Geometric Dynamics of Signal Propagation

• Geometric dynamics of signal propagation is defined by the interplay of physical space, network topology, and internal degrees of freedom that shape phase shifts and transport behavior.
• It employs microlocal, spectral, and manifold-based analytical methods to quantify optical, periodic, and network effects across various systems.
• This framework offers practical insights for robust mode design, computational surrogate modeling, and optimizing deep learning architectures via geometric signal analysis.

The geometric dynamics of signal propagation addresses how geometry—of physical space, media, networks, or internal degrees of freedom—governs the laws, phases, and transport of signals across diverse settings. This unifies phenomena across electromagnetic propagation in curved geometries, spatio-temporal periodic media, network dynamics, vector beams, and even high-dimensional representation learning. The topic encompasses microlocal and spectral analysis, geometric phases, emergent transport regimes, and topological features. This entry surveys fundamental principles, mathematical formalisms, and domain-specific implications.

1. Geometric Phases and Angular Momentum Redirection

When signals, specifically optical fields, are guided along non-planar paths or manipulated via nontrivial geometric configurations, they accumulate geometric phases dependent on angular momentum degrees of freedom. For a vector beam with spin (helicity) σ and orbital angular momentum ℓ, adiabatic transport along a path of solid angle Ω on the momentum sphere endows the beam with a geometric phase

$\varphi_{\text{geo}} = -(\ell + \sigma)\,\Omega.$

The minus sign corresponds to the canonical Berry-phase result for polarization in cyclic adiabatic evolution (McWilliam et al., 2021). Scalar (ℓ) and spin (σ) contributions sum, such that rotations of both transverse intensity and polarization coincide—both rotate by Ω. Special eigenmodes with total quantum number $j=\ell+\sigma=0$ (e.g., radial and azimuthal beams) are invariant under geometric redirection, providing mode-robustness in fiber or mirror systems. These phase mechanisms underpin the photonic spin Hall effect, enable polarization/intensity shaping in vector microscopy, and allow for topologically robust transmission in communicated channels.

2. Geometry-Driven Dynamics in Complex Networks

In networked dynamical systems, the geometry—defined by node connectivity, weighted degree, and shortest paths—controls the spatio-temporal evolution of signal propagation. Propagation may be characterized by a single exponent θ derived from the local dynamic rules, partitioning dynamics into three universality classes (Hens et al., 2018):

• Distance-driven (θ=0): Propagation aligns with graph distance. All nodes exhibit similar intrinsic delays, and shells of equal $d_{ij}$ respond synchronously.
• Degree-driven (θ>0): Nodes with large degree (hubs) are bottlenecks, slowing signal spread. The leading bottleneck along minimal paths dominates arrival times.
• Composite (θ<0): High-degree nodes accelerate propagation. Hubs act as express lanes, and arrival times are reduced proportional to negative powers of degree.

The propagation metric $$\mathcal{L}_{ij} = \min_{\text{paths}}\sum_{p\neq j}{S_p}^\theta$$ defines an effective temporal geometry. This renders propagation predictable from network structure plus a dynamic fingerprint, despite nonlinearities and degree heterogeneity.

3. Geometric Band Structure and Critical Points in Periodic Media

Space–time periodic systems introduce geometric complexity in the frequency-momentum domain through band structure engineering (Salazar-Arrieta et al., 2021). For two-dimensional transmission lines with capacitive modulation, spectral properties are controlled by the space–time geometry of the lattice and the relative synchronization ($\hat\Omega$, $m_c$). The resulting eigenvalue problem in $(k_x, k_y, \omega)$ uncovers:

• Forbidden-ω (stop bands): No propagating modes for certain frequencies.
• Forbidden-k (momentum gaps): No real momenta for certain frequencies.
• Criticality and phase transition: As parameters cross critical values, the topology of allowed regions in the Brillouin zone changes via the formation or closure of stop bands.
• Diabolic (conical) points: At symmetry points (X, M), bands touch conically, organizing local dispersion and supporting topological Berry phases with winding number ±1.

These geometric features direct pulse propagation, generate slow-light conditions at saddle points, and control the onset of evanescently trapped vs. freely propagating states. Modulation mixes Floquet sidebands, introducing anisotropic and beam-steering effects that derive directly from the geometry of the dispersion manifold.

4. Microlocal, Semiclassical, and Manifold-Based Propagation

Propagation in generalized geometric settings—manifolds, curved spaces, or under affine transformations—necessitates microlocal and semiclassical analysis.

