Equivalence of uniform acceleration and gravity

Every smartphone contains a direct, daily-life consequence of the equivalence principle: its accelerometer cannot measure gravity separately from acceleration, only the combination — the specific force $\mathbf{a} - \mathbf{g}$. Sitting on a table it reads $9.81\ \mathrm{m/s^2}$ upward (the normal force), and in free fall it reads zero. Inertial navigation systems in aircraft, submarines, and rockets live with this every second: to recover true acceleration they must subtract a modelled gravity field, and any error in the model is indistinguishable from real motion. Meanwhile GPS is a standing monument to equation (7): satellite clocks run fast by about $45\ \mu\mathrm{s}$ per day from sitting higher in the Earth's potential (minus $7\ \mu\mathrm{s}$ of special-relativistic slowing), and without the general-relativistic correction the system would drift by kilometres within a day. At the precision frontier, the equivalence principle is one of the most stringently tested statements in physics — because any violation would be a crack in general relativity's foundation. The MICROSCOPE satellite compared the free fall of titanium and platinum test masses over two years of orbit and bounded any composition dependence below a part in $10^{15}$; atom interferometers now drop clouds of two different isotopes in the same vacuum chamber and compare their accelerations at the $10^{-12}$ level, with dedicated missions aiming further. The elevator thought experiment of 1907 has become a precision industry.

Everything begins with the transformation to a linearly accelerating frame. Let the frame's origin ride a trajectory $\mathbf{R}(t)$ with acceleration $\mathbf{A} = \ddot{\mathbf{R}}$, axes held parallel to inertial axes (no rotation — that restriction matters, and the last limiting case below shows why). Newton's law then reads, for an observer inside the frame,

where: $\mathbf{a}'$ — acceleration measured in the accelerating frame; $\mathbf{F}_{\text{real}}$ — sum of real (interaction) forces, e.g. contact, spring; $m$ — particle mass (inertial = gravitational); $\mathbf{A}$ — acceleration of the frame's origin relative to an inertial frame. The new term $-m\mathbf{A}$ is the entire price of the frame change: uniform across space, steady if $\mathbf{A}$ is steady, and proportional to $m$. That proportionality is the whole story. Comparing $-m\mathbf{A}$ with Newtonian gravity $m\mathbf{g}$ term by term,

where: $\mathbf{F}_{\text{fict}}$ — fictitious force in the accelerating frame; $\mathbf{g}_{\text{eff}}$ — effective gravitational field that mimics the acceleration. A frame accelerating upward manufactures gravity pointing downward — which is why the rocket cabin feels like home. The identification would fail if gravity coupled to a different mass than inertia does. Writing the two roles separately,

where: $m_{\text{inertial}}$ — mass resisting acceleration in $\mathbf{F}=m_i\mathbf{a}$; $m_{\text{gravitational}}$ — mass coupling to gravity in $\mathbf{F}=m_g\mathbf{g}$. This equality — the weak equivalence principle — is an experimental input, not a theorem of mechanics. Newton tested it with pendulums to a part in $10^3$; Eötvös's torsion balance reached parts in $10^9$; the MICROSCOPE satellite (final results, 2022) found no violation at the $10^{-15}$ level. Put a particle in a true uniform field $\mathbf{g}$ and view it from a frame accelerating at $\mathbf{A}$, and the two uniform fields simply add:

where: $\mathbf{g}$ — true uniform gravitational field; $\mathbf{A}$ — frame acceleration. The mass has cancelled: every body, whatever its composition, follows the same trajectory. Set $\mathbf{A} = \mathbf{g}$ and the right side vanishes — free fall is weightlessness. Set $\mathbf{g} = 0$, $\mathbf{A} = -\mathbf{g}_{\text{home}}$ and you have built a perfect Earth-simulator in deep space. What can not be built is the non-uniformity of a real field. The object that measures this is the tidal tensor,

where: $\Phi$ — gravitational potential, $\mathbf{g}=-\nabla\Phi$; $T^{i}{}_{j}$ — tidal (gravity-gradient) tensor, second derivatives of $\Phi$. A uniform $\mathbf{A}$ is spatially constant, so its gradient is identically zero: whatever survives differentiation is real gravity that no choice of frame removes. Near the Earth's surface the gradient is $2GM/r^3 \approx 3.1\times10^{-6}\ \mathrm{s^{-2}}$ — three microns per second squared of relative acceleration for every metre of separation. For energy bookkeeping, the fictitious force is conservative and comes from a linear potential,

