

(7A.20)G(x,x)=nΨn(x)Ψn(x)λλn.

We will show two examples [1] of series form of Green’s function for two particular problems obtained from Equation 7A.20.

Example 1

(7A.6)d2Gdx2=δ(xx).

Equation 7A.6 is a particular case of Equation 7A.19 where

λ=0,p(x)=1,q(x)=0,r(x)=1,L=d2dx2.

The eigenfunctions are obtained from

(L+λnr(x))ψ(x)=[d2dx2+λn]ψn=0

subject to the boundary conditions

ψn=0,x=0orL.

Thus,

ψn=2LsinnπxL(orthonormal),n=1,2,,,λn=(nπL)2,G(x,x)=n2LsinnπxL2LsinnπxL0(nπL)2,

which is the same as Equation 7A.14. The second example is Green’s function of Helmholtz equation (in series form)

Example 2

(7A.21a)d2Gdx2+β2G=δ(xx),
(7A.21b)G(0)=G(L)=0.

Equation 7A.21a is a particular case of Equation 7A.19, where

λ=β2,p(x)=1,q(x)=0,r(x)=1,L=d2dx2.

The orthonormal equations are

ψn=2Lsinβnx=2LsinnπxL,λn=β2n=(nπL)2.

Hence,

(7A.22)G(x,x)=2Ln=1,2,sin(nπx/L)sinnπx/Lβ2(nπ/L)2.

To complete this section, here we will write down Green’s function in closed form for the Sturm–Liouville problem:

(7A.23a)G1=G(x,x)=h2(x)h1(x)p(x)w(x),0<x<x,
(7A.23b)G2=G(x,x)=h1(x)h2(x)p(x)w(x),x<x<L,

where h1(x) is the solution form of the homogeneous equation, in the interval 0 < x < x′, and h2(x) is the solution form of the homogeneous equation, in the interval x′ < x < a:

(7A.23c)G1(x)=A1h1(x),
(7A.23d)G2(x)=A2h2(x),

and w(x′) is the Wronskian of h1 and h2 at x = x′:

(7A.23e)w(x)=h1(x)h2(x)h2(x)h1(x).

The closed solution of Green’s function given in Equation 7A.9a and 7A.9b can be obtained from the general solution given in (7A.23) by noting

p(x)=1,h1(x)=sinkx,h2(x)=sink(Lx)

and the Wronskian w(x′) is obtained as

w(x)=h1(x)h2(x)h2(x)h1(x)=(sinkx)(k)cosk(Lx)sink(Lx)(k)coskx=[sinkxcosk(Lx)+coskxsink(Lx)]=[sink(x+Lx)]=ksinkL,
(7A.24a)G=G1=sink(Lx)sinkxksinkL,0<x<x,
(7A.24b)G=G2=sinkxsink(Lx)ksinkL,x<x<L.

The difference in sign between Equations 7A.24 and 7A.9 is because Equation 7A.9 is the response when the impulse is of strength −1.

7A.4Two-Dimensional Green’s Function in Rectangular Coordinates

7A.4.1Laplace Equation: Formulation and Closed Form Solution

(7A.25a)d2Gdx2+d2Gdy2=δ(xx)δ(yy),
(7A.25b)x=0ora,G=0,y=0orb,G=0.

Let us first write

(7A.26)G(x,y,x,y)=n=1,2,gm(y,x,y)sinmπxa.

Here, we formulated such that the boundary conditions at the walls x = 0 or a are satisfied.

By substituting Equation 7A.26 into Equation 7A.25a, we obtain

(7A.27)m=1[(mπa)2gm(y;x,y)sinmπxa+sinmπxad2gm(y;x,y)dy2]=δ(xx)δ(yy).

Multiply both sides of Equation 7A.27 by sin nπx/a and integrate with respect to x from 0 to a.

From orthogonality property, the LHS is

a2[(nπa)2gn(y;x,y)+d2dy2gn(y;x,y)]

and the RHS is

a0δ(xx)δ(yy)sinnπxadx=sinnπxaδ(yy).

Thus, the differential equation in y is obtained as

(7A.28)d2dy2gm(y;x,y)(mπa)2gm(y;x,y)=a2sinmπxaδ(yy).

Equation 7A.28 can now be solved by using the recipe of Equations 7A.23a and 7A.23b.

The homogeneous form of Equation 7A.28 is

(7A.29)d2gmdy2(y;x,y)(mπa)2gm(y;x,y)=0

subject to the boundary condition gm = 0, y = 0 or b (Figure 7A.1):

(7A.30a)A1h1(x)Amh(1)m(y)=Am(x,y)sinhmπya,region1,0<y<y,
(7A.30b)A2h2(x)Bmh(2)m(y)=Bm(x,y)sinhmπ(by)a,region2,y<y<b.
image

FIGURE 7A.1
Two-dimensional rectangular area divided into two regions: region 1, 0 < y < y′ − ε; region 2, y′ + ε < y < b.

Now the Wronskian from Equation 7A.23e, after simplification, becomes

(7A.30c)w(y;x,y)=mπasinhmπba.

From Equation 7A.23a, we obtain

(7A.31a)g(1)m(y;x,y)=(2/a)sin(mπx/a)(h(2)m|y)(h(1)m|y)(mπ/a)sinh(mπb/a)=(2/mπ)sin(mπx/a)sinh(mπ/a)(by)sinh(mπy/b)sinh(mπb/a),0<y<y

From Equation 7A.23b, we obtain

(7A.31b)g(2)m(y;x,y)=(2/mπ)sin(mπy/a)sinh[(mπ/a)(by)]sinh(mπb/a),y<y<b.

Now the complete Green’s function is given as

(7A.32a)G(x,y;x,y)=2πm=1sin(mπx/a)sinh[(mπ/a)(by)]msinh(mπb/a)×sinmπxasinhmπya,for0xa,0yy.
(7A.32b)=2πm=1sin(mπx/a)sinh[(mπ/a)(by)]msinh(mπb/a)×sinmπxasinhmπyafor0xa,yyb.

We can develop an alternative form of Green’s function by first satisfying the boundary conditions at the walls y = 0 or b, that is, instead of Equation 7A.2 start with

(7A.33)G(x,y;x,y)=n=1,2,gn(x;x,y)sinnπyb

and develop the solution. The result is given in Equations 14.83a and b of Balanis [1].

7A.4.2Laplace Equation: Series Form (Bilateral) Solution

For series solution of Green’s function, we need the orthonormal eigenfunction ψmn for this two-dimensional problem.

These are the solutions of (for Laplace equation)

(7A.34a)2ψmnx2+2ψmny2+λmnψmn=0

subject to boundary condition

(7A.34b)ψmn=0,x=0ora,=0,y=0orb.

Of course, we know that the orthogonal eigenfunctions that satisfy the boundary conditions (Equation 7A.34b) are

ψmn=Amnsinmπxasinnπyb.

From orthogonality property,

1=ax=0by=0(ψmn)2dxdy=ab4A2mn.

Thus, Amn=2/ab,

(7A.35a)ψmn=2absinmπxasinnπyb,
(7A.35b)λmn=[(mπa)2+(nπb)2]1/2.

Using the bilinear form (7A.20), we get

(7A.36)G(x,y;x,y)=mnψmn(x,y)ψmn(x,y)λλmn=4abm=1n=1sin(mπx/a)sin(nπy/b)[(mπ/a)2+(nπ/b)2]sinmπxasinnπyb.

Note that λ = 0, for the Laplace equation.

One can use Equation 7A.36 to solve Poisson’s equation:

(7A.37a)2Vx2+2Vy2=q(x,y).

By using Equation 7A.36 and the superposition integral, we obtain

(7A.37b)V(x,y)=4abm=1n=1sin(mπx/a)sin(nπy/b)(mπ/a)2+(nπ/b)2×by=0ax=0q(x,y)sin(mπx/a)sin(nπy/b)dxdy.

7A.4.3Helmholtz Equation (Series Form)

Green’s function with homogeneous, Dirichlet boundary condition is given by

(7A.38a)2Gx2+2Gy2+β2G=δ(xx)δ(yy)

with boundary conditions

(7A.38b)G=0,x=0ora,y=0orb.

For this case, λ = β2:

(7A.39)ψmn=2absinmπxasinnπyb

and Equation 7A.36, gets modified as

(7A.40)G(x,y;x,y)=4abm=1n=1sin(mπx/a)sin(nπy/b)β2[(mπ/a)2+(nπ/b)2]2×sinmπxasinnπyb.

7A.5Generalized Green’s Function Method

Till now, we derived Green’s function that satisfied homogenous Dirichlet boundary conditions; in all the examples, the potential also satisfied the homogenous Dirichlet boundary condition.

Let us now investigate whether we can use a more general Green’s function that has an impulse source but not necessarily satisfying the Dirichlet homogenous boundary condition. The scalar Helmholtz equation is

(7A.41)2Φ(r)+β2Φ(r)=f(r)

and Green’s function for the problem is G(r, r′) satisfying the equation

(7A.42)2G(r,r)+β2G(r,r)=δ(rr).

