An Introduction to Ordinary Differential Equations

MATLAB files
James C. Robinson

1
MATLAB files referred to in the book

This document contains listings of all the MATLAB files available on the web,
and brief descriptions of what they do and how to use them. They were all
written to help create the figures and exercises in the book; as such they are
in no way intended to stand up as worthy examples of elegant programming.
Nevertheless, I hope that they will be useful.
Note that many of these programs use the same variable names, so it is
a good idea to use the clear command between different programs to clear
all the variables.

besselseries.m
This program computes the series expansion of the Bessel function of order
that has the form

x
an xn ,
n=0

using the recurrence relation

an =

an2
,
n(n + 2)

as discussed in Section 20.4. The program then plots this series solution,
along with the true solution (the normalisation has to be changed to make
the two coincide, see comments after Example 20.3, and Exercise 20.8).
%% Bessel function of order nu from its series expansion
%% Sum up to power x^(2N)
nu=input(nu = );
N=input(N= );

1 MATLAB files referred to in the book

a(1)=-1/(4*(1+nu));
%% a(i) is coefficient of x^(2i)
for i=2:N;
a(i)=-a(i-1)/(4*i*(i+nu));
end
x=linspace(0,10,1000);
y=1+0.*x;
hold on
for k=1:N;
y=y+a(k)*x.^(2*k);
ay=y.*x.^nu;
if (k/2)~=floor(k/2) & k>1;
plot(x,ay);
end
end
%% Change the normalisation
z=besselj(nu,x)*2^nu*gamma(1+nu);
plot(x,z,linewidth,2)
ylim([1.5*min(z),1.5*max(z)])

bifurcation.m
This short program draws the bifurcation diagram for the logistic map
xn+1 = f (xn ) := rxn (1 xn ). For a selection of values of r between the
values you input it computes 100 iterates, and then plots the next 30 on the
vertical axis. This can take a long time to run.
%% bifurcation.m
%% draw bifurcation diagram for logistic map
clf
hold on
r1=input(Smallest value of r: );
r2=input(Largest value of r: );
%% N is the number of r values used
%% lowering/increasing N will decrease/increase the resolution

MATLAB files referred to in the book

%% with a consequent decrease/increase in the time needed

N=400;
for t=0:N
%% for each value of r start at x_0=0.5
z=0.5;
r=r1+((r2-r1)*t/N);
%% compute one hundred iterates of f
for j=1:100
z=r*z*(1-z);
end
%% and then plot the next 30
for j=1:30
z=r*z*(1-z);
plot(r,z,.,MarkerSize,2)
end
end
xlabel(r,FontSize,20)
xlim([r1 r2])

cylinder.m
This plots animated trajectories of the nonlinear pendulum equation, but
moving in their natural phase space, the cylinder. The * represents the
stable fixed point (the pendulum hanging vertically downwards) and the
x represents the unstable fixed point (the pendulum balanced vertically
upwards). If you tilt the three-dimensional figure correctly you can see
something very like the swings of the pendulum. This program uses the file
pendde.m. The program ends when the initial condition is 999.
%%
%%

Animate trajectories of the nonlinear pendulum

on the phase cylinder

%%

Requires pendde.m

hold off

1 MATLAB files referred to in the book

%% Draw portion of cylinder

[tpx,tpz]=meshgrid(-pi:.1:pi+.1,-3:.1:3);
tpcx=cos(tpx);
tpcy=sin(tpx);
surfl(tpcx,tpcy,tpz), shading flat
hold on
axis off
%% * is pendulum pointing down
%% x is pendulum pointing up
plot3(-1,0,0,*)
plot3(1,0,0,x)
X0=0;
%% program ends when initial condition is 999
while X0~=999
X0=input(Initial condition:
if imag(X0)~=0
X0=[u(2001,1) u(2001,2)];
end
if X0~=999
T=input(Maximum time:
);
if X0~=999

);

tspan=linspace(0,T,2001);
[t,u]=ode45(pendde,tspan,X0,[],1);
c1(:,1)=cos(pi+u(:,1));
c2(:,2)=sin(pi+u(:,1));
c3(:,3)=u(:,2);
comet3(c1(:,1),c2(:,2),c3(:,3))
end
end
end
hold off

MATLAB files referred to in the book

darrow.m
This routine, which is handy for phase diagrams, draws an arrow at the
point (xo, yo) in the direction (nx, ny) of size s.
function darrow(xo,yo,nx,ny,s)
%% Draw an arrow at (xo,yo) in the (nx,ny)
%% direction, of size s
if s==0;

s=0.4;

end

%% theta is (half) the angle the sides of the arrow make

%% to each other
theta=pi/9;
l=sqrt(nx^2+ny^2); if l~=0; nx=nx/l; ny=ny/l; end
if nx~=0
ppx=-(ny/nx); ppy=1;
nl=sqrt(ppx^2+ppy^2);
ppx=ppx/nl; ppy=ppy/nl;
else
ppx=1; ppy=0;
end
plot([xo-s*nx+s*tan(theta)*ppx xo],[yo-s*ny+s*tan(theta)*ppy yo])
plot([xo-s*nx-s*tan(theta)*ppx xo],[yo-s*ny-s*tan(theta)*ppy yo])

euler.m
This is the MATLAB implementation of Eulers method,
xn+1 = xn + hf (xn , tn ),
discussed in Section 21.3, applied as there to the simple differential equation
x = t x2 .
%% Eulers method
T=12;

%% final time

h=0.5;

%% timestep

%% MATLAB does not allow an index 0

%% on a vector, so x_n is x(n+1) here
t(1)=0;
x(1)=0;

