Proper Acceleration and the Comoving Frame

Inside the CERN muon storage ring, a muon whips round a seven-metre circle at $\gamma = 29.33$ . Ask what it is enduring: the transverse proper acceleration is $1.1\times10^{19}\ \mathrm{m/s^2}$ — about $10^{18}$ times the surface gravity of the Earth. Nothing a human has built accelerates anything harder. And the muon does not care. Its lifetime is stretched by exactly $\gamma$ — the factor that depends on its speed and nothing else. Bailey and his collaborators measured it in 1977 and found agreement with the pure-velocity formula to two parts in a thousand, with no detectable acceleration dependence at all. That is a result, and it is not a theorem: the claim that an ideal clock's rate depends only on its instantaneous speed — the clock hypothesis — does not follow from the postulates. It is a separate assumption, and the muon ring is where it was put to the sword. Underneath the experiment sits a definition that has to be got exactly right. Coordinate acceleration $d\mathbf{v}/dt$ is frame-dependent and, worse, is not what anything feels. What an accelerometer reads — what presses the crew into their seats — is the acceleration in the inertial frame momentarily comoving with the object: the proper acceleration, an invariant. And the innocent-looking business of relating it to the lab-frame acceleration produces one of the most frequently mis-typed formulae in special relativity.

Key takeaways

Signature $(+,-,-,-)$ : $\eta_{\mu\nu} = \mathrm{diag}(+1,-1,-1,-1)$ , $ds^2 = c^2dt^2 - dx^2$ , so $U\!\cdot\!U = +c^2$ and the four-acceleration is spacelike: $A\!\cdot\!A = -\alpha^2 < 0$ , hence $\alpha = \sqrt{-A_\mu A^\mu}$ .
Proper acceleration $\alpha$ is the acceleration measured in the momentarily comoving inertial frame (MCIF) — an inertial frame that matches the particle's velocity at one instant only. It is what an accelerometer reads, and it is an invariant.
$A\!\cdot\!U = 0$ always: the four-acceleration is orthogonal to the four-velocity, which is what forces it to be spacelike.
Collinear motion: $\alpha = \gamma^{3}\,a$ . Transverse motion: $\alpha = \gamma^{2}\,a_\perp$ . The same lab-frame acceleration is felt $\gamma$ times more strongly when it lies along the motion.
Owned result — and a correction to this article's base. The general identity is $\;\alpha^{2} = \gamma^{6}\!\left[a^{2} - \dfrac{(\mathbf{v}\times\mathbf{a})^{2}}{c^{2}}\right] = \gamma^{4}a^{2} + \gamma^{6}\dfrac{(\mathbf{v}\cdot\mathbf{a})^{2}}{c^{2}}$ — with $\gamma^{6}$ , not $\gamma^{4}$ . The AllQFT base prints $\gamma^{4}$ , which gives $\gamma^{2}a$ collinear and $\gamma a$ transverse — wrong in both limits, by a factor of $\gamma$ , and contradicting its own limits table. Repaired and asserted symbolically.
The clock hypothesis is an assumption, not a theorem. An ideal clock's rate depends on its instantaneous speed and not on its acceleration. Bailey et al. (Nature 268, 301, 1977) tested it at $\sim10^{18}\,g$ in the CERN muon ring and found the Einstein factor good to $2\times10^{-3}$ at 95% confidence, with no acceleration term.
Key idea: acceleration in relativity has two faces, and only one is real. Coordinate acceleration depends on who is watching and dies as $v\to c$ ; proper acceleration is the invariant magnitude of a four-vector, it is what the accelerometer reads, and it can be held constant forever.

The core insight

The whole subject is the sentence "differentiate with respect to proper time, not coordinate time". $U^\mu = dx^\mu/d\tau$ is a four-vector because $\tau$ is invariant; so is $A^\mu = dU^\mu/d\tau$ . Differentiating the constant $U\!\cdot\!U = c^2$ gives $A\!\cdot\!U = 0$ at once, and since $U$ is timelike, $A$ must be spacelike — so $A\!\cdot\!A$ is negative and $\alpha \equiv \sqrt{-A\!\cdot\!A}$ is real. Everything else here, including the $\gamma^3$ , the $\gamma^2$ and the $\gamma^6$ that ties them together, is that one computation done in components.

Where this lives today

Every synchrotron light source is an application of the transverse rule. The radiated power of a charged particle goes as the square of its proper acceleration, so $\alpha = \gamma^2v^2/R$ says at once that the output scales as $\gamma^4/R^2$ — which is why storage rings are built large, why electrons are the particle of choice, and why the LHC collides protons instead. The clock hypothesis is load-bearing infrastructure. It is what makes proper time a purely geometric quantity, $\tau = \frac1c\int ds$ , depending on the path and not on the forces that produced it — without it there is no unambiguous twin paradox, no well-posed ageing of an accelerated observer, and no way to build the Rindler chart of the following articles. Every atomic clock ever flown quietly assumes it, and every one has been consistent with it.

