c LFV in Tau experiments K Hayasaka Niigata

c. LFV in Tau experiments K. Hayasaka (Niigata) 5/Jun/2019 International School on c. LFV 1

tgℓg Simplest, most promising… 5/Jun/2019 International School on c. LFV 2

Signature of signal and background m signal g t e mm event t Only tag side has neutrino! t t m e m Both sides have no neutrino total energy is equal to beam energy. qq event (q=u, d, s, c) g ISR g m e g ISR g e e SM tt event fake n n e e q e Both sides have neutrinos missing helps us to reject this kind of BG’s. 5/Jun/2019 … q International School on c. LFV … Both sides have many tracks and photons. No. of photons 3 decreases this kind of BG’s.

Concepts for signal selection criteria (tgmg) • Require 1 -1 prong events (2 charged tracks well separated) • Require muon by PID and photon close to muon (using opening angle) • Veto muon for another charged track to reject mm events • Require zero missing mass and large missing momentum, i. e. , only one neutrino is expected. 5/Jun/2019 International School on c. LFV 4

Previous result for tgmg search • 545 fb-1 Belle data sample • Detection eff. : 6. 1% in 5 s region • Obs: XX events • Exp: (88. 4± 7. 4) events • Main BG: tgmnn + ISR g eegmm+ ISR g Upper Limits by 2 d UEML fit on M-DE plane • Expected UL: 7. 8 x 10 -8 @90%CL • Obtained UL: 4. 5 x 10 -8 @90%CL 5/Jun/2019 UEML=Unbinned Extended Maximum Likelihood fit International School on c. LFV 5

Previous result for tgmg search • 545 fb-1 Belle data sample • Detection eff. : 6. 1% in 5 s region • Obs: XX events • Exp: (88. 4± 7. 4) events • Main BG: tgmnn + ISR g eegmm+ ISR g Upper Limits by 2 d UEML fit on M-DE plane • Expected UL: 7. 8 x 10 -8 @90%CL • Obtained UL: 4. 5 x 10 -8 @90%CL 5/Jun/2019 UEML=Unbinned Extended Maximum Likelihood fit International School on c. LFV 6

Previous result for tgmg search • 545 fb-1 Belle data sample • Detection eff. : 6. 1% in 5 s region • Obs: XX events • Exp: (88. 4± 7. 4) events • Main BG: tgmnn + ISR g eegmm+ ISR g Upper Limits by 2 d UEML fit on M-DE plane • Expected UL: 7. 8 x 10 -8 @90%CL • Obtained UL: 4. 5 x 10 -8 @90%CL 5/Jun/2019 UEML=Unbinned Extended Maximum Likelihood fit International School on c. LFV 7

Previous result for tgmg search • 545 fb-1 Belle data sample • Detection eff. : 6. 1% in 5 s region • Obs: 94 events • Exp: (88. 4± 7. 4) events • Main BG: tgmnn + ISR g eegmm+ ISR g Upper Limits by 2 d UEML fit on M-DE plane • Expected UL: 7. 8 x 10 -8 @90%CL • Obtained UL: 4. 5 x 10 -8 @90%CL 5/Jun/2019 UEML=Unbinned Extended Maximum Likelihood fit International School on c. LFV 8

Belle II expectation • 545 fb-1 Belle data sample • Detection eff. : 6. 1% in 5 s region • Obs: 94 events Finally, 50 ab-1 data sample is expected at Belle II. 100 x larger data sample BG: 9400? How to reduce? Br(tgmg)~10 -9 Br(tgmnn)~17% To reject accidental BG from tgmnn, how powerful rejection do we need? 5/Jun/2019 International School on c. LFV 9

Tag side missing mass (tag-side missing mass)2 Hadronic tags m t t g This can be reconstructed from another t leptonic tags (tag side missing)= (reconstructed tag-side t) -(visible tracks and gammas belonging to tag side) 5/Jun/2019 International School on c. LFV 10

mmg BG direction CM frame m e g e m • In the CM frame, m-m opening angle is not back-to-back due to g in mmg event. Cosine of the angle between m+ and m- at CM-g frame cos. Qmm. CM-g>-0. 99 CM-g frame = (Pbeam++Pbeam--Pg=0) frame m g e But, in the CM-g frame, m-m opening angle should be 180 o. 5/Jun/2019 This selection affects only for mm BG. International School on c. LFV 11

Correlation? • To get more sensitivity, one way is to use better calorimeter. Narrow region low BG 5/Jun/2019 International School on c. LFV 12

Comments on beam constrained mass We can use ~ t mass instead of since they always agree thanks to energy conservation law. But, signal region does not get small on 2 D plane. 5/Jun/2019 International School on c. LFV DE/√s~0 Signal MC distribution 13

Two tails for signal distribution Two kinds of tails appear for signal distribution. g res. ISR Signal MC distribution 5/Jun/2019 International School on c. LFV 14

Homework/discussion • LHC searches lepton flavor violating higgs: H tm Sometimes, someone says: Since LHC gives much more stringent limit for t. LFV, no more B-factory search for, at least, tgmg, is necessary. But this is invalid statement. Why? 5/Jun/2019 International School on c. LFV 15

tgℓℓℓ easiest, most clean… 5/Jun/2019 International School on c. LFV 16

Signals • t-gm-m+m • t-ge-m+m • t-gm-e+e • t-ge-e+e • t-gm-e+m • t-ge-m+e 5/Jun/2019 Usual tau LFV decay More complicated tau LFV decay International School on c. LFV 17

Signal and Background tt p- p+ 2 photon process p- f=leptons, quarks n signal n l. Neutrinos in both side l. Missing energy in signal side _ radiative Bhabha/mm process qq e+ l. Neutrino(s) in tag side l. Particle ID lmany tracks 5/Jun/2019 International School on c. LFV eg e/me- e+ e+/m+ 18

SM t decay BG • It seems tgpppn can be BG. • Rate for p is misidentified as m is ~1%. For 3 p, 10 -6. • Requirement for missing has 1/100 rejection power. Br(tgpppn)=0. 1 So effective BR~10 -9. In addition, due to n, most of all events cannot reach signal region, Mmmm<<mt, DE<<0. 5/Jun/2019 International School on c. LFV 19

Selection • 3 -1 prong events are selected. • Basically, 3 times lepton-ID is so strong to reject BGs. • For t-ge-m+m-, m-e+e-, e-e+e-, 2 photon processes, e+ e-m+ m-, e+ e- and radiative Bhabha/mm with g conversion can be background which cannot be avoided by Lepton-ID. So, lepton veto is required for remaining charged track. (tag side) 5/Jun/2019 International School on c. LFV 20

g-conversion veto • Some BG comes from events with g-conversion. We require Mee>0. 2 Ge. V 5/Jun/2019 International School on c. LFV 21

Search for t 3 leptons • Data: 782 fb-1 • No event is found in the signal region. e-e+e- m-m+m- e-m+m- m-e+e- m-e+m- e-m+e- • Almost BG free • Expected # of BG: 0. 01 -0. 21 • Because of good lepton ID • Br<(1. 5 -2. 7)x 10 -8 at 90%CL. most sensitive results Phys. Lett. B 687, 139 (2010) 5/Jun/2019 Mode e (%) e-e+e 6. 0 7. 6 m-m+me-m+m 6. 1 m-e+e 9. 3 m-e+m 10. 1 International School on c. LFV e-m+e 11. 5 NBGEXP 0. 21± 0. 15 0. 13± 0. 06 0. 10± 0. 04± 0. 04 0. 02± 0. 02 0. 01± 0. 01 ssyst (%) 9. 8 7. 4 9. 5 7. 8 7. 6 7. 7 UL (x 10 -8) 2. 7 2. 1 2. 7 1. 8 1. 7 22 1. 5

Comparison of size for signal region 5/Jun/2019 International School on c. LFV 23

Comparison of size for signal region Signal region for mmm is smaller than that for mg. When t is reconstructed from only charged tracks, signal region gets smaller than that reconstructed from charged track(s) and photon(s). Since in general, BG amount is proportional to the area size of signal region, t LFV reconstructed from only charged tracks is expected to be rather clean analysis. 5/Jun/2019 International School on c. LFV 24

t production in LHCb • Mainly t is produced from Ds decay: Ds tn • Referring to Ds fp (and then fgmm), number of Ds tn can be counted. • LHCb s(Ds) ~ 200 mb at 7 Te. V 10 -year later, Belle II will • Br(Ds tn)~5% collect 4. 5 x 10 3 fb-1 x 200 mb x 5% =3 x 1010 t! 10 5/Jun/2019 International School on c. LFV 25

tgmmm search on LHCb • Pros: Huge number of t is available. • Cons: Since of single t production, it is difficult to tag t event. Thus, large BG is expected. M(DS)~2 Ge. V 5/Jun/2019 International School on c. LFV 26

tgℓP 0 (P 0=p 0, h, h’) 5/Jun/2019 International School on c. LFV 27

Signature of signal and background m signal t e t g h g Only tag side has neutrino! t t g ISR g m e m Both sides have no neutrino total energy is equal to beam energy. qq event (q=u, d, s, c) g ISR g m e fake e e SM tt event fake mm event n n e … q e Both sides have neutrinos missing helps us 5/Jun/2019 to reject this kind of International BG’s. School on c. LFV … Both sides have many tracks and photons. No. of photons 28 decreases this kind of BG’s.

To search for tgmh… g m t e t h(p 0)e t t 5/Jun/2019 Br(hgppp 0)~20% Br(hggg)~40% m g t e t Charged tracks：２ Photon： 2 m e Br(h’gpph)~45% r 0 p h’ e p g p p g h(h’)p 0(h) g e Charged track：４ Photon： 2 Br(h’gr 0 g)~30% Charged track：４ Photon：１ Different analysis Different background (Of course, common background is also found. ) International School on c. LFV 29

Distribution for variables used for selection. （tgmh(ggg)） ＋: experimental data ■：Signal simulation ―：BG simlation 5/Jun/2019 International School on c. LFV 30

Result for t→ℓP 0(=p 0, h, h’) Belle preliminary tg Eff. NBGexp mh(ggg) 8. 2% 0. 63± 0. 37 3. 6 mh(gppp 0) 6. 9% 0. 23± 0. 23 mh(comb. ) UL -8) tg UL(x 10 Eff. NBGexp x 10 -8 -8) UL(x 10 mh’(gpph) 8. 1% 0. 00+0. 16 -0. 00 10. 0 8. 6 mh’ (gr 0 g) 6. 2% 0. 59± 0. 41 6. 6 2. 3 mh’ (comb. ) 3. 8 eh(ggg) 7. 0% 0. 66± 0. 38 8. 2 eh’ (gpph) 7. 3% 0. 63± 0. 45 9. 4 eh(gppp 0) 6. 3% 0. 69± 0. 40 8. 1 eh’ (gr 0 g) 7. 5% 0. 29± 0. 29 6. 8 4. 4 eh’ (comb. ) 2. 7 ep 0(ggg) eh(comb. ) mp 0(ggg) 4. 2% 5/Jun/2019 0. 64± 0. 32 International School on c. LFV 3. 6 4. 7% 31 0. 89± 0. 40 2. 2

SM t decay BG for mh, eh • tgphn cannot be dominant BG with mis. ID of p as m. tgphn is strongly suppressed due to Isospin symmetry, called “second class current” by Weinberg. (not yet observed, expected 10 -6 Branching fraction. ) • tgpp 0 p 0 n is main BG. g from p 0 ggg and g from another p 0 construct fake h. Remaining gs are not detected. • For electron mode, since p mis. IDed as e is heavier than e, BG can reach signal region. For m mode, probability to reach signal region is much lower. 5/Jun/2019 International School on c. LFV 32

SM t decay BG for mp 0 • tgpp 0 n is dominant BG with mis. ID of p as m. But, they cannot reach signal region. On the other hand, in electron case, they reach around the signal region. tgep 0 tgmp 0 5/Jun/2019 International School on c. LFV 33

Analysis difficulty • Need to understand background to construct good selection to reduce BG as much as possible Simple • t ℓℓℓ • t ℓKs, Lh • t ℓV 0( hh’) Difficulty of reducing the background • t ℓP 0( gg) • t ℓhh’ • t ℓg Hard • Analyze the modes from simple selection to hard for background reduction • Provide feedback to next analysis of similar final state 5/Jun/2019 International School on c. LFV 34

Search for ℓhh’ • Update with 854 fb-1 data • 14 modes are investigated (h, h’=p± and K±) • t-→ℓ-h+h’-: 8 modes (lepton flavor violation) • t-→ℓ+h-h’-: 6 modes (lepton number violation) e- or m- p- signal generic tt BG si sid gna e l K+ ta sid g e e- n t- e+ n t- e- t+ miss-ID as K+ p p p+ n t- e+ e- t+ n miss-ID as K+ p - e+ e- g p 0 g e+ t+ n Also, qq events can be BG 5/Jun/2019 International School of on c. LFV Missing momentum can help to reject this kind BGs since signal has n only on tag side. 35

To reject g-conversion • To reflect fake signal ep. K events from pp 0(ggg)n with g-conversion, we evaluate Mee, where electron mass is forcibly assigned for K candidate. g-conversion 5/Jun/2019 International School on c. LFV 36

Result for ℓhh’ In the signal region 1 event : in m+p-p- and m-p+Kno events: in other modes ⇒ no significant excess/Expected # of BG: 0. 06 -0. 72 予想される背景事象 (0. 1 -1. 5)events Set upper limits at 90%CL: Br(t ℓhh’)< (2. 0 -8. 6)x 10 -8 most sensitive results 5/Jun/2019 (preliminary) International School on c. LFV 37

Tau LFV at Belle II This is obtained from simple extrapolation by normalizing luminosity. As for rich BG analysis, we need modified analysis to obtain similar sensitivity to those shown here. 5/Jun/2019 International School on c. LFV 38

Why eg UL is worse than mg? • UL evaluation is, in some meaning, “gambling”. No one knows where accidental BG appears. If this BG events moves slightly away from signal dominant region or did not appear, UL drastically goes down. Just I say… Unfortunately… 5/Jun/2019 International School on c. LFV 39

Other kinds of t. LFV decay 5/Jun/2019 International School on c. LFV 40

Lm-Lt? • tgμ+(neutral long life boson) Neutral: undetectable tracker Long life: not-produce particle in detector. Experiment signal: m + missing If new boson’s mass is known, since this is 2 -body decay, m momentum is monochromatic in t rest frame. These days, it is discussed that such a boson can appear in Lm-Lt model without being consistent with other experimental result. 5/Jun/2019 International School on c. LFV 41

Experimental method • Already established by ARGUS m ４７２ｐb-1 For tag side t, 3 pn decay is required. Unfortunately, we cannot take t-rest frame. Instead, we reconstruct pseudo-t p where energy and size of momentum are decided from energy conservation low and the direction of momentum is taken from 3 p system. 5/Jun/2019 International School on c. LFV a t t p p n 42

Why 3 pn decay is required? • Since π mass is 140 Me. V, 3 xπ mass is around 420 Me. V. Much lighter than mt. • But, it is considered that τ→３πν dominantly comes fromτ→a 1ν. • Since a 1 mass is around 1. 2 Ge. V, 70% of mt goes to a 1. It is expected that a 1 direction is close to t. • In addition, Br(tgpppn) is large (~10%). 5/Jun/2019 International School on c. LFV 43

Evaluated distribution of 3 p mass in tgpppn. • PRD 81, 113007 (2010) Red: MC Black: data 5/Jun/2019 • Just distribution is obtained. No fit is performed to evaluate width, mass and so on. • Due to t phase space, peak is shifted from a 1 mass. International School on c. LFV 44

Signal and Background • Main background comes from tgmnn. (nn are also missing. ) tgmnn Since we do not user true-t rest frame, monochromatic momentum distribution slightly spreads in pseudo-t rest frame. Ma=1. 4 Ge. V Ma=0 m momentum in pseudo t rest frame. 5/Jun/2019 International School on c. LFV 45

Result □ for e ■ for m 472 pb-1 data: =4 x 105 t-pair Br(tgmnn/enn)=17% UL<0. 004 @95%CL Neither Belle nor Ba. Bar updates this analysis. It may be a chance… 5/Jun/2019 International School on c. LFV 46

Lower energy t LFV search than that of B-factory 5/Jun/2019 International School on c. LFV 47

t pair production cross section • Threshold at 3. 55 Ge. V • Maximum at 4. 25 Ge. V ( s(U(4 s))x 4) • Then falls like 1/s. • at the lowest order b: velocity of t 5/Jun/2019 International School on c. LFV B-factory 48

ISR Spectrum • At near threshold • Eg for ee ttg background cannot be as high as Eg for t mg. • Background from ee mmg will become more important. good m. ID is essential. Eg (CMS) from t mg and ISR(ttg) U(4 s) maximum s 5/Jun/2019 International School on c. LFV 49

t analysis with the early Belle II data 5/Jun/2019 International School on c. LFV 50

tg 3 pn in Belle II early data • As a demonstration, we analyze this event in Belle II early data and measured tau mass. 5/Jun/2019 International School on c. LFV 51

Event selection • 1 -3 prong topology • To reject electron, for signal side, E/Plab<0. 8 is required. • thrust value > 0. 87 • Visible Energy in CMS < 9. 7 Ge. V Thrust vector • Energy for signal side in CMS < 5. 29 Ge. V • Energy for tag side in CMS < 5. 22 Ge. V For signal side, at most 1 g is allowed to accept ISR or FSR photon. tag For tag side, any number of g are allowed. side 5/Jun/2019 International School on c. LFV Signal side 52

3 p mass distribution in tg 3 pn • After the selection, we can obtain: Good agreement between data and MC is found. Simulation works well. Data is well-understood. 5/Jun/2019 International School on c. LFV 53

Event display of typical event candidate 5/Jun/2019 International School on c. LFV 54

t mass measurement (1) • 5/Jun/2019 International School on c. LFV 55

t mass measurement (2) • Zoom of the previous plot 5/Jun/2019 International School on c. LFV 56

5/Jun/2019 International School on c. LFV 57

ISR cross section • The first order ISR cross section to particular final state ; Photon energy fraction of the beam energy • • is Born cross section to is spectrum function of the ISR photon, which takes account of vertex and self energy correction. 5/Jun/2019 International School on c. LFV 58

ATLAS result for LFV higgs 5/Jun/2019 International School on c. LFV 59