ε-PDO Formulation Framework

• ε-PDO formulation is an analytical framework that expands operator symbols in powers of ε to model wave propagation with slow modulations.
• It employs asymptotic expansions, including the Moyal product, to systematically reduce multidimensional PDEs to effective lower-dimensional models.
• Applications span acoustics, electromagnetics, and hydrodynamics, enabling rigorous treatment of inhomogeneity, dissipation, and boundary effects.

An ε-pseudodifferential operator (ε-PDO) formulation is an analytical framework used to asymptotically describe partial differential equations that exhibit slow spatial or temporal modulations compared to fast scales, often arising in semiclassical limits, wave propagation, and adiabatic perturbation problems. Central to its construction is the systematic expansion of operators and symbols in powers of a small parameter ε, representing the ratio of the slow to fast scales. This technique provides a bridge between full multidimensional wave equations and their effective reduced models, seamlessly incorporating inhomogeneity, dissipation, and operator-valued effects.

1. Mathematical Foundations and Scaling

Consider a general wave field $u(z,t,x,y)$ in a domain characterized by a fast vertical variable $z$ and slow horizontal variables $(x,y)$, typically encountered in shallow-water acoustics or electromagnetics with modulated media (Kaplun et al., 12 Nov 2025, Nittis et al., 2013). The governing PDE, after scaling slow variables via $x \to \epsilon x$, $y \to \epsilon y$, $t \to \epsilon c_{bot}^{-1} t$ ($\epsilon \ll 1$), leads naturally to an operator acting on a composite fast/slow variable structure: $$\hat{\mathcal{H}}(-i\epsilon \nabla_r, -i \partial_z, r, z) u(r, z) = 0$$ Here, $\nabla_r$ denotes derivatives with respect to the slow variables $r = (\tau, x, y)$ and $\tau = \epsilon c_{bot} t$. This scaling isolates $\epsilon$ as the controlling parameter for horizontal inhomogeneity or adiabatic variations.

The operator $\hat{\mathcal{H}}$ has a full symbol

$$\mathcal{H}(r, p, -i \partial_z, z) = \left[ -(\partial_z)^2 + \nu^2(z, r)\, p_\tau^2 \right] + (p_\tau^2 - p_x^2 - p_y^2)$$

where $\nu^2(z, r)$ encodes the stratification, and $p$ are phase-space momenta conjugate to slow variables.

2. ε-Pseudodifferential Operator Calculus

In the Kohn–Nirenberg convention, an ε-PDO with symbol $M(r, p)$ acts as: $$\hat{M}[f](r) = \frac{1}{(2\pi \epsilon)^d} \iint e^{\frac{i}{\epsilon}(r - r') \cdot p} M(r, p) f(r')\, dr' dp$$ This construction extends classical PDO theory to the semiclassical regime, permitting operator-valued symbols, i.e., $M$ may itself be a differential operator in fast variables. An asymptotic expansion in $\epsilon$ of the symbol yields: $$L(r, p; \epsilon) = L^0(r, p) + \epsilon L^1(r, p) + O(\epsilon^2)$$ which forms the basis for systematic perturbative treatments.

For products and compositions, the Moyal product gives

$$a \# b = ab + \frac{i\epsilon}{2} \{a, b\} - \frac{\epsilon^2}{8} \sum_{j, k} \partial_{k_j} \partial_{k_k} a \, \partial_{r_j} \partial_{r_k} b + R_2$$

where $\{a, b\}$ is the Poisson bracket. Norm estimates and Calderón–Vaillancourt theorems guarantee operator boundedness under suitable symbol regularity (Nittis et al., 2013).

3. Operator Separation of Variables and WKB Ansatz

A key analytical advance is the application of operator separation and WKB ansatz for single-mode reduction: $$u(z, r; \epsilon) = \hat{\psi}\left(z, -i\epsilon\nabla_r, r, \epsilon\right) \left[ A(r; \epsilon) e^{i \phi(r)/\epsilon} \right]$$ where $\hat{\psi}$ is an operator-valued ε-PDO and $A(r;\epsilon)$, $\phi(r)$ capture slow amplitude and phase. Expanding $\psi$ and $A$ in powers of $\epsilon$ yields amplitude and eikonal equations. Acting on such ansatz, an ε-PDO produces: $$e^{i \phi/\epsilon} \hat{M}[-i\epsilon \nabla_r, r](A e^{i \phi/\epsilon}) = e^{i\phi/\epsilon} \left( M_0[A] + \epsilon [M_1[A] + M_0[A^1]] + O(\epsilon^2) \right)$$ with leading term $M_0$ evaluated at $p = \nabla_r \phi$.

