Dynamic QM/MM: A Hybrid Approach to Simulating Gas–Liquid Interactions - Multiscale Molecular Methods in Applied Chemistry - page 46

Our motivation for doing these studies is that the structure and dynamics of
chemical interfaces at phase boundaries (e.g., gas-liquid interface) is of significant
importance to chemistry and only recently has the capability been established to
describe these processes (both with modeling and with experiments) at an atomistic
level. Newly developed experimental methods provide the capability to study the
reactivity of a liquid surface with beam-surface scattering, which provides detailed
mechanistic information, but it is often difficult to interpret these experiments
without theoretical modeling. Among the experiments being developed is work
by Nesbitt et al. [ 33 - 35 ] who have resolved quantum-state reaction dynamics via
direct absorption detection of the rovibrational states of the nascent gaseous
products coming from the surface using high resolution infrared spectroscopy.
Related technology has been developed by Nathanson [ 38 , 39 ] and McKendrick
[ 40 , 41 ], who have used laser-induced fluorescence methods to detect the
nascent products in gas-liquid surface reaction dynamics studies. In addition,
Minton and coworkers [ 31 , 32 ] have performed molecular beam experiments
involving hyperthermal oxygen scattering from liquid surfaces, in which a mass
spectrometer is used to determine the angular and translational energy distributions
of the nascent gaseous products. These experiments, as well as others, require
theoretical modeling that includes the flexibility to compute the dynamics of the
reactive events (and sometimes several sequential reactive events), while at the
same time the simulation needs to include a considerable portion of the liquid
(thousands of atoms) so that energy transfer, diffusion, and electrostatic interactions
can be described. Also, the hydrocarbon and ionic liquids are often nanostructured,
which means that a simulation region of sufficient size is needed in order to capture
all of the dynamical complexity that can occur.
2 Description of the Method
2.1 Basic QM/MM
Since most large-scale chemical processes can generally be partitioned into reactive
and nonreactive regions, dividing up the system to be treated with different levels of
sophistication in the theoretical model is logical. As stated above, the atoms in the
reactive parts of the model are treated by direct dynamics QM (electronic structure)
computations, while the remaining nonreactive part is treated with an empirical
MM potential. Our dynamic QM/MM code utilizes various subroutines from
TINKER 4.2 [ 42 ] to compute the MM part of the simulation (forces and potential
of the MM atoms). For the QM part of the code, subroutines from the MSINDO 2.1
electronic structure code were used [ 43 ], which includes the MSINDO (modified
symmetrically orthogonalized intermediate neglect of differential overlap) semi-
empirical Hamiltonian to compute the energy [ 44 , 45 ]. In general, semiempirical
methods are a middle ground between a fully ab initio QM calculation and fully