Relativistic Kinetic Energy

Take an LHC proton. Its kinetic energy is $6.799$ TeV — a number the machine's operators know to many decimal places, because the whole 27-kilometre ring exists to put it there. Now ask Newton how fast it is going. Set $\tfrac12 mv^2 = 6.799$ TeV, solve for $v$ , and out comes

v = c\,\sqrt{\frac{2K}{mc^2}} = 120.4\,c.

A hundred and twenty times the speed of light. This is not a small error, a percent-level correction, or a regime where Newton is "approximately right". It is an answer that is not even the right kind of thing. And the machine, meanwhile, is quietly running the proton at $0.99999999c$ and will never get it any faster, no matter how many more cavities are installed. The relativistic kinetic energy $K = (\gamma - 1)mc^2$ fixes this, and it does so in a way that is more generous to Newton than it first appears. Expand it in powers of $v/c$ and the very first term is $\tfrac12 mv^2$ — exactly, with no fudge factor. Newton was not wrong. Newton had the first term of a series and no way of knowing there were more. The rest of this article is that series, where it comes from, and what happens at the end of it.

Key takeaways

Signature $(+,-,-,-)$ : $\eta_{\mu\nu} = \mathrm{diag}(+1,-1,-1,-1)$ , so $p\cdot p = +m^2c^2$ . $m$ is the invariant rest mass — there is no "relativistic mass".
Owned result: $K = (\gamma - 1)mc^2$ , from the work–energy theorem. Integrate $K = \int \vec F\cdot d\vec x$ with $\vec F = d\vec p/dt$ and $\vec p = \gamma m\vec v$ . The integral is done in full below — it is not quoted.
Owned result: Newton is the FIRST TERM of a series, not a different theory. K = \tfrac12 mv^2 + \tfrac38\frac{mv^4}{c^2} + \tfrac{5}{16}\frac{mv^6}{c^4} + \tfrac{35}{128}\frac{mv^8}{c^6} + \cdots (coefficients sympy-confirmed). The leading term is Newtonian kinetic energy exactly, with no correction factor.
Owned result: $K$ diverges as $v \to c$ , so $c$ is unattainable for $m > 0$ . For an electron: $\beta = 0.99 \Rightarrow K = 3.11$ MeV; $\beta = 0.999 \Rightarrow 10.92$ MeV; $\beta = 0.9999 \Rightarrow 35.62$ MeV. Each extra "9" costs a factor approaching $\sqrt{10} = 3.162$ and buys ever less speed — which is why accelerator physicists quote energy, never speed.
In terms of momentum the series reads $K = \dfrac{p^2}{2m} - \dfrac{p^4}{8m^3c^2} + \dfrac{p^6}{16m^5c^4} - \cdots$ , and that second term is the fine-structure correction of atomic physics — the same operator, arriving from relativity.
Key idea: $\tfrac12 mv^2$ is not an approximation to the kinetic energy in the sense of being a bit off. It is the leading term of a convergent expansion whose radius of convergence is exactly $\beta = 1$ — so the series that contains Newtonian mechanics also contains, in its own divergence, the reason the speed of light cannot be reached.

The core insight

The work–energy theorem survives relativity untouched: the work you do on a body still equals the increase in its kinetic energy. What changes is what "kinetic energy" evaluates to, because $\vec p$ is now $\gamma m \vec v$ rather than $m\vec v$ . Every relativistic effect in this article follows from that single substitution inside a perfectly ordinary integral. And the divergence at $v = c$ is therefore not a mathematical pathology to be regularised away — it is a physical statement about work: pushing a massive body to the speed of light would require you to do an infinite amount of it.

