Nonlinear Wave Interactions

Scrollable formula. Use Left and Right arrow keys to reveal hidden content.

Scrollable table. Use Left and Right arrow keys to reveal additional columns.

Scrollable figure. Use Left and Right arrow keys to inspect hidden content.

Nonlinearity can create a probe at a point no physical source can reach directly. Inject several conormal waves from the observation region, arrange their null hypersurfaces to intersect transversally in the unknown region, and differentiate the response with respect to their amplitudes. A sufficiently high mixed derivative contains a new singularity whose future wavefront is the light cone of the interaction point. Detecting that cone recovers passive light-observation data from active experiments.

Required background. Analysis for Mathematical Relativity supplies conormal distributions and principal symbols; PDEs and Green Functions for Gravitation supplies propagation of singularities; and Lorentzian metrics and signature supplies null bicharacteristics.

Helpful background. The Einstein equation supplies the quasilinear interaction problem, and Causality and Global Structure supplies the accessible diamond.

Multilinearizing the response

Consider in four spacetime dimensions

\Box_gu+a(x)u^2=f, \qquad f=\sum_{j=1}^4\epsilon_j f_j,

with retarded solutions and sources $f_j$ supported in the known observation region $V$ . Expand about the zero solution; the following expression displays selected Taylor terms, with the omitted terms including all other cubic and quartic monomials:

u(\epsilon)= \sum_j\epsilon_jv_j +\frac12\sum_{j,k}\epsilon_j\epsilon_kv_{jk} +\cdots +\epsilon_1\epsilon_2\epsilon_3\epsilon_4v_{1234} +O(\lvert\epsilon\rvert^5).

Here $v_{jk}=\partial_{\epsilon_j}\partial_{\epsilon_k}u\vert_0$ ; the factor $1/2$ is the multivariable Taylor normalization. The first variations solve $\Box_gv_j=f_j$ . Writing $Q_g$ for the retarded inverse, distinct second variations satisfy $v_{jk}=-2Q_g(av_jv_k)$ . Repeating the differentiation shows that $v_{1234}$ is a sum over interaction trees: products of first variations and waves produced by their earlier pair or triple interactions, followed by one last application of $Q_g$ .

For distinct indices this can be written without hidden combinatorial factors:

v_{ijk}=-2Q_g\!\left[a(v_i v_{jk}+v_jv_{ik}+v_kv_{ij})\right],

v_{1234}=-2Q_g\!\left[a\left( \sum_{i=1}^4v_i v_{\{1,2,3,4\}\setminus\{i\}} +v_{12}v_{34}+v_{13}v_{24}+v_{14}v_{23}\right)\right].

The multi-index on each term in the sum denotes the other three source derivatives. The factor two comes from the two ways to distribute a chosen pair of complementary derivatives between the factors of $u^2$ . This also shows why the fourth-order response is not merely $Q_g(av_1v_2v_3v_4)$ : a quadratic equation produces successive interactions.

The mixed derivative

\partial_{\epsilon_1}\partial_{\epsilon_2} \partial_{\epsilon_3}\partial_{\epsilon_4} L(f)\big\vert_{\epsilon=0} =v_{1234}\vert_V

is determined by the nonlinear source-to-solution map. Subtracting the lower derivatives isolates information unavailable to the linearized map.

Four transversal waves

Choose $f_j$ so that $v_j$ is conormal to a narrow null hypersurface $K_j$ : its singular covectors are normal to that surface, as for a wavefront across which transverse derivatives become singular. The waves must have the regularity and wavefront transversality needed to define their products; arbitrary products of distributions are not allowed. Before caustics, propagation of singularities carries its wavefront along the null bicharacteristic flow. Arrange the four $K_j$ to meet transversally at one point $q$ in the unknown region and to have no earlier common intersection.

At $q$ , multiplication combines their conormal covectors. For generic source symbols and geometry, the relevant principal symbol in $v_{1234}$ is nonzero. The retarded inverse then propagates that singularity along $\partial J^+(q)$ . To receivers in $V$ , it looks like radiation emitted by an artificial point source at $q$ even though every actual $f_j$ was supported in $V$ .

Kurylev, Lassas, and Uhlmann carry out the distorted-plane-wave construction for quadratic semilinear equations. Their microlocal result identifies the fourfold interaction and converts its observed earliest front into light observation sets Kurylev, Lassas, and Uhlmann 2018, Proposition 3.4, Theorem 4.5, and §§3.3–4.2. Once those sets are known for enough $q$ , Theorem 1.5 and the passive reconstruction recover the topology, smooth structure, and conformal class of the accessible diamond.

