Transcript B2hhFlavourTagging - Indico

Hbh+h’−
Flavour Tagging Strategy
and Performance
Stefano Perazzini
Vincenzo Vagnoni
INFN Bologna
CERN, 19th February 2009
Introduction


In order to make time dependent CP
measurements with B decays to CP eigenstates
we need of course to tag the initial flavour of the
reconstructed B meson candidate
We aim to make CP fits to the whole Hb → h+h’−
sample


For a specific selected B signal candidate we do not want to
distinguish a priori whether it is a B0, a B0s or a Lb
Flavour tagging in LHCb is performed by means of
different algorithms, notably including


opposite side taggers
same side taggers
2
Reminder:
Opposite Side Tagging

The OS taggers determine the initial flavour of the
signal B meson by means of the charge of the
lepton originating from semileptonic decays of the
opposite B…



…and of the kaon from the bcs chain…


OS muon tagger
OS electron tagger
OS kaon tagger
…or finally by tagging the weighted vertex charge
of the opposite B decay vertex

OS vertex charge tagger
3
Reminder:
Same Side Tagging


SS pion (SS kaon) taggers are based on the
charge of the pion (kaon) produced nearby in
phase space in association to the production of a
B0 (B0s) meson during the hadronization phase of
the b-quark pair
While the OS taggers are equally applicable either
if the signal is a B0 or a B0s meson, the SS pion
(kaon) tagger must be specifically applied to
events where the signal is a B0 (B0s)


The correct SS tagger algorithm should be run according to the
signal B hadron under consideration
For each Hbh+h’− event we cannot decide a priori whether to run a
SS pion tagger (for a B0), a SS kaon tagger (for a B0s ), or whether it
makes no practical sense to run any of the two (for a Lb).
4
Combination of tagging decisions

In the present LHCb Flavour Tagging algorithm, the
combination of all tagger decisions which have fired in a
particular event into a unique tagging decision is
performed by estimating a probability for each tagger with
the employment of a neural network


For each tagger, the neural network combines several kinematical
variables of the tagging tracks and gives as output a quantity
discriminating between the two possible tags
An event-by-event tagging probability is then obtained as a
function of the neural network output for each tagger


All the probabilities are finally combined into a single probability that the
meson is a B or an anti-B
This way it is possible to take a unique decision and, furthermore, to
estimate an event-by-event mistag rate
5
Tagging categories


In the present implementation of the LHCb tagging strategy
5 tagging categories are found to be an optimal choice
Ordered with increasing purity, the five categories are
defined according to the following values of the estimated
event-by-event mistag rate ω:





Category 1, if ω > 0.36;
Category 2, if 0.30 < ω < 0.36;
Category 3, if 0.24 < ω < 0.30;
Category 4, if 0.18 < ω < 0.24;
Category 5, if ω < 0.18.
6
B0(s)h+h’− control channels

One nice thing concerning B0(s)h+h’− decays lies in the fact that the
control channels needed to determine from data the mistag rates are
automatically selected by the same selection algorithm



the B0K+p− is the control channel needed to measure the mistag rate to be used
for the study of the time dependent CP asymmetry of the B0 p+p− decay
the B0sp+K− decay is the control channel for the B0sK+K− decay
While the B0K+p− decay has a very rich event yield, the B0s p+K−
decay is characterized by a much smaller statistics, due to the smaller
branching fraction conspirating with the smaller hadronization
probability fs of the B0s meson with respect to the B0 ones


Nevertheless, we expect, as the simulations clearly confirm, that the OS taggers
give identical tagging efficiencies and mistag rates for B0, B0s and indeed also Lb
decays
It would be convenient then to determine the OS mistag rate for all the channels by
exploiting the larger statistics of the B0 K+p− decay, differentiating amongst the
various B hadron species just the mistags for SS pions and SS kaons
7
Possible tagging strategy for Hbh+h’−

An effective and simple way to deal with all these
considerations, consists in running the OS taggers, the SS
pion tagger and the the SS kaon tagger exclusively


e.g. running the OS tagger, then the SS pion tagger just in case the OS
taggers did not give an answer, then the SS kaon tagger just if both the
previous steps did not give any answer, obtaining this way three mutually
exclusive samples
With such approach, we should decide what is the most
convenient order for running the three types of tagging
algorithms
8
Choice of the best strategy


This table summarizes
the effective tagging
efficiencies calculated
for offline selected events
adopting all the 6 possible
tagging orders.
We define as the best order
the one characterized by
the largest total efficiency
calculated as the sum of the
total efficiencies for the four
B0(s) channels


such order is
OSSS kaonSS pion
By this choice, we obtain

for the B0h+p− modes
eeff = (3.64 ± 0.15)%
for the B0sh+K− modes
eeff = (4.42 ± 0.27)%
9
Results with the OSSS kaonSS pion strategy

The tables show eeff, e
and w for each OS
tagger individually, as
well as for 5 OS
tagging categories and
2 SS tagging
categories


The OS categories for all
the channels are
characterized by the same
e and w as expected
The SS taggers instead
give the same e and w for
the two B0h+p− decays,
for the two B0sh+K−
decays
10
Conclusions

In conclusion, the strategy here
described consists in subdividing
the sample of tagged events in 7
categories



5 for the OS tagging
2 for the SS tagging
In the CP fits e and w for the OS
categories will be the same
quantities for all the Hbh+h’−
decay modes, while those of the
SS pion and kaon categories will
be the same for the three pairs of
channels B0h+p−, B0sh+K− and
Lbph−.

This corresponds to 11 tagging
efficiencies and 11 mistag fractions, i.e.
22 parameters in total in the fit
11