betaFeedback

WorksPhysics and Astronomy

Orbital Magnetization from Uniform and Periodic Magnetic Fields

Abstract

Magnetization is thermodynamically defined as the derivative of the grand potential with respect to a uniform magnetic field. However, a uniform magnetic field makes the kinetic momentum operators noncommuting and Landau-quantizes the electron motion. This changes the zero-field momentum-space to Landau-levels and raises a fundamental question: how can the thermodynamic response to a uniform field be reproduced by a linear-response calculation carried out in the momentum space of the zero-field problem? We address this question analytically in a quantum Hall ferromagnet that allows the orbital magnetization MM to be computed in a closed form. We first compute MM from the local Hartree–Fock projector response to a periodic magnetic field with zero net flux. We then compute MM from the derivative of the grand potential with respect to a uniform magnetic field along the Středa line. The two approaches give the same result, even though the first keeps the Hilbert space fixed while the second changes the Landau-level degeneracy. Their agreement suggests that we should view orbital magnetization as the energy associated with the spectral flow that gives rise to the Středa formula. Our work provides a tutorial introduction to orbital magnetization and its relation to the Středa formula.

Introduction

Orbital magnetization has recently received renewed interest [1, 2, 3, 4, 5, 6, 7, 8, 9], driven in part by the experimental discovery of strongly correlated phases in moiré materials [10, 11, 12, 13, 14] and rhombohedral multilayer graphene [15, 16, 17] that break time-reversal symmetry through spontaneous valley polarization. At the same time, nano-SQUID magnetometry [18, 19] is now able to measure changes in magnetization relative to a reference state with high sensitivity, providing new opportunity to understand this fundamental thermodynamic quantity.

The theory of orbital magnetization in solid-state systems was developed extensively in the mid-2000s, leading to what is called the modern theory of orbital magnetization [20, 21, 22]. For a two-dimensional crystal, the orbital magnetization density is

M=∑n∫d2k(2π)2f(ϵnk)[mn(k)+eℏ(μ−ϵnk)Ωn(k)],(1)M = \sum_n \int\frac{d^2 k}{(2\pi)^2} f(\epsilon_{nk}) \left[m_n(k) + \frac{e}{\hbar}(\mu-\epsilon_{nk})\Omega_n(k)\right], \tag*{(1)}

where f(ϵnk)f(\epsilon_{nk}) is the occupation and μ\mu is the chemical potential. Here mn(k)m_n(k) [is the] intrinsic orbital magnetic moment of a Bloch wave packet (self-rotation) and the Ωn(k)\Omega_n(k) is the Berry-curvature contribution. This formula is remarkable because the thermodynamic definition of orbital magnetization as the derivative of the energy with respect to magnetic field [22], the semiclassical wave-packet dynamics in phase-space [21], and the microscopic definition in terms of the volume average of current loops [20] all lead to the same expression. Most early discussions, as well as this agreement among definitions, were developed primarily for noninteracting electrons.

Hartree–Fock approximation (HFA) provides a useful starting point for understanding interaction-induced spin-valley-polarized phases in two-dimensional materials, in part because it contains the exchange energy responsible for the flavor polarization observed in magnetotransport experiments. Recent work [8, 9, 23] has shown that, within HFA, Eq. (1) retains the same form provided that one uses the quasiparticle eigenstates and eigenvalues of the Hartree–Fock mean-field Hamiltonian.

Uniform BBPeriodic δBq\delta B_q
Vector potentialA(r)=12B×rA(\mathbf{r}) = \frac{1}{2}B \times rA(r)=Aqeiq⋅rA(\mathbf{r}) = A_q e^{i q \cdot r}
Magnetic fluxδΦ=AδB≠0\delta\Phi= A\delta B \ne0∫d2r B(r)=0\int d^2r\,\mathcal{B}(\mathbf{r}) = 0
MethodLandau quantizationFinite-qq perturbation
Středa response1A∂N∂B∣μ=Cϕ0\left.\frac{1}{A}\frac{\partial N}{\partial B}\right|_\mu= \frac{C}{\phi_0}δnq=Cϕ0δBq\delta n_q = \frac{C}{\phi_0}\delta B_q
Orbital magnetizationM=−∂∂B(E−μN)M=-\frac{\partial}{\partial B}(E-\mu N)δΩq=−MδBq\delta\Omega_q=-M\delta B_q

Table 1. Comparison between the uniform-field and periodic-field calculations of the Středa formula and orbital magnetization. A uniform field changes the total magnetic flux, whereas a periodic magnetic perturbation carries zero net flux and therefore preserves the zero-field Hilbert-space. We show that the orbital magnetizations obtained from the two approaches agree with each other.

A subtle point clarified in [23, 24] is that while the vector potential couples microscopically to the bare velocity ∂kTk\partial_k T_k, where TkT_k is the kinetic energy Hamiltonian. Nevertheless, after the required vertex corrections are included, the velocity entering the orbital-magnetization formula is the Hartree–Fock velocity ∂kHkHF\partial_k H_k^{\mathrm{HF}}, as required by the Ward identity.

The purpose of this manuscript is to resolve some subtleties in defining orbital magnetization and to provide a deeper understanding of this fundamental quantity. The problem is the following. A uniform magnetic field enters quantum mechanics through a vector potential which grows linearly with position, e.g. A=B×r/2A = B \times r/2. This field replaces the commuting momentum variables kx,kyk_x,k_y of the zero-field Hamiltonian by non-commuting momentum operators [Πx,Πy]=ieB/ℏ[\Pi_x,\Pi_y] = ieB/\hbar. Thus, a uniform field changes the kinematic structure of the Hamiltonian and requires Landau quantization. For example, in monolayer graphene, the linear dispersion vFkv_F k is reorganized into the well-known Dirac Landau levels, ϵN=±vF2eℏNB\epsilon_N = \pm v_F\sqrt{2e\hbar NB}. The B\sqrt{B} dependence cannot be obtained from ordinary perturbation theory around a zero-field momentum eigenstate in the form ϵn(k)=ϵn0(k)+BΣn(1)(k)\epsilon_n(k) = \epsilon_n^0(k) + B\Sigma_n^{(1)}(k). Moreover, changing $B also changes the number of flux quanta and hence the Landau-level degeneracy. This raises a basic question: how can orbital magnetization be formulated as a linear response to a uniform magnetic field when a uniform field is not an ordinary perturbation of the zero-field Bloch Hamiltonian, but instead reorganizes the spectrum into Landau levels?

