(145e) Property Prediction with Analogous Series: A Continuum View of Discrete Molecular Entities

Property
Prediction with Analogous Series: A Continuum View of Discrete Molecular
Entities

Brian
K. Peterson

ExxonMobil
Research & Engineering

Depending
on the data available and other constraints, the best choice of a method for
predicting the properties of materials might include the use of theory,
molecular simulation, quantum chemistry, various
correlative methods, or combinations of several of these. When faced with
sparse data and limited time, the data-based correlative method of Analogous
Series offers surprising ease and accuracy if data of a certain structure is
available. The method is introduced by several examples and its principles are
explained with reference to a hypothetical continuum of molecular species which
is intuitive for those with experience in statistical mechanical theory and
molecular simulations. We demonstrate that given small amounts of data on two
families of materials related to the material of interest, the method can be
used to predict a wide variety of thermophysical and other properties and is
especially useful for detecting outliers and anomalous data. We have so far
applied the method to predicting critical and boiling temperatures, infinite
dilution activity coefficients, GC retention times, closed-cup flash points, heat
capacities, adsorption in zeolites, and kinetic rate constants.