The CKM matrix

Nature built its quarks in two guises that refuse to line up. The states with definite mass — the ones that hadronise into the pions, kaons and B mesons an experiment actually sees — are not the states the weak force couples to. When a $W$ boson reaches into the quark sector it does not grab a $d$ , an $s$ , or a $b$ ; it grabs a fixed quantum superposition of all three. The three-by-three unitary matrix that spells out that superposition is the Cabibbo–Kobayashi–Maskawa matrix, $V_{\rm CKM}$ , and it is the single object that sets the strength of every quark charged-current transition in the Standard Model. Two facts make it more than bookkeeping. First, the matrix is strongly hierarchical: transitions within a generation are near-unit strength, one step across costs a factor of the Cabibbo parameter $\lambda\approx0.22$ , and the corner elements $V_{ub}$ and $V_{td}$ are down by $\lambda^3\sim0.004$ . This near-diagonal shape is why the top quark decays almost exclusively to bottom and why strangeness-changing decays are slow. Second — the deep fact, the one that won Kobayashi and Maskawa the 2008 Nobel Prize — a three-generation unitary matrix carries one irreducible complex phase that no rephasing of the quark fields can remove. That single phase is the only source of CP violation among quarks in the Standard Model, which is why in 1973, with only three quarks known, its authors were really predicting a third generation. This article builds the matrix, counts its parameters, derives why three generations are the minimum for CP violation, and puts real measured numbers on all nine elements.

Key takeaways

The CKM matrix is the unitary $3\times3$ matrix relating the weak (primed) down-type states to the mass eigenstates, $(d',s',b')^{T}=V_{\rm CKM}(d,s,b)^{T}$ — the three-generation heir of the Cabibbo angle.
Every quark charged-current vertex $u_i\!\to\!d_j$ carries the factor $-i\frac{g_w}{2\sqrt2}V_{ij}\gamma^\mu(1-\gamma^5)$ ; the antiquark process carries $V_{ij}^{*}$ .
Unitarity ( $V^\dagger V=\mathbb 1$ ) leaves exactly three mixing angles and one complex phase $\delta$ — and that phase is the sole Standard-Model source of quark CP violation.
The measured magnitudes are near-diagonal and hierarchical: $|V_{ud}|\approx0.974$ , $|V_{us}|\approx0.225$ , $|V_{cb}|\approx0.041$ , $|V_{ub}|\approx0.0037$ , $|V_{tb}|\approx0.999$ .
CP violation is quantified by the rephasing-invariant Jarlskog $J\approx3.1\times10^{-5}$ , equal to twice the area of every unitarity triangle.
Key idea: the weak interaction does not respect the quark mass basis — each up-type quark couples to a superposition of down-type mass states through a single unitary matrix whose one irreducible complex phase is the only Standard-Model source of CP violation among quarks.

The core insight

The weak interaction does not respect the quark mass basis: when a $W$ couples to quarks it connects each up-type quark to a quantum superposition of down-type mass states, encoded in one unitary matrix whose single irreducible complex phase is the only Standard-Model source of CP violation among quarks.

Where this lives today

The CKM matrix is now the most over-constrained object in flavour physics, and the sport is to test its unitarity to breaking point. Global fits by the CKMfitter and UTfit groups overlay dozens of independent measurements — kaon, $D$ , $B_d$ and $B_s$ decays and mixings — on the single apex $(\bar\rho,\bar\eta)$ , and they still intersect at one point, a triumph for the three-generation Standard Model. But cracks are being probed. The LHCb experiment at CERN and Belle II at KEK are pushing the angle $\gamma$ and the mixing phases to sub-degree precision; a long-standing $\sim3\sigma$ tension between the exclusive ( $B\to D^{*}\ell\nu$ ) and inclusive ( $b\to c\ell\nu$ ) determinations of $|V_{cb}|$ — and a parallel one for $|V_{ub}|$ — refuses to resolve. Meanwhile the first-row "Cabibbo angle anomaly" of Worked Example 1 sits at roughly $3\sigma$ below unity, hinting at either underestimated theory uncertainties or new physics in the very cleanest weak decays. The elements $V_{td}$ and $V_{ts}$ , unreachable in tree decays because the top quark decays before it hadronises, are extracted from $B_d$ and $B_s$ oscillation frequencies — loop (box) diagrams whose measured $\Delta m_s=17.77\,\mathrm{ps}^{-1}$ turns a mixing rate into a matrix element. And looming over all of it is a cosmological verdict: the Jarlskog $J\approx3\times10^{-5}$ delivers CP violation some ten orders of magnitude too feeble to have generated the observed excess of matter over antimatter, so the search for CP-violating physics beyond the CKM matrix — in the lepton sector's PMNS matrix (aq-pp-three-flavour-pmns), in electric dipole moments, in rare kaon decays — is among the central quests of the field.

