betaFeedback

WorksPhysics and Astronomy

Late-time attractors in relativistic spin hydrodynamics in Gubser flow

Abstract

We investigate the late-time asymptotic solutions and attractor structure of the spin density in minimal causal spin hydrodynamics in Gubser flow. After deriving the differential equation governing the spin density, we obtain its late-time asymptotic solutions and identify both attractors and repellers in the corresponding numerical solutions. We then map these solutions back to flat Minkowski space and find parameter regions where the spin density exhibits a power-law decay. We further show that, when the characteristic length scale of the system is much larger than the proper time, several components of the spin density can decay as slowly as conventional thermodynamic variables in relativistic hydrodynamics. In this regime, the spin density behaves as a hydrodynamic mode governed by the late-time scaling laws of the flow.

Introduction

In noncentral relativistic heavy-ion collisions, the colliding nuclei carry a large initial orbital angular momentum. Part of this angular momentum can be transferred to the spin degrees of freedom of produced hadrons, leading to spin polarization of hyperons and spin alignment of vector mesons through spin–orbit coupling [1, 2]. The RHIC-STAR Collaboration has measured the global polarization, i.e., the polarization along the direction of the initial orbital angular momentum, as a function of collision energy, for Λ\Lambda and Λˉ\bar{\Lambda}, and other hyperons [3, 4, 5, 6, 7]. At high and intermediate collision energies, the observed global polarization is primarily driven by thermal vorticity and can be described by a variety of phenomenological models [8, 9, 10, 11, 12, 13, 14, 15]. The global polarization in low-energy collisions [6, 7, 16], in particular near and below the production threshold of the Λ\Lambda hyperon, has also been investigated [12, 17, 18, 19, 20, 21, 22, 23], but a definitive understanding has not yet been reached.

On the other hand, the local polarization of hyperons, namely the polarization along the beam direction and the out-of-plane direction as a function of azimuthal angle or transverse momentum, has not been successfully described for a long time by the same models that reproduce the global polarization. To address this discrepancy, a hydrodynamic effect referred to as shear-induced polarization [24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38] has been identified as playing an important role in the local polarization. In addition, polarization induced by gradients of the baryon chemical potential (often discussed in terms of the spin Hall effect [28, 37, 39, 40] and anomalous spin Hall effect [41, 42]), may become relevant at low collision energies. Partonic-interaction corrections have also been investigated systematically in recent studies, e.g. see Refs. [41, 43, 44] and the references therein. Nevertheless, achieving a quantitative description of local polarization at low energies remains challenging [29, 36, 37, 45]. Recently, the CMS measurement of local polarization in p+Pb collisions [46] has also been reported to be incompatible with standard modeling assumptions [45, 47, 48]. For further details, we refer to recent reviews [49, 50, 51, 52] and the references therein.

To get a better understanding of spin polarization in relativistic heavy-ion collisions, it is essential to investigate the dynamical evolution of spin degrees of freedom in the medium produced in the collisions. One approach is quantum kinetic theory [28, 41, 43, 44, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67], which provides a microscopic effective description of the evolution of quasi-particles with spin. Another approach is to formulate macroscopic evolution equations for spin, namely relativistic spin hydrodynamics [68, 69, 70, 71, 72]. This motivates extending conventional relativistic hydrodynamics by incorporating spin as an additional hydrodynamic degree of freedom. Such formulations have been developed using quantum statistical field theory [73, 74, 75, 76, 77], quantum kinetic theory [78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93], analyses based on the entropy principle [71, 72, 94, 95, 96, 97, 98, 99, 100, 101], and the effective field-theory approach [68, 102]. Very recently, using spin hydrodynamics, an extended Bargmann-Michel-Telegdi equation that incorporates rotation and dissipative effects has been derived [100]. For further discussion of spin hydrodynamics, we refer to recent reviews [70, 103, 104] and the references therein.

Although the basic framework of relativistic spin hydrodynamics has been established, several open issues remain. One issue concerns causality and stability. It has been found that first-order spin hydrodynamics in a gradient expansion, formulated in the canonical form, is generically acausal and unstable [105, 106, 107, 108]. Moreover, even after a minimal extension to second order, spin hydrodynamics can still exhibit unstable modes in a linear-mode analysis, even when the asymptotic causality conditions are satisfied [105]1. To address this issue, Ref. [117] applied the thermodynamic stability method [118] and derived causality and stability conditions. A second issue concerns the choice of pseudo-gauge. In addition to the canonical form [71, 98], the Belinfante form [72] and other pseudo-gauge choices [70, 81, 85, 119] can also be used in constructing spin hydrodynamics. Although there has been recent progress in this direction [95, 120, 121, 122], the problem remains unsettled and should be addressed systematically in future work. A third issue is that several recent studies [90, 123, 124] have reported that the thermodynamic relations commonly employed in spin hydrodynamics at finite spin density and spin chemical potential differ from those obtained in microscopic approaches. By contrast, other works have found consistency between spin-hydrodynamic thermodynamics and microscopic results by carefully accounting for rotational effects [125] or by introducing an appropriate pseudo-gauge transformation [126].

In addition, there is another long-standing issue in spin hydrodynamics, especially in the canonical formulation. In canonical spin hydrodynamics, the spin density is not a conserved quantity [71], in contrast to the conserved charge (number) density. It is commonly expected that the spin density relaxes more rapidly than other macroscopic hydrodynamic fields, which would make its effects difficult to observe on the freeze-out hypersurface. This expectation is supported, for example, by analytic solutions of canonical spin hydrodynamics in Bjorken flow [127] and Gubser flow [128]. Very recently, a work by some of us [129] has shown the existence of late-proper-time attractors [130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144] in minimal causal relativistic spin hydrodynamics within Bjorken flow. In particular, nontrivial attractor solutions were found, accompanied by a parametrically slow decay of the spin density. This suggests that the spin density can persist to late proper times and may still influence observables on the freeze-out hypersurface. On the other hand, Bjorken flow assumes boost invariance and neglects the transverse expansion of the system. A more realistic framework with radial expansion is therefore desirable. In this respect, Gubser flow [145, 146], which incorporates transverse expansion while preserving boost invariance along the longitudinal direction, provides a natural extension.

This motivates the present work. Here we investigate the possible existence of attractor behavior for the spin density in canonical spin hydrodynamics under Gubser flow. We analyze the late-time asymptotic solutions for the spin density and identify the corresponding attractor and repeller branches for different parameter choices. Finally, we determine the conditions under which the spin density exhibits a power-law behavior at late times.

This paper is organized as follows. In § 2, we briefly review the minimal causal spin

discussion in Refs. [109, 110, 111]. Also see Refs. [112, 113, 114, 115, 116] for recent discussions of improved causality and stability criteria in linear-mode analyses for relativistic many-body theories. hydrodynamics in Gubser flow and derive the differential equation governing the spin density. In Sec. III, we derive the asymptotic late-time solutions for the spin density for different parameter choices. In Sec. IV, we present numerical results for the attractor and repeller branches. In § 2.4, we discuss the late-time power-law decay of the spin density. Finally, we summarize in Sec. VI.

We adopt the metric gμν=diag⁡{+,−,−,−}g_{\mu\nu} = \operatorname{diag}\{+,-,-,-\} in a flat Minkowski spacetime with coordinates (t,x,y,z)(t,x,y,z) and the corresponding projector is defined as Δμν=gμν−uμuν\Delta_{\mu\nu} = g_{\mu\nu} - u_{\mu}u_{\nu} with uμu^{\mu} being the fluid velocity. For an arbitrary tensor AμνA^{\mu\nu}, we introduce the following convention for simplicity

A(μν)=12(Aμν+Aνμ),A[μν]=12(Aμν−Aνμ),andA⟨μν⟩=12(ΔμαΔνβ+ΔναΔμβ)Aαβ−13Δμν(AρσΔρσ).A^{(\mu\nu)} = \frac{1}{2}(A^{\mu\nu} + A^{\nu\mu}), \quad A^{[\mu\nu]} = \frac{1}{2}(A^{\mu\nu} - A^{\nu\mu}), \quad\text{and} \quad A^{\langle\mu\nu\rangle} = \frac{1}{2}(\Delta^{\mu\alpha}\Delta^{\nu\beta} + \Delta^{\nu\alpha}\Delta^{\mu\beta})A_{\alpha\beta} - \frac{1}{3}\Delta^{\mu\nu}(A^{\rho\sigma}\Delta_{\rho\sigma}).

Throughout this work, we use standard notation for macroscopic quantities defined in flat Minkowski spacetime R3,1\mathbb{R}^{3,1}. For quantities defined in the dS3×RdS_{3} \times\mathbb{R} spacetime after the Weyl transformation, we add a hat. For example, uμu^{\mu} and u^μ\hat{u}^{\mu} denote the fluid velocity in R3,1\mathbb{R}^{3,1} and in dS3×RdS_{3} \times\mathbb{R}, respectively.

Relativistic spin hydrodynamics with Gubser expansion

We first briefly review the main equations of relativistic spin hydrodynamics in § 2 A. We then review the standard Gubser-flow setup in § 2 B and revisit spin hydrodynamics in dS3×RdS_{3} \times\mathbb{R} in § 2 C.

Brief review on relativistic spin hydrodynamics in canonical form

We first review relativistic spin hydrodynamics in relativistic heavy-ion collisions. The main conservation laws are those of the energy-momentum tensor Θμν\Theta^{\mu\nu} and the total angular momentum tensor JλμνJ^{\lambda\mu\nu},

∇μΘμν=0,\nabla_{\mu}\Theta^{\mu\nu} = 0,
∇λJλμν=0.(1)\nabla_{\lambda}J^{\lambda\mu\nu} = 0. \tag*{(1)}

For simplicity, throughout this paper, we neglect the conservation equations for the conserved currents. In this work, we follow [71, 72] and adopt the relativistic spin hydrodynamics in canonical pesudo-gauge, in which the constitutive relations for Θμν\Theta^{\mu\nu} and JλμνJ^{\lambda\mu\nu} take the form

Θμν=euμuν−(p+Π)Δμν+2h(μuν)+πμν+2q[μuν]+ϕμν,\Theta^{\mu\nu} = e u^{\mu}u^{\nu} - (p+\Pi)\Delta^{\mu\nu} + 2h^{(\mu}u^{\nu)} + \pi^{\mu\nu} + 2q^{[\mu}u^{\nu]} + \phi^{\mu\nu},
Jλμν=xμΘλν−xνΘλμ+Σλμν.(2)J^{\lambda\mu\nu} = x^{\mu}\Theta^{\lambda\nu} - x^{\nu}\Theta^{\lambda\mu} + \Sigma^{\lambda\mu\nu}. \tag*{(2)}

where ee and pp are the energy density and thermodynamic pressure, respectively, uμu^\mu is the fluid four-velocity, and Σλμν\Sigma^{\lambda\mu\nu} is the rank-3 spin current tensor. Here, Π\Pi, πμν\pi^{\mu\nu}, and hμh^\mu denote the bulk viscous pressure, shear viscous tensor, and heat flux, respectively. For simplicity, we set the hμh^\mu, Π\Pi and πμν\pi^{\mu\nu} to vanish in this work. The vector qμq^\mu and the antisymmetric tensor ϕμν=−ϕνμ\phi^{\mu\nu}=-\phi^{\nu\mu} encode spin-related contributions. These quantities can be extracted from Θμν\Theta^{\mu\nu} as

hμ=uνΘ(μν),πμν=Θ⟨μν⟩,qμ=uνΘ[μν],ϕμν=ΔμαΔνβΘ[αβ].(3)\begin{aligned} h^\mu&= u_\nu\Theta^{(\mu\nu)}, \qquad\pi^{\mu\nu}=\Theta^{\langle\mu\nu\rangle},\\ q^\mu&= u_\nu\Theta^{[\mu\nu]}, \qquad\phi^{\mu\nu}=\Delta^{\mu\alpha}\Delta^{\nu\beta}\Theta_{[\alpha\beta]}. \tag*{(3)} \end{aligned}

Substituting Eq. (2) into Eq. (1) yields

∇λΣλμν=−2Θ[μν]=−4q[μuν]−2ϕμν.(4)\nabla_\lambda\Sigma^{\lambda\mu\nu}=-2\Theta^{[\mu\nu]}=-4q^{[\mu}u^{\nu]}-2\phi^{\mu\nu}. \tag*{(4)}

This evolution equation shows that the anti-symmetric part of the energy-momentum tensor determines how spin evolves. We further decompose the spin current tensor Σλμν\Sigma^{\lambda\mu\nu} in analogy with the particle current,

Σλμν=uλSμν+Σ(1)λμν.(5)\Sigma^{\lambda\mu\nu}=u^\lambda S^{\mu\nu}+\Sigma_{(1)}^{\lambda\mu\nu}. \tag*{(5)}

Here we have introduced the spin density tensor Sμν=−SνμS^{\mu\nu}=-S^{\nu\mu} and the residual contribution Σ(1)λμν\Sigma_{(1)}^{\lambda\mu\nu}, which is orthogonal to uμu^\mu. Note that, in this work, Σλμν\Sigma^{\lambda\mu\nu} carries six independent degrees of freedom. One may alternatively consider a totally antisymmetric rank-3 spin tensor; see, e.g., [98, 100].

