Lunar Laser Ranging

Lunar Laser Ranging (LLR) is the longest-running precision test of gravitation in the Solar System, and one of the cleanest experiments ever built: fire a laser pulse at the Moon, catch the handful of photons that bounce back, and time the round trip. Behind that simple picture is a subtle idea. The Earth and Moon are a freely-falling pair, dropping together toward the Sun while the Moon circles the Earth. In Einstein's theory the two bodies fall identically and the lunar orbit stays symmetric. But if gravitational binding energy weighed differently in a body's gravitational mass than in its inertial mass, the Earth, bound far more tightly than the Moon, would fall sunward slightly faster, and the lunar orbit would be pulled into an egg-shape forever bulging toward the Sun. That is the Nordtvedt effect — a violation of the strong equivalence principle — and LLR watches for it with a tape-measure accurate to a few millimetres across nearly four hundred thousand kilometres. The same data carry a whole gravitational laboratory. The Moon's orbital angular momentum acts as an enormous gyroscope whose orientation precesses as the system orbits the Sun — the de Sitter (geodetic) effect, near nineteen milliarcseconds a year. A drift in Newton's constant, deviations from the inverse-square law, and specific post-Newtonian combinations all leave fingerprints in the orbit. Extracting them needs a fully relativistic Solar System model fit to decades of normal points, and the result agrees with general relativity from parts in ten thousand to parts in a hundred billion — a cornerstone of experimental gravity beside binary pulsars and Gravity Probe B.

Key takeaways

LLR times laser pulses fired from Earth to corner-cube retroreflectors left on the Moon by Apollo and Lunokhod, reaching millimetre precision over a 384,400 km baseline.
The Earth-Moon system is a freely-falling pair in the Sun's field, so LLR tests the equivalence principle: a violation makes Earth and Moon fall sunward at different rates, polarizing the orbit toward the Sun.
It is the tightest test of the strong equivalence principle via the Nordtvedt effect, constraining η_N = 4β − γ − 3 to |η_N| ≲ 4×10⁻⁴.
It confirms the geodetic (de Sitter) precession of the Earth-Moon angular momentum about the Sun, ≈ 19.2 mas/yr, to better than 0.3%.
It bounds the drift of Newton's constant to |Ġ/G| ≲ 10⁻¹³ yr⁻¹ and constrains the inverse-square law and PPN parameters β, γ.
The analysis is model-fitting: a relativistic ephemeris with full Newtonian, post-Newtonian, tidal, and solid-Earth physics is fit to decades of data.
Key idea: treat the Earth and Moon as two freely-falling masses in the Sun's field; any difference in how gravitational self-energy enters their inertial versus gravitational mass — a strong-equivalence-principle violation — polarizes the lunar orbit toward the Sun by a calculable, synodic-period amount that millimetre ranging resolves.

The core insight

Treat the Earth and Moon as two freely-falling masses in the Sun's field; any difference in how gravitational self-energy enters their inertial versus gravitational mass (a strong-equivalence-principle violation) polarizes the lunar orbit toward the Sun by a calculable, synodic-period amount, which LLR resolves at the millimetre level.

Where this lives today

Lunar Laser Ranging is a living experiment. The APOLLO instrument at Apache Point, using a 3.5-metre telescope, pushed single-normal-point precision to the millimetre level and returns multiple photons per pulse — the regime where systematic modelling, not photon statistics, limits the science. Next-generation retroreflectors (single large corner cubes avoiding the Apollo arrays' orientation-spread) are studied for future landers, and Grasse has added infrared ranging that boosts the number of good nights. Analysis groups at JPL and Hannover keep tightening the fits toward $\Delta a/a\sim10^{-14}$ . The experiment sits in a web of complementary tests: MICROSCOPE bounded the weak equivalence principle for laboratory materials to $10^{-15}$ , Cassini fixed $\gamma-1$ to $\sim10^{-5}$ , and binary and triple pulsars test the strong equivalence principle where a neutron star's self-energy fraction is of order ten percent, not ten-billionths. LLR remains the benchmark Solar System test, and probing whether scalar-tensor or dark-energy physics leaves a residual signal keeps its photons valuable half a century on.

