Proper Time

Every clock tells the same story about itself and a different story about everyone else. Strap a clock to a rocket, a muon, or a satellite, let it fly a loop through spacetime, and the number it reads on return is not the elapsed coordinate time of any frame — it is the proper time τ, the arc length of that clock's own path measured with the spacetime metric. Coordinate time t is bookkeeping: it depends on which observer slices spacetime into "now" surfaces, and two observers will disagree about it as freely as they disagree about who is moving. Proper time is the opposite kind of object. It is a scalar built directly from the invariant interval, so every observer — inertial, accelerated, or falling through a black hole — computes the same value for the reading of a given clock between two events on its history. That single fact reorganizes relativistic physics. Because τ is invariant, it is the only sensible parameter to differentiate along a worldline, so the four-velocity $u^\mu=\mathrm{d}x^\mu/\mathrm{d}\tau$ and four-acceleration inherit their vector character from it. Because the timelike part of the metric carries a minus sign, adding motion to a worldline subtracts proper time, so the straightest path ages the most — the root of the twin paradox. And because in curved spacetime the metric encodes gravity, a clock's rate depends on where it sits: deeper in a well, τ crawls. This article takes the invariant interval owned by aq-gr-spacetime-interval and the metric structure of aq-gr-minkowski-metric-and-signature, and asks what the timelike arc length of a worldline is, measures, and demands. Time dilation, gravitational redshift, and the geodesic principle all fall out of one line: $c^2\,\mathrm{d}\tau^2=-g_{\mu\nu}\,\mathrm{d}x^\mu\,\mathrm{d}x^\nu$ .

Key takeaways

Proper time τ is the reading of an ideal clock carried along its own worldline — a Lorentz- and diffeomorphism-invariant scalar, identical in every frame.
It is the timelike arc length of the worldline: $c^2\,\mathrm{d}\tau^2=-g_{\mu\nu}\,\mathrm{d}x^\mu\,\mathrm{d}x^\nu$ , so $\mathrm{d}\tau^2=-\mathrm{d}s^2/c^2$ for timelike separations.
In an inertial frame it reduces to $\mathrm{d}\tau=\mathrm{d}t/\gamma$ : a moving clock advances more slowly than coordinate time — the kinematic origin of time dilation.
τ is the natural affine parameter of a timelike worldline and defines the four-velocity $u^\mu=\mathrm{d}x^\mu/\mathrm{d}\tau$ , normalized to $u^\mu u_\mu=-c^2$ .
Free-fall worldlines (timelike geodesics) extremize $\int\mathrm{d}\tau$ ; in flat spacetime the inertial path is the one of maximal proper time — the resolution of the twin paradox.
In curved spacetime the metric carries gravity, so proper time depends on position: a clock at radius r outside a mass reads $\mathrm{d}\tau=\sqrt{1-2GM/c^2 r}\,\mathrm{d}t$ .
Key idea: proper time is the arc length in the timelike direction of a worldline measured by the spacetime metric, $c^2\,\mathrm{d}\tau^2=-g_{\mu\nu}\,\mathrm{d}x^\mu\,\mathrm{d}x^\nu$ — the frame-independent reading of a comoving clock and the parameter along its path.

The core insight

Proper time is the arc length in the timelike direction of a worldline measured by the spacetime metric, $c^2\,\mathrm{d}\tau^2=-g_{\mu\nu}\,\mathrm{d}x^\mu\,\mathrm{d}x^\nu$ ; being a metric invariant it is the frame-independent reading of a comoving clock, and it parametrizes every timelike worldline.

Where this lives today

Proper time long ago stopped being a thought experiment and became an engineering tolerance. The global positioning system runs a live calculation of $\int\mathrm{d}\tau$ along every satellite worldline, correcting the +38 μs/day net rate that would otherwise wreck navigation — the first mass-market technology whose function depends on both special- and general-relativistic time dilation. At the frontier, optical lattice clocks now reach fractional stabilities near $10^{-18}$ , enough to resolve the gravitational proper-time difference between two clocks separated in height by a single centimetre; this has opened "relativistic geodesy", in which a clock's tick rate maps the geopotential to centimetre elevation and tightens tests of whether $\mathrm{d}\tau$ depends on anything besides the metric. And in the strong-field regime, the Event Horizon Telescope and pulsar-timing arrays trace worldlines where the $\sqrt{1-r_s/r}$ factor plunges toward zero, testing the geodesic principle around black holes. From satellite navigation to the centimetre geoid to the shadow of a black hole, the working quantity is always the arc length of a clock's path through spacetime.

