betaFeedback

WorksPhysics and Astronomy

Chiral magnetic effect in a cylindrical domain

Abstract

We compute the chiral magnetic effect (CME) in a cylindrical region coaxial with the external magnetic field. As the boundary condition we require vanishing of the radial component of the electric current on the cylinder side wall. We find that when the magnetic length is comparable to or larger than the cylinder radius, the CME is suppressed compared to the corresponding result in an infinite medium. As a result, for a given cylinder radius, the suppression is stronger in weak fields. We argue that the electric current generated by the CME vanishes at the cylinder wall and monotonically increases toward the symmetry axis.

Introduction

Systems possessing chiral fermions can be found at vastly different energy and distance scales. For example, they include the quark-gluon plasma, Dirac and Weyl semimetals and certain neutron stars. Owing to the chiral anomaly, these chiral systems exhibit remarkable new properties such as nondissipative transport phenomena [1, 2, 3, 4, 5, 6, 7]. Since the chiral anomaly is an inherently quantum effect, the novel chiral properties associated with them are entirely absent in classical theories. Perhaps the most celebrated of them is the chiral magnetic effect (CME) [8, 9, 10] which refers to the induction of electric current along the direction of the external magnetic field:

j^=q22π2μAB^,(1)\hat{\mathbf{j}} = \frac{q^2}{2\pi^2}\mu_A\hat{\mathbf{B}}, \tag*{(1)}

where j^\hat{\mathbf{j}} is the electric current density, B^\hat{\mathbf{B}} is the magnetic field, qq is the electric charge, and μA\mu_A is the axial chemical potential; the hat sign denotes operators. The coefficient in front of B^\hat{\mathbf{B}} is referred to as the CME conductivity. The theoretical description of this effect was developed in many papers over the past decade. With a few notable exceptions, mentioned below, all studies ignore the boundary effects even though all realistic systems are confined in a finite region of space. A simple dimensional analysis can convince one of the importance of the finite-volume effects. The goal of this paper is to initiate study of the boundary conditions on the CME. To this end we examine the role of the cylindrical boundary whose symmetry axis coincides with the external magnetic field direction. The axial symmetry makes this model analytically tractable and allows us to develop keen insights into the CME dynamics in finite systems.

There are three reasons that necessitate imposing the boundary conditions. First, there are physical boundaries: nuclei, stars and material specimens all have finite dimensions. Second, many systems consist of domains of finite μA\mu_A each generating different chiral magnetic currents. For instance, it was argued in [11, 12] that the quark-gluon plasma may contain one large chiral domain. The vanishing of μA\mu_A outside this domain has a dramatic impact on CME [13, 14]. Third, the spacetime geometry can confine a system to a finite volume as happens, for example, in a rapidly rotating chiral fluid. Such a system is realized in off-center relativistic heavy-ion collisions that produce fluid with extremely high vorticity [15], and the causality demands that a particle cannot rotate with angular velocity Ω\Omega further than a distance of R=c/ΩR = c/\Omega. A number of studies have begun to examine the importance of the inclusion of these causal boundaries [16, 17, 18, 19, 20, 21, 22, 23, 24]. Admittedly, a system dynamics may, and usually is, richer than can be encapsulated by a boundary condition. Nevertheless, modeling finite size with boundary conditions is a good starting point.

The problem of consistent boundary conditions for a system of relativistic fermions has been addressed already at the dawn of QCD. The color confinement requires quarks to stay inside hadrons. In the MIT bag model this is accomplished by imposing boundary conditions at a spacelike surface [25, 26, 27]. Specifically, the normal component of the current is required to vanish. Many studies of the MIT bag model considered a spherical symmetry with a sphere as the boundary surface. We instead adjust the model to our problem by imposing the boundary condition on the side surface of a cylinder coaxial with the magnetic field. The height of the cylinder is immaterial so we simply assume that it extends to infinity. Although this model is axially symmetric, it breaks the chiral symmetry [25, 26, 27]. Indeed, the bouncing of a massless fermion off the boundary, flips its momentum and helicity/chirality.

The dynamics in the transverse direction depends on two radial scales. One is the radius of the cylinder RR, and the other one is the magnetic length lB=ℏc/∣qB∣l_B = \hbar c/\sqrt{|qB|}. The relative importance of the boundary condition on the properties of the system is indicated by the dimensionless parameter

ρR=R22lB2.(2)\rho_R = \frac{R^2}{2l_B^2}. \tag*{(2)}

When this ratio is large ρR≫1\rho_R \gg1, the boundary barely modifies the Landau states. However, when ρR∼1\rho_R \sim1, lBl_B is comparable to the radius of the cylinder and the finite size effects strongly deform the Landau levels and the properties of the system.

The role of boundaries on the CME was discussed before in [28, 29, 30] in the context of the two-dimensional slab geometry. When the magnetic field is directed along the normal to the flat slab boundary, analytical calculation yielded vanishing CME current [28, 30]. However, when the magnetic field is parallel to the boundaries of the slab, numerical lattice calculations showed that while the total current vanishes, it is locally finite near the slab boundary [29]. In our model, the axial chemical potential is introduced as a dynamical, albeit adiabatic, quantity and we preserve the axial symmetry of the magnetic field. Thus the generation of a CME current is not forbidden by the boundary conditions, which we confirm by explicit analytical calculation.

This paper is organized as follows. In Sec. II we review the derivation of the CME in thermal equilibrium at finite axial chemical potential using statistical mechanics. In Sec. III we obtain the solutions of the Dirac equation of a massless particle in a constant magnetic field using the MIT boundary conditions in a finite cylinder. We show in Sec. III B that these solutions are compatible with a thermal equilibrium description. Finally, in Sec. IV we derive the CME in a finite cylinder. We will refer to it as the “bound CME” in contrast to the “unbound CME” of the infinite system, given by (1). To compute the electric current along the magnetic field direction we use Vilenkin’s method [31] that consists in considering a system of relativistic fermions in thermal equilibrium at a finite axial chemical potential μA\mu_A, which describes the chiral imbalance.1 Although the CME is an out-of-equilibrium phenomenon [33], the use of thermal equilibrium is justified at timescales shorter than the relaxation time of the CME current. Our main results are Eq. (59) and Figure 2–Figure 6. Our summary and outlook are presented in Sec. V. We use natural units where ℏ=c=kB=1\hbar=c=k_B=1.

Figure 2. The normalized CME (62) at fixed ρR\rho R and different values of the magnetic field ∣qB∣|qB|/radius RR. Solid lines are the total contributions, while dashed lines are the contribution of the LLL. The bottom panel is the same plot as the one in the top, zoomed in on the total results (solid lines). The thickness of the lines is larger than the estimated error.

Figure 6. The characteristic value of the magnetic field at a fixed radius of the cylinder RR.

The CME from statistical mechanics

The CME can be computed using the methods of statistical mechanics assuming thermal equilibrium at finite axial imbalance [31, 34, 35]. Before addressing the effect of a finite boundary, in this section we review the derivation of the CME for a system of free massless fermions in an unbound space and we introduce the notation used in this work.

Thermal states are constructed starting from the solution of the Dirac equation with an external constant homogeneous magnetic field along the z^\hat{z} direction. Using cylindrical coordinates {t,r,ϕ,z}\{t,r,\phi,z\} and the Dirac representation of gamma matrices, the solution of the Dirac equation in an external gauge field Aμ=(0,−By/2,Bx/2,0)A^\mu=(0,-By/2,Bx/2,0), such that qB>0qB>0 with qq the charge

of the fermion, is, in an unbound space [36],

ψ∞=e−iEn,pzteipzz2πeimϕ2π∣qB∣(C1In,a(ρ)e−iϕ/2iC2In−1,a(ρ)eiϕ/2C3In,a(ρ)e−iϕ/2iC4In−1,a(ρ)eiϕ/2)(3)\psi_\infty=e^{-iE_{n,p_z}t}\frac{e^{ip_z z}}{\sqrt{2\pi}}\frac{e^{im\phi}}{\sqrt{2\pi}}\sqrt{|qB|} \begin{pmatrix} C_1 I_{n,a}(\rho)e^{-i\phi/2}\\ iC_2 I_{n-1,a}(\rho)e^{i\phi/2}\\ C_3 I_{n,a}(\rho)e^{-i\phi/2}\\ iC_4 I_{n-1,a}(\rho)e^{i\phi/2} \end{pmatrix} \tag*{(3)}

where n=0,1,2,3,…n=0,1,2,3,\ldots is the principal quantum number, the magnetic quantum number mm is the eigenvalue of the zz-component of the total angular momentum, a=n+m−1/2=0,1,2,3,…a=n+m-1/2=0,1,2,3,\ldots, and we defined the Laguerre functions

In,a(ρ)=a!n!e−ρ/2ρn−a2Lan−a(ρ),(4)I_{n,a}(\rho)=\sqrt{\frac{a!}{n!}}e^{-\rho/2}\rho^{\frac{n-a}{2}}L_a^{n-a}(\rho), \tag*{(4)}

with ρ=∣qB∣r2/2\rho=|qB|r^2/2 and Lan−a(ρ)L_a^{n-a}(\rho) the Laguerre polynomials. The state (3) is an eigenvector of the Hamiltonian with energy

En,pz=2n∣qB∣+pz2.E_{n,p_z}=\sqrt{2n|qB|+p_z^2}.

