Using atomic multipoles and developing machine learning models to investigate transition states

• Grant number: 17/22741-3
• Support type: Scholarships in Brazil - Doctorate (Direct)
• Effective date (Start): February 01, 2018
• Effective date (End): February 02, 2023
• Field of knowledge: Physical Sciences and Mathematics - Chemistry - Physical-Chemistry
• Principal Investigator: Roy Edward Bruns
• Grantee: Leonardo José Duarte
• Home Institution: Instituto de Química (IQ). Universidade Estadual de Campinas (UNICAMP). Campinas , SP, Brazil
• Associated scholarship(s): 18/24844-7 - Using atomic polar tensors and QCT parameters to train a machine learning model and predict Hammett constants, BE.EP.DD

Abstract

The application of the Charge-Charge Transfer-Dipolar Polarization (CCTDP) model quantitatively reproduces the infrared fundamental intensities values of all molecules for which gas phase experimental data have been measured. As such it hás proved to be a powerful tool for extracting information from the dynamics of the electronic density of several classes of molecules during a vibrational movement. Empirical data show a high correlation between reactivity parameters or intermolecular interactions and the infrared intensities so that, the CCTDP model allows identification of which factors mainly contribute to the reactivity or stabilization of intermolecular interactions. Previous work of our group has already shown the application of the model with the stabilization energy of the hydrogen bond.In this project, we will investigate transition states of different reactions, analyzing their vibrational intensities, especially those having imaginary frequency values, since these normal coordinates simulate the trajectories of the atoms for the transition state. In this way, we are going to understand how the dynamics of electronic density influence the physics of the transition state. Because it is based on the QTAIM theory of Professor Bader, the CCTDP parameters have transferability properties, that is, they are characteristic of certain groups. In this way, they can be used for the construction of prediction models, which allowed a quick inference of reactivity without high computational cost. As transferability between molecules is not exact, machine learning techniques will be used for their refinement. (AU)

Scientific publications

(4)

(References retrieved automatically from Web of Science and SciELO through information on FAPESP grants and their corresponding numbers as mentioned in the publications by the authors)

SILVA, ARNALDO F.; DUARTE, LEONARDO J.; POPELIER, PAUL L. A. Contributions of IQA electron correlation in understanding the chemical bond and non-covalent interactions. STRUCTURAL CHEMISTRY, v. 31, n. 2 FEB 2020. Web of Science Citations: 0.

DUARTE, LEONARDO J.; SILVA, ARNALDO F.; RICHTER, WAGNER E.; BRUNS, ROY E. Infrared Intensification and Hydrogen Bond Stabilization: Beyond Point Charges. Journal of Physical Chemistry A, v. 123, n. 30, p. 6482-6490, AUG 1 2019. Web of Science Citations: 0.

DUARTE, LEONARDO J.; BRUNS, ROY E. FTIR and dispersive gas phase absolute infrared intensities of hydrocarbon fundamental bands. SPECTROCHIMICA ACTA PART A-MOLECULAR AND BIOMOLECULAR SPECTROSCOPY, v. 214, p. 1-6, MAY 5 2019. Web of Science Citations: 0.

DUARTE, LEONARDO J.; BRUNS, ROY E. Atomic Polarizations, Not Charges, Determine CH Out-of-Plane Bending Intensities of Benzene Molecules. Journal of Physical Chemistry A, v. 122, n. 51, p. 9833-9841, DEC 27 2018. Web of Science Citations: 1.