Toward Prediction of Nonradiative Decay Pathways in Organic Compounds II: Two Internal Conversion Channels in BODIPYs

Zhou Lin,Alexander Kohn,Troy Van Voorhis
DOI: https://doi.org/10.26434/chemrxiv.9755951
2019-09-03
Abstract:Boron-dipyrromethene (BODIPY) molecules are widely used as laser dyes and have therefore become a popular research topic within recent decades. Numerous studies have been reported for the rational design of BODIPY derivatives based on their spectroscopic and photophysical properties, including absorption and fluorescence wavelengths ( λ abs and λ fl ), oscillator strength ( f ), nonradiative pathways, and quantum yield ( φ ). In the present work, we illustrate a theoretical, semi-empirical model that accurately predicts φ for various BODIPY compounds based on inexpensive electronic structure calculations, following the data-driven algorithm proposed and tested on the naphthalene family by us [Kohn, Lin, and Van Voorhis, J. Phys. Chem. C. 2019 , 123 , 15394]. The model allows us to identify the dominant nonradiative channel of any BODIPY molecule using its structure exclusively and to establish a correlation between the activation energy ( E a ) and the fluorescence quantum yield ( φ fl ). Based on our calculations, either the S 1 → S 0 or L a → L b internal conversion (IC) mechanism dominates in the majority of BODIPY derivatives, depending on the structural and electronic properties of the substituents. In both cases, the nonradiative rate ( k nr ) exhibits a straightforward Arrhenius-like relation with the associated E a . More interestingly, the S 1 → S 0 mechanism proceeds via a highly distorted intermediate structure in which the core BODIPY plane and the substituent at the 1-position are forced to bend, while the internal rotation of the very same substituent induces the La → Lb transition. Our model reproduces k fl , k nr , and φ fl to mean absolute errors (MAE) of 0.16 decades, 0.87 decades, and 0.26, when all outliers are considered. These results allow us to validate the predictive power of the proposed data-driven algorithm in φ fl . They also indicate that the model has a great potential to facilitate and accelerate the machine learning aided design of BODIPY dyes for imaging and sensing applications, given sufficient experimental data and appropriate molecular descriptors.
What problem does this paper attempt to address?