How Matrices Describe Neutrino Oscillations

Physicists love telling the story that neutrinos are strange because they “change identity” as they travel. Electron neutrinos turn into muon neutrinos. Muon neutrinos turn into tau neutrinos. Observations of this effect — first at the Super-Kamiokande detector in 1998, later at SNO, KamLAND, and a long list of others — produced the only confirmed deviation from the original Standard Model of particle physics. The 2015 Nobel Prize was awarded for it.

What is rarely mentioned is that the “strangeness” has an entirely unremarkable mathematical explanation. It is a textbook application of linear algebra — specifically, a change of basis between two orthonormal frames in a complex vector space, followed by diagonal time evolution in one of them. The weird physical behavior is just what that matrix mathematics produces when projected onto an experimenter’s measurement basis.

This article goes through the math. It assumes you’re comfortable with what a matrix is and what it means to multiply one by a vector. Everything else we’ll build from scratch.

Two bases for one particle

A neutrino, in modern physics, is described by a state vector in a three-dimensional complex vector space $\mathbb{C}^3$ . That is, the thing we call “a neutrino” is a unit vector $|\nu\rangle$ with three complex components. There are two natural choices of basis for this space, and the entire phenomenon of oscillation comes from the fact that they don’t agree.

The flavor basis. Particle physicists detect neutrinos by the reactions they trigger. An electron neutrino produces an electron in a detector; a muon neutrino produces a muon; a tau neutrino produces a tau lepton. This defines three physical states — call them $|\nu_e\rangle$ , $|\nu_\mu\rangle$ , $|\nu_\tau\rangle$ — and any detector measurement is, mathematically, a projection onto this basis.

The mass basis. A neutrino has a mass. More precisely, it has a definite energy when its mass is definite, and the time evolution of a state with definite energy is particularly simple: it just picks up a complex phase $e^{-iEt/\hbar}$ . The three states of definite mass — call them $|\nu_1\rangle$ , $|\nu_2\rangle$ , $|\nu_3\rangle$ — evolve independently, each with its own phase.

Here’s the crucial point: these two bases are not the same. A state of definite flavor does not have definite mass, and vice versa. A state $|\nu_e\rangle$ is a specific linear combination of $|\nu_1\rangle$ , $|\nu_2\rangle$ , and $|\nu_3\rangle$ . This single fact is the mathematical origin of neutrino oscillation.

The PMNS matrix

The transformation from the mass basis to the flavor basis is described by a 3×3 matrix called the Pontecorvo–Maki–Nakagawa–Sakata matrix — PMNS for short — named after the four physicists who worked out the theoretical framework between 1957 and 1962.

\begin{pmatrix} \nu_e \\ \nu_\mu \\ \nu_\tau \end{pmatrix} = U \begin{pmatrix} \nu_1 \\ \nu_2 \\ \nu_3 \end{pmatrix}, \qquad U = \begin{pmatrix} U_{e1} & U_{e2} & U_{e3} \\ U_{\mu 1} & U_{\mu 2} & U_{\mu 3} \\ U_{\tau 1} & U_{\tau 2} & U_{\tau 3} \end{pmatrix}.

$U$ is unitary — meaning $U^\dagger U = I$ , where $U^\dagger$ is the conjugate transpose. That’s not an arbitrary property; it’s forced by the geometry. Both bases are orthonormal (the three flavors are mutually exclusive outcomes of an experiment, and so are the three mass states), and any matrix that maps one orthonormal basis to another must preserve inner products. The defining equation of unitary matrices is exactly that.

A general 3×3 unitary matrix has nine real parameters (eighteen real components minus nine unitarity constraints). For the PMNS matrix, five of them can be absorbed into the definition of the states themselves, leaving four physical parameters: three mixing angles $\theta_{12}$ , $\theta_{13}$ , $\theta_{23}$ , and one CP-violating phase $\delta$ . Current measurements put them at roughly $\theta_{12} \approx 33°$ , $\theta_{23} \approx 49°$ , $\theta_{13} \approx 8.6°$ , and $\delta$ somewhere near $\pi$ but with large uncertainty.

If you’ve read our piece on what matrices really are, this should feel familiar. $U$ is a change-of-basis matrix. Its entries are not arbitrary numbers; they are the components of each flavor state expressed in the mass basis.

Time evolution in the mass basis

Here’s where the physics enters — but only a single line of it.