Ask the tutor: Ask me to count the CKM parameters for any number of generations, or to draw the unitarity triangle and turn its area into the Jarlskog measure of CP violation.

The physical picture

The whole matrix lives in one vertex. Send a $u$ quark into a $W$ and what comes out is not a $d$ but the weak partner $d'=V_{ud}d+V_{us}s+V_{ub}b$ — a superposition weighted by the first row of $V_{\rm CKM}$ . Because the experiment detects hadrons, and hadrons are built from mass eigenstates, that single vertex fans into three physically distinct outcomes with amplitudes $V_{ud}$ , $V_{us}$ , $V_{ub}$ . The first figure draws that fan with its real weights and then lays the nine magnitudes out as a grid, so the hierarchy — near-unit diagonal, small off-diagonal, tiny corners — is visible at a glance.

Figure

A single u→W vertex couples not to a d quark but to the superposition d′ = V_ud d + V_us s + V_ub b. Line thickness tracks the amplitude: V_ud = 0.974 dominates, V_us = 0.225 is the Cabibbo-suppressed step across one generation, and V_ub = 0.0037 is the doubly-suppressed corner. The grid at right shows all nine magnitudes — a near-diagonal, strongly hierarchical matrix with powers of λ ≈ 0.22

The grid is nearly the identity, but not quite — and the tiny departures are where all the physics is. The off-diagonal elements are what let strangeness and beauty decay at all; without them the $s$ and $b$ quarks would be stable and the universe would look very different. More subtly, the two corner elements $V_{ub}$ and $V_{td}$ carry the irreducible complex phase, and a complex amplitude is the seed of CP violation. Unitarity forces the nine elements to satisfy six orthogonality relations, each a sum of three complex numbers that closes to zero — a triangle in the complex plane. The second figure is the one every flavour physicist draws: the " $db$ " unitarity triangle, whose non-zero area is the geometric signature that CP is violated, with the apex fixed by the measured Wolfenstein coordinates $(\bar\rho,\bar\eta)$ .

Figure

The db unitarity triangle: V_ud V_ub + V_cd V_cb + V_td V*_tb = 0, normalised so the base runs from 0 to 1. The apex sits at the measured (ρ̄, η̄) ≈ (0.16, 0.35); the interior angles are α ≈ 93°, β ≈ 22°, γ ≈ 66° and sum to 180°. The measured sin2β = 0.699 fixes β, and the triangle area equals J/2 with Jarlskog J ≈ 3.1×10⁻⁵ — a non-zero area is exactly the statement that CP is violated

The core equations

The definition is a change of basis. The down-type quark fields the weak interaction couples to — call them primed — are a unitary rotation of the mass-eigenstate fields, and the rotation matrix is the CKM matrix. By convention the entire misalignment between the up and down sectors is loaded onto the down-type quarks.

\begin{pmatrix} d' \\ s' \\ b' \end{pmatrix} = V_{\rm CKM}\begin{pmatrix} d \\ s \\ b \end{pmatrix},\qquad V_{\rm CKM}=\begin{pmatrix} V_{ud} & V_{us} & V_{ub} \\ V_{cd} & V_{cs} & V_{cb} \\ V_{td} & V_{ts} & V_{tb} \end{pmatrix} \qquad (1)

where: the unprimed $(d,s,b)$ are mass eigenstates (definite mass, they hadronise); the primed $(d',s',b')$ are the weak eigenstates the $W$ couples to; $V_{ij}$ is the amplitude for up-type $i\in\{u,c,t\}$ to convert to down-type $j\in\{d,s,b\}$ . Reading it: the first row tells you the three down-type states a $u$ quark connects to, weighted by $V_{ud},V_{us},V_{ub}$ . That rotation enters physics through the quark charged-current vertex. It is the vertex of aq-pp-weak-charged-current — the same $g_w$ and the same parity-violating $(1-\gamma^5)$ — with one addition: a CKM factor multiplying every up–down pair.

\text{vertex}(u_i\to d_j) = \frac{-ig_w}{2\sqrt2}\,V_{ij}\,\gamma^\mu(1-\gamma^5),\qquad \text{vertex}(d_j\to u_i)=\frac{-ig_w}{2\sqrt2}\,V_{ij}^{*}\,\gamma^\mu(1-\gamma^5) \qquad (2)

where: $g_w\approx0.653$ is the weak coupling; $\tfrac1{2\sqrt2}$ is the standard normalisation; $\gamma^\mu(1-\gamma^5)$ is the V−A structure (recalled from aq-pp-v-minus-a). The complex conjugate $V_{ij}^{*}$ appears when the down-type quark is created rather than destroyed — this is why the phase of $V_{ij}$ flips between a process and its CP conjugate, the origin of all CKM CP violation. Unitarity is not assumed for elegance; it follows because $V_{\rm CKM}$ is a rotation between two orthonormal bases. It gives twelve constraints — three normalisations, three off-diagonal relations, and their conjugates.

