[gmx-users] (no subject)

[Mark Abraham]

Sunil Thapa wrote:
> Respectable David/Justin/Mark/Berk
> In my study of diffusion of oxygen in water, I solved the segmentation 
> fault problem as below
> 1. At first I ran for 16 ps beyond which segmentation fault occured in 
> the previous run. I generated the .gro file of that run
> 2. I then ran for 100 ps using the .gro file of step 1. Also generated 
> the .gro file from this step. I constantly checked the pressure content. 
> It really decreased from 4000's to 1400s. No segmentation fault occurred.

Pressure can't be measured instantaneously, only as a time average. Many 
picoseconds can be required to observe the equilibrium value.

> 3 I than ran again for 500 ps using the same technique.
> 4. When I ran for 1 ns, there was no segmentation fault as well as the 
> pressure was about 1 bar which I needed. The temperature was 298 K as I 
> needed.
>  
> Now the run upto 100 ns went without any segmentation fault.
> Is this the way I should solve the problem?

According to the above, you haven't solved a problem - it just hasn't 
happened again.

> Now when I saw the msd curve, it was nearly straight line and the 
> calculation yielded the diffusion coefficient of oxygen as 40000 x 
> 10(-9), where as the experimental value is nearly 2 x 10(-9).
> Actually, as the time increases it increases, Why is this happening?
> We are doubtful about the force field parameters we have supplied.
> Specially, the nonbonded parameters.

OK, we aren't magically experts on the model you're using. You need to 
read up on what the authors parametrized it to be able to do, and under 
what conditions.

This is also a good lesson on checking on the results of your simulation 
while it's still running - rather than waiting until the end to find out 
you'd made some mistake.

> I have tried to define bare oxygen in .atp file
> Is not that the charge of bare oxygen is zero?

The charge on atomic or molecular oxygen atoms will be necessarily zero 
- but it isn't set in the .atp file. See Chapter 5.

> Could you tell me what is the value for nonbonded parameters for oxygen 
> molecule and spc water molecule?
> OB--------OW
> OB---------HW

No - that's your job before you even get near a simulation package :-)

> In the nb.itp the nonbonded parameter (Linard-Jonnes) between oxygen and 
> water hydrogen is given as zero.
> I am stuck at this step and I need to solve it anyway for my Msc. 
> Thesis. Please help me.

Preparation, planning, and good supervision are all important!

Mark