betaFeedback

WorksPhysics and Astronomy

Relativistic fluid dynamics in a 'hydro' frame

Abstract

In this letter, we investigate how field redefinition influences the spectrum of linearized perturbations in relativistic fluid dynamics. We show that the hydrodynamic modes do not get affected under local field redefinition, whereas the non-hydrodynamic modes do. These non-hydrodynamic modes can be removed through a suitable all-order field redefinition. This process leads to a new frame containing only hydrodynamic modes, which we refer to as the hydro frame. Additionally, we demonstrate that the resulting stress-energy tensor may constitute an infinite series in momentum space, with the radius of convergence associated with the removed non-hydrodynamic mode, highlighting its role in the hydrodynamic expansion's validity.

Introduction:– In relativistic fluid dynamics, any viable model must satisfy critical physical criteria like causality and stability. These are typically assessed by examining linearized perturbations around an equilibrium background. When this background shows translational symmetry, the analysis uses Fourier modes, scrutinizing their spectrum to ensure compliance with physical constraints. To ensure the robustness and observer-independence of these criteria, the analysis must respect the system’s inherent symmetries. For instance, if the background fluid is rotationally invariant, all spatial coordinates related by global rotations must be treated equivalently. This means the observer can choose any coordinate set, and the derived constraints should remain consistent. Incorporating rotational invariance into the perturbation analysis ensures this symmetry is reflected in the Fourier mode spectrum and resulting constraints. Thus, the fluid model’s physical validity remains unaffected by the observer’s coordinate choice, preserving its robustness and universality[1, 2, 3, 4, 5, 6]. Nevertheless, manifest invariance is not assured for all gauge freedoms in a model. When invariance is not explicit, it is important to separate artefacts of choice from the genuine physical spectrum to test stability and causality constraints effectively.

In relativistic hydrodynamics, defining fluid velocity and temperature beyond the perfect fluid case introduces ambiguity. In thermodynamic equilibrium, these variables are naturally defined by uniform thermodynamic potentials. Outside equilibrium, they can be re-defined with arbitrary corrections that become nonzero with spacetime variations. These ambiguities are typically resolved by imposing constraints on the conserved currents, akin to gauge fixing in gauge theory, with common choices being the ‘Landau’ or ‘Eckart’ frame[7, 8]. However, one can always opt for constraints beyond these or even proceed without addressing these field ambiguities, as in BDNK theory[9, 10, 11]. It appears that the spectrum of linearized perturbations is significantly altered when the fluid fields are redefined. The BDNK formalism, in particular, leaves the fluid frame unfixed to establish the frame in which the first-order fluid dynamics preserves causality and stability. This method diverges from the established difficulties related to causality and stability in the traditional Landau or Eckart frame[12, 13, 14, 15, 16, 17]. While rotational and translational symmetries are easily addressed in the analysis, gauge freedoms in defining fluid variables add further complexities. Addressing these complexities is crucial for ensuring that the physical constraints derived from linearized perturbation analysis appropriately represent the underlying physics without any artefacts from arbitrary definitions[18, 19, 20, 21, 22, 23]. It is well understood that the linearized spectrum changes with field redefinition, as the perturbations of the redefined fields correspond to different physical quantities than the original fields[24, 25, 26, 27]. However, the causal nature of a specific fluid model should not be contingent upon the choice of variables used for its description.

The causality of any relativistic fluid dynamic model depends on the nature of non-hydrodynamic modes. If the non-hydrodynamics modes are acausal, then the fluid model is also acausal. Thus, to establish a causal initial value problem for relativistic hydrodynamics, one must incorporate non-hydrodynamic modes[28]. These non-hydrodynamic modes are linked with the hyperbolicity of the equations, and field redefinitions may change the structure of non-hydrodynamic modes. It is possible that a fluid model that is causal in one frame might be acausal after the redefinition of fluid variables. Therefore, one may wonder which of the fluid models should be considered physical. We, in this work, provided a novel technique to find out which fluid model is physical under arbitrary field redefinition. The method is simple: we, through an all-order field redefinition, eliminate all non-hydrodynamic modes from the fluid model and analyze the structure of the resulting modified energy-momentum tensor. This technique works provided the original stress-energy tensor (the one with non-hydrodynamic modes, before the field redefinition) has a finite number of terms in derivative expansion. We observe that if the new modified stress-energy tensor, after eliminating the non-hydrodynamic modes, incorporates higher derivative terms of the fluid variables that include up to infinite orders, then the fluid theory we started with is physical, whereas, if the modified stress-energy tensor turns out to have a finite number of terms, then the non-hydrodynamic modes of the original theory must emerge from field redefinition artefacts solely and should not be considered physical. Another criterion we find is whether the dispersion polynomial neatly factorizes into hydrodynamic and non-hydrodynamic modes. If the non-hydrodynamic modes do not appear as distinct finite factors, then they are not frame artefacts[29]. In this and a companion study[29], we investigate the impact of field redefinition on the spectrum of linearized perturbations within relativistic fluid dynamics. We have found that the spectrum of hydrodynamic modes (those with frequencies that vanish as spatial momentum approaches zero) remains unchanged by local field redefinition. In contrast, the spectrum of non-hydrodynamic modes (those with frequencies that approach a finite nonzero value as momenta diminish) is influenced by field redefinition. We also show that non-hydrodynamic modes can be removed from the spectrum of linearized perturbations through a suitable, generally all-order field redefinition. Once this is applied, the spectrum will be simplified to include only the hydrodynamic modes of the original fluid model, and the stress-energy tensor becomes ready for analyzing the physical validity of the original non-hydrodynamic modes. We will refer to this new fluid frame as the “hydro frame”. Even if we start with a stress tensor that contains a finite number of terms, transforming to the “hydro frame” typically yields a stress tensor with an infinite number of terms, incorporating arbitrarily higher-order derivatives of the fluid variables. In Ref. [29], we show how the non-hydrodynamic modes of the original theory could control the hydrodynamic expansion’s validity of the new stress-energy tensor.

Setup:— Let’s consider a set of variables {Φi}\{\Phi_i\} (such as velocity, temperature, and conserved charges in a fluid) governed by a system of nonlinear coupled PDEs represented as E({Φi})=0\mathcal{E}(\{\Phi_i\})=0 (for fluids, E\mathcal{E} includes equations from stress tensor and charge current conservation). We assume that {Φi}={Φˉi}\{\Phi_i\}=\{\bar{\Phi}_i\} is a precise solution to E\mathcal{E} that is invariant under spacetime translations and spatial rotations. To study the spectrum, we first linearize the equations in E\mathcal{E} around {Φˉi}\{\bar{\Phi}_i\}

Φis=Φˉi+ϵ δΦi(ω,k)e−iωt+ik⃗⋅x⃗,with k=k⃗⋅k⃗, ϵ≪1,\Phi_i^s=\bar{\Phi}_i+\epsilon\,\delta\Phi_i(\omega,k)e^{-i\omega t+i\vec{k}\cdot\vec{x}},\quad\text{with } k=\sqrt{\vec{k}\cdot\vec{k}},\ \epsilon\ll1,
E({Φi=Φis})=0⇒∑jMij(Φˉ,ω,k)δΦj=0,(1)\mathcal{E}(\{\Phi_i=\Phi_i^s\})=0\Rightarrow\sum_j M_{ij}(\bar{\Phi},\omega,k)\delta\Phi_j=0, \tag*{(1)}

where MijM_{ij} is the linearization matrix, and the spectrum is obtained from the zeros of the determinant of MijM_{ij}. If the equation E\mathcal{E} contains a finite number of derivatives, then Det⁡[M]\operatorname{Det}[M] will be a finite polynomial in ω\omega and kk, resulting in a finite number of zeros. Now consider a field redefinition Φi→Ψi=Φi+ΔΦi\Phi_i\to\Psi_i=\Phi_i+\Delta\Phi_i, where the shift ΔΦi\Delta\Phi_i is a nonlinear function of Φi\Phi_i and its derivatives. In the context of fluid dynamics, we assume that at equilibrium, Φi\Phi_i and Ψi\Psi_i coincide, meaning ΔΦi\Delta\Phi_i vanishes when evaluated at Φi=Φˉi\Phi_i=\bar{\Phi}_i. Under this field redefinition, the equations of motion transform as E(Φ)→E~(Ψ)\mathcal{E}(\Phi)\to\widetilde{\mathcal{E}}(\Psi). While the set of PDEs E~\widetilde{\mathcal{E}} can be fully determined from E\mathcal{E} and the transformation ΔΦ\Delta\Phi, it will have a distinctly different structure. We can linearize E~\widetilde{\mathcal{E}} to obtain the spectrum, similarly to how we did for E\mathcal{E} in (1), resulting in a different linearization matrix M~ij\widetilde{M}_{ij}. Our objective is to establish the relationship between MijM_{ij} and M~ij\widetilde{M}_{ij}

