Curvature Invariants for Exact-Spacetime Identification

Scalar curvature invariants can decisively rule out a proposed local isometry, recover parameters and invariantly distinguished surfaces in many exact families, and encode algebraic syzygies. They do not, in general, prove equivalence: nonisometric degenerate Kundt spacetimes can have identical scalar polynomial invariants, including the extreme case in which every such invariant vanishes.

Required background. The Riemann Curvature Tensor supplies the contractions and curvature convention used here; Killing Fields and Isometry Groups provides an independent equivalence discriminator; The Einstein Equation relates Ricci invariants to the matter model.

Helpful background. Frames, Coframes, and Anholonomy prepares the frame-normalized invariants needed when scalar contractions are incomplete.

What scalar polynomial invariants contain

A scalar polynomial curvature invariant (SPI) is a complete contraction of a polynomial in the Riemann tensor and its covariant derivatives. Familiar examples are

$$R, \qquad R_{\mu\nu}R^{\mu\nu}, \qquad K=R_{\mu\nu\rho\sigma}R^{\mu\nu\rho\sigma}, \qquad \nabla_\alpha R_{\mu\nu\rho\sigma} \nabla^\alpha R^{\mu\nu\rho\sigma}.$$

One can also regard curvature as an endomorphism—on tangent vectors, bivectors, or trace-free symmetric tensors—and use the coefficients of its characteristic polynomial. Traces of powers then give eigenvalue invariants. Polynomial relations among invariants are syzygies; they efficiently screen algebraic type or special parameter loci.

Under a local isometry $\psi$, every scalar satisfies $\widetilde I=I\circ\psi$. Therefore different invariant ranges, different functional relations, or different numbers of functionally independent invariants prove inequivalence. Equal numerical values at arbitrarily chosen coordinate points prove very little, because the unknown isometry need not pair points carrying the same coordinate labels.

Schwarzschild as a parameter screen

Use $m=GM$ as a length. In a Schwarzschild region with $r>0$,

$$R=0, \qquad R_{\mu\nu}R^{\mu\nu}=0, \qquad K=\frac{48m^2}{r^6}.$$

The first two vanish because Schwarzschild is Ricci flat, but their vanishing alone does not establish Ricci flatness in Lorentzian signature. For example, a nonzero null-radiation tensor $R_{ab}=\Phi k_a k_b$ with $k^ak_a=0$ has both $R=0$ and $R_{ab}R^{ab}=0$. Thus these contractions do not even distinguish every matter geometry from vacuum. The Kretschmann scalar detects nonflat curvature and its divergence at $r=0$ for $m\ne0$, but $K$ alone mixes $m$ and $r$. Adding its squared gradient,

$$J=\nabla_\alpha K\nabla^\alpha K =\frac{82944m^4}{r^{14}}\left(1-\frac{2m}{r}\right),$$

provides a second relation. The zero of $J$ at $r=2m$ is an invariantly distinguished hypersurface in the $m>0$ family, while $K$ remains finite there. Ratios of $K$ and $J$ can eliminate coordinate dimensions and recover $m/r$. This is screening within a declared family, not a universal horizon detector.

Mergulhão and Batista show why matching a finite list—or even all functional relations among SPIs in exceptional families—is not a complete equivalence proof Mergulhão and Batista 2021, §I, printed pp. 1–3, PDF.

The VSI counterexample

Consider the Brinkmann metric

$$ds^2=-2du dv+dx^2+dy^2+H(u,x,y)du^2.$$

For $\partial_x^2H+\partial_y^2H=0$ it is a vacuum pp-wave. Choosing $H=x^2-y^2$ gives nonzero components such as $R_{uxux}$, so the geometry is not flat. Nevertheless every SPI constructed from the Riemann tensor and its derivatives vanishes. Minkowski spacetime has the same complete list of zero SPIs. The aligned null structure prevents the nonzero negative-boost-weight curvature components from surviving a complete scalar contraction.

This example fixes the logical direction:

$$\text{different SPIs}\Longrightarrow\text{not locally isometric},$$

but equal SPIs need not imply local isometry. Coley, Hervik, and Pelavas prove that a four-dimensional Lorentzian metric not characterized by its SPIs must lie in the degenerate Kundt class; the theorem is local and assumes the smooth four-dimensional Lorentzian setting used in their analysis Coley, Hervik, and Pelavas 2009, §§1–3 and §7, printed pp. 1–7 and 21–24, Open PDF.

A disciplined invariant workflow

First use low-order Ricci and Weyl invariants to reject impossible matches. Next compare syzygies, ranks, invariant ranges, and parameter limits. Count independent invariants to locate cohomogeneity and compare Killing data separately. If the candidate is known to be invariantly nondegenerate, sufficiently rich SPI data can characterize it locally. If it is Kundt, VSI, CSI, or algebraically highly special, escalate to normalized Cartan invariants rather than declaring victory from scalar agreement.

SPIs also do not settle global topology, geodesic completeness, or the nature of every singularity. Those questions require domain, extension, and curve-based information.

Check your understanding

A metric has $R=0$ and $K=0$ everywhere. Which conclusions are valid?

Solution

The vanishing Ricci scalar says only that the Ricci tensor has zero trace; it does not imply $R_{\mu\nu}=0$. The vanishing Kretschmann scalar does not imply $R_{\mu\nu\rho\sigma}=0$ in Lorentzian signature, because nonzero null-aligned tensors can have zero quadratic contraction. Even if every SPI vanishes, a nonflat VSI pp-wave remains possible. Flatness must be checked from the full curvature tensor or complete frame-normalized data. Thus none of “vacuum,” “flat,” or “Minkowski” follows from the two stated equalities.

References

• Coley, Alan, Sigbjørn Hervik, and Nicos Pelavas. “Spacetimes Characterized by Their Scalar Curvature Invariants.” Classical and Quantum Gravity 26 (2009), 025013. doi:10.1088/0264-9381/26/2/025013.
• Mergulhão, Thiago M., and Carlos Batista. “On the Equivalence of Spacetimes, the Cartan–Karlhede Algorithm.” General Relativity and Gravitation 53 (2021), 28. doi:10.1007/s10714-021-02820-4.