The Equivalence Principle as Motivation

Drop a hammer and a feather in a vacuum and they hit the floor together. That single fact — known to Galileo, sharpened by Newton into the puzzling equality of two conceptually unrelated masses, and confirmed today to one part in 10¹³ — is the seed from which all of general relativity grows. Special relativity had already dethroned the global inertial frame and welded space and time into Minkowski spacetime, but it kept silent about gravity: a uniform gravitational field has no home in flat spacetime. The equivalence principle is the bridge across that silence, reading the universality of free fall not as a coincidence about matter but as a statement about the arena. Because gravity accelerates every body identically, an observer who falls with it feels nothing at all — Einstein's "happiest thought" — and inside a small enough freely falling laboratory the whole of physics reduces to special relativity. What cannot be scrubbed away by falling is the difference in the field from place to place: two nearby free-fallers drift together or apart, a frame-independent residue called the tidal field, and it is exactly what a curved spacetime encodes. So the principle does more than motivate the theory; it dictates its kinematics. The local inertial frames are the freely falling ones, no global inertial frame survives in the presence of matter, and the irreducible remainder — tidal acceleration between geodesics — is the Riemann curvature tensor. The whole geometric edifice of gravity is the working-out of one dropped feather.

Key takeaways

The equivalence principle promotes the universality of free fall into a structural claim: a freely falling observer feels no gravity, and a small accelerating laboratory in empty space is locally indistinguishable from one at rest in a gravitational field.
Its empirical seed is the equality of inertial and gravitational mass, $m_i = m_g$ , confirmed from Galileo through Eötvös to modern torsion balances to better than one part in 10¹³.
It comes in graded strengths — weak (universality of free fall), Einstein (add local Lorentz and position invariance), and strong (extend to self-gravitating bodies) — and general relativity satisfies the strong form.
Gravity transforms away locally but never globally: the leftover relative acceleration of neighbouring free-fall paths — tidal force — is real, frame-independent gravity, and it becomes spacetime curvature.
A degree-of-freedom count shows a coordinate change flattens the metric to Minkowski and kills its first derivatives at a point, but 20 second-derivative components survive — precisely the Riemann tensor.
Key idea: inertial and gravitational mass are equal, so a freely falling frame locally erases gravity and recovers special relativity; what cannot be erased — the tidal divergence of nearby free-fall paths — is the curvature of spacetime.

Where this lives today

The equivalence principle is simultaneously the most tested idea in gravitation and a prime suspect in the search for new physics. Eöt-Wash torsion balances and the MICROSCOPE satellite (2016–2018) have pushed the equality of free-fall accelerations to a few parts in 10¹⁵, while lunar laser ranging bounces pulses off Apollo retroreflectors to test whether the Earth and Moon — bodies with real self-gravity — fall alike toward the Sun, probing the strong principle through the Nordtvedt effect. Experimenters press because almost every attempt to unify gravity with quantum physics — moduli, light scalars, dark-energy models — tends to introduce a new long-range interaction that couples to composition and would appear as a tiny violation; a confirmed violation at one part in 10¹⁶ would be a first-magnitude discovery. And the principle is quietly load-bearing in everyday technology: the gravitational redshift amounts to a clock-rate offset of about $+45\ \mu\mathrm{s}$ per day for a GPS satellite, which with the $-7\ \mu\mathrm{s}$ of velocity time dilation must be corrected or positions drift by kilometres within a day. The elevator thought experiment is wired into the phone in your pocket.

The physical picture

Seal an observer in a windowless box and ask two questions they must answer with local apparatus alone: am I accelerating, and is there gravity? The equivalence principle claims these have no separate answers. Set the box at rest on Earth, and released objects fall at $g = 9.8\ \mathrm{m/s^2}$ while the observer feels weight underfoot. Take an identical box into deep space and fire an engine so it accelerates at exactly $g$ : objects again "fall" at $9.8\ \mathrm{m/s^2}$ and the observer is pressed to the floor with the same weight. Nothing measured inside distinguishes the two. Conversely, cut the cable and box and contents accelerate downward together — objects hover, light crosses in a straight line, the observer floats — locally identical to drifting, engines off, in deep space. Gravity, at a point, is a choice of frame.

