# [gmx-users] microcanonical

Mark Abraham Mark.Abraham at anu.edu.au
Thu May 18 11:02:33 CEST 2006

```karamyog singh wrote:
> I placed the atoms randomly means i used random atomic positions, not
> velocities. The velocities were 0 initially and then I generated
> velocities at 0.01 K.(1st MD run to get a stable configuration) After
> that I subjected this structure that I got to NVE.(2nd MD run) Now over
> here I will try with no velocity generation. Thank you for pointing out
> this mistake. I should have realised this. One more thing, is generating
> velocities in the MD run correct?

It isn't meaningful to have velocities before starting an MD run, so you
want to generate them at some point before any MD, then equilibrate as
appropriate.

> I will try what you are saying. I was doing it till the r-min beacuse at
> that point the forces become 0.Minimum potential should imply 0 force.
> So if the forces are zero and I increase my simulation time for 1st MD
> run, I get a stabler structure i.e. the forces between atoms should be
> nearly equal to zero. However my forces wil never be zero. Wil I ever
> get a frozen structure for NVE?

The minimum of the potential is where -F=dU/dx=0, not where U=0 - after
all the latter depends where you take your convention of zero energy. A
normal (e.g. 6,12) LJ potential has a minimum below zero and then
converges up to it (e.g. Fig 4.1 in gromacs manual). You want your
cutoff to be large enough that your potential is getting close to zero
under the standard convention that zero is at infinite separation -
because the cutoff means that the potential effectively is zero. My
point is that a cut-off at the minimum gives you a sharp discontinuity
for your effective potential, which means you are doing something other
than you want to do.

Mark

```