Relativistic Kinematics in Particle Collisions

The $\phi$ meson weighs $1019.461\ \mathrm{MeV}/c^2$ and decays, most of the time, into a $K^+K^-$ pair. The $\rho$ meson weighs $775.26\ \mathrm{MeV}/c^2$ and never does — not once, not rarely, not suppressed: the channel is simply shut. Two kaons weigh $987.354\ \mathrm{MeV}/c^2$ between them, and the $\rho$ does not have it. That much is obvious. What is not obvious is that the same three numbers, read as the sides of a triangle, say it too. Lay out segments of length $775.26$ , $493.677$ and $493.677$ : they close into a perfectly good isosceles triangle, of area $118\,503\ \mathrm{MeV}^2$ . Lay out $1019.461$ , $493.677$ , $493.677$ : the two short sides cannot reach, and there is no triangle at all. The decay that happens is the one whose mass triangle fails to close. That is not a coincidence and not a mnemonic. The function governing every two-body relativistic decay and every threshold — the Källén function $\lambda$ — is, on the nose, minus sixteen times the squared Heron area of the mass triangle, and equally minus four times the Gram determinant of the two four-momenta. One determinant, two readings: in Euclidean signature an area, in Minkowski signature a phase-space volume, and the sign separating them is the sign separating a forbidden reaction from an allowed one.

Key takeaways

Convention lock. $(+,-,-,-)$ , $\eta = \mathrm{diag}(+1,-1,-1,-1)$ . Four-momentum in energy units, $P^\mu = (E,\ \vec p\,c)$ , so $P\cdot P = E^2 - |\vec p\,|^2c^2 = m^2c^4$ . Then $s$ , $t$ , $u$ are in energy² and $\sqrt{s}$ is an energy — the convention `aq-sr-mandelstam-invariants` boxes.
The method. Never boost. Write $\sum_{\rm in} P^\mu = \sum_{\rm out} P^\mu$ , isolate the four-momenta you do not want, and square. Each square kills an unknown angle and returns a Lorentz scalar. Choose a frame only at the very end.
The owned result — $\lambda$ is a determinant, twice over. For two four-momenta with Gram matrix $G_{ij} = P_i\cdot P_j$ : $\ \lambda(s,\,m_1^2c^4,\,m_2^2c^4) = -4\det G$ . And for any three lengths: $\ \lambda(a^2,b^2,c^2) = -16\,A(a,b,c)^2$ , with $A$ the Heron area.
Hence the triangle criterion, proved rather than asserted. $\lambda \geq 0 \iff M$ , $m_1$ , $m_2$ cannot be the sides of a triangle $\iff M \geq m_1 + m_2 \iff$ the decay is allowed. `aq-sr-particle-decay-kinematics` states the triangle fact; here it is a theorem, and $\lambda = 0$ is the degenerate, zero-area triangle — the threshold.
The geometry underneath. $\det G < 0$ means the plane spanned by $P_1$ and $P_2$ contains a timelike direction, so a CM frame exists. Positive $\lambda$ , timelike plane, real $p^{*}$ , imaginary triangle: four names for one fact.
What it buys. $p^{*} = \sqrt{\lambda}/(2c\sqrt{s})$ — the whole two-body phase space, from three masses and no angles. Thresholds, decay momenta and missing-mass reconstruction fall out of the same square.
Key idea: a relativistic reaction is allowed exactly when its mass triangle refuses to close. The Källén function measures the failure, and the failure is a Gram determinant — which is why it is frame-independent, why it is the phase-space volume, and why it changes sign at threshold and nowhere else.

The core insight

Two four-vectors span a plane, and a plane in Minkowski space has a causal character: it either contains a timelike direction or it does not. The Gram determinant $\det G = (P_1\cdot P_1)(P_2\cdot P_2) - (P_1\cdot P_2)^2$ is the invariant that decides which — negative for a timelike plane, positive for a spacelike one, zero on the boundary. Now notice that in Euclidean geometry the very same determinant is $4A^2$ , four times the squared area of the triangle the two vectors span. The Källén function is that determinant carried into Minkowski space, and its sign flip is the entire content of "kinematically allowed": a real reaction spans a timelike plane, its Gram determinant is negative, and so its mass triangle has negative squared area and cannot be drawn. Ask whether the triangle closes; you have asked whether the reaction is forbidden.

Where this lives today

Every kinematic fit at every collider runs on this. When ATLAS or CMS reconstructs a decay chain the software is not boosting between frames — it is computing invariant masses by squaring sums of four-momenta, and evaluating $\lambda$ for the two-body phase-space factor. A Monte Carlo generator sampling a $1 \to 2$ decay draws two angles and gets everything else from $\lambda$ , because everything else is $\lambda$ . The missing-mass move is how the field found half its particles. Measure one outgoing track, square the rest, plot $M_X$ : a peak is a state. The $\Omega^-$ and much of the early hadron spectrum came out of exactly this histogram, and the LHC's search for invisible particles is the same search with better detectors. Threshold physics remains a live design constraint: every accelerator ever built was sized against $\sqrt{s} \geq \sum_f m_f c^2$ , and the Fermi-motion correction of the second worked example is why subthreshold production is a measured effect in nuclear collisions rather than a paradox. With three particles out, $\lambda$ alone no longer settles it: the allowed region becomes a two-dimensional Dalitz plot whose boundary is — inevitably — a Gram determinant, this time $3\times3$ .

The physical picture

Collision kinematics has a reputation for being a swamp of $\gamma$ 's and $\beta$ 's, and it earns that reputation only if you do it wrong. The wrong way is to pick the lab frame, write components, and boost. The right way — the invariant-first method — is to refuse to pick a frame at all. Everything rests on two facts and no others. Four-momentum is conserved, $\sum_{\rm in}P^\mu = \sum_{\rm out}P^\mu$ , and each particle sits on its mass shell, $P\cdot P = m^2c^4$ . Both hold in every frame. So any dot product you build out of the $P$ 's is a number — the same number for every observer — and you may evaluate it in whichever frame makes it trivial, then use it in whichever frame the detector lives. The algorithm has one move. Isolate what you do not know, and square it. Squaring a four-vector destroys its direction and leaves its invariant length, so an unmeasured angle disappears in a single line. In a two-body decay $M \to 1 + 2$ you do not know the emission direction — so write $P_2 = P - P_1$ , square, and the direction is gone; what is left is an equation in the three masses alone. No boost appears anywhere in it. What comes out, over and over, is one function of three arguments:

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

It is called the Källén function, or the triangle function, and everyone who meets it is told the name comes from a triangle and then handed no triangle. This article hands over the triangle — and shows that the triangle is a determinant, and that the determinant is why $\lambda$ knows which reactions are allowed.

The core equations

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

P\cdot P \;=\; \eta_{\mu\nu}P^\mu P^\nu \;=\; E^2 - |\vec p\,|^2c^2 \;=\; m^2c^4 \;>\; 0 \qquad (1)

where: timelike is positive, as the signature demands. Physical reading: every quantity below is in energy² units. $s$ is an energy², $\sqrt{s}$ an energy, $\lambda$ an energy⁴. This is the convention shipped by `aq-sr-mandelstam-invariants`; mixing it with the momentum²-units convention is the commonest way to lose a factor of $c^2$ . The total four-momentum of a system, and its invariant:

P^\mu_{\rm tot} = \sum_i P_i^\mu, \qquad s \;\equiv\; 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: $\sqrt{s}$ is the total energy in the centre-of-momentum frame — proved in `aq-sr-centre-of-momentum-frame`, recalled here. Physical reading: $\sqrt{s}$ is the only energy a reaction can actually spend. The Källén function, and the factorisation that makes it readable:

\lambda(x,y,z) = x^2+y^2+z^2-2xy-2yz-2zx, \qquad \boxed{\;\lambda(a^2,b^2,c^2) = (a{+}b{+}c)(a{+}b{-}c)(a{-}b{+}c)(a{-}b{-}c)\;} \qquad (3)

where: $a, b, c$ are the three lengths whose squares are the arguments — in practice the rest energies $Mc^2$ , $m_1c^2$ , $m_2c^2$ , or $\sqrt{s}$ in place of $Mc^2$ . Physical reading: the four factors are the four triangle inequalities. The first is always positive; the other three vanish on the degenerate configurations. `aq-sr-centre-of-momentum-frame` uses the partially factored $[s-(m_1+m_2)^2c^4][s-(m_1-m_2)^2c^4]$ ; equation (3) is that identity taken all the way. Now the two owned identities. The Gram determinant of the two four-momenta, $G_{ij} = P_i\cdot P_j$ :

\boxed{\;\lambda\big(s,\ m_1^2c^4,\ m_2^2c^4\big) \;=\; -4\,\det G, \qquad \det G = \begin{vmatrix} m_1^2c^4 & P_1\!\cdot\! P_2 \\[2pt] P_1\!\cdot\! P_2 & m_2^2c^4\end{vmatrix}\;} \qquad (4)

where: $s = (P_1+P_2)\cdot(P_1+P_2)$ . Physical reading: $\lambda$ is not an ad-hoc polynomial — it is the invariant volume² of the parallelogram the two four-momenta span, and its sign is the causal character of their plane. No stray factors of $c$ : both sides are energy⁴. And the Heron area of the mass triangle:

\boxed{\;\lambda(a^2,b^2,c^2) \;=\; -16\,A(a,b,c)^2, \qquad A = \sqrt{\sigma(\sigma-a)(\sigma-b)(\sigma-c)},\ \ \sigma = \tfrac12(a{+}b{+}c)\;} \qquad (5)

where: $A$ is Heron's area of a Euclidean triangle with sides $a$ , $b$ , $c$ . Physical reading: a two-body reaction is allowed iff $\lambda \geq 0$ iff $A^2 \leq 0$ iff no such triangle exists. The triangle in the name is a forbidden triangle: when it can be drawn, the physics cannot happen. Finally the payload, recalled from the owner and used throughout:

E_1^{*} = \frac{s + m_1^2c^4 - m_2^2c^4}{2\sqrt{s}}, \qquad p^{*} = \frac{\lambda^{1/2}\big(s,\ m_1^2c^4,\ m_2^2c^4\big)}{2c\sqrt{s}} \qquad (6)

where: the star denotes the centre-of-momentum frame. Physical reading: three masses in, the entire two-body final state out — energies, momentum magnitude, and hence the whole phase-space volume. Only the orientation is left free, and that is the amplitude's business, not the kinematics'. Derived in `aq-sr-centre-of-momentum-frame` (general $s$ ) and `aq-sr-particle-decay-kinematics` ( $\sqrt{s} = Mc^2$ ).

Figure

THE DECAY THAT HAPPENS IS THE ONE WHOSE TRIANGLE FAILS. Both panels lay out three rods whose lengths are the three rest energies, in MeV. LEFT — the ρ(775): the two kaon rods, 493.677 each, total 987.354, and they comfortably over-reach the 775.26 baseline, so they hinge shut into a real isosceles triangle of area 118 503 MeV². A real area means λ = −16A² is NEGATIVE, p is imaginary, and ρ → K⁺K⁻ CANNOT HAPPEN. RIGHT — the φ(1019): the same two kaon rods now fall 32.107 MeV SHORT of the 1019.461 baseline. There is NO triangle, the Heron area is imaginary, λ = −16A² is POSITIVE, and the decay proceeds with p = 126.92 MeV/c. The boundary between the two panels is the degenerate zero-area triangle at exactly 2m_K = 987.354 MeV — which is the threshold, and nothing else is

One determinant, two geometries

Derivation AThe Källén function is a Gram determinant, and its sign is the causal character of the momentum plane

The formula $\lambda = x^2+y^2+z^2-2xy-2yz-2zx$ looks like something a textbook wrote down because it fell out of an algebraic simplification. It is not. It is a determinant — the invariant measuring the parallelogram spanned by the two four-momenta — and once you see that, its sign, its frame-independence and its role as the phase-space volume stop being coincidences and become one fact.

Step 1. Take two four-momenta $P_1$ , $P_2$ and build their Gram matrix, $G_{ij} = P_i \cdot P_j$ — all mutual invariant products. On the mass shell $P_i \cdot P_i = m_i^2c^4$ , so

G = \begin{pmatrix} m_1^2c^4 & P_1\!\cdot\! P_2 \\ P_1\!\cdot\! P_2 & m_2^2c^4 \end{pmatrix}, \qquad \det G = m_1^2m_2^2c^8 - (P_1\!\cdot\! P_2)^2 .

Why: the Gram matrix is built entirely from Lorentz scalars, so $\det G$ is one too — and in any signature it is the squared volume of the parallelogram the vectors span.

Step 2. Express $P_1\cdot P_2$ through $s$ . Expanding $s = (P_1+P_2)\cdot(P_1+P_2)$ and using (1),

s = m_1^2c^4 + m_2^2c^4 + 2\,P_1\!\cdot\! P_2 \qquad\Longrightarrow\qquad P_1\!\cdot\! P_2 = \tfrac12\big(s - m_1^2c^4 - m_2^2c^4\big).

Why: $s$ is the one invariant a two-particle state has beyond its masses, so everything must be expressible through it.

Step 3. Substitute, and compare with $\lambda$ . Writing $y = m_1^2c^4$ , $z = m_2^2c^4$ ,

-4\det G = (s - y - z)^2 - 4yz = s^2 + y^2 + z^2 - 2sy - 2sz - 2yz = \lambda(s, y, z).

\boxed{\;\lambda\big(s,\ m_1^2c^4,\ m_2^2c^4\big) \;=\; -4\det G\;}

Why: exact, with no factors of $c$ left over and no approximation anywhere. $\lambda$ was never a polynomial someone chose; it is the Gram determinant of the process, and that is why it is the same number in every frame.

Step 4. Read the sign. The determinant of the induced metric on the 2-plane $\mathrm{span}(P_1,P_2)$ is $\det G$ , and in signature $(+,-,-,-)$ the induced metric on a plane containing a timelike direction has signature $(+,-)$ , hence negative determinant; a purely spacelike plane has signature $(-,-)$ , hence positive. So

\lambda > 0 \iff \det G < 0 \iff \text{the plane is timelike} \iff \text{a CM frame exists} \iff p^{*} \text{ is real}.