The physical picture

Put an accelerometer on the table. It reads zero when it sits still and zero when it coasts, and something when a rocket pushes it. What it measures is not $d\mathbf{v}/dt$ in any particular frame — it is the deviation of the worldline from a straight line. The instrument does not know which inertial frame you are using, so its reading must be an invariant. Making that precise takes one idea: the momentarily comoving inertial frame. At each instant of the particle's life, choose the inertial frame in which it is at rest at that instant. It will not stay at rest — by the next instant it has drifted — but for that one moment the frame matches, and the acceleration measured in that frame is the proper acceleration. The MCIF is not one frame but a continuous ladder of them, each tangent to the worldline for an instant. The accelerometer rides the ladder. Now the arithmetic that catches everyone. The lab sees the particle accelerating along its motion at $a = dv/dt$ . How big is that in the MCIF? Two effects conspire, and they multiply:

A lab velocity increment $dv$ is a larger increment $dv' = \gamma^2\,dv$ in the MCIF, because velocities compose relativistically — differentiate $\tanh$ and the factor is $\gamma^2$ .
The MCIF's clock runs faster than the particle's, $dt = \gamma\,d\tau$ , so the same increment is achieved in a shorter time — one more factor of $\gamma$ .

Together: $\alpha = \gamma^{2}\cdot\gamma\cdot a = \gamma^{3}a$ . Turn it round and you get the statement that governs every rocket: the lab-frame acceleration is $a = \alpha/\gamma^3$ , and it dies as $v \to c$ even while the accelerometer reading holds steady. That is why constant proper acceleration traces a hyperbola and not a parabola (`aq-sr-relativistic-uniform-acceleration`). But if the acceleration is perpendicular to the velocity, only one effect survives — the transverse velocity transforms with a single factor of $\gamma$ — and the answer is $\alpha = \gamma^{2}a_\perp$ , one power of $\gamma$ smaller. The same lab-frame acceleration is felt $\gamma$ times more strongly along the motion than across it. That asymmetry is why a general formula is needed, and why the general formula is so often mis-copied.

Figure

The ladder of momentarily comoving frames, and what each one measures. The curved worldline is the particle; at each instant there is an inertial frame that matches its velocity for that instant only — the MCIF. Proper acceleration is what an accelerometer reads in the MCIF, and it is an INVARIANT: every inertial observer computes the same α from A·A = −α², even though they all disagree about the coordinate acceleration dv/dt. Note the two orthogonality facts that make it work: U·U = +c² is constant, so differentiating gives A·U = 0, and since U is timelike A must be SPACELIKE — which is why A·A is NEGATIVE and α = √(−A·A). The accelerometer never sees a frame; it sees the bend in the worldline

Figure

PUNCHLINE — the base's formula fails BOTH of its own limits. Plotted is α/a, the factor by which a given lab-frame acceleration is amplified in the comoving frame, against β. The two SOLID curves are correct: γ³ when the push is along the motion, γ² when it is across it — the same lab acceleration is felt γ times harder if collinear. The two DASHED curves are what this article's AllQFT base predicts, its general law carrying γ⁴ where γ⁶ belongs: it returns γ² collinear and γ¹ transverse, each short by exactly one power of γ. Both dashed curves sit a whole factor of γ below the truth, and at the CERN muon ring's γ = 29.33 that is a factor of 29 in the headline proper acceleration — 1.1×10¹⁹ m/s² becomes 3.8×10¹⁷. The base's own limits table prints γ³ and γ². It contradicts itself, and nobody substituted

The core equations

The four-velocity and four-acceleration are objects owned by `aq-sr-four-velocity` and `aq-sr-four-acceleration`; we recall them and go to work.

U^{\mu} = \frac{dx^{\mu}}{d\tau} = \gamma\,(c,\;\mathbf{v}), \qquad A^{\mu} = \frac{dU^{\mu}}{d\tau}, \qquad U\!\cdot\!U = \eta_{\mu\nu}U^{\mu}U^{\nu} = +c^{2} \qquad (1)

where: $\tau$ is proper time. Differentiating the constant $U\!\cdot\!U$ gives at once

\boxed{\;A\!\cdot\!U = 0\;} \qquad\Longrightarrow\qquad A^{\mu}\ \text{is \textbf{spacelike}} \qquad\Longrightarrow\qquad \boxed{\;\alpha \;=\; \sqrt{-\,A_{\mu}A^{\mu}}\;} \qquad (2)

where: the minus sign is forced, not chosen: in $(+,-,-,-)$ a vector orthogonal to a timelike vector is spacelike, so $A\!\cdot\!A < 0$ . Sources written in $(-,+,+,+)$ print $\alpha = \sqrt{+A_\mu A^\mu}$ ; transcribing that into our convention asks for the root of a negative number. Operationally, $\alpha = |\mathbf{a}_{\text{rest}}|$ , the acceleration measured in the MCIF — equation (3). The two special cases everyone actually uses:

\boxed{\;\alpha = \gamma^{3}\,a\quad\text{(collinear, } \mathbf{a}\parallel\mathbf{v})\;} \qquad\qquad \boxed{\;\alpha = \gamma^{2}\,a_{\perp}\quad\text{(transverse, } \mathbf{a}\perp\mathbf{v})\;} \qquad (4)

where: $a = |d\mathbf{v}/dt|$ in the lab. The collinear case picks up $\gamma^2$ from velocity composition and one more $\gamma$ from time dilation; the transverse case picks up $\gamma^2$ in total. Read backwards: a rocket at constant thrust has coordinate acceleration $\alpha/\gamma^3$ , which decays to zero. And the general law, which unifies them. This is the equation this article exists to get right:

\boxed{\;\alpha^{2} \;=\; \gamma^{6}\left[\,a^{2} - \frac{(\mathbf{v}\times\mathbf{a})^{2}}{c^{2}}\,\right] \;=\; \gamma^{4}a^{2} + \gamma^{6}\,\frac{(\mathbf{v}\cdot\mathbf{a})^{2}}{c^{2}}\;} \qquad (5)

where: $\mathbf{a} = d\mathbf{v}/dt$ , and the two forms are algebraically identical (expand $(\mathbf{v}\times\mathbf{a})^2 = v^2a^2 - (\mathbf{v}\cdot\mathbf{a})^2$ ). Physical reading: the exponent is six, not four. Set $\mathbf{v}\times\mathbf{a} = 0$ and (5) gives $\alpha = \gamma^3 a$ ; set $\mathbf{v}\cdot\mathbf{a} = 0$ , so $|\mathbf{v}\times\mathbf{a}| = va$ , and it gives $\alpha^2 = \gamma^6 a^2(1-\beta^2) = \gamma^4 a^2$ , i.e. $\alpha = \gamma^2 a$ . Both limits of (4) fall straight out. A $\gamma^4$ in front of the bracket — which is what this article's AllQFT base prints — reproduces neither. See the warning. Finally, the assumption that lets any of this be measured: > The clock hypothesis. An ideal clock's rate depends on its instantaneous speed and not on its acceleration: $d\tau = dt/\gamma(v)$ , with no $\alpha$ -dependent correction. This does not follow from the postulates. It is a separate hypothesis — Einstein's Uhrenhypothese — and it is an experimental question.

Working it out

Derivation AThe four-acceleration is spacelike, and its magnitude is what the accelerometer reads

The goal is to establish equation (2) and then identify $\sqrt{-A\!\cdot\!A}$ with the reading of a physical instrument — two different claims, and the second is what makes the first useful. The first is a two-line consequence of the constancy of $U\!\cdot\!U$ ; the second requires going into the momentarily comoving frame and looking. Doing both is what turns an index identity into physics an engineer can trust.

Step 1. Start from the normalisation: $U\!\cdot\!U = \gamma^2(c^2 - v^2) = \gamma^2c^2(1-\beta^2) = c^2$ . Why: exactly $+c^2$ , positive, because in $(+,-,-,-)$ a timelike vector has positive norm — and constant along every worldline, which makes the next step free.

Step 2. Differentiate with respect to proper time: $\dfrac{d}{d\tau}(U\!\cdot\!U) = 2\,A\!\cdot\!U = 0$ , so $A\!\cdot\!U = 0$ identically. Why: no assumption about the motion entered. The four-acceleration is orthogonal to the four-velocity for every worldline in special relativity, always.

Step 3. Deduce the causal character. $U$ is timelike, and a nonzero vector orthogonal to a timelike vector must be spacelike. Hence $A\!\cdot\!A < 0$ , and $\alpha \equiv \sqrt{-A_\mu A^\mu}$ is real. Why: the sign guard. If you find yourself writing $\alpha = \sqrt{+A\!\cdot\!A}$ , you have imported a mostly-plus formula. The magnitude is convention-independent; the formula is not.

Step 4. Now go to the MCIF and evaluate it there. In the frame where the particle is momentarily at rest, $\mathbf{v} = 0$ and $\gamma = 1$ , so $U^\mu = (c, \mathbf{0})$ and, since $A\!\cdot\!U = 0$ forces $A^0 = 0$ , the four-acceleration is purely spatial: $A^\mu = (0, \mathbf{a}_{\text{rest}})$ . Then $A\!\cdot\!A = -|\mathbf{a}_{\text{rest}}|^2$ , so $\alpha = |\mathbf{a}_{\text{rest}}|$ . Why: the payoff. The invariant $\sqrt{-A\!\cdot\!A}$ , defined by an index contraction, equals the ordinary three-acceleration in the one frame where the particle is at rest — precisely what an accelerometer strapped to it reports. The abstraction and the instrument agree. *Verified symbolically in verify_aq-sr-proper-acceleration-comoving-frame.py: for a generic symbolic worldline $\mathbf{v}(t)$ , $U\!\cdot\!U = +c^2$ , $A\!\cdot\!U = 0$ and $A\!\cdot\!A < 0$ in $\eta = \mathrm{diag}(+1,-1,-1,-1)$ ; and boosting to the instantaneous rest frame returns $A^\mu = (0,\mathbf{a}_{\text{rest}})$ with $|\mathbf{a}_{\text{rest}}| = \sqrt{-A\!\cdot\!A}$ .*

Go deeper · Derivation B · 2 minThe general law: gamma to the sixth, and why gamma to the fourth fails both limits

Derivation BThe general law: gamma to the sixth, and why gamma to the fourth fails both limits

This card is the article's owned structural result, and it exists because the formula it derives is printed incorrectly in this article's AllQFT base — $\gamma^4$ where $\gamma^6$ belongs. That is not cosmetic: it makes the general law disagree with both of its own special cases, each by a factor of $\gamma$ , while the base's own limits table quotes those cases correctly. The base contradicts itself, and the contradiction is invisible unless somebody substitutes. So: derive from scratch, then substitute.

Step 1. Compute $A^\mu$ in components, with no restriction on $\mathbf{a}$ . From $U^\mu = \gamma(c,\mathbf{v})$ and $d/d\tau = \gamma\,d/dt$ : $A^{\mu} = \gamma\big(\dot\gamma\,c,\;\dot\gamma\,\mathbf{v} + \gamma\,\mathbf{a}\big)$ . Why: two derivatives, one chain rule. Everything hard hides in $\dot\gamma$ .

Step 2. Evaluate $\dot\gamma$ . From $\gamma = (1-v^2/c^2)^{-1/2}$ , $\dot\gamma = \gamma^{3}\dfrac{\mathbf{v}\cdot\mathbf{a}}{c^{2}}$ . Why: note it depends only on the component of $\mathbf{a}$ along $\mathbf{v}$ — a purely transverse acceleration does not change the speed, so it does not change $\gamma$ at all. This single fact is the origin of the entire collinear/transverse asymmetry.

Step 3. Contract. $A^{0} = \gamma^{4}\dfrac{\mathbf{v}\cdot\mathbf{a}}{c}$ and $\mathbf{A} = \gamma^{4}\dfrac{(\mathbf{v}\cdot\mathbf{a})}{c^{2}}\mathbf{v} + \gamma^{2}\mathbf{a}$ . Then, in $(+,-,-,-)$ ,

A\!\cdot\!A = (A^{0})^{2} - |\mathbf{A}|^{2} = \gamma^{8}\frac{(\mathbf{v}\!\cdot\!\mathbf{a})^{2}}{c^{2}}\Big(1 - \frac{v^{2}}{c^{2}}\Big) - 2\gamma^{6}\frac{(\mathbf{v}\!\cdot\!\mathbf{a})^{2}}{c^{2}} - \gamma^{4}a^{2} \;=\; -\gamma^{4}a^{2} - \gamma^{6}\frac{(\mathbf{v}\!\cdot\!\mathbf{a})^{2}}{c^{2}}.

Why: the first term's $(1-\beta^2)$ eats two powers of $\gamma$ , turning $\gamma^8$ into $\gamma^6$ , which then cancels against half the cross term. This is where a careless hand loses a factor of $\gamma^2$ — and losing exactly $\gamma^2$ is precisely how $\gamma^6$ becomes $\gamma^4$ .

Step 4. Read off $\alpha^2 = -A\!\cdot\!A$ and rewrite. $\alpha^{2} = \gamma^{4}a^{2} + \gamma^{6}\dfrac{(\mathbf{v}\cdot\mathbf{a})^{2}}{c^{2}}$ . Using $(\mathbf{v}\times\mathbf{a})^{2} = v^{2}a^{2} - (\mathbf{v}\cdot\mathbf{a})^{2}$ and $\gamma^{6}(1-\beta^{2}) = \gamma^{4}$ , this is identically $\alpha^{2} = \gamma^{6}\big[a^{2} - (\mathbf{v}\times\mathbf{a})^{2}/c^{2}\big]$ — equation (5). Why: both forms are useful — the first shows that only the longitudinal part of $\mathbf{a}$ gets the extra boost; the second is the compact one textbooks quote, and the one the base mangles.