%% initial time
%% initial condition

1 MATLAB files referred to in the book

for n=1:T/h;
t(n+1)=n*h;
x(n+1)=x(n) + h * (t(n) - x(n)^2);
end
[t x]

%% display values

%% Plot crosses at numerical values, and join these

plot(t,x,x,MarkerSize,20)
hold on
plot(t,x)

f2.m
This program draws the pictures for Exercise 24.3, which asks you to investigate the periodic-doubling cascade by looking at a renormalised version
of the graph of f 2 . First it draws graphs of f and f 2 , then finds the righthand point nxr of the small box (where f 2 (nxr) = 1 (1/r)) using an
interval-halving method. It then draws a blown-up version of the function
that appears in this small box.
%%

Renormalisation of logistic map f

r=input(value of r:

);

clf reset
%%

The left-hand plot shows f and f^2

subplot(121)
x0=0:0.025:1;
%% y=f(x0) and z=f^2(x0)
y=r.*x0.*(1-x0);
z=r.*y.*(1-y);
plot(x0,y)
hold on
plot(x0,x0)
plot(x0,z, LineWidth, 2)
axis equal
%% Now find where f^2(nxl)=1-(1/r)

MATLAB files referred to in the book

%% using the interval halving method

nxl=1-(1/r);
g=inline((r^2*x*(1-x)*(1-(r*x*(1-x))))-1+(1/r),x,r);
xl=1-(1/r)+0.00001; xr=1;
d=1;
while d>0.0001
m=(xl+xr)/2;
if g(xl,r)*g(m,r)<0
xr=m;
else
xl=m;
end
d=(xr-xl);
end
nxr=m;
%% Draw the little invariant box
plot([nxr
plot([nxr
plot([nxl
plot([nxl

nxr],[nxr
nxl],[nxl
nxr],[nxr
nxl],[nxr

nxl])
nxl])
nxr])
nxl])

xlim([0 1])
ylim([0 1])
%%

On the right-hand side draw an expanded version of little box

subplot(122)
x2=linspace(nxl,nxr,100);
y2=r.*x2.*(1-x2); y2=r.*y2.*(1-y2);
plot(x2,y2, LineWidth, 2);
hold on
plot(x2,x2);
axis equal
xlim([nxl nxr])
ylim([nxl nxr])

flies.m
This program, inspired by the MMath project of Warwick student James
Macdonald, animates the swarm of flies produced by solving the Lorenz
equations for a cluster of different initial conditions using solvem.m. As
noted below, any more than 63 solutions will strain most computers. [It is
important not to use the clear command between running solvem.m and
flies.m.]

1 MATLAB files referred to in the book

%% Swarm of flies from the Lorenz equations

%% Run M-file solvem.m first
hold on
xlim([-20 20])
ylim([-30 30])
zlim([0 50])
for m=1:(d+1)^3;
hend(m)=plot3(y0(1,m),y0(2,m),y0(3,m),Erase,background);
hstart(m)=plot3(y0(1,m),y0(2,m),y0(3,m),Erase,none);
end
pause=input(Rotate axes now, then press [RETURN]);
for i=floor(length(yi)/3):length(yi)-1,
for m=1:(d+1)^3
set(hstart(m),xdata,yi(i,1,m),ydata,yi(i,2,m),zdata,yi(i,3,m));
drawnow
set(hend(m),xdata,yi(i,1,m),ydata,yi(i,2,m),zdata,yi(i,3,m));
drawnow
end
end

linearde.m
This file contains the linear equation


dx
a b
=
c d
dt
for use in lportrait.m (via the ode45 routine).
function yprime=linearde(t,y,flag,a,b,c,d)
yprime=[a*y(1)+b*y(2); c*y(1)+d*y(2)];

logistic.m
This program will draw cobweb diagrams for the logistic map
xn+1 = rxn (1 xn ),
which forms the subject of Chapter 24. Having chosen r the program draws
the graph of f (x) = rx(1 x) for 0 x 1 and the diagonal (y = x).
Then given an initial condition x 0 it will plot the selected number of iterates

MATLAB files referred to in the book

on the diagram. When asked for the number of iterates, 0 will finish the
program, .1 will clear all lines from the cobweb diagram, .2 will change the
initial condition, and .3 will erase everything and allow you to choose a new
value of r. The clear option is particularly useful, for example, to find a
stable periodic orbit: start anywhere, perform 100 iterations (say), clear the
picture, and then compute a few iterations to show a clean picture of the
orbit.
%%
%%

logistic.m
draw cobweb diagrams for logistic map

hold off
r=input(r=: );
x=0:0.01:1;
y=r.*x.*(1-x);
plot(x,x)
hold on
plot(x,y, LineWidth, 2)
z=input(x_0=: );
n=1; while n~=0
n=input( ...
# iterations (0=stop, .1=clear, .2=change IC, .3=change r): );
if n==0.1
hold off; plot(x,x); hold on; plot(x,y, LineWidth, 2);
end
if n==0.2
z=input(new initial condition: );
end
if n==0.3
r=input(new value of r:
);
hold off
y=r.*x.*(1-x);
plot(x,x)
hold on
plot(x,y)
z=input(z_0=: );
end
if n~=0
for j=1:abs(n)
nz=r*z*(1-z);
if n>0
plot([z z],[z nz])
ms=min(8,25*abs(nz-z));
if z<nz plot(z,(z+nz)/2,^,markersize,ms)
else
plot(z,(z+nz)/2,v,markersize,ms)
end
plot([z nz],[nz nz])
if z<nz plot((z+nz)/2,nz,>,markersize,ms)

