Causal Transport Equations

Updated 9 June 2026

Causal transport equations are mathematical formulations that ensure signal propagation remains within the light cone, preserving relativistic causality.
They impose universal analytic bounds on transport coefficients via dispersion relations and complex analyticity constraints in hydrodynamics and quantum field theory.
Applications span from relativistic hydrodynamic models to stochastic optimal transport, ensuring well-posed evolution of dissipative systems.

Causal transport equations govern the dynamics of physical systems in a manner compatible with relativistic causality—namely, that signals and information cannot propagate outside the light cone. Such equations are essential both in relativistic hydrodynamics and in stochastic processes where temporal order and adaptation are required. Their mathematical formulation imposes constraints not merely for physical realism, but for consistency under boosts and for well-posedness of the associated partial differential or integral equations.

1. Foundations: Causality, Dispersion Relations, and Analyticity

In relativistic quantum field theory (QFT), causality is encoded at the level of commutators of conserved currents. For a conserved current $J^\mu(x)$ in Minkowski space, the retarded two-point function vanishes outside the forward light cone: $G_r^{\mu\nu}(x-y) = -i\,\theta(x^0-y^0)\langle [J^\mu(x), J^\nu(y)] \rangle=0, \quad x \notin V_+$ Transforming to Fourier space, the singularities of $\tilde G_r^{\mu\nu}(\omega,k)$ as functions of complex $\omega$ at fixed $k$ define the dispersive behavior of the system. The locations $\omega=\omega(k)$ solve the dispersion relation for collective excitations in various channels, such as

$\text{Shear (diffusion):}\quad \omega(k) \sim -i D k^2 +\ldots$

$\text{Sound:}\quad \omega(k) = v k - i\frac{\Gamma_s}{2}k^2 + \ldots$

Causality—specifically, microcausality—imposes that $\tilde G_r(\omega,k)$ is analytic in the “forward tube” $\operatorname{Im}\omega > |\operatorname{Im}k|$ , so any dispersion branch must satisfy

$G_r^{\mu\nu}(x-y) = -i\,\theta(x^0-y^0)\langle [J^\mu(x), J^\nu(y)] \rangle=0, \quad x \notin V_+$ 0

If the dispersion function $G_r^{\mu\nu}(x-y) = -i\,\theta(x^0-y^0)\langle [J^\mu(x), J^\nu(y)] \rangle=0, \quad x \notin V_+$ 1 exhibits higher-than-linear terms (e.g., quadratic dissipation), analyticity arguments show it cannot be entire—there must exist a finite radius $G_r^{\mu\nu}(x-y) = -i\,\theta(x^0-y^0)\langle [J^\mu(x), J^\nu(y)] \rangle=0, \quad x \notin V_+$ 2 of convergence for the expansion around $G_r^{\mu\nu}(x-y) = -i\,\theta(x^0-y^0)\langle [J^\mu(x), J^\nu(y)] \rangle=0, \quad x \notin V_+$ 3 (Heller et al., 2022).

2. Universal Causality-Induced Bounds on Transport Coefficients

Using only causality and analyticity in the complex $G_r^{\mu\nu}(x-y) = -i\,\theta(x^0-y^0)\langle [J^\mu(x), J^\nu(y)] \rangle=0, \quad x \notin V_+$ 4-plane, one derives rigorous bounds on the Taylor coefficients $G_r^{\mu\nu}(x-y) = -i\,\theta(x^0-y^0)\langle [J^\mu(x), J^\nu(y)] \rangle=0, \quad x \notin V_+$ 5 of the dispersion relation $G_r^{\mu\nu}(x-y) = -i\,\theta(x^0-y^0)\langle [J^\mu(x), J^\nu(y)] \rangle=0, \quad x \notin V_+$ 6, where $G_r^{\mu\nu}(x-y) = -i\,\theta(x^0-y^0)\langle [J^\mu(x), J^\nu(y)] \rangle=0, \quad x \notin V_+$ 7: $G_r^{\mu\nu}(x-y) = -i\,\theta(x^0-y^0)\langle [J^\mu(x), J^\nu(y)] \rangle=0, \quad x \notin V_+$ 8 For the shear-diffusion (with $G_r^{\mu\nu}(x-y) = -i\,\theta(x^0-y^0)\langle [J^\mu(x), J^\nu(y)] \rangle=0, \quad x \notin V_+$ 9), the diffusion constant is bounded: $\tilde G_r^{\mu\nu}(\omega,k)$ 0 Similarly, for sound attenuation ( $\tilde G_r^{\mu\nu}(\omega,k)$ 1), the bounds read