V^\dagger V = V V^\dagger = \mathbb 1 \;\Longrightarrow\; \sum_k |V_{ik}|^2=1\ \ (\text{rows}),\qquad \sum_k V_{ik}V_{jk}^{*}=0\ \ (i\neq j) \qquad (3)

where: the normalisation $\sum_k|V_{ik}|^2=1$ says each up-type quark's total coupling is fixed (a $u$ decays somewhere); the orthogonality $\sum_k V_{ik}V_{jk}^{*}=0$ is a sum of three complex numbers closing to zero — a unitarity triangle. These relations tie poorly-measured elements to well-measured ones and are themselves a falsifiable test of exactly three generations. Unitarity plus the freedom to rephase the quark fields collapses the nine complex entries to just four real numbers. The Particle Data Group standard parametrisation uses three angles $\theta_{12},\theta_{23},\theta_{13}$ and one phase $\delta$ , with $c_{ij}=\cos\theta_{ij}$ , $s_{ij}=\sin\theta_{ij}$ .

V=\begin{pmatrix} c_{12}c_{13} & s_{12}c_{13} & s_{13}e^{-i\delta} \\ -s_{12}c_{23}-c_{12}s_{23}s_{13}e^{i\delta} & c_{12}c_{23}-s_{12}s_{23}s_{13}e^{i\delta} & s_{23}c_{13} \\ s_{12}s_{23}-c_{12}c_{23}s_{13}e^{i\delta} & -c_{12}s_{23}-s_{12}c_{23}s_{13}e^{i\delta} & c_{23}c_{13} \end{pmatrix} \qquad (4)

where: $\theta_{12}\approx13.0°$ is the Cabibbo angle; $\theta_{23}\approx2.4°$ , $\theta_{13}\approx0.21°$ ; $\delta\approx66°$ is the CP-violating phase, riding on $s_{13}$ in the corners $V_{ub},V_{td}$ . Read the structure: set $\delta=0$ and every entry is real — CP is conserved; the phase lives where the smallest angle lives, which is why CP violation is small. Because the angles are so hierarchical, it is far more transparent to expand in $\lambda=\sin\theta_{12}\approx0.225$ . Wolfenstein's four parameters $\lambda,A,\rho,\eta$ make the powers of $\lambda$ manifest.

V=\begin{pmatrix} 1-\tfrac12\lambda^2 & \lambda & A\lambda^3(\rho-i\eta) \\ -\lambda & 1-\tfrac12\lambda^2 & A\lambda^2 \\ A\lambda^3(1-\rho-i\eta) & -A\lambda^2 & 1 \end{pmatrix}+\mathcal O(\lambda^4) \qquad (5)

where: $\lambda=0.225$ , $A\approx0.83$ , and $(\rho,\eta)\approx(0.16,0.35)$ ; $\eta$ is the imaginary (CP-violating) part, appearing only in the corners. Read the hierarchy directly: diagonal $\mathcal O(1)$ , the Cabibbo step $\mathcal O(\lambda)$ , the $2\!\leftrightarrow\!3$ mixing $\mathcal O(\lambda^2)$ , the corners $\mathcal O(\lambda^3)$ — a factor $\sim0.22$ per generation crossed. Finally, the amount of CP violation is captured by a single rephasing-invariant number, the Jarlskog invariant $J$ , defined so it does not depend on any arbitrary quark-phase convention.

\mathrm{Im}\!\left(V_{ij}V_{kl}V_{il}^{*}V_{kj}^{*}\right)=J\sum_{m,n}\varepsilon_{ikm}\,\varepsilon_{jln},\qquad J=c_{12}c_{23}c_{13}^2\,s_{12}s_{23}s_{13}\sin\delta\approx A^2\lambda^6\eta \qquad (6)

where: $J\approx3.1\times10^{-5}$ is twice the area of every unitarity triangle; $\varepsilon$ is the Levi-Civita symbol. Read it: $J\neq0$ requires all three angles non-zero and $\sin\delta\neq0$ — CP violation switches off if any mixing angle vanishes or the phase is trivial, which is the algebraic heart of the next derivation.

Derivation ACounting the parameters and why three generations are needed for CP violation

The goal is to prove the central structural fact of the CKM matrix: an $N$ -generation unitary mixing matrix, after every allowed rephasing of the quark fields, retains $\tfrac12N(N-1)$ real mixing angles and $\tfrac12(N-1)(N-2)$ physical phases — so that $N=2$ (Cabibbo) has one angle and zero phases and cannot violate CP, while $N=3$ has three angles and exactly one phase, the minimum that permits CP violation. This is precisely the argument Kobayashi and Maskawa made in 1973, and it is why a complex CKM phase is equivalent to the existence of a third generation. We count the free parameters of a general unitary matrix, subtract the phases that are pure convention, and read off what physics is forced to keep.