Why: this is the geometric content of "kinematically allowed". For two physical massive particles the plane always contains a timelike direction — $P_1$ itself is timelike — so $\lambda > 0$ automatically for any real initial state. The condition only bites when you ask for a state: can a parent of mass $M$ produce daughters $m_1$ , $m_2$ ? That is $\lambda(M^2c^4, m_1^2c^4, m_2^2c^4) \geq 0$ , and it is not automatic. *Verified in verify_aq-sr-relativistic-collision-kinematics.py: the identity $\lambda(s,m_1^2c^4,m_2^2c^4) = -4\det G$ is proved symbolically in sympy from the definition of $s$ ; the sign correspondence between $\det G$ and the causal character of the plane is checked on explicit timelike, null and spacelike pairs; and dimensional consistency (both sides energy⁴) is asserted.*

Go deeper · Derivation B · 2 minThe same determinant is Heron's area, and therefore the triangle criterion is a theorem

Derivation BThe same determinant is Heron's area, and therefore the triangle criterion is a theorem

`aq-sr-particle-decay-kinematics` records, in a single sentence, that $\lambda < 0$ precisely when $\sqrt{x}$ , $\sqrt{y}$ , $\sqrt{z}$ can form a triangle. That sentence is true, it is the origin of the name, and it is almost never proved. Here it is — and the proof shows it is the same determinant as Step 3 above, in Euclidean rather than Minkowski signature. The triangle is a triangle, not a metaphor.

Step 1. Write Heron's formula, with semiperimeter $\sigma = (a+b+c)/2$ , and clear the halves:

A^2 = \sigma(\sigma-a)(\sigma-b)(\sigma-c) = \tfrac{1}{16}(a{+}b{+}c)(-a{+}b{+}c)(a{-}b{+}c)(a{+}b{-}c).

Why: each factor after the first is a triangle inequality. If any is negative, $A^2 < 0$ and no triangle exists — Heron's formula is already the triangle test, written as a product.

Step 2. Compare with the factorised Källén function, equation (3):

\lambda(a^2,b^2,c^2) = (a{+}b{+}c)(a{+}b{-}c)(a{-}b{+}c)(a{-}b{-}c).

The two products differ in exactly one factor: Heron carries $(-a+b+c)$ where $\lambda$ carries $(a-b-c) = -(-a+b+c)$ . Every other factor is identical. Hence

\boxed{\;\lambda(a^2,b^2,c^2) \;=\; -16\,A(a,b,c)^2\;}

Why: one sign, and the whole thing turns over. The Källén function is Heron's area formula continued past the point where the triangle stops existing. It does not resemble it; it is it, times $-16$ .

Step 3. Now do the Euclidean Gram determinant, and watch it match. For ordinary vectors $\vec u$ , $\vec v$ spanning a triangle, the Gram determinant is $|\vec u|^2|\vec v|^2 - (\vec u \cdot \vec v)^2 = 4A^2$ — four times the squared area, because the parallelogram has twice the area. Compare Step 3 of the previous derivation, $\lambda = -4\det G$ , with Step 2 here, $\lambda = -16A^2$ : they agree iff $\det G = 4A^2$ . They are the same determinant. Why: the Källén function is the Gram determinant of the momentum plane, and the mass triangle is what that determinant would measure if the metric were Euclidean. Minkowski's minus sign is the whole difference between an area and a phase-space volume.

Step 4. State the criterion — now a theorem, not a mnemonic. With $a = Mc^2$ , $b = m_1c^2$ , $c = m_2c^2$ :

\text{the decay } M \to 1+2 \text{ is allowed} \iff \lambda \geq 0 \iff A^2 \leq 0 \iff \textbf{no triangle with sides } M, m_1, m_2 \textbf{ exists} \iff M \geq m_1 + m_2.

Why: the physics is the failure of the triangle inequality, and the threshold is the degenerate triangle. At $M = m_1 + m_2$ the three rods lie flat in a line: zero area, $\lambda = 0$ , $p^{*} = 0$ . Below it the triangle opens up, acquires real area, and the reaction is dead. $\lambda$ is not loosely "measuring how far above threshold you are" — it is exactly minus sixteen times the squared area of the triangle you failed to build. *Verified in verify_aq-sr-relativistic-collision-kinematics.py: the factorisation of equation (3) and the identity $\lambda(a^2,b^2,c^2) = -16A^2$ are both proved symbolically; the Euclidean identity $\det G = 4A^2$ is proved; and the criterion is checked numerically against the $\phi$ , $\rho$ and $\Lambda$ numbers printed in this article.*

Figure

λ IS THE THRESHOLD DETECTOR, AND IT HAS ONLY ONE ZERO. Plot of λ(s, m_K²c⁴, m_K²c⁴) against the parent rest energy √s, for two kaons in the final state. BELOW √s = 2m_K = 987.354 MeV the curve is NEGATIVE — the mass triangle closes, its Heron area is real, and p = √λ/(2c√s) is imaginary: the channel is SHUT. The ρ(775) sits deep in this region, λ = −2.247×10¹¹ MeV⁴. AT exactly 987.354 MeV the curve crosses zero — the degenerate flat triangle, zero area, p = 0. This crossing IS the threshold; there is no other. ABOVE it λ climbs and the φ(1019) sits just clear, λ = +6.697×10¹⁰ MeV⁴, p* = 126.92 MeV/c. The φ is only 32.107 MeV above the line, which is why it is a narrow resonance that decays overwhelmingly into kaons: it has just barely enough mass to break the triangle

Go deeper · Derivation C · 2 minThe invariant-first method: isolate, square, and never boost — with the missing-mass formula as its sharpest use

Derivation CThe invariant-first method: isolate, square, and never boost — with the missing-mass formula as its sharpest use

The two identities above explain why $\lambda$ works; this one is about how to use it. The point is a technique, not a formula: every standard result of collision kinematics — thresholds, decay momenta, the mass of an unobserved recoiling system — is one application of the same move. Put the unknowns on one side, square, and let the mass shell eat them.