Figure

Einstein's box: standing at rest in a field g equals accelerating at g in empty space, and both give a dropped ball the same parabola at 9.8 m per second squared — no local experiment tells them apart

The moment the box grows large, the spell breaks — and the way it breaks is the whole point. A real gravitational field is the field of a mass, so it points toward that mass and weakens with distance. Two balls released side by side high above the Earth both fall toward the planet's centre, so their paths lean toward each other; a ball released higher falls a touch less hard, so a vertical pair stretches apart. A uniformly accelerating rocket cannot fake this: its inertial "gravity" is strictly parallel and uniform, and parallel lines neither converge nor stretch. The leftover relative acceleration — convergence sideways, stretch vertically — is the tidal field, and unlike the uniform part no choice of frame removes it. This irreducible residue deserves the name gravity proper, and it is precisely what the curvature of spacetime measures.

Figure

The residue you cannot fall away from: a sphere of free-falling particles is stretched radially and squeezed transversely, with tidal eigenvalues plus 2GMr⁻³ and minus GMr⁻³ — at Earth's surface the radial value is about 3.1×10⁻⁶ per second squared

Why is gravity, alone among forces, expressible this way? Because it accelerates every body identically. The "charge" a body presents to gravity — its gravitational mass — equals the inertial mass that resists acceleration, so the two cancel in the equation of motion and composition drops out. An electric field has no such courtesy: it accelerates a proton and an alpha particle differently, and no single frame makes electromagnetism vanish for everything at once. You can transform gravity away locally precisely because $m_i = m_g$ — and that uniqueness is why gravity gets promoted to geometry while the other interactions remain fields living on that geometry.

The core equations

Newton already wrote the equation whose strangeness launches everything. For a body of inertial mass $m_i$ in a field $\mathbf{g}$ , the gravitational force $m_g \mathbf{g}$ equals $m_i \mathbf{a}$ . Nothing in mechanics demands that the mass measuring inertia equal the mass measuring gravitational coupling — yet experiment says they are.

m_i \, \mathbf{a} = m_g \, \mathbf{g} \qquad (1)

where: $m_i$ — inertial mass, resistance to acceleration in $\mathbf{F} = m_i \mathbf{a}$ ; $m_g$ — gravitational mass, the body's "charge" for gravity; $\mathbf{a}$ — acceleration; $\mathbf{g}$ — local field strength. Physical reading: solving gives $\mathbf{a} = (m_g/m_i)\,\mathbf{g}$ , and since $m_g/m_i$ is one universal constant, every body falls the same way — the weak equivalence principle. The consequence is that a freely falling observer can always erect a frame in which, at their location, spacetime looks exactly like special relativity — the metric there is Minkowski.

\eta_{\mu\nu} = \mathrm{diag}(-1,+1,+1,+1) \qquad (2)

where: $\eta_{\mu\nu}$ — Minkowski metric components in signature $-+++$ ; $\mu,\nu$ — spacetime indices $0,1,2,3$ . Physical reading: the metric of the local inertial frame — the tangent-space stand-in for the freely falling laboratory. But "at their location" is doing heavy lifting. The precise form of the principle is the local flatness theorem: at any point $x_0$ one can choose coordinates so the metric is Minkowski and its first derivatives vanish.

g_{\mu\nu}(x_0) = \eta_{\mu\nu}, \qquad \partial_{\alpha} g_{\mu\nu}(x_0) = 0 \qquad (3)

where: $g_{\mu\nu}(x)$ — the general, possibly curved, metric; $x_0$ — the point made inertial; $\partial_\alpha$ — coordinate partial derivative. Physical reading: zeroing the first derivatives makes the Christoffel symbols vanish at $x_0$ , so the freely falling observer feels no gravity right where they are. What (3) does not let you zero is the second derivatives — and that obstruction is curvature. Climbing into a freely falling frame is, in Newtonian terms, subtracting the frame's acceleration from the field.

