Relativistic Boltzmann Theory

A relativistic distribution function resolves matter in position and momentum rather than compressing it immediately into density, temperature, and one velocity. It lives on the future mass shell, is transported by the geodesic flow between interactions, and changes through a collision operator. Integrating that one kinetic equation against collision invariants produces particle-current and stress–energy conservation; higher moments reveal exactly what a fluid closure discards.

Required background. Hilbert Stress–Energy from Metric Variation supplies the stress tensor and conservation laws; Events, Intervals, and Minkowski Geometry supplies the relativistic mass shell; and Differential Equations supplies phase-space transport.

Helpful background. The Einstein Equation provides gravitational backreaction; Relativistic Matter and Compact Objects compares kinetic and continuum descriptions; and Computational Relativity gives numerical contexts for transport equations.

Distribution on the future mass shell

Assume uncharged particles with a dilute one-particle description, binary local collisions, and a specified cross section. At each spacetime event $x$, particles of rest mass $m$ occupy the future mass shell

$$P_x^+=\{p^a:g_{ab}p^a p^b=-m^2, p^a\text{ future directed}\}.$$

In a local orthonormal frame the invariant measure may be written

$$dP=\frac{d^3p}{p^{\hat0}}, \qquad p^{\hat0}=\sqrt{m^2+\delta_{ij}p^{\hat i}p^{\hat j}}.$$

Some conventions include a spin degeneracy and $(2\pi)^{-3}$ in $dP$; here those constants are absorbed into the scalar distribution $f(x,p)$. Physical moments are unchanged provided the same normalization is used consistently. The curved-spacetime mass shell, Liouville operator, invariant measure, and moment definitions are set out in Andréasson 2011, § 2, pp. 11–13, Eqs. (25)–(32).

The particle current and stress–energy tensor are

$$N^a=\int_{P_x^+}p^a f dP, \qquad T^{ab}=\int_{P_x^+}p^a p^b f dP.$$

In an orthonormal frame $N^{\hat0}=\int f d^3p$ is number density measured by that observer, and $T^{\hat0\hat0}=\int p^{\hat0}f d^3p$ is energy density. The factor $1/p^{\hat0}$ in $dP$ is therefore not optional: it makes the measure invariant while the moment factors supply the measured densities.

Geodesic transport and collisions

Using the three independent spatial momentum coordinates on the shell, the relativistic Boltzmann equation is

$$L[f]\equiv p^a\frac{\partial f}{\partial x^a} -\Gamma^i{}_{ab}p^a p^b\frac{\partial f}{\partial p^i} =C[f].$$

The left side is the Liouville derivative along the lifted geodesic flow. If $C[f]=0$, $f$ is constant along each particle trajectory. Curvature changes coordinate momentum through the connection term but does not create or destroy phase-space density.

For binary elastic collisions $p+p_1\leftrightarrow p'+p_1'$, the collision integral compares gain and loss terms weighted by an invariant transition rate. Its detailed kernel depends on the microscopic cross section. The structural statements needed macroscopically are that collision invariants obey

$$\int C[f] dP=0, \qquad \int p^b C[f] dP=0$$

when reactions conserve particle number and four-momentum. Integrating the Boltzmann equation then gives

$$\nabla_aN^a=0, \qquad \nabla_aT^{ab}=0.$$

If reactions change species number, the first identity is replaced by balanced source terms across species, while the sum of their stress tensors remains conserved. Andréasson writes the special-relativistic binary collision operator and its four-momentum conservation law in Andréasson 2011, § 1.1, pp. 4–7, Eqs. (1)–(10). Rocha et al. derive the invariant collision measure, moment identities, detailed balance, and entropy production in Rocha et al. 2024, § V.A, pp. 24–25, Eqs. (76)–(83). They use signature $(+,-,-,-)$; translating the mass-shell and velocity normalizations gives the formulas used here.

