Conservation of Four-Momentum

Newtonian physics hands you two conservation laws and no reason why there should be two. Energy is conserved; momentum is conserved; they are separate statements about separate quantities, and nothing in the theory connects them. Relativity connects them, and the connection is not a philosophical flourish — it is forced. Here is the argument in one move. Suppose momentum were conserved in every inertial frame but energy were not conserved in yours: some collision leaves the total momentum unchanged while the total energy jumps by $\Delta E = 1\ \mathrm{GeV}$ . Now boost to a frame moving at $\beta = 0.6$ . The Lorentz transformation mixes the two, $\Delta p'c = \gamma(\Delta pc - \beta\,\Delta E)$ , and with $\Delta p = 0$ and $\gamma = 1.25$ it hands you $\Delta p'c = -0.75\ \mathrm{GeV}$ .

Momentum is no longer conserved in the new frame. The energy leak has become a momentum leak, purely by changing your point of view. So the assumption collapses: you cannot have momentum conservation in all frames without energy conservation as well. They are not two laws that happen to both hold. They are four components of one law, and the single object that carries it is the total four-momentum

P^\mu = \sum_i p_i^\mu = \left(\frac{E_{\text{tot}}}{c},\; \vec p_{\text{tot}}\right), \qquad P^\mu_{\text{in}} = P^\mu_{\text{out}} .

That is the whole content of this article, and the rest is consequence: the invariant mass of a system, which is not the sum of its parts and which an inelastic collision genuinely changes; and Noether's theorem, which explains where the law came from in the first place. Two conventions are worth stating at once, because the AllQFT base article for this topic gets both wrong. We use the signature $(+,-,-,-)$ , so $p\cdot p = +m^2c^2$ — the base explicitly declares the opposite, $(-,+,+,+)$ with $p\cdot p = -m^2c^2$ . And the base's invariant-mass card is patched with a hand-inserted factor of $\cdot(-1)$ to paper over that choice, leaving an equation whose two sides do not even have the same units. Both are repaired below, in the open.

Key takeaways

Signature $(+,-,-,-)$ , $\eta = \mathrm{diag}(+1,-1,-1,-1)$ , $\;p\cdot p = +m^2c^2$ . Plus. The base declares $(-,+,+,+)$ ; we do not follow it.
The law: for an isolated system, $\boxed{P^\mu_{\text{in}} = P^\mu_{\text{out}}}$ where $P^\mu = \sum_i p_i^\mu$ . Four equations: one for energy, three for momentum.
Owned result: momentum conservation in every frame FORCES energy conservation. A boost mixes $E$ and $\vec p$ , so an energy leak in one frame reappears as a momentum leak in another. Energy and momentum conservation are not two laws — they are one four-vector law.
Conservation is frame-independent: $\Delta P^\mu = 0 \Rightarrow \Delta P'^\mu = \Lambda^\mu{}_\nu \Delta P^\nu = 0$ . A Lorentz transformation is linear, so it maps zero to zero. If it holds in one inertial frame it holds in all of them.
The system's invariant mass, dimensionally sound: $\boxed{M^2c^2 = P\cdot P = \dfrac{E_{\text{tot}}^2}{c^2} - |\vec p_{\text{tot}}|^2}$ , equivalently $E_{\text{tot}}^2 = |\vec p_{\text{tot}}|^2c^2 + M^2c^4$ . No stray $\cdot(-1)$ , and both sides carry the same units.
Key idea: $M \ne \sum_i m_i$ . A system's invariant mass counts the kinetic energy and the relative motion of its parts. It is conserved in any collision — but it is emphatically not the sum of the constituent masses, and in an inelastic collision the sum of the constituent masses changes while $M$ does not.
Two colliding particles of mass $m$ , each at $\gamma = 1.25$ , that stick together form one body of invariant mass $M = 2\gamma m = 2.5m$ — 20% of it created from kinetic energy.

The core insight

A Lorentz transformation is a linear map, and a linear map sends the zero vector to the zero vector. That single fact is the whole proof that four-momentum conservation is frame-independent: if $\Delta P^\mu = 0$ for one observer, then $\Delta P'^\mu = \Lambda^\mu{}_\nu \Delta P^\nu = \Lambda^\mu{}_\nu\, 0 = 0$ for every other. The law does not need to be re-checked frame by frame — it cannot fail in one frame and hold in another, because linearity forbids it.

