Oral Candidacy Presentation

David Doll

1

OutlineThesis topic: b→sγ

MotivationPrevious Analysis

Babar overviewSubdetector introduction and impact on thesis

Previous work: Introduction to Random Forest method

Applications to thesis topic and plans

KB

2

b→sγ Motivation

Flavor changing neutral current decay (absent at tree level in SM)

Precision test of SMDifferent models may enhance or suppress the BF

γ

Source: U. Haisch, FPCP Conference Taipei, 2008

3

For photon energy cut Eγ > 1.6 GeV in B meson rest frameNNLO theoretical calculation

HFAG experimental results (as of March 15, 2007)

b→sγ Motivation

409.010.0 10)03.024.055.3()B(

sXBF

410)23.015.3()B( sXBF

Combined statistical and systematic error Systematic

uncertainty associate with Ecut = [1.8,2.0]

Error associated with subtraction of events

dXB

Source U. Haisch FPCP Violation 2008

4

b→sγ MotivationPhoton energy spectrum

In b quark rest frame (impossible to boost to), this would be a delta function (≈mb/2)

b quark motion within meson smears this spectrum Spectral shape dependent on modeling of spectator

quark In the framework of HQET, the parameters λ1 and

(or equivalently mb) may be determined from the first two moments of the spectrum

Beyond SM theories are not predicted to influence this much

5

Different Photon ModelsShape function of b quark motion is universal

Applicable to all decays involving tranistions to massless states (B→Xs γ, B→ Xd γ, etc.)

Different shape function models exist

A.L. Kagan and M. Neubert propose an exponential shape function (KN model):

xaaexNkF )1()1()(

where: 1

k

x

bB mm

)1(3

1

3

1 2

12

a

Source Eur. Phys. Jour. C7, 5-27 (1999)

Gaussian Ansatz

6

Dependence on Ecut

7

D. Benson, I.I. Bigi, and N. Uraltsev also investigate an exponential and a Gaussian ansatz (with minimal difference between the two)Use purely perturbative spectrum calculated by Z. Ligeti, M. Luke,

A.V. Manohar, and M. Wise as a starting pointAdd nonperturbative pieces to the energy momentsFinally, they investigate the effects of the minimum photon energy

cut to evaluate a bias in mb and μπ2

Complete bias

Without perturbative corrections

difference

Source D. Benson, I.I. Bigi and N. Uraltsev FPCP Violation 2004

Other Photon ModelsB. Lange, M. Neubert, and G. Paz also

present shape functions based on exponential, gaussian, and hyperbolic functions (hep-ph/0504071)They detail how to fit the parameters:

8

),ˆ(ˆˆˆ),ˆ(0ˆ

0

0 iN

iN SdM

min

0 2ˆ EM B

First and second moments are directly relatable to and μπ

2 and:

Is a good model for

)ˆ()]ˆ(/),ˆ([ 0][

000 FMM Fi

Exponential, gaussian, or hyperbolic

),ˆ(ˆiS

Search Strategy for b→sγPerforming a sum of

exclusive states38 states totalUpdate of former analysis

published in Phys. Rev. D (2005) 052004 Based on 89.1 fb-1 of data

collected at the Y(4S)

9Source Babar doc

On CP Asymmetry MeasurementBecause of the final state

reconstruction, a direct CP asymmetry measurement is possible with this strategy in modes of definite flavor (yellow)Recently investigated with

~80% of the total dataPossible source of other

new physicsSearch to be performed in

another analysis

10Source Babar doc

Former Analysis ProcedureReconstruct event candidates into the 38 different decay modesUse a Neural Network to reject continuum events based on event

shape variablesSet the lower cutoff energy at Eγ > 1.9 GeV (or equivalently at

MXs between 0.6-2.8 GeV/c2) to limit peaking B background

Choose the ‘best’ candidate as the one that minimizes ΔE

11

Source Babar doc

)2/(* sEBE

Subtract off the continuum, generic BB, and cross-feed backgrounds by fitting the beam substituted mass, mES and fit to signal on bin-by-bin basis in MXs

The peaking background contribution (cross-feed and generic BB) is fit with a Novosibirsk function:

The continuum contribution is fit with an ARGUS functionAs a default signal model, they use the KN exponential model

with mb= 4.65 GeV/c2 and λ1=-0.30 GeV2/c4

Kagan and Neubert recommend treating only the range between MXs={1.1 GeV/c2, 2.8 GeV/c2} as a non-resonant spectrum

Below MXs= 1.1 GeV/c2, they recommend using K*γ MC below 1.1 GeV/c2

12

Former Analysis Procedure

Bin-by-bin fit to signal gives Partial Branching Fractions, PBF(MXs)

Need to correct PBF(MXs) for fractional coverage of inclusive b→sγ decays to get Total Branching Fractions, TBF(MXs)

Convert TBF(MXs) to TBF(Eγ)

Fit TBF(Eγ) to different expected models, allowing extraction of inclusive Branching Fraction measurement to lower Eγ

Also extract shape function paramters, mb and λ1 from model fit

13

Former Analysis Procedure

Impact of competing modelsIdeally, photon spectrum measurement is

model independentAnalysis strategy forbids this

Need idea of total decay coverage in each MXs bin

Not able to extrapolate to a total BF without introducing some model dependencies

Use a sample with a flat Eγ distribution to reweight to any model chosenMeasure parameters in all models considered

(everyone’s happy)14

Results of Former AnalysisQuote results of KN fit, ‘kinetic’ model (BBU

from above), and ‘shape function’ models (average of 3 BNP shapes from above)

15Source Babar doc

PEP-II at SLAC

e- (at 9.0 GeV) on e+ (at 3.1 GeV)CM energy = 10.58, the mass of the Υ (3S)Lorentz boost of βγ = 0.56B meson lifetime 1.5-1.6 ps → Δz ≈ 250-270 μmTurned off in April with a total of ~485 fb-1 at or just below the

Υ(4S).

Source Babar Doc

16

Babar Detector

DIRC

SVT

DCH

EMC

IFR

Solenoid Magnet (1.5 T)

e +

e -

17

Subsystems OverviewSubsystem Measured

quantityMethod

Silicon Vertex Tracker (SVT)

Particle Tracking, Vertex Location,

dE/dx

Double sided silicon strips

Drift Chamber (DCH)

Particle Tracking, dE/dx

Sense wires in helium-isobutane gas

mixture

Detector of Internally Reflected Cherenkov

light (DIRC)

Particle ID for particles of

momentum greater than 700 MeV/c

Cherenkov light measured on PMTs

Electromagnetic Calorimeter

(EMC)

Energy, Shower Shape

CsI(Tl) crystals, read out by Si Photo-

diodes

Instrumented Flux Return (IFR)

Penetration,Shower Shape

Streamer detection

18

SVT and DCH

SVT 5 layers of double-sided silicon strip sensorsφ measuring strips parallel to the beam, z measuring strips

perpendicular to the beam20-40 μm resolution in all 5 layers.

DCH7,104 small drift cells arranged in 40 cylindrical layersdE/dx measured by total charge deposited in each cell

Source Babar doc

SVT DCH

19

DIRCParticle ID for particles with momentum above 750

MeV/c144 fused, synthetica silica bars arranged in a 12-sided

polygonReadout by 11,000 PMTs

Source Babar doc

20

IFRSegmented steel flux return (later

also brass), instrumented in gapsOriginally used resistive plate

chambers (RPC) to detect streamers from ionizing particles

Upgraded to limited streamer tubes (LST) starting in 2004

Muon efficiency

Pion mis-id rate

RPC data

LST data

Source Babar doc.

21

EMCDesigned to operate over the energy range of 20MeV to 9GeV6,580 CsI(Tl) crystals separated into 5,760 in the barrel, and 820 in the endcap

16.1 X0 in the backward half of the barrel, to 17.6 X0 in endcap

Each crystal read out by two 1cm2 Si photodiodesCalibration at low energy using a 6.13MeV photon source and at high energies

using Bhabha events Studies of the low energy calibrations have shown light yield falloff to total

around 8% or less after the run of the experiment (depending on crystal manufacturer).

Angular resolution vs photon energy

Source Babar doc.

22

B+→K+νν (or the benefits of a multivariate classifier)

Performed search with D. Hitlin, I. Narsky, and B. Bhuyan

Also a FCNC, and therefore highly suppressed in the SM

23

Standard Model BF

Experimental Limit on BF

(90% CL)62.1

6.0 108.3 5104.1 arXiv:0708.4089v2 [hep-ex]

Analysis Procedure, TaggingPerform a ‘semileptonic’ tagged analysis

Fully reconstruct the ‘tag B’ in the decayLook at the rest of the event for our signal

24

νlDB 0

}{ 00 ππ,ππ,ππKD

0D

l

ν ν

ν

KTag B Signal B

BB

Analysis Procedure, CutsSeparately pursued two different techniques

to suppress backgroundStandard Rectangular Cut methodMore sophisticated Multivariate technique with

a Random ForestFor Rectangular Cuts, separated the Monte

Carlo (MC) into 3 sets: train, valid, test; in a 2:1:1 ratioOptimized the ‘Punzi’ Figure of Merit:

25

BNS

2

Where S is the number of signal, Nσ is the sigma level of discovery, and B is the number of background

Rectangular Cut Results

26