betaFeedback

WorksPhysics and Astronomy

Pseudogauge dependence of the spin polarization and of the axial vortical effect

Abstract

The mean spin polarization vector of spin 1/2 particles in a relativistic fluid at local thermal equilibrium (LTE) in different pseudogauges (PGs), i.e., with different choices for the decomposition of orbital and spin angular momentum, is obtained. The spin polarization obtained in the canonical PG differs from the one obtained in the Belinfante PG. It is found that this difference can not be written by replacing the thermal vorticity with the spin potential in the usual polarization formula. Other PG choices affect the contribution of thermal shear to the spin polarization. Therefore, the choice of a PG affects the predictions for the polarization measured in heavy-ion collisions. In general, it is shown that the Wigner function of a noninteracting Dirac field at LTE is affected by the PG transformations. Explicit expressions of the mean axial current are also calculated and it is found that even the axial vortical effect conductivity depends on the PG.

Introduction

The measurement of spin polarization of particles in the quark gluon plasma (QGP) formed in relativistic heavy ion collisions [1, 2] has opened the possibility for new phenomenological investigations of spin physics in relativistic fluids. Predictions of relativistic hydrodynamics are able to explain the measurements of global (integrated over all momenta) spin polarization, but fails to reproduce the data on momentum-dependent polarization [3]; see [4] for a review. Recently, it was found that a previously overlooked spin-thermal shear coupling [5, 6] could restore the agreement between theory and experimental data [7, 8]. However, from the theory side, one should take into account that the choice of a decomposition of the angular momentum into an orbital and spin part could significantly affect the spin polarization predictions of a fluid at local thermal equilibrium.

For this reason, recent studies have begun to examine the role of the spin tensor [9, 10, 11, 12, 13, 14, 15, 16, 17, 18] and also the impact of a spin potential [19, 20, 21, 22, 23] in relativistic hydrodynamics. In more general terms, different decompositions of the boost-angular momentum current are obtained by choosing a particular form of the quantum energy-momentum tensor (EMT) T^μν\widehat{T}^{\mu\nu} and of the spin tensor S^λ,μν\widehat{S}^{\lambda,\mu\nu}. Indeed this choice is not unique in special relativity. Given one pair of operators, one can always generate another equally valid pair by means of a pseudogauge transformation (PGT) [24]:

T^′μν=T^μν+12∇λ(Φ^λ,μν−Φ^μ,λν−Φ^ν,λμ)\widehat{T}^{\prime\mu\nu}=\widehat{T}^{\mu\nu}+\frac{1}{2}\nabla_{\lambda}\left(\widehat{\Phi}^{\lambda,\mu\nu}-\widehat{\Phi}^{\mu,\lambda\nu}-\widehat{\Phi}^{\nu,\lambda\mu}\right)
S^′λ,μν=S^λ,μν−Φ^λ,μν+∇ρZ^μν,λρ\widehat{S}^{\prime\lambda,\mu\nu}=\widehat{S}^{\lambda,\mu\nu}-\widehat{\Phi}^{\lambda,\mu\nu}+\nabla_{\rho}\widehat{Z}^{\mu\nu,\lambda\rho}

where the rank-3 tensor field Φ^\widehat{\Phi} and the rank-4 tensor Z^\widehat{Z} must satisfy the following symmetries:

Φ^λ,μν=−Φ^λ,νμ,Z^μν,λρ=−Z^νμ,λρ=−Z^μν,ρλ.\widehat{\Phi}^{\lambda,\mu\nu}=-\widehat{\Phi}^{\lambda,\nu\mu},\qquad\widehat{Z}^{\mu\nu,\lambda\rho}=-\widehat{Z}^{\nu\mu,\lambda\rho}=-\widehat{Z}^{\mu\nu,\rho\lambda}.

Since such transformations do not affect the equation of motions and the total energy, momentum, and angular momentum, they are generally regarded as physically irrelevant. However, the relativistic hydrodynamic equations are obtained by constraining local - as opposed to global - quantities, therefore the PGTs would lead to different physical results. The role of PGTs in statistical quantum field theory was analyzed here [25, 26, 27, 28], where it was shown that they might bring a contribution only when the system is out of global thermal equilibrium; see also [29].

Despite the fact that this issue was already discussed in the literature, explicit formulas for spin polarization in different pseudogauges have not yet been reported. Indeed, predictions for spin polarization of lambda hyperons produced in heavy-ion collisions are obtained starting from the local thermal equilibrium statistical operator,

ρ^LTEB=1Zexp⁡[−∫ΣdΣμ(T^Bμνβν−j^μζ)],(1)\widehat{\rho}_{\mathrm{LTE}}^{\mathrm{B}}=\frac{1}{Z}\exp\left[-\int_{\Sigma}d\Sigma_{\mu}\left(\widehat{T}_{\mathrm{B}}^{\mu\nu}\beta_{\nu}-\widehat{j}^{\mu}\zeta\right)\right], \tag*{(1)}

obtained by choosing the Belinfante form of EMT T^B\widehat{T}_{\mathrm{B}} with a corresponding vanishing spin tensor and where β\beta is the fluid velocity uu divided by the temperature TT. The resulting spin polarization of a particle with momentum kk and mass mm at first order of thermodynamic gradients is [5, 30, 31]

SBμ(k)≃Sϖμ(k)+Sξμ(k),(2)S_{\mathrm{B}}^{\mu}(k)\simeq S_{\varpi}^{\mu}(k)+S_{\xi}^{\mu}(k), \tag*{(2)}

where

Sϖμ(k)=−18mϵμρστkτ∫ΣdΣ⋅k nF(1−nF)ϖρσ∫ΣdΣ⋅k nF(3)S_{\varpi}^{\mu}(k)=-\frac{1}{8m}\epsilon^{\mu\rho\sigma\tau}k_{\tau}\frac{\int_{\Sigma}d\Sigma\cdot k\,n_{\mathrm{F}}(1-n_{\mathrm{F}})\varpi_{\rho\sigma}}{\int_{\Sigma}d\Sigma\cdot k\,n_{\mathrm{F}}} \tag*{(3)}
Sξμ(k)=−14mϵμλστkτkρϵk∫ΣdΣ⋅k nF(1−nF)t^λξρσ∫ΣdΣ⋅k nF,(4)S_{\xi}^{\mu}(k)=-\frac{1}{4m}\epsilon^{\mu\lambda\sigma\tau}\frac{k_{\tau}k^{\rho}}{\epsilon_{k}}\frac{\int_{\Sigma}d\Sigma\cdot k\,n_{\mathrm{F}}(1-n_{\mathrm{F}})\widehat{t}_{\lambda}\xi_{\rho\sigma}}{\int_{\Sigma}d\Sigma\cdot k\,n_{\mathrm{F}}}, \tag*{(4)}

with ϵk=k2+m2\epsilon_k=\sqrt{k^2+m^2}, and

ϖμν=−12(∂μβν−∂νβμ),\varpi_{\mu\nu}=-\frac{1}{2}\left(\partial_{\mu}\beta_{\nu}-\partial_{\nu}\beta_{\mu}\right),
ξμν=12(∂μβν+∂νβμ).\xi_{\mu\nu}=\frac{1}{2}\left(\partial_{\mu}\beta_{\nu}+\partial_{\nu}\beta_{\mu}\right).

are respectively the thermal vorticity and the thermal shear, and nF=nF(β(x)⋅k)n_{\mathrm F}=n_{\mathrm F}(\beta(x)\cdot k) denotes the Fermi distribution function:

nF(z)=1ez+1.n_{\mathrm F}(z)=\frac{1}{e^z+1}.

The QGP being a fluid out of equilibrium, one then expects PGTs to affect the prediction (2). It is then of phenomenological importance to quantify possible modifications to the previous formula in different pseudogauges. In particular, the statistical operator obtained by choosing the canonical form of EMT T^Cμν\widehat{T}_{\mathrm C}^{\mu\nu} and the associated canonical spin tensor S^Cμ,λν\widehat{S}_{\mathrm C}^{\mu,\lambda\nu} is

ρ^LTEC=1Zexp⁡[−∫dΣμ(T^Cμνβν−12ΩλνS^Cμ,λν)],(5)\widehat{\rho}_{\mathrm{LTE}}^{\mathrm C}=\frac{1}{Z}\exp\left[-\int d\Sigma_\mu\left(\widehat{T}_{\mathrm C}^{\mu\nu}\beta_\nu-\frac{1}{2}\Omega_{\lambda\nu}\widehat{S}_{\mathrm C}^{\mu,\lambda\nu}\right)\right], \tag*{(5)}

where Ω\Omega is the spin potential. In this work I am going to show that if one starts from the statistical operator (5), then the spin polarization becomes

SCμ(k)≃SBμ(k)+ΔΘCSμ(k),(6)S_{\mathrm C}^{\mu}(k)\simeq S_{\mathrm B}^{\mu}(k)+\Delta_{\Theta}^{\mathrm C}S^{\mu}(k), \tag*{(6)}

which differs from (2) by

ΔΘCSμ(k)=ϵλρστt^λ(kμkτ−ητμm2)8mεk∫ΣdΣ(x)⋅k nF(1−nF)(ϖρσ−Ωρσ)∫ΣdΣ⋅k nF,(7)\Delta_{\Theta}^{\mathrm C}S^{\mu}(k)=\frac{\epsilon^{\lambda\rho\sigma\tau}\widehat{t}_{\lambda}(k^{\mu}k_{\tau}-\eta^{\mu}_{\tau}m^2)}{8m\varepsilon_k}\frac{\int_{\Sigma}d\Sigma(x)\cdot k\,n_{\mathrm F}(1-n_{\mathrm F})(\varpi_{\rho\sigma}-\Omega_{\rho\sigma})}{\int_{\Sigma}d\Sigma\cdot k\,n_{\mathrm F}}, \tag*{(7)}

where t^\widehat{t} is the time direction in the fluid frame. I will also show that either the de Groot-van Leeuwen-van Weert and Hilgevoord-Wouthuysen (HW) decompositions of angular momentum result in the spin polarization

SGLW,HWμ(k)≃SCμ(k)+ΔΘGLW,HWSμ(k)+ΔξGLW,HWSμ(k)=−18mϵμρστkτ∫ΣdΣ⋅k nF(1−nF)Ωρσ∫ΣdΣ⋅k nF,S_{\mathrm{GLW,HW}}^{\mu}(k)\simeq S_{\mathrm C}^{\mu}(k)+\Delta_{\Theta}^{\mathrm{GLW,HW}}S^{\mu}(k)+\Delta_{\xi}^{\mathrm{GLW,HW}}S^{\mu}(k) =-\frac{1}{8m}\epsilon^{\mu\rho\sigma\tau}k_{\tau}\frac{\int_{\Sigma}d\Sigma\cdot k\,n_{\mathrm F}(1-n_{\mathrm F})\Omega_{\rho\sigma}}{\int_{\Sigma}d\Sigma\cdot k\,n_{\mathrm F}},

with

ΔΘGLW,HWSμ(k)=−14mϵμλρτt^λkτkρεk∫ΣdΣ⋅k nF(1−nF)(ϖρσ−Ωρσ)∫ΣdΣ⋅k nF,(8)\Delta_{\Theta}^{\mathrm{GLW,HW}}S^{\mu}(k)=-\frac{1}{4m}\epsilon^{\mu\lambda\rho\tau}\widehat{t}_{\lambda}\frac{k_{\tau}k^{\rho}}{\varepsilon_k}\frac{\int_{\Sigma}d\Sigma\cdot k\,n_{\mathrm F}(1-n_{\mathrm F})(\varpi_{\rho\sigma}-\Omega_{\rho\sigma})}{\int_{\Sigma}d\Sigma\cdot k\,n_{\mathrm F}}, \tag*{(8)}
ΔξGLW,HWSμ(k)=−Sξμ(k)=+14mϵμλστkτkρεk∫ΣdΣ⋅k nF(1−nF)t^λξρσ∫ΣdΣ⋅k nF,(9)\Delta_{\xi}^{\mathrm{GLW,HW}}S^{\mu}(k)=-S_{\xi}^{\mu}(k)=+\frac{1}{4m}\epsilon^{\mu\lambda\sigma\tau}\frac{k_{\tau}k^{\rho}}{\varepsilon_k}\frac{\int_{\Sigma}d\Sigma\cdot k\,n_{\mathrm F}(1-n_{\mathrm F})\widehat{t}_{\lambda}\xi_{\rho\sigma}}{\int_{\Sigma}d\Sigma\cdot k\,n_{\mathrm F}}, \tag*{(9)}

and where the expression was simplified by taking advantage of the Schouten identity. Therefore the predictions of spin polarization in heavy-ion collisions depend on the chosen PG.

chosen PG. Notice that modifications (7) and (8) rely on the spin potential and vanish when Ω=ϖ\Omega=\varpi, while modifications such as (9) do not require the presence of a spin potential. Similarly, I will show that the axial vortical effect (AVE) [32, 33, 34, 35], which is the mean axial current of a fermion induced along the rotation of the fluid, is also affected by the PGT in a system out of equilibrium.

The paper is organized as follows: in Sec. II I review how to obtain the statistical density operator in Zubarev’s statistical quantum field theory and how it is affected by PGTs. In Sec. III I consider a free Dirac field and, by using linear response theory, I evaluate the differences between the Wigner function resulting from the Belinfante decomposition of angular momentum and the Wigner functions resulting from other pseudogauges. In Sec. IV I obtain the different predictions of spin polarization in various pseudogauges and I discuss their relevance in heavy-ion collisions. In Sec. V I evaluate the mean axial current of a free Dirac field and I show that the axial vortical effect conductivity also depend on the PGTs. Finally, in Sec. VI I summarize and discuss these findings.

Notation

In this paper I adopt the natural units, with ℏ=c=kB=1\hbar=c=k_{\mathrm B}=1. The Minkowskian metric tensor gg is diag⁡(1,−1,−1,−1)\operatorname{diag}(1,-1,-1,-1); for the Levi-Civita symbol I use the convention ϵ0123=1\epsilon^{0123}=1.

I use the relativistic notation with repeated indices assumed to be saturated. Operators in Hilbert space are denoted by a wide upper hat, e.g. H^\widehat{H}, except the Dirac field operator which is denoted by a Ψ\Psi. The names “mean value”, “(thermal) expectation value” and “(thermal) average” of an operator O^\widehat{O} are all used interchangeably to denote the trace with a statistical operator ρ^\widehat{\rho}: tr⁡(ρ^O^)\operatorname{tr}(\widehat{\rho}\widehat{O}).

