The equivalence principle — weak, strong, Einstein, and experimental tests

Intuition Beginner

The companion unit 13.01.01 introduced the elevator thought experiment: standing in a sealed box on Earth feels identical to standing in a rocket accelerating at $9.8{\text{m/s}}^2$. Here we ask the sharper questions. How precisely is that equivalence true? And what exactly does the word "equivalent" cover?

Physicists split the claim into three nested statements. The weak form says every object, no matter what it is made of, falls at the same rate. The Einstein form adds that all of local physics — clocks, atoms, light, nuclear decays — behaves in free fall exactly as it does in empty space. The strong form goes further: even bodies held together by their own gravity, like the Earth or a neutron star, fall the same way as a feather.

Each level can fail on its own, so each gets its own experiment. The weak form is tested by dropping different materials side by side. The Einstein form is tested by comparing clocks at different heights. The strong form is tested by watching the Earth and Moon fall toward the Sun. Every test so far agrees, to about one part in a thousand trillion.

A failure at any level would break general relativity and point to new physics. That is why the precision matters so much. The MICROSCOPE satellite, launched in 2016, dropped cylinders of platinum and titanium in orbit and confirmed they fall together to one part in $10^{15}$ — the tightest test of any principle in all of physics.

The same effect runs the GPS on your phone. A satellite $20,200\text{km}$ up sits higher in Earth's gravitational well, so its clocks tick faster than ground clocks by a few parts in $10^{10}$. Left uncorrected, that mismatch would drift your reported position by about $10\text{km}$ per day. The system works only because engineers build the correction in.

Visual Beginner

The three principles are nested: the EEP contains the WEP as a special case, and the SEP contains the EEP. Each outer ring adds a claim the inner ring does not make. A theory could satisfy the inner ring and violate an outer one, so each level must be tested separately.

Principle	What it asserts	Best test	Current bound
WEP	All bodies fall the same way regardless of composition	MICROSCOPE (Pt vs Ti)	$
EEP	WEP + local non-gravitational physics is special relativity in free fall	Optical lattice clocks (LPI)	$\sim 10^{-18}$
SEP	EEP + gravitational self-energy falls the same way	Lunar laser ranging	$

Worked example Beginner

In 1959 Robert Pound and Glen Rebka fired a beam of gamma rays up a stairwell at Harvard, $22.5\text{m}$ tall, and measured how gravity changed the photons' energy. The prediction from the equivalence principle is that a photon climbing a height $h$ loses energy by the fraction $gh/c^2$, where $g=9.8{\text{m/s}}^2$ and $c$ is the speed of light.

For $h=22.5\text{m}$ the predicted fractional shift is

$$\frac{\Delta E}{E}=\frac{gh}{c^2}=\frac{9.8\times 22.5}{(3\times 10^8{)}^2}\approx 2.5\times 10^{-15}.$$

This is a staggeringly tiny shift. The $14.4\text{keV}$ gamma ray (emitted by iron-57) changes energy by only $3.6\times 10^{-14}\text{eV}$, about one part in $10^{15}$.

To detect so minute a change, Pound and Rebka used the Mossbauer effect: iron-57 nuclei absorb gamma rays only at exactly their own emission energy. They moved the source at a tiny speed to Doppler-shift the photons back into resonance. The compensating speed was $v=gh/c\approx 7\times 10^{-7}\text{m/s}$, slower than a snail. They measured the shift to within one percent of the prediction.

This is local position invariance in action: the fractional shift depends only on the height difference, not on what the photon is or how energetic it is. The same formula runs every atomic clock and every GPS satellite in orbit. The experiment turned the equivalence principle from a thought experiment into a measured number.

Check your understanding Beginner

Formal definition Intermediate+

The Eotvos parameter and the three principles

For two test bodies $A$ and $B$ with inertial masses $m_i^{A,B}$ and passive gravitational masses $m_g^{A,B}$, the Eotvos parameter is

$$\eta (A,B)=2\frac{(m_g/m_i{)}_A-(m_g/m_i{)}_B}{(m_g/m_i{)}_A+(m_g/m_i{)}_B}.$$

The WEP holds for this pair iff $\eta (A,B)=0$; it holds as a principle iff $\eta (A,B)=0$ for every pair of compositions. A non-zero $\eta$ means the two materials fall with different accelerations, $\Delta a/g=\frac{1}{2}\eta$ to leading order in the external field $g$.

Weak equivalence principle (WEP). The trajectory of a freely falling test body is independent of its internal structure and composition: $\eta (A,B)=0$ for all $A,B$. This is the universality of free fall.