$\tilde G_r^{\mu\nu}(\omega,k)$ 2

and a mixed bound combining $\tilde G_r^{\mu\nu}(\omega,k)$ 3 and $\tilde G_r^{\mu\nu}(\omega,k)$ 4: $\tilde G_r^{\mu\nu}(\omega,k)$ 5 The numerical factor $\tilde G_r^{\mu\nu}(\omega,k)$ 6 is fixed by explicit calculation of angular integrals and the Carathéodory integral representation, independent of any dynamical input or coupling constants. These bounds are universal: they apply in any causal relativistic QFT, regardless of coupling strength, and require only microcausality and temperedness (Heller et al., 2022).

3. Structural Examples, Tightness, and Physical Implications

Several key examples illustrate these bounds and clarify their physical content:

The acausal heat equation ( $\tilde G_r^{\mu\nu}(\omega,k)$ 7 with $\tilde G_r^{\mu\nu}(\omega,k)$ 8) is ruled out by causality, as its entire dispersion violates the finite-radius criterion. Thus, classical (parabolic) diffusion is inconsistent with relativistic causality.
The telegrapher’s equation (arising in Müller–Israel–Stewart-type theories) with $\tilde G_r^{\mu\nu}(\omega,k)$ 9 exhibits a finite radius $\omega$ 0 and saturates $\omega$ 1.
In strongly coupled holographic models (e.g., RN-AdS $\omega$ 2), the product $\omega$ 3 is numerically below the absolute bound, indicating that the $\omega$ 4 bound is sharp but not saturated by known physical theories.

These bounds exclude the possibility of arbitrarily large transport coefficients in any causal theory, establishing a link between macroscopic dissipative scales and the underlying microscopic relaxation time $\omega$ 5 (Heller et al., 2022).

4. Causal Transport in Stochastic and Optimal Transport Frameworks

Beyond QFT and hydrodynamics, causal transport theorems underpin optimal transport problems on stochastic processes. In discrete time, a “causal coupling” between processes $\omega$ 6 and $\omega$ 7 requires for each time $\omega$ 8 that the law of $\omega$ 9 be adapted to $k$ 0. Such couplings generalize classical adapted (non-anticipative) couplings and are characterized by filtration constraints or recursive kernel factorization (Veraguas et al., 2016). The associated causal transport problem minimizes over causal plans: $k$ 1 Dynamic programming principles allow reduction to a sequence of stagewise problems when costs and marginals have product or Markov structure, and the unique optimizer can, in symmetric cases, be expressed via the Knothe–Rosenblatt rearrangement.

The framework admits a robust theory of “nested distance”—a generalization of Wasserstein distance respecting two-sided causality—along with transport-information inequalities (e.g., control by the relative entropy), key for quantifying differences between dynamical stochastic systems (Veraguas et al., 2016).

5. Implementation in Relativistic Hydrodynamics and Beyond

Causal transport equations are realized explicitly in the modern theories of relativistic hydrodynamics and kinetic theory:

In the Israel–Stewart (second-order) formalism, transport equations for dissipative fields (e.g., shear stress $k$ 2, bulk pressure $k$ 3) take hyperbolic form, enforcing finite propagation speeds:

$k$ 4

The presence of relaxation times $k$ 5 restores causality and ensures subluminal signal and group velocities in all frames (Fujii et al., 2023).

