betaFeedback

WorksPhysics and Astronomy

Hydrodynamic Gradient Expansion Diverges beyond Bjorken Flow

Abstract

The gradient expansion is the fundamental organising principle underlying relativistic hydrodynamics, yet understanding its convergence properties for general nonlinear flows has posed a major challenge. We introduce a simple method to address this question in a class of fluids modelled by Israel-Stewart–type relaxation equations. We apply it to (1+1)-dimensional flows and provide numerical evidence for factorially divergent gradient expansions. This generalises results previously only obtained for (0+1)-dimensional comoving flows, notably Bjorken flow. We also demonstrate that the only known nontrivial case of a convergent hydrodynamic gradient expansion at the nonlinear level relies on Bjorken flow symmetries and becomes factorially divergent as soon as these are relaxed. Finally, we show that factorial divergence can be removed using a momentum space cutoff, which generalises a result obtained earlier in the context of linear response.

Introduction

Hydrodynamics plays a pivotal role in the description of nonequilibrium phenomena, with applications ranging from condensed matter systems [1] to scenarios in astrophysics [2, 3, 4] or nuclear physics [5, 6]. The reason is that hydrodynamics captures the infrared behavior of any medium endowed with conserved quantities. For a given set of conserved currents, the expression of hydrodynamic behaviour rests on the derivative expansion in the spirit of an effective field theory [7, 8, 9, 10, 11]. For a neutral relativistic fluid, the natural choice of dynamical variables are the energy density E(x)\mathcal{E}(x) and the unit-normalized fluid velocity U(x)=Uμ(x)∂μU(x) = U^\mu(x)\partial_\mu, with the conserved currents, TμνT^{\mu\nu}, given by the constitutive relation

Tμν=EUμUν+P(E)(gμν+UμUν)+Πμν.(1)T^{\mu\nu} = \mathcal{E} U^\mu U^\nu+ P(\mathcal{E})(g^{\mu\nu} + U^\mu U^\nu) + \Pi^{\mu\nu}. \tag*{(1)}

Here, the first two terms describe ideal flow with gg being the Minkowski metric and Πμν\Pi^{\mu\nu} captures dissipative effects organised as

Πμν=∑n=1∞ϵnΠμν(n)[E,U],(2)\Pi_{\mu\nu} = \sum_{n=1}^{\infty} \epsilon^n \Pi_{\mu\nu}^{(n)}[\mathcal{E},U], \tag*{(2)}

where Πμν(n)\Pi_{\mu\nu}^{(n)} contains nn spacetime derivatives of E,U\mathcal{E},U and we have introduced ϵ\epsilon as a formal derivative-counting parameter. The gradient expansion in Eq. (2) is defined up to redundancies associated with frame choice and current conservation ∇μTμν=0\nabla_\mu T^{\mu\nu}=0.

Understanding the character of the expansion (2) constitutes a fundamental open problem. Is it convergent, in such a way that subsequent truncations are progressively more accurate? If not, how does its divergent nature relate to the empirical success of low-order truncations?

Studies of comoving flows in Refs. [12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31], in which all fluid flow lines can be mapped to each other under symmetry transformations, rendering the problem effectively (0+1)-dimensional, have been instrumental in advancing our understanding of the hydrodynamic expansion (2). Among these, Bjorken flow [32] in conformally invariant theories is the most thoroughly explored example due to its role in studies of quark-gluon plasma. In these cases a particular strategy to solve the dynamical equations is an expansion in the Knudsen number, 1/w1/w [33]. It is possible to compute a sufficient number of terms to assess that these expansions are factorially divergent. The expansion in ϵ\epsilon defined in Eq. (2) encapsulates the expansion in 1/w1/w, as we review in the Supplemental Material. Another well-studied example of a comoving flow is the Gubser flow which is reached by Weyl transformation from a (0 + 1)-flow on dS3×R\mathrm{dS}_3 \times\mathbb{R} [34, 35].

Outside the realm of comoving flows, the only generic result on (2) was restricted to the linear response regime [36]. It showed that depending on how the momentum space support kmax⁡k_{\max} of E\mathcal{E} and UμU^\mu compares to an intrinsic scale of the underlying microscopic theory k∗k_* [37], the gradient-expanded constitutive relations could either be convergent (kmax⁡<k∗k_{\max}<k_*), geometrically divergent (k∗<kmax⁡<∞k_*<k_{\max}<\infty) or factorially divergent (kmax⁡→∞k_{\max}\to\infty).

In this Letter, we break free both from the symmetry constraints of comoving flows and the conveniences of linearisation to address, for the first time, the large-order behavior of the hydrodynamic gradient expansion beyond comoving flows at the fully nonlinear level. Specifically, we ask the following question: Given a generic nonequilibrium, nonlinear configuration of E,Uμ\mathcal{E}, U^\mu arising on-shell, what is nature of the expansion in ϵ\epsilon (2) when evaluated on this solution? We answer this question by introducing a simple method that allows one to calculate Eq. (2) up to high order on a desktop computer. In this Letter we illustrate it in two examples. The first is the model put forward by Baier, Romatschke, Son, Starinets and Stephanov (BRSSS) in Ref. [38], while the second is the model originally introduced by Denicol and Noronha (DN) in Ref. [27]. Both theories are representative examples within a broad class of models employing the Israel-Stewart approach to embed hydrodynamics in a framework compatible with relativistic causality [39, 40]. Our method applies to any member of this class, such as [41], and covers also equations with more than one derivative of Πμν\Pi^{\mu\nu}, such as [42, 43, 44]. Each member of this class generally gives rise to infinitely many transport coefficients in the gradient expansion (2) that are specific to it.

