Machine Learning for Singlet Fission Chromophores Quest
Alia Tadjer a, Lyuben Borislavov b, Julia Romanova a
a Sofia University, Faculty of Chemistry and Pharmacy, new
b Institute of General and Inorganic Chemistry, Bulgarian Academy of Sciences
Materials for Sustainable Development Conference (MATSUS)
Proceedings of MATSUS Spring 2024 Conference (MATSUS24)
#SF-TF - Singlet Fission and Triplet Fusion
Barcelona, Spain, 2024 March 4th - 8th
Organizer: Jonas Sandby Lissau
Poster, Alia Tadjer, 478
Publication date: 18th December 2023

Singlet fission is a photophysical phenomenon, which has the potential to double the efficiency of the solar cells [1]. However, there is a major obstacle on the road toward singlet fission based solar cell technologies – the number of known materials able to undergo singlet fission is relatively limited. This is a consequence of the following problems related to singlet fission materials: 1) they should satisfy a long list of requirements, so they are inherently rare, 2) their experimental characterization is not trivial, 3) their modelling with low cost computational methods like density functional theory is questionable.

Our approach to overcome the above mentioned impediments is the development of statistical models able to find new singlet fission chromophores within large general purpose databases of existing compounds or among newly designed sets of molecules. Here, we demonstrate two models, which rely on efficient combination of quantum chemical calculations and machine learning approaches. The first one is a classification model, which is able to predict potential singlet fission chromophores based on their diradical character. This model is trained with more than 500 000 compounds and makes predictions in seconds [2]. It is implemented in a web application for users convenience. The second one is a regression model, which can predict the feasibility conditions for singlet fission. This model is trained with about 40 molecules and can quickly map DFT excited states data to theoretical estimations obtained with computationally demanding multiconfigurational methods.

The authors acknowledge the financial support of the Bulgarian National Science Fund, contract KP-06-H39/2/2019

© FUNDACIO DE LA COMUNITAT VALENCIANA SCITO
We use our own and third party cookies for analysing and measuring usage of our website to improve our services. If you continue browsing, we consider accepting its use. You can check our Cookies Policy in which you will also find how to configure your web browser for the use of cookies. More info