In Ref. [23], following Ref. [22], we avoided this difficulty by perturbing the system with a periodic magnetic field carrying zero net flux. The new ingredient in our approach is to study the resulting change of the Hartree–Fock projector. Because the perturbing field carries no net flux, it does not change the total number of flux quanta through the sample, and the response can be computed within a fixed zero-field Hilbert-space using standard perturbation theory. Here we show, by an explicit analytical calculation in a simple model, that the two procedures described in Table. 1 give the same orbital magnetization. We then comment on what this equivalence implies.

Model Hamiltonian

The toy model Hamiltonian we consider is inspired by the adiabatic model of twisted homobilayers semiconductor [25]. The spin-up electron experiences a uniform effective magnetic field arising from the layer-skyrmion texture, while the spin-down electron experiences the opposite effective field. The total system is therefore time-reversal invariant. The NN electron Hamiltonian is

H=∑i=1N[pi+esiAsky(ri)+eAext(ri)]22m+12∑i≠jV(rij),(2)H = \sum_{i=1}^{N} \frac{\left[\mathbf{p}_i + e s_i \mathbf{A}_{\mathrm{sky}}(\mathbf{r}_i) + e\mathbf{A}_{\mathrm{ext}}(\mathbf{r}_i)\right]^2}{2m} + \frac{1}{2}\sum_{i\ne j} V(r_{ij}), \tag*{(2)}

where si=+1s_i=+1 for spin up and si=−1s_i=-1 for spin down. Here the layer-skyrmion vector potential is,

Asky(r)=12Bsky×r,(3)\mathbf{A}_{\mathrm{sky}}(\mathbf{r}) = \frac{1}{2}\mathbf{B}_{\mathrm{sky}}\times\mathbf{r}, \tag*{(3)}

where Bsky=Bskyz^\mathbf{B}_{\mathrm{sky}}=B_{\mathrm{sky}}\hat{\mathbf{z}} is uniform and the magnitude depends on the material properties such as twist angles. For simplicity, we do not consider Zeeman coupling. A uniform external magnetic field Bext\mathbf{B}_{\mathrm{ext}} can be introduced in the conventional way,

Aext(r)=12Bext×r.(4)\mathbf{A}_{\mathrm{ext}}(\mathbf{r}) = \frac{1}{2}\mathbf{B}_{\mathrm{ext}}\times\mathbf{r}. \tag*{(4)}

Thus the total effective magnetic field seen by particle ii is

Bieff=siBsky+Bext.(5)\mathbf{B}^{\mathrm{eff}}_i = s_i\mathbf{B}_{\mathrm{sky}}+\mathbf{B}_{\mathrm{ext}}. \tag*{(5)}

This model is convenient because the external magnetic field can be introduced simply by tuning Bext\mathbf{B}_{\mathrm{ext}}.

When Bext=0B_{\mathrm{ext}}=0, each spin species forms Landau levels with degeneracy

Nϕ=BskyAϕ0,(6)N_{\phi} = \frac{B_{\mathrm{sky}}A}{\phi_0}, \tag*{(6)}

Periodic-field calculation

where ϕ0=h/e\phi_0=h/e is the flux quantum. We study the groundstate at electron density

ne=NA=NϕA.(7)n_e = \frac{N}{A} = \frac{N_{\phi}}{A}. \tag*{(7)}

Thus the two spin-projected lowest Landau levels are half filled in total. We refer to this filling as ν=1\nu=1. For a wide range of repulsive interactions [26], including the Coulomb interaction with weak Landau-level mixing [27], the ground state at ν=1\nu=1 is the familiar quantum Hall ferromagnet where electrons completely occupied one spin-projected Landau level,

∣Ψ⟩=∏X=1Nϕc0X↑†∣0⟩,(8)|\Psi\rangle= \prod_{X=1}^{N_{\phi}} c^\dagger_{0X\uparrow}|0\rangle, \tag*{(8)}

Here XX is the guiding-center. For simplicity, we consider the long range Coulomb potential

V(rij)=e2e∣ri−rj∣.(9)V(r_{ij}) = \frac{e^2}{e|\mathbf{r}_i-\mathbf{r}_j|}. \tag*{(9)}

We now review the approach of Ref. [23], where the orbital magnetization was derived within the mean-field approximation. Instead of applying a uniform magnetic field, one studies the response of the Hartree–Fock density matrix to a weak periodic vector potential

A(r^)=Aqeiq⋅r.(10)\mathbf{A}(\hat{\mathbf{r}})=\mathbf{A}_q e^{i\mathbf{q}\cdot\mathbf{r}}. \tag*{(10)}

The corresponding magnetic field is

Bq=i(q×Aq)⋅z^,(11)B_q=i(\mathbf{q}\times\mathbf{A}_q)\cdot\hat{\mathbf{z}}, \tag*{(11)}

which is also periodic and has zero net flux. To leading order in Aq\mathbf{A}_q, the single-particle density matrix changes as

