Multiple Decorrelation in Complex Scattering

• Multiple decorrelation property is a phenomenon that defines how the interplay of surface and volumetric scattering leads to the breakdown of the memory effect in complex wave scattering.
• The framework employs a superposition model combining thin-layer (surface) and volumetric (diffusive) contributions, using integral formulations to predict decorrelation accurately.
• Experimental validations with controlled samples show that the model captures microstructural details and sets bounds on imaging performance in turbid media.

The multiple decorrelation property describes the phenomenon where, in the context of complex wave scattering—such as optical waves in turbid media—the statistical correlations of the backscattered field break down due to multiscattering, leading to a loss of the so-called memory effect. This property captures how the interplay between surface and volumetric scattering dictates the decay of correlations in the scattered intensity, with significant implications for imaging through disordered media.

1. First-Principle Mechanism for Decorrelation

The statistical structure of the backscattered wave field is described by a superposition of two principal contributions: a surface or thin-layer scattering term and a volumetric (diffusive) scattering term. The starting point is a first-principle formulation based on the Lippmann–Schwinger equation, which allows the decomposition of the normalized decorrelation function $C(u)$ (as a function of the aperture displacement $u$) as

$$C(u) = \frac{1}{1+\kappa}\, C_{2d}(u; p) + \frac{\kappa}{1+\kappa}\, C_{3d}(u; p, z_0)$$

Here,

• $C_{2d}(u; p)$ accounts for surface scattering in a thin-layer region of thickness $\Delta = \sqrt{\lambda f / 2}$, with $\lambda$ the wavelength and $f$ the focal length.
• $C_{3d}(u; p, z_0)$ models volumetric, diffusive scattering throughout the full sample with thickness $z_0$.
• $\kappa$ is a dimensionless coefficient quantifying the weight of the volumetric term relative to surface scattering and depends on scattering depth, particle properties, and optical setup.

The thin-layer term retains detailed microstructure information, as the first-Born approximation holds and structure functions (e.g., the scatterer size distribution $p(\rho)$) express themselves in the autocorrelation sidelobes. The volumetric term absorbs the loss of correlation due to random walk–like diffusive transport, suppressing sidelobes as $\kappa$ increases.

2. Mathematical Formulation of the Decorrelation Terms

The statistical contributions are specified by precise integral formulas:

• Thin-layer (surface) term:

$$C_{2d}(u; p) = \frac{\int \rho^2\, \mathrm{Jinc}\left(\frac{\rho k_0 |u|}{f}\right) p(\rho) d\rho}{\int \rho^2\, p(\rho) d\rho}$$

with $k_0 = 2\pi/\lambda$, $p(\rho)$ the size distribution, and $\mathrm{Jinc}(x) = 2J_1(x)/x$.

• Volumetric (diffusive) term:

$$C_{3d}(u; p, z_0) = \frac{1 - \exp\left(-\left[1 + a(p)\frac{|u|^2}{f^2}\right] \chi(p, z_0)\right)}{(1 - \exp(-\chi(p, z_0)))(1 + a(p) |u|^2/f^2)}$$

with

$$\chi(p, z_0) = \frac{\beta z_0}{\rho_z(p)},\quad a(p) = \frac{k_0^2 \langle \tan \Delta\theta \rangle^2 \rho_z^2(p)}{2\beta}$$

and $\kappa$ (mixing coefficient):

$$\kappa(p, z_0) = \frac{k_0 \gamma \epsilon \rho_z(p) (1 - \exp(-\chi(p, z_0)))}{\pi\beta f}$$

where $\rho_z(p)$ is the mean scatterer size, $\gamma$ the filling factor, $\epsilon$ a vertical coherence length, and $\langle \tan \Delta\theta \rangle$ the mean single-scatter deflection.

Fluctuations in a single realization encode additional microstructural information, predicted by the quasi–power law: $$\frac{\sqrt{\langle |\delta\mu|^2 \rangle}}{\langle \mu \rangle} = c + \frac{a^2}{\Delta^2} \int \frac{\Delta}{\rho} p_v(\rho) d\rho$$ where $p_v(\rho)$ is the particle volume–weighted distribution and $a$ (typically constant) represents surface defect depth.

