🔬 optics

Largest eigenvalue statistics of wavefront shaping in complex scattering media

By combining experiments, simulations, and random-matrix theory, this study reveals that while the distribution of wavefront shaping enhancement factors aligns with theoretical predictions in weakly scattering media, strongly scattering systems exhibit giant fluctuations that serve as a sensitive probe for long-range mesoscopic correlations.

Original authors: Grégory Schehr, Hasan Yılmaz

Published 2026-07-14

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Grégory Schehr, Hasan Yılmaz

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to shout a secret message through a thick, foggy forest. Normally, the trees scatter your voice in every direction, and by the time it reaches the other side, it's just a messy jumble of noise. But what if you could tweak your voice before you start shouting, shaping the sound waves so they all bounce off the trees and land perfectly on a single ear? That's the magic of wavefront shaping.

In this study, researchers Gregoire Schehr and Hasan Yilmaz asked a very specific question: If we try to focus light (or sound) onto multiple secret spots in this foggy forest at the same time, how much louder will it get? And more importantly, how much does that "loudness" jump around from one try to the next?

The "Loudness" Meter

The scientists use a number called the enhancement factor to measure this. Think of it as a volume knob. If you just shout randomly, the volume at your target is low. If you shape your wavefront perfectly, the volume spikes. The paper defines this factor, $\eta$ , as the ratio of the focused intensity to the average intensity you'd get without any special shaping.

They found that for a single target, this is well understood. But when you try to hit multiple targets (say, 2, 3, or 5 spots) simultaneously, things get tricky. The researchers wanted to know: Is the "loudness" predictable, or does it go wild?

The "Perfectly Random" Guess

To get a baseline, the team used a mathematical tool called Random Matrix Theory. Imagine you have a giant bag of marbles, and every time you pull one out, it's a completely random number, with no connection to the marble you pulled before. This is the "Laguerre-Wishart" model. It assumes the foggy forest is just a chaotic mess with no hidden patterns.

Using this "perfectly random" model, they calculated exactly what the average loudness should be and how much it should wiggle around. They even wrote down a precise formula for the probability of getting a certain volume.

Here is the big surprise:
When they tested this with real light and a real scattering sample (a layer of zinc oxide nanoparticles about 10 µm thick), the "random" model worked perfectly... but only when they controlled a small number of input channels (around 100 or fewer).

However, as soon as they increased the number of input channels they could control (going up to 1,000), the real world started to act weird. The "loudness" didn't just wiggle a little; it went giant. The fluctuations became much larger than the "perfectly random" model predicted.

The "Hidden Teamwork" in the Fog

Why did the random model fail? The paper argues that the foggy forest isn't just a bag of random marbles. The light waves are actually "talking" to each other over long distances.

Think of it like a chaotic party. In a truly random party, everyone talks to their immediate neighbor. But in this scattering medium, the waves are like guests who can hear conversations happening across the entire room. These are called long-range mesoscopic correlations. They arise because light takes many different paths through the material, and these paths interfere with each other in a coordinated way.

The paper explicitly rules out the idea that the "random marble" model (the Laguerre-Wishart ensemble) is enough to describe what happens when you control many channels. The model underestimates the wild swings in intensity. The "giant fluctuations" are a direct signature of this hidden teamwork among the light waves.

The Detective Work

How sure are they?

The Average: They are very sure that the average loudness doesn't change much, even with these correlations. The "random" model still predicts the average correctly.
The Fluctuations: They are equally sure that the fluctuations (the wiggles) get huge in strongly scattering media.
- They measured this in the lab using a 10 µm thick layer of zinc oxide nanoparticles.
- They simulated this on computers using 2D slabs with thicknesses ranging from k₀L = 1 (very thin) to k₀L = 200 (very thick).
- In the simulations, as the slab got thicker, the fluctuations grew even larger, far exceeding the "random" prediction.

The paper suggests that these giant fluctuations are a new, simple way to detect these long-range correlations. Previously, scientists needed to measure massive transmission matrices (thousands of inputs and outputs) to see these correlations. Now, they can just look at how much the "loudness" fluctuates, even with a small number of targets (like M₂ = 1 or 2) and a moderate number of inputs (M₁ < 200).

What It All Means

The authors conclude that the "enhancement factor" isn't just a scorecard for how well you can focus light. It's also a statistical detective. If the loudness fluctuates wildly, it tells you that the material has these complex, long-range connections.

They don't claim to have solved the math for exactly how to predict these giant fluctuations yet (that's still an open problem), but they have proven that the simple "random" math isn't enough. They've shown that in the real world, light in a scattering medium is more like a coordinated dance than a chaotic brawl, and if you control enough of the dancers, the whole group starts to swing wildly in ways you can't predict with a simple coin toss.

So, the next time you try to focus light through a messy material, remember: if the intensity jumps around like a startled cat, it's not just noise. It's the material whispering its secrets through long-range correlations.

Technical Summary: Largest Eigenvalue Statistics of Wavefront Shaping in Complex Scattering Media

Problem Statement
Wavefront shaping enables the coherent control of waves (light, sound) to focus through complex scattering media, such as biological tissue or disordered materials. The efficacy of this process is quantified by the enhancement factor ( $\eta$ ), defined as the ratio of the focused intensity at target positions to the average intensity of unshaped wavefronts. For a single target, $\eta$ is maximized by digital optical phase conjugation. For multiple targets ( $M_2 > 1$ ), the maximum enhancement corresponds to the largest eigenvalue ( $\lambda_{\text{max}}$ ) of the focusing operator $A^\dagger A$ , where $A$ is the $M_2 \times M_1$ sub-transmission matrix of the medium.