P→P+δPq.(12)P\to P+\delta P_q. \tag*{(12)}

where

δPq=eℏ∑cv(Aq⋅⟨uck∣∇kuvk+q⟩⟨uvk∣+A−q∗⋅⟨uck∣∇kuvk⟩∗∣uvk⟩⟨uck−q∣).(13)\delta P_q=\frac{e}{\hbar}\sum_{cv}\left(\mathbf{A}_q\cdot\langle u_{ck}|\nabla_k u_{vk+q}\rangle\langle u_{vk}|+\mathbf{A}_{-q}^{*}\cdot\langle u_{ck}|\nabla_k u_{vk}\rangle^{*}|u_{vk}\rangle\langle u_{ck-q}|\right). \tag*{(13)}

Here vv and cc label the occupied (unoccupied) Hartree–Fock bands. From δPq\delta P_q, we found that the induced density has the Streda form,

δnq=Tr⁡δPqA=Cϕ0Bq,(14)\delta n_q=\frac{\operatorname{Tr}\delta P_q}{A}=\frac{C}{\phi_0}B_q, \tag*{(14)}

where CC is the Chern number of the occupied Hartree–Fock band,

C=ϵij∑cv∫d2k2πIm⁡[⟨∂kiuvk∣uck⟩⟨uck∣∂kjuvk⟩](15)C=\epsilon_{ij}\sum_{cv}\int\frac{d^2k}{2\pi}\operatorname{Im}\left[\langle\partial_{k_i}u_{vk}|u_{ck}\rangle\langle u_{ck}|\partial_{k_j}u_{vk}\rangle\right] \tag*{(15)}
=12π∫d2k Ωz(k).(16)=\frac{1}{2\pi}\int d^2k\,\Omega_z(\mathbf{k}). \tag*{(16)}

For metal, the Chern number CC should be replaced by integrated Berry curvature of the occupied states.

The same density-matrix response that gives rise to the local charge density also changes the local grand-potential density, thereby defining the orbital magnetization,

δΩq=Tr⁡Ω^ δPqA=−MδBq,(17)\delta\Omega_q = \frac{\operatorname{Tr}\hat{\Omega}\,\delta P_q}{A} = -M\delta B_q, \tag*{(17)}

where Ω^=HMF−μN^\hat{\Omega}=H^{\mathrm{MF}}-\mu\hat{N}. The resulting expression for the orbital magnetization is Eq. (44) of Ref. [23], which we reproduce here:

M=eℏ∑cv∫d2k(2π)2(ξck+ξvk2)×ϵijIm⁡[⟨∂kiuvk∣uck⟩⟨uck∣∂kjuvk⟩].(18)M=\frac{e}{\hbar}\sum_{cv}\int\frac{d^2k}{(2\pi)^2}\left(\frac{\xi_{ck}+\xi_{vk}}{2}\right)\times\epsilon_{ij}\operatorname{Im}\left[\langle\partial_{k_i}u_{vk}|u_{ck}\rangle\langle u_{ck}|\partial_{k_j}u_{vk}\rangle\right]. \tag*{(18)}

We have defined ξnk=ϵnkHF−μ\xi_{nk}=\epsilon_{nk}^{\mathrm{HF}}-\mu, where ϵnkHF\epsilon_{nk}^{\mathrm{HF}} is the Hartree–Fock quasiparticle energy and μ\mu is the chemical potential. As shown in the appendix of Ref. [23], Eq. (18) can be rewritten in the same form as Eq. (1). Comparing Eq. (18) with Eqs. (14) and (15), we see that orbital magnetization is the energy-weighted response of the same spectral flow that creates the local density modulation in the Streda formula, with weight (ξck+ξvk)/2(\xi_{ck}+\xi_{vk})/2.

This perspective provides several useful insights that are not usually emphasized in the literature. First, the orbital magnetization vanishes identically in a particle-hole-symmetric two-band model since ξck=−ξvk\xi_{ck}=-\xi_{vk}. Second, the orbital magnetization has no Fermi surface contribution (at zero temperature) because the energy weight vanishes at the Fermi surface. Orbital magnetization is really the response of the occupied states (Fermi-sea) to external magnetic field. This is different from the intrinsic anomalous Hall conductivity in metal, where the non-quantized part is a Fermi-surface property [28]. Third, in general, computing orbital magnetization requires information about the wave functions and energies of remote unoccupied states. A two-band model is a special case, where the single unoccupied band is fixed once the occupied band is known.

Although Eq. (1) is expressed in terms of sum over occupied states, once mn(k)m_n(k) and Ωn(k)\Omega_n(k) are known, the computation of the self-rotation term

mn(k)=−ie2ℏ⟨∇kunk∣×[Hk−ϵnk]∣∇kunk⟩(19)m_n(k)=-\frac{ie}{2\hbar}\langle\nabla_k u_{nk}|\times[H_k-\epsilon_{nk}]|\nabla_k u_{nk}\rangle \tag*{(19)}

is not determined by the occupied state alone. This is because

Hk∣∇kunk⟩=∑mϵmk∣umk⟩⟨∂kunk∣H_k|\nabla_k u_{nk}\rangle=\sum_m\epsilon_{mk}|u_{mk}\rangle\langle\partial_k u_{nk}|

and mm includes unoccupied states too. In fact, only the interband Berry connection between occupied and unoccupied states enters the orbital magnetization formula, as shown in Eq. (18). Fundamentally, this is because when we evaluate the expectation value of the current-loop operator ⟨n∣r×v∣n⟩\langle n|\mathbf{r}\times\mathbf{v}|n\rangle, one operator (e.g. vv) can take the occupied state to the unoccupied states and then the other operator (e.g. rr) bring it back.

