Boundary Value Problems

Introduction
● Previous discussion covered
– Uniform plane waves
– Transmission line structures
● Special techniques were used to solve such
problems
● New methods are required to solve problems
on wave guides, resonance, and radiation
 Introduction
● Solutions to differential equations subject to
boundary conditions
– Solution of field distributions
– These solutions may be used to solve for other
parameters
● Techniques to be discussed are
– Numerical solutions of boundary value problems
– Method of separation of variables
 Numerical Methods
● Increase in performance of computers allow us
to achieve accurate solutions for EM field
analysis through numerical methods
– May be used to evaluate analytic expressions
● No closed form solution
● Wholly numerical solution
● Only the most basic of methods are considered
● Choice of method depends on
– Accuracy, speed, versatility, hardware requirements
 Numerical Methods
● Finite difference method
– Simple but large range of applications
– For a given ordinary differential equation f(x)', it can
be expressed approximately as:
f  xa− f  x
f  x' ≈
a
– The differential equation may be approximated as
f  xa = f  x a f '  x 
– For a given set of boundary conditions, we may be
able to find the approximate values of the solution
at points with unknown values for f(x) and f'(x)
 Numerical Methods
● Finite Element Method
– More difficult than finite difference method
– More adaptable to complex boundarys (irregular
shapes), and spatially varying medium properties
– Used for solving 3 dimensional full wave problems
 Numerical Methods
● Method of moments
– Based on integral equation
– Efficient for solving planar structures
– Results in knowledge of charge distributions on
surfaces of the structure
– Allows to find the total charge in the structure for a
given potential
● Find capacitance
 Numerical Methods
● Let us first investigate how MoM may be used
to find the static charge distribution and
capacitances for 2 and 3 dimensional planar
structures
● The structure is subdivided into small elements
– Preparation for discrete numerical calculations
● Surface charge density ρsi is assumed to be
uniform for each of the elements
 Numerical Methods
● The total charge for each element is ρsi Δsi.
● The total charge is then treated as a point
charge located at the center of the element
– Point charge is used in making potential
calculations
● For 2D structures, there is no variation in axial
direction
– Charge per unit length ρsi Δsi is used instead of total
charge
– Represented as a line charge located at the center
of the element
 Numerical Methods
● Potentials are calculated using formulas for
charges in free space
– Conductors in the structure are accounted for by
including all charges on the surfaces
● The potential at the center of the element ith
element is the sum of the potentials due to each
of the charges in the structure
N
 sj  S j
V i =V ii ∑
j ≠i 4 ∣ri− rj∣
 Numerical Methods
● The term Vii is the potential at the center of the
ith element due to the charge on the ith element
itself
– Must be handled separately from the summation
term (if i = j, term inside summation is singular)
– This is computed by integrating the charge density
over the element
● Point charge equivalent is not used for solving this
potential
● For convenience, element is assumed to be a disk even if
actual element is a square
2 a
si r dr si
V ii =∫ d  ∫ =0.282   S i
0 4  r
0

 Numerical Methods
● For 2 dimensional structures, the charges
considered are line charges
– The potential at the center of segment i
N

∑  sj  l j ln∣ri− rj∣
j≠i
V i=V ii  C T
2 
– Again, Vii must be handled separately to avoid
singularity

V ii =
− si  l i
2  [
ln
 li
2
−1
]
 Numerical Methods
● Let us try to use MoM to solve a parallel plate
structure

V=Vo
1 2 3 4

5 6 a
V=-Vo 7 8

a
 Numerical Methods
● The equation for element 1 can be written as
a s2 a  s8 v 
4 0.282s1  ... = V0
∣r1−r2∣ ∣r1 −r8∣ a
– Writing equations for each of the elements and
expressing in matrix form

