AMBER Archive (2006)

Subject: Re: AMBER: NTT for production run

From: Vlad Cojocaru (Vlad.Cojocaru_at_eml-r.villa-bosch.de)
Date: Wed Mar 15 2006 - 02:53:46 CST


Dear David,

At least in explicit water, according to tests that I have performed,
ntt=0 in production runs leads to steady decrease in T (and kinetic
energy).

Prior, I equilibrated the system for 100 ps (heating from 10.0 to 300.0
degrees with Langevin dynamics gamma_ln=1.0 for 20 ps - NVT; 30 ps of ct
T at 300K with gamma_ln=0.5 - NVT; and finally 50 ps of ct. pressure
equilibration). The system looked like it equilibrated well (from plots
of rmsd, T, P, density or Energy terms).

After these 100 ps I switched to production runs both with ntt=1,
tautp=5.0 and with ntt=0 ... And I definetely observed a steady decrease
in T and Ek. That's when I decided to go further with ntt=1, tautp=5.0 ....

The system was disolved in a trucated octahedron box with 12 A buffer
distance.

So, since you say that you are not using ntt=1 anymore, I presume that
you did not encounter such problems ... However, I believe that on the
Amber website, such leakages are mentioned at some point....

Maybe you can briefly comment on this.

Best wishes
vlad

David A. Case wrote:

>On Tue, Mar 14, 2006, Vlad Cojocaru wrote:
>
>
>>ntt=1 with a higher tautp (weak coupling) should be preferred ...
>>
>>
>
>ntt=1 has a the least effect on *dynamics*, and probably works pretty
>well with a large tautp (>2.0). However, if you are really interested
>in dynamics, you should set ntt=0 for production; one a system is
>equilibrated, there is often no need for further temperature regulation.
>
>ntt=3 will most reliable get you a canonical distribution of configurations.
>You should prefer this for free-energy simulations, for example.
>
>There are many bad things that can happen with ntt=1, especially with small
>values of tautp. The Amber 9 Users' Manual has the following sentence
>(in italics): "Please don't use ntt=1 unless you are sure you know what you
>are doing." (This is especially true for implicit solvent simulations.)
>
>Your mileage may vary, but I almost never use ntt=1 anymore.
>
>...dac
>-----------------------------------------------------------------------
>The AMBER Mail Reflector
>To post, send mail to amber_at_scripps.edu
>To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu
>
>
>

-- 
Dr. Vlad Cojocaru
EML Research gGmbH
Molecular and Cellular Modeling Group
Schloss-Wolfsbrunnenweg 33
69118 Heidelberg, Germany
Phone: +49-6221-533266
Fax: +49-6221-533298
e-mail: Vlad.Cojocaru_at_eml-r.villa-bosch.de
http://projects.villa-bosch.de/mcm/people/cojocaru/

----------------------------------------------------------------------- The AMBER Mail Reflector To post, send mail to amber_at_scripps.edu To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu