[gmx-users] Use NVT to mimic NVE

[Johnny Lu]

The simulation get worse in a new way.

I see that thermostat scales velocity and may not fix numerical error in
potential energy.

On Mon, Oct 13, 2014 at 2:07 PM, Johnny Lu <johnny.lu128 at gmail.com> wrote:

> I just want something that will patch up the energy lost due to numerical
> error of NVE simulation of the system ran by mixed precision gromacs,
> instead of a thermostat.
>
> I hope the velocity rescaling is sufficiently uncorrelated with the motion
> of the protein.
>
> So far, the simulations ran 2.5 ns, and the fluctuation of total energy
> seems to be within 1% of total energy, with a 100ps or 1ns coupling
> constant.
>
> May be I should try a 10 ns coupling constant, just to see if the
> simulation would go belly up.
>
> On Mon, Oct 13, 2014 at 11:02 AM, Michael Shirts <mrshirts at gmail.com>
> wrote:
>
>> > I guess, if I pick a coupling constant that is just small enough to
>> keep the
>> energy conserved, I would get a NVT simulation that is as close as a NVE
>> simulation as possible.
>>
>> > Is this correct?
>>
>> Yes, but then at that point the thermostat isn't actually thermostatting.
>> The Bussi comment is merely to show that his thermostat correctly reduces
>> to Newton's law in the limit, not that it would be useful to run it in
>> that
>> limit.
>>
>> On Mon, Oct 13, 2014 at 10:28 AM, Johnny Lu <johnny.lu128 at gmail.com>
>> wrote:
>>
>> > On page 014101-3, the Bussi paper (http://dx.doi.org/10.1063/1.2408420)
>> > mentioned: "On the other hand, for coupling constant approaching
>> > infinity,the Hamiltonian dynamics is recovered."
>> > Does that means that for a large enough coupling constant, the
>> velocities
>> > are nearly not rescaled, and the dynamics (like rate of motion) would be
>> > same as that of NVE?
>> >
>> > A larger coupling constant, means a smaller diffusion coefficient in the
>> > axillary dynamics by equation 6.
>> >
>> > While the effects of the velocity rescaling at each step will
>> accumulate, a
>> > larger coupling constant means the thermostat perturb less of the
>> dynamics,
>> > and the resulting dynamics is closer to a NVE simulation.
>> > There is no worry that the thermostat would suddenly rescale the
>> dynamics
>> > every x step, because in the procedure of the thermostat, the velocities
>> > are rescaled every step, regardless of the coupling constant.
>> >
>> > I guess, if I pick a coupling constant that is just small enough to keep
>> > the energy conserved, I would get a NVT simulation that is as close as a
>> > NVE simulation as possible.
>> >
>> > Is this correct?
>> > --
>> > Gromacs Users mailing list
>> >
>> > * Please search the archive at
>> > http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before
>> > posting!
>> >
>> > * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>> >
>> > * For (un)subscribe requests visit
>> > https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or
>> > send a mail to gmx-users-request at gromacs.org.
>> >
>> --
>> Gromacs Users mailing list
>>
>> * Please search the archive at
>> http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before
>> posting!
>>
>> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>>
>> * For (un)subscribe requests visit
>> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or
>> send a mail to gmx-users-request at gromacs.org.
>>
>
>