[gmx-developers] problems with trajectory reversal

[Jarek Juraszek]

Dear list users,

I need to run a trajectory back in time. To acheive this I tried to 
reverse velocities in the gro and later in the trr file as well.
Unfortunatelly this doesn't work properly as it looks like the kinetic 
energy is not beeing conserved. 
My system is trp-cage (1L2Y) solvated in water, dodecahedron box.
I do the following : 

1. short 5 ps run
2. 5ps continuation run (energy seems OK)
3. run with reversed momenta (kinetic energy increases significantly
    over the fluctuations) 

I first thought it might be a problem of precision but when I looked 
at the debug-log file the kinetic energy is actually well calculated
in the first evaluation. After the second evaluation the energy increases. 
I am running without the removal of center of mass motion, without 
pressure and temperature coupling. 

Is there an easy way of reversing my trajectory or should I somehow 
modify the code to get proper results? What is Gromacs doing with the 
kinetic energy at the initialisation that it seems to increase ?

Regards

Jarek Juraszek