Generalized Plane Waves

• Generalized plane waves are extensions of classical plane waves using complex, polynomial, or geometric modifications to address variable coefficients and non-Euclidean settings.
• They are applied in finite elasticity for inhomogeneous waves, in numerical methods as quasi-Trefftz functions, and in Helmholtz theory to capture both propagative and evanescent modes.
• Practical implementations include high-order interpolation strategies in PDE discretizations, leading to stable representations in challenging physical and geometrical regimes.

Searching arXiv for relevant papers on generalized plane waves and closely related usages. Generalized plane waves are extensions of the classical plane-wave ansatz that arise when constant coefficients, affine phases, Euclidean geometry, or purely propagative behavior are no longer adequate. Across the literature, the expression denotes several distinct but structurally related constructions: in finite elasticity it refers to inhomogeneous “longitudinal” waves with complex slowness bivectors; in numerical analysis it denotes quasi-Trefftz basis functions such as $e^{P(x)}$ or $A(x)e^{d\cdot(x-x_c)}$ tailored to variable-coefficient PDEs; in Helmholtz theory it includes propagative and evanescent waves with complex-valued directions; and in mathematical physics it appears in superspace, curved-space, and plane-wave spacetime settings (Destrade et al., 2013, Imbert-Gerard, 2014, Galante, 2023, Adán, 2020, Holland et al., 2024).

1. Terminological scope and basic forms

The common point of departure is the ordinary plane wave, which in constant-coefficient settings is represented by an affine phase. For the scalar Helmholtz equation in a constant medium of refractive index $\epsilon_0$, the standard local solution is $u(x)=e^{i\kappa d\cdot x}$, $|d|=1$. In elasticity, a homogeneous longitudinal wave in an undeformed body has displacement

$u(x,t)=\epsilon\,n\,\cos[k(n\cdot x-c t)],$

with $n\in \mathbb R^3$, $|n|=1$. Generalizations replace the real direction $d$ or $n$ by a complex direction, a complex slowness bivector, a polynomial phase, or a non-Euclidean analogue (Imbert-Gerard, 2014, Destrade et al., 2013, Imbert-Gerard et al., 2015).

Domain	Representative form	Defining feature
Finite elasticity	$A(x)e^{d\cdot(x-x_c)}$0	Complex slowness bivector; propagation and attenuation may differ
Variable-coefficient PDEs	$A(x)e^{d\cdot(x-x_c)}$1 or $A(x)e^{d\cdot(x-x_c)}$2	Local quasi-Trefftz cancellation to order $A(x)e^{d\cdot(x-x_c)}$3
Helmholtz in spherical domains	$A(x)e^{d\cdot(x-x_c)}$4, $A(x)e^{d\cdot(x-x_c)}$5, $A(x)e^{d\cdot(x-x_c)}$6	Propagative and evanescent directions treated in one framework
Superspace and geometric settings	Plane-wave decompositions over superspheres or pp-wave metrics	Extension of the plane-wave concept beyond ordinary Euclidean waves

In the numerical PDE literature, the decisive distinction is between exact Trefftz functions and generalized plane waves. Trefftz methods use exact local solutions of the governing PDE; GPWs are instead approximate Trefftz, or quasi-Trefftz, functions, constructed so that the PDE residual vanishes to a prescribed Taylor order near an element center. In elasticity, the decisive distinction is between homogeneous and inhomogeneous waves, and between classical longitudinal polarization and generalized longitudinal polarization. In Helmholtz approximation, the crucial distinction is between propagative plane waves and evanescent plane waves, the latter carrying high Fourier modes stably in regimes where propagative waves alone do not (Imbert-Gerard et al., 2019, Imbert-Gerard, 2020, Galante, 2023).

2. Generalized longitudinal plane waves in finite elasticity

In finitely deformed isotropic elastic materials, generalized plane waves were formulated for infinitesimal motions superposed on a primary homogeneous pure stretch

$A(x)e^{d\cdot(x-x_c)}$7

with $A(x)e^{d\cdot(x-x_c)}$8. The inhomogeneous longitudinal wave is written

$A(x)e^{d\cdot(x-x_c)}$9

where $\epsilon_0$0 is the slowness bivector, $\epsilon_0$1 gives the phase-normal, and $\epsilon_0$2 gives the attenuation-normal. The generalized longitudinal condition is $\epsilon_0$3: amplitude bivector and slowness bivector are parallel in $\epsilon_0$4, so particle paths are ellipses, or circles in the isotropic case, in the plane spanned by $\epsilon_0$5 and $\epsilon_0$6 (Destrade et al., 2013).