[ [] ]
a a

[]
3.54 ... V0
∣r1− r2∣ ∣r1− r8∣ S1
V0
a a S2 4 
3.54 ... = ...
∣r2−r 1∣ ∣r2− r8∣ ... a
−V 0
a a  s8
... 3.54 −V 0
∣r8 −r 1∣ ∣r8− r2∣
 Numerical Methods
● System of 8 equations and 8 unknowns
– If the potentials at the conductors are know, surface
charge densities may be solved
– From the surface charge densities, the fields may
also be found
– The capacitance is just Qtotal/2V0
– Values found are subject to errors due to the
approximations made
 Numerical Methods
● Assumptions made are:
– Charges due to each of the elements are due to
point charges at the center of the elements
– For the Vii, the shape was assumed to be a circular
disk instead of a square
● Errors are dependent on these assumptions
– For more accurate results, it is required that the
conductor surfaces be divided into very small
elements such that the contributions of these errors
are very small
– Inversion of a large matrix is required
 Method of Separation of Variables
● One of the most powerful techniques to solve
linear partial differential equations
● Leads to solutions which are made up of the
products of n independent functions (for n-
dimensional problems)
– Each function depends only on one variable
F  x , y , z = X  x Y  y  Z  z 
– Though not strictly general, but may be added to
form a series expansion to represent more general
functions
 Method of Separation of Variables
● Single product solutions of the wave equation
represent modes which can propagate
individually
– Different modes of propagation in waveguides
● Let us consider the 2D Laplace's equation in
rectangular coordinates
2 2
∂ V ∂ V
2
 2 =0
∂x ∂y
 Method of Separation of Variables
● We assume that the solution V(x,y) is the
product of two functions, each only a function of
x or y
V  x , y = X  xY  y
● Substituting this solution to the second order
differential equation we have:
X ' ' Y  X Y ' '=0
● Separating the sums into functions of one
variable only X '' Y ' '
 =0
X Y
 Method of Separation of Variables
X '' Y ' '
 =0
X Y
● The previous equation is to hold for all values of
x and y for the solution to be valid
– First term is only a function of x, and second term is
only a function of y
– The second term does not vary with x
– For the equation to hold, the first term must not vary
with x as well
– Therefore, the first and second terms of the
equation must be constants
 Method of Separation of Variables
● Let us denote the constant for the first term as
Kx2 and the constant for the second term as Ky2
● The equation then becomes
2 2
K K =0
x y
● Which can be separated into two second order
equations
X '' 2 Y '' 2
=K x =K y
X Y
2 2
X ' '=K x X Y ' '=K y Y
2 2
X ' '−K X =0
x Y ' '−K y Y =0
 Method of Separation of Variables
● These are two second order ordinary differential
equations.
● The characteristic equation for these equations
are: 2 2
s −K =0
● The roots of this characteristic equation are
s=±K
● The solutions can be expressed either as
exponentials or as hyperbolic functions
Kxx −K x x
X  x= Ae B e Ky y
Y  y =C e D e
−K y y

X x =A cosh K x xB sinh K x x Y  y =C cosh K y yD sinh K y y

 Method of Separation of Variables
● Therefore the general form of our solution
should be:
V  x , y = A cosh K x xB sinh K x xC cosh K y yDsinh K y y 

