Invariant Mass

Take two photons. Each has zero mass — that is not an approximation, it is exact. Send them flying in opposite directions, each with energy $E$ . Now ask what the system of two photons weighs. The answer is not zero. It is $M = 2E/c^2$ . Nothing has been added. There are two massless objects in the box and nothing else, and the box has mass. This is not a trick of definitions: that $M$ is the thing that resists acceleration if you try to push the pair around, the thing that gravitates, the thing a detector reconstructs and calls a particle. Mass has been built out of massless things, and it happened because mass is not something you add up. It is the length of a four-vector — and the length of a sum of arrows is not the sum of their lengths. That single observation is the most useful idea in relativistic dynamics. It is why a proton weighs a hundred times more than the quarks inside it. It is why a bound nucleus weighs less than its parts. It is why you cannot compute the energy needed to make a new particle by adding up laboratory energies, and why every modern accelerator is a collider. And it is how particles are found: the Higgs boson announced itself as a bump in a histogram of exactly this quantity.

Key takeaways

Signature $(+,-,-,-)$ is locked: $\eta_{\mu\nu} = \mathrm{diag}(+1,-1,-1,-1)$ , so $U\cdot U = +c^2$ and $\boxed{p\cdot p = +m^2c^2}$ — not $-m^2c^2$ . Every result below is re-derived from that metric.
Owned result: the invariant mass of a SYSTEM is not the sum of its parts' masses. M^2c^4 = \Big(\textstyle\sum_i E_i\Big)^2 - \Big|\textstyle\sum_i \vec p_i\Big|^2c^2 Four-momenta add; masses do not. Two back-to-back photons of energy $E$ , each with $m = 0$ , form a system of mass $M = 2E/c^2$ .
Owned result: two particles of mass $m$ colliding head-on at $\gamma$ have $M = 2\gamma m > 2m$ . The kinetic energy became mass. Bound systems run the other way: $M < \sum m_i$ , and the deficit is the binding energy.
Owned result: thresholds are computed with the invariant $s$ , NEVER with a lab-frame energy sum. For $p + p \to p+p+p+\bar p$ on a fixed target, $s_{\text{thr}} = (4m_pc^2)^2$ and $s = 2m_pc^2E_{\text{lab}} + 2m_p^2c^4$ give $E_{\text{lab}} = 7m_pc^2 = \mathbf{6.568\ \text{GeV}}$ (TOTAL), i.e. $K_{\text{lab}} = 6m_pc^2 = \mathbf{5.630\ \text{GeV}}$ (KINETIC). Those are different numbers. Say which one you mean.
Owned result: this is why colliders exist. $\sqrt{s}_{\text{collider}} = 2E$ ; $\sqrt{s}_{\text{fixed}} \approx \sqrt{2m_pc^2E}$ . The LHC's $6.8$ TeV beams give $\sqrt{s} = 13.6$ TeV. The same beam on a stationary target gives only $113$ GeV — $120$ times less. To reach $13.6$ TeV on a fixed target you would need a beam of $9.86\times10^{16}$ eV, about $\mathbf{14{,}500\times}$ the LHC's.
Key idea: for two photons, $m^2c^4 = 2E_1E_2(1-\cos\theta)$ — and THIS is how particles are discovered. Two $67.5$ MeV photons back-to-back reconstruct to $135$ MeV: the $\pi^0$ . The Higgs was a bump at $125$ GeV in the same histogram. An invariant-mass peak is a particle, because the invariant is the one number every observer agrees on.

The core insight

Ordinary vectors already teach the lesson: two arrows of length 1, pointing opposite ways, sum to an arrow of length 0. Length is not additive. In Minkowski space the same is true, with a twist — the "length" is $p\cdot p = E^2/c^2 - |\vec p|^2$ , and its indefinite signature means the sum of two null vectors ( $p\cdot p = 0$ ) can be strictly timelike ( $P\cdot P > 0$ ). That is the whole of "mass from massless things", and it is not a paradox but a triangle inequality behaving as the geometry requires. Add the four-momenta first. Take the norm second. Never the other way round.

Where this lives today

Almost everything you weigh is invariant mass, and almost none of it is the mass of the parts. The three quarks in a proton have rest masses summing to a few MeV; the proton weighs $938$ MeV. Over $99\%$ of the proton's mass — and therefore of yours — is the energy of the gluon field and the confined motion of the quarks, bound into a system whose total four-momentum has a norm far larger than the sum of its constituents'. The two-photon example is not an exotic curiosity; it is a toy model of you. The same equation runs the world's detectors in the opposite direction. An LHC event is reconstructed as a list of four-momenta, and every question asked of it is a contraction. To search for a new particle you choose a decay channel, add the four-momenta of the candidate daughters, take the norm, and histogram it across millions of events. Background combinations smear out; genuine decays of a common parent stack at one value of $M$ . The peak's position is the particle's mass and its width is related to the particle's lifetime — and because $M$ is a Lorentz invariant, it does not matter that the parent was produced with some unknown, wildly varying boost. Nothing else in the event has that property. It is the invariance that makes the measurement possible. And it is why the accelerator itself looks the way it does. The $\sqrt{E}$ penalty of Step 6 is unforgiving: fixed-target machines gave the field the antiproton, the pion and a generation of hadron spectroscopy, and then ran out of road. Every energy-frontier machine since — the SPS collider, the Tevatron, LEP, the LHC — collides beams head-on so that the laboratory is the centre-of-momentum frame and $\sqrt{s} = 2E$ . The price is that colliding two beams is vastly harder than hitting a stationary block of metal. Physicists pay it because a square root is not negotiable.

The physical picture

Everything here follows from one rule, and the rule is that four-momenta add. If a system is made of parts with four-momenta $p_i^\mu$ , its total four-momentum is $P^\mu = \sum_i p_i^\mu$ — a plain componentwise sum, because that is what four-vectors do, and because conservation of four-momentum is a statement about this sum. The system's mass is then defined the way any four-vector's mass is defined: by the invariant norm. In our locked signature $\eta_{\mu\nu} = \mathrm{diag}(+1,-1,-1,-1)$ ,

P\cdot P = \left(\frac{E_{\text{tot}}}{c}\right)^2 - \big|\vec P_{\text{tot}}\big|^2 = M^2c^2.

And now the trap springs. Each individual $p_i$ has norm $m_i^2c^2$ . The sum has norm $M^2c^2$ . And these are not related by $M = \sum_i m_i$ , because the norm of a sum is not the sum of the norms. Nothing in the algebra ever suggested it would be. Yet the instinct that mass is additive is so deeply set — it is true in Newtonian mechanics, and it is true to a part in $10^{10}$ in a chemistry laboratory — that this is reliably the single most costly mistake in relativistic dynamics. The sharpest possible demonstration uses objects that have no mass at all, so that there is nothing to add up.

Figure

PUNCHLINE — mass built out of massless things. Two photons, each with p·p = 0 and therefore m = 0 EXACTLY, flying back-to-back with energy E. Their four-momenta ADD: the momenta cancel and the energies do not, so the total is P = (2E/c, 0) — a purely timelike four-vector. Its norm is P·P = 4E²/c² = M²c², so the SYSTEM has mass M = 2E/c². Nothing was added to the box. Mass is the LENGTH of the total four-vector, and the length of a sum of arrows is not the sum of their lengths

Look at what the momenta did. They cancelled, because the photons go opposite ways. The energies could not cancel, because energy is positive for both. So the total four-vector points straight up the time axis — it is purely timelike, it has a rest frame, and in that frame it has energy $2E$ and no momentum at all. A thing with energy and no momentum is a thing at rest with mass $2E/c^2$ . The photons individually have no rest frame; their pair does.

The core equations

The definition, re-derived from the locked metric and from nothing else:

P^{\mu} = \sum_i p_i^{\mu} = \left(\frac{\sum_i E_i}{c},\ \sum_i \vec p_i\right), \qquad \boxed{\;P\cdot P = M^2c^2 \;\Longleftrightarrow\; M^2c^4 = \Big(\sum_i E_i\Big)^2 - \Big|\sum_i \vec p_i\Big|^2 c^2\;} \qquad (1)

where: $\eta_{\mu\nu} = \mathrm{diag}(+1,-1,-1,-1)$ , so the time part enters with a plus sign and the spatial part with a minus. Physical reading: the sign matters and it is easy to get wrong. In the opposite convention $(-,+,+,+)$ one writes $p\cdot p = -m^2c^2$ , and every intermediate sign in this article flips. A formula copied without its metric is a formula copied without its meaning. Here: $p\cdot p = +m^2c^2$ , and a timelike four-vector has positive norm. The Mandelstam invariant, which is the same quantity wearing its collider-physics hat:

s \;\equiv\; \Big(\sum_i E_i\Big)^2 - \Big|\sum_i \vec p_i\Big|^2c^2 \;=\; M^2c^4, \qquad \boxed{\sqrt{s} \;=\; Mc^2} \qquad (2)

where: $\sqrt{s}$ is universally called the centre-of-momentum energy. Physical reading: $\sqrt{s}$ is the total energy of the system in the frame where its momentum vanishes, and it is therefore the energy actually available to make new things. It is a Lorentz scalar, so every observer computes the same value — which is exactly why it, and not any laboratory energy, is what a threshold calculation must use. (The Mandelstam variables $t$ and $u$ belong to Mandelstam Invariants; the centre-of-momentum frame itself to Centre-of-Momentum Frame. Here we need only the invariant.) And the formula that finds particles:

\boxed{\,m_{\gamma\gamma}^2c^4 \;=\; 2E_1E_2\,(1 - \cos\theta)\,} \qquad (3)

where: $E_1, E_2$ — the energies of two photons; $\theta$ — the angle between them. Physical reading: it vanishes only when $\theta = 0$ (collinear photons) and is otherwise strictly positive. Feed a detector's measured photon pairs into it, histogram the result, and a particle that decays to two photons appears as a peak. That is not a metaphor for how discoveries are made. It is literally how they are made.

Deriving the system invariant

Derivation AThe invariant mass of a system, and why it is not the sum of the parts

The claim is easy to state and constantly disbelieved: the mass of a system is not the sum of the masses of its parts. Since this contradicts an intuition that is correct in Newtonian mechanics and very nearly correct in a chemistry laboratory, it needs a derivation that leaves no room to wriggle — and it needs the sharpest possible example, one where the parts have no mass at all, so that no amount of squinting can make the arithmetic come out additive. The metric signature is doing real work in what follows and we re-derive every sign from $\eta = \mathrm{diag}(+1,-1,-1,-1)$ rather than importing it.

Step 1. Add the four-momenta. Conservation of four-momentum is a statement about $P^\mu = \sum_i p_i^\mu$ , and four-vectors add componentwise: $P^\mu = \big(\sum_i E_i/c,\ \sum_i \vec p_i\big)$ . Why: this step is the whole of the physics, and it is the step people skip. Four-momenta add. Masses do not. The mass is not a component; it is a norm, and norms are not linear.

Step 2. Take the norm, with the signature written out. P\cdot P = \eta_{\mu\nu}P^\mu P^\nu = (P^0)^2 - |\vec P|^2 = \Big(\frac{\sum_i E_i}{c}\Big)^2 - \Big|\sum_i \vec p_i\Big|^2 \;\equiv\; M^2c^2. Why: the plus on the time slot and the minus on the space slots come from $\eta$ , and from nowhere else. In $(-,+,+,+)$ this would read $-M^2c^2$ and every sign below would flip — which is why one must never inherit a norm from a source without checking its metric.

Step 3. Check it on one particle. For a single particle $P = p$ , and $p\cdot p = m^2c^2$ recovers the ordinary rest mass. Why: the definition had better reduce to the familiar thing in the familiar case, and it does. This is the sanity check that licenses the generalisation.

Step 4. Now break additivity, with the extreme case. Two photons, each of energy $E$ , back-to-back along $x$ : $p_1 = (E/c,\ +E/c,\ 0,\ 0)$ and $p_2 = (E/c,\ -E/c,\ 0,\ 0)$ . Each is null: $p_i\cdot p_i = (E/c)^2 - (E/c)^2 = 0$ , so $m_1 = m_2 = 0$ exactly. But the sum is $P = (2E/c,\ 0,\ 0,\ 0)$ , whose norm is P\cdot P = \Big(\frac{2E}{c}\Big)^2 - 0 = \frac{4E^2}{c^2} = M^2c^2 \qquad\Longrightarrow\qquad \boxed{M = \frac{2E}{c^2}}. Why: the momenta cancelled and the energies did not — energy is positive for both photons, momentum is a vector and points opposite ways. A system of two massless particles has mass. And note it is not a limit or an approximation: $m_1$ and $m_2$ are exactly zero and $M$ is exactly $2E/c^2$ .

Step 5. Do the general two-photon case, because it is the one that finds particles. With $p_i = (E_i/c)(1, \hat n_i)$ and $\hat n_1\cdot\hat n_2 = \cos\theta$ : M^2c^4 = (E_1+E_2)^2 - \big|E_1\hat n_1 + E_2\hat n_2\big|^2 = \big(E_1^2 + 2E_1E_2 + E_2^2\big) - \big(E_1^2 + E_2^2 + 2E_1E_2\cos\theta\big) = \boxed{2E_1E_2(1-\cos\theta)}. Why: it is zero only for $\theta = 0$ . Two photons travelling in exactly the same direction form a massless system; any opening angle at all gives the pair a mass. Back-to-back ( $\theta = \pi$ ) maximises it at $M^2c^4 = 4E_1E_2$ , recovering Step 4 when $E_1 = E_2 = E$ .

Step 6. Do the massive two-body case, and read off where the mass came from. Two particles of mass $m$ approaching head-on with Lorentz factor $\gamma$ : $p_{1,2} = (\gamma mc,\ \pm\gamma mv,\ 0,\ 0)$ , so $P = (2\gamma mc,\ \vec 0\,)$ and M = 2\gamma m \;>\; 2m. Why: the excess is $2(\gamma-1)mc^2/c^2$ — precisely the two particles' kinetic energy, divided by $c^2$ . The kinetic energy of the constituents has become mass of the system. Run it the other way and you get binding: for a bound system the internal potential energy is negative and $M < \sum_i m_i$ , the deficit being the binding energy. Both effects are the same equation. *Verified symbolically in verify_aq-sr-invariant-mass.py: $\eta = \mathrm{diag}(+1,-1,-1,-1)$ is constructed explicitly and every norm below is contracted against it; $p_i\cdot p_i = 0$ for each photon; $P\cdot P = 4E^2/c^2$ for the back-to-back pair, giving $M = 2E/c^2$ ; the general two-photon norm reduces to $2E_1E_2(1-\cos\theta)$ with sympy returning zero difference; and the head-on massive pair gives $M = 2\gamma m$ .*

Go deeper · Derivation B · 3 minThresholds must be computed with the invariant, and the answer is not what a lab-frame energy sum would suggest

Derivation BThresholds must be computed with the invariant, and the answer is not what a lab-frame energy sum would suggest

Here is the question that makes this the most useful idea in the chapter. To create a proton–antiproton pair you must supply $2m_pc^2 = 1876.5$ MeV of new rest energy. So — fire a proton of a bit more than $1876.5$ MeV at a stationary proton and the pair should appear? No. The true threshold is more than three times higher, and the reason is that most of the beam's energy is not available: it is locked up in the motion of the centre of momentum, which conservation of momentum forbids you to spend. Only the invariant $\sqrt{s}$ is spendable. This derivation shows how to compute it, and ends on a trap that has bitten the corpus before — the difference between a total energy and a kinetic energy.

Step 1. State the reaction and the conservation law. $p + p \to p + p + p + \bar p$ : a beam proton strikes a stationary target proton, and a proton–antiproton pair is created. (Baryon number is conserved: two protons in, three protons and one antiproton out.) Why: you cannot make a lone antiproton — baryon number forbids it — so the cheapest reaction must make a proton alongside it, and the threshold is set by four final particles, not two.

Step 2. Find the threshold condition, in the invariant. At threshold the products are created with the least possible energy, which means they are all at rest in the centre-of-momentum frame — moving together, with no relative motion to waste energy on. In that frame the total energy is just $4m_pc^2$ , so \sqrt{s_{\text{thr}}} = 4m_pc^2 \qquad\Longrightarrow\qquad s_{\text{thr}} = \big(4m_pc^2\big)^2 = 16\,m_p^2c^4. Why: this is the step that must be done with the invariant. "All products at rest" is a statement that is only true in one frame — but $s$ is a scalar, so we may evaluate it in the frame where the statement is easy and then use the number in any frame we like. That licence is the entire technique.

Step 3. Compute the same $s$ in the laboratory. Beam proton $(E_{\text{lab}}/c,\ \vec p_{\text{lab}})$ , target at rest $(m_pc,\ \vec 0\,)$ . Then s = \big(E_{\text{lab}} + m_pc^2\big)^2 - \big(p_{\text{lab}}c\big)^2. Using $(p_{\text{lab}}c)^2 = E_{\text{lab}}^2 - m_p^2c^4$ , the $E_{\text{lab}}^2$ terms cancel and s = 2m_pc^2E_{\text{lab}} + 2m_p^2c^4. Why: notice how little survives. The beam's energy enters only linearly, multiplied by the target's rest energy — which is the mathematical origin of the fixed-target penalty in Step 6. A collider, by contrast, gets $s$ growing as the square of the beam energy.

Step 4. Equate the two expressions for the same invariant. 2m_pc^2E_{\text{lab}} + 2m_p^2c^4 = 16m_p^2c^4 \qquad\Longrightarrow\qquad E_{\text{lab}} = 7\,m_pc^2. Why: this is why $s$ is worth learning. We computed one number in the CM frame, where the threshold condition is trivial, and the same number in the lab frame, where the experiment happens — and equating them solved the problem with no Lorentz transformation performed anywhere.

