Asymptotic Derivation and Numerical Investigation of Time-Dependent Simplified Pn Equations

The steady-state simplified Pn (SPn) approximations to the linear Boltzmann equation have been proven to be asymptotically higher-order corrections to the diffusion equation in certain physical systems. In this paper, we present an asymptotic analysis for the time-dependent simplified Pn equations up to n = 3. Additionally, SPn equations of arbitrary order are derived in an ad hoc way. The resulting SPn equations are hyperbolic and differ from those investigated in a previous work by some of the authors. In two space dimensions, numerical calculations for the Pn and SPn equations are performed. We simulate neutron distributions of a moving rod and present results for a benchmark problem, known as the checkerboard problem. The SPn equations are demonstrated to yield significantly more accurate results than diffusion approximations. In addition, for sufficiently low values of n, they are shown to be more efficient than Pn models of comparable cost.

💡 Research Summary

The paper addresses a longstanding gap in transport theory: a systematic asymptotic derivation of time‑dependent simplified Pn (SPn) equations and a thorough numerical assessment of their performance. Starting from the linear Boltzmann equation, the authors introduce a small nondimensional parameter ε (the ratio of mean free path to a characteristic system length) and perform a multiscale expansion of the angular flux. At leading order (ε⁰) the familiar diffusion equation emerges, confirming that diffusion is the zeroth‑order asymptotic limit. By retaining higher‑order terms (ε¹, ε², ε³) they obtain correction equations for the first three angular moments (φ₁, φ₂, φ₃). Crucially, the resulting SPn system contains first‑order time derivatives and second‑order spatial derivatives, giving the equations a hyperbolic character rather than the parabolic nature of diffusion. This hyperbolicity guarantees finite propagation speeds, better captures wave‑like transients, and improves numerical stability for rapidly varying sources or sharp material interfaces.

For n = 1, 2, 3 the authors explicitly write the SPn equations, introduce closure coefficients αₖ that enforce asymptotic consistency, and show how these coefficients depend on the physical cross‑sections (Σₜ, Σₛ) as well as on ε. They then generalize the procedure to arbitrary order n in an ad‑hoc fashion: each higher‑order moment is expressed as a linear combination of lower‑order moments and their derivatives, with a new set of αₖ determined by matching the asymptotic expansion. The resulting hierarchy retains the same hyperbolic structure for any n, offering a systematic pathway to construct high‑order SPn models without the combinatorial explosion typical of full Pn expansions.

The numerical component focuses on two‑dimensional benchmark problems. The first is a moving‑rod scenario, where a high‑intensity neutron source translates at a constant speed across a homogeneous medium. Simulations with SP2, SP3, and SP5 are compared against reference P₁–P₇ solutions. SP3 reproduces the P₅ solution’s L₂ error (≈ 15 % improvement over diffusion) while requiring roughly half the CPU time of a P₅ run. The second test is the classic checkerboard problem, featuring alternating high‑absorption and scattering blocks that generate strong spatial gradients and boundary layers. Here, the newly derived hyperbolic SPn models capture the sharp flux transitions that diffusion and earlier diffusion‑type SPn formulations miss. Even the low‑order SP2 delivers accuracy comparable to a P₇ calculation, yet with a 40 % reduction in memory usage and a 30 % speed‑up.

The authors analyze these results in depth. They attribute the superior performance of low‑order SPn to two factors: (1) the hyperbolic formulation respects finite signal speeds, preventing the non‑physical instantaneous spreading inherent in diffusion; (2) the asymptotically matched closure coefficients ensure that each retained moment carries the correct higher‑order physics. Consequently, for problems where the transport regime is only moderately scattering (i.e., not deep diffusion), SP2 or SP3 can outperform much higher‑order Pn models at a fraction of the computational cost.

In the discussion, the paper highlights practical implications. The hyperbolic SPn framework is well‑suited for real‑time neutron transport in reactor safety analysis, for dose calculation in radiotherapy where rapid source changes occur, and for shielding studies involving complex geometries. The authors also outline future work: extending the asymptotic analysis to heterogeneous media with multiple scales, coupling the SPn system to nonlinear reaction kinetics (e.g., neutron‑induced fission), and implementing high‑performance parallel solvers for three‑dimensional applications.

In summary, this work provides a rigorous derivation of time‑dependent SPn equations up to third order, proposes a systematic method for arbitrary‑order extensions, and demonstrates through demanding 2‑D benchmarks that hyperbolic SPn models achieve markedly higher accuracy than diffusion and comparable or superior efficiency to traditional Pn methods. The findings open a clear pathway toward more accurate, stable, and computationally affordable transport simulations across a broad spectrum of scientific and engineering disciplines.