The ABINIT Users Mailing List ( CLOSED )

[Fabien Bruneval <fabien.bruneval@polytechnique.fr>]

Dear Corsin,
Indeed when iscf=-2, eigenvalues are computed at fixed self-consistent 
potential, but the algorithm used to diagonalize the Hamiltonian is an 
iterative one, based on conjugate-gradient scheme. It saves memory and 
time with respect to usual non iterative diagonalizations, when the 
number of eigenvalues that you need is much smaller than the size of the 
Hamiltonian (= number of plane-waves). That's why you have to choose 
nstep != 0.

Fabien Bruneval


Corsin Battaglia wrote:

> Dear Matthieu, dear Abinit users
>
> First of all, thanks for your mail. I increased my nstep value and 
> added the
> prtvol command, so my bandstructure looks quite O.K. now.
> What I do not understand at all is the following: According to the manuel
> entry for the iscf command, iscf=-2 results in a non-self consistent
> calculation. Makes sense to me, since eigenvalues are computed at 
> fixed self
> consistent potential, which  was determined in a previous scf-run (prtden
> after scf, getden for bandstructure calculation). But then why did I 
> have to
> increase nstep? And why abinit has to converge the eigenvalues? In my
> opinion, it is sufficient to put the self-consistent potential into the
> Hamiltonian and to diagonalize it.
>
> Well, maybe it is more complicated then I thought it is?
> Thanks for your answer.
>
> Corsin
>
> P.S.: I wrote a little Igor Pro procedure, which automatically reads 
> in the
> abinit output file and plots the bandstructure. Just in case, somebody is
> interested.
>
>
>
> ----- Original Message -----
> From: <verstraete@pcpm.ucl.ac.be <mailto:verstraete@pcpm.ucl.ac.be>>
> To: <forum@abinit.org <mailto:forum@abinit.org>>
> Sent: Monday, February 02, 2004 11:50 AM
> Subject: Re: [abinit-forum] fuzzy bandstructure
>
>
> >
> >
> > Hello Corsin,
> >
> > The (stupid) problem is that by default abinit only prints 50 
> kpoints. The
> > calculation didn't have enough nsteps to converge all the points (hence
> > the fuzziness), but then it didn't print out the data it calculated. Set
> > prtvol 10 to print out everything, and increase nstep (for dtset 2 it is
> > the default of 1 - you only specified nstep1 not nstep2).
> >
> > We should print a separate file with the eigenvalues in good formats
> > (xmgr, with a linear coordinate along the band structure), but 
> someone has
> > to sit down and code it in abinit - not hard. Volunteers?
> >
> > Matthieu
> >
> >
> > On Sun, 1 Feb 2004, Corsin Battaglia wrote:
> >
> > > Dear abinit users
> > >
> > > I am trying to calculate the bandstructure for the metallic compound
> NbS2.
> > > In a first step, I computed the total energy. I used the 
> Hartwigsen-Goed
> ecker-Hutter pseudopotentials for Nb and S. For Nb, I used the one, which
> contains semicore states as well (I read somewhere that semicore 
> states are
> necessary in an electronegative environment).
> > > During the scf-cycle, there seem to be no convergence problems.
> > >
> > >      iter   Etot(hartree)      deltaE(h)  residm     vres2    diffor
> maxfor
> > >
> > >  ETOT 14  -154.22694726148    -3.875E-11 1.106E-04 8.090E-09 1.155E-06
> 2.924E-02
> > >  ETOT 15  -154.22694726148    -2.394E-12 1.110E-04 8.903E-10 2.442E-07
> 2.925E-02
> > >  ETOT 16  -154.22694726148    -9.612E-13 4.247E-05 5.884E-10 3.020E-07
> 2.925E-02
> > >  ETOT 17  -154.22694726148    -4.440E-13 3.888E-05 6.675E-12 1.490E-07
> 2.925E-02
> > >
> > > The Fermi energy converges as well
> > >
> > > 11792: newocc : new Fermi energy is       0.247380 , with nelect=
> 50.000000
> > > 12728: newocc : new Fermi energy is       0.247382 , with nelect=
> 50.000000
> > > 13664: newocc : new Fermi energy is       0.247381 , with nelect=
> 50.000000
> > > 14600: newocc : new Fermi energy is       0.247381 , with nelect=
> 50.000000
> > > 15536: newocc : new Fermi energy is       0.247381 , with nelect=
> 50.000000
> > > 16472: newocc : new Fermi energy is       0.247381 , with nelect=
> 50.000000
> > > 17408: newocc : new Fermi energy is       0.247381 , with nelect=
> 50.000000
> > >
> > > No warnings or errors up to this point.
> > > In a second step, I calculated the bandstructure. When I plot the
> bandstructure, it looks very fuzzy, although the number of k-points was
> quite dense (50 k-points between Gamma and M).
> > >
> > >