Ψis=Φˉi+ϵ δΨi(ω,k)e−iωt+ik⃗⋅x⃗with k=k⃗⋅k⃗, ϵ≪1,\Psi_i^s=\bar{\Phi}_i+\epsilon\,\delta\Psi_i(\omega,k)e^{-i\omega t+i\vec{k}\cdot\vec{x}}\quad\text{with } k=\sqrt{\vec{k}\cdot\vec{k}},\ \epsilon\ll1,
E~({Ψi=Ψis})=0⇒∑jM~ij(Φˉ,ω,k)δΨj=0.(2)\widetilde{\mathcal{E}}(\{\Psi_i=\Psi_i^s\})=0\Rightarrow\sum_j\widetilde{M}_{ij}(\bar{\Phi},\omega,k)\delta\Psi_j=0. \tag*{(2)}

As ΔΦ\Delta\Phi is vanishing when Φ=Φˉ\Phi=\bar{\Phi}, it must be of order O(ϵ)\mathcal{O}(\epsilon) or higher for the field configuration in (1). Thus, ΔΦ\Delta\Phi can be represented by the matrix equation up to O(ϵ2)\mathcal{O}(\epsilon^2)

ΔΦi(Φs)=ϵ∑jSij(Φˉ,ω,k)δΦj(ω,k)e−iωt+ik⃗⋅x⃗.(3)\Delta\Phi_i(\Phi^s)=\epsilon\sum_j S_{ij}(\bar{\Phi},\omega,k)\delta\Phi_j(\omega,k)e^{-i\omega t+i\vec{k}\cdot\vec{x}}. \tag*{(3)}

Under field redefinition, Φis→Ψis\Phi_i^s\to\Psi_i^s

Ψis=Φˉi+ϵ δΨi(ω,k)e−iωt+ik⃗⋅x⃗,=Φis+ΔΦi(Φs),=Φˉi+ϵ∑j[δij+Sij(Φˉ,ω,k)]δΦj(ω,k)e−iωt+ik⃗⋅x⃗,(4)\begin{aligned} \Psi_i^s&=\bar{\Phi}_i+\epsilon\,\delta\Psi_i(\omega,k)e^{-i\omega t+i\vec{k}\cdot\vec{x}},\\ &=\Phi_i^s+\Delta\Phi_i(\Phi^s),\\ &=\bar{\Phi}_i+\epsilon\sum_j\left[\delta_{ij}+S_{ij}(\bar{\Phi},\omega,k)\right]\delta\Phi_j(\omega,k)e^{-i\omega t+i\vec{k}\cdot\vec{x}}, \tag*{(4)} \end{aligned}

implying

δΨi(Φˉ,ω,k)=∑j[δij+Sij(Φˉ,ω,k)]δΦj(Φˉ,ω,k).(5)\delta\Psi_i(\bar{\Phi},\omega,k)=\sum_j\left[\delta_{ij}+S_{ij}(\bar{\Phi},\omega,k)\right]\delta\Phi_j(\bar{\Phi},\omega,k). \tag*{(5)}

Putting (5) in (2) and comparing with (1), we obtain

∑jkM~ij(δjk+Sjk)δΦk=0⇒Mik=∑jM~ij(δjk+Sjk)⇒Det⁡[M]=Det⁡[M~]Det⁡[1+S].(6)\begin{aligned} \sum_{jk}\widetilde{M}_{ij}\left(\delta_{jk}+S_{jk}\right)\delta\Phi_k=0\Rightarrow M_{ik}=\sum_j\widetilde{M}_{ij}\left(\delta_{jk}+S_{jk}\right) \\ &\Rightarrow\operatorname{Det}[M]=\operatorname{Det}[\widetilde{M}]\operatorname{Det}[1+S]. \tag*{(6)} \end{aligned}

The zeros of Det⁡[1+S]\operatorname{Det}[1+S] represent new modes in the Φ\Phi frame that are absent in the Ψ\Psi frame. These new zero modes arise from perturbations in the Φ\Phi frame, where the additional terms coming from the linearized field redefinitions are exactly canceled by the linearized fluctuations. In the Ψ\Psi frame, these new modes indicate no fluctuations, meaning that they are artefacts of the frame transformation with no physical significance. Importantly, depending on the choice of field redefinitions, these artificial modes may appear unstable or acausal, even if the theory in the Ψ\Psi frame is entirely valid. However, in the Φ\Phi frame, there is no straightforward way to distinguish these artefacts from genuine physical modes.

Within the framework of relativistic fluid dynamics:— Here, we concentrate on fluid variables to establish the most general form of the frame transformation matrix SS (see eq. (3)). To simplify our analysis, we will focus on uncharged fluids, where fluid velocity and temperature are the only variables, and the equation of motion is dictated by stress tensor conservation.

Let the velocity and temperature in two different frames, ‘Frame-1’ and ‘Frame-2,’ be denoted as {u^μ,T^}\{\hat{u}^{\mu},\hat{T}\} and {uμ,T}\{u^{\mu},T\}, respectively. The shift functions Δuμ\Delta u^{\mu} and ΔT\Delta T are defined as

uμ=u^μ+Δuμ(u^,T^),T=T^+ΔT(u^,T^).(7)u^{\mu}=\hat{u}^{\mu}+\Delta u^{\mu}(\hat{u},\hat{T}), \qquad T=\hat{T}+\Delta T(\hat{u},\hat{T}). \tag*{(7)}

Fluid variables in different frames are expected to align in a rotationally and translationally invariant equilibrium. Consequently, every term in Δuμ\Delta u^{\mu} and ΔT\Delta T must include at least one spacetime derivative, ensuring that the shift variables are non-zero only for spatially or temporally non-uniform fluid profiles. We focus on terms that contribute at the linear order in fluctuation amplitude when evaluated on equilibrium fluid profiles plus small fluctuations. This holds for single derivative terms, while for multiple derivatives, all must act on a single fluid variable such as (uαuβ∂α∂β)uμ(u^{\alpha}u^{\beta}\partial_{\alpha}\partial_{\beta})u^{\mu}. Therefore, the most general expressions for Δuμ\Delta u^{\mu} and ΔT\Delta T that may influence the spectrum of small fluctuations are[29]

Δuμ=Fu(u^⋅∂)u^μ+FT(P^μα∂αT^T^)+Ru(P^μθP^αβ∂α∂βu^θ),\Delta u^{\mu}=F_{u}\left(\hat{u}\cdot\partial\right)\hat{u}^{\mu}+F_{T}\left(\frac{\hat{P}^{\mu\alpha}\partial_{\alpha}\hat{T}}{\hat{T}}\right)+R_{u}\left(\hat{P}^{\mu\theta}\hat{P}^{\alpha\beta}\partial_{\alpha}\partial_{\beta}\hat{u}_{\theta}\right),
ΔTT^=Gu(∂⋅u^)+GT(u^α∂αT^T^)+RT(P^αβ∂α∂βT^T^).(8)\frac{\Delta T}{\hat{T}}=G_{u}\left(\partial\cdot\hat{u}\right)+G_{T}\left(\frac{\hat{u}^{\alpha}\partial_{\alpha}\hat{T}}{\hat{T}}\right)+R_{T}\left(\frac{\hat{P}^{\alpha\beta}\partial_{\alpha}\partial_{\beta}\hat{T}}{\hat{T}}\right). \tag*{(8)}

where Fu(T)=Fu(T)[(u^⋅∂),P^αβ∂α∂β]F_{u}(T)=F_{u}(T)[(\hat{u}\cdot\partial),\hat{P}^{\alpha\beta}\partial_{\alpha}\partial_{\beta}], Gu(T)=Gu(T)[(u^⋅∂),P^αβ∂α∂β]G_{u}(T)=G_{u}(T)[(\hat{u}\cdot\partial),\hat{P}^{\alpha\beta}\partial_{\alpha}\partial_{\beta}] are linear differential operators

