Lunar Laser Ranging

Scrollable formula. Use Left and Right arrow keys to reveal hidden content.

Scrollable table. Use Left and Right arrow keys to reveal additional columns.

Scrollable figure. Use Left and Right arrow keys to inspect hidden content.

Lunar laser ranging (LLR) times laser pulses from a terrestrial station to a corner-cube reflector on the Moon and back. Its strength is not one exceptionally precise distance: decades of ranges sample many geometries and frequencies, allowing a global dynamical fit to separate solar-directed orbital polarization, anomalous precession, geodetic rotation, and secular phase drift. Those separations remain conditional on the lunar, terrestrial, and instrumental model.

Required background. Weak-Field Gravity supplies the post-Newtonian equations of motion; Geodesics and Jacobi Fields supplies light propagation and endpoint variation. Helpful background. Probability and Statistical Inference supplies correlated fits and priors; Tests of Gravity supplies the theory-to-observable logic.

From a return pulse to a lunar range

Let $t_1$ , $t_2$ , and $t_3$ be transmission, lunar reflection, and reception epochs on declared time scales. After clock and propagation corrections, the basic two-way observable is

\rho_{\rm obs}=\frac{c}{2}(t_3-t_1),

but the computed range evaluates station and reflector positions at different epochs and includes atmosphere, gravitational light time, Earth orientation, lunar libration, and hardware delay. The factor $c/2$ converts a round-trip time into length. A normal point compresses accepted photon returns over a short interval; its epoch, time of flight, scatter, wavelength, detector information, and calibration metadata must remain traceable to the underlying session.

At the evidence cutoff 2026-08-19, the International Laser Ranging Service lists Consolidated Laser Ranging Data Format version 2 as its current exchange format. Its normal-point and calibration records are specified in ILRS 2022, CRD v2.01, §§4–5 and record 11, PDF. A reproducible analysis pins the exact archive files and checksums, format and errata version, retrieval date, station logs, reflector coordinates, terrestrial and celestial frames, Earth–Moon–Sun ephemeris, time-scale realization, and every later correction or superseding release.

Calibration covers event timers, detector and laser delay, meteorology, atmospheric mapping, station motion, Earth orientation, reflector orientation and thermal response, and wavelength-dependent offsets. Selection covers weather and return quality, target and station availability, minimum photons, time-span boundaries, eclipses, maintenance, outlier rules, and session-level rejection. Completeness therefore means completeness relative to the declared stations, reflectors, epochs, and quality rules—not every attempted pulse.

Four signatures in one dynamical solution

Write the relative Earth–Moon dynamics schematically as

\ddot{\mathbf r} =-\frac{\mu(t)}{r^3}\mathbf r +\mathbf a_{\rm Sun} +\mathbf a_{\rm tide} +\mathbf a_{\rm PN} +\mathbf a_{\rm test}, \qquad \mu(t)=\mu_0\left[1+g_\mu(t-t_0)\right],

where $\mu=G(M_\oplus+M_{\rm Moon})$ and $g_\mu=\dot\mu/\mu$ has dimensions of inverse time. Only with constant Earth and Moon masses does $g_\mu$ equal $g=\dot G/G$ . A consistent variation model also states which other masses vary and whether solar mass loss is fitted. The range model may be linearized around a reference solution,

\rho(t)=\rho_0(t;\boldsymbol\xi) +\eta\,A_\eta(t) +\alpha\,A_\alpha(t) +\delta_{\rm dS}A_{\rm dS}(t) +g A_g(t).

Here $\boldsymbol\xi$ contains initial states and nuisance parameters. The dimensionless Nordtvedt parameter $\eta$ multiplies a principally synodic, Sun-directed polarization $A_\eta\propto\cos D$ , where $D$ is lunar elongation from the Sun. Its sign depends on the declared gravitational-to-inertial-mass convention. Composition-dependent weak-equivalence effects can imitate part of this signature, so LLR alone is not a pure strong-equivalence test; Will 2014, §4.3.1 and Eqs. (66)–(67) develops that distinction.

An inverse-square modification $\alpha$ changes radial acceleration and hence mean motion, eccentricity, and apsidal harmonics; its template must be derived for a stated force law and range, not represented by one universal amplitude. The geodetic coefficient $\delta_{\rm dS}$ scales the de Sitter rotation of the lunar orbital plane and pericenter produced by the Earth–Moon system’s solar motion. A nonzero $g$ produces accumulated orbital phase and semimajor-axis changes. Tidal dissipation also creates secular and long-period evolution, making $g$ especially dependent on the lunar-interior and Earth-tide model. These signatures are distinguishable because their directions and time dependences differ, but they are not orthogonal in a finite, uneven dataset.

