Auxiliary material to the lecture

Theorie der Elektrotechnik (engl.)

(437.253)
by

Univ.-Prof. Dipl.-Ing. Dr.techn. Oszkár Bíró

1
Contents
1. Maxwell‘s equations
1.1 Classification of electromagnetics
1.2 Fundamentals of circuit theory
1.3 Energy conversion in electromagnetic field
1.4 Existence of a unique solution to Maxwell’s equations
2. Static fields
2.1 Boundary value problems for the scalar potential
2.2 Analytic methods for solving the Laplace equation
2.2.1 Method of fictitious charges (method of images)
2.2.2 Separation of variables
2.2.3 Conformal mapping
2.3 Numerical methods for solving the boundary value problems for the scalar
potential
2.3.1 Method of finite differences
2.3.2 Variational problem of electrostatics
2.3.3 Ritz‘s procedure
2.3.4 The method of finite elements
2.4 Integral equations for the scalar potential
2.4.1 Fundamentals of potential theory
2.4.2 Integral equation for the surface charge density 2
2.4.3 The method of boundary elements
 2.5 Boundary value problems for the vector potential
2.5.1 Planar 2D problems
2.5.2 Axisymmetric 2D problems
2.5.3 3D problems
3. Quasi-static fields
3.1 Some analytical solutions of the boundary value problem for the magnetic
vector potential
3.1.1 Current flow in an infinite conducting half space
3.1.2 Current flow in an infinite conducting plate
4. Electromagnetic waves
4.1 Planar waves
4.2 Electromagnetic waves in homogeneous, infinite space
4.2.1 Solution of Maxwell’s equations with retarded potentials
4.2.2 Hertz dipole
4.3 Guided waves
4.3.1 TM and TE waves
4.3.2 Waves in rectangular waveguides

3
 1. Maxwell‘s equations

∂D
I . curlH= J + , generalized Ampere‘s law,
∂t
∂B
II . curlE = − , Faraday‘s law,
∂t
III . divB = 0, magnetic flux density is source free,
IV . divD = ρ , Gauss‘ law.

Consequence: law of continuity

(implied by the divergence of the first equation
∂ρ
divJ = − . + the fourth equation:
∂t ∂D 
div J +  = 0, divD = ρ .)
 ∂t  4
 Field quantities in voltage sources
Charges are separated in voltage sources due to non-electric
(e.g. chemical) processes. The impressed field intensity Ee is
a fictitious electric field intensity which would give rise to
the same amount of charge separation:
P1 P2 P2

uq = ∫ E e ⋅ dr = − ∫ E e ⋅ dr, u = ∫ E ⋅ dr
P2 P1 P1
P1
+++++++++
i Γ: Cross section P1
JΓ
P2
J
Ee J E u iRq = ∫ ⋅ dr = − ∫ ⋅ dr
_____
γ : conductivity P2
γΓ P1
γ
P2
J
Non-ideal voltage source: u = uq − iRq ⇒ E = −E e +
γ
Ideal voltage source: u = u q ⇒ E = −E e , γ → ∞
J = γ (E + E e )
5
Material relationships:
D = εE, B = µH, J = γ (E + E e ).

Energy and power density:

1 1  D B

w = we + wm = E ⋅ D + H ⋅ B  = ∫ E ⋅ dD + ∫ H ⋅ dB ,
2 2  0 0 
2
J
p= .
γ
Energy and power loss in a volume Ω:

W = ∫ wdΩ, P = ∫ pdΩ.
Ω Ω

6
 1.1 Classification of electromagnetics

 ∂ = 0
1. Static fields  
 ∂t 
=curlH J ,= curlE 0= divD ρ=
, divB 0,= , divJ 0.

electrostatic field: =
curl =
E 0, divD ρ=
, D ε E.

magnetostatic field: =
curl =
H J , div B µ H.
B 0,=

static current field: =

curl =
E 0, div J γ (E + Ee ).
J 0,=

7
  ∂D 
2. Quasi-static field  J >> 
 ∂t 

curlH = J ,
∂B
curlE = − ,
∂t
divB = 0,

B = µH, J = γ (E + E e ).

8
3. Electromagnetic waves
∂D
curlH= J + ,
∂t
∂B
curlE = − ,
∂t
divB = 0,
divD = ρ ,

D = εE, B = µH, J = γ (E + E e ).

9
 1.2 Fundamentals of circuit theory

Circuit signals:
2
u12 = ∫ E ⋅ dr, iΓ = ∫ J ⋅ ndΓ,
1 Γ

QΩ = ∫ ρdΩ = ∫ divDdΩ = ∫ D ⋅ ndΓ, Φ Γ = ∫ B ⋅ ndΓ.

Ω Ω Γ Γ

Circuit elements:

resistor: u = Ri,
du
ideal capacitor: i=C ,
dt
di
ideal inductor: u=L .
dt
10
 Kirchhoff’s current law:

Γ n ∂D
i2 =0 on Γ, since Γ
i1 ∂t
cuts no capacitor.
Ω
i3

i4
Integrating the law of continuity over Ω:
 ∂ρ   ∂D   ∂D 
∫Ω  divJ + 
∂t 
d Ω
= ∫
Ω
div 

J + 
∂t 
d Ω
= ∫
 
Γ 
J +  ⋅ nd =
∂t 
Γ ∫ J ⋅ nd=
Γ
Γ 0.

∑
ν
iν = 0
11
 Kirchhoff‘s voltage law:

uqµ uL1
uC1
uRµ uR1
uq1
uCµ
Γ
CΓ
uLµ
uLm
uCm
uRm
uqm
Integrating Faraday’s law over Γ:
 ∂B  d
∫Γ  curl E + 
∂t 
⋅ n d=Γ ∫

CΓ
E ⋅ dr + ∫
dt Γ
B ⋅ nd=
Γ 0.

(u µ + u µ + u µ + u µ ) = 0
m

∑
µ=1
q R C L
12
 1.3 Energy conversion in electromagnetic field

∂D
−E ⋅ curlH =−E ⋅ J − E ⋅ I. M.Eq.⋅ (-E)
∂t
∂B
H ⋅ curlE =− H ⋅ II. M.Eq. ⋅ H
+ ∂t
∂B ∂D
H ⋅ curlE − E ⋅ curlH =−H ⋅ − E⋅J − E⋅
∂t ∂t

H ⋅ curlE − E ⋅ curlH= H ⋅ ( ∇ × E ) − E ⋅ ( ∇ × H )=
= ∇ ⋅ (E × H c ) + ∇ ⋅ (E c × H ) = ∇ ⋅ (E × H ) = div(E × H )

13
 Integration over a volume Ω:

 ∂B ∂D 
− ∫H⋅ + E ⋅ dΩ = ∫ E ⋅ JdΩ + ∫ div(E × H )dΩ
Ω
∂t ∂t  Ω Ω

∂B ∂ ∂w ∂1 ∂1
B
H⋅ = ∫ H ⋅ dB = m  = H ⋅B = µ H 
2

∂t ∂t 0 ∂t  ∂t 2 ∂t 2 for

linear
∂D ∂ ∂we  ∂ 1 ∂1
D
2 media
E⋅ = ∫ E ⋅ dD = = E⋅D = εE 
∂t ∂t 0 ∂t  ∂t 2 ∂t 2 
2
J J
J = γ (E + E e ) ⇒ E = − Ee ⇒ E ⋅ J = − Ee ⋅ J = p − Ee ⋅ J
γ γ

∫ div(E × H )dΩ = ∫ (E × H ) ⋅ ndΓ

Ω Γ
(Gauss’ theorem)

14
 Poynting’s theorem:

2
d J
− ∫ (wm + we )dΩ = ∫ dΩ − ∫ E e ⋅ JdΩ + ∫ (E × H ) ⋅ ndΓ
dt Ω Ω
γ Ω Γ

right hand side:

reasons for the decrease of energy in a volume

S = E× H Poynting’s vector: power through unit surface

∫ S ⋅ ndΓ
Γ
is the power leaving the volume Ω through
radiation.

15
 1.4 Existence of a unique solution to Maxwell’s equations

The solution of Maxwell’s equations in a volume Ω with the

boundary Γ is unique for t > t0 , provided that the

1. initial conditions
H (r, t = t0 ) = H 0 anf (r ), E(r, t = t0 ) = E 0 anf (r ), ∀r ∈ Ω, and the
2. boundary conditions for the tangential components
=
H t (r, t ) H 0tan (r, t ) or=
Et (r, t ) E0tan (r, t ), ∀r ∈ Γ, t ≥ t0

are fulfilled. Both the functions H 0 anf (r ), E 0 anf (r ),

H 0tan (r, t ) and E0tan (r, t ) and the impressed field intensity Ee
must be known. The material properties are assumed to be linear.
16
Proof:

Assumption:
two solutions E′, H′ and E′′,H′′ exist.
Their differences: E 0 = E′ − E′′, H 0 = H′ − H′′
satisfy Maxwell’s equations (these are linear). The initial and
boundary conditions for the difference fields are homogenous
and Ee0=0. Therefore, Poynting’s theorem :
2
d 1 1 2 J0
− ∫ µ + ε 0  dΩ = ∫ dΩ + ∫ (E 0 × H 0 ) ⋅ ndΓ
2
 H E
dt Ω  2
0
2  Ω
γ Γ

Since the boundary conditions imply that either E0 or H0

point in the normal direction n, the vector S 0 = E 0 × H 0
has no normal component. Therefore, the surface integral is zero.
17
 2
d 1 1 2 J0
− ∫  µ H 0 + ε E 0  dΩ = ∫ dΩ
2

dt Ω  2 2  Ω
γ
The right hand side is obviously non-negative, therefore
the quantity whose negative time derivative is written in
the left hand side can never increase. The initial
conditions, however, fix it as zero at the time instant t0
and, since it is obviously non-negative, it can never
decrease either. Therefore it is always zero:
 1 µ H 2 + 1 ε E 2  d Ω = 0.
∫Ω  2 0 2 0 
Therefore,=
E0 0= =
, H 0 0, i.e. E′ E′′ und
= H′ H′′.
q. e. d.
18
2. Static fields

Maxwell’s equations  ∂ = 0  : curlH = J ,

 ∂t  curlE = 0,
divB = 0,
divD = ρ ,
divJ = 0.

electrostatic magnetostatic static current

field: field: field:
curlE = 0, curlH = J , curlE = 0,
divD = ρ , divB = 0, divJ = 0,
D = ε E. B = µ H. J γ (E + Ee ).
=

19
2.1 Boundary value problems for the scalar potential

Electrostatic field and static current field:

curlE =⇒
0 E=− gradV , V: electric scalar potential

Magnetostatic field, if J=0:

curlH =⇒ − gradψ , ψ : magnetic scalar potential

0 H=
Differential equations:
divD = ρ ⇒ − div (εgradV ) = ρ ,
generalized Laplace-Poisson
divB = 0 ⇒ − div ( µgradψ ) = 0, and Laplace equation.
divJ = 0 ⇒ − div(γgradV ) = 0,
20
The solution of the Laplace-Poisson equation in
unbounded free space (ε=ε0):

1 ρ (r′)d Ω′
V (r )
4πε 0 ∫=
Ω
r − r′
, V (r → ∞) 0.

In regions free of charges, the electrostatic field is also

described by the generalized Laplace equation:

− div (εgradV ) = 0.

21
Boundary conditions:
=
Dirichlet boundary condition: V V0 (known) on Γ D ,
= ψ ψ 0 (known) on Γ D .
This means the prescription of Et or Ht.

ΓD is typically constituted by electrodes in case of

electrostatic and static current field. Indeed:
E t = 0 ⇒ V = konstant.
Magnetostatic field: interface to highly permeable
regions, to magnetic walls ( µ → ∞).

