Re: Re: NAMD: Exiting Prematurely

From: Vincent Kraeutler (vincent_at_kraeutler.net)
Date: Wed Aug 15 2007 - 14:06:44 CDT

A SHAKE/RATTLE failure is a fairly generic indication that "something
went wrong" and does not indicate much, other than that the positions
between timesteps changed too much for the algorithm to remain stable.
In fact, I've seen it happen spontaneously (once). There are two classes
of possible causes for this:

a) You're straining the implicit assumptions of the program (e.g. too
large timestep, temperature or pressure coupling is going haywire, weird
cutoff/pairlist settings, a bug, or just bad luck).
b) You're using an interaction function which is too stiff for your
integrator (e.g. wrong setting for exclusions, too stiff forces for
restraint potential or angles, bad force-field parameters).

What I think you should do:
1. make sure that at the end of your equilibration phase temperature,
pressure, and box volume are no longer changing rapidly;
2. assert that extending the equilibration protocol is numerically
stable by letting it run for a longer period of time;
3. starting from the equilibration protocol, change one input parameter
at a time, and let it run for 1000 steps. Repeat until you have obtained
the production protocol.

If you've reached the end of step 3: BINGO. If not: you will have a
better indication of where the problem lies.

BTW, in the case where I had a "bad luck" SHAKE failure, changing the
temperature from 300 K to 300.1 K fixed the problem. ;-)

Cheers,
v.

Arun Krishnan wrote:
> Hi All,
>
> I tried to restart my simulation from just before it crashed and it
> seemed to run for more steps but ultimately
> crashed with a
> ERROR: Constraint failure in RATTLE algorithm for atom 726!
> ERROR: Constraint failure; simulation has become unstable.
> ERROR: Exiting prematurely.
>
> I have taken a look at the archives and the usual suggestions are to
> minimize well before dynamics (I have minimized for upto 3ps), and to
> heat slowly and equilibrate (both of which I have done). I have also
> checked the PMEGrid sizes to see if they are
> ok and it looks like they are. According to someone's rule of thumb,
> there should be one atom for every 10 A^3. I have roughly 39,000 atoms
> and am using 80 as my PMEGridSizeX , y and z which means a total
> volume of 512,000 A^3 which seems to be ok according to the rule of
> thumb.
>
> Do I need to minimize the system again after heating to 500K? That is,
> minimizing just before the production run at 500K? If so how can I do
> that? When I tried to minimize using my files from the equilibration
> at 500K, it kept taking the temperature as 0 and not 500. Or am I
> missing something? Any help will be much appreciated.
>
> Cheers,
>
> Arun
> On 8/15/07, * Arun Krishnan* <krishnan_at_ttck.keio.ac.jp
> <mailto:krishnan_at_ttck.keio.ac.jp>> wrote:
>
> Dear All,
>
> I seem to be running across this weird error. I am trying to
> unfold a small protein. I use the following protocol:
>
> a) energy minimization of protein + water box
> b) heat to 300K
> c) equilibration for 50ps
> d) Production run for 1ns
>
> e) Heat to 500K
> f) equilibrate
> g) production run
>
> All the steps from a-f seem to go fine. I followed the
> temperature/pressure/energy profiles for each step
> and they seem to be fine. However for step (g), the program ends
> abruptly with an "Exiting Prematurely"
> message. That's it.. no other error messages. This happens about
> 2ps into the simulation. I am using an NVT
> ensemble.. I checked the temperature/volume/pressure/energy and
> all are stable. There are no sudden drops
> of any sort.. The only sudden change I found was during stage (f)
> where the total energy during the equilibrate stage
> at 500K rose from -71000 to -66500 or thereabouts and then
> remained stable.
>
> I also checked the DCD file for the production run at 500K and I
> don't see anything out of the ordinary. The protein
> is still well within the waterbox. Am at a loss to understand what
> could be causing the sudden crash. Is there a way
> to find out more about what caused the premature exit?
>
> Thanks,
>
> Arun
>
>
>
>
> --
> ***********************************************
> Arun Krishnan, Ph.D,
> Assistant Professor,
> Institute for Advanced Biosciences,
> Keio University,
> Center Building,
> Tsuruoka, Yamagata 997-0035
> Japan
> Phone: +81 (0)235-29-0824
> Email: krishnan_at_ttck.keio.ac.jp <mailto:krishnan_at_ttck.keio.ac.jp>
> URL: http://www.iab.keio.ac.jp/~krishnan
> <http://www.iab.keio.ac.jp/%7Ekrishnan>
> **********************************************


This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:46:48 CST