Band-basis decomposition of superfluid weight in magic-angle twisted bilayer graphene: Quantifying geometric and conventional contributions
Jian Zhou · 2026
Read Band-basis decomposition of superfluid weight in magic-angle twisted bilayer graphene: Quantifying geometric and conventional contributions by Jian Zhou (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-05994.txt.
Introduction
Zhou splits the superfluid weight of magic-angle twisted bilayer graphene into conventional band-velocity and geometric interband-coherence parts. In the flat band, quantum geometry gives 22 to 26 percent of the weight at charge neutrality, rising to about 55 to 58 percent once remote bands join, purely through interband coherence. That share peaks near fillings of plus or minus two, where superconductivity runs strongest.
Catalog record
| CREATOR | Jian Zhou |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 21149 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.05994
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Band-basis decomposition of superfluid weight in magic-angle twisted bilayer graphene: Quantifying geometric and conventional contributions Jian Zhou Abstract We decompose the superfluid weight D s D_{s} of magic-angle twisted bilayer graphene (MATBG) into conventional (band-velocity) and geometric (interband-coherence) contributions using a band-basis current operator splitting applied to the Bistritzer-MacDonald continuum model. In the flat-band subspace, quantum geometry accounts for 22 22 – 26 % 26\% of D s D_{s} at charge neutrality depending on pairing symmetry, with cross terms vanishing to machine precision. Including remote bands raises the geometric fraction to ∼ 55 {\sim}55 – 58 % 58\% , while D s conv D_{s}^{\rm conv} converges to within 2 % 2\% —demonstrating that remote bands contribute exclusively through interband coherence. The geometric fraction peaks at ∼ 27 {\sim}27 – 33 % 33\% near the ν = ± 2 \nu=\pm 2 fillings where superconductivity is strongest, and is insensitive to gap magnitude in the experimentally relevant range. I Introduction The observation of superconductivity in magic-angle twisted bilayer graphene (MATBG) cao_sc_2018 ignited intense interest in the interplay of band flatness, topology, and unconventional pairing in moiré materials cao_insulator_2018 ; yankowitz_2019 . Nearly flat bands at the magic angle \theta\approx 1.05^{\circ} exhibit a large quantum metric owing to the nontrivial band topology protected by C_{2z}\mathcal{T} symmetry song_2019 ; po_2019 . Since conventional Drude-like superfluid transport scales with band velocity—which vanishes for perfectly flat bands—the question of how MATBG sustains a large superfluid stiffness is intimately connected to quantum geometry. Peotta and Törmä peotta_torma_2015 showed that in an isolated flat band, the superfluid weight is bounded from below by the quantum metric integrated over the Brillouin zone (BZ). This result was extended to multiband systems and fragile topology by Xie, Song, Lian, and Bernevig xie_2020_prl , who proved that the C_{2z}\mathcal{T} -protected winding number in MATBG guarantees a finite geometric superfluid weight. Hu et al. hu_2019 and Julku et al. julku_2020 further generalized these bounds to include interband pairing and realistic band structures. On the experimental side, Tian et al. tian_2023 used circuit quantum electrodynamics (cQED) to measure the superfluid stiffness of MATBG, finding it approximately ten times larger than a conventional BCS estimate—a striking signature of geometric enhancement. Despite the theoretical understanding that quantum geometry contributes to D_{s} in flat-band systems, a systematic quantitative decomposition into conventional and geometric parts for MATBG—with explicit dependence on pairing symmetry, chemical potential, gap magnitude, and number of included bands—has not been reported. In this work, we fill this gap by applying a band-basis current decomposition to the Bistritzer-MacDonald (BM) continuum model and reporting the first BZ-averaged, multi-band convergence study of the geometric fraction. II Formalism II.1 Superfluid weight The superfluid weight tensor for a two-dimensional superconductor is scalapino_1993 \displaystyle D_{s}^{\mu\nu}=-\langle K^{\mu\nu}\rangle-\Lambda^{\mu\nu}(\mathbf{q}=0,\omega=0), (1) where K^{\mu\nu}=\partial^{2}H/\partial k_{\mu}\partial k_{\nu} is the diamagnetic (kinetic energy) kernel and \Lambda^{\mu\nu} is the retarded paramagnetic current-current response evaluated in the superconducting ground state. For the BM continuum model, the single-layer Hamiltonian is H_{\ell}(\mathbf{k})=v_{F}\bm{\sigma}\cdot(\mathbf{k}-\mathbf{K}_{\ell}) , which is linear in \mathbf{k} . Consequently, K^{\mu\nu}=\partial^{2}H/\partial k_{\mu}\partial k_{\nu}=0 identically. This is a well-known property of Dirac-type continuum models kopnin_2008 : the vanishing diamagnetic term means D_{s}=-\Lambda_{xx}(0,0) , and the total superfluid weight computed from the paramagnetic response alone grows without bound as more remote bands are included. In a tight-binding lattice model, K^{\mu\nu}\neq 0 and provides the diamagnetic