MATLAB Codes

clear all; format compact;format short g; tic

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% THIS PROGRAM GENERATES HERMITE'S FUNCTIONS HN(X) IN TWO WAYS
USING:
% 1) SERIES EXPANSION
% 2) RECURRENCE RELATION
% THE TWO METHODS ARE COMPARED
% X = ARGUMENT (FIXED IN THIS PROGRAM)
% N = ORDER OF THE FUNCTION
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

X = 0.5; %Argument
NMAX = 15; %Order of Function

% METHOD 1: SERIES EXPANSION FORMULA Eq. (2.190)

for N = 0:NMAX
SUM = 0;
FN = factorial(N);
I = floor(N/2); %Greatest Integer Function
for K = 0:I
M = N - 2*K;
FM = factorial(M);
FK = factorial(K);
SUM = SUM + ( ((-1)^K)*FN*((2*X)^M) )/( FK*FM );
end
HS(N+1) = SUM;
end

% METHOD 2: RECURRENCE FORMULA Eq. (2.191a)

HR(1) = 1;
HR(2) = 2*X;
for k = 2:N
n = k-1; %MATLAB HR vector starts at 1 while equation
% subscript starts at 0
HR(k+1) = 2*X*HR(k) - 2*(n)*HR(k-1);
end

Difference = HS-HR;

hdr = [{'N'},{'Series Expansion'},{'Recurrence'},{'Difference'}];

 output = [hdr; num2cell([(0:N)',HS',HR', Difference'])];
disp([' Values of Hn(x) for x = ',num2str(X),', 0<=n<=',num2str(N)])
disp(output)

Fig. 2.11 Program for Hermite function Hn(x).

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%
% MAIN PROGRAM
%
% FOR SPHERICAL RAIN DROPS, THIS PROGRAM CALCULATES
ATTENUATION IN dB/KM
% AND PHASE SHIFT IN DEG/KM FOR A GIVEN RAIN RATE
%
% R = RAIN RATE IN MM/HR
% D = DROP DIAMETER IN CM
% F = FREQUENCY IN GHZ
% AT = ATTENUATION IN dB/KM
% PH = PHASE SHIFT IN DEG/KM
% V = TERMINAL VELOCITY OF RAIN DROPS
% P = PERCENT OF TOTAL VOLUME AS MEASURED
% M = COMPLEX REFRACTIVE INDEX OF WATER AT T = 20 C
% X = ALPHA = K*A

clear all; format compact; tic

%Inputs:
F = 11.0; %Frequency (GHz)
NMAX = 10; %Ideally inf but 10 is sufficient eqn (2.256)

LAM = 30.0/F; % WAVELENGTH IN CM

% Raindrop Terminal Velocity Data

V=[2.1,3.9,5.3,6.4,7.3,7.9,8.35,8.7,9.0,9.2,9.35,9.5,9.6,9.6];

% Rain rate (mm/hr)

R=[0.25,1.25,2.5,5.0,12.5,25.0,50.0,100.0,150.0];

% Laws and Parsons drop-size distribution for various rain rate (% of total volume)
P =1/100*[28.0, 50.1, 18.2, 3.0, 0.7, 0 0 0 0 0 0 0 0 0;...
10.9, 37.1, 31.3, 13.5, 4.9, 1.5, 0.6, 0.2, 0 0 0 0 0 0;...
7.3, 27.8, 32.8, 19.0, 7.9, 3.3, 1.1, 0.6, 0.2, 0 0 0 0 0;...
4.7, 20.3, 31.0, 22.2, 11.8, 5.7, 2.5, 1.0, 0.5, 0.3, 0 0 0 0;...
2.6, 11.5, 24.5, 25.4, 17.3, 10.1, 4.3, 2.3, 1.2, 0.6, 0.2 0 0 0;...
1.7, 7.6, 18.4, 23.9, 19.9, 12.8, 8.2, 3.5, 2.1, 1.1, 0.5, 0.3 0 0;...
 1.2, 5.4, 12.5, 19.9, 20.9, 15.6, 10.9, 6.7, 3.3, 1.8, 1.1, 0.5, 0.2, 0;...
1.0, 4.6, 8.8, 13.9, 17.1, 18.4, 15.0, 9.0, 5.8, 3.0, 1.7, 1.0, 0.7, 0;...
1.0, 4.1, 7.6, 11.7, 13.9, 17.7, 16.1, 11.9, 7.7, 3.6, 2.2, 1.2, 1.0, 0.3]';

f = [0.6 0.8 1 1.6 2 3 4 6 11 16 20 30 40 60 80 100 160 200 300];

% Refractive Index of Water at 20 degrees C, real and complex portions
mr= [8.96 8.956 8.952 8.933 8.915 8.858 8.78 8.574 7.884 7.148 6.614 5.581 4.886
4.052 3.581 3.282 2.82 2.668 2.481];
mi= [0.1713 0.2172 0.2648 0.4105 0.5078 0.7471 0.9771 1.399 2.184 2.614 2.78 2.848
2.725 2.393 2.100 1.864 1.382 1.174 0.8466];

% Interpolate refractive index of water to target frequency. F must be

% within the range of f.
mr_i = interp1(f,mr,F);
mi_i = interp1(f,mi,F);
M = mr_i - mi_i*i;

N = 1:NMAX;
D = 0.05*(1:length(V));
SO = zeros(1,length(V));
SB = zeros(1,length(V));
for I = 1:length(V) %I sweeps rain drop diameter

X = pi*D(I)/LAM;

% CALCULATE THE SCATTERING COEFFICIENTS an and bn

% Spherical Bessel function of the first kind of order n

J = sqrt(pi./(2*X)).*besselj(N+0.5,X);
JD = -N.*sqrt(pi./(2*X)).*besselj(N+0.5,X)+ X.*sqrt(pi./(2*X)).*besselj(N+0.5-1,X);
JM = sqrt(pi./(2*M*X)).*besselj(N+0.5,M*X);
JMD= -N.*sqrt(pi./(2*M*X)).*besselj(N+0.5,M*X)+
M*X.*sqrt(pi./(2*M*X)).*besselj(N+0.5-1,M*X);
H = sqrt(pi./(2*X)).*besselh(N+0.5,2,X);
HD = -N.*sqrt(pi./(2*X)).*besselh(N+0.5,2,X)+X.*sqrt(pi./(2*X)).*besselh(N+0.5-
1,2,X);

A1 = JM.*JD - J.*JMD;
A2 = JM.*HD - H.*JMD;
B1 = J.*JMD - (M^2)*JM.*JD;
B2 = H.*JMD - (M^2)*JM.*HD;
an = A1./A2;
bn = B1./B2;
cn = i./(X*A2);
dn = -i*M./(X*B2);
 % FORWARD SCATTERING AMPLITUDE FUNCTION S(O)
SO(I) = sum((2*N+1).*(an+bn))/2;
% NORMALIZED RADAR CROSS-SECTION
SB(I)=(abs( sum((2*N+1).*((-1).^N).*( an - bn )) )/X).^2;
end

AT = zeros(1,size(P,2));
PH = zeros(1,size(P,2));
for n = 1:size(P,2) % N sweeps rain rate
% CALCULATE ATTENUATION AND PHASE SHIFT
ATO = real(SO).*P(:,n)'./(6*D.^3.* V);
AT(n) = 4.343*R(n)*(LAM/pi)^2*sum(ATO);

PHO = imag(SO).*P(:,n)'./(6*D.^3.*V);
PH(n) = 90.0*R(n)*(LAM/pi)^2/pi*sum(PHO);
end

disp(['Attenuation and Phase Shift at ',num2str(F),' GHz'])

hdr = [{'Rain rate (mm/hr)'},{'Attenuation (dB/km)'},{'Phase shift (deg/km)'}];
out = [hdr; num2cell([R', AT', PH'])];
disp(out)

figure(1),
semilogy(R,AT,'--'),
xlabel('rain rate (mm/hr)'),ylabel('Attenuation (dB/km)')
legend([num2str(F),' GHz'],'location','southeast')

Fig. 2.16 MATLAB program for Examples 2.13 and 2.14.

% ********************************************************************
% MATLAB code for example 3.2 on one-dimensional wave equation solved
% using an explicit finite difference scheme
%*********************************************************************

clear all; format compact; tic

%Explicit Method
delx = 0.1; % resolution size
r = 1; % 'aspect ratio'
u = 1; % Constant of given wave equation
delt = r^2*delx/u; % time step size
Tsteps = round(1/delt); % Number of time steps

% X1 is the potential grid of the simulation, due to symetry only half

 % of the field is calculated.
X1 = zeros(Tsteps,1/(2*delx)+2); % Initilize X1

%Initial conditions and reflection line defined

x = 0:delx:.5+delx;
X1(1,:) = sin(pi*x);
X1(2,2:end-1) = .5*(X1(1,1:end-2)+X1(1,3:end));
X1(2,end) = X1(2,end-2); %reflection line

for row = 3:size(X1,1)

for col = 2:size(X1,2)-1
X1(row,col) = X1(row-1,col-1)+X1(row-1,col+1)-X1(row-2,col); % eqn. (3.26)
end
X1(row,end) = X1(row,end-2); %reflected line
end

%Use symetry condition to create entire field

X2 = [X1,fliplr(X1(:,1:end-3))];

figure(1),imagesc(0:delx:1,(0:delt:Tsteps*delt),X2),colorbar
ylabel('\leftarrow time (sec)')
xlabel('x')
title('Hyperbolic PDE')

if (delx==.1)
dispmat = [X1(1:8,1:7)];
disp(sprintf('\nCompare to Table 3.5, Solution of the Wave Equation in Example
3.2'))
disp(num2str(dispmat))
end

Fig. 3.8 MATLAB code for Example 3.2.

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%
% FINITE DIFFERENCE SOLUTION OF POISSON'S EQUATION:
% Vxx + Vyy = G
% USING THE METHOD OF SUCCESSIVE OVER-RELAXATION
%
% NX : NO. OF INTERVALS ALONG X-AXIS
% NY : NO. OF INTERVALS ALONG Y-AXIS
% A X B : DIMENSION OF THE SOLUTION REGION
% V(I,J) : POTENTIAL AT GRID POINT (X,Y) = H*(I,J)
% WHERE I = 0,1,...,NX, J = 0,1,....,NY
% H : MESH SIZE
% *******************************************************
 % SPECIFY BOUNDARY VALUES AND NECESSARY PARAMETERS
A=1;B=1;
V1=0;V2=10;V3=20;V4=-10;
NX= 20; %4 12 20
NY= NX;
H = A/NX;
% SET INITIAL GUESS EQUAL TO ZEROS OR TO AVERAGE OF FIXED
VALUES

for I=1:NX-1
for J=1:NY-1
V(I+1,J+1)=(V1 + V2 + V3 + V4)/4.0;
end
end
% SET POTENTIALS AT FIXED NODES
for I = 1:NX-1
V(I+1,1)=V1;
V(I+1,NY+1)=V3;
end
for J=1:NY-1
V(1,J+1)=V4;
V(NX+1,J+1)=V2;
end
V(1,1)=(V1 + V4)/2.0;
V(NX+1,1)=(V1 + V2)/2.0;
V(1,NY+1)=(V3 + V4)/2.0;
V(NX+1,NY+1)=(V2 + V3)/2.0;
% FIND THE OPTIMUM OVER-RELAXATION FACTOR
T = cos(pi/NX) + cos(pi/NY);
W = ( 8 - sqrt(64 - 16*T^2))/(T^2);
disp(['SOR Factor Omega = ',num2str(W)])
W4 = W/4;
% ITERATION BEGINS
NCOUNT = 0;

loop = 1;
while loop == 1;
RMIN = 0;
for I =1:NX-1
X = H*I;
for J = 1:NY-1
Y = H*J;
G = -36.0*pi*X*(Y - 1.0);
 R = W4*( V(I+2,J+1) + V(I,J+1) + V(I+1,J+2) + V(I+1,J)-4.0*V(I+1,J+1) -
G*H*H );
RMIN = RMIN + abs(R);
V(I+1,J+1) = V(I+1,J+1) + R;
end
end
RMIN = RMIN/(NX*NY);
if(RMIN>=0.0001)
NCOUNT = NCOUNT + 1;
if(NCOUNT>100)
loop = 0;
disp('SOLUTION DOES NOT CONVERGE IN 100 ITERATIONS')
end
else
%Then RMIN is less than .0001 and then solution has converged
loop = 0;
disp(['Solution Converges in ',num2str(NCOUNT),' iterations'])
disp(['h = ', num2str(H)])
end
end

Vnum = V;

%Grab original points a through i

abc = zeros(1,9);
a_tic = 1;
vec = [0:H:1];
for ii = .25:.25:.75
for jj = .25:.25:.75
xind = find(vec==ii);
yind = find(vec==jj);
%disp([xind,yind])
abc(a_tic) = Vnum(xind,yind);
a_tic = a_tic + 1;
end
end

% OUTPUT THE FINITIE DIFFERENCE APPROX. RESULTS

% ---------------------------------------------------------
% CALCULATE THE EXACT SOLUTION
%
% POISSON'S EQUATION WITH HOMOGENEOUS BOUNDARY CONDITIONS

% SOLVED BY SERIES EXPANSION

%
for I =1:NX-1
X = H*I;
for J = 1:NY-1
Y = H*J;
SUM = 0;
for M = 1:10 % TAKE ONLY 10 TERMS OF THE SERIES
FM = M;
for N = 1:10
FN = N;
FACTOR1 = (FM*pi/A)^2 + (FN*pi/B)^2;
FACTOR2 = ( (-1)^(M+N) )*144*A*B/(pi*FM*FN);
FACTOR3 = 1 - (1 - (-1)^N)/B;
FACTOR = FACTOR2*FACTOR3/FACTOR1;
SUM = SUM + FACTOR*sin(FM*pi*X/A)*sin(FN*pi*Y/B);
end
end
VH = SUM;

% LAPLACE'S EQUATION WITH INHOMOGENEOUS BOUNDARY

CONDITIONS
% SOLVED USING THE METHOD OF SEPARATION OF VARIABLES

C1=4*V1/pi;
C2=4*V2/pi;
C3=4*V3/pi;
C4=4*V4/pi;
SUM=0;
for K =1:10 % TAKE ONLY 10 TERMS OF THE SERIES
N=2*K-1;
AN=N;
A1=sin(AN*pi*X/B);
A2=sinh(AN*pi*(A-Y)/B);
A3=AN*sinh(AN*pi*A/B);
TERM1=C1*A1*A2/A3;
B1=sinh(AN*pi*X/A);
B2=sin(AN*pi*Y/A);
B3=AN*sinh(AN*pi*B/A);
TERM2=C2*B1*B2/B3;
D1=sin(AN*pi*X/B);
D2=sinh(AN*pi*Y/B);
D3=AN*sinh(AN*pi*A/B);
TERM3=C3*D1*D2/D3;
E1=sinh(AN*pi*(B-X)/A);
E2=sin(AN*pi*Y/A);
E3=AN*sinh(AN*pi*B/A);
 TERM4=C4*E1*E2/E3;
TERM = TERM1 + TERM2 + TERM3 + TERM4;
SUM=SUM + TERM;
end
VI = SUM;
Vexact(I+1,J+1) = VH + VI;
end
end

%Grab original points a through i

abc2 = zeros(1,9);
a_tic = 1;
vec = [0:H:1];
for ii = .25:.25:.75
for jj = .25:.25:.75
xind = find(vec==ii);
yind = find(vec==jj);
%disp([xind,yind])
abc2(a_tic) = Vexact(xind,yind);
a_tic = a_tic + 1;
end
end

figure(1),
imagesc(flipud(Vnum')),
colorbar
ylabel('y'), xlabel('x')
title('Example 3.4: Poisson PDE')

format short g
disp(' numerical exact')
disp([abc' abc2'])

Fig. 3.12 MATLAB code for Example 3.4.