µ = µ0 µ → ∞
since H t = H tEisen
H t = 0 H Eisen → 0
22
 ∂V
=
Neumann boundary condition: ε σ (known) on Γ N ,
∂n
∂V
γ = 0 on Γ N ,
∂n
∂ψ
= µ b (known) on Γ N .
∂n
This means the prescription of Dn, Jn or Bn.

In case of static current field, ΓN is the interface to

the non-conducting region.
If σ=0 in the electrostatic case and b=0 in the
magnetostatic case, ΓN is a surface parallel to the
field lines. Otherwise, Dn, Bn is known on ΓN.

23
Boundary value problems:
− div(ε gradV
= ) ρ in Ω,
electrostatic
∂V
field: V= V0 on Γ D , ε σ on Γ N .
=
∂n

− div( µ grad=
ψ ) 0 in Ω,
magnetostatic
∂ψ
field: ψ= ψ 0 on Γ D , µ =
b on Γ N .
∂n

− div(γ gradV
= ) 0 in Ω,
static current
∂V
field: V= V0 on Γ D , γ =
0 on Γ N .
∂n

24
Uniqueness of the solution of the boundary value problem:
Due to the analogy, it is sufficient to treat the electrostatic
case.
Proof:
Assumption: two solutions V ′ und V ′′ exist.
Their difference: V = V ′ − V ′′ satisfies the boundary value
problem − div(ε gradV= ) 0 in Ω,
∂V
V= 0 on Γ D , ε = 0 on Γ N .
∂n

Identity:

div(VεgradV ) = ε gradV + Vdiv(εgradV ).

2

25
Integration over a domain Ω with boundary Γ:

∫ ε gradV dΩ = − ∫ Vdiv(εgradV )dΩ + ∫ VεgradV ⋅ ndΓ.

2

Ω Ω Γ

According to the boundary value problems for V, the

∂V
right hand side is zero (εgradV ⋅ n = ε ).
∂n

Hence gradV =0 ⇒ V =constant in Ω.

V 0 on Γ D , V = 0 in Ω ⇒ V ′ = V ′′.
Since, however=

q. e. d.

26
2.2 Analytic methods for solving the Laplace equation
∂ 2V ∂ 2V ∂ 2V
Laplace equation: divgradV = ∆V = 2 + 2 + 2 = 0,
∂x ∂y ∂z
(divgradψ = 0).

electrostatic field: ρ 0,=

= ε =
constant ( ε 0 ),
magnetostatic field: = , µ
J 0= =
constant ( µ 0 ),
static current field: γ = constant.

∆V = 0 ⇒ V is a harmonic function.

There exist infinitely many harmonic functions!

27
 Dirichlet or Neumann boundary conditions :

V=constant on ΓD

∉Ω ∉Ω
∆V = 0 in Ω

∂V
= 0 on Γ N
∂n

Analytic methods:
Generation of harmonic functions and selection of the one
satisfying the boundary conditions.
This yields the true solution, since that is unique.
28
2.2.1 Method of fictitious charges (method of images)

The potential function of a point charge is harmonic in all

points in space except in the point where the charge is
located.
Q ′ ′ ′
z ( x , y , z ) r = xe + ye + ze , r′ = x′e + y′e + z ′e
x y z x y z
r′
y
[ ]
1
Q −
r V ( x, y , z ) = ( x − x′)2 + ( y − y′)2 + ( z − z′)2 2
4πε 0
x

29
 Proof:
∂V
[ ]
3
Q −
=− ( x − x′) ( x − x′)2 + ( y − y′)2 + ( z − z′)2 2 =
∂x 4πε 0
Q
=− (x − x )r − r
′ ′
−3

4πε 0
∂ 2V
∂x 2
= −
Q
4πε 0
r [
− r ′
−3
− 3( x − x ′ )2
r − r ′
−5
]
∂ 2V ∂ 2V ∂ 2V Q  3 ( x − x′ ) + ( y − y ′ ) + ( z − z ′ ) 
2 2 2
+ 2 + 2 =−  −3 =
∂x 2
∂y ∂z 4πε 0  r − r′ 3
r − r′
5


Q  1 r −r 
′
2

=
− 3 − 5 
0, if r ≠ r′.
=
4πε 0  r − r′ 3
r − r′ 
q. e. d. 30
In two dimensions (planar problems, ∂ = 0),
∂z
the potential function of an infinitely long line charge is
harmonic.

τ ( x′, y′) r = xe x + ye y , r′ = x′e x + y′e y

y
r′ τ r0
V ( x, y ) = ln
2πε 0
[ ]
1
r
x ( x − x′ ) + ( y − y ′ )
2 2 2

V ( x, y ) =
τ 
 0
2πε 0 
ln r −
1
2
ln ([x − x ′ )2
+ ( y − y ′ )]
2 



31
∂V τ x − x′
=−
∂x 2πε 0 ( x − x′)2 + ( y − y′)2
∂ 2V τ ( x − x′) + ( y − y′) − 2( x − x′)
2 2 2
=− =
∂x 2
2πε 0 [
( x − x′ ) + ( y − y ′ )
2
]
2 2

τ ( x − x′)2 − ( y − y′)2
=
[ ]
2πε 0 ( x − x′)2 + ( y − y′)2 2

∂ 2V ∂ 2V τ ( x − x′)2 − ( y − y′)2 + ( y − y′)2 − ( x − x′)2

+ 2 = =0
∂x 2
∂y 2πε 0 [
( x − x′ ) + ( y − y ′ )
2
]
2 2

q. e. d.

32
Due to its linearity, the Laplace equation is satisfied by the
potential function of an arbitrary charge distribution (in regions
free of charges).
Satisfaction of the boundary conditions: a fictitious charge
distribution within the electrodes should have equipotential
surfaces which coincide with the electrodes.
Examples:
• One point charge: concentric spherical surfaces (spherical
capacitor)
• One infinitely long line charge: concentric cylindrical
surfaces (cylindrical capacitor)
• Line dipole: non-concentric cylindrical surfaces
• Mirror image over a plane: infinite conducting plane
• Multiple mirror images over planes: planar electrodes
forming an angle (α = 180 / n)
33
2.2.2 Separation of variables
Laplacian in a general orthogonal coordinate system:

∆u = div( gradu )
1  ∂  h2 h3 ∂u  ∂  h3 h1 ∂u  ∂  h1h2 ∂u 
∆u =   +  +  
h1h2 h3  ∂x1  h1 ∂x1  ∂x2  h2 ∂x2  ∂x3  h3 ∂x3 

∂ 2u ∂ 2u ∂ 2u
∆u ( x, y, z ) = 2 + 2 + 2
∂x ∂y ∂z

1 ∂  ∂u  1 ∂ 2u ∂ 2u
∆u (r , φ , z ) = r  + 2 + 2
r ∂r  ∂r  r ∂φ 2
∂z

1 ∂  2 ∂u  1 ∂  ∂u  1 ∂ 2u
∆u (r ,θ , φ ) = 2  r + 2  sin θ + 2
r ∂r  ∂r  r sin θ ∂θ  ∂θ  r sin 2 θ ∂φ 2
34
Method of separation:
Assumption: V ( x1 , x2 , x3 ) = X 1 ( x1 ) X 2 ( x2 ) X 3 ( x3 )

It is attempted to reduce the Laplace equation (a partial

differential equation) for V to three ordinary differential equations
for the functions X1, X2 and X3. The general solutions of these
equations yield special solutions of the Laplace equation by
means of the above assumption. Due to the linearity of the
Laplace equation, it is satisfied by any linear combination of
these solutions. The solution of a particular boundary value
problem is the linear combination which satisfies the boundary
conditions of the problem. This is only possible if the boundary
conditions are specified along surfaces xi = constant.

35
Cartesian coordinates, 2D case:
Assumption: V ( x, y ) = X ( x)Y ( y )
∂ 2V ∂ 2V
Laplace equation: + 2 = 0.
∂x 2
∂y
d 2 X ( x) d 2Y ( y )
Y ( y) 2
+ X ( x) 2
=0
dx dy
Division by X(x)Y(y) yields:
1 d 2 X ( x) 1 d 2Y ( y )
2
+ 2
=0
X ( x) dx Y ( y ) dy
      
f ( x) g ( y)

This is only possible if

=
f ( x) constant, g ( y ) constant.
36
Hence, the ordinary differential equations are:

d 2 X ( x) d 2Y ( y )
2
= fX ( x), 2
= gY ( y ),
dx dy
where g = − f .

f = − p 2 , g = p 2 : X ( x) = C1 cos px + C2 sin px
Y ( y ) = C3e py + C4 e − py = C3′ cosh py + C4′ sinh py

=
One solution: V ∑ n
A
n
± ± pn y
e cos pn x + B ± ± pn y
ne sin pn x

A further solution is obtained by interchanging x and y:

=V ∑ n
A
n
± ± pn x
e cos pn y + B ± ± pn x
ne sin pn y
37
 Example: boundary conditions:
y V = f (x) (1) y = b, V = f ( x);
(2) y = 0, V = 0; ⇒ sinh pn y,
b (3) x = 0, V = 0;  sin pn x, pn = nπ / a,
V =0 V =0 ⇒
(4) x = a, V = 0.  n = 1, 2, ... .
V =0 ∞
 nπy   nπx  (2), (3), (4)
x V ( x, y ) = ∑ Bn sinh   sin  :
a n =1  a   a  satisfied.
∞
 nπx 
Fourier series of f (x): f ( x) = ∑ bn sin  .
n =1  a 
∞
 nπb   nπx   nπb 
V ( x, b) = ∑ Bn sinh   sin   ⇒ Bn sinh   = bn
n =1  a   a   a 
 nπy 
sinh  
 a  sin  nπx .
∞
V ( x, y ) = ∑ bn  
π
sinh    a 
n =1
n b
 38
 a 
Cylindrical coordinates, 2D case:
Assumption: V (r , φ ) = R (r )Φ (φ )
1 ∂  ∂V  1 ∂ 2V
Laplace equation: r + 2 = 0.
r ∂r  ∂r  r ∂φ 2

1 d  dR (r )  1 d 2 Φ (φ )
Φ (φ ) r  + R(r ) 2 = 0.
r dr  dr  r dφ 2

Multiplication by r2 and division by R(r)Φ(φ) yield:

1 d  dR (r )  1 d 2 Φ (φ )
r r + = 0.
R (r ) dr  dr  Φ (φ ) dφ 2
     
f (r ) g (φ )

This is only possible if

=
f (r ) constant, g (φ ) constant.
39
Hence, the ordinary differential equations are:

d  dR (r )  d 2 Φ (φ )
r r  = fR (r ), = gΦ (φ ),
dr  dr  dφ 2

where g = − f .

Since Φ(φ) must be periodic with the period 2π , the only

possibility is g = -n2 ( n: integer)
Φ (φ ) = C3 cos nφ + C4 sin nφ (n > 0).
The case n = 0 yields Φ (φ ) = C3 + C4φ .
This is only periodic if C4 = 0. The case C4 ≠ 0,
makes sense only if the problem region is 0 ≤ φ ≤ φmax
where φmax < 2π .
Ω
40
 d  dR (r ) 
r r  = n 2
R (r ).
dr  dr 
dR (r )
a) n = 0 : r = C1 , R (r ) = C1 ln r + C2 .
dr

b) n > 0 : Assumption: R (r ) = r α ,
dR (r ) α −1 dR (r ) α d  dR (r )  2 α −1
= αr , r = αr ,  r  = α r ,
dr dr dr  dr 
α 2 r α = n 2 r α ⇒ α = ± n : R(r ) = C1r n + C2 r − n

Solution:
V= ( C1 ln r + C2 ) ( C3 + C4φ ) + ∑ ( An± r ± n cos nφ + Bn± r ± n sin nφ )
n

41
Example: cylindrical capacitor
Va ∂
Ra Axisymmetry: = 0 ⇒ V = C1 ln r + C2 .
Vi
Ri ∂φ

boundary conditions: Vi = C1 ln Ri + C2 ,
Va = C1 ln Ra + C2 .
Va − Vi Vi ln Ra − Va ln Ri
C1 = , C2 = .
ln Ra Ri ln Ra Ri
Va ln r Ri − Vi ln r Ra
V= .
ln Ra Ri

The same solution is obtained by the method of fictitious

charges.
42
 Example: dielectric cylinder in a homogeneous field

homogeneous field: V = − E0 r cos φ .

