Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://zebu.uoregon.edu/~uochep/talks/talks04/semiex.pdf Äàòà èçìåíåíèÿ: Fri Jul 16 02:44:19 2004 Äàòà èíäåêñèðîâàíèÿ: Tue Oct 2 10:23:14 2012 Êîäèðîâêà: Ïîèñêîâûå ñëîâà: êàñêàä êåìáëà

Semiexclusive



Update

¡

Chris Potter and David Strom University of Oregon

BaBar Collaboration Meeting, 7 July 2004 ­ p.1/16

¡

¢


Analysis Overview
Event selection: A tag meson decaying in a hadronic mode is fully reconstructed using the semiexclusive (where ). We use reconstruction: RecoilAnalysis/baseClass.cc for the interface to the semiexclusive ntuples.
¦¥ ¤ ©¨ § ¡ ¸ ¡ ¢ !

Events with two remaining tracks identify the decays to single track tau modes: or . This accounts for 51% of all tau pair modes.
' #( #



¢

%$ #

& "

Event remainder kaon, electron, muon and neutral pion multiplicity consistent with a tau pair decay are required. Kinematic correlations in momenta and remaining unassigned neutral energy are exploited by a neural network analysis. Changes since the last presentation (February Collaboration Meeting): We now use maximum likelihood rather than least chi-squared fits for background subtraction. A new preselection on signal side tracks and neutral pions has been implemented. The signal selection has been optimized for best upper limit. Fit, tracking, PID and neutrals systematics have been evaluated. BAD #542 version 3 was released.

BaBar Collaboration Meeting, 7 July 2004 ­ p.2/16

%$ #


Monte Carlo and Data Samples
SP4 signal (cocktail and generic) and generic background B0B0bar Monte Carlo simulation (cocktail and generic) samples used in this analysis:

¡ ¢

Sample

¢

¸¤ ¡

¥

Events



¨

¨§ ¦

©

!

¦



¨§ ¦

¦

¦

§¦ §© § ¦
Weighted B0 -> DX and DstarX cocktail This table in BAD #542 version 3 is badly incorrect. It is unknown how many B0B0bar events generated the semiexclusive ntuples, stored in VubAnalysis/chains/generic. SP5 signal samples are not yet included in the analysis. 70K of the generic and 69K of the cocktail have been generated so far. Data samples are Runs 1 and 2, stored in VubAnalysis/chains/data. Runs 3 and 4 will be included in the final version of the analysis.
!


B0->tau+tau- +CC (FSR) ...tau+tau- vs D(star)X cocktail B0B0bar generic



¨



¨§ ¦



¦

¥

¡

¨§ ¦

¦

¦
¨

!! !!

§



"

#

"

¦

¦

!



¨

$

BaBar Collaboration Meeting, 7 July 2004 ­ p.3/16

§



¦

!! !!



¦



¦


Tag
events/bin 25000

Yield
Data 2000, 2001, 2002 Tag B Yield Signal Region: 343564 Argus: 219811 Crystal Ball: 123753

¡


events/bin

25000

Data 2000, 2001, 2002 Tag B Yield Signal Region: 343564 Argus: 224388 Gaussians: 119176

20000

20000

15000

15000

10000

10000

5000

5000

0 5.2

5.21 5.22 5.23 5.24 5.25 5.26 5.27 5.28 5.29 5.3 mES [GeV]

0 5.2

5.21 5.22 5.23 5.24 5.25 5.26 5.27 5.28 5.29 5.3 mES/GeV

At left, a Crystal Ball fit to in the tag sample after requiring GeV . The yield is 123753 in the peak. and the reconstruction mode purity greater than


At right, a fit to the data tag yield using two Gaussians rather than a Crystal Ball. The yield is , giving a relative systematic error of on the data tag yield .
¦

¦

¦

BaBar Collaboration Meeting, 7 July 2004 ­ p.4/16

§©

©¨ § ¨



¦¦

¦

!



