Uploaded September 2021 | Updated November 2026, 19 hours ago
Shown is an initially prepared Gaussian wave packet moving in a cyclotron orbit under exposure of a constant homogeneous magnetic field at two different configurations.
@gpcbass has, as always, produced the fantastic music. The song is called NYH.
Further details:
1. Absolute value squared of a wave packet in a relatively small two-dimensional region for a certain magnetic field strength.
2. Real part of the motion in 1.
3. Absolute value squared of the same wave packet having the same initial momentum in a larger two dimensional region for a weaker magnetic field strength.
Note: I'm still a bit suspicious of the physical validity of the animation. I've tried both Landau gauge and symmetric gauge for the magnetic vector potential with the same result, though.
An ordinary finite-difference approximation is utilized to approximate the time independent Hamiltonian (five bands = very sparse), whereafter the propagation is done with Cayley's approximation for the matrix exponential popping out when straightforwardly solving the Schrödinger equation. The resulting linear system of equations are solved iteratively, allowing systems having Hamiltonians significantly larger than 1Mx1M without memory issues. This is done for each time stamp. The final part corresponds to a 700k x 700k Hamiltonian. Total computation was a couple of hours, where saving of frames take significant time.
Shown is an initially prepared Gaussian wave packet moving in a cyclotron orbit under exposure of a constant homogeneous magnetic field at two different configurations.
@gpcbass has, as always, produced the fantastic music. The song is called NYH.
Further details:
1. Absolute value squared of a wave packet in a relatively small two-dimensional region for a certain magnetic field strength.
2. Real part of the motion in 1.
3. Absolute value squared of the same wave packet having the same initial momentum in a larger two dimensional region for a weaker magnetic field strength.
Note: I'm still a bit suspicious of the physical validity of the animation. I've tried both Landau gauge and symmetric gauge for the magnetic vector potential with the same result, though.
An ordinary finite-difference approximation is utilized to approximate the time independent Hamiltonian (five bands = very sparse), whereafter the propagation is done with Cayley's approximation for the matrix exponential popping out when straightforwardly solving the Schrödinger equation. The resulting linear system of equations are solved iteratively, allowing systems having Hamiltonians significantly larger than 1Mx1M without memory issues. This is done for each time stamp. The final part corresponds to a 700k x 700k Hamiltonian. Total computation was a couple of hours, where saving of frames take significant time.









![Fourier expansion of simple geometries [+MWE Python code]
Shown is some simple geometries drawn by epicycles, i.e., represented by Fourier series. See pinned comment for a minimal working example code (Python).
The ellipse of course only needs two rods to coincide with the exact solution. However, I was a bit surprised that the square and the triangle were so well approximated using only two rods. When representing the three (disconnected) geometries by a single basis set the number of basis functions must be large. The discontinuities naturally induce Gibbs phenomenon. Ive cheated a bit here, and removed a small portion of the traces in connection with the discontinuities.
Reuse of the fantastic song Vanessa, made by @gpcbass. See https://www.youtube.com/watch?v=yE8G0kmfzjc for the full version. Fourier expansion of simple geometries [+MWE Python code]](https://i.ytimg.com/vi/V6bQoEjnCKk/mqdefault.jpg)