Step 1 — the move. Suppose a reaction $1 + 2 \to 3 + X$ , where $X$ is anything at all: one particle, a jet, a shower you never saw. Conservation gives $P_X = P_1 + P_2 - P_3$ . You cannot measure $P_X$ . But you can square it, and its square is by definition the invariant mass of whatever $X$ is:

M_X^2c^4 \;=\; \big(P_1 + P_2 - P_3\big)\cdot\big(P_1 + P_2 - P_3\big).

Why: the left side is a property of the thing you did not detect; the right side is built entirely from things you did. That asymmetry is the whole trick.

Step 2 — expand, in invariants only. Using $P_1 + P_2 = P_{\rm tot}$ with $P_{\rm tot}\cdot P_{\rm tot} = s$ , and $P_3\cdot P_3 = m_3^2c^4$ :

\boxed{\;M_X^2c^4 \;=\; s + m_3^2c^4 - 2\,P_{\rm tot}\!\cdot\! P_3 \;=\; s + m_3^2c^4 - 2\sqrt{s}\,E_3^{*}\;}

where the last form evaluates $P_{\rm tot}\cdot P_3$ in the CM frame, where $P_{\rm tot} = (\sqrt{s},\,\vec 0\,)$ . Why: this is the missing-mass method, and it is how most of the particle zoo was found. Measure one outgoing particle, and the invariant mass of everything you missed drops out. A bump in $M_X$ is a new particle.

Step 3 — the same move gives the threshold. For $1 + 2 \to$ (anything with total rest energy $\sum_f m_f c^2$ ), the products must at minimum be created at rest in the CM frame, so $\sqrt{s_{\min}} = \sum_f m_f c^2$ . Squaring that against the lab expression for $s$ with a stationary target of mass $m_2$ — where $P_1\cdot P_2 = E_1 m_2 c^2$ — gives

s = m_1^2c^4 + m_2^2c^4 + 2 E_1 m_2 c^2 \qquad\Longrightarrow\qquad E_1^{\rm thr} = \frac{\big(\sum_f m_f\big)^2 - m_1^2 - m_2^2}{2m_2}\,c^2 .

Why: one line, and no $\gamma$ appears in it. `aq-sr-threshold-energy` owns this result and works the antiproton case in full; recalled here to show the general formula is a two-line consequence of the invariant-first move, not a separate calculation.

Step 4 — and it gives $p^{*}$ . For $M \to 1 + 2$ : write $P_2 = P - P_1$ , square, use $P\cdot P = M^2c^4$ and $P\cdot P_1 = Mc^2 E_1^{*}$ in the parent rest frame, and $E_1^{*}$ falls out; feeding it back into $p^{*2}c^2 = E_1^{*2} - m_1^2c^4$ produces $\lambda$ , which by equation (5) is minus sixteen times a squared area. The triangle was in the algebra all along. Why: the reason the squaring trick keeps producing $\lambda$ is that squaring is what a Gram determinant is made of — you are computing the volume of the momentum parallelogram every time, without meaning to. *Verified in verify_aq-sr-relativistic-collision-kinematics.py: the missing-mass identity is proved symbolically; the general threshold formula is derived in sympy and checked against the antiproton numbers of aq-sr-threshold-energy ( $\sqrt{s}_{\rm thr} = 4m_pc^2 = 3.753$ GeV, $E_1 = 7m_pc^2 = 6.568$ GeV, $T_1 = 6m_pc^2 = 5.630$ GeV); and the Fermi-motion extension printed below is solved to 50 digits.*

Limiting cases

Case	Result	Meaning
$M = m_1 + m_2$	$\lambda = 0$ , $A = 0$ , $p^{*} = 0$	the degenerate triangle — three rods flat in a line. Threshold, and $\lambda$ 's only zero
$M < m_1 + m_2$	$\lambda < 0$ , $A$ real	the triangle closes and has genuine area. Forbidden
$M > m_1 + m_2$	$\lambda > 0$ , $A$ imaginary	no triangle exists. Allowed, with $p^{*} = \sqrt{\lambda}/(2Mc)$
$m_2 \to 0$	$p^{*} = (M^2-m_1^2)c/(2M)$	one massless daughter; the triangle degenerates to two rods
$m_1 = m_2 = 0$ , or $M \gg m_1, m_2$	$\lambda \to M^4c^8$ , $p^{*} \to Mc/2$	far above threshold the masses stop mattering; phase space saturates
both $P_i$ future-timelike	$\det G < 0$ always	a real two-particle state always has $\lambda > 0$ — you cannot fail to have a CM frame

Common trapThree ways to lose a factor of c, and one way to lose the physics

The units trap. $s$ , $t$ , $u$ can be defined in energy² (this article, and shipped canon) or in momentum². Both appear in the literature and both are self-consistent — but the sum rule reads $s+t+u = \sum_i m_i^2 \mathbf{c^4}$ in the first and $\sum_i m_i^2c^2$ in the second, and $\lambda$ 's arguments change with it. The base article for this page uses both conventions in different equation cards, which is how a factor of $c^2$ goes missing.