§¤ §©
$



¢

¢



¸

¥


Preselection
Four requirements are imposed on the samples before beginning this analysis: tag tag energy-substituted mass

§¤ §©
$

¥

GeV
¦







reconstruction mode purity greater than

¢



¸

two very loose track candidates reconstructed opposite the tag two or fewer neutral pions reconstructed opposite the tag The yields are B0B0bar: 8773 events Data Runs 1,2: 4433 events Signal Generic: 142 events Signal Cocktail: 21923 events


¦

are on the following slide: clockwise from top left are the B0B0bar, data, signal Plots of generic, and signal cocktail samples.


¢



BaBar Collaboration Meeting, 7 July 2004 ­ p.5/16




Preselection
events/bin 1200 1000
B0B0bar Generic Preselection Signal Region: 10146 Argus: 1373 Crystal Ball: 8773

Plots
events/bin 900 800 700 600 500 400
Data 2000, 2001, 2002 Preselection Signal Region: 8288 Argus: 3855 Crystal Ball: 4433

800

600

400



300 200

200 100 0 5.2 5.21 5.22 5.23 5.24 5.25 5.26 5.27 5.28 5.29 5.3 mES [GeV] 0 5.2 5.21 5.22 5.23 5.24 5.25 5.26 5.27 5.28 5.29 5.3 mES [GeV]

events/bin

2500

B + - Cocktail Preselection Signal Region: 22117 Argus: 194 Crystal Ball: 21923

0

events/bin

50

B + - Generic Preselection Signal Region: 154 Argus: 12 Crystal Ball: 142

0

2000

40

1500

30

1000

20

500

10

0 5.2

5.21 5.22 5.23 5.24 5.25 5.26 5.27 5.28 5.29 5.3 mES [GeV]

0 5.2

5.21 5.22 5.23 5.24 5.25 5.26 5.27 5.28 5.29 5.3 mES [GeV]

BaBar Collaboration Meeting, 7 July 2004 ­ p.6/16


Analysis Chain Overview
Preselection (two signal side GoodTracksVeryLoose, two or fewer pi0DefaultMass). Total charge in GoodTracksLoose is zero.GoodTracksLoose multiplicity is two. Kaon veto: zero candidates in KMicroLoose, KsDefault, KlongEmcTight. Unassigned GeV. neutral energy (GoodPhotonsLoose less pi0DefaultMass) less than Signal eMicroTight, muMicroLoose, pi0DefaultMass multiplicities and
¡ § ¢§

§

selection:

Selection Mode 0: 1: 2: 3: 4:
' ' '

BR

¤¸ §

Efficiency







¤

¥

"

"

¸

¸

na na

§



§

¦







©

§

¦



§
GeV GeV GeV

#

§ ¤

'

©

§

¦



¦

¦

§ ¦

!

¥

§ © § § " §

! !

&

#

$

na





§

¦







#



§ ¤ § ¤



§

!



¦

5:





§

©

!





Neural network output greater than between tau daughter candidates in GeV). neutral energy (below

. The inputs are: reconstruction mode, cosine of angle frame, tau daughter candidate momenta and unassigned

§



§

¦

BaBar Collaboration Meeting, 7 July 2004 ­ p.7/16

§ ¦

&

#

$

!

¥

§ ¦

&



#

$

!

¥

"

§

#

"

&


Signal Selection Optimization
We optimize for the lowest upper limit on the branching ratio. The neural network optimization proceeds by applying all requirements in the analysis chain and then varying the neural network output ). The maximum signal presence for is then . We assume to be cut value ( . the equal to the expected mean background
¤ ¦¥©¨ § ¡ ©¨¥ § ¡ § ¢¸ § ¢¸
¡



µ( bkg)/ sig

194 192 190 188 186 184 182 180 178 176 174 0 0.1 0.2 0.3 0.4 0.5 NN

¨

¤

cut

Triangles indicate maximum signal presence versus neural net cut value. Diamonds include the Monte Carlo simulation statistical, fit parameter and fit pdf error; stars include both fit and particle indentification error. the neutrals error; crosses include fit, neutrals and
$



BaBar Collaboration Meeting, 7 July 2004 ­ p.8/16

$

¤



¤ ¨¦¥ § ¨ ¡ ¥©¨ §

¡

¥











§

§ ¢
§

§ ¢¸
§
¨§ ¢ ¤ © ¥ ¦ ¥ ¦ ¥ ¦ ¦ ¤ ¸¢ ¡

§ §

Tag B Yield

Preselection

Requirement

§
! ! © ¤ ¦




¥

¦


¦
GeV

§








!

¦ © "
!

¦ ©
!

¦¦ ""

!

©


© ¦


©
! !

" ¦


Peak Yield

©©


© " § ¦©

©

§



¦¦



" §
!



§ ¦¦ § ©


B0B0bar Monte Carlo

Efficiency

§ ©



¦ §


§" §


§ §


© "
! !

§ ¦ ¦©
blind blind blind blind blind
!

§ § ©
!



§ ¦

!



§ ©


¦





© ""


¦ " ¨§ ¦ ¦
blind

© ¦ © §¦
!

Peak Yield

Efficiencies in B0B0bar and Data

¦¦ ¨§ ¦


¦ §
!



" ¨§ ¦ ¦ §
!

©©"" ¦


¦ © ¦ "
!

©


© ¨§ ¦ ¦


Data 2000,2001,2002

©


Efficiency


BaBar Collaboration Meeting, 7 July 2004 ­ p.9/16

"

§©

§" ¦¦

§" ¦©












§

§ ¢
§

§ ¢¸
§
§ ¢ ¤ © ¥ ¦ ¥ ¦ ¥ ¦ ¦ ¤ ¢¡

§ §

Tag B Yield

Preselection

Requirement

§
! ! © ¤ ¦




¥

¦


¦
GeV

§







¦
! !

¦ "


¦ " " ¦


¦ " ¦


¦
!

¦ ¦© ©


!




¦ ¦ ©


¦ © " © ¦


¦ " © ¦


©" ¦ " "


Peak Yield

Efficiencies in Signal





¦¦ ¦ §
!

§¦
!

§ ¦ § ¦
!

§
!

©


¨§ ¦ §


§ ¦ ¦
!

§ §


§ ¦ " § ¦© §¦


Signal Cocktail

Efficiency

§


¦ §¦
!

¦

§


" ¦ ¦


!

§ © ¦¦ § §


§


¦ " § ©
!

§


¦ ¦© ©


§


¦ ¦ ¦© © ¦





¦ ¦ § §


" ¦¦ © §
!

©©


Peak Yield

¦¦ §"




¦ © § §




¦ © §




Signal Generic

§ §


©

¦


Efficiency

BaBar Collaboration Meeting, 7 July 2004 ­ p.10/16

¨§ ¦" § © ©





The Extra
entries/bin
14 12 10 8 6 4 2 0 0

Control Sample

The peaking components of data (dots), B0B0bar (solid) and the signal cocktail (dashed) are plotted.

0.1

0.2



0.3

0.4

0.5

0.6

0.7

0.8

0.9 1 NN Out

¦¥ ¤

¨§ ©

¡



¸





A typical background event is thus: Events like
¦¥ ¤ ¨§ ¡ © ¸ ¢ ¢

(where is undetected). should very closely mimic the background.
¢ "! # $ %





control sample, the requirement is replaced by the requirement , with the provision that the KshortDefault daughter charged pions are invisible to the analysis chain. All other requirements in the signal analysis chain apply.
! # $ &

In the extra



From Monte Carlo simulation, the expected number of events is . number of events is
&' & 1 & &

1



¢

. The observed

&'(

0

& ) )

%

BaBar Collaboration Meeting, 7 July 2004 ­ p.11/16

2


Control Sample NN Inputs
The peaking components of data (dots), B0B0bar (solid) and the signal cocktail (dashed) are plotted.
entries/bin 60 entries/bin 0 1 2 3 4 5 mode 30

