The Equivalence Principle

Special relativity, built in the preceding modules, is the physics of inertial frames in flat spacetime. Gravity does not fit that frame: there is no way to switch off a planet, and no global inertial frame exists in its presence. The route into general relativity begins with a single experimental fact that Newton already knew and could not explain — that all bodies fall with the same acceleration — and reads it as a statement about geometry rather than about forces. This lesson formalizes that fact into the equivalence principle, isolates exactly how far it reaches, and identifies the residue it cannot remove: the tidal field, which the rest of the module builds into curvature.

Inertial and gravitational mass

Two masses appear in Newtonian mechanics, and nothing in that theory requires them to be equal. The inertial mass $m_{i}$ is the resistance to acceleration in the second law, $F = m_{i} a$ . The gravitational mass $m_{g}$ is the coupling to a gravitational field, the gravitational analogue of electric charge, appearing in $F_{g} = m_{g} g$ . For a body falling under gravity alone,

m_{i} a = m_{g} g ⟹ a = \frac{m _{g}}{m _{i}} g .

The acceleration is independent of the body only if the ratio $m_{g} / m_{i}$ is the same universal constant for every material. Experiment says it is, to extraordinary precision. Newton compared pendulum periods; Eötvös compared the torque on a balance carrying different materials in the combined field of Earth's gravity and rotation; modern torsion-balance and lunar-laser-ranging tests bound the fractional difference

η = 2 \frac{∣ a _{1} - a _{2} ∣}{a _{1} + a _{2}}

for two test bodies at the $1 0^{- 13}$ level, and the MICROSCOPE satellite mission tightened it further. No other pair of charges in physics is observed to be proportional this way; electric charge and inertial mass are utterly unrelated. The universality of free fall is therefore a clue, not a coincidence, and the equivalence principle takes it as foundational: set $m_{g} = m_{i}$ identically and ask what follows.

Two bodies of different composition released together fall along the same trajectory because the ratio of gravitational to inertial mass is universal; their separation stays fixed to the precision of the Eotvos-type experiments.

The three equivalence principles

The single Newtonian equality unfolds into a hierarchy of physical statements of increasing strength. Each says that some class of experiments cannot detect a uniform gravitational field.

Weak equivalence principle (WEP). The trajectory of a freely falling test body depends only on its initial position and velocity, not on its composition. This is the universality of free fall, $m_{g} = m_{i}$ , restricted to the motion of uncharged, non-spinning test masses in a given external field.
Einstein equivalence principle (EEP). In a freely falling laboratory, the outcome of any local, non-gravitational experiment is independent of the laboratory's velocity and location in spacetime. WEP is the special case of mechanical experiments; EEP extends the claim to electromagnetism, weak and strong interactions, and every other non-gravitational law. It bundles WEP with local Lorentz invariance (no preferred velocity) and local position invariance (no preferred place or time).
Strong equivalence principle (SEP). EEP holds for all experiments, including those in which the object's own gravitational binding energy is significant — planets, stars, the Cavendish experiment done with massive source bodies. SEP demands that gravity couple to its own energy exactly as it couples to any other, and general relativity is essentially the unique metric theory that satisfies it. Most competing theories obey EEP but violate SEP, so tests that stress the strong version (lunar laser ranging, binary-pulsar timing) discriminate between theories.

The operative promotion is from EEP to geometry. If special relativity holds in a small patch around every event, then spacetime is locally Minkowskian everywhere — the defining property of a curved manifold equipped with a metric, built in the next lesson. General relativity is the theory that takes EEP seriously and asks what global structure is consistent with local flatness.

The elevator gedanken

Einstein's argument for EEP is a pair of thought experiments that cannot be told apart. Consider a windowless laboratory.

Cabin A sits at rest on the surface of a planet with surface gravity $g$ . An object released inside accelerates downward at $g$ ; the occupant feels a weight $m g$ pressing on the floor.
Cabin B is far from all masses, dragged by a rocket with proper acceleration $a = g$ . An object released inside, now unforced, continues at constant velocity while the floor accelerates up to meet it — it falls at $g$ relative to the cabin; the occupant feels a weight $m g$ from the floor pushing up.

No local experiment inside the sealed cabin distinguishes the two situations. The weight, the fall of dropped objects, the arc of a thrown ball, the reading of a spring scale — all agree. The converse gedanken removes gravity: a laboratory freely falling toward the planet, like a lift with its cable cut, reproduces the physics of an inertial cabin drifting in deep space. Inside the falling lab, released objects float, and the occupant is weightless. Free fall cancels a uniform gravitational field exactly.

The four cabins of the elevator argument. Standing on a planet (A) and accelerating in free space (B) give identical local physics; free fall toward the planet (C) and drifting inertially in deep space (D) are likewise identical.

Two immediate predictions follow from the gedanken alone, before any field equation. A light ray crossing the accelerating cabin B travels a straight line in the inertial frame but, seen from the accelerating floor, bends downward, because the far wall has moved up during the crossing. By EEP the same bending must occur in cabin A: light falls in a gravitational field. And a photon emitted from the floor of cabin B and received at the ceiling is redshifted, because the ceiling recedes during the flight time; by EEP a photon climbing a gravitational field is likewise redshifted. Both effects are derived rigorously later in the course; here they are consequences of equivalence, not of curvature.

In the accelerating cabin a horizontal light ray strikes the far wall below its entry height, because the wall rises during the crossing time; the equivalence principle transfers the same downward bend to a static cabin in gravity.

