Four-Momentum and Energy-Momentum Relation

In Newtonian physics energy and momentum are separate bookkeeping: one scalar, one vector, each conserved by its own law, with no reason to think of them together. Special relativity dissolves that separation. Once time is a coordinate on the same footing as space — the lesson of the Minkowski metric and the Lorentz transformations — the natural object that rides along a particle's worldline is the four-velocity, the derivative of position with respect to proper time. Multiply it by the particle's invariant rest mass and you obtain the four-momentum $p^\mu$ , a single geometric arrow whose time part is energy (over $c$ ) and whose space part is the ordinary momentum. Because this arrow is built from the four-velocity, it transforms as a contravariant four-vector, so its conservation in one inertial frame guarantees conservation in every frame — a boost merely reshuffles energy into momentum and back, exactly as a rotation reshuffles the components of a spatial vector. And because the four-velocity has a fixed Minkowski length, so does the four-momentum: its magnitude is locked to the rest mass alone, $p^\mu p_\mu = -m^2c^2$ . Written out in components, that one invariant statement is the celebrated energy–momentum relation $E^2 = (pc)^2 + (mc^2)^2$ , which contains Einstein's $E=mc^2$ at rest and the null relation $E=pc$ for light as its two extremes. Everything downstream — collision thresholds, particle decays, Compton scattering, the conserved orbital energy of a star falling toward a black hole — is this single arrow and its fixed length, worked with covariant algebra instead of frame-by-frame arithmetic.

Key takeaways

The four-momentum $p^\mu = (E/c,\,\mathbf p)$ fuses energy and three-momentum into one four-vector that boosts exactly like the displacement $\mathrm dx^\mu$ .
It is rest mass times four-velocity, $p^\mu = m\,u^\mu = m\,\mathrm dx^\mu/\mathrm d\tau$ , which makes it manifestly a four-vector with $\tau$ the proper time.
Its Minkowski magnitude is a Lorentz invariant fixed by the mass alone: $p^\mu p_\mu = -m^2c^2$ (signature $-+++$ ).
In components that invariant is $E^2 = (pc)^2 + (mc^2)^2$ , giving $E=mc^2$ at rest and $E=pc$ for massless particles like the photon.
Total four-momentum is conserved in every frame at once, so collisions, decays and thresholds are read off from invariants such as $s = -(\sum p^\mu)(\sum p_\mu)$ .
In curved spacetime $p^\mu$ is the geodesic tangent, and its contraction with a Killing vector $\xi^\mu$ gives a conserved energy or angular momentum.
Key idea: energy and momentum are the four components of one four-vector $p^\mu = m u^\mu$ ; its Lorentz-invariant length $p^\mu p_\mu = -m^2c^2$ is the energy–momentum relation $E^2 = (pc)^2 + (mc^2)^2$ .

The core insight

Energy and momentum are the four components of a single four-vector $p^\mu = m u^\mu$ ; its Lorentz-invariant magnitude $p^\mu p_\mu = -m^2c^2$ is the energy–momentum relation $E^2 = (pc)^2 + (mc^2)^2$ .

Where this lives today

The four-momentum is the daily currency of experimental physics. Every LHC event is reconstructed by adding up the four-momenta of the outgoing tracks and reading the invariant mass of the sum — a bump at 125 GeV is how the Higgs boson announced itself in 2012, and "missing transverse momentum" is how detectors infer neutrinos or dark-matter candidates they cannot see. The same threshold algebra that sized the Bevatron is the central argument for colliders over fixed-target machines: a fixed target wastes energy because $E_{\mathrm{cm}}$ grows only as $\sqrt{2 m c^2 E_{\mathrm{beam}}}$ , whereas two colliding beams put all of it to work, $E_{\mathrm{cm}} = 2E_{\mathrm{beam}}$ . On cosmic scales, four-momentum conservation against the microwave background sets the GZK cutoff near $5\times10^{19}$ eV, above which protons lose energy to pion production and the cosmic-ray spectrum should steepen — a prediction the Pierre Auger Observatory tests. And in curved spacetime the very same $p^\mu$ , now the tangent to a geodesic, delivers the conserved orbital energy and angular momentum of matter spiraling into a black hole, and the gravitational redshift of light climbing out — the bridge from this special-relativistic object to the observational machinery of general relativity.