\mathbf{g}_{\mathrm{eff}} = \mathbf{g} - \mathbf{a}_{\mathrm{frame}} \qquad (4)

where: $\mathbf{g}_{\mathrm{eff}}$ — effective field felt inside the frame; $\mathbf{a}_{\mathrm{frame}}$ — acceleration of the frame. Physical reading: choose $\mathbf{a}_{\mathrm{frame}} = \mathbf{g}$ and the effective field vanishes — you have fallen away gravity, exactly only where $\mathbf{g}$ is uniform. Where $\mathbf{g}$ is not uniform, the leftover relative acceleration of neighbouring free-fallers is governed by the Riemann tensor through geodesic deviation.

\frac{D^2 \xi^{\mu}}{d\tau^2} = -R^{\mu}{}_{\nu\rho\sigma}\, u^{\nu}\, \xi^{\rho}\, u^{\sigma} \qquad (5)

where: $\xi^\mu$ — separation vector between two nearby free-fall worldlines; $D^2/d\tau^2$ — second covariant derivative along the worldline; $R^{\mu}{}_{\nu\rho\sigma}$ — Riemann curvature tensor; $u^\nu$ — four-velocity of the reference geodesic; $\tau$ — proper time. Physical reading: the tidal field written geometrically — nonzero Riemann means nearby geodesics accelerate relative to one another, the residue no frame can remove. The same principle predicts, before any field equation, that clocks at different potentials tick at different rates.

\frac{\Delta\nu}{\nu} = \frac{g\,h}{c^2} = \frac{\Delta\Phi}{c^2} \qquad (6)

where: $\Delta\nu/\nu$ — fractional frequency shift between two heights; $g$ — local acceleration; $h$ — height difference; $c$ — speed of light; $\Delta\Phi = gh$ — potential difference. Physical reading: light climbing out of a potential well loses frequency (redshift); clocks deeper in the well run slow. Finally, a freely falling test particle follows the geodesic equation — the curved-space replacement for "straight line at constant velocity".

\frac{\mathrm{d}^2 x^{\mu}}{\mathrm{d}\tau^2} + \Gamma^{\mu}_{\nu\rho}\,\frac{\mathrm{d}x^{\nu}}{\mathrm{d}\tau}\,\frac{\mathrm{d}x^{\rho}}{\mathrm{d}\tau} = 0 \qquad (7)

where: $x^\mu(\tau)$ — worldline of the test particle; $\Gamma^{\mu}_{\nu\rho}$ — Christoffel symbols, the "gravitational force" sent to zero in a local inertial frame; $\tau$ — proper time. Physical reading: in the freely falling frame $\Gamma = 0$ and the particle moves in a straight line — gravity is not a force but the geometry of the paths.

Derivation AGravitational redshift from the equivalence principle alone

The goal is to extract one of general relativity's signature predictions — that light climbing out of a gravitational potential is redshifted — using nothing but the equivalence of a gravitational field with an accelerating frame, no field equations and no metric. This is essentially Einstein's 1911 argument, and its power is that it fixes gravitational time dilation from special-relativistic Doppler physics plus the single postulate of the principle. We replace a static field of strength $g$ by a rocket accelerating upward at $g$ in empty space, let a photon leave the floor and arrive at the ceiling a height $h$ above, and ask what the ceiling detector measures.

Step 1. Work in the momentary inertial frame coinciding with the rocket at emission; the floor emits light of frequency $\nu$ toward the ceiling a height $h$ away. Why: the equivalence principle says the physics inside the accelerating rocket is, instant by instant, that of an inertial frame — ordinary special relativity applies.

Step 2. The photon crosses the gap in time $t \approx h/c$ to leading order, the rocket moving negligibly during transit. Why: we need only the first-order result; corrections are of order $(gh/c^2)^2$ .

Step 3. In that time the ceiling detector, accelerating at $g$ , acquires an upward velocity $v = g\,t = gh/c$ relative to the emission-instant frame, so it is receding from where the light was emitted. Why: this is the crux — the receiver at reception is a different inertial observer than the emitter, boosted by the acceleration, and a moving receiver produces the shift.

