Stochastic many-body calculations of moir\'e states in twisted bilayer graphene at high pressures

Abstract

Atomistic first principles many-body computational studies were so far limited by the system size. In this work, we apply and expand the stochastic GW method allowing calculations of quasiparticle energies of giant systems. We introduce three new developments to the stochastic many-body formalism: efficient evaluation of off-diagonal self-energy terms, construction of Dyson orbitals, and introduce stochastic cRPA, enabling efficient downfolding onto model Hamiltonians. We exemplify the new methodology by exploring twisted bilayer graphene (tBLG); its moire periodicity is associated with giant unit-cells hosting correlated electrons in flat bands. The resulting behavior of tBLG is governed by the coupling between the weakly and strongly correlated electrons in individual monolayers, controlled by twist or applied pressure. Here we study the latter scenario. The calculations document the formation of charge carrier localization at the Fermi level while most weakly correlated states are merely shifted in energy. This contrasts with the mean-field results. We show how to efficiently downfold the correlated subspace on a model Hamiltonian, namely the Hubbard model with a screened frequency dependent on-site interaction, computed with newly developed stochastic cRPA. For the $6^{\circ}$ tBLG system, the onsite interactions range between 200 and 300 meV under compression. The Dyson orbitals exhibit spatial distribution similar to the mean-field single particle states. Under pressure, the electron-electron interactions in the localized states increase; the dynamical screening does not fully compensate the dominant bare Coulomb interaction. The correlated moir\'e states thus appear only moderately screened at high pressures suggesting the presence of insulating states in compressed tBLG.