Step 5. Substitute the limits — the check the base never performed. Collinear, $\mathbf{v}\times\mathbf{a} = 0$ : (5) gives $\alpha = \gamma^3 a$ ✓. Transverse, $\mathbf{v}\cdot\mathbf{a} = 0$ so $|\mathbf{v}\times\mathbf{a}| = va$ : (5) gives $\alpha^2 = \gamma^6 a^2(1 - \beta^2) = \gamma^4 a^2$ , so $\alpha = \gamma^2 a$ ✓. Now the base's $\gamma^4$ version: collinear $\alpha = \gamma^2 a$ ✗, transverse $\alpha = \gamma a$ ✗. Why: wrong by a factor of $\gamma$ in both limits — and the base's own limits table prints $\gamma^3 a$ and $\gamma^2 a$ . The base disagrees with itself. This is exactly the defect class that survives review: an equation nobody substitutes into. *Verified symbolically and numerically in verify_aq-sr-proper-acceleration-comoving-frame.py: for a generic $\mathbf{v}(t)$ , sympy confirms $A\!\cdot\!A = -\gamma^{6}[a^{2}-(\mathbf{v}\times\mathbf{a})^{2}/c^{2}]$ exactly, and confirms that the base's $\gamma^{4}$ form is not equal to it; both limits are substituted into both forms and the base's failures are asserted.*

Go deeper · Derivation C · 2 minThe clock hypothesis, and the muon that was accelerated at ten to the eighteen g

Derivation CThe clock hypothesis, and the muon that was accelerated at ten to the eighteen g

Everything above computes what an accelerometer reads. Nothing above tells you what a clock does. That these are independent questions is easy to miss and impossible to derive: the assertion that a clock's rate is fixed by its instantaneous speed alone, with no $\alpha$ -dependent correction, is an extra postulate. The goal here is to state it honestly, show why it cannot be proved, and show how it was tested — at an acceleration so large that any plausible correction should have screamed.

Step 1. State what relativity actually gives you. The Lorentz transformations relate two inertial frames. Applied to a momentarily comoving frame they tell you the relation between $d\tau$ and $dt$ at that instant: $d\tau = dt/\gamma(v)$ . Why: a statement about an instant — and an instant has a velocity but no history. Nothing forbids a real clock from also responding to $\alpha$ ; a mechanical one certainly would if you shook it hard enough.

Step 2. Recognise the gap. Integrating $\tau = \int dt/\gamma(v(t))$ along an accelerated worldline needs the extra assumption that the instantaneous relation still holds while the particle accelerates. That is the clock hypothesis: it defines "ideal clock", and it makes $\tau = \frac1c\int ds$ a property of the path, not of the dynamics. Why: without it the twin paradox has no definite answer, the proper time of a curved worldline is not geometric, and none of this chapter's machinery is well posed. It is load-bearing.

Step 3. Design the test. You want a clock that is genuinely a clock, moving fast enough for $\gamma$ to be large and precisely known, and accelerated enormously. A muon in a storage ring is all three: its decay is an ideal clock, $\gamma$ is set by the beam momentum, and circular motion at $\gamma \gg 1$ gives a transverse proper acceleration $\alpha = \gamma^2 v^2/R$ that is astronomical for a small ring. Why: circular motion is the trick — it cranks $\alpha$ to absurd values while holding $|\mathbf{v}|$ , and hence $\gamma$ , exactly constant. The two variables are cleanly separated, which is what a good experiment does.

Step 4. Do the arithmetic for the CERN ring. With $\gamma = 29.33$ , $R = 7.0\ \mathrm{m}$ : $\beta = 0.99942$ , so $a = v^2/R = 1.28\times10^{16}\ \mathrm{m/s^2}$ , and by the transverse rule (4), $\alpha = \gamma^{2}a = 860.2 \times 1.28\times10^{16} = \mathbf{1.1\times10^{19}\ m/s^{2}} = \mathbf{1.1\times10^{18}\,g}$ . Why: note which formula was used. Get the transverse rule wrong — or use the base's $\gamma^4$ -law, which gives $\gamma a$ — and this headline number moves by a factor of $29$ . The significance of the whole experiment rests on the exponent this article is at pains to fix.

Step 5. Read the verdict. Bailey et al. (Nature 268, 301–305, 1977) measured the dilated lifetime of both $\mu^{+}$ and $\mu^{-}$ at that $\gamma$ and found the Einstein factor to agree with experiment to a fractional error of $\mathbf{2\times10^{-3}}$ at 95 % confidence, with no detectable acceleration dependence. The measured proper lifetime, $\tau_0^{-} = 2.1948\ \mu\mathrm{s}$ , dilates to $\gamma\tau_0 = 64.4\ \mu\mathrm{s}$ in the lab. Why: the clock hypothesis survives at $10^{18}\,g$ to two parts in a thousand. It remains an assumption — an experiment bounds a correction, it never abolishes one — but it now has a very sharp fence around it, and that is the honest way to describe it. *Verified in verify_aq-sr-proper-acceleration-comoving-frame.py: $\beta$ , $a = v^2/R$ , $\alpha = \gamma^2 a$ and the conversion to $g$ are recomputed from $\gamma = 29.33$ , $R = 7.0$ m at 50-digit precision, and the dilated lifetime $\gamma\tau_0 = 64.4\ \mu\mathrm{s}$ is checked against the quoted $\tau_0 = 2.1948\ \mu\mathrm{s}$ .*

