Bernd Ensing

Bernd Ensing

PAH spectra from machine learning

Investigating vibrational properties of PAHs using Artificial Neural Networks

Polycyclic Aromatic Hydrocarbons (PAHs) are molecules of great astronomical interests. In this project, we aim to predict anharmonic vibrational spectra for astronomical polycyclic aromatic hydrocarbons (PAHs) using artificial neural networks.

Project type and duration:

Master project 6 - 9 Months

Supervisors:

Bernd Ensing and Alessandra Candian

Group:

Computational Chemistry

Institute:

Van 't Hoff Institute for Molecular Sciences

Project description

Description: Polycyclic Aromatic Hydrocarbons (PAHs) are molecules of great astronomical interests, since their infrared spectrum is detected by space telescope such as the James Webb Space Telescope everywhere in space, from the very early universe to regions where stars and planets are formed (Candian, Zhen and Tielens, 2018)
To fit the astronomical PAH spectrum and learn how these molecules are created and destroyed in space, astronomers use a database containing Density Functional Theory-computed vibrational spectra in the harmonic approximation.
Because PAHs in space are highly excited, this approximation is very crude, and we should improve on it. Anharmonic DFT-based vibrational spectra obtained with generalised 2nd order perturbation theory (GVPT2) can be calculated for PAHs (e.g. Mackie et al, 2018) but the calculations become quickly very expensive as the size increases.
In this project we will use a database of calculated anharmonic vibrational spectra for small PAHs as a train set for an Artificial Neural Network to predict the spectra of larger species and compare with experimental results.

Figure 1: Image from Laurens et al. (2021), Theoretical Chemistry Accounts, 140, 66

Aim: Predict anharmonic vibrational spectra for astronomical polycyclic aromatic hydrocarbons (PAHs) using artificial neural networks.

References:

Candian, Zhen and Tielens, 2018, Physics Today 17, 11.
Mackie et al, 2018, Physical Chemistry Chemical Physics 20 (2), 1189-1197.

Interested? please apply here

Other projects

Master Project
Peter Bolhuis

The emergence of (auto)catalysis

In this fundamental research project the student will elucidate how catalysis emerges by applying a novel method to optimise molecular models so that they have desired kinetic properties, such as catalysis.

Read More »