Where this lives today

The $\gamma^3$ of longitudinal inertia and the divergence of $K$ together determine the architecture of every accelerator on Earth. A linear accelerator adds energy at a roughly constant rate per metre, so the energy grows linearly with the length of the machine while the speed saturates almost immediately. SLAC's three-kilometre linac reached about $50$ GeV — an average of some $17$ MeV per metre — and an electron in it is already past $\beta = 0.9999$ within the first few metres, spending the remaining three kilometres gaining $\gamma$ , not $v$ . The machine is a $\gamma$ -multiplier, and that is why the useful specification is " $50$ GeV", a number that keeps growing, rather than a speed, which stopped being informative almost immediately. The same series has a second life in atomic physics, and it is not an analogy — it is the identical expansion. Take equation (3), keep the first correction, and promote $p$ to an operator: $\hat H'_{\text{rel}} = -\hat p^4/8m^3c^2$ . This is the relativistic kinetic correction to the hydrogen Hamiltonian, and Relativistic Kinetic Energy Correction derives precisely this series and identifies that term as the simplest of the three corrections that together build hydrogen's fine structure. The electron in a hydrogen atom moves at about $\alpha \approx 1/137$ of the speed of light, so $\beta^2 \sim 5\times10^{-5}$ — small, but spectroscopy is far more precise than that, and the correction is plainly visible in the spectrum. The line splittings you see in a good spectrograph are, in part, the second term of the series in this article. And the divergence itself is visible in the sky. Cosmic-ray protons have been observed with kinetic energies above $10^{20}$ eV — that is $16$ joules, the energy of a briskly struck tennis ball, carried by a single subatomic particle — which puts them at $\gamma \approx 1.1\times10^{11}$ . Such a proton's speed falls short of $c$ by about $4$ parts in $10^{23}$ , since $1 - \beta \approx 1/2\gamma^2$ . It is the fastest thing ever measured, and it is still not fast enough. Nothing in the universe has ever been observed to close that gap.

The physical picture

Newtonian kinetic energy comes from an integral, and it is worth remembering which one. Push a body from rest and the work you do is $\int \vec F\cdot d\vec x$ . With $\vec F = m\vec a$ this evaluates to $\tfrac12 mv^2$ , and we call the result the kinetic energy because it is what the work became. Relativity does not touch the definition. Work is still $\int \vec F \cdot d\vec x$ , and it still equals the gain in kinetic energy. What relativity changes is one thing only: the second law reads $\vec F = d\vec p/dt$ with $\vec p = \gamma m \vec v$ — a substitution forced, not chosen, as Relativistic Momentum proves. Put the new momentum into the old integral and turn the handle. The answer that comes out, $K = (\gamma - 1)mc^2$ , is worth reading twice. It is not $\tfrac12 \gamma m v^2$ , which is what you get by naively decorating Newton's formula with a $\gamma$ , and which is wrong. And it is not $\tfrac12 m v^2$ with a small correction bolted on. It is a new function, whose expansion happens to begin with $\tfrac12 mv^2$ — and whose behaviour near $v = c$ has no Newtonian analogue at all.

Figure

PUNCHLINE — the two theories agree beautifully, and then one of them leaves. Below β ≈ 0.3 the curves are indistinguishable. But Newton's ½mv² is BOUNDED by ½mc², while K = (γ−1)mc² has a vertical asymptote at β = 1. Marked: an electron at β = 0.99 has K = 3.11 MeV, at 0.999 it is 10.92 MeV, at 0.9999 it is 35.62 MeV. Each extra 9 costs a factor approaching √10 = 3.162 and buys ever less speed. Run Newton backwards on an LHC proton — set ½mv² = 6.799 TeV — and he reports a speed of 120.4c

The figure carries the whole argument. Newton's parabola is bounded: since $v < c$ always, $\tfrac12 mv^2 < \tfrac12 mc^2$ , so Newtonian mechanics assigns a finite maximum kinetic energy to a particle travelling at the speed of light. Relativity's curve has no maximum at all. That difference — bounded versus unbounded — is the light-speed limit, and it is not an extra postulate bolted on. It is what the work integral says.

The core equations

The kinetic energy, defined as always and evaluated relativistically:

\boxed{K \;=\; E - E_0 \;=\; \gamma mc^2 - mc^2 \;=\; (\gamma - 1)\,mc^2}, \qquad \gamma = \frac{1}{\sqrt{1 - v^2/c^2}} \qquad (1)

where: $m$ — the invariant rest mass; $E = \gamma mc^2$ — the total energy (Relativistic Energy); $E_0 = mc^2$ — the rest energy. Physical reading: kinetic energy is everything a body's energy has beyond what it would have standing still. At $v = 0$ , $\gamma = 1$ and $K = 0$ , as it must. Note what this is not: it is not $\tfrac12\gamma mv^2$ , an expression with no derivation behind it that nonetheless appears distressingly often. Expanded in the speed, the series that contains Newton:

K \;=\; \tfrac12 mv^2 \;+\; \tfrac38\,\frac{mv^4}{c^2} \;+\; \tfrac{5}{16}\,\frac{mv^6}{c^4} \;+\; \tfrac{35}{128}\,\frac{mv^8}{c^6} \;+\; \cdots \qquad (2)

where: the coefficients are those of $(1-\beta^2)^{-1/2}$ . Physical reading: the first term is Newtonian kinetic energy, exactly. Not "approximately", not "up to a factor" — the binomial series hands back $\tfrac12 mv^2$ with coefficient precisely $\tfrac12$ . The series converges for $\beta < 1$ and diverges at $\beta = 1$ , which is the light-speed limit showing up as a radius of convergence. Expanded in the momentum instead — a different-looking series, and the one atomic physics needs:

K \;=\; \sqrt{(pc)^2 + (mc^2)^2} \;-\; mc^2 \;=\; \frac{p^2}{2m} \;-\; \frac{p^4}{8m^3c^2} \;+\; \frac{p^6}{16m^5c^4} \;-\; \cdots \qquad (3)

where: $p = \gamma mv$ . Physical reading: same physics, different variable — and note the signs alternate here while they were all positive in (2). The two series are perfectly consistent (verified below); they look different because $p$ and $v$ are not linearly related. The term $-p^4/8m^3c^2$ is the leading relativistic correction used in atomic fine structure, and Relativistic Kinetic Energy Correction derives exactly this expansion and promotes that term to the operator $\hat H'_{\text{rel}} = -\hat p^4/8m^3c^2$ .

Doing the integral

Derivation AThe work-energy theorem, done in full, gives K equals gamma minus one times mc squared

The theorem itself is not in question and does not change: the work done by the net force on a body equals the increase in its kinetic energy, and that is true in relativity exactly as it is in Newtonian mechanics. What changes is the momentum inside the force law. So the honest way to obtain $K$ is not to quote it but to run the same integral Newton ran, with $\vec p = \gamma m\vec v$ substituted for $m\vec v$ , and see what falls out. It is a short calculation and everything interesting about the article is inside it — including, at the last step, the reason the answer is not $\tfrac12\gamma mv^2$ .

Step 1. Set up the work integral. For one-dimensional motion from rest to speed $v$ , K = \int \vec F\cdot d\vec x = \int \frac{dp}{dt}\,v\,dt = \int_0^{v} v\; d(\gamma m v). Why: the substitution $d\vec x = \vec v\,dt$ turns a spatial integral into an integral over the momentum, and the $dt$ cancels. Nothing relativistic has happened yet — this is Newton's own manipulation.

Step 2. Integrate by parts. K = \Big[\gamma m v^2\Big]_0^{v} - \int_0^{v} \gamma m v\, dv = \gamma m v^2 - m\!\int_0^{v}\frac{v\,dv}{\sqrt{1 - v^2/c^2}}. Why: by parts is the move that converts $\int v\,dp$ into something we can integrate directly, because $\gamma$ is an explicit function of $v$ and $v\,dv$ is exactly the combination that appears when you differentiate the square root.

Step 3. Do the remaining integral. Substituting $s = 1 - v^2/c^2$ , $ds = -2v\,dv/c^2$ : m\int_0^{v}\frac{v\,dv}{\sqrt{1-v^2/c^2}} = mc^2\Big[1 - \sqrt{1 - v^2/c^2}\,\Big] = mc^2\Big(1 - \frac{1}{\gamma}\Big). Why: elementary, and it is where the $c^2$ enters the answer — from the Jacobian of the substitution, not from anywhere exotic. The $mc^2$ in the final formula is manufactured right here.

Step 4. Assemble. K = \gamma m v^2 - mc^2 + \frac{mc^2}{\gamma}. Why: this is the answer, but not yet in recognisable form. The next step is pure algebra and it is the one students most often fail to spot.

Step 5. Collapse it with the $\gamma$ identity. Note that $\gamma m v^2 = \gamma m c^2\beta^2 = \gamma mc^2\big(1 - 1/\gamma^2\big) = \gamma mc^2 - mc^2/\gamma$ , using $\beta^2 = 1 - 1/\gamma^2$ . Substituting into Step 4, the two $mc^2/\gamma$ terms cancel: K = \Big(\gamma mc^2 - \frac{mc^2}{\gamma}\Big) - mc^2 + \frac{mc^2}{\gamma} = \boxed{(\gamma - 1)\,mc^2}. Why: this cancellation is why the answer is not $\tfrac12\gamma mv^2$ . The by-parts boundary term $\gamma mv^2$ is not the kinetic energy; it is a piece of it, and the integral supplies a second piece that partly cancels it. Anyone who "generalises" $\tfrac12 mv^2$ by inserting a $\gamma$ has skipped the integral and got a different — wrong — function.

Step 6. Check the limits. As $v \to 0$ , $\gamma \to 1$ and $K \to 0$ . As $v \to c$ , $\gamma \to \infty$ and $K \to \infty$ . And $K = E - mc^2$ reproduces $E = \gamma mc^2$ , consistent with Relativistic Energy. Why: a cleaner way to see the whole result — from $E^2 = (pc)^2 + (mc^2)^2$ one gets $dE/dp = pc^2/E = v$ , so $dK = dE = v\,dp$ , which is Step 1. The work integral and the energy identification are the same statement, approached from opposite ends. *Verified symbolically in verify_aq-sr-relativistic-kinetic-energy.py: sympy evaluates $\int_0^{v} v\,d(\gamma m v)$ in closed form and the difference from $(\gamma-1)mc^2$ is identically zero; the identity $\gamma m v^2 = \gamma mc^2 - mc^2/\gamma$ is confirmed; and $dE/dp = v$ is derived from the energy–momentum relation.*

Go deeper · Derivation B · 3 minNewton is the first term of a series, and the series diverges exactly at the speed of light

Derivation BNewton is the first term of a series, and the series diverges exactly at the speed of light

Having the closed form is not the same as understanding it, and the expansion is where the understanding is. Two things need establishing, and they pull in opposite directions. The first is continuity with Newton: the relativistic kinetic energy must reduce to $\tfrac12 mv^2$ at low speed, and it must do so exactly, with no leftover constant and no correction factor — otherwise relativity would contradict two centuries of successful mechanics rather than containing it. The second is the break with Newton: the same series must fail, and fail catastrophically, as $v \to c$ . Remarkably, one expansion delivers both.

Step 1. Expand the Lorentz factor binomially. With $\beta = v/c$ and $|\beta| < 1$ , \gamma = (1 - \beta^2)^{-1/2} = 1 + \tfrac12\beta^2 + \tfrac38\beta^4 + \tfrac{5}{16}\beta^6 + \tfrac{35}{128}\beta^8 + \cdots Why: this is the standard binomial series $(1-x)^{-1/2} = \sum_n \binom{2n}{n}(x/4)^n$ with $x = \beta^2$ . The coefficients $\tfrac12, \tfrac38, \tfrac{5}{16}, \tfrac{35}{128}$ are not decoration and are machine-checked in the verify script — a mistyped coefficient here is exactly the sort of defect that survives a hundred readings.

Step 2. Subtract the 1 and multiply by $mc^2$ . K = (\gamma-1)mc^2 = mc^2\Big(\tfrac12\beta^2 + \tfrac38\beta^4 + \cdots\Big) = \tfrac12 mv^2 + \tfrac38\frac{mv^4}{c^2} + \tfrac{5}{16}\frac{mv^6}{c^4} + \cdots Why: the " $-1$ " that we subtract is precisely the rest energy. The famous $mc^2$ is the zeroth term of this series, and $K$ is what is left when you remove it — which is exactly why $K$ , and not $E$ , is the thing that reduces to the Newtonian energy.

Step 3. Read the first term. It is $\tfrac12 mv^2$ , with coefficient exactly $\tfrac12$ . Newtonian kinetic energy is not an approximation to this formula; it is a term in it. Why: this is the article's thesis and it deserves the emphasis. Relativity does not correct Newton in the way that a better fit corrects a worse one. It exhibits Newton as the leading behaviour of something larger — which is why Newtonian mechanics works so well, and why its failures, when they come, are so total.

Step 4. Size the first correction. The ratio of the second term to the first is $\dfrac{\tfrac38 mv^4/c^2}{\tfrac12 mv^2} = \tfrac34\beta^2$ . At $\beta = 0.1$ that is $0.0075$ — three parts in four hundred. At $\beta = 10^{-6}$ , the speed of a fast aircraft, it is $7.5\times10^{-13}$ . Why: this is why nobody found relativity by doing mechanics carefully. The first correction is quadratically small in $\beta$ , so at any speed available before the twentieth century it sits far below experimental error.

