Choosing the correct Ewald sum parameters

I'm currently working with a system where the long range Coulomb interactions are described with the Ewald summation method.

I am using DL_POLY Classic. The Ewald summation method used is described in pages 46-48 in the user manual which I unfortunately can't link because of my low ranking.

I'm working with two papers which use the same computational methods. Again, I can't link anything but you can find the papers in arxiv. The names are

B. Guillot, N. Sator; Carbon dioxide in silicate melts: A molecular dynamics simulation study

B. Guillot, N. Sator; A computer simulation study of natural silicate melts. Part I: Low pressure properties

These papers describe the parameters as follows:

The long range coulombic interactions are accounted for by a Ewald sum with a constant $\alpha L= 5-7$ (where $\alpha$ is the width of the charge distribution on each ion and $L$ is the size of the simulation box) and cut off distance $(r_{cut})$ of 10 - 11 Å, the summation in the reciprocal space being evaluated for all $k$ vectors with $|k|L/2\pi<6-7$.

DL_POLY requires 4 values for ewald-summation, $\alpha~,k_{x},k_{y}$ and $k_{z}$, where (according to the user manual) $\alpha$ is the convergence parameter and $k_{x,y,z}$ are the maximum k-vector indexes in x,y or z-direction.

So now I would need to find out the correct convergence parameter and the values for $k_{x,y,z}$. Can I get these values straight from the data supplied in the paper, and if yes, how?

• Welcome to SciComp.SE. Can you write your equations and provide the reference please? – nicoguaro Feb 25 '16 at 16:12
• I have much more experience with DL_POLY 4 (being a contributor to the code) which is a successor to DL_POLY classic, but doesn't the classic code provide "sensible" defaults for the Ewald summation? Is there any reason you can't use those? – Ian Bush Mar 30 '16 at 10:31

I have limited experience with Ewald, but I found this thesis to be helpful when working with it. I do not know if it contains the answer to your question, but it might contain useful references.

http://kth.diva-portal.org/smash/record.jsf?pid=diva2%3A458575&dswid=9118

I found out the problem. Turns out it wasn't in the Ewald summation parameters. The paper that I'm trying to figure out uses the word "equilibrate", so I assumed (incorrectly) that the writers had used the "equilibration n" -command in DL_POLY. It seems that in this particular case using the equilibration command messes up the simulation and causes it to form structures which are not possible.

I have much more experience with DL_POLY 4 (being a contributor to the code) which is a successor to DL_POLY classic, but you can get "sensible" values for the Ewald parameters by use of the procedure outlined in section 3.2.5.1 "Ewald sum and SPME" on page 89 of the manual where it suggests including a

ewald precision 1d-6