[gmx-users] problem with total energy

[Mark Abraham]

On 9/04/2010 1:25 PM, jampani srinivas wrote:
> Dear Justin,
>
> I am sorry for the poor description of the problem, OK let me explain
> you clearly here.
>
> I have taken a decapeptide and solvated it with box size 6.0 nm, I want
> to create a frozen wall (confined sphere) around protein after a certain
> radius (in this case it is 2.5 nm). To achieve this i made two
> temperature coupling groups, first (Tmp1) one has protein and waters
> within 2.5nm from the center of the box and rest is second temperature
> coupling group (Tmp2). Initially when i run the simulation it was
> creating nrdf = 0 for both groups, Berk has helped me with a file
> readir.c file, i compiled my gromacs again that problem was solved. I
> have submitted simulations again and found that the energy was blowing
> up. I think you know the story after this. Earlier I want protein to
> interact with inner wall of the frozen group and check what happens,
> because of this I have never looked at the energygrps_excl option. I
> have done with both 2 and 4 fs time steps and I took 4fs option to speed
> up the simulation, and I have to still look at the paper you suggested me.
>
> Thanks for your suggestions, I have implemented your suggestion one
> after the other, finally when i use the energygrps_excl option it worked
> out. Now there is no sudden drift in the energy, and also i checked the
> velocities of non-frozen group is mostly zero (except for first frame)
>
> When I use the energygrps_excel option in the following way i am getting
> the below mentioned note i have to still understand what is this
> message. On the other hand if i use only "energygrp_excl  = Tmp2 Tmp2
> Tmp2 Tmp1"  line in mdp file i am getting fatal error. Is it necessary
> to to define the energy groups first and later exclude the energy option?

Yes, energy group exclusions require matching energy group definitions. 
The relevant part of manual 7.3 probably says this.

> +++++++++++++++++++++++++++++
>
> energygrps     = Tmp1 Tmp2
> energygrp_excl  = Tmp2 Tmp2 Tmp2 Tmp1
>
> ++++++++++++++++++++++++++++++++
>
> NOTE
>
> "Can not exclude the lattice Coulomb energy between energy groups"

You seem to be trying to simulate a droplet inside a rigid shell of 
water. If so, why do you want the system to be periodic? You don't want 
periodicity artefacts *and* frozen-water-shell artefacts. There are 
other electrostatics models better suited than PME to such situations. 
You should look in the literature for successful published 
methodologies, rather than risk inventing a square wheel yourself before 
you've got enough experience to be confident with what you're doing.

Here, the reciprocal-space part of the PME calculation must contain 
contributions from your excluded energy-groups, so you've got some 
Frankenstein possible worst-of-all-worlds combination. If I got such a 
paper to review, I'd need to see some serious groundwork justifying such 
a choice.

> I used -maxwarn option here and generated the tpr file. I hope this does
> not harm the simulation.

That's a risky strategy. The programmer didn't put in such a message 
just to see whether he could do it...

Mark