EigenWave: Multi-Domain Spectral Modes

• EigenWave is a polysemous concept that denotes intrinsic spectral modes used in tropical geometry, numerical algorithms, and communication systems.
• In tropical geometry, the eigenwave defines a canonical cohomology class that encodes affine twisting, monodromy, and the construction of tropical intermediate Jacobians.
• In numerical and wave physics contexts, EigenWave underpins efficient eigensolvers and mode decompositions for elliptic operators and multidimensional channel modeling.

EigenWave, together with the lower-case forms eigenwave and eigenwaves, denotes several distinct constructions in contemporary mathematical and computational literature. In tropical geometry, the tropical eigenwave is a canonical cohomology class $\phi \in H^1(X;\,_1)$ attached to a compact tropical space and used to encode affine-geometric twisting, monodromy, and tropical intermediate Jacobians (Mikhalkin et al., 2013). In numerical analysis, EigenWave is an algorithm for computing eigenvalues and eigenvectors of elliptic boundary value problems by solving a time-dependent wave equation and filtering its solution in time (Appelo et al., 24 Jul 2025). In wave physics and communications, eigenwaves are spectral propagation modes associated with operator pencils for Maxwell systems, with bulk eigenmodes of homogenized media, or with multidimensional channel decompositions used as modulation carriers (Smirnov et al., 2012, Chipouline et al., 2012, Zou et al., 2022).

1. Terminological scope and principal usages

The term is therefore polysemous. It does not designate a single universally standardized object, but a family of domain-specific constructions that share a spectral interpretation.

Domain	Meaning	Representative papers
Tropical geometry	Canonical cohomology class and induced action on tropical (co)homology	(Mikhalkin et al., 2013, Mikami, 24 May 2026, Yamamoto, 2021)
Numerical PDEs	Time-filtered wave-equation eigensolver with optimal $O(N)$ scaling	(Appelo et al., 24 Jul 2025, Appelo et al., 30 Jun 2026)
Wave physics and communications	Propagation modes of waveguides, homogenized media, or multidimensional channels	(Smirnov et al., 2012, Chipouline et al., 2012, Zou et al., 2022)

Within these literatures, the common feature is not a shared formal definition but the use of intrinsic modes of a wave, geometric, or transport operator. This suggests that the name emphasizes mode selection or spectral diagonalization rather than a single cross-disciplinary theory.

2. Tropical eigenwave as a canonical cohomology class

In tropical geometry, the tropical eigenwave was introduced as a canonical class $\phi \in H^1(X;\,_1)$ attached to any compact tropical space $X$, with particular significance for tropical manifolds. A tropical manifold is described locally by models

$Y_\alpha=\Sigma_M\times \mathbb{T}^s,$

where $\Sigma_M$ is a Bergman fan of a loopless matroid $M$, and $\mathbb{T}^s$ is the tropical affine factor. The key geometric ingredient is the wave tangent space $W(x)$, obtained from the local affine structure after modding out divisorial directions. For a singular $1$-simplex $O(N)$0, the eigenwave is defined by

$O(N)$1

where $O(N)$2 is chosen as a nearby mobile point when $O(N)$3 has positive sedentarity and $O(N)$4 when $O(N)$5 is mobile. Because the choice at sedentary points is not unique, the construction is canonical only modulo coboundaries coming from divisorial directions, but the cocycle condition

$O(N)$6

shows that the resulting class is well defined in tropical cohomology (Mikhalkin et al., 2013).

The eigenwave is extracted from the integral-affine gluing data of tropical charts. Its role is to measure the displacement between endpoints of paths in affine coordinates, interpreted in the wave tangent bundle. The same paper introduces tropical wave and cowave groups

$O(N)$7

built from exterior powers of the wave tangent geometry, and states that “Other wave classes of similar type are responsible for deformations of the tropical structure.” The eigenwave is the degree-$O(N)$8 distinguished class among these related wave objects.

The class acts on tropical homology by cap product. For compact smooth tropical manifolds of dimension $O(N)$9, this action is part of the construction of tropical intermediate Jacobians. When $\phi \in H^1(X;\,_1)$0 and $\phi \in H^1(X;\,_1)$1, the group $\phi \in H^1(X;\,_1)$2 plays the central role; one sets $\phi \in H^1(X;\,_1)$3, forms $\phi \in H^1(X;\,_1)$4, and defines the tropical intermediate Jacobian as the principally polarized tropical torus

$\phi \in H^1(X;\,_1)$5

equipped with a symmetric bilinear form $\phi \in H^1(X;\,_1)$6 derived from the tropical intersection product

$\phi \in H^1(X;\,_1)$7

In this framework, $\phi \in H^1(X;\,_1)$8 is presented as the tropical analogue of the Hodge component $\phi \in H^1(X;\,_1)$9, while the pairing $X$0 plays the role of a polarization.

