Reconstructed FLiMESolve and FLiMESolver to play more nicely with other QuTiP functions. Wrote FLiMESolver into the correlation tab, but had to add taulist as an input into the _make_solver function.

Trying to rework the time list stuff in FlimeSolve/er to work better with QuTiP. Namely, autogeneration of the time list to create the rate matrix & trying to avoid time list in solver.run so I don't have to do anything kludgey to get the modes in an expedient way

Rewrote the mode table in solver.run() to be found in a slightly slower, but much more generalized fashion

Figured out how to do Expectation operators with my construction of the solvers, and I added a constraint the to the available frequencies in the relaxed secular approx. based on the Nyquist theorem.

Floquet beta

Working

Added flimesolve to dynamics-floquet documentation, and added floquet_ex4 as an example usage (identical to floquet_ex3 with exception of use of  flimesolve instead of fmmesolve)

Added towncrier changelog...I think

…to qutip-5.0.X

Got rid of a few commented lines that were not useful

Deleted folder of scripts I'm using to write a paper about this solver

I fixed linting, extra spacing, and blank line issues. Moved all new functions to flimesolve.py. Replaced scipy functions with Numpy equivalents. Fixed issues in correlation.py by updating to my most current version, which should just introduce flimesolve.py as an additional solver.

Commented out a [0] in line 91 for some reason. I uncommented it now to bring it better in line with correlation.py

Fixing code format

Co-authored-by: Eric Giguère <eric.giguere@calculquebec.ca>

Co-authored-by: Eric Giguère <eric.giguere@calculquebec.ca>

Co-authored-by: Eric Giguère <eric.giguere@calculquebec.ca>

Co-authored-by: Eric Giguère <eric.giguere@calculquebec.ca>

Used built-in autoformat in Spyder and then pycodestyle afterwards to fix code format to PEP8 guidelines.

…to qutip-5.0.X

Made a few changes caught by CodeClimate on Github

Fixed the dimensionality for FLiMESolve to properly deal with tensor product Hamiltonians, added flimesolve_test to the testing folder

Went through the differences between my master branch and the master branch on GitHub, reverted changes that didn't have to do with my code.

Fixed a small formatting issue in flimesolve

Co-authored-by: Eric Giguère <eric.giguere@calculquebec.ca>

Co-authored-by: Eric Giguère <eric.giguere@calculquebec.ca>

Killed/Simplified the Kronecker Deltas in the rate matrix building loop to reduce the number of calculations by...I think a couple orders of magnitude? Much faster now.

Rebuilt the rate matrix for loop to use masks instead of a for loop. Takes roughly 2% as long to construct, now.

Updated some formatting, commenting, and variable explanations to make code more legible and to better align with PeP8

Updated the Rate Matrix to use nothing but masks instead of a kludgy combination of masks and for loops. The end result is slightly faster (not as much as I had hoped, to be honest), and looks a lot cleaner.

For some reason, the Floquet Modes started getting wonky when they were stacked into an array using the old method I had that I think Eric just brought forward. When Nt (the number of points in one period of the Hamiltonian) was brought high enough, the modes ended up transposes around the anti-diagonal of the Hdim x Hdim single-time arrays. No idea why, but using the built in '.to_floquet_basis' function, while slightly slower, seems to have fixed the issue.

Fixed some issues including but not limited to:
Floquet transformation on collapse operators now goes to the MODE basis and not the STATE basis
Incorrect Nyquist limit filter removed - I had the wrong conceptualization here

Moved c_op_rates in the time dependent flimesolve to a more "central" location.

Fixed issue in included_delta to include the absolute value arguments. Fixes it such that TD FLiME no longer returns worse results than TI FLiME

Updated the "create_rhs" method in FLiMESolver so that it actually appends the correct frequencies to each rate matrix slice. Also updated the rate matrix list construction to be slightly less redundant.

Fixed an issue where the state evolution would "switch" (i.e. ground state decaying, excited state growing) in a pulsed system. Also cleaned up after myself a bit.

Took the fftshift of the collapse operator FFT, which seems to have fixed the "wiggly coherences" issue (along with fixing some previous basis transformations to compensate). Also cleaned up after myself a bit and added documentation to _floquet_mode_table and _floquet_state_table

Fixed an issue with a misplaced negative sign in the quasienergy lists used to construct the rate matrix.

Fixed Most of Eric's notes (not sure how to fix the step issue yet), and reformatted flimesolve.py to Black

Changed c_ops so that it should now accept collapse operators in the exact same manner as mesolve()

Changed initial options in flimesolve() from 'None' to '{}' i.e. an empty dictionary. This fixed an issue with trying to set the options indirectly, e.g. through a correlation function.

Removed typing from QObj initialization

Updated the 'FLiMESolver' solver in 'correlation.py' to reflect changes to inputs in 'flimesolve.py'

Fixed issues where 'test_flimesolve.py' was using an older version of the 'flimesolve' inputs

Adding manual workflow option so I can test in my own branch

Fixed documentation issue in dynamics-floquet.rst, and updated it with a short explanation and example of the time_sense argument.

Forgot to turn the returned Floquet States into a Qobj. Fixed that.

Committing all the changes this time

Co-authored-by: Eric Giguère <eric.giguere@calculquebec.ca>

Added flimesolve import

Updates test_flimesolve.py to include a test with nonzero time_sense value and to include a test for the correlation function. Also fixed some typos/added a few words in places.

Fixed some bad documentation on c_ops. Also added a bit more about which method is used depending on time_sense values.

Updated Correlation test to be much cleaner, more streamlined. Also removed a bunch of unnecessary numpy functions and replaced with QuTiP functions where applicable.

Updated Flimesolve with a steady state solver. This uses the rate matrix as an analogue for the Monodromy matrix to find the steady state of the system in the full secular approximation. Not sure if adding in more time-dependence (if possible) would do anything, as I think that would just cause the solution to orbit/rotate around the time independent result.

Attempting to fix the build error.

Co-authored-by: Eric Giguère <eric.giguere@calculquebec.ca>

Co-authored-by: Eric Giguère <eric.giguere@calculquebec.ca>

Co-authored-by: Eric Giguère <eric.giguere@calculquebec.ca>

Co-authored-by: Eric Giguère <eric.giguere@calculquebec.ca>

Co-authored-by: Eric Giguère <eric.giguere@calculquebec.ca>

Co-authored-by: Eric Giguère <eric.giguere@calculquebec.ca>

Co-authored-by: Eric Giguère <eric.giguere@calculquebec.ca>

Removed the steady state solver to add it in as a separate PR later.

Fixed some issues in the correlation function test

Fixed an fftshift causing state decay to move backwards. This was already fixed and made its way back in somehow. I...don't know.

The FFT is the only way to bring FLiMESolve to agree with MESolve. Additionally, FMMESolve displays the same issue as FLiMESolve when the driving amplitude is brought to zero. This makes me think that there's an eigenvalue/vector ordering issue here. Probably degeneracies....somewhere? It's also the only way to get all the indices to have signs that agree with the theoretical FLiMESolve.

Fixed FFTshift issue third time. Needed to flip the FFT'd collapse operators to get decay moving in the proper direction. Now FLiMESolve relaxes to MESolve for both pulsed and CW systems, and it even decays to zero like it should.