Linearization about the finite deformation yields the propagation condition

$\epsilon_0$7

with $\epsilon_0$8, $\epsilon_0$9, and $u(x)=e^{i\kappa d\cdot x}$0 expressed through the response coefficients $u(x)=e^{i\kappa d\cdot x}$1, $u(x)=e^{i\kappa d\cdot x}$2. For a longitudinal wave one sets $u(x)=e^{i\kappa d\cdot x}$3, where $u(x)=e^{i\kappa d\cdot x}$4 is the directional bivector. Decomposition along the principal axes of $u(x)=e^{i\kappa d\cdot x}$5 reduces the characteristic equations, for arbitrary $u(x)=e^{i\kappa d\cdot x}$6 and arbitrary positive $u(x)=e^{i\kappa d\cdot x}$7, to

$u(x)=e^{i\kappa d\cdot x}$8

These conditions both determine the complex slowness magnitude $u(x)=e^{i\kappa d\cdot x}$9 and restrict the constitutive response functions (Destrade et al., 2013).

This leads to a constitutive classification. Requiring that any directional bivector $|d|=1$0 be allowed forces

$|d|=1$1

hence

$|d|=1$2

with $|d|=1$3 constant, and the hyperelastic strain energy

$|d|=1$4

This is the Hadamard material, which allows inhomogeneous longitudinal waves of arbitrary elliptical or circular polarization in any direction. If one requires only isotropic bivectors $|d|=1$5, that is $|d|=1$6, so that only circular polarization is prescribed, the weaker conditions

$|d|=1$7

define the generalized Hadamard material, with stored energy

$|d|=1$8

where $|d|=1$9 and $u(x,t)=\epsilon\,n\,\cos[k(n\cdot x-c t)],$0 are arbitrary. The classical Hadamard model is the special case $u(x,t)=\epsilon\,n\,\cos[k(n\cdot x-c t)],$1, $u(x,t)=\epsilon\,n\,\cos[k(n\cdot x-c t)],$2 constant (Destrade et al., 2013).

A further geometric feature is the family of ellipsoids $u(x,t)=\epsilon\,n\,\cos[k(n\cdot x-c t)],$3. Each such ellipsoid has exactly two central circular sections. A circularly polarized longitudinal wave whose polarization circle lies in one of these planes has directional bivector $u(x,t)=\epsilon\,n\,\cos[k(n\cdot x-c t)],$4 satisfying

$u(x,t)=\epsilon\,n\,\cos[k(n\cdot x-c t)],$5

For principal stretches $u(x,t)=\epsilon\,n\,\cos[k(n\cdot x-c t)],$6, the four corresponding directional bivectors are given explicitly in terms of

$u(x,t)=\epsilon\,n\,\cos[k(n\cdot x-c t)],$7

This construction ties propagation and attenuation directions directly to the finite strain tensor $u(x,t)=\epsilon\,n\,\cos[k(n\cdot x-c t)],$8 (Destrade et al., 2013).

3. Quasi-Trefftz generalized plane waves for variable-coefficient PDEs

In numerical analysis, generalized plane waves were introduced for variable-coefficient wave problems where exact local Trefftz bases are unavailable. A general formulation considers a homogeneous PDE

$u(x,t)=\epsilon\,n\,\cos[k(n\cdot x-c t)],$9

on $n\in \mathbb R^3$0. A GPW of order $n\in \mathbb R^3$1 at $n\in \mathbb R^3$2 is a function

$n\in \mathbb R^3$3

with $n\in \mathbb R^3$4, chosen so that

$n\in \mathbb R^3$5

The resulting nonlinear system is underdetermined, and the construction proceeds through $n\in \mathbb R^3$6 triangular linear subsystems, one for each homogeneous layer, after fixing the coefficients with $n\in \mathbb R^3$7 freely. This “road-map” formulation makes explicit the layered structure of GPW construction and the conditions required for high-order interpolation (Imbert-Gerard et al., 2019).

For the scalar wave equation

$n\in \mathbb R^3$8

a local GPW centered at $n\in \mathbb R^3$9 takes the form $|n|=1$0, where $|n|=1$1 is a polynomial of degree $|n|=1$2 chosen so that

$|n|=1$3

The coefficients are determined by prescribing

$|n|=1$4

and setting $|n|=1$5 for $|n|=1$6, $|n|=1$7. Two normalizations are emphasized: a $|n|=1$8-normalization $|n|=1$9, which recovers exact plane waves when $d$0 is constant and negative, and a constant-normalization $d$1, which remains nonzero at cut-off (Imbert-Gerard, 2014).

For the variable-coefficient Helmholtz equation

$d$2

the GPW ansatz on an element $d$3 is again $d$4, with $d$5 of total degree $d$6, chosen so that

$d$7

The induction formula for the coefficients $d$8 determines all higher-order phase terms after imposing

$d$9

If $n$0 is constant on $n$1, all higher-order coefficients vanish and $n$2 reduces to an ordinary plane wave (Imbert-Gerard et al., 2015).

A later quasi-Trefftz formulation writes the local solution as $n$3, with $n$4, and imposes vanishing of the truncated Taylor polynomial of

$n$5

up to order $n$6. Expanding $n$7 into homogeneous layers yields linear problems

$n$8

which are solved by forward substitution. This connects phase-based GPWs with polynomial quasi-Trefftz spaces and gives a systematic route to finite linearly independent GPW families (Fontana et al., 13 Aug 2025).

4. Interpolation theory and numerical discretizations

The main analytical justification for GPWs is their local interpolation property. For second-order operators under the hypotheses stated in the road-map formulation, if $n$9 satisfies $A(x)e^{d\cdot(x-x_c)}$00, $A(x)e^{d\cdot(x-x_c)}$01, and $A(x)e^{d\cdot(x-x_c)}$02, then there exists

$A(x)e^{d\cdot(x-x_c)}$03

whose Taylor jets agree with those of $A(x)e^{d\cdot(x-x_c)}$04 up to order $A(x)e^{d\cdot(x-x_c)}$05 at the center, hence

$A(x)e^{d\cdot(x-x_c)}$06

In the scalar wave and Helmholtz versions, comparable local estimates are stated as

$A(x)e^{d\cdot(x-x_c)}$07

with $A(x)e^{d\cdot(x-x_c)}$08 and $A(x)e^{d\cdot(x-x_c)}$09 in those conventions (Imbert-Gerard et al., 2019, Imbert-Gerard, 2014, Imbert-Gerard et al., 2015).

These local estimates feed directly into discontinuous Galerkin and UWVF schemes. For variable-coefficient Helmholtz, the GPW-TDG method modifies the usual Trefftz sesquilinear form by adding the volumetric stabilization

$A(x)e^{d\cdot(x-x_c)}$10

because GPWs are only approximate, not exact, local solutions. The resulting formulation satisfies a coercivity statement of the form

$A(x)e^{d\cdot(x-x_c)}$11

and a Cea-type estimate

$A(x)e^{d\cdot(x-x_c)}$12

For smooth solutions and $A(x)e^{d\cdot(x-x_c)}$13 in the penalty $A(x)e^{d\cdot(x-x_c)}$14, the $A(x)e^{d\cdot(x-x_c)}$15-error satisfies

$A(x)e^{d\cdot(x-x_c)}$16

and in practice is observed to be nearly $A(x)e^{d\cdot(x-x_c)}$17 (Imbert-Gerard et al., 2015).

The reported numerical evidence is specific. For the Airy-wave test with $A(x)e^{d\cdot(x-x_c)}$18 and $A(x)e^{d\cdot(x-x_c)}$19, $A(x)e^{d\cdot(x-x_c)}$20 gives at best third-order $A(x)e^{d\cdot(x-x_c)}$21-convergence with $A(x)e^{d\cdot(x-x_c)}$22, whereas $A(x)e^{d\cdot(x-x_c)}$23 gives approximately $A(x)e^{d\cdot(x-x_c)}$24 up to $A(x)e^{d\cdot(x-x_c)}$25. For the Weber-wave example with $A(x)e^{d\cdot(x-x_c)}$26, $A(x)e^{d\cdot(x-x_c)}$27, $A(x)e^{d\cdot(x-x_c)}$28, plane waves stagnate while GPWs with $A(x)e^{d\cdot(x-x_c)}$29 yield high-order convergence and resolve both oscillatory and evanescent regions (Imbert-Gerard et al., 2015).

In the 2D plasma mode-conversion model, GPWs are embedded in a UWVF discretization for

$A(x)e^{d\cdot(x-x_c)}$30

with impedance boundary conditions. A theorem states that if $A(x)e^{d\cdot(x-x_c)}$31 and $A(x)e^{d\cdot(x-x_c)}$32, there is a unique $A(x)e^{d\cdot(x-x_c)}$33. The reported simulations use triangular elements with $A(x)e^{d\cdot(x-x_c)}$34, local GPW order $A(x)e^{d\cdot(x-x_c)}$35, and $A(x)e^{d\cdot(x-x_c)}$36 GPWs per element. The transmission coefficient $A(x)e^{d\cdot(x-x_c)}$37 is approximately $A(x)e^{d\cdot(x-x_c)}$38 at $A(x)e^{d\cdot(x-x_c)}$39 or $A(x)e^{d\cdot(x-x_c)}$40, has a pronounced maximum near the steepest-gradient direction $A(x)e^{d\cdot(x-x_c)}$41, and decreases as the transition-zone width increases (Imbert-Gérard, 2015).

$A(x)e^{d\cdot(x-x_c)}$42	Reported $A(x)e^{d\cdot(x-x_c)}$43
$A(x)e^{d\cdot(x-x_c)}$44	$A(x)e^{d\cdot(x-x_c)}$45, $A(x)e^{d\cdot(x-x_c)}$46
$A(x)e^{d\cdot(x-x_c)}$47	$A(x)e^{d\cdot(x-x_c)}$48, $A(x)e^{d\cdot(x-x_c)}$49

A recurrent misconception is that GPWs are “Trefftz functions with variable coefficients.” The precise statement is weaker: they are quasi-Trefftz functions, and that distinction is what necessitates residual control, interpolation analysis, and, in some formulations, volumetric stabilization (Imbert-Gerard, 2020, Imbert-Gerard et al., 2015).

5. Amplitude-based, phase-based, and evanescent extensions

Phase-based GPWs place higher-order terms in the exponent: $A(x)e^{d\cdot(x-x_c)}$50 Their strength is asymptotic high-order interpolation; their weakness is pre-asymptotic instability, because high-order terms in the exponent can make $A(x)e^{d\cdot(x-x_c)}$51 large when $A(x)e^{d\cdot(x-x_c)}$52. Amplitude-based GPWs alter the ansatz to

$A(x)e^{d\cdot(x-x_c)}$53

where $A(x)e^{d\cdot(x-x_c)}$54 is a polynomial of degree $A(x)e^{d\cdot(x-x_c)}$55, $A(x)e^{d\cdot(x-x_c)}$56 is chosen like a plane-wave direction, and the coefficients of $A(x)e^{d\cdot(x-x_c)}$57 are fixed so that

$A(x)e^{d\cdot(x-x_c)}$58

The local space is built from $A(x)e^{d\cdot(x-x_c)}$59 directions

$A(x)e^{d\cdot(x-x_c)}$60

For $A(x)e^{d\cdot(x-x_c)}$61 solving $A(x)e^{d\cdot(x-x_c)}$62, the span of these amplitude-based GPWs satisfies

$A(x)e^{d\cdot(x-x_c)}$63

as long as $A(x)e^{d\cdot(x-x_c)}$64. Numerically, they remain stable when $A(x)e^{d\cdot(x-x_c)}$65 and yield much smaller errors at moderate $A(x)e^{d\cdot(x-x_c)}$66 than phase-based GPWs (Imbert-Gerard, 2020).

A different extension enlarges the admissible directions rather than the local phase. In three-dimensional Helmholtz theory, generalized plane waves are solutions

$A(x)e^{d\cdot(x-x_c)}$67

Propagative waves correspond to real $A(x)e^{d\cdot(x-x_c)}$68; evanescent waves correspond to complex directions with

$A(x)e^{d\cdot(x-x_c)}$69

The 3D Jacobi–Anger identity extends to these complex directions, and any Helmholtz solution in a ball can be represented as a continuous superposition of evanescent plane waves. The associated Herglotz transform $A(x)e^{d\cdot(x-x_c)}$70 is boundedly invertible and diagonal on suitable bases: $A(x)e^{d\cdot(x-x_c)}$71 This yields stable bounded-coefficient representations that propagative waves alone cannot provide. If one approximates a single spherical mode by propagative plane waves only, the coefficients must grow super-exponentially in the mode index $A(x)e^{d\cdot(x-x_c)}$72; evanescent waves avoid that instability (Galante, 2023).

The numerical constructions follow this theory. Sampling generalized directions from a weighted parameter domain, building the basis $A(x)e^{d\cdot(x-x_c)}$73, and solving a regularized least-squares system produces stable approximations of spherical modes, fundamental solutions, and solutions on non-spherical geometries. The reported experiments show machine-precision accuracy for evanescent sets in regimes where propagative sets fail (Galante, 2023).

6. Superspace and symmetric space-time formulations

In superspace, plane-wave generalization is tied to monogenicity and CK-extension theory. On $A(x)e^{d\cdot(x-x_c)}$74, with super-Dirac operator

$A(x)e^{d\cdot(x-x_c)}$75

a super-function is monogenic if $A(x)e^{d\cdot(x-x_c)}$76. For the bi-axial Dirac operator $A(x)e^{d\cdot(x-x_c)}$77, the generalized Cauchy–Kovalevskaya extension theorem distinguishes two regimes. If the superdimension $A(x)e^{d\cdot(x-x_c)}$78, the extension is a single Bessel-type operator series: $A(x)e^{d\cdot(x-x_c)}$79 If $A(x)e^{d\cdot(x-x_c)}$80, novel structures appear: the extension depends on two initial functions and splits into Appell-type series. These operator series admit plane-wave decompositions over the supersphere through Pizzetti-type integrals, and the same framework yields decompositions of the super-Cauchy kernel into monogenic plane waves (Adán, 2020).

A different generalization arises from enforcing a dimensionless symmetry between space and time in the wave equation. After scaling with $A(x)e^{d\cdot(x-x_c)}$81 and $A(x)e^{d\cdot(x-x_c)}$82, one obtains

$A(x)e^{d\cdot(x-x_c)}$83

with no leftover constants. The usual ansatz

$A(x)e^{d\cdot(x-x_c)}$84

satisfies $A(x)e^{d\cdot(x-x_c)}$85, but the standing-wave decomposition is interpreted as only one geometrical projection of a quaternionic 4D phasor

$A(x)e^{d\cdot(x-x_c)}$86

Alternative projections produce additional constraints: $A(x)e^{d\cdot(x-x_c)}$87 In the reported acoustic cavity experiment, these spatial phase flows support recovery of the incidence angle by a formula involving only spatial integrals over the aperture and cavity, without temporal measurements (Blas et al., 2022).

These constructions suggest that “generalized plane wave” can denote not only a modified ansatz but also a different ambient algebra or symmetry principle. In superspace the extension introduces Bessel and Appell operator series; in the dimensionless wave-equation setting it introduces alternative hypercomplex projections of the same formal wave object (Adán, 2020, Blas et al., 2022).

7. Curved-space, gravitation, and non-Euclidean wave analogues

In the most general shift-symmetric Horndeski theory, an exact plane-wave solution exists with pp-wave metric

$A(x)e^{d\cdot(x-x_c)}$88

Because of shift symmetry, $A(x)e^{d\cdot(x-x_c)}$89 is arbitrary. The only nonzero Ricci component is

$A(x)e^{d\cdot(x-x_c)}$90

and the metric profile splits as

$A(x)e^{d\cdot(x-x_c)}$91

For $A(x)e^{d\cdot(x-x_c)}$92, the solution reduces exactly to the standard plane gravitational wave of general relativity; with Minkowski metric it reproduces known plane-wave solutions of $A(x)e^{d\cdot(x-x_c)}$93-essence and non-covariant Galileon models (Babichev, 2012).

In Lorentzian geometry, plane-wave spacetimes are defined as simply-connected smooth Lorentzian manifolds admitting a complete dilation group whose fixed-point set is a smooth embedded curve. That curve is a null geodesic. In Brinkmann coordinates the metric is

$A(x)e^{d\cdot(x-x_c)}$94

and abreast Jacobi fields satisfy

$A(x)e^{d\cdot(x-x_c)}$95

Writing $A(x)e^{d\cdot(x-x_c)}$96 with $A(x)e^{d\cdot(x-x_c)}$97 leads to the Sachs equation

$A(x)e^{d\cdot(x-x_c)}$98

Every plane wave is locally a Rosen universe, with metric

$A(x)e^{d\cdot(x-x_c)}$99

Rosen coordinate singularities occur where $\epsilon_0$00; these zeros are isolated and removable in the sense that the original Brinkmann metric extends smoothly across them (Holland et al., 2024).

A non-Euclidean analogue of quasi-plane waves appears for a spin-1 field in Lobachevsky space. In quasi-Cartesian coordinates with spatial metric

$\epsilon_0$01

one seeks solutions of the Duffin–Kemmer equation of the separated form

$\epsilon_0$02

An extended helicity operator

$\epsilon_0$03

commutes with the wave operator, permitting exact separation. The reduced equations lead to Bessel equations in

$\epsilon_0$04

so that the components are expressible in Bessel, hence confluent hypergeometric, functions. The effective one-dimensional equation contains the term $\epsilon_0$05, which acts as an exponential potential barrier, and the geometry therefore behaves as a medium with simple reflecting properties (Ovsiyuk et al., 2012).

Taken together, these developments show that generalized plane waves are not a single object but a family of extensions of the plane-wave paradigm. The unifying idea is the preservation of oscillatory, transport, or decomposition structure while modifying the classical ansatz to accommodate finite deformation, variable coefficients, complex directions, supersymmetry, or curved geometry.