On the Electronic Structure of Nanodiamonds for Photocatalysis

Abstract

Nanodiamond photocatalysis has the potential to replace a range of high-energy industrial processes and provide a green alternative for energy harvesting and the production of chemical feedstocks. This thesis investigates the properties and characteristics of nanodiamonds in the context of photocatalysis with a focus on their diverse electronic structures.

First, we characterize a sample of fluorinated nanodiamonds from hydrogen fluoride synthesis via soft X-ray spectroscopy. X-ray photoelectron spectroscopy reveals a fluorine coverage of about 50 % on the nanodiamonds. The analysis is complemented by X-ray absorption spectroscopy (XAS) and resonant inelastic X-ray scattering spectroscopy (RIXS) experiments and supported by theoretical investigations of the underlying systems. The observed XAS and RIXS signatures are verified and explained by the quantum chemical calculations which show that the XAS signals systematically shift upon increasing the surface fluorine content. On the other hand, the same F K-edge RIXS signature is found for a broad range of fluorinated hydrocarbons, with the main signal originating from a selective excitation of the F 1s electron into an antibonding C-F sigma* orbital in all cases.

Furthermore, we investigate the interaction of nanodiamonds with adsorbates in aqueous dispersion, with a focus on charge transfer doping towards oxidative adsorbates. The observed charge transfer is found to increase with the adsorbates' standard reaction potentials in water and can further be tuned by modifying the nanodiamonds' highest occupied molecular orbital energies. The nanodiamond charge transfer excited states are similarly influenced by aqueous oxidative adsorbates, which capture a large fraction of the excited electrons. Nanodiamond cluster formation results in lowering of their optical gaps and preserves the atomic orbital-like shapes in the clusters' lowest unoccupied orbitals.

Finally, we introduce the ND5k data set which consists of 5,089 structures and frontier orbital energies of nanodiamonds. Based on this data, we suggest to consider the use of phosphorous-doped nanodiamonds for sunlight-driven photocatalysis. Furthermore, modern machine learning algorithms are evaluated for molecular property prediction of the ND5k structures. In this context, we propose an extension of graph neural networks using a set of tailored atomic descriptors which we test for the enn-s2s network architecture. The best results are obtained using the PaiNN graph neural network, the second best from our modified enn-s2s variant.

Overall, this work contributes to a better understanding of the electronic structures of nanodiamonds to aid future research in nanodiamond photocatalysis.