Alternating Twisted Trilayer Graphene

• Alternating twisted trilayer graphene is a mirror-symmetric trilayer system where outer layers are twisted in opposite directions relative to the middle layer, yielding a flat-band pair and a Dirac cone per valley.
• It inherits moiré flat-band physics from twisted bilayers while retaining a dispersive Dirac sector that can be tuned via displacement field, lattice relaxation, and symmetry breaking.
• This platform exhibits a rich phase diagram including superconductivity, orbital magnetism, quantum Hall effects, and Kekulé ordering, making it ideal for exploring correlated electronic phenomena.

Searching arXiv for papers on alternating twisted trilayer graphene and closely related mirror-symmetric twisted trilayer graphene. Alternating twisted trilayer graphene (ATTLG) is a trilayer graphene heterostructure in which the top and bottom layers are twisted by equal and opposite angles relative to the middle layer, forming the mirror-symmetric configuration $\theta_{TM}=-\theta_{MB}$; in the language of alternating twisted multilayer graphene, it is the $M=L=N=1$ case. Its defining low-energy feature is the coexistence, per spin and per valley, of one pair of flat bands and a monolayer-like Dirac cone near charge neutrality. That combination places ATTLG between twisted bilayer graphene and more general multi-moiré systems: it inherits moiré flat-band physics, but it also retains a dispersive sector that is tunable by displacement field, lattice relaxation, and symmetry breaking. Across the current literature, ATTLG therefore appears as a platform for superconductivity, orbital magnetism, Hofstadter and quantum Hall phenomena, Kekulé ordering, supermoiré interference, and incommensuration-driven topology (Xie et al., 2021, 1901.10485, Bhardwaj et al., 2024).

1. Geometry, moiré scales, and defining symmetries

ATTLG consists of three graphene layers in which the outer layers are twisted in opposite senses relative to the middle layer. Equivalent conventions appear across the literature: the top and bottom layers may be written as $\vartheta_\ell\in\{+\theta,0,-\theta\}$ relative to the middle layer, or as $\theta_1=+\theta/2$, $\theta_2=-\theta/2$, $\theta_3=+\theta/2$ so that the adjacent interfaces have relative twists $\pm\theta$. In the ideal alternating configuration, the outer layers are aligned to each other and the system is mirror symmetric about the middle layer. Time-reversal symmetry relates the valleys $K$ and $K'$, and in small-angle continuum descriptions a combined $C_{2z}T$ symmetry protects Dirac crossings. Inversion symmetry is generally absent once the alternating twists are applied, whereas mirror symmetry $M=L=N=1$0 is present when the top and bottom layers are related by reflection (Xie et al., 2021, Craig et al., 2023).

For small twist angle $M=L=N=1$1, the moiré length is

$M=L=N=1$2

with graphene lattice constant $M=L=N=1$3, and the moiré area is $M=L=N=1$4. The moiré filling factor is written as $M=L=N=1$5, or experimentally as $M=L=N=1$6 after calibrating the full-filling density $M=L=N=1$7 from Landau-fan analysis. In devices with $M=L=N=1$8, $M=L=N=1$9, and $\vartheta_\ell\in\{+\theta,0,-\theta\}$0, the reported $\vartheta_\ell\in\{+\theta,0,-\theta\}$1 values are approximately $\vartheta_\ell\in\{+\theta,0,-\theta\}$2, $\vartheta_\ell\in\{+\theta,0,-\theta\}$3, and $\vartheta_\ell\in\{+\theta,0,-\theta\}$4, respectively, consistent with $\vartheta_\ell\in\{+\theta,0,-\theta\}$5 (Bhardwaj et al., 2024).

