Collision Kinematics and Center-of-Momentum Frames

The center-of-momentum frame of an isolated system is determined by its total four-momentum, not by averaging particle velocities. When the total momentum is timelike, its invariant length is the system’s center-of-momentum energy and its unit direction is the frame’s four-velocity.

Required background. Events, Intervals, and Minkowski Geometry begins the Special Relativity sequence and supplies causal geometry; Relativistic Energy, Momentum, and Mass Shells supplies four-momentum and $p^2=-m^2$. Helpful background. Calculus and Linear Algebra supplies vector sums and quadratic invariants; Classical and Variational Mechanics supplies momentum conservation and the Newtonian center-of-mass limit.

Invariant mass and the center-of-momentum observer

For an asymptotically free incoming or outgoing particle state, with interactions negligible and every emitted particle or radiation quantum included, let

$$P^\mu=\sum_a p_a^\mu$$

for all particles in that state. Four-momentum conservation states that an isolated process has the same $P^\mu$ in its asymptotic initial and final states.

At finite times an interaction or field can itself carry four-momentum, so the particle sum need not be complete. The full conserved translation charge must then include those degrees of freedom, as developed in Poincaré Symmetry and Relativistic Conservation Laws. This page stays with asymptotically free particle states.

With signature $(-,+,+,+)$, define

$$s=-P^\mu P_\mu.$$

If $P^\mu$ is future timelike, $s\gt0$ and the center-of-momentum four-velocity is

$$U_{\rm CM}^\mu=\frac{P^\mu}{\sqrt{s}}.$$

The energy measured by this observer is

$$E_{\rm CM}=-P\cdot U_{\rm CM}=\sqrt{s},$$

while its projected total spatial momentum vanishes:

$$h^\mu{}_{\nu}(U_{\rm CM}) =\delta^\mu{}_{\nu}+U_{\rm CM}^\mu U_{{\rm CM}\nu}, \qquad h^\mu{}_\nu(U_{\rm CM})P^\nu=0.$$

This proves that $U_{\rm CM}$ is the unique future unit timelike center-of-momentum four-velocity. An adapted coordinate frame still has conventional spatial axes and origin. The invariant mass $\sqrt{s}$ includes internal kinetic and interaction energy; it is generally not the sum of constituent rest masses.

Two incoming particles

For momenta $p_1,p_2$ and masses $m_1,m_2$,

$$s=-(p_1+p_2)^2 =m_1^2+m_2^2-2p_1\cdot p_2.$$

If $m_2\gt0$, particle 2 has a rest frame in which $p_2^\mu=(m_2,\mathbf0)$, and

$$s=m_1^2+m_2^2+2m_2E_1.$$

This fixed-target identity is often more useful than explicitly boosting. It also displays a practical fact: much of a high-energy projectile’s laboratory energy goes into motion of the center of momentum rather than into invariant energy available for new rest mass.

For two beams of equal energy $E$ and opposite momentum, $P^\mu=(2E,\mathbf0)$ and $\sqrt{s}=2E$. A collider uses energy more efficiently than a fixed target because the laboratory can already be the center-of-momentum frame.

Two-body final states

Consider a final state with masses $m_3,m_4$. In the center-of-momentum frame,

$$p_3^{*\mu}=(E_3^*,\mathbf p^*), \qquad p_4^{*\mu}=(E_4^*,-\mathbf p^*).$$

Energy conservation and the two mass shells give

$$E_3^*=\frac{s+m_3^2-m_4^2}{2\sqrt{s}}, \qquad E_4^*=\frac{s+m_4^2-m_3^2}{2\sqrt{s}},$$

and

$$|\mathbf p^*| =\frac{\sqrt{\lambda(s,m_3^2,m_4^2)}}{2\sqrt{s}},$$

where

$$\lambda(x,y,z) =x^2+y^2+z^2-2xy-2xz-2yz.$$

Reality of $|\mathbf p^*|$ requires $\lambda\ge0$, but that algebraic inequality alone also admits the lower branch $s\le(m_3-m_4)^2$. Requiring both final momenta to be future-directed removes that branch: a physical two-body final state exists only for $\sqrt{s}\ge m_3+m_4$. With two positive masses, equality is an attained threshold and both particles are at rest in the center-of-momentum frame. If a daughter is massless, equality would assign it zero four-momentum, so the boundary may be only an infimum of physical final states rather than an attained state.

Worked example: equal masses with a fixed target

A particle of mass $m$ and laboratory energy $E$ strikes an identical particle at rest. Then

$$s=2m^2+2mE.$$

If the projectile momentum magnitude is $p=\sqrt{E^2-m^2}$, its laboratory speed is $v=p/E$, and the center-of-momentum frame moves with

$$\beta_{\rm CM}=\frac{p}{E+m}.$$

Indeed, the total laboratory momentum and energy are $(E+m,\mathbf p)$, so a boost at that speed removes the spatial part. In the nonrelativistic limit $E=m+K$ with $K\ll m$, $\beta_{\rm CM}\simeq v/2$, as expected for equal Newtonian masses.

