Generalized Ray-Based Formulation

• Generalized ray-based formulation is a framework that extends classical ray theory using variational principles and Hamiltonian methods to model energy, field, and information transport.
• It incorporates metric and transformation approaches to describe rays as geodesics on curved manifolds, enabling applications in imaging, cloaking, and wave-optical rendering.
• Computational methods, including finite-element solvers, Hadamard integrators, and GPU-based ray marching, ensure efficient simulation in high-frequency, anisotropic, and quantum settings.

A generalized ray-based formulation refers broadly to the mathematical and computational frameworks that extend classical ray theory to describe the transport, interaction, and reconstruction of energy, fields, or information along families of spatial curves or "rays," with generalizations that incorporate anisotropy, wavefront curvature, wave-optical effects, or nontrivial transformation geometries. These frameworks underpin diverse areas such as high-frequency wave propagation, inverse imaging, wave-optical rendering, transformation design, and even particle statistics, providing both variational and geometrical constructs to relate microscopic physical laws to macroscopic observables.

1. Variational Principles and Ray Equations in General Media

Generalized ray-based theories originate from variational formulations in which rays are extremal curves of an action functional (typically traveltime or optical path length, subject to Fermat's principle) in possibly anisotropic and heterogeneous media. For propagating waves in general anisotropic media, the arclength-related Lagrangian $L(x, r) = 1/V_\mathrm{ray}(x, r)$, where $V_\mathrm{ray}(x, r)$ is the group velocity, yields Euler–Lagrange equations for stationary ray paths: $$\frac{\mathrm{d}}{\mathrm{d}s}\left(\frac{\partial L}{\partial r}\right) = \frac{\partial L}{\partial x}$$ which are equivalent to Hamiltonian ray equations under Legendre transformation $H(x, p) = p \cdot r - L(x, r)$, with $p$ the slowness vector and the Christoffel equation imposing physical constraints (Koren et al., 2020, Koren et al., 2020). The formalism holds in both isotropic and general anisotropic elasticity, provided the stiffness tensor $C_{ijkl}(x)$ is sufficiently smooth and the medium is locally homogeneous.

The framework extends to paraxial (dynamic) ray tracing for field amplitudes and phases via the Jacobi (geodesic deviation) equation, which governs normal deviations (paraxial shifts) from a central ray: $$\frac{\mathrm{d}}{\mathrm{d}s}\left(L_{rx}\cdot u + L_{rr}\cdot \dot{u}\right) = L_{xx}\cdot u + L_{xr}\cdot \dot{u}$$ where $u(s)$ is the paraxial deviation and $L_{ab}$ are second derivatives of $L$ (Ravve et al., 2020). Solutions yield ray-spreading, caustics, and Green’s function amplitude corrections.

2. Metric and Transformation Approaches

The ray-bending properties of structured or designed media can be described via spatial metrics, with rays as geodesics on a Riemannian (possibly pseudo-Riemannian or Finslerian) manifold. The generalized transformation design formalism identifies the effective metric $g_{ij}(x)$ and its inverse $g^{ij}(x)$ with the medium’s anisotropic squared-speed tensor $C^{ij}(x)$, making rays solutions to: $g_{ij}(x)\, \dot{x}^i\, \dot{x}^j = 1$ and, in eikonal/Hamiltonian formulation,

$$H(x, p) = \tfrac{1}{2} (g^{ij}(x) p_i p_j - \omega^2) = 0$$

This viewpoint unifies ray, wave, and diffusive processes, leveraging coordinate transformations for device and cloak design. The geometric structure allows derivation of generalized Snell’s laws, geodesic equations, and diffusion analogs (Kinsler et al., 2015).

3. Generalizations Beyond Classical Rays: Ray Transforms, X-ray Operators, and Inversion

Classical ray-based forward models are generalized to nontrivial integration curves or to ray transforms with weights for imaging and inversion. Examples include:

• Broken-ray (V-line) Radon transform: Integration is along piecewise-linear (V-shaped) curves with corner variables corresponding to scattering events. Inversions are explicit, yielding both absorption and scattering maps, and generalizing classical filtered backprojection (Florescu et al., 2010).
• Geodesic X-ray transforms with connection: On Riemannian manifolds, attenuation or matrix-valued weights arising from $GL(n, \mathbb{C})$ connections yield forward maps as solutions to transport equations along geodesics. The inversion is provided by Fredholm equations with precise injectivity and range characterizations, even in nontrivial curvature or non-unitary settings (Monard et al., 2016).
• Ray transforms on non-straight curves: Complex-analytic generalizations allow inversion on data from curved or conformally related ray families (Hoell et al., 2010).

Such generalizations extend the applicability of ray-based imaging to complex geometries, anisotropy, and non-scalar attenuation.

