betaFeedback

WorksPhysics and Astronomy

Novel hydrodynamic features in dissipative spin-polarized fluids

Abstract

We investigate an extension of perfect spin hydrodynamics that includes dissipation in the spin sector. The presence of dissipation implies that antisymmetric parts of the energy-momentum tensor should be taken into account. Their form follows from the Navier-Stokes or Israel-Stewart theory, which introduces a priori unknown kinetic coefficients and relaxation times. We find that the inclusion of the antisymmetric parts of the energy-momentum tensor has twofold effects: i) they form source terms for the spin tensor allowing for the transfer between the spin and orbital parts of the total angular momentum (spin-orbit interaction) and ii) they appear in the conservation laws for the energy and linear momentum, indirectly affecting the spin dynamics. Numerical results are obtained for two boost-invariant expansion geometries: one-dimensional longitudinal motion (Bjorken model) and two-dimensional boost-invariant and cylindrically symmetric expansion. They indicate (for various choices of the spin kinetic coefficients and the relaxation times) that the spin polarization tensor approaches thermal vorticity. We also observe that Israel-Stewart dynamics on the way to local equilibrium approaches first the Navier-Stokes regime. In general, we find that the Israel-Stewart formalism naturally helps to consistently combine the conservation laws for linear and angular momentum, while the Navier-Stokes framework itself can lead to a disruption in the structure of the equations through the inclusion of terms involving derivatives of an order higher than one.

Introduction

Phenomena connected with spin polarization and alignment attract a lot of attention currently [1, 2, 3, 4, 5, 6]; for a recent experimental review, see [7]. From the theoretical point of view, the spin polarization effects initiated interest in spin hydrodynamics [8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43]. This framework includes spin degrees of freedom into the well-established formalism of relativistic hydrodynamics – the basic tool to interpret the results of heavy-ion experiments [44, 45, 46, 47, 48].

In this work we study an extension of perfect spin hydrodynamics that includes dissipation in the spin sector. The framework of perfect spin hydrodynamics follows our recent formulation summarized in [49] and [50]. The inclusion of dissipation is achieved by adding antisymmetric parts to the energy–momentum tensor [14, 51]. This is done within the Navier–Stokes (NS) [14, 51] or Israel–Stewart (IS) [52] approach. In both cases one has to introduce kinetic coefficients whose forms are generally unknown. We assume that they scale with the entropy density and consider different proportionality constants. Similarly, using the IS theory, we consider in addition different values of the spin relaxation times.

The antisymmetric parts of the energy–momentum tensor have twofold effects. First, they represent source terms for the spin tensor, which allows for the transfer between the spin and orbital parts of the total angular momentum (dissipative spin-orbit interaction). Second, they appear in the conservation laws for the energy and linear momentum, indirectly affecting spin dynamics. In our opinion, such a feedback has not yet been sufficiently analyzed. We find that it is highly restrictive for spin evolution.

Our numerical results are obtained for two boost-invariant geometries: a one-dimensional longitudinal motion and a two-dimensional boost-invariant and cylindrically symmetric expansion. To distinguish between these two cases, we use the acronyms TH (transversely homogeneous system) and CS (cylindrically symmetric system). The antisymmetric part of the energy–momentum tensor gives a contribution to the entropy production that involves a difference between the spin polarization tensor ω\omega and the thermal vorticity tensor ϖ\varpi. As the system approaches local equilibrium, the entropy production rate decreases, and we expect that the spin polarization tensor approaches thermal vorticity. This behavior is confirmed by our numerical results.

In order to concentrate on spin effects mainly, we ignore dissipation connected with heat conduction, shear viscosity, and bulk viscosity. Nevertheless, in the spin sector we find many similarities to the standard dissipative dynamics. For example, on the way to local equilibrium the IS dynamics approaches first the NS regime. On the other hand, it turns out that the IS formalism is particularly convenient for describing dissipative spin dynamics, as it allows for the direct determination of the time derivative of the antisymmetric components of the energy–momentum tensor, which appear in the conservation laws for energy and linear momentum. However, the use of the NS approach alone may lead to an undesired modification of hydrodynamic equations via introduction of terms involving derivatives of an order higher than one.

Our present work can be interpreted as a continuation of a series of previous studies [16, 24, 53, 54, 55, 56, 57, 58, 59, 60] that offered numerical and analytic solutions to the equations of spin hydrodynamics (such studies analyzed, for example, one dimensional boost-invariant expansion [16, 55, 56, 57, 59, 60], two-dimensional Gubser expansion [24, 54], one-dimensional non boost-invariant motion [53], boost-invariant and cylindrically symmetric geometry [58]). Our results shed light on the future more realistic applications of dissipative spin hydrodynamics. In particular, they emphasize the advantages of using the IS theory.

Our paper is organized as follows: In Sec. II we introduce expressions for the energy–momentum and spin tensors that include dissipative terms. In § 1 the entropy production is discussed in the NS and IS approaches. Equations defining the dissipative terms of the energy–momentum tensor are given for these two cases. Boost-invariant and transversely homogeneous systems are discussed in Sec. IV, while boost-invariant and cylindrically symmetric systems are investigated in § 1.3. We present a summary in Sec. VI, and a collection of useful differential identities is given in Appendix A.

Notation and conventions: For the Levi-Civita tensor ϵμναβ\epsilon^{\mu\nu\alpha\beta} we follow the convention ϵ0123=−ϵ0123=+1\epsilon^{0123}=-\epsilon_{0123}=+1. The metric tensor is of the “mostly minuses” form gμν=diag⁡(+1,−1,−1,−1)g_{\mu\nu}=\operatorname{diag}(+1,-1,-1,-1). We use boldface symbols to denote three-vectors, and their scalar products are written as a⋅b\mathbf{a}\cdot\mathbf{b}. The particles are always considered on the mass shell with the energy Ep=m2+p2E_p=\sqrt{m^2+\mathbf{p}^2}. Throughout the text, we make use of natural units, ℏ=c=kB=1\hbar=c=k_{\mathrm{B}}=1.

We also use the standard notation where round (squared) brackets denote symmetrization (antisymmetrization) of the Lorentz indices: A(μν)=12(Aμν+Aνμ)A^{(\mu\nu)}=\frac{1}{2}(A^{\mu\nu}+A^{\nu\mu}) and A[μν]=12(Aμν−Aνμ)A^{[\mu\nu]}=\frac{1}{2}(A^{\mu\nu}-A^{\nu\mu}). Moreover, we introduce an orthogonal projection Tαβ⟨γ⟩δ…=ΔργTαβρδ…T^{\alpha\beta\langle\gamma\rangle\delta\ldots}=\Delta^\gamma_{\rho}T^{\alpha\beta\rho\delta\ldots}, where the operator Δργ=gργ−UγUρ\Delta^\gamma_{\rho}=g^\gamma_{\rho}-U^\gamma U_{\rho} projects on the plane transverse to the flow vector UU. If the symbols ⟨⟩\langle\rangle are used to the right of the differential symbol, the contractions with Δ\Delta should be done after the derivative is calculated first. The transverse gradient ∇μ\nabla^\mu is defined by the contraction Δμν∂ν\Delta^{\mu\nu}\partial_\nu.

Energy–momentum and spin tensors

Asymmetric energy–momentum tensor

In this work, we consider a system whose energy–momentum tensor has a symmetric and an antisymmetric part, Tμν=Teqμν+δTaμνT^{\mu\nu}=T_{\mathrm{eq}}^{\mu\nu}+\delta T_{\mathrm{a}}^{\mu\nu}. The symmetric part, TeqμνT_{\mathrm{eq}}^{\mu\nu}, describes local equilibrium, while the antisymmetric part, δTaμν\delta T_{\mathrm{a}}^{\mu\nu}, is dissipative but includes only the spin dissipation effects. Since we intend to isolate and study the consequences of having an antisymmetric part of the energy–momentum tensor, we neglect standard viscous corrections to TeqμνT_{\mathrm{eq}}^{\mu\nu} that are symmetric.

The tensor TeqμνT_{\mathrm{eq}}^{\mu\nu} has the well known perfect-fluid form

Teqμν=(ε+P)UμUν−Pgμν,(1)T_{\mathrm{eq}}^{\mu\nu}=(\varepsilon+P)U^\mu U^\nu-Pg^{\mu\nu}, \tag*{(1)}

where ε\varepsilon is the energy density, PP is the equilibrium pressure, and UμU^\mu is the hydrodynamic flow vector (with the normalization UμUμ=1U^\mu U_\mu=1). The energy density and pressure are connected through the equation of state. They can be treated as functions of temperature TT and baryon chemical potential μ\mu. In numerical calculations, we use the expressions describing a relativistic gas of massive particles (for their explicit forms see, for example, our previous work [58]). Following earlier studies, we neglect dependence of TeqμνT_{\mathrm{eq}}^{\mu\nu} on the components of the spin polarization tensor ωμν\omega^{\mu\nu}, as such dependence appears only as the quadratic and higher order corrections in the components of ωμν\omega^{\mu\nu}.

The antisymmetric part of TμνT^{\mu\nu} can be in general written in the form [14, 51]

δTaμν=dμUν−dνUμ+eμν.(2)\delta T_{\mathrm{a}}^{\mu\nu}=d^\mu U^\nu-d^\nu U^\mu+e^{\mu\nu}. \tag*{(2)}

The vector dμd^\mu is orthogonal to the flow vector, d⋅U=0d\cdot U=0, while the tensor eμνe^{\mu\nu} is antisymmetric and orthogonal to UU in two indices, eμνUμ=eμνUν=0e^{\mu\nu}U_\mu=e^{\mu\nu}U_\nu=0. There are three independent components in each of dμd^\mu and eμνe^{\mu\nu} – together, they represent the six independent components of δTaμν\delta T_{\mathrm{a}}^{\mu\nu}. The properties of eμνe^{\mu\nu} allow us to represent it as the contraction of the Levi-Civita tensor with the hydrodynamic flow vector UU and a vector ee that is orthogonal to UU, namely

eμν=εμναβUαeβ,e⋅U=0.(3)e^{\mu\nu}=\varepsilon^{\mu\nu\alpha\beta}U_\alpha e_\beta,\qquad e\cdot U=0. \tag*{(3)}

In analogy to magnetohydrodynamics, we may think of the vectors dd and ee as the electriclike and magneticlike contributions to δTaμν\delta T_{\mathrm{a}}^{\mu\nu}.

The energy and linear momentum conservation is expressed by the equation

∂μTμν=∂μTeqμν+∂μδTaμν=0.(4)\partial_\mu T^{\mu\nu}=\partial_\mu T_{\mathrm{eq}}^{\mu\nu}+\partial_\mu\delta T_{\mathrm{a}}^{\mu\nu}=0. \tag*{(4)}

Below, we consider two expansion geometries connected with two different sets of basis vectors X(i)ν\mathcal{X}_{(i)}^\nu (i=0,1,2,3i=0,1,2,3). By projecting Eq. (4) along X(i)ν\mathcal{X}_{(i)}^\nu, we obtain four equations

X(i)ν∂μTμν=0.(5)\mathcal{X}_{(i)}^\nu\partial_\mu T^{\mu\nu}=0. \tag*{(5)}

As the first basis vector, we always choose the flow vector, X(0)ν=Uν\mathcal{X}_{(0)}^\nu=U^\nu.

Spin tensor

Similarly to the energy–momentum tensor, we decompose the spin tensor into the equilibrium part and a correction, Sλ,μν=Seqλ,μν+δSλ,μνS^{\lambda,\mu\nu}=S_{\mathrm{eq}}^{\lambda,\mu\nu}+\delta S^{\lambda,\mu\nu}. The equilibrium part has the form [11]

Seqλ,μν=Uλ[A(kμUν−kνUμ)+A1tμν]+A2(tλμUν−tλνUμ+Δλμkν−Δλνkμ).(6)S_{\mathrm{eq}}^{\lambda,\mu\nu}=U^\lambda\left[A\left(k^\mu U^\nu-k^\nu U^\mu\right)+A_1t^{\mu\nu}\right]+\frac{A}{2}\left(t^{\lambda\mu}U^\nu-t^{\lambda\nu}U^\mu+\Delta^{\lambda\mu}k^\nu-\Delta^{\lambda\nu}k^\mu\right). \tag*{(6)}

where AA and A1A_1 are functions of TT and μ\mu (explicitly defined, for example, in [58]), whereas the vector kk and the tensor tλμt^{\lambda\mu} define the spin polarization tensor

ωμν=kμUν−kνUμ+tμν=kμUν−kνUμ+εμναβUαωβ.(7)\omega_{\mu\nu}=k_\mu U_\nu-k_\nu U_\mu+t_{\mu\nu}=k_\mu U_\nu-k_\nu U_\mu+\varepsilon_{\mu\nu\alpha\beta}U^\alpha\omega^\beta. \tag*{(7)}

The vectors kk and ω\omega satisfy the orthogonality conditions, k⋅U=0k \cdot U = 0 and ω⋅U=0\omega\cdot U = 0.

