Extended Free-Fall Laboratories

A freely falling laboratory is locally inertial only in a controlled small-region limit. Along its central geodesic the metric can be Minkowskian and its first derivatives can vanish, but quadratic curvature terms produce measurable relative acceleration, clock-rate variation, and velocity-dependent effects across any finite apparatus. Laboratory size $L$, duration $T$, internal velocity $v$, curvature magnitude, and curvature-gradient scale together determine the accuracy of local inertial reasoning.

Required background. Affine and Variational Geodesics supplies the central geodesic and neighboring Jacobi fields; The Riemann Curvature Tensor supplies curvature and its derivatives; Observers, Rest Spaces, and Measured Components supplies the laboratory tetrad and clock readings.

Helpful background. Events, Intervals, and Minkowski Geometry supplies the tangent-space comparison; Tensor Fields and Natural Operations reviews covariant expansions; The Newtonian Gravitational Field and Potential supplies an intuitive tidal-gradient estimate.

Fermi coordinates around the central probe

Let the laboratory center follow a timelike geodesic with proper time $T$, and parallel-transport an orthonormal triad along it. Fermi normal coordinates $(T,X^i)$ are constructed by spacelike geodesics emitted orthogonally from the central worldline. On that worldline, $g_{\hat\alpha\hat\beta}=\eta_{\hat\alpha\hat\beta}$ and $\Gamma^{\hat\alpha}{}_{\hat\beta\hat\gamma}=0$.

Translated to the site’s curvature convention, the leading metric terms are

$$\begin{aligned} g_{00}&=-1-R_{0i0j}X^iX^j+O(X^3),\\ g_{0i}&=-\frac23R_{0jik}X^jX^k+O(X^3),\\ g_{ij}&=\delta_{ij}-\frac13R_{ikjl}X^kX^l+O(X^3). \end{aligned}$$

Curvature components are tetrad components evaluated on the central worldline. Manasse and Misner construct these coordinates on pp. 735–743 and display the quadratic coefficients on Manasse and Misner 1963, p. 743, Eqs. (65)–(66). Their curvature sign is opposite to the site’s, so $R^{\rm MM}_{abcd}=-R^{\rm site}_{abcd}$ before translation. Independently, $-\Gamma^i{}_{00}$ derived from the translated $g_{00}$ coefficient is $-R^i{}_{0j0}X^j$, exactly the site’s geodesic-deviation equation.

Size, time, and gradient error parameters

Let $\lVert R\rVert$ bound relevant tetrad curvature components and let $\lVert\nabla R\rVert$ bound their first spatial derivatives across the apparatus. The useful dimensionless controls are

$$\epsilon_L=\lVert R\rVert L^2, \qquad \epsilon_T=\lVert R\rVert T^2, \qquad \epsilon_\nabla=\lVert\nabla R\rVert L^3.$$

The metric differs from Minkowski form by $O(\epsilon_L)$. Two points separated by $L$ acquire differential acceleration of order

$$|\delta a|\sim\lVert R\rVert L,$$

relative velocity $\sim\lVert R\rVert LT$, and relative displacement $\sim\lVert R\rVert LT^2$. The leading curvature-gradient correction to relative acceleration is $O(\lVert\nabla R\rVert L^2)$, a fraction $O(L/L_\nabla)$ of the Jacobi term when $L_\nabla\equiv\lVert R\rVert/\lVert\nabla R\rVert$. Quadratic curvature corrections are relatively $O(\epsilon_L)$. Internal velocities also activate the $g_{0i}$ and spatial-curvature sectors, producing terms schematically of order $\lVert R\rVert Lv$ and $\lVert R\rVert Lv^2$ in the local equations of motion.