Step 5. Now break it. The series converges for $|\beta| < 1$ and diverges at $\beta = 1$ — its radius of convergence is exactly the speed of light. Equivalently, from the closed form, $K = (\gamma-1)mc^2 \to \infty$ as $v \to c$ . Why: the light-speed limit is not an extra postulate imposed on the dynamics. It is a property of the function — and it is the same singularity that terminates the series. The mathematics knew about $c$ before we told it.

Step 6. Say what the divergence means physically. $K = \int \vec F\cdot d\vec x$ , so an infinite $K$ means an infinite amount of work. To accelerate a body of mass $m$ to exactly $c$ you would have to do infinite work on it; every real accelerator does a finite amount, and so every real particle stays below $c$ . Why: the barrier is not a wall in space. It is an unpayable bill. And this is why accelerator physicists quote energy, never speed — as the next section makes uncomfortably concrete. *Verified symbolically in verify_aq-sr-relativistic-kinetic-energy.py: sympy extracts the coefficients of $\beta^2, \beta^4, \beta^6, \beta^8$ from $(\gamma-1)mc^2$ and confirms them as $\tfrac12, \tfrac38, \tfrac{5}{16}, \tfrac{35}{128}$ ; the momentum series $p^2/2m - p^4/8m^3c^2 + p^6/16m^5c^4$ is extracted independently from $\sqrt{(pc)^2 + (mc^2)^2} - mc^2$ ; and the two series are shown to be consistent — substituting $p = \gamma mv$ into the momentum series and re-expanding in $v$ reproduces the velocity series exactly, difference identically zero.*

Figure

The price of a nine. This is the ladder every accelerator physicist lives on. Going from β = 0.99 to 0.999 to 0.9999 buys an electron almost NO extra speed — the three points are visually on top of each other — while its kinetic energy climbs 3.11 → 10.92 → 35.62 MeV. Each extra nine costs a factor approaching √10 = 3.162, because γ ≈ 1/√(2(1−β)) and each nine divides (1−β) by ten. Speed is a useless coordinate up here. Energy is the only one that moves

Limiting cases

Case	Result	Meaning
$v \to 0$	$K \to 0$	as it must — and $\gamma \to 1$
$v \ll c$	$K \to \tfrac12 mv^2$	Newton, exactly, as the leading term
first correction	$\tfrac38 mv^4/c^2$ , i.e. $+\tfrac34\beta^2$ fractionally	$0.75\%$ at $\beta = 0.1$ ; $7.5\times10^{-13}$ for an aircraft
$c \to \infty$	$K \to \tfrac12 mv^2$ exactly, all corrections vanish	the Galilean world
$v \to c$ , $m > 0$	$K \to \infty$	$c$ is unattainable: it would cost infinite work
$K \gg mc^2$	$K \approx pc$	ultrarelativistic: energy and momentum merge
in terms of $p$ , low speed	$K \approx \dfrac{p^2}{2m} - \dfrac{p^4}{8m^3c^2}$	the fine-structure correction of atomic physics
$m = 0$	$K$ is not defined separately; $E = pc$	a photon has no rest energy to subtract from

Common trapThe kinetic energy is not one-half gamma m v squared

This is the single commonest error in the topic, and it comes from "upgrading" $\tfrac12 mv^2$ by sprinkling a $\gamma$ on it — the same instinct that produces "relativistic mass", and wrong for the same reason. The correct result, $K = (\gamma-1)mc^2$ , comes out of the work integral, and the derivation above shows exactly where $\tfrac12\gamma mv^2$ goes astray: $\gamma mv^2$ is the boundary term in the integration by parts, but the integral contributes a second piece that partly cancels it. Check the two against each other at $\beta = 0.99$ for an electron: the truth is $3.11$ MeV, while $\tfrac12\gamma mv^2$ gives $1.77$ MeV — wrong by nearly a factor of two, and getting worse. There is no speed at which the naive formula is right except $v = 0$ , where both are zero.

Worked examples

1 · Climbing the ladder: what a nine costs

Compute the kinetic energy of an electron ( $m_ec^2 = 0.510999$ MeV) at $\beta = 0.99$ , $0.999$ and $0.9999$ . Then explain the pattern.

Step 1 — identify. $K = (\gamma - 1)mc^2$ , and $\gamma = (1-\beta^2)^{-1/2}$ . Nothing else is needed.

