Center-of-Momentum Frame and Invariant Mass

Take two photons. Neither has a mass, neither has a rest frame, and neither can be caught — chase one at $0.999999c$ and it still leaves you at $c$ . Now let them fly apart at $120^\circ$ , one carrying $3.00$ GeV and the other $1.00$ GeV, and ask what the pair weighs. The answer is $3.000$ GeV/ $c^2$ , and the pair has a rest frame: an observer moving at $0.66144c$ sees the two photons flying exactly opposite, with equal and opposite momenta, and measures a total energy of $3.000$ GeV. A system of things that cannot be at rest is, itself, at rest in some frame. That is not a paradox — it is the whole point, and this article is about what makes it work. The usual account says the rest frame is "the frame where $\sum_i \vec p_i = 0$ ", which is true and slightly beside the point, because it makes the rest frame sound like a bookkeeping convenience. It is not. The rest frame is the frame in which the system's total energy is a minimum — a strict, non-degenerate minimum, whose value is $Mc^2$ and whose curvature is $M$ itself. And $M$ , together with the sign of the energy, is the complete invariant of the total four-momentum: two systems are related by a Lorentz transformation if and only if these two numbers agree. Everything else you can say about $P^\mu$ — its energy, its momentum, its direction — is a choice of frame.

Key takeaways

Convention lock. $(+,-,-,-)$ , $\eta = \mathrm{diag}(+1,-1,-1,-1)$ . Four-momentum in energy units: $P^\mu = (E,\ \vec p\,c)$ , so $P\cdot P = E^2 - |\vec p\,|^2c^2 = m^2c^4 > 0$ for a massive particle. Timelike is POSITIVE.
Recalled, not re-derived. `aq-sr-invariant-mass` owns $M$ itself and the massless-pair result. `aq-sr-centre-of-momentum-frame` owns the existence of the frame, $\vec V_{\rm CM} = c^2\vec P/E_{\rm tot}$ , and the collinear-photon exception. `aq-sr-relativistic-inelastic-collisions` owns $M \geq \sum_i m_i$ , equality iff every constituent comoves.
Owned result 1 — the rest frame is a variational object. For an observer who sees the system's centre of momentum moving at speed $u$ , the measured total energy is exactly $E(u) = \gamma(u)\,Mc^2$ . Hence $\boxed{Mc^2 = \min_{\text{all frames}} E_{\rm tot}}$ , the minimum is strict (any $u \neq 0$ costs energy), and $\left.\mathrm{d}^2E/\mathrm{d}u^2\right|_{0} = M$ : the mass is the curvature of the minimum.
Owned result 2 — $M$ is a complete invariant. Two four-momenta $P$ , $P'$ are related by a proper orthochronous Lorentz transformation iff $P\cdot P = P'\cdot P'$ and $\mathrm{sign}(E) = \mathrm{sign}(E')$ . Proof is constructive: the boost $\vec\beta = \vec P c/E$ carries any timelike $P$ to the standard vector $(Mc^2,\ \vec 0)$ .
So "rest mass" is not one property among many. It is the only frame-independent thing $P^\mu$ has. The Lorentz group acts transitively on each sheet of the mass hyperboloid, and $M$ labels the sheet.
The system does not remember its parts. A composite of mass $M$ accelerates, carries momentum and radiates exactly like a point particle of mass $M$ — even when $M$ is a hundred times the sum of its constituents' masses, as it is for the proton.
Key idea: a system has a rest frame not because its parts can be stopped, but because its total four-momentum is timelike. Timelike means there is a frame of least energy, that least energy is $Mc^2$ , and no other number survives the change of observer.

The core insight

Ask what a boost can and cannot do to the total four-momentum $P^\mu$ . It can change $E$ ; it can change $\vec P$ ; it can rotate, reverse and rescale them. What it cannot do is change $P\cdot P$ , and — provided $P$ is timelike — it cannot change the sign of $E$ either. Those two numbers exhaust the invariant content. The consequence is sharper than it looks: since the boost $\vec\beta = \vec Pc/E$ is subluminal precisely when $P$ is timelike, every timelike four-momentum can be carried to $(Mc^2, \vec 0\,)$ , so any two with the same $M$ and the same energy sign can be carried into each other. The set of physically distinguishable total four-momenta is therefore just the positive real line, parametrised by $M$ . A system's rest mass is not a property it has. It is the only property it has.