LOCAL THERMAL EQUILIBRIUM STATISTICAL OPERATORS

The statistical operator of a relativistic quantum fluid that has reached the local thermodynamic equilibrium (LTE) can be obtained by Zubarev’s method [36, 37]. This approach is particularly suitable for the quark gluon plasma produced in heavy-ion collisions and, among other predictions, it has been used to obtain the spin polarization formula (2) [31]. A more detailed review of the method can be found in [38, 39]. Here, I want to highlight the effect of pseudogauge transformations on the form of the statistical operator, as already discussed in [27, 28].

If the system reached LTE at a certain time τ0\tau_0, then one can describe it with an energy-momentum density TμνT^{\mu\nu}, a boost-angular momentum density Jμ,λνJ^{\mu,\lambda\nu}, and an (electric) conserved current jμj^\mu, all lying on a space-like hyper-surface Σ(τ0)\Sigma(\tau_0). The LTE density operator ρ^\hat{\rho} is then obtained by maximizing the entropy S=−tr⁡(ρ^log⁡ρ^)S=-\operatorname{tr}(\hat{\rho}\log\hat{\rho}) with the constraints of the mean conserved currents being equal to the actual ones in the hyper-surface Σ(τ0)\Sigma(\tau_0). These constraints are obtained by projecting the mean values of the quantum operators onto nn, the normalized vector perpendicular to Σ(τ0)\Sigma(\tau_0):

nμtr⁡(ρ^T^μν)=nμTμν,n_\mu\operatorname{tr}\left(\hat{\rho}\hat{T}^{\mu\nu}\right)=n_\mu T^{\mu\nu},
nμtr⁡(ρ^J^μ,λν)=nμJμ,λν,n_\mu\operatorname{tr}\left(\hat{\rho}\hat{J}^{\mu,\lambda\nu}\right)=n_\mu J^{\mu,\lambda\nu},
nμtr⁡(ρ^j^μ)=nμjμ.(10)n_\mu\operatorname{tr}\left(\hat{\rho}\hat{j}^{\mu}\right)=n_\mu j^{\mu}. \tag*{(10)}

where the operators are in the Heisenberg representation. Notice that since the boost-angular momentum current is written in terms of the EMT and of the spin tensor as

J^μ,λν=xλT^μν−xνT^μλ+S^μ,λν,\hat{J}^{\mu,\lambda\nu}=x^\lambda\hat{T}^{\mu\nu}-x^\nu\hat{T}^{\mu\lambda}+\hat{S}^{\mu,\lambda\nu},

the constraint on EMT is redundant and the boost-angular momentum is fixed just by constraining the spin tensor

nμtr⁡(ρ^S^λ,μν)=nλSλ,μν.n_\mu\operatorname{tr}\left(\hat{\rho}\hat{S}^{\lambda,\mu\nu}\right)=n_\lambda S^{\lambda,\mu\nu}.

Then, the resulting statistical operator is

ρ^=1Zexp⁡[−∫Σ(τ0)dΣμ(T^μνβν−12ΩλνS^μ,λν−j^μζ)].\hat{\rho}=\frac{1}{Z}\exp\left[-\int_{\Sigma(\tau_0)}\mathrm{d}\Sigma_\mu\left(\hat{T}^{\mu\nu}\beta_\nu-\frac{1}{2}\Omega_{\lambda\nu}\hat{S}^{\mu,\lambda\nu}-\hat{j}^{\mu}\zeta\right)\right].

where β\beta, Ω\Omega, and ζ\zeta are the Lagrange multipliers of the problem and they have the physical meaning of four-temperature vector, spin potential, and chemical potential divided by the temperature T=1/β2T=1/\sqrt{\beta^2}, respectively [40]. Notice that statistical operator is not time-independent and that it can be written by means of Gauss’s theorem as [5, 31, 40]

ρ^=1Zexp⁡[−∫Σ(τ)dΣμ(T^μνβν−12ΩλνS^μ,λν−j^μζ)+∫VdV∇μ(T^μνβν−12ΩλνS^μ,λν−j^μζ)].(11)\hat{\rho}=\frac{1}{Z}\exp\left[-\int_{\Sigma(\tau)}\mathrm{d}\Sigma_\mu\left(\hat{T}^{\mu\nu}\beta_\nu-\frac{1}{2}\Omega_{\lambda\nu}\hat{S}^{\mu,\lambda\nu}-\hat{j}^{\mu}\zeta\right)+\int_V\mathrm{d}V\nabla_\mu\left(\hat{T}^{\mu\nu}\beta_\nu-\frac{1}{2}\Omega_{\lambda\nu}\hat{S}^{\mu,\lambda\nu}-\hat{j}^{\mu}\zeta\right)\right]. \tag*{(11)}

where Σ(τ)\Sigma(\tau) is the space-like hyper-surface at “time” τ\tau, and VV is the region of space-time encompassed by the space-like hypersurfaces Σ(τ0)\Sigma(\tau_0), Σ(τ)\Sigma(\tau) and the time-like boundaries. The dissipative content of the system is only contained in the second term of the exponent [38]. For a quasi-ideal fluid such as the QGP, one expects the first term to be the predominant one, while the second term is supposedly a correction. This second term requires a careful analysis and is left for future research. In the linear response theory, it is always possible to include these neglected dissipative effects later by adding them together with the non-dissipative ones. The statistical operator is then only given by the first term in the exponent, which is the local thermal equilibrium form at time τ\tau and only contains non-dissipative effects:

ρ^LTE=1Zexp⁡[−∫Σ(τ)dΣμ(T^μνβν−12ΩλνS^μ,λν−j^μζ)].(12)\hat{\rho}_{\mathrm{LTE}}=\frac{1}{Z}\exp\left[-\int_{\Sigma(\tau)}\mathrm{d}\Sigma_\mu\left(\hat{T}^{\mu\nu}\beta_\nu-\frac{1}{2}\Omega_{\lambda\nu}\hat{S}^{\mu,\lambda\nu}-\hat{j}^{\mu}\zeta\right)\right]. \tag*{(12)}

The relativistic nature of the fluid requires that the constraints (10) were given in terms of the local currents. As long as the global equilibrium conditions are not enforced (see discussion below), it is expected that the form of the quantum operators in (10) affects the form of the statistical operator (12). Therefore, I now evaluate the effect of pseudogauge transformations (PGTs) on the LTE statistical operator (12).

Suppose one repeats the procedure above with the symmetric Belinfante EMT T^Bμν\hat{T}^{\mu\nu}_{\mathrm{B}}. The spin tensor associated with the Belinfante pseudogauge is a vanishing one (S^Bλ,μν=0\hat{S}^{\lambda,\mu\nu}_{\mathrm{B}}=0) and the total boost-angular momentum current has the form of an orbital angular momentum:

J^Bμ,λν=xλT^Bμν−xνT^Bμλ.\hat{J}^{\mu,\lambda\nu}_{\mathrm{B}}=x^\lambda\hat{T}^{\mu\nu}_{\mathrm{B}}-x^\nu\hat{T}^{\mu\lambda}_{\mathrm{B}}.

In this particular case, the constraint on the boost-angular momentum current is unnecessary as it is already being taken care of in the constraint on the energy-momentum current. The resulting statistical operator is

ρ^LTEB=1Zexp⁡[−∫dΣμ(T^Bμνβν−j^μζ)].(13)\hat{\rho}^{\mathrm{B}}_{\mathrm{LTE}}=\frac{1}{Z}\exp\left[-\int\mathrm{d}\Sigma_\mu\left(\hat{T}^{\mu\nu}_{\mathrm{B}}\beta_\nu-\hat{j}^{\mu}\zeta\right)\right]. \tag*{(13)}

In a general pseudogauge the EMT is not symmetric and the spin tensor is not vanishing. It is always possible to obtain any generic EMT and spin tensor T^Φμν\hat{T}^{\mu\nu}_{\Phi} and S^Φλ,μν\hat{S}^{\lambda,\mu\nu}_{\Phi} from the Belinfante one with the pseudogauge transformation:

T^Φμν=T^Bμν+12∇λ(Φ^λ,μν−Φ^μ,λν−Φ^ν,λμ),\hat{T}^{\mu\nu}_{\Phi}=\hat{T}^{\mu\nu}_{\mathrm{B}}+\frac{1}{2}\nabla_\lambda\left(\hat{\Phi}^{\lambda,\mu\nu}-\hat{\Phi}^{\mu,\lambda\nu}-\hat{\Phi}^{\nu,\lambda\mu}\right),
S^Φλ,μν=−Φ^λ,μν+∇ρZ^μν,λρ.(14)\hat{S}^{\lambda,\mu\nu}_{\Phi}=-\hat{\Phi}^{\lambda,\mu\nu}+\nabla_\rho\hat{Z}^{\mu\nu,\lambda\rho}. \tag*{(14)}

for some specific Φ^\hat{\Phi} and Z^\hat{Z}. Consider the LTE statistical operator (12) obtained with a specific pseudogauge:

ρ^LTEΦ=1Zexp⁡[−∫dΣμ(T^Φμνβν−12ΩλνS^Φμ,λν−j^μζ)].(15)\hat{\rho}^{\Phi}_{\mathrm{LTE}}=\frac{1}{Z}\exp\left[-\int\mathrm{d}\Sigma_\mu\left(\hat{T}^{\mu\nu}_{\Phi}\beta_\nu-\frac{1}{2}\Omega_{\lambda\nu}\hat{S}^{\mu,\lambda\nu}_{\Phi}-\hat{j}^{\mu}\zeta\right)\right]. \tag*{(15)}

The statistical operator (15) can be written in terms of the Belinfante EMT by taking advantage of the transformations (14):

ρ^LTE=1Zexp⁡{−∫dΣμ[T^Bμνβν+12∇λ(Φ^λ,μν−Φ^μ,λν−Φ^ν,λμ)βν+12Ωλν(Φ^μ,λν−∇ρZ^λν,μρ)−j^μζ]}=1Zexp⁡{−∫dΣμ[T^Bμνβν+12∇λ(βνΦ^λ,μν−βνΦ^μ,λν−βνΦ^ν,λμ)+12ΩλνΦ^μ,λν−12∇λβν(Φ^λ,μν−Φ^μ,λν−Φ^ν,λμ)+12Ωλν∇ρZ^λν,μρ−j^μζ]}.\begin{aligned} \hat{\rho}_{\mathrm{LTE}}=\frac{1}{\mathcal{Z}}\exp\left\{-\int d\Sigma_{\mu}\left[\hat{T}^{\mu\nu}_{B}\beta_{\nu}+\frac{1}{2}\nabla_{\lambda}\left(\hat{\Phi}^{\lambda,\mu\nu}-\hat{\Phi}^{\mu,\lambda\nu}-\hat{\Phi}^{\nu,\lambda\mu}\right)\beta_{\nu}+\frac{1}{2}\Omega_{\lambda\nu}\left(\hat{\Phi}^{\mu,\lambda\nu}-\nabla_{\rho}\hat{Z}^{\lambda\nu,\mu\rho}\right)-\hat{j}^{\mu}\zeta\right]\right\} \\ &=\frac{1}{\mathcal{Z}}\exp\left\{-\int d\Sigma_{\mu}\left[\hat{T}^{\mu\nu}_{B}\beta_{\nu}+\frac{1}{2}\nabla_{\lambda}\left(\beta_{\nu}\hat{\Phi}^{\lambda,\mu\nu}-\beta_{\nu}\hat{\Phi}^{\mu,\lambda\nu}-\beta_{\nu}\hat{\Phi}^{\nu,\lambda\mu}\right)+\frac{1}{2}\Omega_{\lambda\nu}\hat{\Phi}^{\mu,\lambda\nu}-\frac{1}{2}\nabla_{\lambda}\beta_{\nu}\left(\hat{\Phi}^{\lambda,\mu\nu}-\hat{\Phi}^{\mu,\lambda\nu}-\hat{\Phi}^{\nu,\lambda\mu}\right)+\frac{1}{2}\Omega_{\lambda\nu}\nabla_{\rho}\hat{Z}^{\lambda\nu,\mu\rho}-\hat{j}^{\mu}\zeta\right]\right\}. \end{aligned}

The second term in the exponent is a total divergence and is vanishing for suitable boundary conditions imposed on β\beta and/or Φ^\hat{\Phi}. Then, by splitting the β\beta derivative into a symmetric and anti-symmetric parts,

ϖμν=−12(∇μβν−∇νβμ),\varpi_{\mu\nu}=-\frac{1}{2}\left(\nabla_{\mu}\beta_{\nu}-\nabla_{\nu}\beta_{\mu}\right),
ξμν=12(∇μβν+∇νβμ),\xi_{\mu\nu}=\frac{1}{2}\left(\nabla_{\mu}\beta_{\nu}+\nabla_{\nu}\beta_{\mu}\right),

one obtains

\hat{\rho}_{\mathrm{LTE}}=\frac{1}{\mathcal{Z}}\exp\left\{-\int d\Sigma_{\mu}\left[\hat{T}^{\mu\nu}_{B}\beta_{\nu}-\frac{1}{2}\left(\varpi_{\lambda\nu}-\Omega_{\lambda\nu}\right)\hat{\Phi}^{\mu,\lambda\nu}-\xi_{\lambda\nu}\hat{\Phi}^{\lambda,\mu\nu}-\frac{1}{2}\Omega_{\lambda\nabla_{\rho}\hat{Z}^{\lambda\nu,\mu\rho}-\hat{j}^{\mu}\zeta\right]\right\}.} \tag*{(16)}

The anti-symmetric derivative of the β\beta field ϖ\varpi is called the thermal vorticity and contains information about the acceleration and the rotation of the fluid [41]. Instead, the symmetric derivative ξ\xi is called the thermal shear tensor, see [5, 6]. The relation between the statistical operators obtained in different pseudogauges was discussed in [27, 28].

As a notable example, the canonical EMT and spin tensor are related to the Belinfante one with the PGT (14) with

Φ^Cλ,μν=−S^Cλ,μν,Z^Cμν,λρ=0,\hat{\Phi}^{\lambda,\mu\nu}_{C}=-\hat{S}^{\lambda,\mu\nu}_{C},\qquad\hat{Z}^{\mu\nu,\lambda\rho}_{C}=0,

where S^C\hat{S}_{C} is the canonical spin tensor which is obtained directly by the Noether theorem:

S^Cλ,μν=−i∑a,bδLδ(∂λψa)D(Jμν) baψb\hat{S}^{\lambda,\mu\nu}_{C}=-i\sum_{a,b}\frac{\delta\mathcal{L}}{\delta(\partial_{\lambda}\psi^{a})}D(J^{\mu\nu})^{a}_{\ b}\psi^{b}

with D being the irreducible representation matrix of the Lorentz group pertaining to the field. In this case, the statistical operator becomes:

ρ^LTEC=1Zexp⁡{−∫dΣμ[T^Bμνβν+12(ϖλν−Ωλν)S^Cλ,μν+ξλνS^Cλ,μν−j^μζ]}.\hat{\rho}^{C}_{\mathrm{LTE}}=\frac{1}{\mathcal{Z}}\exp\left\{-\int d\Sigma_{\mu}\left[\hat{T}^{\mu\nu}_{B}\beta_{\nu}+\frac{1}{2}\left(\varpi_{\lambda\nu}-\Omega_{\lambda\nu}\right)\hat{S}^{\lambda,\mu\nu}_{C}+\xi_{\lambda\nu}\hat{S}^{\lambda,\mu\nu}_{C}-\hat{j}^{\mu}\zeta\right]\right\}.

