betaFeedback

WorksPhysics and Astronomy

The Flexoelectric Twist: Persistence of Heavy Fermion Picture in Small-Angle Twisted Bilayer Graphene

Abstract

Twisted bilayer graphene (TBG) near the first magic angle (θ≈1.05∘θ\approx1.05^{\circ}) hosts topological flat bands that provide a fertile ground for exploring strongly correlated phases. At smaller twist angles (θ<1∘θ<1^{\circ}), lattice relaxation profoundly alters the moiré landscape. The extent to which moiré lattice reconstruction at different length scales affects the electronic structure through electro-mechanical coupling remains unresolved. Using scanning tunneling microscopy on small-angle TBG devices with different twist angles θ≈0.3−0.7∘θ\approx0.3-0.7^{\circ}, we observe a pronounced dI/dVdI/dV peak at the AA site of the moiré lattice on top of a spectrum that is asymmetric about the charge neutrality point; we visualize lattice reconstruction at moiré lattice length scales (20-45,nm); and we detect a dilation of the atomic lattice at the AA sites at Angstrom length scale. We demonstrate that the inclusion of moiré lattice reconstruction supports compact localized heavy fermion f−f-states at AA sites whereas the inclusion of atomic lattice dilation generates a strong flexo-electric response that breaks particle-hole symmetry, pushing the ff-states down in energy. Consequently, these heavy fermion states energetically coexist with itinerant cc-electrons at the AB/BA sites. Our spatially resolved spectroscopic measurements visualize this coexistence, confirming the persistence of the heavy fermion model in the small-angle regime θ<1∘θ<1^{\circ}. Our findings show that flexo-electric responses can realize qualitatively distinct electronic landscapes in moiré materials beyond the single-particle picture. Looking ahead, small-angle TBG could prove an ideal platform to explore the interacting large-NN Kondo lattice problem with high tunability of the f−f-occupation and c−c-density through twist-angle and electrostatic back gates, respectively.

Introduction

Moiré superlattices composed of two or more layers of rotationally misaligned two-dimensional materials have emerged as a premier platform for examining topological many-body states [1, 2]. In twisted bilayer graphene (TBG) near the first magic angle (θ≈1.05∘\theta\approx1.05^\circ), the moiré potential modulation of the original Dirac electrons generates a pair of nearly particle-hole (PH) symmetric flat bands [3, 4, 5, 6]. These flat bands (spectral width ≈20\approx20 meV [7]) provide a fertile ground for exploring a rich phase diagram that hosts exotic phenomena, including superconductivity, correlated insulators, ferromagnetism and integer and fractional Chern insulators [8, 9, 10, 11, 12].

It has recently been recognized that the effective low-energy electronic structure of magic-angle TBG can be elegantly captured by a topological heavy fermion (HF) model [13, 14]. This model partitions the system into ff-orbitals localized at the AA stacking sites and itinerant cc-electrons residing in the neighboring AB- and BA-stacking regions Figure 1(a). At the first magic angle, owing to the negligible density of states of the cc-electrons near the PH-symmetric flat bands Figure 1(b), the phenomenology of the partially filled flat bands is dominated by on-site Coulomb interactions within the ff-orbitals, whereas the effect of coupling to the cc-states is largely negligible except within a small region around the moiré Γ\Gamma point [13, 14].

Figure 1. The flexoelectric twist–electronic structure of small-angle twisted bilayer graphene: (a,c) Schematic lattice reconstruction near the first magic angle and at smaller twist angles, respectively. Opposite displacements of the two layers enlarge the AB/BA domains relative to the AA regions. In (c), a common in-plane dilation (green arrows) additionally expands the atomic lattice near AA. Curved arrows denote relative distortion; Layer rotations are exaggerated for clarity. (b,d) Schematic spectral distributions near the first magic angle and at small twist angles, respectively. Red and blue curves indicate AA-localized and itinerant spectral weight. Near the first magic angle, the hybridized bands form an isolated flat-band manifold with approximate particle–hole (PH) symmetry about the charge neutrality point (CNP). At small angles, AA-localized resonances persist within a dense itinerant continuum. The scalar deformation potential associated with common dilation shifts the dominant AA resonance below CNP. (e) Real-space heavy-fermion representation common to both regimes. The AA-centered ff-orbitals couple through VfcV_{fc} to the itinerant cc-states, which have substantial weight in the AB/BA domains and domain-wall (DW) network. The color map denotes the local stacking registry.

However, as the twist angle decreases to θ<1∘\theta< 1^\circ, lattice relaxation of the metastable moiré landscape becomes increasingly dominant. This results in a relative reduction of the AA stacking area and extended AB/BA domains with Bernal stacking registry separated by domain wall states Figure 1(c) [15, 16, 17]. Strain gradients cause a flexoelectric response [18, 19], which is expected to redistribute electric charge within the moiré unit cell. Even though continuum models predicted that other magic angles should occur at θ<1∘\theta< 1^\circ [5], the impact of this lattice reconstruction on the low-energy electronic structure through electro-mechanical coupling has remained poorly understood. Consequently, the experimental study of TBG at these smaller twist angles has received comparatively less attention [15, 20, 21, 22, 23].

Here, using scanning tunneling microscopy and spectroscopy (STM/STS) experiments on small twist-angle TBG devices ranging between θ≈0.3∘\theta\approx0.3^\circ and 1.1∘1.1^\circ, we directly examine the effect of lattice reconstruction on the low-energy electronic structure. We detect enlarged AB/BA at moiré length scales as well as an atomic lattice dilation at the AA sites, accompanied by a highly electron-hole asymmetric differential tunnel conductance (dI/dVdI/dV) spectrum with a pronounced peak localized at the AA site. By comparing these observations with quantitative model calculations, we reveal that the inclusion of this lattice dilation [green arrows in Figure 1(c)] introduces a scalar strain field that naturally breaks electron-hole symmetry at small twist angles Figure 1(d).