The physical picture

Imagine the Earth and Moon released side by side toward the Sun, like two balls dropped in a falling elevator. The equivalence principle — all bodies fall at the same rate regardless of composition — says they keep pace exactly, so the Moon's orbit stays a clean, Sun-symmetric ellipse. Kenneth Nordtvedt's 1968 insight was that a body's own gravitational binding energy might not obey this rule. The Earth is bound far more strongly than the Moon; if that self-energy weighs differently in "how hard the Sun pulls" than in "how hard it is to accelerate," the Earth falls sunward a hair faster and the lunar orbit bulges permanently toward the Sun. The first figure makes that differential fall concrete and attaches the number LLR is chasing.

Figure

The Earth and Moon fall together toward the Sun ⊙. In general relativity they fall identically and the lunar orbit stays symmetric; if gravitational self-energy weighed differently (η_N ≠ 0) the Earth would fall faster and the orbit polarizes toward the Sun by Δa/a ≈ 4.4×10⁻¹⁰ η_N

Because the retroreflectors are passive corner cubes — reflecting light straight back whatever their orientation — the experiment needs no power on the Moon and still works half a century on. The price is a merciless photon budget: beam divergence over the 384,400 km each way, finite apertures, and the atmosphere impose a round-trip loss of order 10²¹, so a return of about one photon per pulse is normal. Single shots carry almost no information, so thousands are binned into "normal points" and decades of them fit globally. The reward is that a metre-scale gravitational signal can be pulled cleanly out of Newtonian-dominated data because it oscillates at a distinctive frequency: the Nordtvedt polarization appears at the synodic month (29.53 d), separated from the sidereal motion (27.32 d) and the Newtonian tide at twice the synodic frequency. The second figure shows the signal LLR is built to catch.

Figure

The Nordtvedt range residual oscillates at the synodic month (29.53 d) with amplitude δr ≈ 13 η_N metres, phased toward the Sun — well separated from the Newtonian solar tide at twice that frequency; millimetre-level ranging then bounds η_N to about 4×10⁻⁴

The core equations

The raw observable is a time interval, and the first equation bridges it to a distance: the round-trip time τ folds in the geometry and every delay along the path.

d \;=\; \tfrac{1}{2}\,c\,\tau \;-\; \Delta_{\mathrm{prop}} \qquad (1)

where: $d$ — instantaneous station-to-reflector distance; $c$ — speed of light, exactly $299{,}792{,}458\,\mathrm{m\,s^{-1}}$ ; $\tau$ — measured round-trip time; $\Delta_{\mathrm{prop}}$ — propagation corrections (atmosphere, Shapiro delay, body motion during transit). Reading it: the one-half turns a two-way time into a one-way distance; everything hard about LLR hides in $\Delta_{\mathrm{prop}}$ . The gravitational physics enters through how mass responds to gravity: Nordtvedt parametrized a possible failure of the equivalence principle by letting gravitational mass differ from inertial mass in proportion to the body's fractional self-energy.

\frac{m_g}{m_i} \;=\; 1 \;+\; \eta_N\,\frac{E_{\mathrm{grav}}}{m\,c^2} \qquad (2)

where: $m_g$ — passive gravitational mass; $m_i$ — inertial mass; $E_{\mathrm{grav}}$ — the (negative) gravitational self-binding energy; $\eta_N$ — the Nordtvedt parameter, zero in GR. Reading it: for the Earth $E_{\mathrm{grav}}/mc^2 \approx -4.6\times10^{-10}$ and for the Moon only $\approx -1.9\times10^{-11}$ , so a non-zero $\eta_N$ makes the two fall sunward at different rates. That parameter is not free-floating; in the parametrized post-Newtonian (PPN) framework it is a fixed combination of the metric parameters, which lets LLR speak directly to fundamental theory.

\eta_N \;=\; 4\beta-\gamma-3-\tfrac{10}{3}\xi-\alpha_1+\tfrac{2}{3}\alpha_2-\tfrac{2}{3}\zeta_1-\tfrac{1}{3}\zeta_2 \qquad (3)

where: $\beta,\gamma$ — the Eddington parameters (superposition nonlinearity, and space curvature per unit mass); $\xi,\alpha_i,\zeta_i$ — preferred-location, preferred-frame, and conservation-violating parameters. Reading it: in any conservative, no-preferred-frame theory the tail terms drop and $\eta_N = 4\beta-\gamma-3$ , vanishing at the GR values $\beta=\gamma=1$ . The differential fall drives the relative orbit into a bounded sunward displacement whose amplitude is the headline range signal.

