1
ECE 6382
Green’s Functions
Notes are from D. R. Wilton, Dept. of ECE
David R. Jackson
Fall 2019
Notes 19
Green’s Functions
2
Green's Mill in Sneinton (Nottingham), England, the mill owned by Green's father. The mill was renovated in 1986 and is now a science centre.
George Green (1793-1841)
The Green's function method is a powerful and systematic method for determining a solution to a problem with a known forcing function on the RHS.
The Green’s function is the solution to a “point” or “impulse” forcing function.
It is similar to the idea of an “impulse response” in circuit theory.
Green’s Functions
3
Consider the following second-order linear differential equation:
( ) ( ) ( ) ( ) ( )1 ( )d dP x u x Q x u x f xw x dx dx
− + =
u f=or
( ) ( ) ( )1 d dP x Q xw x dx dx
≡− +
where
( f is a “forcing” function.)
Green’s Functions
4
Problem to be solved:
u f=
( ) ( )0, 0 ( )u a u b= = Dirichlet BCs
( ) ( )0, 0 ( )u a u b′ ′= = Neuman BCs
or
Green’s Functions (cont.)
5
We can think of the forcing function f(x) as being broken into many small rectangular pieces.
Using superposition, we add up the solution from each small piece.
Each small piece can be represented as a delta function in the limit as the width approaches zero.
Note:The ∆(x) function becomes a δ (x) function in
the limit as ∆x → 0.
( ) ( )1 , / 2, / 2
0,
x x xx x
∈ ∆ ∆∆ ≡ ∆ otherwise
( ) ( ) ( )P x x x= ∆ ∆1.0 x∆
ix
( )( ) ( )
( )
( )i i
i i
f x P x x
f x x x x
−
= ∆ ∆ −
x
( )f x
a
x∆ b
( )f x
Green’s Functions (cont.)
6
( ) ( ) ( ) ( ) ( )1 1
, ,N N
i i i ii i
u x f x P x x f x x G x x∆ ∆= =
≈ = ∆∑ ∑
( ) ( ), i i iG x x x x x∆ ∆≡ −solution from single pulse centered at
From superposition:
( ) ( ) ( ) ( )1
, ,bN
i ii a
f x x G x x f x G x x dx∆=
′ ′ ′∆ →∑ ∫
( ) ( ), i i iPP x x x x x∆ ≡ −solution from single pulse centered at
0x∆ →Let
( ) ( ), G x x x x xδ′ ′ ′≡ −solution from centered at
Green’s Functions (cont.)
7
The solution to the original differential equation (from superposition) is then
( ) ( ) ( ),b
a
u x f x G x x dx′ ′ ′= ∫
( ) ( ),G x x x xδ′ ′= −
The Green’s function G(x, x′) is defined as the solution with adelta-function at x = x′ for the RHS.
xx′
( )G x,x′
a a
Green’s Functions (cont.)
8
There are two general methods for constructing Green’s functions.
Method 1:Find the solution to the homogenous equation to the left and right of the delta function, and then enforce boundary conditions at the location of the delta function.
( ) ( )( )
1
2
,,
,Au x x x
G x xBu x x x
′ ≤′ = ′≥
The Green’s function is assumed to be continuous. The derivative of the Green’s function is allowed to be discontinuous.
The functions u1 and u2 are solutions of the homogenous equation.
xx′
( )2Bu x( )1Au x
a b
Green’s Functions (cont.)
9
Method 2:Use the method of eigenfunction expansion.
n n nψ λψ=
( ) ( ), n nn
G x x a xψ′ =∑
Eigenvalue problem:
We then have:
( ) ( )0, 0 ( )a bψ ψ= = Dirichlet BCs
( ) ( )0, 0 ( )a bψ ψ′ ′= = Neuman BCsor
Note:The eigenfunctions
are orthogonal.
Green’s Functions (cont.)