The argument trap. $\lambda$ takes squared quantities: $\lambda(s, m_1^2c^4, m_2^2c^4)$ , never $\lambda(\sqrt{s}, m_1c^2, m_2c^2)$ . The triangle inequality lives on the square roots of the arguments, which is exactly why the triangle connection is easy to state and easy to garble.

And the real one: $\lambda \geq 0$ is necessary, not sufficient. It is a kinematic statement and knows nothing about charge, baryon number, angular momentum, or whether an interaction exists to mediate the decay. A triangle that fails to close means the decay is not forbidden by energy and momentum. Everything else still has a vote.

Worked examples

1 · Why the φ decays to kaons and the ρ cannot

Take the $\phi(1019)$ and the $\rho(775)$ , both of which would like to decay into $K^+K^-$ . Decide, using nothing but the mass triangle, which one can — and get the daughter momentum for the one that does.

Step 1 — identify. The kinematic question is $\lambda(M^2c^4, m_K^2c^4, m_K^2c^4) \geq 0$ , which by equation (5) is $A^2 \leq 0$ : does a triangle with sides $Mc^2$ , $m_Kc^2$ , $m_Kc^2$ exist? PDG masses: $m_\phi c^2 = 1019.461$ , $m_\rho c^2 = 775.26$ , $m_{K^\pm}c^2 = 493.677$ MeV.

Step 2 — the $\rho$ . The two kaon sides sum to $2m_Kc^2 = \mathbf{987.354}$ MeV, and $775.26 < 987.354$ . The triangle closes. Heron:

\sigma = \tfrac12(775.26 + 987.354) = 881.307,\qquad A = \sqrt{\sigma(\sigma - 775.26)(\sigma - 493.677)^2} = \mathbf{118\,503\ MeV^2}.

Real area $\Rightarrow \lambda = -16A^2 = \mathbf{-2.2469\times10^{11}\ MeV^4} < 0 \Rightarrow p^{*}$ imaginary. $\rho \to K^+K^-$ is kinematically shut, and the $\rho$ goes to $\pi\pi$ instead.

Step 3 — the $\phi$ . Now $1019.461 > 987.354$ : the kaon rods fall short by $\mathbf{32.107}$ MeV and no triangle can be built, so $A^2 < 0$ and $\lambda > 0$ . By equation (3) with $a = 1019.461$ , $b = c = 493.677$ :

\lambda = (a+2b)\,a\,a\,(a-2b) = a^2\big(a^2 - 4b^2\big) = \mathbf{+6.6965\times10^{10}\ MeV^4}.

Step 4 — the boundary number. Equation (6) with $\sqrt{s} = m_\phi c^2$ :

p^{*} = \frac{\sqrt{\lambda}}{2\,m_\phi c^2} = \frac{258\,776}{2038.92}\ \mathrm{MeV}/c = \mathbf{126.92\ MeV/}c,

against the PDG's tabulated $127$ MeV/ $c$ . Each kaon carries $126.92$ MeV/ $c$ , back-to-back.

Step 5 — the moral. The $\phi$ clears the kaon threshold by $32$ MeV out of $1019$ — three per cent — and that sliver is why $\phi \to K\bar K$ dominates its decays and why the $\phi$ is a narrow resonance at all. Move its mass down by $33$ MeV and the triangle snaps shut, $\lambda$ changes sign, and the particle becomes something else entirely. The whole of that physics is one inequality between three rods.

2 · The Bevatron in one line — and why a bound target beats a free one

Recall the antiproton threshold from `aq-sr-threshold-energy`, reproduce it with the invariant-first move, then extend it: what if the target proton is not at rest, but a nucleon inside a nucleus carrying Fermi momentum?

Step 1 — recall, do not re-derive. For $p + p \to p + p + p + \bar p$ on a free stationary proton, `aq-sr-threshold-energy` owns the result: $\sqrt{s}_{\rm thr} = 4m_pc^2 = \mathbf{3.753}$ GeV, total beam energy $E_1 = 7m_pc^2 = \mathbf{6.568}$ GeV, kinetic $T_1 = 6m_pc^2 = \mathbf{5.630}$ GeV. Step 3 of the third derivation reproduces it: $E_1^{\rm thr} = [(4m_p)^2 - 2m_p^2]c^2/(2m_p) = 7m_pc^2$ . One line, no boost.