E 0 = E0 e x ε > 1
r

R Proof:
∂V 1 ∂V
y Er = − = E0 cos φ , Eφ = − = − E0 sin φ .
Eφ ∂r r ∂φ
r E E = E cos φ − E φ sin φ = E cos 2
φ + E sin 2
φ = E0 ,
φ
x r 0 0
r
x E = E sin φ + E cos φ = E cos φ sin φ − E sin φ cos φ = 0.
y r φ 0 0

Vi (r , φ ), wenn r < R, q. e. d.

V (r , φ ) = 
boundary Va (r , φ ), wenn r > R.
lim Vi (r , φ ) < ∞ ⇒ Vi = ∑ ( Ain r cos nφ + Bin r sin nφ ),
n n
conditions:
r →0
n ≥1

lim Va (r , φ ) = − E0 r cos φ ⇒
r →∞

Va = − E0 r cos φ + ∑ ( Aan r − n cos nφ + Ban r − n sin

43
nφ ).
n ≥1
Boundary and interface conditions:

Eφi (r = R ) = Eφa (r = R ) ⇒ Vi ( R, φ ) = Va ( R, φ ) ⇒

Ai1 R = − E0 R + Aa1 R −1 , Ain R n = Aan R − n (n > 1), Bin R n = Ban R − n

∂Vi ( R, φ ) ∂Va ( R, φ )
Dri (r = R ) = Dra (r = R ) ⇒ ε r = ⇒
∂r ∂r
ε r Ai1 = − E0 − Aa1 R − 2 , ε r Ain nR n −1 = − Aan nR − n −1 (n > 1),
ε r Bin nR n −1 = − Ban nR − n −1.
Solution: Ai1 =
−2 ε −1
E0 , Aa1 = r E0 R 2 ,
εr +1 εr +1
Ain = Aan = 0 (n > 1), Bin = Ban = 0.

44
 − 2 E0
Vi (r , φ ) = r cos φ ,
εr +1
εr −1 R2
Va (r , φ ) = − E0 r cos φ + E0 cos φ .
εr +1 r

The field within the

cylinder is homogeneous.

45
2.2.3 Conformal mapping

Consider an arbitrary regular complex function:

w( z ) = u ( x, y ) + jv ( x, y ), z = x + jy.

It realizes a mapping of the x-y plane to the u-v plane:

y z v w
w = w(z )
C′
C x u

This mapping is conformal, i.e. it preserves angles locally.

46
Regular complex functions have the following property:

w( z + ∆x) − w( z ) w( z + j∆y ) − w( z )
w′( z ) = lim = lim ,
∆x →0 ∆x ∆y → 0 j∆y
u ( x + ∆x, y ) − u ( x, y ) v( x + ∆x, y ) − v( x, y ) 
lim  +j =
∆x →0 ∆x ∆x 
 u ( x, y + ∆y ) − u ( x, y ) v( x, y + ∆y ) − v( x, y ) 
= lim  +j ,
∆y →0
 j∆y j∆y 
∂u ∂v ∂u ∂v
+ j =−j + .
∂x ∂x ∂y ∂y

47
 ∂u ∂v ∂v ∂u
Cauchy-Riemann equations: = , =− .
∂x ∂y ∂x ∂y

∂ 2u ∂ 2 v ∂ 2u ∂ 2v ∂ 2u ∂ 2u
= , =− ⇒ 2 + 2 = 0,
∂x 2
∂x∂y ∂y 2
∂y∂x ∂x ∂y
∂ 2v ∂ 2u ∂ 2 v ∂ 2u ∂ 2v ∂ 2v
=− , = ⇒ 2 + 2 = 0.
∂x 2
∂x∂y ∂y 2
∂y∂x ∂x ∂y

Both the real part and the imaginary part of a regular

complex function satisfy the two-dimensional Laplace
equation.

48
The real part or the imaginary part of a regular complex function
is the solution of an electrostatic problem if it is constant along the
electrodes. v
V =u:

v = constant lines
y
z

(equipotential lines) (field lines)

u
V = V2 V = V1
field lines

x u = V2 u = V1
u = constant lines
(equipotential lines)
V = v: v
v = V2

v = constant lines
equipotential lines

In general: C1w( z ) + C2 . u
v = V1
u = constant lines 49
(field lines)
 Example: cylindrical capacitor
Va
Ra w( z ) = ln z = ln re jφ = ln r + jφ ,
Ri y
Vi u = ln r = ln x 2 + y 2 , v = φ = arctan .
x
=u constant=
⇒ r constant: circles.

The boundary conditions can be satisfied if w( z ) = C1 ln z + C2 .

V = u = C1 ln r + C2 . Using the boundary conditions:

Va ln r Ri − Vi ln r Ra
V= .
ln Ra Ri
The same solution is obtained by the method of fictitious
charges or the method of separation of variables.
50
 Example: Two planes normal to each other
y
w( z ) = ln z = ln re jφ = ln r + jφ ,
V =U
y
x u = ln r = ln x + y , v = φ = arctan .
2 2

V =0 x

= ⇒ ϕ constant: radial lines.

v constant =
The boundary conditions can be satisfied if w( z ) = C1 ln z + C2 .

V = v = C1φ + C2 .

2U 2U y
Using the boundary conditions: V = φ= arctan .
π π x

51
2.3 Numerical methods for solving the boundary value
problems for the scalar potential

Disadvantages of analytical methods:

• special geometry
• homogeneous materials

Numerical methods:
• geometry discretized
• taking account of non-homogeneous materials

52
 Discretization of geometry:

V=constant on ΓD

∉Ω ∉Ω

div(εgradV ) = 0 in Ω
∂V
ε= σ on Γ N
∂n
Numerical methods:

The potential is approximated in discrete nodes. The field is

computed by numerical differentiation. The Dirichlet
boundary conditions in the nodes are satisfied exactly, the
Neumann boundary conditions can only be approximately
fulfilled. 53
2.3.1 Method of finite differences

Two-dimensional planar problems are treated only, the

generalization to 3D problems is straightforward.
Homogeneous materials are assumed: Laplace equation.
Generalization for piecewise homogeneous materials is
possible.
Uniform mesh,
generalization for non-
uniform mesh is
Ω possible.

Γ h

54
h
 1 ∂V 1 ∂ 2V 2 1 ∂ 3V 3 1 ∂ 4V 4
Va = VA + h+ h + h + h + ...
1! ∂x 2! ∂x 2
3! ∂x 3
4! ∂x 4

1 ∂V 1 ∂ 2V 2 1 ∂ 3V 3 1 ∂ 4V 4
Vb = VA + h+ h + h + h + ...
1! ∂y 2! ∂y 2
3! ∂y 3
4! ∂y 4

1 ∂V 1 ∂ 2V 2 1 ∂ 3V 3 1 ∂ 4V 4
Vc = VA − h+ h − h + h ± ...
1! ∂x 2! ∂x 2
3! ∂x 3
4! ∂x 4

1 ∂V 1 ∂ 2V 2 1 ∂ 3V 3 1 ∂ 4V 4
Vd = VA − h+ h − h + h ± ...
+ 1! ∂y 2! ∂y 2
3! ∂y 3
4! ∂y 4

Va + Vb + Vc + Vd = 4VA +
∆V = 0
∂ 2V ∂ 2V
+ h ( 2 + 2 ) + O(h 4 )
2
c
b
h ∂x ∂y 

A
a ≈0
d
=0
y
x 4VA − Va − Vb − Vc − Vd = 0
55
h
The equation 4VA − Va − Vb − Vc − Vd =
0, i.e.
b
1
VA = (Va + Vb + Vc + Vd ) A
4 c a
d
corresponds approximately to the mean value
theorem of potential theory (see 2.4.1):
Sphere
1
= V (M ) 2 ∫
4π R Sphere
Vd Γ
M R

56
 Taking account of boundary conditions:
Dirichlet boundary condition on ΓD: Vg = V0 (known).
4Vi − Ve − V f − Vg − Vh = 0 ⇒ 4Vi − Ve − V f − Vh = V0
ΓN

Neumann bondary condition on ΓN:

∂V
g
f l
ε0 = σ at the point o.
∂n
i m o
e k
h n

k: fictitious node outside of Ω

h
∂V Vk − Vm 2hσ
ε0 (o ) ≈ ε 0 = σ ⇒ Vk = Vm +
ΓD ∂n 2h ε0
h
2hσ
4Vo − Vk − Vl − Vm − Vn = 0 ⇒ 4Vo − Vl − 2Vm − Vn =
ε0 57
 Example: node 1: 4V1 − V2 − V5 = 2U
h V=U
node 2: − V1 + 4V2 − V3 − V6 = U
1 2 3
4′ 4 5 6 7 8 8′ node 3: − V2 + 4V3 − V7 = 2U
node 4: − V4′ + 4V4 − V5 = U
∂V V=0 ∂V
ε0 = σ1 ε0 =σ2 2 hσ 1
∂n ∂n 4V4 − 2V5 = U +
ε0
V4′ − V5 − V1 − V4 + 4V5 − V6 = 0
ε0 = σ1 node 5:
2h
⇓ node 6: − V2 − V5 + 4V6 − V7 = 0
2hσ 1
V4′ = V5 +
ε0 node 7: − V3 − V6 + 4V7 − V8 = 0
V8′ − V7 − V7 + 4V8 − V8′ = U
ε0 =σ2 node 8:
2h
⇓ 2hσ 2
2hσ 2 − 2V7 + 4V8 = U +
V8′ = V7 + ε0
ε0 58
Equations system in matrix form:

 4 −1 0 0 −1 0 0 0  V1   2U 
 −1 4 −1 0 0 0 −1 0  V2   U 
    
 0 −1 4 0 0 0 −1 0  V3   2U 
 0 0 0 4 −2 0 0 0  V4  U + 2hσ 1 / ε 0 
   =  
− 1 0 0 −1 4 −1 0 0  V5
  
0

0 −1 0 0 −1 4 −1 0  V6   0 
    
0 0 −1 0 0 −1 4 − 1 V7   0 
 0 0 0 0 0 0 − 2 4  V8  U + 2hσ 2 / ε 0 

Sparse matrix.

59
2.3.2 Variational problem of electrostatics

The boundary value problem of electrostatic field

− div(ε gradV
= ) ρ in Ω,
∂V
V= V0 on Γ D , ε = σ on Γ N
∂n
is equivalent to the following variational problem:

Find V (V=V0 on ΓD), so that the functional

1
W (V ) = ∫ εgrad 2VdΩ − ∫ ρVdΩ − ∫ σVdΓ
Ω
2 Ω ΓN

attains a minimum.
60
Physical meaning of the functional

1 1
∫Ω 2 ε Ω = ∫Ω 2 E ⋅ DdΩ : energy of electrical field: We
2
grad Vd

∫ ρVdΩ + ∫ σVdΓ :
Ω ΓN
potential energy of the charges: Wp

= We − W p is the action!
W

The variational problem corresponds to the principle

of least action.

61
 2.3.3 Ritz‘s procedure

Since the variational problem and the boundary value problem

are equivalent, an approximate solution of the variational
problem is simultaneously an approximate solution of the
boundary value problem.
The variational problem can be solved approximately by means
of the Ritz’s procedure.
An approximate solution is sought in the following form:
n
V ≈V (n)
= VD + ∑ V j w j , VD : arbitrary function with=
VD V0 on Γ D ,
j =1
V j , j = 1, 2, ..., n : numerical parameters,
w j , j = 1, 2, ..., n : basis functions with
=
w j 0 on Γ D .
V ( n ) satisfies the Dirichlet boundary conditions for arbitrary V j! 62
The unknown parameters Vj, j = 1, 2, ..., n are determined from
the condition that the approximate solution minimizes the
functional. The necessary conditions are:

∂W (V ( n ) )
= 0, i = 1, 2, ..., n.
∂Vi

These are n equations (the Ritz equations system), allowing

the determination of the n unknowns Vj, j = 1, 2, ..., n.