The differences between this operator obtained in the canonical pseudogauge and the one obtained in the Belinfante one are discussed in ref. [27].

Besides the canonical decomposition, infinite choices for Φ^\hat{\Phi} and Z^\hat{Z} are possible. For instance, if there exist a vector and/or an axial currents j^Vμ\hat{j}^{\mu}_{V} and j^Aμ\hat{j}^{\mu}_{A}, other possible choices for Φ^\hat{\Phi} are given by

Φ^ϵAλ,μν=12ϵλτμνj^Aτ,Φ^Aλ,μν=ηλμj^Aν−ηλνj^Aμ,\hat{\Phi}^{\lambda,\mu\nu}_{\epsilon A}=\frac{1}{2}\epsilon^{\lambda\tau\mu\nu}\hat{j}_{A\tau},\qquad \hat{\Phi}^{\lambda,\mu\nu}_{A}=\eta^{\lambda\mu}\hat{j}^{\nu}_{A}-\eta^{\lambda\nu}\hat{j}^{\mu}_{A},
Φ^ϵVλ,μν=12ϵλτμνσj^Vτ,Φ^Vλ,μν=ηλμj^Vν−ηλνj^Vμ,\hat{\Phi}^{\lambda,\mu\nu}_{\epsilon V}=\frac{1}{2}\epsilon^{\lambda\tau\mu\nu\sigma}\hat{j}_{V\tau},\qquad \hat{\Phi}^{\lambda,\mu\nu}_{V}=\eta^{\lambda\mu}\hat{j}^{\nu}_{V}-\eta^{\lambda\nu}\hat{j}^{\mu}_{V},

or linear combination of all the above. Taking advantage of the derivative operator, one can also choose

Φ^ϵ∂λ,μν=i2ϵλτμν∂^τ,Φ^∂λ,μν=iηλμ∂^ν−iηλν∂^μ.\hat{\Phi}^{\lambda,\mu\nu}_{\epsilon\partial}=\frac{i}{2}\epsilon^{\lambda\tau\mu\nu}\hat{\partial}_{\tau},\qquad \hat{\Phi}^{\lambda,\mu\nu}_{\partial}=i\eta^{\lambda\mu}\hat{\partial}^{\nu}-i\eta^{\lambda\nu}\hat{\partial}^{\mu}.

This work will focus on the noninteracting Dirac field Ψ\Psi. In this case, I will discuss the following choices for Φ^\hat{\Phi}:

Φ^Cλ,μν=Φ^ϵAλ,μν=12ϵλμντΨˉγτγ5Ψ=−i8Ψˉ{γλ,[γμ,γν]}Ψ,\begin{aligned} \hat{\Phi}^{\lambda,\mu\nu}_{C}=\hat{\Phi}^{\lambda,\mu\nu}_{\epsilon A}=\frac{1}{2}\epsilon^{\lambda\mu\nu\tau}\bar{\Psi}\gamma_{\tau}\gamma^{5}\Psi\\ &=-\frac{i}{8}\bar{\Psi}\left\{\gamma^{\lambda},[\gamma^{\mu},\gamma^{\nu}]\right\}\Psi, \end{aligned}
Φ^ϵVλ,μν=12ϵλμντΨˉγτΨ,\hat{\Phi}^{\lambda,\mu\nu}_{\epsilon V}=\frac{1}{2}\epsilon^{\lambda\mu\nu\tau}\bar{\Psi}\gamma_{\tau}\Psi,
Φ^Aλ,μν=Ψˉ(ηλμγνγ5−ηλνγμγ5)Ψ,\hat{\Phi}^{\lambda,\mu\nu}_{A}=\bar{\Psi}\left(\eta^{\lambda\mu}\gamma^{\nu}\gamma^{5}-\eta^{\lambda\nu}\gamma^{\mu}\gamma^{5}\right)\Psi,
Φ^Vλ,μν=Ψˉ(ηλμγν−ηλνγμ)Ψ.(17)\hat{\Phi}^{\lambda,\mu\nu}_{V}=\bar{\Psi}\left(\eta^{\lambda\mu}\gamma^{\nu}-\eta^{\lambda\nu}\gamma^{\mu}\right)\Psi. \tag*{(17)}

where γ\gamma are the Dirac Gamma matrices. I will also consider

Φ^ϵ∂λ,μν=i2ϵλτμνΨˉ∂↔τΨ,\hat{\Phi}^{\lambda,\mu\nu}_{\epsilon\partial}=\frac{i}{2}\epsilon^{\lambda\tau\mu\nu}\bar{\Psi}\overleftrightarrow{\partial}_{\tau}\Psi,
Φ^∂λ,μν=iΨˉ(ηλμ∂↔ν−ηλν∂↔μ)Ψ,\hat{\Phi}^{\lambda,\mu\nu}_{\partial}=i\bar{\Psi}\left(\eta^{\lambda\mu}\overleftrightarrow{\partial}^{\nu}-\eta^{\lambda\nu}\overleftrightarrow{\partial}^{\mu}\right)\Psi,
Φ^ϵ∂Aλ,μν=12ϵλτμνΨˉ∂↔τγ5Ψ,\hat{\Phi}^{\lambda,\mu\nu}_{\epsilon\partial A}=\frac{1}{2}\epsilon^{\lambda\tau\mu\nu}\bar{\Psi}\overleftrightarrow{\partial}_{\tau}\gamma_{5}\Psi,
Φ^∂Aλ,μν=Ψˉ(ηλμ∂↔ν−ηλν∂↔μ)γ5Ψ,\hat{\Phi}^{\lambda,\mu\nu}_{\partial A}=\bar{\Psi}\left(\eta^{\lambda\mu}\overleftrightarrow{\partial}^{\nu}-\eta^{\lambda\nu}\overleftrightarrow{\partial}^{\mu}\right)\gamma_{5}\Psi,
Φ^∂Σλ,μν=imΨˉ∂↔λσμνΨ,(18)\hat{\Phi}^{\lambda,\mu\nu}_{\partial\Sigma}=\frac{i}{m}\bar{\Psi}\overleftrightarrow{\partial}_{\lambda}\sigma^{\mu\nu}\Psi, \tag*{(18)}

where

∂↔μ=(∂→μ−∂←μ),σμν=i2[γμ,γν].\overleftrightarrow{\partial}^{\mu}=\left(\overrightarrow{\partial}^{\mu}-\overleftarrow{\partial}^{\mu}\right),\qquad \sigma^{\mu\nu}=\frac{i}{2}[\gamma^{\mu},\gamma^{\nu}].

Other than the canonical and Belinfante pseudogauges, other valid choices of boost-angular momentum decomposition position for the Dirac field are given by the de Groot-van Leeuwen-van Weert (GLW) and the Hilgevoord-Wouthuysen (HW) decomposition. The de Groot-van Leeuwen-van Weert EMT and spin tensor (see [42]) are obtained from the Belinfante ones with [28]

Φ^GLWλ,μν=−S^Cλ,μν+i4mΨ‾(σλμ∂↔ν−σλν∂↔μ)Ψ,Z^GLWμν,λρ=0.(19)\widehat{\Phi}_{\mathrm{GLW}}^{\lambda,\mu\nu}=-\widehat{S}_{\mathrm{C}}^{\lambda,\mu\nu}+\frac{i}{4m}\overline{\Psi}\left(\sigma^{\lambda\mu}\overleftrightarrow{\partial}^{\nu}-\sigma^{\lambda\nu}\overleftrightarrow{\partial}^{\mu}\right)\Psi, \qquad \widehat{Z}_{\mathrm{GLW}}^{\mu\nu,\lambda\rho}=0. \tag*{(19)}

Instead, the Hilgevoord-Wouthuysen EMT and spin tensor are obtained with [28, 43, 44]

Φ^HWλ,μν=Φ^GLWλ,μν−i4mΨ‾(ηλμσνα−ηλνσμα)∂↔αΨ,Z^HWμν,λρ=−18mΨ‾(σμνσλρ+σλρσμν)Ψ.(20)\widehat{\Phi}_{\mathrm{HW}}^{\lambda,\mu\nu}=\widehat{\Phi}_{\mathrm{GLW}}^{\lambda,\mu\nu}-\frac{i}{4m}\overline{\Psi}\left(\eta^{\lambda\mu}\sigma^{\nu\alpha}-\eta^{\lambda\nu}\sigma^{\mu\alpha}\right)\overleftrightarrow{\partial}_{\alpha}\Psi, \qquad \widehat{Z}_{\mathrm{HW}}^{\mu\nu,\lambda\rho}=-\frac{1}{8m}\overline{\Psi}\left(\sigma^{\mu\nu}\sigma^{\lambda\rho}+\sigma^{\lambda\rho}\sigma^{\mu\nu}\right)\Psi. \tag*{(20)}

Lastly, I will also consider the improved EMT proposed by Callan, Coleman, and Jackiew (CCJ) [45], whose matrix elements are finite at all orders of perturbation theory for a renormalizable interaction. The improved EMT is obtained with

Φ^CCJλ,μν=−16(ηλν∂μ−ηλμ∂ν)Ψ‾Ψ,Z^CCJμν,λρ=−16(ηρμηλν−ηρνηλμ)Ψ‾Ψ.(21)\widehat{\Phi}_{\mathrm{CCJ}}^{\lambda,\mu\nu}=-\frac{1}{6}\left(\eta^{\lambda\nu}\partial^\mu-\eta^{\lambda\mu}\partial^\nu\right)\overline{\Psi}\Psi, \qquad \widehat{Z}_{\mathrm{CCJ}}^{\mu\nu,\lambda\rho}=-\frac{1}{6}\left(\eta^{\rho\mu}\eta^{\lambda\nu}-\eta^{\rho\nu}\eta^{\lambda\mu}\right)\overline{\Psi}\Psi. \tag*{(21)}

so that

S^CCJλ,μν=0,T^CCJμν=T^Bμν−16(∂μ∂ν−ημν∂λ∂λ)Ψ‾Ψ.\widehat{S}_{\mathrm{CCJ}}^{\lambda,\mu\nu}=0, \qquad \widehat{T}_{\mathrm{CCJ}}^{\mu\nu}=\widehat{T}_{\mathrm{B}}^{\mu\nu}-\frac{1}{6}\left(\partial^\mu\partial^\nu-\eta^{\mu\nu}\partial_\lambda\partial^\lambda\right)\overline{\Psi}\Psi.

Global equilibrium

If the system is at the global thermal equilibrium, the statistical operator only depends on the global conserved quantities, i.e. the total momentum, boost-angular momentum and the conserved charge. Since these quantities are not affected by the PGTs, the statistical operator of global equilibrium is also pseudogauge independent [27, 28]. In what follows, I review how this comes to be starting from the local thermodynamic equilibrium statistical operator (16).

To describe a global thermal equilibrium one needs to impose that the statistical operator is actually time-independent, which occurs when the integrals in (16) are independent of the space-like hypersurfaces Σ\Sigma. This is equivalent to imposing that the divergence of the integrand is vanishing [46]:

∇μ[T^Bμνβν−12(ϖλν−Ωλν)Φ^μ,λν−ξλΦ^λ,μν−12Ωλν∇ρZ^λν,μρ−j^μζ]=0.\nabla_\mu\left[\widehat{T}_{\mathrm{B}}^{\mu\nu}\beta_\nu-\frac{1}{2}\left(\varpi_{\lambda\nu}-\Omega_{\lambda\nu}\right)\widehat{\Phi}^{\mu,\lambda\nu}-\xi_\lambda\widehat{\Phi}^{\lambda,\mu\nu}-\frac{1}{2}\Omega_{\lambda\nu}\nabla_\rho\widehat{Z}^{\lambda\nu,\mu\rho}-\hat{j}^{\mu}\zeta\right]=0.

Taking advantage of the operatorial equations,

∇μT^Bμνβν=0,∇μj^μ=0,\nabla_\mu\widehat{T}_{\mathrm{B}}^{\mu\nu}\beta_\nu=0,\qquad\nabla_\mu\hat{j}^{\mu}=0,
∇μΦ^μ,λν=T^Φλν−T^Φνλ,∇μ∇ρZ^λν,μρ=0,\nabla_\mu\widehat{\Phi}^{\mu,\lambda\nu}=\widehat{T}_{\Phi}^{\lambda\nu}-\widehat{T}_{\Phi}^{\nu\lambda},\qquad\nabla_\mu\nabla_\rho\widehat{Z}^{\lambda\nu,\mu\rho}=0,

the divergence of the integrand is written as

T^Bμν∇μβν−12(ϖλν−Ωλν)(T^Φλν−T^Φνλ)+Φ^μ,λν∇μ(ϖλν−Ωλν)−12(∇ρZ^λν,μρ)∇μΩλν−ξλ∇μΦ^λ,μν−Φ^λ,μν∇μξλ−j^μ∇μζ=0.\widehat{T}_{\mathrm{B}}^{\mu\nu}\nabla_\mu\beta_\nu-\frac{1}{2}\left(\varpi_{\lambda\nu}-\Omega_{\lambda\nu}\right)\left(\widehat{T}_{\Phi}^{\lambda\nu}-\widehat{T}_{\Phi}^{\nu\lambda}\right)+\widehat{\Phi}^{\mu,\lambda\nu}\nabla_\mu\left(\varpi_{\lambda\nu}-\Omega_{\lambda\nu}\right)-\frac{1}{2}\left(\nabla_\rho\widehat{Z}^{\lambda\nu,\mu\rho}\right)\nabla_\mu\Omega_{\lambda\nu}-\xi_\lambda\nabla_\mu\widehat{\Phi}^{\lambda,\mu\nu}-\widehat{\Phi}^{\lambda,\mu\nu}\nabla_\mu\xi_\lambda-\hat{j}^{\mu}\nabla_\mu\zeta=0.

The last term of the divergence is vanishing if ∇μζ=0\nabla_\mu\zeta=0. In flat space-time this condition sets ζ\zeta to be a constant. The first term is vanishing if β\beta is a Killing field, i.e., if it solves

∇μβν+∇νβμ=0.\nabla_\mu\beta_\nu+\nabla_\nu\beta_\mu=0.

