betaFeedback

WorksPhysics and Astronomy

Preponderant Orbital Polarization in Relativistic Magnetovortical Matter

Abstract

We establish thermodynamic stability and gauge invariance in the magnetovortical matter of Dirac fermions under the coexistent rotation and strong magnetic field. The corresponding partition function reveals that the orbital contribution to bulk thermodynamics preponderates over the conventional contribution from anomaly-related spin effects. This orbital preponderance macroscopically manifests itself in the sign inversion of the induced charge and current in the magnetovortical matter, and can be tested experimentally as the flip of the angular momentum polarization of magnetovortical matter when the magnetic field strength is increased.

Introduction

While the proper formulation of angular momentum was a cornerstone in quantum physics since the early twentieth century, its new facets continue to emerge in quantum many-body systems. In quantum optics, the angular momentum of light is crucial for understanding optical vortices [1]. In condensed matter physics, angular momentum underpins the quantum Hall effect [2], topological insulators [3], and the physics of Skyrmions [4]. Advancements in orbitronics have opened up novel streams for information processing in addition to spintronics [5]. The recent high-energy physics is inseparable from angular momentum dynamics of quarks, gluons, and hadrons, which poses challenging problems investigated with relativistic colliders [6, 7].

A typical characteristic of angular momentum is its polarization under an external magnetic field or vorticity. A lot of magnetic or rotational responses are empirically understood with the spin polarization. Two of the prominent examples are the Einstein-de Haas effect [8] and the Barnett effect [9]. Also, the energy shift by the spin polarization implies that the chiral condensate is stabilized in a magnetized system [10], and destabilized in a rotating system [11]. A similar argument can be applied to the Chandrasekhar-Clogston limit of a superconductor [12]. Furthermore, the spin-polarization argument underlies the chirality-induced physics of relativistic fermions, such as the chiral magnetic and vortical effects [13, 14, 15].

The following plausible argument is found for magnetovortical matter [16]. While the Zeeman coupling distinguishes positive and negative charges, the spin-rotation coupling does not. As a result, the interplay between a parallel magnetic field (B=Bz^\boldsymbol{B}=B\hat{\boldsymbol{z}}) and a vortical field (Ω=Ωz^\boldsymbol{\Omega}=\Omega\hat{\boldsymbol{z}}) provides an energy imbalance between positive and negative charges, and thermodynamically induces a nonvanishing net charge. Such an intuitive picture has indeed been implemented with the linear response theory; the charge density of chiral fermions under B\boldsymbol{B} and Ω\boldsymbol{\Omega}

is shown to be ρ=eBΩ/(4π2)\rho=eB\Omega/(4\pi^{2}) [17], which has also been reproduced with the chiral kinetic theory [18, 19]. The effective-theory approach with an axial gauge field revealed the connection to a quantum anomaly and an associated induced current j=eE×Ω/(4π2)\boldsymbol{j}=e\boldsymbol{E}\times\boldsymbol{\Omega}/(4\pi^{2}) [20]. Owing to their nondissipative nature, the corresponding transport coefficients should be essential ingredients of the constitutive relations in (chiral) magnetohydrodynamics [21]; see Ref. [22] for a review.

This Letter aims to challenge the conventional spin-dominant argument for the polarization effects as sketched in Figure 1 (a). In magnetovortical matter, the crucial piece is the significant contribution from the orbital angular momentum around the guiding center of each cyclotron orbit, as illustrated in Figure 1 (b). The polarization of the orbital angular momentum is opposite to that of the spin, as understood from Lenz’s law. In quantum theory, we show that this orbital polarization preponderates over the spin polarization in magnitude, and thus the direction of the total angular momentum is flipped. Such a sign inversion definitely affects thermodynamic quantities; for example, the charge density becomes

Figure 1. Spin and orbital polarizations coupled with an external Ω\boldsymbol{\Omega} under (a) a weak B\boldsymbol{B} and (b) a strong B\boldsymbol{B}. The orbital contribution in (a) is suppressed with widely spread cyclotron motion, while the orbital contribution surpasses the spin effects in (b).

ρ=−eBΩ4π2,(1)\rho=-\frac{eB\Omega}{4\pi^{2}}, \tag*{(1)}

with an opposite sign to the aforementioned result. In this Letter, we formulate the thermodynamics of the magnetovortical matter, and derive the charge density Eq. (1) and the associated spatial current with special attention to the overall sign. We reveal that reaching the correct answer requires careful implementation of thermodynamic stability and gauge invariance, which were not encompassed in the previous thermodynamic function [23]. These fundamental requirements turn out to be two sides of the same coin, and essential to resolving the unphysical divergence originating from large angular momenta [24, 25]. Besides, the thermodynamic function constructed in this way elucidates the relation among the family of the anomaly-related transport including Eq. (1).

Angular momentum in the Landau-level basis—In the quantum theory of charged particles, one of the fundamental concepts is the gauge invariance of kinematical variables. For instance, the physical momentum of a charged particle is not p=−i∇\boldsymbol{p}=-i\nabla but Π=p−eA\boldsymbol{\Pi}=\boldsymbol{p}-e\boldsymbol{A} with an external electromagnetic field A\boldsymbol{A} and a charge ee. A similar distinction is found in angular momenta. The gauge-invariant and thus physical quantity is not the canonical angular momentum, L=x×p\boldsymbol{L}=\boldsymbol{x}\times\boldsymbol{p}, but the kinetic angular momentum, Lkin=x×Π\boldsymbol{L}_{\mathrm{kin}}=\boldsymbol{x}\times\boldsymbol{\Pi}. The difference between them may well be interpreted as the electromagnetic contribution. Indeed, in the symmetric gauge, the canonical angular momentum is a Noether conserved charge for the rotational symmetry and is related to the kinetic one as L=Lkin+LEM\boldsymbol{L}=\boldsymbol{L}_{\mathrm{kin}}+\boldsymbol{L}_{\mathrm{EM}} with LEM=x×(eA)\boldsymbol{L}_{\mathrm{EM}}=\boldsymbol{x}\times(e\boldsymbol{A}) [26]. The gauge dependence in L\boldsymbol{L} and LEM\boldsymbol{L}_{\mathrm{EM}} cancels out.

Quantum mechanics under B=Bz^\boldsymbol{B}=B\hat{\boldsymbol{z}} is conveniently described by the Landau-level basis, that is, ∣n,m⟩=(a†)n(b†)m∣0,0⟩/n!m!|n,m\rangle=(a^\dagger)^n(b^\dagger)^m|0,0\rangle/\sqrt{n!m!}, where a=(Πx+isΠy)/2∣eB∣a=(\Pi_x+is\Pi_y)/\sqrt{2|eB|}, b=∣eB∣/2(X−isY)b=\sqrt{|eB|/2}(X-isY), and s=sgn⁡(eB)s=\operatorname{sgn}(eB). Here, the coordinate variables, X=x+Πy/(eB)X=x+\Pi_y/(eB) and Y=y−Πx/(eB)Y=y-\Pi_x/(eB), are the guiding center of the cyclotron motion, and they are constants of motion if further interactions are turned off. The longitudinal component of the kinetic angular momentum is represented as Lkin=xΠy−yΠx=Λ+ΔL_{\mathrm{kin}}=x\Pi_y-y\Pi_x=\Lambda+\Delta with

Λ=(x−X)Πy−(y−Y)Πx=−s(2a†a+1),Δ=XΠy−YΠx=is(a†b†−ab).(2)\begin{aligned} \Lambda&=(x-X)\Pi_y-(y-Y)\Pi_x=-s(2a^\dagger a+1),\\ \Delta&=X\Pi_y-Y\Pi_x=is(a^\dagger b^\dagger-ab). \tag*{(2)} \end{aligned}

The derivation is reviewed in the Supplemental Material [27]. The diagonal operator Λ\Lambda quantifies the local orbital motion around the guiding center, while the purely off-diagonal Δ\Delta describes the global orbital motion around the origin of the coordinate system. The form of Λ\Lambda can be understood as follows. For eB>0eB>0, the sign of the Λ\Lambda eigenvalue should be negative as the cyclotron motion is clockwise on the xyxy plane. The magnitude should be characterized by a†aa^\dagger a so that higher Landau levels have larger LkinL_{\mathrm{kin}}. Also, Λ\Lambda as well as the energy levels are independent of bb and b†b^\dagger, as a result of the translational symmetry that leads to the Landau degeneracy. Contrarily, Δ\Delta obviously breaks the translational symmetry, and thus does not play any role in characterizing conventional magnetized systems without rotation; see also Ref. [28], in which Λ\Lambda enters the kinetic theory and the hydrodynamics describing the guiding center motion, while Δ\Delta does not.

Gauge invariance in the partition function—The traditional approach to installing the rotational effect in thermodynamics is based on the maximum entropy principle. First, let us examine rigidly rotating matter (i.e., systems with rotational symmetry) in the absence of a gauge field. Then, the angular velocity enters a statistical ensemble as the Lagrange multiplier for the angular momentum conservation, in parallel to those of inverse temperature and chemical potential for the conservation of energy and charge, respectively. The partition function is thus given as [29]

Z=tr⁡exp⁡[−β(H−ΩJ)](no gauge field)(3)Z=\operatorname{tr}\exp\left[-\beta(\mathcal{H}-\Omega\mathcal{J})\right]\qquad\text{(no gauge field)} \tag*{(3)}

with the inverse temperature β=1/T\beta=1/T and the angular velocity Ω=Ωz^\boldsymbol{\Omega}=\Omega\hat{\boldsymbol{z}}. Here, H\mathcal{H} and J\mathcal{J} are the field operators of a Hamiltonian and a conserved Noether charge, i.e., the zz component of the angular momentum in our setup, respectively. Although the partition function Eq. (3) was originally formulated in a nonrelativistic form, it yields the correct form of the chiral vortical effect [30], which was later confirmed with the linear response theory under a metric perturbation [15].

However, the partition function Eq. (3) needs refinement for gauge theories. The subtlety stems from the difference between the canonical and kinetic momenta elaborated above. Whereas the conserved J\mathcal{J} is built with the canonical angular momentum, L=xpy−ypxL=xp_y-yp_x, it causes a problem that the partition function can become gauge dependent. Instead, the correct partition function should read

Z=tr⁡exp⁡[−β(H−ΩJkin)],(4)Z=\operatorname{tr}\exp\left[-\beta(\mathcal{H}-\Omega\mathcal{J}_{\mathrm{kin}})\right], \tag*{(4)}

where Jkin\mathcal{J}_{\mathrm{kin}} is the field operator built with the kinetic angular momentum, Lkin=xΠy−yΠxL_{\mathrm{kin}}=x\Pi_y-y\Pi_x. The partition function Eq. (4) is one particular form of the covariant density operator [31].

