betaFeedback

WorksPhysics and Astronomy

Quantum-geometric dipole: a topological boost to flavor ferromagnetism in flat bands

Abstract

Robust flavor-polarized phases are a striking hallmark of many flat-band moiré materials. In this work, we trace the origin of this spontaneous polarization to a lesser-known quantum-geometric quantity: the quantum-geometric dipole. Analogous to how the quantum metric governs the spatial spread of wavepackets, we show that the quantum-geometric dipole sets the characteristic size of particle-hole excitations, e.g. magnons in a ferromagnet, which in turn boosts their gap and stiffness. Indeed, the larger the particle-hole separation, the weaker the mutual attraction, and the stronger the excitation energy. In topological bands, this energy enhancement admits a lower bound within the local-mode approximation, highlighting the crucial role of topology in flat-band ferromagnetism. We illustrate these effects in microscopic models, emphasizing their generality and relevance to moiré materials. Our results establish the quantum-geometric dipole as a predictive geometric indicator for ferromagnetism in flat bands, a crucial prerequisite for topological order.

Introduction

Quantum geometry provides a lens to understand contributions to macroscopic observables—such as conductivity, polarizability, or optical responses—that elude semiclassical explanations because they arise from how quantum states are “sewn” together [1, 2, 3, 4]. To quantify this sewing, a family of quantum geometric quantities has been introduced. The most useful among them share two key features: (i) they are related to physical observables, e.g. the quantum metric sets a bound on Wannier localization [5] and governs the spin stiffness in flat-band superconductors [6, 7]; and (ii) they possess predictive power, offering material design principles when combined with topology [8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20].

In this work, we promote the quantum-geometric dipole to the status of a core quantum-geometric quantity alongside the Berry curvature and quantum metric. Originally introduced in the context of excitonic photoresponses [21, 22, 23, 24, 25, 26], we refine and extend the concept to show that it satisfies both criteria outlined above. Specifically, we first recall that it measures the typical separation between the particle (pp) and hole (hh) in a phph excitation (Figure 1a) [26, 27, 28]; we then relate it to the gap and stiffness of such excitations (Figure 1b); and finally show that topological invariants impose approximate lower bounds on these quantities, which become exact for flat ideal bands [29, 30, 31, 32, 33, 34, 35, 36, 37], within the local-mode approximation.

Figure 1. a) Schematics of the magnon dipole d\boldsymbol{d} measuring the distance between the ↓\downarrow-particle (pp, blue) and ↑\uparrow-hole (hh, orange) forming the flavor-flipping excitation. b) As the average dipole increases, the attraction between the oppositely charged pp and hh weakens, which decreases the magnon’s binding energy (red) and increases its total energy (green) — see Eq. (5). The quantum-geometric dipole Sgeom\mathcal{S}^{\mathrm{geom}} sets the typical amplitude of the magnon dipole, resulting in larger magnon energies (non-shaded area) that can be lower bounded up to O(1)\mathcal{O}(1) factors by topological invariants — see Eqs. (8) and (9).

The relation between the phph excitation energy and dipole strength admits a simple physical interpretation: as the dipole grows, the oppositely charged pp and hh become more spatially separated, their mutual attraction weakens, and the total energy of the excitation, which given by the Hartree Fock contribution and the (negative) mutual attraction energy, rises. The topological lower bound we derive for the phph energy mirrors that of the Wannier function spread from the quantum metric [38]: just as the quantum metric measures the average spread of an electron around its center of mass, and is bounded below (up to an O(1)\mathcal{O}(1) factor) by the electronic Chern number [35, 39]; the quantum geometric dipole estimates the phph distance and is similarly bounded by the difference of pp and hh Chern numbers. This, in turn, results in an approximate topological lower bound on the phph energy via its dependence on the dipole. Our result provides a natural explanation for the widely observed tendency of topological flat bands to spontaneously polarize under interactions [40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53].

Quantum geometric effects are especially pronounced in flat bands, where semiclassical dynamics vanish and interactions dominate—a regime routinely realized in moiré [54, 55, 56, 57, 58] and other superlattice [59, 60, 61, 62, 63, 64, 65, 66, 67, 68] materials, which host a wide range of correlated and topological phases [69, 70, 71, 72, 73, 74, 75, 76]. For instance, fractionalized topological phases have recently been observed without magnetic fields in twisted MoTe2_2 [44, 45, 46, 47] and penta-layer graphene [49, 53]. Crucial to their realization at experimentally accessible temperatures is a robust and spontaneous time-reversal symmetry breaking [77, 78, 79, 80, 81], enabled by extended ferromagnetic phases. Focusing on twisted MoTe2_2, we numerically demonstrate that this robustness originates from the quantum-geometric dipole.

By clarifying its fundamental properties and crucial role in stabilizing topologically ordered phases in moiré materials, our work establishes the quantum-geometric dipole as a core quantum-geometric quantity. While our primary focus is on magnons relevant to moiré ferromagnets, the scope of the quantum geometric dipole extends far beyond, and could shed light on interaction-driven phenomena for excitons [82], plasmons [83, 84], superconductors [7], or multiband systems [85].

Quantum geometric dipole — The central object of our study is the magnon electric dipole, i.e. the distance between the particle and hole forming the spin-flip excitation. While a similar particle–hole dipole has appeared in the context of excitonic photoresponses [21, 22, 23, 24, 25, 26, 27, 86, 87, 88, 89, 90, 91, 92, 93], our definition clarifies two of its key properties: it separates the quantum-geometric contribution from the non-universal spatial part; and it recasts the quantum-geometric part as a Berry-like flux in mixed spin–momentum space, which offers an efficient, gauge-invariant, and numerically-stable expression à la Fukui-Hatsugai [94, 95] (Eq. (3) below).

To set the stage, consider a generic band with dispersion ϵk,σ\epsilon_{k,\sigma} and periodic Bloch states ∣ukσ⟩\lvert u_k^\sigma\rangle annihilated by ck,σc_{k,\sigma}, with kk in the Brillouin zone (BZ) and σ∈{↑,↓}\sigma\in\{\uparrow,\downarrow\} a flavor index. While our treatment extends to any band filling, we focus on the half-filled case and study magnons created over the ferromagnetic state ∣⇑⟩=∏k∈BZck,↑†∣∅⟩\lvert\Uparrow\rangle= \prod_{k\in\mathrm{BZ}} c^\dagger_{k,\uparrow}\lvert\varnothing\rangle with ∣∅⟩\lvert\varnothing\rangle the vacuum. For ∣⇑⟩\lvert\Uparrow\rangle to be an eigenstate, we require (at least) U(1)-flavor symmetry, i.e. that the Hamiltonian be diagonal in σ\sigma. A generic band-projected flavor-flipping excitation with momentum QQ can be written as

∣ψQ⟩=1NQ∑k∈BZψQ,kck+Q/2,↓†ck−Q/2,↑∣⇑⟩,(1)\lvert\psi_Q\rangle= \frac{1}{\sqrt{N_Q}}\sum_{k\in\mathrm{BZ}}\psi_{Q,k}c^\dagger_{k+Q/2,\downarrow}c_{k-Q/2,\uparrow}\lvert\Uparrow\rangle, \tag*{(1)}
ψQ,k=zQ,ksQ,k,sQ,k=⟨uk+Q/2↓∣uk−Q/2↑⟩,\psi_{Q,k}=z_{Q,k}s_{Q,k},\qquad s_{Q,k}=\langle u^\downarrow_{k+Q/2}\vert u^\uparrow_{k-Q/2}\rangle,

with NQ=∑k∈BZ∣ψQ,k∣2N_Q=\sum_{k\in\mathrm{BZ}}\lvert\psi_{Q,k}\rvert^2 a normalization factor, and where we have decomposed the wavefunction coefficients into a gauge-invariant component zQ,kz_{Q,k} capturing the spatial structure of the magnon, and a spin-flipping form factor sQ,ks_{Q,k} originating from band projection.

With these notations, the magnon electric dipole d\boldsymbol{d}, defined as the average distance between the spin-↓\downarrow particle and spin-↑\uparrow hole, takes the form (see Ref. [27] or Supplemental Material (SM) Sec. A)

d=⟨SQ,kspat+SQ,kgeom⟩∣ψQ∣2,SQ,kspat=i∇klog⁡zQ,k,\boldsymbol{d}=\left\langle\boldsymbol{\mathcal{S}}^{\mathrm{spat}}_{Q,k}+\boldsymbol{\mathcal{S}}^{\mathrm{geom}}_{Q,k}\right\rangle_{\lvert\psi_Q\rvert^2},\qquad\boldsymbol{\mathcal{S}}^{\mathrm{spat}}_{Q,k}=i\nabla_k\log z_{Q,k},
SQ,kgeom=Ak−Q/2↑−Ak+Q/2↓+i∇klog⁡sQ,k,(2)\boldsymbol{\mathcal{S}}^{\mathrm{geom}}_{Q,k}=\boldsymbol{\mathcal{A}}^\uparrow_{k-Q/2}-\boldsymbol{\mathcal{A}}^\downarrow_{k+Q/2}+i\nabla_k\log s_{Q,k}, \tag*{(2)}

with Akσ=−i⟨uk,σ∣∇k∣uk,σ⟩\boldsymbol{\mathcal{A}}^\sigma_k=-i\langle u_{k,\sigma}\vert\nabla_k\vert u_{k,\sigma}\rangle the spin-resolved Berry connection, and where ⟨Ok⟩μ=∑kμkOk/∑kμk\langle O_k\rangle_\mu=\sum_k\mu_kO_k/\sum_k\mu_k denotes the μ\mu-weighted average. Compared to the earlier definition introduced in Ref. [26], we have isolated two distinct contributions to the magnon dipole. The first, dubbed the spatial dipole and denoted as Sspat\boldsymbol{\mathcal{S}}^{\mathrm{spat}}, only depends on the zz-coefficients and therefore originates from the structure of the dipole in real space. The second, Sgeom\boldsymbol{\mathcal{S}}^{\mathrm{geom}}, is the quantum-geometric dipole advertised in the introduction. It is manifestly gauge-invariant [96], and depends solely on the Bloch vectors of the underlying bands. From Eq. (2), Sgeom\boldsymbol{\mathcal{S}}^{\mathrm{geom}} encodes the discrepancy between the Berry connections of spin-↑\uparrow and spin-↓\downarrow bands. One intuitively expects this discrepancy to be larger in a topological phase, thereby enhancing the quantum geometric dipole. The log⁡sQ,k\log s_{Q,k} term furnishes an interband compensating contribution that ensures overall gauge invariance.

To emphasize the spatial/geometric characters of the two terms in Eq. (2), consider the following two limits. When the Bloch bundle is geometrically trivial, ∣ukσ⟩=1\lvert u^\sigma_k\rangle=1, Sgeom\boldsymbol{\mathcal{S}}^{\mathrm{geom}} vanishes and the ph dipole entirely comes from the spatial part Sspat\boldsymbol{\mathcal{S}}^{\mathrm{spat}}. On the other hand, the most localized magnon with center-of-mass momentum QQ, given by ∫d2R ei(Q⋅R)cR,↓†cR,↑∣⇑⟩\int d^2R\,e^{i(Q\cdot R)}c^\dagger_{R,\downarrow}c_{R,\uparrow}\lvert\Uparrow\rangle, corresponds to zQ,k=1z_{Q,k}=1 after projection, leading to Sspat=0\boldsymbol{\mathcal{S}}^{\mathrm{spat}}=0 and a purely quantum-geometric dipole d=⟨Skgeom⟩∣ψ∣2\boldsymbol{d}=\langle\boldsymbol{\mathcal{S}}^{\mathrm{geom}}_k\rangle_{\lvert\psi\rvert^2}.

Beyond these limits, the geometric character of Sgeom\boldsymbol{\mathcal{S}}^{\mathrm{geom}} becomes clear when interpreted as a Berry-like flux. To unveil it, we discretize the BZ in steps of δka=∣δk∣ba\delta k_a=\lvert\delta k\rvert b_a along the reciprocal lattice vectors ba=1,2b_{a=1,2} and approximate derivatives by finite differences to obtain at lowest order in ∣δk∣\lvert\delta k\rvert

δka⋅SQ,kgeom=ilog⁡[Λk↑(δka)sQ,k+δkaΛk↓(δka)sQ,k],(3)\delta k_a\cdot\boldsymbol{\mathcal{S}}^{\mathrm{geom}}_{Q,k} =i\log\left[\frac{\Lambda^\uparrow_k(\delta k_a)s_{Q,k+\delta k_a}}{\Lambda^\downarrow_k(\delta k_a)s_{Q,k}}\right], \tag*{(3)}

where Λkσ(q)=⟨uk+q−Q/2σ∣uk−Q/2σ⟩\Lambda^\sigma_k(q)=\langle u^\sigma_{k+q-Q/2}\vert u^\sigma_{k-Q/2}\rangle are spin-resolved form factors. The arrows on the accompanying sketch represent link variables [94, 95] corresponding to the terms in the logarithm (for Q=0Q=0). The closed loop they form emphasizes the gauge invariance of our discretized expression and its interpretation as a Berry flux — or quantum geometric curvature in mixed spin–momentum space.

Magnon interaction energy — In the introduction, we argued that the interaction-induced magnon gap should scale with the amplitude of the magnon dipole, dd: a larger dipole means more separation between the spin-↓\downarrow particle and spin-↑\uparrow hole, which reduces their mutual attraction and thereby increases the overall magnon energy. To rationalize this intuition, we now compute the magnon energy and compare it with two independent approximate treatments that elucidate its connection to the dipole d\boldsymbol{d} and to the band’s underlying topology. The first is a small-qq expansion of the form factors, which links the quantum geometric quantities, and leads to Eq. (5). The second is a local-mode approximation where a local spin flip is projected onto the band (corresponding to zk=1z_k=1 in Eq. (1)), and yields Eq. (7) when combined with the first approximation.

The interaction energy of the magnon relative to the ferromagnetic state is given by Δ(ψQ)=⟨ψQ∣Hint∣ψQ⟩−⟨⇑∣Hint∣⇑⟩\Delta(\psi_Q)=\langle\psi_Q\vert H_{\mathrm{int}}\vert\psi_Q\rangle-\langle\Uparrow\vert H_{\mathrm{int}}\vert\Uparrow\rangle, for the band-projected interaction Hint=H_{\mathrm{int}}=

∑q,k,pσ,τv(q)2NBZ⟨uk+qσ∣ukσ⟩⟨upτ∣up+qτ⟩ck+q,σ†cp,τ†cp+q,τck,σ,\sum_{q,k,p \atop\sigma,\tau} \frac{v(q)}{2N_{\mathrm{BZ}}} \langle u^\sigma_{k+q}|u^\sigma_k\rangle\langle u^\tau_p|u^\tau_{p+q}\rangle c^\dagger_{k+q,\sigma}c^\dagger_{p,\tau}c_{p+q,\tau}c_{k,\sigma},

with NBZN_{\mathrm{BZ}} the number of points in the discretized BZ, and v(q=∥q∥)v(q = \lVert q\rVert) a rotation-symmetric Coulomb potential. Wick’s theorem yields (see SM Sec. B)

Δ(ψ)=⟨∑qv(q)NBZ∣Λk↑(q)∣2[1−Λk↓(q)∗Λk↑(q)∗ψk+qψk]⟩k,∣ψ∣2,(4)\Delta(\psi) = \left\langle\sum_q \frac{v(q)}{N_{\mathrm{BZ}}} \lvert\Lambda^\uparrow_k(q)\rvert^2 \left[1-\frac{\Lambda^\downarrow_k(q)^*}{\Lambda^\uparrow_k(q)^*}\frac{\psi_{k+q}}{\psi_k}\right]\right\rangle_{k,\lvert\psi\rvert^2}, \tag*{(4)}

where we have left the QQ dependence implicit. The total energy of the magnon also includes a kinetic term

K(ψQ)=⟨ϵk+Q/2,↓−ϵk−Q/2,↑⟩∣ψQ∣2.K(\psi_Q)=\left\langle\epsilon_{k+Q/2,\downarrow}-\epsilon_{k-Q/2,\uparrow}\right\rangle_{\lvert\psi_Q\rvert^2}.

