[gmx-users] Strange dgdl-value together with lincs

[Maik Goette]

David, Berk,

 > Hold on, if you are doing this with hardcore, there is a singularity
 > in dG/dlambda. It won't be numerically integrable so whatever you
 > compute will be in error. See my recent JCP paper that I referred you
 > to before.

thats what I thought.
With the latest patched code, I get really good results for the 
three-step process and, for the hardcore LJ a deviation, which really 
comes from the singularities, I guess.
Now, the 10ns hardcore slow growth and the 3-step discrete lambda 
sampling yield perfectly the same number and this is fine.
I will now continue with computing the non-equilibrium values for the 
switching process, but I think, those numbers will be ok, too.

 > For what it's worth, too, a delta_lambda of 0.02 is insanely fast
 > "slow growth".

agreed ;)
This was just for bug tracking issues. The step I will use for 
production will be slightly smaller :)

Thanks for all your help.

Btw, cause I already uploaded the pics, here they are:
The 1-step hardcore process:
http://www.mpibpc.mpg.de/groups/grubmueller/start/people/mgoette/images/tot.png
3-step QQ off:
http://www.mpibpc.mpg.de/groups/grubmueller/start/people/mgoette/images/qqoff.png
3-step QQ on:
http://www.mpibpc.mpg.de/groups/grubmueller/start/people/mgoette/images/qqon.png
3-step LJ and bonded softcore:
http://www.mpibpc.mpg.de/groups/grubmueller/start/people/mgoette/images/vdw_sc.png
3-step LJ and bonded hardcore:
http://www.mpibpc.mpg.de/groups/grubmueller/start/people/mgoette/images/vdw_hc.png

This actually proofs again the problem of singularities while using 
hardcore LJ morphs. Anyway, the softcore part looks fine and the sum of 
the single contributions fits quite well.

As I found out recently, the GROMACS version on the cluster I used, was 
without the november bug fix, so I can't tell which fix helped with the 
problem. Sorry for that.

Regards

Maik Goette, Dipl. Biol.
Max Planck Institute for Biophysical Chemistry
Theoretical & computational biophysics department
Am Fassberg 11
37077 Goettingen
Germany
Tel.  : ++49 551 201 2310
Fax   : ++49 551 201 2302
Email : mgoette[at]mpi-bpc.mpg.de
         mgoette2[at]gwdg.de
WWW   : http://www.mpibpc.gwdg.de/groups/grubmueller/


David Mobley wrote:
> Maik,
> 
>>  Assume, you take the interconversion of ethane to methanol in solvent in
>>  one step.
>>  You sample hardcore at 11 evenly spaced lambda values for, lets say, 2
>>  ns each.
>>  You get the dG/dl mean from every run and integrate via simpson to get a
>>  total DG.
> 
> Hold on, if you are doing this with hardcore, there is a singularity
> in dG/dlambda. It won't be numerically integrable so whatever you
> compute will be in error. See my recent JCP paper that I referred you
> to before.
> 
> I think if you are using hardcore, you *have* to use slow growth and
> the Jarzynski or Crooks expressions, or do some sort of polynomial fit
> to the singularity in dG/dlambda and integrate the polynomial.
> Integrating dG/dlambda directly will simply fail since you can't
> numerically integrate a singularity.
> 
> If you're using softcore, you have more options, and simple numerical
> integration can work.
> 
> For what it's worth, too, a delta_lambda of 0.02 is insanely fast
> "slow growth".
> 
> David
> _______________________________________________
> gmx-users mailing list    gmx-users at gromacs.org
> http://www.gromacs.org/mailman/listinfo/gmx-users
> Please search the archive at http://www.gromacs.org/search before posting!
> Please don't post (un)subscribe requests to the list. Use the 
> www interface or send it to gmx-users-request at gromacs.org.
> Can't post? Read http://www.gromacs.org/mailing_lists/users.php
> 
> .
>