Let us now apply this formalism to the ν=1\nu=1 quantum Hall ferromagnet ∣Ψ⟩|\Psi\rangle defined in previous section. We set Bext=0B_{\mathrm{ext}}=0 and apply a periodic magnetic field. This field induces both a density modulation, δnq=CδBq/ϕ0\delta n_q=C\delta B_q/\phi_0, and a change in the grand-potential density, δΩq=−MδBq\delta\Omega_q=-M\delta B_q. Then, we compute MM using Eq. (18). Since Eq. (18) is written in momentum space, we first can carry out a basis transformation from guiding-center to magnetic Bloch states ∣unk⟩|u_{nk}\rangle. We choose the real-space magnetic unit cell to enclose one flux quantum, so that the number of allowed kk points in the magnetic Brillouin zone is NϕN_\phi. The shape of the magnetic Brillouin zone can be arbitrary. The quantum Hall ferromagnet can then be written equivalently as

∣Ψ⟩=∏kNϕc0k↑†∣0⟩.(20)|\Psi\rangle=\prod_{k}^{N_\phi}c_{0k\uparrow}^{\dagger}|0\rangle. \tag*{(20)}

The area of the magnetic Brillouin zone AMBZA_{\mathrm{MBZ}} is such that,

AMBZ4π2=12πℓB2,ℓB=ℏeBsky(21)\frac{A_{\mathrm{MBZ}}}{4\pi^2}=\frac{1}{2\pi\ell_B^2},\qquad\ell_B=\sqrt{\frac{\hbar}{eB_{\mathrm{sky}}}} \tag*{(21)}

We emphasize that because the perturbing magnetic field BqB_q carries zero net flux, this set of magnetic Bloch states remains fixed throughout the calculation.

The Hartree–Fock quasiparticle energies are “flat-bands”:

ξnk=ϵnHF−μ=(n+12)ℏωc+Σn−μ,(22)\xi_{nk}=\epsilon_n^{\mathrm{HF}}-\mu=\left(n+\frac{1}{2}\right)\hbar\omega_c+\Sigma_n-\mu, \tag*{(22)}

where n=0,1,2,…n=0,1,2,\ldots and

Σn=−∑X′⟨0X,nX′∣V∣nX′,0X⟩,(23)\Sigma_n=-\sum_{X'}\langle0X,nX'|V|nX',0X\rangle, \tag*{(23)}

is the Fock self-energy of an electron in the nn-th same-spin Landau level due to the filled n=0n=0 Landau level.

In the present problem, the occupied band is v=(n=0,↑)v=(n=0,\uparrow), and the unoccupied bands are the higher same-spin Landau levels, c=(n,↑)c=(n,\uparrow) with n=1,2,3,…n=1,2,3,\ldots since the applied vector potential couples does not flip spin. Because the Landau-level does not disperse with kk, we can take the energy factor in Eq. (18) outside the momentum integral and obtain

M=eℏ∑n=1∞[n+12ℏωc+Σn+Σ02−μ]Jn.(24)M=\frac{e}{\hbar}\sum_{n=1}^{\infty}\left[\frac{n+1}{2}\hbar\omega_c+\frac{\Sigma_n+\Sigma_0}{2}-\mu\right]\mathcal{J}_n. \tag*{(24)}

The term in the square bracket is the average energy between nn-th unoccupied Landau level and the n=0n=0 occupied Landau level. The quantity Jn\mathcal{J}_n is the nn-th Landau level contribution to the Berry-curvature integral,

Jn=∫MBZd2k4π2ϵijI[⟨∂kiu0k∣unk⟩⟨unk∣∂kju0k⟩].(25)\mathcal{J}_n=\int_{\mathrm{MBZ}}\frac{d^2k}{4\pi^2}\epsilon_{ij}\mathfrak{I}\left[\langle\partial_{k_i}u_{0k}|u_{nk}\rangle\langle u_{nk}|\partial_{k_j}u_{0k}\rangle\right]. \tag*{(25)}

Because the mean-field approximation for a parabolic-band electron gas does not change the Landau-level wavefunctions, the magnetic Bloch wavefunctions have the same form as in the noninteracting problem. The reason is because the quantum Hall ferromagnet is an isotropic electron-fluid. Therefore, we can use the Feynman–Hellmann relation and Landau-level ladder operators to evaluate the matrix elements. Since the kinetic momentum Π\mathbf{\Pi} is linear in the ladder operators aa and a†a^\dagger, the n=0n=0 Landau level only mixes with the n=1n=1 Landau level:

⟨unk∣∂kxu0k⟩=−ℓB2δn,1,(26)\langle u_{nk}|\partial_{k_x}u_{0k}\rangle=-\frac{\ell_B}{\sqrt{2}}\delta_{n,1}, \tag*{(26)}
⟨unk∣∂kyu0k⟩=+iℓB2δn,1.(27)\langle u_{nk}|\partial_{k_y}u_{0k}\rangle=+\frac{i\ell_B}{\sqrt{2}}\delta_{n,1}. \tag*{(27)}
ϵijℑ[⟨∂kiu0k∣unk⟩⟨unk∣∂kju0k⟩]=−ℓB2δn,1.(28)\epsilon_{ij}\Im\left[\langle\partial_{k_i}u_{0k}|u_{nk}\rangle\langle u_{nk}|\partial_{k_j}u_{0k}\rangle\right]=-\ell_B^2\delta_{n,1}. \tag*{(28)}

It follows that

Jn=−ℓB2δn,1AMBZ4π2=−12πδn,1.(29)\mathcal{J}_n=-\ell_B^2\delta_{n,1}\frac{A_{\mathrm{MBZ}}}{4\pi^2}=-\frac{1}{2\pi}\delta_{n,1}. \tag*{(29)}

Substituting Eq. (29) into Eq. (24), only the n=1n=1 unoccupied Landau level contributes:

Mz=−e2πℏ[ℏωc+Σ1+Σ02−μ]=1ϕ0[μ−ℏωc+34π2e2ϵℓB].(30)\begin{aligned} M_z=-\frac{e}{2\pi\hbar}\left[\hbar\omega_c+\frac{\Sigma_1+\Sigma_0}{2}-\mu\right] \\ &=\frac{1}{\phi_0}\left[\mu-\hbar\omega_c+\frac{3}{4}\sqrt{\frac{\pi}{2}}\frac{e^2}{\epsilon\ell_B}\right]. \tag*{(30)} \end{aligned}