Step 1. Count the parameters of a general $N\times N$ unitary matrix. Unitarity $V^\dagger V=\mathbb 1$ is $N^2$ real conditions on the $2N^2$ real numbers in a complex matrix, leaving $N^2$ real parameters. Split these the way a real orthogonal matrix would: an $O(N)$ rotation has $\tfrac12N(N-1)$ angles, so the remaining $N^2-\tfrac12N(N-1)=\tfrac12N(N+1)$ parameters are phases. Why: separating "rotation-like" angles from "phase-like" parameters is the whole game — only the phases can be complex, and only complex entries can violate CP.

Step 2. Remove the phases that are unphysical. The matrix appears in the Lagrangian sandwiched as $\bar u_i\,V_{ij}\,d_j$ , so rephasing $u_i\to e^{i\alpha_i}u_i$ and $d_j\to e^{i\beta_j}d_j$ sends $V_{ij}\to e^{i(\beta_j-\alpha_i)}V_{ij}$ . There are $2N$ such phases, but one overall common phase leaves every $V_{ij}$ unchanged, so $2N-1$ relative phases can be used to kill entries of $V$ . Why: a phase you can rotate away by redefining a quark field is not observable — it carries no physics, so it must be subtracted from the count.

Step 3. Subtract to get the physical content. Physical angles: $\tfrac12N(N-1)$ (rephasing cannot touch these). Physical phases: $\tfrac12N(N+1)-(2N-1)=\tfrac12(N-1)(N-2)$ . For $N=2$ : one angle, $\tfrac12(1)(0)=0$ phases — the Cabibbo matrix is purely real. For $N=3$ : three angles and $\tfrac12(2)(1)=1$ phase. Why: the formula $\tfrac12(N-1)(N-2)$ is the punchline — it is zero for one and two generations and first becomes non-zero at three, so CP violation in the quark sector demands at least three generations.

Step 4. Connect the surviving phase to CP violation. Under CP the vertex factor $V_{ij}$ is replaced by $V_{ij}^{*}$ (equation 2). If $V$ can be made entirely real by rephasing — as for $N\le2$ — then $V=V^{*}$ and every process equals its CP conjugate. Only the irreducible $N=3$ phase makes $V\neq V^{*}$ unavoidably, so amplitudes and their CP mirrors can differ. Why: this closes the logic — an unremovable complex phase is not merely allowed at three generations, it is the one and only way the Standard Model breaks the quark CP symmetry, and its physical measure is the Jarlskog $J$ . *Verified: physical angles $\tfrac12N(N-1)$ and phases $\tfrac12(N-1)(N-2)$ give $(1,0)$ for $N=2$ and $(3,1)$ for $N=3$ ; $J=0$ whenever any $s_{ij}=0$ or $\sin\delta=0$ .*

Go deeper · Derivation B · 2 minThe unitarity triangle and the Jarlskog measure of CP violation

Derivation BThe unitarity triangle and the Jarlskog measure of CP violation

The aim is to turn the abstract orthogonality relation $\sum_k V_{ik}V_{jk}^{*}=0$ into the geometric object flavour physics actually measures — a triangle in the complex plane whose area is a convention-independent measure of CP violation — and to show that this area equals $J/2$ with $J$ the Jarlskog invariant. We pick the relation whose three terms are comparable in size, plot it as three vectors that must close, read the CP-violating content off its area, and confirm that a flat (zero-area) triangle is exactly a CP-conserving theory. This is why "measuring the angles of the unitarity triangle" and "measuring CP violation" are the same experimental programme.

Step 1. Write down the relevant relation. Orthogonality of the first and third columns of $V$ gives $V_{ud}V_{ub}^{*}+V_{cd}V_{cb}^{*}+V_{td}V_{tb}^{*}=0$ . In Wolfenstein counting each term is $\mathcal O(A\lambda^3)$ , so all three legs are comparable — this " $db$ " triangle is not squashed, unlike the others whose legs differ by powers of $\lambda$ . Why: only a triangle with sides of similar length has large angles that can be measured precisely, which is why the $B_d$ system, governed by this relation, is where CP violation was first seen cleanly.

Step 2. Read it as vectors. Each product $V_{ud}V_{ub}^{*}$ etc. is a complex number — a vector in the plane; the relation says the three vectors placed head to tail return to the origin, forming a closed triangle. Divide through by $V_{cd}V_{cb}^{*}$ to fix one side along the real axis from $(0,0)$ to $(1,0)$ ; the apex then lands at $(\bar\rho,\bar\eta)$ . Why: normalising removes the arbitrary orientation and scale, leaving a shape whose only invariants — its angles and area — are genuine physics.

Step 3. Compute the area. A triangle with vertices $0$ , $z_1$ , $z_1+z_2$ in the complex plane has area $\tfrac12|\mathrm{Im}(z_1^{*}z_2)|$ . Applying this to the un-normalised triangle, the area is $\tfrac12|\mathrm{Im}(V_{ud}V_{ub}^{*}V_{cd}^{*}V_{cb})|=\tfrac12 J$ , and the same $J$ emerges from any of the six triangles — they are all different shapes with the same area. Why: that every triangle shares one area is the signal that $J$ , not any individual angle, is the rephasing-invariant amount of CP violation.

Step 4. Take the CP-conserving limit. If the theory conserved CP, $V$ could be made real, every product $V_{ik}V_{jk}^{*}$ would be real, all vectors would lie on a line, and the triangle would collapse to zero area — $J=0$ . A non-degenerate triangle with finite area is therefore a direct picture of broken CP. Why: this is the geometric restatement of Derivation A's result — non-zero area, non-zero phase, and CP violation are one and the same, now something an experiment can literally draw. *Verified: triangle area $=\tfrac12 J$ ; with $\theta_{12,23,13}=(13.0°,2.40°,0.21°)$ and $\delta=66°$ , $J=c_{12}c_{23}c_{13}^2 s_{12}s_{23}s_{13}\sin\delta=3.08\times10^{-5}\approx A^2\lambda^6\eta$ ; area $\to0$ as $\delta\to0$ .*

Go deeper · Derivation C · 2 minThe Wolfenstein hierarchy from the Cabibbo angle

Derivation CThe Wolfenstein hierarchy from the Cabibbo angle

The goal is to show how the entire nine-element matrix organises itself as a power series in a single small number, $\lambda=\sin\theta_{12}\approx0.225$ , so that the near-diagonal structure and the pattern of suppressions are consequences of one measured quantity — the Cabibbo angle — rather than nine independent accidents. We start from the exact standard parametrisation, feed in the empirical fact that the three angles fall off in powers of $\lambda$ , and expand to third order, recovering equation (5) and its immediate predictions like $|V_{ub}/V_{cb}|\sim\lambda$ . This is why physicists quote four Wolfenstein numbers instead of nine magnitudes and a phase.

Step 1. Encode the empirical hierarchy. Experiment gives $s_{12}=\lambda$ , and the higher angles scale as $s_{23}=A\lambda^2$ and $s_{13}e^{-i\delta}=A\lambda^3(\rho-i\eta)$ , defining $A,\rho,\eta$ as $\mathcal O(1)$ numbers. The single physical input is that each successive mixing angle is smaller by a factor of $\lambda$ . Why: promoting "the angles happen to be small and ordered" into explicit powers of one parameter is what makes every element's size predictable at a glance.

Step 2. Expand the diagonal. With $s_{12}=\lambda$ , $c_{12}=\sqrt{1-\lambda^2}=1-\tfrac12\lambda^2+\mathcal O(\lambda^4)$ , and $c_{13},c_{23}=1+\mathcal O(\lambda^4)$ , the $1$ – $1$ and $2$ – $2$ entries become $c_{12}c_{13}=1-\tfrac12\lambda^2$ . The diagonal is not exactly one — it is depleted at order $\lambda^2$ by the probability leaking into the off-diagonal channels. Why: the $-\tfrac12\lambda^2$ is unitarity at work — what leaves the diagonal as $|V_{us}|^2=\lambda^2$ must be subtracted from $|V_{ud}|^2$ , and the square root gives the $\tfrac12$ .

Step 3. Fill in the corners and check a prediction. To $\mathcal O(\lambda^3)$ the corners are $V_{ub}=A\lambda^3(\rho-i\eta)$ and $V_{td}=A\lambda^3(1-\rho-i\eta)$ , while $V_{cb}=A\lambda^2$ . Their ratio is $|V_{ub}/V_{cb}|=\lambda\sqrt{\rho^2+\eta^2}\approx0.225\times0.39\approx0.088$ , a clean falsifiable consequence. Why: the Wolfenstein form does not just tidy notation — it converts the raw hierarchy into ratios that different experiments can check against each other, closing the triangle independently.