63
∂W (V ( n ) ) ∂ 1 ∂ ∂
∫ ε Ω − ∫ ρ Ω
= − ∫ σ dΓ
2 (n) (n) (n)
grad V d V d V
∂Vi ∂Vi Ω 2 ∂Vi Ω ∂Vi Γ N

∂V ( n ) ∂V (n)
∂V (n)
= ∫ grad ⋅ εgradV ( n ) dΩ − ∫ ρdΩ − ∫ σdΓ.
Ω
∂Vi Ω
∂Vi ΓN
∂Vi

∂V ( n ) ∂ n
= (VD + ∑ V j w j ) = wi .
∂Vi ∂Vi j =1

The Ritz equations system:

n

∑V ∫ gradw ⋅ εgradw dΩ = − ∫ gradw ⋅ εgradV

j =1
j i j i D dΩ + ∫ wi ρdΩ + ∫ wiσdΓ,
Ω Ω Ω ΓN

Symmetric matrix! i = 1, 2, ..., n.

64
2.3.4 The method of finite elements (Finite Element Method=FEM)
Discretization of the geometry
Triangular elements: simplest possible approach
Unknowns: potential values in nodes
Potential in each element: low order polynomial
finite elements

65
Linear interpolation of the potential function within elements

66
 Shape functions V
Vj V
V
VjNj VkNk 1
ViNi
Vi Vk
1 k k 1 k
+ + =

i j i j i j
Ni Nj Nk
ViNi + VjNj +VkNk Vj

Vi Vk
k
=

i j 67
 V
1 1 in node j
Nj Nj = 
0 in all other nodes

nn
V (n) = ∑V j N j nn: number of nodes
j =1
Vj: nodal potential values
68
 V0 VD
Basis functions:
w j = N j , j = 1, 2, ..., n. ΓD
n+1 n+2 ...
n ... nn

1
2

nn nn n n
V (n)
= ∑V j N j = ∑V N + ∑V N
j j j j = VD + ∑ V j N j .
j =1 j = n +1 j =1 j =1
69
Ritz equations:

∑V ∫ gradN
j =1
j i ⋅ εgradN j dΩ =
Ω

= − ∫ gradN i ⋅ εgradVD dΩ + ∫ N i ρdΩ + ∫ N iσdΓ, i = 1, 2, ..., n.

Ω Ω ΓN

In matrix form: [Aij ]{V j } = {bi }.

Aij = ∫ gradN i ⋅ εgradN j dΩ,

Ω

bi = − ∫ gradN i ⋅ εgradVD dΩ + ∫ N i ρdΩ + ∫ N iσdΓ.

Ω Ω ΓN

70
Ni Nj
Aij = ∫ gradN i ⋅ εgradN j dΩ ≠ 0,
Ω

j
i

Ni Nj
Aij = ∫ gradN i ⋅ εgradN j dΩ = 0.
Ω
j

Sparse matrix.

71
8-node quadrilateral elements

6
7
8 5
4
1 2 3

72
20-node hexahedral elements

73
Shape and basis functions for 20-node hexahedral elements

74
2.4 Integral equations for the scalar potential

The boundary value problem for the scalar potential can also be
represented by various integral equations.

Integral equations for functions a single variable:

y ( x) : unknown function, f ( x) : known function,
K ( x, x′) : Kernel of the integral equation (given),
λ : given konstant.
b
Fredholm integral equation of the first
∫a K ( x , x ′) y ( x ′) d x ′ = f ( x )
kind
b
Fredholm integral equation
y ( x) − λ ∫ K ( x, x′) y ( x′)dx′ = f ( x)
of the second kind
a 75
Boundary value problem for the electric scalar potential in
case of homogeneous medium (ε=ε0):
ρ
divgradV = ∆V = − in Ω, Laplace-Poisson equation,
ε0
V = V0 auf ΓD Dirichlet boundary condition,

∂V σ
= auf ΓN Neumann boundary condition.
∂n ε 0

∂V
If V and were known on the entire boundary,
∂n
one could compute the potential in any point in Ω with the
aid of the integral representation.

76
2.4.1 Fundamentals of potential theory

Green function: solution of the special Laplace-Poisson equation

− ∆G (r, r′) = δ (r − r′)

in the entire three-dimensional space.
δ (r − r′) : Dirac impulse function in the point r′, defined by

∫ w(=
r )δ (r − r′)d Ω w(r′) or ∫ w(r=
′)δ (r − r′)d Ω′ w(r )
ℜ3 ℜ3

77
 ρ (r )
The solution of the Laplace-Poisson equation − ∆V (r ) =
ε0
in infinite empty space (ε=ε0) is:

1 ρ (r′′)dΩ′′
V (r ) = ∫
4πε 0 ℜ3 r − r′′
.

Therefore, the Green function is ( ρ (r ) = ε 0δ (r − r′) ) :

1 δ (r′′ − r′) 1
G (r, r′) =
4π ∫ r − r′′
dΩ′′ =
4π r − r′
ℜ3

This is the potential of a point charge of the magnitude ε0 in the

point r´.
A point charge of the magnitude Q in the point r´ corresponds
to the charge density ρ (r ) = Qδ (r − r′). 78
 Green‘s theorem:
div(φgradψ ) = gradφ ⋅ gradψ + φ∆ψ
− div (ψgradφ ) = gradψ ⋅ gradφ + ψ∆φ
φ∆ψ −ψ∆φ = div(φgradψ ) − div(ψgradφ )

 ∂ψ ∂φ 
∫Ω (φ∆ψ − ψ∆φ )dΩ = ∫Γ φ ∂n − ψ ∂n dΓ
φ ⇐ V (r′),ψ ⇐ G (r, r′), dΩ ⇐ dΩ′, dΓ ⇐ dΓ′

79
  
V (r′)∆G (r, r′) − G (r, r′)∆V (r′) dΩ′ =
∫Ω   
 −δ ( r −r ′ ) 
 ∂G (r, r′) ∂V (r′) 
= ∫ V (r′) − G (r, r′) dΓ′
Γ
∂n′ ∂n′ 

Integral representation:

1 1
V (r ) = −
4π ∫Ω r − r′ ∆V (r′)dΩ′ −
1 ∂ 1 1 1 ∂V (r′)
− ∫
4π Γ
V (r′)
∂n′ r − r′
dΓ′ +
4π ∫Γ r − r′ ∂n′ dΓ′

80
Corollary
Mean value theorem of potential theory :
If V(r) is a solution of von ∆V=0, then the mean value of V,
over the surface of a sphere with an arbitrary radius R, equals
the value of V in the centre of the sphere.
Proof
Let the centre be r, and the surface Γ be the sphere with the
radius R: r − r′ = R
1 1 ∂ 1
V (r ) =− ∫ ′ ′
∆V (r ) d Ω −
4π R Ω  ∫

4π Γ
′
V (r )
∂R R

d Γ′ +
=0
1
−
R2

1 ∂V (r′) 1
+
4π R ∫ ∂n′
= d Γ′
4π R Γ∫
2 
V (r′)d Γ′ cf. method of
finite differences

Γ
 
=  ∫ gradV=
⋅nd Γ′ ∫ divgradVd
= Ω′ 0
81
Γ Ω
Inserting the quantities known from the boundary value problem:

1 ρ (r′) known
V (r ) = ∫ r − r′ dΩ′ −
4πε 0 Ω
1 ∂ 1 1 σ (r′)
− ∫
4π ΓD
′
V0 (r )
∂n′ r − r′
′
dΓ + ∫
4πε 0 ΓN r − r′
dΓ′ −

1 ∂ 1 1 σ (r′)
− ∫
4π ΓN
V (r′)
∂n′ r − r′
dΓ′ + ∫
4πε 0 ΓD r − r′
dΓ′

unknown

Various integral equations can be obtained for the unknown

∂.V
quantities V on Γ N and on Γ D .
∂n 82
 2.4.2 Integral equation for the surface charge density

Simplest case: electrodes (Vi = constant, i =1, 2, ..., m).

V = Vi n

V = V1 ... ΓN = 0, ΓD = ∑ Γi , ρ = 0.
Γi i =1

Γ1 Ωi
Ω Integral representation:
Ω1 ∆V = 0
1 m
∂ 1
V (∞ ) = 0 V (r ) = −
4π
∑
i =1
Vi ∫
∂n′ r − r′
dΓ′ +
Γi
V = Vm
Γm 1 m
σ (r′)
Ωm
+
4πε 0
∑ ∫ r − r′ dΓ′.
i =1 Γi

∂ 1 1 1
∫Γ ∂n′ r − r′ dΓ = Γ∫ grad r − r′ ⋅ n dΓ = Ω∫ ∆ r − r′ dΩ′ =
′ ′ ′ ′ ′
i i i

= 0 falls r ∉ Ωi . 83
Integral equation for the surface charge density on the
electrodes:

1 m
σ (r′)
4πε 0
∑ ∫ r −=
i =1 Γi r′
d Γ′ Vi , for r=
∈ Γi , i 1, 2, ..., m.

1 1
In the 2D case, the Green function is ln .
2π r − r′
Let the contours of the electrodes be the curves Ci.
Integral equation for the line charge density on the
electrodes:
m
1 1
∑ ∫

2πε 0 i =1 Ci
τ (r′) ln =
r − r′
d Γ′ Vi , for r ∈
= Ci , i 1, 2, ..., m.

Fredholm integral equations of the first kind. 84

2.4.3 Method of boundary elements (Boundary Element Method = BEM)
Numerical procedure to solve integral equations
Discretization of the electrodes: 3D problems - surfaces
2D problems - curves
Unknowns: surface charge densities in the elements
Boundary elements

85
Simplest assumption: σ is constant in each element.
Let the number of elements be n, the unknowns are
σj (j = 1, 2, ..., n).
In each element, one test point Pi (i = 1, 2, ..., n) is selected in
which the integral equation is required to be satisfied.
... Pn P1 P2 ...

86
Linear equation system for the unknown surface charge densities
(Γj: j-th element, Vi: potential in Pi):
n
1 1
∑ σj∫
4πε 0 j =1 Γ j rPi − r′
dΓ′ = Vi , i = 1, 2, ..., n.

In the 2D case, the unknowns are the values of the line

charge density (Cj: j-th line segment, Vi: potential in Pi):
n
1 1
∑ τ j ∫ ln
2πε 0 j =1 C j rPi − r′
ds′ = Vi , i = 1, 2, ..., n.

Non-symmetric full matrix.

