[Pw_forum] DFT+U and non-collinear calculations

[Paolo Giannozzi]

On Wed, 2014-04-30 at 11:41 +0200, Pietro Bonfa' wrote:

> So this looks like a bug in the 5.1rc1 version. I'll try to figure out
> why this happens with the help of diffs files.

this is not going to be simple, because I made extensive changes to the
implementation of DFT+U in order to speed it up. There is no test for
the DFT+U case with noncolinear magnetization, so problems may go
unnoticed for a long time (thank you for noticing it). If you are
willing to dig into the code: the problem apparently arises after 
rev. 10181,
http://qe-forge.org/gf/project/q-e/scmsvn/?action=browse&path=%
2F&view=rev&revision=10181

Paolo

> Thanks for your help,
> kind regards,
> Pietro Bonfa'
> 
> 
> On 04/30/2014 11:00 AM, Alex Smogunov wrote:
> > Dear Pietro.
> > I have tried your input with QE 5.0, it works fine for me, try it.
> > The problem could also arrise from diagonalization method though it is
> > strange...
> > Try to set diagonalization = 'cg'.
> > good luck,
> > Alexander
> > 
> > 
> > 
> > 2014-04-29 17:18 GMT+02:00 Pietro Bonfa' <pietro.bonfa at fis.unipr.it
> > <mailto:pietro.bonfa at fis.unipr.it>>:
> > 
> >     Dear Dr. Alexander Smogunov,
> > 
> >     the short digression on the non-linear core corrections was intended to
> >     explain why I need DFT+U to correctly reproduce the insulating and
> >     magnetic ground state of La2CuO4.
> > 
> >     I attached a modified version of example 11 (that I called example12).
> >     I set all the parameters to zero just to check if everything is working
> >     properly. I tried also with realistic values for U and J but I obtained
> >     the same errors.
> > 
> >     Thanks,
> >     kind regards,
> >     Pietro Bonfa'
> > 
> > 
> > 
> >     On 04/29/2014 04:13 PM, Alex Smogunov wrote:
> >     > Dear Pietro.
> >     > It should work, in principle, at least for US PPs ... even though
> >     > I do not understand what is the relation between  non-linear core
> >     > corrections and
> >     > the need for DFT+U together with spin-orbit, moreover you set up
> >     all the
> >     > parameters
> >     > essentially to zero. Anyway the code should run even in this case.
> >     Could
> >     > you provide the complete
> >     > input file, please?
> >     > regards,
> >     > Alexander
> >     >
> >     >
> >     >
> >     > 2014-04-28 10:52 GMT+02:00 Pietro Bonfa'
> >     <pietro.bonfa at fis.unipr.it <mailto:pietro.bonfa at fis.unipr.it>
> >     > <mailto:pietro.bonfa at fis.unipr.it
> >     <mailto:pietro.bonfa at fis.unipr.it>>>:
> >     >
> >     >     Dear All,
> >     >
> >     >     I would like to compare the total energies of various
> >     non-collinear
> >     >     magnetic configurations in La2CuO4 (LCO).
> >     >
> >     >     I started working with pseudopotentials from the PSLibrary,
> >     the GBRV
> >     >     library and some home brewed norm conserving Cu pseudos. After
> >     a short
> >     >     analysis I realized that the experimentally observed
> >     insulating and
> >     >     antiferromagnetic ground state is obtained only if I remove
> >     non-linear
> >     >     core corrections. This holds true for both my NC pseudos and
> >     the Cu
> >     >     pseudo from PSL 1.0.0 (I didn't try GBRV).
> >     >
> >     >     Nonetheless, since studying the magnetic ground state of LCO
> >     without
> >     >     using non-linear core corrections does not make much sense to
> >     me, I
> >     >     tried DFT+U with noncolin and lspinorb flags set to true.
> >     >
> >     >     More precisely, I added the following lines to example 11 in the
> >     >     QE-5.1rc1 release
> >     >
> >     >         lda_plus_u=.true.
> >     >         lda_plus_u_kind=1
> >     >         Hubbard_U(1)=1.d-10
> >     >         Hubbard_J(1,1)=1.d-10
> >     >         Hubbard_J(2,1)=0.0
> >     >
> >     >     Depending on the pseudopotential type (PAW, US or NC) I get
> >     errors like:
> >     >
> >     >          from cdiaghg : error #        19
> >     >          eigenvectors failed to converge
> >     >     or
> >     >          from cdiaghg : error #       109
> >     >          S matrix not positive definite
> >     >
> >     >     So my question is: is lda+u and non-collinear magnetism still not
> >     >     implemented, is this a bug or me doing something utterly wrong?
> >     >
> >     >     Thanks,
> >     >     kind regards,
> >     >     Pietro Bonfa'
> >     >
> >     >     --
> >     >     Pietro Bonfa' - PhD student
> >     >     Dipartimento di Fisica e Scienze della Terra "Macedonio Melloni"
> >     >     Viale delle Scienze 7A
> >     >     43124 Parma - Italy
> >     >     _______________________________________________
> >     >     Pw_forum mailing list
> >     >     Pw_forum at pwscf.org <mailto:Pw_forum at pwscf.org>
> >     <mailto:Pw_forum at pwscf.org <mailto:Pw_forum at pwscf.org>>
> >     >     http://pwscf.org/mailman/listinfo/pw_forum
> >     >
> >     >
> >     >
> >     >
> >     > _______________________________________________
> >     > Pw_forum mailing list
> >     > Pw_forum at pwscf.org <mailto:Pw_forum at pwscf.org>
> >     > http://pwscf.org/mailman/listinfo/pw_forum
> >     >
> > 
> > 
> >     --
> >     Pietro Bonfa' - PhD student
> >     Dipartimento di Fisica e Scienze della Terra "Macedonio Melloni"
> >     Viale delle Scienze 7A
> >     43124 Parma - Italy
> > 
> >     _______________________________________________
> >     Pw_forum mailing list
> >     Pw_forum at pwscf.org <mailto:Pw_forum at pwscf.org>
> >     http://pwscf.org/mailman/listinfo/pw_forum
> > 
> > 
> > 
> > 
> > _______________________________________________
> > Pw_forum mailing list
> > Pw_forum at pwscf.org
> > http://pwscf.org/mailman/listinfo/pw_forum
> > 
> 
> 

-- 
 Paolo Giannozzi, Dept. Chemistry&Physics&Environment, 
 Univ. Udine, via delle Scienze 208, 33100 Udine, Italy
 Phone +39-0432-558216, fax +39-0432-558222