Where this lives today

Every collision at every particle accelerator is analysed with equation (3), and the reason is that $M$ is invariant: a detector measures energies and momenta in the lab frame, and the invariant mass of a candidate set of decay products can be computed without ever transforming to the frame where the parent was at rest. Plot that invariant mass for millions of events and a new particle shows up as a bump — a peak at its rest mass sitting on a smooth background. That is how the $\pi^0$ appears at 135 MeV, and it is how the Higgs boson was found in 2012, as a peak near 125 GeV in the two-photon and four-lepton invariant-mass spectra. The technique is a direct application of the paragraph above about pions. The law's other great use is negative: when four-momentum appears not to balance, something has been missed. Pauli's neutrino is the canonical case — the continuous energy spectrum of beta decay looked like an outright violation of energy conservation, and Bohr was prepared to abandon the law. Pauli instead bet on the law and postulated an unseen particle to carry the missing four-momentum, a bet that took twenty-six years to pay off. The same logic runs today: at the LHC, "missing transverse momentum" is precisely a four-momentum imbalance in the plane transverse to the beams, and it is the primary signature for neutrinos and for hypothetical dark-matter particles that leave the detector unseen. And the invariant mass of a system is not a technicality but the origin of most of the mass around you. The proton's three valence quarks account for roughly 1% of its mass; the remainder is the energy of the gluon field and the quarks' relative motion, counted by $M^2c^2 = P\cdot P$ . You are, to a very good approximation, made of binding energy.

The physical picture

Each particle carries a four-momentum $p^\mu = mU^\mu = (E/c,\ \vec p\,)$ , and four-momenta add like vectors. So a system of particles has a total four-momentum $P^\mu = \sum_i p_i^\mu$ , which is itself a four-vector — sums of four-vectors are four-vectors — and the conservation law is the statement that this one object is unchanged by whatever happens inside the system. The power of packaging it this way is that a vector equation cannot half-hold. $\Delta P^\mu = 0$ is four scalar equations at once, and because $\Delta P^\mu$ transforms as a four-vector, its vanishing is a frame-independent fact — not a coincidence that has to be rechecked in every laboratory. The Newtonian picture, where energy and momentum are unrelated scalars and vectors, simply has no way to express that. And there is a second dividend. The length of $P^\mu$ is a Lorentz invariant, and it defines the invariant mass of the system, $M$ . This is the quantity that tells you what the system is, independent of who is looking — and it is not, in general, the sum of the masses of its parts.

Figure

PUNCHLINE — momentum conservation in EVERY frame FORCES energy conservation. Suppose in frame S momentum is conserved (Δp = 0) but energy leaks by ΔE = 1 GeV. Boost to β = 0.6 (γ = 1.25): the Lorentz transformation mixes them, Δp'c = γ(Δpc − βΔE) = −0.75 GeV. The energy leak has BECOME a momentum leak. So you cannot have one law without the other — they are four components of a single four-vector law

The core equations

The total four-momentum of a system is the vector sum of its parts:

\boxed{\;P^\mu \;=\; \sum_i p_i^\mu \;=\; \left(\frac{\sum_i E_i}{c},\;\; \sum_i \vec p_i\right) \;=\; \left(\frac{E_{\text{tot}}}{c},\;\; \vec p_{\text{tot}}\right)\;} \qquad (1)

where: $p_i^\mu = (E_i/c,\ \vec p_i)$ with $E_i = \gamma_i m_i c^2$ , $\vec p_i = \gamma_i m_i \vec v_i$ , and $m_i$ the invariant rest mass of particle $i$ . Physical reading: a sum of four-vectors is a four-vector, so $P^\mu$ transforms with $\Lambda$ like any other. Massless particles are included with no special pleading — a photon simply contributes $p^\mu = (E/c,\ \vec p\,)$ with $|\vec p| = E/c$ . For an isolated system (no external four-force), the law is:

\boxed{\;P^\mu_{\text{in}} \;=\; P^\mu_{\text{out}}\;} \qquad\Longleftrightarrow\qquad \sum_{\text{in}} E_i = \sum_{\text{out}} E_j \;\;\text{and}\;\; \sum_{\text{in}} \vec p_i = \sum_{\text{out}} \vec p_j \qquad (2)

where: "isolated" means $\mathrm{d}P^\mu/\mathrm{d}\tau = K^\mu_{\text{ext}} = 0$ . Physical reading: one four-vector equation, four scalar equations. Energy conservation is the $\mu = 0$ component and momentum conservation is $\mu = 1,2,3$ — and the derivation below shows you cannot keep one without the other. The law holds through any interaction whatsoever: elastic, inelastic, decay, annihilation, particle creation. The invariant mass of the system — the base's broken card, repaired:

