[gmx-users] zero temperature bug revisited when using frozen groups

[ying-hua-chung at uiowa.edu]

I want to run an nve simulation with 2 subsystems. Both subsystems are water.
Some of the water molecules I want frozen and the other water molecules I want 
to run NVE with. My simulation box is divided into two parts: An inside sphere 
and an outside part that occupies the rest of the simulation box.
I want sometimes to freeze the molecules outside the sphere and let the rest 
of the system inside the sphere run NVE or vise versa to keep what is inside 
of the sphere frozen and run NVE on the rest of the simulation box.

My question is about zero temperature while I freeze part of my system in nve
process.

In my system there are 224795 atoms (21903 atoms are FIX, 2892 atoms are SOL)

SOL is defined as the molecules inside the sphere, FIX is defined as the 
molecules outside the sphere

I have velocities in my intial .gro file and my nve.mdp file reads:

energygrps	    = FIX SOL
freezegrps          = FIX
freezedim           = Y Y Y
gen_vel             =  no

In this case I freezed FIX and did not freeze SOL and I got zero temperature 
in the gromacs output (BTW, the molecules do move but the computation of the 
temperature is somehow wrong)  

But if I unfreeze FIX and freeze SOL
 
freezegrps          = SOL
I get a temperature that fluctuates between 315~325K

I read the message by Berk Hess on Mon Apr 5, 2004
and followed the suggestion but I still got zero temperature.

I first tried using 

1. "comm_grps System" in .mdp

and then

2. changed src/kernel/readir.c 
as indicated in the newgroup not finding any difference. Still the temperature 
of the total system was 0.

Could you please offer any ideas? Is there a bug in the computation of the 
temperature when a group of molecules is frozen?
Thanks!!!

YingHua