Dynamic QM/MM: A Hybrid Approach to Simulating Gas–Liquid Interactions - Multiscale Molecular Methods in Applied Chemistry - page 45

are issues concerning the partitioning of charges between QM or MM regions that
keep being refined.
One major difference in the charges in the QM and MM regions is that the MM
charges typically do not change during the simulation as they are predefined in the
potential force field. The development of QM/MM methods in which the charges
from one region dynamically affect those in the other, as would be the case in a true
system, is an active aspect of research [ 18 - 20 ]. This approach is referred to as
electrical embedding, as opposed to the simple mechanical embedding where the
charges are fixed. Evidently, mechanical embedding should be used when charge
transfer or long range polarization effects are minimal for the chemical system
being studied.
Only a small fraction of the QM/MM studies have explicitly considered molecular
dynamics calculations involving both QM and MM regions, as the computational
time associated with a QMmolecular dynamics calculation is quite demanding. Some
examples where this has been done are covered in [ 21 - 26 ]. Recently, our group has
developed a simulation model to specifically study reactions at the gas-liquid
interface that merges classical MM for the entire system with a QM treatment of
the atoms in the reaction region(s) at each time step to form a hybrid QM/MM-MD
approach. This new computational approach extends the direct dynamics models our
group has used to study gas phase collision chemistry previously [ 27 ], and now
includes MM calculations for the liquid that is not involved in reactive events. Our
primary target liquids have been squalane (for which there are many experiments)
and ionic liquids (an unconventional fuel of current interest), and in most cases we
have been interested in studying the collision of reactive atoms such as atomic
oxygen (O) and fluorine (F) with the liquid surfaces in order to simulate gas-surface
molecular beam experiments. In addition, since most of these experiments refer to
reactive atoms that have several eV of energy (hyperthermal energies), our appli-
cations have been concerned with highly nonthermal processes in which reaction
mechanisms that do not normally contribute to thermal kinetics are described.
Projects concerning the O + squalane [ 28 ] and F squalane [ 29 ] reaction

dynamics have been completed, and they provide detailed information about the
spatial distribution of reactive sites and the correlation of the reaction mechanism
with the angular and translational distribution of the scattered products. These
results have been used to give a detailed mechanistic understanding of beam-liquid
surface experiments in the Minton lab [ 30 , 31 ]. In some cases we were able to
identify products not initially considered in the experimental studies, providing
stimulus for subsequent molecular beam measurements where they were seen [ 32 ].
Our F

þ

squalane studies [ 29 ] have been used to interpret experiments done by
Nesbitt and coworkers at thermal collision energies [ 33 - 35 ], showing that there is a
component of the HF product vibrational and rotational distributions which
involves escape of HF from the surface with little relaxation. In more recent work
we have considered the collisions of atomic oxygen (and also hyperthermal argon)
with the surfaces of ionic liquids [ 36 , 37 ], providing detailed information about
a class of chemical reactions that has not been considered previously, and enabling
the interpretation of experiments by Minton, McKendrick, and collaborators.

þ