Arxiv:1808.08016V3 [Hep-Ph] 12 Apr 2019

Eddy magnetization from the chiral Barnett eﬀect

Kenji Fukushima,1 Shi Pu,1, 2 and Zebin Qiu1 1Department of Physics, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-0033, Japan 2Department of Modern Physics, University of Science and Technology of China, Hefei, Anhui 230026, China We discuss the spin, the angular momentum, and the magnetic moment of rotating chiral fermions using a kinetic theory. We find that, in addition to the chiral vortical contribution along the rotation axis, finite circular spin polarization is induced by the spin-momentum correlation of chiral fermions, which is canceled by a change in the orbital angular momentum. We point out that the eddy magnetic moment is nonvanishing due to the g-factors, exhibiting the chiral Barnett effect.

I. INTRODUCTION stead of measuring both two effects for the same physical system. Here, we lay out several examples of experimen- The Barnett effect refers to the magnetization induced tal realization: The Einstein–de Haas effect has been ob- by mechanical rotation of a charge neutral object [1]. The served in thin film deposited on a microcantilever [10]. Einstein–de Haas effect is an inverse phenomenon [2], There are several proposals for experiments in an atomic that is, a finite rotation attributed to a change in the gas with Bose-Einstein condensate [11, 12] and in a fer- magnetization. We can regard these two closely related romagnetic insulator with phonons [13]. In contrast to effects as realization of transmutation between the spin, the Einstein–de Haas effect, the Barnett effect has been S, and the orbital angular momentum, L, via the LS measured in systems such as the magnetic nanostruc- coupling. Because of the conservation law of the total tures [14], nuclear magnetic resonance [15], paramagnetic angular momentum J, a change in the magnetization materials [16], etc. Furthermore, circular spin-current or S must be compensated by a change in L. For a generation has been theoretically predicted as a result of historical summary of the theory and the experiments, spin-orbit interaction and the mechanical rotation [17– Ref. [3] is one of the most comprehensive reviews, in 19], which has an analoguous feature to what we are go- which the gyromagnetic effects including not only the ing to discuss in the present work. Interestingly, the the- above-mentioned two effects but also Maxwell’s exper- oretical predictions have been experimentally confirmed iment and the gyromagnetic magnetization by rotating recently, see Refs. [20–22]. For more details, interested magnetic fields are explained from a general point of readers can consult a recent textbook [23]. view. Possible extension of the Barnett effect to systems with Before relativistic generalization of the Barnett effect, massless or chiral fermions is an intriguing problem, and which is the central subject studied in the current work, theoretical investigations are demanded by recent exper- it would be useful to review key equations briefly for the imental developments. In high-energy experiments with conventional Barnett effect. A finite rotation with the almost massless quarks involved, the most pertinent ef- angular velocity vector ω would shift the one-particle en- fort lies in the measurement of Λ and Λ¯ global polariza- ergy by ω · J. This energy shift should be equated to the tion conducted by the STAR Collaboration of the Rel- magnetic energy of µ · Heff with the magnetic moment ativistic Heavy-Ion Collider (RHIC) [24, 25]. In non- µ and the effective magnetic field Heff corresponding to central collisions, two nuclei collide with a huge orbital the magnetization. We note that the vacuum perme- angular momentum [26–31], creating the “most vortical ¯ ability is µ0 = 1 in our convention of the natural unit. fluid,” and inducing a nonzero value of Λ and Λ global With the magnetic susceptibility, χB, the magnetization polarization. arXiv:1808.08016v3 [hep-ph] 12 Apr 2019 is given as M = χBHeff , and the magnetic moment is Many works have been published to formulate the µ = γJ where γ denotes the gyromagnetic ratio. Com- transfer from the orbital angular momentum to the bining these relations to eliminate Heff and µ, we finally spin carried by hot and dense hadronic matter. Some get the well-known formula, i.e., M = (χB/γ)ω. examples include the microscopic spin-orbital coupling The Barnett/Einstein–de Haas effects have attracted model [26–28], the statistical hydrodynamical model [32– attentions in general physics fields including condensed 38], and the quantum kinetic theory with Wigner func- matter physics for years. Theoretical studies are found, tions [39–41]. Moreover, it was proposed in Refs. [26, 27] for example, in the rotational states of nanostructured that the local polarization of Λ and vector mesons could magnetic systems [4–7]. It has also been pointed out also be experimentally sensitive to the net orbital angu- in Refs. [7–9] that both the Barnett and the Einstein– lar momentum. For more relevant references and discus- de Haas effects are governed by the same gyromagnetic sions, see the review [42]. Although there are theoretical tensor components which satisfy the Onsager reciprocal simulations for the observed polarization of hadrons, to relations, i.e., the gyromagnetic relation. In experiments, deepen our understanding from the fundamental level, it therefore, confirming one of them could be sufficient in- would be instructive to analyze an idealized environment 2 of noninteracting and rotating chiral fermions, as done in fermions. In Sec.III we will write down the expressions this work. of the orbital angular momentum and the spin operators In relativistic systems the chiral anomaly plays an im- in the kinetic theory language. In Sec.IV we will con- portant role for inducing