The alternating branch is distinct from the helical or co-rotated branch. In general twisted trilayer graphene, the alternating condition corresponds to $\vartheta_\ell\in\{+\theta,0,-\theta\}$6, whereas the helical branch has $\vartheta_\ell\in\{+\theta,0,-\theta\}$7. This sign structure controls whether the two moiré patterns interfere into mirror-symmetric $\vartheta_\ell\in\{+\theta,0,-\theta\}$8-type domains or into the distinct domain structures characteristic of chiral trilayers. It also controls the size of the emergent supermoiré period, which is enlarged in the alternating geometry because the two moiré lattices tend to cancel in reciprocal space more efficiently than in the helical case (Nakatsuji et al., 2023, Xia et al., 3 Sep 2025).

2. Continuum descriptions and the magic-angle hierarchy

The standard continuum description of ATTLG is a generalized Bistritzer-MacDonald-type model in which each layer carries a rotated Dirac Hamiltonian and adjacent layers are coupled by three moiré harmonics. For a fixed valley $\vartheta_\ell\in\{+\theta,0,-\theta\}$9, the alternating twisted multilayer Hamiltonian can be written in block form with untwisted intrasequence Dirac sectors $\theta_1=+\theta/2$0 and intersequence moiré tunneling matrices $\theta_1=+\theta/2$1. In the trilayer specialization, a unitary transformation mixes the top and bottom layers into symmetric and antisymmetric combinations and reveals two decoupled sectors: a twisted-bilayer-like block with tunneling enhanced by $\theta_1=+\theta/2$2 and a free Dirac block. The resulting low-energy structure is therefore one flat-band pair plus one Dirac cone per spin and per valley (Xie et al., 2021, 1901.10485).

This decomposition produces a shifted magic-angle condition. In the chiral limit $\theta_1=+\theta/2$3, the effective coupling is

$\theta_1=+\theta/2$4

so the ATTLG magic angle is rescaled upward by $\theta_1=+\theta/2$5 relative to twisted bilayer graphene:

$\theta_1=+\theta/2$6

Within the exact hierarchy derived for alternating twisted multilayers, the trilayer singular value is $\theta_1=+\theta/2$7, giving $\theta_1=+\theta/2$8 and a first trilayer magic angle around $\theta_1=+\theta/2$9–$\theta_2=-\theta/2$0 when the bilayer value is $\theta_2=-\theta/2$1–$\theta_2=-\theta/2$2. Away from the chiral limit, finite $\theta_2=-\theta/2$3 broadens the flat bands and introduces particle-hole asymmetry, but the block structure and the coexistence with a Dirac cone remain (1901.10485, Xie et al., 2021).

A complementary optical formulation reaches the same partitioning from the response side. In mirror-symmetric ATTLG, a unitary transformation maps the trilayer Hamiltonian to an effective twisted bilayer sector plus a single-layer sector, and the layer-resolved optical conductivity decomposes into twisted-bilayer, single-layer, and coupling contributions. The in-plane magnetic response is proportional to the coupling term, $\theta_2=-\theta/2$4, and is found to be negligibly small because of the energy-scale mismatch between the flat-band and Dirac-like sectors (Margetis et al., 2024).

3. Lattice relaxation, supermoiré interference, and local stacking structure

A central feature of ATTLG is that its electronic structure is not set only by the nominal twist angles. Multi-scale lattice relaxation reorganizes the trilayer into domains that are locally closer to particular stacking motifs. In the alternating case with $\theta_2=-\theta/2$5 and $\theta_2=-\theta/2$6, the two moiré patterns are rotated by $\theta_2=-\theta/2$7 relative to one another, and their interference generates a super-long-range moiré-of-moiré structure. After relaxation, alternating trilayers form $\theta_2=-\theta/2$8 domains in which the AA spots of the two moirés overlap, locally reproducing the mirror-symmetric twisted trilayer configuration. This local restoration of mirror symmetry explains why alternating trilayers near $\theta_2=-\theta/2$9 exhibit coexisting flat bands and a monolayer-like Dirac cone, whereas chiral trilayers can instead produce domain-wall-localized one-dimensional bands and wide low-density-of-states windows (Nakatsuji et al., 2023).