Linearized analysis of the relativistic Navier–Stokes equations about the global rest frame demonstrates that, given standard thermodynamic stability, even first-order theories can be causal provided they are applied within their regime of validity (small departures from equilibrium, rest frame) (Sandoval-Villalbazo et al., 2010). However, more general backgrounds or gradients necessitate full causal (second-order) treatment.
In radiative transfer, strict causality and frame-independent stability require that transport coefficients lie within a convex “hydrohedron” determined by the analytic structure of the exact dispersion relation. The calculation for radiative heat flow including photon transit times leads to a hyperbolic Cattaneo-type equation and a full hierarchy of higher-order, causal transport coefficients (Gavassino, 12 Feb 2025).

Within the kinetic theory context, the Chapman–Enskog gradient expansion is insufficient for causality unless time and space derivatives are treated symmetrically, as in the relaxation model approach (Chen et al., 2011). Absence of comoving time derivatives in first-order theories can generically lead to acausality outside the rest frame unless the structure of the equations is modified (Mitra, 2021).

6. Causal Optimal Transport on Path Space and Continuous Time

Recent developments generalize causal transport concepts to path space $k$ 6 and continuous time (Cont et al., 2024, Backhoff et al., 19 May 2026). Causal and bicausal couplings are characterized by progressive measurability and semimartingale structure, leading to explicit martingale representations and complete classification of optimal maps (including Monge-type maps induced by stochastic integrals with rotation-valued integrands). The causal optimal transport value for a finite-state Markov source and continuous diffusion target can be characterized as the solution to a nonlinear parabolic master equation, admitting dual stochastic-control representations with provably convergent numerical schemes (Backhoff et al., 19 May 2026).

7. Methodological and Physical Significance

Causal transport equation theory provides a rigorous, physically consistent framework for connecting microscopic dynamics to macroscopic relaxation in relativistic and stochastic systems, placing strict constraints on allowed forms of dissipation and the admissible sizes of transport coefficients. These constraints are universal, arising from analyticity and symmetry, and cannot be evaded even in strongly coupled, non-perturbative, or highly engineered systems, barring explicit violation of microcausality.

The implications reach from hydrodynamics in heavy-ion collisions (where causality of linearized fluctuations is relevant for interpreting observables) to stochastic programming and quantum walks (where causal optimal transport and nested distances provide new ways to compare the evolution of stochastic processes).

A key controversy resolved by this framework is that analyticity on the real axis alone is insufficient for universal transport bounds: detailed extension into the complex $k$ 7-plane and existence of a hydrodynamic branch in a finite disk are required. Simple kinetic models can, via suitable initialization, mimic arbitrarily pathological (including acausal) macroscopic evolution, highlighting that assumptions about analytic continuation and structure of singularities away from the real axis are essential for rigorous constraints (Gavassino, 8 Apr 2026).

References:

(Heller et al., 2022) Rigorous bounds on transport from causality
(Veraguas et al., 2016) Causal transport in discrete time and applications
(Sandoval-Villalbazo et al., 2010) A causality analysis of the linearized relativistic Navier-Stokes equations
(Gavassino, 12 Feb 2025) Causality constraints on radiative transfer
(Fujii et al., 2023) Causal hydrodynamic fluctuations in a one-dimensional expanding system
(Cont et al., 2024) Causal transport on path space
(Backhoff et al., 19 May 2026) Analytical Approach to Continuous-Time Causal Optimal Transport
(Gavassino, 8 Apr 2026) How acausal equations emerge from causal dynamics
(Chen et al., 2011) Causal Relativistic Fluid Dynamics
(Mitra, 2021) Causality and stability analysis of first-order field redefinition in relativistic hydrodynamics from kinetic theory