an imbalance with respect to sider the CKT in a globally rotating chiral system and chirality [43, 44]. A clear manifestation of the chiral will compute the orbital angular momentum and the spin. anomaly can be found in a system with electromagnetic Next, we will relate our results to the Einstein–de Haas fields and/or finite vorticity; one can easily understand and the Barnett effects and will discuss their chiral ex- the chiral anomaly in such a system in terms of helic- tensions in Sec.V. We will also make a comment on the ity conservation; the helicity of fermions can be inter- anomalous hydrodynamics in Sec.VI. Finally we sum- changed with the magnetic helicity and/or the fluid he- marize our results in Sec.VII. Throughout this paper we licity. Then, helicity transmutation results in the chi- use the natural unit for the speed of light, c = 1, while ral magnetic effect (CME) [45–47], the chiral vortical ef- we retain ¯h. fect (CVE) [48–50], and related topological effects (see Refs. [51–53] for reviews). These topological effects induce nondissipative currents and survive in the hydro- II. ANGULAR MOMENTUM dynamic limit. Thus, the quantum anomaly could be DECOMPOSITION macroscopically manifested. Interestingly, it has been argued within the framework of hydrodynamics [54] that The angular momentum is a conserved quantity, but its the transmutation between the helicity of fermions and decomposition into the spin and the orbital components the fluid helicity, which is related to the CVE, can be re- is not unique in relativistic theories. In this section, we garded as an analogue of the Barnett/Einstein–de Haas clarify our convention and explain its physical interpre- effects. tation. Let us start with a free Dirac field (where the In this paper we will first discuss the properties of ro- generalization to include interaction is not difficult by tating chiral fermions using the kinetic theory and then ∂µ → Dµ), whose Lagrangian density is, address a possible connection to the hydrodynamic coun- L = ψ¯i¯hγµ∂ − mψ . (1) terpart. We take this strategy since in this way a physi- µ cal interpretation of the spin and the orbital parts would This Lagrangian is invariant under an infinitesimal rota- be transparent combined with field-theoretical considera- tion, tions. The Boltzmann equation assumes a quasi-particle µ 0µ µ µ ν approximation, which is a semiclassical treatment of dy- x → x = x + ν x , (2) namics. Recently it has been established how to im- µ where ν is an antisymmetric tensor whose magnitude is plement the spin degrees of freedom in the Boltzmann infinitesimally small. The angular momentum tensor is equation. Such an augmented Boltzmann equation is the Nöther current associated with rotation symmetry. commonly called the chiral kinetic theory (CKT) in the We note that, under Eq. (2), the spinor transforms as high-energy physics community. There are a number of i literature using different ways to derive the CKT; e.g., ef- ψ(x) → ψ0(x0) = ψ(x) − Σµν ψ(x) , (3) fective field theories [55–58], path integrals [59–61], and 2 µν Wigner functions [62–68]. where Σµν = (i/4)[γµ, γν ]. Correspondingly, the Nöther Once we have the CKT, it is straightforward to derive current with current index λ gets two contributions; the the macroscopic currents and the energy-momentum ten- coordinate part from Eq. (2) and the spinor part from sor. By integrating physical observables weighted with Eq. (3) as the distribution function over the momentum, one will obtain the expectation values of the observables such J λµν = Lλµν + Sλµν . (4) as the vector and axial vector currents with quantum We can express the first term, Lλ µν , using the canonical corrections, which are identified as the CME and the energy-momentum tensor, CVE [39, 60–63]. Remarkably, the correct transport coefficient of the CVE contains two origins. The first one ∂L ∂ψ T µν = = ψ¯ i¯hγµ∂ν ψ , (5) comes from a shift in the particle energy dispersion mod- ∂(∂µψ) ∂xν ified by the rotation. The nontrivial Lorentz transfor- mations for massless particles, called the side-jump ef- as the following form, fects [61–64], make the second contribution to the CVE Lλµν = xµT λν − xν T λµ coefficient. Since the description of the CVE in terms of λ µ ν λ ν µ the CKT has been fully established, it is natural for us = ψ¯ i¯h γ x ∂ − γ x ∂ ψ . (6) to employ the CKT for the Barnett effect that is related For the second term of Eq. (4), the explicit form reads, to the CVE. The present paper is organized as follows. In Sec.II we ∂L h i i 1 Sλµν = − Σµν ψ(x) = ψ¯ i¯hγλγµ, γν ψ . will give a brief review on the total angular momentum, ∂(∂λψ) 2 4 the orbital angular momentum, and the spin of chiral (7) 3

One could also obtain another form of the while the canonical ones, Lλµν and Sλµν are not. How- spin tensor from the symmetrized Dirac La- ever, this fact does not mean any superiority of L˜λµν and −→ ←− ih¯ ¯ µ ¯ µ S˜λµν because neither of them is a true symmetry genera- grangian, L = 2 ψγ ∂µψ − ψγ ∂µψ , that is, λµν 1 ¯ λ µ ν tor alone. The situation is quite similar to the decompo- S = 8 ψ i¯h{γ , γ , γ }ψ, but we are interested in S0µν components for later discussions and the differ- sition of the optical spin and the optical orbital angular ence from Eq. (7) is vanishing and the choice of the momentum. For free electromagnetic fields one can gen- Lagrangian is irrelevant for physical quantities as it erally define individually conserved spin and orbital an- should. gular momentum operator, but due to the transversality Now, the total angular momentum tensor is constraint, only their combination, i.e., the total angular momentum is the physically meaningful quantity [70, 71]. λµν λ µ ν λ ν µ 1 λ µ ν Throughout this work we adopt the canonically defined J = ψ¯ i¯h γ x ∂ − γ x ∂ + γ γ , γ ψ , (8) λµν λµν 4 spin S and orbital