The stability of the state ∣⇑⟩\lvert\Uparrow\rangle against magnonic excitations requires that the Stoner criterion, min⁡zQ[K(ψQ)+Δ(ψQ)]>0\min_{z_Q}[K(\psi_Q)+\Delta(\psi_Q)]>0, be fulfilled at all QQ. For flat bands in which the bandwidth is negligible compared to the interaction scale, we can safely ignore K(ψ)K(\psi).

The connection between the interaction gap Δ(ψ)\Delta(\psi) and the quantum-geometric dipole is brought to light by a small-qq expansion of the form factors — a standard approach for identifying geometric contributions to physical observables [97, 98, 99, 100, 101, 102, 103, 104]. The zeroth order term in the expansion vanishes since the bracketed terms in Eq. (4) cancel out (Λkσ(0)=1\Lambda^\sigma_k(0)=1). The linear terms also vanish due to the rotation invariance of v(q)v(q). Hence, the leading contribution is of order q2q^2, which, by dimensional analysis, requires the introduction of a typical length scale. One may deduce this length scale should be the magnon’s dipole by recalling that ψk∝sk\psi_k\propto s_k and observing that the phase of the bracketed term in Eq. (4) is precisely the spin-momentum Berry flux sketched in Eq. (3) up to the replacement δka→q\delta k_a\to q. Confirming this intuition, the complete expansion (detailed in SM Sec. B) yields

Δ(ψ)=a−2⟨U(gk)[∥Skgeom+Skspat∥2]⟩∣ψ∣2,(5)\Delta(\psi)=a^{-2}\left\langle U(g_k)\left[\lVert S^{\mathrm{geom}}_k+S^{\mathrm{spat}}_k\rVert^2\right]\right\rangle_{\lvert\psi\rvert^2}, \tag*{(5)}

with aa the lattice constant. In this calculation, the decay of the form factors for large-qq (not captured by the small-qq expansion) is enforced by re-exponentiating the terms proportional to the quantum metric (gkσ)ab=Re⁡⟨∂kaukσ∣∂kbukσ⟩−Ak,aσAk,bσ(g^\sigma_k)^{ab}=\operatorname{Re}\langle\partial_{k_a}u^\sigma_k\vert\partial_{k_b}u^\sigma_k\rangle-A^\sigma_{k,a}A^\sigma_{k,b}. Considering these effective momentum cutoffs as independent of spin, we obtain the momentum-dependent interaction scale U(gk)=(4NBZ)−1∑q(qa)2v(q)e−qagkabqbU(g_k)=(4N_{\mathrm{BZ}})^{-1}\sum_q(qa)^2v(q)e^{-q_ag_k^{ab}q_b} with gk=(gk↑+gk↓)/2g_k=(g^\uparrow_k+g^\downarrow_k)/2 the spin-averaged quantum metric. This coefficient has an intuitive interpretation: it describes the typical repulsion between two oppositely charged Gaussian wavepackets of width gkg_k, the minimal spread allowed by the band’s quantum geometry. To see this, let us momentarily specialize to the two-body Coulomb potential v(r)=e2/(4πε∣r∣)v(r)=e^2/(4\pi\varepsilon\lvert r\rvert) and assume an isotropic quantum metric with g~=Tr⁡g\tilde g=\operatorname{Tr}g; then

U(g)=πe2a2ε(8πg~)3/2=πa28∫d2r v(r)∣ϕ(r)∣4,(6)U(g)=\frac{\pi e^2a^2}{\varepsilon(8\pi\tilde g)^{3/2}}=\frac{\pi a^2}{8}\int d^2r\,v(r)\lvert\phi(r)\rvert^4, \tag*{(6)}

with ϕ(r)=exp⁡(−r2/2g~)/2πg~\phi(r)=\exp(-r^2/2\tilde g)/\sqrt{2\pi\tilde g} a normalized Gaussian wavepacket, and where a2a^2 is required by dimensionality.

Eq. 5 confirms the intuitive idea that the magnon energy grows with its dipole. Furthermore, within the local-mode approximation (LMA), the magnon gap is entirely determined by the quantum-geometric dipole. Specifically, the LMA assumes that to decrease the total dipole, the lowest energy magnons are well described by projecting local-in-space spin-flips of the form cR,↓†cR,↑∣⇑⟩c^\dagger_{R,\downarrow}c_{R,\uparrow}\lvert\Uparrow\rangle, i.e. setting zk=1z_k=1, so that ψk=sk\psi_k=s_k. This choice eliminates the spatial dipole (see discussion above Eq. (3)) and isolates the geometric contribution

ΔQgeom=a−2⟨U(gk)∥SQ,kgeom∥2⟩∣s∣2,(7)\Delta^{\mathrm{geom}}_Q=a^{-2}\left\langle U(g_k)\lVert S^{\mathrm{geom}}_{Q,k}\rVert^2\right\rangle_{\lvert s\rvert^2}, \tag*{(7)}

after the small-qq expansion.

Topological boost to flat-band ferromagnetism

Having derived the precise relation between the magnon energy and quantum-geometric dipole in flat bands (Eq. (5)), we now describe how topology underpins robust ferromagnetism in topological flat bands.

Let us start with the SU(2)-spin invariant limit ∣uk,↑⟩=∣uk,↓⟩\lvert u_{k,\uparrow}\rangle=\lvert u_{k,\downarrow}\rangle, where a gapless magnon branch with quadratic dispersion must exist [105, 106, 107, 108]. We recover this gaplessness by noting that SQ=0,kgeom=0S^{\mathrm{geom}}_{Q=0,k}=0 for SU(2) invariant systems (see Eq. (2)). Turning to Eq. (5), Taylor expanding ΔQgeom≃ρs∣Q∣2/2\Delta^{\mathrm{geom}}_Q\simeq\rho_s\lvert Q\rvert^2/2 gives a magnon stiffness

ρs=⟨U(gk)Ωk2⟩BZ,\rho_s=\left\langle U(g_k)\Omega_k^2\right\rangle_{\mathrm{BZ}},

with Ωk=∇k×Ak↑\Omega_k=\nabla_k\times A^\uparrow_k the Berry curvature of the band (see SM Sec. C). When fluctuations of the quantum metric about its mean, gˉ\bar g, can be ignored, this formula reduces to

ρs≃U(gˉ)⟨Ωk2⟩≥U(gˉ)C2,(8)\rho_s\simeq U(\bar g)\langle\Omega_k^2\rangle\ge U(\bar g)C^2, \tag*{(8)}

with C=⟨Ωk⟩BZC=\langle\Omega_k\rangle_{\mathrm{BZ}} the Chern number, and where the inequality follows from Cauchy–Schwarz. This topologically induced lower bound on the spin stiffness qualitatively agrees with previous results [98, 109], and becomes exact for bands with uniform quantum metric [30]. It is derived assuming SU(2) symmetry, but is non-trivial when the Chern number is non-vanishing. Thus, it is not useful when opposite spins are related by time reversal symmetry, but yields a nontrivial bound in, e.g. valleys of moiré materials away from time-reversal invariant momenta, such as the KK valley of twisted bilayer graphene for which analytical results have previously been derived in certain limits [110, 111, 112, 113, 114, 115, 116, 117, 118].

An important advantage of our work is its applicability to spin-orbit coupled systems, where SU(2) symmetry is absent. To this end, we now consider a flavor-symmetry lower than SU(2), for which the magnon spectrum is gapped. We lower bound the magnon gap by a sum of ∣Cs∣\lvert C_s\rvert strictly positive contributions, with Cs=C↑−C↓C_s=C_\uparrow-C_\downarrow the spin Chern number of the underlying bands. This again highlights the role of topology in flat-band ferromagnetism.

The lower bound arises because the spin-lowering overlap sQ,k=⟨uk+Q/2↑∣uk−Q/2↓⟩s_{Q,k}=\langle u^\uparrow_{k+Q/2}\vert u^\downarrow_{k-Q/2}\rangle is a complex function in kk whose phase needs to increase by $2\pi C_sasask winds around the BZ. It hence possesses at least ∣Cs∣|C_s| vortices at momenta {kv}\{k_v\} with vorticity τ=sign⁡Cs\tau= \operatorname{sign} C_s (counted without multiplicities), and can be factorized as sQ,k=s~Q,k∏v(kτ−kvτ)s_{Q,k} = \tilde{s}_{Q,k} \prod_v (k^\tau-k_v^\tau) with kτ=kx+iτkyk^\tau= k_x+i\tau k_y and s~\tilde{s} a complex function with no winding. Near these vortices, the quantum-geometric dipole in Eq. (2) is dominated by the singular connection i∇log⁡sQi\nabla\log s_Q [27], which diverges as ∥SQgeom∥∼∣sQ∣−1\lVert S_Q^{\mathrm{geom}}\rVert\sim|s_Q|^{-1} when sQ→0s_Q \to0. Isolating these singular contributions in Eq. (7), we write ΔQgeom=ΔQbulk+ΔQtopo\Delta_Q^{\mathrm{geom}}=\Delta_Q^{\mathrm{bulk}}+\Delta_Q^{\mathrm{topo}}, where ΔQtopo=∑vU(gkv)\Delta_Q^{\mathrm{topo}}=\sum_v U(g_{k_v}) contains the contributions from the vortices, while ΔQbulk=[∑k≠kvU(gk)∥skSkgeom∥2]/[∑ka2∣sk∣2]\Delta_Q^{\mathrm{bulk}}=\left[\sum_{k\ne k_v}U(g_k)\lVert s_kS_k^{\mathrm{geom}}\rVert^2\right]/\left[\sum_k a^2|s_k|^2\right] gathers the contribution from all momenta away from the vortices. The positivity of the bulk part gives the topological bound

ΔQgeom=ΔQbulk+ΔQtopo≥ΔQtopo≃∣Cs∣U(gˉ),(9)\Delta_Q^{\mathrm{geom}}=\Delta_Q^{\mathrm{bulk}}+\Delta_Q^{\mathrm{topo}}\ge\Delta_Q^{\mathrm{topo}}\simeq|C_s|U(\bar{g}), \tag*{(9)}

where the second (approximate) equality holds under the assumption of a near-uniform quantum metric. Note that this bound is not tight: the mean quantum geometric dipole is of the order of a few times the lattice constant aa, leading to a bulk contribution of the order of ΔQbulk∼U(gˉ)\Delta_Q^{\mathrm{bulk}}\sim U(\bar{g}), which is similar, up to a non-universal multiplicative factor, to the topological term. Nonetheless, it demonstrates that a vanishing gap is forbidden in the topological phase.

Although our derivation of the lower bound assumes a homogeneous quantum geometry gk=gˉg_k=\bar{g}, the numerical results we will show demonstrate that even in the inhomogeneous case, where Eq. (9) no longer provides a strict lower bound, Eqs. (5) and (7) still provide an excellent approximation to the computed gap.

The approximate topological bounds derived for the magnon stiffness (Eq. (8)) and gap (Eq. (9)) show the fundamental role of topology on the stability of ferromagnetism in flat bands. They arose from the tight connection between the quantum-geometric dipole and the magnon spectrum (Eq. (5)). They are consistent with the familiar structure of spin excitations in topologically trivial models, such as the Hubbard model, where a spin-flip can be localized to a single point, which forces the quantum-geometric dipole to vanish and leads to a vanishing gap. This gaplessness in the flat-band limit implies that any additional term in the Hamiltonian (e.g. superexchange) can dominate, favoring different spin-ordering at low temperature (e.g. antiferromagnetic).

Microscopic verification and relevance — We now substantiate our theory by studying two microscopic models. We focus on systems that break SU(2) but preserve U(1) spin symmetry and compute the magnon gap at Q=0Q=0, where the magnon spectrum attains its minimum. To assess the validity of previous approximations, we compute in Figure 2 and Figure 3 the magnon energy of two different models using various methods: the Bethe–Salpeter analysis (SM Eq. (26), blue); the LMA before small-qq expansion (Eq. (4), orange); and the geometric contribution formulas with spatial contribution (Eq. (5), magenta) and without spatial contribution (Eq. (7), red).

Figure 2. Magnon gap versus M/AM/A for the 2D BHZ model, shown in units of U0U_0. The magnon gap is plotted using three different methods: the Bethe–Salpeter analysis (SM Eq. (26), blue); the local-mode approximation, with zk=1z_k=1 (SM Eq. (34), orange); and the geometric contribution formulas (Eq. (5), magenta; Eq. (7), red). The shaded blue (red) regions correspond to topological phases with Chern number C=1C=1 (C=−1C=-1) for spin-↑\uparrow. The inset shows dmin⁡2=⟨∥Skgeom∥2⟩∣s∣2d_{\min}^2=\langle\lVert S_k^{\mathrm{geom}}\rVert^2\rangle_{|s|^2}, the average squared quantum-geometric dipole. Parameters: A=1A=1, B=1B=1, and rξ=0.1r_\xi=0.1.

Magnon gap versus interlayer displacement potential for the continuum model of $\theta$-twisted bilayer MoTe$_2$

Figure 3. Same as Figure 2 for the continuum model of θ\theta-twisted bilayer MoTe2_2. Gray dashed line marks the transition from polarized to unpolarized phases, Δεcexp≈13\Delta\varepsilon_c^{\mathrm{exp}} \approx13 meV, extracted from the MCD measurements of Ref. [44]. Using a relaxed Stoner criterion, our theory estimates the transition point to occur at around Δεcth≈16−17\Delta\varepsilon_c^{\mathrm{th}} \approx16-17 meV, corresponding to the crossing points between the maximal kinetic energy gain Kkin=max⁡(∣εk,↓−εk,↑∣)K_{\mathrm{kin}} = \max(\lvert\varepsilon_{\mathbf{k},\downarrow} - \varepsilon_{\mathbf{k},\uparrow}\rvert) (black line) and the magnon interaction energy (colored lines).

Parameters: θ=3.7∘\theta= 3.7^\circ, ξ=30\xi= 30 nm, ϵ=7\epsilon= 7, and ξ0=0.7\xi_0 = 0.7 nm.

red). To highlight the quantitative relationship between these energies and the quantum-geometric dipole (Figure 1), we also evaluate dmin⁡2=⟨∥Skgeom∥2⟩∣s∣2d_{\min}^2=\langle\lVert S_k^{\mathrm{geom}}\rVert^2\rangle_{|s|^2} (shown in insets). These quantities are presented for two representative models. First, we consider the 2D BHZ model [119] and show that the magnon gap is strong in the topological regime and sharply decreases to a near-zero value upon entering the trivial regime, which qualitatively corroborates the effect of topology identified in Eq. (9). Next, we apply our framework to twisted bilayer MoTe2_2, where spin-valley polarized phases have been observed [43, 44]. At filling ν=1\nu=1, our coarse Stoner criterion predicts a transition to an unpolarized phase at an interlayer displacement field that closely matches experimental observations, highlighting the predictive power of our method.

  • Lattice toy model: The noninteracting part of the 2D BHZ model reads H0=diag⁡{H↑,H↓}H_0=\operatorname{diag}\{H^\uparrow,H^\downarrow\}, where Hσ=(M−2B(2−cos⁡kx−cos⁡ky))σz+Asin⁡kxσx+Asin⁡kyσyH^\sigma=(M-2B(2-\cos k_x-\cos k_y))\sigma_z+A\sin k_x\sigma_x+A\sin k_y\sigma_y. Choosing A=B=1A=B=1, the model is symmetric around M=4M=4 and undergoes topological phase transitions at M=0,4M=0,4 and 88 at which the Chern number shifts as C↑=0→1→−1→0C_\uparrow=0\to1\to-1\to0 (see SM Sec. D). We then project the screened Coulomb interaction, v(q)=U0rξ∣q∣tanh⁡(rξ∣q∣)v(q)=\frac{U_0}{r_\xi|q|}\tanh(r_\xi|q|), onto the lower band and neglect the kinetic energy to isolate interaction effects. Figure 2 shows the magnon gap computed using all three methods at rξ=0.1r_\xi=0.1.

Two striking behaviors emerge from this plot. First, the average squared geometric dipole dmin⁡2d_{\min}^2 evolves with the same trend as the magnon gap. This supports the physical picture developed in Figure 1 and formalized in Eq. (7) that a greater dipole, or spatial separation between the spin-↑\uparrow and spin-↓\downarrow forming the magnon, correlates with the ferromagnetic gap. Second, we notice that the ferromagnetic gap is finite across the topological region of the phase diagram, peaking near the C=1→−1C=1 \to-1 transition at M=4M=4, but sharply diminishes upon entering the trivial phase. This underscores the crucial role of topology in stabilizing ferromagnetic order by enhancing the average quantum-geometric dipole, as explained in Eq. (9) and surrounding text.