\boxed{\;M^2c^2 \;=\; P\cdot P \;=\; \frac{E_{\text{tot}}^2}{c^2} \;-\; |\vec p_{\text{tot}}|^2\;} \qquad\Longleftrightarrow\qquad E_{\text{tot}}^2 = |\vec p_{\text{tot}}|^2c^2 + M^2c^4 \qquad (3)

where: $M$ — the invariant mass of the whole system, a Lorentz scalar. Physical reading: check the units and they balance, $\mathrm{kg^2\,m^2\,s^{-2}}$ on both sides. The base article prints $M^2c^2 = \frac{P^\mu P_\mu}{c^2}\cdot(-1) = \frac{E_{\text{tot}}^2}{c^4} - \frac{|\vec p_{\text{tot}}|^2}{c^2}$ , whose left side is $\mathrm{kg^2\,m^2\,s^{-2}}$ and whose right side is $\mathrm{kg^2}$ — an equation off by a factor of $c^2$ , patched with a hand-inserted $\cdot(-1)$ that exists only to undo its $(-,+,+,+)$ signature. In $(+,-,-,-)$ no such factor is needed: $P\cdot P$ is already $+M^2c^2$ , exactly as $p\cdot p = +m^2c^2$ for a single particle. And $M \ne \sum_i m_i$ — see the second warning. Since $P\cdot P$ is invariant, it can be evaluated in whichever frame is easiest and the answer used in any other. In the centre-of-momentum frame $\vec p_{\text{tot}} = \vec 0$ and equation (3) collapses to

M c^2 = E_{\text{tot}}^{\text{CM}} \qquad (4)

where: $E^{\text{CM}}_{\text{tot}}$ — the total energy in the frame where the system is at rest as a whole. Physical reading: the invariant mass of a system is its total energy in the CM frame, divided by $c^2$ — the cleanest possible statement of what mass means for a composite object, and the reason a hot gas weighs more than a cold one.

Why one law replaces two

Derivation AConservation is frame-independent, and linearity is the entire reason

Nothing here is deep, which is exactly the point: the frame-independence of the conservation law is a triviality once the conserved quantity is packaged as a four-vector, and is nearly impossible to see otherwise.

Step 1. Define the change in total four-momentum across the interaction: $\;\Delta P^\mu \equiv P^\mu_{\text{out}} - P^\mu_{\text{in}}$ . Why: this is a difference of four-vectors, hence itself a four-vector — the object we can transform.

Step 2. Transform it. Under a boost, $\;\Delta P'^\mu = \Lambda^\mu{}_\nu\,\Delta P^\nu$ . Why: $\Delta P^\mu$ is a four-vector, so this is just the definition of what that means. No physics has been used yet.

Step 3. Apply linearity. $\Lambda$ is a linear map, so $\Lambda^\mu{}_\nu\, 0 = 0$ . Hence $\;\Delta P^\mu = 0 \;\Longrightarrow\; \Delta P'^\mu = 0$ . Why: conservation in one inertial frame implies conservation in every inertial frame. There is no need to verify the law frame by frame, and no possibility of it holding in the lab but failing on a passing train.

Step 4. Note the converse, which does the real work. $\Lambda$ is invertible ( $\Lambda^{-1}$ is the boost with $\vec\beta \to -\vec\beta$ ), so a non-zero $\Delta P^\mu$ maps to a non-zero $\Delta P'^\mu$ . A violation cannot be hidden by changing frames, and neither can it be manufactured. Why: this is what makes the next derivation bite. A leak in one frame is a leak in all of them — but, crucially, it need not leak out of the same component. *Verified in verify_aq-sr-conservation-of-four-momentum.py: $\Lambda\,\Delta P = 0 \iff \Delta P = 0$ checked against 300 random boosts and random $\Delta P$ ; $\det\Lambda = 1$ and $\Lambda^{\mathsf T}\eta\Lambda = \eta$ confirmed for each.*

Go deeper · Derivation B · 2 minMomentum conservation in every frame FORCES energy conservation

Derivation BMomentum conservation in every frame FORCES energy conservation

This is the article's owned result, and it is the reason relativity has one conservation law where Newton had two. The Newtonian laws are logically independent — you could imagine a universe with one and not the other. Relativity forbids it, and the proof is a single boost.

