Contrasting results of dynamics simulation in Dmol3

Hi everyone.

I’m doing a series of dynamic simulations on a new type of molecule which has two Fe atoms at a certain distance from one another.

At the end of a dynamic simulation this distance will assume a certain value, different from the initial one.

If I launch a new dynamic simulation, using the same .xsd document as starting structure and the same simulation settings, I should expect to obtain an identical distance between the atoms as the one of the previous simulation, or at the very least an increasing or decreasing trend.

But that is not so. For example, with a common starting value of 9,919 Å, I have obtained as final values 10,011 Å; 9,782 Å; 9,855 Å; 10,001 Å.

Why is that?

The .xsd document used as the initial structure contains the optimized molecule, so it should not be a problem of non-optimized structure.

Thanks!