Three-flavour oscillations & the PMNS matrix

Two flavours told the essential story — a flavour state is a superposition of mass states, and interference between their propagation phases makes the flavour content oscillate. But nature runs three. The weak (flavour) eigenstates $(\nu_e,\nu_\mu,\nu_\tau)$ — born beside a definite charged lepton at a $W$ vertex — are unitary rotations of the mass eigenstates $(\nu_1,\nu_2,\nu_3)$ , which propagate as clean plane waves. The rotation is the $3\times3$ Pontecorvo–Maki–Nakagawa–Sakata matrix $U$ , the leptonic twin of the quark CKM matrix, its nine complex entries fundamental constants no theory yet predicts. Three flavours buy three qualitatively new things two could never hold. First, a genuine phase: a $3\times3$ unitary matrix, after absorbing every removable phase, still carries one irreducible complex phase $\delta$ , so leptonic mixing can violate CP — the amplitude for $\nu_\mu\to\nu_e$ need not equal that for $\bar\nu_\mu\to\bar\nu_e$ . Second, two independent mass-squared splittings, a small "solar" $\Delta m^2_{21}$ and a large "atmospheric" $\Delta m^2_{31}$ , whose ratio $\sim1/30$ separates reactor, solar, and accelerator physics. Third, a discrete unknown — the ordering of the masses, normal or inverted — that no vacuum oscillation amplitude can resolve. Everything downstream in the chapter grows from this matrix: matter effects tilt it (MSW), the phase $\delta$ makes it CP-violating, and its existence forces the masses it mixes to be nonzero. This article assembles $U$ , derives the master oscillation probability, counts its parameters, and pins the CP question to one number — the Jarlskog invariant.

Key takeaways

The weak and mass eigenstates are linked by the $3\times3$ unitary PMNS matrix $U$ : $|\nu_\alpha\rangle=\sum_i U_{\alpha i}^*|\nu_i\rangle$ , generalising the two-flavour mixing angle.
The master probability $P(\nu_\alpha\to\nu_\beta)$ is a sum over mass pairs of $\sin^2\Delta_{ij}$ terms weighted by products of $U$ elements, with the practical phase $\Delta_{ij}=1.267\,\Delta m^2_{ij}[\mathrm{eV}^2]\,L[\mathrm{km}]/E[\mathrm{GeV}]$ .
$U$ is parametrised by three mixing angles $(\theta_{12},\theta_{23},\theta_{13})$ and one Dirac CP phase $\delta$ (plus two Majorana phases invisible to oscillations).
Measured: $\sin^2\theta_{12}\approx0.30$ , $\sin^2\theta_{23}\approx0.55$ , $\sin^2\theta_{13}\approx0.022$ ; $\Delta m^2_{21}\approx7.4\times10^{-5}$ , $|\Delta m^2_{31}|\approx2.5\times10^{-3}\,\mathrm{eV}^2$ — leaving the ordering open.
A complex $\delta$ makes $P(\nu_\mu\to\nu_e)\neq P(\bar\nu_\mu\to\bar\nu_e)$ ; all CP violation is captured by one number, the Jarlskog invariant $J\approx0.033\sin\delta$ .
Key idea: flavour oscillation is interference — a flavour state is a coherent mixture of mass eigenstates that propagate with slightly different phases, so the detected flavour content beats in and out with $L/E$ , governed by the PMNS matrix and the two mass-squared differences.

The core insight

Flavour oscillation is interference: a flavour state is a coherent mixture of mass eigenstates that propagate with slightly different phases, so the detected flavour content beats in and out as a function of $L/E$ , governed by the PMNS matrix and the mass-squared differences.

Where this lives today

Three-flavour oscillation is now precision physics with two headline unknowns open. The mass ordering — is $\nu_3$ heaviest (normal) or lightest (inverted)? — will be settled by JUNO, a 20-kton reactor experiment reading the fine interference of $\Delta m^2_{31}$ and $\Delta m^2_{32}$ , and by the matter-enhanced oscillations at DUNE and in atmospheric detectors (Hyper-Kamiokande, KM3NeT/ORCA). The CP phase $\delta$ is the second prize: T2K and NOvA already disfavour CP conservation at modest significance, and DUNE plus Hyper-Kamiokande aim to measure $\sin\delta$ decisively — a possible clue, via leptogenesis, to why the universe is made of matter. Meanwhile the unitarity of $U$ is under test: any deficit in $\sum_i|U_{\alpha i}|^2$ would betray a fourth, sterile neutrino, kept alive by short-baseline anomalies (MicroBooNE, SBN). The matrix Pontecorvo imagined is now a workbench for physics beyond the Standard Model.