Step 1. Assume the contrary, in the most favourable form. Suppose that in every inertial frame the total momentum is conserved, $\Delta\vec p = \vec 0$ , but that in our frame $S$ the energy is not: $\Delta E \ne 0$ . Why: we are granting momentum conservation universally and asking only whether energy conservation can be dropped. If it can, they are independent laws.

Step 2. Boost along $x$ and transform the deficit. Since $\Delta P^\mu = (\Delta E/c,\ \Delta\vec p\,)$ is a four-vector, its components mix exactly as $(E/c, \vec p\,)$ do: \Delta p'_x c = \gamma\left(\Delta p_x c - \beta\,\Delta E\right), \qquad \Delta E' = \gamma\left(\Delta E - \beta\,\Delta p_x c\right) . Why: the transformation of the deficit is the transformation of a four-vector, no matter that it happens to be a deficit.

Step 3. Impose the assumption and read the damage. With $\Delta p_x = 0$ : \Delta p'_x c = -\gamma\beta\,\Delta E \;\ne\; 0 \qquad\text{whenever } \Delta E \ne 0 \text{ and } \beta \ne 0. Momentum is not conserved in $S'$ . Why: this contradicts the assumption that momentum is conserved in every frame. The energy leak in $S$ has become a momentum leak in $S'$ — the boost converted one into the other.

Step 4. Put numbers on it. Take $\Delta E = 1\ \mathrm{GeV}$ (a 1 GeV energy leak, momentum conserved) and boost at $\beta = 0.6$ , $\gamma = 1.25$ : \Delta p'_x c = 1.25\,(0 - 0.6 \times 1\ \mathrm{GeV}) = -0.75\ \mathrm{GeV}. Why: three-quarters of a GeV of momentum appears from nowhere, in a frame reachable by any passing observer. Meanwhile $\Delta E' = 1.25\,(1 - 0) = 1.25\ \mathrm{GeV}$ : the energy leak has grown too.

Step 5. Conclude. The assumption is impossible: momentum conservation in all frames requires $\Delta E = 0$ . And the same argument run the other way — energy conserved in all frames, momentum not — fails identically, since $\Delta E' = -\gamma\beta\,\Delta p_x c \ne 0$ . Neither law can stand alone. They are the four components of the single statement $\Delta P^\mu = 0$ . Why: this is the sharpest available answer to "why does relativity unify energy and momentum?" It does not unify them for elegance. Anything less than the full four-vector law is self-contradictory. *Verified in verify_aq-sr-conservation-of-four-momentum.py: the printed $\Delta p'c = -0.75$ GeV recomputed from $\gamma(\Delta pc - \beta\Delta E)$ ; and, over 400 random boosts, any $\Delta P^\mu$ with $\Delta\vec p = 0$ but $\Delta E \ne 0$ is confirmed to produce $\Delta\vec p\,' \ne 0$ in every boosted frame with $\beta \ne 0$ (zero exceptions).*

Go deeper · Derivation C · 2 minThe invariant mass of a system, and how an inelastic collision changes the mass of its parts

Derivation CThe invariant mass of a system, and how an inelastic collision changes the mass of its parts

Now the repaired card (3), and the subtlety it exists to protect. There is a genuine exception to the rule " $m$ never changes" hiding here, and it is important to state it precisely rather than to fudge it.

Step 1. Define it as the length of the total four-momentum. $\;M^2c^2 \equiv P\cdot P = \left(\dfrac{E_{\text{tot}}}{c}\right)^2 - |\vec p_{\text{tot}}|^2$ . Why: $P\cdot P$ is a Lorentz scalar, so $M$ is the same number for every observer. Units: both sides are $\mathrm{kg^2\,m^2\,s^{-2}}$ . Note there is no factor of $\cdot(-1)$ and no division by $c^2$ — those are artefacts of the base's opposite signature, and one of them broke the units.

Step 2. Evaluate it in the easiest frame. Because it is invariant, compute it wherever you like. In the centre-of-momentum frame, $\vec p_{\text{tot}} = \vec 0$ , so $\;Mc^2 = E_{\text{tot}}^{\text{CM}}$ . Why: the invariant mass of a system is its CM-frame energy. This includes the kinetic energy of the constituents relative to one another — which is why $M$ is not the sum of the parts' masses.

Step 3. Show $M \ne \sum_i m_i$ with the cleanest case. Two photons, each of energy $E$ , flying apart back-to-back: $E_{\text{tot}} = 2E$ , $\vec p_{\text{tot}} = \vec 0$ , so $M = 2E/c^2 \ne 0$ — a system of two massless particles has non-zero invariant mass. Why: mass is not a substance that gets added up. It is the length of a four-vector, and two null vectors can sum to a timelike one.

