Large number of SCFs after optimization step

[Lucas Bandeira]

Dear cp2k group,

I am performing some calculations to find the lattice parameter of Cu. I have noted that there is a moment at which the cell optimization stops and a large number of SCF calculations start. The cell optimization stops, but it has not converged. Find attached the input I used in this calculation and the output printed. Exactly, what is happening is that only six cell optimizations are performed (with no convergence). After the last one, a large number of SCFs start. Can somebody help me to solve this problem? 

Yours sincerely,

Lucas Bandeira

[Krack Matthias (PSI)]

Hello Lucas

 

Did you try already BFGS instead of LBFGS as OPTIMIZER? LBFGS can get “stuck” when the PES is very flat.

 

Matthias

 

--
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/CAM1%3DHZnEnB8B8F4PdFmhHSpWJsQTKZCN6WBhRmqU%3DwJkf0E_Ng%40mail.gmail.com.

[ASSIDUO Network]

I don't see why you need the OUTER_SCF flag on. I did a bulk optimization for both Au and Cu without using it and got pretty close to literature. My &SCF parameters were:

    &SCF

      MAX_SCF    200

      EPS_SCF    1E-6

      SCF_GUESS  ATOMIC

      CHOLESKY INVERSE

      ADDED_MOS 100 ! You could lower this

      &DIAGONALIZATION T

        ALGORITHM STANDARD

      &END DIAGONALIZATION

      &SMEAR T

        METHOD FERMI_DIRAC

        ELECTRONIC_TEMPERATURE 300

      &END SMEAR

      &MIXING T

        METHOD BROYDEN_MIXING

        ALPHA 0.01 !You could increase this

        BETA 0.5

        NBUFFER 8

      &END MIXING

    &END SCF

I would also suggest rather using BFGS instead of LBFGS.