10

1 MATLAB files referred to in the book

else
plot((z+nz)/2,nz,<,markersize,ms)
end
end
z=nz;
end
end

end

logisticn.m
This simple program iterates the logistic map, and plots successive iterates
xn against n. The program finishes when the initial condition lies outside
the interval [0, 1].
%% Plot iterates of logsitic map
%% against n
clf
hold on
r=input(r = );
N=input(N = );
xlim([0 N]); ylim([0 1]);
xlabel(n,FontSize,20); ylabel(x,FontSize,20)
x0=input(initial condition = );
while x0>=0 & x0<=1
x(1)=x0; t(1)=0;
for n=2:N;
t(n)=n-1;
x(n)=r*x(n-1)*(1-x(n-1));
end
plot(t,x)
plot(t,x,x,MarkerSize,15)
x0=input(initial condition = );
end

MATLAB files referred to in the book

11

lorenzdraw.m
This program integrates the Lorenz equations (using MATLABs ode45 command) up to time t = 50, starting at the initial condition yzero. It then
draws the trajectory in R3 . The equations themselves are contained in the
file lorenzde.m.
%% Integrate the Lorenz equations
%% Requires lorenzde.m
tspan=[0 50];
yzero=input(yzero = );
[t,y]=ode45(lorenzde,tspan,yzero);
%% Draw the trajectory in R^3
plot3(y(:,1),y(:,2),y(:,3))
xlabel(x); ylabel(y); zlabel(z);
title(Lorenz attractor)

lorenzde.m
This file contains the Lorenz equations,
x = (x + y)
y = rx y xz
z = bz + xy
for use in lorenzdraw.m, lorenz37.m, and solvem.m (via the ode45 command).
function yprime=lorenzde(t,y)
yprime=[10*(y(2)-y(1)); 28*y(1)-y(2)-y(1)*y(3); y(1)*y(2)-8*y(3)/3];

lorenz37.m
This long program is essentially a presentation of various topics that are
covered in Chapter 37 that discusses the Lorenz equations. It would be
useful to run the program as you read that chapter. Note that the program
requires the lorenzde.m file.
Having set up the parameters it gives explicit values for the non-zero
stationary points, and plots them. It then computes the eigenvalues and
eigenvectors at the origin and at the non-zero stationary points.
After this it plots the direction field near each stationary point, and then,
unhelpfully all these direction fields together on one plot. After this it

12

1 MATLAB files referred to in the book

shows three components of the solution, along with the function V (x, y, z) =
x2 + y 2 + (z r )2 which can be used to show that the solutions remain
bounded for all time.
Then it plots the trajectories in R3 , showing the Lorenz attractor, and
finally draws a graph of successive maxima of z against each other.
%%
%%

Lorenz presentation
Requires lorenzde.m

clf
echo
% parameters
sigma=10; r=28; b=8/3;
echo off
p=input(To continue press [RETURN]); echo
% non-zero fixed points
x=sqrt(b*(r-1)); y=x; z=r-1;
echo off
p=input(draw the fixed points (press [RETURN]));
% draw the three fixed points & some axes
plot3([0 0],[0 0],[-40 40],g)
hold on
plot3([0 0],[-40 40],[0 0])
plot3([-40 40],[0 0],[0 0])
plot3(0,0,0,x,MarkerSize,15)
plot3(x,y,z,x,MarkerSize,15)
plot3(-x,-y,-z,x,MarkerSize,15)
xlabel(x); ylabel(y); zlabel(z);
p=input(To continue press [RETURN]);
echo
% eigenvalues and eigenvectors at the origin
% L gives values, V gives vectors
[V L]=eig([-sigma sigma 0; r -1 0; 0 0 -b])
% two stable directions and one unstable direction
echo off
p=input(look close to origin (press [RETURN]));
echo

MATLAB files referred to in the book

13

% draw local direction field near origin

% the picture is nice if you rotate to -141 30
[x0,y0,z0]=meshgrid(-2:1:2,-2:1:2,-2:1:2);
xd=sigma.*(-x0+y0);
yd=r.*x0-y0-x0.*z0;
zd=-b.*z0+x0.*y0;
quiver3(x0,y0,z0,xd,yd,zd)
xlim([-2.5 2.5]);
ylim([-2.5 2.5]);
zlim([-2.5 2.5])
a1=3.5*V(:,1);
a2=3.5*V(:,2);
a3=3.5*V(:,3);
plot3([-a1(1) a1(1)],[-a1(2)
plot3([-a2(1) a2(1)],[-a2(2)
plot3([-a3(1) a3(1)],[-a3(2)
plot3([-a3(1) a3(1)],[-a3(2)

a1(2)],[-a1(3)
a2(2)],[-a2(3)
a3(2)],[-a3(3)
a3(2)],[-a3(3)

a1(3)],g)
a2(3)],r)
a3(3)],w)
a3(3)],g)

echo off
p=input(To continue press [RETURN]);
echo
% eigenvalues and eigenvectors at the non-zero points
[V L]=eig([-sigma sigma 0; r-z -1 -x; y x -b])
% a stable direction and a 2d unstable focus
echo off
p=input(look close to one of these fixed points (press [RETURN]));
echo
% draw local direction field near one of these fixed points
[x0,y0,z0]=meshgrid(x-2:1:x+2,y-2:1:y+2,z-2:1:z+2);
xd=sigma.*(-x0+y0);
yd=r.*x0-y0-x0.*z0;
zd=-b.*z0+x0.*y0;
quiver3(x0,y0,z0,xd,yd,zd)
xlim([x-3 x+3]);
ylim([y-3 y+3]);
zlim([z-3 z+3])
% draw some local axes to make picture easier to see
a1=3.5*V(:,1); a2=3.5*real(V(:,2)); a3=3.5*imag(V(:,2));
plot3([x-a1(1) x+a1(1)],[y-a1(2) y+a1(2)],[z-a1(3) z+a1(3)],g)

