Design of Molecular Products with Optimal Properties

The competitive edge and market share of many chemical industries manufacturing polymers, refrigerants, solvents and surfactants are ultimately intertwined with the identification of "new" and "better" products. With the rapid growth in optimization theory, algorithm development and high-performance computing, exciting and unprecedented research opportunities are emerging in molecular design to assist in this endeavor. Key research objectives include the identification of the "best" molecular design, as described by the employed structure-property prediction model, with mathematical certainty eliminating the caveat of convergence to suboptimal molecular designs. By removing the possibility of unknowingly identifying suboptimal molecular designs, possible discrepancies between obtained optimal and experimentally verified designs can unequivocally be attributed to uncertainty (imprecision) in property estimation. By describing this uncertainty within a probabilistic framework, answers to questions regarding the likelihood of meeting various design objectives are answered. Research work has focused on the following categories.

Polymers

This research addresses the problem of identifying the polymer repeat unit architecture so that a performance objective that is a function of mechanical, electrical and/or physicochemical properties is optimized. Originally group contribution methods were employed to derive structure-property relations containing only partial information regarding connectivity. Lately, the use of topological indices is being considered to provide full connectivity information. In addition, a systematic method for quantitatively assessing the effect of property prediction uncertainty (i.e., imprecision) in optimal molecular design is introduced. Property prediction uncertainty is explicitly modeled using multivariate normal distribution for the group contribution parameters. Deterministic equivalent formulations are derived based on chance-constraint programming concepts. The developed framework is customized for the design of polymers with high lithium ion conductivity and low water absorption for battery applications.

Refrigerants

Here, the optimal refrigerant selection and simultaneous refrigeration cycle synthesis is pursued which satisfies a single or multiple process cooling duties at different temperatures. For the pure refrigerant case, a superstructure representation is explored encompassing many features from existing multistage refrigeration cascades such as multiple refrigerants and stages, presaturators and economizers. For the mixed refrigerant case the vertical cascade, proposed for pure refrigerants, is generalized to include horizontal stages arising from partial condensation that generates streams of different compositions and thus temperatures. The proposed methodology is customized for LNG problems.

Surfactants

A two-tier modeling and optimization framework has been proposed for the design of surfactants that match a set of performance metrics quantifying their suitability for applications such as cleaning, textile processing, and agrochemical formulations. In the inner stage, free energy models are employed to identify the equilibrium values for geometric (i.e. shape) and size micellar parameters as well as thermodynamic properties such as critical micelle concentration (CMC), aggregation number and surface tension. In the outer stage, the surfactant performance is optimized with respect to structural descriptors such as tail and head sizes as well as dipole moment and separation. So far efforts have concentrated on the design of nonionic oxyethylene surfactants, however, the ionic as well as surfactant mixtures are of interest.

[Lehmann and Maranas (2004); Camarda et al (1999); Camarda and Maranas (1999); Vaidyaraman and Maranas (1999); Raman and Maranas (1998); Maranas (1997b); Maranas (1997a); Maranas (1996)]