Unconventional superconductivity in magic-angle graphene superlattices

This Nature paper asks whether magic-angle twisted bilayer graphene (TBG)—a two-dimensional moiré superlattice formed by stacking two graphene sheets with a small twist angle—can host intrinsic unconventional superconductivity, and how its superconducting phase relates to the correlated insulating states known to occur near “magic” twist angles. The question matters because unconventional superconductivity (e.g., in cuprates and iron-based materials) has long resisted a unified theoretical explanation, and because TBG offers a uniquely tunable, clean, and purely carbon-based platform where carrier density can be tuned in situ with minimal disorder changes between measurements. If superconductivity in TBG is indeed tied to the correlated half-filled insulating state of the flat moiré bands, then TBG could become a controllable laboratory for studying mechanisms relevant to high- T_c superconductors and potentially related quantum phases such as spin liquids.

The authors realize magic-angle TBG devices by encapsulating twisted bilayer graphene between hexagonal boron nitride (h-BN) layers and contacting the stack with edge contacts. Twist angles are controlled to within roughly 0.1°–0.2° and are near the first magic angle (about 1.1°). Carrier density is tuned electrostatically using a metal gate beneath the bottom h-BN. The moiré superlattice produces ultra-flat bands near charge neutrality with bandwidth on the order of 5–10 meV. For twist angles near 1.1°, the flat bands yield correlated insulating behavior at half-filling of the lower flat band, corresponding to a density of approximately $n\approx \pm n_s/2$, where the superlattice density $n_s$ is defined from the moiré unit cell area. The paper focuses on transport signatures of superconductivity (zero resistance, critical current behavior, and magnetic-field response) and on normal-state properties (quantum oscillations and inferred Fermi surface size).

Methodologically, the study is experimental transport spectroscopy across multiple devices (notably two detailed devices: M1 with $\theta =1.16^{\circ }$ and M2 with $\theta =1.05^{\circ }$). The main measurements are four-probe longitudinal resistance $R_{xx}(n,T)$ at dilution-refrigerator base temperature around 70 mK, plus magnetic-field sweeps (perpendicular $B_{\perp }$ and parallel $B_{\parallel }$) to map superconducting domes and extract critical fields. To assess whether the superconducting transition is consistent with two-dimensional physics, the authors measure current–voltage characteristics $I$–$V$ and fit the low-bias scaling to a Berezinskii–Kosterlitz–Thouless (BKT) form, extracting a tentative $T_{\mathrm{BKT}}$ of about 1.0 K at $n\approx -1.44\times 10^{12}\mathrm{c}{\mathrm{m}}^{-2}$ in device M2. They also analyze the temperature dependence of the perpendicular critical field using Ginzburg–Landau theory to obtain a superconducting coherence length ${\xi }_{\mathrm{GL}}\approx 52\mathrm{nm}$ at $T=0$. For the in-plane critical field, they argue that Ginzburg–Landau thin-film orbital effects are insufficient and interpret the behavior as paramagnetic (Zeeman) pair breaking, extrapolating a zero-temperature in-plane critical field around 1.1 T and comparing it to an estimated Pauli limit $B_P\approx 0.93\mathrm{T}$ from a BCS gap estimate.

Key results are presented as a gate-tunable phase diagram with superconducting domes on both sides of the correlated insulating state. At base temperature, both devices show zero resistance in the superconducting regions. The highest reported critical temperature is $T_c=1.7\mathrm{K}$ in device M2 (with $T_c\approx 0.5\mathrm{K}$ in device M1, defined using the 50% normal-state resistance criterion). The superconductivity appears when the Fermi energy is tuned away from charge neutrality toward half-filling of the lower flat band (for $E_F<0$), and notably the authors do not observe comparable superconductivity when tuned into the flat conduction bands ($E_F>0$). The superconducting domes are separated from the correlated insulating phase at half-filling by a region where resistance rises as temperature is lowered (insulating behavior). At intermediate temperatures (roughly 1–4 K), the half-filling region shows insulating temperature dependence, but at the lowest temperatures superconducting behavior emerges, which the authors attribute to possible coexistence of superconducting and insulating phases due to sample inhomogeneity.

