Can I print half time step velocities of a run with leap-frog algorithm?

User discussions
1

GROMACS version: gromacs2021
GROMACS modification: No

Hello,

I am working on something a little bit theoretical where I would like to know the exact velocity of my particles in order to compare different systems with similar configurations except a minor change. When calculating the difference of velocity for each particle between my 2 systems I observe little fluctuations. I would like to know if this comes from the way the algorithm calculates the velocities at each step, from averaging the half time step ones.

Is it possible to have access to the half time-step velocities that the leap-frog algorithm calculates, in order to compare with by hand calculations ?

Thank you for your help.

2

With leap-frog times for the velocities in the output are always offset by -deltat/2.

3

Yes now that you’re saying it I notice that I was thinking of the v-verlet algorithm. Is it always the full time-step in the output ? If yes, can we get access to the half time-step ?

4

Also, what does it mean that the output for the velocity is the half time step in the leap-frog ? When I plot my velocities through time, should I take an average of the half time step velocities ?

5

I think VV outputs the full step and you don’t have access to the half step.

I don’t understand you leap frog question. The velocities in the trr file are simply at t-dt/2 for a frame with time stamp t. You can average the velocities at t-dt/2 and t+dt/2 to get a velocity at t, but that will be less accurate than the half step velocities.

6

Thank you for your answer, I got the clarity that I need. I guess I was just confused by the fact that the output of r is at a full timestep and the output of v at a half timestep. They’re therefore shifted. I was then wondering if I should change my time axis when I plot v, to account for this shift.