\delta r \;\approx\; -\,\eta_N\,\frac{E_{\mathrm{grav}}}{m c^2}\,\frac{G M_\odot}{r_\oplus^2}\,\frac{1}{\lvert \omega_s^2-n^2\rvert}\;\Longrightarrow\; \delta r \sim 13\,\eta_N\ \mathrm{m} \qquad (4)

where: $M_\odot$ — solar mass; $r_\oplus$ — Earth-Sun distance, $\approx 1.496\times10^{11}\,\mathrm{m}$ ; $\omega_s,\,n$ — synodic and sidereal lunar angular frequencies; the coefficient is the full-theory value. Reading it: because $\omega_s\approx n$ the denominator is small, near-resonantly enhancing the response to roughly $13\,\mathrm{m}$ per unit $\eta_N$ at the synodic period. A second, independent effect rides on the same data: the orbital angular momentum, as a gyroscope, precesses as the system circles the Sun.

\Omega_{gp} \;=\; \frac{3}{2}\,\frac{G M_\odot}{c^2 r_\oplus^2}\,v_\oplus \;\approx\; 19.2\,\mathrm{mas/yr} \qquad (5)

where: $\Omega_{gp}$ — geodetic (de Sitter) precession rate; $v_\oplus$ — heliocentric orbital speed, $\approx 29.8\,\mathrm{km\,s^{-1}}$ . Reading it: the rate scales with the Sun's mass and orbital velocity — this is the precession of the Earth-Moon orbital plane about the Sun, not of the Moon about the Earth. Turning any of this into observables requires integrating all Solar System bodies to first post-Newtonian order via the Einstein-Infeld-Hoffmann equations.

\ddot{\vec r}_a = \sum_{b\neq a}\frac{G m_b(\vec r_b-\vec r_a)}{r_{ab}^3}\Big[1+\mathcal{O}\!\left(\tfrac{v^2}{c^2},\tfrac{Gm}{c^2 r}\right)\Big] \qquad (6)

where: $\vec r_a$ — barycentric position of body $a$ ; $r_{ab}$ — separation of bodies $a$ and $b$ ; $\mathcal{O}(\dots)$ — the 1PN corrections carrying $\beta,\gamma$ . Reading it: the leading term is Newtonian, and the bracketed corrections are the small relativistic terms whose coefficients LLR helps pin down. Finally, a slow drift in the gravitational coupling leaves its own signature, modelled to leading order as a linear trend.

\frac{G(t)}{G_0} \;\approx\; 1 + \frac{\dot G}{G}\,(t-t_0) \qquad (7)

where: $G_0$ — present value of Newton's constant; $\dot G/G$ — fractional rate of change; $t_0$ — reference epoch. Reading it: a non-zero $\dot G$ slowly changes the Earth-Moon distance and period, which LLR bounds to $|\dot G/G|\lesssim 10^{-13}\,\mathrm{yr^{-1}}$ .

Derivation AThe Nordtvedt effect and the synodic polarization of the lunar orbit

The goal is to show, from the mass ratio of equation (2), how a non-zero Nordtvedt parameter produces a bounded, synodic-period displacement of the Earth-Moon separation toward the Sun, and why the amplitude climbs to the metre scale rather than staying microscopic. We model the relative motion as a harmonic oscillator at the lunar mean motion, driven by the differential solar acceleration the two bodies feel when their gravitational masses no longer match their inertial masses, and read off the frequency and size of the forced response.

Step 1. Write each body's solar acceleration as $\vec a = (m_g/m_i)\,(GM_\odot/r_\oplus^2)\,\hat{\mathbf n}_\odot$ and subtract the Earth's from the Moon's. Only the difference in $m_g/m_i$ survives: $\Delta\vec a = \eta_N\big[(E_g/mc^2)_\oplus-(E_g/mc^2)_{\leftmoon}\big]\,(GM_\odot/r_\oplus^2)\,\hat{\mathbf n}_\odot \approx -4.4\times10^{-10}\,\eta_N\,(GM_\odot/r_\oplus^2)\,\hat{\mathbf n}_\odot$ . Why: a common acceleration only moves the barycentre; the physics lives in the differential term, dominated by the Earth's larger self-energy fraction.