The resulting electronic landscape provides a direct mapping to the HF model [13, 14] of highly localized ff-orbitals highly localized ff orbitals embedded in a broad, interconnected continuum of itinerant cc states. Figure 1(e). More broadly, our work reveals that the low-energy physics of small-angle TBG can be captured by the HF model over a remarkably large angle range without the need for fine-tuning to a “magic” condition. This establishes small-angle TBG as a highly tunable framework for examining the Kondo lattice phase diagram and exploring complex strongly correlated states in the future.

Electron-hole symmetry breaking and lattice dilation detected with Scanning Tunneling Microscopy

We have used cryogenic (temperature T=4.5 KT = 4.5\,\mathrm{K}) scanning tunneling microscopy (STM) measurements to characterize the impact of lattice relaxation on the topographic and electronic properties of twisted bilayer graphene (TBG). To this end, we have studied TBG devices near the first magic angle (θ≈1.1∘\theta\approx1.1^\circ) and at smaller twist angles (θ≈0.3∘−0.7∘\theta\approx0.3^\circ- 0.7^\circ). We determined θ\theta and small hetero-strain ε\varepsilon, which is present in all devices, by fitting a uniaxial heterostrain model [24] using the experimentally determined moiré lattice wavelengths λ1,λ2,λ3\lambda_1,\lambda_2,\lambda_3 [see Figure 2(a)] as input parameters. Details of the TBG device fabrication, STM experiment, and this fitting procedure are described in the Methods section and Supplementary Section 1.

Figure 2. Visualizing the impact of moiré lattice relaxation on the electronic properties of twisted bilayer graphene: (a) STM topography of small-angle TBG sample with twist angle θ≈0.37∘\theta\approx0.37^\circ (bias voltage VB=−0.6 VV_{\mathrm{B}}=-0.6\,\mathrm{V}, tunnel current I=50 pAI=50\,\mathrm{pA}). (b) STM topography of a TBG sample (θ≈1.16∘\theta\approx1.16^\circ) near the first magic angle (VB=−0.3 VV_{\mathrm{B}}=-0.3\,\mathrm{V}, I=1 nAI=1\,\mathrm{nA}). (c, d) High-resolution STM topographies recorded at the center of an AA and AB site, respectively, within the field of view of panel a (VB=0.3 VV_{\mathrm{B}}=0.3\,\mathrm{V}, I=50 pAI=50\,\mathrm{pA}). (e, f) Corresponding two-dimensional fast Fourier transform (FFT) maps of panels c and d. (g) Normalized dI/dVdI/dV spectra recorded at the center of AA sites of TBG samples with different indicated twist angles (VB=−300 mVV_{\mathrm{B}}=-300\,\mathrm{mV}, I=500 pAI=500\,\mathrm{pA}, Vmod=3 mVV_{\mathrm{mod}}=3\,\mathrm{mV}). (h) Spectroscopic line cut recorded along the AA-AB/BA-DW-BA/AB-AA direction within the field of view of panel a (VB=−300 mVV_{\mathrm{B}}=-300\,\mathrm{mV}, I=500 pAI=500\,\mathrm{pA}, lock-in modulation Vmod=3 mVV_{\mathrm{mod}}=3\,\mathrm{mV}). (i) Twist angle dependence of the bias voltage position of the dI/dVdI/dV peak detected on AA sites in panel g.

In Figure 2(a, Figure 2b), we compare STM topographies recorded on small-angle TBG (θ≈0.37∘\theta\approx0.37^\circ) and on a sample near the magic angle (θ≈1.17∘\theta\approx1.17^\circ). Both topographies exhibit a triangular lattice of AA stacking sites, but the different twist angle strongly affects the moiré lattice characteristics. At small angles, the AA sites are separated by much larger, extended AB stacking regions and connected by domain wall sites [also see previous work on the STM study of the AB and domain wall sites by Huang et al. in [20]]. This finding contrasts with the dense lattice of AA sites that characterizes the well-established topography of magic-angle devices [24, 25, 26], highlighting the profound role of lattice relaxation on the moiré lattice at small twist angles θ<1∘\theta< 1^\circ.

Indications for lattice relaxation effects can also be found on atomic length scales when examining atomically resolved STM topographies recorded at the AA and AB sites of the small-angle TBG sample (θ≈0.37∘\theta\approx0.37^\circ) in Figure 2(c, Figure 2d). Note that the different topographic appearances of these maps reflect the underlying atomic stacking registry of the AA and AB regions. The atomic lattice constants of these regions (aAAa_{\mathrm{AA}} and aABa_{\mathrm{AB}}) can be precisely determined by analyzing the corresponding two-dimensional fast-Fourier transform (2D-FFT) maps shown in Figure 2(e, Figure 2f) in which the periodic atomic lattice in real space transforms into a set of six Bragg vectors qBragg\mathbf{q}_{\mathrm{Bragg}} of length qBragg=2/3aAA/ABq_{\mathrm{Bragg}} = 2/\sqrt{3}a_{\mathrm{AA/AB}}. Our detailed analysis of the FFT maps [see Supplementary Section 2] reveals a relative dilation Δη=100×2(aAA−aAB)/(aAA+aAB)≈2.1%\Delta\eta= 100 \times2(a_{\mathrm{AA}} - a_{\mathrm{AB}})/(a_{\mathrm{AA}} + a_{\mathrm{AB}}) \approx2.1\% of the atomic lattice constant between the AA and AB sites averaged over the field of view of Figure 2(a); that is, the atomic lattice constant of the AA site is larger compared to the lattice constant of the AB site. In contrast, the same analysis applied to STM topographies recorded on the TBG device near the magic angle reveals identical atomic lattice constants at the AA and AB/BA sites [see Supplementary Section 2].