10
Method 1
( ) ( ) ( )
( )( ) ( )
0 0
0
1lim lim
( , ) ( , )1 lim ( ) 1
x x
x x
x
x
d dGG dx P x Q x G dxw x dx dx
P x dG x x P x dG x xx x dx
w x dx dx
ε ε
ε εε ε
ε
εε
δ
′ ′+ +
→ →′ ′− −
+ + − − ′+
→′−
= − + ′ ′ ′ ′ ′ ′
′ = − − = − = ′
∫ ∫
∫
( ) ( ),G x x x xδ′ ′= −
( ) ( ) ( )( )2 1
w xBu x Au x
P x′
′ ′ ′ ′− = −′
( ) ( )( )
1
2
,,
,Au x x x
G x xBu x x x
′ ≤′ = ′≥
Integrate both sides over the delta function:
( )( )
( , ) ( , ) w xdG x x dG x xdx dx P x
+ − ′′ ′ ′ ′ − = − ′
xx′
( )2Bu x( )1Au x
a b
Green’s Functions (cont.)
11
Method 1 (cont.)
Also, we have (from continuity of the Green’s function):
( ) ( )1 2Au x Bu x′ ′=
( ) ( )( )
1
2
,,
,Au x x x
G x xBu x x x
′ ≤′ = ′≥
xx′
( )2Bu x( )1Au x
a b
Green’s Functions (cont.)
12
Method 1 (cont.)
We then have:
( ) ( )( ) ( )
( )( )1 2
1 2 0
w xu x u x A
P xu x u x B
′ ′ ′ ′ ′ −− ′= ′ ′−
( )( )
( ) ( )( )
( )2 1
1 2 1 2
,[ , ] [ , ]
w x u x w x u xA B
P x W u u P x W u u ′ ′ ′ ′
= − = − ′ ′
( )1 2 1 2 1 2[ , ] ( ) ( ) ( ) ( )W u u W x u x u x u x u x′ ′ ′ ′ ′ ′ ′= = − (Wronskian)
where
xx′
( )2Bu x( )1Au x
a b
( )W x′∆ = = −determinantNote :
Green’s Functions (cont.)
13
Method 1 (cont.)
We then have:
( )
( )( )
( ) ( )( )
( )( )
( ) ( )( )
2 1
1 2
,
,
,
w x u x u xx x
P x W xG x x
w x u x u xx x
P x W x
′ ′′− < ′ ′ ′ =
′ ′ ′− > ′ ′
xx′
( )2Bu x( )1Au x
a b
Green’s Functions (cont.)
14
Method 2
xa bx′
n=1n=2
n=3
n=4
The Green’s function is expanded as a series of eigenfunctions.
n n nψ λψ=
( ) ( ), n nn
G x x a xψ′ = ∑
where
The eigenfunctions corresponding to distinct eigenvalues are orthogonal (from Sturm-Liouville theory).
Green’s Functions (cont.)
15
Method 2 (cont.)
xa bx′
n=1n=2
n=3
n=4
( ) ( ),G x x x xδ′ ′= −
( ) ( )n nn
a x x xψ δ ′= −∑
( ) ( )n nn
a x x xψ δ ′= −∑
( ) ( )n n nn
a x x xλ ψ δ ′= −∑
( ) ( ) ( ) ( ), ,n n n m mn
a x x x x xλ ψ ψ δ ψ′= −∑
( ) ( ) ( ) ( )*,n n n m mn
a x x w x xλ ψ ψ ψ′ ′=∑
( ) ( ) ( ) ( )*,m m m m ma x x w x xλ ψ ψ ψ′ ′=
Orthogonality
Delta-function property
Multiply both sides by ψm*(x) w(x) and then integrate from a to b.
Green’s Functions (cont.)
16
Method 2 (cont.)
xa bx′
n=1n=2
n=3
n=4
( ) ( )( ) ( )
*m
mm m m
w x xa
x xψ
λ ψ ψ′ ′
=
Therefore, we have
( ) ( ) ( )( ) ( )
( )*
, nn
n n n n
w x xG x x x
x xψ
ψλ ψ ψ
′ ′′ =
∑
Hence
( ) ( ) ( ) ( )2b
n n na
x x x w x dxψ ψ ψ≡ ∫Note :
Application: Transmission Line
17
A short-circuited transmission line with a distributed current source:
The distributed current source is a surface current.
( )I z
+- ( )V z
( )dsI z0z = z h=
( ) ( ) [ ]A/mds sxI z J z=
x
z
( ) ( ) 1Ads sI z z z Iδ ′= − ⇒ =
Green’s function:
18
An illustration of the Green’s function:
Application: Transmission Line (cont.)
