{SPAM?} [CP2K:3008] Re: DFT-D

[Teodoro Laino]

Joerg,

the script you've been using should be the one we distributed at the CP2K tutorial.
It's strange that Matthias does not know anything about that since he was tutor at that tutorial.
Anyway: the script does exactly what I said in the previous mail.
It takes the linear dimensions of the molecule (atomic distances) and sums up a buffer 3-5 Angstrom.
This will lead to converged results for decouplers like MT. 
Again in CP2K we do not have cheaper decoupler (you may try the others one available like :

MULTIPOLE

WAVELET

)
Refer to the papers cited for each of these methods for constraints about the box size: each of them is slightly different.
For large molecules WAVELET may be definitely much better. In fact for WAVELET is also enough that the density is zero at the border of the box.
But there are other constraints : for non-periodic calculations we have implemented only cubic boxes. So.. you see.. depends a lot.. 
for MT you can use an orthorhombic (but larger in size).. for WAVELET a cubic although a bit smaller.

Up to you the final decision.
Teo


On Dec 16, 2010, at 10:54 PM, Jörg Saßmannshausen wrote:

> Hi Matthias,
> 
> thanks for the reply. 
> I was using the script from the workshop to get the box size. You are not 
> aware of any bugs in that script which could lead to a wrong box size for 
> rathe large but flat molecules?
> 
> I will toy around with the grid size and see if I can get it going.
> Probably moving to a PBC calculation would not help with the problem at all I 
> would guess, or?
> 
> All the best from a rainy London
> 
> Jörg
> 
> 
> 
> On Dienstag 14 Dezember 2010 Matthias Krack wrote:
>> Hi Joerg,
>> 
>> your simulation box size seems to be about 27 Angstrom^3 and using a
>> cutoff 660 Ry for such a box will result in large grids and thus huge
>> memory allocations. Possibly, the extreme memory request is the reason
>> for abnormal program termination. You may check your grid sizes using
>> http://cp2k.berlios.de/manual/CP2K_INPUT/FORCE_EVAL/PRINT/GRID_INFORMATION.
>> html to get an idea of what will be used.
>> Afaik, SAVE_MEM helps only for systems which are large due to the
>> number of atoms employed. This is not the case for your system and
>> thus there is no impact from this keyword.
>> 
>> cheers,
>> 
>> Matthias
>> 
>> On Dec 13, 11:54 pm, Jörg Saßmannshausen <j.sassma... at ucl.ac.uk>
>> 
>> wrote:
>>> Dear Juerg,
>>> 
>>> ok, I am now playing around with the DFT-D in cp2k.
>>> I am currently trying to ascertain how much memory we would require for
>>> the work we are planning to do and for the new cluster. Hence, these are
>>> test runs of the right size of molecule but not with a really good
>>> optimised structure (i.e. not optimised with DFT-D).
>>> 
>>> I have tried both the DFT-D2 and DFT-D3 option and I got 2 problems:
>>> a) the test molecule needs about > 48 GB of RAM, which is significant. Is
>>> there any way to get that into a more handable size? I have tried the
>>> SAVE_MEM in GLOBAL but that did not really had any effect it seems.
>>> Following Matthias' suggestion some time ago, I am using a cutoff of 660,
>>> which was working well for different molecules in the past.
>>> 
>>> b) I get the following error for the DFT-D2 frequency calculation:
>>> 
>>> [ ... ]
>>>  REPLICA| layout of the replica grid, number of groups                  
>>>       2
>>>  REPLICA| layout of the replica grid, size of each group                
>>>       1
>>>  REPLICA| MPI process to grid (group,rank) correspondence:
>>>   (   0 :    0,   0)  (   1 :    1,   0)
>>> 
>>>  VIB| Vibrational Analysis Info
>>> 
>>>   Pair potential vdW calculation
>>>   Dispersion potential type: DFTD2
>>>   Scaling parameter (s6)   1.00000000000000000
>>>   Exponential prefactor     20.000000000000000
>>>   Total vdW energy [au]     : -6.37708902660472643E-002
>>>   Total vdW energy [kcal]   :  -40.016837470249470
>>> 
>>>   Dispersion Forces
>>>   Atom   Kind                            Forces
>>> [ ... ]