Where this lives today

Invariant mass is the single most-plotted quantity in experimental particle physics. Every resonance ever found — the $J/\psi$ , the $Z$ , the Higgs at $125$ GeV — was found as a bump in a histogram of $M$ , reconstructed by summing the four-momenta of the decay products and taking the norm, exactly as in equation (2). The reason the method works is the completeness theorem: $M$ is the one number that every detector, every beam energy and every boost agrees about, so a peak in $M$ is a fact about nature rather than about the apparatus. The variational reading is not decoration either. In lattice QCD one computes a hadron's mass by finding the lowest-energy state in a given quantum-number sector at zero total momentum — which is to say, by locating exactly the minimum of the first derivation. And the completeness statement is why "relativistic mass" was retired: once you know that $(M, \mathrm{sign}\,E)$ exhausts the invariant content of a four-momentum, a velocity-dependent "mass" is visibly a coordinate wearing a physicist's coat. The one genuinely open direction is the origin of $M$ itself. Relativity tells you that a bound system's mass is the invariant norm of its total four-momentum; it does not tell you why that norm comes out at $938$ MeV. That question belongs to QCD, and the 2008 lattice calculations answered it numerically before anyone answered it in closed form.

The physical picture

The centre-of-momentum frame usually enters as a trick: "boost until the momenta cancel, because the algebra is easier there." That is true, it is useful, and it hides the actual theorem. Here is the theorem. Fix a system — a colliding pair, a decaying nucleus, a box of photons, anything with timelike total four-momentum. Every inertial observer measures some total energy $E_{\rm tot}$ , and they all get different answers. Plot those answers against the observer's velocity, and you do not get a monotonic ramp or a family of unrelated numbers. You get a hyperbolic cosine with a single strict minimum, and the value at that minimum is $Mc^2$ . So the CM frame is the frame the system is cheapest in. That is the sense in which it deserves the name "rest frame": not that some coordinate happens to vanish there, but that it is the frame of least energy, the way a ball's rest position is the point of least potential. And the analogy runs deeper than it should be allowed to. Expand about the minimum, and the quadratic coefficient is the mass:

E(u) = \gamma(u)\,Mc^2 = Mc^2 + \tfrac12 M u^2 + \tfrac38 \frac{M u^4}{c^2} + \cdots

The leading correction is exactly Newton's kinetic energy — with $M$ , the system's mass, not the sum of its parts. The system moves like a point particle of mass $M$ , and the internal structure is invisible to the outside. For the proton that is a factor of a hundred: $938.272$ MeV of inertia built out of quarks whose masses sum to about $9$ MeV. The second half of the article asks the obvious follow-up. If $M$ survives every boost, what else does? Nothing. Once you fix $M$ and the sign of the energy, the Lorentz group can take you anywhere on the mass shell and nowhere off it — so $M$ is not merely an invariant, it is the complete invariant, and every other number attached to $P^\mu$ is a statement about the observer rather than the system.

The core equations

The mass shell, re-derived from $\eta$ , not transcribed. With $\eta = \mathrm{diag}(+1,-1,-1,-1)$ and $P^\mu = (E,\ \vec p\,c)$ in energy units:

P\cdot P \;=\; \eta_{\mu\nu}P^\mu P^\nu \;=\; E^2 - |\vec p\,|^2c^2 \;=\; m^2c^4 \qquad (1)

where: timelike is positive. Physical reading: this is the definition of mass, not a consequence of it. Add four-momenta first, take the norm second — never the other way round. The total four-momentum and the invariant mass of a system:

P^\mu_{\rm tot} = \sum_i P_i^\mu, \qquad M^2c^4 \;=\; P_{\rm tot}\cdot P_{\rm tot} \;=\; \Big(\sum_i E_i\Big)^2 - \Big|\sum_i \vec p_i\Big|^2c^2 \qquad (2)

where: $M$ is the system's rest mass. Physical reading: $M \neq \sum_i m_i$ in general — `aq-sr-relativistic-inelastic-collisions` proves $M \geq \sum_i m_i$ , with equality only when every constituent is comoving. The boost to the rest frame, and the condition for it to exist:

\vec\beta_{\rm CM} = \frac{\vec P_{\rm tot}\,c}{E_{\rm tot}}, \qquad |\vec\beta_{\rm CM}| < 1 \iff E_{\rm tot} > |\vec P_{\rm tot}|c \iff P_{\rm tot}\cdot P_{\rm tot} > 0 \qquad (3)

where: the last equivalence is equation (1) read backwards. Physical reading: the rest frame exists exactly when the total four-momentum is timelike — which is not automatic, and fails for collinear massless particles. Owned and proved by `aq-sr-centre-of-momentum-frame`. Now the owned result. Let $u$ be the speed at which an observer sees the system's centre of momentum moving. Then:

\boxed{\;E_{\rm tot}(u) \;=\; \gamma(u)\,Mc^2, \qquad\text{so}\qquad Mc^2 \;=\; \min_{|u|<c} E_{\rm tot}(u), \qquad \left.\frac{\mathrm{d}^2 E}{\mathrm{d}u^2}\right|_{u=0} \!\!= M\;} \qquad (4)

where: the minimum is attained only at $u = 0$ , i.e. in the CM frame. Physical reading: the rest frame is the frame of least energy, and the mass is the curvature of that minimum. Since $\gamma \geq 1$ with equality only at $u=0$ , the minimum is strict; since $\mathrm{d}^2E/\mathrm{d}u^2|_0 = M > 0$ , it is non-degenerate. The system's inertia — the $M$ in $\tfrac12 M u^2$ — is its invariant mass. And the completeness statement:

\boxed{\;P' = \Lambda P \ \text{ for some } \Lambda \in L^{\uparrow}_{+} \iff P\cdot P = P'\cdot P' \ \text{ and } \ \mathrm{sign}(E) = \mathrm{sign}(E')\;} \qquad (5)

where: $L^{\uparrow}_{+}$ is the proper orthochronous Lorentz group. Physical reading: $(M, \mathrm{sign}\,E)$ is a complete set of invariants for a four-momentum. Two systems with the same rest mass are, as far as relativity is concerned, the same object seen from two chairs.

Figure

THE REST FRAME IS THE FRAME OF LEAST ENERGY. Total energy of the two-photon system (3.00 GeV at 0 degrees plus 1.00 GeV at 120 degrees) as measured by an observer who sees the system's centre of momentum moving at speed u. The curve is E = gamma(u) Mc-squared — a hyperbolic cosine in rapidity — and it has ONE strict minimum, at u = 0, where E = Mc-squared = 3.000 GeV EXACTLY. The LAB is not that frame: in the lab the CM is already moving at 0.66144c, so the lab sees 4.000 GeV. Boost the other way and it gets worse, not better: at u = 0.9c the observer sees 6.882 GeV. There is no frame anywhere that sees less than 3.000 GeV. Expand about the minimum and the quadratic coefficient is M itself — the curvature of the well IS the mass, and the leading term is Newton's one-half M u-squared

Two theorems about a timelike vector

Derivation AThe rest frame is the frame of least energy, and the mass is the curvature of the minimum

The centre-of-momentum frame is normally defined by a coordinate condition — the frame where $\sum_i \vec p_i = 0$ — and a definition is not a theorem. This derivation replaces it with a variational characterisation: the CM frame is the unique frame in which the system's total energy is stationary, that stationary point is a strict minimum, its value is $Mc^2$ , and its curvature is $M$ . The rest frame stops being a convenient choice of coordinates and becomes the bottom of a well.

Step 1. Take a system with timelike $P_{\rm tot}$ , so that by equation (3) the CM frame exists. In that frame, by construction, $P_{\rm tot} = (Mc^2,\ \vec 0\,)$ — energy $Mc^2$ , no momentum. Why: this is not an assumption. `aq-sr-centre-of-momentum-frame` proves the boost $\vec\beta_{\rm CM} = \vec Pc/E$ is subluminal exactly when $P$ is timelike, so the frame is reachable.

Step 2. Now let an observer see that CM moving at velocity $\vec u$ . A boost acts on $P_{\rm tot}$ exactly as it acts on any four-vector, so the energy component transforms as

E_{\rm tot}(u) \;=\; \gamma(u)\Big(Mc^2 + \vec u\cdot\vec 0\Big) \;=\; \gamma(u)\,Mc^2 .

Why: every internal detail has vanished. The system's constituents, their masses, their internal motions — none of it appears. Whatever the system is made of, its total four-momentum boosts like the four-momentum of a single particle of mass $M$ .

Step 3. Minimise. $\gamma(u) = (1 - u^2/c^2)^{-1/2}$ is $\geq 1$ for every $|u| < c$ , with equality only at $u = 0$ . Therefore

\boxed{\;Mc^2 \;=\; \min_{|u|<c} E_{\rm tot}(u), \qquad\text{attained uniquely in the CM frame.}\;}

Why: the minimum is strict, not merely attained. Any observer in relative motion — in either direction, along any axis — measures strictly more energy. There is no flat direction and no competing minimum, because $\gamma$ has neither.

Step 4. Expand and read off the curvature. For small $u$ ,

E_{\rm tot}(u) = Mc^2\Big(1 + \frac{u^2}{2c^2} + \frac{3u^4}{8c^4} + \cdots\Big) = Mc^2 + \tfrac12 M u^2 + \tfrac38\frac{Mu^4}{c^2} + \cdots \qquad\Longrightarrow\qquad \left.\frac{\mathrm{d}^2E}{\mathrm{d}u^2}\right|_{0} = M .

Why: the mass is the curvature of the energy minimum, and that is the whole meaning of inertia. The coefficient of $\tfrac12 u^2$ is what Newton called mass, and here it comes out as $M$ — the system's invariant mass, not $\sum_i m_i$ . A proton resists acceleration with $938$ MeV/ $c^2$ of inertia, though its quarks bring only about $9$ . *Verified in verify_aq-sr-center-of-momentum-invariant-mass.py: $E(u) = \gamma(u)Mc^2$ is checked symbolically and, independently, by explicitly boosting the two-photon system and composing velocities relativistically — two routes, same number; the minimum is confirmed by differentiation ( $\mathrm{d}E/\mathrm{d}u = 0$ only at $u=0$ ) and numerically on a scan; and $\mathrm{d}^2E/\mathrm{d}u^2|_0 = M$ is proved.*

Go deeper · Derivation B · 2 minThe invariant mass is a complete invariant: the Lorentz group is transitive on each mass shell

Derivation BThe invariant mass is a complete invariant: the Lorentz group is transitive on each mass shell

It is a commonplace that $M$ is invariant. The stronger and more useful statement is that it is complete: nothing else about a four-momentum survives a change of observer, so two systems with the same rest mass and the same energy sign are Lorentz-equivalent, full stop. The proof is short and entirely constructive — we exhibit the transformation — and the construction is the same subluminal boost that built the CM frame.

Step 1 — necessity. Lorentz transformations preserve the metric, so $P\cdot P$ is unchanged; and the orthochronous ones preserve the sign of the energy of a timelike vector (proved in `aq-sr-reinterpretation-principle-causal-paradoxes`: for timelike $P$ , $E > |\vec p\,|c$ , so $E' = \gamma(E - \vec\beta\cdot\vec p\,c)$ can never reach zero). Why: this half is the easy half, and it says the two numbers are at most what survives.

Step 2 — sufficiency, constructively. Let $P$ be timelike with $E > 0$ and $P\cdot P = M^2c^4$ . Apply the boost with $\vec\beta = \vec P c/E$ . By equation (3) this is subluminal, and it carries $P$ to

P \;\longmapsto\; \big(\,\sqrt{P\cdot P},\ \vec 0\,\big) \;=\; \big(Mc^2,\ \vec 0\,\big) \;\equiv\; \hat P_M ,

the standard vector of mass $M$ . Why: the same boost that defines the CM frame is the one that proves completeness. Any timelike, future-pointing four-momentum can be brought to the standard vector; the construction never fails, because $|\vec\beta| < 1$ is exactly the timelike condition.

Step 3 — compose. If $P$ and $P'$ have the same $M$ and the same energy sign, then $\Lambda_P\,P = \hat P_M = \Lambda_{P'}P'$ , so $P' = \Lambda_{P'}^{-1}\Lambda_P\,P$ , and the composite is again a proper orthochronous Lorentz transformation. Hence

\boxed{\;P \sim P' \iff P\cdot P = P'\cdot P' \ \text{ and } \ \mathrm{sign}(E)=\mathrm{sign}(E')\;}

Why: the Lorentz group acts transitively on each sheet of the mass hyperboloid. The sheet — one number, $M$ , plus a sign — is the entire orbit. There is no finer invariant to find, because there is nothing left to distinguish.