From unequal fall to a synodic range signal

Let $\epsilon_A=E_{g,A}/(M_Ac^2)\lt0$ be the binding-energy fraction. In the weak-self-gravity Nordtvedt parameterization, subtracting the Earth’s and Moon’s responses to the same solar field gives

\Delta\mathbf a_{\rm Moon-Earth} =\eta(\epsilon_{\rm Moon}-\epsilon_\oplus)\mathbf g_\odot,

in addition to the ordinary solar tide and any composition-dependent force. The perturbation points approximately toward the Sun while the Earth–Moon radial direction rotates. Their relative angle is the lunar elongation $D$ , so projecting the forcing into the orbit produces a synodic component. The orbital response requires the three-body dynamics: one cannot obtain its range amplitude by multiplying the acceleration by half the squared flight time of a laser pulse. The range signature and the composition degeneracy are discussed in Will 2014, §4.3.1, Eqs. (66)–(67), PDF.

The instrument and dynamics have very different scales. A mean distance of $3.84\times10^8$ m corresponds to about 2.56 s of round-trip light time; a 1 mm range change corresponds to only $2\times10^{-3}/c\simeq6.7$ ps in that time. This conversion is a timing scale, not a claim that every individual photon determines a millimetre range: reflector orientation, pulse width, atmosphere, calibration, and averaging remain essential.

Likelihood, covariance, and consistency

For residual vector $\mathbf e=\boldsymbol\rho_{\rm obs}-\boldsymbol\rho_{\rm calc}$ , a normalized Gaussian model has

-2\ln\mathcal L =\mathbf e^{\mathsf T}C^{-1}\mathbf e +\ln\det C +N\ln(2\pi).

The covariance $C$ may include session-correlated timing and atmosphere terms; calibration parameters shared across normal points belong in the model or a joint prior, not in independently inflated error bars. Priors on station coordinates, tides, lunar interior, reflector geometry, radiation pressure, and range biases must be declared. Posterior-predictive residuals should be stratified by station, reflector, lunar phase, solar elongation, epoch, wavelength, and weather. Withheld epochs and stations test extrapolation rather than goodness of fit to reused data.

The central systematic risk is parameter absorption: station motion, reflector libration, atmosphere, tides, thermal expansion, solar radiation, ephemeris covariance, weighting, or a long-period range bias can project onto a gravity template. Fit-span, nuisance-basis, and covariance changes therefore measure model dependence. Most importantly, one must integrate each modified equation and refit the full correlated solution. Reinterpreting residuals from an ephemeris constructed under general relativity is inconsistent, as emphasized in Fienga and Minazzoli 2024, introduction, §§2–3, and §5.

Documented rerun benchmark

A numerical reproduction must identify its actual software versions, floating-point mode, random seeds, and input hashes for CRD normal points, station logs, reflector coordinates, and the reference ephemeris. Integrate the dynamics together with variational derivatives of the state with respect to initial conditions and gravity parameters. Nondimensionalization prevents one absolute tolerance from treating metres and metres per second as interchangeable. Compare at least three increasingly strict integration tolerances and check the resulting range and inferred parameters, not merely the integrator’s local error estimate.

The benchmark generates synthetic normal points at the pinned real epochs, injects blinded $\eta$ and finite-range inverse-square signals, and refits states, tides, libration, station and session biases, and correlated noise. Nested 5-, 10-, and 20-year spans test long-period absorption; every fifth lunation and then one station are withheld. Two hundred fixed noise seeds test interval coverage. The finest two resolutions must shift each recovered injection by less than $0.1$ posterior standard deviation, 68% intervals must attain declared binomial coverage, and withheld standardized residuals must remain calibrated. Nonconvergence, a changed numerical branch, silent data loss, or failed coverage produces a nonzero exit status.

Record the executable rerun command with the actual implementation. Output should include fitted parameters, covariance, coverage, convergence and selection tables, residuals, and run metadata. Failure to recover a blinded signal or achieve calibrated coverage invalidates that numerical result. The procedure here is a benchmark design, not a claim that a lunar-ephemeris fit has been executed or provided by this page.

Validity boundary

LLR tests gravity in the Earth–Moon–Sun environment and over the fitted time and frequency support. It does not by itself identify a unique alternative theory, determine $G$ separately from all masses, or transfer an inverse-square constraint to a screened environment. A later data release, calibration revision, ephemeris change, or lunar-interior model that materially changes template covariance requires a new global fit.

Exercises

Suppose a simplified fit contains $\delta\rho=A\eta\cos D+Bq\cos D+Cq\cos2D$ , where $q$ is a nuisance amplitude. Explain when $\eta$ is identifiable and what additional data improve the inference.

Solution

Because the columns for $\eta$ and the first part of $q$ are proportional when the dataset samples only $\cos D$ , only the combination $A\eta+Bq$ is measured. Sampling $\cos2D$ with $C\ne0$ constrains $q$ independently and therefore breaks the exact degeneracy. Broad lunar-phase coverage, multiple reflectors and stations, and external calibration of the nuisance process further reduce covariance. More points at the same phase do not fix the structural degeneracy; they only measure the same combination more precisely.

References

Fienga, A., and Minazzoli, O., “Testing Theories of Gravity with Planetary Ephemerides,” Living Reviews in Relativity 27 (2024) article 1, DOI.
International Laser Ranging Service, Consolidated Laser Ranging Data Format, Version 2.01 (2022), official NASA/ILRS PDF.
Will, C. M., “The Confrontation between General Relativity and Experiment,” Living Reviews in Relativity 17 (2014) article 4, DOI.