Condensed Matter
Moiré Superlattices: How a 1.1° Twist Turns Graphene Into a Superconductor
Stack one sheet of graphene on another and rotate it by exactly 1.1° — the so-called magic angle — and something absurd happens: the material's electrons slow almost to a halt, their kinetic energy collapses to a few meV, and below about 1.7 K the whole thing becomes a superconductor. No new atoms, no chemistry, just a rotational misalignment thinner than the width of a human hair reproduced across a lattice spacing of 0.246 nm.
The trick is a moiré pattern: two nearly-identical lattices overlaid at a small angle produce a giant beat pattern with a period thousands of times larger than the atoms themselves. That long-wavelength superlattice folds graphene's band structure into a tiny Brillouin zone, and at the magic angle it flattens the bands so completely that electron–electron interactions — normally a minor perturbation — take over and dictate the physics.
- Magic angleθ ≈ 1.1°
- Moiré periodL = a / (2 sin(θ/2)) ≈ 13 nm
- Flat-band width≲ 5–10 meV
- T_c (superconducting)≈ 1.7 K
- Predicted / observed2011 / 2018
- Graphene lattice a0.246 nm
Interactive visualization
Press play, or step through manually. The visualization is yours to drive — try it before reading on.
Watch the 60-second explainer
A condensed visual walkthrough — narrated, captioned, under a minute.
The moiré pattern: geometry that magnifies the lattice
A moiré pattern is what you see when two periodic structures overlap slightly out of register — the interference of two window screens, or two combs. For two identical lattices of period a rotated by a small angle θ, the emergent superlattice period is:
- L = a / (2 sin(θ/2)) ≈ a/θ (for small θ in radians)
Plug in graphene's lattice constant a = 0.246 nm and the magic angle θ = 1.1° = 0.0192 rad: L ≈ 0.246 nm / 0.0192 ≈ 13 nm. That is a superlattice roughly 53 times larger than the underlying carbon spacing, containing on the order of 10,000 carbon atoms per moiré unit cell. The moiré acts as a geometric magnifier — a small angular mismatch θ produces a huge real-space period, and the smaller θ, the larger L, diverging as θ → 0.
Locally the pattern cycles through stacking registries: regions of AA stacking (atoms directly above atoms), AB and BA (Bernal) stacking (offset by one bond), arranged in a triangular superlattice. This periodic modulation of the interlayer coupling is the potential landscape that reshapes the electrons.
From Dirac cones to a tiny Brillouin zone
Isolated graphene is a semimetal: near the corners (K, K′) of its hexagonal Brillouin zone the electrons obey a 2D Dirac equation, with energy linear in momentum, E = ± ℏ v_F |k|, and a Fermi velocity v_F ≈ 1 × 10⁶ m/s — about c/300. There is no band gap and no natural energy scale; the physics is essentially non-interacting.
Twisting two layers shifts each layer's Dirac point in momentum space. The moiré Brillouin zone is a small hexagon whose size scales as the momentum mismatch:
- k_θ = 2 k_D sin(θ/2) ≈ K·θ, where k_D = 4π/(3a) is graphene's Dirac momentum.
Because L is large, this mini-Brillouin zone is tiny — for θ ≈ 1.1°, k_θ ≈ 0.033 Å⁻¹, roughly 2% of the original zone. The two layers' Dirac cones now sit close together and hybridize through interlayer tunneling, folding the bands back into the mini-zone and opening a cascade of hybridization gaps.
The magic angle and the Bistritzer–MacDonald model
The quantitative theory was published by Rafi Bistritzer and Allan MacDonald in 2011 (PNAS). Their continuum model treats each layer as a Dirac Hamiltonian and couples them with a periodic interlayer tunneling matrix of amplitude w ≈ 110 meV. The physics collapses onto a single dimensionless parameter:
- α = w / (ℏ v_F k_θ) — the ratio of interlayer tunneling to the intralayer kinetic energy across the mini-zone.
As θ decreases, k_θ shrinks, the kinetic scale ℏ v_F k_θ shrinks, and α grows. At a critical value α ≈ 0.586 the central bands become flat — the renormalized Fermi velocity v_F* is driven essentially to zero. This first magic condition corresponds to θ ≈ 1.05–1.1°. There is a discrete sequence of smaller magic angles (≈ 0.5°, 0.35°, …), but the first is the celebrated one.
Flatness has a dramatic meaning: a flat band means the electron's group velocity v = (1/ℏ) ∂E/∂k ≈ 0. The electrons carry almost no kinetic energy. The bandwidth W of these flat bands drops to just a few meV — compare to the full graphene bandwidth of ~18 eV, a suppression by more than three orders of magnitude.
Why flat bands unleash correlations
In an ordinary metal the electronic kinetic energy (bandwidth W) far exceeds the Coulomb repulsion U between electrons, so interactions are a small correction and band theory works. The magic angle inverts this. With W squeezed to a few meV, the Coulomb energy of two electrons confined to one moiré cell, roughly:
- U ≈ e² / (4π ε₀ ε_r L) ≈ 10–20 meV for L ≈ 13 nm and dielectric screening ε_r ~ 4–10,
becomes comparable to or larger than W. When U ≳ W, the system enters the strongly correlated regime: electrons can no longer be treated independently, and the physics resembles that of the cuprate high-T_c superconductors and heavy-fermion materials. The flat band acts like a nearly dispersionless atomic level whose degeneracy interactions must lift — precisely the setting for Mott insulators, magnetism, and unconventional superconductivity.
A useful mental picture: kinetic energy wants electrons to delocalize (be waves); Coulomb repulsion wants them to localize and stay apart (be particles). At the magic angle the tie is broken in favor of interactions, and the ground state is selected by the subtle interplay of spin, valley, and moiré-band degrees of freedom.