We used Σ1=12Σ0\Sigma_1=\frac{1}{2}\Sigma_0 in the last equation. This relation can be derived from standard gaussian integral,

Σn=−∫d2q(2π)2V(q)∣Fn0(q)∣2,(31)\Sigma_n=-\int\frac{d^2q}{(2\pi)^2}V(q)|F_{n0}(q)|^2, \tag*{(31)}

where the form factors are

∣F00(q)∣2=e−q2ℓB2/2,∣F10(q)∣2=q2ℓB22e−q2ℓB2/2.(32)|F_{00}(q)|^2=e^{-q^2\ell_B^2/2},\qquad|F_{10}(q)|^2=\frac{q^2\ell_B^2}{2}e^{-q^2\ell_B^2/2}. \tag*{(32)}

Therefore

Σ0=−e2ϵ∫0∞dq e−q2ℓB2/2=−π2e2ϵℓB,(33)\Sigma_0=-\frac{e^2}{\epsilon}\int_0^\infty dq\,e^{-q^2\ell_B^2/2}=-\sqrt{\frac{\pi}{2}}\frac{e^2}{\epsilon\ell_B}, \tag*{(33)}
Σ1=−e2ϵ∫0∞dq q2ℓB22e−q2ℓB2/2=−12π2e2ϵℓB.(34)\Sigma_1=-\frac{e^2}{\epsilon}\int_0^\infty dq\,\frac{q^2\ell_B^2}{2}e^{-q^2\ell_B^2/2}=-\frac{1}{2}\sqrt{\frac{\pi}{2}}\frac{e^2}{\epsilon\ell_B}. \tag*{(34)}

Uniform-field calculation

As mentioned in Section II, adding a uniform external magnetic field in our toy model simply changes the effective magnetic field seen by each spin species,

Bieff=siBsky+Bext.(35)B_i^{\mathrm{eff}}=s_iB_{\mathrm{sky}}+B_{\mathrm{ext}}. \tag*{(35)}

For the ν=1\nu=1 quantum Hall ferromagnet considered here where only spin up is occupied, they experience a net magnetic field B↑eff=Bsky+BextB_{\uparrow}^{\mathrm{eff}}=B_{\mathrm{sky}}+B_{\mathrm{ext}}.

We consider the response at fixed chemical potential, with μ\mu lying inside the exchange gap. As the external flux is increased, the degeneracy of the spin-up Landau level increases.

Along the Středa line, ∂n/∂B∣μ=1/ϕ0\left.\partial n/\partial B\right|_\mu=1/\phi_0, the particle number changes in such a way that the lowest Landau level remains completely filled. Thus the ground state is a family of quantum Hall ferromagnets parametrized by the external field,

∣Ψ(λ)⟩=∏X=1N(λ)c0X↑†∣0⟩.(36)|\Psi(\lambda)\rangle=\prod_{X=1}^{N(\lambda)}c^\dagger_{0X\uparrow}|0\rangle. \tag*{(36)}

Here N(λ)N(\lambda) is the Landau-level degeneracy in the presence of the external magnetic field,

N(λ)=(Bsky+B)Aϕ0=Nϕ(1+λ),λ=BBsky.(37)N(\lambda)=\frac{(B_{\mathrm{sky}}+B)A}{\phi_0}=N_\phi(1+\lambda),\qquad\lambda=\frac{B}{B_{\mathrm{sky}}}. \tag*{(37)}

This is what we mean when we say that the uniform-field calculation takes place in a varying magnetic Hilbert space. In a crystal, this is more dramatic. When the flux per unit cell is rational, α=Φcellϕ0=pq\alpha=\frac{\Phi_{\mathrm{cell}}}{\phi_0}=\frac{p}{q}, an ordinary Bloch band splits into qq magnetic subbands. For example, at α=1/5\alpha=1/5, one ordinary band splits into five magnetic Bloch bands. If the flux is decreased slightly to α=9/50\alpha=9/50, each of the magnetic Bloch bands at α=1/5\alpha=1/5 is further split into ten narrower magnetic subbands. The continuity of this spectrum between rational α\alpha and irrational α\alpha is discussed in Ref. [29].

The stability of ∣Ψ(λ)⟩|\Psi(\lambda)\rangle as a function of λ\lambda was studied in detail in Ref. [30]. For parameters relevant to twisted MoTe2_2, the incompressible state along the Středa line pointing away from charge neutrality remains stable. We can therefore compute the orbital magnetization by differentiating the grand potential with respect to the external magnetic field at fixed chemical potential,

M=−1A∂Ω∂B∣μ.(38)M=-\left.\frac{1}{A}\frac{\partial\Omega}{\partial B}\right|_\mu. \tag*{(38)}

The expectation value of the Hamiltonian in the state ∣Ψ(λ)⟩|\Psi(\lambda)\rangle is

E(λ)=⟨Ψ(λ)∣H∣Ψ(λ)⟩(39)E(\lambda)=\langle\Psi(\lambda)|H|\Psi(\lambda)\rangle \tag*{(39)}
=N(λ)[12ℏωc(λ)+12Σ0(λ)],(40)=N(\lambda)\left[\frac{1}{2}\hbar\omega_c(\lambda)+\frac{1}{2}\Sigma_0(\lambda)\right], \tag*{(40)}

where the first term is the zero-point energy and the second term is the exchange energy Σ0(λ)=−π2e2ϵℓ(λ)\Sigma_0(\lambda)=-\sqrt{\frac{\pi}{2}}\frac{e^2}{\epsilon\ell(\lambda)}. The cyclotron frequency and magnetic length depends on λ\lambda,

