We investigate the proximity-induced spin-orbit coupling in heterostructures of twisted graphene and monolayers of transition metal dichalcogenides (TMDCs) ${\mathrm{MoS}}_2, {\mathrm{WS}}_2, {\mathrm{MoSe}}_2$, and ${\mathrm{WSe}}_2$ from first principles. We identify strain, which is necessary to define commensurate supercells, as the key factor affecting the band offsets and thus magnitudes of the proximity couplings. We establish that for biaxially strained graphene the band offsets between the Dirac point and conduction (valence) TMDC bands vary linearly with strain, regardless of the twist angle. This relation allows us to identify the apparent zero-strain band offsets and find a compensating transverse electric field correcting for the strain. The resulting corrected band structure is then fitted around the Dirac point to an established spin-orbit Hamiltonian. This procedure yields the dominant, valley-Zeeman, and Rashba spin-orbit couplings. The magnitudes of these couplings do not vary much with the twist angle, although the valley-Zeeman coupling vanishes for ${30}^{\circ }$ and Mo-based heterostructures exhibit a maximum of the coupling at around ${20}^{\circ }$. The maximum for W-based stacks is at $0^{\circ }$. The Rashba coupling is in general weaker than the valley-Zeeman coupling, except at angles close to ${30}^{\circ }$. We also identify the Rashba phase angle which measures the deviation of the in-plane spin texture from tangential, and find that this angle is very sensitive to the applied transverse electric field. We further discuss the reliability of the supercell approach with respect to atomic relaxation (rippling of graphene), relative lateral shifts of the atomic layers, and transverse electric field.

• Received 13 August 2021
• Revised 8 November 2021
• Accepted 10 November 2021

DOI:https://doi.org/10.1103/PhysRevB.104.195156