definition of cell parameters and fhiaims

[manuperezesc]

Dear users, 

i have been recently trying to perform calculations using i-pi coupled to FHI-aims. I have, however, not succeeded yet because I suspect that I am not providing the cell parameters properly to i-pi.

This are the lattice vectors of my optimized slab, on top of which I can perform FHI-aims calculations normally:

lattice_vector     28.20155506      0.00337267      0.00549278
lattice_vector    -14.10891688     24.43476662      0.01100276
lattice_vector      0.00000000      0.00000000     60.00000000

And this is how I am building the input.xml file for i-pi:

<cell mode="manual" units="angstrom" > [28.20155506, -14.10891688, 0.00000000, 0.00337267, 24.43476662, 0.00000000, 0.00549278, 0.01100276, 60.00000000] </cell>

In the documentation I found that the lattice vectors should be given in upper triangular form in column format. I assumed that the y and z components of a and the z component of b can be neglected. I am encountering issues with the very first scf from FHI-aims that I do not get when using FHI-aims alone, so I guess I am doing something very naively, but I don't see what.

If anyone could help, I would really appreciate it.

Thank you and best regards,

Manu

[Mariana Rossi]

Hi there, I am not sure that this is the problem you are getting - please share inputs and outputs. Check what lattice vector is reported in FHI-aims. it is likely that everything is breaking for another reason. 

[manuperezesc]

Dear,

Thank you for your prompt response. The error that I get in FHI-aims is: *** Ill-conditioned overlap matrix found.

Please find attached all files I used, including the input for i-pi and the output from aims. On it, I do not see any information about the lattice vectors apart from that read and ignored from my geometry.in. If I run a standard calculation, I do not get that error, which makes me feel that it has to be with the coordinates and the lattice vectors, but I am not sure.

Thank you,

Manu

[Mariana Rossi]

Hi there, sorry for the late response - I missed you had sent the files! Thanks for reminding me.

It seems to me that opt.xyz is listing atomic positions in angstroms. If so, you have to specify. If you don't specify, i-PI assumes atomic units (bohr radius). See https://docs.ipi-code.org/units.html#units-conventions

We recommend the best way to do this is to specify directly in opt.xyz. You can do that by adding the string "positions{angstrom}" to the comment line of the xyz file. There are other ways to do it for ASE-type extxyz files or in the input file, check our examples folder.

Let me know if this solves your problem and happy to check back if you still find something wrong.

Cheers,

Mariana