-
We focused on the so-called Rindler-coordinates for which
$ \tau = \sqrt{t^{2}-r^{2}} $ is the proper time and$ \eta_{s} = 0.5\ln [(1+r/t)/(1-r/t)] $ is the space-time rapidity, where$ x^\mu = (t,r_1,\dots,r_d) $ and$ r = \sqrt{\Sigma_i r_i^2} $ [16, 18, 22]. We consider (1+1) dimensional fluid flow in (1+3) dimensional space-time since we focus on the perturbation solutions of a longitudinal expanding fireball with shear viscosity. The flow 4-velocity field$ u^{\mu} $ in the cartesian coordinates (the Minkowski flat space-time) for this system is$ u^{\mu} = (\cosh\Omega, 0, 0, \sinh\Omega), $
(1) where flow rapidity
$ \Omega $ is a function of space-time rapidity$ \eta_{s} $ and is independent of proper time$ \tau $ [18], with the 4-velocity normalized as$ u^{\mu}u_{\mu} = 1 $ . The second-order hydrodynamic equations without external currents are simply given by$ \partial_{\mu} T^{\mu\nu} = 0, $
(2) with the energy-momentum tensor
$ T^{\mu\nu} \!=\! \varepsilon u^{\mu}u^{\nu} \!-\!p \Delta^{\mu\nu} \!+\! \pi^{\mu\nu} $ , where$ \varepsilon $ is the energy density, p is the pressure,$ g_{\mu\nu} $ =$ {\bf{diag}} $ $ (1,-1,-1,-1) $ the metric tensor, and$ \Delta^{\mu\nu} = g^{\mu\nu}-u^{\mu}u^{\nu} $ the projection operator which is orthogonal to the fluid velocity. The shear pressure tensor$ \pi^{\mu\nu} $ represents the deviation from ideal hydrodynamics and local equilibrium, it satisfies$ u^{\mu}\pi_{\mu\nu} = 0 $ and is traceless$ \pi^{\mu}_{\; \mu} = 0 $ in the Landau frame.The energy density and pressure are related to each other by the equation of state (EoS),
$ \varepsilon = \kappa p, $
(3) where
$ \kappa $ is usually related to the local temperature [44], in this case we assume$ \kappa $ to be a constant and independent of the temperature.The fundamental equations of viscous fluid dynamics are established by projecting appropriately the conservation equations of the energy momentum tensor (Eq. (2)). The conservation equations can be rewritten as,
$ D\varepsilon = -(\varepsilon+p)\theta+\sigma_{\mu\nu}\pi^{\mu\nu}, $
(4) and
$ (\varepsilon+p)Du^{\alpha} = \nabla^{\alpha}p+\pi^{\alpha\mu}Du_{\mu}-\Delta^{\alpha\nu}\nabla^{\mu}\pi_{\mu\nu}, $
(5) respectively, where
$ D = u^{\mu}\partial_{\mu} $ is the comoving derivative and$ \theta = \partial_{\mu}u^{\mu} $ is the expansion rate.In terms of the 14-moment approximation result from [7, 45],
$ \partial_{\mu}s^{\mu}\geqslant 0 $ reduces the corresponding thermodynamic forces. The general traceless shear tensor$ \pi^{\mu\nu} $ is [7, 41],$ \begin{split} \pi^{\mu\nu} =& 2\eta\sigma^{\mu\nu}-\tau_\pi \left[ \Delta^\mu_\alpha\Delta^\nu_\beta u^\lambda \nabla_\lambda \pi^{\alpha\beta} +\frac{1}{3}\pi^{\mu\nu} \theta\right] \\&- \lambda_1 \pi^{\langle\mu}_{\ \ \lambda} \pi^{\nu\rangle\lambda}-\lambda_2 \pi^{\langle \mu}_{\ \ \lambda} \Omega^{\nu\rangle\lambda}- \lambda_3 \Omega^{\langle \mu}_{\ \ \lambda}\Omega^{\nu\rangle \lambda}, \end{split} $