Relevance to moiré materials: To explore the interplay between topology and interaction in realistic systems, we turn to twisted bilayer MoTe2_2, where experiments have identified a broad, cone-shaped region of spin-valley polarization in the phase diagram spanned by electron filling and displacement field [43]. A stable spin-valley polarized phase is a prerequisite for realizing exotic correlated states [77, 120], in particular for the fractional Chern insulator recently observed in this system [44, 121].

We focus on the case of filling ν=1\nu=1 at a twist angle of θ=3.7∘\theta=3.7^\circ, using the continuum moiré Hamiltonian (details in SM Sec. D) [122, 123, 124, 125]. An important tunable parameter is the interlayer displacement field DD, which creates an interlayer potential difference Δϵ=Dϵ0ξ0\Delta\epsilon= \frac{D}{\epsilon_0}\xi_0 with ξ0\xi_0 the interlayer spacing and ϵ\epsilon the dielectric constant. The long-range Coulomb interaction is screened by dual metal gates, resulting in an effective potential of the form v(q)=e22ϵ0ϵtanh⁡(ξ∣q∣)∣q∣v(q)=\frac{e^2}{2\epsilon_0\epsilon}\frac{\tanh(\xi\lvert q\rvert)}{\lvert q\rvert}, where ξ\xi is the distance from the bilayer to the gates. We adopt representative experimental parameters: ξ=30\xi=30 nm, ϵ=7\epsilon=7, and ξ0=0.7\xi_0=0.7 nm.

Figure 3 presents the magnon gap versus interlayer displacement potential, and demonstrates a remarkable agreement among the four methods discussed above. Similar to the BHZ model, the difference between the topological and trivial regimes is stark: the magnon interaction energy drops to less than one-third of its initial value at Δϵ=0\Delta\epsilon=0 upon crossing the topological transition indicated by the vertical green dashed line.

Experimentally, a transition from a ferromagnetic to an unpolarized phase is observed as a function of increasing displacement field, which we estimate around Δϵcexp≈13\Delta\epsilon_c^{\mathrm{exp}}\approx13 meV (gray dashed line) using Ref. [44]'s data. To evaluate this transition point theoretically, we employ a relaxed Stoner criterion by approximating the kinetic contribution to the magnon energy as Kkin=max⁡(∣ϵk,↓−ϵk,↑∣)K_{\mathrm{kin}}=\max(\lvert\epsilon_{\mathbf{k},\downarrow}-\epsilon_{\mathbf{k},\uparrow}\rvert), shown as the black solid line in Figure 3, an upper bound on the kinetic energy cost K(ϕ)K(\phi) defined above. The ferromagnetic transition is expected to occur at the intersection between the kinetic energy cost and the magnon gap. This occurs around Δϵcth=16−17\Delta\epsilon_c^{\mathrm{th}}=16-17 meV for all three methods used in Figure 3, which closely matches Δϵcexp\Delta\epsilon_c^{\mathrm{exp}}.

These two microscopic examples underscore the central role of the quantum-geometric dipole in the emergence of ferromagnetism. They support our approximate topological lower bound, which accounts for the widely observed tendency of topological bands to spontaneously polarize under interactions. Moreover, the strong agreement between our predictions and the depolarization transition point in twisted MoTe2_2 highlights the quantitative accuracy and predictive power of our approach. The geometric dipole of the magnon should also give rise to measurable transport phenomena, analogous to excitonic drift and shift currents [26, 27, 126], although these effects would manifest at different excitation energy scales – the energy estimates in Figure 3 indicate the terahertz regime.

Finally, we note that for pure contact interactions the Bethe–Salpeter equation factorizes, allowing the magnon spectrum to be computed directly in orbital space (SM Sec. E). For toy models with only a few orbitals, this leads to drastic numerical gains. In this contact limit, and under the assumption of strictly local spin flips and/or constraints analogous to those used in Refs. [7, 127], one may even expect to derive analytical expressions consistent with our results. We leave such an analysis to future work.

Acknowledgments — VC is indebted to C. Paiva for an insightful presentation of her results [27] while the manuscript was being written. We also acknowledge insightful conversations with S. Divic and M. Goerbig. LC, SAAG and JC acknowledge support from the Air Force Office of Scientific Research under Grants No. FA9550-20-1-0260 and FA9550-24-1-0222. SAAG and JC acknowledge support from the Alfred P. Sloan Foundation through a Sloan Research Fellowship. This work was performed in part at the Aspen Center for Physics, which is supported by National Science Foundation grant PHY-2210452 and by a grant from the Simons Foundation (1161654, Troyer) The Flatiron Institute is a division of the Simons Foundation.

Supplemental Material

Quantum geometric dipole

In this appendix, we derive Eq. (2) of the main text by providing the real-space representation of the magnon wavefunction (Eq. (1)) and computing the average distance between the hole and electron that form this excitation.

Real-space representation: We consider a system whose ground state is a fully polarized ferromagnet, denoted by ∣⇑⟩=∏k∈BZck,↑†∣∅⟩|\Uparrow\rangle= \prod_{k\in\mathrm{BZ}} c_{k,\uparrow}^{\dagger}|\varnothing\rangle as defined in the main text. Let us denote the operator annihilating a fermion of spin σ\sigma at position rr as cr,σc_{r,\sigma}, whose real-space orbital is a delta function at position rr, and write ∣r↓,r↑⟩=cr↓,↓†cr↑,↑†∣⇑⟩|r_{\downarrow},r_{\uparrow}\rangle= c_{r_{\downarrow},\downarrow}^{\dagger}c_{r_{\uparrow},\uparrow}^{\dagger}|\Uparrow\rangle. Recalling that the ck,σc_{k,\sigma} operators used in the main text are the fermionic operator for a specific band corresponding to the eigenstates eik⋅r∣ukσ⟩e^{i k\cdot r}|u_{k}^{\sigma}\rangle, and that the anti-commutation relation between fermionic operators is equal to the scalar product of their respective orbitals, it follows that {cr,σ,ck,σ′†}=δσ,σ′⟨r∣ukσ⟩eik⋅r\{c_{r,\sigma},c_{k,\sigma'}^{\dagger}\}=\delta_{\sigma,\sigma'}\langle r|u_{k}^{\sigma}\rangle e^{i k\cdot r} where ⟨r∣ukσ⟩\langle r|u_{k}^{\sigma}\rangle denotes the real-space representation of the periodic function ∣ukσ⟩|u_{k}^{\sigma}\rangle. For convenience, we repeat here the generic band-projected magnon state introduced in the main text:

∣ψQ⟩=1NQ∑k∈BZψQ,kck+Q/2,↓†ck−Q/2,↑∣⇑⟩,ψQ,k=zQ,ksQ,k,sQ,k=⟨uk+Q/2↓∣uk−Q/2↑⟩(10)|\psi_Q\rangle= \frac{1}{\sqrt{N_Q}}\sum_{k\in\mathrm{BZ}}\psi_{Q,k}c_{k+Q/2,\downarrow}^{\dagger}c_{k-Q/2,\uparrow}|\Uparrow\rangle,\qquad\psi_{Q,k}=z_{Q,k}s_{Q,k},\qquad s_{Q,k}=\langle u_{k+Q/2}^{\downarrow}|u_{k-Q/2}^{\uparrow}\rangle \tag*{(10)}

with NQ=∑k∈BZ∣ψQ,k∣2N_Q=\sum_{k\in\mathrm{BZ}}|\psi_{Q,k}|^2 a normalization factor. This state can be expressed in real-space as

⟨r↓,r↑∣ψQ⟩=∑k∈BZψQ,kNQ⟨⇑∣cr↑,↑†cr↓,↓ck+Q/2,↓†ck−Q/2,↑∣⇑⟩(11)\langle r_{\downarrow},r_{\uparrow}|\psi_Q\rangle=\sum_{k\in\mathrm{BZ}}\frac{\psi_{Q,k}}{\sqrt{N_Q}}\langle\Uparrow|c_{r_{\uparrow},\uparrow}^{\dagger}c_{r_{\downarrow},\downarrow}c_{k+Q/2,\downarrow}^{\dagger}c_{k-Q/2,\uparrow}|\Uparrow\rangle \tag*{(11)}
=∑k∈BZψQ,kNQei(k⋅r+Q⋅R)⟨r↓∣uk+Q/2↓⟩⟨r↑∣uk−Q/2↑⟩∗,R=r↑+r↓2,r=r↓−r↑,(12)=\sum_{k\in\mathrm{BZ}}\frac{\psi_{Q,k}}{\sqrt{N_Q}}e^{i(k\cdot r+Q\cdot R)}\langle r_{\downarrow}|u_{k+Q/2}^{\downarrow}\rangle\langle r_{\uparrow}|u_{k-Q/2}^{\uparrow}\rangle^{*},\qquad R=\frac{r_{\uparrow}+r_{\downarrow}}{2},\quad r=r_{\downarrow}-r_{\uparrow}, \tag*{(12)}

where we have used the anti-commutation relation discussed above, and introduced the center of mass RR and relative rr coordinates.

Average particle-hole dipole: Reproducing the calculation from Ref. [27] for completeness, we can now compute the expectation value of the relative distance as

d=⟨ψ∣r∣ψ⟩=∫d2r↓d2r↑⟨ψ∣r↓,r↑⟩r⟨r↓,r↑∣ψ⟩(13)d = \langle\psi\lvert\mathbf{r} \rvert\psi\rangle= \int d^{2}r_{\downarrow}d^{2}r_{\uparrow}\langle\psi\lvert r_{\downarrow},r_{\uparrow}\rangle\mathbf{r}\langle r_{\downarrow},r_{\uparrow}\rvert\psi\rangle \tag*{(13)}
=1N∑k,k′∫d2r↓d2r↑ψk′∗ψk⟨uk′+Q2,↓∣r↓⟩⟨r↑∣uk′−Q2,↓⟩⟨r↓∣uk+Q2,↑⟩⟨uk−Q2,↑∣r↑⟩e−i(k′⋅r)ei(k⋅r)(14)= \frac{1}{N}\sum_{k,k'}\int d^{2}r_{\downarrow}d^{2}r_{\uparrow}\psi_{k'}^{*}\psi_{k}\langle u_{k'+\frac{Q}{2},\downarrow}\lvert r_{\downarrow}\rangle\langle r_{\uparrow}\lvert u_{k'-\frac{Q}{2},\downarrow}\rangle\langle r_{\downarrow}\lvert u_{k+\frac{Q}{2},\uparrow}\rangle\langle u_{k-\frac{Q}{2},\uparrow}\lvert r_{\uparrow}\rangle e^{-i(k'\cdot r)}e^{i(k\cdot r)} \tag*{(14)}
=1N∑k,k′∫d2r↓d2r↑ψk′∗ψk⟨uk′+Q2,↓∣r↓⟩⟨r↑∣uk′−Q2,↓⟩⟨r↓∣uk+Q2,↑⟩⟨uk−Q2,↑∣r↑⟩e−i(k′⋅r)(−i∇k)ei(k⋅r)(15)= \frac{1}{N}\sum_{k,k'}\int d^{2}r_{\downarrow}d^{2}r_{\uparrow}\psi_{k'}^{*}\psi_{k}\langle u_{k'+\frac{Q}{2},\downarrow}\lvert r_{\downarrow}\rangle\langle r_{\uparrow}\lvert u_{k'-\frac{Q}{2},\downarrow}\rangle\langle r_{\downarrow}\lvert u_{k+\frac{Q}{2},\uparrow}\rangle\langle u_{k-\frac{Q}{2},\uparrow}\lvert r_{\uparrow}\rangle e^{-i(k'\cdot r)}(-i\nabla_{k})e^{i(k\cdot r)} \tag*{(15)}
=1N∑k,k′∫d2r↓d2r↑ψk′∗⟨uk′+Q2,↓∣r↓⟩⟨r↑∣uk′−Q2,↓⟩ei(k−k′)⋅r(i∇k)[ψk⟨r↓∣uk+Q2,↑⟩⟨uk−Q2,↑∣r↑⟩]∗(16)= \frac{1}{N}\sum_{k,k'}\int d^{2}r_{\downarrow}d^{2}r_{\uparrow}\psi_{k'}^{*}\langle u_{k'+\frac{Q}{2},\downarrow}\lvert r_{\downarrow}\rangle\langle r_{\uparrow}\lvert u_{k'-\frac{Q}{2},\downarrow}\rangle e^{i(k-k')\cdot r}(i\nabla_{k})\left[\psi_{k}\langle r_{\downarrow}\lvert u_{k+\frac{Q}{2},\uparrow}\rangle\langle u_{k-\frac{Q}{2},\uparrow}\lvert r_{\uparrow}\rangle\right]^{*} \tag*{(16)}
=iN∑k,k′ψk′∗Ok−k′(uk′+Q2,↓†,uk+Q2,↑)Ok′−k(uk−Q2,↑†,uk′−Q2,↓)∇kψk+iN∑k,k′∣ψk∣2Ok−k′(uk′+Q2,↓†,∇kuk+Q2,↑)Ok′−k(uk−Q2,↑†,uk′−Q2,↓)+iN∑k,k′∣ψk∣2Ok−k′(uk′+Q2,↓†,uk+Q2,↑)Ok′−k(∇kuk−Q2,↑†,uk′−Q2,↓),(17)\begin{aligned} ={}&\frac{i}{N}\sum_{k,k'}\psi_{k'}^{*}\mathcal{O}_{k-k'}\left(u_{k'+\frac{Q}{2},\downarrow}^{\dagger},u_{k+\frac{Q}{2},\uparrow}\right)\mathcal{O}_{k'-k}\left(u_{k-\frac{Q}{2},\uparrow}^{\dagger},u_{k'-\frac{Q}{2},\downarrow}\right)\nabla_{k}\psi_{k}\\ &+\frac{i}{N}\sum_{k,k'}\lvert\psi_{k}\rvert^{2}\mathcal{O}_{k-k'}\left(u_{k'+\frac{Q}{2},\downarrow}^{\dagger},\nabla_{k}u_{k+\frac{Q}{2},\uparrow}\right)\mathcal{O}_{k'-k}\left(u_{k-\frac{Q}{2},\uparrow}^{\dagger},u_{k'-\frac{Q}{2},\downarrow}\right)\\ &+\frac{i}{N}\sum_{k,k'}\lvert\psi_{k}\rvert^{2}\mathcal{O}_{k-k'}\left(u_{k'+\frac{Q}{2},\downarrow}^{\dagger},u_{k+\frac{Q}{2},\uparrow}\right)\mathcal{O}_{k'-k}\left(\nabla_{k}u_{k-\frac{Q}{2},\uparrow}^{\dagger},u_{k'-\frac{Q}{2},\downarrow}\right), \tag*{(17)} \end{aligned}

where we have used integration by parts between lines Eq. (15) and Eq. (16), and have also defined the overlap Oq(ϕ1,ϕ2)=∫d2r⟨ϕ2∣r⟩eir⋅q⟨r∣ϕ1⟩\mathcal{O}_{q}(\phi_{1},\phi_{2})=\int d^{2}r\langle\phi_{2}\lvert r\rangle e^{i\mathbf{r}\cdot\mathbf{q}}\langle r\lvert\phi_{1}\rangle between any unit-cell periodic functions ϕ1,2\phi_{1,2}. Because the periodic functions carry no crystal momentum, it must be that Oq(ϕ1,ϕ2)∝δq\mathcal{O}_{q}(\phi_{1},\phi_{2})\propto\delta_{q}. Then the resolution of the identity yields Oq(ϕ1,ϕ2)=δq⟨ϕ2∣(∫d2r∣r⟩⟨r∣)∣ϕ1⟩=δq⟨ϕ2∣ϕ1⟩\mathcal{O}_{q}(\phi_{1},\phi_{2})=\delta_{q}\langle\phi_{2}\lvert(\int d^{2}r\lvert r\rangle\langle r\rvert)\rvert\phi_{1}\rangle=\delta_{q}\langle\phi_{2}\lvert\phi_{1}\rangle. This gives

d=iN∑kψk∗∇kψk+∣ψk∣2[⟨uk+Q2,↓∣∇kuk+Q2,↓⟩+⟨∇kuk−Q2,↑∣uk−Q2,↑⟩](18)d=\frac{i}{N}\sum_{k}\psi_{k}^{*}\nabla_{k}\psi_{k}+\lvert\psi_{k}\rvert^{2}\left[\langle u_{k+\frac{Q}{2},\downarrow}\lvert\nabla_{k}u_{k+\frac{Q}{2},\downarrow}\rangle+\langle\nabla_{k}u_{k-\frac{Q}{2},\uparrow}\lvert u_{k-\frac{Q}{2},\uparrow}\rangle\right] \tag*{(18)}
=iN∑kψk∗∇kψk+∣ψk∣2[⟨uk+Q2,↓∣∇kuk+Q2,↓⟩−⟨uk−Q2,↑∣∇kuk−Q2,↑⟩](19)=\frac{i}{N}\sum_{k}\psi_{k}^{*}\nabla_{k}\psi_{k}+\lvert\psi_{k}\rvert^{2}\left[\langle u_{k+\frac{Q}{2},\downarrow}\lvert\nabla_{k}u_{k+\frac{Q}{2},\downarrow}\rangle-\langle u_{k-\frac{Q}{2},\uparrow}\lvert\nabla_{k}u_{k-\frac{Q}{2},\uparrow}\rangle\right] \tag*{(19)}
=∑k∣ψk∣2N[i∇klog⁡zk+i∇klog⁡sk−Ak+Q2↓+Ak−Q2↑]=∑k∣ψk∣2N[Skspat+Skgeom],(20)=\sum_{k}\frac{\lvert\psi_{k}\rvert^{2}}{N}\left[i\nabla_{k}\log z_{k}+i\nabla_{k}\log s_{k}-A_{k+\frac{Q}{2}}^{\downarrow}+A_{k-\frac{Q}{2}}^{\uparrow}\right]=\sum_{k}\frac{\lvert\psi_{k}\rvert^{2}}{N}\left[S_{k}^{\mathrm{spat}}+S_{k}^{\mathrm{geom}}\right], \tag*{(20)}

where we have omitted the dependence on the center of mass momentum QQ and used the fact that Bloch vectors are normalized to go from Eq. (18) to Eq. (19). This gives Eq. (2) in the main text.

