SCF convergence stalls and restarts during GEO_OPT calculation

[hamid zabihi]

Dear CP2K community,

I am encountering difficulties with SCF convergence during a geometry optimization. The SCF goes smoothly for a while, with the residual decreasing steadily, but then it suddenly restarts and fails to progress further.

Below is a snippet from my output log showing the behavior:

   363 Broy./Diag. 0.50E-01    8.2     0.00000692      -221.5787867235 -4.05E-07
   364 Broy./Diag. 0.50E-01    7.3     0.00000325      -221.5787969182 -1.02E-05
   365 Broy./Diag. 0.50E-01    7.0     0.00000278      -221.5787817914  1.51E-05
   366 Broy./Diag. 0.50E-01    7.9     0.00000132      -221.5787799812  1.81E-06
   367 Broy./Diag. 0.50E-01    7.4     0.00000058      -221.5787749285  5.05E-06
   368 Broy./Diag. 0.50E-01    7.6     0.07802717      -221.5375627816  4.12E-02
   369 Broy./Diag. 0.50E-01    8.0     0.01514596      -221.6747537534 -1.37E-01
   ...

As you can see, the SCF almost converges (residuals < 1e-6), but then the procedure restarts with much larger residuals and does not settle afterward.

I have attached my input file (geo-opt.inp) and the full log file (geo-opt.log) for reference.

Has anyone experienced a similar issue? Do you have suggestions on how to stabilize SCF convergence in this situation (e.g., settings in &SCF, level shifting, mixing parameters, etc.)?

Any advice would be greatly appreciated.

Best regards,

Hamid Zabihi