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

Strong equivalence principle (SEP). The EEP extended to gravitational self-energy. A body's gravitational binding energy $U_g$ contributes equally to its inertial and passive gravitational mass, and the outcome of local gravitational experiments (including those on self-gravitating bodies) is also independent of velocity and location. Violations are parametrized by the Nordtvedt parameter ${\eta }_N$: a body of mass $M$ falls in an external field $g_{\text{ext}}$ with anomalous acceleration $\delta a={\eta }_N(U_g/M{c}^2)g_{\text{ext}}$.

The three statements are genuinely independent. A theory can satisfy the WEP yet violate LPI (a fundamental "constant" varying with position); or satisfy the EEP yet violate the SEP (gravitational binding energy falling differently from rest mass). Each independence is an experimental lever.

Metric theories of gravity

Following Thorne, Lee, and Lightman (1972) and the systematic development by Will, a theory of gravity is a metric theory iff it satisfies three conditions: (1) spacetime is equipped with a single symmetric Lorentzian metric $g_{\mu \nu }$ of signature $(-,+,+,+)$; (2) freely falling test bodies follow geodesics of $g_{\mu \nu }$; (3) in a local freely falling frame, all non-gravitational physics reduces to special relativity 13.02.01. General relativity, Brans-Dicke theory, and most scalar-tensor and vector-tensor theories are metric theories in this sense.

The decisive structural fact is that the EEP holds in a theory if and only if that theory is a metric theory. The forward direction (metric $\Rightarrow$ EEP) follows from the existence of Riemann normal coordinates at each event, in which $g_{\mu \nu }(p)={\eta }_{\mu \nu }$ and the Christoffel symbols vanish at $p$. The reverse direction (EEP $\Rightarrow$ metric) is the content of the Thorne-Lee-Lightman theorem: once local non-gravitational physics is special relativity in every freely falling frame, a single Lorentz-signature metric is forced on the whole manifold. Non-metric theories — for example the Belinfante-Swihart $b$-field theory, in which the metric that governs photon propagation differs from the metric coupled to massive matter — necessarily violate LLI and are excluded by Michelson-Morley-type experiments at the $10^{-15}$ level.

The parametrized post-Newtonian formalism

Within metric theories, the weak-field, slow-motion regime relevant to the solar system is described by ten PPN parameters: $\gamma ,\beta ,\xi ,{\alpha }_1,{\alpha }_2,{\alpha }_3,{\zeta }_1,{\zeta }_2,{\zeta }_3,{\zeta }_4$. Each measures a specific way a metric theory can deviate from GR; GR is the point $\gamma =\beta =1$ with all other parameters zero. The parameter $\gamma$ measures space curvature per unit rest mass; $\beta$ the nonlinearity in superposition of gravity; ${\alpha }_1,{\alpha }_2,{\alpha }_3$ preferred-frame effects; $\xi$ preferred-location effects; and ${\zeta }_1,\dots ,{\zeta }_4$ violations of total momentum conservation.

Two algebraic relations connect the PPN parameters to EP tests. First, in any fully conservative metric theory (one respecting total momentum conservation, so ${\alpha }_i={\zeta }_i=0$),

$$\eta =4\beta -\gamma -3-{\alpha }_1-\frac{2}{3}{\alpha }_2.$$

Setting $\eta =0$ (the WEP) does not by itself force $\beta =\gamma =1$; pinning those down requires the solar-system tests of 13.05.03. Second, the gravitational redshift is controlled by a separate parameter. A clock moved through a Newtonian potential difference $\Delta U$ ticks at a rate differing by

$$\frac{\Delta \nu }{\nu }=(1+\alpha )\frac{\Delta U}{c^2},$$

where the redshift parameter $\alpha$ vanishes iff LPI holds. Redshift experiments (Pound-Rebka, Gravity Probe A, optical lattice clocks) bound $\alpha$ directly, independent of the $\eta$ bound from free-fall tests. The cleanest statement of the EEP's experimental content is therefore four separate numbers: $\eta$ (WEP), $\alpha$ (LPI), the Kennedy-Thorndike combination (LLI), and ${\eta }_N$ (SEP).

