[gmx-users] about the velocity output from leap-frog integrators

[Mark Abraham]

On 9/11/2011 9:44 PM, Ravi Bhadauria wrote:
> Hello users,
>
> I have few conflicting answers from the user mailing list about the
> following question.
>
> Using leap-frog integrator, what is the velocity that is WRITTEN in
> the trajectory file for a frame corresponding to r(t)? Is it v(t-dt/2)
> or v(t). Has the protocol for writing velocity in trajectory file been
> consistent from version 3.x to 4.5.x?

With leap-frog, it is v(t-dt/2) in 4.5.x. v(t) is never even computed 
(see do_update_md() in src/mdlib/update.c). I don't know if Erik's 
statement was accurate in 2002, but it isn't accurate now.

Mark

>
> Here are few posts hinting its v(t):
> http://lists.gromacs.org/pipermail/gmx-users/2002-July/001969.html
> http://lists.gromacs.org/pipermail/gmx-users/2003-September/006956.html
>
> Here are couple of posts hinting its v(t-dt/2):
> http://lists.gromacs.org/pipermail/gmx-users/2006-June/022261.html
> http://lists.gromacs.org/pipermail/gmx-users/2003-March/004827.html
>
> Please let me know. The problem is with certain class of simulations
> of bounded flows, these two definitions provide contrasting results
> for suitably averaged velocity.
>
> Sincerely
> Ravi Bhadauria