# what is my problem using split operator methods on gaussian wave packet time evolution in free space

5 views (last 30 days)
Daniel Niu on 17 Sep 2022
Commented: William Rose on 19 Sep 2022
dt = 0.005; % Size of time step
tfinal = 2; % Final time
M = round(tfinal/dt); % Total number of time steps
N = 512; % Number of Fourier modes
L = 20; % Space period
h = L/N; % Space step
n = (-N/2:1:N/2-1)'; % Indices
x = n*h; % Grid points
%initial wavefunction
u0=exp(-1.*(x).^2);
u = u0; % Initial Condition
k = 2*n.*pi/L; % Wavenumbers.
Vx=0; %potential function
ur=exp(-1i*dt/2.*Vx); %U(r) operator
uk=exp(-1i*dt.*k.^2/2); %U(k) operator
for m = 1:1:M % Start time loop
u=ur.*u; %advancing in real space dt/2
c = dt * fftshift(fft(fftshift(u))); % Take Fourier transform
%c=fftshift(fft(u))*dt;% Take Fourier transform
c=uk.*c; %advancing in fourier space dt/2
u=ur.*c; %advancing in real space dt/2
end
figure;plot(x,u0,'k',x,u,'r','LineWidth',2)
Warning: Imaginary parts of complex X and/or Y arguments ignored.
The time evolution of a gaussian wave packet in free space should be a gaussian wave with spread width, what is the problem about my code?
Your help would be highly appreciated!

William Rose on 17 Sep 2022
Plot abs(u), instead of u, and you will see that the transformed signal is still Gaussian, as expected.
.
William Rose on 19 Sep 2022
@Daniel Niu, you are welcome. This problem and the solution are interesting. ur is simply unity. uk is a phase shift only, since the magnitude of uk is unity. The amount of phase shift for each dt (i.e. the rate of phase change) is proportional to the wavenumber squared (k^2).
Each sinusoidal component in the initial Gaussian waveform has phase angle=0. As time progresses, the phase angle of each sinusoid changes, but the magnitude stays the same. As the phase angles change, they interfere with each other in a different way than they did initially. Therefore , even though each sinusoid has the same magnitude and frequency as before, their sum results in a Gaussian which broadens with time. And, as you observed, the wavefunction, withich was purely real initially, becomes complex valued. Good luck!