First-Return Statistics in Henyey-Greenstein Scattering: Colored Motzkin Polynomials and the Cauchy Kernel

TL;DR

This work addresses first-return statistics for photons in three-dimensional Henyey–Greenstein scattering by introducing a Boundary Truncation Factor (BTF) that corrects 1D Motzkin–Catalan combinatorics for boundary constraints. The BTF takes a simple Cauchy-kernel form with parameters $A(g)=1 - g(1+g)/2$ and $m_x(g)=4g/(1-g)$, enabling accurate reconstruction of 3D first-return probabilities from 1D Motzkin counts for $g \lesssim 2/3$, validated by extensive Monte Carlo simulations. For higher anisotropy, a Modified Cauchy kernel with a shape parameter $\alpha(g)$ extends accuracy, with $\alpha(g) = 1 + 0.033 (g - 2/3)/(1-g)$ yielding better fits up to $g \approx 0.95$; this remains phenomenological but strongly supported by data. The resulting framework replaces costly 3D Monte Carlo with efficient 1D polynomial evaluations, benefiting inverse problems and tissue optics applications where many transport evaluations are required, and highlights open questions about the theoretical origin of the Cauchy form and the meaning of the integer coefficients observed.

Abstract

We show that first-return statistics in three-dimensional Henyey Greenstein scattering require a Boundary Truncation Factor (BTF) that takes a Cauchy kernel form. In our previous work (6), we established that first-return probabilities in 1D scattering expand in Catalan and Motzkin numbers. Extending this to 3D anisotropic scattering requires a BTF that, as Monte Carlo reveals, follows a Cauchy kernel:

Figures (4)

• Figure 1: The Boundary Truncation Factor versus scattering order $n$ for $g = 0.1, 0.3, 0.5, 0.7, 0.9$. Points: extracted from Monte Carlo. Curves: Cauchy kernel (equation \ref{['eq:btf_cauchy']}). The width $m_x(g)$ increases with $g$; the amplitude $A(g)$ decreases. At $g = 0.9$, deviations from Cauchy become visible. (Figures use "path length $m_s$" following MC convention; $m_s \equiv n$.)
• Figure 2: Deviations from the Cauchy kernel versus scaled path length. At high $g$, Monte Carlo runs high near the peak and low in the tail---the signature of lighter-than-Cauchy tails.
• Figure 3: Path length distributions $P(m_s|g)$ for six anisotropy values. Gray points: Monte Carlo. Blue solid: Cauchy kernel ($\alpha = 1$). Red dashed: Modified Cauchy kernel. At $g \leq 0.7$, the curves are indistinguishable; at $g \geq 0.85$, the modified kernel captures the lighter tails.
• Figure 4: Shape parameter $\alpha$ versus transport mean free path ratio $\ell^*/\ell = 1/(1-g)$. The Cauchy case $\alpha = 1$ occurs at $\ell^*/\ell = 3$.