The correction to the equilibrium part of the spin tensor has the form [51]

δSλ,μν=δiλμUν−δiλνUμ+δjλμν,(8)\delta S^{\lambda,\mu\nu} = \delta i^{\lambda\mu}U^\nu- \delta i^{\lambda\nu}U^\mu+ \delta j^{\lambda\mu\nu}, \tag*{(8)}

where the tensors δiλμ\delta i^{\lambda\mu} and δjλμν\delta j^{\lambda\mu\nu} satisfy the following relations: δiλμUλ=δiλμUμ=0\delta i^{\lambda\mu}U_\lambda= \delta i^{\lambda\mu}U_\mu= 0, δjλμν=−δjλνμ\delta j^{\lambda\mu\nu} = -\delta j^{\lambda\nu\mu}, δjλμνUλ=δjλμνUμ=δjλμνUν=0\delta j^{\lambda\mu\nu}U_\lambda= \delta j^{\lambda\mu\nu}U_\mu= \delta j^{\lambda\mu\nu}U_\nu= 0 (consequently, there are nine independent components in δiλμ\delta i^{\lambda\mu} and also nine independent components in δjλμν\delta j^{\lambda\mu\nu}).

Since the energy–momentum tensor has an antisymmetric part, it acts as the source term for the spin tensor, namely, the conservation of the total angular momentum can be written as

∂λSλ,μν=∂λSeqλ,μν+∂λδSλ,μν=δTaνμ−δTaμν.(9)\partial_\lambda S^{\lambda,\mu\nu} = \partial_\lambda S_{\mathrm{eq}}^{\lambda,\mu\nu} + \partial_\lambda\delta S^{\lambda,\mu\nu} = \delta T_{\mathrm{a}}^{\nu\mu} - \delta T_{\mathrm{a}}^{\mu\nu}. \tag*{(9)}

Entropy production

In the Navier–Stokes limit, the entropy production is given by the expression [14, 51]

∂μsNSμ=−δNμ∂μξ+δTsμλ∂μβλ+δTaμλ(∂μβλ−ωλμ)−12δSμ,αβ∂μωαβ,(10)\partial_\mu s_{\mathrm{NS}}^\mu= -\delta N^\mu\partial_\mu\xi+ \delta T_{\mathrm{s}}^{\mu\lambda}\partial_\mu\beta_\lambda+ \delta T_{\mathrm{a}}^{\mu\lambda}\left(\partial_\mu\beta_\lambda- \omega_{\lambda\mu}\right) - \frac{1}{2}\delta S^{\mu,\alpha\beta}\partial_\mu\omega_{\alpha\beta}, \tag*{(10)}

where δNμ\delta N^\mu is the nonequilibrium correction to the baryon current, δTsμλ\delta T_{\mathrm{s}}^{\mu\lambda} (δTaμλ\delta T_{\mathrm{a}}^{\mu\lambda}) is the nonequilibrium symmetric (antisymmetric) correction to the energy–momentum tensor, and δSμ,αβ\delta S^{\mu,\alpha\beta} is the nonequilibrium correction to the spin tensor.

Since we are only interested in the spin effects, we set δNμ=0\delta N^\mu= 0 and δTsμλ=0\delta T_{\mathrm{s}}^{\mu\lambda} = 0. We also assume here that δSμ,αβ=0\delta S^{\mu,\alpha\beta} = 0. In this case, the formula for the entropy production reduces to

∂μsNSμ=δTaμλ(∂μβλ−ωλμ)=δTaμλ(ϖλμ−ωλμ),(11)\partial_\mu s_{\mathrm{NS}}^\mu= \delta T_{\mathrm{a}}^{\mu\lambda}\left(\partial_\mu\beta_\lambda- \omega_{\lambda\mu}\right) = \delta T_{\mathrm{a}}^{\mu\lambda}\left(\varpi_{\lambda\mu} - \omega_{\lambda\mu}\right), \tag*{(11)}

where we have introduced the thermal vorticity

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

Using the decomposition (2) we find

∂μsNSμ=dμ(2kμ−βaμ−β2∇μT)+eμν(β∇[μUν]+ϵμναβUαωβ),(13)\partial_\mu s_{\mathrm{NS}}^\mu= d^\mu\left(2k_\mu- \beta a_\mu- \beta^2\nabla_\mu T\right) + e^{\mu\nu}\left(\beta\nabla_{[\mu}U_{\nu]} + \epsilon_{\mu\nu\alpha\beta}U^\alpha\omega^\beta\right), \tag*{(13)}

where aμ=DUμ=Uλ∂λUμa^\mu= DU^\mu= U^\lambda\partial_\lambda U^\mu is the four-acceleration of the fluid element. The positivity of the entropy production is guaranteed by the conditions

dμ=dNSμ=λβ−1(βaμ+β2∇μT−2kμ),(14)d^\mu= d_{\mathrm{NS}}^\mu= \lambda\beta^{-1}\left(\beta a^\mu+ \beta^2\nabla^\mu T - 2k^\mu\right), \tag*{(14)}
eμν=eNSμν=γβ−1(β∇[μUν]+ϵμναβUαωβ).(15)e^{\mu\nu} = e_{\mathrm{NS}}^{\mu\nu} = \gamma\beta^{-1}\left(\beta\nabla^{[\mu}U^{\nu]} + \epsilon^{\mu\nu\alpha\beta}U_\alpha\omega_\beta\right). \tag*{(15)}

The coefficients λ\lambda and γ\gamma are the “new” (nonnegative) spin kinetic coefficients introduced for the first time in [14]. We have rescaled them compared to [14] in such a way that their dimension agrees with the dimension of the entropy density σ\sigma (in numerical calculations we assume proportionality between λ\lambda, γ\gamma and σ\sigma, and consider different proportionality constants between these variables). Within our assumptions, Eqs. (14) and (15) should be used in the dynamic equations

∂μ(Teqμν+dμUν−dνUμ+eμν)=0,(16)\partial_\mu\left(T_{\mathrm{eq}}^{\mu\nu}+d^\mu U^\nu-d^\nu U^\mu+e^{\mu\nu}\right)=0, \tag*{(16)}
∂λSeqλ,μν+2(dμUν−dνUμ+eμν)=0.(17)\partial_\lambda S_{\mathrm{eq}}^{\lambda,\mu\nu}+2\left(d^\mu U^\nu-d^\nu U^\mu+e^{\mu\nu}\right)=0. \tag*{(17)}

Israel–Stewart approach

In the Israel–Stewart approach [61, 62], the entropy production is defined by the formula

∂μSISμ=δTaμλ(ϖλμ−ωλμ)+∂μQμ,(18)\partial_\mu S_{\mathrm{IS}}^\mu=\delta T_a^{\mu\lambda}\left(\varpi_{\lambda\mu}-\omega_{\lambda\mu}\right)+\partial_\mu Q^\mu, \tag*{(18)}

where the four-vector QμQ^\mu is defined by the expression [52]4

Qμ=a4Uμdλdλ+a5Uμeαβeαβ.(19)Q^\mu=a_4 U^\mu d^\lambda d_\lambda+a_5 U^\mu e^{\alpha\beta}e_{\alpha\beta}. \tag*{(19)}

In this case, the vector dd satisfies the equation

τdd˙IS⟨μ⟩+dISμ=−λT(2kμ−βaμ−β2∇μT+dISμUλ∂λa4+a4ΘdISμ),(20)\tau_d\dot{d}_{\mathrm{IS}}^{\langle\mu\rangle}+d_{\mathrm{IS}}^\mu=-\lambda T\left(2k^\mu-\beta a^\mu-\beta^2\nabla^\mu T+d_{\mathrm{IS}}^\mu U^\lambda\partial_\lambda a_4+a_4\Theta d_{\mathrm{IS}}^\mu\right), \tag*{(20)}

where Θ≡∂μUμ\Theta\equiv\partial_\mu U^\mu is the expansion scalar, τd=2λTa4\tau_d=2\lambda T a_4 is the relaxation time, while the tensor eμνe^{\mu\nu} fulfills the equation

τee˙IS⟨[μν]⟩+eISμν=γβ−1(β∇[μUν]+ϵμναβUαωβ+eISμνUλ∂λa5+a5ΘeISμν),(21)\tau_e\dot{e}_{\mathrm{IS}}^{\langle[\mu\nu]\rangle}+e_{\mathrm{IS}}^{\mu\nu}=\gamma\beta^{-1}\left(\beta\nabla^{[\mu}U^{\nu]}+\epsilon^{\mu\nu\alpha\beta}U_\alpha\omega_\beta+e_{\mathrm{IS}}^{\mu\nu}U^\lambda\partial_\lambda a_5+a_5\Theta e_{\mathrm{IS}}^{\mu\nu}\right), \tag*{(21)}

where τe=−2γTa5\tau_e=-2\gamma T a_5 is another relaxation time. We note that a4>0a_4>0, while a5<0a_5<0.

Boost-invariant and transversely uniform system

In our numerical analyses, we first consider the Bjorken expansion [63], i.e., a one-dimensional, boost-invariant expansion of a system that is uniform in the transverse plane. We call this geometry of expansion shortly the TH (transversely homogeneous) model.

Basis vectors and basic symmetries

In the TH model, it is convenient to use a basis X(i)μ\mathcal{X}_{(i)}^{\mu} whose vectors have a form invariant under boosts along the zz axis. A natural choice is [64]:

X(0)μ=Uμ=1τ(t,0,0,z)=(cosh⁡η,0,0,sinh⁡η),X(1)μ=Xμ=(0,1,0,0),X(2)μ=Yμ=(0,0,1,0),X(3)μ=Zμ=1τ(z,0,0,t)=(sinh⁡η,0,0,cosh⁡η).(22)\begin{aligned} \mathcal{X}_{(0)}^{\mu} = U^{\mu} &= \frac{1}{\tau}(t,0,0,z) = (\cosh\eta,0,0,\sinh\eta),\\ \mathcal{X}_{(1)}^{\mu} = X^{\mu} &= (0,1,0,0),\\ \mathcal{X}_{(2)}^{\mu} = Y^{\mu} &= (0,0,1,0),\\ \mathcal{X}_{(3)}^{\mu} = Z^{\mu} &= \frac{1}{\tau}(z,0,0,t) = (\sinh\eta,0,0,\cosh\eta). \tag*{(22)} \end{aligned}

Here τ=t2−z2\tau= \sqrt{t^2-z^2} is the longitudinal proper time, and η=12ln⁡t+zt−z\eta= \frac{1}{2}\ln\frac{t+z}{t-z} is the spacetime rapidity. The four-vector UμU^{\mu} is timelike, while the four-vectors XμX^{\mu}, YμY^{\mu}, ZμZ^{\mu} are spacelike. In addition, they are pairwise orthogonal,

U2=1,X2=Y2=Z2=−1,U⋅X=0,U⋅Y=0,U⋅Z=0,Z⋅Y=0,Y⋅X=0,X⋅Z=0.(23)\begin{aligned} U^2 &= 1,\qquad X^2=Y^2=Z^2=-1,\\ U\cdot X &= 0,\qquad U\cdot Y=0,\qquad U\cdot Z=0,\\ Z\cdot Y &= 0,\qquad Y\cdot X=0,\qquad X\cdot Z=0. \tag*{(23)} \end{aligned}

The basis vectors (22) satisfy also the identities:

ϵαβγδUβXγYδ=−Zα,ϵαβγδUβYγZδ=−Xα,ϵαβγδUβZγXδ=−Yα.(24)\begin{aligned} \epsilon_{\alpha\beta\gamma\delta}U^{\beta}X^{\gamma}Y^{\delta} &= -Z_{\alpha},\\ \epsilon_{\alpha\beta\gamma\delta}U^{\beta}Y^{\gamma}Z^{\delta} &= -X_{\alpha},\\ \epsilon_{\alpha\beta\gamma\delta}U^{\beta}Z^{\gamma}X^{\delta} &= -Y_{\alpha}. \tag*{(24)} \end{aligned}

Boost-invariant scalar functions (such as thermodynamic quantities) depend only on the proper time. For the derivatives of scalar functions with respect to τ\tau, we introduce the dot notation

Df(τ)≡Uμ∂μf(τ)=df(τ)dτ≡f˙(τ).(25)Df(\tau) \equiv U^{\mu}\partial_{\mu}f(\tau) = \frac{df(\tau)}{d\tau} \equiv\dot{f}(\tau). \tag*{(25)}

The projector Δμν≡gμν−UμUν\Delta^{\mu\nu} \equiv g^{\mu\nu}-U^{\mu}U^{\nu}, orthogonal to the flow in both indices, is expressed in the basis (22) as Δμν=−XμXν−YμYν−ZμZν\Delta^{\mu\nu}=-X^{\mu}X^{\nu}-Y^{\mu}Y^{\nu}-Z^{\mu}Z^{\nu}. Using this formula, it is easy to check that the transverse gradient of scalar functions vanishes

∇μf(τ)=Δμν∂νf(τ)=0.(26)\nabla^{\mu}f(\tau)=\Delta^{\mu\nu}\partial_{\nu}f(\tau)=0. \tag*{(26)}

Using the properties listed above, one finds that the following conditions are satisfied in the TH model

aμ=0,∇μT=0,∇[μUν]=0,ϖμν=0.(27)a^{\mu}=0,\qquad\nabla^{\mu}T=0,\qquad\nabla^{[\mu}U^{\nu]}=0,\qquad\varpi^{\mu\nu}=0. \tag*{(27)}

Hence, in the TH model the Navier–Stokes expressions (14) and (15) are reduced to simple expressions

dNSTHμ=−2kμλβ−1,(28)d^\mu_{\mathrm{NSTH}}=-2k^\mu\lambda\beta^{-1}, \tag*{(28)}
ϵNSTHμν=γβ−1ϵμναβUαωβ,(29)\epsilon^{\mu\nu}_{\mathrm{NSTH}}=\gamma\beta^{-1}\epsilon^{\mu\nu\alpha\beta}U_\alpha\omega_\beta, \tag*{(29)}

with the entropy production given by the formula

∂μsNSTHμ=−4λβ−1kμkμ−2γβ−1ωμωμ,(30)\partial_\mu s^\mu_{\mathrm{NSTH}}=-4\lambda\beta^{-1}k^\mu k_\mu-2\gamma\beta^{-1}\omega^\mu\omega_\mu, \tag*{(30)}

which is manifestly positive, since kμkμ≤0k^\mu k_\mu\le0 and ωμωμ≤0\omega^\mu\omega_\mu\le0. It is important to notice that on the right-hand sides of Eqs. (28) and (29) no terms with gradients of hydrodynamic variables appear. This is so because the entropy production is proportional to the difference between the spin polarization tensor and the thermal vorticity, and the latter vanishes in the TH model. Consequently, we expect that dissipation in the TH model leads to destruction of the spin polarization, if such is present at the initial stage of the system evolution.

Given the orthogonality conditions, k⋅U=0k\cdot U=0 and ω⋅U=0\omega\cdot U=0, one can introduce the following decompositions [16]:

kμ=CkxXμ+CkyYμ+CkzZμ=(Ckzsinh⁡η,Ckx,Cky,Ckzcosh⁡η),(31)k^\mu=C_{kx}X^\mu+C_{ky}Y^\mu+C_{kz}Z^\mu=(C_{kz}\sinh\eta,C_{kx},C_{ky},C_{kz}\cosh\eta), \tag*{(31)}
ωμ=CωxXμ+CωyYμ+CωzZμ=(Cωzsinh⁡η,Cωx,Cωy,Cωzcosh⁡η),(32)\omega^\mu=C_{\omega x}X^\mu+C_{\omega y}Y^\mu+C_{\omega z}Z^\mu=(C_{\omega z}\sinh\eta,C_{\omega x},C_{\omega y},C_{\omega z}\cosh\eta), \tag*{(32)}

where the coefficient functions CkiC_{ki} and CωiC_{\omega i} (i=x,y,zi=x,y,z) depend on the proper time only. Analogous decompositions hold for the vectors dd and ee:

dμ=dxXμ+dyYμ+dzZμ=(dzsinh⁡η,dx,dy,dzcosh⁡η),(33)d^\mu=d_xX^\mu+d_yY^\mu+d_zZ^\mu=(d_z\sinh\eta,d_x,d_y,d_z\cosh\eta), \tag*{(33)}
eμ=exXμ+eyYμ+ezZμ=(ezsinh⁡η,ex,ey,ezcosh⁡η),(34)e^\mu=e_xX^\mu+e_yY^\mu+e_zZ^\mu=(e_z\sinh\eta,e_x,e_y,e_z\cosh\eta), \tag*{(34)}

where the coefficient functions did_i and eie_i also depend only on the proper time. The properties of the tensor eμνe^{\mu\nu} allow us to represent it also in the form

eμν=−ex(τ)(YμZν−YνZμ)−ey(τ)(ZμXν−ZνXμ)−ez(τ)(XμYν−XνYμ).(35)e^{\mu\nu}=-e_x(\tau)\left(Y^\mu Z^\nu-Y^\nu Z^\mu\right)-e_y(\tau)\left(Z^\mu X^\nu-Z^\nu X^\mu\right)-e_z(\tau)\left(X^\mu Y^\nu-X^\nu Y^\mu\right). \tag*{(35)}

The equivalence of Eq. (35) with Eq. (3) can be directly established with the help of the identities (24). Below, we frequently use the three-vector notation

Ck=(Ckx,Cky,Ckz),Cω=(Cωx,Cωy,Cωz),\mathbf{C}_k=(C_{kx},C_{ky},C_{kz}),\qquad\mathbf{C}_\omega=(C_{\omega x},C_{\omega y},C_{\omega z}),
d=(dx,dy,dz),e=(ex,ey,ez).(36)\mathbf{d}=(d_x,d_y,d_z),\qquad\mathbf{e}=(e_x,e_y,e_z). \tag*{(36)}

From now on, we will refer to the vectors d\mathbf{d} and e\mathbf{e} as to the electric and magnetic dissipation vectors, or shortly, as to the electric and magnetic dissipation.

Conservation laws

Baryon number, energy, linear momentum

Since we consider a system with finite baryon number density nn, we have to include the baryon number conservation law, ∂μ(nUμ)=0\partial_\mu(nU^\mu)=0, which for the TH model takes the form

dndτ+nτ=0.(37)\frac{\mathrm{d}n}{\mathrm{d}\tau}+\frac{n}{\tau}=0. \tag*{(37)}

The energy–momentum conservation equation (16) contracted with Xν(i)X_\nu^{(i)} gives additional four equations,

ε˙+ε+Pτ=0,(38)\dot{\varepsilon}+\frac{\varepsilon+P}{\tau}=0, \tag*{(38)}
d˙x+dxτ=0,(39)\dot{d}_x+\frac{d_x}{\tau}=0, \tag*{(39)}
d˙y+dyτ=0,(40)\dot{d}_y+\frac{d_y}{\tau}=0, \tag*{(40)}
d˙z=0.(41)\dot{d}_z=0. \tag*{(41)}

A new and striking feature of the equations above, which can be regarded as equations of motion of the fluid, is the presence of the coefficients of the vector dd in them. This is a consequence of an asymmetric energy–momentum tensor. Interestingly, the components of the tensor eμνe^{\mu\nu} do not appear in Eqs. (38)–(41), since ∂μeμν=0\partial_\mu e^{\mu\nu}=0. This property follows from the structure of Eq. (35).

We note that the solutions of Eq. (37) and Eqs. (38)–(41) have simple forms

n(τ)=τ0τn0,dx(τ)=τ0τdx0,dy(τ)=τ0τdy0,dz(τ)=dz0,(42)n(\tau)=\frac{\tau_0}{\tau}n^0,\qquad d_x(\tau)=\frac{\tau_0}{\tau}d_x^0,\qquad d_y(\tau)=\frac{\tau_0}{\tau}d_y^0,\qquad d_z(\tau)=d_z^0, \tag*{(42)}

where n0n^0, dx0d_x^0, dy0d_y^0 and dz0d_z^0 are constants, while τ0\tau_0 is the initial proper time for the hydrodynamic evolution. With the help of the thermodynamic identities dP=Tdσ+μdn\mathrm{d}P=T\mathrm{d}\sigma+\mu\mathrm{d}n and ε+P=Tσ+μn\varepsilon+P=T\sigma+\mu n, where σ\sigma is the entropy density, one can show that Eq. (38) is reduced to the entropy conservation in the form dσ/dτ+σ/τ=0\mathrm{d}\sigma/\mathrm{d}\tau+\sigma/\tau=0, again with the scaling solution σ=σ0τ0/τ\sigma=\sigma_0\tau_0/\tau.

Angular momentum

The angular momentum conservation equation (17) contracted with pairs of the basis vectors UμXνU_\mu X_\nu, UμYνU_\mu Y_\nu, UμZνU_\mu Z_\nu, YμZνY_\mu Z_\nu, ZμXνZ_\mu X_\nu, and XμYνX_\mu Y_\nu gives six equations:

\begin{pmatrix} A & 0 & 0 & 0 & 0 & 0 \\ 0 & A & 0 & 0 & 0 & 0 \\ 0 & 0 & A & 0 & 0 & 0 \\ 0 & 0 & 0 & A_{1} & 0 & 0 \\ 0 & 0 & 0 & 0 & A_{1} & 0 \\ 0 & 0 & 0 & 0 & 0 & A_{1} \end{pmatrix} \begin{pmatrix} \dot{C}_{kx} \\ \dot{C}_{ky} \\ \dot{C}_{kz} \\ \dot{C}_{\omega x} \\ \dot{C}_{\omega y} \\ \dot{C}_{\omega z} \end{pmatrix} = \begin{pmatrix} Q_{x} & 0 & 0 & 0 & 0 & 0 \\ 0 & Q_{y} & 0 & 0 & 0 & 0 \\ 0 & 0 & Q_{z} & 0 & 0 & 0 \\ 0 & 0 & 0 & R_{x} & 0 & 0 \\ 0 & 0 & 0 & 0 & R_{y} & 0 \\ 0 & 0 & 0 & 0 & 0 & R_{z} \end{pmatrix} \begin{pmatrix} C_{kx} \\ C_{ky} \\ C_{kz} \\ C_{\omega x} \\ C_{\omega y} \\ C_{\omega z} \end{pmatrix} -2 \begin{pmatrix} d_{x} \\ d_{y} \\ d_{z} \\ e_{x} \\ e_{y} \\ e_{z} \end{pmatrix}. \tag*{(43)}\end{pmatrix}

Here the functions AA and A1A_{1} define spin densities. The other functions are defined through the relations

Qx=Qy=−(A˙+3A2τ),Qz=−(A˙+Aτ).(44)Q_{x}=Q_{y}=-\left(\dot{A}+\frac{3A}{2\tau}\right), \qquad Q_{z}=-\left(\dot{A}+\frac{A}{\tau}\right). \tag*{(44)}
Rx=Ry=−(A˙1+2A1−A2τ),Rz=−(A˙1+A1τ).(45)R_{x}=R_{y}=-\left(\dot{A}_{1}+\frac{2A_{1}-A}{2\tau}\right), \qquad R_{z}=-\left(\dot{A}_{1}+\frac{A_{1}}{\tau}\right). \tag*{(45)}

Equations (37), (38)–(41), and (43) represent a system of eleven equations for fourteen unknown functions of the proper time: TT, μ\mu, dxd_{x}, dyd_{y}, dzd_{z}, exe_{x}, eye_{y}, eze_{z}, CkxC_{kx}, CkyC_{ky}, CkzC_{kz}, CωxC_{\omega x}, CωyC_{\omega y}, and CωzC_{\omega z}. The main question now is, however, how they can be solved together with the constraints following from the entropy production condition. This issue is discussed in the following section.

