(83a) A New Group Contribution Method for Organosilicon Structures and Its Application in the Design of Electronics Cooling Fluids

Group contribution (GC) methods have proved to be indispensable in the design of novel chemical compounds. They allow efficient calculation of target property values for new product discovery. However, existing GC methods, such as in [1], lack predictive capabilities for silicon-based structures. As a result, organosilicon compounds are usually excluded from computer-aided molecular design applications, despite their wide use in commercial products.

In this work, we develop group contribution models for silicon-based compounds with a focus on functional group selection. We formulate an optimization problem to decompose each molecular structure into the smallest number of non-overlapping sub-molecular groups. Each resulting functional group is structurally simple but holds maximum information. Additionally, we propose a hierarchical tree structure to represent the multi-level relationships among different orders of functional groups. Higher-order functional groups in this tree are selected if and only if their corresponding lower-order constituents are selected. For each property data set, we select a model by minimizing an information criterion, thus preventing overfitting, reducing root mean square error over the training data, and increasing generalization. The black-box modeling tool ALAMO [2] is used to solve the constrained regression problem with an objective to minimize Bayesian Information Criterion (BIC). The resulting GC models are embedded in a CAMD framework [3] to generate organosilicon compounds to be used as electronics coolants.

[1] J. Marrero and R. Gani, Group-contribution based estimation of pure component properties, Fluid Phase Equilibria, 183–184, 183–208, 2001

[2] A. Cozad, N. V. Sahinidis and D. C. Miller, Learning surrogate models for simulation-based optimization, AIChE Journal, 60, 2211-2227, 2014.

[3] A. Samudra and N. V. Sahinidis, Optimization-based framework for computer-aided molecular design, AIChE Journal, 59:3686–3701, 2013