Step 4 — the corollary that matters. Every other number you can write down for $P^\mu$ — the energy, the momentum, the speed, the direction, the rapidity — is a frame label, not a property of the system. Two protons, one sitting on a bench and one circulating in the LHC at $6500$ GeV, carry four-momenta that differ by a boost of $\gamma = 6927.6$ and by nothing else. Why: this is why "relativistic mass" is a dead concept. The quantity $\gamma m$ is the time component of a four-vector in a particular frame. It is a coordinate, and coordinates are not properties. The mass is $M$ , and it is the same on the bench and in the ring. *Verified in verify_aq-sr-center-of-momentum-invariant-mass.py: the boost $\vec\beta = \vec Pc/E$ is applied symbolically to a general timelike $P$ and shown to produce $(\sqrt{P\cdot P},\vec 0)$ ; the composed transformation is checked to preserve $\eta$ ; $|\vec\beta| < 1 \iff P\cdot P > 0$ is proved; and the explicit LHC boost ( $\gamma = 6927.63$ , $1-\beta = 1.0418\times10^{-8}$ ) is computed to 50 digits.*

Figure

ONE NUMBER LABELS THE ORBIT. The mass shell E-squared minus p-squared c-squared equals M-squared c-to-the-fourth, drawn in the (pc, E) plane. It has TWO DISCONNECTED SHEETS. A Lorentz boost slides a four-momentum ALONG a sheet and can never leave it — so the sheet, labelled by M and by the sign of E, IS the orbit of the Lorentz group. The proton on a laboratory bench sits at the vertex, (938.272 MeV, 0). The proton circulating in the LHC at 6500 GeV sits far out on the SAME curve. They are the same object: a single boost of gamma = 6927.63 takes one to the other, and nothing else distinguishes them. That is what it means for M to be a COMPLETE invariant — every other number on the diagram, energy and momentum included, is a label for the observer, not a property of the proton

Limiting cases

Case	Result	Meaning
$P_{\rm tot}$ timelike	rest frame exists, $Mc^2 = \min E_{\rm tot}$	the generic case for anything with mass, and for any non-collinear photons
$P_{\rm tot}$ null	$M = 0$ , no rest frame	collinear massless particles — $\vec\beta_{\rm CM} \to c$ . Owned by `aq-sr-centre-of-momentum-frame`
all constituents comoving	$M = \sum_i m_i$	the only case where mass adds. Owned by `aq-sr-relativistic-inelastic-collisions`
slow internal motion	$M \approx \sum_i m_i + (\sum_i T_i^{*} + U)/c^2$	binding energy is negative mass; kinetic energy is positive mass
$u \to 0$	$E \to Mc^2 + \tfrac12 Mu^2$	the Newtonian limit of the well — the curvature is the inertia
$u \to c$	$E \to \infty$	boosting away from the rest frame is unboundedly expensive in both directions
two systems, same $M$ , same $\mathrm{sign}\,E$	Lorentz-equivalent	nothing else about $P^\mu$ is real

Common trapThree ways to say something false about the rest frame

"The rest frame is where the particles are at rest." No — it is where their total momentum vanishes. The constituents are typically flying about at nearly $c$ inside it, and for a two-photon system they are moving at exactly $c$ , in opposite directions. The system is at rest; nothing in it is.

"A massless system has no rest frame." Only if it is collinear. Two photons at any angle other than $0^\circ$ have $M > 0$ and a perfectly good rest frame — the opening example makes $3.000$ GeV/ $c^2$ out of two objects with none. The exception is real but narrow, and it is a statement about $P_{\rm tot}$ being null, not about the constituents being massless.

"Relativistic mass $\gamma m$ is the mass of a fast particle." It is the energy of a fast particle, divided by $c^2$ , and it is a frame label. Equation (5) is the reason: $M$ and $\mathrm{sign}(E)$ exhaust the invariant content of $P^\mu$ , so a quantity that changes when you change chairs is not a property of the system. The proton in the LHC ring has the same mass as the proton on the bench — $938.272$ MeV/ $c^2$ — and a different energy.

Worked examples

1 · Two massless things with a rest frame, and the frame of least energy

Photon 1 carries $E_1 = 3.00$ GeV along $+x$ . Photon 2 carries $E_2 = 1.00$ GeV at $120^\circ$ to it. Find the system's mass, its rest frame, and prove that no observer sees less energy than $Mc^2$ .