Entropy production constraints (Navier–Stokes)

In the TH model, the NS entropy production (13) equals

∂μSNSμ=−2d⋅Ck+2e⋅Cω(46)\partial_{\mu}S^{\mu}_{\mathrm{NS}}=-2d\cdot C_{k}+2e\cdot C_{\omega} \tag*{(46)}

and is positive if

d=−2λTCk,e=γTCω,(47)d=-2\lambda T C_{k}, \qquad e=\gamma T C_{\omega}, \tag*{(47)}

which is consistent with Eqs. (28) and (29). It is readily seen that Eqs. (47) represent additional constraints on the system of Eqs. (37), (38)–(41), and (43). With Eqs. (47) included, the system of Eqs. (37), (38)–(41), and (43) becomes overdetermined – we obtain seventeen equations for fourteen unknowns.

The origin of this difficulty is the appearance of the vector dd in the equations of motion (38)–(41). If the components of dd were neglected in (38)–(41), we would have equal numbers of equations and unknown functions. Neglecting dd in (38)–(41), is, however, inconsistent with keeping it on the right-hand side of Eq. (43). These arguments motivate us to conclude that the functions CkC_k and, consequently, d\boldsymbol{d} should be set equal to zero. This assumption is consistent with the fact that thermal vorticity is zero and the inclusion of dissipation should result in ω\boldsymbol{\omega} approaching ϖ=0\boldsymbol{\varpi} = 0.

It remains for us to explain the behavior in the magnetic sector described by the vectors e\boldsymbol{e} and Cω\boldsymbol{C}_{\omega}. This is done numerically. The coupled system of ordinary differential equations is solved by employing a Runge–Kutta integration scheme. The initial proper time for the hydrodynamic evolution is set to τ0=1.0\tau_0 = 1.0 fm. The hydrodynamic background is initialized with the temperature T0=0.155T_0 = 0.155 GeV and the baryon chemical potential μ0=0.8\mu_0 = 0.8 GeV. The particle mass is set to m=1.116m = 1.116 GeV, which corresponds to the Λ\Lambda hyperon. The initial conditions for the spin polarization components are fixed uniformly at Cωx(τ0)=Cωy(τ0)=Cωz(τ0)=1C_{\omega x}(\tau_0) = C_{\omega y}(\tau_0) = C_{\omega z}(\tau_0) = 1. The numerical integration is performed up to τf=10.0\tau_f = 10.0 fm. The resulting evolution is presented in Figure 1. We explicitly assume that the spin kinetic coefficient γ\gamma is proportional to the entropy density, using the form γ=γ0σ\gamma= \gamma_0 \sigma. The curves demonstrate the relaxation dynamics governed by different values of the dimensionless parameter γ0∈{0.00,0.01,0.02,0.05}\gamma_0 \in\{0.00, 0.01, 0.02, 0.05\}.

Figure 1. Proper-time evolution of (a) transverse spin polarization Cωx(τ)C_{\omega x}(\tau) and (b) longitudinal spin polarization Cωz(τ)C_{\omega z}(\tau). The curves show the relaxation dynamics for different values of the dimensionless kinetic coefficients γ0∈{0.00,0.01,0.02,0.05}\gamma_0 \in\{0.00, 0.01, 0.02, 0.05\} with initial conditions Cωx(τ0)=Cωz(τ0)=1C_{\omega x}(\tau_0) = C_{\omega z}(\tau_0) = 1.

In the nondissipative case (γ0=0.00\gamma_0 = 0.00), the spin polarization continuously increases. This behavior has been already found in the studies of perfect fluid hydrodynamics [16]. The inclusion of the entropy production constraints introduces damping effect on the spin polarization. This damping becomes stronger for larger values of γ0\gamma_0.

Entropy production constraints (Israel–Stewart)

In the IS case and TH geometry, Eqs. (20) and (21) become two differential vector equations:

τdd˙+d=−λT(2Ck+a˙4d+a4τd)(48)\tau_d \dot{\mathbf{d}}+\mathbf{d}=-\lambda T\left(2C_k+\dot{a}_4\mathbf{d}+\frac{a_4}{\tau}\mathbf{d}\right) \tag*{(48)}

and

τee˙+e=γT(Cω+a˙5e+a5τe).(49)\tau_e \dot{e}+e=\gamma T\left(C_\omega+\dot{a}_5e+\frac{a_5}{\tau}e\right). \tag*{(49)}

Similarly to the NS case, we observe that the system is overdetermined, unless the dissipation vector d\mathbf{d} and the related vector CkC_k vanish. Consequently, we set d=Ck=0\mathbf{d}=C_k=0 and consider only the evolution of the vector ee governed by Eq. (49).

The numerical integration scheme and the thermodynamic background parameters remain identical to those used in Sec. IV C. To examine the dynamics of the dissipation vector ee, we first set the relaxation time to τe=1.0\tau_e=1.0 fm and vary the initial condition e0=e(τ0)e_0=e(\tau_0). We explore three distinct scenarios: e0=0e_0=0, e0=0.5 eNSe_0=0.5\,e_{\mathrm{NS}}, and e0=1.0 eNSe_0=1.0\,e_{\mathrm{NS}}. Here, eNS=γ(τ0)T(τ0)Cω(τ0)e_{\mathrm{NS}}=\gamma(\tau_0)T(\tau_0)C_\omega(\tau_0) denotes the initial dissipation vector predicted by the Navier–Stokes limit. The inclusion of an independent dynamic equation for e(τ)e(\tau) leads to the behavior depicted in Figure 2, which shows the proper time evolution of the components ex(τ)e_x(\tau), ez(τ)e_z(\tau), Cωx(τ)C_{\omega x}(\tau), and Cωz(τ)C_{\omega z}(\tau). Due to the transverse homogeneity, the evolutions of the xx and yy components are the same.

Figure 2. Proper-time evolution of the components of the magnetic dissipative vector ee and the components of the spin polarization vector CωC_\omega in the Israel–Stewart formulation with τe=1.0\tau_e = 1.0 fm, compared with the Navier–Stokes asymptotic limit (e=γTCωe = \gamma T C_\omega). The panels show (a) the transverse component ex(τ)e_x(\tau), (b) the transverse spin polarization component Cωx(τ)C_{\omega x}(\tau), (c) the longitudinal component ez(τ)e_z(\tau), and (d) the longitudinal spin polarization component Cωz(τ)C_{\omega z}(\tau). Different line styles correspond to different initial dissipation magnitudes of e0e_0, as described in the legend.

If the system is initialized with zero anisotropy (e0=0e_0=0), the dissipative vector requires certain time to develop. This delay implies that the initial decay of the spin polarization is noticeably slower than in the Navier–Stokes limit – as long as the right-hand side of (43) is negligible, the spin evolution is governed by the perfect-fluid equations. In contrast, if the system is initialized with the fully developed dissipation vector (e0=1.0 eNSe_0=1.0\,e_{\mathrm{NS}}), the spin polarization is subject to strong damping. Regardless of the chosen initial state, all the Israel–Stewart cases converge towards the Navier–Stokes result. This behavior is similar to that found in standard relativistic hydrodynamics, where the IS dynamics reaches local equilibrium by first approaching the NS regime.

To investigate the role of the relaxation time in the Israel–Stewart formulation, we perform computations for τe=0.5\tau_e = 0.5 fm, τe=1.0\tau_e = 1.0 fm, and τe=2.0\tau_e = 2.0 fm, assuming e0=0e_0 = 0. The corresponding evolution of the dissipation and spin polarization vector components is presented in Figure 3. The relaxation time determines how quickly the dissipation vector components develop and approach the NS limit. For larger values of τe\tau_e, the dissipation vector requires more time to reach this limit.

Figure 3. Proper-time evolution of the dissipation and spin polarization vectors in the Israel–Stewart formulation for different relaxation times, τe∈{0.5,1.0,2.0}\tau_e \in\{0.5, 1.0, 2.0\} fm. The system is initialized at e0=0e_0 = 0 with Cω(τ0)=1C_\omega(\tau_0) = 1. The panels display (a) transverse dissipation exe_x, (b) transverse spin polarization CωxC_{\omega x} (with Cωy≡CωxC_{\omega y} \equiv C_{\omega x}), (c) longitudinal dissipation eze_z, and (d) longitudinal spin polarization CωzC_{\omega z}. All results obtained in Israel–Stewart approach are compared to the Navier–Stokes limit (black dashed lines).

Our results for a scenario where the system is initialized in the NS regime, e(τ0)=eNS(τ0)e(\tau_0) = e_{\mathrm{NS}}(\tau_0), are shown in Figure 4. To observe the full relaxation process, numerical integration has been extended to the final time τf=15.0\tau_f = 15.0 fm. We observe in this case that the IS evolution of e(τ)e(\tau) follows closely the NS dynamics, although small differences can be seen. Larger differences between the IS and NS results are seen for the spin polarization. As expected, for larger relaxation times the IS dynamics approaches the NS limit later. Interestingly, for τe=2.0\tau_e = 2.0 fm the components of the spin polarization vector (being positive initially) change sign during the time evolution and approach zero from below – the corresponding NS values approach zero from above.

Figure 4. Proper-time evolution of the dissipation and spin polarization vectors for the case e(τ0)=eNS(τ0)e(\tau_0)=e_{\mathrm{NS}}(\tau_0) and different relaxation times, τe∈{0.5,1.0,2.0}\tau_e \in\{0.5,1.0,2.0\} fm. The final evolution time is τf=15.0\tau_f=15.0 fm. The panels display (a) transverse dissipation ex(τ)e_x(\tau), (b) transverse spin polarization Cωx(τ)C_{\omega x}(\tau), (c) longitudinal dissipation ez(τ)e_z(\tau), and (d) longitudinal spin polarization Cωz(τ)C_{\omega z}(\tau).

Boost-invariant and cylindrically symmetric systems

We turn now to a discussion of the system that is boost-invariant and cylindrically symmetric in the transverse plane. A novel feature, compared to the TH model, is the inclusion of the radial flow. We call this type of expansion shortly cylindrically symmetric (CS). We follow essentially the same steps as in the discussion of the TH model. The basic difference is in the more complicated tensor structures present in the CS formulation. A version of the CS model featuring a perfect spin fluid was recently investigated in [58]; the present work can therefore be viewed as a direct extension of [58] that incorporates dissipative phenomena.

Basis vectors and basic symmetries

The boost-invariant and cylindrically symmetric basis X(i)μ\mathcal{X}_{(i)}^\mu (i=0,1,2,3i=0,1,2,3) consists now of the vectors UU, RR, Φ\Phi, and ZZ defined by the following expressions [64]:

X(0)μ=Uμ=(cosh⁡θcosh⁡η,sinh⁡θcos⁡ϕ,sinh⁡θsin⁡ϕ,cosh⁡θsinh⁡η),(50)\mathcal{X}_{(0)}^\mu= U^\mu= \left(\cosh\theta\cosh\eta,\sinh\theta\cos\phi,\sinh\theta\sin\phi,\cosh\theta\sinh\eta\right), \tag*{(50)}
X(1)μ=Rμ=(sinh⁡θcosh⁡η,cosh⁡θcos⁡ϕ,cosh⁡θsin⁡ϕ,sinh⁡θsinh⁡η),(51)\mathcal{X}_{(1)}^\mu= R^\mu= \left(\sinh\theta\cosh\eta,\cosh\theta\cos\phi,\cosh\theta\sin\phi,\sinh\theta\sinh\eta\right), \tag*{(51)}
X(2)μ=Φμ=(0,−sin⁡ϕ,cos⁡ϕ,0),(52)\mathcal{X}_{(2)}^\mu= \Phi^\mu= \left(0,-\sin\phi,\cos\phi,0\right), \tag*{(52)}

and

X(3)μ=Zμ=(sinh⁡η,0,0,cosh⁡η).(53)\mathcal{X}_{(3)}^\mu= Z^\mu= \left(\sinh\eta,0,0,\cosh\eta\right). \tag*{(53)}

See Figure 5 for a schematic view. Here rr and ϕ\phi are polar coordinates in the transverse plane (radius and azimuthal angle), while θ=θ(τ,r)\theta= \theta(\tau,r) is the transverse rapidity. At z=0z = 0 (η=0\eta= 0) the transverse velocity vr(τ,r)=tanh⁡θ(τ,r)v_r(\tau,r) = \tanh\theta(\tau,r). Similarly to (23), we have

Figure 5. Schematic view of the CS geometry. (a) Transverse plane z=0z = 0: the red circles denote surfaces of constant rr, the black arrows indicate the radial flow with the transverse velocity vr=tanh⁡θv_r = \tanh\theta, and the blue arrows show the basis vectors RR and Φ\Phi, Eqs. (51) and (52), at a point with the azimuthal angle ϕ\phi; the zz axis points out of the plane. (b) Longitudinal (z,t)(z,t) plane at a fixed transverse position: hyperbolas of constant proper time τ\tau (red), lines of constant spacetime rapidity η\eta (gray), and the light cone (dashed). The basis vector UU is tangent to the line η=const\eta= \mathrm{const}, while ZZ, Eq. (53), is tangent to the hyperbola τ=const\tau= \mathrm{const}.

