Operational Meaning of the Riemann Tensor

The Riemann tensor is operational because it controls several distinct local experiments: relative acceleration of freely falling probes, failure of parallel transport around small loops, and the noncommutation of covariant derivatives. No single unboosted gradiometer measures all of its components, but a sufficiently varied family of separation vectors, loop planes, and local frames can reconstruct the full algebraic tensor at an event.

Required background. We assume the algebraic Riemann curvature tensor and the physical scope of equivalence principles and local measurement.

Helpful background. The measurement language uses special relativity, tensor fields, orthonormal frames, and Jacobi fields.

Relative acceleration as a curvature reading

Let $u^\mu$ be the unit tangent of a central timelike geodesic and $\xi^\mu$ the connecting vector to a neighboring geodesic, chosen so that $[u,\xi]=0$. Then

$$\frac{D^2\xi^\mu}{d\tau^2} =-R^\mu{}_{\alpha\nu\beta} u^\alpha\xi^\nu u^\beta.$$

Choose an orthonormal tetrad with $e_{(0)}=u$ and a parallel-transported spatial triad. For a spatial separation $\xi^{(i)}$, the measured equation is

$$\ddot\xi^{(i)}=-\mathcal E^i{}_j\xi^{(j)}, \qquad \mathcal E_{ij}=R_{i0j0}.$$

The matrix $\mathcal E_{ij}$ is symmetric, so one observer’s ideal relative-acceleration experiment yields at most six independent numbers. Repeating the experiment with three linearly independent separations determines that matrix. Reall derives the Jacobi equation from a two-parameter geodesic family and identifies it with the observable tidal field Reall 2013, § 6.5, pp. 60–62, PDF.

These six components do not exhaust four-dimensional curvature. Components involving different time directions or purely spatial two-planes require additional transport protocols or observers with different velocities.

Covariant-derivative commutators

For any vector $V^\rho$,

$$[\nabla_\mu,\nabla_\nu]V^\rho =R^\rho{}_{\sigma\mu\nu}V^\sigma.$$

This identity gives an abstract but direct measurement prescription. Prepare a vector or tetrad, compare covariant changes in the two orders associated with directions $e_{(\mu)}$ and $e_{(\nu)}$, and infer the linear map $R(e_{(\mu)},e_{(\nu)})$ acting on the prepared vector. Varying $V$, the ordered direction pair, and the tetrad orientation samples different components.

The infinitesimal-loop form is equivalent. Parallel transport $V^\mu$ around a small oriented parallelogram with area bivector $\Sigma^{\alpha\beta}$. To leading order,

$$\Delta V^\mu =\frac12R^\mu{}_{\nu\alpha\beta} V^\nu\Sigma^{\alpha\beta} +O(L^3),$$

where the factor $1/2$ accompanies the convention that $\Sigma^{\alpha\beta}$ includes both antisymmetric index orderings. For traversal first along $\delta a\,A$, then $\delta b\,B$, then back along the two opposite sides, take $\Sigma^{\alpha\beta}=\delta a\,\delta b(B^\alpha A^\beta-B^\beta A^\alpha)$. The remainder is at fixed loop shape: curvature gradients first enter at cubic order, with second derivatives and curvature products at quartic order. Reversing the loop orientation sends $\Sigma^{\alpha\beta}\mapsto-\Sigma^{\alpha\beta}$ and reverses the leading change. This sign test distinguishes genuine curvature from a fixed apparatus offset.

Why four dimensions give twenty components

Lower the first index and use the Levi-Civita symmetries:

$$R_{abcd}=-R_{bacd}=-R_{abdc}, \qquad R_{abcd}=R_{cdab}, \qquad R_{a[bcd]}=0.$$

An antisymmetric index pair labels a bivector. In four dimensions there are $\binom42=6$ independent bivectors. Pair exchange makes $R$ a symmetric $6\times6$ bilinear form, initially giving $6(6+1)/2=21$ components. The algebraic Bianchi identity removes one independent pseudoscalar combination, leaving twenty.

Carroll gives this counting and the Ricci–Weyl decomposition that reorganizes the twenty components Carroll 1997, Chapter 3, pp. 74–84, PDF. Ten combinations reside in Ricci curvature and ten in the Weyl tensor in four dimensions. This is an algebraic statement at an event; no field equation has been assumed.

A reconstruction strategy

One timelike observer measures the six entries of $R_{i0j0}$ through relative acceleration. Infinitesimal timelike–spacelike and spacelike–spacelike transport loops probe mixed and purely spatial curvature maps. Equivalently, repeat the tidal experiment with boosted four-velocities: the time direction $u$ changes, and the new matrices $R_{a b c d}u^b u^d$ provide independent linear combinations. Enforcing the Riemann symmetries and Bianchi identity turns the measurements into a twenty-parameter linear inversion.

The reconstruction assumes ideal point probes, calibrated tetrads, negligible curvature gradients during one cycle, and loops small enough that higher powers of area can be ignored. A finite loop measures integrated holonomy rather than curvature at one event, and global topology can produce holonomy even where the local curvature vanishes along parts of the path.

Exercises

Explain why measuring $\mathcal E_{ij}=R_{i0j0}$ for one observer cannot determine all twenty curvature components, even if arbitrarily many separation vectors are used.

Solution

Arbitrarily many separations determine only the linear map $\xi^j\mapsto-R^i{}_{0j0}\xi^j$. Because $\mathcal E_{ij}$ is symmetric, it has six independent entries. Mixed components $R_{0ijk}$ and purely spatial components $R_{ijkl}$ do not enter that map; a component with three time indices vanishes by antisymmetry. One must change the loop plane, transported vector, or observer velocity to obtain the missing combinations.

References

• Carroll, Sean M. “Lecture Notes on General Relativity.” NSF-ITP/97-147, arXiv:gr-qc/9712019v1 (1997). arXiv.
• Reall, Harvey S. General Relativity. Cambridge Part III lecture notes (2013). PDF.