with the symmetric shear tensor
$ \sigma^{\mu\nu} $ and the antisymmetric vorticity tensor$ \Omega^{\mu\nu} $ defined as$ \sigma^{\mu\nu}\equiv\left(\frac{1}{2}(\Delta^{\mu}_{\alpha}\Delta^{\nu}_{\beta}+\Delta^{\mu}_{\beta}\Delta^{\nu}_{\alpha}) -\frac{1}{3}\Delta^{\mu\nu}\Delta_{\alpha\beta}\right)\partial^{\alpha}u^{\beta}, $
(6) $ \Omega^{\mu\nu}\equiv \frac{1}{2}\Delta^{\mu\alpha}\Delta^{\nu\beta}(\nabla_\alpha u_\beta -\nabla_\beta u_\alpha), $
(7) where
$ \eta $ ,$ \tau_{\pi} $ ,$ \lambda_{1} $ ,$ \lambda_{2} $ ,$ \lambda_{3} $ are positive transport coefficients in the flat space time.$ \eta $ is the shear viscosity coefficient, and$ \tau_{\pi} $ is the relaxation time for shear pressure tensor corresponding to the dissipative currents, respectively. Shear viscosity ratio$ \eta/s $ of the QGP is very close to the lower bound of$ 1/4\pi $ computed for a strongly coupled gauge theory ($ {\cal{N}} = 4 $ SYM) in the AdS/CFT correspondence. Relaxation time$ \tau_{\pi} $ is in fact approximately$ (2-\ln 2)/(2\pi T) $ [40-43], where s is the entropy,$ s^{\mu} $ is the entropy four-current, and T is the temperature. It is customary to split$ \pi^{\mu\nu} $ order-by-order in terms of$ \sigma^{\mu\nu} $ into a traceless part. The contribution from higher-order term is suppressed by the relaxation time$ \tau_{\pi} $ , and assuming transport coefficients$ \lambda_{1} = \lambda_{2} = \lambda_{3} = 0 $ [7, 45], and neglecting the contribution from higher order$ \tau_{\pi} $ terms, one can show that Eqs. (4)-(5) can be cast into,$ D\varepsilon = -(\varepsilon+p)\theta+2\eta \sigma^{\mu\nu}\sigma_{\mu\nu}-2 \eta\tau_{\pi}\sigma_{\mu\nu}\left[\Delta_{\alpha}^{\mu}\Delta_{\beta}^{\nu} D \sigma^{\alpha\beta} + \frac{1}{3}\sigma^{\mu\nu} \theta \right], $
(8) and
$ \begin{split} (\varepsilon+p)Du^{\alpha} =& \nabla^{\alpha}p+2\eta\left(\sigma^{\alpha\mu}-\tau_{\pi}\left[\Delta_{\gamma}^{\alpha}\Delta_{\rho}^{\mu} D \sigma^{\gamma\rho} + \frac{1}{3}\sigma^{\alpha\mu} \theta \right]\right)Du_{\mu}\\ & -2\eta\Delta^{\alpha\nu}\nabla^{\mu}\left(\sigma_{\mu\nu}-\tau_{\pi}\left[\Delta_{\mu}^{\gamma}\Delta_{\nu}^{\rho} D \sigma_{\gamma\rho} + \frac{1}{3}\sigma_{\mu\nu} \theta \right]\right). \end{split} $
(9) Here, the CKCJ solutions [22] and the perturbation solutions, both are characterized by the flow velocity field Eq. (1) in the Rindler coordinates. It is straightforward to find that co-moving derivative D and expansion rate
$ \theta $ can be expressed as [23],$ D = {\rm cosh}(\Omega-\eta_{s})\frac{\partial}{\partial\tau}+\frac{1}{\tau}{\rm sinh}(\Omega-\eta_{s})\frac{\partial}{\partial\eta_{s}}, $
(10) and
$ \theta = {\rm sinh}(\Omega-\eta_{s})\frac{\partial\Omega}{\partial\tau}+\frac{1}{\tau}{\rm cosh}(\Omega-\eta_{s})\frac{\partial\Omega}{\partial\eta_{s}}, $
(11) respectively.