Let us now derive the gauge-invariant partition function Eq. (4). The derivation is based on the balance condition within electromagnetism [32]. Suppose that a particle with charge ee is rotating on the circle with a radius r=(x2+y2)1/2r=(x^2+y^2)^{1/2} and an angular velocity Ω=Ωz^\boldsymbol{\Omega}=\Omega\hat{\boldsymbol{z}}. This particle has a velocity v=rΩθ^\boldsymbol{v}=r\Omega\hat{\boldsymbol{\theta}} along the azimuthal angle θ\theta direction. If we apply a magnetic field B=Bz^\boldsymbol{B}=B\hat{\boldsymbol{z}}, the particle feels a radial drift force, Fdrift=ev×B=eBΩr\mathbf{F}_{\mathrm{drift}}=e\mathbf{v}\times\mathbf{B}=eB\Omega\mathbf{r}. Then, the circular motion is no longer closed since the particle flows outwardly (inwardly) when eB>0eB>0 (eB<0eB<0). The particle motion can reach a steady state only when an additional electric field offsets the radial drift force. Such an additional electric field is generated by an effective scalar potential eA0=ΩLEMeA_{0}=\Omega L_{\mathrm{EM}} involved in JkinJ_{\mathrm{kin}}. In the symmetric gauge A=(A1,A2,A3)=(−By/2,Bx/2,0)\mathbf{A}=(A^{1},A^{2},A^{3})=(-By/2,Bx/2,0), one finds that eE=−e∇A0=−e∇(xA2−yA1)=−Fdrifte\mathbf{E}=-e\nabla A_{0}=-e\nabla(xA^{2}-yA^{1})=-\mathbf{F}_{\mathrm{drift}}. In many-body physics, the stability of the circular motion has a close connection to equilibration of magnetovortical systems. That is, while a usual global equilibrium state allows for a global rotation and translation, the magnetovortical systems need to satisfy the additional balance condition along the radial direction (see Table I). The covariant form of the local equilibrium condition in Ref. [33] also results in the present condition between BB and EE in its special case [34].

Total HamiltonianFdriftF_{\mathrm{drift}}eEeEElectric forces
H−ΩJ\mathcal{H}-\Omega JeBΩreB\Omega r00Unbalanced
H−ΩJkin\mathcal{H}-\Omega J_{\mathrm{kin}}eBΩreB\Omega r−eBΩr-eB\Omega rBalanced

Table I. Difference of total Hamiltonians

Hereafter, we focus on relativistic magnetovortical matter, i.e., the rotating and magnetized relativistic medium composed of Dirac fermions. As shown in the Supplemental Material [27], the partition function is given by Eq. (4) with

H−ΩJkin=∫d3x ψˉ[−iγiDi+M−Ωγ0(Lkin+S)]ψ.(5)\mathcal{H}-\Omega J_{\mathrm{kin}}=\int d^{3}x\,\bar{\psi}\left[-i\gamma^{i}D_{i}+M-\Omega\gamma^{0}(L_{\mathrm{kin}}+S)\right]\psi. \tag*{(5)}

where we defined Di=∂i+ieAiD_{i}=\partial_{i}+ieA_{i}, a mass MM, and the spin operator S=iγ1γ2/2S=i\gamma^{1}\gamma^{2}/2 [35].

Orbital preponderance in the strong magnetic field limit

A standard manner to evaluate Eq. (4) with the Hamiltonian [Eq. (5)] is to diagonalize the Dirac operator. However, the Landau-level basis is not the eigenstate of the Dirac operator due to Δ\Delta. In a technical perspective, the difficulty of the present eigenvalue problem is more obvious in the corresponding differential equation of motion in the symmetric gauge; a potential term ∝r4=(x2+y2)2\propto r^{4}=(x^{2}+y^{2})^{2} arises and the anharmonic oscillator problem must be solved. Also the translational noninvariance due to Δ\Delta requires the careful treatment of the boundary effect in the cylindrical coordinate, and this inhomogeneity makes the analysis more complicated.

Nevertheless, in the lowest Landau level (LLL) approximation, we can obtain the analytic expression of Eq. (4) and illustrate the importance of the orbital motion. Inserting the complete set of the Ritus basis for the spin-degenerate Landau levels ∣n,m⟩R=P+∣n,m⟩+P−∣n−1,m⟩|n,m\rangle_{\mathrm{R}}=\mathcal{P}_{+}|n,m\rangle+\mathcal{P}_{-}|n-1,m\rangle with P±=12(1±2sS)\mathcal{P}_{\pm}=\frac{1}{2}(1\pm2sS) (see, e.g., Refs. [36, 37]), one can show that

ln⁡Z=tr⁡ln⁡(γ0∂τ−iγ3∂z+M−γ0ν),(6)\ln Z=\operatorname{tr}\ln(\gamma^{0}\partial_{\tau}-i\gamma^{3}\partial_{z}+M-\gamma^{0}\nu), \tag*{(6)}

where we used Eq. (2) and defined the rotation-induced effective chemical potential

ν=Ω⟨Lkin+S⟩LLL=−sΩ2(7)\nu=\Omega\langle L_{\mathrm{kin}}+S\rangle_{\mathrm{LLL}}=-\frac{s\Omega}{2} \tag*{(7)}

with ⟨O⟩LLL=⟨0,m∣O∣0,m⟩R\langle O\rangle_{\mathrm{LLL}}=\langle0,m|O|0,m\rangle_{\mathrm{R}}.

A crucial remark is that ν\nu involves not only the spin-rotation coupling (spin polarization) but also the orbital-rotation coupling (orbital polarization). This is why the conventional picture in Figure 1 (a) is insufficient in the strong magnetic field case. The negative sign in Eq. (7) is due to the orbital contribution, νorb=Ω⟨Lkin⟩LLL=−sΩ\nu_{\mathrm{orb}}=\Omega\langle L_{\mathrm{kin}}\rangle_{\mathrm{LLL}}=-s\Omega preponderating over the spin contribution, νspin=Ω⟨S⟩LLL=sΩ/2\nu_{\mathrm{spin}}=\Omega\langle S\rangle_{\mathrm{LLL}}=s\Omega/2. Such a sign inversion is a unique feature in the coupling to Ω\Omega. For instance, the LLL energy dispersion incorporates a similar combination, i.e., the magnetization coupling, but this vanishes for the LLL due to the gg factor: eB⟨Lkin+2S⟩LLL=0eB\langle L_{\mathrm{kin}}+2S\rangle_{\mathrm{LLL}}=0.

The evaluation of Eq. (6) is similar to that in the magnetized system at finite density. After performing the Matsubara summation, the pressure is obtained as

P=∣eB∣2π∫dpz2π[ϵ+T∑η=±ln⁡(1+e−β(ϵ−ην))],(8)P=\frac{|eB|}{2\pi}\int\frac{dp_{z}}{2\pi}\left[\epsilon+T\sum_{\eta=\pm}\ln\left(1+e^{-\beta(\epsilon-\eta\nu)}\right)\right], \tag*{(8)}

where ϵ=pz2+M2\epsilon=\sqrt{p_{z}^{2}+M^{2}}. The summation over mm in the Landau-level basis was taken as ∑m=∣eB∣R2/2\sum_{m}=|eB|R^{2}/2 with the radial system size RR, so that the usual Landau degeneracy is reproduced. Equation (8) is derived without considering the boundary effects, which have to be taken into account in usual rotating systems because of the causality constraint R≤∣Ω∣−1R\leq|\Omega|^{-1}. This is justified for bulk thermodynamics as long as we impose ∣eB∣≫R−1\sqrt{|eB|}\gg R^{-1}. The physical meaning of the inequality is that the Landau wave function is localized too tightly to perceive the existence of the boundary [24].

The pressure [Eq. (8)] implies a nontrivial property of the magnetovortical matter. To see this, let us specifically look into the zero temperature limit of Eq. (8):

P=∣eB∣Ω216π2[1−α−αln⁡1+1−αα]θ(1−α),(9)P=\frac{|eB|\Omega^{2}}{16\pi^{2}}\left[\sqrt{1-\alpha}-\alpha\ln\frac{1+\sqrt{1-\alpha}}{\sqrt{\alpha}}\right]\theta(1-\alpha), \tag*{(9)}

where α=(2M/Ω)2\alpha=(2M/\Omega)^{2} and θ(x)\theta(x) is the step function. For ∣Ω∣/2>M|\Omega|/2>M, hence, the rotational effect is visible even at the vacuum. This property emerges from the formation of the Landau levels that remain low-energy modes. Indeed, the rotational effect is unseen in a magnetic-free vacuum, due to the infrared energy gap demanded by the causality constraint [24].

Charge density and angular momenta—The negative sign in Eq. (7) turns out to be crucial in charge-dependent thermodynamic quantities. As a prominent example, let us compute the charge density in the massless limit M→0M \to0. It is now convenient to add a real chemical potential μ\mu through the shift ΩJkin→ΩJkin+μ∫d3x ψˉγ0ψ\Omega J_{\rm kin} \to\Omega J_{\rm kin}+\mu\int d^3x\,\bar{\psi}\gamma^0\psi in Eq. (5), and accordingly ν→ν+μ\nu\to\nu+\mu in Eq. (8). The pressure is then analytically evaluated as

P=∣eB∣4π2[(μ−sΩ2)2+π2T23].(10)P=\frac{|eB|}{4\pi^2}\left[\left(\mu-\frac{s\Omega}{2}\right)^2+\frac{\pi^2T^2}{3}\right]. \tag*{(10)}

The charge density is calculated as ∂P/∂μ=∣eB∣μ/(2π2)+ρ\partial P/\partial\mu=|eB|\mu/(2\pi^2)+\rho with the magnetovortical contribution ρ\rho in Eq. (1). The negative sign in ρ\rho originates in the same way as the effective chemical potential [Eq. (7)]. Namely, the charge density ρ\rho is given as the sum of the spin and orbital contributions:

ρspin=eBΩ4π2,ρorb=−eBΩ2π2.(11)\rho_{\rm spin}=\frac{eB\Omega}{4\pi^2},\qquad\rho_{\rm orb}=-\frac{eB\Omega}{2\pi^2}. \tag*{(11)}

We make comparisons with related studies. First, in Refs. [24, 25], we perform a similar analysis with the partition function [Eq. (3)], instead of Eq. (4). Then, ρspin\rho_{\rm spin} was correctly found, but the orbital contribution corresponding to ρorb\rho_{\rm orb} was divergent in the large RR limit, as ≃eBΩN/(4π2)\simeq eB\Omega N/(4\pi^2) with the magnetic flux N=eBR2/2≫1N=eBR^2/2\gg1 [38]. This unphysical divergence is identified as the thermodynamic instability because of the nonvanishing drift force in the partition function [Eq. (3)]. Second, the Kubo formula based on the diagrammatic approach led to the answer that looked like only ρspin\rho_{\rm spin} [17]. The sign of ρ\rho was skewed due to the lack of clear recognition of the orbital-rotation coupling. As we discuss in the Supplemental Material [27], the careful analysis of the linear response theory agrees with Eq. (1) [39].

