1. Rare Tau Decay at Belle – Search for Lepton Flavor Violation –
Takayoshi Ohshima Nagoya University Belle Collaboration
New data on tmg & tmh
EPS2003, Aachen
We, Belle collaboration, search for LFVing tau-decay at KEKB-factory experiment.
I would like to present the latest results on t->mg and mh decays.
We, Belle collaboration, search for LFVing tau-decay at KEKB-factory experiment. I would like to present the latest results on t->mg and mh.

2. KEKB & Belle spectrometer
KEKB asymmetric e+e– collider
- e+ / e– : 3.5 / 8 GeV
- CM energy: GeV
- Design luminosity: 1034/cm2/s
KEKB is an asymmetric electron-positron collider in Japan which attained the world highest peak luminosity of 1× 10^34 this May.
We now accumulate about 160/fb data, corrsponding to about 140 M tau-pairs produced at an energy of 10.6 GeV, by this general purpose Belle detector.
KEKB is an asymmetric electron-positron collider in Japan which attained
the world highest peak luminosity of 1× 10^34 this May.
We now accumulate about 160/fb data, corrsponding to about 140 M tau-pairs
produced at an energy of 10.6 GeV, by this general purpose Belle detector.
EPS03, Aachen; T. Ohshima, Belle Collaboration

3. Physics of tmg Forbidden in SM, while New physics allows LFV decay.
SUSY predicts LFV ( , e, …, e )
Enhancement due to high mass 
Br() ~ 105-6Br(e)
Let me report first on t->mg, which is forbidden in SM but is allowed in new physics beyond the SM. Some SUSY models predict rather a large branching fraction accessible at Belle.
The best limit is so far achieved by CLEO and Belle as 1×10^-6 in the branching fraction.
This LFVing decay is forbidden in SM but is allowed
in new physics beyond the SM. Some SUSY models
predict rather a large branching fraction accessible
at Belle. The best limit is so far achieved by CLEO and
Belle as 1×10^-6 in the branching fraction.
EPS03, Aachen; T. Ohshima, Belle Collaboration

4. Event selection (mg)+(m + ng >0 + neutrino(s))
Search for
(mg)+(m + ng >0 + neutrino(s))
86.3/fb data analyzed (78.5 M tt)
From the previous studies, we know that
gtt and gmm form the prominent BG;
(2) Non-zero candidate events are found to exist in signal region.
Therefore,
BG reduction & knowledge of its distribution
are essential to extract the number of signal events.
In order to remove BG, we newly introduce a cut, pmissing-Mmissing cut
We look for a tau-pair event using 86.3/fb of data. One t decays to m and g, and the other t decays to a charged particle, but not m, and any number of photons with neutrino(s). => TO PAGE 5
From the previous studies, we know that radiative tau-pair and radiative mu-pair form the prominent BGs and BG is expressed with this formula. Lambda is a small uds continuum and kappa is m-ID inefficiency.
And, non-zero candidate events are found to exist in a signal region. So, reduction of BG and knowledge of its detail distribution are quite essential to extrcat the number of signal events. In order to remove BG, we newly introduce a cut on missing momentum vs missing mass-squared. => TO PAGE 6.
EPS03, Aachen; T. Ohshima, Belle Collaboration

5. Event selection (dominant BG)
Here list the selections and some kinematics. But I have to skip this due to limited time.
Figures show, for instance, opening angle distributions between m-g, and between a tagged-track and a missing track.
Signal is indicated by yellow, and the dominant background of general tau-pair decay by open histogram. Data are by dots. Selected region is indicated by arrows. => RETURN TO PAGE 4.
EPS03, Aachen; T. Ohshima, Belle Collaboration

6. pmissing vs. Mmissing cut & Blind analysis
98% tt and 86% mm removed ,
76% signals survived.
Signal MC is indicated by yellow.
Signal yield is evaluated in DE-vs-Mmg
(DE=EmgCM-EbeamCM)
In order to avoid bias on analysis, we
Blind the signal region
1.70 GeV < Mmg < 1.85 GeV
As indicated by a red lines for general tau-pair and signal MCs.
This cut moves 98% of tau-pair and 86% of mu-pair, but 76% of signals survives.
Signal yield is evaluated in DE-Mmg plane, where DE is the difference of mg energy from the beam energy in the center of mass system. Signal MC is indicated here by yellow.
In order to avoid introducing bias on analysis, we blind the signal region between 1.7 and 1.85 GeV in mg mass, as indicated by soft purple.
EPS03, Aachen; T. Ohshima, Belle Collaboration

7. BG in the signal region BG comprises (1)ttg and
(2) mmg (one m misidentified as m) and (3) small cont.
For BG, ttg and cont. are obtained by MC,
For mmg, mmg from data and multiplied by m-ID inefficiency k.
BG probability density (Si) is expressed by Gaussian and Landau functions.
For BG, ttg and continuum are obtained by MC.
For not-mmg, we get mmg from data and multiplied m-ID inefficiency.
Thus obtained BG spectrum at the blinded region is shown here by the curves.
BG can be also obtained from actual data by averaging their distributions at both side-bands, as indicated by the histogram.
Curve and histogram agree very well.
Finally, we open the blinded area. => QUICKLY GOTO NEXT SLID & RETURN HERE.
Dots are the remaining data. It well agree with the expected BGs.
Yellow shows the expected signal distribution.
Thus obtained B spectrum at the blinded region is shown here by the curves. BG can be also obtained from actual data by averaging
their distributions at both side-bands, as indicated by the histogram. Curve and histogram agree very well. Finally, we open the blinded area.
Dots are the remaining data. It well agree with the expected BGs. Yellow shows the expected signal distribution.
EPS03, Aachen; T. Ohshima, Belle Collaboration

8. DE vs. Mmg 5s region (e = 11.0%) Resolution DE： 65.40.6 MeV
Event distribution where the blinded region is unveiled.
(1)
Due to initial radiation and energy leakage of photon calorimeter, the distribution has a long tail.
Resolution
DE： 65.40.6 MeV
Mmg：20.30.9 MeV/c2
The signal MC is indicated by yellow.
Due to initial radiation and energy leakage of photon calorimeter, the distribution has a long tail. The resolutions are evaluated as listed here.
In order to evaluate the number of signal events, we take 5s region indicated here. It provides 10.3 % of detection efficiency.
(2)
5s region (e = 11.0%)
(3)
In order to evaluate the number of signal events, we take 5s region indicated here,
which provides 10.3 % of detection efficiency.
EPS03, Aachen; T. Ohshima, Belle Collaboration

9. Unbinned EML fit S. Ahmed et. al., PR D61 071101 (200)
An unbinned extended maximum likelihood fit is performed: Likelihood is defined as this.
Result of fit for 54 events gives us 0 signal events.
Upper limit on the number of signal events at 90% CL is obtained as 5.1 events with use of a Toy MC.
Here shows the best fit result.
Probability densities for a sum of BG and signal are displayed by dark and bright pattern, and the data by dots. Right side shows for 5.1 signal events.
From these figures you can see, the events observed are much more characteristic of BG than of signal.
Now, we get a limit on branching fraction.
Detection efficiency is 10.3% and the number of tau-pairs is 78.5 Ms.
BR is 3.0×10^-7.
Probability densities for a sum of BG and signal are displayed by dark and bright pattern, and the data by dots.
From these figures it can be seen, the events observed are much more characteristic of BG than of signal.
EPS03, Aachen; T. Ohshima, Belle Collaboration

10. Br(tmg) Systematic uncertainty on 2eNtt Br(tmg) < 3.2 x 10 -7
Track rec. eff. : 2.0%
Photon rec. eff. : 2.8%
Cut selection : 2.2%
Luminosity : %
Trigger efficiency : 1.6%
MC statistics : 0.8%
Total : 4.7%
(LFV interaction structure & spin correlation: < 0.1%)
Systematic uncertainty
on s0
Continuum & k : +0.06/-0.11 ev.
BG function: 0.3 ev.
Fit region: ev.
Br(tmg) < 3.2 x 10 -7
at 90% CL.
Here lists the systematic uncertainties relating to s0 and detector sensitivity.
Including all of these uncertainties, we finally get the branching fraction of 3.2×10^-7 at 90% CL.
EPS03, Aachen; T. Ohshima, Belle Collaboration

11. Physics of LFV tmh
The constrained MSSM Higgs-mediated model.
An attractive process to give the most strigent bound on Higgs-mediated LFV in MSSM.
Especially, large tanb would provide large Br.
　　 M. Sher, PR D (2002); K.S. Bubu and C. Kolda, PRL 89, (2002);
A. Dedes, J. Ellis, and M. Raidal, PL B549, 159 (2002)
Br(t  mh) is 8.4 times larger than Br(t  mmm)
color facor  3
Higgs coupling  (ms/mm)2
Current upper limit from CLEO　（Ldt = 4.7fb-1）
Using h  gg mode.
Br(tmh) < 9.62×10-6 ; Br(teh) < 8.19×10-6 　
Next is, tau decay to mh, which is an attractive process to give the most strigent bound on Higgs-mediated LFV in MSSM.
Current limit is 1×10^-5 by CLEO.
EPS03, Aachen; T. Ohshima, Belle Collaboration

12. Event selection Essentially, very similar to tmg hgg (Br = 39.4%):
Two h decay-modes
hgg (Br = 39.4%):
2 tracks + ng > 2 + missing
hp+p-p0 (Br = 22.6%):
4 tracks + ng > 2 + missing (m is not required)
84.3/fb data used
h gg mass
h p+p-p0 mass
s12 MeV
s5 MeV
The analysis is essentially very similar to tmg.
We analyze 84/bf of data by detecting h in gg and 3p modes.
Here shows a resolution normalized h-mass in gg mode, and p0 and h-mass for 3p modes.
a resolution normalized h-mass in gg mode, and p0 and h-mass for 3p modes.
EPS03, Aachen; T. Ohshima, Belle Collaboration

13. Backgrounds & Resolution
hgg h3p
From MC
Resolutions:
DE: 2.6 MeV,  2.0 MeV
Mmh:  0.6 MeV/c2,  0.3 MeV/c2
Signal-ellipse: 90% acceptance
BG: from MC mostly tt, and uds cont. and mm.
From data
Remaining events:
(after kinematical cuts) events events
(within 10s but out of ellipse) 7events events
(MC=3.72.4) (MC=0)
Open blind
(within region) events events
(MC=0.9) (MC=0)
Blind analysis is performed as same as the mg case,
and the signal region is defined, this time,
by an ellipse, as shown here,
which gives a 90% acceptance.
DE-Mmh plot
Here lists resolutions for two decay-modes.
Blind analysis is performed as same as the mg case, and the signal region is defined, this time, by an ellipse, as shown here, which gives a 90% acceptance.
After all kinematical cuts, 18 events for gg and 60 events for 3p are found.
Within ±10s region but outside the ellipse, they are 7 and 2 events while MC predicts 3.7 and 0 events.
When we open the blinded region, no events are found for both modes.
EPS03, Aachen; T. Ohshima, Belle Collaboration

14. Number of signals & Br evaluation
hp+p-p0
hgg
s0: upper limit of signal events = 2.3
These are the events distributions in DE vs mhMass plane.
Dots are the data, open circle is tt MC events and square
is continuum. The ellipses are the signal region with
an acceptance of 90%.
Branching fraction
hgg p gg+3p
e  Brh % % %
Ntt  106
Br (10-7) < < < 3.4
These are the events distributions in DE vs mhMass plane.
Dots are the data, open circle is tt MC events and square is continuum.
The ellipses are the signal region with an acceptance of 90%.
Since no event is observed, a 90% CL upper limit of signal events is set as 2.3 events.
Here list the parameters for detection sensitivity (efficiency and number of tt-pairs produced).
We obtaine 90% CL upper limit in Br as 4.5×10-7 for gg mode, 13.6×10^-7 for 3p mode, and combining two modes 3.4×10^-7.
EPS03, Aachen; T. Ohshima, Belle Collaboration

15. Br(tmh) at 90% CL. Br(tmh) < 3.4 10-7
Systematic uncertainties(%)
hgg h3p
Luminosity
Brh
Beam BG
Trigger eff
Tracking eff
p0 veto
h/p0 recon. eff
m ID eff
MC stat
Sum
Br(tmh) < 3.4 10-7
at 90% CL.
Include systematic uncertainty into an upper limit at 90% CL.
S: detection sensitivity, b: BG
R.Cousins and V.Highland, NIM A320, 331 (1992)
Here lists the systematic uncertainties. It amounts to 8 and 7%.
Including these systematic errors into the Br, we obtaine 3.4×10^-7.
EPS03, Aachen; T. Ohshima, Belle Collaboration

16. Summary Br(tmg) < 3.2  10-7 & Br(tmh) < 3.4  10-7
We attain upper limits of
Br(tmg) < 3.2  & Br(tmh) < 3.4  10-7
at 90% CL using 85/fb data.
2. For the first data sensitivity reaches 10-7 level.
3. They provide some constraint on physics beyond the SM.
Dedes, J. Ellis, and M. Raidal, PL B549, 159 (2002)
Let me summarize my talk.
We attain upper limits on tmg as 3.2×10^-7 and on tmh as 3.4×10^-7 at 90% CL.
These are the first data entering to sensitivity of 10^-7.
They provide some constraints on the physics beyond the SM.
For instance, here shows an excluded region by this tmg data on the relation between tanb and SUSY particle mass following the paper by Dedes et. al., and this shows also an excluded region by this tmh data on the relation between tanb and Higgs mass following the paper by Babu and Kolda.
We have an additional available data of 75/fb so that these sensitivities will be soon improved.
K.S. Bubu and C. Kolda, PRL 89, (2002)
4. Additional available data of 75/fb should improve these sensitivities soon.
EPS03, Aachen; T. Ohshima, Belle Collaboration