While the mean enhancement factor $\langle \lambda_{\text{max}} \rangle$ is well-understood to scale as $M_1/M_2$ (where $M_1$ is the number of controlled input channels), the statistical fluctuations of this factor ( $\Delta \lambda_{\text{max}}$ ) and its full probability distribution remain largely unexplored, particularly for finite system sizes and in the presence of long-range mesoscopic correlations. Existing asymptotic theories (e.g., Marčenko–Pastur law) often fail to accurately describe finite-size experimental regimes, and the impact of correlation-induced deviations on reproducibility is unclear.

Methodology
The authors employ a tripartite approach combining experimental measurements, numerical simulations, and exact random-matrix theory (RMT):

Experimental Setup: Transmission matrices were measured using phase-shifting interferometry on a scattering sample consisting of a $\sim 10\,\mu\text{m}$ thick layer of densely packed Zinc Oxide (ZnO) nanoparticles ( $L \gg l_t$ , where $l_t$ is the transport mean free path). The input was modulated by a Spatial Light Modulator (SLM) with $M_1$ degrees of freedom, and output intensities were measured at $M_2$ target positions.
Numerical Simulations: Recursive Green's function algorithms were used to simulate wave propagation through 2D diffusive slabs of varying thicknesses ( $k_0 L = 1, 50, 200$ ) and normalized transport mean free paths.
Theoretical Framework: The authors derived exact finite-size predictions using the Laguerre–Wishart (LW) random-matrix ensemble. They modeled the sub-transmission matrix $A$ as having independent complex Gaussian entries (the standard assumption for open scattering systems without long-range correlations). This allowed for the analytical calculation of the joint probability density function (JPDF) of eigenvalues and the cumulative distribution function (CDF) of the largest eigenvalue, $Q_{\text{max}}(x)$ , expressed via determinants of incomplete gamma functions.

Key Contributions and Results

Exact Finite-Size Theory: The paper provides exact analytical expressions for the mean $\langle \lambda_{\text{max}} \rangle$ and the full probability distribution $P_{\text{max}}(\lambda_{\text{max}})$ for finite $M_1$ and $M_2$ . These predictions, based on the LW ensemble, accurately describe experimental data in the regime where long-range correlations are negligible (e.g., weakly scattering media or small $M_1$ ). This framework supersedes the asymptotic Marčenko–Pastur prediction, which deviates significantly for the finite channel numbers typical of experiments ( $M_2 \in [1, 5]$ ).
Validation of Mean Enhancement: Both experimental and numerical results confirm that the mean enhancement factor is robust against long-range mesoscopic correlations. The scaling $\langle \lambda_{\text{max}} \rangle \sim M_1/M_2$ holds regardless of sample thickness or correlation strength, consistent with the LW baseline.
Discovery of Giant Fluctuations: The primary finding is that while the mean is unaffected, the fluctuations of the enhancement factor are highly sensitive to long-range mesoscopic correlations.
- In weakly scattering regimes (thin slabs or small $M_1$ ), experimental fluctuations match the LW theoretical predictions.
- As the number of controlled input channels ( $M_1$ ) increases or the sample thickness grows (stronger scattering), the experimentally observed standard deviation $\Delta \lambda_{\text{max}}$ systematically exceeds the LW prediction.
- Numerical simulations on thick slabs reveal "giant" fluctuations that can be several times larger than the correlation-free baseline.
Correlation as a Probe: The deviation of fluctuations from the parameter-free LW baseline serves as a direct signature of long-range mesoscopic correlations. The authors demonstrate that these correlations can be detected using relatively small datasets ( $M_1 \lesssim 200$ , $M_2=1$ ) and thin samples ( $\approx 10 l_t$ ), whereas previous methods required measuring massive transmission matrices ( $M_1, M_2 > 2000$ ) through very thick samples.
Independence of Intensity Balance: The study further shows that the magnitude of the enhancement factor ( $\lambda_{\text{max}}$ ) is not correlated with the intensity balance between multiple targets; high enhancement can occur with either equal or highly unequal intensity splits among targets.

Significance and Claims
The paper establishes the enhancement factor not merely as a performance metric for focusing, but as a sensitive statistical observable. The authors claim that:

Statistical Foundation: The finite-size LW framework provides a rigorous, parameter-free baseline for predicting wavefront shaping performance in the absence of correlations.
Diagnostic Tool: Fluctuations in the enhancement factor offer a simple, experimentally accessible probe for detecting long-range mesoscopic correlations in complex scattering media. This bridges the gap between wavefront shaping experiments and the fundamental physics of mesoscopic wave transport (e.g., universal conductance fluctuations, Anderson localization).
Limitations of Current Models: The results highlight that standard random-matrix models (assuming independent entries) are insufficient for strongly scattering media with many controlled channels, as they fail to capture correlation-induced giant fluctuations.
Broad Applicability: While demonstrated with optical waves, the underlying physics relies on multiple scattering and finite-size statistics, suggesting the findings extend to acoustic, elastic, microwave, and matter-wave systems.

The authors conclude that establishing a predictive mapping between mesoscopic multiple-scattering theory and correlated random-matrix theory remains an open problem, but their work successfully identifies the statistical signatures of these correlations in the extreme eigenvalue statistics of wavefront shaping.