>>>  |G| =   5.08859983571732719E-002
>>> 
>>>  Stress Tensor (dispersion)
>>>  -0.911187517690E-03 -0.491992126627E-02  0.140501262854E-01
>>>  -0.491992126627E-02  0.134365275526E-01  0.179254944339E-02
>>>   0.140501262854E-01  0.179254944339E-02 -0.155716224075E-01
>>>    Tr(P)/3 :   -1.01542745750825042E-003
>>> 
>>>  CP2K| condition FAILED at line 253
>>>  CP2K| Abnormal program termination, stopped by process number 1
>>>  CP2K| condition FAILED at line 253
>>>  CP2K| Abnormal program termination, stopped by process number 0
>>> 
>>> For the DFT-D3 calculation I get this error:
>>> [ ... ]
>>> 
>>> |G| =   1.19763995420629592E-002
>>> 
>>>  Stress Tensor (dispersion)
>>>  -0.525748742588E-02 -0.260638877285E-02  0.799232865409E-02
>>>  -0.260638877285E-02 -0.433023621108E-03  0.181088946886E-02
>>>   0.799232865409E-02  0.181088946886E-02 -0.929520614835E-02
>>>    Tr(P)/3 :   -4.99523906511300034E-003
>>> 
>>>  CP2K| condition FAILED at line 236
>>>  CP2K| Abnormal program termination, stopped by process number 0
>>> 
>>> It appears to me I am doing something wrong. I basically have copied and
>>> pasted your section into the force_eval.inc file.
>>> Can you or somebody be so kind and point me in the right direction? It is
>>> possible to do the frequency calculation with DFT-D or am I wrong here?
>>> 
>>> I have attached the input files dft-d-D2.inp, force_eval-D2.inc and
>>> subsys.inc.
>>> 
>>> All the best from London
>>> 
>>> Jörg
>>> 
>>> On Montag 22 November 2010 hut... at pci.uzh.ch wrote:
>>>> Hi
>>>> 
>>>> here is an example for the Grimme D2 method
>>>>       &vdW_POTENTIAL
>>>>          DISPERSION_FUNCTIONAL PAIR_POTENTIAL
>>>>          &PAIR_POTENTIAL
>>>>             TYPE DFTD2
>>>>             R_CUTOFF 15.0
>>>>             SCALING 1.0
>>>>          &END PAIR_POTENTIAL
>>>>       &END vdW_POTENTIAL
>>>> 
>>>> The SCALING parameter refers to the s6 term in this method.
>>>> Default values for some functionals are available in the
>>>> code through
>>>>              REFERENCE_FUNCTIONAL BLYP
>>>> 
>>>> This is an example for the new D3 method
>>>>          DISPERSION_FUNCTIONAL PAIR_POTENTIAL
>>>>          &PAIR_POTENTIAL
>>>>             TYPE DFTD3
>>>>             REFERENCE_FUNCTIONAL BLYP
>>>>             CALCULATE_C9_TERM .TRUE.
>>>>             PARAMETER_FILE_NAME dftd3.dat
>>>>             R_CUTOFF 15.0
>>>>          &END PAIR_POTENTIAL
>>>>       &END vdW_POTENTIAL
>>>> 
>>>> With VERBOSE_OUTPUT TRUE you can get all kinds of detailed
>>>> energy contributions.
>>>> 
>>>> regards
>>>> 
>>>> Juerg Hutter
>>>> 
>>>> --------------------------------------------------------------
>>>> Juerg Hutter                       Phone : ++41 44 635 4491
>>>> Physical Chemistry Institute   FAX   : ++41 44 635 6838
>>>> University of Zurich               E-mail:  hut... at pci.uzh.ch
>>>> Winterthurerstrasse 190
>>>> CH-8057 Zurich, Switzerland
>>>> ---------------------------------------------------------------
>>>> 
>>>> -----cp... at googlegroups.com wrote: -----
>>>> 
>>>> To: cp2k <cp... at googlegroups.com>
>>>> From: Jörg Saßmannshausen <j.sassma... at ucl.ac.uk>
>>>> Sent by: cp... at googlegroups.com
>>>> Date: 11/21/2010 11:23AM
>>>> Subject: [CP2K:2928] DFT-D
>>>> 
>>>> Dear all
>>>> 
>>>> admittingly, my skills in cp2k are a bit rusty.
>>>> I wanted to look into the DFT-D option of cp2k, specially how the
>>>> second derivatives calculation are performing on rather large
>>>> molecules.
>>>> 
>>>>> From the manual I takt it one can do DFT-D calculations so I thought I
>>>>> am
>>>> 
>>>> using the water example and play around with that. I decided, mainly
>>>> for speed reasons, to go with the BP functional and compare that MD run
>>>> with one where DFT-D is switched on, i.e. BP-D.
>>>> However, I am not sure whether I done the setup correctly, so I thought
>>>> I might as well print out whether or not dispension is actually switch
>>>> on. Playing around with various option in
>>>> __ROOT__%FORCE_EVAL%DFT%PRINT%DFT_CONTROL_PARAMETERS
