I use theory, first-principles calculations, atomistic simulation and machine learning to connect the structure of layered materials to their electronic properties.
The themes below summarise the work I have published and the directions I am developing. Most of it involves close collaboration with experimental groups at the National Graphene Institute and elsewhere. If one of these areas interests you, the Recruiting page explains how to get in touch.
01 · Developing direction
Machine-learned interatomic potentials and machine learning for materials
Machine-learned interatomic potentials (MLIPs) are trained on first-principles data and reproduce density functional theory energies and forces at a small fraction of the cost. They make it possible to relax and simulate structures with hundreds of thousands of atoms, such as twisted bilayers with long moiré periods, extended defect networks and interfaces, while staying close to DFT accuracy.
I am developing MLIPs for layered materials and van der Waals heterostructures, with particular attention to the weak interlayer interactions that control stacking, sliding and reconstruction. A second strand uses machine learning beyond potentials: models that predict electronic properties directly from structure, and neural-network methods for correlated electrons such as neural quantum states.
02 · Developing direction
Large-angle moiré heterostructures
Most work on twisted 2D materials concentrates on small twist angles, where the moiré period is long and the lattice reconstructs into domains. At larger angles the moiré period shrinks to a few nanometres and reconstruction is weak. The electronic structure then depends on how the two lattices match up and on interlayer hybridisation at particular points of the Brillouin zone.
I am interested in how twisted bilayers and heterostructures behave across the full range of angles, including commensurate angles and the 30° quasicrystalline limit, and in combinations of different 2D crystals. Accurate modelling of these systems needs large supercells, which links this theme to the machine-learned potentials above.
03 · Moiré materials
Lattice reconstruction in marginally twisted bilayers
When two layers are stacked with a small twist or lattice mismatch, the atoms relax to form domains of low-energy stacking separated by networks of domain walls. I model this reconstruction in transition-metal dichalcogenides and NbSe₂, using multiscale models parametrised with density functional theory.
Reconstruction changes the potential that electrons and excitons experience. In MoSe₂/WSe₂ and MoS₂/WS₂ bilayers the domain network produces arrays of self-organised quantum dots, and the competition between sites of the network decides where carriers localise. With electron-microscopy groups in Manchester, I compare these models with images of twist domains and of reconstruction at buried interfaces.
04 · Ferroelectrics
Sliding ferroelectricity and domain-wall electronics
In bilayers with parallel (rhombohedral) stacking, the two stacking configurations carry opposite out-of-plane polarisation, and sliding one layer by one interatomic distance switches between them. This sliding ferroelectricity appears in materials that are not ferroelectric in bulk, and the stacking and twist determine the pattern of its domains.
I study the electronic states that form at domain walls and twin boundaries in these materials, including two-dimensional electron gases at mirror and twistronic twin boundaries, and I use DFT to compare ferroelectric properties across van der Waals chalcogenides. More recent work extends these ideas to sliding ferroic multipoles in van der Waals antiferromagnets.
05 · Electronic structure
Band-edge physics of 2D semiconductors from first principles
Quantitative predictions for devices need accurate band-edge parameters: effective masses, spin–orbit splittings, valley energies and their response to gating. Standard density functional theory does not always give these correctly, so I work on first-principles recipes that correct them, calibrated against measurements on MoS₂ and other transition-metal dichalcogenides.
Recent work refines DFT for monolayer MoS₂ using the measured spin–orbit splitting, explains the gate-tunable band edge in few-layer MoS₂, and adds extended Hubbard corrections (on-site U and inter-site V) that control the orbital character of band-edge states in MoS₂, MoSe₂, WS₂ and WSe₂.
06 · Defects and mechanics
Defects, dislocations and curvature in layered materials
Layered materials host defects with no direct counterpart in ordinary crystals, such as ripplocations, alongside dislocations whose structure is changed by the weak bonding between layers. I model the structure and energetics of prismatic edge dislocations and partial basal dislocations in graphite and bilayer graphene, and the dislocation networks that form in strained hexagonal boron nitride.
Defects and curvature also change electronic properties and transport. This theme includes charge redistribution across dislocation networks, flexoelectricity in graphene curved at the sub-nanometre scale, the diffusion of fission products and metal nanoclusters on graphene and in graphite, and friction on strained graphene.
Across all themes
Working with experiment
Much of this work is done with experimental groups, particularly in electron microscopy and transport. I provide modelling and DFT for studies of gold species at liquid–solid interfaces, transition-metal diiodides, lithium intercalation in bilayer graphene and quantum transport in MoS₂.