3. Monodromy, Gauss–Manin theory, and radiance obstruction

For realizable tropical limits of one-parameter degenerations of complex projective manifolds, the eigenwave records monodromy. The monodromy operator is written

$X$1

where $X$2 is the classical monodromy on the homology of a smooth fiber. The tropical eigenwave acts on tropical homology in a way that matches the action of $X$3 on the $X$4-page of the Steenbrink–Illusie spectral sequence. On the special tropical chain subcomplex called the konstruktor, the relation

$X$5

is the combinatorial analogue of the monodromy shift by one step. In the realizable smooth projective case, powers of the eigenwave satisfy

$X$6

as an isomorphism for $X$7, identified with the classical operator $X$8 (Mikhalkin et al., 2013).

Later work reformulated this picture in a tropical Steenbrink theory. A construction of strictly semistable reductions of tropical varieties leads to a tropical monodromy-weight spectral sequence and to a tropical Gauss–Manin connection. In that setting, the eigenwave action is recalled as a morphism

$X$9

and the main comparison theorem states that the residue of the tropical Gauss–Manin connection

$Y_\alpha=\Sigma_M\times \mathbb{T}^s,$0

coincides with the eigenwave action, up to sign. This identifies the eigenwave with the tropical analogue of monodromy or Gauss–Manin residue; the paper attributes the realizable case to Mikhalkin–Zharkov and the general case to Amini–Piquerez (Mikami, 24 May 2026).

A related development appears in the Gross–Siebert context of toric degenerations of Batyrev–Borisov Calabi–Yau complete intersections. There, a tropical contraction

$Y_\alpha=\Sigma_M\times \mathbb{T}^s,$1

maps the tropicalization onto the dual intersection complex, preserves tropical cohomology and homology, and sends the eigenwave class of the tropical variety to the radiance obstruction $Y_\alpha=\Sigma_M\times \mathbb{T}^s,$2 of the integral affine manifold with singularities. The eigenwave is thus identified with the affine displacement cocycle that records the failure of local affine data to glue globally (Yamamoto, 2021).

4. EigenWave as a time-filtered eigensolver for elliptic operators

In numerical analysis, EigenWave is an algorithm for computing eigenvalues and eigenvectors of large elliptic boundary value problems by transforming the eigenproblem into a time-dependent wave-equation problem. The governing continuous equation has the form

$Y_\alpha=\Sigma_M\times \mathbb{T}^s,$3

where $Y_\alpha=\Sigma_M\times \mathbb{T}^s,$4 is the elliptic operator whose eigenpairs are sought. If

$Y_\alpha=\Sigma_M\times \mathbb{T}^s,$5

then the wave dynamics oscillate at frequencies related to $Y_\alpha=\Sigma_M\times \mathbb{T}^s,$6, and EigenWave exploits this by choosing a target frequency $Y_\alpha=\Sigma_M\times \mathbb{T}^s,$7 and applying a temporal filter over one period or a fixed time window. Repeated application of the “solve wave equation + filter” map reinforces eigenmodes whose eigenvalues are near the target and suppresses others, producing a power-iteration-like mechanism centered on a chosen spectral window rather than on the largest magnitude eigenvalue (Appelo et al., 24 Jul 2025).

The method extends the WaveHoltz scheme and is designed to compute eigenpairs anywhere in the spectrum without inverting an indefinite shifted matrix of the form $Y_\alpha=\Sigma_M\times \mathbb{T}^s,$8. It can be embedded in a matrix-free Arnoldi algorithm to compute multiple eigenpairs near the target frequency. For efficiency, the wave equation is solved with implicit time-stepping; only about $Y_\alpha=\Sigma_M\times \mathbb{T}^s,$9 time-steps per-period are needed, essentially independent of mesh spacing. When the definite timestep equations are solved by multigrid, the cost scales linearly with the number of grid points $\Sigma_M$0, yielding an optimal $\Sigma_M$1 algorithm. Demonstrations are given for Laplacian eigenpairs in complex geometries, including overset grids, in two and three space dimensions and with second-order and fourth-order accurate approximations (Appelo et al., 24 Jul 2025).

The same eigenpair computation is used as an enabling subroutine in later work on accelerating WaveHoltz for the Helmholtz equation. For energy-conserving boundary conditions, WaveHoltz requires approximately $\Sigma_M$2 iterations in $\Sigma_M$3 dimensions at high frequency. Deflating eigenvectors whose eigenvalues lie nearest the driving frequency substantially reduces iteration counts, and the required eigenpairs can be computed efficiently with EigenWave. In two dimensions, when the number of deflation vectors grows quadratically with $\Sigma_M$4, the asymptotic convergence rate remains essentially constant. On structured, curvilinear, and overset grids, the deflated solver can break even against the undeflated solver after as few as two right-hand sides, once the precomputation cost of the eigenvectors is included (Appelo et al., 30 Jun 2026).