The physical picture

Picture the four-momentum as an arrow drawn in spacetime, anchored at the particle and pointing along its worldline. Its Minkowski length is fixed once and for all by the rest mass: a proton's arrow is exactly as long as any other proton's, no matter how fast it moves. When you change frames — boost to a train, a rocket, a falling elevator — you rotate your view of that arrow. What changes is how much of it points "in time" (the energy $E/c$ ) versus "in space" (the momentum $\mathbf p$ ); what never changes is the arrow's own length. A particle at rest is an arrow aimed purely along the time axis, of length $mc$ , carrying pure rest energy $E=mc^2$ and no momentum at all. The catch, and the reason this is not just a spatial vector in disguise, is that Minkowski length is indefinite: the time part enters with a minus sign. So a longer time-component (more energy) is bought by a longer space-component (more momentum) with the difference $E^2-(pc)^2$ held rigidly fixed. The tips of all the arrows a boosted observer could assign to one particle trace out a hyperbola — the mass shell — not a circle. The first figure is that hyperbola, drawn with real numbers.

Figure

The energy–momentum hyperbola, the mass shell E² = (pc)² + (mc²)² for a particle of rest energy mc² = 1 GeV. Every inertial observer places the particle somewhere on this curve; the rest energy is the intercept at pc = 0, and a fast particle with total energy E = 5 GeV sits at momentum pc = 4.90 GeV. The dashed line E = pc is the light cone, the massless limit the hyperbola never crosses.

The same fixed length also runs the movie of a boost. As you speed past the particle, its arrow does not stretch or shrink — it slides along the hyperbola, trading time-component for space-component like a bead on a wire. That sliding is a hyperbolic rotation (a boost), the exact analogue of an ordinary rotation for a Euclidean arrow, and it is why a quantity conserved in one frame is conserved in all: the total arrow is the same object seen from different angles. Conservation of four-momentum then becomes plain vector addition — the incoming arrows, laid head to tail, equal the outgoing arrows, simultaneously in every frame. The second figure shows one particle's arrow at rest and after a boost, both tips pinned to the same mass shell.

Figure

Boosting slides the four-momentum arrow along the mass hyperbola p⁰ = √(p² + m²c²). The rest arrow points straight up the energy axis with length mc; the boosted arrow is longer in both its time part (energy E/c) and its space part (momentum p), yet its invariant Minkowski length p·p = −m²c² is unchanged. The dashed diagonal is the light cone p⁰ = p.

The core equations

The starting point is a definition, not a law: the four-momentum is the rest mass times the four-velocity. Because $m$ is a Lorentz scalar and $u^\mu = \mathrm dx^\mu/\mathrm d\tau$ is a four-vector (proper time $\tau$ is invariant), the product is guaranteed to transform as a four-vector — the property that makes conservation frame-independent.

p^\mu = m\,u^\mu = m\,\frac{\mathrm dx^\mu}{\mathrm d\tau} \qquad (1)

where: $p^\mu$ — four-momentum, components $(p^0,p^1,p^2,p^3)$ ; $m$ — invariant rest mass, a Lorentz scalar; $u^\mu$ — four-velocity $\mathrm dx^\mu/\mathrm d\tau$ ; $\tau$ — proper time along the worldline. Reading: the four-momentum inherits its vector character wholesale from the four-velocity, and its magnitude wholesale from the mass. Splitting the definition into time and space parts, and using $\mathrm dt/\mathrm d\tau = \gamma$ (time dilation), gives the physical identity of the components. The time part is the energy over $c$ ; the space part is the relativistic three-momentum.

p^\mu = \left(\frac{E}{c},\ \mathbf p\right) = \left(\gamma m c,\ \gamma m \mathbf v\right) \qquad (2)

where: $E = \gamma m c^2$ — total relativistic energy; $\mathbf p = \gamma m \mathbf v$ — relativistic three-momentum; $\gamma = 1/\sqrt{1-v^2/c^2}$ — Lorentz factor; $\mathbf v$ — ordinary three-velocity. Reading: energy is the "time-component of momentum", which is why a boost mixes the two. The heart of the topic is that this four-vector has a Lorentz-invariant length, and that the length is set by the rest mass alone. In signature $-+++$ it is negative because the object is timelike.