Step 4. A receiver receding at speed $v$ sees a first-order Doppler redshift, so

\frac{\Delta\nu}{\nu} = -\frac{v}{c} = -\frac{g h}{c^2} = -\frac{\Delta\Phi}{c^2}.

Why: identifying $gh$ with $\Delta\Phi$ turns a statement about a rocket into one about any weak static field, by the equivalence principle running in reverse.

Step 5. Physically: light climbing through $\Delta\Phi$ is redshifted by $\Delta\Phi/c^2$ , and a clock deeper in the potential runs slow by the same fraction — gravitational time dilation. Pound and Rebka confirmed $\Delta\nu/\nu \approx 2.5\times10^{-15}$ over Harvard's $22.5\ \mathrm{m}$ Jefferson tower in 1959. Why: the same $\Delta\Phi/c^2$ is the leading weak-field term of the Schwarzschild redshift $\nu_\infty/\nu_{\rm emit} = \sqrt{1 - 2GM/rc^2}$ , so elevator and exact metric agree.

Go deeper · Derivation B · 1 minTidal acceleration as the unavoidable residue of gravity

Derivation BTidal acceleration as the unavoidable residue of gravity

The goal is to make precise the claim that free fall erases the uniform part of gravity but cannot erase the whole of it, and to show that the surviving piece — the tidal field — is a second derivative of the potential, the Newtonian shadow of spacetime curvature. This calculation separates "gravity as a coordinate artefact" from "gravity as geometry": the former transforms away, the latter cannot. We take two test masses falling freely in a potential $\Phi$ , separated by a small vector $\boldsymbol{\xi}$ , and follow how that separation evolves.

Step 1. Each particle obeys the same free-fall law $\ddot{x}^i = -\partial_i \Phi(\mathbf{x})$ , with $\Phi = -GM/r$ for a point source. Why: the universality of free fall guarantees both particles obey the identical equation with no mass-dependent factor, so their relative motion isolates the field's structure, not their properties.

Step 2. Subtract the two equations of motion; with the second particle at $\mathbf{x} + \boldsymbol{\xi}$ , the separation obeys

\ddot{\xi}^i = -\big[\,\partial_i\Phi(\mathbf{x}+\boldsymbol{\xi}) - \partial_i\Phi(\mathbf{x})\,\big].

Why: the relative acceleration, not the absolute one, is the frame-independent observable — an accelerating frame shifts both particles equally and cancels in the difference.

Step 3. Taylor expand to first order in the small separation:

\ddot{\xi}^i = -\,\partial_i\partial_j\Phi \; \xi^j.

Why: the uniform part $\partial_i\Phi$ — the very thing a freely falling frame subtracts — cancels exactly, leaving only the symmetric tidal tensor $\partial_i\partial_j\Phi$ : curvature is a second derivative.

Step 4. Evaluate for the point-mass potential, radial and transverse:

\ddot{\xi}^{\,r} = +\frac{2GM}{r^3}\,\xi^{\,r}, \qquad \ddot{\xi}^{\perp} = -\frac{GM}{r^3}\,\xi^{\perp}.

Why: the signs are the physics — positive means radial separations stretch, negative means transverse ones squeeze, and the two transverse eigenvalues sum with the radial one to zero because $\nabla^2\Phi = 0$ in vacuum (trace-free).

Step 5. Match to the covariant equation of geodesic deviation (5): the Newtonian tidal tensor is the weak-field limit of Riemann components,

R^{i}{}_{0j0} \;\longleftrightarrow\; \frac{1}{c^2}\,\partial_i\partial_j\Phi.

Why: the punchline — tidal force is not merely analogous to curvature, it is curvature in the Newtonian limit, so the equivalence principle localizes gravity precisely to the irremovable Riemann tensor.