Results in the BRSSS model— In this Letter, we consider the restriction of the BRSSS model to conformal fluids in dd spacetime dimensions (d=4d=4 in our numerics). We work in the Landau frame and take ΠμνUν=0\Pi_{\mu\nu}U^\nu=0. The model is defined by promoting Πμν\Pi_{\mu\nu} to a set of independent degrees of freedom subject to a relaxation equation,

Πμν=−2ηD⟨μuν⟩−τΠUαDαΠμν+λ1η2Πλ⟨μΠν⟩λ,(3)\Pi^{\mu\nu}=-2\eta D^{\langle\mu}u^{\nu\rangle}-\tau_{\Pi}U^\alpha D_\alpha\Pi^{\mu\nu}+\frac{\lambda_1}{\eta^2}\Pi_\lambda^{\langle\mu}\Pi^{\nu\rangle\lambda}, \tag*{(3)}

where we neglected terms not relevant to the flow we consider. DμD_\mu is the Weyl-covariant derivative [45], and the angle brackets instruct one to take the symmetric, transverse and traceless part of the tensor they act upon. The relaxation time τΠ\tau_\Pi, the shear viscosity η\eta, and λ1\lambda_1 are transport coefficients. Conformal invariance demands that these quantities depend on the local temperature TT, defined by the relation E=E0Td\mathcal{E}=\mathcal{E}_0T^d, as

η=Cη4E3T,τΠ=CτΠT,λ1=Cλ1ηT(4)\eta=C_\eta\frac{4\mathcal{E}}{3T},\qquad\tau_\Pi=\frac{C_{\tau_\Pi}}{T},\qquad\lambda_1=C_{\lambda_1}\frac{\eta}{T} \tag*{(4)}

where CηC_\eta and CτΠ>0C_{\tau_\Pi}>0. The equations of motion of BRSSS theory are given by (3) and the conservation equation ∇μTμν=0\nabla_\mu T^{\mu\nu}=0, where the energy-momentum tensor is specified in terms of E\mathcal{E}, UU and Πμν\Pi_{\mu\nu} as in (1).

The fluid flows we focus on are characterised as follows. We separate the spatial coordinates into one longitudinal direction, xx, and d−2d-2 transverse directions, x1,…,xd−2x_1,\ldots,x_{d-2}, demanding isotropy and translational invariance in the transverse hyperplane spanned by xix_i. Hence, the nontrivial dynamics is confined to the longitudinal plane spanned by tt and xx, and our fluid flows are (1+1)(1+1)-dimensional. We refer to these fluid flows as longitudinal. At the linearized level such flows would correspond to sound wave propagation. See, e.g., Ref. [46] for a study of longitudinal flows in a quark-gluon plasma context.[47]

The most general fluid velocity for a longitudinal flow is parameterized by a single degree of freedom, uu, as

U=Uμ∂μ=cosh⁡u ∂t+sinh⁡u ∂x.(5)U=U^\mu\partial_\mu=\cosh u\,\partial_t+\sinh u\,\partial_x. \tag*{(5)}

Furthermore, any tensor which is symmetric, transverse to UμU^\mu and traceless is described in terms of a single additional degree of freedom that we pick as

Πμν=(2−d)Π⋆Σμν(6)\Pi^{\mu\nu}=(2-d)\Pi_{\star}\Sigma^{\mu\nu} \tag*{(6)}

where Σμν≡gμν+UμUν−d−1d−2PTμν\Sigma^{\mu\nu}\equiv g^{\mu\nu}+U^\mu U^\nu-\frac{d-1}{d-2}P_T^{\mu\nu} and Π⋆=1d−2PTμνΠμν\Pi_\star=\frac{1}{d-2}P_T^{\mu\nu}\Pi_{\mu\nu}, with PTμνP_T^{\mu\nu} being the projector in the transverse directions.

We now consider the expansion (2) applied to the conformal BRSSS model for longitudinal flows. This is facilitated by a numerical algorithm which makes a computation of (2) to large orders tractable. Since ϵ\epsilon counts derivatives it can be introduced by taking (3) and replacing ∇μ→ϵ∇μ\nabla_\mu\to\epsilon\nabla_\mu together with positing a perturbative ansatz for Π⋆\Pi_\star, as follows,

Dα→ϵDα,Π⋆→∑n=1∞Π⋆(n)ϵn.(7)D_\alpha\to\epsilon D_\alpha,\qquad\Pi_\star\to\sum_{n=1}^{\infty}\Pi_\star^{(n)}\epsilon^n. \tag*{(7)}

This leads to the following recursion relation

Π⋆(1)=−2d−2ηPTμνD⟨μUν⟩,(8)\Pi_\star^{(1)}=-\frac{2}{d-2}\eta P_T^{\mu\nu}D_{\langle\mu}U_{\nu\rangle}, \tag*{(8)}
Π⋆(n+1)=−τΠ(U⋅∂)Π⋆(n)−d(∂⋅U)d−1τΠΠ⋆(n)−(d−3)λ1η2∑m=1nΠ⋆(m)Π⋆(n+1−m),n>1.(9)\Pi_\star^{(n+1)}=-\tau_\Pi(U\cdot\partial)\Pi_\star^{(n)}-\frac{d(\partial\cdot U)}{d-1}\tau_\Pi\Pi_\star^{(n)}-\frac{(d-3)\lambda_1}{\eta^2}\sum_{m=1}^{n}\Pi_\star^{(m)}\Pi_\star^{(n+1-m)},\qquad n>1. \tag*{(9)}

Here, E\mathcal{E} and UU are not expanded in ϵ\epsilon. Therefore to proceed to evaluate (8) we must first find E\mathcal{E} and UU for a given choice of flow. These (as well as the exact Π⋆\Pi_\star) can be obtained by numerically solving the BRSSS equations of motion as an initial value problem without invoking an ϵ\epsilon expansion. Once E\mathcal{E} and UU are known, the recursion relation (8) can be efficiently evaluated numerically to high orders. Careful consideration of resolution and precision is required, as this procedure involves high numbers of successive derivatives of the background solution E\mathcal{E} and UU. This is further discussed in the Supplemental Material, where we show that our numerical results are convergent. Our approach applies to the whole class of theories which build on the Israel-Stewart approach.

Note that solving the recursion relation (8) analytically is prohibitively expensive due to the fast growth in the number of terms contributing at each order. In particular, for the case λ1=0\lambda_1=0 we observed an exponential growth of the number of individual contributions at each order. Our method circumvents this difficulty.

We have applied our approach in the BRSSS model across a wide variety of initial conditions, transport coefficients and spacetime points. We find factorial growth in all cases considered. We illustrate this in Figure 1 with one representative example in which we consider a strong Gaussian overdensity for our initial conditions and adopt a periodic compactification of the spatial direction. The left panel shows TT and flow lines of UU and highlights three spacetime point samples. The right panel root plot demonstrates factorial growth at these sampled points. Further details are provided in the Supplemental Material.

Image not extracted

Figure 1. Left panel: Solution to an initial value problem in BRSSS. Colour shows the temperature profile T(t,x)T(t,x), and the black solid lines are flow lines of velocity UU. Here t,xt,x are obtained from standard Minkowski coordinates by periodically identifying x∼x+1x\sim x+1. Initial conditions were provided at t=0t=0 corresponding to a strong Gaussian overdensity. The colour scale ranges from min⁡T=0.76\min T=0.76 (violet) to max⁡T=1.76\max T=1.76 (red). Also included are three marked points (disk, plus, open diamond). Right panel: Root plot for the hydrodynamic expansion of the constitutive relation (2) for BRSSS, evaluated using the recursion relation (8) for the solution shown in the left panel, at each of the marked points with matching marker shapes. To guide the eye, in red we show straight lines fit to the range n∈[20,130]n\in[20,130], which correspond to Π∗(n)∼Γ(n)\Pi_*^{(n)}\sim\Gamma(n) at large nn. Numerical convergence of our results is demonstrated in the Supplemental Material.

Momentum cutoff– In previous work [36] we showed that a momentum-space cutoff gives at most a geometrically divergent hydrodynamic expansion for linear deviations from equilibrium. This result naturally extends to strongly nonlinear scenarios, as we now demonstrate. So far in this Letter we have used a numerical grid simply as a tool to approximate the continuum, but we now push beyond the continuum picture and re-evaluate the grid in a new role as a physical lattice which naturally enforces a momentum-space cutoff. In the BRSSS model, when λ1=0\lambda_1=0 the recursion relation (8) can be written as

Π∗(n+1)=MΠ∗(n)n>1(10)\Pi_*^{(n+1)}=\mathcal{M}\Pi_*^{(n)} \qquad n>1 \tag*{(10)}

where M=−τΠ(U⋅∂)−d(∂⋅U)d−1\mathcal{M}=-\tau_\Pi(U\cdot\partial)-\frac{d(\partial\cdot U)}{d-1} is a differential operator independent of nn, depending only on the background solution E,U\mathcal{E},U. For a grid of dimensions Nx×NtN_x\times N_t, each Π∗(j)\Pi_*^{(j)} can be written as a NxNtN_xN_t-sized vector, and M\mathcal{M} accordingly as a NxNt×NxNtN_xN_t\times N_xN_t square matrix. Thus, on a lattice the expansion is ultimately only geometrically growing at a rate set by the largest eigenvalue of M\mathcal{M}, which scales with the inverse lattice spacing. In Figure 2 this is demonstrated by utilising a deliberately low resolution lattice to allow for evaluating the hydrodynamic expansion to order n=8000n=8000. It shows the transition from factorial growth where the continuum approximation holds, to the geometrically divergent asymptotic behaviour governed by the aforementioned eigenvalue. We have also verified numerically that this result holds at λ1≠0\lambda_1\ne0, with the definition of M\mathcal{M} as given.

Image not extracted

Figure 2. Working on a lattice gives a window of factorial growth–where it successfully approximates the continuum–before yielding to a geometrically growing hydrodynamic expansion regulated by the lattice. The asymptotic value reached is governed by the lattice spacing as discussed in the text. Black disks are those from Figure 1 corresponding to the disk spacetime marker point. Red circles are the same simulation and spacetime point, but on a coarse numerical grid and evaluated to hydrodynamic order n=8000n = 8000. This plot serves also as an indication of a convergence of our approach.

Resolving the DN model tension

Bjorken flow is a boost-invariant longitudinal flow such that the dynamics depends on the proper time τ=t2−x2\tau=\sqrt{t^2-x^2} only. In Ref. [27] the authors analysed a Knudsen number expansion for an ultrarelativistic gas of hard spheres undergoing Bjorken flow. While in all the other models such expansions have been observed to be factorially divergent, in Ref. [27] the terms grow only geometrically, with convergence ensuing for a Knudsen number smaller than a critical value. Our objective is to re-analyse this physical scenario using the expansion in (2). For Bjorken flow we find analogous results, namely geometric growth; however, our method allows us to explore what happens when these symmetries are relaxed.

We work in d=4d = 4. As in the conformal BRSSS model, the energy-momentum tensor in the DN model is traceless and decomposed as in Eq. (1), with Πμν\Pi_{\mu\nu} still obeying Eq. (3) with λ1=0\lambda_1 = 0. Hence, the recursion relations giving Π⋆(n)\Pi^{(n)}_\star still take the form (8), again with zero λ1\lambda_1. The differences start with the inclusion of a conserved current Jμ=ρUμJ^\mu= \rho U^\mu, where ρ\rho is the particle density. Furthermore, τΠ\tau_\Pi and η\eta are not fixed purely in terms of the local temperature TT, but rather obey

E=3ρT,η=aσTT,τΠ=ab4σT1ρ,(11)\mathcal{E} = 3\rho T,\qquad\eta= \frac{a}{\sigma_T}T,\qquad\tau_\Pi= \frac{ab}{4\sigma_T}\frac{1}{\rho}, \tag*{(11)}

where σT\sigma_T is the total cross section and aa, bb are positive dimensionless constants.

For Bjorken flow, the conservation of the particle current JμJ^\mu entails that the particle density ρ\rho decouples from the energy-momentum tensor. One has that

ρ(τ)=ρ0τ0τ,(12)\rho(\tau) = \frac{\rho_0\tau_0}{\tau}, \tag*{(12)}

where ρ0=ρ(τ0)\rho_0 = \rho(\tau_0) is the initial particle density. Hence,

τΠ=14ab Kn τ,(13)\tau_\Pi= \frac{1}{4}ab\,\mathrm{Kn}\,\tau, \tag*{(13)}

where Kn=1/(ρ0τ0σT)\mathrm{Kn} = 1/(\rho_0\tau_0\sigma_T) is the Knudsen number. In the DN model for Bjorken flow it is time independent.

To assess the large-order behavior of the expansion in ϵ\epsilon, Eq. (2), we first note that one can find a closed-form expression for Π⋆(n)\Pi^{(n)}_\star,

Π⋆(n)(τ)=23a Kn ρ0(τ0τ)43(−abKn4)n−1×(τ∂τ)n−1((ττ0)13T(τ)),(14)\Pi^{(n)}_\star(\tau) = \frac{2}{3}a\,\mathrm{Kn}\,\rho_0\left(\frac{\tau_0}{\tau}\right)^{\frac{4}{3}}\left(-\frac{ab\mathrm{Kn}}{4}\right)^{n-1}\times\left(\tau\partial_\tau\right)^{n-1}\left(\left(\frac{\tau}{\tau_0}\right)^{\frac{1}{3}}T(\tau)\right), \tag*{(14)}

a fact that relies crucially on Eq. (13). Second, we recall that TT can also be determined exactly [27]

T(τ)=T0,+(τ0τ)α++T0,−(τ0τ)α−,(15)T(\tau) = T_{0,+}\left(\frac{\tau_0}{\tau}\right)^{\alpha_+}+T_{0,-}\left(\frac{\tau_0}{\tau}\right)^{\alpha_-}, \tag*{(15)}

where T0,±T_{0,\pm} depend on initial conditions and α±\alpha_\pm on aa, bb, and Kn\mathrm{Kn}. Together, Eqs. (14) and (15) entail that Π⋆(n)\Pi^{(n)}_\star cannot grow factorially with nn at fixed τ\tau, since the repeated action of the differential operator τ∂τ\tau\partial_\tau on terms of the form (τ/τ0)13−α±(\tau/\tau_0)^{\frac{1}{3}-\alpha_\pm} only grows geometrically. Furthermore, Eqs. (14) and (15) also imply that Π⋆(1)\Pi^{(1)}_\star (and therefore all Π⋆(n>1)\Pi^{(n>1)}_\star) is a linear combination of eigenfunctions of the differential operator M\mathcal{M} defined as in Eq. (10), providing another perspective on why the gradient expansion grows geometrically in this case. We refer the reader to the Supplemental Material for further details.

The analysis above relies crucially on the symmetry restrictions of Bjorken flow. Empirically, when relaxing these symmetry restrictions in all cases studied we find that the large-order geometric growth is destroyed and the factorial divergence is restored. We illustrate this in Figure 3 for a longitudinal flow corresponding to a small perturbation of Bjorken flow.

Image not extracted

Figure 3. Deviations from Bjorken symmetries restores factorial growth in the DN model. Open circles represent ∣Π∗(n)∣1n\lvert\Pi_*^{(n)}\rvert^{\frac{1}{n}} for Bjorken flow in the DN model evaluated at τ=1\tau=1, where we have chosen initial conditions given by ρ(τ0)=4\rho(\tau_0)=4, T(τ0)=2T(\tau_0)=2 and Π∗(τ0)/(ρ(τ0)T(τ0))=0.436461\Pi_*(\tau_0)/(\rho(\tau_0)T(\tau_0))=0.436461 at τ0=0.1\tau_0=0.1. Crosses represent ∣Π∗(n)∣1n\lvert\Pi_*^{(n)}\rvert^{\frac{1}{n}} at τ=1\tau=1 and zero rapidity for a longitudinal flow in the DN model which obeys the same initial conditions as Bjorken flow, but with a rapidity-even Gaussian overdensity in ρ\rho of amplitude 0.1 and unit width. We have set a=0.5a=0.5, b=10b=10 and σT=1\sigma_T=1. The Bjorken flow coefficients have been obtained from the analytic solution, while the non-Bjorken flow ones from a simulation discussed in the Supplemental Material.

Summary and outlook– Understanding the behaviour of the hydrodynamic gradient expansion at large orders is a challenging question in the foundations of relativistic hydrodynamics. We have proposed a method to compute such series in a large class of models. Applying this to nonlinear longitudinal flows reveals factorially divergent series which we have illustrated with a number of examples. This shows that previously observed instances of factorial growth were not reliant on Bjorken symmetries.

It is natural to ask what are the generic conditions that lead to factorial growth. We have here established that the ability of a system to support arbitrarily large momentum is important; ways around this include working on a lattice, and appropriate restriction of initial data in the linearised case [36], both of which naturally lead to geometric growth. Second, as our analysis of the DN model shows, imposing special symmetries such as boost invariance can also lead to geometric rather than factorial growth. It would be interesting to explore this question further including other natural momentum cut-offs such as microscopic physics and turbulent cascades. For models with recursion relations of the form (10), it may be possible to engineer further examples where the underlying equations of motion give rise to Π⋆(n)\Pi^{(n)}_\star which are eigenfunctions of M\mathcal{M} such that the hydrodynamic gradient expansion grows geometrically. This could form the basis of a rigorous mathematical formulation for investigating the genericity of factorial growth.

The picture that is emerging from this work and results in linear response [36] is that the origin of factorial growth at large nn in the hydrodynamic gradient expansion is the successive action of nn derivatives on the hydrodynamic variables E\mathcal{E}, UU. This is intimately connected with support of a solution in momentum space. It suggests that having a factorial growth in the number of transport coefficients at each order is not necessary.

The factorial divergence of asymptotic series is not an impediment to their practical utility: such series typically provide excellent approximations as long as one does not exceed the so-called order of optimal truncation. Our work makes such investigations possible for a much wider set of flows than previously tractable.

Acknowledgements– The Gravity, Quantum Fields and Information group at AEI is supported by the Alexander von Humboldt Foundation and the Federal Ministry for Education and Research through the Sofja Kovalevskaja Award. A. S. was supported by the Polish National Science Centre grant 2018/29/B/ST2/02457 and by grant CEX2019-000918-M funded by MCIN/AEI/10.13039/501100011033. M. S. is supported by the National Science Centre, Poland, under grants 2018/29/B/ST2/02457 and 2021/41/B/ST2/02909. For the purpose of Open Access, the author has applied a CC-BY public copyright licence to any Author Accepted Manuscript (AAM) version arising from this submission. B. W. is supported by a Royal Society University Research Fellowship.

References

The expansions in ϵ\epsilon and 1/w1/w in Bjorken flow

In the studies of conformal Bjorken flow in the past decade, an important role was played by the following dimensionless clock variable [33]

w≡τT(τ),(16)w \equiv\tau T(\tau), \tag*{(16)}

where τ\tau is the proper time and T(τ)T(\tau) is the local temperature as defined in the main text. The purpose of this Appendix is to review the known relation between the expansion in the Knudsen number in conformal Bjorken flow – the 1/w1/w-expansion mentioned in the Introduction – and the gradient expansion of hydrodynamic constitutive relations as given by Eq. (2) in the main text.

While this relation is general, i.e. it applies to any conformally invariant model, we will present it in the context of the BRSSS model captured by Eq. (3) in the main text in four spacetime dimensions. For Bjorken flow Eq. (3) reduces to

Π⋆′(τ)+(43τ+T(τ)CτΠ)Π⋆(τ)−3Cλ14CηCτΠE0T(τ)3Π⋆(τ)2−8CηE0T(τ)49CτΠτ=0.(17)\Pi_{\star}'(\tau) + \left(\frac{4}{3\tau}+\frac{T(\tau)}{C_{\tau\Pi}}\right)\Pi_{\star}(\tau)-\frac{3C_{\lambda_1}}{4C_{\eta}C_{\tau\Pi}\mathcal{E}_0T(\tau)^3}\Pi_{\star}(\tau)^2-\frac{8C_{\eta}\mathcal{E}_0T(\tau)^4}{9C_{\tau\Pi}\tau}=0. \tag*{(17)}

whereas the conservation equation always (in all theories) reads

T′(τ)+T(τ)3τ−Π⋆(τ)2E0τT(τ)3=0.(18)T'(\tau)+\frac{T(\tau)}{3\tau}-\frac{\Pi_{\star}(\tau)}{2\mathcal{E}_0\tau T(\tau)^3}=0. \tag*{(18)}

