Comparison of multifidelity machine learning models for potential energy surfaces.

Multifidelity modeling is a technique for fusing the information from two or more datasets into one model. It is particularly advantageous when one dataset contains few accurate results and the other contains many less accurate results. Within the context of modeling potential energy surfaces, the low-fidelity dataset can be made up of a large number of inexpensive energy computations that provide adequate coverage of the N-dimensional space spanned by the molecular internal coordinates. The high-fidelity dataset can provide fewer but more accurate electronic energies for the molecule in question. Here, we compare the performance of several neural network-based approaches to multifidelity modeling. We show that the four methods (dual, Δ-learning, weight transfer, and Meng-Karniadakis neural networks) outperform a traditional implementation of a neural network, given the same amount of training data. We also show that the Δ-learning approach is the most practical and tends to provide the most accurate model.

Exploiting Excited-State Aromaticity To Design Highly Stable Singlet Fission Materials.

Singlet fission, the process of forming two triplet excitons from one singlet exciton, is a characteristic reserved for only a handful of organic molecules due to the atypical energetic requirement for low energy excited triplet states. The predominant strategy for achieving such a trait is by increasing ground state diradical character; however, this greatly reduces ambient stability. Herein, we exploit Baird's rule of excited state aromaticity to manipulate the singlet-triplet energy gap and create novel singlet fission candidates. We achieve this through the inclusion of a [4n] 5-membered heterocycle, whose electronic resonance promotes aromaticity in the triplet state, stabilizing its energy relative to the singlet excited state. Using this theory, we design a family of derivatives of indolonaphthyridine thiophene (INDT) with highly tunable excited state energies. Not only do we access novel singlet fission materials, they also exhibit excellent ambient stability, imparted due to the delocalized nature of the triplet excited state. Spin-coated films retained up to 85% activity after several weeks of exposure to oxygen and light, while analogous films of TIPS-pentacene showed full degradation after 4 days, showcasing the excellent stability of this class of singlet fission scaffold. Extension of our theoretical analysis to almost ten thousand candidates reveals an unprecedented degree of tunability and several thousand potential fission-capable candidates, while clearly demonstrating the relationship between triplet aromaticity and singlet-triplet energy gap, confirming this novel strategy for manipulating the exchange energy in organic materials.