Parameterization Strategies

we have detailed how to construct a molecular PES as a sum of energies from chemically intuitive functional forms that depend on internal coordinates and on atomic (and possibly bond-specific) properties. However, we have not paid much attention to the individual parameters appearing in those functional forms (force constants, equilibrium coordinate values, phase angles, etc.) other than pointing out the relationship of many of them to certain spectroscopically measurable quantities. Let us now look more closely at the Art and Science of the parameterization process. 

In an abstract sense, parameterization can be a very well-defined process. The goal is to develop a model that reproduces experimental measurements to as high a degree as possible. Thus, step 1 of parameterization is to assemble the experimental data. For molecular mechanics, these data consist of structural data, energetic data, and, possibly, data on molecular electric moments. We will discuss the issues associated with each kind of datum further below, but for the moment let us proceed abstractly. We next need to define a penalty function , that is, a function that provides a measure of how much deviation there is between our predicted values and our experimental values. Our goal will then be to select force-field parameters that minimize the penalty function. Choice of a penalty function is necessarily completely arbitrary. One example of such a function is 

what steps can be taken to decrease the scope of the problem One approach is to make certain parameters that depend on more than one atom themselves functions of single-atom-specific parameters. For instance, for use in Eq. (2.16), one usually defines 

Yet another way to minimize die number of parameters required is to adopt a so-called united-atom (UA) model. That is, instead of defining only atoms as the fundamental units 

Another approach that is conceptually similar is to make certain constants depend on bond order or bond hybridization. Thus, for instance, in the VALBOND force field, angle bending energies at metal atoms are computed from orbital properties of the metal-ligand bonds in the MM2 and MM3 force fields, stretching force constants, equilibrium bond lengths, and two-fold torsional terms depend on computed n bond orders between atoms. 

Such additions to the force field somewhat strain the limits of a classical model, since references to orbitals or computed bond orders necessarily introduce quantum mechanical aspects to the calculation. There is, of course, nothing wrong with moving the model in this direction - aesthetics and accuracy are orthogonal concepts - but such QM enhancements add to model complexity and increase the computational cost. 

---

Because of the interelectronic repulsion term l/ri2, the electronic Hamiltonian is not separable and only approximate solution of the wave equation can be considered. The obvious strategy would be to use Hj wave functions in a variation analysis. Unfortunately, these are not known in functional form and are available only as tables. A successful parameterization, first proposed by James and Coolidge [89] and still the most successful procedure, consists of expressing the Hamiltonian operator in terms of the four elliptical coordinates 1j2 and 771 >2 of the two electrons and the variable p = 2ri2/rab. The elliptical coordinates 4> 1 and 2, as in the case of Hj, do not enter into the ground-state wave function. The starting wave function for the lowest state was therefore taken in the power-series form... 

The electronic coupling of donor and acceptor sites, connected via a t-stack, can either be treated by carrying out a calculation on the complete system or by employing a divide-and-conquer (DC) strategy. With the Hartree-Fock (HF) method or a method based on density functional theory (DFT), full treatment of a d-a system is feasible for relatively small systems. Whereas such calculations can be performed for models consisting of up to about ten WCPs, they are essentially inaccessible even for dimers when one attempts to combine them with MD simulations. Semiempirical quantum chemical methods require considerably less effort than HF or DFT methods also, one can afford application to larger models. However, standard semiempirical methods, e.g., AMI or PM3, considerably underestimate the electronic couplings between r-stacked donor and acceptor sites and, therefore, a special parameterization has to be invoked (see below). 

In the sequential strategy, a control (manipulated) variable profile is discretized over a time interval. The discretized control profile can be represented as a piecewise constant, a piecewise linear, or a piecewise polynomial function. The parameters in such functions and the length of time subinterval become decision variables in optimization problem. This strategy is also referred to a control vector parameterization (CVP). 

