Dave Case: If you measure charge in units of the electron charge, and distance in Angstroms (as is done in amber), then an electrostatic energy looks like:
E (kcal/mol) = 332 * q1*q2/rwhere q1 and q2 are charges and r is a distance. The square root of 332 is 18.2; hence, to save the multiplication by 332 all of the time, the charges are modified by 18.2, so that
E (kcal/mol) = (18.2*q1) * (18.2*q2) / r
Amber internally uses lengths in angstroms, masses in atomic mass units, and energies in kcal/mol. This means that the unit of time is 1/20.455 ps. Since the set of units is internally consistent, you should be able to compute the kinetic energy in the "normal" fashion:
KE(in kcal/mol) = 1/2 sum mv**2where the masses are in amu and the velocities are from the program.
Note, however, that the velocities stored in the restart file are the velocities at a time 0.5(dt) before the time of coordinate (since a "leap-frog" integration scheme is being used. When amber actually prints kinetic energies, it estimates the velocities at the current time [by averaging those at t - 0.5(dt) and t + 0.5(dt)]. So your kinetic energies won't exactly match those printed unless you do something equivalent.
VDW parameters: R* is in Angstroms, epsilon in kcal/mol.