Besides, the chiral kinetic theory (CKT) up to O(ℏ2)O(\hbar^2) also reproduces only the spin part ρspin\rho_{\rm spin} [18]. This, however, does not contradict our analysis in the LLL approximation. In the usual CKT, background electromagnetic fields are assumed to be small compared with other scales. Since a weak magnetic field cannot yield the Landau quantization [40], only the spin-rotation coupling contributes to the induced charge density [see Figure 1 (a)]. On the other hand, the CKT constructed with the Landau-level basis [41] should pick up the orbital contribution ρorb\rho_{\rm orb}. Our result appears to be consistent with the “total” charge density obtained from the “vortical solution” in Ref. [19].

The above observations elucidate the distinct origins of ρspin\rho_{\rm spin} and ρorb\rho_{\rm orb}. The former origin persists irrespective of the magnitude of the magnetic field. The latter emerges only under a strong magnetic field. Hence, it is deduced that only ρspin\rho_{\rm spin} is related to a quantum anomaly. We confirm this identification with the thermodynamic potential. From the pressure [Eq. (10)], the total angular momentum is derived as ∂P/∂Ω=Jspin+Jorb+∣eB∣Ω/(8π2)\partial P/\partial\Omega=J_{\rm spin}+J_{\rm orb}+|eB|\Omega/(8\pi^2), where the last term is the contribution from the effective chemical potential ν=−sΩ/2\nu=-s\Omega/2 and the others are from the real one μ\mu:

Jspin=eBμ4π2,Jorb=−eBμ2π2.(12)J_{\rm spin}=\frac{eB\mu}{4\pi^2},\qquad J_{\rm orb}=-\frac{eB\mu}{2\pi^2}. \tag*{(12)}

The Maxwell relation for ∂2P/∂Ω∂μ\partial^2P/\partial\Omega\partial\mu shows that ρspin\rho_{\rm spin} and JspinJ_{\rm spin} share the same origin of the coefficient eB/4π2eB/4\pi^2; see also a similar correspondence in Ref. [42]. We recall that the fermionic spin is also connected with the axial current as Jspin=jz5/2J_{\rm spin}=j^5_z/2. Therefore, both ρspin\rho_{\rm spin} and JspinJ_{\rm spin} are anomaly related, and so is the chiral separation current jz5=eBμ/(2π2)j^5_z=eB\mu/(2\pi^2). We note that they receive the mass correction, which is again the same as the chiral separation effect.

Finally, we discuss an associated transport phenomenon in magnetovortical matter. When the induced charge density ρ\rho is rotating with the velocity vθ=rΩv_\theta=r\Omega, the azimuthal current density is given as jθ=ρvθ=−eBΩ2r/(4π2)j_\theta=\rho v_\theta=-eB\Omega^2r/(4\pi^2). The stability condition for electromagnetic fields, E=−BΩr\boldsymbol{E}=-B\Omega r, leads to the vector form j=−eE×Ω/(4π2)\boldsymbol{j}=-e\boldsymbol{E}\times\boldsymbol{\Omega}/(4\pi^2), which agrees with the spatial component of the Chern-Simons current [20], but with the opposite sign for the same reason as ρ\rho. This explains the physical picture behind the Chern-Simons current derived from the discussion about quantum anomaly.

Outlook—Our work sheds light on the orbital angular momentum carried by magnetized Dirac fermions. Lattice QCD simulation is a first-principles method to extend our analysis to the strongly coupled magnetovortical matter with dynamical gluons. A benchmark quantity is magnetization, which has been simulated without rotation [43, 44]. In the magnetovortical matter, magnetization is obtained from our pressure [Eq. (10)] as

∂P∂(eB)=s4π2(μ2−sμΩ+Ω24+π2T23).(13)\frac{\partial P}{\partial(eB)}=\frac{s}{4\pi^2}\left(\mu^2-s\mu\Omega+\frac{\Omega^2}{4}+\frac{\pi^2T^2}{3}\right). \tag*{(13)}

The sign of the term ∝μΩ\propto\mu\Omega provides a crucial test. Our analysis is also inspiring enough to revisit the phase structures of magnetovortical matter [23] [45] with the correct thermodynamic formulation in lattice QCD. In general, numerical simulations for rotating systems depend on details of boundary conditions implemented for causality [46]. Investigating the magnetovortical matter is technically advantageous in this respect because the wave functions in strong magnetic fields are tightly localized within a cyclotron radius.

Our finding in Eq. (12) at Ω=0\Omega=0 implies that the strongly magnetized matter plays a role of the storage of the negative angular momentum J=Jspin+Jorb<0J=J_{\rm spin}+J_{\rm orb}<0 for eBμ>0eB\mu>0. If we adiabatically turn off BB, a stored JJ is distributed in a rigid body so that the total angular momentum is conserved. This evokes the Einstein-de Haas (EdH) effect, but the rotation occurs in an opposite direction to the conventional one, which we call the anti-EdH effect. This novel phenomenon can be tested in parallel to the conventional one by using Weyl or Dirac semimetals, which are good environments for the direct applicability of the present relativistic analysis, the avoidability of the mass correction, and the controllability of the chemical potential [47]. The maximum induced angular momentum is evaluated from Ωind=∣J∣/I\Omega_{\mathrm{ind}} = |J|/I with II being the density of moment of inertia for a cylindrical material. The estimated magnitude is as large as

Ωind=(μ10−1 eV)(B102 T)(ρmass1 g/cm3)(R10−2 cm)2×7.76×10−5 rad/s,(14)\Omega_{\mathrm{ind}} = \frac{\left(\frac{\mu}{10^{-1}\,\mathrm{eV}}\right)\left(\frac{B}{10^{2}\,\mathrm{T}}\right)}{\left(\frac{\rho_{\mathrm{mass}}}{1\,\mathrm{g}/\mathrm{cm}^{3}}\right)\left(\frac{R}{10^{-2}\,\mathrm{cm}}\right)^{2}} \times7.76 \times10^{-5}\,\mathrm{rad/s}, \tag*{(14)}

where ρmass\rho_{\mathrm{mass}} is the mass density and the Fermi velocity is introduced as vF=106 m/sv_{F}=10^{6}\,\mathrm{m/s}. This Ωind\Omega_{\mathrm{ind}} is smaller by only 1 order of magnitude than the conventional EdH, and sufficiently testable; for example, a realistic set of parameters μ=10−1 eV\mu=10^{-1}\,\mathrm{eV}, B=102 TB=10^{2}\,\mathrm{T}, and R=10−2 cmR=10^{-2}\,\mathrm{cm} leads to μ/2ℏvF2eB=0.39\mu/\sqrt{2\hbar v_{F}^{2}eB}=0.39 and eBR2/2ℏ=1.52×109eBR^{2}/2\hbar=1.52\times10^{9}, which justify the LLL approximation with the large degeneracy, and thus the estimate based on Eq. (12). Further studies open up a new avenue to the orbitronics and related phenomena of Dirac electrons.

Acknowledgments

The authors thank Takuya Shimazaki for his contribution to the early stage of this work, and Matteo Buzzegoli, Hao-Lei Chen, Maxim Chernodub, Masaru Hongo, Xu-Guang Huang, Shu Lin, Xin-Li Sheng, Igor Shovkovy, Dam Thanh Son, Qun Wang, Naoki Yamamoto, Di-Lun Yang, and Yi Yin for valuable comments and discussions. The authors also appreciate the hospitality of ECT* during the workshop “Spin and Quantum Features of QCD Plasma,” where this work was discussed and finalized. This work is supported by the Japan Society for the Promotion of Science KAKENHI Grant No. 20K03948, No. 22H01216, No. 22H05118, and No. 24K17052.

References

— Supplemental Material —

Landau quantization

This section is devoted to a review of the nonrelativistic Landau problem (see also Ref. [37]); the following argument is also applied to the energy dispersion of Dirac fermions. The one-particle Hamiltonian under a magnetic field B=Bz^\mathbf{B}=B\hat{\mathbf{z}} is given by

H=12M(Πx2+Πy2),Πi=pi−eAi(x,y),(15)H=\frac{1}{2M}\left(\Pi_x^2+\Pi_y^2\right),\qquad\Pi_i=p_i-eA_i(x,y), \tag*{(15)}

where AiA_i is the gauge field such that ∂xAy−∂yAx=B\partial_x A_y-\partial_y A_x=B. Here Πi\Pi_i is called the kinetic (linear) momentum; ‘kinetic’ means this is the gauge-invariant and thus physical momentum. Another fundamental quantity to characterize the cyclotron motion is (X,Y)(X,Y) defined as

X=x+ΠyeB,Y=y−ΠxeB.(16)X=x+\frac{\Pi_y}{eB},\qquad Y=y-\frac{\Pi_x}{eB}. \tag*{(16)}

From a solution of the classical equation of motion, p˙i=−∂H/∂xi\dot{p}_i=-\partial H/\partial x_i and x˙i=∂H/∂pi\dot{x}_i=\partial H/\partial p_i, one finds that the set of (X,Y)(X,Y) is the center coordinate of the cyclotron motion, which is referred to as the guiding center coordinate.