Having solved the equation system, the potential can be

computed in any point by evaluating the integral representation.
87
 2.5 Boundary value problems for the vector potential

Magnetostatic field:
divB =0 ⇒ B =curlA, A: magnetic vector potential

Static current field:

divJ = 0 ⇒ J = curlT, T: current vector potential

Differential equations:
1
curlH =
J ⇒ curl ( curlA) =
J,
µ
1
curlE =
0 ⇒ curl ( curlT) =
0.
γ
88
2.5.1 Planar 2D problems
Magnetostatic field:
∂
= 0 : J = J ( x, y )e z , B = Bx ( x, y )e x + B y ( x, y )e y .
∂z
A = A( x, y )e z
Field lines of B
ex e y ez
∂ ∂ ∂A ∂A
J = =
B curl A =0 ex − e y .
∂x ∂y ∂y ∂x
0 0 A

89
Differential equation for the single component vector potential in
planar 2D case:

ex ey ez
1 ∂ ∂
=
curl[ curl ( Ae z )] = 0
µ ∂x ∂y
1 ∂A 1 ∂A
− 0
µ ∂y µ ∂x
 ∂  1 ∂A  ∂  1 ∂A   1
=
−   +    ez =
−e z div( gradA).
 ∂x  µ ∂x  ∂y  µ ∂y   µ
1
− div ( gradA) = J , generalized Laplace-Poisson equation.
µ

90
Magnetic flux by means of A:
B

Φ
= ∫ B ⋅ nd Γ= ∫ curlA ⋅ nd Γ= ∫ A ⋅ dr.
Γ Γ C
n dΓ
Γ A
Planar 2D case:
dr
ΓPQ: surface of unit length C
through the points P and Q ez
z=constant
ΓPQ
Φ PQ = ∫ A ⋅ dr = A( P) − A(Q)
C PQ
A( P )e zP Q
A(Q)e z
CPQ 1
y
91
x
Flux lines (magnetic field lines) are parallel to B.

Q For any two points P and Q on a flux line

B one has ΦPQ = A(P) - A(Q) = 0.
P A( P=
) A(Q) ⇒ =
A constant along flux lines!

∂A ∂A dy B y
dA( P ) = dx + dy = − B y dx + Bx dy = 0 ⇒ = .
∂x ∂y dx Bx

Flux lines: lines of constant A(x, y).

If flux lines are drawn so that the difference of the vector

potential between any two neighboring flux lines is
constant, then the density of the lines is proportional to
the magnitude of B. 92
 Current by means of T:
J

I= ∫ J ⋅ nd Γ= ∫ curlT ⋅ nd Γ= ∫ T ⋅ dr.
Γ Γ C
n dΓ
Γ T
Planar 2D case:
dr
ΓPQ: surface of unit length C
through the points P and Q ez
z=constant
ΓPQ
I PQ = ∫ T ⋅ dr = T ( P) − T (Q)
C PQ
T ( P )e zP Q
T (Q)e z
CPQ 1
Current lines: y
Lines of constant T(x,y).
93
x
Static current field:

∂
= 0 : J = J x ( x, y )e x + J y ( x, y )e y .
∂z

T = T ( x, y )e z

ex e y ez
∂ ∂ ∂T ∂T
= =
J curl T =0 ex − ey.
∂x ∂y ∂y ∂x
0 0 T

1
− div ( gradT ) = 0, generalized Laplace equation.
γ 94
Boundary conditions:
=
Dirichlet boundary condition: A A0 (known) on Γ D ,
= T T0 (known) on Γ D .
This means the prescription of Bn or Jn:
e.g. for the magnetic field:
Bn =n ⋅ curlA =n ⋅ curl ( Ae z ) =n ⋅ ( gradA × e z ) =gradA ⋅ (e z × n) =

∂A
=
t ⋅ gradA = : tangential derivative of A!
∂t
ez
In most cases A0=constant or T0=constant:
n Then, the section of ΓD with a plane z=constant is
ez × n = t a flux line or a current line. The differences in the
ΓD values of A0 or T0 yield the flux or current per unit
length between the lines.
95
 1 ∂A
= α (known) on Γ N ,
Neumann boundary condition:
µ ∂n
1 ∂T
= e (known) on Γ N .
γ ∂n
This means the prescription of Ht or Et :
e.g. for the magnetic field:
1 1
H t = (e z × n) ⋅ curl ( Ae z ) = (e z × n) ⋅ ( gradA × e z ) =
µ µ
1 1 1 ∂A
= e z × (e z × n) ⋅ gradA = −n ⋅ gradA = − .
µ µ µ ∂n
On interfaces to highly permeable regions or electrodes,
the Neumann boundary condition is
1 ∂A 1 ∂T
homogeneous:= 0= or 0.
µ ∂n γ ∂n
96
Boundary value problem for the single component vector
potential functions in the planar 2D case:

1
magnetostatic − div( =
gradA) J in Ω,
field: µ
1 ∂A
A=
A0 on Γ D , α on Γ N .
=
µ ∂n

Static current 1
− div( gradT
= ) 0 in Ω,
field: γ
1 ∂T
T=
T0 on Γ D , =
0 on Γ N .
γ ∂n

Similar boundary value problem to the ones for the scalar potential
functions.
97
Duality between the boundary conditions for the scalar potential
and the single component vector potential in the planar 2D case

Flux line Magnetic wall:

ψ = constant,
1 ∂A
= 0.
µ ∂n
Flux line:
∂ψ
µ = 0,= A constant.
∂n

Magnetic wall

98
2.5.2 Axisymmetric 2D problems
Magnetostatic field:
∂
= 0 : J = J (r , z )eφ , B = Br (r , z )e r + Bz (r , z )e z .
∂φ
A = A(r , z )eφ
1 1
e r eφ ez
r r
∂ ∂
= =
B curl A 0 =
∂r ∂z
0 rA 0

∂A 1 ∂ (rA)
=− er + ez .
∂z r ∂r 99
Differential equation for the single component vector potential in the
axisymmetric case:
1 1
er eφ ez
r r
1 ∂ ∂
curl[ curl ( Aeφ )] = 0
µ ∂r ∂z
1 ∂A 1 ∂ (rA)
− 0
µ ∂z µ r ∂r
 ∂  1 ∂( rA )  ∂  1 ∂A   1
= −   +    eϕ ≠ −eϕ div( gradA ).
 ∂r  µ r ∂r  ∂z  µ ∂z   µ
 1 1 ∂  1 ∂A  ∂  1 ∂A  
=
 div( gradA )  r  +  
 µ r ∂r  µ ∂r  ∂z  µ ∂z 
 ∂  1 ∂( rA )  ∂  1 ∂A  
−   +   =J
 ∂r  µ r ∂r  ∂z  µ ∂z   100
 Flux in axisymmetric 2D case:
z
A(Q)eφ
ΓPQ: Conical surface through Q
the points P and Q rQ φ = constant
ΓPQ
A(P )eφ P
CPQ φ = constant
rP

Φ PQ = ∫ A ⋅ dr = 2π [r A( P) − r
C PQ
P Q A(Q)].

Flux lines: Lines of constant rA(r, z).

Current field: Lines of constant rT(r, z) are the

current lines.
101
 2.5.3 3D problems

The vector potential functions are not unique:

= =
B curl A curl ( A + gradu ) u is an arbitrary scalar function,

= =
J curl T curl (T + gradu ) u is an arbitrary scalar function,

Differential equations: Boundary conditions:

1
curlH =
J ⇒ curl ( curlA) =
J, Prescription of Bn or of Ht .
µ
1
curlE =
0 ⇒ curl ( curlT) =
0, Prescription of Jn or of Et .
γ

102
In case of 3D problems, the boundary conditions are
specified for the normal and tangential components of the
field quantities or vector potentials.
. A
nAn
A = nAn + tAt . .
n . tA Γ
t
A×n
nAn = n( A ⋅ n)
tAt = n × (A × n ) = A(n ⋅ n) − n( A ⋅ n) = A − nAn
 
1 An

Instead of tAt we use A × n.

103
Specification of Bn or of Jn with the aid of the vector potentials:

div( A × n) = n ⋅ curlA − A ⋅ curl

 n = n ⋅ B = Bn , div(T × n) = J n .
0

 1  ∂ ∂ ∂ 
 divv ( x1 , x2 , x3 ) =  (v1h2 h3 ) + (v2 h1h3 ) + (v3h1h2 ) 
 h1h2 h3  ∂x1 ∂x2 ∂x3 

A × n and T × n have no normal component

⇓
When building the divergence, no differentiation in the
normal direction occurs.

By specifying A × n or T× n , Bn or Jn are determined:

Dirichlet boundary condition.
104
Specification of Ht or of Et with the aid of the vector potentials:

H × n or E × n are prescribed:

1 1
=
H ×n curlA × n, E=
×n curlT × n : Neumann boundary
µ γ condition.

105
Boundary value problems for the vector potential functions in 3D case:

Static magnetic curl ( 1 curl=

A) J in Ω,
field: µ
1
A×n = a on Γ D , rotA × n
= α on Γ N .
µ

1
Static current curl ( curl=
T) 0 in Ω,
field: γ
1
T × n= τ on Γ D , curlT × n= e on Γ N .
γ

The solution of the boundary value problems is not

unique.
106
Special case: vector potential due to a given current
=
density in infinite free space ( µ µ 0 and Γ → ∞) :

curlcurlA = µ0 J ,
=
A (∞ ) 0 (⇒ A=
(∞) × n 0 or curlA=
(∞) × n 0).

Only B=curlA is defined uniquely, but not A.

A becomes unique if divA is additionally defined: gauging.

The choice divA = 0 is the Coulomb gauge.

107
 The gauge makes A unique:

=
divA div( A + gradu ) ⇒= ∆u 0, 
 ⇒ gradu =
0.
A(∞)= A (∞) + gradu (∞) ⇒ gradu (∞)= 0, 

Boundary value problem for A:

curlcurlA = µ 0 J , 
 curlcurlA − graddivA = −∆A = µ 0 J ,
=
divA 0, ⇒
 A(∞) =0.
A(∞) =0. 
⇓
Vector Laplace-Poisson differential equation.

108
The Coulomb gauge follows from the vector Laplace-
Poisson differential equation:

curlcurlA − graddivA = −∆A = µ 0 J , 

⇒
A(∞) =0. 
div(curlcurlA − graddivA) =
 
µ 0 divJ 

∆ ( divA ) 0  ⇒ divA =
0.
divA(∞) =0, 

−∆A µ 0 J ,=
Solution of= A (∞ ) 0 :

µ 0 J (r′)d Ω′
A(r ) = ∫
4π Ω r − r′
.

109
Biot-Savart‘s law:

µ0 J (r′)d Ω′
= =
B(r ) curl A ∫
4π Ω
curl =
r − r′
µ0 1 µ0 J (r′) × (r − r′)
∫
4π Ω
( grad =
r − r′
) × J (r′)d Ω′ ∫
4π Ω r − r′
3
d Ω′,

1 J (r′) × er′→r
H (r ) = ∫
4π Ω r − r′ 2
dΩ′.

110
3. Quasi-static fields
 ∂D 
Maxwell’s equations  J >> :
 ∂t 

In conducting media (Ωl): In non-conducting media (Ωi):

curlH = J ,
∂B curlH = J ,
curlE = − ,
∂t divB = 0,
divB = 0,
B = µH, J = γE (E e = 0). B = µH.

J is unknown: J is known:
time dependent time dependent
quasi-static field magnetostatic field 111
Example:

Ωi

J (t )
Ωl

112
Boundary and interface conditions:

ΓB : B ⋅n = −b
n
Ωi

ΓE : E × n = 0 non-conducting region

Γli : H × n, B ⋅ n continuous ΓHi : H × n = α

n
J
Ωl
conducting region
ΓHl : H × n = 0 n

113
 Summary
Differential equations in Differential equations in Ωi
Ωl (eddy current region): (eddy current free region):
curlH l = J l curlH i = J i
∂ Bl divBi = 0
curlEl = −
∂t =Bi µ=
H i , H i ν Bi
divBl = 0
=Bl µ H=
l , Hl ν=
B l , J l γ El

boundary conditions: interface conditions on Γli:

H l × n = 0 on ΓHl ,
El × n = 0 on ΓE , Hl × nl + Hi × ni = 0
H i × n = K on ΓHi , Bl ⋅ nl + Bi ⋅ ni = 0
B i ⋅n = −b on ΓB .
initial conditions at t=0: B l = B l 0 in Ω l , B i = B i 0 in Ω i 114
Complex notation for time harmonic quantities:
=
Time function : Bx (r, t ) Bˆ x (r ) cos(ω t + ϕ x (r ))
= By (r, t ) Bˆ y (r ) cos(ω t + ϕ y (r ))
= Bz (r, t ) Bˆ z (r ) cos(ω t + ϕ z (r ))
specially for linear polarization: B(r, t ) = Bˆ (r ) cos(ωt + ϕ (r ))

Complex amplitude: B(r ) = Bˆ (r )e jϕ ( r )

Time derivative:
∂
in time domain → multiplication by jω in frequency domain
∂t
Maxwell’s equations for complex amplitudes
(quasi-static case):
curlH = − jω B , divB =
J , curlE = 0.
115
Poynting’s theorem for complex amplitudes in quasi-static
case:

− div ( E × H* ) =E ⋅ J * + jω B ⋅ H*
1 1 1 1 1
E ⋅ curlH* − H* ⋅ curlE =
2 2 2 2 2

1
∫ E ⋅ J *
dΩ + j ω
1
∫ B ⋅ H *
d Ω = −
1
∫ (E × H *
)⋅ ndΓ = S
2Ω 2Ω 2Γ

S: complex power flowing into the region Ω through the

boundary Γ.

