One scattering event or many

Count interactions along the path

A wave that emerges from a thin cloud may have interacted once, many times or not at all. Single scattering retains only paths with one relevant encounter. Multiple scattering includes sequences in which an outgoing wave from one part of the medium becomes an incoming wave for another. The distinction is about the interaction history, not merely whether the final beam looks diffuse.

Thin and thick are therefore relative descriptions. A physically short sample can be optically thick if its interaction strength and number density are high. A long dilute sample can remain close to a single-scattering regime. The appropriate scale is the distance over which interactions become likely, together with the length scales controlling phase and directional memory. These scales depend on wavelength, target state and geometry. Reporting only the sample thickness leaves out the information needed to decide which transport model is justified.

Optical depth gives a first estimate

In a uniform medium of independent scatterers, let the number density be n and the relevant total scattering cross section be σ. The mean free path is ℓ = 1/(nσ), and the optical depth of a straight path of length L is τ = L/ℓ. Under a Poisson model, the probability of no interaction is P₀ = exp(−τ), while exactly one interaction has probability P₁ = τ exp(−τ).

For τ = 0.10, these probabilities are approximately 0.905 and 0.0905. The probability of two or more encounters is about 0.00468. For τ = 2, P₀ is approximately 0.135, P₁ approximately 0.271 and the remainder approximately 0.594. These original calculations show why a linear single-event approximation becomes unreliable as optical depth increases. They are not a full model of a finite slab: after a deflection, the actual remaining path length depends on direction and the sample boundaries, which a proper transport calculation must include.

Directional memory has its own length scale

The mean free path measures encounter spacing, but each encounter may only slightly deflect the wave. A strongly forward-peaked scattering process can retain directional memory over many events. Transport therefore introduces an angular measure, often the average cosine g = ⟨cos θ⟩. In a common independent-scattering approximation, the transport mean free path is ℓ* = ℓ/(1 − g).

If g = 0, the two scales coincide. If g = 0.90, the transport length is ten times the encounter mean free path. A sample can then contain numerous interactions without fully randomizing direction. These values are illustrative and the formula has a defined transport context; it is not a universal law for coherent structured media. The practical lesson is that measuring total attenuation and measuring angular broadening provide different information. A single transmission number cannot determine both the encounter rate and the loss of directional memory without additional assumptions.

Ballistic, diffuse and absorbed components differ

The ballistic component reaches the detector without the relevant scattering event. Diffuse transmission arrives after one or more redirected paths. Absorption transfers energy into material degrees of freedom or other channels that are not simply another propagating direction of the same wave. A narrow forward detector can lose both scattered and absorbed power, so its attenuation combines processes that an application may need to separate.

An integrating measurement can capture more of the redistributed radiation, but it still requires a calibrated collection geometry and careful treatment of reflection. Time-resolved detection provides another distinction because longer paths can arrive later. That timing is influenced by group velocity and boundary conditions as well as the number of encounters. Avoid calling every delayed photon a particular scattering order unless a model supports that assignment. The useful decomposition is the one your instrument can discriminate, with the unresolved part represented honestly in the uncertainty.

A random walk is an approximation

At a coarse scale, repeated scattering can resemble a random walk. After sufficient directional randomization, a diffusion equation may describe the average transport. The diffusion coefficient depends on an energy-transport velocity and a transport length in the relevant dimensionality. This description is often useful far from boundaries and beyond the initial few encounters, but those conditions need checking.

Near an interface, at early times or in a strongly absorbing medium, a diffusion approximation can fail. Coherent interference among paths can also produce effects absent from an incoherent random walk. A system that is deterministic in each realization may look random after averaging over configurations. Choose the model according to the observable: average intensity transport may tolerate coarse graining, while speckle correlations or phase-sensitive recovery require more detailed field information. A convenient diffusion fit should not be mistaken for proof that all microscopic phase relations have disappeared.

Coherent multiple scattering can be controlled

A static scattering medium can connect incident spatial modes to outgoing modes through a complex transmission matrix. Multiple paths contribute to each matrix element. Their interference generates a complicated output, but a stable relationship can remain between controlled input phases and measured output fields. Wavefront shaping uses that relationship to favor a desired output, such as a focus behind the sample.