U2=1,Z2=R2=Φ2=−1,U⋅Z=0,U⋅R=0,U⋅Φ=0,Z⋅R=0,R⋅Φ=0,Φ⋅Z=0.(54)U^2 = 1,\quad Z^2 = R^2 = \Phi^2 = -1, \\ U \cdot Z = 0,\quad U \cdot R = 0,\quad U \cdot\Phi= 0, \\ Z \cdot R = 0,\quad R \cdot\Phi= 0,\quad\Phi\cdot Z = 0. \tag*{(54)}

The analog of the identities (24) is the following set of equations

ϵαβγδUβRγΦδ=−Zα,ϵαβγδUβΦγZδ=−Rα,ϵαβγδUβZγRδ=−Φα.(55)\epsilon_{\alpha\beta\gamma\delta} U^\beta R^\gamma\Phi^\delta= -Z_\alpha, \\ \epsilon_{\alpha\beta\gamma\delta} U^\beta\Phi^\gamma Z^\delta= -R_\alpha, \\ \epsilon_{\alpha\beta\gamma\delta} U^\beta Z^\gamma R^\delta= -\Phi_\alpha. \tag*{(55)}

In the TH model we deal with scalar functions of the proper time. In the CS model, scalar functions depend on the proper time and the distance rr from the symmetry axis. For such functions, the total time derivative equals

Df(τ,r)≡Uμ∂μf(τ,r)=(cosh⁡θ ∂τ+sinh⁡θ ∂r)f(τ,r)=(cosh⁡θ f˙(τ,r)+sinh⁡θ f′(τ,r)).(56)Df(\tau,r) \equiv U^\mu\partial_\mu f(\tau,r) = \left(\cosh\theta\,\partial_\tau+\sinh\theta\,\partial_r\right)f(\tau,r) = \left(\cosh\theta\,\dot{f}(\tau,r)+\sinh\theta\,f'(\tau,r)\right). \tag*{(56)}

Here we have introduced a dot (a prime) as the symbol of the partial derivative with respect to proper time (radius coordinate). Similarly, we define the transverse radial derivative

∇Rf(τ,r)≡Rμ∂μf(τ,r)=(sinh⁡θ f˙(τ,r)+cosh⁡θ f′(τ,r)).(57)\nabla_R f(\tau,r) \equiv R^\mu\partial_\mu f(\tau,r) = \left(\sinh\theta\,\dot{f}(\tau,r)+\cosh\theta\,f'(\tau,r)\right). \tag*{(57)}

One can notice that the operators DD and ∇R\nabla_R differ by exchange of the hyperbolic functions. The transverse gradient of a scalar function can be written as

∇μf(τ,r)=Δμν∂νf(τ,r)=−(sinh⁡θ f˙(τ,r)+cosh⁡θ f′(τ,r))Rμ=−∇Rf(τ,r)Rμ.(58)\nabla^\mu f(\tau,r) = \Delta^{\mu\nu}\partial_\nu f(\tau,r) = -\left(\sinh\theta\,\dot{f}(\tau,r)+\cosh\theta\,f'(\tau,r)\right)R^\mu= -\nabla_R f(\tau,r)R^\mu. \tag*{(58)}

Furthermore, for the CS model one finds:

aμ=(cosh⁡θ θ˙+sinh⁡θ θ′)Rμ=Dθ(τ,r)Rμ≡ar(τ,r)Rμ,(59)a^\mu= \left(\cosh\theta\,\dot{\theta}+\sinh\theta\,\theta'\right)R^\mu= D\theta(\tau,r)R^\mu\equiv a_r(\tau,r)R^\mu, \tag*{(59)}
∇[μUν]=0.(60)\nabla^{[\mu}U^{\nu]}=0. \tag*{(60)}

Using Eqs. (58) and (60), one finds that the thermal vorticity ϖμν\varpi_{\mu\nu} has the structure of Eq. (7) with the vanishing magneticlike part, namely

ϖμν=kμeqUν−kνeqUμ=Ckreq(RμUν−RνUμ)(61)\varpi_{\mu\nu}=k_\mu^{\mathrm{eq}}U_\nu-k_\nu^{\mathrm{eq}}U_\mu=C_{kr}^{\mathrm{eq}}\left(R_\mu U_\nu-R_\nu U_\mu\right) \tag*{(61)}

where

Ckreq=β2(Dθ−β∇RT)=β2(ar−β∇RT).(62)C_{kr}^{\mathrm{eq}}=\frac{\beta}{2}\left(D\theta-\beta\nabla_R T\right)=\frac{\beta}{2}\left(a_r-\beta\nabla_R T\right). \tag*{(62)}

Following [16], we denote the coefficients of thermal vorticity by the superscript “eq”, since they define the values of the coefficients of the spin polarization tensor at which the entropy production vanishes.

For the spin polarization tensor ω\omega, we use now a parametrization whose form is the same as Eq. (7), however, the vectors kk and ω\omega are expanded in the basis (51)–(53),

kμ=CkrRμ+CkϕΦμ+CkzZμ,ωμ=CωrRμ+CωϕΦμ+CωzZμ.(63)k^\mu=C_{kr}R^\mu+C_{k\phi}\Phi^\mu+C_{kz}Z^\mu,\qquad\omega^\mu=C_{\omega r}R^\mu+C_{\omega\phi}\Phi^\mu+C_{\omega z}Z^\mu. \tag*{(63)}

Similarly, the dissipative vector dμd^\mu and the antisymmetric tensor eaμνe_a^{\mu\nu} can be decomposed as

dμ=drRμ+dϕΦμ+dzZμ,(64)d^\mu=d_rR^\mu+d_\phi\Phi^\mu+d_zZ^\mu, \tag*{(64)}
eaμν=−er(ΦμZν−ΦνZμ)−eϕ(ZμRν−ZνRμ)−ez(RμΦν−RνΦμ).(65)e_a^{\mu\nu}=-e_r\left(\Phi^\mu Z^\nu-\Phi^\nu Z^\mu\right)-e_\phi\left(Z^\mu R^\nu-Z^\nu R^\mu\right)-e_z\left(R^\mu\Phi^\nu-R^\nu\Phi^\mu\right). \tag*{(65)}

Below, following (36), we use the three-vector notation

Ck=(Ckr,Ckϕ,Ckz),Cω=(Cωr,Cωϕ,Cωz),(66)C_k = (C_{kr}, C_{k\phi}, C_{kz}), \qquad C_\omega= (C_{\omega r}, C_{\omega\phi}, C_{\omega z}), \tag*{(66)}
d=(dr,dϕ,dz),e=(er,eϕ,ez).(67)\mathbf{d} = (d_r, d_\phi, d_z), \qquad\mathbf{e} = (e_r, e_\phi, e_z). \tag*{(67)}

We note that all the scalar coefficients depend exclusively on proper time τ\tau and radial coordinate rr.

Similarly as in the TH case, we call the vectors d\mathbf{d} and e\mathbf{e} the electric and magnetic dissipation vectors.

Conservation laws

In this section we present the explicit forms of all the conservation equations used in the CS model.

Baryon number conservation

Implementation of boost invariance and cylindrical symmetry in the baryon conservation equation ∂μ(nUμ)=0\partial_\mu(nU^\mu)=0 gives

(∂∂τ+tanh⁡θ∂∂r)n+n[(1τ+∂θ∂r)+tanh⁡θ(1r+∂θ∂τ)]=0.(68)\left(\frac{\partial}{\partial\tau}+\tanh\theta\frac{\partial}{\partial r}\right)n+n\left[\left(\frac{1}{\tau}+\frac{\partial\theta}{\partial r}\right)+\tanh\theta\left(\frac{1}{r}+\frac{\partial\theta}{\partial\tau}\right)\right]=0. \tag*{(68)}

In the limit θ→0\theta\to0, (68) reduces to (37).

Energy–momentum conservation

Projecting Eq. (4) onto the flow vector UνU_\nu, we obtain the energy evolution equation

Dϵ+Θ(ϵ+P)+∇Rdr+dr(sinh⁡θτ+cosh⁡θr)=0.(69)D\epsilon+\Theta(\epsilon+P)+\nabla_R d_r+d_r\left(\frac{\sinh\theta}{\tau}+\frac{\cosh\theta}{r}\right)=0. \tag*{(69)}

Similarly, contraction of Eq. (4) with the radial vector RνR_\nu yields the momentum conservation equation in the transverse plane

∇RP+(ϵ+P)Dθ−Ddr−dr(cosh⁡θτ+sinh⁡θr)=0.(70)\nabla_R P+(\epsilon+P)D\theta-Dd_r-d_r\left(\frac{\cosh\theta}{\tau}+\frac{\sinh\theta}{r}\right)=0. \tag*{(70)}

The projections onto the azimuthal vector Φν\Phi_\nu and the longitudinal vector ZνZ_\nu receive no contributions from the symmetric equilibrium tensor TeqμνT^{\mu\nu}_{\mathrm{eq}}, isolating the dynamics of the transverse dissipative degrees of freedom. For the azimuthal projection, we find

Ddϕ+dϕ(∇Rθ+cosh⁡θτ)+∇Rez+ez(Dθ+sinh⁡θτ)=0,(71)Dd_\phi+d_\phi\left(\nabla_R\theta+\frac{\cosh\theta}{\tau}\right)+\nabla_R e_z+e_z\left(D\theta+\frac{\sinh\theta}{\tau}\right)=0, \tag*{(71)}

whereas for the longitudinal projection we obtain

Ddz+dz(∇Rθ+sinh⁡θr)−∇Reϕ−eϕ(Dθ+cosh⁡θr)=0.(72)D d_z+d_z\left(\nabla_R\theta+\frac{\sinh\theta}{r}\right)-\nabla_R e_\phi-e_\phi\left(D\theta+\frac{\cosh\theta}{r}\right)=0. \tag*{(72)}

It is interesting to compare Eqs. (69)–(72) with Eqs. (38)–(41). Three out of four equations in (38)–(41) can be interpreted as constraints on spin dynamics. In the CS model, only two out of four equations in (69)–(72) play such a role. We thus see that a more developed flow pattern restricts the spin dynamics less.

In the CS model, we also observe that the dissipation terms appear in Eqs. (69) and (70). This means that we cannot separate a spinless hydrodynamic background from spin evolution, unless the terms involving drd_r are numerically negligible in Eqs. (69) and (70).