( ) ( ) ( )0
,h
dsV z I z G z z dz′ ′ ′= ∫
The total voltage due to the distributed current source is then
+- ( ) ( ),V z G z z′=
z z′=
( )I z0z = z h=
z z′=1AsI =
19
Take the derivative of the first and substitute from the second:
ds
dV j LIdzdI I j CVdz
ω
ω
= −
= −
Telegrapher’s equations for a distributed current source:
(Please see the Appendix.)
Application: Transmission Line (cont.)
( )2
2ds
d V j L I j CVdz
ω ω= − −
Note: j is used instead of i here.
L = inductance/meterC = capacitance/meter
20
Hence
zk LCω=where
Application: Transmission Line (cont.)
( )2
22
dz s
d V k V j L Idz
ω+ = −
( ) ( ) ( )2
22
1 dz s
d V zk V z I z
j L dzω
+ = −
or
( ) ( ) ( )2
22
,1 ,zd G z z
k G z z z zj L dz
δω
′ ′ ′+ = − −
Therefore:
21
Application: Transmission Line (cont.)
Compare with:
( ) ( ) ( ) ( ) ( )1 ( )d dP z u z Q z u z f zw z dz dz
− + =
( ) ( ) ( )2
22
1 dz s
d V zk V z I z
j L dzω
+ = −
Therefore:
( ) ( ) ( ) 2, 1, zw z j L P z Q z kω= = =
Note: The self-adjoint property required the operator to be real, which is not the case here since w(z) is not real. However, we can multiply both sides of the equation by j to make the operator real, and then divide the final answer by jat the end. This process does not affect the final result, so it is not done here.
22
Application: Transmission Line (cont.)
( ) ( ) ( )1 1 sin zV z Au z A k z= =
( ) ( ) ( )( )2 2 sin zV z Bu z B k z h= = −
The general solution of the homogeneous equation is:
Method 1
+- ( ) ( ),V z G z z′=
z z′=
( )I z0z = z h=
1A
22
2 0zd V k Vdz
+ =
Homogeneous equation:
23
Application: Transmission Line (cont.)
( )
( )( )
( ) ( )( )
( )( )
( ) ( )( )
2 1
1 2
,
,
,
w z u z u zz z
P z W zG z z
w z u z u zz z
P z W z
′ ′′− < ′ ′ ′ =
′ ′ ′− > ′ ′
The Green’s function is:
( ) ( )1 sin zu z k z=
( ) ( )( )2 sin zu z k z h= −
+- ( ) ( ),V z G z z′=
z z′=
( )I z0z = z h=
1A
24
Application: Transmission Line (cont.)
The final form of the Green’s function is:
( )( ) ( )( ) ( )
( )
( ) ( ) ( )( )( )
sin sin,
,sin sin
,
z z
z z
k z h k zj L z z
W zG z z
k z k z hj L z z
W z
ω
ω
′−′− <
′′ = ′ − ′− > ′
( ) ( ) ( ) ( ) ( )( ) ( ) ( )( ) ( )1 2[ , ] sin cos cos sin sinz z z z z z zW z W u z u z k k z k z h k z k z h k k h ′ ′ ′ ′ ′ ′ ′= = − − − =
where
+- ( ) ( ),V z G z z′=
z z′=
( )I z0z = z h=
1A
25
Application: Transmission Line (cont.)
The eigenvalue problem is
Method 2
+- ( ) ( ),V z G z z′=
z z′=
( )I z0z = z h=
1A
22
21
zd k
j L dzψ ψ λψ
ω
− + =
22
2ddzψ λ ψ′= − ( )2 2
zj L kλ ω λ′ ≡ +
This may be written as
where
Note: A minus sign in introduced in the
eigenvalue problem for convenience.
26
Application: Transmission Line (cont.)We then have:
( ) ( )2
22 , 0 0d h
dzψ λ ψ ψ ψ′= − = =
( ) ( )sinn
n
z znh
ψ λπλ
′=
′ =
The solution is:
( ) ( ) ( ) ( ) ( ) ( ) ( )* 2
0
sin2
b h
n n n na
n z hz z z z w z dz j L dz j Lhπψ ψ ψ ψ ω ω = = =
∫ ∫Note :
( )2
2
sin
1
n
n z
n zzh
n kj L h
πψ
πλω
=
= −
( )2 21n n zk
j Lλ λ
ω′= −Recall :
27
Application: Transmission Line (cont.)
