I am working on a polarizable water model with polarizability (we got this after fitting). Now my
question is how to select appropriate magnitude of core-shell charges because there are
infinitely many ways to select these charges and spring constant which will satisfy
alpha = q^2 / k
Is there any upper and lower limit for these parameters? I have tested different core-shell
charges ranging form 6e to 12e with corresponding k values in DLPOLY classic.
And strangely it has a strong effect on structural properties like RDF.
And what values on the core-shell will reproduce point dipole results?
Any guidance on this issue.
Best
Ommair
Attachments area
Dear Ommair,
As you state in your letter, you are working on polarisable model, and use Drude approach to model the polarizability. Developing polarisable force field is a complex task that requires good understanding of electrostatic interaction in realistic molecules and extensive fitting based both on good intuition and good luck. For water there is a carefully developed by a MacKerell’s group in Maryland model, SMW4_NDP, so if your target is a polarisable water model you can use that one and read about the model on a Wiki page: https://en.wikipedia.org/wiki/Drude_particle where there is a reference to their water model with a link. If you want to learn about the theory behind selecting charges, read their paper https://doi.org/10.1021/ct049930p .
Your last question about reproducing point dipole result is unclear. The Drude oscillator model is richer that the point dipole, the latter is just a limiting case of infinite harmonic constant. You probably wanted to ask what charges will fit the experimental polarizability of water? – Use SMW4_NDP values, they were fitted to an effective polarizability of water, i.e., its ‘mean’ condensed phase value at normal conditions.
Hope this helps.
Vlad
Thanks,
I understand what you said, so it means if we use values of drude charges and spring constant without fitting it will affect properties of water. Please correct me if I am wrong.
Please see this paper Journal of Molecular Liquids 188 (2013) 245–251 in which they have used GROMACS which also have a core-shell model for induction. Their study shows that values of drude charges should not affect properties of water. But when I used different values of drude charges in DLPOLY classic, and calculated RDF, results were quite different. I am worried if this behavior is correct or not.
Dear Ommair,
so it means if we use values of drude charges and spring constant without fitting it will affect properties of water. Please correct me if I am wrong
That’s correct. It is also supported by the reference you provided, where in the abstract the authors state: “However, we show in this paper that the point-dipole approximation condition is actually not satisfied for a range of Drude oscillator parameters; that leads to a significant dependency of the obtained results from the particular choice of the Drude charge/spring constant”.
The reason for this dependence is that changing the value of Drude charge changes the quadrupole moment and therefore orientational correlations in water. You can look, if interested, at the details of our earlier model of water based on quantum version of Drude charges, https://doi.org/10.1103/PhysRevLett.110.227801 (especially, the scaling of dispersion and polarisation, which was possible in the quantum version). You can also think along the lines of charge equilibration process – another way of treating the polarisation (polarisable charges), where the principle of electronegativity equalisation of Sanderson is used (Google for it and for ‘hardness’). One needs not only fitting the first moment, but although the higher moments (well, at least the second) to describe not only the structure of solid (at zero kelvin) but the fluctuations at finite temperature, important for liquids.
For water, there are literally hundreds of models both polarisable and non-polarisable, and this fact just show how difficult the problem is. Successful fitting is a delicate balance of many factors, and I would put it into the domain of art rather than science.
Good luck with your model,
Vlad