14

1 MATLAB files referred to in the book

plot3([x-a2(1) x+a2(1)],[y-a2(2) y+a2(2)],[z-a2(3) z+a2(3)],r)

plot3([x-a3(1) x+a3(1)],[y-a3(2) y+a3(2)],[z-a3(3) z+a3(3)],r)
% rotate to -85 44 to see rotational part
echo off
% draw local direction field near the other one of these fixed points
[x0,y0,z0]=meshgrid(-x-2:1:-x+2,-y-2:1:-y+2,z-2:1:z+2);
xd=sigma.*(-x0+y0);
yd=r.*x0-y0-x0.*z0;
zd=-b.*z0+x0.*y0;
quiver3(x0,y0,z0,xd,yd,zd)
% draw some local axes to make picture easier to see
plot3([-x-a1(1) -x+a1(1)],[-y-a1(2) -y+a1(2)],[z-a1(3) z+a1(3)],g)
plot3([-x-a2(1) -x+a2(1)],[-y-a2(2) -y+a2(2)],[z-a2(3) z+a2(3)],r)
plot3([-x-a3(1) -x+a3(1)],[-y-a3(2) -y+a3(2)],[z-a3(3) z+a3(3)],r)
p=input(look at whole picture again (press [RETURN]));
echo
% and "join the dots"?
% show whole picture
xlim([-35 35]); ylim([-35 35]); zlim([-2 35])
% oh dear -- try looking at components of solutions
% as functions of time, and also the bounding function
% V(x,y,z)=x^2+y^2+(z-sigma-r)^2
echo off
p=input(To continue press [RETURN]);
figure(2)
tt=1; cla
yzero=[0;1;0];
while yzero~=999
if tt==1; echo; end
tspan=[0 50];
% length of time
yzero=input(initial condition (input 999 to continue):
if yzero~=999

);

MATLAB files referred to in the book

15

[t,y]=ode45(lorenzde,tspan,yzero);
figure(2)
subplot 411
plot(t,y(:,1)); ylabel(x)
subplot 412
plot(t,y(:,2)); ylabel(y)
subplot 413
plot(t,y(:,3)); ylabel(z)
V=sqrt(y(:,1).^2+y(:,2).^2+(y(:,3)-sigma-r).^2);
subplot 414
plot(t,V); ylabel(V(x,y,z)^{1/2})
end
if tt==1; echo off; end
tt=tt+1;
end
echo
% this is no good -- look at trajectories moving in R^3
echo off
figure(1)
yzero=[0;1;0];
while yzero~=999
if tt==1; echo; end
tspan=[0 50];
% length of time
yzero=input(initial condition (input 999 to continue):
if yzero~=999
[t,y]=ode45(lorenzde,tspan,yzero);
comet3(y(:,1),y(:,2),y(:,3))
end
if tt==1; echo off; end
tt=tt+1;
end
echo
% still looks complicated... try plotting successive maximum
% values of z against each other...

);

16

1 MATLAB files referred to in the book

echo off
p=input(To continue press [RETURN]);
yzero=[0 1 0]
figure
hold on
for k=1:25
tspan=[0 20]; [t,y]=ode45(lorenzde,tspan,yzero);
j=0;
for i=2:length(y)-1
if y(i,3)>y(i+1,3) & y(i,3)>y(i-1,3)
j=j+1;
m(j)=y(i,3);
end
end
for i=1:j-1
plot(m(i),m(i+1),x)
end
yzero=y(940,:);
end

lotkade.m
This final contains a general form of the Lotka-Volterra equations
x = x(A + ax + by)
y = y(B + cx + dy)
for use in lotkaplane.m (via the ode45 command). The inclusion of the
extra parameter sn enables lotkaplane.m to integrate the trajectory both
forwards (if sn= 1) and backwards (if sn= 1) in time.
function yprime=lotkade(t,y,flag,sn,A,B,a,b,c,d)
yprime=sn*[y(1)*(A+a*y(1)+b*y(2)); y(2)*(B+c*y(1)+d*y(2))];

lotkaplane.m
This program integrates the Lotka-Volterra model (above) for a given choice
of the parameters, and then plots the trajectory in the phase plane. It
integrates both forwards and backwards from the initial condition for a time

MATLAB files referred to in the book

17

T , and as a default places an arrow indicating the direction of movement

at the initial condition. However, there are other options: if T < 0 then
the equations are integrated for a time |T | but no arrow is used. If T has
an imaginary part (e.g. 10 + i) then the solution is integrated for a time
Re(T ) and drawn as a thicker line (this is useful for highlighting stable
manifolds). If a complex initial condition is entered then the program will
continue integrating from the end of the previous trajectory, while it will
finish when the initial condition is 999.
The program uses lotkade.m (which contains the equations) and darrow.m
(a routine to add arrows to the trajectories).
%%
%%
%%
%%

lotkaplane.m
draw phase plane diagram of the
Lotka-Volterra model of two species
Requires lotakde.m & darrow.m

hold on
%%

input parameters

A=input(A
a=input(a
b=input(b
B=input(B
c=input(c
d=input(d

=
=
=
=
=
=

);
);
);
);
);
);