Note that it is convenient to replace Π⋆\Pi_{\star} by a dimensionless quantity by taking its ratio with the energy density E\mathcal{E}. This is simply related to the pressure anisotropy A\mathcal{A} measuring deviations from local thermal equilibrium in the following way

A=PT−PLE/3=9Π⋆E.(19)\mathcal{A}=\frac{P_T-P_L}{\mathcal{E}/3}=9\frac{\Pi_{\star}}{\mathcal{E}}. \tag*{(19)}

The 1/w1/w-expansion has been primarily discussed for the this quantity and up to second order it takes the form

A=8Cηw+16Cη(CτΠ−Cλ1)3w2+O(1w3).(20)\mathcal{A}=\frac{8C_{\eta}}{w}+\frac{16C_{\eta}(C_{\tau\Pi}-C_{\lambda_1})}{3w^2}+O\left(\frac{1}{w^3}\right). \tag*{(20)}

Using the expansion in ϵ\epsilon of Π⋆\Pi_{\star} at low orders we obtain the following expression for A\mathcal{A}

A=8CητTϵ−8Cη((CτΠ+2Cλ1)T+9CτΠτT′)3τ2T3ϵ2+O(ϵ3),(21)\mathcal{A}=\frac{8C_{\eta}}{\tau T}\epsilon-\frac{8C_{\eta}\left((C_{\tau\Pi}+2C_{\lambda_1})T+9C_{\tau\Pi}\tau T'\right)}{3\tau^2T^3}\epsilon^2+O(\epsilon^3), \tag*{(21)}

where T=T(τ)T=T(\tau). One can now use the conservation equation to replace derivatives of T(τ)T(\tau) at each order in terms of powers of ww, which directly leads to (A5). This shows that knowing Π⋆\Pi_{\star} in the expansion in ϵ\epsilon to order nn allows one to generate the expansion of A\mathcal{A} in 1/w1/w up to and including terms O(1/wn)O(1/w^n). Note however, that the nn-th order of the expansion in 1/w1/w contains in principle contributions from the orders 2 to nn of the expansion in ϵ\epsilon. A key difference between the two expansions is that the functional expansion of Π⋆\Pi_{\star} in derivatives of E,U\mathcal{E},U counted by ϵ\epsilon as in Eq. (2) does not explicitly require invoking conservation laws (beyond the stated redundancies), whereas the 1/w1/w-expansion necessarily utilizes the conservation equation. Finally, note again that the discussion above applies to conformal Bjorken flow, while in Appendix C we discuss a particular non-conformal version of Bjorken flow.

Further details on numerics and convergence

For the BRSSS simulations shown in Figure 1 we use initial data u=14fu=\frac{1}{4}f, T=1+fT=1+f, Π⋆=0\Pi_{\star}=0 where

f(x)=e−cos⁡(2πx)22π2γ2−e−12π2γ2I0(14π2γ2)(22)f(x)=e^{-\frac{\cos(2\pi x)^2}{2\pi^2\gamma^2}}-e^{-\frac{1}{2\pi^2\gamma^2}}I_0\left(\frac{1}{4\pi^2\gamma^2}\right) \tag*{(22)}

on a unit spatial circle (x∼x+1x\sim x+1). ff was chosen to locally reproduce a Gaussian of width γ\gamma near x=1/2x=1/2, respect the spatial periodicity, and the additive constant chosen so that ff has no homogeneous component in a cosine expansion. We used E=T4\mathcal{E}=T^4, CτΠ=1/4C_{\tau\Pi}=1/4, Cη=1/(4π)C_{\eta}=1/(4\pi), λ1=0\lambda_1=0, γ2=1/60\gamma^2=1/60. The initial value problem is solved using RK4 on a uniform spatial grid with Nx=1600N_x=1600 points for a total of Nt=3200N_t=3200 timesteps to reach the arbitrary choice of end time t=1t=1, working with 300 digits of precision. The differential operator for ∂x\partial_x used periodic fourth-order finite differences, and second-order finite differences for ∂t\partial_t one-sided at t=0t=0 and t=1t=1.

Due to the large numbers of successive applications of derivative operators used in evaluating Π⋆(n)\Pi_{\star}^{(n)} we have placed particular importance on testing convergence of our results, both in the resolution of the numerical grid Nx,NtN_x,N_t and in the number of digits of precision used. Tests confirming convergence of ∣Π⋆(n)∣1/n\left|\Pi_{\star}^{(n)}\right|^{1/n} are shown in Figure 4. For a given resolution and precision, when the window of factorial growth ends it can be extended by increasing one, the other, or both.

Image not extracted

Figure 4. Both precision and numerical resolution are important factors in extracting the correct growth of the hydrodynamic expansion using our technique. Here we show convergence of ∥Π⋆(n)∥1n\lVert\Pi_{\star}^{(n)}\rVert^{\frac{1}{n}} for the BRSSS simulations of Figure 1 at the point marked with ‘+’. Upper panel: Convergence to the continuum with increasing resolution at fixed precision. Lower panel: Convergence with increasing precision at fixed resolution.

The results in the DN model had been obtained working in a curvilinear coordinate system in which the Minkowski metric reads

ds2=−e2α(τ,x)dτ2+e2β(τ,x)dx2+dx⃗⊥ 2,(23)ds^2=-e^{2\alpha(\tau,x)}d\tau^2+e^{2\beta(\tau,x)}dx^2+d\vec{x}_{\perp}^{\,2}, \tag*{(23)}

and the fluid velocity is U=e−α∂τU=e^{-\alpha}\partial_\tau, in such a way that a line of constant x,x⃗⊥x,\vec{x}_{\perp} corresponds to a flow line. Besides the conservation equations and the relaxation equation obeyed by Π⋆\Pi_{\star}, the evolution equations include one extra equation enforcing that the metric (B2) is flat. When performing numerical simulations, we worked with a compactified spatial coordinate ζ∈[−1,1]\zeta\in[-1,1] instead of xx. Both coordinates are related as x=γ−1tanh⁡−1(ζ)x=\gamma^{-1}\tanh^{-1}(\zeta), with γ∈R\gamma\in\mathbb{R}. Spatial derivatives were discretized with fourth-order centered finite-difference stencils, while the time evolution was performed with an explicit RK4 method. For the numerical results displayed in Figure 3, we employed a spatial grid of spacing dζ=1/1800d\zeta=1/1800, a time step dτ=0.5dζd\tau=0.5d\zeta, and worked with precision 300. The time derivatives appearing in the recursion relation were computed with fourth-order finite-difference stencils.

