Centre-of-Momentum Frame

In 1954 the Berkeley Radiation Laboratory finished a machine built around a single number. An antiproton must be made in pairs — baryon number forbids making one alone — so the cheapest reaction on hydrogen is $p + p \to p + p + p + \bar p$ , and the four particles coming out weigh $4m_pc^2 = 3753\ \mathrm{MeV}$ . The naive budget is obvious: supply $3753\ \mathrm{MeV}$ of beam and you can afford them. The Bevatron was designed for 6.2 GeV of kinetic energy, nearly four times that. It was not over-engineering. It was arithmetic. Fire a $6.2\ \mathrm{GeV}$ proton at a proton sitting still and the lab ledger reads: beam $7.138\ \mathrm{GeV}$ , target rest energy $0.938\ \mathrm{GeV}$ , total $8.077\ \mathrm{GeV}$ . Yet only $3.893\ \mathrm{GeV}$ of it can become new matter. The other $4.183\ \mathrm{GeV}$ is unspendable — not lost, not radiated, but locked into the forward motion the products must inherit, because the beam brought momentum and momentum does not evaporate. The frame that accounts for this honestly is the one where the total three-momentum vanishes: the centre-of-momentum frame. Its total energy is $\sqrt{s}$ , and $\sqrt{s}$ — not the beam energy, not the sum of lab energies — is what a collision has to spend. For the Bevatron, $\sqrt{s} = 3.893\ \mathrm{GeV}$ against a requirement of $3.753\ \mathrm{GeV}$ : a margin of $140\ \mathrm{MeV}$ . It cleared the bar by less than a sixth of a proton, and the antiproton appeared in 1955.

Key takeaways

Signature $(+,-,-,-)$ : $p\cdot p = (E/c)^2 - |\vec p\,|^2 = +\,m^2c^2$ on the mass shell. Timelike norms are positive.
Convention lock: $s$ carries units of ENERGY², so $\sqrt{s}$ is an ENERGY. $s \equiv (\sum_i E_i)^2 - |\sum_i \vec p_i|^2c^2$ , and $\sqrt{s} = 13\ \mathrm{TeV}$ at the LHC. The four-momentum norm $P\cdot P$ is instead in momentum² units, with $s = (P\cdot P)c^2$ . Both are standard; mixing them inside one calculation is a factor-of- $c^2$ disaster, and the AllQFT base for this article does exactly that.
The CM frame is the frame in which $\sum_i \vec p_i = \vec 0$ . It exists — uniquely, up to rotations — for any system whose total four-momentum is timelike, and it moves at $\vec V_{\mathrm{CM}} = c^2\vec P_{\mathrm{tot}}/E_{\mathrm{tot}}$ relative to the lab.
There $P^\mu = (Mc,\vec 0\,)$ , so $\sqrt{s} = Mc^2$ : the system's invariant mass, times $c^2$ , IS its centre-of-momentum energy. Invariant Mass owns the concept; this article supplies the frame that makes it a reading.
Owned result — the available energy. Only $\sqrt{s}$ can become rest mass, because in every other frame the products must still carry the system's momentum. Fixed target on a stationary $m_2$ : $\sqrt{s} = \sqrt{m_1^2c^4 + m_2^2c^4 + 2m_2c^2E_1^{\mathrm{lab}}}$ , growing only as $\sqrt{E_1}$ . A head-on collider gives $\sqrt{s} = 2E_{\mathrm{beam}}$ .
Key idea: the lab says how much energy is PRESENT; the CM frame says how much is USABLE, and the ratio is exactly $1/\gamma_{\mathrm{CM}}$ . Matching the LHC's 13 TeV with a fixed target would need a 9.0 × 10⁴ TeV beam — ninety thousand TeV to buy thirteen.

The core insight

Momentum conservation is not a constraint applied at the end of a collision calculation; it is a tax levied at the start. A system with $\vec P_{\mathrm{tot}} \neq \vec 0$ must end with $\vec P_{\mathrm{tot}}$ , so its products can never all be at rest, so some of their energy is permanently spoken for. Boost to the frame where $\vec P_{\mathrm{tot}} = \vec 0$ and the tax vanishes — everything there is spendable. That frame is the CM frame, its energy is $\sqrt{s}$ , and since $\sqrt{s}$ is a Lorentz scalar you may compute it wherever is convenient.

Where this lives today

Every event at a modern collider is processed through this arithmetic before anyone looks at it. Detectors measure four-momenta in the lab; physics questions are asked of invariants. A candidate particle's mass is the $\sqrt{s}$ of its decay products. The "hard scatter" energy of an LHC event is the $\sqrt{s}$ of the two colliding partons — not 13 TeV, because each parton carries only a fraction $x$ of its proton's momentum: $\sqrt{\hat s} = \sqrt{x_1x_2}\,\sqrt{s}$ , with the hard collision's CM frame boosted along the beam by an unknown amount. That is exactly why hadron-collider physicists measure transverse momentum and use rapidity differences — those survive an unknown longitudinal boost. The same logic sets the design of every machine ever built. Fixed target: cheap, dense, high event rate — the right choice for statistics on physics you can already reach, and still the workhorse of neutrino production and nuclear physics. Collider: expensive, thin target, low rate — and the only choice when what you need is $\sqrt{s}$ .

