Searches for BSM Physics in Rare B-Decays in ATLAS Iskander Ibragimov on behalf of the ATLAS Collaboration SUSY 14 The nd International Conference on Supersymmetry and Unification of Fundamental Interactions University of Manchester 1-6 July 14, Manchester (England) 1
ATLAS B-Physics Data excellent data taking efficiency and quality of data multiple interactions per bunch crossing (<µ>) > 5 fb recorded in 11 (7 TeV) <µ> = 9 > fb recorded in 1 (8 TeV) <µ> = only 11 data are used in the presented searches topological di-muon triggers require two muons with pt(µ) > 4 GeV or more at the first trigger level full track reconstruction and loose mass selection at the higher trigger levels - no trigger prescales applied in 11 single muon triggers also used (e.g. in Bd K * µ + µ - analysis) high-rate triggers prescaled at higher instantaneous luminosity ( nd half of 11, at the beginning of a run) Bs(5366) µ + µ -?
Bs µ + µ - : Introduction flavor changing neutral current (FCNC) decay strongly suppressed in the SM at tree level, loop-dominated coupling to non-sm particles can affect BR => powerful indirect search for New Physics BR(Bs µ + µ - ) = (3.7.7)x1-9 [Buras et al., Eur.Phys.J. C7 (1) 17] = (3.54.3)x1-9 (time-integrated) [K. De Bruyn et al., Phys.Rev.Lett 19 (1) 4181] = (.9.7)x1-9 (LHCb and CMS) [LHCb: arxiv:137.54, CMS: arxiv:137.55] BR(Bd µ + µ - ) <.7 x1-9 (LHCb) still some room for NP left! LHC Run II data may give an answer... [D. Straub arxiv:15.694] 3
Bs µ + µ - : Analysis Strategy Relative BR measurement: partial cancelation of uncertainties (luminosity, cross-section,...) reference channel B J/ψK BR(B s! µµ) = N Bs!µµ Single Event Sensitivity (SES): BR if one signal event observed 1 N B!J/!K BR(B! J /!K ) f " u B!J/!K f s " Bs!µµ A B!J/!K A Bs!µµ counting in signal region yield from mass fit to data [PDG 1] [LHCb: JHEP 134 (13) 1]] ε x A = NB reco & sel / NB gen on MC tuned to data use MVA technique (BDT) for signal/background discrimination optimize discrimination avoiding biasses => independent datasets used for - BDT training (MC modeling) - selection optimization (5% of sideband data) - background measurement (remaining 5% of sideband data) blind analysis (region 3 GeV around Bs mass blinded) 4
Bs µ + µ - : Background Composition continuum (from real _ muons) dominated by bb µ + µ - X measured by interpolation from sideband data modeled with dedicated MC sample continuum (muon + fake, i.e. hadrons reconstructed as muons due to decays in flight or punch-through) e.g. Bd π - µ + ν, Bs K - µ + ν populate mostly left sideband contribution negligible for 5 fb analysis peaking B hh ( fake + fake ) mainly Bs K + K -, Bd K π, Bs π + π - decays BR x (fake rate) 1-9 (non-negligible due to signal-like topology) b Arbitrary Units _ b µ - c µ + µ + background.6 B s /B s KK B K - + s.5.4.3..1 + - B s K B s /B s B d /B d KK B d B d K - + K + - B d /B d Bs µ - signal ATLAS Preliminary Simulation s=7 TeV B s 171 MeV B s 133 MeV B s 116 MeV 5 51 5 53 54 55 56 Invariant mass [MeV] µ + 5
Bs µ + µ - : Background Discrimination pointing angle: modeling in background MC Entries Data / MC 8 Sideband data MC bb to µµx 7 MC signal 6 5 4 3 1.5 3 1.5.5 1 ATLAS Preliminary s = 7 TeV Ldt = 4.9 fb..4.6.8 1 1. 1.4 D [rad] pointing angle: data/mc agreement MVA for continuous background separation Boosted Decision Trees (BDT) trained with 13 best separation variables pointing angle is one of the most powerful good modeling in background MC, good data/ MC agreement optimization of BDT selection (q) and search window ( m): maximum of P(q, m) = εsig/(1+ Nbkg) εsig from MC, Nbkg from data (5% of events) optimized selection : q >.118, m = 11 MeV candidates /.1 rad Data/MC B 4 1 3 1 1 1 ATLAS Preliminary s = 7 TeV Ldt = 4.9 fb Data (SB subtracted) MC signal.5 1 1.5.5 3 D [rad] P(µµ) µ - αd SV ( x)xy PV y µ + x Entries /.11 3 5 15 1 5 Sideband data MC signal ATLAS Preliminary s = 7 TeV Ldt = 4.9 fb -.6 -.4 -...4 BDT output 6