Key derivation Intermediate+

Derivation A — the gravitational redshift from the EEP

A photon is emitted at the floor of a laboratory that accelerates "upward" with proper acceleration $g$ and is detected at the ceiling, a height $h$ above the floor. The photon climbs at speed $c$ and arrives after a coordinate time $\Delta t=h/c$. During that interval the ceiling has acquired an upward velocity

$$\Delta v=g\Delta t=\frac{gh}{c}$$

relative to the instantaneous rest frame of the floor at the moment of emission. The detector is receding from the source, so the photon suffers a first-order Doppler redshift:

$$\frac{\Delta \nu }{\nu }=-\frac{\Delta v}{c}=-\frac{gh}{c^2}.$$

By the EEP, this identical result must hold for a photon climbing in a static gravitational field of strength $g$. Recognizing $gh=\Delta \Phi$ as the Newtonian potential difference between emission and observation, the general statement is

$$\frac{\Delta \nu }{\nu }=-\frac{\Delta \Phi }{c^2},\text{equivalently}\frac{{\nu }_{\text{obs}}}{{\nu }_{\text{em}}}=1+\frac{{\Phi }_{\text{em}}-{\Phi }_{\text{obs}}}{c^2}.$$

The shift depends only on the potential difference — not on the photon's energy, frequency, or polarization. This frequency-independence is the testable content of local position invariance, and it is exactly what Pound-Rebka, Gravity Probe A, and modern optical-clock comparisons measure. The derivation uses only the EP plus the special-relativistic Doppler formula, so Einstein obtained it in 1911, four years before the field equations existed.

Derivation B — the Eotvos composition-dipole signal

Two masses $A$ and $B$ hang at opposite ends of a horizontal beam suspended by a thin fibre, forming a torsion balance at the surface of the rotating Earth. Each mass feels the Sun's gravitational acceleration $g_{\odot }\approx 6\times 10^{-3}{\text{m/s}}^2$ toward the Sun and the inertial centrifugal acceleration of Earth's rotation. If $A$ and $B$ had different ratios $m_g/m_i$, their net accelerations would point in slightly different directions, and the composition dipole $(A,B)$ would experience a torque about the fibre. The transverse differential acceleration is, to leading order,

$$\Delta a_{\perp }\approx \frac{1}{2}\eta g_{\odot }\sin \theta ,$$

with $\theta$ the angle between the dipole and the Sun. With the Eotvos-Pekar-Fekete sensitivity $\eta \sim 10^{-9}$, this is $\Delta a_{\perp }\sim 10^{-12}{\text{m/s}}^2$, far below seismic noise for a single measurement; the balance is therefore rotated at a known period and the signal is read out at that frequency, lifting it above the $1/f$ seismic background.

The MICROSCOPE satellite replaced the fibre with a drag-free differential accelerometer: two coaxial platinum and titanium cylinders held at fixed separation by electrostatic servos, falling together around Earth. The servo directly measures $\Delta a$, and the space environment removes seismic noise and permits integration over months of free fall. The result, $|\eta |<1.5\times 10^{-15}$, is the tightest WEP bound ever achieved.

Bridge. The redshift derivation builds toward 13.04.01, where the Newtonian potential $\Phi$ is revealed as the weak-field limit of the metric component $g_{00}=-(1+2\Phi /c^2)$, and appears again in 13.05.01 as the Schwarzschild redshift factor $\sqrt{1-2GM/(r{c}^2)}$ that fixes GPS clock rates and the Pound-Rebka shift. The foundational reason the redshift is universal is the metric-theory structure: every clock, photon, and atom couples to one and the same $g_{\mu \nu }$, so the $\Delta \Phi /c^2$ shift cannot depend on composition. This is exactly the content of local position invariance, and it generalises the flat-space Doppler argument to arbitrary curved spacetimes whose curvature is measured in 13.03.01; the bridge is between the kinematic EP arguments of this unit and the dynamical field equations that determine $\Phi$ from matter.

Exercises Intermediate+

Lean formalization Intermediate+

The equivalence principle in its three forms is a physical postulate, and the "metric theories of gravity" framework together with the parametrized post-Newtonian formalism are parametric physics-layer constructions rather than theorems of pure mathematics. None of them is a natural Lean statement without an axiomatised physics layer that Mathlib does not provide.

Mathlib does contain the geometric prerequisites at the linear-algebra and smooth-manifold level: Geometry.Manifold for smooth manifolds and their tangent and cotangent bundles, LinearAlgebra.TensorProduct and LinearAlgebra.BilinearForm for the algebra underlying the metric, and LinearAlgebra.QuadraticForm for indefinite forms at the vector-space level. What is missing is the layer that gives the EEP its mathematical teeth: Lorentzian signature as a first-class structure (Mathlib carries only positive-definite Riemannian metrics), the existence of Riemann normal coordinates on a pseudo-Riemannian manifold — the precise geometric shadow of the claim that a freely falling frame reduces locally to special relativity — and the geodesic equation derived as the Euler-Lagrange equation of the length functional.

The natural formalisation target is therefore the theorem that at each point of a pseudo-Riemannian manifold there exist coordinates in which $g_{\mu \nu }(p)={\eta }_{\mu \nu }$ and ${\Gamma }_{\mu \nu }^{\lambda }(p)=0$; the EEP is the physical interpretation of that result. The experimental parametrisations of this unit — the Eotvos parameter $\eta$, the Nordtvedt parameter ${\eta }_N$, the redshift parameter $\alpha$, and the ten PPN coefficients — have no Mathlib formalisation target at all: they are operational definitions tied to laboratory measurement, and their correctness gate is experimental rather than proof-theoretic. This unit therefore ships without a Lean module, mirroring the gap noted in 13.01.01 and extending it to the metric-theory classification and the PPN straitjacket as explicitly out of scope.

Advanced results Master

The strong-field SEP: pulsars as laboratories

Lunar laser ranging tests the SEP in the weak field of the solar system. The decisive strong-field tests come from neutron stars, where gravitational binding energy is a substantial fraction of the total mass. The cleanest system is the pulsar triple PSR J0337+1715 ^{[Archibald et al. 2018]}: a millisecond pulsar in a 1.6-day inner orbit with an inner white dwarf, the pair in a much wider orbit around an outer white dwarf. The neutron star, with gravitational binding fraction $|U_g|/M{c}^2\sim 0.1$–$0.2$, and the inner white dwarf, with $|U_g|/M{c}^2\sim 10^{-4}$, fall together in the field of the outer white dwarf. Timing the pulsar to microsecond precision over years of observation constrains their differential acceleration to $|\Delta a|/a<2.6\times 10^{-6}$ — the strongest SEP test in the strong-field regime.

This matters because whole classes of alternative gravity theories are indistinguishable from GR in the weak field but diverge for compact objects. In the Damour-Esposito-Farèse family of scalar-tensor theories, a neutron star can undergo spontaneous scalarization, acquiring a scalar charge that makes it fall differently from a white dwarf. The PSR J0337+1715 bound excludes the symmetric version of this effect across most of the coupling-strength parameter space, complementing the weak-field constraints from Cassini on the scalar coupling.

Systematic-error budgets

The precision of an EP test is set not by its headline number but by its control of systematic error. Each instrument has a different floor.

The MICROSCOPE differential accelerometer was limited by thermal gradients across the instrument, which produced differential radiometer forces at the femtonewton level; the Earth-along-track gravity gradient coupled into the signal at the orbital period; and residual gas outgassing produced a stochastic drag. Drag-free control reduced the non-gravitational acceleration on the satellite to below $10^{-12}{\text{m/s}}^2$, and the common-mode rejection of the two coaxial test masses suppressed shared disturbances by a factor of $10^4$. The final systematic floor sat near $10^{-15}{\text{m/s}}^2$ after two years of integration.

The Eot-Wash rotating torsion balance was limited by seismic noise at low frequency (hence rotation to move the signal to a quiet band), by the composition of local topographic mass, and by the need to distinguish a composition-dependent fifth force from ordinary gravitational gradients. Lunar laser ranging must disentangle a putative $13\text{m}\times {\eta }_N$ polarization from solar-radiation-pressure, regolith, and asteroid-perturbation effects over decades of data. Pound-Rebka's floor was the temperature dependence of the Mossbauer linewidth and second-order Doppler shifts from source motion.

The full PPN experimental map

Each PPN parameter is pinned by a specific class of observation, and the map is what makes the formalism powerful as a discriminator.

The parameter $\gamma$ (space curvature per unit mass) is fixed by light deflection and the Shapiro time delay; the Cassini 2003 measurement gives $\gamma -1=(2.1\pm 2.3)\times 10^{-5}$, the tightest solar-system bound. The parameter $\beta$ (nonlinearity) is fixed by perihelion precession of Mercury and lunar orbital precession, yielding $\beta -1=(1.0\pm 1.1)\times 10^{-4}$. The preferred-frame parameters ${\alpha }_1,{\alpha }_2$ are bounded by pulsar proper motions and lunar precession to $\lesssim 10^{-4}$; the parameter ${\alpha }_3$, which would produce a secular acceleration of pulsars, is bounded below $10^{-20}$ and is the most tightly constrained PPN coefficient. The momentum-conservation parameters ${\zeta }_i$ are excluded by Newton's-third-law tests and by the stability of the lunar orbit.

A single consistent picture emerges: GR, the point $\gamma =\beta =1$ with all others zero, lies within every experimental ellipse, and the allowed region around it is tiny. Any viable alternative must pass through that same region, which is why viable beyond-GR theories (Einstein-dilaton-Gauss-Bonnet, dynamical Chern-Simons, carefully tuned massive gravity) are all constructed to reproduce GR at first post-Newtonian order while deviating only at higher order or in the strong field.

Why the EP is the most powerful discriminator

Every proposed alternative to GR falls into one of two categories. Either it preserves the metric-theory structure and the EP, in which case it survives the $10^{-15}$ WEP bound but is then constrained by $\gamma$, $\beta$, gravitational waves, and black-hole shadows; or it violates the EP, in which case it is excluded outright by MICROSCOPE and the torsion-balance results. There is no third option within the Lagrangian-based framework analysed by Lightman and Lee.

This binary is what makes the EP the most powerful single discriminator in gravitational physics. A composition-dependent fifth force coupled to baryon number, lepton number, or electromagnetic binding energy is excluded at couplings stronger than $10^{-15}$ of the gravitational coupling — a bound unreachable by any other measurement. The future experimental programme (STEP, STE-QUEST, atomic-interferometer satellites) targets $|\eta |<10^{-18}$, which would probe energy scales relevant to quantum gravity and to the dilaton couplings predicted by string compactifications.

Synthesis. The three equivalence principles are the foundational reason that gravity admits a unified geometric description: the WEP forces geodesic motion, the EEP (via the Thorne-Lee-Lightman theorem) forces a single Lorentzian metric and reduces local physics to special relativity, and the metric-theory framework generalises every viable theory of gravity into the PPN straitjacket that experiment can parametrise. The central insight is that universality of coupling is the quantity under test — every clock, every atom, every photon must read off the same $g_{\mu \nu }$, and any material-dependent response would immediately violate the EP. This is exactly why Pound-Rebka, GPS, MICROSCOPE, lunar laser ranging, and the pulsar triple all measure the same underlying invariance at wildly different scales and precisions, from $10^{-2}$ down to $10^{-15}$. Putting these together, the experimental programme of the last century is a single coordinated siege on the metric-theory hypothesis, and the bridge is from the kinematic EP of this unit to the dynamical content of 13.04.01, where that single metric is sourced by matter and curved into the Schwarzschild geometry of 13.05.01.

Full proof set Master

Proposition (EEP $\iff$ metric theory). A theory of gravity is a metric theory if and only if it satisfies the Einstein equivalence principle. In particular, a freely falling test body follows a geodesic of the unique Lorentzian metric $g_{\mu \nu }$, and no second metric governing any sector of matter can coexist with local Lorentz invariance.

Proof. ($\Rightarrow$) Assume the theory is metric. At each event $p$ there exist Riemann normal coordinates in which $g_{\mu \nu }(p)={\eta }_{\mu \nu }$ and ${\Gamma }_{\mu \nu }^{\lambda }(p)=0$. In these coordinates the metric and its first derivatives match Minkowski space at $p$, so every non-gravitational experiment performed in a small enough neighbourhood of $p$ returns its special-relativistic value: the WEP holds (test bodies follow geodesics, straight lines at $p$), LLI holds (the local frame is Minkowskian), and LPI holds (the local laws contain no reference to absolute position). Hence the EEP holds.

($\Leftarrow$) Assume the EEP. The WEP supplies, at each event $p$, a family of freely falling trajectories through $p$ that are independent of the test body's composition; these are the candidate geodesics. By LLI, the readings of clocks and rods in a small freely falling laboratory at $p$ are related by Lorentz transformations, so the tangent space $T_{p}M$ carries the Minkowski inner product ${\eta }_{\alpha \beta }=\text{diag}(-1,+1,+1,+1)$. The assignment $p\mapsto {\eta }_{\alpha \beta }(p)$ patches smoothly across overlapping frames (physical trajectories depend smoothly on initial data), producing a single smooth Lorentzian metric $g_{\mu \nu }$ on the manifold, and the freely falling trajectories are its geodesics. This establishes conditions (1)–(3) of the metric-theory definition.