Green’s first and second identities are given below.

Green’s first identity

(7A.43)sΦψnds=VΦ2ψdv+V(ˉΦˉψ)dv.

In the above, s is a closed surface bounding a volume V, Φ and ψ are two scalar functions, and ˆn is the unit vector normal to the surface.

Green’s second identity

(7A.44)s[ΦψnψΦn]ds=V[Φ2ψψ2Φ]dv.

Multiply Equation 7A.41 with G(ˉr,ˉr) and Equation 7A.42 with Φ(ˉr), we obtain

(7A.45)G2Φ+β2ΦG=fG,
(7A.46)Φ2G+β2ΦG=Φδ(ˉr,ˉr).

By subtracting Equation 7A.45 from Equation 7A.46 and integrating over the volume V, we obtain

(7A.47)VΦ(ˉr)δ(ˉr,ˉr)dvVfGdv=V[Φ2GG2Φ]dv.

Applying Equation 7A.44 to the RHS of Equation 7A.47 and also evaluating the first term on the LHS of Equation 7A.47, we get

(7A.48a)Φ(ˉr)=Vf(ˉr)G(ˉr,ˉr)dv+s[ΦGnGΦn]ds.

Since ˉr is an arbitrary point in V and ˉr is a dummy variable, G(ˉr,ˉr)=G(ˉr,ˉr).

We can write (7A.48) as (exchanging ˉr and ˉr)

(7A.48b)Φ(ˉr)=Vf(ˉr)G(ˉr,ˉr)dv+s[Φ(ˉr)GnGΦ(ˉr)n]ds.

In the above, the differentiations are with respect to primed (′) coordinates.

If we have homogeneous Dirichlet boundary conditions, satisfied by Φ as well as G on s, the second integral (surface integral) becomes zero, and hence

(7A.49)Φ(ˉr)=Vf(ˉr)G(ˉr,ˉr)dv.

This is exactly the superposition integral we used in all the previous discussions. In all those discussions, we had the homogeneous Dirichlet boundary condition satisfaction both by Φ and G. Equation 7A.48 is the modified superposition integral for the general case. To use Equation 7A.48, we need to know Φ, ∂Φ/∂n, G, and ∂G/∂n on the closed boundary s.

7A.6Three-Dimensional Green’s Function and Green’s Dyadic

Three-dimensional Green’s function in free space for the scalar Helmholtz equation

(7A.50)2G+k2G=δ(ˉrˉr)

is given by

(7A.51)G(ˉr,ˉr)=14πejk|ˉrˉr||ˉrˉr|.

By making k = 0 in Equations 7A.50 and 7A.51, we obtain the Laplace equation and the associated Green’s function.

Green’s function ˉΓ(ˉr,ˉr), called Green’s dyadic, satisfies the equation

(7A.52)[ˉ×ˉ×k2]ˉΓ(ˉr,ˉr)=ˉuδ(ˉrˉr),

where ˉu is a unit dyadic given as

(7A.53)ˉu=3m=13n=1ˆemˆenδmn.

In Equation 7A.53, ˆem, in Cartesian coordinates are ˆx, ˆy, ˆz for m = 1, 2, 3, respectively. δmn is the Kronecker delta and given by

(7A.54)δmn={1,m=n,0,mn.

A good account of dyads and their properties are given in [2,3]. In this connection, one can also define the double gradient ˉˉ:

(7A.55)ˉˉ=3m=13n=1emenxmxn.

From the properties of dyadic operations [2,3], we can show that dyadic Green’s function

(7A.56)ˉΓ(ˉr,ˉr)=(ˉu+1k2ˉˉ)G(ˉr,ˉr).

Equation 7A.56 relates the scalar Green’s function G given by Equation 7A.51 with dyadic Green’s function given by Equation 7A.56.

Thus, the equation for the electric field

(7A.57)ˉ×ˉ×ˉEk2ˉE=jωμˉJ

can be solved as a superposition integral in terms of dyadic Green’s function ˉΓ:

(7A.58)ˉE(ˉr)=jωμˉΓ(ˉr,ˉr)ˉJ(ˉr)dv.

References

  1. 1.Balanis, C. A., Advanced Engineering Electromagnetics, Wiley, New York, 1989.
  2. 2.Papas, C. H., Theory of Electromagnetic Wave Propagation, McGraw-Hill, New York, 1964.
  3. 3.Van Bladel, J., Electromagnetic Fields, Hemisphere Publishing, New York, 1985.
..................Content has been hidden....................

You can't read the all page of ebook, please click here login for view all page.
Reset
18.117.153.38