With the help of Gibbs thermodynamic relation and CNC solutions [16] for systems without bulk viscosity and net charge current (net baryon, net electric charge or net strangeness), the hydrodynamic conservation equations (Eqs. (8), (9)) for a longitudinal expanding fireball in the presence of shear viscosity in the Rindler coordinate can be written as,
$ \begin{split} \tau\frac{\partial T}{\partial\tau}+\tanh(\Omega-\eta_{s})\frac{\partial T}{\partial\eta_{s}}+\frac{\Omega'}{\kappa}T =& \frac{\Pi_{d}}{\kappa}\frac{\Omega'^{2}}{\tau}\cosh(\Omega-\eta_{s})-\frac{\Pi_{d}\tau_{\pi}}{6\kappa\tau^{2}}\Omega'[-6\cosh(2(\Omega-\eta_{s}))\Omega'\\ & +(1+7\cosh(2(\Omega-\eta_{s})))\Omega'^{2}+ \sinh(2(\Omega-\eta_{s}))\Omega''], \end{split} $
(12) and
$ \begin{split} \tanh(\Omega-\eta_{s})\left[\tau\frac{\partial T}{\partial\tau}+T\Omega'\right]+\frac{\partial T}{\partial\eta_{s}} =& \frac{\Pi_{d}}{\tau}(2\Omega'(\Omega'-1)+\Omega''\coth(\Omega-\eta_{s}))\sinh(\Omega-\eta_{s})\\ & +\frac{\Pi_{d}\tau_{\pi}}{6\kappa\tau^{2}}(-\tanh(\Omega-\eta_{s})\Omega'(12+24\cosh(2(\Omega-\eta_{s}))\\& +\Omega'(-28 -46\cosh(2(\Omega-\eta_{s}))+3(5+7\cosh(2(\Omega-\eta_{s})))\Omega'))\\ & +(18\cosh(2(\Omega-\eta_{s}))+(1-23\cosh(2(\Omega-\eta_{s})))\Omega')\Omega''\\ &-6\Omega''-3\sinh(2(\Omega-\eta_{s}))\Omega^{(3)}), \end{split} $
(13) where
$ \Pi_{d} = \dfrac{4\eta}{3s} $ is related to the shear viscosity ratio,$ \Omega' $ approximately characterizes the longitudinal acceleration of flow element in the medium, and$ \Omega',\; \Omega'',\; \Omega^{(3)} $ are derivative function of flow rapidity with,$ \Omega' = \dfrac{\partial \Omega}{\partial \eta_{s}} $ ,$ \Omega'' = \dfrac{\partial^{2}\Omega}{\partial \eta_{s}^{2}} $ , and$ \Omega^{(3)} = \dfrac{\partial^{3}\Omega}{\partial \eta_{s}^{3}} $ .Case A. Bjorken solution, CNC solution, CKCJ solution under the velocity field
$ \Omega = \lambda\eta_{s} $ .A comprehensive study of the longitudinally expanding fireballs for ideal hydrodynamics was carried out by Csörgő, Nagy, and Csanád (or CNC family of solutions) in Refs. [16, 18]. According to the results from the CNC solutions, the fluid rapidity
$ \Omega(\tau, \eta_{s}) $ depends on the space-time rapidity$ \eta_{s} $ alone. For ideal hydrodynamics, one finds that$ \Omega'' = 0 $ ,$ \Omega^{(3)} = 0 $ ,$ \Pi_{d} = 0 $ and$ \tau_{\pi} = 0 $ , and the conservation equations, i.e. Eqs. (12), (13) reduce to,$ \tau\frac{\partial T}{\partial\tau}+\tanh(\Omega-\eta_{s})\frac{\partial T}{\partial\eta_{s}}+\frac{\Omega'}{\kappa}T = 0, $
(14) and
$ \tanh(\Omega-\eta_{s})\left[\tau\frac{\partial T}{\partial\tau}+T\Omega'\right]+\frac{\partial T}{\partial\eta_{s}} = 0. $
(15) (a) For boost-invariant Hwa-Bjorken flow where
$ \Omega = \eta_s $ ,$ \tanh(\Omega-\eta_{s}) = 0 $ , and Eqs. (14, 15) have following exact solution,$ T(\tau) = T_{0}\left(\frac{\tau_{0}}{\tau}\right)^{\textstyle\frac{1}{\kappa}}, $
(16) where
