Accelerating structure discovery for functional materials

Download Statistics

Downloads

Downloads per month over past year

Dorrell, Jordan Antony ORCID: https://orcid.org/0000-0003-0043-0222 (2024). Accelerating structure discovery for functional materials. University of Birmingham. Ph.D.

Preview

Dorrell2024PhD.pdf
Text - Accepted Version
Available under License All rights reserved.
Download (61MB) | Preview

Abstract

Discovering new functional materials and furthering understanding of known functional materials benefits from computational structure searching approaches that can identify the ground state structure for a particular system. To this end, using both density-functional theory (DFT) and machine learning (ML) models, I have searched for the structures of new materials with potential applications as electrodes for alkali-ion batteries and as single-photon emitters (SPEs).
To search for new Li-Ni-S ternary electrodes with a high gravimmetric charge capacity, I have taken a random structure searching approach. Initial searches were performed with an ab initio model, but the process was accelerated by three orders of magnitude by switching to a machine learning model. Over 150,000 structures were generated and while no ternary compounds were identified on the convex hull, new polymorphs of NiS and NiS\(_2\) were discovered that are lower in formation energy than the structures reported experimentally and on the Materials Project.
Ab initio random structure searching (AIRSS) was also applied to search for point defects responsible for single photon emission in monolayer hBN. Previous studies of hBN as an SPE have focused on only a limited range of intrinsic defects. In this work I have generated over 5000 intrinsic and extrinsic defects and calculated their defect formation energies. Numerous new low energy defects were discovered that have not previously been considered as SPE sources. The optical absorption and emission spectra of these defects can now be modelled so that they can be compared to experiment in hopes of identifying defects that give rise to desirable optical phenomena.
K-intercalated graphite (KGIC) is a potential anode material for K-ion batteries and its electrochemical synthesis can be observed unobtrusively through Raman spectroscopy. To better our understanding of the relationship between the structure and the Raman spectra, I have modelled Raman spectra for stages 1-5 KGIC within the conventional Rüdorff- Hofmann (RH) model and a range of candidate cells within the Daumas-Hérold (DH) model, which were generated with a bespoke machine-learning driven algorithm. DH-type cells were energetically accessible at room temperature but the modelled Raman spectra showed poor correlation with experiment. I conclude that the RH model is more consistent with experimentally observed KGICs, but the material may contain DH-type defects in small quantities.
In addition, two Na-niobates with potential electrode applications, layered Na\(_3\)NbO\(_4\) and NaNb\(_7\)O\(_{18}\), were recently synthesised. To verify the structures of these compounds, I have modelled their \(^{24}\)Na NMR spectra. In the case of NaNb7O18, the positions of the Na\(^+\) ions were unknown and so I have modelled the spectra of a range of candidate cells. The total energy of these candidate cells differed by less than 3.3 meV / atom, suggesting that the Na\(^+\) ions are disorderly distributed between several sites throughout the structure. The modelled \(^{24}\)Na NMR spectrum of layered Na\(_3\)NbO\(_4\) shows exceptional correlation with the experimental spectrum, suggesting a strong match between the candidate structure and the synthesised material. The modelled \(^{24}\)Na NMR spectrum of NaNb\(_7\)O\(_{18}\) contained three peaks which correlate well with peaks in the experimental spectrum. The modelled spectra did not reproduce the fourth peak, which has been attributed to a NaNb3O8 impurity. The model cells of NaNb\(_7\)O\(_{18}\) therefore also seem to be consistent with the experimental sample.

Type of Work:

Thesis (Doctorates > Ph.D.)

Award Type:

Doctorates > Ph.D.

Supervisor(s):