status of the pp ee analysis
DESCRIPTION
Status of the pp ee analysis. Mauro Raggi, LNF INFN 29 th August 2013 NA48/2 rare decay session NA62 Collaboration meeting Liverpool. Outline. Introduction to pp ee decay Status of the MC generators The CMC DE generator vs theory Review of the analysis strategy - PowerPoint PPT PresentationTRANSCRIPT
Status of the ppee analysis
Mauro Raggi, LNF INFN
29th August 2013
NA48/2 rare decay session
NA62 Collaboration meeting Liverpool
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy
Outline Introduction to ppee decay
Status of the MC generators The CMC DE generator vs theory
Review of the analysis strategy List of cuts Data and MC sample
BG evaluation technique Likelihood method Meeg cut and side band evaluation
Data MC comparison (ppee IB) Resolutions and energy scales (charged neutral and global) Mee resolution vs mee
Measurement of the Kaon Flux BG evaluation in K2pDg
Trigger efficiency
BR values and systematic checks
28 August 2013
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy
Introduction to ppee decay
28 August 2013
The interference thanks to the possibility of measuring the plane of polarization of the e+e- pair is splitted into 3 terms: IB/E, IB/M and E/B. IB/M cancels when integrated over f as in ppg, while E/B is only
non-zero if CP violation is allowed.
Short and long distance parity violation contributions
Theoretical paper on ppee are currently: H. Pichl, ``K --> pi pi0 e+ e- decays and chiral low-energy
constants,'’Eur. Phys. J. C 20, 371 (2001) [arXiv:hep-ph/0010284].
L. Cappiello, O. Cata, G. D'Ambrosio and D. GaoEur. Phys. J. C (2012) 72:1872
Never observed so far
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy
BR IB isospin breaking correction
28 August 2013
Cut IB Pub. IB isospin
Cut IB stand IB isospin
The isospin breaking correction only changes the absolute value of the BR (-2%) not the shape of the Q distribution.
Values are not published (G. D’Ambrosio private communication)
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy
DE matrix element
28 August 2013
New set of form factors including unknown parameters
Combination are the same as K±->p±p0g Need to implement the T3 components
From NA48/2 K±->p±p0g
While
Thanks to P. Massarotti
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy
Table of comparison MC theory
28 August 2013
Q>MeV IB th IB MC DE th DE MC
2*me 1 1.000000 1 1
2 MeV 0.736 0.736 0.857 0.865
4 MeV 0.466 0.467 0.689 0.693
8 MeV 0.262 0.263 0.517 0.519
15 MeV 0.134 0.135 0.366 0.363
35 MeV 0.0371 0.0373 0.164 0.167
55 MeV 0.0134 0.0136 0.080 0.082
85 MeV 0.00328 0.00332 0.026 0.026
100 MeV 0.00160 0.00154 0.014 0.013
120 MeV 0.00057 0.00051 0.005 0.005
140 MeV 0.000096 0.00015 0.0007 0.0016
Both IB and DE generators have been implemented in a private version of CMC007For both components all 2003 MC sample has been generated with a statistics which exceeds 10 time the data one.