$ T_{0} $ defines the values for temperature at the proper time$ \tau_{0} $ and coordinate rapidity$ \eta_{s} = 0 $ . This is the famous boost-invariant Hwa-Bjorken solution [12, 13] and other hydrodynamics variables are functions of the proper time$ \tau $ .(b) For a perfect fluid with longitudinal acceleration,
$ \Omega(\eta_{s}) $ , shear viscosity$ \Pi_{d} = 0 $ . Eqs. (14), (15) reduce to,$ \begin{split} &\tanh(\Omega-\eta_{s})\left[\frac{\tau\partial T}{T\partial\tau}+\Omega'\right]+\frac{\partial T}{T\partial\eta_{s}} = 0,\\ \Rightarrow & \frac{\partial\ln T}{\partial\ln\tau} = -\frac{1-\kappa\tanh^{2}(\Omega-\eta_{s})}{1-\tanh^{2}(\Omega-\eta_{s})}\frac{\Omega'}{\kappa}. \end{split} $
(17) For
$ \kappa = 1 $ , one finds that the$ \Omega' $ could be either an arbitrary constant [16] or a function of$ \eta_{s} $ [22, 39]. For the former case, one finds that$ \Omega(\eta_{s}) = \lambda\eta_{s}+\eta_{0} $ , here$ \lambda $ is an arbitrary constant and$ \lambda = \Omega' $ ,$ \eta_{0} $ is the space-time rapidity shift. The solution of Eq. (17) case is$ \frac{\partial\ln T}{\partial\ln\tau} = -\Omega', \; \; \; \Rightarrow\; \; \; T(\tau) = T_{0}\left(\frac{\tau_{0}}{\tau} \right)^{\lambda}\frac{1}{{\cal{V}}(S)}, $
(18) this is the well-known CNC exact solutions case (e) that presented by Csörgő, Nagy, and Csanád in Refs. [16, 18]. Here
$ {{\cal{V}}(S)}$ is an arbitrary function of scaling function S, where S can be obtained from the scaling function definition$ u^{\mu}\partial_{\mu}S = 0 $ . The properties and detailed discussion of this S can be found in Ref. [18] for CNC exact solutions and in Ref. [46] for Hubble-type viscous flow.(c) A finite, accelerating, and realistic 1+1 dimensional solution of relativistic hydrodynamics was recently given by Csörgő, Kasza, Csanád and Jiang (CKCJ) with the condition
$ H(\eta_{s}) = \Omega(\eta_{s})-\eta_{s} $ . For details, see review in Refs. [22, 39].Case B. Perturbation solution with Navier-Stokes approximation.
For a relativistic hydrodynamic in the Navier-Stokes (first-order) approximation, fluid flow rapidity
$\Omega = $ $ \lambda\eta_{s} = (1+\lambda^{*})\eta_{s} $ , shear viscosity tensor$ \pi^{\mu\nu} = 2\eta\sigma^{\mu\nu} $ , shear viscosity ratio$ \Pi_{d} = 4\eta/3s $ , the relaxation time$ \tau_{\pi} = 0 $ . The last terms in the right side of Eqs. (12), (13) disappear automatically as follow,$ \tau\frac{\partial T}{\partial\tau}+\tanh(\lambda^{*}\eta_{s})\frac{\partial T}{\partial\eta_{s}}+\frac{1+\lambda^{*}}{\kappa}T = \frac{\Pi_{d}}{\kappa}\frac{1+2\lambda^{*}+\lambda^{*2}}{\tau}\cosh(\lambda^{*}\eta_{s}), $
(19) and
$ \tanh(\lambda^{*}\eta_{s})\left[\tau\frac{\partial T}{\partial\tau}\!+\!T(1\!+\!\lambda^{*})\right]\!+\!\frac{\partial T}{\partial\eta_{s}} \!=\! \frac{\Pi_{d}}{\tau}(2(1\!+\!\lambda^{*})\lambda^{*})\sinh(\lambda^{*}\eta_{s}). $
(20) However, the reduced conservation equations i.e., Eqs. (19), (20), including the first-order approximation are still a set of nonlinear differential equations, which are notoriously hard to solve analytically. Fortunately, based on the results from the ideal hydro [21, 38], we found that the longitudinal acceleration parameter