Magnon gap and quantum geometric dipole

This appendix contains all details relevant to the numerical simulations presented in Figure 2 and Figure 3 of the main text, and the key steps necessary to derive our main analytical results (Eqs. (4) - (9)). In more details

  • Sec. B 1 derives the Bethe-Salpeter equation for collective excitations.

  • Sec. B 2 details the local-mode approximation that provides an analytically tractable approximation of the lowest-lying magnon excitation.

  • Sec. B 3 performs the small-qq expansion sketched in the main text and derives the explicit expression linking the magnon gap to the quantum geometric dipole (Eq. (5)).

Bethe-Salpeter formula and spin-flip spectrum

We start from a generic interacting Hamiltonian projected onto the bands of interest

H=H0+Hint,H0=∑k,σϵk,σck,σ†ck,σ,Hint=12∑k,p,q,στWkpστ(q)ck+q,σ†cp−q,τ†cp,τck,σ(21)H=H_{0}+H_{\mathrm{int}},\qquad H_{0}=\sum_{k,\sigma}\epsilon_{k,\sigma}c_{k,\sigma}^{\dagger}c_{k,\sigma},\qquad H_{\mathrm{int}}=\frac{1}{2}\sum_{k,p,q,\sigma\tau}W_{kp}^{\sigma\tau}(q)c_{k+q,\sigma}^{\dagger}c_{p-q,\tau}^{\dagger}c_{p,\tau}c_{k,\sigma} \tag*{(21)}

where the interaction matrix elements Wkpστ(q)W^{\sigma\tau}_{\mathbf{k}\mathbf{p}}(\mathbf{q}) depend on the band form factors Γk,σ(q)=⟨uk+qσ∣ukσ⟩\Gamma_{\mathbf{k},\sigma}(\mathbf{q})=\langle u^{\sigma}_{\mathbf{k}+\mathbf{q}}|u^{\sigma}_{\mathbf{k}}\rangle and the Fourier transform of the (rotation invariant) interaction potential v(q=∥q∥)v(q=\lVert\mathbf{q}\rVert) as

Wkpστ(q)=v(q)NBZΓk,σ(q)Γp,τ∗(q)=v(q)NBZ⟨uk+qσ∣ukσ⟩⟨upτ∣up+qτ⟩.(22)W^{\sigma\tau}_{\mathbf{k}\mathbf{p}}(\mathbf{q})=\frac{v(q)}{N_{\mathrm{BZ}}}\Gamma_{\mathbf{k},\sigma}(\mathbf{q})\Gamma^{*}_{\mathbf{p},\tau}(\mathbf{q})=\frac{v(q)}{N_{\mathrm{BZ}}}\langle u^{\sigma}_{\mathbf{k}+\mathbf{q}}|u^{\sigma}_{\mathbf{k}}\rangle\langle u^{\tau}_{\mathbf{p}}|u^{\tau}_{\mathbf{p}+\mathbf{q}}\rangle. \tag*{(22)}

Projecting the Hamiltonian onto the magnon basis spanned by ∣SQ,k−⟩|S^{-}_{Q,\mathbf{k}}\rangle, as introduced in Eq. (10), we have

Hk,k′(Q)=⟨SQ,k−∣H0+Hint∣SQ,k′−⟩−⟨⇑∣H0+Hint∣⇑⟩δk,k′=(Ek+,↓−Ek−,↑)δk,k′−Wk−,k+↑↓(k′−k),k±=k±Q/2,(23)\begin{aligned} \mathcal{H}_{\mathbf{k},\mathbf{k}'}(Q)=\langle S^{-}_{Q,\mathbf{k}}|H_{0}+H_{\mathrm{int}}|S^{-}_{Q,\mathbf{k}'}\rangle-\langle\Uparrow|H_{0}+H_{\mathrm{int}}|\Uparrow\rangle\delta_{\mathbf{k},\mathbf{k}'} \\ &=(E_{\mathbf{k}_{+},\downarrow}-E_{\mathbf{k}_{-},\uparrow})\delta_{\mathbf{k},\mathbf{k}'}-W^{\uparrow\downarrow}_{\mathbf{k}_{-},\mathbf{k}_{+}}(\mathbf{k}'-\mathbf{k}),\qquad\mathbf{k}_{\pm}=\mathbf{k}\pm Q/2, \tag*{(23)} \end{aligned}

where the Hartree-Fock quasiparticle energies are given by

Ek,↓=ϵk,↓+[∑pWkp↓↓(0)],Ek,↑=ϵk,↑+[∑pWkp↑↑(0)−∑pWpp↑↑(k−p)].(24)E_{\mathbf{k},\downarrow}=\epsilon_{\mathbf{k},\downarrow}+\left[\sum_{\mathbf{p}}W^{\downarrow\downarrow}_{\mathbf{k}\mathbf{p}}(0)\right],\qquad E_{\mathbf{k},\uparrow}=\epsilon_{\mathbf{k},\uparrow}+\left[\sum_{\mathbf{p}}W^{\uparrow\uparrow}_{\mathbf{k}\mathbf{p}}(0)-\sum_{\mathbf{p}}W^{\uparrow\uparrow}_{\mathbf{p}\mathbf{p}}(\mathbf{k}-\mathbf{p})\right]. \tag*{(24)}

under the assumption that ∣⇑⟩|\Uparrow\rangle is the ground state. Note that H\mathcal{H} is diagonal in the center of mass momentum QQ of the excitation due to translation invariance. Noticing that the Wp,kστ(0)=v(0)W^{\sigma\tau}_{\mathbf{p},\mathbf{k}}(0)=v(0) is independent of spin and momentum, the Hartree terms cancel, so that:

Ek+,↓−Ek−,↑=ϵk+,↓−ϵk−,↑+∑pWpp↑↑(k−−p),(25)E_{\mathbf{k}_{+},\downarrow}-E_{\mathbf{k}_{-},\uparrow}=\epsilon_{\mathbf{k}_{+},\downarrow}-\epsilon_{\mathbf{k}_{-},\uparrow}+\sum_{\mathbf{p}}W^{\uparrow\uparrow}_{\mathbf{p}\mathbf{p}}(\mathbf{k}_{-}-\mathbf{p}), \tag*{(25)}

The magnon spectrum is obtained by diagonalizing Hk,k′(Q)\mathcal{H}_{\mathbf{k},\mathbf{k}'}(Q), i.e. solving the so-called Bethe-Salpeter equation

∑k′Hk,k′(Q)ψQ,k′=E(Q)ψQ,k,(26)\sum_{\mathbf{k}'}\mathcal{H}_{\mathbf{k},\mathbf{k}'}(Q)\psi_{Q,\mathbf{k}'}=\mathcal{E}(Q)\psi_{Q,\mathbf{k}}, \tag*{(26)}

which yields the collective excitation energies E(Q)\mathcal{E}(Q), and provides the coefficients ψk,Q\psi_{\mathbf{k},Q} of the corresponding magnon wavefunctions (Eq. (10)). In the flat band limit, ϵk+,↓=ϵk−,↑=0\epsilon_{\mathbf{k}_{+},\downarrow}=\epsilon_{\mathbf{k}_{-},\uparrow}=0, the projected Hamiltonian reduced to

Hk,k′(Q)=(∑pWpp↑↑(k−−p))δk,k′−Wk−,k+↑↓(k′−k).(27)\mathcal{H}_{\mathbf{k},\mathbf{k}'}(Q)=\left(\sum_{\mathbf{p}}W^{\uparrow\uparrow}_{\mathbf{p}\mathbf{p}}(\mathbf{k}_{-}-\mathbf{p})\right)\delta_{\mathbf{k},\mathbf{k}'}-W^{\uparrow\downarrow}_{\mathbf{k}_{-},\mathbf{k}_{+}}(\mathbf{k}'-\mathbf{k}). \tag*{(27)}

where the first term corresponds to the interaction-driven Hartree-Fock splitting, while the second term represents the mutual attraction between the particle and hole. In Figure 2 and Figure 3, we focus on the spectrum at Q=0Q=0, and extract the magnon gap from the lowest eigenvalue of Eq. (26) (curves labeled “Bethe Salpeter” in Figure 2 and Figure 3).

Local mode approximation

The lowest magnon energy can also be obtained by minimizing the expectation value of H\mathcal{H} with respect to the coefficients of the magnon wavefunction (we omit the subscript QQ for clarity when there is no ambiguity)

Emin⁡(Q)=min⁡{ψ}Δ(ψ),Δ(ψ)=1N∑kk′ψk∗Hk,k′(Q)ψk′,N=∑k∣ψk∣2.(28)\mathcal{E}_{\min}(Q)=\min_{\{\psi\}}\Delta(\psi),\qquad\Delta(\psi)=\frac{1}{N}\sum_{\mathbf{k}\mathbf{k}'}\psi^{*}_{\mathbf{k}}\mathcal{H}_{\mathbf{k},\mathbf{k}'}(Q)\psi_{\mathbf{k}'},\qquad N=\sum_{\mathbf{k}}|\psi_{\mathbf{k}}|^{2}. \tag*{(28)}

Combing Eqs. (23) – (25) yields Δ(ψ)=K+I\Delta(\psi)=K+I, where

K=1N∑k∣ψk∣2(ϵk+,↓−ϵk−,↑),(29)K=\frac{1}{N}\sum_{\mathbf{k}}|\psi_{\mathbf{k}}|^{2}(\epsilon_{\mathbf{k}_{+},\downarrow}-\epsilon_{\mathbf{k}_{-},\uparrow}), \tag*{(29)}
I=1N∑k[∣ψk∣2∑pWpp↑↑(k−−p)]−1N∑kk′[Wk−,k+↑↓(k′−k)ψk∗ψk′](30)I=\frac{1}{N}\sum_{\mathbf{k}}\left[|\psi_{\mathbf{k}}|^{2}\sum_{\mathbf{p}}W^{\uparrow\uparrow}_{\mathbf{p}\mathbf{p}}(\mathbf{k}_{-}-\mathbf{p})\right]-\frac{1}{N}\sum_{\mathbf{k}\mathbf{k}'}\left[W^{\uparrow\downarrow}_{\mathbf{k}_{-},\mathbf{k}_{+}}(\mathbf{k}'-\mathbf{k})\psi^{*}_{\mathbf{k}}\psi_{\mathbf{k}'}\right] \tag*{(30)}

and k±=k±Q/2\mathbf{k}_{\pm}=\mathbf{k}\pm Q/2. Here, KK captures the kinetic energy from the non-interacting band structure, while all interaction effects are gathered in II. The stability of the ferromagnetic ground state against spin-flip excitations requires a Stoner-like criterion I+K>0I+K>0 for any choice of ψ\psi and all momenta QQ. In Figure 3 of the main text (where we only show Q=0Q=0), we adopt a relaxed version of this criterion by comparing II with the maximal kinetic cost Kkin=max⁡(∣ϵk,↓−ϵk,↑∣)K_{\mathrm{kin}} = \max(\lvert\epsilon_{k,\downarrow} - \epsilon_{k,\uparrow}\rvert), which provides an upper bound of KK.

In the flat-band limit, K=0K = 0, and the gap is entirely given by the interaction contribution II. Expanding Eq. (30) with the explicit form of the interaction kernel (Eq. (22)) and splitting ψk=zksk\psi_k = z_k s_k gives

Δ(ψ)=I=1N∑k,qv(q)NBZ[∣ψk⟨uk−+q↑∣uk−↑⟩∣2−ψk∗ψk+q⟨uk−+q↑∣uk−↑⟩⟨uk−↓∣uk−+q↓⟩],(31)\Delta(\psi) = I = \frac{1}{N} \sum_{k,q} \frac{v(q)}{N_{\mathrm{BZ}}} \left[ \lvert\psi_k \langle u^\uparrow_{k_-+q} \vert u^\uparrow_{k_-} \rangle\rvert^2 - \psi_k^* \psi_{k+q} \langle u^\uparrow_{k_-+q} \vert u^\uparrow_{k_-} \rangle\langle u^\downarrow_{k_-} \vert u^\downarrow_{k_-+q} \rangle\right], \tag*{(31)}
=1N∑k∣ψk∣2∑qv(q)∣⟨uk−+q↑∣uk−↑⟩∣2NBZ[1−ψk+qψk⟨uk+↓∣uk++q↓⟩⟨uk−+q↑∣uk−↑⟩],(32)= \frac{1}{N} \sum_k \lvert\psi_k \rvert^2 \sum_q \frac{v(q) \lvert\langle u^\uparrow_{k_-+q} \vert u^\uparrow_{k_-} \rangle\rvert^2}{N_{\mathrm{BZ}}} \left[ 1 - \frac{\psi_{k+q}}{\psi_k} \frac{\langle u^\downarrow_{k_+} \vert u^\downarrow_{k_++q} \rangle}{\langle u^\uparrow_{k_-+q} \vert u^\uparrow_{k_-} \rangle} \right], \tag*{(32)}
=1N∑k∣ψk∣2∑qv(q)∣Λk↑(q)∣2NBZ[1−ψk+qψkΛk↓(q)∗Λk↑(q)∗],(33)= \frac{1}{N} \sum_k \lvert\psi_k \rvert^2 \sum_q \frac{v(q) \lvert\Lambda_k^\uparrow(q) \rvert^2}{N_{\mathrm{BZ}}} \left[ 1 - \frac{\psi_{k+q}}{\psi_k} \frac{\Lambda_k^\downarrow(q)^*}{\Lambda_k^\uparrow(q)^*} \right], \tag*{(33)}

where we have used the notation Λkσ(q)=⟨uk+q/2−σq/2σ∣uk−q/2σ⟩\Lambda_k^\sigma(q) = \langle u^\sigma_{k+q/2-\sigma q/2} \vert u^\sigma_{k-q/2} \rangle, with σ=±1\sigma= \pm1 for spin-↑\uparrow / ↓\downarrow (introduced in the main text) to match the form of Eq. (4).

