Neutrino Transport in Dense Matter

Neutrinos carry energy, momentum, electron lepton number, and spectral information through hot dense matter. Their mean free paths range from microscopic in opaque cores to macroscopic in transparent ejecta, so neither instantaneous equilibrium nor free streaming is valid everywhere. A defensible transport model must distinguish flavors and energies, use rates consistent with the equation of state, recover detailed balance and diffusion, and conserve the combined matter–neutrino budgets.

Required background. Hilbert Stress–Energy from Metric Variation supplies stress exchange and conserved totals; Differential Equations supplies kinetic transport, stiff sources, and diffusion limits.

Helpful background. The Einstein Equation supplies spacetime coupling; Relativistic Matter and Compact Objects gives the dense-matter context; and Computational Relativity supplies relativistic evolution methods.

Flavor, energy, and weak interactions

Treat the neutrinos as ultrarelativistic and incoherent, so their trajectories are null to the retained accuracy and each species has a scalar occupation. This classical transport approximation does not evolve coherent flavor oscillations. In matter containing neutrons, protons, electrons, and positrons, electron-flavor charged-current reactions include

$$\nu_e+n\leftrightarrow p+e^-, \qquad \bar\nu_e+p\leftrightarrow n+e^+.$$

They exchange electron lepton number and drive the net electron fraction $Y_e=(n_{e^-}-n_{e^+})/n$, where $n$ is baryon number density. All flavors participate in neutral-current scattering and pair processes, but heavy-lepton flavors lack the same low-energy charged-current channels when muons and taus are absent. Opacities depend strongly on neutrino energy, often approximately as its square in simple weak-interaction limits, and also on temperature, density, degeneracy, composition, nuclear correlations, and final-state blocking. Muon-rich matter and coherent flavor conversion require additional species or quantum kinetic variables.

For each species, a distribution $f_s(x,p)$ obeys a general-relativistic Boltzmann equation. Its stress tensor and, where appropriate, number current are momentum moments. Matter exchange is arranged so

$$\nabla_aT_{\rm matter}^{ab}=\sum_s G_s^b, \qquad \nabla_aR_s^{ab}=-G_s^b.$$

The sum is exactly conserved. Let $N_s^a$ be a species number current and $\mathcal R_s=\nabla_aN_s^a$ its net production rate. In the absence of flavor conversion or additional electron-lepton carriers,

$$\begin{aligned} \nabla_a(nY_eu^a)&=-\mathcal R_{\nu_e}+\mathcal R_{\bar\nu_e},\\ \nabla_a\left(nY_eu^a+N_{\nu_e}^a-N_{\bar\nu_e}^a\right)&=0. \end{aligned}$$

For example, absorbing one $\nu_e$ removes one neutrino and creates one electron, so the two changes cancel in electron lepton number. Energy conservation alone cannot enforce this number balance. The same collision kernels must supply both; see Foucart 2023, §§ 3.1–3.2 and 4.2.3.

Detailed balance and equilibrium

In a homogeneous material cell at temperature $T$, the equilibrium occupation is Fermi–Dirac,

$$f_{\rm eq}(\epsilon)= \frac{1}{\exp[(\epsilon-\mu_\nu)/T]+1}.$$

For an emission–absorption contribution to the local material-frame time derivative written

$$\left(\frac{\partial f}{\partial t}\right)_{\rm coll} =j(1-f)-\kappa f,$$

detailed balance requires

$$j(1-f_{\rm eq})=\kappa f_{\rm eq}.$$

Here $j$ and $\kappa$ are emission and true absorption rates, with the displayed neutrino blocking factor kept separate. In the covariant Liouville equation $L[f]=C[f]$, the collision term has the extra factor of locally measured particle energy: $C=\epsilon[j(1-f)-\kappa f]$. The distinction between these normalizations and the effective-opacity rearrangement is explicit in Mezzacappa et al. 2020, § 5.2.1, Eqs. (168)–(169). Detailed balance implies

$$\frac j\kappa=e^{-(\epsilon-\mu_\nu)/T}, \qquad \mu_{\nu_e}=\mu_e+\mu_p-\mu_n$$

for charged-current chemical equilibrium. The chemical potentials must include rest energies consistently. In tables that absorb stimulated-absorption corrections into an effective opacity, the symbol called “absorption” may instead mean $j+\kappa$; those conventions must be translated before inserting a rate in this equation. Foucart explains this rearrangement in Foucart 2023, § 3.2.1, Eq. (22).

The blocking factor $1-f$ matters in degenerate matter. Mixing opacity tables with inconsistent equation-of-state chemical potentials can produce a false equilibrium and secular source terms even if each table is individually smooth.

With matter and rates fixed and $j+\kappa\gt0$, the one-zone equation has the exact solution

$$f(t)=f_{\rm eq}+[f(0)-f_{\rm eq}]e^{-(j+\kappa)t}, \qquad f_{\rm eq}=\frac{j}{j+\kappa}.$$

Nonnegative rates keep $0\leq f(t)\leq1$ whenever the initial occupation lies in that interval. If both rates vanish, $f$ simply retains its initial value and no collision equilibrium is selected. This tests the equilibrium, timescale, and bound preservation independently. A coupled test must instead evolve matter temperature and $Y_e$ while conserving the combined energy and electron lepton number; the fixed-reservoir exponential is then no longer the reference solution.

