[CP2K-user] Convergence with CUTOFF

Augustin Bussy august... at chem.uzh.ch
Mon Jan 27 09:15:32 UTC 2020

Additionally, I would suggest that you tighten the EPS_SCF threshold (to 
1.E-7 or 1.E-8). Even using a really high CUTOFF, you might not observe 
converged results if the SCF cycle itself is not fully converged. 


On Monday, 27 January 2020 09:54:34 UTC+1, Travis wrote:
> Hi,
> Vacuum is expensive with plane wave methods. Your box only needs to be 
> large enough that the density goes to zero at the boundaries (6-10 
> Angstroms) with the wavelet solver. You'll see a warning about non-zero 
> density if you don't use a large enough box. The rule of thumb for the 
> cutoff is to multiply the largest exponent in the basis by REL_CUTOFF. For 
> your basis set and REL_CUTOFF, that's about 1500 Ry. This puts the tightest 
> orbital on the finest grid.
> -T
> On Monday, January 27, 2020 at 3:02:14 AM UTC-5, coko312 wrote:
>> Dear all,
>> I am a cp2k-DFT beginner so the answer to this question may seem obvious 
>> to you, but it would help me a lot!
>> I try to converge the total energy of a single Na+ ion as a function of 
>> the energy CUTOFF, but the convergence is very slow and even using 900 Ry 
>> is not fully satisfactory for what I want to do. Even with no experience 
>> with this system, I would not expect the computation to last several hours 
>> on a single processor for only 1 atom. I have attached my input-output 
>> files. What should I modify to be able to use a more reasonable CUTOFF? 
>> NB : the computation is non-periodic because I would like, next, to add 
>> more atoms around to compute interaction energies between these little 
>> "clusters" and various types of molecules.
>> Best,
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20200127/a5610450/attachment.htm>

More information about the CP2K-user mailing list