X0=0;
%% mX is largest value of x and y in the phase diagram
%% change this number if the stationary points lie
%% outside 0<=x,y<=5
mX=5;
%% X0 is initial condition
%% if X0=999 then program finishes
%% if X0 has non-zero imaginary part then the integration
%%
continues from the end of the previous trajectory
while X0~=999
X0=input(Initial condition:
if imag(X0)~=0
X0=[u(2001,1) u(2001,2)];
end
%% T is maximum time

);

18

1 MATLAB files referred to in the book

%% if T is negative there is no arrow

%% if T has non-zero imaginary part the trajectory is
%%
drawn in bold
T=input(Maximum time:
if X0~=999

);

RX=X0;
%%

trajectory is calculated both forwards and backwards from X0

for k = 1:2,
X0=RX;
sn=(-1)^k;
rT=2*abs(real(T)); rN=4001; u(1,1)=mX*2; u(2,2)=mX*2;

%%

make sure that trajectory doesnt get too large

while (max([max(u(:,1)) max(u(:,2))])>mX)
rT=rT/2; rN=(rN-1)/2+1;
tspan=linspace(0,rT,rN);
[t,u]=ode45(lotkade,tspan,X0,[],sn,A,B,a,b,c,d);
end
if imag(T)==0
plot(u(:,1), u(:,2))
else
plot(u(:,1), u(:,2), LineWidth,2)
end
end

%%

put the arrow on the trajectory (unless T is imaginary)

D=lotkade(0,X0,[],1,A,B,a,b,c,d);
if real(T)>0; darrow(X0(1), X0(2), D(1), D(2), 0.05); end

end
end
hold off

MATLAB files referred to in the book

19

lportrait.m
This lengthy file draws phase portraits for linear equations


dx
a b
=
x.
c d
dt
Once the parameters are fixed it calculates and displays the eigenvalues and
eigenvectors of the matrix, and then (if the eigenvalues are not complex)
these eigenvectors, labelling them with arrows. If the eigenvalues have the
same sign then the eigenvector corresponding to the eigenvalue of larger
modulus has two arrows. The figure window then shows the region 1
x, y 1. The program needs the files linearde.m and darrow.m
%%
%%
%%

lportrait.m
phase portraits for linear equations
Requires linearde.m & darrow.m

mng=5;

% maximum number of integrations from each IC

a=input(a=);
b=input(b=);
c=input(c=);
d=input(d=);
ew=1;
%%

% width of eigenvectors

calculate eigenvalues and eigenvectors

[V,D]=eig([a b;c d])

[x,y]=meshgrid(-1:.25:1,-1:.25:1);
DxDt=a*x+b*y; DyDt=c*x+d*y;
hold on
%%

if the eigenvalues are real then draw the eigenvectors

if imag(D(1))==0
x1=linspace(-1,1,101)*V(1,1);
y1=linspace(-1,1,101)*V(2,1);
plot(x1,y1,LineWidth,ew)
x2=linspace(-1,1,101)*V(1,2);
y2=linspace(-1,1,101)*V(2,2);
plot(x2,y2,LineWidth,ew)
%% draw arrows on eigenvectors
s=0.05;

20

1 MATLAB files referred to in the book

xo=.5*V(1,1); yo=.5*V(2,1); nx=D(1,1)*xo; ny=D(1,1)*yo;
darrow(xo,yo,nx,ny,s)
xo=-xo; yo=-yo; nx=-nx; ny=-ny;
darrow(xo,yo,nx,ny,s)
xo=.5*V(1,2); yo=.5*V(2,2); nx=D(2,2)*xo; ny=D(2,2)*yo;
darrow(xo,yo,nx,ny,s)
xo=-xo; yo=-yo; nx=-nx; ny=-ny;
darrow(xo,yo,nx,ny,s)
if D(1,1)*D(2,2)>0
if abs(D(1,1))>abs(D(2,2))
xo=.55*V(1,1);yo=.55*V(2,1); nx=D(1,1)*xo; ny=D(1,1)*yo;
darrow(xo,yo,nx,ny,s)
xo=-xo; yo=-yo; nx=-nx; ny=-ny;
darrow(xo,yo,nx,ny,s)
else
xo=.55*V(1,2); yo=.55*V(2,2); nx=D(2,2)*xo; ny=D(2,2)*yo;
darrow(xo,yo,nx,ny,s)
xo=-xo; yo=-yo; nx=-nx; ny=-ny;
darrow(xo,yo,nx,ny,s)
end
end

end
axis equal
%%

now draw trajectories

tspan=linspace(0,1,101);
X0=0;
%%

enter 999 for initial condition to stop

while X0~=999
X0=input(Initial condition:
%%
%%

);

trajectories starting at both X0 and -XO are drawn,

both forwards and backwards in time
if X0~=999
for yp=1:2;
xm=(-1)^yp; ym=(-1)^yp;
Z0=[xm ym].*X0;
dir=linearde(0,Z0,[],a,b,c,d);
xo=Z0(1); yo=Z0(2); nx=dir(1); ny=dir(2); s=.05;
darrow(xo,yo,nx,ny,s)

MATLAB files referred to in the book

21

for p=1:2;
m=(-1)^p;
j=102; Y0=Z0; gn=0;
while j==102 & gn<mng
[t,u]=ode45(linearde,tspan,Y0,[],a*m,b*m,c*m,d*m);
u(102,:)=[0 0];
j=2;
while abs(u(j,1))<=1 & abs(u(j,2))<=1 & ...
(abs(u(j,1))+abs(u(j,2)))>=0.001 & j<=101
plot([u(j-1,1); u(j,1)],[u(j-1,2); u(j,2)])
j=j+1;
end
Y0=u(101,:); gn=gn+1;
end
end
end
end
end
axis off
axis equal
xlim([-1 1])
ylim([-1 1])

makematrix.m
The program uses the results of Exercises 28.5, 29.5, and 30.4 to construct
matrices with specified eigenvalues and eigenvectors. After inputting the
first eigenvector and eigenvalue, the program only asks for a second if they
are real. Otherwise it takes the second eigenvalue and eigenvector to be the
complex conjugate of the first. The matrix A is constructed as follows:
if the two eigenvalues 1 and 2 are real and distinct then