The physical picture

A neutrino is born with a flavour but travels with a mass. Where a $W$ boson turns a muon into a neutrino, the outgoing state is $\nu_\mu$ — a fixed column of the PMNS matrix. But $\nu_\mu$ is not a stationary state of the free Hamiltonian; it is a blend of $\nu_1,\nu_2,\nu_3$ , and those are what move. Each mass eigenstate carries its own phase clock $e^{-i\phi_i}$ ticking at a rate set by its mass, so after a flight $L$ the clocks drift out of register and the recombined flavour is no longer pure $\nu_\mu$ . The first figure shows the raw material of that interference: how much of each flavour lives inside each mass state.

Figure

Flavour composition of the three mass eigenstates in normal ordering — each bar is a mass state split into electron blue muon amber and tau green fractions equal to the moduli squared of the PMNS elements. ν₃ is almost muon–tau with only 2 percent electron set by the small angle θ₁₃ while ν₁ is dominantly electron. Columns and rows each sum to 1 by unitarity

The mixing is large — nothing like the near-diagonal quark CKM matrix. Two angles are big ( $\theta_{12}\approx34°$ , $\theta_{23}\approx48°$ ), so $\nu_2$ and $\nu_3$ are genuine three-way blends; only $\theta_{13}\approx8.6°$ is small, which is why $\nu_3$ holds barely 2% electron flavour. This heavy mixing lets a muon neutrino turn almost fully into a tau neutrino over an atmospheric baseline. The second ingredient is the two clocks: the mass states split into a tight solar pair $(\nu_1,\nu_2)$ and a distant third, and it is the spacing of these levels — not their height — that oscillations read out. The next figure is the level diagram, with the measured splittings and the two orderings.

Figure

The neutrino mass-squared spectrum. A small solar splitting Δm²₂₁ ≈ 7.4×10⁻⁵ eV² separates ν₁ and ν₂ while a large atmospheric gap ≈ 2.5×10⁻³ eV² sets off ν₃ — about thirty times wider. Oscillations fix these gaps but not the absolute scale nor whether ν₃ sits above the solar pair normal ordering or below it inverted ordering

The core equations

A neutrino created with a charged lepton of flavour $\alpha$ is, by definition, the flavour eigenstate $|\nu_\alpha\rangle$ . Expanding it over the propagating mass basis is the first line of everything that follows; the conjugate on $U$ appears because the neutrino enters the charged-current vertex as the field that annihilates it (the adjoint spinor).

|\nu_\alpha\rangle=\sum_{i=1}^{3} U_{\alpha i}^{*}\,|\nu_i\rangle,\qquad UU^{\dagger}=I \qquad (1)

where: $\alpha\in\{e,\mu,\tau\}$ labels flavour, $i\in\{1,2,3\}$ mass; $U_{\alpha i}$ is the PMNS element; unitarity enforces $\sum_i U_{\alpha i}U_{\beta i}^{*}=\delta_{\alpha\beta}$ . Reading: each flavour is a fixed weighted sum of the three mass states; the same nine numbers run the reverse expansion $|\nu_i\rangle=\sum_\alpha U_{\alpha i}|\nu_\alpha\rangle$ . Each mass eigenstate propagates as a plane wave with phase $\phi_i=E_i t-\mathbf p_i\!\cdot\!\mathbf x$ . Projecting the evolved state onto flavour $\beta$ and squaring gives the master oscillation probability, valid for any number of flavours.

P(\nu_\alpha\!\to\!\nu_\beta)=\delta_{\alpha\beta}-4\!\sum_{i>j}\!\operatorname{Re}\!\big(U_{\alpha i}^{*}U_{\beta i}U_{\alpha j}U_{\beta j}^{*}\big)\sin^{2}\!\Delta_{ij}+2\!\sum_{i>j}\!\operatorname{Im}\!\big(U_{\alpha i}^{*}U_{\beta i}U_{\alpha j}U_{\beta j}^{*}\big)\sin 2\Delta_{ij} \qquad (2)

where: $\Delta_{ij}\equiv\dfrac{\Delta m^2_{ij}\,L}{4E}$ , $\Delta m^2_{ij}\equiv m_i^2-m_j^2$ , $L$ baseline, $E$ energy. Reading: the real part is the CP-even oscillation; the imaginary part is CP-odd and flips sign for antineutrinos ( $U\!\to\!U^{*}$ ), the origin of leptonic CP violation. The oscillating phase is the most-used formula in the field. Restoring $\hbar$ and $c$ and inserting practical units collapses it to one memorable coefficient.