An important advantage of introducing the guiding center (X,Y)(X,Y) is that with the help of the canonical commutation relation [xi,pj]=iδij[x_i,p_j]=i\delta_{ij}, one can construct the following algebra:

[Πx,Πy]=ieB,[X,Y]=−ieB,[X,Πi]=[Y,Πi]=0.(17)[\Pi_x,\Pi_y]=ieB,\qquad[X,Y]=-\frac{i}{eB},\qquad[X,\Pi_i]=[Y,\Pi_i]=0. \tag*{(17)}

This enables us to introduce the following two types of the ladder operators:

a:=12∣eB∣(Πx+isΠy),b:=∣eB∣2(X−isY),s:=sgn⁡(eB),(18)a:=\frac{1}{\sqrt{2|eB|}}(\Pi_x+is\Pi_y),\qquad b:=\sqrt{\frac{|eB|}{2}}(X-isY),\qquad s:=\operatorname{sgn}(eB), \tag*{(18)}

which satisfies [a,a†]=[b,b†]=1[a,a^\dagger]=[b,b^\dagger]=1 and [a,b]=0[a,b]=0. We inversely solve Eq. (18) in terms of Πi\Pi_i, XX and YY as

Πx=∣eB∣2(a+a†),Πy=−is∣eB∣2(a−a†),X=12∣eB∣(b+b†),Y=is2∣eB∣(b−b†).(19)\Pi_x=\sqrt{\frac{|eB|}{2}}(a+a^\dagger),\qquad\Pi_y=-is\sqrt{\frac{|eB|}{2}}(a-a^\dagger),\qquad X=\frac{1}{\sqrt{2|eB|}}(b+b^\dagger),\qquad Y=\frac{is}{\sqrt{2|eB|}}(b-b^\dagger). \tag*{(19)}

The Landau-level basis is defined as the eigenstate for a†aa^\dagger a and b†bb^\dagger b:

∣n,m⟩:=(a†)n(b†)mn!m!∣0,0⟩,(20)|n,m\rangle:=\frac{(a^\dagger)^n(b^\dagger)^m}{\sqrt{n!m!}}|0,0\rangle, \tag*{(20)}

where the ground state is normalized as ⟨0,0∣0,0⟩=1\langle0,0|0,0\rangle=1.

Let us explain the physical meaning of the quantum numbers nn and mm. It is straightforward to show

Πx2+Πy2=∣eB∣(2a†a+1).(21)\Pi_x^2+\Pi_y^2=|eB|(2a^\dagger a+1). \tag*{(21)}

This representation implies that nn corresponds to the transverse momentum on the xyxy-plane, or equivalently, to the quantum number to specify the Landau level, as is clear from

H∣n,m⟩=∣eB∣M(n+12)∣n,m⟩.(22)H|n,m\rangle=\frac{|eB|}{M}\left(n+\frac{1}{2}\right)|n,m\rangle. \tag*{(22)}

An important observation here is that the Landau level is degenerate in terms of mm. The reason is understandable if we notice the following relation:

X2+Y2=1∣eB∣(2b†b+1).(23)X^2+Y^2=\frac{1}{|eB|}(2b^\dagger b+1). \tag*{(23)}

This indicates that mm characterizes the guiding center position, which is obviously not invariant under the translation on the xyxy-plane. The degeneracy of mm hence reflects the translational invariance of the Landau problem. More precisely, each state ∣n,m⟩\lvert n,m\rangle corresponds to the cyclotron motion where the magnitude of the momentum is ∣eB∣(2n+1)\sqrt{\lvert eB\rvert(2n+1)} and the center position is located at the distance (2m+1)/∣eB∣\sqrt{(2m+1)/\lvert eB\rvert} from the origin. Using the above ladder operators, we represent the kinetic angular momentum as

Lzkin:=xΠy−yΠx=Λ+Δ,(24)L_z^{\mathrm{kin}} := x\Pi_y-y\Pi_x=\Lambda+\Delta, \tag*{(24)}

where Λ\Lambda and Δ\Delta are the diagonal and off-diagonal operator defined as

Λ:=(x−X)Πy−(y−Y)Πx=−s(2a†a+1),(25)\Lambda:= (x-X)\Pi_y-(y-Y)\Pi_x=-s(2a^\dagger a+1), \tag*{(25)}
Δ:=XΠy−YΠx=−is(ab−a†b†).(26)\Delta:= X\Pi_y-Y\Pi_x=-\mathrm{i}s(ab-a^\dagger b^\dagger). \tag*{(26)}

The operator Λ\Lambda corresponds to the usual cyclotron motion, i.e., the local circular motion around the guiding center. The overall sign of Λ\Lambda is met to the classical Lenz’s law, and it is plausible that the magnitude is proportional to the transverse momentum. Another piece Δ\Delta is regarded as the angular momentum due to the global circular motion around the origin. For this reason, Δ\Delta is nothing to do with the usual Landau problem respecting the translational symmetry, while it generally contributes under rotation, which washes out the translation invariance. We emphasize that in the lowest Landau level approximation, the off-diagonal operator Δ\Delta does not enter the thermodynamics of magnetovortical matter. This is physically because the strong magnetic field limit effectively recovers the translation invariance even under rotation.

Field operator of angular momentum

We present the brief derivation of the gauge-invariant angular momentum operator composed of Dirac fields (see also Ref. [48] for more detailed discussion), providing Eq. (5) in the main text. We start from the following Lagrangian:

L=ψˉ(iγμDμ−m)ψ−14FμνFμν.(27)\mathcal{L}=\bar{\psi}(\mathrm{i}\gamma^\mu D_\mu-m)\psi-\frac{1}{4}F_{\mu\nu}F^{\mu\nu}. \tag*{(27)}

with Dμψ=(∂μ+ieAμ)ψD_\mu\psi=(\partial_\mu+\mathrm{i}eA_\mu)\psi and Fμν=∂μAν−∂νAμF_{\mu\nu}=\partial_\mu A_\nu-\partial_\nu A_\mu. Here and hereafter the raising and lowering of indices are performed using the Minkowski metric ημν=diag⁡(1,−1,−1,−1)\eta_{\mu\nu}=\operatorname{diag}(1,-1,-1,-1). The canonical energy-momentum tensor is computed as

Θμν=∂L∂(∂μψ)∂νψ+∂νψˉ∂L∂(∂μψˉ)+∂L∂(∂μAρ)∂νAρ−ημνL=ψˉiγμ∂νψ−Fμρ∂νAρ−ημνL.(28)\begin{aligned} \Theta^{\mu\nu}=\frac{\partial\mathcal{L}}{\partial(\partial_\mu\psi)}\partial^\nu\psi+\partial^\nu\bar{\psi}\frac{\partial\mathcal{L}}{\partial(\partial_\mu\bar{\psi})}+\frac{\partial\mathcal{L}}{\partial(\partial_\mu A_\rho)}\partial^\nu A_\rho-\eta^{\mu\nu}\mathcal{L} \\ &=\bar{\psi}\mathrm{i}\gamma^\mu\partial^\nu\psi-F^{\mu\rho}\partial^\nu A_\rho-\eta^{\mu\nu}\mathcal{L}. \tag*{(28)} \end{aligned}

The above energy-momentum tensor is not gauge-invariant obviously, but the situation can be cured through the Belinfante prescription; the energy momentum tensor can be shifted by ∂ρBρμν\partial_\rho B^{\rho\mu\nu} such that Bμρν=−BρμνB^{\mu\rho\nu}=-B^{\rho\mu\nu} because the conservation law still holds as ∂μ(Θμν+∂ρBρμν)=0\partial_\mu(\Theta^{\mu\nu}+\partial_\rho B^{\rho\mu\nu})=0. One of such Belinfante tensors above is

Bρμν=12(Msρμν−Msμρν+Msνμρ),Msμρσ:=ψˉγμΣρσψ+FμρAσ−FμσAρ,(29)B^{\rho\mu\nu}=\frac{1}{2}\left(M_s^{\rho\mu\nu}-M_s^{\mu\rho\nu}+M_s^{\nu\mu\rho}\right),\qquad M_s^{\mu\rho\sigma}:=\bar{\psi}\gamma^\mu\Sigma^{\rho\sigma}\psi+F^{\mu\rho}A^\sigma-F^{\mu\sigma}A^\rho, \tag*{(29)}

where Σμν=(i/4)[γμ,γν]\Sigma^{\mu\nu}=(\mathrm{i}/4)[\gamma^\mu,\gamma^\nu]. This MsμρσM_s^{\mu\rho\sigma} is a spin-like part of the total angular momentum tensor, as we will show below. Thanks to the identity γμγνγρ=ημνγρ+ηνργμ−ημργν−iϵμνρσγ5γσ\gamma^\mu\gamma^\nu\gamma^\rho=\eta^{\mu\nu}\gamma^\rho+\eta^{\nu\rho}\gamma^\mu-\eta^{\mu\rho}\gamma^\nu-\mathrm{i}\epsilon^{\mu\nu\rho\sigma}\gamma_5\gamma_\sigma with γ5=iγ0γ1γ2γ3\gamma_5=\mathrm{i}\gamma^0\gamma^1\gamma^2\gamma^3 and ϵ0123=1\epsilon^{0123}=1, the Belinfante term is written as

Bρμν=i2ψˉ(ηνργμ−ημργν−i2ϵμνρσγ5γσ)ψ−FρμAν.(30)B^{\rho\mu\nu}=\frac{\mathrm{i}}{2}\bar{\psi}\left(\eta^{\nu\rho}\gamma^\mu-\eta^{\mu\rho}\gamma^\nu-\frac{\mathrm{i}}{2}\epsilon^{\mu\nu\rho\sigma}\gamma_5\gamma_\sigma\right)\psi-F^{\rho\mu}A^\nu. \tag*{(30)}

Then, the improved energy-momentum tensor is

Tμν=Θμν+∂ρBρμν=ψˉiγμD↔νψ+14ϵμνρσ∂ρ(ψˉγ5γσψ)−FμρFνρ+14ημνFρσFρσ.(31)\begin{aligned} T^{\mu\nu}=\Theta^{\mu\nu}+\partial_\rho B^{\rho\mu\nu} \\ &=\bar{\psi}\mathrm{i}\gamma^\mu\overleftrightarrow{D}^{\nu}\psi+\frac{1}{4}\epsilon^{\mu\nu\rho\sigma}\partial_\rho(\bar{\psi}\gamma_5\gamma_\sigma\psi)-F^{\mu\rho}F^\nu{}_{\rho}+\frac{1}{4}\eta^{\mu\nu}F_{\rho\sigma}F^{\rho\sigma}. \tag*{(31)} \end{aligned}

where D↔μ:=(Dμ−D←μ)/2\overleftrightarrow{D}_\mu:=(D_\mu-\overleftarrow{D}_\mu)/2 with ψˉD←μ:=ψˉ(∂←μ−ieAμ)\bar{\psi}\overleftarrow{D}_\mu:=\bar{\psi}(\overleftarrow{\partial}_\mu-\mathrm{i}eA_\mu). In deriving the last line, we utilize the U(1) current conservation ∂μ(ψˉγμψ)=0\partial_\mu(\bar{\psi}\gamma^\mu\psi)=0 and the equations of motions (iγμDμ−m)ψ=0(\mathrm{i}\gamma^\mu D_\mu-m)\psi=0 and ∂μFμν=eψˉγνψ\partial_\mu F^{\mu\nu}=e\bar{\psi}\gamma^\nu\psi.