A=P

1 0
0 2

P1 ,

22

1 MATLAB files referred to in the book

where
P = [v1 v2 ];

(M1.1)

if the eigenvalues and eigenvectors are imaginary, with = i and

= v1 iv2 then

A=P

P1 ,

where P is as in (M1.1);
if the eigenvalues are the same then the first eigenvector is taken as the
eigenvector, and the second eigenvector is taken to be the v2 used in the
coordinate transformation in Section 30.2. In this case

A=P

1
0

P1 ,

with P again given by (M1.1).

%% makematrix.m
%% construct matrix with specified
%% eigenvalues & eigenvectors
l1=input(First eigenvalue = );
v1=input(First eigenvector = ); v1=conj(v1);
if imag(l1)~=0
v2=imag(v1); v1=real(v1);
M=[real(l1) imag(l1); -imag(l1) real(l1)];
else
l2=input(Second eigenvalue = );
v2=input(Second eigenvector = ); v2=v2;
M=[l1 0; 0 l2];
if l1==l2 & norm(v1-v2)~=0;
M=[l1 1; 0 l1];
end
end
P=[v1, v2];
A=P*M*inv(P)
[V D]=eig(A)

MATLAB files referred to in the book

23

newtonde.m
This file contains a variety of Newtonian equations for use in newtonplane.m
(via the ode45 routine). As in lotkade.m, the variable s allows integration
both forwards and backwards in time.
function yprime=newtonde(t,y,flag,s,k)
%% Comment out unwanted equations
%% Particles moving in the following potentials:
%%
V(x) = x-x^3/3
yprime=s.*[y(2); -k*y(2)-1+y(1)^2];
%%
V(x) = x^2/2
%% yprime=s.*[y(2); -k*y(2)-y(1)];
%%
V(x) = x^4/2-x^2
%% yprime=s.*[y(2); -k*y(2)-2*y(1)^3+2*y(1)];
%%
V(x) = x^6/6 - 5x^4/4 + 2x^2
%% yprime=s.*[y(2); -k*y(2)-y(1)^5+5*y(1)^3-4*y(1)];
%% Particles moving on a wire in the following shapes
%%
V(x) = x-x^3/3
%% yprime=s.*[y(2); -k*y(2)-(1-y(1)^2)*(1-2*y(1)*y(2)^2)/(2-2*y(1)^2+y(1)^4)];
%%
V(x) = x^2/2
%% yprime=s.*[y(2); -k*y(2)-y(1)*(1+y(2)^2)/(1+y(1)^2)];
%%
V(x) = x^3/2-x^2
%% n=2*y(1)-2*y(1)^3-(2*y(1)^3-y(1))*(6*y(1)^2-1)*y(2)^2;
%% yprime=s.*[y(2); -k*y(2)+n/(1+(2*y(1)^3-y(1))^2)];

newtonplane.m
Another phase plane program, this time using the file newtonde.m to provide
the equation. It also allows a damping term k x to be added (choose k = 0
for no damping). The program has the same options as lotkaplane.m, and
also requires darrow.m
%%
%%
%%

newtonplane.m
draw phase plane diagrams in whole plane
requires newtonde.m & darrow.m

clf
hold on

24

1 MATLAB files referred to in the book

dmp=input(damping k = );
mX=5;
X0=0;
while X0~=999
X0=input(Initial condition:
if imag(X0)~=0
X0=[u(2001,1) u(2001,2)];
end
T=input(Maximum time:
);
if X0~=999
%
tspan=linspace(0,T,2001);

);

RX=X0;
for k = 1:2,
X0=RX;
sn=(-1)^k;
rT=2*abs(real(T)); rN=4001; u(1,1)=mX*2; u(2,2)=mX*2;
while (max([max(u(:,1)) max(u(:,2))])>mX)
rT=rT/2; rN=(rN-1)/2+1;
tspan=linspace(0,rT,rN);
[t,u]=ode45(newtonde,tspan,X0,[],sn,dmp);
end
if imag(T)==0
plot(u(:,1), u(:,2))
else
plot(u(:,1), u(:,2), linewidth, 2)
end
end
D=newtonde(0,X0,[],1,dmp);
if real(T)>0; darrow(X0(1), X0(2), D(1), D(2), .1); end

MATLAB files referred to in the book

25

end
end

pendde.m
This file has the pendulum equation x
= sin x written in the coupled form
x = y
y = sin x
for use in piphase.m and cylinder.m (via the ode45 routine).
function yprime=pendde(t,y,flag,s)
yprime=s*[y(2); -sin(y(1))];

piphase.m
This phase plane program draws the direction field for the nonlinear pendulum on the phase space between and , and then animated trajectories
on this restricted region of the phase space: if trajectories leave this region
to one side then they will come back on the other. The program requires
pendde.m to run.
%%
%%
%%

piphase.m
phase portrait for the nonlinear pendulum
on -pi<=x<=pi

hold off
[x,y]=meshgrid(-pi:.5:pi,-3:.5:3);
DxDt=y; DyDt=-sin(x);
quiver(x,y,DxDt,DyDt)
hold on
plot([-pi -pi],[-4 4],--)
plot([pi pi],[-4 4],--)
X0=0;
while X0~=999
X0=input(Initial condition [999 to end]:
if imag(X0)~=0
X0=[u(2001,1) u(2001,2)];
end
if X0~=999
T=input(Maximum time:
);
tspan=linspace(0,T,2001);

