1. Extreme Penguins: Tools requirements at the analysis frontier
Jeffrey Berryhill (UCSB)
for the Radiative Penguin AWG
BaBar Analysis Tools Workshop
September 30, 2005

2. Analysis Tools and “Extreme” Analyses
Tool development does not occur in a vacuum
Necessity is the mother of invention
Development driven by analyses with strongest motivation
to improve standard methods (“extreme analyses”)
because it is a limiting factor/systematic OR
because it is a high priority/high manpower result
Extreme analysts: *the* natural source of manpower for tools groups
If you have extremal needs why are you not part of a tools group?
Complaining is evidence of the desire to volunteer

3. Extreme Muon ID
B →K(*)mm requires tight PID requirements (~2-4% fake rate)
to suppress 50X background from B →K(*)pp AND
Highest efficiency possible to measure BF < 10-6
Close interaction between analysts and PID group to
continuously improve muon selection (neural net based selector muNNTight)
We are very happy
with PID killing!
See J. Hollar’s talk
tomorrow
Next step: take maximal advantage of LSTs, retuned matching chi2

4. Extreme Pion ID
Rare b→ dg penguin decays have large (40X) peaking background from
b→sg decays (Ex: B0 → K*0g mimics B0→ r0g )
Requires efficient pion ID with low K fake rate
Run 4 analysis improved upon
pionLHVeryTight by adding
consistency cut for # of
photons in DIRC
80% eff. and <2% fake rate
Should turn this into a selector
In BetaPID!

5. Extreme Single Photon ID
Inclusive b→ sg BF and photon
spectrum requires precisely known
Single photon ID efficiency
Photon energy resolution
Pi0 veto rate
Not limiting factors yet, but will be with 1 ab-1
Beyond neutrals group studies,
we have studied:
DE in K*g
Energy scale/resolution in VCS sample
VCS energy resolution
Data vs.SP4
Need to support these in neutrals group!

6. Extreme Continuum Suppression
For exclusive and inclusive B→ Xg measurements,
continuum background is the limiting factor
Main tool is feed-forward neural net,
as a cut or as a fit variable
Includes detailed event shape and tagging info
Excellent test case for other multivariate
classifier methods (StatPatternRecognition)
Has lead to dramatic improvements, so let’s
keep working at it!
b→sg
B→ r,w g

7. Extreme Background Subtraction
For inclusive b → s g measurements,
Off resonance sample size is leading statistical error (~8%)
S/B is 1/1 with lepton tag to reduce qq background
More restrictive taggings (BRECO, D*ln) have not yet proven to be competitive
Tagging efficiency needs to be precisely understood
(detailed composition of b→cln)
Off resonance running needs to be sustained at the 10% level for continued improvement

8. Extreme Background Subtraction
For inclusive b → sg measurements,
BB background subtraction precision is leading systematic error (~6%)
Requires precise understanding of inclusive particle spectra in B decays:
B→Xp0, B→Xh, B→Xr, B→Xw, B→Xp, B→Xe at p > 1-3 GeV
Does anyone else care about these?

9. Other Extremes Combinatorics in reconstruction
Exclusive reconstruction of B K+np+g or K+np+ll, n=1-4,
is computationally very expensive; can efficiency be improved?
Photon energy resolution improvement
Will improve S/B in DE for rare decays
EMC edge correction helped!
Modelling of fragmentation of Xs, Xd
Limiting systematic for sum-of-exclusive analysis of b→sg, b→sll
Control sample studies in BRECO samples or charmonium samples
may be helpful

10. Other Extremes Kaon tracking
Precision tracking studies typically look at only pions
Dedicated control sample of kaons would be useful for precisely
constraining tracking efficiency and CP asymmetry systematics
Isospin asymmetry High precision measurements in K*g and sum-of-exclusive b →sg
B0/B+ production ratio already an important systematic; can it be
Improved?

11. Charmless Wish List Incremental improvement in hadron PID
Common supported recipe for hadron PID systematics
Incremental improvement in p0 systematic error

12. Conclusions
The radiative penguin group has a variety of analysis concerns which
are limiting factors for their analysis performance
Optimal background suppression for rare decays
Precision efficiency, resolution and background subtraction in b→sg
Primary concerns: PID, Neutrals (single photon)
Secondary concerns: tagging, vertexing, tracking, BRECO
We should be contributing more manpower to improve the tools