For uniqueness, suppose two metrics $g_{\mu \nu }$ and ${\tilde{g}}_{\mu \nu }$ governed different sectors, so that photons propagated under $\tilde{g}$ while massive matter coupled to $g$. The ratio of photon speed to massive-particle speed would then vary with the orientation and velocity of the local frame, producing an anisotropy in the measured speed of light — a direct violation of LLI. Michelson-Morley-type experiments bound such anisotropy below $10^{-15}$, excluding any two-metric construction. Therefore the metric is unique, and the theory is metric. $\square$

Proposition (LPI redshift is species-independent). In any metric theory, the gravitational redshift between two static observers is ${\nu }_{\text{obs}}/{\nu }_{\text{em}}=\sqrt{g_{00}(r_{\text{em}})/g_{00}(r_{\text{obs}})}$, independent of the photon's energy, frequency, and species, and reduces in the weak field to $\Delta \nu /\nu =-\Delta \Phi /c^2$.

Proof. Let the spacetime be static with timelike Killing field $\xi ={\partial }_t$, and let a photon travel on a null geodesic with tangent $k^{\mu }$. The quantity $E=-g_{\mu \nu }{\xi }^{\mu }{k}^{\nu }$ is conserved along the geodesic because $\xi$ is Killing. A static observer at radius $r$ has four-velocity $u^{\mu }(r)={\xi }^{\mu }/\sqrt{-g_{00}(r)}$ (normalised so $g_{\mu \nu }{u}^{\mu }{u}^{\nu }=-1$), and measures frequency $\nu (r)=-g_{\mu \nu }{u}^{\mu }{k}^{\nu }=E/\sqrt{-g_{00}(r)}$. The emitter at $r_{\text{em}}$ measures its own atomic transition frequency ${\nu }_0$, a constant of its internal physics; the conserved energy at emission is therefore $E={\nu }_0\sqrt{-g_{00}(r_{\text{em}})}$. At the observer,

$$\frac{{\nu }_{\text{obs}}}{{\nu }_{\text{em}}}=\frac{E/\sqrt{-g_{00}(r_{\text{obs}})}}{{\nu }_0}=\sqrt{\frac{-g_{00}(r_{\text{em}})}{-g_{00}(r_{\text{obs}})}}.$$

The atomic frequency ${\nu }_0$ cancels, so the ratio depends only on the metric at the two radii and not on the photon species or energy — which is local position invariance. In the weak field $g_{00}=-(1+2\Phi /c^2)$, expanding the square root gives ${\nu }_{\text{obs}}/{\nu }_{\text{em}}=1+({\Phi }_{\text{em}}-{\Phi }_{\text{obs}})/c^2$, i.e. $\Delta \nu /\nu =-\Delta \Phi /c^2$. $\square$

Connections Master

• Riemann curvature 13.03.01: the EP holds only at a point, in a single tangent space; the Riemann tensor measures precisely the failure of the equivalence between gravity and acceleration to extend over any finite region. The tidal accelerations that no frame can eliminate are the components of $R^{\rho }{}_{\sigma \mu \nu }$, so curvature is the obstruction that turns a local principle into a global geometry. This unit tests the EP to $10^{-15}$; the next tests how the geometry bends around it.

• Einstein field equations 13.04.01: the metric-theory theorem shows the EEP fixes the kinematics (geodesic motion under a unique $g_{\mu \nu }$) but says nothing about the dynamics — how matter sources $g_{\mu \nu }$. Brans-Dicke theory satisfies the EEP yet obeys different field equations with an additional scalar field, which is why the PPN parameter $\gamma$ rather than $\eta$ distinguishes it from GR. The field equations are the separate dynamical postulate layered on top of the EEP.

• Schwarzschild solution 13.05.01: the static, spherically symmetric metric $d{s}^2=-(1-2GM/r{c}^2)c^{2}d{t}^2+\dots$ supplies the concrete redshift factor $\sqrt{1-2GM/(r{c}^2)}$ that underlies both GPS clock corrections and the Pound-Rebka shift in the Earth's weak field. Every numerical answer in this unit's exercises is the weak-field expansion of a Schwarzschild quantity.

• Solar-system tests of GR 13.05.03: the PPN parameters $\gamma$ and $\beta$ defined in this unit are exactly the quantities pinned down by the light-deflection, Shapiro-delay, and perihelion-precession measurements catalogued there. The relation $\eta =4\beta -\gamma -3$ links the laboratory WEP bound to the dynamical solar-system tests, so the two programmes constrain the same underlying theory from complementary directions.