Moiré lattice relaxation effects also manifest in the local electronic density of states (LDOS) at low energies, indicating the presence of finite electro-mechanical coupling. Scanning tunneling spectroscopy (STS) measurements of a device near the magic angle (θ=1.04∘\theta= 1.04^\circ) reveal a pair of dI/dVdI/dV peaks that is approximately symmetric about the charge neutrality point at zero bias voltage VBV_{\mathrm{B}}. These dI/dVdI/dV peaks, which give rise to the topographic contrast of the AA sites in Figure 2(b), reflect the LDOS of the two flat bands [4, 24, 25, 26] that preserve approximate PH symmetry [27]. STS measurements conducted on small-angle TBG (θ≈0.37∘\theta\approx0.37^\circ) reveal qualitatively different characteristics. The dI/dVdI/dV spectrum [Fig. 2(g)] features a pronounced peak centered at VB≈−80V_{\mathrm{B}} \approx-80 mV and exhibits a pronounced asymmetry about the charge neutrality point (CNP), which is located at zero bias voltage. A spectroscopic line-cut recorded between two AA sites on the sample surface [white dashed line in Figure 2(a)] illustrates that this dI/dVdI/dV peak is strongly localized to the AA site [Fig. 2(h)]. Moreover, it reveals a dI/dVdI/dV peak at VB≈80V_{\mathrm{B}} \approx80 mV localized at the domain wall site, which was previously discussed in the context of helical edge states [20], as well as a set of spatially extended dI/dVdI/dV features of smaller amplitude between −50-50 mV <VB<50< V_{\mathrm{B}} < 50 mV that appear as sharp spectroscopic features of width ≈10\approx10 mV. Finally, examining dI/dVdI/dV spectra recorded at the AA site of samples with different twist angles [Fig. 2(g)], we find that the position of the dI/dVdI/dV peak shifts to more negative bias voltages as the twist angle decreases [Fig. 2(i)].

Impact of axial strain on the low-energy electronic structure in small-angle TBG

Our experimental observations of lattice reconstruction at moiré length scales, relative atomic lattice dilation at AA sites, and a particle–hole-asymmetric local density of states suggest an important role of lattice deformation effects in small-angle TBG at θ<1∘\theta< 1^\circ. To interpret these observations, we distinguish the relative reconstruction at moiré length scales from atomic-scale dilation effects. From a theoretical viewpoint, toward smaller twist angles, lattice relaxation effects cause the two graphene layers to rotate locally toward Bernal registry over larger AB/BA domains, concentrating the registry changes near the AA centers and a well-developed domain-wall network [Fig. 1(a, c)] [20, 28, 29, 30, 31]. We obtain this relative reconstruction by minimizing the continuum elastic and interlayer adhesion energy [see Supplementary Section 3]. However, within the effective continuum model used here, relative reconstruction preserves the approximate particle–hole (PH) symmetry of the ZLL-derived spectrum [see Ref. [27] and Supplementary Section 4]. Microscopic calculations near the first magic angle report PH-breaking corrections of several meV [32], much smaller than the energy scale corresponding to the approximately 100 mV shift of the experimentally observed dI/dVdI/dV peak in small-angle devices.

Motivated by our experimental observations, we therefore introduce an additional phenomenological common displacement that expands the atomic lattice of both graphene layers near the AA site center relative to the AB/BA regions. We show that the inclusion of this dilation, which effectively introduces a finite electro-mechanical coupling, naturally results in a PH symmetry-breaking LDOS through the scalar deformation potential [Fig. 1(c)]. We assume equal additional displacements in the two layers, i.e., a common displacement that leaves the relative stacking registry unchanged, which gives a lattice constant dilation profile η(r)=2Δη9[Cp(r)−⟨Cp⟩M]\eta(\mathbf{r}) = \frac{2\Delta\eta}{9}[C_p(\mathbf{r}) - \langle C_p \rangle_M], with Cp(r)=∑j=13cos⁡[bj⋅δref(r)]C_p(\mathbf{r}) = \sum_{j=1}^{3} \cos[\mathbf{b}_j \cdot\boldsymbol{\delta}_{\mathrm{ref}}(\mathbf{r})]. Here, Δη=ηAA−ηAB\Delta\eta= \eta_{\mathrm{AA}} - \eta_{\mathrm{AB}} denotes the lattice dilation contrast between AA and AB sites, and ⟨⋅⟩M\langle\cdot\rangle_M denotes the moiré-cell average. The vectors bj\mathbf{b}_j (j=1,2,3j = 1, 2, 3) are chosen from the first shell of the graphene reciprocal lattice and are separated by 120∘120^\circ. The reference stacking registry field δref(r)\boldsymbol{\delta}_{\mathrm{ref}}(\mathbf{r}) denotes the in-plane relative shift between two layers in a reference TBG model, whose twist angle can be tuned to obtain different expansion region sizes centered at AA [see Supplementary Section 3]. Through the scalar deformation potential of graphene [32, 33, 34, 35, 36], this common dilation lowers the orbital energy in the AA stacking region and shifts the dominant ZLL-derived resonance below the CNP.

Figure 3 presents continuum calculations that incorporate lattice relaxation and strain-dependent electronic couplings [28, 29, 32, 33, 37]. Details of the model, displacement-field construction, orbital projectors, and numerical checks are presented in Supplementary Section 4. Figure 3(a,Figure 3b) show the relative and common displacement fields at θ=0.374∘\theta= 0.374^\circ, respectively. The relative displacement produces opposite local rotations in the two layers, enlarging the AB/BA domains and concentrating the distortion near the DW network. The associated strain generates a spatially modulated pseudovector potential [32, 33, 37, 38, 39, 40], which by itself, however, does not shift the DOS peak of ff-orbitals localized at the AA sites.