The physical picture

Take two particles. Each carries a four-momentum $p_i^\mu = (E_i/c,\ \vec p_i\,)$ , transforming with $\Lambda$ and with its norm fixed by the particle's mass, $p_i\cdot p_i = +m_i^2c^2$ (Relativistic Four-Momentum owns this; Conservation of Four-Momentum owns the statement that the sum is conserved). Add them:

P^\mu = p_1^\mu + p_2^\mu = \left(\frac{E_1 + E_2}{c},\ \vec p_1 + \vec p_2\right)

This is a four-vector, because four-vectors add. But it is emphatically not on any particle's mass shell, and treating it as though it were is the source of half the errors in this subject. Its norm, however, is still an invariant — and it is the single most useful number in collision physics. Give it a name: $s \equiv (P\cdot P)c^2$ . Now ask a question that sounds trivial and is not: is there a frame in which the total three-momentum is zero? For a single massive particle, yes — that is the rest frame. For a system, yes precisely when $P^\mu$ is timelike, which is to say when $s > 0$ . And where it exists, the physics is simplest: two-body states become back-to-back, decays symmetric, momentum bookkeeping nil.

Figure

PUNCHLINE — the energy ledger of the Bevatron shot. In the lab, 8.077 GeV of energy is present; only 3.893 GeV of it is SPENDABLE. The remaining 4.183 GeV is locked into the forward motion the products are obliged to inherit from the beam's momentum. The antiproton needs 4·m_p·c² = 3.753 GeV — the machine cleared it by 140 MeV. Available fraction = 1/γ_CM = 48.2 percent

That figure is the whole subject in one picture. Green is $\sqrt{s}$ ; red is everything the lab has but cannot use. As the beam energy climbs, red grows faster, and the available fraction $\sqrt{s}/E_{\mathrm{tot}} = 1/\gamma_{\mathrm{CM}}$ collapses.

Figure

The existence proof, drawn. In the E–pc plane the two particles' four-momenta each sit on their own mass hyperbola (p·p = +m²c²). Their SUM P does not — it sits strictly INSIDE the light cone, at s = 15.155 GeV². A boost slides P along the hyperbola √s = 3.893 GeV until its momentum component is zero: that landing point, P = (√s, 0), IS the CM frame. The one system with no CM frame is a pair of collinear photons, whose sum lies ON the cone (s = 0) and can never be stood upright

The boost that stands $P$ upright is unique up to a rotation, which changes nothing. So "the" CM frame is a well-defined object, and every statement made in it is a Lorentz-invariant statement about the collision.

The core equations

Start from the total four-momentum of an isolated system, which is a four-vector because it is a sum of four-vectors, and conserved because each collision conserves it:

P^{\mu} \;=\; \sum_i p_i^{\mu} \;=\; \left(\frac{E_{\mathrm{tot}}}{c},\ \vec P_{\mathrm{tot}}\right), \qquad \vec P_{\mathrm{tot}} = \sum_i \vec p_i \qquad (1)

where: $E_{\mathrm{tot}}$ , $\vec P_{\mathrm{tot}}$ — the totals in whatever frame you stand in. Physical reading: $P^\mu$ is the only object in a collision that both transforms cleanly and survives it. The CM frame kills its spatial part, and the boost that gets you there is read straight off $P^\mu$ :

\vec P_{\mathrm{tot}}^{\;\mathrm{CM}} = \vec 0 \qquad\Longleftrightarrow\qquad \vec V_{\mathrm{CM}} = \frac{c^2\,\vec P_{\mathrm{tot}}}{E_{\mathrm{tot}}}, \qquad \gamma_{\mathrm{CM}} = \frac{E_{\mathrm{tot}}}{\sqrt{s}} \qquad (2)

where: $\vec V_{\mathrm{CM}}$ — the CM frame's velocity seen from the lab. Physical reading: not a mass-weighted average of particle velocities (that object is not even Lorentz-covariant), but the velocity that makes the momentum vanish. And $|\vec V_{\mathrm{CM}}| < c$ requires $|\vec P_{\mathrm{tot}}|c < E_{\mathrm{tot}}$ — exactly $s > 0$ . Now the convention this article — and the four after it — lives or dies by:

\boxed{\; s \;\equiv\; \Big(\textstyle\sum_i E_i\Big)^2 - \Big|\textstyle\sum_i \vec p_i\Big|^2 c^2 \;=\; (P\cdot P)\,c^2, \qquad \sqrt{s} \;=\; E_{\mathrm{CM}} \;=\; Mc^2 \;} \qquad (3)

where: $M$ — the invariant mass of the system (Invariant Mass owns the concept; this article supplies the frame in which it is just energy over $c^2$ ). Physical reading: $s$ has units of energy², so $\sqrt{s}$ is an energy — not a momentum. Write $\sqrt{s} \geq \sum_f m_fc^2$ with an $s$ defined as $(E/c)^2 - |\vec p\,|^2$ and you have compared a momentum to an energy. Both conventions exist ( $P\cdot P$ is in momentum² units, $s = (P\cdot P)c^2$ converts). Pick one, state it, never drift. Evaluate (3) in the lab for the standard arrangement — beam $m_1$ at lab energy $E_1$ , target $m_2$ at rest:

\sqrt{s} \;=\; \sqrt{\,m_1^2c^4 + m_2^2c^4 + 2\,m_2c^2\,E_1^{\mathrm{lab}}\,} \;\;\xrightarrow[\;E_1 \gg mc^2\;]{}\;\; \sqrt{2\,m_2c^2\,E_1^{\mathrm{lab}}} \qquad (4)

where: $E_1^{\mathrm{lab}}$ — the beam's total energy, not its kinetic energy. Physical reading: the useful energy grows only as the square root of the beam energy. Quadruple the accelerator, double the reach. A head-on collider of equal beams already has $\vec P_{\mathrm{tot}} = \vec 0$ , so $\sqrt{s} = 2E_{\mathrm{beam}}$ — linear. Threshold Energy owns that comparison and plots it. Finally, in the CM frame a two-body system's energies and momenta are fixed entirely by $s$ and the two masses:

E_1^{*} = \frac{s + (m_1^2 - m_2^2)c^4}{2\sqrt{s}}, \qquad p^{*} \;=\; \frac{\sqrt{\big[s - (m_1+m_2)^2c^4\big]\big[s - (m_1-m_2)^2c^4\big]}}{2c\sqrt{s}} \qquad (5)

where: $p^{*} \equiv |\vec p_1^{\;*}| = |\vec p_2^{\;*}|$ — the common momentum magnitude, equal and opposite by construction. Physical reading: the $c$ belongs in the denominator — the bracket product is energy⁴, its root energy², and dividing by $c\sqrt{s}$ gives energy/ $c$ , a momentum. Particle Decay Kinematics sets $\sqrt{s} = Mc^2$ here and reads off the monoenergetic decay momentum.

Deriving the frame

Derivation AThe centre-of-momentum frame exists, is unique, and moves at c squared P over E

Everything below rests on a frame we have so far only asserted. Three things need proving: that a frame with $\sum_i\vec p_i = \vec 0$ exists at all; that its velocity relative to the lab is $c^2\vec P_{\mathrm{tot}}/E_{\mathrm{tot}}$ ; and that this velocity is subluminal, since a "frame" moving at $c$ is not a frame. The three turn out to be one statement, decided entirely by the sign of the invariant $s$ — which is the pattern of the whole subject: a question about frames, answered by a scalar.

Step 1. Boost along the momentum. Let the lab see $E_{\mathrm{tot}}$ and $\vec P_{\mathrm{tot}}$ , and boost with velocity $\vec V = V\hat n$ along $\hat n = \vec P_{\mathrm{tot}}/|\vec P_{\mathrm{tot}}|$ . Since $P^\mu$ is a four-vector, its longitudinal momentum component transforms as $P'_{\parallel} = \gamma_V\big(P_{\parallel} - VE_{\mathrm{tot}}/c^2\big)$ . Why: a boost mixes the time component only with the momentum component along the boost direction, so boosting along $\vec P_{\mathrm{tot}}$ leaves nothing transverse to kill.

Step 2. Demand the momentum vanish. Setting $P'_{\parallel} = 0$ gives $P_{\parallel} = VE_{\mathrm{tot}}/c^2$ , hence $\boxed{\vec V_{\mathrm{CM}} = c^2\vec P_{\mathrm{tot}}/E_{\mathrm{tot}}}$ . Why: $\gamma_V$ is never zero, so the bracket must vanish. Note what this velocity is NOT: a mass-weighted average of constituent velocities. Momentum, not mass, is what gets cancelled.

Step 3. Check it is a legal velocity. $|\vec V_{\mathrm{CM}}|/c = |\vec P_{\mathrm{tot}}|c/E_{\mathrm{tot}}$ , which is below $1$ if and only if $|\vec P_{\mathrm{tot}}|^2c^2 < E_{\mathrm{tot}}^2$ — precisely $s > 0$ . Why: existence of the CM frame is EXACTLY the statement that $P^\mu$ is timelike. Not an extra assumption but a property of the system, and an invariant one, so every observer agrees whether a CM frame exists.

Step 4. Find the exception. The sum of two future-pointing timelike-or-null four-vectors is timelike unless both are null and parallel. Two photons flying the same way have $P^\mu$ null: $s = 0$ , no CM frame, no mass. Turn one around and $s = 4E_\gamma^2 > 0$ — a CM frame appears out of nowhere. Why: the sharpest illustration that invariant mass belongs to the SYSTEM and is not a sum over its parts. Two massless particles can make a massive system; two pointed the same way cannot.