Step 4. Locate the CP violation. Only the corners carry an imaginary part, proportional to $\eta$ ; to leading order CP violation enters at $\mathcal O(\lambda^3)$ and is doubly protected — suppressed both by the corner smallness and by the requirement $\eta\neq0$ . Why: this explains at a glance why Standard-Model CP violation is minute ( $J\sim\lambda^6$ ) even though the phase $\delta\approx66°$ is of order one — the phase is large but it multiplies the tiniest elements. *Verified: $c_{12}=1-\tfrac12\lambda^2$ to $\mathcal O(\lambda^4)$ ; with $\lambda=0.225,A=0.826$ , $|V_{cb}|=A\lambda^2=0.0418$ and $|V_{ub}/V_{cb}|=\lambda\sqrt{\rho^2+\eta^2}=0.088$ , both matching the measured magnitudes.*

Limiting cases

Case	Result	Meaning
One generation	$V=1$	no mixing possible; $u\!\to\!d$ only, no CP violation
Two generations	one angle $\theta_c$ , no phase	Cabibbo matrix, real orthogonal; CP conserved by theorem
Three generations	3 angles $+$ 1 phase	minimal CP violation; the physical CKM matrix
No mixing $\theta_{ij}\to0$	$V\to\mathbb 1$	generations decouple; strangeness and beauty exactly conserved
CP-conserving limit $\delta\to0$	$V$ real, $J\to0$	triangle collapses to a line; no matter–antimatter difference
Degenerate up- or down-quark masses	$J\to0$	a phase becomes rephasable; two equal quarks kill CP violation
Leading Wolfenstein $\lambda\to0$	$V\to\mathbb 1$	all mixing scales as powers of the Cabibbo parameter

Common trapA large CP phase must mean large CP violation

It does not — and this is the most instructive number in the whole subject. The Standard-Model CP phase is nearly maximal, $\delta\approx66°$ , of order one radian; yet the observed CP violation is minuscule. The reason is that the phase never acts alone: the physical, convention-independent measure is the Jarlskog invariant $J=c_{12}c_{23}c_{13}^2\,s_{12}s_{23}s_{13}\sin\delta\approx A^2\lambda^6\eta\approx3\times10^{-5}$ . The six powers of $\lambda\approx0.22$ — three small mixing angles multiplied together — crush a large phase into a tiny effect. Turn any single mixing angle off and $J$ vanishes no matter how large $\delta$ is; make any two up-type or two down-type quarks degenerate in mass and $J$ vanishes again. So CP violation is not measured by $\delta$ but by $J$ , and the smallness of $J$ — not of $\delta$ — is why the Kobayashi–Maskawa mechanism falls short of explaining the cosmic matter–antimatter asymmetry by some ten orders of magnitude, one of the sharpest hints that CP-violating physics beyond the CKM matrix must exist. (A related trap: the CKM matrix does not produce flavour-changing neutral currents at tree level — its unitarity, through the GIM mechanism of aq-pp-gim-mechanism, guarantees the neutral current stays flavour-diagonal.)

Worked examples

1 · Testing first-row unitarity and extracting the Wolfenstein parameters (fully worked)

From the directly measured magnitudes $|V_{ud}|=0.97373$ (superallowed nuclear $\beta$ decay), $|V_{us}|=0.2243$ (kaon $K_{\ell3}$ decay), and $|V_{ub}|=0.00382$ (semileptonic $B$ decay), test the unitarity relation $|V_{ud}|^2+|V_{us}|^2+|V_{ub}|^2=1$ and extract $\lambda$ . Then use $|V_{cb}|=0.0410$ to get $A$ .

Step 1. Square and sum the first row. $|V_{ud}|^2=0.97373^2=0.94815$ ; $|V_{us}|^2=0.2243^2=0.050310$ ; $|V_{ub}|^2=0.00382^2=0.0000146$ . Adding, $0.94815+0.05031+0.0000146=\mathbf{0.99847}$ . Unitarity predicts exactly $1$ , so the row closes to $0.15\%$ .

Step 2. Read the result honestly. The deficit is $1-0.99847=1.53\times10^{-3}$ . With the tiny modern uncertainties on $|V_{ud}|$ and $|V_{us}|$ this sits about $3\sigma$ below unity — the live "Cabibbo angle anomaly." At the level of a first pass, the row is unitary to a fraction of a percent, a genuine confirmation of three-generation structure; the residual $3\sigma$ is a current research thread, not a textbook failure. First-row sum = 0.9985, unitary to 0.15% (with an intriguing ~3σ deficit).

Step 3. Extract $\lambda$ . By definition $\lambda=|V_{us}|$ (to this order), so $\lambda=\mathbf{0.2243}$ , i.e. a Cabibbo angle $\theta_{12}=\arcsin(0.2243)=12.96°$ . Consistency check with the tangent form: $\tan\theta_{12}=|V_{us}|/|V_{ud}|=0.2243/0.97373=0.2304\Rightarrow\theta_{12}=12.97°$ — the two agree, as they must for a unitary row.

Step 4. Extract $A$ . From $|V_{cb}|=A\lambda^2$ , $A=|V_{cb}|/\lambda^2=0.0410/0.2243^2=0.0410/0.050310=\mathbf{0.815}$ . So four measured magnitudes have fixed two of the four Wolfenstein numbers; the corner $|V_{ub}|$ then gives $\sqrt{\rho^2+\eta^2}=|V_{ub}|/(A\lambda^3)=0.00382/(0.815\times0.01129)=0.415$ , pinning the apex distance from the origin. λ = 0.224, A = 0.815, and the apex lies at radius √(ρ²+η²) ≈ 0.41 from the origin.