The physical picture

Picture the clock's history drawn as a curve — a worldline — through spacetime, and think of τ as the odometer wound along that curve. Just as the distance a car logs between two towns depends on the road it takes, the proper time a clock accumulates between two events depends on the worldline that joins them. There is no single "true" elapsed time between two events; there is only the reading of each specific clock that travels between them. Proper time is a property of the path, not merely of its endpoints. The geometry, though, is not the Pythagorean geometry of a road map. The timelike part of the metric carries a minus sign, so spatial wandering subtracts from the accumulated time instead of adding to it. A clock that detours through high speed logs less proper time, not more — the straightest possible worldline, uniform motion, records the most. This inverted, hyperbolic geometry is the whole surprise of relativistic time, and it is what the first figure makes concrete: two twins leave a common event, one coasting inertially and one looping out and back at v = 0.8c, and they reunite reading different totals — the traveller only 0.6 of the stay-at-home's clock.

Figure

Two worldlines join the same events P and Q: the inertial twin coasts straight up the time axis and ages τ = T, while the twin who loops out and back at v = 0.8c ages only τ = 0.6 T. The dashed amber lines are the light cone from P. Spatial detour subtracts proper time, so the straight geodesic ages the most.

Coordinate time t is a label glued to a frame; proper time τ is a physical reading carried by a clock. The two coincide only for a clock sitting at rest at the origin of a single inertial frame in flat spacetime — the special case that lulls us into thinking there is one universal "now". Confusing the two is the root of nearly every relativistic paradox. In curved spacetime the same arc-length construction holds, but now the metric $g_{\mu\nu}(x)$ varies from place to place, so even a clock bolted down and not moving at all ticks at a rate set by the local gravitational potential. The second figure shows this for the simplest curved geometry — a static clock outside a spherical mass — where the tick rate $\mathrm{d}\tau/\mathrm{d}t$ falls from 1 far away toward 0 at the horizon.

Figure

Tick rate of a static clock outside a spherical mass, dτ/dt = √(1 − r_s/r), versus radius in units of the Schwarzschild radius r_s = 2GM/c². The rate rises from 0 at the horizon toward 1 far away; at r = 2 r_s a clock runs at 0.707 of coordinate rate, at r = 4 r_s at 0.866. Deeper in the well means slower proper time.

The core equations

The whole subject is contained in one definition: proper time is the timelike arc length of a worldline, and the metric is the ruler that measures it. In signature $-+++$ the squared line element is negative for timelike separations, so we flip its sign and divide by $c^2$ to get a positive $\mathrm{d}\tau^2$ . Everything below is a specialization, an integral, or a derivative of this single statement.

c^2\,\mathrm{d}\tau^2 = -g_{\mu\nu}\,\mathrm{d}x^\mu\,\mathrm{d}x^\nu \qquad (1)

where: $\tau$ — proper time, a Lorentz- and diffeomorphism-invariant scalar; $g_{\mu\nu}$ — spacetime metric (Minkowski $\eta_{\mu\nu}$ in special relativity, curved in general relativity); $\mathrm{d}x^\mu$ — infinitesimal coordinate displacement along the worldline; $c$ — speed of light. The right side is $-\mathrm{d}s^2$ , so $\mathrm{d}\tau^2=-\mathrm{d}s^2/c^2$ ; the construction only makes sense for timelike $\mathrm{d}s^2<0$ , where a real clock can travel. Specialize to an inertial frame in flat spacetime by setting $g_{\mu\nu}=\eta_{\mu\nu}=\mathrm{diag}(-1,+1,+1,+1)$ and $x^\mu=(ct,x,y,z)$ . The time component's minus sign flips $c^2\mathrm{d}t^2$ to a positive contribution, and the spatial displacements enter with a physical minus. This is the form that shows the competition between "ticking forward in time" and "moving through space".

c^2\,\mathrm{d}\tau^2 = c^2\,\mathrm{d}t^2 - \mathrm{d}x^2 - \mathrm{d}y^2 - \mathrm{d}z^2 \qquad (2)

where: $t$ — inertial coordinate time; $\mathrm{d}x,\mathrm{d}y,\mathrm{d}z$ — spatial coordinate displacements of the clock. Reading it physically: if the clock does not move ( $\mathrm{d}x^i=0$ ) then $\mathrm{d}\tau=\mathrm{d}t$ ; any spatial motion strictly reduces $\mathrm{d}\tau$ below $\mathrm{d}t$ . Factor $c^2\mathrm{d}t^2$ out of equation (2) and identify the coordinate speed $v=|\mathrm{d}\mathbf{x}/\mathrm{d}t|$ . What remains is the most familiar face of proper time — the Lorentz factor relating a moving clock's reading to coordinate time.