%**************************************************************
% Using the finite difference method
% This program calculates the characteristic impedance of the transmission
% line shown in Fig. 3.14.
%**************************************************************
clear all; format compact;

% Output:
%
% H NT Zo
% --------------------------------
% 0.25 700 49.05
% 0.1 500 58.074
% 0.05 500 65.817
% 0.05 700 63.103
% 0.05 1000 61.53

H = 0.05;
NT = 1000;

A = 2.5; B = 2.5; D = 0.5; W = 1.0;

ER=2.35;
EO=8.81E-12;
U=3.0E+8;

NX = A/H;
NY = B/H;
ND = D/H;
NW = W/H;
VD = 100.0;

% CALCULATE CHARGE WITH AND WITHOUT DIELECTRIC

ERR = 1.0;
for L=1:2
E1 = EO;
E2 = EO*ERR;

% INITIALIZATION
V = zeros(NX+2,NY+2);

% SET POTENTIAL ON INNER CONDUCTOR (FIXED NODES) EQUAL TO VD

V(2:NW+2,ND+2) = VD;

% CALCULATE POTENTIAL AT FREE NODES

P1 = E1/(2*(E1 + E2));
P2 = E2/(2*(E1 + E2));
for K=1:NT
for I=0:NX-1
for J=0:NY-1
 if( (J==ND)&&(I<=NW) )
%do nothing
elseif (J==ND)
% IMPOSE BOUNDARY CONDITION AT THE INTERFACE
V(I+2,J+2) = 0.25*(V(I+3,J+2) + V(I+1,J+2)) + ...
P1*V(I+2,J+3) + P2*V(I+2,J+1);
elseif(I==0)
% IMPOSE SYMMETRY CONDITION ALONG Y-AXIS
V(I+2,J+2) = (2*V(I+3,J+2) + V(I+2,J+3) + V(I+2,J+1))/4.0;
elseif(J==0)
% IMPOSE SYMMETRY CONDITION ALONG X-AXIS
V(I+2,J+2) = (V(I+3,J+2) + V(I+1,J+2) + 2*V(I+2,J+3))/4.0;
else
V(I+2,J+2) = (V(I+3,J+2)+V(I+1,J+2)+V(I+2,J+3)+V(I+2,J+1))/4.0;
end
end
end
% Animation of calculation
% figure(1),imagesc(V),colorbar,title([num2str(K),'/',num2str(NT)])
% drawnow
end

% NOW, CALCULATE THE TOTAL CHARGE ENCLOSED IN A

% RECTANGULAR PATH SURROUNDING THE INNER CONDUCTOR
IOUT = round((NX + NW)/2);
JOUT = round((NY + ND)/2);
% SUM POTENTIAL ON INNER AND OUTER LOOPS
for K=1:2
SUM = E1*sum(V(3:IOUT+1,JOUT+2)) ...
+ E1*V(2,JOUT+2)/2 + E2*V(IOUT+2,2)/2;
for J=1:JOUT-1
if(J<ND)
SUM = SUM + E2*V(IOUT+2,J+2);
elseif(J==ND)
SUM = SUM + (E1+E2)*V(IOUT+2,J+2)/2;
else
SUM = SUM + E1*V(IOUT+2,J+2);
end
end
if K==1
SV(1) = SUM;
end
IOUT = IOUT - 1;
JOUT = JOUT - 1;
end
SUM = SUM + 2.0*E1*V(IOUT+2,JOUT+2);
 SV(2) = SUM;
Q(L) = abs( SV(1) - SV(2) );
ERR = ER;
end

% FINALLY, CALCULATE Zo
C0 = 4.0*Q(1)/VD;
C1 = 4.0*Q(2)/VD;
Z0 = 1.0/( U*sqrt(C0*C1) );

disp([H,NT,Z0])

Fig. 3.21 MATLAB code for Example 3.6.

%*******************************************************************
% APPLICATION OF THE FINITE-DIFFERENCE TIME-DOMAIN METHOD
% This program involves the penetration of a lossless dielectric SPHERE
% by a plane wave. The program provides in the maximum absolute value of
% Ey and Ez during the final half-wave of time-stepping
% Assumption:
% +y-directed incident wave with components Ez and Hx.
% I,J,K,NN correspond to X,Y,Z, and Time.
% IMAX,JMAX,KMAX are the maximum values of x,y,z
% NNMAX is the total number of timesteps.
% NHW represents one half-wave cycle.
% MED is the number of different uniform media sections.
% JS is the j-position of the plane wave front.

% THIS PROGRAM WAS DEVELOPED BY V. BEMMEL [80]

% AND LATER IMPROVED BY D. TERRY

clear all; format compact; tic

IMAX=19; JMAX=39; KMAX=19;

NMAX=2; NNMAX=500; NHW=40; MED=2; JS=3;
DELTA=3E-3; CL=3.0E8; F=2.5E9;

% Define scatterer dimensions

OI=19.5; OJ=20.0; OK=19.0; RADIUS=15.0;

ER=[1.0,4.0]; % CONSTITUTIVE PARAMETERS

SIG=[0.1,0.0];

% Statement function to compute position w.r.t. center of the sphere

 E0=(1E-9)/(36*pi);
U0=(1E-7)*4*pi;
DT=DELTA/(2*CL);
R=DT/E0;
RA=(DT^2)/(U0*E0*(DELTA^2));
RB=DT/(U0*DELTA);
TPIFDT = 2.0*pi*F*DT;

%********************************************************
% EP # 1 - COMPUTE MEDIA PARAMETERS
%********************************************************

CA = 1-R*SIG./ER;
CB = RA./ER;
CBMRB = CB/RB;

% (i) CALCULATE THE REAL/ACTUAL GRID POINTS

% Initialize the media arrays.Index (M) determines which

% medium each point is actually located in and is used to
% index into arrays which determine the constitutive
% parameters of the medium.There are separate M determining
% arrays for EX, EY, and EZ. These arrays correlate the
% integer values of I,J,K to the actual position within
% the lattice. Computing these values now and storing them in these
% arrays as opposed to computing them each time they are
% needed saves a large amount of computation time.

x = 0:(IMAX+1); y = 0:(JMAX+1); z = 0:(KMAX+1);

[Mx,My,Mz]=ndgrid(x,y,z);

IXMED = (sqrt((Mx-OI+.5).^2+(My-OJ).^2+(Mz-OK).^2)<=RADIUS)+1;
IYMED = (sqrt((Mx-OI).^2+(My-OJ+.5).^2+(Mz-OK).^2)<=RADIUS)+1;
IZMED = (sqrt((Mx-OI).^2+(My-OJ).^2+(Mz-OK+.5).^2)<=RADIUS)+1;

% *************************************************
% STEP # 2 - INITIALIZE FIELD COMPONENTS
% *************************************************
% components for output

EY1 = zeros(1,JMAX+2);
EZ1 = zeros(1,JMAX+2);

EX=zeros(IMAX+2,JMAX+2,KMAX+2,NMAX+1);
EY=zeros(IMAX+2,JMAX+2,KMAX+2,NMAX+1);
 EZ=zeros(IMAX+2,JMAX+2,KMAX+2,NMAX+1);
HX=zeros(IMAX+2,JMAX+2,KMAX+2,NMAX+1);
HY=zeros(IMAX+2,JMAX+2,KMAX+2,NMAX+1);
HZ=zeros(IMAX+2,JMAX+2,KMAX+2,NMAX+1);

% ********************************************************
% STEP # 3 - USE FD/TD ALGORITHM TO GENERATE
% FIELD COMPONENTS
% ********************************************************
% SINCE ONLY FIELD COMPONENTS AT CURRENT TIME (t) AND PREVIOUS

% TWO TIME STEPS ( t-1 AND t-2) ARE REQUIRED FOR COMPUTATION,

% WE SAVE MEMORY SPACE BY USING THE FOLLOWING INDICES

% NCUR is index in for time t
% NPR1 is index in for t-1
% NPR2 is index in for t-2

% NOTES %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%
% *ind03c.m I incremented the time so it goes 1 2 3, instead of 0 1 2

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%

NCUR = 3;
NPR1 = 2;
NPR2 = 1;
for NN = 1: NNMAX % TIME LOOP
if mod(NN,10)==0
disp(['NN = ',num2str(NN)]) % DISPLAY PROGRESS
end
% Next time step - move indices up a notch.
NPR2 = NPR1;
NPR1 = NCUR;
NCUR = mod( NCUR, 3)+1;
for K=0:KMAX % Z LOOP
for J=0:JMAX % Y LOOP
for I=0:IMAX % X LOOP
% (ii)-APPLY SOFT LATTICE TRUNCATION CONDITIONS

%---x=delta/2
if (I==0)
if ((K~=KMAX)&&(K~=0))
 HY(0+1,J+1,K+1,NCUR) = (HY(1+1,J+1,K-1+1,NPR2) +
HY(1+1,J+1,K+1,NPR2)+ HY(1+1,J+1,K+1+1,NPR2))/3;
HZ(0+1,J+1,K+1,NCUR) = (HZ(1+1,J+1,K-1+1,NPR2) +
HZ(1+1,J+1,K+1,NPR2)+ HZ(1+1,J+1,K+1+1,NPR2))/3;

else
if (K==KMAX)
HY(0+1,J+1,KMAX+1,NCUR) = (HY(1+1,J+1,KMAX-
1+1,NPR2)+ HY(1+1,J+1,KMAX+1,NPR2))/2;
HZ(0+1,J+1,K+1,NCUR)=( HZ(1+1,J+1,K-1+1,NPR2)+
HZ(1+1,J+1,K+1,NPR2) )/2;
else
HY(0+1,J+1,K+1,NCUR) = ( HY(1+1,J+1,K+1,NPR2)+
HY(1+1,J+1,K+1+1,NPR2))/2;
HZ(0+1,J+1,0+1,NPR2)=(HZ(1+1,J+1,0+1,NPR2)+
HZ(1+1,J+1,1+1,NPR2))/2;
end
end
end
% ---y=0
if (J==0)
EX(I+1,0+1,K+1,NCUR)=EX(I+1,1+1,K+1,NPR2);
EZ(I+1,0+1,K+1,NCUR)=EZ(I+1,1+1,K+1,NPR2);
else
%---y=ymax
if (J==JMAX)
EX(I+1,JMAX+1,K+1,NCUR)=EX(I+1,JMAX-1+1,K+1,NPR2);
EZ(I+1,JMAX+1,K+1,NCUR)=EZ(I+1,JMAX-1+1,K+1,NPR2);
end
end
%---z=0
if(K==0)
if ((I~=0)&&(I~=IMAX))
EX(I+1,J+1,0+1,NCUR) = (EX(I-1+1,J+1,1+1,NPR2) +
EX(I+1,J+1,1+1,NPR2)+EX(I+1+1,J+1,1+1,NPR2))/3;
EY(I+1,J+1,0+1,NCUR) = (EY(I-1+1,J+1,1+1,NPR2) +
EY(I+1,J+1,1+1,NPR2)+EY(I+1+1,J+1,1+1,NPR2))/3;
else
if (I==0)

EX(0+1,J+1,0+1,NCUR)=(EX(0+1,J+1,1+1,NPR2)+EX(1+1,J+1,1+1,NPR2))/2;

EY(I+1,J+1,0+1,NCUR)=(EY(I+1,J+1,1+1,NPR2)+EY(I+1+1,J+1,1+1,NPR2))/2;
else
 EX(I+1,J+1,0+1,NCUR)=(EX(I-
1+1,J+1,1+1,NPR2)+EX(I+1,J+1,1+1,NPR2))/2;
EY(I+1,J+1,0+1,NCUR)=(EY(I-
1+1,J+1,1+1,NPR2)+EY(I+1,J+1,1+1,NPR2))/2;
end
end
end

% (iii) APPLY FD/TD ALGORITHM

%-----a. HX generation:

HX(I+1,J+1,K+1,NCUR)=HX(I+1,J+1,K+1,NPR1)+RB*(EY(I+1,J+1,K+1+1,NPR1)-
EY(I+1,J+1,K+1,NPR1)+EZ(I+1,J+1,K+1,NPR1)-EZ(I+1,J+1+1,K+1,NPR1));
%-----b. HY generation:

HY(I+1,J+1,K+1,NCUR)=HY(I+1,J+1,K+1,NPR1)+RB*(EZ(I+1+1,J+1,K+1,NPR1)-
EZ(I+1,J+1,K+1,NPR1)+EX(I+1,J+1,K+1,NPR1)-EX(I+1,J+1,K+1+1,NPR1));
%-----c. HZ generation:

HZ(I+1,J+1,K+1,NCUR)=HZ(I+1,J+1,K+1,NPR1)+RB*(EX(I+1,J+1+1,K+1,NPR1)-
EX(I+1,J+1,K+1,NPR1)+EY(I+1,J+1,K+1,NPR1)-EY(I+1+1,J+1,K+1,NPR1));
%---k=kmax ! SYMMETRY
if (K==KMAX)
HX(I+1,J+1,KMAX+1,NCUR)=HX(I+1,J+1,KMAX-1+1,NCUR);
HY(I+1,J+1,KMAX+1,NCUR)=HY(I+1,J+1,KMAX-1+1,NCUR);
end
% -----d. EX generation:
if ((J~=0)&&(J~=JMAX)&&(K~=0))
M = IXMED( I+1, J+1, K+1 );
EX(I+1,J+1,K+1,NCUR) = CA(M)*EX(I+1,J+1,K+1,NPR1) +
CBMRB(M)*(HZ(I+1,J+1,K+1,NCUR)-HZ(I+1,J-1+1,K+1,NCUR)+HY(I+1,J+1,K-
1+1,NCUR)-HY(I+1,J+1,K+1,NCUR));
end
%-----e. EY generation:
if(K~=0)
M = IYMED( I+1, J+1, K+1 );
if I~=0
EY(I+1,J+1,K+1,NCUR)=CA(M)*EY(I+1,J+1,K+1,NPR1) +
CBMRB(M)*(HX(I+1,J+1,K+1,NCUR)-HX(I+1,J+1,K-1+1,NCUR)+HZ(I-
1+1,J+1,K+1,NCUR)-HZ(I+1,J+1,K+1,NCUR));
else
EY(I+1,J+1,K+1,NCUR)=CA(M)*EY(I+1,J+1,K+1,NPR1) +
CBMRB(M)*(HX(I+1,J+1,K+1,NCUR)-HX(I+1,J+1,K-1+1,NCUR)+ 0 -
HZ(I+1,J+1,K+1,NCUR));
end
 end
%-----f. EZ generation:
if ((J~=0)&&(J~=JMAX))

M = IZMED( I+1, J+1, K+1 );

% sig(ext)=0 for Ez only from Taflove[14]
if(M==1)
CAM=1;
else
CAM=CA(M);
end

if I~=0

EZ(I+1,J+1,K+1,NCUR)=CAM*EZ(I+1,J+1,K+1,NPR1)+CBMRB(M)*(HY(I+1,J+1,K
+1,NCUR)-HY(I-1+1,J+1,K+1,NCUR)+HX(I+1,J-1+1,K+1,NCUR)-
HX(I+1,J+1,K+1,NCUR));
else
EZ(I+1,J+1,K+1,NCUR)=CAM*EZ(I+1,J+1,K+1,NPR1)+CBMRB(M)*(HY(I+1,J+1,K
+1,NCUR)- 0 +HX(I+1,J-1+1,K+1,NCUR)-HX(I+1,J+1,K+1,NCUR));
end

% (iv) APPLY THE PLANE-WAVE SOURCE

if (J==JS)
EZ(I+1,JS+1,K+1,NCUR) = EZ(I+1,JS+1,K+1,NCUR)
+sin( TPIFDT*NN );
end
end
%---i=imax+1/2 ! SYMMETRY
if(I==IMAX)
EY(IMAX+1+1,J+1,K+1,NCUR)=EY(IMAX+1,J+1,K+1,NCUR);
EZ(IMAX+1+1,J+1,K+1,NCUR)=EZ(IMAX+1,J+1,K+1,NCUR);
end
%---k=kmax
if(K==KMAX)
EX(I+1,J+1,KMAX+1+1,NCUR)=EX(I+1,J+1,KMAX-1+1,NCUR);
EY(I+1,J+1,KMAX+1+1,NCUR)=EY(I+1,J+1,KMAX-1+1,NCUR);
end
end % X LOOP
% ********************************************************
% STEP # 4 - RETAIN THE MAXIMUM ABSOLUTE VALUES DURING
% THE LAST HALF-WAVE
% ********************************************************
if ( (K==KMAX)&&(NN>(NNMAX-NHW)) )
TEMP = abs( EY(IMAX+1,J+1,KMAX-1+1,NCUR) );
if (TEMP > EY1(J+1) )
 EY1(J+1) = TEMP;
end
TEMP = abs( EZ(IMAX+1,J+1,KMAX+1,NCUR) );
if (TEMP > EZ1(J+1) )
EZ1(J+1) = TEMP;
end
end
end % Y LOOP
end % Z LOOP
end % TIME LOOP
% *******************************************************
toc

figure(3),plot(6:34,EY1(6:34),'.-')
ylabel('Computed |E_y|/|E_i_n_c|')
xlabel('j')
grid on
figure(4),plot(5:34,EZ1(5:34),'.-')
ylabel('Computed |E_z|/|E_i_n_c|')
xlabel('j')
grid on

Fig. 3.31 Computer program for FDTD three-dimensional scattering problem.