Step 2 — apply. $\gamma(0.99) = 7.0888$ , so $K = 6.0888 \times 0.510999 = \mathbf{3.11\ \text{MeV}}$ . $\gamma(0.999) = 22.366$ , so $K = 21.366 \times 0.510999 = \mathbf{10.92\ \text{MeV}}$ . $\gamma(0.9999) = 70.712$ , so $K = 69.712 \times 0.510999 = \mathbf{35.62\ \text{MeV}}$ .

Step 3 — the pattern. The ratios are $10.92/3.11 = \mathbf{3.51}$ and $35.62/10.92 = \mathbf{3.26}$ . They are drifting downward toward a limit, and the limit is worth naming. Near $\beta = 1$ , write $\varepsilon = 1 - \beta$ ; then $\gamma = 1/\sqrt{(1-\beta)(1+\beta)} \approx 1/\sqrt{2\varepsilon}$ . Each extra "9" divides $\varepsilon$ by ten, so it multiplies $\gamma$ — and, once $\gamma \gg 1$ , the kinetic energy — by $\sqrt{10} = \mathbf{3.162}$ . Check it: $1/\sqrt{2\varepsilon}$ for $\varepsilon = 0.01, 0.001, 0.0001$ gives $7.071$ , $22.361$ , $70.711$ , against the exact $7.089$ , $22.366$ , $70.712$ . The approximation is already good to a quarter of a percent at the first rung.

Step 4 — the boundary number. To go from $\beta = 0.99$ to $\beta = 0.9999$ — an increase in speed of about one percent — costs the electron a factor of 11.4 in energy. This is the fact that shapes every accelerator ever built. Speed is a useless coordinate in this regime: it is pinned against $1$ and moves only in the far decimal places, while energy climbs without limit. Hence the universal convention: a beam is a " $6.8$ TeV beam", never a " $0.99999999c$ beam".

2 · The naive formula, and how badly it fails

An electron has $\beta = 0.99$ . Compare the true $K$ with the two common wrong answers, $\tfrac12 mv^2$ and $\tfrac12\gamma mv^2$ .

Step 1 — identify. All three are computable in one line each; the point is the comparison.

Step 2 — apply. True: $K = (\gamma - 1)m_ec^2 = 6.0888 \times 0.510999 = \mathbf{3.11}$ MeV. Newton: $\tfrac12 m v^2 = \tfrac12 \beta^2 m_ec^2 = \tfrac12(0.9801)(0.510999) = \mathbf{0.250}$ MeV. The " $\gamma$ -decorated" version: $\tfrac12\gamma m v^2 = 7.0888 \times 0.250 = \mathbf{1.77}$ MeV.

Step 3 — compare. Newton is low by a factor of $12.4$ . The $\gamma$ -decorated formula is low by a factor of $1.75$ — better, and still wrong, which is the dangerous kind of wrong. It is close enough to look plausible and far enough to ruin an answer.

Step 4 — the boundary number. The moral is in the ratio $3.11/1.77 = \mathbf{1.75}$ : no amount of decorating Newton's formula with Lorentz factors will produce the right function, because the right function came out of an integral that Newton's formula never performed. You cannot patch $\tfrac12 mv^2$ . You have to redo the work integral, and when you do, you get $(\gamma - 1)mc^2$ .

Physics of the portrait

The portrait is a quantity that has been rewritten more often than almost any other in physics — and whose final rewriting, in 1905, was the gentlest of all, because it did not so much replace the old formula as reveal where it had come from.

Origins · 2 minKinetic energy and how this was discovered

Kinetic energy

1676–1905 · Leibniz's vis viva · Coriolis supplies the half · Thomson gives it its name · Einstein makes it the second term of a series

The quantity began as a weapon in a quarrel. Against the Cartesians, who held that the conserved measure of motion was $mv$ , Gottfried Wilhelm Leibniz argued from about 1676 that it was $mv^2$ — the vis viva, or living force — and the dispute ran, bitterly and largely at cross purposes, for the better part of a century before it became clear that both quantities are conserved and they are simply different things. The modern form arrived with Gaspard-Gustave de Coriolis in 1829, who supplied the factor of one-half that makes the quantity equal to the work done, and thereby tied it to force and displacement rather than to metaphysics. The name came from William Thomson, later Lord Kelvin, around 1849–1851; Rankine's complementary "potential energy" followed in 1853, and Rankine himself credits Thomson and Tait with substituting "kinetic" for his own "actual". So by the 1860s the quantity had its formula, its factor, its name and two centuries of unbroken success. Einstein did not overturn it. He showed that $\tfrac12 mv^2$ is the leading term of $(\gamma - 1)mc^2$ — that the formula three generations had fought over was the first line of a longer expression, correct as far as it went and silent about everything beyond. It is the most courteous revolution in the history of mechanics.

