[gmx-users] RE: Energy conservation in collision

[Janne Hirvi]

>>>>Hello gmx-users!
>>>>
>>>>I am still struggling with the energy conservation in the system where a
>>>>water droplet (consisting of rigid molecules) collides with a frozen
>>>>structured surface. The total energy is well conserved at the beginning of
>>>>NVE simulation when the droplet is approaching the surface and once again 
>>>>when the droplet is stabilized on the surface, but the drift or fluctuation
>>>>in the total energy is observed between these extremes.
>>>>
>>>>The total energy of the system (~300 000kJ/mol) decreases slightly
>>>>(<100kJ/mol) at the collision and on non-wetting surfaces (water droplet
>>>>wont intrude to the pores of the surface) part of the energy reverts
quickly
>>>>causing a sharp peak in the energy curve. This could indicate an
>>>>integration error at the collision due to the large time step, but equal
>>>>behavior is observed with the time steps of 2.0fs and 0.5fs.
>>>>
>>>>On the wetting surfaces (water droplet intrude to the pores of the surface)
>>>>the center of mass motion of the droplet come to a stop at the collision
>>>>as in the case of non-wetting surface and sharp decrease in the total
energy
>>>>is observed. However, now none of this energy is reverted and the total 
>>>>energy stabilize to this value until the water molecules starts to intrude
>>>>to the pores of the surface. The total energy decrease slowly in conjuction
>>>>with the slow sinking of the height of the center of mass of the water
>>>>droplet until the bottom of the pores is reached. After the bottom of the
>>>>pores is reached energy will stabilize to the corresponding value which is
>>>>about 200-300kJ/mol lower than atthe beginning.
>>>>
>>>>Decreasing the time step from 2.0fs to 0.5fs have no influence for the
>>>>energy conservation and somehow the problem seems to be connected to
>>>>water-surface interactions: on the wetting surfaces (more water-surface
>>>>interactions) a drift is observed instead of a sharp peak of non-wetting
>>>>surfaces (less water-surface interactions). Any ideas how I could achieve
>>>>complete energy conservation?
>>>>
>>>>Thanks for your time and help,
>>>>
>>>>Janne
>>>
>>>What kind of interactions are you using?
>>>Plain cut-off for Coulomb and LJ will give bad energy conservation.
>>>For your system you probably want to use shifted LJ and Coulomb potentials.
>>>
>>>Berk.
>>
>>I am using PME with slab correction for Coulombic interactions and plain 
>>cut-off for Lennard-Jones interactions. Neighbor lists are updated every time
>>step due to the use of plain cut-off for LJ interactions. This actually gives
>>really good energy conservation at the beginning of the simulation, when the
>>droplet is approaching the surface and once again when the droplet is
>>stabilized on the surface, but the problems exist between these "stable
>>situations".
> 
>You don't specify how you made your surface,
>but I assume it would be with LJ particles.
>Then you have jumps in the energy with a cut-off, since the LJ potentials
>is not zero at the cut-off. This will result in lower energies at contact.
> 
>Use a shifted potential, in that way you can also set nstlist to something 
>like 10 which will speed up your simulations.
> 
>Berk.

Thanks Berk!

The total energy seems to be conserved at the collision when a shifted or
switch
function is used for LJ interactions. Comparison simulations without an impact
velocity in NVT ensemble used a pure cut-off for LJ interactions but I assume
that this small change in the "shape of potential function" won't affect the
physics of the system but I still have to check it... 

Janne

PS. I have some problems with my email and hope that this isn't a duplicate
message...