( ) ( ) ( )( ) ( )
( )*
, nn
n n n n
w z zG z z z
z zψ
ψλ ψ ψ
′ ′′ =
∑
We then have:
where
( )2
2 2 21 1n n z z
nk kj L j L h
πλ λω ω
′= − = −
( ) sinnn zz
hπψ =
( ) ( ) ( )2n nhz z j Lψ ψ ω=
( )w z j Lω′ =
28
Application: Transmission Line (cont.)
The final solution is then:
( ) ( ) 21 2
sin2, sin
nz
n zn zhG z z j L
h hnkh
ππω
π
∞
=
′ ′ = − −
∑
29
Summary
Application: Transmission Line (cont.)
( ) ( ) 21 2
sin2, sin
nz
n zn zhG z z j L
h hnkh
ππω
π
∞
=
′ ′ = − −
∑
( )( ) ( )( ) ( )
( )
( ) ( ) ( )( )( )
sin sin,
sin,
sin sin,
sin
z z
z z
z z
z z
k z h k zj L z z
k k hG z z
k z k z hj L z z
k k h
ω
ω
′−′− <
′ = ′ − ′− >
zk LCω=+- ( ) ( ),V z G z z′=
z z′=
( )I z0z = z h=
1A
30
Application: Transmission Line (cont.)
Other possible Green’s functions for the transmission line:
We solve for the Green’s function that gives us the current I(z)due to the 1A parallel current source.
We solve for the Green’s function giving the voltage due a 1V series voltage source instead of a 1A parallel current source.
We solve for the Green’s function giving the current to due a 1V series voltage source instead of a 1A parallel current source.
Cavity Model for Patch Antenna
Next we use the cavity model and the method of eigenfunction expansion to solve for the input impedance of the rectangular
microstrip patch antenna.
31
xL
z
h rε
Cavity Model
0 0
0 0
e
e
x x Ly y W= + ∆
= + ∆
Note: ∆L is often chosen from Hammerstad’s formula∆W is often chosen from Wheeler’s formula
32
The coordinates (x0, y0) are measured from the corner of the physical patch.
Accounting for fringing:
0 0( , )e ex yPMC
eL
eW
x
y
PEC
22
e
e
L L LW W W
= + ∆= + ∆
( )
( )
0.3 0.264/ 0.412
0.258 0.8
effr
effr
WhL h
Wh
ε
ε
+ + ∆ = − +
ln 4/W hπ
∆ =
Cavity Model (cont.)
eff0e rck k ε=Let
( )effeff1rc r jlε ε= −
eff eff1 1 1 1 1tan
d c sp sw
lQ Q Q Q Q
δ= = = + + +
Assume no z variation (the probe current is constant in the z direction.)
33
0 0( , )e ex yPMC
eL
eW
x
y
PEC
1tan ddQ
δ =Note :
Cavity Model (cont.)We first derive the Helmholtz equation for Ez.
Substituting Faraday’s law into Ampere’s law, we have
efficH J j E
E j Hωε
ωµ∇× = +∇× = −
( )
( )( )
eff
2
2 2
2 2
1 ic
ie
ie
ie
E J j Ej
E j J k E
E E j J k E
E k E j J
ωεωµ
ωµ
ωµ
ωµ
− ∇× ∇× = +
⇒∇× ∇× = − +
⇒∇ ∇⋅ −∇ = − +
⇒∇ + =34
Cavity Model (cont.)Hence
Denote
2 2 iz e z zE k E j Jωµ∇ + =
( , ) ( , )zx y E x yψ =
( , ) ( , )izf x y j J x yωµ=
Then
35
2 2 ( , )ek f x yψ ψ∇ + =
Introduce eigenfunctions of the 2-D Laplace operator:
For rectangular patch we have, from separation of variables,
( , )mn x yψ2 2( , ) ( , )mn mn mnx y x yψ λ ψ∇ = −
0mnCn
ψ∂=
∂
2 22
( , ) cos cosmne e
mne e
m x n yx yL W
m nL W
π πψ
π πλ
=
= +
36
2mnλ− = eigenvalue
Cavity Model (cont.)