angular momentum L , because these are the definitions with most natural connection to whose λ = 0 component is the conserved charge density, their nonrelativistic counterparts. Another advantage is i.e., the conserved angular momentum. Using the Dirac that S0ij, or S, turns out to be nothing but the axial equation, we can easily check that current, λµν λµν ¯ µ ν ν µ k ∂λL = −∂λS = ψ i¯h(γ ∂ − γ ∂ )ψ , (9) 0ij ijk ¯h ¯ k ijk j5 S = ψγ γ5ψ = , (15) λµν 2 2 from which ∂λJ = 0 immediately follows. If the sur- 1 face term is irrelevant, we can then arrive at the angular Sk ≡ ijkS0ij . (16) 2 momentum conservation law: Thus, it has an interpretation evidently related to the d Z d3x J 0µν = 0 . (10) chiral anomaly. Similarly we define the orbital angular dt momentum vector L as 1 One might have thought that the above identification of Lk ≡ ijkL0ij . (17) L0µν and S0µν as the orbital and the spin components 2 would be the most natural. Indeed, in the nonrelativis- Equation (15) also implies that, if the axial current is a tic limit, L0µν and S0µν amount to the orbital and the measurable physical observable, so will S and then L be. spin components, respectively. Nevertheless, in relativistic theories, no unique decomposition is guaranteed. Actually, the energy-momentum tensor always has am- III. TRANSCRIPTION TO KINETIC THEORY biguity by an arbitrary antisymmetric tensor Σµνλ as Since we will deal with our problem in terms of ki- µν µν µνλ Θ = T + ∂λΣ . (11) netic theory, we should seek for corresponding expressions for Lλµν and Sλµν involving the distribution func- It is obvious that Θµν also satisfies the conservation law, tion, f(p, x, t). We note that the spin and the orbital so it is equally qualified as the energy-momentum tensor. angular momentum are the properties of matter in equi- In particular, with an appropriate choice of Σµνλ, one can librium unrelated to the collisions, once the correspond- make Θµν symmetric as ing operators for Lλµν and Sλµν are identified. Although 1 we discuss the kinetic theory transcription, we are not Θµν = ψ¯ i¯h(γµ∂ν + γν ∂µ)ψ . (12) 2 studying the off-equilibrium dynamics, but we are con- sidering the operators in terms of the kinetic theory in The corresponding “orbital” component of the angular this section and in terms of hydrodynamics in Sec.VI. momentum, deduced from Eq. (6) with T µν replaced by To this end, we consider the single-particle angular mo- Θµν , is mentum tensor as done in Ref. [58, 72], i.e., 1 1 J µν = xµpν − xν pµ + Sµν , (18) L˜λµν = Lλµν + ψ¯ i¯h(xµγν − xν γµ)∂λψ , (13) 2 2 where pµ = (p = |p|, p). Comparing Eq. (18) with µ µ and the “spin” component is inferred from S˜λµν = J λµν − Eq. (8), given the correspondence of i¯h∂ → p , we can 0µν L˜λµν . Interestingly, using the Dirac equation again, we identify the first two terms as the orbital part L of can prove, our choice. Then, the last term represents the spin part, whose specific form, according to Ref. [58, 63, 64], is fixed λµν λµν ∂λL˜ = ∂λS˜ = 0 . (14) up to a frame vector nβ. We choose the laboratory frame with nβ = (1, 0), which simplifies the concrete expression In contrast to Eq. (9), the above relation (14) indicates of Sµν , leading to the following operator decomposition: that, in this construction, the orbital and the spin compo- pˆ nents of the angular momentum are separately conserved L = x × p , S =hλ ¯ pˆ − ¯hλ × ∇ . (19) (see Ref. [69] for a related discussion on electron vortices), p 4

Here λ is the helicity, i.e., λ = ±1/2 and pˆ = p/|p| is the we introduce ω 6= 0 into the system. For this purpose unit momentum vector. we put ourselves into a comoving frame that rotates to- We emphasize the importance of the second term in S gether with matter. We can thereby postulate that the to make the computation consistent with the CVE and local thermal equilibrium is reached after a sufficiently the relation (15). This additional term originates from a long time, so that f = f(εrot) with εrot defined in the co- gyromagnetic effect and is nothing but a familiar Rashba moving frame (which is implicitly assumed in the imple- spin-orbit coupling. Another way to think of the field- mentation of finite-temperature field theory in Ref. [48]). theoretical origin of this term is the current expectation We can solve a free Weyl equation in the rotating frame value as a derivative with respect to the vector potential. to find εrot as Then, as discussed in Ref. [61, 63, 64], the current reads: εrot = p − ω · x × p + ¯hλpˆ (22) Z pˆ j = pˆ − ¯hλ × ∇ f , (20) using the lab-frame (nonrotating) coordinates x and mo- p p menta p. We note that the energy shift in Eq. (22) takes where the second term in the parentheses appears from a standard cranking form, −ω · J. In terms of lab-frame a magnetic dependent term, −λpˆ · B/|p|, in the energy variables f(εrot) is neither homogeneous in coordinate dispersion relation, which is eventually transcribed into space nor isotropic in momentum space due to finite ro- the additional term in S as seen above. An interesting tation, thus the spin and the orbital angular momentum point worth mentioning is that ∇ is the spatial derivative derived from f(εrot) can be nonzero. We begin with cal- and a finite rotation would indeed induce spatial inhomo- culating the spin expectation value under an assumption geneity. that ω is small. Up to the linear order of ω we get, We note that one can understand Eqs. (19) and (20) Z pˆ easily from the well-known Gordon decomposition on the hSi = λ¯h pˆ − λ¯h × ∇ f(εrot) p vector current with Dirac spinors at momentum p [73], p Z p Z i.e., ≈ −¯hλ(ω × x) f 0(p) − ¯h2λ2ω f 0(p) (23) 3 ¯h ¯h p p jˆ =h ¯ψ¯γψ = ψ†∇ψ − (∇ψ†)ψ + ∇ × (ψ†Σψ) , 2ip 2p where f 0(p) = ∂f(p)/∂p. It should be mentioned that (21) our “expectation value” involves only the momentum in- k ijk ij R R 3 3 where Σ = Σ . This is the mathematical back- tegration, p = d p/(2π¯h) , but not the coordinate ground for Eq. (20). Because extra γ5 is irrelevant for a R R 3 integration, which is denoted later by V = d x. system with either left or right handed particles only, the Here, we briefly mention the difference between setups argument on the vector current can be straightforwardly in Refs. [58, 61] and ours. If the rotation effects are intro- translated to the axial current in Eq. (19). duced by local vorticity vector as in Refs. [58, 61], physi- It should be noted that L and S have the same physical cal quantities can be homogeneous. However, to charac- unit but ¯h in L is hidden in the momentum p which looks terize the Einstein-de Haas effect, we implicitly assume a 0 O(¯h ) in a semiclassical treatment. Such ¯h counting is finite size system, for which the center of rotation is well- consistent with our intuition that the spin is a quantum defined. Then, the velocity of rotating particles depends effect but the orbital angular momentum is a macroscopic on the distance from the center of rotation, and phys- observable, while the full consistent treatment would re- ical quantities including the spin expectation value can quire the derivative expansion. be dependent on x as seen in the first term in Eq. (23). We shall make a remark about our power counting of ¯h order. In the last section we found the operators for IV. ROTATING CHIRAL FERMIONS the spin and the orbital angular momentum in a heuristic way. In principle, one could utilize the Wigner function In this work we study the effect of bulk rotation of chi- to take account of quantum corrections systematically in 3 ral matter at constant angular velocity ω rather than a the ¯h expansion. Then, S and εrot may have O(¯h ) and fluid with local vorticity. We turn electromagnetic fields O(¯h2) corrections, respectively, and they contribute to off for simplicity, and if necessary, the generalization in- an O(¯h3) correction to Eq. (23). cluding electromagnetic fields is straightforward. We next turn to the orbital angular momentum. In For an equilibrium state in the absence of rotation, the same way we expand the distribution function with the distribution function f is homogeneous in coordinate space and isotropic in momentum space, which means that f should be a function of single particle energy ε, i.e., f = f(ε) 1. Let us consider what would change if sponds to the “global equilibrium” case because our distribution function is independent of nβ up to the ¯h order. In the “local equilibrium” case the distribution function may depend on nβ which is generally a function of spatial coordinates. For more discussions on polarization effects in the local equilibrium case, 1 According to some references [61, 63, 64] our assumption corre- see Ref. [72]. 5 respect to ω and obtain Z hLi ≈ (x × p) f 0(p)(−ω) · (x × p + ¯hλpˆ) p Z p2 Z p = −x × (ω × x) f 0(p) + ¯hλ(ω × x) f 0(p) . p 3 p 3 (24) Equations (23) and (24) are our central results in this paper. In the following sections we shall expound their physical interpretations.

V. APPLICATIONS – CHIRAL EINSTEIN–DE HAAS / BARNETT EFFECTS FIG. 1. A schematic illustration for an intuitive picture to We utilize our results for hLi and hSi to discuss the understand the circular spin polarization and the associated eddy magnetization µ in a rotating chiral system with the relativistic extension of the Einstein–de Haas eﬀect and angular velocity vector ω. For simplicity we only consider the Barnett eﬀect. the right-handed fermions in the illustration. The red arrows stand for the direction of particle momentum and spin.

A. Chiral Einstein–de Haas Effect we will consider a cylindrically symmetric system which The physical meaning of Eq. (23) becomes transparent rotates rigidly around the z-axis, i.e., ω = ωzˆ. Then in once we add up both left-handed and right-handed con- such a setup the volume integration of hLimech yields tributions. After an integration by parts, the first term in Eq. (23), which is denoted by hSi hereafter, takes Z Z Z ⊥ 2 4 the following form as hLimech = ωzˆ r p fR(p) + fL(p) . (27) V V 3 p Z X p 0 hSi⊥ = −¯h λ(ω × x) fλ(p) Since p is the energy for chiral fermions, the p integra- p 3 R,L tion gives the energy density or the mass distribution, Z ¯h ¯h together with which the volume integration leads to the = (ω × x) fR(p) − fL(p) = (ω × x) n5 . 2 p 2 moment of inertia. To see this clearly, let us assume that (25) the distribution function features Fermi degeneracy to a chemical potential µ, and then the energy density, E, is where fR and fL refer to the distribution functions of 3 calculated as E = 4 µn where n is the number density. right-handed and left-handed particles respectively, and Consequently, 4 R pf (p) + f (p) reduces to a rela- thus n = n − n means the chirality density. Such 3 p R L 5 R L tivistic counterpart of the mass density, µ n + µ n . a rotation-induced spin alignment is intuitively under- R R L L From this argument it is clear that hLi corresponds stood as follows. For massless fermions, the spin and the mech to the mechanically induced orbital angular momentum, momentum directions are locked up. In this way, the an- which is naturally of O(¯h0). gular momentum is related to the translational motion. Therefore, if we macroscopically move our chiral matter Next, we delve into the second term in hLi given by with the velocity u = ω × x, the spin will be tilted along −hSi⊥. This term has an intriguing interpretation as the “Chiral Einstein–de Haas effect.” Let us consider the u. In this sense hSi⊥ is a unique result inherently for chiral fermions. Interestingly, this transverse eddy spin following thinking experiment; we rotate the fermionic alignment requires a finite chiral imbalance. We present system from the initial condition, hLi = hSi = 