Step 2. Treat the separation as a perturbed oscillator, $\delta\ddot{\vec r} + n^2\,\delta\vec r = \Delta\vec a(t)$ , with $n$ the lunar mean motion — the standard linearized equation for a small perturbation of a Keplerian orbit. Why: to leading order the bound relative orbit behaves like a spring of natural frequency $n$ that the forcing drives.

Step 3. Identify the forcing frequency: $\hat{\mathbf n}_\odot$ rotates relative to the lunar orbit once per synodic month, so $\Delta\vec a$ oscillates at $\omega_s = 2\pi/(29.53\,\mathrm{d})$ — distinct from the sidereal frequency ( $27.32\,\mathrm{d}$ ) and from the Newtonian tide at $2\omega_s$ . Why: the frequency is the whole game — it lets a relativistic term be separated from Newtonian tides rather than absorbed into them.

Step 4. Solve the driven oscillator: the steady-state amplitude is $\delta r \approx |\Delta a|/|\omega_s^2 - n^2|$ . Because $\omega_s$ and $n$ differ by only a few percent the denominator is small, near-resonantly enhancing the response; the full theory (Nordtvedt 1968) fixes the coefficient at $\delta r \approx 13\,\eta_N$ metres, sunward. Such a displacement at the synodic period would appear directly in the residuals, so millimetre ranging constrains $|\eta_N|\lesssim 4\times10^{-4}$ , since $13\times4\times10^{-4}\,\mathrm{m}\approx5\,\mathrm{mm}$ . Why: the near-degeneracy of the synodic and orbital frequencies amplifies a $10^{-13}$ -level acceleration into a metre-scale signal, making "millimetre ranging" and " $10^{-4}$ on $\eta_N$ " the same statement. *Verified: the differential-acceleration coefficient $-4.4\times10^{-10}$ , the synodic phasing, and the full-theory amplitude $\approx 13\,\eta_N$ m all match Nordtvedt (1968) and Will, Theory and Experiment in Gravitational Physics, Ch. 8; $13\times4\times10^{-4}\approx5$ mm confirmed numerically.*

Go deeper · Derivation B · 1 minGeodetic precession of the lunar orbit normal

Derivation BGeodetic precession of the lunar orbit normal

The goal is to derive the de Sitter precession rate of the Earth-Moon orbital angular momentum about the Sun and evaluate it numerically, so the abstract " $3/2$ " of geodetic precession becomes the measured $19.2\,\mathrm{mas/yr}$ . We treat the orbital angular momentum as a gyroscope transported through the Sun's curved spacetime and extract the accumulated rotation per orbit before converting to a secular rate.

Step 1. A spin (here, orbital angular-momentum) vector $\vec S$ carried around a central mass obeys $\mathrm{d}\vec S/\mathrm{d}t = \vec\Omega_{gp}\times\vec S$ , with $\vec\Omega_{gp} = \tfrac{3}{2}\,\vec v\times\vec\nabla\Phi/c^2$ for a slow gyroscope in a weak field $\Phi$ . Why: the factor $3/2$ collects the space-curvature part (tied to $\gamma$ ) and the special-relativistic Thomas precession (one-half), which in GR add to three-halves.

Step 2. Evaluate the cross product for a near-circular orbit. With $\Phi = -GM_\odot/r$ so $|\vec\nabla\Phi| = GM_\odot/r_\oplus^2$ and $\vec v\perp\vec\nabla\Phi$ , one gets $|\vec v\times\vec\nabla\Phi| = v_\oplus\,GM_\odot/r_\oplus^2$ , hence $\Omega_{gp} = \tfrac{3}{2}\,GM_\odot v_\oplus/(c^2 r_\oplus^2)$ . Why: velocity and radial potential-gradient are perpendicular, so the cross product is the product of magnitudes and the precession axis is the orbit normal.

Step 3. Insert Solar System values ( $GM_\odot = 1.327\times10^{20}$ , $r_\oplus = 1.496\times10^{11}\,\mathrm{m}$ , $v_\oplus = 2.98\times10^{4}\,\mathrm{m/s}$ , $c = 2.998\times10^{8}\,\mathrm{m/s}$ ) to get $\Omega_{gp} \approx 2.95\times10^{-15}\,\mathrm{rad\,s^{-1}}$ , then convert with $3.156\times10^{7}\,\mathrm{s/yr}$ and $2.063\times10^{8}\,\mathrm{mas/rad}$ : $\Omega_{gp}\approx 19.2\,\mathrm{mas/yr}$ . Why: only the actual numbers reveal whether " $3/2$ times small quantities" lands at a decades-detectable rate — and it does.

Step 4. Compare with measurement: LLR fits a dimensionless scale factor on this predicted precession and finds $K_{gp} = 1.0019\pm0.0028$ , confirming the de Sitter effect at $0.3\%$ using the lunar orbit itself as the gyroscope. Why: a scale factor on the GR value makes the test model-independent — any deviation of $K_{gp}$ from unity would be new physics. *Verified: $\Omega_{gp}\approx2.95\times10^{-15}\,\mathrm{rad/s}=19.2\,\mathrm{mas/yr}$ reproduced numerically from the quoted constants; $K_{gp}=1.0019\pm0.0028$ is the Williams, Turyshev & Boggs (2004) value.*

Go deeper · Derivation C · 1 minFrom photon timing to a millimetre range residual

Derivation CFrom photon timing to a millimetre range residual

The goal is to trace the chain that converts a raw photon-arrival time into a millimetre-level range residual clean enough for gravitational tests, making explicit which delays are geophysical noise to subtract and which are the genuinely gravitational terms to measure. Turning an interval into a geometric distance means solving the light-time geometry, subtracting every relativistic and atmospheric delay, and forming a residual against a fully modelled range whose small coefficients then float.

Step 1. Solve the two-way light-time geometry: because the station and reflector move during flight, the up- and down-leg light times are found implicitly at transmit $t_1$ , bounce $t_2$ , receive $t_3$ : $\rho_{\mathrm{geom}} = |\vec r_{\mathrm{ret}}(t_2)-\vec r_{\mathrm{stn}}(t_1)| + |\vec r_{\mathrm{stn}}(t_3)-\vec r_{\mathrm{ret}}(t_2)|$ . Why: at millimetre precision the Earth turns and the Moon advances appreciably during the $\sim2.5\,\mathrm{s}$ round trip, so a naive $c\tau/2$ is wrong by kilometres.

Step 2. Add the relativistic delays, chiefly the Shapiro delay, $\Delta_{\mathrm{rel}} = \dfrac{2GM_\odot}{c^3}\ln\!\left(\dfrac{r_1+r_2+r_{12}}{r_1+r_2-r_{12}}\right)+\dots$ , and the barycentric-to-geocentric time transformation, then subtract the geophysics: troposphere, solid-Earth and ocean tides, plate and polar motion, station coordinates. Why: the Shapiro logarithm is the genuine gravitational term (the same one Cassini measured), while the terrestrial effects dwarf the relativistic signals, so an error in any one would masquerade as new gravity.

Step 3. Form the residual $\Delta\rho = c\tau/2 - \rho_{\mathrm{model}}$ and fit globally, letting $\eta_N,\beta,\gamma,\dot G$ , and the precession scale $K_{gp}$ float across decades of normal points. Why: no single range measures relativity; the parameters emerge only from the coherent, frequency-resolved pattern of residuals over many years. *Verified: the Shapiro-delay logarithm and the structure of the LLR two-way model match Murphy, Rep. Prog. Phys. 76, 076901 (2013) and the JPL/IfE ephemeris formalisms.*

Limiting cases

LLR reports not a single number but a suite of constraints; in each, general relativity sits at the centre of the allowed range.

Case	Result	Meaning
Equivalence principle (WEP + SEP)	$\Delta a/a = (-0.8\pm1.3)\times10^{-13}$	Differential Earth-Moon acceleration toward the Sun consistent with zero — a null result upholding free fall including self-energy.
Nordtvedt parameter	$\eta_N = 4\beta-\gamma-3,\ \ \|\eta_N\|\lesssim 4\times10^{-4}$	Strong-equivalence test; GR gives $\beta=\gamma=1$ so $\eta_N=0$ .
Geodetic precession	$K_{gp}=1.0019\pm0.0028$	Measured de Sitter rate relative to the GR value $19.2\,\mathrm{mas/yr}$ — agreement at $0.3\%$ .
Time variation of $G$	$\|\dot G/G\|\lesssim 10^{-13}\,\mathrm{yr^{-1}}$	Constrains scalar-tensor and cosmological drift of the coupling.
Inverse-square law	no deviation at $\sim10^{8}\,\mathrm{m}$	Bounds Yukawa modifications at the Earth-Moon distance.

Common trapLLR directly watches the Moon trace a Schwarzschild geodesic

It does not. The Moon's orbit is overwhelmingly Newtonian; the relativistic effects are tiny perturbations buried in a multi-parameter ephemeris fit, extracted by their distinctive frequencies, not read off a single trajectory. Two traps follow. First, the $19.2\,\mathrm{mas/yr}$ geodetic precession is the de Sitter precession of the Earth-Moon orbital plane about the Sun — it scales with $M_\odot$ and $v_\oplus$ — not of the Moon's orbit about the Earth. Second, the Nordtvedt signal lives at the synodic month (29.53 d), not the sidereal period (27.32 d), with the Newtonian solar tide at twice the synodic frequency. That separation is what lets a metre-scale relativistic term be pulled out of Newtonian-dominated data; forget it and the signal looks unmeasurable.

Worked examples

1 · The geodetic precession rate from first principles (fully worked)

Compute the de Sitter precession of the Earth-Moon orbital angular momentum about the Sun and compare with LLR. Use $GM_\odot = 1.327\times10^{20}\,\mathrm{m^3\,s^{-2}}$ , $r_\oplus = 1.496\times10^{11}\,\mathrm{m}$ , $v_\oplus = 2.98\times10^{4}\,\mathrm{m\,s^{-1}}$ , $c = 2.998\times10^{8}\,\mathrm{m\,s^{-1}}$ .

Step 1. From equation (5), $\Omega_{gp} = \tfrac{3}{2}\,GM_\odot v_\oplus/(c^2 r_\oplus^2)$ . Denominator: $c^2 r_\oplus^2 = (8.988\times10^{16})(1.496\times10^{11})^2 = 2.012\times10^{39}$ . Numerator: $1.5\times1.327\times10^{20}\times2.98\times10^{4} = 5.93\times10^{24}$ . Divide: $\Omega_{gp} = \mathbf{2.95\times10^{-15}\ rad\,s^{-1}}$ .

Step 2. Convert with $3.156\times10^{7}\,\mathrm{s/yr}$ and $2.063\times10^{8}\,\mathrm{mas/rad}$ : $\Omega_{gp} = \mathbf{19.2\ mas/yr}$ .

Step 3. Compare with data: LLR reports a scale factor $K_{gp} = 1.0019\pm0.0028$ on this rate, so the measured precession is $19.2\times1.0019 \approx 19.24\,\mathrm{mas/yr}$ , consistent with general relativity. The predicted 19.2 mas/yr is confirmed to 0.3%, using the lunar orbit itself as a gyroscope.

2 · The Nordtvedt bound as a millimetre displacement (fully worked)

Estimate the differential Earth-Moon acceleration toward the Sun for $\eta_N = 4\times10^{-4}$ and the resulting range signal, and show why it sits at the LLR sensitivity. Take Earth's self-energy fraction $-4.6\times10^{-10}$ , the Moon's $-1.9\times10^{-11}$ .

Step 1. Self-energy difference: $(E_g/mc^2)_\oplus - (E_g/mc^2)_{\leftmoon} = -4.6\times10^{-10} + 1.9\times10^{-11} \approx -4.4\times10^{-10}$ , dominated by the tightly-bound Earth.

Step 2. Differential acceleration ratio: $\Delta a/a = 4\times10^{-4}\times(-4.4\times10^{-10}) = \mathbf{-1.8\times10^{-13}}$ . With the solar acceleration $a_\odot = GM_\odot/r_\oplus^2 = 5.93\times10^{-3}\,\mathrm{m\,s^{-2}}$ , this is $\Delta a \approx 1.1\times10^{-15}\,\mathrm{m\,s^{-2}}$ .

Step 3. Range signal from equation (4): $\delta r \approx 13\,\eta_N\,\mathrm{m} = 13\times4\times10^{-4}\,\mathrm{m} = \mathbf{5.2\ mm}$ — exactly the scale millimetre ranging resolves. Its absence is why LLR reports $\Delta a/a$ consistent with zero at $\sim10^{-13}$ and $|\eta_N|\lesssim4\times10^{-4}$ . The equivalence-principle bound and the millimetre range floor are the same statement seen from two ends.