Limiting cases

Case	Result	Meaning
$v \ll c$	$\alpha \to a$	Newton: proper and coordinate acceleration coincide
$\mathbf{a}\parallel\mathbf{v}$	$\alpha = \gamma^{3}a$	the rocket case; read backwards, $d^2x/dt^2 = \alpha/\gamma^3 \to 0$
$\mathbf{a}\perp\mathbf{v}$	$\alpha = \gamma^{2}a_\perp$	the storage-ring case; one power of $\gamma$ less
general	$\alpha^{2} = \gamma^{6}\big[a^{2} - (\mathbf{v}\times\mathbf{a})^{2}/c^{2}\big]$	contains both above; the exponent is six

Common trapThe general law has gamma to the SIXTH — and this article's base says gamma to the fourth

The correct identity is

\alpha^{2} = \gamma^{6}\left[a^{2} - \frac{(\mathbf{v}\times\mathbf{a})^{2}}{c^{2}}\right].

The AllQFT base prints the same bracket with $\gamma^{4}$ in front. Substitute the limits and it collapses: collinear it yields $\alpha = \gamma^{2}a$ (should be $\gamma^{3}a$ ); transverse, $\alpha = \gamma\,a_\perp$ (should be $\gamma^{2}a_\perp$ ). Short by a factor of $\gamma$ in both — while the base's own limits table gives $\gamma^3 a$ and $\gamma^2 a_\perp$ . It is internally inconsistent. At the muon ring's $\gamma = 29.33$ that is a factor of $29$ in the headline proper acceleration. This is the archetype of the defect that survives review: an equation nobody ever substituted a special case into. Substitute. Always. It costs one line.

Common trapThe clock hypothesis is an assumption — and the acceleration is not what dilates the clock

Two errors travel together. The first is to think the clock hypothesis is a theorem: it is not derivable from the postulates, and it must be assumed before the proper time of an accelerated worldline is even well defined. The second, and commoner, is to think the muon's enormous acceleration is somehow responsible for its lifetime dilation. It is not. The dilation is $\gamma(v)$ , and it would be identical for a muon on a straight line at the same speed with zero acceleration. The ring's $10^{18}\,g$ is not the cause of the effect; it is the control variable, cranked as high as possible precisely to show it has none. Bailey et al. bounded any acceleration term at $2\times10^{-3}$ . That is a null result, and the null is the point.

Worked examples

1 · How hard is a muon in a storage ring pushed?

The CERN muon storage ring holds muons at $\gamma = 29.33$ on a circle of radius $R = 7.0\ \mathrm{m}$ . What proper acceleration do they feel, and what does that do to their clocks?

Step 1 — identify. Circular motion means $\mathbf{a}\perp\mathbf{v}$ : this is the transverse case, $\alpha = \gamma^2 a_\perp$ . Getting this exponent right is the whole exercise.

Step 2 — the lab acceleration. $\beta = \sqrt{1-1/\gamma^2} = \sqrt{1 - 1/860.2} = \mathbf{0.99942}$ , so $v = 0.99942c$ and

a = \frac{v^{2}}{R} = \frac{(0.99942 \times 2.998\times10^{8})^{2}}{7.0} = \mathbf{1.28\times10^{16}\ m/s^{2}}.

Step 3 — the proper acceleration. $\alpha = \gamma^{2}a = 860.2 \times 1.28\times10^{16} = \mathbf{1.1\times10^{19}\ m/s^{2}}$ , which in units of surface gravity is $1.1\times10^{19}/9.81 = \mathbf{1.1\times10^{18}\,g}$ .

Step 4 — the boundary number. A billion billion gravities — and the muon's clock does not notice. Its lifetime is dilated by $\gamma = 29.33$ and by nothing else: a proper lifetime of $\tau_0 = 2.1948\ \mu\mathrm{s}$ becomes $\gamma\tau_0 = \mathbf{64.4\ \mu s}$ in the lab, which is what Bailey et al. measured, to $2\times10^{-3}$ at 95 % confidence. The clock hypothesis is not a formality — it has been tested at the edge of what physics can arrange, and it held. Note what the base's $\gamma^4$ -law would have said here: $\alpha = \gamma a = 3.8\times10^{17}\ \mathrm{m/s^2}$ , low by a factor of $29$ .

2 · The same push, felt two ways

A particle at $\beta = 0.99$ is subjected to a lab-frame acceleration of magnitude $a = 1.00\times10^{15}\ \mathrm{m/s^2}$ . How hard does it feel if the push is along its motion? And across it?

Step 1 — identify. Same $|\mathbf{a}|$ , same $\gamma$ , two orientations. Use equation (5), or equivalently the two limits (4).

Step 2 — the Lorentz factor. $\gamma = 1/\sqrt{1-0.99^{2}} = \mathbf{7.089}$ , so $\gamma^{2} = 50.25$ and $\gamma^{3} = 356.2$ .

