DOS, Bandstructures, Relaxations » Geometry extraction?

FLEUR Forum � User support � DOS, Bandstructures, Relaxations � Geometry extraction?

Topic settings

Jump to forum

Announcements New releases, new features, important bugs User support Installation Inpgen -- the FLEUR input generator Problems while running SCF or other basic tasks DOS, Bandstructures, Relaxations Magnetism, Non-collinear calculations, SOC Wannier functions Other problems FUTURE related problems General discussion Feature requests Physics with FLEUR Open discourse

#1 | Geometry extraction? datum

Fri May 21, 2021 3:07 pm

Does anyone have any suggestions for tools to extract the trajectory out of a structural optimisation? I could hack together some tools to pull info out of relax.xml and sym.xml to reconstruct a full set of Cartesian coordinates, but surely there are already suitable tools around...

reply

#2 | RE: Geometry extraction? datum

Sat May 22, 2021 9:37 am (Last edited: Sat May 22, 2021 9:47 am)

1. Let me first warn you that this trajectory is nothing representing real physics. It represents different steps of the respective optimization algorithm. It also hopefully looks boring. But if not it may be used to help understanding problems in the relaxation process.

2. Having said this the most convenient way of extracting the trajectory into something that is easy to understand is to know this before the relaxation and store the struct-relax.xsf file created in every relaxation step somewhere. These files can be opened in a text editor and are easy to understand. The units in it are Angstrom and the atom positions are in cartesian coordinates. You can also open them with some visualization software that can read xsf files (XCrysDen, Vesta, ... ). They contain the complete unit cell description after the displacement of the respective relaxation step. But the file always gets overwritten by the next relaxation step. So knowing that you need these files before is essential. The input generator also generates a struct.xsf file for the initial geometry.

reply

#3 | RE: Geometry extraction? datum

Sat May 22, 2021 1:41 pm

Hi Gregor

Fear not! I was using the term trajectory in the loosest way possible. struct-relax.xsf is good enough for my purposes, thanks.

reply

#4 | RE: Geometry extraction? datum

Tue May 25, 2021 7:40 am (Last edited: Tue May 25, 2021 7:44 am)

Hi Terry,

The way you can/we do this is to go over the out.xml bzw, the inp.xml information. During relaxations everytime you have a new struct-relax.xsf you also have a new inp.xml. We need the atomic groups information to know to which atoms the forces belong to. The symmetries alone won't help you much. Therefore for a real trajectory, the struct-relax.xsf alone won't help you.

The fleur parsers in masci-tools have parsers for fleur files see (remove the space in the link):
https:// masci-tools.readthedocs.io/en/latest/user_guide/fleur_parser.html#parser-for-the-fleur-inp-xml-file

This includes functions 'get_relaxation_information' and 'get_structure_data' which gets you besides the structure (geometry), the atom group information and what is relaxed and what not. From this you can easily create a new ASE, aiida, or pymatgen structure object. The forces we also parse. This is still a bit low level, but we do not have yet implemented putting that back together to a full trajectory.

@Gregor: actually extraction a trajectory of the positions together with the forces is an open todo issue for the relaxation worklfow in aiid-fleur. For AiiDA the tricky thing will be to match the fleur atomtypes back to the original positions (pre inpgen).

@Terry: if you like to contribute you could implement putting this information (forces and structure) together to a full real trajectory in python using the information returned from the parsers, given a list of out.xml files and ggf force files. We would help you of course.

reply

#5 | RE: Geometry extraction? datum

Tue May 25, 2021 11:48 am

@Jens: If you also control the inpgen input you could work with labels for the atoms. Or is there a misunderstanding on my side?

reply

#6 | RE: Geometry extraction? datum

Tue May 25, 2021 3:07 pm

Zitat von JensB im Beitrag #4
Therefore for a real trajectory, the struct-relax.xsf alone won't help you.

But struct-relax.xsf has all the information needed to reproduce the geometry. You don't directly have symmetry, but I don't see why that's not enough for most purposes.

I have my own tools to stitch such things together, with permutations & shifts if necessary. Fortran & awk are my hammers, so I suspect you don't want my code.

reply

Stagnation in the SCF loop after several relaxation steps �