History

1676

— Leibniz argues that the conserved measure of motion is mv², the vis viva, against the Cartesian mv; the dispute runs for decades.

1829

— Coriolis introduces the factor of one-half, defining kinetic energy as ½mv² so that it equals the work done, and connects it to force and displacement.

1849

— William Thomson (Lord Kelvin) coins the term "kinetic energy"; Rankine adds "potential energy" in 1853 and credits Thomson and Tait with the word "kinetic".

1905

— Einstein's "Zur Elektrodynamik bewegter Körper" (Ann. Phys. 17) computes the work needed to accelerate an electron and obtains (γ − 1)mc², diverging as v approaches c.

1906

— Planck's relativistic dynamics (Verh. Deutsch. Phys. Ges. 8, 136) puts the result on the footing of F = dp/dt.

1928

— Dirac's equation supplies the exact relativistic electron, of which −p⁴/8m³c² is the leading correction; it becomes a standard term in the fine-structure Hamiltonian.

Going further

Texts. Rindler, Introduction to Special Relativity, 2nd ed., §§35–38 · Taylor & Wheeler, Spacetime Physics, 2nd ed., ch. 7 · Griffiths, Introduction to Electrodynamics, 4th ed., §12.2 · Morin, Introduction to Classical Mechanics, ch. 12 · Jackson, Classical Electrodynamics, 3rd ed., §11.5 · Griffiths, Introduction to Quantum Mechanics, 3rd ed., §7.3 (the fine-structure correction).

Provenance. Signature $(+,-,-,-)$ , so $p\cdot p = +m^2c^2$ . Everything above is recomputed in verify_aq-sr-relativistic-kinetic-energy.py: sympy evaluates $\int_0^v v\,d(\gamma mv)$ in closed form and confirms it equals $(\gamma-1)mc^2$ identically; the velocity-series coefficients $\tfrac12, \tfrac38, \tfrac{5}{16}, \tfrac{35}{128}$ are extracted from $(\gamma-1)mc^2$ ; the momentum series $p^2/2m - p^4/8m^3c^2 + p^6/16m^5c^4$ is extracted independently from $\sqrt{(pc)^2+(mc^2)^2} - mc^2$ ; the two series are proved consistent (substituting $p = \gamma mv$ and re-expanding gives difference identically zero); the ladder $K = 3.11$ , $10.92$ , $35.62$ MeV at $\beta = 0.99$ , $0.999$ , $0.9999$ with ratios $3.51$ and $3.26$ tending to $\sqrt{10} = 3.162$ ; the approximation $\gamma \approx 1/\sqrt{2(1-\beta)}$ giving $7.071$ , $22.361$ , $70.711$ against exact $7.089$ , $22.366$ , $70.712$ ; the wrong answers $\tfrac12 mv^2 = 0.250$ MeV and $\tfrac12\gamma mv^2 = 1.77$ MeV at $\beta = 0.99$ ; and the Newtonian speed of a $6.799$ TeV proton, $v = c\sqrt{2K/mc^2} = 120.4\,c$ .

Cross-track. The claim about the atomic fine-structure term is verified against the served content of Relativistic Kinetic Energy Correction, which prints the expansion $T = p^2/2m - p^4/8m^3c^2 + p^6/16m^5c^4 - \cdots$ and the operator $\hat H'_{\text{rel}} = -\hat p^4/8m^3c^2$ , describing it as the simplest of the three corrections that together build hydrogen's fine structure.

Scope. $\vec p = \gamma m\vec v$ is derived at Relativistic Momentum; $E = \gamma mc^2$ and the $E$ -versus- $E_0$ distinction at Relativistic Energy; the norm $p\cdot p = +m^2c^2$ and $E^2 = (pc)^2 + (mc^2)^2$ at Four-Momentum. The force law is Relativistic Force and Newton's Second Law's. The fate of "lost" kinetic energy in an inelastic collision — that it becomes mass — is Mass–Energy Equivalence's and Invariant Mass's.

