augmented basis sets and cholesky issues

[Noam Bernstein]

Hi - I'm trying to switch from a DZVP basis to the an augmented

version, specifically aug-DZVP-GTH from the GTH_BASIS_SETS

file in the cp2k/tests/QS directory.

1. Is there a better, or more standard, augmented double-zeta basis I could use?

2. When I do use this basis (instead of DZVP-GTH-BLYP), I get this error

 *****************************************************************************

 *** 12:13:15 ERRORL2 in cp_fm_cholesky:cp_fm_cholesky_decompose processor ***

 *** 0  err=-300 condition FAILED at line 102                              ***

 *****************************************************************************

 ===== Routine Calling Stack ===== 

           11 cp_fm_cholesky_decompose

           10 make_full_all

            9 make_preconditioner

            8 prepare_preconditioner

            7 init_scf_loop

            6 scf_env_do_scf

            5 qs_energies_scf

            4 qs_forces

            3 velocity_verlet

            2 qs_mol_dyn_low

            1 CP2K

 CP2K| condition FAILED at line 102

 CP2K| Abnormal program termination, stopped by process number 0

What if anything can I do about this?  The positions come from an

MM calculation.  Do I have to switch to a different preconditioner?

Is this likely to be an issue with the (well equilibrated) MM initial

positions?

        thanks,

        Noam

[hut...@pci.uzh.ch]

Hi

if you are running a condensed phase system (or just a very
large molecule) augmented basis sets cause problems of
overcompletness. The overlap matrix gets illconditioned and
for example the Cholesky decomposition fails.
What can you do:
- try CHOLESKY OFF in CP2K_INPUT / FORCE_EVAL / DFT / SCF
and tune EPS_EIGVAL to a good value for your application
(same input section)
This is what is usually done in QC codes
- Augmented basis sets are often not needed in condensed
systems (the overcompletness is a hint), maybe it is better
to use triple zeta or higher. Surfaces are tricky as they
have both types of environments.
- MOLOPT basis sets. They were developed to have diffuse
components in the basis and at the same time still have
a well conditioned overlap matrix (see BASIS_MOLOPT)

regards

Juerg Hutter

--------------------------------------------------------------
Juerg Hutter Phone : ++41 44 635 4491
Physical Chemistry Institute FAX : ++41 44 635 6838
University of Zurich E-mail: hut...@pci.uzh.ch
Winterthurerstrasse 190
CH-8057 Zurich, Switzerland
---------------------------------------------------------------

-----cp...@googlegroups.com wrote: -----
To: cp2k <cp...@googlegroups.com>
From: Noam Bernstein
Sent by: cp...@googlegroups.com
Date: 11/13/2012 07:20PM
Subject: [CP2K:4184] augmented basis sets and cholesky issues

--
You received this message because you are subscribed to the Google Groups "cp2k" group.
To post to this group, send email to cp...@googlegroups.com.
To unsubscribe from this group, send email to cp2k+uns...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/cp2k?hl=en.

[Noam Bernstein]

Thanks Juerg - I appreciate the help. I'm trying to compare

with a published calculation that used some sort of augmented

basis (still trying to find out exactly which one), even though

my system is a molecule (with charged groups) in a solvent.  I'll

try turning off Cholesky and the alternative bases you suggested.

      Noam

[c.pignedoli]

dear all sorry for this mistake

On 14/nov/2012, at 16:11, "c.pignedoli" <c.pig...@gmail.com> wrote:

si prospetta lo scenario disegnato dal nostro capo:

se vuoi chiudere una attività difficile da chiudere

devi assumere una persona molto giovane (e brava) che 

assuma solo gente più vecchia di lui

in modo che alla fine resti solo lui

a chiudere la porta

--