● Because of the relationship between Kx and Ky,

we can see that at least one of the two must be
imaginary
K 2x K 2y =0
2 2
K x =−K y
K x = j K y K y= j K x
 Method of Separation of Variables
● Our general solution can be rewritten in terms
of hyperbolic and sinusoidal functions
V  x , y = A cosh K xBsinh K xC cos K yD' sin K y
or
V  x , y = A cos K xB ' sin K xC cosh K yD sinh K y
where K x = j K y =K K y = j K x=K
● The choice between the two possible types of
general solutions are dependent on the
boundary conditions
 Method of Separation of Variables
● If the potential is required to have repeated
zero's as a function of y, the first equation is the
solution
● If the potential is required to have repeated
zero's as a function of x, the second equation is
the solution
● If the boundaries extend to infinity, in one
direction, real exponentials are used in place of
hyperbolic functions
 Method of Separation of Variables
● Let us now consider the solution for the 3
dimensional case
2 2 2
∂ V ∂ V ∂ V
2
 2  2 =0
∂x ∂y ∂z
● The solution may be expressed as the product
of three independent functions
V  x , y , z = X  x Y  y  Z  z 
● Substituting these equations
X x ' ' Y  y  Z  z X  x Y  y ' ' Z  z  X  xY  y Z  z ' ' =0
X '' Y ' ' Z''
  =0
X Y Z
 Method of Separation of Variables
● Using the same argument for the two
dimensional case, the three terms must also be
constants 2 2 2
K x K y K z =0
● The general solution is of the form
V  x , y = A cosh K x xB sinh K x xC cosh K y yDsinh K y y 
 E cosh K z zF sinh K z z 
● For the first equation to hold, at least one of the
squared constants must be negative
– At least one of Kx, Ky or Kz must be imaginary
 Method of Separation of Variables
● Consider a 2 dimensional problem with the
following boundary conditions:
– V=0 at x= 0,a and at y = 0
– V=V0 at x = a/2 and y = b
● Notice that we have periodic zeros along x.
Therefore, the solution that we should use is:
V  x , y = A cos K xB ' sin K xC cosh K yD sinh K y
 Method of Separation of Variables
● We now apply the boundary conditions to solve
for the coefficients of the solution
– @ x =0
V 0, y= A cos K 0B ' sin K 0C cosh K yDsinh K y =0
 AC cosh K yDsinh K y =0
● For this equation to hold for all values of y, A must be
zero
● The solution for the potential field simplifies to
V  x , y = B sin K xC cosh K yD sinh K y 
Method of Separation of Variables
– @ x = a, V=0
V a , y = B sin K aC cosh K yDsinh K y =0
● For this condition to hold without setting B=0 (trivial
solution) K must be equal to π/a
– @ y = 0, V = 0
  
V  x ,0= B sin xC cosh 0D sinh 0=0
a a a

B sin xC=0
a
● For this condition to hold for all values of x, C must be
equal to zero
Method of Separation of Variables
– The solution can be simplified to
   
V  x , y = B sin x  D sinh y =E sin x sinh y
a a a a
– @ x = a/2 y = b, V = V0
a  b
V a /2, b=E sin sinh b=E sinh =V 0
a 2 a a
V0
E=
b
sinh
a
– The solution is therefore:
 
V 0 sin x sinh y
a a
V  x , y =
b
sinh
a
 Fourier Series and Integral
● Preceding discussion shows that single-product
solution can satisfy only very special forms of
boundary conditions
– For general boundaries, a sum of each solution
must be used
– Fourier series or integrals become useful in forming
solutions for these problems
– We review fourier tools that may be used to solve
these problems
 Fourier Series
● Fourier series are used to represent periodic
functions
– A periodic function satisfies the following condition:
● f(x) = f(x + L) where L is the period of the function
– WE assume that the periodic function can be
represented by a constant plus the sum of infinite
series of sine and cosine functions of harmonics of
a fundamental spatial frequency
f  x=a 0a1 cos k xa 2 cos 2 k x....b1 sin k xb 2 sin 2 k x....
 Fourier Series
● The phase factor k is related to the period
through kL = 2π
● We use the orthogonality properties of
sinusoids to evaluate the unknown coefficients
of our infinite series
L L
2 2

∫ cos nkx cos mkx dx=0 ; m≠n ∫ sin nkx cos mkx dx=0 ; m≠n /m=n
−L −L
2 2
L L L
2 2 2
L
∫ sin nkx sin mkx dx=0 ; m≠n ∫ cos nkx dx=∫ sin nkx dx= 2
2 2

−L −L −L
2 2 2
 Fourier Series
● To be able to use these properties, we multiply
each term in the series by cos nkx and integrate
over the period
– Each term on the right vanishes except for the
cos nkx term, which leaves us anL/2
L/ 2
2
a n= ∫
L −L/2
f  x cos nkx dx
● Multiplying the series by sin nkx and integrating
again gives us
L/2
2
b n= ∫
L −L/ 2
f  xsin nkx dx
 Fourier Series
● To get the constant term a0, every term is
integrated over a period
– All cos and sin terms become zero
– a0 becomes the average value of the function f(x)
L/ 2
1
a 0= ∫ f  x dx
L − L/2
● The function f(x)