Builds on. Relativistic Momentum · Relativistic Energy · Four-Momentum. Leads to. Energy–Momentum Relation · Mass–Energy Equivalence · Massless Particles and the Photon · Conservation of Four-Momentum.

Why is the kinetic energy not ½γmv²?

Because that expression is not the result of any derivation — it is $\tfrac12 mv^2$ with a $\gamma$ stuck on, which is guessing rather than physics. The work integral is unambiguous: $K = \int v\,d(\gamma mv)$ , and integrating by parts produces a boundary term $\gamma mv^2$ plus a second integral that partly cancels it. Keep only the boundary term and you get the wrong function. Numerically, at $\beta = 0.99$ an electron's true $K$ is $3.11$ MeV while $\tfrac12\gamma mv^2$ gives $1.77$ MeV — wrong by a factor of $1.75$ , and diverging further as $\beta$ rises.

The velocity series has all-positive coefficients but the momentum series alternates in sign. Are they consistent?

Yes, exactly — and the apparent conflict is a nice test of understanding. The two series are expansions of the same function in different variables, and $p$ and $v$ are not linearly related: $p = \gamma m v$ , which is itself a series in $v$ . Substitute it into $p^2/2m - p^4/8m^3c^2$ and re-expand in powers of $v$ : the $p^2/2m$ term contributes $\tfrac12 mv^2 + \tfrac12 mv^4/c^2$ , the $-p^4/8m^3c^2$ term contributes $-\tfrac18 mv^4/c^2$ , and the $v^4$ pieces combine as $\tfrac12 - \tfrac18 = \tfrac38$ — precisely the velocity-series coefficient. The verify script does this symbolically and gets zero difference. A negative coefficient in $p$ and a positive one in $v$ are describing the same physics.

Does the work–energy theorem still hold in relativity?

Yes, and unchanged: the work done by the net force equals the increase in kinetic energy, $dK = \vec F\cdot d\vec x$ . What changes is only what you substitute for $\vec F$ — namely $d\vec p/dt$ with $\vec p = \gamma m\vec v$ . This is one of the pleasing economies of relativistic mechanics: the structure of the Newtonian laws survives almost intact, and essentially all of the new physics enters through the single replacement of $m\vec v$ by $\gamma m\vec v$ . It is a good habit, when meeting a relativistic formula, to ask which Newtonian statement it is and what got replaced.

What is the kinetic energy of a photon?

The question does not have an answer, and that is the correct response rather than an evasion. $K$ is defined as $E - mc^2$ : the energy a body has beyond its rest energy. A photon has $m = 0$ , so it has no rest energy to subtract, and its total energy $E = pc$ is not usefully split into "rest" and "kinetic" parts. You may of course say that all of a photon's energy is energy of motion, and informally people do — but the formula $K = (\gamma - 1)mc^2$ is $\infty \times 0$ for a photon and must not be evaluated. Use $E = pc$ . Massless particles are Massless Particles and the Photon's subject.

If a fast electron is nearly at c, why does giving it more energy do anything at all?

Because almost nothing that matters about a fast particle is its speed. Its momentum keeps growing, without limit; its energy keeps growing; its $\gamma$ keeps growing, which lengthens its lifetime and stiffens it against deflection; and when it hits a target, what it delivers is energy, not speed. A $50$ GeV electron and a $10$ GeV electron cross the laboratory in almost exactly the same time, and produce entirely different physics on arrival. Speed saturates; everything else does not. This is why the relativistic quantity worth tracking is $\gamma$ — and it is why beams are quoted in GeV.

How much of the LHC's 6.8 TeV per proton is kinetic energy?

Essentially all of it. The proton's rest energy is $938.272$ MeV, so $K = 6\,800\,000 - 938.272 = 6\,799\,061.7$ MeV $= 6.799$ TeV — that is $99.986\%$ of the total. The proton is deep in the ultrarelativistic regime, where $K \approx E \approx pc$ and the mass has all but switched itself off. It is worth noticing what this does to the Newtonian question "how fast is it going": feeding that kinetic energy into $\tfrac12 mv^2$ returns $v = 120.4\,c$ , an answer that is not merely inaccurate but meaningless. The formula has been extrapolated a hundredfold beyond its domain.