The local-mode approximation discussed in the main text takes zk=1z_k = 1, yielding

ΔLMA(ψ)=1N∑k∣sk∣2∑qv(q)∣Λk↑(q)∣2NBZ[1−sk+qskΛk↓(q)∗Λk↑(q)∗],(34)\Delta^{\mathrm{LMA}}(\psi) = \frac{1}{N} \sum_k \lvert s_k \rvert^2 \sum_q \frac{v(q) \lvert\Lambda_k^\uparrow(q) \rvert^2}{N_{\mathrm{BZ}}} \left[ 1 - \frac{s_{k+q}}{s_k} \frac{\Lambda_k^\downarrow(q)^*}{\Lambda_k^\uparrow(q)^*} \right], \tag*{(34)}

which corresponds to the curves labeled “local mode approximation” in Figure 2 and Figure 3.

Relation between magnon gap and quantum geometric dipole

We further approximate Δ(ψ)\Delta(\psi) by neglecting large-qq processes. This is justified because both the form factor magnitudes and the screened Coulomb interaction decay with increasing momentum transfer qq. Specifically, we expand the form factors to second order in qq to capture the small-qq behavior, and re-exponentiate the parts that includes the quantum metric to ensure fast decay for large-qq. For instance, the norm of the spin-conserving form factors approximately behaves as

∣⟨uk+qσ∣ukσ⟩∣2=⟨uk+qσ∣ukσ⟩⟨ukσ∣uk+qσ⟩(35)\lvert\langle u^\sigma_{k+q} \vert u^\sigma_k \rangle\rvert^2 = \langle u^\sigma_{k+q} \vert u^\sigma_k \rangle\langle u^\sigma_k \vert u^\sigma_{k+q} \rangle \tag*{(35)}
≈1+qa2(⟨∂aukσ∣ukσ⟩+⟨ukσ∣∂aukσ⟩)+qaqb2(⟨∂a∂bukσ∣ukσ⟩+⟨ukσ∣∂a∂bukσ⟩−2⟨∂aukσ∣∂bukσ⟩)(36)\approx1 + \frac{q_a}{2} \left( \langle\partial_a u^\sigma_k \vert u^\sigma_k \rangle+ \langle u^\sigma_k \vert\partial_a u^\sigma_k \rangle\right) + \frac{q_a q_b}{2} \left( \langle\partial_a \partial_b u^\sigma_k \vert u^\sigma_k \rangle+ \langle u^\sigma_k \vert\partial_a \partial_b u^\sigma_k \rangle- 2 \langle\partial_a u^\sigma_k \vert\partial_b u^\sigma_k \rangle\right) \tag*{(36)}
=1−qaqb[Re⁡⟨∂aukσ∣∂bukσ⟩−Ak,aσAk,bσ]=1−qaqbgk,abσ(37)= 1 - q_a q_b \left[ \operatorname{Re} \langle\partial_a u^\sigma_k \vert\partial_b u^\sigma_k \rangle- A^\sigma_{k,a} A^\sigma_{k,b} \right] = 1 - q_a q_b g^\sigma_{k,ab} \tag*{(37)}
≈e−qagk,abσqb,(38)\approx e^{-q_a g^\sigma_{k,ab}q_b}, \tag*{(38)}

with summation over repeated spatial indices implied, and where the two approximate equal signs indicate the small-qq expansion and the re-exponentiation, respectively. We have introduced the quantum metric gσg^\sigma and Berry connection AσA^\sigma defined as

gk,abσ=Re⁡⟨∂kaukσ∣∂kbukσ⟩−Ak,aσAk,bσ,Ak,aσ=−i⟨ukσ∣∂ka∣ukσ⟩,(39)g^\sigma_{k,ab} = \operatorname{Re} \langle\partial_{k_a} u^\sigma_k \vert\partial_{k_b} u^\sigma_k \rangle- A^\sigma_{k,a} A^\sigma_{k,b}, \qquad A^\sigma_{k,a} = -i \langle u^\sigma_k \vert\partial_{k_a} \vert u^\sigma_k \rangle, \tag*{(39)}

and have used the normalization of Bloch eigenstates to derive the identities

⟨ukσ∣ukσ⟩=1,∂a⟨ukσ∣ukσ⟩=⟨∂aukσ∣ukσ⟩+⟨ukσ∣∂aukσ⟩=0,(40)\langle u^\sigma_k \vert u^\sigma_k \rangle= 1, \qquad\partial_a \langle u^\sigma_k \vert u^\sigma_k \rangle= \langle\partial_a u^\sigma_k \vert u^\sigma_k \rangle+ \langle u^\sigma_k \vert\partial_a u^\sigma_k \rangle= 0, \tag*{(40)}
∂a∂b⟨ukσ∣ukσ⟩=⟨∂a∂bukσ∣ukσ⟩+⟨ukσ∣∂a∂bukσ⟩+2Re⁡⟨∂aukσ∣∂bukσ⟩=0.(41)\partial_a \partial_b \langle u^\sigma_k \vert u^\sigma_k \rangle= \langle\partial_a \partial_b u^\sigma_k \vert u^\sigma_k \rangle+ \langle u^\sigma_k \vert\partial_a \partial_b u^\sigma_k \rangle+ 2 \operatorname{Re} \langle\partial_a u^\sigma_k \vert\partial_b u^\sigma_k \rangle= 0. \tag*{(41)}

To apply these two steps on Δ(ψ)\Delta(\psi), we will separately consider the following two contributions

Δ(ψ)=Δ1(ψ)−Δ2(ψ),Δ1(ψ)=1N∑qv(q)NBZ∑k∣ψkΛk↑(q)∣2,Δ2(ψ)=1N∑qv(q)NBZ∑kψk∗ψk+qΛk↑(q)Λk↓(q)∗.(42)\Delta(\psi) = \Delta_1(\psi) - \Delta_2(\psi), \qquad\Delta_1(\psi) = \frac{1}{N} \sum_q \frac{v(q)}{N_{\mathrm{BZ}}} \sum_k \lvert\psi_k \Lambda_k^\uparrow(q) \rvert^2, \qquad\Delta_2(\psi) = \frac{1}{N} \sum_q \frac{v(q)}{N_{\mathrm{BZ}}} \sum_k \psi_k^* \psi_{k+q} \Lambda_k^\uparrow(q) \Lambda_k^\downarrow(q)^*. \tag*{(42)}

The approximation of Δ1(ψ)\Delta_1(\psi) follows straightforwardly from our calculation on the form factors above:

Δ1(ψ)=1N∑qv(q)NBZ∑k∣ψk∣2∣⟨uk−+q↑∣uk−↑⟩∣2≃1N∑k,qv(q)NBZ∣ψk∣2e−qagk−,ab↑qb.(43)\Delta_1(\psi)=\frac{1}{N}\sum_q\frac{v(q)}{N_{\mathrm{BZ}}}\sum_k|\psi_k|^2|\langle u^\uparrow_{k_-+q}|u^\uparrow_{k_-}\rangle|^2\simeq\frac{1}{N}\sum_{k,q}\frac{v(q)}{N_{\mathrm{BZ}}}|\psi_k|^2e^{-q_a g^\uparrow_{k_-,ab}q_b}. \tag*{(43)}

Finding the approximate behavior of Δ2(ψ)\Delta_2(\psi) requires more caution and can be split into three steps. The first step is to shift (without loss of generality) k→k−q/2k\to k-q/2 in the kk-integration as above, and expand the integrand

ψk−q/2∗ψk+q/2Λk−q/2↑(q)Λk−q/2↓(q)∗=ψk−q/2∗ψk+q/2⟨uk−+q/2↑∣uk−−q/2↑⟩⟨uk−−q/2↓∣uk−+q/2↓⟩(44)\psi^*_{k-q/2}\psi_{k+q/2}\Lambda^\uparrow_{k-q/2}(q)\Lambda^\downarrow_{k-q/2}(q)^*=\psi^*_{k-q/2}\psi_{k+q/2}\langle u^\uparrow_{k_-+q/2}|u^\uparrow_{k_--q/2}\rangle\langle u^\downarrow_{k_--q/2}|u^\downarrow_{k_-+q/2}\rangle \tag*{(44)}
≃∣ψk∣2+∣ψk∣2qa2[∂alog⁡ψk−∂alog⁡ψk∗+2i(Ak+,a↑−Ak−,a↑)]+∣ψk∣2qaqb4[∂a∂bψk∗2ψk∗+∂a∂bψk2ψk−∂aψk∂bψk∗∣ψk∣2]−∣ψk∣2qaqb2[gk−,ab↑+Ak−,a↑Ak−,b↑+gk+,ab↓+Ak+,a↓Ak+,b↓]+i∣ψk∣2qaqb2[(∂alog⁡ψk−∂alog⁡ψk∗)(Ak+,b↓−Ak−,b↑)]+∣ψk∣2qaqbAk+,b↓Ak−,b↑.(45)\begin{aligned} &\simeq|\psi_k|^2+\frac{|\psi_k|^2q_a}{2}\left[\partial_a\log\psi_k-\partial_a\log\psi_k^*+2i\left(A^\uparrow_{k_+,a}-A^\uparrow_{k_-,a}\right)\right]\\ &\quad+\frac{|\psi_k|^2q_aq_b}{4}\left[\frac{\partial_a\partial_b\psi_k^*}{2\psi_k^*}+\frac{\partial_a\partial_b\psi_k}{2\psi_k}-\frac{\partial_a\psi_k\partial_b\psi_k^*}{|\psi_k|^2}\right]\\ &\quad-\frac{|\psi_k|^2q_aq_b}{2}\left[g^\uparrow_{k_-,ab}+A^\uparrow_{k_-,a}A^\uparrow_{k_-,b}+g^\downarrow_{k_+,ab}+A^\downarrow_{k_+,a}A^\downarrow_{k_+,b}\right]\\ &\quad+i\frac{|\psi_k|^2q_aq_b}{2}\left[(\partial_a\log\psi_k-\partial_a\log\psi_k^*)(A^\downarrow_{k_+,b}-A^\uparrow_{k_-,b})\right]+|\psi_k|^2q_aq_bA^\downarrow_{k_+,b}A^\uparrow_{k_-,b}. \tag*{(45)} \end{aligned}

In Eq. (45), the first line gathers the zeroth and first order terms in qq. The second line contains all quadratic terms arising solely from the expansion of ψk−q/2∗ψk+q/2\psi^*_{k-q/2}\psi_{k+q/2}. The third line includes the quadratic terms from the form factor expansion (given in Eq. (37)). The final line gathers all remaining quadratic terms that result from products of first-order terms not accounted for in the previous lines. An elegant way to regroup these terms is to realize that, with

Sktot=Skgeom+Skspat,S_k^{\mathrm{tot}}=S_k^{\mathrm{geom}}+S_k^{\mathrm{spat}},
Sk,atot,∗Sk,btot=(Ak−,a↑−Ak+,a↓−i∂alog⁡ψk∗)(Ak−,b↑−Ak+,b↓+i∂blog⁡ψk)=Ak−,a↑Ak−,b↑+Ak+,a↓Ak+,b↓+∂alog⁡ψk∗∂blog⁡ψk−2Ak+,a↓Ak−,b↑−i(∂alog⁡ψk−∂alog⁡ψk∗)(Ak+,b↓−Ak−,b↑)(46)\begin{aligned} S^{\mathrm{tot},*}_{k,a}S^{\mathrm{tot}}_{k,b} &=(A^\uparrow_{k_-,a}-A^\downarrow_{k_+,a}-i\partial_a\log\psi_k^*)(A^\uparrow_{k_-,b}-A^\downarrow_{k_+,b}+i\partial_b\log\psi_k)\\ &=A^\uparrow_{k_-,a}A^\uparrow_{k_-,b}+A^\downarrow_{k_+,a}A^\downarrow_{k_+,b}+\partial_a\log\psi_k^*\partial_b\log\psi_k-2A^\downarrow_{k_+,a}A^\uparrow_{k_-,b}\\ &\quad-i(\partial_a\log\psi_k-\partial_a\log\psi_k^*)(A^\downarrow_{k_+,b}-A^\uparrow_{k_-,b}) \tag*{(46)} \end{aligned}

contains most of the terms appearing in Eq. (45) such that

ψk−q/2∗ψk+q/2Λk−q/2↑(q)Λk−q/2↓(q)∗≈∣ψk∣2+∣ψk∣2qa2[∂alog⁡ψk−∂alog⁡ψk∗+2i(Ak+,a↑−Ak−,a↑)]+qaqb8[∂a∂b∣ψk∣2]−∣ψk∣2qaqb2[gk−,ab↑+gk+,ab↓]−∣ψk∣2qaqb2Sk,atot,∗Sk,btot.(47)\begin{aligned} \psi^*_{k-q/2}\psi_{k+q/2}\Lambda^\uparrow_{k-q/2}(q)\Lambda^\downarrow_{k-q/2}(q)^* &\approx|\psi_k|^2+\frac{|\psi_k|^2q_a}{2}\left[\partial_a\log\psi_k-\partial_a\log\psi_k^*+2i(A^\uparrow_{k_+,a}-A^\uparrow_{k_-,a})\right]\\ &\quad+\frac{q_aq_b}{8}[\partial_a\partial_b|\psi_k|^2]-\frac{|\psi_k|^2q_aq_b}{2}\left[g^\uparrow_{k_-,ab}+g^\downarrow_{k_+,ab}\right]-\frac{|\psi_k|^2q_aq_b}{2}S^{\mathrm{tot},*}_{k,a}S^{\mathrm{tot}}_{k,b}. \tag*{(47)} \end{aligned}

The second step is to simplify this expansion by factorization of either the qq or kk sum in the expression of Δ2(ψ)\Delta_2(\psi). For instance, we notice that Eq. (47) consists of products of qq monomials and qq-independent terms, such that the sums over qq can be factorized and lead to simplifications when we assume that a rotation-invariant Coulomb potential. In that case, we indeed have

Fa∑qv(q)NBZqa=0,Gab∑qv(q)NBZqaqb=Gabδab2∑qv(q)NBZq2=Tr⁡G2∑qv(q)NBZq2,(48)F_a\sum_q\frac{v(q)}{N_{\mathrm{BZ}}}q_a=0,\qquad G_{ab}\sum_q\frac{v(q)}{N_{\mathrm{BZ}}}q_aq_b=G_{ab}\frac{\delta_{ab}}{2}\sum_q\frac{v(q)}{N_{\mathrm{BZ}}}q^2=\frac{\operatorname{Tr}G}{2}\sum_q\frac{v(q)}{N_{\mathrm{BZ}}}q^2, \tag*{(48)}

for any vector FF and matrix GG, which gives

Δ2(ψ)=∑k,qv(q)NBZ[∣ψk∣2+qaqb8∂a∂b∣ψk∣2−∣ψk∣2qaqbgk−,ab↑+gk+,ab↓2−q24∣ψk∣2∥Sktot∥2].(49)\Delta_2(\psi)=\sum_{k,q}\frac{v(q)}{N_{\mathrm{BZ}}}\left[|\psi_k|^2+\frac{q_aq_b}{8}\partial_a\partial_b|\psi_k|^2-|\psi_k|^2q_aq_b\frac{g^\uparrow_{k_-,ab}+g^\downarrow_{k_+,ab}}{2}-\frac{q^2}{4}|\psi_k|^2\lVert S_k^{\mathrm{tot}}\rVert^2\right]. \tag*{(49)}

Factorizing the sum over kk gives an additional simplification. Indeed, the ψk\psi_k is periodic, such that the average of any of its derivatives vanishes, and in particular

∑k,qv(q)NBZqaqb8∂a∂b∣ψk∣2=∑qqaqbv(q)8NBZ∑k∂a∂b∣ψk∣2=0.(50)\sum_{k,q}\frac{v(q)}{N_{\mathrm{BZ}}}\frac{q_aq_b}{8}\partial_a\partial_b|\psi_k|^2=\sum_q\frac{q_aq_bv(q)}{8N_{\mathrm{BZ}}}\sum_k\partial_a\partial_b|\psi_k|^2=0. \tag*{(50)}

This leaves us with only three terms

Δ2(ψ)=∑k,qv(q)NBZ∣ψk∣2[1−qaqbgk−,ab↑+gk+,ab↓2−q24∥Sktot∥2],(51)\Delta_2(\psi)=\sum_{k,q}\frac{v(q)}{N_{\mathrm{BZ}}}|\psi_k|^2\left[1-q_aq_b\frac{g^\uparrow_{k_-,ab}+g^\downarrow_{k_+,ab}}{2}-\frac{q^2}{4}\lVert S_k^{\mathrm{tot}}\rVert^2\right], \tag*{(51)}

where we have kept the quadratic term involving the quantum metric in a form similar to that of Eq. (43) to obtain a similar exponentiated form. This exponentiation is the third and last step of the calculation, and yields

