The Equivalence Principle

General relativity begins from a single empirical fact that Newtonian gravity treats as a coincidence: all bodies fall the same way in a gravitational field, regardless of their mass or composition. Einstein elevated this to a principle and drew from it the radical conclusion that gravity is not a force propagating on a fixed spacetime but a manifestation of spacetime curvature. This page motivates that move — from the universality of free fall to the geometrization of gravity — setting up the curved-spacetime arena built in The Geometry of Gravity.

We work in natural units $c = 1$ and keep $G$ explicit.

The crisis: Newtonian gravity and special relativity are incompatible

Special relativity fixed the kinematics of spacetime and demanded that no signal outrun light. Newtonian gravity violates this on its face:

Instantaneous action at a distance. Newton's law $F = - \frac{GM m}{r ^{2}} \hat{r}$ has the force depend on the present separation $r$ ; a change in the source is felt everywhere at once. This is incompatible with the finite signal speed and the relativity of simultaneity of SR.
No preferred frame for "now". The potential $Φ$ obeys $\nabla^{2} Φ = 4 π Gρ$ — an elliptic equation with no time derivative, hence no propagation and no Lorentz-covariant meaning.

A naive fix — writing a Lorentz-invariant scalar or vector field theory of gravity on flat spacetime — fails to reproduce known physics (a scalar theory gets the perihelion precession of Mercury wrong and gives no light bending; a vector theory makes like masses repel). The resolution is not a better field theory on flat spacetime but a change in the spacetime itself.

The universality of free fall

Two distinct notions of mass appear in Newtonian physics:

Inertial mass $m_{i}$ , the resistance to acceleration in $F = m_{i} a$ ;
Gravitational mass $m_{g}$ , the "gravitational charge" in $F = m_{g} g$ .

Equating them in a gravitational field gives $a = (m_{g} / m_{i}) g$ . Experimentally $m_{g} / m_{i}$ is the same for all bodies, so the acceleration is universal:

$a = g independent of mass or composition.$

This is the weak equivalence principle (WEP), or the universality of free fall (UFF) — Galileo's leaning-tower observation, sharpened. It has been tested to extraordinary precision:

Experiment	Bound on $η = 2 \frac{∣ a _{1} - a _{2} ∣}{a _{1} + a _{2}}$
Eötvös (torsion balance, 1908–22)	$\sim 1 0^{- 9}$
Dicke, Braginsky (1960s–70s)	$\sim 1 0^{- 11}$ – $1 0^{- 12}$
Lunar laser ranging	$\sim 1 0^{- 13}$
MICROSCOPE satellite (2017–22)	$\sim 1 0^{- 15}$

That $m_{g} = m_{i}$ to fifteen decimal places is, in Newtonian terms, an unexplained coincidence. In general relativity it is structural: free fall is motion along the geometry, and geometry does not know the composition of the falling body.

Einstein's elevator

Einstein's "happiest thought" (1907) turned UFF into a statement about frames. Consider a closed laboratory with no windows:

Freely falling in a uniform gravitational field. Every object inside — released coins, a beam of light — shares the lab's acceleration, so relative to the lab nothing falls. Locally, the lab is indistinguishable from an inertial lab floating in deep space, far from any mass. Gravity has been transformed away.
Accelerating in gravity-free space. A rocket accelerating at $g$ mimics a lab sitting at rest in a uniform field $g$ : released objects "fall" to the floor at $g$ ; a horizontal light beam bends downward. A uniform gravitational field is locally indistinguishable from acceleration.

graph LR
  A["Freely falling lab<br/>in gravity"] -->|indistinguishable| B["Inertial lab<br/>in deep space"]
  C["Lab at rest<br/>in gravity g"] -->|indistinguishable| D["Lab accelerating<br/>at g in free space"]

The word local is essential: the equivalence holds only over a region small enough that the field's variation is negligible. A real gravitational field, unlike a uniform one, is non-uniform — and that residual, non-removable variation is the true gravitational effect (see tidal forces below).

