Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://nuclphys.sinp.msu.ru/conf/lpp14/240809/Giorgi.pdf Äàòà èçìåíåíèÿ: Wed Sep 16 16:32:49 2009 Äàòà èíäåêñèðîâàíèÿ: Tue Oct 2 00:40:08 2012 Êîäèðîâêà: Ïîèñêîâûå ñëîâà: tail

BABAR:Search for Lepton Flavor Violation in tau decay
Marcello A. Giorgi (on behalf of the BABAR Collaboration) INFN & Universita' di Pisa it

08/15/2009

Marcello A. Giorgi

1


Outline

· Theory Overview · The BaBar Detector

· l · lll · lV0
· Conclusion
08/15/2009 Marcello A. Giorgi 2


SM prediction for LFV in decay
Standard Model allows LFV. Suppressed by (quasi)-degenerate neutrino masses. In charged leptons it can occur in loops with expected branching fractions well below present sensitivities. E.g.: expected BF ( µ)
If detected, LFV would imply New Physics with present (and near future) luminosities. Many New Physismodels predict LFV BFup to [O(10-8)]. [O(10
08/15/2009 Marcello A. Giorgi 3


Some model predictions
In SUSY LFV decays are generated via slepton mixing

Undetectable

08/15/2009

Marcello A. Giorgi

4


Looking at NP flavor structure

· Different models expect different processes leading to LFV in charged sector · If LFV is observed NP flavor structure may be investigated by looking at LFV BF Ratios.

[arxi : hepph/ 0610344v3 and hepph/0702136] iv

08/15/2009

Marcello A. Giorgi

5


PePII and BaBar
Sample Used: l 470 fb1 ((4S)) + 31 fb1 ((3S)) + 15 fb1 ((2S)) lll 470f b1 ((4S)) lV0 451fb1 ((4S))

Signal MC is produced with KK2f +Tauola Tau decay with Tauola, with the radiation in decay simulated with PHOTOS

Highest peak luminosity 4 times design luminosity : 1.2 x 1034 Monthly integrated luminosity 6 times the design : 20 fb1 Resulting in one of the largest data sample Clean environmentMfoelloraGior process searches arc r A. regi 08/15/2009

6


The BaBar Detector

08/15/2009

Marcello A. Giorgi

7


Variables for blind analyses
· All analyses are blind · Signal and backgrounds scattered in a plane defined by the variables: bl

· Resolution is different among channels, worse for channel with electrons:
­ spread dominated by radiative losses, with larger tails in negative regions ­ ec spread dominated by tracking resolution
08/15/2009 Marcello A. Giorgi 8


Search for l

08/15/2009

Marcello A. Giorgi

9


Preselection
e/µ/ tag

l


· 5 tags considered to minimize backgrounds from QED radiative processes. · Further Discrimination:
­ Charge conservation ­ Reconstructed mass compatible with in tag side i ­ 0.76 250 MeV (e/µ); < 250 Mev (,3h) <500 Mev () ­ Cosrecoil <0.975 (e/µ)
08/15/2009 Marcello A. Giorgi



e/µ/
(s)

tag

l



(s)

3h tag

l

10


Neural network selector
· · · NN implemented for each tag. Topological info used for radiative QED and backgrounds reduction Six variables used as input: bl ­ Total tag side momentum [2PTag CM / s], Recoil angle [Cosrecoil], leptonphoton opening angle [Cos l ], ­ Missing Mass [m2 ] Missing pTmiss [ln(2 pTmiss /s )] , E

E

ECM s

-

sin(1 + 2 ) sin(1 ) + sin( 2 ) + sin(1 + 2 )

08/15/2009

Marcello A. Giorgi

11


Neural network selector
· · · NN implemented for each tag. Topological info used for radiative QED and backgrounds reduction Six variables used as input: bl ­ Total tag side momentum [2PTag CM / s], Recoil angle [Cosrecoil], leptonphoton opening angle [Cos l ], ­ Missing Mass [m2 ] Missing pTmiss [ln(2 pTmiss /s )] , E

E

ECM

1 1 s NN Selection power is self evident from plots.