These estimates require a smooth curvature field, $L$ within the normal-coordinate tube, test bodies small enough to neglect multipoles and self-force, and duration short enough that caustics or large tidal displacements do not spoil the single-laboratory description. The curvature may also vary along the central worldline. Treating $R(T)$ as a constant requires $T\lVert\nabla_uR\rVert\ll\lVert R\rVert$ in addition to spatial control; otherwise the time-dependent Jacobi equation must be integrated. Near a zero of the retained curvature component, compare omitted terms with the permitted absolute error rather than divide by that component. At the next spatial order, $\nabla\nabla R$ contributes alongside $R^2$ and must be checked if first gradients happen to vanish.

A laboratory tolerance calculation

Suppose an experiment seeks an acceleration signal $a_{\rm sig}$ and allows unmodeled tidal acceleration no larger than a fraction $\varepsilon_a$. It must satisfy

$$L\lesssim\frac{\varepsilon_a a_{\rm sig}}{\lVert R\rVert}.$$

If the coordinate-metric approximation itself must be accurate to $\varepsilon_g$, then also

$$L\lesssim\sqrt{\frac{\varepsilon_g}{\lVert R\rVert}}.$$

Finally, keeping accumulated relative displacement below a fraction $\varepsilon_x$ of the baseline requires $\lVert R\rVert T^2\lesssim\varepsilon_x$. The admissible laboratory is set by the most restrictive bound, not by an informal declaration that it is “small.” A nearly uniform common acceleration does not enter these criteria; only curvature and its variation do. The persistence of such finite-region effects in free fall is the tidal qualification to the equivalence principle Reall 2013, §1.4, pp. 4–5, PDF.

Clock networks and relative velocity

For a clock held momentarily at fixed Fermi position, the time-time metric gives

$$\frac{d\tau}{dT} =1+\frac12R_{0i0j}X^iX^j+O(X^3).$$

Two clocks at $X_A$ and $X_B$ therefore have a leading fractional rate difference

$$\frac{d\tau_A-d\tau_B}{dT} \simeq\frac12R_{0i0j} \left(X_A^iX_A^j-X_B^iX_B^j\right).$$

Clocks fixed away from the central line generally require support, so this is a short-time or controlled-platform statement rather than a claim that every node is freely falling. Free nodes instead develop Jacobi velocities, and Doppler terms must then be separated from the curvature-dependent clock comparison.

Shrinking $L$ sends all finite-laboratory corrections to zero, recovering local special-relativistic accuracy. It does not send $R_{0i0j}$ itself to zero. The coefficient remains measurable by dividing the diminishing relative acceleration by the diminishing baseline; this is the independent check that “locally inertial” does not mean “locally curvature-free.”

Exercises

1. Baseline bound. If $\lVert R\rVert=10^{-12}\,\mathrm{m}^{-2}$ in units with $c=1$, $a_{\rm sig}=10^{-8}\,\mathrm{m}^{-1}$, and $\varepsilon_a=10^{-3}$, what baseline satisfies the acceleration criterion?

Solution

$L\lesssim\varepsilon_a a_{\rm sig}/\lVert R\rVert=(10^{-3})(10^{-8})/10^{-12}=10\,\mathrm m$. This bound addresses only unmodeled tidal acceleration; duration, gradients, and instrument forces require separate checks.

2. Opposite clocks. Two supported clocks lie at $X^i=\pm L n^i$. Do they have a quadratic Fermi clock-rate difference?

Solution

No. The quadratic contraction $R_{0i0j}X^iX^j$ is even under $X\mapsto-X$, so the two values agree. A difference first appears through asymmetric placement, odd curvature-gradient terms, velocity, or support effects.

3. Diagnose an incomplete error budget. A design has $\lVert R\rVert L^2\ll1$ but $L\lVert\nabla R\rVert/\lVert R\rVert\sim1$. Is the constant-curvature Jacobi model controlled?

Solution

No. The metric may be close to Minkowski, yet curvature changes by order itself across the baseline. The gradient correction $\lVert\nabla R\rVert L^2$ is then comparable to the leading $\lVert R\rVert L$ differential acceleration, so curvature must be modeled as spatially varying or the laboratory shortened.

References

• 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, §1.4, pp. 4–5. PDF.