116
Proof:
Time function of Poynting’s vector:
ˆ cos(ωt + ϕ E ) × H
S(t ) = E(t ) × H (t ) = E ˆ cos(ωt + ϕ H ) =
1ˆ ˆ
= E × H[cos(ϕ E − ϕ H ) + cos(2ωt + ϕ E + ϕ H )] =
2
1ˆ ˆ
= E × H cos(ϕ E − ϕ H ) [1 + cos 2(ω t + ϕ E ) ] +
2
  
Effective part
1ˆ ˆ
+ E × H sin(ϕ E − ϕ H ) [sin 2(ω t + ϕ E ) ]
2
 
Reactive part

1ˆ ˆ
Effective power: P = − ∫ E × H cos(ϕ E − ϕ H )⋅ ndΓ
Γ
2
1ˆ ˆ
Reactive power: Q = − ∫Γ 2 E × H sin(ϕ E − ϕ H )⋅ ndΓ 117
 1ˆ ˆ
S = P + jQ = − ∫ E × H [cos(ϕ E − ϕ H ) + j sin(ϕ E − ϕ H )]⋅ ndΓ =
Γ
2

1 ˆ ˆ j (ϕ E −ϕ H ) 1 ˆ jϕ E ˆ − jϕ H
= −∫ E ×He ⋅ ndΓ = − ∫ E e × He ⋅ ndΓ =
Γ
2 Γ
2

=−
1
∫ (E × H *
)⋅ ndΓ. q. e. d.
2Γ
1
Complex Poynting’s vector: S = E × H *

2
2
1 1 J
Effective power: P = ∫ E ⋅ J *dΩ = ∫ dΩ, (E e = 0),
2Ω 2Ω γ
1 1
Reactive power: Q = ω ∫ B ⋅ H dΩ = ω ∫ µ H dΩ.
* 2

2Ω 2 Ω
118
 3.1 Some analytical solutions of the boundary value problem for
the magnetic vector potential

In conducting region (Ωl): divB =0 ⇒ B =curlA,

∂B ∂curlA ∂A ∂A
curlE + =curlE + =curl (E + ) =⇒
0 E=− gradV − .
∂t ∂t ∂t ∂t

Differential equations:
1 ∂A
curlH − J = 0 ⇒ curl ( curlA) + γ + γ gradV = 0,
µ ∂t

 divJ = 0 ⇒ − div(γgradV ) − div (γ ∂A ) = 0.

 
 ∂t 
boundary conditions: A(∞=
) 0, V (∞=
) 0.
119
Special case: µ = constant, γ = constant, time harmonic case.
Coulomb gauge: divA=0

− div(γ gradV ) − jω div(γ A=

) 0 ⇒ −∆V= jω divA= 0
 
⇓

∆V = 0 
 ⇒V =
0
V (∞) =0 
Differential equation in Ωl:

− ∆A + jωµγA = 0 vector diffusion equation.

Planar 2D problems:

− ∆A( x, y ) + jωµγA( x, y ) = 0 scalar diffusion equation.

120
 3.1.1 Current flow in an infinite conducting half space

∂ ∂
z = 0, = 0 :
∂z ∂x
J = J ( y )e z
x A = A( y )e z ,
y B = B ( y )e x E = − jωA( y )e z = E ( y )e z ,
γ ,µ J = γE = − jωγA( y )e z = J ( y )e z ,
ex e y ez
∂ dA( y )
B =
curl A 0 =0 = e x B ( y )e x ,
∂y dy
0 0 A 1 1 dA( y )
H= B= e x = H ( y )e x .
µ µ dy 121
 d 2 A( y )
Diffusion equation: − 2
+ jωµγA( y ) = 0.
dy

1+ j 1+ j
jωµγ = p , p =
2
jωµγ = = , δ: penetration depth.
2 δ
ωµγ
d 2 A( y )
2
= p 2
A( y )
dy

A( y ) = A1e − py + A2 e py
y y
− −j
lim A( y ) < ∞ ⇒ A2 = 0 : A( y ) = A1e − py = A1e δ e δ
y →∞

122
 Field quantities:

E ( y ) = − jωA( y ) = − jωA1e − py ,

J ( y ) = − jωγA( y ) = − jωγA1e − py ,

dA( y )
B( y ) = = − pA1e − py ,
dy
1 dA( y ) p
H ( y) = = − A1e − py .
µ dy µ

Determination of the constant A1: the current through a

conductor of width b is assumed to be given.

123
 ∫ J ⋅ ndΓ = ∫ H ⋅ dr = H ( y = 0)b = I
z
Γ0 C0

l pb µI
− A1 = I ⇒ A1 = −
I µ pb
Γ0 C0 jωµ − jpy p − j δ −δ
y y
E ( y) = Ie = Ie e ,
b pb γb
x p − j δ −δ
y y
J ( y ) = γE ( y ) = I e e ,
y b
µ −j
y
−
y
I − j δ −δ
y y
Skin effect: B( y ) = I e e , H ( y ) = e e .
δ δ

J ( y) b b
Magnitude of the I −δy
J ( y) = 2 e
current density bδ
decays exponentially
δ
124
y
Impedance of a conductor of width b and length l:

S = I Z = − ∫ (E × H * )⋅ ndΓ
1 2 1
2 2Γ
E × H * = E ( y )e z × H * ( y )e x = E ( y ) H * ( y )e y

n=
−e y for y =
0, otherwise n ey.
*
1 2 1 1 p I 1 2 pl
I Z = E ( y = 0) H ( y = 0)bl =
*
I bl = I .
2 2 2 γb b 2 γb

l
Z = R + jX = (1 + j ) , l
γδb b D. C.
l l resistance
R= , X= . I δ
γδb γδb 125
 3.1.2 Current flow in an infinite conducting plate

I h2

z x
h2
b2 b2
E = − jωA( y )e z = E ( y )e z , J = γE = − jωγA( y )e z = J ( y )e z ,
dA( y ) 1 dA( y )
B= e x = B ( y )e x , H= e x = H ( y )e x .
dy µ dy
Diffusion equation:
d 2 A( y ) 1+ j 1+ j
= p 2
A( y ), p= jωµγ = = .
dy 2 2 δ
ωµγ 126
Solution of the diffusion equation:

A( y ) = A1e − py + A2 e py = C1 cosh( py ) + C2 sinh( py ).

The current density J ( y ) = − jωγA( y ) has to be an even function

( J(y) = J(-y) ): C2 = 0.

A( y ) = C1 cosh( py ).

Field quantities: E ( y ) = − jωA( y ) = − jωC1 cosh( py )

J ( y ) = − jωγA( y ) = − jωγC1 cosh( py )

dA( y )
B( y ) = = pC1 sinh( py )
dy
1 dA( y ) p
H ( y) = = C1 sinh( py )
µ dy µ 127
Determination of the constant C1: Current through a
conductor of width b is assumed to be given.

y
Γ0
y=h 2 I
y = −h 2 z x
C0
b

∫ J ⋅ ndΓ = ∫ H ⋅ dr = − H ( y = h 2)b + H ( y = − h 2)b =

Γ0 C0

2 pb ph
= −2 H ( y = h 2)b = − C1 sinh( ) = I ,
µ 2
µI
C1 = − .
ph
2 pb sinh( ) 128
2
 jωµ I pI
E ( y) = cosh( py ) = cosh( py ),
ph ph
2 pb sinh( ) 2γb sinh( )
2 2

pI
J ( y ) = γE ( y ) = cosh( py ),
ph
2b sinh( )
2
µI
B( y ) = − sinh( py ),
ph
2b sinh( )
2
I
H ( y) = − sinh( py ).
ph
2b sinh( )
2
129
 Impedance of a conductor of width b and length l:
y Γ
S = I Z = − ∫ (E × H * )⋅ ndΓ
1 2 1
2 2Γ l
b
E × H * = E ( y )e z × H * ( y )e x = E ( y ) H * ( y )e y I
z h
x
n=
e y for y ==
h 2, n −e y for y =
− h 2, otherwise n ey.

1 2 1 h * h 1 h * h
I Z = − E ( y = ) H ( y = )bl + E ( y = − ) H ( y = − )bl =
2 2 2 2 2 2 2
ph
p I cosh( ) *
h * h 2 I
= − E ( y = ) H ( y = )bl = bl.
2 2 ph
2γb sinh( ) 2b
2 130
 ph
pl cosh( )
Z= 2 , ph 1 + j h 1 + j ωµγ
= = h .
ph
2γb sinh( ) 2 2 δ 2 2
2
ph ph ph ph
Low frequency: << 1 ⇒ cosh( ) ≈ 1, sinh( ) ≈ .
2 2 2 2

l
Z≈ : like D. C. in the entire height h.
γbh
ph ph ph
High frequency: >> 1 ⇒ cosh( ) ≈ sinh( ).
2 2 2
pl l
Z≈ = (1 + j ) : D. C. resistance of two layers each of
2γb 2γδb
thickness δ, reactance same as
resistance. 131
4. Electromagnetic waves
The full set of Maxwell’s equations:

∂D ∂B
curlH =J+ , curlE =
− ,
∂t ∂t
=
divB 0,=divD ρ , =D ε E,=
B µ H,=
J γ (E + E e )

describes electromagnetic waves.

in vacuum:
∂D ∂B
− The time varying electric and
∂t ∂t
magnetic fields mutually sustain
each other:
H (t ) E(t ) electromagnetic field

132
 4.1 Planar waves

Maxwell’s equations in vacuum (µ = µ0, ε = ε0), in

absence of charges and currents (ρ = 0, J = 0):

∂E ∂H
curlH = ε 0 , curlE = − µ 0 ,
∂t ∂t
divH = 0, divE = 0.
∂ ∂2H
curlcurlH = graddivH − ∆H = ε 0 curlE = −ε 0 µ 0 2 ,
  ∂t ∂t
=0

∂2H
∆H =ε 0 µ0 2 ,
∂t 
 vector 3D wave equation
∂E 
2
Similarly: ∆E =ε 0 µ0 2 .
∂t 133
 ∂ ∂
Assumption:= 0,= 0. The electromagnetic field is
∂x ∂y constant in any plane z = constant:

Planar waves:
∂2H ∂2H ∂ 2E ∂ 2E
= ε 0 µ0 2 , = ε 0 µ0 2 :
∂z 2
∂t ∂z 2
∂t
Vector 1D wave equation.

