Relativistic Gradiometry and Curvature Reconstruction

Curvature reconstruction is a local tensor-tomography problem. A nonrotating free-fall array at one event determines the six components of the electric Riemann tensor for its four-velocity, but a four-dimensional Riemann tensor has twenty independent components. Complete reconstruction therefore requires twenty independent scalar equations assembled from genuinely different contractions—most naturally boosted arrays, supplemented when useful by loop holonomy or gyroscope transport. Spatially rotating one array cannot supply the missing information.

Required background. Affine and Variational Geodesics supplies the deviation experiment; The Riemann Curvature Tensor supplies the algebraic symmetries and Bianchi identity; Observers, Rest Spaces, and Measured Components supplies calibrated tetrads and observer changes.

Helpful background. Events, Intervals, and Minkowski Geometry supplies Lorentz boosts; Tensor Fields and Natural Operations supplies linear reconstruction from contractions; The Newtonian Gravitational Field and Potential provides the six-component low-velocity benchmark.

Why the target has twenty components

At an event, regard $R_{\mu\nu\rho\sigma}$ as a bilinear form on the six-dimensional space of antisymmetric index pairs. Antisymmetry within each pair leaves six bivector labels. Pair exchange makes the resulting $6\times6$ matrix symmetric, giving $6(6+1)/2=21$ entries. The algebraic Bianchi identity removes one combination, leaving

$$21-1=20$$

independent components in four spacetime dimensions. Carroll gives the same four-dimensional count while developing the Riemann symmetries in Carroll 1997, §3, pp. 80–83, PDF.

For a chosen unit timelike vector $u$, an ideal initially comoving array measures

$$\mathcal E_{\mu\nu}(u) =R_{\mu\alpha\nu\beta}u^\alpha u^\beta,$$

a spatial symmetric tensor with six entries. Three independent separation vectors determine those entries through $\ddot\xi^{\hat i}=-\mathcal E^{\hat i}{}_{\hat j}\xi^{\hat j}$. Reall shows more abstractly that deviation data for freely selectable tangent and connecting vectors determine the curvature tensor by polarization and its symmetries Reall 2013, §6.5, pp. 60–62, PDF. A physical protocol must realize enough distinct tangent directions to make that algebraic statement operational.

Why rotations are insufficient and boosts work

Changing only the spatial triad while retaining the same observer sends $\mathcal E\mapsto Q\mathcal E Q^{\mathsf T}$. It produces different displayed numbers but no new tensor information. The mixed time-space and purely spatial curvature blocks remain invisible to that observer’s instantaneous, initially comoving Jacobi experiment.

A Lorentz boost changes the four-velocity itself. Consequently $\mathcal E(u')$ mixes the original $R_{0i0j}$ entries with $R_{0ijk}$ and $R_{ijkl}$. Repeating the array experiment for non-coplanar, distinct boost velocities produces new linear equations for the same twenty unknown curvature components.

There is no observer-independent “minimum number of spacecraft” because one device may return several scalar channels and loop or precession data may replace array channels. The invariant requirement is a design matrix of rank twenty. The following exact electric-array construction makes one successful five-velocity design reproducible.

Fix an orthonormal tetrad with $\eta_{ab}=\operatorname{diag}(-1,1,1,1)$ and choose

$$\begin{aligned} u_1&=(1,0,0,0), &u_2&=\frac14(5,3,0,0),\\ u_3&=\frac14(5,0,3,0), &u_4&=\frac14(5,0,0,3),\\ u_5&=\frac1{3131}(4069,1800,1440,1200). \end{aligned}$$

Every vector is future directed and satisfies $u_n^au_{na}=-1$. The first observer is at rest in the reference tetrad, the next three move at speed $3/5$ along its three axes, and the last has nonzero velocity in all three directions. The awkward-looking final integers are useful rather than arbitrary: they give an exactly normalized rational point on the unit timelike hyperboloid.

To build the twenty columns, use the oriented bivector basis

$$P_A=(01,02,03,23,31,12)$$

and the symmetric matrix $M_{AB}=R_{P_AP_B}$. Its twenty-one symmetric entries obey the single four-dimensional Bianchi relation

$$M_{03}+M_{14}+M_{25}=0.$$

Here the digits on $M_{03}$, $M_{14}$, and $M_{25}$ are bivector-basis labels $A,B$, not spacetime tensor indices. Eliminate $M_{25}=-M_{03}-M_{14}$ and use the other twenty entries as the unknown vector. For each $u=(u^0,u^1,u^2,u^3)$, a convenient exact basis of its rest space is

$$\begin{aligned} s_1&=(u^1,u^0,0,0),\\ s_2&=(u^2,0,u^0,0),\\ s_3&=(u^3,0,0,u^0). \end{aligned}$$

These vectors obey $u\cdot s_i=0$. They need not be orthonormal: passing to an orthonormal apparatus triad applies an invertible congruence to the measured symmetric matrix and cannot change its rank. The six rows contributed by observer $n$ are the coefficients of

$$e_{ij}^{(n)} =R_{acbd}s_i^a u_n^c s_j^b u_n^d, \qquad 1\le i\le j\le3,$$

in the twenty-entry basis above. Substituting the displayed rational vectors and performing exact fraction-free row reduction gives