Step 5. ⚠ Now say which energy you mean. $E_{\text{lab}} = 7m_pc^2$ is the beam proton's TOTAL energy: E_{\text{lab}} = 7 \times 938.272 = \mathbf{6567.9\ \text{MeV}} = \mathbf{6.568\ \text{GeV}}\quad\textbf{(TOTAL)}. An accelerator, however, is specified by the KINETIC energy it imparts, which is smaller by one rest energy: K_{\text{lab}} = E_{\text{lab}} - m_pc^2 = 6\,m_pc^2 = 6 \times 938.272 = \mathbf{5629.6\ \text{MeV}} = \mathbf{5.630\ \text{GeV}}\quad\textbf{(KINETIC)}. Why: these are different numbers and both are correct answers to different questions. Quoting one and labelling it the other is a textbook error, and it is exactly the shape of the transcription defect this corpus hunts. Print both. Label both. The historical check: the Bevatron at Berkeley was designed to deliver $6.2$ GeV of kinetic energy — comfortably above $5.63$ GeV — and in 1955 Chamberlain, Segrè, Wiegand and Ypsilantis used it to find the antiproton.

Step 6. Extract the general lesson. Rearranging Step 3 for $E_{\text{lab}} \gg m_pc^2$ : \sqrt{s}_{\text{fixed}} \approx \sqrt{2m_pc^2\,E_{\text{lab}}} \qquad\text{versus}\qquad \sqrt{s}_{\text{collider}} = 2E. Why: the fixed-target CM energy grows only as the square root of the beam energy. Quadrupling the beam buys you a factor of two. A collider's CM energy grows linearly. That asymmetry — a square root against a straight line — is the single argument that built every major machine of the last fifty years, and the next figure puts a number on it. *Verified in verify_aq-sr-invariant-mass.py: sympy solves $2m_pE_{\text{lab}} + 2m_p^2 = 16m_p^2$ and returns $E_{\text{lab}} = 7m_p$ exactly; the numbers $E_{\text{lab}} = 6567.9$ MeV (TOTAL) and $K_{\text{lab}} = 5629.6$ MeV $= 5.63$ GeV (KINETIC) are both computed and both grep-asserted against the prose; and $s = 2m_pc^2E_{\text{lab}} + 2m_p^2c^4$ is derived from the four-momenta rather than quoted.*

Figure

PUNCHLINE — the one number that is the argument for colliders. Take the SAME 6.8 TeV proton beam. Fire two of them at each other and √s = 2E = 13.6 TeV. Fire one at a stationary proton and √s = √(2mc²E) = only 113 GeV — 120 TIMES LESS, from an identical beam. The fixed-target CM energy grows as the SQUARE ROOT of the beam energy; a collider's grows linearly. To reach 13.6 TeV against a stationary target you would need a beam of 9.86×10¹⁶ eV = 98,600 TeV — about 14,500× the LHC's energy, and a ring that would not fit on the planet

Limiting cases

Case	Result	Meaning
single particle	$M = m$	the definition reduces correctly
all parts at rest, same frame	$M = \sum_i m_i$	the only case where mass is additive
two photons, angle $\theta$	$M^2c^4 = 2E_1E_2(1-\cos\theta)$	zero only if collinear
two photons, back-to-back	$M = 2E/c^2$	mass from massless things
two masses head-on at $\gamma$	$M = 2\gamma m > 2m$	kinetic energy became mass
bound system	$M < \sum_i m_i$	the deficit is the binding energy
collider	$\sqrt{s} = 2E$	grows linearly with beam energy
fixed target, $E \gg mc^2$	$\sqrt{s} \approx \sqrt{2mc^2E}$	grows only as $\sqrt{E}$ — the penalty

Common trapThe signature is locked, and the sign of the norm is not a detail

Throughout this chapter $\eta_{\mu\nu} = \mathrm{diag}(+1,-1,-1,-1)$ , so $U\cdot U = +c^2$ and $p\cdot p = +m^2c^2$ , and a timelike four-vector has positive norm. In the opposite convention $(-,+,+,+)$ every one of those flips, and one writes $p\cdot p = -m^2c^2$ . Both conventions describe the same physics and neither is more correct — but mixing them silently corrupts every invariant mass and every threshold you compute, and no numerical check will catch it, because the error is a sign that propagates consistently through wrong algebra. Sources disagree freely, and some disagree with themselves. Whenever you meet $p\cdot p$ in a book or a paper, find the metric before you trust the sign.