Fu(T)≡∑m,nfm,nu(T)[u^⋅∂]m[P^αβ∂α∂β]n,F_{u}(T)\equiv\sum_{m,n}f_{m,n}^{u(T)}[\hat{u}\cdot\partial]^{m}\left[\hat{P}^{\alpha\beta}\partial_{\alpha}\partial_{\beta}\right]^{n},
Gu(T)≡∑m,ngm,nu(T)[u^⋅∂]m[P^αβ∂α∂β]n,(9)G_{u}(T)\equiv\sum_{m,n}g_{m,n}^{u(T)}[\hat{u}\cdot\partial]^{m}\left[\hat{P}^{\alpha\beta}\partial_{\alpha}\partial_{\beta}\right]^{n}, \tag*{(9)}

and Ru(T)=Ru(T)[P^αβ∂α∂β]R_{u}(T)=R_{u}(T)[\hat{P}^{\alpha\beta}\partial_{\alpha}\partial_{\beta}] do not depend on the operator (u^⋅∂)(\hat{u}\cdot\partial). They are included to account for field redefinitions that remain non-zero as ω→0\omega\to0. These can also be expanded in the small kk limit as Ru(T)=∑mrmu(T)[P^αβ∂α∂β]mR_{u}(T)=\sum_{m}r_{m}^{u(T)}[\hat{P}^{\alpha\beta}\partial_{\alpha}\partial_{\beta}]^{m}. Here, fm,nu(T)f_{m,n}^{u(T)}, gm,nu(T)g_{m,n}^{u(T)}, and rmu(T)r_{m}^{u(T)} are functions of temperature, and P^μν\hat{P}^{\mu\nu} is the projector orthogonal to u^μ\hat{u}^{\mu}, P^μν=ημν+u^μu^ν\hat{P}^{\mu\nu}=\eta^{\mu\nu}+\hat{u}^{\mu}\hat{u}^{\nu}. In the context of linearized analysis, the derivatives (u^α∂α)(\hat{u}^{\alpha}\partial_{\alpha}) and (P^αβ∂α)(\hat{P}^{\alpha\beta}\partial_{\alpha}) commute, so their ordering in the functions

don't matter. For a rotational and translationally invariant equilibrium case, we have uˉμ={1,0,0,0}\bar{u}^{\mu}=\{1,0,0,0\}, Tˉ=\bar{T}= constant, where the perturbed solution is uμ(s)=uˉμ+ϵ δuμe−iωt+ik⃗⋅x⃗u^{\mu(s)}=\bar{u}^{\mu}+\epsilon\,\delta u_{\mu}e^{-i\omega t+i\vec{k}\cdot\vec{x}}, T(s)=Tˉ+ϵ δTe−iωt+ik⃗⋅x⃗T^{(s)}=\bar{T}+\epsilon\,\delta T e^{-i\omega t+i\vec{k}\cdot\vec{x}}.

As uμ(s)u^{\mu(s)} and uˉμ\bar{u}^{\mu} are normalized to unity, hence, δu⋅uˉ∼O(ϵ2)\delta u\cdot\bar{u}\sim O(\epsilon^{2}) or δuμ\delta u^{\mu} must be of the form δuμ={0,β⃗}\delta u^{\mu}=\{0,\vec{\beta}\} where β⃗\vec{\beta} reads β⃗=βk(k⃗/k)+β⃗⊥\vec{\beta}=\beta_{k}(\vec{k}/k)+\vec{\beta}_{\perp} with k=k⃗⋅k⃗k=\sqrt{\vec{k}\cdot\vec{k}}, βk=(β⃗⋅k⃗)/k\beta_{k}=(\vec{\beta}\cdot\vec{k})/k, and k⃗⋅β⃗⊥=0\vec{k}\cdot\vec{\beta}_{\perp}=0.

After putting in Eq. (8) we obtain Δuμ={0,Δu⃗}\Delta u_{\mu}=\{0,\vec{\Delta u}\} with Δu⃗=Δuk(k⃗/k)+Δu⃗⊥\vec{\Delta u}=\Delta u_{k}(\vec{k}/k)+\vec{\Delta u}_{\perp} such that Δu⃗⊥⋅k⃗=0\vec{\Delta u}_{\perp}\cdot\vec{k}=0. Thus, the shift functions we receive are[29]

\begin{bmatrix} \Delta T\\ \Delta u_{k}\\ \Delta u_{\perp} \end{bmatrix} = S_{ij} \begin{bmatrix} \delta T\\ \beta_{k}\\ \beta_{\perp} \end{bmatrix}, \tag*{(10)}\end{bmatrix}

with the matrix (as mentioned in (3))

Sij=[−iωGT−k2RTikGu0ikFT−iωFu−k2Ru000−iωFu−k2Ru]S_{ij}= \begin{bmatrix} -i\omega G_{T}-k^{2}R_{T} & ikG_{u} & 0\\ ikF_{T} & -i\omega F_{u}-k^{2}R_{u} & 0\\ 0 & 0 & -i\omega F_{u}-k^{2}R_{u} \end{bmatrix}

where Δuk\Delta u_{k} and Δu⊥\Delta u_{\perp} are defined the same way as βk\beta_{k} and β⊥\beta_{\perp}. Now we compute Det⁡[1+S]\operatorname{Det}[1+S] as

F(ω,k2)≡Det⁡[1+S]=(1−iωFu−k2Ru)×[(1−iωGT−k2RT)(1−iωFu−k2Ru)+k2FTGu].(11)\mathcal{F}(\omega,k^{2})\equiv\operatorname{Det}[1+S]=(1-i\omega F_{u}-k^{2}R_{u})\times\left[(1-i\omega G_{T}-k^{2}R_{T})(1-i\omega F_{u}-k^{2}R_{u})+k^{2}F_{T}G_{u}\right]. \tag*{(11)}

We have suppressed the arguments of Fu(T)F_{u}(T), Gu(T)G_{u}(T), and Ru(T)R_{u}(T) for simplicity. From Eq. (6), we note that a general field redefinition of the fluid variables introduces an additional factor, F\mathcal{F}, to the dispersion polynomial which is a polynomial in ω\omega and k2k^{2}. This generates new modes in the system. Some remarks on the properties of Fu(T)F_{u}(T) and Gu(T)G_{u}(T): If FuF_{u} and GTG_{T} are vanishing, and FTF_{T} and GuG_{u} do not depend on the operator (u⋅∂)(u\cdot\partial) or (−iω)(-i\omega) in Fourier space, then F\mathcal{F} will have no zeros for any ω\omega, resulting in no new modes. However, depending on FTF_{T} and GuG_{u}, F\mathcal{F} might have zeros for some real values of kk. If every term in FuF_{u} and GTG_{T} includes at least one factor of (Pμν∂μ∂ν)(P^{\mu\nu}\partial_{\mu}\partial_{\nu}) or k2k^{2} in Fourier space, the frequency of these new modes will diverge as k→0k\to0. If either FuF_{u} or GTG_{T} has at least one term without (Pμν∂μ∂ν)(P^{\mu\nu}\partial_{\mu}\partial_{\nu}), at least one new mode will appear with a finite, non-zero frequency as k→0k\to0, resembling a genuine “non-hydrodynamic” mode. None of the zeros of F\mathcal{F} will have the form lim⁡k→0ω(k)=0\lim_{k\to0}\omega(k)=0. Thus, frame redefinitions will not generate new hydrodynamic modes, and redefining fluid variables will not affect the spectrum of hydrodynamic modes in Fourier space.

