m. v. purohit, univ. of s. carolina, ichep 2004, beijing, china _ a search for d 0 d 0 mixing in...

M. V. Purohit, Univ. of S. Carolina, ICHEP 2004, Beijing, China

__ A Search for DA Search for D00DD0 0 mixing in mixing in

Semileptonic decays, A search for CP Semileptonic decays, A search for CP Violation in DViolation in D++ K K++KK-- decays decays

and a Measurement of and a Measurement of the Branching ratiothe Branching ratio

Milind V. Purohit

Univ. of South Carolina

(for the BaBar collaboration)


The BaBar Collaboration

China [1/5]Inst. of High Energy Physics, Beijing

Germany [4/28]Ruhr U BochumTU DresdenPhys. Inst., HeidelbergU Rostock

France [5/60]LAPP, AnnecyEcole PolytechniqueLAL OrsayLPNHE des Universités Paris 6/7CEA, DAPNIA, CE-Saclay

United Kingdom [10/72]U of BirminghamU of BristolBrunel UniversityU of EdinburghU of LiverpoolImperial CollegeQueen Mary & Westfield CollegeRoyal Holloway, University of LondonU of ManchesterRutherford Appleton Laboratory

Italy [13/105]INFN BariINFN FerraraINFN Frascati INFN GenovaINFN MilanoINFN NapoliINFN PadovaINFN PaviaINFN PerugiaINFN PisaINFN RomaINFN TorinoINFN Trieste

Canada [4/19]U of British ColumbiaMcGill UU de MontréalU of Victoria

Norway [1/3]U of Bergen

Russia [1/11]Budker Inst., Novosibirsk

Netherlands [1/5]NIKHEF

USA [38/301]Caltech, PasadenaUC, IrvineUC, Los AngelesUC, San DiegoUC, RiversideUC, Santa BarbaraUC, Santa CruzU of CincinnatiU of ColoradoColorado StateFlorida A&MHarvardU of IowaIowa State ULBNLLLNLU of LouisvilleU of MarylandU of MassachusetsMITU of MississippiMount Holyoke CollegeU of Notre DameOhio StateU of OregonU of PennsylvaniaPrairie View A&MPrincetonSLACUniv. of South CarolinaStanford USUNY, AlbanyU of TennesseeU of Texas at DallasU of Texas, AustinVanderbiltU of WisconsinYale U

(78 institutions,

609 Collaborators)


The BaBar Detector

•Silicon Vertex Tracker z hit resolution 15m•Drift CHamber (pT)/pT = 0.14% pT + 0.45%

•Detector of Internally Reflected Cherenkov light K- separation 4.2 @ 3.0GeV•ElectroMagnetic Calorimeter E/E = 2.3%E-1/4 1.4%

•RPC based Instrumented magnetic field Flux Return 18/19 layers of RPC in 60/65 cm of iron

~250 fb-1 of data collected so far.


Searching for New Physics in Charm CP violation with BaBar data

CP violation in charm is expected to first manifest itself in Singly Cabibbo-suppressed (SCS) decays.

Within the Standard Model, one expects CP violation asymmetries in SCS Decays ~10-3, while New Physics can give CP violation asymmetries ~10-2.

[G. Burdman & I. Shipsey, Ann. Rev. Nucl. Part. Sci., 2003, hep-ph/0310076. See also S. Bianco, F. L. Fabbri, D. Benson & I. Bigi, hep-ex/0309021.]

Current experimental limits are ~ (2 – 5) x 10-2 leaving a considerable window for new physics discovery.

The analysis reported here is based on ~43,000 D+ KK decays from ~80 fb-1. [BaBar has ~250 fb-1 of data.]


Standard Model interfering amplitudes:

A ~Vcs*¢ Vus A ~ Vcs

*¢ Vus, Vcd*¢ Vud, Vcb

*¢ Vub

•In the Wolfenstein parameterization, the CKM matrix (below) clearly gives only a small CP violating asymmetry.


Data Sample and Analysis• Our final sample contains ~43,000 D+ KK decays

from ~80 fb-1.• We measure the asymmetry

where

• We also measure the asymmetry in the & K* regions.


