[CP2K:5389] CUTTOFF and REL_CUTTOFF for large molecular system

[hut... at chem.uzh.ch]

Hi

some info on cutoffs:

- The relative cutoff is the minimal cutoff that is used to
  integrate a Gaussian with exponent 1. This scales linearly.
  Exponent a -> a*RelCutoff.
  40 Rydberg usually gives reasonable results.
  60 Rydberg should be close to 'convergence'.
  80 Rydberg should always be converged. 
  !!! Remember, convergence tests always also depend on other
  factors, like the Max. Cutoff, number of grids, how subgrids are
  calculated.

- Estimate of total cutoff: Max. exponent in your basis set=a
  Cutoff = 2*a*Rel_cutoff
  This is usually way too high but can be used to work your way down
  to smaller cutoffs.

- For larger systems (many atoms and electrons) and dense systems,
  CPU time is not dominated by the cutoff

- For some GGA functionals convergence with Cutoff can be slow.
  Dirty trick: use one of the smoothing options (XC_GRID)
  Better: use higher cutoff for XC calculation only(testing needed)

- Sometimes slow convergence with cutoff points to a problem with
  the GTH pseudopotentials and GGA functionals.
  There are NLCC pseudos available that solve this issue.
  -> more testing needed

regards

Juerg
--------------------------------------------------------------
Juerg Hutter                         Phone : ++41 44 635 4491
Institut für Chemie C                FAX   : ++41 44 635 6838
Universität Zürich                   E-mail: hut... at chem.uzh.ch
Winterthurerstrasse 190
CH-8057 Zürich, Switzerland
---------------------------------------------------------------

-----cp... at googlegroups.com wrote: -----
To: cp... at googlegroups.com
From: Tobias Kraemer <161brun... at gmail.com>
Sent by: cp... at googlegroups.com
Date: 06/10/2014 07:06PM
Subject: [CP2K:5389] CUTTOFF and REL_CUTTOFF for large molecular system

Dear CP2K group,


being a relatively new cp2k user, I was wondering if this group could give me some advice on the choice of (REL) CUTOFF values.
The system I am considering is a somewhat extended molecular crystal  structure, containing organometallic fragments (basically a rhodium  phosphine complex with additional ligand plus large borate counteranions (BArF4-).The full system contains about 600+ atoms in the unit cell, basically 4 cation/anion pairs 
(asymmetric units). I think the system cannot be reduced in size, and the full unit cell has to be taken into account.I have attached a cif file which gives you
an idea about the kind of system (not exactly the one used in the calculation below, but very much related)
 
The basic idea behind determining the CUTOFF is clear to me and I have done the tutorial on bulk Si. I have used the same protocol for my system 
(running a series of calculations with varying CUTOFF and REL_CUTOFF values). In contrast to the suggested method in the tutorial, I have 
fully converged the SCF cycle) Below are the (first) results of this series. The energy appears to be converging towards higher CUTOFF values (some more calculations are currently running)

(1) I wonder if choosing such a large value for the cutoff is reasonable? It should also increase the accuracy of the calculations, but also the cost.
Also, with respect to the number of Gaussian being mapped onto the finest grid, what would one need to look for?  
In the tutorial the decision was  based on the fact that "the distribution of Gaussian functions on the grids are reasonable". Could  somebody elaborate on this 
and give me some guidance how to decide  this?
(2) How does the size of the CUTOFF value correlate with the system size, if there is any correlation?  
(3) Are there some rule of thumbs as how to choose the REL_CUTOFF with respect to the CUTOFF? For now I have tried 30, 60 and 100 Ry.
But I may need to go to higher numbers in this particular case. 



# Grid cutoff vs total energy
# Date: Tue Jun 10 17:26:55 BST 2014
# REL_CUTOFF = 60
# Cutoff (Ry) | Total Energy (Ha) | Charge | SCF | | NG on grid 1 | NG on grid 2 | NG on grid 3 | NG on grid 4
     50.00  -4583.1402739393   -8.4046446120  200  1623260  248362    2112       0
    100.00  -4477.8479306771   -0.0745436930  128  1264867  519816   89051       0
    150.00  -4479.5978060995   -0.0191170935  128  1150415  472845  248362    2112
    200.00  -4478.5030451759    0.0008175249  128  905641  531280  432781    4032
    250.00  -4478.4529157377   -0.0000917804  128  838757  500556  523109   11312
    300.00  -4478.4141842105   -0.0000534995  128  796882  467985  519816   89051
    350.00  -4478.3827319182   -0.0000170792  128  748308  474790  454036  196600
    400.00  -4478.3597765363    0.0000003328  128  679828  499675  489931  204300
    450.00  -4478.3498795935    0.0000000149  128  621687  528728  472845  250474
    500.00  -4478.3399690553    0.0000000147  128  567997  416013  600098  289626
    550.00  -4478.3338555119   -0.0000000007  128  485909  446314  559882  381629
    600.00  -4478.3312330111    0.0000000007  128  442629  463012  531280  436813
    650.00  -4478.3291405214    0.0000000006  128  399089  473084  554024  447537
    700.00  -4478.3308119206    0.0000000008  128  375655  473218  509402  515459
    750.00  -4478.3298871109    0.0000000008  128  365399  473358  500556  534421
    800.00  -4478.3266993684    0.0000000007  128  353459  468098  461722  590455
    850.00  -4478.3266987930    0.0000000007  128  348184  458373  472494  594683
    900.00  -4478.3273848891    0.0000000007  128  336984  459898  467985  608867
    950.00  -4478.3277135782    0.0000000007  128  309758  476872  469877  617227
   1000.00  -4478.3277705414    0.0000000008  128  291089  482269  455857  644519
   1050.00  -4478.3278605666    0.0000000007  128  229036  519272  474790  650636

Some further questions;

(4) Could you recommend some literature (books) which explains the theory of solid state calculations, k-sampling, planewaves?

(5) As a next step I would like to manipulate the structure, i.e. rotate/move ligands around? Is there a good visualizer that can do such 
things for periodic systems? The problem I am facing is that I would like to do these things simultaneously for all 4 asymmetric units.
I think it becomes impracticable to do this for each unit separately. I have started using GDIS but I am a little bit uncertain about its
abilities (great program though!!). Has anyone experience with VESTA? I am particularly looking for free software. 


Thank you very much for your help, this is very much appreciated. There are a lot of questions right a the beginning, and it is not always easy to 
find some answers.   

Best


Tobias


  Dr. Tobias Kraemer
  Research Associate
  Institute of Chemical Sciences
  School of Engineering & Physical Sciences
  Heriot-Watt University 
  Edinburgh EH14 4AS
  United Kingdom
  
  
  -- 
 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... at googlegroups.com.
 To post to this group, send email to cp... at googlegroups.com.
 Visit this group at http://groups.google.com/group/cp2k.
 For more options, visit https://groups.google.com/d/optout.
 

[attachment "input_test.inp" removed by Jürg Hutter/at/UZH]
[attachment "PCCPiBu_NBA.cif" removed by Jürg Hutter/at/UZH]