);

26

1 MATLAB files referred to in the book

[t,u]=ode45(pendde,tspan,X0,[],1);
js=1; je=1;
while je<2001
je=je+1;
if u(je,1)>pi
comet(u(js:je,1),u(js:je,2))
for j=je:2001
u(j,1)=u(j,1)-2*pi;
end
js=je;
end
if u(je,1)<-pi
comet(u(js:je,1),u(js:je,2))
for j=je:2001
u(j,1)=u(j,1)+2*pi;
end
js=je;
end
end
if js~=2001;
comet(u(js:je,1),u(js:je,2))
end
end

end
hold off

solvem.m
1)3

This program integrates (d +

copies of the Lorenz equations on a collection of initial conditions starting in a very small ball (one pixel) near
(0, 1, 0). The output can then be viewed using flies.m. Watching the evolution of 63 solutions will probably push the capabilities of most systems.
The program needs lorenzde.m in order to work.
%%
%%
%%
%%

solvem.m
solve d^3 copies of the Lorenz equations
starting in a small ball
Requires lorenzde.m

clf
clear

MATLAB files referred to in the book

27

d=input(d = );
for i=0:1999;
tspan(i+1)=i*35/1999;
end
for i=0:d;
for j=0:d;
for k=0:d;
m=1+k+(j*(d+1))+(i*(d+1)^2)
y0(:,m)=[(i-4.5)/100000; 1+(j-4.5)/100000; (k-4.5)/100000];
[t,y]=ode45(lorenzde,tspan,y0(:,m));
yi(2000,3,m)=1;
yi(:,:,m)=y;
end
end
end

2
MATLAB files used in the exercise solutions

backeuler.m
This program implements the backwards Euler method
xn+1 = xn + f (xn+1 )
using the iterative method
yj+1 = xn + f (yj )

y0 = xn

(M2.1)

in order to find a good approximation (yj with j large) to xn+1 given xn . It

is applied here to the equation x = x(1 x) (i.e. with f (x) = x(1 x)) as in
Exercise 21.7. You need to be careful to make sure that h is small enough
that the iterative method in (M2.1) converges, see Exercise 21.5.
%% backwards Euler method
T=8;

%% final time

h=0.5;

%% timestep

%% MATLAB does not allow an index 0

%% on a vector, so x_n is x(n+1) here
t(1)=0;
x(1)=.5;

%% initial time
%% initial condition

for n=1:T/h;
t(n+1)=n*h;
%% Use iterative method to find x(n+1) given x(n)
gn=x(n);
g=gn+2*h^3;

28

MATLAB files used in the exercise solutions

29

while abs(gn-g)>h^3;
%% method is O(h) so approximate x(n+1) to within O(h^3)
g=gn;
gn=x(n)+h*g*(1-g);
end
x(n+1)=gn;
end
%% Plot crosses at numerical values, and join these
plot(t,x,x,MarkerSize,20)
hold on
plot(t,x)

exint.m
This is the solution to Exercise 4.1(iii). It draws a graph of
Z

T (t) =

e(ts) sin s ds

against t for 0 t 7.
%% Solution of Exercise 4.1(iii)
f=inline(exp(-(t-s)).*sin(s),t,s);
for j=0:100;
t(j+1)=7*j/100;
T(j+1)=quad(f,0,t(j+1),[],[],t(j+1));
end
plot(t,T)

findk.m
This program finds the solution of the equation


12(k 2 + 2 ) 5k(k cos 6 + sin 6)
k = ln
17(k 2 + 2 ) 5k(k cos 5 + sin 5)

(M2.2)

30

2 MATLAB files used in the exercise solutions

which occurs in Exercise 9.5 by choosing an initial guess k0 = 0 and then

setting


12(kn2 + 2 ) 5kn (kn cos 6 + sin 6)
.
kn+1 = ln
17(kn2 + 2 ) 5kn (kn cos 5 + sin 5)
If kn+1 = kn then this gives a solution of (M2.2). The program performs 20
iterates and ends up with a reasonable approximation.
%% find k for Exercise 9.5(ii)
%% first guess
k=0;
w=pi/12;
%% change N for more iterations and so
%% a more accurate value for k
N=20;
for i=1:N;
dn=12-(5*k*(k*cos(6*w)+w*sin(6*w))/(k^2+w^2));
nn=17-(5*k*(k*cos(5*w)+w*sin(5*w))/(k^2+w^2));
x=dn/nn;
nk=-log(x);
k=nk
end

findt0.m
Also used for Exercise 9.5, this program finds the solution t of the equation


17(k 2 + 2 ) 5k(k cos 5 + sin 5)
1
t = 7 + ln
k
34(k 2 + 2 ) 5k(k cos (t 2) + sin (t 2))
by choosing an initial guess t0 = 2 and then setting


1
17(k 2 + 2 ) 5k(k cos 5 + sin 5)
tn+1 = 7 + ln
.
k
34(k 2 + 2 ) 5k(k cos (tn 2) + sin (tn 2))
Although the first line sets the value of k to the first four decimal places of
that found by findk.m, you could erase this line and run findt0.m straight
after findk.m for a more accurate value.

