Direct methods polynomial approximation

Several methods are used to calculate these coefficients, such as the finite difference approximation, the direct method and its modifications, the Green-function method, and the method of polynomial approximations. 

A true FEA procedure was implemented for the DFT problem in 1988 and shortly thereafter for the HF problem. Triangular elements and up to sixth-order approximating polynomials in the p, v variables (called there s, t) are employed. A logarithmic scaling of element size with distance is used to provide higher accuracy with fewer points. The cutoff is based on distance. The linear equations are solved by a direct method. The largest molecules for which HF results have been reported are CO and N2. 

Starting with the method described above, extensive tables of the numerical values of mean energies, integrals of electrostatic and constant of spin-orbit interactions are presented in [137] for the ground and a large number of excited configurations, for atoms of boron up to nobelium and their positive ions. They are obtained by approximation of the corresponding Hartree-Fock values by polynomials (21.20) and (21.22). Such data can be directly utilized for the calculation of spectral characteristics of the above-mentioned elements or they can serve as the initial parameters for semi-empirical calculations [138]. 

There exist (4, 5, 8, 9, 27) simple direct relations, between isotope effect, structure, and force field, which do not necessarily require a complete knowledge of all molecular parameters and avoid the solution of the secular equation. These relations are, however, approximations restricted to limited ranges of temperature. [Newer approximation methods, based on expansions in Jacobi polynomials, are applicable over wide ranges of temperatures (6, i6).] In the past, before the ready availability of fast digital computers, tests of the validity of these approximations were usually fairly limited in nature, but recent extensive tests on model calculations of kinetic isotope effects have been carried out 23, 28). In addition, extensive tests of power-series approximations (not considered in the present paper) have now been performed (6,16). 

Finite difference methods (FDM) are directly derived from the space time grid. Focusing on the space domain (horizontal lines in Fig. 6.6), the spatial differentials are replaced by discrete difference quotients based on interpolation polynomials. Using the dimensionless formulation of the balance equations (Eq. 6.107), the convection term at a grid point j (Fig. 6.6) can be approximated by assuming, for example, the linear polynomial. 

Note that we obtain the same result (using polynomial of the 2nd degree (4.499)) from an equivalent set of independent reactions, say (4.49). This may be seen directly from (4.503), (4.504) inserting (4.500) then 3, are equilibrium constants of reactions (4.49) respectively. We also note that we can also eliminate Cj, C3 then we must use the preceding method for polynomial of the type (4.484) but in (cp and the result (4.503) may be obtained again. Therefore, chemical kinetics in the system O, O2,03 may be described by two equilibrium and six rate constants when constitutive equations for reaction rates are approximated by a polynomial of the second order (a polynomial of the third order gives 20 rate constants [79] equilibrium constants are again two because of two independent chemical reactions). 

The shift parameter can be used to ensure that the optimization proceeds downhill even if the Hessian has negative eigenvalues. In addition, it can be chosen such that the step size is lower or equal to a predefined threshold. Popular methods using a shift parameter are the rational function optimization (RFO) [48] and Trust Radius (TR) methods [49, 50]. A finer control on the step size and direction can be achieved using an approximate line search method, which attempts to fit a polynomial function to the energies and gradients of the best previous points [51]. 

---