CCL: Wisdom on mixing (combination) rules in MD simulations



Hi,

I'm wondering if someone can help me gain some wisdom about mixing (combination) rules in MD simulations.  The classical simple point charge models for water -- SPC and SPC/E -- seem to be widely used to the present day in MD simulations of aqueous systems.  At the same time, it seems that there are two widely used sets of mixing (combination) rules for calculating Lennard-Jones (LJ) interaction parameters for the interaction of different atomic species.  If I understand correctly, these two most common sets are:

(1) Lorentz-Berthelot mixing rules: arithmetic for sigma [ sigma_ij = 1 / 2 * (sigma_i + sigma_j) ], geometric for epsilon [ epsilon_ij = \sqrt{epsilon_i * epsilon_j} ]
(2) Geometric mixing rules: geometric for sigma [ sigma_ij = \sqrt{sigma_i * sigma_j} ], geometric for epsilon [ epsilon_ij = \sqrt{epsilon_i * epsilon_j} ]

and the Lennard-Jones (LJ) interaction I am referring to is

U_LJ(r_ij) = 4 epsilon_ij * [(sigma_ij / r_ij)^12 - (sigma_ij / r_ij)^6].

Many of the papers reporting MD simulations of SPC-only or SPC/E-only systems (mostly from the 1980s, 1990s, and 2000s) state that they used the Lorentz-Berthelot mixing rules (1).  However, in the case of SPC or SPC/E water, epsilon for the hydrogens is 0, so (1) and (2) will give the same result.

However, the situation becomes less straightforward when one introduces, for example, Na and Cl ions modeled by the OPLS-AA force field.  OPLS-AA uses mixing rules (2) above: geometric combination for both sigma and epsilon, as described in the second page of the Jorgensen group's classic paper (Jorgensen et al., JACS 1996, 118, 45, 11225--11236): https://doi.org/10.1021/ja9621760

So, my question is, if I have an OPLS-modeled solute in SPC or SPC/E water, do I use mixing rules (1) or (2) above?

One possible lead is this paper from the Frenkel group (Espinosa et al., J. Chem. Phys. 2018, 149, 224501): https://doi.org/10.1063/1.5054056  In that paper, they state, "As OPLS/AA has been parameterized for being used in combination with SPC, TIP3P, and TIP4P water models, we have chosen these models for simulating water despite them being less accurate than other models in reproducing the behavior of pure water."  Sure enough, when I look at what appears to be the original OPLS paper -- albeit united-atom, not all-atom -- Jorgensen et al. state, "The potential functions have the simple Coulomb plus Lennard-Jones form and are compatible with the widely used models for water, TIP4P, TIP3P, and SPC" (Jorgensen & Tirado-Rives, JACS 1988, 110, 1657--1666: https://doi.org/10.1021/ja00214a001 )

But I'm still a little confused... does the discussion above mean it is legitimate to use mixing rules (2) when simulating an OPLS solute in SPC or SPC/E water?  As far as I know, most people, and most MD codes, use either mixing rules (1) or mixing rules (2) -- not a "mixture" of mixing rules (1) and (2), so to speak.

Does anyone have wisdom on this?  Thank you very much for your time!

Andrew DeYoung, PhD
Department of Chemistry
Carnegie Mellon University