Step 5. Uniqueness. Two frames both seeing $\vec P = \vec 0$ differ by a Lorentz transformation preserving $(Mc,\vec 0\,)$ — and the stabiliser of a timelike vector is exactly the rotation group. So the CM frame is unique up to orientation; any boost away from it immediately reintroduces momentum. Why: this licenses the phrase "the CM frame", and lets $E_1^{*}$ and $p^{*}$ count as invariants of the collision rather than accidents of a choice. *Verified in verify_aq-sr-centre-of-momentum-frame.py: boosting $(E_{\mathrm{tot}}/c, \vec P_{\mathrm{tot}})$ with $\vec V = c^2\vec P/E$ returns spatial part exactly zero and time component $\sqrt{s}/c$ ; $|V| < c \Leftrightarrow s > 0$ ; $\gamma_{\mathrm{CM}} = E_{\mathrm{tot}}/\sqrt{s}$ ; and the two-photon cases give $s = 0$ (collinear) and $s = 4E_\gamma^2$ (anti-collinear).*

Go deeper · Derivation B · 2 minThe CM energy is the invariant mass, and the fixed-target formula that follows

Derivation BThe CM energy is the invariant mass, and the fixed-target formula that follows

The invariant $s$ is built from energies and momenta, both of which change under a boost; calling it invariant says that this particular combination does not. That turns $s$ from a definition into a tool: evaluate it wherever the algebra collapses, then read the answer wherever you work. We evaluate it twice — in the CM frame, which supplies the meaning of $\sqrt{s}$ , and in the lab, which supplies the number an experimenter needs.

Step 1. Evaluate $s$ in the CM frame. There $\sum_i\vec p_i = \vec 0$ by definition, so $s = \big(\sum_i E_i^{\mathrm{CM}}\big)^2 - 0 = E_{\mathrm{CM}}^2$ , giving $\sqrt{s} = E_{\mathrm{CM}}$ . Why: the momentum term dies. This is the only frame in which $s$ is the perfect square of something physically obvious — which is why it is the frame that gives $s$ its meaning.

Step 2. Identify the invariant mass. In the CM frame $P^\mu = (E_{\mathrm{CM}}/c,\ \vec 0\,) \equiv (Mc,\ \vec 0\,)$ , which defines $M$ . So $\boxed{\sqrt{s} = Mc^2}$ . Why: a four-vector with no spatial part looks exactly like the four-momentum of a single particle at rest, and $M$ is what that fictitious particle would weigh — a genuine, gravitating mass. But it is NOT $\sum_i m_i$ , and Relativistic Inelastic Collisions is built on the difference.

Step 3. Evaluate the same $s$ in the lab: beam ( $m_1$ , energy $E_1$ , momentum $\vec p_1$ ) on a stationary target ( $m_2$ ). Expanding, $s = \big(E_1 + m_2c^2\big)^2 - |\vec p_1|^2c^2$ . Why: nothing is assumed about the collision — this is arithmetic on the initial state, legitimate because $s$ is conserved (the four-momentum is) and invariant (the norm is).

Step 4. Kill $|\vec p_1|^2c^2$ with the mass shell, $|\vec p_1|^2c^2 = E_1^2 - m_1^2c^4$ . Substituting: $s = E_1^2 + 2m_2c^2E_1 + m_2^2c^4 - E_1^2 + m_1^2c^4$ , so $\boxed{s = m_1^2c^4 + m_2^2c^4 + 2m_2c^2E_1^{\mathrm{lab}}}$ . Why: the $E_1^2$ terms cancel identically. That cancellation is why $s$ is LINEAR in the beam energy — hence why $\sqrt{s}$ grows only as $\sqrt{E_1}$ . The entire economics of accelerator physics sits in it.

Step 5. Check both ends. At $E_1 = m_1c^2$ the formula collapses to $s = (m_1+m_2)^2c^4$ , so $\sqrt{s} = (m_1+m_2)c^2$ — two particles at rest side by side have invariant mass $m_1 + m_2$ , as they must. At $E_1 \gg mc^2$ the mass terms drop and $\sqrt{s} \to \sqrt{2m_2c^2E_1}$ . Why: the low-energy limit is the ONLY place where invariant mass is additive; the high-energy limit is where the subject gets expensive.

Step 6. Contrast the collider. Two equal beams of energy $E$ head-on have $\vec P_{\mathrm{tot}} = \vec 0$ in the lab — the lab already IS the CM frame — so $s = (2E)^2$ and $\sqrt{s} = 2E$ , with no square root eating the gain. Why: a collider does nothing clever with the energy. It simply refuses to pay the momentum tax, by arranging for there to be no momentum. *Verified in verify_aq-sr-centre-of-momentum-frame.py: $s$ is recomputed symbolically in both frames and shown equal; the fixed-target formula is derived from the mass shell; both limits are taken; and the Bevatron numbers printed in this article ( $E_1 = 7.138$ , $\sqrt{s} = 3.893$ , $V_{\mathrm{CM}} = 0.876c$ , $\gamma_{\mathrm{CM}} = 2.075$ ) are reproduced to the digits shown.*

Go deeper · Derivation C · 2 minSplitting the CM energy between two bodies

Derivation CSplitting the CM energy between two bodies

Once the CM frame is fixed, a two-body state has no freedom left but its orientation: the momenta are equal and opposite, and their common magnitude — along with both energies — is fixed entirely by $s$ and the two masses. Worth doing carefully, because the result is the workhorse of the next three articles, and because the standard form of it is where factors of $c$ go to die.