0. Ap- parently, the total angular momentum carried by rotating a schematic illustration in Fig.1 to explain how hSi⊥ appears in a rotating chiral system. chiral matter should be hJi = hLimech. However, as men- After similar algebra, we rewrite the orbital angular tioned above, due to the spin and momentum lock-up, momentum Eq. (24) as the transverse motion results in hSi⊥ 6= 0. This nonzero hSi must be canceled by a change in the orbital part Z ⊥ 4 so that the total angular momentum conservation can be hLi = x × (ω × x) p fR(p) + fL(p) − hSi⊥ 3 p satisfied. In this way, a shift by −hSi⊥ should arise in = hLi − hSi . (26) hLi. Such a physical mechanism is comparable to the mech ⊥ Einstein–de Haas effect. In the nonrelativistic case the Here, hLimech represents the first term involving ω ×x in spin is controlled by an external magnetic field, but it the above expression. We shall illuminate the physical in- can be changed by the momentum direction for chiral terpretation of hLimech in what follows. For concreteness fermions, which induces an orbital rotation. 6

We make two comments on the second term in Eq. (23). We can immediately identify the first term of hµLi as the The first one is that this term corresponding to the CVE mechanical contribution. The integration by parts makes can be also exactly canceled in a finite size system by it more visible as surface states not to violate the angular momentum con- 1 servation [74]. The second comment is that, if we con- hµ i = x × (ω × x) n , (32) L mech 2 e sider the zero n5 limit, the second term in Eq. (23) would dominate and lead to the local spin polarization proposed where ne represents the electric charge density. Given in Ref. [39]. that ω × x is the velocity vector associated with the rotating motion, the above expression is exactly the one known as the magnetic dipole moment from the Ampére B. Chiral Barnett Effect loop in classical electromagnetism. The second term of hµLi is at the same order as hµSi, Along similar lines, we can apply our formula to ad- but they do not cancel out. The total magnetization dress the Barnett effect for chiral fermions. That is, a reads: finite magnetization is generated by rotation [1], which Z can be quantified with our results. For this purpose we qe 0 hµi = hµLi + hµSi = hµLimech − ¯hλ (ω × x) f (p) . need the gyromagnetic ratio to convert the angular mo- 6 p mentum into the magnetic moment. For nonrelativis- (33) tic fermions, the gyromagnetic ratio is derived from the This result exhibits the chiral Barnett effect. What is Dirac equation as nontrivial in the relativistic case is the second term. It is proportional to ω × x, and thus has the circular ori- q q µ = µ + µ = g e L + g e S , (28) entation around the rotation axis, just like previously L S L 2m S 2m discussed hSi⊥ as shown in Fig.1. Since the magnetic moment is a source for the magnetic field, we can expect where qe and m are, respectively, the electric charge and the mass of the considered particle. For noninteracting a generation of eddy magnetic field in rotating chiral media. Further exploration on this point and applications Dirac fermions the g-factors are gL = 1 and gS = 2. Since to astrophysical objects will be reported elsewhere [76]. gL 6= gS, the right-hand side of Eq. (28) is not parallel to J = L+S. Once one takes an expectation value with the 2 J and Jz eigenstates, however, one can show that the right-hand side is projected onto the J direction, which VI. COMMENTS ON HYDRODYNAMICAL is guaranteed by the Wigner-Eckardt theorem, and the FORMULATION effective g-factor becomes the Landé g-factor. For chiral fermions Eq. (28) should be modified. In the In this section we briefly address the problem of calcu- chiral limit Eq. (28) turns into (see Ref. [73, 75])2 lating the orbital angular momentum in anomalous hydrodynamics [77]. In the framework of anomalous hy- qe qe drodynamics, the energy-momentum tensor reads (see µ = µL + µS = gL L + gS S . (29) 2p 2p Refs. [39, 78]):

The g-factors remain the same, and from now on we plug µν µ ν µν µ ν ν µ Thydro = (E+P )u u −P g +¯h n5(u ω +u ω ) , (34) gL = 1 and gS = 2 into µL and µS. We note that Eq. (29) is a local relation, and so we compute the expectation where E and P are the energy density and the pressure value as we did in the previous sections. The results up respectively, and uµ = γ(1, u) denotes the fluid velocity. to ¯h order are For simplicity we assume the small velocity limit, i.e., Z |u| 1 and uµ ≈ (1, u). In such limit the vorticity is qe 0 hµ i = − x × (ω × x) p f (p) ων = 1 ναβγ u ∂ u ≈ (0, ∇×u)+O(|u|2) by definition. L 6 2 α β γ p Using this energy-momentum tensor to define the hydro q Z + ¯hλ e (ω × x) f 0(p) , (30) orbital angular momentum, we find, 6 p Z Lij = xiT 0j − xjT 0i qe 0 hydro hydro hydro hµSi = −¯hλ (ω × x) f (p) . (31) i j j 3 p = x [(E + P )u + ¯h n5ω ] − (i ↔ j) . (35)

For a mechanically rotating ﬂuid, we specify uµ = (1, ω× x) and ωµ = (0, ω), which leads to 2 Infrared singularity in Eq. (29) is regularized by the Debye screening in many-body system. In other word, the momentum Lhydro = (E + P )(x × u) − ¯h n5(ω × x) . (36) convoluted with a distribution function has an infrared cutoﬀ by gT , where g is the coupling constant of the theory and T is the To see the connection between Eq. (36) and our results in temperature. a kinetic picture, we remind that massless noninteracting 7 systems has the equation of state as P = E/3, and in a VII. CONCLUSION kinetic framework E can be expressed as