Δ2(ψ)≈∑k,qv(q)NBZ∣ψk∣2[1−q24∣ψk∣2∥Sktot∥2]e−12qa(gk−,ab↑+gk+,ab↓)qb.(52)\Delta_2(\psi) \approx\sum_{k,q} \frac{v(q)}{N_{\mathrm{BZ}}}|\psi_k|^2\left[1-\frac{q^2}{4}|\psi_k|^2\lVert\mathbf{S}^{\mathrm{tot}}_k\rVert^2\right]e^{-\frac{1}{2}q_a(g^{\uparrow}_{k_-,ab}+g^{\downarrow}_{k_+,ab})q_b}. \tag*{(52)}

The main purpose of the exponential decay obtained in Eqs. (43) and (52) is to provide a characteristic cutoff and avoid divergences from large-qq. These cutoffs are approximate, and, in the limit where the spin-↑\uparrow and spin-↓\downarrow quantum metrics do not fluctuate too much around their mean, they are almost identical and uniform across the BZ. To avoid unnecessary complexity, we assume that this is the case from now on and identify

gk−,ab↑≈gk−,ab↑+gk+,ab↓2≈gk,ab↑+gk,ab↓2=gk,(53)g^{\uparrow}_{k_-,ab}\approx\frac{g^{\uparrow}_{k_-,ab}+g^{\downarrow}_{k_+,ab}}{2}\approx\frac{g^{\uparrow}_{k,ab}+g^{\downarrow}_{k,ab}}{2}=g_{k}, \tag*{(53)}

with the spin-averaged quantum metric introduced in the main text. Combining the different pieces, we finally obtain Eq. (5)

Δ(ψ)=1N∑k∣ψk∣2⋅∥Sktot∥2⋅Uk,Uk=[∑qv(q)4NBZe−qagk,abqb].(54)\Delta(\psi)=\frac{1}{N}\sum_k|\psi_k|^2\cdot\lVert\mathbf{S}^{\mathrm{tot}}_k\rVert^2\cdot U_k,\qquad U_k=\left[\sum_q\frac{v(q)}{4N_{\mathrm{BZ}}}e^{-q_ag_{k,ab}q_b}\right]. \tag*{(54)}

Magnon spin stiffness

In this appendix, we assume that the system has an extended spin-SU(2) symmetry and expand the topological dipole around Q=0Q=0 to obtain the spin stiffness of the gapless magnons. We work in the gauge where sQ=0,k=1s_{Q=0,k}=1 and A↑=A↓≡A\mathcal{A}^{\uparrow}=\mathcal{A}^{\downarrow}\equiv\mathcal{A}, and find that

(SQ,kgeom)b=Ak−Q/2−Ak+Q/2+i∇klog⁡sQ,k≃−Qa∂kaAk;b+i∂kblog⁡[1+12Qa(⟨∂kauk∣uk⟩−⟨uk∣∂kauk⟩)](55)\left(S^{\mathrm{geom}}_{Q,k}\right)_b=\mathcal{A}_{k-Q/2}-\mathcal{A}_{k+Q/2}+i\nabla_k\log s_{Q,k}\simeq-Q_a\partial_{k_a}\mathcal{A}_{k;b}+i\partial_{k_b}\log\left[1+\frac{1}{2}Q_a\left(\langle\partial_{k_a}u_k|u_k\rangle-\langle u_k|\partial_{k_a}u_k\rangle\right)\right] \tag*{(55)}
=−Qa∂kaAk;b+i∂kblog⁡[1−Qa⟨uk∣∂kauk⟩]≃Qa[−∂kaAk;b+∂kbAk;a]=QaϵbaΩk,(56)=-Q_a\partial_{k_a}\mathcal{A}_{k;b}+i\partial_{k_b}\log\left[1-Q_a\langle u_k|\partial_{k_a}u_k\rangle\right]\simeq Q_a\left[-\partial_{k_a}\mathcal{A}_{k;b}+\partial_{k_b}\mathcal{A}_{k;a}\right]=Q_a\epsilon_{ba}\Omega_k, \tag*{(56)}

with Ωk\Omega_k the spin-independent Berry curvature of the model. As a result, we get

∥SQ,kgeom∥2≃∣Q∣2Ωk2,(57)\lVert S^{\mathrm{geom}}_{Q,k}\rVert^2\simeq|Q|^2\Omega_k^2, \tag*{(57)}

providing the spin stiffness quoted in the main text (Eq. (8)). A similar expansion can be used to derive the magnons’ effective mass when their spectrum is gapped.

Microscopic models

Interacting BHZ model

We now describe the lattice Hamiltonian presented in the main text. The noninteracting Hamiltonian follows the 2d BHZ model [119], H0=diag⁡[H↑,H↓]H_0=\operatorname{diag}[H^{\uparrow},H^{\downarrow}], where

Hσ(k)=[M−2B(2−cos⁡kx−cos⁡ky)A(σsin⁡kx−isin⁡ky)A(σsin⁡kx+isin⁡ky)−[M−2B(2−cos⁡kx−cos⁡ky)]](58)H^{\sigma}(\boldsymbol{k})= \begin{bmatrix} M-2B(2-\cos k_x-\cos k_y) & A(\sigma\sin k_x-i\sin k_y) \\ A(\sigma\sin k_x+i\sin k_y) & -[M-2B(2-\cos k_x-\cos k_y)] \end{bmatrix} \tag*{(58)}

The wavefunction for ↑\uparrow and ↓\downarrow spins of the lower band are

∣uk↑⟩=12dk(dk−mk)[mk−dkFk+](59)\lvert u_k^{\uparrow}\rangle= \frac{1}{\sqrt{2d_k(d_k-m_k)}} \begin{bmatrix} m_k-d_k \\ F_k^{+} \end{bmatrix} \tag*{(59)}
∣uk↓⟩=12dk(dk−mk)[mk−dk−Fk−](60)\lvert u_k^{\downarrow}\rangle= \frac{1}{\sqrt{2d_k(d_k-m_k)}} \begin{bmatrix} m_k-d_k \\ -F_k^{-} \end{bmatrix} \tag*{(60)}

where mk=M−2B(2−cos⁡kx−cos⁡ky)m_k=M-2B(2-\cos k_x-\cos k_y), Fk±=A(sin⁡kx±isin⁡ky)F_k^{\pm}=A(\sin k_x\pm i\sin k_y), dk=Fk+Fk−+mk2d_k=\sqrt{F_k^{+}F_k^{-}+m_k^2}. Setting A=1A=1 and B=1B=1, the Chern number as a function of M/BM/B is shown in Figure S1. The system exhibits two types of topological phase transitions: at M=0M=0 and M=8M=8, the Chern number changes from C=0C=0 to C=±1C=\pm1, while at M=4M=4, it switches from C=1C=1 to C=−1C=-1.

Chern number of wavefunction $|u_k\rangle$ as function of $M/B$

Figure S1. Chern number of wavefunction ∣uk⟩|u_k\rangle as function of M/BM/B

tMoTe2 and ferromagnetism

An extended fully polarized phase has been observed in twisted MoTe2_2 near a twist angle of 3.7∘3.7^\circ [44]. In this appendix, we present the continuum model used to describe the tMoTe2_2 system. In tMoTe2_2, the low-energy valence bands predominantly originate from the ±K\pm K valleys of the monolayer. Due to strong spin–orbit coupling, the spin and valley degrees of freedom are locked to each other. In addition to the translational symmetry of the moiré lattice, the AA-stacked tMoTe2_2 also respects C3C_3, C2yC_{2y}, and time-reversal (TT) symmetries. By assuming that the top and bottom layers are rotated by θ/2\theta/2 and −θ/2-\theta/2, respectively, the moiré lattice constant is

aM=a02sin⁡(θ2)(61)a_M=\frac{a_0}{2\sin\left(\frac{\theta}{2}\right)} \tag*{(61)}

where a0=3.52a_0=3.52 Å is the monolayer lattice constant. The continuous model Hamiltonian for spin-↑\uparrow (equivalently, at KK valley) takes the form as [122, 123, 124, 125]

H↑=[Ht(r)ΔT(r)ΔT∗(r)Hb(r)](62)\mathcal{H}_{\uparrow}= \begin{bmatrix} H_t(\boldsymbol{r}) & \Delta_T(\boldsymbol{r}) \\ \Delta_T^{*}(\boldsymbol{r}) & H_b(\boldsymbol{r}) \end{bmatrix} \tag*{(62)}

where the layer Hamiltonians are given by

Ht/b(r)=−ℏ22m∗(−i∇−κ±)2+Δ±(r)±Δε2.(63)H_{t/b}(\boldsymbol{r})=-\frac{\hbar^2}{2m^*}(-i\nabla-\boldsymbol{\kappa}_{\pm})^2+\Delta_{\pm}(\boldsymbol{r})\pm\frac{\Delta\varepsilon}{2}. \tag*{(63)}

Here, m∗=0.6mem^*=0.6m_e denotes the effective electron mass, and Δε=Dϵ0ξ0\Delta\varepsilon=\frac{D}{\epsilon_0}\xi_0 represents the potential difference induced by an applied vertical displacement field, with ξ0\xi_0 the interlayer distance of the MoTe2_2 bilayer. Due to the interlayer twist, the KK points of the two layers are displaced and folded into the corners of the moiré Brillouin zone, labeled as κ±\boldsymbol{\kappa}_{\pm}. We define the moiré reciprocal lattice vectors to be Gj=4π3aM(cos⁡π(j−1)3,sin⁡π(j−1)3)\boldsymbol{G}_j=\frac{4\pi}{\sqrt{3}a_M}\left(\cos\frac{\pi(j-1)}{3},\sin\frac{\pi(j-1)}{3}\right), and choose κ+=G1+G23\boldsymbol{\kappa}_{+}=\frac{\boldsymbol{G}_1+\boldsymbol{G}_2}{3}, κ−=G1+G63\boldsymbol{\kappa}_{-}=\frac{\boldsymbol{G}_1+\boldsymbol{G}_6}{3}, respectively. The intralayer moiré potential and the interlayer tunneling terms are given by

Δ±(r)=2w1∑j=1,3,5cos⁡(Gj⋅r±ϕ),ΔT=w2(1+e−iG2⋅r+e−iG3⋅r).(64)\Delta_{\pm}(\boldsymbol{r})=2w_1\sum_{j=1,3,5}\cos(\boldsymbol{G}_j\cdot\boldsymbol{r}\pm\phi),\qquad \Delta_T=w_2(1+e^{-i\boldsymbol{G}_2\cdot\boldsymbol{r}}+e^{-i\boldsymbol{G}_3\cdot\boldsymbol{r}}). \tag*{(64)}

Because the ±K\pm K valleys are related by time-reversal symmetry, the Hamiltonian for spin-↓\downarrow electrons (H↓\mathcal{H}_{\downarrow}) can be obtained as the time-reversal conjugate of H↑\mathcal{H}_{\uparrow}. Under time reversal symmetry, the wavefunction of valley-η\eta and layer-ll transforms as Tcη,l,rT=c−η,l,r†\mathcal{T}c_{\eta,l,\boldsymbol{r}}\mathcal{T}=c^{\dagger}_{-\eta,l,\boldsymbol{r}}. As a result, the spin-↓\downarrow Hamiltonian takes the form

H↓=[Ht∗(r)ΔT∗(r)ΔT(r)Hb∗(r)].(65)\mathcal{H}_{\downarrow}= \begin{bmatrix} H_t^{*}(\boldsymbol{r}) & \Delta_T^{*}(\boldsymbol{r}) \\ \Delta_T(\boldsymbol{r}) & H_b^{*}(\boldsymbol{r}) \end{bmatrix}. \tag*{(65)}

For the interaction term, we consider a dual-gate screened Coulomb interaction v(q)=e22ϵϵ0tanh⁡(ξ∣q∣)∣q∣v(q) = \frac{e^2}{2\epsilon\epsilon_0}\frac{\tanh(\xi|q|)}{|q|}, where ξ\xi is the distance from the middle of the bilayer to the gates. The parameters of the Hamiltonian we used are listed in Table S1.

m∗(me)m^*(m_e)w1w_1 (meV)ϕ\phi (deg)w2w_2 (meV)ξ0\xi_0 (Å)ξ\xi (Å)ϵ\epsilon
0.6016.5-105.9-18.873007

Table S1. Values of the parameters for the continuum model of AA-stacking tMoTe2_2.

Figure S2 (a) and (b) display the Berry curvature Ω(k)\Omega(\boldsymbol{k}) and the trace of the quantum metric Tr⁡g(k)\operatorname{Tr}g(\boldsymbol{k}), respectively, for the case of zero displacement field. In Figure S2 (c), we track how the system evolves with increasing displacement potential: the maximum energy difference between spin bands, Kkin=max⁡(∣εk,↓−εk,↑∣)K_{\mathrm{kin}} = \max(|\varepsilon_{\boldsymbol{k},\downarrow} - \varepsilon_{\boldsymbol{k},\uparrow}|), the bandwidth of the top band, and the energy gap separating the first and second highest bands are shown as a function of Δε\Delta\varepsilon. The topological phase transition is marked by vertical gray dashed line with Δεc=24.5\Delta\varepsilon_c = 24.5 meV.

Berry curvature $\Omega(\boldsymbol{k})$ for spin-$\uparrow$ with zero displacement field

(a)

Trace of quantum metric $\operatorname{Tr}g(\boldsymbol{k})$ for spin-$\uparrow$ with zero displacement field

(b)

Maximum energy difference, bandwidth, and band gap versus displacement potential

(c)

Figure S2. (a) Berry curvature Ω(k)\Omega(\boldsymbol{k}) and (b) trace of quantum metric Tr⁡g(k)\operatorname{Tr}g(\boldsymbol{k}) for spin-↑\uparrow with zero displacement field (Δε=0\Delta\varepsilon= 0). (c) Maximum energy difference Kkin=max⁡(∣εk,↓−εk,↑∣)K_{\mathrm{kin}} = \max(|\varepsilon_{\boldsymbol{k},\downarrow} - \varepsilon_{\boldsymbol{k},\uparrow}|), bandwidth of the top band, and band gap between the first and second highest bands versus the displacement potential Δε\Delta\varepsilon. The gray dashed line marks the topological phase transition point.

Bethe–Salpeter equation in the contact-interaction limit

In this appendix we analyze the BS equation in the contact-interaction limit, where the interaction is momentum independent, v(q)=vv(q) = v. In this case the interaction kernel in Eq. (27) becomes separable in the Bloch wavefunctions. Factoring out the constant interaction kernel vv, the diagonal interaction term in the Bethe-Salpeter Hamiltonian take the form

Σk(Q)≡∑pWpp↑↓(k−−p)=vρˉk−,ρˉk−=⟨uk−↑∣ρ^↑∣uk−↑⟩,ρ^↑=1NBZ∑p∣up↑⟩⟨up↑∣(66)\Sigma_{\boldsymbol{k}}(\boldsymbol{Q}) \equiv\sum_{\boldsymbol{p}} W^{\uparrow\downarrow}_{\boldsymbol{p}\boldsymbol{p}}(\boldsymbol{k}_{-} - \boldsymbol{p}) = v\bar{\rho}_{\boldsymbol{k}_{-}}, \qquad\bar{\rho}_{\boldsymbol{k}_{-}} = \langle u_{\boldsymbol{k}_{-}}^{\uparrow}|\hat{\rho}^{\uparrow}|u_{\boldsymbol{k}_{-}}^{\uparrow}\rangle, \qquad\hat{\rho}^{\uparrow} = \frac{1}{N_{\mathrm{BZ}}}\sum_{\boldsymbol{p}} |u_{\boldsymbol{p}}^{\uparrow}\rangle\langle u_{\boldsymbol{p}}^{\uparrow}| \tag*{(66)}

which acts as a momentum-dependent self-energy and scales with the local density of charge removed from the ground state (as measured by the band-projected ρ^\hat{\rho}). Introducing an orbital basis ∣a=1,⋯ ,Norb⟩|a = 1,\cdots,N_{\mathrm{orb}}\rangle for the original model from which the topological band derives and the spin-mixing projector Pk(Q)=∣uk−↑⟩⟨uk+↑∣P_{\boldsymbol{k}}(\boldsymbol{Q}) = |u_{\boldsymbol{k}_{-}}^{\uparrow}\rangle\langle u_{\boldsymbol{k}_{+}}^{\uparrow}|, we can recast the off-diagonal coefficients as

Wk−,k+↑↓(k′−k)=vNBZ∑a,bPk,ab(Q)Pk′,ab∗(Q),Pk,ab(Q)=⟨a∣Pk(Q)∣b⟩.(67)W^{\uparrow\downarrow}_{\boldsymbol{k}_{-},\boldsymbol{k}_{+}}(\boldsymbol{k}'-\boldsymbol{k}) = \frac{v}{N_{\mathrm{BZ}}}\sum_{a,b} P_{\boldsymbol{k},ab}(\boldsymbol{Q})P_{\boldsymbol{k}',ab}^{*}(\boldsymbol{Q}), \qquad P_{\boldsymbol{k},ab}(\boldsymbol{Q}) = \langle a|P_{\boldsymbol{k}}(\boldsymbol{Q})|b\rangle. \tag*{(67)}

Crucially, this expression factorizes the k\boldsymbol{k} and k′\boldsymbol{k}' contributions, allowing for an efficient solution of the Bethe-Salpeter equation in orbital space after summation over momenta.