Similarly, the canonical angular momentum tensor is derived as

Mλμν=xμΘλν−xνΘλμ−i[∂L∂(∂λψ)Σμνψ+ψˉΣμν∂L∂(∂λψˉ)+∂L∂(∂λAρ)ΞμνρσAσ]=xμΘλν−xνΘλμ+Msλμν.(32)\begin{aligned} M^{\lambda\mu\nu} &= x^\mu\Theta^{\lambda\nu}-x^\nu\Theta^{\lambda\mu}-i\left[\frac{\partial\mathcal{L}}{\partial(\partial_\lambda\psi)}\Sigma^{\mu\nu}\psi+\bar{\psi}\Sigma^{\mu\nu}\frac{\partial\mathcal{L}}{\partial(\partial_\lambda\bar{\psi})}+\frac{\partial\mathcal{L}}{\partial(\partial_\lambda A_\rho)}\Xi^{\mu\nu\rho\sigma}A_\sigma\right] \\ &=x^\mu\Theta^{\lambda\nu}-x^\nu\Theta^{\lambda\mu}+M_{\mathrm{s}}^{\lambda\mu\nu}. \tag*{(32)} \end{aligned}

with Ξμναβ=ημαηνβ−ημβηνα\Xi^{\mu\nu\alpha\beta}=\eta^{\mu\alpha}\eta^{\nu\beta}-\eta^{\mu\beta}\eta^{\nu\alpha}. Again the above form is not gauge-invariant, but can be improved by the Belinfante tensor; the angular momentum can be shifted by ∂ρ(xμBρλν−xνBρλμ)\partial_\rho(x^\mu B^{\rho\lambda\nu}-x^\nu B^{\rho\lambda\mu}) because Bλρν=−BρλνB^{\lambda\rho\nu}=-B^{\rho\lambda\nu} ensures the conservation law ∂λ[Mλμν+∂ρ(xμBρλν−xνBρλμ)]=0\partial_\lambda[M^{\lambda\mu\nu}+\partial_\rho(x^\mu B^{\rho\lambda\nu}-x^\nu B^{\rho\lambda\mu})]=0. From Bμλν−Bνλμ=−MsλμνB^{\mu\lambda\nu}-B^{\nu\lambda\mu}=-M_{\mathrm{s}}^{\lambda\mu\nu}, we find

Jλμν:=Mλμν+∂ρ(xμBρλν−xνBρλμ)=xμTλν−xνTλμ.(33)\begin{aligned} J^{\lambda\mu\nu} &:=M^{\lambda\mu\nu}+\partial_\rho(x^\mu B^{\rho\lambda\nu}-x^\nu B^{\rho\lambda\mu}) \\ &=x^\mu T^{\lambda\nu}-x^\nu T^{\lambda\mu}. \tag*{(33)} \end{aligned}

The conserved energy density and angular momentum density along the zz-axis are T00T^0_0 and J0xyJ^{0xy}, respectively. Keeping only the fermionic contributions, we arrive at the following expressions of the Hamiltonian and the angular momentum:

H:=∫d3x T00=∫d3x ψˉ(−iγiDi−m)ψ,\mathcal{H}:=\int d^3x\,T^0_0=\int d^3x\,\bar{\psi}(-i\gamma^iD_i-m)\psi,
Jkin:=∫d3x J0xy=∫d3x ψˉγ0[−i(xDy−yDx)+i2γ1γ2]ψ.(34)J_{\mathrm{kin}}:=\int d^3x\,J^{0xy}=\int d^3x\,\bar{\psi}\gamma^0\left[-i(xD_y-yD_x)+\frac{i}{2}\gamma^1\gamma^2\right]\psi. \tag*{(34)}

where we utilize the Dirac equation and perform the integration by parts. These gauge-invariant operators correspond to those in Eq. (5) in the main text.

Linear response theory

Correspondence between the flow and metric perturbations

We revisit the linear response theory for the charge density induced in the magnetovortical matter, and clarify the correspondence between our result in the main text and that in Ref. [17]. Within the linear-response regime, the current induced by the Hamiltonian density HextH_{\mathrm{ext}} is given as [49]

δjμ(t,x)=−i∫−∞tdt′∫d3x′⟨[jμ(t,x),Hext(t′,x′)]⟩,(35)\delta j^\mu(t,\boldsymbol{x})=-i\int_{-\infty}^{t}dt'\int d^3x'\langle[j^\mu(t,\boldsymbol{x}),H_{\mathrm{ext}}(t',\boldsymbol{x}')]\rangle, \tag*{(35)}

where the angle brackets represent the thermal average. The perturbation Hamiltonian density is identified as follows. We first check the equivalence between a flow perturbation and a metric perturbation used in Ref. [17]. The Hamiltonian for the rest fluid u˙μ=(1,0)\dot{u}^{\mu}=(1,0) is given as

H˙=T0μu˙μ=T00.(36)\dot{H}=T^{0\mu}\dot{u}_{\mu}=T^{00}. \tag*{(36)}

Let us consider a nontrivial background fluid velocity uμ=(1,v)u^\mu=(1,\boldsymbol{v}), i.e., uμ=(1,−v)u_\mu=(1,-\boldsymbol{v}), where we ignore the terms of O(∣v∣2)O(|\boldsymbol{v}|^2). Then, the Hamiltonian at the comoving frame reads

H=T0μuμ=H˙−T0ivi.(37)H=T^{0\mu}u_\mu=\dot{H}-T^{0i}v^i. \tag*{(37)}

In particular, let us consider a flow v=Ω×x\boldsymbol{v}=\boldsymbol{\Omega}\times\boldsymbol{x} with a constant Ω\boldsymbol{\Omega} that realizes the rotating system with the angular velocity 12∇×v=Ω\frac{1}{2}\nabla\times\boldsymbol{v}=\boldsymbol{\Omega}. Then, the above Hamiltonian becomes

H=H˙−Ω⋅(x×P),(38)H=\dot{H}-\boldsymbol{\Omega}\cdot(\boldsymbol{x}\times\boldsymbol{P}), \tag*{(38)}

where Pi=(P)i=T0iP^i=(P)^i=T^{0i}. This Hamiltonian is mimicked by a background geometry. That is, we consider a static curved spacetime quantified by the metric tensor gμν=ημν+hμνg_{\mu\nu}=\eta_{\mu\nu}+h_{\mu\nu}, where the Minkowskian metric is given as ημν=diag⁡(1,−1,−1,−1)\eta_{\mu\nu}=\operatorname{diag}(1,-1,-1,-1). We suppose that the metric perturbation hμνh_{\mu\nu} has nonzero components only in h0ih_{0i}. The Hamiltonian for the rest fluid in this geometry is specified by uμ=(1,0)u^\mu=(1,0), i.e., uμ=u˙μ+h0μu_\mu=\dot{u}_\mu+h_{0\mu} and reads

H=T0μuμ=H˙+T0ih0i.(39)H=T^{0\mu}u_\mu=\dot{H}+T^{0i}h_{0i}. \tag*{(39)}

According to Eqs. (37) and (39), the perturbation Hamiltonian is now identified as

Hext=−T0ivi=T0ih0i.(40)H_{\mathrm{ext}}=-T^{0i}v^{i}=T^{0i}h_{0i}. \tag*{(40)}

This implies the correspondence between the perturbations

vi=−h0i,(41)v^{i}=-h_{0i}, \tag*{(41)}

and thus that

Ωi=−12ϵijk∂jh0k,(42)\Omega^{i}=-\frac{1}{2}\epsilon^{ijk}\partial_{j}h_{0k}, \tag*{(42)}

where the Levi-Civita symbol is normalized by ϵ123=−ϵ123=1\epsilon^{123}=-\epsilon_{123}=1. The same expression of the vorticity is shown in Ref. [50] and required to reproduce the chiral vortical effect in Ref. [51] with the correct sign. However, above Eq. (8) of Ref. [17], the relation between the perturbations has an opposite sign.

Linear response under a metric perturbation

Let us analyze the induced charge under the metric perturbation. In the following, all indices can be raised and lowered by ημν\eta_{\mu\nu}, granted that we consider the linear response to the metric perturbation. The linear response theory (S21) tells that, under a weak metric perturbation (S25), the induced current is expressed with the retarded correlator as

δjμ(q)=12GRμνρ(q)hνρ(q),GRμνρ(x−x′)=−iθ(t−t′)⟨[jμ(x),Tνρ(x′)]⟩,(43)\delta j^{\mu}(q)=\frac{1}{2}G_{\mathrm{R}}^{\mu\nu\rho}(q)h_{\nu\rho}(q),\qquad G_{\mathrm{R}}^{\mu\nu\rho}(x-x')=-i\theta(t-t')\langle[j^{\mu}(x),T^{\nu\rho}(x')]\rangle, \tag*{(43)}

where we defined qμ=(ω,q)q^{\mu}=(\omega,\boldsymbol{q}) and the Fourier transform

f(x)=∫d4q(2π)4e−iq⋅xf(q).(44)f(x)=\int\frac{\mathrm{d}^{4}q}{(2\pi)^{4}}e^{-iq\cdot x}f(q). \tag*{(44)}

The factor of 1/21/2 is attached in front of the correlator, in order to cancel a doubled contribution from the symmetrized contraction. The microscopic definitions of the current and the energy-momentum tensor are given by

jμ(x)=ψˉ(x)γμψ(x),T0i(x)=i2ψˉ(x)(γ0Di+γiD0)ψ(x).(45)j^{\mu}(x)=\bar{\psi}(x)\gamma^{\mu}\psi(x),\qquad T^{0i}(x)=\frac{i}{2}\bar{\psi}(x)(\gamma^{0}D^{i}+\gamma^{i}D^{0})\psi(x). \tag*{(45)}

The covariant derivative Dμ=∂μ+ieAμD_{\mu}=\partial_{\mu}+ieA_{\mu} contains AμA_{\mu} for a constant magnetic field. Based on the analytic continuation, we relate the retarded correlator to the imaginary-time correlator as

GEμνρ(q0→ω+iδ,q)=GRμνρ(q),GEμνρ(x−x′)=−⟨Tτjμ(x)Tνρ(x′)⟩,(46)G_{\mathrm{E}}^{\mu\nu\rho}(q_{0}\to\omega+i\delta,\boldsymbol{q})=G_{\mathrm{R}}^{\mu\nu\rho}(q),\qquad G_{\mathrm{E}}^{\mu\nu\rho}(x-x')=-\langle T_{\tau}j^{\mu}(x)T^{\nu\rho}(x')\rangle, \tag*{(46)}

where δ\delta is a positive infinitesimal number for the analytic continuation and TτT_{\tau} is the imaginary-time ordering. The momentum in the imaginary-time correlator is qμ=(q0,q)q^{\mu}=(q_{0},\boldsymbol{q}) with the bosonic Matsubara frequency q0=2iπTmq_{0}=2i\pi Tm (mm is an integer). Focusing on the charge density, we have