Z 2 Z In this work we systematically discussed the spin, the (u · p) angular momentum, and the magnetic momentum for ro- E = 0 (fR + fL) ≈ p (fR + fL). (37) p p p tating chiral fermions using the framework of the CKT. First, we gave a brief review of deriving the angular mo- Then, eventually, the hydro angular momentum takes the mentum tensors as Nöther’s currents. Although the de- form of composition into the orbital and the spin components is 4 Z not unique, we adopted the canonical definition in which L = x × (ω × x) p(f + f ) − 2hSi . (38) the spin is directly related to the axial current. Next, we hydro 3 R L ⊥ p considered a globally rotating chiral system. Combin- ing the contributions from a shift in the particle energy Now we can make a direct comparison between Eq. (38) dispersion and an extra spin-orbital coupling term in the and our results in Eq. (26). We find that the first term CKT, we identified the expectation values of the spin [see corresponding to hµ i exactly agrees, but the coeffi- L mech hSi in Eq. (23)] and the orbital angular momentum [see cient for hSi is different. ⊥ hLi in Eq. (24)]. This discrepancy originates from different definitions Based on these two expressions for hSi and hLi, we de- of the energy-momentum tensor. The energy-momentum veloped a physical picture of the relativistic extension of tensor operator T µν obtained from the Nöther theorem in the Einstein–de Haas effect. Up to O(¯h) terms, the circu- Eq. (5) is not symmetric and does not correspond to the lar spin alignment is induced by the mechanical rotation hydrodynamic energy-momentum tensor in Eq. (34). In as illustrated in Fig.1, which can be intuitively under- fact, it is the symmetrized energy-momentum tensor op- stood through the fact that the rotation is accompanied erator Θµν in Eq. (12) that corresponds to Eq. (34). Such by the axial current. Then, a shift in hLi is caused by symmetrized definition is prevalently adopted in anoma- hSi to maintain the conservation law of the total angu- lous hydrodynamics (see, Refs. [58, 63, 64] for examples). lar momentum, which can be regarded as a relativistic Accordingly, for the orbital angular momentum, L hydro counterpart of the Einstein–de Haas effect realized in a corresponds to L˜0µν in Eq. (13) rather than the canonical 0µν chiral medium. one L in Eq. (6) used for our CKT computation. Furthermore, we applied our results to address the Bar- Here, we make a comment on the approximate spin nett effect for chiral fermions. We computed the mag- conservation law. Under such circumstance as massless netic moments, hµ i and hµ i, proportional to the or- fermions and vanishing electromagnetic fields, the axial L S µ bital angular momentum and the spin, respectively. The current j5 is conserved, and so does the spin. In hydro- 0 leading order term in hµLi of O(¯h ) is exactly the one dynamics one can show that this conservation law of hSi from the magnetic dipole moment obtained in classical follows from the expansions with respect to ¯h and ω. In electromagnetism. The next order terms in hµ i and terms of the fluid velocity uµ and the vorticity ωµ, the L µ hµSi of O(¯h) will not cancel out and this nonvanishing conservation law of j5 is [39, 50] magnetic moment exhibits what we call the chiral Bar- nett effect. ∂ jµ = ∂ n uµ + ¯h ξ ωµ = 0, (39) µ 5 µ 5 5 Before closing our discussions in the end, we supple- mented some discussions on the anomalous hydrody- where ξ is the CVE coefficient as a function of the tem- 5 namics. We pointed out that the symmetric energy- perature, the chemical potentials, etc. For a slowly ro- momentum tensor adopted in the anomalous hydrody- tating fluid with uµ ≈ (1, ω × x) Eq. (39) tells us that namics does not correspond to the canonical one derived from the Nöther theorem. Using the hydrodynamical d µ n5 = −∇ · n5ω × x + ¯h ∂µ(ξ5ω ) . (40) energy-momentum tensor, we could define another form dt of the orbital angular momentum Lhydro, which is ap- Then, for small ω, it is easy to verify that proximately an conserved quantity of O(¯h). There will be many possible extensions and applica- Z Z d ¯h d 2 2 2 tions of our work. As discussed in Sec.VB the rotation hSi = (n5ω × x) + O(¯h ) = O(¯h , ω ). dt V 2 V dt may induce the eddy magnetic fields, which could ex- (41) plain the internal structures of the neutron star [79] In We thus conclude that, under the approximation to drop astrophysics, generally, the magnetic field and the rota- terms of O(ω2) and/or O(¯h2), both our hLi in the canon- tion are commonly found in macroscopic objects, and the ical definition and hLihydro are equally qualified as the chiral Barnett effect may play an intriguing role [76]. An- orbital angular momentum. It should be noted that in other interesting direction lies in possible generalization the above argument we implicitly assumed that x is of to nonequilibrium situation. In this study we assumed order of the unity. This means that the system size must only a near equilibrium distribution function to compute not be as large as 1/¯h or 1/ω; otherwise, the surface state hSi and hLi in a steady state. The full real-time evolu- is not negligible [74]. tion of hSi and hLi starting with some initial condition 8 would be a challenging future problem. these effects dropped in the present work. In the future approximations made in the present work should be relaxed. Our treatment of fermions is limited to the massless case only, and the inclusion of finite mass effects would be a crucial step toward phenomenologi- ACKNOWLEDGMENTS cal applications to relativistic heavy-ion collision experiments. In this work, we neglected surface terms, and we should amend this with finite size effects taken into ac- We thank Yoshimasa Hidaka, Xu-Guang Huang, Di- count. So far, our analysis is only up to O(¯h) apart from Lun Yang, and Qun Wang for helpful discussions. K. F. the CVE term. Thus, we have not included higher order was supported by JSPS KAKENHI Grant No. 18H01211. nontrivial effects, e.g., the local spin polarization effect S. P. was supported by JSPS post-doctoral fellowship for [39]. We are currently making progresses to incorporate foreign researchers.

