13.01.01 · gr-cosmology / equivalence

The equivalence principle

draft3 tiersLean: nonepending prereqs

Anchor (Master): Will, Theory and Experiment in Gravitational Physics, 2e (2018); Norton, 'What was Einstein's Principle of Equivalence?' (1985)

Intuition [Beginner]

Drop a hammer and a feather in a vacuum. They hit the ground at the same time. This is not a coincidence -- it is a deep fact about the nature of gravity, and it is the starting point for general relativity.

Newton explained this by declaring that gravitational mass (the $m$ in $F = GM m / r^{2}$ ) and inertial mass (the $m$ in $F = ma$ ) are exactly the same number. He had no explanation for why. He just observed that experiments kept confirming it.

Einstein's insight, which he called "the happiest thought of my life," was to take this equality seriously as a clue about what gravity is. He imagined himself in a freely falling elevator. If the elevator cables snap and the cabin plummets, everything inside floats. A released pencil drifts. A scale reads zero. From inside the cabin, gravity has vanished.

Now run the experiment in reverse. Put an elevator in deep space, far from any planet, and attach a rocket to it that accelerates upward at $9.8 m/s^{2}$ . Stand inside. Your feet press against the floor. A released pencil falls. A scale reads your weight. From inside the cabin, gravity has appeared -- even though there is no mass anywhere nearby.

The equivalence principle says these two situations are physically identical. No experiment performed entirely inside a small enough, sealed elevator can distinguish "standing on Earth" from "accelerating in a rocket." Gravity is not a force in the Newtonian sense. It is the name we give to the fact that reference frames can accelerate.

This has a striking consequence. A light ray crossing the elevator from one wall to the other must bend downward -- because the elevator floor accelerates upward to meet it. If the EP is correct, gravity must bend light. Einstein predicted this in 1907; Eddington confirmed it in 1919. The bending of starlight by the Sun was the first experimental triumph of general relativity.

Visual [Beginner]

The two panels are indistinguishable to anyone inside the elevator. The light ray bends by the same angle in both cases. The EP asserts that no local measurement can tell them apart. Only tidal forces -- the slight difference in gravitational pull between the floor and ceiling -- break the illusion, and only over distances large enough for those differences to matter.

Worked example [Beginner]

A light ray enters a lab-sized elevator from the left wall, traveling horizontally. The elevator accelerates upward at $g = 9.8 m/s^{2}$ . The elevator is $L = 3 m$ wide. How much does the light ray deflect as it crosses?

The light takes time $t = L / c = 3/ (3 \times 1 0^{8}) = 1 0^{- 8} s$ to cross the elevator. In this time, the elevator floor moves upward by

Δ y = \frac{1}{2} g t^{2} = \frac{1}{2} (9.8) (1 0^{- 8})^{2} \approx 5 \times 1 0^{- 16} m .

This is about one hundredth the diameter of a proton. The deflection angle is

θ \approx \frac{g L}{c ^{2}} = \frac{9.8 \times 3}{( 3 \times 1 0 ^{8} ) ^{2}} \approx 3.3 \times 1 0^{- 16} rad .

Far too small to measure in a lab. But the Sun's gravitational field acting on starlight over a path length comparable to the solar radius produces a deflection of about $1.75$ arcseconds -- measurable with a telescope during a solar eclipse. This is exactly what Eddington's 1919 expedition observed.

Check your understanding [Beginner]

Exercise (easy, true/false).

The equivalence principle says that a gravitational field and an accelerating reference frame are physically identical even over large regions of space.

Hint

Think about tidal forces. Does gravity pull equally on the floor and ceiling of a tall building?

Answer

False.

The EP holds only locally -- in a small enough region that tidal effects (differences in gravitational strength between nearby points) can be neglected. Over a large region, tidal forces distinguish a real gravitational field from uniform acceleration. A freely falling elevator near Earth has slightly stronger gravity at the floor than at the ceiling; two freely falling particles released side by side will drift closer together as they fall toward Earth's center. These tidal effects are the hallmark of genuine spacetime curvature and cannot be eliminated by any choice of reference frame.

Formal definition [Intermediate+]

The equivalence principle comes in three levels of strength. Each makes a precise physical claim and each has distinct experimental consequences.

The weak equivalence principle (WEP). The trajectory of a freely falling test body in a gravitational field is independent of its internal structure and composition. Formally: the ratio $m_{grav} / m_{inertial}$ is the same for all bodies. This is the universality of free fall (UFF), tested since Galileo and measured with increasing precision by Eotvos, Dicke, the Eot-Wash group, and the MICROSCOPE satellite mission.

In the parametrized post-Newtonian (PPN) formalism 13.03.01 pending, violations of the WEP are quantified by the Eotvos parameter $η$ , defined so that $η = 0$ if the WEP holds exactly. Current bounds are $∣ η ∣ < 1 0^{- 15}$ (MICROSCOPE, 2022).

The Einstein equivalence principle (EEP). The WEP holds, and in addition: (i) the outcome of any local non-gravitational experiment is independent of the velocity of the freely-falling reference frame in which it is performed (local Lorentz invariance, LLI); (ii) the outcome of any local non-gravitational experiment is independent of where and when in the universe it is performed (local position invariance, LPI).

The EEP is stronger than the WEP because it asserts that all of local non-gravitational physics -- not just the trajectories of test masses, but spectroscopy, clock rates, nuclear decays, electromagnetic measurements -- is the same in a freely falling frame as in an accelerating frame. A violation of LPI would appear, for example, as a spatial or temporal variation of a fundamental "constant" such as the fine-structure constant $α$ .

The strong equivalence principle (SEP). Extends the EEP to include gravitational experiments -- specifically, the internal gravitational binding energy of a body contributes equally to its inertial and gravitational mass. This is a claim about self-gravitating bodies: the Earth-Moon system, neutron stars, black holes. The SEP is tested by lunar laser ranging (the Nordtvedt effect) and by comparing the accelerations of Earth and Moon toward the Sun.

The SEP implies that even a massive, self-gravitating body falls on the same geodesic as a test particle. Violations of the SEP are parametrized by the Nordtvedt parameter $η_{N}$ ; current bounds give $∣ η_{N} ∣ < 1 0^{- 4}$ .

From the EEP to spacetime geometry

The central theorem connecting the EEP to differential geometry is due to works by Thorne, Lee, and Lightman (1973), and refined by Norton (1985) and Will (2018):

Theorem. If the Einstein equivalence principle holds, then spacetime is a pseudo-Riemannian manifold with metric signature $(-, +, +, +)$ , and the equations of motion for freely falling test particles are geodesics of this metric. Furthermore, non-gravitational physics in a local freely falling frame reduces to special relativity.

The logic is as follows. The WEP guarantees the existence of freely falling trajectories along which test particles of all compositions move identically. The EEP adds that local non-gravitational physics in a freely falling frame is special relativity 10.05.01 pending. Special relativity is the physics of Minkowski spacetime, which is a flat Lorentzian manifold. A manifold that is locally Minkowskian at every point is, by definition, a Lorentzian (pseudo-Riemannian) manifold. Test-particle trajectories that reduce to straight lines in each local inertial frame are geodesics of the global metric.

The EEP does not determine the field equations -- it tells you spacetime is curved, but not how matter tells spacetime how to curve. That requires the Einstein field equations 13.03.01 pending, which are a separate physical postulate.

Counterexamples and common slips

The EP is not the statement that "gravity is curvature." Curvature is a geometric property of the manifold; the EP is the physical postulate that links the geometry to gravitational observations. One can have curved Lorentzian manifolds with no matter at all (vacuum solutions) and one can imagine violations of the EP in which gravity is still associated with curvature but different materials fall on different curves.
The EP holds only locally -- in a region small enough that tidal forces are negligible. A gravitational field is equivalent to acceleration in a single tangent space, not across a finite patch of the manifold. The curvature tensor 13.02.01 measures the failure of the EP over finite regions.
The EP does not say that gravity is "not a force." It says that for a single point particle in free fall, the local physics is the same as in an accelerating frame. Standing on the floor of a rocket is indistinguishable from standing on the surface of a planet only if you do not look outside. Gravitational tidal effects, gravitational waves, and cosmological expansion are all real physical phenomena that are not "equivalent to acceleration" in any meaningful sense.

Key theorem with proof [Intermediate+]

Theorem (EEP implies Lorentzian manifold structure). Assume the Einstein equivalence principle. Then there exists a four-dimensional manifold $M$ equipped with a metric $g$ of signature $(-, +, +, +)$ such that:

(i) Freely falling test particles follow geodesics of $g$ . (ii) At each event $p \in M$ , there exist local coordinates $(t, x^{1}, x^{2}, x^{3})$ in which $g_{μν} (p) = η_{μν}$ (Minkowski metric) and $Γ_{μν}^{λ} (p) = 0$ (local inertial frame). (iii) Non-gravitational laws of physics in these local coordinates reduce to those of special relativity.

Proof. We construct the manifold and metric from the physical data guaranteed by the EEP.

Step 1: Events and freely falling frames. Label spacetime events by the set $M$ of all points in space and time. The WEP guarantees that for every event $p \in M$ , there exist freely falling trajectories through $p$ parametrized by proper time, and these trajectories are independent of the test body's composition.

Step 2: Local Minkowski structure. By the EEP, a small enough laboratory centered at $p$ , in free fall, can be equipped with clocks and rulers whose readings are related by the Lorentz transformations of special relativity 10.05.01 pending. This assigns to $p$ a local inertial frame with coordinates $ξ^{α}$ in which the metric is $η_{α β} = diag (- 1, + 1, + 1, + 1)$ .

Step 3: Patching into a global manifold. The freely falling frames at different events are related by smooth coordinate transformations (because physical trajectories depend smoothly on initial conditions). Two overlapping local frames with coordinates $ξ^{α}$ and $ξ^{' α}$ are related by a smooth map. This smooth atlas makes $M$ a four-dimensional smooth manifold.

Step 4: The metric. In any local inertial frame at $p$ , the spacetime interval is $d s^{2} = η_{α β} d ξ^{α} d ξ^{β}$ . In a general coordinate system $x^{μ}$ (e.g., a frame fixed to the surface of a gravitating body), the interval becomes $d s^{2} = g_{μν} (x) d x^{μ} d x^{ν}$ where $g_{μν} = η_{α β} (\partial ξ^{α} / \partial x^{μ}) (\partial ξ^{β} / \partial x^{ν})$ . The matrix $g_{μν}$ is smooth, symmetric, and has signature $(-, +, +, +)$ because the Lorentz-signature property is preserved under smooth, non-degenerate coordinate changes.

Step 5: Geodesic motion. A freely falling test particle at $p$ moves in a straight line in the local inertial frame: $d^{2} ξ^{α} / d τ^{2} = 0$ . Transforming to general coordinates gives

\frac{d ^{2} x ^{λ}}{d τ ^{2}} + Γ_{μν}^{λ} \frac{d x ^{μ}}{d τ} \frac{d x ^{ν}}{d τ} = 0,

which is the geodesic equation with $Γ_{μν}^{λ}$ the Christoffel symbols of $g$ . This holds at every point, so free-fall trajectories are geodesics of $g$ . $□$

The converse -- that geodesic motion on a Lorentzian manifold satisfies the EEP -- follows from the existence of Riemann normal coordinates at each point, in which $g_{μν} = η_{μν}$ and $Γ = 0$ at the point itself (with curvature appearing in second derivatives). The Riemann curvature tensor 13.02.01 then measures the tidal corrections to the EP at finite separation.

Remark on the field equations. The theorem above determines the kinematics (what free fall looks like) but not the dynamics (what $g_{μν}$ is for a given matter distribution). The Einstein field equations $G_{μν} = 8 π G T_{μν} / c^{4}$ 13.03.01 pending are an additional physical postulate, not a consequence of the EEP alone. One can imagine alternative field equations compatible with the same EEP-motivated geometric framework (e.g., Brans-Dicke theory, $f (R)$ gravity); the EEP constrains the kinematic structure but not the dynamic law.

Exercises [Intermediate+]

Exercise 3 (medium, symbolic).

In a uniformly accelerating rocket in flat spacetime (acceleration $a$ in the $z$ -direction), the metric in coordinates co-moving with the rocket is $d s^{2} = - (1 + a z / c^{2})^{2} c^{2} d t^{2} + d x^{2} + d y^{2} + d z^{2}$ . Show that the coordinate speed of light depends on $z$ and compute the ratio of the speed at height $z$ to the speed at $z = 0$ .

Hint

Set $d s^{2} = 0$ for a light ray moving radially (in the $z$ -direction).

Answer

For light, $d s^{2} = 0$ . A ray moving in the $z$ -direction ( $d x = d y = 0$ ) satisfies

0 = - (1 + a z / c^{2})^{2} c^{2} d t^{2} + d z^{2},

so $d z / d t = \pm c (1 + a z / c^{2})$ . At $z = 0$ the coordinate speed is $c$ . At height $z$ it is $c (1 + a z / c^{2})$ . The ratio is $1 + a z / c^{2}$ . This is the gravitational redshift in disguise: clocks at height $z$ run faster by the same factor, which is a direct consequence of the EP. Einstein derived this result in 1907 from the EP alone, years before the full field equations.

Exercise 4 (medium, symbolic).

Two test masses are released from rest at the same height above Earth, separated horizontally by distance $d = 1 m$ at altitude $h = 100 km$ . Estimate the rate at which they drift closer together due to tidal forces. Earth's radius is $R_{E} \approx 6370 km$ .

Hint

The tidal acceleration in the transverse direction is approximately $- g \cdot d / R_{E}$ where $g$ is the gravitational acceleration at height $h$ .

Answer

At height $h$ , $g \approx GM / R_{E}^{2} \approx 9.8 m/s^{2}$ (the correction from $h = 100 km$ is about 3%). The transverse tidal acceleration between two particles separated by $d$ is approximately $a_{tidal} \approx g \cdot d / R_{E}$ . With $d = 1 m$ and $R_{E} = 6.37 \times 1 0^{6} m$ :

a_{tidal} \approx \frac{9.8 \times 1}{6.37 \times 1 0 ^{6}} \approx 1.5 \times 1 0^{- 6} m/s^{2} .

The relative velocity after time $t$ is $v = a_{tidal} \cdot t$ . After 100 seconds, $v \approx 1.5 \times 1 0^{- 4} m/s$ . This is the tidal effect that distinguishes a real gravitational field from uniform acceleration -- and the reason the EP holds only locally.

Exercise 5 (medium, symbolic).

The Eotvos parameter $η$ measures WEP violation: if two materials have gravitational-to-inertial mass ratios $(m_{g} / m_{i})_{A}$ and $(m_{g} / m_{i})_{B}$ , then $η = 2∣ (m_{g} / m_{i})_{A} - (m_{g} / m_{i})_{B} ∣/ ((m_{g} / m_{i})_{A} + (m_{g} / m_{i})_{B})$ . The MICROSCOPE mission measured $∣ η ∣ < 1 0^{- 15}$ . If $η$ were instead $1 0^{- 3}$ , estimate the differential acceleration between a platinum and a titanium test mass (each $1 kg$ ) in Earth's field ( $g = 9.8 m/s^{2}$ ).

Hint

The differential acceleration is $Δ a = η \cdot g /2$ .

Answer

The differential acceleration is $Δ a \approx η \cdot g /2 = 1 0^{- 3} \times 9.8/2 \approx 5 \times 1 0^{- 3} m/s^{2}$ . This would be easy to detect with a torsion balance (Eotvos measured $η$ to about $1 0^{- 9}$ in 1891). The fact that $η < 1 0^{- 15}$ means that if the WEP is violated, the violation is extraordinarily small -- a tribute to how precisely nature enforces the equality of gravitational and inertial mass.

Exercise 7 (hard, symbolic).

Derive the gravitational redshift formula from the EP alone, without using the field equations. A photon is emitted at the floor of a rocket accelerating at $a$ and detected at the ceiling, a height $h$ above. Show that the received frequency is redshifted by $Δ ν / ν = - g h / c^{2}$ .

Hint

In the accelerating frame, the ceiling has velocity $v = ah / c$ relative to the floor at the time the photon arrives (to first order). Use the Doppler shift.

Answer

The photon travels upward at speed $c$ , taking time $Δ t = h / c$ to reach the ceiling. During this time the ceiling has acquired velocity $v = a Δ t = ah / c$ relative to the floor (in the instantaneously comoving frame of the floor at emission). The detector at the ceiling is moving away from the source, so there is a first-order Doppler redshift $Δ ν / ν = - v / c = - ah / c^{2}$ .

By the EP, the same result must hold for a gravitational field $g$ : a photon climbing height $h$ in a uniform gravitational field is redshifted by $Δ ν / ν = - g h / c^{2}$ . This was Einstein's 1907 derivation -- obtained from the EP and the Doppler formula alone, a decade before the field equations existed. The Pound-Rebka experiment (1959) confirmed this prediction by measuring the redshift of gamma rays over a vertical distance of $22.5 m$ at Harvard.

Exercise 8 (hard, symbolic).

Show that if the SEP holds, a black hole and a neutron star of the same mass, initially at rest in an external gravitational field, fall on identical trajectories. Then explain why lunar laser ranging tests the SEP, not just the WEP.

Hint

The SEP extends the WEP to self-gravitating bodies. Earth and Moon have very different gravitational binding energy fractions.

Answer

The SEP says that gravitational binding energy contributes equally to inertial and gravitational mass. A black hole's mass is entirely gravitational binding energy (in the classical limit); a neutron star's is partly gravitational, partly nuclear. If the SEP holds, both have $m_{g} = m_{i}$ exactly, so both follow the same geodesic in the external field.

Lunar laser ranging tests the SEP because the Earth and Moon have very different gravitational self-energy fractions ( $\sim - 4.6 \times 1 0^{- 10}$ for Earth vs $\sim - 1.9 \times 1 0^{- 11}$ for Moon). If the SEP were violated, the Earth and Moon would fall at slightly different rates toward the Sun, producing a measurable anomaly in the lunar orbit (the Nordtvedt effect). Current data constrain $η_{N} < 1 0^{- 4}$ , meaning any SEP violation is smaller than one part in ten thousand. This is the strongest test of the SEP to date.

Exercise 9 (hard, symbolic).

The Schiff conjecture states that any viable theory of gravity that satisfies the WEP must also satisfy the full EEP (and hence predict a curved Lorentzian manifold). Outline the argument: if $m_{g} / m_{i}$ is universal, why must non-gravitational constants be spacetime-independent?

Hint

Consider what would happen to energy conservation if a "constant" varied with position. Connect this to the WEP via the mass-energy equivalence $E = m c^{2}$ .

Answer

The argument (following Will 2018, Ch. 2) goes through energy conservation. If a fundamental parameter such as the fine-structure constant $α$ varied with position, then the binding energy of an atom would depend on where it is located. By $E = m c^{2}$ , this means the inertial mass of the atom depends on position. The WEP says $m_{g} / m_{i}$ is the same for all bodies at all locations, so the gravitational mass would have to vary in lockstep -- but there is no reason it should, unless the variation of $α$ feeds back into the gravitational coupling in a very specific way.

More precisely: consider two bodies $A$ and $B$ whose internal energies depend differently on $α$ . If $α$ varies with position, then $m_{i}^{A}$ and $m_{i}^{B}$ develop different position-dependences. The WEP demands $(m_{g} / m_{i})_{A} = (m_{g} / m_{i})_{B}$ everywhere, which forces the $α$ -dependence of $m_{g}$ to track $m_{i}$ exactly. This is an extremely fine-tuned requirement that almost all theories violating LPI also violate the WEP. Lightman and Lee (1973) proved this rigorously in the PPN framework: the parameters measuring WEP violation ( $η$ ) and LPI violation ( $α_{1}, α_{2}$ ) are linked by algebraic relations, and setting $η = 0$ forces the others to vanish as well in any "reasonable" (Lagrangian-based) theory. The Schiff conjecture thus says the WEP and the full EEP rise or fall together.

Exercise 10 (hard, symbolic).

In the PPN formalism, the Eotvos parameter is related to the PPN parameters by $η = 4 β - γ - 3 - α_{1} - \frac{2}{3} α_{2}$ where $β$ and $γ$ measure nonlinearity and space curvature per unit mass, and $α_{1}, α_{2}$ measure preferred-frame effects. Given that observation gives $γ - 1 = (2.1 \pm 2.3) \times 1 0^{- 5}$ (Cassini, 2003) and $β - 1 = (0.6 \pm 1.1) \times 1 0^{- 4}$ (from Mercury perihelion), compute the bound on $η$ implied by these measurements alone (ignoring $α_{1}, α_{2}$ for simplicity).

Hint

Set $α_{1} = α_{2} = 0$ and substitute the central values plus uncertainties.

Answer

With $α_{1} = α_{2} = 0$ , the formula reduces to $η = 4 β - γ - 3 = 4 (β - 1) - (γ - 1)$ . Using central values: $η = 4 (0.6 \times 1 0^{- 4}) - (2.1 \times 1 0^{- 5}) \approx 2.4 \times 1 0^{- 4} - 2.1 \times 1 0^{- 5} \approx 2.2 \times 1 0^{- 4}$ . Including uncertainties, $∣ η ∣ ≲ 1 0^{- 3}$ from these measurements alone. This is much weaker than the direct MICROSCOPE bound of $∣ η ∣ < 1 0^{- 15}$ . The point: direct free-fall tests constrain $η$ far more tightly than indirect PPN parameter measurements. The PPN relation becomes important for constraining the SEP (via $β$ and $γ$ ) rather than the WEP.

Lean formalization [Intermediate+]

The equivalence principle is a physical postulate and cannot be formalized as a mathematical theorem. However, the geometric structure it motivates -- Lorentzian manifolds, geodesics, local inertial frames -- is within the scope of Mathlib's differential geometry layer.

Mathlib already contains: smooth manifolds (Geometry.Manifold), tangent bundles, smooth maps, and the beginnings of affine connections. It does not yet contain: Lorentzian metrics as a first-class structure (only positive-definite Riemannian metrics); the proof that Riemann normal coordinates exist at any point of a pseudo-Riemannian manifold; or the geodesic equation derived from the Euler-Lagrange equations of the length functional.

A natural formalization target would be: given a pseudo-Riemannian manifold $(M, g)$ with signature $(-, +, +, +)$ , prove that at each point $p$ there exist coordinates in which $g_{μν} (p) = η_{μν}$ and $Γ (p) = 0$ . This is a theorem of pure differential geometry (existence of normal coordinates) and does not require the EP as an axiom. The EP enters as the physical interpretation of this mathematical result.

lean_status: none reflects the absence of the relevant Mathlib infrastructure. This unit ships without a lean_module and is reviewer-attested.

Advanced results [Master]

Schiff conjecture

The Schiff conjecture (1960) asserts that any self-consistent theory of gravity that satisfies the WEP must also satisfy the full EEP. This is not a mathematical theorem -- it depends on what counts as a "self-consistent" theory -- but it has been verified rigorously within the PPN framework.

Lightman and Lee (1973) proved that in any Lagrangian-based metric theory of gravity, the WEP parameter $η$ and the LPI/LLI violation parameters $α_{1}, α_{2}, α_{3}$ are related by algebraic identities. Setting $η = 0$ forces $α_{1}, α_{2}, α_{3} \to 0$ in all theories satisfying conservation of momentum and energy. The converse also holds: a theory violating the WEP necessarily violates LPI or LLI (or both).

The practical consequence is that experimental tests of the WEP (torsion balances, MICROSCOPE) simultaneously test LLI and LPI. One does not need separate experiments for each component of the EEP -- a result that would have surprised Einstein, who treated the universality of free fall and the local validity of special relativity as independent hypotheses.

Experimental tests of the EP

Torsion-balance experiments (Eotvos-Wash). The Eot-Wash group at the University of Washington has set the gold standard for laboratory WEP tests. Their rotating torsion balance compares the accelerations of test masses of different composition toward a laboratory source mass (or the Galactic center, or a hypothetical fifth-force source). The technique exploits the fact that a torsion balance is sensitive to torque differences of order $1 0^{- 15} N \cdot m$ . Current bounds: $∣ η ∣ < 1.8 \times 1 0^{- 13}$ (Adelberger et al. 2003; Schlamminger et al. 2008).

MICROSCOPE satellite (2016-2018). A space-based differential accelerometer comparing cylindrical test masses of platinum and titanium in free fall around Earth. The space environment eliminates seismic noise and allows long integration times. Final result: $∣ η ∣ < 1 0^{- 15}$ (95% confidence; Berge et al. 2022). This is the best current bound on WEP violation and confirms the EEP to extraordinary precision.

Lunar laser ranging (Nordtvedt test). The Earth and Moon have different gravitational self-energy fractions. If the SEP were violated, the two would fall at different rates toward the Sun, producing a polarization of the lunar orbit along the Earth-Sun line with amplitude $δ r \sim 13 m \cdot η_{N}$ . Decades of ranging data constrain $∣ η_{N} ∣ < 1 0^{- 4}$ (Williams, Turyshev and Boggs 2012). This tests the SEP specifically, not just the WEP, because it involves self-gravitating bodies.

Atomic interferometry. Cold-atom interferometers compare the free-fall accelerations of different atomic species (e.g., Rb-85 and Rb-87). Current bounds: $∣ η ∣ < 1 0^{- 12}$ , with prospects for $1 0^{- 15}$ or better in space-based experiments (STE-QUEST, proposed). The advantage is quantum control over the test masses; the disadvantage is that current measurements are less precise than torsion balances.

Gravitational redshift tests. The Gravity Probe A experiment (1976) compared a hydrogen maser clock on a rocket at $10, 000 km$ altitude to one on the ground. The measured redshift agreed with the EP prediction to $7 \times 1 0^{- 5}$ . More recently, comparisons of optical lattice clocks at different heights in the same laboratory (Bothwell et al. 2022, Tokyo Skytree experiment) have tested LPI at the $1 0^{- 18}$ level, corresponding to height sensitivity of about $1 cm$ .

The Nordtvedt effect

Kenneth Nordtvedt (1968) showed that in theories of gravity where the SEP is violated, a massive self-gravitating body does not follow the same geodesic as a test particle. The anomalous acceleration of a body with gravitational binding energy $U_{grav}$ and total mass $M$ is

a_{Nordtvedt} = η_{N} \frac{U _{grav}}{M c ^{2}} g_{external},

where $η_{N}$ is the Nordtvedt parameter and $g_{external}$ is the external gravitational field. For Earth, $U_{grav} / M c^{2} \approx - 4.6 \times 1 0^{- 10}$ ; for the Moon, $\approx - 1.9 \times 1 0^{- 11}$ . The difference drives the lunar-orbit polarization measured by lunar laser ranging.

In general relativity, $η_{N} = 0$ exactly -- the SEP holds, and all bodies, regardless of their internal structure or gravitational binding, follow identical geodesics. This is a nontrivial prediction of GR; scalar-tensor theories (Brans-Dicke and descendants) generically predict $η_{N} \neq = 0$ .

EP tests and alternatives to GR

The EP is the most precisely tested principle in gravitational physics. Every alternative theory of gravity that has been proposed either satisfies the EP (and is then constrained by other tests -- perihelion precession, gravitational waves, black hole shadows) or violates it at a level already ruled out by experiment.

This makes the EP an extraordinarily powerful discriminator. The constraints from MICROSCOPE ( $∣ η ∣ < 1 0^{- 15}$ ) rule out, for example, any theory in which a new force couples to baryon number, lepton number, or electromagnetic binding energy with coupling strength greater than $1 0^{- 15}$ times the gravitational coupling. The only viable modifications to GR are those that preserve the EP to this precision -- a severe theoretical constraint that has driven the development of "metric-compatible" alternatives (Einstein-dilaton-Gauss-Bonnet, dynamical Chern-Simons, massive bigravity) in which the EP is built in by construction.

Future directions

Several next-generation EP experiments are in development:

STEP (Satellite Test of the Equivalence Principle): a proposed space mission targeting $∣ η ∣ < 1 0^{- 18}$ using SQUID-based differential accelerometry.
MICROSCOPE 2: a proposed follow-up aiming for $∣ η ∣ < 1 0^{- 17}$ with improved thermal control.
STE-QUEST: a proposed ESA mission using atom interferometry in space, targeting $∣ η ∣ < 1 0^{- 15}$ with quantum test masses.
Atomic clock networks: comparing clocks at different gravitational potentials to test LPI at the $1 0^{- 20}$ level, which would probe Planck-scale variations of fundamental constants.

A detection of EP violation at any level would be revolutionary. It would indicate either a new force coupling to composition, a spatial variation of a fundamental constant, or a breakdown of the metric description of gravity -- any of which would require new physics beyond GR.

Connections [Master]

SR postulates 10.05.01 pending: The EEP extends special relativity from inertial frames to freely falling frames. In the language of the key theorem, the EEP says that at each point of the spacetime manifold, the tangent space carries the Minkowski metric. The EP is thus the bridge from the flat spacetime of SR to the curved spacetime of GR.
Tensors on manifolds 13.02.01: The EP motivates the manifold structure of spacetime, but the actual differential-geometric machinery -- the metric tensor, the Riemann curvature tensor, covariant derivatives -- lives in 13.02.01. The EP is the physical reason for the mathematical structure; the tensor calculus is the language in which the reason is expressed.
Einstein field equations 13.03.01 pending: The EP determines the kinematics (geodesic motion) but not the dynamics. The field equations $G_{μν} = 8 π T_{μν} / c^{4}$ are the separate physical postulate that specifies how matter determines the metric. A violation of the EP would invalidate the field equations, but the field equations could be wrong even if the EP holds (as in Brans-Dicke theory, which satisfies the EP but has different field equations).
Newton's laws 09.01.02 pending: The EP is implicit in Newton's observation that $m_{grav} = m_{inertial}$ . Newton tested this to about one part in $1 0^{3}$ with pendulums; Eotvos reached $1 0^{- 9}$ ; MICROSCOPE has reached $1 0^{- 15}$ . The EP elevates Newton's empirical observation to a founding principle of spacetime geometry.
Philosophy of physics -- equivalence of gravitational and inertial mass 20.03.01 pending: The equality $m_{g} = m_{i}$ raises a question that the EP answers physically but not philosophically: why are they equal? Mach's principle (inertial mass arises from interaction with distant matter) is one proposed answer; the geometric interpretation (both are aspects of the same thing, namely the geodesic-affinity of the body) is another. The philosophy-of-physics unit explores whether the EP explains the equality or merely encodes it.
Gravitational redshift and time dilation: The gravitational redshift formula $Δ ν / ν = - g h / c^{2}$ follows from the EP alone (Exercise 7). This makes the redshift a test of the EP, not of the full field equations -- a point that is often confused in textbook presentations. The Pound-Rebka experiment and the Gravity Probe A experiment are, strictly speaking, EP tests.
Geodesic deviation and tidal forces: The geodesic deviation equation 13.02.01 quantifies how tidal forces -- the failure of the EP over finite regions -- arise from spacetime curvature. Two nearby geodesics separate at a rate proportional to the Riemann curvature tensor. This is the geometric content of the statement "tidal forces measure curvature."

Historical & philosophical context [Master]

The story of the equivalence principle begins with Galileo, who (apocryphally) dropped objects from the Leaning Tower of Pisa and (verifiably) measured their descent rates using inclined planes. His conclusion that all bodies fall at the same rate regardless of weight was the first statement of what we now call the WEP.

Newton systematized this in the Principia (1687) by distinguishing gravitational mass (which appears in the law of gravitation $F = GM m / r^{2}$ ) from inertial mass (which appears in the second law $F = ma$ ). He tested their equality using pendulums of different compositions, reaching a precision of about $1 0^{- 3}$ . Newton had no explanation for the equality; it was an unexplained coincidence in his theory.

The first precision test was performed by Roland von Eotvos in 1885 using a torsion balance. His device compared the gravitational and inertial forces on test masses of different materials by measuring the torque on a horizontal beam suspended by a thin fiber. The Eotvos experiment reached a precision of about $1 0^{- 9}$ and confirmed the equality of gravitational and inertial mass to that level. The Eotvos experiment was refined by Dicke (1964) and by the Eot-Wash group (Adelberger, Heckel, and collaborators, 1990s-2000s), reaching $1 0^{- 13}$ .

Einstein's contribution, in his 1907 review paper "On the relativity principle and the conclusions drawn from it," was not a new measurement but a new interpretation. His "happiest thought" -- that a freely falling observer does not feel gravity -- turned the Eotvos result from a coincidence into a founding principle. If a freely falling frame is indistinguishable from an inertial frame in special relativity, then gravity is not a force but a property of spacetime geometry. This insight took eight years (1907-1915) to develop into the full field equations of general relativity.

The distinction between the weak, Einstein, and strong equivalence principles was articulated in the 1960s and 1970s, primarily by Dicke, Nordtvedt, and Will, as the theoretical framework for testing alternatives to general relativity. The PPN formalism (Will and Nordtvedt 1972) systematized this by parametrizing all possible deviations from GR in a common language. The Schiff conjecture (1960) -- that the WEP implies the full EEP -- provided theoretical support for the strategy of testing the EP as a whole rather than component by component.

John Norton's 1985 paper "What was Einstein's Principle of Equivalence?" clarified the historical and philosophical confusion surrounding the EP. Norton showed that Einstein used at least three distinct versions of the principle at different stages of his work: an early version (1907) asserting the equivalence of a uniform gravitational field and a uniformly accelerating frame; a middle version (1911-1912) identifying the gravitational redshift as a consequence; and the final version (1915-1916) asserting local Lorentz invariance in freely falling frames. These are not equivalent statements, and the modern EEP corresponds most closely to the final version.

The MICROSCOPE satellite mission (2016-2018) represents the current experimental frontier, confirming the WEP to $∣ η ∣ < 1 0^{- 15}$ . No violation has been observed at any scale. The EP remains one of the most precisely tested principles in all of physics, and its continued validity is both the foundation of general relativity and a powerful constraint on theories beyond it.

Bibliography [Master]

Primary and secondary literature:

Galilei, G., Discorsi e dimostrazioni matematiche intorno a due nuove scienze (1638). [Originator of the universality of free fall.]
Newton, I., Philosophiae Naturalis Principia Mathematica (1687), Def. I and Book III, Prop. VI. [Distinction of gravitational and inertial mass.]
Eotvos, R. v., "Uber die Anziehung der Erde auf verschiedene Substanzen," Math. Naturwiss. Ber. Ungarn 8 (1891), 65-68. [First precision torsion-balance test.]
Einstein, A., "Uber das Relativitatsprinzip und die aus demselben gezogenen Folgerungen," Jahrb. Radioakt. Elektron. 4 (1907), 411-462. [The "happiest thought" paper; EP stated for the first time.]
Einstein, A., "Uber den Einfluss der Schwerkraft auf die Ausbreitung des Lichtes," Ann. Phys. 35 (1911), 898-908. [Gravitational redshift derived from the EP.]
Dicke, R. H., "Theoretical significance of experimental tests of general relativity," in Evidence for Gravitational Theories, ed. Moller (Academic Press, 1962). [Modern reformulation of EP tests.]
Schiff, L. I., "On experimental tests of the general theory of relativity," Am. J. Phys. 28 (1960), 340-343. [The Schiff conjecture.]
Nordtvedt, K., "Equivalence principle for massive bodies, II: Theory," Phys. Rev. 169 (1968), 1017-1025. [The Nordtvedt effect.]
Lightman, A. P. and Lee, D. L., "Restricted proof that the weak equivalence principle implies the Einstein equivalence principle," Phys. Rev. D 8 (1973), 364-376. [Rigorous PPN proof of Schiff conjecture.]
Will, C. M., Theory and Experiment in Gravitational Physics, 2nd ed. (Cambridge, 2018). [The definitive reference on EP tests and PPN formalism.]
Will, C. M., "The Confrontation between General Relativity and Experiment," Living Rev. Rel. 17 (2014), 4. [Living review; updated periodically.]
Adelberger, E. G., Heckel, B. R. and Nelson, A. E., "Tests of the Gravitational Inverse-Square Law," Ann. Rev. Nucl. Part. Sci. 53 (2003), 77-121. [Eot-Wash torsion-balance results.]
Schlamminger, S., et al., "Test of the Equivalence Principle Using a Rotating Torsion Balance," Phys. Rev. Lett. 100 (2008), 041101.
Williams, J. G., Turyshev, S. G. and Boggs, D. H., "Lunar Laser Ranging Tests of the Equivalence Principle," Class. Quantum Grav. 29 (2012), 184004. [Best SEP test to date.]
Touboul, P., et al. (MICROSCOPE collaboration), "MICROSCOPE Mission: Final Results of the Test of the Equivalence Principle," Phys. Rev. Lett. 129 (2022), 121102. [Best WEP test: $∣ η ∣ < 1 0^{- 15}$ .]
Norton, J. D., "What was Einstein's Principle of Equivalence?," Stud. Hist. Phil. Sci. 16 (1985), 203-246. [Historical and philosophical clarification.]
Pound, R. V. and Rebka, G. A., "Apparent Weight of Photons," Phys. Rev. Lett. 4 (1960), 337-341. [First laboratory gravitational redshift measurement.]
Vessot, R. F. C., et al., "Test of Relativistic Gravitation with a Space-Borne Hydrogen Maser," Phys. Rev. Lett. 45 (1980), 2081-2084. [Gravity Probe A redshift test.]
Bothwell, T., et al., "Resolving the gravitational redshift across a millimetre-scale atomic sample," Nature 602 (2022), 420-424. [Centimeter-scale redshift measurement with optical clocks.]

Wave 3 physics unit, produced 2026-05-19. Both hooks_out targets (13.02.01, 20.03.01) are proposed; no receiving unit yet exists to confirm them. Status remains draft pending Tyler's review and the GR chapter retro per PHYSICS_PLAN.

Prerequisites

10.05.01 pending
09.01.02 pending

Tier anchors

beginner: Hartle, Gravity: An Introduction to Einstein's General Relativity (2003), Ch. 1-2
intermediate: Schutz, A First Course in General Relativity, 2e (2009), Ch. 1, 5
master: Will, Theory and Experiment in Gravitational Physics, 2e (2018); Norton, 'What was Einstein's Principle of Equivalence?' (1985)

References

TODO_REF
Schutz — A First Course in General Relativity, 2e (Cambridge, 2009) · Ch. 1, 5
TODO_REF
Hartle — Gravity (Addison-Wesley, 2003) · Ch. 1-2
TODO_REF
Will — Theory and Experiment in Gravitational Physics, 2e (Cambridge, 2018) · Ch. 2-3
TODO_REF
Einstein — Uber das Relativitatsprinzip und die aus demselben gezogenen Folgerungen (1907) · Jahrb. Radioakt. Elektron. 4, 411-462
TODO_REF
Eotvos — Uber die Anziehung der Erde auf verschiedene Substanzen (1891) · Math. Naturwiss. Ber. Ungarn 8, 65-68
TODO_REF
Norton — What was Einstein's Principle of Equivalence? (1985) · Stud. Hist. Phil. Sci. 16, 203-246
TODO_REF
Will — The Confrontation between General Relativity and Experiment (2014) · Living Rev. Rel. 17, 4
TODO_REF
Adelberger, Heckel and Nelson — Tests of the Gravitational Inverse-Square Law (2003) · Ann. Rev. Nucl. Part. Sci. 53, 77-121
TODO_REF
Schlamminger et al. — Test of the Equivalence Principle Using a Rotating Torsion Balance (2008) · Phys. Rev. Lett. 100, 041101
TODO_REF
Williams, Turyshev and Boggs — Lunar Laser Ranging Tests of the Equivalence Principle (2012) · Class. Quantum Grav. 29, 184004

Reviewer

Tyler (pending external GR reviewer per PHYSICS_PLAN 6)

Estimated time

beginner: 15m
intermediate: 30m
master: 45m