 [∫ ] [ ] 
L/ 2 n=∞ L/ 2 L/ 2
1 2 2
f  x= ∫ f  x dx ∑ f  x cos nkx dx cos nkx ∫ f  x sin nkx dx sin nkx
L −L /2 n=1 L −L/ 2 L −L/2
 Fourier Series
● For a general function, an infinite number of
terms is required in the Fourier series
representation
– When other terms are discarded, the function
retains the desired shape, but is somewhat
distorted
– For functions with sharp discontinuities, large
number of terms may be required to approximate
sharp corners properly
– Series does not converge to the function near the
discontinuity (Gibbs phenomenon)
 Fourier Integral
● In some problems, the function of interest is
aperiodic and is defined over the entire range
● An example is a square pulse with value C in
the interval -a <x< a and zero otherwise
– Limiting case, where L approaches infinity
– Spacing between components becomes vanishingly
small
– The infinite summation becomes an integral
 Fourier Integral
The function is now represented as
●

1 ∞
f  x= ∫
2  −∞
jkx
g k e dk
● The function g(k) which replaces the an and bn terms
of the infinite series become
∞
g k = ∫ f  xe
− jkx
dx
−∞
● For the integral pair to work it is required that the
function be continuous or have only a finite number
of finite discontinuities, and must be absolutely
integrable ∞
∫
−∞
∣ f  x∣dx∞
 Fourier Applications
● The fourier series may be used to give the
complete solution of a boundary value problem
which can not be expressed as only one
solution
– Express the solution as a series of solutions
– Fourier series may be used
 Fourier Applications
● Consider a structure with the following
boundary conditions
– V=0 at x = 0,a and y=0
– V= V0 at y = b
– Assume that there is an infinitesimal gap between
the boundaries at x =0,a and y = b
● Form of solution to be selected must have
sinusoidal variations along x
– Repeated zeros along x (boundary condition)
 Fourier Applications
● The form of the solution must be:
V  x , y = A cos KxBsin KxC cosh KyD sinh Ky

● From the boundary conditions, the solution

must be zero at y = 0 for all values of x
– Evaluating at y = 0
V  x ,0= A cos KxB sin KxC =0
– For the condition to hold, C = 0
 Fourier Applications
● From the boundary conditions, at x = 0, V=0 for
all y
– Substituting into the solutions
V 0, y= A D sinh Ky=0
– For the condition to hold, A must be zero
● From the boundary conditions, at x = a, V=0 for
all y
V a , y = B sin K a Dsinh Ky=0
– The only way for this condition to hold without
giving us a trivial solution is when K= m
a
 Fourier Applications
● Combining the constants B and C, the form of
the solution is:
m m
V  x , y =C m sin x sinh y
a a
● This form of the solution satisfies the boundary
condition at x=0,a and y =0, but a single term of
this form can not satisfy the condition at y =b
– we may express the solution as an infinite series to
satisfy the condition at y =b
∞
m x m y
V  x , y = ∑ C m sin sinh
m=1 a a
 Fourier Applications
● This infinite series satisfies the boundary
conditions for x=0,a and y = 0
● For the solution to hold, it must satisfy the
condition at y = b
∞
m x m b
V 0= ∑ C m sin sinh 0xa
m=1 a a
– This is the fourier series expansion in sines of a
constant with value V0 over the interval 0<x<a
(pulse)
 Fourier Applications
● The sinh term is constant and may be included
in the coefficient
∞
m x
V 0= ∑ bm sin
m=1 a
– The coefficient bm is given by
a a
2V 0
2
b m= ∫ V 0 sin
a0
m x
a
dx=
m [
−cos
nx
a ]
0
4V0
bm = for m odd
m
=0 for m even
 Fourier Applications
● Comparing the two equations we have
m b
b m=C m sinh
a
4V 0
C m= for m odd
mb
m sinh
a
● The complete solution is therefore:
4 V 0 sinh m y /a
V  x , y = ∑ m sinh m b/ a 
sin m x/a 
m odd
 Cylindrical Harmonics
● Previous discussion concerns the solution of
boundary value problems in rectangular
coordinates
● There are a large variety of other boundary
value problems where the boundaries are better
defined in cylindrical coordinates
● Method of Separation of Variables may also be
applied to such problems, taking note that the
equations must be in cylindrical coordinates
 Cylindrical Harmonics
● A variety of solutions are found depending on
the symmetries that were assumed
● Laplace's equation in cylindrical coordinates is