\Delta_{ij}=\frac{\Delta m^2_{ij}\,L}{4E}=1.267\;\frac{\Delta m^2_{ij}\,[\mathrm{eV}^2]\;L\,[\mathrm{km}]}{E\,[\mathrm{GeV}]} \qquad (3)

where: $1.267=(4\hbar c)^{-1}$ in these units. Reading: an experiment is "tuned" to a splitting when $\Delta_{ij}\sim\pi/2$ ; short-baseline reactors sit on the atmospheric scale, KamLAND ( $\sim180$ km) on the solar scale. The nine entries of $U$ are not independent. After using unitarity and absorbing the phases that can be rotated into the lepton fields, $U$ reduces to three angles and one physical phase, conventionally factored as an atmospheric, a reactor, and a solar rotation.

U=\underbrace{\begin{pmatrix}1&0&0\\0&c_{23}&s_{23}\\0&-s_{23}&c_{23}\end{pmatrix}}_{\text{atmospheric}}\underbrace{\begin{pmatrix}c_{13}&0&s_{13}e^{-i\delta}\\0&1&0\\-s_{13}e^{i\delta}&0&c_{13}\end{pmatrix}}_{\text{reactor}}\underbrace{\begin{pmatrix}c_{12}&s_{12}&0\\-s_{12}&c_{12}&0\\0&0&1\end{pmatrix}}_{\text{solar}} \qquad (4)

where: $c_{ij}=\cos\theta_{ij}$ , $s_{ij}=\sin\theta_{ij}$ , $\delta$ the Dirac CP phase. Reading: $\theta_{23}$ governs atmospheric $\nu_\mu\!\leftrightarrow\!\nu_\tau$ , $\theta_{12}$ the solar $\nu_e$ survival, $\theta_{13}$ the small reactor amplitude gating $\delta$ — no $\theta_{13}$ , no observable CP violation. All CP violation, whatever the channel, is controlled by one rephasing-invariant combination, the Jarlskog invariant $J$ — the "area" of the unitarity triangle.

J=\operatorname{Im}\!\big(U_{e1}U_{\mu 2}U_{e2}^{*}U_{\mu 1}^{*}\big)=s_{12}c_{12}\,s_{23}c_{23}\,s_{13}c_{13}^{2}\,\sin\delta \qquad (5)

where: the same $J$ (up to sign) governs every flavour pair. Reading: $J$ vanishes if any angle is $0$ or $90°$ or if $\delta=0,\pi$ — CP violation needs all three angles nonzero and a genuinely complex phase. Numerically $J\le 0.033$ , so leptonic CP violation, if present, is near-maximal in structure and awaits only a measurement of $\sin\delta$ .

Derivation AFrom the flavour state to the master oscillation formula

The goal is to turn the physical story — born as flavour, propagate as mass, detected as flavour — into the exact probability of equation (2), being careful about where each PMNS factor and its conjugate enter, and to show the result depends only on mass-squared differences and on $L/E$ . We start from the flavour expansion, evolve each mass eigenstate by its phase, project onto the final flavour, and square, keeping the interference cross-terms that carry the physics.

Step 1. Write the produced state as $|\nu_\alpha\rangle=\sum_i U_{\alpha i}^{*}|\nu_i\rangle$ and evolve each mass eigenstate by its plane-wave phase, $|\nu_i(L)\rangle=e^{-i\phi_i}|\nu_i\rangle$ with $\phi_i=E_i t-p_iL$ . Why: the mass eigenstates are the energy eigenstates of the free Hamiltonian, so they and only they evolve by a simple phase — this is exactly why we change basis to them.

