# MATLAB Help!, solving simple PDE with ODE45 or ODE23 solver in matlab

1. Jan 22, 2010

### kribosgiros

The equation i'm trying to solve is

dq(x,t)/dt=-c*dq(x,t)/dx

with initial condition for example q(x,0)=exp(-(x-5)^2)

c is a constant. What i want to do is to first discritize the initial condition with interval dx then because i have to know the value of -c*dq/dx, i use fourier transform in matlab then i multiply it by -c*i*k after that i invert the fourier transform back so i get the value of -c*dq/dx at every point in my grid. Then i solve it with matlab ODE solver such ODE45 but result is still incorrect, i don't know where i might have gone wrong, here is my code

******the main code********
N=2^14;c0=0.5;
dx=0.01;x=[0:dx:(N-1)*dx];%defining the dx for x
tspan=linspace(0,20,100);%defining the time span for the solution
%defining the initial condition
q0=exp(-(x-5).^2);
dqx = diskx(q0,dx,N,c0);
%calculating the solution
ode=@(t,q) rhs(t,diskx(q0,dx,N,c0)*q);
options = odeset('RelTol',1e-5,'AbsTol',1e-6);
[t,q] = ode23(ode,tspan,q0);
%animating the result
for i=1:length(tspan)
h=plot(x,q(i,;axis([0 164 -10 11]);
name=['Simpangan q(x,t) saat t=',num2str(t(i))];
title(name);
pause(0.1)
set(h,'EraseMode','xor');
end

******function diskx********
function dxq = diskx(q0,dx,N,c0)
dk=2*pi/((N-1)*dx);
k=[-(N-1)/2*dk:dk:(N-1)/2*dk];
Q0=fft(q0);
Q0T=fftshift(Q0);
dQ0T=-c0*i*k.*Q0T;
dQ0=ifftshift(dQ0T);
dxq=real(ifft(dQ0));

******function rhs********
function dqdt = rhs(t,dqx,q)
dqdt = [dqx'];

please give me hint or correct me if i'm wrong, because i've checked and rechecked that the result suppose to be correct and yet i'm still not getting what i want which is just a moving function with the speed c. thx guys

2. Jan 22, 2010

### kosovtsov

Wherefore such complexity? Your equation has well-known general solution

q(x,t) =F(x-t*c), where F is an arbitrary function.

Substituting t=0 you obtain that

F(x)=exp(-(x-5)^2)

so solution with your initial condition is

q(x,t) =exp(-(x-t*c-5)^2)

3. Jan 22, 2010

### kribosgiros

that is correct, but this is just the first step from several steps that i want to do. Later on i want to define c as a function of k, which as we know only exists in fourier domain this is why i try to solve it through fourier doamain . And this kind of relation is called the dispersion relation in water waves

4. Feb 4, 2010

### matematikawan

So how's going? Have you sort out the problem?

Your problem looks interesting. But I have figure out anything yet. Need to have some time to understand it.

5. Feb 18, 2010

### kribosgiros

i got it hehe.. finally, the problem is i just compute my diskx function onece when i should compute it for each time step, and i notice that i don't really need the rhs function so i erase it and put my diskx directly inside my ode45 or 23 bracket and it works perfectly.

6. Feb 18, 2010

### kribosgiros

oh and don't forget to change q0 in diskx to q so it compute the dynamics