Chemistry Reference
In-Depth Information
was a critical and somewhat surprising result, leading the authors to suggest
that including polarizability in the model effectively results in more screening
taking place between ions, thereby enabling the ions to move past each other
with greater ease than when polarizability is absent. This is a reasonable con-
clusion but has not been investigated further by any other groups. It is an issue
that is certain to be investigated in the near future.
The transverse current correlation function was integrated by Voth and
co-workers
63
to compute the shear viscosity. Viscosities of 4.74 cP were calcu-
lated for the polarizable model, while the fixed-charge model gave a value of
6.84 cP. Both results are consistent with the experimental value of 4.42 cP.
Again, the fixed-charge model results in slower dynamics than does the polar-
izable model, and the polarizable model agrees better with the experiments
than does the fixed-charge model. This study provides convincing evidence
that the inclusion of polarizability will yield faster dynamics in a particular
model, and many authors have since commented on this. The problem is
that polarizable models are much more demanding computationally than
are fixed-charge models. It is desirable to determine if polarizable force fields
are absolutely necessary for such simulations, especially given that the
transport properties for other liquids can be computed adequately without
resorting to polarizable models.
At this point, a comment is warranted on how Voth and co-workers
63
computed the viscosity because it has relevance to the other studies described
below. As discussed earlier, there is an implicit assumption in the use of equi-
librium fluctuation formulas when computing transport coefficients. It is
assumed that the time scale over which the transport coefficient is evaluated
is longer than the correlation times of the quantity being evaluated. In the
case of the work by Voth and co-workers, this was verified for the self-
diffusivity, where the slope of the mean-square displacement was taken over
1 ns. For the shear viscosity, however, integrals of the transverse current
correlation function were taken over time scales ranging from 1 to 10 ps.
The reason such short times were used is that the correlation function decays
to zero very rapidly, and so long-time integration was numerically impossible.
This does not mean, however, that the time scale relevant for viscosity is only
1-10 ps. We know that the viscosity of a liquid depends on many orientational
and rotational relaxation processes, all of which occur on time scales orders of
magnitude longer than 10 ps, calling into question exactly what is being com-
puted from integrals over such a short time. Is it the ''global'' shear viscosity?
Or, might it instead reflect some type of ''local'' apparent shear viscosity for
molecules? This is not to criticize the work of Voth and co-workers; indeed, as
shown below, several other studies have computed the viscosity in the same
way. Rather, this issue is raised for the novice modeler because it needs further
exploration by the modeling community.
Bhargava and Balasubramanian
103
used equilibrium MD to compute the
self-diffusivity, shear viscosity, and electrical conductivity for [C
1
mim][Cl] at