KK Yields

21632 ± 228 D+ events 20940 ± 226 D- events

23066 ± 217 Ds+ events 22928 ± 214 Ds

- events

•Likelihood ratio cut uses p* and beam-spot constrained 2

•Kaons have kaon ID, pions must not have kaon ID


ACP in , Regions of KK Dalitz Plot

5452 ± 87 D+ events 5327 ± 86 D- events

• mass is required to lie within 10 MeV/c2 of nominal mass

• |cos(H)| is required to be ≥ 0.2 ; H is helicity angle in rest frame

1. The region:


5247 ± 96 D+ events 5113 ± 96 D- events

• K* mass is required to lie within 50 MeV/c2 of nominal K* mass

• |cos(H)| is required to be ≥ 0.3 ; H is helicity angle in K* rest frame

2. The region:


Systematic Errors on ACP

• Systematic errors on ACP estimated as:– Largest difference using other normalizations (0.8%)– Largest uncorrected asym. in control samples (1.1%)– From table (in units of 10-2):

Source KK K*0K

MC simulation 0.06 0.06 0.06

Background estimate 0.63 0.32 0.49

Event Selection 0.51 0.56 0.54

Total 0.81 0.65 0.73


ACP Results


The Branching Ratio (D+ K-K+) / (D+ K-)

• Using measured yields and efficiencies in bins of the Dalitz plots, the total branching ratio is determined: 0.107 ± 0.001 ± 0.002

• Sources of systematic errors:Source Error (10-2)

PID, tracking 0.21

Background estimate 0.05

Event Selection 0.02

Total 0.22


The Branching Ratio Compared


Searching for New Physics in mixing with BaBar data

For decays of neutral Ds Rmix – The standard model predicts a low rate (~10-7) of for the

box diagram which goes up to ~10-3 when long distance effects are included; nevertheless we would like to observe mixing as a first step.

– New physics effects can easily produce > 10-7 rates of mixing, and are the only way we can get CP violation in mixing.

[G. Burdman & I. Shipsey, Ann. Rev. Nucl. Part. Sci., 2003, hep-ph/0310076. See also S. Bianco, F. L. Fabbri, D. Benson & I. Bigi, hep-ex/0309021.]

The analysis reported here is based on ~50,000 semi-electronic decays of neutral D mesons from ~87 fb-1 with mixing rate sensitivity down to ~10-3.

DD M ixing


(A. Petrov, hep/ph 0311371)

• Possible enhancements due to new particles and interactions in new physics models:

•gluinos, squarks•fourth generation quarks

•lepto-quarks, etc.• No CP-violating effects

expected in SM -- CP violation in mixing would be unambiguous signal of new physics

x=M/y=/2

x=M/

mix

ing

ra

te =

|am

pli

tud

e|2

mix

ing

ra

te =

|am

pli

tud

e|2

New Physics Mixing Predictions

current experimental sensitivity

Charm Mixing, continuedSM Mixing Predictions


cc continuum event topology

K

e,

slow

other charmed hadron(will hadronically reconstruct in next iteration of analysis)

interaction point

beamspot D0

D*+ D0 +tag

D0 K+e-

D*- D0 -tag

D0 K-e+

Wrong-sign mixed decays

D*+ D0 +tag

K-e+

D*- D0 -tag

K+e-

Right-sign unmixed decays


The Analysis Strategy

4exp)(

222

00

yxtttDD

WS

exponentialenvelope

Quadratic time dependence

mixing rate

• We use neutral D mesons from D*+ decays: D*+ D0

• Flavor at birth is tagged by pion from D* decay• Flavor at decay is tagged by electron: D0 (K-/K*-) e+e

Clearly, an e+ signifies a RS (Right Sign, or unmixed) D0

while an e- would signify a WS (Wrong Sign or mixed) D0

• The mixing rate is given by

• where x M / , y / 2 • No DCS decay background exists


Semileptonic Mixing Analysis Technique

• A neural network event selector is used.• Another neural network is used for p*(D0) reconstruction.

(p* is the c.m. momentum)• An unbinned extended maximum likelihood fit

using M & transverse lifetime is then done, where M m(D0) – m(D0)• First, a fit to high-statistics RS sample gets

– signal M shape and – unmixed D0 lifetime

for use in WS pdf and to get N(unmixed), the normalization for Rmix.

• Subsequent fit to WS sample for N(mix)• Rmix ≈ N(mix) / N(unmixed)• Data sample: 80 fb-1 on resonance, 7.1 fb-1 off-resonance


• Unbinned extended maximum likelihood fit to transverse lifetime and M = M(D*)-M(D0) with 15 floated parameters

RS Unmixed Fit to 87 fb-1 of Data

D+D0 bkgdzero life

M signal region

M sideband

D+

D0 bkgdzero life

lifetime tails

D0 sgnlD0 sgnl

Results of fit:

• Unmixed D0 yield: 49620 ± 324 evts (stat)

• M and lifetime pdfs


WS Mixed D0 Fit to 87 fb-1 of Data

•Mixed signal pdf parameters taken from high-statistics unmixed data fit– Ni and zero lifetime triple Gaussian

parameters floated (11 parameters)

Random D+

Random D0

Zero Life

M projection D0 signal

Peaking D+

•Distribution of N(mix) from fits to an ensemble of 170 WS generic MC datasets with zero embedded mixed signal events

(~5% probability of getting a larger result for Rmix=0)

•Unblinded N(mix): 114 ± 61 evts


WS Mixed D0 Fit to 87 fb-1 of Data

zero life

Lifetime projection showing mixed signal

random D+

random D0

D0 signal

pkng D+

Full lifetime projection


stat stat + syst

90% Confidence limiton number of mixingsignal.

