[Pw_forum] Total forces not converging: relax calculation

[Daniel Reta]

Dear Josue,

Thank you for your answer. If I may, Ill answer point by point.

Concerning the representation of the system in XCrysden. I am using crystal_sg to specify the coordinates, indicating the appropriate space_group. When I open the output file of a converged scf with XCrysden, the cell contains the atoms as the original cif files. The only difference is that when I try to switch the Unit of Repetition from Unit Cell to translational asymmetric unit, nothing changes and the atoms outside the unit cell do not appear in the interior. I assumed this had to do with the fact that I'm reading the output; Is there something wrong in this?

Concerning the type of calculation. I want to compute phonon dispersion at the gamma point. That is why I am using norm-conserving pseudopotentials. And in a first approach I do not want to change the cell parameters, only to relax the atomic coordinates in the unit cell. So yes, I would stick to 'relax' calculation.

The system contains a lanthanide atom, which might be the cause for the behaviour different from the standard cases. Is there any particular keyword I could play with to affect the convergence of the total forces?

Cheers

========================

Dr. Daniel Reta,
Post-doctoral Research Associate,
Computational Chemistry, School of Chemistry
University of Manchester

Email: daniel.reta at manchester.ac.uk

========================

________________________________________
From: pw_forum-bounces at pwscf.org [pw_forum-bounces at pwscf.org] on behalf of Josue Itsman Clavijo Penagos [jiclavijop at unal.edu.co]
Sent: 24 April 2017 16:41
To: PWSCF Forum
Subject: Re: [Pw_forum] Total forces not converging: relax calculation

Dear Daniel,

It would be more useful if you post the input file you're using, or at least the atomic coordinates. Also, check if the system is well represented by the coordinates you are intending to relax, since a well-defined system usually converges at E_cutoff values as low as 60 Ryd for simple SCF calculations.  (Does he system look as you expect when viewed in XCrysden, for example?) , unless you include rather demanding features like spin polarization, electric fields, defects and so on. By the way, Are norm-conserving PP's the only choice?(how about Ultrasoft?)

Moreover, what type of relax calculation do you want to study? judging by the info you sent, it is 'relax', but, Wouldn't be that your system need some cell parameters/angles relaxing (vc-relax), not just atomic coordinates one?

Best regards,


Prof. Josué Clavijo, PhD
Cristiano, born again

Assistant Professor
Universidad Nacional de Colombia
Sede Bogotá