Local spectroscopy of atomically thin MoS2: electronic states at 1D defects, charge transfer and screening

The last decade has seen immense research effort dedicated to atomically thin transition metal dichalcogenide layers. This owes itself to their wide range of exotic properties and to the possibility of combining them with other two-dimensional materials, such as graphene, in a heterostructure. There is particular interest in MoS2 and its analogues MoSe2, WS2 and WSe2 because they are intrinsically semiconducting, opening up new possibilities for electronic and optoelectronic devices at the nanoscale. Transition metal dichalcogenide layers are typically prepared by exfoliation from the bulk, chemical vapour deposition or molecular beam epitaxy. The last offers samples of superlative quality and of scalable size, but for many years the community struggled to grow transition metal disulfides - including the paradigmatic MoS2 - by this method. Thanks to recent advances by our group, we can synthesise high-quality MoS2 layers on graphene on Ir(111) by a molecular beam epitaxy variant known as van der Waals epitaxy. Performed under ultra-high vacuum conditions, the MoS2 surface is kept free of adsorbates and local spectroscopic investigation is facilitated. Furthermore, the growth method leads to excellent epitaxial alignment of the MoS2 layer with its weakly interacting substrate. This results in the only common defects being well-defined and one-dimensional: atomically straight island edges, and mirror twin boundaries which form due to the reduced symmetry of MoS2 (three-fold) compared to graphene (six-fold). As the fabrication of transition metal dichalcogenides improves, such defects will become increasingly relevant for devices. This thesis studies the electronic properties of two-dimensional MoS2 and its one-dimensional defects, and how these properties are dependent on their environment. This is principally performed with scanning tunnelling microscopy and spectroscopy at 5K, on the substrate graphene/Ir(111) which only weakly influences the MoS2 layer. The main focus is on 4|4E mirror twin boundaries, which represent metallic, one-dimensional wires. Their one-dimensionality, their electronic isolation from the semiconducting layer, and the lack of screening in two dimensions leads to extraordinary electron-interaction behaviour, both in and around the boundary. The thesis is composed of four manuscripts, each presented in their entirety. They are bookended by an introduction of relevant theoretical and experimental background information, and by a discussion of the outcomes, possible future experiments, and a brief look at the state of play within the community. First, the electronic band structures of mono-, bi- and trilayer MoS2 on graphene/Ir(111) are characterised by `comprehensive' tunnelling spectroscopy, whereby three different spectroscopic modes are used and compared. This shows monolayer MoS2 to be quasi-freestanding on this substrate, with a 2.53eV band gap size which is larger than that reported for other metallic substrates. Through this comprehensive spectroscopy, we can also identify critical point energies within the MoS2 band structure and compare them with theoretical predictions. It additionally allows us to follow the dynamics of interlayer coupling as the MoS2 thickness is increased from mono- up to trilayer. Through the MoS2 case study and thorough technical discussion, this work also demonstrates that traditional tunnelling spectroscopy (only using constant height mode) is inadequate for accurate estimation of the electronic band gap. Second, electronic states which exist along the axis of MoS2 4|4E mirror twin boundaries are examined. These boundaries can be viewed as one-dimensional wires of finite length. In high-resolution tunnelling spectroscopy they exhibit a wire-length-dependent energy gap around the Fermi level, quantised energy levels beyond this, and periodic modulation of the electron density along their axis. Together with density functional theory and model Hamiltonian calculations, we explain our observations to be unambiguous evidence for the existence of a confined Tomonaga-Luttinger liquid. Due to the lack of electrostatic screening, the electrons experience a strong Coulomb interaction and act as coherent, bosonic excitations. Moreover, the electronic excitations in the wire are split into collective modes of spin and charge density, consistent with our spectroscopic data. This represents the first real-space observation of spin-charge separation, and provides a rare opportunity to probe Tomonaga-Luttinger liquid physics locally. In the third manuscript, we investigate how the surrounding MoS2 layer is influenced by its line defects, such as island edges and the 4|4E and 4|4P mirror twin boundary types. Focussing on the 4|4E boundary again, we find that it causes a large upwards bending of the MoS2 valence and conduction bands over a 5nm range. Remarkably, this occurs stepwise in the valence band, producing characteristic quantised states. Using tunnelling spectroscopy, density functional theory and electrostatic continuum modelling, we elucidate the origin of the bending and quantisation. Inherent interface polarisation charge is counteracted by charge from the surrounding layer and charge donated by the substrate, often leaving the line defects with an excess electron density. The resultant electrostatic potential bends the MoS2 bands, and creates a hole-confining quantum well. Additionally, the boundary is found to suppress the transport of holes across it, leaving independent quantum wells on either side. Finally, the three previous manuscripts are built upon when we modify the graphene/Ir(111) substrate through Eu- or O- intercalation. This changes the electronic environment of the MoS2 layer, but keeps it chemically pristine with graphene acting as a buffer layer. Via tunnelling spectroscopy, the electrostatic response of MoS2 can be observed in its band structure, the states confined along mirror twin boundaries and the quantised band bending around them. The effects are strongest in MoS2 on graphene/Eu/Ir(111), where the Eu strongly n-dopes the graphene. This leads to n-gating of MoS2 and large band gap renormalisation due to the increased screening efficiency in graphene. Meanwhile, more electrons are donated to the 4|4E boundary -- this is evidenced by a changed periodicity of its confined states. In the band bending, the boundary charge is better screened by the heavily doped graphene, leading to a much narrower potential well. The O-intercalation layer has the opposite effect; the boundary charge experiences weaker screening due to decoupling from Ir, leading to a wider potential well. This in turn affects the quantised valence band states. In both cases, the band bending and resultant quantised states can be reproduced by electrostatic modelling. In the Scientific Appendix we show that the graphene/Ir(111) moire causes a superperiodic potential which locally modulates the MoS2 band structure by 30meV, motivating further studies with more strongly corrugated substrates. Additionally, we characterise the electronic properties of two point defect types in MoS2, revealing their distinctly different in-gap states and their electrostatic influence on their surroundings