Step 1 — identify. Add four-momenta first, norm second (equation 2). For a photon $E = |\vec p\,|c$ , so

\vec P_{\rm tot}\,c = \big(3.00 + 1.00\cos 120^\circ,\ \ 1.00\sin 120^\circ\big) = (\mathbf{2.500},\ \mathbf{0.8660})\ \mathrm{GeV}, \qquad E_{\rm tot} = \mathbf{4.000}\ \mathrm{GeV}.

Step 2 — the mass. $|\vec P_{\rm tot}|c = \sqrt{2.500^2 + 0.8660^2} = \sqrt{7.000} = \mathbf{2.6458}$ GeV, so

M^2c^4 = E_{\rm tot}^2 - |\vec P_{\rm tot}|^2c^2 = 16.000 - 7.000 = 9.000\ \mathrm{GeV^2} \qquad\Longrightarrow\qquad Mc^2 = \mathbf{3.000\ GeV}.

Cross-check against `aq-sr-invariant-mass`, which owns the two-photon formula: $M^2c^4 = 2E_1E_2(1 - \cos\theta) = 2(3)(1)(1.5) = 9.000$ GeV². Two massless particles, a $3$ GeV mass.

Step 3 — the rest frame. Equation (3): $\beta_{\rm CM} = |\vec P|c/E_{\rm tot} = 2.6458/4.000 = \mathbf{0.66144}$ , along $\hat P$ . Then $\gamma_{\rm CM} = E_{\rm tot}/Mc^2 = 4.000/3.000 = \mathbf{1.3333}$ , and independently $1/\sqrt{1-0.66144^2} = 1.3333$ ✓. In that frame the two photons fly exactly back-to-back with $1.500$ GeV each.

Step 4 — the boundary number. Now scan every observer. The energy they measure is $E(u) = \gamma(u)\,Mc^2$ where $u$ is the speed at which they see the CM move:

observer	sees the CM at	measures $E_{\rm tot}$
the CM frame	$u = 0$	$\mathbf{3.000}$ GeV ← the minimum
the lab	$u = 0.66144c$	$\mathbf{4.000}$ GeV
moving at $0.3c$ along $\hat P$	$u = 0.4509c$	$\mathbf{3.3611}$ GeV
moving at $0.9c$ along $\hat P$	$u = 0.5895c$	$\mathbf{3.7138}$ GeV

The third and fourth rows are computed twice — once by boosting the lab four-momentum directly, once by composing velocities to get $u$ and using $E = \gamma(u)Mc^2$ — and they agree. Nothing anywhere sees less than $3.000$ GeV, and the lab's $4.000$ GeV is exactly $\gamma_{\rm CM}$ times it.

Step 5 — the moral. The lab is not privileged and neither is any other frame — except one. The rest frame is the bottom of the energy well, and $Mc^2$ is the depth of the well. Everything else is a choice of chair.

2 · The proton: a hundredfold mass out of almost nothing, and the boost that proves it is one object

The proton weighs $938.272$ MeV/ $c^2$ . Its three valence quarks weigh, in the $\overline{\rm MS}$ scheme at $2$ GeV, about $2.16 + 2.16 + 4.67 = 8.99$ MeV/ $c^2$ between them. Check the mass inequality, then show that the LHC proton and the bench proton are the same object.

Step 1 — identify. `aq-sr-relativistic-inelastic-collisions` owns $M \geq \sum_i m_i$ , with equality iff all constituents comove. Recall it; do not re-derive it.

Step 2 — apply.

\frac{M}{\sum_i m_i} = \frac{938.272}{8.99} = \mathbf{104.4}.

The inequality is satisfied — with a factor of a hundred to spare. About $99\%$ of the proton's mass is the invariant mass of a violently moving, confined system, not the mass of its ingredients. Quark masses are scheme- and scale-dependent, so the exact ratio moves with the convention; the order of magnitude does not.

Step 3 — and yet it moves like a point particle. By equation (4), the proton's response to a boost is $E = \gamma Mc^2$ — with $M = 938.272$ MeV/ $c^2$ , the whole thing, and no trace of the $8.99$ MeV. The internal structure contributes exactly nothing to how the system boosts. That is the content of Step 2 of the first derivation, and it is why particle physics can treat the proton as a single four-vector.