In flat space-time the most general solution of the Killing equation is given by

βμ=bμ+ϖμνxν\beta^\mu=b^\mu+\varpi^{\mu\nu}x_\nu

with bb a constant four-vector and ϖ\varpi a constant anti-symmetric rank-2 tensor. This global equilibrium form of β\beta implies ξλ=0\xi_\lambda=0 and ∇μξλ=0\nabla_\mu\xi_\lambda=0. Then, the integrand is vanishing under the following conditions:

{∇μ(ϖλν−Ωλν)=∇μΩλν=0,(ϖλν−Ωλν)=0,\left\{ \begin{aligned} \nabla_\mu\left(\varpi_{\lambda\nu}-\Omega_{\lambda\nu}\right)&=\nabla_\mu\Omega_{\lambda\nu}=0,\\ \left(\varpi_{\lambda\nu}-\Omega_{\lambda\nu}\right)&=0, \end{aligned} \right.

which are both satisfied for Ωμν=ϖμν\Omega_{\mu\nu}=\varpi_{\mu\nu}. The form of the statistical operator at global equilibrium is then obtained by replacing these global equilibrium forms for the thermal fields β\beta, Ω\Omega, and ζ\zeta in the statistical operator (16). This form is the following:

ρ^GE=1Zexp⁡{−∫dΣμ[T^Bμν(bν+ϖνλxλ)−j^μζ]}=1Zexp⁡{−bν∫dΣμT^Bμν−ϖνλ∫dΣμT^Bμνxλ+ζ∫dΣμj^μ}.\begin{aligned} \hat{\rho}_{\mathrm{GE}}=\frac{1}{Z}\exp\left\{-\int d\Sigma_\mu\left[\widehat{T}^{\mu\nu}_{\mathrm{B}}\left(b_\nu+\varpi_{\nu\lambda}x^\lambda\right)-\hat{j}^{\mu}\zeta\right]\right\} \\ &=\frac{1}{Z}\exp\left\{-b_\nu\int d\Sigma_\mu\widehat{T}^{\mu\nu}_{\mathrm{B}}-\varpi_{\nu\lambda}\int d\Sigma_\mu\widehat{T}^{\mu\nu}_{\mathrm{B}}x^\lambda+\zeta\int d\Sigma_\mu\hat{j}^{\mu}\right\}. \end{aligned}

In the first and last terms of the last line, one can readily recognize the total momentum operator P^\widehat{P} and the charge Q^\widehat{Q}, respectively. In the second term, by taking advantage of the anti-symmetric indices, the total boost-angular momentum J^\widehat{J} is obtained. The global equilibrium statistical operator is

ρ^GE=1Zexp⁡[−b⋅P^+12ϖ:J^+ζQ^].(22)\hat{\rho}_{\mathrm{GE}}=\frac{1}{Z}\exp\left[-b\cdot\widehat{P}+\frac{1}{2}\varpi\mathbin{:}\widehat{J}+\zeta\widehat{Q}\right]. \tag*{(22)}

This derivation shows that no matter what the form of the pseudogauge fields Φ^\widehat{\Phi} and Z^\widehat{Z} is, the global equilibrium statistical operator will always have the same form; therefore ρ^GE\hat{\rho}_{\mathrm{GE}} is pseudogauge independent. Notice that if one chooses to constraint only the EMT but not the spin tensor, one would have to set Ω=0\Omega=0. Therefore, when imposing the global equilibrium, one would obtain only the particular case of the previous operator with ϖ=0\varpi=0. The only exceptions are the pseudogauges with a vanishing spin current, such as the Belinfante one. In these cases, since all the spin information is contained in the EMT, one still obtains the global equilibrium with non-vanishing thermal vorticity without introducing a spin potential [39].

It follows that the PGTs do not affect the predictions of spin polarization and of the axial vortical effect (AVE) when the system is at global thermal equilibrium. The spin polarization at the first order of thermal vorticity of a Dirac particle with momentum kk in a system at global thermal equilibrium is

Sμ(k)=−12mϵμρστkτ∫ΣdΣ⋅k nF(1−nF)ϖρσ∫ΣdΣ⋅k nFS^\mu(k)=-\frac{1}{2m}\epsilon^{\mu\rho\sigma\tau}k_\tau\frac{\int_\Sigma d\Sigma\cdot k\,n_{\mathrm{F}}(1-n_{\mathrm{F}})\varpi_{\rho\sigma}}{\int_\Sigma d\Sigma\cdot k\,n_{\mathrm{F}}}

which has the same form of Eq. (3), the only difference being that at global equilibrium the thermal vorticity must be a constant tensor. For the spin polarization at all orders of thermal vorticity see [47].

Similarly, for the Dirac field, the mean axial current resulting from the Zubarev statistical operator is discussed both for massive [48, 49, 50] and massless fermions [39, 49, 51]. The result is that the mean axial current is

⟨j^Aμ⟩GE=WAωμ+O(ϖ2)(23)\langle\widehat{j}^{\mu}_{A}\rangle_{\mathrm{GE}}=W^{A}\omega^{\mu}+O(\varpi^{2}) \tag*{(23)}

where the four-vector ww is defined by

ωμ=−12ϵμρσλϖρσuλ=wμT,(24)\omega^\mu=-\frac{1}{2}\epsilon^{\mu\rho\sigma\lambda}\varpi_{\rho\sigma}u_\lambda=\frac{w^\mu}{T}, \tag*{(24)}

which is the local rotation of the fluid,

ωμ=−12ϵμνρσ∂νuρuσ,\omega^\mu=-\frac{1}{2}\epsilon^{\mu\nu\rho\sigma}\partial_\nu u_\rho u_\sigma,

divided by the temperature. The thermal coefficient WAW^A is the AVE conductivity, and for a free Dirac field is given by [51, 52]

WA=∫0∞dk2π2βεk2+k2εk[nF(βεk−ζ)+nF(βεk+ζ)](25)W^A=\int_0^\infty\frac{dk}{2\pi^2\beta}\frac{\varepsilon_k^2+k^2}{\varepsilon_k}\left[n_{\mathrm{F}}(\beta\varepsilon_k-\zeta)+n_{\mathrm{F}}(\beta\varepsilon_k+\zeta)\right] \tag*{(25)}

with β=β2\beta=\sqrt{\beta^2}, εk2=k2+m2\varepsilon_k^2=\mathbf{k}^2+m^2, and mm being the mass of the field. For a massless field, the AVE conductivity above becomes

WA=T36+μ22π2T,W^A=\frac{T^3}{6}+\frac{\mu^2}{2\pi^2}T,

where μ=ζT\mu=\zeta T is the chemical potential. The mean axial current (23) and the conductivity (25) were obtained with the global equilibrium statistical operator (22) and do not depend on the pseudogauge. In Sec. V I obtain how the AVE (23) is modified in a system at local thermal equilibrium with a generic pseudogauge.

Pseudogauge Dependence of Wigner Function

The Wigner function extends the concept of the classical distribution function to a quantum system. Like the distribution function, from the Wigner function one can obtain many relevant expectation values in statistical quantum field theory. As it is generally expected that different distribution functions result in different averages, a difference in the Wigner function may result in a variation in the final expectation values. In this section I derive the effect of the pseudogauge transformations (PGTs) in the Wigner function of the free Dirac field. I will show that different choices of Φ^\widehat{\Phi} and Z^\widehat{Z} in (16) affect the Wigner function. In Secs. IV and V I will show that indeed these different Wigner functions give different results for the spin polarization and the mean axial current.

The covariant Wigner function of a free Dirac field is defined as [42]

WAB(x,k)=tr⁡(ρ^ W^AB(x,k)),(26)W_{AB}(x,k)=\operatorname{tr}\left(\widehat{\rho}\,\widehat{W}_{AB}(x,k)\right), \tag*{(26)}

where A,BA,B denote the spinorial indices, and W^\widehat{W} denotes the Wigner operator,

W^AB(x,k)=∫d4y(2π)4e−ik⋅y:Ψ‾B(x+y2)ΨA(x−y2): ;\widehat{W}_{AB}(x,k)=\int\frac{d^4y}{(2\pi)^4}e^{-ik\cdot y}:\overline{\Psi}_{B}\left(x+\frac{y}{2}\right)\Psi_{A}\left(x-\frac{y}{2}\right):\ ;

and the colons “::::” denotes the normal ordering. The mean axial current can be obtained once the Wigner function is known by performing the integral

jAμ(x)=⟨j^Aμ(x)⟩=∫d4k tr⁡4[γμγ5W(x,k)],j^\mu_A(x)=\langle\widehat{j}^\mu_A(x)\rangle=\int d^4k\,\operatorname{tr}_4\left[\gamma^\mu\gamma^5W(x,k)\right],

where tr⁡4\operatorname{tr}_4 indicates the trace over the spinorial indices. Similarly, the mean spin polarization of a particle state with momentum kk is obtained through [31]

Sμ(k)=12∫ΣdΣ⋅k tr⁡4[γμγ5W+(x,k)]∫ΣdΣ⋅k tr⁡4[W+(x,k)],S^\mu(k)=\frac{1}{2}\frac{\int_\Sigma d\Sigma\cdot k\,\operatorname{tr}_4\left[\gamma^\mu\gamma^5W^+(x,k)\right]}{\int_\Sigma d\Sigma\cdot k\,\operatorname{tr}_4\left[W^+(x,k)\right]},

where W+W^+ is the future time-like part (that is the particle part) of the Wigner function, obtained by the decomposition

W(x,k)=θ(k2)θ(k0)W+(x,k)+θ(k2)θ(−k0)W−(x,k)+θ(−k2)WS(x,k).(27)W(x,k)=\theta(k^2)\theta(k^0)W^+(x,k)+\theta(k^2)\theta(-k^0)W^-(x,k)+\theta(-k^2)W^S(x,k). \tag*{(27)}

Notice that the Wigner function (26) is affected by the statistical operator ρ^\widehat{\rho} that describes the thermal state of the system. I am now going to plug the pseudogauge dependent statistical operator (16) in Eq. (26). Then, I will quantify the pseudogauge dependence of the Wigner function using the linear response theory.

In this work I am going to consider the pseudogauge potentials Φ^\widehat{\Phi} given in Eqs. (17, 18, 19, and 20). All of these operators are bilinears in the Dirac fields. Therefore, recalling that

i2Ψ‾(x)(∂→μ−∂←μ)Ψ(x)=∫d4k kμW^(x,k),\frac{i}{2}\overline{\Psi}(x)\left(\overrightarrow{\partial}^{\mu}-\overleftarrow{\partial}^{\mu}\right)\Psi(x)=\int d^4k\,k^\mu\widehat{W}(x,k),

they can be obtained from the Wigner operator through the integral

Φ^λ,μν(x)=∫d4k′ Φ^B′A′λ,μν(k′)W^A′B′(x,k′)\widehat{\Phi}^{\lambda,\mu\nu}(x)=\int d^{4}k'\,\widehat{\Phi}^{\lambda,\mu\nu}_{B'A'}(k')\widehat{W}_{A'B'}(x,k')

where Φ^B′A′λ,μν\widehat{\Phi}^{\lambda,\mu\nu}_{B'A'} is a function that might depend on the gamma matrices and on the metric tensors, but that does not contain the derivative operator. See the Appendix A for the list of all the functions Φ\Phi used in this work. Among the gauge-potentials (17 - 20) only the Hilgevoord-Wouthuysen (HW) decomposition has a non-vanishing Z^\widehat{Z} potential. The pseudogauge-dependent statistical operator (16) only depends on the derivative of Z^\widehat{Z}, which for the form in (20) can be written as

∂αZ^HWρσ,λα(x)=−18m(σρσσλα+σλασρσ)B′A′∫d4k′ ∂αW^A′B′(x,k′)\partial_{\alpha}\widehat{Z}^{\rho\sigma,\lambda\alpha}_{\mathrm{HW}}(x)=-\frac{1}{8m}\left(\sigma^{\rho\sigma}\sigma^{\lambda\alpha}+\sigma^{\lambda\alpha}\sigma^{\rho\sigma}\right)_{B'A'}\int d^{4}k'\,\partial_{\alpha}\widehat{W}_{A'B'}(x,k')

Lastly, the improved Callan, Coleman, and Jackiew (CCJ) pseudogauge potentials (21) can be obtained from the Wigner operator by

Φ^CCJλ,μν(x)=−16(ηλν∂μ−ηλμ∂ν)∫d4k′ tr⁡4(W^(x,k′)),\widehat{\Phi}^{\lambda,\mu\nu}_{\mathrm{CCJ}}(x)=-\frac{1}{6}\left(\eta^{\lambda\nu}\partial^{\mu}-\eta^{\lambda\mu}\partial^{\nu}\right)\int d^{4}k'\,\operatorname{tr}_{4}\left(\widehat{W}(x,k')\right),
∂αZ^CCJρσ,λα=−16(ηαρηλσ−ηασηλρ)∫d4k′ ∂αtr⁡4(W^(x,k′))(28)\partial_{\alpha}\widehat{Z}^{\rho\sigma,\lambda\alpha}_{\mathrm{CCJ}}=-\frac{1}{6}\left(\eta^{\alpha\rho}\eta^{\lambda\sigma}-\eta^{\alpha\sigma}\eta^{\lambda\rho}\right)\int d^{4}k'\,\partial_{\alpha}\operatorname{tr}_{4}\left(\widehat{W}(x,k')\right) \tag*{(28)}

However, eventually it is found that the pseudogauge potentials Φ^CCJ\widehat{\Phi}_{\mathrm{CCJ}} and Z^CCJ\widehat{Z}_{\mathrm{CCJ}} in (21) do not affect the Wigner function at first order in linear response theory.

As mentioned before, the pseudogauge-dependent Wigner function at local thermal equilibrium is obtained by

⟨W^AB(x,k)⟩Φ=tr⁡(ρ^LTEΦW^AB(x,k))\langle\widehat{W}_{AB}(x,k)\rangle_{\Phi}=\operatorname{tr}\left(\widehat{\rho}^{\Phi}_{\mathrm{LTE}}\widehat{W}_{AB}(x,k)\right)

where, setting ζ=0\zeta=0 for clarity, and introducing the shorthand notation

Θμν≡ϖμν−Ωμν(29)\Theta_{\mu\nu}\equiv\varpi_{\mu\nu}-\Omega_{\mu\nu} \tag*{(29)}

one has

ρ^LTEΦ=1Zexp⁡{−∫dΣλ[T^Bλνβν−12ΘρσΦ^λ,ρσ−ξρσΦ^ρ,λσ−12Ωρσ∂αZ^ρσ,λα]}(30)\widehat{\rho}^{\Phi}_{\mathrm{LTE}}=\frac{1}{Z}\exp\left\{-\int d\Sigma_{\lambda}\left[\widehat{T}^{\lambda\nu}_{B}\beta_{\nu}-\frac{1}{2}\Theta_{\rho\sigma}\widehat{\Phi}^{\lambda,\rho\sigma}-\xi_{\rho\sigma}\widehat{\Phi}^{\rho,\lambda\sigma}-\frac{1}{2}\Omega_{\rho\sigma}\partial_{\alpha}\widehat{Z}^{\rho\sigma,\lambda\alpha}\right]\right\} \tag*{(30)}

and the operators Φ^\widehat{\Phi} and ∂αZ^ρσ,λα\partial_{\alpha}\widehat{Z}^{\rho\sigma,\lambda\alpha} can be written as described above. For a fluid in the hydrodynamic regime, such as the QGP, the thermodynamics field β\beta slowly varies compared to the correlation lengths between two operators. Furthermore, since no effect of the spin potential Ω\Omega was ever observed, one assumes that it is not larger than the thermal vorticity ϖ\varpi. Therefore the terms in the exponent of (30) which are coupled to Θ\Theta, ξ\xi, and Ω\Omega can be treated as “small” perturbations.