All components Ex, Ey, Ez, Hx, Hy, Hz of the electromagnetic field
satisfy the scalar 1D wave equation:

∂2 f ∂2 f z 1
= ε 0 µ0 2 . Solution: f (t = ), v = c.
∂z 2
∂t v µ 0ε 0

134
Waves propagating with light velocity:

z z z
Ex ( z , t ) = Ex (t  ), E z
y( , ) t = E y( t  ), E z ( z , t ) = E z (t  ),
c c c
z z z
H x ( z , t ) = H x (t  ), H y ( z , t ) = H y (t  ), H z ( z , t ) = H z (t  ).
c c c
Relationship between E and H:

ex e y ez
∂ ∂ ∂ ∂E
curlH = ε0 ,
∂x ∂y ∂z ∂t
Hx Hy Hz

∂ ∂ ∂ 1∂
Planar waves: = 0,= 0,=  .
∂x ∂y ∂z c ∂t
135
 ex ey ez
ex e y ez
1∂ 1∂ 1∂
=
curlH 0 0 =  0 =
0 1  (e z × H ).
c ∂t c ∂t c ∂t
Hx Hy Hz
Hx Hy Hz

1∂ ∂E 1 ε0
 ) ε0
(e z × H= ) ε 0 E; (e z × H=
⇒  (e z × H= )  E.
c ∂t ∂t c µ0

µ0
Similarly, from Faraday’s law: (e z × E) =
± H.
ε0

Sign above: propagation in the positive z-direction,

Sign below: propagation in the negative z-direction.

136
 z z
E( z ,=
t ) E(t − ) H ( z= , t ) H (t + )
c c
S direction of propagation S direction of propagation
ez ez
z z
H ( z=
, t ) H (t − ) E( z ,=
t ) E(t + )
c c

The direction of Poynting’s vector coincides with the

direction of propagation.

µ0 µ0 2 1
S =×
E H= (e z × H ) × H =±e z H =±e z µ0 H ,
2

ε0 ε0 µ 0ε 0
ε0 ε0 2 1
S =E × H =E × [± (e z × E)] =±e z E =±e z ε0 E ,
2

µ0 µ0 µ 0ε 0
1
±e z c (ε 0 E + µ 0 H ).
S=
2 2

2 137
More general material properties: µ , ε , γ = konstant.

Maxwell’s equations in absence of charges (ρ = 0):

∂E ∂B
γE+ε
curlH = , curlE =
− ,
∂t ∂t
divH 0,= divE 0.
∂ ∂H ∂2H
curlcurlH = graddivH − ∆H = γ curlE + ε curlE = −γµ − εµ 2 ,
  ∂t ∂t ∂t
=0

∂H ∂2H
∆H − γµ − εµ 2 =0,
∂t ∂t
∂E ∂ 2E
Similarly: ∆E − γµ − εµ 2 = 0.
∂t ∂t
138
 ∂ ∂ ∂2
For planar waves: = 0, = 0 ⇒ ∆ = .
∂x ∂y ∂z 2

∂ 2E ∂ 2E ∂E ∂2H ∂2H ∂H
= µε 2 + µγ ,= µε 2 + µγ .
∂z 2
∂t ∂t ∂z 2
∂t ∂t

Total analogy with the transmission line equations:

∂ 2u ∂ 2u ∂u
= LC + ( RC + LG ) + RGu ,
∂z 2
∂t 2
∂t
∂ 2i ∂ 2i ∂i
= LC + ( RC + LG ) + RGi.
∂z 2
∂t 2
∂t

u ⇔ E, i ⇔ H, R ⇔ 0, L ⇔ µ , G ⇔ γ , C ⇔ ε .

139
Time harmonic case, complex notation

E( z ), H ( z ) : complex amplitudes.

Solution (due to analogy):

+ −
E E
E( z ) E+ e − pz + E− e pz , =
= H( z ) e − pz − e pz .
Z0 Z0
Propagation coefficient: p =jωµ (γ + jωε ) =
α + jβ ,

jωµ
Wave impedance: Z0 = .
γ + jωε
Attenuated waves propagating in the positive and
negative z-direction.
140
Lossless medium: γ = 0.

p= ( jωµ )( jωε )= jω µε ⇒ α= 0, β= ω µε .

jωµ µ
=Z0 = .
jωε ε

ω 1 1 1 c
=
v = = = ≤ c.
β µε µrε r µ 0ε 0 µrε r
c
Optics: v = , n: refraction index
n
=
Maxwell’s relationship: n ε r ; n2 ε r ,
=
since for optically transparent media µ r = 1.
141
4.2 Electromagnetic waves in homogeneous, infinite space

Assumptions:
• current density J and charge density ρ are known everywhere at
any time instant: J (r, t ) and ρ (r,t) are given,
• material properties µ and ε are constant everywhere, e. g. µ = µ0,
ε = ε0 ,
• lossless medium: γ = 0.
e. g. antenna: homogeneous medium
(e. g. vacuum or air: µ 0 , ε 0 )

J (r, t ) electromagnetic field: E(r,t), H(r,t)

142
4.2.1. Solution of Maxwell’s equations with retarded potentials

Maxwell’s equations:

∂D ∂B
curlH= J + , curlE = − ,
∂t ∂t
divB = 0, divD = ρ ,=B µ=
0 H, D ε 0 E.

1
=
Potentials: =
B curl A, H curlA,
µ0
∂A ∂A
E=
− −ε 0
− gradV , D = − ε 0 gradV .
∂t ∂t

143
 ∂2A ∂V
curlcurl= A µ 0 J − µ 0ε 0 2 − µ 0ε 0 grad
A graddivA − ∆= .
∂t ∂t
The divergence of A can be freely chosen:

∂V
divA = − µ 0ε 0 : Lorenz gauge.
∂t
∂2A
−∆A + µ 0ε 0 2 = µ 0 J.
∂t
∂A ∂divA ρ
div(− gradV − ) = −∆V − = . Non-homogeneous 3D
∂t ∂t ε0 wave equations
Using the Lorenz gauge:
∂ 2V ρ
−∆V + µ 0ε 0 2 = .
∂t ε0
144
 ∂
In the static case ( = 0) these equations reduce to the
∂t
ρ
Laplace-Poisson equations −∆A = µ 0 J , −∆V =
ε0
whose solutions are known to be

µ 0 J (r′)d Ω′ 1 ρ (r′)d Ω′
A(r ) = ∫
4π Ω r − r′
, V (r ) = ∫
4πε 0 Ω r − r′
.

In regions with vanishing current density and charge density,

one obtains the wave equations

∂2A ∂ 2V
−∆A + µ 0ε 0 2 = 0, −∆V + µ 0ε 0 2 = 0.
∂t ∂t

145
The solutions of the non-homogeneous wave equations in
infinite free space are

r − r′
′ ′
µ 0 J (r , t − c )d Ω
A(r, t ) = ∫
4π Ω r − r′
,

r − r′
1
ρ (r′, t − )d Ω′ 1
∫
V (r, t ) = c , (c = ).
4πε 0 Ω r − r′ µ 0ε 0

A(r, t ) and V (r, t ) are the retarded potentials.

146
Time harmonic case:
A(r ), V (r ), J (r′) and ρ (r′) are complex amplitudes.
r − r′ r − r′
In time domain: J (r′, t − = ) J (r′) cos[ω (t −
ˆ ) + ϕ (r′)].
c c
r −r ′
− jω − jk0 r −r ′
In frequency domain: Jˆ (r′)e jϕ ( r′) e c ′
= J (r )e ,
ω
k=
0 = ω µ 0ε 0 : wave number (phase factor).
c
′ ′
µ 0 J (r′)e − jk r −r d Ω′
0
1 ρ (r′)e − jk r −r d Ω′ 0

A(r ) = ∫
4π Ω r − r′
, V (r ) = ∫
4πε 0 Ω r − r′
.

147
Using the Lorenz gauge, V can be eliminated:

∂V
divA = − µ 0ε 0 is in the frequency domain divA = − jωµ 0ε 0V .
∂t
1
V= − divA.
jωµ 0ε 0

1
− jω A − gradV =
E= − jω A + graddivA =
jωµ 0ε 0
1
(ω 2 µ 0ε 0 A + graddivA).
jωµ 0ε 0
1
B ==
curlA, E (k02 A + graddivA).
jωµ 0ε 0

148
4.2.2. Hertz dipole

2π
l i (t ) = ˆ
I cos(ω t ) lλ = , d  l.
k0 Q(t ) = Qˆ sin(ω t )
d i (t )
Equivalent to a dipole: l
−Q(t )

Iˆ can be considered to be the complex amplitude of the current.

Iˆ = jω Qˆ ( r ,θ , φ )
z
θ
r y
r
Spherical coordinate system:
Iˆ φ
x 149
 − jk0 r −r ′
′
µ 0 J (r )e d Ω′
A(r ) = ∫
4π Ω r − r′
= , r ′ 0, J ( 0=) d Ω′ ˆ e=
Il z, r r.

z ez er =e z e r cos θ − eθ sin θ .
θ
eθ

No integration is necessary, since the integrand is constant:

µ 0 Ilˆ e − jk r
0

A ( r ,θ , φ ) (e r cos θ − eθ sin θ ).
4π r

150
 Magnetic field:
1 1 1
er eθ eφ
r sin θ
2
r sin θ r
1 1 ∂ ∂ 1 ∂ ∂Ar
H = rotA = 0 eφ [ (rAθ ) − ].
µ0 µ0 ∂r ∂θ µ 0 r ∂r ∂θ
Ar rAθ 0

Ilˆ − jk0 r 1 k0
H e ( 2 + j ) sin θ eφ .
4π r r

Ilˆ 1 e − jk0 r
= Hθ =
H r 0,= 0, H φ jk0 (1 + ) sin θ .
4π jk0 r r

151
 1
=
Electric field: E (k02 A + graddivA).
jωµ 0ε 0
1 ∂ (r 2 Ar ) 1 ∂ (sin θ Aθ ) µ 0 Ilˆ − jk0 r 1 jk0
=
divA + = e (− 2 − ) cos θ ,
r 2
∂r r sin θ ∂θ 4π r r
∂divA µ 0 Ilˆ − jk0 r 2 jk0 k02
( graddivA )r = = e ( 3 + 2 2 − ) cos θ ,
∂r 4π r r r
1 ∂divA µ 0 Ilˆ − jk0 r 1 jk0
( graddiv
= A )θ = e ( 3 + 2 ) sin θ ,
r ∂θ 4π r r
Ilˆ µ0 1 e − jk r 0

Er (1 + ) 2 cos θ ,
2π ε0 jk0 r r
Ilˆ µ0 1 1 e − jk r
0

=Eθ jk0 [1 + + ] sin θ , Eφ = 0.

4π ε0 jk0 r ( jk0 r ) 2
r
152
 r
Near field: k0 r = 2π  1.
λ
Ilˆ 1 e − jk0 r Ilˆ e − jk0 r
Hφ = jk0 (1 + ) sin θ ≈ sin θ .
4π jk0 r r 4π r 2
 eφ

1 J (r′) × er′→r Ilˆ e z × e r

=
Biot-Savart: H (r ) ∫
4π Ω r − r′
=
2
d Ω′ ⇒ H (r )
4π r 2
.

Ilˆ µ 0 1 e − jk0 r ˆ e − jk0 r

Ql
Er = (1 + ) 2 cos θ ≈ cos θ ,
2π ε 0 jk0 r r 2πε 0 r 3

Ilˆ µ0 1 1 e − jk r
0 ˆ e − jk r
Ql 0

=Eθ jk0 [1 + + ] sin θ ≈ sin θ .

4π ε0 jk0 r ( jk0 r ) 2
r 4πε 0 r 3

p cos θ p sin θ
Static dipole field: Er = , Eθ = .
2πε 0 r 3
4πε 0 r 3

153
 r
Far field: k0 r = 2π  1.
λ
Ilˆ 1 e − jk0 r Ilˆ e − jk0 r jIˆ l e − jk0 r
Hφ = jk0 (1 + ) sin θ ≈ jβ sin θ = sin θ .
4π jk0 r r 4π r 2 λ r

Ilˆ µ 0 1 e − jk0 r
Er = (1 + ) 2 cos θ ≈ 0,
2π ε 0 jk0 r r

Ilˆ µ0 1 1 e − jk r
0
Ilˆ µ0 e − jk r
0

Eθ jk0 [1 + + ] sin θ ≈ jk0 = sin θ

4π ε0 jk0 r ( jk0 r ) 2
r 4π ε0 r

jIˆ l µ 0 e − jk r
0

= sin θ .
2 λ ε0 r
Eθ( Fern ) µ0
= = Z 0 ≈ 120π Ω ≈ 377 Ω.
Hφ( Fern )
ε0
154
Eθ( Fern ) (θ ) = E0 (r ) sin θ ,

Eθ( Fern ) (θ )
= sin θ : radiation pattern.
( Fern )
Eθ (θ max )

Eθ
θ θ
Emax

155
 Radiated power:

Sphere R  λ : far field.