p^\mu p_\mu = \eta_{\mu\nu}\,p^\mu p^\nu = -m^2 c^2 \qquad (3)

where: $p_\mu = \eta_{\mu\nu}p^\nu$ — covariant four-momentum; $\eta_{\mu\nu} = \mathrm{diag}(-1,1,1,1)$ — Minkowski metric. Reading: this single number is the same in every inertial frame and along every geodesic; it is the mass, geometrically. Writing that invariant out in the energy–momentum components turns the geometric statement into the algebraic relation every free particle obeys — the relativistic dispersion relation.

E^2 = (pc)^2 + (mc^2)^2 \qquad (4)

where: $E$ — total energy; $p = |\mathbf p|$ — magnitude of three-momentum; $mc^2$ — rest energy. Reading: energy and momentum are the two legs of a right triangle whose hypotenuse-like invariant is the rest energy; $E=mc^2$ at $p=0$ and $E=pc$ at $m=0$ are the two corners. Boosting the four-momentum along $x$ is the same Lorentz matrix that acts on $\mathrm dx^\mu$ — a hyperbolic rotation mixing energy and momentum. This is the algebra behind "sliding along the hyperbola".

\begin{pmatrix} E'/c \\ p'_x \end{pmatrix} = \begin{pmatrix} \gamma & -\gamma\beta \\ -\gamma\beta & \gamma \end{pmatrix} \begin{pmatrix} E/c \\ p_x \end{pmatrix} \qquad (5)

where: $\beta = v/c$ — boost speed in units of $c$ ; $E',\,p'_x$ — energy and momentum in the boosted frame. Reading: an observer who boosts sees a different split of the same arrow into energy and momentum. For a system of particles the total four-momentum is conserved, and the invariant built from its square is the squared center-of-momentum energy — the master quantity of relativistic kinematics, used for every threshold and resonance.

s = -\Big(\sum_i p_i^\mu\Big)\Big(\sum_i p_{i\,\mu}\Big) = \frac{E_{\mathrm{cm}}^2}{c^2} \qquad (6)

where: $s$ — Mandelstam invariant (total four-momentum squared); $E_{\mathrm{cm}}$ — total energy in the center-of-momentum frame; $p_i^\mu$ — four-momentum of particle $i$ . Reading: whatever the frame, $s$ is the same, so it can be evaluated in the lab and set equal to its center-of-momentum value at threshold.

Derivation AFrom four-velocity to the fixed length of the four-momentum

The goal is to show that the four-momentum is a genuine four-vector whose Minkowski magnitude is nailed to the rest mass, $p^\mu p_\mu = -m^2c^2$ , and to read off its components as energy and momentum. We start from the definition $p^\mu = m u^\mu$ , expand the four-velocity into ordinary time and velocity using the proper-time chain rule, and then contract the four-momentum with itself using the norm of the four-velocity that proper time forces upon it. Every step leans only on results the prerequisite articles own — proper time, the Minkowski metric, and the four-velocity.

Step 1. Expand the four-velocity by the chain rule through coordinate time: $u^\mu = \mathrm dx^\mu/\mathrm d\tau = (\mathrm dt/\mathrm d\tau)(\mathrm dx^\mu/\mathrm dt) = \gamma\,(c,\mathbf v)$ , using $\mathrm dt/\mathrm d\tau = \gamma$ . Why: proper time is the invariant clock, and its rate relative to coordinate time is exactly the time-dilation factor $\gamma$ , which injects the velocity dependence.

Step 2. Multiply by the scalar mass to get the components of the four-momentum: $p^\mu = m u^\mu = (\gamma m c,\ \gamma m \mathbf v)$ . Identifying $E \equiv \gamma m c^2$ makes the time part $p^0 = E/c$ and the space part $\mathbf p = \gamma m \mathbf v$ . Why: this is where "energy" is defined so that the time-component of the momentum is $E/c$ ; the low-speed limit $\gamma \to 1 + v^2/2c^2$ then gives $E \to mc^2 + \tfrac12 mv^2$ , tying it to Newton.