Angular momentum

(a) The radial electriclike sector obtained from the contraction of Eq. (17) with UμRνU_\mu R_\nu,

D(ACkr)+ACkr(Θ+cosh⁡θ2τ+sinh⁡θ2r)=−2dr.(73)D(AC k_r)+AC k_r\left(\Theta+\frac{\cosh\theta}{2\tau}+\frac{\sinh\theta}{2r}\right)=-2d_r. \tag*{(73)}

(b) The radial magneticlike sector obtained from the contraction of Eq. (17) with ΦμZν\Phi_\mu Z_\nu,

D(A1Cωr)+Cωr[A1Θ−A2(cosh⁡θτ+sinh⁡θr)]=−2er.(74)D(A_1C\omega_r)+C\omega_r\left[A_1\Theta-\frac{A}{2}\left(\frac{\cosh\theta}{\tau}+\frac{\sinh\theta}{r}\right)\right]=-2e_r. \tag*{(74)}

(c) The longitudinal magneticlike sector obtained from the contractions of Eq. (17) with UμΦνU_\mu\Phi_\nu and RμΦνR_\mu\Phi_\nu,

D(ACkϕ)−12∇R(ACωz)+A[32Ckϕ(Θ−sinh⁡θ3r)−Cωz2(Dθ+sinh⁡θτ)]+A1CωzDθ+2dϕ=0,(75)D(AC k_\phi)-\frac{1}{2}\nabla_R(AC\omega_z)+A\left[\frac{3}{2}C k_\phi\left(\Theta-\frac{\sinh\theta}{3r}\right)-\frac{C\omega_z}{2}\left(D\theta+\frac{\sinh\theta}{\tau}\right)\right]+A_1C\omega_zD\theta+2d_\phi=0, \tag*{(75)}
D(A1Cωz)+12∇R(ACkϕ)+A[32Ckϕ(Dθ+cosh⁡θ3r)−Cωz2(Θ−cosh⁡θτ)]+A1CωzΘ+2ez=0.(76)D(A_1C\omega_z)+\frac{1}{2}\nabla_R(AC k_\phi)+A\left[\frac{3}{2}C k_\phi\left(D\theta+\frac{\cosh\theta}{3r}\right)-\frac{C\omega_z}{2}\left(\Theta-\frac{\cosh\theta}{\tau}\right)\right]+A_1C\omega_z\Theta+2e_z=0. \tag*{(76)}

(d) The longitudinal electriclike sector obtained from the contractions of Eq. (17) with UμZνU_\mu Z_\nu and ZμRνZ_\mu R_\nu,

D(ACkz)+12∇R(ACωϕ)+A[32Ckz(Θ−cosh⁡θ3τ)+Cωϕ2(Dθ+cosh⁡θr)]−A1CωϕDθ+2dz=0,(77)D(AC k_z)+\frac{1}{2}\nabla_R(AC\omega_\phi)+A\left[\frac{3}{2}C k_z\left(\Theta-\frac{\cosh\theta}{3\tau}\right)+\frac{C\omega_\phi}{2}\left(D\theta+\frac{\cosh\theta}{r}\right)\right]-A_1C\omega_\phi D\theta+2d_z=0, \tag*{(77)}
D(A1Cωϕ)−12∇R(ACkz)−A[32Ckz(Dθ+sinh⁡θ3τ)+Cωϕ2(Θ−sinh⁡θr)]+A1CωϕΘ+2eϕ=0.(78)D(A_1C\omega_\phi)-\frac{1}{2}\nabla_R(AC k_z)-A\left[\frac{3}{2}C k_z\left(D\theta+\frac{\sinh\theta}{3\tau}\right)+\frac{C\omega_\phi}{2}\left(\Theta-\frac{\sinh\theta}{r}\right)\right]+A_1C\omega_\phi\Theta+2e_\phi=0. \tag*{(78)}

Entropy production constraints (Navier–Stokes)

Using Eqs. (58) and (60) and the decompositions (66), (64), and (65), we find that in the CS model the NS entropy production (13) equals

∂μSNSμ=dr(βDθ−β2∇RT)−2 d⋅Ck+2 e⋅Cω(79)\partial_\mu S_{\mathrm{NS}}^\mu= d_r \left(\beta D\theta-\beta^2\nabla_R T\right)-2\,d\cdot C_k+2\,e\cdot C_\omega \tag*{(79)}

and is positive if

dr=λT(βDθ−β2∇RT−2Ckr),dϕ=−2λTCkϕ,dz=−2λTCkz,(80)d_r=\lambda T\left(\beta D\theta-\beta^2\nabla_R T-2C_{kr}\right),\quad d_\phi=-2\lambda T C_{k\phi},\quad d_z=-2\lambda T C_{kz}, \tag*{(80)}
e=γTCω,(81)e=\gamma T C_\omega, \tag*{(81)}

which is consistent with Eqs. (14) and (15). Equations (80) and (81) are the CS analogs of Eq. (47). The only difference is the appearance of the terms βDθ\beta D\theta and −β2∇RT-\beta^2\nabla_R T in the expression for drd_r. They originate from thermal vorticity (12), which, in contrast to the TH model, does not vanish in the CS geometry. Thus the first equation in (80) can be written as dr=2λT(Ckreq−Ckr)d_r=2\lambda T(C_{kr}^{\mathrm{eq}}-C_{kr}), and the entropy production (79) becomes

∂μSNSμ=4λT[(Ckreq−Ckr)2+Ckϕ2+Ckz2]+2γTCω⋅Cω≥0.(82)\partial_\mu S_{\mathrm{NS}}^\mu=4\lambda T\left[\left(C_{kr}^{\mathrm{eq}}-C_{kr}\right)^2+C_{k\phi}^2+C_{kz}^2\right]+2\gamma T C_\omega\cdot C_\omega\geq0. \tag*{(82)}

Consequently, the entropy production vanishes if Ckr=CkreqC_{kr}=C_{kr}^{\mathrm{eq}}, while all other coefficients of the spin polarization tensor vanish. Hence, in the CS model we expect that dissipation drives the radial electriclike coefficient CkrC_{kr} towards CkreqC_{kr}^{\mathrm{eq}}, while the other components approach zero or should be equal to zero.

In fact, if the functions TT, μ\mu, and θ\theta are known, Eqs. (75) and (76) together with Eq. (71) become three equations for four unknowns: CkϕC_{k\phi}, CωzC_{\omega z}, dϕd_\phi, and eze_z. Including the NS conditions connecting dϕd_\phi with CkϕC_{k\phi} and eze_z with CωzC_{\omega z}, we obtain five equations for four unknowns, i.e., an overdetermined system. A natural solution in this case is to set all these functions equal to zero. Similar analysis of Eqs. (77) and (78) together with Eq. (72) leads us to the conclusion that CkzC_{kz}, CωϕC_{\omega\phi}, dzd_z, and eϕe_\phi are zero. Thus, only radial components drd_r and ere_r (or CkrC_{kr} and CωrC_{\omega r}) exhibit a nontrivial dynamics.

One more important remark should be made about the presence of the time derivative of drd_r in Eqs. (69) and (70). In the NS approach, this derivative should be obtained by differentiation of Eq. (80), which introduces second order proper-time derivatives of θ\theta and TT. Such a procedure spoils the structure of hydrodynamic equations, indicating strong limitations of the NS approach in spin physics. Due to such issues, we do not perform a numerical analysis of the NS case. We show below that a natural solution to this problem is to use the IS approach.

Entropy production constraints (Israel–Stewart)

In the CS system, the expansion scalar is given by Θ≡∂μUμ=cosh⁡θ/τ+sinh⁡θ/r+∇Rθ\Theta\equiv\partial_{\mu}U^{\mu} = \cosh\theta/\tau+ \sinh\theta/r + \nabla_{R}\theta, with ∇Rθ=sinh⁡θ θ˙+cosh⁡θ θ′\nabla_{R}\theta= \sinh\theta\,\dot{\theta} + \cosh\theta\,\theta'. By projecting Eqs. (20) and (21) onto the basis vectors RμR^{\mu}, Φμ\Phi^{\mu}, and ZμZ^{\mu}, and using the relations DRμ=Dθ UμDR^{\mu}=D\theta\,U^{\mu} and DΦμ=DZμ=0D\Phi^{\mu}=DZ^{\mu}=0 (see App. A), we obtain two differential equations for the vectors d\mathbf{d} and e\mathbf{e} defined in (67):

Dd+d(1τd+Da42a4+Θ2)=1a4(Ckeq−Ck)(83)D\mathbf{d}+\mathbf{d}\left(\frac{1}{\tau_{d}}+\frac{Da_{4}}{2a_{4}}+\frac{\Theta}{2}\right)=\frac{1}{a_{4}}\left(C_{k}^{\mathrm{eq}}-C_{k}\right) \tag*{(83)}

and

De+e(1τe+Da52a5+Θ2)=−Cω2a5.(84)D\mathbf{e}+\mathbf{e}\left(\frac{1}{\tau_{e}}+\frac{Da_{5}}{2a_{5}}+\frac{\Theta}{2}\right)=-\frac{C_{\omega}}{2a_{5}}. \tag*{(84)}

Since 1/(2a4)=λT/τd1/(2a_{4})=\lambda T/\tau_{d} and −1/(2a5)=γT/τe-1/(2a_{5})=\gamma T/\tau_{e}, the right-hand side of the radial component of Eq. (83) equals 2λT(Ckreq−Ckr)/τd2\lambda T(C_{kr}^{\mathrm{eq}}-C_{kr})/\tau_{d}. Hence, in the limit τd→0\tau_{d}\to0 and τe→0\tau_{e}\to0, Eqs. (83) and (84) formally reproduce the Navier–Stokes expressions (80) and (81).

Before we switch to a discussion of our numerical results based on the IS method, let us come back to the issues discussed in the end of the last section. First of all, we notice that similar problems with overdetermined systems of equations occur in the IS approach, which forces us to assume that Ckϕ=Cωz=dϕ=ez=0C_{k\phi}=C_{\omega z}=d_{\phi}=e_{z}=0 and Ckz=Cωϕ=dz=eϕ=0C_{kz}=C_{\omega\phi}=d_{z}=e_{\phi}=0. Thus, below we restrict ourselves to study the dynamics in the radial sector only. Second, we observe that the IS theory gives us the time derivative of drd_{r} that can be directly used in Eqs. (69) and (70). In this way, we naturally avoid problems of the NS method.

Let us turn now to a discussion of our numerical results obtained for the CS model in the IS version. The initial proper time is τ0=1.0\tau_{0}=1.0 fm, and the evolution is followed up to the final time τf=10.0\tau_{f}=10.0 fm. As in Ref. [58], the initial transverse profiles follow the Woods–Saxon distribution ρ(r)=[1+exp⁡((r−RWS)/aWS)]−1\rho(r)=[1+\exp((r-R_{\mathrm{WS}})/a_{\mathrm{WS}})]^{-1} with RWS=7.0R_{\mathrm{WS}}=7.0 fm and aWS=0.54a_{\mathrm{WS}}=0.54 fm, namely

T(τ0,r)=(T0−Tvac)ρ(r)1/3+Tvac,μ(τ0,r)=μ0ρ(r),Cωr(τ0,r)=ρ(r),(85)T(\tau_{0},r)=(T_{0}-T_{\mathrm{vac}})\rho(r)^{1/3}+T_{\mathrm{vac}},\qquad\mu(\tau_{0},r)=\mu_{0}\rho(r),\qquad C_{\omega r}(\tau_{0},r)=\rho(r), \tag*{(85)}

with T0=0.3T_{0}=0.3 GeV, μ0=0.03\mu_{0}=0.03 GeV, and Tvac=0.03T_{\mathrm{vac}}=0.03 GeV. The fluid is initially at rest, θ(τ0,r)=0\theta(\tau_{0},r)=0, and Ckr(τ0,r)=dr(τ0,r)=er(τ0,r)=0C_{kr}(\tau_{0},r)=d_{r}(\tau_{0},r)=e_{r}(\tau_{0},r)=0. As in the TH case, m=1.116m=1.116 GeV, λ=λ0σ\lambda=\lambda_{0}\sigma and γ=γ0σ\gamma=\gamma_{0}\sigma with λ0=γ0=0.02\lambda_{0}=\gamma_{0}=0.02, and τd=τe=1.0\tau_{d}=\tau_{e}=1.0 fm. We note that the electriclike sector (Ckr,dr)(C_{kr},d_{r}) is not coupled to the magneticlike one (Cωr,er)(C_{\omega r},e_{r}); therefore, the results for CkrC_{kr} and drd_{r} depend neither on the initial value of CωrC_{\omega r} nor on the relaxation time τe\tau_{e}.