4. Amplitude and Phase Evolution: Eikonal and Transport Equations

At $O(\epsilon^0)$, the eikonal (dispersion) equation emerges: $$L^0(r, \nabla \phi) A^0 = 0 \implies H(\nabla \phi(r), r) = 0$$ where $H(p, r)$ is the effective Hamiltonian, involving vertical-mode eigenvalues $\lambda(p_\tau, r)$.

At $O(\epsilon^1)$, the transport equation for amplitude is: $$\nabla_r \cdot [A^2 k(p, r)] = 2A^2 \left( \nabla_x \phi, \frac{\langle \nabla_x \psi^0, \xi^0 \rangle}{\langle \psi^0, \xi^0 \rangle} \right)$$ and $k(p, r) = \nabla_p H(p, r)$. Additional terms from non-self-adjoint operators (bottom leakage, complex modes) modify amplitude evolution and account for energy dissipation.

5. Hamiltonian Ray Formalism and Characteristics

Eikonal equations naturally lead to a Hamiltonian ray framework. The phase-space flow

$\frac{df}{d\sigma} = J \nabla_f H(f)$

captures the propagation of rays, where $J$ is symplectic. Along rays, the phase accumulates as

$$\phi(\sigma) - \phi(0) = \int_0^\sigma p(\sigma') \cdot \dot{r}(\sigma') d\sigma'$$

Amplitude evolution is

$$A(r(\sigma)) = A(r(0)) \sqrt{ \frac{ \det[ \partial_\mu r(0) ] }{ \det[ \partial_\mu r(\sigma) ] } } \exp\left\{ \int_0^\sigma g(\sigma') d\sigma' \right\}$$

with leakage prefactor $g$ and Jacobi matrix $\partial_\mu r$.

6. Boundary Conditions and Non-Self-Adjoint Operators

Boundary conditions at interfaces, especially the seabed ($z = h(r)$), fundamentally determine the operator structure:

• $\alpha = 0$ (Neumann): self-adjoint, real eigenvalues; ideal waveguide.
• $\alpha = 1$: matching densities, full transmission, self-adjoint.
• $0 < \alpha < 1$: partial reflection, non-self-adjoint, complex eigenvalues $\lambda$; biorthogonal modal pairs $(\psi^0, \xi^0)$; amplitude equation right-hand side encodes leakage.

This formalism generalizes mode matching and dissipative effects, handling exponential leakage via the complex spectrum (Kaplun et al., 12 Nov 2025).

7. Analytical and Numerical Application

For constant stratification $\nu^2(z) = \nu_0^2$ and simple geometry, the mode equations reduce to transcendental spectral conditions: $$\cot(\sqrt{\nu_0^2 p_\tau^2 - k^2}) = -\frac{\alpha k}{\sqrt{\nu_0^2 p_\tau^2 - k^2}}$$ yielding explicit expressions for vertical wavenumbers and mode leakage rates. Hamiltonian ray-tracing for various $\alpha$ demonstrates amplitude decay and front distortion, with numerical experiments quantifying energy leakage near critical depths.

8. Relationships to Other Fields and Generalizations

The ε-PDO strategy is closely connected to space-time canonical operator theory, Maslov–Babich asymptotics, and adiabatic dynamical reduction schemes. In electromagnetics, analogous ε-PDO constructions enable derivation of effective Maxwell operators and band-structure perturbation theory (Nittis et al., 2013). For hydrodynamics, EPDiff-type ε-PDOs encode regularized, elliptic metrics on diffeomorphism groups, yielding globally well-posed geodesics and in the singular limit, classical Euler flow (Mumford et al., 2012). These methods unify semiclassical, ray-based, and operator-theoretic wave modeling with a rigorous $\epsilon$-expansion formalism.

A plausible implication is that ε-PDO models provide systematic means to rigorously derive reduced-dimensional, energy-dissipative, and inhomogeneous wave equations from first principles, with broad applicability across acoustics, optics, and fluid mechanics.