The coefficients C1,…,C4C_1,\ldots,C_4 can be fixed by requiring that the state (3) is also an eigenvector of a chosen spin operator and that it is normalized according to

∫d3x ψ∞†γ0ψ∞=∑i=14Ci2=1.(5)\int d^3x\,\psi_\infty^\dagger\gamma^0\psi_\infty=\sum_{i=1}^{4}C_i^2=1. \tag*{(5)}

As the spin operator we choose the zz-component of the magnetic spin μ^z\hat{\mu}_z defined as [36]

μ^=Σ−iγ0γ52Σ×(p−qA).(6)\hat{\mu}=\boldsymbol{\Sigma}-\frac{i\gamma^0\gamma^5}{2}\boldsymbol{\Sigma}\times(\mathbf{p}-q\mathbf{A}). \tag*{(6)}

It is convenient for imposing the boundary conditions, as we show in the next section. Its eigenstates are referred to as the transverse polarizations. Solving the secular equation

μ^zψ∞=ζ(E−Ωm)2−pz2 ψ∞,(7)\hat{\mu}_z\psi_\infty=\zeta\sqrt{(E-\Omega m)^2-p_z^2}\,\psi_\infty, \tag*{(7)}

with ζ=±\zeta=\pm, and imposing (5), we obtain

C1,3=−122(A±+ζA∓),C2,4=122(A∓+ζA±),(8)C_{1,3}=-\frac{1}{2\sqrt{2}}(A^\pm+\zeta A^\mp),\qquad C_{2,4}=\frac{1}{2\sqrt{2}}(A^\mp+\zeta A^\pm), \tag*{(8)}

with the upper signs referring to the indexes 1, 2 and the lower ones to 3, 4, and

A±=1±pzE.A^\pm=\sqrt{1\pm\frac{p_z}{E}}.

To compute the CME it is convenient to split the state into the right- and left-handed parts

ψχ,∞=1+χγ52ψ∞,χ=±=R/L.(9)\psi_{\chi,\infty}=\frac{1+\chi\gamma^5}{2}\psi_\infty,\qquad\chi=\pm=R/L. \tag*{(9)}

where χ\chi denotes the chirality of the state. The chiral state is of the same form as in Eq. (3) but replacing the coefficients C1,…,C4C_1,\ldots,C_4 with

C1,3χ=−122(1+χ2A+±ζ1−χ2A−),C^{\chi}_{1,3}=-\frac{1}{2\sqrt{2}}\left(\frac{1+\chi}{2}A^{+}\pm\zeta\frac{1-\chi}{2}A^{-}\right),
C2,4χ=122(1+χ2A−∓ζ1−χ2A+).C^{\chi}_{2,4}=\frac{1}{2\sqrt{2}}\left(\frac{1+\chi}{2}A^{-}\mp\zeta\frac{1-\chi}{2}A^{+}\right).

Considering only the particle part, the Dirac field with a definite chirality is

ψ^χ,∞(x)=∑p^ψχ,∞(x)a^χ(p^)(10)\widehat{\psi}_{\chi,\infty}(x)=\sum_{\widehat{p}}\psi_{\chi,\infty}(x)\widehat{a}_{\chi}(\widehat{p}) \tag*{(10)}

where p^={n,m,pz}\widehat{p}=\{n,m,p_z\}. The thermodynamic equilibrium with a chiral imbalance μA=μR−μL\mu_A=\mu_R-\mu_L is described with the density operator

ρ^=1Ze−(βH^−βμRQ^R−βμLQ^L).(11)\widehat{\rho}=\frac{1}{Z}e^{-\left(\beta\widehat{H}-\beta\mu_R\widehat{Q}_R-\beta\mu_L\widehat{Q}_L\right)}. \tag*{(11)}

Since we are considering a free-field, it readily follows that

⟨a^χ′†(p^′)a^χ(p^)⟩=δχ′,χδp^′,p^nF(Ep^−μχ)(12)\left\langle\widehat{a}^{\dagger}_{\chi'}(\widehat{p}')\widehat{a}_{\chi}(\widehat{p})\right\rangle=\delta_{\chi',\chi}\delta_{\widehat{p}',\widehat{p}}n_F(E_{\widehat{p}}-\mu_{\chi}) \tag*{(12)}

where nFn_F is the Fermi-Dirac distribution function with temperature T=1/βT=1/\beta

nF(x)=1eβx+1.(13)n_F(x)=\frac{1}{e^{\beta x}+1}. \tag*{(13)}

The CME is obtained evaluating the thermal expectation value of the electric current.

We first compute the thermal expectation value of the right and left currents. The electric current is then obtained from the right and left currents as a simple sum: j^=j^R+j^L\widehat{j}=\widehat{j}_R+\widehat{j}_L. The thermal expectation value of the particle contribution of the right current is

⟨j^Rμ(x)⟩+=tr⁡[ρ^j^R+μ(x)]=∑p^nF(Ep^−μR)jRμ(x∣p^)=∑n,a=0∞∫−∞∞dpz nF(Ep^−μR)jRμ(x∣p^),(14)\begin{aligned} \left\langle\widehat{j}^{\mu}_R(x)\right\rangle_{+}=\operatorname{tr}\left[\widehat{\rho}\widehat{j}^{\mu}_{R+}(x)\right] \\ &=\sum_{\widehat{p}}n_F(E_{\widehat{p}}-\mu_R)j^{\mu}_R(x|\widehat{p}) \\ &=\sum_{n,a=0}^{\infty}\int_{-\infty}^{\infty}\mathrm{d}p_z\,n_F(E_{\widehat{p}}-\mu_R)j^{\mu}_R(x|\widehat{p}), \tag*{(14)} \end{aligned}

where the matrix element is obtained with

jRμ(x∣p^)=qψ‾R,∞(x)γμψR,∞(x).(15)j^{\mu}_R(x|\widehat{p})=q\overline{\psi}_{R,\infty}(x)\gamma^{\mu}\psi_{R,\infty}(x). \tag*{(15)}

The antiparticle contribution can be added at the end of the calculation by changing the sign of the chemical potential in the particle contribution. Using the solution of the Dirac equation above, the component along the magnetic field is

jRz(x∣n,a,pz)=q2B2π2[C1RC3RIn,a2(ρ)−C2RC4RIn−1,a2(ρ)].(16)j^z_R(x|n,a,p_z)=\frac{q^2B}{2\pi^2}\left[C^R_1C^R_3I^2_{n,a}(\rho)-C^R_2C^R_4I^2_{n-1,a}(\rho)\right]. \tag*{(16)}

with

C1,3R=1221+pzEn,pz,C2,4R=1221−pzEn,pz.(17)C^R_{1,3}=\frac{1}{2\sqrt{2}}\sqrt{1+\frac{p_z}{E_{n,p_z}}},\qquad C^R_{2,4}=\frac{1}{2\sqrt{2}}\sqrt{1-\frac{p_z}{E_{n,p_z}}}. \tag*{(17)}

Considering that the odd terms in pzp_z are vanishing once integrated, and there is no ζ\zeta dependence, we have

⟨j^R3(x)⟩+=q2B8π2∑n,a=0∞∫−∞∞dpz nF(En,pz−μR)×[In,a2(ρ)−In−1,a2(ρ)].\begin{aligned} \left\langle\widehat{j}^3_R(x)\right\rangle_{+}=\frac{q^2B}{8\pi^2}\sum_{n,a=0}^{\infty}\int_{-\infty}^{\infty}\mathrm{d}p_z\,n_F(E_{n,p_z}-\mu_R) \\ &\times\left[I^2_{n,a}(\rho)-I^2_{n-1,a}(\rho)\right]. \end{aligned}

Including the antiparticles we obtain

⟨j^R3(x)⟩=q2B4π2∑n,a=0∞[In,a2(ρ)−In−1,a2(ρ)]×∫0∞dpz[nF(En,pz−μR)−nF(En,pz+μR)].\begin{aligned} \left\langle\widehat{j}^3_R(x)\right\rangle=\frac{q^2B}{4\pi^2}\sum_{n,a=0}^{\infty}\left[I^2_{n,a}(\rho)-I^2_{n-1,a}(\rho)\right] \\ &\times\int_{0}^{\infty}\mathrm{d}p_z\left[n_F(E_{n,p_z}-\mu_R)-n_F(E_{n,p_z}+\mu_R)\right]. \end{aligned}

The left current is obtained in the same fashion, and the electric current is the sum of the two currents

⟨j^3(x)⟩=q2B4π2∑n,a=0∞[In,a2(ρ)−In−1,a2(ρ)]In(T,μ,μA),(18)\left\langle\widehat{j}^3(x)\right\rangle=\frac{q^2B}{4\pi^2}\sum_{n,a=0}^{\infty}\left[I^2_{n,a}(\rho)-I^2_{n-1,a}(\rho)\right]\mathcal{I}_n(T,\mu,\mu_A), \tag*{(18)}

where

In=∫0∞dpz[nF(En,pz−μR)−nF(En,pz+μR)−nF(En,pz−μL)+nF(En,pz+μL)].(19)\mathcal{I}_n=\int_{0}^{\infty}\mathrm{d}p_z\left[n_F(E_{n,p_z}-\mu_R)-n_F(E_{n,p_z}+\mu_R) -n_F(E_{n,p_z}-\mu_L)+n_F(E_{n,p_z}+\mu_L)\right]. \tag*{(19)}

In view of the exponential falloff of the Fermi-Dirac distribution one can prove that In\mathcal{I}_n is finite and that the sum over nn converges absolutely. Therefore, we can first sum over aa and then over nn. Using the following properties of the II functions (see, for instance, [36]):

I−1,a(ρ)=0,∑aIn,a2(ρ)=1,I_{-1,a}(\rho)=0,\qquad\sum_a I^2_{n,a}(\rho)=1,

and noticing that In\mathcal{I}_n does not depend on aa, it follows that

⟨j^3(x)⟩=q2B4π2∑n,a=0∞[In,a2(ρ)−In−1,a2(ρ)]In=q2B4π2∑a=0∞δn,0[In,a2(ρ)−In−1,a2(ρ)]In+q2B4π2∑n=1∞∑a=0∞[In,a2(ρ)−In−1,a2(ρ)]In=q2B4π2I0∑a=0∞I0,a2(ρ)+q2B4π2∑n=1∞In[∑a=0∞In,a2(ρ)−∑a=0∞In−1,a2(ρ)]=q2B4π2I0+q2B4π2∑n=1∞In[1−1]=q2B4π2I0=q2B2π2μA.\begin{aligned} \langle\hat{j}^{3}(\boldsymbol{x})\rangle &=\frac{q^{2}B}{4\pi^{2}}\sum_{n,a=0}^{\infty}\left[I_{n,a}^{2}(\rho)-I_{n-1,a}^{2}(\rho)\right]\mathcal{I}_{n}\\ &=\frac{q^{2}B}{4\pi^{2}}\sum_{a=0}^{\infty}\delta_{n,0}\left[I_{n,a}^{2}(\rho)-I_{n-1,a}^{2}(\rho)\right]\mathcal{I}_{n}\\ &\quad+\frac{q^{2}B}{4\pi^{2}}\sum_{n=1}^{\infty}\sum_{a=0}^{\infty}\left[I_{n,a}^{2}(\rho)-I_{n-1,a}^{2}(\rho)\right]\mathcal{I}_{n}\\ &=\frac{q^{2}B}{4\pi^{2}}\mathcal{I}_{0}\sum_{a=0}^{\infty}I_{0,a}^{2}(\rho)\\ &\quad+\frac{q^{2}B}{4\pi^{2}}\sum_{n=1}^{\infty}\mathcal{I}_{n}\left[\sum_{a=0}^{\infty}I_{n,a}^{2}(\rho)-\sum_{a=0}^{\infty}I_{n-1,a}^{2}(\rho)\right]\\ &=\frac{q^{2}B}{4\pi^{2}}\mathcal{I}_{0}+\frac{q^{2}B}{4\pi^{2}}\sum_{n=1}^{\infty}\mathcal{I}_{n}\left[1-1\right]\\ &=\frac{q^{2}B}{4\pi^{2}}\mathcal{I}_{0}=\frac{q^{2}B}{2\pi^{2}}\mu_{A}. \end{aligned}

where we used I0=I0(T,μ,μA)=2μA\mathcal{I}_{0}=\mathcal{I}_{0}(T,\mu,\mu_{A})=2\mu_{A}. This is how the CME (1) is derived from the exact solutions of the Dirac equation. To include the finite-volume effects we follow the same procedure as described in this section but impose the boundary conditions.

Solution of the Dirac equation in a finite cylinder

In this section, we solve the Dirac equation for a free massless fermion in a constant homogeneous magnetic field inside a cylinder of radius RR with the MIT boundary conditions. The Dirac equation is the same as in the unbound case, but the requirements on the radial part of the solutions are different. In the unbound case, one has to require that the wave function is regular at the origin ρ=0\rho=0 and vanishing at infinity ρ→∞\rho\to\infty. In a cylinder of finite radius RR the wave function does not have to vanish at infinity. Rather, we impose the boundary condition at r=Rr=R. Solutions to the Dirac equation that are regular at the origin r=0r=0 have the general form [21, 22]

ψ=e−iEλ,pzteipzz2πeimϕ2π∣qB∣(C1F1(ρ)e−iϕ/2iC2F2(ρ)eiϕ/2C3F1(ρ)e−iϕ/2iC4F2(ρ)eiϕ/2).(20)\psi=e^{-iE_{\lambda,p_z}t}\frac{e^{ip_zz}}{\sqrt{2\pi}}\frac{e^{im\phi}}{\sqrt{2\pi}}\sqrt{|qB|} \begin{pmatrix} C_{1}F_{1}(\rho)e^{-i\phi/2}\\ iC_{2}F_{2}(\rho)e^{i\phi/2}\\ C_{3}F_{1}(\rho)e^{-i\phi/2}\\ iC_{4}F_{2}(\rho)e^{i\phi/2} \end{pmatrix}. \tag*{(20)}

where F1,2F_{1,2} are given in terms of the confluent hypergeometric function:

F1(ρ)=N1Γ(m+12)e−ρ/2ρm2−14 1F1(−λ,m+12;ρ),F_{1}(\rho)=\frac{N_{1}}{\Gamma\left(m+\frac{1}{2}\right)}e^{-\rho/2}\rho^{\frac{m}{2}-\frac{1}{4}}\,{}_1F_{1}\left(-\lambda,m+\frac{1}{2};\rho\right),
F2(ρ)=N2Γ(m+32)e−ρ/2ρm2+14 1F1(−λ+1,m+32;ρ).(21)F_{2}(\rho)=\frac{N_{2}}{\Gamma\left(m+\frac{3}{2}\right)}e^{-\rho/2}\rho^{\frac{m}{2}+\frac{1}{4}}\,{}_1F_{1}\left(-\lambda+1,m+\frac{3}{2};\rho\right). \tag*{(21)}

and N1N_{1} and N2N_{2} are such that

∫0ρRdρ [F1(ρ)]2=1,∫0ρRdρ [F2(ρ)]2=1.(22)\int_{0}^{\rho_{R}}d\rho\,[F_{1}(\rho)]^{2}=1,\qquad\int_{0}^{\rho_{R}}d\rho\,[F_{2}(\rho)]^{2}=1. \tag*{(22)}

In an unbound case, the requirement that the wave function is vanishing at infinity is fulfilled only if nn and a=n+m−1/2a=n+m-1/2 are non-negative integers. In that case the hypergeometric functions reduce to the Laguerre functions (4) and we recover the unbound solutions (3). In a finite cylinder the principal quantum number nn does not have to be an integer. To stress this point, from now on, we denote the principal quantum number as λ\lambda. The magnetic quantum number mm is not affected by the boundary and runs over all semi-integer values. The energy in a finite cylinder is

Eλ,pz=2λ∣qB∣+pz2,(23)E_{\lambda,p_z}=\sqrt{2\lambda|qB|+p_z^{2}}, \tag*{(23)}

with the values of λ\lambda to be determined according to the boundary condition.

At first glance it may seem that the simplest boundary condition is the Cauchy one that requires vanishing of the wave function (20) on the surface ρ=ρR=∣qB∣R2/2\rho=\rho_{R}=|qB|R^{2}/2. However, it is easy to realize, that for a finite ρR\rho_{R} the radial wave functions in Eq. (21) cannot both be set to zero for the same value of λ\lambda and mm. A more elaborate boundary condition must be used. In this work we use the MIT boundary conditions which allow us to keep the system in a finite region without including an (explicit) confining potential in the Hamiltonian. In this way the fields remain free.

The coefficients C1,…,C4C_{1},\ldots,C_{4} in the transverse polarization are obtained as described in the previous section:

C1,3=12N1N2λN12+N22A+±ζA−N1,C_{1,3}=\frac{1}{2}\frac{N_{1}N_{2}}{\sqrt{\lambda N_{1}^{2}+N_{2}^{2}}}\frac{A_{+}\pm\zeta A_{-}}{N_{1}},
C2,4=12N1N2λN12+N22A−∓ζA+N2λ,(24)C_{2,4}=\frac{1}{2}\frac{N_{1}N_{2}}{\sqrt{\lambda N_{1}^{2}+N_{2}^{2}}}\frac{A_{-}\mp\zeta A_{+}}{N_{2}}\sqrt{\lambda}, \tag*{(24)}

where ζ=±\zeta=\pm and

A±=1±pzEλ,pz.(25)A_{\pm}=\sqrt{1\pm\frac{p_{z}}{E_{\lambda,p_z}}}. \tag*{(25)}

The exact same coefficients are obtained if we replace the condition in Eq. (7) with the requirement that

C4C1=−C3C2.(26)\frac{C_{4}}{C_{1}}=-\frac{C_{3}}{C_{2}}. \tag*{(26)}

As we show below, this is a necessary condition for a state to satisfy the MIT boundary condition. Therefore, the MIT boundary condition can be applied directly to states in the transverse polarization. Different spin states can be chosen, such as helicity states, but in that case, only a specific linear combination of spin states can satisfy the MIT boundary condition [17].

The MIT boundary condition

The MIT bag model [25, 27] allows us to find solutions of the Dirac equation in a bound region without introducing discontinuities in the wave function and preserving the conserved quantities of the system, such as energy, momentum and electric charge, thanks to the vanishing of the fluxes of the corresponding currents at the boundary. Ultimately, all these features are consequences of the self-adjointness of the Hamiltonian with respect to the Dirac inner product [17, 30].

