Branched Flow of Light in Disordered Media

• Branched flow of light is the formation of filamentary caustics when coherent optical waves propagate through weak, smoothly varying disorder, characterized by universal power-law scaling laws.
• The phenomenon is analyzed using ray and wave approaches, with models predicting caustic distances scaling as ⟨zc⟩ ∝ ε⁻²/³ ℓc^(4/3) and non-Gaussian intensity statistics.
• Experimental realizations in disordered waveguides, liquid crystals, and gradient-index media highlight its applications in photonic channeling, secure communications, and enhanced imaging.

Branched flow of light is the emergence of narrow, filamentary intensity channels—termed branches or caustics—when coherent or partially coherent optical waves propagate through weakly disordered, smoothly varying refractive-index landscapes. This phenomenon unifies nonlinear dynamics, wave-optics, and random media, and has direct analogues in electronic, acoustic, and fluid systems. It is characterized by a universal power-law scaling for the typical distance to the first branch (caustic), non-Gaussian statistics of intensity, and remarkable robustness to disorder. Branched flow persists in both linear and weakly nonlinear regimes and can be controlled, shaped, and engineered for diverse applications in photonics and wave physics.

1. Physical Mechanism and General Properties

Branched flow arises when a wave or ray, initially launched from a point source or a quasi-plane wave, traverses a medium whose refractive index varies randomly but smoothly on a scale $\ell_c$ much greater than the optical wavelength $\lambda$ (Mattheakis et al., 2017, Heller et al., 2019, Lin et al., 2021). Individual rays receive a sequence of weak refractive "kicks," leading to the folding of the propagating wavefront and concentration of energy into filamentary caustics (branches). These branches are regions of strongly enhanced intensity, interleaved by wide, low-intensity voids.

Formation of branches is reproducible within a given realization and occurs at distances far shorter than those required for angular diffusion. The caustics are described by singularities in the mapping from initial angle or position to final transverse coordinate, producing fold and cusp catastrophes. In the presence of strong scattering, branches overlap and rare, extreme intensity events akin to optical rogue waves can appear (Mattheakis et al., 2017).

Branched flow is universal for linear waves propagating in smooth, weak random potentials, with analogous phenomena observed for electrons in 2DEGs, microwaves, acoustic waves, and even surface gravity waves in oceans (Heller et al., 2019, Mattheakis et al., 2017).

2. Mathematical Formulation and Scaling Laws

2.1 Ray and Wave Descriptions

In the geometrical-optics (ray) picture, the trajectory $\mathbf{r}(z)$ for propagation along the $z$ direction satisfies

$$\frac{d^2\mathbf{r}}{dz^2} = -\frac{1}{n_0}\nabla_{\perp} \delta n(x,y,z)$$

where $n_0$ is the background index and $\delta n(x,y,z)$ is the small random fluctuation.

The paraxial wave equation for the optical envelope $\psi(x,y,z)$ reads

$$2ik_0 \partial_z \psi + \nabla_{\perp}^2\psi + k_0^2 \delta n^2(x,y,z)\psi = 0$$

where $k_0 = 2\pi/\lambda$. The refractive-index fluctuation $\delta n^2$ acts as a random potential in this Schrödinger-type equation (Mattheakis et al., 2017, Heller et al., 2019, Lin et al., 2021).

2.2 Universal Scaling for the Branching (Caustic) Length

For a zero-mean, stationary Gaussian random $\delta n(x, z)$ with rms amplitude $\varepsilon$ and correlation length $\ell_c$, the mean distance $\langle z_c \rangle$ from the source to the first caustic obeys a universal scaling law: $$\langle z_c \rangle \propto \ell_c\, \varepsilon^{-2/3}$$ or more generally, with statistical prefactors (Heller et al., 2019, Lin et al., 2021, Mattheakis et al., 2017): $$\langle z_c \rangle = C\, D^{-1/3} \sim \varepsilon^{-2/3} \ell_c^{4/3}$$ where $D\sim \varepsilon^2 / \ell_c^4$. For branched flow in higher dimensions and with anisotropic or rotated disorder, generalized forms $$d_f \propto \varepsilon^{-2/3}\, A^{4/3}\, \ell_x\,[\cos^2\alpha + ( \sin^2\alpha)/A^2 ]^{5/6}$$ capture directional dependencies (Chang et al., 20 Jan 2026).

Key statistical signatures include a heavy-tailed distribution for intensity (non-Rayleigh), with the probability of high intensities $P(I) \sim I^{-\alpha}$, $\alpha \approx 3$ for large $I$ (Heller et al., 2019).

3. Regimes: Linear, Nonlinear, and Incoherent Branched Flow

3.1 Linear and Weak Scattering

In linear, weak-scattering regime ($\varepsilon \ll 1$), branching is dominated by single-scattering per correlation length, producing well-defined filaments and moderate intensity fluctuations. The inter-branch spacing follows the universal scaling, and the scintillation index peaks at the first branching distance (Mattheakis et al., 2017, Lin et al., 2021).