One should now study the difference between the Wigner function resulting from a generic pseudogauge and the one obtained in the Belinfante decomposition, which is denoted as

⟨W^AB(x,k)⟩B=tr⁡(ρ^BW^AB(x,k))\langle\widehat{W}_{AB}(x,k)\rangle_{B}=\operatorname{tr}\left(\widehat{\rho}_{B}\widehat{W}_{AB}(x,k)\right)

For the sake of clarity, I report only the steps for the particle part of the Wigner function W+W^{+}; see Eq. (27). The other parts are carried out in a similar fashion. Denoting the thermal average of an operator O^\widehat{O} made with the statistical operator

ρ^β=1Zexp⁡{−βν(x)∫dΣλT^Bλν}=1Zexp⁡{−β(x)⋅P^}\widehat{\rho}_{\beta}=\frac{1}{Z}\exp\left\{-\beta_{\nu}(x)\int d\Sigma_{\lambda}\widehat{T}^{\lambda\nu}_{B}\right\}=\frac{1}{Z}\exp\left\{-\beta(x)\cdot\widehat{P}\right\}

as ⟨O^⟩β(x)=tr⁡(ρ^βO^)\langle\widehat{O}\rangle_{\beta(x)}=\operatorname{tr}(\widehat{\rho}_{\beta}\widehat{O}) and using the linear response theory as described in [5, 37, 39, 41], one obtains:

ΔΦWAB+(x,k)=⟨W^AB+(x,k)⟩Φ−⟨W^AB+(x,k)⟩B≃ΔΘWAB+(x,k)+ΔξWAB+(x,k)+ΔΩZWAB+(x,k),\Delta_{\Phi}W^{+}_{AB}(x,k)=\langle\widehat{W}^{+}_{AB}(x,k)\rangle_{\Phi}-\langle\widehat{W}^{+}_{AB}(x,k)\rangle_{B} \simeq\Delta_{\Theta}W^{+}_{AB}(x,k)+\Delta_{\xi}W^{+}_{AB}(x,k)+\Delta_{\Omega Z}W^{+}_{AB}(x,k),

where

ΔΘWAB+(x,k)=12∫01dz∫ΣdΣλ(y)Θρσ(y)∫d4k′Φ^B′A′λ,ρσ⟨W^AB+(x,k)W^A′B′(y+izβ(x),k′)⟩β(x),\Delta_{\Theta}W^{+}_{AB}(x,k)=\frac{1}{2}\int_{0}^{1}dz\int_{\Sigma}d\Sigma_{\lambda}(y)\Theta_{\rho\sigma}(y)\int d^{4}k'\widehat{\Phi}^{\lambda,\rho\sigma}_{B'A'}\langle\widehat{W}^{+}_{AB}(x,k)\widehat{W}_{A'B'}(y+i z\beta(x),k')\rangle_{\beta(x)},
ΔξWAB+(x,k)=∫01dz∫ΣdΣλ(y)ξρσ(y)∫d4k′Φ^B′A′ρ,λσ⟨W^AB+(x,k)W^A′B′(y+izβ(x),k′)⟩β(x),\Delta_{\xi}W^{+}_{AB}(x,k)=\int_{0}^{1}dz\int_{\Sigma}d\Sigma_{\lambda}(y)\xi_{\rho\sigma}(y)\int d^{4}k'\widehat{\Phi}^{\rho,\lambda\sigma}_{B'A'}\langle\widehat{W}^{+}_{AB}(x,k)\widehat{W}_{A'B'}(y+i z\beta(x),k')\rangle_{\beta(x)},
ΔΩZWAB+(x,k)=12∫01dz∫ΣdΣλ(y)Θρσ(y)∫d4k′Z^B′A′ρσ,λα∂yα⟨W^AB+(x,k)W^A′B′(y+izβ(x),k′)⟩β(x).\Delta_{\Omega Z}W^{+}_{AB}(x,k)=\frac{1}{2}\int_{0}^{1}dz\int_{\Sigma}d\Sigma_{\lambda}(y)\Theta_{\rho\sigma}(y)\int d^{4}k'\widehat{Z}^{\rho\sigma,\lambda\alpha}_{B'A'}\partial_{y^{\alpha}}\langle\widehat{W}^{+}_{AB}(x,k)\widehat{W}_{A'B'}(y+i z\beta(x),k')\rangle_{\beta(x)}.

Using standard thermal field theory techniques (see for instance [5, 53]), one finds

\left\langle\widehat{W}^{+}_{ab}(x,k)\widehat{W}_{cd}(y+iz\beta(x),k')\right\rangle_{c,\beta(x)}=\frac{1}{(2\pi)^6}\int\frac{\mathrm{d}^3p}{2\varepsilon_p}\int\frac{\mathrm{d}^3p'}{2\varepsilon_{p'}}\delta^4\left(k-\frac{p+p'}{2}\right)\delta^4\left(k'-\frac{p+p'}{2}\right) \left(\slashed{p}'+m\right)_{ad}\left(\slashed{p}+m\right)_{cb}e^{i(p-p')(x-y)}e^{z(p-p')\beta}n_{\mathrm{F}}(\beta(x)\cdot p)\left(1-n_{\mathrm{F}}(\beta(x)\cdot p')\right)

from which it straightforwardly follows that

\Delta_{\Theta}W^{+}_{AB}(x,k)=\frac{1}{2}\int_0^1\mathrm{d}z\int_{\Sigma}\mathrm{d}\Sigma_{\lambda}(y)\Theta_{\rho\sigma}(y)\frac{1}{(2\pi)^6}\int\frac{\mathrm{d}^3p}{2\varepsilon_p}\int\frac{\mathrm{d}^3p'}{2\varepsilon_{p'}}\delta^4\left(k-\frac{p+p'}{2}\right) \left[(\slashed{p}'+m)\Phi^{\lambda,\rho\sigma}(\slashed{p}+m)\right]_{AB}e^{i(p-p')(x-y)}e^{z(p-p')\beta}n_{\mathrm{F}}(\beta(x)\cdot p)\left(1-n_{\mathrm{F}}(\beta(x)\cdot p')\right),
\Delta_{\xi}W^{+}_{AB}(x,k)=\int_0^1\mathrm{d}z\int_{\Sigma}\mathrm{d}\Sigma_{\lambda}(y)\xi_{\rho\sigma}(y)\frac{1}{(2\pi)^6}\int\frac{\mathrm{d}^3p}{2\varepsilon_p}\int\frac{\mathrm{d}^3p'}{2\varepsilon_{p'}}\delta^4\left(k-\frac{p+p'}{2}\right) \left[(\slashed{p}'+m)\Phi^{\rho,\lambda\sigma}(\slashed{p}+m)\right]_{AB}e^{i(p-p')(x-y)}e^{z(p-p')\beta}n_{\mathrm{F}}(\beta(x)\cdot p)\left(1-n_{\mathrm{F}}(\beta(x)\cdot p')\right),
\Delta_{\Omega Z}W^{+}_{AB}(x,k)=-\frac{i}{2}\int_0^1\mathrm{d}z\int_{\Sigma}\mathrm{d}\Sigma_{\lambda}(y)\Omega_{\rho\sigma}(y)\frac{1}{(2\pi)^6}\int\frac{\mathrm{d}^3p}{2\varepsilon_p}\int\frac{\mathrm{d}^3p'}{2\varepsilon_{p'}}\delta^4\left(k-\frac{p+p'}{2}\right) \left[(\slashed{p}'+m)Z^{\rho\sigma,\lambda\alpha}(\slashed{p}+m)\right]_{AB}(p-p')_{\alpha}e^{i(p-p')(x-y)}e^{z(p-p')\beta}n_{\mathrm{F}}(\beta(x)\cdot p)\left(1-n_{\mathrm{F}}(\beta(x)\cdot p')\right). \tag*{(31)}

Notice that the integration over the space-like hypersurface Σ\Sigma, i.e., in the yy coordinates, only involves the exponential ei(p−p′)(x−y)e^{i(p-p')(x-y)} and the thermodynamic fields Θ\Theta, ξ\xi, and Ω\Omega. As mentioned above, in the hydrodynamic regime the thermodynamic fields are slowly varying, and the integral over Σ\Sigma can be approximated by Taylor expanding the thermodynamic fields around the point y=xy=x and retaining only the leading order. For instance, the expansion of Θ\Theta is

Θρσ(y)=Θρσ(x)+∂κΘρσ(x)(y−x)κ+⋯ ,\Theta_{\rho\sigma}(y)=\Theta_{\rho\sigma}(x)+\partial_{\kappa}\Theta_{\rho\sigma}(x)(y-x)^{\kappa}+\cdots,

and one obtains (see [5, 53])

∫ΣdΣλ(y)Θρσ(y)ei(p−p′)(x−y)≃Θρσ(x)t^λ(2π)3δ3(p−p′).(32)\int_{\Sigma}\mathrm{d}\Sigma_{\lambda}(y)\Theta_{\rho\sigma}(y)e^{i(p-p')(x-y)}\simeq\Theta_{\rho\sigma}(x)\hat{t}_{\lambda}(2\pi)^3\delta^3(\mathbf{p}-\mathbf{p}'). \tag*{(32)}

where t^\hat{t} is the time direction in the fluid frame. After this approximation, thanks to the delta function, it is straightforward to integrate the (31) in p′p'. As a result, ΔΩZW+\Delta_{\Omega Z}W^{+} is vanishing because of the factor p−p′p-p', while in the other contributions the dependence on zz goes away. Lastly, taking advantage of the transformation

∫d3p2εpδ4(k−p)f(p)=∫d4kδ(k2−m2)θ(k0)δ4(k−p)f(p)=θ(k0)δ(k2−m2)f(k),\int\frac{\mathrm{d}^3p}{2\varepsilon_p}\delta^4(k-p)f(p)=\int\mathrm{d}^4k\delta(k^2-m^2)\theta(k_0)\delta^4(k-p)f(p)=\theta(k_0)\delta(k^2-m^2)f(k),

these final expressions are obtained:

\Delta_{\Theta}W^{+}_{AB}(x,k)=\frac{\theta(k_0)\delta(k^2-m^2)}{4(2\pi)^3}\frac{\hat{t}^{\lambda}\Theta_{\rho\sigma}(x)}{\varepsilon_k}\left[(\slashed{k}+m)\Phi^{\lambda,\rho\sigma}(\slashed{k}+m)\right]_{AB}n_{\mathrm{F}}(\beta(x)\cdot k)\left(1-n_{\mathrm{F}}(\beta(x)\cdot k)\right),
\Delta_{\xi}W^{+}_{AB}(x,k)=\frac{\theta(k_0)\delta(k^2-m^2)}{2(2\pi)^3}\frac{\hat{t}^{\lambda}\xi_{\rho\sigma}(x)}{\varepsilon_k}\left[(\slashed{k}+m)\Phi^{\rho,\lambda\sigma}(\slashed{k}+m)\right]_{AB}n_{\mathrm{F}}(\beta(x)\cdot k)\left(1-n_{\mathrm{F}}(\beta(x)\cdot k)\right),
ΔΩZWAB+(x,k)=0.(33)\Delta_{\Omega Z}W^{+}_{AB}(x,k)=0. \tag*{(33)}

(33) clearly shows that the Wigner function of local thermal equilibrium does depend on the choice of pseudogauge potentials. In some cases, these corrections might be vanishing. For instance, if Φ^ρ,λσ\widehat{\Phi}^{\rho,\lambda\sigma} is completely anti-symmetric, such as in Φ^C\widehat{\Phi}_{C}, Φ^ϵA\widehat{\Phi}_{\epsilon A}, Φ^ϵV\widehat{\Phi}_{\epsilon V}, Φ^ϵ∂\widehat{\Phi}_{\epsilon\partial}, Φ^ϵ∂A\widehat{\Phi}_{\epsilon\partial A}, then one readily has that ΔξWAB+(x,k)=0\Delta_{\xi}W^{+}_{AB}(x,k)=0. One could also repeat the argument above for the improved Callan,

Coleman, and Jackiew pseudogauge potentials (21) using (28). As in ΔΩZW+\Delta_{\Omega Z}W^{+}, the derivatives of Wigner function in (28) create factors p−p′p-p' which vanish when integrating the Dirac delta from the approximation (32). Therefore, at first order in linear perturbation theory, the Callan, Coleman, and Jackiew decomposition gives the same thermal expectation values as the Belinfante decomposition. In the next sections I show that this difference in the Wigner function actually affects physical observables.

Pseudogauge dependence of spin polarization

The spin polarization is obtained from the Wigner function by [31]

Sμ(k)=12∫ΣdΣ⋅k tr⁡4[γμγ5W+(x,k)]∫ΣdΣαkαtr⁡4[W+(x,k)].S^\mu(k)=\frac{1}{2}\frac{\int_\Sigma d\Sigma\cdot k\,\operatorname{tr}_4\left[\gamma^\mu\gamma^5W^+(x,k)\right]}{\int_\Sigma d\Sigma_\alpha k^\alpha\operatorname{tr}_4\left[W^+(x,k)\right]}.