1 1 1
S= E× H =
*
Eθ H φ*eθ × eφ = Eθ H φ*e r .
R 2 2 2
2
n = er 1
ˆ
I l 2 µ 1
Eθ H φ =
*
( ) 0
sin 2
θ.
2 8 λ ε0 R 2

P= ∫ = ⋅ Γ Γ =θ θ φ sin θ dθ dφ .
2
S n d . d Rd R sin d R
Kugel
2 2
Iˆ 2π π ˆ π π
l 2 µ0 I l µ
(= ∫ ∫ θ θ φ = ∫ θ dθ
3 2 0 3
P ) sin d d ( ) sin
8 λ ε0 0 0 4 λ ε0 0
 
2 4
ˆI π
l 2 µ0 1
3
2 l
= ( ) Rs Iˆ . Rs ≈ 80π 2 ( ) 2 : Radiation resistance.
3 λ ε0 2 λ 156
 4.3 Guided waves

Transmission lines: transversal (x, y) dimensions are much

smaller than the wave length.
z
2π 2π 1 v
=
D  λ == = .
ω µε k f µε f

If this condition is not fulfilled: waveguide.

e. g.: cylindrical Metallic tube

waveguide (γ → ∞)
b µ ,ε
157
a
Assumptions:

• sinusoidal time dependence: all quantities are

complex amplitudes,
• material properties are constant:
µ, ε = constant,
• Lossless medium: γ = 0,
• No free charges: ρ = 0.

158
4.4.1 TM and TE waves

Wave equations for the potentials A and V:

In time domain:
∂2A ∂ 2V
−∆A + µε 2 = 0, −∆V + µε 2 = 0.
∂t ∂t

In frequency domain:

−∆A − ω 2 µε A = 0, −∆V − ω 2 µεV = 0,

k = ω µε : −∆A − k 2 A = 0.

159
 Electromagnetic field in case A( x, y, z ) = A( x, y, z )e z :

ex e y ez
1 1 ∂ ∂ ∂ 1 ∂A 1 ∂A
=H curl=
( Ae z ) = ex − ey ,
µ µ ∂x ∂y ∂z µ ∂y µ ∂x
0 0 A
1
E= [k 2 Ae z + graddiv( Ae z )] =
jωµε
1 ∂2 A ∂2 A ∂2 A
= [ ex + e y + ( 2 + k 2 A)e z ].
jωµε ∂x∂z ∂y∂z ∂z

H z = 0 : the longitudinal component of the magnetic field is zero.

The magnetic field is transversal: TM waves.

160
Alternative to the potentials A and V:
1
=
D curl F, E curlF,
ε
curlH − jω D =curl (H − jω F ) =0 ⇒ H = jω F − gradψ ,
=B jωµ F − µ gradψ .

F : electric vector potential, ψ : magnetic scalar potential.

curlE = − jω B : curlcurlF = graddivF − ∆F = − jωµε ( jω F − gradψ ).

Lorenz gauge: divF = jωµεψ .

−∆F − ω 2 µε F = 0, divB= 0 ⇒ −∆ψ − ω 2 µεψ = 0.

−∆F − k 2 F = 0 : wave equation.

161
Using the Lorenz gauge, ψ can be eliminated:
1
=
divF jωµεψ ⇒
= ψ divF.
jωµε

1
H=jω F − gradψ =jω F − graddivF =
jωµε
1
=
− (ω 2 µε F + graddivF ).
jωµε

1
D = curlF, H=
− (k 2 F + graddivF ).
jωµε

162
Electromagnetic field in case F ( x, y, z ) = F ( x, y, z )e z :

ex e y ez
1 1 ∂ ∂ ∂ 1 ∂F 1 ∂F
E curl=( Fe z ) = ex − ey ,
ε ε ∂x ∂y ∂z ε ∂y ε ∂x
0 0 F
1
H=− [k 2 Fe z + graddiv( Fe z )] =
jωµε
1 ∂2 F ∂2 F ∂2 F
=
− [ ex + e y + ( 2 + k 2 F )e z ].
jωµε ∂x∂z ∂y∂z ∂z

Ez = 0 : the longitudinal component of the electric field is zero.

The electric field is transversal: TE waves.

163
The general solution of Maxwell’s equations in
homogeneous media can be written as the superposition
of TM and TE waves.
Hence, the single component vector potentials allow the
description of the electromagnetic field by means of two
scalar functions.

164
4.4.2 Waves in rectangular waveguides
z
Assumption:
0, 0 x=a
x Wave propagation in the
positive z-direction.
y=b
y
− jβ z
TM waves: =
A ( x , y , z ) =
A( x , y , z )e z A( x , y ) e ez ,
TE waves: = ( x, y, z )e z F ( x, y )e − j β z e z .
F ( x, y, z ) F=
Boundary conditions: E × n =0 on the walls of the waveguide.
=
E y 0,=
Ez 0 = =
at x 0 and x a,

=
Ex 0,=
Ez 0 = =
at y 0 and y b.
165
 ∂ ∂2
Boundary conditions for the potentials: = − jβ , 2 =
−β 2.
∂z ∂z
∂2 A
1 ∂2 A ∂2 A
=
TM waves: E [ ex + e y + ( 2 + k= 2
A)e z ]
jωµε ∂x∂z ∂y∂z ∂z
1 ∂A ∂A
=− [ jβ ex + jβ e y + ( β 2 − k 2 ) Ae z ].
jωµε ∂x ∂y
=A 0 at= x 0,=
x a= , y 0 and= y b, Dirichlet B.C.
1 ∂F 1 ∂F
=
TE waves: E ex − ey.
ε ∂y ε ∂x
∂F
= 0= =
at x 0 and x a,
∂x
Neumann B.C.
∂F
= 0= =
at y 0 and y b,
∂y
166
TM waves: −∆A − k 2 A = 0, k = ω µε .
A( x, y, z ) = A( x, y )e − j β z e z :

∂2 A ∂2 A
− 2 − 2 + β 2A− k2A =
0.
∂x ∂y

Solution by separation: A( x, y ) = X ( x)Y ( y ).

d 2 X ( x) d 2Y ( y )
−Y ( y ) 2
− X ( x ) 2
+ ( β 2
− k 2
) X ( x)Y ( y ) =
0,
dx dy
1 d 2 X ( x) 1 d 2Y ( y )
− 2
− 2
+ β 2
− k 2
= 0.
X ( x) dx Y ( y ) dy
    
f ( x) g ( y)

=
This is only possible =
if f ( x) constant, g ( y ) constant.
167
 The following ordinary differential equations are obtained:

d 2 X ( x) d 2Y ( y )
2
=
fX ( x), 2
gY ( y ).
dx dy
Solution:
f = − k=
− k x2 , g = 2
y :
X ( x) C1x cos k x x + C2 x sin k x x,
=
Y ( y ) C1 y cos k y y + C2 y sin k y y.
π
boundary conditions: X (0) =X (a ) =⇒
0 C1x =0, k x =m , m=
1, 2,... ,
a
π
Y (0) =Y (b) =⇒
0 C1 y =0, k y =n ,n=
1, 2,... .
b

168
 TMmn waves:
mπ nπ
C = C2 x C2 y : A( x, y, z ) = C sin( x) sin( y )e − j β z .
a b
1 ∂A 1 ∂A C nπ mπ nπ − jβ z
H ex − ey : H = sin( x ) cos( y ) e ,
µ ∂y µ ∂x µ b
x
a b
C mπ mπ nπ
Hy = − cos( x) sin( y )e − j β z ,
µ a a b
H z = 0.
1 ∂A ∂A
E=
− [ jβ ex + jβ e y + ( β 2 − k 2 ) Ae z ] :
jωµε ∂x ∂y
C β mπ mπ nπ
Ex = − cos( x) sin( y )e − j β z ,
ωµε a a b
C β nπ mπ nπ
Ey = − sin( x) cos( y )e − j β z ,
ωµε b a b
C (k 2 − β 2 ) mπ nπ
Ez = sin( x) sin( y )e − j β z .
jωµε a b 169
TE waves: −∆F − k 2 F = 0, k = ω µε .
F ( x, y , z ) = F ( x, y )e − j β z e z :

∂2 F ∂2 F
− 2 − 2 + β 2F − k 2F =
0.
∂x ∂y

Solution by separation: F ( x, y ) = X ( x)Y ( y ).

d 2 X ( x) d 2Y ( y )
−Y ( y ) 2
− X ( x ) 2
+ ( β 2
− k 2
) X ( x)Y ( y ) =
0,
dx dy
1 d 2 X ( x) 1 d 2Y ( y )
− 2
− 2
+ β 2
− k 2
= 0.
X ( x) dx Y ( y ) dy
    
f ( x) g ( y)

=
This is only possible =
if f ( x) constant, g ( y ) constant.
170
The following ordinary differential equations are obtained:

d 2 X ( x) d 2Y ( y )
2
=
fX ( x), 2
gY ( y ).
dx dy
Solution:
f = − k=
− k x2 , g = 2
y :
X ( x) C1x cos k x x + C2 x sin k x x,
=
Y ( y ) C1 y cos k y y + C2 y sin k y y.

boundary conditions:

dX (0) dX (a ) π
= =⇒
0 C2 x =0, k x =m , m=0,1, 2,... ,
dx dx a

dY (0) dY (b) π
= =⇒
0 C2 y =0, k y =n ,n=0,1, 2,... .
dy dy b 171
TEmn waves:
mπ nπ
C = C1x C1 y : F ( x, y, z ) = C cos( x) cos( y )e − j β z .
a b
1 ∂F 1 ∂F C nπ mπ nπ
E ex − e y : Ex = − cos( x) sin( y )e − j β z ,
ε ∂y ε ∂x ε b a b
C mπ mπ nπ
Ey = sin( x) cos( y )e − j β z ,
ε a a b
Ez = 0.
1 ∂F ∂F
H [ jβ ex + jβ e y + ( β 2 − k 2 ) Fe z ] :
jωµε ∂x ∂y
C β mπ mπ nπ
Hx = − sin( x ) cos( y )e − j β z ,
ωµε a a b
C β nπ mπ nπ
Hy = − cos( x) sin( y )e − j β z ,
ωµε b a b
C(β 2 − k 2 ) mπ nπ
Hz = cos( x) cos( y )e − j β z .
jωµε a b 172
Satisfaction of the separation equation: − f − g + β 2 − k 2 =0.

mπ nπ 2
k +k +β −k =
2
x
2
y 0, =
2
kx 2
=
, ky , k ω 2 µε .
=
a b
mπ 2 nπ 2
( ) + ( ) + β 2 − ω 2 µε =
0, m, n = (0), 1, 2, ... .
a b

At given values of n and m (wave modes), the angular

frequency is not arbitrary: β cannot be negative!
2

If β 2 < 0, one had β = α ⇒ e − j β z =

± jα , − j β = e α z :

attenuation only, no wave propagation.

173
 mπ 2 nπ 2 π m 2 n 2
β= ω µε − ( ) − ( ) ≥ 0 ⇒ ω ≥
2 2
( ) +( ) .
a b µε a b
 
2π f g
1 m 2 n 2
=
f ≥ fg ( ) +( ) .
2 µε a b

fg: cut-off frequency.

Any particular mode is only propagable above the cut-off

frequency.

174