The three forms of the principle

The equivalence principle comes in strengthening versions:

Weak (WEP). The trajectory of a freely falling test body (uncharged, negligible self-gravity) depends only on its initial position and velocity, not on its internal structure or composition. Equivalent to $m_{g} = m_{i}$ .
Einstein (EEP). In any freely falling frame, the outcome of any local non-gravitational experiment is independent of where and when it is performed and of the frame's velocity. WEP + local Lorentz invariance + local position invariance. This is the assertion that special relativity holds locally in every freely falling frame.
Strong (SEP). EEP extended to include gravitational experiments and self-gravitating bodies — the local physics is that of SR even for systems whose own gravity matters (planets, stars). General relativity satisfies the SEP; most alternative (e.g. scalar–tensor) theories satisfy only the EEP, which is why SEP tests (Nordtvedt effect via lunar laser ranging) discriminate between theories.

The EEP is the operative one for building the theory: at each event there exists a local inertial frame in which the laws of physics take their special-relativistic form. This is the "minimal coupling" bridge — take an SR law and demand it hold in the local freely falling frame.

From a principle to geometry

The EEP says local inertial frames exist at every event, but a global inertial frame generally does not: you cannot make freely falling coordinates cover all of spacetime at once, because the field varies from place to place. This is exactly the situation of a curved manifold, where local flatness (a tangent plane at every point) coexists with global curvature.

The analogy is precise. On the curved surface of the Earth:

every point has a local tangent plane where Euclidean geometry holds to first order;
but no single flat map covers the globe without distortion;
the failure to build a global flat chart is measured by the curvature.

Replacing "tangent plane / Euclidean" with "local inertial frame / Minkowski", the same structure describes gravity: spacetime is a manifold that is locally Minkowskian (EEP) but globally curved, and gravity is the curvature. Free-falling bodies follow the straightest available lines — geodesics — and what Newton called the gravitational force is the failure of these geodesics to stay parallel.

Tidal forces: the irreducible content

If a uniform field can be transformed away by free fall, what cannot? The answer is the non-uniformity of a real field. Two test masses released side by side above the Earth both fall, but they fall toward the Earth's center, so they slowly converge; two masses at different heights separate, because the lower one falls faster. Inside a freely falling lab these tidal accelerations remain — they are the observable, frame-independent signature of gravity.

Tidal effects are exactly what a single accelerating frame cannot mimic, and they are what curvature encodes. The relative acceleration of nearby freely falling particles is governed by the geodesic deviation equation, in which the Riemann curvature tensor appears as the coefficient. Newtonian tidal tensor $\partial_{i} \partial_{j} Φ$ becomes, in the full theory, a component of the Riemann tensor — the precise sense in which "tidal force = curvature".

Two immediate physical consequences

Even before the field equations, the EEP alone predicts effects absent from Newtonian gravity + SR:

Gravitational redshift. Compare two clocks at heights separated by $h$ in a field $g$ . By the elevator argument (equivalence to acceleration) plus the relativistic Doppler shift, light climbing out of a potential well is redshifted by $\frac{Δ ν}{ν} = - \frac{g h}{c ^{2}} = - \frac{ΔΦ}{c ^{2}} .$ Confirmed by Pound–Rebka (1959) and, at high precision, by optical-clock comparisons over meter-scale height differences. Clocks deeper in a potential well run slow — gravitational time dilation.
Bending of light. A light beam crossing the accelerating elevator follows a curved path in the lab frame; by equivalence, light bends in a gravitational field. The full theory doubles the naive value, giving the $1.7 5^{''}$ deflection at the solar limb confirmed by Eddington (1919). See Schwarzschild solution.

Both effects say the same thing: the presence of gravity affects the rates of clocks and the paths of light, i.e. it deforms the metric that measures spacetime intervals.

Summary

Newtonian gravity (instantaneous, on flat space) is incompatible with special relativity.
All bodies fall identically ( $m_{g} = m_{i}$ to $\sim 1 0^{- 15}$ ): the universality of free fall.
Hence a freely falling frame is locally indistinguishable from an inertial frame (EEP): special relativity holds locally at every event.
Global inertial frames do not exist because the field varies — the hallmark of a curved manifold. Gravity is spacetime curvature; free fall is geodesic motion; tidal forces are curvature.

The Geometry of Gravity makes the curved-manifold picture precise: spacetime as a Lorentzian manifold with a dynamical metric $g_{μν}$ . The mathematics of geodesics and parallel transport follows in Covariant Derivative and Parallel Transport, and curvature in Curvature of Spacetime.

youyuanwu