Step 1. Write the constraints. Energy: $E_1^{*} + E_2^{*} = \sqrt{s}$ . Momentum: $\vec p_1^{\;*} = -\vec p_2^{\;*}$ , so $|\vec p_1^{\;*}| = |\vec p_2^{\;*}| \equiv p^{*}$ . Why: two equations, two unknowns. The problem is already solved in principle; the rest is algebra.

Step 2. Subtract the mass shells. $E_1^{*2} - E_2^{*2} = \big(p^{*2}c^2 + m_1^2c^4\big) - \big(p^{*2}c^2 + m_2^2c^4\big) = (m_1^2 - m_2^2)c^4$ , since $p^{*}$ is common and cancels. Why: this is the trick — subtracting the mass shells eliminates the momentum entirely and leaves a pure statement about energies.

Step 3. Solve the linear system. Factor: $(E_1^{*} + E_2^{*})(E_1^{*} - E_2^{*}) = \sqrt{s}\,(E_1^{*} - E_2^{*})$ , so $E_1^{*} - E_2^{*} = (m_1^2 - m_2^2)c^4/\sqrt{s}$ . Add $E_1^{*} + E_2^{*} = \sqrt{s}$ and halve: $\boxed{E_1^{*} = \dfrac{s + (m_1^2 - m_2^2)c^4}{2\sqrt{s}}}$ , with $E_2^{*}$ following by swapping labels. Why: the difference of squares turns a quadratic system linear. Read the structure — each body gets half of $\sqrt{s}$ , plus or minus a term measuring which is heavier.

Step 4. Extract the common momentum. Insert $E_1^{*}$ into $p^{*2}c^2 = E_1^{*2} - m_1^2c^4$ and factor the resulting difference of squares twice. The Källén (triangle) function $\lambda(x,y,z) = x^2+y^2+z^2-2xy-2yz-2zx$ appears, and it factorises: $\lambda\big(s, m_1^2c^4, m_2^2c^4\big) = \big[s - (m_1+m_2)^2c^4\big]\big[s - (m_1-m_2)^2c^4\big]$ , giving $\boxed{p^{*} = \dfrac{\lambda^{1/2}\big(s,\,m_1^2c^4,\,m_2^2c^4\big)}{2c\sqrt{s}}}$ . Why: the factorised form beats the symmetric one because it makes the physics visible — $p^{*}$ vanishes exactly when $\sqrt{s} = (m_1+m_2)c^2$ , the threshold, and goes imaginary below it, which is how the algebra tells you the state is forbidden.

Step 5. Audit the units. $\lambda$ takes energy² arguments, so $\lambda$ is energy⁴ and $\lambda^{1/2}$ energy²; dividing by $2c\sqrt{s}$ (a velocity times an energy) leaves energy/velocity — a momentum. Put the $c$ in the numerator instead and $p^{*}$ comes out as velocity × energy, wrong by $c^2$ . Why: the AllQFT base writes exactly that, having defined $s$ in momentum² units in one card and used it in energy² units in the next four. Units are the cheapest error-detector you own. *Verified in verify_aq-sr-centre-of-momentum-frame.py: $E_1^{*} + E_2^{*} = \sqrt{s}$ and $E_1^{*2} - p^{*2}c^2 = m_1^2c^4$ hold symbolically; the Källén factorisation is proved by expansion; $p^{*}$ is dimensionally asserted to be an energy over $c$ ; and the equal-mass case gives $E^{*} = \sqrt{s}/2$ and the printed Bevatron value $p^{*} = 1.705$ GeV/ $c$ .*

Limiting cases

Case	Result	Meaning
$\vec P_{\mathrm{tot}} = \vec 0$ already	$\vec V_{\mathrm{CM}} = \vec 0$ , $\sqrt{s} = E_{\mathrm{tot}}$	the lab is the CM frame — what a collider buys
all particles at rest	$\sqrt{s} = \big(\sum_i m_i\big)c^2$	the only case where invariant mass is additive
$E_1 \gg mc^2$ (fixed target)	$\sqrt{s} \to \sqrt{2m_2c^2E_1}$	available energy $\propto \sqrt{E_{\mathrm{beam}}}$ — the root colliders exist to defeat
head-on equal beams	$\sqrt{s} = 2E_{\mathrm{beam}}$	available energy $\propto E_{\mathrm{beam}}$ — linear
$v \ll c$ throughout	$\vec V_{\mathrm{CM}} \to \sum m_i\vec v_i/\sum m_i$	Newton's centre-of-mass velocity, recovered because $E_i \to m_ic^2$
$s \to (m_1+m_2)^2c^4$	$p^{*} \to 0$	threshold: the two bodies are at rest relative to each other in the CM frame
two photons, back-to-back, $E_\gamma$ each	$s = 4E_\gamma^2 > 0$ , $M = 2E_\gamma/c^2$	a system of two massless particles has mass
two photons, collinear	$s = 0$ , $M = 0$	no CM frame exists — $P^\mu$ is null and no boost stands it upright

