[CP2K-user] [CP2K:18369] Re: Guided AIMD reaction, CONSTRAINT, RESTRAINT, METADYNAMICS
Dobromir A Kalchevski
dobromirak at gmail.com
Sat Jan 21 00:00:24 UTC 2023
Dear Marcella Iannuzzi,
Thank you again for the clarifications on sampling final state, finding
reaction mechanism and potential minimal energy pathway :D This was very
useful.
Best Regards,
Dobromir
On Friday, January 20, 2023 at 4:00:59 PM UTC+2 Dobromir A Kalchevski wrote:
> This is the colvar:
>
> &COLVAR
> &DISTANCE
> AXIS XYZ !DEF XYZ, ANY COMBINATION
> ATOMS 50 81
> &END DISTANCE
> &END COLVAR
>
> and with restraint it moves the C atom from methane (reacting molecule) to
> a Si atom in the Si lattice, the way it's supposed to, it just goes too far.
>
> Best Regards,
> Dobromir
>
> On Friday, January 20, 2023 at 3:32:09 PM UTC+2 Marcella Iannuzzi wrote:
>
>>
>> Most probably you defined the COLVAR in the wrong way.
>> If there is an activation barrier in this reaction, this will never
>> happen in the right way just crashing the molecule into the surface
>>
>> Regards
>> Marcella
>>
>>
>> On Friday, January 20, 2023 at 2:14:26 PM UTC+1 Dobromir A Kalchevski
>> wrote:
>>
>>> Still, my supervisor doesn't want me to study the FES, he simply wants
>>> me to make the reaction happen in dynamics. He wants the molecules to move
>>> around and eventually a reaction to occur.
>>>
>>> Even if I try
>>>
>>> &CONSTRAINT
>>> &COLLECTIVE
>>> COLVAR 2
>>> INTERMOLECULAR
>>> TARGET [angstrom] 5.0
>>> !initial position in the xyz geometry file
>>> TARGET_GROWTH [angstrom*fs^-1] 0.01
>>> TARGET_LIMIT [angstrom] 2.0
>>> &END COLLECTIVE
>>> &END CONSTRAINT
>>>
>>> cp2k drives the two atoms from the COLVAR to a distance of 2.0 angstrom
>>> IMMEDIATELY. Is this normal or is it a bug ?
>>>
>>> And once again, why does RESTAINT with 1 kcal/mol K drive the molecule
>>> into the surface lattice, pushing lattice atoms around and integrating the
>>> C atom from the methane into the lattice ? Could it be due to DFTB
>>> parameterization or is it a bug or am I missing something ?
>>>
>>> Best Regards,
>>> Dobromir
>>> On Friday, January 20, 2023 at 2:35:31 PM UTC+2 Dobromir A Kalchevski
>>> wrote:
>>>
>>>> Dear Marcella Iannuzzi,
>>>>
>>>> Thank you for your reply.
>>>>
>>>> Best Regards,
>>>> Dobromir
>>>>
>>>> On Friday, January 20, 2023 at 2:21:37 PM UTC+2 Marcella Iannuzzi wrote:
>>>>
>>>>> Dear Dobromir
>>>>>
>>>>> The TARGET is the initial value that is imposed to the variable, as it
>>>>> is defined.
>>>>> The TARGET_GROWTH is the rate of change of that target value, which in
>>>>> the example you posted is ridiculously high:
>>>>>
>>>>> TARGET_GROWTH {Real}
>>>>> *Specifies the growth speed of the target value of the constrained
>>>>> collective variable.*
>>>>>
>>>>> The constraint with the value determined by the growth rate will be
>>>>> imposed, no matter which are the other forces in place.
>>>>> At least until the calculation does not converge anymore and
>>>>> everything explodes.
>>>>>
>>>>> Nothing of this is a sampling in the NVT ensemble, anyway.
>>>>>
>>>>> If the goal is to sample the final state after the reaction has
>>>>> occurred, just start a simulation from the configuration of the reacted
>>>>> molecule.
>>>>> If the goal is to find the reaction mechanism, there is no way out of
>>>>> studying the free energy surface or, in simpler cases, the potential energy
>>>>> surface.
>>>>> An efficient way to find the potential minimal energy pathway in
>>>>> cases like yours is in general the nudged elastic band approach.
>>>>>
>>>>> Regards
>>>>> Marcella
>>>>>
>>>>>
>>>>> On Thursday, January 19, 2023 at 9:42:15 PM UTC+1 Dobromir A
>>>>> Kalchevski wrote:
>>>>>
>>>>>> Hello everyone!
>>>>>>
>>>>>> I want to make a molecule occasionally approach a surface and react
>>>>>> with it in a NVT ensemble production run. I don't want it fixed in
>>>>>> predetermined coordinates and I don't want it to fly away or to wait for
>>>>>> too long of a trajectory.
>>>>>> For now I don't want to study the reaction's thermodynamics.
>>>>>> Experimentally the reaction occurs.