Step 2. In the ultrarelativistic limit $E_i=\sqrt{p^2+m_i^2}\approx p+m_i^2/2E$ with common momentum $p\approx E$ and $t\approx L$ , so the relative phase between two mass states is $\phi_i-\phi_j\approx (m_i^2-m_j^2)L/2E=\Delta m^2_{ij}L/2E=2\Delta_{ij}$ . Why: only phase differences survive squaring, and the absolute masses drop out — the observable is $\Delta m^2$ , never $m_i$ itself, and it enters purely through $L/E$ .

Step 3. Project onto the detected flavour using the reverse expansion $\langle\nu_\beta|=\sum_i U_{\beta i}\langle\nu_i|$ , giving the amplitude $A_{\alpha\beta}=\sum_i U_{\alpha i}^{*}U_{\beta i}e^{-i\phi_i}$ . Why: detection fixes the outgoing charged-lepton flavour, so the final projection uses $U$ (not $U^{*}$ ); the mismatch of conjugation between production and detection is what makes the $\operatorname{Im}$ term in (2) physical.

Step 4. Square, $P=|A_{\alpha\beta}|^2=\sum_{i,j}U_{\alpha i}^{*}U_{\beta i}U_{\alpha j}U_{\beta j}^{*}e^{-i(\phi_i-\phi_j)}$ , split into diagonal ( $i=j$ , giving $\delta_{\alpha\beta}$ by unitarity) and off-diagonal terms, and use $e^{-i2\Delta_{ij}}=1-2\sin^2\Delta_{ij}-i\sin2\Delta_{ij}$ to reach equation (2). Why: the diagonal piece is the "no oscillation" baseline; the real cross-terms give the familiar $\sin^2$ oscillation and the imaginary cross-terms give the CP-odd asymmetry, cleanly separated. *As a check: setting the third state aside ( $U_{\alpha3}=U_{\beta3}=0$ ) reduces (2) to the two-flavour $P=\sin^2 2\theta\,\sin^2\Delta$ , and $\sum_\beta P(\nu_\alpha\to\nu_\beta)=1$ by unitarity.*

Go deeper · Derivation B · 1 minCounting the parameters of the PMNS matrix

Derivation BCounting the parameters of the PMNS matrix

The aim is to explain why a $3\times3$ mixing matrix, seemingly carrying nine complex numbers, collapses to just three angles and one CP phase for Dirac neutrinos — and why two extra "Majorana" phases hide from oscillations. The bookkeeping is a subtraction: real parameters of a unitary matrix, minus the phases removable by redefining the lepton fields, equals the physical content.

Step 1. A general $3\times3$ unitary matrix has $9$ real parameters: an $N\times N$ unitary matrix has $N^2$ , split into $N(N-1)/2=3$ rotation angles and $N(N+1)/2=6$ phases. Why: the angles are the "how much mixing" and the phases are the "complex character"; only some of the latter are physical.

Step 2. Remove phases by rephasing the six lepton fields (three charged leptons $e,\mu,\tau$ and three neutrinos). For Dirac neutrinos, $2N-1=5$ of the six phases can be absorbed, leaving $6-5=1$ physical phase. Why: multiplying a charged-lepton or neutrino field by $e^{i\varphi}$ shifts entries of $U$ but is unobservable; one overall phase is common and cannot remove anything, hence $2N-1$ , not $2N$ .

Step 3. The survivor is the Dirac phase $\delta$ ; combined with the three angles this gives $3+1=4$ physical parameters, matched to the standard factorisation of equation (4). Why: writing $U$ as three sequential rotations with $\delta$ attached to the $\theta_{13}$ block isolates each angle in the sector that measures it, and puts $\delta$ where only $\theta_{13}\neq0$ can activate it.

Step 4. If neutrinos are Majorana, they cannot be rephased freely (the Majorana mass term is not phase-invariant), so two of the previously-absorbed phases become physical: $U\to U\,\mathrm{diag}(1,e^{i\alpha_{21}/2},e^{i\alpha_{31}/2})$ . Why: these Majorana phases multiply each amplitude by a common factor in oscillations and cancel in $|A|^2$ — they are invisible to oscillation experiments and show up only in lepton-number-violating processes like neutrinoless double beta decay. *As a check: the parameter count is $3$ angles $+1$ Dirac phase for Dirac neutrinos, with $+2$ Majorana phases for Majorana neutrinos, none of the latter affecting $P(\nu_\alpha\to\nu_\beta)$ .*

Go deeper · Derivation C · 1 minCP violation and the Jarlskog invariant

Derivation CCP violation and the Jarlskog invariant

The goal is to show that the difference between a neutrino oscillation and its antineutrino mirror is governed by one rephasing-invariant number $J$ times a product of all three phase factors, so that leptonic CP violation requires every mass splitting and every mixing angle nonzero at once. We form the CP asymmetry from the master formula, use unitarity to collapse the sum, and expose the triple-sine structure.