 
2 2
1 ∂ ∂V 1 ∂ V ∂ V
  2  2 =0
 ∂ ∂  ∂ 2
∂z
● For a system with no variations wrt φ and z

V  r=C 1 ln C 2
 Cylindrical Harmonics
● For systems with no variation wrt z, we have
circular harmonics of the form
V  , = A B 
n −n
 C cos nD sin n  
● For systems with axial symmetry (no variations
with φ), Laplace's equation becomes
2 2
∂ V 1 ∂V ∂ V
  2 =0
∂ 2  ∂ ∂z
 Cylindrical Harmonics
● We assume that the solution is of the form
V  , z=P  Z  z
● Substituting into the differential equation
1
P ' ' Z  P ' Z PZ ' ' =0

● Dividing the entire equation by PZ
P'' 1 P' Z ' '
  =0
P  P Z
Z''
Z
=−
[ P'' 1 P'
P

 P ]
 Cylindrical Harmonics
● Again for the equality to hold, both the left and
right side of the equation must be equal to a
constant
– We let the constant be T2 we now have 2 differential
equations
2
1 d P 1 dP 2
 =−T
P d 2  P d 
2
1 d Z 2
=T
Z d z2
 Cylindrical Harmonics
● The first equation can be rewritten as
d2P 1 d P 2
 T P=0
d  d 
2

– This equation is the simplest form of the Bessel

equation
– A method for finding a solution is to substitute an
infinite series and find the conditions on the terms
of the series for it to be a valid solution of the
differential equation
2 n
P=a0 a1 a 2  ... a n 
∞
P= ∑ a p 
p

p=0
 Cylindrical Harmonics
● Substitution of this series into the differential
equation shows that it is a solution if the
constants are
2m
T / 2
m
a p =a 2m =C1 −1
m !2
– The following equation is a solution to the
differential equation
∞ m 2m
−1 T  /2
P=C 1 ∑
m=0 m! 2
 Cylindrical Harmonics
● The series is convergent, and values may be
calculated for any value of the argument Tρ
– Such calculations have been made over a wide
range of values, and the results are tabulated
● If T2 is positive, the function defined by the
series is denoted by J0(Tρ) and called the
Bessel function of the first kind and zero order
∞ m 2m
−1 v/ 2
J 0 v= ∑
m =0 m! 2
 Cylindrical Harmonics
● The solution to the differential equation can be
written simply as
P=C 1 J 0 T 
● The differential equation of P is of second
order, and therefore must have a second
solution with a second arbitrary constant
– The second solution can not be obtained using
power series method
– At least one of the 2 solutions must be singular at
ρ=0
 Cylindrical Harmonics
● One form for the second solution is of the form