The time evolution of the temperature and baryon chemical potential is shown in Figure 6. The opaque lines show the results with dissipation, and the semitransparent lines describe our previous results obtained for the perfect spin fluid [58]. The fact that the two types of lines overlap and are, in practice, indistinguishable means that the dissipative terms involving drd_r can be neglected in (69) and (70). Hence, in the considered case, one can separate a spinless hydrodynamic background from the spin dynamics.5

Figure 6. Radial profiles of (a) the temperature TT and (b) the baryon chemical potential μ\mu, normalized to their initial values at the center, T0T_0 and μ0\mu_0. The initial conditions are given by Eq. (85). Different line styles correspond to different values of the proper time τ\tau, as described in the legend. Opaque lines show the results of the dissipative evolution with λ0=γ0=0.02\lambda_0=\gamma_0=0.02 and τd=τe=1.0 fm\tau_d=\tau_e=1.0\ \mathrm{fm}, while semitransparent lines show the evolution without dissipation, λ0=γ0=0\lambda_0=\gamma_0=0. The two types of lines overlap, and no difference between dissipative and perfect-fluid evolutions can be noticed in this case.

Our numerical results for the spin electric radial sector are shown in Figure 7. Panel (a) shows spatial profiles of the coefficient CkrC_{kr} for different evolution times. Panels (b) and (c) show the two contributions to the coefficient CkreqC_{kr}^{\mathrm{eq}}; see Eq. (62): the term βDθ\beta D\theta, connected with the acceleration of the fluid, and the term −β2∇rT-\beta^2\nabla_r T, connected with the radial temperature gradient. Both terms are positive and of a similar magnitude. They are localized in the region of the expansion front, where the fluid is accelerated and the temperature drops, and they move together with the front towards larger values of rr. Panel (d) shows the spatial dependence of the combination

Figure 7. Radial profiles of (a) the radial electriclike spin polarization component CkrC_{kr}, (b) the term βDθ\beta D\theta, (c) the term −β2∇RT-\beta^{2}\nabla_{R}T, and (d) the combination βDθ−β2∇RT−2Ckr\beta D\theta-\beta^{2}\nabla_{R}T-2C_{kr} that appears on the right-hand side of Eq. (83). Different line styles correspond to different values of the proper time τ∈{1.0 fm,4.0 fm,7.0 fm,10.0 fm}\tau\in\{1.0\ \mathrm{fm}, 4.0\ \mathrm{fm}, 7.0\ \mathrm{fm}, 10.0\ \mathrm{fm}\}. The system is initialized with Ckr(τ0,r)=dr(τ0,r)=0C_{kr}(\tau_{0},r)=d_{r}(\tau_{0},r)=0, T0=0.3 GeVT_{0}=0.3\ \mathrm{GeV}, and μ0=0.03 GeV\mu_{0}=0.03\ \mathrm{GeV}; the kinetic coefficient is λ=λ0σ\lambda=\lambda_{0}\sigma with λ0=0.02\lambda_{0}=0.02, and the relaxation time is τd=1.0 fm\tau_{d}=1.0\ \mathrm{fm}.

βDθ−β2∇RT−2Ckr=2(Ckreq−Ckr).(86)\beta D\theta-\beta^{2}\nabla_{R}T-2C_{kr}=2\left(C_{kr}^{\mathrm{eq}}-C_{kr}\right). \tag*{(86)}

At the initial time, Ckr=0C_{kr}=0, and the right-hand side of Eq. (83), shown in panel (d), is determined by thermal vorticity alone. This source generates the electric dissipation drd_r, which in turn builds up the coefficient CkrC_{kr} through the spin equation (73). As a result, the profile of CkrC_{kr} in panel (a) quickly adopts the shape of CkreqC_{kr}^{\mathrm{eq}}, and the combination βDθ−β2∇RT−2Ckr=2(Ckreq−Ckr)\beta D\theta-\beta^{2}\nabla_{R}T-2C_{kr}=2\left(C_{kr}^{\mathrm{eq}}-C_{kr}\right) becomes small compared to its initial value and continues to decrease during the evolution. Thus, CkrC_{kr} does not relax to a constant value but follows the time-dependent thermal vorticity. This behavior agrees with the form of the entropy production (82), which vanishes for Ckr=CkreqC_{kr}=C^{\mathrm{eq}}_{kr}.

Figure 8 shows the time dependence of the same quantities as in Figure 7 but obtained here with a smaller relaxation time value, τd=0.5\tau_d=0.5 fm. Once again, we observe the spin polarization tending toward the value corresponding to thermal vorticity; this process occurs more rapidly due to the smaller value of τd\tau_d assumed.

It is worth noting that the similar behavior of the functions βDθ\beta D\theta and −β2∇RT-\beta^2\nabla_R T, shown in Figure 7 and Figure 8, is not accidental. In the case of the perfect fluid, the hydrodynamic equations have the form ∂μ(Tuν)=∂νT\partial_\mu(Tu^\nu)=\partial^\nu T, which implies an exact equality between these two functions. The case considered here includes a small chemical potential and spin effects, which leads to differences between βDθ\beta D\theta and −β2∇RT-\beta^2\nabla_R T, although they are rather small.

Figure 8. Same as Figure 7, except for the relaxation time τd=0.5\tau_d = 0.5 fm.

Finally, in Figure 9 we show the time evolution of the coefficients ere_r and CωrC_{\omega r}. As expected, both ere_r and CωrC_{\omega r} approach zero. For comparison, we added semitransparent lines showing the perfect spin-fluid results. If dissipation is not taken into account, the value of the coefficient CωrC_{\omega r} increases slowly with time.

Figure 9. Radial profiles of (a) the radial component ere_r of the magnetic dissipation vector and (b) the radial magneticlike spin polarization coefficient CωrC_{\omega r}. The initial conditions are given by Eq. (85) with er(τ0,r)=0e_r(\tau_0,r)=0. Different line styles correspond to different values of the proper time τ\tau, as described in the legend. Opaque lines show the results of the dissipative evolution with γ0=0.02\gamma_0=0.02 and τe=1.0\tau_e=1.0 fm, while semitransparent lines in panel (b) show the evolution without dissipation, λ0=γ0=0\lambda_0=\gamma_0=0, for which er=0e_r=0.

Summary

To summarize our findings, we make the following observations:

  1. Inclusion of the antisymmetric parts of the energy–momentum tensor implies that dissipative dynamics drives the spin polarization tensor ω\omega toward thermal vorticity ϖ\varpi. The speed of this process depends on the values of the spin kinetic coefficients and spin relaxation times.

  1. At the same time, a transfer between the spin and orbital parts of the total angular momentum takes place. This process can be interpreted as a dissipative spin-orbit interaction.

  1. Antisymmetric parts of the energy–momentum tensor appear in the equations of motion (conservation laws for energy and momentum), which, in the cases under consideration, implies that certain components of the spin-polarization tensor must vanish. These constraints become increasingly stringent as the symmetries imposed on the hydrodynamic flow become more restrictive.

  1. The presence of the time derivatives of the antisymmetric parts of the energy–momentum tensor in the equations of motion leads to problems with the application of the Navier–Stokes prescription. They are automatically solved if the Israel–Stewart method is applied.

  1. The experience gained from the analyses of boost-invariant systems considered in this work can be used in more realistic calculations based on spin hydrodynamics.

Acknowledgments

This work was supported in part by the National Science Centre, Poland (NCN) Grant No. 2022/47/B/ST2/01372.

Useful identities for boost-invariant and cylindrically symmetric systems

∂μUμ=cosh⁡θτ+sinh⁡θr+sinh⁡θθ˙+cosh⁡θθ′,∂μRμ=sinh⁡θτ+cosh⁡θr+cosh⁡θθ˙+sinh⁡θθ′,∂μZμ=0,∂μΦμ=0(87)\begin{aligned} \partial_\mu U^\mu&= \frac{\cosh\theta}{\tau}+\frac{\sinh\theta}{r}+\sinh\theta\dot{\theta}+\cosh\theta\theta',\\ \partial_\mu R^\mu&= \frac{\sinh\theta}{\tau}+\frac{\cosh\theta}{r}+\cosh\theta\dot{\theta}+\sinh\theta\theta',\\ \partial_\mu Z^\mu&= 0,\\ \partial_\mu\Phi^\mu&= 0 \tag*{(87)} \end{aligned}
Uμ∂μf(τ,η,r,ϕ)=(cosh⁡θ ∂τ+sinh⁡θ ∂r)f(τ,η,r,ϕ),Rμ∂μf(τ,η,r,ϕ)=(sinh⁡θ ∂τ+cosh⁡θ ∂r)f(τ,η,r,ϕ),Zμ∂μf(τ,η,r,ϕ)=1τ∂ηf(τ,η,r,ϕ),Φμ∂μf(τ,η,r,ϕ)=1r∂ϕf(τ,η,r,ϕ)(88)\begin{aligned} U^\mu\partial_\mu f(\tau,\eta,r,\phi)&=(\cosh\theta\,\partial_\tau+\sinh\theta\,\partial_r)f(\tau,\eta,r,\phi),\\ R^\mu\partial_\mu f(\tau,\eta,r,\phi)&=(\sinh\theta\,\partial_\tau+\cosh\theta\,\partial_r)f(\tau,\eta,r,\phi),\\ Z^\mu\partial_\mu f(\tau,\eta,r,\phi)&=\frac{1}{\tau}\partial_\eta f(\tau,\eta,r,\phi),\\ \Phi^\mu\partial_\mu f(\tau,\eta,r,\phi)&=\frac{1}{r}\partial_\phi f(\tau,\eta,r,\phi) \tag*{(88)} \end{aligned}
Uμ∂μUν=(cosh⁡θθ˙+sinh⁡θθ′)Rν,Uμ∂μRν=(cosh⁡θθ˙+sinh⁡θθ′)Uν,Uμ∂μZν=0,Uμ∂μΦν=0(89)\begin{aligned} U^\mu\partial_\mu U^\nu&=(\cosh\theta\dot{\theta}+\sinh\theta\theta')R^\nu,\\ U^\mu\partial_\mu R^\nu&=(\cosh\theta\dot{\theta}+\sinh\theta\theta')U^\nu,\\ U^\mu\partial_\mu Z^\nu&=0,\\ U^\mu\partial_\mu\Phi^\nu&=0 \tag*{(89)} \end{aligned}
Rμ∂μUν=(sinh⁡θθ˙+cosh⁡θθ′)Rν,Rμ∂μRν=(sinh⁡θθ˙+cosh⁡θθ′)Uν,Rμ∂μZν=0,Rμ∂μΦν=0(90)\begin{aligned} R^\mu\partial_\mu U^\nu&=(\sinh\theta\dot{\theta}+\cosh\theta\theta')R^\nu,\\ R^\mu\partial_\mu R^\nu&=(\sinh\theta\dot{\theta}+\cosh\theta\theta')U^\nu,\\ R^\mu\partial_\mu Z^\nu&=0,\\ R^\mu\partial_\mu\Phi^\nu&=0 \tag*{(90)} \end{aligned}
Zμ∂μUν=cosh⁡θτZν,Zμ∂μRν=sinh⁡θτZν,Zμ∂μΦν=0(91)\begin{aligned} Z^\mu\partial_\mu U^\nu&=\frac{\cosh\theta}{\tau}Z^\nu,\\ Z^\mu\partial_\mu R^\nu&=\frac{\sinh\theta}{\tau}Z^\nu,\\ Z^\mu\partial_\mu\Phi^\nu&=0 \tag*{(91)} \end{aligned}
Φμ∂μUν=sinh⁡θrΦν,Φμ∂μRν=cosh⁡θrΦν,Φμ∂μZν=0(92)\begin{aligned} \Phi^\mu\partial_\mu U^\nu&=\frac{\sinh\theta}{r}\Phi^\nu,\\ \Phi^\mu\partial_\mu R^\nu&=\frac{\cosh\theta}{r}\Phi^\nu,\\ \Phi^\mu\partial_\mu Z^\nu&=0 \tag*{(92)} \end{aligned}
∇μ=Δμν∂ν=−Rμ(sinh⁡θ ∂τ+cosh⁡θ ∂r)−Zμ1τ∂η−Φμ1r∂ϕ(93)\nabla^\mu= \Delta^{\mu\nu}\partial_\nu= -R^\mu\left(\sinh\theta\,\partial_\tau+\cosh\theta\,\partial_r\right)-Z^\mu\frac{1}{\tau}\partial_\eta-\Phi^\mu\frac{1}{r}\partial_\phi \tag*{(93)}
∇μT(τ,r)=−(sinh⁡θ T˙+cosh⁡θ T′)Rμ(94)\nabla^\mu T(\tau,r)=-\left(\sinh\theta\,\dot{T}+\cosh\theta\,T'\right)R^\mu \tag*{(94)}
aμ=Uα∂αUμ=(cosh⁡θ θ˙+sinh⁡θ θ′)Rμ(95)a^\mu=U^\alpha\partial_\alpha U^\mu=\left(\cosh\theta\,\dot{\theta}+\sinh\theta\,\theta'\right)R^\mu \tag*{(95)}