• The equivalence principle 13.01.01: this unit is the depth complement of its parent. The parent establishes the elevator thought experiment and the theorem that the EEP forces a Lorentzian manifold; this unit unpacks the three nested forms, the metric-theory classification, the PPN formalism, and the experimental machinery that tests each level to its current bound.

• Special relativity 10.05.01: local Lorentz invariance is the demand that special relativity hold in every local freely falling frame, not merely in global inertial frames. The Michelson-Morley and Kennedy-Thorndike lineage that confirmed SR on Earth is reinterpreted here as a test of the EEP, because the EEP promotes SR from a global statement to a local one valid at every event of a curved spacetime.

Historical & philosophical context Master

The quantitative history of the equivalence principle is a story of precision, measured in powers of ten. Galileo's inclined-plane experiments around 1604 established the universality of free fall to about one part in a hundred. Newton, distinguishing gravitational from inertial mass in the Principia (1687), tested their equality with pendulums of different composition to about one part in $10^3$, and noted the equality as an unexplained coincidence.

The first precision measurement was Roland von Eotvos's torsion balance, begun in 1885 and refined through the 1920s ^{[Eotvos, Pekar & Fekete 1922]}. Eotvos hung two masses of different composition at opposite ends of a beam and compared the gravitational pull of the Earth with the inertial centrifugal force of Earth's rotation; he reached $|\eta |\sim 10^{-9}$, a thousand-fold improvement on Newton. Roll, Krotkov, and Dicke (1964) modernized the apparatus, and the Eot-Wash group (Adelberger, Heckel, and collaborators, from 1990 onward) pushed the laboratory limit to $|\eta |<1.8\times 10^{-13}$ using a rotating torsion balance with the Sun as the source.

Einstein's contribution, in his 1907 review and the 1911 paper ^{[Einstein 1911]}, was interpretive rather than instrumental. He derived the gravitational redshift $\Delta \nu /\nu =-gh/c^2$ from the EP plus the Doppler formula alone, four years before the field equations. The prediction sat untested until 1959–1960, when Pound and Rebka ^{[Pound & Rebka 1960]} used the newly discovered Mossbauer effect to measure the redshift of $14.4\text{keV}$ gamma rays over $22.5\text{m}$ at Harvard, agreeing with the EP prediction to about one percent. The Gravity Probe A hydrogen-maser rocket experiment (Vessot and Levine, 1976) extended the test to $7\times 10^{-5}$ using a clock at $10,000\text{km}$ altitude; modern optical lattice clocks now reach $\sim 10^{-18}$ over centimetre height differences.

The conceptual tripartition into weak, Einstein, and strong principles, and the PPN formalism that parametrizes alternatives, were the achievement of Dicke, Nordtvedt, and Will in the 1960s and 1970s ^{[Will 2018]}. Nordtvedt's 1968 prediction that scalar-tensor theories would make the Earth and Moon fall differently toward the Sun — the Nordtvedt effect — turned lunar laser ranging into a SEP test; half a century of ranging data now bound $|{\eta }_N|<5\times 10^{-4}$. The MICROSCOPE satellite mission ^{[Touboul et al. 2022]}, launched in 2016, delivered the present world record $|\eta |<1.5\times 10^{-15}$. The strong-field frontier has moved to pulsars: PSR J0337+1715 ^{[Archibald et al. 2018]} tests the SEP for a neutron star to $2.6\times 10^{-6}$, probing compact-object gravity that no laboratory experiment can reach.

Philosophically, the EP raises the question John Norton pressed in 1985: what exactly did Einstein assert? The principle evolved through at least three inequivalent versions between 1907 and 1916, from the equivalence of a uniform field and uniform acceleration, through the redshift consequence, to the modern statement of local Lorentz and position invariance in freely falling frames. The experimental programme of this unit tests the modern (strongest) version, and its survival to $10^{-15}$ is the deepest empirical fact gravity theory possesses.

Bibliography Master

@article{eotvos1922, author = {E{"o}tv{"o}s, R. v. and P{'e}k{'a}r, D. and Fekete, E.}, title = {Beitr{"a}ge zum Gesetze der Proportionalit{"a}t von Tr{"a}gheit und Gravit"at}, journal = {Annalen der Physik}, volume = {68}, pages = {11--66}, year = {1922}, note = {The definitive torsion-balance test; $|\eta |\sim 10^{-9}$.} }

@article{einstein1911, author = {Einstein, A.}, title = {{"U}ber den Einfluss der Schwerkraft auf die Ausbreitung des Lichtes}, journal = {Annalen der Physik}, volume = {35}, pages = {898--908}, year = {1911}, note = {The gravitational redshift derived from the equivalence principle.} }

@article{poundrebka1960, author = {Pound, R. V. and Rebka, G. A.}, title = {Apparent Weight of Photons}, journal = {Physical Review Letters}, volume = {4}, pages = {337--341}, year = {1960}, note = {First laboratory gravitational-redshift measurement, 22.5,m, Fe-57 M{"o}ssbauer line.} }

@article{vessot1980, author = {Vessot, R. F. C. and Levine, M. W. and others}, title = {Test of Relativistic Gravitation with a Space-Borne Hydrogen Maser}, journal = {Physical Review Letters}, volume = {45}, pages = {2081--2084}, year = {1980}, note = {Gravity Probe A redshift test to $7\times 10^{-5}$.} }

@article{nordtvedt1968, author = {Nordtvedt, K.}, title = {Equivalence Principle for Massive Bodies, {II}: Theory}, journal = {Physical Review}, volume = {169}, pages = {1017--1025}, year = {1968}, note = {The Nordtvedt effect and the strong equivalence principle.} }

@article{lightmanlee1973, author = {Lightman, A. P. and Lee, D. L.}, title = {Restricted Proof that the Weak Equivalence Principle Implies the Einstein Equivalence Principle}, journal = {Physical Review D}, volume = {8}, pages = {364--376}, year = {1973}, note = {Rigorous PPN proof of the Schiff conjecture within Lagrangian metric theories.} }

@book{will2018, author = {Will, C. M.}, title = {Theory and Experiment in Gravitational Physics}, edition = {2nd}, publisher = {Cambridge University Press}, year = {2018}, note = {The canonical reference on the metric-theory framework, the PPN formalism, and EP tests.} }

@article{will2014, author = {Will, C. M.}, title = {The Confrontation between General Relativity and Experiment}, journal = {Living Reviews in Relativity}, volume = {17}, pages = {4}, year = {2014}, note = {Living review, updated periodically; comprehensive summary of bounds.} }

@article{schlamminger2008, author = {Schlamminger, S. and Choi, K.-Y. and Wagner, T. A. and Gundlach, J. H. and Adelberger, E. G.}, title = {Test of the Equivalence Principle Using a Rotating Torsion Balance}, journal = {Physical Review Letters}, volume = {100}, pages = {041101}, year = {2008}, note = {E{"o}t-Wash laboratory WEP bound $|\eta |<1.8\times 10^{-13}$.} }

@article{williams2012, author = {Williams, J. G. and Turyshev, S. G. and Boggs, D. H.}, title = {Lunar Laser Ranging Tests of the Equivalence Principle}, journal = {Classical and Quantum Gravity}, volume = {29}, pages = {184004}, year = {2012}, note = {Best SEP test from lunar ranging; $|{\eta }_N|<5\times 10^{-4}$.} }

@article{touboul2022, author = {Touboul, P. and others}, title = {{MICROSCOPE} Mission: Final Results of the Test of the Equivalence Principle}, journal = {Physical Review Letters}, volume = {129}, pages = {121102}, year = {2022}, note = {Best WEP test: $|\eta |<1.5\times 10^{-15}$ from a drag-free Pt/Ti differential accelerometer.} }

@article{archibald2018, author = {Archibald, A. M. and others}, title = {Universality of Free Fall from the Orbital Motion of a Pulsar in a Stellar Triple System}, journal = {Nature}, volume = {559}, pages = {73--76}, year = {2018}, note = {PSR J0337+1715 strong-field SEP test; $|\Delta a|/a<2.6\times 10^{-6}$.} }

@book{carroll2004, author = {Carroll, S. M.}, title = {Spacetime and Geometry: An Introduction to General Relativity}, publisher = {Addison-Wesley}, year = {2004}, note = {Intermediate treatment of the EP, metric theories, and redshift derivation.} }

@book{hartle2003, author = {Hartle, J. B.}, title = {Gravity: An Introduction to Einstein's General Relativity}, publisher = {Addison-Wesley}, year = {2003}, note = {Undergraduate operational treatment of the three EP forms and GPS.} }

@book{mtw1973, author = {Misner, C. W. and Thorne, K. S. and Wheeler, J. A.}, title = {Gravitation}, publisher = {W. H. Freeman}, year = {1973}, note = {Master-level reference on the metric-theory theorem and strong-field gravity.} }

---

Depth complement of 13.01.01. Catalog stub gr-cosmology.equivalence-depth registered before validation. Status remains shipped pending Tyler's GR-chapter review.