Further perspectives on geometric growth in the DN model

We start by noting that the DN model, as originally defined in Ref. [27], features an additional term in the relaxation equation obeyed by Πμν\Pi_{\mu\nu}. This term is linear in Πμν\Pi_{\mu\nu} and, in this work, we have set to zero the transport coefficient associated to it since this restriction does not modify our main conclusions regarding Bjorken flow. Hence, in the formulation of the DN model we have considered, the gradient expansion (2) for Bjorken flow is given by the following recursion relations,

Π⋆(1)=2a3σTTτ,(24)\Pi_{\star}^{(1)}=\frac{2a}{3\sigma_T}\frac{T}{\tau}, \tag*{(24)}
Π⋆(n+1)=−abKn3Π⋆(n)−abKn4τ∂τΠ⋆(n).(25)\Pi_{\star}^{(n+1)}=-\frac{abK\mathrm{n}}{3}\Pi_{\star}^{(n)}-\frac{abK\mathrm{n}}{4}\tau\partial_\tau\Pi_{\star}^{(n)}. \tag*{(25)}

The closed-form expression (14) can be shown to be a solution of the recursion relations (C1) by induction. The parameters appearing in Eq. (15) are

T0,+=c0E03n0(1+c0)=c0T0,−,(26)T_{0,+}=\frac{c_0\mathcal{E}_0}{3n_0(1+c_0)}=c_0T_{0,-}, \tag*{(26)}
α±=13+2abKn±234b+9a2b2Kn2,(27)\alpha_\pm=\frac{1}{3}+\frac{2}{abK\mathrm{n}}\pm\frac{2}{3}\sqrt{\frac{4}{b}+\frac{9}{a^2b^2K\mathrm{n}^2}}, \tag*{(27)}

where E0\mathcal{E}_0 is the energy density at τ=τ0\tau=\tau_0. Combining Eq. (14) with Eq. (15), one finds the final form of the gradient expansion coefficients,

Π⋆(n)=∑i=+,−Π⋆,i(n)=∑i=+,−Ai(τ0τ)4abKn+3γi3abKnγin,(28)\Pi_{\star}^{(n)}=\sum_{i=+,-}\Pi_{\star,i}^{(n)}=\sum_{i=+,-}A_i\left(\frac{\tau_0}{\tau}\right)^{\frac{4abK\mathrm{n}+3\gamma_i}{3abK\mathrm{n}}}\gamma_i^n, \tag*{(28)}
A±=−8T0,±bτ0σTKn(1−3α±),(29)A_\pm=-\frac{8T_{0,\pm}}{b\tau_0\sigma_TK\mathrm{n}(1-3\alpha_\pm)}, \tag*{(29)}
γ±=12±169+4a2b2Kn2.(30)\gamma_\pm=\frac{1}{2}\pm\frac{1}{6}\sqrt{9+4a^2b^2K\mathrm{n}^2}. \tag*{(30)}

According to Eq. (C3), Π⋆(n)\Pi_{\star}^{(n)} is given by the sum of two contributions which only grow geometrically at rates γ±\gamma_\pm. No large-order factorially growing behavior is present.

For illustrative purposes, let us focus on the c0=0c_0=0 solution. This choice of initial conditions corresponds to the attractor solution for the normalized pressure anisotropy, which played a prominent role in the original analysis of Ref. [27]. In this case, the terms in Eq. (C3) labeled by a ++ vanish and the gradient expansion converges provided that

Kn<Kn∗≡32ab.(31)K\mathrm{n}<K\mathrm{n}^{*}\equiv\frac{3\sqrt{2}}{a\sqrt{b}}. \tag*{(31)}

This existence of a critical Knudsen number is qualitative agreement with the findings of Ref. [27]. The quantitative difference in the value of Kn∗\mathrm{Kn}^{*} reported here and the one originally quoted in Ref. [27] is explained by the fact that our gradient expansion is different from the one considered there.

We conclude this Appendix by arguing in two different ways that the large-order geometric growth displayed by Π⋆(n)\Pi_{\star}^{(n)} can be understood as a fine-tuned phenomenon.

The first way starts from the realisation that, had one chosen to write the closed-form solution (14) in terms of an expansion in the longitudinal derivative ∂τ\partial_\tau instead of τ∂τ\tau\partial_\tau, one would had gotten an expression of the form

Π⋆(n)=∑k=0n−1cn,kτk−1∂τkT(τ),(32)\Pi_\star^{(n)}=\sum_{k=0}^{n-1}c_{n,k}\tau^{k-1}\partial_\tau^kT(\tau), \tag*{(32)}

where, in particular,

cn,n−1=(−14abKn)n−12a3σT.(33)c_{n,n-1}=\left(-\frac{1}{4}ab\mathrm{K}n\right)^{n-1}\frac{2a}{3\sigma_T}. \tag*{(33)}

For a T(τ)T(\tau) given by a linear combination of power-laws τ−λ\tau^{-\lambda}, the contribution to the gradient expansion of the form

Π⋆⊃∑n=1∞cn,n−1τn−2∂τn−1T(τ),(34)\Pi_\star\supset\sum_{n=1}^{\infty}c_{n,n-1}\tau^{n-2}\partial_\tau^{n-1}T(\tau), \tag*{(34)}

is a factorially divergent asymptotic series when evaluated at a given time since according to Eq. (33) the coefficients cn,n−1c_{n,n-1} grow geometrically and

∂τkτ−λ=(−1)kτ−λ−kΓ(λ+k)Γ(λ).(35)\partial_\tau^k\tau^{-\lambda}=(-1)^k\tau^{-\lambda-k}\frac{\Gamma(\lambda+k)}{\Gamma(\lambda)}. \tag*{(35)}

The fact that the whole gradient expansion only grows geometrically when adding the remaining contributions is indicative of a very delicate cancellation between different factorially growing subsectors.

