Fully periodic QM/MM simulations

[Jadzia]

Hi Matt, 

Thank you very much for your answer. 

Since previously I did not have coupled the (DFTB3) QM region (which is the 
same size of the MM region) to the periodic MM images, if I understand you 
correctly this could be the cause that some of my water molecules fall 
apart since I did not take many measures against boundary effects as you 
have written. The only thing I have done is to add standard LJ pontentials 
for the MM/QM atoms close to each other. What other measures are there 
which one could take?

I have now tried to set up fully periodic QM/MM:


  &QMMM
    &CELL
      ABC 41.576 41.369 42.526
      PERIODIC XYZ
    &END CELL
    ECOUPL Coulomb
    &PERIODIC
      GMAX 5.0E-01
      &MULTIPOLE off
      &END MULTIPOLE
      &POISSON
        &EWALD
          EWALD_TYPE spme
          GMAX 1
        &END EWALD
      &END POISSON
    &END PERIODIC
    CENTER setup_only
    CENTER_GRID FALSE
    USE_GEEP_LIB 7
    NOCOMPATIBILITY
    &FORCEFIELD
      &NONBONDED14
        @include ../../../cp2k.in.qmmm.lj.system1
      &END NONBONDED14
    &END FORCEFIELD
  &END QMMM


However, I got:

 *******************************************************************************
 *   ___                                                                   
    *
 *  /   \                                                                   
   *
 * [ABORT]                                                                 
    *
 *  \___/     QM-QM long range correction not possible with COULOMB 
coupling   *
 *    |                                                                     
   *
 *  O/|                                                                     
   *
 * /| |                                                                     
   *
 * / \                                                   
 qmmm_per_elpot.F:379 *
 *******************************************************************************

Thus it seems I can either
1) switch to mechanical embedding (i.e. ECOUPL NONE)
2) switch off the QM/MM electrostatic periodicity as I had before, in which 
case I would need to add more of the special measures you mentioned it 
seems.
And recommendations?

I used DFTB3 without hydrogen corrections. Could that be the problem, or 
maybe the dispersion correction enhances the electron spillout problem?

Best wishes,
Jadzia



On Tuesday, September 5, 2017 at 11:41:30 AM UTC+2, Matt W wrote:
>
> Hi,
>
> you want to keep the periodic section, if you want a periodic system. This 
> _should_* make the QM region feel a potential that comes from the 
> periodically replicated MM region (so essentially Ewald sum). Without that 
> section you just get the potential from the point charges surrounding the 
> region, that needs lots of care about boundary effects otherwise you can 
> get stray fields and other artifacts.
>
> Multipole correction is (normally a fairly small correction) that allows 
> you to use a smaller QM box but still be periodic. You can ignore it if 
> your QM and QMMM cells are the same size.
>
> * I don't know if the periodic functionality works with DFTB ...
>
> Matt
>
> On Monday, September 4, 2017 at 3:27:49 PM UTC+1, Jadzia wrote:
>>
>> Dear All, 
>>
>> I'm relatively (but not completely) new to CP2K, and am currently trying 
>> to run a QM/MM simulation (at the moment just a test system with water) 
>> which is fully periodic, using DFTB3 with dispersion correction, and 
>> electrostatic embedding (Coulomb).
>>
>> As far as I could find out, with DFTB one can set the QM box size to the 
>> system size, and in this case one does not need to decouple the QM images 
>> in the QMMM section. Does this mean that I can leave away the &PERIODIC 
>> section completely in the QM/MM region? Because if I have &PERIODIC 
>> sections in the MM and in the DFT sections, then the periodicity should 
>> alraedy be set up correctly, right? I have already tried to switch off only 
>> the &MULTIPOLE feature, but then I get an error about to a missing keyword 
>> related to EWALD:
>>
>> Value requested, but no value set getting value from keyword GMAX of 
>>> section EWALD. 
>>
>>
>> Therefore I guess I need to renmove the &PERIODIC keyword in the QMMM 
>> section, and if I remove it the simulation seems to run fine. But I can't 
>> tell if the periodicity works as it should. I also read a lot of topics in 
>> the mailing list, but I couldn't figure things completely out. My input 
>> files are attached, and about any comments I would be grateful. 
>>
>> Also, there is one other problem. After a while of simulation time, the 
>> QM water molecules are slowly falling apart. I.e. a hydrogen would detach 
>> from one water molecule and goes to the oxygen of another one. This happens 
>> almost always when MM water molecules interact with QM water molecules, but 
>> very few times I saw also a QM water molecule falling apart which was not 
>> directly near an MM water. I suspect that the* electron spillout problem* 
>> (as it is called in your 2005 article) is the cause. In the article is 
>> written that a Gaussian charge distribution is used to solve the electron 
>> spillout problem. Is this mechanism always active in QM/MM force 
>> evaluations in CP2K, or do I need to active it? (I could not find much 
>> about that topic.) Or do you have other recommendations on how to solve the 
>> water instability problem?
>>
>> If you see anything other unusual in my input file I would be glad to as 
>> well. I have checked things as best as possible, but don't have much 
>> experience with CP2K. 
>>
>> Best wishes,
>> Jadzia
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20170906/84364975/attachment.htm>