The ABINIT Users Mailing List ( CLOSED )

["Allan, Douglas C Dr" <AllanDC@corning.com>]

Whether you break the symmetry or not, make sure you choose occupancies
"occ" that are correct for the final set of degeneracies appropriate to
the final eigenstates.  These will be different depending on whether you
break symmetry or not.  You will definitely get convergence problems if
you choose occupancies that are incorrect for the final degeneracies.
That is, degenerate states have to have the same occupancy or else the
code can't decide which of the equivalent states should magically get a
different occupancy -- this is a recipe for convergence trouble.  If you
can guess or iterate to determine which states are degenerate in the
solution then, in my experience, you need not artificially break the
symmetry of the cell.  However, breaking the symmetry artificially is no
more artificial than using a supercell anyway and it is quicker to break
degeneracies if you can't guess them in advance.

Good luck,
Doug Allan 

-----Original Message-----
From: Chol-Jun Yu [mailto:yucj@ghi.rwth-aachen.de] 
Sent: Wednesday, January 02, 2008 2:49 PM
To: forum@abinit.org
Subject: Re: [abinit-forum] Convergence problem of isolated Ni atom

Hello,

thank you very much for your helpful comments.

Kris Andersen wrote:
> For other isolated atoms, I have found it is necessary to lower the 
> symmetry of the supercell. For example, use
> 
>     rprim  1 0 0 0 1 0 0 0 1.1
> 
> or even lower symmetry. This breaks degeneracies and allows the atom  
> to converge to a ground state with a higher spin moment. That said, I 
> haven't done a d10 atom like Ni, where you expect a ground state with 
> no spin moment, so this may not be the problem in your case.
> 
> Also, I'm not sure why you are including the symmetry operations 
> explicitly in the input file (nsym, symrel, and tnons) instead of 
> letting Abinit find them?

I thought that the symmetry operations would enforce the low symmetry to
the system.

By the way, don't you think the parameters such as diemac, diemix or occ
even iscf have the certain effect on the convergence in the case of
isolated atoms?

Best regards,
Chol-Jun
--
Chol-Jun Yu
Computational Materials Engineering (CME) Center for Computational
Engineering Science (CCES) Institute of Mineral Engineering (GHI) RWTH
Aachen University Mauerstr. 5, D52064 Aachen

email: yucj@ghi.rwth-aachen.de
phone: 49-0241-80-94989