The 2018 experiments: correlated insulators and superconductivity
In March 2018 Pablo Jarillo-Herrero's group at MIT (Cao et al., two back-to-back Nature papers) reported the smoking guns. Cooling magic-angle twisted bilayer graphene and tuning the carrier density with a gate voltage, they found:
- Correlated insulating states at commensurate fillings — notably at ±2 electrons per moiré cell (half-filling) — where band theory predicts a metal but the sample instead becomes insulating, a hallmark of Mott-like physics.
- Superconductivity emerging on doping away from those insulators, with a critical temperature T_c ≈ 1.7 K at carrier densities around n ≈ 1.5 × 10¹¹ cm⁻².
The phase diagram — superconducting domes flanking an insulator, tuned by carrier density — looks strikingly like that of the copper-oxide high-temperature superconductors, but here with everything tunable in situ by a knob (the gate) rather than by chemical substitution. The T_c is low in absolute terms, but the ratio T_c / T_F (Fermi temperature) is enormous — of order 0.1 — placing TBG among the most strongly-coupled superconductors known.
Twistronics: a tunable platform, and its subtleties
This whole field, dubbed twistronics, turns twist angle into a materials-design parameter as powerful as chemical composition. Beyond bilayer graphene, moiré flat bands and correlated states appear in twisted trilayer graphene (higher T_c, ~2.9 K), twisted transition-metal dichalcogenides (WSe₂, MoTe₂ — where fractional quantum anomalous Hall states have been seen), and graphene aligned to hexagonal boron nitride (which gives Hofstadter's-butterfly fractal spectra in a magnetic field).
Two subtleties are essential to state correctly:
- The magic angle is not a resonance you can be sloppy about. Device performance is exquisitely sensitive: a mismatch of even 0.1° destroys the flat bands. Real samples also relax structurally — the lattice reconstructs into sharp AB/BA domains separated by narrow AA regions rather than a smooth moiré, which shifts the effective magic angle.
- Moiré is not a real crystal. The superlattice period L is incommensurate with the atomic lattice in general; it is a slowly-varying effective potential, not a periodic array of real potential wells. The 'unit cell' is a coarse-grained continuum object, which is exactly why the Bistritzer–MacDonald continuum theory works.
Finally, a common misconception: the electrons are not literally trapped in the AA regions like marbles in a box. The flatness is a band-structure effect from destructive hybridization of Dirac cones, and while the wavefunction weight does concentrate near AA sites, the bands remain genuinely delocalized moiré Bloch states.
| Property | Single-layer graphene | Magic-angle TBG (θ ≈ 1.1°) |
|---|---|---|
| Real-space period | a = 0.246 nm | L ≈ 13 nm (moiré) |
| Band dispersion | Linear (Dirac cone) | Flat bands (~few meV wide) |
| Fermi velocity | v_F ≈ 1 × 10⁶ m/s | → near 0 at magic angle |
| Dominant physics | Non-interacting (kinetic) | Strongly correlated (interactions) |
| Ground states | Semimetal | Superconductor, correlated insulator |
| Bandwidth W vs Coulomb U | W ≫ U | W ≲ U (Mott-like) |
Frequently asked questions
Why is 1.1° the 'magic' angle and not some other value?
It is the angle at which the dimensionless ratio α = w/(ℏv_F k_θ) reaches ≈ 0.586, the value that drives the renormalized Fermi velocity of the central bands to zero. Since k_θ ∝ θ, decreasing the twist increases α; the first flat-band condition falls at θ ≈ 1.05–1.1° for graphene's parameters (w ≈ 110 meV, v_F ≈ 10⁶ m/s). There is actually a sequence of smaller magic angles (~0.5°, 0.35°…), but the first is by far the most robust and experimentally accessible.
How can twisting two ordinary sheets create a superconductor with no new atoms?
The twist doesn't change the chemistry — it changes the band structure. By flattening the bands to a few meV, it suppresses electron kinetic energy until the Coulomb interaction U (~10–20 meV per moiré cell) dominates. In that strongly-correlated regime the electrons organize into collective quantum states, including a superconducting condensate below T_c ≈ 1.7 K, much as they do in the cuprates.
How big is a moiré unit cell compared to the atoms?
For θ ≈ 1.1° the moiré period is L = a/(2 sin(θ/2)) ≈ 13 nm, about 53 times the 0.246 nm carbon lattice constant. Each moiré cell contains on the order of 10,000 carbon atoms. The pattern is a geometric magnifier: halving the angle roughly doubles L, which diverges as θ → 0.
Is the moiré superlattice a real crystal with real potential wells?
Not in the usual sense. The moiré is an interference pattern of two incommensurate atomic lattices — a slowly varying effective potential, not an array of physical wells with trapped electrons. This is why a continuum theory (the Bistritzer–MacDonald model) describes it so well: the relevant length scale L is thousands of times the atomic spacing, so the atomic detail averages out.
Why is magic-angle graphene so hard to make?
The flat bands exist only in a razor-thin window around 1.1° — a deviation of just 0.1° washes them out. Fabricators use the 'tear-and-stack' method to control the relative angle, but the layers also structurally relax into AB/BA domains, and twist-angle disorder across a device can smear the correlated states. Reproducibility of the magic angle remains a central experimental challenge.
How does this connect to high-temperature superconductivity?
The phase diagram is analogous: correlated (Mott-like) insulating states at integer moiré fillings, flanked by superconducting domes as you dope away from them, all controlled by carrier density. Because everything is set by a gate voltage rather than chemical doping, TBG is a clean, tunable model system for studying the interplay of correlations, magnetism, and unconventional pairing that underlies cuprate physics.