Re: [AMBER] nmode and newton optimization divergence

From: David A Case <>
Date: Tue, 5 Nov 2013 07:34:56 -0500

On Tue, Nov 05, 2013, maghtar wrote:

> I have a molecule consist of 81 atoms in vacume and I want to calculate the
> normal modes of it. The first conjgrad minimization is done successfully
> but in Newton-Raphson after the first step the energy goes to infinity.
> Does Anyone have any idea what the problem can be? Following you can see
> the input and output.
> nblist_box, number of atoms= 81

You say that your molecule is in vacuum, but it looks like the prmtop file
has box information, and nab is trying to use it. However, normal modes
(or Newton Raphson, which also uses second derivatives) in nab do not
support periodic boundary conditions (and, in fact, the periodic support
in nab is rather limited).

I will add a test to make sure that the code gives a error message if you
request second derivatives for a periodic system. But for now, you need to
build a non-periodic system.

Also, you may need to to get a gradient lower than 0.005 before starting

Finally, I almost never use Newton Raphson any more (although it should be
fine for such a small system). The xmin minimizer will generally get you to
an acceptable minimum faster.

...good luck...dac

AMBER mailing list
Received on Tue Nov 05 2013 - 05:00:03 PST
Custom Search