4. Wave-Optical Extensions: Ray Theory of Waves and Generalized Rays

Moving beyond geometric optics, generalized ray-based formalisms such as the Ray Theory of Waves (RTW) incorporate local wavefront curvature (WFC) as a ray property. Each ray carries a second fundamental form tensor $Q$ encoding principal curvatures. The full wavefront equation, derived via differential geometry and phase continuity at interfaces, relates incident and refracted curvatures: $$k_r P_r Q_r P_r^\mathrm{T} = k_i P_i Q_i P_i^\mathrm{T} + (n\cdot(k_r - k_i))C$$ Amplitude, phase, and local divergence are transported along rays, with caustic singularities regularized by replacing classic polynomial field expansions with numerically computed VCRM fields in uniform approximations (Ren et al., 2024).

In wave-optics rendering, a generalized ray is a minimum-uncertainty Gaussian-correlated wavepacket, realized as an eigenstate of the photodetection process and characterized by a phase-space Wigner distribution: $$\hat{g}_\beta(\mathbf{r}, \mathbf{k}|\mathbf{r}_0, \mathbf{k}_0) = (2\pi)^{-3} \exp\left(-\frac{\|\mathbf{r}-\mathbf{r}_0\|^2}{\beta^2}\right) \exp\left(-\beta^2\|\mathbf{k}-\mathbf{k}_0\|^2\right)$$ Transport, scattering, and detection are described by linear, weakly-local operators, admitting backward path integration (sensor-to-source) and simulation of arbitrary coherence and spectral properties. Classical MC path tracing strategies transfer directly to this setting, enabling scalable and physically accurate computations (Steinberg et al., 2023).

5. Computational Implementations: Finite-Element, Hadamard, and Efficient Ray Marching

Variational ray frameworks are discretized using advanced numerical methods:

• Finite-element solvers for two-point ray boundary problems in anisotropic 3D media, employing Hermite interpolation, weak (Galerkin) forms, and target function minimization to robustly solve for stationary rays, including minimum and saddle-point (caustic) solutions (Koren et al., 2020).
• Hadamard integrator for time-dependent wave equations using a Gelfand-Shilov ansatz for the Green’s function, Lagrangian ray tracing in geodesic polar coordinates, and low-rank Chebyshev interpolants for wavefront and coefficient storage. This allows frequency-independent, dispersion-free, and caustic-robust time marching (Wei et al., 2023).
• Branchless generalized Joseph projector (GJP): A highly performant GPU-friendly algorithm for volume ray casting, where ray marching and interpolation are designed to maximize computational efficiency, coherence, and minimize memory bandwidth (Graetz, 2016).

6. Ray-Based Formulations in Structured Light and Quantum Theory

• Ray-optical Poincaré sphere for structured beams: Gaussian beams and their generalizations (HG, LG, HLG) are constructed as orbits on a Poincaré sphere, with the mapping to the physical transverse plane dictating the elliptic caustics and enabling direct calculation of modal structures, caustics, and geometric phases (Gouy, Pancharatnam–Berry). Field reconstruction proceeds via double integration over Poincaré parameter space, effectively replacing explicit diffraction integrals (Alonso et al., 2016).
• Ray spaces in parastatistics: In quantum theory, generalized ray spaces encode higher-dimensional irreducible representations for paraparticle statistics (order $p \ge 2$), via bracketed states labeled by Young diagrams. These constructions formalize the external sector of S-matrix elements in paraparticle theory and provide explicit formulas for creation-operator-based orthonormal basis states for parabosons, parafermions, and composite para-families (Nelson, 2019).

7. Applications and Extensions

The generalized ray-based paradigm finds widespread application and ongoing extension:

• Inverse imaging and tomography: Explicit inversion for systems with curved, broken, or attenuated rays.
• Optics and transformation design: Engineering of cloaks, lenses, and diffusive devices via metric selection.
• Structured light and beam shaping: Unified geometrical optics and modal design.
• Efficient large-scale computation: High-performance algorithms for simulation and field reconstruction in graphics, seismology, or non-destructive testing.
• Quantum information and parastatistics: Construction of ray spaces and external sectors for exotic particle theories.

These developments collectively demonstrate the depth and versatility of generalized ray-based formulations for analysis, simulation, and inversion in complex, high-frequency, or nonlocal physical systems.

---

Key References:

• Eigenray variational and dynamic formulation: (Koren et al., 2020, Koren et al., 2020, Ravve et al., 2020, Koren et al., 2020)
• Transformation optics/diffusion: (Kinsler et al., 2015)
• Advanced imaging and X-ray transforms: (Florescu et al., 2010, Monard et al., 2016)
• Ray-theory of waves and wavefront curvature: (Ren et al., 2024)
• Structured beams and Poincaré sphere: (Alonso et al., 2016)
• Wave-optical generalized rays: (Steinberg et al., 2023)
• Hadamard integrator and ray-based time propagation: (Wei et al., 2023)
• Ray spaces for paraparticles: (Nelson, 2019)
• Efficient ray casting: (Graetz, 2016)
• Paraxial equations and metric representations: (Bergman, 2015)
• Generalized Lau condition: (Yan et al., 2019)