where: $\Phi_{\text{eff}}$ — effective potential energy per unit mass; $\mathbf{r}$ — position in the accelerating frame. Liquid surfaces and equilibrium shapes arrange themselves perpendicular to $\nabla\Phi_{\text{eff}}$ — the working principle behind the tilted-train example below. Einstein's 1907–1911 move was to feed light through the equivalence. A photon climbing a height $h$ inside a rocket accelerating at $A$ arrives at a receiver that has gained speed $At = Ah/c$ during the transit, and the ordinary Doppler effect then dictates a fractional frequency drop. If acceleration and gravity are indistinguishable, the same shift must occur in a gravitational field:

where: $\Delta\nu/\nu$ — fractional frequency shift of light between heights; $h$ — height difference along the acceleration; $c$ — speed of light. For a $22.5\ \mathrm{m}$ tower on Earth this is $2.5\times10^{-15}$ — and it was measured, by Pound and Rebka in 1959, exactly where Einstein's argument put it.

Stand in an elevator that suddenly accelerates upward: you feel heavier, pressed into the floor. Your body cannot tell whether the floor pushed up harder or whether gravity strengthened — and that is not a failure of your inner ear, it is a symmetry of mechanics itself.

Drop a ball and a feather (in vacuum) inside the accelerating rocket: both “fall” to the floor at the same rate, because the floor rushes up to meet them equally. What looks like the universality of free fall is, in this frame, just the geometry of one rigid floor accelerating toward everything at once. The same universality on Earth is a genuine physical fact — cannonball and apple accelerate identically — and it holds precisely because the $m$ in $\mathbf{F} = m\mathbf{a}$ and the $m$ in $\mathbf{F} = m\mathbf{g}$ are the same number. If they differed by even a part per billion for different materials, the elevator disguise would fail, and torsion-balance experiments would have caught it long ago. Now run the argument backwards: jump out of a plane (physics-thought-experiment style), and in free fall everything around you floats. Gravity has not gone anywhere — the Earth still pulls — but your frame accelerates with the pull, and the fictitious force $-m\mathbf{A} = -m\mathbf{g}$ cancels the real one exactly. This is why astronauts float: an orbiting station is not “beyond gravity” (at $400\ \mathrm{km}$ altitude, $g$ is still $8.7\ \mathrm{m/s^2}$, almost $90\%$ of its surface value); it is permanently falling. The catch — and it is the most important fine print in this corner of physics — is that the cancellation is only exact at a single point:

Watch two dust motes released $10\ \mathrm{m}$ apart in real free fall above the Earth: each falls toward the centre of the planet, so their paths converge — after four seconds of falling they are closer by about $0.06\ \mathrm{mm}$. A truly uniform acceleration of the frame can never produce that relative drift. Tides are gravity's unfakeable signature, and the reason every statement of equivalence carries the word local.

1 · The bathroom scale in an elevator (confidence builder)

A $70\ \mathrm{kg}$ person stands on a scale in an elevator accelerating upward at $A = 3.0\ \mathrm{m/s^2}$. What does the scale read, and what uniform gravitational field would produce the same reading in a stationary elevator?

Step 1. Work in the elevator frame. The person is at rest there ($\mathbf{a}' = 0$), acted on by real gravity $-mg$, the normal force $N$, and the fictitious force $-mA$ (downward, since $\mathbf{A}$ is up). Equation (1) gives $0 = N - mg - mA$.

Step 2. Solve: $N = m(g + A) = 70 \times (9.81 + 3.0) = \boxed{897\ \mathrm{N}}$, compared with $687\ \mathrm{N}$ at rest — the scale, calibrated for ordinary gravity, displays an apparent mass of $897/9.81 = 91.4\ \mathrm{kg}$.

Step 3. Equivalence reading: by equation (2) the cabin is indistinguishable from one at rest in a field $g_{\text{eff}} = g + A = 12.81\ \mathrm{m/s^2}$. No experiment confined to the cabin — springs, pendulums, dropped keys — can tell “stronger planet” from “accelerating elevator.”

Step 4. Sanity check the two extremes: $A = 0$ returns $687\ \mathrm{N}$, and $A = -g$ (cable cut, free fall) returns $N = 0$ — weightlessness, the scale reads zero while gravity still acts.

2 · Einstein's tower: redshift from the equivalence, then checked against Pound–Rebka (exam level)

Use the equivalence principle to predict the fractional frequency shift of gamma rays climbing the $h = 22.5\ \mathrm{m}$ Harvard tower used by Pound and Rebka in 1959, and estimate the tidal acceleration across a $2\ \mathrm{m}$ laboratory in free fall near the Earth's surface to show why the prediction is safe from tidal corrections.

Step 1. Replace the tower by a rocket of height $h$ accelerating at $A = g$ in empty space. A photon emitted at the base takes $t \approx h/c$ to reach the top, during which the receiver gains speed $\Delta v = g h / c$ away from the source.