\mathrm{d}\tau = \mathrm{d}t\,\sqrt{1-\frac{v^2}{c^2}} = \frac{\mathrm{d}t}{\gamma} \qquad (3)

where: $v$ — coordinate speed of the clock; $\gamma=1/\sqrt{1-v^2/c^2}$ — Lorentz factor. Physical reading: a clock moving at speed v advances at rate $1/\gamma<1$ relative to coordinate time. This is not an illusion of the clock but a statement about the length of its worldline; it is time dilation seen as geometry. To get a finite reading between two events, integrate along the worldline against any convenient parameter λ. The result is manifestly reparametrization-invariant — change λ and the Jacobian cancels — so it depends only on the geometric path, exactly as arc length should.

\tau = \frac{1}{c}\int_{\lambda_1}^{\lambda_2}\sqrt{-g_{\mu\nu}\,\frac{\mathrm{d}x^\mu}{\mathrm{d}\lambda}\,\frac{\mathrm{d}x^\nu}{\mathrm{d}\lambda}}\;\mathrm{d}\lambda \qquad (4)

where: $\lambda$ — any parameter labeling points along the worldline; $\mathrm{d}x^\mu/\mathrm{d}\lambda$ — tangent vector to the worldline. Physical reading: τ is a functional of the whole path between the endpoints, the timelike analogue of the length of a curve — which is why different worldlines between the same events give different totals. Because τ is a scalar, differentiating position with respect to it produces a genuine four-vector, the four-velocity. Dividing the definition (1) through by $\mathrm{d}\tau^2$ fixes its invariant magnitude once and for all.

u^\mu \equiv \frac{\mathrm{d}x^\mu}{\mathrm{d}\tau},\qquad u^\mu u_\mu = -c^2 \qquad (5)

where: $u^\mu$ — four-velocity, the unit tangent (per unit proper time) to the worldline. Physical reading: every massive particle moves through spacetime at the same "speed" $c$ in the sense $u^\mu u_\mu=-c^2$ ; motion merely rotates that fixed four-vector between the time and space directions. (In signature $+---$ the norm reads $+c^2$ .) In general relativity nothing about the definition changes except that $g_{\mu\nu}$ now varies with position. For a clock held fixed outside a spherically symmetric mass, only the time-time component survives, and the clock's rate carries the gravitational potential directly.

\mathrm{d}\tau = \sqrt{1-\frac{2GM}{c^2 r}}\;\mathrm{d}t \qquad (6)

where: $G$ — Newton's constant; $M$ — mass of the central body; $r$ — areal radial (Schwarzschild) coordinate. Physical reading: the tick rate falls as r shrinks, reaching zero at the horizon $r_s=2GM/c^2$ ; expanding for weak fields gives $\mathrm{d}\tau\approx(1+\Phi/c^2)\,\mathrm{d}t$ with $\Phi=-GM/r$ , the gravitational-redshift law.

Derivation AWhy proper time is the same in every inertial frame

The entire claim that τ deserves to be called a physical clock reading rests on one property: its value cannot depend on who computes it. If two inertial observers, related by an arbitrary Lorentz boost or rotation, disagreed about the elapsed proper time on a given clock, then τ would be just another frame-dependent coordinate and no better than t. The goal here is to show, from the defining property of Lorentz transformations alone, that an infinitesimal proper-time interval computed in one inertial frame equals the interval computed in any other — so that integrating along the worldline yields a single invariant number.

Step 1. Take a timelike segment $\mathrm{d}x^\mu$ and transform to a second inertial frame by $\mathrm{d}x'^\alpha=\Lambda^\alpha{}_\mu\,\mathrm{d}x^\mu$ . Why: Lorentz transformations act linearly on coordinate displacements, so the segment in the new frame is a fixed matrix acting on the old one.

Step 2. Invoke the defining property of a Lorentz matrix, $\Lambda^\alpha{}_\mu\,\Lambda^\beta{}_\nu\,\eta_{\alpha\beta}=\eta_{\mu\nu}$ . Why: this is precisely the statement that Λ preserves the Minkowski metric — it is what makes a transformation "Lorentz" in the first place, the matrix form of $\Lambda^{\mathsf T}\eta\Lambda=\eta$ .