Step 3. Compute the norm of the four-velocity. Along any worldline the interval is $\mathrm ds^2 = -c^2\,\mathrm d\tau^2$ , and dividing by $\mathrm d\tau^2$ gives $u^\mu u_\mu = -c^2$ , a fixed number independent of speed. Why: proper time is defined so that the four-velocity is a unit timelike vector (of length $c$ ); this constancy is what will pin the four-momentum's length.

Step 4. Scale by $m^2$ : $p^\mu p_\mu = m^2\,u^\mu u_\mu = -m^2 c^2$ . Because $m$ is a scalar and $u^\mu u_\mu$ is invariant, the result is a Lorentz invariant — the same in every frame and, in curved spacetime, conserved along a geodesic. Why: this is the geometric content of "mass" — not a component but the invariant length of the momentum arrow. *The three results to carry forward: $u^\mu u_\mu = -c^2$ (from $\mathrm ds^2 = -c^2\mathrm d\tau^2$ ); $p^\mu p_\mu = -m^2c^2$ ; and components $(E/c,\mathbf p)=(\gamma mc,\gamma m\mathbf v)$ with $E\to mc^2+\tfrac12mv^2$ at low speed.*

Go deeper · Derivation B · 1 minThe energy–momentum relation and its two extreme corners

Derivation BThe energy–momentum relation and its two extreme corners

The aim is to convert the invariant $p^\mu p_\mu = -m^2c^2$ into the working algebraic relation $E^2 = (pc)^2 + (mc^2)^2$ , and then to extract from it the two limits that make it famous — the rest energy $E=mc^2$ and the massless relation $E=pc$ — together with the Newtonian and ultra-relativistic expansions. This is pure component bookkeeping in the $-+++$ metric, but it is the single most-used equation in relativistic mechanics, so it is worth doing slowly and watching the minus sign.

Step 1. Lower an index with $\eta = \mathrm{diag}(-1,1,1,1)$ and contract: $p^\mu p_\mu = -\,(p^0)^2 + \mathbf p\cdot\mathbf p = -\,(E/c)^2 + p^2$ . Set this equal to the invariant $-m^2c^2$ . Why: the time-component enters with a minus sign in this signature, which is the whole reason energy and momentum combine as a difference of squares rather than a sum.

Step 2. Rearrange: $(E/c)^2 = p^2 + m^2c^2$ , and multiply through by $c^2$ to reach $E^2 = (pc)^2 + (mc^2)^2$ . Why: this is the mass shell of the first figure — for fixed $m$ it is the hyperbola that every observer's $(pc,E)$ must lie on.

Step 3. Read off the corners. At $p=0$ the particle is at rest and $E = mc^2$ , the rest energy. At $m=0$ the term vanishes and $E = pc$ , so the arrow lies on the light cone with $p^\mu p_\mu = 0$ — a null vector with energy and momentum but no rest frame. Why: the same formula holds Einstein's $E=mc^2$ and the photon's $E=pc$ as its two boundaries, which is why one relation governs both matter and light.

Step 4. Expand between the corners. For $pc \ll mc^2$ , $E = mc^2\sqrt{1+(p/mc)^2} \approx mc^2 + p^2/2m - p^4/8m^3c^2$ , recovering rest energy plus Newtonian kinetic energy plus corrections; for $pc \gg mc^2$ , $E \approx pc + (mc^2)^2/2pc$ , so fast particles behave almost like light. Why: the interpolation between Newton and the null limit shows the relation is not two special cases glued together but one smooth curve. *The four results to carry forward: $E^2=(pc)^2+(mc^2)^2$ ; $E=mc^2$ at $p=0$ ; $E=pc$ at $m=0$ ; low-speed $E\approx mc^2+p^2/2m$ ; ultra-relativistic $E\approx pc+(mc^2)^2/2pc$ .*

Go deeper · Derivation C · 1 minThe Compton shift from four-momentum conservation alone

Derivation CThe Compton shift from four-momentum conservation alone

The goal is to obtain the Compton wavelength shift $\Delta\lambda = (h/m_ec)(1-\cos\theta)$ without ever writing a component equation for the recoiling electron — using only conservation of total four-momentum and the fact that squaring a four-vector gives a frame-independent invariant. This is the showcase of the formalism: a messy two-body collision collapses to one line of invariant algebra because the electron's mass is unchanged.

