jp9b08292_si_002.zip (506.97 kB)
Download fileToward Prediction of Nonradiative Decay Pathways in Organic Compounds II: Two Internal Conversion Channels in BODIPYs
dataset
posted on 2020-02-05, 15:07 authored by Zhou Lin, Alexander W. Kohn, Troy Van VoorhisBoron-dipyrromethene
(BODIPY) molecules are widely used as laser
dyes and have therefore become a popular research topic within recent
years. Numerous studies have been reported for the rational design
of BODIPY derivatives based on their photophysical properties, including
absorption and fluorescence wavelengths (λabs and
λfl), oscillator strength (f), nonradiative
pathways, and quantum yield (Φ). In the present work, we illustrate
a theoretical, semiempirical model that accurately predicts Φ
for various BODIPY compounds on the basis of inexpensive electronic
structure calculations, following the data-driven algorithm proposed
by us in a previous study [Kohn
et al. 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 (Ea) and the fluorescence
quantum yield (Φfl). On the basis of our calculations,
either the S1 → S0 or La → Lb internal conversion
(IC) mechanism dominates in the majority of BODIPY derivatives, depending
on the structural and electronic properties of the substituents. In
either case, the nonradiative rate (knr) exhibits a straightforward Arrhenius-like relation with the associated Ea. More interestingly, the S1 →
S0 mechanism proceeds via a highly distorted intermediate
structure in which the core BODIPY plane and the substituent at the
1-position are twisted, while the internal rotation of the very same
substituent induces the La → Lb transition. Our model reproduces kfl, knr, and Φfl to mean absolute errors (MAEs) 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 Φfl. 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.