%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% MATLAB CODE FOR EXAMPLE 3.9:
% AN AXISYMMETRIC PROBLEM OF AN EARTHED CYLINDER PARTIALLY
FILLED WITH
% CHARGED LIQUID SOLVED USING FINITE DIFFERENCE SCHEME
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

A=1; % Radius of cylindrical tank (meters)

B=1; % Height of liquid in tank
C=1; % Height of gas in tank

ER1=1;ER2=2;EO=8.854E-12; %dielectric parameters

H = 0.05; %spacial step size

NA = A/H; %number of points along A
NB = B/H; %number of points along B
NC = C/H; %number of points along C
NBC = NB + NC; %%number of points along B & C
NMAX = 500; %number of iterations
RHOV = 1E-5;
G = -RHOV/(ER2*EO);
GH2 = G*H^2;

%INITIALIZE - THIS ALSO TAKES CARE OF DIRICHLET CONDITIONS

V = zeros(NA+1,NBC+1); %V(radius, height)

%NOW, APPLY FINITE DIFFERENCE SCHEME

for N = 1:NMAX
for I = 2:NA %step through radius
FM = (2*(I-1) - 1)/(2*(I-1)); %note that indicie starts at 1 instead of 0
FP = (2*(I-1) + 1)/(2*(I-1));

%step through liquid (z+ direction)

for J = 2:NB
V(I,J) = 0.25*( V(I,J-1) + V(I,J+1)...
+ FM*V(I-1,J) + FP*V(I+1,J) - GH2 );
end

%step through gas (z+ direction)

for J = NB+2:NBC
V(I,J) = 0.25*( V(I,J-1) + V(I,J+1)...
+ FM*V(I-1,J) + FP*V(I+1,J) );
end
%ALONG THE GASS-LIQUID INTERFACE
V(I,NB+1) = 0.5*( V(I,NB+2)*ER1/(ER1+ER2)+...
V(I,NB)*ER2/(ER1+ER2) )...
+ 0.25*(FM*V(I-1,NB+1)+FP*V(I+1,NB+1) );
end
%IMPOSE NEUMANN CONDITION ALONG THE Z-AXIS
for J = 2:NBC
V(1,J) = ( 4.0*V(2,J) + V(1,J-1)+V(1,J+1))/6.0;
end
end

% OUTPUT THE POTENTIAL ALONG RHO = 0.5, 0 < Z < 1.0

NR5 = 0.5/H;

radius = 0:H:A;
height = 0:H:(B+C);

figure(1),
subplot(211),
plot(height,V(NR5+1,(0:NBC)+1)/1000.0)
ylabel('V(kV)')
title('Potential distribution in the tank of Fig. 3.38 along \rho=0.5m')
 subplot(212)
plot(radius ,V((0:NA)+1,NB+1)/1000.0)
title('Potential distribution in the tank of Fig. 3.38 along gas-liquid interface')
ylabel('V(kV)')

figure(2),surf(height,radius,V/1000)
ylabel('radius in tank')
xlabel('height in tank')
zlabel('Potential (kV)')

Fig. 3.39 MATLAB code for Example 3.9.

clear all
% Compare MATLAB Bessel function to a customly created Bessel function
% function nc_method is needed by this program

% Inputs:
x = 0.1:0.1:2; % input to bessel function
m = 0; %order of bessel function
n = 6; % Newton-Cotes rule
N = 840; % Newton-Cotes N

% Compare with built in MATLAB

J = besselj(m,x);

% Create custom Bessel function

lwr = 0; %lower limit of integration
uppr= pi; %upper limit of integration
theta = linspace(0,pi,N+1);

for ii = 1:length(x)
% Create function to be integrated
A = cos(x(ii)*sin(theta)-m*theta)/pi;

%Integrate with newton-cotes method

J_m(ii) = nc_method(A,lwr,uppr,n);
end

% Difference between MATLAB and custom Bessel

compare = J-J_m;

% eps is the Spacing of floating point numbers in MATLAB.

% Any number smaller than in magnitude than eps is essentially zero
% Try 1 + eps, this is equal to 1.
v_eps = [eps, eps];
x_eps = [x(1),x(end)];

figure(1),
subplot(211),plot(x,J,x,J_m,':')
title('Comparison between MATLAB and Custom Bessel function')
legend('MATLAB','Custom')
subplot(212),plot(x,compare,'-o',x_eps,v_eps,x_eps,-v_eps)
title('Difference between MATLAB and Custom Bessel')
legend('difference','+eps','-eps')

%figure(2),plot(theta,A)

(a)

Fig. 3.46 (a) Main program for Example 3.10.

function I = nc_method(f,a,b,n)

% I = nc_method_fun(f,a,b,n)
%
% Calculates Newton-Cotes Method Quadrature
%
% Inputs:
% f : data vector from function to be integrated
% a : lower integration limit
% b : upper integration limit
% n : order of approximating polynomial
%
% Output:
% I : Value of Integral
%
% NOTE that m (the number of sections) must be a multiple of the order of
% the approximation polynomial

m = length(f)-1; %Divide the data into m sections

%Below checks if the requirement that m is a multiple of n

if rem(m,n) ~= 0
disp('Error: length(f)-1 must be a multiple of n')
I = NaN; return
end

h = (b-a)/m; %step size.

C = nc_weight (n); %Get Newton-Cotes Number.
I = 0; %intitialize I
for ii = 1:n:m,
ind2 = (ii:ii+n); %select indicies of function
Ai = n*h*f(ind2)*C; %Equation (3.138)
I = I+Ai; %Equation (3.142)
end

function q = nc_weight(n)
% returns in q the weights for the N-point Newton-Cotes quadrature rule.
% N must be between 1 and 8.

if (n<1)
disp('error: newton-cotes order must be at least 1!')
elseif (n==1) % Trapezoidal rule:
q = [1 1]'/2;
elseif (n==2) % Simpson's rule:
q = [1 4 1]'/6;
elseif (n==3) % Simpson's 3/8 rule:
q = [1 3 3 1]'/8;
elseif (n==4)% Boole's rule:
q = [7 32 12 32 7]'/90;
elseif (n==5)
q = [19 75 50 50 75 19]'/288;
elseif (n==6)
q = [41 216 27 272 27 216 41]'/840;
elseif (n==7)
q = [751 3577 1323 2989 2989 1323 3577 751]'/17280;
elseif (n==8)
q = [989 5888 -928 10496 -4540 10496 -928 5888 989]'/28350;
else
disp('error: N must be no more than 8!')
end

(b)

Fig. 3.46 (b) Function for the main program in (a); for Example 3.10.

clear all; format compact; tic

% Output of this program

%W=H N Nx Ny Zo
%1 7 5 2 67.735
%1 11 8 3 64.963
% 0.5 7 5 2 96.758
% 0.5 11 8 3 99.098
% *****************************************************************
% THIS PROGRAM CALCULATES THE CHARACTERISTIC IMPEDANCE
% Zo OF A BOXED MICROSTRIP LINE USING COLLOCATION
% POINT-MATCHING METHOD
% *****************************************************************

X=[5 5 5, 4.0, 2.0, 1.0, 0.5, 0.0, -0.5, -1.0 -1];

Y=[0.0,0.5,1 1 1 1 1 1 1 1 ,0.5];
EO=8.8541E-12;VL=3.0E+8;

% INPUT DATA
% W : Width of conductor
% H : Vertical separation of conductor and ground planes
% V0 : Potential at the conductor
% W1 : Horizontal separation of conductor and ground plane
% NMAX : Number of matching points
% NBC : NO. OF POINTS ON BC OR X-AXIS
% ER : Dielectric parameter

for qq = 1:4
switch qq
case 1
W = 1; NMAX = 7; NBC = 5;
str = sprintf('\n\nW=H N Nx Ny Zo ');
disp(str)
case 2
W = 1; NMAX = 11; NBC = 8;
case 3
W =.5; NMAX = 7; NBC = 5;
case 4
W =.5; NMAX = 11; NBC = 8;
end

H = W;
W1 = 5.0;
ER = 1.0;
V0 = 1.0;
NCD = NMAX - NBC; % NO. OF POINTS ON DC OR Y-AXIS
DX = (W + W1)/NBC;
DY = H/NCD;
EPS = EO*ER;

% CALCULATE R & PHI FOR EACH POINT (X,Y)

for N = 1:NMAX
 R(N) = sqrt( X(N)^2 + Y(N)^2 );
if (X(N))<0
PHI(N) = pi - atan( Y(N)/abs(X(N)) );
elseif X(N)==0
PHI(N) = pi/2;
elseif X(N)>0
PHI(N) = atan( Y(N)/X(N) );
end
end

% CALCULATE MATRICES F(I,J) AND B(I)

for I = 1:NMAX
B(I) = -V0;
M = 0;
for J = 1:NMAX
M = 2*J - 1;
FM = (M)/2.0;
if (I==NMAX)
CC = cos(PHI(I))*cos(PHI(I)*FM);
SS = sin(PHI(I))*sin(PHI(I)*FM);
F(I,J) = ( R(I)^(FM-1.) )*(CC + SS)*FM;
B(I) = 0.0;
else
F(I,J) = ( R(I)^(FM) )*cos(FM*PHI(I));
end
end
end

% DETERMINE THE EXPANSION COEFFICIENTS A(I)

IDM = 30;
F = inv(F);
for I = 1:NMAX
A(I) = 0.0;
for J = 1:NMAX
A(I) = A(I) + F(I,J)*B(J);
end
end

% NOW, CALCULATE CHARGE ON THE X-SIDE, i.e. BC

RHO = 0.0;
YC = H;
XC = -W - DX/2.0;
for I = 1:NBC
 XC = XC + DX;
RC = sqrt(XC^2 + YC^2);
if(XC)<=0
PC = pi - atan(YC/abs(XC));
else
PC = atan(YC/XC);
end

for K = 1:NMAX
FK = (2*K - 1)/2.0;
RRO = sin(PC)*cos(FK*PC) - cos(PC)*sin(FK*PC);
RHO = RHO - A(K)*FK*(RC^(FK-1))*RRO*DX*EPS;
end
end

% NEXT, CALCULATE THE CHARGE ONT THE Y-SIDE, i.e. DC

XC = W1;
YC = -DY/2;
for I = 1:NCD
YC = YC + DY;
RC = sqrt(XC^2 + YC^2);
PC = atan(YC/XC);
for K = 1:NMAX
FK = (2*K - 1)/2;
RRO = cos(PC)*cos(FK*PC) + sin(PC)*sin(FK*PC);
RHO = RHO - A(K)*FK*(RC^(FK-1))*RRO*DY*EPS;
end
end

% CALCULATE THE CHARACTERISTIC IMPEDANCE Zo

Q = 4.0*RHO;
C = abs(Q)/V0;
Z0 = sqrt(ER)/( C*VL );
%disp([C,Z0])
%disp([W N NBC NCD C Z0])
str = sprintf('%3g %3g %3g %3g %g',W, N, NBC, NCD, Z0);
disp(str)
end

Fig. 4.7 Computer program for Example 4.11.

% *****************************************************************
% USING MOMENTS METHOD,
% THIS PROGRAM DETERMINES THE CHARACTERISTIC IMPEDANCE
% OF A STRIP TRANSMISSION LINE
% WITH CROSS-SECTION W X H
% THE STRIPS MAITAINED AT 1 VOLT AND -1 VOLT.
%
% ONE STRIP IS LOCATED ON THE Y = H/2 PLANE WHILE THE OTHER
% IS LOCATED ON THE Y = -H/2 PLANE.
%
% ALL DIMENSIONS ARE IS S.I. UNITS
%
% N IS THE NUMBER OF SUBSECTIONS INTO WHICH EACH STRIP IS
DIVIDED
% *****************************************************************

% FIRST, SPECIFY THE PARAMETERS

NN = [3 7 11 18 39 59];

disp(' ')
disp('n Zo')

for N = NN
CL=3.0e8; % SPEED OF LIGHT IN FREE SPACE
ER=1.0;
EO=8.8541878176E-12;
H=2.0;
W=5.0;

NT=2*N;
DELTA = W/(N);

% SECOND, CALCULATE THE MATCH POINTS

% AND THE COEFFICIENT MATRIX [A]

for K=1:N
X(K) = DELTA*(K - .5);
Y(K) = -H/2.0;
X(K+N) = X(K);
Y(K+N) = H/2.0;
end

FACTOR = DELTA/(2.0*pi*EO);

for I=1:NT
for J=1:NT
if(I==J) %eqn (5.98)
 A(I,J) = -(log(DELTA) - 1.5)*FACTOR;
else
R = sqrt( (X(I) - X(J))^2 + (Y(I) - Y(J))^2 );
A(I,J) = -log(R)*FACTOR;
end
end
end

% NOW DETERMINE THE MATRIX OF CONSTANT VECTOR [B]

for K=1:N
B(K)=1.0;
%The line below was commented!
B(K+N)= -1.0; %rja -this line needed to be un-commented
end

% INVERT MATRIX A(I,J) AND CALCULATE MATRIX RO(N)

% CONSISTING OF THE UNKNOWN ELEMENTS
% ALSO CALCULATE THE TOTAL CHARGE Q,
% THE CAPACITANCE C, AND THE CHARACTERISTIC IMPEDANCE Zo.

NIV=NT;
NMAX=100;
A = inv(A);
for I=1:NT
RO(I)=0.0;
%The line below had an error N instead of NT
for M=1:NT
RO(I)=RO(I) + A(I,M)*B(M);
end
end
SUM=0.0;
for I=1:N
SUM = SUM + RO(I);
end
Q=SUM*DELTA;
VD=2.0;
C=abs(Q)/VD;
ZO = sqrt(ER)/(CL*C);

disp(num2str([N ZO ]))

end

Fig. 5.9 MATLAB program for Example 5.7.

%====================================================
% THIS PROGRAM CALCULATES THE SCATTERING PATTERN
% OF AN ARRAY OF PARALLEL WIRES
% ====================================================

clear; format compact;

thearray = 1; % thearray equals 1 for a plane array and 2 for a semicircular array

%PHIO and THETAO define the axis of propagation

THETAO = pi/2.0; %elevation angle pi/2
LAMBDA=1.0; %wavelength
EO = 1.0; %E field
R = 1.125*LAMBDA; %Radius of semicircle array
K = 2.0*pi/LAMBDA; %Propagation constant
AA = 0.05/K; %wire radius
S = 1.0/K; %wire spacing
H = K*cos(THETAO); %H & G are part of the incident wave definition
G = sqrt( K^2 - H^2 );

% DEFINE WIRE LOCATIONS

if thearray ==1
NN = 15; %Number of wires
PHIO = 40;
X = zeros(1,NN);
Y = linspace(-1,1,NN);
elseif thearray ==2
NN = 30;
PHIO = 0.0;
PHI2 = linspace(-pi/2,pi/2,NN);
X = R*cos(PHI2);
Y = R*sin(PHI2);
end
% CALCULATE RHO
[Mx,My]=meshgrid(X,Y);
RHO = sqrt((Mx-Mx').^2+(My-My').^2)+diag(AA*ones(1,NN));

% CONSTRUCT MATRIX [A]

A = besselh(0,2,G*RHO);

% CONSTRUCT MATRIX [B]

ALPHA = X*sin(THETAO)*cos(PHIO)+ Y*sin(THETAO)*sin(PHIO);
B = EO*exp( -i*K*ALPHA).';
% SOLVE FOR MATRIX[I] CONSISTING OF "MODIFIED CURRENT" OR
CURRENT COEFFICIENTS
A = inv(A);
I = A*B;

% FINALLY, CALCULATE THE SCATTERING PATTERN E(PHI)

PHI = linspace(0,pi,128);
ALP = cos(PHI.')*X + sin(PHI.')*Y;
E = abs(exp( i*G*ALP)*I);

% Plot E(PHI)
if thearray ==2
figure(1),plot(PHI*180/pi,E)
title('Scattering Pattern')
xlabel('\phi (Degrees)')
ylabel('E(\phi)')
grid on
elseif thearray ==1

% ang = PHI*180/pi-90;
% field = fftshift(E);

figure(2),plot(PHI*180/pi,E)
title('Example 5.8 compare to Fig. 5.16')
xlabel('\phi (Degrees)')
ylabel('E(\phi)')
grid on

end

Fig. 5.15 Computer program for Example 5.8.

%ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo
% This program determines the power absorption of bilogical bodies.
% bodies are assumed to have the permeability of freespace
% eo = permittivity of free space
% er = relative permittivity of the body
% f = frequency in Hz, w in rad/s
% sig = conductivity
% xd, yd and zd have the x, y z coordiantes of the cells
% N is the number of cells
% Ei = incident electric field
% Rmn = distance between cell m and n.
% cxpmn and cxqmn = cos(theta_xp_mn) and cos(theta_xq_mn)
% We compute the 9 block matrices separately although one might have a
% function to that and combine them into the G matrix.
% For the diagonal blocks we compute the diagonal elements.
% for the off-diagonal blocks they are zeros.
clear all; close all
R=3e-2; H = R/4; an = H*(0.75/pi)^(1/3);
eo = 8.854e-12; muo = pi*4e-7;
f = 918e6; er = 35; sig=0.7;
w = 2*pi*f; ko = w*sqrt(muo*eo);
N=40; %

xd = [0.5*H*ones(1,8) 1.5*H*ones(1,6) 2.5*H*ones(1,4) 3.5*H 3.5*H ...

0.5*H*ones(1,6) 1.5*H*ones(1,4) 2.5*H*ones(1,2) ...
0.5*H*ones(1,4) 1.5*H*ones(1,2) ...
0.5*H*ones(1,2)];

yd= [0.5*H*ones(1,20) 1.5*H*ones(1,12) 2.5*H*ones(1,6)

3.5*H*ones(1,2)];

zd=[(3.5*H:-H:-3.5*H) (2.5*H:-H:-2.5*H) (1.5*H:-H:-1.5*H) 0.5*H -0.5*H

...
(2.5*H:-H:-2.5*H) (1.5*H:-H:-1.5*H) 0.5*H -0.5*H ...
(1.5*H:-H:-1.5*H) 0.5*H -0.5*H ...
0.5*H -0.5*H];
r=[xd(:) yd(:) zd(:)];

tow=[(sig+j*w*eo*(er-1))*ones(N,1)];

A=-j*w*muo*ko*H^3/4/pi;
C=-2*j*w*muo/3/ko^2*(exp(-j*ko*an)*(1+j*ko*an)-1)-1/3/j/w/eo;
Ei = [86.83*exp(-j*ko*zd(:));0*exp(-j*ko*zd(:));0*exp(-j*ko*zd(:))];

%Gxx
for m=1:N
for n=1:N
if m==n
Gxx(m,n) = C*tow(n)-1;
else
Rmn=norm(r(m,:)-r(n,:));
almn = ko*Rmn;
cxpmn =(r(m,1)-r(n,1))/Rmn;
cxqmn =(r(m,1)-r(n,1))/Rmn;

B = (almn^2-1-j*almn)+cxpmn*cxqmn*(3-almn^2+3*j*almn);
Gxx(m,n)= A* tow(n)*exp(-j*almn)/almn^3*B;
end
end
end
%Gyy
for m=1:N
for n=1:N
if m==n
Gyy(m,n) = C*tow(n)-1;
else
Rmn=norm(r(m,:)-r(n,:));
almn = ko*Rmn;
cxpmn =(r(m,2)-r(n,2))/Rmn;
cxqmn =(r(m,2)-r(n,2))/Rmn;

B = (almn^2-1-j*almn)+cxpmn*cxqmn*(3-almn^2+3*j*almn);
 Gyy(m,n)= A* tow(n)*exp(-j*almn)/almn^3*B;
end
end
end

%Gzz
for m=1:N
for n=1:N
if m==n
Gzz(m,n) = C*tow(n)-1;
else
Rmn=norm(r(m,:)-r(n,:));
almn = ko*Rmn;
cxpmn =(r(m,3)-r(n,3))/Rmn;
cxqmn =(r(m,3)-r(n,3))/Rmn;

B = (almn^2-1-j*almn)+cxpmn*cxqmn*(3-almn^2+3*j*almn);
Gzz(m,n)= A* tow(n)*exp(-j*almn)/almn^3*B;
end
end
end

%Gxy
for m=1:N
for n=1:N
if m~=n
Rmn=norm(r(m,:)-r(n,:));
almn = ko*Rmn;
cxpmn =(r(m,1)-r(n,1))/Rmn;
cxqmn =(r(m,2)-r(n,2))/Rmn;

B = cxpmn*cxqmn*(3-almn^2+3*j*almn);
Gxy(m,n)= A* tow(n)*exp(-j*almn)/almn^3*B;
end
end
end

%Gxz
for m=1:N
for n=1:N
if m~=n
Rmn=norm(r(m,:)-r(n,:));
almn = ko*Rmn;
cxpmn =(r(m,1)-r(n,1))/Rmn;
cxqmn =(r(m,3)-r(n,3))/Rmn;

B = cxpmn*cxqmn*(3-almn^2+3*j*almn);
Gxz(m,n)= A* tow(n)*exp(-j*almn)/almn^3*B;
end
end
end

%Gyx
for m=1:N
for n=1:N
if m~=n
 Rmn=norm(r(m,:)-r(n,:));
almn = ko*Rmn;
cxpmn =(r(m,2)-r(n,2))/Rmn;
cxqmn =(r(m,1)-r(n,1))/Rmn;

B = cxpmn*cxqmn*(3-almn^2+3*j*almn);
Gyx(m,n)= A* tow(n)*exp(-j*almn)/almn^3*B;
end
end
end

%Gyz
for m=1:N
for n=1:N
if m~=n
Rmn=norm(r(m,:)-r(n,:));
almn = ko*Rmn;
cxpmn =(r(m,2)-r(n,2))/Rmn;
cxqmn =(r(m,3)-r(n,3))/Rmn;

B = cxpmn*cxqmn*(3-almn^2+3*j*almn);
Gyz(m,n)= A* tow(n)*exp(-j*almn)/almn^3*B;
end
end
end

%Gzx
for m=1:N
for n=1:N
if m~=n
Rmn=norm(r(m,:)-r(n,:));
almn = ko*Rmn;
cxpmn =(r(m,3)-r(n,3))/Rmn;
cxqmn =(r(m,1)-r(n,1))/Rmn;

B = cxpmn*cxqmn*(3-almn^2+3*j*almn);
Gzx(m,n)= A* tow(n)*exp(-j*almn)/almn^3*B;
end
end
end

%Gzy
for m=1:N
for n=1:N
if m~=n
Rmn=norm(r(m,:)-r(n,:));
almn = ko*Rmn;
cxpmn =(r(m,3)-r(n,3))/Rmn;
cxqmn =(r(m,2)-r(n,2))/Rmn;

B = cxpmn*cxqmn*(3-almn^2+3*j*almn);
Gzy(m,n)= A* tow(n)*exp(-j*almn)/almn^3*B;
end
end
end
G = [Gxx Gxy Gxz;Gyx Gyy Gyz;Gzx Gzy Gzz];
E = -G\Ei;
Nx=[4 11 16 19];
Ny=[4 23 34 39];
Nz= 8:-1:1;;

Exaxis = (abs(E(Nx)).^2+abs(E(N+Nx)).^2+abs(E(2*N+Nx)).^2);
Exaxis = [Exaxis(4:-1:1);Exaxis];
Eyaxis = (abs(E(Ny)).^2+abs(E(N+Ny)).^2+abs(E(2*N+Ny)).^2);
Eyaxis = [Eyaxis(4:-1:1);Eyaxis];
Ezaxis = 2*(abs(E(Nz)).^2+abs(E(N+Nz)).^2+abs(E(2*N+Nz)).^2);

EMAX=max([Exaxis;Eyaxis;Ezaxis]);
Exaxis=Exaxis/EMAX;
Eyaxis=Eyaxis/EMAX;
Ezaxis=Ezaxis/EMAX;
pd=(-2.625:.75:2.625);
plot(pd,Exaxis,'ok');hold on
plot(pd,Eyaxis,'ok','MarkerFaceColor','k');
plot(pd,Ezaxis,'sk');
pdi=(-2.625:5.25/199:2.625);
Exaxisi=interp1(pd,Exaxis,pdi,'spline');
Eyaxisi=interp1(pd,Eyaxis,pdi,'spline');
Ezaxisi=interp1(pd,Ezaxis,pdi,'spline');
plot(pdi,Exaxisi,'k','LineWidth',2);
plot(pdi,Eyaxisi,'-.k','LineWidth',2);
plot(pdi,Ezaxisi,':k','LineWidth',2);
xlabel('cm','FontSize',12)
axis([-3 3 0 1.1]);grid
text(-1.5,.28,'y')
text(-1.5,.45,'z')
text(-1.5,.92,'x')

Fig. 5.26 Computer program for Example 5.11.

%OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO
% FINITE ELEMENT SOLUTION OF LAPLACE'S EQUATION FOR
% TWO-DIMENSIONAL PROBLEMS
% TRIANGULAR ELEMENTS ARE USED
%
% THE UNKNOWN POTENTIALS ARE OBTAINED USING
% ITERATION METHOD
%
% ND = NO. OF NODES
% NE = NO. OF ELEMENTS
% NP = NO. OF FIXED NODES (WHERE POTENTIAL IS PRESCRIBED)
% NDP(I) = NODE NO. OF PRESCRIBED POTENTIAL, I = 1,2,...NP
% VAL(I)) = VALUE OF PRESCRIBED POTENTIAL AT NODE NDP(I)
% NL(I,J) = LIST OF NODES FOR EACH ELEMENT I, WHERE
% LF(I) = LIST OF FREE NODES I = 1,2,...,NF=ND-NP
% J = 1, 2, 3 IS THE LOCAL NODE NUMBER
% CE(I,J) = ELEMENT COEFFICIENT MATRIX
% ER(I) = VALUE OF THE RELATIVE PERMITTIVITY FOR ELEMENT I
% C(I,J) = GLOBAL COEFFICIENT MATRIX
% X(I), Y(I) = GLOBAL COORDINATES OF NODE I
% XL(J), YL(J) = LOCAL COORDINATES OF NODE J = 1,2,3
% V(I) = POTENTIAL AT NODE I
% MATRICES P(I) AND Q(I) ARE DEFINED IN EQ.(6.1.1)

% ***************************************************
% FIRST STEP - INPUT DATA DEFINING GEOMETRY AND
% BOUNDARY CONDITIONS
% ***************************************************
clear;
NI = 50; %! NO. OF ITERATIONS
NE=25; %Number of Elements
ND=21;
NP=15;

X = [0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0 0.2 0.4 0 0.2 0];
Y = [0 0 0 0 0 0 0.2 0.2 0.2 0.2 0.2 0.4 0.4 0.4 0.4 0.6 0.6 0.6 0.8 0.8 1.0];

NDP = [1 2 3 4 5 6 7 11 12 15 16 18 19 20 21];
VAL = [0 0 0 0 0 50 0 100 0 100 0 100 0 100 50];

NL = [ 1 2 7 ;2 8 7;2 3 8; 3 9 8; 3 4 9; 4 10 9; 4 5 10; 5 11 10; 5 6 11;...

7 8 12;8 13 12; 8 9 13;9 14 13;9 10 14; 10 15 14; 10 11 15;...
12 13 16; 13 17 16; 13 14 17; 14 18 17; 14 15 18; 16 17 19; 17 20 19; 17 18 20; 19 20
21];

EO = 1.0E-9/(36.0*pi);
ER = ones(1,NE);
% ***************************************************
% SECOND STEP - EVALUATE COEFFICIENT MATRIX FOR EACH
% ELEMENT AND ASSEMBLE GLOBALLY
% ***************************************************
C = zeros(ND,ND);
for I = 1: NE
% FIND LOCAL COORDINATES XL, YL FOR ELEMENT I
XL = X(NL(I,:));
YL = Y(NL(I,:));

P = [YL(2) - YL(3),YL(3) - YL(1),YL(1) - YL(2)];

Q = [XL(3) - XL(2),XL(1) - XL(3),XL(2) - XL(1)];
AREA = 0.5*abs( P(2)*Q(3) - Q(2)*P(3) );
% DETERMINE COEFFICIENT MATRIX FOR ELEMENT I
 CE = ER(I)*(P'*P+Q'*Q)/(4*AREA);
% ASSEMBLE GLOBALLY - FIND C(I,J)
J=1:3;
L=1:3;
IR = NL(I,J);
IC = NL(I,L);
C(IR,IC) = C(IR,IC) + CE(J,L);
end

% *************************************************
% THIRD STEP - SOLVE THE RESULTING SYSTEM
% ITERATIVELY
% *************************************************
%
% DETERMINE LF - LIST OF FREE NODES
LF = setdiff(1:ND, NDP);
V = zeros(1,ND);
V(NDP) = VAL;
NF = length(LF);

for N = 1:NI
for ii = 1:NF
rowC = C(LF(ii),:);
rowC(LF(ii)) = 0;
V(LF(ii)) = -1/(C(LF(ii),LF(ii)))*rowC*V.';
end
figure(1),stem(V),drawnow
end

disp([{'Node','X','Y','Potential'};num2cell([(1:ND)',X',Y',V'])]);

Fig. 6.10 Computer program for Example 6.2.

clear;
% *****************************************************
% FINITE ELEMENT SOLUTION OF THE WAVE EQUATION
% TRIANGULAR ELEMENTS ARE USED
%
% ND = NO. OF NODES
% NE = NO. OF ELEMENTS
% NL(I,J) = LIST OF NODES FOR EACH ELEMENT I, WHERE
% CE(I,J) = ELEMENT COEFFICIENT MATRIX
% C(I,J) = GLOBAL COEFFICIENT MATRIX
% X(I), Y(I) = GLOBAL COORDINATES OF NODE I
% XL(J), YL(J) = LOCAL COORDINATES OF NODE J = 1,2,3
% MATRICES P(I) AND Q(I) ARE DEFINED IN
% LF(I) = LIST OF FREE NODES
% ALAM(I) = CONTAINS EIGENVALUES

% ***************************************************
% FIRST STEP - INPUT DATA DEFINING GEOMETRY AND
% BOUNDARY CONDITIONS (USE SUBROUTINE GRID)
% ***************************************************

NX = 10;
select = 1; %select=1 for square, select=2 for rectangular

if select == 1 %For Square Waveguide

NY = 1*NX; AA = 1.0; BB = 1.0;
else %For Rectangular Waveguide
NY = 2.0*NX; AA = 1.0; BB = 2.0;
end

DELTAX = AA/NX;
DELTAY = BB/NY;
DX = ones(1,NX)*DELTAX;
DY = ones(1,NY)*DELTAY;

[NE,ND,NP,NL,X,Y,NDP]=gridmesh(NX,NY,DX,DY);

% Plot Mesh - Grab X and Y values of element nodes. Note: inserting NaN
% keeps the plot function from connecting different elements together!
mx = reshape([X([NL,NL(:,1)]),NaN*zeros(NE,1)]',1,NE*5);
my = reshape([Y([NL,NL(:,1)]),NaN*zeros(NE,1)]',1,NE*5);

figure(2),plot(X,Y,'o',mx,my,':'),xlabel('x'), ylabel('y')

% ***************************************************
% SECOND STEP - EVALUATE COEFFICIENT MATRIX FOR EACH
% ELEMENT AND ASSEMBLE GLOBALLY
% ***************************************************

C = zeros(ND,ND);
T = zeros(ND,ND);

for I = 1: NE
% FIND LOCAL COORDINATES XL, YL FOR ELEMENT I
XL = X(NL(I,:));
YL = Y(NL(I,:));
P = [YL(2) - YL(3),YL(3) - YL(1),YL(1) - YL(2)];
Q = [XL(3) - XL(2),XL(1) - XL(3),XL(2) - XL(1)];
 AREA = 0.5*abs( P(2)*Q(3) - Q(2)*P(3) );

% DETERMINE COEFFICIENT MATRIX FOR ELEMENT I

CE = (P'*P+Q'*Q)/(4*AREA);

% ASSEMBLE GLOBALLY - FIND C(I,J) AND T(I,J)

J=1:3; L=1:3;
IR = NL(I,J);
IC = NL(I,L);
C(IR,IC) = C(IR,IC) + CE(J,L);
T(IR,IC) = T(IR,IC) + AREA/12*[2 1 1;1 2 1;1 1 2];
end

% *************************************************
% THIRD STEP - SOLVE THE RESULTING SYSTEM
% *************************************************
% DETERMINE LF(I) - LIST OF FREE NODES

LF = setdiff(1:ND, NDP);
NF = length(LF);

% FROM GLOBAL C AND T, FIND C_ff AND T_ff (f for free nodes)
C_ff = C(LF,LF);
T_ff = T(LF,LF);

A = inv(T_ff)*C_ff;

ALAM = eig(A);
% *************************************************
% FOURTH STEP - OUTPUT THE RESULTS
% *************************************************

Kcalc = sqrt(min(ALAM));

Kc = sqrt( (pi/AA)^2+(pi/BB)^2 );
theerror = 100*(Kcalc-Kc)/Kc;

s1 = sprintf('\n Nx Ne Kcalc %% error');

s2 = sprintf('\n %-3.5g %-3.5g %-6.5g %-3.5g',NX,NE,Kcalc,theerror);

disp([s1,s2])

Fig. 6.11 Computer program for Example 6.3.

function [NE,ND,NP,NL,X,Y,NDP]=gridmesh(NX,NY,DX,DY)
% function [NE,ND,NP,NL,X,Y,NDP]=gridmesh(NX,NY,DX,DY)
% ******************************************************
% THIS PROGRAM DIVIDES A RECTANGULAR DOMAIN INTO
% TRIANGULAR ELEMENTS (NY BY NY NONUNIFORM
% MESH IN GENERAL)
% NX & NY ARE THE NUMBER OF SUBDIVISIONS ALONG X & Y AXES
% NE = NO. OF ELEMENTS IN THE MESH
% ND = NO. OF NODES IN THE MESH
% NP = NO. OF BOUNDARY (PRESCRIBED) NODES
% X(I) & Y(I) ARE GLOBAL COORDINATES OF NODE I
% DX(I) & DY(I) ARE DISTANCES BETWEEN NODES ALONG X & Y AXES
% NL(I,J) IS THE LIST OF NODES FOR ELEMENT I, J = 1, 2, 3 ARE
% LOCAL NUMBERS
% NDP(I) = LIST OF PRESCRIBED NODES I
%
% REF: J. N. REDDY, “AN INTRODUCTION TO THE FINITE ELEMENT
% METHOD.” NEW YORK: MCGRAW-HILL, 1984, P. 436

% CALCULATE NE, ND, AND NP

NE = 2*NX*NY; %Number of elements
NP = 2*(NX + NY); %Number of prescribed nodes (edges)
NX1 = NX+1; %Number of nodes in the X direction
NY1 = NY +1; %Number of nodes in the Y direction
ND = NX1 * NY1; %Number of nodes in mesh

% Determine Node List which follows a regular pattern utilized below.

