Measurement of |Vub|with B → ℓ

Masahiro Morii

Harvard University

Laboratory for Particle Physics and Cosmology

August 2006 M. Morii, Harvard 2

Motivation |Vub| determines the left side of the UT

Precise |Vub| and sin2 strong constraint on the UT

Uncertainty on |Vub| is dominated by theory errors Measurements with different methods important Inclusive B → Xuℓ

Use difference in kinematics to separate uℓ from cℓ Theory (OPE, SCET) must predict signal spectra

Current theory error ~5% of |Vub|

Exclusive B → ℓ, ℓ, ℓ, … Better S/B, esp. if we’ve tagged one B Theory (LCSR, LQCD, etc.) must predict form factors

Theory error hard to quantify

* 2( , , )Xp m ql

August 2006 M. Morii, Harvard 3

B → ℓ Form Factors

Two (among many) types of calculations Light-Cone Sum Rules

Latest Ball/Zwicky (PRD71:014015) quote 10-13% error at q2 = 0 Not valid above q2 ~ 14 GeV2

Lattice QCD Older calculations were “quenched” extra 15% error Unquenched calculations from HPQCD (PRD73:074502) and

Fermilab (hep-lat/0409116) quote ~11% systematic error at high q2

Not valid below q2 ~ 15 GeV2

Theory errors on |Vub| comparable to the inclusive approach We must measure partial rates in q2 bins

222

02 3

2 3

( )

24( )F

ub

Gd BV f qp

dq

−

+

+Γ →=

l 222

02 3

2 3

( )

24( )F

ub

Gd BV f qp

dq

−

+

+Γ →=

l 01 for

2B + +× → l

August 2006 M. Morii, Harvard 4

Previous Measurements

Untagged measurements have better statistics Background and cross-feed (from ℓ) higher

Tagged-B measurements have better S/B Statistics limited Binning in q2 requires large statistics

Semileptonic recoil Balance between efficiency and purity

Measurement Reference B(B0 → −ℓ+v) × 104

CLEO untagged PRD68:072003 (2003) 1.33 ± 0.18 ± 0.11 ± 0.07

BABAR untagged PRD72:051102 (2005) 1.38 ± 0.10 ± 0.16 ± 0.08

Belle semileptonic tag hep-ex/0604024 1.38 ± 0.19 ± 0.14 ± 0.03

BABAR hadronic tag hep-ex/0408068 1.08 ± 0.28 ± 0.16

Errors are statistical, systematic, and FF dependence

August 2006 M. Morii, Harvard 5

Analysis Flow

Event preselection

Find D(*)ℓv tag(s)

Find ℓv candidate(s)

Pick the best candidate

Extract signal yield

Divide by efficiency

Reconstruct D and D*

Combine with lepton

Recoil of the tag containsℓ and little else

Allow one candidate/event

Fit cos2B distirubtion

Double-tag sample determines data/MCBranching fraction

ℓ D(*)

v

v

ℓ

signal Btag B

a.k.a. Y Data sample contains232 M BB events

Data sample contains232 M BB events

August 2006 M. Morii, Harvard 6

D(*) Reconstrction Reconstruct D mesons

Reconstruct D*+ mesons

Channel D mass window (MeV)

D0 → K −+ 1863.8 ± 15.6

D0 → K

−+−+

1863.8 ± 12.5

D0 → K −+0 1863.8 ± 29.1

D0 → KS+− 1863.8 ± 15.1

D+ → K −++ 1868.8 ± 13.0

Twice as wide as the other channels

Little statistics in this channel

Channel mD* − mD window (MeV)

D*+ → D0+ 145.5 ± 3.0

D*+ → D+0 140.7 ± 3.0

We use mD sidebands to subtract combinatoric background, assuming linear distribution

August 2006 M. Morii, Harvard 7

B → D(*)ℓ Tag Combine a D(*) candidate with a lepton candidate with p* >

0.8 GeV in the CMS

Calculate

For correct tags, BY = anglebetween B and D(*)ℓ momenta

Signal should peak in−1 < cosBY < +1

Background is broad

(*) (*)

(*)

* * 2 2

* 2

2cos

2B BD D

BYB D

E E m m

p p

− −= l l

l

on-peak datab → uℓv MCB0B0 MCB+B− MCoff-peak data

cosBY

August 2006 M. Morii, Harvard 8

B → ℓ Signal Look for a lepton and a pion in the recoil side

Lepton p* > 0.8 GeV Pion with opposite charge

Nothing else left in the event No tracks in the drift chamber No cluster in the calorimeter

Calculate

Signal between ±1

* * 2 2

* 2

2cos

2B B

BB

E E m m

p p

− −= l l

ll

on-peak datasignal MCb → uℓv MCB0B0 MCB+B− MCoff-peak data

cosB l

August 2006 M. Morii, Harvard 9

Signal Kinematics Tag B and recoil B are back-to-back

Combine kinematical information into a single variable

cos2B < 1 for correctly-reconstructed signal events

2 22

2

cos cos 2cos cos coscos

sinBY B BY B

B γφ

γ+ +

= l l

Angle between the B momentum and the plane

defined by the D(*)ℓ and ℓ momenta

August 2006 M. Morii, Harvard 10

cos2B

Use cos2B distribution to distinguish signal from background

Background distributions are nearly flat Tested using sideband control samples

Perform unbinned maximum likelihood fitto extract signal yields in 3 bins of q2

on-peak datasignal MCBB-bar MC combinatoric background

q2 < 8 GeV2 8 < q2 < 16 GeV2 q2 > 16 GeV2

Nsig =23.8−7.2+7.8 Nsig =18.2−6.4

+7.2 Nsig =15.1−7.2+8.0

August 2006 M. Morii, Harvard 11

Double-Tag Sample Events with two non-overlapping tags

Number of double-tags (Tag efficiency)2

Selection of double-tag events reproducethe signal selection as closely as possible Not perfect – e.g., the number of remaining

neutral clusters depend on both sides

Compare data and MC

Error includes statistics, backgroundnormalization, Ncluster cut dependence, etc.

data

MC=

Ndata

NMC

=1.004 ±0.073

data

MC=

Ndata

NMC

=1.004 ±0.073 on-peak datasignalincorrect tagsbackground

August 2006 M. Morii, Harvard 12

B(B0 → –ℓ+ We measure the partial and total BFs (in 10-4)

q2 < 8GeV 2 8 < q2 <16GeV 2 q2 >16GeV 2 Total

0.50 ±0.16 ±0.05 0.33±0.14 ±0.04 0.29 ±0.15±0.04 1.12 ±0.25±0.10

August 2006 M. Morii, Harvard 13

Systematic Errors Main systematics are:

Tagging efficiency cos2B distribution of BB

background →ℓ and other Xuℓ

background in high-q2 bin Monte Carlo statistics

Still small comparedwith the stat. error Some of the errors are

intentionally conservative

August 2006 M. Morii, Harvard 14

Combining Analyses We combine results of the analyses by 3 groups

B0 semileptonic tag

B+ semileptonic tag

B0 hadronic tag

B+ hadronic tag

August 2006 M. Morii, Harvard 15

Measurement Reference B(B0 → −ℓ+v) × 104

CLEO untagged PRD68:072003 (2003) 1.33 ± 0.18 ± 0.11 ± 0.07

BABAR untagged PRD72:051102 (2005) 1.38 ± 0.10 ± 0.16 ± 0.08

Belle semileptonic tag hep-ex/0604024 1.38 ± 0.19 ± 0.14 ± 0.03

BABAR semileptonic + hadronic tag

hep-ex/0607089 1.33 ± 0.17 ± 0.11

Errors are statistical, systematic, and FF dependence

How We Compare

Competitive and statistics-limited result Paper has been submitted to Phys. Rev. Lett. Next steps:

Update with 211 400 fb-1 data Include other light hadrons (′)

August 2006 M. Morii, Harvard 16

Extraction of |Vub|

We use four calculations of the FF and find

c.f. HFAG average of inclusive measurements is

0

ub

B

Vζ τΔ

=Δ ⋅

B

0

ub

B

Vζ τΔ

=Δ ⋅

B 2max

2min

222 3 2

3( )

24

qF

q

Gf q p dqζ

+Δ = ∫2max

2min

222 3 2

3( )

24

qF

q

Gf q p dqζ

+Δ = ∫where

FF calculation q2 range Vub

(10−3)

Ball-Zwicky <16GeV 2 3.2 ±0.2stat. ±0.1syst.−0.4+0.5

FF

HPQCD >16GeV 2 4.5±0.5stat. ±0.3syst.−0.5+0.7

FF

Fermilab >16GeV 2 4.0 ±0.5stat. ±0.3syst.−0.5+0.7

FF

APE >16GeV 2 4.1±0.5stat. ±0.3syst.−0.7+1.6

FF

PRD71:014015

PRD73:074502

hep-lat/0409116

NPB619:565

Vub =(4.49 ±0.19exp ±0.27theo)×10−3

August 2006 M. Morii, Harvard 17

Summary |Vub| is a critical piece of the CKM “puzzle”

Harvard group makes strong contribution in this area We pursue two analyses based on complementary theoretical

approaches We measured B(B0 → –ℓ+) in the recoil of B0 → D(*)+ℓ– and

extracted |Vub|

Result (hep-ex/0607089) has been presented at ICHEP 2006 and submitted to Phys. Rev. Lett.

V

ub=(4.5±0.5stat. ±0.3syst.−0.5

+0.7FF ) ×10

−3using FF from a LQCD calculation