δj0=h0k(q)lim⁡ω→0GE00k(q0→ω+iδ,q).(47)\delta j^{0}=h_{0k}(q)\lim_{\omega\to0}G_{\mathrm{E}}^{00k}(q_{0}\to\omega+i\delta,\boldsymbol{q}). \tag*{(47)}

The limit of ω→0\omega\to0 is required in order to obtain a static transport coefficient. We compute the above correlator in the lowest Landau level (LLL) limit at the one-loop order. One should start with the coordinate-space representation

GE00k(x−x′)=(−1)⋅(−1)i2tr⁡[P+γ0SLLL(x,x′)(γ0Dx′k+γkDx′0)SLLL(x′,x)]=i2tr⁡[P+γ0SLLL(x,x′)γ0Dx′kSLLL(x′,x)].(48)\begin{aligned} G_{\mathrm{E}}^{00k}(x-x')&=(-1)\cdot(-1)\frac{i}{2}\operatorname{tr}\left[P_{+}\gamma^{0}S_{\mathrm{LLL}}(x,x')(\gamma^{0}D_{x'}^{k}+\gamma^{k}D_{x'}^{0})S_{\mathrm{LLL}}(x',x)\right]\\ &=\frac{i}{2}\operatorname{tr}\left[P_{+}\gamma^{0}S_{\mathrm{LLL}}(x,x')\gamma^{0}D_{x'}^{k}S_{\mathrm{LLL}}(x',x)\right]. \tag*{(48)} \end{aligned}

where the two minus signs are due to the definition of GEμνρG_{\mathrm{E}}^{\mu\nu\rho} in Eq. (46) and the fermion loop. The LLL propagator SLLL(x′,x)S_{\mathrm{LLL}}(x',x) is given by

SLLL(x′,x)=⟨ψˉLLL(x′)ψLLL(x)⟩=eiϕ(x′,x)S~LLL(x′−x).(49)S_{\mathrm{LLL}}(x',x)=\langle\bar{\psi}_{\mathrm{LLL}}(x')\psi_{\mathrm{LLL}}(x)\rangle=e^{i\phi(x',x)}\widetilde{S}_{\mathrm{LLL}}(x'-x). \tag*{(49)}

with ψLLL(x)\psi_{\mathrm{LLL}}(x) is the field operator for the Dirac fermion in the LLL. The translational invariance is broken by the Schwinger phase

ϕ(x′,x)=−e∫xx′dzμ[Aμ(z)+12Fμν(zν−x′ν)],(50)\phi(x',x)=-e\int_x^{x'} dz^\mu\left[A_\mu(z)+\frac{1}{2}F_{\mu\nu}(z^\nu-x'^\nu)\right], \tag*{(50)}

which fulfills

Dx′μS~LLL(x′,x)=eiϕ(x′,x)[∂x′μ−ie2Fμν(xν′−xν)]S~LLL(x′−x).(51)D^\mu_{x'}\widetilde{S}_{\mathrm{LLL}}(x',x)=e^{i\phi(x',x)}\left[\partial^\mu_{x'}-\frac{ie}{2}F^{\mu\nu}(x'_\nu-x_\nu)\right]\widetilde{S}_{\mathrm{LLL}}(x'-x). \tag*{(51)}

The translation-invariant part of the imaginary-time propagator is given as

S~LLL(p)=2e−∣p⊥∣2/∣eB∣S1+1(p∥),S1+1(p∥)=P+γ∥μpμ=P+γ0p0+γ3p3,(52)\widetilde{S}_{\mathrm{LLL}}(p)=2e^{-\lvert p_\perp\rvert^2/\lvert eB\rvert}S_{1+1}(p_\parallel),\qquad S_{1+1}(p_\parallel)=\frac{\mathcal{P}_+}{\gamma^\mu_\parallel p_\mu}=\frac{\mathcal{P}_+}{\gamma^0p_0+\gamma^3p_3}, \tag*{(52)}

where pμ=(p0,p)p^\mu=(p_0,\mathbf{p}), p0=iπT(2n−1)p_0=i\pi T(2n-1), p=(p1,p2,p3)\mathbf{p}=(p^1,p^2,p^3), p∥μ=(p0,0,0,p3)p^\mu_\parallel=(p^0,0,0,p^3), p⊥μ=(0,p1,p2,0)p^\mu_\perp=(0,p^1,p^2,0) and P+=(1+iγ1γ2)/2\mathcal{P}_+=(1+i\gamma^1\gamma^2)/2. The field strength is now specified as F21=∂2A1−∂1A2=∂1A2−∂2A1=BF^{21}=\partial^2A^1-\partial^1A^2=\partial_1A_2-\partial_2A_1=B, namely,

Fkl=−ϵ⊥klB.(53)F^{kl}=-\epsilon^{kl}_\perp B. \tag*{(53)}

The transverse Levi-Civita symbol is defined by ϵ⊥12=−ϵ⊥21=1\epsilon^{12}_\perp=-\epsilon^{21}_\perp=1, i.e., ϵ⊥jk=ϵ3jk\epsilon^{jk}_\perp=\epsilon^{3jk}.

From the above preparation, the correlation function is written as

GE00k(x−x′)=i2tr⁡[P+γ0S~LLL(x−x′)γ0(∂lx+i(eB/2)ϵ⊥kl(x′l−xl))S~LLL(x′−x)]=i2T∑n∫d3p(2π)3T∑l∫d3q(2π)3e−i(q−p)⋅(x−x′)tr⁡[P+γ0S~LLL(q)γ0(−ipk+(eB/2)ϵ⊥kl∂lp)S~LLL(p)]=12T∑n∫d3p(2π)3T∑l∫d3q(2π)3e−iq⋅(x−x′)tr⁡[P+γ0S~LLL(p+q)γ0(pk+i(eB/2)ϵ⊥kl∂lp)S~LLL(p)].(54)\begin{aligned} G^{00k}_{E}(x-x')&=\frac{i}{2}\operatorname{tr}\left[\mathcal{P}_+\gamma^0\widetilde{S}_{\mathrm{LLL}}(x-x')\gamma^0\left(\partial_l^x+i(eB/2)\epsilon^{kl}_\perp(x'^l-x^l)\right)\widetilde{S}_{\mathrm{LLL}}(x'-x)\right]\\ &=\frac{i}{2}T\sum_n\int\frac{d^3p}{(2\pi)^3}T\sum_l\int\frac{d^3q}{(2\pi)^3}e^{-i(q-p)\cdot(x-x')}\operatorname{tr}\left[\mathcal{P}_+\gamma^0\widetilde{S}_{\mathrm{LLL}}(q)\gamma^0\left(-ip^k+(eB/2)\epsilon^{kl}_\perp\partial_l^p\right)\widetilde{S}_{\mathrm{LLL}}(p)\right]\\ &=\frac{1}{2}T\sum_n\int\frac{d^3p}{(2\pi)^3}T\sum_l\int\frac{d^3q}{(2\pi)^3}e^{-iq\cdot(x-x')}\operatorname{tr}\left[\mathcal{P}_+\gamma^0\widetilde{S}_{\mathrm{LLL}}(p+q)\gamma^0\left(p^k+i(eB/2)\epsilon^{kl}_\perp\partial_l^p\right)\widetilde{S}_{\mathrm{LLL}}(p)\right]. \tag*{(54)} \end{aligned}

The last line is obtained with the change of the variable q−p→qq-p\to q. Here, we define ∂pj=∂/∂pj=−∂/∂pj\partial_p^j=\partial/\partial p_j=-\partial/\partial p^j, and accordingly ∂jp=−∂pj\partial_j^p=-\partial_p^j. From Eq. (54), the correlation function in the momentum space reads

GE00k(q)=12T∑m∫d3p(2π)3tr⁡[P+γ0S~LLL(p+q)γ0(pk+i(eB/2)ϵ⊥kl∂lp)S~LLL(p)].(55)G^{00k}_{E}(q)=\frac{1}{2}T\sum_m\int\frac{d^3p}{(2\pi)^3}\operatorname{tr}\left[\mathcal{P}_+\gamma^0\widetilde{S}_{\mathrm{LLL}}(p+q)\gamma^0\left(p^k+i(eB/2)\epsilon^{kl}_\perp\partial_l^p\right)\widetilde{S}_{\mathrm{LLL}}(p)\right]. \tag*{(55)}

Hereafter, we focus only on k=1,2k=1,2, that is, the transverse component of GE⊥00k(q)G^{00k}_{E\perp}(q). The sign in front of eBeB is opposite to that in Eq. (15) of Ref. [17] for k=2k=2.

The phase space integral in GE⊥00k(q)G^{00k}_{E\perp}(q) is decomposed into those over the transverse and longitudinal momenta, as follows:

GE⊥00k(q)=I⊥k(q⊥)I∥(q),(56)G^{00k}_{E\perp}(q)=I^k_\perp(q_\perp)I_\parallel(q), \tag*{(56)}
I⊥k(q⊥)=2∫d2p⊥(2π)2(p⊥k+isϵ⊥lkp⊥l)e−(∣p⊥+q⊥∣2+∣p⊥∣2)/∣eB∣,(57)I^k_\perp(q_\perp)=2\int\frac{d^2p_\perp}{(2\pi)^2}\left(p^k_\perp+is\epsilon^{lk}_\perp p^l_\perp\right)e^{-(\lvert p_\perp+q_\perp\rvert^2+\lvert p_\perp\rvert^2)/\lvert eB\rvert}, \tag*{(57)}
I∥(q∥)=T∑n∫dpz2πtr⁡[P+γ0S1+1(p∥+q∥)γ0S1+1(p∥)],(58)I_\parallel(q_\parallel)=T\sum_n\int\frac{dp_z}{2\pi}\operatorname{tr}\left[\mathcal{P}_+\gamma^0S_{1+1}(p_\parallel+q_\parallel)\gamma^0S_{1+1}(p_\parallel)\right], \tag*{(58)}

where s=sgn⁡(eB)s=\operatorname{sgn}(eB). The former integration can be performed as

