Re: [AMBER] SHAKE flucatuation

From: Thomas Cheatham <>
Date: Thu, 10 Oct 2013 13:19:09 -0600 (Mountain Daylight Time)

> I ran a MD simulation with pmemd.cuda for a pure water box of 1994 water
> molecules with SHAKE algorithm turned on. But when I checked the bond
> length and H-O-H angle of the water molecules, they still fluctuated for
> around 0.1~0.2. In my case I need to make the water molecule perfectly
> rigid.

If the O-H or O-O bond length fluctuates by 0.1-0.2 Angstroms, then this
is a real problem as there would be no energy conservation, likely the
integrator would fail, etc...

I just checked one of my simulations and see that the fluctuations
(standard deviation) of O-H or H1-H2 bond lengths are ~0.0004 A and
~0.04 degrees for the H1-O-H2 angle for a salt solution that was run with a
SHAKE tolerance of 0.00001, so how are you estimating the ranges?

If I plot the time dependent bond/angle length, the range of values is
approximately +/-0.0015 A or +/-0.1 degree.

If you want it to be "more" rigid, increase the SHAKE tolerance. However,
it is numerically difficult (effectively infinitely small tolerance) to
get it completely rigid.


AMBER mailing list
Received on Thu Oct 10 2013 - 12:30:03 PDT
Custom Search