[gmx-users] problem in cluster determination : g_clustsize and trjconv -pbc cluster problem

[Sanku M]

Hi,
  I have run a simulation using gromacs 4.0.7 with 50 copies of amphiphilic surfactant ( in water) using Martini coarse-grained force-field. Mere visualization of trajectories shows that starting with 50 discrete copies, they gradually aggregate and finally two well-separated large micelles are formed. But, my problem starts in quantifying the cluster-size-distributions . 
Here is what I tried:

1. Initially, I was trying to directly use g_clustsize tools to calculate the cluster-size distribution. Here is the command line I used :
   g_clustsize -f traj.xtc -s topol.tpr -mol 

But, It turned out that for all the time-frame, it gives me only single  cluster having all the molecules , giving a flat line in the time-evolution of cluster-size. 
This looks to be wrong as the starting configuration has all the 50 copies of surfectatnt well dispersed( at least from visualization) and at the end of the simulation the visualization shows two clusters.  Here, since  I am interested in the molecules that is part of a cluster , hence I used -mol option.  Without this option also, I get almost all the atoms in a single cluster .

Then I tried  
  trjconv  -pbc nojump -f traj.xtc -o traj_nojump.xtc 
 option to take care of periodicity. But, It does not change the result g_clustsize_4mpi is giving: it always gives me only 1 cluster. use of -pbc whole also gives me same result.

2. I also tried to play with cut-off in g_clustsize . Initially I was using default cut-off value ( 0.35 nm) . I tried smaller value but result does not change. The most awkward part is that even if I make cut-off distance 0 as follows:
   g_clustsize -f traj.xtc -cut 0.0 -s topol -mol 
it still gives me a single cluster !  How can it be possible ? I am pretty sure there is something wrong but not able to figure this out. 

3.  Later, I found that in gromacs, there is a description of how to determine micelle clusters ( perhaps by Justin Lemkul ? )
   Here is the steps it was recommended : 
  	1. trjconv -f a.xtc -o a_cluster.gro -e 0.001 -pbc cluster
	2. grompp -f a.mdp -c a_cluster.gro -o a_cluster.tpr
	3. trjconv -f a.xtc -o a_cluster.xtc -s a_cluster.tpr -pbc nojump
But, when I tried the first step  as follows:
    trjconv -f traj.xtc -o traj_cluster.gro -dump 0 -pbc cluster,
  it looks like it does not converge and enter an infinite loop: Here is an excerpt of the output:
COM:    0.000     0.000    19.999  iter = 19526  Isq = 12671033.000
COM:    0.000     0.000     0.000  iter = 19527  Isq = 12628369.000
COM:    0.000     0.000    19.999  iter = 19528  Isq = 12671033.000
COM:    0.000     0.000     0.000  iter = 19529  Isq = 12628369.000
COM:    0.000     0.000    19.999  iter = 19530  Isq = 12671033.000
COM:    0.000     0.000     0.000  iter = 19531  Isq = 12628369.000
COM:    0.000     0.000    19.999  iter = 19532  Isq = 12671033.000
COM:    0.000     0.000     0.000  iter = 19533  Isq = 12628369.000
COM:    0.000     0.000    19.999  iter = 19534  Isq = 12671033.000
COM:    0.000     0.000     0.000  iter = 19535  Isq = 12628369.000
COM:    0.000     0.000    19.999  iter = 19536  Isq = 12671033.000
COM:    0.000     0.000     0.000  iter = 19537  Isq = 12628369.000
COM:    0.000     0.000    19.999  iter = 19538  Isq = 12671033.000

I am not sure why I am getting an infinite loop ( looks like lsq is only fluctuating between two numbers). I also saw similar kind of report in the gromacs mailing list , but no proper solution came to my sight .



Overall, none of the attempts is giving me an appreciable result? 
Can someone help me on this so as to get  right cluster size distribution ? Any help will be appreciated.

Thanks
Sanku


      
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20100705/8e089bc7/attachment.html>