Integer doping in reconstructed graphene is attributed to strain-induced pseudo-Landau levels (PLLs) localized at the vertices of moiré domains. Scanning tunnelling microscopy (STM), molecular-dynamics (MD) calculations and density-functional-theory (DFT) modelling show that the AB and AA configurations develop different strain patterns and pseudomagnetic fields (PMFs). These fields create localized, nearly flat bands whose interaction with local work-function variations releases holes in AB regions and electrons in AA regions, producing the observed integer charge transfer per moiré cell.
Strain reconstruction and pseudomagnetic fields
The reconstructed AB and AA configurations were directly visualized using STM. STM images of different regions, together with a reference unidirectional graphene/hBN region on the same flake, are shown in Supplementary Figs. 12 and 13. The structures were also reproduced in MD calculations (Fig. 5a,b). The calculated atomic displacements were used to determine the strain distribution in the reconstructed moiré unit cell and the associated PMF.7,8,9,10,11,34
The two configurations produce distinct strain fields. In the AB configuration, AB-1 and AB-3 stacking have substantially lower energy than AB-2. Reconstruction therefore produces triangular AB-1 and AB-3 domains, while AB-2 develops strong distortions and out-of-plane corrugation. In the AA configuration, AA-3 has the lowest energy, producing hexagonal AA-3 moiré domains. Strain is concentrated at the AA-1 and AA-2 corners of the moiré plaquette.
The predominance of in-plane strain in AA and out-of-plane deformation in AB follows from the mirror symmetry of the AA configuration, which produces a negligible net out-of-plane force. In both configurations, however, the structural distortion generates a strong PMF.
Panels a and b compare STM images of AB and AA regions with corresponding MD simulations, confirming strong reconstruction into an AB triangle and an AA hexagon. Panels c and d show the calculated strain distributions. In AB, strain is dominated by out-of-plane deformation and is strongest at AB-2; in AA, in-plane forces compress AA-1 and AA-2 and expand AA-3. Panels e and f show the PMF and work-function landscapes, with one pinch point in AB and two in AA. The work-function contrast between the pinch point and the bulk region has opposite signs in the two configurations. Panels g–j show calculated band structures without and with work-function variations. The localized flat bands have bandwidths of 24 meV for AB and 17 meV for AA. Including the work-function variation shifts the flat band downward in AB and upward in AA. Panels k and l show ARPES spectra and energy-distribution curves for dual-oriented mhBN/graphene/hBN and reference graphene/hBN regions. The scale bar is 0.02 Å−1; au denotes arbitrary units.
Localized flat bands and work-function variations
The reconstruction affects the band structure through two main components: the PMF and local work-function changes across the moiré plaquette.35 When the work-function variation is initially excluded, the difference between the on-site potentials of the two carbon sublattices opens a substantial band gap.35,36,37 Several fourfold spin- and valley-degenerate flat bands also appear inside the gap.
The corresponding real-space wavefunctions are localized at the AB-2, AA-1 and AA-2 vertices, where the strain and PMF are strongest. These states can be understood as PLLs generated by the PMF, or as strain-induced quantum dots at the apices of the moiré regions. In reciprocal space, they appear as flat bands containing strongly localized electrons.
The zeroth PLL in graphene lies at the neutral point and is initially half-filled. Its energy and doping cannot change unless it hybridizes with either the valence or conduction band. Work-function variations across the moiré plaquette provide that hybridization.
At the curved AB-2 site, hybridization is reduced for the carbon pz orbitals, in a manner similar to the orbitals in carbon nanotubes.38 The local work function increases, shifting the corresponding flat band downward. In contrast, in-plane strain at the AA-1 and AA-2 vertices reduces the distance between carbon atoms, lowering the local work function and shifting the associated flat bands upward.39 The van der Waals environment and other effects contribute to the same trends.35,37
How the reconstruction produces p-type and n-type doping
In AB, the work-function change moves the flat band toward the valence band. In AA, it moves the flat band toward the conduction band. When the initially empty or filled flat band in the AB or AA configuration crosses the corresponding dispersive band, holes or electrons trapped in the quantum dots are released into the conducting part of the moiré domain.
This redistribution shifts the Fermi level from the neutral point toward the valence band in AB and toward the conduction band in AA. The two configurations therefore produce p-doping and n-doping, respectively. In this picture, charge is redistributed between a conducting moiré domain and an insulating quantum dot at its vertex.
Hole-side angle-resolved photoelectron spectroscopy (ARPES) supports this interpretation. Compared with the conventional graphene/hBN reference region on the same flake, the mhBN/graphene/hBN heterostructure shows a flat band only in the mhBN region. That band hybridizes with an adjacent dispersive band, as expected from the calculated band structure.
Integer charge per moiré plaquette
In the simplest model, each PLL has a degeneracy of g = gsgv = 4, from spin and valley degeneracy. In the AB configuration, only one vertex per plaquette satisfies the required condition. Because the zero-energy PLL is initially half-filled, each vertex accommodates two additional electrons or holes, giving n/n0 = +2 and an initially p-doped state, represented by the blue sweep in Fig. 2d.
In the AA configuration, two vertices contribute, giving n/n0 = −4 and an n-doped state, represented by the red sweep. This accounts for the consistently smaller p-doping before the transition compared with the n-doping after the transition.
Energetic stabilization of the quantized state
Integer quantization also requires an energetic explanation. In systems with discrete energy levels, configurations in which states are completely filled or empty and separated by a gap are generally favorable. Examples include Peierls transitions in one-dimensional crystals and the filled shells of noble gases.
Here, the discrete levels do not come from atomic orbitals or an externally applied magnetic field. They arise from PLLs generated by the strong periodic distortions of the moiré reconstruction. Unlike an atomic system, this solid-state structure can be tuned electrostatically using a gate. The system can therefore adjust its energy spectrum to maintain integer filling.
As in flux optimization in GaAs quantum rings,40 the system can adjust the PMF through small changes in strain around the moiré vertices so that the PLL degeneracy matches the electron occupancy. The energy gained by completely filling a Landau level offsets the additional elastic cost of deformation, stabilizing the integer-quantized state.
This self-optimization is local and involves only small strain variations near the moiré vertices. It is distinct from the global structural transition, which results from competition among van der Waals, elastic and electrostatic energies.
Additional evidence that the discrete PLL spectrum causes the integer doping, and that alternative explanations—including charge trapping, Coulomb blocking in moiré-potential quantum dots, interfacial ferroelectricity and contact electrostatics—do not account for the observations, is provided in Supplementary Information Section 39.
Source: www.nature.com












