Status Report on C. Di Donato, B. Di Micco, M. Jacewicz Phi-Decay WG Meeting February 9 2010
Jan 13, 2016
Status Report on
C. Di Donato, B. Di Micco, M. Jacewicz
Phi-Decay WG Meeting
February 9 2010
Outline
• Analysis
• Data-MC comparison
• Momenta Smearing
• Evaluation of systematics
• Discussion with referee
• New smearing method
• Track efficiency correction applied on MC (as in the Memo343);
• Cross-check with Kinematic fit
• PCV EVCL vs tracks selected in the analysis
Analysis Event Signature:
2 Prompt Neutral Clusters: |tcl-lcl/c|<5t
Recoil photon: most energetic cluster with E250 MeV2 tracks closest to IP (using PCA, no vertex requirement)
Kinematical Constraints:Two body decay kinematics to calculate E recoil kinematics to calculate : |Et-Pt|<10 MeV (EtPt)Best Photon: we choose one PNC with <0.13 rad to the
calculated (OPAN) Background: main one is 0: M(M
in the 0 rest frame cos
Only barrelTOF(Time Of Flight cut to reject bhabha background)-ANGLE
DATA-MC comparison and MC smearing
Let’s go back to the smearing, starting from control sample selected to check Data-MC: selected using reversed cut on cos
we look at missing mass from +-
Data – MC and smearing: first approach
Missing mass on control sample
We find discrepancy and try to solve smearing the Pt
OLD APPROACH
After Smearing
OLD APPROACH
Only background is included
After Smearing
OLD APPROACH
All background is included
Residual discrepancy on the right tail.
Smearing method?
Background estimation?
Data-MC comparison
Opening angle
between
Data-MC comparison
Invariant Mass
M(MeV) M(MeV)
M(MeV)
Data-MC
PP(MeV) P(MeV)
P(MeV)
Data-MC
Summary on Systematics
After smearing systematics for angular cuts remains practically unchanged.
We look at cosand cos(OpAn) for cluster also in the transverse variable
Summary on Systematics
Full cos
Only transverse angle
xy plane
Summary on SystematicsCos(Full OpAn ) Only transverse angle
The situation is much better, systematic reduced from 1-2% to 1-2 per mil, the price we pay is the
background contamination:
OPAN–Cos–ANGLE–TOF–onlybarrel: Cut Eff = 25.67%
Signal events = 564765, PHI Bkg = 6.7% ETA Bkg = 30%
OPAN–Cos–ANGLE–TOF–onlybarrel–EtPt: Cut Eff = 24.89%
Signal events = 547759, PHI Bkg = 3% ETA Bkg = 0.2%
TransOPAN–CosTrans–ANGLE–TOF–onlybarrel: Cut Eff = 24.26%
Signal events = 533938, PHI Bkg = 25% ETA Bkg = 120%
TransOPAN–CosTrans–ANGLE–TOF–onlybarrel–EtPt: Cut Eff = 22.98%
Signal events = 505692, PHI Bkg = 11% ETA Bkg = 1%
NEW CUTS:
OLD CUTS:
New SystematicsTransverse cos
OK
New Systematics
Cos(Transv. OpAn )
OK
New Systematics
Et-Pt
The systematic on Et-Pt doesn't change.
New Systematics
E-P
Spectra after smearing
Spectra before smearing
Unbiased sample: only preselection
Outline (II)
New data sample to “play with” 560 pb-1 preselected with basic conditions: to study
smearing effect we introduce less kinematic constraints, and ask only ≥2 Prompt Neutral Cluster, one with E≥250MeV
This sample was used for the recent study of systematics and new smearing
What’s new:Different approach for the momenta:we take into account
curv from DTFS cov matrix
Better fitting
We use the smearing function:
tSmearett
Scalett
PfPGausP
fPP
,
1
Full fit with 3 params:
1) Total Scale: 0.88082) shift: 337.8 • 10-6
3) smearing: 0.1470
Smearing: Old method versus New
OLD: simple method each parameterFitted separately
NEW: Full fit with 3 parameters:Total Scale, shift, smearing
E – P cut problemEven with new smearing we do not get satisfactory
comparison with MC
eeg continuum
Unbiased sample: only preselection
Now is clear the problem is on the background estimation
We try to fit the Et-Pt (E–P) spectrum to see if the discrepancy is due to a bad
Signal/Background estimation Data – MC.
E-P: with all MC contributionssatisfying agreement
Output from TFractionFitter (1200bins in the histograms)
2 = 1728 Ndf = 1196
eeg continuum
Process Exp. Int.=602260
Parameter from fit
Original integral
Integral from fit
0.431598 193340 259934
0.3767 245648 226872
eeg 0.0013846 1520 8333.89
0.190317 104500 114620
After smearing and fit…
• Data-MC satisfying agreement
• Branching ratio stability and Systematic seems to be under control, we are reevaluating the systematics.
Other checks:
• kinematic fit
• Prompt Charged Vertex from EVCL
Kinematic fit 21 Input Parameters:
2 PNC: 2 5 = 10 parameters
2 tracks at PCA: 2 3 = 6 parameters
IP position: 3 parameters
Beam info: 2 parameters
7 Constraints:
Time of flight of photons: 2
4-momentum conservation: 4
invariant mass
RAD Stream
DATA-MC comparison 2 distribution
Analysys cut:•RAD•OPAN•Cos•ANGLE•TOF•Onlybarrel•EtPt
RPI Stream
DATA-MC comparison 2 distribution Analysys cut:•RAD•OPAN•Cos•ANGLE•TOF•Onlybarrel•EtPt
RAD StreamP+
P-
True-Rec True-Fit
(Ptrue-Prec)
(Ptrue-Prec)
(Ptrue-Pfit)
(Ptrue-Pfit)
RAD
RPI StreamTrue-Rec True-Fit
P+
P-
(Ptrue-Prec)
(Ptrue-Prec)
(Ptrue-Pfit)
(Ptrue-Pfit)
RPI
PCV Check: EVCL requirement on VTX
MC sgn Efficiency MC BKG Efficiency EXP EfficiencyALL 85619 269870 355371FITOK 77226 90.20% 171795 63.66% 355371 100.00%PCV 76910 99.59% 148768 86.60% 336859 94.79%Match 2 65300 84.90% 123165 82.79% 276182 81.99%Match 1 1254 1.63% 11438 7.69% 26104 7.75%Match 0 430 0.56% 993 0.67% 1406 0.42%
MC sgn Efficiency MC BKG Efficiency EXP EfficiencyALL 13080 1171479 985282FITOK 12179 93.11% 1171270 99.98% 985282 100.00%PCV 12179 100.00% 1171270 100.00% 985282 100.00%Match 2 11009 90.39% 1099563 93.88% 912749 92.64%Match 1 140 1.15% 9883 0.84% 11686 1.19%Match 0 50 0.41% 2195 0.19% 1835 0.19%
RAD
RPI
•PCV means there is the VTX as from EVCL requirement•Match means the tracks we choose are the one connected to PCV
Conclusions
• DATA-MC: the agreement is satisfying; systematics and BR seems to be under control: we are recomputing
• For us everithing is ok and we are ready to produce final number and documentation, if we are below 1%: TOF systematic still to be studied.
• Kinematic fit does not improve resolution and background reduction, moreover the 2 distribution shows disagrement at 2-5% level (well beyond our target)
• We can apply PCV requirement, in RAD stream, we do not use extra VTX info; use Roberto and Antonio work on tracks/vtx
Spare
Fitting with fraction fitter seems better (but fit probability still low)
E