Reading in dynamic (polarization) charges from a trajectory?

[hsantila]

Problem

I have a trajectory where charges change dynamically in time in a polarizable surface. The charges are printed out in a lammps dump trajectory. Is there a way of reading these in and assigning to the atom charges?

Expected behaviour

The manual seems to indicate that there is a possibility to read in charges from a lammps trajectory. The source code suggest that this could happen via additional_columns keyword in the lammps dump reader.

Actual behaviour

After trying multiple ways of reading in the trajectory via mda universe, the charges remain constant at the values they are in the topology. When using the additional_columns keyword, the ts.data field does not contain any additional columns. This is true for all ways of formatting the trajectory (that is, ordering the trajectory columns) I tried out.

Code to reproduce behaviour

#u = mda.Universe('graphene_water.lmp','cpm_full.xyz', topology_format="DATA", format="LAMMPSDUMP",atom_style="id resid type charge x y z", additional_columns=['q'])
#u = mda.Universe('graphene_water.lmp','cpm.dcd', topology_format="DATA")
u = mda.Universe('graphene_water.lmp','cpm.lammpstrj', topology_format="DATA", format="LAMMPSDUMP", additional_columns=True)
for ts in u.trajectory:
    charges = u.atoms.charges
    pos = u.atoms.positions
    print(f"Frame {ts.frame}, First 5 charges: {charges[:5]}")
    print(f"Frame {ts.frame}, First 5 coords: {pos[:5]}")
    print(ts.data)

Versions

MDAnalysis 2.7.0
Python 3.11.2
Distributor ID: Debian
Description: Debian GNU/Linux 12 (bookworm)
Release: 12
Codename: bookworm

Grateful for any tips & insights :)

[orbeckst]

Maybe try MDAnalysis 2.9.0? I see in the docs for LAMMPS.DumpReader that the arbitrary columns were supposedly implemented in 2.7.0 but the CHANGELOG says that this actually happened in 2.8.0 — sorry for the confusion!

[hsantila]

Thanks! Updating the version indeed did the trick.

On a semi-related note: the arbitrary columns seem to only work on numerical values. Reading in atom names doesn't work because there is a string to float conversion assumed upon reading.

[orbeckst]

If this worked, can you mark as answer? This helps anyone else having the same problem (see also my general comment #4558 ).

To enable reading of string-attributes from arbitrary columns I suggest you raise an issue. I am not sure what the best way to proceed is (likely having to autodetect first or setting the data type — but this is something better discussed on an issue).

[hsantila]

Will do! Thanks again!