Step 1. Conserve total four-momentum: $p_\gamma^\mu + p_e^\mu = p_\gamma'^\mu + p_e'^\mu$ , and isolate the unknown recoil electron, $p_e'^\mu = p_\gamma^\mu + p_e^\mu - p_\gamma'^\mu$ . Why: we never actually want the electron's energy and angle, so we hide them in a single four-vector and eliminate them by squaring.

Step 2. Square both sides and use the invariant masses: $p_e'^\mu p_{e'\mu} = -m_e^2c^2$ (unchanged) and the photon norms $p_\gamma^\mu p_{\gamma\mu} = p_\gamma'^\mu p_{\gamma'\mu} = 0$ . Expanding the right side, the $-m_e^2c^2$ terms cancel and leave $p_\gamma^\mu p_{e\mu} - p_\gamma'^\mu p_{e\mu} - p_\gamma^\mu p_{\gamma'\mu} = 0$ . Why: the electron's invariant mass is a fixed scalar, so squaring turns "what happened to the electron" into a known number and it drops out.

Step 3. Insert components with the electron initially at rest, $p_e^\mu = (m_ec,\mathbf 0)$ , and photon energies $E_\gamma = p_\gamma c$ : the dot products give $m_ec\,(E_\gamma - E_\gamma') = (E_\gamma E_\gamma'/c)\,(1-\cos\theta)$ , where $\theta$ is the photon scattering angle. Why: the rest frame of the target makes $p_e^\mu$ trivial, and the photon dot product $p_\gamma^\mu p_{\gamma'\mu}$ carries the scattering angle through $1-\cos\theta$ .

Step 4. Divide by $E_\gamma E_\gamma'$ and use $E = hc/\lambda$ : $1/E_\gamma' - 1/E_\gamma = (1-\cos\theta)/m_ec^2$ , i.e. $\Delta\lambda = \lambda' - \lambda = (h/m_ec)(1-\cos\theta)$ . Why: the shift is set purely by the electron's Compton wavelength $h/m_ec$ , independent of the incident wavelength — the signature that this is relativistic kinematics, not classical scattering. *The results to carry forward: $\Delta\lambda=(h/m_ec)(1-\cos\theta)$ with Compton wavelength $h/m_ec = 2.426\times10^{-12}$ m; the shift vanishes at $\theta=0$ and reaches its maximum $2h/m_ec$ at $\theta=\pi$ .*

Limiting cases

The energy–momentum relation interpolates smoothly between the regimes below; each is one point or tail of the mass hyperbola.

Case	Result	Meaning
Rest (massive)	$E = mc^2,\quad \mathbf p = 0$	pure rest energy; the four-momentum is $(mc,\mathbf 0)$ , straight up the time axis
Massless	$E = pc,\quad p^\mu p_\mu = 0$	photons and gravitons: null four-momentum, no rest frame, energy set by frequency $E=\hbar\omega$
Non-relativistic ( $pc \ll mc^2$ )	$E \approx mc^2 + \dfrac{p^2}{2m} - \dfrac{p^4}{8m^3c^2}$	rest energy plus Newtonian kinetic energy plus the first relativistic correction
Ultra-relativistic ( $pc \gg mc^2$ )	$E \approx pc + \dfrac{(mc^2)^2}{2pc}$	mass is a small correction; the particle hugs the light cone like light

Common trapThe mass in E² = (pc)² + (mc²)² is a "relativistic mass" that grows with speed

It is not. The $m$ in the energy–momentum relation is the invariant rest mass — a single scalar fixed once and for all, the same in every frame and the same whether the particle sits still or races through the LHC. The old "relativistic mass" $\gamma m$ is nothing but $E/c^2$ in disguise; substituting it into $E^2=(pc)^2+(mc^2)^2$ double-counts the velocity factor and gives nonsense. The velocity dependence lives entirely in $E$ and $\mathbf p$ , never in $m$ . Two related traps: mind the signature (with $-+++$ the norm is $-m^2c^2$ , with $+---$ it is $+m^2c^2$ — same physics, opposite sign), and remember that the invariant mass of a system is not the sum of the parts' masses.

