I've looked in the documentation, and I've looked in the forums (and charmmtutorial.org, for that matter) but I haven't been able to find the answer so I thought I'd ask here.

When starting dynamics using, for example,

DYNA LEAP VERLET STARt -
NSAVC 1 NSAVV 1 ...

from what time step are the first velocities that are written to the VEL trajectory?

Based on my understanding of the leapfrog integrator, it is likely

v(t0 + 0.5h), where

h is the timestep ... in other words, a half (time) step ahead of the starting position coordinates and a half step behind the first set of position coordinates written to the DCD. Is that correct?

If that is not correct, please tell me what time step the first velocities written to the VEL trajectory are from, when using the LEAP VERLET options as specified above. (And read no further.)

If indeed the first set of velocities written in the .vel trajectory are the velocities at time

t=t0+0.5h, then it would seem that to reverse dynamics, using the negative of those velocities (i.e.,

-v(t0+0.5h)) with the first set of coordinates from the dcd (i.e.,

x(t0+h)) and

not the starting coordinates (i.e.,

x(t0)) would be the way to extend the trajectory backwards in time. The first step in the backwards direction would bring you back to

x(t0). However, it is necessary to use

x(t0+h) as the starting point for integration backwards in time because the earliest velocities that you know are from

t=t0+0.5h. Subsequent steps would bring you into the

t < t0 range. Does that follow?