Given any two solutions of the Dirac equation (ψ,Φ)(\psi, \Phi), the Dirac inner product is defined as

⟨ψ,Φ⟩=∫Vd3x ψˉ(x)γ0Φ(x).(27)\langle\psi, \Phi\rangle= \int_{V} \mathrm{d}^{3}x\,\bar{\psi}(x)\gamma^{0}\Phi(x). \tag*{(27)}

The Hamiltonian is a self-adjoint operator if

⟨ψ,H^Φ⟩=⟨H^ψ,Φ⟩,(28)\langle\psi, \hat{H}\Phi\rangle= \langle\hat{H}\psi, \Phi\rangle, \tag*{(28)}

or, equivalently, for H^=i∂t\hat{H}=i\partial_{t}, when the inner product is time independent

∂t⟨ψ,Φ⟩=0.(29)\partial_{t}\langle\psi,\Phi\rangle=0. \tag*{(29)}

Integrating by parts, this condition can be written as an integral over the boundary of the volume

−∫∂VdΣi ψˉ(x)γiΦ(x)=0,(30)-\int_{\partial V}\mathrm{d}\Sigma_{i}\,\bar{\psi}(x)\gamma^{i}\Phi(x)=0, \tag*{(30)}

which in the case of a cylinder of radius RR reads

R∫0∞dz∫02πdϕ ψˉ(R)γrΦ(R)=0,∂r(ψˉΦ)(R)=−Θ(R),(31)\begin{aligned} R\int_{0}^{\infty}\mathrm{d}z\int_{0}^{2\pi}\mathrm{d}\phi\,\bar{\psi}(R)\gamma^{r}\Phi(R)&=0,\\ \partial_{r}(\bar{\psi}\Phi)(R)&=-\Theta(R), \tag*{(31)} \end{aligned}

where ψ(R)=ψ(z,ϕ,r=R)\psi(R)=\psi(z,\phi,r=R), Θ(R)<∞\Theta(R)<\infty and γr=n^ ⁣ ⁣ ⁣/ =n^μγμ=r^μγμ\gamma^{r}=\hat{n}\!\!\!/\ =\hat{n}_{\mu}\gamma^{\mu}=\hat{r}_{\mu}\gamma^{\mu}, with n^=r^\hat{n}=\hat{r} the radial direction in cylindrical coordinates.

In order to have a self-adjoint Hamiltonian we require that any solution ψ\psi of the Dirac equation is also solving the MIT boundary condition

iγrψ(R)=−ψ(R),(32)i\gamma^{r}\psi(R)=-\psi(R), \tag*{(32)}

Indeed, this condition and its conjugate iψˉ(R)γr=ψˉ(R)i\bar{\psi}(R)\gamma^{r}=\bar{\psi}(R) imply [25, 27]

ψˉ(R)Φ(R)=ψˉ(R)[−iγrΦ(R)]=−[iψˉ(R)γr]Φ(R)=−ψˉ(R)Φ(R)=0,(33)\begin{aligned} \bar{\psi}(R)\Phi(R)&=\bar{\psi}(R)[-i\gamma^{r}\Phi(R)]\\ &=-[i\bar{\psi}(R)\gamma^{r}]\Phi(R)=-\bar{\psi}(R)\Phi(R)=0, \tag*{(33)} \end{aligned}

from which it follows that the integrand of Eq. (31) is vanishing at any point z,ϕz,\phi:

ψˉ(R)γrΦ(R)=−iψˉ(R)Φ(R)=0.(34)\bar{\psi}(R)\gamma^{r}\Phi(R)=-i\bar{\psi}(R)\Phi(R)=0. \tag*{(34)}

Using the MIT boundary conditions is not the only way one can require the vanishing of Eq. (31). In [21, 22] this is ensured requiring that

F1(λ,m,ρR)F2(λ′,m,ρR)=0(35)F_{1}(\lambda,m,\rho_{R})F_{2}(\lambda',m,\rho_{R})=0 \tag*{(35)}

for arbitrary values of mm, λ\lambda, and λ′\lambda', which is the generalization to the case of the magnetic field of the spectral boundary conditions discussed in [17].

From the self-adjointness of the Hamiltonian we can also show that the normalized eigenvectors of the Hamiltonian are orthogonal with respect to the Dirac inner product. Take two eigenvectors of the Hamiltonian Φ\Phi and ψ\psi with quantum numbers p^={λ,m,pz,ζ}\hat{p}=\{\lambda,m,p_{z},\zeta\} and p^′={λ′,m′,pz′,ζ′}\hat{p}'=\{\lambda',m',p'_{z},\zeta'\} respectively. Using

∂t⟨ψ(t),Φ(t)⟩=∂t[e−i(Eλ−Eλ′)⟨ψ(t=0),Φ(t=0)⟩]=−i(Eλ−Eλ′)⟨ψ(t),Φ(t)⟩=0,(36)\begin{aligned} \partial_{t}\langle\psi(t),\Phi(t)\rangle&=\partial_{t}\left[e^{-i(E_{\lambda}-E_{\lambda'})}\langle\psi(t=0),\Phi(t=0)\rangle\right]\\ &=-i(E_{\lambda}-E_{\lambda'})\langle\psi(t),\Phi(t)\rangle=0, \tag*{(36)} \end{aligned}

it follows that when Eλ≠Eλ′E_{\lambda}\ne E_{\lambda'} we must have ⟨ψ,Φ⟩=0.\langle\psi,\Phi\rangle=0.

The integrals over zz and ϕ\phi immediately give

⟨ψ,Φ⟩=δ(pz−pz′)δm,m′∫0ρRdρ e−i(Eλ−Eλ′)[H1F1F1′+H2F2F2′](37)\langle\psi,\Phi\rangle=\delta(p_{z}-p'_{z})\delta_{m,m'}\int_{0}^{\rho_{R}}\mathrm{d}\rho\,e^{-i(E_{\lambda}-E_{\lambda'})}\left[H_{1}F_{1}F'_{1}+H_{2}F_{2}F'_{2}\right] \tag*{(37)}

where

H1=N2N2′λN12+N22λ′N1′2+N2′2A+A+′+ζζ′A−A−′2,H_{1}=\frac{N_{2}N'_{2}}{\sqrt{\lambda N_{1}^{2}+N_{2}^{2}}\sqrt{\lambda'N_{1}'^{2}+N_{2}'^{2}}}\frac{A_{+}A'_{+}+\zeta\zeta'A_{-}A'_{-}}{2},
H2=λλ′N1N1′λN12+N22λ′N1′2+N2′2A−A−′+ζζ′A+A+′2.(38)H_{2}=\frac{\sqrt{\lambda\lambda'}N_{1}N'_{1}}{\sqrt{\lambda N_{1}^{2}+N_{2}^{2}}\sqrt{\lambda'N_{1}'^{2}+N_{2}'^{2}}}\frac{A_{-}A'_{-}+\zeta\zeta'A_{+}A'_{+}}{2}. \tag*{(38)}

From what we showed above the remaining integral must give δj,j′δζ,ζ′\delta_{j,j'}\delta_{\zeta,\zeta'}. Because the wave functions have been normalized, we have

⟨ψ,Φ⟩=∫Vd3x ψˉp^γ0Φp^=δ(pz−pz′)δm,m′δζ,ζ′δj,j′.(39)\langle\psi,\Phi\rangle=\int_{V}\mathrm{d}^{3}x\,\bar{\psi}_{\hat{p}}\gamma^{0}\Phi_{\hat{p}}=\delta(p_{z}-p'_{z})\delta_{m,m'}\delta_{\zeta,\zeta'}\delta_{j,j'}. \tag*{(39)}

We already found that the general form of the solution of the Dirac equation is given by (20). Imposing the condition (32) to the general solution of the Dirac equation (20), we obtain

{C1F1(ρR)−C4F2(ρR)=0,C3F1(ρR)+C2F2(ρR)=0,(40)\begin{cases} C_{1}F_{1}(\rho_{R})-C_{4}F_{2}(\rho_{R})=0,\\ C_{3}F_{1}(\rho_{R})+C_{2}F_{2}(\rho_{R})=0, \tag*{(40)} \end{cases}

whence

F1(ρR)=C4C1F2(ρR)=−C2C3F2(ρR).(41)F_{1}(\rho_{R})=\frac{C_{4}}{C_{1}}F_{2}(\rho_{R})=-\frac{C_{2}}{C_{3}}F_{2}(\rho_{R}). \tag*{(41)}

where ρR\rho_{R} is defined in (2). The effect of the boundary on the energy spectrum increases as ρR\rho_{R} decreases. Since (41) coincides with (26), our choice of the transverse polarization basis is the most convenient, as mentioned above.

If F1F_{1} or F2F_{2} vanish identically, we cannot impose the MIT boundary condition. In particular, this is the case for λ=0\lambda=0 because then F2=0F_{2}=0. In the unbounded case the CME is caused entirely by the lowest Landau Level (LLL) n=0n=0 which is also the chiral state. With the MIT boundary condition the LLL is not realized for λ=0\lambda= 0, and all eigenstates are not chiral. Therefore, at this point of the analysis, it is not clear whether the CME can also be realized in a finite region with this type of boundary condition. Solutions with λ\lambda very close to zero but still positive are instead allowed, and the new ground state replacing the lowest Landau level is discussed below.

