[CP2K-user] SCF_GUESS: using restart with atomic guess for additional atoms

[Fabian Ducry]

Hi Sergey,

The code works well and can indeed reduce the number of required SCF 
iterations considerably. There is one caveat though, I had to add the 
following lines to qs_initial_guess.F before the call to

CALL calculate_atomic_orbitals(atomic_kind, qs_kind, iunit=ounit, &
                                        pmat=pmat(ikind)%mat, nocc=nocc)

in SUBROUTINE calculate_atomic_block_dm

         write(*,*) ikind, nkind
         write(*,*) nocc
         write(*,*) TRIM(atomic_kind%name)

Without them I always got an undefined reference error and cp2k crashed, 
with them everything works as intended. This happens only with the 
ATOMIC_RESTART. Using the ATOMIC guess cp2k runs without the print 
statements.

To compile cp2k I use the toolchain with gfortran both on Piz Daint and a 
local cluster.

Thanks again for the feature,
Fabian
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20200331/125d9698/attachment.htm>