Worked examples

1 · The Bevatron antiproton threshold (fully numeric)

Antiprotons are made by firing protons at a stationary hydrogen target: $p + p \to p + p + p + \bar p$ . Baryon number forces three protons plus one antiproton in the final state. What minimum beam energy is needed? Take $m_p c^2 = 938.3$ MeV.

Step 1. Write the invariant $s$ two ways. In the lab, the beam proton has $(E_{\mathrm{beam}}, \mathbf p_{\mathrm{beam}})$ and the target is $(m_pc^2, \mathbf 0)$ , so $s c^2 = (E_{\mathrm{beam}} + m_pc^2)^2 - (p_{\mathrm{beam}}c)^2$ . Using the beam's own dispersion $E_{\mathrm{beam}}^2 - (p_{\mathrm{beam}}c)^2 = (m_pc^2)^2$ , this simplifies to $s c^2 = 2 m_pc^2 E_{\mathrm{beam}} + 2(m_pc^2)^2$ .

Step 2. Evaluate $s$ at threshold. The cheapest final state has all four particles at rest together in the center-of-momentum frame, so the minimum invariant mass is the sum of the rest masses: $\sqrt{s}\,c = 4 m_pc^2$ , hence $s c^2 = 16(m_pc^2)^2$ .

Step 3. Equate and solve: $2 m_pc^2 E_{\mathrm{beam}} + 2(m_pc^2)^2 = 16(m_pc^2)^2$ , giving $E_{\mathrm{beam}} = 7\,m_pc^2$ . Numerically $E_{\mathrm{beam}} = 7 \times 0.9383 = 6.57$ GeV, a kinetic energy $K = E_{\mathrm{beam}} - m_pc^2 = 5.63$ GeV and a beam momentum $p_{\mathrm{beam}}c = \sqrt{E_{\mathrm{beam}}^2 - (m_pc^2)^2} = 6.50$ GeV. This is exactly the design energy of Berkeley's Bevatron, which discovered the antiproton in 1955. The invariant $s$ turns a four-body threshold into one line of algebra: $E_{\mathrm{beam}}^{\mathrm{thr}} = 7 m_pc^2 \approx 6.57$ GeV.

2 · Compton backscatter of a 511 keV gamma ray (fully numeric)

A 511 keV annihilation photon scatters straight back ( $\theta = 180°$ ) off a free electron at rest. Find the wavelength shift and the energy of the scattered photon. Use $h/m_ec = 2.426$ pm and $m_ec^2 = 511$ keV.

Step 1. The shift is maximal at backscatter: $\Delta\lambda = (h/m_ec)(1-\cos 180°) = 2\,(h/m_ec) = 2 \times 2.426 = \mathbf{4.85}$ pm. The incident wavelength is $\lambda = hc/E_\gamma = (1240\ \mathrm{eV\,nm})/(511\,000\ \mathrm{eV}) = 2.427$ pm — by coincidence essentially the Compton wavelength itself.

Step 2. So $\lambda' = \lambda + \Delta\lambda = 2.427 + 4.85 = 7.28$ pm, and the scattered energy is $E_\gamma' = hc/\lambda' = 1240/7.28\times10^{-3} = 170$ keV.

Step 3. Cross-check with the energy form $E_\gamma' = E_\gamma /\big[1 + (E_\gamma/m_ec^2)(1-\cos\theta)\big] = 511/(1 + 1\times 2) = 511/3 = \mathbf{170}$ keV, agreeing exactly. The electron carries off the missing $511 - 170 = 341$ keV. Notice the whole result followed from one invariant equation, never solving for the electron's recoil angle. Backscatter: $\Delta\lambda = 4.85$ pm and $E_\gamma' = 170$ keV.

3 · Momentum and speed of a 5 GeV particle (fully numeric)

A particle of rest energy $mc^2 = 1$ GeV is measured to have total energy $E = 5$ GeV. Find its momentum, its speed, and its Lorentz factor.

Step 1. From $E^2 = (pc)^2 + (mc^2)^2$ : $(pc)^2 = 25 - 1 = 24\ \mathrm{GeV}^2$ , so $pc = \sqrt{24} = 4.90$ GeV, i.e. $p = 4.90$ GeV/ $c$ .