In analogy with the charge chemical potential, which is conjugate to the particle density, one can introduce a spin chemical potential ωμν=−ωνμ\omega_{\mu\nu}=-\omega_{\nu\mu} conjugate to the spin density SμνS^{\mu\nu}. The thermodynamic relations of spin hydrodynamics then read [71, 72],

e+p=Ts+ωμνSμν,de=Tds+ωμνdSμν,dp=sdT+Sμνdωμν.(6)\begin{aligned} e+p &= Ts+\omega_{\mu\nu}S^{\mu\nu},\\ de &= Tds+\omega_{\mu\nu}dS^{\mu\nu},\\ dp &= sdT+S^{\mu\nu}d\omega_{\mu\nu}. \tag*{(6)} \end{aligned}

where TT and ss denote the temperature and entropy density, respectively. Recently, some works [90, 123, 124] have reported that the thermodynamic relations derived from microscopic theory do not coincide with Eqs. (6). On the other hand, other studies [125, 126] have found agreement between kinetic-theory results and Eqs. (6). The appropriate form of the thermodynamic relations in spin hydrodynamics therefore remains an open question. For simplicity, we adopt the conventional thermodynamic relations in this work.

Next, we introduce the power counting used in this work. We consider a gradient expansion in spin hydrodynamics and follow Ref. [72] by taking ωμν∼O(∂)\omega^{\mu\nu} \sim\mathcal{O}(\partial) and Sμν∼O(1)S^{\mu\nu} \sim\mathcal{O}(1). It was shown in Refs. [106, 107] that first-order spin hydrodynamics generally violates causality and stability. We therefore employ the minimal causal theory proposed in Ref. [105]. Here, minimal means that we retain only the essential terms in the second-order gradient expansion required to ensure causality. In this setup, the constitutive equations for qμq^\mu and ϕμν\phi^{\mu\nu} read,

τqΔμνuα∇αqν+qμ=λ(T−1Δμα∇αT+uα∇αuμ−4ωμνuν),\tau_q \Delta^{\mu\nu}u^\alpha\nabla_\alpha q_\nu+ q^\mu= \lambda\left(T^{-1}\Delta^{\mu\alpha}\nabla_\alpha T + u^\alpha\nabla_\alpha u^\mu- 4\omega^{\mu\nu}u_\nu\right),
τϕΔμαΔνβuρ∇ρϕαβ+ϕμν=2γsΔμαΔνβ(∇[αuβ]+2ωαβ),(7)\tau_\phi\Delta^{\mu\alpha}\Delta^{\nu\beta}u^\rho\nabla_\rho\phi_{\alpha\beta} + \phi^{\mu\nu} = 2\gamma_s\Delta^{\mu\alpha}\Delta^{\nu\beta}\left(\nabla_{[\alpha}u_{\beta]} + 2\omega_{\alpha\beta}\right), \tag*{(7)}

where λ\lambda and γs\gamma_s are two positive-definite transport coefficients, and τq,τϕ>0\tau_q,\tau_\phi> 0 are the relaxation times for qμq^\mu and ϕμν\phi^{\mu\nu}, respectively. When these transport coefficients satisfy the relations given in Ref. [117], the theory is causal and linearly stable. For second-order spin hydrodynamics with additional dissipative terms, we refer to Refs. [99, 108]. Later, we will apply a Weyl transformation to Eq. (7) and show that the constitutive equation for ϕμν\phi^{\mu\nu} breaks conformal symmetry. To restore conformal covariance, we introduce an additional term and write

τϕΔ ρμΔ βνuρ∇ρϕαβ+λθϕμν∇αuα+ϕμν=2γsΔμαΔνβ(∇[αuβ]+2ωαβ),(8)\tau_\phi\Delta^\mu_{\ \rho}\Delta^\nu_{\ \beta}u^\rho\nabla_\rho\phi^{\alpha\beta} + \lambda_\theta\phi^{\mu\nu}\nabla_\alpha u^\alpha+ \phi^{\mu\nu} = 2\gamma_s\Delta^{\mu\alpha}\Delta^{\nu\beta}\left(\nabla_{[\alpha}u_{\beta]} + 2\omega_{\alpha\beta}\right), \tag*{(8)}

where λθ\lambda_\theta is a transport coefficient that controls the strength of the coupling between ϕμν\phi^{\mu\nu} and the expansion rate ∇⋅u\nabla\cdot u. Such second-order coupling terms are generally allowed.

Coordinates for Gubser flow

In this section, we briefly review the conventional Gubser-flow setup introduced in Ref. [145, 146]. The Gubser flow is constructed to be invariant under the symmetry group SO(3)×SO(1,1)×Z2SO(3)\times SO(1,1)\times Z_2. Here, the SO(1,1)SO(1,1) factor corresponds to boost invariance, while the Z2Z_2 factor corresponds to reflection symmetry across the collision plane. The SO(3)SO(3) factor, in turn, encodes a specific combination of transverse translations and rotations in the transverse plane. Because the fluid four-velocity is tightly constrained by these symmetries, it is nontrivial to construct a velocity profile that realizes them directly in Minkowski spacetime R3,1\mathbb{R}^{3,1}. Exploiting conformal invariance, the dynamics are invariant under a Weyl rescaling of the metric, gμν→Λ−2gμνg_{\mu\nu} \to\Lambda^{-2}g_{\mu\nu}, where Λ\Lambda may vary across spacetime. This allows one to realize the Gubser symmetry on a conformally related manifold and then map the solution back to R3,1\mathbb{R}^{3,1}. To make this concrete, we start from the

Minkowski metric on R3,1\mathbb{R}^{3,1},

ds2=dt2−dx2−dy2−dz2=dτ2−dr2−r2dφ2−τ2dη2.(9)\begin{aligned} ds^2 = dt^2 - dx^2 - dy^2 - dz^2 \\ &= d\tau^2 - dr^2 - r^2d\varphi^2 - \tau^2d\eta^2. \tag*{(9)} \end{aligned}

where τ\tau is the proper time, η\eta is the spacetime rapidity, rr is the transverse radius, and φ\varphi is the azimuthal angle. These coordinates are defined as

τ=t2−z2,η=12ln⁡(t+zt−z),\tau= \sqrt{t^2-z^2}, \qquad\eta= \frac{1}{2}\ln\left(\frac{t+z}{t-z}\right),
r=x2+y2,φ=arctan⁡(yx).(10)r = \sqrt{x^2+y^2}, \qquad\varphi= \arctan\left(\frac{y}{x}\right). \tag*{(10)}

We then perform a Weyl rescaling to map the fluid dynamics from R3,1\mathbb{R}^{3,1} to the conformally related spacetime dS3×RdS_3 \times\mathbb{R}. The rescaled line element ds^2d\hat{s}^2 can be brought into the Gubser coordinates (ρ,θ,φ,η)(\rho,\theta,\varphi,\eta) as

ds^2=1τ2(dτ2−dr2−r2dφ2)−dη2=dρ2−cosh⁡2ρ(dθ2+sin⁡2θdφ2)−dη2.(11)\begin{aligned} d\hat{s}^2 = \frac{1}{\tau^2}(d\tau^2-dr^2-r^2d\varphi^2)-d\eta^2 \\ &= d\rho^2-\cosh^2\rho(d\theta^2+\sin^2\theta d\varphi^2)-d\eta^2. \tag*{(11)} \end{aligned}

The coordinate transformation is defined by

sinh⁡ρ=−L2−τ2+r22Lτ,\sinh\rho= -\frac{L^2-\tau^2+r^2}{2L\tau},
tan⁡θ=2LrL2+τ2−r2.(12)\tan\theta= \frac{2Lr}{L^2+\tau^2-r^2}. \tag*{(12)}

Here LL is a parameter with dimensions of length that sets the characteristic transverse size of the system. To satisfy the Gubser symmetry, the velocity of the fluid in the coordinate (ρ,θ,φ,η)(\rho,\theta,\varphi,\eta) is found to be,

u^μ=(1,0,0,0).(13)\hat{u}^{\mu} = (1,0,0,0). \tag*{(13)}

According to the Weyl rescaling in Eq. (11), the metric gμνg_{\mu\nu} in Minkowski spacetime and the metric g^μν\hat{g}_{\mu\nu} in dS3×RdS_3 \times\mathbb{R} are related by

g^μν=1τ2gμν,g^μν=τ2gμν.(14)\hat{g}_{\mu\nu} = \frac{1}{\tau^2}g_{\mu\nu}, \qquad\hat{g}^{\mu\nu} = \tau^2 g^{\mu\nu}. \tag*{(14)}

The Christoffel symbols transform as

Γμνλ=Γ^μνλ+τ−1(δνλ∇^μτ+δμλ∇^ντ−g^μνg^λα∇^ατ).(15)\Gamma^{\lambda}_{\mu\nu} = \hat{\Gamma}^{\lambda}_{\mu\nu}+\tau^{-1}\left(\delta^{\lambda}_{\nu}\hat{\nabla}_{\mu}\tau+\delta^{\lambda}_{\mu}\hat{\nabla}_{\nu}\tau-\hat{g}_{\mu\nu}\hat{g}^{\lambda\alpha}\hat{\nabla}_{\alpha}\tau\right). \tag*{(15)}

The general relation between a physical tensor Aν1ν2⋯νnμ1μ2⋯μm(x)A^{\mu_1\mu_2\cdots\mu_m}_{\nu_1\nu_2\cdots\nu_n}(x) in R3,1\mathbb{R}^{3,1} and its counterpart A^ν1ν2⋯νnμ1μ2⋯μm(x)\hat{A}^{\mu_1\mu_2\cdots\mu_m}_{\nu_1\nu_2\cdots\nu_n}(x) in dS3×RdS_3 \times\mathbb{R} is

A^ν1ν2⋯νnμ1μ2⋯μm(x)=τΔAAν1ν2⋯νnμ1μ2⋯μm(x).(16)\hat{A}^{\mu_1\mu_2\cdots\mu_m}_{\nu_1\nu_2\cdots\nu_n}(x) = \tau^{\Delta_A} A^{\mu_1\mu_2\cdots\mu_m}_{\nu_1\nu_2\cdots\nu_n}(x). \tag*{(16)}

where ΔA=[A]+m−n\Delta_A = [A] + m - n, and [A][A] denotes the mass dimension of AA.

