CP violation

CP violation is the failure of a theory to be invariant under the combination of charge conjugation and parity. Under CPT invariance it is equivalent to time-reversal violation.

CP violation matters experimentally because a symmetry predicts a forbidden transition, not merely a small correction to a permitted one. The Christenson-Cronin-Fitch-Turlay kaon CP violation experiment found evidence for precisely such a transition: a long-lived neutral kaon decaying into two pions. Understanding that inference requires both two-pion kaon event reconstruction and the distinction between kaon CP eigenstates and the physical states of neutral kaon mixing.

James H. Christenson, James Watson Cronin, Val Logsdon Fitch and René Turlay reported the result in 1964. At Brookhaven's AGS, 30 GeV protons struck beryllium; collimators selected a neutral particle beam at 30 degrees. A downstream helium decay volume reduced interactions.^[ref]

The flight distance suppressed the short-lived kaon component. Each arm of a two-arm magnetic spectrometer had spark chambers before and after its magnet. Cherenkov detectors and scintillation counters triggered photographic recording in coincidence.^[ref]

The helium sample yielded $45\pm 9$ signal events; $(2.0\pm 0.4)\times 10^{-3}$ was the ratio to all charged decay channels.^[ref] That denominator must be retained when comparing this result with a total branching fraction.

The evidence was kinematic. Later regenerated-beam quantum interference established that the short- and long-lived states reached the same two-pion final state, excluding alternatives such as distinct pion-like particles.^[ref] The 1964 observation established evidence for CP violation; it did not determine every parameter of the CKM matrix. Inferring time-reversal symmetry violation from CP violation invokes CPT symmetry, rather than a direct comparison of time-reversed transitions in that experiment.

A two-track invariant mass and a direction test answer different questions. In units $c=1$, assign each track the pion mass $m_{\pi }$ and reconstruct its momentum before magnetic deflection. With $p_{\pm }=|{\mathbf{p}}_{\pm }|$ and opening angle $\alpha$,This follows by expanding the squared sum of the two four-momenta. The angle $\theta$ between the reconstructed parent momentum $\mathbf{P}={\mathbf{p}}_{+}+{\mathbf{p}}_{-}$ and the neutral particle beam direction $\hat{\mathbf{n}}$ satisfiesA fully reconstructed two-body decay channel gives the parent mass and $\theta \simeq 0$, within experimental resolution. A three-body decay channel with an unseen neutrino generally misses both constraints; an accidental correct invariant mass alone is insufficient. Notice that the track opening angle $\alpha$ need not be small even when the parent direction angle $\theta$ vanishes.

The 1964 analysis selected $490<M_{\pi \pi }<510$ MeV and remeasured events with $\cos \theta >0.9995$, comparing adjacent mass bands.^[ref]

For an original dimensionless example, take $m_{\pi }=1$, $\hat{\mathbf{n}}=(0,0,1)$ and ${\mathbf{p}}_{\pm }=(\pm \sqrt{3},0,3)$. Each energy is $\sqrt{13}$; the opening angle is $60$ degrees since $\cos \alpha =(9-3)/(9+3)=1/2$. Yet $\mathbf{P}=(0,0,6)$ points exactly along the beam, and $M_{\pi \pi }^2=4(13)-36=16$. Thus a parent of mass $4$ passes both ideal tests. These chosen numbers illustrate four-momentum conservation, not detector performance.

Neutral kaon regeneration is the production of a short-lived component when a neutral kaon beam passes through matter. In the phase convention of kaon CP eigenstates, let the forward transmission amplitudes of $K^0$ and ${\bar{K}}^0$ be $t$ and $\bar{t}$. Expanding $t|K^0⟩+\bar{t}|{\bar{K}}^0⟩$ in the kaon CP eigenstates gives the displayed equation. Different forward scattering amplitudes, including absorption, can therefore generate a two-pion signal even with CP symmetry. Ordinary matter is not a CP-symmetric preparation.

Tungsten regeneration calibrated the signal; scanning its position sampled acceptance. Helium regeneration was insufficient, bag walls lay outside acceptance, vertex distributions excluded collimator regeneration, and hydrogen data cross-checked the effect.^[ref] These controls are central: the question is whether the long-lived component itself reaches two pions, rather than whether material has prepared a fresh short-lived component.

Use the kaon CP eigenstates convention $CP|K^0⟩=-|{\bar{K}}^0⟩$, matching kaon CP mixing parameter. Then $|K_1⟩=(|K^0⟩-|{\bar{K}}^0⟩)/\sqrt{2}$ is CP even, while $|K_2⟩=(|K^0⟩+|{\bar{K}}^0⟩)/\sqrt{2}$ is CP odd. With CPT symmetry and a common small mixing parameter, the physical states can be writtenHere $S,L$ mean short and long mean lifetimes; the kaon CP mixing parameter $ϵ$ depends on the flavor-state phase convention. A positive flavor-exchange convention instead makes the symmetric flavor combination CP even, as explained in exchange-positive CP convention for neutral kaon mixing. Mixing these conventions would reverse the interpretation of the written sums.

Why does the two-pion final state test CP symmetry? A spin-zero kaon decaying into two spin-zero pions has relative orbital angular momentum $\ell =0$ by angular momentum conservation. Two negative intrinsic parities multiply to plus, and $(-1{)}^{\ell }=1$. Charge conjugation exchanges ${\pi }^{+}$ and ${\pi }^{-}$; the $\ell =0$ state is even under this exchange. Hence this two-pion state has CP eigenvalue $+1$. If a decay transition operator $T$ preserves CP symmetry, inserting CP on both sides of $⟨\pi \pi |T|K_2⟩$ changes its sign without changing its value: that decay amplitude must vanish.

Assume for illustration that CP violation occurs only through the state admixture and the decay transition preserves CP symmetry. Put $A_1=⟨\pi \pi |T|K_1⟩$. ThenNeglecting the tiny difference in two-body Lorentz-invariant phase space between the physical masses gives ${\Gamma }_L^{\pi \pi }/{\Gamma }_S^{\pi \pi }\simeq |ϵ{|}^2$. A branching fraction, however, divides a partial decay width by the total decay width. With $\hslash =1$, ${\tau }_i=1/{\Gamma }_i$, soFor chosen values $|ϵ|=0.01$ and ${\tau }_L/{\tau }_S=100$, the branching fraction ratio is $0.01$, whereas the partial decay width ratio is $0.0001$. This is original arithmetic, not a historical fit. The longer observation time amplifies the integrated fraction relative to the instantaneous partial decay rate.

For a coherently prepared state proportional to $\rho |K_S⟩+|K_L⟩$, evolve each component with ${\lambda }_j=m_j-i{\Gamma }_j/2$. The proper time is $t$, and $\hslash =1$. Since both components reach the same final state, their decay amplitudes add before squaring. Factoring out $|A_S{|}^2$ and a common beam normalization gives the two-pion intensityThe last term is quantum interference. It vanishes upon averaging over an independently random preparation phase, but cannot be discarded from a coherent preparation. Summing the two positive contributions alone would lose the phase-sensitive information.

At late proper time, the short-lived contribution and the quantum interference term die away, leaving the long-lived tail $|{\eta }_{\pi \pi }{|}^2{e}^{-{\Gamma }_{L}t}$. Such a tail is easier to identify after the initially abundant short-lived component has disappeared. The useful experimental balance is between eliminating that component and retaining enough long-lived decays to observe a rare decay channel.

Direct CP violation concerns the decay transition itself. At leading order, the charged and neutral two-pion decay amplitude ratios separate a common $ϵ$ from channel-dependent ${ϵ}^{\prime }$.^[ref]

For the corresponding double ratio of partial decay widths,The expansion follows from $(1-2x)/(1+x)=1-3x+O(x^2)$. A single nonzero long-lived two-pion yield does not isolate $x$: it can already occur with ${ϵ}^{\prime }=0$. Thus the 1964 evidence for CP violation must be distinguished from the later experimental establishment of direct CP violation through ${ϵ}^{\prime }$.

A direct CP asymmetry compares decay widths for a process and its CP symmetry conjugate. Let two decay amplitudes contribute with real nonnegative magnitudes $a_1,a_2$, absorptive phases ${\delta }_i$ and weak phases ${\varphi }_i$. Then $\mathcal{A}=a_1{e}^{i({\delta }_1+{\varphi }_1)}+a_2{e}^{i({\delta }_2+{\varphi }_2)}$ and $\overline{\mathcal{A}}=a_1{e}^{i({\delta }_1-{\varphi }_1)}+a_2{e}^{i({\delta }_2-{\varphi }_2)}$. Expanding the absolute squares givesThus two nonzero contributions and both kinds of relative phase are required in this example. Equal total decay widths under CPT symmetry constrain the sum over channels and do not forbid individual channel asymmetries.

A weak phase is a phase in a coupling or decay amplitude that reverses sign in the CP symmetry conjugate process. A physically relevant difference must survive allowed field rephasings; a complex coupling alone does not prove CP violation.

The fermion electric dipole moment operator couples a fermion pseudotensor bilinear to the electromagnetic field tensor. Its nonrelativistic interaction is proportional to $\mathbf{S}\cdot \mathbf{E}$, so it is odd under parity and time reversal and signals CP violation.

For real $a$, the Hermitian effective interaction $ia\bar{\psi }{\sigma }^{\mu \nu }{\gamma }^5\psi F_{\mu \nu }$ is odd under time-reversal symmetry. The spinor tensor and gauge field strength both acquire $-t_{\mu }{t}_{\nu }$, while the chirality matrix is even. Their signs cancel, leaving the minus sign from the explicit $i$ under antiunitarity. By the CPT theorem this is also CP violation. It is not a renormalizable tree-level Standard Model term, but a radiatively generated low-energy dipole is allowed because of the CP-violating phase in the CKM matrix.

A CP-violating phase is a complex phase in couplings or mixing matrices that cannot be removed by field redefinitions and changes sign under CP conjugation.

Multiplying individual fields by constant phases changes the apparent phases of interaction coefficients. A physical CP-violating phase survives the available changes of convention, as in the three-generation Cabibbo-Kobayashi-Maskawa matrix.