Re: Equilibration when splitting an ABF calculation

From: Jérôme Hénin (jhenin_at_ifr88.cnrs-mrs.fr)
Date: Fri Jan 28 2011 - 06:44:32 CST

Dear Ajasja,

Usually, the colvar spends a small amount of time in the "wall"
region, hence, only a few samples are missing from the ABF grid. Note
that you could define the colvar boundaries to be _wider_ than the
interval between the walls: then samples would be collected throughout
the simulation. In principle, samples from the wall region are
statistically valid for ABF purposes (as long as the colvars are
mutually orthogonal). The tradition is for the walls and grid
boundaries to coincide, but there are no strong rational grounds to
this.

Cheers,
Jerome

On 28 January 2011 13:26, Ajasja Ljubetiè <ajasja.ljubetic_at_gmail.com> wrote:
> Dear Jerome,
> The pulling went OK, the equilibration as well. The ABF is running smoothly,
> only do you think I would I need a higher wall constant? Some points are
> outside the regions, but since the wall potential is a harmonic potential,
> there are always some points outside the region, right?
> Best regards,
> Ajasja
>
> On Thu, Jan 27, 2011 at 15:48, Ajasja Ljubetiè <ajasja.ljubetic_at_gmail.com>
> wrote:
>>
>> Dear Jerome,
>>
>> >Even simpler: you can enable the colvars, including their wall
>> >potentials, but not ABF (e.g. comment out the abf block).
>> Aha, so the upper and lower wall potentials are active as soon as a colvar
>> is defined?  That's good to know, thanks.
>> Best regards,
>> Ajasja
>> On Thu, Jan 27, 2011 at 15:08, Jérôme Hénin <jhenin_at_ifr88.cnrs-mrs.fr>
>> wrote:
>>>
>>> Hi Ajasja,
>>>
>>> On 27 January 2011 13:25, Ajasja Ljubetiè <ajasja.ljubetic_at_gmail.com>
>>> wrote:
>>> > Dear all,
>>> > I partitioned the FES of my system into eight smaller parts. For each
>>> > part I
>>> > wanted the system to be inside the region of interest (ROI)  when
>>> > starting
>>> > the calculation, so puled the each part to the desired location (using
>>> > the
>>> > harmonic keyword of the colvars module - fig I). I then slowly relaxed
>>> > the constraint and let each part equilibrate for a while. For some
>>> > parts the
>>> > system escaped from the ROI during equlibration (fig II - at least F
>>> > and B
>>> > are out of ROI). So when the ABF simulation started the system was not
>>> > inside the boundaries defined for the colvars for some parts.
>>> > Two questions:
>>> >
>>> > Is this a problem? When the ABF started the wall potential will push
>>> > the
>>> > system to the right region, but probably with some input of energy.
>>>
>>> Indeed, and that defeats the purpose of equilibrating before that...
>>>
>>> > How could I have ensured the the system stays in the ROI during
>>> > equlibration? Should I have ABF turned on during equilibration, but
>>> > with a
>>> > very high fullsamples, so only the wall potential is active?
>>>
>>> Even simpler: you can enable the colvars, including their wall
>>> potentials, but not ABF (e.g. comment out the abf block).
>>>
>>> Cheers,
>>> Jerome
>>
>
>

This archive was generated by hypermail 2.1.6 : Mon Dec 31 2012 - 23:19:45 CST