ωc(λ)=e(Bsky+B)m=ωc(1+λ)(41)\omega_c(\lambda)=\frac{e(B_{\mathrm{sky}}+B)}{m}=\omega_c(1+\lambda) \tag*{(41)}
ℓ(λ)=ℏe(Bsky+B)=ℓB1+λ,(42)\ell(\lambda)=\sqrt{\frac{\hbar}{e(B_{\mathrm{sky}}+B)}}=\frac{\ell_B}{\sqrt{1+\lambda}}, \tag*{(42)}

Here ωc=eBsky/m\omega_c=eB_{\mathrm{sky}}/m and ℓB=ℏ/(eBsky)\ell_B=\sqrt{\hbar/(eB_{\mathrm{sky}})} are the values at zero external field. The grand potential changes with λ\lambda as

Ω(λ)=E(λ)−μN(λ)(43)\Omega(\lambda)=E(\lambda)-\mu N(\lambda) \tag*{(43)}

Explicitly, the grand-potential density is

Ω(λ)A=Bskyϕ0[12ℏωc(1+λ)2−12π2e2ϵℓB(1+λ)3/2−μ(1+λ)].(44)\frac{\Omega(\lambda)}{A}=\frac{B_{\mathrm{sky}}}{\phi_0}\left[\frac{1}{2}\hbar\omega_c(1+\lambda)^2-\frac{1}{2}\sqrt{\frac{\pi}{2}}\frac{e^2}{\epsilon\ell_B}(1+\lambda)^{3/2}-\mu(1+\lambda)\right]. \tag*{(44)}

This expression is valid for all external field. The orbital magnetization is then obtained by evaluating the slope of the grand potential at λ=0\lambda=0,

M=−∂∂BΩA∣μ,B=0=−1Bsky∂∂λΩ(λ)A∣μ,λ=0=−1ϕ0[ℏωc(1+λ)−34π2e2ϵℓB(1+λ)1/2−μ]∣λ=0=1ϕ0[μ−ℏωc+34π2e2ϵℓB].(45)M=-\left.\frac{\partial}{\partial B}\frac{\Omega}{A}\right|_{\mu,B=0} =-\left.\frac{1}{B_{\mathrm{sky}}}\frac{\partial}{\partial\lambda}\frac{\Omega(\lambda)}{A}\right|_{\mu,\lambda=0} =-\left.\frac{1}{\phi_0}\left[\hbar\omega_c(1+\lambda)-\frac{3}{4}\sqrt{\frac{\pi}{2}}\frac{e^2}{\epsilon\ell_B}(1+\lambda)^{1/2}-\mu\right]\right|_{\lambda=0} =\frac{1}{\phi_0}\left[\mu-\hbar\omega_c+\frac{3}{4}\sqrt{\frac{\pi}{2}}\frac{e^2}{\epsilon\ell_B}\right]. \tag*{(45)}

The orbital magnetization obtained from the ground-state energy calculation in a uniform field agrees with the periodic-field result in the previous section Eq. (30). The agreement is nontrivial, in my view. For example, the factor of 3/43/4 in front of the self-energy π2e2ϵℓB\sqrt{\frac{\pi}{2}}\frac{e^2}{\epsilon\ell_B} appears in two completely different ways.

Discussion

The formula derived in Ref. [23] perturbs the zero-field Hartree-Fock density matrix with a periodic magnetic field. To leading order, the perturbation induces particle-hole components in the density matrix, mixing occupied states of the Hartree-Fock Slater determinant with unoccupied states. This occupied–unoccupied mixing is the microscopic origin of the local Středa response. Regions with δB(r)>0\delta B(\mathbf{r})>0 has an enhanced local density, while regions with δB(r)<0\delta B(\mathbf{r})<0 has a reduced local density. For parabolic-dispersion electron gas, this is generated by mixing the occupied n=0n=0 Landau level with the unoccupied n=1n=1 Landau level. The orbital magnetization is equivalent to the energy associated with this field-induced mixing of the wavefunction [23]. The energy-weight factor is given by the average Hartree-Fock energy of the occupied n=0n=0 and unoccupied n=1n=1 level. The exchange part of this average energy is Σ0+Σ12=34Σ0\frac{\Sigma_0+\Sigma_1}{2}=\frac{3}{4}\Sigma_0.

By contrast, in the uniform-field calculation, increasing BB increases the Landau-level degeneracy. At fixed chemical potential inside the exchange gap, this increases the number of electrons in the majority-spin lowest Landau level. The information about the unoccupied Landau levels do not enter explicitly in this method. Instead, the same factor of 3/43/4 arises from differentiating the total exchange-energy density with respect to BB.

The agreement between the two calculations is therefore nontrivial. In the periodic-field calculation, the Hilbert space is held fixed, and the magnetic perturbation induces Landau-level mixing between the n=0n=0 and n=1n=1 Landau level. In the uniform-field calculation, by contrast, we evaluate the groundstate energy in the Hilbert space appropriate to the new value of BB, where the Landau-level degeneracy has changed. Their agreement shows that the local projector response δPq\delta P_{\mathbf{q}} encodes the same physics as the global change in Landau-level degeneracy.

The spirit of linear response is to compute things without re-diagonalizing the Hamiltonian but nevertheless uniform BB-field described by unbounded vector potential necessarily involves Landau-quantization of the zero field Hamiltonian. However, if we understand orbital magnetization as the energy associated with spectral flow, one is global spectral flow, the other is local spectral flow, then it is not so surprising that they give rise to the same result. The microscopic details of spectral flow, say mixing 0 and 1 Landau level, are system dependent, but the resulting orbital magnetization is the same.

A similar comparison was made in Ref. [8] for the Rashba model. Since the free energy is known for arbitrary external field in this model, it can also be used to test the magnetic susceptibility by computing ∂2Ω/∂λ2\partial^2\Omega/\partial\lambda^2 from the uniform-field free energy and compare it with a zero-field momentum space calculation.