-

sin(1 + 2 ) sin( ) + sin( 2 ) + sin( + 2 )

08/15/2009

Marcello A. Giorgi

12


Neural network selector
· · · NN implemented for each tag. Topological info used for radiative QED and backgrounds reduction Six variables used as input: bl ­ Total tag side momentum [2PTag CM / s], Recoil angle [Cosrecoil], leptonphoton opening angle [Cos l ], ­ Missing Mass [m2 ] Missing pTmiss [ln(2 pTmiss /s )] , E

E

ECM

After NN: For e backgrounds dominated by in µ & tags. For µ backgrounds dominated by in µ & e tags

1 1 s NN Selection power is self evident from plots.

-

sin(1 + 2 ) sin( ) + sin( 2 ) + sin( + 2 )

e
08/15/2009 Marcello A. Giorgi

µ
13


Results
Data sample used : (9.676±7 x 10 8 ) 's, that includes runs @ Y(3s) [30f b1] and Y(2s)[15 fb1]

UL measured using Feldman Cousin method. Background estimation validated looking at larger signal boxes, efficiency estimated from MC studies for each tag and for all tags together
08/15/2009 Marcello A. Giorgi 14


Results
Data sample used : (9.676±7 x 10 8 ) 's, that includes runs @ Y(3s) [30f b1] and Y(2s)[15 fb1]

UL measured using Feldman Cousin method. Background estimation validated looking at larger signal boxes, efficiency estimated from MC studies for each tag and for all tags together
08/15/2009 Marcello A. Giorgi

No evidence of signal in the 2 signal elipses!

15


Search for lll Search for lV0 (Had Vector Meson)

08/15/2009

Marcello A. Giorgi

16


Preselection
Babar lab. reference - system e
-

e

+

e
lll
- - - - - -



e
+

+

+

lV0
- -

Center of mass system

e-e+eµ-e+e-

µ- e- µ- e e


e-µ+e- µ-e+µ
-

- - - - - -

µ-µ+e-
08/15/2009

µ-

µ-µ+µ



µ- e

In total limits for 14 channels presented
Marcello A. Giorgi 17


Preselection
Babar lab. reference - system e
-

e

+

e
lll
- - - - - -



e
+

+

+

lV0
- -

Center of mass system

e-e+eµ-e+e-

µ- e- µ- e e


e-µ+e- µ-e+µ
-

- - - - - -

µ-µ+e-
08/15/2009

µ-

µ-µ+µ



µ- e

In total limits for 14 channels presented
Marcello A. Giorgi 18


Preselection
Babar lab. reference - system e
-

e

+

e
lll
- - - - - -



e
+

+

+

lV0
- -

Center of mass system

e-e+eµ-e+e-

µ- e- µ- e e


e-µ+e- µ-e+µ
-

- - - - - -

µ-µ+e-
08/15/2009

µ-

µ-µ+µ



µ- e

In total limits for 14 channels presented
Marcello A. Giorgi 19


Preselection
Babar lab. reference - system e
-

1 prong Q=-1 prong selection

e

+

e
lll
- - - - - -



e
+

+

+

lV0
- -

e-e+eµ
-e+e-

µ- e- µ- e e


e-µ+e- µ-e+µ
-

- - - - - -

3 prong Q=+1 selection
Hemisphere#2

l+ l- l

Center of mass system

Hemisphere#1

+

µ-µ+e-
08/15/2009

µ-

µ-µ+µ



µ- e

In total limits for 14 channels presented
Marcello A. Giorgi 20


Background Rejection
· · 5 main background categories We use ­ Combinatorial uds background: flat on mass distribution ­ cc peaking at D0 mass, may contain true leptons for lV0 ­ QED radiative events for lll ­ backgrounds with real meson and pions, with neutrinos ­ Irreducible background: D V0l for lV0

Background is parameterized in PDF bidimensionally to account for the correlations seen in the plots. eK
*

eK

*

08/15/2009

Marcello A. Giorgi

21


Three region

Selection Strategy s defined in (ec, ) plane
Signal Box (SB): different for each channel, dimension optimized to give the best UL for each channel. Data events in this region are BLIND

Large Box (LB): identical for all channels. Almost all signal events lie in this region

ecVs

Grand Sideband (GS): is the unblinded region of the LB. Background estimation made extrapolating data from GS to SB

08/15/2009

Marcello A. Giorgi

22


UL Calculation
UL obtained using Cousins and Highland. Data sample used : 8.6 x 10 8 's
PRL 95:251803, 2007

Previous babar: e a 376f b1

Major Improvements: µ eff. 66% 77% e eff. 89% 91% Smaller PID syst. Better selection and
UL improved by factor ~23 Statistics increase~25% 8 times better than Lumi Increase only
08/15/2009 Marcello A. Giorgi 23


UL Calculation
UL obtained using Cousins and Highland. Data sample used : 8.6 x 10 8 's

No evidence of signal in the signal box!
PRL 95:251803, 2007

Previous babar: e a 376f b1

Major Improvements: µ eff. 66% 77% e eff. 89% 91% Smaller PID syst. Better selection and
UL improved by factor ~23 Statistics increase~25% 8 times better than Lumi Increase only
08/15/2009 Marcello A. Giorgi 24


UL Calculation lV

0

UL obtained using Cousins and Highland method . Data sample used : : 8.3 x 10 8 's

08/15/2009

Marcello A. Giorgi

25


UL Calculation lV

0

UL obtained using Cousins and Highland method . Data sample used : : 8.3 x 10 8 's

No evidence of signal in the signal box!

08/15/2009

Marcello A. Giorgi

26


Conclusion

Undetectable

Some predictions are strongly constrained by present results BFactories is one of the cleanest environment for LFV searches, great efforts from the BaBar collaboration made it possible to improve measurement more than N thanks to new tools , that make the analysis more efficient.
08/15/2009 Marcello A. Giorgi 27


BACKUP

08/15/2009

Marcello A. Giorgi

28


Backgrounds e · Backgrounds after NN dominated by in µ & tags

08/15/2009

Marcello A. Giorgi

29


Backgrounds µ · Backgrounds after NN dominated by in µ & e tags

08/15/2009

Marcello A. Giorgi

30


Particle Identification: Event Weighting
· For MC events, PID selection applies a weight rather than rejecting events:
­ Each MC Track in signal hemisphere is weighted with a probability, corresponding to the bilit (mis)identification efficiency in data for the truth matched particle.
· For 0.1% of tracks not having MC truth match, average weight is used for tracks.

­ Each MC Event is reweighted by the product of the three tracks weights ­ This is a much more efficient use of MC Statistics
08/15/2009 Marcello A. Giorgi 31


Bidimensional Bkg Estimation
Hadronic (uds) backgrounds estimated using a twodimensional (M, E) PDF as product of two onedimensional (PMh, PEh) PDF. To avoid correlations a choice of rotated variables is made (empirical parametrization): PM' is a bifurcated Gaussian. E0' [and (E')] , are free parameters. Tau backgrounds are estimated with a similar likelihood fit but with same parameterization , but now 2 parameters : and . ti (ee+e) and (eµ+µ) have large QED background contributions the estimation is done by using a Reverse PID approach. cc backgrounds for lV0 channels are divided into subsamples depending on

M ' = cos( )M + sin( )E

E' = - sin( )M + cos( )E

the number of misid particles, and fitted with the shape best fitting the category among those used for other categories Control sample is built with events in the great sideband (LB) passing selections,but PID in 1 prong side. but PQED is defined as product of two one dimensional PDF on rotated variables PM' and PE' . PM' is a third order polynomial while PE' is a crystal ball function. A single rotation angle is used .
08/15/2009 Marcello A. Giorgi 32


Neural network selector
· · · NN implemented for each tag. Topological info used for radiative QED and backgrounds reduction Six variables used as input: ­ Total tag side momentum, Cosrecoil, leptonphoton opening angle Cos l, ­ missing pt, E

E

ECM s

-

sin(1 + 2 ) sin(1 ) + sin( 2 ) + sin(1 + 2 )

08/15/2009

Marcello A. Giorgi

33