One goal of theoretical chemistry is to give quantitative estimates of chemical properties of unknown compounds. In spite of the unprecedented availability of computer resources solving the electronic structure of all possible species without a systematic guidance is phohibitive.Novel techniques based on chemical reactivity indexes in the framework of Density Functional Theory have been developed to search and design new compounds with desirable properties. Changes in electron density for constant external potential have been widely used to evaluate many properties intrinsic to the nature of the electronic structure. Thus, searching new compounds is possible by rationalizing and mimicking chemical groups studied previously.In analogy to changes in electron number, ``alchemical'' changes and derivatives, i.e. changes of energies due to structural and stoichiometric variations, have been proposed for the prediction of chemical properties of new molecular species, and more generally to rationally navigate chemical compound space. In this work we present development and application of such prediction models for atomization energies of unknown molecules resulting from transmutation of a single known reference molecule. Other electronic properties are also investigated. The performance of our models is assessed for a set of two thousand isoelectronic species with 54 electrons that were chosen from a pool of stable organic molecules with 7 heavy atoms (C,O,N). We will also discuss the effect of molecular alignment, addition of reference molecules, and usage of molecular descriptors.

Predicting molecular properties with alchemical derivatives

245th National Meeting (2013)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.