Limiting cases

Case	Result	Meaning
Vanishing field (deep space)	$g \to 0 \;\Rightarrow\; g_{\mu\nu} \to \eta_{\mu\nu}$	With no gravity the freely falling frame is a genuine global inertial frame; physics is exactly special relativity everywhere.
Local limit (small region)	$g_{\mu\nu}(x) = \eta_{\mu\nu} + O\big((x - x_0)^2\big)$	At a point the metric is Minkowski and its first derivatives vanish; only second derivatives (tidal/curvature) remain, and those cannot be removed.
Uniform field idealization	$R^{\rho}{}_{\sigma\mu\nu} = 0$	A perfectly uniform field has zero curvature and is globally an accelerated frame (Rindler spacetime); its "gravity" is a pure coordinate effect.
Test-mass limit	$m_g/m_i = 1 \pm 10^{-13}$	The weak principle holds when self-gravity is negligible; Eötvös and lunar laser ranging confirm equality to this level. Strongly self-gravitating bodies probe the strong principle.

Common trapTreating "gravity equals uniform acceleration" as a statement true over an extended region

The equivalence is strictly local, and forgetting that word is the classic error. A uniformly accelerating frame produces strictly parallel, uniform inertial accelerations, but the field of any real mass points toward the mass and weakens with distance. Over any extended region, neighbouring free-fall trajectories therefore converge or diverge — tidal effects — and no single accelerated frame reproduces that pattern everywhere at once. The rocket-field indistinguishability holds only in a region small enough, and a time short enough, that these tidal variations stay below your measurement precision. The failure of the equivalence over finite regions is not a defect; it is gravity, and curvature is the exact measure of the obstruction.

Worked examples

1 · Gravitational redshift in the Pound–Rebka tower (fully numeric)

Gamma rays ( $14.4\ \mathrm{keV}$ ) from a $^{57}\mathrm{Fe}$ source climb the $h = 22.5\ \mathrm{m}$ Jefferson tower at Harvard. With $g = 9.81\ \mathrm{m/s^2}$ and $c = 3.00\times10^{8}\ \mathrm{m/s}$ , find the fractional frequency shift.

Step 1. The principle gives $\Delta\nu/\nu = -gh/c^2$ for light climbing up. Insert $gh = (9.81)(22.5) = 220.7\ \mathrm{m^2/s^2}$ — the potential difference $\Delta\Phi$ the photon must "pay for".

Step 2. Divide by $c^2 = 9.00\times10^{16}\ \mathrm{m^2/s^2}$ :

\frac{\Delta\nu}{\nu} = -\frac{220.7}{9.00\times10^{16}} = -2.45\times10^{-15}.

The gamma rays arriving at the top are redshifted by about 2.5 parts in 10¹⁵ — a fantastically small number, which is why the Mössbauer effect (recoil-free emission with an extraordinarily sharp line) was essential to detecting it.

Step 3. Interpret the size: over $22.5\ \mathrm{m}$ a clock at the top runs faster by this same fraction, about $2.5\times10^{-15}\times(3.15\times10^{7}\ \mathrm{s/yr}) \approx 80\ \mathrm{ns}$ per year. Pound and Rebka (1959), and Pound and Snider (1964) to about $1\%$ , confirmed exactly this. The tiny 2.5×10⁻¹⁵ shift, predicted from the elevator argument with no field equations, is a real measured effect.

2 · Curvature is the irreducible residue — the local-flatness count (fully worked)

Count degrees of freedom to show a coordinate change flattens the metric to $\eta_{\mu\nu}$ and kills its first derivatives at a point, but must leave exactly 20 second-derivative components — the Riemann tensor — untouched.

Step 1. Expand the coordinate transformation about the point. Its first-derivative array $\partial x'^{\mu}/\partial x^{\nu}$ has $4\times4 = 16$ free components; use these to set $g_{\mu\nu} = \eta_{\mu\nu}$ ( $10$ conditions, a symmetric matrix). This leaves $16 - 10 = 6$ unused — exactly the dimension of the Lorentz group, the residual local symmetry.

Step 2. The transformation's second derivatives, symmetric in the lower pair, number $4\times10 = 40$ ; use them to impose $\partial_\alpha g_{\mu\nu} = 0$ , also $40$ conditions ( $10\times4$ ). The counts match, so first derivatives always vanish — the mathematics of "the falling observer feels no gravity".