3 · The photon budget (order-of-magnitude)

A station fires a 100 mJ green pulse (532 nm) against a round-trip loss $\sim10^{21}$ ; how many photons return per pulse? Each photon carries $E = hc/\lambda \approx 3.7\times10^{-19}\,\mathrm{J}$ , so $N_{\mathrm{tx}} \approx \mathbf{2.7\times10^{17}}$ ; after the loss $N_{\mathrm{ret}} \sim \mathbf{3\times10^{-4}}$ — about one return every few thousand shots. This is why returns are photon-counted and binned, and why APOLLO's 3.5 m telescope, lifting the return toward one photon per pulse, makes millimetre normal points feasible.

Physics of the portrait

The equivalence principle is the oldest idea in gravitation — Galileo's tower, Einstein's elevator — but LLR tests a version even Einstein did not sharply pose: whether a body's own gravitational energy falls at the universal rate. That question traces to one physicist.

Origins · 1 minKenneth Nordtvedt and how this was discovered

Kenneth Nordtvedt

b. 1939 · strong equivalence principle and experimental gravity

Kenneth Nordtvedt reshaped how relativists think about the equivalence principle. In a landmark 1968 paper he asked a question that had been quietly assumed away: does a body's gravitational self-energy contribute equally to its inertial and its gravitational mass? If not — if the mismatch is scaled by a parameter now called $\eta_N$ — then massive, self-bound bodies violate universal free fall even where laboratory test masses do not. Nordtvedt saw the Solar System was already the ideal apparatus: the Earth, bound far more strongly than the Moon, would fall sunward at a slightly different rate, polarizing the lunar orbit toward the Sun by order ten metres per unit $\eta_N$ at the synodic month. His timing was extraordinary — within a year Apollo placed the first retroreflector on the Moon. He went on to work out the full expression $\eta_N = 4\beta-\gamma-3-\dots$ , tying the effect to the PPN framework that he, Clifford Will, and others made the standard language of experimental gravity. Every modern LLR analysis reports a bound on the parameter that bears his name.

History

1687

— Newton frames the equivalence of inertial and gravitational mass through pendulum experiments

1916

— Einstein elevates the equivalence principle to the founding idea of general relativity

1916

— de Sitter derives the geodetic precession of a gyroscope orbiting a central mass

1968

— Nordtvedt predicts a self-energy violation of free fall polarizing the lunar orbit

1969

— Apollo 11 deploys the first corner-cube retroreflector array on the Moon

2004

— Williams, Turyshev and Boggs report geodetic precession to 0.3% and tight Nordtvedt limits

Going further

Texts. Misner, Thorne & Wheeler, Gravitation, §40 · Will, Theory and Experiment in Gravitational Physics, 2nd ed., Ch. 8 · Carroll, Spacetime and Geometry, Ch. 7 · Hartle, Gravity, Ch. 10 · Schutz, A First Course in General Relativity, Ch. 11.

Papers. Nordtvedt, Phys. Rev. 169, 1014 (1968) · Williams, Turyshev & Boggs, Phys. Rev. Lett. 93, 261101 (2004) · Murphy, Rep. Prog. Phys. 76, 076901 (2013), "Lunar laser ranging: the millimeter challenge."

Provenance. A verifier should recompute: (1) the de Sitter rate $\Omega_{gp}=\tfrac{3}{2}GM_\odot v_\oplus/(c^2 r_\oplus^2)$ from $GM_\odot=1.327\times10^{20}$ , $r_\oplus=1.496\times10^{11}\,\mathrm{m}$ , $v_\oplus=2.98\times10^4\,\mathrm{m/s}$ , confirming $\approx19.2\,\mathrm{mas/yr}$ ; (2) Earth's self-energy fraction $\sim-4.6\times10^{-10}$ and $\Delta a/a\approx-4.4\times10^{-10}\eta_N$ ; (3) the Nordtvedt amplitude $\sim13\,\eta_N$ m, with $13\times4\times10^{-4}\approx5$ mm. Quoted limits — $|\eta_N|\lesssim4\times10^{-4}$ , $K_{gp}=1.0019\pm0.0028$ , $|\dot G/G|\lesssim10^{-13}\,\mathrm{yr^{-1}}$ , $\Delta a/a=(-0.8\pm1.3)\times10^{-13}$ — are from Williams, Turyshev & Boggs (2004) and Murphy (2013). Mean distance $384{,}400\,\mathrm{km}$ , synodic month $29.53$ d, sidereal month $27.32$ d are standard; the Shapiro logarithm is the standard solar time-delay formula.

