[gmx-users] SD and CG EM

[Martina Bertsch, PhD]

Dear Luciano,

When a message appears during an attempt to minimize a system using SD:

Stepsize too small... Converged to machine precision..."

it essentially means that the SD algorithm has problems and that it is time to switch to the CG.

It may also be helpful to delete the SOL molecules that experience excessive forces during both SD and CG and reminimize. Monitor forces in the standard (screen) output during EM.

You may also find that EM performs better with double precision fftw libraries. To recompile gmx with double precision: 

make clean
./configure --disable-float
make
make install

For the fftw with double precision it is just:

./configure
etc.

Finally, it may be necessary to submit the system to multiple iterations of SD and CG, gradually decreasing emtol, followed by deletion of unwanted atoms with large forces after each iteration.

Regards,
Martina Bertsch, Ph.D.
Northwestern University
Feinberg School of Medicine
Molecular Pharmacology and Biological Chemistry
303 East Chicago Avenue
Chicago, IL 60611



Luciano wrote:

>Hi gmx-users and specially David and Anton very very thanks for advise.
>
>I have been read the last mails about EM (SD and CG) and doing EM to my
>system (
>protein + water) without pbc it have showed to emtol = 10000:
>
>Conjugate Gradients:
>   Tolerance         =  1.00000e+04
>   Number of steps   =         1000
>   F-Norm            =  1.21548e+05
>
>Step 1, E-Pot = 7.2018726562e+04, Fmax =  8.01764e+03, atom = 1155
>
>writing lowest energy coordinates.
>
>Back Off! I just backed up octa20orig.trr to ./#octa20orig.trr.15#
>
>Back Off! I just backed up after_em_pr.gro to ./#after_em_pr.gro.15#
>
>
>But when I set emtol = 7000 to next step the result was:
>
>Conjugate Gradients converged to 10000 in 2 steps
>  Potential Energy  =  7.20187e+04
>Maximum force:  8.01764e+03
>
>
>Negative w:  -5.389869190490e+21
>z=  -2.909245276160e+11
>gpa=   3.167263553258e+11, gpb=   2.842420549736e+11
>a=   0.000000000000e+00, b=   1.130758611745e-10
>EpotA=   7.016328906250e+04, EpotB=   7.019690625000e+04
>Negative number for sqrt encountered (-5389869190490223542272.000000)
>Terminating minimization
>
>writing lowest energy coordinates.
>
>Back Off! I just backed up octa20orig.trr to ./#octa20orig.trr.16#
>
>Back Off! I just backed up after_em_pr.gro to ./#after_em_pr.gro.16#
>
>Conjugate Gradients did not converge in 3 steps
>  Potential Energy  =  7.01633e+04
>Maximum force:  3.67083e+05
>What are the terms z, gpa, gpb, a and b? Why the negative number for sqrt?!
>What I should do?! About the message of Martina Bert, Sun, 16 Nov 2003
>01:53:35 -0600, what the matter when the message comes up?
>
>"Stepsize too small (... nm) Converged to machine precision"
>
>Thanks in advance...
>
>Luciano
>
>
>######################################################
>               Luciano Tavares da Costa
>             Mestre em Engenharia Química
>   icq #232852036 ; ltcosta at universiabrasil.net  
>    Doutorando Laboratório de Biofisicoquímica
>              Instituto de Química - USP
>             
>######################################################
>
>
>------------------------------------------------------------------------
>Vote no Universia para o Prêmio IBest 2004 e concorra a uma
>viagem para Trancoso.
>www.universiabrasil.net
>
> Part 1.1
>
> Content-Type:
>
> text/plain
> Content-Encoding:
>
> quoted-printable
>
>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20031120/a4f2b15f/attachment.html>