Step 2 — now let the target move. Inside a nucleus a nucleon has Fermi momentum, typically $|\vec p_F|c \approx 250$ MeV. Suppose it is moving head-on into the beam. Then $\vec p_1 \cdot \vec p_2 = -|\vec p_1||\vec p_F|$ and equation (2) gives

s = 2m_p^2c^4 + 2\big(E_1 E_2 + |\vec p_1||\vec p_F|c^2\big), \qquad E_2 = \sqrt{|\vec p_F|^2c^2 + m_p^2c^4} = \mathbf{971.007\ MeV}.

The cross term now adds. A head-on target hands the collision extra invariant energy for free.

Step 3 — set the threshold and solve. Impose $s = (4m_pc^2)^2 = 16m_p^2c^4$ . The $2m_p^2c^4$ cancels down and, with $|\vec p_1|c = \sqrt{E_1^2 - m_p^2c^4}$ and $m_pc^2 = 938.272$ MeV, the threshold condition is

E_1E_2 + |\vec p_1|c\,|\vec p_F|c = 7\,m_p^2c^4 \quad\Longrightarrow\quad 971.007\,E_1 + 250\sqrt{E_1^2 - m_p^2c^4} = \mathbf{6\,162\,481\ MeV^2},

a single scalar equation with the root $E_1^{\rm thr} = \mathbf{5065.0\ MeV}$ .

Step 4 — the boundary number. Kinetic threshold $T_1 = 5065.0 - 938.3 = \mathbf{4126.7\ MeV} = \mathbf{4.127}$ GeV, against $5.630$ GeV on a free proton. Fermi motion lowers the antiproton threshold by $\mathbf{1503}$ MeV — a full $1.5$ GeV, or $27\%$ .

Step 5 — the moral. This is subthreshold production, and it is why heavy-ion and proton–nucleus experiments see particles at beam energies a free-target calculation forbids. Nothing exotic happened: we wrote $s$ , we squared, we never chose a frame. The target's internal motion enters as one dot product, and one dot product is all it takes.

Physics of the portrait

Origins · 2 minKällén function and how this was discovered

Källén function

1926–1968 · Gunnar Källén — the triangle function, the spectral representation, and a Piper Cherokee over Hanover

Gunnar Källén was born in Kristianstad, Sweden, on 13 February 1926, took his doctorate at Lund in 1950 under Torsten Gustafson, and spent five formative years (1952–1957) in CERN's theory division — which in those days was not in Geneva at all but housed at Niels Bohr's institute in Copenhagen. He was a ferociously exact man, the kind of theorist who would not let a renormalisation argument past unless every limit had been taken honestly, and who said so, loudly, to people senior to him. His enduring contribution is the Källén–Lehmann spectral representation (1952), which writes an exact interacting propagator as a positive-weighted superposition of free ones and thereby makes non-perturbative statements possible about a theory nobody can solve. The function that carries his name is smaller and more useful. It appears in his 1964 textbook Elementary Particle Physics as the combination governing two-body phase space, and generations of experimentalists have used it without asking why it is called the triangle function. The answer — that it is Heron's area formula, proved in the Metrica around 60 CE, analytically continued into Minkowski signature, and equally the Gram determinant of the momentum plane — is a fair reflection of the man: an exact identity, hiding under a formula everyone was using approximately. Källén learned to fly in 1964. On 13 October 1968, aged 42, he was flying his Piper PA-28 Cherokee Arrow from Malmö to CERN when the aircraft came down during an emergency landing at Hanover. His wife and his other passenger survived. He did not.

History

1883

— Jorgen Pedersen Gram, a Danish actuary and mathematician, publishes the determinant of mutual inner products that now carries his name. He dies in 1916, struck by a cyclist on his way to a meeting of the Royal Danish Academy.

1926

— Gunnar Kallen is born in Kristianstad, Sweden, on 13 February.

1952

— Kallen, at CERN's theory division (then housed at Niels Bohr's institute in Copenhagen), publishes the spectral representation of the exact propagator; Harry Lehmann finds it independently in 1954.

1955

— Chamberlain, Segre, Wiegand and Ypsilantis observe the antiproton at the Berkeley Bevatron: "Observation of Antiprotons", Physical Review 100, 947. Chamberlain and Segre share the 1959 Nobel Prize.

1958

— Stanley Mandelstam introduces the invariants s, t and u.

1964

— Kallen's Elementary Particle Physics presents the triangle function lambda as the standard two-body phase-space factor. He also takes his first flying lessons.

1968

— Kallen dies on 13 October, aged 42, when his Piper Cherokee comes down during an emergency landing at Hanover. His wife and his other passenger survive.

Going further

Texts. Byckling & Kajantie, Particle Kinematics (Wiley, 1973) — the definitive treatment of $\lambda$ , Gram determinants and physical regions · Halzen & Martin, Quarks and Leptons, §4 · G. Källén, Elementary Particle Physics (Addison-Wesley, 1964). Primary: O. Chamberlain, E. Segrè, C. Wiegand and T. Ypsilantis, "Observation of Antiprotons", Phys. Rev. 100, 947 (1955); G. Källén, Helv. Phys. Acta 25, 417 (1952); Heron of Alexandria, Metrica I.8.