• Wave Propagator on Riemannian Manifolds: The global propagator is constructed as a single Fourier integral operator with phase determined by the geodesic flow (Levi–Civita phase function), yielding an invariant expansion with curvature-coupled subprincipal symbol terms (Capoferri et al., 2019).
• Tilted Hyperplane Propagation: For high-frequency scalar waves, restriction onto affine (tilted) hyperplanes is described by explicit semiclassical Fourier integral operators, whose canonical relations map input to output data along geometrically computed rays and encode amplitude corrections via Jacobians and Egorov-type theorems (Gioia et al., 18 Apr 2025).
• Green Function Transform: In 3D, the solution to the Helmholtz equation decomposes into homogeneous (on-shell, propagating) and inhomogeneous (off-shell, evanescent/near-field) contributions in $k$-space, providing a geometric interpretation of causal signal transport and evanescent field formation (Sheppard et al., 2014).

These methods expose how geometric optics (ray tracing), curvature, and spectral data cooperate to determine propagation amplitudes, phase accumulation, and the emergence of caustics or Maslov index effects.

5. Geometry-Based Surrogate Models in Complex Domains

In bounded or scattering domains with complex geometry, geometric decomposition of signal propagation provides computational tractable and interpretable models:

• Field decomposition: The solution is approximated as a sum of direct, reflected, and diffracted components, each corresponding to a particular geometric path (ray) and localized field (Pradovera et al., 2023).
• Automated component generation: Reflection and diffraction rules leverage domain geometry, edges, and vertices via method-of-images and diffraction coefficients, producing an efficient sum-of-functions representation for the spatiotemporal field.
• Error guarantees: Under high-frequency or geometric optics limits, approximation errors are controlled by the accuracy of local solutions (e.g., UTD wedge coefficients) and by the omission of negligible amplitude components (satisfying causality).

This approach allows for rapid evaluation compared to full finite element models, with the geometric structure dictating signal arrival patterns, shadowing, and interference.

6. Manifold and Grassmannian Geometry in Signal Processing

The internal geometry of signal representations—emanating from subspace structures or learning architectures—determines both detection and trainability:

• Grassmannian Models in MIMO: Signals are mapped to points on the Grassmann manifold $G(T,N)$, and domain shifts correspond to geodesic flows between subspaces (PCA bases). Integrating over this geodesic yields a geodesic flow kernel that encodes domain-invariant distances for efficient classification (G-SVM), robust to nonstationary channel conditions (Shelim et al., 2024).
• Transformer Particle Geometry: The evolution of $n$-token representation vectors through a deep transformer is equivalently described as a discrete dynamical system of $n$ particles subjected to geometry-driven update maps. The system exhibits order-chaos phase transitions controlled by initialization hyperparameters, captured by analytic Lyapunov exponents governing both forward signal geometry (token alignment) and backward gradient propagation (Cowsik et al., 2024). The edge-of-chaos, defined by vanishing of both angle and gradient exponents, coincides precisely with the regime of optimal trainability.

7. Geometry-Induced Couplings, Topology, and Spin-Orbit Effects

In confined or topologically structured systems, geometry not only shapes the phase but also induces nontrivial coupling between internal degrees of freedom:

• Spin-Orbit Coupling in Curved Trajectories: For electromagnetic waves in curved waveguides or confined to space curves, curvature and torsion couple spin and orbital angular momentum, inducing geometric (Berry) phase, as well as spin- and orbital-Hall effects and even curvature-induced helicity inversion (Lai et al., 2018).
• Topological Features and Band Engineering: In magnonic crystals (e.g., vortex-antivortex lattices), spatial arrangement and topological alternation create bi-sublattice structures, enhancing bandwidth and enabling ultrafast, field-tunable signal propagation (Kim et al., 2014).
• Polarization Transport and Real Principal Type: The polarization vector of waves in anistropic elastic media or at interfaces evolves under a transport equation reflecting both the principal symbol of the differential operator and subprincipal geometrical corrections, effecting polarization rotation, reflection, and mode conversion (Hansen, 2021).

Summary Table: Geometric Control Mechanisms in Signal Propagation

Context	Geometric Effect	Governing Quantity/Observable
Non-planar optical beam path	Spin+orbital Berry phase	$\varphi = -(\ell+\sigma)\Omega$
Space–time periodic media	Band structure topology	Appearance/closure of ω,k gaps, cones
Networked oscillators	Temporal distance metric	$\mathcal{L}_{ij}=\min_\text{paths}\sum S_p^\theta$
High-frequency manifold propagation	Curvature, caustics	Subprincipal symbol, Maslov index
MIMO/Grassmannian	Domain-invariant kernel	Geodesic flow on $G(T,N)$
Deep transformer dynamics	Order-chaos transition	Lyapunov exponents $\lambda_a$, $\lambda_g$

Geometric dynamics unifies diverse manifestations of signal behavior as consequences of underlying spatial, topological, and internal geometry. It provides not only interpretive power but also quantitative control of mode robustness, transport speed, phase response, and interdisciplinary design paradigms across physics, engineering, and machine learning.