In quantum mechanics, a state of definite energy $E$ evolves in time by picking up a phase:

$|\psi(t)\rangle = e^{-iEt/\hbar} |\psi(0)\rangle.$

The mass eigenstates $|\nu_i\rangle$ have energies $E_i$ (determined by their masses $m_i$ and momentum), so their time evolution is trivial:

$|\nu_i(t)\rangle = e^{-iE_i t/\hbar} |\nu_i(0)\rangle.$

Each mass state evolves independently. Their phases tick at different rates because their energies are different.

This is the only physical input we need. The rest is linear algebra.

The two-flavor derivation

To see the oscillation cleanly, it helps to drop one dimension and work with a 2×2 version. (Nature actually does have three flavors, but the two-flavor case is nearly identical in essence and vastly less cluttered.)

In the 2×2 case, the mixing matrix is just a rotation:

U = \begin{pmatrix} \cos\theta & \sin\theta \\ -\sin\theta & \cos\theta \end{pmatrix}.

Now suppose at time $t = 0$ we produce an electron neutrino. That is, the initial state is $|\psi(0)\rangle = |\nu_e\rangle$ , which in the mass basis reads $\cos\theta |\nu_1\rangle + \sin\theta |\nu_2\rangle$ .

Let the mass states evolve:

$|\psi(t)\rangle = \cos\theta \, e^{-iE_1 t/\hbar} |\nu_1\rangle + \sin\theta \, e^{-iE_2 t/\hbar} |\nu_2\rangle.$

The state is still a valid neutrino state — still a unit vector in our complex 2D space — but it is no longer a pure $|\nu_e\rangle$ , because the two phase factors in front of $|\nu_1\rangle$ and $|\nu_2\rangle$ are no longer equal.

Now ask: what is the probability that, if we detect this state at time $t$ , it registers as a muon neutrino? Quantum mechanics says:

$P(\nu_e \to \nu_\mu) = |\langle \nu_\mu | \psi(t) \rangle|^2.$

Compute the inner product:

$\langle \nu_\mu | \psi(t) \rangle = -\sin\theta \cos\theta \, e^{-iE_1 t/\hbar} + \cos\theta \sin\theta \, e^{-iE_2 t/\hbar}$

$= \sin\theta \cos\theta \left( e^{-iE_2 t/\hbar} - e^{-iE_1 t/\hbar} \right).$

Take the modulus squared. Using the identity $|e^{ix} - e^{iy}|^2 = 4\sin^2\!\left(\frac{x-y}{2}\right)$ and $\sin\theta\cos\theta = \frac{1}{2}\sin(2\theta)$ :

$\boxed{P(\nu_e \to \nu_\mu) = \sin^2(2\theta) \cdot \sin^2\!\left(\frac{(E_2 - E_1)\,t}{2\hbar}\right).}$

For relativistic neutrinos traveling a distance $L = ct$ , the energy difference simplifies to $E_2 - E_1 \approx \Delta m^2 / (2E)$ , where $\Delta m^2 = m_2^2 - m_1^2$ . Substituting (and setting $\hbar = c = 1$ in natural units):

$P(\nu_e \to \nu_\mu) = \sin^2(2\theta) \cdot \sin^2\!\left(\frac{\Delta m^2 \, L}{4E}\right).$

This is the neutrino oscillation formula. The probability of flavor change is a sinusoidal function of distance — literally an oscillation. The amplitude is set by the mixing angle $\theta$ ; the wavelength is set by the mass-squared difference $\Delta m^2$ and the neutrino energy $E$ .

Notice what just happened mathematically. Two phases at slightly different frequencies produced a beat pattern. The beat pattern, viewed through the fixed-basis lens of flavor measurement, looks like a particle changing identity. Nothing is “really” changing — the full quantum state is evolving smoothly. The appearance of identity change is an artefact of measuring in a basis that is not the one where time evolution is diagonal.

Why unitarity matters

A subtle but important feature of the math: because $U$ is unitary, the total probability of detecting the neutrino as some flavor is always $1$ . Whatever its initial flavor, at any later time:

$P(\nu_e \to \nu_e) + P(\nu_e \to \nu_\mu) + P(\nu_e \to \nu_\tau) = 1.$

Neutrinos don’t disappear or multiply; they just shuffle probability between flavors. This conservation law is a direct consequence of the matrix being unitary. Any non-unitary matrix would break it.

Physicists search very carefully for tiny deviations from unitarity in the PMNS matrix, because any such deviation would signal the existence of additional, as-yet-undiscovered neutrino states beyond the three we know about.

The three-flavor case

Everything above generalises cleanly to three flavors. The $2\times 2$ rotation matrix is replaced by the full $3\times 3$ PMNS matrix. The two-phase interference becomes a three-phase interference, and the probability formula gains additional oscillation terms controlled by the three independent mixing angles and the CP phase.

One qualitatively new feature appears in the three-flavor case: if the CP-violating phase $\delta$ is nonzero, then the probability of $\nu_e \to \nu_\mu$ is different from the probability of $\bar\nu_e \to \bar\nu_\mu$ for antineutrinos. This asymmetry between matter and antimatter in the neutrino sector is currently being measured by experiments like T2K and NOvA. It may turn out to be connected to why the universe contains more matter than antimatter — one of the deepest open questions in cosmology.

Mathematically, CP violation in the three-flavor PMNS matrix is exactly analogous to CP violation in the quark sector’s CKM matrix, which won Kobayashi and Maskawa the 2008 Nobel Prize in Physics. Both are consequences of the same underlying fact: a 3×3 unitary matrix, after phase redefinitions, still has one irreducible phase parameter. A 2×2 unitary matrix, after the analogous redefinitions, has none. This is a theorem of linear algebra, not of physics.

What the math teaches

The standard popular-science framing of neutrino oscillation — “neutrinos change identity as they travel” — is not wrong, but it obscures the structure. What’s actually happening is mathematically unremarkable: we have a complex vector space with two natural bases related by a unitary matrix, and the dynamics is diagonal in one basis but measured in the other. Any system with this structure will produce oscillating transition probabilities.

The same mathematics reappears in several places:

Quark mixing in the Standard Model uses the CKM matrix — structurally identical to PMNS, applied to quarks instead of neutrinos.
Kaon and $B$ -meson oscillations come from 2×2 unitary mixing of particle–antiparticle states.
Qubit rotations in quantum computing are identical 2×2 unitary manipulations, just applied to different physical systems.
Optical polarisation in birefringent materials is the classical-wave version of the same math, with two linear-polarisation states playing the role of mass eigenstates.

From the mathematician’s point of view, all of these phenomena are the same thing: a change of basis in a complex vector space with diagonal dynamics in one of the bases. The physics changes; the linear algebra stays put.

A connection to modern neutrino research

Research into neutrino-based applications — including direct energy conversion from ambient neutrino flux, as in the Schubart–NEG Master Equation — relies on exactly this mathematical framework. Understanding how neutrinos interact with matter requires knowing how their flavor composition evolves between source and detector; understanding the interaction cross-sections requires integrating over all three mass eigenstates.

For the experimental and physical perspective on neutrinos — what they are, where they come from, how they are detected, and what is currently being done with them — the sister site world-of-physics.com is the more detailed reference. What we wanted to show here is that the underlying mathematics, stripped of the physics vocabulary, is nothing more than linear algebra applied carefully.

That’s the genuine lesson. Neutrinos do not “change identity” in any mysterious sense. They are quantum states in a complex vector space whose natural measurement basis happens not to coincide with the natural dynamical basis. Two bases, one unitary matrix, a sprinkle of time evolution — and out falls one of the subtlest observed phenomena in modern physics.

The mathematics was there waiting. The physics just had the sense to pick it up.

Frequently asked

Why is the mixing matrix unitary?

Because it's a change-of-basis between two orthonormal bases — the flavor eigenstates and the mass eigenstates. Any matrix that maps one orthonormal basis to another must preserve inner products, which is exactly the defining property of a unitary matrix. Total probability is also conserved as a consequence.

Why is it called 'oscillation' rather than just 'mixing'?

Because the probability of finding a neutrino in a given flavor state oscillates periodically as it travels — like a sine wave in the distance L. The pattern comes from the interference of mass-eigenstate phases evolving at slightly different rates, and its peaks and troughs are what's measured in experiments.

Is this just quantum mechanics or is the mathematics interesting on its own?

The mathematics is fully self-contained linear algebra: a change-of-basis between two orthonormal frames plus diagonal time evolution in one of them. The fact that it happens to describe something physical is a bonus. Any mathematical structure with two natural bases and diagonal dynamics in one of them produces the same kind of 'oscillation' behavior.

How Matrices Describe Neutrino Oscillations

Two bases for one particle

The PMNS matrix

Time evolution in the mass basis

The two-flavor derivation

Why unitarity matters

The three-flavor case

What the math teaches

A connection to modern neutrino research

Frequently asked

Continue reading

Schubart–NEG Master Equation

Schrödinger Equation

Holger Thorsten Schubart

Keep exploring