The second way builds upon an observation already performed in the main text: Π⋆(1)\Pi_\star^{(1)} is a linear superposition of eigenfunctions of the differential operator M\mathcal{M}. Let us elaborate on this. According to the definition (10) and the recursion relation (C1b),

M=−abKn3−abKn4τ∂τ,(36)\mathcal{M}=-\frac{ab\mathrm{K}n}{3}-\frac{ab\mathrm{K}n}{4}\tau\partial_\tau, \tag*{(36)}

in such that a way that its eigenfunctions ψi\psi_i, Mψi=ξiψi\mathcal{M}\psi_i=\xi_i\psi_i, are given by

ψi=τ−43−4abKnξi.(37)\psi_i=\tau^{-\frac{4}{3}-\frac{4}{ab\mathrm{K}n}\xi_i}. \tag*{(37)}

For a Π⋆(1)\Pi_\star^{(1)} corresponding to a finite linear superposition of eigenfunctions,

Π⋆(1)=∑iciψi,(38)\Pi_\star^{(1)}=\sum_i c_i\psi_i, \tag*{(38)}

the gradient expansions grows geometrically with rates given by ξi\xi_i,

Π⋆(n+1)=MnΠ⋆(1)=∑iξinciψi.(39)\Pi_\star^{(n+1)}=\mathcal{M}^n\Pi_\star^{(1)}=\sum_i \xi_i^n c_i\psi_i. \tag*{(39)}

Eqs. (C11) and (C1a) imply that

T=32aρ0τ0Kn∑iciτ−13−4abKnξi.(40)T=\frac{3}{2a\rho_0\tau_0\mathrm{K}n}\sum_i c_i\tau^{-\frac{1}{3}-\frac{4}{ab\mathrm{K}n}\xi_i}. \tag*{(40)}

Inserting the expression above in the conservation equation fixes Π⋆\Pi_\star,

Π⋆=−9a2bKn2∑iciξiτ−43−4abKnξi.(41)\Pi_\star=-\frac{9}{a^2b\mathrm{K}n^2}\sum_i c_i\xi_i\tau^{-\frac{4}{3}-\frac{4}{ab\mathrm{K}n}\xi_i}. \tag*{(41)}

Finally, for the whole procedure to be self-consistent, the Π⋆\Pi_\star given by Eq. (41) has to solve the relaxation equation. This can only be achieved provided that the linear superposition (C14) is restricted to two contributions with eigenvalues

ξ1,2=12±169+4a2bKn2.(42)\xi_{1,2}=\frac{1}{2}\pm\frac{1}{6}\sqrt{9+4a^2b\mathrm{K}n^2}. \tag*{(42)}

The eigenvalues ξ1,2\xi_{1,2} agree with the growth rates quoted in Eq. (C3c) and, when introduced in Eq. (40) (resp. Eq. (41)), match the exponents appearing in Eq. (C2b) (resp. Eq. (C3a)).

In a generic model with a recursion relation of the form (10) M\mathcal{M} depends nontrivially on the hydrodynamic fields. In this situation, while a Π⋆(1)\Pi_\star^{(1)} given by a finite superposition of eigenfunctions would result in a geometrically growing gradient expansion, demanding that the on-shell hydrodynamic fields conspire to produce precisely a Π⋆(1)\Pi_\star^{(1)} of this form is a tall order. It is natural to expect that this is not the case for a generic flow, providing another sense in which the geometric growth of the gradient expansion for Bjorken flow in the DN model is a fine-tuned phenomenon.