Step 3. Compute the interval in the primed frame and collapse the two Λ factors using Step 2:

-\eta_{\alpha\beta}\,\mathrm{d}x'^\alpha\,\mathrm{d}x'^\beta = -\eta_{\alpha\beta}\,\Lambda^\alpha{}_\mu\,\Lambda^\beta{}_\nu\,\mathrm{d}x^\mu\,\mathrm{d}x^\nu = -\eta_{\mu\nu}\,\mathrm{d}x^\mu\,\mathrm{d}x^\nu.

Why: the metric-preserving identity converts the transformed interval straight back into the original one, with no leftover dependence on the boost velocity.

Step 4. Hence $c^2\,\mathrm{d}\tau'^2=c^2\,\mathrm{d}\tau^2$ , so $\mathrm{d}\tau$ — and, integrating along the worldline, the total $\Delta\tau$ — is identical in both frames. Why: an invariant infinitesimal integrates to an invariant total, so all inertial observers agree on the reading of the clock. Verified against $\Lambda^{\mathsf T}\eta\Lambda=\eta$ ; cf. Schutz §2.2, Carroll §1.3.

Go deeper · Derivation B · 2 minThe twin paradox and the maximal aging of the inertial path

Derivation BThe twin paradox and the maximal aging of the inertial path

The most famous consequence of proper time is that two clocks which part and reunite generically read different totals, with the unaccelerated one always ahead. This is not a puzzle to be explained away but a theorem about which timelike path between two events is longest. The goal is to compute the proper time along each of two worldlines joining a departure event P and a reunion event Q at the same spatial location — an inertial "stay-at-home" twin A and a "traveller" twin B who flies out at speed v and returns — and to show that the inertial worldline strictly maximizes the accumulated proper time, resolving the apparent symmetry.

Step 1. For the stay-at-home twin A, $v=0$ throughout, so equation (3) gives $\mathrm{d}\tau_A=\mathrm{d}t$ and $\tau_A=\int_P^Q\mathrm{d}t=T$ , the full coordinate time between the events. Why: a clock at rest at the origin of an inertial frame is the one case where proper and coordinate time coincide, setting the benchmark all other paths are compared against.

Step 2. Twin B travels at constant speed v on the outbound and inbound legs (the turnaround is taken to be negligibly short), so on each leg $\mathrm{d}\tau_B=\sqrt{1-v^2/c^2}\,\mathrm{d}t$ and

\tau_B=\int_0^{T/2}\!\!\sqrt{1-\tfrac{v^2}{c^2}}\,\mathrm{d}t+\int_{T/2}^{T}\!\!\sqrt{1-\tfrac{v^2}{c^2}}\,\mathrm{d}t=T\sqrt{1-\frac{v^2}{c^2}}.

Why: the same Lorentz factor multiplies both legs, so the traveller's total is just T scaled down by $\sqrt{1-v^2/c^2}<1$ — the traveller ages less.

Step 3. There is no contradiction from "each sees the other's clock slow", because the situation is not symmetric: twin B's worldline is bent — it accelerates at turnaround and is therefore non-inertial — while twin A's is straight. Why: only inertial worldlines are geodesics; the acceleration that B must feel to return is the physical asymmetry that singles out which twin is younger.

Step 4. Geometrically, among all timelike worldlines joining two events, the unaccelerated (geodesic) one maximizes $\int\mathrm{d}\tau$ — the timelike mirror of "a straight line is the shortest distance", with the inequality reversed by the metric's minus sign. Why: spatial detour adds negative-signed terms to the interval, subtracting proper time. Verified: at v = 0.8c, $\tau_B=0.6\,T$ ; MTW §1.5, Hartle ch. 5.

Go deeper · Derivation C · 1 minTimelike geodesics extremize proper time

Derivation CTimelike geodesics extremize proper time

In general relativity, free fall is not a force but a statement about worldlines: a test body released in a gravitational field follows the path that makes its own clock reading stationary. Making this precise recovers the geodesic equation directly from the proper-time functional, tying the measurable quantity τ to the dynamical law of motion. The goal is to vary the proper-time action $S=-mc\int\mathrm{d}\tau$ for a test particle of mass m over all nearby worldlines with fixed endpoints, and to show the stationary paths obey the geodesic equation of the Levi-Civita connection.

