Recent results on kaon rare decays from the NA48/2 experiment

High Energy Physics in the LHC EraDecember 11-15 2006, Valparaiso – Chile

Massimiliano Fiorini(Università degli Studi di Ferrara – INFN

Ferrara)

on behalf of the NA48/2 Collaboration:Cambridge, CERN, Chicago, Dubna, Edinburgh, Ferrara,

Firenze, Mainz, Northwestern, Perugia, Pisa, Saclay, Siegen, Torino, Vienna

Outline The NA48/2 experiment at CERN SPS The K± π± π0 γ decay (Kπ2γ)

Matrix element contributions Experimental status NA48/2 preliminary measurement of Direct

Emission and Interference Terms

The K± π0 π0 e± ν decay (Ke400)

Introduction and events selection NA48/2 preliminary measurement of Branching

Ratio and Form Factors Conclusions

The NA48/2 simultaneous beams

K+

K−

BM

PK spectra, 603 GeV/c

54 60 66

Width ~ 5 mm

K+/K- ~ 1.8

p/p ~ 1 %

x,y ~ 100 m

SPS protons @ 400 GeV/c

Simultaneus, unseparated, focused beams

NA48/2 detector

Spectrometer:Spectrometer:

LKR calorimeterLKR calorimeter

HodoHodo, AKL MUVMUV, HAC HAC KabesKabes

0.044%1.02%

p

p p

3.2 0.2 % 0.09 0.01

0.42 0.05 %

0.40.05

300t

E

E EE

cmx y cm

Eps

NA48/2 data taking periods 20032003 run: ~ 50 days

20042004 run: ~ 60 days

Total statistics in 2 years: K π+π-π± : > 3·109

K π0π0π± : > 1·108

Rare K± decays:BR’s down to 10–9

can be measured

>200 TB of data recorded

A view of the NA48/2 beam line

Primary goal:

Search for CP-violating charge asymmetries in K± 3 decays (see Dmitri Madigozhin talk on Friday Plenary Session)

K±→π±π0γ decay

Gamma production mechanism

DEDEIBIB

IBIB DEDE

IBIB INTINT DEDE

Γ± depends on 2 variables (W e T*π), reduced to only W by

integration

With this parametrization the ratio data/MC(IB) has the form 1+αW2+βW4

P*K = 4-momentum of the K±

P*π = 4-momentum of the π±

P*γ = 4-momentum of the

radiative γ

2

****2

)(

))((

mm

PPPPW

k

K Lorentz invariant

variable:

W distributions for IB, DE, INT

IB and DE are well separated in W

Inner Bremsstrahlung(IB) : (2.75Inner Bremsstrahlung(IB) : (2.75±0.15)±0.15)××1010-4 -4 PDG 2006 PDG 2006

(55<T(55<T**ππ<90 MeV)<90 MeV)

Direct Emission (DE)Direct Emission (DE) : (4.4 : (4.4±0.8±0.8))××1010-6 -6 PDG 2006PDG 2006

(55<T(55<T**ππ<90 MeV)<90 MeV)

Interference (INT)Interference (INT) : not yet : not yet measuredmeasured

Kπ2γ amplitudes Two amplitudes:

Inner Bremsstrahlung (IB) Calculable: QED corrections to K± π±π0

Suppressed by ΔI=1/2

Direct Emission (DE) Insight on weak vertex structure Electric contribution (E) comes from L4 ChPT lagrangian and loops

L2 (non predictable) Magnetic contribution (M) has contributions from reducible chiral

anomaly (WZW calculable) and also direct contributions (non predictable)

Interference (INT) possible between IB and electric part of DE Measuring at the same time DE and INT gives measurement of

both M and E In addition CPV could appear in INT

Present experimental results seem to suggest a M dominated DE

Experimental measurements

All the measurements have been performed:

in the T*π region 55-90 MeV to avoid π± π0 π0 and π± π0 BG

assuming INT term = 0

BNL E787

KEK E470

Interference measurements:

Experiment

Year

# Events

BR(DE) × 106

E787E470E787E470ISTRA

2000

2003

2005

2006

2006

198364434

2057110154

930

4.7 ± 0.8 ± 0.3

3.2 ± 1.3 ± 1.0

3.5 ± 0.6 +

0.3-0.4

3.8 ± 0.8 ± 0.7

3.7 ± 3.9 ± 1.0

History of DE Branching ratio

NA48/2 has analyzed > 5 times more events

What’s new in NA48/2 measurement

In-flight Kaon decays Both K+ and K- in the beam (possibility to check CP

violation) Very high statistics (220K π±π0γ candidates of which

124K used in the fit ) Enlarged T*π region in the low energy part (0 < T*π < 80

MeV) Negligible background contribution < 1% of the DE

component Good W resolution mainly in the high statistic region More bins in the fit to enhance sensitivity to INT Order ‰ γ mistagging probability for IB, DE and INT Fit with free interference term

Enlarged T*region

T*π(IB) T*

π(DE) T*π(INT)

Standard region Standard

region

Standard region

Using standard region 55 < T*π < 90 MeV is a safe choice for background rejection (π±π0π0 and π±π0)

But the region below 55 MeV is the most interesting for DE and INT measurements

The presented measurement is performed in the region 0 < T*π< 80 MeV to improve statistics and sensitivity to DE

K±→π±π0γ selectionTrack Selection

• # tracks = 1.

• Pπ+ > 10 GeV

• E/p < 0.85

• No muon hits

• 0 MeV < T*π+ < 80 MeV

Gamma selection

• Nγ = 3. (well separated in time LKr clusters)

• Minimum γ energy > 3 GeV (>5 for the fit)

Gamma tagging optimization

• CHA and NEU vertex compatibility

• Only one compatible NEU vertex

BG rejection cuts

• COG < 2 cm

• Overlapping γ cuts

• |MK-MKPDG| < 10 MeV

200K events

Reconstruction strategy

ZV(CHA)

γ1

γ3

γ2

ZV(2-3)

ZV(1-3)

ZV(NEU)

LKr

We can get two independent determination of the K decay vertex: - The charged vertex ZV(CHA) using the K and π flight directions (DCH) - The neutral vertex ZV(NEU) imposing π0 mass to gamma couples (LKr)

For the neutral vertex we have 3 values: ZV(12), ZV(23), ZV(13)

We evaluate the kaon mass for all of them and then choose the vertex giving the best kaon mass.

Once the neutral vertex has been chosen we also know what the radiated γ is.

Main BG sources

Physical BG rejection: For π±π0 we can rely on the cut in T*

π < 80 MeV, MK and COG cuts

For π±π0π0 we have released the T*π cut, but we can anyway

reach required rejection with kaon mass cut (missing γ) and overlapping γ cut (fused γ)

Accidental BG rejection: mainly π±π0, Ke3(π0e±ν) and Kμ3(π0μ±ν) Clean beam, very good time, space, and mass resolutions.

Decay BR Background mechanism

K±→π±π0 (21.13±0.14)%

+1 accidental or hadronic extra cluster

K±→ π±π0π0 (1.76±0.04)%

-1 missing or 2 overlapped gammas

K±→π0e±ν (4.87±0.06)%

+1 accidental γ and e misidentified as a π

K±→ π0μ±ν (3.27±0.06)%

+1 accidental γ and μ misidentified as a π

K±→ π0e± ν(γ) (2.66±0.2)·10−4

e misidentified as a π

K±→ π0μ±ν(γ) (2.4±0.85)·10−5

μ misidentified as a π

Fused rejection:overlapping

cut

Split 1 out of the 3 clusters in two γ of energies:εγ1=xECL εγ2=(1-x)ECL

now we got 4 γ’s and we start reconstructing the event as a π± π0 π0.Evaluate the ZV(x) pairing the gammas and extract x imposing that: Zv(π0

1)= Zv(π02) same K decay vertex: the 2π0 came from

the same K!

Zv(π02)

Zv(π01)

Put x back in the Zv(π02) to get the real ZV(neu)

If the |Zv(CHA)-ZV(neu)| < 400 cm the γ are really fused and so we discard the event

Fused gamma events are very dangerous BG source: - MK and COG cuts automatically satisfied! - Releasing the T*

π cut they can give sizable contributionNA48 LKr is very good to reject them: - very high granularity (2×2 cm) cells - very good resolution in Z vertex

Multi step algorithm looped over clusters:

BG rejection performance

Source %IB %DE

π± π0 π0 ~1·10-4 ~0.61·10-2

Accidentals

<0.5·10-4

~0.3·10-2

total background < 1% DE component

Selected

region 220Kevents

K± π±π0π0 K± π±π0γ

All physical background can be explained in terms of the π±π0π0 events only.

Very small contribution from accidentals

0,000

0,500

1,000

1,500

2,000

2,500

3,000

0 50 100 150 200 250 300 350 400 450 500 550

Mistagging cut (cm)

mis

tagg

ing

prob

%

■ DE▼IB

The mistagged events: a self BG

The mistagging probability has been evaluated in MC as a function of the mistagging cut to be 1.2‰ at 400 cm

Mistagging problem: mistagged gamma events

behave like BG because they can induce fake shapes in the W distribution

they tend to populate the region of high W simulating DE events

must keep mistagging probability as small as possible

Simply demanding compatibility between zvc and best zvn gives 2.5% mistagging probability

Solution: Reject events with a second solution for neutral vertex close to best one

Require |zvn (second)-zvn (best)| Dzvn> xx cm

The K± π±π0 decay. Trigger

L1 trigger Require one track and LKr

information (peaks) compatible with at least 3 clusters

This introduces an energy dependence distortion of W distribution

Correction found using all 3 events (K± π± π0π0 with lost) and applied to Monte Carlo

L2 trigger (rejects K± π±π0)

Using DCH information and assuming 60 GeV kaon along z axis on-line processors compute a sort of missing mass of the K-π system

Cut events whose missing mass is compatible with π0. Equivalent to T*π < 90 MeV cut

To avoid edge resolution effects require T*π < 80 MeV in analysis

L1 requires nx>2 or ny>2

Data MC comparison

Data MC for Eγ (W<0.5)W(Data)/W (IBMC)

Fitting region

IB dominatedregion

The IB dominated part of the data W spectrum is very well reproduced by MC

The radiated γ energy (for the IB part of the spectrum) is very well reproduced

Eγmin cut

Fitting algorithm

nbin

kkkkkkkkkkkkk IdbnL

1

lnlnlnlnln

kkkk j

To get the fractions of IB(α), DE(β), INT(γ), from data we use an extended maximum likelihood approach:

1

The fit has been performed in 14 bins in W, between 0.2-0.9, with a minimum γ energy of 5 GeV, using a data sample of 124K events.

To get the fractions of DE and INT the raw parameter are corrected for different acceptances

Systematic uncertainties

EffectSyst. DE

Syst. INT

Energy scale +0.09 -0.21

Fitting procedure 0.02 0.19

L1 trigger ±0.17 ±0.43

Mistagging _ ±0.2

L2 Trigger ±0.17 ±0.52

Resolutions difference <0.05 <0.1

LKr non linearity <0.05 <0.05

BG contributions <0.05 <0.05

TOTAL ±0.25 ±0.73

Many systematic checks have been performed using both data and MC

Trigger efficiency dominates In 2004 both L1 and L2 triggers have been

modified in order to reduce systematics

Fit results

First measurement in the region 0 < T*

π< 80 MeV of the DE term with free INT term.

First evidence of a non zero INT term

*

*

0 80

0 80

( ) (3.35 0.35 0.25 )%

( ) ( 2.67 0.81 0.73 )%

stat systT MeV

stat systT MeV

Frac DE

Frac INT

NA4

8/2

Prel

imin

ary

The error on the results is still dominated by statistics and we could profit of 2004 data (and the remaining fraction of 2003 data) to reduce statistical uncertainties.

High correlation (ρ=-0.92)

INT=0 fit: just for comparisonFor comparison with other experiments, we have also extracted the

fraction of DE, with the INT term fixed to 0 in the region 55-90 MeV. A likelihood fit using IB and DE MC only has been performed in the region 0 < T*

π < 80 MeV. Extrapolating to the 55-90 MeV region using MC we get:

Fraction of DE in % of IB (INT=0)

0.00

0.50

1.00

1.50

2.00

2.50

1999 2000 2001 2002 2003 2004 2005 2006 2007year

% D

E BNL E787

KEK E470

NA48/2

The analysis of fit residuals shows a bad χ2

Description in term of IB+DE only is unable to reproduce W data spectrum.