2 · The Jarlskog invariant and the area of the unitarity triangle (fully worked)

Using the standard-parametrisation angles $\theta_{12}=13.0°$ , $\theta_{23}=2.40°$ , $\theta_{13}=0.21°$ and the phase $\delta=66°$ , compute the Jarlskog invariant $J$ , and hence the area of the normalised unitarity triangle. Compare with $A^2\lambda^6\eta$ .

Step 1. Evaluate the sines and cosines. $s_{12}=\sin13.0°=0.2250$ , $s_{23}=\sin2.40°=0.04188$ , $s_{13}=\sin0.21°=0.003665$ ; the cosines are $c_{12}=0.9744$ , $c_{23}=0.99912$ , $c_{13}=0.99999\approx1$ , and $\sin\delta=\sin66°=0.9135$ .

Step 2. Assemble $J=c_{12}c_{23}c_{13}^2\,s_{12}s_{23}s_{13}\sin\delta$ . Multiply the small factors first: $s_{12}s_{23}s_{13}=0.2250\times0.04188\times0.003665=3.454\times10^{-5}$ . Then the cosines $c_{12}c_{23}c_{13}^2=0.9744\times0.99912\times1=0.9735$ . So $J=0.9735\times3.454\times10^{-5}\times0.9135=\mathbf{3.07\times10^{-5}}$ .

Step 3. Get the triangle area. Every unitarity triangle has area $J/2$ , so the un-normalised area is $1.5\times10^{-5}$ . For the normalised triangle (base scaled to 1) the area is $\bar\eta/2$ ; with apex $\bar\eta\approx0.35$ that area is $0.175$ . Both encode the same physics — a non-zero area is CP violation. J = 3.1×10⁻⁵; the six unitarity triangles all share area J/2 = 1.5×10⁻⁵.

Step 4. Cross-check with Wolfenstein. Using $\lambda=0.225$ , $A=0.826$ , $\eta=0.355$ : $A^2\lambda^6\eta=0.826^2\times0.225^6\times0.355=0.6823\times1.297\times10^{-4}\times0.355=3.14\times10^{-5}$ , agreeing with the exact value to a few percent — the small mismatch is the $\mathcal O(\lambda^4)$ truncation of the Wolfenstein form. The two routes agree: J ≈ 3.1×10⁻⁵, confirming the parametrisations are consistent.

Physics of the portrait

The CKM matrix is one of the rare cases where a piece of pure parameter-counting predicted a whole generation of matter. In 1973 two theorists in Kyoto asked what it would take to fit CP violation into the weak interaction — and the answer, stunningly, was more quarks.

Origins · 2 minMakoto Kobayashi and Toshihide Maskawa and how this was discovered

Makoto Kobayashi and Toshihide Maskawa

b. 1944 and 1940–2021 · the third generation and the CP-violating phase

In 1973, working at Kyoto University, Makoto Kobayashi and Toshihide Maskawa confronted a puzzle nine years old: CP violation, discovered by Cronin and Fitch in the neutral-kaon system in 1964, had no home in the theory of the day. The weak interaction was understood, Cabibbo's quark mixing angle was a decade old, and charm had just been proposed by Glashow, Iliopoulos and Maiani — but with the four quarks then imagined, the mixing matrix was a real $2\times2$ rotation, and Kobayashi and Maskawa proved by direct parameter-counting that a real matrix cannot violate CP. Their paper, "CP Violation in the Renormalizable Theory of Weak Interaction," worked through the possibilities systematically and reached a startling conclusion: the simplest way to accommodate CP violation was to enlarge the quark sector to three generations, because a $3\times3$ unitary mixing matrix carries exactly one irreducible complex phase that no field redefinition can remove. They wrote this in an era when only three quarks were confirmed and the charm quark itself was still hypothetical. The charm quark arrived in November 1974, the tau lepton in 1975, the bottom quark in 1977, and finally the top quark in 1995 — each discovery vindicating a structure the two had deduced from the algebra of unitary matrices. For a generation the phase remained an untested hypothesis until the BaBar and Belle experiments observed large CP violation in $B$ mesons in 2001, drawing the unitarity triangle exactly as predicted. Kobayashi and Maskawa shared the 2008 Nobel Prize in Physics; Nicola Cabibbo, whose 1963 angle is the first row of their matrix and whose name completes the acronym, was not included — a widely-noted omission. Their result stands as one of physics' cleanest demonstrations that a symmetry principle, pursued rigorously, can foretell the contents of the universe.

