[Pw_forum] Re : problem with scf in ruthenium - textbook example

[Alexander Kulesza]

Dear developers and users,
first of all: thanks for your response Emine and Gabriele, i have tried 
out your suggestions.

-changing cutoffs does not affect the (really numerically unstalbe, not 
simply not converging) scf failure (see below for an example)
- using differnet modifications works: ibrav=3 bcc  Ruthenium gives 
converged scf energies.
- turning off symmetry does not affect failure
- ihave tried cutoffs from 40/400 to 80/800 with no effect
- trying Gabriele's suggestion (cutoff 200/800 ; only modified to 
coarser kpoint grid ) solved the problem (smaller cutoffs returned the 
"negative dr2" error)
- i switched to the pslibrary-0.2.3  -> this easily gives converged scf 
energies

would you recommend in this case to use the "old" PP
(pslibrary 0.2.3  : Ru.pbe-spn-rrkjus_psl.0.2.3.UPF) ?
Additionally, here also much more feasible cutoffs are suggested as i 
would like to simulate 4*4*2 surface  cells of hcp Ru(0001) ?
Thank you for investing the time to respond!
All the best
Alexander

Dr. Alexander Kulesza
Post-Doc with R. Mitric        CTO of CreativeQuantum GmbH
Department of Physics       Wegedornstraße 32
Free University Berlin          12524 Berlin
Arnimallee 14



      iteration #  1     ecut=    30.00 Ry     beta=0.80
      Davidson diagonalization with overlap
      ethr =  1.00E-02,  avg # of iterations =  5.4

      Threshold (ethr) on eigenvalues was too large:
      Diagonalizing with lowered threshold

      Davidson diagonalization with overlap
      ethr =  1.61E-04,  avg # of iterations =  3.2

      negative rho (up, down):  0.578E-03 0.000E+00

      total cpu time spent up to now is    448.59 secs

      total energy              =    -410.75163143 Ry
      Harris-Foulkes estimate   =    -410.68899116 Ry
      estimated scf accuracy    <       0.04497447 Ry

      iteration #  2     ecut=    30.00 Ry     beta=0.80
      Davidson diagonalization with overlap
      ethr =  1.41E-04,  avg # of iterations =  4.1

      negative rho (up, down):  0.824E+03 0.000E+00

      total cpu time spent up to now is    668.86 secs

      total energy              =   18020.31776757 Ry
      Harris-Foulkes estimate   =    -603.61433948 Ry
      estimated scf accuracy    <    3005.49358409 Ry