3.2 Strong Scattering and Rogue Events

In the strong-scattering regime ($\varepsilon \sim O(1)$), multiple strong kicks lead to branching coalescence, increased interference, and the emergence of isolated, high-magnitude peaks ("rogue waves"), with statistics exhibiting non-Rayleigh long tails (Mattheakis et al., 2017).

3.3 Nonlinear and Nonlocal Effects

With intense optical fields in nonlinear media (e.g. plasma with photo-ionization or Kerr-type media), the index fluctuations can be modified by the propagating wave, producing intensity-dependent branching. Moderately intense waves locally amplify disorder, shortening the caustic distance and enhancing branching, whereas ultra-intense waves (relativistic regime or strong nonlocality) can homogenize the medium and suppress branching (Jiang et al., 2021, Zhao et al., 12 Sep 2025).

In nonlocal Kerr-type media, the range of nonlocality $d$ screens local self-focusing: increased $d$ delays the appearance of branches and broadens filaments, ultimately recovering the linear branching regime as $d \to \infty$. The first branching distance interpolates between the local (accelerated branching) and strongly nonlocal (linear scaling) regimes: $$z_b(d) \approx \alpha\, \ell_c\, \varepsilon^{-2/3}\, \left[1 + \frac{ \sigma\,I_0\,\ell_c^2 }{w_0^2 + 8d } \right]^{-1/3}$$ (Zhao et al., 12 Sep 2025).

3.4 Incoherent Branched Flow

Even with incoherent or partially coherent illumination, branched flow persists, although the peak intensity fluctuations decrease and interference features are suppressed. The scintillation index and intensity-correlation functions can be quantitatively predicted by stochastic partial differential equations without free parameters; the maximum of the scintillation index marks the dynamic onset of branching (Garnier et al., 10 Feb 2025).

4. Experimental Realizations and Control of Branching

Branched flow of light has been observed in diverse optical systems:

• Weakly disordered soap-film waveguides, enabling quantitative comparison between theory and experiment via direct refractive-index mapping and caustic imaging (Lin et al., 2021).
• Random lens arrays and gradient-index (GRIN) media, producing 2D branching networks and rogue-wave events under strong scattering (Mattheakis et al., 2017).
• Anisotropic nematic liquid crystal slabs programmed via photoalignment, yielding deterministic, highly configurable branched flow with tunable directionality, scaling laws, and controlled channeling in chip-scale platforms (Chang et al., 20 Jan 2026).
• Tailored random media and multimode waveguides, where wavefront shaping and partial transmission-matrix knowledge select specific branches for robust, broadband transmission, with applications to imaging and communications (Brandstötter et al., 2019).

The ability to program branching potentials allows for reconfigurable guidance, optical channeling, and adaptation to environmental fluctuations. The statistical position and width of caustics remain governed by the universal scaling with respect to disorder amplitude and correlation length in all cases.

5. Connections to Other Systems, Stability, and Theoretical Frameworks

The universality of branched flow extends across physical systems:

• Electronic branched flow in two-dimensional electron gases, with identical caustic statistics and scaling laws; robust superwires arise in periodic lattice potentials when the transverse Lyapunov exponent vanishes, creating lossless, dynamically-confined channels (Daza et al., 2021).
• Ocean wave dynamics and freak events, where waves focus into branches and develop rogue events over random seafloor topography (Heller et al., 2019).
• Microwave cavity and acoustic propagation, exhibiting branched caustic networks and extreme-intensity statistics.

The stability of branched flow against strong, localized perturbations is provided by "piecewise classical stable paths" (PCSPs): quantum and wave interference is locally reconstructed along the most stable segments, allowing branch structure to recover after local damage or disorder, in contrast to classical chaos where individual rays diverge exponentially (Liu, 2014).

Stochastic PDE analysis, Fokker–Planck approaches, and time-sliced semiclassical approximations rigorously formalize these properties, and yield closed-form predictions for caustic statistics and intensity fluctuations (Lin et al., 2021, Garnier et al., 10 Feb 2025, Liu, 2014).

6. Applications, Control Protocols, and Future Directions

Branched flow provides unique opportunities for disorder-guided photonic circuits, robust energy delivery, imaging, and information transport:

• Adaptive optical interfaces that dynamically reconfigure the branched flow in response to environmental changes, leveraging programmable liquid-crystal media (Chang et al., 20 Jan 2026).
• Secure, multiplexed communications in multimode fibers or through scattering media, exploiting high-transmission channels along selected branches (Brandstötter et al., 2019).
• Scattering-resilient endoscopic imaging via dynamically formed bright caustics, enhancing depth penetration and image fidelity (Chang et al., 20 Jan 2026).
• Engineering nonlocality or nonlinear response in optical media to tune the appearance and suppression of branches, with possible mitigation of rogue intensity events (Zhao et al., 12 Sep 2025).
• Exploitation of PCSPs for robust transport and wavefront delivery through complex, disordered environments (Liu, 2014).

Emerging research directions involve studying the 3D structure and magnetic field effects in plasma branching, the role of partial coherence and time-varying media, and the application of these principles in device-scale photonic systems with fully programmable disorder (Chang et al., 20 Jan 2026, Jiang et al., 2021).