[gmx-users] Conserved energy ("Conserved En.") in NVT simulation (Wade)

[Johnny Lu]

Does double precision cause a less drop/rise in tilde H over time, when
compared to a mixed precision run under same condition ?

On Sat, Oct 18, 2014 at 11:28 AM, Johnny Lu <johnny.lu128 at gmail.com> wrote:

> Oh, integrating with rectangles under the curve causes error. I guess
> larger time-steps can cause error in tilde H by this method of integration
> alone.
>
> On Sat, Oct 18, 2014 at 11:25 AM, Johnny Lu <johnny.lu128 at gmail.com>
> wrote:
>
>> For V-rescale thermostat, the bussi paper seems to say the tilde H is the
>> integral of the auxiliary dynamics. Not sure if this is same as the
>> conserved energy in gromacs, which is also the integral of the auxiliary
>> dynamics.
>>
>> I guess numerical error causes the system to lose/gain total energy in a
>> linear way. Rise/drop of tilde H may reflect the extra energy that the
>> thermostat has to put in to compensate for this numerical error. [just my
>> guess]
>>
>> If the energy of the system is conserved without the thermostat, tilde H
>> should be constant... I guess.
>>
>> The Bussi paper at [
>> marge.uochb.cas.cz/~roesel/teach/Clanky/paper5_Bussi_v_rescale.pdf] says
>> tilde H is the conserved quantity.
>>
>>
>>
>