Magic-Angle Twisted Bilayer Graphene

• Magic-angle tBLG is a two-layer graphene system where a ~1.1° twist creates a moiré superlattice that produces nearly flat bands, driving correlated insulating states and superconductivity.
• Advanced spectroscopic and transport studies reveal suppressed Dirac velocities, van Hove singularities, and intricate topological features linked to the moiré geometry.
• Recent experiments using displacement-field tuning, heterostrain, and screening strategies are key to understanding and engineering the unconventional superconductivity and many-body phases in tBLG.

Magic-angle twisted bilayer graphene (tBLG) is a two-layer graphene system in which a small relative twist generates a moiré superlattice whose low-energy minibands become exceptionally narrow near the first magic-angle regime around $1.1^\circ$. In that regime, the Dirac velocity is strongly renormalized, kinetic energy is quenched, and the electronic structure becomes a platform for correlated insulating behavior, superconductivity, nematicity, orbital ferromagnetism, and topological transport. The subject is defined as much by the moiré geometry and its symmetry structure as by the many-body phases built on top of the flat bands (Utama et al., 2019, Andrei et al., 2020).

1. Moiré geometry, flat-band formation, and the meaning of the “magic angle”

Twisted bilayer graphene consists of two graphene monolayers rotated by a relative angle $\theta$. The moiré period follows

$L_m=\frac{a}{2\sin(\theta/2)},$

with $a\approx 0.246\,\text{nm}$, and for $\theta\ll 1$ in radians this reduces to $L_m\approx a/\theta$. As $\theta$ decreases, the moiré real-space cell expands into the tens-of-nanometers range and the mini Brillouin zone contracts correspondingly (Utama et al., 2019).

In the Bistritzer–MacDonald picture, twisting shifts the two monolayer Dirac cones in momentum space, while interlayer tunneling hybridizes them. Near the first magic angle, approximately $1.1^\circ$, the low-energy moiré bands become nearly flat and the Fermi velocity is strongly suppressed. This is the canonical single-particle origin of the strong-correlation regime in tBLG (Utama et al., 2019, Feraco et al., 2024).

Later ab initio-informed modeling revised the idealized “single magic angle with vanishing velocity” narrative. In the minimal $(2+2)$-band analysis based on the exact $k\cdot p$ model with relaxation, pseudogauge fields, and nonlocal interlayer tunneling corrections, the low-energy bands are never perfectly flat and the Fermi velocity never vanishes. Instead, a finite “magic range” appears. Using the shifted angle $\theta$0 with $\theta$1, the lower low-energy band is maximally flat in

$\theta$2

with midpoint features near $\theta$3 (Bennett et al., 2023).

This more realistic description is tied to atomic relaxation. Below about $\theta$4, and especially below a critical $\theta$5, the moiré texture is better understood as shrunken AA regions, enlarged AB/BA domains, and domain walls separating them. That real-space decomposition is not secondary: it controls which low-energy orbitals dominate and why nearby dispersive domain-wall-derived bands remain important even in the magic-angle regime (Bennett et al., 2023).

A practical implication is that magic-angle phenomenology is not confined to a single sharply tuned angle. Transport on a device at $\theta$6, about $\theta$7 below the nominal magic angle, still showed a Mott-like correlated insulator and superconductivity, motivating the language of a broader “magic range” rather than a single singular value (Codecido et al., 2019).

2. Direct electronic structure and spectroscopic visualization of the flat bands

A major milestone was the direct momentum-resolved observation of the flat moiré miniband by nanoARPES. In an uncapped tBLG on hBN on doped Si, local ARPES on a domain with

$\theta$8

revealed a sharp, weakly dispersing band near $\theta$9 around the graphene $L_m=\frac{a}{2\sin(\theta/2)},$0 points at room temperature. The key point was not merely a van Hove singularity in the density of states, but a direct momentum-space visualization of a flat miniband near charge neutrality (Utama et al., 2019).

NanoARPES was essential because conventional large-spot ARPES averages over twist-angle inhomogeneity, strain, reconstruction, and disorder. The experiment used a beam spot of about $L_m=\frac{a}{2\sin(\theta/2)},$1, photon energy $L_m=\frac{a}{2\sin(\theta/2)},$2, net energy resolution about $L_m=\frac{a}{2\sin(\theta/2)},$3, and UHV base pressure better than $L_m=\frac{a}{2\sin(\theta/2)},$4. The flat-band signal extended over about $L_m=\frac{a}{2\sin(\theta/2)},$5 along $L_m=\frac{a}{2\sin(\theta/2)},$6, which was interpreted as consistent with strong real-space localization of the associated states (Utama et al., 2019).

The same measurements showed that the low-energy feature is embedded in a richer reconstructed band structure. At higher binding energy, the spectra exhibit multiple hybridized Dirac cones, moiré-zone repetition, and avoided crossings or gaps produced by the periodic moiré potential. The comparison with spectral-function simulations based on an ab initio-informed tight-binding model at $L_m=\frac{a}{2\sin(\theta/2)},$7, including relaxation and band unfolding, reproduced the flat band at $L_m=\frac{a}{2\sin(\theta/2)},$8, its broad momentum-space extent, small outgoing branches, and multiple hybridized Dirac cones (Utama et al., 2019).

Local spectroscopy provided a complementary picture. In STM/STS at the magic angle, a single flat-band DOS peak appears when the band is full or empty, while at partial filling it reconstructs into lower-band and upper-band features separated by a pseudogap. In one study at $L_m=\frac{a}{2\sin(\theta/2)},$9, the lower-band–upper-band separation near charge neutrality was about $a\approx 0.246\,\text{nm}$0, and the $a\approx 0.246\,\text{nm}$1 signal at $a\approx 0.246\,\text{nm}$2 displayed dips near $a\approx 0.246\,\text{nm}$3, linking local spectroscopy directly to the correlated filling sequence (Jiang et al., 2019).

3. Correlated phases: insulating states, superconductivity, and symmetry breaking

The many-body importance of the flat bands appears most clearly in transport. In a device with $a\approx 0.246\,\text{nm}$4, the low-energy bandwidth remained $a\approx 0.246\,\text{nm}$5, comparable to magic-angle devices, and the system exhibited a Mott-like correlated insulating state at $a\approx 0.246\,\text{nm}$6 together with superconductivity on the electron-doped side for

$a\approx 0.246\,\text{nm}$7

with optimal doping near $a\approx 0.246\,\text{nm}$8. The reported superconducting transition scale was $a\approx 0.246\,\text{nm}$9–$\theta\ll 1$0, with $\theta\ll 1$1, showing that correlated and superconducting behavior survives well below the nominal first magic angle (Codecido et al., 2019).

That same work also found additional higher-filling structure: narrow activated resistance peaks at $\theta\ll 1$2 with activation gap $\theta\ll 1$3, and resistance peaks at $\theta\ll 1$4 attributed to high-energy Dirac points. This broadened the scope of magic-angle phenomenology beyond the first miniband and suggested that partial flatness in higher-energy bands can also generate correlation effects (Codecido et al., 2019).

A more explicit link between superconductivity and band structure was established by displacement-field tuning in a dual-gated near-magic-angle device at $\theta\ll 1$5. There, the filling factor was defined as

$\theta\ll 1$6

and the superconducting dome at zero displacement field occupied approximately $\theta\ll 1$7–$\theta\ll 1$8, with optimal doping $\theta\ll 1$9–$L_m\approx a/\theta$0. The BKT criterion $L_m\approx a/\theta$1 with $L_m\approx a/\theta$2 gave $L_m\approx a/\theta$3, and the perpendicular critical field near the strongest dome was $L_m\approx a/\theta$4–$L_m\approx a/\theta$5 (Dutta et al., 2024).

The displacement-field dependence showed a pronounced competition between superconductivity and symmetry-broken order. At $L_m\approx a/\theta$6, superconductivity appeared without a half-filling resistance peak. As $L_m\approx a/\theta$7 increased, superconductivity weakened, while a resistance peak emerged near $L_m\approx a/\theta$8 once $L_m\approx a/\theta$9. Hall-density analysis, using

$\theta$0

showed that superconductivity at $\theta$1 formed near a van Hove singularity around $\theta$2 in an isospin-unpolarized $\theta$3 regime, whereas larger $\theta$4 shifted the van Hove singularity toward $\theta$5 and drove a $\theta$6 symmetry-broken state near half-filling that suppressed superconductivity (Dutta et al., 2024).

