General least squares, molecular

An experiment using thermobalance yields a record of the cell mass with sample as a function of time. From the slope of this curve, the rate of mass loss is determined. The inert gas pressure in the furnace is read on a manometer, and is lowered stepwise in the course of the experiment. Thus a set of corresponding values for n2 and P, which may be considered as knowns in Eq. (7.15) is obtained. On the other hand, the parameters A, B, C, P, and y are generally unknown, since the molecular mass of the vapor, which is included in the parameter y, is not known a priori. The problem may be handled by means of a suitable, non-linear least-squares analysis computer program, which fits Eq. (7.16) for the observed set of data. 

Due to the effects of molecular size and shape and pore structure on the kinetics, the model cannot be used for general predictive purposes. In practice, in order to predict PAC adsorption, a series of experiments must first be carried out using the compound of interest, the activated carbon to be applied, and the water in which it is to be used. Equilibrium parameters, determined from the Freundlich adsorption isotherm equation, are used as input into a computer-based HSDM, which uses the method of least squares to minimize the difference between the experimental kinetic data points and the HSDM fit of the data [10]. When the best fit is achieved, the resultant kinetic parameters (liquid film mass transfer coefficient, k(, and the surface diffusion coefficient, DJ can then be used for the prediction of adsorption behavior under different conditions. 

Common chemometric tools may be applied to deal with similarity matrices. Particularly, partial least squares (PLS) [73,74] stands as an ideal technique for obtaining a generalized regression for modeling the association between the matrices X (descriptors) and Y (responses). In computational chemistry, its main use is to model the relationship between computed variables, which together characterize the structural variation of a set of N compounds and any property of interest measured on those N substances [75-77]. This variation of the molecular skeleton is condensed into the matrix X, whereas the analyzed properties are recorded into Y. In PLS, the matrix X is commonly built up from nonindependent data, as it usually has more columns than rows hence it is not called the independent matrix, but predictor or descriptor matrix. A good review, as well as its practical application in QSAR, is found in Ref. 78 and a detailed tutorial in Ref. 79. 

In general, there is an art and a science to molecular mechanics parameterization. On one extreme, least-squares methods can be used to optimize the parameters to best fit the available data set, and reviews on this topic are avail-able. Alternatively, parameters can be determined on a trial-and-error basis. The situation in either case is far from straightforward because the data usually available come from a variety of sources, are measured by different kinds of experiments in different units, and have relative importances that need subjective assessment. Therefore, straightforward applications of least-squares methods are not expected to give optimum results. 

Molecular structures may be described and compared in terms of external or internal coordinates. The question of which is to be preferred depends on the type of problem that is to be solved. For example, one problem that is much easier to solve in a Cartesian system is that of finding the principal inertial axes of a molecule indeed, if only internal coordinates are given then, in general, the first step is to convert them to Cartesian ones and then proceed as described in Section 1.2.4. Similarly, the optimal superposition of two or more similar molecules or molecular fragments, i.e. with the condition of least-squared sums of distances between all pairs of corresponding atoms, is best done in a Cartesian system. On the other hand, systematic trends in a collection of molecular structures and correlations among their structural parameters are more readily detectable in internal coordinates. 

Once least-squares methods came into general use it became standard practice to refine not only atomic positional parameters but also the anisotropic thermal parameters or displacement parameters (ADPs), as they are now called [22]. These quantities are calculated routinely for thousands of crystal structures each year, but they do not always get the attention they merit. It is true that much of the ADP information is of poor quality, but it is also true that ADPs from reasonably careful routine analyses based on modem point-by-point or area diffractometer measurements can yield physically significant information about atomic motions in solids. We may tend to think of crystal structures as static, but in reality the molecules undergo translational and rotational vibrations about their equilibrium positions and orientations, as well as internal motions. Cruickshank taught us in 1956 how analysis of ADPs can yield information about the molecular rigid-body motion [23], and many improvements and modifications have been introduced since then. In particular, various computer programs are available to estimate the amplitudes of simple postulated types of internal molecular motion e.g., torsion-... 

Molecular weights are determined by solving the equilibrium form of the Lamm equation, generally by nonlinear least-squares fitting of the data to a presumptive model. For a monomeric solute (single ideal species), eqn [6] is a typical model. For associating systems, eqn [6] is extended by terms for each species present in the model. A monomer- -mer system is described by the model... 

---