[1] S. J. Barnett, Phys. Rev. 6, 239 (1915). K. Nagao, and J.-y. Kim, Scientiﬁc Reports 8, 1974 [2] A. Einstein and W. J. de Haas, Verh. d. Deutsch. Phy. (2018). Ges. 17, 152 (1915). [23] S. Maekawa, S. O. Valenzuela, T. Kimura, and E. Saitoh, [3] S. J. Barnett, Rev. Mod. Phys. 7, 129 (1935). Spin current, Vol. 22 (Oxford University Press, 2017). [4] A. A. Kovalev, G. E. W. Bauer, and A. Brataas, Phys. [24] L. Adamczyk et al. (STAR), Nature 548, 62 (2017), Rev. B 75, 014430 (2007). arXiv:1701.06657 [nucl-ex]. [5] R. Jaafar, E. M. Chudnovsky, and D. A. Garanin, Phys. [25] B. I. Abelev et al. (STAR), Phys. Rev. C76, 024915 Rev. B 79, 104410 (2009). (2007), [Erratum: Phys. Rev.C95,no.3,039906(2017)], [6] E. M. Chudnovsky and D. A. Garanin, Phys. Rev. B 81, arXiv:0705.1691 [nucl-ex]. 214423 (2010). [26] Z.-T. Liang and X.-N. Wang, Phys. Rev. Lett. 94, 102301 [7] G. E. W. Bauer, S. Bretzel, A. Brataas, and (2005), [Erratum: Phys. Rev. Lett.96,039901(2006)], Y. Tserkovnyak, Phys. Rev. B 81, 024427 (2010). arXiv:nucl-th/0410079 [nucl-th]. [8] G. E. Bauer, E. Saitoh, and B. J. Van Wees, Nature [27] Z.-T. Liang and X.-N. Wang, Phys. Lett. B629, 20 materials 11, 391 (2012). (2005), arXiv:nucl-th/0411101 [nucl-th]. [9] L. Landau, E. Lifshitz, and L. Pitaevskii, Electrodynam- [28] J.-H. Gao, S.-W. Chen, W.-t. Deng, Z.-T. Liang, ics of continuous media. Course of theoretical physics Q. Wang, and X.-N. Wang, Phys. Rev. C77, 044902 (Butterworth-Heinemann, Oxford U.K., 1995). (2008), arXiv:0710.2943 [nucl-th]. [10] T. M. Wallis, J. Moreland, and P. Kabos, Applied [29] S. A. Voloshin, (2004), arXiv:nucl-th/0410089 [nucl-th]. physics letters 89, 122502 (2006). [30] B. Betz, M. Gyulassy, and G. Torrieri, Phys. Rev. C76, [11] Y. Kawaguchi, H. Saito, and M. Ueda, Phys. Rev. Lett. 044901 (2007), arXiv:0708.0035 [nucl-th]. 96, 080405 (2006). [31] F. Becattini, F. Piccinini, and J. Rizzo, Phys. Rev. C77, [12] U. Ebling and M. Ueda, (2017), arXiv:1701.05446 [cond- 024906 (2008), arXiv:0711.1253 [nucl-th]. mat.quant-gas]. [32] C. G. Van Weert, Annals of Physics 140, 133 (1982). [13] L. Zhang and Q. Niu, Phys. Rev. Lett. 112, 085503 [33] D. N. Zubarev, A. V. Prozorkevich, and S. A. Smolyan- (2014). skii, Theoretical and Mathematical Physics 40, 821 [14] S. Bretzel, G. E. Bauer, Y. Tserkovnyak, and A. Brataas, (1979). Applied Physics Letters 95, 122504 (2009). [34] F. Becattini and L. Tinti, Annals Phys. 325, 1566 (2010), [15] H. Chudo, M. Ono, K. Harii, M. Matsuo, J. Ieda, arXiv:0911.0864 [gr-qc]. R. Haruki, S. Okayasu, S. Maekawa, H. Yasuoka, and [35] F. Becattini, Phys. Rev. Lett. 108, 244502 (2012), E. Saitoh, Applied Physics Express 7, 063004 (2014). arXiv:1201.5278 [gr-qc]. [16] M. Ono, H. Chudo, K. Harii, S. Okayasu, M. Matsuo, [36] F. Becattini, V. Chandra, L. Del Zanna, and E. Grossi, J. Ieda, R. Takahashi, S. Maekawa, and E. Saitoh, Phys. Annals Phys. 338, 32 (2013), arXiv:1303.3431 [nucl-th]. Rev. B 92, 174424 (2015). [37] F. Becattini and E. Grossi, Phys. Rev. D92, 045037 [17] M. Matsuo, J. Ieda, E. Saitoh, and S. Maekawa, Phys. (2015), arXiv:1505.07760 [gr-qc]. Rev. Lett. 106, 076601 (2011). [38] T. Hayata, Y. Hidaka, T. Noumi, and M. Hongo, Phys. [18] M. Matsuo, J. Ieda, K. Harii, E. Saitoh, and S. Maekawa, Rev. D92, 065008 (2015), arXiv:1503.04535 [hep-ph]. Phys. Rev. B 87, 180402 (2013). [39] J.-H. Gao, Z.-T. Liang, S. Pu, Q. Wang, and [19] M. Matsuo, Y. Ohnuma, and S. Maekawa, Phys. Rev. B X.-N. Wang, Phys. Rev. Lett. 109, 232301 (2012), 96, 020401 (2017). arXiv:1203.0725 [hep-ph]. [20] R. Takahashi, M. Matsuo, M. Ono, K. Harii, H. Chudo, [40] R.-h. Fang, L.-g. Pang, Q. Wang, and X.-n. Wang, Phys. S. Okayasu, J. Ieda, S. Takahashi, S. Maekawa, and Rev. C94, 024904 (2016), arXiv:1604.04036 [nucl-th]. E. Saitoh, Nature Physics 12, 52 (2015). [41] L.-G. Pang, R.-H. Fang, H. Petersen, Q. Wang, and [21] D. Kobayashi, T. Yoshikawa, M. Matsuo, R. Iguchi, X.-N. Wang, Proceedings, 16th International Conference S. Maekawa, E. Saitoh, and Y. Nozaki, Phys. Rev. Lett. on Strangeness in Quark Matter (SQM 2016): Berkeley, 119, 077202 (2017). California, United States, J. Phys. Conf. Ser. 779, 012069 [22] A. Hirohata, Y. Baba, B. A. Murphy, B. Ng, Y. Yao, (2017). 9