Common trapTOTAL energy and KINETIC energy are different numbers. Say which one you mean

The antiproton threshold is $E_{\text{lab}} = 7m_pc^2 = 6.568$ GeV of total energy, which is $K_{\text{lab}} = 6m_pc^2 = 5.630$ GeV of kinetic energy. Both are right; they answer different questions. Accelerators are conventionally specified by the kinetic energy they deliver, so "the antiproton threshold is 5.63 GeV" is the statement a machine builder wants — and indeed the Bevatron was built for $6.2$ GeV of kinetic energy to clear it. But a four-momentum calculation naturally produces the total. The two differ by exactly one rest energy, $938.272$ MeV, and quoting one under the other's name is a small, silent, and thoroughly typical error.

Worked examples

1 · Find the π⁰ in a pile of photons

A detector records two photons, each of energy $67.5$ MeV, travelling in exactly opposite directions. What produced them?

Step 1 — identify. A single particle decaying at rest into two photons must give them equal energies and opposite momenta. So reconstruct the invariant mass of the pair and see what it matches.

Step 2 — apply. Back-to-back means $\theta = 180°$ , so $\cos\theta = -1$ and $1 - \cos\theta = 2$ : M^2c^4 = 2E_1E_2(1-\cos\theta) = 2 \times 67.5 \times 67.5 \times 2 = 4 \times 67.5^2 = 18225\ \text{MeV}^2.

Step 3 — take the root. $Mc^2 = \sqrt{18225} = \mathbf{135.0\ \text{MeV}}$ .

Step 4 — the boundary number. The neutral pion has a measured mass of $134.977$ MeV — so this pair came from a $\pi^0$ , and we identified it without ever seeing it, from two massless particles it left behind. This is the central technique of experimental particle physics. Collect millions of photon pairs, compute $M$ for each, histogram the results: combinatorial background spreads out smoothly, while pairs that really came from a common parent pile up at that parent's mass. A bump in the histogram is a particle. The Higgs boson was announced in July 2012 as exactly such a bump, in exactly this variable, at $125$ GeV.

2 · The antiproton threshold, done two ways

What kinetic energy must a proton beam have to create an antiproton on a stationary hydrogen target, via $p + p \to p+p+p+\bar p$ ?

Step 1 — identify (and reject the naive answer). The tempting answer is " $2m_pc^2 = 1876.5$ MeV, the rest energy of the pair you are making". This is wrong, and it is wrong because momentum conservation forbids the products from being at rest in the laboratory: the beam brought in momentum, so the products must carry it away, and that momentum costs kinetic energy which is therefore unavailable for creation. The invariant $s$ is what accounts for this automatically.

Step 2 — the invariant at threshold. In the CM frame, threshold means all four products are produced at rest together: $\sqrt{s_{\text{thr}}} = 4m_pc^2$ , so $s_{\text{thr}} = 16\,m_p^2c^4$ .

Step 3 — the same invariant in the lab. $s = (E_{\text{lab}} + m_pc^2)^2 - (p_{\text{lab}}c)^2 = 2m_pc^2E_{\text{lab}} + 2m_p^2c^4$ . Setting the two equal: $2m_pc^2E_{\text{lab}} = 14\,m_p^2c^4$ , hence $E_{\text{lab}} = 7m_pc^2$ .

Step 4 — the boundary numbers, both of them. E_{\text{lab}} = 7 \times 938.272 = \mathbf{6567.9\ \text{MeV}} = \mathbf{6.568\ \text{GeV}}\ \ \textbf{(TOTAL)} K_{\text{lab}} = 6 \times 938.272 = \mathbf{5629.6\ \text{MeV}} = \mathbf{5.630\ \text{GeV}}\ \ \textbf{(KINETIC)} So a machine must deliver $5.63$ GeV of kinetic energy — three times the naive $1.88$ GeV, and the factor of three is entirely the cost of the momentum the beam forced the products to carry. History confirms the arithmetic: the Bevatron was designed to reach $6.2$ GeV of kinetic energy, deliberately above this threshold, and in 1955 Chamberlain, Segrè, Wiegand and Ypsilantis found the antiproton with it. A relativistic invariant, computed on a blackboard, sized a machine.

Physics of the portrait

The portrait is a scalar — one number, computed the same way by every observer in the universe, and the single most productive quantity in experimental physics. Its story is a straight line from a piece of geometry to a bump in a histogram.