Removal of mode from the spectrum via field redefinition:— In the previous section, we noticed that “Frame-2” dispersion polynomial has more non-hydrodynamic modes (coming from the zeros of F\mathcal{F}) than in ‘Frame-1’ due to frame transformation.

Now, let’s reverse this process—starting with the ‘Frame-2’ equations (using the hatted fluid variables u^μ\hat{u}^{\mu} and T^\hat{T}) and implementing the inverse transformations from (7) and (8) to return to ‘Frame-1’ equation with fluid variables uμu^\mu and TT. This inversion would naturally eliminate the extra factor F\mathcal{F} from the ‘Frame-2’ dispersion polynomial. From the viewpoint of ‘Frame-2’, this effectively removes a non-hydrodynamic mode from the spectrum through field redefinition. The possibility to fully absorb a non-hydrodynamic mode through a field redefinition implies that such a mode may not be physical. A truly physical non-hydrodynamic mode should never be completely erased by a frame transformation.

In this section, we will investigate an “inverse transformation” to remove a mode from the spectrum and analyze if physical non-hydrodynamic modes persist as expected. The dispersion polynomials in ‘Frame-1’ and ‘Frame-2’, indicated as P(ω,k2)P(\omega,k^2) and P^(ω,k2)\hat{P}(\omega,k^2), respectively, are related as

P^(ω,k2)=P(ω,k2)F(ω,k2),(12)\hat{P}(\omega,k^2)=P(\omega,k^2)\mathcal{F}(\omega,k^2), \tag*{(12)}

where F\mathcal{F} is given in (11), with

Fu(T)≡∑m,nfm,nu(T)[−iω]m[−k2]n,F_u(T)\equiv\sum_{m,n}f^{u(T)}_{m,n}[-i\omega]^m[-k^2]^n,
Gu(T)≡∑m,ngm,nu(T)[−iω]m[−k2]n,G_u(T)\equiv\sum_{m,n}g^{u(T)}_{m,n}[-i\omega]^m[-k^2]^n,
Ru(T)≡∑mrmu(T)[−k2]m.(13)R_u(T)\equiv\sum_m r^{u(T)}_m[-k^2]^m. \tag*{(13)}

Due to the isotropy of the background fluid profile and the constitutive relations, the linearized dynamics in the shear and sound sectors will get disentangled. Consequently, both P^(ω,k2)\hat{P}(\omega,k^2), P(ω,k2)P(\omega,k^2) and F(ω,k2)\mathcal{F}(\omega,k^2) must factor into contributions from the shear and sound sectors

P^sh(ω,k2)=Psh(ω,k2)(1−iωFu−k2Ru),(14)\hat{P}_{\mathrm{sh}}(\omega,k^2)=P_{\mathrm{sh}}(\omega,k^2)(1-i\omega F_u-k^2R_u), \tag*{(14)}
P^snd(ω,k2)=Psnd(ω,k2)[(1−iωGT−k2RT)(1−iωFu−k2Ru)+k2FTGu],\hat{P}_{\mathrm{snd}}(\omega,k^2)=P_{\mathrm{snd}}(\omega,k^2)\left[(1-i\omega G_T-k^2R_T)(1-i\omega F_u-k^2R_u)+k^2F_TG_u\right],

where P^(ω,k2)=P^sh(ω,k2)P^snd(ω,k2)\hat{P}(\omega,k^2)=\hat{P}_{\mathrm{sh}}(\omega,k^2)\hat{P}_{\mathrm{snd}}(\omega,k^2), and P(ω,k2)=Psh(ω,k2)Psnd(ω,k2)P(\omega,k^2)=P_{\mathrm{sh}}(\omega,k^2)P_{\mathrm{snd}}(\omega,k^2). Assume that Psh(ω,k2)P_{\mathrm{sh}}(\omega,k^2) has N1shN^{\mathrm{sh}}_1 hydrodynamic modes with frequencies ωsha(k)\omega_{\mathrm{sh}}^{a}(k) and N2shN^{\mathrm{sh}}_2 non-hydrodynamic modes with wsha(k2)\mathfrak{w}_{\mathrm{sh}}^{a}(k^2) where, respectively, a={1,2,…,N1sh}a=\{1,2,\ldots,N^{\mathrm{sh}}_1\} and a={1,2,…,N2sh}a=\{1,2,\ldots,N^{\mathrm{sh}}_2\}. And in the sound channel, it has N1sndN^{\mathrm{snd}}_1 hydrodynamic modes with ωsnda(k)\omega_{\mathrm{snd}}^{a}(k) (a={1,2,…,N1snd})(a=\{1,2,\ldots,N^{\mathrm{snd}}_1\}) and N2sndN^{\mathrm{snd}}_2 non-hydrodynamic modes with wsnda(k2)\mathfrak{w}_{\mathrm{snd}}^{a}(k^2) (a={1,2,…,N2snd})(a=\{1,2,\ldots,N^{\mathrm{snd}}_2\}). Thus, Psh(ω,k2)P_{\mathrm{sh}}(\omega,k^2) and Psnd(ω,k2)P_{\mathrm{snd}}(\omega,k^2) can be factored as

Psh(ω,k2)=[∏a=1N1sh(ω−ωsh(a))][∏a=1N2sh(ω−wsh(a))],P_{\mathrm{sh}}(\omega,k^2)=\left[\prod_{a=1}^{N^{\mathrm{sh}}_1}(\omega-\omega_{\mathrm{sh}}^{(a)})\right]\left[\prod_{a=1}^{N^{\mathrm{sh}}_2}(\omega-\mathfrak{w}_{\mathrm{sh}}^{(a)})\right],

and

Psnd(ω,k2)=[∏a=1N1snd(ω−ωsnd(a))][∏a=1N2snd(ω−wsnd(a))].P_{\mathrm{snd}}(\omega,k^2)=\left[\prod_{a=1}^{N^{\mathrm{snd}}_1}(\omega-\omega_{\mathrm{snd}}^{(a)})\right]\left[\prod_{a=1}^{N^{\mathrm{snd}}_2}(\omega-\mathfrak{w}_{\mathrm{snd}}^{(a)})\right].

Both ωsh/snd(a)\omega_{\mathrm{sh/snd}}^{(a)} and wsh/snd(a)\mathfrak{w}_{\mathrm{sh/snd}}^{(a)} are generally complex, non-polynomial functions of kk with limits: lim⁡k→0ωsnd/sh(a)(k)=0\lim_{k\to0}\omega_{\mathrm{snd/sh}}^{(a)}(k)=0, lim⁡k→0wsh/snd(a)(k)=csh/snd(a) ∀a\lim_{k\to0}\mathfrak{w}_{\mathrm{sh/snd}}^{(a)}(k)=c_{\mathrm{sh/snd}}^{(a)}\ \forall a.

Our aim is to identify a frame transformation that eliminates non-hydrodynamic modes while remaining precisely equivalent to the original system in the hydrodynamic sector. This exact equivalence requires that both the frequencies and eigenvectors in the hydrodynamic sector align perfectly before and after the transformation. It’s essential to recognize that the permissible non-trivial profiles of linearized perturbations also cleanly factorize into two subspaces: the 2D space of the sound channel, characterized by the velocity perturbation aligned with the wave vector (k⃗\vec{k}) and the temperature perturbation, and the 1D subspace of the shear channel, represented by the velocity perturbation perpendicular to k⃗\vec{k}. As the hydrodynamic eigenspace in the shear channel is one-dimensional, the frame transformation will automatically match the eigenvector after removing the shear non-hydrodynamic frequency from the spectrum without impacting the hydrodynamic frequency. In contrast, in the sound channel, the eigenvector is a specific vector in the two-dimensional subspace of temperature and longitudinal velocity perturbations, defined by the ratio R\mathcal{R} of these two components: R(ω,k)=δT/δuk\mathcal{R}(\omega,k)=\delta T/\delta u_k. In the sound channel, after eliminating the non-hydrodynamic mode, we must ensure that both the hydrodynamic frequencies and R\mathcal{R} remain unchanged before and after the frame transformation. Hence, we obtain

(1−iωFu−k2Ru)[∏a=1N2sh(ω−wsha)]=Csh,(1-i\omega F_u-k^2R_u)\left[\prod_{a=1}^{N^{\mathrm{sh}}_2}(\omega-\mathfrak{w}_{\mathrm{sh}}^{a})\right]=C_{\mathrm{sh}},
[(1−iωGT−k2RT)(1−iωFu−k2Ru)+k2FTGu][∏a=1N2snd(ω−wsnda)]=Csnd,(15)\left[(1-i\omega G_T-k^2R_T)(1-i\omega F_u-k^2R_u)+k^2F_TG_u\right]\left[\prod_{a=1}^{N^{\mathrm{snd}}_2}(\omega-\mathfrak{w}_{\mathrm{snd}}^{a})\right]=C_{\mathrm{snd}}, \tag*{(15)}

where Csh=lim⁡k→0∏a=1N2sh[−wsha]C_{\mathrm{sh}}=\lim_{k\to0}\prod_{a=1}^{N^{\mathrm{sh}}_2}[-\mathfrak{w}_{\mathrm{sh}}^{a}] and Csnd=lim⁡k→0∏a=1N2snd[−wsnda]C_{\mathrm{snd}}=\lim_{k\to0}\prod_{a=1}^{N^{\mathrm{snd}}_2}[-\mathfrak{w}_{\mathrm{snd}}^{a}]. If Fu(T)F_u(T), Gu(T)G_u(T), and Ru(T)R_u(T) are finite polynomials in ω\omega and k2k^2, the condition, (15), cannot be fulfilled, by permitting an infinite-order derivative expansion in Fourier space (around ω=k=0\omega=k=0), we can, in principle, eliminate all non-hydrodynamic modes from the dispersion polynomial through an appropriate selection of expansion coefficients fm,nu(T)f^{u(T)}_{m,n}, gm,nu(T)g^{u(T)}_{m,n}, rmu(T)r^{u(T)}_m. Moreover, the radius of convergence for this infinite-order frame transformation in Fourier space will be dictated by the position of the nearest non-hydrodynamic mode to the origin in the multi-dimensional space of real momenta and complex frequencies. Let’s normalize the perturbed solution in the sound channel as: uμ(s)=uˉμ+ϵ{0,k/k⃗}e−iωt+ik⃗⋅x⃗u_\mu^{(s)}=\bar{u}_\mu+\epsilon\{0,k/\vec{k}\}e^{-i\omega t+i\vec{k}\cdot\vec{x}}, T(s)=Tˉ+ϵδTe−iωt+ik⃗⋅x⃗T^{(s)}=\bar{T}+\epsilon\delta T e^{-i\omega t+i\vec{k}\cdot\vec{x}}, where δT\delta T is dictated by R(ω,k)\mathcal{R}(\omega,k) before the frame transformation and R^(ω,k)\hat{\mathcal{R}}(\omega,k) after the transformation. The equivalence of the hydrodynamic sector requires that:

R(ωsnda,k)=R^(ωsnda,k)∀a={1,2,…,N1snd}.(16)\mathcal{R}(\omega_{\mathrm{snd}}^{a},k)=\hat{\mathcal{R}}(\omega_{\mathrm{snd}}^{a},k)\quad\forall a=\{1,2,\ldots,N^{\mathrm{snd}}_1\}. \tag*{(16)}

Let’s examine whether we can obtain a solution for this set of coupled nonlinear equations (15) and (16)) and to what degree the solution would be unique. Naively, it appears that there are (N1snd+2)(N^{\mathrm{snd}}_1+2) equations for six unknown functions. However, RTR_T and RuR_u depend only on kk. Taking the limit ω→0\omega\to0 in the two equations of (15) effectively yields two additional equations involving RTR_T, RuR_u, and [FTGu](0)[F_TG_u](0) (defined as [FTGu](0)≡lim⁡ω→0[FT(ω,k)Gu(ω,k)][F_TG_u](0) \equiv\lim_{\omega\to0}[F_T(\omega,k)G_u(\omega,k)]).

Using these additional equations, we can uniquely determine RTR_T and RuR_u in terms of [FTGu](0)[F_TG_u](0) and other known functions of kk, such as ωsnda(k)\omega_{\mathrm{snd}}^a(k) and ωsha(k)\omega_{\mathrm{sh}}^a(k). Next, we substitute the solutions for RTR_T and RuR_u and reintroduce ω\omega into (15). We can now uniquely solve for FuF_u and GTG_T in terms of the product FTGuF_TG_u and [FTGu](0)[F_TG_u](0). At this point, both equations in (15) will be solved for all values of ω\omega and kk. Next, we have N1sndN^{\mathrm{snd}}_1 equations from (16) to determine FTF_T and GuG_u. For an uncharged hydrodynamic theory like those discussed here, N1snd=2N^{\mathrm{snd}}_1=2, giving us two equations for the two remaining unknowns, FTF_T and GuG_u. However, the equations in (16) are evaluated at ω=ωsnda(k)\omega=\omega_{\mathrm{snd}}^a(k), which prevents us from uniquely determining FTF_T and GuG_u for all values of ω\omega. In particular, we can add two arbitrary functions of the product ∏a(ω−ωsnda)\prod_a(\omega-\omega_{\mathrm{snd}}^a) to any solutions for FTF_T and GuG_u, and they will still satisfy (16). Despite the minor ambiguity in FTF_T and GuG_u noted above, we can nearly uniquely determine the functions in the linearized frame transformation by imposing that the spectrum in the transformed frame contains only the hydrodynamic modes[29].

Field redefinition and non-hydrodynamic mode in BDNK

Let us now apply the above algorithm to eliminate the non-hydrodynamic modes in the BDNK theory[9, 10]. The stress-energy tensor for an uncharged conformal system reads

Tμν=T4{(4uμuν+ημν)+4[θ(uμqν+uνqμ)+χE(4uμuν+ημν)−2λσμν]}(17)T^{\mu\nu}=T^4\left\{(4u^\mu u^\nu+\eta^{\mu\nu})+4\left[\theta(u^\mu q^\nu+u^\nu q^\mu)+\chi E(4u^\mu u^\nu+\eta^{\mu\nu})-2\lambda\sigma^{\mu\nu}\right]\right\} \tag*{(17)}

with qμ≡(u⋅∂)uμ+Pμν∂νTTq^\mu\equiv(u\cdot\partial)u^\mu+\frac{P^{\mu\nu}\partial_\nu T}{T}, E≡(u⋅∂)TT+(∂⋅u)3E\equiv\frac{(u\cdot\partial)T}{T}+\frac{(\partial\cdot u)}{3}, and

σμν≡(PμαPνβ+PμβPνα2−PμνPαβ)(∂αuβ).\sigma^{\mu\nu}\equiv\left(\frac{P^{\mu\alpha}P^{\nu\beta}+P^{\mu\beta}P^{\nu\alpha}}{2}-P^{\mu\nu}P^{\alpha\beta}\right)(\partial_\alpha u_\beta).

Here, {χ(T),θ(T),λ(T)}\{\chi(T),\theta(T),\lambda(T)\} are constants. The dispersion polynomial is1

PBDNK(ω,k)=[ω(1−iχω)+iλk2]×[3ω(1−iθω)−iχk2][3ω(1−iχω)−i(θ−4λ)k2]+3ik2[1−i(χ+θ)ω].(18)P_{\mathrm{BDNK}}(\omega,k)=\left[\omega(1-i\chi\omega)+i\lambda k^2\right]\times\left[3\omega(1-i\theta\omega)-i\chi k^2\right]\left[3\omega(1-i\chi\omega)-i(\theta-4\lambda)k^2\right]+3ik^2[1-i(\chi+\theta)\omega]. \tag*{(18)}

It has one hydrodynamic and one non-hydrodynamic mode in the shear channel, and two hydrodynamic and two non-hydrodynamic modes in the sound channel. To determine the frame transformation as an expansion in ω\omega and kk, it suffices to know the modes in a similar expansion in kk (up to O(k4)\mathcal{O}(k^4))

ωsh=−iθ+iλk2,ωsh=−iλk2,\omega_{\mathrm{sh}}=-\frac{i}{\theta}+i\lambda k^2,\qquad\omega_{\mathrm{sh}}=-i\lambda k^2,
ωsnd(1)=−iθ+i3[4λ−θχθ−χ]k2,\omega_{\mathrm{snd}}^{(1)}=-\frac{i}{\theta}+\frac{i}{3}\left[4\lambda-\frac{\theta\chi}{\theta-\chi}\right]k^2,
ωsnd(2)=−iθ+i3(θχθ−χ)k2,\omega_{\mathrm{snd}}^{(2)}=-\frac{i}{\theta}+\frac{i}{3}\left(\frac{\theta\chi}{\theta-\chi}\right)k^2,
ωsnd(1)=k3−(2iλ3)k2−(2λ233)k3,\omega_{\mathrm{snd}}^{(1)}=\frac{k}{\sqrt{3}}-\left(\frac{2i\lambda}{3}\right)k^2-\left(\frac{2\lambda^2}{3\sqrt{3}}\right)k^3,
ωsnd(1)=−k3−(2iλ3)k2+(2λ233)k3.(19)\omega_{\mathrm{snd}}^{(1)}=-\frac{k}{\sqrt{3}}-\left(\frac{2i\lambda}{3}\right)k^2+\left(\frac{2\lambda^2}{3\sqrt{3}}\right)k^3. \tag*{(19)}

Then using equations (15) and (16), we solve for Fu(T)F_u(T), Gu(T)G_u(T), and Ru(T)R_u(T). Taking ω→0\omega\to0 limit in the 1st equation of (15), we obtain the solution for Ru(k)R_u(k)

Ru(k)=1k2(ωsh+iθωsh).(20)R_u(k)=\frac{1}{k^2}\left(\frac{\omega_{\mathrm{sh}}+\frac{i}{\theta}}{\omega_{\mathrm{sh}}}\right). \tag*{(20)}

Now we put (20) in (15) keeping ω\omega finite to solve Fu(ω,k)F_u(\omega,k)

Fu(ω,k)=−1/θωsh(ω−ωsh).(21)F_u(\omega,k)=-\frac{1/\theta}{\omega_{\mathrm{sh}}(\omega-\omega_{\mathrm{sh}})}. \tag*{(21)}

Then we turn to the second equation of (15) and use Eqs. (20) and (21). Again, taking ω→0\omega\to0 limit of the resulting equation, we receive the solution for RT(k)R_T(k)

RT(k)=iθωsh[FTGu](0)+1k2[1+iχ(ωshωsnd(1)ωsnd(2))].(22)R_T(k)=i\theta\omega_{\mathrm{sh}}[F_TG_u](0)+\frac{1}{k^2}\left[1+\frac{i}{\chi}\left(\frac{\omega_{\mathrm{sh}}}{\omega_{\mathrm{snd}}^{(1)}\omega_{\mathrm{snd}}^{(2)}}\right)\right]. \tag*{(22)}

Finally, putting (22) in the same equation with ω\omega finite, we solve for GT(ω,k)G_T(\omega,k)

GT(ω,k)=θωsh(k2θ)(FTGu−[FTGu](0))G_T(\omega,k)=\theta\omega_{\mathrm{sh}}\left(\frac{k^2}{\theta}\right)\left(F_TG_u-[F_TG_u](0)\right)
−θk2(FTGu)−1χ[ωsnd(1)ωsnd(2)+ωsh(ω−ωsnd(1)−ωsnd(2))ωsnd(1)ωsnd(2)(ω−ωsnd(1))(ω−ωsnd(2))].(23)-\theta k^2(F_TG_u)-\frac{1}{\chi}\left[\frac{\omega_{\mathrm{snd}}^{(1)}\omega_{\mathrm{snd}}^{(2)}+\omega_{\mathrm{sh}}(\omega-\omega_{\mathrm{snd}}^{(1)}-\omega_{\mathrm{snd}}^{(2)})}{\omega_{\mathrm{snd}}^{(1)}\omega_{\mathrm{snd}}^{(2)}(\omega-\omega_{\mathrm{snd}}^{(1)})(\omega-\omega_{\mathrm{snd}}^{(2)})}\right]. \tag*{(23)}

We have two unknowns, FT(ω,k)F_T(\omega,k) and Gu(ω,k)G_u(\omega,k), and two equations (16) evaluated at ωsnd(1)\omega_{\mathrm{snd}}^{(1)} and ωsnd(2)\omega_{\mathrm{snd}}^{(2)}. First we determine R^(ω,k)\hat{\mathcal{R}}(\omega,k) and R(ω,k)\mathcal{R}(\omega,k) as

R^(ω,k)=k[Gu(θk2+3γω2+3iω)+i(1−iωFu−k2Ru)(iθω+γω)]k2[iθω+γω]FT+i(1−k2RT−iωGT)(θk2+3γω2+3iω),\hat{\mathcal{R}}(\omega,k)=\frac{k[G_u(\theta k^2+3\gamma\omega^2+3i\omega)+i(1-i\omega F_u-k^2R_u)(i\theta\omega+\gamma\omega)]}{k^2[i\theta\omega+\gamma\omega]F_T+i(1-k^2R_T-i\omega G_T)(\theta k^2+3\gamma\omega^2+3i\omega)},
R(ω,k)=k[i+(θ+χ)ωk2θω+3i(ω+χω)].\mathcal{R}(\omega,k)=k\left[\frac{i+(\theta+\chi)\omega}{k^2\theta\omega+3i(\omega+\chi\omega)}\right].
R(ωsnd(1),k)=R^(ωsnd(1),k),R(ωsnd(2),k)=R^(ωsnd(2),k).(24)\mathcal{R}(\omega_{\mathrm{snd}}^{(1)},k)=\hat{\mathcal{R}}(\omega_{\mathrm{snd}}^{(1)},k),\qquad\mathcal{R}(\omega_{\mathrm{snd}}^{(2)},k)=\hat{\mathcal{R}}(\omega_{\mathrm{snd}}^{(2)},k). \tag*{(24)}

Note that (24) must be fulfilled only at the frequencies of the hydrodynamic sound modes, ωsnd(1)(k)\omega_{\mathrm{snd}}^{(1)}(k) and ωsnd(2)(k)\omega_{\mathrm{snd}}^{(2)}(k). Solving these equations does not determine the unknown functions FTF_T and GuG_u for all ω\omega and kk. However, our aim is to demonstrate the existence of at least one frame transformation that eliminates the non-hydrodynamic modes. By assuming FTF_T and GuG_u do not depend on ω\omega, equation (24) uniquely solves these functions, providing a suitable frame transformation. With this assumption, the solutions for RTR_T and GTG_T simplify

RT(k)=iθωshFTGu+1k2[1+iχωshωsnd(1)ωsnd(2)](25)R_T(k)=i\theta\omega_{\mathrm{sh}}F_TG_u+\frac{1}{k^2}\left[1+\frac{i}{\chi}\frac{\omega_{\mathrm{sh}}}{\omega_{\mathrm{snd}}^{(1)}\omega_{\mathrm{snd}}^{(2)}}\right] \tag*{(25)}
GT(ω,k)=−θk2(FTGu)−1χ[ωsnd(1)ωsnd(2)+ωsh(ω−ωsnd(1)−ωsnd(2))ωsnd(1)ωsnd(2)(ω−ωsnd(1))(ω−ωsnd(2))].G_T(\omega,k)=-\theta k^2(F_TG_u)-\frac{1}{\chi}\left[\frac{\omega_{\mathrm{snd}}^{(1)}\omega_{\mathrm{snd}}^{(2)}+\omega_{\mathrm{sh}}\left(\omega-\omega_{\mathrm{snd}}^{(1)}-\omega_{\mathrm{snd}}^{(2)}\right)}{\omega_{\mathrm{snd}}^{(1)}\omega_{\mathrm{snd}}^{(2)}\left(\omega-\omega_{\mathrm{snd}}^{(1)}\right)\left(\omega-\omega_{\mathrm{snd}}^{(2)}\right)}\right].

Even after assuming FTF_T and GuG_u are independent of ω\omega, eq. (24) remains a set of coupled nonlinear algebraic equations, with nonlinearity due to RTR_T depending on the product of FTF_T and GuG_u. While solving is straightforward, the exact expressions for FTF_T and GuG_u are cumbersome compared to other functions FuF_u, RuR_u, RTR_T, and GTG_T. The complete details are presented in [29]. The final leading solutions, up to O(k2)O(k^2), are as follows:

Gu(k)=g0,FT(k)=3g0+χ−θ.(26)G_u(k)=g_0,\qquad F_T(k)=3g_0+\chi-\theta. \tag*{(26)}

with g0=−16(5θ2+16λ(λ−θ)+8λχ−3χ(χ+2θ)−θ−4λ+χ)g_0=-\frac{1}{6}\left(\sqrt{5\theta^2+16\lambda(\lambda-\theta)+8\lambda\chi-3\chi(\chi+2\theta)}-\theta-4\lambda+\chi\right). After solving for all the functions in the frame transformation in Fourier space, we can convert them to position space using the following simple replacement: ω→−i(u⋅∂)\omega\to-i(u\cdot\partial), k2→−Pμν∂μ∂νk^2\to-P^{\mu\nu}\partial_\mu\partial_\nu.

At this point, we note that although we provided formal exact solutions for FuF_u, RuR_u, RTR_T, and GTG_T, these should be understood as expansions in non-negative powers of ω\omega and kk. Polynomials of kk and ω\omega in the denominator imply an infinite power series with a convergence radius computed by the zeros of these polynomials. Our study relies on the derivative expansion, and replacing {ω,k2}\{\omega,k^2\} with derivatives doesn’t make sense if ω\omega and kk appear in the denominator. All functions Fu(T)F_u(T), Gu(T)G_u(T), and Ru(T)R_u(T) remain finite as ω→0\omega\to0 and/or k→0k\to0. This ensures no negative powers of (u⋅∂)(u\cdot\partial) or (Pμν∂μ)(P^\mu{}_\nu\partial_\mu) arise after expansion and replacement. Furthermore, the final frame transformation solution must depend only on even powers of kk to apply the replacement without using non-analytic expressions like the square root of a derivative. Since we’re concerned with the spectrum of linearized perturbations, the order of derivatives (u⋅∂)(u\cdot\partial) and Pμν∂μP^{\mu\nu}\partial_\mu is irrelevant. The frame transformation incorporates terms of all orders in the derivative expansion, thus, the final transformed stress-energy tensor also have terms of all orders in the derivative expansion.

Conclusion:– In this letter, we have shown that the spectrum of hydrodynamic modes is unaffected by any redefinition of fluid fields, while the spectrum of non-hydrodynamic modes can change under such redefinitions. Notably, it is possible to eliminate a non-hydrodynamic mode from the spectrum through an appropriate all-order frame redefinition, as illustrated by our example on the first-order BDNK theory. However, this redefinition cannot fully erase the information about any physical non-hydrodynamic mode. The infinite series of the final stress-energy tensor in the ‘hydro’ frame has a radius of convergence located at the lowest non-hydrodynamic mode in the complex frequency and momentum space[30, 31], suggesting that this mode acts as a cutoff in the hydrodynamic expansion[32, 33, 34].

Our analysis implies that in a causal physical theory, removing physical non-hydrodynamic modes requires including all higher-order terms in the energy-momentum tensor without truncation. In Fourier space, the data concerning the physical non-hydrodynamic mode is encapsulated in the radius of convergence of this infinite series[28, 35, 36, 37, 38, 39]. If the energy-momentum tensor is truncated to a finite order, physical non-hydrodynamic modes are necessary for causality. Recent observations indicate that a theory without non-hydrodynamic modes when truncated, may lack causality but remain stable[1, 40, 41, 42, 43]. Our analysis focused on uncharged fluids, where the conservation of the stress tensor is the only hydrodynamic equation of motion. Future work may explore other fluids with various charge conservation equations and MIS-type theories that involve auxiliary fields, such as the shear tensor, which can also be redefined.

We have found that a typical linearized frame redefinition introduces new non-hydrodynamic modes corresponding to the zero eigenspace of the linearized frame transformation. Perturbations with frequencies matching these new modes will vanish after the transformation, trivially satisfying the linearized equations of motion. In the transformed frame, these appear as new modes that modify the UV behavior of the theory without affecting its low-energy physics. We have utilized this characteristic of frame redefinition to construct an equivalent hydrodynamic theory with the non-hydrodynamic modes eliminated from the spectrum of linearized perturbations.

Acknowledgments:– S.B., S.M. and S.R. acknowledge the Department of Atomic Energy, India, for the funding support. R.S. acknowledges kind hospitality and support of INFN Firenze, Department of Physics and Astronomy, University of Florence, ECT* Trento, NCBJ Warsaw, TU Darmstadt, ITP Goethe University where part of this work is completed and is supported partly by postdoctoral fellowship of West University of Timișoara. We acknowledge enlightening discussions with L. Gavassino, M. Spalinski, and A. Yaron.

References

  1. [1]W. A. Hiscock and L. Lindblom, “Generic instabilities in first-order dissipative relativistic fluid theories,” Phys. Rev. D 31 (1985) 725–733.pubmed.ncbi.nlm.nih.gov/9955752
  2. [2]R. P. Geroch and L. Lindblom, “Dissipative relativistic fluid theories of divergence type,” Phys. Rev. D 41 (1990) 1855.
  3. [3]R. P. Geroch, “Relativistic theories of dissipative fluids,” J. Math. Phys. 36 (1995) 4226.DOI
  4. [4]L. Gavassino, M. Antonelli, and B. Haskell, “Thermodynamic Stability Implies Causality,” Phys. Rev. Lett. 128 no. 1, (2022) 010606, arXiv:2105.14621 [gr-qc].DOI
  5. [5]P. Arnold, P. Romatschke, and W. van der Schee, “Absence of a local rest frame in far from equilibrium quantum matter,” JHEP 10 (2014) 110, arXiv:1408.2518 [hep-th].DOI
  6. [6]L. Lehner, O. A. Reula, and M. E. Rubio, “Hyperbolic theory of relativistic conformal dissipative fluids,” Phys. Rev. D 97 no. 2, (2018) 024013, arXiv:1710.08033 [gr-qc].DOI
  7. [7]W. Israel, “Nonstationary irreversible thermodynamics: A Causal relativistic theory,” Annals Phys. 100 (1976) 310–331.DOI
  8. [8]W. Israel and J. M. Stewart, “Transient relativistic thermodynamics and kinetic theory,” Annals Phys. 118 (1979) 341–372.DOI
  9. [9]F. S. Bemfica, M. M. Disconzi, and J. Noronha, “Causality and existence of solutions of relativistic viscous fluid dynamics with gravity,” Phys. Rev. D 98 no. 10, (2018) 104064, arXiv:1708.06255 [gr-qc].DOI
  10. [10]P. Kovtun, “First-order relativistic hydrodynamics is stable,” JHEP 10 (2019) 034, arXiv:1907.08191 [hep-th].DOI
  11. [11]F. S. Bemfica, M. M. Disconzi, and J. Noronha, “Causality of the Einstein-Israel-Stewart Theory with Bulk Viscosity,” Phys. Rev. Lett. 122 no. 22, (2019) 221602, arXiv:1901.06701 [gr-qc].DOI
  12. [12]G. S. Denicol, T. Kodama, T. Koide, and P. Mota, “Stability and Causality in relativistic dissipative hydrodynamics,” J. Phys. G 35 (2008) 115102, arXiv:0807.3120 [hep-ph].DOI
  13. [13]S. Floerchinger and E. Grossi, “Causality of fluid dynamics for high-energy nuclear collisions,” JHEP 08 (2018) 186, arXiv:1711.06687 [nucl-th].DOI
  14. [14]R. Biswas, A. Dash, N. Haque, S. Pu, and V. Roy, “Causality and stability in relativistic viscous non-resistive magneto-fluid dynamics,” JHEP 10 (2020) 171, arXiv:2007.05431 [nucl-th].DOI
  15. [15]R. E. Hoult and P. Kovtun, “Causality and classical dispersion relations,” Phys. Rev. D 109 no. 4, (2024) 046018, arXiv:2309.11703 [hep-th].DOI
  16. [16]J. Sammet, M. Mayer, and D. H. Rischke, “Linear stability analysis of Israel-Stewart theory in the case of a nonzero background charge,” Phys. Rev. D 107 no. 11, (2023) 114028, arXiv:2302.01070 [hep-th].DOI
  17. [17]T. S. Domingues, R. Krupczak, J. Noronha, T. N. da Silva, J.-F. Paquet, and M. Luzum, “The effect of causality constraints on Bayesian analyses of heavy-ion collisions,” arXiv:2409.17127 [nucl-th].arxiv.org/abs/2409.17127
  18. [18]J. Noronha, M. Spaliński, and E. Speranza, “Transient Relativistic Fluid Dynamics in a General Hydrodynamic Frame,” Phys. Rev. Lett. 128 no. 25, (2022) 252302, arXiv:2105.01034 [nucl-th].DOI
  19. [19]S. Mitra, “Acausality in truncated Müller-Israel-Stewart-type theory,” Phys. Rev. D 109 no. 12, (2024) L121501, arXiv:2312.10972 [gr-qc].
  20. [20]S. Bhattacharyya, S. Mitra, and S. Roy, “Frame transformation and stable-causal hydrodynamic theory,” arXiv:2312.16407 [nucl-th].arxiv.org/abs/2312.16407
  21. [21]S. Bhattacharyya, S. Mitra, and S. Roy, “Causality and stability in relativistic hydrodynamic theory - a choice to be endured,” Phys. Lett. B 856 (2024) 138918, arXiv:2407.18997 [nucl-th].arxiv.org/abs/2407.18997
  22. [22]S. Mitra, “Why do newer degrees of freedom appear in higher-order truncated hydrodynamic theory?,” arXiv:2408.15088 [nucl-th].arxiv.org/abs/2408.15088
  23. [23]X.-Q. Xie, D.-L. Wang, C. Yang, and S. Pu, “Causality and stability analysis for the minimal causal spin hydrodynamics,” Phys. Rev. D 108 no. 9, (2023) 094031, arXiv:2306.13880 [hep-ph].arxiv.org/abs/2306.13880
  24. [24]S. Grozdanov, A. Lucas, and N. Poovuttikul, “Holography and hydrodynamics with weakly broken symmetries,” Phys. Rev. D 99 no. 8, (2019) 086012, arXiv:1810.10016 [hep-th].DOI
  25. [25]S. Grozdanov, P. K. Kovtun, A. O. Starinets, and P. Tadić, “The complex life of hydrodynamic modes,” JHEP 11 (2019) 097, arXiv:1904.12862 [hep-th].DOI
  26. [26]Y. Bea and P. Figueras, “Field redefinitions and evolutions in relativistic Navier-Stokes,” arXiv:2312.16671 [hep-th].arxiv.org/abs/2312.16671
  27. [27]R. E. Hoult and A. Shukla, “Causal and Stable Superfluid Hydrodynamics,” arXiv:2410.22855 [hep-th].DOI
  28. [28]M. P. Heller, A. Serantes, M. Spaliński, and B. Withers, “Rigorous Bounds on Transport from Causality,” Phys. Lett. 130 no. 26, (2023) 261601, arXiv:2212.07434 [hep-th].DOI
  29. [29]S. Bhattacharyya, S. Mitra, S. Roy, and R. Singh, “Field redefinition and its impact in relativistic hydrodynamics,” arXiv:2409.15387 [nucl-th].DOI
  30. [30]M. P. Heller, R. A. Janik, and P. Witaszczyk, “Hydrodynamic Gradient Expansion in Gauge Theory Plasmas,” Phys. Rev. Lett. 110 no. 21, (2013) 211602, arXiv:1302.0697 [hep-th].DOI
  31. [31]M. P. Heller, A. Serantes, M. Spaliński, V. Svensson, and B. Withers, “Hydrodynamic gradient expansion in linear response theory,” Phys. Rev. D 104 no. 6, (2021) 066002, arXiv:2007.05524 [hep-th].DOI
  32. [32]S. Grozdanov, P. K. Kovtun, A. O. Starinets, and P. Tadić, “Convergence of the Gradient Expansion in Hydrodynamics,” Phys. Rev. Lett. 122 no. 25, (2019) 251601, arXiv:1904.01018 [hep-th].DOI
  33. [33]M. P. Heller, A. Serantes, M. Spaliński, V. Svensson, and B. Withers, “Convergence of hydrodynamic modes: insights from kinetic theory and holography,” SciPost Phys. 10 no. 6, (2021) 123, arXiv:2012.15393 [hep-th].DOI
  34. [34]C. Cartwright, M. G. Amano, M. Kaminski, J. Noronha, and E. Speranza, “Convergence of hydrodynamics in a rotating strongly coupled plasma,” Phys. Rev. D 108 no. 4, (2023) 046014, arXiv:2112.10781 [hep-th].arxiv.org/abs/2112.10781
  35. [35]F. S. Bemfica, F. S. Bemfica, M. M. Disconzi, M. M. Disconzi, J. Noronha, and J. Noronha, “Nonlinear Causality of General First-Order Relativistic Viscous Hydrodynamics,” Phys. Rev. D 100 no. 10, (2019) 104020, arXiv:1907. 12695 [gr-qc]. [Erratum: Phys.Rev.D 105, 069902 (2022)].
  36. [36]L. Gavassino, “Can We Make Sense of Dissipation without Causality?," Phys. Rev. X 12 no. 4, (2022) 041001, arXiv:2111. 05254 [gr-qc].DOI
  37. [37]L. Gavassino, “Bounds on transport from hydrodynamic stability,” Phys. Lett. B 840 (2023) 137854, arXiv:2301. 06651 [hep-th].DOI
  38. [38]L. Gavassino, “Infinite Order Hydrodynamics: An Analytical Example,” Phys. Rev. Lett. 133 no. 3, (2024) 032302, arXiv:2402. 19343 [nucl-th].DOI
  39. [39]D.-L. Wang and S. Pu, “Stability and causality criteria in linear mode analysis: Stability means causality,” Phys. Rev. D 109 no. 3, (2024) L031504,arxiv.org/abs/2309.11708
  40. [40]I. Novak, J. Sonner, and B. Withers, “Hydrodynamics without boosts,” JHEP 07 (2020) 165, arXiv:1911. 02578 [hep-th].DOI
  41. [41]J. Armas and A. Jain, “Effective field theory for hydrodynamics without boosts,” SciPost Phys. 11 no. 3, (2021) 054, arXiv:2010. 15782 [hep-th].DOI
  42. [42]G. Başar, J. Bhambure, R. Singh, and D. Teaney, “Stochastic relativistic advection diffusion equation from the Metropolis algorithm,” Phys. Rev. C 110 no. 4, (2024) 044903, arXiv:2403. 04185 [nucl-th].DOI
  43. [43]S. Pu, T. Koide, and D. H. Rischke, “Does stability of relativistic dissipative fluid dynamics imply causality?,” Phys. Rev. D 81 (2010) 114039, arXiv:0907.3906 [hep-ph].DOI

Paper details

vlnc:W_gy74g33u
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 6

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