I have been tunning the timestep until very short ones (i think 0.01 fseg) but the problem appears (more or less in the same time point if you repeat the test with a short timestep). checking some parameters in the potential code it seems that the bondorder term in the potential file (bij) (if you can write the main part of the potential as Vr + bij*Va the bij is the bondorder term) takes bad (or far from you can expect) values. I am searching in this way for a posible Spline fail but i am no completly sure. Anyway is sure the temperature blows up with very short timesteps.
Diego
2009/2/6 <[email protected]>