3. Experimental Validation and Quantitative Agreement

Experimental measurements with potassium chloride powders of well-controlled size distributions and various thicknesses validated the model across a range of scenarios. A custom-shaped beam illuminated the turbid samples, and the recorded speckle intensity autocorrelations were quantitatively compared to the theoretical predictions.

• The superposition model $C(u)$, incorporating both $C_{2d}$ and $C_{3d}$, fits the measured data with high accuracy (e.g., combined $L_1$ loss $<0.002$ across 16 datasets and only three global parameters needed).
• Single-realization fluctuation levels were observed to follow the predicted quasi–power law, supporting the theoretical description of fluctuation statistics.

Neither previous pure-diffusion models nor partial approaches (such as the PEACE model) captured the empirically observed visibility and curvature of the autocorrelation sidelobes. The new framework's explicit handling of both surface and volumetric contributions provides superior predictive power.

4. Implications for Imaging and Applications

The decorrelation framework rigorously bounds the performance of imaging techniques relying on the memory effect. Any multiple scattering not confined to the thin layer fundamentally restricts the ability to maintain correlation between different illumination and detection points.

• Memory Effect: The range of the memory effect—i.e., the angular span over which correlations persist—shrinks with increasing $\kappa$ as volumetric scattering dominates.
• Turbid Media Imaging: Strong volumetric decorrelation limits the field of view and resolution in imaging through turbid media, as readable information is effectively "scrambled" beyond a critical thickness.
• Wavefront Shaping: Quantification of the tradeoff between microstructure-resolved contrast (from $C_{2d}$) and decorrelated background (from $C_{3d}$) enables more robust calibration and optimization of wavefront-shaping and imaging protocols.

The explicit quantification of how scatterer statistics (distribution, size, filling factor) and sample thickness control the breakdown of correlation can inform practical parameter choices in experimental design.

5. Significance of Multiple Decorrelation in Complex Scattering

The comprehensive model links the multiple decorrelation property—here, the concurrent suppression of sidelobe visibilities over many independent field directions—to both micro- and macro-structural features of the scattering medium. The model's key insight is that:

• The multiple decorrelation phenomenon is governed by the balance between single-scattering (memory preserving) and multi-scattering (decorrelating) processes, with volumetric diffusion ultimately overwhelming the localized angular memory as thickness increases.
• The extent and detailed shape of the decorrelation can be predicted for arbitrary multiscale structures, reducing a purely empirical or phenomenological process to a deterministic function of sample statistics.

In addition, the framework shows that fluctuations in a single experimental realization are not mere statistical noise, but in fact encode microstructural details according to a predicted scaling law.

6. Limitations and Future Directions

The presented approach is based on well-characterized, essentially non-absorbing and non-resonant, isotropic scatterers. Extensions to regimes involving strong absorption, anisotropic or resonant systems, or exotic near-field effects require further adaptation of the model. Additionally, the quantitative use of the model in biological tissue or engineering applications with unknown or broad scatterer distributions may be limited by the lack of accurate input parameters for $p(\rho)$ or other statistical properties.

A plausible implication is that, by linking decorrelation directly to scatterer statistics, the approach could be extended to provide diagnostic information or to engineer media with tailored decorrelation properties for specific imaging applications.

Conclusion

The multiple decorrelation property in complex wave scattering, as rigorously formulated in this work (Zhang et al., 15 Apr 2025), provides a quantitative, physically grounded explanation of the limitations imposed on the memory effect by multiple scattering. The framework decomposes the overall ensemble statistics of the scattered intensity into explicit surface and volumetric terms, mathematically characterizes their relative influence, and validates these predictions with experimental evidence across a range of parameters. The property highlights a fundamental transition: as volumetric scattering dominates, correlations are lost across all directions, thereby bounding the information content accessible to any memory-effect-based imaging technique. This theoretical and experimental integration establishes decorrelation as both a practical limitation and a diagnostic tool in complex optical systems.