>>>> where somehow unsuccesful in this respect. Would somebody just briefly
>>>> comment on whether my input file (attached below) is correct and how to
>>>> print out the DFT parameters?
>>>> I have omitted the coordination section to save some space.
>>>> 
>>>> All the best from London!
>>>> 
>>>> Jörg
>>>> 
>>>> Input file:
>>>> &FORCE_EVAL
>>>>   METHOD QS
>>>>   &DFT
>>>>     BASIS_SET_FILE_NAME ../BASIS_SET
>>>>     POTENTIAL_FILE_NAME ../POTENTIAL
>>>>     &MGRID
>>>>       CUTOFF 280
>>>>     &END MGRID
>>>>     &QS
>>>>       EPS_DEFAULT 1.0E-12
>>>>       WF_INTERPOLATION PS
>>>>       EXTRAPOLATION_ORDER 3
>>>>     &END QS
>>>>     &SCF
>>>>       SCF_GUESS ATOMIC
>>>>       &OT ON
>>>>         MINIMIZER DIIS
>>>>       &END OT
>>>>     # SCF_GUESS        RESTART
>>>>     # EPS_SCF      1.0E-7
>>>>       &PRINT
>>>>         &RESTART OFF
>>>>         &END
>>>>       &END
>>>>     &END SCF
>>>>     &XC
>>>>       &XC_FUNCTIONAL BP
>>>>       &END XC_FUNCTIONAL
>>>>       &VDW_POTENTIAL
>>>>        &PAIR_POTENTIAL
>>>>         Type Grimme
>>>>        &end PAIR_POTENTIAL
>>>> #      POTENTIAL_TYPE DISPERSION_FUNCTIONAL
>>>>       &END VDW_POTENTIAL
>>>>     &END XC
>>>>    &print
>>>>     &DFT_CONTROL_PARAMETERS high
>>>>     &end DFT_CONTROL_PARAMETERS
>>>>    &end print
>>>>   &END DFT
>>>>   &SUBSYS
>>>>     &CELL
>>>>       ABC 9.8528 9.8528 9.8528
>>>>     &END CELL
>>>>     # 32 H2O (TIP5P,1bar,300K) a = 9.8528
>>>>     &COORD
>>>> [ ... ]
>>>>     &END COORD
>>>>     &KIND H
>>>>       BASIS_SET TZV2P-GTH
>>>>       POTENTIAL GTH-PADE-q1
>>>>     &END KIND
>>>>     &KIND O
>>>>       BASIS_SET TZV2P-GTH
>>>>       POTENTIAL GTH-PADE-q6
>>>>     &END KIND
>>>>   &END SUBSYS
>>>> &END FORCE_EVAL
>>>> &GLOBAL
>>>>   PROJECT H2O-32-vdw
>>>>   RUN_TYPE MD
>>>>   PRINT_LEVEL low
>>>>   &TIMINGS
>>>>      THRESHOLD 0.000001
>>>>   &END
>>>> &END GLOBAL
>>>> &MOTION
>>>>   &MD
>>>>     ENSEMBLE NVE
>>>>     STEPS 300
>>>>     TIMESTEP 0.5
>>>>     TEMPERATURE 300.0
>>>>   &END MD
>>>> &END MOTION
>>>> 
>>>> --
>>>> *************************************************************
>>>> Jörg Saßmannshausen
>>>> University College London
>>>> Department of Chemistry
>>>> Gordon Street
>>>> London
>>>> WC1H 0AJ
>>>> 
>>>> email: j.sassma... at ucl.ac.uk
>>>> web:http://sassy.formativ.net
>>>> 
>>>> Please avoid sending me Word or PowerPoint attachments.
>>>> Seehttp://www.gnu.org/philosophy/no-word-attachments.html
>>>> 
>>>> --
>>>> You received this message because you are subscribed to the Google
>>>> Groups "cp2k" group. To post to this group, send email to
>>>> cp... at googlegroups.com. To unsubscribe from this group, send email to
>>>> cp2k+uns... at googlegroups.com. For more options, visit this group
>>>> at http://groups.google.com/group/cp2k?hl=en.
>>> 
>>> --
>>> *************************************************************
>>> Jörg Saßmannshausen
>>> University College London
>>> Department of Chemistry
>>> Gordon Street
>>> London
>>> WC1H 0AJ
>>> 
>>> email: j.sassma... at ucl.ac.uk
>>> web:http://sassy.formativ.net
>>> 
>>> Please avoid sending me Word or PowerPoint attachments.
>>> Seehttp://www.gnu.org/philosophy/no-word-attachments.html
>>> 
>>>  dft-d-D2.inp
>>> < 1KViewDownload
>>> 
>>>  subsys.inc
>>> 4KViewDownload
>>> 
>>>  force_eval-D2.inc
>>> 1KViewDownload
> 
> 
> 
> -- 
> *************************************************************
> Jörg Saßmannshausen
> University College London
> Department of Chemistry
> Gordon Street
> London
> WC1H 0AJ 
> 
> email: j.sassma... at ucl.ac.uk
> web: http://sassy.formativ.net
> 
> Please avoid sending me Word or PowerPoint attachments.
> See http://www.gnu.org/philosophy/no-word-attachments.html
> 
> -- 
> You received this message because you are subscribed to the Google Groups "cp2k" group.
> To post to this group, send email to cp... at googlegroups.com.
> To unsubscribe from this group, send email to cp2k+uns... at googlegroups.com.
> For more options, visit this group at http://groups.google.com/group/cp2k?hl=en.
>