Direct structural imaging by interferometric 4D-STEM has refined this picture. In near-magic alternating twisted trilayers, AAA regions contract substantially relative to rigid models, and AB/BA-like domains expand. The energetic hierarchy inferred from the reconstructed stacks is strong: in the AtA geometry, AAA stacking is $\theta_3=+\theta/2$0/unit cell higher in energy than A-SP-A and $\theta_3=+\theta/2$1/unit cell higher than ABA, while in co-rotated tAB trilayers AAB is $\theta_3=+\theta/2$2/unit cell higher than ABC and BAB. These measurements imply a reconstruction-driven reduction of the effective $\theta_3=+\theta/2$3 ratio and show that global $\theta_3=+\theta/2$4 is degraded by reconstruction into non-inversion stackings even though local AAA pockets retain approximate $\theta_3=+\theta/2$5. The same 4D-STEM work also found sizable AtB regions persisting down to $\theta_3=+\theta/2$6, revising earlier assumptions that slightly misaligned alternating trilayers relax almost entirely into AtA domains (Craig et al., 2023).

Atomistic tight-binding plus molecular-dynamics studies of double-moiré interference reach a related conclusion from the electronic side. In the symmetric alternating case $\theta_3=+\theta/2$7, the supermoiré period grows as $\theta_3=+\theta/2$8 at small angle. For rigid structures, two van Hove singularities near charge neutrality approach one another but remain split, with a minimum gap of approximately $\theta_3=+\theta/2$9 near $\pm\theta$0. After relaxation, however, the same alternating geometry exhibits van Hove singularity merger at charge neutrality at $\pm\theta$1. The same framework also shows that a slight angle disorder $\pm\theta$2 can strongly suppress van Hove singularity peaks in rigid structures, whereas relaxation protects the positions, widths, and amplitudes more effectively (Meng et al., 2022).

4. Spectroscopic, optical, and phononic signatures

The spectroscopic phenomenology of twisted trilayers was first established in a non-alternating geometry. Scanning tunneling microscopy and spectroscopy on a co-rotated trilayer with $\pm\theta$3 and $\pm\theta$4 resolved two moiré patterns, two sets of van Hove singularities, and a much larger interference pattern associated with $\pm\theta$5. Because the inferred outer-layer relative angle equaled $\pm\theta$6 rather than $\pm\theta$7, that sample was not an alternating $\pm\theta$8 trilayer. Even so, it established several diagnostics later used in ATTLG: two-interface spectral superposition, low-energy van Hove singularities close to the Fermi level, splitting of each van Hove singularity by $\pm\theta$9, and real-space symmetry breaking of the local density of states near the singularities, consistent with enhanced electron-electron interactions (Zuo et al., 2017).

In alternating geometries, transport-based spectroscopy and atomistic calculations identify both constructive and destructive double-moiré interference. The rigid symmetric alternating case shows a decreasing then increasing van Hove singularity separation with a minimum near $K$0, while relaxed structures display van Hove singularity merging at the charge-neutrality point around $K$1. Real-space local density-of-states patterns can become strongly layer selective: in some regimes the middle layer carries the most pronounced supermoiré modulation, whereas in others the localization is dominated by the larger-period interface. The same calculations predict interference-generated superstructures, including a Kagome-like network for $K$2 and $K$3 (Meng et al., 2022).

The optical response inherits the TBG-plus-Dirac decomposition. In mirror-symmetric ATTLG, the layer-resolved conductivity can be written in terms of effective twisted-bilayer and single-layer conductivities together with a coupling term, while global optical activity vanishes at normal incidence because of mirror symmetry. The remaining magneto-electric response is local rather than global: it couples the in-plane electric field to vertical gradients of magnetic field and magnetic moment through the chiral interlayer conductivity of the effective twisted-bilayer sector. Because the in-plane magnetic susceptibility is proportional to the small coupling conductivity $K$4, it is predicted to be negligibly small compared with the large in-plane orbital response known in twisted bilayers (Margetis et al., 2024).

