Quantum embedding for functional nanodevice design

Interfaces of metals and molecules are important in a wide range of technologies: There is remarkable progress in increasing the sensitivity of analytical and non-destructive medical imaging techniques by using so-called plasmonic metal nanoparticles functionalised with light-sensitive molecules to enable sensing single molecules and profiling cancerous tissues, for example. Single organic molecules can also be connected to nanostructured metallic tips to create miniaturised electronics components such as diodes and switches, study chemical reaction mechanisms and understand how light interacts with matter. For many of these applications, there is in particular a search for molecules with unique quantum properties based on how they interact with light and charges, as well as how they can be anchored to metallic surfaces. Plasmonic metal nanostructures also provide a way to utilise solar radiation to catalyse chemical reactions that can help us to move towards the sustainable production of high-value chemicals and fuels.
However, there is still a lack of understanding of the fundamental physical and chemical processes behind these technologies which makes it extremely hard to establish design principles for finding good candidate molecules or increasing the efficiency of these devices and catalysts. There is an urgent need for accurate computational methods based on quantum mechanics that can model light- and charge-driven processes for molecule-metal interfaces.
My aim is to combine the strengths of computational techniques used in three different fields: materials modelling, molecular modelling, and machine learning to develop new computational methods that can give us fundamental insights into the mechanisms of light- and charge-driven processes at molecule-metal interfaces.
My work will widen access to a highly accurate quantum chemistry technique, Density Functional Embedding Theory, that employs molecular and materials modelling techniques in a multiscale fashion, by handling the molecular neighbourhood and the extended metallic slab separately, with different levels of approximate quantum chemistry methods. Despite its demonstrated predictive power, due to its high computational cost and complexity so far this methodology has been quite limited in the number and type of processes it can model. I will present a novel machine learning-based acceleration of this quantum embedding technique, that will enable me to study optimal reaction pathways and dynamics to study processes that are currently out of reach of the method. I will focus on modelling latest ultrafast surface spectroscopic measurements and surface catalytic reactions to understand light-matter interaction and charge transfer across molecule-metal surfaces. By predicting the selectivity and rate of surface catalysed reactions, these simulations will help us to improve current technology, for example, for carbon dioxide reduction and solar fuel production. The developed methodology will have impact well beyond molecule-metal interfaces; it will offer an alternative way to model, for example, defects in solids, spectroscopy or reactions in solution, and on non-metallic surfaces.
The developed methods and the fundamental insights into light- and charge-driven processes will help me establish molecular design principles and develop machine learning-based screening and design techniques for finding molecules with optimal quantum properties for single-molecule electronics, bioimaging nanoparticles, and other optoelectronic devices. By enhancing the sensitivity and stability of such devices, this will enable improved reaction monitoring and biosensing, and open up new applications of plasmonic nanostructures.