Study of decay $B^0(\bar{B^{0}})\to K^+K^-\pi^0$

😀😄🧑‍💻😺

basf2 version: release-06-01-12

TO DO

2023.8

• use git to manager my analysis files
• add flavor tag
• study the correlation between variables

2023.9

• give a brief report on charmless group
• generate more signal MC sample
• toy 2d fit
• fit to data and get branching fraction
• Amplitude analysis

Update logs

Time	Description
2023	The first month to log

Selection criteria

Preliminary selections

The distribution of different variables

• kaon selection
  • kaonID>0.2
• $\pi^0$ selection
  • gamma
    • clusterNHits > 1.5 and 0.2967 < clusterTheta < 2.6180
    • gamma in different range:
      • clusterReg=1: E>0.1 GeV
      • clusterReg=2: E>0.06 GeV
      • clusterReg=3: E>0.1 GeV
    • cluster shape: cluseterE9E21>0.9
  • $\pi^0$
    • the mass window used now: $0.111 \lt M_{\pi_{0}} \lt 0.153 \rm~GeV/c^2$, invariant mass between m_mean +- 3sigma.
    • $|\cos\theta_{helicity}|&lt;0.95$
• $B^0$
  • vertex fit
  • delta E
  • Mbc

Best $B^0$ candidate selection

We used two variables to select the best $B^0$ candidate:

• $\pi^0$ mass constrained chi2
• chiProb of $B^0$ vertex fit chi2

Selection strategy:

• Firstly, we select the $B^0$ candidate whose $\pi^0$ with lowest chi2.
• If there are still multi $B^0$ candidate, we select the $B^0$ candidate with higher chiProb.

Continuum suppression

MC sample: MC 15 run-indepent qqbar sample

We used FBDT to do continuum suppression, which is introduced in the basf2 website.

We used the MVA package in basf2 to mix signal and background sample, and evaluate the result of training.

The cut of continuum suppression output: $ContProb &lt; 0.4$.

To do fit process, transform this variable: $ContProb_{TRAN} = log(\frac{ContProb-0}{0.4-ContProb})$.

Flavor Tag

We also used the method indroduced in basf2 sofeware website

We add a cut on the output of flavor tag: $q \cdot r_{FBDT} &lt; 0.89$

Study of background from generic BBbar

Following the note from belle, we will remove the background from D meson decay, and do the fit.

charm veto

But in the B2Kpipi0 analysis of BARBAR experiment, they do not remove the background from D, and considered this source of background in amplitude fit.

Cut Flow

cut	sigMC efficiency/%	SCF/%	Background number ($1~ab^{-1}$)
fundemental	35.70	8.76	$3.93\times10^6$
continuum suppression	30.97	6.72	$2.93\times10^5$
charm veto	29.93	6.84	$2.81\times10^5$
flavor tag	29.72	6.85	$2.76\times10^5$

Fit to MC sample

correlation between $\Delta E$ and $ContProb_{TRAN}$

Component	Correlation factor
Truth-matched signal	-0.51%
SCF signal	-0.10%
Continuum $q\bar{q}$	0.99%
Generic B sample	-1.66%
peaking from B sample	0.88%

fit model of different sample

Component	$\Delta E$	$ContProb_{TRAN}$
Truth-matched signal	-0.51%
SCF signal	-0.10%
Continuum $q\bar{q}$	0.99%
Generic B sample	-1.66%
peaking from B sample	0.88%