Step 4. The sticky collision. Two particles of mass $m$ approach head-on, each with $\gamma = 1.25$ , and stick together. Before: $E_{\text{tot}} = 2\gamma mc^2 = 2.5\,mc^2$ and $\vec p_{\text{tot}} = \vec 0$ (they are equal and opposite), so $M = 2\gamma m = 2.5\,m$ . After: the single lump is at rest with invariant mass $M_{\text{lump}}$ , and conservation forces $M_{\text{lump}}c^2 = E_{\text{tot}} = 2.5\,mc^2$ . Why: so $M_{\text{lump}} = 2.5\,m$ , while the constituents contributed only $2m$ .

Step 5. State the exception precisely. The sum of the constituent rest masses went from $2m$ to $2.5m$ : it is not conserved, and $0.5m$ — 20% of the final mass — was created out of kinetic energy. But the system's invariant mass $M$ never changed: it was $2.5m$ before the collision and $2.5m$ after. Why: both statements are true and they are not in conflict. $M$ is conserved because $P^\mu$ is conserved; $\sum_i m_i$ is not conserved because it is not the length of anything. The kinetic energy did not vanish — it became rest mass, and it was already counted in $M$ all along. This is the one legitimate sense in which "mass changes" in relativity: a system's invariant mass differs from the sum of its parts', and an inelastic collision converts between the two. It is not, and never was, a licence for a velocity-dependent "relativistic mass" of a single particle. *Verified in verify_aq-sr-conservation-of-four-momentum.py: DIMENSIONAL AUDIT of the base's card (LHS $\mathrm{kg^2m^2s^{-2}}$ vs RHS $\mathrm{kg^2}$ — asserted UNEQUAL) and of the repaired card (asserted EQUAL); $M = 2\gamma m = 2.5m$ and the 20% created fraction; and the two-photon $M = 2E/c^2$ .*

Figure

The invariant mass of a SYSTEM is not the sum of its parts. Two particles of mass m, each at γ = 1.25, collide head-on and stick. The system's invariant mass M = 2γm = 2.5m is CONSERVED throughout — but the sum of the constituent rest masses jumps from 2m to 2.5m. 20% of the final lump's mass was created out of kinetic energy. This is the one legitimate sense in which mass changes: it is a statement about a SYSTEM, never a velocity-dependent mass of a particle

Limiting cases

Case	Result	Meaning
$v \ll c$	$E_{\text{tot}} \approx \sum_i m_ic^2 + \sum_i \tfrac12 m_iv_i^2$	Newton's two laws re-emerge as separate-looking pieces of one
all $\vec p_i$ parallel, massless	$M = 0$	co-moving photons: two null vectors, still null
two photons back-to-back	$M = 2E/c^2 \ne 0$	a system of massless particles has mass
CM frame	$Mc^2 = E_{\text{tot}}$	the invariant mass is the CM energy
elastic collision	$M$ conserved, $\sum_i m_i$ conserved	nothing converts
inelastic collision	$M$ conserved, $\sum_i m_i$ changes	kinetic energy $\leftrightarrow$ rest mass
non-isolated system	$\mathrm{d}P^\mu/\mathrm{d}\tau = K^\mu_{\text{ext}}$	the four-force is the leak
signature $(-,+,+,+)$	$P\cdot P = -M^2c^2$	same physics, opposite sign — the base's choice

Common trapThe base article declares the opposite signature, and its invariant-mass card is dimensionally broken.

Two separate defects in the AllQFT base for this topic, and the second is a consequence of the first.

One — the signature. The base states, verbatim: "Using metric signature $(-,+,+,+)$ so that $p^\mu p_\mu = -m^2c^2$ ." This article uses $(+,-,-,-)$ throughout, so $p\cdot p = +m^2c^2$ and $P\cdot P = +M^2c^2$ . Both conventions are legitimate in isolation. Mixing them is not, and mixing is what happened next.

Two — the units. The base's invariant-mass card reads

M^2c^2 = \frac{P^\mu P_\mu}{c^2}\cdot(-1) = \frac{E_{\text{tot}}^2}{c^4} - \frac{|\vec p_{\text{tot}}|^2}{c^2} .

