Does pppm/disp support triclinic simulation boxes?

LAMMPS LAMMPS Development
1

Hello LAMMPS developers,

Thank you for developing and maintaining LAMMPS.

I am using LAMMPS 22 Jul 2025 - Update 3, and I would like to ask about the current triclinic-box support for pppm/disp.

For an orthogonal simulation box, the following setup works well:

kspace_style    pppm/disp 1.0e-6
kspace_modify   disp/auto yes

However, when I switch to a triclinic simulation box, I get the following error:

PPPMDisp initialization ...
ERROR: KSpace style does not yet support triclinic geometries (../kspace.cpp:182)

From the documentation, I understand that Coulomb PPPM supports triclinic simulation boxes, and ewald/london also appears to work with triclinic boxes. However, pppm/disp seems to be restricted to orthogonal boxes.

My questions are:

  1. Is this still the expected/current limitation of pppm/disp?

  2. Are there any plans to support triclinic boxes for pppm/disp in the future?

For my application, I am studying molecular-crystal phase transitions using a machine-learning potential combined with analytical long-range dispersion. A triclinic NPT simulation would be physically useful, but ewald/london is too slow for large production MD. Therefore, triclinic support for pppm/disp would be extremely valuable.

Thank you very much for your time.

2

yes.

there is currently no open feature request issue for that on GitHub and all
core developers are currently very busy to process a very large amount of
pending pull requests. So nobody is really looking for something to work on.

I can try to let an AI coding agent give it a try, but I cannot make any promises
that this will work. Our experiences are mixed. The coding agent has to find
enough reference material in the other kspace styles to come up with a viable
solution. It did work recently to enable the slab correction for PPPM with triclinic boxes.

3

Thank you very much for the clarification and for taking the time to answer.

I really appreciate that you are even willing to let an AI coding agent try this. Triclinic support for pppm/disp would be extremely useful for molecular-crystal simulations, so I would be very happy if it eventually becomes available.

Thank you again!

4

Please check out the pull request branch that was submitted earlier today and let us know how it pans out: Add triclinic-box support to pppm/disp (and its omp/tip4p variants) by akohlmey · Pull Request #5060 · lammps/lammps · GitHub

5

Thank you very much for implementing this so quickly.

I tested the pull request (#5060) on my system, where I had previously encountered

ERROR: KSpace style does not yet support triclinic geometries

The new implementation worked successfully, and I was able to run triclinic NPT simulations using pppm/disp without any issues. The results also appear physically reasonable.

I’ll continue testing it, and if I encounter any issues, I’ll report them here. So far, everything looks good.

Thank you again for adding this feature. It will be very useful for molecular-crystal simulations.