I am looking for a way to study the interaction of a single mobile ion with a fixed charged lattice (using Lennard Jones and long range electrostatic interactions using ewald). From the simulations I calculate the potential of mean force and compare the results with theoretically calculated values.
I realize its hard to stabilize the temperature using a single ion for the simulation!
So I tried using multiple ions in the same system and I shut off interactions between these ions using the "neigh_modify exclude" command.
This ignores the short range interactions but I am not sure if the interactions between these ions at long range is excluded?
Is there a way to work around that, as in shut off the long range interactions between the ions as well, so that in effect I can have multiple copies of the single ion system and the average temperature remains close to the desired value?
I am looking for a way to study the interaction of a single mobile ion with
a fixed charged lattice (using Lennard Jones and long range electrostatic
interactions using ewald). From the simulations I calculate the potential of
mean force and compare the results with theoretically calculated values.
I realize its hard to stabilize the temperature using a single ion for the
simulation!
So I tried using multiple ions in the same system and I shut off
interactions between these ions using the "neigh_modify exclude" command.
This ignores the short range interactions but I am not sure if the
interactions between these ions at long range is excluded?
it is not.
Is there a way to work around that, as in shut off the long range
interactions between the ions as well, so that in effect I can have multiple
copies of the single ion system and the average temperature remains close to
the desired value?
the only approach i can think of, is to have multiple
concurrent *complete* simulations using partitions.
that will require more compute power, but is at least
embarrassingly parallel.
Hi
Thanks for the quick reply!
However I am not sure if I understand what is meant by : " multiple concurrent complete simulations using partitions."
I apologize for my ignorance, if its a commonly used term among lammps users.
I have a two dimensional charged lattice with the mobile ions and a layer of background charge (sufficiently far away from the charged lattice) to neutralize the charge of the system. I use slab geometry for the 2D ewald.
Can you please direct me to the place where I can find out more about using partitions to run multiple concurrent simulations!
Thanks steve. I shall take a look at this section.
However for my problem, I tried a different approach.
I tried running a simulation of the same system (multiple mobile ions interacting with a charged lattice, with the excess charge appropriately neutralized).
What I do differently now is I have only one of the mobile ions charged!
I used the neigh_modify command in this case too, to exclude interactions between the mobile ions. I thought this would eliminate electrostatic interactions among mobile ions since only one of the mobile ions is now charged.
However the electrostatic energy values that I get from these runs is a little misleading. Now, as far as I understand, since only one of the mobile ions is charged, the total number of mobile ions in the system must not affect the total electrostatic energy of the system. However when I run simulations with different number of mobile ions, the electrostatic energy increases with an increase in the number of mobile ions.
Thanks for the reply Josh.
However in the system that you have in the link, if I understand right, the discrepancy arises from using both “special_bonds” and “neigh_modify exclude” with long range electrostatics.
I do not have any bonds (or special_bonds) in my system, so there is no subtraction term that interferes with the long range electrostatic calculations!
Special bonds are a separate issue. If I understand your Q,
you are hoping to turn off all interactions between 2 charged
particles by excluding that type pair via neigh_modify exclude.
That will work for the short-range part, but not for
the long-range part. The PPPM solver
have no conception of interactions between pairs of particles.
Maybe it would be possible in Ewald. I'm including Stan
in this CC - maybe he has comments on this question.
In theory, it is possible to exclude interactions between 2 charged
particles in both Ewald and PPPM, as is done in compute group/group.
However, this option is not currently available in the regular PPPM or
Ewald kspace solvers, and all charged particles will interact. The
only options right now that I can see are to use particles with zero
charge (which turns off all coulombic interactions) or to add a
correction term afterwards that cancels the force and energy due to
the excluded pair interaction. For example, the kspace part includes
spurious self-interactions which must be removed, and you could remove
other interactions in a similar manner.
And maybe the "neigh_modify exclude" command already does this
correction--I am not sure. Paul would know better. One thing that
you have to remember though is that when you use Ewald/PPPM, not only
are you interacting all particles, but also all of their periodic
images out to infinity. So even if you correct for the interaction of
the particles in the closest periodic image, you still are interacting
in all other periodic images, unless you modify the Ewald/PPPM code as
was done for compute group/group.
Neigh_modify exclude doesn't do anything other
than remove the interaction from the neighbor list,
i.e. the short range piece is not included. But the long-range
piece (infinite images) that Stan refers to will still
be computed in Ewald or PPPM. What kind of model
are you envisioning where somehow 2 charged particles
have no interaction?
Thanks Steve, Stan
Please refer to the first email in the thread where I wrote about the system I am looking at.
Interaction of a single positively charged ion with a negatively charged fixed square lattice. (using Lennard Jones and long range electrostatic interactions using ewald)
I found a way to work around it by using multiple mobile ions to stabilize the temperature with only one of the ions charged! ( this shuts down electrostatic interactions between the ions)
However what is confusing me is, the electrostatic energies that I get as a result of these simulations is dependent on the number of mobile ions.
One would think that the electrostatic energy should be an exact match as long as the single charged mobile ion is at the same position.
But I find it changes as the number of uncharged mobile ions are changed. Am i missing something here?
Ewald uses the total number of atoms (including uncharged) to estimate the g_ewald parameter that divvies up the calculation between real space and k-space. The total number of atoms also figures in the accuracy estimates, and the number of k-space vectors that are used. So you’ll see slight variations in your total electrostatic energy (elong + ecoul) when changing the number of uncharged mobile ions. You should be able to minimize these variations by cranking up the accuracy (use a smaller number, like 1e-12 or so).