[gmx-users] g_energy sum

[Berk Hess]

>From: "merc mertens" <merc.mertens at gmx.net>
>Reply-To: Discussion list for GROMACS users <gmx-users at gromacs.org>
>To: Discussion list for GROMACS users <gmx-users at gromacs.org>
>Subject: Re: [gmx-users] g_energy sum
>Date: Mon, 30 Apr 2007 10:00:11 +0200
>
>hello,
>
>you are right but "summing" over 1 term was done only for testing purposes. 
>first i tried to sum up all lj terms but ran into the same problem.
>
>nevertheless i finally found the reason for the difference. the energy file 
>is from a rerun. during the setup mdrun complained that not all frames of 
>the trajectory used for rerunning contained velocities. therefore i reran a 
>converted trajectory containing only the frames with velocities (i.e. only 
>every 100th frame). the resulting energy.xvg contained 21 energy values but 
>the sum was averaged over 2000001 steps (see below). this explains the 
>small average value from "-sum".
>
>this leads to the next question: was it wrong to only use the frames 
>containing velocities in the rerun??
>
>thanks for the help and suggestions,
>
>merc

Ah, that explains it indeed.

Unfortunately mdrun -rerun produces incorrect averages and fluctuations
in the energy file when you do not use EVERY frame.
I think this problem is currently not documented.
Only using every frame (nstxout=1 for the original run) will solve this 
problem.

But anyhow, your average from the xvg file is fine and accurate.
For energies you can use all frames, not only the ones with velocities.
Only the kinetic energy and the virial are affected by the velocities.

Berk.

_________________________________________________________________
FREE pop-up blocking with the new Windows Live Toolbar - get it now! 
http://toolbar.msn.click-url.com/go/onm00200415ave/direct/01/