The left side has units $\mathrm{kg^2\,m^2\,s^{-2}}$ . The right side has units $\mathrm{kg^2}$ . The equation is off by a factor of $c^2$ and cannot be right in any signature. The hand-inserted $\cdot(-1)$ is there purely to undo the $(-,+,+,+)$ sign, and while chasing that sign the author lost track of a $c^2$ . The repair in $(+,-,-,-)$ needs no fudge factor at all:

M^2c^2 = P\cdot P = \frac{E_{\text{tot}}^2}{c^2} - |\vec p_{\text{tot}}|^2, \qquad\text{or equivalently}\qquad M^2 = \frac{E_{\text{tot}}^2}{c^4} - \frac{|\vec p_{\text{tot}}|^2}{c^2} .

Both are dimensionally sound; the second is the base's right-hand side with the correct left-hand side attached. The lesson is the reliable one: a stray $\cdot(-1)$ in a formula is a signature bug wearing a disguise, and it is worth a dimensional check every time.

Common trapThe system's invariant mass is not the sum of the parts. And no, this is not "relativistic mass".

$M \ne \sum_i m_i$ . A system's invariant mass counts the kinetic energy of its constituents and the energy of their interaction; a box of hot gas is genuinely heavier than the same box cold, and two back-to-back photons have $M = 2E/c^2$ despite having no mass at all individually. There is therefore one true statement of the form "mass changes": in an inelastic collision the system's invariant mass differs from the sum of its constituents' rest masses, and kinetic energy converts into rest mass. That is a fact about a system. It is not relativistic mass. Each particle's own $m$ is an invariant scalar and does not change with speed — there is no $\gamma m$ masquerading as a mass anywhere in this article, and the correct way to say that a fast particle is hard to accelerate is with the response coefficients of Relativistic Force, never with a velocity-dependent mass. If you find yourself writing $m(v)$ , you have conflated the two statements.

Worked examples

1 · The opening angle of the two photons from a neutral pion

A neutral pion decays at rest, $\pi^0 \to \gamma\gamma$ , with $m_{\pi^0}c^2 = 134.9768\ \mathrm{MeV}$ . Find each photon's energy and the angle between them, then show how the invariant mass is recovered from the photons alone.

Step 1 — identify. An isolated decay: $P^\mu$ before equals $P^\mu$ after. Before, the pion is at rest, so $P^\mu = (m_{\pi^0}c,\ \vec 0\,)$ .

Step 2 — conserve momentum. $\vec p_{\text{tot}} = \vec 0$ before, so the two photons must fly exactly back-to-back with equal and opposite momenta, hence equal energies. The opening angle is $\theta = \mathbf{180^\circ}$ , and no other angle is possible for a pion at rest.

Step 3 — conserve energy. $E_1 + E_2 = m_{\pi^0}c^2$ with $E_1 = E_2$ , so each photon carries E_\gamma = \tfrac12 \times 134.9768 = \mathbf{67.4884\ \mathrm{MeV}} .

Step 4 — recover the mass from the photons. For two massless particles, equation (3) reduces to $M^2c^4 = 2E_1E_2(1 - \cos\theta)$ . At $\theta = 180^\circ$ : M c^2 = \sqrt{2(67.4884)^2(1 - (-1))} = \sqrt{4 \times 67.4884^2} = 2 \times 67.4884 = \mathbf{134.9768\ \mathrm{MeV}} \;\checkmark — the pion mass, reconstructed from two massless particles.

Step 5 — the boundary number. Notice what the formula says about the opening angle: at $\theta = 0$ the two photons are collinear and $Mc^2 = 0$ , while at $\theta = 90^\circ$ we would get $\sqrt{2}\times 67.4884 = \mathbf{95.4430\ \mathrm{MeV}}$ . The invariant mass of a two-photon system is carried entirely by the angle between them, which is exactly how a detector reconstructs $\pi^0$ mesons: measure two photon energies and one opening angle, and the invariant-mass peak appears at 135 MeV. (A moving pion has its opening angle squeezed below $180^\circ$ by the boost, but $M$ — being invariant — comes out at 134.9768 MeV regardless of how fast the pion was going. That is the whole point of using an invariant.)

2 · The sticky collision, and where the missing mass went

Two particles of rest mass $m$ collide head-on, each with $\gamma = 1.25$ ( $\beta = 0.6$ ), and stick together. Find the invariant mass of the resulting lump, and account for the bookkeeping.

Step 1 — identify. Perfectly inelastic, in the CM frame (the momenta are equal and opposite, so this frame is the CM frame). Kinetic energy is not conserved — but $P^\mu$ is, and that is all we need.