Step 4 — the boundary number. An LHC beam proton has $E = 6500$ GeV. Then $\gamma = E/(Mc^2) = 6500/0.93827208816 = \mathbf{6927.63}$ , and

|\vec p\,|c = \sqrt{E^2 - M^2c^4} = \mathbf{6499.9999323\ GeV}, \qquad 1 - \beta = \mathbf{1.0418\times10^{-8}} .

By equation (5), that single boost of $\gamma = 6927.63$ is the entire difference between the LHC proton and the one on the bench. Same $M$ , same $\mathrm{sign}(E)$ , therefore the same orbit, therefore one Lorentz transformation apart — and no further invariant distinguishes them.

Step 5 — the moral. Ask what is real about the LHC proton's $6500$ GeV and the answer is: nothing intrinsic. It is a coordinate on the mass hyperboloid. The proton's mass, $938.272$ MeV/ $c^2$ , is real, and it is the only thing about $P^\mu$ that is.

Physics of the portrait

Origins · 2 minRest frame and how this was discovered

Rest frame

1905–1939 · Einstein's three pages · Minkowski's hyperboloid · and the frame that has to be found rather than chosen

The rest frame arrives in physics three times, and each time it means something larger. The first is Albert Einstein, in three pages of Annalen der Physik 18, 639 (received 27 September 1905, published 21 November), under the title "Does the inertia of a body depend upon its energy content?" The argument is a body emitting two equal light pulses in opposite directions, computed in two frames; the mismatch in the energy bookkeeping is a mismatch in inertia, and Einstein concludes that the body's mass has fallen by $E/c^2$ . It is a paper about the rest frame: the whole point is that a quantity you measure there — the rest energy — is what the rest of physics has to agree about. He did not write $E = mc^2$ ; he wrote $L/c^2$ , and he called it $m$ afterwards. The second is Hermann Minkowski, on 21 September 1908, at the 80th Assembly of German Natural Scientists and Physicians in Cologne, opening with the line everyone quotes and few finish: space by itself and time by itself are doomed to fade into shadows. What "Raum und Zeit" gave the rest frame was a geometry. The mass shell became a surface, the boost became a motion along it, and the rest frame became the point where that surface meets the energy axis — a vertex, not a convention. Minkowski was dead of appendicitis within four months, aged 44. The third is Eugene Wigner, in 1939, who turned the surface into an orbit. Once you see the mass hyperboloid as the set the Lorentz group sweeps out — transitively, missing nothing — the rest frame is no longer somewhere you choose to go. It is the distinguished point of an orbit, and the orbit is labelled by exactly one number. That is the theorem this article proves in its elementary form, and `aq-sr-wigner-rotation` is where the group theory begins in earnest.

History

1905

— Einstein, in Annalen der Physik 18, 639, argues that a body radiating energy E loses mass E/c-squared. The paper is three pages long and is received on 27 September.

1908

— Minkowski delivers "Raum und Zeit" in Cologne on 21 September, at the 80th Assembly of German Natural Scientists and Physicians. The mass shell becomes a geometric surface.

1909

— Minkowski dies of appendicitis in Gottingen, aged 44. "Raum und Zeit" is printed posthumously.

1939

— Wigner classifies the unitary irreducible representations of the inhomogeneous Lorentz group. The mass hyperboloid is recognised as an orbit, labelled by mass and by the sign of the energy.

1955

— the Bevatron reaches the antiproton threshold, and invariant mass becomes an everyday experimental tool rather than a theoretical one.

2008

— Duerr and collaborators compute the light-hadron spectrum from lattice QCD (Science 322, 1224), confirming from first principles that the proton's mass is overwhelmingly not the mass of its quarks.

Going further

Texts. Rindler, Introduction to Special Relativity, 2nd ed., §5–6 · Weinberg, The Quantum Theory of Fields I, §2.5 (the mass hyperboloid as an orbit; the standard vector $\hat P_M$ ) · Griffiths, Introduction to Elementary Particles, §3. Primary: A. Einstein, "Ist die Trägheit eines Körpers von seinem Energieinhalt abhängig?", Annalen der Physik 18, 639 (1905); H. Minkowski, "Raum und Zeit", Cologne, 21 September 1908; E. P. Wigner, Ann. Math. 40, 149 (1939); S. Dürr et al., Science 322, 1224 (2008).