25

50

40

20

30

15

20

10

10

5

0

0

-1

-0.5

0

0.5

1 cos( +, -)

entries/bin

20 18 16 14 12

entries/bin 0.5 1 1.5 2 p 2.5 [GeV]

22

14 12 10 8

10 8 6 4 2 2 0 0 0 0 0.1 0.2 0.3 E 0.4 -E 0.5 [GeV] 6 4

+

CN

pDM

BaBar Collaboration Meeting, 7 July 2004 ­ p.12/16


Signal Analysis NN Inputs (Blind)
The peaking components of data (dots), B0B0bar (solid) and the signal cocktail (dashed) are plotted.
entries/bin 200 180 160 140 120 100 80 30 60 20 40 20 0 0 1 2 3 4 5 mode 10 0 -1 -0.5 0 0.5 1 cos( +, -) entries/bin entries/bin 0.5 1 1.5 2 2.5 80 70 60 50 40

entries/bin

60

50

50

40 40 30 30 20

20

10

10

0 0

0 p
+

0

0.1

0.2

0.3 E

CN

0.4 -E

pDM

0.5 [GeV]

BaBar Collaboration Meeting, 7 July 2004 ­ p.13/16


Sources of Systematic Error
Several sources of systematic error affect the four quantities ( . will determine the limit on
¡

and

) which

¨§ ¨

©¨¥ §

¤



¨




$



¡





For the error due to the choice of probability density function, we refit the with two Gaussians and an Argus rather than a Crystal Ball and an Argus.

"¨

¢

"



$

distributions

For two signal side tracks, we take the track reconstruction error to be twice the single track recommended by the tracking group. systematic error

§

¦

The systematic errors for reconstructing and are obtained by comparing the efficiencies before and after the baseClass switches -pe, -pm and -pk were employed and calculating the relative difference in efficiency. The PID tables are RecoilAnalysis/PidTables-200x.txt. The systematic error for neutral energy resolution was obtained by comparing the efficiencies before and after the baseClass switch -SN was employed. The systematic error due to the remaining neutral energy we derive from the extra control sample: the difference between the Monte Carlo simulation expectation and observation we take to be an upper bound on the error due to the remaining neutral energy systematic. An additional track multiplicity error originates in the mismatch between the Monte Carlo simulation of extra junk tracks and the true junk track rate.
$

BaBar Collaboration Meeting, 7 July 2004 ­ p.14/16



¤



¢

¢



¥






Systematic Errors
¡
¦

¡



¦

¡ ¨





Systematic Effect MC Stat and Fit Parameters Fit PDF Choice Electron Particle ID Muon Particle ID Kaon Particle ID Neutrals Energy Resolution Tracking Efficiency Track Multiplicity Remaining Neutral Energy Signal Decay Model Total Error





¤

§

¤

§

¤

§

¤

©¨¥ §

©¨¥ §

©¨¥ §

¥©¨ §

§





¨

§

©

§ ¦

¨

¨§ ¦

¦

§

©

!

¦

"

"

§ §© ¨§ ¦ § § §©
!

§

"

§

§

§" ¦"

§



"

§ ©

§



¨§ ¦

!

¨§ ¦







§"

Relative systematic errors for the signal cocktail efficiency, signal generic efficiency, B0B0bar efficiency. Errors which are set to zero have not yet been evaluated.

¢¡

¡

BaBar Collaboration Meeting, 7 July 2004 ­ p.15/16

§ ¦

"







!



¨§ ¦



©

§ §
!







§



¨





¦


Outlook
Near term plans: Finish evaluating systematic errors. Finalize limit setting procedure. Obtain AWG approval to... Meet with our Review Committee. Unblind. Long term plans: Include SP5 signal Monte Carlo, include Runs 3 and 4 and begin work on :
¢



$"

"

¤

10000

entries/bin

8000

6000

4000

2000

0 0

0.5

1

1.5

2

2.5

p

+

BaBar Collaboration Meeting, 7 July 2004 ­ p.16/16