[gmx-users] Re: mdrun -rerun and box size bug not fixed?

[Michel Cuendet]

Dear David and Berk,

I have performed further tests on a box of 1136 waters with one sodium 
atom, to see if mdrun -rerun was able to reproduce the Lennard-Jones 
energies calculated during the run from frames stored in an xtc 
trajectory. I used three .gro files with grompp :
    - confin.gro, corresponds to step 0
    - smallbox.gro, smaller box size
    - bigbox.gro, very large box size
I checked the behavior of gromacs 3.3.0 with the bug fix provided in 
bugzilla (3.3.0_patch) and gromacs 3.3.1. Here are my observations:

1) Neither 3.3.0_patch nor 3.3.1 reproduce the LJ energies of the run.
2) 3.3.0_patch gives results that are independent of the .gro file.
3) 3.3.1 results depend on the .gro file. A small box makes energies 
explode.
4) 3.3.1 with a .gro corresponding to the initial condition gives the 
same results as 3.3.0_patch.
5) 3.3.1 with a very large box (no interaction between images ?) gives a 
good correlation with the run.

I attach an eps figure illustrating the points above. The deviations 
between 3.3.0_patch and the energies from the run are about 8% of the 
total value, which can not be explained by the imprecisions of the 
positions in the .xtc file.

Below is the .mdp file that I used for the run. For the rerun, I changed 
nstlist and nstenergy to 1. I also set tcoupl and pcoupl to "no".

Tell me what you think about this. I am puzzled, especially about the 
behavior of 3.3.1 with confin.gro . Also point (5) above hints to the 
fact that LJ interactions between images are handeled wrongly by both 
3.3.0_patch and 3.3.1.

Cheers,
Michel

------------------------------------------------------------------------
title                    = Na
cpp                      = /lib/cpp
include                  = -I../top
integrator               = md
dt                       = 0.001
nsteps                   = 100000
comm_mode               = Linear
nstcomm                 = 1
comm_grps               = System
 
nstxout                  = 10000
nstvout                  = 0
nstlog                   = 1000
nstenergy                = 1000
nstxtcout                = 1000
xtc-grps                 = System
 
nstlist                  = 5
ns_type                  = grid
rlist                    = 0.8
coulombtype              = reaction-field
rcoulomb                 = 1.4
epsilon_rf               = 80
vdw-type                 = cut-off
rvdw                     = 1.4
 
tcoupl                   = nose-hoover
tc-grps                  = System
tau_t                    = 0.15
ref_t                    = 300
Pcoupl                   = parrinello-rahman
tau_p                    = 0.5
compressibility          = 4.5e-5
ref_p                    = 1.0
 
gen_vel                  = no
gen_temp                 = 300
gen_seed                 = 173529
------------------------------------------------------------------------

>Dear David and Berk,
>
>I am investigating further to understand why I don't get the same 
>energies with -rerun. I will report next week.
>
>Thanks for your replies,
>Michel
>
>  
>
>>>
>>>I just checked the 3.3.1 source code in md.c and the box is copied
>>    
>>
>>from the xtc file.
>  
>
>>>I don't know where your problems could come from.
>>>
>>>Berk.
>>>
>>>
>>>  
>>>
>>    
>>
>>>>>From: "David Mobley" <dmobley at gmail.com>
>>>>>Reply-To: Discussion list for GROMACS users <gmx-users at gromacs.org>
>>>>>To: "Discussion list for GROMACS users" <gmx-users at gromacs.org>
>>>>>Subject: Re: [gmx-users] mdrun -rerun and box size bug not fixed?
>>>>>Date: Thu, 1 Nov 2007 06:12:56 -0700
>>>>>
>>>>>Dear Michel,
>>>>>
>>>>>Are you sure you're using 3.3.1? I have not checked this specific
>>>>>issue -- but I have checked that if I run hydration free energy
>>>>>calculations in patched 3.3.0, and again in patched 3.3.1, and do my
>>>>>usual free energy analysis using mdrun -rerun, I get the same values.
>>>>>That either tells me that (a) the patch is in 3.3.1 also, or possibly
>>>>>(b) that the problem is small enough for hydration free energies that
>>>>>I'm missing it.
>>>>>
>>>>>Anyway, hopefully (a) is true, otherwise I'm going to need to repeat a
>>>>>LOT of work (and be very grumpy). Keep me posted on this.
>>>>>
>>>>>Thanks,
>>>>>David
>>>

-- 

==========================================================
Michel Cuendet, Ph.D
Molecular Modeling Group
Swiss Institute of Bioinformatics
CH-1015 Lausanne, Switzerland
www.isb-sib.ch/groups/Molecular_Modeling.htm
==========================================================


-------------- next part --------------
A non-text attachment was scrubbed...
Name: LJ-rerun.eps
Type: application/postscript
Size: 23173 bytes
Desc: not available
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20071106/ddf856c7/attachment.eps>