The lattice sector adds further ATTLG-specific signatures. Mirror-symmetric twisted trilayer graphene supports, in addition to ordinary acoustic phonons, two shear-mode classes distinguished by mirror parity. The mirror-even modes are gapless phasons and are equivalent to twisted-bilayer phasons after the adhesion amplitude is rescaled by $K$5. The mirror-odd shear modes have no analogue in twisted bilayers and are gapped, with

$K$6

where $K$7 measures the degree of lattice relaxation. Across the experimentally relevant angle range, the reported odd-mode gap decreases from $K$8 to $K$9. A perpendicular displacement field mixes the even and odd sectors and reduces the odd-mode gap, while broken time-reversal symmetry can endow the phonon branches with finite angular momentum through a Hall-viscosity term (Samajdar et al., 2022).

5. Superconductivity, orbital magnetism, and interaction-driven ordering

The most direct experimental demonstration of competing broken symmetries in ATTLG comes from intermediate-angle devices with $K'$0–$K'$1. In this regime, orbital magnetism is strongest near the charge-neutrality point on both electron and hole sides, while superconductivity is suppressed close to neutrality and emerges at larger moiré fillings. The competition is also displacement-field dependent: increasing $K'$2 suppresses orbital magnetism and enhances superconductivity. In a device at $K'$3, the strongest superconducting pocket occurs on the electron-doped side with $K'$4 at $K'$5 and $K'$6, whereas orbital magnetism peaks near $K'$7 and fades for $K'$8. The inferred orbital ferromagnetic ordering temperature is $K'$9, about half $C_{2z}T$0 (Bhardwaj et al., 2024).

Gate-defined Josephson junctions provide a phase-sensitive probe of that competition. In the same devices, narrow electrostatic links between adjacent Hall bars can be tuned into superconducting, normal, or orbital-magnetic weak-link states while the reservoirs remain superconducting at $C_{2z}T$1. The superconducting-normal-superconducting configuration exhibits a symmetric Fraunhofer pattern with high-field periodicity of about $C_{2z}T$2 and a central lobe width of about $C_{2z}T$3, consistent with $C_{2z}T$4 for $C_{2z}T$5. By contrast, superconducting-orbital-magnetic-superconducting junctions show pronounced asymmetry, $C_{2z}T$6, and the asymmetry vanishes near $C_{2z}T$7. Its temperature dependence is consistent with a Curie-Bloch form, $C_{2z}T$8, with $C_{2z}T$9 and $M=L=N=1$00. The same temperature scale governs a superconducting-diode-like nonreciprocity at $M=L=N=1$01, linking the Josephson asymmetry directly to the magnetic weak link (Bhardwaj et al., 2024).

The microscopic interpretation advanced for this magnetic phase is orbital, not spin, ferromagnetism. The Hall-slope jumps appear only for out-of-plane field $M=L=N=1$02, not for in-plane $M=L=N=1$03, and remain unchanged up to $M=L=N=1$04. The proposed origin is spontaneous valley polarization near charge neutrality, which generates an orbital magnetic moment tied to valley index and Berry curvature. That identification differentiates the phase from the anomalous Hall effect widely discussed at integer fillings in other twisted graphene systems: in the ATTLG experiment, the strongest magnetism occurs at fractional fillings near $M=L=N=1$05, is tunably suppressed by displacement field, and is diagnosed by asymmetric Fraunhofer interference rather than by quantized anomalous Hall transport (Bhardwaj et al., 2024).

Additional interaction-driven phases have been proposed in the same mirror-symmetric band structure. Mean-field calculations at $M=L=N=1$06 find robust incommensurate Kekulé spiral order under experimentally reasonable heterostrain and, at zero strain but large interlayer potential $M=L=N=1$07, a commensurate Kekulé spiral that is absent in twisted bilayer graphene. The same analysis identifies charge-transfer cascades between the mirror-even TBG-like sector and the mirror-odd graphene-like sector, so that the most insulating densities need not coincide with integer total filling. Earlier work on heavy-fermion emulation in twisted trilayers further showed that interlayer bias can sharpen the distinction between localized flat-band modes and dispersive valley-helical modes, yielding a Kondo-lattice description with $M=L=N=1$08, hybridization $M=L=N=1$09, and a bias-tunable exchange that changes sign near $M=L=N=1$10 (Wang et al., 2023, Ramires et al., 2021).