NL1 = 1:(NX1*NY);
NL1(NX1:NX1:end) = [];

NL3 = (NX+2):(NX+NX1*NY+1);
NL3(NX1:NX1:end) = [];

NL = zeros(NE,3);
NL(1:2:end,1) = NL1;
NL(2:2:end,1) = NL1;
NL(1:2:end,2) = NL3+1;
NL(2:2:end,2) = NL1+1;
NL(1:2:end,3) = NL3;
NL(2:2:end,3) = NL3+1;

% DETERMINE X AND Y (look saw tooth waves)

x1 = [0 cumsum(DX)]'*ones(1,NY1);
X = reshape(x1,1,numel(x1));

y1 = ones(NX1,1)*[0 cumsum(DY)];
Y = reshape(y1,1,numel(y1));
% DETERMINE NDP, [bottom, right, top, left]
NDP = [1:(NX1) , 2*(NX1):(NX1):ND , (ND-1):-1:(ND-NX) , ((ND-2*NX-1)):-(NX1):
(NX+2)];

Fig. 6.13 Program GRID.

function ch6_FIG6_16

fd=inline('ddiff( dunion(drectangle(p,0,5,0,10),drectangle(p,5,8,5,10)) ,dcircle(p,2.5,7.5,

1.5) )','p');

box=[-.1 -.1; 8.1 10.1];

fix=[0 0; 0 5; 0 10; 5 0; 5 5; 5 10; 8 5; 8 10];
[p,t]=distmesh2d(fd,@huniform,0.7,box,fix);
disp('Mesh Complete')

function [p,t]=distmesh2d(fd,fh,h0,bbox,pfix,varargin)
%DISTMESH2D 2-D Mesh Generator using Distance Functions.
% [P,T]=DISTMESH2D(FD,FH,H0,BBOX,PFIX,FPARAMS)
%
% P: Node positions (Nx2)
% T: Triangle indices (NTx3)
% FD: Distance function d(x,y)
% FH: Scaled edge length function h(x,y)
% H0: Initial edge length
% BBOX: Bounding box [xmin,ymin; xmax,ymax]
% PFIX: Fixed node positions (NFIXx2)
% FPARAMS: Additional parameters passed to FD and FH
%
% Example: (Uniform Mesh on Unit Circle)
% fd=inline('sqrt(sum(p.^2,2))-1','p');
% [p,t]=distmesh2d(fd,@huniform,0.2,[-1,-1;1,1],[]);
%
% Example: (Rectangle with circular hole, refined at circle boundary)
% fd=inline('ddiff(drectangle(p,-1,1,-1,1),dcircle(p,0,0,0.5))','p');
% fh=inline('min(4*sqrt(sum(p.^2,2))-1,2)','p');
% [p,t]=distmesh2d(fd,fh,0.05,[-1,-1;1,1],[-1,-1;-1,1;1,-1;1,1]);
%
% See also: MESHDEMO2D, DISTMESHND, DELAUNAYN, TRIMESH.

% Copyright (C) 2004-2006 Per-Olof Persson.

dptol=.001; ttol=.1; Fscale=1.2; deltat=.2; geps=.001*h0; deps=sqrt(eps)*h0;

% 1. Create initial distribution in bounding box (equilateral triangles)
[x,y]=meshgrid(bbox(1,1):h0:bbox(2,1),bbox(1,2):h0*sqrt(3)/2:bbox(2,2));
x(2:2:end,:)=x(2:2:end,:)+h0/2; % Shift even rows
p=[x(:),y(:)]; % List of node coordinates

% 2. Remove points outside the region, apply the rejection method

p=p(feval(fd,p,varargin{:})<geps,:); % Keep only d<0 points
r0=1./feval(fh,p,varargin{:}).^2; % Probability to keep point
p=[pfix; p(rand(size(p,1),1)<r0./max(r0),:)]; % Rejection method
N=size(p,1); % Number of points N

pold=inf; % For first iteration

while 1
% 3. Retriangulation by the Delaunay algorithm
if max(sqrt(sum((p-pold).^2,2))/h0)>ttol % Any large movement?
pold=p; % Save current positions
t=delaunayn(p); % List of triangles
pmid=(p(t(:,1),:)+p(t(:,2),:)+p(t(:,3),:))/3; % Compute centroids
t=t(feval(fd,pmid,varargin{:})<-geps,:); % Keep interior triangles
% 4. Describe each bar by a unique pair of nodes
bars=[t(:,[1,2]);t(:,[1,3]);t(:,[2,3])]; % Interior bars duplicated
bars=unique(sort(bars,2),'rows'); % Bars as node pairs
% 5. Graphical output of the current mesh
trimesh(t,p(:,1),p(:,2),zeros(N,1))
view(2),axis equal,axis off,drawnow
end

% 6. Move mesh points based on bar lengths L and forces F

barvec=p(bars(:,1),:)-p(bars(:,2),:); % List of bar vectors
L=sqrt(sum(barvec.^2,2)); % L = Bar lengths
hbars=feval(fh,(p(bars(:,1),:)+p(bars(:,2),:))/2,varargin{:});
L0=hbars*Fscale*sqrt(sum(L.^2)/sum(hbars.^2)); % L0 = Desired lengths
F=max(L0-L,0); % Bar forces (scalars)
Fvec=F./L*[1,1].*barvec; % Bar forces (x,y components)
Ftot=full(sparse(bars(:,[1,1,2,2]),ones(size(F))*[1,2,1,2],[Fvec,-Fvec],N,2));
Ftot(1:size(pfix,1),:)=0; % Force = 0 at fixed points
p=p+deltat*Ftot; % Update node positions

% 7. Bring outside points back to the boundary

d=feval(fd,p,varargin{:}); ix=d>0; % Find points outside (d>0)
dgradx=(feval(fd,[p(ix,1)+deps,p(ix,2)],varargin{:})-d(ix))/deps; % Numerical
dgrady=(feval(fd,[p(ix,1),p(ix,2)+deps],varargin{:})-d(ix))/deps; % gradient
p(ix,:)=p(ix,:)-[d(ix).*dgradx,d(ix).*dgrady]; % Project back to boundary

% 8. Termination criterion: All interior nodes move less than dptol (scaled)
if max(sqrt(sum(deltat*Ftot(d<-geps,:).^2,2))/h0)<dptol, break; end
end

function h=huniform(p,varargin)
h=ones(size(p,1),1);

(a)

function d=drectangle(p,x1,x2,y1,y2)

d=-min(min(min(-y1+p(:,2),y2-p(:,2)),-x1+p(:,1)),x2-p(:,1));

function d=dcirc(p,xc,yc,r)

d=sqrt((p(:,1)-xc).^2+(p(:,2)-yc).^2)-r;

function d=dunion(d1,d2), d=min(d1,d2);

function d=ddiff(d1,d2), d=max(d1,-d2);

(b)

Fig. 6.16 MATLAB code for automatic mesh generation: (a) main program, (b)
functions that the main function will need.

%=============================================================
% USING THE TLM METHOD, THIS PROGRAM CALCULATES THE RELATIVE
% CURRENT DENSITY JR IN A ROUND COPPER WIRE
% THE EXACT SOLUTION JRE IS ALSO INCLUDED
% =============================================================

clear

% SPECIFY INPUT DATA

F = 1.0E+9;
SIGMA = 5.8E+7;
PIE = 4.0*atan(1.0);
MIU = 4.0*PIE*1.0E-7;
OMEGA = 2.0*PIE*F;
DELTA = sqrt( 2.0/(SIGMA*MIU*OMEGA) );
H = 0.1*DELTA;
A = 4.0*DELTA;
NMAX = A/H;

% INITIALIZE AND CALCULATE RELATIVE CURRENT DENSITY JR USING

TLM

V(1) = 0;
V(2) = 0.1*(SIGMA*( (1) - 0.5 )*(H^2));

for N = 3:NMAX+1
V(N) = (SIGMA*(N-1-0.5)*H^2)/(SIGMA*(N-2-0.5)*H^2) *( V(N-1) - V(N-2) )...
+ i*((OMEGA*MIU/(N-2)))*(SIGMA*(N-1-0.5)*H^2)*V(N-1) + V(N-1);
end

N = 2:NMAX+1;
I = ( V(N) - V(N-1) )./(SIGMA*(N-1-0.5)*H^2);
JR = abs(I/I(NMAX));

% CALCULATE THE CURRENT DENSITY JRE FROM THE EXACT SOLUTION

K = 0:20;
NRO = 1:NMAX;
X = ( (NRO)/(NMAX) )*A*sqrt(2.0)/DELTA;
[K2,X2] = meshgrid(K,X);

FB = ( (X2/2).^(2*K2) )./(factorial(K2).^2);
BEO = FB*exp(i*K*pi/2).';

JRE = abs(BEO);
JRE = JRE/JRE(NMAX); % RELATIVE CURRENT DENSITY

RHO = (1:NMAX)/(NMAX);

%Output Result

disp(sprintf('\n%s %s %s','RADIAL POSITION','TLM J','EXACT J'));

for N=4:4:NMAX
s2 = sprintf('%10g %16g %12g',RHO(N), JR(N), JRE(N));
disp(s2)
end

figure(1),plot(RHO,JR,RHO,JRE)
ylabel('Relative current Density');
xlabel('Radial position')
Fig. 7.7 Computer program for Example 7.1.

clear
% *****************************************************************
% THIS PROGRAM APPLIES THE TLM METHOD TO SOLVE
% ONE-DIMENSIONAL WAVE PROBLEMS. THE SPECIFIC EXAMPLE
% SOLVED HERE IS DESCRIBED AS FOLLOWS:-
%
% THE TEM WAVES ON A 25 x 11 MATRIX
% THE BOUNDARES AT ARE AT X = 2 AND X = 10
% INITIAL IMPULSE EXCITATION IS ALONG Z=4 AT t = 0
% AND SUBSEQUENTLY THIS LINE IS SET TO ZERO. THE GRID
% IS TERMINATED AT Z=25. OUTPUT IS TAKEN AT Z = 14,
% X = 6 FOR Ey AND Hx FOR 100,150,200 ITERATIONS
% VI(IT,I,J,K) -- ARRAY FOR INCIDENT VOLTAGE
% VR(IT,I,J,K) -- ARRAY FOR REFLECTED VOLTAGE
% IT = 1 -- FOR PREVIOUS PULSE VALUE
% =2 -- FOR CURRENT PULSE VALUE
% I,J -- CORRESPOND TO NODE LOCATION (Z,X)
% K = 1..4 -- FOR TERMINALS
% NX -- INDEX OF NODES IN X-DIRECTION
% NZ -- INDEX OF NODES IN Z-DIRECTION
% NX/NZ B,E -- INDEX OF BEGINNING, END NODE
% NX/NX O -- INDEX OF OUTPUT NODE
% GAMMA -- RELFLECTION COEFFICIENT AT THE BOUNDARY C
% DELTA -- MESH SIZE DIVIDED BY LAMBA
% ITRATE -- NO. OF ITERATIONS
% *****************************************************************

VI = zeros(2,25,11,4); VR = zeros(2,25,11,4);
EFI = zeros(1,20); EFR = zeros(1,20);
HFI = zeros(1,20); HFR = zeros(1,20);

CEF = zeros(1,20); CHF = zeros(1,20);

OUT = zeros(20,9);

NXB=2;NXE=10;NZB=4;NZE=24;NT=4;ITRATE=200;
NXO=6;NZO=14;PIE=3.1415927;GAMMA=0;DELTA=.002;

% STEP #1 *********************************************************

% INSERT INITIAL PUSLE EXCITATION ALONG LINE Z = 4

VI(1,NZB+1,NXB:NXE,2) = 1.0;

% STEP #2 *********************************************************

% UsinG EQS. (7.40) TO (7.42), CALCULATE THE

% REFLECTED VOLTAGE AND SUBMIT IT DIRECTLY
% TO THE NEIGHBORING NODE.

K = 1:20; II = 0;
DEL = DELTA:DELTA:DELTA*20;
OUT(:,1) = DEL;

for ITIME = 1: ITRATE

IT = 2;
for I = (NZB+1):NZE
for J = NXB:NXE

VR(IT,I,J,1:NT) = 0.5*sum(VI(IT-1,I,J,1:NT)) - VI(IT-1,I,J,1:NT);

VI(IT,I,J-1,3) = VR(IT,I,J,1);
VI(IT,I-1,J,4) = VR(IT,I,J,2);
VI(IT,I,J+1,1) = VR(IT,I,J,3);
VI(IT,I+1,J,2) = VR(IT,I,J,4);

% STEP #3 *********************************************************

% UsinG EQS. (7.43) AND (7.44), INSERT BOUNDARY CONDITIONS

if (J==NXE), VI(IT,I,NXE,3) = VR(IT,I,NXE,1); end

if (J==NXB), VI(IT,I,NXB,1) = VR(IT,I,NXB,3); end
if (I==NZE), VI(IT,NZE,J,4) = GAMMA*VR(IT,NZE,J,4); end
end
end

% STEP #4 *********************************************************

% UsinG EQS. (7.46) - (7.49), CALCULATE IMPULSE RESPONSE

% OF Ey and Hx AT Z=NZO,X=NXO

EI = sum(VI(IT,NZO,NXO,1:NT)) * 0.5;
HI = VI(IT,NZO,NXO,2) - VI(IT,NZO,NXO,4);

% SUM THE FREQUENCY RESPONSE FOR DIFFERENT VALUES OF MESH-SIZE

DIVIDED
% BY WAVELENGTH
 CEF = CEF + EI*exp(j*2*pi*ITIME*DEL); %Complex E Field
CHF = CHF + HI*exp(j*2*pi*ITIME*DEL); %Complex H Field

% SAVE THE CURRENT PULSE MAGNITUDE FOR NEXT ITERATION

VI(IT-1,:,:,:) = VI(IT,:,:,:);
VR(IT-1,:,:,:) = VR(IT,:,:,:);

IT = ITIME;
if((IT==100)||(IT==150)||(IT==200))
II = II + 2;
% STEP #5 *********************************************************

% CALCULATE MAGNITUDE & ARGUMENT OF IMPEDANCE

OUT(K,II) = abs(CEF)./abs(CHF); % MAGNITUDE
ZARG = CEF./CHF;
OUT(K,II+1) = -atan(imag(ZARG)./real(ZARG));
end
end

% STEP #6 *********************************************************

% CALCULATE EXACT VALUE OF IMPEDANCE [REF. 13]

R2 = 1./(pi*DEL./(asin(sqrt(2).*sin(pi*DEL))));
R3 = tan(21.0*R2*pi.*DEL);
CNUM = R2+R3*i;
CDEM = 1 + j*real(CNUM).*imag(CNUM);
OUT(K,8) = abs(CNUM)./(abs(CDEM).*real(CNUM));
ZARG = CNUM./CDEM;
OUT(K,9) = atan(imag(ZARG)./real(ZARG));
% END

% disp('Compare to Table 7.2')

% disp(OUT(1:9,:))

hdr2 = {'','|Z|','Arg(Z)','|Z|','Arg(Z)','|Z|','Arg(Z)','|Z|','Arg(Z)'};
hdr3 = {'iteration','100','','150','','200','','Exact',''};
disp([hdr3;hdr2;num2cell(OUT(1:9,:))])

Fig. 7.14 Computer program for Example 7.2.