Step 2. Speed from the ratio $\beta = v/c = pc/E = 4.90/5 = \mathbf{0.980}$ — 98% of $c$ . This uses $\mathbf p/E = \mathbf v/c^2$ , a direct consequence of $p^\mu = (\gamma mc, \gamma m\mathbf v)$ .

Step 3. Lorentz factor $\gamma = E/mc^2 = 5/1 = 5$ , and as a check $1/\sqrt{1-0.980^2} = 1/\sqrt{0.0396} = 5.0$ . $p = 4.90$ GeV/ $c$ , $v = 0.980\,c$ , $\gamma = 5$ — a triangle whose legs are $pc = 4.90$ and $mc^2 = 1$ and whose hypotenuse is $E = 5$ .

Physics of the portrait

The four-momentum is the crispest fruit of one idea: that space and time are a single arena through which a particle carries one arrow. No one made that idea concrete more decisively than the man who named "spacetime".

Origins · 1 minHermann Minkowski and how this was discovered

Hermann Minkowski

1864–1909 · the geometer who fused space and time into spacetime and made energy and momentum one four-vector

Hermann Minkowski was Einstein's own mathematics teacher at the Zurich Polytechnic — he once called the young Einstein a "lazy dog" — yet it was Minkowski who supplied the geometric language in which Einstein's 1905 relativity became inevitable rather than paradoxical. In his 1908 Cologne lecture Raum und Zeit he opened with the now-famous manifesto that "henceforth space by itself, and time by itself, are doomed to fade away into mere shadows, and only a kind of union of the two will preserve an independent reality." That union is Minkowski spacetime, and in it the separate Newtonian conservation laws for energy and for momentum become a single statement: the conservation of one four-vector whose time part is energy and whose space part is momentum. The invariant length of that vector — what we now write $p^\mu p_\mu = -m^2c^2$ — is the mass, a frame-independent number rather than a component. Minkowski built the four-dimensional tensor calculus, the light cone, the worldline, and the notion of proper time that organize everything here; his indefinite metric is exactly why a nonzero null four-momentum can exist and photons can be massless yet energetic. He died suddenly of appendicitis in 1909, at 44, only months after the lecture, and never saw general relativity — which rests entirely on the spacetime geometry he founded — take shape from his student's hands.

History

1905

— Einstein derives mass–energy equivalence, E = mc², from special relativity

1906

— Planck recasts relativistic dynamics with momentum p = γmv

1908

— Minkowski's lecture Raum und Zeit unifies energy and momentum into one four-vector

1928

— Dirac's relativistic electron equation is built on E² = (pc)² + (mc²)²

1955

— Chamberlain and Segrè discover the antiproton at the Bevatron's 7mₚc² threshold

Going further

Texts. Misner, Thorne & Wheeler, Gravitation, §2.4–2.8 (four-momentum as a geometric object) · Wald, General Relativity, §4.2 · Carroll, Spacetime and Geometry, §1.3–1.6 (energy–momentum and kinematics) · Hartle, Gravity, Ch. 5 (relativistic collisions and decays) · Schutz, A First Course in General Relativity, Ch. 2 (four-vectors and four-momentum). Key numerical benchmarks used throughout this article: $p^\mu p_\mu = -m^2c^2$ in $-+++$ signature follows from $u^\mu u_\mu = -c^2$ ; the Bevatron antiproton threshold is $E_{\mathrm{beam}} = 7 m_pc^2 \approx 6.57$ GeV (with $m_pc^2 = 938.3$ MeV); the Compton shift is $\Delta\lambda = (h/m_ec)(1-\cos\theta)$ with Compton wavelength $h/m_ec = 2.426\times10^{-12}$ m, and 511 keV backscatter gives $E_\gamma' = 170$ keV; the GZK cutoff is of order $\sim 5\times10^{19}$ eV. All numbers are CODATA/PDG textbook values.

Prerequisites. Four-Vectors · Proper Time · Minkowski Metric and Signature · Lorentz Transformations · Lorentz Boosts and Rapidity. Leads to. Causal Structure and Light Cones · Spacetime Interval.

Ask the tutor: Ask me to walk through any relativistic collision or decay by adding up four-momenta and reading off the invariant mass.

Why call it a four-momentum rather than just tracking energy and momentum separately?

