Strangeness oscillations

A neutral kaon leads a double life. Struck into being by the strong interaction — say in $\pi^- p \to K^0 \Lambda$ — it is born a $K^0$ , a definite quark state $d\bar s$ carrying strangeness $S=+1$ . But strangeness is not conserved by the weak interaction that lets the kaon decay, so the states that propagate and decay with definite mass and lifetime are not $K^0$ and $\overline{K^0}$ at all. They are two superpositions, $K_S$ and $K_L$ — "short" and "long" — with slightly different masses and very different lifetimes. A beam that starts as pure $K^0$ is therefore a coherent sum of $K_S$ and $K_L$ , and because those components tick at slightly different rates, their relative phase winds forward and the flavour content beats: a pure $K^0$ periodically regenerates a $\overline{K^0}$ out of nothing but time. This is the interference that drives neutrino oscillations, transplanted into a system where the two eigenstates also decay — at wildly different rates. The mass splitting $\Delta m = m(K_L) - m(K_S)$ sets the oscillation frequency; the width difference $\Gamma_S - \Gamma_L$ damps it. Their competition is quantified by one dimensionless number, $x = \Delta m\,\tau_S/\hbar \approx 0.47$ , which happens to be less than one — so the short-lived component fades before even a quarter of an oscillation completes, leaving the pattern muted but unmistakably present. The whole story unfolds whether or not CP is violated; strangeness oscillation is the stage on which the far subtler drama of CP violation is later played, but needs none of its own.

Key takeaways

Neutral kaons are produced and decay as flavour eigenstates $K^0,\overline{K^0}$ (strangeness $\pm1$ ) but propagate as the mass eigenstates $K_S,K_L$ .
The mass splitting $\Delta m = m(K_L)-m(K_S) = 3.484\times10^{-12}\,$ MeV drives oscillation with period $T_{\rm osc}=2\pi\hbar/\Delta m \approx 1.19\times10^{-9}\,$ s.
The width difference damps the beat: with $x\equiv\Delta m\,\tau_S/\hbar \approx 0.47 < 1$ , the $K_S$ component dies before a full oscillation, so the structure is real but muted.
Semileptonic decays tag the flavour at decay: only $K^0\to\pi^-e^+\nu_e$ and $\overline{K^0}\to\pi^+e^-\bar\nu_e$ occur, so the lepton charge fixes $s$ vs $\bar s$ (the $\Delta S=\Delta Q$ rule).
CPLEAR at CERN tagged production flavour with the accompanying charged kaon and measured the strangeness asymmetry $A_S(t)=\cos(\Delta m\,t)/\cosh(\Delta\Gamma\,t/2)$ , extracting $\Delta m$ and the CP parameters.
Key idea: production and decay select flavour eigenstates, but propagation is governed by the mass eigenstates $K_S,K_L$ — their mass difference makes the flavour content beat in time, while their decay-rate difference damps the beat.

The core insight

Production and detection select flavour (or CP) eigenstates, but propagation is governed by the mass eigenstates $K_S$ and $K_L$ ; their mass difference makes the flavour content beat in time, while their decay rates damp the beat — and none of it requires CP violation.

Where this lives today

The neutral-kaon oscillation is no longer a curiosity to be discovered — it is a precision instrument. The CPLEAR experiment at CERN's Low-Energy Antiproton Ring (1990–1996) produced tagged $K^0$ and $\overline{K^0}$ from $\bar p p\to K^\mp\pi^\pm K^0/\overline{K^0}$ , using the accompanying kaon's charge to fix strangeness at birth and the semileptonic lepton charge to tag it at decay; from the time-dependent asymmetries it extracted $\Delta m$ , the CP parameters $|\varepsilon|$ and $\arg\varepsilon$ , and — most famously — the first direct observation of time-reversal violation, $A_T=(6.6\pm1.3)\times10^{-3}$ . At Frascati's DA $\Phi$ NE $\phi$ -factory, KLOE turns $\phi\to K_S K_L$ into entangled neutral-kaon pairs and uses their quantum correlations to test CPT symmetry and quantum mechanics itself. And the framework transplants wholesale to heavier flavours: $B^0$ – $\overline{B^0}$ mixing ( $x_d\approx0.77$ ) at BaBar, Belle, and LHCb, the fast $B_s$ oscillations ( $\Delta m_s\approx17.8\ \mathrm{ps}^{-1}$ ) resolved by LHCb, and $D^0$ mixing — all measured with the same $\cos(\Delta m\,t)$ interference first written down for kaons. The open problem these systems bear on is cosmic: the CP violation glimpsed in kaon mixing is a prototype of the mechanism that must explain why the universe is made of matter and not antimatter.