Provenance. Signature $(+,-,-,-)$ , $P^\mu = (E,\vec p\,c)$ , $P\cdot P = +m^2c^4$ ; $s$ in energy². De-overlap, and a refusal. The lane spec asked this article to own $M \geq \sum_i m_i$ . It cannot: `aq-sr-relativistic-inelastic-collisions` already ships that theorem, with the comoving-equality condition and the $M = \sum m_i + T^{*}_{\rm tot}/c^2$ refinement. Under the earlier-serial-owns rule it is recalled here by slug and not re-derived, and this article owns instead the variational characterisation of the rest frame (equation 4) and the completeness of $M$ (equation 5), neither of which appears anywhere in the shipped corpus. Likewise the base article's DERIV[0] and DERIV[1] (fixed-target vs collider $\sqrt{s}$ scaling) are serial 1984's payload and are not developed here, and its DERIV[2] (threshold) is owned by `aq-sr-threshold-energy`. The base is otherwise sound: correct signature, correct energy²-units $s$ . All numbers above are asserted in verify_aq-sr-center-of-momentum-invariant-mass.py.

Builds on. `aq-sr-invariant-mass` · `aq-sr-centre-of-momentum-frame` · `aq-sr-relativistic-inelastic-collisions` · `aq-sr-relativistic-collision-kinematics`. Leads to. `aq-sr-mandelstam-variables` · `aq-sr-wigner-rotation`.

How can a system of massless particles have a mass?

Because mass is the norm of the total four-momentum, and norms in Minkowski space do not add. Each photon has $P_i \cdot P_i = 0$ , but the cross term does not vanish: $P_{\rm tot}\cdot P_{\rm tot} = 2P_1\cdot P_2 = 2E_1E_2(1-\cos\theta)$ , which is strictly positive unless the photons are collinear. Add the four-momenta first, take the norm second. The two-photon example in this article makes $3.000$ GeV/ $c^2$ out of $3$ GeV and $1$ GeV of pure light, and it is not a trick — that pair has a rest frame, a rest energy, and the inertia to match.

Is the rest frame really a minimum, or just a stationary point?

A strict minimum, and non-degenerate. $E(u) = \gamma(u)Mc^2$ , and $\gamma(u) \geq 1$ with equality only at $u = 0$ — so every other observer, in every direction, measures strictly more. There is no flat direction and no second minimum. The curvature is $\mathrm{d}^2E/\mathrm{d}u^2|_0 = M > 0$ , which is both the confirmation that it is a minimum and the statement that the mass is the system's inertia.

If M is a complete invariant, what happened to spin?

Spin is an invariant of the particle, not of its four-momentum. This article's completeness theorem is about $P^\mu$ alone: the orbit of a four-vector under the Lorentz group is labelled by $P\cdot P$ and $\mathrm{sign}(E)$ , and that is all. A particle is more than its four-momentum — it also carries a state in the representation of the subgroup that fixes $P^\mu$ , and that is where spin lives. The classification is Wigner's; `aq-sr-wigner-rotation` is the entry point.

Why is "relativistic mass" wrong if it gives the right answers?

It gives the right answers because it is $E/c^2$ , and $E$ is a real measurable quantity — just not an invariant one. The objection is not to the arithmetic but to the word mass, which in every other context means a property of the system rather than of the observer. Equation (5) makes the objection precise: the invariant content of $P^\mu$ is exhausted by $M$ and the sign of $E$ , so anything frame-dependent is a coordinate. Calling a coordinate a mass invites exactly the error of thinking a fast proton is somehow heavier — it is not; it is the same proton, seen from a chair that is moving.

Does the minimum-energy characterisation still work for an unbound system?

Yes, and that is the point of using $P_{\rm tot}$ rather than any statement about forces. The derivation assumes only that $P_{\rm tot}$ is a four-vector and that it is timelike. Two photons on their way out of a collision are not bound to anything, have no interaction and will never meet again — and they still have a rest frame, a rest energy, and a strict energy minimum there. "System" means nothing more than "this list of four-momenta, added up".

What breaks if the total four-momentum is null?

Everything, and cleanly. $\vec\beta_{\rm CM} = \vec Pc/E \to c$ , so the boost you would need is luminal and does not exist; $M = 0$ , so there is no rest energy; the energy function $E(u)$ has no minimum, and can be boosted arbitrarily close to zero without ever reaching it. Physically this is the case of exactly collinear massless particles — a set of measure zero, but a real exception, and `aq-sr-centre-of-momentum-frame` owns the proof. Note what is not the condition: massless constituents are fine, as long as they are not all pointing the same way.