$ \lambda^{*} $ extracted from the experimental data is pretty small ($ 0<\lambda^{*}\ll1 $ ), which results in a simple antsaz or perturbation solution here.We assume the longitudinal rapidity perturbation
$ \lambda^{*}\eta_{s} $ is a pretty small numbers here. Using the Taylor series expansion$ \tanh(\lambda^{*}\eta_{s})\approx\lambda^{*}\eta_{s}-\dfrac{(\lambda^{*}\eta_{s})^{3}}{3} $ ,$ \cosh(\lambda^{*}\eta_{s})\approx1+ \dfrac{(\lambda^{*}\eta_{s})^{2}}{2} $ and$ \sinh(\lambda^{*}\eta_{s})\approx\lambda^{*}\eta_{s}+\dfrac{(\lambda^{*}\eta_{s})^{3}}{6} $ , up to the leading order$ {\cal{O}}(\lambda^{*}) $ , Eqs. (19), (20) yields a partial differential equation depending on$ \tau $ only,$ \tau\frac{\partial T}{\partial\tau}+\frac{(1+\lambda^{*})T}{\kappa} = \frac{\Pi_{d}}{\kappa}\frac{1+2\lambda^{*}}{\tau}, $
(21) and the exact temperature solution
$ T(\tau,\eta_{s}) $ of above equation is$ \begin{split} T(\tau,\eta_{s}) = & T_{1}(\eta_{s})\left(\frac{\tau_{0}}{\tau}\right)^{\textstyle\frac{1+\lambda^{*}}{\kappa}}+\frac{(2\lambda^{*}+1)\Pi_{d}}{(\kappa-1)\tau_{0}}\left(\frac{\tau_{0}}{\tau}\right)^{\textstyle\frac{1+\lambda^{*}}{\kappa}} \\ & \times \left[1-\left(\frac{\tau_{0}}{\tau}\right)^{\textstyle 1-\frac{1+\lambda^{*}}{\kappa}}\right], \end{split} $
(22) where
$ \tau_{0} $ is the value of proper time,$ T_{1}(\eta_{s}) $ is an unknown function constrained by Eq. (20).Putting Eq. (22) back to the Euler equation Eq. (20), up to the leading order
$ {\cal{O}}(\lambda^{*}) $ , one gets the exact expression of$ T_{1}(\eta_{s}) $ as follow,$ T_{1}(\eta_{s})\! =\! T_{0}\exp\left[\!-\!\frac{1}{2}\lambda^{*}\left(1\!-\!\frac{1}{\kappa}\right)\eta_{s}^{2}\right] \!-\!\frac{\left(1\!\!-\!\!\exp\left[\!-\dfrac{1}{2}\lambda^{*}\left(1\!-\!\dfrac{1}{\kappa}\right)\eta_{s}^{2}\right] \right)\Pi_{d}}{(\kappa-1)\tau_{0}}, $
(23) where
$ T_{0} $ defines the value of temperature at the proper time$ \tau_{0} $ and coordinate rapidity$ \eta_{s} = 0 $ . Here if one places the perturbation solution Eq. (22) back to the energy conservation equation Eq. (19), up to the leading order$ {\cal{O}}(\lambda^{*}) $ , one can obtain the same results.Finally, substituting Eq. (23) into Eq. (22), the perturbation solution of the
$ 1+1 $ D embeding$ 1+3 $ D relativistic viscous hydrodynamics can be written as,$\begin{split} T(\tau,\eta_{s}) =& T_{0}\left(\frac{\tau_{0}}{\tau}\right)^{\textstyle\frac{1+\lambda^{*}}{\kappa}} {\bigg [}\exp\left(-\frac{1}{2}\lambda^{*}\left(1-\frac{1}{\kappa}\right)\eta_{s}^{2}\right)\\&+\frac{R_{0}^{-1}}{\kappa-1}{\bigg (}2\lambda^{*}+\exp\left[-\frac{1}{2}\lambda^{*}\left(1-\frac{1}{\kappa}\right)\eta_{s}^{2}\right]\\&-(2\lambda^{*}+1) \left(\frac{\tau_{0}}{\tau}\right)^{\textstyle\frac{\kappa-\lambda^{*}-1}{\kappa}} {\bigg )} {\bigg ]}, \end{split}$
(24) where the Reynolds number is