+++++++++++++++++++++++++++++++++++++++++++++++++++++++
 :
:
:

DATA NXB,NXE,NZB,NZE,NT,ITRATE/2,10,2,24,4,200/
DATA NXO,NZO,PIE,GAMMA,DELTA/6,12,3.1415927,0,.02/

:
:
:

C STEP #3 ********************************************
C USING EQS. (7.43) AND (7.44), INSERT BOUNDARY
CONDITIONS
C
IF(J .EQ. NXE)VI(IT,I,NXE,3)= - VR(IT,I,NXE,1)
IF(J .EQ. NXB)VI(IT,I,NXB,1)=VR(IT,I,NXB,1)
IF(I .EQ. NZE)VI(IT,NZE,J,4)=GAMMA*VR(IT,NZE,J,4)

:
:
:

70 DEL = DEL + 0.001

:
:
:

(in Fortran)
Fig. 7.15 Modification in the program in Fig. 7.14 for simulating waveguide
problem in Example 7.3.

clear
% =======================================================
% THIS PROGRAM SOLVES A TYPICAL TWO-DIMENSIONAL
% WAVE PROPOGATION PROBLEM AT STATED BELOW:
% A WAVE GUIDE 6cm X 13cm IS FILLED WITH A
% DIELECTRIC OF RELATIVE PERMITTIVITY EQUAL TO 4.9
% AND CONDUCTIVITY 0.05 MHO/M. THIS GEOMETRY IS
% BEING SIMULATED ON A MATRIX OF 12 X 26 NODES
% THE MATRIX WAS EXCITED AT ALL POINTS ALONG THE
% LINE Z = 1 WITH IMPULSES CORRESPONDING TO Ey.
% THE IMPLUSE FUNCTION OF THE OUTPUT WAS TAKEN FROM
% THE POINT (Z = 6, X=6) AFTER 750 ITERATIONS.
% THE BOUNDARIES WERE SHORT CIRCUITED AT X = 0.5
% AND X = 12.5 AND TERMINATED AT Z = 26.5 IN AN
% OPEN CIRCUIT DISCONTINUITY.
% VI -- ARRAY FOR INCIDENT VOLTAGE
% VR -- ARRAY FOR REFLECTED VOLTAGE
% NX -- INDEX OF NODES IN X-DIRECTION
% NZ -- INDEX OF NODES IN Z-DIRECTION
% RO -- REFLECTION COEFFICIENT
% DELTA -- MESH SIZE (SPACING)
% DELTA -- MESH SIZE DIVIDED BY LAMBDA

VI=zeros(2,27,13,5);
VR=zeros(2,27,13,5);
EFI=zeros(1,50);
EFR=zeros(1,50);
HFI=zeros(1,50);
HFR=zeros(1,50);
NXB=1;
NXE=12;
NZB=1;
NZE=26;
NT=5;
ITRATE=750;
NXI=0;
NZI=1;
NXO=6;
NZO=6;
RoS=-1;
RoC=1;
Eo=8.854E-12;
Er=4.9;
SIGMA=0.05;
DELTA=0.005;

Yo = 4.0 * (Er/2 -1);

Ur = 1.0;
Uo = 4.0 * pi * 1.0E-07;
Go = SIGMA * DELTA * sqrt( Uo/Eo );
Y = 4.0 + Yo + Go;

% SINCE REFLECTION COEFFICIENT, RO, DEPENDS ON

% FREQUENCY, THE ITERATIONS MUST BE REPEATED FOR
% EACH VALUE OF THE MESH-SIZE DIVIDED BY WAVELENGTH.

DELTA = 0.0;
for L = 1:10
DELTA = DELTA + 0.003;

% INITIALIZE ALL NODES (BOTH FREE AND FIXED)

VI = zeros(2,NZE,NXE,NT);
VR = zeros(2,NZE,NXE,NT);

% START THE ITERATION

for ITime = 0: ITRATE

IT = 1;
for I = NZB: NZE
for J = NXB: NXE
SUM = 0.0;
for K = 1:4
SUM = SUM + VI(IT-1+1,I,J,K);
end
SUM = (SUM + Yo*VI(IT-1+1,I,J,5)) * 2/Y;

% INSERT THE INITIAL CONDITION AT I = 1

if ((ITime==0)&&(I==NZI)), SUM = 1.0; end

for K = 1: NT
VR(IT+1,I,J,K) = SUM - VI(IT-1+1,I,J,K);
end

if (J~=NXB), VI(IT+1,I,J-1,3) = VR(IT+1,I,J,1);end

if (I~=NZB), VI(IT+1,I-1,J,4) = VR(IT+1,I,J,2);end
if (J~=NXE), VI(IT+1,I,J+1,1) = VR(IT+1,I,J,3);end
if (I~=NZE), VI(IT+1,I+1,J,2) = VR(IT+1,I,J,4);end
VI (IT+1,I,J,5) = VR(IT+1,I,J,5);

% INSERT THE BOUNDARY CONDITIONS

% FOR THE SHORT CIRCUIT AT X = 0.5
if (J==NXB), VI(IT+1,I,1,1) = RoS*VR(IT+1,I,1,1);end

% FOR THE SHORT CIRCUIT AT X = 12.5

if (J==NXE), VI(IT+1,I,J,3) = RoS*VR(IT+1,I,J,3);end

% FOR THE OPEN CIRCUIT DISCONTINUITY AT Z = 26.5

if (I==NZE), VI(IT+1,I,J,4) = RoC*VR(IT+1,I,J,4);end

end
end

% IN ORDER TO CONSERVE SPACE, THE ARRAYS

% HAVE TO BE UPDATED

VI(IT-1+1,1:NZE,1:NXE,1:NT) = VI(IT+1,1:NZE,1:NXE,1:NT);
 VR(IT-1+1,1:NZE,1:NXE,1:NT) = VR(IT+1,1:NZE,1:NXE,1:NT);

% CALCULATE IMPULSE RESPONCE AT Z=NZO, X=NXO

EI = (sum(VI(IT+1,NZO,NXO,1:4)) + Yo * VI(IT+1,NZO,NXO,5)) * 2.0/Y;
HI = -(VI(IT+1,NZO,NXO,2)-VI(IT+1,NZO,NXO,4));

% SUM THE FREQUENCY RESPONSE (imaginary and

% real parts) FOR DIFFERENT VALUES OF
% MESH-SIZE DIVIDED BY WAVELENGTH

T = (ITime);
EFI(L) = EFI(L)+EI*sin(2.0 * pi * T * DELTA);
EFR(L) = EFR(L)+EI*cos(2.0 * pi * T * DELTA);
HFI(L) = HFI(L)+HI*sin(2.0 * pi * T * DELTA);
HFR(L) = HFR(L)+HI*cos(2.0 * pi * T * DELTA);
OUT(L,1) = DELTA;
if (ITime==ITRATE)
CEF = EFR(L)+EFI(L)*i;
CHF = HFR(L)+HFI(L)*i;
OUT(L,2) = abs(CEF)/abs(CHF);

% CALCULATE ARGUMENT Z

ZARG = CEF/CHF;
XZ = real(ZARG);
YZ = imag(ZARG);
XYZ = YZ / XZ;
OUT(L,3) = -atan(XYZ);
end
end
end

format short g
disp(' DELTA |Z| Arg(Z)')
disp(OUT)

Fig. 7.19 Computer program for Example 7.4.

C********************************************************
C THIS PROGRAM ANLYSIZES THREE-DIMENSIONAL WAVE PROBLEMS
C USING THE TLM METHOD
C THE SPECIFIC EXAMPLE SOLVED HERE IS THE DETERMINATION
C OF THE TM DOMINANT NODE OF RECTANGULAR LOSSLESS CAVITY
C OF DIMENSION 12 X 8 X 6
C E ... E - field
C H ... H - field
C X ... X - component
C Y ... Y - component
C Z ... Z - component
C I ... incident impulse
C R ... reflected impulse
C NX ... number of nodes in X-direction
C NY ... number of nodes in Y-direction
C NZ ... number of nodes in Z-direction
C SHUNT and SERIES ... scatterring matrix
C DELTA = DELTAL/LAMDA (FREE SPACE)

IMPLICIT REAL*8(A,B,D-H,O-Z),COMPLEX*8(C),
1 INTEGER*2(I-N)
DIMENSION EXR(2,7,9,13,5), EXI(2,7,9,13,5),
1 EYR(2,7,9,13,5), EYI(2,7,9,13,5),
2 EZR(2,7,9,13,5), EZI(2,7,9,13,5),
3 HXR(2,7,9,13,5), HXI(2,7,9,13,5),
4 HYR(2,7,9,13,5), HYI(2,7,9,13,5),
5 HZR(2,7,9,13,5), HZI(2,7,9,13,5),
6 EX(2,1000),EY(2,1000),EZ(2,1000),
7 HX(2,1000),HY(2,1000),HZ(2,1000),
8 OUT(12,1000),SHUNT(5,5),SERIES(5,5)

COMMON / PART1 /NT,NXB,NXE,NYB,NYE,NZB,NZE,IT

SQMAG(CM) = ( CABS(CM) ) ** 2
DATA ITRATE,NXB,NXE,NYB,NYE,NZB,NZE,NT,DELTAL
1 /1000, 1, 12, 1, 8, 1, 6, 4, 1.0/
DATA NXI,NYI,NZI,NXO,NYO,NZO,RoH,NUMPT,DEL
1 /11, 7, 0, 2, 2, 2,-1.0,1000,0.0001/
DATA PI, Eo, Er, Ur,SIGMA,DELTA
1 /3.1415927,8.854E-12,1.0,1.0,0.0,0.01 /
DATA SERIES /-1.0,1.0,1.0,-1.0,-1.0,
1 1.0,-1.0,-1.0,1.0,1.0,
2 1.0,-1.0,-1.0,1.0,1.0,
3 -1.0,1.0,1.0,-1.0,-1.0,
4 -1.0,1.0,1.0,-1.0,-1.0/
DATA SHUNT / 1.0,1.0,1.0,1.0,1.0,
1 1.0,1.0,1.0,1.0,1.0,
2 1.0,1.0,1.0,1.0,1.0,
3 1.0,1.0,1.0,1.0,1.0,
4 1.0,1.0,1.0,1.0,1.0/