Středa formula and the quantum Hall effect

We have presented orbital magnetization as an energy associated with changing particle via Streda formula. It is therefore natural to ask whether the same viewpoint can be related to the quantum Hall effect. It turns out that it can, provided the bulk charge response is gapped so that the long-wavelength and low-frequency limits commute.

Let us assume, for the moment, that the Středa formula in Eq. (14) remains valid at finite frequency,

δn(q,ω)=Cϕ0δBz(q,ω).(46)\delta n(q,\omega)=\frac{C}{\phi_0}\delta B_z(q,\omega). \tag*{(46)}

We take q=qx^\mathbf{q}=q\hat{x} and choose a transverse vector potential

A(r,t)=Ay(q,ω)ei(qx−ωt)y^,(47)\mathbf{A}(r,t)=A_y(q,\omega)e^{i(qx-\omega t)}\hat{y}, \tag*{(47)}

with vanishing scalar potential. Then,

δBz(q,ω)=iqAy(q,ω),Ey(q,ω)=iωAy(q,ω).(48)\delta B_z(q,\omega)=iqA_y(q,\omega),\quad E_y(q,\omega)=i\omega A_y(q,\omega). \tag*{(48)}

Therefore the “finite-frequency” Středa formula relates particle density to electric field,

δn(q,ω)=Cϕ0iqAy(q,ω)=Cϕ0qωEy(q,ω).(49)\delta n(q,\omega)=\frac{C}{\phi_0}iqA_y(q,\omega)=\frac{C}{\phi_0}\frac{q}{\omega}E_y(q,\omega). \tag*{(49)}

We can next use the continuity equation for particle number,

−iω δn(q,ω)+iqJx(q,ω)=0,(50)-i\omega\,\delta n(q,\omega)+iqJ_x(q,\omega)=0, \tag*{(50)}

to arrive at the quantized Hall response,

jx(q,ω)=ωqδn(q,ω)=CΦ0Ey(q,ω).(51)j_x(q,\omega)=\frac{\omega}{q}\delta n(q,\omega)=\frac{C}{\Phi_0}E_y(q,\omega). \tag*{(51)}

Multiplying by the electron charge gives the quantized Hall conductivity,

σxy=Ce2h.(52)\sigma_{xy}=C\frac{e^2}{h}. \tag*{(52)}

There is a very important assumption hidden in this derivation. The Středa formula is a thermodynamic response: it is derived at ω=0\omega=0 and then taking the long-wavelength limit q→0q\to0. The Hall conductivity is a transport response: we first take the uniform limit q→0q\to0 and then ω→0\omega\to0 later. The derivation above is valid only if these two limits commute. These two limits are expected to commute for many-body states with no gapless bulk charge excitations. In that case the electromagnetic response is regular near (q,ω)=(0,0)(q,\omega)=(0,0), and the same coefficient controls both the static density response and the Hall current. In a metal, these two limits clearly do not commute due to the Fermi surface [31], and the Hall conductivity derived from our argument above is not the full Hall conductivity. In fact, our argument can only give the σxyII\sigma_{xy}^{II} contribution in Ref. [32]. A proper calculation for a metal must also recover the Fermi-surface contribution, σxyI\sigma_{xy}^{I} in Ref. [32].

Acknowledgment I am grateful to Giovanni Vignale for helpful discussions and for encouraging me to publish this note. This work is supported by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences, under Award Number DE-SC-0024346.