Figure 3. Particle-hole symmetry breaking from AA-site lattice dilation in small-angle twisted bilayer graphene: (a, b) Relative and common displacement fields, u−=u(2)−u(1)\mathbf{u}^{-}=\mathbf{u}^{(2)}-\mathbf{u}^{(1)} and u+=u(2)+u(1)\mathbf{u}^{+}=\mathbf{u}^{(2)}+\mathbf{u}^{(1)}, respectively, at θ=0.374∘\theta=0.374^\circ. The relative displacement is obtained from elastic relaxation, and the common displacement is introduced phenomenologically with an AA–AB dilation contrast Δη=2%\Delta\eta=2\% and a half-contrast equivalent-area radius wη=0.107λmw_{\eta}=0.107\lambda_m. (c) Continuum band structure colored by the AA-localized ff-orbital weight. The left inset provides a zoomed-in view of the hybridized resonance, while the right inset shows the bare f−f^{-} and c−c^{-}-bands. (d) Registry-resolved local density of states (LDOS) at the AA, AB/BA, and DW sites. (e) Calculated AA site LDOS at three indicated twist angles as the lattice dilation contrast Δη\Delta\eta increases from 0 to 2.5%. Spectra are vertically offset for clarity. (f) Representative calculated LDOS spectrum at the AA site at θ=0.323∘\theta=0.323^\circ, 0.374∘0.374^\circ, and 0.438∘0.438^\circ, using Δη=2.5%\Delta\eta=2.5\%, 2%2\%, and 1.5%1.5\%, respectively. Both (e) and (f) use wη/λm=0.107w_{\eta}/\lambda_m=0.107, with vertical dashed lines marking the CNP. (g, h) Hybridized total and f/cf/c-projected DOS, and bare f/cf/c DOS at Vfc=0V_{fc}=0, respectively.

The calculated spectral function [Fig. 3(c)] reveals a pronounced resonance near −85-85 meV within a dense continuum of weakly dispersive [band width ≈10\approx10 meV] mini bands that occupy the entire displayed energy interval. These mini bands arise from band folding into the small moiré Brillouin zone and energetically overlap, collectively forming a broad itinerant background. Their narrow individual bandwidths therefore do not imply the same localization as the nearly dispersionless bare ff-orbital bands [13]. To identify the AA-localized component across the hybridized bands, we use the most localized zero-mode doublet of the continuum Hamiltonian expanded near an AA center [see Supplementary Section 5]. This local Hamiltonian has a pseudo-Landau-level form, providing AA-centered orbital projectors and a common construction for comparison with magic-angle TBG [13]. The orbital character of this resonance is resolved using the AA-site localized zeroth Landau level (ZLL) doublet by the projector onto the primary m=0,1m = 0, 1 ZLL doublet, comprising two orbitals per spin and valley and defining the ff sector of the heavy-fermion representation [13, 14]. The band coloring resolves this doublet’s weight across several hybridized bands near the resonance [left inset of Figure 3(c)]. Examining the registry-resolved LDOS [Figure 3(d)] confirms that the resonance at E≈−86E \approx-86 meV is strongly localized at the AA site, displaying much weaker amplitude at AB/BA and DW regions.

To further verify the impact of lattice dilation on the low-energy electronic structure, we analyze the LDOS at the AA site as a function of the lattice dilation contrast Δη\Delta\eta [Figure 3(e)]. In the absence of lattice dilation effects, the resulting LDOS is centered about the CNP and has approximate PH-symmetry at all twist angles. This case contrasts with the PH-breaking LDOS in the presence of lattice dilation (Δη>0\Delta\eta> 0), which is asymmetric about the CNP. Here, the dilation effect causes a twist-angle dependent shift of the AA resonance peak to lower energies. This finding highlights the critical role of common lattice dilation at the AA site, as detected in STM measurements, for inducing a PH-symmetry breaking band structure.

Finally, we compare the calculated LDOS at the AA site with our experimental observations from STM measurements [Figure 2(g)]. In Figure 3(f), we compare calculated LDOS spectra at the AA site for θ=0.323∘\theta= 0.323^\circ, 0.374∘0.374^\circ, and 0.438∘0.438^\circ, using Δη=2.5%\Delta\eta= 2.5\%, 2%2\%, and 1.5%1.5\%, respectively, at the common fractional width wη/LM=0.107w_\eta/L_M = 0.107. At each angle, the lattice dilation contrast Δη\Delta\eta was selected within the estimated strain range to match the measured energy E=eVBE = eV_B of the experimentally detected dI/dVdI/dV peak. Given these parameter choices, the calculated AA LDOS peaks occur at E=−121E = -121, −84-84, and −59-59 meV, providing a good approximation of the experimentally observed peak positions at −117-117, −90-90, and −51-51 meV, respectively.

Heavy fermions embedded in an itinerant continuum in small-angle TBG

The inclusion of the experimentally detected lattice dilation into our continuum model provides a phenomenological framework that successfully captures the impact of electro-mechanical coupling on the low-energy LDOS. In the following, we further show that the localized resonance at the AA site and the dense continuum of bands can be directly mapped into the heavy fermion model basis,

H(k)=(Hff(k)Vfc(k)Vfc†(k)Hcc(k)).(1)H(\mathbf{k}) = \begin{pmatrix} H_{ff}(\mathbf{k}) & V_{fc}(\mathbf{k}) \\ V_{fc}^{\dagger}(\mathbf{k}) & H_{cc}(\mathbf{k}) \end{pmatrix}. \tag*{(1)}

