Machine learning-based screening of complex molecules for polymer solar cells

Abstract

Polymer solar cells admit numerous potential advantages including low energy payback time and scalable high-speed manufacturing, but the power conversion efficiency is currently lower than for their inorganic counterparts. In a Phenyl-C_61-Butyric-Acid-Methyl-Ester (PCBM)-based blended polymer solar cell, the optical gap of the polymer and the energetic alignment of the lowest unoccupied molecular orbital (LUMO) of the polymer and the PCBM are crucial for the device efficiency. Searching for new and better materials for polymer solar cells is a computationally costly affair using density functional theory (DFT) calculations. In this work, we propose a screening procedure using a simple string representation for a promising class of donor-acceptor polymers in conjunction with a grammar variational autoencoder. The model is trained on a dataset of 3989 monomers obtained from DFT calculations and is able to predict LUMO and the lowest optical transition energy for unseen molecules with mean absolute errors of 43 and 74 meV, respectively, without knowledge of the atomic positions. We demonstrate the merit of the model for generating new molecules with the desired LUMO and optical gap energies which increases the chance of finding suitable polymers by more than a factor of five in comparison to the randomised search used in gathering the training set.

Guest Editorial: Special Topic on Data-Enabled Theoretical Chemistry

“…The position of the LUMO and the width of the optical gap in polymers for solar cells are important for power conversion efficiency. Jørgensen et al 69 perform first-principles calculations on about 4 000 monomers and show that a grammar variational autoencoder using a simple string representation makes quite accurate predictions, reducing the cost of a search by up to a factor of 5. Afzal et al 70 model the refraction index of organic polymers by combining first-principles calculations with ML to predict packing fractions of the bulk polymers.…”

Deep Generative Models for Molecular Science

“…In this case the SMILES grammar formulation may be unnecessarily complex and can be replaced with a simpler application specific grammar that is easier to handle for the grammar VAE. This approach is used by Jørgensen et al [23] for screening of materials for polymer solar cells where each material is composed from a library of acceptor, donor and side group substructures. If the application allows it, the grammar can be formulated such that a syntactically valid string implies a semantically valid molecule, such that the model only generates valid molecules.…”

“…The problem of interest in [23] is to find new materials for polymer solar cells. The polymer units are composed by one of 13 acceptor units, one of 10 donor units and a number of side groups.…”

“…[19] Instead they train a feedforward neural network on the latent representation and then optimize the position by taking a few gradient steps starting from the latent representation of a molecule with good properties. [19] Jørgensen et al [23] continuing the example above) sample a large number of random points from the latent space and this set of points is decoded and encoded repeatedly to find well-behaved regions of the latent space. After a number of iterations the best points according to the regression model are selected for further studies.…”

Computational high throughput screening of inorganic cation based halide perovskites for perovskite only tandem solar cells

“…This far, to the best of our knowledge, there has been no experimental literature that explored the prospects of tandem solar cell efficiencies based on hybrid halide perovskites. There are, however, a number of theoretical publications available, that concentrated in high-throughput screening of perovskite materials, in terms of their energy band gaps using density functional theory calculations [11][12][13][14][15][16][17][18][19][20][21][22]. In 2018, we performed a high-throughput study on hybrid perovskite materials based on DFT calculations that aimed in finding novel perovskite materials for hybrid perovskite-only tandem solar cells [23].…”