Data-Driven Design Boosts Two-Photon Absorption in Light-Driven Nanomotors

Light-driven molecular nanomotors represent a potentially transformative technology for fields ranging from materials science to biomedicine, but realising their full potential requires optimising their light-harvesting and energy conversion capabilities. Alexander Mielke, Alexander Scrimgeour, and Enrico Tapavicza, from the University of Regensburg and California State University, Long Beach, present a systematic investigation into the chemical space of these nanomotors, focusing on enhancing their performance for both conventional and two-photon activation. Their work analyses a dataset of over two thousand molecular designs, identifying key structural features that maximise light absorption and drive efficient molecular motion, achieving increases in two-photon absorption strengths of up to two orders of magnitude compared to existing designs. By introducing a novel ‘photoreactivity score’ and employing machine learning models, the researchers demonstrate a powerful new approach to accelerate the discovery of advanced nanomotors, potentially replacing time-consuming computational chemistry with rapid, accurate predictions.

Molecular Nanomotor Photochemical Properties and Prediction

Researchers employed advanced computational techniques, including quantum chemical calculations and machine learning, to understand and improve the efficiency of molecular nanomotors, tiny machines powered by light. A large dataset of nanomotor designs and their calculated properties was created to train and validate predictive models, aiming to optimize their ability to rotate in response to light. Quantum chemical calculations, such as time-dependent density functional theory, determined how each nanomotor absorbs light and transitions between energy states, while non-adiabatic molecular dynamics simulations modeled their movement. Machine learning algorithms, including XGBoost and deep learning, were trained on these calculated properties to predict the performance of new designs.

Data-Driven Design of Light-Powered Nanomotors

Researchers adopted a data-driven approach to design highly efficient nanomotors, systematically altering the molecular structures of over 2000 potential candidates to explore a wide range of possibilities. This process began with a foundational structure, modified by introducing different chemical groups at various positions to generate a diverse library of potential nanomotors. To accurately assess each nanomotor, the team employed quantum mechanical calculations, including density functional theory and time-dependent density functional theory, predicting how each molecule would respond to light. A key innovation was the development of a “photoreactivity score”, which gauged how closely the excited state of each molecule resembled the ideal state for efficient photoisomerization, a critical step in the nanomotor’s operation. To streamline analysis and accelerate future discoveries, the researchers also explored machine learning models, training them on calculated properties to predict the performance of new designs.

Data-Driven Design Boosts Nanomotor Efficiency

Researchers are making significant strides in the development of light-driven molecular nanomotors, with a new data-driven approach yielding substantial improvements in their efficiency. These nanomotors, which convert light into mechanical motion at the nanoscale, hold promise for applications ranging from advanced materials to targeted drug delivery. The team analyzed a vast dataset of over 2000 nanomotor designs, varying the chemical groups attached to the core structure to influence how they absorb and respond to light. By carefully considering the excited state properties of these molecules, they developed a “photoreactivity score” to predict how effectively a given design would undergo the necessary structural changes to drive rotation. The results demonstrate a remarkable increase in two-photon absorption strengths, with the best designs exhibiting up to a hundredfold improvement compared to existing nanomotors. Furthermore, the team developed machine learning models capable of accurately predicting a molecule’s photochemical properties, potentially replacing computationally intensive chemistry calculations.

Photoreactivity Predicts Efficient Nanomotor Design

This study systematically investigated the design of light-driven molecular nanomotors, generating and analyzing a dataset of over 2000 potential candidates. Researchers identified key criteria for efficient photoisomerization, including strong light absorption, suppression of reverse rotation, and preservation of specific excited state characteristics. A novel photoreactivity score (PRS) was developed to assess the likelihood of successful photoisomerization following light absorption, allowing for targeted molecular design. The results demonstrate that incorporating push-pull functional groups significantly enhances two-photon absorption strengths, improving the potential of these nanomotors for biomedical applications requiring near-infrared light. Machine learning models accurately predicted key photochemical properties, offering a computationally efficient method for screening potential candidates. Future work will focus on non-adiabatic molecular dynamics simulations to verify the predictive power of the PRS and to assess the quantum yield of the candidate molecules.