Step 2 — total four-momentum before. $E_{\text{tot}} = 2\gamma mc^2 = 2 \times 1.25\, mc^2 = \mathbf{2.5\,mc^2}$ , and $\vec p_{\text{tot}} = \vec 0$ by symmetry. So $P^\mu = (2.5\,mc,\ \vec 0\,)$ .

Step 3 — apply conservation. The lump inherits $P^\mu$ exactly: it is at rest (since $\vec p_{\text{tot}} = \vec 0$ ) with energy $2.5\,mc^2$ . By equation (4), its invariant mass is M = \frac{E_{\text{tot}}}{c^2} = \mathbf{2.5\,m} .

Step 4 — the bookkeeping. The two particles brought $2m$ of rest mass and $2(\gamma - 1)mc^2 = 0.5\,mc^2$ of kinetic energy. The lump has $2.5\,m$ of rest mass and no kinetic energy. The kinetic energy did not disappear — it became rest mass, and the sum of the constituent masses rose from $2m$ to $2.5m$ .

Step 5 — the boundary number. The created mass is $0.5m$ out of a final $2.5m$ : 20% of the lump's mass did not exist as mass before the collision. And yet the system's invariant mass $M$ was $2.5m$ the entire time, before and after — because $P^\mu$ never changed. The mass that "appeared" was in $M$ all along, carried as the kinetic energy of the parts. This is precisely how a proton gets its mass: the three valence quarks contribute only about 1% of it, and the rest is the energy of the gluon field and the quarks' motion — invariant mass, not summed mass.

Physics of the portrait

Conservation laws feel like brute facts of nature until you learn where they come from. They come from symmetries, and it was Emmy Noether who proved it — in a paper written in Göttingen while she was being denied a salary, a title, and the right to lecture under her own name.

Origins · 2 minConservation law and how this was discovered

Conservation law

1918 — the symmetry origin · Noether's Invariante Variationsprobleme: every continuous symmetry is the source of a conserved quantity

A conservation law had, for 250 years, been a fact you discovered: momentum is conserved, energy is conserved, angular momentum is conserved — each established separately, essentially from Newton's third law plus observation. Emmy Noether turned that list into a single structural statement, and the person is inseparable from the idea. Amalie Emmy Noether was born in Erlangen on 23 March 1882 and took her doctorate there in 1907, at a time when German universities admitted women only by special dispensation. Hilbert and Klein brought her to Göttingen in 1915 to help with the conservation-of-energy problem then troubling general relativity, and the faculty refused to appoint her — Hilbert's reported retort, that the senate was not a bathhouse, did not carry the day, and for four years she lectured under his name. Her answer to the problem, "Invariante Variationsprobleme", was presented to the Göttingen society on 26 July 1918 — presented by Felix Klein, since she was not a member and could not present it herself. Its first theorem states that every continuous symmetry of a physical system's action yields a conserved quantity, and it is the reason this article's law exists at all: invariance under time translation gives conservation of energy, invariance under spatial translation gives conservation of momentum. In relativity these are unified — spacetime translation is a single symmetry group, and its conserved current is the single four-vector $P^\mu$ of this page. Noether went on to found modern abstract algebra; Einstein called her the most significant creative mathematical genius produced since the higher education of women began. She was dismissed by the Nazi regime in 1933 for being Jewish, emigrated to Bryn Mawr, and died there on 14 April 1935, aged 53.

History

1905

— Einstein's relativity forces energy and momentum into a single framework, though the four-vector language does not yet exist to say so.

1908

— Minkowski's four-dimensional formalism supplies that language, and the total four-momentum of a system becomes expressible as one object.

1909

— Lewis and Tolman derive relativistic momentum conservation from the principle of relativity alone (Phil. Mag. 18, 510), showing that Newtonian momentum cannot be conserved in every frame.

1918

— Emmy Noether's "Invariante Variationsprobleme" is presented to the Göttingen society on 26 July by Felix Klein, proving that every continuous symmetry yields a conserved quantity — spacetime translation invariance yields conservation of four-momentum.

1930

— Pauli's letter of 4 December to the Tübingen meeting, addressed to "Dear Radioactive Ladies and Gentlemen", postulates the neutrino as a "desperate remedy" to save energy and momentum conservation in beta decay.

1956

— Reines and Cowan detect the neutrino and telegraph Pauli on 14 June; the conservation law had been right, and the missing four-momentum was a particle.