Observer arrays included	Design rows	Rank	Undetermined combinations
1	6	6	14
2	12	11	9
3	18	15	5
4	24	18	2
5	30	20	0

The shared information has a geometric origin. The electric tensor for $u$ is the restriction of the curvature bilinear form on bivectors to the three-dimensional subspace $S_u=u\wedge u^\perp$. For distinct $u$ and $v$, the line spanned by $u\wedge v$ lies in both $S_u$ and $S_v$, so their two six-entry restrictions cannot be independent in all twelve entries. Exact elimination establishes the full sequence for this design; a nonzero rank minor remains nonzero for nearby velocities, while special coplanar, repeated, or symmetry-related choices can have lower rank. In particular, simply counting $4\times6>20$ misses the two curvature directions that four arrays leave undetermined here.

Loop and gyroscope channels

Parallel transport around an infinitesimal oriented loop supplies a complementary contraction. If $\Sigma^{\alpha\beta}$ is its area bivector, then, with loop orientation fixed consistently,

$$\Delta V^\mu =\frac12R^\mu{}_{\nu\alpha\beta} V^\nu\Sigma^{\alpha\beta} +O(\Sigma^{3/2}).$$

Reversing the loop reverses the leading change. Varying the transported vector and loop plane probes the curvature operator on bivectors directly, including sectors that a single electric array misses. Comparing gyroscopes transported along nearby paths realizes the same holonomy principle through relative orientation. These channels can replace some boosted-array equations, but only if their calibrated contractions raise the combined design-matrix rank to twenty.

Loops must be small enough that curvature variation is negligible, and gyroscope torques and tetrad rotation must be separated from geometric holonomy. A loop with finite area measures an integral contaminated by $\nabla R$ and higher powers of size, not a point value automatically.

A reconstruction workflow and its checks

First choose a reference event and orthonormal tetrad, synchronize the release data, and verify that the central reference probes are geodesic. Second, use three independent baselines to infer the six entries of $\mathcal E(u)$. Third, add boosted arrays or calibrated loop and gyro channels, writing every reading as a linear equation in a twenty-component Riemann basis. Fourth, solve only after confirming rank twenty and report conditioning, since nearly redundant boosts can amplify noise.

Finally test agreement with channels withheld from the fit, consistency of overlapping readings, reversal of loop orientation, and stability as baselines and loop areas shrink. Antisymmetry, pair exchange, and the algebraic Bianchi identity check the tensor implementation, but they are already built into the twenty-column basis above and are not independent tests of the measured data. Accelerated platforms require force-gradient, frame, and synchronization corrections, as explained on the generalized-deviation page; finite arrays require extrapolation past $O(\nabla R L^2)$ and higher corrections. The quadratic Fermi metric that organizes these finite-size limits is given in Manasse and Misner 1963, p. 743, Eqs. (65)–(66).

As a limiting check, retaining only the leading slow-velocity order in a stationary Newtonian field leaves $R_{i0j0}=\partial_i\partial_j\Phi$. This supplies six symmetric Hessian components. At nonzero boost velocity the other sectors can still enter, but their suppressed coefficients make reconstruction progressively ill-conditioned as all velocities approach the same rest frame. Full rank and useful precision are separate requirements.

Exercises

1. Derive the component count. Starting from antisymmetric index pairs, reproduce the number twenty.

Solution

There are $\binom42=6$ antisymmetric pairs. Pair exchange makes Riemann a symmetric $6\times6$ bilinear form, with $6\cdot7/2=21$ components. The algebraic Bianchi identity imposes one independent four-dimensional relation, leaving $20$.

2. Rotate or boost? Why can three spatial rotations of one array not replace a second observer velocity?

Solution

All rotations preserve $u$ and therefore sample the same contraction $R_{\mu\alpha\nu\beta}u^\alpha u^\beta$ in different bases. A boost replaces $u$ by $u'$, so the quadratic contraction mixes electric, mixed, and spatial curvature blocks and can add independent equations.

3. Diagnose incomplete rank. The first four electric arrays in the displayed design give a combined rank of eighteen. Two loop-derived scalar equations are then added. When is the reconstruction complete?

Solution

Only when the two loop rows are linearly independent of each other and of the existing eighteen rows, raising the combined rank to twenty. Merely collecting two more numbers is insufficient. One must compute the rank in the same constrained Riemann basis and then test the reconstructed tensor against unused symmetry or repeated-channel data.

References

• Carroll, Sean M. “Lecture Notes on General Relativity.” arXiv:gr-qc/9712019 (1997), §3, pp. 77–88. PDF.
• Manasse, F. K., and C. W. Misner. “Fermi Normal Coordinates and Some Basic Concepts in Differential Geometry.” Journal of Mathematical Physics 4, no. 6 (1963): 735–745. DOI.
• Reall, Harvey S. General Relativity. Cambridge Part III lecture notes, 2013, §6.5, pp. 60–62. PDF.