Abstract
A fully defined and molecular-size agnostic Gibb’s free energy function that uses strictly structural parameters as input would permit real-time energetics determination and optimization of molecules heretofore intractable at ab initio levels. Here we present, the first part of a linear function for Gibbs free energy (GiFE Function) that covers the elements {H,N,O,C,F} and is molecular-size agnostic, using only atomic structure to generate the input variables. Critically, the GiFE function is capable not only of producing the value of Gibbs free energy for a given complex in constant time, but also may serve as a function over which molecules may be optimized in O(sqrt{n}) time. The prediction of individual and reaction free energies are demonstrated, as well as explanations of chemical understanding generated from the function and an outlining of how the rest of this function may be constructed, with examples covering {H,B,C,N,O,F,S,Si,Cl,Br,I}.
Supplementary weblinks
Title
ChemArchItect
Description
A program library for machine learning data generation of chemical systems. These programs were used in the construction and analysis of the dataset.
Actions
View