I⊥k(q⊥)=2e−∣q⊥∣2/(2∣eB∣)∫d2p⊥(2π)2(p⊥k+isϵ⊥lkp⊥l)e−2∣p⊥+q⊥/2∣2/∣eB∣=−e−∣q⊥∣2/(2∣eB∣)(q⊥k+isϵ⊥lkq⊥l)∫d2p⊥′(2π)2e−2∣p⊥′∣2/∣eB∣=−∣eB∣8πe−∣q⊥∣2/(2∣eB∣)(q⊥k+isϵ⊥lkq⊥l).(59)\begin{aligned} I^k_\perp(q_\perp)&=2e^{-\lvert q_\perp\rvert^2/(2\lvert eB\rvert)}\int\frac{d^2p_\perp}{(2\pi)^2}\left(p^k_\perp+is\epsilon^{lk}_\perp p^l_\perp\right)e^{-2\lvert p_\perp+q_\perp/2\rvert^2/\lvert eB\rvert}\\ &=-e^{-\lvert q_\perp\rvert^2/(2\lvert eB\rvert)}\left(q^k_\perp+is\epsilon^{lk}_\perp q^l_\perp\right)\int\frac{d^2p'_\perp}{(2\pi)^2}e^{-2\lvert p'_\perp\rvert^2/\lvert eB\rvert}\\ &=-\frac{\lvert eB\rvert}{8\pi}e^{-\lvert q_\perp\rvert^2/(2\lvert eB\rvert)}\left(q^k_\perp+is\epsilon^{lk}_\perp q^l_\perp\right). \tag*{(59)} \end{aligned}

For the latter integral, one can replace the Matsubara summation by a contour integral. By the use of identities tanh⁡z=±[1−2/(e±2z+1)]\tanh z = \pm[1 - 2/(e^{\pm2z} + 1)], one can separate the vacuum and thermal parts as I∥=I∥vac+I∥matI_{\parallel}=I_{\parallel}^{\mathrm{vac}}+I_{\parallel}^{\mathrm{mat}} with

I∥vac(q∥)=∫dpz2π∫−i∞i∞dp02πitr⁡[P+γ01γ∥μ(pμ∥+qμ∥)γ01γ∥νpν∥],I∥mat(q∥)=−∫dpz2π∑η=±∫Cηdp02πinF(ηp0)tr⁡[P+γ01γ∥μ(pμ∥+qμ∥)γ01γ∥νpν∥].(60)\begin{aligned} I_{\parallel}^{\mathrm{vac}}(q_{\parallel})&=\int\frac{dp_z}{2\pi}\int_{-i\infty}^{i\infty}\frac{dp_0}{2\pi i}\operatorname{tr}\left[\mathcal{P}_{+}\gamma^{0}\frac{1}{\gamma_{\parallel}^{\mu}(p_{\mu\parallel}+q_{\mu\parallel})}\gamma^{0}\frac{1}{\gamma_{\parallel}^{\nu}p_{\nu\parallel}}\right],\\ I_{\parallel}^{\mathrm{mat}}(q_{\parallel})&=-\int\frac{dp_z}{2\pi}\sum_{\eta=\pm}\int_{C_{\eta}}\frac{dp_0}{2\pi i}n_{\mathrm{F}}(\eta p_0)\operatorname{tr}\left[\mathcal{P}_{+}\gamma^{0}\frac{1}{\gamma_{\parallel}^{\mu}(p_{\mu\parallel}+q_{\mu\parallel})}\gamma^{0}\frac{1}{\gamma_{\parallel}^{\nu}p_{\nu\parallel}}\right]. \tag*{(60)} \end{aligned}

where we defined pz=p3p_z=p^3 and the thermal distribution function nF(p0)=1/[eβp0+1]n_{\mathrm{F}}(p_0)=1/[e^{\beta p_0}+1]. The contours C±C_{\pm} go along −i∞±ϵ→i∞±ϵ-i\infty\pm\epsilon\to i\infty\pm\epsilon with a positive infinitesimal number ϵ\epsilon, and pick up the residues at the positive- and negative-energy poles, respectively.

It is here important to notice that the thermal contribution I∥mat(q∥)I_{\parallel}^{\mathrm{mat}}(q_{\parallel}) vanishes totally, as shown in the following. First, we compute the spinor trace as tr⁡[P+γ0γ∥μγ0γ∥ν]=2(−η∥μν+2η∥μ0η∥ν0)\operatorname{tr}[\mathcal{P}_{+}\gamma^{0}\gamma^{\mu}_{\parallel}\gamma^{0}\gamma^{\nu}_{\parallel}]=2(-\eta_{\parallel}^{\mu\nu}+2\eta_{\parallel}^{\mu0}\eta_{\parallel}^{\nu0}) with η∥μν\eta_{\parallel}^{\mu\nu} being the longitudinal components of the Minkowskian metric tensor. The contour integral is then evaluated with the residues as

I∥mat=∑η=±η∫dpz2π[nF(∣pz∣)ηpz(ηpz+q0)+pz(pz+qz)2ηpz[(ηpz+q0)2−(pz+qz)2]+nF(∣pz+qz∣)(η(pz+qz)−q0)η(pz+qz)+pz(pz+qz)2[(η(pz+qz)−q0)2−pz2]η(pz+qz)]=12∑η=±1qz−ηq0∫dpz2π[nF(∣pz+qz∣)−nF(∣pz∣)].(61)\begin{aligned} I_{\parallel}^{\mathrm{mat}}={}&\sum_{\eta=\pm}\eta\int\frac{dp_z}{2\pi}\left[n_{\mathrm{F}}(|p_z|)\frac{\eta p_z(\eta p_z+q_0)+p_z(p_z+q_z)}{2\eta p_z[(\eta p_z+q_0)^2-(p_z+q_z)^2]} \right.\\ &\left. +n_{\mathrm{F}}(|p_z+q_z|)\frac{(\eta(p_z+q_z)-q_0)\eta(p_z+q_z)+p_z(p_z+q_z)}{2[(\eta(p_z+q_z)-q_0)^2-p_z^2]\eta(p_z+q_z)}\right]\\ ={}&\frac{1}{2}\sum_{\eta=\pm}\frac{1}{q_z-\eta q_0}\int\frac{dp_z}{2\pi}\left[n_{\mathrm{F}}(|p_z+q_z|)-n_{\mathrm{F}}(|p_z|)\right]. \tag*{(61)} \end{aligned}

where we used eηβq0=e2iπm=1e^{\eta\beta q_0}=e^{2i\pi m}=1 for q0=2iπTmq_0=2i\pi Tm. Since the above pzp_z-integrals are convergent, we can freely shift the integral variable. This manipulation leads to the cancellation between the two integrals in Eq. (61), and thus the conclusion that I∥mat(q∥)=0I_{\parallel}^{\mathrm{mat}}(q_{\parallel})=0.

The remaining piece is the vacuum contribution. This can be computed with the standard Feynman-parameter trick as [52, 53]

I∥(q∥)=I∥vac(q∥)=−1i∫d2p∥(2π)2tr⁡[P+γ0iγ∥μ(pμ∥+qμ∥)γ0iγ∥νpν∥]=−1πq∥2−q02q∥2=1πqz2q02−qz2.(62)\begin{aligned} I_{\parallel}(q_{\parallel})=I_{\parallel}^{\mathrm{vac}}(q_{\parallel})={}&-\frac{1}{i}\int\frac{d^{2}p_{\parallel}}{(2\pi)^{2}}\operatorname{tr}\left[\mathcal{P}_{+}\gamma^{0}\frac{i}{\gamma_{\parallel}^{\mu}(p_{\mu\parallel}+q_{\mu\parallel})}\gamma^{0}\frac{i}{\gamma_{\parallel}^{\nu}p_{\nu\parallel}}\right]\\ ={}&-\frac{1}{\pi}\frac{q_{\parallel}^{2}-q_{0}^{2}}{q_{\parallel}^{2}}=\frac{1}{\pi}\frac{q_{z}^{2}}{q_{0}^{2}-q_{z}^{2}}. \tag*{(62)} \end{aligned}

where we rotate the contour of the p0p_0-integral from −i∞→i∞-i\infty\to i\infty to −∞→∞-\infty\to\infty. This result is equal to the polarization function in the (1+1)(1+1)-dimensional QED divided by a factor of (−ie)2(-ie)^2 [53]. However, Eq. (19) of Ref. [17] has an opposite sign.

Performing the analytic continuation and taking the static limit, we have

lim⁡ω→0I∥(q0→ω+iδ,qz)=−1π.(63)\lim_{\omega\to0}I_{\parallel}(q_0\to\omega+i\delta,q_z)=-\frac{1}{\pi}. \tag*{(63)}

Finally, the correlation function reads

lim⁡ω→0GE0k(q0→ω+iδ,q)=(−1π)⋅(−eB8π)(q⊥k+iϵ3jkq⊥j)+O(∣q⊥∣/∣eB∣).(64)\lim_{\omega\to0}G_{\mathrm{E}}^{0k}(q_0\to\omega+i\delta,\boldsymbol{q})=\left(-\frac{1}{\pi}\right)\cdot\left(-\frac{eB}{8\pi}\right)\left(q_{\perp}^{k}+i\epsilon^{3jk}q_{\perp}^{j}\right)+O(|q_{\perp}|/\sqrt{|eB|}). \tag*{(64)}

Therefore, the vorticity-induced term reads

δjvortical0=eB8π2ϵ3jk∂jh0k=−eBΩ4π2.(65)\delta j_{\mathrm{vortical}}^{0}=\frac{eB}{8\pi^{2}}\epsilon^{3jk}\partial_{j}h_{0k}=-\frac{eB\Omega}{4\pi^{2}}. \tag*{(65)}

with Ω=−ϵ3jk∂jh0k/2\Omega=-\epsilon^{3jk}\partial_{j}h_{0k}/2. This is consistent with the isothermal transport coefficient that is derived from the partition function in the main text. In summary, the three opposite signs in Eqs. (42), (S40), and (S48) explain why Eq. (65) has the extra minus sign compared with the main result in Ref. [17]. This concludes that the linear-response analysis in the strong magnetic field contains both the orbital and spin contributions.

Linear response under a perturbative rotation

We show that the same charge density is reproduced from the linear-response theory in terms of a weak angular velocity Ω\Omega based on the correspondence in Eq. (41). The kinetic angular momentum operator explicitly appears in this analysis.