In magnetic field, superconductivity is suppressed by vortices and shows a maximum critical field around 70 mT (perpendicular). The critical field varies strongly with carrier density and forms domes aligned with the superconducting domes. Near the boundary between superconducting and insulating phases, the authors observe periodic oscillations in resistance and critical current as a function of $B_{\perp }$. They interpret these oscillations as phase-coherent Josephson transport through an inhomogeneous landscape—effectively a SQUID-like interference between superconducting regions separated by insulating islands. Using the oscillation period $\Delta B$ and the superconducting flux quantum ${\Phi }_0=h/2e$, they estimate effective loop areas of about $0.09\mu {\mathrm{m}}^2$ and $0.5\mu {\mathrm{m}}^2$ for two representative density cuts.

A central part of the paper is the normal-state quantum oscillation analysis, which probes the Fermi surface near the correlated insulating state. Using Shubnikov–de Haas oscillations in $R_{xx}$ at high perpendicular fields, the authors reconstruct Landau fans. In addition to the expected Landau fans associated with the charge neutrality point and the superlattice density $n=-n_s$, they find an anomalous Landau fan that emanates from the correlated insulating state at $n\approx -n_s/2$. This fan shows a sequence of halved filling factors (e.g., $-2,-4,-6,-8,\dots$) and an associated parameter $s=-1/2$, which they argue cannot be explained by single-particle Hofstadter butterfly physics (which would require much stronger magnetic fields) or by simple unit-cell doubling scenarios that would typically break spin or valley degeneracies.

From the oscillation frequency and Lifshitz–Kosevich fits, the authors infer small Fermi surfaces in the normal state just above the superconducting dome. They report that the superconducting dome corresponds to a record-low two-dimensional carrier density of approximately $n^{\prime }\approx 1.5\times 10^{11}\mathrm{c}{\mathrm{m}}^{-2}$ (their notation indicates that the effective carrier density contributing to the Fermi surface is $n^{\prime }=|n|-n_s/2$, rather than $|n|$ itself). The effective mass extracted from the temperature dependence of oscillation amplitudes is about $m^{\ast }\approx 0.2\text{–}0.4m_0$ for the anomalous oscillations near the correlated insulating region, compared with $\sim 0.1m_0$ near charge neutrality and $\sim 0.05m_0$ beyond the superlattice gap. The presence of small Fermi pockets not directly corresponding to the single-particle Fermi surface is compared to similar phenomenology in underdoped cuprates.

The authors then place these results in a strong-coupling context by comparing $T_c$ to characteristic Fermi temperatures $T_F$ and to a Bose–Einstein condensation temperature estimate $T_{\mathrm{BEC}}$. Using $n_{2D}\approx 0.15\times 10^{12}\mathrm{c}{\mathrm{m}}^{-2}$ and $m^{\ast }\approx 0.2m_0$ at optimal doping (and accounting for a halved degeneracy factor $g=1$), they estimate $T_c/T_{\mathrm{BEC}}$ up to about 0.37, which they interpret as evidence for very strong interactions and proximity to the BCS–BEC crossover. They also note that the coherence length $\xi \sim 50\mathrm{nm}$ is comparable to the average inter-particle spacing estimated from $n^{\prime }$, supporting the crossover interpretation.

Limitations are not exhaustively quantified, but several are evident from the methodology and discussion. First, the superconducting transition is not sharp: the superconductor–insulator evolution is gradual, making extraction of $T_c$ and critical fields somewhat uncertain. Second, the authors acknowledge that inhomogeneity likely plays a role in the coexistence of superconducting and insulating phases, which complicates interpretation of the phase diagram and the Josephson-like oscillations. Third, the paper provides transport-based evidence for superconductivity and for the nature of quasiparticles (via quantum oscillations), but it does not directly measure the superconducting pairing symmetry; the authors explicitly call for further experiments such as tunneling and Josephson heterostructures to determine pairing symmetry and to clarify whether the superconductivity is tied to the correlated half-filling state in a way analogous to cuprates.

Practically, the results establish magic-angle TBG as the first purely carbon-based two-dimensional superconductor and as a highly tunable platform for exploring strongly correlated phases. Researchers in condensed matter physics should care because the system provides a controllable route to study the interplay of flat bands, correlated insulating states, superconducting domes, and anomalous Fermi surface reconstruction. Materials scientists and device engineers should care because superconductivity is achieved with gate-tuned carrier densities on the order of $10^{12}\mathrm{c}{\mathrm{m}}^{-2}$ (the authors emphasize record-low densities compared with many other 2D superconductors), enabling systematic studies without ionic-liquid gating or chemical doping. The broader implication is that TBG may help identify which ingredients—strong correlations, reduced carrier density, proximity to Mott-like states, and possible frustration from moiré lattice geometry—are essential for unconventional superconductivity and potentially for related quantum spin liquid physics.