Spin hydrodynamics in dS3×RdS_3 \times\mathbb{R} spacetime

In this subsection, we briefly review the spin hydrodynamic with Gubser flow following our previous studies [128]. Using this rule in Eq. (16), it is straightforward to derive that

Θ^μν=τ6Θμν.(17)\hat{\Theta}^{\mu\nu}=\tau^{6}\Theta^{\mu\nu}. \tag*{(17)}

and the energy-momentum conservation equation (1) in dS3×RdS_3 \times\mathbb{R} spacetime,

0=∇μΘμν=τ−6(∇^μΘ^μν−2τ−1Θ^[μν]∇^μτ)−τ−7Θ^ αμg^αν∇^ατ.(18)0=\nabla_{\mu}\Theta^{\mu\nu}=\tau^{-6}\left(\hat{\nabla}_{\mu}\hat{\Theta}^{\mu\nu}-2\tau^{-1}\hat{\Theta}^{[\mu\nu]}\hat{\nabla}_{\mu}\tau\right)-\tau^{-7}\hat{\Theta}^{\mu}_{\ \alpha}\hat{g}^{\alpha\nu}\hat{\nabla}_{\alpha}\tau. \tag*{(18)}

Since we set the bulk pressure to vanish, we have Θ^ μμ=0\hat{\Theta}^{\mu}_{\ \mu}=0. The above equation therefore reduces to

∇^μΘ^μν−2τ−1Θ^[μν]∇^μτ=0.(19)\hat{\nabla}_{\mu}\hat{\Theta}^{\mu\nu}-2\tau^{-1}\hat{\Theta}^{[\mu\nu]}\hat{\nabla}_{\mu}\tau=0. \tag*{(19)}

Similarly, the conservation equation for the total angular momentum tensor in Eq. (4) becomes

∇^λ(u^λS^μν)+2ϕ^μν−τ−1∇^στ(g^μσu^αS^αν+g^νσu^αS^μα+u^μS^νσ−u^νS^μσ)=0.(20)\hat{\nabla}_{\lambda}(\hat{u}^{\lambda}\hat{S}^{\mu\nu})+2\hat{\phi}^{\mu\nu}-\tau^{-1}\hat{\nabla}_{\sigma}\tau\left(\hat{g}^{\mu\sigma}\hat{u}_{\alpha}\hat{S}^{\alpha\nu}+\hat{g}^{\nu\sigma}\hat{u}_{\alpha}\hat{S}^{\mu\alpha}+\hat{u}^{\mu}\hat{S}^{\nu\sigma}-\hat{u}^{\nu}\hat{S}^{\mu\sigma}\right)=0. \tag*{(20)}

Here, we drop Σ(1)λμν\Sigma^{\lambda\mu\nu}_{(1)} and qμq^{\mu} to simplify the subsequent steps in searching for analytic solutions. The constitutive equation for ϕμν\phi^{\mu\nu} in Eq. (7) now becomes

0=τϕΔ^μαΔ^νβu^σ∇^σϕ^αβ−4τϕ−1u^σϕ^μν∇^στ−τ4γsΔ^μαΔ^νβ(∇^αu^β−∇^βu^α+4ω^αβ)+τϕ^μν+λθϕ^μν(∇^αu^α+3τ−1u^σ∇^στ),(21)\begin{aligned} 0={}&\tau_{\phi}\hat{\Delta}^{\mu\alpha}\hat{\Delta}^{\nu\beta}\hat{u}^{\sigma}\hat{\nabla}_{\sigma}\hat{\phi}_{\alpha\beta}-4\tau^{-1}_{\phi}\hat{u}^{\sigma}\hat{\phi}^{\mu\nu}\hat{\nabla}_{\sigma}\tau\\ &-\tau^{4}\gamma_{s}\hat{\Delta}^{\mu\alpha}\hat{\Delta}^{\nu\beta}\left(\hat{\nabla}_{\alpha}\hat{u}_{\beta}-\hat{\nabla}_{\beta}\hat{u}_{\alpha}+4\hat{\omega}_{\alpha\beta}\right)+\tau\hat{\phi}^{\mu\nu}\\ &+\lambda_{\theta}\hat{\phi}^{\mu\nu}\left(\hat{\nabla}_{\alpha}\hat{u}^{\alpha}+3\tau^{-1}\hat{u}^{\sigma}\hat{\nabla}_{\sigma}\tau\right), \tag*{(21)} \end{aligned}

where we have used that

∇μuν=τ∇^μu^ν−u^μ∇^ντ+g^μνu^α∇^ατ,∇σϕαβ=τ−2∇^σϕ^αβ−τ−3(ϕ^σβ∇^ατ+ϕ^ασ∇^βτ+4ϕ^αβ∇^στ)+τ−3(g^σαg^λμϕ^λβ+g^σβg^λμϕ^αλ)∇^μτ.(22)\begin{aligned} \nabla_{\mu}u_{\nu}&=\tau\hat{\nabla}_{\mu}\hat{u}_{\nu}-\hat{u}_{\mu}\hat{\nabla}_{\nu}\tau+\hat{g}_{\mu\nu}\hat{u}^{\alpha}\hat{\nabla}_{\alpha}\tau,\\ \nabla_{\sigma}\phi_{\alpha\beta}&=\tau^{-2}\hat{\nabla}_{\sigma}\hat{\phi}_{\alpha\beta}-\tau^{-3}\left(\hat{\phi}_{\sigma\beta}\hat{\nabla}_{\alpha}\tau+\hat{\phi}_{\alpha\sigma}\hat{\nabla}_{\beta}\tau+4\hat{\phi}_{\alpha\beta}\hat{\nabla}_{\sigma}\tau\right)\\ &\quad+\tau^{-3}\left(\hat{g}_{\sigma\alpha}\hat{g}^{\lambda\mu}\hat{\phi}_{\lambda\beta}+\hat{g}_{\sigma\beta}\hat{g}^{\lambda\mu}\hat{\phi}_{\alpha\lambda}\right)\hat{\nabla}_{\mu}\tau. \tag*{(22)} \end{aligned}

From the above equation, we find that conformal symmetry is preserved for the constitutive equation of ϕμν\phi^{\mu\nu} when

λθ=43τϕ.(23)\lambda_{\theta}=\frac{4}{3}\tau_{\phi}. \tag*{(23)}

For simplicity, we restrict ourselves to this choice in the remainder of this work. In this case, Eq. (21) reduces to

0=τϕτΔ^μαΔ^νβu^σ∇^σϕ^αβ+4τϕ3τϕ^μν∇^αu^α+ϕ^μν−τ3γsΔ^μαΔ^νβ(∇^αu^β−∇^βu^α+4ω^αβ),(24)\begin{aligned} 0={}&\frac{\tau_{\phi}}{\tau}\hat{\Delta}^{\mu\alpha}\hat{\Delta}^{\nu\beta}\hat{u}^{\sigma}\hat{\nabla}_{\sigma}\hat{\phi}_{\alpha\beta}+\frac{4\tau_{\phi}}{3\tau}\hat{\phi}^{\mu\nu}\hat{\nabla}_{\alpha}\hat{u}^{\alpha}+\hat{\phi}^{\mu\nu}\\ &-\tau^{3}\gamma_{s}\hat{\Delta}^{\mu\alpha}\hat{\Delta}^{\nu\beta}\left(\hat{\nabla}_{\alpha}\hat{u}_{\beta}-\hat{\nabla}_{\beta}\hat{u}_{\alpha}+4\hat{\omega}_{\alpha\beta}\right), \tag*{(24)} \end{aligned}

To close the system, we consider the equations of state for conformal fluid [105, 129],

e^=cs2p^,(25)\hat{e}=c_s^2\hat{p}, \tag*{(25)}
S^μν=χ^ω^μν+χ^′u^λ(ω^λ μu^ν−ω^λ νu^μ).(26)\hat{S}^{\mu\nu}=\hat{\chi}\hat{\omega}^{\mu\nu}+\hat{\chi}'\hat{u}^{\lambda}\left(\hat{\omega}_{\lambda}^{\ \mu}\hat{u}^{\nu}-\hat{\omega}_{\lambda}^{\ \nu}\hat{u}^{\mu}\right). \tag*{(26)}

where csc_s is the speed of sound and χ,χ′\chi,\chi' are two spin susceptibilities satisfying χ′>χ>0\chi' > \chi> 0.

Next, we insert the dS3×RdS_3 \times\mathbb{R} metric g^μν=diag⁡{1,−cosh⁡2ρ,−cosh⁡2ρsin⁡2θ,−1}\hat{g}_{\mu\nu}=\operatorname{diag}\{1,-\cosh^2\rho,-\cosh^2\rho\sin^2\theta,-1\} to derive the differential equations of spin hydrodynamics in dS3×RdS_3 \times\mathbb{R}. Eq. (19) splits into four distinct component equations, as shown below,

0=∂^ρe^+2tanh⁡ρ(e^+p^),(27)0=\hat{\partial}_{\rho}\hat{e}+2\tanh\rho(\hat{e}+\hat{p}), \tag*{(27)}

and

0=1cosh⁡2ρ∂^θp^+∂^μϕ^μθ−2τ−1ϕ^μθ∇^μτ,0=1cosh⁡2ρsin⁡2θ∂^ϕp^+∂^μϕ^μϕ+1tan⁡θϕ^θϕ−2τ−1ϕ^μϕ∇^μτ,0=∂^ηp^+∂^μϕ^μη+1tan⁡θϕ^θη−2τ−1ϕ^μη∇^μτ.(28)\begin{aligned} 0&=\frac{1}{\cosh^2\rho}\hat{\partial}_{\theta}\hat{p}+\hat{\partial}_{\mu}\hat{\phi}^{\mu\theta}-2\tau^{-1}\hat{\phi}^{\mu\theta}\hat{\nabla}_{\mu}\tau,\\ 0&=\frac{1}{\cosh^2\rho\sin^2\theta}\hat{\partial}_{\phi}\hat{p}+\hat{\partial}_{\mu}\hat{\phi}^{\mu\phi}+\frac{1}{\tan\theta}\hat{\phi}^{\theta\phi}-2\tau^{-1}\hat{\phi}^{\mu\phi}\hat{\nabla}_{\mu}\tau,\\ 0&=\hat{\partial}_{\eta}\hat{p}+\hat{\partial}_{\mu}\hat{\phi}^{\mu\eta}+\frac{1}{\tan\theta}\hat{\phi}^{\theta\eta}-2\tau^{-1}\hat{\phi}^{\mu\eta}\hat{\nabla}_{\mu}\tau. \tag*{(28)} \end{aligned}

where we have used u^μϕ^μν=0\hat{u}_{\mu}\hat{\phi}^{\mu\nu}=0 from ϕ^ρν=ϕ^νρ=0\hat{\phi}^{\rho\nu}=\hat{\phi}^{\nu\rho}=0 and inserted the u^μ\hat{u}^{\mu} in Eq. (13). Eq. (20) can be expressed as six independent equations for the components of S^μν\hat{S}^{\mu\nu},

0=∂^ρS^ϕη+3tanh⁡ρS^ϕη+2ϕ^ϕη.(29)0=\hat{\partial}_{\rho}\hat{S}^{\phi\eta}+3\tanh\rho\hat{S}^{\phi\eta}+2\hat{\phi}^{\phi\eta}. \tag*{(29)}

and

0=∂^ρS^ρθ+3tanh⁡ρS^ρθ,0=∂^ρS^ρϕ+3tanh⁡ρS^ρϕ−τ−1S^ϕθ∇^θτ,0=∂^ρS^ρη+2tanh⁡ρS^ρη−τ−1S^ηθ∇^θτ,0=∂^ρS^θϕ+4tanh⁡ρS^θϕ+2ϕ^θϕ+τ−11cosh⁡2ρS^ρϕ∇^θτ,0=∂^ρS^θη+3tanh⁡ρS^θη+2ϕ^θη+τ−11cosh⁡2ρS^ρη∇^θτ.(30)\begin{aligned} 0&=\hat{\partial}_{\rho}\hat{S}^{\rho\theta}+3\tanh\rho\hat{S}^{\rho\theta},\\ 0&=\hat{\partial}_{\rho}\hat{S}^{\rho\phi}+3\tanh\rho\hat{S}^{\rho\phi}-\tau^{-1}\hat{S}^{\phi\theta}\hat{\nabla}_{\theta}\tau,\\ 0&=\hat{\partial}_{\rho}\hat{S}^{\rho\eta}+2\tanh\rho\hat{S}^{\rho\eta}-\tau^{-1}\hat{S}^{\eta\theta}\hat{\nabla}_{\theta}\tau,\\ 0&=\hat{\partial}_{\rho}\hat{S}^{\theta\phi}+4\tanh\rho\hat{S}^{\theta\phi}+2\hat{\phi}^{\theta\phi}+\tau^{-1}\frac{1}{\cosh^2\rho}\hat{S}^{\rho\phi}\hat{\nabla}_{\theta}\tau,\\ 0&=\hat{\partial}_{\rho}\hat{S}^{\theta\eta}+3\tanh\rho\hat{S}^{\theta\eta}+2\hat{\phi}^{\theta\eta}+\tau^{-1}\frac{1}{\cosh^2\rho}\hat{S}^{\rho\eta}\hat{\nabla}_{\theta}\tau. \tag*{(30)} \end{aligned}

We find that the components S^ρϕ\hat{S}^{\rho\phi}, S^ρη\hat{S}^{\rho\eta}, S^ϕθ\hat{S}^{\phi\theta}, and S^ηθ\hat{S}^{\eta\theta} are mutually coupled and Eqs. (28) are not automatically satisfied. Equivalently, the standard Gubser-flow velocity profile in dS3×RdS_3 \times\mathbb{R} spacetime in Eq. (13) no longer remains a solution of the dynamical equations during time evolution unless we set S^ρϕ\hat{S}^{\rho\phi}, S^ρη\hat{S}^{\rho\eta}, S^ϕθ\hat{S}^{\phi\theta}, and S^ηθ\hat{S}^{\eta\theta} to zero [128]. Interestingly, we find that the evolution equation for S^ρθ\hat{S}^{\rho\theta} decouples from that for ϕ^ρθ\hat{\phi}^{\rho\theta}. As a result, the solution for S^ρθ\hat{S}^{\rho\theta} contains no additional information about the relaxation time or dissipative effects. We therefore set the initial condition S^ρθ(ρ0)=0\hat{S}^{\rho\theta}(\rho_0)=0 and do not consider this component in the following. Therefore, following our previous treatment in Ref. [128], we retain only a finite S^φη\hat{S}^{\varphi\eta} and set all other components of S^μν\hat{S}^{\mu\nu} to zero.

