[gmx-users] T-coupling ...

[Senthil Kandasamy]

Hi all, 

I had posted an earlier message about the strange T-coupling behavior.
After analyzing the energy files, I realize that the culprit is water.
Eventhough the lipid, protein and water are independently coupled at
310K, the water for some reason seems to heat up to ~650K. Thus, the
system temperature fluctuates at about ~650 K or so. This was using a
tau-t of 0.1ps and berendsen coupling. Then I played around with
different coupling constants and these are the results

Run	Tcoupl  	Ref-t	Tau-t	T-water	 RMSD	T-System RMSD
1	Berendsen	310	0.1	785	 79	668	60
2	Berendsen	310	0.01	348	 6.2	338	4.6
3	Berendsen	310	0.005	329	3.83	324	2.87
4	nose-hoover	310	0.005	310	15	310	11

All these systems had a gen_vel =yes at 310K.
They are all averages over atleast 300ps (300 frames)

The nose-hoover seems to work alright, but still the magnitude of the
fluctuations bother me. 

Now, coming back to the possible reasons for this weird error. Now that
I think about it, the gromacs on this machine was compiled using pgcc.
Could that make a real difference?  I think during compilation, there
were some warnings, but I ignored them. Could some subroutines be badly
compiled and thus lead to the weird behavior? 

Any input will be greatly appreciated. I am currently working with the
system administrators to get gromacs compiled with gcc to see if it
would make any difference.

Senthil