Another numerical usage of the name refers to an eigenfunction-expansion method for an extra-wide angle parabolic equation in acoustic wave propagation. There the square-root evolution operator is handled by diagonalizing a one-dimensional indefinite Schrödinger operator with a piecewise constant potential, estimating eigenvalues from approximate eigenfunctions, and refining them by the Secant Method. The solution is then reconstructed as

$\Sigma_M$5

and the cost of obtaining each eigenpair is independent of the grid size (Wright et al., 2022).

5. Eigenwaves in waveguides and homogenized electromagnetic media

In the mathematical theory of waveguides with dielectric inclusions, eigenwaves are the spectral objects associated with time-harmonic electromagnetic propagation in a shielded cylindrical guide. The Maxwell boundary-value problem for the longitudinal wave number $\Sigma_M$6 is reduced to an eigenvalue problem for a quartic nonselfadjoint operator pencil

$\Sigma_M$7

The longitudinal scalar fields $\Sigma_M$8 and $\Sigma_M$9 satisfy a variational problem in Sobolev spaces, and an eigenvector of that boundary problem is a pair

$M$0

for which the variational relation holds. In this setting, eigenwaves are the normal waves corresponding to eigenvalues of $M$1, while associated waves are generalized root vectors forming Jordan-chain-like structures (Smirnov et al., 2012).

The spectrum of these eigenwaves has a precise geometry. It lies in a vertical strip

$M$2

satisfies the symmetry

$M$3

and, outside a real interval of degeneration, consists of isolated eigenvalues of finite algebraic multiplicity with accumulation only at infinity. The paper concludes that the spectrum of normal waves is nonempty, forms a countable set of isolated points, is localized symmetrically in a strip, and contains at most finitely many real points (Smirnov et al., 2012).

A different electromagnetic use appears in homogenization theory for bulk materials. There the relevant regime is propagation far from interfaces, “where the eigenwave in the medium has been already formed and stabilized.” The problem is to derive macroscopic constitutive relations from averaged microscopic Maxwell equations and then determine the bulk dispersion relation of the equivalent medium. The paper distinguishes three admissible representations—Casimir (“C”), Landau–Lifshitz (“L&L”), and Anapole (“A”)—and stresses that effective parameters are not always uniquely defined because of Serdyukov–Fedorov transformations. What remains invariant is the dispersion relation governing the bulk eigenmodes. In the L&L form, for example, the macroscopic eigenmode condition is written

$M$4

The paper also emphasizes that several modes may exist at the same frequency, with excitation coefficients determined by boundary conditions that lie outside the scope of the bulk theory (Chipouline et al., 2012).

6. Channel-adapted eigenwaves in multidimensional wireless modulation

In wireless communications, the term denotes multidimensional orthogonal carriers derived from the channel itself. The framework is called Multidimensional Eigenwaves Multiplexing (MEM) and is based on a Higher Order Generalized Mercer’s Theorem decomposition of a multidimensional channel kernel

$M$5

Here $M$6 are transmit-side eigenfunctions, $M$7 are receive-side eigenfunctions, $M$8 are gains, and both families are orthonormal. These eigenfunctions are interpreted as eigenwaves: physically meaningful signaling waveforms over multiple degrees of freedom, including space, time-frequency, and delay-Doppler (Zou et al., 2022).

The motivation is explicitly non-stationary channels. OFDM mitigates time-domain interference caused by path difference, and OTFS mitigates frequency-domain interference due to velocity difference, but relative acceleration produces Inter-Doppler Interference in non-stationary channels. MEM therefore uses jointly orthogonal eigenwaves decomposed from the full multidimensional channel as subcarriers. Data are modulated by

$M$9

and after transmission the received field takes the separated form

$\mathbb{T}^s$0

Matched filtering against the receive eigenwaves yields

$\mathbb{T}^s$1

The intended consequence is orthogonality across each degree of freedom, including users or antennas in MU-MIMO, time-frequency, and delay-Doppler (Zou et al., 2022).

The framework assumes channel state information at both transmitter and receiver. The paper also mentions a zero-padded MEM variant in which symbols are not placed on the weakest eigenwaves. Its capacity discussion is given in the stationary setting, while the non-stationary discussion is cast primarily in diversity terms. Within that literature, eigenwaves are therefore not spectral objects of a geometric space or PDE operator but channel-adapted carriers that diagonalize the communication kernel itself (Zou et al., 2022).

Across these usages, the recurrent idea is the extraction of intrinsic modes from a structured operator, whether the operator encodes tropical affine gluing, an elliptic boundary value problem, a Maxwell waveguide, a homogenized medium, or a non-stationary communication channel. This suggests that the shared label “EigenWave” marks a common spectral intuition—mode decomposition, resonance targeting, or propagation along preferred directions—without erasing the substantial differences in formal definition and mathematical setting.