It is important to stress that the form of this formula does not depend on the pseudogauge. The pseudogauge dependence of spin polarization from the previous formula is completely contained inside the Wigner function, which has itself inherited the pseudogauge dependence from the statistical operator. As mentioned, the spin polarization in the Belinfante PG (1) is given by

SBμ(k)≃Sϖμ(k)+Sξμ(k);(34)S^\mu_{\mathrm{B}}(k)\simeq S^\mu_{\varpi}(k)+S^\mu_{\xi}(k); \tag*{(34)}

see Eqs. (3) and (4). Under the hypothesis that linear response theory is a good approximation, the difference between the spin polarization in a generic pseudogauge SΦμ(k)S^\mu_{\Phi}(k) and in the Belinfante SBμ(k)S^\mu_{\mathrm{B}}(k) can be obtained from the results of the previous section. Using the above formula, it is given by:

ΔΦSμ(k)=SΦμ(k)−SBμ(k)=12D∫ΣdΣ⋅k tr⁡4[γμγ5ΔΦW+(x,k)],(35)\begin{aligned} \Delta_{\Phi}S^\mu(k)=S^\mu_{\Phi}(k)-S^\mu_{\mathrm{B}}(k) \\ &=\frac{1}{2\mathcal{D}}\int_\Sigma d\Sigma\cdot k\,\operatorname{tr}_4\left[\gamma^\mu\gamma^5\Delta_{\Phi}W^+(x,k)\right], \tag*{(35)} \end{aligned}

with

D=∫ΣdΣαkαtr⁡4[⟨W^+(x,k)⟩β(x)]=4m(2π)3∫ΣdΣ⋅k δ(k2−m2)θ(k0)nF,\begin{aligned} \mathcal{D}=\int_\Sigma d\Sigma_\alpha k^\alpha\operatorname{tr}_4\left[\left\langle\widehat{W}^+(x,k)\right\rangle_{\beta(x)}\right] \\ &=\frac{4m}{(2\pi)^3}\int_\Sigma d\Sigma\cdot k\,\delta(k^2-m^2)\theta(k_0)n_{\mathrm{F}}, \end{aligned}

where nF=nF(β(x)⋅k)n_{\mathrm{F}}=n_{\mathrm{F}}(\beta(x)\cdot k). Plugging in the results of (33) into (35), we obtain

ΔΦSμ(k)=ΔΘSμ(k)+ΔξSμ(k),\Delta_{\Phi}S^\mu(k)=\Delta_{\Theta}S^\mu(k)+\Delta_{\xi}S^\mu(k),
ΔΘSμ(k)=AΘ,Φμρσλtλ32mϵk∫ΣdΣ⋅k ΘρσnF(1−nF)∫ΣdΣ⋅k nF,\Delta_{\Theta}S^\mu(k)=\frac{A^{\mu\rho\sigma\lambda}_{\Theta,\Phi}t_\lambda}{32m\epsilon_k}\frac{\int_\Sigma d\Sigma\cdot k\,\Theta_{\rho\sigma}n_{\mathrm{F}}(1-n_{\mathrm{F}})}{\int_\Sigma d\Sigma\cdot k\,n_{\mathrm{F}}},
ΔξSμ(k)=Aξ,Φμρσλtλ16mϵk∫ΣdΣ⋅k ξρσnF(1−nF)∫ΣdΣ⋅k nF,(36)\Delta_{\xi}S^\mu(k)=\frac{A^{\mu\rho\sigma\lambda}_{\xi,\Phi}t_\lambda}{16m\epsilon_k}\frac{\int_\Sigma d\Sigma\cdot k\,\xi_{\rho\sigma}n_{\mathrm{F}}(1-n_{\mathrm{F}})}{\int_\Sigma d\Sigma\cdot k\,n_{\mathrm{F}}}, \tag*{(36)}

where

AΘ,Φμρσλ=tr⁡4[γμγ5(k̸+m)Φλ,ρσ(k̸+m)],A^{\mu\rho\sigma\lambda}_{\Theta,\Phi}=\operatorname{tr}_4\left[\gamma^\mu\gamma^5(\not{k}+m)\Phi^{\lambda,\rho\sigma}(\not{k}+m)\right],
Aξ,Φμρσλ=tr⁡4[γμγ5(k̸+m)Φρ,λσ(k̸+m)].A^{\mu\rho\sigma\lambda}_{\xi,\Phi}=\operatorname{tr}_4\left[\gamma^\mu\gamma^5(\not{k}+m)\Phi^{\rho,\lambda\sigma}(\not{k}+m)\right].

These are ordinary gamma matrix traces and result in

AΘ,Φμρσλ={4ϵλρστ(kμkτ−η τμm2),ΦC=ΦϵA,16(kμkσηλρ+m2ημρηλσ),ΦA,−16ϵμρστkτkλ,Φ∂Σ,AΘ,ΦCμρσλ+8ϵλμρτkτkσ,ΦGLW and ΦHW,(37)A^{\mu\rho\sigma\lambda}_{\Theta,\Phi}= \begin{cases} 4\epsilon^{\lambda\rho\sigma\tau}(k^\mu k_\tau-\eta^\mu_{\ \tau}m^2), & \Phi_{\mathrm{C}}=\Phi_{\epsilon_{\mathrm{A}}},\\ 16(k^\mu k^\sigma\eta^{\lambda\rho}+m^2\eta^{\mu\rho}\eta^{\lambda\sigma}), & \Phi_{\mathrm{A}},\\ -16\epsilon^{\mu\rho\sigma\tau}k_\tau k^\lambda, & \Phi_{\partial\Sigma},\\ A^{\mu\rho\sigma\lambda}_{\Theta,\Phi_{\mathrm{C}}}+8\epsilon^{\lambda\mu\rho\tau}k_\tau k^\sigma, & \Phi_{\mathrm{GLW}}\text{ and }\Phi_{\mathrm{HW}}, \end{cases} \tag*{(37)}

and

Aξ,Φμρσλ={0,ΦC=ΦϵA,8(kμkσ−m2ημσ)ηλρ−8(kμkλ−m2ημλ)ηρσ,ΦA,16ϵλμστkτkρ,Φ∂Σ,−4ϵλμστkτkρ,ΦGLW and ΦHW,(38)A^{\mu\rho\sigma\lambda}_{\xi,\Phi}= \begin{cases} 0, & \Phi_{\mathrm{C}}=\Phi_{\epsilon_{\mathrm{A}}},\\ 8(k^\mu k^\sigma-m^2\eta^{\mu\sigma})\eta^{\lambda\rho}-8(k^\mu k^\lambda-m^2\eta^{\mu\lambda})\eta^{\rho\sigma}, & \Phi_{\mathrm{A}},\\ 16\epsilon^{\lambda\mu\sigma\tau}k_\tau k^\rho, & \Phi_{\partial\Sigma},\\ -4\epsilon^{\lambda\mu\sigma\tau}k_\tau k^\rho, & \Phi_{\mathrm{GLW}}\text{ and }\Phi_{\mathrm{HW}}, \end{cases} \tag*{(38)}

and are vanishing for the other pseudogauges considered in this work, (17 - 21).

Canonical decomposition

It is worth discussing some special cases of the general result (36). If one chooses to describe the local thermodynamic equilibrium with the canonical decomposition, one should use the statistical operator

ρ^LTEC=1Zexp⁡[−∫dΣμ(T^Cμνβν−12ΩλνS^Cμ,λν)].\widehat{\rho}^{\mathrm{C}}_{\mathrm{LTE}}=\frac{1}{\mathcal{Z}}\exp\left[-\int d\Sigma_\mu\left(\widehat{T}^{\mu\nu}_{\mathrm{C}}\beta_\nu-\frac{1}{2}\Omega_{\lambda\nu}\widehat{S}^{\mu,\lambda\nu}_{\mathrm{C}}\right)\right].

which, as shown in Sec. II, can be rewritten by means of the pseudogauge transformation (14) with

Φ^Cλ,μν=−S^Cλ,μν=−i8Ψ‾{γλ,[γμ,γν]}Ψ\widehat{\Phi}_{C}^{\lambda,\mu\nu}=-\widehat{S}_{C}^{\lambda,\mu\nu}=-\frac{\mathrm{i}}{8}\overline{\Psi}\left\{\gamma^{\lambda},\left[\gamma^{\mu},\gamma^{\nu}\right]\right\}\Psi

and Z^C=0\widehat{Z}_{C}=0, as2

ρ^LTEC=1Zexp⁡[−∫dΣμ(T^Bμνβν+ϖλν−Ωλν2S^Cλ,μν)].(39)\widehat{\rho}_{\mathrm{LTE}}^{C}=\frac{1}{Z}\exp\left[-\int\mathrm{d}\Sigma_{\mu}\left(\widehat{T}^{\mu\nu}_{B}\beta_{\nu}+\frac{\varpi_{\lambda\nu}-\Omega_{\lambda\nu}}{2}\widehat{S}_{C}^{\lambda,\mu\nu}\right)\right]. \tag*{(39)}

The spin polarization resulting from this operator is

SCμ(k)≃Sϖμ(k)+Sξμ(k)+ΔΘCSμ(k).(40)S_{C}^{\mu}(k)\simeq S_{\varpi}^{\mu}(k)+S_{\xi}^{\mu}(k)+\Delta_{\Theta}^{C}S^{\mu}(k). \tag*{(40)}

The first two terms are given by Eqs. (3) and (4) and are the contributions coming from the Belinfante form, which appears in the first term in the exponent of Eq. (39). These two are the terms that have been considered so far to predict the spin polarization of particles emitted by the quark gluon plasma. When adopting the canonical decomposition, in addition to these terms we should also add the contribution from the canonical spin tensor Δ^ΘCSμ(k)\widehat{\Delta}_{\Theta}^{C}S^{\mu}(k). This one can be read from (36) and it is given by

ΔΘCSμ(k)=ϵλρστt^λ(kμkτ−η τμm2)8mεk∫ΣdΣ(x)⋅k nF(1−nF)(ϖρσ−Ωρσ)∫ΣdΣ⋅k nF.(41)\Delta_{\Theta}^{C}S^{\mu}(k)=\frac{\epsilon^{\lambda\rho\sigma\tau}\widehat{t}_{\lambda}(k^{\mu}k_{\tau}-\eta^{\mu}_{\ \tau}m^{2})}{8m\varepsilon_{k}}\frac{\int_{\Sigma}\mathrm{d}\Sigma(x)\cdot k\,n_{F}(1-n_{F})(\varpi_{\rho\sigma}-\Omega_{\rho\sigma})}{\int_{\Sigma}\mathrm{d}\Sigma\cdot k\,n_{F}}. \tag*{(41)}

To the best of my knowledge, this is the first time that an explicit formula for the contribution of (canonical) spin tensor is reported. Notice that in this case there is no additional contribution from the thermal shear ξ\xi, as the canonical spin tensor is completely anti-symmetric.

The contribution to spin polarization in Eq. (41) is the first important and necessary step to study the impact of pseudogauge transformations to the spin polarization in heavy-ion collisions. For instance, the result (41) reveals that the contribution of spin potential is not simply obtained by replacing the thermal vorticity with the spin potential in Eq. (3). An other important conclusion that one draws from Eq. (41) is that it is impossible to remove the difference between the spin polarization in the Belinfante and canonical PGs by choosing different values of the β(x)\beta(x) and Ω(x)\Omega(x) fields for the two PGs. Indeed, since the kk dependence makes the Eqs. (3), (4), and (41) linearly independent, the only way to satisfy the equivalence SB[βB](k)=SC[βC,ΩC](k)S_{B}[\beta^{B}](k)=S_{C}[\beta^{C},\Omega^{C}](k) for all the values of kk, is to choose βB=βC\beta^{B}=\beta^{C} and ΩC=ϖC\Omega^{C}=\varpi^{C}, that is the same β\beta field for both the PGs and the condition for global equilibrium.

Clearly, a quantitative estimate requires a numerical analysis. To assess the relevance of the contribution (41) one should know the magnitude of the difference between the thermal vorticity and the spin potential. While one can evaluate the thermal vorticity, for instance with hydrodynamic simulations, the magnitude of the spin potential Ω\Omega, which appears as a parameter to be fixed with observation, is not known. However, as discussed in Sec. II A, the quantity Θ=ϖ−Ω\Theta=\varpi-\Omega must reach zero as the system approaches global equilibrium and it is therefore expected to be significant only for systems far from equilibrium. Also note that the manifest breaking of Lorentz covariance in (41) by the presence of the unit vector t^\widehat{t} is expected as the canonical spin tensor is not a conserved quantity and the statistical operator (39) depends on the particular 3D hypersurface of integration; see also the discussion in [5].

de Groot-van Leeuwen-van Weert and Hilgevoord-Wouthuysen decompositions

Consider now the de Groot-van Leeuwen-van Weert (GLW) and Hilgevoord-Wouthuysen (HW) decompositions. Both these decompositions result in the same predictions for the spin polarization. From the results (36) the spin polarization reads:

SGLW,HWμ(k)≃Sϖμ(k)+Sξμ(k)+ΔΘCSμ(k)+ΔΘGLW,HWSμ(k)+ΔξGLW,HWSμ(k).S_{\mathrm{GLW,HW}}^{\mu}(k)\simeq S_{\varpi}^{\mu}(k)+S_{\xi}^{\mu}(k)+\Delta_{\Theta}^{C}S^{\mu}(k)+\Delta_{\Theta}^{\mathrm{GLW,HW}}S^{\mu}(k)+\Delta_{\xi}^{\mathrm{GLW,HW}}S^{\mu}(k).

The first three terms were discussed above, the other two are given by:

ΔΘGLW,HWSμ(k)=−14mϵμλρτt^λkτkσεk∫ΣdΣ⋅k nF(1−nF)(ϖρσ−Ωρσ)∫ΣdΣ⋅k nF.(42)\Delta_{\Theta}^{\mathrm{GLW,HW}}S^{\mu}(k)=-\frac{1}{4m}\epsilon^{\mu\lambda\rho\tau}\widehat{t}_{\lambda}\frac{k_{\tau}k^{\sigma}}{\varepsilon_{k}}\frac{\int_{\Sigma}\mathrm{d}\Sigma\cdot k\,n_{F}(1-n_{F})(\varpi_{\rho\sigma}-\Omega_{\rho\sigma})}{\int_{\Sigma}\mathrm{d}\Sigma\cdot k\,n_{F}}. \tag*{(42)}

and

ΔξGLW,HWSμ(k)=−Sξμ(k)=+14mϵμλστkτkρεk∫ΣdΣ⋅k nF(1−nF)t^λξρσ∫ΣdΣ⋅k nF.(43)\Delta_{\xi}^{\mathrm{GLW,HW}}S^{\mu}(k)=-S_{\xi}^{\mu}(k)=+\frac{1}{4m}\epsilon^{\mu\lambda\sigma\tau}\frac{k_{\tau}k^{\rho}}{\varepsilon_{k}}\frac{\int_{\Sigma}\mathrm{d}\Sigma\cdot k\,n_{F}(1-n_{F})\widehat{t}_{\lambda}\xi_{\rho\sigma}}{\int_{\Sigma}\mathrm{d}\Sigma\cdot k\,n_{F}}. \tag*{(43)}

The contribution in Eq. (42) has a different form compared to both Eq. (3) and Eq. (41), and its numerical investigation requires the knowledge of the spin tensor Ω\Omega. Instead, the contribution of (43) only requires the knowledge of the thermal shear ξ\xi and it must be included even if the spin potential is not introduced at all. Remarkably, it cancels exactly the contribution of Eq. (4). The remaining terms can be simplified using the Schouten identity, obtaining the more compact formula:

SGLW,HWμ(k)=−18mϵμρστkτ∫ΣdΣ⋅k nF(1−nF)Ωρσ∫ΣdΣ⋅k nF.S_{\mathrm{GLW,HW}}^{\mu}(k)=-\frac{1}{8m}\epsilon^{\mu\rho\sigma\tau}k_{\tau}\frac{\int_{\Sigma}\mathrm{d}\Sigma\cdot k\,n_{F}(1-n_{F})\Omega_{\rho\sigma}}{\int_{\Sigma}\mathrm{d}\Sigma\cdot k\,n_{F}}.

