On Fri, May 17, 2013, Anselm Horn wrote:
>
> in their book "A guide to biomolecular simulations" by Becker and
> Karplus, a general convergence criterion for MD simulations is presented
> (p. 79 in 1st edition) without any further reference:
> log_10 (<E(t)-<E>> / <E>) < -3
I remember this, but never liked the formula very much, since it depends on
the zero of energy, which is an arbitrary value. To me, a more useful measure
of energy convergence (not "general convergence") is the ratio of fluctuations
in the total energy divided by fluctuations in the kinetic energy.
It is also of considerable importance to examine "drift" in the average value
of the kinetic energy, which can be systematic and considerable, even when
there is an apparent conservation of energy over local time slices.
Both measure of energy conservation seem to be roughly proportional to the
size of the system, and so are often reported as values per degree of freedom.
...dac
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri May 17 2013 - 17:30:02 PDT