Diffusion and free streaming

Define an energy-dependent optical depth along a path by

$$\tau_\epsilon=\int \kappa_{\rm tr}(\epsilon,s) ds,$$

where $ds$ is path length in the material frame and $\epsilon$ is the local neutrino energy along the path, including redshift or Doppler changes. Transport opacity includes the angular inefficiency of forward scattering. In a stationary flat-space material background, take the local mean free path to be short compared with the radiation-gradient scale, remain away from a boundary layer, and wait longer than the angular-relaxation time. The distribution is then nearly isotropic and the flux approaches

$$F^i_\epsilon\simeq-\frac{1}{3\kappa_{\rm tr}} \nabla^iE_\epsilon.$$

Velocity gradients, gravitational redshift, and energy-group coupling add terms to the general relativistic diffusion problem. The displayed formula isolates the local stationary limit.

For $\tau\ll1$, neutrinos free-stream along null geodesics. This does not imply $|F|=E$: that equality holds only for a single propagation direction. Opposed collisionless beams have $F=0$ even in vacuum, as the moment-closure example shows. Far from one localized source, the angular cone narrows and $|F|/E$ can approach one. Decoupling is not a sharp physical surface and differs by energy and flavor; a single grey neutrinosphere hides that structure.

Optical thickness and trapping also answer different questions. With mean free path $\lambda=\kappa_{\rm tr}^{-1}$, the diffusion estimate across scale $L$ is $t_{\rm diff}\sim L^2/D\sim3\tau L$, up to geometry-dependent factors. Neutrinos are advected as trapped particles when this escape time is long compared with fluid evolution. Rapid elastic scattering can isotropize them without establishing thermal or chemical equilibrium, which require energy-changing and number-changing reactions respectively.

Diffusion changes more than propagation speed. In equilibrium, trapped neutrinos contribute to pressure and thermodynamics; in transparent matter, they are an escaping radiation field. Leakage schemes interpolate cooling times but do not evolve the full angular radiation stress. Two-moment methods evolve energy and flux but require a closure. Monte Carlo methods retain angular and energy information statistically but introduce sampling noise. Foucart 2023, §§ 4.1–4.3 compares these approximations.

Benchmark suite and rerun boundary

The continuum inputs are a fixed metric, material state $(\rho,T,Y_e)$, flavor-dependent energy-group opacities and emissivities, collision terms, initial neutrino distributions or moments, and boundary conditions. A minimal validation suite contains:

• homogeneous relaxation to Fermi–Dirac equilibrium with detailed balance;
• a constant-opacity diffusion sphere or slab compared with the analytic diffusion solution;
• a transparent pulse propagating at light speed;
• a coupled cell in which charged-current reactions conserve total energy and electron lepton number.

A possible implementation uses binary64 arithmetic, finite-volume spatial transport, logarithmic energy groups, and an implicit–explicit source integrator. Reproduction requires the exact executable and software versions, flavor definitions, weak-rate and equation-of-state tables, energy-bin edges, quadrature or moment closure, source tolerances, and input data. This page supplies analytic checks and a proposed workflow, without claiming an executed transport benchmark.

Use $N=32,64,128$ spatial cells and at least three nested energy grids, such as 12, 24, and 48 groups over the same declared energy interval. Report spatial and spectral $L_1$ errors, equilibrium residuals, maximum $f$ bound violation, diffusion-rate error, free-streaming speed, and fractional total-energy and lepton-number drift. Halve the time step and nonlinear source tolerance independently. In stochastic transport, replace deterministic refinement order by ensemble uncertainty, packet-number scaling, and repeat seeds.

One failure mode is an emission–absorption pair that does not share the same equilibrium distribution. Others are negative occupation or $f>1$, energy-group boundary loss, an incorrect diffusion coefficient from using total instead of transport opacity, explicit-source instability, closure artifacts in crossing beams, and double counting neutrino pressure in both the equation of state and transported stress. Near decoupling, locating a single optical-depth surface can hide the dominant systematic uncertainty.

Exercises

For fixed $j$ and $\kappa$, derive the equilibrium occupation and relaxation time. At $\epsilon=\mu_\nu$, what relation between the rates is required, and what equilibrium would result if blocking were mistakenly omitted?

Solution

Writing the equation as $\dot f=j-(j+\kappa)f$ gives

$$f_{\rm eq}=\frac{j}{j+\kappa},\qquad t_{\rm relax}=\frac1{j+\kappa}.$$

At $\epsilon=\mu_\nu$, Fermi–Dirac equilibrium has $f_{\rm eq}=1/2$, so detailed balance requires $j=\kappa$. Omitting blocking instead gives $\dot f=j-\kappa f$, whose equilibrium is $f=1$ for these same rates. It changes the physical equilibrium rather than merely the notation.

References

• Foucart, François. “Neutrino Transport in General Relativistic Neutron Star Merger Simulations.” Living Reviews in Computational Astrophysics 9, 1 (2023). DOI.
• Mezzacappa, Anthony, Eirik Endeve, O. E. Bronson Messer, and Stephen W. Bruenn. “Physical, Numerical, and Computational Challenges of Modeling Neutrino Transport in Core-Collapse Supernovae.” Living Reviews in Computational Astrophysics 6, 4 (2020). DOI.