As found for the canonical PG, even allowing for different values of the thermodynamic fields, the only way to obtain the same predictions for the spin polarization in the HW/GLW PG as in the Belinfante or canonical PG is to have the same β\beta field in all PGs and to impose the global equilibrium conditions ξ=0\xi=0 and Ω=ϖ\Omega=\varpi.

The importance of including the thermal shear term in Eq. (4) in heavy-ion collisions has been discussed in [7, 8], where it was found that such a term is able to restore the agreement between the predictions and the experimental data for the momentum dependent spin polarization of Lambda hyperons. The above studies were (tacitly) carried out in the Belinfante decomposition. From the analysis above, if the GLW or HW decomposition were used instead, there would have been no thermal-shear corrections and hence no agreement with the data (with the predictions of local thermal equilibrium at first order in the gradients). This shows how the choice of the pseudogauge can indeed have sizable effects on observables.

Other pseudogauges

In special relativity, the choice of the pseudogauge is arbitrary. In this work, I considered only a few choices for the pseudogauges of the Dirac field, see Eqs. (17–21), and I evaluated the resulting predictions for the spin polarization (36). It was found that in the GLW and HW decompositions, the contribution of the thermal shear is increased compared to the Belinfante. If instead one chooses the Φ∂Σ\Phi_{\partial\Sigma} pseudogauge in Eq. (18), one would obtain

ΔξΦ∂ΣSμ(k)=−1mϵμλστkτkρεk∫ΣdΣ⋅k nF(1−nF)t^λξρσ∫ΣdΣ⋅k nF(44)\Delta_{\xi}^{\Phi\partial\Sigma}S^\mu(k)=-\frac{1}{m}\epsilon^{\mu\lambda\sigma\tau}\frac{k_\tau k^\rho}{\varepsilon_k}\frac{\int_\Sigma d\Sigma\cdot k\,n_F(1-n_F)\hat{t}_\lambda\xi_{\rho\sigma}}{\int_\Sigma d\Sigma\cdot k\,n_F} \tag*{(44)}

which is larger compared to the contribution (4). Adding the two terms (44) and (4) together will enhance the effects discussed in [7]. However, both the sign and the numerical factor of the pseudogauge Φ∂Σ\Phi_{\partial\Sigma} are arbitrary. Indeed, in special relativity, one could have used any linear combinations of the pseudogauges in (17–21). It follows that it should be easy to pick an ad-hoc pseudogauge which is able to explain a certain set of data. Therefore, such an approach should be avoided unless one is dealing with a large set of measurements. In the lack of experimental evidence, one can also look at theoretical arguments that favor some of the pseudogauges.

The canonical, Belinfante, GLW and HW pseudogauges do not constitute an arbitrary or a posteriori choice, as they are the result of the direct application of Noether theorem or a specific choice in the symmetries and properties of the EMT tensor or the spin tensor. When we have a strong case for using a specific pseudogauge, then we can use measurements made at local equilibrium, such as the spin polarization in heavy-ion collisions, to discern what pseudogauge better describes the system.

Pseudogauge dependence of mean axial current

Let us now turn to the mean axial current, which is obtained from the Wigner function with

jAμ(x)=⟨j^Aμ(x)⟩=∫d4k tr⁡[γμγ5W(x,k)].j_A^\mu(x)=\langle\widehat{j}_A^\mu(x)\rangle=\int d^4k\,\operatorname{tr}\left[\gamma^\mu\gamma^5W(x,k)\right].

As previously discussed, at global equilibrium one obtains the axial current in Eq. (23). Instead, the mean axial current evaluated using the local equilibrium statistical operator in the Belinfante pseudogauge (13) is

⟨j^Aμ⟩B=WAwμ+ϵμστρξρσt^τΞρ+O(∂β2),(45)\langle\widehat{j}_A^\mu\rangle_{\mathrm B}=W^A w^\mu+\epsilon^{\mu\sigma\tau\rho}\xi_{\rho\sigma}\widehat{t}_\tau\Xi^\rho+\mathcal{O}\left(\partial\beta^2\right), \tag*{(45)}

where WAW^A and ww are defined as in Eqs. (25) and (24), except that the thermal vorticity does not have to be constant anymore. The second term in (45) is the axial current induced by the thermal shear ξ\xi, and from the Wigner function evaluated in [5] one can show that

Ξρ=4∫d4k(2π)3εkθ(k0)δ(k2−m2)nF(1−nF)kτkρ.(46)\Xi^\rho=4\int\frac{d^4k}{(2\pi)^3\varepsilon_k}\theta(k^0)\delta(k^2-m^2)n_F(1-n_F)k_\tau k^\rho. \tag*{(46)}

However, the above formula must be modified in other pseudogauges. In a different pseudogauge one should add the difference:

ΔΦjAμ(x)=⟨j^Aμ(x)⟩Φ−⟨j^Aμ(x)⟩B≃ΔΘjAμ(x)+ΔξjAμ(x).\Delta_{\Phi}j_A^\mu(x)=\langle\widehat{j}_A^\mu(x)\rangle_{\Phi}-\langle\widehat{j}_A^\mu(x)\rangle_{\mathrm B}\simeq\Delta_{\Theta}j_A^\mu(x)+\Delta_{\xi}j_A^\mu(x).

Using the results in Eq. (33) for the particle part and similar expressions for the anti-particle, one obtains

ΔΘjAμ(x)=Θρσt^λ∫d4k2(2π)3εkθ(k0)δ(k2−m2) nF(1−nF)AΘ,Φμρσλ,\Delta_{\Theta}j_A^\mu(x)=\Theta_{\rho\sigma}\widehat{t}_\lambda\int\frac{d^4k}{2(2\pi)^3\varepsilon_k}\theta(k^0)\delta(k^2-m^2) \ n_F(1-n_F)A_{\Theta,\Phi}^{\mu\rho\sigma\lambda},
ΔξjAμ(x)=+ξρσt^λ∫d4k(2π)3εkθ(k0)δ(k2−m2) nF(1−nF)Aξ,Φμρσλ.(47)\Delta_{\xi}j_A^\mu(x)=+\xi_{\rho\sigma}\widehat{t}_\lambda\int\frac{d^4k}{(2\pi)^3\varepsilon_k}\theta(k^0)\delta(k^2-m^2) \ n_F(1-n_F)A_{\xi,\Phi}^{\mu\rho\sigma\lambda}. \tag*{(47)}

where the quantities AA are the same as Eqs. (37) and (38). These modifications are generally non-vanishing. In particular, the thermal field Θ\Theta is proportional to the thermal vorticity, so ΔΘjAμ\Delta_{\Theta}j_A^\mu contains a contribution along the rotation of the fluid, affecting the axial vortical effect conductivity.

As a notable example consider a massless field and the canonical decomposition where Φ^=Φ^C\widehat{\Phi}=\widehat{\Phi}_{\mathrm C}, that is

ΔΘCjAμ(x)=ϵλρστt^λΘρσ∫d4k4π3εkθ(k0)δ(k2)kμkτ×nF(β(x)⋅k)(1−nF(β(x)⋅k))=ϵλρστt^λΘρσ(13ημτ+23uμuτ)T3(x)12=uμϵλρστt^λΘρσuτT3(x)36−12ϵμρσλΘρσt^λT3(x)9,\begin{aligned} \Delta_{\Theta}^{\mathrm C}j_A^\mu(x)=\epsilon^{\lambda\rho\sigma\tau}\widehat{t}_\lambda\Theta_{\rho\sigma}\int\frac{d^4k}{4\pi^3\varepsilon_k}\theta(k^0)\delta(k^2)k^\mu k_\tau\\ &\times n_F\bigl(\beta(x)\cdot k\bigr)\left(1-n_F\bigl(\beta(x)\cdot k\bigr)\right) \\ &=\epsilon^{\lambda\rho\sigma\tau}\widehat{t}_\lambda\Theta_{\rho\sigma}\left(\frac{1}{3}\eta^{\mu\tau}+\frac{2}{3}u^\mu u^\tau\right)\frac{T^3(x)}{12} \\ &=u^\mu\epsilon^{\lambda\rho\sigma\tau}\widehat{t}_\lambda\Theta_{\rho\sigma}u_\tau\frac{T^3(x)}{36} -\frac{1}{2}\epsilon^{\mu\rho\sigma\lambda}\Theta_{\rho\sigma}\widehat{t}_\lambda\frac{T^3(x)}{9}, \end{aligned}
endsmid=1{ends_mid=1}

where I took advantage of the Lorentz invariance of the integral measure and I chose the unit vector u=β/β2u = \beta/\sqrt{\beta^2} as the time direction for the kk four-vector. Furthermore, if one decomposes t^\hat{t} in a orthogonal and parallel parts respect to uu as

t^λ=(t^⋅u)uλ+t^⊥λ\hat{t}_{\lambda} = (\hat{t} \cdot u)u_{\lambda} + \hat{t}_{\perp\lambda}

and uses Θρσ=ϖρσ−Ωρσ\Theta_{\rho\sigma} = \varpi_{\rho\sigma} - \Omega_{\rho\sigma} and the definition (24), one obtains

ΔΘCjAμ(x)=uμϵλρστt^λΘρσuτT3(x)36+wμ(t^⋅u)T3(x)9−12ϵμρσλΘρσt^⊥λT3(x)9+12ϵμρσλΩρσt^λT3(x)9.\Delta_{\Theta}^{\mathrm{C}} j_{\mathrm{A}}^{\mu}(x) = u^{\mu}\epsilon^{\lambda\rho\sigma\tau}\hat{t}_{\lambda}\Theta_{\rho\sigma}u_{\tau}\frac{T^{3}(x)}{36} + w^{\mu}\frac{(\hat{t}\cdot u)T^{3}(x)}{9} - \frac{1}{2}\epsilon^{\mu\rho\sigma\lambda}\Theta_{\rho\sigma}\hat{t}_{\perp\lambda}\frac{T^{3}(x)}{9} + \frac{1}{2}\epsilon^{\mu\rho\sigma\lambda}\Omega_{\rho\sigma}\hat{t}_{\lambda}\frac{T^{3}(x)}{9}.

The second term describes a mean axial current flowing along the rotation of the fluid and therefore gives a modification of the AVE conductivity of

ΔΘCWA=(t^⋅u)T3(x)9.\Delta_{\Theta}^{\mathrm{C}}\mathcal{W}^{\mathrm{A}} = \frac{(\hat{t}\cdot u)T^{3}(x)}{9}.

In general, the modification of the AVE conductivity for the other pseudogauges in both the massive and massless cases can be obtained from the general expressions (47) by projecting along ww:

ΔΦWA=wμw2(ΔΘjAμ(x)+ΔξjAμ(x)).\Delta_{\Phi}\mathcal{W}^{\mathrm{A}} = \frac{w_{\mu}}{w^{2}}\left(\Delta_{\Theta}j_{\mathrm{A}}^{\mu}(x) + \Delta_{\xi}j_{\mathrm{A}}^{\mu}(x)\right).

Discussion

In summary, I showed that the pseudogauge (PG) dependent part of the statistical operator describing a system at local thermal equilibrium (LTE) results in a non-vanishing contribution to the Wigner function of the free Dirac field. I then evaluated the spin polarization at LTE with different choices of pseudogauge potentials and I found that they generally give different results. In the canonical case, the contribution from the canonical spin tensor is different form what one might expect by just replacing the thermal vorticity with the spin potential in Eq. (2), see Eq. (41). The de Groot-van Leeuwen-van Weert (GLW) and the Hilgevoord-Wouthuysen (HW) decompositions result in the same spin polarization. This spin polarization differs from the ones resulting from the Belinfante and from the canonical decomposition. In particular, there is no contribution from thermal-shear in the GLW and HW PG, see Eq. (43). It is important to stress that this difference does not depend on the spin potential. Therefore, the choice of the pseudogauge significantly affects the spin polarization predictions used in heavy-ion collisions. I also showed that the PG dependence in the spin polarization can not be removed by choosing different values for the four-temperature β(x)\beta(x) and for the spin potential Ω(x)\Omega(x) in the different PGs. On the contrary, as suggested by the analysis of the statistical operator itself, I found that the different PGs give the same spin polarization only if the same β\beta field is chosen and if the conditions of global thermal equilibrium are satisfied.

In principle, the matching of different predictions with experimental measurements should reveal what pseudogauge must be used and if this choice is universal for all systems. However, given the arbitrariness of pseudogauge transformations, this method alone does not seem compelling. I advocate instead for the search of a theoretical argument in favor of a particular pseudogauge, which one can thereafter put to test. When these predictions are compared with experiments, one should also keep in mind that they are the result of the approximation of the “true” statistical operator (11) with the local equilibrium one (12) which only account for non-dissipative phenomena. The dissipative effects for the spin polarization are yet unknown. Future studies on that topic are therefore recommended.

More stringent requirements on the pseudogauge transformations might come from the theory of general relativity, where the form of the energy-momentum tensor is strictly related to the geometry. The Einstein-Cartan theory allows the inclusion of a spin tensor in general relativity and requires the use of Riemann-Cartan geometry [24], which has non-vanishing torsion. Then, the comparison of results from quantum hydrodynamics, gravitation, and proton spin decomposition [54] might shed light on this long standing problem.

Furthermore, I showed that the pseudogauge transformations also affect the axial vortical effect (AVE) conductivity for a system out of global equilibrium. This might pose a problem to the interpretation of the AVE as a consequence of the gravitational anomaly [55], as one might expect the gravitational anomaly to be universal and not pseudogauge dependent. Further investigation in this direction might also clarify the role of spin in gravity and hydrodynamics.

Acknowledgments

Very useful discussions with F. Becattini and K. Tuchin are gratefully acknowledged. I am also immensely grateful to N. Weickgenannt for her comments, that helped me identify some mistakes in the calculations. M.B. is supported by the U.S. Department of Energy under Grant No. DE-FG02-87ER40371.

