Boundary Distance and Lorentzian Time Separation

Lorentzian time separation records the greatest proper time between causally ordered events. It resembles Riemannian boundary distance only superficially: timelike geodesics maximize locally, null-related points have zero separation, unrelated points also have zero separation, and cut or conjugate points make the maximizing branch nonsmooth. Reconstruction therefore depends on which pairs are timed and on how their maximizing geodesics cover the unknown region.

Required background. Analysis for Mathematical Relativity supplies regularity of value functions; PDEs and Green Functions for Gravitation supplies inverse-boundary methods; and Lorentzian metrics and signature supplies causal length and geodesics.

Helpful background. The Einstein equation supplies rigidity in stationary solution classes, and Causality and Global Structure supplies global hyperbolicity and cut loci.

Time separation as data

For a time-oriented spacetime $(M,g)$, define

$$\tau_g(p,q)= \sup_{\gamma:p\rightsquigarrow q} \int_\gamma\sqrt{-g(\dot\gamma,\dot\gamma)}\,ds$$

when $q\in J^+(p)$, with the supremum over future causal curves, and set $\tau_g(p,q)=0$ otherwise. In a globally hyperbolic spacetime the supremum is finite and is attained. If $q\in I^+(p)$ and the maximizing timelike geodesic is unique and $q$ is outside its timelike cut locus (in particular, there is no conjugate point at or before $q$), then $\tau_g$ is smooth near $(p,q)$ and satisfies the Lorentzian eikonal equation

$$g^{\mu\nu}(p)\,\partial_{p^\mu}\tau_g\,\partial_{p^\nu}\tau_g=-1.$$

Thus first derivatives give the initial unit tangent of the maximizer, and second derivatives encode its Jacobi fields. The singular set detects where maximizers bifurcate or acquire conjugate points. Unlike Riemannian distance, the zero set alone mixes spacelike and null-related pairs; its boundary in the causal ordering must also be known.

The measurement geometry might provide $\tau_g\vert_{\Sigma\times\Sigma}$ for a timelike or higher-dimensional submanifold $\Sigma$, or only selected source–receiver pairs on a boundary. No theorem can recover a region through which no relevant maximizing or null geodesic passes.

Local jet recovery

The local comparison theorem concerns two smooth Lorentzian manifolds with codimension-one timelike hypersurfaces $\Sigma$ and $\widetilde\Sigma$. Their closures must lie in simply convex neighborhoods. A diffeomorphism $\Phi:\Sigma\to\widetilde\Sigma$ identifies the apparatus, and the two time-separation functions must agree on every pair of apparatus points after this identification. At a chosen $x_0\in\Sigma$ and timelike $\xi_0\in T_{x_0}\Sigma$, both hypersurfaces must also satisfy the paper’s timelike-convexity condition in compatible normal directions.

Under precisely these assumptions, equality of the data implies equality of the full $C^\infty$ metric jets at $x_0$ and $\Phi(x_0)$ in the associated semigeodesic, or Fermi, coordinates Lassas, Oksanen, and Yang 2016, Theorem 1 and §4. The proof architecture is:

1. use short timelike separations to recover the metric restricted to $T\Sigma$;
2. use timelike convexity to obtain nearby timelike geodesics that leave and return to the measurement hypersurface;
3. compare their length identities in semigeodesic coordinates and recover successive normal derivatives of the metric inductively.

This is a boundary determination theorem. Smooth jet recovery is not by itself global rigidity: two smooth metrics may share every boundary jet yet differ in the interior.

For example, in a normal coordinate $x$ with measurement hypersurface $x=0$, the function $b(x)=e^{-1/x^2}$ for $x\gt0$ and $b(x)=0$ for $x\le0$ is smooth and has every derivative zero at $0$, yet is nonzero in the interior. A sufficiently small perturbation $b(x)k_{\mu\nu}$ preserves Lorentzian signature and leaves every metric jet on that hypersurface unchanged. This does not assert identical full travel-time data; it explains exactly why jet recovery alone needs another argument to become interior recovery.

Extending into the interior

Global recovery requires additional structure. If two connected, real-analytic Lorentzian manifolds satisfy the local theorem and are geodesically complete modulo scalar-curvature singularities, the same data determine only their universal Lorentzian covering spaces up to isometry. This is Theorem 2; the manifolds themselves need not be isometric when their fundamental groups differ.

The stationary Einstein–scalar application has two additional steps. Proposition 3 proves real analyticity when the smooth metric and scalar fields solve the stated Einstein–scalar system with real-analytic potential, admit a smooth timelike Killing field, and the scalar fields are invariant under that field. Corollary 4 then obtains an isometry of the original spacetimes by also assuming simple connectivity, completeness modulo scalar-curvature singularities, and the local measurement-identification and timelike-convexity hypotheses. Its further recovery of the scalar fields and Killing field needs the extra matching and transversality data written in the corollary Lassas, Oksanen, and Yang 2016, Theorem 2, Proposition 3, Corollary 4, and §§3.2, 5.

The hypotheses are consequential. Analyticity replaces the missing unique continuation for arbitrary smooth metrics. Completeness controls premature termination of the geodesic probes. Stationarity is a structural reduction, not a coordinate convenience. The theorem’s counterexamples show that dropping such assumptions can hide an interior modification from the measured time data.

Cut loci and coverage

At a timelike cut point, several maximizing geodesics can give the same separation; at a conjugate point, the exponential map loses rank. The value function remains geometrically meaningful but ceases to provide one smooth coordinate branch. A proof may avoid this set, use multiple branches, or extract it as additional data, but cannot differentiate through it as though nothing happened.

For null scattering data, a related nonconjugacy hypothesis makes a defining function for null-related boundary pairs act like a Lorentzian boundary distance: its linearization is the light ray transform of the metric perturbation Stefanov 2024, Theorem 2.1(c) and §§2.3, 3. This determines only the portion visible to the measured null geodesics and is subject to diffeomorphism and conformal gauges. It is not equivalent to proper- time data, which can fix a conformal factor.

Accordingly, the strongest reconstructible region is the one swept by controlled geodesic branches connecting the observed set, augmented only when analyticity, field equations, or another propagation theorem justifies an extension. The limitation is coverage, not merely proof technique.

Minkowski check

In Minkowski space with signature $(-,+,+,+)$,

$$\tau(p,q)^2=(q^0-p^0)^2-\lvert\mathbf q-\mathbf p\rvert^2$$

for timelike-related points. Show how measurements with $p=(0,\mathbf0)$ and $q=(T,\mathbf x)$ determine the conformal scale that a light-cone observation would miss.

Solution

The light cone determines only $T^2=\lvert\mathbf x\rvert^2$, which is unchanged by a constant rescaling of the metric. Proper-time data give the numerical value $\tau^2=T^2-\lvert\mathbf x\rvert^2$. Under $g\mapsto c^2g$, this value becomes $c^2\tau^2$. Hence any nonzero timelike separation fixes the constant scale relative to the clock units on the measurement set. A variable conformal factor requires enough geodesic coverage to reconstruct its variation.

References

• Lassas, Matti, Lauri Oksanen, and Yang Yang. “Determination of the Spacetime from Local Time Measurements.” Mathematische Annalen 365 (2016): 271–307. DOI.
• Stefanov, Plamen. “The Lorentzian Scattering Rigidity Problem and Rigidity of Stationary Metrics.” arXiv:2212.13213v3 (2024). arXiv.