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy
Analysis strategy Use only the full 2003 data sample (SS0-1-2-3)
Use p+p0D(g)IB as normalization channel
Use: (1Vtx or 2Vtx or 1TrkP) as trigger sample
Perform the BRIB and possibly charge asymmetry measurements
Leave the enlargement of the sample and P violation for final result
28 August 2013
Correction name Data MC
Non linearity (1) X
Projectivity X X
Alpha and Beta X X
Blue Field X X
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy
Analysis definitions cuts (presel) Good Vertex
-1000 < ZVTX < 8000
NVTXtrk = 3
Good Track TrackQual >0.75 new 2 GeV <TrackP< 60 GeV 12 cm <RDCH1<135 cm 12 cm <RDCH4<135 cm Ddead > 2 cm Track to track dist > 2cm
Good Clusters 2Gev < Ecl <60 GeV accep(LKR) routine for geometrical acceptance Cluster to cluster > 10cm Cluster status <4 new
28 August 2013
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy
Common preselection
Preselection cuts (common to PPD and signal analysis) NgoodVertex=1 3 < NgoodCluster < 8 3 < NgoodTracks < 8 Ellipse 3pc cut 116ns < Track time <154ns new! 3 good track are the same used in vertex fitting
28 August 2013
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy
Signal selection ngoodCl ≥4
N electrons =2 E/P>0.85 new
N pions =1 E/P<0.85
N gammas =2 cluster with no associated track &Ecl>3 GeV
COG < 2 cm
abs(ETOT-pk)< 6 GeV
abs(Mpi0-MPI0PDG)<10 MeV
Mee > 0.001022 GeV
Distance of electrons @ DCH1 > 0.25 against conversions
Total charge of electrons = 0
abs(Meeg-MPI0PDG)>0.005) cut on Dalitz decay mass
abs(MK-MKPDG) < 10 MeV
28 August 2013
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy
Data MC comparison ppee
28 August 2013
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy
MK for data and MC (no BG sub)
Data MC comparison is non good due to 10% BG in the data sample
MK(MC) media = 493.30 MeV Sigma = 5.6 MeV
MK(data) media = 492.69 MeV Sigma = 5.85 MeV
28 August 2013
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy
MK for data and MC (BG sub)
Data MC comparison is good after BG subtraction in the data sample
MK(MC) media = 493.30 MeV Sigma = 5.6 MeV
MK(data) media = 493.29 MeV Sigma = 5.37 MeV
28 August 2013
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy
ZV data MC for signal
28 August 2013
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy
Data MC Mee no BG sub
28 August 2013
Region above 100 MeV dominated by p+p0p0D BG
Bad agreement between data and MC
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy
Data MC Mee BG sub
28 August 2013
Good agreement in between D’Ambrosio and data
Data peak at 120MeV due to 3PD BG almost disappeared
Is the remaining discrepancy due to DE events? Or only insufficient MC statistics in 3pD?
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy
BG estimate signal
28 August 2013
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy
Final set of cuts except MK
28 August 2013
Integrating in the mass plot BG(3PD)= 190.5±11
BG(2PD)= 100±5
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy
Flux normalization
28 August 2013
Use the flux measurement and BRs to estimate BG value
Scale the MC to: NToT(BG)=Kflux x Acc(BG) x Eff x BR(BG)
Compute the integral in the region MK±10MeV
BG(3PD)= 192±12.5stat ± 6ext
BG(2PD)= 75±10stat ±2.4ext
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy
BG estimate summary table
Method 3PD 2PD Total
TfractionFitter
190.5±11 100±12 290.5±23
Flux norm. 195± 13 sta ± 6 ext
75 ± 10 stat ± 2.2ext
270.0 ±17stat±7ext
28 August 2013
The BG measurements agree very well on the 3PD
The value for PPD is not soo good due to low MC statistics in the plot
The value we will use for the BG subtraction is:
280±17stat±7syst
The systematic assigned is the 0.5*difference between the 2 estimates 10
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy
Flux measurement
28 August 2013
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy
The Kaon flux measurement
Used the decay K±->p±p0D(g) (PPD) means K±->p±ge+e-(g)
Same trigger chain of the signal Very similar final state in the charged trigger (e+e-p+)
MonteCarlo generator used: According to Evgueni suggestion using the KLOE generator (CMC
51) 1 times full 2003 data set 222 Million events generated run by run 165 in the ZV region 1.3 time the data
Normalization BR used in the calculation: BR(K->2p(g))xBR(p->Dalitz)=(20.66*1.174)x10-
2=(2.425±0.073)x10-3
No cut applied to minimum E* g assuming that generator is fine
Flux measurement formula: (NPPD - NBGPPD)/(ePPD x AccPPD x BRPPD)
28 August 2013
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy
PPD selection cuts N electrons = 2 (0.9<E/p) with different charges
Pion charged = 1 (E/p<0.85)
N gammas = 1 (no ass cluster)
COG < 2 cm
abs(M(eeg)-MPI0(PDG)) < 10 MeV
Distance of e+e- tracks at DCH1 > 0.25 cm
abs(ETOT-PK)< 6 GeV
Track and clusters in 5ns (data only)
abs(MK-MK(PDG)) < 10 MeV
T*p > 85 MeV
Trigger (2VTX or 1VTX or 1TRKP)
28 August 2013
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy
T*p PPD data BG estimate
28 August 2013
Unknown BG source: NBG=65*(130-85)*4+30*(140-130)*4+734=13634 < 0.1% Systematic on the flux measurement (0.1%)
Km3D BG from MC estimate (subtracted from data) NBG=7109
65 events/binKm3D
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy
Data MC for PPD selection
28 August 2013
Data Mass fit resultsMean 493.09 MeV Sigma 3.83 MeV
MC Mass fit resultsMean 493.18 MeV Sigma 3.93 MeV
Mass are very close to each other <100 KeVThey are apart from the PDG by 0.5 and 0.4 MeV
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy
T*p data MC for PPD
28 August 2013
Data T*p fitMean 1085.56 ±1.23x10-6 Sigma 3.83 MeV
MC T*p fitMean 108.587 MeVSigma 3.91 MeV
The T*p is perfectly on it’s place 108.5The difference in data to MC is order 30KeV only
The scale of the spectrometer is really in the right place!
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy
Flux calculation results 2003
Quantity Value Rel Error
BR(p+p0D(g)) (2.425±0.073)x1
0-3 3%
Acceptance (8.00± 0.002)% 0.02%
Trigger efficiency 97.15 ± 0.042 0.046%
BG in PPD sample ?? 13130 0.1% (Sys)
Ntot events 14654994 0.026%
28 August 2013
KFlux=(NPPD - NBGPPD)/(ePPD x AccPPD x BRPPD)=(7.766±0.23)x1010
Error completely dominated by external error dBR(p0
D)=3%BG in PPD sample set to 0 and added 0.1% systematic
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy
Trigger efficiency measurement
28 August 2013
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy
General remarks on trigger
Control trigger : CPRE & !WDOG (TRGW)
Sample Trigger : 1VTX or 2VTX or 1TRK-P 1TRK-P: NT-PK (LvL 0) and 1TRK-P (Mfake<475 MeV)
We expect the trig eff to be different for the 2 decays due to 1TRK-P component in fact: The ppee has 1 more cluster (NT-PK more efficient) The ppee has lower value of T*p (Mfake) 1TRK-P is more efficient In the MC the efficiency of LvL0 NT-PK is not simulated correctly
Problems of low statistics for signal measurement (27 events) Try to use BG enriched sample (330 events sample) Cannot use efficiency measured using PPD
28 August 2013
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy
Trigger efficiency analysis Control trigger : CPRE & !WDOG (TRGW)
Sample Triggers: (1VTX or 2VTX or 1TRKP) (1VTX or 2VTX)
Measured with both the signal and the normalization in data and MC with loose and tight cuts
Difference in between loose and tight cuts is < 0.1% in MC
Data shows the same agreement at least in PPD
Can we use the data loose cuts determination for trigger efficiency?
28 August 2013
Trigger Loose cuts PPD PPD MC ppee ppee MC
1VTX or 2VTX 94.97±0.055
97.03±0.0046
97.7±0.85
97.00±0.027
1VTX or 2VTX or 1TRK-P
97.15±0.042
98.48±0.0033
98.7±0.65
98.95±0.016Trigger Tight cuts PPD PPD MC ppee ppee MC
1VTX or 2VTX 95.08±0.056
97.10±0.0047
92.6±5.0 97.44±0.032
1VTX or 2VTX or 1TRK-P
97.25±0.042
98.55±0.0033
96.3±3.6 99.07±0.019
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy 31
Trigger efficiency MC vs data PPD
Different absolute value due to bad simulation of NT-PK but same behavior as function of Mee (Mee>2MeV) Difference increasing at low Mee
23 April 2013
DataMC
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy 32
Trigger efficiency ppee vs PPD MC
This difference is expected and due to MBX cut is small due to electrons in the MC NT-PK is not simulated The difference in data comes from NT-PK (3 clus PPD 4clus ppee)
23 April 2013
PPD(MC)ppee(MC)
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy
Trigger efficiency loose to tight
28 August 2013
Efficiency comparison for ppee MC.
Tight 98.95Loose 99.07
389197 control sample events247440 control sample events
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy
Total BR result for 2003 raw value
Quantity Value
Acceptance (7.80±0.028)x10-3
Trigger efficiency (raw data value) (98.7±0.65) see systematics
N background 270±16
Nppee BG subtracted 2540±49
Kflux (7.766±0.23ext)x1010
28 August 2013
Stat: Includes only the error on Nppee
External error: only Kflux coming from BR(p0)D
Trig: 0.65% error on 98.7% measured from data loose selection
What about DE?
