Strangeness Oscillations and Regeneration · Elementary Particle Physics

Removing part of a beam creates something that was not there. Regeneration is what makes a superposition impossible to explain away as bookkeeping.

🎯 Why this matters

Every later argument that a superposition is physical rather than a confession of ignorance has this shape: take one component away and watch the other appear. Nothing about ignorance behaves like that.

§8.2 set up three bases for one two-state system: strangeness, CP, and mass. This section is what happens when you prepare a state in the first and let it propagate in the third.

Strangeness oscillation was worked out by Gell-Mann and Pais in 1955, before anyone had seen a $K_L$ . The prediction is startling enough that it is worth doing the algebra rather than quoting the answer.

§8.3 The oscillation, derived

🪜 From a pure K⁰ to a beat — Eqs. (8.13)–(8.18)

Step 1 of 5 — the propagating states

|K_{1,2}^0\rangle = \tfrac{1}{\sqrt2}\left(|K^0\rangle \pm |\bar K^0\rangle\right)

Why you may do this: Only states of definite mass have simple time dependence, e^(−i(m − iΓ/2)t). The flavour states do not — they are not eigenstates of anything that propagates. Provided CP is conserved, the propagating states are the CP eigenstates.

This is the step that decides everything. A state is "simple" only in the basis where the Hamiltonian is diagonal, and that is not the basis strong interactions produce.

Bettini pp. 320–321. The mesons are treated as stable here; damping is put back below.

The angular frequency of the beat is $\Delta m$ itself, in natural units. So the period is

T = \frac{2\pi}{\htmlClass{t-dm}{\Delta m}} = \frac{2\pi \times \htmlClass{t-hb}{6.582\times10^{-7}\;\text{eV ns}}}{\htmlClass{t-dm}{3.481\times10^{-6}\;\text{eV}}} = \htmlClass{t-t}{1.19\;\text{ns}}

The oscillation period, from the mass difference alone. In natural units the beat frequency IS the mass splitting — no other quantity enters.

Every symbol, one at a time

Hover or tap a symbol above — it lights up in the equation and its meaning, units and type appear here.

and for a 10 GeV beam ( $\gamma = 20.1$ ) the first maximum of the $\bar K^0$ appearance sits at $\gamma c T/2 = 3.6$ m from the target. The apparatus is a few metres of empty space.

⚠️ The beat gives |Δm|, never its sign

Every expression above contains $\Delta m$ only inside a cosine or a sine squared, both of which are even. No oscillation measurement can tell you which of $K_S$ and $K_L$ is heavier.

The sign comes from somewhere else entirely. A kaon beam crossing matter has a refractive index — the same physics as light, with forward scattering amplitudes in place of polarizability — and that index depends on $\Delta m$ in both magnitude and sign. The answer is $\Delta m = m_L - m_S > 0$ : the heavier neutral kaon is the longer-lived one.

That is not a definition and not a convention. It is a measurement, and for the $B^0$ the corresponding statement genuinely is a convention ( $\Delta m_B \equiv m_H - m_L > 0$ by construction), which is exactly the kind of thing that gets copied across systems incorrectly.

How do you even see it?

You cannot use the pion channels. $2\pi$ and $3\pi$ select CP, not strangeness — a $K_S \to \pi\pi$ tells you the beam contained a $K_1^0$ , which is true of $K^0$ and $\bar K^0$ equally. To watch strangeness oscillate you need a decay that couples to the flavour basis, and the semileptonic decays do, because of the ΔS = ΔQ rule , which follows directly from having one W vertex:

Bettini Eq. (8.19). The rule is not an extra postulate — it is what a single W vertex can do, and §7.9's problem 7.24 is the same rule bounding Σ⁺ → n e⁺ν.

