For more than thirty years, the people who model how light travels through fog, tissue and paint have disagreed about a coefficient. The disagreement sounds like bookkeeping. It is not: it determines what numbers you get when you shine light into a body and try to read absorption and scattering back out.
The setup
In a strongly scattering medium, tracking every photon is hopeless. After enough scattering events, the radiative transfer equation — the exact but unwieldy description — is replaced by the diffusion equation, which needs one parameter: the diffusion coefficient D. The most commonly used expression for D includes an explicit dependence on the material's absorption.
That is the problem. Absorption is not supposed to appear in D at all, because doing so violates a scaling property of the radiative transfer equation itself. For decades the field has used an approximation while quietly discarding one of the properties of the equation it approximates. Alternative definitions were proposed; no consensus followed, largely because the competing expressions differ by amounts too small to separate in an experiment or a simulation.
The approach
André Liemert and colleagues avoided the comparison problem by going analytic. Rather than fitting expressions to numerically similar curves, they used solutions of the radiative transfer equation to test the diffusion description directly, in two regimes: a short pulse of light, and continuous illumination.
The pulsed case behaves as the theory requires. At long times, the solution converges exactly onto the diffusion description only if D is independent of absorption — the condition the radiative transfer equation imposes.
The continuous case does not. Even at arbitrarily large distances, where diffusion should become increasingly accurate, no single value of D can reproduce both the rate at which the signal decays and its amplitude. The two cannot be matched simultaneously. That is the paradox: steady-state diffusion is a useful approximation that never becomes the exact long-distance limit of light transport in an absorbing material.
What it changes
Diffuse optics underpins near-infrared spectroscopy, diffuse optical tomography, and a wide range of materials measurements in powders, foams and disordered solids. Every one of those techniques solves an inverse problem: measure light at the surface, infer absorption and scattering inside. If the forward model's diffusion coefficient is defined inconsistently, the inferred values inherit the inconsistency — in ways that are hard to notice, because the errors are small and often partly absorbed into calibration.
The practical consequences are uneven. In time-domain work the answer is now unambiguous: use a D that does not depend on absorption. In continuous-wave work the result is a warning rather than a recipe — a single D can be tuned to match decay or amplitude, not both, so any quantitative claim about long-distance steady-state transport has to state which one it matched and why.
The finding does not overturn diffuse optics, and the numerical correction in weakly absorbing tissue may be negligible for many applications. What it does is remove the ambiguity that let the field treat the two definitions as interchangeable, and explain why thirty years of measurement could not settle the question: the difference is not something the apparatus was ever going to resolve. Theory had to do it.
