Transcript Slide

Measurement of Tau hadronic branching
ratios in DELPHI experiment at LEP
Dima Dedovich (Dubna)
DELPHI Collaboration
1. Final results on exclusive hadronic branchings
(π/K blind) – submitted to E.Phys.J. C
2. Preliminary results on inclusive single-prong
branching to charged kaons
14 Sept 2004
D.Dedovich
Tau04
1
The DELPHI detector
14 Sept 2004
D.Dedovich
Tau04
2
The first stage (common for both studies):
the tau pair selection
• Almost full LEP-1 statistic was used (1992-1995)
• Analysis was restricted to the barrel region
• Standard LEP-1 tau selection based on kinematic criteria
was used: low multiplicity events with large missing
energy
• Selection efficiency was about 52% (85% within
acceptance) with background 1.5%
• In total, about 80,000 tau pairs were selected
14 Sept 2004
D.Dedovich
Tau04
3
Exclusive hadronic branchings
Track counting
• “Track counting” – event classification into 1- , 3- , and
5-prong tau decays. Method was the same as in the
published paper on topological branchings
• Charged pions from Ks decays were not counted due to
requirement of Vertex Detector measurement on track
• The number of selected tau decay candidates was
• 134421 for 1-prong
• 23847 for 3-prong
• 112
for 5-prong
14 Sept 2004
D.Dedovich
Tau04
4
Exclusive hadronic branchings
charged hadron selection
• 3- and 5- prong decays all are hadronic
• For 1-prong the leptonic decays were rejected using:
dE/dx, EM calorimeter, Hadron calorimeter and muon
chambers
DELPHI
DELPHI
Electron rejection
14 Sept 2004
Muon rejection
D.Dedovich
Tau04
5
Exclusive hadronic branchings
π0 counting
• 4 types of reconstructed π0 were accepted:
–
–
–
–
2 separated photon showers
Photon shower and converted e+e- pair
Single energetic shower (overlapped photons)
Neutral shower + shower wrongly assigned to charged
track
• Neural networks was used to separate π0 and
single photons
• Efficiency to reconstruct π0 was about 70% with
purity of about 90%
14 Sept 2004
D.Dedovich
Tau04
6
Exclusive hadronic branchings
π0 invariant mass
14 Sept 2004
D.Dedovich
Tau04
7
Exclusive hadronic branchings
decay mode identification
• 2 analyses were performed for 1- and 3-prong samples:
one was based on sequential cuts and the other on
neural network approach
• The final results were based on the NN ( trained on
simulation) which provided better precision
• Only sequential cuts was used for 5-prong sample
• The following semi-exclusive decay mode were
identified:
– 1-prong: h±ν ; h± π0 ν ; h±2π0 ν ; h±≥3π0 ν
– 3-prong: 3h ± ν ; 3h± π0 ν; 3h± ≥2π0 ν
– 5-prong: 5h± ν ; 5h±≥1π0 ν
14 Sept 2004
D.Dedovich
Tau04
8
Exclusive hadronic branchings
invariant masses of hadronic systems
14 Sept 2004
D.Dedovich
Tau04
9
Exclusive hadronic branchings
neural network outputs
e
μ
h
hπ0
h2π0
h3π0
3h
14 Sept 2004
3h2π0
3hπ0
D.Dedovich
Tau04
10
Exclusive hadronic branchings
calibration and systematic errors
• Careful checks of data/simulation agreement were
performed using clean test samples selected from real
data : ee→ee; ee→μμ ; ee→eeγ; ee→μμγ ; τ→hπ0ν
• When necessary, corrections were applied on simulation
• Response of calorimeters, track momentum, dE/dx ,
secondary interactions, track and π0 reconstruction
efficiency and muon chamber response were calibrated
• The uncertainties of these calibrations were the main
source of systematic errors
14 Sept 2004
D.Dedovich
Tau04
11
Exclusive hadronic branchings
RESULTS
14 Sept 2004
D.Dedovich
Tau04
12
Inclusive branching to kaons
• DELPHI is the only LEP experiment capable to identify
kaons using not only dE/dx but also with RICH detector
• So far only 1992 results on τ→K±Xν were published.
• Current preliminary results cover full LEP-1 statistics
(1992-1995) and are supposed to replace the old
results
• Only inclusive branching ratio is being presented
14 Sept 2004
D.Dedovich
Tau04
13
Inclusive branching to kaons
hadronic sample selection
• To reduce systematic effects we actually measure the
ratio Br(τ→K±Xν )/Br(τ→ π±Xν ). Many biases are
canceled as kaons and pions are both hadrons
• As a first stage, a sample of 1-prong hadronic tau
decays was selected using calorimeters and muon
chambers.
• The efficiency of the hadronic selection was about 89%,
the background was about 0.3% from non-tau events,
and 3.7% from leptonic and multiprong tau decays
14 Sept 2004
D.Dedovich
Tau04
14
Inclusive branching to kaons
Kaon identification
• At LEP1 kaons from tau decays are allowed to have
momentum in the range 3.6-45 GeV/c
• Measurements of dE/dx in TPC provide π/K separation in
the full kinematic range at the level of 1.6-2.2 σ
• For momenta below 8.5 GeV/c kaons are also identified
by VETO in DELPHI RICH detector
• For momenta between 8.5 and about 25 GeV/c
identification is based on Cherenkov angle measurement
in RICH (Ring measurements)
14 Sept 2004
D.Dedovich
Tau04
15
Inclusive branching to kaons
Kaon identification
π
π
K
K
14 Sept 2004
D.Dedovich
Tau04
16
Inclusive branching to kaons
Pull variables
The K identification was based on pull variables ΠH for
hypothesis H=π/K/e/μ
For Cherenkov angle measurements a similar
variables ΠRING was constructed
14 Sept 2004
D.Dedovich
Tau04
17
Inclusive branching to kaons
dE/dx calibration
• dE/dx pull position and width were carefully calibrated as
a function of particle velocity and direction using test
sample of pions, muons and kaons selected from real
data using RICH.
• Small discrepancy was found between pions and muons
of same velocity. Therefore for final calibration clean
pions sample was used.
• dE/dX of kaons and pions of same velocity was found in
agreement, and the uncertainty of this comparison (2.4%
of pull width) was assigned to systematic error
14 Sept 2004
D.Dedovich
Tau04
18
Inclusive branching to kaons
Clean sample of pions (kaons suppressed by RICH)
14 Sept 2004
D.Dedovich
Tau04
19
Inclusive branching to kaons
Kaon-enriched sample
dE/dx kaon pull
14 Sept 2004
D.Dedovich
Tau04
20
Inclusive branching to kaons
All hadronic tau decay candidates
14 Sept 2004
D.Dedovich
Tau04
21
Inclusive branching to kaons
Ring pull calibration
• Unlike the case of dE/dx, the ring pull has significant
non-Gaussian tails. Therefore the following calibration
procedure was adopted :
• Small corrections (few % of pull width) depending on
velocity were applied to simulation to get agreement
with the real data (clean pion samples selected using
dE/dx)
• The pull distribution shapes obtained for simulation were
used as probability density function in further fits
• The far parts of tails were combined into 2 single bins to
avoid problems with shape description
14 Sept 2004
D.Dedovich
Tau04
22
Inclusive branching to kaons
Clean sample of pions (kaons suppressed by dE/dx)
14 Sept 2004
D.Dedovich
Tau04
23
Inclusive branching to kaons
Kaon-enriched sample
14 Sept 2004
D.Dedovich
Tau04
24
Inclusive branching to kaons
All hadronic tau decay candidates
14 Sept 2004
D.Dedovich
Tau04
25
Inclusive branching to kaons
VETO identification
• The main source of
systematic is the
rate of false VETO
identifications
• The data/simulation
agreement was
checked using clean
samples of muons
and pions
14 Sept 2004
D.Dedovich
Tau04
26
Inclusive branching to kaons
The fit procedure
• The measured pulls were used to construct the
probability W that the particle is a kaon: W=FK/(Fπ+FK)
• Here FK(ΠK) and Fπ(Ππ) are the probability density
functions for a given hypothesis
• Gaussian PDF was used for dE/dx and the shapes
predicted by simulation in the case of RICH
• Distribution of W in real data was fitted by a linear
combination of simulated pions and kaons
• The results of dE/dX and RICH were fitted either
separately or combined into a single probability W
14 Sept 2004
D.Dedovich
Tau04
27
Inclusive branching to kaons
fit to dE/dx probability
14 Sept 2004
D.Dedovich
Tau04
28
Inclusive branching to kaons
fit to Ring probability
14 Sept 2004
D.Dedovich
Tau04
29
Inclusive branching to kaons
combined fit : Ring+dE/dx
14 Sept 2004
D.Dedovich
Tau04
30
Inclusive branching to kaons
Systematic errors
• The main source of systematic errors is the uncertainties
in calibration of pull position and width. Even small bias
results in large error in estimation of pion background
• However this error reduced dramatically if RICH and
dE/dx are used in combination
• Therefore our results were obtained using combined
measurement when possible (RICH was not always
operational)
• Individual measurements were used for a cross-check
14 Sept 2004
D.Dedovich
Tau04
31
Inclusive branching to kaons
Systematic errors
The uncertainty of residual pion background (colored)
Is strongly redused if pions were already suppresed
by another detector
14 Sept 2004
D.Dedovich
Tau04
32
Inclusive branching to kaons
Systematic errors in %
RING pull
Position
width
dE/dx pull
Moment.
depend.
dE/dx
Ring
4.6
3.1
VETO
Position
width
Moment.
depend.
K/π MIP
agreem.
3.3
5.3
2.0
2.4
4.2
veto
8.4
dE/dx+VETO
dE/dx+Ring
False
ID
1.5
0.7
1.1
0.3
0.7
0.6
0.6
0.7
1.3
0.1
1.0
2.4
Other sources of systematic errors are MC statistics (1.2%)
and tau decay branchings (1.9%)
14 Sept 2004
D.Dedovich
Tau04
33
Inclusive branching to kaons
The results (in %)
χ2 = 3.26/3
χ2 = 1.99/2
14 Sept 2004
D.Dedovich
Tau04
34
Inclusive branching to kaons
Results of Individual measurements in %
Ring
1.745 ± 0.170 (0.126 stat 0.115 syst)
dE/dx
1.455 ± 0.131 (0.068 stat 0.105 syst)
VETO
1.685 ± 0.272 (0.231 stat 0.144 syst)
Total
1.579 ± 0.097
14 Sept 2004
D.Dedovich
Tau04
35
Inclusive branching to kaons
Results of combined measurements in %
Ring+dE/dx
1.639 ± 0.112
VETO+dE/dx
1.594 ± 0.184 (0.172 stat 0.066 syst)
dE/dx only
1.346 ± 0.139 (0.082 stat 0.106 syst)
Ring only
1.871 ± 0.489 (0.462 stat 0.165 syst)
Total
1.545 ± 0.078
14 Sept 2004
D.Dedovich
(0.097 stat 0.054 syst)
Tau04
36
Summary
• We have measured tau semi-exclusive hadronic
branching ratios. Some of them are at the level of
world best.
• We also presented preliminary result for inclusive tau
to kaons branching : 1.545±0.078%
14 Sept 2004
D.Dedovich
Tau04
37