Note:The eigenvalues are
real and the eigenvalues are
orthogonal.
Cavity Model (cont.)Assume an “eigenfunction expansion”:
Hence
,( , ) ( , )mn mn
m nx y A x yψ ψ=∑
2 2 ( , )ek f x yψ ψ∇ + =
2 2
, ,( , )mn mn e mn mn
m n m nA k A f x yψ ψ∇ + =∑ ∑
Using the properties of the eigenfunctions, we have
( )2 2
,( , ) ( , )mn e mn mn
m nA k x y f x yλ ψ− =∑
This must satisfy
37
( ), 0,1, 2m n =
Cavity Model (cont.)Multiply the previous equation by and integrate.
Note that the eigenfunctions are orthogonal, so that
Denote
( , )m n x yψ ′ ′
*( , ) ( , ) 0 ( , ) ( , )mn m nS
x y x y dS m n m nψ ψ ′ ′ ′ ′= ≠∫
2, ( , )mn mn mnS
x y dSψ ψ ψ< > = ∫
( )2 2 , ,mn e mn mn mn mnA k fλ ψ ψ ψ− < > = < >
We then have
38
Note: The eigenfunctions are real, so we can drop the conjugate here.
Cavity Model (cont.)Hence, we have
Using
2 2
, 1,
mnmn
mn mn e mn
fAk
ψψ ψ λ
< >= < > −
2 2
, 1,
iz mn
mnmn mn e mn
JA jk
ψωµψ ψ λ
< >= < > −
The field inside the patch cavity is then given by
,( , ) ( , )z mn mn
m nE x y A x yψ=∑
39
( , ) ( , )izf x y j J x yωµ=
Cavity Model (cont.)
2
2
2
2
*
*
*
,
*
,
2
,
1 ( , )21 ( , )212
1 ,2
,1
1
12
,
,
,
iin z z
V
iz z
S
imn z
m nS
imn z
m n
imn z
m n mn mn
mn
imn z
mn
e
mn e
m
mn
n
A
P E x y J dV
h E x y J dS
h J dS
h J
Jh j
k
Jjk
λ
ψω
ψ
ψ
ψωµ
ψ
µψ ψ λ
ψ
< >
= −
= −
= −
= − < >
< > = −
−
< >
< >
−
∫
∫
∑∫
∑
∑
To calculate the input impedance, we first calculate the complex power going into the patch as
40
The integral is over the surface of the probe current.
Cavity Model (cont.)
Also, from circuit theory,
212in in inP Z I=
so
22 in
inin
PZI
=
41
2
2 2 2,
,1 1,
imn z
inm n mn mn e mnin
JZ j h
kI
ψωµ
ψ ψ λ
< > = − < > − ∑
Hence we have
, 0 0m n m n
∞ ∞
= =
=∑ ∑ ∑
Cavity Model (cont.)For the rectangular patch:
where
2 22
eff0
cos cosmne e
mne e
e rc
m x n yL W
m nL W
k k
π πψ
π πλ
ε
=
= +
=
( )effeff1rc r jlε ε= −
2 2
0 0
, cos cose eL W
mn mne e
m x n ydx dyL Wπ πψ ψ
=
∫ ∫
42
We need:
Cavity Model (cont.)
so
( )( )0 0, 1 12 2
e emn mn m n
W Lψ ψ δ δ = + +
0
1, 00, 0m
mm
δ=
= ≠
43
Cavity Model (cont.)
We also need . , imn zJψ
44
To calculate this, assume a strip model for the probe feed:
0 0( , )e ex y
eL
eW
x
y
0 0( , )e ex ypWpa
0 0( , )e ex y
( )( )
0 022
0
, ,2 2
2
4
p pe einsz
p e
p p
W WIJ y y y yW
y y
W a
π
= ∈ − +
− −
=
For a “Maxwell” strip current assumption, we have:
Note: The total probe current is Iin.
45
0 0( , )e ex ypW
Cavity Model (cont.)
( ) 0 0
32
, ,2 2
4.482
p pe einsz
p
p p p
W WIJ y y y yW
W a e a
= ∈ − +
=
For a uniform strip current assumption, we have:
Note: The total probe current is Iin.