With this simplification, the full set of differential equations reduces to three differential equations: Eqs. (27), (29) and the constitutive equation for ϕ^φη\hat{\phi}^{\varphi\eta} from Eq. (24),

0=∂^ρϕ^φη+(113tanh⁡ρ+ττϕ)ϕ^φη−4τ^4γ^sχ^τ^ϕS^φη(31)0=\hat{\partial}_{\rho}\hat{\phi}^{\varphi\eta}+\left(\frac{11}{3}\tanh\rho+\frac{\tau}{\tau_{\phi}}\right)\hat{\phi}^{\varphi\eta}-\frac{4\hat{\tau}^{4}\hat{\gamma}_{s}}{\hat{\chi}\hat{\tau}_{\phi}}\hat{S}^{\varphi\eta} \tag*{(31)}

Eq. (27) gives exactly the same analytic solution for e^\hat{e} as in the standard Gubser flow. We combine Eqs. (29) and (31) and derive the equation for S^φη\hat{S}^{\varphi\eta}:

0=d2dρ2S^+(203tanh⁡ρ+τ^ϕ−1)ddρS^+(3+8tanh⁡2ρ+3τ^ϕ−1tanh⁡ρ+8γ^sτ^ϕ−1χ^−1)S^(32)0=\frac{d^{2}}{d\rho^{2}}\hat{S}+\left(\frac{20}{3}\tanh\rho+\hat{\tau}_{\phi}^{-1}\right)\frac{d}{d\rho}\hat{S}+\left(3+8\tanh^{2}\rho+3\hat{\tau}_{\phi}^{-1}\tanh\rho+8\hat{\gamma}_{s}\hat{\tau}_{\phi}^{-1}\hat{\chi}^{-1}\right)\hat{S} \tag*{(32)}

where

τ^ϕ=τϕτ,γ^s=τ3γs,(33)\hat{\tau}_{\phi}=\frac{\tau_{\phi}}{\tau},\qquad\hat{\gamma}_{s}=\tau^{3}\gamma_{s}, \tag*{(33)}

are the Weyl-rescaled relaxation time associated with ϕμν\phi^{\mu\nu} and the Weyl-rescaled transport coefficient γs\gamma_s, respectively. After the Weyl transformation, we assume that τ^ϕ\hat{\tau}_{\phi} and γ^s\hat{\gamma}_{s} depend on ρ\rho only. From now on, we denote S^φη\hat{S}^{\varphi\eta} by S^\hat{S}. It is then clear that S^\hat{S} is a function of ρ\rho only, consistent with our Gubser-flow assumption.

Asymptotic solutions at late time

In the previous section, we derived the main differential equation for S^\hat{S}, given in Eq. (32). We now study the asymptotic behavior of S^\hat{S} at late times, i.e., in the limit ρ→+∞\rho\to+\infty. Since Eq. (32) contains several factors of tanh⁡ρ\tanh\rho, which approach unity as ρ→∞\rho\to\infty, it is convenient to introduce the variable

w=11−tanh⁡ρ,(34)w=\frac{1}{1-\tanh\rho}, \tag*{(34)}

so that w→∞w\to\infty as ρ→∞\rho\to\infty. Let us discuss the physical meaning of ww. For this purpose, we recall Eq. (12), which yields

τL=−1sinh⁡ρ−cos⁡θcosh⁡ρ.(35)\frac{\tau}{L}=-\frac{1}{\sinh\rho-\cos\theta\cosh\rho}. \tag*{(35)}

At fixed θ\theta, one finds

τL=2w−1,when cos⁡θ=1.(36)\frac{\tau}{L}=\sqrt{2w-1},\qquad\text{when }\cos\theta=1. \tag*{(36)}

Therefore, the large ww limit corresponds to late proper time limit, i.e., τ2→∞\tau^2 \to\infty at fixed cos⁡θ\cos\theta.

Next, we introduce an auxiliary function ff defined by

f(w)=(2w−1)S^dS^dw.(37)f(w)=\frac{(2w-1)}{\hat{S}}\frac{d\hat{S}}{dw}. \tag*{(37)}

Once f(w)f(w) is determined, S^\hat{S} can be reconstructed from f(w)f(w) via

ln⁡[S^(w)S^(w0)]=∫w0wf(w′)2w′−1 dw′.(38)\ln\left[\frac{\hat{S}(w)}{\hat{S}(w_0)}\right]=\int_{w_0}^{w}\frac{f(w')}{2w'-1}\,dw'. \tag*{(38)}

To study the dynamical evolution of f(w)f(w), we follow Refs. [129, 143] and parameterize the transport coefficients as

τ^ϕ−1=αwΔ1,8γ^sτ^ϕ−1χ^−1=βwΔ2,(39)\begin{aligned} \hat{\tau}_{\phi}^{-1}&=\alpha w^{\Delta_1},\\ 8\hat{\gamma}_s\hat{\tau}_{\phi}^{-1}\hat{\chi}^{-1}&=\beta w^{\Delta_2}, \tag*{(39)} \end{aligned}

where α\alpha, β\beta, and Δ1,2\Delta_{1,2} are constants. Since the transport coefficients γ^s\hat{\gamma}_s and τ^ϕ−1\hat{\tau}_{\phi}^{-1}, as well as the spin susceptibility χ^\hat{\chi}, are positive-definite as required by the entropy principle (see, e.g., the discussion below Eqs. (7), (26) and the references therein), the constants α\alpha and β\beta must also be positive-definite. On the other hand, according to Eqs. (33), (36), one finds τϕ∼τ1−2Δ1\tau_{\phi}\sim\tau^{1-2\Delta_1}. Returning to the constitutive equation for ϕμν\phi^{\mu\nu} in Eq. (8), the relaxation-time term τϕΔαμΔβνuρ∇ρϕαβ∼τϕddτϕμν\tau_{\phi}\Delta^\mu_\alpha\Delta^\nu_\beta u^\rho\nabla_\rho\phi^{\alpha\beta}\sim\tau_{\phi}\frac{d}{d\tau}\phi^{\mu\nu} should, in general, be of higher order than ϕμν\phi^{\mu\nu} in the hydrodynamic gradient expansion, i.e., ∣τϕddτϕμν∣/∣ϕμν∣≪1\left|\tau_{\phi}\frac{d}{d\tau}\phi^{\mu\nu}\right|/|\phi^{\mu\nu}|\ll1. This implies

Δ1>0.(40)\Delta_1>0. \tag*{(40)}

By contrast, there is no strict constraint on Δ2\Delta_2. After substituting into Eq. (32), we obtain

0=w2dfdw+w22w−1f2+w2w−1[203(w−1)+αwΔ1+1]f+w2w−1(11w−16+8w−1+3αwΔ1+1−3αwΔ1+βwΔ2+1).(41)\begin{aligned} 0={}&w^2\frac{df}{dw}+\frac{w^2}{2w-1}f^2+\frac{w}{2w-1}\left[\frac{20}{3}(w-1)+\alpha w^{\Delta_1+1}\right]f\\ &+\frac{w}{2w-1}\left(11w-16+8w^{-1}+3\alpha w^{\Delta_1+1}-3\alpha w^{\Delta_1}+\beta w^{\Delta_2+1}\right). \tag*{(41)} \end{aligned}

Next, we implement the slow-roll expansion method [147, 148] to derive the asymptotic solutions. For convenience, we redefine f(w)f(w) in terms of a new function v(w)v(w) as

v(w)=12w−1f(w)+12A(w).(42)v(w)=\frac{1}{2w-1}f(w)+\frac{1}{2}\mathcal{A}(w). \tag*{(42)}

where

A(w)=26−20w−1+3αwΔ13(2w−1).(43)\mathcal{A}(w)=\frac{26-20w^{-1}+3\alpha w^{\Delta_1}}{3(2w-1)}. \tag*{(43)}

With this change of variables, Eq. (41) reduces to the Riccati-type equation

dv(w)dw+v(w)2=I(w),(44)\frac{dv(w)}{dw}+v(w)^2=\mathcal{I}(w), \tag*{(44)}

where the source term reads

I(w)=1(2w−1)2[−89−29w−2+49w−1+(13+Δ1)αwΔ1−(12Δ1+13)αwΔ1−1+14α2w2Δ1−βwΔ2].(45)\mathcal{I}(w)=\frac{1}{(2w-1)^2}\left[-\frac{8}{9}-\frac{2}{9}w^{-2}+\frac{4}{9}w^{-1}+\left(\frac{1}{3}+\Delta_1\right)\alpha w^{\Delta_1}-\left(\frac{1}{2}\Delta_1+\frac{1}{3}\right)\alpha w^{\Delta_1-1}+\frac{1}{4}\alpha^2w^{2\Delta_1}-\beta w^{\Delta_2}\right]. \tag*{(45)}

We now note that Eq. (44) contains no term linear in v(w)v(w). Next, we introduce an auxiliary parameter ϵ≪1\epsilon\ll1. We emphasize that ϵ\epsilon is merely a book-keeping parameter, which will be set to 1 at the end. We then expand v(w)v(w) in a power series in ϵ\epsilon,

v(w)=v0(w)+ϵv1(w)+ϵ2v2(w)+O(ϵ3).(46)v(w)=v_0(w)+\epsilon v_1(w)+\epsilon^2v_2(w)+\mathcal{O}(\epsilon^3). \tag*{(46)}

In the slow roll expansion, each derivative is counted as an additional power of ϵ\epsilon. Accordingly, Eq. (44) is rewritten as

ϵdv(w)dw+v(w)2=I(w).(47)\epsilon\frac{dv(w)}{dw}+v(w)^2=\mathcal{I}(w). \tag*{(47)}

Substituting Eq. (46) into the above equation, we obtain

[v02−I(w)]+ϵ[dv0dw+2v0v1]+ϵ2[dv1dw+v12+2v0v2]+O(ϵ3)=0.(48)\left[v_0^2-\mathcal{I}(w)\right]+\epsilon\left[\frac{dv_0}{dw}+2v_0v_1\right]+\epsilon^2\left[\frac{dv_1}{dw}+v_1^2+2v_0v_2\right]+\mathcal{O}(\epsilon^3)=0. \tag*{(48)}

Since ϵ\epsilon is arbitrary, the above equation can be satisfied order by order only if the coefficients of each power ϵn\epsilon^n vanish. This yields

v02−I(w)=0,dv0dw+2v0v1=0,dv1dw+v12+2v0v2=0.(49)\begin{aligned} v_0^2-\mathcal{I}(w)&=0,\\ \frac{d v_0}{d w}+2v_0v_1&=0,\\ \frac{d v_1}{d w}+v_1^2+2v_0v_2&=0. \tag*{(49)} \end{aligned}