[42] Q. Wang, Proceedings, 26th International Conference D89, 094003 (2014), arXiv:1312.2032 [hep-th]. on Ultra-relativistic Nucleus-Nucleus Collisions (Quark [61] J.-Y. Chen, D. T. Son, M. A. Stephanov, H.-U. Yee, Matter 2017): Chicago, Illinois, USA, February 5-11, and Y. Yin, Phys. Rev. Lett. 113, 182302 (2014), 2017, Nucl. Phys. A967, 225 (2017), arXiv:1704.04022 arXiv:1404.5963 [hep-th]. [nucl-th]. [62] J.-W. Chen, S. Pu, Q. Wang, and X.-N. Wang, Phys. [43] S. L. Adler, Phys. Rev. 177, 2426 (1969). Rev. Lett. 110, 262301 (2013), arXiv:1210.8312 [hep-th]. [44] J. S. Bell and R. Jackiw, Nuovo Cim. A60, 47 (1969). [63] Y. Hidaka, S. Pu, and D.-L. Yang, Phys. Rev. D95, [45] A. Vilenkin, Phys. Rev. D22, 3080 (1980). 091901 (2017), arXiv:1612.04630 [hep-th]. [46] K. Fukushima, D. E. Kharzeev, and H. J. Warringa, [64] Y. Hidaka, S. Pu, and D.-L. Yang, Phys. Rev. D97, Phys. Rev. D78, 074033 (2008), arXiv:0808.3382 [hep- 016004 (2018), arXiv:1710.00278 [hep-th]. ph]. [65] Y. Hidaka and D.-L. Yang, Phys. Rev. D98, 016012 [47] K. Fukushima, D. E. Kharzeev, and H. J. Warringa, (2018), arXiv:1801.08253 [hep-th]. Phys. Rev. Lett. 104, 212001 (2010), arXiv:1002.2495 [66] J.-h. Gao, S. Pu, and Q. Wang, Phys. Rev. D96, 016002 [hep-ph]. (2017), arXiv:1704.00244 [nucl-th]. [48] A. Vilenkin, Phys. Rev. D20, 1807 (1979). [67] J.-H. Gao, Z.-T. Liang, Q. Wang, and X.-N. Wang, [49] J. Erdmenger, M. Haack, M. Kaminski, and A. Yarom, (2018), arXiv:1802.06216 [hep-ph]. JHEP 01, 055 (2009), arXiv:0809.2488 [hep-th]. [68] A. Huang, S. Shi, Y. Jiang, J. Liao, and P. Zhuang, [50] D. T. Son and P. Surowka, Phys. Rev. Lett. 103, 191601 (2018), arXiv:1801.03640 [hep-th]. (2009), arXiv:0906.5044 [hep-th]. [69] S. M. Barnett, Phys. Rev. Lett. 118, 114802 (2017). [51] D. E. Kharzeev, J. Liao, S. A. Voloshin, and G. Wang, [70] S. J. van Enk and G. Nienhuis, EPL (Europhysics Let- Prog. Part. Nucl. Phys. 88, 1 (2016), arXiv:1511.04050 ters) 25, 497 (1994). [hep-ph]. [71] S. M. Barnett, Journal of Modern Optics 57, 1339 (2010). [52] J. Liao, Pramana 84, 901 (2015), arXiv:1401.2500 [hep- [72] D.-L. Yang, (2018), arXiv:1807.02395 [nucl-th]. ph]. [73] M. Stone, Int. J. Mod. Phys. B30, 1550249 (2015), [53] A. A. Burkov, J. Phys. Condens. Matter 27, 113201 arXiv:1507.01807 [physics.optics]. (2015), arXiv:1502.07609 [cond-mat.mes-hall]. [74] H.-L. Chen, K. Fukushima, D. Kharzeev, Y. Hirono, X.- [54] V. I. Zakharov, J. Exp. Theor. Phys. 120, 428 (2015). G. Huang, and K. Mameda, under preparation. [55] D. T. Son and N. Yamamoto, Phys. Rev. Lett. 109, [75] D. E. Kharzeev, M. A. Stephanov, and H.-U. Yee, Phys. 181602 (2012), arXiv:1203.2697 [cond-mat.mes-hall]. Rev. D95, 051901 (2017), arXiv:1612.01674 [hep-ph]. [56] M. Stone and V. Dwivedi, Phys. Rev. D88, 045012 [76] K. Fukushima, S. Pu, and Z. Qiu, under preparation. (2013), arXiv:1305.1955 [hep-th]. [77] D. T. Son and P. Surowka, Phys. Rev. Lett. 103, 191601 [57] D. T. Son and N. Yamamoto, Phys. Rev. D87, 085016 (2009), arXiv:0906.5044 [hep-th]. (2013), arXiv:1210.8158 [hep-th]. [78] S. Pu, J.-h. Gao, and Q. Wang, Phys. Rev. D83, 094017 [58] J.-Y. Chen, D. T. Son, and M. A. Stephanov, Phys. Rev. (2011), arXiv:1008.2418 [nucl-th]. Lett. 115, 021601 (2015), arXiv:1502.06966 [hep-th]. [79] K. Makishima, T. Enoto, J. S. Hiraga, T. Nakano, [59] M. A. Stephanov and Y. Yin, Phys. Rev. Lett. 109, K. Nakazawa, S. Sakurai, M. Sasano, and H. Murakami, 162001 (2012), arXiv:1207.0747 [hep-th]. Phys. Rev. Lett. 112, 171102 (2014), arXiv:1404.3705 [60] J.-W. Chen, J.-y. Pang, S. Pu, and Q. Wang, Phys. Rev. [astro-ph.HE].