Equilibrium and the perfect-fluid limit

For a dilute classical gas, local equilibrium has the Maxwell–Jüttner form

$$f_{\mathrm{eq}}=\exp(\alpha+\beta_a p^a), \qquad \beta_a=\frac{u_a}{T}.$$

Here $\alpha$ fixes particle number; apart from the constant normalization absorbed into $f$, it is $\mu/T$. In the rest frame, $\beta_ap^a=-p^{\hat0}/T$, so high-energy particles are exponentially suppressed. For a number-conserving elastic collision, the two incoming and two outgoing factors contain the same $2\alpha$, and four-momentum conservation makes their remaining exponents equal. Gain and loss cancel: $C[f_{\mathrm{eq}}]=0$. Quantum statistics modifies the occupation factors and replaces the exponential by Fermi–Dirac or Bose–Einstein forms.

Local collision equilibrium does not automatically solve the complete transport equation. Along a geodesic,

$$\frac{L[f_{\rm eq}]}{f_{\rm eq}} =p^a\nabla_a\alpha+p^ap^b\nabla_{(a}\beta_{b)}.$$

Thus constant $\alpha$ and a Killing inverse-temperature vector $\beta^a$ are sufficient for an exact global equilibrium. Arbitrary spatially varying $T$ and $u^a$ generally make the right-hand side nonzero even though $C[f_{\rm eq}]=0$. In hydrodynamics, small departures from local equilibrium allow collisions to balance that slow transport. This is why a perfect-fluid approximation is a scale-separation statement, not the assertion that the distribution is an exact collision equilibrium everywhere.

Isotropy in the $u^a$ rest frame makes the first two moments

$$N^a=n u^a, \qquad T^{ab}=(\varepsilon+p)u^a u^b+p g^{ab}.$$

Thus a perfect fluid is a special moment closure, not a different theory of matter. Departures from equilibrium generate heat flux, shear stress, bulk stress, and an infinite hierarchy of higher angular and energy moments. A closure approximates the unresolved moments in terms of a finite set; its validity depends on the Knudsen number, the ratio of mean free path to macroscopic scale.

The classical entropy current

$$S^a=-\int p^a(f\ln f-f) dP$$

satisfies $\nabla_aS^a\geq0$ for a collision kernel with microscopic reversibility. Equality holds at local equilibrium. This is the classical-statistics specialization of the entropy current and H-theorem in Rocha et al. 2024, § V.A, Eqs. (82)–(83). It explains why smooth perfect-fluid flow advects entropy and why dissipative corrections produce it.

Scope and limiting checks

For $T\ll m$ in the rest frame, $p^{\hat0}=m+\lvert\boldsymbol p\rvert^2/(2m)+\cdots$ gives

$$f_{\rm eq}\simeq e^{\alpha-m/T} \exp\!\left[-\frac{\lvert\boldsymbol p\rvert^2}{2mT}\right].$$

Absorbing the constant rest-energy Boltzmann factor into the normalization leaves the Maxwell distribution; there is no oscillating phase to remove. Expanding the transport equation consistently gives the nonrelativistic Boltzmann equation. One may use $p^{\hat0}\simeq m$ in a leading-order number moment, but the kinetic correction must be retained to compute internal energy. The massless limit instead uses $p^{\hat0}=\lvert\boldsymbol p\rvert$; its momentum dependence cannot be dropped from the measure.

Kinetic theory is essential when the mean free path is long, the distribution has multiple beams, or reaction rates compete with dynamics. A one-particle distribution still assumes that correlations entering the collision term can be closed, commonly through molecular chaos. Strong correlations, coherent quantum transport, and pair creation beyond a specified reaction network require a more general description.

Exercises

Suppose a collision operator conserves four-momentum but allows pair creation, so $\int p^bC[f]dP=0$ while $\int C[f]dP\neq0$. Which macroscopic conservation law survives, and which does not?

Solution

Multiplying the Boltzmann equation by $p^b$ and integrating gives $\nabla_aT^{ab}=0$, because the four-momentum moment of the collision term vanishes. Integrating with weight one gives $\nabla_aN^a=\int C[f]dP$, which is nonzero. Energy–momentum conservation survives, but the chosen particle-number current has a reaction source.

References

• Andréasson, Håkan. “The Einstein–Vlasov System/Kinetic Theory.” Living Reviews in Relativity 14, 4 (2011). DOI.
• Rocha, Gabriel S., David Wagner, Gabriel S. Denicol, Jorge Noronha, and Dirk H. Rischke. “Theories of Relativistic Dissipative Fluid Dynamics.” Entropy 26, 189 (2024). DOI. Open preprint.