Dear Ross,
Thanks! This explains a lot.
On Wed, 25 Aug 2004, Ross Walker wrote:
> Hi Hang,
> ...
> > Does this also mean that one has no control nor
> > prediction over how long this minimizing process will go?
> To do what you mentioned above you would just set Maxcyc to some very high
> number - like 100,000. In this case, yes, you would have to either wait
> until 100,000 steps had elapsed or the RMS had decreased below the
> convergence criteria. Note, the steepest descent method will not normally
> get you down much below 10^-2. Conjugate gradient will get you further but
> you probably won't ever get to very very tight convergence like you could
> with a second derivative method.
Ah, I see. If I understand correctly, the default in Amber8 is the
conjugate gradient method (switch on/off using ncyc). The manual also has
reference and intruction for LMOD method. However, I haven't found the
second derivative method anywhere. Could you please give me some
reference? Is it possible to try in Amber8?
> Failing to minimise here would probably have led to very large initial
> forces and hence velocities which could have led to the system "blowing up".
> Typically you should make an estimate at how many minimisation steps to run.
> The bigger your system the more degrees of freedom so the more steps you
> need. Running too many steps before MD won't do you any harm, it will just
> waste cpu time.
I totally agree!
Thanks,
--Hang
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Thu Sep 02 2004 - 01:15:11 PDT