Step 3. Now the metric's second derivatives:

\partial_\alpha\partial_\beta g_{\mu\nu}: \quad 10 \times 10 = 100 \ \text{components}.

The transformation's third derivatives, symmetric in three lower indices, supply only $4\times20 = 80$ parameters. More metric second-derivatives to kill than knobs to turn.

Step 4. The shortfall is $100 - 80 = 20$ components that cannot be transformed away. These 20 are exactly the independent components of the Riemann tensor in four dimensions. Curvature is the irreducible remainder after every coordinate freedom has been spent flattening spacetime at a point. The equivalence principle erases gravity to first order and no further — and the "no further" is general relativity.

Physics of the portrait

The equivalence principle is Einstein's, root and branch — the thread he pulled on for eight years, from a clerk's daydream in Bern to the field equations of 1915.

Origins · 1 minAlbert Einstein and how this was discovered

Albert Einstein

1879–1955 · the equivalence principle and the road to general relativity

In 1907, asked to review relativity, Einstein was struck by a thought he later called "the happiest of my life": a person falling freely from a roof does not feel their own weight. He saw at once that this was not a curiosity but a principle — that a gravitational field and an accelerating frame are locally the same thing — opening a way to bring gravity into relativity. From it, still in 1907, he deduced that clocks run slow deep in a potential and that light is redshifted climbing out of one. In 1911 he sharpened the light-bending prediction from the elevator argument alone, obtaining half the eventual value — the "missing" factor of two would come only once he understood that space itself is curved. Turning a local equivalence into a global theory of a curved manifold consumed him from 1912, working with Marcel Grossmann on Riemannian geometry, to the field equations of November 1915. General relativity is the strong equivalence principle made into a dynamical theory of geometry. Einstein's genius here was not mathematical firepower but physical taste: the refusal to treat $m_i = m_g$ as a coincidence, and the insistence that a fact true of every falling body must be a fact about spacetime.

History

1907

— Einstein's "happiest thought": a freely falling observer feels no gravity; he states the equivalence of gravitation and acceleration

1911

— Einstein predicts gravitational redshift and light bending from the equivalence principle, getting half the final deflection

1915

— the field equations complete the program, turning the strong equivalence principle into a dynamical theory of curved spacetime

1919

— Eddington's eclipse expedition measures light deflection at the solar limb, favouring the full general-relativistic value

1959

— Pound and Rebka measure the gravitational redshift over 22.5 m using the Mössbauer effect, confirming the elevator prediction

2017

— the MICROSCOPE satellite confirms the weak equivalence principle to a few parts in 10¹⁵, the tightest test to date

Going further

Texts. Misner, Thorne & Wheeler, Gravitation (1973), Chs. 1 and 16 (the box, tidal forces, and local flatness) · Wald, General Relativity (1984), Ch. 4 · Carroll, Spacetime and Geometry (2004), Ch. 4 (the local-flatness count) · Hartle, Gravity (2003), Ch. 6 · Schutz, A First Course in General Relativity, 2nd ed. (2009), Ch. 5 · Will, Theory and Experiment in Gravitational Physics (2018) for the modern test landscape.

Provenance. Mass-equality and the weak principle follow MTW Ch. 1 and Schutz Ch. 5, with the Eötvös bound $\sim 10^{-13}$ from Will (2018); MICROSCOPE (2017) tightens it to a few parts in 10¹⁵. The local-flatness counting (16/10/6, 40/40, 100/80/20 leaving 20 Riemann components) is standard in Wald Ch. 4 and Carroll Ch. 4. The redshift derivation reproduces Einstein (1911) and Pound–Rebka (1959), $gh/c^2 \approx 2.5\times10^{-15}$ over $22.5\ \mathrm{m}$ . The tidal coefficients $+2GM/r^3$ (radial) and $-GM/r^3$ (transverse), trace-free in vacuum, follow from $\Phi = -GM/r$ and match the Newtonian limit of $R^i{}_{0j0}$ (Carroll Ch. 4, MTW §1.6). The GPS net rate $+38\ \mu\mathrm{s/day}$ ( $+45$ gravitational, $-7$ velocity) is standard (Ashby, Living Reviews in Relativity, 2003); the $1.75''$ solar deflection follows from Schwarzschild geodesics.

Builds on. Postulates of Special Relativity · Minkowski Spacetime · Minkowski Metric and Signature · Proper Time. Leads to. Geodesics and Free Fall · Riemann Curvature Tensor · Gravitational Time Dilation · Einstein Field Equations.

Ask the tutor: Want to see the local-flatness count turned into an explicit second-order coordinate transformation, or watch the tidal tensor of a specific spacetime become Riemann components step by step? Ask and we'll work it through.

Is the equivalence principle just Galileo's observation that everything falls the same?

That is the seed, but the principle is much stronger. Galileo's fact — the universality of free fall — is the weak equivalence principle, a statement about test-body trajectories. Einstein's leap was that in a freely falling frame all of physics reduces to special relativity: electromagnetism, atomic clocks, particle collisions, everything non-gravitational behaves as in flat spacetime. That Einstein equivalence principle is what forces gravity to be geometry rather than a force field.

If a freely falling observer feels no gravity, is gravity real?

Locally, no — and that is the whole point. At a single event you can always choose a frame in which the gravitational force vanishes, so "the force of gravity at a point" is not frame-independent; it is like a centrifugal force, an artefact of coordinates. What is real is the tidal field: the way nearby free-fallers accelerate relative to one another, which no frame removes. So gravity is real, but its reality lives in the second derivatives of the metric — the curvature — not the first derivatives a falling observer transforms away.

Why does the naive light-bending estimate come out a factor of two too small?

Einstein's 1911 elevator argument treats only the effect of the field on time — the redshift/potential piece — and predicts about $0.87''$ at the solar limb. A full treatment must also account for the curvature of space, which in the weak field contributes equally, doubling the deflection to $1.75''$ . The equivalence principle, being purely local, captures the time part but cannot see the spatial curvature of the extended field. Eddington's 1919 eclipse measurement favoured the full value — a decisive early sign that general relativity, not just the equivalence principle, was needed.

What is the difference between the weak, Einstein, and strong equivalence principles?

They are a ladder of increasing strength. The weak principle (WEP) is the universality of free fall for test bodies: composition does not affect the trajectory. The Einstein principle (EEP) adds local Lorentz invariance and local position invariance, so all non-gravitational physics in a freely falling frame is special-relativistic. The strong principle (SEP) extends the same universality to self-gravitating bodies and gravitational experiments themselves — for example, whether a planet's own binding energy falls the same way as ordinary mass. General relativity uniquely satisfies the strong form; most alternatives violate it, which is why the Nordtvedt effect and lunar laser ranging are such sharp discriminators.

How exactly does tidal force become the Riemann tensor?

In the Newtonian picture the tidal field is $\partial_i\partial_j\Phi$ , driving nearby particles through $\ddot{\xi}^i = -\partial_i\partial_j\Phi\,\xi^j$ . In general relativity the same relative acceleration of neighbouring geodesics is governed by geodesic deviation, $D^2\xi^\mu/d\tau^2 = -R^\mu{}_{\nu\rho\sigma}u^\nu\xi^\rho u^\sigma$ . Comparing the two in the weak, slow-motion limit gives $R^i{}_{0j0} \leftrightarrow \partial_i\partial_j\Phi/c^2$ . So the Riemann tensor is the exact, fully relativistic generalization of the Newtonian tidal tensor — tidal force is curvature, made covariant.

Does the equivalence principle say a uniform gravitational field is exactly a coordinate effect?

Yes, and this is a useful sharp case. A perfectly uniform, static field has zero Riemann tensor, carries no tidal effects, and can be produced entirely by an accelerating coordinate system in flat spacetime — Rindler space, where "gravity" is nothing but a coordinate artefact, removable globally by going to an inertial frame. Real fields are never uniform: the field of any mass falls off with distance, so its curvature is nonzero and cannot be globally transformed away. The uniform-field idealization is the boundary between the transformable (coordinate gravity) and the irreducible (curvature).