Figures and Tables for Search for the B_s→μ⁺μ^- and B_d→μ⁺μ^- decay

Primary Authors: Douglas Glenzinski, Matthew Herndon, Walter Hopkins, Teruki Kamon, DaeJung Kong, Vyacheslav Krutelyov, Cheng-Ju Lin, David Sperka, Julia Thom, Satoru Uozumi

Data Validation
Baseline Requirements
Neural Network (NN) Signal Discrimination
Neural Network Cross Checks
Combinatoric Background
B→ hh Background
Cross Checks of Background Estimates with Control Samples
Systematics of ε_NN
Signal Expectations
Unblinding
Old NN Cross Check
Results: B_d Signal Window
Results: B_s Signal Window
Statistical Interpretation of excess in B_s signal window

Data Validation

Validation plots comparing kinimatic variable distribution for B⁺ sample of miniskirt data with first 3.7 fb^-1 and last 3.3 fb^-1: eps, png
Validation plots comparing NN input variables distribution for B⁺ sample of miniskirt data with first 3.7 fb^-1 and last 3.3 fb^-1: eps, png
Validation plots comparing NN input variables (cont.) distribution for B⁺ sample of miniskirt data with first 3.7 fb^-1 and last 3.3 fb^-1: eps, png
Validation plots comparing kinimatic variable distribution for B⁺ sample of COT spacer data with first 3.7 fb^-1 and last 3.3 fb^-1: eps, png
Validation plots comparing NN input variables distribution for B⁺ sample of COT spacer data with first 3.7 fb^-1 and last 3.3 fb^-1: eps, png
Validation plots comparing NN input variables (cont.) distribution for B⁺ sample of COT spacer data with first 3.7 fb^-1 and last 3.3 fb^-1: eps, png

Baseline Requirements

Yields after baseline requirements for B⁺: eps, png
Yields after baseline requirements for di-muon sample: eps, png

Neural Network (NN) Signal Discrimination

Comparison of old and new NN signal and background efficiencies: eps, png
Table of NN input variables ranked in order of significance: png
Signal discrimination for NN input variables (variables ranked highest by NN): eps, png
Signal discrimination for NN input variables and additional kinematic variables: eps, png
Signal discrimination for NN input variables: eps, png
Signal discrimination for NN input variables (cont.): eps, png
NN output for MC Signal and mass sidebands for CC channel: eps, png
NN output for MC Signal and mass sidebands for CF channel: eps, png
NN output for MC Signal and mass sidebands for CC and CF combined: eps, png

Neural Network Cross Checks

Mass bias check (NN output vs Mass): eps, png
Mass bias check (Average NN output vs Mass) for OS- control sample: eps, png
Mass bias check (Averag NN output vs Mass) for FM+ control sample: eps, png
Training on inner and out mass sidebands and compare NN output for CC: eps, png
Training on inner and out mass sidebands and compare NN output for CF: eps, png
Signal and background efficiency for different SB training sets: eps, png
Background efficiency for NN's trained on different fractions of the data: eps, png
Background efficiency for NN's trained on different fractions of the data (zoomed into the NN>0.7): eps, png

Combinatoric Background

Blinded 2D plot of NN output vs mass: eps, png
Mass sideband distribution for all NN bins combined (used for slope of combinatoric background): eps, png
Blinded mass distributions for first 4 NN bins (0.7<NN<0.94) in CC: eps, png
Blinded mass distributions for last 4 NN bins (0.94<NN<1.0) in CC: eps, png
Blinded mass distributions for first 4 NN bins (0.7<NN<0.94) in CF: eps, png
Blinded mass distributions for last 4 NN bins (0.94<NN<1.0) in CF: eps, png
Tables summarizing combinatoric background estimates for CC and CF: png
Full combinatoric table with information for all NN bins and mass bins for both CC and CF in Bs mass window: png
Full combinatoric table with information for all NN bins and mass bins for both CC and CF in Bd mass window: png

B→ hh Background

D0 mass plots for specific p_T bin for CC: eps, png
D0 mass plots for for higher p_T bin with wider distributions for CC: eps, png
Fake Rates for pions and kaons as a function of p_T for CC: eps, png
Fake Rates for pions and kaons as a function of p_T for CF: eps, png
Cross check of NN distribution for fake rate p_T dependence for CC: eps, png
Cross check of NN distribution for fake rate p_T dependence for CF: eps, png
Table of final B→ hh estimates: png
Full B→ hh table with information for all NN bins and mass bins for both CC and CF in Bs mass window: png
Full B→ hh table with information for all NN bins and mass bins for both CC and CF in Bd mass window: png

Cross Checks of Background Estimates with Control Samples

Fake Rates in FM+ sample for pions and kaons as a function of p_T for CC: eps, png
Fake Rates in FM+ sample for pions and kaons as a function of p_T for CF: eps, png
FM+ Dimuon mass distributions for last 4 NN bins (0.94<NN<1.0) in CC: eps, png
FM+ Dimuon mass distributions for last 4 NN bins (0.94<NN<1.0) in CF: eps, png
Table with B→ hh contribution to FM+ background estimates: png
Control Table for CC channel: png
Control Table for CF channel: png
Control Table combining all NN bins: png

Systematics of ε_NN

Comparison of kinematic variables and NN input variables for B⁺ data and MC: eps, png
Comparison of NN input variables for B⁺ data and MC: eps, png
Comparison of NN input variables for B⁺ data and MC (cont.): eps, png
Comparison of NN ouput for B⁺ sideband subtracted data and B⁺ MC for CC: eps, png
Comparison of NN ouput for B⁺ sideband subtracted data and B⁺ MC for CF: eps, png
Comparison of NN ouput for B⁺ sideband subtracted data and B⁺ MC for CC (linear scale): eps, png
Comparison of NN ouput for B⁺ sideband subtracted data and B⁺ MC for CF (linear scale): eps, png
ε_NN cross check table, comparig MC with data for B⁺ normalization mode: png

Signal Expectations

Summary table of acceptances and efficiencies: png
Table of Standard Model expectation for B_s: png
Summary table of expected backgrounds and SM yields: png

Unblinding

NN output vs mass for all NN bins: eps, png
NN output vs mass for highest NN bins: eps, png
Unblinded mass distributions for first 4 NN bins (0.7<NN<0.94) in CC: eps, png
Unblinded mass distributions for last 4 NN bins (0.94<NN<1.0) in CC: eps, png
Unblinded mass distributions for first 4 NN bins (0.7<NN<0.94) in CF: eps, png
Unblinded mass distributions for last 4 NN bins (0.94<NN<1.0) in CF: eps, png
Table of signal information (run, event number, NN output, and mass): html

Old NN Cross Check

Mass distribution for old NN for CC: eps, png
Mass distribution for old NN for CF: eps, png
Distribution of old NN output for different new NN output section for B_s: eps, png
Distribution of old NN output for different new NN output section for B_d: eps, png

Results: B_d Signal Window

Di-muon Mass Plots: The hashed area represents the total uncertainty on the background estimate calculated by adding in quadrature the systematic uncertainty on the mean expected background in each bin with the associated poisson uncertainty.
- Comparison of background and observations, combining CC and CF: eps, png
- Comparison of background and observations, separating CC and CF, but combining 5 lowest NN bins: eps, png
Di-muon Mass Plots: The hashed area represents the systematic uncertainty on the mean expected background while the error bars on the points represent the associated poisson uncertainty.
- Comparison of background and observations, combining CC and CF: eps, png
- Comparison of background and observations, separating CC and CF, but combining 5 lowest NN bins: eps, png
Full table of expectations and observations: png
Table of expectations and observations for the 3 highest NN bins: png
1-CL_s plot: eps, png

Results: B_s Signal Window

Di-Muon Mass Plots: The hashed area represents the total uncertainty on the background estimate calculated by adding in quadrature the systematic uncertainty on the mean expected background in each bin with the associated poisson uncertainty.
- Comparison of background, 5.6xSM, and observations, combining CC and CF: eps, png
- Comparison of background, 5.6xSM, and observations, separating CC and CF, but combining 5 lowest NN bins: eps, png
- Comparison of background, SM, and observations, combining CC and CF: eps, png
- Comparison of background, SM, and observations, separating CC and CF, but combining 5 lowest NN bins: eps, png
Di-muon Mass Plots: The hashed area represents the systematic uncertainty on the mean expected background while the error bars on the points represent the associated poisson uncertainty.
- Comparison of background, 5.6xSM, and observations, combining CC and CF: eps, png
- Comparison of background, 5.6xSM, and observations, separating CC and CF, but combining 5 lowest NN bins: eps, png
- Comparison of background, SM, and observations, combining CC and CF: eps, png
- Comparison of background, SM, and observations, separating CC and CF, but combining 5 lowest NN bins: eps, png
Full table of expectations and observations: png
Table of expectations and observations for the 3 highest NN bins: png
Event Displays for NN highest scoring events: pdf
1-CL_s plot: eps, png

Statistical Interpretation of excess in B_s signal window

p-value plot for background only null-hypothesis: eps, png
p-value plot for background only null-hypothesis using only the two highest NN bins (cross check): eps, png
p-value for background only null-hypothesis using only the 3rd highest NN bin (cross check): 0.46%
p-value for background only null-hypothesis using all NN bin in CC (cross check): 0.15%
p-value for background only null-hypothesis using all NN bin in CF (cross check): 89%
p-value plot for SM+background null-hypothesis: eps, png
p-value plot for SM+background null-hypothesis using only the two highest NN bins (cross check): eps, png
p-value plot SM+background null-hypothesis using only the 3rd highest NN bin (cross check): 0.55%
p-value for SM+background null-hypothesis using all NN bin in CC (cross check): 0.97%
p-value for SM+background null-hypothesis using all NN bin in CF (cross check): 85%
p-value plot for B_d search window and background only null-hypothesis: eps, png
p-value plot for pseudo experiments with test hypothesis with BR(B_s→μ⁺μ^-)=1.8× 10^-8: eps, png
Central value from Δχ²: eps, png
Central value from Bayesian Posterior (as cross check): eps, png
Central value from Δχ² using only 2 highest NN bins: eps, png
- BR(Bs→ μ⁺μ^-) = 1.4^+1.0_-0.8× 10^-8
- 3.3× 10^-9 < BR(Bs→ μ⁺μ^-) < 3.3× 10^-8 @ 90 C.L.
- 1.2× 10^-9 < BR(Bs→ μ⁺μ^-) < 6.2× 10^-8 @ 95 C.L.
To compare the consistency between the CC and CF channels, we fit each of them separatel
- Δχ² curve using all NN bins for CC: eps, png
  - 6.8× 10^-9 < BR(Bs→ μ⁺μ^-) < 5.4× 10^-8 @ 90 C.L.
- Δχ² curve using all NN bins for CF: eps, png
  - 90% upper bound: 3.3× 10^-8.
Comparison of 95% bounds of LHC experiments and CDF: png

Date: 2011-09-29 17:53:27 CDT

Figures and Tables for Search for the Bs→μ+μ- and Bd→μ+μ- decay

Table of Contents

Data Validation

Baseline Requirements

Neural Network (NN) Signal Discrimination

Neural Network Cross Checks

Combinatoric Background

B→ hh Background

Cross Checks of Background Estimates with Control Samples

Systematics of εNN

Signal Expectations

Unblinding

Old NN Cross Check

Results: Bd Signal Window

Results: Bs Signal Window

Statistical Interpretation of excess in Bs signal window

Figures and Tables for Search for the B_s→μ⁺μ^- and B_d→μ⁺μ^- decay

Systematics of ε_NN

Results: B_d Signal Window

Results: B_s Signal Window

Statistical Interpretation of excess in B_s signal window