Step 2. First-order Doppler shift: $\Delta\nu/\nu = -\Delta v / c = -gh/c^2$ — a redshift, since the receiver recedes. By the equivalence, a tower at rest in gravity $g$ must show the same shift, equation (7).

Step 3. Numbers: $\dfrac{\Delta\nu}{\nu} = \dfrac{9.81 \times 22.5}{(2.998\times10^8)^2} = \boxed{2.46\times10^{-15}}$ — a redshift of two and a half parts in $10^{15}$. Pound and Rebka resolved it using the Mössbauer effect and confirmed the prediction within experimental error (about ten percent then, one percent in the 1964 refinement).

Step 4. Tidal safety check: across a lab of size $\ell = 2\ \mathrm{m}$, the field varies by $\Delta g \approx (2GM/r^3)\,\ell = 3.1\times10^{-6} \times 2 = 6.2\times10^{-6}\ \mathrm{m/s^2}$, a fractional non-uniformity $\Delta g / g \sim 6\times10^{-7}$. The equivalence-based prediction is uniform-field physics; tidal corrections enter only at this $10^{-7}$ relative level — completely negligible at the experiment's precision.

Step 5. The moral: equation (7) is a classical-frame argument that makes a correct statement about light and time. Following it seriously — clocks lower in a gravitational field run slow — leads out of Newtonian mechanics entirely, into general relativity.

References. Morin, Introduction to Classical Mechanics, Ch. 10 (Accelerating frames of reference) · Goldstein, Poole & Safko, Classical Mechanics, 3rd ed., Ch. 1 & 7 (equivalence principle remarks) · Landau & Lifshitz, Mechanics, Vol. 1, §39 (motion in a non-inertial frame) · Taylor, Classical Mechanics, Ch. 9 (Mechanics in noninertial frames) · Einstein, 'On the Influence of Gravitation on the Propagation of Light' (1911)

Provenance. Equations follow from the standard linearly-accelerating-frame transformation $\mathbf{a}_{\text{in}}=\mathbf{A}+\mathbf{a}'$ (Morin Ch.10; Taylor Ch.9; Landau-Lifshitz §39). Verifier should recompute: (1) the rearrangement to $m\mathbf{a}'=\mathbf{F}-m\mathbf{A}$; (2) cancellation of $m$ giving $\mathbf{a}'=\mathbf{g}-\mathbf{A}$; (3) tilt result $\tan\theta=A/g$ for the accelerating train; (4) tidal scaling $2GM/r^3$ for a $1/r^2$ field. The redshift heuristic $\Delta\nu/\nu=gh/c^2$ is Einstein's 1911 equivalence-principle result. No fabricated empirical constants used; equality of inertial and gravitational mass cited as the experimentally established weak equivalence principle.

Builds on: Non-inertial reference frames · Fictitious (inertial) forces · Linearly accelerating frames Leads to: Effective gravity on the rotating Earth · Equatorial bulge and the figure of the Earth · Centrifugal force

Tidal acceleration across a lab of size $\ell$: $\Delta g \approx (2GM/r^3)\,\ell$ — grows linearly with lab size, falls as $1/r^3$. Fractional redshift grows linearly with height.

$2GM_\oplus/R_\oplus^3 \approx 3.1\times10^{-6}\ \mathrm{s^{-2}}$ (tidal gradient at Earth's surface) · $gh/c^2 = 2.46\times10^{-15}$ for $h = 22.5$ m · WEP verified to $10^{-15}$ (MICROSCOPE, 2022).

Step 1. Identify the frame acceleration $\mathbf{A}$ and add the single force $-m\mathbf{A}$ to every free body in the frame.

Step 2. For statics (bobs, surfaces, scales), replace $\mathbf{g}$ everywhere by $\mathbf{g}_{\text{eff}} = \mathbf{g} - \mathbf{A}$ and solve as an ordinary gravity problem.

Step 3. For scale readings, remember the scale measures the real contact force: $N = m|g + A_{\text{up}}|$ components along the vertical.

Step 4. Ask whether tidal terms matter: estimate $(2GM/r^3)\ell$ against the effect you compute; if comparable, the uniform-field model has expired.

• The fictitious force has no reaction partner — never apply Newton's third law to it.
• Plumb bobs tilt backward (opposite $\mathbf{A}$), not forward.
• “No gravity in orbit” is false: $g \approx 8.7\ \mathrm{m/s^2}$ at ISS altitude; orbit = perpetual free fall.
• Equivalence is local and non-rotating: tides and Coriolis forces are the two things it cannot imitate.
• An accelerometer reads specific force $\mathbf{a}-\mathbf{g}$, never gravity alone.