4
$\begingroup$

For classical Molecular Dynamics (cMD) simulations, the usual methodology is to use force fields (FFs) with fixed point charges (examples include CHARMM, AMBER and Gromos). However, this approach means that you are unable to capture the effect of local electrostatics (since fixed point charges cannot respond to the changes in the environment). There are multiple literature reports (including ours) which has shown that accurate capture of local electrostatic interactions are necessary for biological systems and surface interactions.

To include local electrostatics into the simulations, we have two methods:

  1. Use a polarizable FF description: This is the methodology that is followed in classical Drude polarizable FF description in CHARMM, where a massless Drude particle is attached to the heavy atom, and the polarizability is modified based on the vibration of the Drude particle). This approach leads to better capture of electrostatic interactions, even though the charges on the heavy atom and the Drude particle are not changing over time.
  2. The second approach is called charge equilibration (Qeq) where you start with a set of initial charges, and the charges on each atom are recomputed every nth step to account for the changes in the local environment. This approach is usually followed in the case of reactive FFs; but can in-principle be used in any case.

As far as my understanding goes, both Drude and Qeq comes with significant performance hits when compared with non-polarizable additive FFs. For Drude, this performance hits comes from the increase in the number of particles, while for Qeq, it is the recalculation of the charges that is expensive.

In the context of these, what are the advantages or disadvantages that one approach has over the other?

$\endgroup$

1 Answer 1

4
$\begingroup$

The three main polarization models are Qeq, Drude and induced dipole. The Qeq is a charge-polarization, while the other two are dipole-polarizations. The Drude and ID are essentially equivalent, and parameters can be moved from one model to the other. The advantage of Drude is that it can be run in existing codes written for fixed-charge models. All polarization models are 5-10 times more expensive than fixed-charge models, as either the charge parameters must be iterated to self-consistency in each timestep, or treated as fictive small-mass particles in an extended Lagrange formulation, which requires smaller timesteps. Qeq comes with a lot a other acronyms, FQ, ENN, ..., but it has two non-physical artifacts: charge-transfer over infinite distances and cubic scaling of the polarizability with (non-interacting) system size. There are a number of other charge-polarization models (also called charge-flow or charge-flux) that do not have these artifacts, but very few programs (if any) have implemented (and parameterized) these in application type simulation programs. The different charge-polarization models can be shown to be mathematically equivalent, under certain conditions, but their parameterizations differs. JCTC 19, 4047 Unifying Charge-Flow Polarization Models will give you the full details.

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .