AMBER Archive (2004)

Subject: Re: AMBER: Nmode convergence

From: ML (
Date: Wed Feb 11 2004 - 02:14:31 CST

Dear Dr. Case and the AMBER community,
I've just found how to compile nmode so as to be able to increase
nonbond cutoff to 99.0. I had to increase integer memory to cca.
10MWords (for a 3000-atom protein).

I assume now that the convergence problem was due to both, imposing
non-bonded cutoff and re-creating pairlist. Now I hope to succeed just
by using conjug grad mini and normal modes.

Thank you very much for your help.


 Martin Lepsik

David A. Case wrote:

>On Tue, Feb 10, 2004, ML wrote:
>>I minimize a 3000-atom protein with Newton-Raphson (after conjugate
>>grad.) as I want to calculate normal modes. I use convergence criterion
>>of E10-4 or 5E10-5 but for the normals modes it's much higher (E10-3 to
>>5E10-4). I understand that this has to do with re-creating the pairlist
>>when restarting but how to proceed? (the same thing happened after
>>conj.grad. mini)
>Set your cutoff to 9999. This will make the minimization converge
>more smoothly, and will remove any effect of changing cutoffs.
>I generally find that conjugate gradient minimization is need
>for Newton-Raphson. Just be patient (i.e. run lots of cycles of
>...good luck...dac

Martin Lepsik
Dept. of Molecular Modeling
Institute of Organic Chemistry & Biochemistry
Czech Academy of Sciences
phone: +420-220 183 540
fax:+420-220 183 292

----------------------------------------------------------------------- The AMBER Mail Reflector To post, send mail to To unsubscribe, send "unsubscribe amber" to