Finding predictive models for singlet fission by machine learning

Abstract

Singlet fission (SF), the conversion of one singlet exciton into two triplet excitons, could significantly enhance solar cell efficiency. Molecular crystals that undergo SF are scarce. Computational exploration may accelerate the discovery of SF materials. However, many-body perturbation theory (MBPT) calculations of the excitonic properties of molecular crystals are impractical for large-scale materials screening. We use the sure-independence-screening-and-sparsifying-operator (SISSO) machine-learning algorithm to generate computationally efficient models that can predict the MBPT thermodynamic driving force for SF for a dataset of 101 polycyclic aromatic hydrocarbons (PAH101). SISSO generates models by iteratively combining physical primary features. The best models are selected by linear regression with cross-validation. The SISSO models successfully predict the SF driving force with errors below 0.2 eV. Based on the cost, accuracy, and classification performance of SISSO models, we propose a hierarchical materials screening workflow. Three potential SF candidates are found in the PAH101 set.

Inverse Design of Tetracene Polymorphs with Enhanced Singlet Fission Performance by Property-Based Genetic Algorithm Optimization

“… 57 For tetracene, the best dimer extracted from the T1 structure is predicted to yield a SF rate higher than that of the T2 structure (see Figure 6 ), contrary to experimental observations. 57 We note that in the future Simple may be replaced by more advanced models that go beyond the dimer approximations and/or by machine learning models 36 for the evaluation of the SF-based fitness function. The GW+BSE method can capture the many-body effects in a molecular crystal with periodic boundary conditions.…”

“…Because of the versatility of the genetic algorithm, different fitness functions can be easily implemented. The SF-based fitness function may be improved in the future by using more advanced models that go beyond the dimer approximation and/or machine learning models . Furthermore, GA fitness functions may be tailored to search for any property or combination of properties of interest.…”

“…A singlet exciton with more electron delocalization on neighboring molecules can potentially improve the coupling between singlet and triplet states. − Thus, a higher degree of charge transfer character (%CT) in the lowest singlet excited state is considered beneficial to SF. ,, In addition to efficient SF, practical applications in solar cells require several desirable properties such as photostability, conductivity, and exciton diffusibility. , This has motivated the search for new classes of SF materials that are commercially viable. Recently, several experimental , and computational ,− studies have focused on identifying new candidate materials for SF.…”

pyGWBSE: a high throughput workflow package for GW-BSE calculations

“…Despite its obvious indispensability, high-throughput computational material discovery studies for light-matter interaction related applications have rarely incorporated the QP or excitonic properties of materials using GW-BSE formalism [22][23][24][25] mostly due to the unavailability of an automated workflow implementation that can perform such calculations. Two main challenges for such an implementation is the efficient convergence of multiple parameters and the tractability of the huge computational cost associated with the multi-step GW-BSE formalism.…”

Comparative Analyses of Data Driven Machine Learning Models for TADF Emitters

“…[6][7][8][9][10][11][12][13] Based on the experimental and ab initio data set, several research works have reported excellent ML models on the fast prediction of the band gap, evaluation of singlet-triplet energy gap, and fluorescence and phosphorescence rates in reference to the excited state properties. [14][15][16][17][18] The examples are persistent and continue to be applications in various fields. [19][20][21][22] Thermally activated delayed fluorescence (TADF) materials have been regarded as the most promising approach to harnessing 100% efficiency theoretically from the singlet and triplet exciton in organic electronics.…”