I am currently simulating the oxidation of silicon nanowires using ReaxFF with Parameters by Buehler et al. (Phys. Rev. Lett. 96, 2006) in LAMMPS. My first approach is to replicate calculations by Khalilov et al. (Nanoscale 5, 2013) carried out using the Reax Code. When comparing my data to the data obtained by Khalilov et al. I noticed a difference that is most likely related to charges. Charge equalization in LAMMPS is realized via QEq while in the Reax Code it is realized via EEM. In the original QEq paper by Rappe and Goddard (J. Phys. Chem. 95, 1991) it is mentioned that the charges obtained via EEM are a factor 3-6 small in comparison to charges obtained via QEq. In the original ReaxFF paper by van Duin et al. (J. Phys. Chem. A 105, 2001) it is argued that QEq and EEM are very similar and that the main difference between the two can be compensated by adjusting a shielding parameter in the Coulomb Term of ReaxFF. The question I have is whether and how the difference between QEq and EEM was considered when ReaxFF was implemented in LAMMPS.
I am still wondering about the different charge equilibration methods in LAMMPS and the Reax Code. The main difference between my results and the results of Khalilov et al. is long range Coulomb repulsion between oxygen molecules approaching a silicon nanowire and oxygen atoms adsorbed on the wire. This repulsion is happening in my simulation because of the global, instantaneous charge redistribution done by QEq. The results of Khalilov et al. do not seem to show the repulsion even though EEM also works globally and instantaneously.
Does anybody of you know how one can explain this difference?
Zitat von Georg Heinze <[email protected]>:
The original QEq proposed by Rappe and Goddard uses the slater type orbital to describe charge-charge interactions (resulted from orbital overlap), but ReaxFF uses the slightly modified approach that describes charge-charge interactions via a shielded Coulomb (the EEM approach). The EEM approach for ReaxFF is performed via fix qeq/reax.
If you want to use Slater type orbital for charge-charge interactions, there is a fix qeq/slater that you can use. You might need to find the necessary parameters from the Literature or even “guess” them. Please see http://lammps.sandia.gov/doc/fix_qeq.html for more info.
thank you for your reply. So what you are saying is that fix qeq/reax is actually not QEq but EEM, just as in the Reax Code, correct?
Also I found an old correspondence between you and Suleiman Oloriegbe. There a tapering function for the coulombic interaction term was mentioned. Is this function controlled by the tapering parameter of fix qeq/reax and am I correct in assuming that changing this term will not change the charges in my system but rather taper the coulombic interaction between the charged atoms for long distances?
Finally, do you know whether this is done the same way in LAMMPS as in the Reax Code?
Zitat von Ray Shan <[email protected]>:
Yes, strictly speaking fix qeq/reax implements the EEM method described by van Duin et. Al. Yes, it is the same as Adri van Duin’s standalone Reax code.
Tapering function approximates the charge-charge interaction near the cutoff to avoid discontinuities. It is parameterized to approximate the results from the Ewald Sum. Any changes to the Coulombic interactions will somehow affect the equilibrium charges since charge equilibration, albeit QEQ or EEM, equalizes the derivatives of charge-depenedent terms wrt charge for all the atoms. Changing the energy expression affects the derivatives hence the equilibrium charges.
Yes, LAMMPS’ implementation of ReaxFF is exactly the same as van Duin’s standalone reax code.