[Paper Review] MATBG as Topological Heavy Fermion: I. Exact Mapping and Correlated Insulators

Zhida Song, B. Andrei Bernevig|arXiv (Cornell University)|Nov 10, 2021

Graphene research and applications4 citations

TL;DR

This paper proposes a topological heavy fermion model for magic-angle twisted bilayer graphene (MATBG) that unifies localized flat-band states at AA-stacking regions with extended topological conduction bands, explaining both STM localization and transport delocalization. The model reproduces U(4) and U(4)×U(4) symmetries, predicts correlated insulator phases with quantized Chern numbers, and provides a first-principles explanation for the minimal gap at ΓM and large dispersion of charge-±1 excitations.

ABSTRACT

Magic-angle ($θ=1.05^\circ$) twisted bilayer graphene (MATBG) has shown two seemingly contradictory characters: the localization and quantum-dot-like behavior in STM experiments, and delocalization in transport experiments. We construct a model, which naturally captures the two aspects, from the Bistritzer-MacDonald (BM) model in a first principle spirit. A set of local flat-band orbitals ($f$) centered at the AA-stacking regions are responsible to the localization. A set of extended topological conduction bands ($c$), which are at small energetic separation from the local orbitals, are responsible to the delocalization and transport. The topological flat bands of the BM model appear as a result of the hybridization of $f$- and $c$-electrons. This model then provides a new perspective for the strong correlation physics, which is now described as strongly correlated $f$-electrons coupled to nearly free topological semimetallic $c$-electrons - we hence name our model as the topological heavy fermion model. Using this model, we obtain the U(4) and U(4)$ imes$U(4) symmetries as well as the correlated insulator phases and their energies. Simple rules for the ground states and their Chern numbers are derived. Moreover, features such as the large dispersion of the charge $\pm1$ excitations and the minima of the charge gap at the $Γ_M$ point can now, for the first time, be understood both qualitatively and quantitatively in a simple physical picture. Our mapping opens the prospect of using heavy-fermion physics machinery to the superconducting physics of MATBG.

Motivation & Objective

• To resolve the apparent contradiction between STM-localized and transport-delocalized behavior in MATBG.
• To construct a first-principles-compatible model that captures both flat-band localization and topological conduction bands.
• To explain the emergence of U(4) and U(4)×U(4) symmetries and correlated insulator phases in MATBG.
• To provide a physical picture for the minimal gap at the ΓM point and large dispersion of charge-±1 excitations.
• To establish a framework linking MATBG's strong correlation physics to heavy-fermion theory for future superconductivity studies.

Proposed method

• Introduce two sets of orbitals: local f-electrons centered at AA-stacking regions and extended c-electrons forming topological semi-metallic bands.
• Model the hybridization of f- and c-electrons to reproduce the topological flat bands of the Bistritzer-MacDonald model.
• Use a low-energy effective Hamiltonian combining strongly correlated f-electrons and nearly free c-electrons to define the topological heavy fermion model.
• Derive symmetry structures (U(4), U(4)×U(4)) from the f-c hybridization and spin-valley locking.
• Apply mean-field and perturbative techniques to analyze ground states, Chern numbers, and excitation spectra.
• Use position-dependent spectral analysis to connect c-electron dominance at AB-stacking sites to enhanced Landau level quantization and gap minima at ΓM.

Experimental results

Research questions

• RQ1 How can the coexistence of localized STM behavior and delocalized transport in MATBG be consistently explained within a single effective model?
• RQ2 What is the origin of the U(4) and U(4)×U(4) symmetries observed in MATBG's correlated phases?
• RQ3 Why does the charge gap exhibit a minimum at the ΓM point, and how does this relate to the underlying electronic structure?
• RQ4 How do the large dispersions of charge-±1 excitations arise from the f-c hybridization mechanism?
• RQ5 Why are Landau levels more readily observable at AB-stacking sites, and how does this reflect the c-electron character?

Key findings

• The model successfully reproduces the U(4) and U(4)×U(4) symmetries observed in prior studies, linking them to f-c hybridization and spin-valley locking.
• Correlated insulator phases with Chern numbers ±1 are predicted at fillings ν=±3, consistent with experimental observations under magnetic fields.
• The minimal gap at the ΓM point arises from c-electron contributions and is explained by the hybridization mechanism, not by symmetry alone.
• Charge-±1 excitations exhibit large dispersion due to the coupling of f-electrons to nearly free c-electrons, resolving a long-standing puzzle.
• Position-dependent STM spectra and Landau level quantization are naturally explained: AB-stacking sites show enhanced spectral weight at the minimal gap and clearer LLs due to c-electron delocalization.
• The model predicts a gapless spectrum at ν=±3 with realistic parameters, consistent with DMRG and ED studies, suggesting that the Chern insulator is metastable rather than the true ground state.