Step 1. Under CP, $U\to U^{*}$ , so the real part of (2) is unchanged while the imaginary part flips sign: $P(\bar\nu_\alpha\to\bar\nu_\beta)$ has the same $\operatorname{Re}$ term but $-\operatorname{Im}$ . Hence $\Delta P_{CP}\equiv P(\nu_\alpha\to\nu_\beta)-P(\bar\nu_\alpha\to\bar\nu_\beta)=4\sum_{i>j}\operatorname{Im}(U_{\alpha i}^{*}U_{\beta i}U_{\alpha j}U_{\beta j}^{*})\sin2\Delta_{ij}$ . Why: only the CP-odd imaginary interference survives the subtraction; the bulk $\sin^2$ oscillation is CP-even and cancels.

Step 2. For $\alpha\neq\beta$ the three imaginary parts are equal up to sign and each equals $\pm J$ , the common Jarlskog invariant, so $\Delta P_{CP}=4J(\sin2\Delta_{21}+\sin2\Delta_{32}+\sin2\Delta_{13})$ with the cyclic labelling. Why: unitarity forces all off-diagonal $\operatorname{Im}(\cdots)$ for a given $(\alpha,\beta)$ to share one magnitude — there is a single independent measure of CP violation, not three.

Step 3. Because $\Delta_{21}+\Delta_{32}+\Delta_{13}=0$ , the identity $\sin2A+\sin2B+\sin2C=-4\sin A\sin B\sin C$ (for $A+B+C=0$ ) collapses the bracket, giving $\Delta P_{CP}=16\,J\,\sin\Delta_{21}\,\sin\Delta_{32}\,\sin\Delta_{31}$ . Why: this is the punchline — the asymmetry needs all three $\Delta_{ij}\neq0$ , so if any two masses were degenerate CP violation would vanish even with $\delta\neq0$ .

Step 4. Insert the parametrisation to get $J=s_{12}c_{12}s_{23}c_{23}s_{13}c_{13}^{2}\sin\delta$ , whose numerical ceiling with measured angles is $J_{\max}\approx0.033$ . Why: this ties the abstract invariant to the four physical parameters and shows $\theta_{13}\neq0$ (discovered only in 2012) is precisely what reopened the leptonic CP question. *As a check: $\Delta P_{CP}\to0$ if any $\Delta_{ij}\to0$ , if any angle $\to0$ , or if $\delta\to0,\pi$ ; and $J=\operatorname{Im}(U_{e1}U_{\mu2}U_{e2}^{*}U_{\mu1}^{*})$ evaluates to $s_{12}c_{12}s_{23}c_{23}s_{13}c_{13}^2\sin\delta$ .*

Limiting cases

Case	Result	Meaning
One splitting dominates ( $\Delta_{21}\to0$ , $\theta_{13}$ small)	$P(\nu_\mu\!\to\!\nu_\tau)\approx\sin^2 2\theta_{23}\sin^2\Delta_{31}$	atmospheric two-flavour limit; Super-K's original signal
Short-baseline reactor	$P(\bar\nu_e\!\to\!\bar\nu_e)\approx1-\sin^2 2\theta_{13}\sin^2\Delta_{31}$	isolates $\theta_{13}$ (Daya Bay, RENO, Double Chooz)
Long-baseline reactor (KamLAND)	$P\approx1-\cos^4\theta_{13}\sin^2 2\theta_{12}\sin^2\Delta_{21}$	solar sector at reactor energies; measures $\theta_{12},\Delta m^2_{21}$
$\delta=0$ or $\pi$	$J=0$ , $\;P(\nu_\alpha\!\to\!\nu_\beta)=P(\bar\nu_\alpha\!\to\!\bar\nu_\beta)$	CP conserved even with three flavours
Any two masses degenerate	$\Delta P_{CP}\to0$ , oscillation into that pair frozen	CP violation needs three distinct masses
$L/E\to0$	$P\to\delta_{\alpha\beta}$	no flight, no interference — the mixing is invisible

Common trapOscillations do not measure neutrino masses

The single most common error is to read an oscillation result as a neutrino mass. Oscillation probabilities depend only on the phase differences $\Delta m^2_{ij}=m_i^2-m_j^2$ , so they measure mass-squared splittings and mixing angles — never an absolute mass, and (in vacuum) not even the sign of the large splitting, leaving the normal-versus-inverted ordering open. The absolute scale comes from elsewhere: the tritium beta-decay endpoint (KATRIN) bounds $m_\beta\lesssim0.8\text{ eV}$ , cosmology bounds $\sum m_i\lesssim0.1\text{ eV}$ . It is likewise meaningless to speak of "the mass of the electron neutrino": $\nu_e$ is a superposition of three mass eigenstates, so it has no single mass any more than a chord has a single pitch. What you can state is its mass content — $\nu_e$ is $68\%\,\nu_1$ , $30\%\,\nu_2$ , $2\%\,\nu_3$ — a statement about the PMNS matrix, not a mass.

Worked examples

1 · Tuning a beam to the oscillation maximum, and the reactor θ₁₃ deficit (confidence builder)

(a) At what $L/E$ does the atmospheric oscillation first peak? Use it to explain why the T2K beam ( $L=295$ km) is centred at $E\approx0.6$ GeV. (b) A reactor experiment sits at $L=1.6$ km with $\bar\nu_e$ of mean energy $E\approx4$ MeV; estimate the $\bar\nu_e$ deficit.

Step 1. The first maximum of $\sin^2\Delta_{31}$ is at $\Delta_{31}=\pi/2$ . Using equation (3) with $\Delta m^2_{31}=2.5\times10^{-3}\text{ eV}^2$ : $1.267\times2.5\times10^{-3}\times(L/E)=\pi/2$ , so $L/E=\pi/(2\times1.267\times2.5\times10^{-3})\approx\mathbf{496\ km/GeV}$ .

Step 2. For T2K's fixed baseline $L=295$ km, the matching energy is $E=295/496\approx\mathbf{0.60\ GeV}$ — exactly the peak of T2K's off-axis narrow-band beam. The baseline and beam energy are co-designed to park the experiment on the atmospheric oscillation maximum, where the appearance signal is largest.

Step 3. For the reactor, first evaluate the phase: $\Delta_{31}=1.267\times2.5\times10^{-3}\times1.6/(4\times10^{-3})=\mathbf{1.27\ rad}$ (near maximal, $\sin^2\Delta_{31}=0.91$ ). With $\sin^2\theta_{13}=0.022$ , $\sin^2 2\theta_{13}=4\sin^2\theta_{13}\cos^2\theta_{13}=0.086$ .

Step 4. The survival deficit is $1-P(\bar\nu_e\to\bar\nu_e)\approx\sin^2 2\theta_{13}\,\sin^2\Delta_{31}=0.086\times0.91\approx\mathbf{0.078}$ , i.e. about a 7.8% disappearance of reactor $\bar\nu_e$ at $1.6$ km. This few-percent dip, resolved by comparing near and far detectors, is exactly what Daya Bay measured in 2012 to nail down the last mixing angle.

2 · Muon-to-electron appearance and the CP term at T2K

Estimate the leading $P(\nu_\mu\to\nu_e)$ at the T2K oscillation maximum, and gauge the size of the CP-violating contribution.

Step 1. At $\Delta_{31}=\pi/2$ the leading (atmospheric) appearance term is $P(\nu_\mu\to\nu_e)\approx\sin^2\theta_{23}\,\sin^2 2\theta_{13}\,\sin^2\Delta_{31}$ . With $\sin^2\theta_{23}=0.55$ , $\sin^2 2\theta_{13}=0.086$ , $\sin^2\Delta_{31}=1$ : $P\approx0.55\times0.086\times1\approx\mathbf{0.047}$ , a 4.7% appearance probability.

Step 2. The full formula adds a CP-interference term $\propto\sin\delta$ from the Jarlskog structure. With $\Delta_{21}/\Delta_{31}\approx7.4\times10^{-5}/2.5\times10^{-3}\approx0.03$ it is a sub-dominant few-percent modulation of the 4.7% signal — small, but the only term that differs between $\nu$ and $\bar\nu$ .

