Plans for Updating the sin2beta_s Measurement

1
Plans for Updating the sin(2beta_s) Measurement

• Sin(2betas) Group
• B meeting 02/14/08

2
People involved (students in red) Karlsruhe
M. Feindt, M. Kreps, Th. Kuhr, J.
Morlock, A. Schmidt Pittsburgh J.
Boudreau, K. Gibson, C. Liu Madrid
J.P. Fernandez, L. Labarga Hopkins
G. Giurgiu, P. Maksimovic Carnegie-Mellon M.
Paulini MIT G. Gomez
Ceballos Paris student
3
Schedule and Things
to Do
Schedule - Shoot for an updated publication for
summer conferences (use gen6 data, 2.4
fb-1) Things to do - Detailed to-do list
presented by Diego in previous B
meeting (Thursday, January 31, 2008, B Physics
Group Meeting) - Will try to address each
point in the list, identify what is doable by
summer and who will do the work
4
- BstNtuples and/or BltNtuples ? - Will use one
ntuple. Will decide asap which one. - Could use
the second one for cross-checks - Which NN ?
RootSNNS or Neurobayes ? - This comes down to
deciding which ntuple to use. - Will decide
shortly. - Use TTT ? - Do NOT plan to use it
for next analysis - Use JPsi -gt ee ? - No
trigger. Probably not worth. - Can we use
tagging info in signal optimization - We are not
aware of a way of doing it, do not plan to do
it - This was tried previously for the mixing
analysis (Nuno Leonardo) and was found that the
S/sqrt(SB) gave the same results as the
optimization with tagging information
5

• Better figure of merit for sin(2beta) ?
• - Ideally would optimize on toy with
  the error on beta_s as the figure of merit. The
  optimization quantity would be the NN cut
• - For each NN cut, generate a set of
  toys with S and B fractions corresponding to the
  NN cut.
• - Unfortunately the errors are
  asymmetric ? must run MINOS which is very slow
  on a fit with many parameters
• - Will probably try a simplified fit
  version (only a few freely floating parameters)
• - For each NN cut determine the average
  error and select the cut with the smallest error
• - This can be done only after the ntuple
  and the NN is chosen
• - plots

6

• Better figure of merit for sin(2beta) ? (cont)
• - As part of the 1.3 fb-1 analysis we have
  varied the NN cuts to check the
• stability of the result
• - We find that the result is indeed stable,
  but also learn that by releasing
• the NN cut
• - might be able to improve sensitivity to DG
  significantly
• - might be able to improve sensitivity on bs
  marginally
• NN cut 0.3 0.26 (used)
  0.2 0.1

• Of course, the procedure has to be done on toy
  and best cut determined
• from toy

7
- Produce a full CDF simulation Bs -gt J/Psi Phi
sample with correct polarization and CP violation
? - We were not aware of a generator which can
do that - Will try, but need help from Marco
Rescigno who has more information on this -
Extract angular efficiencies from data ? - Could
do that by simply taking the shape of the
combinatorial background - However, this
assumes that the combinatorial background is
absolutely flat in angles - Drawback with data
is the limited statistics in the Bs
sidebands (we need a high statistics 3D
histogram to evaluate angular efficiency) -
Could compare efficiency from data with
efficiency from MC as consistency check -
this was already done and found consistent
results (see next page)
8
- Efficiency from MC
- Background angular distributions shows similar
shapes as efficiency from MC
9
- Establish the effect of angular efficiency on
measurement - Will do it on toy and confirm on
data - Plans for calibrated SSKT beyond a.3
fb-1 ? - dE/dx calibration (see talk by Diego
and Louise) - TOF calibration (J. Morlock -
showed great progress in sub-group meetings) -
Karlsruhe group working on a combined NN OST/SSKT
tagger that will be calibrated by measuring
mixing in Bs -gt Ds pi - can investigate tagging
asymmetries in Bs mixing - will start
presenting progress in subgroup next week -
this is just one bullet in this list but it is a
huge and crucial job ! - work has started
and Karlsruhe students will show first results
at upcoming subgroup meetings - Quote
2beta_s or sin(2beta_s) ? - Will change the
fitter to study advantages from fitting
sin(2beta_s)
10
- Why do we have ct resolution of 106 fs in
Bs-gtJ/Psi Phi compared to 87 fs from the
hadronic mixing analysis Bs-gtDs pi ? - The mean
is of the sideband subtracted sigma(ct)
resolution for a 4-track vertex is 25.05 microns.
This has to be multiplied by thect resolution
scale factor determined from the prompt peak s
1.262 /- 0.020. - We get 31.6 microns or
105.4 fs - The most probable value is at 18.5
microns. This means 23.3 microns after sigma ct
scaling or 77.8 fs. - The number quoted by the
mixing analysis is probably MPV, so in fact
J/Psi Phi seems to have better resolution.
11
- What about direct CP violation ? - Joe
Boudreau is studying the possibility of adding
direct CP violation in the fit - Alternative
strong phases boxing for faster extraction of
confidence regions ? (Gavril) - cos(phi_perp -
phi_para) gt 0, cos(phi_perp) lt 0 -
Re-parameterized the fit in terms of (phi_perp,
phi_perp - phi_para) instead of (phi_perp,
phi_para) and box -pi/2 lt phi_perp - phi_para
lt pi/2 pi/2 lt phi_perp lt
3pi/2 - Currently under study, running
fits - Other ideas ? - Can we use less than 5
sigma to test the nuisance parameters ? -
Giovanni ? - Sensitivity to Dms ? - Will
repeat our previous studies with all data
12
- Can we devise fast/standard procedure for
combining our results with CMK/UTfit groups
? - Have been working with Giovanni on this
issue and came up with a method that can be
used by CKM people (frequentists) - Still need
to come up with a procedure for UTfit (baesyans)
13
- Can we float only 6 parameters instead of 27
? - Have investigated this possibility and it
looks promissing - Giovanni points out that we
would have to run more (than 16) "alternative
Universes ? this will take more time but might
be still worth confidence regions in 1.3 fb-1
likelihood
profile in 2.4 fb-1
continuous all float
continuous 6 float dotted
6 float
dotted - all float
14
Questions that we still have to figure out how to
address
- Do we want to combine with A_sl ? - Of
course, will try that but dont have a procedure
yet - Can we have reliable point estimate on
lifetime, strong phases ? - lots of work, need
manpower - Set up a D0/CDF combination effort
? ?? - Can we try to see which strong phase
is favored by data (a la BaBar) ? ?? -
Systematics that are not nuisance parameters in
fit ? - Is there a way to include such
systematics in a Feldan-Cousins procedure ?