Uo = 4.0 * PI * 1.0E-07
Yo = 4.0 * (Er - 1.0)
Zo = 4.0 * (Ur - 1.0)
Go = SIGMA * DELTAL * SQRT(Uo/Eo)
Y = 4.0 + Yo + Go
Z = 4.0 + Zo
DO 5 I = 1,5
SHUNT(I,5) = SHUNT(I,5) * Yo
SERIES(5,I) = SERIES(5,I) * Zo
5 CONTINUE
C
C INITIALIZE ALL THE NODES
C
DO 10 IT = 1,2
DO 10 K = NZB, NZE
 DO 10 J = NYB, NYE
DO 10 I = NXB, NXE
DO 10 L = 1, NT
EXI(IT,K,J,I,L) = 0.0
EYI(IT,K,J,I,L) = 0.0
EZI(IT,K,J,I,L) = 0.0
HXI(IT,K,J,I,L) = 0.0
HYI(IT,K,J,I,L) = 0.0
HZI(IT,K,J,I,L) = 0.0
EXR(IT,K,J,I,L) = 0.0
EYR(IT,K,J,I,L) = 0.0
EZR(IT,K,J,I,L) = 0.0
HXR(IT,K,J,I,L) = 0.0
HYR(IT,K,J,I,L) = 0.0
10 HZR(IT,K,J,I,L) = 0.0
C
C INSERT THE INITIAL EXITATION
C
DO 15 K = NZB, NZE
DO 15 L = 1, 4
EZR(2,K,NYI,NXI,L) = 1.0
15 CONTINUE
IT = 2
C
C START ITERATION
C
DO 60 ITime = 1, ITRATE
IF(ITIME .EQ. 1) GOTO 25
C
C COMPUTE THE REFLECTED E-FIELD AT ALL NODES
C
ISIGN = -1
CALL COMPUTE(EXR,EYR,EZR,SHUNT
1 ,Y,EXI,EYI,EZI,ISIGN)
C
C RE-INITIALIZE MATRIX AND FORCE TANGENTIAL
C E-FIELD TO ZERO AT BOUNDARY
C
DO 20 K = NZB, NZE
DO 20 J = NYB, NYE
DO 20 I = NXB, NXE
DO 20 L = 1, NT
HXI(IT-1,K,J,I,L) = 0.0
HYI(IT-1,K,J,I,L) = 0.0
HZI(IT-1,K,J,I,L) = 0.0
IF(I .EQ. NXB) THEN
EYR(IT,K,J,I,L) = 0.0
EZR(IT,K,J,I,L) = 0.0
HXR(IT,K,J,I,L) = 0.0
END IF
IF(J .EQ. NYB)THEN
EXR(IT,K,J,I,L) = 0.0
EZR(IT,K,J,I,L) = 0.0
HYR(IT,K,J,I,L) = 0.0
END IF
IF(K .EQ. NZB)THEN
EXR(IT,K,J,I,L) = 0.0
 EYR(IT,K,J,I,L) = 0.0
HZR(IT,K,J,I,L) = 0.0
END IF
20 CONTINUE
25 CONTINUE
DO 30 K = NZB, NZE
DO 30 J = NYB, NYE
DO 30 I = NXB, NXE
IF(J .NE. NYB)HZI(IT-1,K,J-1,I,4)=EXR(IT,K,J,I,1)
IF(K .NE. NZB)HYI(IT-1,K-1,J,I,4)=EXR(IT,K,J,I,2)
HZI(IT-1,K,J,I,2) =EXR(IT,K,J,I,3)
HYI(IT-1,K,J,I,2) =EXR(IT,K,J,I,4)
IF(I .NE. NXB)HZI(IT-1,K,J,I-1,3)=EYR(IT,K,J,I,1)
IF(K .NE. NZB)HXI(IT-1,K-1,J,I,4)=EYR(IT,K,J,I,2)
HZI(IT-1,K,J,I,1) =EYR(IT,K,J,I,3)
HXI(IT-1,K,J,I,2) =EYR(IT,K,J,I,4)
IF(I .NE. NXB)HYI(IT-1,K,J,I-1,3)=EZR(IT,K,J,I,1)
IF(J .NE. NYB)HXI(IT-1,K,J-1,I,3)=EZR(IT,K,J,I,2)
HYI(IT-1,K,J,I,1) =EZR(IT,K,J,I,3)
HXI(IT-1,K,J,I,1) =EZR(IT,K,J,I,4)
30 CONTINUE
C
C INSERT THE BOUNDARY CONDITIONS AT ALL BOUNDARIES
C
DO 35 K = NZB, NZE
DO 35 J = NYB, NYE
DO 35 I = NXB, NXE
IF(I .EQ. NXB)THEN
HYI(IT-1,K,J,I,1) = RoH * HYR(IT,K,J,I,1)
HZI(IT-1,K,J,I,1) = RoH * HZR(IT,K,J,I,1)
END IF
IF(I .EQ. NXE)THEN
HZI(IT-1,K,J,I,3) = RoH * HZR(IT,K,J,I,3)
HYI(IT-1,K,J,I,3) = RoH * HYR(IT,K,J,I,3)
END IF
IF(J .EQ. NYB)THEN
HXI(IT-1,K,J,I,1) = RoH * HXR(IT,K,J,I,1)
HZI(IT-1,K,J,I,2) = RoH * HZR(IT,K,J,I,2)
END IF
IF(J .EQ. NYE)THEN
HXI(IT-1,K,J,I,3) = RoH * HXR(IT,K,J,I,3)
HZI(IT-1,K,J,I,4) = RoH * HZR(IT,K,J,I,4)
END IF
IF(K .EQ. NZB)THEN
HXI(IT-1,K,J,I,2) = RoH * HXR(IT,K,J,I,2)
HYI(IT-1,K,J,I,2) = RoH * HYR(IT,K,J,I,2)
END IF
IF(K .EQ. NZE)THEN
HXI(IT-1,K,J,I,4) = RoH * HXR(IT,K,J,I,4)
HYI(IT-1,K,J,I,4) = RoH * HYR(IT,K,J,I,4)
END IF
35 CONTINUE
C
C COMPUTE THE H-FIELDS AT ALL THE NODES
C
ISIGN = 1
CALL COMPUTE(HXR,HYR,HZR,SERIES,
 1 Z,HXI,HYI,HZI,ISIGN)
C
C RE-INITIALIZE ALL THE NODES
C
DO 40 K = NZB, NZE
DO 40 J = NYB, NYE
DO 40 I = NXB, NXE
DO 40 L = 1, NT
EXI(IT-1,K,J,I,L) = 0.0
EYI(IT-1,K,J,I,L) = 0.0
EZI(IT-1,K,J,I,L) = 0.0
40 CONTINUE
DO 45 K = NZB, NZE
DO 45 J = NYB, NYE
DO 45 I = NXB, NXE
EZI(IT-1,K,J,I,4) =HXR(IT,K,J,I,1)
EYI(IT-1,K,J,I,4) =HXR(IT,K,J,I,2)
IF(J .NE. NYE)EZI(IT-1,K,J+1,I,2)=HXR(IT,K,J,I,3)
IF(K .NE. NZE)EYI(IT-1,K+1,J,I,2)=HXR(IT,K,J,I,4)
EZI(IT-1,K,J,I,3) =HYR(IT,K,J,I,1)
EXI(IT-1,K,J,I,4) =HYR(IT,K,J,I,2)
IF(I .NE. NXE)EZI(IT-1,K,J,I+1,1)=HYR(IT,K,J,I,3)
IF(K .NE. NZE)EXI(IT-1,K+1,J,I,2)=HYR(IT,K,J,I,4)
EYI(IT-1,K,J,I,3) =HZR(IT,K,J,I,1)
EXI(IT-1,K,J,I,3) =HZR(IT,K,J,I,2)
IF(I .NE. NXE)EYI(IT-1,K,J,I+1,1)=HZR(IT,K,J,I,3)
IF(J .NE. NYE)EXI(IT-1,K,J+1,I,1)=HZR(IT,K,J,I,4)
45 CONTINUE
C
C CALCULATE THE IMPULSE RESPONSE AT NXO,NYO,NZO
C
EXT = 0.0
EYT = 0.0
EZT = 0.0
HXT = 0.0
HYT = 0.0
HZT = 0.0
DO 50 L = 1, NT
EXT = EXT + EXI(IT-1,NZO,NYO,NXO,L) * (2.0/Y)
EYT = EYT + EYI(IT-1,NZO,NYO,NXO,L) * (2.0/Y)
EZT = EZT + EZI(IT-1,NZO,NYO,NXO,L) * (2.0/Y)
HXT = HXT + HXI(IT-1,NZO,NYO,NXO,L) * (2.0/Y)
HYT = HYT + HYI(IT-1,NZO,NYO,NXO,L) * (2.0/Y)
50 HZT = HZT + HZI(IT-1,NZO,NYO,NXO,L) * (2.0/Y)
C
C SUM THE FREQUENCY RESPONSE(imaginary and real
C parts) FOR DIFFERENT VALUES OF MESH-SIZE DIVIDED
C BY WAVELENGTHBUT FIRST CONVOLVE IMPULSE RESPONSE
C WITH HANNING PROFILE
C
DINCR = DELTA
DO 55 L = 1, NUMPT
T = DFLOAT(ITIME)
AMT = DFLOAT(ITRATE)
EXTH = EXT * (1.0 + DCOS(PI * T/AMT)) * 0.5
EYTH = EYT * (1.0 + DCOS(PI * T/AMT)) * 0.5
EZTH = EZT * (1.0 + DCOS(PI * T/AMT)) * 0.5
 HXTH = HXT * (1.0 + DCOS(PI * T/AMT)) * 0.5
HYTH = HYT * (1.0 + DCOS(PI * T/AMT)) * 0.5
HZTH = HZT * (1.0 + DCOS(PI * T/AMT)) * 0.5
EX(1,L) = EX(1,L) + EXTH * DCOS(2.0*PI*T*DINCR)
EX(2,L) = EX(2,L) + EXTH * DSIN(2.0*PI*T*DINCR)
EY(1,L) = EY(1,L) + EYTH * DCOS(2.0*PI*T*DINCR)
EY(2,L) = EY(2,L) + EYTH * DSIN(2.0*PI*T*DINCR)
EZ(1,L) = EZ(1,L) + EZTH * DCOS(2.0*PI*T*DINCR)
EZ(2,L) = EZ(2,L) + EZTH * DSIN(2.0*PI*T*DINCR)
HX(1,L) = HX(1,L) + HXTH * DCOS(2.0*PI*T*DINCR)
HX(2,L) = HX(2,L) + HXTH * DSIN(2.0*PI*T*DINCR)
HY(1,L) = HY(1,L) + HYTH * DCOS(2.0*PI*T*DINCR)
HY(2,L) = HY(2,L) + HYTH * DSIN(2.0*PI*T*DINCR)
HZ(1,L) = HZ(1,L) + HZTH * DCOS(2.0*PI*T*DINCR)
HZ(2,L) = HZ(2,L) + HZTH * DSIN(2.0*PI*T*DINCR)
OUT(1,L) = DINCR
55 DINCR = DINCR + DEL
60 CONTINUE
DO L = 1, NUMPT
CEX = CMPLX(EX(1,L),EX(2,L))
CEY = CMPLX(EY(1,L),EY(2,L))
CEZ = CMPLX(EZ(1,L),EZ(2,L))
CHX = CMPLX(HX(1,L),HX(2,L))
CHY = CMPLX(HY(1,L),HY(2,L))
CHZ = CMPLX(HZ(1,L),HZ(2,L))
OUT(2,L) = CABS(CEX)
OUT(3,L) = CABS(CEY)
OUT(4,L) = CABS(CEZ)
OUT(5,L) = CABS(CHX)
OUT(6,L) = CABS(CHY)
OUT(7,L) = CABS(CHZ)
OUT(8,L)=SQRT(SQMAG(CEX)+SQMAG(CEY)+SQMAG(CEZ))
OUT(9,L)=SQRT(SQMAG(CHX)+SQMAG(CHY)+SQMAG(CHZ))
END DO
C
C PICK THE MODES
C
DO 65 L = 2, 9
DO 65 K = 1, NUMPT
IF(K .NE. 1 .AND. K .NE. NUMPT) THEN
IF(OUT(L,K) .GT. OUT(L,K-1) .AND.
1 OUT(L,K) .GT. OUT(L,K+1)) THEN
WRITE(6,80) L,K,(OUT(J,K),J=1,9)
END IF
END IF
65 CONTINUE
C
C WRITE OUT DATA FOR PLOTTING
C
DO 70 L = 2, 9
DO 70 K = 1, NUMPT
WRITE(6,75) K, OUT(1,K), OUT(L,K)
70 CONTINUE
75 FORMAT(I5,4F15.8)
80 FORMAT(2I5,10F8.4)
STOP
END
C*************************************************
C THIS SUBROUTINE COMPUTES THE REFLECTED PULSES
C
SUBROUTINE COMPUTE(AXR,AYR,AZR,SCATER,
1 W,AXI,AYI,AZI,ISIGN)
IMPLICIT REAL*8(A-H,O-Z), INTEGER*2(I-N)
DIMENSION AXR(2,NZE,NYE,NXE,NT),
1 AYR(2,NZE,NYE,NXE,NT),SCATER(5,5),
2 AZR(2,NZE,NYE,NXE,NT),AXI(2,NZE,NYE,NXE,NT),
3 AYI(2,NZE,NYE,NXE,NT),AZI(2,NZE,NYE,NXE,NT)

COMMON / PART1 /NT,NXB,NXE,NYB,NYE,NZB,NZE,IT

DO 20 K = NZB, NZE
DO 20 J = NYB, NYE
DO 20 I = NXB, NXE
C INSERT FEW LINES HERE FOR INHOMOGENEOUS CAVITY
DO 15 L = 1, NT
AXRS = 0.0
AYRS = 0.0
AZRS = 0.0
DO 10 M = 1, NT
AXRS = AXRS+2./W*SCATER(L,M)*AXI(IT-1,K,J,I,M)
AYRS = AYRS+2./W*SCATER(L,M)*AYI(IT-1,K,J,I,M)
AZRS = AZRS+2./W*SCATER(L,M)*AZI(IT-1,K,J,I,M)
10 CONTINUE
AXR(IT,K,J,I,L)=AXRS+ISIGN*AXI(IT-1,K,J,I,L)
AYR(IT,K,J,I,L)=AYRS+ISIGN*AYI(IT-1,K,J,I,L)
AZR(IT,K,J,I,L)=AZRS+ISIGN*AZI(IT-1,K,J,I,L)
15 CONTINUE
20 CONTINUE
RETURN
END

(in Fortran)
Fig. 7.26 Computer program for Example 7.5.

++++++++++++++++++++++++++++++++++++++++++++++++++++++

:
:
:

C*************************************************
C THIS SUBROUTINE COMPUTES THE REFLECTED PULSES
C
SUBROUTINE COMPUTE(AXR,AYR,AZR,SCATER,
1 WW,AXI,AYI,AZI,ISIGN)

:
:
:
C INSERT FEW LINES HERE FOR INHOMOGENEOUS CAVITY
IF ((I.GE.9).AND.(I.LE.13)
1 .AND.(ISIGN.EQ.-1)) THEN
NT = 5
W = WW
ELSE
NT = 4
W = 4.0
END IF

:
:
:

(in Fortran)
Fig. 7.28 Modification in the function COMPUTE for the inhomogeneous cavity of
Fig. 7.27.

000000000000000000000000000000000

for k = 1:100
r = rand;
theta(k) = acos( 1 – 2*r);
end;

Fig. 8.2 Random variable generator; for Example 8.1.

0000000000000000000000000000000

% Integration using crude Monte Carlo

% and antithetic methods
%
% Only few lines need be changed to use this
% program for any multi-dimensional integration
%
% the function fun.m is to be on a separate file

a =0; b= 1.0; c = 0; % limits of integration

d=2*pi;
alpha = 5;
nrun = 10000;
sum1 = 0; sum2 = 0;
for i=1:nrun
u1 = rand;
 u2 = rand;
x1 = a + (b-a)*u1;
x2 = c + (d-c)*u2;
x3 = (b-a)*(1-u1);
x4 = (d-c)*(1-u2);
sum1 = sum1 + fun(x1,x2);
sum2 = sum2 + fun(x1,x2) + fun(x1,x4) + fun(x3,x2) +
fun(x3,x4);
end
area1 = (b-a)*(d-c)*sum1/nrun
area2 = (b-a)*(d-c)*sum2/(4*nrun)
(a)

function y=fun(rho,phi)
alpha = 5;
y=rho*exp(j*alpha*rho*cos(phi));

(b)

Fig. 8.3 MATLAB program for Monte Carlo evaluation of a two-dimensional

integral: (a) main program, (b) function fun.m kept in a separate file.

% Using fixed random walk MCM to solve a potential

% problem involving Laplace's equation; for Example 8.5

nrun = 2000; % no. of runs, N

delta = 0.05; % step size
A=1.0; B=1.0;
v1=0; v2=0; v3=100; v4=0;
xo=0.5;
yo=0.5;
io=xo/delta; % location at which U is to be determined
jo=yo/delta;
imax=A/delta;
jmax=B/delta;
sum=0;
ms = 0; % no. of steps before reaching the boundary
m1 = 0; % no. of walks terminating at v1
m2 = 0; % no. of walks terminating at v2
m3 = 0; % no. of walks terminating at v3
m4 = 0; % no. of walks terminating at v4
for k=1:nrun
i=io;
j=jo;
while i<=imax & j<=jmax
 ms = ms + 1;
r=rand; %random number between 0 and 1
if (r >= 0.0 & r <= 0.25)
i=i+1;
end
if (r >= 0.25 & r <= 0.5)
i=i-1;
end
if (r >= 0.5 & r <= 0.75)
j=j+1;
end
if (r >= 0.75 & r <= 1.0)
j=j-1;
end
% check if (i,j) is on the boundary
if(i == 0)
sum=sum+ v4;
m4 = m4 +1;
break;
end
if(i == imax)
sum=sum+ v2;
m2 = m2 +1;
break;
end
if(j == 0)
sum=sum+ v1;
m1 = m1 + 1;
break;
end
if(j == jmax)
sum=sum+ v3;
m3 = m3 + 1;
break;
end
end % while
end
v=sum/nrun
m=ms/nrun

Fig. 8.13 MATLAB code for Example 8.5.

% Using floating random walk MCM

% to solve Laplace's equation; for Example 8.6
nrun = 500;
ntrials = 5; % no. of trials
tol=0.005; % tolerance
xo = 1.5; % location at which potential
yo = 0.5; % is to be determined
for n=1:ntrials
sum = 0.0;
m=0; % no. of steps before reaching the boundary
for k=1:nrun
x=xo;
y=yo;
while x >=0 & x <=2 & y >=0 & y<=2.5
u=rand; % generate a random no. and move to the next
point
phi=2.0*pi*u;
% find the shortest distance r;
rc = sqrt( (x-1)^2 + (y-1)^2 );
if (x >1 & x < 2 & y >0 & y < 1)
rx = min(x-1,2-x);
r = min(rx,y);
end
if (x >0 & x < 1 & y >1 & y < 2.5)
rx = min(x,y-1);
r = min(rx,2.5-y);
end
if (x >1 & x < 2 & y >1 & y < 2.5)
rx = min(2-x,2.5-y);
r = min(rx,rc);
end
x=x+r*cos(phi);
y=y+r*sin(phi);
m=m+1;
% check if (x,y) is on the boundary
if ( x < (1+tol) & y < (1+tol) ) % corner
sum=sum +0;
break;
end
if ( x >= (2-tol) )
sum=sum +30;
break;
end
if ( y >= (2.5-tol) )
sum=sum +20;
break;
end
if ( x < tol & y >1 & y < 2.5)
 sum=sum +0;
break;
end
if ( x > tol & x < (1-tol) & y <= (1+tol) )
sum=sum +0;
break;
end
if ( x < (1+tol) & y >= tol & y <= 1 )
sum=sum +0;
break;
end
if ( x >= (1+tol) & x <= (2.5-tol) & y < tol )
sum=sum +0;
break;
end
end %while
end % nrun
vv(n)=sum/nrun;
steps(n)=m/nrun;
end % ntrials
% find the mean value of V and mean no. of steps
sum1 = 0.0;
sum2 = 0.0;
for n=1:ntrials
sum1 = sum1 + vv(n);
sum2 = sum2 + steps(n);
end
vmean=sum1/ntrials
stepm=sum2/ntrials
% calculate error
sum3=0.0;
for n=1:ntrials
sum3 = sum3 + ( vv(n) - vmean )^2;
end
std=sqrt( sum3/(ntrials-1) );
error = std*2.776/sqrt(ntrials)

Fig. 8.15 MATLAB code for Example 8.6.

% EXAMPLE 9.1 SOLVED USING METHOD OF LINES

AA = 1;
BB = 1;
N = 15;
% DETERMINE VECTOR ALPHA
H = AA/(N+1);
LAM = 2*sin ((1:N)*pi*0.5/(N+1) );
ALPHA = LAM/H;
% CALCULATE THE TRANSFORMATION MATRIX AND COEFFICIENT B
K = sqrt (2/(N+1));
T = zeros(N,N);
for I=1:N
for J=1:N
T(I,J) = K*sin (I*J*pi/(N+1));
end
end
V = 100*ones(N,1);
C = inv(T)*V;
A = ALPHA';
B = C./sinh(BB*A);
% CALCULATE V AT THE GIVEN POINTS
V1 = 0; V2 = 0; V3 = 0;
for K=1:N
V1 = V1 + T(4,K)*B(K)*sinh(ALPHA(K)*0.75);
V2 = V2 + T(8,K)*B(K)*sinh(ALPHA(K)*0.5);
V3 = V3 + T(12,K)*B(K)*sinh(ALPHA(K)*0.25);
end
diary
V1, V2, V3
diary off

Fig. 9.2 MATLAB code for Example 9.1.

