Embedded Cluster Modelling for Realistic Solid-State Systems

点击翻译按钮获取中文摘要

The advent of modern highly-parallelised computing (HPC) infrastructure, coupled with development of scalable software packages, has led to unprecedented growth in the application of materials chemistry modelling - it is now unthinkable to perform high-impact research without including simulations to either understand observations or predict novelty. Within the materials chemistry modelling community, the most widely-used technique is periodic density functional theory (DFT). Such an approach is highly efficient for systems with high-symmetry (i.e. few atoms in the unit cell); however, a major challenge exists when expanding the model to tackle problems such as surface reactivity on close-packed ionic materials. The typical workaround is to create a repeating surface model (i.e. a surface supercell), which is big enough in the surface norm and across the vacuum region to ensure the removal of spurious "image" interactions, and has several layers of "inactive" sub-surface atoms included to ensure chemical validity. Whilst pragmatic, this approach is throttling the impact of computational simulation on applied catalytic chemistry, because the increased model size results in computational overheads that limit computational accuracy to the lower levels of DFT. Therefore, new approaches need to be realised that enable higher accuracy, realistic simulation for solid-state systems.In this work, my aim is to extend the embedded-cluster hybrid quantum-/molecular-mechanics (QM/MM) approach in order to challenge the working norm and offer a viable option instead of periodic-DFT. The embedded-cluster approach removes periodic boundary conditions, and QM/MM can allow the reduction of the electronic space of interest to just the atoms around an active site, thus reducing computational cost without compromising chemical accuracy. To achieve this goal, significant development work is needed to make this technique accessible for solid-state modelling of surface reactions, including streamlining of the setup procedures (cluster design, forcefield parameterisation). Additionally, I propose extensions of QM/MM to accurately model magnetic materials: accurate embedding environments, which apply appropriate potentials to QM atoms at the QM/MM boundary, will be realised through development of novel pseudopotentials and wavefunction embedding approaches. The development outcomes will be validated by investigations of industrially-relevant green catalytic processes for H2 synthesis, which use metal oxides, with our highest-level benchmark being quantum chemical simulations of catalysis on cation-doped iron oxide polymorphs. These investigations will be followed with extension into previously inaccessible fields of materials simulation, such as elucidating reactivity of Mn- and Fe- containing perovskites for the oxygen evolution reaction, and simulating defect properties and reactivity for contemporary 2D magnetic materials. Accurate, high-level DFT and post-Hartree Fock approaches will be realised for extended systems through the Fellowship outcomes, and their application will allow unprecedented insight into chemical properties of emergent materials, as well as opening up a range of further exciting scientific areas beyond the solid-state.