Search for very rare decays of beauty mesons into four muons Tobias Tekampe on behalf of the LHCb collaboration LHC days, Split, 19th September 216
Introduction Rare B meson decays are suitable processes to look for new particles. B s,d µ + µ µ + µ is a b s, dll flavour changing neutral current decay. Loop suppressed in the Standard Model (SM). b d, s t γ W γ, Z W µ + µ µ + µ b s, d SM branching fraction predicted 1 to be O(1 11 ). So far only limits have been set by LHCb 2. B(Bs µ + µ µ + µ ) < 1.6 1 8 B(B µ + µ µ + µ ) < 6.6 1 9 Improvements are made regarding a more optimal choice of the normalisation, employing a multivariate classifier and having three times more data. P S µ + µ µ + µ 1 Dincer and Sehgal, Phys.Lett. B556 (23) 169-176 2 LHCb, Phys. Rev. Lett. 11, 21181 (213) Tobias Tekampe (TU Dortmund) B s,d µ+ µ µ + µ LHC days, Split, 19th September 216 2 / 13
The LHCb detector LHCb has the largest bb dataset with ~1 12 events recorded in LHC Run 1. 1 fb 1 at 7 TeV 3 1 11 bb. 2 fb 1 at 8 TeV 6 1 11 bb. Int. J. Mod. Phys. A 3, 15322 (215) Tobias Tekampe (TU Dortmund) B s,d µ+ µ µ + µ LHC days, Split, 19th September 216 3 / 13
Strategy for measuring a branching fraction B(Bs,d 4µ) = N(B s,d 4µ) ε(b+ J/ψ K + ) f u N(B + J/ψ K + ) ε(bs,d 4µ) f B(B + J/ψ K + ) s,d Measure the branching fraction by normalising to B + J/ψ K + with J/ψ µ + µ. Precisely measured branching fraction with ~3% uncertainty 1 and high event yield. Hadronisation probabilities f u,s,d. Determine signal yields, N, and efficiencies, ε, of each decay channel. Candidates / ( 3.5 MeV/c 2 ) 5 1 4 1 3 1 2 1 LHCb-PAPER-216-43 + + B J/ψ K Combinatorial Partially Reconstructed 52 54 56 + m(j/ψ K ) [MeV/c 2 ] Preliminary result: N(B + J/ψ K + ) = 68789 ± 92 1 PDG, Chin. Phys. C38 (214) Tobias Tekampe (TU Dortmund) B s,d µ+ µ µ + µ LHC days, Split, 19th September 216 4 / 13
Data-Simulation agreement Some distributions in LHCb simulation are known to deviate from data. Correction is needed to ensure accurate behaviour of multivariate classifier (MVA) and correct efficiencies. Perform gradient boosting re-weighting 1 in most important MVA features and variables correlated to particle identification (PID) variables. Correction function trained on B s J/ψ ( µ + µ )φ( K + K ) background subtracted data and simulation. Candidates / (12). 8. 7. 6. 5. 4. 3. 2. 1 LHCb-PAPER-216-43 1 2 reweighted MC MC Data(sPlot). 1 2 3 4 5 6 Number of tracks 1 A. Rogozhnikov, arxiv:168.586 (216) Candidates / (.6). 14. 12. 1. 8. 6. 4. 2 LHCb-PAPER-216-43 reweighted MC MC Data(sPlot). 5 1 15 2 25 3 χvtx 2 Tobias Tekampe (TU Dortmund) B s,d µ+ µ µ + µ LHC days, Split, 19th September 216 5 / 13
Matrix Net classifier Proprietary algorithm developed by Yandex 1. Trained on re-weighted B s,d µ+ µ µ + µ signal simulation and B mass sidebands from data. Training features: kinematical and geometrical variables 1-fold cross validation is performed, such that all data can be used without any bias. A. Gulin, I. Kuralenok, D. Pavlov; JMLR Proceedings 14 pp. 63-76 (211) Tobias Tekampe (TU Dortmund) B s,d µ+ µ µ + µ LHC days, Split, 19th September 216 6 / 13
Selection optimisation Grid search for maximum of figure of merit 1 in DLL µπ and MVA response. FoM = ε signal 5/2 + N expected background ε background Expected number of background events from fit to sidebands. Background efficiency measured on sidebands. Signal efficiency measured on simulated events. Candidates / (11 MeV/c 2 ) 25 2 15 1 5 LHCb-PAPER-216-43 data Bs search region B search region signal region optimisation 5 51 52 53 54 55 56 1 G. Punzi, PhyStat23, arxiv:physics/3863 (23) Tobias Tekampe (TU Dortmund) B s,d µ+ µ µ + µ LHC days, Split, 19th September 216 7 / 13
Event selection - - Combinatorial background: Suppressed by 99.7% by using a multivariate classifier. Misidentified particles: Cut on DLLµπ. Resonance vetoes: φ:.95 < m(µ + µ ) < 1.9 GeV/c 2 J/ψ : 3. < m(µ + µ ) < 3.2 GeV/c 2 ψ(2s): 3.6 < m(µ + µ ) < 3.8 GeV/c 2 B s,d µ+ µ µ + µ signal Vetoes are 35% efficient on signal. ) [MeV/c 2 ] m(µ + µ 4 3 2 LHCb-PAPER-216-43 4 3 2 ) [MeV/c 2 ] m(µ + µ 4 3 2 LHCb-PAPER-216-43 1 1 1 1 45 5 55 6 - - m(µ + µ µ + µ ) [MeV/c 2 ] 45 5 55 6 - - m(µ + µ µ + µ ) [MeV/c 2 ] Tobias Tekampe (TU Dortmund) B s,d µ+ µ µ + µ LHC days, Split, 19th September 216 8 / 13
Systematic uncertainties For the final results only the ratios of efficiencies are relevant. Systematic uncertainties partially cancel. The selection efficiency is measured on simulation and crosschecked on data. Dominant sources of systematic uncertainty are: Effect B Bs f u /f q - 5.2% 1 Data-simulation re-weighting 3.6% 3.6% B(B + J/ψ K + ) 3.% 3.% Total 7.2% 9.2% Systematic uncertainty is small compared to statistical. 1 LHCb-CONF-213-11 Tobias Tekampe (TU Dortmund) B s,d µ+ µ µ + µ LHC days, Split, 19th September 216 9 / 13
Expected Limits Candidates / (34 MeV/c 2 ) 6 5 4 3 2 1 LHCb-PAPER-216-43 data Bs search region B search region 45 5 55 6 signal region optimisation evaluation Fit sidebands to extract expected amount of background in m(b s,d ) ± 4 MeV/c2. N b (B s ) =.47 ±.23 ±.18 N b (B ) =.55 ±.24 ±.2 Assuming to observe no signal candidates, Use CLs method 1 to compute upper limits on branching fractions. B(B s µ + µ µ + µ ) < 2.6 1 9 B(B µ + µ µ + µ ) < 6.9 1 1 More than a factor 6 smaller than the limits expected by previous LHCb measurement. 1 A.L. Read, J. Phys. G28 (22) Tobias Tekampe (TU Dortmund) B s,d µ+ µ µ + µ LHC days, Split, 19th September 216 1 / 13
Sensitivity in a NP scenario µ + b s, d P S µ µ + µ Calculate efficiencies from B s,d S(µ+ µ )P(µ + µ ) simulation. Motivated by HyperCP search for Σ + pµ + µ 1. Found 3 candidates, all at m(µ + µ ) = 214 MeV/c 2, compatible with a pseudoscalar. Arbitrarily set m(s) = 2.5 GeV/c 2. Scalar and pseudoscalar new particles decay almost instantaneously. 1 HyperCP, Phys. Rev. Lett. 94, 2181 (25) Tobias Tekampe (TU Dortmund) B s,d µ+ µ µ + µ LHC days, Split, 19th September 216 11 / 13
NEW: Measured Limits ] 9 ) [ 1 µ + µ µ + µ Β(B 2 1.8 1.6 1.4 1.2 1.8.6.4.2 LHCb-PAPER-216-43 95% CL Exclusion observed expected ±1 σ expected ±2 σ 1 2 3 4 5 6 9 Β(B µ + µ µ + µ ) [ 1 ] s ] 9 ) [ 1 ) Β(P µ + µ µ + µ S,P) Β(S Β(B 2 1.8 1.6 1.4 1.2 1.8.6.4.2 LHCb-PAPER-216-43 B S,P; B S,P s + + S µ µ ; P µ µ m S =2.5 GeV; m P =214 MeV 95% CL Exclusion observed expected ±1 σ expected ±2 σ 1 2 3 4 5 6 9 Β(B S,P) Β(S µ + µ ) Β(P µ + µ ) [ 1 ] s No signal candidates have been found in the blind region. Preliminary observed limits at 95% CL are: B(B s 4µ) < 2.5 1 9 B(B 4µ) < 6.9 1 1 B(Bs SP 4µ) < 2.2 1 9 B(B SP 4µ) < 6. 1 1 Tobias Tekampe (TU Dortmund) B s,d µ+ µ µ + µ LHC days, Split, 19th September 216 12 / 13
Summary Preliminary, improved limits have been presented. Table shows comparison to previously measured results. f gives the ratio between previous and new limits. Mode 1 fb 1 3 fb 1 f Bs µ + µ µ + µ < 1.6 1 8 < 2.5 1 9 6.4 B µ + µ µ + µ < 6.6 1 9 < 6.9 1 1 9.5 Bs S(µ + µ )P(µ + µ ) < 1.6 1 8 < 2.2 1 9 7.3 B S(µ + µ )P(µ + µ ) < 6.3 1 9 < 6. 1 1 1.5 Documented in LHCb-PAPER-216-43, which will be available soon. Tobias Tekampe (TU Dortmund) B s,d µ+ µ µ + µ LHC days, Split, 19th September 216 13 / 13
BACKUP Tobias Tekampe (TU Dortmund) B s,d µ+ µ µ + µ LHC days, Split, 19th September 216 14 / 13