Provenance. Signature $(+,-,-,-)$ , $P^\mu = (E, \vec p\,c)$ , $P\cdot P = +m^2c^4$ ; $s$ , $t$ , $u$ in energy², matching the shipped owner `aq-sr-mandelstam-invariants`. Base defect corrected: the base for this page is self-contradictory about units — eq[2] and eq[3] put $s$ in momentum² and eq[8] prints the sum rule as $\sum_i m_i^2 c^2$ , while eq[5] ( $\sqrt{s_{\min}} = \sum_f m_f c^2$ ) and eq[9] use energy². Both cannot be right; the shipped owner settles it as energy². The base's eq[2] is additionally garbled — it reads $s = E_{\rm cm}^2/c^2 \cdot c^2 = E_{\rm cm}^2/c^2$ , multiplying by $c^2$ and then asserting equality with the unmultiplied quantity. The base's Bevatron derivation (DERIV[0]) is CORRECT and was not "fixed": $7m_pc^2$ total, $6m_pc^2 = 5.63$ GeV kinetic, exactly as `aq-sr-threshold-energy` ships it. Its DERIV[2] (fixed-target vs collider scaling) is serial 1984's payload and is deliberately not developed here. All numbers above are asserted in verify_aq-sr-relativistic-collision-kinematics.py.

Builds on. `aq-sr-centre-of-momentum-frame` · `aq-sr-particle-decay-kinematics` · `aq-sr-threshold-energy` · `aq-sr-center-of-momentum-invariant-mass`. Leads to. `aq-sr-mandelstam-variables` · `aq-sr-bmt-equation` · `aq-sr-wigner-rotation`.

Why does a triangle that CAN be drawn mean the decay CANNOT happen? That feels backwards.

It feels backwards because the name "triangle function" invites you to imagine the allowed process as the triangle. It is the opposite. A parent can only decay into daughters it can pay for: $M \geq m_1 + m_2$ . But that is precisely the violation of the triangle inequality for sides $M$ , $m_1$ , $m_2$ — the long side is at least as long as the other two together, so they cannot reach around to meet it. Threshold, $M = m_1 + m_2$ , is the degenerate case where the three rods lie flat in a line: zero area, $\lambda = 0$ . Allowed decays live on the far side of the triangle inequality, where the area has gone imaginary.

What is the Gram determinant actually doing for me?

It tells you the causal character of the plane your momenta span, and that is the whole kinematic question. In $(+,-,-,-)$ , a 2-plane containing a timelike direction has induced signature $(+,-)$ and hence negative determinant; a purely spacelike plane has $(-,-)$ and positive. Since $\lambda = -4\det G$ , positive $\lambda$ means a timelike plane, hence a centre-of-momentum frame exists, hence $p^{*}$ is real. It also explains, with no calculation, why $\lambda$ is frame-independent: a Gram matrix is built from dot products, and dot products are scalars.

Which units convention should I use for s, t and u?

Use energy²: $P^\mu = (E, \vec p\,c)$ , $s = (P_1+P_2)^2$ , $\sqrt{s}$ an energy, $s+t+u = \sum_i m_i^2c^4$ . That is what the shipped owner `aq-sr-mandelstam-invariants` boxes and what every experimental paper means when it quotes $\sqrt{s} = 13$ TeV. The momentum²-units convention ( $s = (\sum E)^2/c^2 - |\sum\vec p|^2$ , sum rule $\sum m_i^2c^2$ ) is equally self-consistent and appears in older texts; the base article for this page unhappily uses both. The only fatal error is mixing them, which costs a factor of $c^2$ and is very hard to see afterwards.

Does λ ≥ 0 guarantee the decay actually occurs?

No. It guarantees only that energy and momentum conservation do not forbid it. Charge, baryon number, lepton number, angular momentum and parity all still have a vote, and even with every conservation law satisfied there must be an interaction mediating the process with a non-zero matrix element. $\lambda$ is a kinematic gate: when it is shut nothing else matters, but its being open is permission to ask the dynamical question, not an answer to it.

Why does the invariant-first method work so much better than boosting?

Because a boost carries four components and an angle, and every one of them is a place to drop a sign; a dot product carries one number and no orientation. The method's power is that the unknowns are exactly the things a square destroys. You never know the emission angle in a decay, and squaring $P_2 = P - P_1$ removes it in one line. You never see the recoiling system in a missing-mass measurement, and squaring $P_X = P_1 + P_2 - P_3$ removes it in one line. The frame you would have boosted to is a frame you never needed.

What happens with three particles in the final state?

$\lambda$ stops being sufficient. A two-body decay has zero free invariants once the masses are fixed — $p^{*}$ is determined outright — so the answer is one number and one triangle. Three bodies have two free invariants, and the allowed region becomes a two-dimensional area in the plane of those invariants: the Dalitz plot. Its boundary is again a determinant condition, now a $3\times 3$ Gram determinant, whose vanishing says the three momenta have become linearly dependent. The geometry generalises exactly; only the dimension goes up.