Step 1. Write the action as a length functional with Lagrangian $L=\sqrt{-g_{\mu\nu}(x)\,\dot x^\mu\dot x^\nu}$ , where $\dot x^\mu=\mathrm{d}x^\mu/\mathrm{d}\lambda$ for an arbitrary parameter λ, so that $\tau\propto\int L\,\mathrm{d}\lambda$ . Why: because τ is reparametrization-invariant, we may extremize it with any parameter and fix a convenient one only at the end.

Step 2. Impose stationarity through the Euler–Lagrange equations,

\frac{\mathrm{d}}{\mathrm{d}\lambda}\frac{\partial L}{\partial\dot x^\mu}-\frac{\partial L}{\partial x^\mu}=0.

Why: extremizing the arc length between fixed events is exactly the variational condition the free-fall worldline must satisfy.

Step 3. After varying, choose proper time itself as the parameter, $\lambda=\tau$ , so that $L=c$ is constant along the path and the square-root's awkward derivatives simplify to an affine form. Why: an affine parameter removes the reparametrization slack and puts the equation in its cleanest, second-order shape.

Step 4. The equations collapse to the geodesic equation

\frac{\mathrm{d}^2x^\mu}{\mathrm{d}\tau^2}+\Gamma^\mu{}_{\alpha\beta}\,\frac{\mathrm{d}x^\alpha}{\mathrm{d}\tau}\,\frac{\mathrm{d}x^\beta}{\mathrm{d}\tau}=0, \qquad \Gamma^\mu{}_{\alpha\beta}=\tfrac12 g^{\mu\sigma}\!\left(\partial_\alpha g_{\sigma\beta}+\partial_\beta g_{\sigma\alpha}-\partial_\sigma g_{\alpha\beta}\right).

Why: the connection Γ is generated entirely by derivatives of the metric, so gravity enters purely through how the proper-time ruler varies from point to point; in Minkowski coordinates Γ = 0 and geodesics are straight lines. Verified: Wald §3.3, Carroll §3.3.

Limiting cases

How the proper-time increment $\mathrm{d}\tau$ reduces in the regimes that matter:

Case	Result	Meaning
Clock at rest, inertial frame	$\mathrm{d}\tau=\mathrm{d}t$	With $\mathrm{d}x^i=0$ and $g_{\mu\nu}=\eta_{\mu\nu}$ , proper time equals coordinate time — the only case where they agree.
Slow motion (Newtonian)	$\mathrm{d}\tau\approx\mathrm{d}t\left(1-\dfrac{v^2}{2c^2}\right)$	First term of $\sqrt{1-v^2/c^2}$ for $v\ll c$ ; the tiny deficit is the leading kinematic correction to a clock's rate.
Photon / null worldline	$\mathrm{d}\tau=0$	Along a light ray $\mathrm{d}s^2=0$ , so no proper time elapses — light has no rest frame and no clock can ride it.
Weak static gravity	$\mathrm{d}\tau\approx\mathrm{d}t\left(1+\dfrac{\Phi}{c^2}\right)$	Clock at rest in potential $\Phi<0$ ; deeper in the well (Φ more negative) the clock ticks slower — gravitational time dilation.
Horizon limit	$\mathrm{d}\tau\to 0$ as $r\to r_s$	The static tick rate $\sqrt{1-r_s/r}$ vanishes at $r_s=2GM/c^2$ ; an infalling clock still reads finite τ crossing it.

Common trapThere is one true elapsed time between two events, and it is just Δt.

There is not. Proper time is path-dependent: two clocks connecting the same pair of events generically read different totals, so "the" elapsed time between events is undefined — only the reading along a specified worldline is. Nor is proper time the difference of coordinate times, $\Delta\tau\ne\Delta t$ except for a clock at rest in flat spacetime. And beware the sign: because the timelike interval carries a minus on the spatial part, the elapsed proper time is maximized along the inertial or geodesic path, not minimized — the twin who feels no acceleration is the one who ages most, the reverse of the "shortest path" intuition from ordinary geometry. Finally, beware of splicing legs measured in different frames: proper time itself is additive along a worldline — each leg's Δτ is invariant, so successive legs simply sum — but coordinate times and surfaces of simultaneity belonging to different frames do not splice, which is how the naive twin-paradox accounting loses time; integrate $\mathrm{d}\tau$ continuously along the one worldline.