The word local and its limits

The equivalence of a freely falling frame with an inertial one is exact only in an infinitesimal neighbourhood. A genuine gravitational field is not uniform: it points toward the source and weakens with distance. Two test particles released side by side above the Earth fall along lines that both aim at the Earth's centre, so they converge; two particles released one above the other fall with different accelerations, so they separate vertically. In a single freely falling lab, objects do not merely float — they drift relative to one another. This relative acceleration is the tidal field, and no choice of frame removes it.

The distinction is sharp. A uniform field can be transformed away globally: the transformation to a frame accelerating at $g$ cancels it everywhere at once. A real field can be transformed away only at a point (or, to first order, along a single worldline). Expand the Newtonian potential $Φ$ about the lab's centre $x_{0}$ :

\partial_{i} Φ (x) = \partial_{i} Φ (x_{0}) + \partial_{i} \partial_{j} Φ (x_{0}) (x - x_{0})^{j} + \dots .

Choosing the freely falling frame subtracts the first term, the uniform part, for every particle at once. The second term, the gradient of the field, survives; it drives the relative acceleration of neighbouring particles,

\frac{d ^{2} ξ ^{i}}{d t ^{2}} = - \partial_{i} \partial_{j} Φ ξ^{j},

where $ξ$ is the separation between two nearby free-fallers. The matrix of second derivatives $\partial_{i} \partial_{j} Φ$ is the tidal tensor. Its trace vanishes in vacuum by Laplace's equation $\nabla^{2} Φ = 0$ , so tidal distortion conserves volume there: what stretches along the field line squeezes transverse to it. This equation is the Newtonian shadow of geodesic deviation, and $\partial_{i} \partial_{j} Φ$ is the shadow of the Riemann curvature tensor, both derived later in the module.

A ring of free particles falling toward a mass deforms into an ellipse: radial pairs separate as the near ones fall faster, transverse pairs converge as both aim at the centre. This tidal stretch is the field variation the free-fall frame cannot cancel.

The size of the tidal effect sets the meaning of sufficiently small in EEP. Over a region of extent $L$ and a time $T$ , the residual relative acceleration a freely falling observer cannot remove is of order $(\partial^{2} Φ) L$ , and the displacement it produces is of order $(\partial^{2} Φ) L T^{2}$ . The frame counts as inertial only to the accuracy set by that residue: the more precise the experiment, the smaller the patch over which gravity has been abolished. A laboratory is local when tidal displacements stay below the resolution of its instruments.

Why one patch is not enough

Local flatness holds at every event, yet the collection of local inertial frames cannot be stitched into one global inertial frame. The obstruction is the tidal field itself. If a single global inertial frame existed, the second derivatives $\partial_{i} \partial_{j} Φ$ would vanish everywhere, meaning no tidal forces and no gravitating source. A nonzero curvature is precisely the statement that the local frames at different events are rotated and boosted relative to one another in a way that cannot be undone globally.

Local inertial frames exist at every event (small flat patches), but a real field twists neighbouring patches relative to one another. The mismatch that survives all patch-to-patch matching is curvature; it cannot be combed flat.

The programme of the module is now set. EEP guarantees a Minkowski metric in a small neighbourhood of every event, which is the local structure of a four-dimensional manifold. The failure of the local frames to align globally is measured by curvature, built from the way vectors change under parallel transport. Free-fall worldlines become geodesics of that geometry, and the tidal equation above becomes the equation of geodesic deviation, whose coefficient is the Riemann tensor. The Einstein field equations then tie that curvature to the matter that sources it.

The Rindler frame as a flat-spacetime rehearsal

Before curvature enters, the uniformly accelerated frame in flat spacetime already displays two hallmarks usually attributed to gravity: a position-dependent clock rate and a horizon. It is the exact special-relativistic version of cabin B, and treating it carefully separates the effects of acceleration (present already in flat spacetime) from the effects of curvature (genuinely new).

An observer with constant proper acceleration $α$ traces the hyperbolic worldline derived in the accelerated-motion lesson, $x^{2} - c^{2} t^{2} = (c^{2} / α)^{2}$ . A whole family of such observers, one for each value of a spatial label, fills a wedge of Minkowski spacetime — the Rindler wedge — with coordinates $(ρ, τ)$ related to the inertial $(t, x)$ by

c t = ρ sinh (\frac{α τ}{c}), x = ρ cosh (\frac{α τ}{c}) .

The flat metric in these coordinates is

d s^{2} = - (\frac{α ρ}{c ^{2}})^{2} c^{2} d τ^{2} + d ρ^{2},

which is not the Minkowski form: the coefficient of $d τ^{2}$ depends on position $ρ$ . Clocks deeper in the wedge (smaller $ρ$ , larger proper acceleration) run slow relative to clocks higher up, mimicking gravitational time dilation exactly. The wedge is bounded by the null lines $x = \pm c t$ , where the $d τ^{2}$ coefficient vanishes: this is the Rindler horizon, a surface the accelerating observers can never receive signals from beyond. Yet the spacetime is flat — every curvature component is zero. The lesson is that horizons and gravitational redshift are consequences of the observer's motion and can occur without curvature; what curvature adds, and Rindler lacks, is tidal force. This distinction organizes the black-hole module, where the Schwarzschild horizon is locally a Rindler horizon while the singularity at the centre is genuine curvature.

The Rindler wedge. A family of uniformly accelerated observers traces nested hyperbolae bounded by the null lines x = ct and x = -ct, which form a horizon; lines of constant Rindler time radiate from the origin. The geometry is flat, yet it carries a horizon and a position-dependent clock rate.

The equivalence principle has taken the universality of free fall and turned it into a demand: physics must be written so that at every event a freely falling observer sees special relativity, while the failure of these local views to agree globally encodes gravity. Meeting that demand requires the geometry of manifolds and a metric, the subject of the next lesson.