Here HffH_{ff} acts within the localized doublet, HccH_{cc} acts in its orthogonal complement Pc=1−PfP_c = 1 - P_f, and VfcV_{fc} couples the two sectors. The complement is constructed using the orthogonalized-plane-wave (OPW) approach of Ref. [13]. This transformation retains the full plane-wave space used in the calculation and leaves the spectrum unchanged. Setting Vfc=0V_{fc} = 0 allows us to independently examine the two sectors of this model [right inset of Fig. 3(c)]. In this limit, this projection yields two HffH_{ff} bands around −88-88 meV, effectively dispersionless on the resolved energy scale, and a dispersive HccH_{cc} background. At small twist angles, the increased separation between AA-centered ff orbitals of nearly unchanged spatial extent suppresses intercell hopping, reducing their bandwidths far below the ∼0.5\sim0.5 meV scale reported at twist angles near the magic angle [13]. However, restoring VfcV_{fc} redistributes the localized weight among the continuum bands, accompanied by a small spectral shift to approximately −85-85 meV. The resulting ff-orbital and cc-state projected DOS in the bare and hybridized case are shown in Figure 3(g, Figure 3h).

This analysis reveals that the HF model, previously derived for describing the low-energy electronic structure of magic-angle TBG, naturally extends to smaller twist angles. We verify this understanding by considering spatially resolved maps of the differential tunnel conductance recorded on device A0 (θ≈0.37∘\theta\approx0.37^\circ). The strong localization of the dI/dVdI/dV peak at the AA site (VB=−85V_B = -85 mV) is successfully captured by the hybridized ff-orbital state in our calculation [Figure 4(a, Figure 4b)]. Meanwhile, our measurements near Fermi energy (VB=−17V_B = -17 mV) reveal that the narrow dI/dVdI/dV peaks [width ≈10\approx10 mV] seen in our spectroscopic measurements [Figure 2(h)] arise from spatially extended electronic states with triangular sub-structure at the AB sites [Figure 4(c)]. This observation is closely mirrored by our model calculation [Figure 4(d)], indicating that the electronic structure of small-angle TBG comprises strongly localized ff-electrons embedded in itinerant cc-electronic states.

Figure 4. Heavy fermions embedded in an itinerant continuum in small-angle twisted bilayer graphene: (a, b) Differential conductance (dI/dVdI/dV) maps recorded on small-angle TBG sample (θ≈0.37∘\theta\approx0.37^\circ) at the bias voltage of the dI/dVdI/dV peak (VB=−85 mVV_{\mathrm{B}} = -85\,\mathrm{mV}) and near Fermi energy (VB=−17 mVV_{\mathrm{B}} = -17\,\mathrm{mV}), respectively. (c, d) Corresponding simulated local density of states (LDOS) maps at the energy of the AA site peak (E−EF=−85 meVE-E_{\mathrm{F}} = -85\,\mathrm{meV}) and near Fermi energy (E−EF=−4 meVE-E_{\mathrm{F}} = -4\,\mathrm{meV}), respectively. (e) Ratio of AA region radius rAAr_{\mathrm{AA}} and the moiré wavelength rAA/λmr_{\mathrm{AA}}/\lambda_{\mathrm{m}} as a function of twist angle θ\theta extracted from STM topographic images of different devices. Inset: The value of rAAr_{\mathrm{AA}} is determined by using super-Gaussian fitting (red line) of topographic line profiles along AA sites (black line). Details of this procedure are described in Supplementary Section 6. (f) Calculated equivalent-area radii of the dilation half-contrast region (wηw_{\eta}, blue) and the AA stacking-registry region (wregistryw_{\mathrm{registry}}, orange), normalized by the moiré period λm\lambda_{\mathrm{m}}. Here the dilation profile is generated from the relaxed registry at each physical twist angle [see Supplementary § 0.3].

We have also experimentally characterized the evolution of the HF picture at small twist angles by examining the spatial profile of the AA-localized features. To this end, we analyze the width of the AA site in STM topographies. We determine its effective spatial radius rAAr_{AA} by fitting a topographic line cut across AA sites using a Super-Gaussian profile [Figure 4(e), see Supplementary Sec. 6 for details]. Our analysis reveals that the ratio rAA/λmr_{AA}/\lambda_m decreases approximately linearly as the twist angle is reduced. This observation is qualitatively reproduced by the calculated equivalent-area radii of the dilation half-contrast region and the AA stacking-registry region [Figure 4(f)] and is consistent with the theoretical understanding that in this strongly reconstructed regime, the absolute area of an AA region is nearly angle independent, while its fraction of the expanding moiré cell steadily decreases and the AB/BA domains occupy most of the cell [30, 41]. Each moiré cell contributes a fixed integrated ff-state spectral weight of two states per spin and valley, whereas the broad cc-state DOS retains a nearly angle-independent overall scale per unit area. The cc-state spectral weight per cell within a fixed energy window therefore increases with the moiré-cell area [see Figure 5 for twist-angle dependent DOS]. Consequently, while the local confinement scale of the ff-orbitals remains comparable to that near the first magic angle, as shown by the local-orbital calculations, the decreasing AA area fraction reduces the relative contribution of AA-localized states to the low-energy manifold, leaving them embedded in the growing continuum of itinerant cc-states.

Figure 5. Toward Kondo lattice physics in small-angle twisted bilayer graphene: (a) Energy (EE) dependent density of states (DOS) per moiré cell of the itinerant electrons at different indicated twist angles. Energies are measured relative to the charge-neutrality point ECNPE_{\mathrm{CNP}}, and the DOS is given per spin and valley. (b) DOS per moiré cell of itinerant cc-states (extracted from panel a) at the energy of the dI/dVdI/dV peak at the AA site as a function of the twist angle θ\theta. The spectral position of the dI/dVdI/dV peaks [see Figure 2(i)] is indicated by color-coded markers in panel a. Inset: Shown is the schematic dependence of the Kondo (red line) and RKKY (blue line) energy (EE) scale as a function of the itinerant density of states.