The physical picture

The engine of the effect is a mismatch of bases. The strong interaction that makes a kaon speaks the language of flavour: it produces a clean $K^0$ ( $d\bar s$ , $S=+1$ ) or a clean $\overline{K^0}$ ( $\bar d s$ , $S=-1$ ). The weak interaction that lets a kaon live and die speaks a different language: its eigenstates — the states with a definite mass and a definite exponential lifetime — are the near-CP combinations $K_S$ and $K_L$ . Neither basis is wrong; they are simply rotated relative to each other, and a state prepared in one basis is a superposition in the other. The first figure lays the two bases side by side and shows the $1/\sqrt2$ rotation that connects them.

Once launched together the two mass components evolve at almost the same rate — but "almost" is the whole story. Each carries a phase $e^{-i m t}$ and an amplitude decay $e^{-\Gamma t/2}$ ; the mismatch $\Delta m$ winds their relative phase forward, and the probability of finding a $\overline{K^0}$ — zero at birth — swells and subsides as $\cos(\Delta m\,t)$ turns over. What makes the kaon unlike a textbook two-level oscillator is that the amplitude bleeds away the whole time, overwhelmingly through the $K_S$ channel. Because $x=\Delta m\,\tau_S/\hbar\approx0.47$ is below one, the phase reaches only about half a radian per $K_S$ lifetime, so the short-lived component vanishes before the oscillation completes even a quarter turn. The second figure plots the consequence: the strangeness asymmetry, an almost-cosine whose amplitude is throttled by the $\cosh$ of the width difference.

Figure

The strangeness asymmetry A_S(t) = cos(Δm t)/cosh(Γ_S t/2) versus proper time in units of the K_S lifetime τ_S. With x = Δm τ_S/ℏ = 0.47 the cosine reaches its first zero only at t ≈ 3.3 τ_S (Δm t = π/2), by which point the cosh damping has cut the amplitude to a tenth: less than one oscillation survives, which is why the beat is real but muted

The core equations

Everything begins with the rotation between the two bases. Neglecting the tiny CP-violating admixture (the CP-conserving limit in which $K_S=K_1$ and $K_L=K_2$ are exact CP eigenstates), the flavour states are the symmetric and antisymmetric combinations of the mass states, and vice versa.

|K^0\rangle=\tfrac{1}{\sqrt2}\big(|K_S\rangle+|K_L\rangle\big),\qquad |\overline{K^0}\rangle=\tfrac{1}{\sqrt2}\big(|K_S\rangle-|K_L\rangle\big) \qquad (1)

where: $|K^0\rangle,|\overline{K^0}\rangle$ — flavour (strangeness) eigenstates, $S=\pm1$ ; $|K_S\rangle,|K_L\rangle$ — mass eigenstates, short- and long-lived. Reading it: a definite-strangeness kaon is an equal, coherent mixture of the two propagating states — which is exactly why a single beam contains both lifetimes. Each mass eigenstate propagates as a plane wave in proper time $t$ with its own mass and width. In the kaon rest frame, using natural units where the phase advances as $m t$ and the amplitude falls as $e^{-\Gamma t/2}$ , the two states evolve independently.

|K_S(t)\rangle=e^{-i m_S t}\,e^{-\Gamma_S t/2}|K_S\rangle,\qquad |K_L(t)\rangle=e^{-i m_L t}\,e^{-\Gamma_L t/2}|K_L\rangle \qquad (2)

where: $m_{S,L}$ — masses; $\Gamma_{S,L}=\hbar/\tau_{S,L}$ — total decay widths; $\tau_S=0.895\times10^{-10}\,$ s, $\tau_L=5.12\times10^{-8}\,$ s. Reading it: the intensity of each component decays as $e^{-\Gamma t}$ , while the relative phase of the two accumulates at the rate $\Delta m=m_L-m_S$ — the seed of the oscillation. Projecting the evolved state back onto the flavour basis and squaring gives the probabilities that a kaon born as $K^0$ is found, at proper time $t$ , as a $K^0$ or as its antiparticle. The signs of the interference term are the only difference between them.

P(K^0\!\to\!K^0)=\tfrac14\!\left[e^{-\Gamma_S t}+e^{-\Gamma_L t}+2e^{-(\Gamma_S+\Gamma_L)t/2}\cos(\Delta m\,t)\right] \qquad (3a)

P(K^0\!\to\!\overline{K^0})=\tfrac14\!\left[e^{-\Gamma_S t}+e^{-\Gamma_L t}-2e^{-(\Gamma_S+\Gamma_L)t/2}\cos(\Delta m\,t)\right] \qquad (3b)

where: the first two terms are the decaying populations of the two mass states; the third is the interference — the actual oscillation, damped by the geometric-mean width $(\Gamma_S+\Gamma_L)/2$ . Reading it: at $t=0$ the interference makes $P(K^0)=1$ and $P(\overline{K^0})=0$ ; as $\cos(\Delta m\,t)$ turns over, $\overline{K^0}$ is regenerated out of a beam that started pure. The cleanest observable divides out the overall decay and isolates the beat. The strangeness asymmetry — built from flavour-tagged decay rates — collapses the two probabilities into a single ratio.

A_S(t)=\frac{P(K^0\!\to\!K^0)-P(K^0\!\to\!\overline{K^0})}{P(K^0\!\to\!K^0)+P(K^0\!\to\!\overline{K^0})}=\frac{\cos(\Delta m\,t)}{\cosh\!\big(\tfrac12\Delta\Gamma\,t\big)},\qquad \Delta\Gamma=\Gamma_S-\Gamma_L \qquad (4)

where: $\Delta\Gamma=\Gamma_S-\Gamma_L\approx\Gamma_S$ since $\Gamma_L\ll\Gamma_S$ . Reading it: the numerator is a pure cosine whose frequency is $\Delta m$ , and the denominator is the width-difference damping. The first zero at $\Delta m\,t=\pi/2$ measures $\Delta m$ directly. Finally, the dimensionless ratio that decides whether the oscillation is visible at all — and that resurfaces as the CP-violation "superweak phase" — compares one phase-winding to one $K_S$ lifetime.

x\equiv\frac{\Delta m}{\Gamma_S}=\frac{\Delta m\,\tau_S}{\hbar}\approx0.47,\qquad \tan\phi_{SW}=\frac{2\Delta m}{\Delta\Gamma}\approx 2x\approx0.95 \qquad (5)

where: $x$ — mixing parameter; $\phi_{SW}$ — superweak phase, the argument of the CP parameter $\varepsilon$ . Reading it: $x<1$ means less than a radian of phase accrues per $K_S$ lifetime, so the beat is muted; the same numbers fix $\phi_{SW}\approx43.5^\circ$ .

Derivation AFrom two mass eigenstates to a beating flavour content

The goal here is to start from nothing but the basis rotation (1) and the free evolution (2) and derive the oscillation probabilities (3a)–(3b), making explicit where the interference term comes from and why it carries the frequency $\Delta m$ and the damping $(\Gamma_S+\Gamma_L)/2$ . This is the neutral-kaon analogue of the neutrino oscillation calculation, with the essential new feature that both eigenstates decay, so amplitudes are complex numbers with shrinking modulus, not just winding phase.

Step 1. Take a kaon born at $t=0$ as pure $K^0$ and write it in the mass basis using (1): $|\psi(0)\rangle=\tfrac{1}{\sqrt2}(|K_S\rangle+|K_L\rangle)$ . Why: the production process fixes the flavour, but only the mass states have simple time evolution, so we must express the initial condition in the basis that diagonalises the Hamiltonian.

Step 2. Evolve each component with (2): $|\psi(t)\rangle=\tfrac{1}{\sqrt2}\big(e^{-im_St-\Gamma_St/2}|K_S\rangle+e^{-im_Lt-\Gamma_Lt/2}|K_L\rangle\big)$ . Why: the two mass states are eigenstates of the effective (non-Hermitian) Hamiltonian, so each simply picks up its own complex phase — the non-Hermiticity is what makes $\Gamma$ appear alongside $m$ .

Step 3. Project onto $\langle K^0|$ using $\langle K^0|K_S\rangle=\langle K^0|K_L\rangle=\tfrac{1}{\sqrt2}$ from (1): the amplitude is $\langle K^0|\psi(t)\rangle=\tfrac12\big(e^{-im_St-\Gamma_St/2}+e^{-im_Lt-\Gamma_Lt/2}\big)$ . Why: the observable is a flavour, so we must return to the flavour basis at the moment of detection — the two bases meet only at production and decay.

Step 4. Take the modulus squared. The cross term is $2\,e^{-(\Gamma_S+\Gamma_L)t/2}\,\mathrm{Re}\,e^{-i(m_S-m_L)t}=2\,e^{-(\Gamma_S+\Gamma_L)t/2}\cos(\Delta m\,t)$ , giving (3a); the projection onto $\langle\overline{K^0}|$ carries a relative minus sign from (1), flipping the cross term to give (3b). Why: squaring a sum of two complex amplitudes always produces a cross term, and that cross term — surviving only while both components are still present — is the oscillation itself. *Verified: $P(K^0\!\to\!K^0)+P(K^0\!\to\!\overline{K^0})=\tfrac12(e^{-\Gamma_St}+e^{-\Gamma_Lt})$ (interference cancels in the sum, as it must); at $t=0$ , $P(K^0)=1$ , $P(\overline{K^0})=0$ .*

Go deeper · Derivation B · 1 minThe strangeness asymmetry and reading off the mass difference

Derivation BThe strangeness asymmetry and reading off the mass difference

The aim is to reduce the two decaying, oscillating probabilities to the single clean observable (4), $A_S(t)=\cos(\Delta m\,t)/\cosh(\tfrac12\Delta\Gamma\,t)$ , and to show how its first zero converts a measured proper time into the mass splitting $\Delta m$ — the quantity that the CPLEAR experiment ultimately reported. The trick is that forming the normalised difference cancels the common decay envelope and repackages the two exponentials as a hyperbolic cosine.

Step 1. Form the difference of (3a) and (3b): the two decaying-population terms cancel, leaving $P(K^0)-P(\overline{K^0})=e^{-(\Gamma_S+\Gamma_L)t/2}\cos(\Delta m\,t)$ . Why: the populations are flavour-blind — they carry no information about which flavour survives — so subtracting isolates the interference that does.

Step 2. Form the sum: $P(K^0)+P(\overline{K^0})=\tfrac12(e^{-\Gamma_St}+e^{-\Gamma_Lt})=e^{-(\Gamma_S+\Gamma_L)t/2}\cosh\!\big(\tfrac12\Delta\Gamma\,t\big)$ , using $e^{-\Gamma_St}+e^{-\Gamma_Lt}=e^{-(\Gamma_S+\Gamma_L)t/2}(e^{-\Delta\Gamma t/2}+e^{+\Delta\Gamma t/2})$ . Why: the sum is the total surviving population, and rewriting it as a cosh exposes the same geometric-mean envelope that multiplies the cosine, so the envelope will cancel in the ratio.

Step 3. Divide: the envelope $e^{-(\Gamma_S+\Gamma_L)t/2}$ cancels top and bottom, leaving $A_S(t)=\cos(\Delta m\,t)/\cosh(\tfrac12\Delta\Gamma\,t)$ . Why: this is the cleanest possible observable — a bare cosine of the oscillation phase, throttled only by the width difference, with the overall lifetime scaled away.

Step 4. Read off $\Delta m$ : the cosine (hence $A_S$ ) first vanishes when $\Delta m\,t=\pi/2$ . Measuring that proper time and inverting gives $\Delta m$ with no need to know the absolute normalisation. Why: a zero-crossing is a phase condition, immune to the overall rate and efficiency uncertainties that plague an amplitude measurement — which is why oscillation experiments quote frequencies so precisely. *Verified: $A_S(0)=1$ ; first zero at $\Delta m\,t=\pi/2$ ; with $x=0.47$ this lands at $t\approx3.3\,\tau_S$ , matching the figure.*

Go deeper · Derivation C · 1 minThe superweak phase from the width difference

Derivation CThe superweak phase from the width difference

This derivation ties the same two numbers $\Delta m$ and $\Delta\Gamma$ to the phase $\phi_{SW}$ of the CP-violation parameter $\varepsilon$ , showing that a quantity introduced for oscillations reappears, unbidden, in the theory of CP violation. The point is structural: the effective Hamiltonian's off-diagonal elements $M_{12}$ and $\Gamma_{12}$ set both the eigenvalue splitting and the mixing parameter $\varepsilon$ , so the observed splitting alone predicts the phase that CP measurements must respect.

Step 1. Write the complex eigenvalue splitting of the effective Hamiltonian $H=M-\tfrac{i}{2}\Gamma$ : the two eigenvalues differ by $\Delta\lambda=\Delta m-\tfrac{i}{2}\Delta\Gamma$ . Why: $M$ and $\Gamma$ are the Hermitian and anti-Hermitian parts of the same $2\times2$ matrix, so the physical splitting is a single complex number whose real and imaginary parts are $\Delta m$ and $-\tfrac12\Delta\Gamma$ .

Step 2. The CP parameter satisfies $\varepsilon\propto \dfrac{\mathrm{Im}\,M_{12}-\tfrac{i}{2}\mathrm{Im}\,\Gamma_{12}}{\Delta m-\tfrac{i}{2}\Delta\Gamma}$ ; its phase is dominated by the denominator, $\arg\varepsilon\approx\arg\big(\Delta m-\tfrac{i}{2}\Delta\Gamma\big)^{-1}=\arctan\!\dfrac{\tfrac12\Delta\Gamma}{\Delta m}$ inverted, giving $\tan\phi_{SW}=2\Delta m/\Delta\Gamma$ . Why: the "superweak" phase is a pure kinematic consequence of the eigenvalue denominator, independent of the detailed CP-violating dynamics in the numerator — so it is fixed by oscillation data alone.

Step 3. Insert the numbers using $\Delta\Gamma\approx\Gamma_S$ and $x=\Delta m/\Gamma_S=0.47$ : $\tan\phi_{SW}=2x=0.95$ , so $\phi_{SW}=\arctan(0.95)\approx43.5^\circ$ . Why: this famous value is not measured from CP physics — it is predicted from the mass and width differences, and its agreement with the measured $\arg\varepsilon$ is a nontrivial consistency check of the whole picture. *Verified: $\tan\phi_{SW}=2\times0.474=0.948\Rightarrow\phi_{SW}=43.5^\circ$ , the standard PDG superweak phase.*

Limiting cases

Case	Result	Meaning
Just born, $t=0$	$P(K^0)=1,\ P(\overline{K^0})=0,\ A_S=1$	pure flavour eigenstate; no mixing yet
Very short times, $t\ll\tau_S$	$P(\overline{K^0})\approx\tfrac14(\Delta m\,t)^2$	$\overline{K^0}$ grows quadratically — the leading regeneration of the antiparticle
CP-conserving limit	$K_S=K_1,\ K_L=K_2$ exactly, $\varepsilon=0$	oscillation still occurs — it needs only $\Delta m,\Delta\Gamma$ , not CP violation
Late times, $t\gg\tau_S$	$P(K^0)\approx P(\overline{K^0})\approx\tfrac14 e^{-\Gamma_L t}$	$K_S$ gone; a $50/50$ flavour mix of pure $K_L$ remains
Degenerate masses, $\Delta m\to0$	$A_S\to 1/\cosh(\tfrac12\Delta\Gamma t)$ , no beat	no phase winding; flavour content decays but does not oscillate
Equal widths, $\Delta\Gamma\to0$	$A_S\to\cos(\Delta m\,t)$ , undamped	an ideal neutrino-like oscillation — the kaon's damping is what mutes it

Common trapStrangeness oscillations require CP violation

They do not — this is the most common misconception about the neutral-kaon system. The oscillation $K^0\rightleftharpoons\overline{K^0}$ arises purely from the mass and width differences of the two propagating eigenstates, and survives intact in the CP-conserving limit where $K_S=K_1$ and $K_L=K_2$ are exact CP eigenstates with $\varepsilon=0$ . Equations (3)–(4) contain only $\Delta m$ and $\Delta\Gamma$ ; nowhere does $\varepsilon$ appear. CP violation is a separate, far smaller effect ( $|\varepsilon|\approx2.2\times10^{-3}$ ) layered on top — but the beat itself is CP-blind. The historical order confirms it: oscillations and regeneration were predicted (1955–56) years before CP violation was discovered (1964), because they are logically prior to it.

Worked examples

1 · The oscillation period and the mixing parameter (fully worked)

Given $\Delta m = m(K_L)-m(K_S)=3.484\times10^{-12}\,$ MeV, $\tau_S=0.8954\times10^{-10}\,$ s, and $\hbar=6.582\times10^{-22}\,$ MeV·s, find the oscillation period $T_{\rm osc}$ , the mixing parameter $x=\Delta m\,\tau_S/\hbar$ , and how many $K_S$ lifetimes fit into one oscillation.

Step 1. The oscillation period is set by the phase winding at rate $\Delta m/\hbar$ : one full cycle takes $T_{\rm osc}=2\pi\hbar/\Delta m$ . Substituting, $T_{\rm osc}=2\pi(6.582\times10^{-22})/(3.484\times10^{-12})=4.135\times10^{-21}/3.484\times10^{-12}=\mathbf{1.19\times10^{-9}\ s}$ . This is about one nanosecond — long on nuclear timescales.

Step 2. The mixing parameter compares one radian of phase to one $K_S$ lifetime: $x=\Delta m\,\tau_S/\hbar=(3.484\times10^{-12})(0.8954\times10^{-10})/(6.582\times10^{-22})=3.120\times10^{-22}/6.582\times10^{-22}=\mathbf{0.474}$ . Because $x<1$ , less than one radian of oscillation phase accumulates before the short-lived component has largely decayed.

Step 3. Compare the two clocks: $T_{\rm osc}/\tau_S=(1.19\times10^{-9})/(0.8954\times10^{-10})=\mathbf{13.3}$ . So a full oscillation would take about thirteen $K_S$ lifetimes — but the $K_S$ population falls by $e^{-13}\approx2\times10^{-6}$ over that span. The beat is real but heavily muted: the $K_S$ amplitude is gone long before one oscillation completes, which is exactly why a metre-scale kaon beam becomes a nearly pure $K_L$ beam.

2 · The superweak phase (fully worked)

Using $\Delta\Gamma\approx\Gamma_S=\hbar/\tau_S$ , predict the phase $\phi_{SW}$ of the CP parameter $\varepsilon$ from the oscillation data alone.

Step 1. Compute the $K_S$ width: $\Gamma_S=\hbar/\tau_S=(6.582\times10^{-22})/(0.8954\times10^{-10})=\mathbf{7.35\times10^{-12}\ MeV}$ . Since $\tau_L$ is $570\times$ longer, $\Gamma_L=1.29\times10^{-14}\,$ MeV is negligible and $\Delta\Gamma\approx\Gamma_S$ .

Step 2. Form the tangent from equation (5): $\tan\phi_{SW}=2\Delta m/\Delta\Gamma=2(3.484\times10^{-12})/(7.35\times10^{-12})=\mathbf{0.948}$ . Equivalently $\tan\phi_{SW}=2x=2(0.474)=0.948$ — the same number two ways.

Step 3. Invert: $\phi_{SW}=\arctan(0.948)=\mathbf{43.5^\circ}$ . This matches the measured argument of $\varepsilon$ to well within errors — a striking consistency check, because $\phi_{SW}$ is derived from masses and lifetimes yet must be respected by an independent set of CP-violation measurements.

3 · How much antiparticle appears at early times

A beam is pure $K^0$ at $t=0$ . Estimate the probability of finding a $\overline{K^0}$ after one-tenth of a $K_S$ lifetime, $t=0.1\,\tau_S$ .

Step 1. For $t\ll\tau_S$ the exponentials are all $\approx1$ and $\cos(\Delta m\,t)\approx1-\tfrac12(\Delta m\,t)^2$ , so (3b) gives $P(\overline{K^0})\approx\tfrac14(\Delta m\,t)^2=\tfrac14(x\,t/\tau_S)^2$ — the antiparticle content grows quadratically from zero.

Step 2. Substitute $x=0.474$ , $t/\tau_S=0.1$ : $P(\overline{K^0})\approx\tfrac14(0.0474)^2=\mathbf{5.6\times10^{-4}}$ . Only about one part in two thousand of the beam has oscillated into $\overline{K^0}$ this early — the quadratic onset is why the effect needed sensitive tagging experiments to see cleanly.

Physics of the portrait

The neutral-kaon story has two founders who realised, before any data demanded it, that a particle and its antiparticle could quantum-mechanically mix. One of them put his name on both landmark predictions — the two-lifetime mixing and its regeneration in matter.

Origins · 1 minAbraham Pais and how this was discovered

Abraham Pais

1918–2000 · neutral-kaon mixing, the prediction of two kaon lifetimes, and regeneration

Abraham Pais was the theorist who, more than anyone, taught particle physics to think of the neutral kaon as a two-state quantum system. Born in Amsterdam, he survived the Nazi occupation in hiding and after 1946 became a fixture at the Institute for Advanced Study in Princeton and later Rockefeller University. In 1955, with Murray Gell-Mann, he made the audacious prediction that because the weak interaction does not conserve strangeness, the physical particles of definite mass and lifetime cannot be $K^0$ and $\overline{K^0}$ but must be their orthogonal combinations $K_1$ and $K_2$ — one short-lived and decaying to two pions, one long-lived and forbidden (by CP, as later understood) from doing so. This was the first time anyone proposed that a neutral particle mixes with its own antiparticle, and it predicted a second, long-lived neutral kaon that had never been seen; it was duly found at Brookhaven the following year. Pais then, with Oreste Piccioni, predicted regeneration: a $K_L$ beam passing through matter would coherently rebuild a $K_S$ component, because $K^0$ and $\overline{K^0}$ scatter differently from nuclei. Both ideas rest on the same insight — that flavour and mass are different bases — and both are the direct ancestors of the oscillation formulae here. Pais also coined the term "baryon" and in later life became the field's pre-eminent historian. The neutral-kaon system he opened remains, seventy years on, one of the most sensitive laboratories for CP violation, CPT symmetry, and the foundations of quantum mechanics.

History

1947

— Rochester and Butler photograph neutral "V-particle" decays in cosmic rays

1955

— Gell-Mann and Pais predict K⁰–K̄⁰ mixing and two distinct kaon lifetimes

1956

— Pais and Piccioni predict regeneration; the long-lived K_L is found at Brookhaven

1964

— Christenson, Cronin, Fitch and Turlay discover CP violation via K_L → ππ

1998

— CPLEAR at CERN measures the time-dependent strangeness asymmetry and Δm

Going further

Texts. Thomson, Modern Particle Physics, §14.5 (strangeness oscillations, CPLEAR, and CP violation in the kaon system) · Griffiths, Introduction to Elementary Particles, 2nd ed., §4.4 and §12 (neutral kaons and CP violation) · Perkins, Introduction to High Energy Physics, 4th ed., §7 · Bigi & Sanda, CP Violation, Ch. 6 (the neutral-kaon complex) · Branco, Lavoura & Silva, CP Violation, Ch. 15.

Provenance. The basis rotation (1) and free evolution (2) are the Gell-Mann–Pais construction recalled from aq-pp-neutral-kaon-system; the interference structure of (3)–(4) mirrors the treatment developed in aq-pp-two-flavour-oscillations, with decaying amplitudes added; the semileptonic $\Delta S=\Delta Q$ tag rests on the underlying transitions of aq-pp-weak-quark-interactions. A verifier should confirm: (1) $x=\Delta m\,\tau_S/\hbar=0.47$ from $\Delta m=3.484\times10^{-12}\,$ MeV, $\tau_S=0.8954\times10^{-10}\,$ s; (2) $T_{\rm osc}=2\pi\hbar/\Delta m=1.19\times10^{-9}\,$ s; (3) $\phi_{SW}=\arctan(2\Delta m/\Gamma_S)=43.5^\circ$ ; (4) $A_S(t)=\cos(\Delta m\,t)/\cosh(\tfrac12\Delta\Gamma\,t)$ with first zero at $\Delta m\,t=\pi/2$ ( $t\approx3.3\,\tau_S$ ). All masses and lifetimes are standard PDG values.

Prerequisites. The Neutral Kaon System · Two-Flavour Oscillations · Weak Quark Interactions · Antiparticles. Leads to. CP Violation in Kaons · B-Meson Physics · CP Violation in the Standard Model.

Ask the tutor: Ask me to walk through why the strangeness asymmetry is a cosine damped by a cosh, or to redo the whole calculation for $B^0$ mesons where $x\approx0.77$ and a full oscillation is actually visible.

Do strangeness oscillations require CP violation?

No, and this is the point most worth internalising. The oscillation is generated entirely by the mass difference $\Delta m$ and the width difference $\Delta\Gamma$ between the two propagating states $K_S$ and $K_L$ . In the idealised CP-conserving world, $K_S$ and $K_L$ would be exact CP eigenstates $K_1$ and $K_2$ , the CP parameter $\varepsilon$ would be zero, and yet equations (3)–(4) would be completely unchanged: a beam born as $K^0$ would still beat into $\overline{K^0}$ and back. CP violation is a separate, roughly thousand-times-smaller effect layered on top. Historically the oscillation and its regeneration were predicted in 1955–56, nearly a decade before CP violation was found in 1964.

Why is the kaon oscillation so muted compared with neutrino oscillations?

Because the kaon's two eigenstates decay, and at very different rates. A neutrino's mass eigenstates are (essentially) stable, so their interference persists over enormous baselines and many full cycles. The kaon's short-lived component, $K_S$ , is gone after about one $K_S$ lifetime, while a full oscillation takes some thirteen $K_S$ lifetimes ( $x=\Delta m\,\tau_S/\hbar\approx0.47<1$ ). The interference term in the asymmetry is therefore choked by a $\cosh(\tfrac12\Delta\Gamma\,t)$ in the denominator, and the beat barely reaches its first zero before it damps away. The physics is identical; the difference is that kaons are unstable and neutrinos are (for these purposes) not.

What is the difference between K_S/K_L and K_1/K_2?

$K_1$ and $K_2$ are defined by CP: $K_1$ is CP-even (allowed to decay to two pions), $K_2$ is CP-odd (must decay to three, hence longer-lived). $K_S$ and $K_L$ are defined by mass and lifetime — they are the actual eigenstates of the effective Hamiltonian, the states that propagate cleanly. If CP were an exact symmetry the two pairs would coincide, $K_S=K_1$ and $K_L=K_2$ . Because CP is very slightly violated, the true mass eigenstates are tiny admixtures, $K_S\approx K_1+\varepsilon K_2$ and $K_L\approx K_2+\varepsilon K_1$ , with $|\varepsilon|\approx2.2\times10^{-3}$ . Strangeness oscillations work with the CP-conserving identification; CP violation lives in the $\varepsilon$ correction.

How does a semileptonic decay tell you the flavour at decay?

Through the $\Delta S=\Delta Q$ rule of the weak interaction. A $K^0$ is $d\bar s$ ; its strange antiquark decays as $\bar s\to\bar u\,e^+\nu_e$ , and with the spectator $d$ this yields $K^0\to\pi^-e^+\nu_e$ — a positive lepton. A $\overline{K^0}$ is $\bar d s$ ; here $s\to u\,e^-\bar\nu_e$ gives $\overline{K^0}\to\pi^+e^-\bar\nu_e$ — a negative lepton. The Standard Model forbids the "wrong-sign" combinations at tree level, so the charge of the lepton is a perfect, instantaneous flavour tag: see a $\pi^-e^+\nu$ and the kaon was a $K^0$ at that instant; see $\pi^+e^-\bar\nu$ and it was a $\overline{K^0}$ . That single fact is what makes the oscillation measurable.

What is regeneration, and how does it relate to oscillation?

Regeneration is the matter-induced cousin of oscillation. Far downstream, a beam has decayed to essentially pure $K_L$ — an equal mix of $K^0$ and $\overline{K^0}$ . Passing it through a slab of material, the two flavour components scatter differently from nuclei (the $\overline{K^0}$ has extra channels, like producing hyperons), so they are attenuated and phase-shifted unequally, and a coherent $K_S$ component is regenerated — announced by two-pion decays behind the regenerator. Oscillation is driven by the mass difference in vacuum; regeneration by the scattering difference in matter — but both exploit the same fact that flavour and propagation are different bases.

Is this the same physics as B-meson and D-meson mixing?

Yes — the neutral-kaon system is the template. Any neutral meson that is not its own antiparticle ( $K^0$ , $B^0$ , $B_s^0$ , $D^0$ ) mixes through the same effective $2\times2$ Hamiltonian, evolves by the same equation (2), and shows the same $\cos(\Delta m\,t)$ interference. Only the numbers differ: for $B^0$ the mixing parameter is $x_d\approx0.77$ (a clean oscillation visible before decay), for $B_s$ it is very large ( $x_s\approx27$ , many fast cycles), and for $D^0$ it is tiny. The kaon sits at the awkward middle, $x\approx0.47$ , where the width difference is comparable to the mass difference — which makes it both historically first and formally the richest of the four.