46
0 0( , )e ex ypW
Cavity Model (cont.)
0
0
20
2
20
0
2
0 0 0
, cos cos
cos cos '
cos cos cos sin
pe
pe
p
p
Wy
ei in
mn zW e e p
y
We
ein
Wp e e
e e ein
p e e e e
m x In yJ dyL W W
I m x n y y dyW L W
I m x n y n yn yW L W W W
π πψ
π π
π π ππ
+
−
+
−
=
′ = +
′= −
∫
∫
2
2
0 0
sin
cos cos sinc2
p
p
W
W e
e epin
pp e e e
n y dyW
n WI m x n y WW L W W
π
ππ π
+
−
′′
=
∫
Assume a uniform strip current model:
0ey y y′= +
Use
Integrates to zero
47
Cavity Model (cont.)
Uniform Model (cont.)
Hence
0 0, cos cos sinc2
e epi
mn z ine e e
n Wm x n yJ IL W W
ππ πψ
=
sinc2
p
e
n WWπ
Note: It is the term that causes the series for Zin to converge.
Note:We cannot assume a probe of zero radius,
or else the series will not converge – the input reactance will be infinite.
48
Cavity Model (cont.)Summary
2
2 2 20 0
,1 1,
imn z
inm n mn mn e mnin
JZ j h
kI
ψωµ
ψ ψ λ
∞ ∞
= =
< > = − < > − ∑∑
49
( )( )0 0, 1 12 2
e emn mn m n
W Lψ ψ δ δ = + +
0 0, cos cos sinc2
e epi
mn z ine e e
n Wm x n yJ IL W W
ππ πψ
=
eff0e rck k ε= ( )eff
eff1rc r jlε ε= −
where
2 2
mne e
m nL Wπ πλ
= +
eff 1/l Q=
32 4.482p p pW a e a=
Cavity Model (cont.)CAD Formulas for Q Factors*
50
( )effeff1rc r jlε ε= − eff eff
1 1 1 1 1tand c sp sw
lQ Q Q Q Q
δ= = = + + +
1tand
d
Qδ
=
where
0 0ave
( )2c
s
k hQR
η =
( )ave patch ground / 2s s sR R R= +
1 0
3 116 /
ersp
e
LQpc W hε
λ
≈
hed
hed1r
sw spr
eQ Qe
= − *Derived in ECE 6345.
1sR
σδ=
2δωµσ
=
Cavity Model (cont.)CAD Formulas for Q Factors (cont.)
51
( )
hed3
01
1
3 1 11 14
r
r
e
k hc
πε
=
+ −
1 2
1 2 / 51r r
cε ε
= − +
( ) ( ) ( ) ( )
( ) ( )
2 4 2220 2 4 0 2 0
2 22 2 0 0
3 11 210 560 5
170
ap k W a a k W c k L
a c k W k L
= + + + +
+
2 0.16605a = −
4 0.00761a =
2 0.0914153c =−
Appendix
52
Allow for distributed sources
+−L z∆
C z∆dsi z∆
dsv z∆
v+v -
+
-
+
-
∆z
i+i-
This appendix presents a derivation of the telegrapher’s equations with distributed sources.
53
( ) ( )d ds sv v v z L z i i z
t+ − −∂− = ∆ − ∆ + ∆
∂
ds
v iv Lz t∂ ∂
= −∂ ∂
0 :z∆ →
+−L z∆
C z∆dsi z∆
dsv z∆
v+v -
+
-
+
-
∆z
i+i-
Appendix (cont.)
54
0 :z∆ →
+−L z∆
C z∆dsi z∆
dsv z∆
v+v -
+
-
+
-
∆z
i+i-
( )ds
vi i i z C zt
++ − ∂− = ∆ − ∆
∂
ds
i vi Cz t∂ ∂
= −∂ ∂
Appendix (cont.)
55
In the phasor domain:
ds
ds
v iv Lz ti vi Cz t
∂ ∂= −
∂ ∂∂ ∂
= −∂ ∂
ds
ds
dV V j LIdzdI I j CVdz
ω
ω
= −
= −
jt
ω∂→
∂
Appendix (cont.)