Local spectroscopy revealed that the correlated normal state is itself ordered. At partial filling of the magic-angle flat band, STM/STS found a pseudogap phase accompanied by a global stripe charge order and broken rotational symmetry. Inside a single moiré cell, the lower-band and upper-band LDOS maps became elliptical with roughly orthogonal principal axes, and the inferred local filling formed a quadrupolar charge pattern. Over larger areas, those quadrupoles aligned into stripe order, lowering the approximate moiré $\theta$7 symmetry to $\theta$8 (Jiang et al., 2019).

4. Topology, symmetry, and low-energy effective descriptions

Flatness is not the whole story. In small-angle tBLG, the low-energy moiré Dirac bands also carry nontrivial topology protected by symmetry. Nonlocal transport in hBN-encapsulated devices with twist angles mainly between about $\theta$9 and $1.1^\circ$0 revealed pronounced nonlocal resistance peaks in both electron and hole superlattice gaps, while similar responses were absent in lower-angle samples near $1.1^\circ$1 and $1.1^\circ$2 where the relevant superlattice gaps closed. The key symmetry is

$1.1^\circ$3

which trivializes Berry curvature but leaves two $1.1^\circ$4 invariants for the isolated two-band moiré Dirac subspace (Ma et al., 2019).

One of those invariants is the quantized Berry phase

$1.1^\circ$5

which protects the moiré Dirac points. The other is encoded in the Wilson-loop or Wannier-center counterflow winding

$1.1^\circ$6

and implies one pair of counter-propagating edge states per spin and valley in each superlattice gap. Experimentally, the resulting nonlocal response followed

$1.1^\circ$7

with $1.1^\circ$8 and $1.1^\circ$9 at $(2+2)$0 (Ma et al., 2019).

At the model-building level, realistic tBLG is not captured adequately by a perfectly flat isolated two-band picture. The minimal $(2+2)$1-band model places two AA-derived orbitals and two domain-wall-derived dispersive orbitals on a honeycomb lattice and fits the low-energy bands with 13 physically motivated parameters that vary smoothly across the magic regime. Its exact Schur-complement reduction gives an effective two-band Hamiltonian,

$(2+2)$2

providing a compact starting point for Hubbard-like many-body modeling while retaining the influence of nearby dispersive bands (Bennett et al., 2023).

First-principles wavefunction calculations sharpened the real-space interpretation. Fully relaxed DFT down to $(2+2)$3 identified four characteristic wavefunction textures in the low-energy sector: AA-centered states forming a triangular lattice, ring-like AA$(2+2)$4 states, domain-wall states forming a Kagome lattice, and AB/BA states forming a honeycomb lattice. By tuning interlayer coupling, the calculations tracked the emergence of the flat bands, the associated band inversion, and a further likely topological phase transition at stronger coupling, where the upper and lower flat bands exchange mirror representations and wavefunction character. The transition was associated with $(2+2)$5 to $(2+2)$6 Å, corresponding to roughly $(2+2)$7–$(2+2)$8 GPa (Zhu et al., 4 Jul 2025).

5. Perturbations and band-structure engineering

Magic-angle tBLG is unusually susceptible to perturbations that are modest in other Dirac materials. One route is heterostrain. STM/STS on graphene bilayers grown on $(2+2)$9-Mo$k\cdot p$0C showed that a moderate heterostrain of only about $k\cdot p$1 can evolve a small-angle TGB from roughly $k\cdot p$2 with $k\cdot p$3 into a strained magic-angle TGB near $k\cdot p$4 with $k\cdot p$5, accompanied by the characteristic merger of two low-energy VHSs into a single flat-band peak. With still larger heterostrain, the system enters highly strained tiny-angle deformed tetragonal superlattices supporting topological helical domain-wall networks and localized domain-wall modes arranged into a hexagon-triangle-mixed frustrated lattice derived from Kagome geometry (Qiao et al., 2018).