History

1963

— Cabibbo introduces the quark mixing angle for strangeness-changing weak decays

1973

— Kobayashi and Maskawa show three generations yield one irreducible CP-violating phase

1977

— discovery of the bottom quark confirms a third generation of matter

2001

— BaBar and Belle observe CP violation in B mesons and measure sin2β

2008

— Nobel Prize awarded to Kobayashi and Maskawa for the origin of CP violation

Why is the top quark's coupling V_tb so close to 1?

Because unitarity leaves it almost nowhere else to go. The third row must satisfy $|V_{td}|^2+|V_{ts}|^2+|V_{tb}|^2=1$ , and the first two elements are tiny — $|V_{td}|\approx0.009$ and $|V_{ts}|\approx0.041$ , of order $\lambda^3$ and $\lambda^2$ . Their squares sum to only about $0.0018$ , so $|V_{tb}|^2\approx0.9982$ and $|V_{tb}|\approx0.999$ . Physically this means the top quark decays essentially always to a bottom quark, $t\to bW$ , with only per-mille branchings to $d$ or $s$ . That near-unit coupling is why top physics and bottom physics are so tightly linked, and why single-top production at hadron colliders — whose rate is proportional to $|V_{tb}|^2$ — provides the most direct measurement of this element without assuming three-generation unitarity.

How can V_td and V_ts be measured if the top quark never appears in a decay?

Through quantum loops. The top quark decays far too fast to hadronise, so there is no top-flavoured meson whose semileptonic decay could expose $V_{td}$ or $V_{ts}$ the way $K_{\ell3}$ exposes $V_{us}$ . Instead these elements appear inside "box diagrams" — second-order weak processes in which virtual top quarks mediate the mixing of a neutral $B$ meson with its antiparticle. The oscillation frequency $\Delta m_d$ of the $B_d$ system is proportional to $|V_{td}|^2$ , and $\Delta m_s$ of the $B_s$ system to $|V_{ts}|^2$ ; their ratio cancels much of the theoretical uncertainty and yields $|V_{td}/V_{ts}|$ cleanly. So the very elements that no tree decay can reach are pinned down by how fast a neutral $B$ meson turns into its own antiparticle — a purely quantum-mechanical stopwatch.

Is the CKM phase enough to explain why the universe is made of matter?

No — and the shortfall is dramatic. Generating a matter–antimatter asymmetry requires CP violation (one of Sakharov's three conditions), and the CKM matrix does supply it. But the amount is set by the Jarlskog invariant $J\approx3\times10^{-5}$ , and when propagated through the physics of the early universe it produces a baryon asymmetry roughly ten orders of magnitude smaller than the value cosmology observes. The mixing angles are simply too small: CP violation is suppressed by six powers of $\lambda$ . This is not a defect in the CKM description of laboratory physics, which is superbly accurate, but it is a loud signal that additional CP-violating sources must exist beyond the Standard Model — in the lepton sector, in as-yet-unseen heavy particles, or in interactions that leave their mark in electric dipole moments. The search for that extra CP violation is one of the field's defining goals.

What exactly is the Cabibbo angle's relation to the CKM matrix?

The Cabibbo angle is the CKM matrix's grandparent, and its largest angle. In 1963, with only up, down and strange quarks, Cabibbo explained why strangeness-changing decays are slower than strangeness-conserving ones by positing that the weak partner of the up quark is a rotated combination $d'=\cos\theta_c\,d+\sin\theta_c\,s$ , with $\theta_c\approx13°$ . That single angle is exactly $\theta_{12}$ in the modern parametrisation, and $\lambda=\sin\theta_c\approx0.225$ is the expansion parameter of the whole Wolfenstein hierarchy. When Kobayashi and Maskawa extended the picture to three generations, Cabibbo's $2\times2$ rotation became the upper-left block of a $3\times3$ matrix, gaining two more small angles and the crucial phase. So the CKM matrix is not a replacement for the Cabibbo angle but its natural completion — which is why the acronym keeps his initial first.

Does the CKM matrix allow flavour-changing neutral currents?

Not at tree level, and this is one of its deepest features. The charged current changes flavour freely — that is what the off-diagonal $V_{ij}$ are for. But the neutral current, mediated by the $Z$ or the photon, stays rigidly flavour-diagonal: there is no tree-level $s\to d$ transition via a $Z$ . The reason is unitarity. When you write the neutral current in the weak basis and rotate back to the mass basis, the CKM factors assemble into $V^\dagger V=\mathbb 1$ , and the identity has no off-diagonal entries — the flavour-changing pieces cancel exactly. This is the GIM mechanism (aq-pp-gim-mechanism), and historically the smallness of the observed $K_L\to\mu^+\mu^-$ rate is what forced the charm quark's existence. Flavour-changing neutral currents do occur, but only at the loop level and only suppressed, which is precisely why rare decays like $b\to s\gamma$ are such sensitive probes of new physics.