Origins · 2 minInvariant mass and how this was discovered

Invariant mass

1908–2012 · Minkowski's norm · the Bevatron's antiproton · the π⁰ peak · the Higgs bump at 125 GeV

The quantity begins as pure geometry. In 1908 Hermann Minkowski told the Cologne assembly that space and time had merged into a four-dimensional continuum, and gave that continuum an indefinite metric — one in which the "length" of a vector could be positive, negative or zero. The invariant mass is nothing more than the length of the energy–momentum vector in that geometry, and every strange property it has traces back to the minus signs Minkowski put in the metric. Its usefulness took a generation to emerge. By the 1950s it had become the working tool of accelerator design: the Bevatron at Berkeley was built to exceed the $5.63$ GeV threshold that this article computes, and in 1955 Owen Chamberlain, Emilio Segrè, Clyde Wiegand and Thomas Ypsilantis used it to produce the antiproton — a particle whose existence Dirac's equation had implied and whose price a Lorentz invariant had set. Then the technique inverted. Instead of using the invariant to decide what energy a machine needed, experimentalists began using it to ask what a detector had seen: add up the four-momenta of the debris, take the norm, and plot. Short-lived particles that no detector can ever touch appear as peaks. The $\pi^0$ , decaying to two photons, is the textbook case; the whole zoo of resonances was mapped this way; and on 4 July 2012 the ATLAS and CMS collaborations announced the Higgs boson as a bump near 125 GeV in the diphoton invariant-mass spectrum — the same $2E_1E_2(1-\cos\theta)$ that identifies a $\pi^0$ , applied a thousand times further up the energy scale. Minkowski's norm has been finding particles for a century.

History

1908

— Minkowski's "Raum und Zeit" gives spacetime an indefinite metric; the norm of the energy–momentum four-vector becomes a Lorentz scalar.

1954

— The Bevatron at Berkeley is commissioned, designed for 6.2 GeV of kinetic energy — deliberately above the 5.63 GeV antiproton threshold set by the invariant s.

1955

— Chamberlain, Segrè, Wiegand and Ypsilantis produce and identify the antiproton at the Bevatron; Segrè and Chamberlain receive the 1959 Nobel Prize.

1958

— Mandelstam introduces the invariants s, t and u for two-body scattering; s becomes the standard name for the squared centre-of-momentum energy.

2012

— ATLAS and CMS announce the Higgs boson on 4 July as a peak near 125 GeV in the diphoton invariant-mass spectrum, reconstructed from exactly the formula in this article.

Going further

Texts. Rindler, Introduction to Special Relativity, 2nd ed., §§5.4–5.6 · Taylor & Wheeler, Spacetime Physics, 2nd ed., ch. 7–8 · Griffiths, Introduction to Elementary Particles, 2nd ed., ch. 3 · Griffiths, Introduction to Electrodynamics, 4th ed., §12.2 · Hartle, Gravity, ch. 5 · Okun, "The concept of mass", Physics Today 42(6), 31 (1989).

Provenance. Signature $(+,-,-,-)$ , $\eta = \mathrm{diag}(+1,-1,-1,-1)$ , so $U\cdot U = +c^2$ and $p\cdot p = +m^2c^2$ . The AllQFT base for this article is in the opposite signature and is internally inconsistent with itself — it writes $m^2c^2 = -p_\mu p^\mu$ in one card and $s = (\sum p)\cdot(\sum p)c^2 = E_{\text{cm}}^2$ in another. Nothing was inherited from it: every norm above is re-derived by explicit contraction against $\eta$ , and verify_aq-sr-invariant-mass.py constructs $\eta$ and performs the contractions. Also recomputed there: $p_i\cdot p_i = 0$ for each photon and $P\cdot P = 4E^2/c^2 \Rightarrow M = 2E/c^2$ ; the general $M^2c^4 = 2E_1E_2(1-\cos\theta)$ (sympy, zero difference); $M = 2\gamma m$ head-on; $s = 2m_pc^2E_{\text{lab}} + 2m_p^2c^4$ derived, not quoted; sympy solving $s = 16m_p^2c^4$ for $E_{\text{lab}} = 7m_pc^2$ exactly; $E_{\text{lab}} = 6567.9$ MeV $= 6.568$ GeV (TOTAL) and $K_{\text{lab}} = 5629.6$ MeV $= 5.630$ GeV (KINETIC), both asserted against the prose; $\sqrt{s} = 13.6$ TeV for the collider against $113$ GeV for the same beam on a fixed target (a factor of $120$ ); the fixed-target energy needed for $13.6$ TeV, $9.86\times10^{16}$ eV $= 98{,}600$ TeV, a factor of $14{,}500$ ; and $2 \times 67.5$ MeV back-to-back $\Rightarrow 135.0$ MeV against the measured $\pi^0$ mass of $134.977$ MeV.