These equations can be solved sequentially. The resulting slow-roll expansion for vv is

v(w)=±I−ϵ4dln⁡Idw∓ϵ232I5/2[−4Id2Idw2+5(dIdw)2]+O(ϵ3).(50)v(w)=\pm\sqrt{\mathcal{I}}-\frac{\epsilon}{4}\frac{d\ln\mathcal{I}}{d w}\mp\frac{\epsilon^2}{32\mathcal{I}^{5/2}}\left[-4\mathcal{I}\frac{d^2\mathcal{I}}{d w^2}+5\left(\frac{d\mathcal{I}}{d w}\right)^2\right]+\mathcal{O}(\epsilon^3). \tag*{(50)}

Since we are interested in the large ww behavior, we further simplify I(w)\mathcal{I}(w) in the w→∞w\to\infty limit. In general, we may expand I(w)\mathcal{I}(w) in Eq. (45) as

I(w)=I1wk1+I2wk2+…,(51)\mathcal{I}(w)=\mathcal{I}_1w^{k_1}+\mathcal{I}_2w^{k_2}+\ldots, \tag*{(51)}

where k1>k2k_1>k_2. We note that, given the explicit form of I(w)\mathcal{I}(w) in Eq. (45), the leading power in the large ww expansion satisfies k1>−2k_1>-2. Here, I1wk1\mathcal{I}_1w^{k_1} and I2wk2\mathcal{I}_2w^{k_2} denote the leading and next-to-leading terms, respectively, in the large ww expansion. The exponents k1,2k_{1,2} depend on the specific values of Δ1,2\Delta_{1,2}. Setting ϵ→1\epsilon\to1 and substituting Eq. (51) into Eq. (50), we obtain

v(w)=±I1wk1/2∓I22I1wk2−k1/2−k14w−1+…,(52)v(w)=\pm\sqrt{\mathcal{I}_1}w^{k_1/2}\mp\frac{\mathcal{I}_2}{2\sqrt{\mathcal{I}_1}}w^{k_2-k_1/2}-\frac{k_1}{4}w^{-1}+\ldots, \tag*{(52)}

where …\ldots denotes terms that are high order terms in the large ww expansion.

After some algebra, and using Eq. (42), we derive the corresponding asymptotic solutions for f(w)f(w) for different choices of Δ1,2\Delta_{1,2}, which are summarized in Table I. Let us briefly discuss the results summarized in Table I. Recalling Eq. (37), the late time behavior of S^\hat{S} can be obtained by integrating the corresponding branches f±(w)f_{\pm}(w). In particular, complex (imaginary) asymptotic forms of f±(w)f_{\pm}(w) imply oscillatory behavior of S^\hat{S}. For this reason, we keep the complex solutions for f(w)f(w) in Table I.

I1\mathcal{I}_1k1k_1I2\mathcal{I}_2k2k_2f±(w)f_{\pm}(w)
Δ1>Δ2\Delta_1 > \Delta_2116α2\frac{1}{16}\alpha^22Δ1−22\Delta_1-214(13+Δ1)α\frac{1}{4}\left(\frac{1}{3}+\Delta_1\right)\alphaΔ1−2\Delta_1-2{f+=−3f−=−αwΔ1\begin{cases} f_+ = -3 \\ f_- = -\alpha w^{\Delta_1} \end{cases}
Δ1=Δ2\Delta_1 = \Delta_2116α2\frac{1}{16}\alpha^22Δ1−22\Delta_1-214[(13+Δ1)α−β]\frac{1}{4}\left[\left(\frac{1}{3}+\Delta_1\right)\alpha-\beta\right]Δ1−2\Delta_1-2{f+=−(3+βα)f−=−αwΔ1\begin{cases} f_+ = -\left(3+\frac{\beta}{\alpha}\right) \\ f_- = -\alpha w^{\Delta_1} \end{cases}
Δ1<Δ2<2Δ1\Delta_1 < \Delta_2 < 2\Delta_1116α2\frac{1}{16}\alpha^22Δ1−22\Delta_1-2−β4-\frac{\beta}{4}Δ2−2\Delta_2-2{f+=−βαwΔ2−Δ1f−=−αwΔ1\begin{cases} f_+ = -\frac{\beta}{\alpha}w^{\Delta_2-\Delta_1} \\ f_- = -\alpha w^{\Delta_1} \end{cases}
Δ1<Δ2=2Δ1\Delta_1 < \Delta_2 = 2\Delta_1α2−4β16\frac{\alpha^2-4\beta}{16}2Δ1−22\Delta_1-214(13+Δ1)α\frac{1}{4}\left(\frac{1}{3}+\Delta_1\right)\alphaΔ1−2\Delta_1-2f±=12(−α±α2−4β)wΔ1f_{\pm}=\frac{1}{2}\left(-\alpha\pm\sqrt{\alpha^2-4\beta}\right)w^{\Delta_1}
Δ1<2Δ1<Δ2\Delta_1 < 2\Delta_1 < \Delta_2−β4-\frac{\beta}{4}Δ2−2\Delta_2-2116α2\frac{1}{16}\alpha^22Δ1−22\Delta_1-2f±=±iβwΔ22−α2βwΔ1f_{\pm}=\pm i\sqrt{\beta}w^{\frac{\Delta_2}{2}-\frac{\alpha}{2\beta}w^{\Delta_1}}

Table I. Asymptotic solutions for f(w)f(w) in the large ww limit for different choices Δ1,2\Delta_{1,2}. Here, k1,2k_{1,2} and I1,2\mathcal{I}_{1,2} are defined in Eq. (51).

We can also cross-check the above results by a different method. As an example, consider the case Δ1>Δ2\Delta_1>\Delta_2 with Δ1>0\Delta_1>0. In this regime, keeping only the leading contributions to Eq. (41) in the large ww limit yields

2w2dfdw+wf2+αwΔ1+1f+3αwΔ1+1=0.(53)2w^2\frac{d f}{d w}+wf^2+\alpha w^{\Delta_1+1}f+3\alpha w^{\Delta_1+1}=0. \tag*{(53)}

Under these assumptions, there are two consistent dominant balances in the large ww expansion. First, if the leading terms are wf2+αwΔ1+1fwf^2+\alpha w^{\Delta_1+1}f, one obtains f−→−αwΔ1f_-\to-\alpha w^{\Delta_1}. Second, if the last two terms αwΔ1+1f+3αwΔ1+1\alpha w^{\Delta_1+1}f+3\alpha w^{\Delta_1+1} dominate, one finds f+→−3f_+\to-3. We further assess the stability of the two asymptotic solutions f±f_\pm. Writing f→f±+δf±f\to f_\pm+\delta f_\pm and substituting into Eq. (53), we obtain δf±∼e∓12Δ1αwΔ1\delta f_\pm\sim e^{\mp\frac{1}{2\Delta_1}\alpha w^{\Delta_1}}. Hence δf+\delta f_+ decreases with ww, whereas δf−\delta f_- grows with ww. In other words, f+f_+ is stable and f−f_- is unstable. In the next section, we will show that the stable asymptotic solutions correspond to possible attractors, whereas the unstable asymptotic solutions correspond to repulsive behavior (i.e., repellers). The analysis can be extended straightforwardly to other parameter choices.

Late time attractor

We show the large ww asymptotic behavior of f(w)f(w) for different choices of α\alpha, β\beta, Δ1\Delta_1, and Δ2\Delta_2 in Figure 1. The red solid and blue dashed curves correspond to the asymptotic solutions summarized in Table I, while the gray solid curves represent numerical solutions obtained from different initial conditions. We observe that, for a wide range of initial conditions, the numerical solutions approach the red curves and move away from the blue dashed ones. We therefore identify the red solid and blue dashed curves as the attractors and repellers, respectively.

Image not extracted

Figure 1. Asymptotic behavior of f(w)f(w) at large ww for different parameter choices. The red solid and blue dashed curves denote the attractors and repellers summarized in Table I, respectively. The gray solid curves show numerical solutions for different initial conditions.

Panels (a) and (b) correspond to the two cases with constant asymptotic solutions, namely Δ1>Δ2\Delta_1>\Delta_2 with f(w)→−3f(w)\to-3, and Δ1=Δ2\Delta_1=\Delta_2 with f(w)→−(3+βα)f(w)\to-\left(3+\frac{\beta}{\alpha}\right). Panels (c) and (d) correspond to cases in which f(w)f(w) exhibits a power-law decaying behavior in the large ww limit, namely Δ1<Δ2<2Δ1\Delta_1<\Delta_2<2\Delta_1 with f+=−βαwΔ2−Δ1f_+=-\frac{\beta}{\alpha}w^{\Delta_2-\Delta_1}, and Δ1<Δ2=2Δ1\Delta_1<\Delta_2=2\Delta_1 with f+=12(−α+α2−4β)wΔ1f_+=\frac{1}{2}\left(-\alpha+\sqrt{\alpha^2-4\beta}\right)w^{\Delta_1}. The numerical solutions are consistent with the asymptotic analysis in the previous section. In the next section, we transform f(w)f(w) back to S^\hat{S} and to the spin density in conventional (t,x,y,z)(t,x,y,z) coordinates. In particular, the cases shown in panels (a) and (b) lead to a power-law decay of S^\hat{S} at late times.

From Figure 1, we observe that, for each fixed set of parameters (α,β,Δ1,Δ2)(\alpha,\beta,\Delta_1,\Delta_2), there exists a unique attractor and a unique repeller in the large ww regime. Moreover, in the large ww limit, the attractor branch takes larger values of f(w)f(w) than the repeller branch for the parameter sets shown. It can be understood from the structure of our evolution equation for S^\hat{S} in Eq. (32). As shown in the previous section, after an appropriate change of variables, the differential equation (32) can be cast into the Riccati-type form given in Eq. (44).

We can perform an additional transformation of v(w)v(w) in Eq. (44) by introducing

v(w)=V′(w)V(w),(54)v(w)=\frac{\mathcal{V}'(w)}{\mathcal{V}(w)}, \tag*{(54)}

where the prime denotes differentiation with respect to ww, i.e., V′(w)=dV/dw\mathcal{V}'(w)=d\mathcal{V}/dw. Since Eq. (44) contains no term linear in v(w)v(w), it can be mapped to the second-order linear equation

V′′(w)−I(w)V(w)=0.(55)\mathcal{V}''(w)-\mathcal{I}(w)\mathcal{V}(w)=0. \tag*{(55)}

In the large ww limit, the asymptotic solutions of Eq. (55) can be written formally as

V(w)=exp⁡[±I0(w)],I0(w)=∫w0wduI0(u),(56)\mathcal{V}(w)=\exp\left[\pm I_0(w)\right],\qquad I_0(w)=\int_{w_0}^{w}du\sqrt{\mathcal{I}_0(u)}, \tag*{(56)}

where I0(w)\mathcal{I}_0(w) is the leading order of I(w)\mathcal{I}(w) as w→∞w\to\infty. Note that I0′′(w)\mathcal{I}_0''(w) does not contribute at leading order in the large-ww expansion. For real f(w)f(w) shown in Figure 1, I0(w)\mathcal{I}_0(w) should be strictly positive. Furthermore, Eq. (45) implies I0(w)∼wδI_0(w)\sim w^\delta with δ>0\delta>0 since I0(w)≫w−2\mathcal{I}_0(w)\gg w^{-2} as w→∞w\to\infty. The general solution for V(w)\mathcal{V}(w) is then

V(w)=c1eI0(w)+c2e−I0(w),(57)\mathcal{V}(w)=c_1e^{I_0(w)}+c_2e^{-I_0(w)}, \tag*{(57)}

where c1,2c_{1,2} are constants determined by the initial conditions. Accordingly, v(w)v(w) can be written as

v(w)=I0(w)c1−c2e−2I0(w)c1+c2e−2I0(w).(58)v(w) = \sqrt{I_0(w)}\frac{c_1-c_2e^{-2I_0(w)}}{c_1+c_2e^{-2I_0(w)}}. \tag*{(58)}

If c1≠0c_1 \ne0, then e−2I0(w)e^{-2I_0(w)} decays rapidly at large ww, and therefore v(w)→I0(w)v(w) \to\sqrt{I_0(w)}. This implies that, for generic initial conditions (i.e., c1≠0c_1 \ne0), all trajectories approach the branch v(w)→I0(w)v(w) \to\sqrt{I_0(w)} at late times, identifying it as the attractor. In contrast, if c1=0c_1=0, one has v(w)=−I0(w)v(w)=-\sqrt{I_0(w)}. A small perturbation away from this branch generates a small but finite c1c_1, after which the trajectory departs from v(w)=−I0(w)v(w)=-\sqrt{I_0(w)} and approaches the attractor v(w)→I0(w)v(w) \to\sqrt{I_0(w)}. This branch therefore corresponds to the repeller. This analysis shows that there exists a unique attractor and a unique repeller in the large ww regime.