counterterm that ensures convergence. We address this UV issue below through a controlled band truncation strategy. The paramagnetic response in the Bogoliubov–de Gennes (BdG) eigenbasis reads \displaystyle D_{s}=-\Lambda_{xx}=\frac{1}{A}\sum_{\mathbf{k}}\sum_{a\neq b}\frac{|\langle a,\mathbf{k}|J_{x}|b,\mathbf{k}\rangle|^{2}(f_{a}-f_{b})}{E_{b,\mathbf{k}}-E_{a,\mathbf{k}}}, (2) where |a,\mathbf{k}\rangle are BdG eigenstates with energies E_{a,\mathbf{k}} , f_{a} are Fermi-Dirac occupation factors (we work at T=0 ), A is the system area, and J_{x} is the BdG current operator constructed from the normal-state velocity v_{x}=\partial H/\partial k_{x} . II.2 Band-basis current decomposition We diagonalize the normal-state Hamiltonian at each \mathbf{k} : \displaystyle H(\mathbf{k})=U(\mathbf{k})\,\varepsilon(\mathbf{k})\,U^{\dagger}(\mathbf{k}), (3) with \varepsilon=\mathrm{diag}(\varepsilon_{1},\ldots,\varepsilon_{N}) and U the matrix of eigenvectors. The velocity operator in the band basis is \displaystyle v_{x}^{\rm band}=U^{\dagger}\frac{\partial H}{\partial k_{x}}U=v_{x}^{\rm intra}+v_{x}^{\rm inter}, (4) where [v_{x}^{\rm intra}]_{nn}=\partial\varepsilon_{n}/\partial k_{x} (diagonal, band velocities) and [v_{x}^{\rm inter}]_{nm}=(\varepsilon_{n}-\varepsilon_{m})\mathcal{A}_{nm}^{x} for n\neq m , with \mathcal{A}_{nm}^{x}=\langle u_{n}|\partial_{k_{x}}|u_{m}\rangle the Berry connection. The BdG current operator inherits this decomposition: J_{x}=J_{x}^{\rm intra}+J_{x}^{\rm inter} , where J_{x}^{\rm intra} is diagonal in the normal-state band index and J_{x}^{\rm inter} is off-diagonal. Substituting into Eq. ( 2 ) yields \displaystyle D_{s}=D_{s}^{\rm conv}+D_{s}^{\rm geom}+D_{s}^{\rm cross}, (5) where D_{s}^{\rm conv} contains only intra-band (band velocity) matrix elements, D_{s}^{\rm geom} contains only inter-band (Berry connection) matrix elements, and D_{s}^{\rm cross} contains interference terms. This decomposition is well defined for any finite set of bands, any pairing symmetry, and any chemical potential. It reduces to the Peotta-Törmä formula peotta_torma_2015 in the isolated flat-band, uniform-pairing limit. The interband term D_{s}^{\rm geom} is directly related to the quantum metric: in the flat-band limit ( v_{x}^{\rm intra}\to 0 , band-diagonal pairing), D_{s}^{\rm geom}=(\Delta_{0}/2)\int_{\rm BZ}\mathrm{tr}\,g(\mathbf{k})\,d^{2}k/(2\pi)^{2} peotta_torma_2015 ; torma_review_2022 . III Model and computational details We use the BM continuum model bistritzer_2011 with the relaxation-corrected tunneling parameters of Xie and Bernevig xie_2020_prl : interlayer AA tunneling w_{0}=87.2 meV, AB tunneling w_{1}=109.0 meV, ratio w_{0}/w_{1}=0.80 . These are consistent with the parametrizations of Nam and Koshino nam_2017 ( w_{0}/w_{1}=0.82 ) and Koshino et al. koshino_2018 ( w_{0}/w_{1}=0.70 ). We use Fermi velocity v_{F}=2.135 eV Å and twist angle \theta=1.05^{\circ} . The moiré unit cell area is A_{M}=15{,}606 Å 2 with moiré reciprocal lattice vector G_{M}=0.054 Å -1 . The BM Hamiltonian is constructed with N_{\rm shell}=3 reciprocal lattice shells, yielding a 196\times 196 matrix (per valley, per spin). At the magic angle, this produces two flat bands per valley with bandwidth W=11.2 meV, in good agreement with experiment cao_sc_2018 and previous calculations bistritzer_2011 ; koshino_2018 . For the superconducting state, we consider three pairing symmetries at the mean-field BdG level with gap magnitude \Delta_{0} : (i) uniform s -wave, \Delta_{nm}(\mathbf{k})=\Delta_{0}\delta_{nm} ; (ii) sublattice s -wave, \Delta_{nm}(\mathbf{k})=\Delta_{0}(P_{A})_{nm} with P_{A} projecting onto sublattice A of both layers; (iii) nematic d -wave, \Delta_{nm}(\mathbf{k})=\Delta_{0}[\cos(\mathbf{k}\cdot\mathbf{a}_1)-\cos(\mathbf{k}\cdot\mathbf{a}_2)]\delta_{nm} , where \mathbf{a}_{1,2} are the moiré lattice vectors. All BZ integrals are evaluated on a 14\times 14 Monkhorst-Pack mesh centered at the \Gamma point to avoid divergences of the quantum metric. Band velocities are computed by finite differences ( \delta k=10^{-6} Å -1 ). k-mesh convergence is verified in Sec. IV.5 . Band truncation strategy. —To control the UV divergence inherent in the continuum model, we employ two approaches: 1. Flat-band projection ( n_{\rm keep}=2 ): We retain only the two bands closest to the Fermi energy. In this subspace, both D_{s}^{\rm conv} and D_{s}^{\rm geom} are individually finite. This provides a UV-safe, physically transparent decomposition that captures the intrinsic flat-band physics. 2. Extended truncation scan : We increase n_{\rm keep} from 2 to 6 in steps of 2 and perform BZ-averaged calculations at each truncation level. This reveals how remote bands modify the decomposition and allows us to monitor the convergence of D_{s}^{\rm conv} separately from D_{s}^{\rm geom} . IV Results IV.1 Flat-band decomposition at charge neutrality Table 1 presents the BZ-averaged decomposition at the charge neutrality point ( \mu=0 ) with \Delta_{0}=1 meV and n_{\rm keep}=2 . Results are computed on a converged 14\times 14 k-mesh per valley per spin; the total superfluid weight scales by a factor of 4 (two valleys,