Step 3 — apply. Collinear: $\alpha_\parallel = \gamma^{3}a = 356.2 \times 10^{15} = \mathbf{3.56\times10^{17}\ m/s^{2}}$ . Transverse: $\alpha_\perp = \gamma^{2}a = 50.25 \times 10^{15} = \mathbf{5.03\times10^{16}\ m/s^{2}}$ .

Step 4 — the boundary number. The ratio is exactly $\gamma = \mathbf{7.089}$ : the identical lab-frame acceleration is felt seven times harder along the motion than across it. That is why the direction of $\mathbf{a}$ relative to $\mathbf{v}$ must appear in the general law — as it does, through $(\mathbf{v}\times\mathbf{a})^2$ . It is also the sharpest check on equation (5): substitute the two orientations and you must recover $\gamma^3$ and $\gamma^2$ . The base's $\gamma^4$ form recovers $\gamma^2$ and $\gamma^1$ , and is caught in a single line of arithmetic.

Physics of the portrait

Proper acceleration is the one number about a worldline's bending that everybody agrees on — and the one an instrument actually reports.

Origins · 2 minProper acceleration and how this was discovered

Proper acceleration

1907–1977 · Einstein's uniformly accelerated frame · Minkowski's proper time makes the four-vector possible · Born's hyperbolic worldline · Bailey's muons test the clock hypothesis at 10^18 g

The idea arrives with the geometry. Minkowski's 1908 "Raum und Zeit" introduced Eigenzeit — proper time — and with it the possibility of differentiating a worldline with respect to an invariant, which is what makes $U^\mu$ and $A^\mu$ four-vectors at all. Before that there was no clean way to say what an accelerating particle's acceleration was: every frame gave a different answer and none was privileged. Einstein had already, in his 1907 Jahrbuch review, used a uniformly accelerated reference frame as a device for thinking about gravity, and Born in 1909 wrote down the worldline of constant proper acceleration and named the motion hyperbolic. What none of them could do was prove that a clock on such a worldline ticks at the rate its instantaneous speed dictates and no other. Einstein simply assumed it — the Uhrenhypothese — and said so. It stayed an assumption for seventy years, respectable but untested at any interesting acceleration, until the CERN Muon Storage Ring made it answerable. In 1977 J. Bailey and a large collaboration published in Nature (268, 301–305) the lifetimes of positive and negative muons circulating at $\gamma = 29.33$ , under a transverse proper acceleration of order $10^{18}\,g$ — more than anything in any laboratory before or since. The Einstein factor agreed to $2\times10^{-3}$ at 95 % confidence, with no acceleration-dependent term; as a by-product they delivered the most accurate proper lifetime of the negative muon then known, $2.1948\ \mu\mathrm{s}$ . An assumption, tested where it could most plausibly have failed, and vindicated.

History

1907

— Einstein, in the Jahrbuch der Radioaktivitat review, analyses a uniformly accelerated reference frame and adopts the clock hypothesis as an explicit assumption.

1908

— Minkowski's "Raum und Zeit" introduces proper time (Eigenzeit), making the four-velocity and four-acceleration well-defined four-vectors.

1909

— Born writes down the worldline of constant proper acceleration and names it hyperbolic motion (Ann. Phys. 335, 1-56).

1963

— Frisch and Smith measure muon time dilation on Mount Washington, confirming the velocity factor for muons in free flight.

1977

— Bailey et al. measure relativistic time dilation for muons in the CERN storage ring at a transverse proper acceleration of about 10^18 g, confirming the Einstein factor to 2 parts in 1000 at 95 percent confidence with no acceleration dependence (Nature 268, 301-305).

Going further

Texts. Rindler, Introduction to Special Relativity, 2nd ed. · Misner, Thorne & Wheeler, Gravitation, §6 (accelerated observers, the clock hypothesis) · Bailey, J. et al., Nature 268, 301–305 (1977) · Born, M., Ann. Phys. 335, 1–56 (1909).

Provenance. Signature $(+,-,-,-)$ ; $U\!\cdot\!U = +c^2$ , $A\!\cdot\!A = -\alpha^2$ (negative, spacelike). Every equation and number above is recomputed in verify_aq-sr-proper-acceleration-comoving-frame.py at 50-digit precision, which also greps this article for every printed value. Base defect found and repaired. The AllQFT base's eq[5] prints $a^{\mu}a_{\mu} = -\gamma^{4}\big[\mathbf{a}^{2} - (\mathbf{v}\times\mathbf{a})^{2}/c^{2}\big]$ . The exponent must be 6, not 4. The base's form yields $\alpha = \gamma^{2}a$ for collinear motion and $\alpha = \gamma a_\perp$ for transverse — short by a factor of $\gamma$ in both, and in direct contradiction with the base's own limits table, which correctly lists $\gamma^{3}a$ and $\gamma^{2}a_\perp$ . The verify script derives the correct identity symbolically from $A^\mu = \gamma\,d[\gamma(c,\mathbf{v})]/dt$ and asserts that the $\gamma^4$ form is not equal to it. Note for the programme: the LANE_SPEC's GATE-0 table cleared this equation as "right but ugly"; that assessment is refused here, on the sympy derivation above.