$ R^{-1}_{0} = \dfrac{\Pi_{d}}{T_{0}\tau_{0}} $ [7, 47].From the perspective of the perturbation solution's structure, the above conditional perturbation solution is very nontrivial because it is only satisfied when the longitudinal accelerating parameter
$ \lambda^{*}\ll1 $ and$ \eta_{s} $ is not very large ($ |\eta_{s}|\ll 5 $ ); furthermore, it involves two different transport coefficients and many nonvanishing components of the longitudinal expanding properties. To investigate the stability of perturbation solution Eq. (24), we numerically solved the energy equation Eq. (19) and Euler equation Eq. (20) with the conditions$ \Pi_{d} = 0 $ and$ \tau_{\pi} = 0 $ , longitudinal accelerating parameter$ \lambda^{*} = 0.05 $ , grid length of proper time$ \Delta\tau = 0.05 $ , grid of space-time rapidity$ \Delta\eta_{s} = 0.05 $ , and range space-time rapidity$ \eta_{s} $ from 0.0 to 5.0. The perturbation solution and the numerical solution are compared in Fig. 1 left panel. The difference between the above two solutions appears to be small, which implies that the perturbation solution is special but a stable one.Figure 1. (color online) Temperature profile in the Navier-Stokes approximation for different longitudinal acceleration parameters
$ \lambda^{*} $ . Equation of state$ \varepsilon = 3p $ , shear viscosity ratio$ \eta/s = 1/4\pi $ . Black solid curve is the ideal Bjorken flow for reference, blue solid curve is the 1st order Bjorken flow. Left panel: The proper time$ \tau $ evolution with temperature for$ \eta_{s} = 0 $ . Right panel: space rapidity$ \eta_{s} $ evolution with temperature for$ \tau = 2 $ fm/c. Purple curve shows the comparison between the perturbation solution (dashed) and the numerical solution (solid) for ideal flow with$ \lambda^{*} = 0.05 $ , the accuracy is acceptable in the range$ 0.0\leqslant\eta_{s}\leqslant 5.0 $ . Results from the perturbation solution Eq. (24) and numerical solution for Eqs. (12), (13).The profile of Eq. (24) is a (1+1) dimensional scaling solution in (1+3) dimensions, and the
$ \eta_{s} $ dependence of temperature density is of the Gaussian form, see Fig. 1 right panel. Such perturbation solutions implies that for a non-vanishing longitudinal acceleration parameter$ \lambda^{*} $ , the cooling rate is larger than that for the ideal case. Meanwhile, a non-zero shear viscosity$ \eta $ makes the cooling rate smaller than the ideal case [38], see Fig. 1 left panel. Note that, when$ \lambda^{*} = 0 $ and$ R_{0}^{-1} = 0 $ , one obtains the same solutions as same as the ideal hydrodynamic Bjorken solution [13]; when$ \lambda^{*} = 0 $ and$ R_{0}^{-1}\neq0 $ , one obtains the first-order Bjorken solutions [7, 45]; if$ \lambda^{*}\neq0 $ and$ R_{0}^{-1} = 0 $ , one obtains a special solution which is consistent with the CNC solutions' case (e) in [16, 18], and when Eqs. (14), (15) are solved directly with$ R_{0}^{-1} = 0 $ and$ \lambda^{*}\neq0 $ , one obtains the CKCJ solutions [22].Case C. Perturbation equations with Israel-Stewart approximation.