In Figure 1, we also observe that, in each panel, some gray trajectories lie below the blue dashed curve (the repeller) and keep decreasing toward −∞-\infty. We emphasize that this behavior does not indicate a physical singularity in S^\hat{S}. Recalling the definition of f(w)f(w) in Eq. (37), one can see that f(w)f(w) necessarily passes through ±∞\pm\infty when S^→0\hat{S} \to0. Therefore, trajectories of f(w)f(w) that run to −∞-\infty simply correspond to S^\hat{S} crossing zero: once S^\hat{S} changes sign, the ratio defining f(w)f(w) becomes positive again, and the solution for f(w)f(w) re-enters from +∞+\infty (i.e., f(w)f(w) effectively flips sign by passing through ±∞\pm\infty) and subsequently approaches the red solid curve (the attractor). A similar discussion can be found in our previous study [129] of attractors in spin hydrodynamics with Bjorken flow.

Based on the discussion above, we can characterize the basin of attraction. For generic initial conditions, the system converges to the attractor at late times; the only exception is the measure-zero set of initial conditions that lie exactly on the repeller branch.

Late-time power-law decay of spin density

In the conventional relativistic hydrodynamics, the typical decay behavior of the charge number density nn is n∼τ−1n \sim\tau^{-1} in a Bjorken flow. Therefore, if the spin density exhibits a power-law decay, it can remain comparable at late time. By contrast, if the spin density exhibits an exponential decay, it can be neglected at late time in comparison with the charge number density. Therefore, in this section, we discuss the power-law decay of S^\hat{S} at late time and the physical picture behind it.

Eq. (37) shows that if f(w)∼−wmf(w) \sim-w^m with m≠0m \ne0, then S^∼e−wm\hat{S} \sim e^{-w^m} decays exponentially in ww, which is much faster than the typical power-law dilution of a number density. In such cases, any accumulation effect in S^\hat{S} would be difficult to observe. By contrast, when f(w)∼const.f(w) \sim\mathrm{const.}, S^\hat{S} exhibits a power-law behavior at late times. We find only two possible late time power-law behaviors for S^\widehat{S} according to the Table I,

  1. Δ1>Δ2\Delta_1 > \Delta_2, f+(w)→−3f_+(w) \to-3, S^∝w−3/2\widehat{S} \propto w^{-3/2},

  2. Δ1=Δ2\Delta_1 = \Delta_2, f+(w)→−(3+βα)f_+(w) \to-\left(3+\frac{\beta}{\alpha}\right), S^∝w−(3α+β)/(2α)\widehat{S} \propto w^{-(3\alpha+\beta)/(2\alpha)}.

(i)Δ1>Δ2,f+(w)→−3,S^∝w−3/2,(ii)Δ1=Δ2,f+(w)→−(3+βα),S^∝w−(3α+β)/(2α).(59)\begin{aligned} \text{(i)}\quad&\Delta_1 > \Delta_2,\quad f_+(w) \to-3,\quad\widehat{S} \propto w^{-3/2},\\ \text{(ii)}\quad&\Delta_1 = \Delta_2,\quad f_+(w) \to-\left(3+\frac{\beta}{\alpha}\right),\quad\widehat{S} \propto w^{-(3\alpha+\beta)/(2\alpha)}. \tag*{(59)} \end{aligned}

We also summarize it in Figure 2.

Image not extracted

Figure 2. Late-time asymptotic behavior of S^\widehat{S} for different choices of Δ1,2\Delta_{1,2}. Here, we take τ/τφ\tau/\tau_{\varphi} as an effective Knudsen number, denoted by Kn.

In case (i), using Eq. (16), we obtain

Sφη=τ−5S^φη∝1τ2L3(L2−τ2+r2)3.(60)S^{\varphi\eta}=\tau^{-5}\widehat{S}^{\varphi\eta}\propto\frac{1}{\tau^2}\frac{L^3}{\left(L^2-\tau^2+r^2\right)^3}. \tag*{(60)}

We note that there are two distinct limits in which one can extract the asymptotic behavior of SφηS^{\varphi\eta} at late proper time τ\tau. If L≫τL\gg\tau, Eq. (60) reduces to

Sφη∼1L3τ2.(61)S^{\varphi\eta}\sim\frac{1}{L^3\tau^2}. \tag*{(61)}

We can further transform back to the (t,x,y,z)(t,x,y,z) coordinates in flat spacetime and find that the non-vanishing components of the spin density behave as

Stx,Sty,Szx,Szy∼1τL2.(62)S^{tx},S^{ty},S^{zx},S^{zy}\sim\frac{1}{\tau L^2}. \tag*{(62)}

which is consistent with our previous results in [128]. In terms of the proper-time scaling, this power-law decay is comparable to that of the conventional charge number density in Bjorken flow.

On the other hand, if τ≫L\tau\gg L, Eq. (60) gives

Sφη∼L3τ8,(63)S^{\varphi\eta} \sim\frac{L^3}{\tau^8}, \tag*{(63)}

which corresponds to the non-vanishing components

Sτx,Sτy,Szx,Szy∼L4τ7.(64)S^{\tau x}, S^{\tau y}, S^{zx}, S^{zy} \sim\frac{L^4}{\tau^7}. \tag*{(64)}

Although the spin density in this limit still exhibits a power-law decay, it decreases much more rapidly as a function of τ\tau than other macroscopic hydrodynamic variables. Consequently, its contribution is expected to be strongly suppressed at late proper time.

In case (ii), following a similar analysis, we obtain in the L≫τL \gg\tau limit

Sφη∼τ(β−2α)/αL(β+3α)/α,(65)S^{\varphi\eta} \sim\frac{\tau^{(\beta-2\alpha)/\alpha}}{L^{(\beta+3\alpha)/\alpha}}, \tag*{(65)}

which implies that the non-vanishing components of the spin density scale as

Sτx,Sτy,Szx,Szy∼τ(β−α)/αL(β+2α)/α.(66)S^{\tau x}, S^{\tau y}, S^{zx}, S^{zy} \sim\frac{\tau^{(\beta-\alpha)/\alpha}}{L^{(\beta+2\alpha)/\alpha}}. \tag*{(66)}

In the opposite limit τ≫L\tau\gg L, we find

Sφη∼L(β+3α)/ατ(β+8α)/α,(67)S^{\varphi\eta} \sim\frac{L^{(\beta+3\alpha)/\alpha}}{\tau^{(\beta+8\alpha)/\alpha}}, \tag*{(67)}

which gives the non-vanishing components of the spin density as

Sτx,Sτy,Szx,Szy∼L(β+4α)/ατ(β+7α)/α.(68)S^{\tau x}, S^{\tau y}, S^{zx}, S^{zy} \sim\frac{L^{(\beta+4\alpha)/\alpha}}{\tau^{(\beta+7\alpha)/\alpha}}. \tag*{(68)}

Interestingly, Eq. (66) shows that the spin density scales with proper time as τ(β−α)/α\tau^{(\beta-\alpha)/\alpha}. As mentioned around Eq. (39), the physical constraint only requires that α\alpha and β\beta are positive. We note that when β=α\beta= \alpha the spin density in Eq. (66) is proper time independent, whereas for β>α\beta> \alpha it can even increase with proper time. Here we emphasize that Eq. (66) is derived in the L≫τL \gg\tau limit. Interpreting LL as the characteristic size of the fireball produced in the collision, this regime may not be realized in phenomenologically relevant situations.

To explain the physics in cases (i) and (ii), we return to Eqs. (29), (31). Eq. (31) follows from Eq. (24). Using Eqs. (39) and τ^ϕ∼α−1w−Δ1\hat{\tau}_{\phi} \sim\alpha^{-1}w^{-\Delta_1}, if ∣τ^ϕΔ^μαΔ^νβu^σ∇^σϕ^αβ∣,∣43τ^ϕϕ^μν∇^αu^α∣≪∣ϕ^μν∣\left|\hat{\tau}_{\phi}\hat{\Delta}^{\mu\alpha}\hat{\Delta}^{\nu\beta}\hat{u}^{\sigma}\hat{\nabla}_{\sigma}\hat{\phi}_{\alpha\beta}\right|,\left|\frac{4}{3}\hat{\tau}_{\phi}\hat{\phi}^{\mu\nu}\hat{\nabla}_{\alpha}\hat{u}^{\alpha}\right| \ll\left|\hat{\phi}^{\mu\nu}\right|, then Eq. (24) reduces to the algebraic relation ϕ^μν≃−4τ^ϕ3γsΔ^μαΔ^νβω^αβ\hat{\phi}^{\mu\nu} \simeq-4\hat{\tau}_{\phi}^{3}\gamma_s\hat{\Delta}^{\mu\alpha}\hat{\Delta}^{\nu\beta}\hat{\omega}_{\alpha\beta}. This corresponds to a regime in which the effective relaxation time τ^ϕ\hat{\tau}_{\phi} decreases more rapidly than the characteristic evolution time of ϕ^μν\hat{\phi}^{\mu\nu}, so that the dissipative term proportional to γs\gamma_s dominates the evolution of ϕ^μν\hat{\phi}^{\mu\nu}. This implies that, when the perturbation is small and driven by dissipative effects, the effective response time τ^ϕ\hat{\tau}_\phi is too short for ϕ^μν\hat{\phi}^{\mu\nu} to develop through its own relaxation dynamics. Instead, ϕ^μν\hat{\phi}^{\mu\nu} rapidly approaches an algebraic form determined by the dissipative source term. We then insert this relation into the evolution equation for S^ϕη\hat{S}^{\phi\eta} in Eq. (29). Compared with Eq. (4), Eq. (29) can be interpreted as a local balance equation implied by total angular-momentum conservation. The first two terms, ∂^ρS^ϕη+3tanh⁡ρ S^ϕη\hat{\partial}_\rho\hat{S}^{\phi\eta} + 3\tanh\rho\,\hat{S}^{\phi\eta}, describe the evolution of the spin tensor, while the source term −2ϕ^ϕη-2\hat{\phi}^{\phi\eta} represents the local exchange with orbital angular momentum, i.e., the spin–orbit transfer that redistributes angular momentum between spin and orbital sectors while preserving the total. Next, we compare these two types of variations. From the scaling of the source term, one finds −2ϕ^ϕη∼βαwΔ2−Δ1S^ϕη-2\hat{\phi}^{\phi\eta} \sim\frac{\beta}{\alpha}w^{\Delta_2-\Delta_1}\hat{S}^{\phi\eta}. If Δ1>Δ2\Delta_1 > \Delta_2, i.e. in case (i), the exchange with orbital angular momentum is parametrically faster than the spin-tensor evolution and dominates over both ∂^ρS^ϕη\hat{\partial}_\rho\hat{S}^{\phi\eta} and the expansion term 3tanh⁡ρ S^ϕη∼S^ϕη(∇^⋅u^)3\tanh\rho\,\hat{S}^{\phi\eta} \sim\hat{S}^{\phi\eta}(\hat{\nabla}\cdot\hat{u}). At late times, the remaining evolution of the spin density is then primarily controlled by the expansion term, 3tanh⁡ρ S^ϕη3\tanh\rho\,\hat{S}^{\phi\eta}, which leads to the power-law decay in Eq. (60).

