From owner-chemistry ":at:" ccl.net Fri Nov 18 14:19:00 2005
From: "Konrad Hinsen khinsen::cea.fr" <owner-chemistry[A]server.ccl.net>
To: CCL
Subject: CCL: Tinker: harmonic torsion potential
Message-Id: <-30000-051118141443-2613-WHMdOwdiAMMuaDJQp+4Mzg[A]server.ccl.net>
X-Original-From: "Konrad  Hinsen" <khinsen(!)cea.fr>


Sent to CCL by: "Konrad  Hinsen" [khinsen*cea.fr]
On 18.11.2005, at 16:10, Mark Thompson mark-x-arguslab.com wrote:

> One trick that I've used is to verify MM first derivatives is to write a 
> numerical gradient routine to go along with analytical first derivative 
> code.  Then choose high symmetry cases (small test systems) centered on 
> the origin and tricks like that to test that numerical and analytical 

Another useful tool is automatic derivatives, in particular for non-trivial energy terms, as they 
permit a simple inspection of intermediate values.

For the Python users out there, I provide a nicely packaged potential energy derivative module in 
my ScientificPython package, which is available at

	http://dirac.cnrs-orleans.fr/ScientificPython/

An example code for that module is:

	from Scientific.Physics.Potential import PotentialWithGradients
	from Scientific.Geometry import Vector

	def _harmonic(k,r1,r2):
	    dr = r2-r1
	    return k*dr*dr
	harmonic = PotentialWithGradients(_harmonic)

	energy, gradients = harmonic(1., Vector(0,3,1), Vector(1,2,0))
	print energy, gradients

This prints

3.0 [Vector(-2.0,2.0,2.0), Vector(2.0,-2.0,-2.0)]

There is also a version for second derivatives. ScientificPython also has "raw" automatic derivatives 
of arbitrary order in Scientific.Functions.Derivatives.
--
-------------------------------------------------------------------------------
Konrad Hinsen
Laboratoire Leon Brillouin (CEA-CNRS), CEA Saclay,
91191 Gif-sur-Yvette Cedex, France
Tel.: +33-1 69 08 79 25
Fax: +33-1 69 08 82 61
E-Mail: khinsen*_*cea.fr
-------------------------------------------------------------------------------