Search for B (s,d) μμ in LHCb

Transcription

1 Search for B (s,d) μμ in LHCb Diego artinez Santos (CERN) (on behalf of the LHCb Collaboration) CERN-PH-EP , submitted to PLB 1

2 Outline Observables: Branching Ratios (BR) of B s,d μμ 1 Why do we want to measure that? indirect probe for New Physics(NP) How do we do it? (Analysis strategy) How to find such a rare decay and disentangle from background Normalization and Calibration to get a correct BR 3 What did we get with 010 data? Interesting results Amazing prospects

3 Indirect Approach B s,d μμ can access NP through new virtual particles entering in the loop indirect search Indirect approaches can access higher energy scales and see NP effects earlier: 3 rd quark family inferred by Kobayashi and askawa (1973) to explain CP V in K mixing (1964). Directly observed in 1977 (b) and 1995 (t) Neutral Currents discovered in 1973, Z 0 directly observed in 1983 Roundness of Earth (Eratosthenes, c.iii B.C) discovered ~300 years before direct observation ~.3 k years till the direct observation Eratosthenes 3

4 S and New Physics This decay is very suppressed in S : BR(B s µµ) = (3. ± 0.)x10-9 BR(B d µµ) = (1.0 ± 0.1)x10-10 Experimental upper limit still one order of magnitude above such 95% CL: (q = u, c, t) BR(B s µµ) < 4.3x10-8 BR(B d µµ) < 0.76x10-8 (CDF, 3.7 fb -1, prel.) BR(B s µµ) < 5.1x10-8 (D0, 6.1 fb -1, publ.) But in NP models it can take any value from << S (e.g, some NSS) up to current experimental upper limit (e.g. SUSY at high tanβ) +? Whatever the actual value is, it will have an impact on NP searches 4

5 Best fit contours in tanβ vs A plane in the NUH1 model [O. Buchmuller et al, Eur.Phys.J.C64: ,009] B s µµ x10-8 1x10-8 (5σ ~fb -1 ) 5x10-9 S-like CS H/A ττ, 5σ, ~30-60fb -1 Regions compatible with BR(B s μμ) ~ x10-8, 1x10-8,5x10-9 and S-like. Private calculation using SuperIso program, (F. ahmoudi, arxiv: ) and SoftSusy (B.C. Allanach, Comput. Phys. Commun. 143 (00) ) 5

6 LHCb Low angle spectrometer Very efficient trigger Good particle identification performance Precise reconstruction: TRACKING/ VERTEXING pp interaction ECAL HCAL Separation production vertex decay vertex σ(ip)~ 5 μm Invariant mass Δp/p ~ % B s,d μμ signature: Hits in muon detector µµ pair has B invariant mass geometrical & kinematical signature: pt, detachment of decay vertex μ+ μ- 6

7 Analysis strategy Selection cuts in order to reduce the amount of data to analyze. LHCb trigger selects > 90% of the signal that is interesting for the offline analysis. Classification of B s,d μμ events in bins of a D space Invariant mass of the μμ pair ultivariate discriminant variable combining geometrical and kinematical information about the event: Geometrical Likelihood (GL) Flat distributed for signal, background peaks at 0 Control channels to get signal and background expectations w/o relying on simulation Bd search window Bs search window Compare expectations with observed distribution. Results combined using CL s method 7

8 Geometrical Likelihood S-B separation relies strongly on this variable Trained using C samples of B s μμ signal and bb μμ background. Distributions taken from data to not rely on the accuracy of the simulation Distribution of real signal obtained by looking at B h + h - in real data. Similar to C expectation. Background distribution is obtained from data by interpolating from mass sidebands in GL bins 8

9 Invariant ass Signal distribution depends on the actual mass resolution of LHCb in the B mass region (resolution depends on mass, almost linearly) easured in data by interpolating from dimuon resonances (J/ψ (m<mb), Y (m>mb) ) and looking at B h + h - (B d,s K + π -, B d π + π -, B s K + K - ) μμ background yield in mass bins is interpolated from mass sidebands B h + h - (muon identification not required) σ = 6.71±0.95 ev/c 9

10 Normalization Three channels are used, each one with different (dis)advantages: B + J/ψ( μμ)k + : Ratio of probabilities of b quark to hadronize into the different mesons. f d /f + = 1, f d /f s = 3.71 ± 0.47 (from HFAG) Similar trigger (muon triggers) to the signal, similar particle identif. N(B + J/ψ( μμ)k + ) ~ 13k Well known BR, but is B + and not B s ~13% systematic for B s μμ Different number of tracks in the final state Normalization factors: α(b d ) = ( )x10-9 α(b s ) = (8. 1.3)x

11 Normalization Three channels are used, each one with different (dis)advantages: B s J/ψ( μμ)φ ( K + K - ) : Ratio of probabilities of b quark to hadronize into the different mesons. f d /f + = 1, f d /f s = 3.71 ± 0.47 (from HFAG) Similar trigger (muon triggers) to the signal, similar particle identif. It s a B s, but BR known only with 6% precision Different number of tracks in the final state Normalization factors: α(b d ) = ( )x10-9 α(b s ) = (10.5.9)x

12 Normalization Three channels are used, each one with different (dis)advantages: B d K + π - Ratio of probabilities of b quark to hadronize into the different mesons. f d /f + = 1, f d /f s = 3.71 ± 0.47 (from HFAG) Different trigger (used triggered on the underlying event/other b used) Same kinematics, number of tracks in final state Well known BR, but is B d and not B s ~13% systematic for B s μμ Normalization factors: α(b d ) = ( )x10-9 α(b s ) = ( )x10-9 1

13 Observed pattern and Result (B s ) B mass (ev/c) GL bin 1 GL bin GL bin 3 [-60, -40] [-40,-0] [-0,0] [0,0] [0,40] [40,60] (TOTAL) GL bin 4 (bkg exp.) BR(B s μμ) < 4.3 (5.6) 90 (95% CL) BR(B s µµ) < 95% CL(CDF, prelim) BR(B s µµ) < 95% CL(D0, publish.) Expected are: 5.1 (6.5) CL s 13

14 Observed pattern and Result (B d ) B mass (ev/c) GL bin 1 GL bin GL bin 3 [-60, -40] [-40,-0] [-0,0] [0,0] [0,40] [40,60] (TOTAL) GL bin 4 (bkg exp.) CL s BR(B d μμ) < 1. (1.5) 90 (95% CL) BR(B s µµ) < 95% CL(CDF, prelim) Expected are: 1.4 (1.8) 14

15 BR(B s -> ) (x10-9 ) BR(B s -> ) (x10-9 ) Extrapolated sensitivity %CL 90% CL Exclusion 10 S PRED. S PRED. s 1/ =7 TeV Luminosity (fb -1 ) s 1/ =7 TeV Luminosity(fb -1 ) LHCb can provide VERY interesting results in one year from now! 15

16 Conclusions B s,d μμ is an interesting probe of physics beyond the Standard odel First LHCb result on BR(B s,d μμ ) BR(B s μμ) < 4.3 (5.6) 90 (95% CL) BR(B d μμ) < 1. (1.5) 90 (95% CL) Those are comparable with current best ones Extrapolation to fb -1 shows that LHCb can find/exclude BR(B s,d μμ) from ~10-8 to quite close to S prediction 16

17 Backup 17

18 18

19 Background yield Interpolation in the 4 GL bins gives in one shot the D distribution GL vs ass in the search window (60 ev around the B mass) Peaking background (B h + h - wrongly identified as muons) negligible for current amount of data GL bin [0,0.5] GL bin (0.5,0.75] GL bin (0.5,0.5] GL bin (0.75,1.0] 19

20 0

21 1

22 Decay Physics in S Branching Ratio (BR) as a function of Wilson Coefficients ( effective theory) is: BR( B q 1 4m GF ) 64 C S 3 V V * tb C tq P m 3 C 10 f 1 4m (q = u, c, t) C S, P scalar and pseudo scalar are negligible in S C 10 gives the only relevant contribution This decay is very suppressed in S (BR very small, but precisely predicted): BR(B s µµ) = (3. ± 0.)x10-9 BR(B d µµ) = (1.0 ± 0.1)x10-10 Experimental upper limit still one order of magnitude above such 95% CL: BR(B s µµ) < 4.3x10-8 BR(B d µµ) < 0.76x10-8 (CDF, 3.7 fb-1, prelim.) BR(B s µµ) < 5.1x10-8 (D0, 6.1 fb-1, publ.)

23 3

24 4

25 5

26 6

27 Geometrical Likelihood How the Geometry likelihood is built: 1. Input variables: min Impact Parameter Significance (µ +,µ - ), DOCA, Impact Parameter of B, lifetime, iso - µ +, iso- µ - Isolation: Idea: muons making fake Bs μμ might came from another SV s For each muon; remove the other μ and look at the rest of the event: How many good - SV s (forward, DOCA, pointing) can it make? The precise criteria used is inherited from Hlt Generic Isolation Red: signal Blue: bb inc. Black: b μ b μ Green: Bc+ J/Ψμν Bs IP (mm) less μips 7

28 Geometrical Likelihood How the Geometry likelihood is built: 1. Input variables: min Impact Parameter Significance (µ +,µ - ), DOCA, Impact Parameter of B, lifetime, iso - µ +, iso- µ -. They are transformed to Gaussian through cumulative and inverse error function 3. In such space correlations are more linear-like rotation matrix, and repeat 45 o 8

29 IPS Geometrical Likelihood How the Geometry likelihood is built: 1. Input variables: min Impact Parameter Significance (µ +,µ - ), DOCA, Impact Parameter of B, lifetime, iso - µ +, iso- µ -. They are transformed to Gaussian through cumulative and inverse error function 3. In such space correlations are more linear-like rotation matrix, and repeat 4. Transformations under signal hyp. χ S, under bkg. χ B. 5. Discriminating variable is χ S -χ B, made flat for better visualization. lifetime G S G B χ B Sensitive region GL> 0.5 χ S t (ps) G S1 G B1 9

30 30

31 31

32 Wilson coefficients Hadronic weak decays are often studied in terms of effective hamiltonians of local operators Q i : H eff i C Qˆ i i effective local theory Degrees of freedom of exchanged particles are integrated out giving rise to the Wilson coefficients C i. underlying fundamental theory (S) An example of similar approach: Fermi s theory of neutron decay BR(B s µµ) expressed in eff. th. as: BR( B q ) GF 64 3 V V * tb tq 3 f 1 4m C P,S,10 (pseudoscalar, scalar and axial) depend on the underlying model (S, SUSY ) 1 4m C S C P m C 10 3

33 Computing CLs Reference: Thomas Junk, CERN-EP/ arch 1999 (Used at LEP for Higgs searches) For each bin: si = expected signal events in bin bi = expected bkg. events in bin di = measured events in bin X i Poisson( di, Poisson( d di, d i i si bi ) b ) i For a configuration {Xi}: X N i X i (it is a binned likelihood ratio) CL CL s b b b P s b ( X X P ( X X CLs = CLs+b/CLb OBSERVED OBSERVED ) ) High CLb observed excess w.r.t bkg expectation signal (CLb> sigma) Small CLs too few events w.r.t prediction from signal hypothesis 33

34 34 ' ' ' 3 * sin 64 ) ( A A q P q P q S q S tq tb W F q C C m C C C C m m f V V G B BR

35 BR(B s -> ) (x10-9 ) s 1/ =7 TeV 10 S PRED Luminosity(fb -1 ) 35