> > > For every k-point, abinit writes the following warning into the 
> log file
> > >
> > > 17759: vtowfk: WARNING -
> > > 17760-  Wavefunctions not converged for nnsclo,ikpt=    1    1 max
> resid=  1.22992E+00
> > >
> > > Another problem is that abinit stops after 50 k-points, although I
> intended to compute 125 k-points.
> > >
> > > I have no clue, what I am doing wrong.
> > > Thanks in advance for your help
> > >
> > > Corsin
> > >
> > >
> > >
> > > Here's are the files I used.
> > >
> > > NbS2.files
> > >
> > > NbS2.in
> > > NbS2.out
> > > NbS2i
> > > NbS2o
> > > NbS2
> > > ../../Pseudopotentials/41nb.13.hgh
> > > ../../Pseudopotentials/16s.6.hgh
> > >
> > >
> > > NbS2.in
> > >
> > > # 2H-NbS2 : computation of the total energy and bandstructure
> > >
> > > #Definition of the unit cell
> > > acell   3.31   3.31    11.89 angstrom
> > > #rprim  0.866 -0.500   0.000     # It is better to define
> > > #       0.000  1.000   0.000     # the primitive vectors
> > > #       0.000  0.000   1.000     # using rprim
> > > angdeg 90 90 120
> > >
> > > #Definition of the atom types
> > > ntypat 2          # There are two type of atoms
> > > znucl 41 16       # The keyword "znucl" refers to the atomic number of
> the
> > >                   # possible type(s) of atom. The pseudopotential(s)
> > >                   # mentioned in the "files" file must correspond
> > >                   # to the type(s) of atom.
> > >
> > >
> > > #Definition of the atoms
> > > natom 6           # There are six atoms
> > > natrd 2           # Reads two atoms
> > > typat 1 2         # type 1 is Nb, type 2 is S
> > > xred              # This keyword indicate that the location of the 
> atoms
> > >                   # will follow, one triplet of number for each atom
> > >    0.0  0.0  1/4  # Triplet giving the REDUCED coordinate of atom 1.
> > >    1/3  2/3  1/8  # Triplet giving the REDUCED coordinate of atom 2.
> > >                   # Note the use of fractions (remember the limited
> > >                   # interpreter capabilities of ABINIT)
> > > spgroup 194
> > >
> > > #Definition of the occupation numbers
> > > occopt 4
> > > tsmear 0.01
> > >
> > > #Read psp
> > > npsp 2            # Read 2 psp files
> > > ixc 1             # Nb is of type ixc 1. S is of type ixc 1.
> > >                   # LDA. Nb contains semicores.
> > >
> > >
> > > #Definition of the planewave basis set
> > > ecut 10.0         # Maximal kinetic energy cut-off, in Hartree
> > >
> > > ndtset 2
> > >
> > > #Dataset 1: SCF GS calculation
> > >
> > > #Definition of the k-point grid
> > > kptopt1 1         # Option for the automatic generation of k points,
> taking
> > >                   # into account the symmetry
> > > ngkpt1 8 8 4      # This is the grid based on the primitive vectors
> > >                   # of the reciprocal space
> > >
> > >
> > > #Definition of the SCF procedure
> > > iscf1 3           # SCF cycle, CG based on the minim of the energy
> > > nstep1 250        # Maximal number of SCF cycles
> > > toldfe1 1.0d-12    # Will stop when, twice in a row, the difference
> > >                   # between two consecutive evaluations of total 
> energy
> > >                   # differ by less than toldfe (in Hartree)
> > > #diemac 12.0      # For metals, we use the default 10^6.
> > > nband1 35         # nband=nb of electrons in unit cell/2+(20% for
> metals)
> > >                   # more bands are needed with semicore states
> > > prtden1 1         # Print the density for use by dataset 2
> > >
> > > #Dataset 2: band structure
> > >
> > > iscf2 -2          # Non-SCF calculation
> > > getden2 -1
> > > kptopt2 -4        # Bandstructure with 4 lines
> > > nband2 35         # 35 bands
> > > ndivk2 50 25 50 10
> > > kptbounds2      0.0  0.0  0.0 # Gamma point
> > >                 0.5  0.0  0.0 # M point
> > >                 2/3  2/3  0.0 # K point
> > >                 0.0  0.0  0.0 # Gamma point
> > >                 0.0  0.0  0.5 # A point
> > > tolwfr2 1.0d-12   # Only admitted convergence criterion for non-SCF
> calculations
> > > enunit2 1         # Output the eigenvalues in eV
> > >
> > >
> >
> > --
> > ===================================================================
> > Matthieu Verstraete               mailto:verstraete@pcpm.ucl.ac.be
> > PCPM, Boltzmann, pl. Croix du Sud, 1            tel: 010/ 47 33 59
> > B-1348 Louvain-la-Neuve  Belgium                fax: 010/ 47 34 52