Final result for semileptonic mixing• The fit gives a yield of N

mix=114 ± 61 wrong sign signal events.

• Systematics evaluated as fraction of statistical error:

• Normalising to number of right-sign events gives result on mixing.

Rmix 0.0023 0.0012 stat 0.0004 syst

Rmix 0.0042 90% C.L.

Systematic ErrorMixed m PDF 0.27Mixed decay time PDF 0.06Combinatoric m PDF 0.13Bkg D0 decay model 0.13Bkg D+ decay model 0.10Total systematic 0.35Total stat. + syst. 1.06


Comparison to other results

Summary of all BaBar D mixing measurements:

•Hadronic (K) analysis:

Rmix < 1.3 x 10-3

•Lifetime difference for D0 KK

and D0

yCP = 0.8 ± 0.4 +0.5 -0.4 %

•This (semileptonic) analysis: Rmix < 4.2 x 10-3 (90% CL)


Conclusions

• We have demonstrated that

ACP is consistent with zero. The measured values are:

D+ KK+0.0136 ± 0.0103 ± 0.0110

D+ +0.0024 ± 0.0152 ± 0.0080

D+ K*K+0.0088 ± 0.0177 ± 0.0080

• We find that the Branching Ratio for D+ KKis:

0.1070 ± 0.0009 ± 0.0022

• We find, using semi-electronic neutral D decays, that

Rmix < 4.2 x 10-3 (90% CL)


Acknowledgements

• Thanks to all BaBar collaborators including, particularly:

• Francisco Yumiceva, South Carolina graduate student (D+ K+K- decays)

• Kevin Flood, graduate student at the

Univ. of Massachusetts, Amherst

(semileptonic mixing analysis)


Backup Slides


The Status of CP Violation

CP violation is necessary if the universe is indeed matter-antimatter asymmetric but the Big Bang is not

Standard Model CP violation cannot explain CP violation in the universe

New physics is needed, SM is not enough

Charm may be a good place to look


Raw CP asymmetries in control samples

• Ds+ K+K- decays: +3.0 x 10-3

• D+ K-+ decays: -2.9 x 10-3


Event Sample

• Nearly all Run 1, 2 on/off-resonance (R10 processing)

–80.0 fb-1 on-resonance

–7.1 fb-1 off-resonance

mode N events (x106) equiv lumi (fb-1)b0 154 294b+ 156 298cc 185 142uds 293 140

• SP4 generics

• Wrong-sign mixed D0 lifetime: 100k events


Track Selection• Pion candidates

– GoodTracksVeryLoose– .45 < lab < 2.5– p* < 0.45 GeV/c– track fit prob. > 0.001– beamspot refit prob. > 0.01– ≥ 2 SVT r-phi (z) hits with at

least 1 hit on inner 3 r-phi (z) layers

– ≥ 6 total SVT hits

• Multiple D0 candidates– Veto events with more than

one RS or WS D0 candidate passing all selection criteria (~11% of RS signal lost after other cuts)

• Kaon candidates– GoodTracksVeryLoose– KLHVeryTight

• plus KMicroVeryTight (plab > 2.1 GeV/c)

– .45 < lab < 2.5

• Electron candidates– GoodTracksVeryLoose– PidLHElectrons (default)– .45 < lab < 2.409– 0.8 < E/p < 1.05– conversion veto

• K/e vertex– GeoKin vertex prob. > 0.01– M(KeVtx) < 1.82 GeV/c2

– Lifetime error < 2 D0 lifetimes


Neural Network Event Selection

• Events are selected using a neural network with the following inputs taken from signal and generic MC:

– p*(K/e vertex)– p*()– thrust magnitude (w/o K, e)– Opening angle between p*(K/e) and

thrust (w/o K, e)– opening angle between p*(K) and p*(e)

• RS/WS signal selection efficiencies are identical

• RS/WS signal/bkgd NN event selector output

signal

bkgd

Lumi-scaled NN output

signal

bkgd

Normalized NN output

final cut optimized for best statistical sensitivity

final cut optimized for best statistical sensitivity

final cut


•Novel use of two hidden-layer neural network to map NN input vector to signal p*(D0)

• Momentum resolution fit to double gssn– phi core sigma/fraction: 82 mrad / 0.81– theta core sigma/fraction: 80 mrad / 0.94– total magnitude RMS: 371 MeV/c– transverse magnitude RMS: 350 MeV/c

•Trained with inputs from signal MC:– p*(K/e vertex)– p*()– thrust vector (w/o K, e)– Opening angle between p*(K/e) and thrust (w/o

K,e)– Opening angle between p*() and thrust (w/o

K,e)– opening angle between p*(K) and p*(e)– opening angle between p*(K/e) and p*()

Neural Network D0 Reconstructionp*(D0) Neural Network Residuals

GeV/c

GeV/c

rad

rad


zero lifetime

D+

WS mixed D0

RS unmixed D0 mis-reco’d D0

1-d lifetime pdfs

M, Lifetime PDFs

signal

D+ (RS only)

1-d M pdfs

random comb.

signal (zoom)

off-resdatafit

fit

fit

shapefromMC

shapefromMC

fitfromRS fit

M = M(D*) - M(D0)

MC/fit


Testing the Mixed Fit with Generic MC and Embedded Mixed Events

• WS mixed fit tested with full fit on ~540 lumi-scaled generic MC datasets with different levels of embedded mixed events

• N(mix) pull plots (right) show no evidence of bias or improperly scaled errors in the fit number of N(mix) with or w/o the presence of mixed events

N(mix) R(mix) fit mean err stat mean0 0 0.008 0.089 -0.06650 0.001 0.05 0.097 0.091

100 0.002 0.012 0.07 0.012

N(mix) R(mix) fit sigma err stat rms0 0 0.965 0.079 0.99450 0.001 0.995 0.093 0.999

100 0.002 0.994 0.052 0.994


Goodness-of-Fit to Run 1,2 Data

•Toy MC datasets generated from RS/WS data fit pdfs and fit with unmixed/mixed fit models to determine goodness-of-fit

•Both RS/WS NLL values from data fits lie well within range predicted by the fits to toy MC

NLL Distribution for WS Toy Datasets

NLL distribution for RS toy datasets

WS data fit NLL value

RS data fit NLL value


• No significant variations in the mixing rate were found when making the following changes:– TwoTrksVtx → Hadronic beamspot– GeoKin K,e vertex → FastVtx– KLHVeryTight → KMicroVeryTight– E/p <1.05 → 1.10– Lifetime error cut +/- 10%– Separate fitting of initial D0 and D0 – Different NN event selector cuts

Evaluation of Systematics

• Systematic checks fall into two categories:– reasonableness or “sanity”

checks– systematic variations which

encode lack of knowledge/ understanding and biases in the fit model

• First class demonstrates robustness of result

• Latter class determines quantitative estimation of the systematic error

Reasonableness Checks


RS Unmixed M, Lifetime PDF Fit Classes

RS random M D0 lifetime

RS D+

RS random M zero lifetime

RS peaking M D0 unmixed lifetime

full M range

full M range

full M range

M M M

M M

M M

c


WS Mixed M, Lifetime PDF Fit Classes

Random comb. M D0 lifetime

WS random M D+ lifetime

Random comb. M zero lifetimePeaking M D+ lifetime

WS peaking M D0 mixed lifetime

full M range

full M range

full M range

MM

M M

M M

M M

c

c


stat

•Floated unmixed fit RS signal M shape parameters correlated, so N(mix) systematic error from this source calculated using RS fit correlation matrix

Quantitative Systematic Error

• Total systematic error

Delta M shape 0.00032Unmixed D0 lifetime 0.00008Random comb. shape 0.00015Lifetime resolution model 0.00000Bkgd D0 lifetime model 0.00016Bkgd D+ lifetime model 0.00012

sys

•List of systematic errors:


Semileptonic Mixing Summary

Rmix = 0.0023 ± 0.0012 (stat) ± 0.0004 (sys)• Rmix consistent with no mixing

• CP fit was performed and no CP-violating effects were found

• Rmix < 0.0047 (95% C.L.) from NLL scan

• E791: rmix < 0.50% @ 90% CL

• FOCUS: rmix < 0.05% @ 90% CL

rmix < 0.10% @ 90% CL(using Feldman-Cousins)

Stat error is ~0.1%FOCUS result is UNPUBLISHED

(Plot courtesy of G. Burdman & I. Shipsey, hep/ph 0310076)

N(mix) NLL Scan

m. v. purohit, univ. of s. carolina, ichep 2004, beijing, china _ a search for d 0 d 0 mixing in...

Documents

charm cp violation

cp violation asymmetries

babar datacp violation

d kkp decays

semileptonic decays

p decays

small cp

babar collaborationm