The temperature profile from Eq. (24) shows a peak at earlier proper time
$ \tau $ in the Navier Stokes approximation, see Fig. 1. The source of this acausality can be understood from the constitutive relations satisfied by the dissipative currents$ \pi^{\mu\nu} = 2\eta\sigma^{\mu\nu} $ . The linear relationship between the dissipative currents and the gradients of the primary fluid-dynamical variables imply that any inhomogenity of$ u^{\mu} $ , immediately results in dissipative currents. This instantaneous effect causes the first-order theory to be unstable at earlier times.The Israel-Stewart (second-order) approximation was found to be suitable for describing the physical process happening at earlier times; it describes the counteract of the acceleration effect and viscosity effect well. However, it is difficult to solve the the differential equations Eqs. (12), (13) analytically with the Israel-Stewart approximation. Therefore, we numerically solve the temperature time dependence Eq. (12) first at
$ \eta_{s} = 0.0 $ with the initial condition$ T(0.2, 0.0) = 0.65 $ GeV; here the grid length of$ \tau $ is$ \Delta\tau = 0.05 $ fm. Then, for each$ \eta_{s} $ , we solve the temperature rapidity dependence Eq. (13) step by step with the results from the Eq. (12), and solve these equations together; the grid length of$ \tau $ is$ \Delta\tau = 0.05 $ fm. The temperature distribution of the thermodynamic quantities ($ \varepsilon,\; T,\; p $ ) in whole$ (\tau,\; \eta_{s}) $ coordinates with initial condition$ T(\tau_{0},\; \eta_{s0}) $ is a Gaussian shape, see Fig. 2. Furthermore, to compare with the perturbation results from the first-order approximation, Eqs. (12), (13) can be rewritten up to the leading order$ {\cal{O}}(\lambda^{*}) $ as follows,Figure 2. (color online) Temperature profile in the Israel-Stewart approximation. Left panel: proper time
$ \tau$ evolution of temperature for$ \eta_{s}=0$ . Black solid curve is the ideal Bjorken flow for reference. Right panel: The space-time evolution of temperature in ($ \tau-\eta_{s}$ ) coordinates, the longitudinal acceleration parameter$ \lambda^{*}=0.1$ . Numerical results from Eqs. (14), (15).$ \tau\frac{\partial T}{\partial\tau} = \frac{(2\lambda^{*}+1)\Pi_{d}}{3\tau}-\frac{(\lambda^{*}+1)T}{3}-\frac{(2-\ln 2)\Pi_{d}(1+6\lambda^{*})} {9 \pi T \tau^{2}}, $
(25) $ \frac{\partial T}{\partial\eta_{s}} = \lambda^{*}\eta_{s}\left[-\frac{2T}{3}-\frac{\Pi_{d}}{3\tau}+\frac{(2-\ln 2)\Pi_{d}}{9\pi T \tau^{2}} \right]. $
(26) The above differential equations (25), (26) cannot be solved analytically. Using the same numerical method as for the Eqs. (14), (15), we solve the above second-order viscous hydrodynamic equations (25), (26) with the conformal equation of state
$ \varepsilon\; = \; 3p $ and relaxation time$ \tau_{\pi} = \dfrac{2-\ln2}{2\pi T} $ [40-43] directly in the Rindler coordinates. The numerical results are presented in Fig. 3.Figure 3. (color online) Proper time evolution of temperature density for given primary initial conditions. Left panel: Perturbation results of temperature profile in the Navier-Stokes approximation (1st) and in the Israel-Stewart theory (2nd). The longitudinal acceleration parameter
$ \lambda^{*}=0.10$ . Black solid curve is the ideal Bjorken flow for reference. Right panel: Temperature profile comparsion between the completely numerical solution (solid curve) with perturbation solutions (dashed curve) in the Israel-Stewart theory for different$ \lambda^{*}$ , the grid of$ \tau$ is$ \Delta\tau=0.05$ in the numerical code.
Perturbation solutions of relativistic viscous hydrodynamics for longitudinally expanding fireballs
- Received Date: 2020-01-29
- Accepted Date: 2020-04-26
- Available Online: 2020-08-01
Abstract: The solutions of the relativistic viscous hydrodynamics for longitudinally expanding fireballs are investigated with the Navier-Stokes theory and Israel-Stewart theory. The energy and the Euler conservation equations for the viscous fluid are derived in Rindler coordinates, by assuming that the longitudinal expansion effect is small. Under the perturbation assumption, an analytical perturbation solution for the Navier-Stokes approximation and numerical solutions for the Israel-Stewart approximation are presented. The temperature evolution with both shear viscous effect and longitudinal acceleration effect in the longitudinal expanding framework are presented. The specific temperature profile shows symmetric Gaussian shape in the Rindler coordinates. Further, we compare the results from the Israel-Stewart approximation with the results from the Bjorken and the Navier-Stokes approximations, in the presence of the longitudinal acceleration expansion effect. We found that the Israel-Stewart approximation gives a good description of the early stage evolutions than the Navier-Stokes theory.