Fully periodic QM/MM simulations

[Jadzia]

A short update: If I run the water simulation with the same QM settings but 
QM only (instead of QM/MM), then it seems to run stable and the water is 
not falling apart. Therefore DFTB3 and water seems to work generally with 
my setup, and the problem seems to arise only when QM/MM is switched on. 
Any suggestions or comments are greatly appreciated on how to stabilize the 
water molecules when using QM/MM. 

Many thanks,
Jadzia




On Friday, September 8, 2017 at 11:04:02 AM UTC+2, Jadzia wrote:
>
> Hi everyone, 
>
> @Matt: Thank you very much for the clarification, it was helpful.
>
> Since full QM/MM periodicity is not possible, I now would like to get the 
> QM/MM simulation of water working without QM/MM periodicity, but with 
> electrostatic embedding (E_COUPL = COULOMB). The QM method is still DFTB3. 
> And I have tried almost everything to get it working, but my water 
> molecules still fall apart. 
> I have tried...
>
>    - different GMAX values in the EWALD section
>    - different ALPHA values in the EWALD section
>    - different O_SPLINE values in the EWALD section
>    - the QM part completely without electrostatics (no POISSON section)
>    - DFTB3 with and without dispersion correction
>    - different time step sizes
>    - Standard BOMD and second generation CPMD
>    - ...
>
> Nothing seems to really solve the problem. I have also tried more things 
> such as mechanical embedding with and without QM/MM periodicity. 
>
> Here is a link to a sample input package which can be started directly, 
> all files included:
>
> https://drive.google.com/open?id=0B-Ij17NhUNZoeUpETFJhS1hpbFE
>
> Any help I would greatly appreciate. 
>
> Many thanks,
> Jadzia
>
>
>
> On Thursday, September 7, 2017 at 12:26:24 PM UTC+2, Matt W wrote:
>>
>> Hi again,
>>
>> sorry I don't really have time to give long enough replies to be clear, 
>> so perhaps a bit confusing.
>>
>> QM-QM coupling is not needed in your case - your QM box is the same size 
>> as the QMMM box. It calculates multipole moments of the QM charge 
>> distribution and subtracts artificial interactions if the QM box is smaller 
>> than QMMM and adds back in correct ones.
>>
>> QM-MM long range interactions as you say are the about including the 
>> ewald summed potential from the classical part - I think that this is not 
>> possible except using E_COUPL = GAUSS, which in turn requires DFT. See the 
>> paper from Laio et al for details.
>>
>> There are several different QMMM methods available, and they have 
>> different ranges of applicability.
>>
>> Matt
>>
>> On Thursday, September 7, 2017 at 10:28:58 AM UTC+1, Jadzia wrote:
>>>
>>> Thank you for your answer Matt. 
>>>
>>> Since error messages can be very helpful if they are precise, I would 
>>> like summarize my understanding so far related to the error message I have 
>>> gotten:
>>>
>>> QM-QM long range correction not possible with COULOMB coupling.
>>>
>>>
>>> Firstly, from what you wrote I implicitly conclude that the periodic 
>>> section which I have added to my input file went into the direction of what 
>>> you had in mind when you wrote about including the coupling of the QM 
>>> region with the periodic MM images. In this case the above error message 
>>> which I got seems to be not fully suitable since I had disabled the QM-QM 
>>> long range correction (i.e. the MULTIPOLE section). To my understanding 
>>> (which is admittedly incomplete) the problem might be of a more general 
>>> nature and summarized as follows: *QM/MM long range correction not 
>>> possible with COULOMB coupling*
>>> because it seems to cover both cases QM-MM and optionally also QM-QM 
>>> long term corrections (if the multipole section is not switched off) for 
>>> the QM region. 
>>>
>>> Secondly, the error message seems to be too general in another aspect, 
>>> because you wrote that QM/MM periodic couplings are possible with DFT (and 
>>> Coulomb coupling I guess), in which case the error message does not hold in 
>>> its current general form. 
>>>
>>>
>>>
>>>
>>>
>>>
>>> On Thursday, September 7, 2017 at 9:26:55 AM UTC+2, Matt W wrote:
>>>>
>>>> As I said, I wasn't sure if there was a periodic correction implemented 
>>>> using Coulomb coupling for DFTB/semi-empirical.
>>>>
>>>> Looks like you just have to run without periodic coupling, or switch to 
>>>> DFT to use periodic coupling.
>>>>
>>>> On Thursday, September 7, 2017 at 4:41:31 AM UTC+1, Jadzia wrote:
>>>>>
>>>>> Another thing I wonder about: Matt you said that I should keep the 
>>>>> section QMMM periodic section so that the QM region feels the potential of 
>>>>> the periodically replicated MM region. Which I would call a "MM-QM long 
>>>>> range correction". But the error message of my last post states that 
>>>>> a QM-QM long range correction is not possible with Coulomb coupling when I 
>>>>> add this section. Therefore I wonder if the periodic section which I set up 
>>>>> for the MM-QM long range correction was indeed what you had in mind...
>>>>>
>>>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20170911/d5bcced5/attachment.htm>