state↕	quark decay↕	allowed↕	forbidden↕
$K^0 = d\bar s$	$\bar s \to \bar u\, \ell^+ \nu$	$K^0 \to \pi^-\ell^+\nu$	$K^0 \to \pi^+\ell^-\bar\nu$
$\bar K^0 = \bar d s$	$s \to u\, \ell^-\bar\nu$	$\bar K^0 \to \pi^+\ell^-\bar\nu$	$\bar K^0 \to \pi^-\ell^+\nu$

A positive lepton means the meson was a $K^0$ at the moment it decayed; a negative lepton means it was a $\bar K^0$ . And the channels are generous — 39 % and 27 % of $K_L$ decays go to $\pi e\nu$ and $\pi\mu\nu$ — so there is no statistics problem.

💡 What this really says — the same system, two incompatible readouts

It is worth being explicit about what just happened, because it recurs constantly and it is the practical face of the whole two-basis story.

There is one particle in the beam. Which question you can answer about it depends entirely on which decay you look at:

watch $\pi\pi$ or $\pi\pi\pi$ , and you learn its CP;
watch $\pi\ell\nu$ , and you learn its strangeness.

These are not two measurements of the same quantity with different precision. They are measurements of non-commuting observables, and each destroys the information the other would have given. A kaon that decayed to two pions has told you nothing about whether it was a $K^0$ or a $\bar K^0$ — not because the apparatus was inadequate, but because the state it decayed from was an equal superposition of both.

So “how do I observe the oscillation?” is really “which decay channel has an eigenbasis aligned with the question I am asking?” — and the answer, given by the $\Delta S = \Delta Q$ rule, is the semileptonic one.

The same logic governs §8.6’s B-meson measurements, where the tagging decay and the CP decay must be different decays of different mesons, precisely because one meson cannot supply both answers.

Putting the damping back

Real kaons decay, and the shorter lifetime kills the oscillation quickly. The full expressions are Eq. (8.21), which §8.1 already stated:

Survival and appearance — one equation, four systems

survival — still K⁰
appearance — now K⁰-bar
P₊ + P₋ — decay alone, no mixing

Δm ×1.00 — the measured value · drag to see what changing the period alone would do

Δm

5.29e-3 ps⁻¹

period T = 2π/Δm

1.19e+3 ps

cτ (short)

2.68 cm

T / τ — the whole story

13.3

K⁰. The only system whose two eigenstates have wildly different lifetimes — 89.5 ps against 51.2 ns, a factor 571. The oscillation is damped away long before one period is complete, so you see less than a quarter of a cycle.

The dashed curve is pure exponential decay — what would happen with no mixing at all. The two solid curves always sum to it: mixing redistributes the survivors between the two flavours, it does not change how many there are.

The K⁰ preset. Compare with the undamped cos²/sin² above: the appearance probability never gets near 1, because τ_S = 89.5 ps is 13 times shorter than the 1.19 ns period. You are seeing less than a quarter of a cycle.

What is actually measured is the charge asymmetry — the difference between the numbers of positive and negative leptons, normalised to their sum:

\delta(t) = \htmlClass{t-def}{\frac{N(\pi^-\ell^+\nu) - N(\pi^+\ell^-\bar\nu)}{N(\pi^-\ell^+\nu) + N(\pi^+\ell^-\bar\nu)}} \;\simeq\; 2\htmlClass{t-env}{e^{-\Gamma_S t/2}}\htmlClass{t-osc}{\cos(\Delta m\, t)}

What a detector actually records. The flavour of the decaying meson is not observable, but the CHARGE of the lepton it produces is — and that is enough.

Every symbol, one at a time

Hover or tap a symbol above — it lights up in the equation and its meaning, units and type appear here.

a damped oscillation whose envelope is set by half the short width — because the interference term involves one amplitude of each eigenstate, not two.

δ(t) = 2e^(−Γ_S t/2) cos(Δm t)
envelope ±2e^(−Γ_S t/2)
the residual: 2 Re ε = 3.2 × 10⁻³

Fig. 8.2 redrawn (data: Gjesdal et al. 1974). The envelope has collapsed to ±0.08 by t ≈ 0.6 ns, which is why the published plot uses that scale. Roughly two zero crossings fit inside the measurable range — enough to fix both Γ_S and |Δm|.

what the charge asymmetry gives you

import numpy as np
GS, GL, dm = 1/89.54, 1/51.16e3, 5.293e-3        # ps^-1

print("the envelope 2 exp(-Gamma_S t/2), and the asymmetry itself:")
for t_ns in (0.1, 0.3, 0.5, 0.7, 1.0):
    t = t_ns*1000
    a, b = np.exp(-GS*t), np.exp(-GL*t)
    x = 2*np.exp(-(GS+GL)*t/2)*np.cos(dm*t)
    print(f"  t = {t_ns:3.1f} ns   envelope {2*np.exp(-GS*t/2):6.4f}   delta = {x/(a+b):+.3f}")

print("\nso the +-0.08 vertical scale of Fig. 8.2 is simply where the envelope")
print("sits once the K_S has mostly gone -- around t = 0.5 to 0.7 ns.")
print("\nthe fit gives two numbers at once:")
print(f"  the ENVELOPE decay rate   -> Gamma_S, hence tau_S = {1/GS:.1f} ps")
print(f"  the oscillation frequency -> |Delta m| = {dm:.3e} ps^-1")
print(f"  period T = 2 pi / |Delta m| = {2*np.pi/dm/1000:.3f} ns")
print("\nand one number the fit did NOT expect:")
print( "  at late times only K_L survive, both flavour amplitudes should be")
print( "  equal and delta should go to ZERO.  it does not.  it settles at")
print(f"  2 Re(eps) = {2*1.596e-3:.1e} -- the K_L is not a CP eigenstate.")

prints

the envelope 2 exp(-Gamma_S t/2), and the asymmetry itself:
t = 0.1 ns   envelope 1.1442   delta = +0.744
t = 0.3 ns   envelope 0.3745   delta = -0.006
t = 0.5 ns   envelope 0.1226   delta = -0.108
t = 0.7 ns   envelope 0.0401   delta = -0.034
t = 1.0 ns   envelope 0.0075   delta = +0.004

so the +-0.08 vertical scale of Fig. 8.2 is simply where the envelope
sits once the K_S has mostly gone -- around t = 0.5 to 0.7 ns.

the fit gives two numbers at once:
the ENVELOPE decay rate   -> Gamma_S, hence tau_S = 89.5 ps
the oscillation frequency -> |Delta m| = 5.293e-03 ps^-1
period T = 2 pi / |Delta m| = 1.187 ns

and one number the fit did NOT expect:
at late times only K_L survive, both flavour amplitudes should be
equal and delta should go to ZERO.  it does not.  it settles at
2 Re(eps) = 3.2e-03 -- the K_L is not a CP eigenstate.

That last line is the whole of §8.5 arriving early. Look at the green line in the plot: at times long compared with $\tau_S$ the beam is pure $K_L$ , and if the $K_L$ were the CP eigenstate $K_2^0$ it would contain $K^0$ and $\bar K^0$ in exactly equal measure, so $\delta$ would relax to zero. It does not. It settles at $3.3\times10^{-3}$ .

§8.4 Regeneration — the experiment that made it credible

A prediction this strange needs a test that cannot be explained any other way, and Pais and Piccioni proposed one in 1955: regeneration . Piccioni and collaborators performed it in 1960.

The logic is four steps of bookkeeping, and the arithmetic is exact:

🪜 Regeneration, in four projections

Step 1 of 5 — produce a K⁰

|K^0\rangle = \tfrac{1}{\sqrt2}\left(|K_1^0\rangle + |K_2^0\rangle\right) \qquad I = 1

Why you may do this: π⁻p → K⁰Λ is a strong reaction, so it makes a state of definite strangeness — which is an equal superposition of the two things that propagate.