Worked examples

1 · The twin paradox in years (confidence builder)

Twin A stays on Earth; twin B flies to a star and back at constant speed v = 0.8c, returning after T = 10 years of Earth (coordinate) time. How much does each age, and by how much do they differ on reunion?

Step 1. Twin A is (approximately) inertial and at rest, so proper time equals coordinate time: $\tau_A=T=\mathbf{10\ \text{years}}$ . This is the benchmark path — the longest timelike worldline between departure and reunion.

Step 2. Twin B moves at v = 0.8c, giving $\gamma=1/\sqrt{1-0.64}=1/\sqrt{0.36}=1/0.6=1.667$ . Proper time along B's worldline is $\tau_B=T\sqrt{1-v^2/c^2}=10\times0.6=\mathbf{6\ \text{years}}$ .

Step 3. On reunion the traveller has aged 6 years against the stay-at-home's 10, a difference of $\mathbf{4\ \text{years}}$ . The asymmetry is physical, not verbal: B fired engines at turnaround and felt the acceleration; A never did. The stay-at-home clock, being unaccelerated, logs the maximal proper time. τ_A = 10 yr, τ_B = 6 yr, traveller younger by 4 yr.

2 · Why muons reach the ground

Cosmic-ray muons are created high in the atmosphere and decay with a rest-frame lifetime τ₀ = 2.2 μs. Take a muon born at 10 km altitude moving straight down at v = 0.98c. What proper time does it experience on the trip, how many lifetimes is that, and what fraction survives — versus the naive non-relativistic estimate?

Step 1. Lorentz factor: $\gamma=1/\sqrt{1-0.98^2}=1/\sqrt{0.0396}=\mathbf{5.03}$ . The coordinate (lab-frame) travel time for the 10 km descent is $\Delta t=L/v=10\,000/(0.98\times2.998\times10^8)=\mathbf{34.0\ \mu s}$ .

Step 2. The muon's own clock reads the proper time $\Delta\tau=\Delta t/\gamma=34.0/5.03=\mathbf{6.77\ \mu s}$ . Decay is governed by this proper time, not by lab time: in units of the rest lifetime that is $6.77/2.2=\mathbf{3.08\ \text{lifetimes}}$ , so the survival fraction is $e^{-3.08}=\mathbf{0.046}$ — about 4.6 % make it to the ground.

Step 3. The naive estimate ignores dilation and uses the full 34.0 μs against the 2.2 μs lifetime: $34.0/2.2=15.5$ lifetimes, $e^{-15.5}=1.9\times10^{-7}$ — essentially none. The relativistic answer is larger by a factor $e^{12.4}\approx2.4\times10^5$ . The gap between $\tau$ and $t$ is not a curiosity here; it is why cosmic-ray muons are detected at sea level at all. Δτ = 6.77 μs ≈ 3.08 lifetimes, 4.6 % survive, a 10⁵× enhancement over the naive prediction.

3 · The two clock effects in GPS

A GPS satellite orbits at r = 26 560 km from Earth's centre (Earth's radius R = 6371 km) with orbital speed v = 3.87 km/s. Compute the special-relativistic and gravitational rate shifts of its clock relative to a ground clock, their net, and the positioning error if left uncorrected for a day.

Step 1. Special-relativistic (velocity) effect. The satellite moves, so its clock runs slow by $\Delta(\text{rate})_{\text{SR}}=-v^2/2c^2=-(3874)^2/(2\times8.988\times10^{16})=-8.35\times10^{-11}$ . Over one day ( $86\,400$ s) that is $-8.35\times10^{-11}\times86\,400=\mathbf{-7.2\ \mu s/day}$ — the orbiting clock loses about 7 μs each day.

Step 2. Gravitational (potential) effect. The satellite sits higher in the well, where clocks run fast: $\Delta(\text{rate})_{\text{GR}}=(GM/c^2)(1/R-1/r)=(4.435\times10^{-3})(1.570\times10^{-7}-3.77\times10^{-8})=+5.29\times10^{-10}$ , i.e. $+5.29\times10^{-10}\times86\,400=\mathbf{+45.7\ \mu s/day}$ — the high clock gains about 46 μs each day.

Step 3. Net and consequence. The two effects add: $-7.2+45.7=\mathbf{+38.5\ \mu s/day}$ , so an uncorrected satellite clock runs fast by ~38.5 μs daily. Since positioning multiplies timing error by c, that is a range error of $c\times38.5\ \mu s\approx\mathbf{11\ km\ per\ day}$ — GPS would be useless within minutes. The onboard clocks are deliberately offset in frequency before launch to cancel exactly this $\int\mathrm{d}\tau$ mismatch between orbiting and ground worldlines. SR −7.2, GR +45.7, net +38.5 μs/day, ≈ 11 km/day if uncorrected.