%++++++++++++++++++++++++++++++++++++++++++++++++
% EXAMPLE 9.3 SOLVED USING METHOD OF LINES

% OUR OBJECTIVE IS TO DETERMINE THE POTENTIAL

% DISTRIBUTION IN A COAXIAL CABLE OF INNER RADIUS a
% AND OUTER RADIUS a ASSUMING A POTENTIAL DIFFERENCE OF Vo

a= 0.01; b = 0.02; Vo = 100;

N = 15;
h = 2*pi/N;
K = 1/sqrt (N);
% COMPUTE THE TRANSFORMATION MATRIX T AND MIU
T = zeros(N,N);
miu = zeros(N,1);
for I=1:N
miu(I) = 2*sin (I*h*0.5 )/h;
for J=1:N
alpha = I*J*h;
T(I,J) = K*( cos(alpha) + sin(alpha) );
end
end
% CALCULATE THE MATRIX D IN EQS. (9.3.2) AND (9.3.3)
D = zeros(2*N,2*N);
for i=1:2*N
for j=1:2*N
%Do the upper part of the matrix
 if (i <= N & j < N)
D(i,j) = T(i,j)*a^miu(j);
end
if (i <= N & j == N)
D(i,j) = T(i,j);
end
if (i <= N & j > N & j < 2*N)
D(i,j) = T(i,j-N)*a^(-miu(j-N));
end
if (i <= N & j == 2*N)
D(i,j) = T(i,j-N)*log(a);
end
%Now do the lower part of the matrix
if (i > N & i <= 2*N & j < N)
D(i,j) = T(i-N,j)*b^miu(j);
end
if (i > N & i <= 2*N & j == N)
D(i,j) = T(i-N,j);
end
if (i > N & i <= 2*N & j > N & j < 2*N)
D(i,j) = T(i-N,j-N)*b^(-miu(j-N));
end
if (i > N & i <= 2*N & j == 2*N)
D(i,j) = T(i-N,j-N)*log(b);
end
end
end
% DETERMINE THE BOUNDARY POTENTIAL MATRIX
% AND THE COEFFICIENT MATRIX
F = zeros(2*N,1);
for i=1:2*N
if i > N
F(i) = Vo;
end
end
C = inv(D)*F;
% WITH THE COEFFICIENTS DETERMINED,
% NOW FIND TRANSFORMED POTENTIAL U
% AND FINALLY DETERMINE THE POTENTIAL V USING EQ. (9.30)
% WE MAY SELECT ANY VALUE OF phi, say, phi = 0^o, i.e. i=1
%del = (a -b)/M;
%rho = a:del:b;
rho = 0.01:0.001:0.02;
M = 10; % no. of divisions along rho
Vmol = zeros(M+1,1);
for k=1:M+1
Vmol(k) = 0.0;
for j=1:N
if (j <N)
U(j) = C(j)*(rho(k))^miu(j) + C(j+N)*(rho(k))^(-miu(j));
end
if (j == N)
U(j) = C(j) + C(j+N)*log(rho(k));
end
Vmol(k) = Vmol(k) + T(10,j)*U(j);
% Vmol(k) = Vmol(k) + T(1,j)*U(j);
end
end
% ALSO, CALCULATE THE EXACT VALUE OF V
Vex = Vo* (log(rho/a))/log(b/a);
diary
Vmol, Vex'
diary off
hold off
plot (rho,Vmol);
title( 'Fig. 9.5 Comparison of exact and method of lines solutions.')
xlabel('rho'), ylabel('V')
hold on
plot (rho,Vex);
hold off

Fig. 9.4 MATLAB code for Example 9.3.

%+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
% EXAMPLE 9.4 SOLVED USING METHOD OF LINES
% THIS M-FILES REQUIRES TWO OTHER M-FILES
% "FUN.M" AND "ROOT.M" TO WORK
% FUN.M DETERMINES THE DETERMINANT OF MATRIX [F]
% ROOT.M FINDS THE ROOT(S) OF THE det(F)
global d

Nmax = 50; tol = 1E-5;

eeff1 = 5; eeff2 = 10;% initial/guessed values
for i=1:15
d = 0.01*i;
x(i) = d;
[eeff] = root('fun', eeff1, eeff2, tol, Nmax)
y(i) = eeff;
%diary a:test.out
end
diary
plot(x,y)
diary off

Fig. 9.9(a) Main MATLAB code for Example 9.4.

%+++++++++++++++++++++++++++++++
function determinant = fun(eeff)
% THIS FUNCTION IS NEEDED FOR EXAMPLE 9.4
% IT DETERMINES THE DETERMINANT OF MATRIX [F]
% FOR A GIVEN EFFECTIVE DIELECTRIC CONSTANT EEFF
global d

N = 9;
%d = 0.001;
a = 7*d; b = 3*d; w = 2*d;
er = 9; h = a/N; M = w/h;
lambdao = 1; %assumed
ko = 2*pi/lambdao;
% First calculate the transformation matrices
% Te (= T_ND) and Th (=T_DN)
cons=sqrt(2/(N+0.5));
for i=1:N
for j=1:N
alj = (j - 0.5)*pi/(N + 0.5);
if (j==1)
te(i,j) = 1/sqrt(N);
else
te(i,j) = cons*cos( (i-0.5)*alj);
end
end
end
% Calculate matrices: chi, eta, delta, and gamma
beta = ko*sqrt(eeff);
tau = (1 - eeff)/(er - eeff);
chi = zeros(N,1);
eta = zeros(N,1);
%gammae1 = zeros(N,N)
for i=1:N
x = ((i -0.5)*pi)/(2*N + 1)
chi(i) = sqrt( 4*sin(x)*sin(x) -h^2*(ko^2 - beta^2) );
eta(i) = sqrt( 4*sin(x)*sin(x) -h^2*(er*ko^2 - beta^2) );
end
for i=1:N
for j=1:N
if(i==j)
del(i,i) = 2*sin (((i - 0.5)*pi)/(2*N + 1));
gammae1(i,i) = chi(i)*coth(b*chi(i)/h);
gammah1(i,i) = chi(i)*tanh(b*chi(i)/h);
gammae2(i,i) = eta(i)*coth(d*eta(i)/h);
gammah2(i,i) = eta(i)*tanh(d*eta(i)/h);
else
del(i,j) = 0;
gammae1(i,j) = 0; gammae2(i,j) = 0;
gammah1(i,j) = 0; gammah2(i,j) = 0;
end
end
end
% calculate rho matrix
rho = inv(gammae1 + er*tau*gammae2 - eeff*(1 - tau)^2*del'*(inv(gammah1
+ tau*gammah2))*del);
F = [te(1:3,1:3)*rho(1:3,1:3)*te(1:3,1:3)'];
determinant = det(F);
% Next, solve for the root det(X) = 0 using 'root.m'

%+++++++++++++++++++++++++++++++++++++++++++++++++
function [x2] = root(F, x0, x1, tol, Nmax)
% THIS FUNCTION FINDS THE ROOTS OF fun(x) USING
% THE SECANT METHOD
% fun(x) - EXTERNAL FUNCTION THAT COMPUTES THE VALUES OF f(x)
% x0, x1 - LIMITS OF THE INITIAL RANGE
% x2 - ROOT RETURNED TO THE CALLING ROUTINE
% tol - TOLERANCE VALUE USED IN DETERMINING CONVERGENCE
% Nmax - MAXIMUM NO. OF ITERATIONS
%
% AUTHORS: M. V. ZEIGLER AND E. CHAMPION [37a]

%F0 = eval( [F, '(x0)']);

F0 = fun(x0);
dx = x1 - x0;
for I=1:Nmax
F1 = fun(x1);
%F1 = eval( [F, '(x1)']);
dx = F1*dx/(F1 - F0);
x2 = x1 - dx;
% CHECK IF TOLERANCE HAS BEEN MET
if (abs(dx) <= tol)
fprintf('root at x = %5.g\n', x2);
fprintf('root found after this no. of iterations %5.g\n',I);
break
end
dx = x2 - x1;
x0 = x1;
x1 = x2;
F0 = F1;
end

Fig. 9.9 For Example 9.4.

APPENDIX C

% A set of simultaneous equations [A][X]=[B] is solved by a simple Gaussian

% elimination scheme (without pivoting) and by the MATLAB backward slash "\"
% operation. The MATLAB backward slash "\" operation performs Gaussian
% elimination with pivoting as explained in "Numerical Computing with
% MATLAB" by Cleave Moler. This excellent reference is available for
% download at the mathworks website.

clear;

%Input matrix
A = [10 -7 0;-3 2 6;5 -1 5];
B = [7;4;6];

%MATLAB's backward slash operation

x = A\B;

%Begin simple Gaussian Elimination

N = size(A,1);
%Forward Elimination
for kk = 1:(N-1);
for ii = (kk+1):N
q = A(ii,kk)/A(kk,kk);
%Calculate Elements of the new matrix
jj = (kk+1):N;
A(ii,jj) = A(ii,jj) - q*A(kk,jj);
A(ii,kk) = q;
B(ii) = B(ii) - q*B(kk);
end
end
%Backward Substituion
X(N) = B(N)/A(N,N);
for ii = (N-1):-1:1
sum = A(ii,:)*X(:);
X(ii) = ( B(ii)-sum )/A(ii,ii);
end

thedifference = x-X(:)

Fig. C.1 Gauss elimination method of solving [A][X] = [B].

% This program applies the Cholesky elimination method to solve the system
% [A][X]=[B] where [AA] = augmented matrix [A,B] and compares the result to
% the MATLAB backslash operation.

clear;

%Define A and B
A = [10 -7 0;-3 2 6;5 -1 5];
B = [7;4;6];

%MATLAB backslash operation

x = A\B;

%Begin Cholesky elimination

AA = [A,B];
M = size(AA,2);
N = size(AA,1);

%Calculate the first row of the upper triangular matrix [T]

for jj = 2:M
AA(1,jj) = AA(1,jj)/AA(1,1);
end
%Calculate the other elements of [T] and lower triangular [L] matrices
for ii = 2:N
for qq = ii:N
sum = 0;
for kk = 1:(ii-1)
sum = sum + AA(qq,kk)*AA(kk,ii);
end
AA(qq,ii) = AA(qq,ii)-sum; %calculates [L]
end
for jj = ii+1:M
sum = 0;
for kk = 1:ii-1
sum = sum + AA(ii,kk)*AA(kk,jj);
end
AA(ii,jj) = (AA(ii,jj)-sum )/AA(ii,ii); %calculates [T]
end
end
%Solve for [X] by back substitution
X(N) = AA(N,M);
for nn = 1:N-1
sum = 0;
ii = N-nn;
for jj = ii+1:N
sum = sum + AA(ii,jj)*X(jj);
end
X(ii) = AA(ii,M)-sum;
end

thedifference = x-X(:)

Fig. C.2 Cholesky’s elimination method of solving [A][X]=[B].

% This program employs Gauss-Seidel iterative method to solve a set of

% simultaneous equations [A][X] = [B] and compares the result to the MATLAB
% backslash operator "\". This method will converge if the matrix A is
% diagonally dominant (i.e. the absolute value of the diagonal term is
% greater than the sum of absolute values of the other terms in each row).

clear;

%Define A and B
A = [10 -7 0;-3 6 1;2 -1 5];
B = [7;4;6];

%MATLAB backslash operation

x = A\B;
%Begin Gauss-Seidel iterative method
N = size(A,1);
X = zeros(N,1);
K = 0; % iteration count
TOL = 1e-3; % tolerance for zero
converged = 0; %while loop condition
Xnew = zeros(N,1);

while converged == 0
for ii = 1:N
Xnew(ii) = 1/A(ii,ii)*( B(ii) - A(ii,1:(ii-1))*Xnew(1:(ii-1))...
- A(ii,(ii+1):N)*X((ii+1):N));
end
%Convergent condition
d = sum(abs(Xnew-X))/sum(abs(Xnew));
%Replace old value with newly computed value
X = Xnew;

K = K + 1;
disp([num2str(K),' ',num2str(X(:)'),' ',num2str(d)])
if d<=TOL %convergence test
converged = 1;
end
end

thedifference = x-X(:)

Fig. C.3 Gauss-Seidel iterative method of solving [A][X]=[B].

% This program applies the conjugate gradients method to solve a set of

% simultaneous equations [A][X]=[B] and compares the result to the MATLAB
% backslash operator "\".

clear;

%Define A and B
A = [10 -7 0;-3 6 1;2 -1 5];
B = [7;4;6];

%MATLAB backslash operation

x = A\B;

%Begin Conjugate Gradients Method

N = size(A,1);
kk = 0;% iteration count
giveup = 20;
TOL = 1e-3; %tolerance for zero
X = zeros(N,1);
P = B;
R = B;
converged = 0; %while loop condition

%Iteration begins here

while ~converged
U = A*P;
alpha = (P.'*R)/(P.'*U);
X = X+alpha*P;
R = R-alpha*U;

%calculate residuals and direction vectors

if R.'*R<TOL
 converged = 1;
disp(['Result: X = ',num2str(X')]);
elseif (kk>giveup)
disp('max number of iterations reached. exiting.')
converged = 1;
end
beta = -(R.'*U)/(P.'*U);
P = R+beta*P;

kk = kk + 1;
end

thedifference = x-X

Fig. C.4 The program applies the conjugate gradients method to solve [A][X]=[B].

%This program performs matrix inversion using Gauss-Jordan elimination

%method and the result is compared to the MATLAB function 'inv'

clear;

%Define [A]
A = [10 -7 0;-3 6 1;2 -1 5];
%MATLAB operation
matinv = inv(A);

%Begin Gauss-Jordan Elimination method

N = size(A,1);
%Calculate elements of reduced matrix
for kk = 1:N
ii = [1:(kk-1),(kk+1):N]; %non-pivot row index
jj = [1:(kk-1),(kk+1):N]; %non-pivot column index

%Calculate new elements of pivot row

A(kk,jj) = A(kk,jj)/A(kk,kk);

%Calculate element replacing pivot element

A(kk,kk) = 1/A(kk,kk);

%Calculate new elements not in povot row or pivot column

A(ii,jj) = A(ii,jj) - A(ii,kk)*A(kk,jj);

%Calculate replacement elements for pivot colum-except pivot element

A(ii,kk) = - A(ii,kk)*A(kk,kk);
end

anydiff = matinv-A

Fig. C.5 Matrix inversion using Gauss-Jordan simultaneous method.

% This example finds the largest Eigenvalue and the associated

% eigen vector of a matrix equation [A][X]=Lambda*[X] and
% compares this to the largest eigenvalue obtained by the
% MATLAB eig function.

clear;

%Define [A]
A = [10 -7 0;-3 6 1;2 -1 5];

%MATLAB eig function

eigval = eig(A);

%Begin Iterative Method

maxiter = 100;
epsi = 1e-3;
kk = 0;
X = ones(size(A,1),1);
valuefound = 0;

while ~valuefound
D = A*X;
Z = D/D(1);

%Check if result is sufficient

test = abs((X-Z)-epsi*abs(Z));
if max(abs(X-Z))<epsi
valuefound = 1;
elseif maxiter<kk
disp('Failed to find solution')
valuefound = 1;
end

X = Z;
disp(D')
kk = kk + 1;
end

lambda = D(1);

thedifference = lambda - eigval(1)

Fig. C.6 A program for finding the largest eigenvalue of equation [A][X] =
LAMBDA [X].
% The following program attempts to implement the Jacobi's method of
% finding eigenvalues and eigenvectors. Just for illustration we have
% taken a random 8x8 symmetric matrix. If the actual matrix is stored %
in a .mat file it has to be loaded using load command in the second
%line and the first line will be commented out. We have decided to have
% 10 iterations or less, when the determinant of the original matrix is
% close to that of the modified matrix. The difference E is treated as %
error. We wish that the error E is less than eps or # of iterations % =
10. These parametes have to be appropriately chosen based on one's %
experience and nature of the problem under consideration.
%
% *********** Description of the variables *********
% A = original matrix saved in Ao to start with.
% R is the rotation matrix.
% Rp is the product of all R's.
% At the end of the while loop Rp has the eigen vectors
% in its columns and A has the eigen values along the main diagonal.
% In fact MATLAB has a built-in routine called eig that finds the eigen
% values and the eigen vectors. So we wish to compare both the results;
% hence we have added a command at the bottom of the code
% asking to dispaly both. Note that the eigen values are not sorted in %
this code.
%%%%%%%%%%%%%%%

clear all;close all;

A = 5*rand(8); A = A+A';
%load Afile.mat
N = length(A(:,1)); % N = 8 in this case
Ao=A;
Rp=eye(N);
E=10; Niter=1;
while (E > eps)|(Niter<10)
for p = 1:N
for q = p+1:N
th = .5*atan(2*A(p,q)/(A(p,p)-A(q,q)));
R = eye(N);
R([p,q],[p,q]) = [cos(th) -sin(th);sin(th) cos(th)];
A=R'*A*R;
Rp=Rp*R;
end
end
Niter = Niter+1;
E=det(A)-det(Ao);
end
Rp
A
[V,D]=eig(A)
E

Fig. C.7 MATLAB code for finding all the eigenvalues and eigenvectors of the
equation [A][X] = LAMDBA [x]