Discussion

Our study demonstrates that lattice relaxation effects in small-angle TBG (θ<1∘\theta< 1^\circ) manifests in a strongly particle-hole breaking electronic structure, qualitatively distinct from that observed at the first magic angle (θ=1.05∘\theta= 1.05^\circ). Consequently, the minimal Bistritzer–MacDonald continuum model [5]—which provides a reasonable approximation near the first magic angle but predicts a series of higher-order magic angles at θ<1∘\theta< 1^\circ—becomes insufficient without considering effects of relative moiré lattice reconstruction and atomic lattice dilation in this regime, as shown in this work. Moreover, while initially constructed to describe the physics near the first magic angle, we show that the HF model naturally extends to the small-angle regime where the appearance of compact, localized ff-states at the AA-sites is found to be independent of the specific twist angle in experiment and theory.

Thus, small-angle TBG provides a versatile platform to examine correlation-driven physics between partially filled ff-orbitals without the need for strict fine-tuning to specific magic angles. This regime presents a physically distinct phenomenology compared to systems with spectrally isolated flat bands. Near the first magic angle, the ff-orbital is approximately spectrally separated from the cc-electrons over most of the moiré Brillouin zone [c.f. Figure 1(b)], except near the Γ\Gamma point [13, 14], meaning correlation effects within the partially filled flat band are predominantly governed by on-site Coulomb interactions [14, 25, 42, 43]. Similarly, the flat bands in homobilayers of twisted transition metal dichalcogenides (TMD) are spectrally isolated, with low-energy physics dominated by local and non-local Coulomb interactions [44, 45, 46, 47].

Conversely, in small-angle TBG, flexo-electric coupling shifts the ff-orbital to lower energies where they coexist with the itinerant cc-states, as revealed in our study. The analysis of the calculated bare (not hybridized with ff-orbitals) cc-state DOS spectra at different twist angles Figure 5(a) reveals that the DOS of the itinerant electronic states at the energy of the ff-orbital scales approximately exponentially with decreasing twist angle [Fig. 5(b)], also see Supplementary Section 5.D. The itinerant background can effectively screen Coulomb interactions [48, 49, 50], leading to qualitatively different expected low-energy ground states than near the magic angle. Hence, the electronic structure of small-angle TBG bears similarity to the Kondo lattice in TMD hetero-bilayers [51], however, with significantly enhanced cc-DOS and larger ff-orbital degeneracy (eightfold). Although a microscopic theoretical framework for the electro-mechanical coupling in TBG remains to be established, our study showcases that flexo-electric responses can realize novel electronic landscapes in moiré lattice materials beyond the single-particle picture.

The interacting Kondo lattice problem of ff-orbitals embedded in a cc-continuum has been extensively studied in bulk heavy fermion compounds [52, 53, 54] and more recently in TMD hetero-bilayers [55]; however, revealing the physics of the quantum critical point separating Kondo-screened and long-range ordered phases has remained challenging. While the Ruderman-Kittel-Kasuya-Yosida interaction—which favors long-range magnetic order—depends linearly on the DOS at the Fermi energy, the Kondo energy scale depends exponentially on it [inset Fig. 5(b)]. Looking ahead, by utilizing the twist angle and electrostatic gating as tuning knobs for the cc-DOS and ff-orbital occupation (forming an effective SU(8)SU(8) space by spin, valley, and orbital-degeneracy), small-angle TBG could thus provide a model testbed for studying emergent ground states within the large-NN model [56, 57].

References

Acknowledgments

We appreciate valuable discussions with Jonah Herzog-Arbeitman. This work was primarily supported by the Hong Kong RGC (Grant Nos. 26304221, 16302422, 16302624, 16304525, and C6033-22G) and the Croucher Foundation (Grant No. CIA22SC02) awarded to B.J. B.J. and X.D acknowledge support through the Area of Excellence Scheme by the Hong Kong RGC (Grant No. AoE/P-604/25-R).

Methods

Device fabrication

All devices were fabricated using dry-transfer method. Graphene and h-BN (thickness 15 nm-30 nm) flakes were mechanically exfoliated onto a 285 nm SiO2\mathrm{SiO}_2/Si wafer using Scotch tape. The monolayer graphene was cut into half using an atomic-force-microscopy tip. Then, the TBG/h-BN heterostructure was picked up by a polyvinyl alcohol (PVA)/polycarbonate (PC)/polydimethylsiloxane (PDMS) stamp. The first half of the graphene was picked up by the stamp. The other part of the graphene (on the SiO2\mathrm{SiO}_2/Si wafer) was rotated by a target twist angle relative to the first half of the graphene and then picked up by the stamp. Subsequently, the h-BN flake was picked up and finally the TBG/h-BN heterostructure was released onto a SiO2\mathrm{SiO}_2/Si chip with prepatterned Cr/Au(5nm/50nm) electrodes. The polymer on the devices was then removed in subsequent organic solvent cleaning steps using N-methyl-2-pyrrolidone (NMP), deionized water, and isopropanol. The TBG surface was further cleaned with atomic-force-microscopy tip cleaning. Finally, devices were annealed in ultrahigh vacuum at 170°C for 12 h before being transferred to the STM.

STM measurements

STM measurements were performed using a home-built STM with a base pressure of ≈1.4×10−10\approx1.4 \times10^{-10} mbar and base temperature of T=4.5T = 4.5 K. The STM tip was prepared on a Cu(111) surface by field emission and controlled indention and calibrated against the Cu(111) Shockley surface state before each set of measurements. The differential conductance (dI/dVdI/dV) spectra and maps were captured by a standard lock-in technique with a small bias modulation VacV_{\mathrm{ac}} ranging from 0.5 mV to 3 mV at a frequency of f=719f = 719 Hz. dI/dVdI/dV maps were recorded using the multipass mode, where the topographic height profile was first recorded in constant current mode at a suitable sample bias voltage and then the position-dependent dI/dVdI/dV amplitude was recorded at a chosen sample bias voltage VBV_{\mathrm{B}} in a set of multiple passes along the same topographic line.