6. High-field, large-angle, and incommensurate extensions

In high magnetic field, mirror-symmetric twisted trilayer graphene decomposes into even-parity twisted-bilayer-like Hofstadter bands and odd-parity Dirac-like bands. That parity structure enables a quantum parity Hall state in which mirror symmetry protects counter-propagating edge channels belonging to different parity sectors. The proposed transport signature is simultaneous quantization of Hall and longitudinal resistances; for one representative edge-channel configuration with $M=L=N=1$11 and $M=L=N=1$12, the six-terminal values are $M=L=N=1$13, $M=L=N=1$14, and $M=L=N=1$15. A perpendicular displacement field hybridizes the parity sectors and creates a weakly dispersive zero-energy band localized on the middle layer, isolated from the rest of the Hofstadter spectrum by a gap that scales approximately as $M=L=N=1$16 (Imran et al., 2022).

The alternating geometry also supports a qualitatively different large-angle regime. At $M=L=N=1$17, moiré reconstruction is negligible, the large momentum mismatch suppresses coherent single-particle tunneling, and the low-energy structure is effectively three weakly hybridized Dirac-like bands rather than flat moiré minibands. Magnetotransport in that regime reveals three low-resistance states along $M=L=N=1$18, identified by Hartree-Fock analysis with the layer fillings $M=L=N=1$19, $M=L=N=1$20, and $M=L=N=1$21 and interpreted as spin-resolved helical edge transport. At $M=L=N=1$22 and $M=L=N=1$23, a narrow resistance dip appears when the middle and bottom layers are each half filled while the top layer is inert at $M=L=N=1$24, consistent with an interlayer excitonic quantum Hall state (Kim et al., 13 Sep 2025).

Later work has emphasized that ATTLG need not be confined to isolated magic-angle points. In the multi-moiré perspective, correlated behavior can organize along continuous branches in the two-angle plane. On the alternating branch, where $M=L=N=1$25, the supermoiré period is enlarged and moiré quasicrystals can support superconductivity with $M=L=N=1$26 and Berezinskii-Kosterlitz-Thouless scales such as $M=L=N=1$27 and $M=L=N=1$28 in representative devices, while polycrystalline relaxation near simple angle ratios can favor anomalous Hall behavior in topological domains (Xia et al., 3 Sep 2025). In a distinct theoretical extension, intrinsic incommensuration has been proposed as a stabilizer of fractional Chern insulators in ATTLG: in a patchwise supermoiré description with outer-layer hBN sublattice potential $M=L=N=1$29 and filling $M=L=N=1$30 in one valley, strong-incommensuration patches host Laughlin-like phases with many-body gaps of order $M=L=N=1$31 even as conventional quantum-geometric indicators worsen, because incommensuration suppresses competing charge-density-wave order more strongly than it suppresses topological order (Song et al., 27 Nov 2025).

Taken together, these results define ATTLG as a structurally simple but electronically nontrivial moiré platform. Its mirror-symmetric alternating geometry yields a flat-band sector and a Dirac sector in the same low-energy window; relaxation converts that symmetry into a domain-wise, multi-scale object; and electrostatic tuning of density, displacement field, and weak-link geometry can move the system among superconducting, orbital-magnetic, excitonic, Hofstadter, and incommensurate topological regimes. The remaining open problems in the literature concern the microscopic energetics of valley polarization near charge neutrality, the quantitative role of strain and domain dynamics, the universality of competing orbital magnetism and superconductivity across angle and disorder windows, and the extent to which the supermoiré description can unify the small-angle, intermediate-angle, and large-angle regimes within a single framework (Bhardwaj et al., 2024, Nakatsuji et al., 2023, Xia et al., 3 Sep 2025).