References

References

  1. [1]K. J. Vidarte, H. P. Veiga, J. M. Lopes, R. Cardias, A. Ferreira, T. P. Cysne, and T. G. Rappoport, “Real-space spectral approach to orbital magnetization,” arXiv preprint arXiv:2512.01575, 2025.DOI
  2. [2]Saati, S. A., Hur, K. L., & Piéchon, F. (2025). Theory of local orbital magnetization: local berry curvature. In arXiv preprint. arXiv:2512.12343
  3. [3]X. Chen and Z.-D. Song, “Orbital magnetization of interacting electrons,” arXiv preprint arXiv:2602.02478, 2026.arXiv
  4. [4]P. Li, K. Jiang, Z. Wang, J. Kang, and Y. Zhang, “Contribution of remote bands to orbital magnetization in twisted bilayer graphene,” arXiv preprint arXiv:2603.03702, 2026.arXiv
  5. [5]M. Das and C. Huang, “Unconventional metallic ferromagnetism: Nonanalyticity and sign-changing behavior of orbital magnetization in rhombohedral trilayer graphene,” Physical Review B, vol. 109, no. 6, p. L060409, 2024.DOI
  6. [6]M. Das and C. Huang, “Microscopic theory of multiferroic behavior in rhombohedral multilayer graphene,” Physical Review B, vol. 111, no. 16, p. L161106, 2025.DOI
  7. [7]M. Ye, “A quantum many-body approach for orbital magnetism in correlated multiband electron systems,” arXiv preprint arXiv:2601.14372, 2026.arXiv
  8. [8]J. Kang, M. Wang, and O. Vafek, “Orbital magnetization and magnetic susceptibility of interacting electrons,” arXiv preprint arXiv:2509.20626, 2025.DOI
  9. [9]X. Liu, C. Wang, H. Chen, X.-W. Zhang, T. Cao, and D. Xiao, “Orbital magnetization of correlated states in twisted bilayer transition metal dichalcogenides,” Physical Review Letters, vol. 136, no. 16, p. 166606, 2026.DOI
  10. [10]A. L. Sharpe, J. E. Fox, A. W. Barnard, J. Finney, K. Watanabe, T. Taniguchi, M. Kastner, and D. Goldhaber-Gordon, “Emergent ferromagnetism near three-quarters filling in twisted bilayer graphene,” Science, vol. 365, no. 6453, pp. 605–608, 2019.DOI
  11. [11]Zeng, Y., Xia, Z., Kang, K., & al., . et . (2023). Thermodynamic evidence of fractional Chern insulator in moiré mo te 2. Nature, 622, 69–73.
  12. [12]J. Cai, E. Anderson, C. Wang, X. Zhang, X. Liu, W. Holtzmann, Y. Zhang, F. Fan, T. Taniguchi, K. Watanabe, Y. Ran, T. Cao, L. Fu, D. Xiao, W. Yao, and X. Xu, “Signatures of fractional quantum anomalous hall states in twisted mote2,” Nature, vol. 622, no. 7981, pp. 63–68, 2023.DOI
  13. [13]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, vol. 622, pp. 74–79, Oct 2023.DOI
  14. [14]Xu, F., Sun, Z., Jia, T., Liu, C., Xu, C., Li, C., Gu, Y., Watanabe, K., Taniguchi, B., Jia, J., Shi, S., Jiang, Y., Zhang, X., Liu, X., & Liu, T. (2023). Observation of integer and fractional quantum anomalous hall effects in twisted bilayer mote2. Physical Review X, 13, 031037.
  15. [15]Zhou, H., Holleis, L., Saito, Y., Cohen, L., Huynh, W., Patterson, C. L., Yang, F., Taniguchi, T., Watanabe, K., & Young, A. F. (2022). Isospin magnetism and spin-polarized superconductivity in bernal bilayer graphene. Science, 375(6582), 774–778.
  16. [16]Han, T., Lu, Z., Scuri, G., Sung, J., Wang, J., Han, T., Watanabe, K., Taniguchi, T., Fu, L., Park, H., & al., . et . (2023). Orbital multiferroicity in pentalayer rhombohedral graphene. Nature, 623(7985), 41–47.
  17. [17]J. Deng, J. Xie, H. Li, T. Taniguchi, K. Watanabe, J. Shan, K. F. Mak, and X. Liu, “Superconductivity and ferroelectric orbital magnetism in semimetallic rhombohedral hexalayer graphene,” arXiv preprint arXiv:2508.15909, 2025.arXiv
  18. [18]Auerbach, N., Dutta, S., Uzan, M., Vituri, Y., Zhou, Y., Meltzer, A. Y., Grover, S., Holder, T., Emanuel, P., Huber, M. E., & al., . et . (2025). Isospin magnetic texture and intervalley exchange interaction in rhombohedral tetralayer graphene. Nature Physics, 21(11), 1765–1772.
  19. [19]Arp, T., Sheekey, O., Zhou, H., Tschirhart, C., Patterson, C. L., Yoo, H., Holleis, L., Redekop, G., Babikyan, C., Xie, T., & al., . et . (2024). Intervalley coherence and intrinsic spin–orbit coupling in rhombohedral trilayer graphene. Nature Physics, 20(9), 1413–1420.
  20. [20]Thonhauser, T., CeresoIi, D., Vanderbilt, D., & Resta, R. (2005). Orbital magnetization in periodic insulators. Physical Review Letters, 95(13), 137205.
  21. [21]D. Xiao, J. Shi, and Q. Niu, “Berry phase correction to electron density of states in solids,” Physical review letters, vol. 95, no. 13, p. 137204, 2005.pubmed.ncbi.nlm.nih.gov/16197171
  22. [22]J. Shi, G. Vignale, D. Xiao, and Q. Niu, “Quantum theory of orbital magnetization and its generalization to interacting systems,” Physical review letters, vol. 99, no. 19, p. 197202, 2007.pubmed.ncbi.nlm.nih.gov/18233109
  23. [23]J. Zhu and C. Huang, “Orbital magnetization and streda formula of interacting electrons in the mean-field approximation,”DOI
  24. [24]G. Vignale, J. Shi, D. Xiao, and Q. Niu, “Ward identities and orbital magnetization in current density functional theory,” arXiv preprint arXiv:2605.14336, 2026.DOI
  25. [25]N. Morales-Durán, N. Wei, J. Shi, and A. H. MacDonald, “Magic angles and fractional chern insulators in twisted homobilayer transition metal dichalcogenides,” Phys. Rev. Lett., vol. 132, p. 096602, Mar 2024.DOI
  26. [26]I. Sodemann Villadiego, “Halperin states of particles and holes in ideal time reversal invariant pairs of chern bands and the fractional quantum spin hall effect in moiré mote 2,” Physical Review B, vol. 110, no. 4, p. 045114, 2024.DOI
  27. [27]G. Xu, N. Wei, I. S. Villadiego, and C. Huang, “Localized excitons and landau-level mixing in time-reversal symmetric pairs of chern bands,” Physical Review B, vol. 113, no. 16, p. L161121, 2026.arxiv.org/abs/2509.18438
  28. [28]F. Haldane, “Berry curvature on the fermi surface: Anomalous hall effect<? format?> as a topological fermi-liquid property,” Physical review letters, vol. 93, no. 20, p. 206602, 2004.pubmed.ncbi.nlm.nih.gov/15600949
  29. [29]D. R. Hofstadter, “Energy levels and wave functions of bloch electrons in rational and irrational magnetic fields,” Physical review B, vol. 14, no. 6, p. 2239, 1976.DOI
  30. [30]G. Xu and C. Huang, “Two-body solution and instabilities along streda lines in moire flat bands,” arXiv preprint arXiv:2603.11007, 2026.arxiv.org/abs/2603.11007
  31. [31]Giuliani, G., & Vignale, G. (2008). Quantum theory of the electron liquid. In Cambridge university press.
  32. [32]P. Streda, “Theory of quantised hall conductivity in two dimensions,” Journal of Physics C: Solid State Physics, vol. 15, no. 22, pp. L717–L721, 1982.DOI

Paper details

vlnc:W_wsqq6utq
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