[QE-users] PBE0 convergence issue for SCF calculation

[Rameswar Bhattacharjee]

Hi Everyone,
I have a question regarding SCF convergence using PBE0. I am trying to
run an SCF calculation for a system (no metal) but cannot converge it.
The calculation runs infinite time with the message that convergence
is achieved in 2 or 3 iterations several times., like below:

----------------------------------------------
     highest occupied, lowest unoccupied level (ev):    -4.2577   -2.5888
!    total energy              =    -261.58720112 Ry
     estimated scf accuracy    <       0.00000373 Ry
     convergence has been achieved in   1 iterations
     Using ACE for calculation of exact exchange
     ACE projected onto    60 (nbndproj) and applied to    60 (nbnd) bands

     BEWARE: negative dexx:  -0.671E-01

     total energy              =    -261.52011068 Ry
     Harris-Foulkes estimate   =    -261.52011419 Ry
     est. exchange err (dexx)  =       0.06709044 Ry
     - averaged Fock potential =      33.05795250 Ry
     + Fock energy (ACE)       =     -16.49568794 Ry

     EXX: now go back to refine exchange calculation

     total cpu time spent up to now is    73072.0 secs
---------------------------------------------------------------------------

However, the EXX is not converging whatever number of interactions I
run. The output prints these two lines at the end and terminates.

------------------------------------------
     EXX: now go back to refine exchange calculation

     EXX convergence NOT achieved after 150 iterations: stopping

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

The geometry is fine and easily converges with PBE.

Can anyone suggest a way to converge this calculation? Which keywords
I can alter to smooth the convergence? Any advice on this will be
highly appreciated.



-- 
Rameswar Bhattacharjee
Department of Chemistry
Georgetown University