Physics of the portrait

Proper time is where nineteenth-century geometry became twentieth-century physics. The reading of a moving clock had been a paradox in Einstein's 1905 kinematics; it became a length — the arc length of a worldline — the moment spacetime was recognized as a geometry with its own metric, and that recognition belongs to one mathematician.

Origins · 1 minHermann Minkowski and how this was discovered

Hermann Minkowski

1864–1909 · the geometry of spacetime and proper time

Hermann Minkowski was Einstein's former mathematics teacher at the Zurich Polytechnic — a man who had thought the young Einstein "a lazy dog". Yet it was Minkowski who, in his 1908 Cologne address Raum und Zeit, gave special relativity the geometric form it has worn ever since, declaring that "henceforth space by itself, and time by itself, are doomed to fade away into mere shadows, and only a kind of union of the two will preserve an independent reality." He recast the Lorentz transformations as rotations of a four-dimensional manifold carrying the indefinite interval $\mathrm{d}s^2=-c^2\mathrm{d}t^2+\mathrm{d}x^2+\mathrm{d}y^2+\mathrm{d}z^2$ , and identified the proper time — his Eigenzeit — of a moving body as the arc length of its worldline. This was the decisive move: time dilation stopped being a strange fact about clocks and became the elementary geometry of timelike curves, exactly as distance is the geometry of spacelike ones. Minkowski died suddenly of appendicitis in 1909, at forty-four, months after the lecture. Einstein first dismissed the reformulation as "superfluous learnedness", but within a few years it had become indispensable: general relativity is unthinkable without it, since curving Minkowski's flat metric into $g_{\mu\nu}(x)$ is precisely what turns proper time into a probe of gravity. The four-vectors, the light cone, the invariant interval, and the very word "worldline" are all his legacy.

History

1905

— Einstein's special relativity establishes the relativity of simultaneity and time dilation

1908

— Minkowski recasts special relativity as spacetime geometry and defines proper time (Eigenzeit)

1911

— Einstein predicts gravitational time dilation from the equivalence principle

1959

— Pound and Rebka measure the gravitational shift of clock rate over 22.5 m at Harvard

1971

— Hafele and Keating fly atomic clocks around the world, confirming the velocity and potential effects together

1977

— CERN muon storage ring confirms time dilation of the muon lifetime to better than one percent

Going further

Texts. Misner, Thorne & Wheeler, Gravitation, §1.5–1.6 and §2.8 · Wald, General Relativity, §3.3 and §4.2 · Carroll, Spacetime and Geometry, §1.2–1.3 and §3.3 · Hartle, Gravity, ch. 5 · Schutz, A First Course in General Relativity, §1.6 and §2.2. All equations and results use the $-+++$ signature throughout. Key self-consistency checks: $\mathrm{d}\tau=\mathrm{d}t/\gamma$ follows from factoring the Minkowski line element; frame-invariance follows from $\Lambda^{\mathsf T}\eta\Lambda=\eta$ ; the four-velocity satisfies $u^\mu u_\mu=-c^2$ and differentiating gives $a^\mu u_\mu=0$ ; the twin result is $\tau_B=0.6\,T$ at v = 0.8c; muon survival is 4.6 % at v = 0.98c over 10 km ( $\Delta\tau=6.77\ \mu$ s); GPS clock rates are SR $-7.2$ , GR $+45.7$ , net $+38.5$ μs/day.

Prerequisites. Spacetime Interval · Minkowski Metric and Signature · Lorentz Transformations · Causal Structure and Light Cones. Leads to. Four-Vectors · Time Dilation · Proper Length.

Ask the tutor: ask me to compute the proper time along any worldline you like — a twin's round trip, an infalling clock near a black hole, or the two competing clock effects that keep GPS honest.

What is the difference between proper time and coordinate time?

Coordinate time t is a label attached to a chosen reference frame — one of the four numbers you use to name an event — and different observers assign different t-values to the same pair of events. Proper time τ is the reading of a specific physical clock that actually travels between the events along a definite worldline; every observer, whatever their motion, computes the same τ for that clock because it is built from the invariant interval. They coincide only in one situation: a clock sitting at rest at the origin of a single inertial frame in flat spacetime. In every other case — a moving clock, an accelerated clock, or any clock in a gravitational field — $\Delta\tau\ne\Delta t$ , and it is τ, never t, that a real clock displays.