References

  1. [1]D. A. Bandurin, A. V. Shytov, L. S. Levitov, R. K. Kumar, A. I. Berdyugin, M. Ben Shalom, I. V. Grigorieva, A. K. Geim and G. Falkovich, Fluidity onset in graphene, Nature Communications 9 (Oct, 2018). [arXiv:1205.5040].
  2. [2]M. Shibata, K. Kiuchi and Y.-i. Sekiguchi, General relativistic viscous hydrodynamics of differentially rotating neutron stars, Phys. Rev. D 95 (2017), no. 8 083005 [arXiv:1703.10303].DOI
  3. [3]M. G. Alford, L. Bovard, M. Hanauske, L. Rezzolla and K. Schwenzer, Viscous Dissipation and Heat Conduction in Binary Neutron-Star Mergers, Phys. Rev. Lett. 120 (2018), no. 4 041101 [arXiv:1707.09475].
  4. [4]F. S. Bemfica, M. M. Disconzi and J. Noronha, Causality of the Einstein-Israel-Stewart Theory with Bulk Viscosity, Phys. Rev. Lett. 122 (2019), no. 22 221602 [arXiv:1901.06701].DOI
  5. [5]U. Heinz and R. Snellings, Collective flow and viscosity in relativistic heavy-ion collisions, Ann. Rev. Nucl. Part. Sci. 63 (2013) 123–151 [arXiv:1301.2826].DOI
  6. [6]W. Busza, K. Rajagopal and W. van der Schee, Heavy Ion Collisions: The Big Picture, and the Big Questions, Ann. Rev. Nucl. Part. Sci. 68 (2018) 339–376 [arXiv:1802.04801].DOI
  7. [7]P. Kovtun, Lectures on hydrodynamic fluctuations in relativistic theories, J. Phys. A 45 (2012) 473001DOI
  8. [8]S. A. Hartnoll, A. Lucas and S. Sachdev, Holographic quantum matter, arXiv:1612.07324.
  9. [9]W. Florkowski, M. P. Heller and M. Spaliński, New theories of relativistic hydrodynamics in the LHC era, Rept. Prog. Phys. 81 (2018), no. 4 046001 [arXiv:1707.02282].DOI
  10. [10]P. Romatschke and U. Romatschke, Relativistic Fluid Dynamics In and Out Equilibrium. Cambridge Monographs on Mathematical Physics. Cambridge University Press, 5, 2019.
  11. [11]A. Soloviev, Hydrodynamic attractors in heavy ion collisions: a review, arXiv:1909.10581.
  12. [12]M. P. Heller, R. A. Janik and P. Witaszczyk, Hydrodynamic Gradient Expansion in Gauge Theory Plasmas, Phys.Rev.Lett. 110 (2013), no. 21 211602 [arXiv:1302.0697].DOI
  13. [13]A. Buchel, M. P. Heller and J. Noronha, Entropy Production, Hydrodynamics, and Resurgence in the Primordial Quark-Gluon Plasma from Holography, Phys. Rev. D 94 (2016), no. 10 106011 [arXiv:1603.05344].DOI
  14. [14]J. Casalderrey-Solana, N. I. Gushterov and B. Meiring, Resurgence and Hydrodynamic Attractors in Gauss-Bonnet Holography, arXiv:1712.02772.
  15. [15]M. Baggioli and A. Buchel, Holographic Viscoelastic Hydrodynamics, JHEP 03 (2019) 146 [arXiv:1805.06756].
  16. [16]A. Buchel, Non-conformal holographic Gauss-Bonnet hydrodynamics, JHEP 03 (2018) 037 [arXiv:1801.06165].DOI
  17. [17]I. Aniceto, B. Meiring, J. Jankowski and M. Spaliński, The large proper-time expansion of Yang-Mills plasma as a resurgent transseries, JHEP 02 (2019) 073 [arXiv:1810.07130].
  18. [18]M. P. Heller and M. Spaliński, Hydrodynamics Beyond the Gradient Expansion: Resurgence and Resummation, Phys. Rev. Lett. 115 (2015), no. 7 072501 [arXiv:1503.07514].DOI
  19. [19]G. Başar and G. V. Dunne, Hydrodynamics, resurgence, and transasymptotics, Phys. Rev. D92 (2015), no. 12 125011 [arXiv:1509.05046].
  20. [20]I. Aniceto and M. Spaliński, Resurgence in Extended Hydrodynamics, Phys. Rev. D93 (2016), no. 8 085008 [arXiv:1511.06358].DOI
  21. [21]G. S. Denicol and J. Noronha, Divergence of the Chapman-Enskog expansion in relativistic kinetic theory, arXiv:1608.07869.
  22. [22]W. Florkowski, R. Ryblewski and M. Spaliński, Gradient expansion for anisotropic hydrodynamics, Phys. Rev. D 94 (2016), no. 11 114025 [arXiv:1608.07558].DOI
  23. [23]M. P. Heller, A. Kurkela, M. Spaliński and V. Svensson, Hydrodynamization in kinetic theory: Transient modes and the gradient expansion, Phys. Rev. D97 (2018), no. 9 091503 [arXiv:1609.04803].DOI
  24. [24]G. S. Denicol and J. Noronha, Analytical attractor and the divergence of the slow-roll expansion in relativistic hydrodynamics, arXiv:1711.01657.
  25. [25]M. P. Heller and V. Svensson, How does relativistic kinetic theory remember about initial conditions?, Phys. Rev. D98 (2018), no. 5 054016 [arXiv:1802.08225].
  26. [26]J.-P. Blaizot and L. Yan, Emergence of hydrodynamical behavior in expanding ultra-relativistic plasmas, Annals Phys. 412 (2020) 167993 [arXiv:1904.08677].
  27. [27]G. S. Denicol and J. Noronha, Another hydrodynamic attractor of an ultrarelativistic gas of hard spheres, Phys. Rev. Lett. 124 (2020), no. 15 152301 [arXiv:1908.09957].
  28. [28]A. Behtash, C. N. Cruz-Camacho and M. Martinez, Far-from-equilibrium attractors and nonlinear dynamical systems approach to the Gubser flow, Phys. Rev. D 97 (2018), no. 4 044041 [arXiv:1711.01745].
  29. [29]G. S. Denicol and J. Noronha, Hydrodynamic attractor and the fate of perturbative expansions in Gubser flow, Phys. Rev. D 99 (2019), no. 11 116004 [arXiv:1804.04771].
  30. [30]A. Behtash, S. Kamata, M. Martinez and H. Shi, Global flow structure and exact formal transseries of the Gubser flow in kinetic theory, JHEP 07 (2020) 226 [arXiv:1911.06406].
  31. [31]Z. Du, X.-G. Huang and H. Taya, Hydrodynamic attractor in a Hubble expansion, Phys. Rev. D 104 (2021), no. 5 056022 [arXiv:2104.12534].DOI
  32. [32]J. Bjorken, Highly Relativistic Nucleus-Nucleus Collisions: The Central Rapidity Region, Phys.Rev. D27 (1983) 140–151.
  33. [33]M. P. Heller, R. A. Janik and P. Witaszczyk, The characteristics of thermalization of boost-invariant plasma from holography, Phys.Rev.Lett. 108 (2012) 201602 [arXiv:1103.3452].
  34. [34]S. S. Gubser, Symmetry constraints on generalizations of Bjorken flow, Phys. Rev. D 82 (2010) 085027 [arXiv:1006.0006].
  35. [35]S. S. Gubser and A. Yarom, Conformal hydrodynamics in Minkowski and de Sitter spacetimes, Nucl. Phys. B 846 (2011) 469–511 [arXiv:1012.1314].
  36. [36]M. P. Heller, A. Serantes, M. Spaliński, V. Svensson and B. Withers, Hydrodynamic gradient expansion in linear response theory, Phys. Rev. D 104 (2021), no. 6 066002 [arXiv:2007.05524].
  37. [37]B. Withers, Short-lived modes from hydrodynamic dispersion relations, JHEP 06 (2018) 059 [arXiv:1803.08058].DOI
  38. [38]R. Baier, P. Romatschke, D. T. Son, A. O. Starinets and M. A. Stephanov, Relativistic viscous hydrodynamics, conformal invariance, and holography, JHEP 04 (2008) 100 [arXiv:0712.2451].
  39. [39]W. Israel, Nonstationary irreversible thermodynamics: a causal relativistic theory, Annals of Physics 100 (1976), no. 1-2 310–331.DOI
  40. [40]W. Israel and J. Stewart, Transient relativistic thermodynamics and kinetic theory, Annals Phys. 118 (1979) 341–372.
  41. [41]G. S. Denicol, H. Niemi, E. Molnar and D. H. Rischke, Derivation of transient relativistic fluid dynamics from the Boltzmann equation, Phys. Rev. D 85 (2012) 114047 [arXiv:1202.4551]. [Erratum: Phys.Rev.D 91, 039902 (2015)].
  42. [42]A. Jaiswal, Relativistic third-order dissipative fluid dynamics from kinetic theory, Phys. Rev. C 88 (2013) 021903 [arXiv:1305.3480].
  43. [43]J. Noronha and G. S. Denicol, Transient Fluid Dynamics of the Quark-Gluon Plasma According to AdS/CFT, arXiv:1104.2415.
  44. [44]M. P. Heller, R. A. Janik, M. Spaliński and P. Witaszczyk, Coupling hydrodynamics to nonequilibrium degrees of freedom in strongly interacting quark-gluon plasma, Phys.Rev.Lett. 113 (2014), no. 26 261601 [arXiv:1409.5087].
  45. [45]R. Loganayagam, Entropy Current in Conformal Hydrodynamics, JHEP 0805 (2008) 087 [arXiv:0801.3701].DOI
  46. [46]W. Florkowski, R. Ryblewski, M. Strickland and L. Tinti, Non-boost-invariant dissipative hydrodynamics, Phys. Rev. C 94 (2016), no. 6 064903 [arXiv:1609.06293].DOI
  47. [47]Note that longitudinal flows are in general distinct from comoving flows such as Bjorken and Gubser flow. While Bjorken flow can be found as a special case of a longitudinal flow, Gubser flow cannot since it features transverse dynamics as dictated by symmetries.

Paper details

vlnc:W_69ktzsw8
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