   
∞ m 2m
2 v 2 −1 v /2 1 1
N 0 v = ln J 0 v − ∑ 1 ...
 2  m=1 m! 2 2 m

– The constant γ = 0.5772... is eulers constant

● The general form of the solution to bessel's
equation is
P=C 1 J 0 T C 2 N 0 T 

● The solution to the differential equation in Z is

Z=C 3 sinh TzC 4 cosh Tz
 Cylindrical Harmonics
● Note that the Bessel function of the second kind
is singular at ρ=0, and cannot be present in any
problem that includes ρ=0 in the region where
the solution applies
● If T2 is negative, we let T2 = -τ2 or T = jτ, where τ
is real
– A new series may be obtained which is also real
and convergent
J 0  j  =I 0  
– A second solution must also exist and is denoted by
K0(τρ)
 Cylindrical Harmonics
● The complete solution for negative T2 is given
by
P=C 1 J 0  C 2 K 0  
– The second solution is also singular at ρ=0
● The solution to Z when T2 is negative is given
by Z=C sin  zC cos  z
3 4
● The complete solutions to laplace's equation
are
V  , z=C 1 J 0 T C 2 N 0 T C 3 sinh TzC 4 cosh Tz 
V  , z=C 1 I 0  C 2 K 0  C 3 sin  z C 4 cos  z 
 Cylindrical Harmonics
● Some problems require variations in all
coordinates to be included
– Field between two halves of a longtitudinally split
cylinder
– Solution may again be assumed to be of the form
PΦZ
– The solutions for Z and are Φ:
Z=C cosh TzDsinh Tz
=E cos v F sin v 
 Cylindrical Harmonics
● The differential equation for P becomes
2 2
d P 1 dP 2 v
 T − 2  P=0
d  d 
2

● The series solution to this equation is the more
general form of the Bessel function
∞ m v2m
−1 T /2
J v T =∑
m=0 m!  vm1

– Where Γ(v+m+1) is the gamma function of (v+m+1)

● For integral values of v, is equivalent to (v+m)!
● For non integral values, the tables of the gamma function
are used
 Cylindrical Harmonics
● If v is an integer n, we have
∞ m v2m
−1 T /2
J n T = ∑
m=0 m! nm!
– It can be shown that J −n=−1n J n
● The second independent solution to the Bessel
equation is
J v T  cos v −J −v T 
N v T =
sin v 
– And N −n=−1n N n
 Cylindrical Harmonics
● The complete solution to the general Bessel
equation may be written as
P= A J v T B N v T 
● The constant v is known as the order of the
equation
– Jv is the Bessel function of the first kind, of order v
– Nv is the Bessel function of the second kind, of
order v
 Function as a Series of Bessel
Functions
● From Fourier series, we have seen that a
function may be expressed as a series of
sinusoids
– Coefficients of the sinusoids may be evaluated due
to the orthogonality of sinusoids
● Orthogonality is also seen for Bessel functions

  
a
pm  pq 
∫ J 0 a
J0
a
d =0
0

– Where pm and pq are the mth and qth roots of J0(v)

=0
 Function as a Series of Bessel
Functions
● A function can be expressed as an infinite sum
of zero order bessel functions as

 
pm
∞
P= ∑ b m J 0
m=1 a

 
a
2 pm 
b m= 2 2 ∫  f  J 0 a d 
a J 1 pm 0
 Example
● Consider a cylindrical system with the following
boundary conditions:
– V= 0 for z = 0,l V = V0 for ρ = a
● For this problem, the potential does not have
any variations with respect to φ
– The Bessel function required is of the zeroth order
V  , z=C 1 J 0 T C 2 N 0 T C 3 sinh TzC 4 cosh Tz 
V  , z=C 1 I 0  C 2 K 0  C 3 sin  z C 4 cos  z 
 Example
● Since we have periodic zeros along z, the
solution must be of the sinusoidal form
V  , z=C 1 I 0  C 2 K 0  C 3 sin  z C 4 cos  z 

– The z axis is included in the region where there

must be a solution, therefore, C2 must be zero
● K0 is singular at the z axis
– We have a zero at z = 0, therefore C4 must also be
zero
– We must have a zero at z = l, therefore, τ = m π/l
 Example
● So far, our solution is
m m
V  , z=C m  I 0  sin z
l l

– This solution satisfies the boundaries at z = 0, l

● The last boundary condition is that V=V0 at ρ=a
– Previous equation by itself can not satisfy this
condition
– We must use Fourier series
 Example
● We have:
∞
m m ∞
m
V 0= ∑ C m I 0  a sin z = ∑ b m sin z
 m=1 l l m =1 l
l l
2V0
2
b m= ∫ V 0 sin
l 0
m
l
z dz=
m
−cos [
m z
l ]
0

4V 0
b m=
m
=C m I 0
 
m a
l
4V0
C m=
m I 0 m a /l
4 V 0 I 0  m/ l m z
V  , z= ∑ sin
m odd m I 0 m a/ l l