The Vellekoop–Mosk experiment provides a concrete example of that possibility in a strongly scattering optical system. Its significance here is the survival of useful control despite complicated propagation. Recovery is limited by the number of modes addressed, calibration quality, absorption and the time over which the sample remains stable. It does not imply that a small detector can reconstruct every environmental degree of freedom. Nor does it turn the scattering medium into a source of extra energy. The improved focus redistributes available transmitted energy toward the chosen observation point.

Thickness and time are discriminating controls

A thickness series can reveal the transition from an approximately linear single-event signal to nonlinear transport. Keep concentration, particle state and illumination comparable while changing path length. If those quantities change with thickness, the experiment no longer isolates the desired control. Monitor both forward transmission and a broader angular or total-output measure to distinguish simple removal from redistribution.

Time-resolved data add the path-length distribution, while polarization-resolved data can probe how repeated encounters change orientation information. A useful analysis predicts several measurements together. Fitting only a mean transmitted intensity may leave scattering and absorption parameters highly correlated. Report that identifiability problem rather than assigning precise values to both. A second observable can break the ambiguity more effectively than acquiring many more repeats of the same one. The experiment should be designed around the parameters it aims to identify, not merely around the signal that is easiest to record.

Choose a model with a clear domain

Use a single-scattering model when repeated events are demonstrably negligible for the chosen signal. Use transport or radiative-transfer methods when many paths contribute and an intensity-level description is appropriate. Use coherent field models when interference, correlations or phase recovery are central. These are levels of description, not a ranking in which the most elaborate calculation is automatically the best.

Validate the selected level against controls it did not fit. A model that predicts angular broadening but misses the arrival-time distribution may be omitting relevant path or material physics. A diffusion fit that works only at late times should be reported with that time window. Such boundaries make the result portable and useful. They also protect the central interpretation: a wave's original narrow pattern can become unrecognizable through repeated ordinary interactions while energy and, in suitable conditions, coherent information continue to occupy other accessible channels.

Distinguish an event count from a transport length

The transport mean free path measures directional randomization, while the ordinary scattering mean free path measures how frequently events occur. Their difference has observable consequences. A strongly forward-scattering sample can remove light from a narrow ballistic beam after only a few events while still transporting much of its energy in roughly the original direction. Treating the two lengths as identical would predict an isotropic diffuse field too early.

In the illustrative case g = 0.9, one event preserves an average directional cosine of nine tenths. If successive deflections obey the independent angular model, the directional correlation after n events scales as g to the power n. Ten events leave approximately 0.349 of the initial correlation; twenty leave about 0.122. These numbers are model calculations, not universal measurements of turbid matter. Anisotropy, boundaries or correlations can change them.

Diffusion is a later approximation, appropriate when directional memory has sufficiently relaxed and observations average over many paths. Near a source, near an interface, or at very early times, a diffusion equation can give misleading answers even when the bulk sample is strongly scattering. A careful experiment therefore compares ballistic transmission, angular spreading and arrival-time distributions. Together they constrain event frequency and directional persistence, which a single attenuation coefficient generally cannot disentangle.

This distinction remains essential when the detector resolves only the earliest arriving component.

Go to the original work.

These links identify the sources behind specific claims. Verification scope varies by source and is stated below.

1. 1 / University teachingQuantum Physics III: Scattering and Identical Particles ↗

   Scattering states, differential cross sections, phase shifts and their assumptions.

   Verification scope: MIT course page and lecture topic descriptions inspected; this is a teaching reference, not a new experiment.

2. 2 / Primary experimentVellekoop and Mosk, Focusing coherent light through opaque strongly scattering media ↗

   The reported control of incident wavefronts to focus transmitted multiply scattered light.

   Verification scope: University author-repository abstract and publication metadata inspected; no claim of a full-paper methods audit.

3. 3 / University teachingMIT Physics III: Vibrations and Waves ↗

   Wave equations, interfaces, optics and propagation in media.

   Verification scope: MIT course outline inspected; article derivations are independent textbook-level explanations.