Re: Rise in Total Energy in NVE

From: Leandro Martínez (leandromartinez98_at_gmail.com)
Date: Thu Jul 10 2008 - 07:28:18 CDT

Hi Sourav,
Using a water box is generally better because then you can use
periodic boundary conditions. Your 15A solvation shell would
become about 30A solvation shell (the distance to the image protein image),
and furthermore the whole environment is more biologically
realistic than one in which a spherical boundary was introduced.

Of course this has a computational cost. Therefore, if you would
like to simulate in a "drop" of water (I have done that already) to
save computer time, you actually don't need to use spherical
boundary conditions. Just put the protein in the water sphere
and let it run. The water molecules won't escape from the drop
of water (only a very few will). This approach (the drop or spherical
water approach) will limit the simulation times, because the protein
will diffuse out of the water box. From my experience, after
a few nanoseconds a 20A solvation shell is not enough. Or you
could restrain your protein to be in the center of the water
sphere, but than it will be yet another artificial force in the system.

Leandro.

On Thu, Jul 10, 2008 at 1:30 PM, S.K. Ghosh <skg30_at_cam.ac.uk> wrote:

> Hi NAMD users,
>
> Thanks to Chris for his suggestion.
>
> And is use of water sphere OK? I am aware that for a water sphere I am
> using an artificial surface tension and also not using long range
> electrostatics, but, should the values of force for unbinding vary for these
> two cases (sphere vs box) - given that I am using a 15A padding for a
> protein of radius 35A?
>
> Could someone please comment and/or point to some literature/discussion on
> justificatons of use of water sphere vs water box.
>
> Thanks,
> Sourav
>
> On Jul 10 2008, Chris Harrison wrote:
>
> Sourav,
>> 1 fs should be adequate. If significant drift continues to occur, you
>> might attempt 0.5fs to see if it decreases drift, but 1fs should be
>> sufficient.
>>
>> -Chris
>>
>>
>> On Wed, Jul 9, 2008 at 12:33 PM, S.K. Ghosh <skg30_at_cam.ac.uk> wrote:
>>
>>> Peter,
>>>
>>> Thanks for your suggestion. Is 1 fs OK to work with non-rigidbonds? or
>>> would 0.5 fs be more appropriate?
>>>
>>> Thanks,
>>> Sourav
>>>
>>> On Jul 9 2008, Peter Freddolino wrote:
>>>
>>> Hi Sourav,
>>>> spherical boundary conditions do work on the system (since they're
>>>> applying forces). A rise in energy should not be surprising.
>>>> Also, you may need to avoid multiple time stepping (ie, change
>>>> fullElectFrequency to 1) to get perfect energy conservation.
>>>>
>>>> Peter
>>>>
>>>> S.K. Ghosh wrote:
>>>>
>>>>>
>>>>> Hi Peter,
>>>>>
>>>>> I am now sending this to the group.
>>>>>
>>>>> Hi All,
>>>>>
>>>>> I find the total energy rising in an NVE run:
>>>>>
>>>>> First I ran an NVT (1 fs timestep and rigidbonds off) and then an NVE
>>>>> (again 1 fs timestep and rigidbonds off) starting from where the NVT run
>>>>> ended. For this NVE run, I find that the Total Energy is slightly increasing
>>>>> initially (steady rise but approx 30 kcal/mol in total) for 1 ns and looks
>>>>> to be almost flat for the last 1 ns (please find attached). The temperature,
>>>>> however, does not rise at all.
>>>>>
>>>>> I am curious to find out why the energy rises. Could someone comment on
>>>>> that please. Is 1 fs timestep high with rigidbonds turned off? Should I use
>>>>> 0.5 fs?
>>>>>
>>>>> Or is this because of the following multistepping that I am using.
>>>>>
>>>>> timestep 1
>>>>> #rigidBonds all
>>>>> nonbondedFreq 1
>>>>> fullElectFrequency 2 stepspercycle 10
>>>>>
>>>>> I have attached the energy values for 1st and last timestep here. I can
>>>>> see that DIHED, IMPRP and ELECT have increased and VDW and BOUNDARY have
>>>>> decreased.
>>>>>
>>>>> In both the runs, I have used spherical boundary conditions.
>>>>>
>>>>> Your comments would be much appreciated.
>>>>>
>>>>> Thanks,
>>>>> Sourav
>>>>>
>>>>>
>>>>> On Jul 7 2008, Peter Freddolino wrote:
>>>>>
>>>>> Hi Sourav, this kind of question is really better addressed to namd-l,
>>>>>> so that other users can learn from your question (and search for it later).
>>>>>> For this increase in total energy, what other simulation parameters
>>>>>> (timestep, etc) are you using? Is the energy rise continuous or does it
>>>>>> fluctuate? If you use 1 fs timesteps without multiple timestepping, you
>>>>>> should conserve energy quite well over time. Peter
>>>>>>
>>>>>> S.K. Ghosh wrote:
>>>>>>
>>>>>>>
>>>>>>> Hi Peter,
>>>>>>>
>>>>>>> Starting from where an NVT ended, I ran an NVE i.e. turned off the
>>>>>>> constant temperature control (the pressure control was turned off for both).
>>>>>>> I found that the temperature was pretty stable but the Total Energy curve
>>>>>>> showed slight increase (around 30 kcal/mol). There was no change in timestep
>>>>>>> or rigidbonds or any other parameters between the two runs. Though the
>>>>>>> energy rise is comparatively quite low, is this something normal? I would be
>>>>>>> interested to know why the Total Energy rises even though I am running an
>>>>>>> NVE run. Could you please comment on this.
>>>>>>>
>>>>>>> Thanks,
>>>>>>> Sourav
>>>>>>>
>>>>>>
>>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>
>>>
>>>
>>
>>
>>
>>
>
>
>

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:49:38 CST