>>>>>>
>>>>>> I tried the following approaches without success:
>>>>>>
>>>>>> I defined a COLVAR for the distance between one surface atom (Si) and
>>>>>> the center atom (C) of the gas phase reagent molecule (methane)
>>>>>>
>>>>>> &COLVAR
>>>>>> &DISTANCE
>>>>>> AXIS XYZ !DEF XYZ, ANY COMBINATION
>>>>>> ATOMS 50 81
>>>>>> &END DISTANCE
>>>>>> &END COLVAR
>>>>>>
>>>>>> and then I defined a CONSTRAINT
>>>>>>
>>>>>> &CONSTRAINT
>>>>>> &COLLECTIVE
>>>>>> COLVAR 2
>>>>>> INTERMOLECULAR
>>>>>> TARGET [angstrom] 1.87
>>>>>> TARGET_GROWTH [angstrom*fs^-1] 0.25
>>>>>> TARGET_LIMIT [angstrom] 5.0
>>>>>> &END COLLECTIVE
>>>>>> &END CONSTRAINT
>>>>>>
>>>>>> The result is that both atoms get closer too quickly and the system
>>>>>> gets an unphysical geometry - the Si kind of moves too far away from it's
>>>>>> lattice and the C abandons its hydrogens. The remperature rises too quickly
>>>>>> to unbelievable degrees, the calculation slows down and eventually the
>>>>>> program dies.
>>>>>> My cell is defined just fine, with enough space for the gas reagent
>>>>>> to move around. I use a XYZ PBC with space left in one dimension so it
>>>>>> effectively becomes 2D. In that space is the reacting molecule.
>>>>>> I tried playing with TARGET_GROWTH with many values from 0.01 to 5.0
>>>>>> and the result is always the same.
>>>>>> In this group I read that the moment TARGET is set the atoms are also
>>>>>> set at that distance, so I tried switching TARGET and TARGET_LIMIT, but
>>>>>> even then the result is the same - wrong geometry, they get too close too
>>>>>> quickly even with very low TARGET_GROWTH, too high temperature - program
>>>>>> dies.
>>>>>> - Is this a bug, because it makes no sense as expected behavior ?
>>>>>>
>>>>>> I also tried using a RESTRAINT:
>>>>>>
>>>>>> &CONSTRAINT
>>>>>> &COLLECTIVE
>>>>>> COLVAR 2
>>>>>> INTERMOLECULAR
>>>>>> TARGET [angstrom] 3.0
>>>>>> &RESTRAINT
>>>>>> K [kcalmol] 1.0
>>>>>> &END RESTRAINT
>>>>>> &END COLLECTIVE
>>>>>> &END CONSTRAINT
>>>>>>
>>>>>> with the COLVAR being a plane between 3 surface atoms:
>>>>>>
>>>>>> &COLVAR
>>>>>> &DISTANCE_POINT_PLANE
>>>>>> &POINT
>>>>>> TYPE GEO_CENTER
>>>>>> ATOMS 8
>>>>>> &END POINT
>>>>>> &POINT
>>>>>> TYPE GEO_CENTER
>>>>>> ATOMS 21
>>>>>> &END POINT
>>>>>> &POINT
>>>>>> TYPE GEO_CENTER
>>>>>> ATOMS 28
>>>>>> &END POINT
>>>>>> &POINT
>>>>>> TYPE GEO_CENTER
>>>>>> ATOMS 81
>>>>>> &END POINT
>>>>>> ATOMS_PLANE 1 2 3
>>>>>> ATOM_POINT 4
>>>>>> &END DISTANCE_POINT_PLANE
>>>>>> &END COLVAR
>>>>>>
>>>>>> and the gas phase reagent does not stop at the TARGET distance,
>>>>>> instead it gets inserted into the Si lattice, pushing Si atoms away,
>>>>>> inserting both C and H from the methane and one of the hydrogens flied far
>>>>>> away. Temperature goes anomalously high, program slows down, dies. At least
>>>>>> it took some time and steps to get to the wrong point.
>>>>>> - Why didn't the molecule stop at the surface, while experiencing the
>>>>>> forces of the lattice atoms ?
>>>>>> - Why did it do beyond the TARGET distance ?
>>>>>> - Is this a bug ?
>>>>>>
>>>>>> Can I use metadynamics in any way to push the molecule towards the
>>>>>> surface for a reaction to occur ? Once again, I'm not interested in
>>>>>> studying the FES or the rest of the thermodynamics.
>>>>>>
>>>>>> I just need a way to get the gas phase reagent to occasionally get
>>>>>> close enough for a reaction.
>>>>>>
>>>>>> I think the answers to those question can be very useful for many.
>>>>>>
>>>>>> I attach some pictures of the anomalous results. The first two are
>>>>>> from the CONSTRAINT, the 3rd is from the RESTRAINT.
>>>>>>
>>>>>> Best Regards,
>>>>>> Dobromir
>>>>>>
>>>>>
--
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+unsubscribe at googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/a81ee2b6-d739-458c-b167-995e4cda8049n%40googlegroups.com.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20230120/28e5081f/attachment.htm>
More information about the CP2K-user
mailing list