[Pw_forum] cpmd2upf.x for Norm Conserving pseudopotentials

[Ali KACHMAR]

Dear all,



Using and working with As_MT_GIA_BLYP.psp , Se_MT_BLYP, and 
Ge_MT_BLYP avalaible in  CPMD pseudopotential library and interested in 
d-channel
pseudopotential, I converted the As_MT_GIA_BLYP.psp with cpmd2upf tool
in QE. I have not seen somewhere else so far an Arsenic or Selenuim
with d-channel pseudopotential in QE format or in any other format
could be converted to upf, either in Norm Conserving (NC) or ultrasoft
(US) pseudopotentials.



1) Using the cpmd2upf.x tool to convert the As pseudo, If I take the
wavefunctions occupancy as they are given in the pseudopotential
report  below and trying to run a simple scf calculation at the Gamma
point  with the pseudopotential converted using input_dft='BLYP' and
with a good ecutwfc

./cpmd2upf.x

Input file > As_MT_GIA_BLYP.pspRadial grid r(i) has  168 points

Assuming log radial grid: r(i)=exp[(i-1)*amesh]*r(1), with:

amesh = value read from file =   0.062500



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

    |    Pseudopotential Report                                |

    ------------------------------------------------------------

    |  Atomic Symbol                   :  As                   |

    |  Atomic Number                   :  33                   |

    |  Number of valence states        :   3                   |

    |  Trouiller-Martins normconserving PP                     |

    |  Exchange-Correlation Functional :                       |

    |     Slater exchange :  0.6667                            |

    |     LDA correlation : Lee-Yang-Parr                      |

    |     Exchange GC     : Becke (1988)                       |

    |     Correlation GC  : LYP                                |

    |  Electron Configuration :                                |

    |  As 4s1 4p1.75 4d.25 Rc=1.45 1.65 2.56 MT BLYP           |

    |  USE L=0 AS LOCAL  and LMAX=D                            |

    |  Generated by P. Giannozzi, 25 Nov. 1998                 |

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



I have seen at the



End of self-consistent calculation



           k = 0.0000 0.0000 0.0000 ( 12053 PWs)   bands (ev):



   ***************************  -1.5792  -0.3039



  This is something not clear to me !



2) Ignoring the wavefunction occupancy of d, I can see at the end of

the scf calculation all the band energies for k.



End of self-consistent calculation



           k = 0.0000 0.0000 0.0000 ( 12053 PWs)   bands (ev):



    -14.1389  -4.8651  -4.8651  -4.8651  -0.3376



These band energies match the band energies for an Arsenic US BLYP
with kkrjrus from A. Dal Corso pslibrary.



So before going far a simple postprocessing test works for the second
case and not for the first one. I am curious to understand why I am
not getting all the band energies in the first case, so it doesn't
arrange me to ignore the d-channel since I am interested in it and no
pseudopontial in QE with NC or US with d-channel for As and Se.



Thus before going to generate NC pseudopotentials for As and Se, I
thought it would be a good or not to try the use the coverted
pseudopotentials as a first step. As it has been already discussed 
some years ago by Paolo and Axel, I  understand now why converting 
pseudopotentials is not as good and easy task as pseudopotential 
generation in some cases. As a second, I am  going to generate the NC 
for As and Se with ld1.x However  my guessing is the d-channel will need a deep understanding for ld1.x due to their
bonding and nonbonding character. Morever, I am not going to invent
something new I will just take Paolo's value for Rcut in pseudopotential report.



Saying that maybe an old version of QE could do this scf behavior ( I
mean at the end of the self-consistent calculation) but the scf
behavior are similar for pw.x in the QE.4.2.1 and the svn version.
Note that the scf doesn't fluctuate in the case 1 that much after the fourth digit
of the total energy.



Hoping that the cpmd2up in the QE svn version could change something in the scf band energies but the same result as reported above. Is
epsilon.x is compatible with pw.x in the svn version? it Doesn't work
for me for QE pseudopotentials.


Thanks,

Ali

Research Associate

LPTMC-UMPC
 		 	   		  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20130118/0f0c2bb7/attachment.html>