Because they are not separate — a boost mixes them, so any split into "energy" and "momentum" is frame-dependent bookkeeping of one underlying object. Treating $p^\mu$ as a single four-vector buys two things. First, manifest covariance: if the total $p^\mu$ is conserved in one inertial frame, it is automatically conserved in all of them, because a four-vector equation keeps its form under Lorentz transformations. Second, invariants: the square $p^\mu p_\mu$ is a number every observer agrees on, so you can compute it in whichever frame is easiest and equate it across a process. Together these replace tedious frame-by-frame arithmetic with one line of covariant algebra, as the threshold and Compton examples show.

What exactly is the invariant mass, and how can a system of massless particles have one?

The invariant mass of anything — one particle or a whole collection — is defined from the total four-momentum by $M^2c^2 = -P^\mu P_\mu$ , where $P^\mu = \sum_i p_i^\mu$ . For a single particle this reduces to its rest mass. For a system it is generally not the sum of the parts' masses, because you are squaring a sum of four-vectors, and the cross terms depend on relative directions and energies. Two photons of energy $E$ flying in opposite directions have total $P^\mu = (2E/c,\mathbf 0)$ , whose square gives $M = 2E/c^2$ even though each photon is massless. Physically, mass measures the energy content in the frame where the total momentum vanishes — and a pair of oppositely-directed photons already has zero net momentum.

Can a single photon spontaneously turn into an electron–positron pair in empty space?

No, and four-momentum conservation is exactly why. The photon's four-momentum is null, $p^\mu p_\mu = 0$ , so the invariant mass of the "before" state is zero. An electron–positron pair, however, has invariant mass at least $2m_e$ (their combined rest mass, more if they move relative to each other), so the "after" state has $p^\mu p_\mu \le -(2m_ec)^2 < 0$ . Since the total four-momentum — and hence its invariant square — cannot change, a null vector cannot become a timelike one, and the decay is forbidden. Pair production does happen, but only near a nucleus (or another photon), which absorbs the recoil four-momentum needed to balance the books.

Does E = mc² mean mass and energy are the same thing?

Not quite — it says the rest energy of a body equals its mass times $c^2$ , a specific corner of the fuller relation $E^2 = (pc)^2 + (mc^2)^2$ . Mass is the invariant length of the four-momentum, the energy measured in the body's own rest frame; total energy $E = \gamma mc^2$ is only the time-component, and it grows without bound as the body speeds up while the mass stays fixed. So mass and energy are not identical: mass is one particular, frame-independent value of energy. What $E=mc^2$ really asserts is that even a motionless body has energy $mc^2$ locked in its mass — the source tapped by fission, fusion, and annihilation.

How does the four-momentum carry over into curved spacetime and general relativity?

Almost unchanged, which is the point. In curved spacetime a free particle follows a geodesic, and $p^\mu = m\,\mathrm dx^\mu/\mathrm d\tau$ is the tangent to that worldline — for photons, parametrized by an affine parameter instead of proper time. Its norm is still $p^\mu p_\mu = -m^2c^2$ (or $0$ for light), now contracted with the local metric $g_{\mu\nu}$ rather than $\eta_{\mu\nu}$ . The powerful new feature is conserved charges: if the spacetime has a symmetry, encoded by a Killing vector $\xi^\mu$ , then $\xi_\mu p^\mu$ is constant along the geodesic. The timelike Killing vector of a static spacetime yields conserved orbital energy; the rotational one yields angular momentum — the quantities that let you solve black-hole orbits and derive gravitational redshift.

What is the Mandelstam s, and why evaluate it in two different frames?

The Mandelstam variable $s = -(\sum_i p_i^\mu)(\sum_i p_{i\mu})$ is just the invariant mass squared of the whole system, and it equals $E_{\mathrm{cm}}^2/c^2$ , the total energy in the center-of-momentum frame. Its value is the same in every frame — that is the entire trick. In a collision you know the initial state in the lab (a beam hitting a target), so you compute $s$ there; but the physics of what can be produced is cleanest in the center-of-momentum frame, where at threshold all products sit at rest and $\sqrt{s}\,c$ is simply the sum of their rest masses. Setting the lab expression equal to the threshold value gives the beam energy needed, as in the Bevatron example — no explicit frame transformation required.