The parameters of SCREENING

huan... at huan... at
Tue Jan 27 13:04:03 UTC 2015

Dear CP2K developers and users,

I have a question on the parameters of SCREENING keyword. 

Recently, I read the reference "*Ab initio molecular dynamics using hybrid 
density functionals.  J. Chem. Phys., 128 (21), 214104 (2008)*". The 
authors tested three parameters when performing the SCREENING, e.g. 
epsilon_Schwarz, epsilon_Box and epsilon_Storage, and showed in the Table 
III that by using the epsilong_Storage = 1/10*epsilon_Schwarz, both CPU 
time and the RAM consumption reduced about 2-times, while retained the 
similar accuracy. The epsilon_Box parameter seems do not significantly 
improve the performances. 

However, in CP2K manual, the SCREENING section lists following 5 keywords.  


Only the first one, EPS_SCHWARZ, seems related to the epsilon_Schwarz in 
the reference. *What the relationship between the EPS_SCHWARZ and EPS_SCF? 
Shall I set the same value for them?*

I am also not very clear about the rest keywords. Does the 
SCREEN_ON_INITIAL_P and SCREEN_P_FORCES corresponding to the post-GGA 
screening in the reference? *Shall I set them as TRUE*, if I read the 
wavefunction_restart files? 

If the SCREEN_P_FORCES is set to TRUE, the EPS_SCHWARZ_FORCES should be 
specified, as the CP2K manual instructed. Then *shall I set the value 
of EPS_SCHWARZ_FORCES equals to the value of EPS_SCF*? 

The most important parameter, I think, is the epsilon_Storage, which does 
not shown in the SCREENING section. Is it the keyword of 
EPS_STORAGE_SCALING in MEMORY section? The explanation for this key word is 
"*Scaling factor to scale eps_schwarz. Storage threshold for compression 
will be EPS_SCHWARZ * EPS_STORAGE_SCALING*". I am not sure about the 
relationship between this expression and the memory usage. *How to set the 
appropriate value of memory*? 

I would greatly appreciate it if any of you kindly give me some suggestion.


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <>

More information about the CP2K-user mailing list