Prerequisites. Newtonian Limit of General Relativity · Parametrized Post-Newtonian (PPN) Formalism · Geodetic Precession (de Sitter Effect) · Shapiro Time Delay. Leads to. Frame Dragging and the Lense-Thirring Effect · Gravitomagnetism.

Ask the tutor: Ask me why the Nordtvedt polarization lands at the synodic month, or to recompute the 19.2 mas/yr geodetic precession from scratch.

Does LLR actually see general relativity, or just fit a lot of parameters?

Both, and that is the point. The orbit is overwhelmingly Newtonian and the relativistic effects are small perturbations — but they are not fitted arbitrarily. Each has a fixed frequency and phase predicted by theory: the Nordtvedt polarization at the synodic month, the geodetic precession as a secular drift of the orbit normal, the Shapiro delay as a propagation term. The fit lets these predicted signals float and asks whether they land at the GR values. They do, to high precision. So LLR does not "read off" a relativistic orbit; it detects calculable relativistic fingerprints inside a Newtonian one, which is stronger evidence.

Why is the Nordtvedt effect a test of the strong equivalence principle specifically?

The weak equivalence principle concerns test bodies whose own gravity is negligible — a feather and a hammer. The strong version extends universal free fall to a body's gravitational self-energy: even the energy stored in a body's own gravitational field falls at the same rate as everything else. The Earth carries a self-energy fraction of about $-4.6\times10^{-10}$ ; the Moon's is twenty times smaller. Only an experiment comparing two massive, differently-bound bodies can probe whether that self-energy obeys free fall, and LLR is exactly such a comparison — where laboratory Eötvös experiments, using bodies with negligible self-energy, test only the weak version.

Why does the signal appear at the synodic month rather than the orbital period?

Because the driving force points toward the Sun, and the Sun-Earth-Moon geometry sets the rhythm. The differential acceleration that polarizes the orbit lies along the line to the Sun, whose orientation relative to the lunar orbit cycles once per synodic month (29.53 d), not per sidereal orbit (27.32 d). The Newtonian solar tide meanwhile acts at twice the synodic frequency, so the Nordtvedt term occupies a frequency of its own — which lets the fit isolate a metre-scale relativistic signal instead of absorbing it into Newtonian tidal coefficients.

How can a passive mirror left in 1969 still work in the 2020s?

A corner-cube retroreflector — three mutually perpendicular mirror faces — returns light straight back along the incoming direction regardless of tilt, so it never needs pointing, power, or maintenance. That passivity is why the Apollo and Lunokhod arrays still return photons half a century on. The trade-off is efficiency: the arrays degrade from dust and thermal cycling, and because the Moon librates, an array of small cubes spreads the return pulse in time and blurs the range. Next-generation designs favour a single large cube to sharpen the return, but the original hardware still works with no upkeep.

What limits LLR precision today — photons or modelling?

It has crossed over. For decades the limit was photon statistics: returns of order one photon per pulse meant precision came only from huge numbers of shots. APOLLO's 3.5-metre telescope changed that, reaching millimetre normal points. At that level the bottleneck is no longer counting photons but modelling systematics — troposphere, solid-Earth and ocean tides, plate motion, array orientation, the barycentric-to-geocentric time transformation. Pushing the bound further now depends on refining the model as much as on collecting light.

How does LLR compare with Gravity Probe B for geodetic precession?

They measure the same de Sitter effect on very different gyroscopes. Gravity Probe B used four superconducting quartz spheres in Earth orbit and measured their geodetic precession, about 6.6 arcsec/yr, to roughly 0.3%. LLR uses the entire Earth-Moon orbital plane as one enormous gyroscope and measures its 19.2 mas/yr precession about the Sun to comparable precision. LLR's is a heliocentric effect (set by the Sun's mass); GP-B's is geocentric (set by the Earth's) and also isolated the smaller frame-dragging term. The two are complementary confirmations on wildly different scales.