If Δ1=Δ2\Delta_1 = \Delta_2, i.e. in case (ii), the exchange with orbital angular momentum remains comparable to the spin-tensor evolution at late times. Combining the source term with the expansion term then yields ∂^ρS^ϕη∼−(3+βα)S^ϕη\hat{\partial}_\rho\hat{S}^{\phi\eta} \sim-(3+\frac{\beta}{\alpha})\hat{S}^{\phi\eta} at late times, which gives the power-law behavior in Eq. (65). We then briefly discuss the nontrivial behavior shown in Eq. (66). Recalling Eqs. (33), (39), we find that, when Δ1=Δ2\Delta_1 = \Delta_2, β/α=8γ^s/χ^=8γsτ/χ\beta/\alpha= 8\hat{\gamma}_s/\hat{\chi} = 8\gamma_s\tau/\chi where χ\chi is the spin susceptibility in flat Minkowski space. Since χ\chi is fixed by the equations of state, we take it to be a constant in the following analysis for simplicity. This relation implies the late-time scaling γs∼(β/α)(χ/8)τ−1\gamma_s \sim(\beta/\alpha)(\chi/8)\tau^{-1}. Inserting this scaling into Eq. (8) shows that, for β/α>1\beta/\alpha> 1, the orbital angular-momentum exchange term (encoded in the source ϕμν\phi^{\mu\nu}) decreases more slowly than the typical late-time scaling τ−1\tau^{-1}. Consequently, the source term remains sufficiently large at late times and can counteract the expansion-driven dilution of the spin density. As a result, the spin density can become nearly constant or even increase with τ\tau, consistent with the nontrivial static or growing behavior found in Eq. (66). In contrast, if β/α<1\beta/\alpha< 1, then γs\gamma_s (and thus ϕμν\phi^{\mu\nu}) is suppressed, so the source term becomes subleading and the spin density follows the usual power-law decay at late times.

Conclusion

In this work, we have investigated the late-time asymptotic solutions and attractor structure of minimal causal spin hydrodynamics in Gubser flow. We first derive the governing differential equation for the spin density S^\hat{S} in Eq. (32) and introduce the auxiliary function f(w)f(w) in Eq. (37). We then obtain the large ww asymptotic solutions of f(w)f(w), summarized in Table I. Next, we present numerical solutions for f(w)f(w) for different parameter choices, shown in Figure 1. We identify both attractors and repellers in Figure 1 and analyze their properties. Finally, we map the asymptotic solutions back to the spin density in flat Minkowski space with coordinates (t,x,y,z)(t,x,y,z), and we delineate the parameter regions associated with different decay behaviors in Figure 2. In particular, by considering the two late-time regimes, case (i) Δ1>Δ2\Delta_1 > \Delta_2 and case (ii) Δ1=Δ2\Delta_1 = \Delta_2, we identify parameter regions in which the spin density exhibits a non-vanishing power-law decay rather than an exponential decay. This result indicates that, for suitable parameter choices, the spin density can evolve as a hydrodynamic mode governed by the asymptotic scaling of the underlying flow, rather than as a rapidly damped non-hydrodynamic variable.

Acknowledgments

This work is supported in part by the National Key Research and Development Program of China under Contract No. 2022YFA1605500, by the Chinese Academy of Sciences (CAS) under Grant No. YSBR-088 and by National Natural Science Foundation of China (NSFC) under Grants No. 12135011. D.-L. Wang is supported in part by the National Natural Science Foundation of China under Grant No. 125B2110.

