J. C. Dobrowolsk et al.
104
5.4.3.3 Combined Molecular Dynamics-Quantum Approach (MD/QM)
For systems containing multiple conformations with unknown equilibrium structures and/or for systems with hydrogen-bonded structures between solute and solvent molecules, the hybrid model is needed. there are three ways to use these methods:
1. the simplest method is the supermolecular calculations for solvent molecules
from the first solvation shell combined with PCm model for the remaining part
of the solvent. It is frequently referred as hybrid method.
2. the majority of classical molecular dynamics (md) or Carr-Parrinello-molecular dynamics (CP-md) for modelling chiroptical properties in solution
employ this approach to generate solute-solvent clusters for supermolecular
calculations. here, the first step is to find the number of solvent molecules
hydrogen-bonded to the solute molecule. this can be determined from the analysis of atom-atom radial distribution functions (RdF) or from the generation
of solute-solvent clusters as md “snapshots” [66, 137]. the second step is the
conformational search for the cluster and the simulation of a spectrum. moreover, explicit supermolecular microsolvation calculations can be combined with
the implicit PCm method.
3. the most satisfactory and more rigorous approach to account for solvation is to
perform dynamic simulations, i.e., by employing md or CP-md directly and
by simulating the spectra at each time step of the md trajectories (“snapshots”
corresponding to, for instance, 10 ns of simulation). this accounts for the inhomogeneous band broadening and Boltzmann averaging of many conformations
provided by the conformer distribution [137–141].
At this point, let us mention an important problem. When the geometry of the solute-solvent complexes is optimised, the transformation of the hessian to normal
coordinates in the frame of the harmonic approximation is correct and the energy
gradient with respect to geometric coordinates is zero. however, using “snapshots”
leads to a number of imaginary frequencies and thus introduces errors and artifacts
in the simulated spectra. A solution to this problem was proposed by Bouř and Keiderling [142] using the scheme of separating the low- and high-frequency regions. At
present, such simulations are still very demanding and the investigation of solvent
effects on chiroptical properties, using this technique, is limited.
Concluding this part of the review, we will restrict our discussion to the implicit and explicit solvent methods that actually are used to model solute-solvent
interactions in vCd and/or RoA spectroscopy for model systems. there are examples showing that PCm models, which despite predicted reasonable energies and
geometries [127], do not sufficiently reproduce the vibrational frequencies [143].
Nowadays, most of the vCd and RoA spectra calculations are carried out for biomolecules. modelling of the hydrogen-bond interactions between solute and solvent molecules requires explicit solvent molecules included in the supermolecular
scheme to model the first hydration shell, and then use the PCm approach for the
remaining part of the solvent. A good example of the hybrid model can be found
104
5.4.3.3 Combined Molecular Dynamics-Quantum Approach (MD/QM)
For systems containing multiple conformations with unknown equilibrium structures and/or for systems with hydrogen-bonded structures between solute and solvent molecules, the hybrid model is needed. there are three ways to use these methods:
1. the simplest method is the supermolecular calculations for solvent molecules
from the first solvation shell combined with PCm model for the remaining part
of the solvent. It is frequently referred as hybrid method.
2. the majority of classical molecular dynamics (md) or Carr-Parrinello-molecular dynamics (CP-md) for modelling chiroptical properties in solution
employ this approach to generate solute-solvent clusters for supermolecular
calculations. here, the first step is to find the number of solvent molecules
hydrogen-bonded to the solute molecule. this can be determined from the analysis of atom-atom radial distribution functions (RdF) or from the generation
of solute-solvent clusters as md “snapshots” [66, 137]. the second step is the
conformational search for the cluster and the simulation of a spectrum. moreover, explicit supermolecular microsolvation calculations can be combined with
the implicit PCm method.
3. the most satisfactory and more rigorous approach to account for solvation is to
perform dynamic simulations, i.e., by employing md or CP-md directly and
by simulating the spectra at each time step of the md trajectories (“snapshots”
corresponding to, for instance, 10 ns of simulation). this accounts for the inhomogeneous band broadening and Boltzmann averaging of many conformations
provided by the conformer distribution [137–141].
At this point, let us mention an important problem. When the geometry of the solute-solvent complexes is optimised, the transformation of the hessian to normal
coordinates in the frame of the harmonic approximation is correct and the energy
gradient with respect to geometric coordinates is zero. however, using “snapshots”
leads to a number of imaginary frequencies and thus introduces errors and artifacts
in the simulated spectra. A solution to this problem was proposed by Bouř and Keiderling [142] using the scheme of separating the low- and high-frequency regions. At
present, such simulations are still very demanding and the investigation of solvent
effects on chiroptical properties, using this technique, is limited.
Concluding this part of the review, we will restrict our discussion to the implicit and explicit solvent methods that actually are used to model solute-solvent
interactions in vCd and/or RoA spectroscopy for model systems. there are examples showing that PCm models, which despite predicted reasonable energies and
geometries [127], do not sufficiently reproduce the vibrational frequencies [143].
Nowadays, most of the vCd and RoA spectra calculations are carried out for biomolecules. modelling of the hydrogen-bond interactions between solute and solvent molecules requires explicit solvent molecules included in the supermolecular
scheme to model the first hydration shell, and then use the PCm approach for the
remaining part of the solvent. A good example of the hybrid model can be found