Bettini §8.4, with an idealised absorber that removes K̄⁰ completely and transmits K⁰ untouched. The real one is imperfect, which changes the numbers but not the argument.

🛠️ Fig. 8.3 — the Pais–Piccioni scheme

Click a numbered marker for what that piece does.

Bettini Fig. 8.3, redrawn. Three elements: make a flavour state, wait for the CP state to filter itself, then act on flavour again.

🔬 Experiment card — Pais and Piccioni (proposed 1955); Muller, Birge, Fowler, Good, Hubbard, Matthews, Powell, Ross and Piccioni (1960)

Apparatus

\pi^-

beam on a thin target producing

K^0\Lambda

; a drift space of order a metre in which the short-lived component decays away; a slab of material as an absorber; and a detector downstream watching for two-pion decays. The 1960 realisation used a hydrogen bubble chamber, so production, flight and regeneration were all visible in one photograph.

What is measured

The rate of $2\pi$ decays as a function of distance. It is large just after the target, falls to zero over a few centimetres of

c\tau_S

, stays at zero through the drift space — and then, immediately downstream of the absorber, comes back.

The result

Two-pion decays reappear after the absorber, at roughly an eighth of their original intensity. The short-lived state, which had entirely decayed, is present again.

What it proved

That

K_S

and

K_L

are coherent superpositions of

K^0

and

\bar K^0

, and not a statistical mixture of two kinds of particle. A mixture cannot be regenerated: removing half of a bag of red and blue balls leaves you with fewer balls, not with a new colour. Only amplitudes can be rotated by subtraction.

It also established the Gell-Mann–Pais picture as physics rather than bookkeeping, and it is the reason the whole two-basis language of this chapter is taken literally.

⚙️ Engineer’s bridge — blocking one arm of an interferometer makes a dark fringe bright

This is the deepest idea in the section and it is not really about kaons.

Ask what the beam is in the drift region, after the two-pion decays have stopped. Two descriptions:

(a) a coherent superposition, $\tfrac{1}{\sqrt2}(|K^0\rangle - |\bar K^0\rangle)$ — a pure $K_2^0$ ;
(b) a statistical mixture, half $K^0$ and half $\bar K^0$ , decided particle by particle.

Every measurement that asks only ” $K^0$ or $\bar K^0$ ?” gives the same answer for both. Count semileptonic decays and you get equal numbers of each sign either way.

But the two-pion rate already separates them. In description (b) every single particle is a flavour eigenstate, and a flavour eigenstate is half $K_1^0$ — so a mixture would go on producing $2\pi$ decays for ever, at a rate set by $\tau_S$ and never reaching zero. It is observed to reach zero. In description (a) the two $K_1^0$ amplitudes carried by $|K^0\rangle$ and $-|\bar K^0\rangle$ cancel exactly, and the two-pion channel closes.

So the absence of $2\pi$ decays in the drift region is not an absence of particles. It is an interference null.

And that is what the absorber proves, by destroying it. Remove the $\bar K^0$ amplitude and there is nothing left to cancel against, so the $K_1^0$ component reappears and two-pion decays come back. The sequence — present, absent, present again — is impossible for a mixture, because a mixture that gave $2\pi$ after the absorber would have been giving $2\pi$ before it.

An engineer has seen exactly this. Set up a two-arm interferometer, find a dark fringe, and then block one arm: the dark fringe gets brighter. Nothing was added; a cancellation was removed. Anyone who has debugged a null in a differential measurement knows the same move — you break the symmetry deliberately to prove the null was cancellation and not just absent signal.

The same distinction, in other clothes: a random bit versus a qubit in $|+\rangle$ , indistinguishable in the computational basis and trivially different in the rotated one; incoherent versus coherent light; noise versus signal of unknown phase. In every case, phase information survives in a superposition and is absent in a mixture, and the experiment that recovers it is the one that proves which you had.

Pais and Piccioni’s proposal is that argument made macroscopic, in 1955, with a lump of matter and a metre of empty space.

Where it breaks: an interferometer arm is blocked by a shutter you can open again, and the “which-path” information is destroyed rather than the beam.

Here the absorber works only because nuclear matter treats K⁰ and K̄⁰ differently — the K̄⁰ can make a hyperon on a nucleon and the K⁰ cannot — so a flavour-blind absorber would regenerate nothing at all. The process is also irreversible and lossy in a way a shutter is not: most of the beam is simply absorbed, and the regenerated K_S amplitude is small. So the analogy is exact about what is being demonstrated — superposition versus mixture — and misleading about the cost. You do not block one arm; you destroy most of the beam and read the interference in what survives.

Aside — Question 8.2, and the three-polarizer paradox

Is the analogy with two pairs of perpendicular polarization axes at 45° correct? Design a regeneration experiment with linear polarizers.

The analogy is exact for the basis structure. Take $x, y$ for $K^0, \bar K^0$ and $x', y'$ at 45° for $K_1^0, K_2^0$ ; the relation $|x\rangle = (|x'\rangle + |y'\rangle)/\sqrt2$ is Eq. (8.14) verbatim.

The experiment is the three-polarizer demonstration, run in exactly the order of the StepThrough above:

Send $x$ -polarized light — a pure " $K^0$ ".
Pass it through a polarizer along $y'$ (45°). This removes the $x'$ component, the analogue of the $K_1$ decaying away. Intensity drops to 1/2, and the light is now purely $y'$ .
Now insert a polarizer along $x$ — the analogue of the absorber that removes $\bar K^0$ . Naively this should block everything, since $y'$ light “contains no $x'$ ”. It does not: $y' = (x - y)/\sqrt2$ , so the $x$ -polarizer transmits 1/4 of the original.
Analyse along $x'$ : you find 1/8 of the original intensity — light in a polarization that was completely removed at step 2. Regenerated.

That is the same 1 → ½ → ¼ → ⅛ chain, and it is the classic “insert a third polarizer between two crossed ones and light gets through” demonstration.

Where the analogy stops. In vacuum both light polarizations travel at the same speed, so nothing oscillates — the optical analogue of §8.3 needs a birefringent medium, where the two polarizations have different refractive indices and hence different phase velocities. Different phase velocity is exactly what different mass means here. So: birefringence ↔ $\Delta m$ , and a wave plate is a length of kaon beam.

🔑 If you remember only three things

A metre of vacuum is enough. The oscillation length here is laboratory scale, which is why this system and no other became the testbed.
Absorption is not destruction. Blocking one component leaves a state that is an eigenstate of nothing, so decays reappear that had already stopped.
The experiment was proposed five years before it was performed. The idea was complete in 1955 and the beams were not.

Where this goes next

The charge-asymmetry plot has already given the game away. §8.5 takes the residual seriously: Christenson, Cronin, Fitch and Turlay look for $K_L \to \pi^+\pi^-$ — a decay that CP forbids outright — and find it, two times in a thousand. The $K_S$ and $K_L$ are therefore not the CP eigenstates after all, and the size of the impurity is the parameter $\varepsilon$ .

✅ Check yourself — oscillation and regeneration

0/6 answered · 0 correct

1.Why can't you observe strangeness oscillation by watching the pion decay channels?
2.The measured charge asymmetry δ(t) does not relax to zero at late times. What does that mean?
3.Track the intensity through the Pais–Piccioni experiment. Starting from 1, what fraction ends up as regenerated K₁⁰?
4.Why can no statistical mixture of K⁰ and K̄⁰ reproduce the observed sequence of 2π decay rates?
5.Question 8.2 asks you to design a regeneration experiment with polarizers. What is it, and where does the analogy break down?
6.Every expression for the oscillation contains Δm inside a cosine or a squared sine. What follows?