Another route is substrate-induced spin-orbit coupling. In tBLG on a TMD, proximity-induced Rashba, Ising, and sublattice terms,

$k\cdot p$6

can reconstruct the eight near-magic-angle flat bands into a topological spin-orbit-coupled moiré manifold. For realistic values $k\cdot p$7, $k\cdot p$8, and $k\cdot p$9, the theory predicts valley Chern insulating regimes and, in even or one-sided structures, a time-reversal-protected topological insulator at $\theta$00 (Wang et al., 2020).

Charged defects provide a third form of local engineering. Tight-binding calculations with Coulomb impurities showed that in near-magic-angle tBLG the effect depends strongly on impurity position within the moiré cell. An impurity in the AA region can induce additional flattening of the low-energy bands; in the AB region it can open a gap at the moiré Dirac points by breaking the effective moiré sublattice symmetry; and in a bridge region it removes the Dirac points while leaving the system metallic through a band crossing along $\theta$01-to-$\theta$02 (Ramzan et al., 2022).

Magnetic fields can also reshape the single-particle structure. In an atomistic calculation at $\theta$03, a perpendicular field produced dispersive Landau levels when the magnetic length was comparable to the AA and AB region size, while a strong in-plane field modified the low-energy bands and gap through orbital minimal coupling. In the symmetric gauge, the layer-dependent momentum shift is

$\theta$04

and the effect becomes appreciable because the intrinsic moiré momentum scale is already small near the magic angle (Bigeard et al., 2023).

6. Debates, limitations, and current directions

Several central issues remain unresolved. One is the microscopic origin of superconductivity. Screening experiments point in different directions. In a device where magic-angle tBLG was placed $\theta$05 from a Bernal bilayer graphene screening layer, increasing screening weakened the correlated insulating states but enhanced superconductivity, with $\theta$06 reaching about $\theta$07 in the screened configuration; this was interpreted as evidence that electron-phonon coupling is the dominant pairing mechanism and Coulomb repulsion a competing interaction (Liu et al., 2020).

A later double-moiré screening experiment reached the opposite conclusion. There, a magic-angle TBG at $\theta$08 was placed only $\theta$09 from an electronically decoupled small-angle TBG screening layer with $\theta$10. Increasing the screening-layer density completely suppressed both the $\theta$11 correlated insulator and superconductivity in the adjacent magic-angle layer, which was interpreted as strong support for an unconventional, electronically mediated pairing mechanism (Barrier et al., 2024).

The state dependence of screening helps explain why these results are not trivially incompatible. A separate screening analysis found that interlayer coupling already enhances internal screening in magic-angle tBLG and that external dielectric engineering depends decisively on the electronic state. In a metallic RPA-like state, changing the external dielectric has little effect on the local interaction scale, whereas in a cRPA-like or insulating state it can alter the effective interaction by roughly $\theta$12–$\theta$13 (Pizarro et al., 2019). This suggests that screening experiments probe different interaction channels depending on geometry, layer separation, and the density of states of the screening medium.

A second unresolved issue is how literally the phrase “magic angle” should be taken. The Bistritzer–MacDonald paradigm remains the organizing principle, but realistic modeling and experiments below $\theta$14 support a broadened practical magic range rather than a single singular angle (Bennett et al., 2023, Codecido et al., 2019). A third issue concerns topology beyond the original fragile-band narrative. First-principles wavefunction work identified strong indicators of an additional topological transition under pressure or reduced angle, but did not compute topological invariants directly, so the precise classification remains open (Zhu et al., 4 Jul 2025).

Finally, there are important experimental limitations. Room-temperature nanoARPES directly established the flat-band platform, but it did not observe the low-temperature ordered phases themselves (Utama et al., 2019). Local probes reveal pseudogaps, charge order, and broken rotational symmetry, while transport reveals superconductivity and correlated insulators, yet a unified microscopic theory connecting flat-band topology, displacement-field band renormalization, screening environment, and pairing remains incomplete. What is firmly established is the platform: near the first magic-angle regime, tBLG realizes a moiré electronic structure in which single-particle flatness, topology, and externally tunable interaction scales become inseparable.