Quantum transport in 2D materials: Theoretical and computational optimisation of large heterostructures with spintronics properties

Abstract

Graphene is a monolayer of carbon atoms arranged in a hexagonal lattice, making it the thinnest and strongest material known to man. Its exceptional electronic and thermal properties have generated great interest for its potential use in various applications, particularly in electronics and spintronics. Since its recent isolation, graphene and related 2D materials have been in the scientific spotlight owing to various fundamental discoveries regarding their synthesis and exceptional properties. In particular, graphene nanoribbons (GNRs) are narrow strips of graphene with widths ranging from a few nanometers to several tens of nanometers. GNRs also have improved charge transport properties compared to graphene, making them attractive for use in electronic applications. GNRs have potential applications in spintronics, a field that exploits the intrinsic spin of electrons for information storage and processing. The two symmetry-breaking edges in GNRs are known for being host to spin-polarized edge states, which can enable the creation of spin-based devices and enable the transport of spin currents. Despite impressive advances in fabrication techniques, it is an ongoing challenge to produce and control the desired transport properties in GNR devices. Therefore, characterising the effects of realistic disorders on device behaviour remains crucially important. When dealing with a realistic system size, theoretical predictions of spin properties can be intractable in terms of computational resources. Machine learning (ML) techniques have been employed in various fields, such as consumer recommendation systems, protein folding and chemistry, to exploit patterns in data and make predictions. In this thesis, we address ML techniques to accurately estimate the transport properties and magnetic moment profiles for arbitrarily large and disordered systems. Alongside conventional techniques, developing a neural network tool that accurately estimates the magnetic profile for large and disordered GNRs, we have conducted a thorough analysis on how the edge disorder impacts the robustness of spin-currents in GNRs. The robustness of spin-currents in zigzag graphene nanoribbons (ZGNRs) is highly intertwined with the edge roughness profile at low energies. Whereas spin current is persistent in smooth-edged ribbons due to the absence of back-scattering possibilities, short-ranged scatterers in rough-edged profiles curtail the establishment of edge spin-polarised currents. Our results highlight how ML, by predicting quickly and accurately moment profiles for realistic systems, complements conventional transport techniques to study magnetism and spin transport in 2D materials.