Scope. The four-velocity and four-acceleration as objects belong to `aq-sr-four-velocity` and `aq-sr-four-acceleration` — recalled, not re-derived. The hyperbolic worldline that follows from constant $\alpha$ is derived in `aq-sr-relativistic-uniform-acceleration` and only named here. The Unruh temperature, which the base attaches to this article, is a signpost; it belongs to `aq-sr-unruh-effect`.

Builds on. Four-Velocity · Proper Time · The Minkowski Metric. Leads to. Relativistic Uniform Acceleration and the Rindler Hyperbola · Rindler Coordinates · The Equivalence Principle · The Relativistic Rocket Equation and Traveler Time.

Why does the collinear case get three powers of gamma and the transverse case only two?

Because $\gamma$ responds only to changes in speed, and a transverse acceleration does not change the speed: $\dot\gamma = \gamma^3(\mathbf{v}\cdot\mathbf{a})/c^2$ vanishes identically when $\mathbf{a}\perp\mathbf{v}$ . So the whole $\dot\gamma$ machinery switches off, leaving only the two factors from converting a lab time interval and a lab transverse velocity into the comoving frame — giving $\gamma^2$ . Collinear, $\dot\gamma$ is nonzero and contributes the extra power. The asymmetry is not algebraic bookkeeping; it is the statement that speeding a particle up is harder than turning it.

Is proper acceleration the same as the force divided by the mass?

Only in the rest frame — which is precisely where it is defined. The four-force is $F^\mu = mA^\mu$ with $m$ the invariant mass, so $\alpha = |F_{\text{rest}}|/m$ : rest-frame force per unit rest mass, exactly as Newton would have it. What is not true is that the lab three-force equals $m\,d\mathbf{v}/dt$ with constant $m$ — the lab relation carries those same $\gamma^3$ and $\gamma^2$ factors, which is the origin of the old "longitudinal mass" $\gamma^3m$ and "transverse mass" $\gamma^2m$ . That vocabulary tried to save $F = ma$ by making mass direction-dependent. Better to keep the mass invariant and put the $\gamma$ 's where they belong: in the kinematics.

Could a real clock be affected by acceleration, in violation of the clock hypothesis?

A real mechanical clock certainly can — shake a wristwatch hard enough and it stops. The clock hypothesis is a statement about ideal clocks, and it amounts to defining an ideal clock as one whose rate is unaffected by acceleration. The physical content is the empirical claim that such clocks exist, or that the systems we use approximate them to a stated precision. A muon's decay is an excellent candidate: it is governed by internal physics on a timescale far shorter than any over which the external acceleration varies, so corrections should be of order $\alpha\ell/c^2$ with $\ell$ the internal size — utterly negligible. Bailey et al. bounded them at $2\times10^{-3}$ , and there they remain.

If the muon feels 10^18 g, why does its lifetime depend only on its speed?

Because time dilation is a statement about the geometry of the worldline, not about the forces bending it. The factor $\gamma$ is fixed by the worldline's instantaneous tangent, and the proper time is the invariant arc length $\tau = \frac1c\int ds$ , which knows nothing about the path's curvature. The acceleration is that curvature, and the clock hypothesis is exactly the assertion that the arc length is what the clock measures. Be blunt about the direction of the inference: the muon experiment does not explain why acceleration is irrelevant. It tests whether it is, at the largest acceleration anyone has arranged, and finds nothing at the two-parts-in-a-thousand level.

Where does the base's gamma-to-the-fourth error actually come from?

Almost certainly from mishandling the $(1-\beta^2)$ that appears when the time component is contracted. In Step 3 of the general derivation, $(A^0)^2 = \gamma^{8}(\mathbf{v}\cdot\mathbf{a})^2/c^2$ while the spatial part contributes $\gamma^{8}(\mathbf{v}\cdot\mathbf{a})^2v^2/c^4$ ; their difference carries $\gamma^{8}(1 - \beta^2) = \gamma^{6}$ . Drop that step and you land on $\gamma^4$ . The error is easy to make and completely invisible — until you substitute a special case, at which point it fails immediately in both limits. Substituting limits into every general formula is not pedantry; it is the cheapest error-detection there is, and it is what caught this one.

Does the Unruh temperature belong here, since it is proportional to alpha?

It is proportional to the proper acceleration — $k_BT_U = \hbar\alpha/2\pi c$ — and this article's base does list it. But it is not a result of special relativity at all: it is quantum field theory in a Rindler wedge, requiring a mode decomposition of the vacuum with respect to the accelerated observer's time coordinate. Quoting it here would be quoting a conclusion whose derivation lives elsewhere, so this article names it and stops; it belongs to `aq-sr-unruh-effect`. What does belong here is that $\alpha$ — the invariant this article defines — is the quantity the temperature depends on, which is one more reason the exponent in (5) has to be right.