Rotations

[tfamprikis]

adding functions to track orientations issue #167

[stefsmeets]

Hi @tfamprikis thanks for this! This looks like it will be quite a useful starting point. Do you plan on adding something more? Otherwise we will start integrating this into gemdat.

[tfamprikis]

hey, this was just a test to see how this works.

i still have a couple of issues that havent been able to fix. mainly

• I cant seem to be able to symmetrize my directions (created by gr.bond_direct() ). this is what I'm trying to do in lines 28-59 of test4.py. It works in matlab but not in python...
• I've trouble with normalizing the probability plot (in polar coordinates) by the corresponding areas of the sphere, it becomes problematic near the poles, creating artifacts.

otherwise some other functions from test4.py can also be moved to gemdat_rotation (autocorr(), autocorr_par(), a function to plot distribution of bond lenghts)

I can add the data (.xml) i've been using to test to test_data and some figures of what the symmetrization looks like when it works.

all that said, if you think you can still work with what I have please go ahead...

[stefsmeets]

Okay, no worries! The test at leas worked 😅

There is a bunch of othes stuff I still want to work on as well. I'll let you know when we get back to this.

[SCiarella]

@tfamprikis Can you share the data that you used to run test4.py?
Also, have you made any further progress about the rotations that is not already here?

[tfamprikis]

@tfamprikis Can you share the data that you used to run test4.py? Also, have you made any further progress about the rotations that is not already here?

Hi @SCiarella

I've uploaded the .xml + a version of the test code that actually (half)works here:
https://surfdrive.surf.nl/files/index.php/s/vVIhHHMsOU9wFT5
(let me know if you have trouble accessing)

I have not managed to make meaningful progress against the issues i had described yet...

[SCiarella]

This pull request should fulfil the following objectives:

• compute the orientation vectors and their trajectory given the input expected structure
• apply the correct symmetry group to enhance the trajectory
• produce the rotolinear plot
• measure and plot the autocorrelation

The analysis of rotational transitions will be the object of the next pull request

[SCiarella]

You can look at this notebook to get an overview of this addition.

[stefsmeets]

Hi @SCiarella thanks for fixing up this PR! It looks like all the bits are in place, the plots look great, but I think it could use some polish before merging. I made a first pass through the code, please see my comments.

This seems to be the basic usage:

bd = Orientations(trajectory, central_atoms='S', satellite_atoms='O', n_expected_neigh=8)

direction_cart = bd.get_conventional_form(normalize=False)
direction_spherical = bd.cartesian_to_spherical(direct_cart=direction_cart, degrees=True)

plots.rectilinear_plot(trajectory=direction_spherical, normalize=False)

sg = Oh_point_group()
direction_cart_sym =  bd.get_symmetric_traj(sg.sym_ops_Oh)
direction_spherical_sym = bd.cartesian_to_spherical(direct_cart=direction_cart_sym, degrees=True)

plots.rectilinear_plot(trajectory=direction_spherical_sym, normalize=False)

• I'm wondering if we can clean up the api a little bit, so it's a bit more inviting to use
• Have you looked at the pymatgen symmetry module? There seems to be a lot of code that can help with generating symmetry operations for point groups. This could replace Oh_point_group with familiar pymatgen code.
• Can you also add some unit tests for the new code?
• Can you refactor the Orientations class to get rid of some of the superfluous methods(see comments)?

[stefsmeets]

We could use:

from pymatgen.symmetry.groups import PointGroup
g = symmetry.groups.PointGroup('m-3m')  # Oh
len(g.symmetry_ops)
# 48
g.symmetry_ops[0]
# SymmOp(affine_matrix=array([[ 0.,  0., -1.,  0.],
#       [ 0., -1.,  0.,  0.],
#       [-1.,  0.,  0.,  0.],
#       [ 0.,  0.,  0.,  1.]]))

[SCiarella]

I have cleaned the api, so now the basic usage looks like:

bd = Orientations(trajectory, central_atoms='S', satellite_atoms='O', n_expected_neigh=8, normalize_trajectory=False)

plots.rectilinear_plot(data=bd, symmetrize=False)

bd.set_symmetry_operations(sym_group='m-3m')

plots.rectilinear_plot(data=bd, symmetrize=True)

which is now using pymatgen.PointGroup to get the symmetry, so I have removed Oh_point_group.
I have also refactored and removed some superfluous methods as suggested.

[stefsmeets]

Nice work, using the symmetries from pymatgen makes this a lot cleaner. A few small final comments from my side.

[tfamprikis]

Hi guys, I've been playing with this a while. some implementations don't really make sense to me rn. should I comment here or wait for this to be merged and make issues?

[SCiarella]

@tfamprikis I think this pull request will be closed relatively soon (since it is also quite long), so it would be nice if you could open new issues to address all the problems that you see.

[stefsmeets]

LGTM 🚀

@tfamprikis I agree with @SCiarella that it's good to get this merged now and fix remaining issues later.