*55 90

0( ) (0.85 0.05 0.02 )%T MeV

INTstat systFrac DE

K±→ π0π0e±ν decay

Introduction to Ke400

Ke4 decays form factors are good constraints on ChPT Lagrangian

parameters clean sources of π-π pairs at low energy (extraction of

scattering lengths) theoretically related to each other by isospin

arguments

Ke400 is the simplest because decay kinematics

is described by only one form factor (2 identical π0’s in final state no P wave)

Best measurement so far by KEK-E470 (stopped kaon decays) based on 216 events Measured branching ratio (2.29 0.33) × 10-5

Signal selection Ke4

00 Signal Topology: 1 charged track 2 π0’s (reconstructed from 4γ’s in LKr) 1 electron some missing energy and pT (neutrino)

Analysis cuts: ≥ 4 good γ clusters in LKr 2 good π0’s with similar vertex positions Cut on E/p > 0.95 and on shower width for e/π

separation Assume 60 GeV/c Kaon momentum traveling on z axis

for neutrino reconstruction Elliptical cut in plane mK (hypothetical π±π0π0 mass)

and pT in order to eliminate K± π±π0π0 decays

Background estimate Background: main sources

ππ0π0 decay + π misidentified as e (dominant)

π0eνγ (Ke3γ) decay + accidental γ In total 9642 selected events (2003 data only) with

the following BG: 260 ± 94 from K±→π±π0π0

16 ± 2 accidental BG form Ke3(γ)

Total contamination: 3%

Branching Ratio Measurement

BR Ke400

1,5

1,7

1,9

2,1

2,3

2,5

2,7

2,9

2003 2004 2005 2006 2007

Year

BR

*10-

5 PDG 2005

KEK E470

NA48/2 The statistical error can be further reduced using the 2004 data set (28K events)

Using only a fraction of 2003 data (9642 signal events with 276 94 background events) and ππ0π0 as normalization channel, the Branching Ratio of the Ke4

00 decay has been measured:

The result has been crosschecked using also Ke3 as normalization channel leading to consistent result.

5004 10029.0019.0026.0587.2 extsyststateKBR

NA4

8/2

Prel

imin

ary

The Ke400 form factors

In this case we have only 1 form factor F (2 identical 0):

The fit has been performed using both 2003 and 2004 data (~38K events)

No sensitivity to fe reached → fe set to 0.Under this assumption we get:

Those values are consistent with the ones measured by NA48/2 in “charged” Ke4 see detailed presentation by Dmitri Madigozhin

..4 242 m/SfqfqffF ee''s

'sss

syststatss

syststatss

ff

ff

020.0034.0040.0

020.0036.0129.0"

'

NA4

8/2

Prel

imin

ary

Conclusions NA48/2 has performed the first measurement of the DE

and INT terms in the region 0 < T*π< 80 MeV for the

decay K±→π±π0γ:

The results seem to indicate the presence of a negative, non vanishing, interference and therefore a non negligible contribution of E terms to the DE

An improved measurement of the Ke400 BR has been

achieved, to be compared with recent published value:

In addition Form Factors are measured, consistent with the charged Ke4 measurement

The results can be further improved using the remaining fraction of 2003 data and the full 2004 statistics

00 54 2.587 0.026 0.019 0.029 10e stat syst extBR K N

A48/

2

Prel

imin

ary

*

*

0 80

0 80

( ) (3.35 0.35 0.25 )%

( ) ( 2.67 0.81 0.73 )%

stat systT MeV

stat systT MeV

Frac DE

Frac INT

NA4

8/2

Prel

imin

ary

SPARE SLIDES

Aim to reject K±→π±π0 events and get K±→π±π0π0.It’s based on the online computation of Mfake:

Only events with MFAKE < 475 MeV, are collected by the trigger

MBX1TR-P LVL2 trigger

Ke4 neutral decays : formalism

Branching fraction and Form factors measurements:

Two identical π0 (no P wave) only ONE form factor Fs

..4 242 m/SfqfqffF ee''s

'sss

Se/4m2π0

Sπ/4m2 π0

background