[BUG] Why does the DOS go less then zero and why do they have negative frequencies?

[ChiCheng45]

Describe the expected result As far as I understand it the DOS should never be negative but sometimes you can get negative DOS in MDANSE. Also in MDANSE the DOS includes negative frequencies, why is this done?

Additional details Here is the DOS calculated from the CASTEP test trajectories (croco.md)

[ChiCheng45]

This also occurs on the old MDANSE

[ChiCheng45]

Negative values of the DOS are due to the details of the FFT. Similar results are seen with function like exp(-ax^2)cos(x) with small sampling windows. This issue is probably expected and usually goes away with the instrument resolution setting applied.