Step 3. Evaluate $J$ at the best-fit $\delta\approx197°$ : $J=s_{12}c_{12}s_{23}c_{23}s_{13}c_{13}^2\sin\delta=0.033\times\sin(197°)\approx\mathbf{-0.0097}$ . Nonzero $J$ means $P(\nu_\mu\to\nu_e)\neq P(\bar\nu_\mu\to\bar\nu_e)$ — the signature of leptonic CP violation.

Step 4. The asymmetry $A_{CP}=[P(\nu_\mu\!\to\!\nu_e)-P(\bar\nu_\mu\!\to\!\bar\nu_e)]/[\,\text{sum}\,]$ can reach tens of percent, because the small CP term is divided by the small appearance rate, not the large survival rate. That amplification is why long-baseline appearance, not disappearance, is the route to $\delta$ .

Physics of the portrait

Neutrino mixing is a two-country, two-decade idea: an Italian émigré in the Soviet Union who dared to let a chargeless particle change identity, and a Japanese trio who wrote the change as a matrix — vindicated only forty years later underground.

Origins · 1 minBruno Pontecorvo and how this was discovered

Bruno Pontecorvo

1913–1993 · neutrino oscillations and leptonic mixing

Bruno Pontecorvo was Enrico Fermi's youngest student — one of the "Via Panisperna boys" in 1930s Rome — before a career through Paris, wartime North America, and, in a 1950 defection that stunned the West, Dubna in the Soviet Union. His physics was as bold as his biography. In 1957, inspired by the newly found $K^0$ – $\bar K^0$ oscillations, he proposed that neutrinos might oscillate too, first as $\nu\leftrightarrow\bar\nu$ and then, once the muon neutrino was known, as $\nu_e\leftrightarrow\nu_\mu$ flavour change — radical when neutrinos were universally assumed massless. He also conceived the chlorine radiochemical method Ray Davis used to catch solar neutrinos, and read the resulting "solar neutrino problem" as a hint of oscillation, not a failure of solar models. The matrix carrying his name was written in 1962 by Ziro Maki, Masami Nakagawa, and Shoichi Sakata, who mixed the neutrino "mass" and "weak" states to keep lepton bookkeeping consistent. For decades the idea sat as elegant speculation; when Super-Kamiokande (1998, atmospheric) and SNO (2002, solar) finally proved neutrinos change flavour — and therefore have mass — they vindicated a picture Pontecorvo had sketched a lifetime earlier and did not live to see confirmed.

History

1957

— Pontecorvo proposes that neutrinos can oscillate

1962

— Maki Nakagawa and Sakata write down neutrino flavour mixing

1998

— Super-Kamiokande discovers atmospheric muon-neutrino oscillation

2002

— SNO and KamLAND confirm solar and reactor oscillations

2012

— Daya Bay measures the last mixing angle θ₁₃

Going further

Texts. Thomson, Modern Particle Physics, §13.5–13.6 (three-flavour oscillations, matter effects) · Griffiths, Introduction to Elementary Particles, 2nd ed., Ch. 11 (neutrino oscillations) · Schwartz, Quantum Field Theory and the Standard Model, §29 (neutrino masses and mixing) · Zuber, Neutrino Physics, 3rd ed., Chs. 5–8 · Giunti & Kim, Fundamentals of Neutrino Physics and Astrophysics, Chs. 13–14 · PDG Review, Neutrino Masses, Mixing, and Oscillations. Key numerical and structural results to keep in mind: the master formula reduces to the two-flavour case; $1.267=(4\hbar c)^{-1}$ in eV², km, GeV; the parameter count is $3+1$ (Dirac), with $+2$ Majorana phases invisible to oscillations; $\Delta P_{CP}=16J\prod\sin\Delta_{ij}$ with $J\le0.033$ ; and $|U_{ei}|^2=(0.68,0.30,0.02)$ from the measured angles.

Prerequisites. Two-flavour oscillations · Mass vs weak eigenstates · Chiral weak structure. Leads to. MSW matter effects · CP violation in the Standard Model · Neutrino mass mechanisms.

Ask the tutor: Ask me to walk through how the MSW matter effect tilts these vacuum probabilities and finally breaks the mass-ordering degeneracy.

Why do production and detection use complex conjugates of each other?

