Triclinic box skew is large. How to decide the replicate parameter

Dear all,

I am attempting to simulate a MOF with a triclinic box and have encountered the following warning:

triclinic box = (0 0 0) to (62.8399 54.42095 51.308563) with tilt (31.41995 31.41995 18.140317)

WARNING: Triclinic box skew is large. LAMMPS will run inefficiently.

My LAMMPS version is: LAMMPS (7 Feb 2024 - Update 1)

The parameters for my box are as follows:

0.000000 62.839900 xlo xhi
0.000000 54.420950 ylo yhi
0.000000 51.308563 zlo zhi
31.419950 31.419950 18.140317 xy xz yz

The cell lengths are:

  • a: 62.8399
  • b: 62.8399
  • c: 62.8399

And the cell angles are:

  • alpha: 60
  • beta: 60
  • gamma: 60

Initially, I considered transforming the triclinic box to orthogonal. As suggested here, I can’t.

Upon further investigation, I discovered in the LAMMPS documentation that the xy and xz tilt factors should not exceed half the distance of the parallel box length, here, xy, xz < 31.41995.

I also explored discussions on similar issues here and here, where it was suggested to build a supercell by replicating the unit cell and adjusting the tilt factor. However, I’m puzzled as to how this would resolve the problem. My understanding is that using the replicate command copies the unit cell in three directions, without altering its structure.

I attempted to use “replicate 2 1 1” and “replicate 3 3 3”, but they did not solve the issue.

As geometry is not my forte, I would appreciate it if you could explain how to choose the replicate parameters and why changing the tilt factor is necessary. Alternatively, if you could suggest a book or website where I could learn more about this topic, it would be greatly appreciated.

Below are my input files.

data.MIL-101-primitive (1.4 MB)
in.MIL-101-primitive (664 Bytes)

Thank you,
Best regards,

You cannot use replicate for your system, since you have bonds that reach across periodic boundaries. The replicate command does not handle this kind of situation.

The simplest approach would be to construct a conventional cell instead of a primitive cell. How to do that, would be explained in textbooks on crystallography.

The reason why LAMMPS prints a warning is that with a tilted box you may need to have a larger communication cutoff to make sure you have all atoms within the cutoff. Also, due to the tilt you have more surface per volume and thus need to do more communication. For as long as you don’t change the tilt (e.g. with fix npt where all cell parameters are free), you can run simulations at the current tilt correctly. The risk of missing interactions (or having bond atom missing errors) grows the flatter the system gets.

I will only do NVT simulation, so I guess this is acceptable. And I will try to learn more about crystallography. Thank you for your suggestion!