Re: I cannot make MD simulation constant in pressure

From: Peter Freddolino (petefred_at_umich.edu)
Date: Mon May 23 2016 - 22:56:48 CDT

Dear Faramarz,
Have you looked at the volume of your system over the course of the simulation? It should not be constant, of course, because the way pressure control works is to change the volume. The key is that you should see it change and then reach an equilibrium value. You should only include data in the RMS calculation from after the point where the volume stabilizes. I would expect the system to shrink somewhat…

Regarding the observed average, is it really significantly different from 1 atm when you have such a large stdev? To even answer that, you’d have to first figure out how many independent samples you have in your trajectory, which means figuring out the correlation time of the instantaneous pressure…

Best,
Peter

> On May 23, 2016, at 11:44 PM, Faramarz Joodaki <fjoodaki_at_my.uri.edu> wrote:
>
> Dear Peter!
>
> Thank you very much for your guidance!
>
> The RMS (or stdev) is 1402.12.
> The extension from average (7.65) is +/-2000.
> The size of my box (in A) is:
> cellBasisVector1 24.818 0.000 0.000
> cellBasisVector2 0.000 24.872 0.000
> cellBasisVector3 0.000 0.000 28.207
> cellOrigin 0.000 0.000 0.000
>
> So the volume of the box is 17.41 nm^3.
>
> I checked the temperature and total energy of the system which are constant over MD simulation. It seems that I am in the equilibrium.
> Do I need to make my box smaller? Actually I have one TRP molecule and water molecules in my box.
> It seems that I need to be sure more about the equilibrium.
>
> Another point: My desire pressure is 1atm but I reached to average of 7.65.
>
> Thanks again!
> Faramarz
>
> On Mon, May 23, 2016 at 10:01 PM, Peter Freddolino <petefred_at_umich.edu> wrote:
> The observed fluctuations are also dependent on the water model, which may be different in the assumptions made by gromacs. See http://www.ks.uiuc.edu/Research/namd/2.9/ug/node37.html for a discussion and formula. For TIP3P water, with an isothermal compressibility of 6*10^-5 atm^-1 at 25 C (Jorgensen and Jensen, JCC 1997), and assuming you have roughly the right density (which gives a box with a volume of 15.2 nm^3), you should get an *RMS* fluctuation of ~200 bar if the theory holds perfectly, but practicalities in the approximations needed to run MD are why, as noted there, "Much larger fluctuations are regularly observed in practice”. So, before going any further — are you sure that your system has equilibrated and you’re only considering points after equilibration in your analysis? And, is the +/- 2000 atm the maximum extent or the RMS? You should be calculating the latter to match up with expectations.
> Best,
> Peter
>
>
> > On May 23, 2016, at 7:29 PM, Faramarz Joodaki <fjoodaki_at_my.uri.edu> wrote:
> >
> > Hi Jeff,
> >
> > Thank you for your response!
> >
> > I read that link. I agree fluctuation for pressure is necessary. But I think there is a big fluctuation in my system:
> > My system has 509 water molecules. The pressure of my system is fluctuating between -2000 and +2000. The number of steps are 1M and I set to report pressure each 100 steps. The average of pressure during this MD simulation is 7.65atm.
> > According to the link which you sent me, the fluctuation should be less than 500. Because the number of water molecules is 509.
> > I would be so grateful if you could give me some suggestions to decrease this amount of fluctuation.
> >
> > Best regards,
> > Faramarz
> >
> >
> > On Mon, May 23, 2016 at 5:06 PM, Jeff Comer <jeffcomer_at_gmail.com> wrote:
> > See: http://www.gromacs.org/Documentation/Terminology/Pressure
> > ​
> >
>
>

This archive was generated by hypermail 2.1.6 : Sun Dec 31 2017 - 23:20:29 CST