Perihelion Advance and Orbital Dynamics

Perihelion advance is not read directly from a planet’s orbit. Modern tests fit range, Doppler, and astrometric observations with a many-body ephemeris in which relativistic dynamics, solar shape, asteroid perturbations, initial conditions, reference frames, and instrument calibrations are correlated. A gravity conclusion is justified only after the tested modification is included and the entire ephemeris is refitted.

Required background. Weak-Field Gravity supplies post-Newtonian metric potentials; Geodesics and Jacobi Fields supplies orbital deviation and invariant trajectory reasoning. Helpful background. Probability and Statistical Inference supplies covariance and model comparison; Tests of Gravity supplies the theory–observable–measurement chain.

Relativistic apsidal advance

For a test body with semimajor axis $a$ and eccentricity $e$ orbiting a spherical mass $M$, general relativity gives the leading advance per orbit

$$\Delta\varpi_{\rm GR} = \frac{6\pi GM}{a(1-e^2)}.$$

The numerator and denominator both have length dimensions in units with $c=1$, so the result is an angle. It vanishes in the Newtonian weak-field limit $GM/a\to0$ and grows for eccentric orbits through the semilatus rectum $p=a(1-e^2)$. In the standard PPN reduction,

$$\Delta\varpi_{\rm PPN} = \Delta\varpi_{\rm GR} \frac{2+2\gamma-\beta}{3},$$

before solar oblateness and other perturbations are added. Setting $\beta=\gamma=1$ recovers the GR coefficient. The derivation and its solar-quadrupole degeneracy are reviewed in Will 2014, §4.2 and Eqs. (63)–(65).

The formula is a diagnostic, not the modern observable. Planets exchange perturbations with other planets, asteroids, trans-Neptunian objects, and the nonspherical Sun. The solar quadrupole $J_{2\odot}$ also drives an apsidal precession with an inclination dependence, while uncertain asteroid masses generate long-period signatures. Spacecraft range data constrain the planet’s orbit through an orbit-determination model whose maneuvers, nongravitational forces, station delays, and media corrections are part of the measurement.

The size of Mercury’s relativistic advance

Restore $c$ by replacing $GM$ in the numerator by $GM/c^2$. Using illustrative rounded orbital values $GM_\odot/c^2=1.477$ km, $a=5.79\times10^7$ km, and $e=0.2056$ gives

$$\Delta\varpi_{\rm GR}\simeq5.02\times10^{-7}\ {\rm rad/orbit} \simeq0.1035\ {\rm arcsec/orbit}.$$

With an 87.97-day orbital period, about 415 orbits occur in a Julian century, giving approximately 43 arcseconds per century. This is the relativistic contribution calculated from the ideal formula, not Mercury’s total observed precession: Newtonian planetary perturbations and the reference-frame definition also contribute. It illustrates why a recognizable secular effect still needs a global ephemeris fit.

Build and refit the ephemeris

The evidence basis is current through 2026-08-19. A result must identify its ephemeris family and version, observation archive and calibration, time scale, barycentric reference frame, dynamical model, solve-for parameters, selection, covariance, and software. DE440/441 provide one documented example: they numerically integrate planetary and lunar dynamics and fit ground and spacecraft observations, with ICRF3 orientation and TDB time handling described in Park et al. 2021, §§1–2 and Eqs. (1)–(7).

For observations $\mathbf y$ and parameters $\boldsymbol\theta$, linearization about a reference solution gives

$$\mathbf y-\mathbf h(\boldsymbol\theta_0) \simeq H\,\delta\boldsymbol\theta+\boldsymbol\epsilon, \qquad \widehat{\delta\boldsymbol\theta} = \left(H^{\mathsf T}C^{-1}H+P\right)^{-1} \left(H^{\mathsf T}C^{-1}\mathbf r+P\delta\boldsymbol\theta_{\rm aux}\right).$$

Here $\mathbf r=\mathbf y-\mathbf h(\boldsymbol\theta_0)$. The design matrix $H$ contains derivatives with respect to planetary initial states, masses, asteroid parameters, $J_{2\odot}$, station and spacecraft biases, and the tested gravity parameter. $P$ is the precision matrix of Gaussian auxiliary information centered at $\delta\boldsymbol\theta_{\rm aux}$ relative to the linearization point. The auxiliary term on the right vanishes only when that center is zero. The solution follows by differentiating the sum of the data and auxiliary quadratic forms. Large off-diagonal elements in the inverse normal matrix reveal why an apparent anomalous precession can be absorbed by other parameters.

Selection must list tracking arcs, astrometric epochs, station and spacecraft quality flags, solar-elongation cuts, maneuvers, and outlier rules. Calibration includes clocks and time conversion, antenna and transponder delay, station coordinates, orbit reconstruction, media corrections, and catalog frame. The likelihood should retain correlated errors by instrument and epoch rather than assigning every normal point an independent variance. Withheld arcs and alternative asteroid or solar-$J_2$ models provide posterior-predictive checks.

Why a residual is not a bound

There are two consistent procedures. One may introduce the alternative acceleration directly, integrate it, and fit the modified ephemeris to the same observations. Or one may fit a phenomenological supplementary precession simultaneously with the other solve-for parameters and interpret it only within that parameterization. In either case, all correlated ephemeris parameters must adjust.

It is inconsistent to compute an alternative precession, compare it with a residual or uncertainty obtained from a GR-built ephemeris, and call the comparison an exclusion. The masses and initial conditions that minimized the GR residuals need not minimize the alternative-model residuals. Fienga and Minazzoli 2024, §§5.1–5.4, especially Table 10 shows how derived supplementary advances and unfitted residuals can give misleading constraints. The proper comparison is between fully adjusted models using a stated statistic and common data.

Reproduction should recover the GR advance from synthetic range data, then inject a shift in the PPN combination, refit every ephemeris parameter, and verify recovery on withheld epochs. Repeat with alternate asteroid populations and $J_{2\odot}$ treatments. A signal that exists only when those nuisance parameters are frozen is evidence of parameter absorption, not a new orbital law.

Exercises

Suppose a fit holds $\gamma=1$ and reports a constraint on $\beta$ from Mercury’s perihelion. Explain why importing a different $\gamma$ constraint afterward does not reproduce a simultaneous two-parameter fit.

Solution

The perihelion observable depends on the combination $2+2\gamma-\beta$, and both parameters also correlate with the solar quadrupole, masses, and initial conditions. Fixing $\gamma$ removes one direction from the design matrix and lets the remaining parameters adjust under that condition. Substituting an external $\gamma$ value afterward leaves those adjustments and covariances unchanged, so it cannot reconstruct the joint likelihood. A consistent analysis includes the external information as an auxiliary likelihood or prior while refitting $\beta$, $\gamma$, $J_{2\odot}$, and the orbital parameters together.

References

• Fienga, A., and Minazzoli, O., “Testing Theories of Gravity with Planetary Ephemerides,” Living Reviews in Relativity 27 (2024) article 1, DOI.
• Park, R. S., Folkner, W. M., Williams, J. G., and Boggs, D. H., “The JPL Planetary and Lunar Ephemerides DE440 and DE441,” Astronomical Journal 161 (2021) 105, DOI, official JPL PDF.
• Will, C. M., “The Confrontation between General Relativity and Experiment,” Living Reviews in Relativity 17 (2014) article 4, DOI.