As we already indicated, the boundary condition can only be satisfied by states such that C2/C3=−C4/C1C_2/C_3 = -C_4/C_1, which is always the case for the transverse polarization. Therefore, using states in the transverse polarization with coefficients (24), the MIT boundary condition becomes

F1(ρR)N1=ζλF2(ρR)N2.(42)\frac{F_1(\rho_R)}{N_1} = \zeta\sqrt{\lambda}\frac{F_2(\rho_R)}{N_2}. \tag*{(42)}

Solving this equation for λ\lambda gives the energy spectrum (23) and the whole set of normal modes (20). Equation (42) has an infinite countable number of roots {λi}\{\lambda_i\}, i=1,2,3,…i = 1,2,3,\ldots, where λi<λi+1\lambda_i < \lambda_{i+1}. These roots generally depend on mm, ζ\zeta, and ρR\rho_R. As ρR\rho_R approaches infinity the roots approach the values λ=n=0,1,2,…\lambda= n = 0,1,2,\ldots independently of mm and ζ\zeta.

It is instructive to find the analytic value of the principal quantum number λ\lambda corresponding to the LLL as the radius of the cylinder is increased. As said, when increasing the parameter ρR\rho_R, λ\lambda becomes an integer and one can check that the functions F1F_1 and F2F_2 vanish for a<0a < 0. For λ≪1\lambda\ll1, this implies that m≥+1/2m \ge+1/2. In this case, we can expand the constraint (42) in the power series of λ\lambda and obtain the equation

λ=ζe−ρRρRmΓ(m+12)−Γ(m+12,ρR).(43)\sqrt{\lambda} = \zeta\frac{e^{-\rho_R}\rho_R^m}{\Gamma\left(m+\frac{1}{2}\right)-\Gamma\left(m+\frac{1}{2},\rho_R\right)}. \tag*{(43)}

It has a solution only when ζ=+1\zeta= +1:

λLLL≈e−2ρRρR2mΓ(m+12)2.(44)\lambda_{\mathrm{LLL}} \approx\frac{e^{-2\rho_R}\rho_R^{2m}}{\Gamma\left(m+\frac{1}{2}\right)^2}. \tag*{(44)}

This solution is finite, even though it approaches zero as the radius is increased. This is different from the unbound case where the LLL λ=0\lambda= 0 exists for both polarizations ζ=±\zeta= \pm. Nevertheless, the unbound CME is recovered in the large radius limit. This occurs because the energy levels are weighted more in the bound case compared to the unbound one. Indeed, we can compute analytically the normalization constants N1N_1 and N2N_2 for λ=0\lambda= 0 and derive that in the infinite cylinder limit, the bound modes become

lim⁡ρR→∞ψa>0,ζ={ψζ,∞,λ≥1,0,λ=0, ζ=−1,2ψλ=0,a,ζ=+1,∞,λ=0, ζ=+1.\lim_{\rho_R\to\infty}\psi_{a>0,\zeta} = \begin{cases} \psi_{\zeta,\infty}, & \lambda\ge1,\\ 0, & \lambda= 0,\ \zeta= -1,\\ \sqrt{2}\psi_{\lambda=0,a,\zeta=+1,\infty}, & \lambda= 0,\ \zeta= +1. \end{cases}

where ψζ,∞\psi_{\zeta,\infty} are the unbound solutions given in Eq. (3).

To obtain the energy spectrum, we numerically compute the roots {λk}\{\lambda_k\} of (42) at fixed ρR\rho_R, ζ\zeta, and mm. Figure 1 shows the energy levels

Figure 1. The energy levels (45) as a function of mm for different values of ρR\rho R at β2∣qB∣=0.1\beta^{2}|qB|=0.1, and pz=0p_z=0. Different markers correspond to the ii’th root of Eq. (42) as indicated in the legends. Solid markers correspond to positive polarization ζ=+1\zeta=+1, and open markers to ζ=−1\zeta=-1. Solid lines are the energy levels of the unbound case, βEi,m∞=θ(a)2(i−1)β2∣qB∣\beta E_{i,m}^{\infty}=\theta(a)\sqrt{2(i-1)\beta^{2}|qB|}, corresponding to the Landau levels plotted in the same panel. The ordinate range varies for different panels.

βEn,i=2λiβ2∣qB∣.(45)\beta E_{n,i} = \sqrt{2\lambda_i\beta^2|qB|}. \tag*{(45)}

with pz=0p_z = 0 and β2∣qB∣=0.1\beta^2|qB| = 0.1, where β\beta is a length scale, for instance the inverse temperature, and ii denotes the ii-th root of λ\lambda. These results are in agreement with those of [21, 22]; see also [37] for a detailed discussion, where solutions of free fermions in a magnetic field in a finite cylinder were presented with the boundary condition (35).

Thermodynamics in a finite cylinder

In this section, we show that the energy, momentum, electric and axial charge are conserved inside the cylinder when considering solutions of the Dirac equation with the MIT boundary condition. This ensures that the equilibrium density matrix describes the thermal ensemble of the system and we can use statistical mechanics to evaluate the chiral magnetic effect as done in the case of the unbound region.

The equilibrium thermal state of a system is described by the equilibrium statistical operator

ρ^=1Zexp⁡{−H/T+μQ^/T+μAQ^A/T}.(46)\hat{\rho} = \frac{1}{Z}\exp\left\{-H/T+\mu\widehat{Q}/T+\mu_A\widehat{Q}_A/T\right\}. \tag*{(46)}

provided that the following quantities are conserved:

P^ν=∫Vd3x T^0ν,Q^=∫Vd3x j^0,\widehat{P}^{\nu}=\int_V d^3x\,\widehat{T}^{0\nu},\qquad\widehat{Q}=\int_V d^3x\,\widehat{j}^{0},
Q^A=∫Vd3x j^A0.(47)\widehat{Q}_A=\int_V d^3x\,\widehat{j}_A^{0}. \tag*{(47)}

which are, respectively, the total four-momentum, the electric charge and the axial charge. Even if the region is finite, these quantities are conserved if the fluxes at the boundary vanish

∫∂VdS T^νin^i=0,∫∂VdS j^in^i=0,\int_{\partial V}dS\,\widehat{T}^{\nu i}\widehat{n}_i=0,\qquad\int_{\partial V}dS\,\widehat{j}^{i}\widehat{n}_i=0,
∫∂VdS j^Ain^i=0.(48)\int_{\partial V}dS\,\widehat{j}_A^{i}\widehat{n}_i=0. \tag*{(48)}

In the case of a cylinder, n^\widehat{n} is the radial direction.

We now explicitly show that the MIT boundary condition ensures that all the fluxes are vanishing, not just the electric current. We first remind the reader that the MIT boundary condition implies

in ⁣ ⁣ ⁣/ψ(R)=−ψ(R),ψ‾ψ(R)=0,∂r(ψ‾ψ)(R)=−Θ(R).(49)i n\!\!\!/\psi(R)=-\psi(R),\qquad\overline{\psi}\psi(R)=0,\qquad\partial_r\left(\overline{\psi}\psi\right)(R)=-\Theta(R). \tag*{(49)}

From the general form of the solutions (20), we also have that

ψ‾γ5ψ=0.(50)\overline{\psi}\gamma^5\psi=0. \tag*{(50)}

Using these relations it is easy to prove that the electric current is conserved:

n^μj^μ(R)=ψ‾(R)n^ ⁣ ⁣ ⁣/ψ(R)=iψ‾ψ(R)=0.(51)\widehat{n}_\mu\widehat{j}^{\mu}(R)=\overline{\psi}(R)\widehat{n}\!\!\!/\psi(R)=i\overline{\psi}\psi(R)=0. \tag*{(51)}

Similarly, for the axial current we obtain

n^μjAμ(R)=ψˉ(R)n^γ5ψ(R)=−ψˉ(R)γ5n^ψ(R)=−iψˉγ5ψ(R)=0.(52)\hat n_\mu j_A^\mu(R)=\bar\psi(R)\hat n\gamma^5\psi(R)=-\bar\psi(R)\gamma^5\hat n\psi(R) =-i\bar\psi\gamma^5\psi(R)=0. \tag*{(52)}

Finally, we also check that the flux of energy and momentum is vanishing. Using the canonical energy-momentum tensor of a Dirac field interacting with an external gauge field we have:

∫∂VdS Tμνn^μ=∫∂VdS[i2(ψˉn^∂νψ−∂νψˉn^ψ)−ψˉn^Aνψ]=∫∂VdS[12(ψˉ∂νψ+∂νψˉψ)−iψˉψ(R)Aν]=12∫∂VdS ∂ν(ψˉψ)=12Θ(R)∫∂VdS n^ν=0,(53)\begin{aligned} \int_{\partial V}dS\,T^{\mu\nu}\hat n_\mu=\int_{\partial V}dS\left[\frac{i}{2}\left(\bar\psi\hat n\partial^\nu\psi-\partial^\nu\bar\psi\hat n\psi\right)-\bar\psi\hat n A^\nu\psi\right] \\ &=\int_{\partial V}dS\left[\frac{1}{2}\left(\bar\psi\partial^\nu\psi+\partial^\nu\bar\psi\psi\right)-i\bar\psi\psi(R)A^\nu\right] \\ &=\frac{1}{2}\int_{\partial V}dS\,\partial^\nu(\bar\psi\psi)=\frac{1}{2}\Theta(R)\int_{\partial V}dS\,\hat n^\nu=0, \tag*{(53)} \end{aligned}

where we used ψˉψ(R)=0\bar\psi\psi(R)=0.