Going further

Texts. Rindler, Introduction to Special Relativity, 2nd ed., §§5.6–6.4 · Griffiths, Introduction to Elementary Particles, 2nd ed., §3.3 · Taylor and Wheeler, Spacetime Physics, 2nd ed., ch. 7 · Noether, "Invariante Variationsprobleme", Gött. Nachr. (1918), 235–257 · Particle Data Group, Review of Particle Physics (2024), for $m_{\pi^0}c^2 = 134.9768 \pm 0.0005\ \mathrm{MeV}$ .

Provenance. Signature $(+,-,-,-)$ : $\;p\cdot p = +m^2c^2$ , $\;P\cdot P = +M^2c^2$ , and every $m$ is an invariant rest mass — no relativistic mass appears anywhere. Base-defect repairs (logged), two of them. (i) The base declares "metric signature $(-,+,+,+)$ so that $p^\mu p_\mu = -m^2c^2$ ", the opposite of the convention locked for this corpus; everything here is re-derived in $(+,-,-,-)$ . (ii) The base's invariant-mass eqCard, $M^2c^2 = \frac{P^\mu P_\mu}{c^2}\cdot(-1) = \frac{E^2}{c^4} - \frac{|\vec p|^2}{c^2}$ , is dimensionally broken — LHS $\mathrm{kg^2\,m^2\,s^{-2}}$ , RHS $\mathrm{kg^2}$ , off by $c^2$ — and its $\cdot(-1)$ is a signature artefact. Repaired to $M^2c^2 = P\cdot P = E_{\text{tot}}^2/c^2 - |\vec p_{\text{tot}}|^2$ , with the dimensional audit itself machine-checked. All numbers are recomputed in verify_aq-sr-conservation-of-four-momentum.py, which parses this article's own printed digits.

Scope. Threshold energy for particle production and the Compton shift are not derived here — they are the business of Relativistic Collisions and Compton Scattering respectively, and the base article's derivation slots for those topics belong to those pages. Noether's theorem is stated, not proved; its proof is the business of the Lagrangian-mechanics band.

Builds on. Four-Momentum · Relativistic Energy · Relativistic Momentum · Energy–Momentum Relation · Invariant Mass. Leads to. Four-Force · Massless Particles · Mass–Energy Equivalence.

Is energy conservation a separate law from momentum conservation in relativity?

No — and that is the central claim of this article. They are the $\mu = 0$ and $\mu = 1,2,3$ components of the single equation $\Delta P^\mu = 0$ . The derivation above shows the two cannot be separated even in principle: assume momentum is conserved in every frame but energy is not, and a boost turns the energy leak into a momentum leak, contradicting the assumption.

Is mass conserved in a collision?

Two different questions hide in that one. The system's invariant mass $M$ , defined by $M^2c^2 = P\cdot P$ , is conserved in any collision, because $P^\mu$ is. The sum of the constituent rest masses is not conserved — in an inelastic collision it grows, as kinetic energy converts to rest mass. Neither statement licenses a velocity-dependent "relativistic mass" for a single particle; each particle's $m$ is invariant throughout.

How can two massless photons have a non-zero invariant mass?

Because invariant mass is the length of the total four-momentum, not a sum of the parts' masses, and two null vectors can add to a timelike one. Two photons of energy $E$ flying back-to-back have $\vec p_{\text{tot}} = \vec 0$ and $E_{\text{tot}} = 2E$ , so $M = 2E/c^2$ . Only if they travel in the same direction does the sum stay null and $M = 0$ .

Why is the invariant mass so useful experimentally?

Because it is the same number in every frame, so a detector can compute it directly from lab-frame energies and momenta without knowing anything about the frame in which the parent particle was at rest. Every particle discovery of the last century is, in effect, a bump in an invariant-mass plot — the $\pi^0$ at 135 MeV, the Higgs at about 125 GeV.

What does Noether's theorem actually add?

It explains why the law holds rather than merely asserting it. Every continuous symmetry of the action yields a conserved quantity: invariance under time translation gives energy conservation, invariance under spatial translation gives momentum conservation. In relativity these are one symmetry — invariance under spacetime translation — and its single conserved current is exactly the four-momentum $P^\mu$ of this page.

The base article uses (−,+,+,+). Was it simply wrong to?

Not in itself — $(-,+,+,+)$ is a perfectly standard convention, used by MTW and much of the general-relativity literature, and the physics is identical. What is wrong is the consequence: the base's invariant-mass card carries a hand-inserted $\cdot(-1)$ to undo its own sign, and in the course of that patch a factor of $c^2$ was lost, leaving an equation whose two sides have different units. A stray $\cdot(-1)$ is nearly always a signature bug in disguise, and a dimensional check catches it every time.