Let us consider the charge density δj0=λΩ\delta j^0=\lambda\Omega generated by the perturbation Hamiltonian (S24), i.e.,

Hext(x)=−ΩJkin(x),Jkin=12ϵ3jkxjψˉ(γ0Dk+γkD0)ψ.(66)H_{\mathrm{ext}}(x)=-\Omega J_{\mathrm{kin}}(x),\qquad J_{\mathrm{kin}}=\frac{1}{2}\epsilon^{3jk}x^j\bar{\psi}(\gamma^0D^k+\gamma^kD^0)\psi. \tag*{(66)}

As in the above analysis, the transport coefficient λ\lambda is computed from the imaginary-time correlator as

λ=lim⁡q→0[lim⁡ω→0GE(q0→ω+iδ,q)],GE(x,y)=−⟨Tτj0(x)(−Jkin(y))⟩.(67)\lambda=\lim_{q\to0}\left[\lim_{\omega\to0}G_{\mathrm{E}}(q_0\to\omega+i\delta,q)\right],\qquad G_{\mathrm{E}}(x,y)=-\left\langle T_\tau j^0(x)\left(-J_{\mathrm{kin}}(y)\right)\right\rangle. \tag*{(67)}

We compute the above correlator in the LLL limit at the one-loop order. A similar correlator has been directly evaluated with the covariant derivative DμD_\mu in the above and in [17]. Here, it is more useful to evaluate the covariant derivative as the kinetic angular momentum operator. That is, we can utilize the relation

12ϵ3jkxjψˉLLLγ0DkψLLL=12ψˉLLLγ0LkinψLLL=−s2ψˉLLLγ0ψLLL.(68)\frac{1}{2}\epsilon^{3jk}x^j\bar{\psi}_{\mathrm{LLL}}\gamma^0D^k\psi_{\mathrm{LLL}}=\frac{1}{2}\bar{\psi}_{\mathrm{LLL}}\gamma^0L_{\mathrm{kin}}\psi_{\mathrm{LLL}}=-\frac{s}{2}\bar{\psi}_{\mathrm{LLL}}\gamma^0\psi_{\mathrm{LLL}}. \tag*{(68)}

The kinetic angular momentum is evaluated as LkinψLLL=−sψLLL+(higher LL)L_{\mathrm{kin}}\psi_{\mathrm{LLL}}=-s\psi_{\mathrm{LLL}}+(\text{higher LL}) based on Eq. (2) in the main text. Thus, the angular momentum is expressed with the charge density:

Jkin(y)=−s2j0(y)+i2ϵ3jkxjψˉγkD0ψ.(69)J_{\mathrm{kin}}(y)=-\frac{s}{2}j^0(y)+\frac{i}{2}\epsilon^{3jk}x^j\bar{\psi}\gamma^kD^0\psi. \tag*{(69)}

Since the second term should vanish due to the spinor trace, the correlator GEG_{\mathrm{E}} takes the same form as the density-density correlator:

GE(x,y)=(−1)⋅(−s2)tr⁡[P+γ0SLLL(x,y)γ0SLLL(y,x)],(70)G_{\mathrm{E}}(x,y)=(-1)\cdot\left(-\frac{s}{2}\right)\operatorname{tr}\left[\mathcal{P}_{+}\gamma^0S_{\mathrm{LLL}}(x,y)\gamma^0S_{\mathrm{LLL}}(y,x)\right], \tag*{(70)}

where the additional minus sign is from the fermion loop. The momentum representation of the correlator reads

GE(q)=s2T∑n∫d3p(2π)3tr⁡[P+γ0S~LLL(p+q)γ0S~LLL(p)].(71)G_{\mathrm{E}}(q)=\frac{s}{2}T\sum_n\int\frac{\mathrm{d}^3p}{(2\pi)^3}\operatorname{tr}\left[\mathcal{P}_{+}\gamma^0\widetilde{S}_{\mathrm{LLL}}(p+q)\gamma^0\widetilde{S}_{\mathrm{LLL}}(p)\right]. \tag*{(71)}

One can further decompose the integral as GE(q)=I~⊥(q⊥)I∥(q∥)G_{\mathrm{E}}(q)=\widetilde{I}_{\perp}(q_{\perp})I_{\parallel}(q_{\parallel}) with the same I∥(q∥)I_{\parallel}(q_{\parallel}) as in Eq. (58). The transverse part I~⊥(q⊥)\widetilde{I}_{\perp}(q_{\perp}) is slightly different from Eq. (59), but is readily evaluated in the same manner as

I~⊥(q⊥)=2s∫d2p⊥(2π)2e−(∣p⊥∣2+∣p⊥+q⊥∣2)/∣eB∣=eB4πe−∣q⊥∣2/∣eB∣.(72)\widetilde{I}_{\perp}(q_{\perp})=2s\int\frac{\mathrm{d}^2p_{\perp}}{(2\pi)^2}e^{-\left(\lvert p_{\perp}\rvert^2+\lvert p_{\perp}+q_{\perp}\rvert^2\right)/\lvert eB\rvert}=\frac{eB}{4\pi}e^{-\lvert q_{\perp}\rvert^2/\lvert eB\rvert}. \tag*{(72)}

Plugging Eqs. (63) and (S58) into Eq. (67), we finally arrive at

λ=−eB4π2.(73)\lambda=-\frac{eB}{4\pi^2}. \tag*{(73)}

This result again agrees with the induced charge density derived in the main text. The origin of the negative sign is clear from Eq. (68). We reiterate that, according to the correspondence in Eq. (41), this analysis is equivalent to the analysis leading to Eq. (65) with the metric perturbation.

References

  1. [1]L. Allen, M. W. Beijersbergen, R. J. C. Spreeuw, and J. P. Woerdman, Phys. Rev. A 45, 8185 (1992).
  2. [2]R. E. Prange and S. M. Girvin, eds., The Quantum Hall Effect (Springer-Verlag New York Inc., New York, 1990).
  3. [3]M. Z. Hasan and C. L. Kane, Rev. Mod. Phys. 82, 3045 (2010).
  4. [4]N. Nagaosa and Y. Tokura, Nat. Nanotechnol. 8, 899 (2013).
  5. [5]A. Manchon, I. M. Miron, T. Jungwirth, J. Sinova, J. Zelezný, J. Wunderlich, C. H. Back, and E. V. Gomonay, Rev. Mod. Phys. 91, 035004 (2019).
  6. [6]A. Accardi et al., Eur. Phys. J. A 52, 268 (2016), arXiv:1212.1701 [nucl-ex]; E. Leader and C. Lorcé, Phys. Rep. 541, 163 (2014), arXiv:1309.4235 [hep-ph]; R. Abdul Khalek et al., Nucl. Phys. A 1026, 122447 (2022), arXiv:2103.05419 [physics.ins-det].
  7. [7]L. Adamczyk et al. (STAR), Nature 548, 62 (2017), arXiv:1701.06657 [nucl-ex].
  8. [8]A. Einstein and W. de Haas, Proc. KNAW 181, 696 (1915).
  9. [9]S. J. Barnett, Phys. Rev. 6, 239 (1915).
  10. [10]S. P. Klevansky, Rev. Mod. Phys. 64, 649 (1992).
  11. [11]Y. Jiang and J. Liao, Phys. Rev. Lett. 117, 192302 (2016), arXiv:1606.03808 [hep-ph].DOI
  12. [12]B. S. Chandrasekhar, Appl. Phys. Lett. 1 (1962); A. M. Clogston, Phys. Rev. Lett. 9, 266 (1962).
  13. [13]D. E. Kharzeev, L. D. McLerran, and H. J. Warringa, Nucl. Phys. A 803, 227 (2008), arXiv:0711.0950 [hep-ph]; K. Fukushima, D. E. Kharzeev, and H. J. Warringa, Phys. Rev. D 78, 074033 (2008), arXiv:0808.3382 [hep-ph].DOI
  14. [14]D. T. Son and P. Surowka, Phys. Rev. Lett. 103, 191601 (2009), arXiv:0906.5044 [hep-th].DOI
  15. [15]K. Landsteiner, E. Megias, and F. Pena-Benitez, Phys. Rev. Lett. 107, 021601 (2011), arXiv:1103.5006 [hep-ph].DOI
  16. [16]K. Hattori and X.-G. Huang, Nucl. Sci. Tech. 28, 26 (2017), arXiv:1609.00747 [nucl-th].DOI
  17. [17]K. Hattori and Y. Yin, Phys. Rev. Lett. 117, 152002 (2016), arXiv:1607.01513 [hep-th].DOI
  18. [18]S.-Z. Yang, J.-H. Gao, Z.-T. Liang, and Q. Wang, Phys. Rev. D 102, 116024 (2020), arXiv:2003.04517 [hep-ph]; K. Mameda, Phys. Rev. D 108, 016001 (2023), arXiv:2305.02134 [hep-th]; S.-Z. Yang, J.-H. Gao, and S. Pu, Corrections from space-time dependent electromagnetic fields to Wigner functions and spin polarization (2024), arXiv:2409.00456 [hep-ph].DOI
  19. [19]S. Lin and L. Yang, JHEP 06 (2021), 054, arXiv:2103.11577 [nucl-th].DOI
  20. [20]N. Yamamoto and D.-L. Yang, Phys. Rev. D 103, 125003 (2021), arXiv:2101.13208 [hep-th].
  21. [21]D. E. Kharzeev and H.-U. Yee, Phys. Rev. D 84, 045025 (2011), arXiv:1105.6360 [hep-th]; J. Hernandez and P. Kovtun, JHEP 05 (2017), 001, arXiv:1703.08757 [hep-th]; K. Hattori, Y. Hirono, H.-U. Yee, and Y. Yin, Phys. Rev. D 100, 065023 (2019), arXiv:1711.08450 [hep-th]; M. Hongo and K. Hattori, JHEP 02 (2021), 011, arXiv:2005.10239 [hep-th]; S. Wang and X.-G. Huang, Phys. Rev. D 109, L121302 (2024), arXiv:2307.06746 [nucl-th].DOI
  22. [22]K. Hattori, M. Hongo, and X.-G. Huang, Symmetry 14, 1851 (2022), arXiv:2207.12794 [hep-th].DOI
  23. [23]H.-J. Chen, K. Fukushima, X.-G. Huang, and K. Mameda, Phys. Rev. D93, 104052 (2016), arXiv:1512.08974 [hep-ph].DOI
  24. [24]S. Ebihara, K. Fukushima, and K. Mameda, Phys. Lett. B 764, 94 (2017), arXiv:1608.00336 [hep-ph].
  25. [25]K. Fukushima, T. Shimazaki, and L. Wang, Phys. Rev. D 102, 014045 (2020), arXiv:2004.05852 [hep-ph].DOI
  26. [26]C. R. Greenhelds, R. L. Stamps, S. S. Franke-Arnold, and S. M. Barnett, Phys. Rev. Lett. 113, 240404 (2014).
  27. [27]See Supplemental Material for detailed derivations, which includes Refs. [48–53].
  28. [28]D. T. Son and M. Stephanov, Phys. Rev. Lett. 133, 145201 (2024), arXiv:2405.08073 [nucl-th].DOI
  29. [29]L. D. Landau and E. M. Lifshitz, Statistical Physics, 3rd ed., Vol. 5 (Butterworth-Heinemann, Oxford, 1980).
  30. [30]A. Vilenkin, Phys. Rev. D20, 1807 (1979); Phys. Rev. D 21, 2260 (1980).
  31. [31]H. A. Weldon, Phys. Rev. D 26, 1394 (1982); F. Becattini, Phys. Rev. Lett. 108, 244502 (2012),

Paper details

vlnc:W_rzttxa8d
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