In fact, the distinction between two-step and direct dynamics is rather fuzzy. The basic issue is what kind and amount of preliminary work is needed before starting a dynamical calculation. Direct ab initio dynamics [90,97-101] requires a minimum of preparation some tests to choose basis sets and other options may suffice. For large systems, however, fully ab initio calculations are impractical, and one has to resort to QM/MM or PCM approaches but then, a host of empirical parameters are introduced, which may need some readjustement to avoid artefacts and to improve the accuracy before starting the dynamical calculations. The same holds for the semiempirical methods in order to represent at best the excited states, one has to re-parameterize the hamiltonian. In particular, our FOMO-SCF-CI method [56-58] differs considerably from the normal SCF or SCF+CIS procedures, so that the standard parameters need to be modified. However, the parameter sets are fairly transferable, and their optimization can be limited to the atoms belonging to the chromophore. In the two-step strategies one fits the ab... 

We have adopted a strategy similar to that used in our development of polymer force fields in parameterization of an atomistic potential function for HMX. Specifically, we have undertaken a systematic investigation of conformational and intermolecular binding energies in model nitramine compounds (i.e., those containing the C2N-NO2 moiety ) using high-level QC calculations. In the case of HMX, a QC-based force field is the only realistic option due to insufficient spectroscopic data that would facilitate force field parameterization. 

The electrostatic solvation energy is only a part of the total solvation energy. Cavitation, dispersion and repulsion terms must be added. We show below that the MPE method leads to similar electrostatic energies than the polarizable continuum model (PCM) of Tomasi and co-workers [10], provided the same cavities are used. Therefore, non-electrostatic terms in these methods may be computed using the same computational strategies [15]. We emphasize the fact that accurate non-electrostatic contributions are often difficult to compute since they are based on parameterized formulae that cannot be directly compared to experiment. The obtained data must therefore be used with prudence, especially if they are expected to play a major role in the process under study. Fortunately, in many circumstances, non-electrostatic terms are small and/or vary little, so that they can be neglected. Tunon et al. [80] developed a parameterized expression for the MPE method using an expression of the type... 

This review will focus on theoretical calculations to advance understanding of gas phase oxidafion of gaseous elemental mercury (GEM) by halogen species. Computational and experimental studies to help parameterize models have been performed to make a more reliable description of the dynamics of mercury in the atmosphere so that the consequences of abatement strategies can be assessed. Quantum chemical calculations are the only way to viably investigate the mechanisms and advance what is observed in field and laboratory studies. 

The accuracy to which these points are treated determines the transferability of the TB scheme. For low-level TB schemes, at most only the qualitative trends should be trusted. A pitfall in this context is overly parameterized TB schemes, i.e. too many fitted input parameters — this strategy will tend to hide inherent limitations. In principle, carefully worked-out TB schemes can achieve almost the accuracy of self-consistent LCAO-DFT ° without use of empirical fitting parameters. 

The modeling procedure can be sketched as follows. First an approximate description of the velocity distribution in the turbulent boundary layer is required. The universal velocity profile called the Law of the wall is normally used. The local shear stress in the boundary layer is expressed in terms of the shear stress at the wall. From this relation a dimensionless velocity profile is derived. Secondly, a similar strategy can be used for heat and species mass relating the local boundary layer fluxes to the corresponding wall fluxes. From these relations dimensionless profiles for temperature and species concentration are derived. At this point the concentration and temperature distributions are not known. Therefore, based on the similarity hypothesis we assume that the functional form of the dimensionless fluxes are similar, so the heat and species concentration fluxes can be expressed in terms of the momentum transport coefficients and velocity scales. Finally, a comparison of the resulting boundary layer fluxes with the definitions of the heat and mass transfer coefficients, indiates that parameterizations for the engineering transfer coefficients can be put up in terms of the appropriate dimensionless groups. 

The abatement strategy for reduction of N emission and deposition as well as decreasing of undesirable losses from agroecosystems due to excessive application of fertilizers has to be based on the calculations of its regional biogeochemi-cal budgets with quantitative parameterization of different fluxes in terrestrial and aquatic ecosystems. 

Parameterization and Tracking of Optimization of Synthesis Strategy Using Computer Spreadsheet Algorithms... 

---