Why does the inertial twin age more, not less?

Because the timelike interval carries a minus sign on the spatial part, so adding spatial motion to a worldline subtracts from its accumulated proper time. The stay-at-home twin follows a straight (geodesic) worldline, which is the longest timelike path between the two events, and therefore ages the most. The traveller's worldline is bent — it detours through space and must accelerate to turn around — so it logs less proper time. This is the exact opposite of ordinary geometry, where a straight line is the shortest distance; the sign flip in the metric reverses the inequality. The acceleration the traveller feels is the physical marker that breaks the apparent symmetry between the twins.

Can proper time ever be zero or negative?

Proper time is zero along a null (light-like) worldline, where $\mathrm{d}s^2=0$ : a photon accumulates no proper time between any two points on its path, which is one way of saying light has no rest frame and cannot carry a clock. Proper time is never negative — it is defined as a positive square root, and by convention it increases toward the future along any timelike worldline. For spacelike separations the quantity $-g_{\mu\nu}\mathrm{d}x^\mu\mathrm{d}x^\nu$ turns negative and no real proper time exists; there the natural invariant is proper length instead. So the construction is reserved for the timelike curves that physical clocks can actually follow.

Why is proper time the right parameter for four-velocity?

Because differentiating with respect to an invariant produces an object with clean transformation properties. The four-velocity $u^\mu=\mathrm{d}x^\mu/\mathrm{d}\tau$ is the ratio of a four-vector displacement $\mathrm{d}x^\mu$ to a scalar $\mathrm{d}\tau$ , so it transforms as a four-vector — every component mixes correctly under a Lorentz transformation. If you instead divided by coordinate time t, which is itself frame-dependent, the result (the ordinary three-velocity extended naively) would not transform as a four-vector and the whole tensor machinery of relativistic mechanics would break. Using τ also fixes the normalization $u^\mu u_\mu=-c^2$ automatically, which in turn forces the four-acceleration to be orthogonal to the four-velocity — the backbone of relativistic dynamics.

How does gravity change proper time if the formula looks the same?

The defining equation $c^2\mathrm{d}\tau^2=-g_{\mu\nu}\mathrm{d}x^\mu\mathrm{d}x^\nu$ is identical in special and general relativity; what changes is the metric $g_{\mu\nu}$ . In special relativity it is the constant Minkowski metric, so only motion affects τ. In general relativity $g_{\mu\nu}(x)$ varies from place to place, carrying the gravitational field, so even a clock at rest ticks at a position-dependent rate. For a static clock outside a spherical mass only $g_{tt}$ survives, giving $\mathrm{d}\tau=\sqrt{1-2GM/c^2 r}\,\mathrm{d}t$ : the deeper in the well (smaller r), the slower the clock. Expanding for weak fields recovers $\mathrm{d}\tau\approx(1+\Phi/c^2)\mathrm{d}t$ — gravitational time dilation and redshift, from the same one line.

Is proper time actually measurable, or just a bookkeeping device?

It is directly measurable — it is the most directly measurable time there is, because it is literally what a clock reads. The Hafele–Keating experiment flew caesium clocks around the world and compared their readings against a ground clock, finding exactly the predicted proper-time differences from combined velocity and altitude effects. GPS satellites carry clocks whose proper time is continuously monitored and corrected. Optical lattice clocks now resolve proper-time differences from a height change of a centimetre. In each case there is no abstraction: two clocks part, follow different worldlines, reunite, and their face values differ by $\int\mathrm{d}\tau_1-\int\mathrm{d}\tau_2$ . Coordinate time, by contrast, is the bookkeeping device; proper time is the physical reading.

What does it mean that geodesics extremize proper time?

A timelike geodesic — the worldline of a freely falling test body — is a stationary point of the proper-time functional $\int\mathrm{d}\tau$ among nearby worldlines with the same endpoints. In flat spacetime and over ordinary separations this stationary point is a genuine maximum: the inertial path ages the most. In curved spacetime it is more precisely a stationary path (locally maximal for nearby variations), and varying $\int\mathrm{d}\tau$ yields the geodesic equation with the Levi-Civita connection. Physically this means free fall is the "natural" motion that a clock's own reading singles out — no force is needed to explain it, only the geometry encoded in how proper time varies from point to point. Gravity, in this language, is just the curvature of the proper-time ruler.