References

  1. [1]Z.-T. Liang and X.-N. Wang, Phys. Rev. Lett. 94, 102301 (2005), nucl-th/0410079, [Erratum: Phys.Rev.Lett. 96, 039901 (2006)].
  2. [2]Z.-T. Liang and X.-N. Wang, Phys. Lett. B 629, 20 (2005), nucl-th/0411101.
  3. [3]STAR, L. Adamczyk et al., Nature 548, 62 (2017), 1701.06657.
  4. [4]STAR, J. Adam et al., Phys. Rev. C 98, 014910 (2018), 1805.04400.
  5. [5]STAR, J. Adam et al., Phys. Rev. Lett. 126, 162301 (2021), 2012.13601.
  6. [6]STAR, M. S. Abdallah et al., Phys. Rev. C 104, L061901 (2021), 2108.00044.
  7. [7]STAR, M. I. Abdulhamid et al., Phys. Rev. C 108, 014910 (2023), 2305.08705.
  8. [8]I. Karpenko and F. Becattini, Eur. Phys. J. C 77, 213 (2017), 1610.04717.
  9. [9]H. Li, L.-G. Pang, Q. Wang, and X.-L. Xia, Phys. Rev. C 96, 054908 (2017), 1704.01507.
  10. [10]Y. Sun and C. M. Ko, Phys. Rev. C 96, 024906 (2017), 1706.09467.
  11. [11]D.-X. Wei, W.-T. Deng, and X.-G. Huang, Phys. Rev. C 99, 014905 (2019), 1810.00151.
  12. [12]O. Vitiuk, L. V. Bravina, and E. E. Zabrodin, Phys. Lett. B 803, 135298 (2020), 1910.06292.
  13. [13]B. Fu, K. Xu, X.-G. Huang, and H. Song, Phys. Rev. C 103, 024903 (2021), 2011.03740.
  14. [14]A. Lei, D. Wang, D.-M. Zhou, B.-H. Sa, and L. P. Csernai, Phys. Rev. C 104, 054903 (2021), 2110.13485.
  15. [15]S. Ryu, V. Jupic, and C. Shen, Phys. Rev. C 104, 054908 (2021), 2106.08125.
  16. [16]HADES, R. Abou Yassine et al., Phys. Lett. B 835, 137506 (2022), 2207.05160.
  17. [17]Y. Guo, J. Liao, E. Wang, H. Xing, and H. Zhang, Phys. Rev. C 104, L041902 (2021), 2105.13481.
  18. [18]X.-G. Deng, X.-G. Huang, Y.-G. Ma, and S. Zhang, Phys. Rev. C 101, 064908 (2020), 2001.01371.
  19. [19]X.-G. Deng, X.-G. Huang, and Y.-G. Ma, (2021), 2109.09956.
  20. [20]K.-J. Sun et al., Phys. Rev. Lett. 134, 022301 (2025).
  21. [21]D.-N. Liu et al., (2025), 2508.12193.
  22. [22]Y.-P. Zheng et al., (2025), 2509.15286.
  23. [23]J. Xu, (2026), 2602.23793.
  24. [24]S. Y. F. Liu and Y. Yin, JHEP 07, 188 (2021), 2103.09200.
  25. [25]B. Fu, S. Y. F. Liu, L. Pang, H. Song, and Y. Yin, Phys. Rev. Lett. 127, 142301 (2021), 2103.10403.
  26. [26]F. Becattini, M. Buzzegoli, and A. Palermo, Phys. Lett. B 820, 136519 (2021), 2103.10917.
  27. [27]F. Becattini, M. Buzzegoli, G. Inghirami, I. Karpenko, and A. Palermo, Phys. Rev. Lett. 127, 272302 (2021), 2103.14621.
  28. [28]Y. Hidaka, S. Pu, and D.-L. Yang, Phys. Rev. D 97, 016004 (2018), 1710.00278.
  29. [29]C. Yi, S. Pu, and D.-L. Yang, Phys. Rev. C 104, 064901 (2021), 2106.00238.
  30. [30]C. Yi, S. Pu, J.-H. Gao, and D.-L. Yang, Phys. Rev. C 105, 044911 (2022), 2112.15531.
  31. [31]W. Florkowski, A. Kumar, A. Mazeliauskas, and R. Ryblewski, Phys. Rev. C 105, 064901 (2022), 2112.02799.
  32. [32]S. Alzhrani, S. Ryu, and C. Shen, Phys. Rev. C 106, 014905 (2022), 2203.15718.
  33. [33]A. Palermo, F. Becattini, M. Buzzegoli, G. Inghirami, and I. Karpenko, EPJ Web Conf. 276, 01026 (2023), 2208.09874.
  34. [34]M. Buzzegoli, F. Becattini, G. Inghirami, I. Karpenko, and A. Palermo, Acta Phys. Polon. Supp. 16, 39 (2023), 2208.04449.
  35. [35]X.-Y. Wu, C. Yi, G.-Y. Qin, and S. Pu, Phys. Rev. C 105, 064909 (2022), 2204.0218.
  36. [36]C. Yi et al., Phys. Rev. C 109, L011901 (2024), 2304.08777.
  37. [37]X.-Y. Wu, C. Yi, G.-Y. Qin, and S. Pu, Global and local polarization of Λ hyperons across RHIC-BES energies, in 30th International Conference on Ultrarelativistic Nucleus-Nucleus Collisions, 2023, 2312.09068.
  38. [38]A. Palermo, E. Grossi, I. Karpenko, and F. Becattini, Eur. Phys. J. C 84, 920 (2024), 2404.14295.
  39. [39]B. Fu, L. Pang, H. Song, and Y. Yin, (2022), 2201.12970.
  40. [40]Y. B. Ivanov and A. A. Soldatov, Pisma Zh. Eksp. Teor. Fiz. 116, 137 (2022), 2206.06927.
  41. [41]S. Fang and S. Pu, Phys. Rev. D 111, 034015 (2025), 2408.09877.
  42. [42]J.-R. Wang, S. Fang, D.-L. Yang, and S. Pu, (2025), 2507.15238.
  43. [43]S. Fang, S. Pu, and D.-L. Yang, Phys. Rev. D 106, 016002 (2022), 2204.11519.
  44. [44]S. Fang, S. Pu, and D.-L. Yang, Phys. Rev. D 112, 014038 (2025), 2503.13320.
  45. [45]C. Yi, X.-Y. Wu, J. Zhu, S. Pu, and G.-Y. Qin, Phys. Rev. C 111, 044901 (2025), 2408.04296.
  46. [46]CMS, A. Hayrapetyan et al., Phys. Rev. Lett. 135, 132301 (2025), 2502.07898.
  47. [47]C. Yi, S. Fang, D.-L. Wang, and S. Pu, Local spin polarization of Λ hyperons and its interaction corrections, in 31st International Conference on Ultra-relativistic Nucleus-Nucleus Collisions, 2025, 2509.00377.arxiv.org/abs/2509.00377
  48. [48]C. Yi, X.-Y. Wu, J. Zhu, S. Pu, and G.-Y. Qin, Hydrodynamic effects on spin polarization along the beam direction in Au+Au and p+Pb collisions, in 31st International Conference on Ultra-relativistic Nucleus-Nucleus Collisions, 2025, 2509.00380.arxiv.org/abs/2509.00380
  49. [49]J.-H. Gao, G.-L. Ma, S. Pu, and Q. Wang, Nucl. Sci. Tech. 31, 90 (2020), 2005.10432.
  50. [50]F. Becattini, Rept. Prog. Phys. 85, 122301 (2022), 2204.01144.
  51. [51]F. Becattini et al., (2024), 2402.04540.
  52. [52]T. Niida and S. A. Voloshin, Int. J. Mod. Phys. E 33, 2430010 (2024), 2404.11042.
  53. [53]J.-H. Gao, Z.-T. Liang, S. Pu, Q. Wang, and X.-N. Wang, Phys. Rev. Lett. 109, 232301 (2012), 1203.0725.
  54. [54]J.-W. Chen, S. Pu, Q. Wang, and X.-N. Wang, Phys. Rev. Lett. 110, 262301 (2013), 1210.8312.
  55. [55]Y. Hidaka, S. Pu, and D.-L. Yang, Phys. Rev. D 95, 091901 (2017), 1612.04630.
  56. [56]J.-H. Gao and Z.-T. Liang, Phys. Rev. D 100, 056021 (2019), 1902.06510.
  57. [57]N. Weickgenannt, X.-L. Sheng, E. Speranza, Q. Wang, and D. H. Rischke, Phys. Rev. D 100, 056018 (2019), 1902.06513.
  58. [58]Y.-C. Liu, K. Mameda, and X.-G. Huang, Chin. Phys. C 44, 094101 (2020), 2002.03753, [Erratum: Chin.Phys.C 45, 089001 (2021)].
  59. [59]N. Weickgenannt, E. Speranza, X.-l. Sheng, Q. Wang, and D. H. Rischke, Phys. Rev. Lett. 127, 052301 (2021), 2005.01506.
  60. [60]N. Weickgenannt, E. Speranza, X.-l. Sheng, Q. Wang, and D. H. Rischke, Phys. Rev. D 104, 016022 (2021), 2103.04896.
  61. [61]X.-L. Sheng, N. Weickgenannt, E. Speranza, D. H. Rischke, and Q. Wang, Phys. Rev. D 104, 016029 (2021), 2103.10636.
  62. [62]Y. Hidaka, S. Pu, Q. Wang, and D.-L. Yang, Prog. Part. Nucl. Phys. 127, 103989 (2022), 2201.07644.
  63. [63]W.-B. Dong, Y.-L. Yin, and Q. Wang, Phys. Rev. C 106, 054909 (2022), 2209.12402.
  64. [64]W.-B. Dong, Y.-L. Yin, X.-L. Sheng, S.-Z. Yang, and Q. Wang, Phys. Rev. D 109, 056025 (2024), 2311.18400.
  65. [65]S. Fang, S. Pu, and D.-L. Yang, Phys. Rev. D 109, 034034 (2024), 2311.15197.
  66. [66]Y.-L. Yin, W.-B. Dong, J.-Y. Pang, S. Pu, and Q. Wang, (2024), 2402.03672.
  67. [67]S. Bhadury, Z. Drogosz, W. Florkowski, S. K. Kar, and V. Mykhaylova, (2025), 2505.02657.
  68. [68]D. Montenegro, L. Tinti, and G. Torrieri, Phys. Rev. D 96, 056012 (2017), 1701.08263, [Addendum: Phys.Rev.D 96, 079901 (2017)].
  69. [69]W. Florkowski, B. Friman, A. Jaiswal, and E. Speranza, Phys. Rev. C 97, 041901 (2018), 1705.00587.
  70. [70]W. Florkowski, A. Kumar, and R. Ryblewski, Prog. Part. Nucl. Phys. 108, 103709 (2019), 1811.04409.
  71. [71]K. Hattori, M. Hongo, X.-G. Huang, M. Matsuo, and H. Taya, Phys. Lett. B 795, 100 (2019), 1901.06615.
  72. [72]K. Fukushima and S. Pu, Phys. Lett. B 817, 136346 (2021), 2010.01608.
  73. [73]F. Becattini, W. Florkowski, and E. Speranza, Phys. Lett. B 789, 419 (2019), 1807.10994.
  74. [74]J. Hu, Phys. Rev. C 107, 024915 (2023), 2209.10979.
  75. [75]S. Dey and A. Das, Phys. Rev. D 111, 074037 (2025), 2410.04141.
  76. [76]A. Tiwari and B. K. Patra, Phys. Rev. D 112, 036014 (2025), 2408.11514.
  77. [77]D. She, Y.-W. Qiu, and D. Hou, Phys. Rev. D 111, 036027 (2025), 2410.15142.
  78. [78]W. Florkowski, E. Speranza, and F. Becattini, Acta Phys. Polon. B 49, 1409 (2018), 1803.11098.
  79. [79]S. Bhadury, W. Florkowski, A. Jaiswal, A. Kumar, and R. Ryblewski, Phys. Lett. B 814, 136096 (2021), 2002.03937.
  80. [80]S. Shi, C. Gale, and S. Jeon, Phys. Rev. C 103, 044906 (2021), 2008.08618.
  81. [81]E. Speranza and N. Weickgenannt, Eur. Phys. J. A 57, 155 (2021), 2007.00138.
  82. [82]S. Bhadury, W. Florkowski, A. Jaiswal, A. Kumar, and R. Ryblewski, Phys. Rev. D 103, 014030 (2021), 2008.10976.
  83. [83]R. Singh, G. Sophys, and R. Ryblewski, Phys. Rev. D 103, 074024 (2021), 2011.14907.
  84. [84]H.-H. Peng, J.-J. Zhang, X.-L. Sheng, and Q. Wang, Chin. Phys. Lett. 38, 116701 (2021), 2107.00448.
  85. [85]N. Weickgenannt, D. Wagner, E. Speranza, and D. H. Rischke, Phys. Rev. D 106, 096014 (2022), 2203.04766.
  86. [86]N. Weickgenannt, D. Wagner, and E. Speranza, Phys. Rev. D 105, 116026 (2022), 2204.01797.
  87. [87]N. Weickgenannt, D. Wagner, E. Speranza, and D. H. Rischke, Phys. Rev. D 106, L091901 (2022), 2208.01955.
  88. [88]S. Bhadury, W. Florkowski, A. Jaiswal, A. Kumar, and R. Ryblewski, Phys. Rev. Lett. 129, 192301 (2022), 2204.01357.
  89. [89]D. Wagner, N. Weickgenannt, and E. Speranza, Phys. Rev. D 108, 116017 (2023), 2306.05936.
  90. [90]W. Florkowski and M. Hontarenko, Phys. Rev. Lett. 134, 082302 (2025), 2405.03263.
  91. [91]D. Wagner, M. Shokri, and D. H. Rischke, (2024), 2405.00533.
  92. [92]D. Wagner, Phys. Rev. D 111, 016008 (2025), 2409.07143.
  93. [93]A. Daher, X.-L. Sheng, D. Wagner, and F. Becattini, Phys. Rev. D 112, 094020 (2025), 2503.03713.
  94. [94]A. D. Gallegos and U. Gürsoy, JHEP 11, 151 (2020), 2004.05148.
  95. [95]S. Li, M. A. Stephanov, and H.-U. Yee, Phys. Rev. Lett. 127, 082302 (2021), 2011.12318.
  96. [96]A. D. Gallegos, U. Gürsoy, and A. Yarom, SciPost Phys. 11, 041 (2021), 2101.04759.
  97. [97]D. She, A. Huang, D. Hou, and J. Liao, Sci. Bull. 67, 2265 (2022), 2105.04060.
  98. [98]M. Hongo, X.-G. Huang, M. Kaminski, M. Stephanov, and H.-U. Yee, JHEP 11, 150 (2021), 2107.14231.
  99. [99]R. Biswas, A. Daher, A. Das, W. Florkowski, and R. Ryblewski, (2023), 2304.01009.
  100. [100]S. Fang, K. Fukushima, S. Pu, and D.-L. Wang, (2025), 2506.20698.
  101. [101]Z.-H. Zhang, X.-H. Lv, and X.-G. Huang, (2026), 2603.17794.
  102. [102]D. Montenegro, L. Tinti, and G. Torrieri, Phys. Rev. D 96, 076016 (2017), 1703.03079.
  103. [103]P. Shi and H. Xu-Guang, Acta Phys. Sin. 72, 071202 (2023).
  104. [104]X.-G. Huang, Nucl. Sci. Tech. 36, 208 (2025), 2411.11753.
  105. [105]X.-Q. Xie, D.-L. Wang, C. Yang, and S. Pu, Phys. Rev. D 108, 094031 (2023), 2306.13880.
  106. [106]G. Sarwar, M. Hasanujjaman, J. R. Bhatt, H. Mishra, and J.-e. Alam, Phys. Rev. D 107, 054031 (2023), 2209.08652.
  107. [107]A. Daher, A. Das, and R. Ryblewski, Phys. Rev. D 107, 054043 (2023), 2209.10460.
  108. [108]A. Daher, W. Florkowski, R. Ryblewski, and F. Taghinavaz, (2024), 2403.04711.
  109. [109]G. S. Denicol, T. Kodama, T. Koide, and P. Mota, J. Phys. G 35, 115102 (2008), 0807.3120.
  110. [110]S. Pu, T. Koide, and D. H. Rischke, Phys. Rev. D 81, 114039 (2010), 0907.3906.
  111. [111]P. Kovtun, JHEP 10, 034 (2019), 1907.08191.
  112. [112]M. P. Heller, A. Serantes, M. Spaliński, and B. Withers, Phys. Rev. Lett. 130, 261601 (2023), 2212.07434.
  113. [113]L. Gavassino, Phys. Lett. B 840, 137854 (2023), 2301.06651.
  114. [114]L. Gavassino, M. M. Disconzi, and J. Noronha, Phys. Rev. Lett. 132, 162301 (2024), 2307.05987.
  115. [115]D.-L. Wang and S. Pu, Phys. Rev. D 109, L031504 (2024), 2309.11708.
  116. [116]R. E. Hoult and P. Kovtun, Phys. Rev. D 109, 046018 (2024), 2309.11703.
  117. [117]X. Ren, C. Yang, D.-L. Wang, and S. Pu, (2024), 2405.03105.
  118. [118]L. Gavassino, M. Antonelli, and B. Haskell, Phys. Rev. Lett. 128, 010606 (2022), 2105.14621.
  119. [119]S. Dey, W. Florkowski, A. Jaiswal, and R. Ryblewski, (2023), 2303.05271.
  120. [120]M. Buzzegoli, Phys. Rev. C 105, 044907 (2022), 2109.12084.
  121. [121]M. Buzzegoli and A. Palermo, Phys. Rev. Lett. 133, 262301 (2024), 2407.14345.
  122. [122]F. Becattini and C. Hoyos, (2025), 2507.09249.
  123. [123]F. Becattini, A. Daher, and X.-L. Sheng, Phys. Lett. B 850, 138533 (2024), 2309.05789.
  124. [124]F. Becattini and R. Singh, Eur. Phys. J. C 85, 1338 (2025), 2506.20681.
  125. [125]V. E. Ambrus and A. Gečić, (2025), 2509.17640.
  126. [126]J. Armas and A. Jain, (2026), 2601.14421.
  127. [127]D.-L. Wang, S. Fang, and S. Pu, Phys. Rev. D 104, 114043 (2021), 2107.11726.
  128. [128]D.-L. Wang, X.-Q. Xie, S. Fang, and S. Pu, Phys. Rev. D 105, 114050 (2022), 2112.15535.
  129. [129]D.-L. Wang, L. Yan, and S. Pu, Phys. Rev. D 111, 034033 (2025), 2408.03781.
  130. [130]M. P. Heller and M. Spalinski, Phys. Rev. Lett. 115, 072501 (2015), 1503.07514.
  131. [131]P. Romatschke, Phys. Rev. Lett. 120, 012301 (2018), 1704.08699.
  132. [132]J.-P. Blaizot and L. Yan, Phys. Lett. B 780, 283 (2018), 1712.03856.
  133. [133]G. S. Denicol and J. Noronha, Phys. Rev. D 97, 056021 (2018), 1711.01657.
  134. [134]M. Spaliński, Phys. Lett. B 776, 468 (2018), 1708.01921.
  135. [135]M. Strickland, JHEP 12, 128 (2018), 1809.01200.
  136. [136]J.-P. Blaizot and L. Yan, Annals Phys. 412, 167993 (2020), 1904.08677.
  137. [137]J. Brewer, L. Yan, and Y. Yin, Phys. Lett. B 816, 136189 (2021), 1910.00021.
  138. [138]D. Almaalol, A. Kurkela, and M. Strickland, Phys. Rev. Lett. 125, 122302 (2020), 2004.05195.
  139. [139]J.-P. Blaizot and L. Yan, Phys. Lett. B 820, 136478 (2021), 2006.08815.
  140. [140]M. P. Heller, R. Jefferson, M. Spaliński, and V. Svensson, Phys. Rev. Lett. 125, 132301 (2020), 2003.07368.
  141. [141]M. P. Heller, A. Serantes, M. Spaliński, V. Svensson, and B. Withers, Phys. Rev. D 104, 066002 (2021), 2007.05524.
  142. [142]M. P. Heller, A. Serantes, M. Spaliński, V. Svensson, and B. Withers, Phys. Rev. Lett. 128, 122302 (2022), 2110.07621.
  143. [143]J.-P. Blaizot and L. Yan, Phys. Rev. C 104, 055201 (2021), 2106.10508.
  144. [144]Z. Chen, D. Teaney, and L. Yan, Phys. Rev. C 108, 064911 (2023), 2206.12778.
  145. [145]S. S. Gubser, Phys. Rev. D 82, 085027 (2010), 1006.0006.
  146. [146]S. S. Gubser and A. Yarom, Nucl. Phys. B 846, 469 (2011), 1012.1314.
  147. [147]M. Strickland, J. Noronha, and G. Denicol, Phys. Rev. D 97, 036020 (2018), 1709.06644.
  148. [148]G. S. Denicol and J. Noronha, Phys. Rev. D 99, 116004 (2019), 1804.04771.

Paper details

vlnc:W_aj94sqgq
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