Common trapThe units of s: the single most expensive convention in this chapter

Two conventions circulate and both are correct. The norm $P\cdot P = (E_{\mathrm{tot}}/c)^2 - |\vec P_{\mathrm{tot}}|^2$ carries momentum², so there $s \equiv P\cdot P = (Mc)^2$ and $\sqrt{s}$ is a momentum. The universal high-energy convention — used here, behind every " $\sqrt{s} = 13\ \mathrm{TeV}$ " — carries energy²: $s \equiv (\sum E_i)^2 - |\sum\vec p_i|^2c^2 = (Mc^2)^2$ , and $\sqrt{s}$ is an energy. They differ by $c^2$ . The danger is not that one is wrong. It is that the algebra stays internally consistent in both, so a mixed calculation raises no error — it returns a number that is silently off by $c^2$ . The AllQFT base mixes them inside one list of cards: it defines $s = P\cdot P = (Mc)^2$ , then three cards later writes $\sqrt{s} = Mc^2$ and $\sqrt{s} = \sqrt{m_1^2c^4+\cdots}$ ; its $p^{*}$ card inherits the confusion and puts $c$ in the numerator, so the "momentum" comes out as velocity × energy. Lock the convention in line one and check that both sides of $\sqrt{s} \geq \sum_f m_fc^2$ are energies.

Worked examples

1 · The Bevatron: was 6.2 GeV enough?

A proton beam of kinetic energy $T = 6.2\ \mathrm{GeV}$ strikes protons at rest in a liquid-hydrogen target. Find the CM energy, the CM frame's velocity, and decide whether the reaction $p+p\to p+p+p+\bar p$ (which requires four protons' worth of rest mass in the final state) is possible. Take $m_pc^2 = 938.272\ \mathrm{MeV}$ .

Step 1 — identify. "Is this reaction possible" is a question about $\sqrt{s}$ and nothing else. The lab energies matter only as inputs to $s$ .

Step 2 — the beam's four-momentum. $E_1 = T + m_pc^2 = 6.200 + 0.938 = \mathbf{7.138\ GeV}$ ; from the mass shell, $p_1c = \sqrt{E_1^2 - m_p^2c^4} = \sqrt{50.955 - 0.880} = \mathbf{7.076\ GeV}$ .

Step 3 — apply the invariant. $s = 2m_p^2c^4 + 2m_pc^2E_1 = 0.880 + 0.880 + 2(0.938)(7.138) = \mathbf{15.155\ GeV^2}$ , so $\sqrt{s} = \mathbf{3.893\ GeV}$ . Cross-check from the totals: $E_{\mathrm{tot}} = 8.077\ \mathrm{GeV}$ , $|\vec P|c = 7.076\ \mathrm{GeV}$ , and $8.077^2 - 7.076^2 = 65.230 - 50.075 = 15.155\ \mathrm{GeV^2}$ . Same number, two routes — that is what "invariant" buys.

Step 4 — the frame itself. $V_{\mathrm{CM}}/c = 7.076/8.077 = \mathbf{0.876}$ and $\gamma_{\mathrm{CM}} = E_{\mathrm{tot}}/\sqrt{s} = 8.077/3.893 = \mathbf{2.075}$ , agreeing with $1/\sqrt{1-0.876^2}$ as it must. There each proton carries $E^{*} = \sqrt{s}/2 = 1.947\ \mathrm{GeV}$ and $p^{*}c = \sqrt{1.947^2 - 0.880} = 1.705\ \mathrm{GeV}$ , back to back.

Step 5 — the verdict. The final state needs $\sum_f m_fc^2 = 4m_pc^2 = \mathbf{3.753\ GeV}$ , and $\sqrt{s} = 3.893 > 3.753$ . Yes — by 140 MeV. The 6.2 GeV was not a round number; it was chosen to sit above this threshold with room for Fermi motion in nuclear targets.

Step 6 — the boundary number. Of the $6.200\ \mathrm{GeV}$ of kinetic energy poured in, the part that could become new rest mass is $\sqrt{s} - 2m_pc^2 = 3.893 - 1.877 = \mathbf{2.017\ GeV}$ , or 32.5 %. Nearly two-thirds is spent dragging the collision forward at $0.876c$ .

2 · What a fixed target would cost at LHC energies

The LHC collides protons head-on at $6.5\ \mathrm{TeV}$ per beam. What is $\sqrt{s}$ ? And what beam energy would a fixed-target machine need to match it?

Step 1 — the collider is easy. Head-on equal beams: $\vec P_{\mathrm{tot}} = \vec 0$ , the lab already IS the CM frame, and $\sqrt{s} = 2E_{\mathrm{beam}} = \mathbf{13\ TeV}$ . Every joule is spendable.

Step 2 — invert the fixed-target formula. From $s = m_1^2c^4 + m_2^2c^4 + 2m_2c^2E_1$ with $m_1 = m_2 = m_p$ : $E_1 = \big(s - 2m_p^2c^4\big)/(2m_pc^2)$ .

Step 3 — apply. $s = (13\ \mathrm{TeV})^2 = 1.69\times10^8\ \mathrm{GeV^2}$ and $2m_pc^2 = 1.8765\ \mathrm{GeV}$ ; the $2m_p^2c^4 = 1.76\ \mathrm{GeV^2}$ correction is negligible against $1.69\times10^8$ . So $E_1 = 1.69\times10^8/1.8765 = \mathbf{9.006\times10^7\ GeV}$ .

