[CP2K-user] Extremely slow SCF convergence on O2/metal

[Y Lu]

Dear Matt and other CP2K gurus,

I hope you are well! We are developing a periodic embedding method in 
ChemShell to perform QM/MM MD simulations, typically on adsorbate/metal 
systems, using CP2K.

The attached O2/Pd calculation (PBE) converges extremely slowly. And it 
doesn't converge unless the density mixing ratio (&MIXING ALPHA) is only 
2%-5%. Any higher ratio results in strong SCF fluctuation and it would 
never converge (I tried 2000 cycles). Does this imply the initial guess is 
too far away from a stationary point?

Various other options I tried but didn't seem to help:

   - smearing
   - level shift
   - initial guess
   - other mixing methods
   - multi-grid cutoff and real cutoff
   - smooth XC grid methods

I hope the supercell parameters are correct (because CP2K doesn't accept 
redundant atoms in the cell). Any idea how to speed up the SCF convergence 
to make QM/MM MD simulation more useful, please?

Thank you very much.

Kind regards,

You
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20201113/2657a5a2/attachment.htm>