Scope. The single-particle norm $p\cdot p = +m^2c^2$ and $E^2 = (pc)^2 + (mc^2)^2$ are derived at Four-Momentum. $E = \gamma mc^2$ and $\vec p = \gamma m\vec v$ are Relativistic Energy's and Relativistic Momentum's. This article owns the system invariant and $\sqrt{s}$ , and uses $s$ by name only: the Mandelstam variables $t$ and $u$ belong to Mandelstam Invariants, the centre-of-momentum frame and its kinematics to Centre-of-Momentum Frame, collision kinematics as a subject to Relativistic Elastic Collisions, Relativistic Inelastic Collisions and Threshold Energy for Particle Production, and the photon to Massless Particles and the Photon. What $\Delta m = \Delta E/c^2$ means is Mass–Energy Equivalence's.

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

How can two massless particles have mass? Is this a trick of words?

No — it is the ordinary behaviour of lengths. Two ordinary arrows of length 1 pointing opposite ways sum to an arrow of length 0: the length of a sum is not the sum of the lengths, and nobody finds that troubling. Minkowski space does the same thing in reverse, because its metric is indefinite: two null vectors ( $p\cdot p = 0$ ) can sum to a strictly timelike one ( $P\cdot P > 0$ ). The mass $M = 2E/c^2$ is real in every sense that matters — the pair has that much energy in its rest frame, resists acceleration accordingly, and gravitates accordingly. What is not real is the intuition that mass is a substance you can count.

Is the invariant mass of a system conserved?

It is both invariant (every observer computes the same value) and conserved (unchanged through an interaction), and those are logically different properties that happen to coincide here. Invariance follows because $M$ is a Lorentz scalar. Conservation follows because $P^\mu = \sum p_i^\mu$ is conserved, so its norm is too. Note carefully what this does not say: the sum of the constituents' masses, $\sum_i m_i$ , is neither invariant-in-the-useful-sense nor conserved — it changes freely in any reaction that creates or destroys particles. It is $M$ , the norm of the total, that is protected.

Why can I not just add up the laboratory energies to find a threshold?

Because momentum conservation will not let you spend all of it. If the incoming particles carry net momentum, the outgoing ones must carry the same net momentum — and that motion costs kinetic energy which is therefore not available to create rest mass. The invariant $s$ automatically subtracts this unavailable part: $\sqrt{s}$ is the energy in the frame where the total momentum is zero, and that is precisely the energy you may spend. In the antiproton example the naive lab-frame sum suggests $1.88$ GeV and the truth is $5.63$ GeV of kinetic energy — a factor of three, all of it the cost of momentum you were forced to carry.

If a bound nucleus weighs less than its parts, is that the same equation?

The very same one, run in the other direction. $Mc^2 = \sum_i m_ic^2 + \sum_i K_i + U$ , where $U$ is the internal potential energy. For a bound system $U$ is negative and large enough to outweigh the internal kinetic energies, so $M < \sum_i m_i$ , and the deficit is the binding energy divided by $c^2$ . For two particles flying apart there is no binding, only kinetic energy, so $M > \sum_i m_i$ . Same formula, opposite signs of the internal energy. This is why fusion releases energy — helium weighs less than its nucleons — and it is Mass–Energy Equivalence's subject.

What exactly is √s, and why does everyone use it instead of the beam energy?

$\sqrt{s} = Mc^2$ is the total energy of the colliding system in the frame where its momentum is zero — the energy that is actually available to make new particles. It is a Lorentz scalar, so it is the same number in the laboratory, in the CM frame, and in the frame of a passing spacecraft. A beam energy has none of that: it is a frame-dependent component, and two machines with the same beam energy can have wildly different physics reach depending on what they collide it with. When a physicist says "the LHC runs at $13.6$ TeV" they mean $\sqrt{s}$ , not the beam energy, which is $6.8$ TeV.

The base article for this topic writes m²c² = −pμpμ. Is that wrong?

It is wrong in this chapter's convention, and it is a good illustration of why the convention must be stated. In the signature $(-,+,+,+)$ , $p\cdot p = -m^2c^2$ is perfectly correct. In our signature $(+,-,-,-)$ it is $+m^2c^2$ . Neither is more true than the other; they are different bookkeeping schemes for the same physics. What is not acceptable is what the base does next — it uses the $(-,+,+,+)$ form in one card and the $(+,-,-,-)$ form in another, so its own signs do not agree with each other. Nothing in this article was inherited from it: every norm here is contracted explicitly against $\eta = \mathrm{diag}(+1,-1,-1,-1)$ . Always find the metric before you trust the sign.