Figures

References

  1. [1]E. Y. Andrei, D. K. Efetov, P. Jarillo-Herrero, A. H. MacDonald, K. F. Mak, T. Senthil, E. Tutuc, A. Yazdani, and A. F. Young, Nature Reviews Materials 6, 201 (2021).
  2. [2]K. F. Mak and J. Shan, Nature Nanotechnology 17, 686 (2022).
  3. [3]J. Lopes dos Santos, N. Peres, and A. Castro Neto, Physical review letters 99, 256802 (2007).
  4. [4]G. Li, A. Luican, J. Lopes dos Santos, A. Castro Neto, A. Reina, J. Kong, and E. Andrei, Nature physics 6, 109 (2010).
  5. [5]R. Bistritzer and A. H. MacDonald, Proc. Natl. Acad. Sci. U.S.A. 108, 12233 (2011).
  6. [6]E. Y. Andrei and A. H. MacDonald, Nature materials 19, 1265 (2020).
  7. [7]S. Lisi, X. Lu, T. Benschop, T. A. de Jong, P. Stepanov, J. R. Duran, F. Margot, I. Cucchi, E. Cappelli, et al., Nature Physics 17, 189 (2021).
  8. [8]Y. Cao, V. Fatemi, S. Fang, K. Watanabe, T. Taniguchi, E. Kaxiras, and P. Jarillo-Herrero, Nature 556, 43 (2018).
  9. [9]Y. Cao, V. Fatemi, A. Demir, S. Fang, S. L. Tomarken, J. Y. Luo, J. D. Sanchez-Yamagishi, K. Watanabe, T. Taniguchi, E. Kaxiras, et al., Nature 556, 80 (2018).
  10. [10]A. L. Sharpe, E. J. Fox, A. W. Barnard, J. Finney, K. Watanabe, T. Taniguchi, M. Kastner, and D. Goldhaber-Gordon, Science 365, 605 (2019).
  11. [11]K. P. Nuckolls, M. Oh, D. Wong, B. Lian, K. Watanabe, T. Taniguchi, B. A. Bernевig, and A. Yazdani, Nature 588, 610 (2020).
  12. [12]Y. Xie, A. T. Pierce, J. M. Park, D. E. Parker, E. Khalaf, P. Ledwith, Y. Cao, S. H. Lee, S. Chen, P. R. Forrester, et al., Nature 600, 439 (2021).
  13. [13]H. Shi and X. Dai, Phys. Rev. B 106, 245129 (2022).
  14. [14]Z.-D. Song and B. A. Bernevig, Phys. Rev. Lett. 129, 047601 (2022).
  15. [15]J. S. Alden, A. W. Tsen, P. Y. Huang, R. Hovden, L. Brown, J. Park, D. A. Muller, and P. L. McEuen, Proceedings of the National Academy of Sciences 110, 11256 (2013).
  16. [16]P. San-Jose and E. Prada, arXiv preprint arXiv:1304.5344 (2013).
  17. [17]F. Zhang, A. H. MacDonald, and E. J. Mele, Proceedings of the National Academy of Sciences 110, 10546 (2013).
  18. [18]L. J. McGilly, A. Kerelsky, N. R. Finney, K. Shapovalov, E.-M. Shih, A. Ghiotto, Y. Zeng, S. L. Moore, W. Wu, Y. Bai, et al., Nature Nanotechnology 15, 580 (2020).
  19. [19]Y. Li, X. Wang, D. Tang, X. Wang, K. Watanabe, T. Taniguchi, D. R. Gamelin, D. H. Cobden, M. Yankowitz, X. Xu, et al., Advanced Materials 33, 2105879 (2021).
  20. [20]S. Huang, K. Kim, D. K. Efimkin, T. Lovorn, T. Taniguchi, K. Watanabe, A. H. MacDonald, E. Tutuc, and B. J. LeRoy, Physical review letters 121, 037702 (2018).
  21. [21]P. Rickhaus, J. Wallbank, S. Slizovskiy, R. Pisoni, H. Overweg, Y. Lee, M. Eich, M.-H. Liu, K. Watanabe, T. Taniguchi, et al., Nano letters 18, 6725 (2018).
  22. [22]S. Xu, A. Berdyugin, P. Kumaravadivel, F. Guinea, R. Krishna Kumar, D. Bandurin, S. Morozov, W. Kuang, B. Tsim, S. Liu, et al., Nature communications 10, 4008 (2019).
  23. [23]K. P. Nuckolls, M. G. Scheer, D. Wong, M. Oh, R. L. Lee, J. Herzog-Arbeitman, K. Watanabe, T. Taniguchi, B. Lian, and A. Yazdani, Nature 639, 60 (2025).
  24. [24]A. Kerelsky, L. J. McGilly, D. M. Kennes, L. Xian, M. Yankowitz, S. Chen, K. Watanabe, T. Taniguchi, J. Hone, C. Dean, et al., Nature 572, 95 (2019).
  25. [25]Y. Xie, B. Lian, B. Jäck, X. Liu, C.-L. Chiu, K. Watanabe, T. Taniguchi, B. A. Bernevig, and A. Yazdani, Nature 572, 101 (2019).
  26. [26]Y. Choi, J. Kemmer, Y. Peng, A. Thomson, H. Arora, R. Polski, Y. Zhang, H. Ren, J. Alicea, G. Refael, et al., Nature physics 15, 1174 (2019).
  27. [27]B. A. Bernevig, Z.-D. Song, N. Regnault, and B. Lian, Phys. Rev. B 103, 205413 (2021).