Bs µ + µ - : Reference Channel Yield and εxa Ratio N(B J/ψK ) extraction: selection as close as possible to Bs (to minimize overall systematics) use BDT trained on Bs use the same BDT selection unbinned max. likelihood fit use per-event mass resolution δm systematics accessed by varying background fit model N(B J/ψK ) = 1514 1.1% (stat.).4% (syst.) ε x A measurement: Events / 5 MeV Events / 3 MeV 45 Preliminary Data 4 35 3 5 15 1 5 3 5 15 1 ATLAS s = 7 TeV Ldt = 4.9 fb 5 51 5 53 54 55 56 ATLAS Preliminary s = 7 TeV Signal + background fit B J/ K B J/ Ldt = 4.9 fb Data Signal + background fit B J/ K B J/ signal m J/K signal background Other backgrounds background Other backgrounds [MeV] systematics from data/mc discrepancies 5 1 3 4 5 6 7 8 m J/K [MeV] 7
Bs µ + µ - : Result Single Event Sensitivity: SES = (.7.6) 1-9 1.5 % error dominated by systematics, mainly: BR(B ), fu/fs : 8.5 % ε x A ratio : 6.9 % absolute K tracking efficiency: 5 % Events / 6 MeV 1 ATLAS Preliminary s = 7 TeV 8 6 4 Data - B µ + s µ MC (1x) blinded region optimized search region Ldt = 4.9 fb CL s 1 Observed CLs 48 5 5 54 56 58 m µµ [MeV] 1-1 Expected CLs - Median Expected CLs 1 Expected CLs total Nbkg expected in search region: 6.75 events (with.3 events from B hh ) => BR(Bs µ + µ - ) < 1.6 x1-8 (@ 95% CL) -3 1-4 1 ATLAS Preliminary s = 7 TeV Ldt = 4.9 fb 1 3 4 5 BR(B s - -8 µ + µ ) [1 ] Nµµ observed in search region: 6 events => BR(Bs µ + µ - ) < 1.5 x1-8 (@ 95% CL) 8
Bd K * µ + µ - : Introduction exclusive final state for b sl + l - transition b sl + l - can occur only via loop-suppressed W-exchange: SM expectation BR = (1.6.1) 1-6 [PDG 13] contribution from new particles can affect BR angular observables sensitive to NP: AFB - muon forward-backward asymmetry FL - fraction of longitudinally polarized K * mesons 9
Bd * + K µµ : Analysis Strategy kinematic observables dimuon mass q 3 angles: θl, θk, ϕ => decay rate φ µ+ θl K+ Bd µ limited statistics => use ϕ symmetry integrate over ϕ, cos θl : integrate over ϕ, cos θk : θk π extract AFB (q) and FL (q) via unbinned max. likelihood fit 1
Bd K * µ + µ - : Signal Selection Background _ contributions: _ bb µ + µ - X (main), cc µ + µ - X, Drell-Yan => require τ/στ > 1.75 and cos θpointing >.999 (selections optimized on MC) radiative decays of charmonium in Bd K * J/ψ, Bd K * ψ(s) and J/ψ, ψ(s) tails => veto mass regions in (m(bd )rec - m(bd )PDG) - (mµµ, rec - mj/ψ, PDG) < m (13 MeV) Additional selections: K * ( K + π - ) mass acceptance range [846, 946] MeV veto Bd K * J/ψ, Bd K * ψ(s) decays: 8.68 < q < 1.9 (J/ψ µµ) 1.86 < q < 14.18 (ψ(s) µµ) Bd mass likelihood fit: signal model: Gaussian with per-event errors background model: exponential Nsig = 466 34 Nbkg = 113 43 Events / 4 MeV ATLAS Preliminary s = 7 TeV.4 GeV < q < 19. GeV " Ldt = 4.9 fb 18 16 14 1 1 8 6 4 49 5 51 5 53 54 55 56 57 Data Signal fit Background fit Total fit m(k!µµ) [MeV] 11
Bd K * µ + µ - : Angular Fits unbinned max. likelihood fits using sequential approach for each of six q bins: 1. fit mass distribution => get PDF mass parameters and signal fraction.fit angular distributions with parameters of the first step fixed => extract AFB and FL (using q bin definitions of BELLE) 1
Bd K * µ + µ - : Results A FB 1.8.6.4. F L 1.9.8.7.6 <FL(q )> Theory ATLAS BaBar Belle CDF CMS LHCb.5 -. -.4 -.6 -.8 <AFB(q )> Theory ATLAS BaBar Belle CDF CMS LHCb.4.3..1 4 6 8 1 1 14 16 18 4 6 8 1 1 14 16 18 q [GeV ] q [GeV ] measurements agree with SM and other experiments statistical uncertainties dominate in high q bin ATLAS results are competitive 13
Summary high-quality results with full s = 7 TeV dataset (5 fb ) rare decay Bs µ + µ - (ATLAS-CONF-13-76) angular analysis of Bd K * µ + µ - (ATLAS-CONF-13-38) no signs of BSM Physics so far Bs µ + µ - : observed BR < 1.5 x 1-8, consistent with SM Bd K * µ + µ - : AFB and FL measurements consistent with theoretical predictions and other measurements analyses with s = 8 TeV dataset (ca. fb ) in preparation data of LHC Run will bring us even more statistical power! 14