It comes from the chiral structure of the charged-current weak interaction. At production a $W$ vertex turns $\ell_\alpha^-$ into a neutrino through the field that annihilates it, carrying $U_{\alpha i}^{*}$ , so $|\nu_\alpha\rangle=\sum_i U_{\alpha i}^{*}|\nu_i\rangle$ . At detection the neutrino is absorbed and $\ell_\beta^-$ comes out through a vertex that uses $U_{\beta i}$ . The opposite conjugations are not cosmetic — the mismatch is exactly what lets the imaginary interference term be nonzero, the entire origin of CP violation. For antineutrinos the roles swap, $U\leftrightarrow U^{*}$ , flipping that term's sign.

What is the difference between the "solar" and "atmospheric" splittings?

They are the two independent mass-squared differences among three states. The solar splitting $\Delta m^2_{21}\approx7.4\times10^{-5}\text{ eV}^2$ separates the close pair $\nu_1,\nu_2$ ; pinned down by solar experiments and KamLAND, it drives slow $\nu_e$ oscillations governed by $\theta_{12}$ . The atmospheric splitting $|\Delta m^2_{31}|\approx2.5\times10^{-3}\text{ eV}^2$ — about thirty times larger — sets off the third state; first seen in atmospheric $\nu_\mu$ by Super-Kamiokande, it drives fast $\nu_\mu\leftrightarrow\nu_\tau$ oscillations governed by $\theta_{23}$ . Because they differ so much, an experiment usually sees one splitting oscillating while the other is frozen ( $\Delta\approx0$ ) or averaged out.

What exactly is the mass ordering, and why can't oscillations in vacuum fix it?

The ordering asks whether the lone state $\nu_3$ sits above the solar pair (normal, $m_3>m_2>m_1$ ) or below it (inverted, $m_2>m_1>m_3$ ). Vacuum probabilities depend on $\sin^2\Delta_{ij}$ , even in $\Delta m^2_{ij}$ , so flipping the sign of the large splitting leaves them unchanged — both orderings predict the same vacuum spectrum. The tie is broken either by matter effects (the MSW resonance enhances one ordering and suppresses the other, exploited by DUNE and atmospheric detectors) or by the fine interference of $\Delta m^2_{31}$ against $\Delta m^2_{32}$ in a high-resolution reactor spectrum (JUNO). Both are underway.

Why is the PMNS matrix so different from the quark CKM matrix?

The CKM matrix is nearly the identity — quark angles are small, the largest (Cabibbo) only about $13°$ — so quark flavour barely mixes. The PMNS matrix is the opposite: two of its angles are large ( $\theta_{12}\approx34°$ , $\theta_{23}\approx48°$ ), so neutrino flavours mix strongly and a $\nu_\mu$ can convert almost entirely into $\nu_\tau$ . Only $\theta_{13}\approx8.6°$ is "small", and even that dwarfs the quark angles. Why the two sectors look so different is unsolved; ideas range from flavour symmetries to the seesaw mechanism behind the tiny neutrino masses. No accepted theory predicts either set of angles.

What are Majorana phases and why don't they show up here?

If neutrinos are their own antiparticles (Majorana fermions), the mass term is not rephasing-invariant, so two phases that were rotated away in the Dirac case survive: $U\to U\,\mathrm{diag}(1,e^{i\alpha_{21}/2},e^{i\alpha_{31}/2})$ . But in an oscillation amplitude $A_{\alpha\beta}=\sum_i U_{\alpha i}^{*}U_{\beta i}e^{-i\phi_i}$ each Majorana phase multiplies a term and its conjugate, cancelling in $|A|^2$ — oscillations are blind to them. They surface only in lepton-number-violating processes, above all neutrinoless double beta decay, whose rate depends on $\sum_i U_{ei}^2 m_i$ including these phases. That is why $0\nu\beta\beta$ searches are the complementary probe.

How near-maximal is leptonic CP violation, and is it established?

The structure is near-maximal: the Jarlskog ceiling from the measured angles, $J_{\max}\approx0.033$ , is over three orders of magnitude above the quark value ( $J_{\text{CKM}}\approx3\times10^{-5}$ ). So if $\sin\delta$ is near $\pm1$ , leptonic CP violation is large. Whether it is actually there is unsettled: T2K and NOvA together disfavour the CP-conserving values $\delta=0,\pi$ only at modest significance, with a mild tension between them. A decisive measurement of $\delta$ is the primary goal of DUNE and Hyper-Kamiokande, with implications for leptogenesis and the matter–antimatter asymmetry of the universe.