seemingly-wrong coordinates

[Elliot Menkah]

Hello Everyone,

I'm interfacing i-pi with a linear-scaling DFT code, ONETEP, and i'm getting seemingly-wrong coordinates sent to the client code(ONETEP) from i-pi.

When i print out the data in from sendpos function in i-pi before sending, the data doesn't correspond with what I'm expecting, yet, the data written into the "RESTART" under the "q" tag corresponds to what's expected.
So the question is, besides converting the read atomic coordinates(if in angstrom) into bohr units, does it perform other preprocessing to the coordinates before sending them to the client code.

If it does, what does it do?

thank you.

kind Regards,

Elliot.

Master in High Performance Computing(MHPC)
International School for Advanced Studies
Scuola Internazionale Superiore di Studi Avanzati(SISSA)
The Abdus Salam - www.ictp.it
International Centre for Theoretical Physics
Strada Costiera, 11 - 34151 Trieste - IT
Italy

[Elliot Menkah]

Hi Everyone,
I've got a solution to the problem. It was a pbc switch which was controlling that. So i set pbc="false" under the socket tag in the xml input file and I get the expected coordinates, thanks to Venkat Kapil.

Thank you all.

[Mariana Rossi]

Here it may be worth setting the default of this flag to false? Except for LAMMPS is there any other code that requires this to be true?

[Michele Ceriotti]

Open for discussion. Actually we could change the driver side in LAMMPS so that it applies PBC from within, at that point it would totally make sense to make pbc=false the default. 

[Igor Poltavsky]

Hi,

pbc=true does not work also for any force field where a unit particle consists of several atoms since it breaks molecules into parts. An example is qtip4pf model which is already implamented in i-Pi.

I think the default "false" option would generate less errors.

Best wishes

Igor