The chiral magnetic effect in a finite cylinder

The CME is obtained by evaluating the thermal average of the electric current with the statistical operator (46), as done in Sec. II. This time we use the bound solutions (20) in the transverse polarization (24), and satisfy the MIT boundary conditions (42). We first obtain the thermal expectation value of the right-handed current. The right-handed projection of the solution of the Dirac equation with the MIT boundary condition is

ψRζ=e−iϵEteipzz2πeimϕ2π∣qB∣2N1N2λN12+N22(1+pzEλ,pzF1(ρ)N1e−iϕ/2i1−pzEλ,pzλF2(ρ)N2e+iϕ/21+pzEλ,pzF1(ρ)N1e−iϕ/2i1−pzEλ,pzλF2(ρ)N2eiϕ/2).(54)\psi_{R\zeta}=e^{-i\epsilon_E t}\frac{e^{ip_z z}}{\sqrt{2\pi}}\frac{e^{im\phi}}{\sqrt{2\pi}}\frac{\sqrt{|qB|}}{2}\frac{N_1N_2}{\sqrt{\lambda N_1^2+N_2^2}} \begin{pmatrix} \sqrt{1+\frac{p_z}{E_{\lambda,p_z}}}\frac{F_1(\rho)}{N_1}e^{-i\phi/2}\\ i\sqrt{1-\frac{p_z}{E_{\lambda,p_z}}}\frac{\sqrt{\lambda}F_2(\rho)}{N_2}e^{+i\phi/2}\\ \sqrt{1+\frac{p_z}{E_{\lambda,p_z}}}\frac{F_1(\rho)}{N_1}e^{-i\phi/2}\\ i\sqrt{1-\frac{p_z}{E_{\lambda,p_z}}}\frac{\sqrt{\lambda}F_2(\rho)}{N_2}e^{i\phi/2} \end{pmatrix}. \tag*{(54)}

Therefore, the matrix element of the right-handed current is

jR3(x∣λ,m,pz,ζ)=qψˉRζ(x)γμψRζ(x)=q2B8π2Nˉ22Γ(1+a)Γ(1+λ)[F12(ρ)N12−λF22(ρ)N22+pzEλ,pz(F12(ρ)N12+λF22(ρ)N22)],(55)j_R^3(x|\lambda,m,p_z,\zeta)=q\bar{\psi}_{R\zeta}(x)\gamma^\mu\psi_{R\zeta}(x) =\frac{q^2B}{8\pi^2}\frac{\bar{N}^2}{2}\frac{\Gamma(1+a)}{\Gamma(1+\lambda)} \left[\frac{F_1^2(\rho)}{N_1^2}-\frac{\lambda F_2^2(\rho)}{N_2^2}+\frac{p_z}{E_{\lambda,p_z}}\left(\frac{F_1^2(\rho)}{N_1^2}+\frac{\lambda F_2^2(\rho)}{N_2^2}\right)\right], \tag*{(55)}

where

Nˉ2=2N12N22λN12+N22Γ(1+λ)Γ(1+a)(56)\bar{N}^2=\frac{2N_1^2N_2^2}{\lambda N_1^2+N_2^2}\frac{\Gamma(1+\lambda)}{\Gamma(1+a)} \tag*{(56)}

and a=n+m−1/2a=n+m-1/2. The thermal expectation value of the particle part of the right-handed current is

⟨jR3(x)⟩+=tr⁡[ρ^j^R+3(x)]=∑m=−∞∞∑ζ=±∑k=1∞∫−∞∞dpz nF(Eλk,pz−μR)jR3(x∣λk,m,pz,ζ).(57)\langle j_R^3(x)\rangle_+=\operatorname{tr}\left[\hat{\rho}\hat{j}_{R+}^3(x)\right] =\sum_{m=-\infty}^{\infty}\sum_{\zeta=\pm}\sum_{k=1}^{\infty} \int_{-\infty}^{\infty}dp_z\,n_F(E_{\lambda_k,p_z}-\mu_R)j_R^3(x|\lambda_k,m,p_z,\zeta). \tag*{(57)}

The summation over the roots {λk}\{\lambda_k\} must be done for specific values of mm and ζ\zeta since the energy spectrum depends on these quantities. Removing the pzp_z-odd part of the matrix element, which vanishes upon integration, and including the antiparticle contribution, we obtain

⟨jR3(x)⟩=q2B4π2∑m,ζ,kNˉ22Γ(1+a)Γ(1+λk)[F12(ρ)N12−λkF22(ρ)N22]∫0∞dpz[nF(Eλk,pz−μR)−nF(Eλk,pz+μR)].(58)\langle j_R^3(x)\rangle=\frac{q^2B}{4\pi^2}\sum_{m,\zeta,k}\frac{\bar{N}^2}{2}\frac{\Gamma(1+a)}{\Gamma(1+\lambda_k)} \left[\frac{F_1^2(\rho)}{N_1^2}-\frac{\lambda_kF_2^2(\rho)}{N_2^2}\right] \int_0^\infty dp_z\left[n_F(E_{\lambda_k,p_z}-\mu_R)-n_F(E_{\lambda_k,p_z}+\mu_R)\right]. \tag*{(58)}

We compute the left-handed current in a similar way. The sum of the right- and left-handed currents yields the vector current:

⟨jV3(x)⟩=q2B4π2∑m,ζ,kNˉ22Γ(1+a)Γ(1+λk)[F12(ρ)N12−λkF22(ρ)N22]Iλk(T,μ,μA),(59)\langle j_V^3(x)\rangle=\frac{q^2B}{4\pi^2}\sum_{m,\zeta,k}\frac{\bar{N}^2}{2}\frac{\Gamma(1+a)}{\Gamma(1+\lambda_k)} \left[\frac{F_1^2(\rho)}{N_1^2}-\frac{\lambda_kF_2^2(\rho)}{N_2^2}\right]\mathcal{I}_{\lambda_k}(T,\mu,\mu_A), \tag*{(59)}

where we denoted

Iλ(T,μ,μA)=∫0∞dpz[nF(Eλ,pz−μR)−nF(Eλ,pz+μR)−nF(Eλ,pz−μL)+nF(Eλ,pz+μL)]=IλR(T,μR)−IλL(T,μL).(60)\mathcal{I}_{\lambda}(T,\mu,\mu_A)=\int_0^\infty dp_z\left[n_F(E_{\lambda,p_z}-\mu_R)-n_F(E_{\lambda,p_z}+\mu_R)-n_F(E_{\lambda,p_z}-\mu_L)+n_F(E_{\lambda,p_z}+\mu_L)\right] =\mathcal{I}_{\lambda}^{R}(T,\mu_R)-\mathcal{I}_{\lambda}^{L}(T,\mu_L). \tag*{(60)}

In the infinite radius limit, (59) reduces to (18) because λk=n\lambda_k=n, Nˉ2=1\bar{N}^2=1, and

a!n!F12(ρ)N12=Ia,n2(ρ)=In,a2(ρ),\frac{a!}{n!}\frac{F_1^2(\rho)}{N_1^2}=I_{a,n}^2(\rho)=I_{n,a}^2(\rho),
a!n!F22(ρ)N22=Ia,n−12(ρ)=In−1,a2(ρ).(61)\frac{a!}{n!}\frac{F_2^2(\rho)}{N_2^2}=I_{a,n-1}^2(\rho)=I_{n-1,a}^2(\rho). \tag*{(61)}

Equation (59) expresses the CME in a finite volume and is our main analytical result. The summation and the integration are converging and finite because the boundary condition ensures that the energy spectrum is bound from below. In contrast from the infinite-volume case, where the CME is completely determined by the lowest Landau level, we observe in (59) that in the finite volume the higher Landau levels (k>1k>1) also contribute to the CME.

Since the analytical roots of (42) are unknown, the functional dependence of the bound CME (59) on the model parameters can only be determined numerically. It is convenient to study the effect of the boundary by considering the normalized value of the CME, which we define as the electric current in a finite region obtained from (59) divided by the CME conductivity in an unbound space:

Normalized CME=⟨jV3(x)⟩q2BμA/(2π2).(62)\text{Normalized CME}=\frac{\langle j_V^3(x)\rangle}{q^2B\mu_A/(2\pi^2)}. \tag*{(62)}

Results

We computed the bound CME by truncating the sums in mm and kk, and numerically evaluating (60). We estimated the accuracy of the calculation as the next term in the summation and we stopped the calculation when the error was smaller than the thickness of the lines used to plot the results. We found that the dependence on the axial chemical potential μA\mu_A is linear as in the unbound chiral magnetic effect. It therefore cancels out in the ratio (62). With this observation in mind we set βμA=0.1\beta\mu_A=0.1 throughout the numerical analysis.

We first report the value of the bound CME at the center of the cylinder ρ=0\rho=0 where the CME reaches its peak. Figure 2 and Figure 3 show the normalized CME at the center of the cylinder obtained by fixing the parameter ρR\rho R and varying the magnetic field intensity ∣qB∣|qB|. Having fixed ρR\rho R, an increase in the magnetic field corresponds to a decrease in the radius of the cylinder R=2ρR/∣qB∣R=\sqrt{2\rho R/|qB|}. For small magnetic fields (large radius), the bound CME has the same value as the unbound one. However, as we commented earlier, its value is not completely determined by the ground state, which can loosely be referred to as the LLL of the bound system. Contributions from higher “Landau levels” (HLL) are larger for smaller ρR\rho R; in Figure 3 the HLL contributions are negligible already at ρR=10\rho R=10. As we increase the magnetic field (decrease the radius) the bound CME decreases and eventually vanishes. We also find that CME depends on the temperature. If we decrease the temperature while keeping the radius and the magnetic field fixed, then β2∣qB∣\beta^2|qB| increases and the CME is suppressed at very low temperatures. Figure 2 and Figure 3 do not reach the limit ∣qB∣=0|qB|=0 because lowering the magnetic field becomes numerically expensive as it requires the inclusions of higher Landau levels.