References

References

  1. [1]Z.-T. Liang and X.-N. Wang, Phys. Rev. Lett. 94, 102301 (2005), [Erratum: Phys.Rev.Lett. 96, 039901 (2006)], arXiv:nucl-th/0410079.DOI
  2. [2]Z.-T. Liang and X.-N. Wang, Phys. Lett. B 629, 20 (2005), arXiv:nucl-th/0411101.DOI
  3. [3]L. Adamczyk et al. (STAR), Nature 548, 62 (2017), arXiv:1701.06657 [nucl-ex].
  4. [4]J. Adam et al. (STAR), Phys. Rev. C 98, 014910 (2018), arXiv:1805.04400 [nucl-ex].DOI
  5. [5]J. Adam et al. (STAR), Phys. Rev. Lett. 123, 132301 (2019), arXiv:1905.11917 [nucl-ex].DOI
  6. [6]S. Acharya et al. (ALICE), Phys. Rev. Lett. 125, 012301 (2020), arXiv:1910.14408 [nucl-ex].
  7. [7]T. Niida and S. A. Voloshin, Int. J. Mod. Phys. E 33, 2430010 (2024), arXiv:2404.11042 [nucl-ex].DOI
  8. [8]F. Becattini and L. Tinti, Annals Phys. 325, 1566 (2010), arXiv:0911.0864 [gr-qc].DOI
  9. [9]F. Becattini, Phys. Part. Nucl. Lett. 8, 801 (2011).
  10. [10]D. Montenegro, L. Tinti, and G. Torrieri, Phys. Rev. D 96, 056012 (2017), [Addendum: Phys.Rev.D 96, 079901 (2017)], arXiv:1701.08263 [hep-th].DOI
  11. [11]W. Florkowski, B. Friman, A. Jaiswal, and E. Speranza, Phys. Rev. C97, 041901 (2018), arXiv:1705.00587 [nucl-th].DOI
  12. [12]W. Florkowski, B. Friman, A. Jaiswal, R. Ryblewski, and E. Speranza, Phys. Rev. D 97, 116017 (2018), arXiv:1712.07676 [nucl-th].DOI
  13. [13]W. Florkowski, A. Kumar, and R. Ryblewski, Phys. Rev. C 98, 044906 (2018), arXiv:1806.02616 [hep-ph].DOI
  14. [14]K. Hattori, M. Hongo, X.-G. Huang, M. Matsuo, and H. Taya, Phys. Lett. B 795, 100 (2019), arXiv:1901.06615 [hep-th].DOI
  15. [15]N. Weickgenannt, X.-L. Sheng, E. Speranza, Q. Wang, and D. H. Rischke, Phys. Rev. D 100, 056018 (2019), arXiv:1902.06513 [hep-ph].DOI
  16. [16]W. Florkowski, A. Kumar, R. Ryblewski, and R. Singh, Phys. Rev. C 99, 044910 (2019), arXiv:1901.09655 [hep-ph].DOI
  17. [17]S. Bhadury, W. Florkowski, A. Jaiswal, A. Kumar, and R. Ryblewski, Phys. Lett. B 814, 136096 (2021), arXiv:2002.03937 [hep-ph].DOI
  18. [18]D. Montenegro and G. Torrieri, Phys. Rev. D 102, 036007 (2020), arXiv:2004.10195 [hep-th].DOI
  19. [19]N. Weickgenannt, E. Speranza, X.-L. Sheng, Q. Wang, and D. H. Rischke, Phys. Rev. Lett. 127, 052301 (2021), arXiv:2005.01506 [hep-ph].DOI
  20. [20]S. Shi, C. Gale, and S. Jeon, Phys. Rev. C 103, 044906 (2021), arXiv:2008.08618 [nucl-th].DOI
  21. [21]S. Bhadury, W. Florkowski, A. Jaiswal, A. Kumar, and R. Ryblewski, Phys. Rev. D 103, 014030 (2021), arXiv:2008.10976 [nucl-th].DOI
  22. [22]K. Fukushima and S. Pu, Phys. Lett. B 817, 136346 (2021), arXiv:2010.01608 [hep-th].DOI
  23. [23]S. Li, M. A. Stephanov, and H.-U. Yee, Phys. Rev. Lett. 127, 082302 (2021), arXiv:2011.12318 [hep-th].DOI
  24. [24]R. Singh, G. Sophys, and R. Ryblewski, Phys. Rev. D 103, 074024 (2021), arXiv:2011.14907 [hep-ph].DOI
  25. [25]A. D. Gallegos, U. Gürsoy, and A. Yarom, SciPost Phys. 11, 041 (2021), arXiv:2101.04759 [hep-th].DOI
  26. [26]N. Weickgennant, E. Speranza, X.-I. Sheng, Q. Wang, and D. H. Rischke, Phys. Rev. D 104, 016022 (2021), arXiv:2103.04896 [nucl-th].
  27. [27]D. She, A. Huang, D. Hou, and J. Liao, Sci. Bull. 67, 2265 (2022), arXiv:2105.04060 [nucl-th].DOI
  28. [28]M. Hongo, X.-G. Huang, M. Kaminski, M. Stephanov, and H.-U. Yee, JHEP 11, 150, arXiv:2107.14231 [hep-th].DOI
  29. [29]J. Hu, Phys. Rev. D 105, 076009 (2022), arXiv:2111.03571 [hep-ph].DOI
  30. [30]R. Singh, M. Shokri, and S. M. A. T. Mehr, Nucl. Phys. A 1035, 122656 (2023), arXiv:2202.11504 [hep-ph].DOI
  31. [31]N. Weickgennant, D. Wagner, E. Speranza, and D. H. Rischke, Phys. Rev. D 106, 096014 (2022), arXiv:2203.04766 [nucl-th].
  32. [32]D. Wagner, N. Weickgennant, and D. H. Rischke, Phys. Rev. D 106, 116021 (2022), arXiv:2210.06187 [nucl-th].DOI
  33. [33]S. Dey, W. Florkowski, A. Jaiswal, and R. Ryblewski, Phys. Lett. B 843, 137994 (2023), arXiv:2303.05271 [hep-th].DOI
  34. [34]N. Weickgennant and J.-P. Blaizot, Phys. Rev. D 109, 056012 (2024), arXiv:2311.15817 [hep-ph].
  35. [35]A. Kumar, D.-L. Yang, and P. Gubler, Phys. Rev. D 109, 054038 (2024), arXiv:2312.16900 [nucl-th].DOI
  36. [36]D. Wagner, M. Shokri, and D. H. Rischke, Phys. Rev. Res. 6, 043103 (2024), arXiv:2405.00533 [nucl-th].DOI
  37. [37]D. Wagner, Phys. Rev. D 111, 016008 (2025), arXiv:2409.07143 [nucl-th].DOI
  38. [38]S. Dey and A. Das, Phys. Rev. D 111, 074037 (2025), arXiv:2410.04141 [nucl-th].arxiv.org/abs/2410.04141
  39. [39]D. She, Y.-W. Qiu, and D. Hou, Phys. Rev. D 111, 036027 (2025), arXiv:2410.15142 [nucl-th].DOI
  40. [40]X.-G. Huang, Nucl. Sci. Tech. 36, 208 (2025), arXiv:2411.11753 [nucl-th].DOI
  41. [41]S. Bhadury, Phys. Rev. C 112, L021901 (2025), arXiv:2503.08428 [hep-ph].arxiv.org/abs/2503.08428
  42. [42]S. Dey, Phys. Rev. D 114, 054043 (2026), arXiv:2504.18388 [hep-th].arxiv.org/abs/2504.18388
  43. [43]A. Daher, D. Wagner, M. Shokri, and D. H. Rischke, (2026), arXiv:2609.31819 [nucl-th].arxiv.org/abs/2609.31819
  44. [44]J.-Y. Ollitrault, Eur. J. Phys. 29, 275 (2008), arXiv:0708.2433 [nucl-th].DOI
  45. [45]P. Romatschke, Int. J. Mod. Phys. E 19, 1 (2010), arXiv:0902.3663 [hep-ph].DOI
  46. [46]W. Florkowski, Phenomenology of Ultra-Relativistic Heavy-Ion Collisions (Singapore: World Scientific, 2010).DOI
  47. [47]C. Gale, S. Jeon, and B. Schenke, Int. J. Mod. Phys. A 28, 1340011 (2013), arXiv:1301.5893 [nucl-th].DOI
  48. [48]A. Jaiswal and V. Roy, Adv. High Energy Phys. 2016, 9623034 (2016), arXiv:1605.08694 [nucl-th].DOI
  49. [49]W. Florkowski, J. Subatom. Part. Cosmol. 3, 100028 (2025), arXiv:2411.19673 [hep-ph].arxiv.org/abs/2411.19673
  50. [50]S. Bhadury, Z. Drogosz, W. Florkowski, and V. Mykhaylova, J. Subatom. Part. Cosmol. 4, 100218 (2025), arXiv:2507.01864 [hep-ph].DOI
  51. [51]Z. Drogosz, W. Florkowski, and M. Hontarenko, Phys. Rev. D 110, 096018 (2024), arXiv:2408.03106 [hep-ph].DOI
  52. [52]R. Biswas, A. Daher, A. Das, W. Florkowski, and R. Ryblewski, Phys. Rev. D 108, 014024 (2023), arXiv:2304.01009 [nucl-th].DOI
  53. [53]W. Florkowski, R. Ryblewski, R. Singh, and G. Sophys, Phys. Rev. D 105, 054007 (2022), arXiv:2112.01856 [hep-ph].DOI
  54. [54]D.-L. Wang, X.-Q. Xie, S. Fang, and S. Pu, Phys. Rev. D 105, 114050 (2022), arXiv:2112.15535 [hep-ph].DOI
  55. [55]R. Singh, M. Shokri, and R. Ryblewski, Phys. Rev. D 103, 094034 (2021), arXiv:2103.02592 [hep-ph].DOI
  56. [56]D.-L. Wang, S. Fang, and S. Pu, Phys. Rev. D 104, 114043 (2021), arXiv:2107.11726 [nucl-th].DOI
  57. [57]Z. Drogosz, W. Florkowski, N. Łygan, and R. Ryblewski, Phys. Rev. C 111, 024909 (2025), arXiv:2411.06154 [hep-ph].DOI
  58. [58]Z. Drogosz, W. Florkowski, and J. Witkowski, Phys. Rev. C 114, 024910 (2026), arXiv:2605.01857 [hep-ph].arxiv.org/abs/2605.01857
  59. [59]S. Singh, S. Dey, A. Das, H. Mishra, and A. Jaiswal, Phys. Rev. D 114, 054023 (2026), arXiv:2604.04533 [nucl-th].DOI
  60. [60]Z. Drogosz and N. Łygan, Acta Phys. Polon. B 57, 9 (2026), arXiv:2604.17392 [hep-ph].arxiv.org/abs/2604.17392
  61. [61]W. Israel and J. M. Stewart, Phys. Lett. A 58, 213 (1976).
  62. [62]W. Israel and J. M. Stewart, Annals Phys. 118, 341 (1979).
  63. [63]J. D. Bjorken, Phys. Rev. D 27, 140 (1983).
  64. [64]W. Florkowski and R. Ryblewski, Phys. Rev. C 85, 044902 (2012), arXiv:1111.5997 [nucl-th].DOI

Paper details

vlnc:W_4t7bza7j
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