  28. [28]N. N. T. Nam and M. Koshino, Phys. Rev. B 96, 075311 (2017).
  29. [29]S. Carr, D. Massatt, S. B. Torrisi, P. Cazeaux, M. Luskin, and E. Kaxiras, Phys. Rev. B 98, 224102 (2018).
  30. [30]N. P. Kazmierczak, M. Van Winkle, C. Ophus, K. C. Bustillo, S. Carr, H. G. Brown, J. Ciston, T. Taniguchi, K. Watanabe, and D. K. Bediako, Nat. Mater. 20, 956 (2021).
  31. [31]N. R. Walet and F. Guinea, 2D Mater. 7, 015023 (2020).
  32. [32]J. Kang and O. Vafek, Phys. Rev. B 107, 075408 (2023).
  33. [33]O. Vafek and J. Kang, Phys. Rev. B 107, 075123 (2023).
  34. [34]S.-M. Choi, S.-H. Jhi, and Y.-W. Son, Phys. Rev. B 81, 081407 (2010).
  35. [35]D. Grassano, M. D’Alessandro, O. Pulci, S. G. Sharapov, V. P. Gusynin, and A. A. Varlamov, Phys. Rev. B 101, 245115 (2020).
  36. [36]L. Wang, A. Baumgartner, P. Makk, S. Zihlmann, B. S. Varghese, D. I. Indolese, K. Watanabe, T. Taniguchi, and C. Schönenberger, Commun. Phys. 4, 147 (2021).
  37. [37]M. Koshino and N. N. T. Nam, Phys. Rev. B 101, 195425 (2020).
  38. [38]B. Xie and J. Liu, Phys. Rev. B 108, 094115 (2023).
  39. [39]H. Shi, Z. Zhan, Z. Qi, K. Huang, E. van Veen, J. A. Silva-Guillén, R. Zhang, P. Li, K. Xie, H. Ji, M. I. Katsnelson, S. Yuan, S. Qin, and Z. Zhang, Nat. Commun. 11, 371 (2020).
  40. [40]A. Ceferino and F. Guinea, 2D Mater. 11, 035015 (2024).
  41. [41]V. H. Nguyen, D. Paszko, M. Lamparski, B. Van Troeye, V. Meunier, and J.-C. Charlier, 2D Mater. 8, 035046 (2021).
  42. [42]D. Wong, K. P. Nuckolls, M. Oh, B. Lian, Y. Xie, S. Jeon, K. Watanabe, T. Taniguchi, B. A. Bernevig, and A. Yazdani, Nature 582, 198 (2020).
  43. [43]U. Zondiner, A. Rozen, D. Rodan-Legrain, Y. Cao, R. Queiroz, T. Taniguchi, K. Watanabe, Y. Oreg, F. von Oppen, A. Stern, et al., Nature 582, 203 (2020).
  44. [44]F. Wu, T. Lovorn, E. Tutuc, and A. H. MacDonald, Physical review letters 121, 026402 (2018).
  45. [45]Y. Tang, L. Li, T. Li, Y. Xu, S. Liu, K. Barmak, K. Watanabe, T. Taniguchi, A. H. MacDonald, J. Shan, et al., Nature 579, 353 (2020).
  46. [46]C. Xu, J. Li, Y. Xu, Z. Bi, and Y. Zhang, Proceedings of the National Academy of Sciences 121, e2316749121 (2024).
  47. [47]N. Morales-Durán, N. C. Hu, P. Potasz, and A. H. MacDonald, Phys. Rev. Lett. 128, 217202 (2022).
  48. [48]J. M. Pizarro, M. Rösner, R. Thomale, R. Valentí, and T. O. Wehling, Phys. Rev. B 100, 161102 (2019).
  49. [49]J. Barrier, L. Peng, S. Xu, C. De Beule, V. I. Fal’ko, K. Watanabe, T. Taniguchi, A. K. Geim, S. Adam, and A. I. Berdyugin, Phys. Rev. X 16, 031040 (2026).
  50. [50]X. Gao, A. Jimeno-Pozo, P. A. Pantaleon, A. Rajesh, E. Codecido, D. L. Sharifi, Z. Zhang, Y. Liu, K. Watanabe, T. Taniguchi, M. W. Bockrath, F. Guinea, and C. N. Lau, Nature Physics 22, 692 (2026).
  51. [51]W. Zhao, B. Shen, Z. Tao, Z. Han, K. Kang, K. Watanabe, T. Taniguchi, K. F. Mak, and J. Shan, Nature 616, 61 (2023).
  52. [52]H. v. Löhneysen, A. Rosch, M. Vojta, and P. Wölfle, Reviews of Modern Physics 79, 1015 (2007).
  53. [53]P. Gegenwart, Q. Si, and F. Steglich, nature physics 4, 186 (2008).
  54. [54]Q. Si and F. Steglich, Science 329, 1161 (2010).
  55. [55]W. Zhao, B. Shen, Z. Tao, S. Kim, P. Knüppel, Z. Han, Y. Zhang, K. Watanabe, T. Taniguchi, D. Chowdhury, et al., Nature Physics 20, 1772 (2024).
  56. [56]H. Hu, G. Rai, L. Crippa, J. Herzog-Arbeitman, D. Călugăru, T. Wehling, G. Sangionnani, R. Valentí, A. M. Tsvelik, and B. A. Bernevig, Physical review letters 131, 166501 (2023).
  57. [57]Y.-Z. Chou and S. Das Sarma, Physical Review Letters 131, 026501 (2023).

Paper details

vlnc:W_39unn779
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 2

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents