betaFeedback

WorksPhysics and Astronomy

Gaussian fluctuating Generally covariant diffusion

Abstract

We extend the previously developed [1] generally covariant formalism to include diffusion of conserved charges, and comment on the seming difference between the chemica potential term and the diffusion term

The problem of relativistic diffusion

Relativistic diffusion is of considerable phenomenological and theoretical interest, being the leading non-equilibrium correction in transport of conserved charges in both heavy ion collisions and neutron stars [2, 3, 4, 5]. It is also an important standalone aspect of the theoretical study of non-equilibrium relativistic statistical mechanics.

It is often said [6, 7, 8] that diffusion is the simplest hydrodynamic-like theory, because it is governed by a conservation law equation, combined with thermodynamics. However, diffusion has no flow but just, as the name says, diffusive motion, A conserved current JνJ_\nu's evolution is defined by

∂νJν=0 ,J⃗i=−D∇⃗μQ(1)\partial_\nu J^\nu= 0 \,, \qquad\vec{J}_i = -D\vec{\nabla}\mu_Q \tag*{(1)}

where μ\mu is the chemical potential, a Lagrange multiplier enforcing charge conservation of the equilibrium distribution. Just like in hydrodynamics, conservation laws and an expression derivable from statistical mechanics, namely the dependence of the diffusion coefficient DD from the chemical potential μQ\mu_Q, is enough to close the equations and make the system solvable from arbitrary initial conditions. The form of DD is usually thought to arise from a Kubo type formula

D=lim⁡k→01kIm⁡∫eikx⟨[Jx(x,t),Jx(0,0)]⟩ d4x(2)D = \lim_{k\to0} \frac{1}{k}\operatorname{Im}\int e^{ikx}\langle[J_x(x,t),J_x(0,0)]\rangle\,d^4x \tag*{(2)}

In the full local thermalization limit D→0D \to0 no diffusion occurs (physically “many“ microscopic collisions stop the particle).

This theory is much simpler than hydrodynamics (where Jμ=Tρβμ+O(∂)J_\mu= T\rho\beta_\mu+ O(\partial) ... and ρ\rho is the density while βμ=uμ/T\beta_\mu= u_\mu/T is the flow divided by temperature), as, because of the absence of flow, it‘s leading order form is linear, and solvable exactly with Greens functions techniques.

It also has the usual issues of theories based on thermalization have with Lorentz invariance and causality[9], and,because of its simplicity, is a good laboratory to overcome them. A usual way to address them is [10–15] via a relaxational dynamics of the Maxwell-Cattaneo type

τQdJ⃗dt+J⃗=J⃗diffusion=−D∇⃗μ(3)\tau_Q \frac{d\vec{J}}{dt} + \vec{J} = \vec{J}_{diffusion} = -D\vec{\nabla}\mu \tag*{(3)}

which restores causality for a long enough τQ\tau_Q w.r.t. D\sqrt{D} at the price of promoting JνJ_\nu to an independent degree of freedom w.r.t. μQ\mu_Q, albeit one with relaxational dynamics only (if hydrodynamics is included, J⊥J_\perp, the perpendicular component of JJ w.r.t. flow, haa the role of such a degree of freedom). However the Lorentz invariance issue never really goes away, since there is a “special“ frame in which the chemical potential is defined. At least in the microcanonical limit, the underlyng dynamics should however be Lorentz invariant (and in the canonical and grand canonical limit dynamics should be in terms of the relative velocity between the system and the bath).

In [1] [16, 17, 18] a way to restore Lorentz invariance was proposed in the context of fluid dynamics: Since ideal fluid behavior is equivalent to local ergodicity in any foliation [17], one should be able to restore Lorentz invariance by assuming any dissipative dynamics is locally indistinguishable from a fluctuation. Thus, if one treats fluctuations and average values on the same footing, one can impose Lorentz invariance by imposing such a local equivalence relation. In this work we shall apply this framework to the simplified problem of diffusion. In (§ 0.2) we qualitatively describe the relationship between non-perturbative fluctuations and general covariance, in Section (III) we show that the issues described above can b clarified in this approach. Finally in Section (IV) we discuss how our calculation fits in the results of [1] as well as other approaches of relativistic statistical mechanics. Finally, in the appendix we make a detailed comparison between the approach presented in this work and the more usual expansion around a thermostatic background on which gradient and Schwinger-Keldysh effective theories are based on

Fluctuations and general covariance

The physical interpretation of this general covariance is that ultimately the “microscopic physics“ is ergodic for a perfectly thermalized system, but, because of the “andromeda paradox“ [19] such ergodicity is frame dependent. Only by promoting Lorentz invariance to invariance under all possible foliations does a Lorentz-invariant definition of local equilibrium emerge[17]. This definition is microcanonical, but can be generalized to the canonical [1] ensemble provided fluctuations in the relative velocity between the system and the bath are accounted for in a Lorentz-invariant way. For dynamics to be generally covariant, it has to be governed by the Ward identity. In [1] it was shown that this identity, together with linear response, was enough to specify the dynamical evolution of the system provided the partition function defined at every cell has a Gaussian form. Note that this does not mean that 3-point functions and higher are not there, but just that they are exclusively functions of 2-point functions, as shown quantitatively in [14].

A simple justification of the Gaussian ansatz is that, since general covariance requires fluctuations, this ansatz is the simplest that can explicitly accomodate it. Furthermore, Gaussian partition functions seem to work well even in strongly coupled dynamics [20] and are also the only finite width distribution which is a fixed point of renormalization group flows [21] (a result similar to Donsker’s theorem in mathematics) we believe it to be a reasonable ansatz in the absence of critical points. What this means physically is that taking the lattice spacing on which to define the Gaussian partition functions to zero will not result in additional terms but in the “flow“ of the average and width to a specific fixed point [21]. It is worth noting that, as shown in the appendix, existing thermostatic backreaction approaches [8] [11, 14] are also Gaussian in the sense that ”non-Gaussian terms” either regularize the two leading cumulants (as in stochastic diffusion), or just the first cumulant (when an effective deterministic theory is written out as a gradient expansion, reflecting the spectral analysis of the Kubo correlator). Where our approach differs is that fluctuations are non-perturbative, and the general covariance underlying the local Lorentz invariance of the ”ergodic” microscopic dynamics is exact. This makes for a much more complicated dynamics, where the whole ensemble needs to be evolved from initial conditions (In [8] [11, 14] it is directly computable from the action), but avoids a thermostatic background thereby respecting the microscopic symmetries of the dynamics. Note also that going beyond the Gaussian approximation is possible, provided higher order Ward identities and response relations are available.

Here we want to apply the methods of [1] to diffusion of conserved charges. We would like to immediately clarify that in a sense this is an academic exercise: The physical applicability would be a system whose energy-momentum fluctuation and dissipation scales are very short, but conserved charge fluctuation and dissipation scales are very long. We are not aware of such a system in nature, for instance in the cited nuclear and astrophysic application [2, 3, 4, 5] diffusion of conserved charge is of the same order as diffusion of momentum.

Nevertheless, the importance of diffusion as the simplest hydrodynamic-like theory motivates us to see how it looks like in this new ansatz, and also to see how the simplification of diffusion w.r.t. full hydrodynamics reflects here. As we shall see, indeed the simpler structure of diffusion means the dynamics will have some qualitative differences.

Gaussian fluctuating generally covariant diffusion

We consider JμJ_\mu to be an operator, endowed with a partition function, whose conserved charge part is

Z=Tr⁡exp⁡[−∫ΣdΣν(μQJ^ν)],(4)Z = \operatorname{Tr} \exp\left[-\int_{\Sigma} \mathrm{d}\Sigma_\nu\left(\mu_Q \hat{J}^{\nu}\right)\right], \tag*{(4)}

(where the trace denotes a functional integral over all microscopic field configurations)

moreover, we postulate that this partition function is approximately gaussian

ZJ≃exp⁡[−(Jμ−⟨Jμ(x,t)⟩)(Jν−⟨Jν(x,t)⟩)Dμν(x,t)](5)Z_J \simeq\exp\left[-\frac{\left(J_\mu-\langle J_\mu(x,t)\rangle\right)\left(J_\nu-\langle J_\nu(x,t)\rangle\right)}{D^{\mu\nu}(x,t)}\right] \tag*{(5)}

Thus we obtain equations of motion for ⟨Jμ(x,t)⟩\langle J_\mu(x,t)\rangle, Dμν(x,t)D^{\mu\nu}(x,t) from the following requirements

General covariance: ⟨Jμ(x,t)⟩\langle J_\mu(x,t)\rangle, Dμν(x,t)D^{\mu\nu}(x,t) should transform covariantly under general space-time refoliations

Fluctuation/dissipation relation: representing thermal equilibrium. Locally, one cannot tell whether a deviation from equilibrium is an ergodic fluctuation or dissipative dynamics.

Charge conservation: for every element of the ensemble. This will be enforced by a Ward identity

The first requirement comes from the definition of the global partition function using the Zubarev formalism Eq. (4) must hold for arbitrary cell choices d3Σμd^{3}\Sigma_{\mu}.

if Σμ\Sigma_{\mu} is a space-like 4-vector [17, 18]

Σμ=(t(x),x⃗)=(t(ϕI),ϕI=1,2,3),(6)\Sigma_{\mu} = (t(x),\vec{x}) = (t(\phi_I),\phi_{I=1,2,3}), \tag*{(6)}

(in its own rest frame Σμ=(0,x,y,z)\Sigma_{\mu}=(0,x,y,z)) its exterior derivative (in its own rest-frame dΣμ=(dV,0,0,0)d\Sigma_{\mu}=(dV,0,0,0)) is time-like. which also implies a general metric,

gμν=∂Σμ∂Σν,d3Σμ=dϕIdϕJdϕKϵμναβϵIJK∂Σν∂ϕI∂Σα∂ϕJ∂Σβ∂ϕK,(7)g_{\mu\nu} = \frac{\partial\Sigma_{\mu}}{\partial\Sigma^{\nu}}, \qquad d^{3}\Sigma_{\mu}=d\phi_I d\phi_J d\phi_K\epsilon_{\mu\nu\alpha\beta}\epsilon_{IJK}\frac{\partial\Sigma^{\nu}}{\partial\phi_I}\frac{\partial\Sigma^{\alpha}}{\partial\phi_J}\frac{\partial\Sigma^{\beta}}{\partial\phi_K}, \tag*{(7)}

It then follows that, since gμν=∂Σμ/∂Σνg_{\mu\nu}=\partial\Sigma_{\mu}/\partial\Sigma_{\nu}, that general covariance automatically implies invariance under d3Σμd^{3}\Sigma_{\mu}.

The linear response equation, written in a covariant form, is therefore

J^μ(Σ0)=⟨Jμ(Σ0)⟩+∫dΣ0′Dμν(Σ,Σ′)δA^ν(Σ0′)dΣ0′,Dμν(Σ,Σ′)=⟨[Jμ(Σ0)Jν(Σ0′)]⟩(8)\hat{J}_{\mu}(\Sigma_0)=\langle J_{\mu}(\Sigma_0)\rangle+\int d\Sigma'_0D_{\mu\nu}(\Sigma,\Sigma')\delta\hat{A}^{\nu}(\Sigma'_0)d\Sigma'_0,\qquad D_{\mu\nu}(\Sigma,\Sigma')=\langle[J_{\mu}(\Sigma_0)J_{\nu}(\Sigma'_0)]\rangle \tag*{(8)}

note that if Σ0\Sigma_0 is a space-like direction, one can regard this as an auto-correlation, if it is in a time-like direction, it will be a linear response evolution subject to causality. The two situations can be incorporated in the same object by analytically continuing Dμν(Σ,Σ′)D_{\mu\nu}(\Sigma,\Sigma') from Σ0\Sigma_0 to Σ0(1+iϵ)\Sigma_0(1+i\epsilon) in the time-like direction, as in [6].

The LHS of Eq. (8) has a hat, because we have an ensemble of configurations of JμJ_{\mu} at every point of spacetime. We can therefore think of Eq. (8) as a representation of a “typical“ ensemble element. It will be the average, plus a random term represented by a random excitation of an auxiliary Gauge field, represented by δA^ν(Σ0′)\delta\hat{A}^{\nu}(\Sigma'_0). But now we have to remember that conservation of charge is exact, true for every element of the ensemble but not just the average.

We therefore contract Eq. (8) with a volume element, and enforce charge conservation

ddΣ0[J^μ(Σ0)d3Σμ]=0ddΣ0Q=0(9)\frac{d}{d\Sigma_0}\left[\hat{J}_{\mu}(\Sigma_0)d^{3}\Sigma^{\mu}\right]=0 \qquad\frac{d}{d\Sigma_0}Q=0 \tag*{(9)}

This means that there is a locally conserved quantity δQ(x)\delta Q(x), defined and conserved at the ensemble level, which is also a scalar under general refoliations

δQ=⟨Jμ(Σ0)⟩d3Σμ+d3Σμ∫dΣ0′Dμν(Σ,Σ′)δA^ν(Σ0′)(10)\delta Q=\langle J_{\mu}(\Sigma_0)\rangle d^{3}\Sigma^{\mu}+d^{3}\Sigma^{\mu}\int d\Sigma'_0D_{\mu\nu}(\Sigma,\Sigma')\delta\hat{A}^{\nu}(\Sigma'_0) \tag*{(10)}

note that locally d3ΣμdΣ0d^{3}\Sigma_{\mu}d\Sigma_{0} is a scalar, so d3ΣμdΣ0′d^{3}\Sigma_{\mu}d\Sigma'_{0} reflects “tidal“ effects of metric deformations, it would be a second derivative. However, δA^ν\delta\hat{A}^{\nu} is simply an auxiliary field (In case of hydrodynamics[1] it is a physical metric perturbation )

Physically, Eq. (10) means that in an arbitrarily defined “fluid cell“ one might observe momentary violations of charge, which could be statistical fluctuations (you did not know what the charge was in the first place!) or diffusive exchanges with the rest of the system. Changing the cell definition (i.e. the foliation) changes the preferred interpretation, but it should have no effect on the dynamics (this interpretation ambiguity is at the heart of why the ”Andromeda paradox” ambiguity [19] does not affect dynamics. It would if the system was purely deterministic).

So, QQ is the constraint between the average and fluctuation of the ensemble that enforces QQ conservation for each element of the ensemble. For instance, the classic linear response result eq. 2.5 of [8] is reproduced for δQ=0\delta Q = 0 and a thermal background, d3Σμ=(d3x,0⃗)d^{3}\Sigma_{\mu} = (d^{3}x,\vec{0}).

Eq. Eq. (8) together with Eq. (10),together with a Gaussian ansatz (which will mean ⟨[Jμ(Σ0)Jν(Σ0′)]⟩\langle[J_{\mu}(\Sigma_{0})J_{\nu}(\Sigma'_{0})]\rangle is a function of Dμν(x,t)D_{\mu\nu}(x,t)) will allow ⟨Jμ⟩(x,t)\langle J_{\mu}\rangle(x,t), Dμν(x,t)D_{\mu\nu}(x,t) to be integrated from an initial condition ⟨Jμ⟩(x,t0)\langle J_{\mu}\rangle(x,t_{0}), Dμν(x,t0)D_{\mu\nu}(x,t_{0}). The dynamics will be foliation independent, although of course the directions of ⟨Jμ⟩(x,t0)\langle J_{\mu}\rangle(x,t_{0}), Dμν(x,t0)D_{\mu\nu}(x,t_{0}) will transform covariantly with the foliation.

To see how it works, we note that δAμ\delta A^{\mu} is an auxiliary field, whose energy density must be positive. Without loss of generality, we therefore introduce a foluation metric1

d3Σμ′=d3xδAμAαAα,gμν=δΣμ′δΣν(11)d^{3}\Sigma'_{\mu}=d^{3}x\frac{\delta A_{\mu}}{\sqrt{A_{\alpha}A^{\alpha}}}, \qquad g_{\mu\nu}=\frac{\delta\Sigma'_{\mu}}{\delta\Sigma^{\nu}} \tag*{(11)}

Eq. (10) then becomes independent of auxiliary gauge fields

d3Qd3Σμ=⟨Jμ(Σ0)⟩+∫dΣ0′d3Σμ′Dμν(Σ,Σ′)=0(12)\frac{d^{3}Q}{d^{3}\Sigma_{\mu}}=\langle J_{\mu}(\Sigma_{0})\rangle+\int d\Sigma'_{0}d^{3}\Sigma'_{\mu}D_{\mu\nu}(\Sigma,\Sigma')=0 \tag*{(12)}

DμνD_{\mu\nu} can be obtained from D~μν\tilde{D}_{\mu\nu} by the usual procedure [1, 6] (f~\tilde{f} is the Fourier transform of ff)

D~μν=12i(D~μν(Σ0,k)D~μν(−iϵΣ0,k)−1)(13)\tilde{D}_{\mu\nu}=\frac{1}{2i}\left(\frac{\tilde{D}^{\mu\nu}(\Sigma_{0},k)}{\tilde{D}_{\mu\nu}(-i\epsilon\Sigma_{0},k)}-1\right) \tag*{(13)}

Since dΣ0dΣ1dΣ2dΣ3d\Sigma_0d\Sigma_1d\Sigma_2d\Sigma_3 is a Lorentz scalar, one can see that the integration measure is a Lorentz scalar times a unit 4-vector, ensuring the right transformation properties.

Thus the system of equations can be closed without the need for volume-preserving diffeomorphisms (as in [1]). Also, in a sense the Ward identity and the fluctuation-dissipation relation needed for the propagation of the ensemble of JJ coincide, a feature of the much simpler structure of JμJ_\mu w.r.t. TμνT_{\mu\nu}. Because flow and co-moving frames do not appear, the violation of general covariance by the integration measure pointed out in [1] (and argued heuristically there to be small) does not appear here.

Discussion

We have achieved the aim set forward in § 0.2 Given an ensemble {Jμ}\{J_\mu\}, characterized by an initial value of ⟨Jμ(x,t0)⟩\langle J_\mu(x,t_0)\rangle, Dμν(x,t0)D_{\mu\nu}(x,t_0) Eq. (12) together with an analytical continuation can be evolved (i.e. ⟨Jμ(x,t)⟩\langle J_\mu(x,t)\rangle, Dμν(x,t)D_{\mu\nu}(x,t) can be obtained for arbitrary t>t0t>t_0) in a generally covariant way (i.e. the evolution of the ensemble as a whole will not depend on the foliation chosen). For such a calculation, however, the thermal spectral function of ⟨[Jμ(k),Jν(k′)]⟩\left\langle\left[J_\mu(k),J_\nu(k')\right]\right\rangle is needed to arbitrary order (the thermal average can then be transformed to arbitrary foliations using the tensor technique, see Fig. 2 and Fig. 3 of [1]). This is a good illustration of the fact that the cumulant and gradient expansion do not commute.

A numerical implementation of this model, just as of [1], is therefore well beyond the scope of this work. As shown in [1], however, such a dynamics has a simple physical interpretation in terms of Crooks fluctuation theorem [18]. Given a proper time dτ2∝d3Σμd3Σμd\tau^2 \propto d^3\Sigma_\mu d^3\Sigma^\mu (the constant of proportionality is a scale of the dimension of a 4-volume, hence a Lorentz scalar), the probability of a configuration going forward P({μQ},τ+dτ∣{μQ′},τ)P(\{\mu_Q\},\tau+d\tau\mid\{\mu_Q'\},\tau) rather than in reverse P({μQ′},τ+dτ∣{μQ},τ)P(\{\mu_Q'\},\tau+d\tau\mid\{\mu_Q\},\tau) is the exponential of the relative entropy, which in the case of diffusion alone is related trivially to the chemical potential via the susceptibility χQ\chi_Q

P({μQ},τ+dτ∣{μQ′},τ)P({μQ′},τ+dτ∣{μQ},τ)=exp⁡[S({μQ′})−S({μQ})](14)\frac{P(\{\mu_Q\},\tau+d\tau\mid\{\mu_Q'\},\tau)}{P(\{\mu_Q'\},\tau+d\tau\mid\{\mu_Q\},\tau)}=\exp\left[S(\{\mu_Q'\})-S(\{\mu_Q\})\right] \tag*{(14)}

where the entropy of a field of {μQ}\{\mu_Q\} defined locally in Σ\Sigma is defined in terms of the volume hypersurface element d3Σμd^3\Sigma_\mu and it's normal unit vector nμn_\mu

S({μQ})=∫d3Σμnμ1T(Σ)∫0ρQ(Σ)ρQ(T,μQ′(Σ))dμQ′(15)S(\{\mu_Q\})=\int d^3\Sigma_\mu n^\mu\frac{1}{T(\Sigma)}\int_0^{\rho_Q(\Sigma)}\rho_Q(T,\mu'_Q(\Sigma))d\mu'_Q \tag*{(15)}

(In hydrodynamics βμ\beta_\mu would take the place of nμn_\mu).

In the full local equilibration limit, Dμν→0D_{\mu\nu} \to0 and ⟨J⟩\langle J\rangle is conserved exactly, as expected. Note that in this limit both fluctuations and deviations from equilibrium vanish together, as expected from the general covariance requirement. Close to local equilibrium, for small fluctuations phenomena such as long-time tails should appear [13]. However, in parallel to [1], in a strongly coupled regime with non-negligible fluctuations long-time tails might become less likely, since fluctuations of the mean background foliation can not be neglected. Such fluctuations would need to be subtracted from any long-time tail signal and, analogously to ghosts in gauge theory correlators, could weaken the “physically observable“ tails discussed in [13].

One can also view Eq. Eq. (12) as a Wiener process

Xt=μt+σW^t(16)X_t = \mu t + \sigma\widehat{W}_t \tag*{(16)}

with the role of Xt=Qτ→Jμd3ΣμX_t = Q_\tau\to J_\mu d^3\Sigma_\mu the drift μ→⟨Jμ⟩d3Σμ\mu\to\langle J_\mu\rangle d^3\Sigma_\mu and the stochastic variation Σ→Dμνd3Σμd3Σν′\Sigma\to D_{\mu\nu}d^3\Sigma^\mu d^3\Sigma'_\nu. In such a picture, the long-time evolution of ⟨Jμ⟩\langle J_\mu\rangle must be, because of general covariance, independent of the exact slicing of Σμ\Sigma_\mu. It is clear that this requirement is exactly equivalent to the property of hyperbolicity for the partial differential equation governing ⟨Jμ⟩\langle J_\mu\rangle. Thus the requirement of strong hyperbolicity [22, 23] is more general than kinetic theory embedding [24] and is inextricably linked with the Lorentz invariance of the underlying dynamics. We note that the non-Relativistic average only diffusion equation, much like the non-relativistic Navier-Stokes equation lacks this property, as expected since there absolute time and space have quite different roles. It is thus not surprising that historic development of diffusion and hydrodynamic equations found it so difficult to accommodate local Lorentz invariance, which nevertheless must be there since microscopic properties that lead to approximate ergodicity are Lorentz invariant. The missing ingredient was that fluctuations are an integral part of the dynamics.

Note that hydrodynamic flow makes no appearance in these equations. This is because hydrodynamic flow (or more correctly βμ\beta_\mu) can be thought-of as Lagrange multiplier for 4-momentum current. But here we are only concerned with a conserved scalar charge. In this limit, because there is no flow, there is also no freedom to redefine the flow frame, which is the basis for the general covariance symmetry [16] on which this work is based. However, the discussion here makes it clear that, as long as the conversed charge is a Lorentz scalar and fluctuations are included, general covariance is realized without considering the flow frame, just with the foliation.

However the absence of flow also implies that a notion of local equilibrium in a purely diffusion theory is not possible. This is not surprising: As shown in [9] the zeroth law of thermodynamics, on which the notion of equilibrium is based, can be defined relativistically via the existence of a killing vector, which is equivalent to a hydrodynamic flow velocity. [9] also notes that this killing vector can be defined only if the two volume cells exchange energy momentum vis uncharged mediators, while diffusion presupposes a Conserved charge.

Away from the purely diffusive limit, in case both flow and diffusion are present, and the partition function has the form

Z=Tr⁡exp⁡[−d3Σν(βμTμν(ϕ)+μQJμ(ϕ))]≃≃exp⁡[−Cμν,α,β(Σ,Σ′)(Tμν(Σ)−⟨Tμν(Σ′)⟩)(Tαβ(Σ′)−⟨Tαβ(Σ′)⟩)]×exp⁡[−Dμν(Jμ(Σ)−⟨Jμ(Σ)⟩)(Jν(Σ′)−⟨Jν(Σ′)⟩)](17)\begin{aligned} \mathcal{Z} &= \operatorname{Tr}\exp\left[-d^{3}\Sigma_{\nu}\left(\beta_{\mu}T^{\mu\nu}(\phi)+\mu_{Q}J^{\mu}(\phi)\right)\right]\simeq\\ &\simeq\exp\left[-C_{\mu\nu,\alpha,\beta}(\Sigma,\Sigma')\left(T^{\mu\nu}(\Sigma)-\langle T^{\mu\nu}(\Sigma')\rangle\right)\left(T^{\alpha\beta}(\Sigma')-\langle T^{\alpha\beta}(\Sigma')\rangle\right)\right]\\ &\quad\times\exp\left[-D_{\mu\nu}\left(J^{\mu}(\Sigma)-\langle J^{\mu}(\Sigma)\rangle\right)\left(J^{\nu}(\Sigma')-\langle J^{\nu}(\Sigma')\rangle\right)\right] \tag*{(17)} \end{aligned}

The calculations goes much as before, but then the fluctuation tensor DαβD_{\alpha\beta} will be in the frame at rest with the flow, just like the object Cαβμγ′C'_{\alpha\beta\mu\gamma} on which dynamics depends in [1].

The energy-momentum and conserved current term decouple but, because this frame is not in general the same as where DαβD_{\alpha\beta} is diagonal, we will have [1]

\ln\mathcal{Z}(\Sigma)=\int_{\Sigma}d\ln\mathcal{Z}(\Sigma),\qquad d\mathcal{Z}(d\Sigma)=\left(\sqrt{\prod_i\lambda_i^2\sqrt{C'_{\alpha\beta}^{\alpha\beta}}}\right)^{-1}\times\left(\sqrt{\prod_i\langle J_i\rangle^2\times\det_{\alpha\beta}|D_{\alpha\beta}|}\right)^{-1} \tag*{(18)}

where C′C', as in [1], is the correlator in the rest frame w.r.t. βμ\beta_{\mu} and λi\lambda_i are the eigenvalues (related to the energy density and pressure). In the Wiener picture Eq. (16) we need two correlated 4-vectors dΣμd\Sigma_{\mu} and βμ\beta_{\mu} to maintain Lorentz covariance. However, as long as these vectors are chosen appropriately (as argued in [1]) the evolution will be independent of the choice of Σμ\Sigma_{\mu}. This is a demonstration that βμ\beta_{\mu} can be somewhat arbitrary as long as long term averages in fluctuating hydrodynamics must obey hyperbolic partial differential equations, something argued from the purely deterministic theory [22, 23]. Note that in the non-relativistic limit hyperbolicity is not necessary since there is a preferred foliation (that going forward in the absolute time) and indeed non-relativistic hydrodynamics is non-hyperbolic.

As in [1] the language seems very different from usual hydrodynamics (the appendix provides a matching), but it is in fact representing the same thing, with the addition that fluctuations are included non-perturbatively. In particular the component of DμνD_{\mu\nu} perpendicular to velocity is the one associated with the diffusion constant in the usual Kubo formulae.

D=lim⁡k,ω→01kIm⁡∫eiωt−kxD⊥(Σ=(t,x),Σ′(0,0)),βμD⊥μν=0(19)D = \lim_{k,\omega\to0} \frac{1}{k}\operatorname{Im}\int e^{i\omega t-kx}D_{\perp}\left(\Sigma=(t,x),\Sigma'(0,0)\right), \qquad\beta_{\mu}D_{\perp}^{\mu\nu}=0 \tag*{(19)}

However, in the full theory fluctuations in energy and conserved charge will intertwine in a way depending on the equation of state, leading to a complicated correlation pattern between fluid cells not readily captured by such asymptotic expansions. In analogy to [1], it is reasonable to suppose that in the strongly coupled but fluctuating limit D⊥D_{\perp} is smaller than the Kubo valu because the flow fluctuations and diffusion fluctuations cancel out to some extent when the physical J,T correlators are calculated (this is a similar argument to the one used above to say that long-time tails are suppressed).

The discussion above makes it apparent why no equivalent of the pseudo-gauge symmetry exists for 1-currents (this issue was discussed in [25]). A redefinition Jμ→Jμ+∂νBμνJ_{\mu}\to J_{\mu}+\partial_{\nu}B^{\mu\nu} (as made in [25]), where BμνB^{\mu\nu} is a fully antisymmetric tensor, would be irrelevant since ΓμαβBμν=0\Gamma^{\mu\alpha\beta}B_{\mu\nu}=0; Ultimately this reflects the metricity condition in Riemannian geometry, so changing Σμ\Sigma_{\mu} would produce no effect on the dynamics. This is of course not true if the energy-momentum tensor is thought to have an anti-symmetric component, representing a near-equilibrium spin densityThe analysis of this situation has been conducted elsewhere [26].

The extension of the Gaussian approach to chemical potential opens up the perspective to study systems with phase transitions and critical points [27]. For this, however, one needs to explicitly go beyond Gaussian dynamics [28], and consider parititon functions containing non-trivial Skewness and Kurtosis. This is in principle possible along a more involved procedure of the one outlined here, using 4th order Ward identities and quadratic response theories (4 parameters, corresponding to the 4 cumulants, 2nd and 4th order Ward identities and linear and quadratic transport coefficients should close the stochastic equations). However, in this case the partition function can not be integrated in closed form, and the evolution equations are likely to be coupled integro-differential equations. This is a much more involved problem, left for future work.

In conclusion, we extended the formalism of [1] to include diffusion of conserved charge quantities. This reinforces the conclusions made in that work, namely that for full covariance average quantities and fluctuations must be considered on an equal bases, and for idal behavior the two must go to zero together respecting conservation laws. As a result, it is not automatic that systems with more degrees of freedom are also the closest to local equilibrium.

We are in theprocess to extend this formalism further, encompassing spin [26], multiple conserved currents with non-trivial symmetry groups, and gauge symmetries [29],and eventually to a non-trivial phase diagram.

This work was was initiated while on a visit to the School of Nuclear Science and Technology, Lanzhou University, as an answer to questions from Navid Abbasi. it would have been impossible without Navid‘s hospitality and fruitfuil,insightful and deep discussions. GT thanks Bolsa de produtividade CNPQ 305731/2023-8 and FAPESP 2023/06278-2as well as participation in the tematico 2023/13749-1 for support.

Stochastic model for linearized hydrodynamics in a hydrostatic background

Diffusion Correlation function without noise

The linear response equation, written in a covariant form, is

Dμν(Σ,Σ′)=−iΘ(Σ0′−Σ0)Gμν(Σ,Σ′)=⟨[Jμ(Σ),jν(Σ′)]⟩=∫d4k(2π)4χμν′′eikr−iωt(20)D_{\mu\nu}(\Sigma,\Sigma')=-i\Theta(\Sigma'_0-\Sigma_0)G_{\mu\nu}(\Sigma,\Sigma')=\langle[J^\mu(\Sigma),j^\nu(\Sigma')]\rangle=\int\frac{d^4k}{(2\pi)^4}\chi''_{\mu\nu}e^{ikr-i\omega t} \tag*{(20)}

where χμν′′\chi''_{\mu\nu} is the absorptive part. It is convenient to work with the full complex dynamic susceptibility χμν=χμν′+iχμν′′\chi_{\mu\nu}=\chi'_{\mu\nu}+i\chi''_{\mu\nu}. We employ the Kramers-Kronig relation to evaluate the real and imaginary part

χμν(k,ω)=−P∫dω′πχμν′′(k,ω′)ω′−z,χμν′′(k,ω)=−P∫dω′πχμν′(k,ω′)ω′−ω(21)\chi_{\mu\nu}(k,\omega)=-\mathcal{P}\int\frac{d\omega'}{\pi}\frac{\chi''_{\mu\nu}(k,\omega')}{\omega'-z},\qquad\chi''_{\mu\nu}(k,\omega)=-\mathcal{P}\int\frac{d\omega'}{\pi}\frac{\chi'_{\mu\nu}(k,\omega')}{\omega'-\omega} \tag*{(21)}

The equation (8) becomes

J^μ(Σ0)=⟨Jμ(Σ0)⟩I^+∫dΣ0′∫dω(2π)∫d3k(2π)3χμν′′eikr−iωΣ0′δA^ν(Σ0′)(22)\hat{J}_\mu(\Sigma_0)=\langle J_\mu(\Sigma_0)\rangle\hat{I}+\int d\Sigma'_0\int\frac{d\omega}{(2\pi)}\int\frac{d^3k}{(2\pi)^3}\chi''_{\mu\nu}e^{ikr-i\omega\Sigma'_0}\delta\hat{A}^\nu(\Sigma'_0) \tag*{(22)}

Where I^\hat{I} is the unit operator. Before applying the Laplace transformation, we perform the transformation indicated below

dΣ0=e^0×dt,δA^ν(Σ′)=∫dω′2πeiω′Σ0′δA^ν(ω′)(23)d\Sigma_0=\hat{e}_0\times dt,\qquad\delta\hat{A}_{\nu}(\Sigma')=\int\frac{d\omega'}{2\pi}e^{i\omega'\Sigma'_0}\delta\hat{A}_{\nu}(\omega') \tag*{(23)}

In this way, we find

J^μ(Σ0)=⟨Jμ(Σ0)⟩I^+∫dt∫dω2π∫d3k(2π)3χμν′′eikr−iωt∫dω′2πeiω′tδA^ν(ω′)=⟨Jμ(Σ0)⟩I^+e^0∫dω2πiχμν′′(k,ω)ω(ω−z)δA^ν(ω)(24)\begin{aligned} \hat{J}_{\mu}(\Sigma_0)&=\langle J_{\mu}(\Sigma_0)\rangle\hat{I}+\int dt\int\frac{d\omega}{2\pi}\int\frac{d^3k}{(2\pi)^3}\chi''_{\mu\nu}e^{i\mathbf{k}\mathbf{r}-i\omega t}\int\frac{d\omega'}{2\pi}e^{i\omega't}\delta\hat{A}^{\nu}(\omega')\\ &=\langle J_{\mu}(\Sigma_0)\rangle\hat{I}+\hat{e}_0\int\frac{d\omega}{2\pi i}\frac{\chi''_{\mu\nu}(k,\omega)}{\omega(\omega-z)}\delta\hat{A}^{\nu}(\omega) \tag*{(24)} \end{aligned}

Assuming χμν\chi_{\mu\nu} as a smooth function of ω\omega, we can apply the identity

lim⁡ϵ→01ω′−ω−iϵ=P1ω′−ω+iπδ(ω′−ω)(25)\lim_{\epsilon\to0}\frac{1}{\omega'-\omega-i\epsilon}=\mathcal{P}\frac{1}{\omega'-\omega}+i\pi\delta(\omega'-\omega) \tag*{(25)}

where P\mathcal{P} is the principal value since the poles zz are above the real axis

Re⁡[δJ^μ(k,ω+iϵ)δA^ν(ω)]=χμν′′ω(26)\operatorname{Re}\left[\frac{\delta\hat{J}^{\mu}(k,\omega+i\epsilon)}{\delta\hat{A}^{\nu}(\omega)}\right]=\frac{\chi''_{\mu\nu}}{\omega} \tag*{(26)}

Rewriting (A7) as an expansion for large values of zz, we get the analytical result

lim⁡z→∞δJ^μ(k,z)δA^ν(z)=izχμν(k)+iz2∫dωπχμν′′(k,ω)+iz2∫ω dωπχμν′′(k,ω)+iz3∫ω2z−ωdωπχμν′′(k,ω)+⋯(27)\begin{aligned} \lim_{z\to\infty}\frac{\delta\hat{J}^{\mu}(k,z)}{\delta\hat{A}_{\nu}(z)} ={}&\frac{i}{z}\chi_{\mu\nu}(k)+\frac{i}{z^2}\int\frac{d\omega}{\pi}\chi''_{\mu\nu}(k,\omega)+\frac{i}{z^2}\int\frac{\omega\,d\omega}{\pi}\chi''_{\mu\nu}(k,\omega)\\ &+\frac{i}{z^3}\int\frac{\omega^2}{z-\omega}\frac{d\omega}{\pi}\chi''_{\mu\nu}(k,\omega)+\cdots \tag*{(27)} \end{aligned}

Following this idea, it is straightforward to determine the dynamical susceptibility

χμν′′(k,ω)=Dξ2+χμνDk2ω2+D2[k2−(χp2/p02/⟨n⟩ξ2)]+χμνωω2−(D/ξ2)Jμ(x,0)Jμ(x,0)+χμνω2ξ2ω3+iω2(D2k2+ξ2)−ω(Dk2−(Jμ(x,0)Jμ(x,0))3/2)(28)\begin{aligned} \chi''_{\mu\nu}(k,\omega)={}&\frac{D}{\xi^2}+\frac{\chi_{\mu\nu}Dk^2}{\omega^2+D^2\left[k^2-\left(\chi\sqrt{p^2/p_0^2}/\langle n\rangle\xi^2\right)\right]}+\frac{\chi_{\mu\nu}\omega}{\omega^2-(D/\xi^2)\sqrt{J^\mu(x,0)J_\mu(x,0)}}\\ &+\frac{\chi_{\mu\nu}\omega^2\xi^2}{\omega^3+i\omega^2(D^2k^2+\xi^2)-\omega\left(Dk^2-\left(J^\mu(x,0)J_\mu(x,0)\right)^{3/2}\right)} \tag*{(28)} \end{aligned}

Which matches standard derivations, for example [6].

Correlation function with relaxation time

To avoid violation of stability and causality, we assume diffusive forces to not turn on instantaneously and include the relaxation time for a gradient to become a current. By enforcing charge conservation in (9), we can represent (3) in a convenient form

d3Q∣−∞∞=∫d3Σμd3Qd3Σμ=∫dΣ0⟨Jμ(Σ0)⟩+∫d3Σμ′∫dΣ0′Dμν(Σ,Σ′)(29)\left.d^3 Q\right|_{-\infty}^{\infty}=\int d^3\Sigma_\mu\frac{d^3Q}{d^3\Sigma_\mu}=\int d\Sigma_0\langle J_\mu(\Sigma_0)\rangle+\int d^3\Sigma'_\mu\int d\Sigma'_0D_{\mu\nu}(\Sigma,\Sigma') \tag*{(29)}

which yields

∫d3Σμ∫dΣ0⟨∣Jμ(0),Jν(Σ)∣⟩=∫dΣ0∫d3Σμ[∂2∂Σ02−1τQ(∂∂Σ0−D∇2)]⟨J^μ(Σ0)⟩=0(30)\int d^3\Sigma_\mu\int d\Sigma_0\langle|J^\mu(0),J^\nu(\Sigma)|\rangle=\int d\Sigma_0\int d^3\Sigma_\mu\left[\frac{\partial^2}{\partial\Sigma_0^2}-\frac{1}{\tau_Q}\left(\frac{\partial}{\partial\Sigma_0}-D\nabla^2\right)\right]\left\langle\hat{J}^\mu(\Sigma_0)\right\rangle=0 \tag*{(30)}

Applying the Laplace transform and integrating by parts, we obtain

0=δAν(Σ0)[(−iz+Dk2−τz2)J^μ(k,z)−J^μ(x,0)]+J^μ(Σ0)(−iz−τz2)δAν(z)(31)0=\delta A_\nu(\Sigma_0)\left[\left(-iz+ Dk^2-\tau z^2\right)\hat{J}^\mu(k,z)-\hat{J}^\mu(x,0)\right]+\hat{J}^\mu(\Sigma_0)\left(-iz-\tau z^2\right)\delta A_\nu(z) \tag*{(31)}

We write the current in function of the source

J^μ(k,z)δAν(z)=χμν(iz+z2τ)(Dk2−iz−τz2)−(τ/z)Jμ(x,0)Jμ(x,0)(32)\frac{\hat{J}^\mu(k,z)}{\delta A_\nu(z)}=\frac{\chi_{\mu\nu}(iz+z^2\tau)}{(Dk^2-iz-\tau z^2)-(\tau/z)\sqrt{J^\mu(x,0)J_\mu(x,0)}} \tag*{(32)}

According to (A7), the dynamic susceptibility of (A13) can be determined by performing an expansion for large values of zz

J^μ(k,z)δAν(z)=[iz+iDk2τz3+iτz2Jμ(x,0)Jμ(x,0)+iDττz3(Jμ(x,0)Jμ(x,0))3/2+O(1z4)](33)\frac{\hat{J}^\mu(k,z)}{\delta A_\nu(z)}=\left[\frac{i}{z}+\frac{iDk^2}{\tau z^3}+\frac{i}{\tau z^2}\sqrt{J^\mu(x,0)J_\mu(x,0)}+\frac{iD\tau}{\tau z^3}\left(J^\mu(x,0)J_\mu(x,0)\right)^{3/2}+O\left(\frac{1}{z^4}\right)\right] \tag*{(33)}

which matches similar theories developed for example in [10].

Correlation function with noise

We introduce random currents in equilibrium, so that the linearized constitutive relation for the current takes the form when τQ≠0\tau_Q\ne0

J⃗(ω,k)=−D∇⃗ν+∂kr⃗ k1−iτQω.(34)\vec{J}(\omega,k)=\frac{-D\vec{\nabla}\nu+\partial_k\vec{r}^{\,k}}{1-i\tau_Q\omega}. \tag*{(34)}

where r⃗ k\vec{r}^{\,k} is the Gaussian stochastic vector in space-time with the ensemble average ⟨r⃗ k⟩=0\langle\vec{r}^{\,k}\rangle=0.

Since we can separate the noise from the low energy effective action, the effective action rises by integrating out the fast modes

e−Seff[xslow]=∫Dxfast e−S[xslow+xfast](35)e^{-S_{\mathrm{eff}}[x_{\mathrm{slow}}]}=\int\mathcal{D}x_{\mathrm{fast}}\,e^{-S[x_{\mathrm{slow}}+x_{\mathrm{fast}}]} \tag*{(35)}
S[xslow+xfast]=S[xslow]+12∫dt dt′ xfast(t) O(t,t′) xfast(t′)+O(xfast3).(36)S[x_{\mathrm{slow}}+x_{\mathrm{fast}}]=S[x_{\mathrm{slow}}]+\frac{1}{2}\int dt\,dt'\,x_{\mathrm{fast}}(t)\,\mathcal{O}(t,t')\,x_{\mathrm{fast}}(t')+\mathcal{O}(x_{\mathrm{fast}}^3). \tag*{(36)}

The generating functional now becomes

Z=∫Dξ Dϕ Dϕˉ e−Seff[ϕ,ϕˉ]e−Sξ[rk](37)Z = \int D\xi\,D\phi\,D\bar{\phi}\,e^{-S_{\mathrm{eff}}[\phi,\bar{\phi}]}e^{-S_{\xi}[r^k]} \tag*{(37)}

where ϕ=(vi,δT,μ)\phi=(v_i,\delta T,\mu). Assuming the initial conditions are uncorrelated. Since the noise is a Gaussian functional integral, its probability distribution is

⟨…⟩=∫Dξ e−Sξ[rk]…,Sξ[r]=12∫dt ddx(riσˉij−1rj2T)(38)\langle\ldots\rangle=\int D\xi\,e^{-S_{\xi}[r^k]}\ldots,\qquad S_{\xi}[r]=\frac{1}{2}\int dt\,d^d x\left(\frac{r_i\bar{\sigma}_{ij}^{-1}r_j}{2T}\right) \tag*{(38)}

where σˉij\bar{\sigma}_{ij} controls the strength of stochastic fluctuations. The rrrr two-point function is

⟨ri(x,t)rk(x′,t′)⟩=Gδ(x−x′)δ(t−t′)(39)\langle r_i(x,t)r_k(x',t')\rangle=\mathcal{G}\delta(x-x')\delta(t-t') \tag*{(39)}

where G\mathcal{G} determines the strength of the noise. The representation of the correlation function in A18 after the integration of the noise A19 is

GJJ(Σ,Σ′)=⟨J(Σ)J(Σ′)⟩=∫Dϕ Dϕˉ⟨δ(e.o.m.)⟩∣J∣e−Seff[ϕ,ϕˉ]J(Σ)J(Σ′)(40)G_{JJ}(\Sigma,\Sigma')=\langle J(\Sigma)J(\Sigma')\rangle=\int D\phi\,D\bar{\phi}\langle\delta(\mathrm{e.o.m.})\rangle\lvert J\rvert e^{-S_{\mathrm{eff}}[\phi,\bar{\phi}]}J(\Sigma)J(\Sigma') \tag*{(40)}

and the stress tensor and the current are given by the classical linear constitutive relations. Splitting the velocity perturbations into δu=δuL+δuT\delta u=\delta u_L+\delta u_T, we have δuLi(ω,k)≡k⋅δu/∣k∣\delta u_L^i(\omega,\mathbf{k})\equiv\mathbf{k}\cdot\delta\mathbf{u}/\lvert k\rvert and δuT(ω,k)≡δu(ω,k)−δuL(ω,k)\delta u_T(\omega,\mathbf{k})\equiv\delta u(\omega,\mathbf{k})-\delta u_L(\omega,\mathbf{k}). Evaluating the effective action in A21 and read off the correlation functions

GJtJtR(ω,k)=−σk2iω(1−iωτQ)−Dnk2−n2k2(1−iωτQ)(ϵ+p)((ω2−vs2k2)(1−iωτQ)+iωD∥k2)+2σω2∣1−iωτQ∣2(41)G^R_{J_tJ_t}(\omega,k)=-\frac{\sigma k^2}{i\omega(1-i\omega\tau_Q)-D_n k^2}-\frac{n^2k^2(1-i\omega\tau_Q)}{(\epsilon+p)((\omega^2-v_s^2k^2)(1-i\omega\tau_Q)+i\omega D_{\parallel}k^2)}+\frac{2\sigma\omega^2}{\lvert1-i\omega\tau_Q\rvert^2} \tag*{(41)}
GJ⊥iJ⊥jR(ω,k)=(iωσ1−iωτQ−n2iω(1−iωτQ)iω(ϵ+p)(1−iωτQ)−D⊥k2)kij.(42)G^R_{J_\perp^iJ_\perp^j}(\omega,k)=\left(\frac{i\omega\sigma}{1-i\omega\tau_Q}-\frac{n^2i\omega(1-i\omega\tau_Q)}{i\omega(\epsilon+p)(1-i\omega\tau_Q)-D_{\perp}k^2}\right)k^{ij}. \tag*{(42)}

which can be matched to fluctuation-dissipation perturbative calculations [8].

Relation to the linear response approach with thermostatic background

The current JμJ^\mu can generically be parametrized as

Jμ=J0μ+Jμ(43)J^\mu=J_0^\mu+\mathcal{J}^\mu \tag*{(43)}

where J0μJ_0^\mu is the bare current and Jμ\mathcal{J}^\mu is a mixing between fluctuations and perturbation not necessarily small. The geometry conservation implies

∂νJν=0.(44)\partial_\nu\mathcal{J}^\nu=0. \tag*{(44)}

It is important to note that considering only the exponential associated with DμνD_{\mu\nu}, we can rewrite the partition function (17) as

Z=Z0⟨exp⁡−(J−J0)⟩0(45)Z = Z_0 \left\langle\exp^{-(J-J_0)} \right\rangle_0 \tag*{(45)}

where Z0=exp⁡[−Cμν,αβ(Σ,Σ′)(Tμν(Σ)−⟨Tμν(Σ′)⟩)(Tαβ(Σ′)−⟨Tαβ(Σ′)⟩)]Z_0 = \exp\left[-C_{\mu\nu,\alpha\beta}(\Sigma,\Sigma')\left(T^{\mu\nu}(\Sigma)-\left\langle T^{\mu\nu}(\Sigma')\right\rangle\right)\left(T^{\alpha\beta}(\Sigma')-\left\langle T^{\alpha\beta}(\Sigma')\right\rangle\right)\right]. Instead of (A18), we are not gonna separate the fast and slow modes as (A16). It follows that the free energy F=−ln⁡ZF=-\ln Z can be expanded as

F=∑k=0∞Fk;F0≡−ln⁡Z0 ,(46)F = \sum_{k=0}^{\infty} F_k \quad; \quad F_0 \equiv-\ln Z_0 \, , \tag*{(46)}
Fk≡−(−1)kk!⟨(J−J0)k⟩(for k≥1) ,(47)F_k \equiv-\frac{(-1)^k}{k!}\left\langle(J-J_0)^k \right\rangle\quad(\text{for } k \ge1) \, , \tag*{(47)}

Potential perturbation ⟷\longleftrightarrow Gauging a symmetry ⟷\longleftrightarrow Current emerges naturally. The nn-point Green’s function computed using the variational formulae

⟨Jμ(t1,x1)…Jμ(xn,tn)⏟n⟩=δnln⁡ZδAμ∣x1,t1…δAμ∣xn,tn ,(48)\left\langle\underbrace{J^\mu(t_1,x_1)\ldots J^\mu(x_n,t_n)}_{n} \right\rangle= \frac{\delta^n \ln Z}{\delta A_\mu\vert_{x_1,t_1}\ldots\delta A_\mu\vert_{x_n,t_n}} \, , \tag*{(48)}

We implement a geometric constraint on the kinematics of the Green function to restore the general covariance

{∇μD(Σ,Σ′)−δ(Σ′−Σ)−g(Aλ∇μ⟨∂λFμν⟩Σ+Aμ∇ν⟨∂μFνλ⟩Σ+Aν∇λ⟨∂νFλμ⟩Σ)}=0(49)\left\{\nabla_\mu D(\Sigma,\Sigma')-\frac{\delta(\Sigma'-\Sigma)}{\sqrt{-g}}\left(A^\lambda\nabla_\mu\left\langle\partial_\lambda F_{\mu\nu}\right\rangle_\Sigma+A^\mu\nabla_\nu\left\langle\partial_\mu F_{\nu\lambda}\right\rangle_\Sigma+A^\nu\nabla_\lambda\left\langle\partial_\nu F_{\lambda\mu}\right\rangle_\Sigma\right)\right\}=0 \tag*{(49)}

where as usual Fμν=∂μAν−∂νAμF_{\mu\nu}=\partial_\mu A_\nu-\partial_\nu A_\mu. We first examine the analytical characteristics of (4) and since we are able to use the ansatz (5), the propagator takes the form

Dμν=Pμν(Σμ)f(β,Jμ,Jν′,Σ) ,(50)D_{\mu\nu}=\mathcal{P}_{\mu\nu}\left(\Sigma^\mu\right)f\left(\beta,J_\mu,J'_\nu,\Sigma\right) \, , \tag*{(50)}

where the projector is Pμν∝Dμν\mathcal{P}_{\mu\nu}\propto D_{\mu\nu} and the function ff has a Gaussian form

f(…)∼∏Σ(x),Σ(x′)exp⁡[−12(Jμ(Σ(x))−⟨Jμ(Σ(x))⟩)Dμν(Σ(x),Σ(x′))(Jν(Σ(x′))−⟨Jν(Σ(x′))⟩)] ,(51)f(\ldots)\sim\prod_{\Sigma(x),\Sigma(x')} \exp\left[-\frac{1}{2}\left(J^\mu(\Sigma(x))-\left\langle J^\mu(\Sigma(x))\right\rangle\right)D_{\mu\nu}(\Sigma(x),\Sigma(x'))\left(J^\nu(\Sigma(x'))-\left\langle J^\nu(\Sigma(x'))\right\rangle\right)\right] \, , \tag*{(51)}

To compute the transport coefficient, we perform a perturbative expansion in (5) step-by-step

ln⁡Z≃ln⁡Z∣0−12!∫Σ,Σ′∂2ln⁡Z∂Aμ∂Aν∣0[(Jμ(Σ)−⟨Jμ(Σ′)⟩)(Jν(Σ)−⟨Jν(Σ′)⟩)+…\ln Z \simeq\ln Z|_0-\frac{1}{2!}\int_{\Sigma,\Sigma'}\frac{\partial^2\ln Z}{\partial A_\mu\partial A_\nu}|_0[(J^\mu(\Sigma)-\langle J^\mu(\Sigma')\rangle)(J^\nu(\Sigma)-\langle J^\nu(\Sigma')\rangle)+\ldots

Since the two currents are not symmetric and obey different directions between Σ\Sigma and Σ′\Sigma'.

Curvature arises from the global geometric structure of the manifold instead of torsion.

Dαβ(Jα−⟨Jα⟩)(Jβ−⟨Jβ⟩)→Dαβ′×diag⁡(λ1,2,3,4)γ×εαβγρ×diag⁡(ξ1,2,3,4)ρ(52)D_{\alpha\beta} \left(J^\alpha-\langle J^\alpha\rangle\right)\left(J^\beta-\langle J^\beta\rangle\right)\to D'_{\alpha\beta}\times\operatorname{diag}(\lambda_{1,2,3,4})^\gamma\times\varepsilon^{\alpha\beta\gamma\rho}\times\operatorname{diag}(\xi_{1,2,3,4})^\rho \tag*{(52)}

where εαβγρ\varepsilon^{\alpha\beta\gamma\rho} is the Levi-Civita tensor. Performing a lorentz transformation in the eigenvalues of the current

Jμ−⟨Jμ⟩=Λαμdiag⁡(λ1,2,3,4)α,Λνμ=exp⁡[∫i2dωαβ(Mαβ)μν],(53)J^\mu-\langle J^\mu\rangle=\Lambda^\mu_\alpha\operatorname{diag}(\lambda_{1,2,3,4})^\alpha,\qquad\Lambda^\mu_\nu=\exp\left[\int\frac{i}{2}d\omega_{\alpha\beta}(M^{\alpha\beta})^\mu{}_\nu\right], \tag*{(53)}

where Λ∈O(1,3)\Lambda\in O(1,3). Let us change the solutions from lab to the comoving frame

λ0≃ξ0≃no,λ1−3≃ξ1−3≃0,Λμν≃(1,0⃗)(54)\lambda_0\simeq\xi_0\simeq n_o,\qquad\lambda_{1-3}\simeq\xi_{1-3}\simeq0,\qquad\Lambda^\mu{}_\nu\simeq(1,\vec{0}) \tag*{(54)}

where non_o is the equilibrium charge density

Z≃∫∏μνdωμνexp⁡[−ωμνωμνD′μμ],Dμν′=(Λμα)(Λνβ)Dαβ,(55)\mathcal{Z}\simeq\int\prod_{\mu\nu}d\omega_{\mu\nu}\exp\left[-\omega_{\mu\nu}\omega^{\mu\nu}D'{}^\mu{}_\mu\right],\qquad D'_{\mu\nu}=(\Lambda^\alpha_\mu)(\Lambda^\beta_\nu)D_{\alpha\beta}, \tag*{(55)}

Integrating in both sides to obtain the Green function

∫dΣ′∇μDμν(Σ,Σ′)=∫dΣ′δ(Σ′−Σ)−g(Aλ∇μ⟨∂λFμν⟩Σ+Aμ∇ν⟨∂μFνλ⟩Σ+Aν∇λ⟨∂νFλμ⟩Σ)(56)\int d\Sigma'\nabla_\mu D^{\mu\nu}(\Sigma,\Sigma')=\int d\Sigma'\frac{\delta(\Sigma'-\Sigma)}{\sqrt{-g}}\left(A^\lambda\nabla_\mu\langle\partial_\lambda F_{\mu\nu}\rangle_\Sigma+A^\mu\nabla_\nu\langle\partial_\mu F_{\nu\lambda}\rangle_\Sigma+A^\nu\nabla_\lambda\langle\partial_\nu F_{\lambda\mu}\rangle_\Sigma\right) \tag*{(56)}

where g=det⁡(gμν)g=\det(g_{\mu\nu}). The divergence of an antisymmetric tensor and and the covariant divergence of an antisymmetric tensor can be written as

∇μFμν=∂μ(−gFμν)−g=μ0Jν,∇μJμ=∂μ(−gJμ)−g=0,(57)\nabla_\mu F^{\mu\nu}=\frac{\partial_\mu\left(\sqrt{-g}F^{\mu\nu}\right)}{\sqrt{-g}}=\mu_0J^\nu,\qquad\nabla_\mu J^\mu=\frac{\partial_\mu\left(\sqrt{-g}J^\mu\right)}{\sqrt{-g}}=0, \tag*{(57)}

Using Γμλμ=∂λln⁡−g\Gamma^\mu_{\mu\lambda}=\partial_\lambda\ln\sqrt{-g}. Applying in (A30), we finally find

Qν=∫dΣ[∇μDμν(Σ,Σ′)+μ0−g∑i1λiξiAμ∇αJβεμαβν]Q^\nu=\int d\Sigma\left[\nabla_\mu D^{\mu\nu}(\Sigma,\Sigma')+\mu_0\sqrt{-g}\sum_i\frac{1}{\lambda_i\xi_i}A_\mu\nabla_\alpha J_\beta\varepsilon^{\mu\alpha\beta\nu}\right]

where QνQ^\nu is a constant of motion. The Helmholtz free energy (A27) can be expanded in a Taylor series. The thermodynamic form is

F(N)≈F(N0)+12(∂2F∂N2)T,V(δN)2+O(δN)3(58)F(N)\approx F(N_0)+\frac{1}{2}\left(\frac{\partial^2F}{\partial N^2}\right)_{T,V}(\delta N)^2+O(\delta N)^3 \tag*{(58)}

and defining δN=N−N0\delta N=N-N_0. The density fluctuation is

⟨N2⟩−⟨N⟩2≡kTV∂2P∂μ2=kTN‾kTv⇒⟨ΔJαΔJβ⟩∼∂2ln⁡Z∂Aα∂Aβ∣0dΣαdΣβ,(59)\langle\mathcal{N}^2\rangle-\langle\mathcal{N}\rangle^2\equiv kTV\frac{\partial^2P}{\partial\mu^2}=k_T\overline{\mathcal{N}}\frac{kT}{v}\Rightarrow\langle\Delta J_\alpha\Delta J_\beta\rangle\sim\left.\frac{\partial^2\ln\mathcal{Z}}{\partial A_\alpha\partial A_\beta}\right|_0d\Sigma_\alpha d\Sigma_\beta, \tag*{(59)}

The covariantized version is

∂Dμν′∂Dμν+∂Σμ∂Σν∂Aμ∂Aν=1(60)\frac{\partial D'_{\mu\nu}}{\partial D_{\mu\nu}}+\frac{\partial\Sigma_{\mu}\partial\Sigma_{\nu}}{\partial A_{\mu}\partial A_{\nu}}=1 \tag*{(60)}

The transport coefficients in terms of our previous results are given via the usual Kubo formulae

μ∼lim⁡k→0k−1Im⁡D~xt′,τQμ∼lim⁡k→0k−2Im⁡∂2∂k2D~xt′.(61)\mu\sim\lim_{k\to0} k^{-1}\operatorname{Im}\widetilde{D}'_{xt}, \qquad\tau_Q\mu\sim\lim_{k\to0} k^{-2}\operatorname{Im}\frac{\partial^2}{\partial k^2}\widetilde{D}'_{xt}. \tag*{(61)}

This shows how the usual calculations on a thermostatic background match with the non-perturbaive approach developed in this work. Of course, in a hydrostatic background thermal field theory techniques can be used to calculate ”non-gaussian” corrections in the form of higher order Feynman diagrams, as was done in [8] [11, 14] and others. In such calculations however non-Gaussianities appear as corrections to Gaussian parameters, since they enter into the propagator,

References

References

  1. [1]G. M. Sampaio, G. Rabelo-Soares and G. Torrieri, Phys. Rev. D 112 (2025) no.5, 056002 doi:10.1103/67x1-knzr [arXiv:2504.17152 [hep-th]].DOI
  2. [2]M. Nahrgang, M. Bluhm, T. Schäfer and S. A. Bass, Nucl. Phys. A 967 (2017), 824-827 doi:10.1016/j.nuclphysa.2017.04.021 [arXiv:1804.02976 [nucl-th]].DOI
  3. [3]G. S. Denicol, C. Gale, S. Jeon, A. Monnai, B. Schenke and C. Shen, Phys. Rev. C 98 (2018) no.3, 034916 doi:10.1103/PhysRevC.98.034916 [arXiv:1804.10557 [nucl-th]].DOI
  4. [4]A. W. Steiner, M. Prakash and J. M. Lattimer, Phys. Lett. B 509 (2001), 10-18 doi:10.1016/S0370-2693(01)00434-8 [arXiv:astro-ph/0101566 [astro-ph]].DOI
  5. [5]J. H. Applegate, C. J. Hogan and R. J. Scherrer, Phys. Rev. D 35 (1987), 1151-1160 doi:10.1103/PhysRevD.35.1151DOI
  6. [6]D. Forster, ”hydrodynamic fluctuations, broken symmetry and correlation functions”, Addison-Wesley (1990)
  7. [7]S. Jeon and U. Heinz, Int. J. Mod. Phys. E 24 (2015) no.10, 1530010 doi:10.1142/S0218301315300106 [arXiv:1503.03931 [hep-ph]].DOI
  8. [8]P. Kovtun, J. Phys. A 45 (2012), 473001 doi:10.1088/1751-8113/45/47/473001 [arXiv:1205.5040 [hep-th]].DOI
  9. [27]M. A. Stephanov, Prog. Theor. Phys. Suppl. 153 (2004), 139-156 doi:10.1143/PTPS.153.139 [arXiv:hep-ph/0402115 [hep-ph]].DOI
  10. [28]M. A. Stephanov, Phys. Rev. Lett. 102 (2009), 032301 doi:10.1103/PhysRevLett.102.032301 [arXiv:0809.3450 [hep-ph]].DOI
  11. [29]G. Torrieri, Eur. Phys. J. A 56 (2020) no.4, 121 doi:10.1140/epja/s10050-020-00121-z [arXiv:1810.12468 [hep-th]].

Paper details

vlnc:W_9e6crhu4
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