Substituting into Eq. (26), the Bethe-Salpeter equation becomes

[εk+,↓−εk−,↑+vρˉk↑−E]ψQ,k−v∑abPk,ab(Q)Ψab(Q)=0,Ψab(Q)=1NBZ∑k′Pk′,ab∗(Q)ψQ,k′.(68)\left[\varepsilon_{k+,\downarrow}-\varepsilon_{k-,\uparrow}+v\bar{\rho}^{\uparrow}_{k}-\mathcal{E}\right]\psi_{Q,k}-v\sum_{ab}P_{k,ab}(Q)\Psi_{ab}(Q)=0,\qquad\Psi_{ab}(Q)=\frac{1}{N_{\mathrm{BZ}}}\sum_{k'}P^{*}_{k',ab}(Q)\psi_{Q,k'}. \tag*{(68)}

Multiplying the equation by Pk,m=cd∗(Q)P^{*}_{k,m=cd}(Q) and summing over kk gives the following equation

[δm,n+Πm,n(E,Q)]Ψn(Q)=0,Πm,n(E,Q)=vNBZ∑kPk,m∗(Q)Pk,n(Q)E−[εk+,↓−εk−,↑+vρˉk↑].(69)\left[\delta_{m,n}+\Pi_{m,n}(\mathcal{E},Q)\right]\Psi_{n}(Q)=0,\qquad\Pi_{m,n}(\mathcal{E},Q)=\frac{v}{N_{\mathrm{BZ}}}\sum_{k}\frac{P^{*}_{k,m}(Q)P_{k,n}(Q)}{\mathcal{E}-\left[\varepsilon_{k+,\downarrow}-\varepsilon_{k-,\uparrow}+v\bar{\rho}^{\uparrow}_{k}\right]}. \tag*{(69)}

over the variable Ψn=ab(Q)\Psi_{n=ab}(Q), which at fixed QQ is a bilinear of the orbital indices and corresponds to a vector with Norb2N_{\mathrm{orb}}^{2} independent entries. The Bethe-Salpeter equation therefore reduces to a finite-dimensional matrix problem of size Norb2N_{\mathrm{orb}}^{2}, and its eigenenergies are determined by the condition

Det⁡[IdNorb2+Π(E,Q)]=0,(70)\operatorname{Det}\left[I_{d_{N_{\mathrm{orb}}^{2}}}+\Pi(\mathcal{E},Q)\right]=0, \tag*{(70)}

with IdpI_{d_p} denoting the identity matrix of size pp. Thus, the infinite momentum-space integral eigenvalue equation collapses into a finite-rank algebraic eigenvalue problem, i.e. the Sherman–Morrison reduction. For the BHZ model, the projector Pk(Q)P_k(Q) has only 4 independent coefficients, speeding up the obtention of the spectrum (although we now need to solve a non-linear equation in the energy E\mathcal{E}). At Q=0Q=0, the projector takes the explicit form

Pk(0)=∣uk−↓⟩⊗∣uk+↑⟩∗=12dk(dk−mk)[(mk−dk)(mk−dk)−Fk+(mk−dk)Fk−(mk−dk)−Fk+2],(71)\begin{aligned} P_k(0)=\lvert u_{k-}^{\downarrow}\rangle\otimes\lvert u_{k+}^{\uparrow}\rangle^{*} \\ &=\frac{1}{2d_k(d_k-m_k)} \begin{bmatrix} (m_k-d_k)(m_k-d_k)\\ -F_k^{+}(m_k-d_k)\\ F_k^{-}(m_k-d_k)\\ -F_k^{+2} \end{bmatrix}, \tag*{(71)} \end{aligned}

where dkd_k, mkm_k and Fk±F_k^{\pm} are defined below Eq. (60). For the twisted MoTe2_2 model, in contrast, the dimension depends on the number of reciprocal lattice vectors included in the basis. In our simulations, this is typically in the few hundreds, similar to the number of discretized momenta along each direction of our lattice. In that case, the numerical efficiency should be similar for both the Bethe-Salpeter equation with discretized momenta (main text) and the Sherman-Morrison method (described in this appendix).