Step 4 — the boundary number. That is $\mathbf{9.0\times10^4\ TeV} = 90\ \mathrm{PeV}$ — about 6,900 times the LHC's per-beam energy, for the same physics. At LHC magnet technology (27 km for 6.5 TeV) such a ring would be of order $10^5\ \mathrm{km}$ around: several Earth diameters. This is why the collider was invented; Threshold Energy puts the whole curve on a plot.

Figure

Available energy versus beam energy for a proton on a stationary proton. The curve is √s = √(2 m_p c² E_beam + 2 m_p²c⁴) — a SQUARE ROOT. Quadrupling the beam doubles the reach. Marked: the Bevatron shot (7.14 GeV of beam buys 3.89 GeV of √s, just clearing the 3.753 GeV antiproton line) and the brutal high-energy points. A head-on collider instead gives √s = 2E_beam, a straight line off the top of this frame

Physics of the portrait

The frame has no discoverer, and that is the interesting thing about it. It arrives twice — in 1687 as a theorem about the centre of mass, and around 1908 as a statement about a four-vector — and the second arrival quietly destroys the first while keeping its answer.

Origins · 2 minCentre-of-momentum frame and how this was discovered

Centre-of-momentum frame

1687–1955 · Newton's Corollary IV · Minkowski's four-vector · the invariant s · the Bevatron's design energy

Newton's Principia (1687) already contains the essential move: Corollary IV to the Laws states that the common centre of gravity of a system of bodies does not alter its state of motion or rest by the mutual actions of those bodies. That is momentum conservation, dressed as a statement about a point, and it licenses the trick of solving a collision in the frame where that point stands still. It works because in Newtonian mechanics a point and a frame are interchangeable. Relativity severs them: once simultaneity is frame-dependent, the average position $\sum m_i\vec r_i/\sum m_i$ stops being observer-independent, because different frames evaluate the average at different sets of events and get genuinely different answers. What survives is not the centre of mass but the condition $\sum_i\vec p_i = \vec 0$ , which mentions no positions and needs no simultaneity convention. Minkowski's 1908 four-vector formalism made that natural: a system's total four-momentum is one geometric object, and asking for the frame where its spatial part vanishes is just asking to look along it. Planck, Lewis and Tolman, and later Møller worked out what a relativistic "centre" could even mean; the answer — that the frame is well defined while the point is not — is why CM is best read as centre of momentum. The idea earned its keep in 1954, when the Bevatron was built to a beam energy chosen by exactly this calculation, and in 1955 the antiproton appeared where $\sqrt{s}$ said it would.

History

1687

— Newton's Principia, Corollary IV: the common centre of gravity of an isolated system does not change its state of motion — momentum conservation, and the licence to work in the frame where it is at rest.

1908

— Minkowski's "Raum und Zeit" gives the four-vector formalism in which a system's total energy and momentum are one geometric object, and the CM condition becomes a statement about direction.

1943

— Rolf Widerøe patents the idea of colliding two beams head-on rather than firing one at a stationary target, precisely to avoid paying the momentum tax.

1954

— the Bevatron at Berkeley is commissioned with a design proton energy of 6.2 GeV kinetic, set by the requirement that √s clear 4·m_p·c² = 3.753 GeV.

1955

— Chamberlain, Segrè, Wiegand and Ypsilantis observe the antiproton at the Bevatron (Physical Review 100, 947); Chamberlain and Segrè share the 1959 Nobel Prize.

1971

— CERN's Intersecting Storage Rings collide proton on proton for the first time on 27 January, reaching √s = 62 GeV — a fixed-target machine would have needed a 2000 GeV beam.

Going further

Texts. Rindler, Introduction to Special Relativity, 2nd ed., §31 · Griffiths, Introduction to Elementary Particles, 2nd ed., §3.3–3.5 · Hagedorn, Relativistic Kinematics (1963), ch. 4 · Byckling & Kajantie, Particle Kinematics (1973), ch. 1–4 · Particle Data Group, Review of Particle Physics, "Kinematics" · Chamberlain, Segrè, Wiegand & Ypsilantis, "Observation of Antiprotons", Phys. Rev. 100, 947 (1955).

Provenance. Signature $(+,-,-,-)$ : $p\cdot p = +m^2c^2$ , $s > 0$ for a timelike total four-momentum, and $s$ in energy² units so that $\sqrt{s}$ is an energy (equation 3). The AllQFT base is internally inconsistent by a factor of $c^2$ and its $p^{*}$ card puts $c$ in the numerator; both are corrected above. verify_aq-sr-centre-of-momentum-frame.py parses this article and re-derives every number printed in it — $E_1 = 7.138$ , $p_1c = 7.076$ , $E_{\mathrm{tot}} = 8.077$ , $s = 15.155$ , $\sqrt{s} = 3.893$ , $V_{\mathrm{CM}} = 0.876c$ , $\gamma_{\mathrm{CM}} = 2.075$ , $E^{*} = 1.947$ , $p^{*}c = 1.705$ , $4m_pc^2 = 3.753$ , margin $140\ \mathrm{MeV}$ , $32.5\%$ , LHC equivalent $9.006\times10^7\ \mathrm{GeV}$ .

