I am having some difficulty fully understanding the cutoff in the COMB potential. The doc page says:
thus for a zero-charge pure element system with no van der Waals interaction, the COMB potential reduces to Tersoff potential, typically truncated at a short cutoff, e.g. 3 to 4 Angstroms. The long-range Coulombic term uses the Wolf summation method described in Wolf, spherically truncated at a longer cutoff, e.g. 12 Angstroms.
If I do not have a zero-charge pure element system, what is the cutoff for non-Coulombic forces?
Additionally, can the cutoff’s be changed for the comb potential, as they are coded in LAMMPS?
For instance, can I change the non-Coulomb cutoff and/or the cutoff used int he Wolf Summation method?
Best Regards,
Ben
I am having some difficulty fully understanding the cutoff in the COMB
potential. The doc page says:
thus for a zero-charge pure element system with no van der Waals
interaction, the COMB potential reduces to Tersoff potential, typically
truncated at a short cutoff, e.g. 3 to 4 Angstroms. The long-range
Coulombic term uses the Wolf summation method described in Wolf
<http://www.qft.iqfr.csic.es/docs/lammps/pair_comb.html#Wolf>,
spherically truncated at a longer cutoff, e.g. 12 Angstroms.
If I do not have a zero-charge pure element system, what is the cutoff for
non-Coulombic forces?
All non-Coulombic terms have cutoffs around 3-4 Angstroms.
Additionally, can the cutoff's be changed for the comb potential, as they
are coded in LAMMPS?
They are read in from the potential file ffield.comb. Chaning cutoffs
require refitting of the potential.
For instance, can I change the non-Coulomb cutoff and/or the cutoff used
int he Wolf Summation method?
Not recommended, as you require refitting.
Ray
Thank you for your prompt response. A follow-up:
For the ReaxFF potential, in the qeq/reax, they refer to a Taper radius cutoff. The default cutoff is 10. If I am trying to determine the computational efficiency on my machine of both potentials (COMB10 and ReaxFF) and want to compare them, do I need to set this Taper radius to 12 to be comparable to COMB10? I have tried searching for Taper radius and have not found anything yet, so I am not even fully sure what it is that I would be changing.
I am not sure if it is proper to compare two potentials with different cutoffs and thought I would perhaps need to either change the COMB10 cutoff to 10 or change ReaxFF Taper radius to 12. Though my understanding of Taper radius is lacking, so there is a good chance I am completely off base.
Ben
You shouldn’t change any cutoffs in any of the potentials. There is no problem in comparaing the efficiency of different potentials with different cutoffs. See benchmark (http://lammps.sandia.gov/bench.html#potentials).
Ray
When you say “any” of the potentials, I am confused. Isn’t it find to say, change the cutoff of the BKS potential? It might run-longer but I can still get good results.
Ben
Perhaps I should clarify, not just simply BKS. But with buck/coul/long, I put in parameters for BKS and can change the cutoff, depending on how accurate I want the results to be.
Ben
Perhaps I should clarify, not just simply BKS. But with buck/coul/long, I
put in parameters for BKS and can change the cutoff, depending on how
accurate I want the results to be.
This does not sound right at all. What do you mean by "depending on how
accurate I want the results to be"? There can only be one result - the
result that the original authors published with their original parameters.
With the published cutoff and published parameters, you get the published
results (most of the time). If you change any of the cutoffs or
parameters, you are not using their potentials anymore - instead you are
using something you invented. Once you use your own, you are responsible
for your own results.
If you want to use "published" potentials, you should always stick with
published cutoffs/parameters. Without fitting experience, you should
always not think about changing cutoffs/parameters. This is not a LAMMPS
question anymore.
Ray
I understand, but the BKS potential did not publish any cutoff values, they just published the coefficients for the Buckingham potential. Other people on this forum have suggested to me in the past changing the cutoff values to see how they effect results and efficiency. I have never seen a “right” cutoff for the BKS potential and I have extensively searched for one.
I know you have used the BKS potential before. Where did you find the right cutoff values?
Ben
The cutoff that reproduces published results. For potentials that do publish cutoffs, you should not change them.
Ray