A thin layer, many routes
A beam crossing a cloudy material reaches the far side by many paths. In the peer-reviewed experiment, the material was a roughly 10-micrometre layer of zinc-oxide particles, illuminated with 532-nanometre laser light. A spatial light modulator changed the incoming wavefront so that light from those paths could meet at selected points beyond the layer. The team measured a transmission matrix, a map of how changes at the input appeared at the output. Their question was as much about variation as about focus: if the same method is applied to different parts of a scattering sample, how stable is the gain?[1]
The researchers compared the intensity at a focus with an average reference intensity; that ratio is the enhancement factor. They selected between 100 and 1,000 controlled input channels and one to five output targets. Three independent measurements of the transmission matrix helped bound experimental uncertainty. An exact finite-size random-matrix calculation captured the measured average enhancement. A simpler formula intended for very large numbers of channels missed part of that finite-size behaviour. The size of the apparatus matters here: a mathematical limit can be accurate in its own regime and still misdescribe a laboratory's reachable one.[1]
The spread carries the clue
The more revealing comparison concerned the spread of possible gains. With fewer controlled channels, the measured distribution agreed with a model that treats scattering paths as independent for this purpose. As the team controlled more channels under strongly scattering conditions, the observed distribution broadened beyond that model's prediction. Numerical simulations of weak and strong scattering reinforced the distinction. The researchers interpret the extra spread as a sign of long-range correlations created when many wave paths interfere. An average can remain well described while the full distribution exposes a missing physical connection.[1]
That reading narrows the next experiment. The paper argues that fluctuations can flag these correlations without measuring an almost complete and much larger transmission matrix; under suitable conditions, roughly 200 or fewer controlled inputs may be informative. The statistic is a probe, not a picture of every route through the material. It also does not establish medical imaging through living tissue, although such imaging motivates parts of the field. The unsolved task is a predictive account of the broad distribution once correlations become strong. A repeat on other scattering samples could test how far this laboratory result travels.[1]