Figure 3. Same as Fig. 2 for a larger range of β2∣qB∣\beta^2|qB|.

Let us now look at the CME in a common experimental setup: a fixed radius of the cylinder and a varied magnetic field. The results are shown in Figure 4. The bound CME approximately equals the unbound one in strong magnetic fields, whereas it is suppressed in weak fields. The parameter that distinguishes the two regimes is B(ρR)\mathcal{B}(\rho R) that we define below in (64). The difference between the normalized CME at strong and weak magnetic fields is seen in Figure 5 which displays the radial dependence of the effect. We can see that in both cases, the CME is larger at the center of the cylinder and is vanishing at its surface ρ=ρR\rho=\rho R. The difference between the two regimes is that the value of the normalized CME at the symmetry axis ρ=0\rho=0 is close to unity for ∣qB∣<B(ρ∣qB∣,R)|qB|<\mathcal{B}(\rho|qB|,R) and is smaller than one otherwise. We also notice the different functional form of the ρ\rho-dependence in the two cases.

Figure 4. The normalized CME as a function of the magnetic field keeping the radius of the cylinder RR constant. The inset shows the details of the top-left part of the larger plot.

Figure 5. The normalized CME as a function of the radial distance. The values of ∣qB∣|qB| are chosen with no particular criteria, and that is why the CME at the origin goes to different values for ∣qB∣>B|qB| > B.

The vanishing of the CME current at the surface of the cylinder might be surprising at first glance because it is not forbidden by the symmetry of the problem nor by the MIT boundary conditions. Moreover, one might expect that the normal modes which are not present in the unbound case are pushed toward the boundary of the cylinder and might cause an accumulation of chiral charge similar to what is observed in the slab geometry [29]. However, on closer inspection, we realize that the settings of the system forbid the presence of a chiral charge at the surface. This can be seen evaluating the axial charge matrix element using the solutions (20), which reads

jA0(x∣λk,m,pz,ζ)=qB2π2[C1C3F12(ρ)+C2C4F22(ρ)].(63)j_A^0(\mathbf{x}|\lambda_k,m,p_z,\zeta)=\frac{qB}{2\pi^2}\left[C_1C_3F_1^2(\rho)+C_2C_4F_2^2(\rho)\right]. \tag*{(63)}

At the boundary ρ=ρR\rho=\rho R, using (41), the axial charge must vanish jA0(R)=0j_A^0(R)=0. That is why the CME always vanishes at the surface of the cylinder. The vanishing of the current at the boundary was also found for the chiral vortical effect with MIT boundary conditions in the absence of a magnetic field [17].

Qualitative analysis

The effect of the boundary on CME can be gauged by examining the contribution of the LLL, which is the level of the unbound system with λ=n=0\lambda= n = 0. Indeed, we have already demonstrated that bound CME consists of contributions from the entire energy spectrum, whereas the unbound CME is driven exclusively by LLL (see, for example, Figure 2). As the cylinder radius decreases, the LLL energy increases [38]. Additionally, the boundary condition lifts the degeneracy in the angular momentum mm. In our model the levels that converge onto the LLL of the unbound system are obtained from Eq. (42). Among them is the lowest energy level with m=1/2m = 1/2, which we will call the LLL of a bound system. It scales with ρR\rho_R which in turn depends on RR and BB. In particular, at large ρR\rho_R, the corresponding principal quantum number is given by (44).

The energy of LLL is given by (23). Since pz∼Tp_z \sim T, the effect of the boundary on CME is essential only if 2∣qB∣λLLLβ2|qB|\lambda_{\mathrm{LLL}}\beta is of order unity or larger. Define the characteristic value of the magnetic field as

β2B(ρR)=12λLLL(ρR).(64)\beta^2\mathcal{B}(\rho_R)=\frac{1}{2\lambda_{\mathrm{LLL}}(\rho_R)}. \tag*{(64)}

The effect of the boundary is negligible if ∣qB∣≪B(ρR)|qB| \ll\mathcal{B}(\rho_R). We note that the characteristic value of the magnetic field is the increasing exponential function of ρR\rho_R.

Although (64) reflects the particular boundary condition that we imposed, the qualitative picture is fairly general. That the energy levels of a system increase as its size decreases follows from the uncertainty relation. An increase in energy results in a suppression of thermal quantities, because they are related to the Fermi-Dirac thermal distribution function. The occupation number nF(βE)n_F(\beta E) is small for high values of βE\beta E, leading to suppression of thermal quantities. In the unbound case the lowest Landau level has the same energy independently of the magnetic field because λ\lambda is exactly zero. The CME conductivity is given completely by the LLL and remains the same at any value of the magnetic field. Instead in the presence of a boundary, we can always find a value of the magnetic field such that 2β2λLLL∣qB∣=12\beta^2\lambda_{\mathrm{LLL}}|qB| = 1.

We can now understand the results obtained for the bound CME. If we keep ρR\rho_R fixed, then the characteristic value B(ρR)\mathcal{B}(\rho_R) is also fixed. Based on the above argument, when ∣qB∣|qB| is close to or larger than B\mathcal{B} the CME would be suppressed. This is what we see in Figure 3. Instead, if we fix the radius of the cylinder RR, the value of ρR\rho_R changes with the magnetic field, and so does the characteristic value B\mathcal{B}. As shown in Figure 6, if we fix the size of the cylinder, then ∣qB∣<B(ρR)|qB| < \mathcal{B}(\rho_R) is always satisfied above a certain value of the magnetic field ∣qB∣0|qB|_0. Namely, ∣qB∣>B(ρR)|qB| > \mathcal{B}(\rho_R) when ∣qB∣<∣qB∣0|qB| < |qB|_0 (weak field), and ∣qB∣<B(ρR)|qB| < \mathcal{B}(\rho_R) when ∣qB∣>∣qB∣0|qB| > |qB|_0 (strong field).

In the weak field (at fixed RR), there is suppression of the CME, while in the strong field there is only minor effect as compared to the unbound case. This explains the results of Figure 4.

What is described above is summarized in Figure 7 showing the dependence of the LLL energy on various parameters. When ∣qB∣≪B(ρR)|qB| \ll\mathcal{B}(\rho_R), the transverse momentum is negligible and the energy is only given by the longitudinal momentum pzp_z. When ∣qB∣=B(ρR)|qB| = \mathcal{B}(\rho_R), the LLL energy is βELLL=1+β2pz2\beta E_{\mathrm{LLL}} = \sqrt{1+\beta^2p_z^2}. For this energy and for energies above this, the effect of the boundary becomes strong and the CME is suppressed. Starting at any values of ∣qB∣|qB| and ρR\rho_R, if we increase the magnetic field ∣qB∣|qB| keeping ρR\rho_R fixed, the LLL energy will increase resulting in further suppression of the CME. Instead, if we increase ∣qB∣|qB| keeping the radius RR fixed we increase ρR\rho_R, we decrease the LLL energy and the CME will be less suppressed. The LLL energy is lowest for λ=0\lambda= 0, i.e. in an unbound system.

Figure 7. The effect of the finite size of the cylinder on the LLL energy for different values of the magnetic field. When ∣qB∣<B(ρR)|qB| < \mathcal{B}(\rho_R) the lowest Landau level lays in the light shaded area and for ∣qB∣≪B(ρR)|qB| \ll\mathcal{B}(\rho_R) it has negligible transverse momentum 2∣qB∣λ≈02|qB|\lambda\approx0 (the lowest solid line). In the opposite limit ∣qB∣>B(ρR)|qB| > \mathcal{B}(\rho_R) the LLL lays in the dark shaded area. The arrows indicate the shift of the LLL energy if we increase the magnetic field at fixed ρR\rho_R or at fixed RR.

Conclusions

We studied a system of free massless fermions confined in a cylinder with a finite radius in the presence of a constant magnetic field using MIT boundary conditions. The energy spectrum is different from the Landau levels obtained in an unbound space. The degeneracy with respect to the longitudinal total angular momentum mm is lifted due to the boundary conditions and the principal quantum number λ\lambda is not an integer. The energy ground state of the bound system, which we also called LLL because it reduces to the ground state of the unbound system, depends on the system size. It increases when the cylinder radius RR decreases, in agreement with the uncertainty relation. At finite RR, the LLL does not have a definite chirality and has a corresponding principal quantum number given by Eq. (44) which tends to zero when lB≪Rl_B \ll R. In general, the effect of the boundary on the spectrum and by extension on all statistical quantities is important when the magnetic length lBl_B is of the order of or larger than the radius of the cylinder RR.

At finite temperature and finite chiral imbalance, we computed the CME. The main result is shown in Figure 4. For cylinders of the same size, the CME is suppressed at very small magnetic fields and at very low temperatures. The CME current is maximal at the center of the cylinder and it is vanishing at the boundary of the cylinder. Even when the CME current is the same as in the unbound case, we found that different from the unbound case where the CME is completely determined by the lowest Landau level, in a finite region, all the levels contribute to the CME.

A suppression of the CME conductivity might seem incompatible with the non-renormalization of the chiral anomaly, which ultimately causes the CME. However, the anomaly only dictates that the CME current must satisfy the relation (1) at the operator level, but does not prevent the CME conductivity to be renormalized when the electric current operator is taken over a physical state [39, 40, 41]. As a consequence, it is possible that the CME conductivity differs when evaluated in bound states rather than unbound states as done in this work.

