@Carlos, thanks for your advice and GLUP. And I checked the list of GULP potential models. I did not find S&M potential what I need.

@Axel @Steve, thanks for your attention. In S&M paper, to get charge, eq. 12 needs be solved. In it, two items-- \chi and \v are variables as in equations 9’ and 10’. Top-left paragraph gives a concise and not clear explanation (Expansion of eq.10 contains a lone item --1/r_ij and how to get it-- standard Ewald techniques). In other words, this potential achieves that charge can be calculated by atom configuration. \chi, \v are main variables related to atom configuration and is not related to charge. Expansion of \chi and \v can be found in J. Phys. Chem. C 2007, 111. 11776-11783 Effect of Particle Size on Melting of Aluminum at Nano Scales eq.16. Expansion of \sum[f_i|f_j] in 10’ equals 1/r_ij plus v_ij. Here 1/r_ij is.

How to get 1/r_ij by standard Ewald techniques?

Meihai

Meihai,

@Carlos, thanks for your advice and GLUP. And I checked the list of GULP

potential models. I did not find S&M potential what I need.

Are you sure? The S&M potential is not listed on the potentials model

page but on the "Libraries of potentials" capabilities. The library

might be an external feature though. The 1994 paper you cite is also

included in the list of GULP references. I am not a GULP user but why

would the GULP team list these features if they were not to be

implemented. Did you try to ask GULP users or developers? Might be

worth going the extra mile here as it has the potential to save you

tons of headaches and TIME.

Carlos

@Axel @Steve, thanks for your attention. In S&M paper, to get charge, eq.

12 needs be solved. In it, two items-- \chi and \v are variables as in

equations 9' and 10'. Top-left paragraph gives a concise and not clear

explanation (Expansion of eq.10 contains a lone item --1/r_ij and how

i don't have access to the specific paper you are referring to.

i looked at: http://prb.aps.org/pdf/PRB/v50/i16/p11996_1

in the end, for as long as you insist that you are

doing the right thing, then please be my guest and

prove me wrong by implementing the potential, that

you are after, the way you want. having open access

to the source code includes the freedom to shoot

yourself in the foot any which way you like.

[...]

How to get 1/r_ij by standard Ewald techniques?

the same way you'd do (q_i*q_j)/r_ij for q_i = q_j = 1.0 ??

ewald summation is discussed in too many text books

on MD and publications to list them all here.

as was mentioned before, if you want to implement

something, simply looking for a cut-n-paste solution

without understanding the underlying principles is a

recipe for doom.

axel.