Inverse Problems and Geometric Reconstruction

Inverse problems ask how much spacetime geometry can be recovered from signals recorded in a restricted region. The answer is never simply “the metric.” Diffeomorphisms fixed on the apparatus are invisible, null paths often see only conformal geometry, and causal coverage limits the reconstructible region. This chapter moves from those obstructions through passive and active data to nonlinear interaction probes, then separates uniqueness from stability and computation.

Helpful background. Analysis for Mathematical Relativity supplies microlocal and stability tools; PDEs and Green Functions for Gravitation supplies wave response maps; Lorentzian metrics and signature supplies null and timelike geometry; and Causality and Global Structure supplies accessible diamonds.

From invariance to reconstruction

The pages follow the logical order of a rigorous inverse theorem.

1. Gauge Obstructions in Geometric Inverse Problems identifies boundary-fixing diffeomorphisms, conformal freedom, internal field gauge, and constrained source relabeling. It defines the strongest invariant output before a reconstruction is attempted.
2. Boundary Distance and Lorentzian Time Separation shows how proper-time data recover local metric jets and why cut loci, conjugate points, coverage, analyticity, or stationarity matter for global continuation.
3. Lens and Light Observation Data compares entry–exit relations with families of earliest light observations and reconstructs topology, smooth structure, and null cones in a visible region.
4. Active Wave Measurements defines source-to-solution and Dirichlet-to-Neumann maps, then explains how propagation, controllability, and unique continuation expose the interior.
5. Nonlinear Wave Interactions differentiates the response four times to create an artificial point source from transversal conormal waves.
6. Stability, Partial Data, and Reconstruction Algorithms gives a dated evidence matrix, records a relevant withdrawn proof claim, and identifies the gap between qualitative uniqueness and a robust finite algorithm.

Kurylev, Lassas, and Uhlmann establish the core passive and active reconstruction architecture in Kurylev, Lassas, and Uhlmann 2018, Theorem 1.2, Corollary 1.3, Theorem 1.5, and §§2–5. The pages here keep each hypothesis and output attached to the exact data model rather than treating that architecture as one universal theorem.

The causal region of an experiment

If sources are confined to $S$ and receivers to $R$, finite propagation makes

$$\mathcal V(S,R)=J^+(S)\cap J^-(R)$$

the first candidate reconstruction region. A point outside it cannot influence the measurements. Inside it, enough null directions must still be visible and their singularities separable. Trapping, event horizons, conjugate points, short observation time, or vanishing interaction symbols can reduce the usable region.

Different measurements determine different structures. Unparametrized null paths determine causal cones and hence the conformal metric. Timelike separation supplies scale through proper time. Wave amplitudes can encode geometric spreading and lower-order coefficients. Field equations can sometimes fix a conformal factor, but only with additional structure. For example, the Einstein–scalar result assumes a globally hyperbolic background, nondegenerate background scalar fields in the source region, and adaptive sources satisfying the conservation law; its Ricci-flat full-metric upgrade also imposes an internal connectivity condition on the target region Kurylev, Lassas, Oksanen, and Uhlmann 2022, Theorem 1.1, Corollary 1.2, and §§1.2–1.3, 5. None of these upgrades follows merely by renaming light data as “metric data.”

The nonlinear advantage

The linear response propagates injected singularities independently. Its canonical relation reveals null geodesics, but a linear inverse problem may have a microlocal kernel or insufficient angular coverage. For a nonlinear wave equation, mixed Fréchet derivatives contain products of linear waves. Four null hypersurfaces in four dimensions can meet transversally at an isolated event $q$, producing a new singularity that propagates on $\partial J^+(q)$.

Observed from the original apparatus, that new cone behaves as though a point source had been placed at $q$. Repeating the construction generates the family of light observation sets needed to rebuild an atlas and cone field. The method does not send information outside causal reach, and its high-order differentiation is not automatically noise-stable. Its achievement is to turn nonlinearity from a nuisance into additional geometric information.

What a complete result must state

For every theorem in this chapter, record:

• dimension, topology, global hyperbolicity, boundary, and asymptotic ends;
• source and receiver regions, time interval, controllable field components, and observation operator;
• metric, coefficient, or coupled matter unknowns and their regularity class;
• visibility, convexity, nontrapping, nonconjugacy, curvature, or analyticity hypotheses;
• the diffeomorphism, conformal, and internal gauge quotient;
• whether the conclusion is qualitative uniqueness, a constructive exact-data procedure, a stability estimate, or a convergent noisy-data algorithm.

This checklist also protects the literature record. A current arXiv entry can replace an earlier theorem claim with a withdrawal notice. The final page uses a 2026-08-12 cutoff precisely so that theorem status is part of the evidence.

Review the chapter

An experiment knows the exact nonlinear source-to-solution map on a worldtube $V$. Four controlled waves meet transversally at $q$ and their artificial cone returns to $V$. What can this observation determine, and what does it not yet prove?

Solution

The fourth mixed response derivative can identify the earliest light observation set of $q$. Repeating this for enough points reconstructs the topology and smooth structure of the accessible diamond; tangent light fronts recover its conformal cones, all modulo diffeomorphisms fixed on $V$. A full metric needs a separate determination of conformal scale, for example from amplitudes or field equations under their hypotheses. The exact operator does not give a stability bound for noisy measurements, and the microlocal construction does not itself provide a convergent finite discretization.

Connections

The preceding asymptotic radiation chapter constructs the null and scattering data that can serve as measurements at infinity. The Einstein initial-value chapters explain gauge reduction and constraint propagation needed by active Einstein–matter experiments. This chapter closes the volume by reversing the forward problem: from selected consequences of evolution back to geometry.

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.