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy
DE component fitting
28 August 2013
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy
DE and IB in the Mee spectrum
Generated 13 Million events of ppee DE run by run 9.6M in ZV acceptance
Acceptance measurement: Acc(DE) = (3.876± 0.006)% Acc(IB) = (0.782± 0.0015)%
The ratio of acceptance is Acc(DE)/Acc(IB)~5 which meansthat the DE is a order ~6% in the data (according to D’Ambrosio prediction 1/77*BR(IB) )
Using G. D’Ambrosio theoretical BR leads to a BG evaluation of: N(ppee)DE =155±20 events ( includes a 10% error on the BR(DE)
)
This number can be subtracted as a BG to get the N(ppee)IB
N(ppee)IB =N(ppee)TOT - N(ppee)DE = 2540-155 = 2385 IB events 28 August 2013
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy
Fitting the DE
Fitting the DE using Mee is quite hard for different reasons: Distributions of Mee is really very similar much more that in ppg Regions in which the DE is dominant are populated by 10% BG The low statistics does’nt allow to have a very hard selection
cuts
Seems that our acceptance spoils the difference even more
28 August 2013
IB MC GENDE MC GEN
Before the selection After the selection
IB MC GENDE MC GEN
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy
Real life is even harder…
28 August 2013
IB MC RECDE MC REC
Reconstructed MC after correct DE normalization including higher acceptance factor 5
Mee
T*p
Situation in T*pi looks a bit betterbut BG has to be taken into account…
IB MC DE MC
IB MC DE MC
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy
T*p BG distributions
28 August 2013
3pD MC DE MC
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy
Results of fitting attempts
28 August 2013
Fit with 3 MC only: IBMC, 3pD, 2pDgMC IB frac (86.9±2.2)% 2445±62 ev BG 3pD frac (9.39±1.2)% 264±34 ev BG 2pDg frac (3.7±1.0)% 104±28 evEstimate from BG fit measurement are:NBG 3PD 190.5 ± 13.0 NBG 2PD 100.0 ± 9.8
Residuals look quite good even without the DE. The enhancement of the 3PD bg +74 events is masking the presence of the DEFits with DE does’nt converge!
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy
Systematic checks
28 August 2013
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy
Cut variation systematics
I did many cuts variation that you can see in my April talk MK, M , p COG, Mee, Mee ,g Egmin, PeMin
No effect has been seen except for the PeMIN
This has been redone This misses still the correction of trigger efficiency point to point
28 August 2013
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy
Minimum electron energy
28 August 2013
Max difference (4.34-4.24)/4.24=2.3% max differencePart of this will be due to absence of trigger efficiency I propose to use the 0.5*Max Difference=1.2% systematicI’ll investigate this more soon.
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy
Errors summary table
28 August 2013
Systematic Value
N of signal events (2450) 2.09% sqrt(2811)/2540
Signal acceptance 0.2%
Total Statistical 2.1%
Difference of 2 analysis 0.45%
Back ground subtraction 0.72% (error on BG 18.4/2540)
Trigger efficiency 0.65% (From data loose sample)
Non linearity <1%
Energy scale No effect
Cut Variation 1.2% (PeMIN) see april talk
Total systematics 1.74%
BR(p0D) external 3%
Total external 3%
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy
Preliminary BR result
Theoretical prediction from G. D’Ambrosio et al (Eur. Phys. J. C (2012) 72:1872):
BR(ppee)IB= 4.19 x 10-6 IB only no isospin correction (PUBLISHED)
BR(ppee)IB= 4.10 x 10-6 IB only isospin breaking correction (PRIVATE)
In the IB only BR we considered DE as a BG we subtracted: DE = Kflux*AccDE*Eff*BRDE(Th) =155±20 events (+0.8% syst or
EXT)
28 August 2013
Mauro Raggi - I.N.F.N. - Laboratori Nazionali di Frascati - Italy
Conclusions
We performed the first measurement of the BR for K±->ppee We reached less than 1% agreement between the two analysis Total and IB only BR can be measured
The Background is very well under control (~9.6%)
First attempt to fit DE is showing many obstacles Small difference in the spectrum BG domination in favored DE regions Low statistics in the data sample
The result is in good agreement with isospin breaking corrected theory
First systematic check does’nt reveal important contribution Trigger efficiency to be added to systematic check routine
28 August 2013
Thank you for your attention28 August 2013