It is important to stress that although we derived the result using thermal equilibrium with a conserved axial charge, the axial chemical potential is treated as a dynamical quantity. In our model μA\mu_A only affects the thermal distribution functions but the ground state of the system is not realized with a nonvanishing μA\mu_A. The use of thermal equilibrium is justified when looking at the system at timescales shorter than the relaxation time of the axial chemical potential. Therefore, a nonvanishing CME does not violate the conservation of energy as it does not describe a true steady state. The finite-volume effects of the CME were also studied in [28, 29, 30] in a slab geometry at actual equilibrium with chiral imbalance using a model where the ground state does have a fixed nonvanishing μA\mu_A. Accordingly, the total CME current must vanish in order to conserve energy [42] in accordance with the Bloch theorem [43]. This was confirmed by [28, 29, 30]. The complete absence of the CME current in [28, 30] is then understood as a consequence of the boundary conditions explicitly preventing the flow of charges in that direction [28], while the different geometry in [29] allows for a local CME current.

In this paper we considered only one possible boundary condition. Different conditions are likely to give different values near the boundary. However, as we argued above, the main cause of the CME suppression is that the LLL energy increases as we decrease the cylinder radius as dictated by the uncertainty principle. For this reason, it seems to us that our result should not significantly change if we impose different boundary conditions. For phenomenological assessments, it is important to show that this result is indeed not model dependent. It would also be interesting to examine models with dynamical confinement instead of boundary conditions.

In practice, an accurate measurement of the boundary effects on CME may be possible in Dirac/Weyl semimetals. This requires lB>Rl_B > R, where RR is the specimen linear size and the magnetic length is given by lB=0.8μm/B(mT)l_B = 0.8\mu\mathrm{m}/\sqrt{B(\mathrm{mT})}. For example, for a semimetal with an observable CME in B∼mTB \sim\mathrm{mT} magnetic fields, RR should be of the order of a μm\mu\mathrm{m}. In the relativistic heavy-ion collisions, R∼10 fmR \sim10\ \mathrm{fm} is the order of the system size, while lBl_B can range from about a fm\mathrm{fm} to pm\mathrm{pm} at different collision energies. Therefore, the boundary effects are expected to be a significant factor in the CME phenomenology.

Acknowledgments

We are grateful to Semeon Valgushev for many fruitful discussions. This work was supported in part by the U.S. Department of Energy under Grants No. DE-FG02-87ER40371 and No. DE-SC0023692.

References

References

  1. [1]D. E. Kharzeev, Prog. Part. Nucl. Phys. 75, 133 (2014), arXiv:1312.3348 [hep-ph].DOI
  2. [2]D. E. Kharzeev, K. Landsteiner, A. Schmitt, and H.-U. Yee, Lect. Notes Phys. 871, 1 (2013), arXiv:1211.6245 [hep-ph].DOI
  3. [3]D. E. Kharzeev, J. Liao, S. A. Voloshin, and G. Wang, Prog. Part. Nucl. Phys. 88, 1 (2016), arXiv:1511.04050 [hep-ph].DOI
  4. [4]V. A. Miransky and I. A. Shovkovy, Phys. Rept. 576, 1 (2015), arXiv:1503.00732 [hep-ph].
  5. [5]K. Landsteiner, Acta Phys. Polon. B 47, 2617 (2016), arXiv:1610.04413 [hep-th].DOI
  6. [6]D. E. Kharzeev and J. Liao, Nature Rev. Phys. 3, 55 (2021), arXiv:2102.06623 [hep-ph].DOI
  7. [7]D. E. Kharzeev and A. Zhitnitsky, Nucl. Phys. A 797, 67 (2007), arXiv:0706.1026 [hep-ph].DOI
  8. [8]D. Kharzeev, Phys. Lett. B 633, 260 (2006), arXiv:hep-ph/0406125.DOI
  9. [9]D. E. Kharzeev, L. D. McLerran, and H. J. Warringa, Nucl. Phys. A 803, 227 (2008), arXiv:0711.0950 [hep-ph].DOI
  10. [10]K. Fukushima, D. E. Kharzeev, and H. J. Warringa, Phys. Rev. D 78, 074033 (2008), arXiv:0808.3382 [hep-ph].DOI
  11. [11]A. R. Zhitnitsky, Nucl. Phys. A 897, 93 (2013), arXiv:1208.2697 [hep-ph].DOI
  12. [12]A. R. Zhitnitsky, Annals Phys. 336, 462 (2013), arXiv:1301.7072 [hep-ph].DOI
  13. [13]K. Tuchin, Phys. Rev. C 97, 064914 (2018), arXiv:1802.09629 [hep-ph].DOI
  14. [14]K. Tuchin, (2020), arXiv:2007.02822 [hep-ph].arxiv.org/abs/2007.02822
  15. [15]L. Adamczyk et al. (STAR), Nature 548, 62 (2017), arXiv:1701.06657 [nucl-ex].
  16. [16]G. Duffy and A. C. Ottewill, Phys. Rev. D 67, 044002 (2003), arXiv:hep-th/0211096.DOI
  17. [17]V. E. Ambrus and E. Winstanley, Phys. Rev. D 93, 104014 (2016), arXiv:1512.05239 [hep-th].DOI
  18. [18]S. Ebihara, K. Fukushima, and K. Mameda, Phys. Lett. B 764, 94 (2017), arXiv:1608.00336 [hep-ph].
  19. [19]M. N. Chernodub and S. Gongyo, Phys. Rev. D 95, 096006 (2017), arXiv:1702.08266 [hep-th].DOI
  20. [20]M. N. Chernodub and S. Gongyo, Phys. Rev. D 96, 096014 (2017), arXiv:1706.08448 [hep-th].DOI
  21. [21]H.-L. Chen, K. Fukushima, X.-G. Huang, and K. Mameda, Phys. Rev. D 96, 054032 (2017), arXiv:1707.09130 [hep-ph].DOI
  22. [22]N. Sadooghi, S. M. A. Tabatabaee Mehr, and F. Taghinavaz, Phys. Rev. D 104, 116022 (2021), arXiv:2108.12760 [hep-ph].DOI
  23. [23]V. E. Ambrus and M. N. Chernodub, arXiv preprint (2023), arXiv:2304.05998 [hep-th].DOI
  24. [24]S.-Y. Yang, R.-H. Fang, D.-F. Hou, and H.-C. Ren, Chin. Phys. C 47, 034106 (2023), arXiv:2111.13053 [hep-th].DOI
  25. [25]A. Chodos, R. L. Jaffe, K. Johnson, C. B. Thorn, and V. F. Weisskopf, Phys. Rev. D 9, 3471 (1974).
  26. [26]A. W. Thomas, Adv. Nucl. Phys. 13, 1 (1984).
  27. [27]R. Shanker, D. Vasak, C. S. Warke, W. Greiner, and B. Muller, Z. Phys. C 18, 327 (1983).
  28. [28]E. V. Gorbar, V. A. Miransky, I. A. Shovkovy, and P. O. Sukhachov, Phys. Rev. B 92, 245440 (2015), arXiv:1509.06769 [cond-mat.mes-hall].DOI
  29. [29]S. N. Valgushev, M. Puhr, and P. V. Buividovich, PoS LATTICE2015, 043 (2016), arXiv:1512.01405 [cond-mat.str-el].arxiv.org/abs/1512.01405
  30. [30]Y. A. Sitenko, Phys. Rev. D 94, 085014 (2016), arXiv:1606.08241 [hep-th].DOI
  31. [31]A. V. Vilenkin, Phys. Rev. D 22, 3080 (1980).
  32. [32]A. Vilenkin, Phys. Rev. D 22, 3067 (1980).
  33. [33]N. Yamamoto, Phys. Rev. D 92, 085011 (2015), arXiv:1502.01547 [cond-mat.mes-hall].DOI
  34. [34]A. Vilenkin, Phys. Rev. D21, 2260 (1980).
  35. [35]M. Buzzegoli, Lect. Notes Phys. 987, 59 (2021), arXiv:2011.09974 [hep-th].DOI
  36. [36]I. M. T. A. Sokolov, Radiation from Relativistic Electrons (American Institute of Physics Translation Series), 2nd ed. (American Institute of Physics, 1986).
  37. [37]K. Mameda, Chiral symmetry and vacuum structure in rotating relativistic systems, Ph.D. thesis, University of Tokyo (2016), doi:10.1508/00075586.DOI
  38. [38]M. Buzzegoli and K. Tuchin, Nucl. Phys. A 1030, 122577 (2023), arXiv:2209.03991 [nucl-th].DOI
  39. [39]A. A. Anselm and A. A. Johansen, JETP Lett. 49, 214 (1989).
  40. [40]S. L. Adler, “Anomalies to all orders,” in 50 years of Yang-Mills theory, edited by G. ’t Hooft (World Scientific, Singapore, 2005) pp. 187–228, arXiv:hep-th/0405040.
  41. [41]B. Feng, D. Hou, and H. Ren, Phys. Rev. D 99, 036010 (2019), arXiv:1810.05954 [hep-ph].DOI
  42. [42]D. E. Kharzeev, in Chiral Matter (WORLD SCIENTIFIC, 2023).
  43. [43]D. Bohm, Phys. Rev. 75, 502 (1949).

Paper details

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

views · 6

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