Scope. Thresholds are stated here and owned by Threshold Energy, which derives $\sqrt{s} \geq \sum_f m_fc^2$ and plots the asymmetry. Mandelstam $t$ and $u$ are not introduced; this article uses $s$ only. Invariant mass is owned by Invariant Mass, four-momentum conservation by Conservation of Four-Momentum.

Builds on. Relativistic Four-Momentum · Conservation of Four-Momentum · Invariant Mass · Relativistic Energy. Leads to. Relativistic Elastic Collisions · Relativistic Inelastic Collisions · Particle Decay Kinematics · Threshold Energy · Mandelstam Invariants.

Is the CM frame the frame in which the centre of mass is at rest?

Non-relativistically yes, and that is where the name comes from. Relativistically the two come apart and only one survives. The Newtonian centre of mass is a point, $\vec R = \sum m_i\vec r_i/\sum m_i$ , and computing it means evaluating all the positions at the same time — so it depends on a simultaneity convention, so different frames locate it at genuinely different places. The CM frame is defined by a condition on the total four-momentum alone, $\sum_i\vec p_i = \vec 0$ : no positions, no simultaneity. "CM" survives as an abbreviation but should be read as centre of momentum; when a particle-physics paper says "centre-of-mass frame", the momentum condition is what it means.

Why is √s and not the total lab energy the available energy?

Because momentum is conserved and therefore cannot be spent. If the incoming system carries $\vec P_{\mathrm{tot}}$ , whatever comes out must carry $\vec P_{\mathrm{tot}}$ too, so the products cannot all be at rest — collectively they must retain enough kinetic energy to supply that momentum, and that part cannot become rest mass. In the CM frame the constraint vanishes: the products may all be at rest, and the whole of $\sqrt{s}$ is convertible. Since $\sqrt{s}$ is invariant you may compute it in the lab and still trust it as the budget. The unavailable part is exactly $E_{\mathrm{tot}} - \sqrt{s} = \sqrt{s}\,(\gamma_{\mathrm{CM}} - 1)$ — the kinetic energy of the CM frame's own motion.

Can the invariant mass of a system be less than the sum of its parts' masses?

Yes — every bound system does it. A helium nucleus weighs less than two protons plus two neutrons, and the deficit is the binding energy. Nothing above forbids that: $\sqrt{s} = Mc^2$ is the total CM energy, and a bound system's negative potential energy is part of that total. What is forbidden is a system of free particles with $M < \sum_i m_i$ , since there $\sqrt{s} = \sum_i E_i^{\mathrm{CM}} \geq \sum_i m_ic^2$ , with equality only if every one is at rest. Binding Energy and Mass Defect develops the bound case; Relativistic Inelastic Collisions the free one, where $M$ strictly exceeds the sum.

The formula V = c²P/E looks like it could exceed c. Can it?

Not for any physical system. $|\vec V_{\mathrm{CM}}|/c = |\vec P_{\mathrm{tot}}|c/E_{\mathrm{tot}}$ , and since every particle obeys $E_i \geq |\vec p_i|c$ (equality only if massless), the triangle inequality gives $E_{\mathrm{tot}} = \sum E_i \geq \sum|\vec p_i|c \geq |\sum\vec p_i|c = |\vec P_{\mathrm{tot}}|c$ . So $|\vec V_{\mathrm{CM}}| \leq c$ always, with equality only in the degenerate case where every particle is massless and they all point the same way — which is exactly the one configuration with no CM frame. The formula is safe precisely because nothing with mass in it can saturate the bound.

If s is conserved AND invariant, are those the same statement?

No, and keeping them apart matters. Invariance means all observers of the same state agree on $s$ — it is a Lorentz scalar. Conservation means the initial and final states have the same $s$ — it follows from $\sum p^\mu$ being conserved. The two are independent: a quantity can be invariant but not conserved (a particle's rest mass, in an inelastic collision), or conserved but not invariant (the total energy, conserved in every frame but with a different value in each). $s$ is both — which is what lets you evaluate it in the lab before the collision and spend it as the CM budget after.

Why is the fixed-target √s formula only linear in the beam energy under the root?

Because of a cancellation worth seeing. Expanding $s = (E_1 + m_2c^2)^2 - |\vec p_1|^2c^2$ produces an $E_1^2$ from the square — and the mass shell, $|\vec p_1|^2c^2 = E_1^2 - m_1^2c^4$ , produces a $-E_1^2$ . They cancel exactly, leaving only the cross term $2m_2c^2E_1$ . The cancellation is not an accident: at high energy the beam four-momentum is nearly null, and a null vector contributes nothing to a norm by itself. All of $s$ comes from the cross term between beam and target — which is why $s$ is proportional to the target mass, and why a heavier target buys more $\sqrt{s}$ only as $\sqrt{m_2}$ .

Centre-of-Momentum Frame

Where this lives today

The physical picture

The core equations

Deriving the frame

Limiting cases

Worked examples

1 · The Bevatron: was 6.2 GeV enough?

2 · What a fixed target would cost at LHC energies

Physics of the portrait

Going further

Before this

Where this leads