MATLAB files used in the exercise solutions

31

%% find time of death for Exercise 9.5(iv)

%% value of k from findk.m
k=0.3640;
%% initial guess
t0=2;
w=pi/12;
%% change N for more accurate value
N=20;
for i=1:N;
dn=17-(5*k*(k*cos(5*w)+w*sin(5*w)))/(k^2+w^2);
nn=34-(5*k*(k*cos(w*(t0-2))+w*sin(w*(t0-2))))/(k^2+w^2);
x=dn/nn;
nt=7+(log(x)/k);
t0=nt
end

param.m
This is the solution to Exercise 4.1(v). It draws a number of curves given
parametrically by
x(t) = Bet + Atet

and

y(t) = Aet ,

for A and B taking integer values between 3 and 3.

%%

Solution to Exercise 4.1(v)

t=linspace(0,5);
hold on
for A=-3:3;
for B=-3:3;
x=B*exp(-t)+A.*t.*exp(-t);
y=A*exp(-t);
plot(x,y)
end

32

2 MATLAB files used in the exercise solutions

end
hold off

renormalised.m
This program combines parts of f2.m and logistic.m in order to draw cobweb diagrams for the renormalised version of the logistic map (see Exercise
24.3 and its solution).
%% Draw cobweb diagrams for renormalised version of logistic map
hold off
r=input(r=: );
plot([0 1],[0 1])
nxl=1-(1/r);
g=inline((r^2*x*(1-x).*(1-(r*x*(1-x))))-1+(1/r),x,r);
xl=1-(1/r)+0.00001; xr=1;
d=1;
while d>0.0001
m=(xl+xr)/2;
if g(xl,r)*g(m,r)<0
xr=m;
else
xl=m;
end
d=(xr-xl);
end
nxr=m;
xp=linspace(0,1,100);
xv=nxl+xp*(nxr-nxl);
yv=r.*xv.*(1-xv); yv=r.*yv.*(1-yv);
yp=(yv-nxl)/(nxr-nxl);
hold on
plot(xp,yp,linewidth,2)
axis equal; xlim([0 1]); ylim([0 1]);
z=input(z_0=: );
n=1;
while n~=0
n=input(number of iterations (0 to stop, .1 to clear, .2 to change IC): );
if n==0.1
hold off; plot(x,x); hold on; plot(xp,yp, LineWidth, 2); axis equal;

MATLAB files used in the exercise solutions

xlim([0 1]); ylim([0 1])
end
if n==0.2
z=input(new initial condition: );
end
if n~=0
for j=1:abs(n)
za=nxl+z*(nxr-nxl);
nza=r*za*(1-za); nza=r*nza*(1-nza);
nz=(nza-nxl)/(nxr-nxl);
if n>0
plot([z z],[z nz])
ms=min(8,25*abs(nz-z));
if z<nz plot(z,(z+nz)/2,^,markersize,ms)
else
plot(z,(z+nz)/2,v,markersize,ms)
end
plot([z nz],[nz nz])
if z<nz plot((z+nz)/2,nz,>,markersize,ms)
else
plot((z+nz)/2,nz,<,markersize,ms)
end
end
z=nz;
end
end
end

rungekutta.m
This file implements the Runge-Kutta scheme
xn+1 = xn +

h
(f1 + 2f2 + 2f3 + f4 )
6

with f1 , . . . , f4 given by
f1 = f (xn , tn )
f2 = f (xn + 12 hf1 , tn + 12 h)
f3 = f (xn + 12 hf2 , tn + 12 h)
f4 = f (xn + hf3 , tn + h),
when f (x, t) = t x2 , as required by Exercise 21.8.
%% Runge-Kutta scheme
T=12;
h=0.5;

33

34

2 MATLAB files used in the exercise solutions

t(1)=0;
x(1)=0;

%% initial time
%% initial condition

for n=1:T/h;
t(n+1)=n*h;
f1=t(n)-x(n)^2;
f2=(t(n)+(h/2))-(x(n)+(h*f1/2))^2;
f3=(t(n)+(h/2))-(x(n)+(h*f2/2))^2;
f4=(t(n)+h)-(x(n)+h*f3)^2;
x(n+1)=x(n) + h * (f1+(2*f2)+(2*f3)+f4)/6;
end
[t x]

%% display values

%% Plot crosses at numerical values, and join these

plot(t,x,x,MarkerSize,20)
hold on
plot(t,x)

temperature.m
This program, also for use with Exercise 9.5 (as are findk.m and findt0.m)
will calculate the temperature at time t2 given the temperature at a time
t1 with a specified value of the parameter k, given by the formula


k

T (t2 ) = + ak 2
cos (t2 ) + 2
sin (t2 )
k + 2
k +


ak
+
T (t1 ) 2
(k
cos
(t

)
+

sin
(t

))
ek(t2 t1 ) .
1
1
k + 2
(which is the solution of Exercise 9.4). This can be used to check the value
of k found in part (ii), and then that the initial temperature found in part
(iv) is also correct.
%% Calculate temperature when given by solution of Exercise 9.4
%% input is k, t1 (initial time), and t2 (final time), and T(t1)
m=3; a=5; f=2; w=pi/12;
% k=0.3640; % this is value from Exercise 9.5(ii)
k=input(k = );
t1=input(t1 = );
t2=input(t2 = );

MATLAB files used in the exercise solutions

35

Tt1=input(T(t1) = );
u=k^2+w^2;
T=m+a*k*(k*cos(w*(t2-f))+w*sin(w*(t2-f)))/u;
T=T+((Tt1-m-(a*k*(k*cos(w*(t1-f))+w*sin(w*(t1-f)))/u))*exp(-k*(t2-t1)))