References

  1. [1]D. Xiao, M.-C. Chang, and Q. Niu, Berry phase effects on electronic properties, Reviews of modern physics 82, 1959 (2010).
  2. [2]P. Törmä, Essay: Where can quantum geometry lead us?, Physical Review Letters 131, 240001 (2023).DOI
  3. [3]J. Yu, B. A. Bernevig, R. Queiroz, E. Rossi, P. Törmä, and B.-J. Yang, Quantum geometry in quantum materials, arXiv preprint arXiv:2501.00098 (2024).arxiv.org/abs/2501.00098
  4. [4]T. Liu, X.-B. Qian, H.-Z. Lu, and X. Xie, Quantum geometry in condensed matter, National Science Review 12, nwae334 (2025).
  5. [5]N. Marzari, A. A. Mostofi, J. R. Yates, I. Souza, and D. Vanderbilt, Maximally localized wannier functions: Theory and applications, Reviews of Modern Physics 84, 1419 (2012).
  6. [6]P. Törmä, S. Peotta, and B. A. Bernevig, Superconductivity, superfluidity, and quantum geometry in twisted multilayer systems, Nature Reviews Physics 4, 528 (2022).
  7. [7]J. Herzog-Arbeitman, A. Chew, K.-E. Huhtinen, P. Törmä, and B. A. Bernevig, Many-body superconductivity in topological flat bands, arXiv preprint arXiv:2209.00007 (2022).
  8. [8]S. Peotta and P. Törmä, Superfluidity in topologically nontrivial flat bands, Nature communications 6, 8944 (2015).
  9. [9]F. Xie, Z. Song, B. Lian, and B. A. Bernevig, Topology-induced superfluid weight in twisted bilayer graphene, Physical review letters 124, 167002 (2020).
  10. [10]J. Herzog-Arbeitman, V. Peri, F. Schindler, S. D. Hub­er, and B. A. Bernevig, Superfluid weight bounds from symmetry and quantum geometry in flat bands, Physical review letters 128, 087002 (2022).
  11. [11]S. Kwon and B.-J. Yang, Quantum geometric bound and ideal condition for euler band topology, Physical Review B 109, L161111 (2024).
  12. [12]Y. Onishi and L. Fu, Fundamental bound on topological gap, Physical Review X 14, 011052 (2024).
  13. [13]I. Komissarov, T. Holder, and R. Queiroz, The quantum geometric origin of capacitance in insulators, Nature communications 15, 4621 (2024).
  14. [14]N. Verma and R. Queiroz, Instantaneous response and quantum geometry of insulators, arXiv preprint arXiv:2403.07052 (2024).DOI
  15. [15]Y. Onishi and L. Fu, Topological bound on the structure factor, Physical Review Letters 133, 206602 (2024).
  16. [16]I. Esin, É. Lantagne-Hurtubise, F. Nathan, and G. Refael, Quantum geometry and bounds on dissipation in slowly driven quantum systems, Physical Review Letters 134, 146603 (2025).
  17. [17]P. Froese, M. R. Hirshbrunner, and Y. B. Kim, Probing quantum geometry with two-dimensional nonlinear optical spectroscopy, arXiv preprint arXiv:2506.05462 (2025).
  18. [18]J. G. Kattan, A. H. Duff, and J. E. Sipe, Linear response of a chern insulator to finite-frequency electric fields, Phys. Rev. B 111, 075202 (2025).
  19. [19]Y. Zeng and A. J. Millis, Superfluid stiffness bounds in time-reversal symmetric superconductors, arXiv preprint arXiv:2506.18081 (2025).
  20. [20]H.-Y. Xie, P. Ghaemi, M. Mitrano, and B. Uchoa, Theory of topological exciton insulators and condensates in flat chern bands, Proceedings of the National Academy of Sciences 121, e2401644121 (2024).
  21. [21]R. von Baltz and W. Kraut, Theory of the bulk photovoltaic effect in pure crystals, Physical Review B 23, 5590 (1981).
  22. [22]B. I. Sturman and V. M. Fridkin, Photovoltaic and photo-refractive effects in noncentrosymmetric materials, Routledge 10.1201/9780203743416 (1992).
  23. [23]J. L. Hughes and J. Sipe, Calculation of second-order optical response in semiconductors, Physical Review B 53, 10751 (1996).
  24. [24]J. Sipe and A. H. Shkrebtii, Second-order optical response in semiconductors, Physical Review B 61, 5337 (2000).
  25. [25]D. Pesin, Two-particle collisional coordinate shifts and hydrodynamic anomalous hall effect in systems without lorentz invariance, Physical Review Letters 121, 226601 (2018).
  26. [26]J. Cao, H. A. Fertig, and L. Brey, Quantum geometric exciton drift velocity, Physical Review B 103, 10.1103/physrevb.103.115422 (2021).
  27. [27]C. Paiva, T. Holder, and R. Ilan, Shift and polarization of excitons from quantum geometry, arXiv preprint arXiv:2408.10300 (2024).
  28. [28]H. Fertig and L. Brey, Many-body quantum geometric dipole, Physical Review B 111, 035158 (2025).
  29. [29]B. Estienne, N. Regnault, and V. Crépel, Ideal chern bands as landau levels in curved space, Physical Review Research 5, L032048 (2023).
  30. [30]R. Roy, Band geometry of fractional topological insulators, Physical Review B 90, 165139 (2014).
  31. [31]S. A. Parameswaran, R. Roy, and S. L. Sondhi, Fractional quantum hall physics in topological flat bands, Comptes Rendus Physique 14, 816 (2013).
  32. [32]V. Crépel, A. Dunbrack, D. Guerci, J. Bonini, and J. Cano, Chiral model of twisted bilayer graphene realized in a monolayer, Physical Review B 108, 075126 (2023).
  33. [33]J. Wang, J. Cano, A. J. Millis, Z. Liu, and B. Yang, Exact landau level description of geometry and interaction in a flatband, Physical review letters 127, 246403 (2021).
  34. [34]V. Crépel, N. Regnault, and R. Queiroz, Chiral limit and origin of topological flat bands in twisted transition metal dichalcogenide homobilayers, Communications Physics 7, 146 (2024).
  35. [35]M. Claassen, C. H. Lee, R. Thomale, X.-L. Qi, and T. P. Devereaux, Position-momentum duality and fractional quantum hall effect in chern insulators, Physical review letters 114, 236802 (2015).
  36. [36]V. Crépel, P. Ding, N. Verma, N. Regnault, and R. Queiroz, Topologically protected flatness in chiral moiré heterostructures, Physical Review X 15, 021056 (2025).DOI
  37. [37]J. Shi, J. Cano, and N. Morales-Durán, Effects of berry curvature on ideal band magnetorotons, arXiv preprint arXiv:2503.15900 (2025).
  38. [38]N. Marzari and D. Vanderbilt, Maximally localized generalized wannier functions for composite energy bands, Physical review B 56, 12847 (1997).
  39. [39]Q. Li, J. Dong, P. J. Ledwith, and E. Khalaf, Constraints on real space representations of chern bands, arXiv preprint arXiv:2407.02561 (2024).arxiv.org/abs/2407.02561
  40. [40]H. Zhou, T. Xie, A. Ghazaryan, T. Holder, J. R. Ehrets, E. M. Spanton, T. Taniguchi, K. Watanabe, E. Berg, M. Serbyn, et al., Half-and quarter-metals in rhombohedral trilayer graphene, Nature 598, 429 (2021).
  41. [41]H. Zhou, L. Holleis, Y. Saito, L. Cohen, W. Huynh, C. L. Patterson, F. Yang, T. Taniguchi, K. Watanabe, and A. F. Young, Isospin magnetism and spin-polarized superconductivity in bernal bilayer graphene, Science 375, 774 (2022).DOI
  42. [42]S. C. de la Barrera, S. Aronson, Z. Zheng, K. Watanabe, T. Taniguchi, Q. Ma, P. Jarillo-Herrero, and R. Ashoori, Cascade of isospin phase transitions in bernal-stacked bilayer graphene at zero magnetic field, Nature Physics 18, 771 (2022).
  43. [43]E. Anderson, F.-R. Fan, J. Cai, W. Holtzmann, T. Taniguchi, K. Watanabe, D. Xiao, W. Yao, and X. Xu, Programming correlated magnetic states with gate-controlled moiré geometry, Science 381, 325 (2023).DOI
  44. [44]J. Cai, E. Anderson, C. Wang, X. Zhang, X. Liu, W. Holtzmann, Y. Zhang, F. Fan, T. Taniguchi, K. Watanabe, et al., Signatures of fractional quantum anomalous hall states in twisted mote2, Nature 622, 63 (2023).DOI
  45. [45]H. Park, J. Cai, E. Anderson, Y. Zhang, J. Zhu, X. Liu, C. Wang, W. Holtzmann, C. Hu, Z. Liu, T. Taniguchi, K. Watanabe, J.-H. Chu, T. Cao, L. Fu, W. Yao, C.-Z. Chang, D. Cobden, D. Xiao, and X. Xu, Observation of fractionally quantized anomalous Hall effect, Nature 622, 74 (2023).DOI
  46. [46]F. Xu, Z. Sun, T. Jia, C. Liu, C. Xu, C. Li, Y. Gu, K. Watanabe, T. Taniguchi, B. Tong, J. Jia, Z. Shi, S. Jiang, Y. Zhang, X. Liu, and T. Li, Observation of integer and fractional quantum anomalous hall effects in twisted bilayer mote2, Phys. Rev. X 13, 031037 (2023).
  47. [47]Y. Zeng, Z. Xia, K. Kang, J. Zhu, K. Nüppel, C. Vaswani, K. Watanabe, T. Taniguchi, K. F. Mak, and J. Shan, Thermodynamic evidence of fractional chern insulator in moiré mote2, Nature 622, 69 (2023).DOI
  48. [48]T. Han, Z. Lu, G. Scuri, J. Wang, T. Wang, J. Han, K. Watanabe, T. Taniguchi, L. Fu, H. Park, et al., Orbital multiferroicity in pentalayer rhombohedral graphene, Nature 623, 41 (2023).DOI
  49. [49]Z. Lu, T. Han, Y. Yao, A. P. Reddy, J. Yang, J. Seo, K. Watanabe, T. Taniguchi, L. Fu, and L. Ju, Fractional quantum anomalous hall effect in multilayer graphene, Nature 626, 759 (2024).
  50. [50]T. Han, Z. Lu, G. Scuri, J. Sung, J. Wang, T. Han, K. Watanabe, T. Taniguchi, H. Park, and L. Ju, Correlated insulator and chern insulators in pentalayer rhombohedral-stacked graphene, Nature Nanotechnol-DOI
  51. [51]B. A. Foutty, C. R. Kometter, T. Devakul, A. P. Reddy, K. Watanabe, T. Taniguchi, L. Fu, and B. E. Feldman, Mapping twist-tuned multiband topology in bilayer wse2, Science 384, 343 (2024).DOI
  52. [52]B. Gao, M. Ghafariasl, M. J. Mehrabad, T.-S. Huang, L. Zhang, D. Session, P. Upadhyay, R. Ma, G. Alshalan, D. G. S. Forero, et al., Probing quantum anomalous hall states in twisted bilayer wse2 via attractive polaron spectroscopy, arXiv preprint arXiv:2504.11530 (2025).DOI
  53. [53]J. Xie, Z. Huo, X. Lu, Z. Feng, Z. Zhang, W. Wang, Q. Yang, K. Watanabe, T. Taniguchi, K. Liu, Z. Song, X. C. Xie, J. Liu, and X. Lu, Tunable fractional chern insulators in rhombohedral graphene superlattices, Nature Materials 10.1038/s41563-025-02225-7 (2025).
  54. [54]D. M. Kennes, M. Claassen, L. Xian, A. Georges, A. J. Millis, J. Hone, C. R. Dean, D. Basov, A. N. Pasupathy, and A. Rubio, Moiré heterostructures as a condensed-matter quantum simulator, Nature Physics 17, 155 (2021).
  55. [55]K. P. Nuckolls and A. Yazdani, A microscopic perspective on moiré materials, Nature Reviews Materials 9, 460 (2024).
  56. [56]V. Crépel and A. Millis, Bridging the small and large in twisted transition metal dichalcogenide homobilayers: A tight binding model capturing orbital interference and topology across a wide range of twist angles, Physical Review Research 6, 033127 (2024).DOI
  57. [57]Y. Zeng, V. Crépel, and A. J. Millis, Keldysh field theory of dynamical exciton condensation transitions in nonequilibrium electron-hole bilayers, Physical Review Letters 132, 266001 (2024).DOI
  58. [58]D. Guercio, K. P. Lucht, V. Crépel, J. Cano, J. Pixley, and A. Millis, Topological kondo semimetal and insulator in ab-stacked heterobilayer transition metal dichalcogenides, Physical Review B 110, 165128 (2024).DOI
  59. [59]L.-k. Shi, J. Ma, and J. C. Song, Gate-tunable flat bands in van der waals patterned dielectric superlattices, 2D Materials 7, 015028 (2019).DOI
  60. [60]S. A. Ghorashi, A. Dunbrack, A. Abouelkomsan, J. Sun, X. Du, and J. Cano, Topological and stacked flat bands in bilayer graphene with a superlattice potential, Physical Review Letters 130, 196201 (2023).DOI
  61. [61]S. A. A. Ghorashi and J. Cano, Multilayer graphene with a superlattice potential, Physical Review B 107, 195423 (2023).
  62. [62]Z. Krix and O. P. Sushkov, Patterned bilayer graphene as a tunable strongly correlated system, Physical Review B 107, 165158 (2023).DOI
  63. [63]Q. Gao, J. Dong, P. Ledwith, D. Parker, and E. Khalaf, Untwisting moiré physics: Almost ideal bands and fractional chern insulators in periodically strained monolayer graphene, Phys. Rev. Lett. 131, 096401 (2023).DOI
  64. [64]X. Wan, S. Sarkar, S.-Z. Lin, and K. Sun, Topological exact flat bands in two-dimensional materials under periodic strain, Physical Review Letters 130, 216401 (2023).DOI
  65. [65]Y. Zeng, T. M. Wolf, C. Huang, N. Wei, S. A. A. Ghorashi, A. H. MacDonald, and J. Cano, Gate-tunable topological phases in superlattice modulated bilayer graphene, Physical Review B 109, 195406 (2024).DOI
  66. [66]D. Seleznev, J. Cano, and D. Vanderbilt, Inducing topological flat bands in bilayer graphene with electric and magnetic superlattices, Physical Review B 110, 205115 (2024).DOI
  67. [67]J. Sun, S. A. Akbar Ghorashi, K. Watanabe, T. Taniguchi, F. Camino, J. Cano, and X. Du, Signature of correlated insulator in electric field controlled superlattice, Nano Letters 24, 13600 (2024).DOI
  68. [68]D. Ault-McCoy, M. N. Y. Lhachemi, A. Dunbrack, S. A. A. Ghorashi, and J. Cano, Optimizing superlattice bilayer graphene for a fractional chern insulator, arXiv preprint arXiv:2505.05551 (2025).
  69. [69]N. Regnault and B. A. Bernevig, Fractional chern insulator, Physical Review X 1, 021014 (2011).DOI
  70. [70]T. Neupert, L. Santos, C. Chamon, and C. Mudry, Fractional quantum hall states at zero magnetic field, Physical Review Letters 106, 10.1103/physrevlett.106.236804 (2011).
  71. [71]D. Sheng, Z.-C. Gu, K. Sun, and L. Sheng, Fractional quantum hall effect in the absence of landau levels, Nature Communications 2, 10.1038/ncomms3180 (2011).
  72. [72]V. Crépel and A. Millis, Spinon pairing induced by chiral in-plane exchange and the stabilization of odd-spin chern number spin liquid in twisted mote 2, Physical Review Letters 133, 146503 (2024).DOI
  73. [73]M. Levin and A. Stern, Fractional topological insulators, Physical Review Letters 103, 10.1103/physrevlett.103.196803 (2009).
  74. [74]V. Crépel and N. Regnault, Attractive haldane bilayers for trapping non-abelian anyons, Physical Review B 110, 115109 (2024).DOI
  75. [75]F. Xie, L. Chen, S. Sur, Y. Fang, J. Cano, and Q. Si, Superconductivity in twisted vse 2 from topology-induced quantum fluctuations, Physical Review Letters 134, 136503 (2025).DOI
  76. [76]A. Stern, Fractional topological insulators: A pedagogical review, Annual Review of Condensed Matter Physics 7, 349–368 (2016).DOI
  77. [77]V. Crépel and L. Fu, Anomalous hall metal and fractional chern insulator in twisted transition metal dichalcogenides, Physical Review B 107, L201109 (2023).
  78. [78]N. Morales-Durán, J. Wang, G. R. Schleder, M. Angeli, Z. Zhu, E. Kaxiras, C. Repellin, and J. Cano, Pressure-enhanced fractional chern insulators along a magic line in moiré transition metal dichalcogenides, Phys. Rev. Res. 5, L032022 (2023).
  79. [79]C. Repellin, Z. Dong, Y.-H. Zhang, and T. Senthil, Ferromagnetism in narrow bands of moiré superlattices, Physical Review Letters 124, 187601 (2020).DOI
  80. [80]D. Muñoz-Segovia, V. Crépel, R. Queiroz, and A. J. Millis, Twist-angle evolution of the intervalley-coherent antiferromagnet in twisted wse _2, arXiv preprint arXiv:2503.11763 (2025).
  81. [81]N. Nakatsuji, J. Cano, and V. Crépel, High-throughput discovery of moir’e homobilayers guided by topology and energetics, arXiv preprint arXiv:2512.15851 (2025).arxiv.org/abs/2512.15851
  82. [82]N. Verma, D. Guerci, and R. Queiroz, Geometric stiffness in interlayer exciton condensates, Physical Review Letters 132, 236001 (2024).DOI
  83. [83]J. Cao, H. Fertig, and L. Brey, Quantum internal structure of plasmons, Physical review letters 127, 196403 (2021).
  84. [84]J. Cao, H. Fertig, and L. Brey, Plasmonic transverse dipole moment in chiral fermion nanowires, Physical Review B 106, 165125 (2022).DOI
  85. [85]J. Kang, T. Oh, J. Lee, and B.-J. Yang, Quantum geometric bound for saturated ferromagnetism, arXiv preprint arXiv:2402.07171 (2024).arxiv.org/abs/2402.07171
  86. [86]S. M. Young and A. M. Rappe, First principles calculation of the shift current photovoltaic effect in ferroelectrics, Physical review letters 109, 116601 (2012).
  87. [87]T. Morimoto and N. Nagaosa, Topological aspects of nonlinear excitonic processes in noncentrosymmetric crystals, Physical Review B 94, 035117 (2016).DOI
  88. [88]R. Fei, L. Z. Tan, and A. M. Rappe, Shift-current bulk photovoltaic effect influenced by quasiparticle and exciton, Physical Review B 101, 045104 (2020).DOI
  89. [89]D. Kaplan, T. Holder, and B. Yan, Twisted photovoltaics at terahertz frequencies from momentum shift current, Physical Review Research 4, 013209 (2022).DOI
  90. [90]S. Chaudhary, C. Lewandowski, and G. Refael, Shift-current response as a probe of quantum geometry and electron-electron interactions in twisted bilayer graphene, Physical Review Research 4, 013164 (2022).
  91. [91]H. Lee, Y. B. Kim, J. W. Ryu, S. Kim, J. Bae, Y. Koo, D. Jang, and K.-D. Park, Recent progress of exciton transport in two-dimensional semiconductors, Nano Convergence 10, 57 (2023).
  92. [92]M. Nakamura, Y.-H. Chan, T. Yasunami, Y.-S. Huang, G.-Y. Guo, Y. Hu, N. Ogawa, Y. Chiew, X. Yu, T. Morimoto, et al., Strongly enhanced shift current at exciton resonances in a noncentrosymmetric wide-gap semiconductor, Nature Communications 15, 9672 (2024).DOI
  93. [93]J. B. Haber, D. Y. Qiu, F. H. da Jornada, and J. B. Neaton, Maximally localized exciton wannier functions for solids, Phys. Rev. B 108, 125118 (2023).arxiv.org/abs/2308.03012
  94. [94]T. Fukui, Y. Hatsugai, and H. Suzuki, Chern numbers in discretized brillouin zone: efficient method of computing (spin) hall conductances, Journal of the Physical Society of Japan 74, 1674 (2005).DOI
  95. [95]T. Fukui and Y. Hatsugai, Topological aspects of the quantum spin-hall effect in graphene: Z 2 topological order and spin chern number, Physical Review B—Condensed Matter and Materials Physics 75, 121403 (2007).DOI
  96. [96]Gauge invariance of the quantum-geometric dipole is straightforwardly checked by substituting |u_{k,σ}⟩ → e^{iθ_{k,σ}}|u_{k,σ}⟩ in Eq. 2.
  97. [97]M. V. Berry, Quantal phase factors accompanying adiabatic changes, Proceedings of the Royal Society of London. A. Mathematical and Physical Sciences 392, 45 (1984).
  98. [98]F. Wu and S. D. Sarma, Quantum geometry and stability of moiré flatband ferromagnetism, Physical Review B 102, 165118 (2020).DOI
  99. [99]V. Crépel, A. Hackenbroich, N. Regnault, and B. Estienne, Universal signatures of dirac fermions in entanglement and charge fluctuations, Physical Review B 103, 235108 (2021).DOI
  100. [100]V. Crépel, B. Estienne, and N. Regnault, Microscopic study of the coupled-wire construction and plausible realization in spin-dependent optical lattices, Physical Review B 101, 235158 (2020).DOI
  101. [101]A. Abouelkaam, K. Yang, and E. J. Bergholtz, Quantum metric induced phases in moiré materials, Physical Review Research 5, L12015 (2023).DOI
  102. [102]Y. Zeng, D. Guerci, V. Crépel, A. J. Millis, and J. Cano, Sublattice structure and topology in spontaneously crystallized electronic states, Physical Review Letters 132, 236601 (2024).DOI
  103. [103]Y. Fang, J. Cano, and S. A. A. Ghorashi, Quantum geometry induced nonlinear transport in alternamagnets, Physical Review Letters 133, 106701 (2024).
  104. [104]V. Crépel and J. Cano, Efficient prediction of superlattice and anomalous miniband topology from quantum geometry, Physical Review X 15, 011004 (2025).DOI
  105. [105]Y. Nambu, Quasi-particles and gauge invariance in the theory of superconductivity, Physical Review 117, 648 (1960).
  106. [106]J. Goldstone, Field theories with superconductor solutions, Il Nuovo Cimento (1955-1965) 19, 154 (1961).
  107. [107]H. B. Nielsen and S. Chadha, On how to count goldstone bosons, Nuclear Physics B 105, 445 (1976).
  108. [108]H. Watanabe and H. Murayama, Unified description of nambu-goldstone bosons without lorentz invariance, Physical Review Letters 108, 251602 (2012).
  109. [109]Y.-Z. Chou, Y. Tan, F. Wu, and S. Das Sarma, Topological flat bands, valley polarization, and interband superconductivity in magic-angle twisted bilayer graphene with proximized spin-orbit couplings, Physical Review B 110, L041108 (2024).DOI
  110. [110]J. Kang and O. Vafek, Strong coupling phases of partially filled twisted bilayer graphene narrow bands, Physical review letters 122, 246401 (2019).DOI
  111. [111]K. Seo, V. N. Kotov, and B. Uchoa, Ferromagnetic mott state in twisted graphene bilayers at the magic angle, Physical review letters 122, 246402 (2019).
  112. [112]R. Pons, A. Mielke, and T. Stauber, Flat-band ferromagnetism in twisted bilayer graphene, Physical Review B 102, 235101 (2020).
  113. [113]N. Bultinck, S. Chatterjee, and M. P. Zaletel, Mechanism for anomalous Hall ferromagnetism in twisted bilayer graphene, Physical review letters 124, 166601 (2020).
  114. [114]Y. Alavirad and J. Sau, Ferromagnetism and its stability from the one-magnon spectrum in twisted bilayer graphene, Physical Review B 102, 235123 (2020).DOI
  115. [115]N. Bultinck, E. Khalaf, S. Liu, S. Chatterjee, A. Vishwanath, and M. P. Zaletel, Ground state and hidden symmetry of magic-angle graphene at even integer filling, Physical Review X 10, 031034 (2020).DOI
  116. [116]B. A. Bernevig, B. Lian, A. Cowsik, F. Xie, N. Regnault, and Z.-D. Song, Twisted bilayer graphene. v. exact analytic many-body excitations in coulomb hamiltonians: Charge gap, goldstone modes, and absence of cooper pairing, Physical Review B 103, 205415 (2021).DOI
  117. [117]S. Becker, L. Lin, and K. D. Stubbs, Exact ground state of interacting electrons in magic angle graphene, Communications in Mathematical Physics 406, 1 (2025).DOI
  118. [118]F. Wu and S. Das Sarma, Collective excitations of quantum anomalous hall ferromagnets in twisted bilayer graphene, Physical review letters 124, 046403 (2020).DOI
  119. [119]B. A. Bernevig, T. L. Hughes, and S.-C. Zhang, Quantum spin hall effect and topological phase transition in hg te quantum wells, science 314, 1757 (2006).
  120. [120]W.-X. Qiu and F. Wu, Topological magnons and domain walls in twisted bilayer mote _2, arXiv preprint arXiv:2502.11010 (2025).DOI
  121. [121]M. Gonçalves, J. F. Mendez-Valderrama, J. Herzog-Arbeitman, J. Yu, X. Xu, D. Xiao, B. A. Bernevig, and N. Regnault, Spinless and spinful charge excitations in moir’e fractional chern insulators, arXiv preprint arXiv:2506.05330 (2025).
  122. [122]F. Wu, T. Lovorn, E. Tutuc, I. Martin, and A. MacDonald, Topological insulators in twisted transition metal dichalcogenide homobilayers, Physical review letters 122, 086402 (2019).
  123. [123]T. Devakul, V. Crépel, Y. Zhang, and L. Fu, Magic in twisted transition metal dichalcogenide bilayers, Nature communications 12, 6730 (2021).
  124. [124]C. Wang, X.-W. Zhang, X. Liu, Y. He, X. Xu, Y. Ran, T. Cao, and D. Xiao, Fractional chern insulator in twisted bilayer mote2, Phys. Rev. Lett. 132, 036501 (2024).
  125. [125]Y. Jia, J. Yu, J. Liu, J. Herzog-Arbeitman, Z. Qi, H. Pi, N. Regnault, H. Weng, B. A. Bernevig, and Q. Wu, Moiré fractional chern insulators. i. first-principles calculations and continuum models of twisted bilayer mote2, Phys. Rev. B 109, 205121 (2024).
  126. [126]J. Ahn, G.-Y. Guo, N. Nagaosa, and A. Vishwanath, Riemannian geometry of resonant optical responses, Nature Physics 18, 290 (2022).DOI
  127. [127]M. Tovmasyan, S. Peotta, P. Törmä, and S. D. Huber, Effective theory and emergent su (2) symmetry in the flat bands of attractive hubbard models, Physical Review B 94, 245149 (2016).

Paper details

vlnc:W_qmpyjw9s
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

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