D1D2 lifetime difference using D0KK, D0and D0K

1
D1-D2 lifetime difference usingD0?K?K?,
D0??????and D0?K???

• W. Lockman, C. Chavez, J. Coleman, R. Cowan, K.
  Flood, B. Petersen
• BaBar Physics Workshop
• June 30 - July 2, 2008

• Introduction
• Analysis
• Fit results
• Cross checks and validations
• Systematics

2
Introduction

• Mixing among the lightest neutral meson flavor
  eigenstates provides important information about
• electroweak interactions, including CP violation
• the CKM matrix
• mixing loop virtual constituents
• D0 system exhibits the smallest mixing
• Short distance Standard Model (SM) suppression
• D mixing loop involves down-type quarks
• b quark loop suppressed
• s and d quark loops GIM suppressed
• mass difference amplitude O(10-5) or less
• long distance mixing amplitudes predominant
  but hard to quantify

3
Mixing between Flavor States

• Flavor eigenstates can mix through weak
  interaction
• Mass eigenstates D1 and D2 ? flavor eigenstates
• If weak interaction splits the masses or widths
  of mass eigenstates, flavor state mixing will
  occur, as seen from the time evolution
• mixing parameters

Schroedinger eqn governs time evolution (off
diagonal M and ? elements determine mixing)
D1 CP? D2 CP?
In the limit of CP conservation
4
Mixing Observable

• This analysis measures the lifetimes of
• Cabibbo favored D0?K????(CF) and
• singly Cabibbo suppressed D0?K?K???SCS??decays
• This allows a determination of by measuring
• In the limit of CP conservation,
• No D flavor tag used
• Higher background to signal than in the tagged
  analysis, but 4x the statistics
• compared to tagged analysis, expect smaller
  statistical, larger systematic errors
• No measurement of CP violating quantity, ?Y
• Construct the untagged sample to be disjoint from
  the tagged sample
• allows tagged and untagged results to be
  trivially combined

5
Previous yCP Measurements

• yCP world average from HFAG

A. Schwartz, arXiv0803.0082
540 /fb tagged (BELLE)
384 /fb tagged and 91 /fb untagged (BaBar)
(1.132 ? 0.266)
6
Previous Mixing Measurements

• Best evidence for D0-D0 mixing to date (mainly
  BaBar, BELLE and CDF)

yCP (1.132 ? 0.266)
consistent? (y includes yCP)
7
Untagged Analysis Strategy

• Samples
• Untagged D0?K???
• Untagged D0?K?K?
• Systematics considerations
• signal systematics correlated and mostly cancel
  in yCP
• background composition different between samples
  
• background systematics do not cancel in yCP
• restrict sample to narrow D0 mass region
  symmetric about D0 peak
• 1.8495 lt D0 lt 1.8795 GeV/c2
• estimate background from sideband regions
• 1.81 lt D0 lt 1.83 GeV/c2 and 1.90 lt D0 lt 1.92
  GeV/c2
• Background composition
• mainly combinatoric, small admixture of broken
  charm decays

removed events containing a D tagged decay
8
Untagged Selection

• Selection requirements
• ??? tracks from a common point
• KLHTight, piLHTight PID selectors
• D0 invariant mass m, reconstructed decay time t
  and its error ?t from beam constrained TreeFitter
  vertex fit
• P(?2) gt 0.1
• 1.80 lt m lt 1.93 GeV/c2
• -25 lt t lt 25 psec
• ?t lt 0.5 psec
• Remove B decays using D0 center of mass momentum
  cut
• P gt2.5 GeV/c
• D0 daughter track number of DCH hits
• NDCH ? 12
• Reduce uds backgrounds using helicity angle
  (angle between ?track in D0 rest frame and D0
  boost direction) cut
• cos?h lt 0.7
• Remove events containing a selected D tagged
  D0?????, K??, K?K? decay
• For multiple D0 candidates sharing tracks, keep
  the one with highest P(?2)

9
2008 untagged analysis

• Data yields and purity in 1.8495 lt m lt 1.8795
  GeV/c2

K?
KK
10
Untagged Analysis Fit Strategy

• Perform fit in stages to determine lifetimes of
  D0?K????and D0?K?K? decays
• Fit D0 invariant mass distribution to determine
  signal yields in signal band
• Background decay time fit
• D0 sidebands fit to determine dominant
  combinatorial background
• use MC to determine signal and charm decay time
  distributions in sidebands
• signal region backgrounds
• charm component from MC
• combinatorial shape and yield parameters
  weighted average from sidebands
• Signal decay time fit
• Signal decay time PDF
• Exponential convolved with triple Gaussian with
  common offset t0 and scaled event-by-event
  errors
• Two different signal fits
• baseline combined fit to K? and KK samples with
  shared resolution function parameters
• scale KK width relative to K?
• independent fits to K? and KK samples with
  independent resolution parameters
• signal region decay time error histogram PDFs
• combinatorial decay time error summed
  distribution from sidebands
• Combinatorial ? total decay time error signal
  decay time error distribution

11
D0 Mass Data MC Comparison

• KK fit
• 2 Gaussians signal PDF
• 2nd order polynomial background PDF
• K? fit
• 2 GaussiansBifurcated Gaussian signal PDF
• 2nd order polynomial background PDF
• Data-MC fit comparison
• disagreement between data and MC in sideband
  background yields
• good agreement between data and MC in signal
  region background yields (next)

K? MC
KK MC
K? data
KK data
12
Data MC Mass Fit yields
K?
MCTruth
K? yields agree to within 5 in the signal box
MC fit
Data fit
KK
MCTruth
KK yields agree to within 2 in the signal box
MC fit
Data fit
13
KK MC combinatorial fits
lower sideband
upper sideband
signal region
double Gaussianwith common mean and CB function
14
K? combinatorial fits
lower sideband
upper sideband
signal region
double Gaussianwith common mean and CB function
15
Comb. Fit Validation in Signal Region
KK MC
K? MC

• Signal region PDF parameters weighted average of
  parameters from sidebands
• Yields scaled to signal region yields
• Weighted averaging of parameters gives a PDF
  which matches truth matched combinatorial
  distribution

16
KK MC Sideband Decay time fit
lower
Upper

• Lower and upper sideband KK MC PDFs overlaid on
  Decay time distributions

17
K? MC Sideband Decay time fit
lower
Upper

• Lower and upper sideband K? MC PDFs overlaid on
  Decay time distributions

18
MC Signal Region fits
KK
K?

• KK and K? MC PDFs overlaid on Decay time
  distributions

19
Data Signal Region Fit
KK
K?
combined

• Independent fit results

Combined fit results
20
Validation and Cross Checks

• Perform unblinded fit to determine K? lifetime
• results are close to world average
• Fit truth decay time Monte Carlo samples
• no deviation seen from exponential
• Compare resolutions between K? and KK
• (t-ttrue)/(SKK?t) distributions are universal
  with SK?1, SKK1.018
• Compare fit to truth matched combinatorial
  distribution in signal region with PDF determined
  from weighted average of sideband fits
• results are nearly identical
• Compare combinatorial decay time error
  distributions from truth matched MC to
  distributions obtained from sidebands
• results are nearly identical

21
Validations and Cross Checks

• Fit reconstructed Monte Carlo samples
• compare with lifetime from fitting MC truth
  lifetime
• Problem fit to KK to generic Monte Carlo sample
  indicates a possible bias or a cruel statistical
  fluctuation
• both fitted and truth lifetimes are low,
  (409.60.6) fsec, dialed 411.6 fsec
• Construct an independent sample using KK signal
  and generic MC for background
• fitted and truth lifetime analysis still under
  investigation, but preliminary results indicate
  lifetimes are coming out a bit higher than
  dialed, 412 fsec
• Kp fitted and dialed lifetimes are consistent
  with each other

22
Systematic Sources Under Investigation

• Fit procedure
• Difference between combined and independent fits
• Difference between combined fits with and without
  mass dependence
• Signal systematics
• Offset in combined signal fit
• Signal yield from mass fit
• Signal box position
• Signal box size
• Detector
• SVT misalignment
• Beam spot
• Magnetic field uncertainties

• Background systematics
• Combinatorial yield
• Sideband positions
• Sideband widths
• Charm yields in signal, sidebands
• Charm lifetime
• Selection
• Decay time error selection
• Decay time selection
• Multiple overlapping candidates

• My Guess
• largest systematics in red
• total systematic error 1 to 2 fsec

23
Status of Untagged Analysis

• Decay time fits have been validated on luminosity
  weighted cocktail MC
• can reproduce the true lifetime to within 0.2-0.3
  fs
• Blind decay time fits to data are completed
• statistical errors appear to be sensible
• Systematic and cross check studies are underway
• BAD 1876 is completed up to systematics
• Still striving to have a preliminary result for
  ICHEP

24
Extra material
25
True decay time distributions

• MC Truth decay time distributions

KK standard generic
KK embedded generic

• Dialed and post selection MCtruth lifetime
  comparisons
• K???generic
• KK embedded
• KK signal
• KK generic

KK signal sample
K? standard generic
26
Decay time pulls vs true lifetime
Decay time pull (t-ttrue)/?t of signal events in
1.80 lt m lt 1.93 GeV/c2 mass region
KK
K?

• Decay time pulls consistent with no true decay
  time dependence

27
Scaled Pull Comparisons

• Kolmogorov comparisons of scaled pull ??
  (t-ttrue)/(S?t) distributions

• Compare truth matched signal scaled pull
  distributions
• K? from generic sample, SK? 1
• KK from signal Monte Carlo sample, SKK 1.018
• Scaled pull distributions statistically
  consistent with each other
• KS probabilities are 25-45
• Motivates combined fit anzatz

28
Mass-lifetime correlation Vertexing bias
t-ttrue vs reconstructed D0 mass

• Tagged K?
• Untagged generic K?
• Untagged generic KK
• Untagged signal KK

• Cross over point
• (m, t-ttrue) (mD0, 0)
• same in all distributions
• Steeper dependence in KK than in K?

29
Combinatorial Background Error distrib.
30
Signal Error Distribution
31
KK MC signal region decay time fit
Combinatorial background Shape is taken
from Weighted average of lower and Upper
sidebands
Charm background Shape and yields taken from MC
Charm fit in Signal Region
?true??fitted 0.3 fs
32
K? MC signal region decay time fit
Combinatorial background Shape is taken
from Weighted average of lower and Upper
sidebands
Charm background Shape and yields taken from MC
Charm fit in Signal Region
?true??fitted 0.2 fs
33
KK data signal region decay time fit
Combinatorial background Shape is taken
from Weighted average of lower and Upper
sidebands
Charm background Shape and yields taken from MC
Charm fit in Signal Region
Lifetime xxx.xxx ? 1.26 fs KK?tagged sample
error 1.8 fs untagged sample offset -7.4 ? 1.2
fs tagged sample offset -4.8 fs
34
K? data signal region decay time fit
Combinatorial background Shape is taken
from Weighted average of lower and Upper
sidebands
Charm background Shape and yields taken from MC
Charm fit in Signal Region
Lifetime 409.88 ? 0.36 fs K??tagged sample
error 0.7 fs untagged sample offset -4.64 ?
0.28 fs tagged sample offset -4.8 fs