Pseudogauge operators

In this appendix I report the different form of pseudogauge fields used to perform the calculations. For the Dirac field, the pseudogauge operators in (17 - 20) can be obtained from the Wigner operator

W^AB(x,k)=∫d4y(2π)4e−ik⋅yΨ‾B(x+y2)ΨA(x−y2).\widehat{W}_{AB}(x,k) = \int\frac{\mathrm{d}^{4}y}{(2\pi)^{4}}e^{-ik\cdot y}\overline{\Psi}_{B}\left(x+\frac{y}{2}\right)\Psi_{A}\left(x-\frac{y}{2}\right).

by performing the integration

Φ^λ,μν(x)=∫d4k′ Φ^B′A′λ,μν(k′)W^A′B′(x,k′).\widehat{\Phi}^{\lambda,\mu\nu}(x)=\int d^{4}k'\,\widehat{\Phi}^{\lambda,\mu\nu}_{B'A'}(k')\widehat{W}^{A'B'}(x,k').

For the gaugepotentials in Eq. (17) the function Φ(k)\Phi(k) is given by

ΦCλ,ρσ=ΦϵAλ,ρσ=12ϵλρστγτγ5,\Phi_{\mathrm{C}}^{\lambda,\rho\sigma}=\Phi_{\epsilon A}^{\lambda,\rho\sigma}=\frac{1}{2}\epsilon^{\lambda\rho\sigma\tau}\gamma_{\tau}\gamma^{5},
ΦϵVλ,ρσ=12ϵλρστγτ,\Phi_{\epsilon\mathrm{V}}^{\lambda,\rho\sigma}=\frac{1}{2}\epsilon^{\lambda\rho\sigma\tau}\gamma_{\tau},
ΦAλ,ρσ=(ηλργσ−ηλσγρ)γ5,\Phi_{\mathrm{A}}^{\lambda,\rho\sigma}=(\eta^{\lambda\rho}\gamma^{\sigma}-\eta^{\lambda\sigma}\gamma^{\rho})\gamma^{5},
ΦVλ,ρσ=ηλργσ−ηλσγρ.\Phi_{\mathrm{V}}^{\lambda,\rho\sigma}=\eta^{\lambda\rho}\gamma^{\sigma}-\eta^{\lambda\sigma}\gamma^{\rho}.

for those in Eq. (18) it is

Φϵ∂λ,ρσ=2ϵλτρσkτ,\Phi_{\epsilon\partial}^{\lambda,\rho\sigma}=2\epsilon^{\lambda\tau\rho\sigma}k_{\tau},
Φ∂λ,ρσ=2(ηλρkσ−ηλσkρ),\Phi_{\partial}^{\lambda,\rho\sigma}=2(\eta^{\lambda\rho}k^{\sigma}-\eta^{\lambda\sigma}k^{\rho}),
Φϵ∂Aλ,ρσ=2(ηλρkσ−ηλσkρ)γ5,\Phi_{\epsilon\partial A}^{\lambda,\rho\sigma}=2(\eta^{\lambda\rho}k^{\sigma}-\eta^{\lambda\sigma}k^{\rho})\gamma^{5},
Φ∂Aλ,ρσ=ϵλτρσkτγ5,\Phi_{\partial A}^{\lambda,\rho\sigma}=\epsilon^{\lambda\tau\rho\sigma}k_{\tau}\gamma^{5},
Φ∂Σλ,ρσ=2mkλσρσ.\Phi_{\partial\Sigma}^{\lambda,\rho\sigma}=\frac{2}{m}k^{\lambda}\sigma^{\rho\sigma}.

for the de Groot-van Leeuwen-van Weert (GLW) decomposition (19) it is

ΦGLWλ,ρσ=ΦϵAλ,ρσ+12m(σλρkσ−σλσkρ)\Phi_{\mathrm{GLW}}^{\lambda,\rho\sigma}=\Phi_{\epsilon A}^{\lambda,\rho\sigma}+\frac{1}{2m}(\sigma^{\lambda\rho}k^{\sigma}-\sigma^{\lambda\sigma}k^{\rho})

and for the Hilgevoord-Wouthuysen decomposition of Eq. (20) it is

ΦHWλ,ρσ=ΦGLWλ,ρσ−12m(ηλρσσα−ηλσσρα)kα.\Phi_{\mathrm{HW}}^{\lambda,\rho\sigma}=\Phi_{\mathrm{GLW}}^{\lambda,\rho\sigma}-\frac{1}{2m}(\eta^{\lambda\rho}\sigma^{\sigma\alpha}-\eta^{\lambda\sigma}\sigma^{\rho\alpha})k_{\alpha}.

The improved Callan, Coleman, and Jackiew (CCJ) pseudogauge potentials (21) are reported in Eq. (28).

In order to obtain the difference between the Wigner function in a generic pseudogauge and the Wigner function in the Belinfante pseudogauge, one just has to replace the right expression above in Eq. (33). To obtain the corrections to spin polarization (36) and to the mean axial current (47), the expression above must be replaced inside the traces,

AΘ,Φμρσλ=tr⁡4[γμγ5(k̸+m)Φλ,ρσ(k̸+m)],A_{\Theta,\Phi}^{\mu\rho\sigma\lambda}=\operatorname{tr}_{4}\left[\gamma^{\mu}\gamma^{5}(\not{k}+m)\Phi^{\lambda,\rho\sigma}(\not{k}+m)\right],
Aξ,Φμρσλ=tr⁡4[γμγ5(k̸+m)Φρ,λσ(k̸+m)],A_{\xi,\Phi}^{\mu\rho\sigma\lambda}=\operatorname{tr}_{4}\left[\gamma^{\mu}\gamma^{5}(\not{k}+m)\Phi^{\rho,\lambda\sigma}(\not{k}+m)\right],

whose results are given in Eqs. (37) and (38).

References

  1. [1]L. Adamczyk et al. (STAR), Nature 548, 62 (2017), arXiv:1701.06657 [nucl-ex].
  2. [2]J. Adam et al. (STAR), Phys. Rev. C 98, 014910 (2018), arXiv:1805.04400 [nucl-ex].DOI
  3. [3]J. Adam et al. (STAR), Phys. Rev. Lett. 123, 132301 (2019), arXiv:1905.11917 [nucl-ex].DOI
  4. [4]F. Becattini and M. A. Lisa, Ann. Rev. Nucl. Part. Sci. 70, 395 (2020), arXiv:2003.03640 [nucl-ex].DOI
  5. [5]F. Becattini, M. Buzzegoli, and A. Palermo, Phys. Lett. B 820, 136519 (2021), arXiv:2103.10917 [nucl-th].DOI
  6. [6]S. Y. F. Liu and Y. Yin, JHEP 07, 188 (2021), arXiv:2103.09200 [hep-ph].DOI
  7. [7]F. Becattini, M. Buzzegoli, G. Inghirami, I. Karpenko, and A. Palermo, Phys. Rev. Lett. 127, 272302 (2021), arXiv:2103.14621 [nucl-th].DOI
  8. [8]B. Fu, S. Y. F. Liu, L. Pang, H. Song, and Y. Yin, Phys. Rev. Lett. 127, 142301 (2021), arXiv:2103.10403 [hep-ph].DOI
  9. [9]W. Florkowski, B. Friman, A. Jaiswal, and E. Speranza, Phys. Rev. C 97, 041901 (2018), arXiv:1705.00587 [nucl-th].DOI
  10. [10]W. Florkowski, B. Friman, A. Jaiswal, R. Ryblewski, and E. Speranza, Phys. Rev. D 97, 116017 (2018), arXiv:1712.07676 [nucl-th].DOI
  11. [11]W. Florkowski, A. Kumar, and R. Ryblewski, Phys. Rev. C 98, 044906 (2018), arXiv:1806.02616 [hep-ph].DOI
  12. [12]K. Hattori, M. Hongo, X.-G. Huang, M. Matsuo, and H. Taya, Phys. Lett. B 795, 100 (2019), arXiv:1901.06615 [hep-th].DOI
  13. [13]S. Bhadury, W. Florkowski, A. Jaiswal, A. Kumar, and R. Ryblewski, Phys. Lett. B 814, 136096 (2021), arXiv:2002.03937 [hep-ph].DOI
  14. [14]N. Weickgenannt, E. Speranza, X.-I. Sheng, Q. Wang, and D. H. Rischke, Phys. Rev. Lett. 127, 052301 (2021), arXiv:2005.01506 [hep-ph].DOI
  15. [15]L. Tinti and W. Florkowski, Lect. Notes Phys. 987, 137 (2021), arXiv:2007.04029 [nucl-th].DOI
  16. [16]K. Fukushima and S. Pu, Phys. Lett. B 817, 136346 (2021), arXiv:2010.01608 [hep-th].DOI
  17. [17]M. Hongo, X.-G. Huang, M. Kaminski, M. Stephanov, and H.-U. Yee, JHEP 11, 150 (2021), arXiv:2107.14231 [hep-th].DOI
  18. [18]S. Bhadury, J. Bhatt, A. Jaiswal, and A. Kumar, Eur. Phys. J. ST 230, 655 (2021), arXiv:2101.11964 [hep-ph].DOI
  19. [19]W. Florkowski, A. Kumar, and R. Ryblewski, Prog. Part. Nucl. Phys. 108, 103709 (2019), arXiv:1811.04409 [nucl-th].DOI
  20. [20]H.-Z. Wu, L.-G. Pang, X.-G. Huang, and Q. Wang, Phys. Rev. Research 1, 033058 (2019), arXiv:1906.09385 [nucl-th].DOI
  21. [21]W. Florkowski, A. Kumar, and R. Ryblewski, Acta Phys. Polon. B 51, 945 (2020), arXiv:1907.09835 [nucl-th].DOI
  22. [22]A. D. Gallegos and U. Gürsoy, JHEP 11, 151 (2020), arXiv:2004.05148 [hep-th].DOI
  23. [23]A. D. Gallegos, U. Gürsoy, and A. Yarom, SciPost Phys. 11, 041 (2021), arXiv:2101.04759 [hep-th].DOI
  24. [24]F. W. Hehl, Rept. Math. Phys. 9, 55 (1976).
  25. [25]F. Beccattini and L. Tinti, Phys. Rev. D 84, 025013 (2011), arXiv:1101.5251 [hep-th].
  26. [26]F. Beccattini and L. Tinti, Phys. Rev. D 87, 025029 (2013), arXiv:1209.6212 [hep-th].
  27. [27]F. Beccattini, W. Florkowski, and E. Speranza, Phys. Lett. B 789, 419 (2019), arXiv:1807.10994 [hep-th].
  28. [28]E. Speranza and N. Weickgenannt, Eur. Phys. J. A 57, 155 (2021), arXiv:2007.00138 [nucl-th].DOI
  29. [29]S. Li, M. A. Stephanov, and H.-U. Yee, Phys. Rev. Lett. 127, 082302 (2021), arXiv:2011.12318 [hep-th].DOI
  30. [30]F. Beccattini, V. Chandra, L. Del Zanna, and E. Grossi, Annals Phys. 338, 32 (2013), arXiv:1303.3431 [nucl-th].
  31. [31]F. Beccattini, Lect. Notes Phys. 987, 15 (2021), arXiv:2004.04050 [hep-th].
  32. [32]A. Vilenkin, Phys. Rev. D 20, 1807 (1979).
  33. [33]K. Landsteiner, E. Megias, L. Melgar, and F. Pena-Benitez, JHEP 09, 121 (2011), arXiv:1107.0368 [hep-th].DOI
  34. [34]J.-H. Gao, Z.-T. Liang, S. Pu, Q. Wang, and X.-N. Wang, Phys. Rev. Lett. 109, 232301 (2012), arXiv:1203.0725 [hep-ph].DOI
  35. [35]D. E. Kharzeev, J. Liao, S. A. Voloshin, and G. Wang, Prog. Part. Nucl. Phys. 88, 1 (2016), arXiv:1511.04050 [hep-ph].DOI
  36. [36]D. N. Zubarev, A. V. Prozorkevich, and S. A. Smolyanskii, Theoretical and Mathematical Physics 40 (1979), 10.1007/BF01015755.
  37. [37]C. van Weert, Annals of Physics 140, 133 (1982).
  38. [38]F. Beccattini, M. Buzzegoli, and E. Grossi, Particles 2, 197 (2019), arXiv:1902.01089 [cond-mat.stat-mech].
  39. [39]M. Buzzegoli, Lect. Notes Phys. 987, 53 (2021), arXiv:2011.09974 [hep-ph].DOI
  40. [40]F. Beccattini, L. Bucciantini, E. Grossi, and L. Tinti, Eur. Phys. J. C 75, 191 (2015), arXiv:1403.6265 [hep-th].
  41. [41]F. Beccattini and E. Grossi, Phys. Rev. D 92, 045037 (2015), arXiv:1505.07760 [gr-qc].
  42. [42]S. D. Groot, W. van Leeuwen, and C. van Weert, Relativistic Kinetic Theory (North Holland, Amsterdam, 1980).
  43. [43]F. W. Hehl, A. Macias, E. W. Mielke, and Y. N. Obukhov, in Symposia to honor Engelbert Schucking (1997) arXiv:gr-qc/9706009.
  44. [44]I. Kirsch, L. H. Ryder, and F. W. Hehl, (2001), arXiv:hep-th/0102102.
  45. [45]C. G. J. Callan, S. Coleman, and R. Jackiw, Annals of Physics 59, 42 (1970).
  46. [46]F. Beccattini, Phys. Rev. Lett. 108, 244502 (2012), arXiv:1201.5278 [gr-qc].
  47. [47]A. Palermo, M. Buzzegoli, and F. Beccattini, JHEP 10, 077 (2021), arXiv:2106.08340 [hep-th].DOI
  48. [48]M. Buzzegoli, E. Grossi, and F. Beccattini, JHEP 10, 091 (2017), [Erratum: JHEP 07, 119 (2018)], arXiv:1704.02808 [hep-th].DOI
  49. [49]G. Y. Prokhorov, O. V. Teryaev, and V. I. Zakharov, JHEP 02, 146 (2019), arXiv:1807.03584 [hep-th].DOI
  50. [50]M. Buzzegoli and D. E. Kharzeev, Phys. Rev. D 103, 116005 (2021), arXiv:2102.01676 [hep-th].DOI
  51. [51]M. Buzzegoli and F. Beccattini, JHEP 12, 002 (2018), arXiv:1807.08701 [hep-th].
  52. [52]A. Flachi and K. Fukushima, Phys. Rev. D 98, 096011 (2018), arXiv:1702.04753 [hep-th].DOI
  53. [53]F. Beccattini, M. Buzzegoli, A. Palermo, and G. Prokhorov, Phys. Lett. B 822, 136706 (2021), arXiv:2009.13449 [hep-ph].
  54. [54]X. Ji, F. Yuan, and Y. Zhao, Nature Rev. Phys. 3, 27 (2021), arXiv:2009.01291.arxiv.org/abs/2009.01291
  55. [55]K. Landsteiner, E. Megias, and F. Pena-Benitez, Phys. Rev. Lett. 107, 021601 (2011), arXiv:1103.5006 [hep-ph].DOI

Paper details

vlnc:W_wq45hh5x
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents