A mathematical study has overturned a long-standing expectation about gravitational waves: a spacetime can remain geodesically complete even when its wave profile is far more complicated than the simple quadratic form traditionally associated with plane waves. The result challenges the Ehlers–Kundt conjecture, proposed in 1962 as a possible defining principle for exact gravitational waves. According to the new analysis, the conjecture fails when the wave is concentrated into an idealized, infinitely sharp impulse. In that setting, even smooth profile functions with arbitrary spatial complexity can produce spacetimes through which freely falling objects continue along complete paths. The work, by Moriz L. Frauenberger, James D. E. Grant and Roland Steinbauer, places mathematical regularity at the center of one of general relativity’s most persistent classification problems.

The Ehlers–Kundt conjecture concerns a family of exact solutions to Einstein’s equations known as pp-waves, short for “plane-fronted waves with parallel rays.” These spacetimes can be written using coordinates (u), (v), (x) and (y), with a metric of the form (ds^2=2du,dv+dx^2+dy^2+H(x,y,u)du^2). The coordinates (x) and (y) describe directions transverse to the wave, while (u) acts as a null wave coordinate and (v) completes the pair of null directions. The function (H) encodes the gravitational-wave profile. In vacuum, the Ricci tensor vanishes precisely when (H) is harmonic in the transverse spatial coordinates, meaning that its two-dimensional Laplacian is zero. This condition removes ordinary matter sources but still permits a rich variety of gravitational fields.

A particularly important subclass is the plane wave, in which the profile is quadratic in the transverse coordinates: (H=h_{ij}(u)x^ix^j). The matrix (h_{ij}) must be trace-free for the spacetime to satisfy the vacuum Einstein equations. These solutions are mathematically tractable because their geodesic equations reduce to linear second-order differential equations, ensuring that freely falling particles can be followed for an unlimited range of their affine parameter. Geodesic completeness is the relativistic analogue of saying that no particle trajectory abruptly ends because of a hidden singularity. Ehlers and Kundt conjectured that this property might uniquely identify plane waves among Ricci-flat pp-waves: if the spacetime is complete, they argued, its profile should be a polynomial of degree no greater than two.

For decades, the conjecture resisted a full proof. A major advance came in 2020, when José Luis Flores and Miguel Sánchez established it under an important restriction: the profile, or equivalently the associated Newtonian-like potential, must be bounded by a polynomial in the transverse coordinates. The connection to Newtonian dynamics is central. Along a geodesic that crosses the wave, the transverse coordinates behave like the position of a particle moving under a potential (V=-H/2). If the potential grows too aggressively in some directions, trajectories can escape to infinity in finite affine time, making the spacetime incomplete. Harmonic functions with polynomial growth are themselves polynomials, which helps explain why the polynomially bounded case leads back to quadratic plane waves.

The new study focuses on a case that lies outside those smooth assumptions: impulsive gravitational waves. Instead of allowing the profile to vary smoothly with the null coordinate (u), the authors represent it as (H(x,u)=f(x)delta(u)), where (f(x)) is a smooth function of the transverse coordinates and (delta(u)) is the Dirac delta distribution. Physically, this idealizes a pulse whose duration shrinks toward zero while its strength increases so that its integrated effect remains finite. Impulsive waves have long been used to model short, violent gravitational disturbances, including shockwaves associated with ultrarelativistic limits of massive spacetimes. They also appear in theoretical work on gravitational-wave memory, quantum scattering and the interaction of quantum systems with curved spacetime.

The delta function makes the geometry singular in a precise mathematical sense. The metric is not an ordinary smooth tensor field at the wavefront (u=0), and the conventional geodesic equations contain products and derivatives of distributions that cannot simply be manipulated as if they were ordinary functions. To avoid that problem, the researchers first replace the delta function with a family of smooth functions (delta_varepsilon(u)). Each member of this “strict delta net” is concentrated inside a narrow interval around (u=0), has an integral approaching one as (varepsilon) tends to zero and obeys a uniform bound on its absolute integral. The resulting regularized metric is smooth for every nonzero (varepsilon), allowing standard differential geometry and ordinary differential-equation methods to be used before taking the impulsive limit.

The geodesic equations reveal why the impulse does not destroy completeness. The null coordinate evolves linearly with the affine parameter, so it can be used as the parameter for any geodesic that crosses the wave. Away from the narrow regularization zone, the transverse motion is simply a geodesic of the underlying Riemannian space (N). If (N) is complete, those background trajectories already exist for all parameter values. Inside the wave zone, the transverse equation acquires a force proportional to (delta_varepsilon(u)nabla f). Although this force can become very large as the pulse narrows, its integrated strength remains controlled by the uniform (L^1) bound of the delta net. A fixed-point argument then provides a solution across the wave zone for sufficiently small (varepsilon), with the required interval of existence independent of the regularization width.

The longitudinal coordinate (v) is easier to recover once the transverse trajectory is known. Its equation is linear and can be solved by integration, even though it includes a term proportional to the derivative (dot{delta}_varepsilon). The authors show that for every set of initial data, there is a threshold (varepsilon_0) such that all narrower regularizations yield a geodesic defined on the entire real line. This is a deliberately data-dependent statement: the proof does not claim that one fixed regularized metric is complete for every possible initial condition with a common (varepsilon_0). Instead, it establishes global solvability for each chosen geodesic family as the impulse becomes sufficiently sharp.

To give the singular limit a rigorous meaning, the researchers use Colombeau’s nonlinear theory of generalized functions. This framework represents a singular object by a whole family of smooth functions and identifies families that differ by quantities vanishing faster than every power of the regularization parameter. Unlike ordinary distribution theory, it permits nonlinear operations such as multiplying generalized functions and constructing generalized tensor fields, connections and curvature. A generalized geodesic is likewise represented by a net of smooth curves. The team verifies that the regularized geodesics are compactly bounded and moderate, meaning their derivatives grow no faster than allowed powers of (1/varepsilon). They also prove uniqueness: if the equations and initial data are perturbed by negligible quantities, the resulting curves differ from the original family by negligible quantities as well.

The conclusion is strikingly broad. For impulsive (N)-fronted waves with parallel rays, any smooth transverse profile (f) produces a geodesically complete generalized spacetime when the Riemannian component (N) is complete. The profile need not be quadratic, polynomial or even polynomially bounded. Completeness, which acts as a powerful rigidity condition for sufficiently regular pp-waves, loses that role when the wave is distributional in the null direction. The result does not imply that every physically realistic gravitational wave has arbitrary structure without consequences; an ideal Dirac impulse is a limiting model, and the regularity of spacetime remains essential to the interpretation. But it does show that the original conjecture cannot be extended unchanged into the impulsive regime. In general relativity, a wave’s sharpness may matter as much as its shape.