Why lower interactions do not suffice

A pair of hypersurfaces typically intersects in codimension two, not at an isolated event. Its product generates a conormal source spread along that intersection. Triple interactions can still have extended or polarization- dependent artifacts. In four dimensions, four transversal hypersurfaces have codimension four and generically select a point.

The proof must distinguish the new light cone from singularities already present in the individual waves, from pairwise interaction fronts, and from caustics. It does so by restricting to observation directions outside the known lower-order wavefront sets and by computing the principal symbol. A mere nonzero fourth derivative is not enough; it must have the canonical relation of a point-generated light cone.

The independent check is microlocal: verify that the four input covectors span $T_q^*M$ , that their chosen signed sum is characteristic for the outgoing branch, and that the symbol coefficient does not cancel. If transversality fails, the artificial source is extended. If the coefficient $a(q)$ or a polarization factor vanishes, this interaction may be silent.

The word signed is essential. In Minkowski cotangent coordinates, take $\xi_1=dt+dx$ , $\xi_2=dt-dx$ , $\xi_3=dt+dy$ , and $\xi_4=dt+dz$ . They are four independent null covectors, but their positive sum is timelike, not characteristic. The signed combination

\xi_1+\xi_2+\xi_3-\frac43\xi_4 =\frac53dt+dy-\frac43dz

is null, since $-25/9+1+16/9=0$ . Conormal wavefronts admit signed normal frequencies, allowing such a characteristic interaction direction. This linear-algebra check establishes neither the required accessible source configuration nor nonvanishing of its interaction symbol; those remain separate parts of the theorem.

Einstein–matter interactions

For a quasilinear system, the principal metric itself depends on the solution, so differentiating produces interactions in both coefficients and fields. Diffeomorphism constraints and stress-energy conservation restrict admissible sources. One also has tensor polarizations rather than one scalar symbol.

Kurylev, Lassas, Oksanen, and Uhlmann implement the method for the coupled Einstein–scalar equations about a smooth globally hyperbolic background. In the source neighborhood, gradients of four background scalar fields must be linearly independent. Adaptive sources enforce the conservation law, and the data consist of small source–solution triples written in source-dependent Fermi coordinates. Under those hypotheses, four-wave interactions determine the conformal geometry of the optimal chronological diamond. The full-metric Ricci-flat conclusion is a separate corollary and requires every point of the target region to be connectible to the observation neighborhood by a path inside that region Kurylev, Lassas, Oksanen, and Uhlmann 2022, Theorem 1.1, Corollary 1.2, and §§1.2–1.3, 5.

This does not mean arbitrary laboratory control of stress-energy is assumed. The hypotheses specify a source family satisfying the model’s microlocal surjectivity condition. Nor does the result reconstruct points outside the send-and-return region. Those are genuine limitations of the theorem.

Proof architecture

The full argument has five separable layers:

well-posedness and smooth dependence on the source amplitudes;
construction of four controlled conormal input waves;
symbol calculus proving a new nonvanishing point interaction;
propagation and separation of its earliest observable light front;
geometric reconstruction from the resulting family of observation sets.

Each layer has a different failure mode. This structure is also why the method is a uniqueness mechanism before it is a practical algorithm: differentiating four times and locating singular support are highly noise-sensitive operations.

Diagnose a degenerate interaction

Four null hypersurfaces meet along a smooth curve rather than at one point. What would receivers see, and why does this fail to identify a unique source event?

Solution

Their conormals are not independent, so the intersection has positive dimension. The nonlinear product is conormal to that curve and the retarded solution is a superposition of fronts emitted from all its points. The observed singular set is therefore not one light cone $\partial J^+(q)$ and cannot be assigned a unique earliest-light observation set. Transversality, or a different collection of inputs, is needed to localize the artificial source.

References

Kurylev, Yaroslav, Matti Lassas, and Gunther Uhlmann. “Inverse Problems for Lorentzian Manifolds and Non-linear Hyperbolic Equations.” Inventiones Mathematicae 212 (2018): 781–857. DOI.
Kurylev, Yaroslav, Matti Lassas, Lauri Oksanen, and Gunther Uhlmann. “Inverse Problem for Einstein–Scalar Field Equations.” Duke Mathematical Journal 171 (2022): 3215–3282. DOI.