In the ultrarelativistic fixed-target limit $E\gg m$,

$$\sqrt{s}\simeq\sqrt{2mE},$$

which grows only as the square root of projectile energy. Two opposed beams with the same single-beam energy instead give $\sqrt{s}=2E$.

The figure makes the four-momentum bookkeeping visible. Inspect the total spatial momentum: it survives in the fixed-target laboratory but cancels between opposed beams.

For equal masses and equal single-beam total energy $E$, a fixed target has $P=(E+m,p)$ and $\sqrt{s}=\sqrt{2m(E+m)}$, while opposed beams have $P=(2E,0)$ and $\sqrt{s}=2E$. The comparison is schematic and not to scale.

When a center-of-momentum frame fails to exist

For a finite collection of nonzero future-directed causal momenta, the sum is timelike unless every summand is null and all are proportional to the same future null direction. In that exact exceptional case every constituent is massless and all momenta point the same way. Then $P^2=0$, no boost can make the spatial momentum vanish, and dividing by $\sqrt{s}$ is impossible. Two photons that are not parallel in the same direction—including a head-on pair—have a timelike sum and do possess a center-of-momentum frame.

External fields or radiation crossing the system boundary complicate the accounting: particle momentum alone may not be conserved. The system must include the relevant field or environmental momentum flux.

Checks and common mistakes

• With the site signature, $s=-P^2$; references using $(+,-,-,-)$ write $s=P^2$. The numerical invariant is the same.
• A frame transformation cannot change $s$. If two frame calculations give different values, a sign, energy, or momentum component is wrong.
• The two-body formulas determine momentum magnitudes, not angular distributions. Scattering dynamics supplies the latter.
• “Center of mass” is potentially misleading for photons or interacting systems; center of momentum states the invariant condition.
• Adding rest masses gives the minimum possible invariant mass only when all constituents can be mutually at rest and binding energy is absent.

The center-of-momentum and fixed-target relations are reviewed in Miller, Tovey, and Jackson 2025, §49.2, printed p. 1, Eqs. (49.2)–(49.7), and the two-body energies and Källén formula in §49.4.2, printed p. 3, Eqs. (49.16)–(49.18). The source uses $(+,-,-,-)$ and $s=P^2$; both signs have been translated to the site convention.

Exercises

Two-photon invariant mass

Two photons have energies $E_1,E_2$ and propagation directions separated by an angle $\psi$ in one inertial frame. Show that

$$s=2E_1E_2(1-\cos\psi),$$

and interpret the parallel and head-on limits.

Solution

Write $k_a^\mu=E_a(1,\mathbf n_a)$ with $|\mathbf n_a|=1$. Since each photon is null,

$$s=-(k_1+k_2)^2=-2k_1\mathbin{\cdot}k_2 =2E_1E_2(1-\mathbf n_1\mathbin{\cdot}\mathbf n_2),$$

and $\mathbf n_1\mathbin{\cdot}\mathbf n_2=\cos\psi$. For $\psi=0$, $s=0$: the total momentum is null and no center-of-momentum frame exists. For $\psi=\pi$, $s=4E_1E_2$; equal energies give a laboratory center-of-momentum frame with $\sqrt{s}=2E_1$.

Verify the center-of-momentum observer

Assume $P$ is future timelike and set $U=P/\sqrt{-P^2}$. Verify $U^2=-1$, $-P\mathbin{\cdot}U=\sqrt{-P^2}$, and $h(U)P=0$.

Solution

Let $s=-P^2\gt0$. Then

$$U^2=\frac{P^2}{s}=-1, \qquad -P\mathbin{\cdot}U=-\frac{P^2}{\sqrt{s}}=\sqrt{s}.$$

Finally,

$$h^\mu{}_{\nu}(U)P^\nu =P^\mu+U^\mu(U\mathbin{\cdot}P) =P^\mu-\frac{P^\mu}{\sqrt{s}}\sqrt{s}=0.$$

Continue to Decays, Thresholds, and Invariant Phase Space to turn these kinematic constraints into allowed final-state measures.

References

• Miller, D., D. R. Tovey, and J. D. Jackson. “Kinematics.” In S. Navas et al. (Particle Data Group), Review of Particle Physics, Physical Review D 110, 030001 (2024), 2025 update. Official PDF.

Further reading

• Tong, David. Lectures on Dynamics and Relativity, §7.5, printed pp. 145–149, University of Cambridge, 2013.